25 European Young Geotechnical Engineers Conference

Romanian Society for Geotechnical and Foundation Engineering

Technical University of Civil Engineering Bucharest
International Society for Soil Mechanics and Geotechnical Engineering

Proceedings of the
th
25 European Young Geotechnical Engineers Conference
Proceedings of the

21-24 June 2016, Sibiu, Romania

th

Editors
Ernest OLINIC, Sanda MANEA
 Romanian Society for Geotechnical and Foundation Engineering
Technical University of Civil Engineering Bucharest
International Society for Soil Mechanics and Geotechnical Engineering

Proceedings of the
25th European Young Geotechnical Engineers Conference

21-24 June 2016, Sibiu, Romania

Editors
Ernest OLINIC, Sanda MANEA

EDITURA CONSPRESS
2016
 Copyright © 2016, Editura Conspress úi editorii

EDITURA CONSPRESS
este recunoscută de
Consiliul NaĠional al Cercetării ùtiinĠifice din ÎnvăĠământul Superior

Descrierea CIP a Bibliotecii NaĠionale a României

EUROPEAN YOUNG ENGINEERS CONFERENCE (25 ; 2016 ; Sibiu)

Proceedings of the 25th European Young Engineers Conference : Sibiu, 21-

24 iunie 2016 / ed.: Ernest Olinic, Sanda Manea. – Bucureúti : Conspress, 2016
ISBN 978-973-100-421-1

I. Olinic, Ernest (ed.)

II. Manea, Sanda (ed.)

Carte universitară

CONSPRESS
B-dul Lacul Tei nr.124, sector 2,
cod 020396, Bucureúti
Tel.: (021) 242 2719 / 300; Fax: (021) 242 0781

ii
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

ORGANIZING COMMITTEE
Ernest OLINIC – Conference Chairperson
Technical University of Civil Engineering Bucharest
Nicoleta – Maria ILIES
Technical University of Cluj – Napoca
Iancu – Bogdan TEODORU
“Gh. Asachi” Technical University of Iasi
Daniel MANOLI
Technical University of Civil Engineering Bucharest
Catalin BURLACU
Technical University of Civil Engineering Bucharest

SCIENCIFIC COMMITTEE
Sanda MANEA –Chairperson
Technical University of Civil Engineering Bucharest
Ernest OLINIC – Conference Chairperson
Technical University of Civil Engineering Bucharest
Loretta BATALI
Technical University of Civil Engineering Bucharest
Nicoleta – Maria ILIES
Technical University of Cluj – Napoca
Irina LUNGU
“Gh. Asachi” Technical University of Iasi
Luiza ROMAN
"Politehnica" University of Timisoara

SPONSORS

iii
iv
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Table of contents

KEYNOTE LECTURES
General presentation of Eurocode 7 on ‘Geotechnical design’
Roger FRANK 3
Modern Design of Landfill Liners
Andrea DOMINIJANNI, Mario MANASSERO, Giacomo BOFFA 21
Difficult foundation conditions in Romania
Ernest OLINIC 37

Session 1. In-situ and laboratory tests

Permanent strain behavior of cyclically loaded cohesive soil in undrained conditions
Andrzej GàUCHOWSKI 53
Geotechnical Characterization of Very Soft Deep-Sea Sediments by In-Situ Penetrometer
Testing
Bert LIETAERT, Francois CHARLET, Peter STAELENS 61
Determination of characteristic values of geotechnical parameters
Monica LÖFMAN 73
Evaluation of ground deformation modulus definition by in situ and laboratory test methods
Sergii POKLONSKYI 83
Geotechnical investigation for determination of causes for occurred deformations of the
Skopje Aqueduct
Borche RUMENOVSKI, Katerina TASEVSKA, Bojan JAKIMOVSKI, Teodor
DIMITRIEVSKI 87

Session 2. Researches in geotechnical engineering

Settlements of Foundations in Rock Fill due to Cyclic Loading and Creep Effects
Madeleine BRANDT, Terje STRØMME 99
Discussion on selected results obtained for fine grained soil treated with fly ash
Karolina KNAPIK 105
Barrettes under negative friction of soil surrounding
Iryna MIESHKOVA 113
Engineering geological investigation of a cave spa cut into rhyolite tuff
Ádám VINCZE, Péter GÖRÖG 119
Continuous Horizontal Reinforcement Diaphragm Wall Application
Nejla YILDIZ, Alp GÖKALP 123

v
Session 3. Underground structures
Energy walls for an underground car park
Alice DI DONNA 133
Use of the Hardening Soil Model for urban tunnels design
Nicolas GILLERON 139
Investigations on energy sheet pile walls for the exploitation of renewable thermal energy
David KOPPMANN 147
Wallslotrobot: a new method of deep braced excavation in compact spaces
Sahin MONSEREZ 153
Dual cell shaft for a 19.5m deep excavation. Design, execution and monitoring
Cristian RADU 163
Eurasia Tunnel Project: The Geotechnical Challenges of the Asian Transition Box
Alexis ROSE 171
Back-analysis of a deep excavation in Budapest
Attila SZEPESHÁZI 177

Session 4. Retaining structures

Design of Retaining Walls at Metro Nordhavn, Copenhagen
Asbjørn M. DAM 191
Eurocode 7 based retaining wall design in Finland – shortcomings in current method
Mika KNUUTI 201
Reliability-based assessment of a retaining wall using FEM
Katerina RIPPI, Ana TEIXEIRA 209
Dry Dock at the Merwede River Banks
Leon SCHADEE 217
Design of the Temporary Cofferdam and Permanent Intake Channel for the Dublin Waste to
Energy Facility
Brian G. SEXTON 225

Session 5. Numerical/physical modelling

Investigation of the influence of gravity on granular flow using silo centrifuge model
Olja BARBIR, John MATHEWS 237
Selected geotechnical applications of deformation sensing by fibre optics and problems with
gathered data evaluation
Kristýna ýÁPOVÁ 247
Finite Deformation Modelling of Cyclic Sand Behaviour using Bounding Surface Plasticity
Christian CAROW 257
Finite element limit analysis of offshore mudmat foundations subject to combined loads
Helen P. DUNNE 267
Jack-up vessel foundation stability during installation next to a Wind Turbine Generator
Carlos MOLINA, Lindita KELLEZI 275

vi
Numerical analysis of penetration problems in clay with the Particle Finite Element Method
Lluís MONFORTE 285
Modelling Spudcan penetration using Abaqus CEL method
Aleksander WORREN 293

Session 6. Slope stability and lanslides

Stabilization of railway landslide with bored pile retaining wall
Marko BISCAN, Goran DIZDAR 303
Methodology for preparation of thematic geotechnical maps for urbanization purposes
Natasha NEDELKOVSKA 309
Experimental and numerical models in landslide behaviour
Ferran PARERA 317
Combined geotechnical and structural numerical analysis for cut slope stabilization
Katerina LAKOýEVIû, Mario STOJANOVSKI 325
Erosion of cuts in flysch: field monitoring by TLS and modelling of slope geometry
development
Goran VLASTELICA 331
Geotechnics in Belarus: problems and solutions
Sergey IGNATOV, Igor BOIKO 341

Session 7. Soil dynamics

4D seismic and velocity evolution in monitoring heated heavy oil sand reservoirs
Hong DOAN 349
Cohesive soil laboratory investigation using resonant column and torsional shear device
Alexandra Alisa GĂINĂ 355
Combined analysis of single station ambient vibration measurements and MASW method
Fátima GOUVEIA, Isabel LOPES, Rui Carrilho GOMES 361
The influence of dynamic loads on postglacial lacustrine fine-grained deposits
Michael HAVINGA 371
Prevention of soil liquefaction using stone columns
Jamal HLEIBIEH 377
Application of Offshore Seismic CPT data for Soil Stiffness Interpretation
Olsi KORETA, Lindita KELLEZI 383
Numerical analyses of the performance of seismically isolating buried barriers
Valeria NAPPA 393

Session 8. Geosynthetics
Swelling ability, volume of cracks and hydraulic conductivity of a polymer modified bentonite
subjected to wet and dry cycles
Michela DE CAMILLIS 401

vii
Physical and numerical modeling of an unpaved road structure liable to void forming. The
effect of a biaxial geosynthetic reinforcement.
Andor-Csongor NAGY, Dorin Vasile MOLDOVAN 409
Influence of prehydration condition on the hydraulic conductivity of GCLs permeated with
landfill leachates
Tu÷çe ÖZDAMAR KUL, Havva DEMøRKIRAN 417
Geotextile encased columns – verification of the analytical design method
Katarina SIRK, Boštjan PULKO 423

viii
25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Keynote lectures
2
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

General presentation of Eurocode 7 on ‘Geotechnical design’

Roger FRANK1*
1
Université Paris-Est, Ecole des Ponts, Lab Navier-Geotechnical group (CERMES)

ABSTRACT
Eurocode 7 on 'Geotechnical design' is now being used in many countries throughout Europe.
Part 1 devoted to the 'General rules' has been published in 2004. National Annexes and
accompanying documents have been prepared for implementation in the various European
countries. Part 2 on 'Ground investigation and testing' was published in 2007. In 2015, the drafting
of a revised version was launched. After describing shortly the history of the development of
Eurocode 7, the contents of the two present documents are given and the main concepts are
described (verification procedures and geotechnical categories, characteristic values, derived
values, ULS verifications, SLS verifications and allowable movements of foundations).

Keywords: Eurocode 7, design values, limit states, foundations, retaining strucutres

x EN 1997 Eurocode 7: Geotechnical

1. INTRODUCTION
The system of Structural Eurocodes
includes 10 following sets of standards
(EN for ‘European Norm’):
x EN 1990 Eurocode: Basis of structural
design
x EN 1991 Eurocode 1: Actions on
structures
x EN 1992 Eurocode 2: Design of
concrete structures
x EN 1993 Eurocode 3: Design of steel
structures
x EN 1994 Eurocode 4: Design of
composite steel and concrete structures
x EN 1995 Eurocode 5: Design of timber
structures
x EN 1996 Eurocode 6: Design of
masonry structures
* presenting author
design
x EN 1998 Eurocode 8: Design of
structures for earthquake resistance
x EN 1999 Eurocode 9: Design of
aluminium structures.
The Structural Eurocodes are design
codes for buildings and civil engineering
works. They are based on the Limit State
Design (LSD) approach used with a partial
factor method.
Except for EN 1990, all Eurocodes are
subdivided into several parts. Eurocodes
2, 3, 4, 5, 6 and 9 are ‘material’
Eurocodes, i.e. relevant to a given
material. EN 1990 (Basis of design),
Eurocode 1 (Actions), Eurocode 7
(Geotechnical design) and Eurocode 8
(Earthquake resistance) are relevant to all
types of construction, whatever the
material.
Altogether, 58 parts in total have been
published and the official deadline for
withdrawing all conflicting national
standards was April 2010.
Eurocode 7 should be used for all the
problems of interaction of structures with
the ground (soils and rocks), through
foundations or retaining structures. It
addresses not only buildings but also
bridges and other civil engineering works.
It allows the calculation of the
geotechnical actions on the structures, as
well the resistances of the ground
submitted to the actions from the
structures. It also gives all the
prescriptions and rules for good practice
required for properly conducting the
geotechnical aspect of a structural project
or, more generally speaking, a purely
geotechnical project.
Eurocode 7 consists presently of two
parts:
x EN 1997-1 Geotechnical design - Part
1: General rules (CEN, 2004, 2013)
x EN 1997-2 Geotechnical design - Part
2: Ground investigation and testing (CEN,
2007)
After describing shortly the history of
the development of Eurocode 7, and
giving the main contents of the two parts,
the main concepts are described, without
recalling all the principles of LSD and of
the partial factor method used.
2. HISTORY OF EUROCODE 7 AND
IMPLEMENTATION
The first Eurocode 7 Group, in charge of
drafting an European standard on
geotechnical design, was created in 1981.
It was composed of representatives of the
Member Societies of the International
Society for Soil mechanics and
Geotechnical Engineering (ISSMGE) of
the 10 countries forming the European
Community at that time. A first model code
on general rules for geotechnical design
(corresponding to Eurocode 7- Part 1) was
published in 1990 (EC7, 1990).
4
In 1990, the task of drafting design
codes for buildings and civil engineering
works was transferred to the Comité
Européen de Normalisation (CEN,
European Committee for Standardization)
and CEN/TC 250 (Technical Committee
250) in charge of all the ‘Structural
Eurocodes’ was created. In particular, SC
7, Sub-Committee 7, is in charge of
Eurocode 7 on ‘Geotechnical design’.
Note that CEN is composed of the national
standard bodies of a number of European
countries (since 2013, 33 countries are
members, i.e. the present 28 countries of
EU, plus 3 countries of EFTA, FYR
Macedonia and Turkey). N. Krebs Ovesen
(Denmark) was the first Chairman of
CEN/TC 250/SC 7, from 1990 until 1998.
The author was the Chairman of SC 7
from 1998 to 2004. From 2004 to 2010,
Bernd Schuppener (Germany) was the
Chairman. The new Chairman, since
2010, is Andrew Bond (UK).
In 1993, SC 7 adopted the ENV 1997-1
pre-standard: ‘Geotechnical design - Part
1: General Rules’ (CEN, 1994). It was
clear, at that time, that (much) more work
still needed to be done before reaching a
full European standard (EN) acceptable to
all members of CEN. An important fact
helped in obtaining, in 1997, a positive
vote for the conversion into an EN. It was
the recognition by CEN/TC 250 that
geotechnical design is unique and cannot
be considered to be the same as other
design practices needed in the
construction industry. The models
commonly used vary from one country to
the other and cannot be harmonised
easily, simply because the geologies are
different and form the rationale for the so-
called ‘local traditions’… This recognition
is confirmed by a resolution taken by TC
250 (Resolution N 87, 1996): ‘CEN/TC
250 accepts the principle that EN 1997-1
might be devoted exclusively to the
fundamental rules of geotechnical design
and be supplemented by national
standards’.
The work for the conversion of ENV
1997-1 into EN 1997-1 ‘Geotechnical
design – Part 1: General rules’ was
performed from 1997 to 2003. The formal
positive vote by CEN members was
obtained early 2004 and CEN finally
published Eurocode 7 – Part 1 (EN 1997-
1) in November 2004 (CEN, 2004).
Two other ENVs, devoted to
geotechnical design assisted by laboratory
testing and by field (in situ) testing were
drafted rather quickly, facing no serious
controversy. They were published in 1999
(CEN, 1999a and 1999b) and, in 2001, the
members of CEN voted positively for their
conversion into a European Norm. During
the conversion phase, the two documents
were merged into the single document
called ‘Eurocode 7 Geotechnical design -
Part 2: Ground investigation and testing’.
The formal positive vote was obtained in
May 2006 and the document was
published in March 2007 (CEN, 2007).
The publication of a Eurocode Part by
each national standardisation body with its
National Annex (in the official language(s)
of the country) had to be completed within
two years after publication by CEN. The
role of the National Annex is to indicate
the decisions corresponding to the so-
called "Nationally Determined Parameters
(NDPs)" (like values of partial factors,
choice of Design Approach, status of
informative Annexes, etc. – see below).
The ‘legal’ status of standards/norms is
different in each country and the
regulatory bodies of the various countries
have an important role to play for the
implementation of the Eurocodes. A
‘Guidance Paper’ has been elaborated by
the European Commission to co-ordinate
the implementation of the Eurocodes into
the national regulations (EC, 2003a). The
European Commission has also issued a
strong recommendation to the Member
States inviting them to adopt the
Eurocodes in their regulations (EC,
2003b). The paper by Anagnostopoulos
and Frank (2012) gives more details about
all the implementation process.
In 2015, the drafting of the ‘Second
Generation of Structural Eurocodes’ was
launched by CEN with the support of the
European Commission. The documents
5
consisting of the whole set of revised
structural Eurocodes should be ready by
2020 and the implementation in the
various countries could follow shortly
afterwards. With regard to Eurocode 7, it
is intended to organise the revised code in
three parts. Part 1 would include and
extend all the general rules included in the
present Part 1. Part 2 would remain
essentially the same as the present Part 2.
Part 3 will be created from the sections of
the present Part 1 dealing with specific
geotechnical structures and will include
more details on the design of some of
them. The various sections would then
deal respectively with: Slopes, cuttings,
and embankments, Spread foundations,
Pile foundations, Retaining walls, Anchors,
Reinforced soil structures and Ground
improvement.
3. CONTENTS OF THE PRESENT
DOCUMENTS
3.1. Part 1: General rules
Eurocode 7 - Part 1 is a rather general
document giving only the principles for
geotechnical design inside the general
framework of LSD. These principles are
relevant to the calculation of the
geotechnical actions on structures
(buildings and civil engineering works) and
to the design of the structural elements
themselves in contact with the ground
(footings, piles, basement walls, etc.).
Detailed design rules or calculation
models, i.e. precise formulae or charts are
only given in informative Annexes. As
already mentioned, the main reason is that
the design models in geotechnical
engineering differ from one country to the
other, and it was not possible to reach a
consensus, especially when many of
these models still need to be calibrated
and adapted to the LSD approach…
Eurocode 7 – Part 1 includes the
following sections (CEN, 2004, 2013):
x Section 1 General
x Section 2 Basis of geotechnical
design
x Section 3 Geotechnical data
x Section 4 Supervision of construction,
monitoring and maintenance
x Section 5 Fill, dewatering, ground
improvement and reinforcement
x Section 6 Spread foundations
x Section 7 Pile foundations
x Section 8 Anchors
x Section 9 Retaining structures
x Section 10 Hydraulic failure
x Section 11 Overall stability
x Section 12 Embankments
A number of Annexes are included.
They are all informative, except for Annex
A which is 'normative' (i. e. mandatory).
The list of the Annexes of EN 1997-1 is
the following:
x Annex A (normative) Partial factors for
ultimate limit states
x Annex B Background information on
partial factors for Design Approaches 1, 2
3 foundations, shallow foundations and
x Annex C Sample procedures to
determine limit values of earth pressures
on vertical walls
x Annex D A sample analytical method
for bearing resistance calculation
x Annex E A sample semi-empirical
method for bearing resistance estimation
x Annex F Sample methods for
settlement evaluation
x Annex G A sample method for deriving
presumed bearing resistance for spread
foundations on rock
x Annex H Limiting foundation
movements and structural deformation
x Annex J Checklist for construction
supervision and performance monitoring
Annex A is important, as it gives the
partial factors for ULS in persistent and
transient design situations (fundamental
combinations), as well as correlation
factors for the characteristic values of pile
bearing capacity. But the numerical values
for the partial or correlation factors given
in Annex A are only recommended values.
The exact values of the factors can be
changed by each national standardisation
6
body in the so-called National Annex. All
other Annexes are informative (i. e. not
mandatory in the normative sense). Some
of them, though, contain valuable material
which can be accepted, in the near future,
by most of the countries. The National
Annex can give a 'normative(s)' status to
one or to several of the 'informative'
Annexes, i.e. it (they) will be mandatory in
the corresponding country.
As mentioned above, each country is
also free to supplement the general rules
of Eurocode 7 by national application
standards, in order to specify the
calculation models and design rules to be
applied in the country. Whatever their
contents they have to respect in all
aspects the principles of Eurocode 7. In
France, for instance, 5 national standards
have been issued for the application of
Eurocode 7. They deal with the design of
embedded walls, of reinforced and soil
nailing retaining structures, deep
gravity walls, respectively (AFNOR,
2009a, 2009b, 2012, 2013 and 2014).
3.2. Part 2: Ground investigation and
testing
The role of this part of Eurocode 7
devoted to laboratory and field testing is to
give the essential requirements for the
equipment and test procedures, for the
reporting and the presentation of results,
for their interpretation and, finally, for the
derivation of values of geotechnical
parameters for the design. It complements
the requirements of Part 1 in order to
ensure a safe and economic geotechnical
design.
It makes the link between the design
requirements of Part 1, in particular
Section 3 ‘Geotechnical data’, and the
results of a number of laboratory and field
tests.
It does not cover the standardisation of
the geotechnical tests themselves.
Another Technical Committee (TC) on
'Geotechnical investigation and testing'
has precisely been created by CEN to
consider this matter (TC 341). In this
respect the role of Part 2 of Eurocode 7 is
to ‘use’ and refer to the detailed rules for
test standards covered by TC 341.
Eurocode 7 – Part 2 includes the
following Sections (CEN, 2007):
Section 1 General
Section 2 Planning of ground
investigations
Section 3 Soil and rock sampling and
groundwater measurements
Section 4 Field tests in soils and rocks
Section 5 Laboratory tests on soils and
rocks
Section 6 Ground investigation report
The Section on field tests in soils and
rocks includes:
- cone penetration tests CPT(U)
- pressuremeter tests PMT
- rock dilatometer tests RDT
- standard penetration tests SPT
- dynamic penetration tests DP
- weight sounding tests WST
- field vane tests FVT
- flat dilatometer tests DMT
- plate loading tests PLT
The Section on laboratory testing of
soils and rocks deals with:
- preparation of soil specimens for testing
- preparation of rock specimens for testing
- tests for classification, identification and
description of soils
- chemical testing of soils and
groundwater
- strength index testing of soils
- strength testing of soils
- compressibility and deformation testing
of soils
- compaction testing of soils
- permeability testing of soils
- tests for classification of rocks
- swelling testing of rock material
- strength testing of rock material
There are provisions on how to
establish and use the so-called ‘derived
values’ from the tests (see paragraph 4.3
below). Some of these provisions are
7
meant to give guidance for using the
sample calculation models in the Annexes
of Part 1. Part 2 also includes a number of
informative Annexes with precise
examples of derived values of
geotechnical parameters and coefficients
commonly used in design.
As is the case in Part 1, most of the
derivations or calculation models given are
informative, but there is also fairly good
agreement about using them in the future
throughout Europe. In any case, they are
a clear picture of the approaches existing
on the continent for the use of in situ or
laboratory test results in the design of
geotechnical structures.
4. SOME ASPECTS OF EUROCODE 7
4.1. Verification procedures and
geotechnical categories
The discussions about verifications of
geotechnical design usually focus on
approaches performed through
calculations. Nevertheless, it should be
stressed that calculations are not the only
means for checking that the basic
requirements are fulfilled.
Eurocode 7 – Part 1 offers, in fact,
various possibilities (clause 2.1 in EN
1997-1) :
‘(4) Limit states should be verified by one
or a combination of the following:
- use of calculations […];
- adoption of prescriptive measures, […];
- experimental models and load tests, […];
- an observational method, […].’
This paragraph is clear enough.
However, it may be useful to add that :
- the adoption of prescriptive measures
indicates that, in some circumstances (see
the geotechnical categories below), one
may avoid calculations which may look
long and cumbersome with regard to the
problem under consideration ;
- the use of experimental models and load
tests recalls that the fundamentals of
geotechnical design and of its calculation
rules are the monitoring of the behaviour
of real structures, with recourse, when
necessary, to full scale tests ;
- finally, mentioning the observational
method, shows one of the directions
devoted to contemporary geotechnical
design (with full consistency with the
fundamentals mentioned above).
With regard to the observational
method, Eurocode 7 adds that (clause 2.7
in EN 1997-1) :
'(2)P The following requirements shall be
met before construction is started:
acceptable limits of behaviour shall be
established;
the range of possible behaviour shall be
assessed and it shall be shown that there
is an acceptable probability that the actual
behaviour will be within the acceptable
limits;
a plan of monitoring shall be devised,
which will reveal whether the actual
behaviour lies within the acceptable limits.
The monitoring shall make this clear at a
sufficiently early stage, and with
sufficiently short intervals to allow
contingency actions to be undertaken
successfully;
the response time of the instruments and
the procedures for analysing the results
shall be sufficiently rapid in relation to the
possible evolution of the system;
a plan of contingency actions shall be
devised, which may be adopted if the
monitoring reveals behaviour outside
acceptable limits.'
(note that, in the Eurocodes, when the
letter 'P' accompanies the number of a
paragraph, it means that it is a principle,
i.e. a fundamental requirement;
paragraphs not marked with 'P' are only
'application rules').
The use of the observational method
should grow considerably in the coming
years (see Huybrechts et al., 2005).
In order to define the design
requirements and the levels needed for
the geotechnical investigation, Eurocode 7
introduces three geotechnical categories
(clause 2.1 in EN 1997-1). It is a way of
introducing, one can say, 'consequences
8
classes ' (see Annex B of EN 1990, CEN,
2002).
Geotechnical category 1 corresponds to
the simple structures that can be designed
and executed, with negligible risk, only on
the basis of experience and with a
qualitative geotechnical investigation. One
can place in this category retaining walls
of moderate height or direct foundations
for individual houses, in simple
geotechnical conditions (no neither
stability nor water problems, etc.).
Geotechnical category 2 covers
conventional geotechnical structures,
without exceptional risk (i.e. without
difficult geotechnical conditions or
loadings). Eurocode 7 requirements
concerning calculations and ground
investigations fully apply to category 2
structures (clause 2.1 in EN 1997-1) :
'(18) Designs for structures in
Geotechnical Category 2 should normally
include quantitative geotechnical data and
analysis to ensure that the fundamental
requirements are satisfied.
(19) Routine procedures for field and
laboratory testing and for design and
execution may be used for Geotechnical
Category 2 designs.
NOTE The following are examples of
conventional structures or parts of
structures complying with Geotechnical
Category 2:
- spread foundations;
- raft foundations;
- pile foundations;
- walls and other structures retaining or
supporting soil or water;
- excavations;
- bridge piers and abutments;
- embankments and earthworks;
- ground anchors and other tie-back
systems;
- tunnels in hard, non-fractured rock and
not subjected to special water tightness or
other requirements.'
Category 3 includes all geotechnical
structures with abnormal risks, for which
Eurocode 7 requirements may not be
sufficient to ensure an acceptable level of
safety. The risks can derive from the
ground conditions or from the loading
conditions. Examples of structures falling
into this category are large dams,
foundations of nuclear power plants,
structures on unstable ground, etc.
Eurocode 7 clearly indicates that (clause
2.1 in EN 1997-1) :
'(21) Geotechnical Category 3 should
normally include alternative provisions and
rules to those in this standard [EN 1997-
1].'
In the Eurocode system, as mentioned
earlier, the calculation method prescribed
is the LSD approach used in conjunction
with a partial factor method. Problems
encountered in geotechnical engineering
projects are often due to reasons not
linked to design calculations. For
geotechnical practice, Eurocode 7 – Part 1
also mentions that (clause 2.4.1 in
EN 1997-1) :
‘(2) It should be considered that
knowledge of the ground conditions
depends on the extent and quality of the
geotechnical investigations. Such
knowledge and the control of
workmanship are usually more significant
to fulfilling the fundamental requirements
than is precision in the calculation models
and partial factors.’
4.2. Characteristic values
The present ‘philosophy’ with regard to the
definition of characteristic values of
geotechnical parameters is contained in
the following clauses of Eurocode 7 – Part
1 (clause 2.4.5.2 in EN1997-1):
‘(2)P The characteristic value of a
geotechnical parameter shall be selected
as a cautious estimate of the value
affecting the occurrence of the limit state.’
‘(7) […]the governing parameter is often
the mean of a range of values covering a
large surface or volume of the ground. The
characteristic value should be a cautious
estimate of this mean value.’
These paragraphs in Eurocode 7 – Part
1 reflect the concern that one should be
able to keep using the values of the
9
geotechnical parameters that were
traditionally used (the determination of
which is not standardised, i.e. they often
depend on the individual judgment of the
geotechnical engineer, one should
confess). However two remarks should be
made at this point: on the one hand, the
concept of 'derived value' of a
geotechnical parameter (preceding the
determination of the characteristic value),
has been introduced (see Figure 1 and
paragraph 4.3) and, on the other hand,
there is now a clear reference to the limit
state involved (which may look evident,
but is, in any case, a way of linking
traditional geotechnical engineering and
the new limit state approach) and to the
assessment of the mean value (and not a
local value; this might appear to be a
specific feature of geotechnical design
which, indeed, involves 'large' areas or
'large' ground masses).
Statistical methods are mentioned only
as a possibility:
‘(10) If statistical methods are employed
[…], such methods should differentiate
between local and regional sampling […].’
‘(11) If statistical methods are used, the
characteristic value should be derived
such that the calculated probability of a
worse value governing the occurrence of
the limit state under consideration is not
greater than 5%. NOTE In this respect, a
cautious estimate of the mean value is a
selection of the mean value of the limited
set of geotechnical parameter values, with
a confidence level of 95%; where local
failure is concerned, a cautious estimate
of the low value is a 5% fractile.’
The general feeling is that the
characteristic value of a geotechnical
parameter cannot be fundamentally
different from the value that was
traditionally used. Indeed, for the majority
of projects, the geotechnical investigation
is such that no serious statistical treatment
of the data can be performed. Statistical
methods are, of course, useful for very
large projects where the amount of data
justifies them.
4.3. Derived values
Many geotechnical tests, particularly field
tests, do not allow basic geotechnical
parameters or coefficients, for example for
strength and deformation, to be
determined directly. Instead, values of
these parameters and coefficients must be
derived using theoretical or empirical
correlations.
The concept of 'derived values' has
been introduced in ENV 1997-3 (CEN
1999b), in order to give a status to
correlations and models commonly used
to obtain, from both results of field tests
and results of laboratory tests,
geotechnical parameters and coefficients
which enter directly into the design. Their
use is intended, primarily, for the design of
pile and shallow foundations as mentioned
in the Annexes D, E, F, and G of
Eurocode 7 - Part 1.
The definition of derived values is given
in Eurocode 7 – Part 2 as:
‘Derived values of geotechnical
parameters and/or coefficients, are
obtained from test results by theory,
correlation or empiricism.’
Type of test
F= field L= laboratory F1
From field test results, the geotechnical
parameter obtained is either an input for
an analytical or indirect model, or a
coefficient for use in a semi-empirical or
direct model of foundation design.
Derived values of a geotechnical
parameter then serve as input for
assessing the characteristic value of this
parameter in the sense of Eurocode 7 -
Part 1 (clause 2.4.5.2 of EN 1997-1) and,
further, its design value, by applying the
partial factor JM ('material factor', clause
2.4.6.2).
The role played by the derived values
of geotechnical parameters can be
understood with the help of figure 1, taken
from Eurocode 7 - Part 2. The borderline
between Part 1 (EN 1997-1) and Part 2
(EN 1997-2) of Eurocode 7 is also shown
on the figure. It can be seen that the
requirements concerning the
measurements of geotechnical properties,
as well as their derived values are covered
by Part 2: 'Ground investigation and
testing', while those concerning the
determination of characteristic values and
design values are given by Part 1:
'General rules'.
F2 L1 L2

Correlations C1 C2
Information
from other
Test results and sources on
1 2 3 4
derived values the site, the
soils and
EN 1997-2 rocks and
the project
EN 1997-1

Cautious selection

Geotechnical model and characteristic

value s of geotechnical properties

Application of
partial factors

Design values of geotechnical

properties

Figure 1. General framework for the selection of derived values, characteristic values and design
values of geotechnical properties (CEN, 2007)
10
4.4. Verifications of ultimate limit states
(ULS)
The ultimate limit states (ULS) to be
checked are defined, in the following
manner, by Eurocode 7 – Part 1,
consistently with ‘Eurocode: Basis of
structural design’ (CEN 2002) (clause
2.4.7.1 in EN 1997-1):
‘(1)P Where relevant, it shall be verified
that the following limit states are not
exceeded:
– loss of equilibrium of the structure or
the ground, considered as a rigid body, in
which the strengths of structural materials
and the ground are insignificant in
providing resistance (EQU);
– internal failure or excessive deformation
of the structure or structural elements,
including footings, piles, basement walls,
etc., in which the strength of structural
materials is significant in providing
resistance (STR);
– failure or excessive deformation of the
ground, in which the strength of soil or
rock is significant in providing resistance
(GEO);
– loss of equilibrium of the structure or
the ground due to uplift by water pressure
(buoyancy) or other vertical actions (UPL);
– hydraulic heave, internal erosion and
piping in the ground caused by hydraulic
gradients (HYD).
NOTE Limit state GEO is often critical to
the sizing of structural elements involved
in foundations or retaining structures and
sometimes to the strength of structural
elements.’
The ultimate limit states should be
verified for the combinations of actions
corresponding to the following design
situations (see EN 1990, CEN, 2002):
x permanent and transient (the
corresponding combinations are called
'fundamental'); in the following these
design situations are noted 'p&tds' for
convenience;
x accidental;
x seismic (see also Eurocode 8 - Part 5,
i.e. EN 1998-5).
11
The design values of the actions and
the combinations of actions are defined in
EN 1990 (partial factors J for the actions
and factors \ for the accompanying
variable actions).
The debate about the format for
checking the GEO and STR ultimate limit
states (ULS) was relevant to the persistent
and transient design situations ('p&tds').
This debate follows from the ENV 1997-1
(CEN, 1994) formulation which inferred
that ULS in persistent and transient design
situations had to be checked for two
formats of combinations of actions, i.e. for
Cases B and C, as they were called at that
time. B was aimed at checking the
uncertainty on the loads coming from the
structure, and C the uncertainty on the
resistance of the ground. Some
geotechnical engineers were in favour of
this double check, as others preferred
having to use only one single format of
combinations of actions (more details can
be found in Frank and Magnan, 1999).
The consensus reached between
structural and geotechnical engineers
opened the way to three different Design
Approaches (DA 1, DA 2 and DA 3). The
choice is left to national determination, i.e.
each country will have to state in its
National Annex, the Design Approach(es)
to be used for each type of geotechnical
structure (spread foundations, pile
foundations, retaining structures, slope
stability).
Generally speaking, for checking ULS-
p&tds, three sets of partial factors to be
applied to characteristic values of actions
are introduced in EN 1990: Sets A, B & C:
 set A is used for checking the static
equilibrium of the structure (EQU);
 set B is relevant to the design of
structural members (STR) not involving
geotechnical actions;
 sets B and C are relevant to the design
of structural members involving
geotechnical actions and the resistance of
the ground (STR/GEO).
Tables 1, 2 and 3 give, in a simplified
manner, the recommended values for
buildings for Sets A, B and C, taken from recommended values of the factors
Tables A1.2 (A), A1.2(B) and A1.2(C) of corresponding to expression 6.10 (table
EN 1990 (CEN, 2002). The recommended A.3 in the note to paragraph A.3(1)P of
values given may be modified by National Annex A in EN 1997-1). This derives from
decision. the fact that the recommended
For STR/GEO ULS in p&tds, the three geotechnical values come from a few
Design Approaches are the following calibration studies performed using the
(clause A1.3.1 in EN 1990): values of expression 6.10, while, on the
‘(5) Design of structural members other hand, there is no experience on the
use of expressions 6.10a et 6.10b in
(footings, piles, basement walls, etc.)
geotechnical engineering.
(STR) involving geotechnical actions and
the resistance of the ground (GEO) should
be verified using one of the following three Table 1. Recommended values for partial
approaches supplemented, for factors for actions (Set A) after EN 1990
geotechnical actions and resistances, by (CEN, 2002) – ULS in p&tds
EN 1997: Action Symbol Value
Approach 1: Applying in separate Permanent actions
calculations design values from Table
- unfavourable JG,sup 1.10(1)
A1.2(C) and Table A1.2(B) to the
- favourable JG,inf 0.90(1)
geotechnical actions as well as the other
actions on/from the structure. In common Variable actions
cases, the sizing of foundations is - unfavourable JQ 1.50
governed by Table A1.2(C) and the - favourable 0
structural resistance is governed by Table
A1.2(B);Note: In some cases, application of (1) As an alternative, the favourable part may
these tables is more complex, see EN 1997. be multiplied by JG,inf = 1.15 and the
Approach 2: Applying design values unfavourable part by JG,sup = 1.35
from Table A1.2(B) to the geotechnical Table 2. Recommended values for partial
actions as well as the other actions factors for actions (Set B) after EN 1990
on/from the structure; (CEN, 2002) – ULS in p&tds
Approach 3: Applying design values Action Symbol Value
from Table A1.2(C) to the geotechnical Eq. Eq. Eq.
actions and, simultaneously, applying
(6.10) (6.10a) (6.10b)
partial factors from Table A1.2(B) to the
other actions on/from the structure. Permanent
Note: The use of approaches 1, 2 or 3 is -unfavourable(1) JGsup 1.35 1.35 1.15(2)
chosen in the National annex.’ - favourable(1) JGinf 1.00 1.00 1.00
In other words, Design Approach 1
Variable
(DA1) is the double check procedure
coming from the ENV 1997-1 (B+C - unfavourable JQ 1.50 1.5\0 1.50
verifications) and Design Approaches 2 - favourable 0 0 0
(DA 2) and 3 (DA 3) are new procedures (1) all permanent actions from one source are
using a single format of combinations of multiplied by JGsup or by JGinf.
actions. DA 2 is elaborated with (2) value of [ is 0.85, so that 0.85JGsup = 0.85 u
‘resistance factors’ for the ground (RFA), 1.35 # 1.15.
as DA 3 makes uses of ‘material factors’ Note 1: choice between expression 6.10 or
for the ground (MFA). expressions 6.10a and 6.10b used together, is
by National decision
With regard to the choice between
Note 2: JG and JQ may be subdivided into Jg
expression 6.10 or expressions 6.10a and and Jq and the model uncertainty factor JSd. JSd
6.10b of EN 1990 (see table 2 for set B), = 1.15 is recommended.
Eurocode 7 only mentions the
12
 Table 3. Recommended values for partial have recommended values equal to 1.0
factors for actions (Set C) after EN 1990 (Sets M1 and R1) or near 1.0 (Set R1 in
(CEN, 2002) – ULS in p&tds the case of axially loaded piles and
Action Symbol Value anchorages); in the second format
Permanent actions imposed by DA 1 (combination 2, former
- unfavourable JG,sup 1.00 case C), the elementary properties of the
- favourable JG,inf 1.00 ground (shear strength parameters) are
Variable actions always factored for the calculation of
- unfavourable JQ 1.30 geotechnical actions and sometimes
- favourable 0 factored for the calculation of resistances
Furthermore, Eurocode 7 allows (Set M2); in the case of axially loaded
applying the partial factors either on the piles and anchorages, the total resistance
actions themselves ("at the source") or on is directly factored by applying Set R4.
the effects of the actions (they are noted In DA 2, safety is applied both on the
ȖF and ȖE , respectively). In principle, for actions (Set B) and on the total ground
DA 1 they are applied "at the source". For resistance (Set R2).
DA 2 and DA 3, both options are allowed.
In DA 3, safety is applied both on the
This is relevant to the factors of set B and
actions (Set B for the actions coming from
of set C (unfavourable variable actions).
the structure and Set M2 for the
Table 4 gives the link between Sets B elementary properties of the ground acting
and C and the corresponding sets of on the structure, i.e. for the geotechnical
factors for geotechnical actions and actions) and on the geotechnical
resistances: Sets M1 and M2 for material resistances (Set M2 for the elementary
properties (e.g. c', ij', cu, etc.) and Sets properties; the recommended values for
R1, R2, R3 and R4 for total resistances Set R3 for the total geotechnical
(e.g. bearing capacity, etc.). These sets resistance is always equal to 1.0, except
are defined in Annex A of Eurocode 7 – for piles in tension and anchorages for
Part 1. As mentioned above, Annex A which they are equal to 1.1).
also gives recommended values for the
Figures 2, 3 and 4, as well as their
partial factors; these values may be set
captions, illustrate the situation for each of
differently by the National Annex. Note
the three Design Approaches. On these
that the recommended values for the
figures, index 'd' indicates a design value
partial factors ȖM on material properties in
different from the characteristic value
Set M1 are always equal to 1.0.
(application of a partial factor J different
Table 4. STR/GEO – ULS in p&tds. Partial
from 1.0) and index 'k' indicates a design
factors to be used according to EN 1990 and
EN 1997-1
value equal to the characteristic value
(application of a partial factor J equal to
Actions
1.0).
Design on/from
Geotechnical It should be mentioned that ‘model
approach the
structure factors’ can also be introduced (clause
Actions Resistances 2.4.7.1 in EN 1997-1):
1 B B and M1 M1 and R1 ‘(6) When calculating the design value
M2 and R1 of the resistance, (Rd ), or the design
C C and M2 or value of the effect of actions, (Ed ), model
M1 and R4*
factors, (JR;d ) or (JS;d ) respectively, may
2 B B and M1 M1 and R2
be introduced to ensure that the results of
3 B C and M2 M2 and R3
the design calculation model are either
*for piles and anchorages
accurate or err on the safe side.’ An
In DA 1, the first format (combination 1, example of the use of a model factor for
former case B) applies safety mainly on the bearing capacity of piles is given by
actions, while the factors on resistances Burlon et al. (2014)
13
 More details on the use of the three
Design Approaches are given, for
instance, in Frank et al. (2004).
With regard to the design values for
accidental situations, Eurocode 7 only
states that (clause 2.4.7.1 in EN 1997-1):
'(3) All values of partial factors for
actions or the effects of actions in
accidental situations should normally be
taken equal to 1,0. All values of partial
factors for resistances should then be
selected according to the particular
circumstances of the accidental situation.
NOTE The values of the partial factors may be
set by the National annex. '

a) DA 1: Combination 1 b) DA 1: Combination 2
Note: for simplicity, only vertical equilibrium is considered and only unfavourable actions are shown.
Figure 2. ULS in p&tds. Design Approach 1 - introduction of partial factors (recommended values)
in the checking of ground bearing capacity (Frank et al., 2004).

a) Factoring actions at the source (DA 2) b) Factoring effects of actions (DA 2*)
Note: for simplicity, only vertical equilibrium is considered and only unfavourable actions are shown.
Figure 3. ULS in p&tds. Design Approach 2 - introduction of partial factors (recommended values)
in the verification of ground bearing capacity (Frank et al., 2004).

14
 Note: for simplicity, only vertical equilibrium is considered and only unfavourable actions are shown.
Figure 4. ULS in p&tds. Design Approach 3 - introduction of partial factors (recommended values)
in the verification of ground bearing capacity (Frank et al., 2004).
4.5. Verification of serviceability limit
states (SLS)
The main discussions during the
development of Eurocode 7 were about
the format for verifying ULS in permanent
and transient situations. However, the
verification of serviceability limit states
(SLS) is an issue equally important in
contemporary geotechnical design. This
issue is fully recognised by Eurocode 7
which indeed often refers to displacement
calculations of foundations and retaining
structures, while common geotechnical
practice mainly sought so far to master
serviceability by limiting the bearing
capacity or by limiting the shear strength
mobilisation of the ground to relatively low
values.
The verification of SLS in the real sense
proposed by Eurocode 7 (prediction of
displacements of foundations) is certainly
going to gain importance in the near
future. For the time being, it is an aspect
which is too often neglected in common
geotechnical practice.
Eurocode 7 – Part 1 repeats the
formulation of EN 1990 (clause 2.4.8, EN
1997-1):
'(1)P Verification for serviceability limit
states in the ground or in a structural
section, element or connection, shall
either require that:
15
Ed d Cd (2.10)
or be done through the method given in
2.4.8(4).
(2) Values of partial factors for
serviceability limit states should normally
be taken equal to 1,0.
NOTE The values of the partial factors
may be set by the National annex.'
with Ed the design value of the effect of
actions and Cd the limiting value
(serviceability criterion) of the design value
of effect of actions.
At the same time, Eurocode 7
introduces immediately the possibility to
keep the traditional approach mentioned
above (clause 2.4.8 in EN 1997-1) :
'(4) It may be verified that a sufficiently low
fraction of the ground strength is mobilised
to keep deformations within the required
serviceability limits, provided this
simplified approach is restricted to design
situations where:
a value of the deformation is not required
to check the serviceability limit state;
established comparable experience exists
with similar ground, structures and
application method.'
This clause is to be linked with the one
dealing with the design methods of spread
foundations (paragraph 6.4(5)P in EN
1997-1) :
'(5)P One of the following design methods
shall be used for spread foundations:
a direct method, in which separate
analyses are carried out for each limit
state. When checking against an ultimate
limit state, the calculation shall model as
closely as possible the failure mechanism,
which is envisaged. When checking
against a serviceability limit state, a
settlement calculation shall be used;
an indirect method using comparable
experience and the results of field or
laboratory measurements or observations,
and chosen in relation to serviceability
limit state loads so as to satisfy the
requirements of all relevant limit states;
a prescriptive method in which a
presumed bearing resistance is used (see
2.5).'
Indeed, the indirect method 'chosen in
relation to serviceability limit state loads'
comes to applying the traditional method
of designing the bearing capacity of
spread foundations, i.e. a simple
calculation comparing the applied loads
for serviceability limit states to a limit load
divided by a global factor of safety high
enough (usually around 3). Of course, as
indicated in Eurocode 7, this can only be
valid if there is no need to assess the
settlement of the foundation.
Paragraph 2.4.8(2) of Eurocode 7 –
Part 1, reproduced above, indicating that
partial actors for SLS are normally taken
equal to 1.0 (in other words that the
design values of the various quantities are
taken equal to their characteristic values),
applies to the actions in the characteristic,
frequent or quasi-permanent combinations
(see EN 1990), as well as to the
geotechnical properties, such as the
modulus of deformation. It should be
noted that, for determining the differential
settlement for instance, sets of lower
characteristic values and upper
characteristic values can be chosen in
order to take account of the ground
variability.
With regard to the use of the
combinations of actions for SLS, EN 1990
provides (in editorial notes) some
16
guidelines which are summarised in table
5 (clause 6.5.3 in EN 1990).
Table 5. Recommended combinations of
actions for checking serviceability limit states
SLS
Combination Use according to
of actions EN 1990
Characteristic Irreversible limit states
Frequent Reversible limit states
Quasi Long term effect and
permanent appearance
When applying equation 2.10 of clause
2.4.8(1)P (see above), it appears that the
frequent and quasi-permanent should be
recommended ; on the contrary, in the
case of the alternative method allowed by
2.4.8(4), it seems that the characteristic
(or 'rare') combination should be used,
because the experience gained in the past
was rather for loads near this type of
combination.
The last general paragraph in
Eurocode 7 – Part 1 about SLS states that
(clause 2.4.8 in EN 1997-1):
'(5)P A limiting value for a particular
deformation is the value at which a
serviceability limit state, such as
unacceptable cracking or jamming of
doors, is deemed to occur in the
supported structure. This limiting value
shall be agreed during the design of the
supported structure.'
The application of these general
clauses is detailed further down in
Eurocode 7 – Part 1 for each geotechnical
structure (in the Sections for spread
foundations, pile foundations, retaining
structures, overall stability and
embankments). It is interesting to note that
the document insists several times on the
difficulty to predict displacements with
accuracy (in the present state of
geotechnical engineering knowledge, of
course!).
4.6. Limiting values of displacements of
foundations
The knowledge of limiting allowable
displacements of foundations is a subject
of prime importance, even though it is not
often explicitly addressed. These limiting
values depend primarily, of course, on the
nature of the supported structure, but it
has also been a point of interest for
geotechnical engineering for a long time,
as well (a summary of data collected for
buildings and bridges is given e.g. by
Frank, 1991).
The limiting values of movements of
foundations are the subject, in particular,
of clause 2.4.9, as well as of Annex H
(informative) of Eurocode 7 – Part 1. It is
noted that clause 2.4.9 contains 4 rather
strong principles, i.e. paragraphs (1)P to
(4)P. The first one says:
'(1)P In foundation design, limiting values
shall be established for the foundation
movements.
NOTE Permitted foundation movements
may be set by the National annex.'
Furthermore, it seems that not only
SLS are concerned (see above) but also
ULS…(because movements of
foundations can trigger an ULS in the
supported structure).
Eurocode 7 gives a list of a certain
number of factors which should be
considered when establishing the limiting
values of movements. It is important that
these limiting values are established in a
realistic manner, by close collaboration
between the geotechnical engineer and
the structural engineer. If the values are
too much severe, they will usually lead to
uneconomical designs.
Figure 5 defines the parameters used
to quantify movements and deformations
of structures. This figure, originally due to
Burland and Wroth (1975) is reproduced in
Annex H (informative) of Eurocode 7 –
Part 1. Annex H quotes the following
limits after Burland et al. (1977):
- for open framed structures, infilled
frames and load bearing or continuous
brick walls: maximum relative rotations
between about 1/2000 and about 1/300 to
prevent the occurrence of a SLS in the
structure;
17
- for many structures, a maximum relative
rotation ȕ = 1/500 is acceptable for SLS
and ȕ = 1/150 for ULS;
- for normal structures with isolated
foundations, total settlements up to 50 mm
are often acceptable.
These values can serve as a guide, in
the absence of other indications on the
limiting values for the deformations of the
structures.
5. LIAISONS WITH OTHER CEN
COMMITTEES
Inside the Eurocode system itself, there
are, of course, many links between the
different standards or parts of them.
Eurocode 7 on Geotechnical design is
more precisely linked to the following
ones:
- EN 1990: ‘Eurocode: Basis of structural
design’ which defines the various limit
states and design situations to be
checked, and gives the general rules for
taking into account the actions on/from the
structures and the geotechnical actions;
- EN 1998-5: Design of structures for
earthquake resistance. Foundations,
retaining structures and geotechnical
aspects.
The other Technical Committees of
CEN working on standards of interest for
Eurocode 7, and for which coordination
must be ensured
are: CEN/TC 341 on ‘Geotechnical
investigation and testing’, as mentioned
earlier; CEN/TC 288 on ‘Execution of
geotechnical works’; CEN/TC 189 on
‘Geotextiles and geotextile-related
products’; CEN/TC 227 on ‘Road
materials’.
The standards on execution (TC 288) and
on geotechnical tests (TC 341) are
particularly important as they complement
Eurocode 7, which is devoted only to
design.

smax
Gs max
a) definitions of settlement s, differential
settlement Gs, rotation T and angular
strain D
b) definitions of relative deflection ' and
deflection ratio '/L
c) definitions of tilt Z and relative rotation
(angular distortion) E
Figure 5. Definitions of foundation movements
and deformations of structures (CEN, 2004,
after Burland and Wroth, 1975)
6. CONCLUDING REMARKS
The work for the elaboration of a common
framework for geotechnical design
throughout Europe, i.e. Eurocode 7,
started over 30 years ago. Given the
progress achieved, the corresponding
standards/codes are now being enforced
in the various countries.
Whatever the precise legal status of
Eurocode 7 in the various countries, it will
prove to be very important for the whole
construction industry. It is meant to be a
tool to help European geotechnical
engineers speak the same technical
language and also a necessary tool for the
dialogue between geotechnical engineers
and structural engineers.
18
Eurocode 7 helps promote research.
Obviously, it stimulates questions on
present geotechnical practice from ground
investigation to design models.
It is our belief that it will also be very
useful to many geotechnical and structural
engineers all over the world, not only in
Europe.
REFERENCES
AFNOR (2009a) Calcul géotechnique -
Ouvrages de soutènement - Écrans,
norme NF P94-282, mars 2009, 182
pages.
AFNOR (2009b) Calcul géotechnique -
Ouvrages de soutènement - Remblais
renforcés et massifs en sol cloué,
norme NF P94-270, juillet 2009, 186
pages.
AFNOR (2012) Justification des ouvrages
géotechniques - Normes d’application
nationale de l’Eurocode 7 - Fondations
profondes, norme NF P 94-262, juillet
2012 (2e tirage, janvier 2013), 206
pages.
AFNOR (2013) Justification des ouvrages
géotechniques - Normes d'application
nationale de l'Eurocode 7 - Fondations
superficielles, norme NF P 94-261, juin
2013, 124 pages.
AFNOR (2014) Justification des ouvrages
géotechniques - Normes d'application
nationale de l'Eurocode 7 - Ouvrages
de soutènement – Murs, norme NF P
94-281, avril 2014, 93 pages.
Anagnostopoulos, A. and Frank, R. (2012)
EC7: Eurocode 7: an “umbrella code”
its implementation, evolution and
impact worldwide, Proc 6th Panhellenic
Conf on Geotechnical and
Geoenvironmental Engng (6°
Panhellinio Synedrio Geotechnikis kai
Geoperivallondikis Michanikis), Volos,
29 September-1st October 2010, vol. 4,
2012, pp. 170-178. .
Burland, J.B., Broms, B.B. and De Mello,
V.F.B. (1977). Behaviour of foundations
and structures. Proc. 9th Int. Conf. Soil
Mechs & Fdn Engng, Tokyo 2: 495-546.
Burland J.B. and Wroth C.P. (1975)
Settlement of buildings and associated
damage, Review Paper, Session V.
Proc. Conf. Settlement of Structures,
Cambridge: 611-654. Pentech Press,
London.
Burlon S., Frank R., Baguelin F., Habert
J., Legrand S. (2014). "Model factor for
the bearing capacity of piles from
pressuremeter test results. A Eurocode
7 approach", Géotechnique 64, No. 7,
513–525
[http://dx.doi.org/10.1680/geot.13.P.061
].
CEN (1994). Eurocode 7 Geotechnical
design - Part 1: General Rules. Pre-
standard ENV 1997-1. European
Committee for Standardization (CEN):
Brussels.
CEN (1999a). Eurocode 7 Geotechnical
design - Part 2: Geotechnical design
assisted by Laboratory Testing. Pre-
standard ENV 1997-2. European
Committee for Standardization:
Brussels.
CEN (1999b). Eurocode 7 Geotechnical
design - Part 3: Geotechnical design
assisted by Field Testing. Pre-standard
ENV 1997-3. European Committee for
Standardization: Brussels.
CEN (2002). Eurocode: Basis of structural
design. European standard, EN 1990 :
2002. European Committee for
Standardization: Brussels.
CEN (2004, 2013). Eurocode
7: Geotechnical design - Part 1:
General rules, EN 1997-1:2004 (E), (F)
and (G), November 2004, EN 1997-
1:2004/AC (corrigendum), February
2009, NF EN 1997-1/A1 (Amendment
for Section 8), 2013, European
Committee for Standardization:
Brussels.
CEN (2007). Eurocode 7: Geotechnical
design - Part 2: Ground investigation
and testing. EN1997-2:2007 (E), March
2007, EN 1997-2:2004/AC
(corrigendum), June 2010, European
Committee for Standardization:
Brussels.
19
EC (2003a). Guidance Paper L.
Application and use of the Eurocodes,
ref.: CONSTRUCT 03/629 Rev.1 (27
November 2003), European
Commission, Brussels.
EC (2003b). Commission recommendation
of 11 December 2003 on the
implementation and use of Eurocodes
for construction works and structural
construction products (2003/887/EC),
Official Journal of the European Union,
19.12.2003, EN, L 332/62 &63.
EC 7 (1990). Eurocode 7: Geotechnics.
Preliminary draft for the European
Communities, Geotechnik, 1990/1.
Frank R. (1991). Quelques
développements récents sur le
comportement des fondations
superficielles. Rapport général, Session
3, Comptes rendus 10ème Cong.
Européen Méca. Sols et Tr. Fond.,
Florence, 26-30 mai, vol. 3, pp. 1003-
1030. (English version: Some recent
developments on the behaviour of
shallow foundations. General report,
Proc. 10th European Conf. Soil Mechs
& Fdn Engng, Florence, 26-30 May,
vol. 4, pp. 1115-1141, 1994).
Frank R., Bauduin C., Driscoll R.,
Kavvadas M., Krebs Ovesen N., Orr T. ,
Schuppener B. (2004). Designer's
guide to EN 1997 Eurocode 7 –
Geotechnical design, Thomas Telford,
London, 216 pages.
Frank, R. & Magnan J.P. (1999).
Quelques réflexions sur la vérification
des états limites ultimes suivant
l’Eurocode 7 (in French - A few
thoughts about ultimate limit states
verifications following Eurocode 7).
Workshop on the Eurocodes, Proc.
12th European conf. soil mechs. &
geot. engng, 7-10 June, Amsterdam,
vol. 3: 2179-2183.
Huybrechts N., Patel D., De Vos M.
(2005). The use of the observational
method. Final Report WP3 on
Innovative design methods in
geotechnical engineering European
network 'GeoTechNet'.
20
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Modern Design of Landfill Liners

Andrea DOMINIJANNI1, Mario MANASSERO1*, Giacomo BOFFA1

1
Politecnico di Torino, Department of Structural, Geotechnical and Building Engineering,
Torino, ITALY

ABSTRACT
The design of landfill bottom barrier systems on the basis of their performances requires the
modelling of the pollutant transport processes, while taking into consideration both advective and
dispersive-diffusive phenomena. This modelling can be developed in the context of an
environmental and health risk assessment, which allows the specific vulnerability of the site to be
taken into account. A calculation procedure, based on closed-form analytical solutions, which can
be considered suitable for a second-level risk analysis type, according to the ASTM and USEPA
classifications, is presented in this paper. This procedure, although based on a very simple and
intuitive conceptual model, is able to provide useful indications for design choices, at least in the
preliminary dimensioning phase of a project.

Keywords: landfill, liner, aquifer, risk analysis, analytical solution

The main problem in the environmental

1. INTRODUCTION
Geotechnical engineers first began to
dedicate their main efforts towards
problems related to the protection of the
environment from anthropological pollution
in the 1970's. A remarkable impetus in the
development of studies on the interaction
between soils and polluting substances
emerged from the need to manage the
disposal of radioactive materials produced
by nuclear stations with solutions that
would be able to guarantee an adequate
level of safety in the long term (Daniel,
1993). Attention was soon extended to
industrial waste landfills and to the
pollution of the sub-soil by products from
chemical and petrochemical industries.
field that geotechnical engineers were
called on to resolve was that of designing
pollutant barrier systems. Initially, the
materials that were used to make up these
systems were in fact of a mineral type
(drainage layers in sand and gravel,
compacted clay liners, soil-bentonite and
cement-bentonite cutoff walls), as it was
believed that they were characterized by a
suitable durability of their properties for
long-term use. An important innovation
that modified the approach to the design
of these structures was the introduction of
the so-called geosynthetic materials in the
1980's. Mineral materials were integrated
with polymeric materials, thus obtaining
much higher performances with lower

* presenting author
realization costs (Koerner, 1986; Rowe,
1998).
Although the basic expertise in the field
of soil mechanics resulted to be very
useful for dealing with many
environmental applications, the
geotechnical engineers that dedicated
themselves to this issue had to explore
new theoretical problems, such as the
study of solute diffusion phenomena in
soils, chemical osmosis in bentonites or
the effect of vacuum application on the
speed of volatile pollutant extraction from
the subsoil (Manassero et al., 2009;
Shackelford, 2014).
Apart from these fundamental novelties
in the field of theoretical studies, it is also
important to mention the development of
specific design methodologies, whose
distinctive features derive from the specific
nature of the objective that has to be
reached, of the materials that are used, of
the boundary conditions and of the
reference regulations.
Risk analysis is one example of these
new methodologies, and an application is
illustrated in this paper which has the aim
of preliminarily determining the
performances of landfill bottom barrier
systems related to the vulnerability of
aquifers located underneath controlled
disposal sites.
2. RISK ANALYSIS FOR LANDFILL
DESIGN
The necessity of determining the targets of
remediation systems for polluted sites has
led to the development of an analysis
methodology for the assessment of the
corrective actions to be conducted in order
to protect human health and environment.
Although this analysis methodology refers
explicitly to the (probabilistic) risk concept,
this procedure can also be developed
through a deterministic approach, in a
similar way to pseudo-probabilistic
methods in other engineering applications
(e.g. structural engineering).
The risk that affects human health and
environment in geoenvironmental
22
applications can be defined as the
quantification of the toxicological damage
produced on man or on the environment
due to the effects of the presence of a
polluting source, whose releases can
reach, through different migration paths, a
potentially exposed receptor. The
evaluation of the environmental and health
risk therefore postulates the quantitative
definition of the source-path-target
relational system (Di Molfetta and Sethi,
2012).
Risk itself is defined as the product of
the probability of a determined event
taking place times the consequences
caused by the event occurring, and it is
generally quantified in terms of costs. As it
is difficult to quantify the consequences, it
is normal practice, in many engineering
applications, to make do with the
verification of the probability of the event
being lower than an established threshold
value. In many cases, the analysis is
carried out according to a conservative
deterministic approach, and it simply has
the aim of verifying the respect of the
threshold condition, the exceeding of
which involves a risk that is conventionally
considered unacceptable.
Within most of the regulations and
related guidelines (e.g. ASTM, 1995,
1998; USEPA, 1989, 1991a, 1996;
APAT/ISPRA, 2008) risk analysis
methodology is based on three levels of
analysis (a 3-tiered approach):
x the first level (Tier 1) consists of a
comparison of the concentrations of the
potential pollutants present in the site with
the threshold values stated in the
regulations in force.
x the second level (Tier 2) is based on
the adoption of simplified transport
conceptual models, to which it is possible
to associate mathematical models that can
be resolved in a closed-form analytical
manner. The simplicity of the analysis can
be justified by the adoption of
conservative boundary conditions and
parameter values. It is important to note
that the objective of these models is not to
supply a detailed forecast of the
phenomena, but to ascertain the existence
of a potential risk.
x the third level (Tier 3) represents the
most detailed part of the analysis, in which
the most representative and therefore
complex conceptual models of the
physical reality are adopted. These
conceptual models can generally be
associated to mathematical models that
cannot be resolved analytically, but can
only be resolved numerically. The
possibility to reach this level of analysis
depends on the availability of chemical,
physical and biological data of the
investigated site, as well as of an accurate
hydrogeological characterization. In this
context, the aim of the analysis is to
provide a forecast, whose reliability
depends on the uncertainties in the
experimental determination of the
parameters.
The basic conceptual scheme of all the
aforementioned levels of analysis is
relatively simple and is founded on the
identification of:
1. a pollution source, which represents the
source of the risk;
2. the migration paths, that is, the
mechanisms that can determine the
movement of the pollutants from the
source;
3. an exposure point, which represents
the target of the risk determined by the
pollution.
A first risk assessment can be
conducted by comparing the expected
concentration values at the exposure point
with the threshold values established in
the regulations in force (maximum
contaminant level, MCL).
Therefore, for a given substance, one
has:
c PoE d MCL (1)
where:
c PoE = the concentration at the point of
exposure (mol/m3 or kg/m3).
In the case of landfills, it is possible to
assume, as a first step, that the exposure
point is located within the aquifer below
23
the landfill, just downstream the landfill
footprint.
The verification required by the risk
analysis procedure therefore involves
evaluating whether the concentration of
the leachate, c0, which is expected to be
larger than the MCL, is reduced at the
exposure point because of attenuation
phenomena that occur in the migration
process, until lower values than the MCL
are reached.
Hereafter, reference will only be made
to the criterion defined by equation (1) to
establish the acceptability or non-
acceptability of the risk.
2.1. Thin aquifer
Adopting the risk analysis procedure, even
in its simplest formulation referring to the
second level of analysis (Tier 2), it is
possible to obtain an indication of the
efficiency of the landfill barrier system, by
comparing, for example, the concentration
at the exposure point with the MCL value
under steady state conditions (Manassero
et al., 2000; Olinic et al., 2002; Guyonnet
et al., 2001).
In the case in which the aquifer
thickness is small enough not to allow any
significant concentration variation along
the vertical section of the aquifer, but only
along the longitudinal flow of groundwater
below the landfill (the x direction), as
shown in Figure 1, the water volumetric
balance inside an element of infinitesimal
volume within the aquifer is given by (see
Figure 2):
dqh ˜ haq q ˜ dx (2)
from which, through separation of the
variables and integration
qh x
q
³
qh0
dqh
haq ³0
˜ dx ,
we obtain
x
qh qh0  q . (3)
haq
where:
qh = groundwater horizontal volumetric
flux in the aquifer below the landfill (m/s);
 d

"
haq ˜ (qhc x ) ˜ dx Js ˜ dx (4)
dx
where Js is the vertical mass flux of the
WASTE solute coming from the landfill.
From equation (4):
haq dc x dq
haqqh  haqc x h Js
x=0 x=" dx dx
dx , (5)
dc
haqqh x  q ˜ c x Js
Figure 1. Sketch of a thin aquifer below a dx
landfill with identification of a representative dq q
volume of length dx. since h from equation (2).
dx haq
The vertical solute mass flux under
steady-state conditions can be obtained

q ˜ dx  d(qh ˜ haq ) from the following equation (Manassero et
qh ˜ haq  dx al., 2000):
dx
dq c 0 ˜ ePL  c x
qh ˜ haq  qh ˜ haq  h haqdx Js q (6)
dx ePL  1
haq
where PL is the Peclet number of a
multilayer barrier and cx is the solute
concentration at the bottom of the multi-
dx 
layer barrier.
Figure 2. Water volume balance in an element The Peclet number of the multilayer
of length dx of the aquifer. barrier can be expressed as follows
(Manassero et al., 2000):
q
PL (7)
qh0 = groundwater horizontal volumetric /
flux in the aquifer upstream the landfill where:
(m/s);
/ = equivalent diffusivity of the multilayer
q = vertical water volumetric flux coming system (m/s).
from the landfill (m/s);
The / parameter is given by:
haq = thickness of the aquifer (m).
1
The evaluation of the solute mass / (8)
N
Li
balance, again referring to the element of ¦ n
i 1 i iDhi
S
the aquifer of length dx, is obtained from
(see Figure 3): where:
Li = thickness of the i-th layer (m);
Si = saturation degree of the i-th layer (-);

Js ˜ dx d(qhhaqc x )
ni = porosity of the i-th layer (-);
qhhaqc x  dx Dhi = hydrodynamic dispersion coefficient
dx
d(qhc x ) of the i-th layer (m2/s).
qhhaqc x qhhaqc x  haqdx
dx The Dhi coefficient is determined from
ha the sum of two contributions, that is, the
mechanical dispersion coefficient (Dmi)
and the effective diffusion coefficient ( Di* ):
dx
Dhi Dmi  Di* (9)
Figure 3. Solute mass balance in an element
of length dx of the aquifer.

24
 The mechanical dispersion coefficient
can be expressed as a function of the
seepage velocity inside the multybarrier
q
system, vi , in the following way:
ni
D mi Di ˜ v i (10)
where:
Di = longitudinal dispersivity coefficient
(m), which, in the absence of data
obtained from laboratory or in situ tests,
can be assumed equal to a tenth of the
thickness of the barrier layer (i.e.
Di # 0.1˜ Li ).
The effective diffusion coefficient can
be expressed as a function of the free
solution diffusion coefficient through the
following relation:
Di* Wi ˜ D0 (11)
where:
Wi = tortuosity factor (-);
D0 = free solution diffusion coefficient
(m2/s).
When the seepage velocity tends to
zero, as in the case of a low permeability
mineral layer underlying a geomembrane,
it is possible to assume Dhi | Di as a first
* value in the lower layer, k2, which is
approximation.
The volumetric flow, q, in the case of a
multilayer mineral barrier, can be
expressed in the following way:
hp  L  hb
q k eq (12)
L
where:
keq = the equivalent hydraulic conductivity
of the multilayer system (m/s);
hp = height of ponded leachate above the
barrier (m);
hb height of water level at the bottom of
the barrier (m);
L = total thickness of the barrier (m).
The equivalent hydraulic conductivity
can be calculated as the harmonic mean
of the hydraulic conductivity of the
individual layers:
25
L
k eq (13)
N
Li
¦
i 1 ki
where:
ki = the hydraulic conductivity of the i-th
layer (m/s);
N = the number of mineral layers that
make up the barrier (-).
The total thickness of the barrier is
obtained by summing the thicknesses of
the individual layers, that is:
N
L ¦L .
i 1
i (14)
In the case of overlapping of a layer
with low hydraulic conductivity (layer 1)
above a layer with higher hydraulic
conductivity (layer 2), equation (12) is only
valid if saturated flow conditions (or, more
precisely, positive hydraulic pressure
conditions) are verified. If the difference in
hydraulic conductivity between the two
layers is significant, unsaturated flow
conditions can be reached in the more
permeable layer below (Giroud et al.,
1997).
The maximum hydraulic conductivity
necessary to maintain saturated flow
conditions, is equal to (Giroud et al.,
1997):
§ hp · § L 2 ·
max(k 2 ) k1 ˜ ¨ 1  ¸˜¨ ¸. (15)
© L1 ¹ © L 2  hb ¹
where k1 and L1 represent the hydraulic
conductivity and the thickness of the
upper layer, respectively, L2 is the
thickness of the lower layer and hb is the
height of the water level at the bottom of
the lower layer.
When the hydraulic conductivity of the
lower layer is higher than the value given
by equation (15), the flow through the
barrier, neglecting capillary effects, is only
determined by the properties of the upper
layer, and it can be calculated as follows
when L2 > hb (Giroud et al., 1997):
hp  L1
q k1 . (16)
L1
 In the case in which the capillary
phenomena inside the second layer are
not negligible, the volumetric flux falls
between the minimum values given by
equation (16) and the maximum value
obtained from equation (12).
In the presence of a geomembrane,
placed above a mineral layer, the
evaluation of the volumetric flux through
the composite barrier requires a specific
treatment. As geomembranes are made
up of polymeric sheets with very low
permeability to water (< 1˜1012 m/s), the
flow occurs above all through the defects,
wrinkles or holes, created during the
installation procedures and the
subsequent working operations.
The volumetric flux through a circular
hole, with or without wrinkles, can be
assessed using analytical solutions
(Rowe, 1998; Rowe, 2005; Rowe et al.,
2004) or semi-empirical ones (Giroud,
1997). As analytical solutions require the
determination of parameters, such as the
hydraulic transmissivity between the
geomembrane and the underlying mineral
layer, which are difficult to evaluate, it may
be easier in a preliminary analysis to
resort to semi-empirical solutions.
Giroud (1997), on the basis of the
analysis of numerous experimental
investigations, proposed the following
equation for the determination of the flow
rate through an individual circular hole:
ª §h ·
0.95
Q Cq ˜ aF0.1 ˜ hp0.9 ˜ k0.74 ˜ «1  0.1˜ ¨ p ¸
«¬ ©L¹
where:
Q = flow rate through a single circular area
hole (m3/s);
Cq = quality coefficient of the contact
between the geomembrane and the
underlying mineral layer (-);
aF = area of the circular hole (m2);
hp = height of leachate in the drainage
layer above the composite barrier (m);
k = hydraulic conductivity of the mineral
layer (m/s);
L = thickness of the mineral layer (m).
The Cq coefficient takes into account
the contact conditions that occur between
the geomembrane and the underlying
mineral layer. Giroud (1997) suggested
the following values for such a coefficient:
­0.21 good contact conditions
Cq ® (18)
¯1.15 poor contact conditions
Once the flow rate through a single hole
has been determined, it is possible to
obtain the volumetric flux on the entire
landfill area by multiplying Q by the
number of holes per unit of area (generally
expressed in m2 or hectares):
q nF ˜ Q (19)
where:
nF = the number of holes per unit of area
(1/m2).
On the basis of observations conducted
on 6 case studies, Giroud and Bonaparte
(1989) recommended adopting a hole
area equal to 1 cm2 (20 mm u 5 mm) in
order to evaluate the flux through a
composite barrier in a conservative
manner for the case of landfills installed
with elevated controls during the working
operations.
The number of holes per hectare
depends on the quality of the installation
and on the controls conducted during the
subsequent phases of installing the layers
above the geomembrane. The American
EPA indicates values of between 2.5 and
75 holes/hectare as possible interval
º values, that is, from 1 to 30 holes/acre
» (17)
»¼ (USEPA, 1991b).
Coming back to equations (5) and (6),
combining them and rearranging, the
following first-order differential equation is
obtained for the solute concentration, cx,
inside the aquifer versus the horizontal
distance, x:
dc x § q ·
 cx ˜ F ˜ ¨
dx ¨ qh0haq  q ˜ x ¸¸
© ¹
, (20)
§ q ·
c0 ˜ F ˜ ¨
¨ q h  q ˜ x ¸¸
© h0 aq ¹
where
26
 ePL 1 cPoE c aq  (c 0  c aq0 ) 
F . (21)
e 1 PL
1  e PL 1
(26)
§ K · 1exp( PL )
Equation (20) can be solved by adding (c 0  c aq0 ) ¨ ¸
a particular integral of the equation to the © 1 K ¹
solution of the associated homogeneous where:
equation. qh0haq
The associated homogeneous equation K . (27)
q˜"
can be solved by using the method of
separation of variables: This parameter takes into consideration
the vulnerability of the aquifer, which
dc x § q ·
results to be as high as the aquifer velocity
³ cx
F ˜ ³ ¨
¨ qh0haq  q ˜ x ¸¸
˜ dx
qh0 is low.
© ¹
ln(c x ) F ˜ ln(qh0haq  q ˜ x)  C1 The efficiency of the landfill lining
system can be checked by comparing the
from which
theoretically assessed concentration at the
cx C2 ˜ (qh0haq  q ˜ x)F , (22) PoE (cPoE) with the maximum acceptable
where C2 is an integration constant that contaminant concentration, MCL as
must be determined as a function of the established by the in force local rules and
shown in equation (1).
initial condition c x (x 0) caq0 , being caq0
In the case of the presence of a blank
the blank groundwater concentration concentration of a given contaminant in
coming from landfill upstream. the groundwater upstream the landfill, caq0,
By adding the particular integral the limit condition given by equation (1)
c x c0 (23) can be also expressed in terms of relative
concentration (RC) or attenuation factor
to equation (22), the complete solution is
(AF) that are defined as follows:
obtained:
1 c PoE  c aq0
cx C2 ˜ (qh0haq  q ˜ x)F  c 0 . RC (28)
AF c 0  c aq0
The integration constant C2 is
and therefore:
determined by means of the upstream
boundary condition: MCL  c aq0
RC d , (29)
F c 0  c aq0
c x (x 0) C2 ˜ (qh0haq )  c 0 c aq0
c 0  c aq0
from which AF t . (30)
MCL  c aq0
c 0  c aq0
C2  . (24) On the basis of equation (26), the
(qh0haq )F
relative concentration at the PoE is equal
Therefore, the final solution can be to:
expressed as follows: 1

c x (x) c aq  (c 0  c aq0 )  1 § K ·1exp( PL )

RC 1 ¨ ¸ . (31)
1 AF © 1 K ¹
§ qh0haq · 1 exp(  PL ) . (25)
(c 0  c aq0 ) ¨ The attenuation factor results to be a
¨ q h  q ˜ x ¸¸
© h0 aq ¹ growing function of parameter K, for which
If we assume that the exposure point, the dependence on the Peclet number is
only significant when PL < 4 (see Figure
PoE, is located at x = ", that is, the first
4).
point of the aquifer downstream the landfill
footprint, the following expressions are
obtained for the concentration at the
exposure point, cPoE:

27
 1.E+6 K Although solute concentration is varying
105 with depth continuously, so that a true
1.E+5
plume border cannot be determined
Attenuation Factor, AF

1.E+4
104 rigorously, Charbeneau et al. (1995)
proposed a theoretical expression to
103
1.E+3 estimate the depth of penetration of the
102
contaminant into the aquifer beneath the
1.E+2
landfill:
1.E+1
10
hp (x) hp,adv (x)  hp,dis (x) (32)
1.E+0 ª § q ˜ x ·º
haq «1  exp ¨  (33)
¨ qh0haq ¸¸ »»
1.E-5 1.E-4 1.E-3 1.E-2 1.E-1 1.E+0 1.E+1 1.E+2 1.E+3 hp,adv (x)
Peclet number, PL ¬« © ¹¼
Figure 4. Attenuation factor as a function of
the Peclet number and the dimensionless hp,dis (x) 2Dv x (34)
parameter K for the solution for thin aquifers. where
hp = plume thickness, varying in the
horizontal distance beneath the landfill
(see Figure 5);
2.2. Thick aquifer
hp,adv = penetration depth due to vertical
The previously developed basic solution
advection of water entering the aquifer;
(Manassero et al. 2000) can be applied to
the cases in which an aquifer of limited hp,dis = penetration depth due to vertical
thickness is found below the landfill, under dispersion in the aquifer, assumed to be
the assumption that the solute is one standard deviation of the
distributed in a homogeneous manner concentration distribution;
within each vertical section of the aquifer. Dv = vertical dispersivity of the aquifer.
In the case in which the aquifer has a If the linear variation of groundwater
significant depth (e.g. a greater depth than horizontal velocity beneath the landfill,
1/100 of the length of the landfill), the given by equation (3) is taken into
vertical distribution of the solute inside the account, then the advective component of
aquifer cannot be neglected. In fact, the the plume thickness can be determined as
concentration of the solute coming from follows:
the landfill tends to reduce with depth and
to form what is known as a contaminant § qh0haq ·
hp,adv (x) haq ˜ ¨ 1  . (35)
“plume”, whose depth depends on the ¨ qh0haq  q ˜ x ¸¸
© ¹
velocity of groundwater and vertical
Moreover, if the contribution to the
dispersion (Figure 5).
plume thickness due to the vertical
dispersion of the contaminant, induced by
the horizontal flow component of the
groundwater, is assessed by a rigorous

" CONTAMINANT
mass balance in a semi-infinite aquifer,
PLUME then the following equation is obtained:
WASTE
BARRIER 4
hp,dis Dv x . (36)
Zonadi
S
haq > "/100 ACQUIFERO
hp(x)
Equations (35) and (36) can be used as
x
an alternative to equations (33) and (34)
proposed by Charbeneau et al. (1995).
Figure 5. Sketch of a landfill above a thick
aquifer. Function hp(x) describes the variation In order to obtain a first estimate of the
in the depth of the contaminant plume below risk that is related to the contaminant
the landfill barrier. penetration into a thick aquifer, it is
28
possible to equate the plume to a perfect where the term in the right-hand side of
mixture zone on the inside of which the equation (39) represents the contaminant
contaminant concentration is constant. vertical mass flux coming from the landfill.
The concentration inside the mixing Since transverse concentration
zone (i.e. the contaminant plume) can be gradients greatly exceed longitudinal
determined from the expression of the gradients and qh0 is expected to be much
concentration in the aquifer cx, which was larger than q, equation (37) can be
obtained in the previous section (i.e. reduced as follows:
equation 25) by multiplying it by the haq/hp wc w 2c
ratio. qh0 naqDh,v . (40)
wx wy 2
This procedure is equivalent to evaluate
The vertical hydrodynamic coefficient is
the contaminant average concentration in
expected to be controlled by the
the aquifer assuming an infinite dispersion
transverse dispersion induced by the
in the vertical direction and, then, to
horizontal water velocity, so that it can be
redistribute the contaminant mass in such
assumed:
a way to concentrate it into the estimated
plume thickness. qh0
Dh,v Dv . (41)
In order to obtain a more accurate naq
evaluation of contaminant distribution, the If equation (41) is combined with
bi-dimensional transport that occurs in the equation (40), the mass balance can be
aquifer needs to be considered. Under written as follows:
steady state conditions, the contaminant wc w 2c
mass balance in a semi-infinite aquifer can Dv , (42)
wx wy 2
be expressed as follows:
wc w 2c w 2c
which is a standard diffusion equation.
naq naqDh,h  naqDh,v 2  An analytical solution can be found for
wt wx 2
wy
(37) equation (42) and the boundary conditions
wc wc
qh0 q 0 given by equations (38) and (39) (Carslaw
wx wy and Jaeger, 1959; Crank, 1975):
where: c(x,y)  c aq0 § y ·
c c(x,y) = contaminant concentration in RC erfc ¨ ¸
c 0  c aq0 ¨2 D x ¸
the aquifer as a function of the horizontal © v ¹
distance (x) and the vertical depth (y); 
 exp J ˜ y  J 2D v x ˜  (43)
naq = aquifer porosity; § y ·
Dh,h = horizontal hydrodynamic dispersion ˜erfc ¨  J Dv x ¸
¨2 D x ¸
coefficient of the aquifer; © v ¹
Dh,v = vertical hydrodynamic dispersion where:
coefficient of the aquifer; q ePL
J . (44)
qh0 = horizontal water Darcy’s velocity in Dv qh0 (ePL  1)
the aquifer; The results, in terms of RC trend, of the
q = vertical water Darcy’s velocity in the proposed closed form solution analytical
aquifer, which is equal to the water vertical model are compared with the results of a
volumetric flux coming from the landfill. two dimensional numerical code
The mass balance given by equation (MIGRATE, 1995) in Figure 7.
(37) is related to the following boundary It is possible to appreciate the good
conditions: approximation of the analytical solution
c c aq0 at x 0 (38) with the well-established numerical one, in
the investigated range of q/qh0 values,
wc c 0ePL  c lower than 0.01. Moreover, the use of the
q ˜ c  naqDh,v q at y 0 (39)
wy ePL  1 final equations of the proposed approach
29
is very simple and direct, giving the Geomembrane Layer (GML)

possibility to appreciate, in a rapid and 0.5 m Drainage Layer

Compacted Clay Layer (CCL)
effective way the influence of the different 1m
k = 1˜10-9 m/s; n = 0.4; W = 0.3 3m
input parameters.
Finally, the use of this model within 3m Attenuation Layer (AL)
k = 1˜10-7 m/s; n = 0.3; W = 0.4
probabilistic approaches for risk
assessment is strongly encouraged since
it is easy, cost and time effective, also Aquifer (AQ) (a)

when it is coupled with very complex

algorithms for statistical analyses that
adopt sophisticated random distributions Geomembrane Layer (GML)
0.5 m Drainage Layer
of the input data.
Geosynthetic Clay Liner (GCL)
k = 1˜10-11 m/s; n = 0.7; W = 0.1 3m
3. EXAMPLES
4m Attenuation Layer (AL)
In order to validate the proposed k = 1˜10-7 m/s; n = 0.3; W = 0.4

procedure from the practical point of view,

the performance of two different bottom Aquifer (AQ) (b)
barriers has been assessed.
Figure 6. Scheme of two different barriers: (a)
The first barrier is constituted of a
Compacted Clay Layer (CCL) that can
compacted clay liner (CCL) 1 m thick and underlies a Geomembrane (GML); (b)
with hydraulic conductivity equal to 1˜109 Geosynthetic Clay Liner (GCL) that can
m/s, while the second barrier is a underlies a Geomembrane (GML).
geosynthetic clay liner 1 cm thick and with
hydraulic conductivity equal to 1˜1011 m/s.
The volumetric flow, q, in the case of a
The vertical profile of the two alternative multilayer mineral barrier (e.g. CCL+AL,
barriers is reported in Figure 6. In both GCL+AL) can be calculated using
cases, the mineral barrier underlies a expression (12), in which the equivalent
drainage layer with a ponded leachate hydraulic conductivity can be expressed
having head equal to 0.5 m and overlies as the harmonic mean of the hydraulic
an attenuation layer (that can be natural or conductivity of the individual layers.
artificial) having hydraulic conductivity
equal to 1˜107 m/s. The hydraulic head
Table 1. Characteristic properties of the
difference between the top of the mineral
example barriers.
layer and the bottom of the attenuation
layer, 'h hp + L  hb, is assumed equal Free solution diffusion D0 = 20.3˜10-10 m2/s
to 3 m. coefficient
The aquifer below the barrier systems Volumetric flow in the qho = 31.6 m/y
is characterized by a horizontal volumetric aquifer
flow, qh0, of 31.6 m/y (Table 1) and the Hydraulic conductivity
length of the landfill footprint in the kaq = 1˜10-4 m/s
of the aquifer
groundwater flow direction, ", is taken Hydraulic gradient of
equal to 1000 m. iaq = 1%
the aquifer
The risk analysis procedure is developed
Landfill length "= 1000 m
considering two different thicknesses of
the aquifer, i.e. a thin aquifer (haq = 3 m) Number of holes of
and a thick aquifer (haq = 100 m), in order the geomembrane nF = 20 holes/hectare
to determine the RC and/or AF at an per unit area
exposure point located downstream the Area of the circular
landfill. aF = 1 cm2
hole

30
 The difference between the volumetric
fluxes of the two barriers is negligible (i.e.
q = 2.91˜10-9 m/s for CCL+AL barrier, q =
2.88˜10-9 m/s for GCL+AL barrier).
A significant reduction of the vertical
volumetric flux is obtained if a
geomembrane is used. In the presence of
a geomembrane, placed above the CCL
or the GCL, the volumetric flow can be
determined combining equations (17) of
Giroud (1997) and (19). Assuming good
contact conditions (i.e. Cq = 0.21) between
the geomembrane and the underlying
CCL, and excellent contact conditions (i.e.
Cq = 0.096, as suggested by Touze-Foltz
and Barroso, 2006) between the
geomembrane and the GCL, the
volumetric fluxes, calculated assuming 20
circular holes/hectare and an area of the
holes equal to 1 cm2, decrease of more
than two orders of magnitude (i.e. q =
2.06˜10-11 m/s for GML+CCL+AL barrier, q
= 1.52˜10-12 m/s for GML+GCL+AL
barrier).
The volumetric flow, q, is used to
calculate the Peclet number, PL, of the
barrier, adopting equation (7), in which the
equivalent diffusivity of the multilayer
system, /, is given by equation (8). For
each layer, the hydrodynamic dispersion
coefficient Dh can be determined from the
sum of the mechanical dispersion
coefficient, Dm, and the effective diffusion
coefficient, D*, using equations (9)-(11).
Tortuosity factors and porosities of
different layers are reported in Figure 6.
In the case of the CCL+AL system, the
Peclet number, PL is equal to 13.3, while it
is equal to 8.5 for the GCL+AL system. If a
geomembrane overlies the mineral layer,
the Peclet number, PL decreases to 0.33
for the GML+CCL+AL system, and to
0.025 for the GML+GCL+AL system.
The Relative Concentration, RC, and
the Attenuation Factor, AF, can be
determined in different ways depending
from the different procedures of risk
analysis.
If we consider a thin aquifer (i.e. 3 m
thick), RC and/or AF can be calculated
through equation (31).
31
In the case of deep aquifer (i.e. 100 m
thick), a first assessment of RC and/or AF
within the contaminant plume can be
obtained by multiplying the result of
equation (31) by the (haq/hp) ratio.
In Table 2 the RC values that have
been calculated for the different cases are
reported.
For the case of thick aquifer, the RC
values calculated by equation (31) are
reported as RCthin-aq, while the RC values
obtained by multiplying equation (31) by
the (haq/hp) ratio are reported as RCthick-aq.
In the case of thin aquifer, the uni-
dimensional aquifer model provides an RC
value for the CCL+AL system equal to
RCthin-aq= 0.4926, which is close to that
obtained for the GCL+AL system (i.e.
RCthin-aq = 0,4902).
In the case of thick aquifer, the bi-
dimensional procedure gives a RCthick-aq
value, which is more than double of the
RCthin-aq value that has been calculated for
haq 100 m neglecting the vertical
distribution of the contaminant
concentration, i.e. using the solution for
thin aquifer given by equation (31) (for the
CCL+AL system, RCthick-aq = 0.0735 and
RCthin-aq = 0.0283; for the GCL+AL
system, RCthick-aq = 0.0730 and RCthin-aq =
0.0280).
The presence of the geomembrane
gives RC values which are even lower
than the previous cases: in the case of
thin aquifer for the GML+CCL+AL system,
RCthin-aq = 0.0240, while for the
GML+GCL+AL system RCthin-aq = 0.0203;
in the case of thick aquifer for the
GML+CCL+AL system, RCthick-aq = 0.0020
and RCthin-aq = 0.0007, for the
GML+GCL+AL system, RCthick-aq = 0.0017
and RCthin-aq = 0.0006.
The simplified approach proposed in
the present paper lets to obtain a
conservative, but rather reliable,
evaluation of the relative concentration
distribution of the contaminant within a
vertical planar section of the aquifer.
Table 2. Results of the example calculations (RCthin-aq = Relative Concentration for thin aquifer
case; RCthick-aq = Attenuation Factor for thick aquifer case; Cq quality coefficient of the contact
between the geomembrane and the underlying mineral layer, haq aquifer thickness; q vertical
volumetric flux through landfill barrier; PL = Peclet number of landfill barrier).

COMPACTED CLAY GEOSYNTHETIC CLAY

LAYER + LINER +
COMPACTED GEOSYNTHETIC GEOMEMBRANE GEOMEMBRANE
CLAY LAYER CLAY LINER
Cq = 0.21 Cq = 0.096
q = 2.91˜109 m/s q = 2.88˜109 m/s (good contact) (excellent contact)
PL = 13.3 PL = 8.5 q = 2.06˜1011 m/s q = 1.52˜1012 m/s

PL = 0.33 PL = 0.025

THIN AQUIFER
RCthin-aq = 0.4926 RCthin-aq = 0.4902 RCthin-aq = 0.0240 RCthin-aq = 0.0203
haq = 3 m

THICK AQUIFER RCthick-aq = 0.0735 RCthick-aq = 0.0730 RCthick-aq = 0.0020 RCthick-aq = 0.0017
Average concentration
within the reference
thickness of the plume hp(x)
hp(x = ") = 38.5 m hp(x = ") = 38.5 m hp(x = ") = 35.7 m hp(x = ") = 35.7 m
at the PoE

haq = 100 m
(Dv = 1 m) RCthin-aq = 0.0283 RCthin-aq = 0.0280 RCthin-aq = 0.0007 RCthin-aq = 0.0006

A more accurate evaluation of Landfill barriers have the purpose of

contaminant concentration can be containing the contaminating substances
obtained from the analytical solution given that are produced by waste and, as a
by equation (43). The results, in terms of consequence, their performances must be
RC varying in the aquifer with the quantified in terms of ability to control
horizontal distance x and the vertical solute migration.
depth y, of the proposed closed form
solution analytical solution, for the cases
of CCL+AL and GML+CCL+AL barriers,
have been compared with the results of a
two dimensional numerical code 0
Relative Concentration, RC
0.05 0.1 0.15
Relative Concentration, RC
-0.001 0 0.001 0.002 0.003
(MIGRATE, 1995), as shown in Figure 7. If 0
x = 10 m
0
x = 10 m
we consider the case of x=1000 m for the 10 10

CCL+AL system, the analytical solution 20 20

x = 1000 m
gives RC = 0.097 at y = 0 m and RC lower x = 1000 m

than 0.03 for depths larger than the 30

x = 500 m
30
Aquifer Depth, y (m)
Aquifer Depth, y (m)

x = 500 m
estimated plume thickness (i.e. hp = y = 40 40

38.5 m), while for the GML+CCL+AL 50 50

system RC = 0.0025 at y = 0 m and RC 60 60

lower than 0.0008 for depths larger than
the estimated plume thickness (i.e. hp = y 70 70

= 35.7 m). 80 80

MIGRATE MIGRATE
90 90
4. CONCLUSIONS ANALYTICAL
SOLUTION (a)
ANALYTICAL
SOLUTION (b)
100 100

The design of landfill bottom barrier Figure 7. Two dimensional analytical solution,
systems is an example of an applicative given by equation (43), and numerical solution
problem in which it is necessary to adopt obtained by MIGRATE (1995) for the following
new and specific analysis methods. cases: (a) CCL+AL barrier; (b) GML+CCL+AL
barrier.
32
 This evaluation requires basic
knowledge of the transport phenomena in
the subsoil, and the adoption of a
conceptual model that takes into account
the migration paths towards possible
points that are exposed to a risk for
human health.
A risk analysis, which allows the
efficiency of the barrier system to be
quantified, in terms of attenuation of the
risk at an exposure point located
downstream the landfill, has been applied
in this paper to the specific problem of the
evaluation of the performances of the
bottom barrier of a landfill.
The illustrated calculation procedure
should be considered as a second-level
one (Tier 2), i.e. it is based on a simple
conceptual model that leads to closed-
form analytical solutions. It is necessary to
point out that this is a very conservative
procedure that does not take into
consideration important attenuation
factors, such as the finite mass of
contamination produced by a given mass
of waste or the sorption or degradation
phenomena that can occur during the
migration towards a point of exposure. In
some cases, the adoption of more
advanced conceptual models could be
indispensable, such as those, for example,
that take into consideration transient
conditions, in order to obtain results more
close to the physical reality. On the other
hand, the risk analysis methodology
considers a third level of analysis (Tier 3),
in which the transport phenomenon is
modelled in detail through resort to
numerical type solutions. A presentation of
advanced models for the study of the
performances of a landfill bottom liner can
be found in Rowe et al. (2004), who
describe transient solutions in which the
uni-dimensional solute migration from the
landfill in a vertical direction to the
underlying aquifer is associated with a
rigorous bi-dimensional numerical analysis
of solute migration within the aquifer itself.
Another aspect that could be important
to point out is the possibility of applying in
a simple way the proposed procedure not
33
only to a deterministic analysis, but also to
a probabilistic analysis in which the
boundary conditions and the parameters
of the model have a random nature. A
significant difficulty in the application of the
procedure in fact arises from the
uncertainty that is encountered in the
evaluation of different parameters, such
as, for example, the leachate
concentration, the hydraulic conductivity of
the mineral layers and the number of
holes per hectare in the geomembrane. In
the deterministic approach, the designer
must trust in his own good judgement to
make the most opportune choices of the
values that have to be assigned to the
parameters, but cannot, however, know
the combined effect of the variability of the
various parameters on the result of the
analysis. For this reason, the adoption of a
probabilistic approach can be considered
a useful tool to help understand the
studied phenomena and the effective
representativeness of the results. In this
approach, the parameters affected by
uncertainty are considered random
variables and a probability distribution is
associated to them. The analysis can be
conducted using approximate analytical
methods (e.g. the First Order Second
Moment method) or numerical methods
(e.g. the Monte Carlo method) and it
makes it possible to evaluate, for example,
the probability of exceeding a threshold
concentration value at a point of exposure.
REFERENCES
APAT/ISPRA (2008). Criteri metodologici
per l’applicazione dell’analisi assoluta di
rischio ai siti contaminati.
ASTM (1995). Emergency Standard Guide
for Risk Based Corrective Actions
Applied at Petroleum Release Sites,
Report E-1739-95.
ASTM (1998). Standard Provisional Guide
for Risk-Based Corrective Action,
Report PS104-98.
Carslaw, H.S., and Jaeger, J.C. (1959).
Conduction of heat in solids, Clarendon
Press, Oxford.
Charbeneau, R.J., Weaver, J.W., Lien,
B.K., (1995). The hydrocarbon spill
screening model (HSSM), Volume 2:
Theoretical background and source
codes, Robert S. Kerr Environmental
Research Laboratory, United States
Environmental Protection Agency,
EPA/600/R-94/039 b.
Crank, J. (1975). The Mathematics of
Diffusion. Second Edition. Clarendon
Press, Oxford.
Daniel, D.E. (1993). Chapter 1.
Introduction, in: Daniel, D.E. (Ed.),
Geotechnical practice for waste
disposal, Chapman & Hall, London, pp.
3-14.
Di Molfetta, A., Sethi, R. (2012).
Ingegneria degli acquiferi, Springer,
Milano.
Giroud, J.P. (1997). Equations for
calculating the rate of liquid migration
through composite liners due to
geomembrane defects, Geosynthetics
International, 4(3-4), 335-348.
Giroud, J.P., Badu-Tweneboah, K.,
Soderman, K.L. (1997). Comparison of
leachate flow through compacted clay
liners and geosynthetic clay liners in
landfill liner systems, Geosynthetics
International, 4(3-4), 391-431.
Giroud, J.P., Bonaparte, R. (1989).
Leakage through liners constructed with
geomembrane liners--parts I and II and
technical note, Geotextiles and
Geomembranes, 8(1), 27-67, 8(2), 71-
111, 8(4), 337-340.
Guyonnet, D., Perrochet, P., Côme, B.,
Seguin, J.-J., Parriaux, A. (2001). On
the hydro-dispersive equivalence
between multi-layered mineral barriers,
Journal of Contaminant Hydrology, 51,
215-231.
Koerner, R.M. (1986). Design with
Geosynthetics, Prentice-Hall, Upper
Saddle River, New Jersey.
Manassero, M., Benson, C.H., Bouazza,
A. (2000). Solid waste containment
systems, in: Proceedings of the
International Conference GeoEng2000,
Melbourne, Australia, Technomic
34
Publishing Company, Lancaster, USA,
Volume I, pp. 520–642.
Manassero, M., Dominijanni, A., Musso,
G. (2009). Modeling applied to
environmental geotechnics, in:
Advances in Environmental
Geotechnics, Proceedings of the
International Symposium on
Geoenvironmental Engineering (ISGE
2009), Hangzhou, 8-10 September
2009, Zhejiang University Press -
Springer, Hangzhou, pp. 57-89.
MIGRATEv9 (1995). User’s Guide, GAEA,
Windsor, Ontario, Canada.
Olinic, E., Dominijanni, A., Manassero, M.,
(2002). Steady-state advective and
diffusive pollutant transport through
landfill barrier systems, Proceedings of
7th International Conference on
Geosynthetics, Nice, France, 22-27
September 2002, Dalmas, Gourc and
Girard, Eds., Swetz and Zeitlinger,
Lisse, 1577-1580.
Rowe, R.K. (1998). Geosynthetics and the
minimization of contaminant migration
through barrier systems beneath solid
waste. Proc. 6th Int. Conf. on
Geosynthetics, Atlanta 1, 27–103.
Rowe, R.K. (2005). Long-term
performance of contaminant barrier
systems (45th Rankine Lecture),
Géotechnique, 55(9), 631-678.
Rowe, R.K, Quigley, R.M., Brachman,
R.W.I., Booker, J.R. (2004). Barrier
Systems for waste disposal, 2nd
Edition, Spon Press, London.
Shackelford, C.D. (2014). The ISSMGE
Kerry Rowe Lecture: The role of
diffusion in environmental geotechnics,
Canadian Geotechnical Journal, NRC,
51(11), 1219-1242.
Touze-Foltz, N., and Barroso, M. (2006).
Empirical equations for Calculating the
Rate of Liquid Flow through GCL-
Geomembrane Composite Liners,
Geosynthetics International, 13(2), 73-
82.
USEPA (1989). Risk Assessment
Guidance for Superfund: volume 1;
 Human Health Evaluation Manual
(PART A), EPA/540/1-89/002.
USEPA (1991a). Risk Assessment
Guidance for Superfund: volume 1;
Human Health Evaluation Manual
(PART B, Development of Risk-Based
Preliminary Remediation Goals),
EPA/540/R-02/003.
USEPA (1991b). Design and construction
of RCRA/CERCLA final covers,
EPA/625/491/025.
USEPA (1996). Soil Screening Guidance:
User’s Guide.

35
36
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Difficult foundation conditions in Romania

Ernest OLINIC1*
1
Technical University of Civil Engineering Bucharest, Department of Geotechnical and
Foundation Engineering, Bucharest, ROMANIA

ABSTRACT
Significant areas of the Romanian territory are covered by soils that impose special foundations.
From the category of difficult soils the following have a particular behaviour in relation to water:
loessoid soils and swelling -shrinkage soils. Loessoid soils collapse irreversibly when are
saturated, leading to significant settlements, while swelling - shrinkage soils significantly change
their volume when moisture variations occur. In both cases, measures should be taken either to
avoid water infiltration or for desensitization of these soils in relation to water. The paper presents
significant aspects from the technical norms related to elaboration of geotechnical documentations
(NP 074-2014) and foundation of constructions on loessoid soils (NP 125-2010) and swelleing -
shrinkage soils (NP 126-2010). The paper also presents the results of several tests performed on:
expansive clay mixed with sand, gravel and slag (foundry waste), loess mixed with sand, bentonite
mixed with sand (bentonite enhanced sand) and loess mixed with bentonite enhanced sand.

Keywords: collapsible soils, loess, expansive soils, soil stabilization

1. INTRODUCTION 2. GEOTECHNICAL
Significant areas of the Romanian territory DOCUMENTATIONS FOR
are covered by soils that impose special CONSTRUCTIONS
foundation solutions. These solutions may In Romania, the geotechnical research of
consist in adopting a deep foundation the foundation soil performed in order to
system (pilots, barrettes), improving the elaborate geotechnical documentations
natural foundation soil, taking special for-constructions is done according to NP
measures regarding water seepage in the 074-2014 "Technical norm on
foundation soil, taking measures to ensure geotechnical documentations for
the stability of the site, etc ... constructions ".
For these sites, the adequate study of In order to establish the requirements of
the physical and mechanical properties of field investigations, laboratory testing and
the foundation soil represents a key issue geotechnical design, this regulation
in the geotechnical design. introduces 3 geotechnical categories for
which geotechnical risks are associated.

* presenting author
 Table 1. Geotechnical categorization ƒ exceptional dewatering system - 4
Geotechnical Points Geotechnical points
risk limits category 3. classification of buildings by
Low 6...9 1
importance:
ƒ low - 2 points
Moderate 10...14 2
ƒ normal - 3 points
Major 15...22 3
ƒ special, exceptional - 5 points
4. neighborhoods:
The geotechnical risk depends on two ƒ no risk - 1 point
categories of factors: on the one hand
ƒ moderate risk - 3 points
factors related to the soil, of which the
most important are the soil conditions and ƒ major risk - 4 points
groundwater, and on the other hand, 5. seismic zone:
factors related to the characteristics of the ƒ ag < 0.15g - 1 point
construction and its neighborhoods. ƒ ag = (0.15...0.25)g - 2 points
In order to define the geotechnical ƒ ag > 0.25g - 3 points
category / risk are evaluated the following
factors:
1. soil conditions:
ƒ good soils - 2 points
ƒ average soils - 3 points
ƒ difficult soils - 6 points

Table 2. Difficult soils according to the

Romanian legislation (NP 074-2014)
No. Type of soil Figure 1. Map of collapsible soils in Romania

1 Sandy soils, including silty sands, in

loose state
2 Saturated sandy soils, sensitive to
liquefaction under seismic loads
3 Fine soils with consistency index IC < 0.5
4 Loessoid soils in group B of sensitive to
wetting soils defined by NP 125
5 Swelling – shrinkage soils (active clays)
defined by NP 126 Figure 2. Map of swelling – shrinkage soils in
Romania
6 Soils with high content of organic matter
(>6%)
7 Slope sites with potential to landsliding
8 Uncontrolled soil fills, less than 10 year
old
9 Waste fills, no matter how old
2. groundwater:
ƒ no dewatering - 1 point
ƒ normal dewatering system - 2
Figure 3. Territory macrozonation concerning
points
landsilde hazard

38
 It can be observed that the existence in
a site of a difficult foundation soil (6 points)
cumulated with the other 4 factors, will
lead to a minimum score of 11 points,
which will frame the site in geotechnical
category 2 / moderate risk.
Geotechnical categories 2 and 3
impose in situ researches, laboratory tests
and design methods more elaborated than
those recommended for geotechnical
category 1.
In the figures above are presented
maps of the Romanian territory with
collapsible soils (Figure 1), swelling –
shrinkage soils (Figure 2) and potential of
landslide production (Figure 3).
3. COLLAPSIBLE SOILS
In Romania the loessoid (collapsible) soils
covers about 17% of the territory as it is
indicated in the map from Figure 1.
The soils sensitive to wetting are
defined as unsaturated macroporic
cohesive soils, which in contact with water
are subjected to sudden and irreversible
changes of internal structure, reflected by
additional settlements and decreases of
mechanical geotechnical parameters.
Additional settlement may occur under
the own weight of wetted layer (Img) and
under the action of compressive loads
transmitted by the foundations (Imp).
In terms of how the settlement occurs,
the loess is classified in two groups:
- Group A: loess having additional Img less
than 5 cm;
- Group B: loess having additional Img
equal to or greater than 5 cm.
The minimum specific geotechnical
data necessary to classify a soil as
sensitive to wetting are related to
composition and compressibility in natural
and saturated conditions.
In this connection were imposed
physical (I) and mechanical (II)
identification criteria as follows:
I.1. cohesive soil with silt 50 ÷ 80% in
unsaturated state (Sr <0.8) and the natural
porosity n> 40%.
39
eL  e
I.2. Index I with values between
1 e
0.10 ÷ 0.30 depending on the plasticity
index Ip = 10 ÷ 22%, where e is the void
ratio in natural state and eL is the void
ratio at the liquid limit of plasticity, wL of
the soil.
II.1. the index of additional settlement to
wetting under the load of 300 kPa (in
oedometric test im300 > 2%).
II.2. the indexes K and G related to soil
settlement in natural and flooded state (in
plate load test) have values:
s
K i t 5 úi G si  sn t 3 cm,
sn
where si is submerged soil settlement and
sn is the settlement at natural moisture
content as determined by plate load test
under the pressure of 300 kPa.
To characterize a soil as sensitive to
wetting should have at least one criteria
related to physical properties and one
criteria related to mechanical behavior.
It follows that for the soils sensitive to
wetting the oedometric tests should be
made after a specific methodology, being
recommended double tests - on samples
with natural moisture content and on
saturated samples - and/or samples at
natural moisture content, saturated under
the pressure of 300 kPa. (Figure 5).
Figure 4. Specific oedometric tests on loess.
Double curves method.
 3.1. Settlement calculation
In the case of loessoid foundation soil it is
compulsory to calculate the settlement in
the hypothesis of foundation soil wetting.
Additional settlement may occur under
the own weight of wetted layer (Img) and
under the action of compressive loads
transmitted by the foundations (Imp).
Settlement calculation is performed on
entire thickness of the layer sensitive to
Figure 5. Specific oedometric tests on loess. wetting by dividing it into elementary
Single test method layers.
Thus, are obtained the necessary For an elementary layer “I” it is
parameters for geotechnical design, evaluated the vertical stress under own
namely, the index of additional settlement
weight at the natural state (Vgn) and, from
to wetting imV, including im300, and the the stress – strain oedometric curve on
structural resistance V0 which represents natural soil sample, the specific settlement
the pressure at which imV is equal to 1%. (Hgn). The vertical stress under own weight
It is also necessary to determine the of saturated layer (Vgi) is evaluated and,
shear strength parameters on samples from the stress – strain oedometric curve
with natural moisture content and of the saturated sample, the specific
saturated. In Table 1 are presented settlement (Hgi) is determined. The
indicative characteristic values for loess difference between the two settlements
and loess-like soils in Romania for the represents the specific settlement of the
main geotechnical parameters. elementary layer "i" under its own
Table 3. Typical values for geotechnical saturated weight (img) (Figure 6).
parameters of loess and loessoid soils The vertical stress under the vertical
Characteristic load transmitted by the foundations (Vz)
Geotechnical parameter
value cumulated with the stress under own
weight in saturated conditions (Vgi)
Skeleton density, Us
2,52 - 2,67 represents the total stress that will act on
[g/cm3]
the elementary layer “I”. The difference
Unit weight of the soil, between the specific settlement of the
12,0 - 18,0
J>kN/m3] saturated sample (Hpi) under this effort and
Dry unit weight of the soil, the specific settlement under own
11,0 - 16,0 saturated weight (Hgi) represents the
Jd [kN/m3]
settlement of the saturated soil under the
Porosity, n [%] 40 - 55 vertical load transmitted by the
Plasticity index, IP [%] 5 – 22 foundations (imp) (Figure 7).
Index of additional If Img is the settlement under own weight
settlement to wetting of the saturated layer, Imp+Imp represents
2 – 14 the total settlement under own weight of
V= 300 kPa, im300 [%] saturated layer and the load transmitted
Oedomertic modulus, by the foundation. This calculation
5000 - 15000 imposed by the technical norm NP 125-
Eoed 200-300 [kPa]
2010 doesn’t indicate the settlement of the
Internal friction angle,M [°] 5 – 25 foundation in the hypothesis the
Cohesion, c [kPa] 10 - 30 foundation soil is not wetted (s), which is
presented in Figure 8.

40
 Figure 6. Additional settlement to wetting due to own weight of wetted layer (Img)

Figure 7. Additional settlement to wetting under external loads (Imp)

Figure 8. Settlement in natural conditions (without wetting) (s) and settlement after wetting of the
foundation soil (si)
41
3.2. Geotechnical design by calculation (V) bigger than the structural resistance
For the design of the foundation solutions (V0). In Figure 9 is indicated a middle
on collapsible soils, the following will be zone, named inert zone because the total
taken into account: vertical load is less than the structural
resistance, therefore no additional
x for the verification at normal exploitation settlements due to soil wetting will occur.
limit state, differential settlements of the
foundations will be limited in order to avoid
appearance of any limit state in the
structure;
x the compatibility of the deformations
reached in ultimate limit state will be taken
into account, by analyzing the relative
rigidity of the structure and soil;
x the choice of the geotechnical actions,
based on the destination and lifetime of
the construction, will be considered those
resulted from wetting (saturation) of the
soil taking into account:
• the source and the type of the wetting
(local, general);
• the direction of wetting, which can be
gravitational or generated by the rising of
the groundwater table; Figure 9. Characteristic zones in the
foundation soil composed of collapsible soils
• speed and direction of the
groundwater flow, that can have With respect to the foundation width (B)
alternately different directions (irrigational and the value of the net pressure on the
canal, shore); foundation raft (pnet), the thickness of the
layer sensitive to water (H) and the value
x in the case of pile foundations
embedded in a layer non-sensitive to of the structural resistance (V0) other
water, beneath a loess layer, if the wetting situations can occur (Figure 10 a…e).
is possible and the settlement under the
weight of the soil can occur, if will be
considered the negative skin friction on
the piles.
3.3. Calculation of the foundation soil
By knowing the structural resistance V0 it
can be defined the zones in the foundation
soil where deformations occur.
Therefore, the deformable upper zone
extends until the depth where the vertical
effort (V) of the foundation load (Vz) and
the soil weight (Vgz) becomes equal to V0
(Figure 9).
On the other hand, for some
thicknesses of the collapsible soil layer, Figure 10. Characteristic situations for the
additional settlements can occur also at foundation soil composed of a layer sensitive
the base of the layer, defined as to water
deformable lower zone, where Vgz is
bigger enough to result a total vertical load
42
 The calculations at the normal
exploitation limit state imply verification of
the settlements. It will be taken into
account: additional settlements to wetting
under the weight of the soil (Img) and under
the external loads (Imp) according to the
type of the loess (group A or B) and the
desensitization measurements.
The calculations at the ultimate limit
state refers to the evaluation of the
bearing capacity based on the shear
strength parameters (M and c) at natural
moisture content and saturated, according
to the measurements for the foundation
soil.
3.4. Measurements for the choice of the
foundation solutions
The following measures should be taken
into account for the choice of the
foundation solution on a foundation soil
consisting of loessoid soils:
x prevention of soil wetting;
x soil improvement by different
technologies following the formation of a
new internal structure for the entire layer
(desensitization to wetting). Can be
considered:
• intensive compaction;
• injection by silication
• thermic treatment;
• compacted columns of concrete or
local materials; it is forbidden to use only
granular permeable materials.
x construction of a compacted cushion
above the layer of collapsible soil; it is
forbidden to use only granular permeable
materials;
x replacement of the collapsible soil layer
by excavation and controlled soil fill with
adequate materials;
x consuming of the additional settlements
by wetting through:
• controlled wetting;
• saturation under supplementary load;
• deep explosions.
x selection of indirect foundation system
(piles, barrets, etc…) embedded in a non-
sensitive to water layer.
43
4. LOESSOID SOILS MIXED WITH
SAND AND BENTONITE
In the experimental programme, various
mixtures of loessoid material with different
natural mineral materials have been
proposed, in view of eliminating moisture
sensitiveness, improving geotechnical
parameters of mechanical behaviour and
limiting permeability.
To this purpose, a series of mixtures
have been proposed: loess with sand 1-2
mm (Cu = 1.5) and loess with sand and
bentonite powder addition in two variants
of mixture. The obtained mixtures are
presented below:
x Mixture 1: 80% loess + 20% sand (1-2
mm);
x Mixture 2: 60% loess + 40% sand (1-2
mm);
x Mixture 3: 50% loess + 40% sand (1-2
mm) + 10% bentonite;
x Mixture 4: 50% loess + mixture from
(40% sand (1-2 mm) + 10% bentonite).
The difference between the last two
mixtures consisted in the way they were
mixed. In the first case, all the three
materials were simultaneously mixed and
then water was added to reach different
degrees of humidity in order to perform the
normal Proctor test. In case of the last
mixture, the sand was first mixed with the
bentonite and with water and then, after
this mixture had dried, it was also mixed
with the loess (Burlacu et al. 2013).
As a result of the Proctor test outcome
analysis (Figure 6), it has been observed
that along with adding up and increasing
the percentage of sand in the mixture
(from 20% to 40%), the maximum density
in dry condition increases. At the same
time, the optimal compaction moisture of
the mixtures decreases.
In case of mixture 4, the Proctor curve
doesn’t have a peak but a constant zone
for the maximum dry density, which was
obtained for moisture content values
ranging between 11 and 15%. In order to
validate the results, tests on this sample
were carried out again and similar values
were obtained (Figure 7). The moisture
content plays a key role in the real scale of
compaction process. Given that, the last
indication regarding mixture 4 is important
because it allows compaction at wider
domain of moisture content.
Figure 11. The results of the Proctor trial for
all the mixtures obtained
include smectite, vermiculite, illite and
chlorite. Generally, the larger the amount
of these minerals is present in the soil, the
greater the expansive potential. These
expansive effects may become “diluted”
by the presence of other non-swelling
mineral such as quartz and carbonate
(Ivasuc et al., 2013).
The activity in relation to water of the
expansive soils can be estimated based
on physical and mechanical properties
which are determined in the laboratory
according to the legislation in force. The
physical properties that characterize the
activity of expansive soils are: content of
colloidal clay, plasticity index, activity
index and free swell. The mechanical
property which indicates more accurately
the activity of expansive soils is the
swelling pressure, determined in
consolidation tests on saturated samples.

Figure 12. Results of Proctor test for mixtures

3 and 4
As to the values of the permeability
coefficient, these have been of the order
of 10-5 cm/s for the average loess sample
rising up to values of 10-4 cm/s in case of
the mixture containing 40% sand, while in
case of the mixtures containing an
addition of bentonite, the measured values Figure 13. Determination of the swelling
were below 10-9 cm/s. pressure. Simple method

5. EXPANSIVE CLAYS
Clayey soils have the property to
significantly modify their volume when
moisture changes: they shrink when
moisture reduces and they swell when
moisture increases. Due to their shrink-
swell behaviour, these soils could create
many problems for engineering structures
and for this reason, direct foundation is
not allowed by the legislation in force,
being mandatory to be replaced or Figure 14. Determination of the swelling
improved by stabilization. pressure. Double method
Shrink-swell behavior is caused by the In Romania, the foundations laid on
mineralogical composition of clay soils with high shrink-swell potential must
minerals. These minerals determine the comply with the requirements of NP
natural expansiveness of the soil, and 126:2012. In order to classify a soil in the

44
category of soils with high shrink-swell
potential, the following geotechnical
parameters are mandatory: A2ȝ - the
percentage clay content with a diameter
less than 0.002 mm (%); IP - plastic index
(%); IA - activity index; CP - plasticity
criteria (%); UL - free swelling (%) and SP
- swelling pressure (kPa) (NP126-2010).
6. MECHANICAL BEHAVIOR OF
DESTRUCTURATED EXPANSIVE CLAY
6.1. Natural soil properties
The studied site is located in the
Transylvanian plateau; it is approximately
19.5 ha and the level difference ranged
from 265 m to 315 m nMN. In the initial
state, the site was not affected by
landslides (Olinic et al., 2014).
According to STAS 1913/5-85, the
analyzed soils were clayey soils with the
grain size distribution composed of 50-70
% clay (A2ȝ = 40 – 48 %), 25-40 % silt and
1-10% sand. Determining the plastic and
liquid limits, we observed that the plasticity
index (IP) showed values between 26.7 –
57.2 %.
Oedometric compression tests were
performed on samples that were initially
saturated to determine the swelling
pressure (SP) with recorded values
between 40…200 kPa.
The shearing resistance parameters
were determined in CU and CD conditions
performed on samples with natural
humidity and initially saturated samples.
Table 4 shows the variation of the shear
strength parameters.
Table 4. Shear strength parameters
Parameter/ Internal Cohesion,
Shearing test friction angle, [kPa]
type ĭ [o]
CU 11 ÷ 29 59 ÷ 160
CUsat 19 ÷ 28 23 ÷ 81
CD 19 ÷ 30 33 ÷ 80
CDsat 17 ÷ 23 25 ÷ 55
According to NP 126-2010, the studied
soils are classified in the category of very
active clays which are considered to be
regarded as difficult foundation soils in
compliance with NP 074-2014. According
to STAS 2914-84, all materials intercepted
in the investigated depth were included in
category of ‘bad’ quality soils. In this case,
they cannot be used (in their natural state)
as filling materials for the body of
perimetral embankments for the municipal
solid waste landfill or for other fillings.
If these materials are used as filling
materials, they must be desensitized in
relation to water, to undertake excavations
and reshape the slopes to 1:3 and to
construct berms with a width of 4...5 m for
every 10...12 m on vertical (Ivasuc, 2013).
6.2. Clay destructuration
Clayey soils are very sensitive to
environmental conditions, especially to
variation in humidity and temperature.
Clay destructuration occurs after a series
of cycles of hydration-dehydration and
freezing-thawing; this modification
produces irreversible effects on the
structure and texture of the expansive
soils causing important damages such as
cracks, differential settlements and loss of
stability (one of the major problems in
earthwork aplications).
All these damages are highly influenced
by the water content and degree of
saturation: during dehydration the
frequency of macropores increases, which
during hydration the macropores do not
close-up perfectly and hence cause the
soil to bulk-out slightly, and also allow
enhanced access to water for the swelling
process.
Engineering practice has shown that
water may enter into every soil structure:
from precipitation (rainfall and snow) or
c
from the ground. In order to avoid the
exposure to humidity and temperature
variations (by preventing the access of
water to the embankments) it is
recommended to cover the slopes with a
layer of top soil and to vegetize it with
grass and shrubs (Olinic et al., 2014).
Instability phenomena have appeared on
site during the winter of 2013, after the
excavation works which have been
performed in autumn and the slopes
45
remained unprotected (Figure 15), (Olinic sample destructuration can be seen in
et al., 2014). Figure 16.
To verify the effect of destructuration of
a clay sample and to explain the loss of
the stability appeared on the site (Figure
15), the shear strength parameters ࢥ
(internal friction angle) and c (cohesion)
have been determinated by the direct
shearing tests which simulated the
landsliding phenomena.
November 2012

08.11.2012 27.12.2012
February 2013

November 2012
February 2013
Figure 15. Loss of stability on the site
6.3. Laboratory tests on destructurated
clay
In order to explain the instability
phenomena occurring on the site, a
representative sample of clay (glomerular
clay) was chosen for submission to
freezing-thawing cycles (at natural
humidity and temperature variations) from
November 2012 until February 2013
(Olinic et al., 2014). The phases of clay
46
25.01.2013 02.02.2012
Figure 16. Clay destructuration at different
periods of time
The landsliding phenomena correspond
to an unconsolidated–undrained (UU)
direct shear strength test performed on a
saturated sample. For this reason, the test
consisted in the application of an 18 kPa
contact load, the saturation of the sample
for approximately 2h and application of a
pressure of 67, 121, 175, 230 and 285
kPa. Direct shearing was performed with
the speed of 1 mm/min, the test
stimulating the real conditions from the
site (Olinic et al., 2014).The shearing
strength parameters resulting from the
destructured clay analysis are: internal
friction angle (ĭ = 17.48°) and cohesion (c
= 6.94 kPa).
The values of the shear strenght
parameters, especially the cohesion,
confirm the fact that after destructuration
the internal soil structure is affected by the
weakening of cohesion between clay Usually, natural clay has swelling
particles. pressures that differ significantly from their
stabilized samples. In this case, for some
7. EXPANSIVE CLAY STABILIZATION addition percentages of non-cohesive
BY MIXING WITH GRANULAR materials, the effect of compaction is
MATERIAL higher that the desensitization effect,
In order to realize the perimetral resulting materials considered even
embankments from the municipal solid ‘worse’ than the natural sample - materials
waste landfill, placed on a slope area and with swelling pressure more higher than
built from expansive soils, it was swelling pressure of the natural sample
attempted to stabilize the existing clayey (Figure 17).
material from the site by adding different
granular materials.
It was proposed an experimental
program consisting of the determination of
the following geotechnical properties:
optimal compaction parameters, swelling
pressure, compressibility and
consolidation parameters and shearing
resistance parameters in consolidated-
undrained conditions on saturated
samples (CUsat) around the optimal
compaction parameters (Ivasuc, 2013).
Figure 17. Swelling pressure variation
In order to stabilize the expansive clay according to the percentage of filler materials
by mixing with a non-cohesive material, it
was proposed three types of granular To develop some reduced swelling
materials: S - slag (foundry sand), SG – pressures, in the case of desensitization
sand with gravel and G - gravel with with granular material is recommended
particles of 4-8 mm in diameter. that the humidity of the material should
have a moisture content with 1...3 %
Based on the compaction tests (normal higher than the optimal humidity of
Proctor test) performed for the natural clay compaction (w = woc + 1...3 %) (Ivasuc,
samples and for the mixtures with granular 2013, Olinic et al., 2014).
materials, the optimal parameters of
compaction showed a decrease in the Laboratory tests revealed that the
optimal moisture content and an increase following mixtures were optimal mixtures
in the dry density with the increasing the in relation to the natural sample: 40% slag,
percentage of granular material. 50% sand with gravel and 30% gravel
(Table 5).
Table 5. Soil characteristics of optimal mixtures with granular materials
Material / Optimal Permeability Shearing
Compressibility
characteristics of parameters of coefficient characteristic
characteristics
compacted compaction s
samples
ȡdmax wopt k Eoed200-300 pu ĭ c
[g/cm3] [%] [cm/s] [kPa] [kPa] [q] [kPa]
Natural clay (C) 1.62 21.0 8.22*10-9 8403 200 24 43
60% C + 40% S 1.77 14.5 7.63*10-9 11905 110 27 35
-8
50% C + 50% SG 1.93 11.2 1.07*10 11111 110 33 17
-8
70% C + 30% G 1.87 11.9 4.89*10 7380 75 *25 *35

47
7.1. Destructuration of stabilized clay
samples subjected to temperature and
humidity variations
It is mandatory for designers to take into
account the hydration-dehydration and
freeze-thaw behavior of soils to select
proper materials for constructing
embankments exposed to temperature
and humidity variations.
In order to explain the solution chosen
for the optimal mixtures with granular
materials were chosen representative
samples of clay and mixtures between
clay and non-cohesive material (sample 1
– Natural clay, sample 2 – 60% C+40% S,
sample 3 – 50% C + 50% SG, sample 4 –
70% C + 30% G) which was submitted to
hydration-dehydration and to freezing- Figure 19. Clay samples flooded at the bottom
thawing cycles from October 2013 until
April 2014 (Olinic et al., 2014).
Samples were subjected to a
dehydration cycle by drying in normal
conditions of temperature and humidity.
The phases of drying may be seen in
Figure 18.

Figure 20. Clay samples dehydration after

Figure 18. Clay samples dehydration
In order to observe the effect of
swelling when the expansive clay is in
contact to water the samples were flooded
at the bottom (hydration cycle - Figure 19)
and subjected to drying (Figure 20).
48
flooding
From October 2013 to April 2014 the
samples were exposed to natural weather
conditions, being subjected to several
freezing - thawing cycles. The samples
destructuration effect can be seen in
Figure 21.
The simulations reveal the influence
that destructuration has on the behavior of
an unprotected embankment constructed
from expansive clays or stabilized clays
after a number of hydration-dehydration
and freezing-thawing cycles. It can be
noticed that minimum differences of
destructuration is observed for the
samples from the middle, samples which
are represented by the mixture of 40%
slag and 50% sand with gravel.
a. October 2013
b. January 2014
c. January 2014
d. February 2014
e. February 2014
f. April 2014
Figure 21. Clay destructuration at different
periods of time
Even the sample composed of 30%
gravel (sample 4) which developed the
lowest swelling pressure value (75 kPa)
was destructured, after the several freeze
- thaw cycles, as much as the natural
sample.
49
8. CONCLUSIONS
In completion to the European technical
norms, Romania elaborated several
technical norms and guides for the local
specific site conditions.
All the Romanian technical norms are
accordingly to the Eurocode. All the
European Norms are fully applied in
Romania.
Significant areas of the Romanian
territory are covered by soils classified as
“difficult foundation soils”.
The paper is focused on the
stabilization of loessoid and expansive
soils by mixing them with different
percentages of granular materials in order
to improve their engineering properties to
use them as soil fillings or foundation
soils. On the basis of the laboratory tests
and field researches, the following
conclusions can be drawn:
• Each soil acts differently depending on
its mineralogical and granulometric
composition: for this reason there is no
‘recipe’ for the improvement of difficult
soils;
• Adding granular materials, for some
addition percentages, the effect of
compaction is higher that the
desensitization effect, resulting in
materials considered even ‘worse’ than
the natural sample;
• The mixture of loess and granular
material has better mechanical
characteristics and reduced permeability
compared to the one the loess has in its
natural state. From all the solutions
proposed (compacted loess, mixture of
loess and sand and mixture of loess, sand
and bentonite) the one with sand and
bentonite, mixed with loess after drying,
seems to be the optimal one due to the
wide domain in which optimal compaction
parameters are reached;
• Concerning mechanical characteristics,
no significant differences seem to exist
between the analysed mixtures, but one
can notice that water sensitivity is
significantly reduced and that, compared
to the flooded loess, the values obtained 2014. ISBN 978-619-7105-08-7. ISSN
are significantly better; 1314-2704.
• In the case of expansive soil mixtures Olinic, E., Manea, S. & Ivasuc, T. (2014)
with granular materials, the swelling “Design of a municipal solid waste
pressure is the only property that changes landfill in difficult geotechnical
and also categorizes the activity of a soil conditions: slope area and expansive
in contact with water; clays. Case study from Romania”,
• In order to develop some reduced Proceedings of 15th Danube-European
swelling pressures, in the case of Conference on Geotechnical
desensitization with granular materials, it Engineering (Eds: Brandl, H. & Adam,
is recommended to assure some D.), Novographic Druck GmbH, Vienna.
compaction degrees of 95-98% to a Olinic, E., Ivasuc, T., Manea, S., (2015)
compaction moisture content 1...3% "Improvement of difficult soils by mixing
higher than the optimum moisture content. with mineral materials and inorganic
waste. Experimental projects and case
studies from Romania", Proceedings of
REFERENCES the XVIth European Conference on Soil
Mechanics and Geotechnical
Burlacu, C., Olinic, E., Manea, S. & UĠă, P.
Engineering, 2437 - 2442, Edinburgh,
(2013) “Compacted soil columns for
Scotland, ICE Publishing, ISBN 978-0-
foundations on collapsible soils.
7277-6067-8.
Laboratory and in-situ experimental
study”, Proceedings of the 18th NP 074-2014 “Technical norm on
International Conference on Soil geotechnical documentations for
Mechanics and Geotechnical constructions”
Engineering (Eds: Delage, P., Desrues, NP 125-2010 „Technical norm on
J., Frank, R., Puech, A. & Schlosser, foundation of constructions on loessoid
F.), 2433 – 2437, Paris, France. soils”
Ivasuc, T., Olinic, E., Manea, S., Soare, B. NP 126-2010 „Technical norm on
(2013) “Studies on the stabilization of foundation of constructions on swelling-
expansive soils treated with granular shrinkage soils”
materials”, 13th International STAS 2914-84 „Road works. Earthworks.
Multidisciplinary Scientific General technical requirements for
Geoconference SGEM 2013, Science quality”
and Technologies in Geology,
Exploration and Mining – Conference
Proceedings, Vol. II, 403-411. Albena,
2013. ISBN 978-945-91818-8-3. ISSN
1314-2704.
Ivasuc T. (2013) PhD Thesis "Foundation
solutions on difficult soils for local
material constructions", Technical
University of Civil Engineering
Bucharest, Romania.
Olinic, E., Ivasuc, T. & Manea, S. (2014)
“Mechanical behaviour of
destructurated expansive clay”, 14th
International Multidisciplinary Scientific
Geoconference SGEM 2014, Science
and Technologies in Geology,
Exploration and Mining – Conference
Proceedings, Vol. II, 581 - 588. Albena,
50
25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Session 1

In-situ and laboratory tests

52
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Permanent strain behavior of cyclically loaded cohesive soil in

undrained conditions

Andrzej GàUCHOWSKI1*
1
Warsaw University of Life Sciences – SGGW, Department of Geotechnical Engineering,
Warsaw, POLAND

ABSTRACT
The road structures, industrial and wind turbine foundations are cyclically loaded in a long period of
time and with a numerous repetitions. The knowledge of the soil subbase response to repeated
loads is important for designers when serviceability limit state is calculated. The calculated
modulus from static triaxial tests provides nonsufficient information for proper calculations of
settlements. The long term cyclic loading should be conducted in purpose not only to obtain such
parameters as resilient modulus but also for characterizing the phenomena of permanent strain
accumulation. The cyclic hardening and softening phenomena reported by many researches
highlights the behavior of soils under repeating loading. Nevertheless, such occurrence as the
stiffness degradation in cyclically loaded after numerous of load repetitions should be studied more
closely. The previous studies concerning unbound aggregates, lead to recognize the long term soil
response to cyclic loading as an element of shakedown theory. The shakedown concept should be
also studied and possibility to adopt of part the shakedown concept to cohesive soil behavior
description may be considered. In this paper results of cyclic triaxial test are presented. The
cohesive soil was characterized by series of tests consists of physical and static examinations in
first stage. The long term repeating loading was performed for isotopically consolidated samples of
sandy clay loaded in one way test method in undrained conditions. The results have led to
characterize permanent strain development and fatigue behavior which occur as stiffness
degradation. The cyclic triaxial test results was later analyzed and an empirical formula considering
permanent strain development, which considers the fatigue phenomena, is proposed. The article
ends with conclusion about undrained cyclically loaded cohesive soil and some remarks
concerning future work.

Keywords: cyclic loading, degradation index, Young modulus

rutting, uneven road surface, more severe

1. INTRODUCTION
The cohesive soils in environment
subjected to human activities are exposed
to cyclic loads generally induced by , e.g.,
machine vibrations or traffic. The reaction
of subgrade soils is usually observed as
* presenting author
settlement. (Guo et al. 2013) This
phenomena is caused mostly by soft
subsoils deformations. In case of road
engineering, the unbound subbase layers
and asphalt layers, can be considered as
buffer zone which reduces the cyclic
stress on subgrade soils. (Tang et al.
2015) When buffer zone thickness is low,
the cyclic stress. The dynamic and cyclic
stresses for design purposes are
converted to equivalent static load. The
cyclic loads differs from static loads. The
studies under dynamic and cyclic loads
influence depths have shown that a fully
loaded truck (44.5kPa) causes additional
vertical stresses at 1.5m equal to 18kPa.
Also, the cyclic stress accumulation was
observed during this tests which plateaus
around 50th cycle. The phenomena of
stress accumulation comes from micro-
scale perspective. The small rotations
within the soil structure, causes residual
stresses but not leads to collapse. The
residual stresses accumulate in soil
skeleton and later are a part of further
cycles (Tang et al. 2015, Cai et al. 2013).
The compaction process which occurs
during cyclic loading as volumetric strain
and frictional sliding of particles leads to
increase of soil stiffness and bearing
capacity. During this process some of the
external loads energy is dissipated. The
amount of the energy which is needed to
conduct this process decreases with
number of repetitions and stabilises to a
constant value in further cycles (Suiker et
al. 2005, Abdelkrim et al. 2003).
Previous studies have shown that
cohesive soils subjected to cyclic loading
in undrained conditions tends to generate
excess pore water pressure which causes
the decrease of the stiffness and strength
of clay (Guo et al 2013, Li et al. 2011).
The total strain caused by cyclic loading
can be divided to resilient strain and
permanent strain. In undrained cyclic
loading conditions od cohesive soil where
the degradation of mechanical parameters
in observed, the plastic and resilient
strains development indicates the size and
rate of changes (Wang et al. 2013, Sun et
al. 2015).
The stress-strain relationship during the
cyclic loading tests is nonlinear. Proposed
by Seed (1955) resilient modulus Mr
characterises the unloading process. The
Mr is defined as a quotient of cyclic stress
54
and resilient strain. The Figure 1 presents
stress-strain characteristics of the cyclic
loading.
Figure 1. The schema of stress-strain
characteristics in one cycle of repeated
loading.
The permanent strain which occurs
during soil mechanical properties
degradation accumulates with decreasing
rate as the cyclic loading is proceeded.
Irrecoverable strains in in situ conditions
are deformations. The deformation with
number of cycles was studied by
Goldscheider and Gudehus (1976). The
deformations were divided to three
categories, stepwise failure, shakedown
and ablation. In first case, stepwise failure
is characterised by deformation
accumulation in every cycle. The
deformation characteristic is linear with
number of cycles. The shakedown states
for permanent deformations which occurs
in first few cycles. After this stage,
irrecoverable deformations vanishes
completely and only resilient strains can
be observed during repeated loading. The
abation is characterised by decrease of
plastic deformation rate. But in opposite to
the shakedown phenomena, the abation
process never ends and some plastic
deformations can be seen after numerous
of repetitions. Figure 2 presents schema
of the three deformation characteristics
versus the number of cycles (Wichtmann
et al. 2005).

Figure 2. The schema of displacement u
/strain İ versus number of cycles N
characteristics, a) stepwise failure, b)
shakedown and c) abation.
The degradation of soil parameters can
be defined by degradation index į. The į
value is defined as the ratio of dynamic
secant modulus in the Nth cycle to
dynamic secant modulus in the first cycle
(Idriss 1978), presents Eq. (1):
ı d ,N
E d ,N İ d ,N
į(N )
E d ,1 ı d ,1
İ d ,1
where, ıd is maximal axial stress in cycle
and İd is axial strain at maximal stress.
The degradation properties of soft soils
under repeated loading depends on axial
strain amplitude. The proposed
exponential model of degradation of soil
properties was later examined in terms of
consolidation ratio (Vucetic 1988, 1990).
The results show that the degradation
model is also applicable to over
consolidated soils and the rate of cyclic
modulus degeneration decreases with the
OCR value increase. The degradation
index į(N) can be later recalculated to
degradation parameter t which is defined
as negative quotient of common logarithm
from degradation index į(N) to common
logarithm of number of cycles, Eq. (2)
(Jiong and Cai 2012):
log į(N )
t  (2)
log N
The degradation phenomena in
cohesive soil depends on the initial
conditions. The void ratio, pore pressure
and the effective stress p’ impacts on
clayey soil response to cyclic loading.
Resilient modulus can be obtained from
uniaxial or triaxial tests. The Equation (3)
is employed to calculate Mr value:
ıd
Mr (3)
İr
where ıd is deviator stress calculated as
(4):
P
ıd (4)
A
where, P is force and A is the area of
sample. Elastic strain İr is calculated as
(5):
'H r
İr (5)
Hi
where ǻHr is elastic displacement and Hi
is height of sample.
2. MATERIAL AND METHODS
Material in this study was classified on the
basis of sieve analysis as sandy clay
(1) (saCl) in accordance with EUROCODE 7 .
Figure 3 presents the results of above-
mentioned test. The material was taken
from road construction site. Figure 4
presents results of Proctor's test using
55
standard energy of compaction, which
equals 0.59J/cm3 PN–EN 13242:2004.
Optimal moisture content amounts to
10.5% and dry density is equal 2.16g/cm3.
Liquid limit obtained from Casagrande test
was equal 19%

Figure 3. Gradation curve of tested soil.

Figure 4. Results of optimal moisture on

the basis of Proctor's test.

The advanced cyclic triaxial device

used in this study (GDS) can vary the
deviatoric stress simultaneously within a
range of frequencies between 0.00001
and 2Hz. All samples were remolded and
then first saturated to guarantee B values
greater than 0.95 for all tests. The
specimens were isotopically consolidated
to the initial effective mean principal stress
of 275kPa. After that, the specimens were
sheared cyclically under various stress
paths in undrained conditions.

3. RESULTS
Results of cyclic triaxial tests presents
Figures 5 and 6. Figure 5 presents axial
strain development during the test.
56
Figure 5. Results of cyclic loading triaxial
tests, axial strain characteristics versus
number of cycles.
 The tests was performed for soil
sample where four stages of cyclic loading
were conducted. Each of the stages was
characterized by maximal stress ıd and
stress amplitude ıa. The detailed
information about the test conditions
presents Table 1. The cyclic loading
frequency was equal to 1Hz. For each
stage 1000 loading cycles was performed.

Table 1. The test conditions for four

stages of cyclic loading.
stage 1 stage 2 stage 3 stage 4
ıd 30.25 68.35 162.35 319.25
ıa 5.5 26.9 52.1 205.9

The axial strain development was

depended on deviator stress value. The
higher applied stress caused higher plastic
strain during cyclic loading. The strain in
all four stages were characterized by
abation behavior. The soil mechanical
properties degeneration can be observed
on Figure 6. The comparison of axial
strain in logarithm scale of cycle number
presents Figure 7.

Figure 6. Plot of the axial stress–strain Figure 6. Plot of the axial strain vs. cycle
comparison of hysteretic loops among number, comparison of four stages -
different cycle numbers. abation phenomena.
57
 The plots of selected cycles compares Table 2. The Young modulus values (MPa) for
the stress-strain relationship during the tested sandy clay.
tests. The stage 1, where low stress levels no. of
stage 1 stage 2 stage 3 stage 4
(maximal stress and stress amplitude) cycle
were applied results in low axial strain. 1 433 147 65.1 0.44
Nevertheless the cyclic loading leads to 10 405 132 93.2 0.61
soil mechanical properties degeneration 100 392 128 82.8 0.91
what can be observed as Young’s 1000 374 127 83.0 1.01
modulus decreasing. The same behavior
was observed during the stage 2.
When stress level increases the change The degradation index į was calculated
of respond can be observed. The plastic for cycles 1, 10, 100 and 1000. The results
strains can be in greater amount in first presents Figure 8. The į value for stages
cycle. Therefore, the Young’s modulus is 1 and 2 follows pattern of degradation
smaller at first cycle than after 10 behaviour. The stage 3 after a few cycles
repetitions. After this stage, the of hardening process, starts to degrade
degradation of mechanical parameters when 100 cycles were applied. The stage
can be observed as well (stage 3). 4 is characterised by degradation index
increase. In this stage the material loosed
The deviator stress value near the
its mechanical properties.
static failure line causes great plastic
strain and no stabilization of this
phenomena can be observed. The Young
modulus increases (stage 4).
Figure 7 presents the view of
abovementioned test results in plot of
change the Young’s modulus versus
number of cycles.

Figure 8. Plot of degradation index į

change during the tests versus number of
cycles.

The detailed view on degradation index

change presents Table 3.

Table 3. The degradation index į (-) for tested

Figure 7. Plot of Young’s modulus change sandy clay.
during the tests versus number of cycles. no. of
stage 1 stage 2 stage 3 stage 4
cycle
The Young modulus values were varied 1 1 1 1 1
from 433MPa to 0.44MPa. The detailed 10 0.93 0.89 1.43 1.38
data presents Table 2. 100 0.90 0.87 1.27 2.08
1000 0.86 0.86 1.27 2.31

58
 The change of Young modulus and
degradation index follows the same
pattern. Figure 9 presents 3D view of
degradation index value under number of
cycles and Young modulus value. The
graph surface analysis led to an equation
being created, which contained these
variables. Equation (6) presents formula
for calculating the degradation index į (-)
(z), basing on varying Young modulus
(MPa) (y) and number of cycles (-) (x):
z a  b ˜ ln( x )  c ˜ ln( y )  d ˜ (ln( x )) 2 
(6)
 e ˜ (ln( y )) 2  f ˜ ln( x ) ˜ ln( y )
where, letters from a to f are constants:
a=-0.987587689; b=0.231129066;
c=0.027590142; d=-0.00508219; e=-
0.00447899; f=-0.03914599. For this
equation the R2 value was equal to 0.899.
Figure 8. 3D view of degradation index
change with number of cycles and Young
modulus value.
From Figure 8 analysis the conclusion can
be drown. The degradation index į
presents appropriate characteristics when
material has high Young modulus at first
loading cycle. In other words the stiffer
material is the more degradation of
mechanical properties will be observed. In
opposite, when sandy clay characterises
low mechanical parameters the
degradation phenomena would be not
observed or will be observed in further
cycles of loading. This occurrence is
similar to the degradation mechanism
59
described in literature. The degradation of
mechanical properties is bounded with
plastic strain development. The abation
phenomena observed in all four stages
differs the magnitude of plastic strain
accumulation.
4. CONCLUSIONS
The research conducted in this paper and
analysis of the degradation of mechanical
properties phenomena of sandy clay
during cyclic loading lead to the following
conclusions:
x Cohesive soils can behave various
response to cyclic loading. The plastic
strain accumulation may be followed one
of three possible patterns. In this study the
abation of plastic strain was observed.
x The Young modulus values for sandy
clay varied from 433MPa to 0.44MPa.
Such wide range of this mechanical
properties was caused by applied axial
stress.
x The degradation index į after 1000
cycles was equal to 0.86 for stages 1 and
2. The į value for stages 3 and 4 was
equal to 1.27 and 2.31 respectively.
x The degradation index change versus
Young modulus and number of cycles
Equation was estimated (6).
x The degradation of soil properties in
this studies lead to conclusion that, the
stiffer soil is the grater degradation of its
initial mechanical properties will be
observed. When the initial conditions
allows for soil mechanical properties
improvement, the degradation will be
observes in further cycles.
ACKNOWLEDGEMENTS
The author of the paper would like to
thank the Polish Society for Soil
Mechanics and Geotechnical Engineering
for nomination for delegation to 25th
EYGEC. The author would like also to
thank the rector of Warsaw University of
Life Sciences –SGGW for financing of the
author participation in this event.
REFERENCES
Abdelkrim, M., Bonnet, G., & De Buhan, P.
(2003). A computational procedure for
predicting the long term residual
settlement of a platform induced by
repeated traffic loading. Computers and
Geotechnics, 30(6), 463-476.
Cai, Y., Gu, C., Wang, J., Juang, C. H.,
Xu, C., & Hu, X. (2012). One-way cyclic
triaxial behavior of saturated clay:
comparison between constant and
variable confining pressure. Journal of
Geotechnical and Geoenvironmental
Engineering, 139(5), 797-809.
Goldscheider M. and Gudehus G. (1976)
Einige bodenmechanische Probleme
bei Kusten- und Offshore-Bauwerken.
In Vortr¨ age zur Baugrundtagung.
DGEG, 1976
Guo, L., Wang, J., Cai, Y., Liu, H., Gao,
Y., & Sun, H. (2013). Undrained
deformation behavior of saturated soft
clay under long-term cyclic loading. Soil
Dynamics and Earthquake
Engineering, 50, 28-37.
Idriss I M, Dobry R, Singh R D. (1978)
Nonlinear behavior of soft clays during
cyclic loading. Journal of Geotechnical
Engineering Division, ASCE, Vol.104,
No.GT12, pp 1427-1447. 5.
Jiang, M., & Cai, Z. (2012). Stiffness
Degradation of Soft Marine Clay under
Uniaxial Cyclic Loading. Electronic
Journal of Geotechnical
Engineering, 17.
Li, L. L., Dan, H. B., & Wang, L. Z. (2011).
Undrained behavior of natural marine
clay under cyclic loading. Ocean
Engineering, 38(16), 1792-1805.
Polish Committee for Standardization.
Geotechnical Investigations – Soil
Classification – Part 2: Classification
Rules; PN–EN ISO 14688-2:2004;
Polish Committee for Standardization:
Warsaw, Poland, 2004.
Polish Committee for Standardization.
Motor roads. Earthworks. Requirements
and tests; PN–EN 13242:2004; Polish
Committee for Standardization:
Warsaw, Poland, 1988 (In polish)
60
Seed, H. B., Chan, C. K., & Monismith, C.
L. (1955). Effects of repeated loading
on the strength and deformation of
compacted clay. In: Highway Research
Board Proceedings (Vol. 34).
Suiker, A. S., Selig, E. T., & Frenkel, R.
(2005). Static and cyclic triaxial testing
of ballast and subballast. Journal of
Geotechnical and Geoenvironmental
Engineering, 131(6), 771-782.
Sun, L., Gu, C., & Wang, P. (2015).
Effects of cyclic confining pressure on
the deformation characteristics of
natural soft clay. Soil Dynamics and
Earthquake Engineering, 78, 99-109.
Tang, L., Chen, H., Sang, H., Zhang, S., &
Zhang, J. (2015). Determination of
traffic-load-influenced depths in clayey
subsoil based on the shakedown
concept. Soil Dynamics and
Earthquake Engineering, 77, 182-191.
Vucetic M, Dobry R. (1988) Degradation of
marine clays under cyclic loading.
Journal of Geotechnical Engineering,
ASCE, Vol.114, No.2, pp 133-149. 6.
Vucetic M. (1990) Normalized behavior of
clay under irregular cyclic loading.
Canadian Geotechnical Journal, Vol.27,
No.1, pp 29-46.
Wang, J., Guo, L., Cai, Y., Xu, C., & Gu,
C. (2013). Strain and pore pressure
development on soft marine clay in
triaxial tests with a large number of
cycles. Ocean Engineering, 74, 125-
132.
Wichtmann, T., Niemunis, A., &
Triantafyllidis, T. (2005). Strain
accumulation in sand due to cyclic
loading: drained triaxial tests. Soil
Dynamics and Earthquake
Engineering, 25(12), 967-979.
 25th European Young Geotechnical Engineers Conference
21st – 24th of June 2016, Sibiu, Romania

Geotechnical Characterization of Very Soft Deep-Sea

Sediments by In-Situ Penetrometer Testing

Bert LIETAERT1*, Francois CHARLET2, Peter STAELENS3

1
Dredging, Environmental and Marine Engineering (DEME), Geotechnical Engineer at the
Research, Methods and Engineering Department (RMPE), Zwijndrecht, BELGIUM.
2
Dredging, Environmental and Marine Engineering (DEME), marine geologist and
exploration manager at Global Sea Mineral resources NV (GSR), Zwijndrecht, BELGIUM.
3
DotOcean NV, Technical director, Brugge, BELGIUM

ABSTRACT
Existing onshore resources of raw materials are becoming more and more depleted. If the current
demands remain or increase over the coming years, alternative resource areas will become highly
interesting. Given these circumstances and the recent technological advances in the field of deep-
sea mining, the interest in the deep-sea is growing. Global Sea Mineral Resources NV (GSR), part
of the DEME-group, has signed a contract with the International Seabed Authority (ISA) in 2013,
giving GSR the exclusive rights to do exploration for polymetallic nodules in a license area in the
Clarion Clipperton Fracture Zone (CCFZ), part of the Central Pacific Ocean.
The assigned area is situated around 125°W and 15°N and covers over 75000 km² divided over
three areas named B2, B4 and B6. Since 2014, GSR has organised and executed two
multidisciplinary research cruises in the assigned license area, mainly focussing on biological,
geological and geotechnical research.
Geotechnical characterization of deep-sea bottom sediments is crucial for the development of
new mining technology. In order to assess the workability of a tracked vehicle in a deep-sea
environment, bearing capacity calculations are needed. This assessment requires input of soil
strength parameters over several meters of depth. Such data is not publically available. As such,
GSR/DEME and DotOcean NV joined forces and started the development of a penetrometer for a
deep-sea environment, over 4000 m deep: the Deep-Sea GraviProbe.
This publication starts with an elaboration on the design process of the GraviProbe. Main
challenges for the design are related to the testing environment, such as dealing with high
pressures (over 400 bar), high accuracy measurements in soft sediments, workability and
repeatability. Testing procedures of the equipment, test results and a number of highlights of the
first deployment in the GSR license area at more than 4000 m depth, are also presented.

Keywords: Deep-sea mining, penetrometer, soft soil, polymetallic nodules, deep-sea

sediments
* presenting author

1. INTRODUCTION
In 2013, the International Seabed
Authority (ISA) granted Global Sea
Mineral Resources NV (GSR) a 15 year
exclusive exploration right for a 75000 km²
license area in the Clarion Clipperton
Fracture Zone (CCFZ) in the Pacific
Ocean (Figure 1). The CCFZ is a
submarine fracture zone of ca. 7000 km
long, well-known for the presence of
polymetallic nodules at its abyssal plains
at over 4000 m depth. Since the
acquisition of the exploration license, two
scientific research cruises have been
organised and executed by GSR.
Figure 1: Exploration areas assigned by the
ISA in the CCFZ (from www.isa.org.jm).
The first research cruise, named
GSRNOD14A, was a 66 days cruise in
July-September 2014. It was a GSRNOD14A and GSRNOD15A.
multidisciplinary cruise, combining
biological, geological, geophysical,
geochemical and geotechnical research.
The main objective of this cruise was to
acquire a complete overview of the
bathymetry in the concession area (over
80000 km² of multibeam echosounder
data were gathered) and to collect the first
nodule and sediment samples. Sediment
samples were obtained by box core
sampling. Part of the sediment samples
was preserved for geotechnical testing.
After the cruise, onshore laboratory testing
was performed on some of the acquired
samples to determine their basic
geotechnical properties (classification and
strength parameters).
62
After analysis of the bathymetry and
backscatter signal data gathered during
GSRNOD14A, three areas of interest
inside the license area, each ca. 200 km²,
were selected in consultation with GSR’s
scientific partners and experts for more in
depth investigation. High-resolution
geophysical imagery and physical
sampling inside these areas was executed
during the 2015 expedition, named
GSRNOD15A, which took place in
September-October 2015. This
multidisciplinary cruise had many scopes,
including an important geotechnical one:
detailed investigation of the strength of the
upper soft seabed sediments.
One of the possible exploitation
methods for the nodules involves a
seabed mining vehicle with a crawling
system. As such, it is important for GSR to
determine the bearing capacity of the
seabed in the areas of interest. Bearing
capacity analysis requires the input of soil
strength parameters and the definition of
strength variation with depth. Therefore,
the geotechnical scope of the
GSRNOD15A campaign primarily
consisted of determining these strength
characteristics by doing in-situ tests.
These in-situ test results could then be
compared with more traditional
geotechnical tests on sediment samples
from box cores, collected both during
2. THE NEED FOR IN-SITU DATA
2.1. Which data is needed?
For mining operations with a tracked
vehicle on the seabed, the workability of
that vehicle needs to be evaluated. The
design of the tracks and the vehicle itself
will be in function of the encountered soil
conditions in the upper seabed sediments.
As a starting point, a tracked vehicle with
2 m wide tracks, operating directly on the
sea bottom is assumed. To estimate the
extent of the zone underneath the tracks
that will be affected by the weight of the
vehicle, a reference is made to traditional
soil mechanics.
 In traditional soil mechanics, the vertical
extent of the soil affected by a shallow
foundation is defined by the well-known
Terzaghi failure surface. An active zone
directly below the foundation is linked at
both sides to a passive zone by a
transition zone. Shearing takes place
along this failure surface in the subsoil.
The encountered shearing resistance is
determined by the strength parameters of
the soil along this failure surface. It is
expected that the failure surface below the
tracked vehicle will be similar in shape. If it
is assumed that the 2 m wide tracks are
comparable to the width B of the
foundation, it can be expected that the soil
below the tracks will be affected by the
weight of the vehicle up to 2 x B = 4 m
depth. Therefore, the goal was set to
achieve in-situ strength information up to 4
m below the seabed surface.
In practice, the tracked vehicle should
be employable over a relatively large area.
Therefore, the objective was not only to
collect strength data up to 4 m deep, but
also to collect these data over a large
area, in an efficient, safe and fast way.
Unfortunately, traditional sampling and
testing techniques do not comply with this
goal. As a result, collecting these data by
means of penetrometer testing logically
came up as the most beneficial option at
that time.
2.2. Penetrometer design requirements
The application of penetrometers in
shallow and deep-sea environments is not
uncommon. A lot of research has been
done on this subject (e.g. Stegmann
(2007), Steiner (2013)) and several
companies provide commercial services
with penetrometers in deep-sea
environments. However, for the
application in this specific case, there
were some additional challenges.
First of all, the depth of the three areas
of interest ranges between 4500 m to
5000 m. Applications of penetrometers at
such great depths are extremely scarce, if
even existing at all. These depths also
involve water pressures up to 500 bar,
63
under which the apparatus needs to
remain functional.
A second challenge was the expected
strength of the sediments. Literature study
(e.g. Tisot (1986), Rey (1988)) and results
of laboratory tests on samples from
GSRNOD14A, indicated only very soft
sediments in the upper soil layers of the
abyssal plains in the CCFZ. To acquire
high accuracy measurements in these soft
sediments, a penetrometer cone capable
of registering very low forces during
penetration of the soil is required.
Thirdly, the great depths at which the
testing takes place require a penetrometer
and corresponding operational procedure
that allows executing several
measurements during a single
deployment. This eliminates the need to
retrieve the penetrometer after each test,
which results in a very high efficiency.
Another challenge was related to the
specific dimensions and available support
equipment on board of the R/V Mount -
Mitchell, the research vessel for the 2015
expedition cruise. It was of critical
importance that the penetrometer could be
deployed with the available A-frame and
winch, without major modifications.
Last but not least, when working with
commercially available deep-sea
penetrometers, whether or not adapted to
the specific requirements of this area, the
price factor and short term availability also
have an important role. Considering the
above, GSR and DotOcean decided to
join forces and started the development of
a deep-sea penetrometer from scratch,
but optimised for this specific environment.
In summary, the most important design
requirements for the penetrometer were:
(1) strength data collection up to 4 m
deep, (2) operational depths of 4000 m to
5000 m and related water pressures, (3) a
high sensitivity cone for strength
measurements in very soft soils, (4) a
flexible system to anticipate altering
ground conditions, (5) a system capable of
executing high amounts of measurements
during a single deployment, (6)
deployment from the available research
vessel and (7) a ‘one-person system’; as
in system needs to be ‘simple’ in set-up,
deployment and data interpretation. The
result of this design process was a
penetrometer called the Deep-sea
GraviProbe (Figure 2). Two complete
GraviProbe sets were made available for
the 2015 GSR campaign; one for
deployment and a spare one.
3. THE DEEP-SEA GRAVIPROBE
3.1. Components
A distinction is made between passive and
active hardware components. The active
components are responsible for the actual
measurements. The most important
passive components are shown in Figure
2.
The main body of the GraviProbe
consists of a thin, hollow tube and a wider,
solid base. The wider base has six
threaded holes, which allow securing six
threaded rods on one side and the bottom
plate on the other side. Three lifting eyes
are attached to the main body for hoisting
operations. The hollow tube of the main
body is sealed on top by a closing lid,
which has a fourth lifting eye.
Semi-circular weights can be slid over
the threaded rods of the main body to vary
the total weight of the GraviProbe.
Weights are secured onto the body with
the protector plate and nuts on top.
The bottom plate is secured to the
wider base of the main body by means of
six hex cap screws. The shaft holder is
screwed into the centre of the bottom plate
and holds the shaft. The large difference
in diameter between the bottom
plate/body and the shaft prevents
excessive penetration in softer soils.
At one side, the 4 m long shaft has a
screw thread which allows connecting it to
the bottom plate and main body via the
shaft holder. The shaft has a diameter of 5
cm. The cone is connected to the other
side of the shaft.
A custom-made camera and lighting
system was also fitted onto the
GraviProbe. The position of the camera
64
and lights focused on the part of the shaft
directly below the body. This camera
position allows to verify whether, and in
which way, 4 m penetration is achieved.
The camera and lights were supplied with
extra battery packs and had an operation
time of ± 4 h.
Figure 2: Sketch of the Deep-sea GraviProbe.
The active hardware parts of the
GraviProbe are responsible for actual
measurements and data registration. The
most important components are the
pressure bottle, the water pressure
sensor, the penetrometer cone and the
accelerometers.
The pressure bottle is the heart of the
GraviProbe and is located in the hollow
tube of the main body. It contains the
printed circuit board to which the sensors
are connected, the microSD card and the
battery pack. The pressure sensor is
connected to the bottom of the pressure
bottle. The cone is connected to the top of
the pressure bottle. There are two more
connections at the top: a connector to
switch the GraviProbe on/off and the data-
output connection.
The GraviProbe’s cone is filled, in vacuum
circumstances, with a carnation mineral oil
and contains the strain gage inside. The
oil reservoir is sealed from the outside by
a nitrile rubber membrane. The detailed
inside configuration of the cone is part of
DotOcean’s intellectual property and will
not be discussed in this publication.
3.2. Data up to 4 m deep
One of the requirements for the
GraviProbe was to collect strength data up
to 4 m below seabed. As a consequence,
the GraviProbe should be capable of
penetrating the cone and the 4 m long
shaft to which it is connected, over this
distance into the soil. The GraviProbe’s
(GP) energy to penetrate the soil over 4 m
consists of a kinetic term and a potential
energy term:
in which:
- Epot = potential energy of the GP
- Ekin = kinetic energy of the GP
- m = buoyant mass GP
- g = gravitational acceleration
- z = penetration depth GP
- v = velocity GP
The equation indicates that there are
two parameters which can contribute
positively to the available energy: (1) an
increased penetration speed v and (2) an
increased buoyant mass m. Due to
operational and safety limitations, and in
order to eliminate dynamic effects as
much as possible, the penetration speed v
was fixed and assumed to be equal to the
speed of the winch used on board of the
research vessel. Consequently, the only
variable parameter is the mass m. The
“naked” GraviProbe has specific
dimensions and is attached to a ca. 4,5
km long cable. This configuration defines
the minimum buoyant mass of the
GraviProbe. However, the design with
modular weights allows adding (or
removing) weights to the body, in order to
vary the available (potential) energy of the
GraviProbe.
The energy of the GraviProbe is
necessary to overcome the resistance
encountered during penetration of the soil.
The soil resistance is calculated using the
65
traditional bearing capacity formulas for
piles. The total soil resistance Qt is split up
into a tip resistance Qb and resistance
related to friction along the shaft Qf:
in which:
- Qt = total soil resistance
- Qb = tip resistance
- Qf = friction resistance
- Ab = tip area
- Nc = bearing capacity factor
- As = shaft area
- Į = friction coefficient
- cu = undrained shear strength
The soil resistance is a function of its
strength. As a starting point for a soil
resistance calculation, a design soil
profile, indicating variation of undrained
shear strength with depth, is needed. An
extensive literature study was conducted
to gather the most important available
geotechnical information in the CCFZ (e.g.
Tisot (1986), Rey (1988)). Although the
amount of data is very limited, especially
in comparison to the extent of the area,
there were no indications that the
undrained shear strength of the soil is
higher than 15 kPa at 4 m depth. With this
input, a design shear strength profile was
selected and applied as a starting point for
soil resistance calculations.
To calculate the soil resistance, a low
impact velocity is considered. In doing so,
it is assumed that phenomena related to
dynamic penetration (e.g. strain rate
effect) are of negligible importance. As a
lower bound approach, one could even
neglect the dynamic effect and assume
that the probe penetrates under its self-
weight when placed on the sea bottom.
Based on the assumptions above,
equilibrium was searched between energy
of the GraviProbe and encountered soil
resistance during 4 m penetration in the
design soil profile. It turned out that the
weight of the GraviProbe should be 950
kg in order to penetrate 4 m of the upper
soft soil.
3.3. Depth and high pressures
The areas of interest in the GSR license
area are situated at more than 4000 m
depth. At these depths, water pressures
can be over 400 bar. During testing, these
are registered using a water pressure
sensor which is installed at the bottom of
the GraviProbe’s body. To deal with these
high pressures, the GraviProbe has
several modifications.
First of all, electronic components are
grouped in the pressure bottle; a small
pressure vessel container located in the
hollow tube of the main body, and capable
of withstanding pressures up to 500 bar.
Another important modification is the
vacuum filling of the cone with a carnation
mineral oil. This makes the cone
unaffected by the surrounding high water
pressures and makes it possible for the
strain gage inside to register relatively
small forces on the cone during
penetration of the soft sediments. Detailed
design of the interior of the cone is part of
DotOcean’s intellectual property and will
not be discussed in this publication.
Evidently, prior to departure and
deployment, the pressure bottle, cone and
water pressure sensor were subjected to
different kind of pressure tests in
hyperbaric chambers (see paragraph 4.1).
3.4. Cone sensitivity
The cone will penetrate very soft soil at a
certain speed. It was expected that the 4
m penetration would not take more than 4
to 5 seconds. For a qualitative strength
characterisation over the entire 4 m, the
cone has to be able to execute lots of
measurements during a relatively short
period of time. A high sampling frequency
is thus preferred. As such, DotOcean has
selected a cone with sampling frequency
of 2 kHz; 2000 measurements per second
are registered and stored.
In order to register low forces on the
cone tip, a strain gage was selected with
66
an optimal measuring range between 20 N
and 2000 N. With this accuracy range it
was assumed that the cone is able to
measure undrained shear strengths (cu)
between cu = 0,5 kPa and cu = 50 kPa.
This range completely covers the strength
values as defined in the design strength
profile and has sufficient margin to
anticipate for stronger soils. The
configuration of the strain gage inside the
cone is part of DotOcean’s intellectual
property and will not be discussed in this
publication.
3.5. Flexible system
The total weight, and thus the (potential)
energy, of the GraviProbe can be modified
by adding or removing weights. These
weights are semi-circular elements in lead
alloy and weigh 25,4 kg per piece.
The “naked” GraviProbe, ready to use,
but without any extra weights attached to
it, weighs 460 kg. As a starting set-up, 18
extra modular weights were attached,
bringing the total weight of the GraviProbe
to 917 kg. This is in line with the mass
which was determined from the energy vs.
soil resistance equilibrium in paragraph
3.2.
In total, a maximum of 62 semi-circular
weights, equalling 1575 kg, can be added
to the “naked” GraviProbe. This brings the
maximum total weight of the GraviProbe at
2035 kg. This modular weight system
allows adjusting to softer of higher
strength soils compared to the design
strength profile. If the soils are softer, the
GraviProbe will penetrate more than 4 m
deep and also its base plate will start to
penetrate the soil. Because of the large
difference in diameter between shaft and
base plate, excessive penetration will be
prevented. In this case, weights can be
removed to a minimum value of ca. 460
kg. If the GraviProbe does not penetrate 4
m deep because of higher encountered
soil resistance, weights can be added, to a
maximum of 2035 kg, to increase the
energy of the GraviProbe and overcome
the extra resistance. With the weight
flexibility, it is definitely possible to operate
within the range of the strain gage
sensitivity; i.e. 20 N to 2000 N or assumed brings the GraviProbe from its horizontal
undrained shear strength (cu) of the soil position on deck to a vertical position
between ca. cu = 0,5 kPa - 50 kPa. required for deployment and testing,
without damaging the cone. The LARS
3.6. Efficiency and simplicity of the
can be easily and quickly installed on any
system
ship’s deck.
With the given winch speed, it takes the
The operational procedure itself is also
GraviProbe 1,5 h to 2 h in this setting to
simple; there is no need for specialized
reach the sea bottom and the same
equipment on board. A standard winch
amount of time to come back up. This
and A-frame are sufficient to start
makes the time available for testing rather
operations for testing. The only parameter
limited and it should therefore be used
to control during testing is winch payout:
very efficiently. As such, the equipment
the GraviProbe can be moved up or down.
and testing procedure should allow
During operation it is assumed that full
multiple penetrations and data registration
penetration has occurred as soon as slack
without retrieval after each penetration.
on the cable is visible on deck. Once this
There are several technical characteristics
is observed, the probe is retrieved over a
of the GraviProbe which are specially
couple of tens of meters, moved to the
designed to accomplish this.
next location and ready for another
For example, the system works on a penetration.
lithium-ion based battery which is located
Also the visualisation and related
in the pressure bottle. The power source
interpretation of the data is relatively easy.
goes down with the probe. The battery
The data is recorded and stored as a time
has 24h autonomy once fully charged and
series. At the time the cone encounters
therefore maximizes the time available for
the seabed, the GraviProbe decelerates
testing once the probe is at the sea
and simultaneously, the forces on the
bottom. Another feature is the data
cone increase. The moment at which the
storage (accelerometers, water pressure
wider base plate hits the sea bottom, and
sensor and cone) on a micro-SD card in
the shaft and cone have thus penetrated 4
SEG-Y format. The micro-SD card is also
m into the soil, is associated with a very
located inside the pressure bottle. The
sharp deceleration and eventually a full
data is stored in such way that every SEG-
stop of the GraviProbe (v=0).
Y file contains 20 min. of data. This is a
Accelerations and decelerations are
safety measure: if one file is corrupted,
continuously registered with
only 20 minutes of data is lost. And since
accelerometers. The full stop is the
the data is directly stored inside the
reference point in the time series for
pressure bottle, there is no need for
‘backward’ data selection; the data during
retrieval after each test.
the 4 m penetration. Based on a double
It was also requested that the system integration procedure, this part of the time
would be as ‘simple’ as possible, both in series can be converted to a depth series.
terms of assembly, operation, retrieval of This eventually results in 4 different
data and interpretation. The GraviProbe graphs for each penetration; (1)
has several adaptations to meet this acceleration vs depth, (2) speed vs depth,
request. (3) dynamic cone resistance vs depth and
First of all, it is very simple to assemble also (4) the water pressure at the sea
the GraviProbe. It is a “plug and play” bottom. This procedure is executed
system with a limited amount of cables automatically by specially developed
and connectors. Secondly, the overboard software programs and results in a simple
procedure is also relatively simple; the visualization of the data. The 4 graphs
ship’s A-frame places the GraviProbe into allow easy interpretation and data
a launch and recovery system (LARS) comparison on board and, if necessary,
installed on the back deck. The LARS
67
they allow adjustments of the GraviProbe
set-up for the next operation.
4. TESTING CAMPAIGN
Despite a limited time frame, the
GraviProbe was subjected to several tests
before it was shipped to the departure
location of the 2015 research cruise.
4.1. Pressure tests
The water pressure sensors were
subjected to a pressure test in a closed
hyperbaric chamber. Pressure was
increased up to 600 bar in steps of 150
bar. This was also performed with the
pressure bottles, containing all the
electronics. Pressures were increased up
to 500 bar in the hyperbaric chamber.
The two cones were also subjected to
tests in hyperbaric chambers. During the
test, the cones were already filled with oil
and the strain gage was already mounted
inside the cone. A maximum pressure of
500 bar, in steps of 50 bar with a 10 s hold
at each step, was applied to the cones.
The cones were also loaded to calibrate
the strain gage. For one of the cones, the
load swings were performed under
hyperbaric circumstances. Unfortunately,
due to time constraints, the load swings
for the other cone were performed under
atmospheric pressure circumstances. This
cone will be post-calibrated after the
GSRNOD15A cruise.
The strain gage was subjected to forces
varying between 0 N and 2500 N.
Because very soft soils were expected,
more calibration points were scheduled in
the lower region, i.e. 0 N – 250 N, in order
to increase the accuracy at small loadings.
According to the corresponding calibration
file, the strain gage, at a gain of 64 bit, is
capable of registering a load increase of
377 Pa on the cone.
4.2. Full-scale tests
In July 2015 two full-scale in-situ tests with
the GraviProbe were executed at the
DEME-HQ harbor in Antwerp, Belgium. A
bathymetric survey in the small harbor
indicated at least 2 m to 3 m of very soft
sediments on top of a more sandy layer,
68
within a reachable distance from the quay
wall. The same survey also indicated that
within that same distance, the water
depths could be sufficiently large (tide
dependent) in order to fully submerge the
GraviProbe. As the targeted area was
close to the quay wall, the GraviProbe can
be put into the water with a crane standing
on the quay wall.
The first full-scale test focused on the
assembly of the GraviProbe (Figure 3). It
was the first time that all (passive)
components were put together. For this
test, the pressure bottle, the water
pressure sensor and the cone were not
installed on the GraviProbe. At the tip of
the shaft a dummy cone was attached.
Not only did this test allow us to familiarize
with the different GraviProbe components
and its assembly, it was also a preparation
of the planned second test (with active
components).
After all, for this second test, in which a
test of all the electronics was foreseen,
some items still needed to be checked.
For example, how could the GraviProbe
efficiently be brought from its horizontal
assembly position into its vertical testing
position without damaging the cone? It
was also very important to verify the
presence of the 2 m to 3 m soft sediments
at the foreseen testing location, exclude
the presence of obstacles or very hard
substratum which could damage the cone
and verify the water depth for a certain
tide.
Figure 3: First full-scale test: impressions from
site.
In the end, this first test was considered
successful and the scope for the second
test was finalized.
 The aim of the second full-scale test
was to perform actual measurements.
Therefore the pressure sensor and cone
were connected to the pressure bottle and
all components were installed into the
GraviProbe. This was also the moment to
test the GraviProbe software, which is
needed to switch the electronics on/off
and retrieve and visualize the data after
testing.
The test was executed according to the
same operational procedure as the first
full-scale test. The crane lowered the
GraviProbe through the water column at a
controlled speed. As soon as the cone
and the shaft started to penetrate the soft
sediment, a clear decrease in weight on
the load cell of the crane was observed.
After 2 m to 3 m of penetration, an even
sharper decrease in weight on the load
cell of the crane occurred. This was
identified as the moment when the very
soft soil layer was fully penetrated by the
cone and the harder substratum (sandy
layer) below was encountered. At that
moment, the crane stopped lowering the
GraviProbe and started retracting it in
order not to damage the cone. Next, the
Figure 4: GraviProbe DeepSea Raw Data Viewer with cone-signal (black).
The second program is the “Deep-sea
GraviProbe software”. This is a more
advanced program which automatically
generates different graphs for further
interpretation. The program reads the
SEG-Y files and automatically identifies
the peaks corresponding to a penetration
of the GraviProbe within the time series.
69
crane repositioned its boom and a new
location was tested using the same
procedure. After testing, the pressure
bottle was connected to a computer and
the data was retrieved. DotOcean has
provided two software programs to read
the produced SEG-Y data files.
The first program is the “Raw Data
Viewer”. This tool can read a SEG-Y file
and visualize the different sensor
registrations, i.e. forces on the cone and
water pressure, in time. An example of a
SEG-Y file obtained during this second
full-scale test is given in Figure 4. The x-
axis indicates the time in seconds and the
y-axis indicates the integer values of the
sensor registrations. For example, the
black line in Figure 4 depicts the
registration of the strain gage, thus the
forces on the cone. The three peaks, at
regular moments in time, correspond to
penetrations in the soil and thus a sharp
increase of the forces acting on the cone.
Based on the calibration file of the strain
gage, these integer values can be
translated to actual forces in Newton and
to stresses in Pascal.
The program selects all the sensor
registrations within that time frame and
combines them with the accelerometer
data during that same time frame. As
explained before, the combination of these
data allows the production of 4 different
graphs for each penetration; (1)
acceleration vs depth, (2) speed vs depth,
(3) dynamic cone resistance vs depth and
also (4) water pressure at the sea bottom.
Graphs presented in Figure 5 belong to
one of the penetrations executed during
the second full-scale test. Because of the
very shallow depth in this testing area, the
fourth graph of the water pressure is
omitted.
Figure 5: Automatic generation
acceleration, velocity and dynamic cone
resistance vs. depth at a penetration location.
5. RESEARCH CRUISE GSRNOD15A
5.1. Scope GraviProbe
The GraviProbe’s scope was to collect
as much strength data as possible up to 4
m deep, over a relatively large area.
Ideally, the penetrations are relatively
close to each other - in order to make
interpolations from one point to another -
and the results can be verified and
confirmed by other data.
The deployment location for the
GraviProbe was determined taking into
account several considerations. The
preferred testing location should be an
area where only soft sediments occur.
Hitting hard outcrops or other obstacles
could damage the GraviProbe’s cone.
Therefore, deployment preferably takes
place in the so-called “high-resolution
box”; a selected small area where high-
resolution side scan sonar measurements
and photographs are available (=another
type of data acquired during the
GSRNOD15A research cruise). These
data allow eliminating as much as possible
the presence of hard rock outcrops or
other obstacles.
The preferred testing area should also
be situated relatively close to a
GSRNOD15A box core sampling location.
This enables the comparison between the
in-situ and laboratory strength
measurements of GSRNOD15A.
Furthermore, the preferred testing area
should be close to a location where a
GSRNOD14A box core sample was taken.
This enables the comparison between the
in-situ strength measurements of
GSRNOD15A and laboratory strength
measurements of GSRNOD14A.
Also from an operational point of view
some considerations were taken into
account. First of all, good weather
conditions are required in order to get the
GraviProbe overboard safely and observe
correctly slack on the cable during
penetration. Secondly, the GraviProbe
operations should be aligned with the
of
deployment of other types of equipment
and their testing locations. It is also
important to note that the R/V Mount
Mitchell is not equipped with a dynamic
positioning system. Therefore, the testing
locations were located along a linear
trajectory and not in a grid. To follow the
scheduled trajectory, one had to make the
best use of the existing currents and wind
at that time.
Eventually, keeping in mind the
considerations as formulated above, a ± 3
km to 4 km straight line was selected in
the designated high-resolution area
(Figure 6). Depending on the testing area,
12 to 15 points were selected along this
line at a distance of ± 250 m from each
other. At each point 5 penetrations with
the probe were planned.
Figure 6: (left) License area B6, (middle) area
of interest B6 South, (right) high-resolution
box inside B6 South and proposed
GraviProbe trajectory for testing.
70
 It was assumed that these 5
penetrations, at very short distances from
each other, would result in similar strength
profiles and would confirm the testing
results at a specific location. On the other
hand, testing at points located about 250
m from each other would allow to observe
possible regional variations within a
testing area. Comparison of the
measurements between the three different
testing areas enables the potential
identification of larger scale regional
variations.
5.2. Operational procedure
The GraviProbe is moved using the ship’s
A-frame and tugger lines from its storage
position on deck into the launch and
recovery system (LARS). Then, the
GraviProbe is “switched on” and camera
and lights are installed. While sailing to the
first testing location, the GraviProbe is put
overboard and starts descending (Figure
7).
Figure 7: GraviProbe ready for descent.
During the descent, 10 to 20 second
stops were executed at regular depth
intervals. These depths were located with
the cable counter on the winch. The
purpose of these stops was two-fold: (1)
recognition point in the data time series
and (2) opportunity to match water depth
with water pressure. The descent was
stopped before the GraviProbe was
expected to touch the seabed. As soon as
the ship reached the first testing location,
the descent continued until the first
penetration took place; i.e. until slack on
the cable was observed on deck. Then the
GraviProbe was retracted over ± 30 m to
50 m above seabed and released again
for the next penetration. This operation
was repeated 5 times in total at each
location. Considering 12 to 15 different
locations per operation, in total 60 to 75
71
penetrations were realized during a single
deployment.
During the operation itself, metadata
were registered, both automatically and
manually. These metadata, for example
time of important events or observations,
are simplifying data processing
afterwards.
5.3. Results
The first deployment of the GraviProbe
took place on 21/09/2015. Once the
GraviProbe was back on board, the
camera material was analysed and
confirmed that the GraviProbe penetrated
the sea bottom each time as planned
(Figure 8). Preliminary analysis of the data
on board also indicated consistent results,
in the expected range (Figure 9).
The GraviProbe has been deployed in
the three areas of interest. For each
deployment, the scope as defined above
was successfully achieved making use of
the operational procedure as described
above.
Figure 8: Impact of the wider GraviProbe body
at the seabed, after 4 m penetration of the
shaft & cone. Notice the perfectly straight
alignment during penetration.
In the first area of interest, 12 different
locations have been tested with a total of
56 penetrations observed on deck. In both
the second and third area of interest, 15
different locations have been tested with
for each area 75 penetrations observed on
deck. At the moment, these results are still
being processed, interpreted and
discussed.
6. CONCLUSION
In order to determine in-situ strength
parameters of very soft soils up to 4 m
below the seabed in a challenging deep-
sea environment, GSR and DotOcean
developed the Deep-Sea GraviProbe. This
penetrometer is user friendly and has
several modifications to deal with
pressures over 400 bar, execute high
accuracy measurements in soft sediments
and optimise workability, repeatability and
data interpretation. Before deployment in
the Pacific Ocean, the equipment has
been thoroughly tested. During the 2015
GSR research cruise, the penetrometer
has been successfully deployed in three
different areas within the GSR license
area. In total, more than 200 penetrations
have been executed. Camera footage and
preliminary analysis of the penetrations on
board indicated 4 m penetration and
consistent results. At the moment, detailed
processing, analysis and interpretation of
the data is executed.

Figure 9: GraviProbe data example of penetration during first deployment.

.
Géologie de Nancy, 314 p., Nancy,
ACKNOWLEDGEMENTS France.
The authors would like to thank the Steiner, A. (2013) “Sub-Seafloor
“Belgische Groepering voor Characterization and Stability of
Grondmechanica en Geotechniek Submarine Slope Sediments using
(BGGG)”, the Belgian Member Society of Dynamic and Static Piezocone
the ISSMG, to select this publication as Penetrometers”, Doctoral Thesis,
one of the two Belgian representatives Faculty of Geosciences, Bremen
and by doing so, giving us the opportunity University, 325 p., Bremen, Germany.
to attend the 2016 EYGEC-conference at Stegmann, S. (2007) “Design of a free-fall
Sibiu, Romania. penetrometer for geotechnical
characterisation of saturated sediments
REFERENCES
and its geological application”, Doctoral
Rey, S. (1988) “Environnements Thesis, Faculty of Geosciences,
sédimentaires et propriétés Bremen University, 181 p., Bremen,
géotechniques de sediments marins Germany.
profound de deux zones du Pacifique Tisot, J.P. (1986) “Propriétés
Tropical nord-est”, Doctoral Thesis, géomécaniques des sediments des
Centre de recherches en mécanique et grands fronds océaniques”, Mémoires
hydraulique des sols et des roches, Sciences de la Terre, Foundation
Ecole Nationale Superieure de scientifique de la géologie et de ses
applications, 351 p., Nancy, France.
72
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Determination of characteristic values of geotechnical

parameters

Monica LÖFMAN1*
1
Aalto University School of Engineering, Department of Civil Engineering, Espoo,
FINLAND

ABSTRACT

According to Eurocode, statistical methods may be used in the determination of characteristic

values of geotechnical parameters. In this paper, different statistical methods represented by
Eurocode and Finnish norm are described and compared using undrained shear strength data
from Perniö test site. Both the current version of Eurocode and the suggested new Annex HS are
considered. The effect of number of observations to the ‘cautious mean’ characteristic value of
each method is studied via simulations. Both COV-unknown and COV-known -methods of
Eurocode are considered. In addition, factors in Annex HS affecting the derived characteristic
value are studied using sensitivity analysis. Based on the results, it is clear that when the number
of observations is low, all these methods must be treated with cautiousness. The COV-known –
method seems to be the most reliable one, assuming that appropriate ‘known’ COV is adopted.

Keywords: characteristic value, COV, coefficient of variation, Eurocode

hand, the process is systematic and the

1. INTRODUCTION
According to Eurocode, statistical methods
may be used in determination of
characteristic values of geotechnical
parameters. Even so, in practical
geotechnical design, deterministic
approach is commonly used instead due
to its simplicity and traditions in design
(Lee et al. 1983, p. 58). However,
deterministic approach leads to
characteristic values that are highly
subjective and thus uncertain as the
selection of cautious estimate is based on
engineering judgement (Phoon 2008, pp.
3-8). In statistical methods on the other
* presenting author
uncertainty in the soil property, all the
available prior information considered, is
defined quantitatively.
In this paper, different statistical
methods represented by Eurocode and
Finnish RIL-guideline are described and
compared based on undrained shear
strength results from Perniö test site. Both
the current version of Eurocode and the
suggested new Annex HS are considered.
Any complementary information and a
priori knowledge can be taken into
account in the determination of
characteristic values. In Eurocode, this
can be done by assuming that the
coefficient of variation (COV) of the soil The scatter in the test results is caused
property is known (so called COV-known - by both inherent variability of soil and
method) (EN 1990 2005, Annex D). Even measurement error. (Lee et al. 1983, pp.
though not covered in Eurocode, another 57-58, Phoon & Kulhawy 1999). The
possibility would be to use Bayesian uncertainty caused by inherent variability,
approach, which is a very powerful tool measurement error or other factors can be
especially in reliability-based design quantified by using coefficient of variation
(RBD). or COV:
In the analysis, the effect of number of
observation to the ‘cautious mean’ (1)
characteristic value of each method is
studied via simulations. Both COV-
unknown and COV-known -methods of where ıx is the standard deviation of the
Eurocode are studied and compared. In random variable x; ȝx is the expected
addition, factors in Annex HS affecting the value; SDx is the standard variable of the
derived characteristic value are studied sample and; xm is the mean of the sample.
using sensitivity analysis. High COV implies high uncertainty.
Inherent soil variability is represented in
Figure 1. The actual value of the soil
2. DETERMINATION OF CHARACTE- property ȟ(z) varies through depth, but one
RISTIC VALUE can determine the trend function t(z). The
fluctuating component w(z) represents the
2.1. Definition according to Eurocode inherent soil variability. This spatial
EN-1997-1 defines the characteristic value variability is one of the main reasons for
as being “selected as a cautious estimate the need of conservatism in the definition
of the value affecting the occurrence of of characteristic value. In the figure,
the limit state”. In the definition “selected” correlation length ș is the distance within
emphasizes the importance of engineering which points are significantly correlated
judgement, and “cautious estimate” means (Fenton & Griffiths 2008, p. 103, Phoon &
that some conservatism is required and Kulhawy 1999).
finally the selected value must relate to the
limit state (EN 1997-1 2004, Frank et al.
2004, pp. 24-28).
According to Frank et al. (2004), when
selecting the characteristic value, two
major aspects are (i) the amount of
knowledge of the parameter values and
the degree of confidence in the knowledge
and (ii) the soil volume involved in the limit
state considered and ability of the
structure to transfer loads from weak to
strong zones in the ground (Frank et al.
2004, pp. 24-28).
The amount and degree of confidence
in the information depends on (i) the
amount of information (local test results
and other relevant information such as a
Figure 1. Inherent soil variability (Phoon &
priori knowledge and (ii) the scatter of the
Kulhawy 1999).
results, which is caused by the variability
of soil (Frank et al. 2004, pp. 24-28).

74
2.2. Eurocode 7
The soil volume involved in the limit state
affects the definition of characteristic
value, which can be either (a) cautious
mean, which is the estimated mean value
corresponding to a 95 % confidence level
or (b) local low value, which is the
estimated 5 % fractile (EN 1997-1 2004,
Frank et al. 2004, pp. 46-49).
Thus according to definition, there is a
probability of 95 % that the (unknown)
mean value governing the occurrence of
a limit state in the ground is more
favourable than the selected mean
characteristic value. 5 % fractile on the
other hand means that there will be only 5
% probability that somewhere in the layer
considered there is an element of soil
having property values lower than the
characteristic value (Frank et al. 2004, pp.
46-49).
So when is it required to use a cautious
mean (large soil volume) and when a local
low value (small soil volume)? The
difference between these two cases is
related to correlation length ș: When ș is
small compared with the dimensions of the
soil volume, low and high local values
compensate, and cautious mean is
adopted. This phenomenon is often
referred to as “averaging” of weak and
strong soil layers. If, on the other hand, ș
is larger compared with the dimensions of
the soil volume involved, the local low
value might affect the occurrence of the
limit state instead. Thus 5 % fractile or a
value somewhere between 5 % fractile
and cautious mean must be used (Frank
et al. 2004, pp. 46-49).
Suggested values for both horizontal
and vertical correlation lengths can be
found from the literature for different soil
properties (Phoon & Kulhawy 1999).
However, it is usually much more
challenging to estimate the dimensions of
the soil volume involved in limit state.
Especially when it comes to the stability of
an embankment, one rarely knows where
the failure surface would occur. However,
if brittle failure or strain softening
behaviour can be expected, usage of the
75
5 % fractile is recommended (Frank et al.
2004, pp. 46-49).
In Eurocode, if statistical methods are
used, characteristic value is defined as
(EN 1990 2005, Annex D, Frank et al.
2004, p. 46):
(2)
where COVx is either the coefficient of
variation calculated for the sample or
tabulated “known” COV for the whole
population and; kn is a statistical
coefficient.
The value of statistical coefficient
depends on the number n of test results
(observations), on the 'type' of
characteristic value (mean or fractile), the
statistical level of confidence required, and
a priori knowledge about the COV (known
or unknown) (Frank et al. 2004, p. 29).
In the case of 95 % reliable mean
value, kn is defined as:
(3)
(4)
where n is number of observations; t0.95n-1
is the value of the t factor of Student’s
distribution (with degree of freedom being
n – 1) corresponding to a probability of 95
%. If COV is known, normal distribution
can be used and the corresponding 95 %
value is 1.645.
However, this method is valid only for
cases where the soil is relatively
homogeneous and there is no significant
trend in the soil property. In addition,
normal distribution is assumed. For some
geotechnical parameters, such as
undrained shear strength, lognormal
distribution is a better fit (Lacasse &
Nadim 1996). Furthermore, lognormal
distribution is non-negative whereas
normal distribution can contain
unrealistically low or even negative values.
If lognormal distribution is used, before
applying the formulae represented above,
the parameter value X must be
transformed to its logarithm Y = ln X
(Frank et al. 2004, p. 29). In Annex D of
Eurocode, determination of Xk for log-
normally distributed properties is provided
in further detail (EN 1990 2005, Annex D).
However, if COV < 30 %, there is not a
significant difference between normal and
lognormal distributions.
2.3. Finnish RIL-guideline
As for statistical methods, Finnish RIL-
guideline only covers the determination of
cautious estimate, but it is based in
Eurocode otherwise. In this method
however, a priori knowledge cannot be
taken into account as in Eurocode 7. In
RIL 207-2009 the characteristic value Xk is
suggested to be calculated as follows (RIL
207-2009 2009):
(5)
where n is the amount of test results and
others are as earlier defined.
In the equation, 1.645 is yet again the
value of normal cumulative distribution
function corresponding to a probability of
95 %.
In this method, it is assumed, that the
calculated standard deviation SD
represents the distribution of the whole
population; if the standard deviation of the
population is unknown, Student t-
distributions should be used instead of
normal distribution. If n = ’, normal and
Student’s distribution actually yield the
same value of 1.645. As a matter of fact,
this method produces the same result as
COVknown -method of Eurocode, if one
uses the calculated sample COV instead
of tabulated ‘known’ COV.
2.4. Suggested Annex HS
‘TC250/SC7/EG11: Characterization’ is a
project, which aims to provide a user
friendly and consistent guidance on
determining characteristic values of
ground parameters, using both the
traditional approach and statistics.
Proposed Annex HS represents a
simplified method which is based on
76
statistics and which considers level of
experience, the amount and quality of test
results and the zone of influence. In this
method, no statistical terms are introduced
at all, which supposedly makes the
method easier to use in practice. The
method fulfills all relevant aspects of the
requirements of Eurocode 7 and is also
open to be adjusted (via factor a)
according to different national experiences
(EG11 2015).
In Annex HS, characteristic value is
defined as (EG11 2015):
(6)
where xm is the mean of the derived
values based on field or lab tests and/or
estimated mean value from comparable
experience and/or estimated mean value
from tabulated soil properties; xextr is the
extreme soil value recorded or estimated
corresponding to an expected extreme
(unfavourable) value for the hypothetical
case of large number of tests; Lv
represents the zone of ground governing
the behaviour of a geotechnical structure
at a limit state. As such, Lv is the vertical
dimension of the zone of influence. Inside
the square root 1 stands for a typical
vertical correlation length of 1 m, and as
such, the term is dimensionless (EG11
2015).
In the equation a is a factor to account
for extent and quality of field and
laboratory investigations or estimation
method, type of tests for selecting derived
values, sampling methods and level of
experience. The suggested range of a is
0.5-1.0. Smallest value of 0.5 is proposed
to be used in the case of several high
quality test values and reliable, good local
site information based on excellent
comparable experiences. Value of a =
0.75 could be used for average quality.
For example, in the calculation example of
Annex HS, a = 0.7 for field vane is
proposed. The most conservative a = 1.0
is suggested to be used when the derived
values are estimated from general
experience or tabulated values (no local
site investigation) (EG11 2015).
Suggested Annex HS states that
derived values (xm and xextr) should be
corrected from uncertainties of the testing
methods as well from the transformation
model used to arrive at derived values
from raw data of lab or field tests. xextr can
also be estimated or confirmed by using
tabulated values of COV (EG11 2015).
3. COMPARISON OF THE METHODS
3.1. The effect of the number of
observations
In the first analysis, fall cone test results
from 8 sampling points near each other
were studied. The studied site is located in
Perniö, Finland. In this site, full-scale
embankment failure test was conducted in
2009 (Lehtonen et al. 2015). The data
used in this study was provided and
studied by Igor Mataic (Mataic 2016).
All the test results are represented in
Figure 2. In the analysis, only the soft
layer at depth of 2.5…5 m is studied.
There is overall n = 28 test results. For
this layer, ‘cautious mean’ characteristic
value of undrained shear strength is
determined using all the methods
described above, and the results are
compared and further analysed.
analysis, the following parameters are
kept constant: (i) ‘known’ COV is 30 %,
which is the suggested standard value for
undrained shear strength (Müller 2013) (ii)
a = 0.75. Since in this analysis the number
of test results varies from n = 3…28, factor
a for average test quality is the most
suitable (iii) Lv = 2.5, which is the
thickness of the layer. Thus it is assumed,
that the whole layer affects the occurrence
of the ultimate limit state. In addition,
normal distribution is assumed.
The determined characteristic values Xk
and the calculated mean values in each
case of n is represented in Figure 3.
Figure 3. The effect of number of observations
to the calculated mean and Xk.
When the number of observations is less
than 8, there is great amount of scatter
which is mostly caused by changes in the
calculated mean. However, when it comes
to COVunknown - and RIL -methods, the
scatter is also partly caused by changes in
the standard deviation SD and COV (see
Figure 4).

Figure 2. Fall cone test results at Perniö site.

First, the effect of number of

observations n is studied. From the
studied layer, samples consisting of n =
3…27 results are pulled from the data
using random number generator in Excel.
For each sample, characteristic value is Figure 4. The effect of number of observations
determined using each method. In this to the calculated SD and COV.
77
Most of the characteristic values and
mean stabilize after n>13, when SD and
COV stabilize as well. Annex HS yields
the highest value when n<10, but as the
number of observations increase, the
value remains approximately constant and
intersects with COVknown –value. All in all,
Annex HS is the most stable because the
results does not depend on the number of
observations at all (factor a being Figure 5. Calculated COV-values for each
constant). sample.
After n>10, RIL and COVunknown yield
approximately the same values, because Thus at low values of n there is a
the value of factor t of Student’s considerable risk that COVunknown - and RIL
distribution gets closer to the value of -methods either under- or overestimate Xk.
1.645 as the number of observations When the determined highest and
increases. lowest characteristic values Xk and means
Overall, the COVknown -method yields are compared (max-min) within each case
the lowest value since the assumed COV of n (Figure 6), it is clear, that COVunknown
(0.30) is much larger than the one and RIL yield the most uncertain values at
calculated from the samples. Note that the low values of n. COVknown, on the other
calculated SD and COV are abnormally hand, is the most stable in almost every
low when n<. Indeed, according to case. Annex HS is relatively stable as
Schneider (1999), statistical methods can well, since Xk is only affected by
be applied successfully only if n>10. Thus calculated mean and observed minimum
it is highly recommended to use COVknown value in this analysis.
-method if there is not enough data (Frank
et al. 2004, pp. 46-47). In this analysis,
COVunknown yielded higher values than
COVknown -method, but this is only due to
the uncertainty brought by small n values
and because the selected ‘known’ COV
was too conservative.

3.2. The uncertainty caused by small

amount of observations
In the second analysis, the effect of small
number of observation is studied. The
parameters are the same as in previous
analysis. In this analysis however, five
random samples are collected from the
data for cases n = 3, 5, 7, 10, and 20. The
characteristic values and means are
calculated for each sample and compared.
For reference, results for n = 28 are
represented as well.
When the values of the calculated COV
in each case of n are plotted, it can be
observed that if n<8, COV is highly
uncertain (Figure 5).
Figure 6. The difference between min and
max values of mean su and Xk at different
cases of n.
The calculated mean and mean±SD
values of the mean su and characteristic
values Xk are represented in Figure 7
(above). The data contains results of five
samples from cases n = 5, 7, 10 and 20
(overall 20 values of mean su and Xk). In
Figure 7 (below) the same results are
represented, the only difference being that
in COVknown -method the ‘known’ COV is
15 %.

78

Figure 7. Mean and mean±SD values of mean
su and Xk. ‘known’ COV being 30% or 15%.
The means of characteristic values are
approximately the same in all methods
expect COVknown when ‘known’ COV is 30
%, which is much larger than the COV
calculated for the whole sample of 28
results (12%, Figure 4). If the ‘known’
COV is adjusted to 15 %, COVknown yields
higher mean and smaller SD. Thus if a
suitable ‘known’ COV is selected, usage of
COVknown leads to more reliable Xk than
COVunknown.
The scatter in Annex HS is yet again
small, since factor a and Lv where not
varied. Next, the effect of varying ‘known’
COV and parameters of Annex HS are
studied.
3.3. Sensitivity analysis – ‘known’ COV
and parameters in Annex HS
In this analysis, the data from the first
analysis is adopted again and the ‘known’
COV is varied based on reported values
for undrained shear strength (Müller
2013). The calculated COVknown -based
values of Xk are shown in Figure 8. For
comparison, values calculated using
COVunknown are represented as well,
containing both the results from the first
79
analysis and the minimum values from 5
samples in studied cases of n.
Figure 8. The effect of n to the calculated
mean and Xk with varying ‘known’ COV.
If COV-values greater than 0.30 are used,
COVknown yields unrealistically low values
of Xk at low values of n. In addition,
‘known’ COV must be over 0.20 in order to
reach the same values as in COVunknown –
method at higher values of n since the
calculated COV for the data is 0.12 (12%)
as discussed in the previous section.
Next, the effect of the parameters in
Annex HS are studied via simulations and
sensitivity analyses. For each uncertain
parameter, a probability density function
(PDF) is defined based on both results
from the simulations and assumptions.
PDF for mean undrained shear strength
is determined by fitting a normal
distribution to the results of the second
analysis (overall 20 results). The
parameters of the PDF are expected value
ȝ = 11.8929 and standard deviation ı =
0.3506. The fitted normal distribution is
represented in Figure 9. In Simulations 1
and 2, this PDF for mean is used, but in
Simulation 3 the input mean su is kept
constant at its expected value.
In the distribution figures, there are two
bars with percentiles and two values. The
gray bar states that based on the input,
the studied variable is between these two
values at a probability of 90 %.
Respectively, these two values taken, the
corresponding probability based on the
fitted distribution is typed in the middle of
the black bar.

Figure 11. Simulation 2: Annex HS Xk

(Lv = 1.0…10 m).

Figure 9. Histogram and fitted normal As can be seen from the histograms, the
distribution of the mean su. highest values of Xk (Figures 10 and 11)
are even higher than the smallest values
In all simulations, for factor a, a triangular of the mean (Figure 9).
distribution with minimum of 0.5, expected
value of 0.75 and maximum of 1.0 is When the effect of each parameter on
assumed. In Simulation 1, for Lv, a uniform the Xk is studied via sensitivity analysis,
distribution with a minimum of 1.0 and a the tornado graphs based on Simulations
maximum of 2.5 (the thickness of the 1 and 2 are acquired (Figures 12 and 13).
studied layer) is assumed. According to
Annex HS, Lv = 1.0 is suggested for
shallow foundations (EG11 2015).
Depending on the problem, Lv can be as
large as 10 m (for example for pile
foundations), and this case is studied in
Simulations 2 and 3. Minimum value xextr is
kept constant at the observed minimum
9.8 kPa.
The simulations consists of 1000
iterations, and the resulted histogram and Figure 12. Simulation 1: Tornado graph based
fitted normal distribution from Simulations on the sensitivity analysis of Annex HS Xk (Lv
1 and 2 of Xk are represented in Figures = 1.0…2.5 m).
10 and 11.

Figure 13. Simulation 2: Tornado graph based

on the sensitivity analysis of Annex HS Xk
(Lv = 1.0…10 m).
Figure 10. Simulation 1: Annex HS Xk (Lv =
1.0…2.5 m).
80
As can be seen from the tornado diagram Yet again Lv has greater influence than
of Simulation 1 (Figure 12), factor a affects factor a. In addition, the highest values of
the derived Xk the most. Interestingly, the Xk are the result of high Lv (Figure 15).
selection of Lv affects the results almost as The statistics of all the simulations are
much as factor a and mean. In Simulation listed in Table 1.
2 Lv has the biggest influence whereas
factor a affects the derived Xk the least
Table 1. Statistics of the simulations
(Figure 13).
In Simulation 3, mean is kept constant Simulation Mean SD min max
at the expected value of 11.89 kPa. Other (Mean su) 11.89 0.351 10.98 12.58
parameters are the same as in Simulation
Sim. 1: Xk
2. This approach enables more accurate 10.68 0.281 9.890 11.76
comparison between the effects of factor a (Lv = 1.0-2.5)
and Lv. Sim. 2: Xk
11.14 0.355 10.05 12.29
The histogram and fitted triangular (Lv = 1.0-10)
distribution of Xk is shown in Figure 14 and Sim. 3: Xk
the corresponding tornado graph in Figure 11.14 0.270 10.02 11.55
(mean constant,
15. The distribution is heavily skewed to Lv = 1.0-10)
the left, implying that the highest values of
Xk are more probable (Figure 14).
According to the statistics, increase in Lv
will result in higher mean values of Xk.
Note that in the Simulation 3 in which the
mean is kept constant the mean and
maximum values of Xk are extremely close
to the observed mean.

Figure 14. Simulation 3: Xk Annex HS (input
mean constant, Lv = 1.0…10 m).
Figure 15. Simulation 3: Tornado graph based
on the sensitivity analysis of Annex HS Xk
(input mean constant, Lv = 1.0…10 m).
81
4. CONCLUSIONS
When the number of observations n is less
than 8, all the described methods should
be treated with cautiousness. Because
calculated standard deviation and COV
are extremely uncertain at n<8,
COVunknown- and RIL -method should not
be used at all in these conditions. Not only
is there a risk of underestimating the
characteristic value Xk, but also a risk of
overestimation.
Frank et al. (2004, pp. 46-47)
recommend the usage of COVknown -
method if a priori information is available
due to the fact that n is usually low in
geotechnical problems. However, the
results indicate that the selected ‘known’
COV must be suitable for the studied soil
property and regional characteristics. As
such, the authors recommend further
investing in the research on COV in order
to provide reliable a priori information for
designers. Furthermore, research on soil
variability provides tools for RBD as well.
In RBD, distributions of the soil properties
are used instead of fixed value of Xk,
which enables qualitative estimation of the
reliability of the design (Phoon 2008, pp.
7-8).
Unlike other discussed methods, Annex
HS -method does not directly depend on
n. Results show that high values of Lv lead
to excessively high values of Xk which
hardly are ‘cautious means’ anymore.
Furthermore, the effect of Lv to the Xk is
the same or even greater than of factor a,
which the authors consider problematic.
Why would a larger zone of influence
increase the reliability more than the
number of results and the quality of the
testing (factor a)?
The method of Annex HS should be
modified so that Lv has less influence on
the derived value of Xk. Alternatively,
Annex HS could provide a maximum value
of Lv in order to ensure that factor a has
greater influence in all conditions.
To conclude, if n is low, the authors
recommend using COVknown -method.
Since COVknown -method yields extremely
conservative values if n<8, a feasible
option would be to promptly assume n>8
while using an appropriate COV.
Nonetheless, if n is low, the best option
would be to apply a priori knowledge on a
distribution based on typical values and
use Bayesian updating in order to acquire
the most reliable distribution for the soil
property for RBD (Phoon & Kulhawy
1999).
REFERENCES
EG11. (2015) "TC250/SC7/EG11:
Characterization. Final report (Oct
2015)" 13 p.
EN 1990. (2005) "Eurocode - Basis of
structural design" European
Committee for Standardization,
Brussels.
EN 1997-1. (2004) "Eurocode 7:
Geotechnical design - Part 1: General
rules" 168 p., European Committee for
Standardization, Brussels.
Fenton, G.A. & Griffiths, D.V. (2008) "Risk
assessment in geotechnical
82
engineering" 461 p., Wiley, New
Jersey.
Frank, R., Bauduin, C., Driscoll, M.,
Kavvadas, M., Krebs Ovesen, N., Orr,
T. & Schuppener, B. (2004)
"Designer's guide to EN 1997-1
Eurocode 7: Geotechnical design -
general rules" 216 p., Thomas Telford
Publishing, Bodmin.
Lacasse, S. & Nadim, F. (1996)
"Uncertainties in characterising soil
properties", Geotechnical Special
Publication, 58 I, 49-75.
Lee, I.K., White, W. & Ingles, O.G. (1983)
"Geotechnical Engineering" 508 p.,
Pitmans Books Limited, London.
Lehtonen, V., Meehan, C., Länsivaara, T.
& Mansikkamäki, J. (2015) "Full-scale
embankment failure test under
simulated train loading",
Géotechnique, Vol. 65:12, 961-974.
Mataic, I. (2016) "On structure and rate
dependence of Perniö clay", Doctoral
thesis, 350 p., Aalto University School
of Engineering, Espoo.
Müller, R. (2013) "Probabilistic stability
analysis of embankments founded on
clay", Doctoral thesis, 92 p., KTH
Royal Institute of Technology,
Stockholm.
Phoon, K. (2008) "Reliability-based design
in geotechnical engineering:
computations and applications" 530
p., CRC Press, New York.
Phoon, K. & Kulhawy, F.H. (1999)
"Characterization of geotechnical
variability", Canadian Geotechnical
Journal, Vol. 36:4, 612-624.
RIL 207-2009. (2009) "Geotekninen
suunnittelu. Eurokoodi, Geotechnical
design. Eurocode" 244 p., Hansaprint
Oy, Helsinki.
Schneider, H.R. (1999) "Determination of
characteristic soil properties"
Proceedings of the 12th European
Conference on Soil Mechanics and
Geotechnical Engineering,
Amsterdam, Netherlands, 7-10 June
1999, pp. 273-282, A.A Balkema,
Rotterdam.
 25th European Young Geotechnical Engineers Conference
21st – 24th of June 2016, Sibiu, Romania

Evaluation of ground deformation modulus definition by in situ

and laboratory test methods

Sergii Poklonskyi

Post graduate student in State Government “State Research Institute of Buildings

Constructions”, Kyiv city, Ukraine

ABSTRACT.
For today, the obtaining responsibility of baseline data for the soil base increases, which
characterizes their strength and deformability. Deformation modulus E is one of the deformability
parameters. There are various methods of determining E. This method has several disadvantages.
In oedometer soil test a ring of small size is used. It causes a number of factors affecting the test
results. The number of these factors reaches thirty according to Professor A.K. Larionov. The
greatest influence on the formation of deformations in the sample have zones of plastic
deformation at the contact with the upper and lower stamps of the device. According to our data
these zones at the top and bottom of the sample have limited distribution - 2 ... 3 mm.

Key words: deformation modulus, oedometer test, in situ test.

according to Professor A.K. Larionov [1].

INTRODUCTION
Today, the load on the buildings and
structures base is significantly increased.
It increases the responsibility of obtaining
the source data for soil foundation base
that characterize their strength and
deformability. The deformation modulus E
is one of the parameters of deformability.
There are various methods for
deformation modulus E determining. In
Ukraine soil deformation in most cases is
determined based on oedometer soil test
in laboratory because of their relative
simplicity and availability [3].
However, this method has several
disadvantages. In the oedometer soil test
a ring of small size is used. This raises a
number of factors affecting the test results.
The number of these factors reaches thirty
* presenting author
Special tests shows the greatest effect on
the formation of the deformation zone in
the sample have the plastic deformation at
the contact with the upper and lower
device stamp. These deformations are
accompanied by the formation of the
crumled structure and transition (buffer)
zones, as well as provide a smooth
transfer of pressure on the soil sample.
According to our researches, these zones
at the top and bottom of the sample have
limited distribution - 2 ... 3 mm.
It can be argued that the difference in E
values observed in the comparison of
oedometer and other methods, due to
these factors. As numerous tests shows
the greatest difference observed in the
results of in situ plate loading and
oedometer tests.
 This difference depends on the soil type
and condition can be 2 ... 10 times.
It is confirmed by the alluvial sands
tests on the experimental site. The
resulting difference in deformation
modules of sand obtained by oedometer
tests and in situ was 1.7 ... 2.3 times.
Addiction comparison of values EPLT
and Eoed from pressure for most types of
soil shows (Figure 1) that the oedometer
tests have greatest uncertainty in the
definition of the load range, although in
theory it is difficult to study. Very often, in
practice, as the characteristic values of the
examined deformation modulus as defined
by the oedometer curve in the pressure
range of 0.1 ... 0.2 MPa, thus meant that
when the ground occurrence depth
changes and pressure under the
foundation changes, the pressure range
for the definition of E should be updated.
Figure 1. Generalized comparative evaluation
of in situ deformation modulus (1) and
oedometer (2) test modulus.
Obviously, oedometer tests do not
account the possibility of the development
of plastic deformation in the actual basis,
and therefore exclude the bearing
capacity the pressure increasing p may
even lead to higher values of E.
On the other hand, it is known that
loading history can cause corresponding
changes in the deformations that cannot
be uniquely determined without regard to
the nature of the build-up pressure
underside foundation base. That’s why it’s
important to change the standard
procedure of soil tests of samples in the
oedometer. This methodology should be
consistent with the nature of the base load
in the real consequences.
84
1. IN SITU TEST.
Static load tests of the round stamp
type I area of 5000 cm2 carried out
according to the Ukrainian standards on
the central axis of the pits with a step of 10
m in the bottom surface of trenches, 24
points in two sectors to obtain the soil
deformation modulus. Calculation of soil
deformation modulus based on the results
of in situ test with load stages from 50 to
300-350 kPa.
According to in situ tests the graphs of
dependence of deformations from
pressure are built S = f (p). In the graph
with help of the averaging line the values
ǻp and ǻS are calculated.
The ground deformation modulus E,
MPa calculated for the linear portion of the
graph by the formula: E = (1-Ȟ2) * Cr * K1
* D * Ap / ǻS. The testing scheme by the
static load on the stamp is shown on
Fgure 2.
Figure 2. Soil testing scheme by the static
load on the stamp
Typical plots of the deformation
dependance from the pressure are shown
in Figure 3.
Figure 3. Characteristic curves of in situ soil
tests. a – sector I, b – sector II.
2. LABORATORY TESTS. a)
For the laboratory tests soil were
sampled at the points of in situ tests. The
selection and storage of soil samples for
laboratory tests conformed with Ukrainian
standards. Samples weighing up to 1.5 kg
transferred to the geotechnical laboratory.
Laboratory tests carried out in
accordance with normative requirements,
governing the procedure of laboratory
tests.
The carried out classification
parameters is 6 grain-size distribution of
filled soil. b)
Oedometer test (determination of
oedometer deformation modulus)
performed on the disturbed samples with
the set (as determined in the field) density
and moisture content in the odometer
loading up to 0.4 MPa at a natural
moisture (6 determinations).
Determination of shear strength
parameters (angle of internal friction and
cohesion) is also performed. The
compression and shear soils test results
are presented in Table 1. The Figure 4. a and b – typical oedometer
characteristic compression curves are curves
shown in Fig. 4

85
3. THE RESULT OF THE WORK. crumple zone in the odometer. When
The average values of the soil dry calculating the foundation drafting soil
density defined by the cutting ring method crumpling can be ignored, but in the
at predetermined points in the field of in oedometer test, due to the small sample
situ testing are: Pd = 1.79 g / cm3 Pd = size, crumple zone deformation
1.78 g / cm3 (sector II.) significantly underestimate the value of
deformation modulus and overstate the
High soil dry density submitted by fine
and dusty sand, due to the inclusion in drafting of the building and consequently
them of rubble up to 15-30%. increase the cost of construction.
Average values of in situ test 6. In Eurocode 7 [2] in order to avoid
deformation modulus are as follows: EPLT malfunctioning determineing the
deformation of the module differentiated
= 31.0 MPa (section I), EPLT = 28.0 MPa
(section II). definition accepted, and in odometer tests
are often used with some oedometer
The average oedometer soil curve unloading branch approximation.
deformation modulus values Eoed =18,6
MPa. LITERATURE
The correlation coefficient between the 1. Larionov A.K., Kislova L.V. The
compression deformation modulus and question of study in oedometer
modulus obtained by in situ test, which compression soil test errors. Buildings
depends on the diameter of the stamp and construction on the loess soils. Voronezh:
the depth of investigation is mk = 1,6. VISI. 1963.
The territory of the planned research, in 2. Eurocode 7 EN 1997-2:2007: (E) : -
general is flat, the absolute level of the Geotechnical design - Part 2: Ground
ground surface are ~ 113.55 m. investigation and testing – (together with
The geological structure includes United Kingdom National Application
technogenic soils (tIV), represented by fine Document), 1997.-196 s.
and dusty sand with gravel inclusion up to 3. Foundation analysis and design /
15-30%. During investigations Joseph E. Bowles. - 1996.-624 s.
groundwater levels are not met, it is Buildings on expansive soils.
located approximately at a depth of 3.0 -
3.5 m from the level of 113.50 m.
The average values soil dry density is
ȡd = 1,78-1,79 g / cm3.
4. CONCLUSIONS
1. Average values soil deformation
modulus, some of in situ tests at up EPLT =
29.5 MPa oedometer Eoed = 18.6 MPa.
2. The deformation of the soil sample is
uneven on the adjustment of the sample;
3. The crumpling of sample significantly
affects the deformation modulus value;
4. The coefficients mk correlation
established in Ukraine as an overall
average for the clay loam and sandy loam,
do not account the structural strength and
depend on the type, density and soil
conditions;
5. Crumple zone occurs also during in
situ plate test, it is commensurate with
86
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Geotechnical investigation for determination of causes for

occurred deformations of the Skopje Aqueduct

B. RUMENOVSKI1*, K. TASEVSKA2, B. JAKIMOVSKI3, T. DIMITRIEVSKI4

1
BSc.geotech.eng., GEING Krebs und Kiefer International and others, Skopje,
R.Macedonia
2
BSc.geotech.eng., GEING Krebs und Kiefer International and others, Skopje,
R.Macedonia
3
BSc.geotech.eng., GEING Krebs und Kiefer International and others, Skopje,
R.Macedonia
4
BSc.geotech.eng., GEING Krebs und Kiefer International and others, Skopje,
R.Macedonia

ABSTRACT
This paper deals with geotechnical investigation works for assessment of factors which led to
occurrence of deformations of Skopje Aqueduct – significant monument for Macedonian cultural
heritage.
The period of construction of the Aqueduct is still undetermined. According some data, the
monument was built in the 6th century in the time of the Byzantine emperor Justinian I, while
according to other data, it was built in the 15th century. It is presumed that it was still in use by the
end of the 19th century.
Based on the performed geotechnical investigations and geophysical surveys on the site,
emphasizing the deformed zone of the Aqueduct, as factors for occurrence of deformations were
noticed the relatively weak stress-strain properties of the foundation soil, the shallow foundation of
the structure, the permanently flooded site, local tectonic and the modifications in the spreading of
surface clayey layer, as possible agents for different damaging (sloping) of the Aqueduct pillars.

Keywords: Aqueduct, geotechnical investigations and geophysical surveys,

deformed zone

* presenting author

1. INTRODUCTION
The Aqueduct is located on the north of
the city centre between the Skopje
Fortress – Kale and the Skupi
archaeological site, in vicinity of the city of
Skopje, Republic of Macedonia. As a
linear building with an angle in the center
of the construction, the Aqueduct is a
distinctive landscape feature.
It is built over a small valley of the Serava
River, with the altitude close to 300
meters, as a part of water supply system
that carried water from Skopska Crna
Gora Mountains to Skopje Fortress.
The construction date of the Aqueduct is
still undetermined. According some data,
the monument was built in the 6th century
in the period of Roman or Byzantine
Empire i.e. the time of the Justinian I (527-
554 A.D.). But due to similarities in the
construction method between this
structure and Kurshumli An in Skopje it is
considered that the Aqueduct is built in the
15th century, period in which the city of
Skopje intensively started to grow along
with building of new Islamic structures –
Turkish baths and mosques. For
construction and functioning of such
facilities huge quantities of pure water
were required. It is presumed that it was
still in use by the end of the 19th century.
Figure 1. The oldest overview of the Aqueduct
in Skopje, drawing from year 1883-1884
(published by the English archaeologist Arthur
John Evans, 1885)
Overall, the Aqueduct have a double role -
construction with strictly defined function
88
into water supply system, and at the same
time with high architectural values.
1.1. The Aqueduct construction
generals
The Aqueduct is preserved in its entire
length which is 385,80 meters long, and it
has 54 main vaults and 42 so called
“relaxing” vaulted holes placed over the
pillars for reducing the structure weight.
The lower part of the Aqueduct consists of
53 square or rectangular pillars whit width
of 2,15 meters to 2,24 meters, and 2 side
ramps – northern and southern, with their
foundations.
Method of building is rich and colourful,
from different kinds of stone and brick.
The pillars from the oldest construction
phase are built of river stone in
characteristic rows of brick old format and
partly stone tiles and other fragments in
lime mortar. The facades are of mixed
masonry, stone and brick in lime mortar,
with archivolts and outside arches of brick.
The Aqueduct as original civil structure,
provides the opportunity to studying the
methods and techniques of building this
type of construction and used building
materials.
Few construction phases and repairs can
be noticed on the building. Most of those
originate from the end of 19th century and
the beginning of 20th century. After the
Second World War, three vaults were
destroyed by mines in a military exercise
to demonstrate the destruction of bridge
vaults. It is presumed that these vaults
were reconstructed in 1968 yr.
Figure 2. Aqueduct after the dynamites’
exercise
The last restoration of the Aqueduct was
carried out after the earthquake in Skopje
in 1963, repairing three arcs and two
pillars destroyed due to the earthquake
activity.
In the central part of the Aqueduct the
construction is not vertical, it is with an
inclination to the west which is up to 53
cm, and there is a very serious problem of
disturbance the stability and demolition
the section between pillars no. 9 and
pillars no. 15.
Also, a part of the foundations are
threatened by high groundwater level, so
in rainy periods form small swamp at the
lowest point along the construction.
Figure 3. Panoramic view of the Aqueduct in Skopje
2. GEOTECHNICAL
INVESTIGATIONS
In March 2014 yr. by the company "Geing
Krebs und Kiefer International" Ltd Skopje,
was performed field investigations and
laboratory testing on the location of the
Aqueduct. They are performed according
to methodology which is in accordance
with the technical norms wildly applied in
Macedonia and in foreign countries, the
Macedonian standards as well as the
instructions and recommendations from
the technical literature.
The purpose of performed investigations
and testing is determination the lithological
structure, with specifically defining the
physical - mechanical and stress - strain
properties of soil materials, for
determination the reason of occurred
deformations, as well as establishing the
appropriate activities recommended for
remediation of the object.
For more detailed presentation of the
lithological structure on the location were
applied the following investigation works:
x Engineering - geological and hydro -
geological site mapping;
89
Special threat to the monument is a risk of
earthquakes. The Skopje valley is a
seismically active zone. The estimated risk
indicates ɚ risk of earthquakes from 7 to 9
degrees by MCS (risk area of 9 MCS is
located at the bottom of the valley).
The aim of this paper is to present the
performed geotechnical investigations,
which would determine the causes of
occurred deformations of construction, as
well as to recommend appropriate
activities for its remediation.
x Performance of geophysical seismic
investigations;
x Performance of investigation boreholes;
x Standard penetration test – SPT test for
each borehole;
x Mapping of the investigation boreholes;
x Measuring and registering the
appearance and level of groundwater;
x Performance of trail pits;
x Sampling non-disturbed, semi-disturbed
and disturbed samples from the
investigated boreholes;
x Trail pits mapping;
x Cone penetration test – CPT test;
2.1. Engineering - geological and hydro
- geological site mapping
The terrain is composed of younger
proluvium deposed sediments, composed
of sands and gravels, silty gravel, covered
with clay (Q2prsk). The thickness of these
sediments is 5÷10 m;
Under the proluvium sediments there are
Pliocene sediments composed of sands
and gravels, silty clayey, marls,
sandstones and conglomerates to depth
of 60.00 m. Under these sediments there
are Miocene - Pliocene marl sediment
deeper than 1000 m.
The geological investigations on the
terrain are showing that there is fault with
direction NE-SW. This profile is not given
on the engineering-geological map for
Skopje, but is given on the engineering-
geological profiles VIII-VIII’ and XII-XII’ (in
the mentioned report from 1963-1964). It
is estimated that this fault cuts the location
on the damaged part of the aqueduct.
2.2. Performance of geophysical
seismic investigations;
The seismic investigations are performed
with shallow refraction and reflective
profiles, placed according the investigation
purposes.
The refraction profiling are performed with
4 profiles with total length of 325 m. The
profile RF-1 is performed with length of
125 m on distance of 1m (east) from the
pillars’ line of the aqueduct, continuously
along the complete length on the
investigated location. The profile RF-2 is
made at the beginning, RF-3 in the middle
(zone of modification of the pillars
inclination) and RF-4 at the end of the
investigated location. These profiles are
performed normally on the line of the
aqueduct pillars, each with length of 65 m.
The profiles are performed with geophone
distance on 5 m and stimulation of seismic
waves on 15 m.
For more detailed perception of data
modification from refraction investigations
it is performed refraction interpretation of
reflective profiles, given in the refraction
profiles RF-5 (located in parallel with RF-1
on distance of 5 km toward west) and RF-
6 located in parallel with RF-3 on distance
of 4 m toward north. The complete length
of these profiles is 190 m.
With the refraction profiles is reached
depth of 20-30 m.
The reflective profiling are performed in
order to provide comparison if the
interpretation of the geological structure
from refraction surveys with surveys with
90
higher resolution. These surveys are
performed with continuous following of the
litho-physical modifications in the ground
and noticing the local tectonic properties
of the location. The surveys provide more
realistic defining the data for physical –
mechanical and the geotechnical
assemble of the investigated area.
For registering seismic waves are used
vertical geophones of 14 Hz.
2.3. The interpretation of data from
refraction profiles
The interpretation of data from refraction
profiles is performed with hodochrones of
P and S waves, thus were determinate the
following parameters of registered elastic
environments:
x The values of Vp and Vs seismic
velocities;
x The physical-mechanical and elastic
dynamical properties, determinate
according the seismic Vp and Vs
velocities;
x The dynamical properties of materials
during earthquake;
The refraction survey determinates the
following contoured and average (in the
bracket) values of seismic waves Vp and
Vs of the foundation environments of the
aqueduct and its geological basis.
x High plastic clay, on depth of 1-3 m with
values of Vp=340-450 (400) m/s and
Vs=125-180 (150) m/s;
x Sands and gravels, silty clayey
materials on depth of 2-5 m and values
of Vp=400-550 (500) m/s and Vs=180-
250 (220) m/s;
x Sands, gravels, sandy clayey silt on
depth of 8-12 m with values of
Vp=910-1360 (1100) m/s and Vs=400-
570 (460) m/s;
x Sands, gravels and clayey silt on depth
of 8-23 m, with Vp=1750-2350 (2000)
m/s Vs=650-950 (800) m/s;
x Sandstones, marlstones and
conglomerates deeper than 23 m, with
Vp=2420-2760 (2600) m/s and
Vs=1000-1100 (1000) m/s.
The first three environments belong into
Quarter younger proluvial sediments
(Q2prsk), and the average 4 and 5 of
Pliocene sediments (Pl). We should stress
out that some of the materials in the
middle 3 belong into the surface weaken
Pliocene materials.
According the data of refraction profile RF-
1 and RF-5, the deeper depth of Quarter
proluvial sediments (around 12 m) are in
the beginning – the north part of the
location (the pillars S1, S2 and S3) near
the channel Serava. At the final – south
art of the location (S20 and S2) its depth is
around 6-8 m. Comparing the data from
the profiles RF-1 and RF-5 (on distance of
5 m), we can conclude that the depth of
the quarter sediments is reduced for
around 2 m in the west part, actually for
the same distance is increased in the east
part of the location.
This conclusion is confirmed with the
cross section profile RF-2 and RF-6. On
the profile RF-3 is obtained increasing of
the thickness at the Quarter sediments in
the west part, and on the profile RF-4 in
the east part of the location. The
appearance of RF-4 can be explained with
possible surface or tectonic damaging of
the surface Pliocene sediments.
According the values of seismic waves,
the first elastic environment – the surface
layer with depth of 2-3 m can be classified
as archeological cultural layer.
The interpretation of reflective profiles
determinate:
x Elastic litho-physical boundaries
between the surface clay layer and
proluvial sandy gravel material and
between the materials of proluval and
Pliocene sediments;
x The tectonic disturbing and dislocation
(faults) in geological environment of
the terrain;
From point of view the local tectonic here
are interpreted two tectonic dislocations,
cutting the investigated location of the
aqueduct. The first dislocation (P1) has
spreading NW-SE and cuts the aqueduct
in the zone of the third and the fourth
pillar. This dislocation is agent for the
inclination of the pillars (1-9) of the
aqueduct toward east in the first half of the
location. The second dislocation (P2) has
direction NE-SW, cutting the zone in the
tenth and eleventh pillar causing
inclination of the pillars (10-21) toward
west in the second part of the location.
The dislocation (P2) is confirmed with the
geological investigations after Skopje
earthquake in 1963-1964.
The interpreted local tectonic and the
modifications for spreading the surface
clay layer are the possible reason for the
variable damaging (hobbling) of the
aqueduct pillars. Due to this it should pay
attention to the modifications of the
physical – mechanical properties of the
clay during the different weather
conditions and the effects of clay swelling.

Table 6. Interval values of elastic dynamical parameters

Proluvium Proluvium Proluvium
Pliocene sediments
sediments sediments sediments
Parameter (Pl)
(Q2prsk) (Q2prsk) (Q2prsk)
H (m) 1-3 2-5 8-12 25-60
Vp (m/s) 340-450 400-550 910-1360 1750-2750
Vs (m/s) 125-180 180-250 400-570 650-1100
J (kN/m3) 15-16 17-18 19-20 21-23
Pdin 0.42-0.40 0.38-0.37 0.38-0.39 0.42-0.40
Edin (MPa) 68-148 165-315 855-1840 2570-7950
Gdin(MPa) 25-55 60-115 375-660 905-2840
Kdin(MPa) 142-247 230-405 1100-2790 5350-13250

91
2.4. Cone penetration test – CPT test; x Resistance of pipe attrition;
On the investigated location are performed x Resistance of attrition of cone and pipe;
tests of static penetration (CPT – test).
The test of static penetration is performed The classification of soil materials is
with penetrometer type Holland 10 t with performed according the
cone Type: Dutch Mantle Cone D=35.7 recommendations of Robertson et al, 1990
mm, A=10 cm2. The measurements are which is showed in following pictures and
performed in the following phases: tables.

x Cone resistance;
Zone
1 Sensitive, fine grained
2 Organic materials
3 Clay
Cone resistance Q (kN)

4 Silty clay to clay

5 Clayey silt to silty clay
6 Sandy silt to clayey silt
7 Silty sand to sandy silt
8 Sand to silty sand
9 Sand
10 Gravely sand to sand
11 Very stiff fine grained*
12 Sand to clayey sand*

Attrition ratio (%), Rf * Over consolidated or cemented

Table 7. Classification of soil materials

2
Depth [m] Mv [kN/m ] Classification Description
Robertson (1990) Robertson (1990)
CPT-1
0.25-2.00 4500 CI/MI Clay
Very compact, fine grained, over consolidated or
2.00-2.50 17700 ML
cemented materials
CPT-2
0.25-1.75 6500 CI/MI Clay
Very compact, fine grained, over consolidated or
1.75-2.50 12000 ML
cemented materials
CPT-3
0.50-3.00 4300 CI/MI Clay
Very compact, fine grained, over consolidated or
3.00-3.75 12000 ML
cemented materials
CPT-4
0.25-3.50 2618 H, CI/MI Clay, organic clay, humus
3.50-4.00 57750 CI/MI Clay
Very compact, fine grained, over consolidated or
4.00-8.50 16125 SFs
cemented materials
CPT-5

92
0.75-1.00 750 H Organic materials
1.00-2.50 8250 CI/MI Clay
Very compact, fine grained, over consolidated or
2.50-4.00 12375 ML
cemented materials
Very compact, fine grained, over consolidated or
4.00-4.50 37500 SFs
cemented materials
CPT-6
0.00-1.00 300 CI/MI Clay, organic clay
1.00-3.25 3000 CI/MI Clay
3.25-4.00 52500 SFs Sandy silt to clayey silt
CPT-7
0.25-1.00 1750 CI/MI Clay
Very compact, fine grained, over consolidated or
1.00-2.25 39300 SFs
cemented materials
CPT-8
0.25-0.75 1500 CI/MI Clay
Very compact, fine grained, over consolidated or
1.75-1.50 65000 SFs
cemented materials
2.5. Performance of investigation BH-1 to BH-10. The investigation drilling is
boreholes; performed geomechanical point of view,
On the investigated site are performed 10 with dry drilling that allow getting quality
(ten) investigation boreholes, each with core 100%.
10.00 meters depth. The investigation The data for the depth of the investigation
drilling is performed with two drilling boreholes, type, number and depth of
machines (GAK 300 and GDR 150) on each sampling, as well as the ground
previously defined locations marked from water appearance are given in table no. 8.

Table 8. Review of investigation boreholes

Borehole Final depth Appearance of Level of Sampling
[m] groundwater groundwater
Sampling
BH-2 10.0 / / 3
BH-3 10.0 6.30 / 3
BH-4 10.0 / 6.30 4
BH-5 10.0 / 8.30 4
BH-6 10.0 / 6.30 6
BH-7 10.0 8.30 / 2
BH-8 10.0 8.30 / 4
BH-9 10.0 / / 3
BH-10 10.0 / / 2

During the investigation drilling are the laboratory tests. The compression
preformed SPT tests, for quality estimation module for the incoherent materials is
of the foundation compactness. The tests determinate through the SPT (Standard
are used for orientation qualification of the penetration test) in table no. 9.
mechanical parameters of the materials
(ij, c and Mv) as comparative indicators of
Table no. 9 Review of performed SPT tests

93
 Compactness/c

Compressibility

internal friction
number of hits
Hits number
penetration

onsistency
Corrected
Borehole

modulus

Angle of
Depth

Cone
Mv
(m) (cm) N N’ N’’ Nkor Ndef / ij (°)
(kPa)
1.600÷1.904 30.4 29 20 / / 20 Solid 12500 /
Very
BH-1 3.800÷3.900 10 30 21 / 64 64 48300 43
compact
7.700÷8.004 30.4 21 15 / / 15 semisolid 10000 /
1.700÷2.004 30.4 30 21 / / 21 solid 13000 /
3.200÷3.330 13 30 21 / 49 49 compact 37800 40
BH-2
5.200÷5.504 30.4 14 10 / / 10 semisolid 7500 /
7.500÷7.804 30.4 17 12 / / 12 semisolid 8500 /
2.300÷2.450 15 30 21 / 43 43 compact 33600 39
medium
4.000÷4.210 21 30 21 / 30 30 24500 36
compact
BH-3
6.200÷6.504 30.4 13 9 / / 9 semisolid 7000 /
medium
8.500÷8.804 30.4 26 18 / / 18 16100 33
compact
1.600÷1.904 30.4 27 19 / / 19 solid 16800 /
2.600÷2.770 17 30 21 / 38 38 compact 30100 38
4.800÷5.104 30.4 11 8 / / 8 soft 6500 /
BH -4 medium
7.000÷7.230 23 30 21 / 28 28 23100 36
compact
Medium
8.000÷8.210 21 30 21 / 30 30 24500 36
compact
2.200÷2.504 30.4 30 21 / / 21 solid 13000 /
5.000÷5.280 28 30 21 / 23 23 solid 14000 /
BH -5
medium
7.800÷8.104 30.4 24 17 / / 17 15400 32
compact
1.800÷2.104 30.4 22 15 / / 15 semisolid 10000 /
BH 6 3.500÷3.804 30.4 24 17 / / 17 semisolid 11000 /
7.200÷7.504 30.4 20 14 / / 14 semisolid 9500 /
1.700÷2.004 30.4 30 21 / / 21 solid 13000 /
BH -7 medium
7.000÷7.240 24 30 21 / 27 27 22400 35
compact
BH -8 1.600÷1.880 28 30 21 / 23 23 solid 14000 /
1.500÷1.630 13 30 21 / 49 49 compact 37800 40
BH -9 medium
5.600÷5.840 24 30 21 / 27 27 22400 35
compact
1.500÷1.804 30.4 13 9 / / 9 semisolid 7000 /
BH-10
5.000÷5.304 30.4 15 11 / / 11 semisolid 8000 /
2.6. Physical-mechanical features of the laboratory testing, i.e. defining the physical
materials and its classification – mechanical properties of the present
In order to obtain more representative materials on the field and the location
physical – mechanical properties of the under the aqueduct.
terrain depending on depth, it were According the performed classification
performed certain type and scope of based on the site mapping and the
investigations boreholes from which were laboratory tests we can conclude that on
taken representative samples for further the mentioned location there are the

94
following types of materials with physical – As reasons for appearance the
mechanical properties as given bellow. deformations of the Aqueduct are the
following:
Surface layer of clays – CI/MI
x The local tectonic and the modification
This material is registered on the terrain in the clayey surface layer are the
surface along the complete profile on the possible reason for the variable
investigated location with maximal layer inclination to the west which is up to 53
thickness of 3.00 m. This layer is cm. Thus, it should pay attention to the
composed of medium plastic silty clays in modification of physical – mechanical
semi-solid to solid consistent condition properties of the clay during different
and covered with humus layer with weather conditions and to the effects of
maximal thickness of 1.00 m. The physical clay swelling;
– mechanical properties are given as it
follows: x The relatively low deformable
U M c Mv properties of the soil layer where is the
Mg/m3 [°] kPa kPa foundation (especially during presence
1.81 16.3÷16.6 13.8÷15.0 7400 of water);
Silts and silty sands – ML, CL/SFs, x The relatively shallow foundation of the
facility;
SFs/ML, SFs
x Constantly flowing terrain (creation of
These materials are registered under the “lakes” and swamp zones)
surface layer of clays. They are classified
as low plastic silts, clayey and silty sands 4. CONCLUSIONS
and sandy silts. The physical –
mechanical properties are given as it Although the Aqueduct in Skopje is not
follows: generally known and appreciated by the
U M c Mv public, the consensus is growing that the
Mg/m3 [°] kPa kPa Aqueduct and its surrounding area are
1.81 16.3÷16.6 13.8÷15.0 7400 unique and have an important historical,
architectural and esthetical value.
Silts and silty sands – GP, GFs
From geological point of view the terrain
In the lower layers on the location to the on the investigated location is composed
final depth there are materials like low of quarter sediments represented with
grained and silty sandy medium to well clayey silty fraction creating the upper
compact gravels. layers and silty sand and gravel, partially
U M c Mv clayey deposits in the lower layers. Under
Mg/m3 [°] kPa kPa the Quarter sediments there are Pliocene
1.8 16.3÷16.6 13.8÷15.0 7400 sediments composed of sands and
gravels, silty clayey, marls, sandstones
3. CAUSES FOR APPEREANCE and conglomerates to depth of 60 meters
THE DEFORMATIONS OF THE and under that there are Miocene
FACILITY Pliocene marl sediments on depth of 1000
Based on the performed geotechnical and meters;
geo-physical investigations on the From hydro-geologic point of view the
location, with special review of the investigated site belongs into the group of
deformed zone of the Aqueduct we can low to well yield terrains where is
conclude the following: developed compact type of aquifers in
The aqueduct is founded on foundations sediments characterized with inter
with average width of 2 x 2 meters in the granular porosity. Considering the
layer of silty clay with medium plasticity, geological structure of the field, especially
and foundation depth of 0.70 meters to the upper parts, we should stress out on
1.30 meters from terrain elevation. the location and the wider region is

95
present contemporary geologic process of rehabilitation of the Aqueduct in Skopje
permanently washed terrain (creation of and its environment”.
“lakes” and swamp zones); Master Urban Plan Skopje, Institute of
From geo-tectonic point of view the town planning and architecture –
investigated location belongs into area Skopje; Polservice Consultants,
with maximal intensity of VIII and IXº Warsaw; Doxiadis Associates –
according MCS. The strongest earthquake Consultants on Development and
is noticed in 1963 with intensity of IXº Ekstics, (1965)
according MCS in Skopje epicentre area, Lenche Jovanova, “Skupi – Colonia Flavia
followed with aftershocks with own local Scupinorum”, Skopje (2008)
epicentre on depth of 5 km; Kosta Balabanov, Antonie Nikolovski,
On the location are performed 10 (ten) Dimitar Kjornakov, “Monuments of
investigation boreholes each with depth of culture in Macedonia”, Skopje (1980)
10.00 meters, 8 (eight) tests of static
penetration SPT-tests, 8 (eight) trial pits
near the most damaged pillars and
geophysical seismic survey with total
length of 325.00 meters, refraction
surveys and 180.00 meters reflective
surveys;
The location has relatively heterogeneous
composition concerning the classification
and physical – mechanical properties of
the present materials. The surface layer is
composed of silty sandy medium plastic
clays covered with humus and clayey
humus cover. The maximal spreading
depth of the layer is 3 m. Under this layer
to the final investigation depth there are
sandy silts, silty sands and gravels, as
well as low grained gravels in the zone
around river Serava;

ACKNOWLEDGEMENTS
The authors of this paper would like to
thank the company "Geing Krebs und
Kiefer International" Ltd Skopje, and to the
Ministry of Culture, Skopje, for the support
while making the researches for the
Aqueduct and the opportunity to
participate on this Conference.

REFERENCES
Geing Krebs und Kiefer International Ltd
Skopje (2014) “Elaborate for
geotechnical investigations and
laboratory testing on the location of the
Aqueduct in Skopje”.
Ministry of Culture, Skopje (2011)
“Strategic plan for preservation and

96
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Session 2

Researches in geotechnical engineering

98
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Settlements of Foundations in Rock Fill due to Cyclic Loading

and Creep Effects

Madeleine BRANDT*1, Terje STRØMME1

1
Multiconsult ASA, Geotechnics Industry, Oil & Gas, NORWAY

ABSTRACT

Many of the installations and structures at industrial onshore plants are more sensitive than other
structures to deformations, differential settlements and rotations of the foundations. The
foundations are often supported by a compacted rock-fill. The geotechnical design must therefore
determine the displacements and rotations of the foundations on rock-fill under static and cyclic
loads (such as loads from machinery, cranes and traffic) and ensure that the displacements are
within the limits required for the plant functionality.
This paper presents a method for estimating foundation rotations and settlements due to
volumetric and shear strains in the rock-fill down to bedrock. The rock-fill strains include static
strains, accumulated cyclic strains and creep strains, developed with time. The theoretical basis of
the calculation methods are presented, together with a step-by-step procedure of the analysis. The
paper aims to focus on the practical implementation of the procedure in a real industrial project.

Keywords: Rock-fill, settlements, shallow foundations

industrial project consisted of a large

1. INTRODUCTION
This article presents a method of
estimating rotations and settlements for
rigid rectangular foundations on rock-fill
over bedrock. A step-by-step procedure of
the complete analysis is presented
together with the theoretical basis of the
implemented calculation methods. An
example analysis is presented at the end
of the paper.
The calculation procedure was developed
in connection with an industrial project
with strict requirements to foundation
rotations and settlements because of
deformation sensitive installations. The
* presenting author
factory hall with dimensions of
approximately 80 x 200 metres with a
great number of foundations supporting
the main construction and various
installations. The strict requirements were
connected to differential settlements
between certain foundations supporting
the same installations. Knowledge of the
ground conditions were limited, but old site
investigations indicated that there could
be expected to be up to approximately 2.5
meters of from bottom of foundation level
to bedrock. The fill material was assumed
to be well-graded and compacted crushed
rock.
2. STEP-BY-STEP PROCEDURE
The procedure for analysis of settlements
and rotations of rectangular foundations
on rock-fill consists of the following steps:
1. Assess the relative stiffness between
the foundation and the rock-fill, Kr
(e.g. Selvadurai, 1979).
2. Select the method used to calculate
deformations. If the foundation is stiff
relative to the rock-fill it is proposed to
use the Multiconsult in-house program
SPLATE (Athanasiu, 2004). If the
foundation is flexible relative to the
rock-fill it is proposed to use the
Multiconsult in-house program
FPLATES (Athanasiu, 1994).
3. Calculate the deformations from
permanent loads.
4. Calculate deformations due to cyclic
loads (crane passing, traffic etc.).
5. Calculate accumulated cyclic
deformations from cyclic loads (e.g.
Wichtmann et al, 2010).
6. Calculate creep deformations from
permanent loads (e.g. Athanasiu et
al., 2005)
7. Calculate the total deformation as the
sum of contributions from permanent
loads, cyclic loads, accumulated cyclic
loading and creep.
3. TEORETICAL BASIS
3.1. Relative foundation stiffness
The relative stiffness of a rectangular
foundation can be estimated using the
following equations suggested by
Selvadurai (1979):
(Eq. 1)
In which Ȟc and Ȟs are Poisson’s ratios for
concrete and rock-fill, respectively. Ec and
Es are the elasticity moduli for concrete
and rock-fill, respectively. L and B are the
dimensions of the rectangular plate (L>B)
and t is the thickness of the plate. If Kr is
less than:
(Eq. 2)
100
the foundation is stiff relative to the
support (rock-fill).
3.2. Calculation of strains in rock-fill
The estimation of rock-fill strains is
performed using the Multiconsult in-house
program SPLATE or FPLATES,
depending on the stiffness of the
foundation relative to the rock-fill. This
article focuses on analysis of rigid
foundations that are relatively stiff
compared to the rock-fill, thus the program
SPLATE is used. In SPLATE, the plate is
assumed infinitely rigid, i.e. the settled
position of the plate is a plane defined by
three displacements (vertical displacement
and two rotations). The ground is
modelled as an elasto-plastic half-space.
The contact stresses at failure are
calculated based on shear strength
parameters of the soil. The contact
stresses between the plate and the soil
are limited to the failure stresses and the
excess stresses are redistributed. The
input data for SPLATE consists of vertical
load and eccentricities about the two axes;
x and y in the plate plane. The plate area
is divided into a number of discrete
elements. The results consists of contact
stresses and settlements under each
element. The loads are specified about
the origin of the coordinate system, which
is selected at the corner of the foundation
as seen in Figure 1. The vertical load, Fz,
acts with the eccentricities, ex (about y-
axis) and ey (about x-axis). The input
eccentricities in SPLATE, ex and ey, are
defined with respect to the corner of the
foundation. The vertical spring stiffness of
the foundations are dependent on the
elastic properties of the rock-fill, the
dimensions of the foundation and on the
depth H from the bottom of the foundation
plate to the bedrock. As the program
SPLATE assumes an elasto-plastic half-
space under the foundation, the elasticity
modulus entered in the input file of
SPLATE must be corrected for the effect
of the depth down to bedrock by a factor,
fH:
(Eq. 3)
SPLATE calculates both static settlement
at the corner of foundation and the
rotations. The settlement at the center of
foundation is also printed in the results
file.
Figure 1. Input data for geometry and loads
3.3. Settlements and rotations due to
permanent loads
In the assessment of deformations from
permanent loads, all permanent loads are
summarized and the resultant force Fpz,
the resultant moments Mpx and Mpy are
used in the calculations. The eccentricities
relative to the center of the foundation,
epxc and epyc and the input eccentricities in
SPLATE are calculated.
(Eq. 4)
(Eq. 5)
(Eq. 6)
(Eq. 7)
The resultant vertical load and
eccentricities are used as input in SPLATE
for calculating deformations. The result
from this step is the permanent settlement
įpz and rotations șpx and șpy.
3.4. Settlements and rotations due to
cyclic and permanent loading
In the assessment of deformation from
cyclic loading (crane passing, traffic loads
etc.) all cyclic loads are summarized with
the permanent loads. The resultant cyclic
force Fcz and the resultant cyclic moments
Mcx and Mcy are added to the resultant
permanent loads:
(Eq. 8)
101
(Eq. 9)
(Eq. 10)
Calculate the eccentricities relative to the
centre of the foundation, etotcycxc and
etotcycyc for permanent and cyclic loading
and the input eccentricities in SPLATE:
(Eq. 11)
(Eq. 12)
(Eq. 13)
(Eq. 14)
The result from this step is the total cyclic
settlement įtotcycz and rotations ștotcycx and
ștotcycx. These results are not used directly,
only to calculate the settlement from the
first cycle loading which is described in
section 3.5.
3.5. Accumulated cyclic settlements
and rotations
Accumulated cyclic settlements and
rotations are additional deformations
depending on the number of cycles of the
cyclic loading, N. The calculation of
accumulated cyclic settlement based on
the high cyclic accumulation (HCA) model
proposed by Wichtmann et al (2010). The
accumulated cyclic settlement and
rotations are calculated as:
(Eq. 15)
(Eq. 16)
In which Gcycl1 and șcycl1 are settlement and
rotations from first cycle of loading,
respectively. fN is the accumulation
function, depending on the number of
cycles, N:
(Eq. 17)
CN1, CN2 and CN3 are material parameters
depending on the rock-fill compaction
energy, grain size distribution curve and
on the coefficient of uniformity, Cu=d60/d10.
For well-graded gravel (0.05 mm < d < 20
mm) the values of CN1=5.2E-04, CN2=0.03
and CN3=1.3E-05 can be used as
conservative estimates to calculate (Eq. 23)
accumulated cyclic settlements.
Settlement and rotation from first cycle of Conservatively, it could be assumed that
loading (Gcycl1 and șcycl1) are extra the unloading is infinitely rigid and that the
deformations occurring during the first first cycle displacements are equal to total
cycle of any cyclic load. The first cycle cyclic minus permanent settlements:
deformations can be calculated using
(Eq. 24)
SPLATE (or FPLATES) for the same
eccentricities as in total cyclic load 3.6. Settlements and rotations due to
analysis (etotcycx and etotcycy) but for a load creep
that is very small (i.e. Creep settlements of foundations can be
Ftotcyczo=Ftotcycz/10000). The result will be a calculated based on the model presented
settlement for the very small load which by Athanasiu et. al. (2005):
provides us with the initial
unloading/reloading stiffness of the
foundation for total cyclic load (Eq. 25)
configuration (eccentricities). The
Gc is the creep settlement in linear phase
permanent settlement after the first cycle
(linear increase of settlement with
Gcycl1 is calculated as the total cyclic
logarithm of time), E is creep parameter for
settlement įtotcycz minus the permanent
rock-fill material, Aı is the area of effective
settlement įpz and minus the elastic
vertical stress diagram under the
(unloading) part (see Figure 2). (Eq. 18) to
foundation, and sf is a bedrock slope
(Eq. 20) is used for calculation of Gcycl1. factor, see Figure 3. For non-sloping
bedrock surface under the foundation, sf =
(Eq. 18) 1, which is assumed in this paper. t is
time elapsed from installation of
foundation to the construction life time and
(Eq. 19)
t0 is the reference time (adjusted so that
the settlements in the diffusion phase are
(Eq. 20) included). Creep parameter choice is
based on Multiconsult experience with
rock-fills (Athanasiu et. al., 2005), see
Table 1. The creep rotations are
calculated as:
(Eq. 26)

Table 1. Rock-fill creep settlement parameters

(Athanasiu et. al., 2005).
Parameter Explanation Recommended
value
Figure 2. Settlement from first cycle of cyclic
loading. t0 Reference to=0.004 years *
time
In (Eq. 18) to (Eq. 20) the paramenters
ȕ Creep 12E-06 m2/kN **
KGtotcyco, KTtotcycxo and KTtotcycyo are the initial parameter
vertical stiffness calculated by
SPLATE/FPLATES in the small load run: *Adjusted so that the settlements in the diffusion
phase is included
**In Athanasiu et. al. (2005) a range of ȕ is
(Eq. 21) proposed (6E-6 to 25E-6 m2/kN). A best estimate
value of 12E-06 m2/kN is used in this paper.

(Eq. 22)

102
 Figure 4. Extract of plan of foundations
supporting main construction and crane rail

Figure 3. Vertical effective stress diagram

under foundation
3.7. Resulting settlements and rotations
Total settlements are calculated as the
sum of settlements from permanent loads,
settlements from cyclic loading and creep
settlements:
įtot = įpz + įacc + įc
Figure 5. Simplified section of factory hall with
= įpz + (1+fN)įcyc1 + įc (Eq. 27) foundations supporting main construction and
Similarly the rotations are calculated: crane rail.
ștot = șpz + șacc + șc (Eq. 28)

4. EXAMPLE ANALYSIS
In this example analysis, the differential
settlements and rotations between two
foundations (A and B) supporting the
same crane rail will be assessed (see
Figure 6. Geometry of foundation A
Figure 4 and Figure 5). It is assumed that
foundation B is founded to bedrock while Table 2. Loads at center of bottom of
foundation
foundation A is founded on a 2.5-meter
thick rock-fill layer above bedrock. The Load Fz (kN) Mx (kNm) My (kNm)
requirements to differential settlements Permanent -4880 -6670 1
between foundation A and foundation B is
Cyclic -1740 2540 2.9
set by the structural discipline to 20 mm.
Total -6620 -4130 3.9
4.1. Geometry and loads
Figure 6 show the geometry of foundation 4.2. Calculations
A with dimensions 3.8 x 8.0 x 1.5 meters. The first step is to check if foundation A is
Loads in the centre under the foundation relatively stiff compared to the rock-fill, as
are summarized in Table 2. The described in section 3.1: Kr § 0.51 < Krlim =
eccentricities relative to the centre of the 5.51. Since Kr is less than Krlim the
foundation and the input eccentricities in foundation plate is relatively stiff, and we
SPLATE are calculated as described in can proceed using SPLATE for calculation
sections 3.3 and 3.4. The cyclic loading of deformations. Settlements and rotations
consist of a crane load theoretically are calculated for permanent loads and
passing the foundation 35 times per day. total cyclic loads (cyclic + permanent) as
The lifetime of the factory is 50 years. The described in section 3.3 and 3.4. The
total number of cycles during the lifetime is accumulated cyclic settlement is
approximately N = 639 000. calculated from the first cycle settlement
103
which is conservatively calculated Table 3. Resulting settlements (in mm) and
assuming that the unloading is infinitely rotations (in radians)
rigid, and that the first cycle displacements Permanent įpz= 8.0 șox=1E-3 șpy= 4E-7
are equal to total cyclic settlements minus
permanent settlements as described in Weigth of įst= 2.7 șst= 0 șst = 0
foundation
section 3.5. Elastic deformation from
weight of the foundation is calculated and Perm+cycl įtotcycz= ștotcycx= ștotcycy=2E-3
subtracted, since this settlement 1.1 6E-4
contribution is assumed to be completed First cycle įcyc1= 2.9 șcyc1x= șcyc1y=1E-6
before the crane rail is installed, and thus -4E-4
can be adjusted for. Creep settlements Acc. cyclic įacc= 2.9 șaccx= șaccy=1E-6
and rotations are calculated for the -4E-4
permanent loads, using a creep parameter Creep įc= 19.4 șcx=3E-3 șcy=1E-6
ȕ=12E-06 m2/kN and reference time
t0=0.004 as proposed in section 3.6.
Creep settlements are calculated for a CONCLUSIONS
lifetime of 50 years. This article has presented a method for
4.3. Result estimating foundation rotations and
settlements due to volumetric and shear
Deformation results from all steps are strains in a rock-fill down to bedrock. The
summarized in Table 3. As seen from the method considers a realistic
results, foundation A can experience a representation of the rock-fill, and provides
total settlement of about 2.8 cm: a conservative and efficient way of
įtot = įpz + įacc + įc - įst = 27.6 mm analysing a large number of foundations
Since foundation B is assumed founded to without the need for using advanced, time-
bedrock, the differential settlements is consuming computer programs.
equal to the total settlement for foundation
A. Thus, the differential settlements ACKNOWLEDGEMENTS
exceed the requirements set by the The authors would like to express their
structural discipline (20 mm). As seen gratefulness towards Dr. Corneliu
from Table 3, creep settlements contribute Athanasiu and Arne Schram Simonsen for
the most to the total deformation. The valuable academic support.
creep calculation is however very
parameter sensitive, especially with REFERENCES
regards to the creep parameter ȕ. This is
Athanasiu, C. (2004) “SPLATE user manual”.
due to the parameter being based on Multiconsult report no. 39109-1.
empirical deformation data from large and
Athanasiu, C. (1994) “FPLATES user
high rock-fills and even rock-fill dam
manual”. Multiconsult report no. 39108-1.
constructions. The creep settlement
Wichtmann, T., Rondón, H.A., Niemunis, A.,
calculation is therefore considered
Triantafyllidis, Th. & Lizcano, A.(2010)
conservative. “Prediction of permanent deformations in
The differential settlements can be pavements using a high-cycle
reduced by limiting the difference in depth accumulation model”. Journal of
to bedrock under the two foundation Geotechnical and Geoenvironmental
types. The foundations should be Engineering, Vol.136, No.5, 2010
instrumented, enabling surveillance of Athanasiu, C., S. Simonsen, A., Søreide, O.K.
deformations during operation of the plant. & Tistel, J. (2005) “Elastic and Creep
This way, measures can be taken when Settlements of Rock-Fills”. ISSMFE
settlements and rotations are approaching Conference, Osaka, Japan, 2005.
the specified operational limits of the Selvadurai, A. (1979) “Elastic analysis of soil-
installations. foundation interaction”, Elsevier,
Amsterdam

104
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Discussion on selected results obtained for fine grained soil

treated with fly ash

Karolina KNAPIK1*
1
The Silesian University of Technology, Department of Geotechnics and Roads, Gliwice,
POLAND

ABSTRACT
The paper presents part of the results obtained during research carried out in order to
complement the knowledge about phenomena occurring in a mix of fine grained soil and fly ash
from fluidized bed combustion. Mentioned phenomena include chemical reactions ongoing in fly
ash and modified behaviour of fine grained soil treated with fly ash. Changes in physical properties,
physical-chemical properties and mechanical behaviour are presented in this paper based on the
example of soil–fly ash mix prepared with the addition of 40% of fly ash.

Keywords: Fly ash, Speswhite kaolin, Mechanical behaviour, Physical behaviour

properties of fly ash are not permanent as

1. INTRODUCTION
Various areas of human activity are
accompanied by generation of wastes.
Effective wastes management helps to
reduce the scale of landfills and creates
an opportunity to reduce costs of
investments. Complete recognition of
materials properties allows for their safe
and optimal use in economy. One of the
wastes showing great potential in the
construction sector is fly ash, created as a
by-product during production of electric
energy in power plants working based on
coal combustion. Poland is one of the
countries in which fly ash is produced
annually in large quantities and is
commonly used, for example as
a component of cement or concrete, in
roads construction and geoengineering.
However, it should be highlighted that
* presenting author
they are a resultant of boiler type,
combustion parameters and the
characteristic of burned materials. Based
on the literature review the approximated
characteristics of the fly ash produced in
conventional pulverized boilers or boilers
using fluidal bed combustion can be
determined, which enable materials pre-
selection for a specific application.
Technology of Fluidized Bed Combustion
(FBC) gained popularity as it gives an
opportunity for efficient combustion of
fuels with low calorific value. Properties of
FBC fly ash were described by many
authors, among others by Sulovský
(2002), Giergiczny (2006), Brandt et al.
(2010) and Zapotoczna-Sytek et al.
(2013). Coal combustion in fluidized bed
takes place in relatively low temperature
(around 850qC), which affects the form of
constituents forming fly ash. Clay minerals
contained in carbon rock undergo the
partial or complete dehydroxylation, which
makes that the material gains pozzolanic
properties. Moreover, ground sorbent
(limestone or dolomite) is added into the
bed in order to bind sulphur compounds,
which results in the presence of anhydrite
CaSO4 and calcium compounds in fly ash.
Part of calcium can be found in the form of
free calcium oxide influencing binding
properties of this waste material. FBC fly
ash comes up against difficulties in
application as a component of concrete or
cement, but its valuable binding properties
can be used in the field of geotechnical
engineering. Paper presents selected
results of a study carried out in order to
investigate the properties of fine grained
soil treated with FBC fly ash, entirely
presented in the PhD dissertation of the
author of this work (Knapik 2016).
2. METHODS AND PROCEDURES
Depending on the test, mixes of soil and
fly ash were stored for 0 day (not longer
than 21 hours), 7 days or 28 days in
closed containers, oedometer rings or
molds. Attention was paid in order to
reduce the contact of mix with
atmospheric air and to limit the variations
of ambient temperatures.
In order to test grain size distribution of
a mix, samples were stored for specified
curing time. Then, they were mixed with
solution of hexametaphosphate in
mechanical mixer for 20 minutes and
sieved. Equivalent grains diameter was
calculated according to Formula (1),
provided by DoáĪyk and Szypcio (2014).
KH r
d 0,05462 (1)
( U s  1)t
where:
K – dynamic viscosity of solution (mm)
Hr – hydrometer effective depth (mm)
Us – specific density (Mg/m3)
t – time (s)
106
The pH value of mixes cured in closed
containers was tested according to the
standard ASTM–D 4972–01 1998 with the
use of handheld pH meter Hanna
Instruments HIí991001.
For the purpose of sedimentation tests,
prepared samples were stored for 16
hours in small closed containers. Then,
they were mixed with the use of laboratory
mixer for 10 minutes with additional
distillate water, poured into the measuring
cylinder and hand-mixed by rotations.
During the whole procedure a special
attention was put in order to avoid any
material loss. Height of suspension was
measured in time intervals selected to
obtained well defined sedimentation
curve.
In order to characterize materials in terms
of mechanical behaviour samples were
tested in oedometer according to ASTM-
D2435-96 and in direct shear strength
apparatus according to PKN-CEN ISO/TS
17892-10:2004. Oedometer rings were
filled just after preparation of a mix and
material was cured in this way for
specified time. Samples tested in direct
shear strength apparatus were prepared
in rectangle metal molds, protected with
plastic foil and greased.
Swelling was measured in oedometer.
Samples were prepared in saturated state
in oedometer ring and the test was started
immediately. During the test samples were
constantly soaked (oedometer cell was
filled with distillate water within measuring
time). Upper surface of samples was
loaded with pressure 0.4 kPa, generated
by loading cup.
3. MATERIALS
Two types of materials were used in
experimental work presented in this paper:
Speswhite kaolin and selected FBC fly
ash.
Speswhite is a highly refined kaolin
obtained from deposits in Cornwall,
distributed by Imerys Materials. The
research based on well-known soil
allowed to reduce the scale of
investigation and to avoid the risk of
variability, typical for natural soils.
This fine grained soil is characterized by
grain size distribution presented in the
Figure 1.
Figure 1. Grain size distribution of Speswhite
kaolin
The pH value of tested sample was 4.2.
X-Ray analysis showed that crystalline
phases present in Speswhite kaolin are
kaolinite, muscovite and quartz (Vitale et
al. 2015). Kaolinite clay belongs to the
group of 1:1 clay minerals in which silica
sheet and alumina sheet are held by
hydrogen bonding between them. Soil
particle can be formed by many platelets.
As described by Mitchell and Soga (2005)
bonding between successive layers is
sufficiently strong that there is no
interlayer swelling in the presence of
water. Depending on water chemistry (the
pH value, ions type, valence and
concentration) kaolinite layers can
deflocculate or flocculate, which influence
soil properties.
Speswhite kaolin was investigated by
Vitale et al. (2015) in terms of pozzolanic
activity. Very low rate of hydrates
formation was confirmed for first four
weeks of curing time, as presented in the
Figure 2.
Fly ash used in experimental study was
obtained in power plant placed in Silesia
region of Poland. Combustion of coal and
coal slime took place in fluidized bed
boiler. The amount of main components of
fly ash are listed below:
x SiO2 40÷42%
x Al2O3 20%
x Fe2O3 6÷8%
107
x CaO 17÷19% (in which CaOfree 5%)
x SO3 4÷6%
Figure 2. Results of thermal analysis obtained
for Speswhite kaolin treated with lime
(Vitale et al. 2015)
Loss on ignition measured for fly ash was
2% and the pH value was equal to 12.6.
Grain size distribution is presented in the
Figure 3.
Figure 3. Grain size distribution of fly ash
Mass of fly ash in the mix FA40 was
calculated with respect to mass of kaolin –
40 g of fly ash for 100g of soil. Samples
were prepared at two different initial water
content, namely 50% and 100%. Initial
water content of mix, as well as the
samples prepared from pure Speswhite
kaolin was equal to 50% (samples signed
with symbol “w50”) or 100% (samples
signed with symbol “w100”).
4. RESULTS
Samples of the mix was tested in order to
investigate the changes in the pH value
over the curing time. The results are
presented in the Figure 4. It can be
noticed, that initially highly alkaline
environment is changing. A rapid drop in
the pH value within first week of curing
time is followed by progressive but more
gentle decrease within the next days.
Characteristic of the pH value over the
time is not significantly different for the
both tested mixes.
Figure 4. The pH values measured in curing
time for the mixes FA40w50 and FA40w100
Sedimentation curves obtained for the mix
FA40 and for the sample of kaolin are
shown in the Figure 5. The curves are
characterized by two different trends. It
can be observed, that in case of the mix
FA40w100 the time before settlement
starts is shorter and the rate of
sedimentation is higher in comparison with
the sample of kaolin. Additionally, the
trend of the curve obtained for the mix
FA40w100 in the time range of 10÷1000
min is divided for two parts. Initially high
slope is lower in second stage. It can be
assumed that this shape is a result of
following stages: sedimentation and self-
weight consolidation. In case of the
sample of kaolin the slope of the curve is
approximately stable.
Figure 5. Sedimentation curves obtained for
the sample of kaolin and the mix FA40w100
The grain size distribution of the mix FA40
was tested in three different curing times –
108
0, 7 and 28 days. The results are
presented in the Figure 6. The changes in
grain size distribution are most evident
within the first week of curing time.
Figure 6. Grain size distribution obtained for
the mix FA40w100 tested after 0, 7 and 28
days of curing time
The oedometer tests were performed for
the samples with 50% and 100% of initial
water content. The results are
summarized in the Figures 7÷9.
Figure 7. Consolidation curves obtained for
kaolin and the mix FA40 prepared at initial
water content 50% and 100%, tested at 0 day
of curing time
In the Figure 7 it can be observed, that the
consolidation curves obtained for samples
tested at 0 day of curing time, presented
as a plot of log(1+e) to logV, converge at
high vertical stress level (2450 kPa).
Oedometer tests performed after 7 days
(Figure 8) evidenced, that the vertical
stress used during the test was not
sufficient to obtain the crossing point. In
the Figure 9 it can be observed, that at
vertical stress equal to 2450 kPa, the void
ratio of both samples reached almost the stress is higher for the sample tested after
same value. 7 days, however, it is also somewhat
higher than volume change obtained for
the sample tested after 28 days. The last
observation could be connected with the
differences in the value of void ratio (2.6
for the sample tested after 7 days and 2.7
for the sample tested after 28 days).

Figure 8. Consolidation curves obtained for

the mix FA40 prepared at initial water content
50% and 100%, tested after 7 days Figure 10. Volume changes under vertical
of curing time load obtained for kaolin and the mix FA40
prepared at initial water content 50% tested at
0, 7 and 28 days of curing time

Figure 11. Volume changes under vertical

load obtained for kaolin and the mix FA40
prepared at initial water content 100% tested
Figure 9. Consolidation curves obtained for
at 0, 7 and 28 days of curing time
the mix FA40 prepared at initial water content
50% and 100%, tested after 28 days For both mixes, prepared at initial water
of curing time content 50% and 100%, an successive
In order to facilitate observations of increase of the yield stress in curing time
changes in compressibility ongoing in can be observed, as presented in the
curing time, the results are compiled in Figure 12.
a different way in the Figures 10 and 11.
The curves are grouped with respect to
initial water content. The volume change
at the highest vertical stress is decreasing
with curing time for the mix prepared at
initial water content 50%. In case of mix
prepared at 100% of water content, the
volume change at the highest vertical Figure 12. Calculated yield stress
109
Direct shear strength tests were
performed for kaolin and the mix FA40
prepared at initial water content 50%. The
results are presented in the Figures
13÷15. The mix of kaolin and fly ash is
characterized by somewhat higher shear
strength at 0 day of curing time in
comparison to the sample of not treated
soil. After 28 days of curing time, the
strength of the mix is increasing mainly
due to high value of cohesion.
Figure 13. Determination of strength
parameters for kaolin
Figure 14. Determination of strength
parameters for the mix FA40 tested at 0 day
of curing time
Figure 15. Determination of strength
parameters for the mix FA40 tested after
28 days of curing time – peak and residual
conditions
110
The measurements of volume changes
during curing in contact with water are
plotted in the Figure 16.
Figure 16. Volume changes measured for
kaolin, fly ash and the mix FA40
It was observed, that the sample of kaolin
settled under pressure 0.4 kPa. Sample of
fly ash shown swelling behaviour. The
volume changes lasting 12 days ended
with the final volume change of the sample
equal to 2.25%. Volume changes
measured for the mix FA40 significantly
exceeded the values measured for pure
components of the mix. The void ratio
increased from the initial value equal to
1.32 up to the value of 1.71 and the water
content increased up to the value 71%.
5. DISCUSSION
Presented results evidence, that addition
of fly ash into the fine grained soil,
represented by Speswhite kaolin, causes
changes in physical properties, physical-
chemical properties and mechanical
behaviour. Observed effects are time-
dependent.
Highly alkaline environment is necessary
to provide dissolution of silica and alumina
(Keller 1964). Fly ash, used in quantity
40%, changes the pH value of pore water
from acid to highly alkaline, which is
favourable in terms of formation of
hydrates such as calcium silicate
hydrates, calcium aluminate hydrates and
ettringite. Decrease in the pH value of the
mix FA40 observed during the first week
of curing time is a result of ongoing
chemical reactions – formation of hydrates
and possibly some carbonation (the
contact with atmospheric air was limited
but some air was trapped in container
together with the sample).
Sedimentation tests evidenced, that the
presence of fly ash changes the behaviour
of fine grained soil. The parameters of
sedimentation curve (Figure 17) were
discussed in the paper of Palomino et al.
(2008).
Figure 17. Characteristic parameters of
sedimentation curve (Palomino et al. 2008)
In the mix FA40 the time needed to obtain
clear water above the sediment is shorter.
Due to the fact that the length of induction
period specifies the degree of particles
association, it can be assumed that fine
grained particles present in the mix formed
bigger and heavier flocks in comparison to
pure kaolin. This phenomena is connected
with changed pore water chemistry.
The initial grain size distribution of the mix
results from the grain size distribution of
the components – Speswhite kaolin and
fly ash. It was observed, that the grain size
distribution of the mix is changing mainly
during first week of curing time. Due to the
fact that the test was performed with the
use of dispersant, observed changes in
grain size distribution should be
connected with the formation of new
products, which bind solid particles
together in a permanent way.
Addition of fly ash causes immediate
changes in mechanical behaviour of a soil,
which is a summarized result of changes
in pore water chemistry and introduction
into the soil coarser fractions. Shearing
strength of the mix is increasing during
curing time mainly due to higher values of
cohesion. Changed pore water chemistry
can influence the compressibility
behaviour in short term. Formation of new
products results in changed
compressibility behaviour with curing time,
causing decreased values of volume
111
changes upon loading and increased
values of yielding stress. Compressibility
behaviour observed for Speswhite kaolin
and for the mix FA40 tested at 0 day of
curing time is consistent with the
observations of Lopes et al. (2015).
Authors noticed that the compressibility
curves obtained for the samples prepared
at two different water content tend to
converge at higher stress level as a result
of gradual changes in soil structure
(Figure 18).
Figure 18. Consolidation curves, log(1+e)-
logı’v plot, dashed lines indicate the projection
of the curves for greater loads
(Lopes et al. 2015)
Tests performed by the author of this
paper evidenced that the value of vertical
stress for which the compressibility curves
reach the same point is not stable in
curing time in case of soil treated with fly
ash. This aspect can be explained with the
fact, the initial water content can influence
the rate of new products formation.
Measurements of volume changes
evidenced, that the mix FA40 shows
substantially higher swelling than the
sample of fly ash. It can be assumed, that
hydration and formation of new products is
not the only mechanism causing volume
changes. Having regard to the influence of
initial conditions of the sample on the
swelling it should be highlighted that the
samples were in saturated state from the
beginning of the test. Observed
phenomena is the subject of the future
study planned by author of this paper.
REFERENCES
ASTM-D2435-96 1998 Standard test
method for one-dimensional
consolidation properties of soils.
Brandt A. M., Czarnecki L., Deja J.,
Domagaáa K., JóĨwiak–NiedĨwiedzka
D., KaĔka S., Kasperkiewicz J.,
Koáodziej à., Kowalska D., Litorowicz
A., àagosz A., Maáolepszy J., Marks M.,
Mossakowski P., Mróz R., Radomski
W., ĝliwiĔski J., Tracz T.,
Woyciechowski P., Zybura A. (2010):
Zastosowanie popioáów lotnych z
kotáów fluidalnych w betonach
konstrukcyjnych (Application of CFBC
fly ash in structural concretes).
Collective work edited by A. M. Brandt.
Copyright by Komitet InĪynierii Lądowej
i Wodnej PAN, Warsaw.
DoáĪyk K., Szypcio Z. (2014): Analiza
areometryczna w Ğwietle teorii PKN–
CEN ISO/TS 17892–4 i PN–88/B–
04481 (in English: Hydrometer analysis
in response to the theory, PKN-CEN
ISO/TS 17892-4 and PN-88/B-04481).
Journal of Civil Engineering,
Environment and Architecture, JCEEA,
vol. XXXI, No 61 (4/14), October–
December 2014, p. 65-79.
Giergiczny Z. (2006): Rola popioáów
lotnych wapniowych i krzemionkowych
w ksztaátowaniu wáaĞciwoĞci
wspóáczesnych spoiw budowlanych i
tworzyw cementowych (in English: The
role of calcium and silicous fly ash in
the formulation of modern binders and
cementous materials’ properties).
Cracow University of Technology,
Cracow.
Keller W.D. (1964): The origin of high–
alumina clay minerals. Proceedings of
the Twelfth National Conference: Clay
and Clay Minerals 19, p. 129–151.
Knapik K. (2016): Experimental and
numerical analyses of fly ash from
fluidized bed combustion applications
for selected ground improvement. PhD.
Dissertation. The Silesian University of
Technology, Gliwice.
Lopes B.C.F.L., Tarantino A.,
Cordão Neto M.P. (2015):
Microstructural interpretation of
compression behaviour of Kaolin clay.
Proceedings of the XVI ECSMGE
Geotechnical Engineering for
112
Infrastructure and Development, ICE
publishing, p. 3207-3212.
Mitchell J., Soga K. (2005): Fundamentals
of soil behavior. Third edition. JOHN
WILEY & SONS INC., New Jersey
(published simultaneously in Canada)
Palomino A.M., Burns S.E., Santamarina
J.C. (2008): Mixtures of fine–grained
minerals – kaolinite and carbonate
grains. Clay and Clay Minerals 56, No
6, p. 599–611.
PKN–CEN ISO/TS 17892–10:2004
Badania geotechniczne – Badania
laboratoryjne gruntów. CzĊĞü 10:
Badanie w aparacie bezpoĞredniego
Ğcinania (in English: Geotechnical tests
– Laboratory testing of soils. Part 10:
Direct shear strength tests).
Sulovský P. (2002): Mineralogy and
chemistry of conventional and fluidised
bed coal ashes. Bulletin of the Czech
Geological Survey, Vol. 77, No. 1, p. 1–
11.
Zapotoczna – Sytek G., àaskawiec K.,
GĊbarowski P., Maáolepszy J.,
Szymczak J. (2013): Popioáy lotne
nowej generacji do produkcji
autoklawizowanego betonu
komórkowego (in English: Fly ashes of
new generation to the production of
autoclaved aerated concrete).
Monograph. Silesian Institute, Warsaw.
Vitale E., Deneele D., Russo G. (2015):
Effetti di breve termine sulla
microstruttura di un terreno stabilizzato
a calce. Incontro Annuale dei
Ricercatori di Geotecnica 2015 - IARG.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Barrettes under negative friction of soil surrounding

Iryna MIESHKOVA
1
State Research Institute of Building Constructions, Kyiv, UKRAINE

ABSTRACT
Problem issues of negative friction force in barrettes and piles of a large diameter were analysed.
Tests of such piles on some construction sites in Kyiv were considered. Barrettes were tested by
Ostenberg’s method and scale tests using the bored piles.
Cases of water saturation of soil bases above and below were considered. Comparative tables
were made for the negative friction forces and bearing capacity values for barrettes and piles of a
large diameter in special soil conditions were obtained.

Keywords: barrettes, negative friction forces, piles of a large diameter

very important issue in compliance with

1. RELEVANCE OF CONSTRUCTION
the requirements adopted for this facility.
ON UNSTABLE SOILS IN UKRAINE
Almost more than 70% of the territory of
Construction in Ukraine by increasing the Ukraine is covered by loess soils.
load on base requires the use of pile
This subsidence of soils is observed at
foundations. In such cases barrettes and
water saturation and requires further study
large diameter piles are used. In
of their physical and mechanical
complicated geological conditions (special
properties at the design.
soils and areas of geological processes
impact manifestation) the number of such Construction on unstable soils (loess,
foundations in the design of multi-storey swelling, alluvial, bulk and arranging
buildings is 80-90%. bases in flooded and landslide areas)
require special approaches for each case
Barrettes are kind of piles arranged in
of choosing the optimal design of
cavities using a grapple in coherently-
foundations.
dispersed soils by technology "wall in
ground" [1]. Large diameter piles are piles
2. DETERMINATION METHODS AND
of different mode of arranging with a
CAUSES OF THE NEGATIVE FRICTION
diameter greater than 600 mm.
FORCES
Design of a reliable base for the
building is a guarantee of a durable In unstable soils there can be soil
reliable operation. Therefore, the choice of subsidence under the action of its own
the optimal solution of the foundation is a weight and the load caused by external
factors. For example, there is a water
* presenting author
saturation of soils at raising the level of
ground water during the flooding accident
or due to loss of water in the pipes.
Experience shows that the consequences
of accidents at pipelines damages are
found after a long time.
Unstable soils have a higher porosity.
The diameters of pores in these soils often
exceed the size of the particles of soil by
tens or hundreds times. Water saturation
process occurs very quickly in time. There
is a soil subsidence as a result [4].
2.1 The causes of the negative friction
forces
Loading friction forces are forces arising
on the lateral piles surfaces and piles
foundations if there is a subsidence of soil
base around them. If subsidence of soil
around the pile exceeds the subsidence of
piles and is directed vertically down, it has
a load to the pile [2].
The forces of negative friction may
occur in different circumstances:
x at mineral granules in the planning
area;
x when highly compressible soils are
deposited on the surface;
x at artificial or natural water lowering
which causes an increase of its own soil
weight and incomplete consolidation of
soil foundations;
x vibration compaction of soil during the
movement of transport;
x during the construction of shallow
foundations which have a load to the pile
base near the pile foundations;
The presence of the weak soil layer
within the depth of the pile also increases
the possibility of negative friction forces.
Deformation of the weak layer can be so
great that the soil layers above the
bedding hang on piles and have additional
load to them [3].
2.2 Methods for determining the forces
of negative friction
There are many ways to determine the
negative friction forces.
The first way which was included to the
standards in Ukraine in 1976 recommends
114
taking a specific value of negative friction
forces as 0.01 MPa. But in fact the
negative friction value depending on the
soil conditions is changing. Therefore, the
data in most cases exceed the values that
were in norms.
The second method is static tests on
the construction site. Work of the pile
without the load of buildings and with the
load is different. Consequently, the
bearing capacity of the pile is lowered,
and the value of friction forces is
overstated.
The third way is definition of friction
forces for theoretical solutions of V.V.
Torhashov and P.I. Salnikov. A high
precision of measurements by the
proposed method is achieved by the fact
that to the main pile the load of
PG=0.5·FRK ·A (A is an area upon which
the pile stands, FR is a calculated
resistance at lower end) is applied. When
soil subsidence is defined, the subsidence
of the pile under the effects of load PG and
forces of negative friction DGK is
determined. The value of subsidence is
put on the graphs and the value of total
subsidence SSUM is determined. Forces of
negative friction DGK acting on the base of
the pile are determined as the difference
between the total load FFK and load PG.
The fourth way is performed by the
results of the pile testing by the static load
with accuracy sufficient for practical
purposes and using the graphs of S = f
(PG) which consider the rate of subsidence
increase in time.
The fifth way is performed by the
results of the pile testing for compression
and pulling. The force of negative friction
DGK is approximately equal to limiting
resistance Fu of soil shift on the lateral
surface of the pile.
The sixth way. According to the old
normative documents the definition of
negative friction DGK using experiential
data is recommended. In the absence of
such data it is necessary to take the
tabulated values of negative friction under
existing rules [2]. But the results of field
tests taken in accordance with the
standards in most cases reduce their real
values by one or two times.
2.3 Examples of water saturation of the
soil above and below
During barrettes testing it was found
that on construction sites ʋ1 and ʋ2 the
upper layers of soil are sands of medium
density saturated with water. Therefore, it
is possible to have lowering of
groundwater level or the effect of vibration
which can lead to additional subsidence
within these layers.
When planning the construction site
ʋ3 the unstable upper soil layers (loess
and sandy loam) were cut off.
On this area ground waters lie under
the lower end of the barrette. They are
powered at the expense of the Dnieper
River. When there are seasonal variations
of groundwater levels it is possible to have
a raise of them and as a result the water
saturation of the lower layers of sandy soil.
Let`s consider the case of soil water
saturation above and below (see Fig. 1).
Fig. 1. Scheme of soil water saturation:
a) water saturation above, b) water saturation
below
The general formula for calculation of
the bearing capacity of the pile at negative
friction forces:
FGK=FRK+FFK-DGK,
where FGK and FFK are respectively
characteristic values of resistance of the
pile on the lower end and on the lateral
115
surface; DGK is a characteristic value of
negative friction effecting the pile.
At water saturation above the
calculation of the pile capacity is
performed according to the scheme a) and
at water saturation below it is performed
according to the scheme b).
3. DEFINITION OF NEGATIVE
FRICTION FORCES BY BARRETTES
TESTING RESULTS AND
CALCULATIONS
On the first construction site the testing
was conducted by the Osterberg’s
method. Three o-cells placed at the
bottom of barrettes were used. With their
help the movement under the lower end of
the barrette and on the lateral surface was
determined.
Tables of normative documents [2]
were designed for piles at a depth of 40
m. They consider the regional peculiarities
of soil in Ukraine. Considering the general
nature of changing the characteristics of
the soil resistance R and f the author
continued these tables to a depth of 85
meters, what provides all possible variants
of piles placement by Ukrainian standards.
Thus, the bearing capacity of barrettes
with the length of 63.3 meters was
calculated.
On the second construction site the
tests were conducted on root piles with a
diameter of 620 mm. According to
calculations [2] the bearing capacity of
piles is 4500 ɤN. On the basis of these
calculations the bearing capacity of
barrettes is determined.
Tests using the bored piles with a
diameter of 820 mm were performed on
the third construction site. According to
their results it was concluded about the
increase of deformation modulus of the
last five layers of soil by three times for
calculation of barrettes.
When determining the bearing capacity
of large diameter piles it is necessary to
use the processed test results of bored
piles and values of scale factors.
 European norms [5] allow the use of rise of water level increases the base
model piles to determine the bearing subsidence which leads to the increase of
capacity in conditions when their diameter negative friction forces; on other hand, the
differs from the diameter of field ones by water saturated base thickness is
no more than in two times. decreased. It has more than water
Barrettes calculation results at the saturated contact friction which reduces
water-saturated soil above and below are the rate of growth of negative forces. At
given in Table 1. achievement of a certain level of water
Table 1. The results of barrettes testing by saturation front these processes are
static load on the construction sites of Kyiv balanced and at the subsequent rise of
ʋ Object Bea- Nega- Bea- Bea- Bea- Dif- ground waters the negative friction force is
name, ring tive ring ring ring fere- decreased [3].
length l, capa- friction capa- capacit ca- nce,
m, city of on the city on y pa- % Negative friction forces are inextricably
cross- the lateral the accordi city
section lower surfa- lateral ng to accor
connected with the determination of the
or end of ce DGK, surfa- DBN, ding pile bearing capacity. In accordance with
diamet the pile kN ce FFK, FGK, kN to the
er of FRK, kN MN test the calculations this difference can
the pile, re- constitute a minimum 2.78% (water
m sults
Fd, saturation below) and a maximum 34.4%
kN (water saturation above).
1 2 3 4 5 6 7 8
1 Sky Natural soil moisture In accordance with the obtained results
tower 58.62 - 20.2 78.82 84
9.92
the determined value of barrette bearing
l=63.3; Soil water-saturation above
ȼ23-
capacity is influenced by a large number
1.2×2.8 58.62 3.92 16.3 71 of factors:
2 Parus Natural soil moisture
x piles diameter and depth of laying;
l=28,6; 11.06 - 13.0 24.06 The
B1- 18.17 - 19.1 37.27 test x the bearing layer of the base;
18.17 - 19.1 37.27 resul
2.8×1 ts of x water saturation and unstable soil
B2, B3-
2.8×2.8
root
piles
34.4 layers;
32.4
Soil water-saturation above
32.4 x pile testing methods (static scale tests
11.06 4.15 8.87 15.78
18.17 6.11 13.1 25.16 or Osterberg’s method).
18.17 6.11 13.1 25.16
3 Natural soil moisture
Mirax
12.32 - 15 27.32 The
4. CONCLUSIONS
plaza,
test
l=32.7 resul According to the results of tests and
ȿ1- ts of calculations of large diameter piles and
2.8×0.8 bore
d 2.78 barrettes it was found that the negative
piles friction forces on the lateral surface are
Water saturation of the soil below
inextricably connected with the value of
12.32 0.38 14.6 26.56 the bearing capacity and quite significantly
Note: barrette B1 is rectangular sectional, barrette B2 is affect it.
tee-sectional, and barrette B3 is cross-sectional. When designing deep foundations it is
Water saturation of the soil above and necessary to pay attention to construction
below has different negative friction site soil conditions, namely the presence
forces. At water saturation above the of unstable layers deep in the soil and
thickness of soaked base layer is groundwater level and adopt the
gradually increased and therefore, the guaranteed value of bearing capacity of
subsidence of the pile is increased piles taking into account the possibility of
causing the increase of negative friction manifestation of negative friction.
force. The obtained values DGK for soil
conditions in Kyiv confirm this conclusion. ACKNOWLEDGEMENTS
Slightly different picture is observed at The author expresses gratitude to the
water saturation above: on one hand, the management of the State Research
116
Institute (SE NIISK) and the International
Society for Soil Mechanics and
Geotechnical Engineering for the provision
of opportunities to participate in testing of
pile foundations on construction sites of
Kyiv and for promoting the participation in
international conferences.

REFERENCES
1.DBN B.2.2-24-2009 Buildings. Designing
of high-rise residential and public
buildings/ Ministry of Construction in
Ukraine, Kiev 2009. - 154 p.
2.DBN 2.1-10-2011Bases and foundations
of buildings. General design provisions.
/ Piles and pile foundations / Ministry of
Construction in Ukraine, Kiev 2011. - 54
p.
3.Karpenko D.A. (2005) “Investigation of
bearing capacity of pile in loess soils
depending on the type of water
saturation of soil base” (Eds: Kornienko
N.V.), Promstroyniiproeject, Donetsk.
4.Dolmatov B.I. “Mechanics of soils, bases
and foundations“ (Eds: Bronin V.N.,
Karlov V.D., Mangushev R.A.),
stroyizdat, Leningrad, Moscow.
5.Eurocode 7: EN 1997-1:2004.
Geotechnical design – Part 1: General
rules.
6. Patent ʋ1158676, Russian Federation,
SU 1158676 A. The friction piles in
thawing soils / V.V. Torgashov, P.I.
Salnikov.

117
118
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Engineering geological investigation of a cave spa cut into

rhyolite tuff

Ádám VINCZE1*, Péter GÖRÖG1

1
Budapest University of Technology and Economics, Department of Engineering Geology
and Geotechnics, Budapest, HUNGARY

1. ABSTRACT
The goal of the paper is to determine the properties of the host rock and perform stability analysis
of an artificially formed underground thermal bath in Demjén, Hungary.
The host formation is rhyolite tuff which is abundant in the area. Cellars and similar underground
facilities were cut in this rock for hundreds of years. What makes this case special, is that the
rhyolite tuff is quite water sensitive and have a considerable reduction in strength in saturated
state. Furthermore, the layout of the underground openings is very complex compared to a cellar
which usually cut into this formation. The aim was to determine, if the planned support was
sufficient either in construction phase and in case of pool leaking which means reduced rock mass
strength. The physical parameters of the rhyolite tuff were determined under laboratory conditions.
The complex layout of the facility made it necessary to carry out calculations both in 2D&3D. For
modelling the underground openings finite element software products (Rocscience, Cesar) were
used. The paper shows the underground thermal bath in construction phase and the experience of
the tunnel driving also concluded since the spa is now open.

Keywords: rhyolite tuff, underground opening, stability analysis, finite-element

method

2. INTRODUCTION expensive for many people to afford. The

owner of the (now built) cave spa saw the
In 1961 exploratory drillings for oil and gas
potential in drilling another well thus
were conducted in the area of Demjén,
creating a simpler, but cheaper spa just
Hungary. In one particular well, hot karst
3km-s south of the existing one. There
water came to the surface. It was found,
was one problem: the well was too far
that the water has unique chemical
from the road, so they built a gravity-fed
composition, similar to Pamukkale in
pipeline which brought it closer. In order
Turkey, which is a hot spring spa since the
to do this they had to cut a small tunnel
roman times, so it was obvious how to
through a small hill with rhyolite tuff inside.
make use of it.
Then came the idea to create a cave spa
By the end of 2010 they built a spa into the hill next to the planned outdoor
complex in the valley which became too spa.
* presenting author
For centuries, thousands of cellars and
other underground caverns have been cut
into the rhyolite tuff in the surrounding
regions without any particular support and
many are still in use, so it was not an
impossible idea. However the planned
facility was far more complex.

3. GEOLOGICAL ENVIRONMENT
The spa is located south of the Bükk-
mountains in its foreland, in a transition
zone between the lowlands and
mountainous area. The oldest known
formation is Triassic limestone which is
considered to be the basement rock in the
area, with a depth of 700-800 m. This is
covered by Eocene conglomerates with a
thickness of around 100m. On top of this
Oligocene sandstone and marl was
deposited with a considerable 500-600m
thickness. The last important layer was
formed during the Miocene due to volcanic
activity in the area. Several cycles of
eruption produced 300-400m thick
blankets of rhyolite tuff. This appears on
the surface in a 30 km wide region. The
uppermost layers are: Pleistocene clay
which was created through the erosion of
tuff. In the valleys Pleistocene and
Holocene sediments can be found: silt,
clay sometimes coarse rock debris.
3.1. Hydrogeology
Groundwater does not affect the planned
facility since it’s level is several meters
below the floor level.
The water which supplies the entire spa
comes from a well located a few hundred
meters away from pools. It has a depth of
696m-s (Eocene layer) with a flow rate of
500l/min. The water is 69°C. As it comes
to the surface and pressure is lost, the
dissolved mineral content is precipitated.
This can sometimes occlude the duct
pipes. Near the well mentioned in the
introduction the water is let to run down on
the hillside, this way beautiful white
formations are created.
120
Figure 1: „Salt hill” precipitated minerals at
Egerszalók
3.2. Description of the rhyolite tuff
The rhyolite tuff is a basic pyroclastic rock.
It has a cream color when dry, and
greenish-yellow when wet. Three main
components can be distinguished: rock
matrix, phenocrysts, rock blocks and
volcanic bombs. The typical composition
of the first is 20% silica 55% feldspar 25%
biotite and amphibole. The structure is
spongy and porous. This property affects
the water resistance negatively. It can be
stated for rhyolite tuffs in the region of
Eger based on experience and many
studies, that the water saturation leads to
a considerable drop in rock strength.
The tuff can be easily carved, therefore
it was widely used in this region as a
building material and host rock for any
underground cavern throughout history. It
has also an important role in vine culture.
4. GEOTECHNICAL INVESTIGATION
The aim of this part was to determine the
input parameters for further calculations
and modeling. This process was divided
into four main parts: gathering existing
literature from the vicinity of the spa, field
tests and sample collection, laboratory
tests, evaluation of the information.
4.1. Field and literature exploration
Since the construction was in progress
when these investigations were to be
carried out sample collection and field
tests were easy. Some of the tunnels were
already cut, so the blocks could be
retrieved directly from the tunnel walls and
crown. It was also easier to get an idea on
the rock mass properties, how fractured it 0,4MPa. This can be a problem in an
was, compared to individual drill samples. occasional pool leak in the future.
Preliminary tests and calculations were The input parameters for modeling are:
carried out by a company. This included Table 1. Elastic rock mass properties
mapping of discontinuities, compressive Elasticity Possion’s
tests, preliminary stability analysis. Some Name
Unit weight
kN/m3
module ratio
of their results were used as well. MPA -
original state
16 270 0.22
tuff
saturated tuff 17 40 0.23
Table 2. Plastic rock mass properties
Tensile
Name Friction Cohesion
strenght
angle deg MPa
MPa
original
43 0,054 0,124
state tuff
saturated
37 0,034 0,028
tuff
Elastic properties were obtained from
the ı-İ diagrams of uniaxial compressive
tests. Plastic properties are form triaxial
tests which were evaluated in Roc-lab.

Figure 2: Layout of the planned facility
4.2. Laboratory tests
The collected samples were prepared and
tested at the university’s1 department lab.
The carried out experiments were: water
content and saturation tests, indirect
tensile strength test, uniaxial compression
test, triaxial compression test. All of these
were conducted according to the ISRM
blue book.
To verify the results, values from the
contractor’s samples were used. These
were tested in ÉMI’s accredited lab. The
numbers I-III are marking our samples 1-
17 are ÉMI’s on figure 2.
4.3. Results
The rock mass is mainly homogenous the
entrance area is jointed (marked with lines
on fig 2.) due to erosion and movement
towards the valley. The homogenous part
has a water content around 17-20m/m%
with an average of 1,5MPa uniaxial
compressive strength. Compared to
literature data these values correspond to
results of rhyolite tuff in the area.
When fully saturated the rock has 24-
25% water content and the compressive
strength is reduced to an average of
121
5. STABILITY ANALYSIS
It can be seen, from the layout, that the
task is hard to be simplified to a two
dimensional problem. One cross section
was chosen for two dimensional-, and one
region in the middle for three dimensional
analysis. (fig. 2) The tunnel profiles are
reverse U shaped 3.8m wide and 3.8m
high. The goal was to determine stability
during construction phase and to give an
estimate to an emergency case: leaking
pool.
5.1. 2D analysis
For the modeling Rocscience’s Phase 2
finite element method software was used.
Input parameters according to Table 1 and
2. As these are characteristic values,
during a shear strength reduction analysis
strength reduction factor 1.5 or greater
was acceptable.
During the construction phase injected
anchors were installed. By an analytic
calculation we determined that the lowest
of the possible failure modes was pull out
resistance, because of the tuffs properties.
These were also taken into account in the
calculations. In construction case analysis
there was no further support. In
emergency case analysis a further 20cm reduction factor of 1.02 (no anchors) was
thick reinforced concrete lining was added. determined.

 also recommended.
Figure 3. Cross section in Phase2
The conclusion was that in construction
phase there was a safety factor of 1.53
which is just above the expected. In
emergency case the area under the
tunnels was modeled with saturated
properties and the concrete lining was
applied. According to the results the safety
factor rises only to 1.84, and the
displacements are 2-3 cms. At some
places the concrete lining’s capacity is
exceeded.
5.2. 3D analysis
For the modeling CESAR-LCPC v4.0 3D
geotechnical finite element method
software was used. The modeling with this
software was quite time consuming and
complicated, therefore only one the
construction phase was analyzed.
6. CONCLUSION
The rock mass has a sufficient stability
during construction. In the emergency
case when the tuff gets saturated it is
imperative that a drain system is created
so if there is any leaking water it won’t be
absorbed by the rock. As a preventive
action a moisture monitoring system is
Today the spa is fully operational
without any geotechnical or structural
problems.
REFERENCES
Kleb, B. (1977) – Engineering geological
evaluation of the region of Eger
R. Ulusay, J.A. Hudson. eds. (1974-2006)
The Complete ISRM Suggested
Methods for Rock Characterization,
Testing and Monitoring
http://www.cesar-lcpc.com/
https://www.rocscience.com/


Figure 4. Full model in CESAR
In this version it strength reduction
analysis it can only be done manually by
reducing the material properties. Neither
the anchors can be taken into account.
This way after five iterations a strength

122
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Continuous Horizontal Reinforcement Diaphragm Wall

Application

Nejla YILDIZ1*, Alp GÖKALP2,

1
KASKTAù A.ù., østanbul, Technical Office Engineer, TURKEY
2
KASKTAù A.ù., østanbul, Projects Coordinator, TURKEY

ABSTRACT
This paper describes the design concept, construction and testing of a box type foundation system
composed of a 100 cm thick diaphragm wall used to support foundations of the South Approach
Viaducts; which is a part of Izmit Bay Bridge Project and located at one of the most seismically
active places in the world. However, since the project location crosses North Anatolian Fault and it
lies on the secondary fault zone, different approaches are implemented for the construction of
these diaphragm walls. This paper describes the construction methodology and application stages
of project in order to maintain continuous horizontal reinforcement; which is quite different from the
traditional methods.

Keywords: diaphragm wall, continuous reinforcement, earthquake,

the North Anatolian Fault Zone (source of

1. INTRODUCTION
The application which is the subject of this
paper has been implemented within the
scope of Izmit Bay Bridge, a 3-km-long
suspension bridge that crosses Izmit Bay
in Turkey. Project site is in one of the most
seismically active places in the world.
When the bridge is completed, it will be
the fourth-longest suspension bridge in
the world by the length of its central span.
The bridge will connect the Diliskelesi
peninsula to the North with the Hersek
peninsula on the south. The proposed
project site spans the plate boundary
between the Anatolian plate on the south
and the Eurasian plate on the north and
will experience significant earthquakes on
* presenting author
the 1999 Magnitude Mw 7.4 Izmit and Mw
7.2 Düzce earthquakes).
The bridge is the critical link of the 420km
Gebze–øzmir Motorway awarded through a
Build Operate Transfer (BOT) model to the
NÖMAYG Joint Venture in 2009. Nurol
ønúaat ve Ticaret A.ù. was awarded by
NÖMAYG for the design and construction
of the on-going South Approach Viaducts
of the Izmit Bay Bridge (Figure 1). The
structural design was carried out by
Wiecon; geotechnical, geological, and
seismological evaluations of the proposed
project region was performed by Fugro
and diaphragm wall construction works
were carried out by Kasktaú A.ù.
2. DESCRIPTION OF THE PROJECT
2.1. Project Area
The South Approach Viaduct (SAV) is
located along the western margin of
Hersek peninsula, within 5 km from the
North Anatolian Fault and within a zone of
secondary deformation around the primary
trace of the North Anatolian Fault. The
alignment crosses the western Hersek
peninsula shoreline approximately 500
meters south of the Northern tip of the
peninsula, and continues south with the
centerline within about 80 meters of the
shoreline.
South
Approach
Viaducts
Figure 1. Location and Alignment of the South
Approach Viaducts (SAV)
The South Approach Viaduct of the Izmit
Bay Bridge brings the bridge down from
on the order of El. 60 meters at the South
Anchorage to an elevated embankment
approximately 1.3km farther south. The
proposed viaduct consists of 11 Piers and
10 standard spans with lengths varying
between 136 m and 100 m and two bank
spans one of 125 m, attached to the main
bridge and the other of 74 m attached to
the south embankment (Figure 2). The
South Approach Viaduct is located
between km 7+084.26 (Pier P0) and km
8+462.43 (Pier A12). The SAV piers are
numbered consecutively starting from Pier
P1 which is located south of the South
Anchorage of the main Bridge to Pier P11
and terminating at the south embankment
A12, located south of Pier P11. The
interface between the main bridge and the
SAV is at the South Anchorage of the
main bridge (Pier P0).
124
Figure 2. General Layout of South Approach
Viaduct Footings
2.2. Subsoil Conditions and Geology
Prior to the construction of the foundations
of the South Approach Viaduct, a detailed
geophysical and geotechnical survey
program was executed between 2011 to
2012 by Fugro Sial. The site investigation
consisted of 10 no. of boreholes down to
60 m depth; 126 no. of Menard
Pressuremeter tests in all boreholes; 48
no. of CPT; and laboratory tests. Since the
site is underlain by deep deposits of soft
soils, and areas of unstable and
liquefiable soils, characterizing the
geological, seismological and
geotechnical setting, foundation soil
conditions, fault locations, as well as other
geo-hazards and developing appropriate
design criteria was the most critical
component for the project. Therefore; a
sophisticated and extensive site
investigation program of the Izmit Bridge
was carried out by Fugro in order to
provide geotechnical engineering services
for the final design of the proposed
viaducts.
The soil layers encountered at the site are
presented below:
Loose to Medium Dense Sand Layer: The
thickness of the layer changes between
2m and 5m and it is prone to liquefaction.
The color is grey and SPT N value is
between 5~10; n=18~18.8 kN/m3;
Ø=25~35°.
Stiff Clay: The thickness of the layer
changes between 2m and 5m and in
yellowish brown color. SPT N value is
between 10~20; Cu= 75~200 kPa;
3
n=18.8~19.0 kN/m .
Very stiff to hard clay: The thickness of the
layer changes between 18m and 35m and
in greenish gray to dark grey color. SPT N
value is N>30; n§19.0 kN/m3, Cu •200
kPa
The ground water depth is about ~2 m.
The idealized soil profile according to the
existing boreholes is shown in Figure 3
below:
Figure 3. Generalized Soil Profile
Having evaluated the results of the
shallow water geophysical and
geotechnical survey data shows presence
of several secondary faults within the area
and as a result the location of the south
anchorage of the main bridge was moved
approximately 150 m north from the
originally proposed location to an apparent
area of no recent faulting.
This shift to the North necessitated the
extension of the South Approach Viaduct
by about 150 meters to the North into an
area of identified secondary faults.
Consequently; Piers P01 and P02 of the
Viaduct are located within number of
active fault traces of North Anatolian Fault
(Figure 4)
125
Figure 4. Secondary Faulting Zone between
P01-P02 footings of SAV
3. FOUNDATION CONCEPT
The performance evaluation was focused
on the following aspects:
• Foundation axial capacity
• Foundation performance under static
and earthquake loads
• Foundation performance against fault
rupture induced displacements
Three dimensional finite difference
analyses were conducted by Fugro using
the computer program FLAC (Itasca 2011)
to develop the axial load-deflection curve
for different foundation types and sizes.
The shallow footing dimensions in plan
view that were analyzed are 30 m x 30 m
(longitudinal x transverse), 28 m x 33 m,
26 m x 33 m, 25 x 25 m, and 20 m x 20 m.
The base of the shallow foundations is at
elevation -5.5 m.
The diaphragm wall foundation system
dimensions that were analyzed are 8 m x
21 m x 13 m (longitudinal x transverse x
depth), 8 m x 21 m x 23 m, 13 m x 21 m x
23 m, and 15 m x 27 m x 23 m. The cap
size thickness was 3 meters, with the cap
bottom at -2 m. (Table 1)
Analyses were performed to evaluate
the performance of two different
foundation types for the SAV piers, a
shallow foundation and a diaphragm wall
foundation system. Due to the severity of
the design ground motions, the
superstructure introduces significant
moments on the foundation
 Table 1. Diaphragm Wall Distribution As a result; a caisson-type of foundation
Pier Width Length Depth
was selected as the most suitable
foundation system; which consists of four
No (m) (m) (m)
perimeter diaphragm walls, a concrete
P01 21 8 22 cap, and a diaphragm wall constructed
along the bridge transverse direction
P02 21 8 22
under each Pier legs. The thickness of the
P03 21 8 22 diaphragm walls is selected as 1 meter
whereas the cap thickness is selected as
P04 21 8 17
3 meters. The foundation concept is
P05 21 8 17 shown schematically in Figure 6.
P06 21 8 14
P07 21 8 14
P08 21 8 14
P09 21 8 10
P10 21 8 10
P11 21 8 10

For a shallow footing solution, the size of

the footing is driven by the overturning
resistance to the superstructure loads
rather than the vertical bearing capacity.
Additionally, since fault rupture through a
pier foundation is a concern for this
project, the foundation system has been
found to play a key role in the response of
structures subjected to fault induced
ground movement. Structures resting on
rigid and continuous foundation systems
(such as a raft, or a box-type foundation)
have demonstrated to be capable of
achieving a very satisfactory performance,
irrespective of the faulting type/system.
“The Box Type Foundation Layout Plan” is
shown in Figure 5; whereas “The Box
Type Foundation Concept” is shown in
Figure 6.
Figure 5. Diaphragm Wall Panel Layout Plan
126
Figure 6. Box Type Foundation Concept
4. CONSTRUCTION
The design of box-type foundation
necessitates a job specific special
diaphragm wall methodology; which
requires special tools and techniques; in
order to maintain continuous horizontal
reinforcement; which is quite different from
the traditional methods. The success of
this application in achieving and
maintaining the required design loads has
paved the way for further applications
within the region with a high seismicity.
4.1. Test Panels
By bearing on mind the fact that the
selected diaphragm wall methodology was
not only the first one to be applied in
Turkey but also the number of similar
projects completed worldwide was very bentonite quality control tests prior to
limited. Therefore; before commencing concreting. Then, reinforcement cages
construction, a number of trial test panels and special tools appropriate for each type
having the same features as the box-type of panel were placed in to the excavated
foundation diaphragm wall panels were trench.
constructed down to the design depth to
calibrate the construction procedures.
These trial test panels were executed so
as to verify implemented construction
methodology for continuous horizontal
reinforcement diaphragm wall in the
vicinity of the working location.
4.2. Construction Stages

4.2.1 Excavation
Panel excavations were executed by
using hydraulic grab. During the
construction of test panels, instability of
panels due to the uppermost loose sandy
layers was observed. In order to take
precautions to prevent this situation in
application project; jet grout columns were
implemented under the guide walls for
ground improvement purposes.
Panel layout plan and sequence of
construction works were prepared for each
footing. The diaphragm wall panels are
classified according to the construction
procedures as Starter (S), Intermediate (I)
and Closer (C). The sequence of work is
shown schematically in Figure 7.
Figure 7. Diaphragm Wall – Sequence of
Construction Works
Figure 8. Diaphragm Wall Panel Excavation
4.2.2 Preparation and Placement of
Reinforcement Cages
Since the diaphragm wall construction
required a special technique for providing
the continuity of the horizontal
reinforcement, special reinforcement
cages in different shapes and dimensions
for each panel type were used in
accordance to the dimensions of the
excavated diaphragm wall panels. Besides
as requested by the Designer all
horizontal reinforcements of the adjacent
panels are overlapped by 700 mm as
shown in Figure 9.
During preparation of starter and
intermediate panel reinforcement cages,
steel end-plates were welded to cages to
provide a barrier between the concreted
and un-concreted sections.

L and T shape starter panels (S) and

intermediate panels (I) were excavated in
two stages, whereas closer panels were
excavated in one stage. (Figure 8)
After diaphragm wall excavations were
finished, bentonite samples were taken
from the bottom of the excavated trench
and taken to the laboratory for standard Figure 9. Horizontal Overlap Detail

127
While the reinforcement cage was lowered
into the excavated trench, concreted
section of the cage was covered with
geotextile at the mouth of the trench in
order to prevent concrete leachate to
adjacent panel and thus, to preclude the
problems that would occur during
construction of adjacent panel. (Figure 10)
This fabric also surrounded the bottom of
the cage to provide maximum protection
against leakage of concrete beyond the
partition steel end-plate. The secondary
panel cage (female cage) was designed to
allow a proper splicing between the
subsequent panel cages.
Figure 10. Covering the reinforcement cage
with geotextile
Each steel reinforcement cage was
assembled horizontally on the ground in a
single longitudinal section. Each cage was
stiffened, using convenient stiffening
elements placed between the main
reinforcing bars, to provide the cage
rigidity needed to avoid deformations
during lifting and lowering into the trench.
In addition, proper welding was performed
to increase the stiffness of the cage during
handling and lifting. Appropriate spacers
were placed on the faces of the cage to
ensure the correct concrete cover. All the
cages were equipped with a number of
60-mm-diameter steel sonic pipes to
measure the concrete integrity. The
distance between the sonic pipes was not
greater than 2.0 m.
The maximum weight of each cage was
about ~40 tons. The handling process for
reinforcing cages was performed by two
service cranes. Each cage, assembled in
128
a horizontal position due to its exceptional
shape and dimension, was lifted from its
horizontal position and suspended
vertically by means of the two service
cranes. A special lifting frame was used to
avoid any localized overstress of the
assembled cage. (Figure 11)
Figure 11. Special lifting frame used in the
Project
Once the cage was in the vertical position,
crawler crane supported the cage from the
top and moved it to the open trench. The
cage was slowly lowered into the trench
and, once down, was suspended on the
guide walls through bars welded to the
main longitudinal bars. Therefore, the
reinforcement cage did not rest on the
bottom of the trench, and the clear
distance between the reinforcement and
the bottom of the trench was not less than
150 mm.
4.2.3 Concreting
Concreting works were executed by using
tremie method. Prior to starter and
intermediate panel concreting works, gaps
which existed outside of steel end plates
were filled with a suitable fill material.
Concreting and filling processes were
performed in a simultaneously manner
one after another until concrete overflew
at the head of trench was observed.
(Figure 12)
Figure 12. Simultaneously Concreting and
Filling Processes
5. QUALITY CONTROL TESTS AFTER
CONSTRUCTION WORKS
As part of a strict quality assurance
program pursued in the project, cross hole
sonic logging tests (CSL) were performed
at each footing in 100 different points
(0.8 nos/m2) to investigate continuity and
integrity of diaphragm wall panels along
their whole depths as shown in Figure 13.
This test enables information about the
discontinuities and defects that could be
occurred during concreting works.
Moreover, tests were executed by
analyzing the sound waves delivered
between transmitter and the receiver
probes placed on the steel pipes welded
to the reinforcement cages while
preparation of them. By examining the
delivery time and magnitude of signals,
continuity of the panels which are
constructed can be interpreted. (Figure
14)
Figure 13. Layout of Cross Hole Sonic
Logging Tests
129
Figure 14. A sample output of a Cross Hole
Sonic Logging Test
6. CONCLUSIONS
In general, when designing structures in
seismically active areas, foundations of
critical structures are typically located
away from known faults. However, for long
structures such as bridges, tunnels and
pipelines, a fault maybe unavoidable, and
fault rupture risk impossible to preclude. In
the subject project interpreted
geotechnical and geophysical data
collected during site investigation revealed
numerous traces of the secondary fault
zone on the entire area near the south
anchorage of the main suspension bridge
and the south approach viaduct. Therefore
foundation design for the approach
structures was optimized by performing
probabilistic fault displacement hazard
analyses in combination with advanced
numerical soil structure interaction studies.
In order to cope with the above mentioned
difficulties, a box-type diaphragm wall
system with continuous horizontal
reinforcement was selected as the most
suitable system.
 This paper describes a new
construction method of diaphragm wall
successfully carried out in Turkey using a
continuous horizontal reinforcement
according to the strict HSE and Quality
Control / Quality Assurance Programs
implemented in the Project. Significant
data related to the construction and
design of the proposed system was
documented, and the technical details of
the design approach were highlighted.
This project is an important model, which
verifies that major iconic structures can be
constructed at one of the most seismically
active places in the world.

ACKNOWLEDGEMENTS
I would like to thank you KASKTAù A.ù.
for both encouraging and supporting me to
participate in this conference.

REFERENCES
Fugro (2012) “Soil Structure Interaction
Evaluations for the South Approach
Structures, Izmit Bay, Turkey”,
prepared for Otoyol YatÕrÕm ve øúletme
A.ù.
Fugro-Sial, (2012)-a, Additional Site
Investigation Results for South
Approach Viaduct
Fugro-Sial, (2012)-b, Performance
Evaluation of Different Pier Foundation
Types
Fugro-Sial, (2013), Report on Crosshole
Sonic Logging Nondestructive Testing
of Foundation No:09

130
25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Session 3

Underground structures
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Energy walls for an underground car park

Alice DI DONNA1
1
Politecnico di Torino, Department of Structural, Geotechnical and Building Engineering,
Torino, ITALY.

ABSTRACT

Energy geostructures are increasingly considered as a new technology coupling the structural role
of retaining structures and foundations with the possibility of exchanging heat with the ground to
cover heating and cooling demand of buildings. In this paper, the possibility of applying such
technology to the diaphragm walls of an underground car park in Torino was investigated through
numerical analyses. It was found that this promising technology could cover a significant portion of
the energy demand of a residential building in the proximity of the car park. The influence of the
thermal activation of the walls on the underground temperature was also investigated and it was
found to be acceptable.

Keywords: geothermal energy, energy walls, FEM

structural reasons, such as piled

1. INTRODUCTION
The energy requirement is increasing
worldwide and a significant portion of it is
represented by the heating and cooling
needs of buildings. In most countries, this
is currently mainly provided by gas, oil and
coal. Among the possible alternatives,
conventional geothermal systems are
recognised as clean, renewable and local
sources. However, such systems require
an initial investment and sometimes a
large area for installation, which make
them practically ineffective from the
economical point of view. To face this
issue, the so-called energy geostructures
are rapidly spreading in Europe and
around the World [1]. They are
underground structures designed for
* presenting author
foundations, slabs, diaphragm walls and
tunnel linings and anchors, which are
equipped to exchange heat with the
ground similarly to standard geothermal
boreholes. Using geostructures which
would be constructed in any case the
initial cost of installation is significantly
reduced with respect to conventional
geothermal systems. This paper focuses
mainly on the application of this
technology to the diaphragm walls of an
underground car park in the city of Torino
(Italy). Efforts have already been devoted
to similar applications [2–6] and a number
of real operational systems exist [7–9].
The work presented hereafter was part of
a research project founded by the
European found for Piedmont Region
(Italy) through the Innovation Pole Temporaryactiveanchors,
Enermhy, in collaboration between the 45° inclined,totallength
25m,activelength15m,
Politecnico di Torino, Resolving srl and spacing2,5m
Teknema Progetti srl.
Sewer

2. THE CASE STUDY

Sewer
The considered case study is a three level
underground car park designed but not yet
constructed in Torino, located in the SE
district of the city. The geotechnical and
Energydiaphragm
structural project, provided by Teknema wallsequipped
Progetti srl, included retaining structures withpolyethylene
pipes,D=25mm,
all over the external perimeter. The th=2.3mm
possibility to transform them into energy
Groundwater
walls would represent an incentive to build flowdirection
the car park and an undeniable added
point for the city municipality. The
designed car park has a rectangular Figure 2 – Vertical section of the energy wall.
shape of 93.15 x 52 m2. On one side, the
At the site, the water table surface is
car park is in contact with the basement of
approximately 5 m below the ground level
another structure, but on the three others
and the thickness of the aquifer is
the retaining walls are in contact with the
estimated in 22–23 m. The water in the
ground and can be reasonably equipped
aquifer has an average temperature of
as heat exchangers (Figure 1). According
14ƕC and flows toward the Po River with
to the project, the retaining walls are 15.5
an average velocity of 1.5 m/day (towards
m depth (Figure 2). The subsoil conditions
SE). The average hydraulic, hydro-
in Torino are characterised by the
dispersive and thermal parameters of the
presence of a sand and gravel deposit,
aquifer (Table 1) are known from in situ
ranging from medium to highly dense,
pumping tests and monitoring performed
down to a depth of 8–10 m. Below this
in the city [12].
depth lenses of cemented soil (in cases a
conglomerate) are often present [10,11]. Table 1. Torino subsoil properties.
Horizontal hydraulic
GROUND kh [m/s] 4.15·10í3
conductivity
Vertical hydraulic
kv [m/s] 0.21·10í3
Energy conductivity
diaphragm
walls
Porosity n [í] 0.25
Bulk ȡc
2.55
[MJ/m3/K]
ADJACENTBUILDING

Ground heat capacity

GROUND

waterflow
93,25m

direction Bulk thermal

Ȝ [W/m/K] 2.26
conductivity
Longitudinal
ĮL [m] 3.1
Possible dispersivity
Energyslab
Transversal
ĮT [m] 0.3
dispersivity

3. FROM CONVENTIONAL TO ENERGY

GROUND DIAPHRAGM WALLS
52,0m
In order to transform the diaphragm walls
Figure 1 – Plan view of the car park. into heat exchangers, polyethylene pipes
134
have to be installed and attached to the
reinforcing steel cage, before concrete
cast. In this study, the pipes were
supposed to have diameter of 25 mm. The
position of the pipes inside the walls was
selected based on a preliminary
optimisation study. The pipes are installed
only on the wall side towards the ground
(Figure 3). The inlet/outlet pipes of each
panel are assumed to be connected to the
main circuit, which links them to the heat
pumps.
Groundside Parkside
one wall panel, having height of 15.5 m,
thickness of 0.8 m and width of 2.5 m
(Figure 2). The model was checked for
mesh sensitivity. According to the Torino
subsoil conditions, the initial temperature
was fixed to 14 °C for the whole domain
and the water table was positioned 5 m
below the ground level.
2,5m outlet
inlet
9,5m
25,5m

polyethylene
pipes, 35,8m
D=25mm,
th=2.3mm Figure 4 – 3D model.
10,0 0,8 25,0

Pipes
5,0 5,5 position

Wall 9,5
A
15,5
7,5 Slab
1,0

B 5,0

7.0
Soil (Tini =14°C)
10,0 C 10,0
Groundside polyethylene
pipes,
D=25mm, External airtemperature
th=2.3mm Fixed temperature14°C
Referencepoints

Parkside
Figure 3 - Pipes position and reinforcing cage.
4. NUMERICAL MODEL
A thermo-hydraulic mathematical
formulation was required to simulate the
thermal exchange between the fluid
circulating through the pipes, the concrete
and the surrounding soil. To this end, the
finite element software FEFLOW© was
selected. The absorber pipes installed in
the wall panels were simulated through
the 1D discrete features elements
provided in FEFLOW©. The 3D model
adopted is presented in Figure 4 and
Figure 5. It reproduces the geometry of
135
Figure 5 - Geometry of the model.
Constant hydraulic head and temperature
were fixed on the left, right and bottom
sides of the model, which were checked to
be far enough not to affect the results.
External air temperature was fixed on the
top boundary, according to Torino average
annual temperature variation (Figure 6).
The establishment of the most appropriate
boundary condition to be applied on the
internal car park wall and excavation plane
was a complex task because, with respect
for instance to energy piles, energy walls
are exposed to the air on that side [2].
Two main approaches have already been
suggested in the literature, either a
constant temperature [5,8,13] or a
convective heat flux determined by a heat
transfer coefficient [14,15]. The second
one was mainly used for metro and train
tunnels where air circulation
predominant, while the first one for
basements, underground stations and car
parks, as in this case study. In the
absence of monitoring data related to the
considered park internal temperature three
different conditions were tested:
temperature fixed to 18°C, to 14°C and
adiabatic boundary. The imposed thermo-
hydraulic properties of the soil were
representative of Torino (Table 1), while
those of the concrete and heat carrier fluid
were those collected in Table 2. The inlet
velocity of the heat carrier fluid was
imposed equal to 0.2 m/s, while the inlet
temperature was assumed according to
Figure 6. The temperature of the model
was initialized by running one-year
simulation without activating the
geothermal plant. The activation of the
system was then simulated for a three-
year duration.
30
25
Temperature,T[°C]
induced temperature variation in the
subsoil are discussed in the following.
5.1. Heat exchange
is
Figure 7 shows the inlet and outlet
temperature for the different internal wall
boundary conditions considered. It is clear
that the boundary condition on the wall
side has a remarkable influence on the
heat exchange. From the difference
between the outlet, Tout, and the inlet, Tin,
temperatures, the exchanged heat Q,
measured in Watt, can be computed as:
Q=mācā(Tout-Tin) 1
where m is the mass fluid rate in the pipes
in kg/s and c the specific heat capacity of
the circulating fluid in J/kg/K. The results
showing the exchanged heat in W per
meter of wall depth are presented in
Figure 8 (positive means heat extraction,
i.e. winter mode). If the wall internal
temperature is fixed, the system can
exchange heat not only with the ground
but also with the internal park air and it
results into a higher efficiency.
Externalair Reasonably, between the two
Inlet
configurations that assume constant
temperature, the case of 18 °C is more
efficient in winter and less efficient in

20
summer with respect to the one at 14 °C.
15 If the wall is considered as adiabatic, the
heat exchange occurs only on the ground
10
side and it is consequently less efficient.
5
This is conservative with respect to the
other configurations. The peak and steady
0 state values of heat exchange obtained for
0 200 400 600 800 1000 1200
Time,t[day] this conservative condition are collected in
Figure 6 – External air and inlet temperature. Table 3.
Table 2. Concrete and fluid properties. 30
Inlet
Adiabaticwall
Concrete Water 25
Twall=18°C
Hydraulic í16 Twall=Tsoil=14°C
k [m/s] 10 -
Temperature,T[°C]

conductivity 20

Heat ȡc 15
2.2 4.2
capacity [MJ/m3/K]
10
Thermal
Ȝ [W/m/K] 2.3 0.65
conductivity
5

5. RESULTS 0
0 200 400 600 800 1000 1200

The results related to the energy Time,t[day]

performance of the system and the Figure 7 – Outlet temperature.

136
 100
Table 4 – Exchanged heat with ground water
80
flow of 1.5 m/day.
60
Steady Steady Steady
Exchangedheat,Q[W/m]

40 Peak Peak Peak

State State State
20

0
W/m W/m W/m2 W/m2 kW kW

Ͳ20 Winter 98.8 70.4 39.5 28.1 104.1 74.2

Ͳ40 Summer 123.5 90.4 49.4 36.2 130.1 95.3
Ͳ60
Adiabaticwall
Ͳ80 Twall=18°C
Twall=Tsoil=14°C
5.2. Soil temperature variation
Ͳ100
0 200 400 600 800 1000 1200
Figure 9 shows the evolution of the
Time,t[day] ground temperature 5 m far from the wall
Figure 8 – Exchanged heat. at different depths. The first year
The first two columns express these represents the situation before the
values per meter of wall depth, while the activation of the geothermal plant,
second and third ones per unit wall followed by three years of heating-cooling
surface. The last two columns indicate the mode. Up to about 5.5 m depth (point A in
values of kW that could be Figure 5) the ground temperature is
extracted/injected by the activation of the affected by the external air temperature
whole diaphragm wall according to Figure fluctuation (see first year). The activation
1, i.e. including 68 panels activated as the of the geothermal system induces a
one considered in the numerical analysis. variation of temperature in both the points
It has to be noticed that these analyses A and B of +/- 1.5 °C with respect to the
assume no ground water flow. From first year of simulation. The point C at 20
previous studies, it is known that the m depth is not affected by the thermal
ground water flow in Torino is about 1.5 activation of the wall.
Externalair
m/day. In order to consider also this 30 activationofthe
geothermal system
A(5.5mdepth)
B(13mdepth)
aspect an additional simulation was C(20mdepth)
25
performed, assuming the ground water
flow in the perpendicular direction to the
Temperature,T[°C]

20
wall panel. The obtained results are
collected in Table 4 and clearly show that 15

the energy efficiency of the system is

significantly improved. Considering that a 10

new Italian residential building has an 5

annual energy need for heating of about
50 kWh/m2 [16] and assuming that the 0

plant would work in steady state for 1800 0 200 400 600 800 1000
Time,t[day]
1200 1400 1600

h/year in winter mode, it can be concluded Figure 9 – Soil temperature in the ground.
that the proposed system could cover the
heating demand of between 9 and 38 6. CONCLUSION
apartments of 70 m2, depending on the
presence of the ground water flow. Equipping the diaphragm walls of the
considered underground car park as
Table 3 – Exchanged heat with no ground
water flow.
energy walls could cover the heating need
of up to 38 apartments of 70 m2
Peak
Steady
Peak
Steady
Peak
Steady considering Torino underground
State State State
conditions. It has to be noticed that this
W/m W/m W/m2 W/m2 kW kW figure is based on conservative
Winter 51.0 17.2 20.4 6.9 53.8 18.1 assumptions. The energy efficiency could
Summer 63.0 24.0 25.2 9.6 66.4 25.3 be improved by equipping also the
basement slab (Figure 1). The induced

137
underground temperature variation was
found to be in acceptable limits.
ACKNOWLEDGEMENTS
The work was part of the Feasibility study
ENERWALL, funded by the Regione
Piemonte (Polo di Innovazione Enermhy)
in collaboration between Politecnico di
Torino, Resolving srl and Teknema
Progetti srl and coordinated by Marco
Barla, adjunct professor at the Politecnico
di Torino.
REFERENCES
[1] Laloui L and Di Donna A. Energy
geostructures: innovation in underground
engineering. ISTE Ltd and John Wiley &
Sons Inc.; 2013. 320 p.
[2] Bourne-Webb PJ, da Costa
Goncalves RA and Bodas Freitas TM.
Retaining walls as heat exchangers: a
numerical study. Proceedings of the XVI
ECSMGE Geotechnical Engineering for
Infrastructure and Development. 2015. p.
2499–504.
[3] Xia C, Sun M, Zhang G, Xiao S and
Zou Y. Experimental study on
geothermal heat exchangers buried in
diaphragm walls. Energy and Buildings.
Elsevier B.V.; 2012;52:50–5.
[4] Sun M, Xia C and Zhang G. Heat
transfer model and design method for
geothermal heat exchange tubes in
diaphragm walls. Energy and Buildings.
Elsevier B.V.; 2013;61:250–9.
[5] Sterpi D, Angelotti A, Corti D and
Ramus M. Numerical analysis of heat
transfer in thermo-active diaphragm
walls. Numerical methods in
geotechnical engineering. 2014;1043–8.
[6] Di Donna A, Cecinato F, Loveridge F
and Barla M. Energy performance of
diaphragm walls used as heat
exchangers. Geotechnical Engineering.
2016;(under revision).
[7] Amis T, Robinson C and Wong S.
Integrating Geothermal Loops into the
Diaphragm Walls of the Knightsbridge
Palace Hotel Project. Proceeding of
EMAP-Basements and Underground
138
Structures,. 2010;10.
[8] Soga K, Qi, H., Rui Y and Nicholson
D. Some considerations for designing
GSHP coupled geotechnical structures
based on a case study. 7th International
Congress onEnvironmental Geotechnics.
Melbourne, Australia; 2014.
[9] Brandl H. Energy foundations and
other thermo-active ground structures.
Géotechnique. 2006;56(2):81–122.
[10] Bottino G and Civita M. Engineering
geological features and mapping of
subsurface in the metropolitan area of
Turin, North Italy. 5th International IAEG
Congress. Buenos Aires, Argentina;
1986. p. 1741–53.
[11] Barla M and Barla G. Torino subsoil
characterisation by combining site
investigations and numerical modelling.
Geomechanics and Tunnelling.
2012;5:214–31.
[12] Barla G, Barla M, Bonini M,
Debernardi D, Perino A, Antolini F and
M. G. 3D thermo-hydro modeling and
real-time monitoring for a
geothermalsystem in Torino, Italy. Proc
of the XVI ECSMGE Geotechnical
Engineeringfor Infrastructure and
Development. Edinburgh; 2015. p. 2481–
6.
[13] Kürten S. Use of geothermal energy
with thermo-active seal panels.
Geotechnical Engineering: New
Horizons. 2011;
[14] Zhang G, Xia C, Sun M, Zou Y and
Xiao S. A new model and analytical
solution for the heat conduction of tunnel
lining ground heat exchangers. Cold
Regions Science and Technology.
Elsevier B.V.; 2013;88:59–66.
[15] Nicholson D, Chen Q, de Silva M,
Winter A and Winterling R. The design of
thermal tunnel energy segments for
Crossrail, UK. Proceedings of the
Institution of Civil Engineers -
Engineering Sustainability.
2014;167(ES3):118–34.
[16] Corrado V, Ballarini I and Corgnati SP.
National scientific report on the TABULA
activities in Italy. 2012.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Use of the Hardening Soil Model for urban tunnels design

Nicolas GILLERON1,2*
1
Université Paris-Est, IFSTTAR, COSYS, Marne-La-Vallée, FRANCE
2
Egis, Paris, FRANCE

ABSTRACT
The Hardening Soil Model introduced in the early 2000’s has been widespread in the geotechnical
engineering practices as well as in several finite element packages. It is now currently used to
design all types of geotechnical structures. We put forward some useful points of vigilance for the
geotechnical engineers in the use of this model for studying urban tunnels.

Keywords: Tunnel, Settlement, Finite Element, Constitutive model

structures: the Hardening Soil Model. This

1. INTRODUCTION
model shares some characteristics with
The constitutive model adopted for the soil others: that’s why some points discussed
in numerical simulations aiming at hereafter can be extended to other
predicting deformations under service models. First we introduce the main
conditions has a strong influence on the mechanisms, and then we discuss their
results. Taking into account the consequences on the maximum
constructive methods and the geology, the settlement and the width of the settlement
settlement induced by shallow tunnelling is trough. We finally discuss the impact of a
highly dependent on the constitutive variant of the HSM: the HSsmall. All the
model. From linear elastic constitutive modelling have been made with PLAXIS
models to more complicated ones, 2D 2015.
numerous formulations have been
proposed in the last fifty years. 2. 2D MODELLING OF URBAN
Nevertheless, it is generally admitted that TUNNELS WITH THE HSM
numerical model tends to predict too wide
2.1. The Hardening Soil Model
settlement troughs (ITA, 2007) and
therefore to minimize differential The Hardening Soil Model (HSM) has
settlements on building. This paper been introduced by (Schanz et al., 1999).
presents a critical discussion on the It has been integrated in PLAXIS since
constitutive model most commonly used more than ten years and in other
by geotechnical engineers to design softwares more recently. PLAXIS, ZSOIL
and FLAC advise the use of the HSM for
* presenting author
tunnelling modelling. In particular Several parameters have hand in more
(PLAXIS 2D, 2015) highlights the than one mechanism, which makes it
complementarities between mechanisms difficult to interpret their role in each
which allows to treat a wide domain of soil mechanism and predict the result of a
(graves, sands, silts and overconsolidated sensitivity analysis.
clays) as a lot of geotechnical cases
2.2. Definition of a reference case
(foundations, excavations, tunnels, dams
or embankments). We consider, in plane strain, a circular
tunnel with a diameter of 10 m, the axis
Regarding the tunnelling modelling and depth being 20 m, in a homogeneous soil
the settlement prediction, a lot of case in drained conditions. The mechanical
studies in the literature are based on this parameters, typical for a sand of the Paris,
model ((Janin et al., 2015) or (Möller and
are defined in Table 1 for the HSM and
Vermeer, 2008) for example). Table 2 for a linear elasticity with Mohr-
Furthermore, this model is used by French Coulomb failure criterion. The stress
design firms to design urban tunnels. dependent stiffness mechanism has been
The HSM is an isotropic elastoplastic deactivated by chosen a value of 0 for the
constitutive model with a double m parameter. The soil unit weight is equal
hardening. It constitutes a synthesis of to 20 kN/m3. The lateral extension of the
several models developed in the second mesh is equal to 45 m and the model
half of the twentieth century such as depth is equal to 35 m. The tunnelling
(Duncan and Chang, 1970) or the Cam- process is simulated with the stress
Clay (Potts and Zdrazkovic, 2001). relaxation method, which is the most used
Hereafter is the list of the main for settlements predictions (Wedekin et al.,
mechanism as described in (Plaxis, 2015): 2012). We increase the stress relaxation
- Plastic straining due to primary factor O by successive steps until the
compression (compression calculation no longer converges. We didn’t
hardening) model the lining influence, because the
- Plastic straining due to primary largest part of the settlements is created
deviatoric loading (shear before the lining is built.
hardening) and hyperbolic stress-
strain relationship Table 1. HSM parameters
- Elastic unloading/reloading K0 c ij \ Eurref
- Failure according to the Mohr- (-) (kPa) (deg) (deg) (MPa)
Coulomb failure criterion
0,7 10 25 5 100
- Dependency of the stiffness with
ref ref
the stress following a power law E50 =Eoed Rf Ȟur m pref
It considers three elastic stiffness (MPa) (-) (-) (-) (kPa)
parameters: the unloading/reloading 33 0,9 0.2 0 100
stiffness Eurref, the secant stiffness in
standard drained triaxial test E50ref, the
tangent stiffness for primary oedometer Table 2. Linear elastic parameters
loading Eoedref. The other parameters are K0 c ij \ E Ȟ
the failure ratio Rf, Poisson’s ratio for
(-) (kPa) (deg) (deg) (MPa) (-)
unloading/reloading Qur, a reference value
for stresses pref, the power for stress-level 0,7 10 25 5 100 0.2
dependency m and the three parameters
of Mohr-Coulomb failure criterion: the
cohesion c, the angle of internal friction M, Figure 1 describes the initial state of
the angle of dilatancy \. soil regarding the two hardening
mechanisms. The plain straight line, with

140
“+” symbols, shows the initial stress state
at different depths. The two yield loci are
drawn for a point located at the depth of
the tunnel axis (20 m). Since the angle of
internal friction is equal to 30°, K0NC is
equal to 0.5 (following Jaky’s formula as
recommended). K0NC is lower than K0
meaning the soil is lightly
overconsolidated. The initial stress state is
not located on the initial shear yield locus;
on the other hand, it is located on the
initial compression yield locus.
400
350
Deviatoric stress q
At 10 % of stress relaxation, almost the
totality of soil is concerned by the
compression hardening. This results from
the fact that the initial stress state is on the
cap yield locus, so that a slight increase of
mean or deviatoric stress is enough. At
the crown and at the invert the soil
remains elastic because the deviatoric
stress and the mean stress are
decreasing.
At 30 % of stress relaxation, the shear
hardening has begun at sidewall. It is
combined with the compression
hardening. This hardening is associated
MCfailureline with a strong acceleration in maximum

300
K0 NC
settlement (+ 3.2 between 0 to 10 % and
250 + 5.8 between 20 to 30 %).
200 K0 At 50 % of stress relaxation, there is
150
20
25 still an elastic zone under the invert and
100 15 above the crown. From the sidewall the
10
50
5 shear hardening spreads towards the
0 0 surface in an “ear” shape.
0 100 200 300 400 500
(a) Mean effective stress p' After 65 % of stress relaxation, the
400 Mohr-Coulomb failure criterion is reached
350 on the side wall. The shear hardening is
Deviatoric stress q

300 now also developing above the tunnel.

250
Compression
200 Shear Table 3. Settlement trough profile
hardening
Hardening
150
100
v O (%) 10 20 30 40
50 Elastic behaviour Smax (mm) 3.2 7.2 13.0 21.6
0
0 100 200 300 400 500 'Smax (mm) +3.2 +4.8 +5.8 +8.5
(b) Mean effective stress p'
K 0.70 0.69 0.69 0.68
Figure 1. Yield loci at initial stress state in
(p’,q) plane
O (%) 50 60 65
In Table 3, the maximum settlement and
the trough width parameter K are given for Smax (mm) 35.6 62.8 89.0
each relaxation step. K is defined as the 'Smax (mm) +14.0 +17.3 +26.2
ratio between the distance where
K 0.63 0.60 0.56
settlement is equal to 60 % and the depth.
The maximum settlement is highly non-
linear and the settlement trough is wider
than the recommended value of 0.35 for
sand by (O’Reilly and New, 1982).
Figure 2 shows the area where plastic
strains occur and the type of hardening for
four values of the relaxation factor: 10 %,
30 %, 50 %, 65 %.

141
(a)

White: Elastic behavior

Light gray: Compression hardening
Gray: Shear hardening
Dark gray: Comp. + shear hardening
Black: Failure
Figure 2. Plastic area for different values of
the stress relaxation factor (a) 10% (b) 30 %
(c) 50 % (c) 65 %

(b)
3. DISCUSSION
3.1. Compression hardening
The initial stress state is located on the
yield locus by construction of the model. A
slight increase of the hardening parameter
peq by an increase of shear or mean stress
activates this mechanism (eq. (1)).
q²
p eq  ( p ' )² (1)
M²
In the case discussed here, plastic
straining starts from the first step of
relaxation. The settlements calculated with
(c) the HSM are more than twice larger than
those calculated with the elastic perfectly
plastic model (Table 4). The settlement
trough is also wider with the HSM. Indeed,
strains are higher where the mechanism is
activated, on the sidewall. Near the crown
and the invert, the soil has an elastic
behavior, associated with the
unloading/reloading stiffness which is
three times higher than the tangent
stiffness at the early stage of the plastic
regime.
The model permits to consider a
(d) previous overburden pressure. With a very
small overburden pressure of 50 kPa,
deformations are divided by two, very
close to those obtained in linear elasticity
(Table 4).
Deformations at this step are small
compared to next steps and final ones.
The effect of volumetric hardening is no
more visible when the stress relaxation
increases and the shear hardening
becomes more significant.

142
 Table 4. Influence of volumetric hardening stiffness by derivation of the hyperbolic
Smax K formula (eq. (6)):
1
(mm) § dH · q
Et ¨¨ 1 ¸¸ Ei (  1)² (6)
Hardening Soil Model 3.2 0.70 © dq ¹ qa
Linear elasticity 1.5 0.625 The initial tangent stiffness is equal to
25 MPa, four times lower than the elastic
HSM + overburden 1.5 0.625 unloading/reloading stiffness, and
pressure of 50 kPa decreases with the increase of deviatoric
stress until 5 MPa when q is equal to 80 %
of qf. In soils where the initial pressure at
3.2. Shear hardening
rest is low, the initial deviatoric stress high,
The shear hardening mechanism has a great care has to be managed when
been built to reproduce the hyperbolic defining parameter E50. It is not equal to
relation introduced by (Duncan and the tangent stiffness and the tangent
Chang, 1970) between the axial strain H 1 stiffness will be also very different from the
and the deviatoric stress q in the context unloading/reloading stiffness.
of a triaxial test as: It is remarkable that this formulation
qa q involves significant variations of stiffness
H1 (2)
in the soil but with little consequences on
Ei q a  q
the width of the settlement trough, even if
The initial tangent stiffness Ei is given by: it modifies significantly the maximum
2 E50 settlement.
Ei (3)
2  Rf
and the asymptotic deviatoric stress by: qf
400 Ei Et
qf
qq (4)
Rf 300
q (kPa)

where qf, is the failure stress:

2 q*
q f (c cos M  V ' 3 sin M ) (5)
1  sin M
E50 is the secant stiffness at 50 % of the
deviatoric stress qf (when q=qf/2,
H 1 q / E 50 ).
It is possible to get an estimation of the
tangent stiffness at the activation of the
shear hardening, when the yield locus is
reached at 20 m depth. Figure 3 shows
the hyperbolic shape and the result of a
simulation of a triaxial test using HSM with
PLAXIS. We highlight the initial value of
the deviatoric stress qi, 120 kPa, and the
deviatoric stress q*, 169 kPa, when the
yield locus is reached. The failure stress qf
is equal to 441 kPa and initial tangent
stiffness Ei is equal to 60 MPa. The
numerical and theoretical curves are
sufficiently close to evaluate the tangent
200
qi
100 HSM
Eur Hyperbola
0
0 2 4 H 1 (%) 6
Figure 3. Triaxial test - HSM and hyperbola
3.3. Failure mechanism
The shear hardening limits the increase of
the deviatoric stress close to the tunnel.
Consequently, the Mohr-Coulomb failure
criterion is reached for larger values of the
stress relaxation factor. The criterion is
reached at 60 % with the HSM and at
35 % with the linear elasticity with Mohr-
Coulomb failure criterion (Figure 4). Near
the crown, since the behavior is mainly
elastic, the HSM has no significant impact
(Figure 5).

143
 The plastic strains have no significant 3.5. Power law stiffness dependency
consequences on the settlement trough Until this point, the power law stress
width. dependency has been deactivated by
setting the parameter m of the model to
400 zero. Here we discuss the influence of this
LE + MC MC mechanism and why it leads to difficulties
300 HSM 35 in the shallow tunnel context.
50
ı1 -ı3 (kPa)

20
Following classical geotechnical
200 65 considerations, the model aims at
40
60 20 K0
expressing the dependency between the
100
0 soil stiffness and the confining stress. The
three moduli of the HSM are concerned
0 with the type of dependency, hereafter
0 100 200 300 400 500 written for the unloading/reloading
(ı1 +ı3 )/2 (kPa)
stiffness (eq.(7)):
Figure 4. Stress path at the sidewall
m
400 ª c. cos M  V 3 sin M º
LE + MC MC E ur E ref
« » (7)
¬« c. cos M  p ref sin M ¼»
ur

300 HSM
ı1 -ı3 (kPa)

The unloading/reloading stiffness and the

200
65
100 40 0
20
0 The parameter pref gives the stress level
0 100 200 300
(ı1 +ı3 )/2 (kPa)
Figure 5. Stress path near the crown
3.4. Unloading/reloading modulus
The unloading/reloading stiffness is at
least equal to the tangent initial modulus
Ei to permit the calculation of the shear
hardening parameter.
Two zones are concerned by
unloading/reloading. A small one is
located just above the tunnel, the second
one under the raft.
The presence of a stiff zone above the
crown does not imply a lower level of
settlement. The settlements are mainly
due to the shear hardening starting from
the sidewall. Under the raft the
unloading/reloading mechanism permits to
avoid the uplift of the tunnel which occurs
with linear elastic law, which is very
relevant. This remark is also one strong
argument to use HSM for designing
shafts.
shear stiffness depend of the minor
K0 principal stress V3; the oedometric
stiffness depends of the major principal
stress (at least at the initial state).
400 500 for which the modulus is equal to its
reference value. PLAXIS recommends pref
equal to 100 kPa. This value brings to at
initial stiffness equal to the reference value
at a depth between 5 and 10 m depending
on K0. The power law parameter m is
advised between 0.5 and 1. The
parameters c and M have also an
influence. If the cohesion is very high, or if
the internal friction ratio is low, the
modulus is equal to the reference value.
This means that the stiffness profile
with stress is highly dependent from the
failure parameters. We cannot discuss
their role independently. We could easily
manage the dependence with stress with
a power law with more simple coefficients.
Furthermore the dependency with the
minor principal stress is redundant with
the shear hardening in the case of shallow
tunnel. Indeed in the vicinity of the
excavation the increase of deviatoric
stress is due to the decrease of minor
principal stress. So both mechanism are
concerned and both leads to a reduction
of stiffness.

144
 On the settlement trough width, it is
noticed that an increase of stiffness with
depth, or confining stress, leads to
narrower settlement trough. Yet this effect
is not sufficient to obtain trough as narrow
as the empirical ones.
3.6. HSsmall version
The Hardening Soil Model with small-
strain stiffness (HSsmall) has been
introduced by (Benz, 2007). It permits to
take into account a higher stiffness for
very low strains.
To test this model, we followed the
values of (Hejazy et al., 2008) to define
the two new parameters, the initial shear
modulus G0 and the shear strain level J0.7.
G0 is equal to 1.5 times the Gur value
65 MPa and J0.7 is equal to 10-4.
At 10% of relaxation, the maximum
settlement is reduced (Smax is equal to
2.6 mm against 3.2 mm) and the
settlement trough is slightly narrower
(K = 0.68 instead of 0.70). At this stage
the strains are globally low in soil and the
secant stiffness is higher, from 10 m from
the sidewall thanks to HSsmall (Figure 6).
From 30 % of relaxation the impact is
negligible on both the maximum
settlement and the trough width.
The model does not converge at 65 %,
this formulation is less stable numerically.
With much higher values of G0, such as
those adopted by (Möller & Vermeer,
2008) the same effect occurs qualitatively
for higher values of the relaxation factor.
Nevertheless this formulation has two
main difficulties for an engineering use in
shallow tunnel. The determination of the
two parameters is very uncertain and even
if the low strain have undoubtedly a role in
the maximum settlement, the settlement
have to be managed when strains are
significant namely when HSsmall has no
impact.
145
Figure 6. Isovalues of GS / Gur after 10 % of
stress relaxation
4. CONCLUSIONS
The formulation of Hardening Soil Model
has several consequences for the study of
settlements induced by urban tunnelling.
The compression hardening is activated
as soon as relaxation forces are applied
even if the solicitation is mainly shear. The
shear hardening conducts to a tangent
modulus much lower than the input
parameters E50 and Eur and to highly non-
linear displacements regarding the stress
relaxation factor. With this constitutive
model, the failure according to Mohr-
Coulomb criterion is also delayed. The
formulation of the dependency of stiffness
with stress with a power law has a
complex influence which is difficult to
anticipate on stiffness. The HSsmall does
not improve significantly the shape of the
settlement trough and it seems difficult to
define the additional parameters.
It could be interesting to propose a
constitutive model which avoids having too
much interdependence between
mechanisms. The more relevant aspects
of HSM could be used for the study of
urban tunnelling: decrease in stiffness with
increase in shear stress, taking into
account unloading/reloading stiffness and
the dependency of stiffness on confining
level.
 formulation and verification”. Beyond
ACKNOWLEDGEMENTS 2000 in Computational Geotechnics,
Thanks to Emmanuel Bourgeois from Balkema, 281-290, Rotterdam.
IFSTTAR and Adrien SAITTA from EGIS Wedekin, V., Kastner, R., Guilloux, A.,
for their support and advices. Thanks to Bezuijen, A., Standing, J., & Negro Jr,
the members of the NEWTUN project A. (2012). “Urban tunnels in soft ground
funded by the Fond Unique : Review of current design practice”.
Interministériel. Geotechnical Aspects of Underground
Construction in Soft Ground. London:
REFERENCES Taylor & Francis Group.1047-1064 eds:
Benz, T.(2007) “Small-strain stiffness of Viggiani
soils its numerical consequences”, Phd
Thesis, Universität Stuttgart.
Duncan J. M., Chang C. Y. (1970).
“Nonlinear analysis of stress and strain
in soil” Journal of the Soil Mechanics
and Foundation Division, vol. 96, n°5,
1629-1653.
ITA, (2007) “Settlements induced by
tunneling in Soft Ground”. Tunnelling
and Underground Space Technology,
vol. 22, 119-149.
Janin J., Dias D., Emeriault F., Kastner R.,
Le Bissonnais H., Guilloux A. (2015).
“Numerical back-analysis of the
southern Toulon tunnel measurements :
A comparison of 3D and 2D
approaches”. Engineering Geology,
vol. 195, 42-52.
Hejazi, Y., Dias, D., Kastner, R. (2008)
“Impact of constitutive models on the
numerical analysis of underground
constructions”. Acta Geotechnica, vol.3,
251-258.
Möller S.C., Vermeer P.A. (2008). “On
numerical simulation of tunnel
installation” Tunnelling and
Underground Space Technology,
vol.23, 461-475.
O’Reilly M.P., New B.M. (1982).
“Settlements above tunnels in the
United Kingdom – their magnitude and
Prediction”. Tunnelling, vol. 82, 173-
181.
PLAXIS 2D (2015). “Material Models
Manual”.
Potts D. M., Zdravkovic L. (2001) “Finite
element analysis in geotechnical
engineering : application”. Thomas
Telford, London.
Schanz T., Vermeer P.A., Bonnier, B.G.
(1999) “The hardening soil model:
146
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Investigations on energy sheet pile walls for the exploitation of

renewable thermal energy

David KOPPMANN, M.Sc.1

1
RWTH Aachen University, Chair of Geotechnical Engineering, Aachen, GERMANY

ABSTRACT
Energy sheet pile walls offer the possibility to exploit thermal energy stored in the ground and
especially in open waters. In an ongoing research project the potential of energy sheet pile walls is
investigated. As a first step numerical simulations have been performed to analyse the impact of a
water flow on the heat extraction rate. In this context the groundwater flow was identified as one of
the most important influence parameters together with the thermal conductivity of the structure.
Furthermore, a large scale laboratory test station has been planned with the technical monitoring to
enable the measurement of heat extraction rates of energy sheet pile walls for various boundary
conditions, which is also presented in this paper. Another objective of the tests is to establish
additional influencing parameters and to serve as a basis to develop a numerical design tool.

Keywords: geothermal, energy sheet pile wall, heat extraction laboratory test

The achievable heat extraction rate is

1. INTRODUCTION
Due to the thermal properties and the
prevailing temperature regime, open
waters have a large renewable energy
potential. In contrast to the ground the
energy potential of open waters still
remains almost unexploited. Energy sheet
pile walls produced by the company SPS
Energy GmbH offer the opportunity for the
future exploitation of energy stored in
open waters. Within the scope of an
ongoing research project, characteristic
values for the heat extraction rate of
energy sheet pile walls will be determined.
Furthermore, a suitable numerical tool for
the dimensioning of energy sheet pile
walls will be developed based on the
results of large scale laboratory tests.
* presenting author
significantly affected by existing water flow
in either the groundwater, the open water
or both. For this reason these factors have
to be considered in the laboratory tests, as
well as in the numerical tool. In this paper,
the impact of a water flow on the heat
extraction rate of plane energy
geostructures is demonstrated by
numerical simulations. Moreover the large
scale test station is described in detail
together with the measurement technology
which will be used to establish a relation
between the heat extraction rate and its
most influencing parameters.
2. ENERGY SHEET PILE WALLS
2.1. Principle
The incorporation of heat exchanger pipes
in foundation piles, ground slabs and
diaphragm or basement walls is
increasingly used for the exploitation of
geothermal energy in the last years. The
so called “energy geostructures” take
advantage of statically required structures
for the thermal utilization of the ground.
Basically every structural element with a
direct contact to the ground can be used
as an energy geostructure. Plane
elements are especially advantageous
because of their lager surface enabling a
better heat transfer. The energy sheet pile
walls rank among these plane energy
geostructures.
2.2. Energy sheet pile walls
Energy sheet pile walls consist of
conventional sheet pile profiles which are
equipped with heat extraction pipes
(Figure 1).
inflow
return flow
sheet pile wall
heat extraction pipe
Figure 1: Principle of an energy sheet pile wall
Because of the sealing effect, a broad
application area opens for the
implementation of energy sheet pile walls
for heat extraction e.g. at riversides or port
facilities. Depending on the area of
application, energy sheet pile walls allow
the exploitation of energy stored at
shallow depths and particularly in open
waters (Figure 2). Results of previous
small scale test performed at a river by the
SPS Energy GmbH have shown that heat
extraction rates of 800 W/m² can be
achieved (Puttke, 2013).
2.3. Application at open waters
The exploitation of heat from water
requires large surface areas with a high
148
thermal conductivity (Puttke, 2013). As a
matter of fact, sheet pile walls possess
these characteristics. The thermal
conductivity of steel (52 W/(mK)) is
significantly higher compared to the
conductivities of concrete (2.3 W/(mK)) or
Polyethylene (0.4 W/(mK)) used as pipe
material in conventional ground heat
exchangers. Furthermore, sheet pile walls
are usually installed in great quantities at
coastal infrastructures. By virtue of the
modular design of sheet pile walls great
heat exchange surfaces can be
constructed in these areas. The aesthetics
of these structures will not be affected,
because the heat exchanger pipes are
located at the side facing the ground.
v = 0 - 1,0 m/d convection &
conduction
(groundwater, soil)
v = 0 - 1,0 m/d
convection &
v = 0,01 - 6,0 m/s
conduction (open
water)
convection &
conduction
(groundwater, soil)
convection &
conduction v = 0 - 1,0 m/d
(groundwater, soil)
Figure 2: Areas of application and occurring
heat transfer mechanisms
3. EFFECT OF GROUNDWATER FLOW
3.1. General
The mathematical description of the heat
transfer occurring in the ground includes
phenomena such as transfer mechanisms
(i.e. conduction, convection, dispersion) or
heat sources. Among these conduction
and convection may be identified as the
most relevant mechanisms for the heat
transfer in the ground. If a groundwater
flow is present the heat transfer by forced
convection ensures a permanent supply of
warm water masses. Thereby the energy
potential of the ground increases which is
why the heat transfer capacity of energy
geostructures is significantly influenced by
a water flow. The impact of the
groundwater depends on the flow velocity
and the flow direction in relation to the
geostructure. For these reasons plane
energy geostructures should be installed
with a flow-parallel orientation
(Koppmann et. al., 2014). Energy sheet
pile walls constructed at open waters can
take advantage of these flow conditions.
However, in contrast to groundwater
conditions, in open waters there is a
continuous water flow initiated by wind,
tidal movement or a natural gradient.
3.2. Calculation Approach
At the Chair of Geotechnical Engineering
a suitable calculation approach for the
dimensioning of plane energy
geostructures, in particular for thermo-
active seal panels, was developed. In this
approach the geostructure itself is
modelled as a multi-layer system with the
help of thermal resistances. This allows to
consider decisive parameters like the pipe
flow as well as geometric properties of the
pipe system and the geostructure. By
implementing this approach into the finite-
differences software SHEMAT-Suite
(Rath et. al. 2006, Clauser 2003,) the soil
and groundwater conditions can be taken
into account. Overall, all decisive
parameters and their interactions are
considered in one model enabling a
realistic simulation of plane energy
geostructures. More details to this
approach can be found in Kürten (2014).
Due to the universal multi-layer modeling,
this calculation approach is also suitable
for plane geostructures like energy
diaphragm walls.
In contrast to diaphragm walls energy
sheet pile walls differ in their geometrical
configuration, especially in their small
thickness. Therefore adjustments to the
calculation approach have to be made and
validated on the basis of measurement
results. Nevertheless, the calculation
approach can be used first for a qualitative
estimation of the potential of energy sheet
pile walls in comparison to common
energy diaphragm walls.
3.3. Numerical simulations
In order to determine the influence of a
water flow on the heat extraction rate
numerous numerical simulations have
been carried out using the aforementioned
calculation approach. The following results
are based on a 3D-model of a sand body
149
centrally divided by a diaphragm wall of
0.5 m thickness. This wall is equipped with
five u-shaped heat extraction pipes on the
right side of the wall under a concrete
cover of 0.05 m operating in series
connection. As heat carrier fluid, water
was used with a constant inflow
temperature of 2.5 °C. The groundwater
velocities were simulated independently
on the left (vf,l) and on the right (vf,r) side of
the wall. To determine the heat extraction
rate, the simulation was carried out until
either a steady state or a time period of
180 days was reached. To identify the
impact of a steel structure like an energy
sheet pile wall the thermal properties have
been adjusted in additional simulations.
The used thermal properties are given in
Table 1.
Table 1. Thermal properties of wall and sand
thermal
2.3 W/(mK)
conductivity
concrete
wall thermal
conductivity of 0.4 W/(mK)
pipes
thermal
52 W/(mK)
conductivity
steel
wall thermal
conductivity of 52 W/(mK)
pipes
thermal
2.5 W/(mK)
sand conductivity
(10 °C) thermal
2.7 MJ/(m³K)
capacity
Figure 3 shows the heat extraction rate for
different groundwater flow velocities (Pi)
divided by the maximum heat extraction
rate without groundwater flow (P0). A
distinction is made between an identical
velocity on both sides of the wall (vf,l = vf,r)
and a change of velocity on only one side
of the wall. This figure shows that the heat
extraction rate can increase up to 80 %
due to a groundwater flow. Both
diaphragm and steel wall show that the
heat extraction rate increases enormously
for velocities smaller than 0.25 m/d.
However, this increment reduces
considerably and remains almost constant
for velocities greater than 0.5 m/d. In case should be adapted to the groundwater
of a high groundwater velocity the conditions.
temperature in the contact area of the wall Although, the simulation of open water
reaches a nearly constant value. and the corresponding high flow velocities
Therefore, the influence of the is not yet possible with SHEMAT-Suite, a
groundwater flow on the heat extraction first approximation can be made by
rate decreases while the thermal replacing the thermal ground properties by
conductivity of the geostructure becomes those of water (table 2).
more relevant. Table 2: Thermal properties of water
2
thermal
1,9 0.582 W/(mK)
water conductivity
1,8
1,7
(10 °C) thermal
4.2 MJ/(m³K)
1,6 capacity
Pi / P0 [-]

1,5 Figure 4 compares the normalized heat

1,4 extraction rate (Pi/P0) of two different
1,3 systems for different flow velocities. The
1,2 first system has sand on both sides and
1,1 the second one has sand on one side and
1
water on the other side of the wall.
0 0,25 0,5 0,75 1 2
groundwater flow velocity vf [m/d]
1,9
diaphragm wall vf,l = vf,r
1,8
diaphragm wall vf,l = 0
diaphragm wall vf,r = 0 1,7
steel wall vf,l = vf,r 1,6
Pi / P0 [-]

steel wall vf,l = 0

1,5
Figure 3: Heat extraction rate versus 1,4
groundwater flow velocity
1,3
Although in Figure 3 the diaphragm wall 1,2
shows a stronger increase than the steel 1,1
wall, the absolute heat extraction rates of
1
the steel wall in this model are about 2.5 0 0,5 1 1,5 2 2,5 3
times higher. Hence, the smaller increase groundwater vf / open water vw flow velocity [m/d]
of the heat extraction rate of the steel wall steel wall vf,l = vf,r (sand/sand)
must account for the great energy steel wall vf,l = vf,w (sand/water)
withdrawal causing a faster cooling of the Figure 4: Heat extraction rate versus water
ground. For this reason the water flow flow velocity due to different contact material
velocity gets even more relevant for the
Figure 4 shows that for the second system
heat extraction rate of energy sheet pile
(sand/water) there is an even stronger
walls because of the faster supply of warm
increase of the heat extraction rate
water masses. This is why energy sheet
compared to the first system (sand/sand).
pile walls benefit from high flow velocities
This is due to the higher thermal capacity
at open waters.
enabling the storage of large amounts of
Figure 3 also shows that as a result of energy in water.
different velocities on both sides of the
Taking everything into account, the
wall, the heat extraction rate varies.
thermal capacity of water leads to a great
Naturally the groundwater flow velocity on
energy potential and the influence of
the side of the heat extraction pipes is of
convective heat transfer has to be
greater influence. Therefore the
considered when dimensioning energy
construction of plane energy geostructures
geostructures.
150
4. LABORATORY TEST STATION
4.1. General
To determine the efficiency of energy
sheet pile walls and to identify important
influencing parameters on the heat
extraction rate, laboratory test will be
carried out in an ongoing research project.
To simulate a realistic application, an
existing large scale test station is rebuilt
lower 2.0 meters. Every loop has a total
length of about 4.2 m. The distance
between the in and return flow pipe is
0.25 m. The space on the side of the
pipes is filled with sand, while the other
side is filled with water reflecting the
installation of an energy sheet pile wall at
a riverside.
groundwater steel grid
Tsoil
cycle with fleece

and adapted to the requirements for heat

50
Qsoil

extraction tests with energy sheet pile

1550
walls. water sand
container energy sheet
4.2. Important influencing parameters circular
pile wall

3200
sealing
The maximum heat extraction rate of curcular Tsheet pile wall
isolation
energy sheet pile walls is affected by the

1550
steel
heat transport mechanisms occurring in frame
water

the geostructure, the soil and the water.

The thermal properties controlling these

50
Qwater

processes are in turn affected by the open water Twater

temperature. Furthermore, the numerical 50
150
cycle
2800 150
50 [mm]

simulations showed the influence of the 3200

flow velocity on both sides of the wall. cooler pump temperature sensor

Hence, next to the temperature the flow Figure 5: Schematic view of the test station
velocities in soil and water must be
variably adjustable and measureable. The
heat extraction rate is additionally
influenced by geometric and operational heat carrier fluid T
cycle in Tout Tin Tout Tin Tout Tin Tout
parameters. While the geometric energy sheet
parameters are defined by the selected pile wall Qpipe Qpipe Qpipe Qpipe
steel grid
sheet pile profile, the operating with fleece
parameters such as fluid temperature and water
container
flow rate are variable and must be circular
measured during the laboratory tests. sealing 2500

Consequently, in the following sections curcular

isolation
the focus is centered on the setting of the steel
frame
flow on both sides of the wall as well as
the technical monitoring of the laboratory
tests. 50 50

4.3. Geometry 150 2800 150

The test station is installed in a 2.5 m cooler pump flow meter

deep waterproof and circular isolated
basin with a base area of 3.2 m x 3.2 m. A
scheme of the complete test station and
the installed measuring equipment is
shown in Figure 5 and Figure 6. The
energy sheet pile wall consists of four cap
profiles of type HP 140S, which are
installed in the middle of the test station.
Each of the profiles is equipped with u-
shaped steel heat extraction pipes in the
151
temperature sensor shut-off valve
Figure 6: Schematic section of the test station
The groundwater flow in the sand body
and the flow in the open water can be
adjusted by two separate water cycles.
The groundwater flow velocity is set by
different hydraulic potentials in lateral
water containers. The open water flow
velocity is set by the delivery rate of the
pump in the open water cycle. The
groundwater velocity can be calculated
using Darcy’s law via the hydraulic
gradient and the sands coefficient of
permeability. The open water velocity can
in turn easily be determined by the
delivery rate of the pump in the water
cycle Qwater and the cross-section. The
flow rate of the heat carrier fluid in the
extraction pipes is regulated by a third
cycle. The four loops can either be
operated in a parallel or in series
connection with flow rates between 100 l/h
and 2000 l/h.
4.4. Technical monitoring
During the laboratory test, the temperature
in the inflow and return flow pipes, the
sand, the open water and at the sheet pile
wall is measured by
thermometers Pt100 with an accuracy of
1/10 DIN B (DIN EN 60751). They have a
precision of about ǻT = ± 0.03 °C allowing
a highly accurate temperature
measurement. In order to determine the
flow rate of the heat carrier fluid Qpipe
every loop is provided with a magnetic
inductive flow meter installed in the return
flow. The specific heat transfer capacity P
[W/m²] of the single loops can
subsequently be calculated
equation 1.
P (Tout  Tin ) ˜ c v , w ˜ Q pipe / A wall
Where cv,w is the volumetric heat capacity
and Awall is the heat extraction surface of
the single loops.
5. CONCLUSIONS AND OUTLOOK
In numerical simulations the impact of a
groundwater or open water flow on the
heat extraction rate of plane energy
geostructures has been investigated. Thus
it has been identified that the water flow
velocity is causing a significant increase of
the heat extraction rate. For this reason
the water flow has to be considered when
dimensioning energy geostructures.
Further investigations showed the
influence of a steel structure. Although the
influence of the water flow velocity is
smaller, the absolute heat extraction rate
is higher than with a diaphragm wall due
to the higher thermal conductivity of steel.
To investigate the efficiency of energy
sheet pile walls a large scale test station
has been designed. It allows the control
and the accurate measurement of all
influence parameters. By future test series
characteristic values for the heat
extraction rate will be determined.
Furthermore, the measurement results will
be used to validate necessary adaptions
to the calculation approach.
6. ACKNOWLEDGEMENTS
The research project is funded by the
Federal Ministry for Economic Affairs and
Energy (BMWi) in Germany. Furthermore
the author would like to thank the DGGT
resistance Deutsche Gesellschaft für Geotechnik e.V.
for the financial support.
REFERENCES
Clauser, C. (Ed.) (2003) „Numerical
simulation of reactive flow in hot
aquifers using SHEMAT and
processing SHEMAT”, Springer.
DIN EN 60751:2009-05. (2009) “Industrial
platinum resistance thermometers and
platinum temperature sensors”.
with
Koppmann, D., Kürten, S., Mottaghy, D.
Ziegler, M. (2014) „Auslegung und
(1) Berechnung von Thermo-aktiven
Bauteilen bei einer vorhandenen
Grundwasserströmung, BBR
Jahresmagazin, 12-2014, p. 64-69.
Kürten, S. (2014) „Zur thermischen
Nutzung des Untergrunds mit flächigen
thermo-aktiven Bauteilen”, Dissertation,
Faculty of Civil Engineering, RWTH
Aachen University.
Puttke, B. (2013) “Costal Waters as
profitable Renewable Energy Source”,
Coastal & Marine, Vol. 22(1), p. 16.
Rath, V., Wolf, A., Bücker, M. (2006) „Joint
three-dimensional inversion of coupled
groundwater flow and heat transfer
based on automatic differntiation:
sensitvity calcuclation, verification and
synthetic examples“, Geophyscial
Journal International 167(1), p. 453-
466.
152
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Wallslotrobot: a new method of deep braced excavation in

compact spaces

Sahin MONSEREZ

DENYS NV, R&D Department, Ghent, BELGIUM

ABSTRACT
This paper outlines the first development stages of Wallslotrobot: a new method of deep
braced excavation. Wallslotrobot is a technological response to the established manual technique
called “beschoeide sleuven” (BE) or “fouilles blindées” (FR) which is geographically limited to the
Benelux countries. The result is a prototype machine which is basically a vertical tunnel boring
machine. The device consists of a ground cutter that is pushed vertically into the ground by a
jacking system. The shaft lining consists of hollow rectangular precast concrete segments.
At present the technical feasibility of Wallslotrobot has been successfully demonstrated. The
prototype machine and lining segments have been tested on the field. Three 15m deep adjacent
trenches were excavated from a 2,5m diameter tunnel. In response to this positive evaluation, the
joint venture Future Foundations was set up to commercialize Wallslotrobot by 2019

Keywords: deep braced excavation, deep sheeted trenching, deep foundations

1. INTRODUCTION A particular example where this

technique was used is the Schuman-
Deep braced excavation or deep sheeted
Josaphat rail tunnel (2008-2012), during
trenching is a construction technique to
which the idea for mechanizing the
build deep reinforced concrete walls in
technique arose. From 2011 until 2013 the
dry, soft soil conditions. These are
general conception of the idea was
manually created by specialized
developed and named Wallslotrobot.
personnel, using a shovel, bucket and
pulley system. This process is slow and Subsequently a nationally funded
workers are exposed to significant safety project was set up to demonstrate the
risks. technical feasibility of constructing these
deep sheeted trenches mechanically and
Its field of application is mainly situated
with an economical advantage compared
on construction sites where there is little
to the existing method. The main project
clear working height and space available,
objective was to design, build a prototype
like tunnels or basements. Thus where
machine and demonstrate it in a realistic
other competing techniques as slurry
operating environment.
walls, soilmix or secant piling are not
applicable.
* presenting author
 In June 2015 the first field tests were
conducted, the overall test program
consisted of the excavation of three 15m
deep adjacent trenches from a 2.5m
diameter concrete pipe anchored at
ground level.
2. TRADITIONAL METHOD OF DEEP
BRACED EXCAVATION
2.1. Description of execution
Deep braced excavations occur typically in
adjacent slots of 3m in length, dimensions
range over 15m in depth and between 1m
and 1.5m in width. The trench is
sequentially dug out, by shovel in soft soil
or pneumatic tools in case of hard
discrepancies, over a height of 0.4-0.5m.
The soil is removed by a bucket and pulley
system installed at the top of the trench.
Hereafter the exposed soil is sheeted by
precast concrete elements which are fixed
in place by horizontal struts at predefined
distances, see Figure 1.
Figure 1: Deep braced excavation by hand
Once the desired depth is reached,
rebar is put in place. Especially where little
clear working height is available (<3m) this
is done manually by workers located
inside the trench (see Figure 2). To
conclude the trenches are poured with
concrete.
Figure 2: Rebarred trenches
154
Due to the complexity of this specialized
technique it is mainly used and
established in the Benelux countries:
Belgium, the Netherlands and Luxemburg.
2.2. Case study: Schuman-Josaphat
project
To illustrate the practical use of deep
braced excavations the case study of the
Schuman-Josaphat rail tunnel (Brussels:
2008-2012) is briefly described next.
The Schuman-Josaphat rail tunnel is
a mined tunnel and junction to an
operational line beneath an avenue,
residential and commercial buildings.
Because of the low soil cover of only 3-4m
normal TBM driving was not applicable. In
addition the environmental consequences
of cut-and-cover were unacceptable.
Therefore it was decided to adopt
sequential excavation combining,
microtunneling with traditional deep
braced excavations and compensation
grouting to construct a tunnel envelope as
described next and presented in Figure 3.
1. Construction of an access pit
(6mx6m) with pipe jacking galleries.
2. Horizontal pipe jacking (DN3000)
aligned with the top corners of the
tunnel.
3. Deep braced excavations (20m
depth), forming the tunnel walls.
4. Transverse pipe jacking, forming
the tunnel roof.
Following the completion of the
boundary structure (concrete filled pipe-
roof and reinforced concrete walls) the
intermediate soil is removed and the
concrete floor is installed.
Figure 3: Josaphat Schuman rail tunnel
construction
3. WALLSLOTROBOT CONCEPT
3.1. Idea
During the construction of the
Schuman-Josaphat rail tunnel over
10,000m³ of deep braced excavations
were performed. The execution of these
excavations was determinative for the total
project duration because of the slow rate
of progress and shortage in specialized
personnel.
Because these excavations were
highly repetitive, the idea arose to
mechanize this technique. This would not
only be beneficial for the project duration
and cost but also a higher overall safety
and better ergonomic working conditions Figure 4: Wallslotrobot concept
for the workers would be achieved.
When the desired trench depth is
3.2. Design goals reached the soil cutter is recovered and
Following general design goals for rebar cages are installed, whereafter the
the Wallslotrobot development were trench is poured with concrete. Multiple
predefined: interlocked adjacent trenches eventually
x Excavation time: 5x faster
form a reinforced concrete wall.
x Maximal depth: 30m The method is patented and
published with number EP2535462B1 on
x Width: 1.200mm
the ninth of March 2016.
x Min. clear working height: Ø2.5m
x Geology: silt layers on top and Tertiary 4. PHASE 1: TECHNICAL FEASIBILITY
dense sand below, including
sandstone concretions. Situated above 4.1. Goal
ground-water level. Following the conception stage a
x Safety: minimal human presence in the nationally funded IWT project was set up
trench in order to assess the technical feasibility
x Cost effectivity: competitive with (2014-2015). The final goal was to
manual method construct three adjacent 15m deep
trenches in a realistic operating
x Quality: excellent water tightness and
high surface finishing.
environment and relevant geology. To
achieve this goal a prototype machine and
3.3. Conception concrete sheeting element had to be
The general conception was designed and built, with a predominant
developed during 2011-2013 according to focus on technical feasibility and a minor
the design goals described in 3.2. focus on process speed and/or
The result is a method, illustrated by automation. This design process included
Figure 4 , in which the trench is formed by multiple iterations.
pressing precast concrete sheeting 4.2. Test setting
elements downwards while loosening the The test setting is discussed first
soil at the base by means of a soil cutter. since the design of the prototype machine
The removal of loosened soil from the is partly based on this.
trench is done by a grab.

155
Testsite geology
As Brussels is the primary market for
this new technique it was chosen to obtain
a test location with a similar geology as for
the Schuman-Josaphat project. Thus a
silty and clayey Quaternary top layer with
underneath sandy Tertiary layers
commonly classified as Brusseliaan. It
typically contains sandstone concretions
and possible decalcified zones.
The testing station of the Belgium
Building research Institute at Limelette is
an ideal location with a comparable
geology. Clayey tertiary layers down to 8m
below ground level with underneath the
sandy Brusseliaan. A CPT diagram and
soil classification is given by Figure 5.
Figure 5: Soil classification and CPT (C3-
BBRI) at the test location for slot 1
In most of the CPT’s that were
available, sandstone concretions were
encountered between 12m-14m below
ground level. Taking this information into
account it was decided to perform tests to
a depth of 15m below ground level in
order to maximize the probability of
encountering the sandstone concretions.
The groundwater level is located well
below this depth so no additional
measures for drainage were necessary.
156
Test setup
To demonstrate the Wallslotrobot
method it was essential to create a
realistic setting. In this case four aligned
concrete pipes, with an internal diameter
of 2.5m and a total length of 11.5m, were
anchored at ground level on a bed of
stabilized sand. Four anchors were
installed on either side of the pipe at an
angle of 17° away from the longitudinal
symmetry plane in order to minimise soil
disturbance prior to excavation. A side-
and cross view of the test setup is given
by Figure 6 and Figure 7 respectively.
Figure 6: Side view of the test setup
Figure 7: Cross section of the test setup
After installation, the anchors were
connected to a steel plate mantle resting
on top of the pipes and prestressed to fix
the pipe in place. Subsequently a 1.3m by
10m slot was cut out and strutted in the
bottom of the pipe to allow passage of the
machine. Lastly rails were installed to
facilitate horizontal transport of the
machine itself, sheeting elements, and
excavated soil.
4.3. Wallslotrobot prototype
The design of the prototype machine
is subdivided in three main components:
the press unit, the ground cutter and a
method for vertical soil removal. Figure 8
illustrates the prototype assembly in
operation, here the mechanism for soil
removal is not displayed since multiple
systems were designed and tested.
Figure 8: Assembly of the prototype machine
in operation
Press unit
The press unit can translate in the
longitudinal axis of the pipe on the rails.
When positioned on a desired location the
press is fixed and aligned in the pipe by a
rail locking system. Four hydraulic
cylinders are installed in order to press the
concrete sheeting elements downwards.
The reaction forces are transferred to the
pipe roof via the reaction frame.
The presence of underground
cavities due to decalcification for example
has been taken into account in the design.
To anticipate on this loss of frontal
resistance, possibly resulting
uncontrolled descent, the ground cutter,
the sheeting elements and the press are
continuously interconnected mechanically.
This link is broken only during the
placement of a new sheeting element,
therefore a static retention system is
installed to temporarily hold the concrete
shaft and the cutter.
Ground Cutter
The ground cutter is designed to
minimize frontal resistance and loosen the
soil to facilitate vertical removal. In the first
stage the soil needs to be transported
from below the cutting drums to above
them. This is achieved by the combination
of the rotation of the cutting drums and the
presence of soil scrapers as displayed on
Figure 9. In addition the cutter needs to be
capable to cut through local sandstone
concretions (UCS <50MPa).
Figure 9: Cross section of the ground cutter
The cutter itself consists of six
identical modules, each equipped with two
hydraulic powered cutting drums, which
are mounted on the cutting shoe. An
overcutting value of 3mm was empirically
determined, based on the following
considerations:
x A minimal overcutting is needed to
allow for steering.
x Settlements need to be within an
acceptable range, resulting in an upper
limit for the value of overcutting.
x When the overcutting value is higher,
soil relaxation is higher resulting in a
decrease of lateral earth pressure on
the sheeting. This has two advantages:
a decrease in thrust forces due to
in lower wall friction and a decrease in
strength specifications of the sheeting
elements.
In order to anticipate on deviations in
verticality each cutting drum is designed to
be individually adjustable in rotational
speed. By adjusting the relative speeds of
specific drums one can theoretically
provoke a steering motion in any direction.
157
 Eventually, once the trench has the
desired depth, the cutter is disassembled.
Ideally this is done automatically which is
not yet incorporated in this prototype
design. On disassembly the cutting shoe
is left behind, functioning as sheeting.
Method for vertical soil removal
Regardless of the grabbing concept,
multiple options for vertical soil removal
were considered. Eventually it was chosen
to develop and test two independent
methods: the original grabbing method
and soil removal by air transport as
presented in Figure 10. The airflow is
invoked by means of a rental vacuum
suction excavator.
Figure 10: Soil removal by air transport,
airflow represented by arrows
Process control and instrumentation
The prototype machine is controlled
via an operating console which allows
control of diverse process parameters and
observation of process data registered by
multiple measurement sensors. The most
important sensors and their function are
listed in Table 1.
158
Table 1: Installed measurement sensors
1. Press: cylinder pressure sensors (4x)
ĺ Total thrust force
2. Press: cylinder stroke sensor (1x)
ĺ Thrust frame position, depth and
progress rate
3. Press: inclinometers (2x)
ĺ Yaw and pitch of the thrust frame
4. Cutter: inclinometers (2x)
ĺ Yaw and pitch of the cutter
5. Cutter: force sensors (4x)
ĺ Front pressure and wall friction
6. Cutter: drum speed sensors (12x)
ĺ Rotational speeds
7. Cutter: drum pressure sensors (12x)
ĺ Cutting torque
8. Cutter: cameras (3x)
ĺ Process observation
9. General: Target laser (1x)
ĺ Depth and lateral drift of the cutter
4.4. Sheeting element prototype
The main functionality of the
sheeting element consists in retaining the
soil, thus preventing the trench from
collapsing during construction. Because
these elements are left behind their
function is temporary and therefore a
minimal production cost is a major design
condition. This resulted in the use of
hollow prefabricated concrete blocks as
sheeting elements.
Design load
To assess the technical feasibility it
was necessary to determine a “worst
case” design value for the lateral earth
pressure. The design assumption was
made to use the standard lateral earth
pressure envelope for braced excavations
according to Terzaghi and Peck (1967).
However, due to a difference in execution
and the presence of an annular space in
contrast with standard braced excavations
there is no 1-to-1 relation. Finally the
design load was set at 300kPa uniform
pressure, for a worst case scenario of 30m
deep trenches in dry, fine grained soils
with low cohesion.
Geometrical design Destructive testing
In Figure 11 the prototype sheeting The sheeting element is designed for
element is presented. The length and a uniform lateral pressure of 300kPa.
width of the elements are based on the However, the hypothesis of uniformity will
current practice. The height was chosen in in practice only occur for very soft soils
function of the available working height (e.g. peat) and hydrostatic pressures.
and was set to 0.75m for this project. Instead a favorable 3D soil arching
mechanism will emerge horizontally
between the struts, the size of the arching
effect is dependent on the soil properties,
depth, strut center distance and the
deformation of the sheeting (Piaskowski &
Kowalewski, 1965).
To assess the influence of the
arching effect, the sheeting elements were
subjected to real scale destructive testing
in the Magnel laboratory for concrete
research. Two test setups were built, one
to simulate uniform loading where the
element is placed on a mortar bed and
subjected to five equal line loads (Figure
Figure 11: Concrete sheeting element 12a). For the other setup the element was
placed on a 0.90m sand bed where it is
The H-shape results in an optimal force pushed against the bed by introducing
distribution and a low wall thickness. In forces via the struts (Figure 12b).
addition there is no need for decoupling
power lines for the cutter every time a new
sheeting element is placed. Furthermore
the shape contributes to a better water
tightness and makes structural continuity
of adjacent slots possible.
The sheeting elements are equipped with
a tooth-groove connection to ensure
alignment of neighboring trenches. The
elements themselves are vertically
centered and interconnected by self-
closing dowels used in the tunneling
segment industry. Vertical pressure forces
are transmitted through wooden slats.
The polystyrene insert at the end face has
a double functionality. On the one hand it
retains the soil during excavation. On the
other hand it divides adjacent slots, acting
as formwork for the concrete in the final
construction stage, whereafter it is
removed by cutting. In a later design stage
the polystyrene inserts will be replaced by
a more sustainable and practical solution.

Figure 12: Real scale test setup for (a)

uniform pressure simulation and (b) soil
arching simulation
159
 The results showed that the arching
effect in highly compacted sands has a
significant advantage in force distribution.
Because the maximal thrust force of the
setup was reached before structural failure
was observed, the exact advantage could
not be quantified. However, it can be
concluded that the deflection capacity of
the sheeting elements is adequate for
arching to occur. A minimal load capacity
increase of 45% was observed, which is in
line with the 3D-arching theory of
Piaskowski & Kowalewski, 1965.
4.5. In situ testing
In total the test program consisted of
five stages: three test stages (A,C,E) and
two design iteration stages (B,D) as
presented by Figure 13. The total program
had a duration of six months starting from
June until November 2015. A detailed
description of the test geology and test
setup can be found in paragraph 4.2.
Figure 13: Test program Wallslotrobot
Before testing, the following operational
risks were identified:
x Need for excessively high thrust force.
x Insufficient steering capacity to ensure
verticality.
x Inefficient vertical soil removal,
resulting in too low productivity
x Inefficient soil cutting mechanism,
resulting in low progress rates.
x The capability to cut through
sandstone concretions.
x Bad alignment of adjacent trenches.
Evaluation of the thrust force
The necessary thrust force was well
below the expected value’s which were
estimated according to EC7, pile design.
An explanation lies in the presence of the
annular space, the occurrence of soil
arching and the short execution time,
160
which all contribute to a lower lateral earth
pressure and thus a lower wall friction.
Moreover the thrust force tends to be
quasi constant, regardless of the
execution depth. This means that the
increase in dead weight of the sheeting
tends to be in equilibrium with the increase
in shaft friction. Thus the lateral earth
pressure doesn’t seem to increase at
greater depths, which corresponds to the
theory of arching according to Piaskowski
& Kowalewski, 1965.
Ground cutter performance
The excavation effectiveness of the
ground cutter was well above expectation
in soft soils. It literally digs itself in even at
very low rotational speeds. Peak progress
rates of 48 mm/min were recorded during
testing. When encountering sandstone
concretions progress rate dropped, but the
cutter eventually got through. It was
concluded that by optimizing the teeth
configuration and tooth type, the efficiency
can be increased significantly.
Effectivity of vertical soil removal
The performance of the air transport
system was very variable, depending on
soil conditions. In course grained soils a
good productivity was achieved. However,
for fine grained, slightly cohesive soils
multiple problems were encountered. The
main issue was clogging of cohesive
particles, obstructing the horizontal suction
channel in the cutter. During the first
design iteration, improvements were made
to the geometry of the suction channel
and air inlets were installed to add
compressed air. This improved the
efficiency but not enough, the clogging
problem was displaced to the vertical
suction tube (Figure 14).
Figure 14: Free 10’’ suction tube (a), clogged
suction tube (b)
 The grabbing mechanism was tested
in stage C. The performance was good
and small cycle times could be achieved.
Unfortunately, the tests were aborted
prematurely due to insufficient torque of
the drum cutters. In comparison to air
transportation the soil is stacked above
the cutter and is funneled from the sides to
the center, requiring a higher torque.
It can be concluded that the method
for vertical soil removal determines the
overall productivity. Air transport is only
viable for course grained soils. However,
the power requirement and installation
size are a major downside. The grabbing
mechanism is efficient and independent of
soil conditions. However, the ground cutter
design needs to be attuned with this
system of vertical soil transport.
Steering performance
Due to the high effectivity of the
ground cutter, there was too little frontal
resistance to induce a steering motion by
adjusting the cutting drum speeds. Most of
the time the frontal resistance was even
negative, indicating that the ground cutter
hung on the shaft. The verticality of the
trenches is given by Graph 1.
Graph 1: Verticality of trenches T1, T2 & T3
The second and third trench (T2 and
T3) were equipped with a steering
compartment with four cylinders, situated
above the cutter. By retracting or
extending these cylinders one creates an
angle between the cutter and the
161
sheeting, introducing a steering motion.
However, there are still teething problems
to overcome since steering is a delicate
operation which effect is dependent on
local soil conditions.
A good alignment was achieved
between adjacent trenches T2 and T3,
which were not placed directly next to T1
because of the excessive longitudinal
deflection of T1.
Evaluation of the Sheeting elements
As stated, there was little to no
frontal resistance. This in combination with
a low wall friction lead to higher tensile
forces in the sheeting dowels than
anticipated. During in situ testing these
connections broke, leading to uncontrolled
descent (Figure 15). This was fixed by
installing screw plates. The top view of T1
is given in Figure 16.
Figure 15: Disconnection of sheeting dowels
Figure 16: Final result of T1, top view
General performance
The technical feasibility has been
successfully demonstrated. Production
rate was doubled for each trench, starting
from 20 days for T1 down to 5 days for T3.
The design goal of quintupling the
excavation time is certainly feasible.
5. PHASE 2: COMMERCIALIZATION
5.1. Goal
Future Foundations, a joint venture
between Denys and the Participation
Company Flanders (PMV) has been set
up to commercialize Wallslotrobot. This
phase consists of a three year action plan
for increasing the technology readiness
level (TRL) from 6 to 9 (Figure 17) to
finally bring Wallslotrobot on the market.
Figure 17: Technology readiness levels (TRL)
In the first year, which started on
January 2016, a beta version of the
Wallslotrobot will be developed and tested
(TRL 7). In this stage the focus lies on
increasing the efficiency, productivity and
overall cost effectiveness.
The second and third year will focus
on further fine-tuning of the technique,
standardisation (TRL8), marketing and
actual execution of Wallslotrobot trenches
in a first commercial project (TRL9). In
parallel new applications will be developed
for Wallslotrobot including trench
excavation below groundwater level.
5.2. Market and applications
The main market for the
Wallslotrobot lies in underground project
development in dense urban areas. Here
surface area is getting really scarce and
the socio-economic impact of open-pit
techniques is no longer sustainable.
Future applications for Wallslotrobot
are: tunnels (e.g. Schuman-Josaphat),
parkings, stations, multi-functional spaces
including shopping malls, cinema’s,
nightlife activities and sports infrastructure
(e.g. the AMFORA project Amsterdam),
and many more.
162
6. CONCLUSION
In this paper, a new method of deep
braced excavation, Wallslotrobot, is
presented. The conduct and result of the
first development phase, i.e. the
assessment of the technical feasibility, are
described. It can be concluded that
Wallslotrobot method is technically viable
and that a fivefold productivity increase in
comparison with the manual method is
achievable. The joint venture Future
foundations is set up to commercialize
Wallslotrobot by 2019.
7. ACKNOWLEDGEMENTS
The author wishes to acknowledge
the technical support of his colleagues at
Denys, with in particular A. Janssens and
K. Van Royen for their significant
contributions and dedication.
The support and advice of the
technical bureau of construction (Seco),
Durabuild materials (Ghent University) and
the Belgian building research institute
(BBRI) is greatly appreciated.
The author gratefully acknowledges
the creativity, expertise and advice of the
originators of the Wallslotrobot concept J.
Van Wassenhove and P. Afschrift.
The Wallslotrobot feasibility project
was financially supported by the Belgian
agency for innovation by science and
technology (IWT).
8. REFERENCES
Terzaghi, K. & Peck, R.B. (1967) ”Soil
Mechanics in Engineering Practice,
Second Edition”, John Wiley & Sons,
New York.
Piaskowski, A. & Kowalewski, Z. (1965)
“Application of Thixotropic Clay
Suspensions for Stability of Vertical
Sides of Deep Trenches without
Strutting”. Proc. of 6th ISSMFE, Vol.II,
pp. 526-529, Montreal.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Dual cell shaft for a 19.5m deep excavation.

Design, execution and monitoring

Cristian RADU
1
Technical University of Civil Engineering Bucharest, Department of Geotechnical and
Foundation Engineering / SBR Soletanche Bachy FundaĠii – Bucharest, ROMANIA

ABSTRACT
FREELO is the first entertainment resort in Romania and eastern Europe that will host an
aerodynamic wind tunnel, comprised of a 17m high superstructure and an underground structure,
laid at a depth of -19.50m below natural ground level. In order to ensure a proper retaining system
for the excavation, several aspects had to be regarded and studied from the process of concept
design to monitoring works and data interpretation.
The paper presents the case history of an unique deep excavation retaining system, a dual cell
shaft comprised of two circular diaphragm walls disposed around the future structure. Special
technological aspects that generate geotechnical risk and issues from correlating structural and
geotechnical engineering will also be treated.

Keywords: circular, analysis, deep excavation, monitoring

which will be embedded in two rectangular

1. INTRODUCTION
FREELO is an ambitious and innovative
concept of entertainment developed in
Bucharest, Romania, with a flagship
attraction comprised of an indoor
aerodynamic wind tunnel. The equipment
which are part of the tunnel’s ensemble
will be incorporated in the interior of an
underground concrete structure laid at
approximately 19.50m below natural
ground level. The necessity of such depth
imposes the need for an excavation
retaining system, adapted to the structural
particularities and to the geotechnical
conditions of the site.
The specific structural configuration is
conditioned by the size of the equipment
* presenting author
concrete cells with 1.20m (or 0.80m) thick
walls, unsupported on their entire depth
during the exploitation phase of the
structure, with no horizontal bracings that
will support lateral earth pressures, aspect
which have an important influence on the
construction staging and works.
Taking into consideration all the
particularities of the project, usual
solutions adopted for retaining systems
could not have been considered due to a
lack of performance in costs and
productivity. Thus, the design and
configuration of the retaining system
became a difficult engineering challenge
because of the depth, which exceeds with
50-80% common deep excavations in
 Bucharest and because of the necessity to
adopt an alternative solution, rarely used
or documented in Romania.
+17.0m
NGL
Ͳ19.50m
Figure 1. Architectural concept, longitudinal
section
2. GEOTECHNICAL CONDITIONS
The geotechnical report available for the
site indicates an uniform alternation of soil
layers, as following:
±0.00m ... -2.95m – fillings, silty sand;
-2.95m ... -13.00m – sand and gravel;
-13.00m ... -14.00m – clayey silt;
-14.00m ... -16.00m – sand and gravel;
-16.00 m ... -25.00 m – stiff silty clay;
-24.00 m... -31.00 m – sandy silt.
The hydrogeology of the site is
characterized by on-site dewatering works
and pumping tests. Based on the available
information, two groundwater levels have
been identified, the first one having its
piezometric level at approximately -4.00m
below ground level. The inferior water
level, captive in the sandy silt situated
below 24.00m depth, is under pressure
and raises to -6.00m, which is translated
in 19m of water pressure. The cover of the
second aquifer, the stiff silty clay, is a soil
layer with low permeability which isolates
communication between the two water
tables.
164
The Romanian norm NP 047-2014 “Norm
regarding geotechnical documentations
for constructions” describes the project’s
geotechnical risk as major, characterized
with 19/22 points - 6 points for ground
conditions, 4 points for dewatering works,
3 points for seismicity, 3 points for
neighbouring structures and 3 points for
structure importance.
3. CONCEPT DESIGN OF THE
CIRCULAR SHAFT
3.1. General information
Due to the project’s major geotechnical
risk, many key aspects had to be regarded
in the design of the retaining system.
Thus, multiple solutions for supporting the
excavation have been studied, such as
the “top-down” method, prestressed
ground anchors or horizontal struts, none
of the options being sufficiently performant
or being difficult to implement.
The adopted and implemented solution is
represented by the disposal of circular
diaphragm walls, self-supporting, which
inscribe the geometry of the future
structure. The main advantage of this
solution is represented by the possibility of
building inside an open excavation by
avoiding the congestion of the pit with
metallic struts.
The interior structure is divided in two
rectangular areas named “the flight tube”
and “the engine tube”, which will be
inscribed in two circles with radiuses of
approximately 7.50m and 8.95m
respectively.
PERIMETRAL
WALL
D-WALL CONTOUR
CONTOUR
Figure 2. Concept of the retaining system
The two circular cells are composed of a
60cm thick diaphragm wall with the toe
level at -23.00m connected by a 80cm
section, executed at -26.00m, parallel with
the interior structure’s walls, which will be
supported on the entire depth of the
excavation with 4 (four) levels of horizontal
metallic struts.
After completing the contour diaphragm
walls, under the excavation level at the toe
of the 80cm thick d-wall, concrete
barrettes were executed in order to ensure
a supplementary rigidity of the system and
for soil improvement purposes. Also, at an
intermediate level in the excavation depth,
a perimetral concrete beam has been cast
with the purpose of distributing non-
uniform loads that may appear, being
generated from surcharges caused by
equipment at ground level or discontinuity
in soil layers.
The equipment inside the “engine tube”
require a high execution precision and
minimum mounting tolerances in order to
avoid any possible dimensional conflicts.
In order to achieve this condition, the
diaphragm walls were distanced with
approximately 10cm from the interior
structure, distance that represents the
total deviations that can occur from the
execution’s verticality and the horizontal
displacements of the retaining system.
Based on all the presented particularities,
the design of the structural elements has
been elaborated for the ultimate limit
states STR and GEO with respect to the
provisions of Design Approach 1 from EN
1997-1:2004.
Because of the depth of the excavation
and the presence of the inferior under
pressure aquifer, it is necessary to
attribute a special attention to
supplementary checks as UPL for the
interior concrete structure and HYD in
order to protect the bottom of the
excavation against hydraulic lift and to
properly dimension the dewatering
system.
Regarding the staging of the construction
works, excavation follows an usual
sequence. In the case of building the
165
interior structure, a correlation is required
between the removal of the metallic struts
and executing the structure in separate
sections through a correct detailing of the
reinforcement.
Surcharges considered in the initial design
of the excavation’s retaining system were
modified taking into consideration
technological necessities and limitations
and terrain adjusting around the
excavation. These limitations imposed a
rigorous study, a conclusive example
being represented by the crane position
next to the pit, limited by the inclination of
the mobile lever and the distance from the
excavation.
3.2. Evaluation of lateral earth pressure
for soils in the case of circular shafts
In current geotechnical engineering,
circular shaft applications are used for
deep excavations utilized as access tubes
for TBMs, tunnels, reservoirs and tanks.
In the case of circular shafts, lateral
pressure generated by the soil massif
acting on the exterior surface of the
diaphragm wall are transmitted through
the arching effect as compression forces
(hoop forces). Based on this effect, the
dimension of the d-wall is dependent on
the maximum excavation depth that
generates the hoop stress.
The diaphragm wall’s thickness is
dependent on the diameter of each
circular cell with a direct impact on the
arching effect of the lateral loads. A higher
diameter of the circular cell generates an
amplified soil pressure, caused by the
decreasing arching effect. For large
diameter shafts, the distribution of lateral
earth pressure becomes equal to the one
obtained in plane strain design.
Circular areas of the diaphragm walls are
consisting of panels which are tangent to
the circles that will inscribe the structure.
In order to obtain a proper arching effect,
the joint alignment is mandatory and
necessary, obtained through CWS
(Continuous Water Stop) beams used as
guiding, patented by Soletanche Bachy.
Another important aspect in the design of software for geotechnical applications,
the diaphragm walls is the verticality of the Plaxis 2D, Plaxis 3D and Midas GTS.
excavation. Two adjacent panels executed An initial 2D analysis was carried out, both
deviated in opposite directions can in axisymmetric and plane strain models,
generate a lack of continuity in the in order to fully comprehend the
assumed calculated thickness, reducing differences between the two distinct
the theoretical thickness of the final sections.
diaphragm wall from 60cm to The axisymmetric model is used in cases
approximately 41cm obtained on site. where circular structures are modelled,
Considering the mentioned scenario has with uniform sections and loads distributed
an impact on the hoop force check around the central rotation axis. The basic
through comparison with the maximum
premise for this model is the identical
axial bearing capacity of the diaphragm configuration in any radial direction
wall. regarding soil lithology and loads.
Joints between panels are areas of
discontinuity in the reinforcement, where
hoop forces have an important role
against openings.
Numerous studies from technical literature
have addressed the influence of soil on
circular shafts for deep excavations,
through different means as the finite
element method (2D and 3D), the limit
equilibrium method or centrifuge
modelling.
Tests and calibrations of centrifuge
models with the finite element method
reveal that evaluating lateral pressure of Figure 3. Axisymmetric model
soils by applying traditional design
methods (for example, the Rankine
theory) generates overestimated values.
This evaluation is based on parametric
studies, the arching effect between soil
particles being predominantly dependent
on the internal friction angle and the
circular cell diameter.
Determining the real values of pressure
generated by the soil on the entire
excavation, as well as studying the
behaviour of such retaining systems is Figure 4. Plane strain model
essential for an economical dimensioning
Material parameters applied for soils
of the retaining system as well as for a
correspond to the Mohr Coulomb and the
judicious design taking into consideration
Hardening Soil constitutive models and
the geotechnical risk, implications on the
were determined based on the available
project’s success being significant.
geotechnical report and by consulting
4. STUDY OF THE ADOPTED specialty literature and information
SOLUTION available from other excavations with
similar lithologies in Bucharest.
Investigating the adopted solution is Although the 2D models can simulate both
based on numerical modelling using finite excavation and construction staging which
element method through dedicated
166
are relevant to the retaining system, as
well as stiffness and structural rigidity, in
order to fully estimate the structural
response of the retaining system, a full 3D
analysis was carried out by use of Midas
GTS NX and Plaxis 3D.
Figure 5. 3D Model – executed retaining
system
Additional studies of the retaining system
have been carried out by usual structural
analysis, with manually applied external
loads on the diaphragm wall. Although
usual structural software do not offer
several modelling capabilities (for example
soil models or realistic embedment
conditions) a global behaviour of the
retaining system can be assessed. The
study confirms the expected behaviour
obtained by means of geotechnical finite
element software and offers
supplementary design precision.
Because of its geometry, the shape of the
retaining system is governed by two
disadvantages. On the connection area,
diaphragm walls which are parallel to the
167
interior structure generate different angles
for unloading the lateral earth pressures of
the soil, transmitted as hoop forces
through the circular diaphragm walls.
The two circular shafts have different
diameters acting on the connection area
with different soil pressures. Altough the
difference between the two diameters is
not significant the excavation generates a
global rotational effect of the entire
structure, proven by measurements
obtained from monitoring the excavation.
Figure 6. 3D Model – Global deformation of
the retaining system
5. MONITORING OF THE RETAINING
STRUCTURE
The main role of monitoring geotechnical
retaining systems is to maintain a safety
factor to the project during the
construction works due to the evolution of
efforts and deformations in the soil massif
which require a constant surveillance.
Thus in-time controlling of the structure is
necessary to validate design hypotheses
considered and to follow-up on the
differences between real site conditions
and conditions taken into account in
design, especially given the particular
features of the project.
The role of monitoring is to certify the
design’s efficiency and to ensure control of
the execution process, allowing
adjustments based on real conditions
such as deviations, structural flaws or
correcting deficient execution to be
applied.
Relevant phases of monitoring for the
retaining system respect technological
stages considering modifications in
support conditions and work progress,
frequency of the measurements being
correlated with direct on-site observations
and situations occurred.
5.1. Inclinometric measurements
Measuring horizontal displacements
through the inclinometric method was
adopted by implementing 4 (four) systems,
embedded in the diaphragm walls, two
inclinometers being characteristic to each
relevant area for providing a general
overview of the retaining system’s
behaviour.
On the instrumented depth, values for
deformations (positive or negative) and Figure 8. Comparison between estimated and
manifestation direction indicate rotations measured horizontal displacements for
and evolutions of the cumulated inclinometer I3
displacements. In the case of the section supported by
horizontal struts, measurements reveal a
similar displacement trend, a notable
I1
difference being obtained in the overall
I2 values, estimated double or triple than on-
site measurements.
This aspect is commonly studied and
approached in different parametric studies
I4 I3
of similar soil conditions in Bucharest.
Advanced constitutive models require
special parameters that can offer realistic
Figure 7. Positioning of the inclinometers,
estimations for displacements.
implemented by the Technical University of
Civil Engineering of Bucharest
In regard to safety factors, by studying
values obtained for displacements,
reference values were set for several
levels: attention, warning and alarm.
Values obtained for the I3 inclinometer are
relevant to the section of d-walls
supported by metallic struts, and the
values obtained for the I4 inclinometer are
compared to the axisymmetric model.
Diagrams present composed values for
the two directions, A0-A180 and B0-B180.
A comparison between measurements for
the two inclinometers positioned on the Figure 9. Comparison between estimated and
contour of the retaining system, I3 and I4, measured horizontal displacements for
and diagrams obtained through numerical inclinometer I4
analysis indicate a similar tendency for In the case of inclinometer I4, although
displacements. measured values exceed designed
values, on-site measurements are situated
within the attention limit, considered as
acceptable. The diagrams offer a global
image of the retaining system’s behaviour.
168
By comparison, values obtained by
studying the 3D models reveal similar
values as the ones measured.
A cause for this difference may be
generated by technological limitations for
constructing diaphragm walls which
cannot be considered through the design
software. The finite element model
considers the entire soil massif as an
isotropic medium on a radial direction, as
in reality these discontinuity areas modify
the static scheme of a panel and generate
a horizontal bending moment between two
joints.
As excavation works inside the retaining
system advance, efforts and strains in the
soil massif and in the concrete elements
suffer reorientations and repositioning due
to the two joints between two consecutive
panels, which are areas of discontinuity.
These hypotheses can be considered
correct by comparing results obtained
after measuring values corresponding to
the first major excavation stage, which
reveal a difference of 1-3 mm between two
successive monitoring steps.
5.2. Strain gauges
In collaboration with Soldata Romania, in
order to determine axial forces in the
metallic elements, strain gauges were
mounted on the struts from the third level
(-13.00m) and on the forth level (-16.50m).
Instrumentedmetallicstruts
Ͳ13.00m
Ͳ16.50m
Figure 10. 3D Model – proposed instrumented
elements
The principle mechanism of the strain
gauges consists in transforming
frequencies measured by the vibrating
169
wire into deformations and efforts. The
positioning of the strain gauges was
established depending on the strut length
and layout, instrumented elements being
considered relevant to the global
behaviour of the retaining system.
Results obtained were compared with
design values generated by numerical
analysis.
Figure 11. Comparison between estimated
and measured axial forces in the strut at -
13.00m
Resulted values obtained from monitoring
the third level of struts (-13.00m) indicate
a 10% variation of stresses around the
dimensioning value (maximum design
value). Although the design value is
exceeded by 10%, due to the applied
safety coefficients, the additional load is
considered acceptable.
Figure 12. Comparison between measured
axial forces in the struts at -16.50m
For the fourth level of struts (-16.50m), a
10-15% difference between the two
instrumented elements can represent a
confirmation of the different loading
conditions of the connected area,
generated by the two different diameters
of the circular cells and can confirm the
global deformation tendency of the
retaining system.
6. CONCLUSIONS
The FREELO project is unique in the
context of Romanian geotechnical
engineering due to the limitations and
constraints encountered during the
project’s evolution.
Approaching a particular and atypical
project with rigorous study and by
consulting technical literature and other
specialists from the Soletanche Bachy
Group represented key components of the
design. However, a systematic approach
and a correlation with all entities involved
in the design and execution process
results in an important impact on the
developments’ success.
Monitoring and geotechnical investigations
represent important aspects of the project,
and back-analysis data obtained during
the project’s execution represent valuable
resources for future similar developments,
applications for circular shafts being
various.
The investors of FREELO sustain
innovative development and delivering a
successful iconic project ensures the
context for further valuable and ambitious
perspectives for Romanian geotechnical
engineering.
Figure 13. Photo taken during the execution
(photo credit KEI Development)
170
Figure 14. Photo taken at final excavation
level (photo credit KEI Development)
REFERENCES
Jaeyeon Cho, Hyunsung Lim, Sangseom
Jeong, Kyoungyul Kim, Analysis of
lateral earth pressure on a vertical
circular shaft by considering the 3D
arching effect, Tunnelling and
Underground Space Technology 48
(2015) 11–19.
Nadine Miller, Risk Analysis for a
Proposed Underground Library at MIT -
Diaphragm Wall, Shaft and Tunnel
System (B.A.Sc. Geological and
Mineral Engineering Thesis),
Department of Civil and Environmental
Engineering, University of Toronto,
2001.
Tan Rwe Yun, Numerical study of a large
diameter shaft in old alluvium (Master
Of Engineering Department Of Civil
Engineering Thesis), National
University Of Singapore, 2004.
Artemis Ioanni Theophilou, Structural
design of an underground cylindrical
shell, (Master of Engineering in Civil
and Environmental Engineering
Thesis), Massachusetts Institute of
Technology, 2001.
Hsiung, B.C.B., Tsai, Y.Y., Tsai, C.C.,
Analysis and Construction of Cross
Passage of Delhi Metro, Indian
Geotechnical Conference – 2010,
GEOtrendz, December 16–18, 2010,
IGS Mumbai Chapter & IIT Bombay.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Eurasia Tunnel Project: The Geotechnical Challenges of the

Asian Transition Box

Alexis ROSE*

WSP | Parsons Brinckerhoff, Cardiff, UK

ABSTRACT
The Asian Transition Box (ATB) is part of the Eurasia Tunnel Project and was constructed in 2013.
It is 168m in length, up to 40m deep and was the launch pit for the tunnel boring machine. The
design of the excavation support system posed some significant challenges and it proved
necessary to redesign the system once excavation had begun following unexpected ground
movements and the trigger of amber alert levels. The redefined parameters were an extension to
the traditional guidance for the locally well-known Trakya Formation and required new descriptions
and parameters to those that are usually used. The new approach required robust justification to
allow variation from conventional practice. This paper gives an overview of the Eurasia Tunnel
Project, with a detailed focus on the ATB construction and the redesign work undertaken.

Keywords: Deep excavations, Plaxis 2D, geotechnical parameters

partnership of YapÕ Merkezi and SK

1. INTRODUCTION
The Istanbul Strait Road Tube Crossing
Project, also known as the ‘Eurasia Tunnel
Project’ is under construction between the
Asian and European sides of Istanbul. The
project is being constructed along a
14.6km route, comprising 5.4km of twin-
deck tunnels and 9.2km of approach
roads. The principal aim of the project is to
relieve Istanbul’s transcontinental traffic
pressure, but the project is also driven by
economic and environmental benefits.
The Build-Operate-Transfer contract for
the project was awarded to the Turkish-
Korean JV, ATAù, formed through the
Engineering and Construction.
This paper focuses on a specific aspect
of the project, the Asian Transition Box
(ATB), which is where the 13.7m diameter
tunnel boring machine (TBM) was
launched from and also where the tunnel
form transitions from TBM to mined tunnel
(NATM). Parsons Brinckerhoff were the
Designers for the ATB and IGT
Muhendislik, based in Istanbul, were their
Sub-designers.
The ATB is located less than 70m from
the edge of the Bosphorus; its
approximate location is shown in Figure 1.

* presenting author

Figure 1. Site location
2. DESIGN ELEMENTS
The main elements for the 14.6km route
are shown in Figure 2. The approximate
lengths of the alignment components are
as follows:
x 5.4km European approach roads
x 3.8km Asian approach roads
x 3.4km TBM tunnel
x 1.0km NATM tunnel
x 1.0km cut and cover tunnel
The maximum depth of the TBM tunnel is
106.4m below the Bosphorus surface
water level, meaning that the TBM will be
Figure 2. Eurasia Tunnel design elements
172
exposed to pressures in excess of 11bar,
the highest ever for TBM construction.
The TBM tunnel is a twin-deck
arrangement with traffic on the upper deck
travelling towards Asia and traffic on the
lower deck travelling towards Europe.
3. GEOLOGY
The geological setting is within a large
scale fault system called the North
Anatolian Fault Zone. The Marmara Fault
System is the part of this fault zone that
runs beneath Istanbul and many strong
earthquakes occur along these fault lines.
The fault map is shown in Figure 3.
The geology within the region
comprises Carboniferous and Devonian
sandstones, claystones, limestones and
greywacke. The greywacke is the principal
rock type at the location of the ATB and is
locally known as the Trakya Formation.
The Trakya is a weathered argillaceous
sandstone; deposition of these sediments
together occurs from turbidity currents or
submarine avalanches. Close to the
Bosphorus the rock is even more intensely
folded, faulted and fractured and as a
result it has highly variable strength and
stability properties.

Figure 3. Fault map
4. GROUND INVESTIGATION
Three main phases of ground
investigation were completed for the
project, between 2010 and 2012.
A total of 12 boreholes were installed
for the ATB, with an average depth of
36m. The majority of rock core was logged
as sandstone, with thick zones of
mudstone and thin zones of fault breccia,
as well as diabase (dolerite) intrusions (a
subvolcanic mafic rock).
Rock mass classification highlighted
that the rock was poor quality, but this was
not fully reflected in the derived
geotechnical parameters. There were
numerous iterations of the geotechnical
parameter table put forward for design; a
clear indication that it was challenging to
confidently assign parameters to the
materials.
The final version showed the rock
divided into four zones; fair, poor, very
poor and extremely poor, with effective
friction angle (I’) ranging between 20o and
35o and the effective cohesion (c’) ranging
between 60kPa and 160kPa.
5. ATB DESIGN
The ATB is 168m in length and up to 40m
deep. Various options were considered for
the support of excavation design, although
all centred on the use of piled walls with
prestressed ground anchor tiebacks,
which was considered to be the preferred
construction method of local contractors.
173
The option taken forward to final design
involved an upper secant piled wall and a
lower contiguous piled wall, as shown in
Figures 4 and 5. The secant piles were up
to 21m in length and the contiguous piles
were up to 25m in length, all reinforced to
full depth.
Figure 4. Pile dimensions
The ground anchors were up to 38m in
length. They were double corrosion
protection (DCP) anchors and were pre-
stressed to 60 tonnes.
It was preferred to construct the wall in
a two-phase approach because there
were concerns over buildability and the
installation of ~50m long piles into rock.
This solution would provide a stiff ground
control system and allow for flexibility in
adapting the design if necessary, via the
observational approach.
The secant piled wall was required for
the top half of the excavation, where
ground conditions and water seepage was
considered more likely to be a problem.
Accordingly, the design included inclined
drain holes within the lower part of the
secant piled wall to prevent build-up of
water pressures. A contiguous piled wall
was considered appropriate for the lower
level, where the ground conditions were
believed to be more favourable.
 Figure 6. Fault breccia

Figure 5. Interface between piled walls

The design was undertaken using

Eurocodes Design Approach 2 and
Design Approach 3. Finite Element
analysis was used to design the support of
excavation system, to allow scrutiny of the
ground stresses and deflections and the
forces in the piles and anchors. Limit
Equilibrium analysis was used to check
the overall stability.
Figure 7. Fault gouge
The contract requirements set out that
the deflection tolerances should follow the In the first week of June 2013,
guidance from Clough & O’Rourke. From monitoring data was beginning to indicate
this, it was considered that the horizontal a potential problem with the support of
deflections should be limited to 0.1-0.2% excavation. On the 13th June, unexpected
(approx.) of the excavation depth. lateral displacements were measured from
the inclinometers in the western section of
6. EXCAVATION the south wall, where excavation was at a
depth of 12m. Corresponding increases in
Excavation began in April 2013. Examples
anchor loads were also recorded and an
of the ground conditions revealed during
amber alert level was reached in one
excavation are shown Figures 6 and 7. As
instance. This raised concern over the
indicated in Figure 7, the material at some
adequacy of design and the amber alert
locations was extremely weak and clayey
triggered the temporary suspension of
and coud be excavated by hand. It was
excavation operations. The excavation
clear that in these instances, the material
was partially re-filled as a safety measure.
did not accord with the geotechnical
design parameters. Also, groundwater 7. REDESIGN (PARAMETERS)
was a problem and was observed flowing
from the excavation walls and collecting in The redesign considered the live
the base of the excavation. monitoring data in parallel with revisiting
the original geotechnical parameter

174
reports and also the core boxes obtained
during the ground investigation.
On closer inspection of the core boxes
and the borehole logs, it was noted that
there was likely to have been mis-
interpretation of the strength of the rock.
For example, a description of “moderately
strong, slightly weathered rock” was noted
in one case; the sections of rock that were
recovered were indeed moderately strong,
but, were accompanied by a TCR (total
core recovery) of 25% and an RQD (rock
quality designation) of 7%.
In this example, the strength of the
ground will be governed by the
unrecovered material, likely to be a heavily
fractured fault breccia material or a soft
clayey fault gouge material, both of which
would probably have been washed out
with the borehole flush fluid.
The site of the ATB is very close to the
Bosphorus and potentially closer to fault
zones, which demanded a change from
common practice and the introduction of a
completely weathered rock, which may not
be commonly observed.
8. REDESIGN (ANALYSIS)
Site observations showed that in places
the material had properties more akin to a
soil, and consequently the stronger
sandstone layers could not be relied upon.
The ground model and parameters were
amended accordingly, with the most
fundamental change being the
introduction of a cohesionless material.
The back analysis was undertaken
using Plaxis 2D software (Figure 8) and
aimed to replicate the observed
deflections. The Mohr-Coulomb soil model
was unable to reproduce the observed
deflections and it was proposed to use the
Hardening Soil model.
The Hardening Soil model allows stress
dependency of soil stiffness, where
different stiffnesses can be applied to
primary and unloading moduli. It was not
originally considered appropriate to use
this soil model for the initial design
because the added degree of complexity
175
could not be supported by the available
data. However, with the addition of the
monitoring data to allow model calibration,
this approach was now regarded as the
preferred solution. The analysis revealed a
failure surface behind the wall within the
free anchor length zone.
Figure 8. Plaxis model example
The design was completed in stages
and on completion of each stage the
construction drawings were detailed and
issued progressively so as to minimise
delay and standing time on site.
9. COMPLETION
Following the successful reanalysis and
revised design of the ATB, the excavation
recommenced. Work was only halted for
two weeks before the new design started
to be implemented on site. The stages of
progress are shown in Figures 9 to 11.
Final excavation depth was reached in
January 2014.
The 13.7m diameter, 120m long TBM
was launched from the Asian side of the
city in April 2014 and broke through on the
European side in August 2015.
10. CONCLUSIONS
Ground conditions encountered during
construction can often differ to those
anticipated from ground investigations or
existing knowledge. The experiences of
the ATB illustrate the real consequences
of this problem.
A far greater importance must be
placed on comprehensive GI and
understanding of complexities in data
interpretation. Similarly, communication
between engineering geologists and
geotechnical engineers, both locally and in
remote design teams, is essential whilst
formulating design parameters.
Figure 9. Progress 1st July 2013
Figure 10. Progress 4th October 2013
Figure 11. ATB completion January 2014
Effective implementation and
management of the observational
approach, along with proactive use of
monitoring data to reappraise ground
conditions and allow dynamic redesign,
was key to the success of the project.
176
ACKNOWLEDGMENTS
The Author would like to thank the
following people for their invaluable
assistance: Adrian Dolecki and Mark
Glendinning from Parsons Brinckerhoff
Cardiff, Tolga To÷an and Ray Castelli
from Parsons Brinckerhoff New York, and
AnÕl Kurban from IGT Muhendislik.
REFERENCES
Clough, G. W. and O’Rourke, T. D. (1990).
Construction induced movements of
insitu walls. ASCE Special Publication
25.
Keskin, H. B. (2008). Analysis of the
performance of retaining systems in
deep excavations in greywackes. PhD
Thesis, Bo÷aziçi University.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Back-analysis of a deep excavation in Budapest

Attila SZEPESHÁZI*

Budapest University of Technology and Economics, Department of Engineering Geology

and Geotechnics, Budapest, HUNGARY

ABSTRACT
A ~14m deep excavation was constructed in the city centre of Budapest, Hungary in 2013 by HBM
Ltd., the local subsidiary of Soletanche Bachy. The speciality of the construction project was that
the ~25m high historical facade of the existing building had to be preserved along the three side of
the jobsite. Due to that a detailed monitoring system was applied, providing a good basis for
scientific research. Using PLAXIS 2D & 3D finite element method software and GEO5 2D
subgrade reaction method software detailed back-analysis of the multi-anchored diaphragm wall
was done. This way some design aspects, which is usually conservatively neglected or simplified
in current design practice, could be evaluated. For instance the influence of small strain stiffness of
stiff clay and 3D stiffness of the diaphragm wall was examined. In addition a comparison of internal
forces of the diaphragm wall was compared calculated by different techniques.

Keywords: diaphragm wall, back-analysis, finite element analysis, small-strain

stiffness, soil-structure interaction

1. INTRODUCTION The monitoring system of the project had

two pillars: the daily monitoring of the
A ~14m deep excavation was constructed
preserved facade and the observation of
in Budapest in 2013 by HBM Ltd., the
the D-wall movement by an inclinometer.
Hungarian subsidiary of Soletanche
This study summarizes the back-analysis
Bachy. The historical facade of the former
of the diaphragm wall deformations using
press building had to be preserved while
conventional subgrade reaction
the building core was planned to be
calculations and 2D&3D PLAXIS analysis
demolished. The 4 storey deep
with sophisticated soil models.
underground car park had to be prepared
in such conditions. The 60cm thick, 16-
2. CONCEPT OF BACK-ANALYSIS
19m deep diaphragm wall was temporarily
supported by 2-3 rows of ground anchors. In local practice deep excavations are
Wall was embedded to the impermeable generally designed by modelling the soil-
clayey bedrock of Budapest to provide a structure interaction with elastic spring
dry pit for reinforced concrete construction support (Ray et al, 2010). However, with
works. increasing excavation depth the calculated
deformations of such models can be lower
* presenting author
than the realistic as the method neglect
the deformation of the active and passive
mobilized soil mass. (Szepesházi R. et al,
2009)
The release of Hardening Soil (hereinafter
referred as HS) model in PLAXIS made it
possible to make more realistic finite
element calculations even for very deep
excavations where unloading-reloading
process has a great significance.
Promising results (Kardos et al. 2014)
were published in recent years in Hungary
providing a good bases for further
analysis. Nevertheless PLAXIS
calculations where usually found
conservative by local practitioners as it
usually provides higher deformations and
bending moments than realistic.
Figure 1. The analysed excavation pit
Figure 2. The historical facade protected
The following improvements of PLAXIS
gives really good prospects for analysis of
deep excavations:
- The release of so-called hardening soil
small (hereinafter referred as HSsmall)
soil model taking into account small
strain stiffness of soils.
178
- The evolution full 3D module with more
sophisticated possibilities for modelling
of structures.
Both improvement of PLAXIS were
testified by back-analysis of measured
settlements of a sugar silo by LĘdör et al
(2016). Other software was tested by the
author (Szepesházi et al, 2016) for 3D
analysis of deep excavations previously.
The new, promising tools of PLAXIS and
the latter researches led to this study
which compares the results of the listed
calculation methods to each other and to
the inclinometer measurements.
2.1. The analysed structure
The geometry of the analysed section of
excavation can be seen in Figure 3. The
inclinometer was installed approximately
in the middle of the ~32m wide wall. The
soil investigation consisted some drillings,
lab tests for soil classification and 1 CPT
test and explored the typical stratification
of city centre in Budapest: the top ~6m is
pour fill and silty sand. Below that a ~7-8m
thick very dense sandy gravel overlay the
impermeable, stiff clay. The groundwater
is in the sandy gravel layer.
Figure 3. The analysed cross-section
2.2. Elastic spring models
The regionally widespread GEO5 software
was used to prepare calculations based
on elastic spring theory. Two methods
were used to calculate the coefficient of
subgrade reaction: diagrams of Monnet
and the Schmitt-formula. The derived
coefficients are shown in Figure 4.
2.3. FEM models the underground part of the protected
PLAXIS 2D and 3D modules were used facade, the diaphragm wall as a plate
for the calculations. Short summary of the element and the ground anchors as node-
most important model features are given to-node anchors and geogrid elements.
here. Two different mechanical model were
prepared: a simple 2D plain strain model
2.3.1. Soil model
as it was introduced in the latter paragraph
HS and the more promising HSsmall and a modified one, having a horizontal
models were used. The theoretical support in the top of the diaphragm wall to
background of the latter, its relevance in simulate the capping beams influence on
geotechnics and the suggestions for input wall behaviour.
parameter derivation can be found in Benz
Soilmodel T1ͲT2ͲT3
(2006). LayerName siSa saGr Cl
2 Toplevel mRel 0 Ͳ6,6 Ͳ13,05
c[MN/m /m] Monnet Schmitt Bottomlevel mRel Ͳ6,6 Ͳ13,05 Ͳ50
Fill/SiltySand 20 3,5 model Ͳ HS HS/HSsmall HS/HSsmall
type Ͳ Drained Drained UndraindedB
SandyGravel 50 130 ɶunsat kN/m3 19 20,5 20,5
MioceneClay 15 30 ɶsat kN/m3 20 21 21
einit Ͳ 0,8 0,45 0,55
Figure 4. The coefficients of the subgrade
E50 MN/m2 5 75 25
reaction Eoed MN/m2 5 75 25
Three soil layers were differentiated Eur MN/m2 15 300 125
conform to Figure 3. Three different model m Ͳ 0,6 0,5 1

were prepared to testify the significance of ʆUR Ͳ 0,2 0,2 0,2

c kN/m2
small-strain stiffness as shown in Figure 5. ʔ ම
6
26
1
38
Ͳ
Ͳ
The T2 model was built as a mixture of HS ʗ ම 0 8 Ͳ
and HSsmall because our expectation was su,REF kN/m2 Ͳ Ͳ 577
that the small-strain stiffness model is only su,INC
kN/m2/
m Ͳ Ͳ 10
significant in the layer were the passive
K0,x Ͳ 0,56 0,38 1,27
pressure on the retaining wall arise. pREF kN/m2 55 140 312
Modeltype T1 T2 T3 tensilestrength kN/m2 10 5 200
G0 MN/m2 48 140 300
Fill/SiltySand HS HS Hssmall
ɶo,7 Ͳ 0,00025 0,0003 0,001
SandyGravel HS HS Hssmall Rinter Ͳ 0,667 0,667 0,667
MioceneClay HS Hssmall Hssmall Ͳ Ͳ OK OK OK
OCR Ͳ 1 1 2,5
Figure 5. Defined soil models in T1-2-3 POP kN/m2 0 0 0
model
Figure 6. Soil parameters
The input data of the soil models are given
in Figure 6. The highlighted specific
parameters of HSsmall model was
naturally inputted only in T2-T3 versions.
The strength and stiffness values was
mainly derived from the CPT results using
formulas offered by the Eurocodes. The
HSsmall parameters were calculated
based calculations by formulas of Hardin
and Black, Seed and Idriss, Weiler, Rix
and Stokoe, Mayne and Rix available in
Kramer (1996).
2.3.2. 2D PLAXIS model
Cross section of the 2D model is given in Figure 7. 2D PLAXIS model with support at
Figure 7. The plane strain model consists top of wall
179
All together 6 different 2D PLAXIS models
were created to analyse significance of
small-strain stiffness and the influence of
horizontally fairly stiff capping beam on the
soil-structure interaction.
2.3.3. 3D PLAXIS model
Zdravkovic (2009) showed the relevance
of 3D wall behaviour on deep excavations’
models. Our goals with present 3D
modelling were to analyse the significance
of 3D wall stiffness and 3D arching effect
in soil on soil-structure interaction.
The preparation needed much higher
efforts compared to 2D models. The
following issues could be highlighted
which made 3D modelling difficult:
- Geometry input – high amount of data
to be handled accurately, importing
geometry is hardly feasible from
drawing softwares.
- 3D plate element models – 3D stiffness
needs several parameters neglected in
2D.
- 3D meshing – need for simplification of
geometry to avoid low quality slender
finite elements.
As Figure 8 shows one half of the
excavation pit could be modelled with
about ~100.000 elements due to capacity
limits of available IT tools.
Figure 8. 3D PLAXIS model mesh
A key issue of such calculations is the 3D
stiffness of the wall. 4 different wall
models have been examined using
Young’s modulus E=19.50GPa:
- F1 - isotropic wall
180
- F2 - orthotropic wall, where
Ehorizontal=0.2×Evertical
- F3 - orthotropic wall, where
Ehorizontal=0.001×Evertical
- F4 - isotropic wall, where free rotation
hinges are defined in the diaphragm
wall panel joints
Figure 9 shows the 3D model with F4 wall
model highlighting the free rotation hinges.
All together 12 different 3D models were
examined using the 3 different soil models
and the 4 different wall models.
Figure 9. 3D PLAXIS model with F4 wall
model
3. RESULTS
The deformation and internal forces of the
diaphragm wall, the maximum anchors
forces and steel strut forces had been
examined. This study highlights only the
wall deformation and internal force results
calculated by the introduced techniques.
The following figures shows diagrams with
several curves. The legend containing the
name of the curve refers to the way of
calculation to help the interpretation.
3.1. Influence of small strain stiffness
on deformations
Figure 10 shows the measured
deformation of the diaphragm wall
orthogonal to the wall plane at the final
phase when the excavation is finished. It
can be seen that the curve calculated with
GEO5 Monnet model (and GEO5 Schmitt
model providing strictly speaking the same
curve) correlates quite well but the top
~5m of the wall tilt back more than the
measured. 2D PLAXIS model with T1 soil
model shows much higher deformation,
however the curvature of the wall is
slightly closer the measured shape. The
2D PLAXIS calculations with T2 and T3
soil models show practically identical
deformations fitting the measured ones
better. Nevertheless the top part of the
wall moves towards the pit more than it
was measured. As a conclusion of Figure
10, it can be stated the small strain
stiffness behaviour has higher relevance
in the clay bedrock providing the passive
earth pressure on the wall. In layers which
acts only as an active zone, the
consideration of small strain stiffness has
lower significance. The difference between
GEO5 Monnet and 2D PLAXIS T2 and T3
model arose from the fact phenomenon of
the disk-like deformation of the retained
earth neglected by the preceding and
considered by the latter.
Figure 10. Wall deformations by different 2D
models
3.2. Drawbacks of 2D models
None of the models analysed in chapter
3.1 takes into account the following
favourable 3D effects which can be
181
deterministic for such deep but narrow
excavations:
- Retaining action of the head beam on
the top of the diaphragm wall.
- 3D arching of soil around the
excavation pit effect.
2D PLAXIS models have been prepared
to try to consider the head beam by a
horizontal elastic support. Its deformation
for T2 soil model is given in Figure 11
providing very good fit with slightly lower
curvature than the measured one.
Nevertheless further similar analysis is
needed to determine rules for such
intervention in modelling for design of
walls.
Figure 11. 2D PLAXIS results with and without
horizontal support at top level
3.3. Comparison of 2D and 3D models
As chapter 3.2 showed, 3D effects can
highly influence the wall deformations. 3D
models could naturally take these effects
into account. Figure 12 shows 3D arching
effect like phenomenon: the total
displacements around the pit dissipates
close to the corners of the pit, where the
wall provides much stiffer retention. This
behaviour implies that plain strain
conditions can hardly be realistic for
narrow deep excavation like this. Figure
13 shows the total displacements of the
diaphragm wall displaying the retaining
influence of the head beam and the
corners on the wall.
Figure 12. Total displacement of the pit in a
3D model from top view
Figure 13. Total displacement of wall
Figure 14 shows the results of 3D models
with T2 soil model and F1-2-3-4 wall
model compared to PLAXIS 2D results
and measured deformations. As it can be
seen, 3D models provide much smaller
deformations compared to 2D models and
measured values. However, as F1-F2-F3-
F4 models provided practically the same
curve covering each other on Figure 14.
To summarize that can be said, the wall
stiffness has no real influence on
deformations of the wall in halfway of the
span where the highest deformations shall
occur. The head beam implemented in the
3D models served its designed role and
retained the diaphragm wall on its top.
However some unrealistic back tilt of the
top of the wall could be observed similarly
to 2D GEO5 models.
182
3.4. Analysis of soil behaviour of 3D
PLAXIS calculations
Soil parameters were correlated from the
available CPT results and the clay
bedrock was modelled as undrained
material with OCR=2,5 resulting in an
earth pressure coefficient at rest of 1,27.
This approach was assumed realistic for a
stiff clay and for short construction period.
Literatures can be found dealing with
overconsolidation of Budapest clays.
Some studies have been published
providing values usually between 1 and 3
for OCR values and some in-situ test were
testified to measure earth-pressure on
diaphragm wall (Deli et al, 2009).
Nevertheless bedrock behaviour is still
under research and most sensitive
question of excavation design.
Figure 14. 3D PLAXIS model results
Based on results shown in Figure 14, an
analysis of latter parameters have been
done with 4 different soil models listed in
Figure 15. Figure 16 shows the resulting
deformations from these models
compared to the measured ones.
As Figure 16 shows the drainage
behaviour has higher significance on soil-
structure interaction than the K0 value.
The graph implies that the clay bedrock
behaves as a drained material. However analysed. Beside the control of
these models still slightly underestimates movements, the internal forces arise in the
the deformations of the wall. wall are a crucial question for
practitioners. Present chapter gives a
CLAYMODEL K0 Drained/Undrained
short comparison of these calculated by
3DPLXT2F2 1,27 Undrained the different methods.
3DPLXT2F2MOD1 1,27 Drained
Bending moments and shear forces
3DPLXT2F2MOD2 1,00 Drained calculated by GEO5, PLAXIS 2D and
3DPLXT2F2MOD3 1,00 Undrained PLAXIS 3D models for soil model T2 at
Figure 15. 3D PLAXIS models with different stage when excavation is finished are
clay models shown in Figure 17 and 18 respectively.
These figures show only F2 type wall
models’ results of 3D calculation for better
visibility. Analysing the graph of the
bending moments, it can be diagnosed
that the 2D models by GEO5 and PLAXIS
with T1 soil model are fitted to each other
very well.

Figure 16. 3D PLAXIS model results with

different soil models
The reason of it could be that, based on
CPT qc values quite high shear strength of
the clay was determined. The derived
value of cu=577kPa was added as
c=577kPa and ੮=0° for the drained soil
models leading to the fact that the
unloading action by excavation of the pit
could not cause the reduction of soil Figure 17. Bending moment results by
strength. Another reason could be that the different methods
K0=1 value is still higher than realistic. Comparison of 2D PLAXIS models’ results
Possibly better fit could be found with implies that the introduction of small strain
more information on clay strength and stiffness behaviour change the resulting
earth pressure values. bending moments significantly. The
3.5. Internal forces bending moments of the 3D model are
In chapter 3.1-3.4 the deformation shifted upwards: the top half part of the
characteristics of the wall had been wall are bended more while there is a

183
relevant decrease of the bending
moments along the bottom. All in all, the
total bending moments are much lower in
the 3D models due to the two-way load-
bearing behaviour. Regarding the shear
forces it is hard to find clear tendencies,
however it can be seen the forces are in a
similar range for all the models.
the wall are resulting from the 3D stiffness
of the wall and the presence of head
beam in the 3D models. Beside this, the
bending moment reduction influence of
HSsmall soil model is even more
significant in 3D models.

Figure 19. Bending moment results of 3D

models with different wall stiffness
Figure 18. Shear force results by different
methods
Figure 19 presents the bending moments
calculated by 3D models with T2 soil
parameters with F1-2-3-4 wall models. It
should be mentioned the slight
disturbance of the curves by 3D models is
because of the limitations of the finite
element mesh refinement capacity of
available computers. As Figure 18
indicates, the orthotropic behaviour of the
diaphragm wall has some influence on
principal bending moments above the
excavation level but it is hard to find clear
tendencies. Possibly finer finite element
mesh could clear the view.
Figure 20 shows the influence of the
application of HSsmall soil model on
bending moments by comparing T1 and
T2 models’ results in PLAXIS 2D and 3D, Figure 20. Bending moment results with
as well. This graph clearly shows that the different soil models by PLAXIS 2D&3D
higher bending moments in the top part of

184
3.6. Tendencies of deformations and moments rotating around the horizontal
forces in 3D models axis are practically zero, as Figure 23
According to present practice, 2D plain displays, while the bending moments
strain models are used for design. It is rotating around the vertical axis are
well-known that closer to pit corners it is a increased.
fairly conservative approach as soil
pressures shall be lower while walls has
better stability. Figure 22 and 23 shows
the results of PLAXIS 3D model with T2
soil model and F2 wall model in 5 sections
along the wall span. Similar tendencies
could be found for all the wall models. The
location of the 5 section are shown in
Figure 21 with the following notes:
- A section: in the mid span (same
analysed in previous chapters)
- B section: halfway between A and C
- C section: ~1m from the anchored
corner
- D section: halfway between A and E
- E section: ~1m from the strutted corner

Figure 22. Wall deformations in section A-B-

C-D-E

Figure 21. Location of analysed sections

As Figure 22 represents, the deformations
are the highest in section A. In the
anchored corner, in section C about half of
the maximum deformation could be
observed. Looking on the strutted corner,
a different deformation shape was found
as the steel struts were not pre-stressed.
Therefore the curvature of the wall are
much smaller. Even the maximum
deformation of the wall in section E was
about 25% of the deformation in section A.
Similar tendencies were found in the same
section regarding the bending moments. It
must be emphasized, the strutted corner
Figure 23. Bending moment in section A-B-C-
suffered much lower moments because of
D-E
the bi-directional bending moments: the
185
4. CONCLUSIONS Kardos N., Mahler. A, Móczár B. (2014)
“Back Analysis of the retaining structure
The back-analysis of the inclinometer
behaviour at Budapest Metro Stations”.
measurements of the introduced pit by
Proceedings of Conference of
GEO5, PLAXIS 2D and 3D lead to the
Geotechnics, Ráckeve, Hungary.
following concluding remarks:
Deli Á., Kaltenbacher T., Havas P. (2009)
- HSsmall model provides more realistic
“Role of monitoring in execution of deep
soil-structure interaction compared to
excavations”. Proceedings of ISSMGE
HS model by reducing the extent of the
International Seminar on Deep
interacting soil body. However further
excavations and Retaining Structure,
research is needed to obtain correct
Budapest. (Eds.: Mahler A., Nagy L.)
parameters regarding overconsolidation
ratio, earth pressure coefficients and Kramer, S. L. (1996) “Geotechnical
soil strength. Based on the research’s Earthquake Engineering”. Prentice Hall,
experiences, the models are not so Upper Saddle River, New Jersey.
sensitive for the specific values of LĘdör K., Mahler A., Móczár B., Bán Z.
HSsmall model (G0 and ɶ0.7) but their (2016) “Back analysis of a sugar silo by
proper range shall be found. 3D finite element method”. PLAXIS
- 3D models provides the possibility to Bulletin 39.
eliminate the inherent deficits of 2D Ray, R., Scharle, P., Szepesházi, R.
models of diaphragm walls by the (2010) "Numerical modelling in the
introduction of head beam and the 3D geotechnical design practice".
stiffness of the wall. Proceedings of 14th Danube-European
- The input of the 3D stiffness values of Conference on Geotechnical
wall structure has low influence on soil- Engineering, Bratislava.
structure interaction but the use of Szepesházi A., Mahler A., Móczár B.
orthotropic material seems reasonable. (2016) “Three Dimensional Finite
The very similar results of F2 and F4 Element Analysis of Deep Excavations’
wall models implied that about 20% of Concave Corners”. Periodica
vertical stiffness shall be applied Polytechnica - Civil Engineering,
horizontally. Budapest (under edition).
- According to the 3D PLAXIS results, Szepesházi R., Meszlényi Zs., Radványi
the plain strain conditions are not valid L. (2009) “Hungarian practice for
and conservative closer to pit corners. retaining structure design according to
EC7”. Proceedings of ISSMGE
ACKNOWLEDGEMENTS International Seminar on Deep
excavations and Retaining Structure,
I would like to thank the support of my
Budapest. (Eds.: Mahler A., Nagy L.)
PhD tutor, Dr. Balázs Móczár. Beside I am
very grateful GergĘ Csapody, our MSc Zdravkovic, L. (2009) "Modelling deep
student for his valuable contributions to excavations in 3D analysis".
the research. Finally I would like to Proceedings of ISSMGE International
express my thanks for the engineers of Seminar on Deep excavations and
HBM Ltd. providing input data for this Retaining Structure, Budapest. (Eds.:
research. Mahler A., Nagy L.)

REFERENCES
Benz, T. (2006) “Small-Strain Stiffness of
Soils and its Numerical
Consequences”. PhD Thesis. Institut für
Geotechnik der Universität, Stuttgart.

186
25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Session 4

Retaining structures
 Executa lucrari de:
x Incinte din pereti mulati, piloti secanti, palplanse sau sprijiniri berlineze;
x Piloti / coloane cu diametre intre 400 ÷ 2000mm; Piloti cu baza largita pana la 2600mm;
x Barete;
x Minipiloti armati si injectati cu diametre intre 133 ÷ 250mm;
x Ancore cu bara rigida, ancore autoforante, step anchors, ancore lita;
x Injectari cu supensie de ciment si prin metoda jetgrouting;
x Protectii de mal verticale, cu peree din anrocamente, gabioane;
x Consolidari de terenuri prin silicatizari;
x Consolidari de terenuri in adancime prin coloane vibropresate, piloti de var/ciment, snec
invers si compactare cu mai foarte greu, coloane de piatra sparta cu preindesare;
x Torcretari si stabilizari de taluzuri;
x Epuizmente si depresionari;
x Incercari statice de proba la capacitate portanta pe piloti / minipiloti /ancore si barete;
x Incercari prin metoda carotajului sonic si metoda ultrasonica de impuls;
x Determinari pe suspensie bentonita la piloti/pereti mulati in laborator „in situ”;

_______________________________________________________________________
ZUBLIN ROMANIA SRL
Sediu social: Strada Calea 13 Septembrie nr. 90, Sector 5, Bucuresti, cod 050726
Punct de lucru: Strada Domnita Ruxandra nr.12, Sector 2, Bucuresti, cod 020562
Tel: +40-21-212-08-89; +40-21-212-08-91; +40-37-275-30-03;
Fax: +40-21-212-57-47
www.zublin.ro Email: office@zublin.ro
 Staggered anchors - Züblin Romania performs step anchors in Romania
Given the growing demand for office buildings and shopping malls in big cities across the country, the market
looking for new solutions, more efficient and economic for execution the enclosure strutting system of the basement.
Zublin Romania has performed the underground works for a new multifunctional center built in Floreasca
district of Bucharest; the project also includes the Sky Tower, a high-rise building with a height of 131m. With 37
upper floors and 5 basement floors, this tower will be the highest building in Bucharest.
The shopping and entertainment center has a gross floor area of approx. 110.000 m2 and it's divided into 3
basement floors with approx. 1300 parking spaces and 3 upper floors with sales areas.
For the performance of the pit the project designer has chosen multiple civil engineering technologies (see
Figure 1. Site layout and anchor position):
x Diaphragm walls for the basement enclosure
x Support with steel tubes for the N-W and S-W execution pit, in the corners area
x Top down method for the North part of the site
x Tension rods support for the exposed corner of the diaphragm wall
x Temporary strand anchors for approx. 60% of the length of the diaphragm wall

Figure 1. Site layout and anchor position

The contract objective was to assure the stability of the diaphragm walls through temporary strand anchorages.
This solution is best suited for the work schedule taking into consideration the urban density of the area, the office
buildings surrounding the site and the work costs.
It has been done a single level of anchorage. According to the designer’s choice, it has been executed the
strand anchors with staggered anchorage length (see Figure 2. Anchors with staggered anchorage length).


Figure2.Anchorswithstaggeredanchoragelength

The anchor was 22.5 to 26 meters length, the bulb length of 8-10m and angles between 10o and 17o with a
lock-off load by 1000kN.
The anchors were performed in sand, clay and sometimes performed in sand with underwater pressure.
During the testing phase, according to the designer requirements, we tension the anchor with a coefficient of
1.5 the working tension –about 1500kN- which is rather impressive if we look at the anchor's body.
 Steps for the execution of the temporary strand anchors (see the Figure 3.):
1. Drilling of the borehole and filling with grout
2. Installing the anchors into the borehole
3. Primary injection with grout
4. Secondary injection with grout after 24 hours
5. Stressing the anchorages up to 1.25 their working tension and blockage at 0.8 as the designer
demanded
6. The total number of executed anchorages was 155

Figure 3. Steps for the execution of the staggered anchors

Advantages of the strand anchors:

x Anchors without joints;
x Relatively low net weight and small diameter;
x Easy installation because of high flexibility of anchors;
x High bearing capacity because of optimized transfer of efforts from anchors to the ground;

For a normal anchor the load transfer from the anchor body to the ground is not distributed uniformly along
the bulb length. This may result in progressive failure of anchors in poor soil conditions such as loose to medium dense
sands as silty / clayed soils, similar to the ones encountered on site.
By staggering the bond length, the load transfer is more uniformly along the bond length, thus better making
use of the ground’s capacities. This property gives us the opportunity to achieve a higher bearing capacity. (see Figure
4. Effort distribution)


`

Figure 4. Effort distribution

Disadvantages of the strand anchors:

For the anchors with staggered bond length the disadvantage is stressing procedure witch is more complicated
as the different free length has to be compensated to achieve equal loads in different strands. (see Figure no. 5 Tension
– Elongation)

Figure 3. Tension - Elongation

In conclusion we can say that the tension gain we achieved with these strand anchors with staggered anchorage
length diminished the total number of normal anchors needed for this job.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Design of Retaining Walls at Metro Nordhavn, Copenhagen

Asbjørn M. Dam1*
1
Züblin A/S

ABSTRACT
Zûblin A/S is a part of the MetNord JV who design and construct the Metro Cityringen – Branch off
to Nordhavnen. The project consists of the Nordhavn Station, a Cut and Cover tunnel, a ramp
which takes the trains up to ground level and a bored tunnel from Nordhavn Station to
Østersøgade, where it is connected to the Cityringen Metro. Züblin A/S has been responsible for
the design of the temporary structures which is mainly carried out as multiple supported secant pile
walls, supported by pre-stressed ground anchors.
The first design was carried out by modelling the secant pile walls with SPOOKS which uses the
theory of J. Brinch Hansen, commonly used in Denmark, for the ultimate limit state and the finite
element program Plaxis for the serviceability limit state. Due to the limitation of Brinch Hansen’s
theory and discussions about stress-strain compatibility, the Employer doubted that the
deformations necessary for activate active and passive earth pressure were sufficient in the
ultimate limit state. Therefore the final design ended up being a combination of SPOOKS for the
ultimate limit state and a PLAXIS model for ultimate limit state to verify the results of SPOOKS and
a serviceability limit state calculation also with PLAXIS.
This paper tries to investigate the influence by introducing PLAXIS into an ULS calculation by
making different variations of inputs that changes the stiffness of the system and thereby
influences the stress-strain compatibility. The results will be compared with the results found with
Brinch Hansen theory and discussed.

Keywords: Retaining walls, Brinch Hansens earth pressure, SPOOKS, PLAXIS,

parameter study

The retaining walls for the underground

1. INTRODUCTION
As a part of the Metro Cityringen in
Copenhagen the Joint Venture (MetNord
JV) consisting of Züblin and Hochtief are
building the first part of the branch off to
Nordhavnen, consisting of an
underground Station, a Cut & Cover
tunnel and a Ramp going up to the
surface level and ending as an elevated
track, see Figure 2.
* presenting author
Station are made as permanent secant
piles supported mainly by 3 layers of
temporary pre-stressed ground anchors.
The retaining walls for the Cut & Cover
tunnel are temporary secant piles
supported by 1 or 2 layers of pre-stressed
ground anchors. The Ramp area consists
of permanent sheet piles supported by 1
layer of permanent pre-stressed ground
anchors.
The deepest excavation, where the TBM
drive will start, is around 18 meter deep
and decreases along the entire alignment
towards the end of the ramp section.
This paper will focus on the design of the
secant pile walls of the underground
Station and the Cut & Cover Tunnel in the
temporary situation, where one cross
section will be used for the analyses
carried in this paper. The secant piles
installed have a diameter of 1.2 meter and
c/c distance of 0.9 – 1 meter and
reinforced by reinforcement cages. The
wall is used as cut off wall for the ground
water and is therefore drilled and casted
until level -20 (DVR90), with reinforcement
going to the toe of the pile.
2. DESIGN OF RETAINING WALLS
The secant piles are designed in both
ultimate limit state (ULS) and serviceability
limit state (SLS). In Denmark, the ULS
calculation is mostly carried out in the
program SPOOKS, which uses the earth
pressure theory developed by J. Brinch
Hansen, (Hansen J.B, 1953). His theory is
using a combination of zone and line
rupture where the earth pressure
distribution is a modified compared to
Figure 1.Location of the new underground Nordhavn Station, Cut & Cover tunnel and Ramp
192
normal failure mechanism and depending
on the rotation of the wall. The theory
states, that the necessary displacement to
mobilize active and passive earth pressure
will be present independent of the failure
mechanism of the wall. The failure
mechanism from the theory of Brinch
Hansen is only a statically possible
solution, if ij > 0, meaning it is a lower
bound solution.
From SPOOKS it is possible to obtain
anchor forces, bending moment, and
necessary toe level to fulfil the equilibrium
requirements.
The program has certain limitation, as it is
only possible to include one real support
force. When designing multiple supported
walls, it is then necessary to add some of
the anchor levels as, so called, additional
pressures. As the values for the additional
pressures, in principle, can be chosen
arbitrary it is necessary to check if the
anchor force chosen is feasible compared
to the earth pressure distribution. This can
be done with a simple beam calculation,
where the anchor levels are modelled as
simple supports and the earth and water
pressure is applied as a line load.
 Due to the fact that the program does the analysis, which is positioned at ground
not take soil-structure interaction into level.
account, the Employer was concerned that
there is insufficient compatibility between
stresses and strain meaning that the
deformations necessary to mobilize full
active earth pressure is not present and
the program therefore calculate too small
section and anchor forces.
Different approaches was used to
convince the Employer that the design
carried out was sufficient, but without luck.
Therefore it was decided to model the wall
and anchor with Finite Element, by means
of a PLAXIS model. With the PLAXIS
model it is possible to include all
construction sequences, wall and soil
stiffeners, and therefore also built-in forces
coming from pre-stressing of anchors and
from the different excavation stages.

3. SOIL AND GROUND WATER

CONDITION
The following soil and ground water Figure 3.Soil profile used for the calculations
conditions are used for the calculations in in this paper
this paper.
3.1. Soil condition Table 1.Soil properties used for the MetNord
The boreholes made on the location project and calculations in this paper
shows fill layers with a thickness variating Soil Ȗ/Ȗ’m ij’ c’ Eoed Ȟ
from 3.5 to 11 meter. That matches the layer
fact that it is an old backfilled harbour kN/m
3
° kN/m
2
MN/m
2
-
area. Below the fill a 2.5 to 10 meter thick Fill 17/19 30 0 3 0.3
clay till layer with lenses of sand till and
Clay till 12+1500xı’
melt water sand is registered on top of the 22/22 34 20
red
1) 0.3
limestone. The top level of the limestone Dist. 0.25-
varies throughout the area and has a Limestone 22/22 45 50 750
0.30
glacially disturbed zone of around 1-3 Limestone 0.25-
meter. A 2 meter disturbed limestone zone 22/22 45 100 900
0.30
has been used in the design. The soil 1)
ı’red is the vertical effective stress corresponding to
profile used in this paper is shown in the minimum stress level the soil has experienced
Figure 3.
The characteristic drained strength 4. SPOOKS CALCULATION
parameters used for the calculation are
For the investigations in this paper a
shown in Table 1.
SPOOKS calculation of a cross section at
3.2. Ground water level the station is included. The section has
In the design of the retaining walls, a two anchor levels and an excavation
primary and secondary ground water table depth of 17.5 m. Figure 4 shows the
are used. For simplification only one results at the final excavation stage giving
ground water level is taken into account in the bending moment, anchor force and
necessary toe level.

193
 x Over-consolidation ratio
x Type of soil model – Mohr Coulomb
or Hardening Soil?
x Pre-stressing level for the anchors
A sensitivity study of the different
parameters have been carried out with
focus on anchor load and bending
moment. The different investigation is
done on the same model and with the
same reference result to compare with.
The results of the SPOOKS calculation
from section 4 are included in the
comparison. Eroare! Fără sursă de
referinĠă. shows the PLAXIS model used
in the investigation. The pre-stressed
ground anchors are modelled as node-to-
node tension elements for the free length
and the bond length is modelled as
geogrid element. The upper anchor layer
is installed with an inclination of 30° and
lower anchor layer with 40°. The secant
Figure 4.Graphically output from SPOOKS pile wall is modelled with plate element
The results are listed in Table 2. with elastic properties. The characteristic
Table 2.Design values from SPOOKS surface load is 20 kPa.
calculations After each main excavation steps an ULS
Bending [kNm/m] 1287 step is calculated. This is done with a Phi-
moment C reduction with a target value of 1.32 and
an increased load to the design load of 30
Upper anchor [kN/m] 196 kPa, which correspond to the partial safety
load factor for the effective soil parameters. A
Lower anchor [kN/m] 868 Mohr-Coulomb material model with
load drained parameters is used for all soil
layers. The influence on the deformations
Toe level [m] -18.95 will not be considered in this paper.
The Spooks calculation shows a field 5.1. Stiffness of wall
bending moment which is identical to the
embedment moment. For the stiffness of the wall the Employer
required that the stiffness of a cracked
5. PARAMETER STUDY section is used. The argument was that
using a cracked stiffness of the wall the
When introducing PLAXIS into an ULS calculation would give higher anchors
calculation, it gives many options and forces but less bending moment in the
different inputs that can be adjusted. For wall. To see what influence it would have
the secant piles at MetNord a number of to use a cracked or uncracked bending
inputs will have an influence on results. moment, the influence of the wall stiffness
That is: is investigated. The bending stiffness, EI,
x Stiffness of the wall – cracked or is increased with intervals of 100E3
uncracked concrete properties? kNm2/m, starting from 600E3 up to the
x Stiffness of the soil layer in this double, 1.2E6 kNm2/m, corresponding to a
case the clay till wall 4-5 times stiffer than the largest AZ -
x Interface reduction factor, Rinter. sheet pile profile on the market. The
results of the analysis are shown in Figure
194
6; the positive bending moment is from the moment towards the excavation pit. The
embedment in the soil, while the negative tendency of the results is as expected –
bending moment is the field bending almost no change occurs.

Figure 5.Plaxis model used for the different analyzes

The increase in stiffness of 100% gives an
increase of approx. 7.7% for the positive
bending moment and 6.6% for the
negative bending moment. The decrease
in anchor load is at the same time 4.2%
for the upper anchor layer and 5.2% for
lower anchor layer. It is therefore only
have marginal and neglectable influences
which wall stiffness is used, so the
difference in using cracked or uncracked
stiffness would only be marginal for the
design.
5.2. Variation of clay till stiffness
The stiffness of the clay till is derived from
a number of oedometer tests. The Danish
practice for presenting the oedometer
modulus is by following equation:

In this way it represents the in-situ

conditions after a pre-consolidation and an
Figure 6 Comparison of results with different
wall of the analysis
unloading to ı’red, which is the minimum

195
effective vertical stress the soil have
experienced after the pre-consolidation.
The chosen Eoed for MetNord is found by
comparison of other clay tills in the
Copenhagen area and the Great Belt Link,
as shown in Figure 7.
Figure 7.Compiled oedometer modules for
different locations
The chosen value is also compared to
measured values from oedometer tests
carried out on site, and shows that the
chosen value is conservative. By
introduction FEM in an ULS calculation it
means that the stiffness of the surrounding
soil will have influence on the section
forces and anchor forces. Therefore an
analysis with the different Eoed values from
Figure 7 is carried out. To take into
account the increase of stiffness with the
depth (effective vertical stresses), the Einc
function in Mohr-Coulomb model is used.
The relation between E and Eoed for MC is
given by:
196
The poissons ratio, Ȟ, for the clay till is 0.3.
It is only the stiffness which is changed;
the effective strength parameters are not
changed. The results of the analysis are
shown in Figure 8 and Figure 9. The
comparison is based on the Eoed in the
middle of the layer.
Figure 8.Bending moments from the analysis
with different clay till stiffness’
As for the wall stiffness the influence of
the clay till stiffness is marginal for the
bending moment and anchor forces.
Comparing the results for “Malmø – lower”
and “Great Belt Link” the increase in Eoed
in the middle of the layer is 180% but the
increase in wall bending moment is only
3% for both the positive bending moment
and the negative bending moment of 3%.
The upper and lower anchor forces are
decreased by 8% and 1% respectively. As
verified the stiffness of the soil has only
marginal to no influence on the design
values.
 used until 10-15 years ago, where today r
is chosen so vertical equilibrium is
obtained.
To see if the interaction has any significant
influence on the bending moment in a
PLAXIS calculation a variation of Rinter
from 0.5 to 1.0 is carried out. In the project
a value of 0.9 has been used, as it is
concrete and in general till layers. The
results are shown in Figure 10.

Figure 9.Anchor loads from the analysis with

different clay till stiffness’
5.3. Variation of roughness/interaction
The interaction between soil and structure
is controlled in PLAXIS by the strength
reduction factor, Rinter. Rinter is connected
to the friction angle and adhesion of the
soil, but also reduces the interface
stiffness. If the Rinter is equal to 1 there is
full interaction between soil and structure
and maximum possible shear stresses
along the wall is mobilized. The reduction
factor depends on the soil type and Figure 10.Results of the PLAXIS calculations
structural material. According to with variation of the reduction factor, Rinter.
(Brinkgreeve, R.B.J., 2015) the value can
be assumed to be 2/3 if no further The reduction factor does not have any
information is given. A similar interaction significant influence on the bending
factor is also used in Denmark when moment and anchor forces. The decrease
determining the skin friction capacity of in bending moment is approximately 1%
piles. For piles in cohesive soil the values and the anchor forces are 4% and 2%
is 0.7 for steel piles and 1.0 for concrete respectively, and as so Rinter are of no
piles (Ovesen N.K et. al, 2009). influence of the design. A similar analysis
A roughness factor, r, is also used in has been carried out in SPOOKS, where r
SPOOKS, which is the ratio between wall is changed for the clay till. The results are
friction angle and friction angle of the soil, shown in Figure 11, where the results from
or ratio between wall adhesion and PLAXIS are included as comparison.
cohesion in the soil. In SPOOKS the
roughness factor has an influence on the
bending moment, anchor forces and toe
level. In Denmark r = 1.0 was normally
197
 Table 3.Govern hardening soil parameters
Erefur Eref50 Erefoed m
[kPa] [kPa] [kPa] [-]
HS 106E3 35E3 35E3 0.8
The results of the analysis are shown in
Table 4.
Table 4.Results of the MC and HS
calculations
Bending Bending Anchor Anchor
moment moment force- force -
- M- - M+ A1 A2
[kNm/m] [kNm/m] [kN/m] [kN/m]
MC -1314 1531 194 768
HS -1343 1565 212 769
The difference between the two soil
models is marginal, which was expected.
Figure 11.Comparison of SPOOKS and The increase in negative and positive
PLAXIS results, when variating r and Rinter. bending moment, when using HS, is 2.2%
and 2.2% respectively. The increase in
anchor load is marginally. The results
The tendency is similar for the verify that using the simpler Mohr-
SPOOKS calculation as for the PLAXIS Coulomb material model is sufficient.
calculation. The decrease in bending
moment when going from a roughness 0.5 5.5. Variation in pre-stress of anchors
to 1.0 is 3% and 7% for the upper anchor The previous calculations show a
and 3% for the lower anchor. difference of approximate 25% for the
positive bending moment, which
5.4. Mohr Coulomb vs. Hardening Soil
correspond fine with the fact that the lower
Due to the stress dependent stiffness anchor is approximately 10% lower. In
behavior of the clay till another soil model SPOOKS it is not possible to add any pre-
that could be feasible to use is the stressing to the anchor, but in PLAXIS it is
Hardening Soil (HS) or Hardening Soil possible to control the anchor force with
Small Strain (HS-ST) model. For the the pre-stressing force. A normal pre-
design it was assessed that the Mohr stress factor used is 0.6 to 0.8 of the ULS
Coulomb is sufficient as the surroundings anchor load found from SPOOKS. The
was not sensitive to deformations. Another calculations have been carried out with a
issue was the amount of inputs that pre-stress factor of 0.6 compared to the
should be determined by assumptions as SPOOKS calculations, which is in the low
the needed test for obtaining the correct end. Therefore the pre-stressing is
inputs was not available. To see what changed in intervals from 0.1 up to 1.0 of
influence the choice of soil model has on the SPOOKS anchor load. The results are
an ULS calculation a set of HS parameters shown in Figure 12. As the design load for
is determined, as shown in Table 3. The the upper anchor does not change the
stiffness parameters are determined result is left out.
based on the EMCoed and an assumed m-
value of 0.8 which is based on experience.

198

Figure 12.Results of the PLAXIS calculation
with pre-stress variation.
In contrary to the other analyzes carried
out the pre-stressing shows a significant
impact on the bending moment and lower
anchors load. The negative and positive
bending moment decreases with 7% and
24% respectively. At the same time the
lower anchor increases with 26%. That is
due to the fact that the anchor takes more
loads and thereby decreases the bending
moment in the embedded part of the wall.
The plot shows that a pre-stress factor
between 0.8 and 0.9 (1632 kN – 1835 kN)
will give results similar to SPOOKS. As so
it is recommended to use a value in that
region for the pre-stressing of anchors, if
PLAXIS is used for ULS verifications and
both positive and negative bending
moments are expected in the design.
6. INFLUENCE OF STRENGTH
PARAMETERS
To show that the strength parameters
have a bigger influence on the bending
moment and anchor forces than most of
the stiffness inputs investigated in the
previous section, 4 calculations with
different strength parameters is carried
out. The 4 combinations are shown in
199
Table 5 and the results of the calculations
are shown in Figure 13.
Table 5.The 4 combinations used for the
analysis
Combination ij’ [°] c’ [kPa]
1 30 10
2 30 20
3 34 20
4 34 30
Figure 13.Results of the PLAXIS calculation
with variating strength parameters.
The calculations show that small changes
in strength parameters decreases the
negative and positive bending moment
with 18% and 17% respectively. The
upper anchor force decreases with 23%
and the lower anchor force with 10%. The
calculations show that the bending
moment and anchor forces are more
sensitive to the changes of strength
parameters than the stiffness properties of
the wall and soil and as so more
emphasize should be put on determining
correct strength parameters.
7. CONCLUSIONS
From the four different analyses it can be
concluded that the stiffness dependent
parameters do not influences the section
forces and anchor forces, when the soil
strati mainly consist of clay till. The
increases and decreases are marginally
and other inputs like the pre-stressing and
strength parameters are assessed to be
more decisive for the design, when having
retaining walls with 2 anchor layers.
Comparing the Spooks calculation and
PLAXIS calculations shows large
difference, especially in respect of the
positive bending moment and second
anchor layer. The difference is due to the
pre-stressing force of the second anchor,
as shown in section 5.5 an increase of
pre-stress factor to a value between 0.8
and 0.9 will give similar bending moment
and anchor force.
The calculations show that the results
from SPOOKS are reliable and sufficient
when designing a retaining wall with 2
anchor layers. The calculations also show
that it is more relevant to obtain more
precise strength parameters than stiffness
parameters deformation is not decisive for
the design.

ACKNOWLEDGEMENTS
The author acknowledges the Danish
Geotechnical Society for the economic
support to participate at the conference.
The author also acknowledges Züblin A/S
for the economic support and giving the
opportunity to participate at the
conference. Morten S. Rasmussen also
needs a big thank for guidance and
support.

REFERENCES
Hansen J.B. (1953), “Earth pressure
Calculation”, Teknisk Forlag,
Copenhagen, Denmark
Brinkgreeve, R.B.J., Engin, E., Swolfs,
W.M.(2015), PLAXIS 2D version 2015
manual, Delf, the Netherlands,
Ovesen N.K., Fuglsang L.D., Bagge G.,
Krogsbøll A., Sørensen C.S., Hansen
B., Bødker K., Thøgersen L., Galsgaard
J., Augustesen A.H, (2009), „Lærebog i
Geoteknik”, 1st edition, Denmark,
Haarh F., GEO (2008), „SPOOKS/W
User’s Manual”, Denmark

200
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Eurocode 7 based retaining wall design in Finland –

shortcomings in current method

Mika KNUUTI
1
Tampere university of technology, Department of civil engineering, Tampere, FINLAND

ABSTRACT
In Finland retaining walls are designed by using design approach 2 according to SFS-EN 1997.
National choices for load combinations and for partial factors have been made based on limited
number of comparison calculations. Recently, few flaws have been detected in current design. In
this paper, improvements for retaining wall design in Finland are proposed. Performance of the
current design approach and EN1997-1 design approaches were tested with three calculation
examples, by variating soil strength and loads. DA1 gave the most constant design with its two
combinations. Partial factors should be applied to loads and soil strength parameters, depending
on design situation. Finnish DA2 gave the poorest results. As a result, author suggests that
minimum change would be to use DA2 with recommended values and with one load combination
(6.10). Better choice would be to use modified DA1, where the partial factors for permanent and
variable loads are equal.

Keywords: eurocodes, supported excavation, partial safety factors

countries, although the meaning of the

1. INTRODUCTION
The work toward next generation of
eurocodes is going on, and thus all
eurocodes are under a critical review. This
also includes EN1997-1 for geotechnical
design, which gives the design instructions
for i.e. shallow foundation design,
retaining wall design, slope design and
pile design. Moreover, every country can
determine their own National Annex,
where they choose for example which
design approach to use (three
possibilities) in design and what the values
for different partial factors are. This has
been led to big variations in the design
procedures and design results between
* presenting author
Eurocode was to have more continuous
design between European countries. Now
for the next generation of eurocodes,
different work groups are trying to simplify
and harmonize the current code, EN1997-
1.
For retaining wall design, all three
design approaches are equally used in
Europe (Bond (2013)). This means that
there is no unanimous opinion how to
apply safety in retaining wall calculations.
In this paper we focus only on retaining
wall design in Finland. In Finland retaining
walls are designed by using design
approach 2 according to SFS-EN 1997. In
design approach 2, partial factors are
applied to loads or their effects and to
earth resistance. National choices for load
combinations and for partial factors have
been made based on limited number of
comparison calculations. The aim was to
end up having the same safety level as
with the total factor of safety approach.
Recently, it has been acknowledged, that
this has not been achieved.
2. EUROCODE 7 BASED DESIGN
2.1. Design approaches according to
EN 1997
Eurocode enables three different design
approaches: DA1, DA2 and DA3. These
design approaches differs on how the
partial factors are applied to design
parameters.
In DA 1, two combinations must be
checked. The first combination aims to
govern uncertainties related to actions, or
their effects, from their characteristic
values (ɶG=1.35; ɶG,inf=1.0; ɶQ=1.5),
whereas the design values of strength
parameters are equal to their
characteristic values (ɶʔ’=ɶc’=ɶcu=1.0). In
DA1 combination 2, the partial factors are
mainly applied to strength parameters,
governing unfavorable deviations in these
(ɶʔ’=ɶc’=1.25; ɶcu=1.4), whereas only
variable actions are factored (ɶG=1.0;
ɶQ=1.3).
DA 2 requires only one verification,
where the same values of partial factors
are used for geotechnical and structural
actions (ɶG=1.35; ɶG,inf=1.0; ɶQ=1.5). The
partial factors have been taken from
structural engineering. DA 2 can be
checked in two ways; depending on when
the partial factors are applied. If referred to
DA2, the partial factors are applied to the
characteristic actions at the beginning of
the calculation, and if referred to DA2*, the
calculation is done with characteristic
values and the partial factors are applied
to action effects at the final state of the
calculation. In both ways, for the
resistance side of the limit state equation,
the partial factors are applied to ground
resistance (ɶR,e=1.4).
202
DA3 is similar to DA1 combination 2.
Partial factors are applied to strength
parameters (ɶʔ’=ɶc’=1.25; ɶcu=1.4 and
variable load (ɶG=1.0; ɶQ=1.3) at the start
of the calculation. However, actions
coming from the structure are multiplied by
the partial factors of DA2 (ɶG=1.35;
ɶG,inf=1.0; ɶQ=1.5). Only one verification is
necessary.
2.2. Current practice in Finland
In Finland retaining walls are designed by
using design approach DA2. For loads,
the load combinations 6.10a and 6.10b
are used. These are given in two National
Annexes of SFS-EN 1990 determined by
Ministry of Environment (2007) and
Finnish transport agency (2010). These
two national annexes are used
simultaneously, but depending on the
nature of the project. The former is used
when one is designing bearing
geotechnical structures and the latter
when the design concerns infrastructure
(roads, railways, waterways etc.). In the
latter annex, partial factors for road and
railway (ɶQ=1.25-1.45) loads has been re-
defined from recommended value for
variable load (ɶQ=1.5).
The use of two load combinations
means that safety on actions or effect of
actions are applied according to one
combination with only permanent actions
included (ɶG=1.35) and another including
both permanent and variable actions
(ɶG=1.15; ɶQ=1.5). Load combinations are
presented below:
6.10a:
1.35·KFI·Gk,sup+0.9·Gk,inf
6.10b:
1.15·KFI·Gk,sup+0.9·Gk,inf+1.5·KFI·Qk,1+1.5·K
FI ·Ȉȥ0,i·Qk,i
KFI- factor takes into account the reliability
class (consequence class) of the design.
3. OBSERVED PROBLEMS IN THE
CURRENT DESIGN APPROACH IN
FINLAND
3.1. Load combinations 6.10a and 6.10b
The first problem relates to the use of two
load combinations 6.10a and 6.10b shown
before. These load combinations have
been adopted to geotechnical design from
the field of structural design. The problem
is the partial factor for permanent load in
load combination 6.10b (ɶG=1.15 in
Finnish NA), which has been derived from
the uncertainties related to self-weight of
bearing structures in structural design
(steel, concrete, timber). Now in retaining
wall design the earth pressure forms the
permanent load and the uncertainty
related to this is much larger than for
“ordinary” building materials. It is thus fair
to ask if the partial factor of ȖG=1.15 is
suitable to use in retaining wall design?
It has been proven that by using these
two load combinations the total factor of
safety can, in certain cases, be really low.
If the variable load is 12% from total loads,
the factor of safety coming from the loads
to the structure is around 1.2 (red circle in
figure 1). Now if we have steel structure,
which material factor is ɶsteel=1.0, we may
end up into a design with total factor of
safety of 1.2. The required safety values in
Finland have been 1.5 for temporary
structure and 1.8 for permanent structure.
Currently, this lack of safety has been
fixed by introducing model factors
(ɶmod=1.15 for temporary structures;
ɶmod=1.35 for permanent structures), which
are applied to design stresses in order to
increase the overall safety.
Fig.1 Safety coming from the load
combinations 6.10a and 6.10b (Finnish NA)
203
Also to compensate the reduced safety
from actions compared to recommended
values, the partial coefficient for resistance
has been raised to ɶR,e=1.5 in Finnish NA.
3.2. Numerical applications
Unequal partial factors for permanent and
transient loads are complicated to use in
numerical analysis because of the
nonlinear nature of the calculation.
However, there are two generally
accepted ways to do this. Either of these
are not theoretically correct, but the error
is rather small.
The first approach is to increase the
transient load by ɶQ /ɶG and then multiply
the actions effects by ɶG. The second
approach is to do one calculation with only
permanent loads and one with all loads
(characteristic values). Then separate the
action effects to the effect of transient
loads and multiply received permanent
actions effects with ɶG and transient action
effects with ɶQ. This needs lot of work, but
is it worth it?
4. UNCERTAINTY OF LOADS
In the eurocode the applied load factors
for geotechnical design comes from
structural design. One should then keep in
mind, that the uncertainties involved might
be quite different in geotechnical
engineering. In the retaining wall design
the permanent load is formed by the earth
pressure. The uncertainty of earth
pressure consists at least uncertainties in
strength parameters, earth pressure
theory and calculation model. Most of the
calculation processes are nonlinear
functions (like Jaky, Mohr-Coulomb),
where a little deviation in the starting
values can lead to big deviation in the
results. Obviously, a load factor of 1.15
(Finnish NA) can be kept rather insufficient
for retaining wall design. Moreover, the
use of small partial factor for permanent
loads leads to small overall safety factor,
because its importance to total factor of
safety is high.
It can also be discussed if it is justified
to require a higher load factor for a
variable load than for a load coming from
i.e. an historic building behind the wall.
The permanent load coming from the
building can be very difficult to determine,
so the uncertainty is high. Now we multiply
this with factor of 1.15 (6.10b), whereas, if
we have some sort of crane on the top of
the excavation, which weight we know
exactly, we multiply this with factor of 1.5.
Is this reasonable? This kind of factoring
can lead to overdesign, if our load
uncertainty is low and we use high partial
factor, and to underdesign, if our load
uncertainty is high and we use small
partial factor.
Especially for DA3 and DA1-2 it may be Figure 2. Case 1, excavation in dense sand.
asked if it is justified to factor a load
coming from a variable load, but leave the Case 2
load from an existing building unfactored? The second calculation example is an
anchored sheet pile wall in soft clay (figure
5. CALCULATION EXAMPLES 3). The excavation is 10 meters deep,
reaching the rock surface. The sheet pile
5.1. Basis of the study
wall is supported by three rows of pre-
A detailed study has been performed to stressed rock anchors placed at 1, 3.5 and
study the different design approaches in 7 meters below ground level. Installation
retaining wall design (Knuuti 2015, Knuuti angle of the anchors is 45 degrees. The
and Länsivaara 2015). Three different toe of the wall is anchored to rock with
calculation examples were created in rock bolts in the excavation phase. For
order to compare how different DA’s can final condition a concrete beam is made at
match up with the assumed variations in the toe of the wall.
ground properties and variable actions.

Case 1
The first calculation example is an
embedded sheet pile wall retaining a 4 m
deep excavation in sand (figure 2). The
characteristic unit weight of the sand is
ɶk=18kN/m3, friction angle ʔk=40° and
effective cohesion c’k=0 kPa. The wall is
supported by a single row of struts
installed at one meter below ground level.
A variable imposed surcharge of 10kPa
acts at the top of the wall. Groundwater
level is assumed to be deep.
Figure 3. Case 2, excavation in soft clay.

The soil consists of two meter thick fill

layer, followed by seven meters of soft
clay and a one meter thick till layer above
bedrock. Jet grouting is performed in the
till back of the wall to prevent water flow,
204
but this is not included in the calculation.
Characteristic undrained shear strength of
the clay is suk=7kPa, increasing with depth
with ȴsuk=1.2kPa/m. Unit weight is
assumed to be ɶk=16 kN/m3. A variable
imposed surcharge of 10kPa is acting at
ground level of the retained side.
Groundwater level is below the fill layer.
The wall is made watertight.
Case 3
The third calculation example is a two-
storey underground car park where the
sheet pile wall forms a permanent wall
structure (figure 4). The wall is supported
by three reinforced concrete slabs at
levels 3.8, 0.7 and -2m. Soil consists of
two meters of fill followed by 20 meter
thick, stiff clay layer. Depth dependent
characteristic undrained shear strength of
the clay is suk=35+1.2kPa/m and unit
weight Ȗk=15 kN/m3. A variable imposed
surcharge of 20kPa acts at the head of the
wall. Groundwater level is at ground level
on the retained side and maintained at
formation level on the restraining side. The
excavation is made using top-down
method with step wise exaction for
temporary support of the wall.
Figure 4. Case 3, excavation in stiff clay.
5.2. Variation of the input parameters
In order to investigate the sensitivity of the
DA’s for misinterprets in characteristic soil
strength parameters, variation of the
determining soil parameters was allowed.
205
Suitable COV- values were chosen
according to Phoon et al (1995, 1999), for
friction angle and undrained shear
strength. Chosen COV- values are 0.1
and 0.2 respectively. Based on these
COV- values and formerly presented
corresponding “mean” strength values,
standard deviations were calculated. For
comparison purposes, reduced strength
values were chosen one and two
deviations away from the mean value, see
Table 1.
Table 1. Variation of the strength parameters
of determining soil layers in calculation
examples. Tables are in respective order from
case 1 to case 3.
Also the variable surcharge load was
varied. Values of 0, 10 and 20 kPa were
used for surcharge in all design examples.
5.3. Conclusion of the study
In case 1, differences between obtained
design bending moments and prop loads
were rather negligible between design
approaches. Moreover, regardless of the
used DA, design values (calculated with
ʔ’=40°) were on the safe side even if the
angle of shearing resistance had been
reduced to its lowest value (40°ї33.9°).
What proved to be more important is the
overall stability of the wall. For example, if
the angle of shearing resistance had been
taken to be 40o and in reality would be
33.9o, the overall factor of safety would
reduce to near 1.0 for all DA’s or even
below for some. This addresses the
importance to also lower the level of
resisting soil in SLS calculation below
expected level when the stability depends
on ground resistance. The poorest stability
(under 1.0) was obtained with Finnish
DA2, which gave the shortest embedment
depth of the wall.
The design situation in case 2, is typical
in Finland. In DA2, partial factor applied to
ground resistance is now useless,
because there was no earth on passive
side of the wall at the final stage. Then the 35 kPa to 21 kPa, but the differences are
whole safety to the design comes from substantial when comparing DA1-2 and
load combinations if the material factor for DA3 to DA1-1 and DA2. For DA1-2 and
steel is equal to 1.0. Using the load DA3 the ULS bending moments
combinations 6.10 a) and b) given in (su=35kPa) are only 12% lower than SLS
Finnish NA gives a quite low overall moments (su=21kPa), whereas the ULS
safety. When variable actions are about moments obtained with DA1-1 and DA2
10 % from the total actions, the total factor are over 60% lower than SLS moments.
of safety for the design is near 1.2. This Same kind of behavior can be found from
can be avoided by using separate model prop loads at the second support level, but
factors. at support levels 1 and 3 the ULS values
The highest safety for case 2 was are more or less equal to SLS values
obtained with DA2 (recommended values) when calculating with DA1-2 or DA3.
and DA1 combination 1, which in this case However, this can be avoided in design
are analogous. DA1 combination 2 and approach 1, where combination 1 is
DA3 do not work for this design case. That dominating for these loads.
is, because when the strength of the clay
is low, the influence of the partial factor 6. CONCLUSION
applied to it is also small. For DA3, the Based on presented design cases the
design bending moments and anchor design approach 1, with its two
forces were much smaller than those combinations, seems to produce the most
obtained by DA 1 combination 1 and DA2. consistent design. In the design case 2,
For design bending moments the where the soil strength was low, it is
difference was 15% and for the first recommended to put the partial safety
anchor level 6%. Moreover, design anchor factors to loads rather than soil strength.
forces at support levels 2 and 3 were more In design cases 1 and 3, the situation was
or less equal to corresponding SLS opposite. This has been illustrated in
values, and the differences compared to figures 5 and 6.
DA1-1 and DA2 were 25%. However, all
DA’s were able to govern the unfavorable
deviations in the strength parameters (su).
The lowest total factor of safety was
obtained with DA3.
Case 3 is quite difficult to make
comparisons with high variability in
strength, as that would influence much on
the design. To be able to compare the
results reasonably, the embedment depth
Fig. 5. Partial factors should be placed on
was taken as 1 m for all cases.
strength parameters if the strength is high.
Opposite to the results obtained for
case 2, in case 3, DA1 combination 2 and
DA3 gave the largest design bending
moments and prop loads. Now the effects
of the partial factors applied to the ground
properties are of high importance,
because the strength of the ground is
high.
In this design case, none of the DA’s
can take into account the increases in
design moments and prop loads if the Fig. 6. Partial factors should be placed on
earth pressure and loads if the strength is
clays undrained shear strength drops from
small.
206
Design approach 2 (with recommended
values in EN1997-1) was also able to
govern uncertainties related to ground
properties rather well, but when the
unfavorable deviations in strength are
high, problems may occur. Moreover, in
some design cases the partial factor for
ground resistance can be useless,
whereas in other cases the situation is
opposite.
The DA2 used in Finland (Finnish NA)
with the two load combinations gave the
lowest total factors of safety for every
design case. This can be explained by
looking figure 1. In retaining wall design
we usually are in the “safety pit”
(proportion of variable load is around
10%), which leads to low total factor of
safety, especially for steel sheet pile walls.
We can improve the total factor of safety
by introducing model factors, but these
factors are not applied on every structure.
This makes the calculation even more
complex. Moreover, no additional benefit
was obtained by using two load
combinations, but instead, the amount of
work increased. Also the numerical
calculation was negatively affected by the
unequal partial factors for permanent and
variable loads.
Based on calculation examples, the
design approach 3 is seen problematic for
cases where the soil strength is low.
7. RECOMMENDATIONS
At minimum, Finland should start to use
design approach 2 with recommended
values and with only one load
combination, 6.10 (ɶG=1.35; ɶQ=1.5). The
other possibility would be to use modified
design approach DA1, where permanent
and transient loads are treated equally in
combination 1(same partial factor for both)
and in combination 2, the safety is applied
only to soil strength parameters. This
option would need more investigations.
Also DA2 would most likely benefit from
equal partial factors for permanent and
variable loads.
207
REFERENCES
Bond, A. (2013). Simplifying Eurocode 7’s
design approaches. Task for Evolution
Group 8 Harmonization at the SC7
meeting in Madrid in April 2013
CEN. 2002. EN 1990: Eurocode - Basis of
structural design. Comité Européen de
Normalisation
CEN. 2004. EN 1997-1: Geotechnical
design – Part 1: General rules. Comité
Européen de Normalisation
DS/EN 1997-1 DK NA:2013 Nationalt
anneks til Eurocode 7: Geoteknik- Del
1:Generelle regler
SFS-EN 1990: Eurocode – Basis of
structural design. Finnish National
Annex
SFS-EN 1997-1: Geotechnical design –
Part 1: General rules. Finnish National
Annex
Knuuti, M. (2015). Retaining wall design
according to EN 1997 – uncertainties in
design and application of safety,
Master of Science Thesis, Tampere
University of Technology, Tampere,
2015.
Knuuti, M. and Länsivaara, T. (2015).
Varmuuden kohdentamisesta
tukiseinien mitoituksessa.
Liikenneviraston tutkimuksia ja
selvityksiä, Finnish Transport Agency,
2015.
Länsivaara, T. and Poutanen, T. (2013).
Safety concepts for slope stability.
Tampere University of Technology.
Phoon, K.-K. and Kulhawy, F.H and
Grigoriu, M.D. (1995). Reliability-based
design of foundations for
transmission lines structures. Electric
Power Research Insti tute, Palo Alto,
Report TR-105000.
Vianova. 2011. Novapoint GeoCalc
Supported Excavation Theory. Versio
2.2., 25.3.2011. Vianova Systems
Finland Oy
208
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Reliability-based assessment of a retaining wall using FEM

Katerina RIPPI 1*, Ana TEIXEIRA1

1
Deltares, Geo-unit, Department of Dike Safety, Delft, THE NETHERLANDS

ABSTRACT
The presented research topic explores the probabilistic assessment of a retaining wall, dealing with
reliability-based analysis of soil-structure interaction using the Finite Element Method (FEM). This
is not only important in the context of an assessment, but also to gain insight and develop the
instruments that allow engineers to use reliability/probabilistic approaches for decision making. The
developments that link FEM and probabilistic methods started a few years ago, and this solution
seems ideal to quantify the hidden conservatism in the usually carried out semi-probabilistic
assessments, since (usually) the employed safety factors are derived from simple models and
limited probabilistic calculations together with several conservative assumptions. The main goals of
this paper are to present the potential of a reliability-based analysis, and guide the reader through
all the implementation, coupling of the FEM model and the reliability analysis, for a retaining wall
example.

Keywords: retaining wall, FEM, probability, assessment, decision making

The current Dutch approach, as

1. INTRODUCTION
The assessment of an existing structure is
a common concept for engineers. In the
particular case of the Netherlands, flood
defense structures are periodically tested
against legal safety standards. These
standards are in terms of maximum
allowable probabilities of flooding, and as
a result, a set of instruments for assessing
the safety of flood defenses were, and still
are, being developed within the project
‘Legal safety assessment 2017’ (Dutch
acronym WBI 2017). The WBI 2017 will
include probabilistic and semi-probabilistic
assessment procedures for the different
types of flood defenses, such as dikes,
dunes and hydraulic structures.
* presenting author
required by the Dutch Flood Defense Act
from 1996, employs these periodic safety
assessments of the existing structures to
warrant an appropriate protection level
from flooding (Jongejan et al., 2013), with
the main advantage of obtaining a
uniformly generated overall picture of the
flood defense system, including
information on the weak spots and the
locations where re-design and
reinforcement is most urgent.
At Deltares, the assessment
framework, corresponding guidelines and
the instrument(s) for the assessment itself,
have been studied and developed for the
next legal assessment, i.e. WBI 2017. In
this context, the presented research
explores the probabilistic assessment of a
hydraulic structure, dealing with reliability-
based analysis of soil-structure interaction
using the Finite Element Method (FEM).
Usually, safety factors derived based
on simple models and limited probabilistic
calculations are employed together with
several conservative assumptions for the
design or the assessment of hydraulic
structures. The developments that link
advanced FEM and probabilistic methods
started few years ago, and they seem an
ideal solution to quantify the hidden
conservatism in the usually carried out
semi-probabilistic assessments (Schwe-
ckendiek, 2007; Wolters, 2012; Rippi,
2015; Teixeira, 2015).
Probabilistic methods are the basis to
develop proper assessment tools to
explicitly handle the different types of
uncertainties (CUR, 1997). These can be
used for various types of engineering
structures, and this paper shows its
application to a case study of a retaining
wall (hydraulic structure) – see case study
in Figure 1.
The paper presents the implementation
and results of a reliability-based
assessment, carried out with Directional
Sampling and First Order Reliability
Method (FORM), for the case study of a
retaining wall (as presented in Figure 1).
Note that the case study is mainly based
on a real structure, which is important for
the lessons learned during the analysis.
The main goals are:
Figure 1. Case study: retaining wall [deformations]
210
(2) to present the potential of a
reliability-based analysis, for example for
an optimized design procedure (when
compared to the current used safety
factors), or for a more rational
quantification of the uncertainties,
avoiding hidden conservatism in the
usually carried out semi-probabilistic
assessments,
(1) and to guide the reader through
some of the challenges of the
implementation of such an analysis,
including the coupling of the FEM model
and the reliability analysis (framework),
choosing the uncertainties and important
failure modes to be taken into account.
2. UNCERTAINTY AND RELIABILITY
BACKGROUND
Most of the times, designs are based on a
semi-probabilistic approach, using safety
factors for load (S) and resistance (R)
parameters. Latest developments and the
possibility to perform numerous
computations, allow reliability based
designs or assessments introducing
uncertainties from all the involved
parameters. It is, therefore, possible to
determine the probability of failure (or the
reliability) of the system, which will lead
towards a more "rational" design, without
the need of safety factors, which
sometimes are not specified on design
codes. In order to obtain a reliability
assessment/design we set up the
performance function(s), define the
random variables and the target reliability.
2.1. Performance Function(s)
A performance function, or limit state
function, defines the ultimate or
serviceability state of a mechanism, which
is the boundary between desirable and
undesirable performance. In general
terms, a limit state function can be written
as shown in eq.(1). When multiple
mechanisms can affect the failure, the
system failure probability is given by eq.(2)
for a series system.
Z RS (1)
pf pf1 ‰ pf 2
(2)
P > Z1  0@ ‰ P > Z 2  0@
Herein, pf is the probability of failure, Z the
limit state function for the sub-
mechanisms and P[i] is defined as the
probability of the event i.
2.2. Reliability methods
A Monte-Carlo approach can be used for
the simulation of the combination of the
mechanisms to determine the probability
of failure, but also the Hohenbichler and
Rackwitz (1983) approach can be used
within FORM (first order reliability
method).
FORM is an approximate method, while
ordinary Monte-Carlo or Directional
Sampling are pure probabilistic methods
with higher accuracy. The Monte-Carlo is
a very straightforward method, while
FORM has some limitations when
complex Z=0 are necessary and/or it is not
possible to approximate with Normal
distributions. On the other hand, influence
coefficients and design point values are
an important and useful by-product of
FORM. With these, one can assess the
influence of each random variable and
choose the necessary number of basic
variables of a problem (random variables
can be reduced without compromising the
the random variables and getting quite
accurate results). During this research,
methods FORM and DS are applied. For
more information on these and more
reliability methods please refer to Rippi
(2015).
2.3. Random Variables
A random variable is a parameter within
the limit state function for which a specific
uncertainty is admitted. Some background
is given for example in Phoon and
Kulhawy (1999a,b) and Jongejan et al.
(2013). General statistical information
regarding soil properties have been
published there and the VNK2-project
(2013) also provides information about the
parameters to be used for water levels
and for soil properties. Usual random
variables (or uncertainties) are associated
to the different soil parameters, their
strength properties, the spatial variability,
the hydraulic or other loads, the theoretical
approaches used to model the failure
mechanism's behaviour and also with the
geometry of the problem.
2.4. Target Reliability
The base of a reliability analysis process
is that the system should achieve an
acceptable reliability target (ȕT). In the
Netherlands, this target is based on the
safety standards formulated in terms of
the “exceedance probability of the
normative load event” (DPV, 2014). For
example, a safety standard of
Pnorm=1/2,000 yr-1 would mean that the
design water level corresponds to an
annual exceedance frequency of 1/2,000
(or a return period of 2,000 years). Since
the reliability is a measure of the
probability that our system does not fail, ȕT
is related with the probability mentioned
above, eq.(3).
°­ pf P > Z  0@  pfT
® (3)
accuracy of the reliability calculation). °̄ ET  ) 1  pfT 
Therefore, Directional Sampling seems
However, some considerations need to
to gather two important advantages which
be made in order to transform the
are the computational time (very optimised
allowable probability of failure/flooding
in comparison with crude Monte-Carlo
(Pnorm) to the allowable probability of failure
simulations) and fully probabilistic method
(taking into account all characteristics of
211
for a specific element of your flood
defence (pfT) – Jongejan et al. (2013).
3. THE COUPLING
3.1. FEM software
The conventional method for a another software program, the reliability
geotechnical stability analysis in a soil
body is represented by the Limit
Equilibrium methods (LEM) although FEM
is increasingly used by researchers. The
latter had been proved quite realistic for
the progressive behaviour (i.e. stress-
strain development in different
construction phases) of a soil system
under the effect of stress redistribution in
comparison with LEM (Johansson &
Sandeman, 2014). Moreover, FEM
seemed to perform also better for the
simulation of retaining walls in comparison
with the classical methods (Blum, Engel,
Krey etc.) (González et al., 2013).
Therefore, in this research, the FEM
simulation is carried out with the
commercial software Plaxis 2D 2015.
3.2. Reliability package
The latest features of Plaxis allow Python
scripting and connection and thus
coupling was carried out with an
uncertainty package available in Python
language: OpenTURNS (OT).
Via OT, the reliability analysis can be
carried out with several reliability methods
(OpenTURNS 1.5, 2015). In this research
Directional Sampling (DS) and the First
Order Reliability Method (FORM) have
been utilized. As referred before, DS is
considered to be a fully probabilistic
method whereas FORM is a probabilistic
method with approximations.
Recall that DS is a method in which
directions are sampled. For each direction,
it is evaluated whereas along a certain
direction, the limit state function equals to
zero or not. This is carried out in
successive steps that are defined by the
user. As far as the FORM is concerned,
the random variables are transformed into
the normalized space and the concept is
based on the approximation of the limit
state function with a linear polynomial.
212
3.3. Coupling FEM-RA
The coupling of reliability analysis (RA)
and FEM requires an interface for the
communication between each other.
When a reliability tool is coupled with
program carries out the whole reliability
analysis and it uses the other program
only for the evaluation of the limit state
function. More precisely, OT interface
should be able to amend the input random
variables and read Plaxis output for
important variables, such as material
parameters, pore pressures generation
and stresses development and
corresponding deformations inside the
dike. Respectively, Plaxis has to be also
capable of obtaining the new values that
have been set by OT for the variables that
are treated as stochastic, during an
iterative process according to the reliability
assessment. In Figure 2, an illustration of
the coupling methodology and its function
is shown.
- RA method
- Random variables
- Correlation matrix
- Limit state function (LSF)
Reliability module
interface
Figure 2. Coupling OpenTURNS <> Plaxis
In principle, an input file is firstly
required where the user sets the
preferable reliability method, the
stochastic input parameters and their
probability distributions, the joint
probability distribution, the corresponding
correlation matrix and finally the limit state
function. However, such input file should
be interpreted so as to be readable by
both Plaxis and OT. Therefore, an “input”
interpreter that helps OT to start up the
reliability analysis according to the
assigned aforementioned definitions, is
necessary.
Keep in mind that the evaluation of the
limit state function is conducted by Plaxis.
For that purpose, an interface should allow
sending the next set of input parameters
from OT to Plaxis. This interpreter also
allows the transfer of the required value of
the limit state function to OT.
4. THE IMPLEMENTATION: RETAINING
WALL
The case study of the retaining wall is
presented in the following figure:
Focus is given to the ultimate limit
state, which describes the situation
wherein the acting extreme loads are just
balanced by the strength of the
construction. If that limit state is exceeded
the construction will lose its functionality
and thus collapse or fail.
Here, the detected elements prone to
failure are the sheet pile, the anchor and
the soil body. The analytical LSF (limit
state functions) are set as in eq.(4). For
multiple limit states the event will be
defined for the minimum value (series
system) of the separate limit states. Note
that the information/values highlighted in
the formulas correspond to outputs of
Plaxis.
­ Z sheet pile
°
Z Z sheet pile ‰ Z anchor ‰ Z soil min ® Z anchor
°Z
¯ soil
Where:
Figure 3. Case study: retaining wall [geometry]
(4.1)
(4.2)
(4.3)
Although in reality all parameters have
a stochastic nature, a high number of
stochastic variables leads to less robust
probabilistic calculation and a significant
increase in necessary computation time.
Furthermore, it is also unnecessary to
consider uncertainties that do not have
significant influence in the probability of
failure. Therefore, in the final RA
calculations a choice is made for a limited
number of soil parameters to be included,
which should correspond to the
uncertainties that mostly influence the
probability of failure result. This choice is
based on expert judgement but also on
previous similar studies.
Random variables within this analysis
included:
x Thickness reduction of the sheet
pile and anchor due to corrosion,
x Soil bottom level at the water side,
(4) x Loading value,
x Soil parameters, e.g. c, phi,
x Structural parameters, e.g. EA, EI.
Also, correlations between e.g. soil
213
parameters are taken into account with a [reliability results]
correlation matrix. Uncertainties that are,
in this case study, not taken into account
include the modelling error, the water level 5. LESSONS LEARNED AND
and the sub-soil division/layering. CONCLUSIONS
Initially, the different mechanisms were Probabilistic reliability analyses provide
studied individually, with the intention to results beyond the metrics strictly required
combine them in a system failure for design and assessment purposes,
probability afterwards (Hohenbichler’s such as the probability of failure or the
method). However, due to several reliability index. A typical by-product is the
reasons, it was decided to proceed with so-called design point and the associated
the reliability analysis that considers the influence coefficients. The latter
three mechanisms simultaneously (i.e. essentially indicates the contribution of
system reliability analysis with DS). One of each (random) variable to the total
the main reasons for this choice was that it uncertainty. This information is useful to
is actually difficult, within a FEM designers because they can infer from it if
environment, to totally separate the failure their focus should be on improving the
mechanisms in an analysis. structural design or reducing uncertainties
It was clear in the analysis that in subsoil conditions. The influence
numerical settings, as in all FEM coefficients can also give clear indications
computations, are very important to of which subsoil strata are dominant in the
achieve realistic and correct modelling failure mechanisms at hand.
results. Furthermore, the computational This paper presented the
time for the reliability analysis was in the implementation of probabilistic analysis to
order of 1 to 2 days for a common i7, a quite complex system, including different
4/8Gb RAM computer. soil and structural elements and an
In the curve shown in Figure 4, one can advanced 2D failure evaluation based on
see the ‘water level vs reliability index’ FEM.
relation, including the confidence interval In the case study of a retaining wall, as
(computed based on the coefficient of studied in this paper, the limit state
variation result of the DS reliability (failure) evaluations are carried out with
analysis). The graph presents 4 points Plaxis, while the reliability analysis is
which represent 4 carried out carried out through a probabilistic and
computations. reliability analysis library (OpenTURNS).
As expected, the higher the reliability The connection between the two is
index is (lower probability of failure) the explained and successfully implemented.
less accurate the DS analysis is. Overall, instability of the structure could
WaterlevelNAP+[m]
not be (robustly) analysed with FORM and
-1.5 -1 -0.5 0 0.5 we resorted to Directional Sampling (DS)
4.4 for the system failure. DS is almost as
4.2 robust in producing results as Crude
Monte-Carlo sampling, with substantially
Reliabilityindex

4
lower computational effort.
3.8
The main conclusion regarding the
3.6 feasibility of reliability analysis of retaining
3.4
walls with FEM from this study is that
results can be produced with some
3.2
simplification in the modelling and that
3 future efforts should address (a) structural
Figure 4. Case study: retaining wall modelling closer to design practice and (b)
further increasing the robustness of the
214
reliability analysis while decreasing
computation times, for example by using
response surfaces.
It is well-known that design recipes in
codes of practices, including the
corresponding partial factors, are
conservative in terms of the actual
reliability they produce on average,
because they need to cover a wide range
of structures and subsoil conditions. As a
consequence, a fully probabilistic design
should (in most cases) lead to a more
economic design than the semi-
probabilistic one. This implies substantial
potential savings. Also, in assessment
situations it may happen that a structure is
not considered safe based on a semi-
probabilistic approach, or there is no
assessment method available at all.
Meanwhile, a fully probabilistic analysis
shows that the target reliability may be
met. Of course, the presumption here is
that the semi-probabilistic design or
assessment method is appropriately
calibrated using fully probabilistic
analyses. For many structures this is not
yet the case. For these structures it is
highly recommendable to carry out such
comparisons or calibration studies to
make sure the designs produced actually
achieve the intended target reliability
levels.
Finally, it is important to mention that
such studies are not only important for this
type of assessments, but it is in general
also important to gain insight and develop
the instruments that allow practicing
engineers to use reliability/probabilistic
approaches for prioritization in the
replacement or reinforcement task(s) of
hydraulic structures, for retrofitting design
as well as for maintenance planning.
REFERENCES
CUR-publication 190. (1997). Probabilities
in civil engineering , Part 1: Probabilistic
design in theory. Stichting CUR,
Gouda, The Netherlands.
DPV (Deelprogramma Veiligheid) (2014).
Synthesedocument achterground-
rapport B1 bij Deltaprogramma 2015. In
215
Dutch. Dutch Ministry of Infrastructure
and the Environment.
González, A., Cuadrado, A., Josa, A. and
Olivella, S. 2013. Safety assessment of
limit equilibrium methods for the design
of cantilever and anchored sheet pile
walls. Escola de Camins, Barcelona,
Spain.
Hohenbichler, M, Rackwitz, R. (1983).
First-order concepts in system
reliability. Structural Safety, 1(3):177–
188.
Johansson, E., and Sandeman, E. 2014.
“Modelling of deep excavation in soft
clay: a comparison of different
calculation methods to in-situ
measurements”, MSc thesis.
Department of Civil and Environmental
Engineering, Chalmers University of
Technology, Sweden.
Jongejan, R. B., Stefess, H., Roode, N.,
Ter Horst, W., & Maaskant, B. (2013).
The VNK2-project: a fully probabilistic
risk analysis for all major levee systems
in The Netherlands. IAHS Publication,
357: Flood.
OpenTURNS 1.5. (2015). Reference
guide: OpenTURNS’ methods for step
C: uncertainty propagation. France.
Phoon, K.-K., Kulhawy, F. H. (1999a).
Characterization of geotechnical
variability. Canadian Geotechnical
Journal, 36:612–624.
Phoon, K.-K., Kulhawy, F. H. (1999b).
Evaluation of geotechnical property
variabaility. Canadian Geotechnical
Journal, 36:625–639.
Rippi, A. (2015). “Structural reliability
analysis of a dike with a sheet pile wall:
Coupling Reliability methods with Finite
Elements”. MSc thesis. Department of
Hydraulic Engineering, TU Delft, The
Netherlands.
Schweckendiek, T (2007). “Structural
Reliability Applied To Deep Excavations
– Coupling Reliability Methods with
Finite Elements”. MSc Thesis, TU Delft,
The Netherlands.
Teixeira, A. et al. (2015). Soil-structure
interaction: Reliability analysis of a
 retaining wall. Report. Project ‘Natte
Kunstwerken van de Toekomst’.
VNK2-project (2013). Van ruwe data tot
overstromingsrisico - Handleiding ter
bepaling van het overstromingsrisico
van dijkringen binnen het project
VNK2,. Tech. Rep. Versienummer: 2.5,
RWS Waterdienst.
Wolters, H. J. (2012). “Reliability of Quay
Walls”. MSc Thesis, TU Delft &
Gemeentewerken Rotterdam, The
Netherlands.

216
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Dry Dock at the Merwede River Banks

Leon SCHADEE*

Fugro GeoServices Geo-Consultancy, the NETHERLANDS

ABSTRACT
The market perspective for the Dutch shipbuilding market is positive. To be competitive and for
future growth Neptune Shipyards wants to build a new dry dock for large river cruise ships. The
location for this dry dock is situated at the river bank of the important shipping route Merwede,
which is a logistically beneficial location. The project location is located in a densely built area. At a
20 m distance a dike is present, as primary flood defence. Due to the dimensions of the dry dock
and the densely built location many boundary conditions have to be taken into account. These
boundary conditions lead to many geotechnical challenges.

This paper highlights the predesign of the dry dock and discusses the key value of the
Geotechnical Engineer in this project.

Keywords: dry dock, sheet piles, deep excavation, PLAXIS, flood defence

1. INTRODUCTION 2. PROJECT LOCATION

Often construction and large maintenance
of ships are performed in a dry
environment. A conventional way for dry
construction is a ship ramp. A more
modern and delicate method is the use of
a dry dock. The geotechnical engineer
plays an important role in the design of
this dry dock. Many critical aspects are
related to soil conditions and soil-structure
interaction. The role and influence of the
geotechnical engineer is crucial in this
predesign phase, i.e. in designing the
walls of the dock, resisting the uplift forces
and analysing the stability of the dike.
The project is located on the banks of the
river Merwede. This river is one of the
important shipping routes in the
Netherlands and is part of the Rhine–
Meuse–Scheldt delta. The Merwede
connects the river Rhine with the river
Meuse. These rivers serve as major
transportation routes to Germany,
Belgium, France and the rest of Europe.
Due to the beneficial location many
companies are located on the banks of the
river Merwede. One of the shipbuilders
wants to build a dry dock for constructing
and repairing ships. Figure 1 shows an
aerial photo of the project location.

* presenting author
 office

PrimaryFloodDefence

Figure 1. Project location with the dry dock marked in green, surrounding structures in yellow
and the primary flood defence in red Bing Maps (2016).
2.1. Site Characterisation Friction, fs [MPa] Friction ratio, Rf [%]

Throughout the centuries the Merwede Cone resistance, qc [MPa]

Inclination

river meandered at this location and

deposited sands and clays. The soil profile
Depth rel. to NAP [m]

consists of soft soils with a dense sand predrilled

layer at depth. In the past this terrain was

created by reclamation. In the “man-made”
top soil layer rubble was encountered.
The current surface level is around NAP
+3.0 m, where NAP is the general
reference level in the Netherlands. Table 1
shows the soil stratigraphy. These soil
conditions are commonly found in this part
of the Netherlands. A typical CPT result is
shown in Figure 2. The future grade level
is at NAP +4.0 m.
Table 1. General Soil Stratigraphy
Top of layer Soil Type
[rel. to NAP] Figure 2. Typical CPT at the project
location.
+3.0 m Sand, with some rubble /
man-made soil From piezometers at ca. 400 m and
800 m behind the dike the groundwater
-1.6 to -2.0 m Sand, medium dense table and the hydraulic head of the first
-4.8 to -6.8 m Peat and Clay water bearing strata are determined. The
groundwater table is at NAP +0.9 m. This
-7.2 to -10.3 m Sand, dense
is the average water level of the Merwede.
-36.5 to -40.7 m Clay, silty At depth a water bearing sand layer is
(overconsolidated) encountered, between approximately
-38.3 to -42.9 m Sand, dense
NAP -7 m to NAP -40 m. The hydraulic
head in this sand layer is connected to the
-45 m Maximum explored depth water level in the Merwede. High water
levels in the river can rise up to
NAP +4.15 m. During these high water
levels the head in first water bearing layer
218
layer can reach a level up to NAP +3.25 explains the excavation methods in more
m. This head corresponds to a once per detail. In both excavation methods a strut
50 years condition. is applied at 3.2 m depth. For the dry
excavation a second strut is used at 7 m
2.2. Soil parameters
depth below surface level. The use of the
The soil investigation is limited as a second strut during the wet excavation is
predesign is made. Soil strength and not necessary due to the beneficial water
stiffness parameters are derived from the pressure inside the building pit during
CPT results. Table 2 presents the effective excavation.
strength parameters of this project. For the
The sheet pile walls of the dock have a
final design additional soil investigation will
different construction. The wall at the
be performed, i.e. CPTs, borings and lab
riverside has to retain only water, where
tests.
the wall at the landside retains both water
2.3. Geometry and soil. In the cross section of the
The dimensions of the dry dock will be 30 excavation the asymmetrical loading can
m wide, 130 m long and 12 m deep. The be seen, see Figure 3 and Figure 4. The
entrance, to float ships into the dock, will sheet piles on the landside will be
be at the south west side of the dock. The anchored using grout anchors. This is a
walls of the dock consist of permanent commonly used anchoring system in the
steel sheet piles. Due to the position of the Netherlands. At the riverside wall no
dock, one wall is retaining soil and water, anchoring is possible, due to lack of a soil
whereas the other wall retains only water. body adjacent to the retaining wall. The
Consequently the lateral load on the dock lateral loads on the riverside wall are
is not symmetrical, which is a somewhat lower compared to the landside
disadvantage in the design. of the dock, as the retaining height of the
soil is much lower.
On top of the dry dock a steel
superstructure will be constructed. The
steel superstructure is 20 m high and
contains a gantry crane for handling large
and heavy objects.
Two excavation methods were
considered, dry and wet. Section 2.5
Table 2. Effective strength parameters for the predesign of the dry dock, based on DKM4
Top of layer Soil Type Junsat / Jsat ij’ (phi) c’
in m depth* in kN/m3 in deg.[°] in kPa
0 Sand, with traces of rumble/ 18 / 20 32,5 -
manmade
-4.6 Sand, medium dense 18 / 20 32,5 -
-8.1 Clay, sandy 18 / 18 22,5 5
-8.6 Sand, medium dense 18 / 20 32,5 -
-9.6 Peat 12 / 12 15,0 2,5
-11.4 Clay, silty 17 / 17 17,5 5
-12.7 Sand, dense 18 / 20 32,5 -
Junsat / Jsat are the dry and wet unit wheight of the soil.
ij’ (phi) is the internal friction angle of the soil.
c’ is the effective cohesion, i.e. shear strength, of the soil.
*surface level equals NAP +4.0 m.
219

Foundation of superstructure
Tension and compression piles
Sheet Piles
Anchors
Surface level
9.15 m below surface level
11.75 m below surface level
Figure 3. Cross section at the landside of
the dry dock, with superstructure, foundation
elements and underwater concrete.
Figure 4. Cross section at the riverside of
the dry dock.
2.4. Special Conditions/Topics
At the Northern part of the project a
primary flood defence is present, i.e. dike.
The Merwede may not be obstructed by
the dry dock. Due to the dry dock’s length
and the nearby waterway a part of the
220
dock will intersect with this dike. The
safety of the polder protected by this dike
may not be impaired. Therefore special
attention is given to the interaction
between the structure and dike. This
challenge is discussed in more detail in
Section 3.
Currently there are several structures
present at the location of the future dry
dock, see Figure 1. Most of the existing
structures will be demolished. The nearby
structures which will be maintained are
marked yellow in Figure 1. The Southern
structures are founded on shallow
foundations. The office at the Northern
side is constructed on a piled foundation.
The buildings on a shallow foundation are
sensitive to deformations and vibrations.
Deformations have to be limited as the
existing structure may not be damaged by
the execution of the dry dock.
2.5. Construction Method
In the predesign phase two construction
methods are considered. One of these
methods is a wet excavation. After the
final depth is reached, underwater
concrete is used to construct a water tight
sealing. The excavation will then be
pumped dry and the dry dock construction
can be finalized.
At the project site no natural watertight
layer was encountered. In order to perform
a dry excavation a horizontal soil injection
technique is proposed to create a water
tight sealing. After the sheet piles are
installed the soil is injected at a minimum
depth of 22,5 m (eq. NAP-18.5 m). This
artificial horizontal waterthight layer is
created from surface level. The horizontal
injection is located between the sheet
piles to create a confined space.
Therefore the sheet piles have to reach
below the minimum injection depth. Using
this artificial layer it is possible to perform
a dry excavation.
Underwater concrete is a large amount
of the construction cost. Compared to
underwater concrete the horizontal soil
injection technique is very competitive, if
the soil conditions meet certain criteria. A
horizontal soil injection can be applied in
sands with sufficient permeability to inject
the water glass mixture which creates a
watertight layer. The sheet piles need to
have a minimum depth to reach beyond
the depth of the horizontal soil injection.
3. CHALLANGES AND
OPPERTUNITIES
One of the important geotechnical
challenges is the asymmetrical loading
condition of the dry dock. In Figure 1 it can
been seen that the dock replaces an
existing quay wall. Figure 3 and Figure 4
show the cross section of the dock. The
different load conditions by the soil and
water are clearly shown.
To limit deformations it is preferable to
apply anchors. The anchors transfer
tension loads into the deep sand layer.
However, only the retaining walls at the
landside can be supplied with anchors. At
the riverside anchors are not possible. The
only lateral support in the final situation of
the retaining wall at the riverside is
through the dock’s floor at 10 m depth.
Besides the asymmetrical loading
conditions other high loads act on the
structure as well. These high loads are
caused by the inclined anchors, the gantry
crane and alternating wind loads acting on
the superstructure. These loads are
transferred to the retaining walls and to
foundation elements at short distance from
the retaining walls. The loads cause a
complex load distribution in the structure.
Another challenge is the interaction with
the primary flood defence. In 2000 the
dike was reinforced by a cofferdam of
sheet piles. The cofferdam consists of two
rows of AZ26 sheet piles. At the outer
dike, i.e. south side, the sheet piles reach
down to NAP -13,0 m. At the inner dike
section the sheet piles reach to NAP
-15,0 m.
Due to the hinterland which is an urban
area, the safety of this dike must always
be guaranteed. In addition large lateral
loads on the retaining walls are caused by
the dike. Furthermore, in the design a
future dike reinforcement has to be
included without additional measures for
221
the dry dock during the reinforcement.
Also, the installation of grout anchors for
the retaining wall underneath the dike,
may not impair the dike’s safety.
In the design process there is a lot of
interaction with the stakeholders, who all
have different interests. In this project
there are four main stakeholders. The
shipbuilder, i.e. future owner of the dock,
wants to build the dry dock as beneficial
as possible. The river Merwede is
controlled and maintained by
Rijkswaterstaat (Dutch Ministry of
Infrastructure and the Environment). The
Water Board is responsible for the
maintenance and safety of the primary
flood defence. The last stakeholder is the
local municipality of Gorichem. The
municipality is responsible for the urban
environment and special planning of the
area. In the end they all have to agree
with the design and construction method
of the dry dock. During the predesign
phase of the project these parties were
already involved, to resolve as many
difficulties as possible in an early stage.
4. DESIGN
This section discusses not all design
aspects, but highlights the most interesting
topics. Besides these topics the
foundation, consisting of compression and
tension piles, is calculated using the
Eurocode 7, the sheet piles are designed
according the Dutch design guideline for
sheet piles (CUR166) and the hydrologic
study is performed using the software
package Watex.
4.1. Retaining Walls
To simulate the interaction due to the
asymmetrical excavation an interaction
calculation was performed. This study is
performed using the software package
D-Sheetpiling of Deltares, which is based
on a spring model to simulate the soil
behaviour. In a later stage the calculations
are verified using FEM-software
PLAXIS 2D.
 Iterative Procedure
The sheet pile walls on the landside and
riverside have been modelled in a
separate model. On the landside anchors
are applied. The supports, i.e. dock floor
and struts if applicable, will transfer the
loads from the riverside to the landside.
To investigate the interaction between
both walls in the cross section the
following iterative procedure is used:
1. Both side walls are analysed without
interaction, using the same
constructive phasing. Supports are
modelled with springs.
2. The loads in the supports on the
riverside are derived and applied on
the landside in the corresponding
phase. This is the first iteration step.
3. From this calculation the
deformation at the level of the
support is derived at the landside.
On the riverside the support’s
stiffness is reduced to meet the
deformation of the support on the
landside.
4. The model on the riverside is
calculated with the less stiff support.
From this new calculation the forces
in the support are derived and again
applied on the landside. This is the
second iteration step.
5. If the difference between the
deformations is sufficiently small the
iteration is stopped. Otherwise the
steps from no. 3. to no. 5. are
repeated until the displacements
meet the criteria.
In this case the displacements on the
riverside should be within 5% of the
previous iteration. This exercise is
performed in the serviceability limit state
(SLS) as displacements are considered.
Figure 5 shows the final model with the
spring supports and corresponding loads.
222
Stempeldruk haven F4-1 Stempel +0,8 (F4)
Stempeldruk haven F4-2 Stempel -4,0 (F4)
K
Ø1220 mm + 2xAZ17-700 S460
Veen
Klei, matig vast
Ø1220 mm + 2xAZ17-700
Zand, matig vast
Injectielaag
Figure 5. Model across the dock, with
spring supports on the riverside (right) and
loads on the landside (left).
Verification with Finite Element
Model
The Finite Element Model (FEM) and
calculation is made using PLAXIS 2D
2015. The geometry, loads and phasing
are equal to the calculation in the spring
model with D-Sheetpiling. In PLAXIS the
soil is modelled using the Hardening Soil
Small Strain Stiffness model. The
geometry is shown is Figure 6.
Figure 6. Plaxis model in final stage, with
structure of dry dock, surface loads and loads
from superstructure.
Table 3 shows the comparison of the
results of both calculations. The spring
model matches or gives an underestimate
of the forces and deformations in the FEM.
The benefit of the FEM is the prediction of
soil deformations behind next to the
excavation as shown in Figure 7
 Table 3. Comparison of results In the final design a stability and
Springmodel and FEM in final stage for wet interaction study will be performed using
excavation. FEM. This study has to prove the safety of
Spring FEM the dike during the different construction
FEM stages and in the final situation. During
model Spring.
this design stage the Water Board will be
Mmax;riverside
478 474 0.99 involved to be certain the design is save
in kNm/m1
and all safety aspects are taken into
umax;riverside account.
29 17 0.60
in mm During the construction phase of the
Mmax;landside dry dock sufficient monitoring has to be
503 287 0.57 done. The aim of the monitoring is to
in kNm/m1
control the safety level of the construction
umax;landside works and the dike.
77 57 0.74
in mm
Sustainability
The anchors will reach into the stability
core of the dike. During future dike
reinforcements the anchors can be an
obstacle. As mentioned before it shall be
possible to reinforce the dike without
severe mitigating measures. The following
aspects are considered in the design:
x If the dike is reinforced by soil, the
anchors do not obstruct the
construction method regarding the
Figure 7. Final result of the finite element reinforcement.
calculation. x If the dike is reinforced by a new
cofferdam construction or retaining wall
4.2. Dike
the anchor do not obstruct this
The use of a free sheet pile wall is not construction method as they are
feasible at the landside or next to the dike. located to deep.
Using anchoring the deformations, sheet
The “design” dike reinforcement is
pile length and size, and thereby the
excpecte to consist of an additional table
hindrance due to installation effects, are
geidht of the dike by approximately a
reduced by a considerable amount.
meter.
As anchoring system a grout anchor
was chosen. This type of anchor can be 5. OUTCOME AND CONCLUSION
installed with non to low relaxation of the
subsoil. Using a specific casing the From this predesign the feasibility is
grouted anchor body can be installed with proven and a general design basis is
a high overpressure, to avoid leakage made. Both excavation methods are
through the dike along the anchor rod. technically possible to execute. The
Properly installed anchors may improve decision whether the horizontal injection or
the stability and reduces the risk of a the underwater concrete is applied will be
leakage. To generate enough tensile based on financial aspects and
strength the anchoring body is located (geotechnical) risks.
approx. 9 m in the Pleistocene sand layer This paper shows a case study where
(15 m below surface level, i.e. NAP complex interactions between structural
-40 m). elements are modelled in a relatively
simplified manner. This approach shows
that the results are comparable with the
223
more advanced finite element models.
However the results from the finite
element model are more comprehensive,
though the simplified method is sufficient
for a predesign.
In this project the geotechnical engineer
was involved in an early stage. Due to the
early involvement of the geotechnical
engineer a number of engineering topics
could be based on proper design
assumptions. Engineering topics are for
example the sheet piles, the foundation
design, the hydrologic conditions of the
area, which are very important regarding
the risk of uplift and the soil-structure
interaction. In this early stage less
conventional building methods could be
investigated and geotechnical risks could
be identified. This all contributed to a more
optimised predesign and an optimal cost
estimate of the building costs. The key
value of the geotechnical engineer during
this project is the insight of geotechnical
risks, mitigation options and proper
assumptions where needed. Which results
in a realistic predesign given the available
information.

ACKNOWLEDGEMENTS
The financial support offered by the
Geotechnical Department of the KIVI
Engineering Society, i.e. Netherlands
Society for Soil Mechanics and
Geotechnical Engineering, is gracefully
acknowledged

REFERENCES
Bing Maps (2016),
http://www.bing.com/maps
Eurocode 7 (EC7), NEN9997-1:2012
Geotechnical
CUR166, Sheet Pile Retaining Structures,
6th edition 2012, CUR SBR.

224
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Design of the Temporary Cofferdam and Permanent Intake

Channel for the Dublin Waste to Energy Facility

Brian G. SEXTON1*
1
AGL Consulting, Suite 2, The Avenue, Beacon Court, Sandyford, Dublin 18, IRELAND

ABSTRACT
The thermal treatment plant for municipal waste in Ringsend is expected to be operational by late
2017. This paper discusses the design of a temporary cofferdam to facilitate construction of the
pump station structure for the plant and the construction of the permanent sheet-pile walls for the
cooling water intake channel, both located within Dublin Port. The cofferdam has been designed as
a fully-interlocked sealed cofferdam partially in open water with the sheet-piles penetrating into
Dublin Port Clay so that the excavation/construction works can be carried out in the dry. Two levels
of temporary internal propping frames have been used to construct the 12m deep cofferdam; the
lower propping arrangement is the biggest ever used by Groundforce in Ireland. The construction
stages have been modelled using the PLAXIS 2D finite element program accounting for soil-
structure interaction and out-of-balance forces across the cofferdam and intake channel.

Keywords: cofferdam, sheet-pile, propping frame, intake channel, anchor

1. INTRODUCTION walls for the cooling water intake channel.

The Dublin Waste to Energy Project is a
Public Private Partnership between Dublin
City Council and Covanta, one of the
world's largest owners/operators of
infrastructure for the conversion of waste
to energy (WtE). The thermal treatment
plant for municipal waste in Poolbeg,
Ringsend is expected to be operational by
late 2017 and will generate energy from
up to 600,000 tonnes of waste per year
(http://dublinwastetoenergy.ie/).
In this paper, the design of two sheet-
pile retaining structures are discussed,
namely a temporary cofferdam to facilitate
construction of the pump station structure
for the plant, and the permanent sheet-pile
* presenting author
Both are located within Dublin Port. The
majority of the paper focuses on the
cofferdam design.
The cofferdam has been designed as a
fully-interlocked sealed cofferdam partially
in open water with the sheet-piles
penetrating into Dublin Port Clay so that
the excavation/construction works can be
carried out in the dry. Penetration into the
Port Clay was necessary to provide a
groundwater cut-off due to the presence of
high permeability gravels above.
The construction stages have been
modelled using the PLAXIS 2D finite
element (FE) program. FE analyses were
required because the cofferdam and
intake channel designs were complicated
by the differential ground and groundwater
levels between the quay and the existing
channel leading to out-of-balance forces
across the cofferdam and intake channel.
FE analyses allow the interaction between
the walls to be considered.
Two levels of temporary internal
propping frames have been used to
construct the 12m deep cofferdam; the
lower propping arrangement is the biggest
ever used by Groundforce in Ireland.
A single level of anchors and an
accompanying waler beam has been used
to support the inner (west) wall of the
intake channel with the outer (east) wall
supported by a permanent propping frame
connecting to the inner (west) wall. The
excavation is up to 15m deep at the
intake. Both structures had to be designed
to the relevant standards for waterfront
structures. Wake waves and sheet-pile
corrosion were considered for the
permanent condition.
2. DESIGN APPROACH
Design has been carried out in
accordance with IS EN 1997-1:2005
(referred to as EC7 hereafter) and the
accompanying Irish National Annex. An
ultimate limit state analysis using Design
Approach 1 of EC7 has been adopted.
This requires two combinations of partial
factors to be considered, namely DA1.C1,
which considers the effect of load
uncertainties, and DA1.C2, which
considers the uncertainty in the
geotechnical parameters. The “single
source” principle has been used.
Design has been carried out using the
finite element computer program PLAXIS
2D 2012 (Brinkgreve et al., 2012) and
checked (for simple conditions only) using
the limit equilibrium program Geocentrix
REWARD 2.7 (Bond & Spencer, 2013).
The REWARD output is not presented in
this paper. Allowance has been made for
a temporary overdig of 0.5m for the
Ultimate Limit State (ULS) during
construction for both the cofferdam and
the intake channel. Phi-c reduction
analyses have also been carried out to
226
verify overall stability and ascertain
possible failure modes. Both undrained
(cohesive soils only) and drained
conditions have been considered. For
both structures, design has been carried
out based on a surcharge load of 20kN/m2
on retained ground.
3. GROUND MODEL & PARAMETERS
3.1. Site Location
The cooling water intake channel and
pump station structures are located in
Ringsend on the south side of the River
Liffey as shown in Figure 1. The ground
conditions in the area typically comprise
3-6m of made ground over recent
estuarine and alluvial deposits (comprising
soft estuarine silts/clays and post glacial
sands and gravels) overlying stiff
laminated glacial clay (Port Clay) over
limestone bedrock.
Site Location
Figure 1. Site Location (Google Maps)
The most recent aerial photograph of
the area available (see Figure 2) shows
two rows of sheet-piles in the channel, i.e.
on the western bank of the channel and
within the channel on the eastern side.
The flow in the channel (cooling water
from the gas-fired power station) can be
clearly observed, and can be seen to flow
through the eastern wall due to the many
holes in it (the eastern wall is in the direct
line of the channel flow and are heavily
corroded). These sheet-piles were
removed where they obstructed the
installation of the sheet-piles for the new
structures (shown indicatively on Figure 2
using yellow shading), thus making it
possible to obtain corroded section
measurements (see Sexton et al., 2016).
Existing
eastern
sheet-pile
Existing wall
western
sheet-pile
wall
Figure 2. Aerial Photograph 2013 (Google
Maps)
3.2. Model Parameters
The soil stratigraphy adopted in the
analyses is summarised in Table 1. The
characteristic properties for each stratum
have been derived using the relevant
corresponding lab test and field test data
(e.g. measured and design standard
penetration test NSPT values are shown in
Figure 3). Where no data were available,
conservative assumed values have been
assigned based on experience. The
properties are given in Table 2. The Mohr
Coulomb model has been used.
Figure 3. Measured and design standard
penetration test NSPT values
227
3.3. Ground, Bed, & Water Levels
The ground level behind the inner (west)
wall of the cofferdam is approximately
+4.0mOD and the channel bed level
outside the outer (east) wall of the
cofferdam was -2.0mOD initially. However,
subsequent to clearing obstructions from
the channel bed along the line of the outer
(east) sheet-pile wall, the bed level was
lowered below -2.0mOD in places (down
to -4.0mOD).
This necessitated the construction of a
berm (to -2.0mOD with a minimum crest
width of 3m) to ensure stability of the
cofferdam and to minimise wall
deflections. This berm required
maintenance and monitoring during
construction due to scouring in the
channel bed.
Tide levels in Dublin Port range from a
Highest Astronomical Tide (HAT) level of
+2.0mOD to a Lowest Astronomical Tide
(LAT) level of -2.6mOD. These tidal
variations were considered in the analyses
in combination with water level
differentials and wake waves.
4. PUMP STATION COFFERDAM
The design of the pump station cofferdam
is as follows:
x The internal dimensions of 35.31m x
12.89m give a minimum clearance of 2m
to the external walls of the pump station,
see Figure 4a.
x Two levels of temporary internal
propping frames (at +2.0mOD and
-3.0mOD) have been used to construct the
12m deep cofferdam, see Figure 5.
x The lower propping arrangement is the
biggest ever used by Groundforce in
Ireland, see Figure 4b.
x The sheet-piles (AZ38-700N for all four
walls, see Figure 6) were 24m long with
top of pile at +5.0mOD (ground level at
+4.0mOD) driven to a minimum toe level
of -19.0mOD in order to give 1.0-2.0m of
embedment into Dublin Port Clay below
an elevation of -17.0mOD to provide a cut-
off for groundwater control.
 Table 1. Design Stratigraphy
Layer Depth to Depth to Elevation at Elevation at Layer
top of stratum bottom of top of bottom of Thickness
(mbgl) stratum stratum (m)
stratum
(mbgl) (mOD)
(mOD)
Made Ground 0.0 4.5 4.0 -0.5 4.5
Upper Sands &
4.5 7.0 -0.5 -3.0 2.5
Gravels
Soft Clay & Silt 7.0 7.9 -3.0 -3.9 0.9
Lower Sands &
7.9 21.0 -3.9 -17.0 13.1
Gravels
Port Clay 21.0 35.0 -17.0 -31.0 14.0
Glacial Gravels 35.0 39.0 -31.0 -35.0 4.0
Limestone 39.0 - -35.0 - -
Table 2. Characteristic Geotechnical Design Parameters
Layer Young’s At-Rest Peak Effective Poisson’s
Modulus, Coefficient Friction Cohesion / Bulk Unit ratio,
E' / Eu of Lateral Angle, Undrained Weight, Q'/ Qu
Earth I'p ( )
o Shear
(MPa) J
Pressure, Strength,
K0 c' / cu (kN/m3)
(kPa)
Made 40 / - 1-sinI'p 34 1/- 21 0.2 / -
Ground
Upper
Sands & 15 / - 1-sinI'p 32 1/- 21 0.2 / -
Gravels
Soft Clay 7.5 / 9 1-sinI'p 30 1 / 30 18 0.2 / 0.495
& Silt
Lower
Sands & 50 / - 1-sinI'p 37 1/- 22 0.2 / -
Gravels
Port Clay 35 / 45 0.6 32 1 / 100 20.5 0.2 / 0.495

Glacial 100 / - 1.0 38 1/- 22 0.2 / -

Gravels
Limestone 2000 / - 1.0 50 250 / - 25 0.2 / -

x Sheet-pile interlocks were sealed with
bituminous sealant to significantly reduce
groundwater infiltration and seawater
ingress.
x The water temperatures in the channel
were much higher than expected due to
228
the Power Station Cooling Water Outfall.
This resulted in softening of the
bituminous sealant between the clutches
in places. To combat this, a gravel drain
was introduced in place of approximately
0.5m of the base slab inside the east,
south, and north walls of the cofferdam
(see Figure 7). Additional PLAXIS
analyses were carried out to verify the
stability of the cofferdam. To reduce the
east-west lateral movement of the base
slab, a 10m length of concrete propping
slab has been used at the east wall over
the central section of the cofferdam.
x 2 No. dewatering/relief wells (well
diameter = 300mm, well base at -
22.0mOD) were installed so that
dewatering could be maintained both
during excavation and until the cofferdam
was backfilled around the completed
Pump Station.
x Excavation within the cofferdam was
carried out in stages to base slab
formation level (-7.35mOD). The base slab
is 1.1m thick.
x An alternative scenario was considered
for the cofferdam whereby the upper
propping frame was replaced with a series
of ground anchors and an accompanying
waler beam. However, the former option
was chosen because it allows for a
quicker construction time, despite lower
ground movements for anchor/waler
solution.
x The sheet-pile walls, props, and base
slab have been modelled as plate
elements. The props have been modelled
with “pinned” connections to the sheet-pile
walls (i.e. with no moment transfer).

Figure 4. Pump Station Cofferdam (a) Aerial Photograph (BAM Civil) (b) Groundforce Props

Prop @ +2mOD Prop @ -3mOD

Berm
HAT @ +2.0mOD

-7.85mOD (including 0.5m overdig)

-19.0mOD

Internal Wall
External Wall

Figure 5. Pump Station Cofferdam: Excavation to -7.85mOD (including 0.5m overdig)

229
 Figure 6. AZ38-700N Sheet-Pile
Figure 7. Gravel drain to east, south, and
north walls of cofferdam
5. FE ANALYSES RESULTS
The results of the numerical analyses
were used to determine the maximum
design shear forces, axial forces, and
bending moments in the sheet-pile walls
and the maximum design loads in the
props. The maximum horizontal wall
deflections and maximum
horizontal/vertical ground movements
have also been determined. A selection of
the outputs are presented in this section.
5.1. Sheet-Pile Walls
The maximum design bending moments
and horizontal wall deflections (for the
most onerous design situation) for the
outer (east) wall of the cofferdam are
presented in Figures 8 and 9. Plots of
maximum design shear forces and axial
forces are not presented due to space
limitations. The analyses outputs were
checked against the design bending
moment, shear force, and axial force
resistances to confirm that the AZ38-700N
sheet-pile section was acceptable.
230
Max. = 1437.0kNm/m
Figure 8. Pump Station Cofferdam Outer
(East) Wall Bending Moment Profile
Max. = 243.9mm
Figure 9. Pump Station Cofferdam Outer
(East) Wall Horizontal Deflection Profile
5.2. Ground Movements
Displacement contour plots were used to
establish the maximum vertical and
horizontal ground movements behind the
cofferdam and intake channel walls. The
horizontal ground movements behind the
inner (west) wall of the cofferdam are
shown in Figure 10. The deformed mesh
is shown in Figure 11. This shows that the
critical mode of deformation is a “racking-
over” failure mechanism. The inner wall is
supported by the propping frame
connecting to the outer wall; FE analyses
were necessary to capture the soil-
structure interaction behaviour.
For the cofferdam, the maximum SLS
ground surface settlements and lateral
movements in the ground retained by the
inner (west) wall were found to be 341mm
and 279mm, respectively. Horizontal and
vertical ground movements 1m below
ground level are plotted against distance
behind the inner (west) wall in Figures 12a
and 12b respectively. The maximum
horizontal ground movements occur at the
cofferdam walls and reduce to <10mm at
a distance of about 15m from the
cofferdam west wall. The nearest structure
to the cofferdam is at a distance of ~30m.
The vertical ground movements at
distances greater than 15m from the
cofferdam are caused by the 20kPa
surcharge.

Figure 10. Pump Station Cofferdam - Horizontal Displacement Contour Plot

Figure 11. Pump Station Cofferdam - Deformed Mesh

Figure 12. Ground movements behind the inner (west) wall 1m below ground level (a) horizontal
(b) vertical
231
6. COMPARISON OF FE ANALYSES
OUTPUTS WITH WALL DEFLECTION
MEASUREMENTS
Measurements of horizontal wall deflection
for the outer (east) wall of the cofferdam
have been taken by both BAM Civil and
AGL Consulting. The measurements were
taken with an inclinometer accurate to
±1mm. The wall deflection measurements
are compared to the FE output in Figure
13 for the relevant construction stage. It
was found that the wall deflection
measurements were in good agreement
with the PLAXIS output when the berm of
granular material to -2mOD (see Section
3.3) was intact. However, the wall
deflections exceeded the design
predictions where scouring of the berm
had occurred. Corrective measures were
taken by filling outside the outer (east) wall
of the cofferdam to the design level of
-2.0mOD.
Figure 13. Pump Station Cofferdam Outer
(East) Wall - Comparison of FE Output with
Wall Deflection Measurements
232
7. COOLING WATER INTAKE
CHANNEL
The design of the cooling water intake
channel is as follows:
x A single level of anchors (drilled
perpendicular to the wall at 30o to the
horizontal) and an accompanying waler
beam above high water level (at
+2.5mOD) have been used to support the
inner (west) wall of the intake channel with
the outer (east) wall supported by a
permanent propping frame (at +2.85mOD)
connecting to the inner (west) wall, see
Figure 14. The excavation is up to 15m
deep at the intake.
x AZ46-700N sheet-piles have been used
for the inner (west) wall (which has to
retain soil and water) and AZ38-700N
sheet-piles have been used for the outer
(east) wall (water only).
x The ground anchors (with double
corrosion protection for the permanent
condition) have been designed as 57.5mm
GEWI Plus Bars drilled at max. 2.1m
centres. They have been designed in
accordance with EC7, BS8081:1999
(Code of practice for ground anchorages),
and IS EN1537:2013 (Execution of special
geotechnical works. Ground anchors).
x The ground anchors have been
modelled as “node-to-node” anchors over
their free (unbonded) length of 13m with a
stiffness (EA) equivalent to the anchor bar
tendon divided by the anchor spacing. The
fixed length has been modelled as a
“geotextile” with a stiffness (EA) equivalent
to the anchor bar tendon divided by the
anchor spacing.
x The sheet-piles, waler beam, and
propping frame have been designed with
a sacrificial steel thickness for corrosion
allowance.
x Note that, based on the site
investigation information, slightly different
stratigraphies have been adopted for the
cofferdam and intake channel analyses.
 Prop @ +2.85mOD
HAT @ +2.0mOD
Anchor @ +2.5mOD

-10.8mOD (including 0.5m overdig)

-19.0mOD

Internal Wall External Wall

Figure 14. Cooling Water Intake Channel: Excavation to -10.8mOD (including 0.5m overdig)

8. COOLING WATER INTAKE for a wave 1.5m above a mean water level
CHANNEL ANALYSES of +2.0mOD (i.e. HAT). This simplified
distribution of water pressure has been
8.1. FE Analyses Outputs
compared (and shown to be conservative)
As for the cofferdam, the results of the with pressure distributions for waves
numerical analyses were used to establish calculated using the method of Sainflou
the maximum design loads in the props (1928), assuming that the mean water
and to confirm that the AZ46-700N and level during the wave occurrence
AZ38-700N sheet-pile sections were corresponds to HAT.
satisfactory for shear force, axial force,
and bending moment. In this case, the 8.3. Corrosion Requirements
maximum design anchor loads were also Corrosion is typically accounted for in the
determined. design of sheet-piled wall structures by
The maximum horizontal wall allowing for a reduction in thickness of the
deflections and maximum sheet-pile section during the lifetime of the
horizontal/vertical ground movements structure. The corrosion rates depend on,
were also checked. In this case, the wall among other factors,
deflections and ground movements were groundwater/seawater conditions and soil
much lower than for the cofferdam owing type. Different corrosion rates are
to the anchor/waler solution. The results applicable in different zones of the sheet-
will not be presented in this paper. pile, e.g. Figure 15.
Instead, some particular design issues
relating to wave loading and to long-term
corrosion design will be mentioned.
8.2. Wave Loading
Due to its more exposed location in the
port, an additional water pressure from a
1.5m high wake wave has been analysed
for the outer (east) wall of the permanent
channel. This has been considered in the
PLAXIS analyses as an increase in
hydrostatic pressure based on an
additional 1.5m of head on the external
(eastern) side of the outer (east) wall. The Figure 15. Example of corrosion rate
most onerous condition considered was distribution (IS EN 1993-5:2007)
233
General corrosion rates are given in
Tables 4.1 and 4.2 of IS EN 1993-5:2007.
The works requirements specified a
minimum design life of 50 years for the
new permanent intake channel and so the
relevant corrosion rates were accounted
for in the design. For the long-term
condition, accounting for corrosion, the
design resistances have calculated based
on corroded section moduli which were
interpolated/extrapolated from charts by
Arcelor (Arcelor Mittal Piling Handbook, 8th
Edition), e.g. Figure 16.
Figure 16. Elastic section modulus versus loss
of thickness for AZ piles (Arcelor Mittal Piling
Handbook, 8th Edition)
9. CONCLUSIONS
In this paper, the design of a temporary
cofferdam and permanent anchored
sheet-pile wall for the cooling water intake
channel for the Ringsend thermal
treatment plant are discussed. The
designs necessitated FE analyses as the
soil structure interactions and the
differential ground and groundwater levels
between the quay and the existing
channel leads to out-of-balance forces
across the cofferdam and intake channel.
Measurements of horizontal wall
deflection for the outer (east) wall of the
234
cofferdam were found to be in good
agreement with the PLAXIS output when
the berm of granular material to -2.0mOD
that was specified in the design was intact.
However, the wall deflections exceeded
the design predictions where scouring of
the berm had occurred. Corrective
measures were taken by filling outside the
outer (east) wall of the cofferdam to the
design level of -2.0mOD.
ACKNOWLEDGEMENTS
The author would like to thank BAM Civil
and Covanta for permission to use the
data presented in the paper. The author is
also very grateful to Dr. David Gill of AGL
Consulting for his valuable input.
REFERENCES
Arcelor Mittal (2008) Arcelor Mittal Piling
Handbook, 8th Edition.
Bond, A. & Spencer, I. (2013) Geocentrix
REWARD 2.7, Geocentrix Ltd., UK.
Brinkgreve, R.B.J., Engin, E. & Swolfs,
W.M. (2012) PLAXIS 2D 2012, PLAXIS
B.V., The Netherlands.
BS8081:1999 (Code of practice for ground
anchorages).
http://dublinwastetoenergy.ie/ Accessed:
29/12/2015.
IS EN1537:2013 (Execution of special
geotechnical works. Ground anchors).
IS EN 1993-5:2007 (Eurocode 3: Design
of Steel Structures - Part 5: Piling).
IS EN:1997-1:2005 (Eurocode 7:
Geotechnical Design - Part 1: General
Rules).
Sainflou, M. (1928) "Essai sur les Digues
Maritimes Verticales. Annales des
Ponts et Chaussees", Vol. 98, as
described in Recommendations of the
Committee for Waterfront Structures
Harbours and Waterways, EAU 2004.
Sexton, B.G., Gill, D.R. & O'Donnell, C.J.
(2016) "Sheet-pile corrosion rates
within an existing outfall channel in
Dublin Port", Civil Engineering
Research in Ireland 2016 (CERI2016),
National University of Ireland, Galway,
29-30 August 2016, In Press.
25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Session 5

Numerical/physical modelling
236
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Investigation of the influence of gravity on granular flow using

silo centrifuge model

Olja BARBIR 1*, John MATHEWS2

1
Vienna University of Technology, Institute of Geotechnics, Vienna, AUSTRIA
2
University of Natural Resources and Life Sciences, Institute of Geotechnics, Vienna,
AUSTRIA

ABSTRACT
The flow of granular material through an orifice has been a subject of numerous studies for
decades due to its importance for proper industrial silo design. The most widely accepted law that
predicts the flow rate of grains trough an orifice was proposed by Beverloo et al. (1961).
The velocity field within the silo during discharge cannot be reliably predicted yet, and the
mechanisms controlling discharge from silos remain unclear. In this work an investigation into
gravity discharge from silos in increased gravity conditions is presented. Increased gravity
conditions were modelled in the geotechnical centrifuge, where the gravity was increased up to the
factor of 15. Materials differing in particle sizes were tested using two different silo model
geometries. The results were then compared to Beverloo equation for a slit orifice of a quasi-two-
dimensional silo, as well as with the equation presented by Rose and Tanaka (1956) regarding
influence of hopper angle on the flow rate.
It was shown that the mass flow rate as well as the local velocity of discharging is proportional
to the square root of gravity and that the time required for a silo model to discharge cohesionless
material scales with gravity. Analysis of the time required to discharge a silo leads to the
observation of a scaling law for silo centrifuge models.

Keywords: granular flow, silo design, centrifuge modelling, Beverloo equation

information that was needed for describing

1. INTRODUCTION
Silos are thin-walled structures used for
storing granular materials of various kinds,
and are as such subject to many different
and erratic loading conditions which can
lead to damage or even failure of the
entire structure. At the beginning of the
19th century, the grains stored in wooden
forerunners of silos were considered to
behave like liquids, and the only
* presenting author
their behaviour was the material’s specific
weight. This incorrect assumption (as
suggested in Figure 1) led to a great deal
of silo malfunctions.
Granular materials consist of a large
number of individual particles, whose
mechanical properties can usually be fairly
easily obtained, whilst the collective
behaviour of the system remains highly
complex due to the interaction between
the particles. No universally applicable
theory that can precisely describe the
movement of bulk solids and its interaction
with the silo construction has yet been
developed. A good definition of the flow
pattern is crucial for a safe and economic
silo design.
The empirical relationship describing
the flow of granular solids through an
orifice by Beverloo (1961) has been
verified for different granular materials in
the literature. However, most publications
deal with flow under Earth's gravitation,
i.e. in a 1g1 environment.
Figure 1. Difference between: a) pressure on
silo wall caused by liquid material, b) pressure
caused by bulk solids (Janssen’s theory2)
In this work, the validity of Beverloo's
relationship under elevating gravity in a
geotechnical centrifuge is studied. The
primary objectives of this work are as
follows:
x Observing the flow of fine and
course sand in a flat-bottom and
30° hopper silo model designed
especially for this purpose.
x Exporting data of the flow in term of
load cell values for later
comparison with other methods.
x Comparing test results with the
Beverloo equation.
x Showing the discharge time
dependence on gravity factor.
1
Gravitational acceleration of 9.80665 m/s2
2
Janssen’s theory (Janssen 1895) is widely
accepted for pressures during filling state of silo,
assuming static equilibrium, and stating that with
granular materials the rate of increase of pressure
decreases with depth.
238
x Showing the discharge velocity
dependence on gravity factor.
x Showing the flow rate dependence
on the hopper angle and material
particle size.
2. STATE OF THE ART
2.1. Silo design
There are three major aspects that need
to be considered when designing silos for
storage of bulk solids. These are the bulk
material, geometric and structural
considerations. Frictional and cohesive
properties of bulk solids are different from
one solid to another, which influences the
bulk solid’s flow properties. Problems
related to geometric design include
ratholing through the material and arching
across an outlet. The optimal geometric
design would maximize the usable
capacity and minimize costs and silo
overall height. The distribution of
pressures and stresses on silo wall and
the changes that occur during static and
dynamic state of the stored material
represent the main problems of structural
design. The objective of silo design is to
predict a proper flow profile and determine
optimum hopper angle and the minimum
outlet dimension for trouble-free flow
(Carson and Jenkyn, 1993).
Main reasons from which most of the
problems with silo behavior arise are:
x The interaction of the silo wall and
the material stored inside.
x The complexity of bulk solid
materials.
x “Silo music” – noise pollution
caused by the vibration during silo
discharge that can cause structural
failure.
x Dust explosion - Movement of the
bulk material (depending on the
dust amount stored inside of a silo)
leads to the formation of dust
clouds that pose an explosion
hazard when in presence of an
ignition source.
2.2. Flow pattern
The flow pattern may be defined as
description of the zones of moving and of
stationary solid when the silo is still
effectively full, but the discharge process
is well established (Zhong et al. 2001).
The pattern of the granular material flow
affects both functional and structural
integrity of the silo, as well as the
pressures exerted on the silo wall during
discharge, and is, because of that, one of
the most important predictions to be made
during the silo design. Flow patterns
during discharge of the material can be
divided into two basic categories (acc. To
EN 1991-4 (2006)):
x Mass flow – whole silo content (every
particle) moves during discharge. Flow
is uniform and reliable, eliminating
stagnant materials and the effects of
segregation.
x Funnel flow - only a portion of the bulk
solid in the silo moves downwards
through a flow channel during
discharge while the rest of the bulk
solid remains stationary thus forming
stagnant zones. Funnel flow can be
subdivided into pipe and mixed flow.
2.3. Centrifuge testing
Geotechnical centrifuge modeling is a
technique used for testing geotechnical
engineering models by rotating the
observed model and thereby increasing
the gravity so that the stresses in the
model and in the prototype are
approximately equal. Models are usually
made for particular field of observation like
slope stability, shallow tunneling and silo
construction. This way the real in-situ
behavior is simulated.
2.4. Beverloo’s flow rate equation
The most widely used method for
predicting silo discharge rate was
developed by Beverloo (1961):
5
W CUb g (D0  kd ) 2
This equation is valid for the outpouring of
grains due to gravity, where W is the mass
flow rate, Ub is the density of granular
239
material, g is the acceleration due to
gravity, D0 is the orifice diameter, d is the
average grain diameter, and C and k are
fitting parameters. This correlation
includes (D0-kd) factor, consistent with the
“empty annulus concept” by Brown and
Richards (1970), where no particle center
can approach the orifice edge within the
distance of d/2 and therefore all particles
centers must pass through a circle of
diameter (D0-kd).
2.5. Particle image velocimetry
Particle image velocimetry (PIV) is defined
as a technique for velocity measurement
that measures the motion of marked
regions of fluids by observing the locations
of the images of markers at different
times. When measuring the velocity of the
granular materials, the optical surface
structure, which is gained by illumination
of the surface of the granular flow, is
sufficient to determine the movement of
the particles. This method is referred to as
“granular PIV”.
3. EUROCODE
With the development of silo design, a
need for regulation became apparent. The
first structural standards were not
produced until the 20th century, and only
in the second half the first silo design
standards were introduced based on
Janssen’s, Reimbert brothers’, Jenike’s
and other models. Current standards used
in silo design proposed by the European
Union are:
x EN 1991-4 (2006), Action on
structure - Silos and tanks
x EN 1993-4-1 (2007), Design of
steel structures - Steel silos
x EN 1998-4 (2006), Design of
structures for earthquake
resistance - Silos, tanks and
pipelines
4. EXPERIMENTAL SETUP
Experiments described in this work took
place in the beam centrifuge at the IGT,
University of Natural Recourses and
Applied Sciences in Vienna, manufactured
by Trio-tech, California in 1989. Table 1
lists its specifications. For safety reasons,
the centrifuge was placed below ground
level and contained within a metal shell to
add protection in case a component or
test specimen becomes loose during flight.
Experiments are conducted in order to
investigate the influence of particle size
and hopper angle on the flow rate. The
silo model is equipped with a transparent
front wall and a high-speed camera, which
allow the usage of the PIV technique.
Additionally, load cells are used for
tracking the mass of material as it is
discharges from the model. The
investigation of flow rate dependence on
the hopper angle is conducted using two
different silo model geometries. The test
results are obtained at high gravity levels
of up to 15g simulated by the centrifuge.
Table 1. Technical specification of IGT Beam
Centrifuge
This scaling law was investigated by
comparison with the model test results
and shown to be true.
4.2. Silo model design
The silo centrifuge model is designed to
behave as a quasi-two-dimensional silo
with the following dimension - 15 cm
width,10 cm depth and 29 cm height.
Figure 3 shows a sketch of the model and
its constitutive parts:
(1) Silo model with the acrylic window
(2) Side wall
(3) Filling funnel
(4) High-speed camera
(5)(6) LED Illumination
(7) Camera stand
(8) Collection bucket
In order to initiate the discharge, a
servo motor is used by pulling a pin which
releases a spring loaded sliding door and
opens the silo outlet. The location of the
outlet was chosen as the lowest possible
height where the silo could be expected to
discharge completely and only under the
influence of gravity for a range of granular
materials tested. The model can
accommodate a hopper of any angle, but
only flat bottom and a 30° hopper were
used for this research (Figure 2).

4.1. Verification of centrifugal

acceleration
The Beverloo correlation predicts that if
two otherwise identical silos are
discharging freely under the action of
gravity then the discharge time at
increased gravities will be equal to the
discharge time at 1g multiplied by the
square root of the factor of gravity
Figure 2. Silo geometries used in the tests;
increase: silo with a 30° hopper (left), flat bottom silo
W1 g1 m (right)
, W
W2 g2 t The rate of the material discharging from
the silo model is measured using load
cells located beneath the collection bucket
and the high speed-camera records the
t1 moving of the particles behind the
t2 
N transparent front wall for a later PIV
analysis.

240

Figure 3. The silo model sketch
Figure 4. Fine sand "DIN EN 12904"
(scale in mm)
Figure 5. Grain size distribution of fine sand
241
4.3. Granular materials
In order to investigate the influence of the
particle size on the flow rate two types of
sand are used for the centrifugal
experiments: poorly graded fine sand (DIN
EN 12904) and a poorly graded course
sand (DIN 1164/58). Practical usage of
fine sand is water filtration, filling
compound between the borehole wall and
the construction pipe and for the support
of bored holes for water extraction. Figure
4 presents the grain size of the material;
Table 2 describes its properties and
Figure 5 the grain size distribution.
Second material is course grained silica
sand “DIN EN 1164/58”, seen in Figure 6,
with the specifications presented in Table
3 and the grain size distribution in Figure
7. Both materials are poured in through
the funnel and are therefore in a loose
condition, not compacted.
Figure 6. Course sand "DIN EN 1164/58"
(scale in mm)
Figure 7. Grain size distribution of course
sand
 Table 2. Properties of fine sand
4.4. Experimental procedure
The silo model is filled with
previously weighted material trough a
funnel at 1g. Every experiment was
labelled, and recorded with the high-speed
camera. After filling the model, the
centrifuge safety switch can be switched
on, and the CATMAN software recording
load cells data can be started. At this
point, the centrifuge room needs to be
vacated and all the proceedings have to
be performed from the control room for
safety reasons. After the centrifuge is
accelerated to the desired angular
velocity, the needle can be pulled out of
the trap door, and the silo model can be
emptied. After the discharge is completed,
the centrifuge is stopped, and the
procedure can be repeated. Tests at a
specific acceleration are repeated until 3
tests with less than 5% variation in
discharge rate are obtained. The load
cells, as well as the PIV values were
calibrated prior to usage.
4.5. Load cells
Two C9B 1 kN miniature load cells from
Hottinger Baldwin Messtechnik (HBM) are
Figure 8. Typical test data recorded by the load cells
242
Table 3. Properties of course sand
used for controlling the mass flow rate.
The load cells are located beneath the
collecting bin on each end measuring the
mass of the bin and the sand discharging
from the model. After discharge, the mass
of the collecting bin with the sand must
correspond to the value of mass recorded
with the load cells. This way, the accuracy
of the load cell results is verified. The load
cells are stabilized in order to ensure the
vertical loads to be transferred directly.
Typical test data recorded by the load
cells are shown in Figure 8.
4.6. Results analysis using PIV
PIV open-source software PIVlab 1.32
running in MATLAB 2011 is used for
results analysis. High speed camera fixed
in front of transparent acrylic front wall
allows the movement of the material inside
of the silo to be recorded. This is a second
method of calculating the discharge rate.
Velocity vectors across the width of the
silo are calculated using image processing
algorithms. The processing time is
additionally reduced by observing only the
discharge interval sufficient for valid
analysis due of the constant flow rate.
 3
5. CENTRIFUGE TEST RESULTS
W CUb g (b  kd )(D0  kd ) 2
Two materials are each tested at 4
gravities corresponding to 1g, 5g, 10g and Where b is the thickness of the silo.
15g. Experiments are conducted in a silo
with a flat bottom and a silo with a hopper
inclined 30 degrees to the vertical. Each
test was repeated at least 3 times so that
the repeatability of the test is assured and
the results are than compared to the
Beverloo correlation.
5.1. PIV results
By using the PIV software it is possible to
extract the following data:
x Velocity magnitude [m/s]
x “u” component of material velocity
[m/s] (vertical to the outlet)
x “v” component of material velocity
[m/s] (horizontal to the outlet)
For the mass flow rate of the material
that is observed, the most important thing
is the velocity of the material in the vertical
direction. This gives a better insight of only
the vertical motion of sand particles
without the horizontal component that
does not influence the mass flow. The flow
profiles can be observed and compared at Figure 9. Contour map of material velocity
any desired height of the silo. [m/s]
The increase in gravity between tests Discharge rates for each test are
occurs due to different angular velocities gained from load cell and PIV results, and
and can be visually as well as analytically compared with Beverloo´s correlation. The
derived from the contour maps (Figure 9). two empirical coefficients C and k are
The velocity of the material flowing from determined experimentally for every kind
the top towards the outlet of the silo of grains and container properties.
increases with the increase of gravity: Discharge coefficient C depends on bulk
density and is taken with a value of 1. The
W v v˜A
value of k has been found to be
A1 A2 , v 2gh independent of the size of the particle but
depends on particle shape and hopper
v1 g1 properties. Parameter k is experimentally
N determined to have a value of 1.5 for
v2 g2
spherical particles, but due to the “empty
5.2. Discharge flow rates annulus concept” and in order to gain the
effective diameter, a value of 1 was taken.
As stated before, the flow rate during
This gives good concurrence of the results
discharge is one of the most important
as can be observed in Figures 10-13.
aspects of the silo behavior, as it was an
important subject of early research. The Beverloo gives a maximum of 8%
Beverloo equation, originally only valid for deviation from observed test results.
large circular orifices can be adjusted for a
slot orifice of a quasi-two-dimensional silo:
243
 Due to friction between the material and
the front silo wall, it can be observed that
the values gained from PIV are mostly
lower than those from the load cells.
5.3. Flow rate dependence on the
particle size
In order to compare the flow rate of two
materials with two different particle sizes,
Beverloo equation is used, which takes in
consideration the density of the material
and therefore also the material particle
Figure 10. Discharge rates, 30° hopper,
size. Results gained with Beverloo
course sand
correlation are compared with experiments
results. The observed difference between
each material and its Beverloo equation
has a maximum value of 6%, with an
average difference of 2.7%, which shows
a good prediction made by Beverloo as
well as reliable test results (Figures 15
and 16).
In general, higher results gained from
Beverloo equation in comparison to test
results can be attributed to several
simplifications that are made by proposing
Figure 11. Discharge rates, flat bottom, course
that k has a value of 1:
sand
x The effective diameter is D 0  d

x The center of mass of a particle is

its geometric center
x The particle is effectively spherical
The argument that no particle center
can approach orifice edge within the
distance of d/2, and because of that all
particle centers must pass through a
region of diameter (D0-d), is not valid
Figure 12. Discharge rates, 30° hopper, fine
because it presumes that the mass of the
sand
particle is concentrated at the center, and
does not allow the value to be greater
than 1.
Each of these simplifications creates an
assumption that the material is uniform,
isotropic, and most importantly,
geometrically homogeneous which cannot
be stated for any of the real materials that
are usually stored inside a silo.
In order to get a better insight of the
differences between Beverloo and the
Figure 13. Discharge rates, flat bottom, fine measured discharge rates, all of the
sand
244
results given above are plotted again.
Load cell and PIV results are presented as
fractions of the value gained with
Beverloo`s correlation (e.g. Figure 14).
Figure 14. Discharge rates in respect to
Beverloo (flat bottom, course sand)
Figure 15. Comparison of mass flow rates for
30° hopper, fine and course sand
Figure 16. Comparison of mass flow rates for
flat bottom silo, fine and course sand
245
5.4. Flow rate dependence on the
hopper angle
Beverloo´s equation, however, does not
consider the influence of the hopper angle
on the flow rate. When replacing the
stagnant zone of the material with wedges
(Figure 17) that allow the material to slide
in a more uniform way due to decrease in
friction, the flow rate increases.
Figure 17. Definition of symbols: 30° silo (left),
flat bottom silo (right), D is the half-angle of
the hopper, D0 is the diameter of the orifice
Rose and Tanaka (1956) investigated
the increase in discharge rate for silos with
different hopper half angles. They
presented an empirical correlation, stating
that the mass flow rate is proportional to
(tan D, tan Md ) 0,35 , provided D  90º Md :
F(D, Md ) (tan D, tan Md ) 0,35
for D  90º Md
F 1,0
for D t 90º Md
The correlation can be incorporated into
the Beverloo correlation in following form:
W WBeverloo ˜ F(D, Md )
Where WBeverloo is the mass flow rate
predicted by Beverloo correlation and Md
is the angle between the stagnant zone
boundary and the horizontal that cannot
be predicted and should therefore be
directly observed.
Taking the geometry of the silo used in
test execution this equation would give a
21.2 % increase of the flow rate for the silo
with a 30° hopper in comparison with the
flat bottom silo. Experiment results
comparing flat bottom and 30° hopper also
show an increase of flow rate but with a
smaller percentage of 11% for fine sand,
and 5% for course sand.
When observing PIV results
regarding velocity magnitude of the
material, it can be seen that the hopper
angle increases the velocities of the
discharging material at the sides of the silo
but does not increase the peak velocity,
which is governed by gravity.
6. CONCLUSIONS
Silo discharge rate is shown to be
proportional to the square root of the
gravity. Strong correlation with Beverloo
equation can be seen in the test results
regarding flow rate of different materials
with variation in particle sizes through the
orifice, even in higher gravity levels of up
to 15g. The results are quantified using
two independent instrumentation types
and therefore confirm that the equation
still gives good results when gravity is
changed. Deviations between results are
most likely present due to the
simplifications made by Beverloo
regarding particle shape and size, and the
homogeneity of the material.
Disagreements between the physical
model results and results gained through
Rose and Tanaka equation can be
attributed to the fact that the value of Md
must be predicted experimentally since
there is no reliable method of its
determination. This is certainly an issue
that requires more work, and can be
considered a challenge for further
investigation and discussion.
ACKNOWLEDGEMENTS
The authors of the paper would like to
express their gratitude to their supervisor
during the writing of their thesis, Professor
Wei Wu.
The presenting author would like to
thank her current supervisor, Professor
246
Dietmar Adam, for giving her the
possibility to present this work.
REFERENCES
Barbir, O., Štimac, B., “Influence of gravity
on granular flow in silo centrifuge
models in comparison with Beverloo
equation”, Institute of Geotechnics,
University of Natural Resources and
Life Sciences, Vienna
Beverloo, W.A., Leniger, H.A. and Velde,
J. (1961), „The flow of granular solids
through an orifice”, Chemical
engineering science, vol.15 (3-4),
September 1961, 260-269
Brown, R.L., Richards, J.C. (1970),
„Principles of powder mechanics”,
Pergamon, Oxford
Carson, J.W., Jenkyn, R.T. (1993), “Load
development and structural
considerations in silo design”, Available
from: http://www.jenike.com, [Accessed
on June 20, 2013]
Choi, J., Kudrolli, A., and Bazant, M. Z.
(2005), “Velocity profile of granular
flows inside silos and hoppers”,
Physics-Condensed Matter, 17-24
Großstück, M. and Schwedes, J. (2005),
“Application of model tests on cohesive
bulk solids using silo centrifuge
technique”, Powder technology, 157(1–
3), September 2005, 149-155
Janssen, H.A. (1895), „Versuche über
Getreidedruck in Silozellen“, Z. Ver. Dt.
Ing., 39, August 1895, 1045-1049
Jenike, A. W., and Johanson, J. R. (1968),
"Bin Loads," Journal of the Structural
Division, Proceedings of the ASCE, 94,
1011
Mathews, J., (2013) “Investigation of
granular flow using silo centrifuge
models”, Institute of Geotechnics,
University of Natural Resources and
Life Sciences, Vienna
Rose, H. F., Tanaka, T. (1956), “The
Engineer”, London, page 208
Zhong, Z., Ooi, J. Y., and Rotter, J. M.
(2001), “The sensitivity of silo flow and
wall stresses to filling method”,
Engineering structures, 23(7), 756-767.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Selected geotechnical applications of deformation sensing by

fibre optics and problems with gathered data evaluation

Kristýna ýÁPOVÁ1*
1
Czech Technical University in Prague, Faculty of Civil Engineering, Department of
Geotechnics, Prague, CZECH REPUBLIC

ABSTRACT
The paper presents the basic principles of the application of two optic fibre methods for
deformation measurement in geotechnics - Brillouin Optical Time Domain Analysis (BOTDA) and
Fibre Bragg Gratings (FBG).
It briefly describes the possible and tested utilization of these methods (concerning both
laboratory testing and on site applications) and in more details it deals with several issues and
problems concerning specific adjustments of sensors, data interpretation and evaluation.

Keywords: geotechnical monitoring, BOTDA, FBG, fibre optics

From a physical point of view, light can

1. INTRODUCTION
Geotechnical monitoring in the soil /
rock massive should give reliable
information of deformation development
as a product of long-term natural
processes and human construction
activities. High accuracy, detailed
distribution of at least deformation within
the area of interest and long-term stability
of measurement and warning system are
often required. Fibre optics can be good
alternative to electric sensors using cheap
telecommunication cables for connection
in several kilometres distances to the
analyzer to cover large areas. The
functioning of optical fibres depends on
basic principles of optics and the
interaction of light with matter.
* presenting author
be seen either as electromagnetic waves
or as photons (quanta of electromagnetic
energy). Both possibilities are valid and
equal, but the most useful viewpoint for
optics is to consider light as rays travelling
in straight lines between optical elements,
which can reflect or refract (bend) them.
The paper presents two methods
working on different sensing principles.
The first one is based on Fibre Bragg
Gratings (FBG) and could replace
standard local electric or other sensors.
The second one uses Brillouin Time
Domain Analysis (BOTDA), which offers
new possibilities in strain sensing. These
methods were initially studied and tested
within two research projects in 2011 –
2013 and their research continues to this
day.
2. BRILLOUIN OPTICAL TIME DOMAIN
ANALYSIS
Sensing methods based on the effect of
Brillouin scattering utilize the fact, that
continuous distributed information of time
and spatial domain can be measured
simultaneously by running an optical fibre
to each location. Brillouin scattered light is
caused by non-linear interaction between
the incident light and photons that are
thermally excited within the light
propagation medium (Zhang et al. 2008).
BOTDA instruments can measure
temperature and mechanical strain
distribution along the fibre length.
Figure 1. Brillouin light frequency shift due to
change of strain and temperature, after (Klar
et al. 2010)
The Brillouin Optical Time Domain
Analysis (BOTDA) utilizes the effect of
stimulated Brilloiun scattering, which
arises from the interaction of glass fibre
molecules and transmitted light. The shift
of Brillouin frequency corresponds to the
deformation or temperature change
(Figure 1). The location of the frequency
change along the optical fibre is calculated
from the travel time of the backscattered
light; therefore the strain / temperature
change value and the position of the
change can be determined by this
method.
Laboratory testing of BOTDA
technology started in 2011 and included
testing with several test beams, which had
been developed by the research team. All
the test beams are made of steel, have
measurement interval of 1m or 2m length
and vary in the number of measuring
intervals. The tested BOTDA sensing
cables are attached on both sides of L-
shaped cantilevers, the distance between
248
those cantilevers can be adjusted and the
elongation / shortening of the measuring
base is therefore obtained (Figure 2). The
principles and construction of test beams
are described in more detail for example in
(Záleský et al. 2015).
Figure 2. BOTDA test beam with one 1m
long measuring interval, after (ýápová 2012)
Independent measurements of
deformation were provided in two different
ways in order to have comparative /
control information about deformation to
the fibre optic method. The first
comparative method was utilization of high
resolution potentiometric displacement
transducers in connection to data-logger
and the other was measurement by sliding
micrometer (Solexperts AG, Switzerland),
which is commonly used for geotechnical
monitoring (measurement of axial
deformations in instrumented borehole).
2.1. Laboratory testing – problem with
splice connection
First sets of laboratory measurements
included measurements of changes of
length of measurement base and stability
of measurement in time, then the
development of optimal system of fixing of
sensing cables to monitored structure
followed. The sensing cable was attached
to the short test beam in four 1m long
measuring sections, the strain sensing
fibres were connected by splice
connection into a close loop (measured
deformation distribution is therefore
mirrored along the point of splice).
At Figure 3 an example of the output
from the analyzing unit is presented. Each
sensing cable includes two sensing fibres;
there is a splice connection in the middle
of the measuring line (the fibre goes
through four measuring parts, then is
spliced and goes through the same four
measuring parts in the opposite direction).
In Figure 3 there are clearly visible four
strained parts with almost constant strain
distribution. Due to the defect in the splice
connection, the other four strained parts
with non-constant strain distribution
appear.
Figure 3. Screenshot of strain distribution
during laboratory testing with BOTDA analyzer
– poorly provided splice connection
A new set of tests was provided on a
new test beam in the lab. The new test
beam is very similar to the previous one,
but the measuring section (base) is 2m
long. The issue that need to be solved
was the poorly provided splice connection
of the fibres. Therefore the execution of
the splice connection was extremely
careful and the results showed that the
effects of a poor splice were suppressed
quite well, see Figure 4 (the figure
presents the zero measurement of an
unloaded fibre, so called baseline).
Figure 4. Screenshot of strain distribution
during new laboratory testing – precisely
provided splice connection
249
2.2. Attachment system of BOTDA
The sensing cable is attached to the
test beam between two duralumin plates
(50x50x5mm), which are screwed to L
profiles on the upper flange of the IPE
beam, see Figure 2.
At Figure 3 there are visible unexpected
peaks both in strained and unstrained
parts. These peaks were formed due to
the attachment system and related large
sudden changes of strain applied to the
sensing cable; the sensing cable may be
too much tight between duralumin plates.
These effects are quite severe; they
occur not only during laboratory testing,
but also in field instrumentation (see
chapter 2.3). As already mentioned, they
are expected to be caused by the local
cable compression at the attachment
points. These effects were further
investigated and tested on a laboratory
model of an instrumented borehole and
also on a test beam. These tests resulted
in careful cable coverage by thick plastic
tape in the area of duralumin plates before
fixing of clamps using controlled torque
moment. The effects were partly reduced
using this procedure (see Figure 4).
Nevertheless, this solution is functional,
but it is not perfect – the issue needs to be
analyzed in more detail.
2.3. Borehole measurement
In the area observed by one member of
the research team (Rabenov, Czech
Republic) there is a borehole used for
monitoring of 3D subsoil deformations.
The monitoring in this borehole (VB01)
indicated shear deformations in the depth
of 14-15m below the ground level and also
showed activity, which kept slowly
developing since 2006. This had been the
reason for providing the locality by a new
borehole (VB01A) in the vicinity of VB01.
The new borehole has been instrumented
by combined casing for measurement by
sliding deformeter and inclinometer and
also by BOTDA sensing cables. Optical
sensing cables have been attached to the
measuring casing in 1m distances – the
strain is measured along this length. The
results of BOTDA measurements are
presented in Figure 5 and the results of
control measurements by sliding
deformeter in Figure 6, both BOTDA and
control measurements were carried out on
the same dates.
The strain was measured along 1m
base using a couple of sensing cables
labelled as 1-2 and 3-4. Each cable
consists of two sensing fibres in closed
loops (1 and 2, 3 and 4) which are spliced
at the toe of the measuring casing. The
resulting figure is a plot of calculated strain
from sensing fibres 1 and 4 along the
whole depth of the borehole, because of a
disturbing phenomenon of poorly provided
splice connection – only a half of each
sensing cable has been used for
calculation of axial strain distribution.
Figure 5. BOTDA measurement in VB01A –
left: differential plot, right: integrated plot
Figure 6. Sliding deformeter measurement in
VB01A – left: differential plot, right: integrated
plot
At Figure 5 high values of strain near
the ground level are clearly visible; these
250
values are expected to be caused by
temperature changes at the ground level
and they do not characterize the
development of deformations. The results
of the BOTDA measurements give us in
points of attachment larger values of
developed strain than the control
measurement, nevertheless the shapes
(trends) of both plots are similar. The
effect of strain increase at the attachment
points is discussed in chapter 2.2. and is
probably caused by too high compression
of the cable by the attachment system.
The solution of the problem is described in
the chapter 2.2. as well. It is also
necessary to provide more deep analysis
of the comparison of BOTDA and control
measurements together with the
development and laboratory testing of a
new attachment system. The rest of the
plot demonstrates overall stability; on the
other hand several major issues occur
there and need to be tested and studied
more deeply.
2.4. Analysis of the measured data
In several measurements, particularly in
the borehole monitoring, there are
unexpected increases of strain not related
to the attachment system issues (too
much tight cable in the attachment
system). These measurements are
provided with four sensing fibres at the
same time, but the strain increases occur
only at some of them. These effects may
be caused by some defect in the sensing
cable or fibres. Therefore it is necessary to
provide a consistent check of any sensing
cable prior to its instrumentation.
Figure 7. Brillouin scattering with Gaussian
distribution, after (ýápová 2016)

Figure 8. Brillouin scattering with non-
Gaussian distribution, after (ýápová 2016)
The measuring / analysing unit used for
the basic BOTDA data evaluation enables
more deep analysis of the measured data
at each point of the measurement. It can
calculate and display a plot of the Brillouin
scattering in comparison with the
Gaussian distribution. This analysis was
provided in several cases and the
example of Gaussian and non-Gaussian
distributions are presented in Figure 7 and
Figure 8 (ýápová 2016). According to this
analysis it also seems that the problem is
accented with strain increase.
3. FIBRE BRAGG GRATINGS
The Fibre Bragg Gratings (FBG) strain
sensors are able to measure the average
value of strain which has been developed
on the optical fibre. The fibre needs to be
provided with the Bragg grating (Figure 9)
and also has to be attached between two
fixed points. In order to keep the sensor in
operation, it is necessary to keep the fibre
tight during the whole measurement.
The measurement of reflected
wavelengths is provided by the measuring
unit; all the results presented in this article
have been obtained by measurement with
the unit FBGuard (developed by the
Safibra, Ltd. within the R&D project).
Figure 9. Principle of Bragg Grating, after
(Sugimoto)
251
Every measuring unit operates only in
specific range of the optical spectrum, for
example 1510 – 1590nm, therefore it is
necessary to design in advance Bragg
wavelengths of specific sensors
considering assumed strain changes. The
FBGuard unit allows to measure on up to
four channels (four optical fibres provided
by FGB sensors) in the wavelength range
around 1550nm. The calculations
provided by (Vopiþka 2012) show that less
the strain change, higher the maximum
number of sensors per one channel.
3.1. Interrogation unit FBGuard 1550
The interrogation unit FBGuard 1550
which is used for all the FBG laboratory
and in-situ measurements performed and
presented by the research team has been
developed by the company Safibra, Ltd.
within the research project. The unit is
described in details in (Záleský, M. et al.
2013).
The unit is fully industrial and can
perform static and dynamic measurements
of FBG sensors with high accuracy. The
measuring system contains a broadband
light source in the range of 1550nm and
performs spectral analysis by means of a
linear CCD-array spectrometer platform.
The monitored input channel can be
selected by an optical switch; it is also
possible to adjust scan frequency and
number of channels.
FBGuard unit is fully controlled by an
internal embedded PC and is independent
on external devices and control
procedures. All the measured data are
stored in the internal SSD memory. The
unit is designed for permanent operation
using Linux system. The configuration of
the interrogation unit, alarms, scan rate
(frequency of measurements), events and
data logging is possible via SSH and web
interface, because FBGuard 1550 is
designed as a web server.
3.2. Borehole monitoring by FBG strain
sensors
From the analysis of the local
investigations, the area, where active
movements are expected and the
instrumentation may help to complete the
knowledge about the mechanism of slope
movements, has been chosen for the
instrumentation of FBG sensors into a
borehole. Geological conditions and
previous geotechnical monitoring of the
area (Rabenov, Czech Republic) is
described in more details in (ýápová et al.
2014) or (Záleský et al. 2014)
Considerations about the most
appropriate sensor lengths, as well as
their deployment in the borehole cross-
section were supported by parametric
studies using geometric models of
measuring casing deformed in a shear
zone by the sliding soil mass. The aim
was the determination of the maximal
horizontal deformation, which can be
measured, when the limit strain in the
optical fibres is reached. Further, the
appropriate lengths of the FBG
deformation sensors had to be stated with
respect to awaited horizontal deformation.
The shape of deformed measuring casing
was estimated as two small circular
sections connected by their mutual non-
horizontal tangent line, Figure 10. The
parametric study was carried out for
several assumed heights of the shear
zone together with selected radii of the
couple of circular parts on deformed
casing. The presented results are based
on analyses published in (ýápová 2012).
Figure 10. Scheme of deformed measurement
casing (left) and schematic cross-section
through instrumented borehole with FBG
sensors (right), after (ýápová 2012)
For sensor lengths of 1000mm, the
maximal calculated horizontal deformation
is about 200mm/1m and in case of 3m
long sensors, the limit deformation is more
than 600mm. This is more than
satisfactory for the intended slope
deformation monitoring. There are two
possible ways of FBG sensors attachment
to the measuring casing in the monitoring
252
borehole, Figure 10. In both cases, the
fibres are attached along both sides of
measuring casing. In the first case, the
fibres are fixed in the neutral axis plane
(perpendicular to the assumed slope
movement direction).
Figure 11. Detail of a chain with a
cylindrical clamp of the FBG sensor in GFRP
tube connected to protective plastic tube, after
(ýápová et al. 2014)
This arrangement provides the
measurement of the relative axial
deformation only and indicates the
location of the slip surface in the borehole
with the development of slope
movements. One fibre is fixed at 1.0m
distances to ensure sensitivity of
instrumentation and the second fibre
attached at 3.0m distances provides
longer instrumentation lifetime in case of
large displacements. The second way of
optical fibres fixing is in the slope
direction, so that deflection curve of the
casing can be calculated together with
values of horizontal and vertical
deformations. Both arrangements are
described in more detail for example in
(Záleský, J. et al. 2013).
This instrumentation emphasized
detailed resolution of the location of shear
deformation; therefore the distance
between FBG strain sensors has been
designed as 1.0m. The strain resolution in
this case can be up to 1ȝm/m.
Figure 12. Left – part of 3D measuring
casing with uncoiled FBG chains, center –
lower end of the chain fitted to the casing,
right – connection of two FBG chains, after
(Záleský et al. 2014)
The FBG sensors chains are in this
case placed into thin protective tubes
made of Glass Fibre Reinforced Polymer
(GFRP). Attachment of GFRP tubes is
provided by gluing them into grooves in
brass cylinders, Figure 11. Chains are
attached to measuring casing which is
used for geotechnical 3-D monitoring; the
casing is placed in a borehole which is
injected by cement-bentonite mixture,
Figure 12.
Figure 13. Example of stable reference
measurement of FBG - GFRP, after (Záleský
et al. 2015)
The reference set of measurements
was performed in December 2013; the
example of results of this measurement is
presented in Figure 13. FBG sensors
demonstrated sufficient overall level of
stability during measurements, although in
several cases there were observed
unexpected increases of mechanical
strain. The sensors in those cases were
not subjected to real loading; the instability
of the measurement probably occurred
either due to defect in the sensor (fault
during its production) or due to faulty
optical fibre used for the manufacturing of
the sensor.
3.3. FBG laboratory testing
New laboratory testing has been
performed since 2015 due to observed
unexpected instabilities during
measurements with FBG sensors.
Therefore the decision had been made to
test two new different glues used for FBG
sensor manufacturing – glue “UV” and
“UHU”. The first kind of glue needs to be
penetrated by the UV light to work. The
UHU glue does not need that because it is
a mixture of two different substances in
order to allow chemical reactions which
make the light penetration not necessary.
253
Each FBG chain had two sensors and
each of these had its own Bragg
Wavelength, one was 1536nm and the
other 1560nm. The two sensors of the two
chains were placed between two L-
consoles, Figure 14. The control
independent deformation measurement
was performed by the displacement
transducers with data-logger; the
transducers measure displacement in the
range of 0-25mm with ±1ȝm/m resolution.
Figure 14. L-consoles with displacement
transducer and tested optical FBG fibres
The test consisted of a sequence of
steps of console displacements. The fibres
were loaded and unloaded with a series of
cycles in order to check the change in its
properties / behaviour. After each
enforcing of displacement, a time interval
of 24h had to pass in order to let the
wavelengths stabilize and to apply the
temperature compensation.
The temperature of the air in the lab
was measured and it changed periodically.
For this reason the temperature
compensation was carried out fixing the
instants where the temperature of the air
was the same, so that only wavelengths
measured in these instants were
considered.
The test consisted of 3 kinds of cycles.
For the first cycle four loading - unloading
steps were done (enforced displacement
of 0.05mm); the aim of the first kind of
cycle is checking the fibres durability. The
second kind of cycle consisted of five
loading – unloading steps (enforced
displacement of 0.1mm); this kind of cycle
is useful to test the stability of anchor
system and especially of glue. The third
cycle is still in progress and it consists of
loading - unloading steps which last more
than 24h to check the durability of glues in
time.
The first cycle proved that the fibre
chain with UHU glue became stable more
quickly than UV fibre chain. Moreover both
kinds of fibre chains showed a plastic
behaviour probably because of the
mechanical properties of both glues (the
modulus of elasticity of the glue may be
much higher than the modulus of elasticity
of the fibre itself; when the fibre is
elongated, there may be significant strain
distribution along the glued part).
In the second cycle the fibre with UHU
glue showed a minor relaxation after the
sequence of loading and unloading steps
(Figure 15); therefore the UHU fibre
improved its behaviour. On the other
hand, the UV fibre did not – it could not be
fixed well to anchor and this can cause
fibre sliding. The reason of that could be
the fact that the glue needed to be
penetrated by the UV light to work but
entering into fibre is difficult for the light
because the space where the glue is not
protected by the anchor is too narrow. The
UHU glue does not need to be penetrated
by any light because it is a mixture of two
different substances in order to allow
chemical reactions which make the light
penetration not necessary.
Figure 15. Example of one step of UHU
testing; blue – deformation, left vertical axis,
ȝm; red – temperature, right vertical axis, °C;
green – humidity, right vertical axis, %;
horizontal axis - time
All the testing steps showed also some
relevant differences between
displacements recorded by the two
sensors of fibres with the same glue which
254
were probably caused by some irregularity
between glues of two sensors. Finally,
both fibre chains seemed significantly
influenced by sudden temperature and
especially humidity changes – the fibres
have to be well protected against humidity
penetration and this aspect should not be
underestimated.
4. CONCLUSIONS
BOTDA analyzer is able to recognize
locations of active strain development on
the sensing cable without any prior
estimation. The sensing cable is very
sensitive to break on larger cracks or
fissured zones and in specific cases it
needs arrangements (in order to spread
deformation to longer base). BOTDA
sensing cables are extremely sensitive
during instrumentation in boreholes,
therefore some centralizers or spacers
shall be used to keep optic cables in
reasonable clearance to borehole casings
and the injection grouting should be
provided in two stages by injection tubes
with sleeves (cuffs) minimizing uplift
impact.
The issue concerning strain increases
in the vicinity of attachment points was
partly solved by careful cable coverage by
thick plastic tape before fixing of clamps
using controlled torque moment.
Nevertheless, this solution is functional,
but it is not perfect – this issue needs to
be analyzed in more detail. The
unexpected increase of measured strain
during borehole measurements showed in
several cases a relationship with the non-
Gaussian distribution of Brillouin
scattering. This issue may be caused by
some defects in the sensing cable / fibre,
but it is necessary to provide further
research into the problem. It is also
necessary to provide more deep analysis
of the comparison of BOTDA and control
measurements together with the
development and laboratory testing of a
new attachment system. Once well
deployed, fibre optic sensing can offer
long-term stable service more safe than
contact electro-mechanical monitoring.
 The Bragg grating sensors are in
principle point-wise and they can be used
in series. If they are deployed in chains,
the reliable assessment of magnitudes of
deformation shall be provided in advance.
More Bragg gratings on one channel are
sensitive to overlapping when expected
range of deformation is exceeded even in
case of a single grating. This results in the
loss of information. In that case it is not
possible to clearly automatically evaluate
deformations measured these gratings.
Concerning test beam results, the first
cycle proved that the fibre chain with UHU
glue became stable more quickly than UV
fibre chain. Moreover, both kinds of fibre
chains showed a plastic behaviour
probably because of the mechanic
characteristics of the glue. Maybe its
modulus of elasticity E is much higher
than the modulus of elasticity of the fibre
and when the fibre is elongated, a strain
distribution along the anchor is made. The
recommendation to cure plastic behaviour
is to carry out a series of loading and
unloading cycles in order to decrease the
difference between the two modules. In
this way the behaviour of the anchor
system could improve because the strain
distribution would not be made into anchor
and the plastic behaviour could be
minimized. After that finding the most
fitting kind of glue, as the glue which
shows the minor plastic behaviour, would
be useful. When the behaviour of the glue
is elastic, doing a calibration of results
using a fitting simple function will be
possible.
The second cycle tested, that the fibre
with UHU glue showed a minor relaxation
after the sequence of loading and
unloading steps. The fibre with UV glue
could not be fixed well to anchor and this
can cause fibre sliding. The reason of that
could be the fact, that the glue needs to
be penetrated by the UV light to work but
entering into fibre is difficult for the light
because the space where the glue is not
protected by the anchor is too narrow. On
the other hand, UHU glue does not need
to be penetrated by any light because it is
a mixture of two different substances in
255
order to allow chemical reactions which
make the light penetration not necessary.
So the UHU fibre could be preferred
because it is fixed better showing to be
more rigid then UV fibre.
The other thing to note is that all steps
showed some relevant differences
between displacements recorded by the
two sensors of fibres with the same glue
which were probably caused by some
irregularity between glues of two sensors.
Finally both chain fibres look significantly
influenced by sudden temperature and
humidity changes. Temperature
compensation is possible to adjust results
but humidity compensation is not. So in
this case the recommendation is to make
the fibre more protected against humidity
(for example using another protective
tube).
ACKNOWLEDGEMENTS
The paper is based on results and
experiences gathered in the frame of two
completed projects “TA01011650
Research and development of
applications of dielectric strain sensors in
geotechnics” and FR-TI3/609 “Research
and development of detection and
monitoring of critical areas of geotechnical
constructions mainly in underground
structures, mining industry and other
engineering constructions” financed by the
Technology Agency and by the Ministry of
Trade and Industry of the Czech Republic.
The paper was created with the support of
the internal grant of CTU
SGS16/050/OHK1/1T/11 “Optimization of
geotechnical monitoring and calculation of
ensuring the stability of foundation pit”.
I would like to acknowledge all the
members of research teams of both
projects, particularly Ing. Jan Záleský,
CSc., who is also my supervisor. I would
also like to thank MSc. Lucrezia Digiorgio,
who performed a traineeship at CTU in
Prague and was very supportive during
laboratory testing.
REFERENCES 1878, Albena, Bulgaria. ISBN 978-619-
7105-08-7. ISSN 1314-2704.
ýápová, K. (2012) “Fibre Optic Monitoring
of Slope Stability”, Master thesis, tutor Záleský, J., Záleský, M., Šašek, L.,
Záleský, J., CTU in Prague, Faculty of ýápová, K. (2015) “Fibre Optics
Civil Engineering, Department of Applied for Slope Movements
Geotechnics. Monitoring”, Proceedings of the XVI
ECSMGE 2015, p. 1699 – 1704,
ýápová K., Záleský J. & Záleský M.
Edinburgh, United Kingdom of Great
(2014) “Comparison of results of two
Britain and Northern Ireland. ISBN 978-
methods of fiber optics deformation
0-7277-6067-8.
monitoring in geotechnics”, 14th
International Multidisciplinary Scientific Záleský, M., Záleský, J., Šašek, L.,
GeoConference SGEM 2014, p. 179 – ýápová, K. (2013) “In-situ testing of
186, Albena, Bulgaria. ISBN 978-619- FBG deformation sensors with use of a
7105-08-7. ISSN 1314-2704. new test beam”, Proceedings of the 5th
International Young Geotechnical
ýápová, K. (2016) “Fibre optic sensing in
Engineers Conference, p. 603 – 606,
geotechnics – analysis and evaluation
Paris, France. ISBN 978-61499-296-7.
of measured data”, Proceedings of
ISSN 2212-781X.
Juniorstav 2016, Brno, Czech Republic.
ISBN 978-80-214-5311-1. Zhang, J.-S., Li, Y.-Q., Zhang, S. et al.
(2008) “Performance analysis of
Klar A., Goldfeld Y. & Charas Z. (2010)
temperature and strain simultaneous
“Measures for identifying cracks within
measurement system based on
reinforced concrete beams using
heterodyne detection of Brillouin
BOTDR”. SPIE - smart structures/NDE,
scattering”, Journal of electronic
San Diego, paper no. 7647-90.
science and technology, Vol. 6, Issue 4,
Sugimoto H. “Landslide Monitoring by p. 381 – 384. ISSN 1672-6464.
Optical Fiber Sensor”, Public Works
Research Institute, Erosion and
Sediment Control Research Group,
Ibaraki, Japan.
Vopiþka, M. (2012) “Use of fiber optic
sensors for deformation monitoring in
geotechnics”, Master thesis, tutor
Záleský J., CTU in Prague, Faculty of
Civil Engineering, Department of
Geotechnics.
Záleský, J., Pospíšil, J., ýápová, K.,
Koska, B., Jon, J., Záleský, M.,
Demuth, J. (2013) “Landslide Mapping
and Monitoring Concept”, 2nd Joint
International Symposium on
Deformation Monitoring (JISDM),
Nottingham, United Kingdom of Great
Britain and Northern Ireland.
Záleský, J., ýápová, K. & Záleský M.
(2014) “Concept of fiber optics
geotechnical deformation monitoring
and in-situ deployment in open cast
mine”, 14th International
Multidisciplinary Scientific
GeoConference SGEM 2014, p. 171 –
256
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Finite Deformation Modelling of Cyclic Sand Behaviour using

Bounding Surface Plasticity

Christian Carow1*
1
Technische Universitaet Berlin,
Chair of Soil Mechanics and Geotechnical Engineering, Berlin, GERMANY

ABSTRACT
Precise and efficient numerical simulation of contemporary geotechnical problems requires
elaborate constitutive models for the stress-strain behaviour of sands. The models are supposed to
incorporate the finite deformation theory as well as sophisticated mechanisms for cyclic loading.
With regard to that, a short review of existing constitutive models is given. The bounding surface
model of Li can be considered one of the most comprehensive approaches and further
development of it seems appropriate. The model is presented thoroughly and two leverage points
for improvements are identified. First, the model is reformulated in a manner consistent with finite
deformation theory. The reformulation is outlined and the consequences for the practical
application of the model. Second, a mechanism for cyclic compression is added to the model in
order to improve simulation results. The general methodology of the additional mechanism is
examined here. Further experimental investigation is required as a basis for the desired extensions
of the model.

Keywords: constitutive modelling, sand, bounding surface plasticity, cyclic loading

danger zones as well as during pile

1. INTRODUCTION
Analytical and numerical models of
complex geotechnical processes have
been developed at the soil mechanics
division of Technische Universitaet Berlin
for several decades. Recent work aims to
design advanced instruments for
numerical modelling. They are meant to
predict displacement and maximum
bearing capacity in situations where
subsoil experiences cyclic loading as well
as large inelastic deformations. This
happens in offshore-areas and earthquake
* presenting author
installation
In a first step the Arbitrary Lagrangian
Euler Method (ALE-method) was adapted
to meet geotechnical requirements by
Aubram (2015). ALE allows to perform
finite element calculations in cases where
conventional Lagrangian FEM ceases to
function due to severe element distortion.
On this basis, the Multi Material
Arbitrary Lagrangian Euler Method
(MMALE) method was developed by
Aubram et al. (2015) and Savidis et al.
(2015). In MMALE the material is able to
move freely through the finite element
mesh. As a result, there are no limitations
with respect to the magnitude of
deformation that can be taken into
account.
There are of course several other
methods that could be used in
geotechnics if large deformations are
expected to occur, as shown in (Wang et
al., 2015). Nevertheless, in all those
methods the quality of the results that are
obtained relies heavily on the constitutive
model that is used for the soil skeleton.
The model to be employed must be
consistent with the finite deformation
theory. Furthermore, it should be able to
capture most characteristics of the soil's
stress-strain behaviour in order to account
for the complex loading paths which are
encountered in the aforementioned
problems.
This paper makes an attempt to explain
how to find such a model for sandy soils.
2. CONSTITUTIVE MODELS FOR
SANDY SOILS
Constitutive models for cyclically loaded
non-cohesive soils have been subject of
myriads of scientific treatises. These days
it seems that the most promising models
either are of the bounding surface
plasticity type or descend from the
Karlsruhe school of hypoplasticity.
Hypoplastic models, as was first
outlined in (Kolymbas, 1977), consist of a
single tensorial equation. Its shape has
been derived deductively on the basis of
general axioms of material theory.
Gathering from (Nübel & Cudmani, 2000;
Tsegaye et al., 2000; von Wolffersdorff &
Schwab, 2009), the hypoplastic model of
(von Wolffersdorff, 1996) in conjunction
with the intergranular strain extension of
(Niemunis & Herle, 1997) can be used
successfully for problems with cyclic
loading. However, there is room for further
improvements. Currently, volumetric
strains and excess pore pressure
associated with shearing through dilatancy
are not predicted satisfactorily. This is an
important issue since dilatancy is the
258
cause of accumulation of settlements and
liquefaction under cyclic loading.
Li & Dafalias (2000) revealed a new
state-dependent dilatancy model in the
context of elastoplasticity. It predicts
shear-induced volumetric strains based on
the state-parameter \ . The scalar number
\ , which is attributed to (Been & Jefferies,
1985), quantifies the distance of the
current material state from the critical state
line (CSL). Since the CSL is a curve in
pressure/void ratio space, the stress- as
well as the density-dependency of the
behaviour of sands are inherently
captured through the concept of (Li &
Dafalias, 2000). As a consequence, a
single set of material constants can be
used to achieve good simulation results
over a wide range of stress-levels and
densities.
The ideas of (Li & Dafalias, 2000)
gained much positive resonance.
Consequently they have been
incorporated into the popular SANISAND-
Model by Dafalias & Manzari (2004).
SANISAND uses the concepts of
bounding surface plasticity in a singular
and very efficient manner to model
hardening and softening of non-cohesive
soils under complex loading/unloading-
paths. Thereby, SANISAND is well-suited
to be used in earthquake-simulations, as
has been proven in (Taiebat et al., 2010).
Another popular model which employs
the approach of (Li & Dafalias, 2000) is
the one developed by Li (2002) on the
base frame of (Wang et al., 1990). It will
be called Critical State Sand (CSSA-)
model throughout this paper. It has been
validated successfully for the use in cases
were cyclic loading is present by Ming & Li
(2003) and matches SANISAND in various
respects. One thing that makes the CSSA-
Model unique is its ability to compute
plastic strains caused by changes in
average effective stress p ' . This is
accomplished by using an additional cap-
shaped bounding surface besides the
usual cone-surface. A similar mechanism
has been incorporated into SANISAND by
Taiebat & Dafalias (2008). Nevertheless,
according to (Prada Sarmiento, 2009) it
causes numerical instabilities there.
This (very) short review of the state-of-
the-art in modelling the cyclic behaviour of
sands may be summarized as follows.
There are several highly developed
models. All have certain advantages and
drawbacks. The CSSA-Model is equipped
with a reliable and powerful dilatancy
model, which is of major importance when
cyclic loading is being considered. Due to
its cap-mechanism, the CSSA-Model
seems to provide the most comprehensive
formulation in the elastoplastic realm.
Therefore, the model may be used in
conjunction with the aforementioned high-
class numerical methods to solve
contemporary boundary value problems
involving cyclic loading and large soil
deformation. Hence, it seems expedient to
take a closer look at the CSSA-model and
outline strategies for removing it’s still
remaining deficiencies.
3. CRITICAL STATE MODEL
SANDS
The CSSA-Model designed by Li (2002)
is of the elastoplastic rate-type and has
been formulated within the geometrically
linearized small-strain theory.
As a member of the bounding surface
family set up by Dafalias (1986), the
CSSA-Model is endowed with two
bounding surfaces in principal stress
space and corresponding loading
surfaces. They are shown here in Figure
1. A cone-shaped bounding surface
Laboratory test
15
Shear-stress τ in kN/m²
10
5
0
-5
-10
-15
0 20 40 60 80 100120 140 160
Effective vertical stress σ'v in kN/m² Effective vertical stress σ'v in kN/m²
Figure 2 Stress path of cyclic simple shear test on Berlin Sand ( ID0
σ'2
p'
ne
Co
p
Ca
σ'1
σ'3
Bounding Surfaces Loading Surfaces
Figure 1: Bounding and loading surfaces of
the CSSA-Model in principal stress space
processes changes of the so called stress
ratio. In general, the stress ratio is
represented by the second order tensor
r s / p ' , with s being the deviatoric part
of the effective stress ı ' . Under
axisymmetric conditions, the stress ratio is
represented by the scalar K q / p ' , in
which q is the equivalent deviatoric
stress.
FOR The cone is being closed off by a flat
cap. As has been indicated in the previous
section, the cap is meant to model plastic
strains caused by pure compression of the
soil along a stress path with constant
stress ratio.
The major benefit of the bounding
surface concept appears when repeated
loading and unloading is to be rendered.
Then, the distance between the current
stress point and its image on the bounding
surfaces serves as a measure for the
influence of the loading history on the
CSSA-Model
0 20 40 60 80 100120 140 160
65 % )
259
current stiffness. That has been proven to
be quite efficient if cyclic loading is being
considered.
Both bounding surfaces are equipped
with dilatancy functions which originate in
the work of Li & Dafalias (2000). As
mentioned before, their most important
input is the state parameter \ . It provides
information about the amount of dilatancy
to be computed with regard to the stress
level as well as to the soil’s density.
In cases where cyclic shearing is
predominant, the functional interaction of
the dilatancy model and the bounding
surface concept works out rather well.
This is being illustrated by back-
calculating a cyclic simple shear test, the
result of which is presented in Figure 2.
In the following, the application of the
CSSA-Model to the solution of a boundary
value problem will be demonstrated. The
problem to be investigated is a reservoir
dam which encounters seismic loading by
the 1971 San Fernando Earthquake. The
finite element model, shown in Figure 3,
utilizes the fully coupled two-phase-
formulation of Zienkiewicz & Shiomi
(1984). It captures the influence of the soil
skeleton’s deformation on the pore
pressure, which is of major importance
here because the sandy soil that forms the
dam’s body is expected to compact due to
the shear loading of the earthquake. This
may, under certain circumstances, result
in considerable pore-pressure-buildup and
in turn reduce effective stress and shear
strength considerably.
Selected results of the simulations are
shown in Figure 4 and Figure 5. The sand
of the fill material that forms the main dam
6m
15 m
nd)
y (Sa
25 m am Bod
y D Clay-core
20 m Sand (dense)
ux(t)
100 m Figure 3: Finite
200Element
m Model of the reservoir dam
body was initially assumed to be of
medium relative density. This leads to a
moderate buildup of excess pore
pressure. The crest settles only slightly.
Compared to that are results for loosely
deposited fill material. Herein, significant
horizontal displacements occur in the
upstream slope. This is supposed to be an
indicator for approaching slope-failure.
The observed behaviour in the second
case is naturally attributed to the fact that
stiffness and shear strength of loose
sands are lower than those of medium
dense sands. In addition to that, the lower
density induces a stronger tendency of the
soil to compact when being sheared. This
results in excess pore-pressure if cycling
loading is applied rapidly, as during
earthquakes. As a consequence, the
effective stress level decreases and the
shear strength is reduced rapidly, as can
be observed in Figure 4. This explains the
comparatively large displacements in
second case.
Excess pore pressure in loose, non-
cohesive dam fill material may even
reduce the effective stress to zero in a
state of liquefaction. This happened at the
Lower San Fernando Dam during the
1971 earthquake, as has been
documented in (Seed et al., 1975).
Bottom line, the simulation results show
qualitatively that the model of Li (2002) is
able to describe the influence of soil
density on the cyclic behaviour of sands
quite well. Still, the model has two weak
spots that prevent it from being used in
large deformation cyclic analyses out of
stock. First, the model is not compatible
with finite deformation theory. Second,
12 m
Berm
(dens
e San
d)
x
60 m 100 m
260
there is one aspect of cyclic sand significant error if such magnitudes of
behaviour which is not captured well. strain are encountered.
These issues are treated in the following To extend a conventional plasticity

Medium dense dam fill Loose dam fill

|u|max 3,4 m
1,1 m
2,8 m
1,5 m
1,1 m

6,3 m
A A
|u|=0 |u|=0

Figure 5: Contour plot of displacement caused by the applied earthquake

Medium dense dam fill Loose dam fill

200
Stress in kN m-2

150 Mean effect. stress Mean effect. stress

100
50
0 Excess pore pressure Excess pore pressure
-50
0 5 10 15 20 25 30 0 5 10 15 20 25 30
Time in s Time in s

Figure 4: Evolution of mean effective stress and excess pore pressure during the earthquake

two sections. model into the finite deformation range,

4. INCORPORATION
DEFORMATION THEORY
As explained in the first section, an
elaborate constitutive model is needed
which is compatible with the finite
deformation theory. The CSSA-Model
however, in accordance with the general
practice in traditional soil mechanics, has
been formulated within the geometrically
linearized infinitesimal theory. Therefore,
the model equations have to be
reformulated in order to be theoretically
compatible with MMALE & Co.
There is another important reason to do
so. During calibration of the CSSA-model,
triaxial compression tests featuring axial
strains of 20 % and more have to be back-
analyzed in order to determine critical
state parameters. The assumption of
infinitesimal deformation induces a
the following issues have to be
OF FINITE considered:
x The finite deformation theory
distinguishes different placements of
the material body in space-time. Which
one should the model refer to?
x How to treat superposed rigid body
motions properly?
x How can the deformation be split into
elastic and inelastic parts?
x How can the elastic stress-strain
relationship be put up?
Those issues were solved in plasticity
theory long ago, but the solutions are not
common in soil mechanics. The general
concepts of finite deformation plasticity are
for example described in (Belytschko et
al., 2000). By making use of them, the
reformulation of the CSSA-Model can be
outlined as follows.
261
 The reformulated CSSA-Model refers to J
§ ·
p'n : r 2
the material placement at current time t . ¨¨ n 
p
d D1I ¸
That is called a spatial or Eularian K p1 © 27 ¸¹
formulation. Superposed rigid body
p ' § 2 ·
motions are accounted for by employing  ¨¨ m  D2 I ¸
K p2 © 27 ¸
objective tensors as stress and strain ¹
measures only. The usual Cauchy-Stress In this equation, the hardening modulus
is an example of an objective tensor.
K p1 controls the evolution of the cone-
Objectivity means here that rigid body
motion may alter the components of the shaped bounding surface, the modulus
stress tensor with respect to a fixed Kp2 that of the cap. The flow direction n
coordinate system, but does not induce is normal to the cone. The flow direction
additional stress. m of the cap is parallel to the current
The symmetric part d of the velocity stress ratio. D1 and D2 are the dilatancy
gradient serves as a means to measure
functions of cone and cap, respectively.
deformation in the reformulated CSSA-
Modell. It is objective, too and called rate- As a result of the foregoing, the
of-deformation. Its additive decomposition elastoplastic stress-strain relationship of
the reformulated CSSA-Model can be
d de  dp stated as
J
into an elastic part d e and a plastic part
ı ' cep (ı,e, O1) : d .
dp is the starting point for the
reformulation. With respect to d e , the The fourth-order elastoplastic stiffness
elastic stress-strain-relationship is tensor c ep depends on stress, void ratio e
postulated. It has the shape of the and the accumulated plastic strain O1
hypoelastic rate-model associated with the cone. In the end, c ep
O O
has the same shape as the stiffness
ı ' c e (ı , e ) : d e , (V ') ij ( c e ) ijkl d kle .
tensor developed by Li (2002), but it is
The elastic moduli c e may use the formed by finite deformation quantities.
same dependency on stress and void ratio The reformulation leads to a series of
e as in the infinitesimal model. However, consequences regarding the practical
since the material time derivative of the application of the CSSA-Modell.
Cauchy-stress ı is not an objective x The corotating stress-rate has to be
tensor, an appropriate objective stress integrated over the computational time
O
rate ı has to be chosen. Here, we pick increment by a so called incrementally
J objective algorithm, like for example the
the corotating rate ı ı  ı ˜ Ȧ  Ȧ ˜ ı one developed by Hughes (1984).
proposed by Jaumann (1911). According x In finite-deformation theory, the
to (Guo, 1963), it has some important relationship between displacement and
theoretical advantages over other rates. It strain induces an additional
is based on the skew-symmetric part Ȧ of nonlinearity. This is supposed to
the rate-of-deformation. influence convergency during iterative
The rate-of-deformation d resembles solution of boundary value problems.
the infinitesimal strain-rate İ in many x The results of laboratory tests that are
ways. Therefore, the plastic mechanisms used to calibrate the model have to be
of the CSSA-model may be transferred analyzed utilizing consistent finite
into the reformulated model to compute deformation stress- and strain
d p , which results in measures.

262
5. MODELLING OF CYCLIC
COMPRESSION
This section concerns an open question
regarding the modelling of cyclic stress-
paths with the CSSA-Modell. As has been
demonstrated in section 3, the model
leads to excellent simulation results so
long as cyclic shearing is concerned only.
However, there is room for improvement
with respect to the computation of plastic
strains induced by cyclic changes of mean
effective stress at constant stress ratio,
e.g. cyclic compression.
The present situation is depicted in
Figure 6. It shows the stress - void ratio
path of a cyclic oedometer test and its
back-calculation with the CSSA-Model. In
the laboratory test, repeated un- and
reloading induces only little compaction
compared to the state after first loading.
Opposed to this, the model predicts
undiminished densification even after 20
cycles.
The reason for this obviously unrealistic
behaviour of the model can be found in
the cap-mechanism. It does not account
for the influence of cyclic loading history
and current material state on the
computed stiffness properly.
To iron out this issue, the hardening
modulus Kp2 of the cap (which was
introduced by the equation for d p ) needs
to be augmented by a cyclic mechanism
hc . Thereby, the hardening modulus takes
on the following form:
Laboratory test
0.62
Void ratio e
0.61
0.60
0.59
0 50 100 150 200 250 300 350
Figure 6: Cyclic oedometer test on Toyoura sand; test results due to (Bauer, 1992)
Effective vertical stress σ'v in kN/m²
263
§ M g (T ) · § U 2 · p
K p2 hcG(e, p ')h4 ¨ c ¸¨ ¸
© R ¹ © U 2 ¹ p
Here, h4 and Mc are material
constants. R and T are invariants of the
stress ratio tensor r . U2 and U2 quantify
the stress history in pursuance of the
bounding surface concept. The function
g (T ) defines the width of the cone-shaped
bounding surface. The new constitutive
function hc is scales Kp2 to increase the
stiffness during cyclic loading realistically.
Elgamal et al. (2003) observed that few
experimental investigations have been
conducted with regard to the development
of a cyclic compression cap for plasticity.
At least some qualitative conclusions can
be drawn from the available experimental
data of (Bauer, 1992; Ko & Scott, 1967;
Mallwitz & Holzlöhner, 1996; Sawicki &
Swidzinski, 1995; Wichtmann et al., 2004).
Accordingly, a sand sample in oedometric
or constant-stress-ratio conditions reacts
as follows to cyclic compression with
amplitude ǻ p ' and maximum pressure p0 .
x Compared to the p0 -state after first
loading, every subsequent cycle
induces further compaction.
x The larger the amplitude, the stress
ratio and the void ratio, the more
compaction is induced per cycle.
x The compaction per cycle decreases
with increasing cycle count.
x The rate of compaction decreases with
increasing volumetric plastic strain.
CSSA-Model
0 50 100 150 200 250 300 350
Effective vertical stress σ'v in kN/m²
x Starting from an isotropic initial state,
isotropic stress cycles induce
volumetric strains only.
x Starting from an anisotropic initial state,
isotropic stress cycles may induce
deviatoric strains, too.
x If a sample is loaded beyond p0 after
the cyclic loading stage, the resulting
stress-strain curve extends the virgin
loading curve more or less seamlessly.
The foregoing list indicates which
material state quantities probably have to
be accounted for by the cyclic mechanism
function hc . These are the stress ratio, the
mean effective stress, the void ratio and
the accumulated plastic volume-strain.
The available qualitative experimental
findings, on the other hand, are not
sufficient to derive a specific form for hc .
To this end, a comprehensive series of
cyclic triaxial tests with constant stress
ratio is currently being conducted at the
soil mechanics laboratory of TU Berlin.
The results and their evaluation with
regard to the constitutive functions will be
published in following papers.
6. CONCLUSIONS
The paper highlighted that elaborate
constitutive models for sand are needed to
solve contemporary boundary value
problems which feature cyclic loading and
large soil deformations.
After selecting the bounding surface
model of Li (2002) as one of the most
comprehensive approaches and outlining
strategies for its further development, the
following conclusions can be drawn.
Complex elastoplastic models like the
CSSA-Model can be extended into the
finite deformation theory without having to
re-compile the underlying model
mechanisms. Nevertheless, some
consequences regarding the practical
application of the extended model have to
be considered.
In elastoplasticity, material behavior in
compression is accounted for by cap-like
surfaces in stress space. They work quite
264
well for monotonic loading, but fail to
capture cyclic material behavior properly.
To resolve this, additional constitutive
functions should be developed in
accordance with findings from
experimental investigations. A short
review of the state of the art showed that
existing experimental data is sparse and
additional tests have to be performed.
Such tests are being conducted
currently. On the basis of their results, the
new cyclic function hc for the CSSA-
Model’s cap will be designed.
Subsequently the reformulated and
extended CSSA-Model will be validated by
applying it to benchmark boundary value
problems.
REFERENCES
Aubram, D.: Development and experimen-
tal validation of an arbitrary Lagrangian-
Eulerian (ALE) method for soil
mechanics. In: geotechnik Vol. 38
(2015), No. 3, p. 193–204
Aubram, D., Rackwitz, F., Savidis, p. A.:
Vibro-Injection Pile Installation in Sand:
Part I - Interpretation as Multi-material
Flow. In: Triantafyllidis, T. (Hrsg.):
Holistic Simulation of Geotechnical
Installation Processes, Lecture Notes in
Applied and Computational Mechanics.
Vol. 77 : Springer, 2015 — ISBN 978-3-
319-18169-1, p. 73–102
Bauer, E.: Zum mechanischen Verhalten
granularer Stoffe unter vorwiegend
ödometrischer Beanspruchung,
Universität Karlsruhe, Dissertation,
1992. — Veröffentlicht in Institutsreihe
Been, K., Jefferies, M. G.: A state
parameter for sands. In: Géotechnique
Vol. 35 (1985), No. 2, p. 99–112
Belytschko, T., Liu, W. K., Moran, B.:
Nonlinear Finite Elements for Continua
and Structures : John Wiley & Sons Ltd,
2000
Dafalias, Y. F.: Bounding Surface
Plasticity. I: Mathematical Foundation
and Hypoplasticity. In: Journal of
Engineering Mechanics Vol. 112
(1986), No. 9, p. 966–987
Dafalias, Y. F., Manzari, M. T.: Simple
Plasticity Sand Model Accounting for
Fabric Change Effects. In: Journal of
Engineering Mechanics Vol. 130
(2004), No. 6, p. 622–634
Elgamal, A., Yang, Z., Parra, E., Ragheb,
A.: Modeling of cyclic mobility in
saturated cohesionless soils. In:
International Journal of Plasticity Vol.
19 (2003), No. 6, p. 883–905
Guo, Z.: Time Derivatives of Tensor Fields
in Non-Linear Continuum Mechanics.
In: Archivum Mechaniki Stosowanej
Vol. 15 (1963), No. 1, p. 131–161
Hughes, T. J. R.: Numerical
Implementation of Constitutive Models:
Rate-Independent Deviatoric Plasticity.
In: Nemat-Nasser, S., Asaro, R. J.,
Hegemier, G. A. (Hrsg.): Theoretical
foundation for large-scale computations
for nonlinear material behavior,
Mechanics of elastic and inelastic solids
6 : Martinus Nijhoff Publishers, 1984 —
ISBN 978-94-009-6215-6, p. 29–63
Jaumann, G.: Geschlossenes System
physikalischer und chemischer
Differentialgesetze (I. Mitteilung). In:
Sitzungsberichte / Akademie der
Wissenschaften in Wien,
Mathematisch-Naturwissenschaftliche
Klasse, Abteilung 2a Vol. 120 (1911), p.
385–530
Ko, H.-Y., Scott, F.: Deformation of sand
in hydrostatic compression. In: Journal
of the Soil Mechanics and Foundations
Division Vol. 93 (1967), No. SM3, p.
137–156
Kolymbas, D.: A rate-dependent constitu-
tive equation for soils. In: Mechanics
Research Communications Vol. 4
(1977), No. 6, p. 367–372
Li, X. S.: A sand model with state-
dependent dilatancy. In: Géotechnique
Vol. 52 (2002), No. 3, p. 173–186
Li, X. S., Dafalias, Y. F.: Dilatancy for
cohesionless soils. In: Géotechnique
Vol. 50 (2000), No. 4, p. 449–460
Mallwitz, K., Holzlöhner, U.: Verfahren zur
Ermitlung der Setzung von
Fundamenten infolge zyklischer
265
Beanspruchung. In: Bautechnik Vol. 73
(1996), No. 3, p. 175–186
Ming, H. Y., Li, X.-S.: Fully coupled
analysis of failure and remediation of
lower San Fernando Dam. In: Journal
of Geotechnical and Geoenvironmental
Engineering Vol. 129 (2003), No. 4, p.
336–349
Niemunis, A., Herle, I.: Hypoplastic model
for cohesionless soils with elastic strain
range. In: Mechanics of Cohesive-
frictional Materials Vol. 2 (1997), p.
279–299
Nübel, K., Cudmani, R.: Examples of finite
element calculations with the
hypoplastic law. In: Kolymbas, D.
(Hrsg.): Constitutive Modelling of Gra-
nular Materials. Berlin, Heidelberg :
Springer, 2000 — ISBN 978-3-642-
63115-3, p. 523–538
Prada Sarmiento, L. F.: Implementation of
the elastoplastic Model Sanisand 2004
into Incremental Driver (Presentation
given at the Karlsruhe Institute of
Technology, slides available at
http://geotecnia.uniandes.edu.co/Desca
rgas.aspx), 2009
Savidis, p. A., Aubram, D., Rackwitz, F.:
Vibro-Injection Pile Installation in Sand:
Part II - Numerical and Experimental
Investigation. In: Triantafyllidis, T.
(Hrsg.): Holistic Simulation of
Geotechnical Installation Processes,
Lecture Notes in Applied and
Computational Mechanics. Vol. 77 :
Springer, 2015 — ISBN 978-3-319-
18169-1, p. 103–131
Sawicki, A., Swidzinski, W.: Cyclic
compaction of soils, grains and
powders. In: Powder Technology Vol.
85 (1995), No. 2, p. 97–104
Seed, H. B., Lee, K. L., Idriss, I. M.,
Makdisi, F. I.: The Slides in the San
Fernando Dams During the Earthquake
of Febuary 9, 1971. In: Journal of the
Geotechnical Engineering Division Vol.
101 (1975), No. 7, p. 651–688
Taiebat, M., Dafalias, Y. F.: SANISAND:
Simple anisotropic sand plasticity mo-
del. In: International Journal for
 Numerical and Analytical Methods in The generalized Biot formulation and its
Geomechanics Vol. 32 (2008), No. 8, p. numerical solution. In: International
915–948 Journal for Numerical and Analytical
Taiebat, M., Jeremiü, B., Dafalias, Y. F., Methods in Geomechanics Vol. 8
Kaynia, A. M., Cheng, Z.: Propagation (1984), No. 1, p. 71–96
of seismic waves through liquefied
soils. In: Soil Dynamics and Earthquake
Engineering Vol. 30 (2010), No. 4, p.
236–257
Tsegaye, A. B., Molenkamp, F.,
Brinkgreve, R., Bonnier, P. G., de
Jager, R., Galavi, V.: Modeling
liquefaction behavior of sands by
means of hypoplastic model. In: Benz,
T., Nordal, p. (Hrsg.): Numerical
Methods in Geotechnical Engineering.
London : CRC Press, 2000 —
ISBN 978-0-415-59239-0, p. 81–87
Wang, D., Bienen, B., Nazem, M., Tian,
Y., Zheng, J., Pucker, T., Randolph, M.
F.: Large deformation finite element
analyses in geotechnical engineering.
In: Computers and Geotechnics Vol. 65
(2015), p. 104–114
Wang, Z., Dafalias, Y. F., Shen, C.:
Bounding Surface Hypoplasticity Model
for Sand. In: Journal of Engineering
Mechanics Vol. 116 (1990), No. 5, p.
983–1001
Wichtmann, T., Niemunis, A.,
Triantafyllidis, Th.: The effect of volu-
metric and out-of-phase cyclic loading
on strain accumulation. In: Cyclic
Behaviour of Soils and Liquefaction
Phenomena : Balkema, 2004. —
Proceedings of CBS04, Bochum,
ISBN 90-5809-620-3, p. 247–256
von Wolffersdorff, P.-A.: A hypoplastic
relation for granular materials with a
predefined limit state surface. In:
Mechanics of Cohesive-frictional
Materials Vol. 1 (1996), No. 3, p. 251–
271
von Wolffersdorff, P.-A., Schwab, R.: The
Uelzen I Lock - Hypoplastic finite-
element analysis of cyclic loading. In:
Bautechnik Vol. 86 (2009), No. S1, p.
64–73
Zienkiewicz, O. C., Shiomi, T.: Dynamic
behaviour of saturated porous media;

266
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Finite element limit analysis of offshore mudmat foundations

subject to combined loads

Helen P. DUNNE1*
1
University of Oxford, Department of Engineering Science, UK

ABSTRACT
Rectangular shallow foundations, termed mudmats, are commonly used in the offshore oil and gas
industry to support pipeline end terminations and pipeline end manifolds. Here, they are subjected
to large horizontal forces that also give rise to overturning and torsional moments. This paper
focuses on analysing the capacity of a rigid rectangular mudmat resting on the surface of
undrained clay, when realistic three dimensional loading combinations are applied. A semi-
analytical limit analysis method is derived, with more complex loading arrangements analysed
using finite element limit analysis. With this technique it is possible to obtain lower and upper
bounds on the exact load factor, and automative adaptive remeshing is used to reduce the bound
gap over successive iterations of the solution. Adaptive mesh refinement helps visualisation of the
failure mechanism and relationships between loading combinations and failure mechanisms are
discussed.

Keywords: limit analysis; clays; numerical modelling; offshore engineering

1. INTRODUCTION significantly influenced by torsional

A typical subsea arrangement of risers
and flowlines for the offshore oil and gas
industry can be seen in Figure 1. Shallow
foundations, termed mudmats, are used to
support subsea infrastructure such as pipe
line end terminations and pipe line end
manifolds. These connect pipelines
through jumpers, which can be situated at
a height above the mudmat, and are often
not aligned along the plan view axes. The
jumper experiences horizontal expansion
and contraction forces from the pipeline,
which can induce biaxial bending, torsion
and combined horizontal loads on the
mudmat (depending on the jumper
arrangement). Failure of mudmats can be
* presenting author
loading (McDonald et al., 2014). A
horizontal force that is not applied along
one of the plan view axes causes a
torsional moment on the mudmat, while a
horizontal force applied at a height above
the mudmat causes an overturning
moment.
Classical bearing capacity theory,
stemming from the Brinch Hansen (1970)
method is adopted in the DNV (1992) and
ISO (2003) design codes. This approach
focuses on vertical bearing capacity. It has
been found to be inaccurate in the
analysis of foundations where the primary
loading is not vertical, and foundations on
undrained clay where short term
 (a)
(b) (c)
Figure 1. Shallow foundations as used for the
offshore oil and gas industry: (a) typical
subsea arrangement; (b) rectangular mudmat
foundation; (c) underside view of mudmat
showing shallow underbase skirts
detachment of the foundation is prevented
through underbase skirts (Ukritchon et al.,
1998; Bransby & Randolph, 1998). A
common alternative is the use of three-
dimensional (3D) finite element analysis
(FEA) (Yun et al., 2009; Nouri et al., 2014;
Feng et al., 2014). This improves accuracy
but can be time consuming as a mesh
refinement study is required to ensure
convergence of the solution.
Offshore pipelines are subjected to
increasing levels of temperature and
pressure change, leading to larger loads
for mudmats to resist. However, mudmat
dimensions are restricted by the size of
their installation vessels – increasing the
importance of an accurate bearing
capacity analysis. Carrying out both lower
bound (LB) and upper bound (UB) finite
element limit analysis (FELA)
advantageous as the accuracy of the
solution is directly quantifiable.
Visualisation of the critical failure
mechanism is also useful in the design
process.
In this paper, analytical calculations and
3D FELA are used to analyse the bearing
Figure 2. Notation and sign conventions
capacity of surface mudmats subject to
combined loading. The notation and sign
conventions used in this study are shown
in Figure 2. The mudmat is modelled as a
rigid body of breadth, B, along the x-axis
and length, L, along the y-axis. The
mudmat length to breadth aspect ratio is
fixed at L/B = 2, as this is typically adopted
in practice (Feng et al., 2014).
Combined horizontal and torsional
loading (Fx, Fy, Mz) is considered first. As it
is improbable that a mudmat would be
loaded from outside its plan area,
emphasis is given to horizontal loads
generating feasible levels of torsion
(Figure 3(a)). The largest feasible level of
torsion that can be applied to a mudmat
occurs when a horizontal load is applied at
its corner. Next, the paper considers
horizontal loads applied at a height, h,
above the mudmat, resulting in combined
horizontal and moment loading (Figure
3(b)). Forces applied at an eccentricity
from, and a height above, the mudmat
centre result in complex five degree of
freedom loading (Fx, Fy, Mx, My, Mz). This
paper explores the interaction between
loading components and examines the
corresponding mudmat failure
mechanisms.
is
2. FINITE ELEMENT LIMIT ANALYSIS
All analyses were performed using OxLim,
a FELA program developed at the
University of Oxford. OxLim has been
used extensively for the analysis of plane
strain problems (Martin, 2011; Martin &
White, 2012; Mana et al., 2013; Dunne et
268
Figure 3. Footing loading arrangement: (a) feasible eccentricity due to non-concentric plan view
loading; (b) horizontal loading applied at a height above the footing
al., 2015), and recently for 3D problems
(Martin et al., 2015).
When using OxLim to solve a 3D
problem, the soil domain is first discretised
into a mesh of tetrahedral elements with
relevant boundary conditions specified.
The program then sets up two separate
constrained optimisation problems that
together allow rigorous bracketing of the
exact collapse load multiplier. For this
study, the LB analyses used a piecewise
linear stress field, and the UB analyses
used a piecewise linear velocity field. The
bracketing error associated with the
bounds is and the
average of the bounds, , is
used as the best estimate solution.
Undrained failure was assumed and the
soil was modelled as a rigid–plastic von
Mises material. This allowed both the LB
and UB analyses to be cast as standard
second-order cone programming
problems, and solved using specialised
software (Makrodimopoulos & Martin,
2006; 2007; MOSEK ApS, 2014).
269
Adaptive mesh refinement was used (if
required) to improve the bracketing of the
exact load multiplier. An adaptivity
algorithm based on the spatial variation of
the maximum shear strain rate in the UB
velocity field was implemented, and all
mesh generation was performed using the
open source code TetGen (Si, 2013).
Sheaves of ‘singularity facets’ were
attached to the edges of the footing base
(Figure 4). When these facets are forced
into the mesh, the bounds converge much
more rapidly when compared to using a
completely unstructured mesh.
The von Mises yield strength in pure
shear, k, was equated with the Tresca
shear strength, su, such that the two
criteria were matched for deformation in
plane strain. This has been found to
provide improved accuracy when
compared with using the shear strength
under triaxial conditions in the analysis of
3D foundations (Gourvenec et al., 2006).
The value of k was assumed as
homogeneous throughout the soil domain.
The footing/soil interface was modelled as
fully rough with unlimited tensile capacity.
Figure 4. Initial FELA mesh of quarter model
showing prescribed facets in red
This is consistent with the assumption that
sufficient interior skirts are present in the
mudmat to fully confine a soil plug (Mana
et al., 2013). The soil was modelled as
weightless, as soil weight has no effect on
the results for this problem.
3. RESULTS
3.1. Horizontal and torsional loading
A surface footing subject to combined
horizontal and torsional loading was
analysed using a semi-analytical method
based on the LB and UB plasticity
theorems, and using FELA. LB values for
Fx, Fy and Mz were found by numerically
integrating the shear stress components Wx
and Wy that would be induced by footing
rotation about a prescribed point (x0,y0),
as shown in Figure 5(a):
(1)
(2)
(3)
where Wx = su cosD, Wy = su sinD, with cosD
and sinD calculated as shown in Figure 5.
It is acknowledged that this lower bound
270
Figure 5. Semi-analytical solution for
combined horizontal and torsional loading: (a)
LB shear stresses; (b) UB velocities
solution is not fully rigorous as it does not
extend the interface stress field into the
remainder of the semi-infinite soil domain.
However, this is considered a formality.
The UB solution was formulated by
equating the internal and external work
rates when the footing rotates with a
virtual angular velocity, Z, about (x0, y0),
as shown in Figure 5(b):
(4)
(5)
with r calculated as shown in Figure 5.
The 3D failure surface obtained by solving
these equations numerically for a range of
rotation centres is shown in Figure 6(a). It
should be noted that the failure surface is
symmetrical about the xy, yz, and xz
planes. The LB equations produced
results which matched exactly with the UB
equations, indicating an exact theoretical
solution. The circular failure surface when
Mz = 0 highlights that the capacity of the
footing in this case is always equal to the
interface shear capacity, irrespective of
the relative magnitudes of Fx and Fy.

0.6 Infeasible torsion
M z/B 2 Lsu
0.4
0.2
0
0
0.2
0.4
Fx/BLsu 0.6
0.8
0.4
1 0.2
0
Fy/BLsu
(a)
Figure 6. Combined horizontal and torsional loading failure surface showing feasible torsion limits
(Fx > 0, Fy > 0, Mz > 0)(a) 3D failure surface from semi-analytical solution; (b) 2D contour plot
comparing semi-analytical solution with FELA solution
Figure 6(b) shows that excellent
agreement was found between the semi-
analytical solution and the results obtained
using FELA. FELA results for horizontal
loading (Mz = 0) solved to ± 0.01% error
without the need for mesh refinement,
while in pure torsion (Fx, Fy = 0) an
unrefined mesh solved to ±2% error. Each
analysis completed in less than 10
seconds, and the critical failure
mechanism was always interface shearing
between the footing and the soil. The
feasible torsion limit increased from
0.48B2Lsu when T = 0 (Fx) to its maximum
value 0.5 B2Lsu, corresponding to T = tan-
1 (B/L) (FT). At T = 90° (Fy) the feasible
torsion limit is 0.38B2Lsu.
3.2. Horizontal, torsional and moment
loading
In Figure 7, bearing capacities are plotted
for a footing subject to horizontal forces
applied at a range of eccentricities from,
and heights above, the centre point. When
h/B = 0 and e/B = 0 the bearing capacity is
the same whether the footing is loaded by
Fx, Fy or FT (corresponding to the circular
failure locus when Mz = 0 in Figure 6).
When these forces are applied at h/B =
0.5, there is no reduction in capacity
Semi-analytical solution
FELA solution
Infeasible torsion
M /B 2Lsu = 0, 0.1, 0.2, 0.3, 0.4, 0.5
1
Feasible torsion
0.8
0.6
Fy/BLsu
0.4
0.2
0.8 1
0.6 0
0 0.2 0.4 0.6 0.8 1
Fx/BLsu
(b)
compared with h/B = 0, and failure still
occurs though interface shearing between
the footing and the soil. When Fx and FT
are applied at h/B = 1 there is a reduction
in capacity compared with the same forces
applied at h/B = 0. When h/B = 2 the
reduction in capacity is substantial for Fx
and FT, but there is still no significant
change in capacity for a footing subject to
Fy at this height.
The reduction in capacity when the
footing is subject to Fx and FT at h/B > 0.5
can be attributed to the increased
overturning moment acting on the
footing, which can be visualised in the
failure mechanism. Figure 8 shows the
failure mechanisms when a footing is
subject to forces in various directions at a
height h/B = 2. Fx and FT cause failure by
‘rolling’ about the y-axis. Fy predominantly
fails through translation in the y direction
with very little rotation. This highlights the
increased moment capacity associated
with ‘pitching’ rotation about the x-axis.
The bearing capacity generally reduces
as the eccentricity, e, increases due to the
additional torsional moment acting on the
footing. This is less apparent when the
footing is loaded at h/B = 2 by Fx or FT, as
in this case, failure is still dominated by
271
rolling about the y-axis and not twisting the bearing capacity when h/B = 2 and 0 ”
about the z-axis. Failure mechanisms are e/B ” 0.5 in Figure 7(c). When e/B =1.12
shown in Figure 9 for a footing loaded by twisting is visible in the failure mechanism,
a corner force FT. When FT is applied at
h/B = 0 and e/B = 0, failure occurs through 4. CONCLUSIONS
translation with no rotation. When FT is This study has focused on likely loading
applied at increasing eccentricities, the scenarios for offshore mudmats used to
footing fails through translation as well as support pipeline end terminations and
increasing amounts of twisting. This manifolds. Rigorous 3D bearing capacity
corresponds to the steep reduction in
capacity when 0 ” e/B ” 1.118 in Figure
7(c). A footing subject to FT at e/B = 0 and
h/B = 2 fails through rotation about the y- e/B = 0
axis. The failure mechanism does not h/B = 2
change significantly as e/B increases to Fx/BLsu = 0.44
0.5, which corresponds to no change in

Infeasible torsion
1.2
h/B = 0, 0.1, 0.2, 0.5
1
0.8
Fx/BLsu

h/B = 1
0.6
0.4 (a)
0.2 h/B = 2 e/B = 0
θ=0
0 h/B = 2
0 0.2 0.4 0.6 0.8 1 1.2 Fy/BLsu = 0.95
e/B
(a)
1.2
h/B = 0, 0.1, 0.2, 0.5, 1
1
0.8
Fy/BLsu

h/B = 2
0.6
0.4
0.2
θ = 90 °
0 (b)
0 0.2 0.4 0.6 0.8 1 1.2
e/B e/B = 0
(b) h/B = 2
1.2 FT /BLsu = 0.48
1 h/B = 0, 0.1, 0.2, 0.5

0.8
Fθ /BLsu

h/B = 1
0.6
0.4
h/B = 2
0.2
θ = tan-1(B/L)
0
0 0.2 0.4 0.6 0.8 1 1.2
e/B
(c) (c)
Figure 7. Bearing capacity of a footing subject Figure 8. Mechanisms for a footing subject to
to F at various eccentricities, e, and heights h: F in various directions applied at h/B = 2: (a)
(a) Fx; (b) Fy; (c) FT Fx; (b) Fy; (c) FT. Shaded by UB velocities
272
analyses have been carried out using both
analytical calculations and FELA.
Horizontal forces were applied to a
mudmat at eccentricities from, and heights
above, its centre. Particular attention was
given to identifying the levels of torsion
and overturning moment that may
realistically arise from non-concentric
loading applied within the envelope of the
mudmat and the equipment that it
supports.
Applying a horizontal force at an
eccentricity from the footing centre
generates a torsional load component.
The results quantify the gradual reduction
in bearing capacity as the eccentricity of
the horizontal force increases. The critical
failure mechanism in this case changes
from pure translation to combined
translation and twisting.
A horizontal force applied at a height
above the footing causes no reduction in
bearing capacity until a certain threshold is
reached, whereupon the capacity reduces
rapidly (Figure 7). This threshold was
e/B = 0 e/B = 0.5
h/B = 0 h/B = 0
FT/BLsu = 1.00 FT/BLsu = 0.70
(a)
e/B = 0 e/B = 0.5
h/B = 2 h/B = 2
FT/BLsu = 0.48 FT/BLsu = 0.47
(d)
Figure 9. Mechanisms for a footing subject to FT . Shaded by UB velocities
273
found to be higher when the force is
aligned parallel to the longer side of the
footing. The critical failure mechanism
changes from pure translation to
combined translation and rotation once
the threshold has been reached.
3D FELA with adaptive mesh
refinement was found to be an effective
method for analysing the bearing capacity
of shallow foundations on undrained clay
subject to complex 3D loading. FELA
results for a footing subject to combined
horizontal and torsional loading showed
excellent agreement with results from a
semi-analytical solution. Combining LB
and UB FELA proved beneficial as the
bracketing error was often sufficiently
small after the first iteration of the solution
that mesh refinement was not necessary.
This research is part of a more
extensive ongoing study aimed at
optimizing the design and sizing of
offshore mudmat foundations. Aspects
such as footing embedment, footing/soil
interface properties, vertical loading and
e/B = 1.118
h/B = 0
FT/BLsu = 0.43
(b) (c)
e/B = 1.118
h/B = 2
FT/BLsu = 0.41
(e) (f)
soil weight are also being considered in
this work.
ACKNOWLEDGEMENTS
The author is grateful for financial support
received from Subsea 7 for her DPhil, and
for helpful guidance and insight from her
DPhil supervisor, Dr Chris Martin.
REFERENCES
Bransby, M.F. & Randolph, M.F. (1998)
“Combined loading of skirted
foundations”, Géotechnique 48(5): 637–
655.
Brinch Hansen, J. (1970) “A revised and
extended formula for bearing capacity”,
Danish Geotechnical Institute Bulletin.
No. 28: 5-11.
DNV (Det Norske Veritas) (1992)
Classification Notes No. 30.4,
Foundations, Oslo, Norway.
Dunne, H.P., Martin, C.M., Muir, L.,
Brown, N. & Wallerand, R. (2015)
“Undrained bearing capacity of skirted
mudmats on inclined seabeds”, Proc.
3rd Int. Symp. on Frontiers in Offshore
Geotech., Oslo 2, 789-794.
Feng, X., Randolph, M.F., Gourvenec, S.
& Wallerand, R. (2014) “Design
approach for rectangular mudmats
under fully three-dimensional loading”,
Géotechnique 64(1): 51–63.
Gourvenec, S. Randolph, M. & Kingsnorth,
O. (2006) “Undrained bearing capacity
of square and rectangular footings”, Int.
J. Geomech. 6(3): 147-157.
ISO (International Organisation for
Standardisation) (2003) ISO 19901-4:
Petroleum and Natural Gas Industries –
Specific Requirements for Offshore
Structures – Part 4: Geotechnical and
Foundation Design Considerations, 1st
edn. Geneva, Switzerland.
Mana, D.S.K., Gourvenec, S.M. & Martin,
C.M. (2013) “Critical Skirt Spacing for
Shallow Foundations under General
Loading”, J. Geotech, Geoenv. Eng.
139(9): 1554-1566.
Makrodimopoulos, A. & Martin, C.M.
(2006) “Lower bound limit analysis of
274
cohesive-frictional materials using
second-order cone programming”, Int.
J. Num. Meth. Eng. 66(4): 604-634.
Makrodimopoulos, A. & Martin, C.M.
(2007) “Upper bound limit analysis
using simplex strain elements and
second-order cone programming”, Int.
J. Num. Anal. Meth. Geomech. 31(6):
835-865.
Martin, C.M., Dunne, H.P., Wallerand, R.
& Brown, N. (2015) “Three-dimensional
limit analysis of rectangular mudmat
foundations”, Proc. 3rd Int. Symp. On
Frontiers in Offshore Geotech., Oslo:
789-794.
Martin, C.M. & White, D.J. (2012) “Limit
analysis of the undrained bearing
capacity of offshore pipelines”,
Géotechnique. 62(9): 847-863.2.
Martin, C.M. (2011) “The use of adaptive
finite element limit analysis to reveal
slipline fields”, Géotechnique Letters 1:
23-29.
McDonald, S., Malachowski, J. & Wang,
Q. (2014) “Analysis of subsea
structures subject to significant torsion”,
Proc. 33rd Int. Conf. on Ocean,
Offshore and Arctic Eng. San Fran-
cisco, 1-10.
MOSEK ApS. (2014) “The MOSEK C
Optomizer API manual”, Version 7.1.
Online at www.mosek.com.
Nouri, H., Biscontin, G. & Aubeny, C.P.
(2014) “Undrained Sliding Resistance
of Shallow Foundations Subject to
Torsion”, J. Geotech. Geoenviron. Eng.
140(8).
Si H. (2013) “TetGen User’s manual”,
Version 1.5. Online at www.tetgen.org.
Ukritchon, B., Whittle, A. J. & Sloan, S. W.
(1998) “Undrained limit analysis for
combined loading of strip footings on
clay”, J. Geotech. Geoenviron. Eng.
124(3): 265-276.
Yun, G.J., Maconochie, A., Oliphant J., &
Bransby, F. (2009). “Undrained
Capacity of Surface Footings Subjected
to Combined V-H-T Loading”, Proc.
19th Int. Offshore and Polar Eng. Conf.,
Osaka, 9-14.
 25th European Young Geotechnical Engineers Conference
21st – 24th of June 2016, Sibiu, Romania

Jack-up vessel foundation stability during installation next to a

Wind Turbine Generator

Carlos MOLINA & Lindita KELLEZI

Geo, Copenhagen, DENMARK

ABSTRACT
After the commissioning of an Offshore Wind Farm, proper maintenance of the wind turbine
generators (WTG) takes a main role to ensure the energy supply during their life span. For doing
this work, jack-up vessels are required, just applied for the installation of WTGs and their
foundations. Thus, if the soil conditions are critical, multiple problems could be encountered during
the jack-up vessel installation and operation next to the WTG foundation. For a jack-up vessel
performing maintenance work near a WTG with gravity base foundation, critical soil / seabed
conditions are interpreted with regard to vessel installation. The multi-beam / bathymetry survey
indicated that one of the legs was placed on a slope that could compromise the vessel structure
stability. Conventional and finite element (FE) analyses are performed for analysing the risks
associated with the leg penetrations and stability during the jack-up vessel installation. A back-
analysis of the measured penetrations is commented and implemented. Useful conclusions and
recommendations for practitioners and engineers working with similar projects are drawn.

Keywords: Jack-up vessel, leg penetration analysis, foundation stability, finite

element (FE) analysis.

completely stable and well above the

1. INTRODUCTION
Geotechnical engineering assessments
are carried out for the offshore wind
industry, not only during the development
of an Offshore Wind Farm (OWF)
assessing the type of foundations of the
Wind Turbines Generators (WTG), but
also during the installation and
maintenance of the WTGs themselves.
The jack-up vessels are entitled to
assembly the necessary pieces of the
WTGs. In this process, they need to be
* presenting author
water currents. They are constructed with
legs, which penetrate into the seabed soil
and elevate the vessel ensuring stability
and a safe operation.
From a geotechnical engineering point of
view, multiple risks can appear. The most
recognized is the so called punch-through.
However, other risks such as leg sliding or
interaction with boulders, pipelines or
other nearby seabed structures, could also
appear. For reducing or eliminating such
risks, multiple mitigation measures can be
implemented e.g. gravel pads, excavation, positions were examined in the vicinity of
swiss cheesing etc. the WTG.
For an OWF with gravity base The final position is shown in Figure 1.
foundations, a quick turnover was needed
in order to assess a safe position to install
a jack-up vessel. Multiple positions and
different vessel configurations were
examined. The results from the
assessment are therefore explained and
analysed in this paper.

2. PROJECT DETAILS
2.1. Scope of work
It was requested to provision with
consulting services to ensure a safe
installation of the vessel during the
expected operation. The scope of work
was: Figure 1 Location of the WTG
x Geotechnical interpretation of the 2.3. Jack-up geometry and loads
available soil data The multi-purpose jack-up has a length of
x Correlation of geotechnical and 32 m and a width of 20 m, being the larger
geophysical data if available distance between legs 24m. Its size gives
x Assessment of suitability of the flexibility in this kind of short operations.
vessel for doing the installation, The jack-up system is based on four legs,
reporting possible risks for punch- each of them with a length of 50.6m,
through, rapid penetration or leg equipped with spud piles with a conical
extraction, including leg penetration shape at the bottom. They are circular with
analysis a diameter of 1.42 m and an area of 1.58
x Comment on other possible risks m2. The distance from tip to base is
(e.g. seabed features or leg sliding) 0.57m.
2.2. Location
The OWF is located offshore Denmark
within a shallow area with water depths
ranged from (2-6) m. A close monitoring of
water depths, made with a recent
topographic map of the seabed
(bathymetry), was the first requirement for
doing the expected maintenance work.
According to a received bathymetry, made
prior to the assessment, the water depth
within 100 m from the WTG varies from
(3.5 - 5.8) m DVR90.
In order to perform the maintenance and
reparation works, the vessel needed to be
close enough to the WTG, so that the Figure 2 Spudcan configuration (left) and spud
crane could safely operate under the leg / pile (right)
requirements defined in (Guidelines for
marine lifting & lowering operations
0027/ND). During the assignment, multiple

276
Alternatively, the vessel has the capability
of working with circular spudcans of 3 m
diameter, leading into an area of 7.07 m2.
For this project, both cases were primarily
studied. During the development of the
project, the client decided that the vessel
would not be assembled with spudcans.
Therefore, this paper will explain both
cases, but only provide feedback for the
configuration without spudcans.
The maximum expected preload was 475
tons / leg.
3. GEOTECHNICAL DATA
3.1. Geotechnical Investigations
The geotechnical investigations and
laboratory testing used for this
assessment were made by Geo in
different phases during 1997-2000. The
purpose of these investigations was
foundation design of the OWF. These
investigations were carried out with Geo´s
combined Cone Penetration Test (CPT)
and Vibrocore sampling (VC) seabed rig.
3 CPT/VCs were performed for each WTG
location, which would be used in this
assignment. Furthermore, laboratory
testing was made by Geo mainly to
confirm the strength of the chalk, common
in this site, by means of triaxial tests.
The relevant CPTs for the assessment
have a depth below seabed ranging from
3.4 m to 3.8 m. For the final location, the
distance from the nearest CPT to the
closest leg is 17 m whilst the furthest is
about 38 m (Figure 3).
The geotechnical data shows a soil profile
consisting of CLAY with various strengths,
underlain by CHALK starting from about
(4.0 – 5.0) m below the seabed.
3.2. Interpreted soil conditions
For the installation of jack-ups in the
offshore wind industry is frequent the lack
of soil data for each leg, different from
offshore oil and gas industry, where it has
become common to have one CPT for
each jack-up rig spudcan.
In an OWF, the geotechnical data used for
the installation assessments is usually the
obtained for the foundation design of the
WTG. Sometimes the distance to the
place of installation might be too large that
it is needed to correlate geotechnical and
geophysical data; like obtained by a Sub
Bottom Profiler (SBP). A SBP generates
an acoustic wave, which reflects into the
subsoil, discerning between layers.
Based on the available soil data, lower /
upper bound soil profiles applicable to leg
penetration analysis are interpreted and
summarized in Table 1.

Soil Depth J' M cu

Layer of [kN/m3] L/U L / U Bound
Layer Bound 2
[m] [kN/m ]
[q]

CLAY, 0.0 7.5 - 10-90 / 20-180

very soft –
to very
stiff 1.7

CLAY, 1.7 9.5 - 150 / 300

very stiff –
to hard
2.2
CLAY, 2.2 8.0 - 35-100/70-200
soft to –
very stiff
4.0/5.0
Figure 3 Location of the CPT/VCs, distance to CHALK 4.0/5.0 12.0 39 / 44 -
the jack-up and CPT/VCs logs –
10

Table 1 Interpreted soil profile

277
The upper / lower bound characteristic soil
parameters are selected as a cautious
estimate of the value affecting the
occurrence of the relevant limit state. The
undrained shear strength for the clay
layers is based on the CPT data applying
cone factor Nk = (10 - 20) for the upper
and lower bound soil parameters,
respectively.
The Nk factor was chosen considering
previous experience in the area,
laboratory tests and accounting the
distance from the furthest CPTs to the leg
location (38 m).
The soil profile exhibits a very soft to stiff
upper top clay layer, underlain by a thin
layer of very stiff to hard clay, overlying a
soft to very stiff layer of clay.
The precise depth to the chalk is
uncertain; it has been considered a level
of variation of 1 meter. The chalk has
been interpreted as drained with lower
bound friction angle of 39 degrees. This
was decided considering all the results
from the triaxial test along the OWF.
4. LEG PENETRATION ANALYSIS
4.1. Bearing capacity formulation and
use in jack-up foundations
The limited shear resistance or ultimate
bearing capacity has been discussed and
developed along the years, since the early
Terzaghi equations (1943). Geotechnical
engineers such as J. Bowles affirm that
‘there is currently no method of obtaining
the ultimate bearing capacity of a
foundation other than as an estimate’.
(Bowles, 2001)
J. Brinch Hansen published in 1970 the
revised and extended formula for bearing
capacity. This extended formula included
shape, depth, load inclination, base
inclination and ground inclination factors.
These factors were implemented to the
original formula (Hansen, 1970). The
general formulation is:
Q/A=0.5ȖBNȖsȖdȖiȖbȖgȖ+qNqsqdqiqbqgq+cNc
scdcicbcgc (1)
278
Where: Q = bearing capacity of foundation
base; A = fundamental area, B =
fundamental width; Ȗ = soil density; q =
vertical overburden; c = cohesion; sȖ, sq,
sc = shape factors; NȖ, Nq, Nc = bearing
capacity factor; dȖ, dq, dc = depth factors;
iȖ, iq, ic = load inclination factors, bȖ, bq, bc
= base inclination factors; gȖ, gq, gc =
ground inclination factors.
In the case of undrained conditions, it
would be more correct to introduce
additive constants, resulting in a simpler
equation:
Q/A=(ʌ+2)cu(1+sca+ dca- ica- bca- gca) (2)
Where: Q = bearing capacity of foundation
base; A = fundamental area, cu =
undrained shear strength; sca = shape
factor with additive constant; dca = depth
factor with additive constant; ica = load
inclination factor with additive constant, bca
= base inclination factor with additive
constant; gca = ground inclination factor
with additive constant.
The bearing capacity with depth can
assess the risks related with installation of
jack-ups. The most described are the
‘rapid penetration’ and ‘punch-through’
failures. These failures occur when the
bearing pressure dramatically reduces,
typically during the penetration of a stiff
layer overlying a softer layer (e.g. sand
over clay). This sudden penetration may
cause damage to the jack-up structure by
cause of the large displacement of the leg.
In critical cases, an accurate leg
penetration prediction is extremely
important for ensuring safety.
4.2. General considerations
The limit state analyses of the circular
conical spud leg / pile follow the guidelines
given in SNAME & ISO. The calculations
are based on design soil parameters with
partial coefficients Ȗm = 1.0.
In the assessment, the applied preload of
up to 475 tons/leg is considered as static
load. To conventionally define footing
penetration depth versus load, the
calculation of static bearing capacity of the expected up to the placement of the
spud leg / pile at various depths is carried spudcan over the chalk.
out. Different failure mechanisms are
assumed during the footing penetration in
multi-layered plastic medium. The spud
leg / pile bearing capacity is based on
Brinch Hansen’s theory and Geo in-house
program developed from the experience
with spudcan penetration predictions such
as (Kellezi, L., Xu, L., Molina, C., 2015).
To account for the backflow conditions, full
backflow is considered in the conventional
lower bound assessment and no backflow
in the upper bound assessment as
indicated in the conventional predictions
(Figure 4, Figure 5, Figure 17).
The spud leg / pile is simplified to a
circular footing with a flat bottom. The
effect of the shape is taken into account.
4.3. LPA results - Spudcan
The versatility of the vessel allows it to
work with or without spudcans. In an initial
assessment, the vessel configuration was
with spudcans. As preliminary assessment
it was decided to choose a predefined
depth of the chalk of 5 m, taking the
behaviour of the Chalk as undrained, with Figure 4 LPA results for the vessel with
unit weight (J') = 12 kN/m3 and undrained spudcan
shear strength (cu) over 800 kPa. Before
4.4. LPA results - Spud leg / pile
issuing a final penetration curve, it was
needed to further investigate the chalk After assessing the results, the final
properties within the OWF. At the same vessel’s configuration, without spudcans,
time this desk study was initiated, the was chosen.
configuration of the vessel changed to A new assessment was therefore needed.
without spudcans. In a first view, a reduction in the total base
From the lower / upper penetration curves, area could lead into further penetration,
assuming a generalized soil profile to giving a clear indication of the need of fully
represent the soil conditions at all 4 legs, assessing the chalk.
the spudcan were estimated to penetrate After a new review of the data available
(3.2 – 4.8) m. and Geo’s database from projects within
The curves (Figure 4) exhibit an increase the area. It was decided that the chalk
in the capacity from the seabed to two would behave as a drained material. Thus,
meters below seabed (bsb), while the accounting with lower bound soil
spudcan penetrates through a layer of parameter of 39 degrees, the leg could
clay with increased strength with depth. penetrate into the chalk.
After that, it is estimated a punch through / From the lower / upper penetration curves,
rapid penetration of around half a meter assuming a generalized soil profile to
during the penetration of the stiff clay represent the soil conditions at all 4 legs,
layer. Posteriorly, a similar soil mechanism the spud leg / piles were estimated to
as the one occurred within the first layer is penetrate (4 - 6) m.

279
The curves show a faster penetration into
the soil with less applied load than in the
previous case, which was expected as a
result of a smaller total area. An increment
in the penetration resistance is done in
two steps as the interpreted soil strength
also increments with depth in two different
layers. In the boundary between layers
can be shown a pick in the capacity prior
to a fast decrease of less than a meter, as
shown in Figure 5.
Figure 5 LPA results for the vessel without
spudcan (spud leg / pile)
The large range of penetration has its
origin in multiple uncertainties such as
lack of data at the leg locations or
ascertain the top of the chalk situated
beneath the clay. Some risks for punch
through at low load level are expected,
however not critical due to vessel draft
conditions.
The results from the leg penetration
analysis did not compromise the safety
during installation. However, the stability
280
of the vessel could be endanger in a
sloping soft clay soil.
5. FOUNDATION STABILITY
5.1. The Stability and seabed instability
problems
The stability of a jack-up unit installing
WTG is one of the requirements for getting
a certificate of approval from a Marine
Warranty Surveyor (MWS). During long
operations, it is needed to establish the
survival airgap range and whether there
are any limiting factors for a vessel
restricting the time that can be spent on
location. Vertical-Horizontal (V-H) capacity
envelopes are required in these
assessments. A bespoke communication
between structural and geotechnical
engineers is important when issuing Site
Specific Assessments (SSA), where the
considered problems are not only for
installation but also for operation. The
problems of stability during operation are
further described in code of practices as
ISO, SNAME or (Guidelines for site
specific assessments of jack-ups
0009/ND).
Seabed instability can result in slope
failures caused by multiple mechanisms.
Cyclic mobility or liquefaction starts when
progressive pore pressure is build up
caused by cyclic stresses within the soils,
as described in SNAME. During the
installation process, seabed instability
might cause risk of sliding, and is precisely
that fact studied in this paper.
5.2. Case
After a review of several positions, a final
vessel location was chosen, which would
be more adequate for the arrangement of
the cranes entitled to move the
components. However, for the chosen
vessel position, the bathymetry indicated
that the Starboard (SB) Leg should be
jacking on a slope. The fact of having SB
leg on top of a soft soil that might already
fail due to seabed mobility was a concern
and it was decided that a further study
should be performed.
For having a clear idea of the inclination of
the slope and the total dimensions of the
problem it was drawn a cross section
showing the most critical slope situation.
The first uncertainty was to know the
seafloor level at the base of the
foundation. For doing that, in absence of
the corresponding foundation drawing,
after reviewing the CPTs, was assumed
the proportional depth of the chalk where
the foundation was resting.
The final section is shown in Figure 6.
Figure 6 Schematic foundation section
5.3. Computational Finite Element
Method (FEM)
The FEM have been used in geotechnical
engineering for a large range of problems,
both onshore and offshore. The large
range of available FE software packages
could confuse the user on its decision of
which fits better the purpose. The increase
in the efficiency of the hardware used for
FE, has also augment the popularity of
advance programs with a higher
computational cost such as two-
dimensional (2D) and three-dimensional
(3D) Large Deformation FE (LDFE). For
certain geotechnical processes which
involve large displacements, new recent
developments anticipate the combination
of FE with other methods such as the
Material Point Method (MPM) (Brinkgreve
et al, 2015)
The use of FE together with conventional
calculations gives a more accurate
solution that is specially needed in critical
281
multi-layered soil conditions with high risk
of punch-through failure, where the peak
bearing capacity prior to failure can be
calculated and compare with conventional
methods. However, when the soil
conditions are favourable, there is no
need to carry out other calculations than
conventional.
In this case, to calculate the peak bearing
capacity of each of the layers was not
considered necessary. However, the risk
for sliding and therefore how the soil
would behave under such circumstances
could not be dismissed. Doing a FE model
of the situation would help to understand
the behaviour of the soil stratigraphy.
5.4. The FE model – Plaxis 2D
The stability (risk for sliding) of the SB Leg
towards the WTG is investigated by
applying the FE with Plaxis 2D 2015.
It is used a plane strain model with 15-
noded elements. The soil material models
are based on Mohr-Coulomb following the
low bound parameters as in Table 1.
The spud pile is simplified as a circular
footing and the soil immediately in contact
with the leg is removed from the sides to
avoid soil failure problems with origin in
the side friction, investigated in previous
models.
The geometry of the problem was ready
from the CAD section showed in Figure 6
Schematic foundation section; and
therefore, imported to Plaxis.
Two FE analyses are carried out
assuming SB Leg in-placed at two
different depths. One placed on an initial
depth of 1.0 m bsb (Figure 7) and another
with full base contact at the very stiff to
hard clay layer 2.2-4.5 m bsb (Figure 8).
Figure 7 Plaxis geometry, model 1. Leg over
first layer of soft to stiff clay.
Figure 8 Plaxis geometry, model 2. Leg over
second layer of stiff clay.
Figure 12 Model 1. Total deviatoric strains
The first model assumed that after an
initial penetration the spud pile would be
Analysing the deformations, the total
into a soft to stiff layer of clay.
displacements showed how the leg could
move towards the gravity base foundation
and the failure figure goes towards the
slope base (Figure 12). The results
showed in (Figures 9-12) are for an
approximate load equal to 94 tons/leg.

The second model aim to prove the

reaction of the soil in a deeper state,
which means in place in the stiff boundary
layer that separates both softer layers.
Figure 9 Model 1. Deformed mesh.

Figure 13 Model 2. Deformed mesh

Figure 10 Model 1. Total displacements

Figure 14 Model 2. Total displacements

Figure 11 Model 1. Plastic Points

282
 accommodated on virgin seabed, being
further away from the slope.

6. BACK ANALYSIS OF MEASURED

PENETRATIONS
After installation, corresponding feedback
was received and implemented in order to
check the results from the analysis.
The feedback received was 4.5 m for the
Figure 15 Model 2. Plastic points
maximum preload. The following
description of the leg penetration was
delivered by the towmaster: ‘a meter of
soft material followed by a stiffer material
and having another penetration in softer
material similar as the first one, until the
maximum preload was reached’.

Figure 16 Model 2. Total deviatoric strains

In terms of deformations, both: total

displacements (Figure 14) and total
deviatoric strains (Figure 16) showed a
more stable response with a further
penetration affecting the soil below, as
expected. The results showed in Figures
13-16 are for an approximate load equal
to 144 tons/leg.

Based on these analyses, it can be

concluded that there is risk for sliding of
SB Leg until penetrations of approximately
(2.5 - 4.0) m are reached. In order to avoid
this, due to low level of applied loads,
stomping of this leg was recommended,
while monitoring the rack phase difference
(RPD), until a vertical position is ensured
while the leg reaches penetrations of (4.0 -
4.5) m. Further preloading after this Figure 17 LPA – Back analysis from
achieved penetration depth is not measured penetrations - Lower / Upper / Best
expected to be associated with risk for Estimate
sliding.
However, to increase the safety against This description indicates that the first clay
sliding for SB Leg, it was recommended (if layer was shallower than expected and
possible), the vessel could be slightly gives an idea of the variability of the
moved to the northeast, having SB Leg stratigraphy within an area.

283
A best estimate (BE) curve has been
traced averaging both upper / lower bound
and adjusting the uncertainty related with
the depth of the chalk.
7. CONCLUSIONS
After several attempts, a final vessel
position was chosen, close enough to the
WTG in order to make possible the
maintenance work and taking into account
the seabed conditions (slope) at the
surrounding of WTG gravity foundation.
The current assessment covers many
geotechnical engineering challenges,
starting with the: interpretation of the soil
data and correlation of in situ and
laboratory test; leg penetration analysis for
a generalized interpreted soil profile
(representing the soil conditions at all four
legs); FE calculations of the stability of SB
Leg located near the sloping seabed at
the vicinity of a gravity base foundation.
The vessel was installed without risk,
taking into consideration the suggested
conclusions such as monitoring of the
RPD during penetration into the top clay
layer.
ACKNOWLEDGEMENTS
The authors would like to acknowledge
the assistance of Kenneth Andersson,
from Svensk Sjöentreprenad (SSE) AB
during the scope of this project and for
providing the feedback from the
installation. The participation of the
presenting author in this conference is
supported by the Danish Geotechnical
Society (DGF) and Geo.
REFERENCES
Bowles, J. E. (2001). Foundation Analysis
and Design, 5th edition.
Brinkgreve et al. (2015). Beyond the Finite
Element Method in Geotechnical
Analysis.
Eurocode 7: Geotechnical Design – Part
1; General Rules. EN 1997-1 2004.
(n.d.).
Guidelines for site specific assessments of
jack-ups 0009/ND.
284
Hansen, J. (1970). A revised and
extended formula for bearing
capacity. Bulletin No. 28. The
Danish Geotechnical Institute.
ISO 19905-1 Petroleum and Natural Gas
Industries Site Specific Assessment
of Mobile Offshore Part 1 Jack-ups
(ISO 19905-1:2012).
Kellezi, L., Kudsk, G. (2009). ‘Jack-up
Foundation, FE Modelling of Punch
Through for Sand over Clay’. 12th
International Conf. on Jack-up
Platform. , (pp. page 1-12.). London
UK.
Kellezi, L., Kudsk, G., Hofstede, H. (2007).
’Seabed Instability and 3D FE Jack-
up Soil-Structure Interaction
Analysis’. 14th European Conf. on
Soil Mech. & Geotech. Eng.
ECSMGE, (pp. Volume 5 page 247
- 252). Madrid, Spain.
Kellezi, L., Stadsgaard, H. (2012). ‘Design
of Gravel Banks – a Way to Avoid
Jack-Up Spudcan Punch Through
Type of Failure’. OTC 2012, (p.
Paper no. OTC 23184). Houston,
USA.
Kellezi, L., Stromann, H. (2003). ‘FEM
Analysis of Jack-up Spudcan
Penetration for Multi-Layered
Critical Soil Conditions’. Proceeding
of BGA International Conference on
Foundations, (pp. page 410-420).
Dundee, Scotland.
Kellezi, L., Xu, L., Molina, C. (2015).
Seabed Remediation for Safe
Sequential Jack-Up Vessel
Installations. 15th International
Conf. ‘The Jack-up Platform
Design, Construction & Operation’,
(Paper no. 18). London, UK.
SNAME. (2008). Guidelines for Site
Specific Assessment of Mobile
Jack-Up Units, Technical &
Research Bulletin 5-5A, January
2002.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Numerical analysis of penetration problems in clay with the

Particle Finite Element Method

Lluís MONFORTE1*
1
Universitat Politècnica de Catalunya - BarcelonaTech, Departament d’Enginyeria del
Terreny, Barcelona, SPAIN

ABSTRACT
This paper highlights a computational framework for the numerical analysis of fluid saturated soil-
structure interaction problems. The Particle Finite Element method is used to solve the linear
momentum and mass of the mixture balance equations at large strains. The robustness and
accuracy of the proposal is numerically demonstrated by means of the analysis of benchmark
examples. Indeed, in the oedomeneter test example, it is shown that using a large deformation
theory may reflect results that are artificially excluded by the linear theory. The paper concludes
with a parametric analysis of a cone penetration test, where the influence of the contact roughness
and permeability to the shaft friction and cone resistance are assessed.

Keywords: CPT, porous media, large strain, penetration test, PFEM.

element simulations; however, there is still

1. INTRODUCTION
The Cone Penetration Test (CPT) is one
of the most widely used in situ methods to
characterize soil properties. The test
method consists of pushing an
instrumented cone into the ground at a
controlled rate. During the penetration,
measurements of the tip resistance and
the friction at the sleeve are recorded. In
addition, the pore water pressure may be
measured at different locations.
Based on these measurements,
constitutive soil parameters may be
estimated with correlations. A large
number of interpretation approaches have
been obtained using a multitude of
techniques, ranging from bearing capacity
analogies to large deformation finite
* presenting author
a large spread between the various
proposals.
A large amount of numerical and
analytical work has been devoted to the
analysis of Cone Penetration test in
undrained clay using a total stress
approach. More recently, hydro-
mechanical numerical analyses have been
performed; the effect of partially drained
conditions in the cone resistance and
cone penetration consolidation test has
been evaluated (Yi et al, 2012, Sheng et
al, 2014).
The majority of hydromechanical
numerical proposals only consider smooth
interfaces between the soil and the
structure. This hypothesis may be
unrealistic: as pointed out by Tsubakihara
& Kishida (1993), that presented results of
interface direct shear tests between clay
and steel, the friction angle of the interface
may be high: the clay friction angle was
estimated at 27º whereas the angle of the
interface was 23º.
Using a realistic description of soil
behaviour (e.g. elasto-plastic effective
stress response) increases both the range
of conditions for and the precision of the
interpretation. The downside is that a
number of material non-linearities are
added to a problem that already features
severe geometric non-linearity.
Finite element method is well suited to
include all sorts of non-linearities.
However, when a Lagrangian formulation
is employed –typically when a path
dependent material model is used- the
mesh may experience severe distortion,
leading to numerical inaccuracies and
even rendering the calculation impossible.
In order to alleviate this problem several
numerical techniques based on FEM have
been proposed: adaptive methods,
Arbitrary Lagrangian-Eulerian Material
Point Method, among others.
In this work, the Particle Finite Element
Method is employed to simulate the
penetration of a rigid probe into the soil.
The method is characterized by a particle
discretization of the domain: every time-
step a finite element mesh –whose nodes
are the particles- is build using a
Delaunay’s tessellation and the solution is
evaluated using a well shaped, low order
finite element mesh (Oñate et al, 2004).
The soil-water mixture is modelled as a
two-phase continuum employing a finite
deformation formulation: the balance of
linear momentum and mass of the mixture
are written following the movement of the
solid skeleton. The water flow is assumed
to obey a generalization of the Darcy’s
Law whereas a multiplicative hyperelastic-
plastic constitutive response is assumed
for the solid skeleton.
The proposed approach is assessed
against numerical benchmark examples.
Finally, preliminary results of a parametric
286
analysis of the Cone Penetration test are
presented: the effect of the permeability
and the friction angle of the cone-soil
interface are evaluated.
2. NUMERICAL MODEL
This section outlines the numerical
procedures used in this work. First the
Particle Finite Element method (PFEM) is
briefly reviewed; then, the balance
equations are highlighted. Finally, the
constitutive equations are described.
2.1. Particle Finite Element method
In the PFEM the continuum is modelled
using an Updated Lagrangian formulation;
that is, a Lagrangian description of the
motion is used and all variables and their
derivatives are referred to the deformed
configuration. The nodes discretizing the
analysis domain are treated as material
particles whose motion is tracked during
the transient solution; the interaction of
these particles is computed using the finite
element method and the particles serves
as nodes of the FE mesh. Another
particularity of PFEM is that only low order
elements are used –linear triangles in 2D.
Periodically, the FE mesh is re-
triangulated in order to alleviate problems
that arise when the mesh becomes highly
distorted. In addition, h-adaptive
techniques are employed to obtain a
better discretization in areas of the domain
with large plastic deformations.
A typical solution algorithm involves the
following steps (Oñate et al, 2004):
a. Discretize the domain with a Finite
Element mesh. Define the shape
and movement of the rigid
structure.
b. Identify the external boundaries.
Search the nodes that are in
contact with the rigid structure.
c. Compute some time-steps of the
coupled hydro-mechanical problem.
d. Construct a new mesh. This step
may include a re-triangulation of the
domain, introduce new particles in
an adaptive fashion and interpolate
 the state variables between the
previous mesh and the new one.
e. Go back to step b. and repeat the
solution process for the next time-
steps.
2.2. Balance equations
The soil-water mixture is modelled as a
two-phase continuum employing a finite
deformation formulation. The equations of
balance of linear momentum and mass of
the mixture are written following the
movement of the solid skeleton, where the
unknown fields are the solid skeleton
displacements and the fluid pressure (u-pw
formulation).
Excluding inertial effects, the Updated
Lagrangian form of the governing
equations read:
This system is completed with the
appropriate initial and boundary
conditions. stands for the Cauchy total
stress, b are the external volumetric loads,
is the solid skeleton velocity and
the Darcy’s velocity.
The system of equations is non-linear
geometrically since balance equations are
imposed in the new (unknown)
configuration. The proposed approach
slightly differs in the treatment of some
terms with respect to (Borja and Alarcón,
1995; Larsson and Larsson, 2002).
In the present implementation both
equations are solved in a monolithic
approach and an implicit time-marching
shceme is employed. The space is
discretized with linear triangles and the
same shape functions are used both for
displacements and water pressure. This
type of elements may produce spurious
oscillations in the pore pressure field in
nearly undrained conditions as a result of
failure to satisfy the inf-sup conditions. In
order to alleviate these oscillations, a
Pressure Laplacian stabilization term is
added to the mixture mass balance
equation (Presig and Prévost, 2011). This
way, the weak form of the problem is
slightly modified so that improved
numerical stability is achieved without
compromising consistency.
2.3. Constitutive relations
In this work, a hyperelastic-based finite
strains elasto-plastic constitutive response
is assumed for the solid skeleton. The
elastic model is defined by the Houlsby
(1985) hyperelastic model; then, the
effective volumetric and deviatoric
Kirchhoff stresses -
where is the Jacobian (a measure of the
volume change) and is the effective
Cauchy stress tensor - are obtained as:
where is the elastic Hencky Strain,
, is the swelling slope, is the
constant part of the shear modulus, is
a reference pressure and is a
parameter that controls the coupling of the
is volumetric and deviatoric response and
the Poisson’s ratio.
The problem is completed with the
Modified Cam Clay yield surface and
hardening law (Borja et al, 2002):
where is the second stress invariant,
is the Lode’s angle, is the
preconsolidation pressure, is the
Hencky volumetric plastic strain,
and is the slope of the normal
compression line.
The integration of stresses is performed
with an explicit integration scheme with
adaptive sub-stepping and a correction for
the yield surface drift is applied (Monforte
et al, 2014).
287
2.4. Contact Constraints the coefficient of consolidation, ,
The interaction between multiple bodies
produces a set of normal and tangential where is the constrained modulus. In
forces in the interface. Mathematically, order to evaluate large strains effects, the
contact conditions are expressed as a set same problem has been computed with
of geometrical restrictions to the solution. two sets of parameters maintaining
Indeed, one of the contacting bodies –the constant the coefficient of consolidation,
structure- is assumed rigid; this hypothesis , and the Poisson’s ratio,
is valid when the ratio between the soil
and structure Young’s moduli is large . Two Young’s modulus have been
(Sheng et al, 2005). used: E = 2·102 kPa and E = 2·105 kPa.
In this work, the contact constraints are
imposed to the solution with the penalty
method; this way, the restrictions are only
approximately fulfilled.
The tangential part of the contact is
idealized with an elasto-plastic analogy,
where the so-called stick conditions –no
tangential relative displacement between
the bodies- is modelled as the elastic
loading part and the slip conditions –which
is characterized by the relative tangential
movement- is represented by plastic flow. Figure 1: Oedometer test. Dissipation curves
for two sets of parameters.
A Coulomb law is used as the yield
surface of the tangential part of the Figure 1 shows the variation of the
contact: water pressure at the bottom of the
sample as a function of the time factor for
the two sets of parameters. In the larger
where is the normal effective stress Young’s modulus case, both
acting at the interface, is the interface displacements and deformations are small
and the solution agrees well with the small
tangential stress, and is the strains analytical solution. The other case
interface friction angle. is different due to the severe geometric
non-linearity: as consolidation takes place,
3. NUMERICAL ANALYSIS the height of the domain decreases; thus,
In this section, first a benchmark example the draining path length reduces, see
–the oedometer test- is presented to Figure 2.
assess the implementation; then, the
constitutive model is exemplified with a
constant volume shear test. Finally, a
parametric analysis of the CPT is
presented.
3.1. Oedometer test
The first example corresponds to an
oedometer test in a weightless soil. For
the sake of simplicity, in this example the
soil is assumed to obey a linear
hyperelastic behaviour.
Small strains analytical solutions states
that the key constitutive parameter that Figure 2: Oedometer test. Isochrones of the
controls the pore pressure dissipation is water pressure for two sets of parameters.
288

Figure 3: Constant volume simple shear test.
Effective stress trajectory in the p-30.5J2 plane
for OCR =1 and 4 for different temporal
discretizations.
3.2. Constant volume simple shear test
The current example consists on a
constant volume simple shear test and
serves to validate the numerical
implementation of the constitutive model.
The same constitutive parameters than a
reference solution are used (Rouainia and
Muir Wood, 2000).
This example consists on the
evaluation at one Gauss point of a
displacement-driven problem. The
displacement field is parameterized by a
pseudo-time variable, t, and is written as:
, for . friction sleeve, at 7.5 cone radii above the
Two different analyses have been
performed with different overconsolidation
ratios (OCR). In both cases the
preconsolidation pressure is equal to 80
kPa and in one case the analysis starts at
normally consolidated conditions whereas
in the other the OCR is equal to 4. In both
cases the analysis begins at hydrostatic
stress.
Figure 3 shows the effective stress
trajectories; both tests tend to the critical
state line. As it can be seen in the
overcosolidated test, there exists a
change in mean stress at constant volume
in elastic regime due to the coupling in the
non-linear hyperelastic model. Then, the
289
stress state tends to the critical state line
in elasto-plastic regime. Note than these
effective stress trajectories are the same
to an undrained triaxial test.
The obtained results are in good
agreement with those reported by the
reference solution using an implicit stress
integration technique. In addition, Figure 3
also shows the stress trajectories
computed with different time discretization:
the problem has been computed with a
different number of steps, ranging from 10
to 1000 steps. In both cases, the solution
computed with a small number of steps
converges towards that obtained with a
larger resolution and minimal
discrepancies between them exists; this is
due to the use of an adaptive substepping
algorithm.
3.3. Cone Penetration Test
The last numerical analysis consists on
the penetration of a CPT in a Modified
Cam Clay soil. Several simulations are
presented with different permeabilities and
interface friction angles and attention is
paid at variations of the net cone
resistance, the friction sleeve resistance
and the water pressure at three positions:
u1 position (at the midface of the cone), u2
position (at the apex between the cone
and the shaft) and u3 position (above the
apex). Four different permeabilities have
been considered, K = 3.3·10-6, 5·10-7, 10-7,
10-9, m/s and two different interface
friction angles, and . The
eight combinations have been considered.
The basic constitutive parameters are
listed in Table 2; the selected values try to
mimic the example reported by Sheng et
al (2014), where a parametric analysis of
the effect of the penetration velocity for
smooth CPT is presented. In this work, the
self weight of the soil has been omitted
and the initial effective stress and water
pressure have been chosen similar those
encountered at the final penetration depth
of the reference solution.
 Figure 4: Cone penetration test. Water pressure (kPa) profiles at the final penetration
depth for different permeabilities: K = 10-7 m/s (left), K = 5·10-7 m/s (center), K =
3.3·10-6 m/s (right) for smooth interface.
Table 1: Cone penetration test. Constitutive
parameters adopted for the Modified Cam
Clay model
(kPa)
0.016 0.1 23.5 400
(kPa) (kPa) (kPa)
70 1 57.5 28.9
In the beginning of the computation the
cone is assumed in a wished-in-place
situation at a depth of 3 cone radii. Then
cone is pushed at the standard velocity
( m/s). The domain is 28 times
the cone radius in with and 56 times in
depth. The bottom boundary of the soil
domain is assumed to be fully drained
(fixed water pressure). A constant vertical
stress is applied at the top boundary. The
radial displacement is fixed at the left and
right boundaries whereas null
displacement in all directions is imposed
at the bottom boundary.
Figure 4 shows the water pressure
profiles at the final penetration depth for
the smooth case. When low permeabilities
are considered, the maximum of the water
pressure is found below the apex between
the tip and the shaft and decreases in all
directions. In addition, large excess water
pressures are found along the shaft of the
structure. In the three cases with the
lowest permeability, a negative excess
water pressure occurs below the cone tip
and the magnitude is larger with lower
permeabilities.
The penetration in the lowest
(kPa)
permeability case takes place in almost
undrained conditions: results reveal that
10 the volumetric deformation is small, less
than 0.1% in the entire domain. In
(kPa)
contrast, in the other scenarios with grater
80 permeabilities largest volume
deformations are found whereas lower
excess pore water pressure are
encountered with respect to the previous
case, confirming that penetration takes
place in partially drained conditions. These
observations are common for both
interface friction angles.
Figure 5 shows the evolution of the net
cone resistance, the water pressure at the
measurement positions and the sleeve
friction resistance as a function of
penetration depth. In all the cases a clear
stationary state of the net cone resistance
and water pressure at the u1 and u2
position is achieved after a penetration of
approximately of 7 radii. At the first 6 radii
of penetration the sleeve friction
resistance increases rapidly; at this depth
all the friction sleeve is in contact with the
soil. Then, the friction sleeve resistance
varies until a stationary value is found. In
some of cases, it is unclear if the water
pressure at the u3 position and the sleeve
friction reach a stationary value.
290
 Figure 6: Cone penetration test. Smooth are depicted with continuous lines whereas
rough (14) are shown with discontinuous lines.

Figure 5: CPT. Net cone resistance (qn), water pressure at the three measurement
positions and friction sleeve resistance in terms of the penetration depth. Smooth
cases (continuous lines) and rough cases (discontinuous lines).
The reported curves present some minimum and maximum are observed in
oscillations that are more pronounced the two lowest permeability scenarios.
when a rough contact is employed and in
the lowest permeability case. These
oscillations are interpreted as the mesh-
dependence of the solution and error
introduced in the transfer of variables
between different finite element meshes.
For the smooth case, the obtained results
compare well with those reported by the
reference solution (Sheng et al, 2014) that
uses a different implementation of the
Modified Cam Clay.
Figure 6 shows the mean value at the
last 10 penetration radii of the relevant
CPT reactions in terms of the interface
friction angle and the dimensionless
penetration velocity, . The net cone
resistance increases with the permeability;
the lowest cone resistance corresponds to
the undrained case. When the interface Figure 6: CPT. Steady state net cone
between the clay and the soil is assumed resistance, water pressure and friction
rough, the cone resistance increases at a resistance fs and ratio fr in terms of the
steeper rate as the permeability increases. dimensionless velocity. Smooth (continuous)
The smallest friction sleeve resistance and rough (discontinuous).
is found in the undrained analysis and it The water pressure at the three
increases with larger permeabilities. On locations decreases as permeability
increases. The largest water pressure is
the other hand, the friction ratio, ,
found in the u1 position whereas the
varies between 2.9% and 6.4%; the lowest one is in the u3 position. In addition,
in the simulations that assume a rough
291
interface, the water pressure at the u1
position is slightly larger than the one
encountered with a smooth interface; on
the other hand, the water pressure in the
u2 position is lower considering a rough
interface.
4. CONCLUSIONS
In this work, a numerical framework for the
analysis of saturated porous media
undergoing large deformations has been
presented. By means of the analysis of the
oedometer test is has been shown that the
obtained results are accurate; indeed,
using a large deformation theory may
reflect results that are artificially excluded
by the linear theory.
Preliminary results of a parametric
analysis of the Cone Penetration Test in a
Modified Cam Clay soil have been
reported. The effect of the interface friction
angle and the permeability of the soil on
the measured reactions has been
assessed.
The developed numerical scheme
appears to be a promising tool for the
simulation of penetration problems in
geotechnics.
ACKNOWLEDGEMENTS
The author would like to thank his thesis
supervisors, Prof Antonio Gens, Dr
Marcos Arroyo and Dr Josep Maria
Carbonell, for their guidance during this
work.
The financial support of the Sociedad
Española de Mecánica de Suelos to cover
the expenses of the trip is acknowledged.
The support of the Ministry of Education
of Spain through research grant BIA2011-
27217 is gratefully appreciated.
REFERENCES
Borja, R.I. & Alarcón, E (1995) “A
mathematical framework for finite
strains elastoplastic consolidation. Part
I”, Comput. Methods in App. Mech. and
Eng., 122(1-2):145-171.
Houlsby, G.T. (1985) ”The use of variable
shear modulus in elasto-plastic models
292
for clays”, Computers and Geotechnics,
1(1):3-13.
Larsson, J & Larsson, R (2002) ”Non-
linear analysis of nearly saturated
porous media: theorical and numerical
formulation”, Comput. Methods in App.
Mech. and Eng., 191(1):3885-3907.
Monforte, L., Arroyo, M., Gens, A. &
Carbonell, J.M. (2014) ”Explicit finite
deformation stress integration of the
elasto-plastic constitutive equations”,
Computers Methods and Recent
Advances in Geomechanics –
Proceedings of the 14th Int. Conf. of
IACMAG, 267-272.
Oñate, E., Idelsohn, S.R., del Pin, F. &
Aubry, R. (2004) ”The Particle Finite
Element Method. An overview”,
International Journal of Computational
Methods, 1(2):267-304.
Preisig, M. & Prévost, J.H. (2011)
“Stabilization procedures in coupled
poromechanics problems: A critical
assessment”, Int. J. Numer. Anal. Meth.
Geomech., 35:1207-1225.
Rounainia, M. & Muir Wood, D. (2000) “An
implicit constitutive algorithm for finite
strain cam clay elasto-plastic model”,
Mechanics of Cohesive-frictional
materials, 5(6):469-489.
Sheng, D., Eigenbrod, K.D. & Wriggers, P.
(2005) “Finite element analysis of pile
installation using large-slip frictional
contact”, Computers and Geotechnics,
32(1):17-26.
Sheng, D., Kelly, R., Pineda, J. & Lachlan,
B. (2014), “Numerical study of rate
effects in cone penetration test”, 3rd
International symposium on Cone
Penetration Testing, 419-428.
Tsubakihara, Y. & Kishida, H (1993)
“Frictional behaviour between normally
consolidate clay and steel by two direct
shear type apparatuses”, Soils and
Foundations, 33(2):1-13.
Yi, J.T., Goh, S.H., Lee, F.H. and
Randolph, M.F. (2012) “A numerical
study of cone penetration in fine-
grained soils allowing for consolidation
effects”, Geotechnique, 62(8):707-719.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Modelling Spudcan penetration using Abaqus CEL method

Aleksander Worren1*
1
Technical Norwegian Geotechnical Institute, department of computational geomechanics,
Oslo

ABSTRACT
A hazard during installation of jack-up spudcans is punch-through, which is characterized by a
peak resistance, followed by a significant reduction in spudcan resistance. This might lead to an
uncontrolled rapid leg penetration as the installation generally is load-controlled. The problem is
typical for sites where a stiff soil layer is overlying a soft clay layer. Accurate calculation of the
expected displacement-resistance curve for these soil conditions is therefore important in order to
reduce the risk of uncontrolled punch-through conditions.
Numerical simulation of spud-can penetration into seabed during installation of jack-up platforms is
a complex problem involving both large strains and large displacements where the geometry
changes during penetration e.g. interface between layers. The Coupled-Eulerian-Lagrangion (CEL)
method available in the finite element program package ABAQUS is suitable for this type of
problem. The main aim of this Master Thesis is therefore to use the CEL method to analyse some
published examples of spudcan penetration.

Keywords: Spudcan, Offshore, Abaqus, CEL,

conical footings. Spudcans have a

1. INTRODUCTION
Jack-up rigs are the most common type of
mobile platforms. They operate at shallow
and up to moderate depths (167 meters
(World Fleet of Jack-Up Drilling Rigs,
2012)). The spudcan may be penetrated
up to tens of meters into the seabed if the
soil is soft. The installation is performed by
applying vertical load from the jack-rig and
water ballast in the hull.
Jack-up rigs have movable legs, which
may be jacked down into the seabed to
give stabilization under operation, hence
the name jack-up. The legs often stand on
spudcan foundations, which are steel
* presenting author
diameter of 10 to around 20 meters. The
main objective to the spudcan is to
distribute the load from the jack-up rig and
give stability. The spudcan may be
penetrated up to tens of meters into the
seabed if the soil is soft. The installation is
performed by applying vertical load from
the jack-rig and water ballast in the hull.
This means that the penetration is load
controlled, and the average penetration
rate is often around 1m/hour (Tjahyono,
2011).
The legs and jack-up rig may be damaged
if the penetration rate becomes too large.
It is therefore important to know the soil
characterization and expected load-
response curve. A typical hazard is punch-
through during the installation. It is
characterized by a peak resistance during
the installation, followed by a fast
reduction in spudcan resistance. This
might lead to a rapid penetration because
the installation is load-controlled. This
problem is typical for sites where a stiff
soil layer is overlying a soft clay layer.
Punch-trough of a jack-up leg will cause
the platform to tilt which consequently will
give rise to large bending moments. The
Figure 2-1 Lagrangian element deformation
bending moment may lead to failure in the on the top, and eulerian deformation at the
jack-up legs and connection between the bottom. (Nonlinear finite elements/Lagrangian
legs and rig. This can endanger personnel and Eulerian descriptions, 2010)
and result in huge economic loss.
The main aim of this master thesis is to It is possible to have more than one
use the CEL method in Abaqus/Explicit to material in the eulerian mesh using
analyze some published examples of Abaqus/Explicit. The materials are
spudcan-penetration. Special focus will be assigned using initial conditions in the
to consider the effect of large strain in start of the analysis. The elements are by
clay, e.g. gradual reduction of the default empty (volume fraction = 0), while
undrained shear strength with increasing the volume fraction is one when the
strain. element is completely filled with material
(Figure 2-2). The material is tracked as it
2. THEORY moves through the mesh by calculating
All of the finite element calculations will be the eulerian volume fraction (EVF) of each
executed in Abaqus/Explicit. element. The eulerian material will
2.1. Coupled Eulerian-Lagrangian disappear from the simulation if it moves
outside the mesh. The material
Lagrangian elements have material fixed boundaries (interfaces) are tracked using
to the nodes. As a result, the elements will the computed eulerian volume fraction by
deform as the material deforms. This may each incrementation. Abaqus/Explicit uses
lead to numerical problems in large an interface reconstruction algorithm that
deformation problems. Heavily distorted approximates a planar boundary within
elements will not work well, and problems each element (Abaqus 6.12 Analysis
like mesh locking, and numerical instability User’s Manual). The simplification with
can follow. planar boundary may lead to discontinues
In eulerian element formulation, the nodes in the interfaces if a coarse mesh is used.
are fixed in space (coordinate system),
while the material is free to move (flow)
through the mesh (Abaqus 6.12 Analysis
User’s Manual). The mesh will
consequently not deform as the material
deforms. Figure 2-1 shows the
deformation of a lagrangian mesh over the
deformation of a eulerian mesh. This is
particularly useful when dealing with large
deformation, where the lagrangian Figure 2-2 Volume fractions (Abaqus/CAE
elements would get heavily distorted and User's Manual)
encounter numerical problems.
294
2.2. T-bar penetration
The preliminary analyses included T-bar
penetration tests.
The T-bar penetration test is similar to the
CPT (Cone Penetration Test) except that it
is a horizontal cylinder that is pushed
through the soil. It is used to define the
cohesion for soft clays. The resistance is
measured during the penetration, and the
undrained resistance force is calculated.


P is the force per unit length acting on the
cylinder, is the undrained shear
strength, d is the diameter of the cylinder,
and is the bar factor. The bar factor is
dependent on the roughness of the
cylinder. The theoretical value is
approximately 12 for rough contact, and 9
for smooth contact (Randolph, M.F. &
Houlsby, G.T., 1984) (Stewart, D.P. &
Randolph, M.F., 1994). This factor is
theoretical, and is based on a plastic
solution with a soil model which is elastic-
perfectly plastic. Effects like strain-rate
dependency, strain-softening and
anisotropy are not included. Strain-rate
dependency and strain-softening is shown
to have a significant effect on the bar
factor (Liyanapathirana, 2008), while the
bar factor is relatively insensitive to
anisotropy (Randolph, M.F. & Andersen,
K.H., 2006).
3. RESULTS
3.1. T-bar test
First a mesh convergence test was
performed in order to study the effect of
the element size for soil-flow around the T-
bar. The speed of the T-bar was not
studied in these analyses, and it is not the
same for all the models. It might, however,
have been more efficient to study the
speed of the T-bar in advance in order to
know which speed to use to avoid
oscillations in the result.
Figure 3-1 Mesh convergence for the T-bar
test
The calculations show that the T-bar factor
is significantly affected by the element
size. Convergence to a constant value
was not possible as the calculation time
become too great for smaller element size.
However, the models with the smallest
element size performed well, and the error
for element size 0.0010 is less than 5 %.
The flow around the T-bar displays why
the models with denser mesh give less
resistance and therefor more accurate
results. In Figure 3-2 is the flow pattern of
the model with element size 0.0015
displayed over the model with element
size 0.0070. A smaller amount of soil flow
around the T-bar in the model with denser
mesh, and consequently less resistance is
measured. The edge of the soil that flows
around the T-bar is located 0.05 meters
from the side of the T-bar for element size
0.0015, while the distance is 0.07 meters
for element size 0.0070. The shear band
which is established along the edge of the
soil flow is also narrower and more distinct
for smaller element size (Figure 3-3),
which will result in lower resistance.

295

Figure 3-2 Flow: element size 0.0015 and
0.0070
Figure 3-3 Shear bands, Element size 0.0035
and 0.0015
The mesh convergence test indicated that
the penetration speed affected how much
oscillation that occurs in the results. It is
also of interest to see if the bar factor
changes for increased speed, or if the
oscillations may be smoothed out (in order
to get the right value). This is important to
know for the spudcan penetration test, as
the deformations are much greater, and
increased speed may reduce the
calculation time significantly. The model
with element size 0.0015 was used for this
study, and the speed is constant for each
analysis. Several different speeds of
penetration were tested, and a selection of
two of them is shown in Figure 3-4.
296
Figure 3-4 Speed convergence
The result indicates that the resistance is
more influenced by the speed at small
deformations than large deformations. The
fast calculation seem to stabilize around
the right value as the deformation become
larger (>0.005 meters). And a T-bar factor
with only a minor error is possible to
obtain by filtering out the oscillations in the
resistance for the fast calculation, for
example by taking the average resistance
over a given displacement. This approach
may be used for the spudcan penetration
analysis, as the deformation is large.
However, it will lead to some errors if the
resistance changes a lot during the
penetration which may be the case around
the punch-through depth.
3.2. Spudcan penetration
H/B refferes to the height of the first layer
divided by the diameter of the spudcan,
while D is the penetration depth.
Single layer
The first analysis on spudcan penetration
was executed in order to verify the
method. The soil is weightless in this case,
so that none back-flow occurs. The
spudcan is modeled as a flat cylinder with
a rough base. The diameter of the cylinder
is 15 meters. All the penetrations are
preformed undrained, and poisons ratio of
0.495 is used. The cohesive yield strength
is 10kPa for all the soil. Only 45 degrees
of the problem is modeled, and symmetry
boundary conditions are used on the
sides. The height of the initially active
elements (soil) is 3.5D=52.5 meters, while
the radius is 2.5B=37.5 meters.
Approximately 228 000 elements were
used, and a penetration rate of 1m/s.
Figure 3-5 Mesh: single-layer penetration
The reaction force was measured, and
oscillations were filtered out. The bearing
capacity factor were calculated by N=q/cu.
The results were compared with Martin
and Randolph’s lower and upper bound
solutions (Martin, C.M. and Randolph,
M.F., 2001).
The calculation is in reasonable
agreement with Martin & Randolph’s
bearing capacity theory. The result shows
some error for shallow penetration, which
is likely due to the dynamic inertia effects,
or not small enough elements.
297
Figure 3-6 Bearing capacity factor for single-
layer clay
Spudca penetation in two-layered clay
First, a case with no strain-softening was
calculated, where the cohesive yield
strength in the upper layer is 100kPa, and
20 kPa in the lower layer. The Young’s
modulus is 500cu, and the effective soil
weight is 8.5kN/m3 for the upper layer and
7kN/m3 for the lower layer. The diameter
of the spudcan is 15 meters. The case
was found in Tjahyono’s doctor thesis
(Tjahyono, S. 2011).
The “decimateFilter” function in
Abaqus/Explicit was used to filter out the
oscillations in the test data. The results
are shown in Figure 3-7, where the
spudcan resistance is plotted against the
normalized penetration H/B. It is evident
that the oscillations are large in the
beginning of the penetration. However, the
result is not necessarily inaccurate. After
filtering out the oscillations and comparing
the resistance with Tjahyono’s result, we
get the result as shown in Figure 3-8. The
result seem to be fairly good, in the sense
that it does not deviate much from
Tjayhono’s FEM calculation, with the
exception of the dynamic oscillations at
the beginning, and at the end where
backflow initiates. The deviation is at
maximum around 25 kPa, for both the
beginning and end of the penetration. The
back-flow is initiated at around D/B=1.40,
and the flow pattern and material
boundaries are shown in Figure 3-9. The
right picture shows the upper material in
red and the lower material in blue. The
eulerian mesh seem to preform very well
in managing the boundaries, as we can
see the upper layer is pushed into the
lower layer.
Figure 3-9 Flow pattern and material
boundaries at initiation of back-flow
( .

The result from FEM calculation was also

compared with the theoretical solutions
from SNAME (2002), Hossain & Randolph
and Tjahyono (2009) in Figure 3-10. The
SNAME method overestimates the
potential for punch-through, and
underestimates the spudcan resistance.
The reason for the apparently bad
performance of SNAME method is
because of the whished-in-place
approach. This case involves a large
reduction in the cohesive yield strength
Figure 3-7 Response filter
from the upper to the lower layer, and
SNAME will consequently underestimate
the resistance and overestimate the
potential for punch-through as it does not
account for the geometric chances in the
soil. The much stronger upper layer will be
pushed into the lower layer, and this
deformation will increase the spudcan
resistance and reduce the potential for
punch-through. Hossain & Randolph’s
method performed better, but is still
underestimating the spudcan resistance if
we compare to the FEM calculation and
Tjahyono’s (2011) result. The FEM
calculation is close to Tjahyono’s theory
(two different results is presented, as it
was some uncertainties in the back-flow
depth). The resistance is dependent on
the element size for flow around an object
as shown in the T-bar test, and from the
Figure 3-8 Spudcan resistance non-soft
comparison with Tjahyono’s numerical and
material, compared with Tjahyono’s result
theoretical work is it possible to conclude
that the elements are too large in the
lower layer. However, the error was within
acceptable range, and further reduction in
element size would increase the
calculation time which was already
relatively long. The same mesh was
298
therefore used for the rest of
calculations.
Figure 3-10 Comparison of spudcan
resistance with hand calculations
Finally, strain-softening was implemented
in the spudcan penetration test. Only the
upper layer was subject to softening. The
sensitivity of the soil was 2, which means
that the softened soil have half the
strength of the original soil. The absolute
plastic strain for which the reduced
cohesive yield strength was established
varied from 2% to 20%. The upper bound
solution ( ) and lower bound
solution ( ) were also plotted in
the result.
 for the lower bound solution. Strain-
Figure 3-11 H/B=1. Spudcan resistance with
strain-softening behavior
the
The results are plotted in Figure 3-11. All
the calculations are pretty similar to the
lower bound solution. It is suspected that
this is due to that the shear bands are
established at relative shallow penetration
depth. The strain-softening behavior has
less effect when the shear bands are
established. This is due to the fact that the
strains in the shear bands are much
greater than the absolute plastic strain for
which the soil is softened (μ). The models
with μ=10% and μ=20% showed a spike in
the resistance for shallow penetration
depth ( ), but follow close to
the lower limit for the rest of the
penetration. The spike in resistance may
be because the shear bands have yet to
be established for the shallow penetration.
The depth of initiated backflow is similar
for all the models, except for the upper
bound solution. The depth is between
D/B=1.1 and D/B=1.21 for the softened
soil and lower bound solution, while it is
between D/B=1.37 and D/B=1.47 for the
upper bound solution. The depth of
initiated backflow is a bit inaccurately
determined because the output frames of
the deformation are a bit too scarce.
Strain-softening may increase the punch-
through danger. Indication of this can be
observed for the calculation with μ=20%,
the punch-through danger is greater than
softening is therefore an important
parameter in the material model. However,
these calculations are hard to interpret
because of the dependency of the
element size regarding the shear band
thickness. It would be interesting to know
how large μ (absolute plastic strain for the
softened condition) is needed to increase
the punch-through potential drastically. It
seems to be between 10-20% for this
analysis, but it is not possible to conclude
this, as a different mesh would probably
give a different result. But Tjahyono’s
result showed a similar result, and it was
concluded that μ < 5% could be calculated
using the lower bound solution (no strain-
299
softening), while this is not possible for μ > Nonlinear finite elements/Lagrangian and
5%. Eulerian descriptions. (2010,
September 2). Retrieved April 23, 2013,
4. CONCLUSIONS from Wikiversity:
The CEL method in Abaqus/Explicit has http://en.wikiversity.org/wiki/Nonlinear_f
proven to be suitable for spudcan inite_elements/Lagrangian_and_Euleria
penetration problems. The penetration n_descriptions
speed affected mainly how much Randolph, M.F. & Andersen, K.H. (2006).
oscillation that occurs in the results. The Numerical analysis of T-bar penetration
oscillation may be filtered out as the in soft clay. International journal of
resistance oscillated around a mean geomechanics(6), 411-420.
value. The computational cost for these Randolph, M.F. & Houlsby, G.T. (1984).
types of problems are large, and it is The limiting pressure on a circular pile
therefore of interest to find the highest loaded laterally in cohesive soil.
penetration rate for which oscillations may Geotechnique, 34(4), 613-623.
be filtered out. However, there were some Stewart, D.P. & Randolph, M.F. (1994). T-
difficulties regarding the effects from the bar penetration testing in soft clay.
element size, especially when trying to Journal of Geotechnical
include strain-softening behavior. It is Engineering(12), 2230-2235.
important to address this problem, as non-
Tjahyono, S. (2011). Experimental and
conservative results might be obtained.
numerical modelling of spudcan
ACKNOWLEDGEMENTS penetration in stiff clay overlying soft
clay. PhD Thesis, National university of
I would like to thank Dr. Hans Petter Singapore, Department of civil
Jostad for guidance trouhout the master engineering, Singapore.
thesis. I would also like to thank Dr. Khoa World Fleet of Jack-Up Drilling Rigs.
Huyan for learning me how to model (2012, October 10). Retrieved June 3,
spudcan penetration using the CEL 2013, from
method in Abaqus/Explicit. http://shipbuildinghistory.com/today/hig
hvalueships/offshorejackups.htm
REFERENCES
Abaqus 6.12 Analysis User’s Manual.
(n.d.). Eulerian Analysis, Chapter 14.1.
Abaqus/CAE User's Manual. (n.d.). The
volume fraction tool, Chapter 28.5.
Hossain, M.S. & Randolph, M.F. (2009).
New mechanism-based design
approach for spudcan foundations on
stiff-over-soft clay. Offshore Technology
Conference. Houston.
Liyanapathirana, D. M. (2008). Numerical
simulation of T-bar penetration in soft
clay. GeoCongress.
Martin, C.M. and Randolph, M.F. (2001).
Application of the lower and upper
bound theorems of plasticity to collapse
of circular foundations. Proc. 10th Int.
Conf. of IACMAG, Vol 2., pp. 1417-
1428. Tucson.

300
25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Session 6

Slope stability and landslides

302
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Stabilization of railway landslide with bored pile retaining wall

Marko BISCAN1*, Goran DIZDAR1*

1
Institut IGH d.d., Zagreb, CROATIA

ABSTRACT
On the RAILWAY ZAGREB-RIJEKA, instability (landslide) has been noticed from km 571+270 to
571+700 (line is in the cut), but since 1958 mentioned part of railway was never fully repaired. For
improvement was chosen 140 m long bored pile retaining wall, 10 m deep, with horizontally dug
drains. Plaxis 2D and GeoStudio-Slope/W software were used for simulation of the current
situation and for designing stabilization elements. The solution is implemented and there are
several measurements of pile retaining wall displacements that will be presented.
Piles has diameter of ĭ 88 cm on the axial spacing of 140 cm. Piles were performed by drilling
using casing. The length of the piles are 6 m and 10 m. Carrying beam of pile retaining wall has
cross section B x H = 100 x 60 cm. Dug drains are carried out with concrete drainage pipes I 20
cm and set at the appropriate depth in the trenches. In total, 10 drains were performed, with length
of 12 m dug maximum 3 m from ground surface, and are connected in the existing concrete
channels along the railway line.

Keywords: landslide, bored piles, retaining wall

railway track, damaging drainage channel

1. INTRODUCTION
The Zagreb-Rijeka railway, Moravice -
Skrad section (built 140 years ago) at the
Lokvice location passes through cut up to
15 m high, approximately from km
571+270 to km 571+700 of railway line.
According to the existing documents, the
stability of the cut and the slopes above
were considered in 1958. The first solution
was a concrete trench and stone facing.
Over time, cracking and deformation of the
concrete trench and stone facing were
noticed. Also, on the slope below the right
side of the cut, was still active an old
landslide that was at the creeping stage.
Landslide constantly causes lifting of the
* presenting author
along the track and stone facing in the cut.
Because of slope instability, in 1992,
geotechnical investigation and conceptual
design were conducted. Later, during
1995, detail design was finished.
According to detail design, slope stability
was solved with retaining wall along the
existing concrete trench on the right side
of the railway track.
During 2012 was decided to rehabilitate
the landslide in cut “Lokvice”. Because of
that, engineering geological mapping was
performed. All geotechnical data obtained
by previous studies was used to create
complete picture of stability and geological
conditions.

Figure 1. Previous solution – drainage
channel and stone facing
2. PREVIOUS STUDIES
Written records about stability problems of
Lokvice cut goes back in 1958. During
mentioned year geological map of the
wider area (Mariü and Crnkoviü, 1958)
was made. In that map material, softening
and sliding on the location was mentioned.
In the same year, another author (Sarajliü,
1958) has writing about the problems of
rehabilitation of Lokvice cut. Also, location
was mentioned as an example, where any
of technical intervention did not solve
problem completely. After each
intervention (many of them were
performed) there was still caution for slow
driving through section of Lokvice cut.
Authors (Mariü and Crnkoviü, 1968), after
field investigation, found that with
performance of the longitudinal drainage
channel, partly solved drainage system,
although they proposed building lateral
drains connected to the longitudinal
channel. In 1992, for the solution of
landslide reconstruction, detailed
geotechnical investigation was performed
and preparation of conceptual design of
slope mitigation was made. Later on, in
1995 detailed design was made for the
landslides mitigation proposing reinforced
concrete retaining wall.
3. GEOTECHNICAL SOIL
CHARACTERISTICS
To determine the geotechnical soil
characteristics, all previously performed
geotechnical investigation were analysed.
304
Investigation was mainly based on
boreholes with average depth of 8-10 m.
In boreholes, disturbed and undisturbed
samples were extracted, after which
samples were tested in laboratory. In each
layer SPT tests was also performed.
Figure 2. Sliding body contours and borehole
locations
According to the performed testing
following can be conclude; natural
moisture content of the samples is
between 9.2 to 21.1%. Liquid limit range is
between 25.5 to 53.2% but most of the
samples have medium plasticity (35-50%).
Plastic limit range is between 12.1 to
16.5% (with two exceptions 20.5 and
34.7%) and calculated plasticity index is
between 13.4 and 33.5%, which together
with liquid limit classified material as a clay
with mainly medium plasticity.
Geotechnical soil model is:
x To a depth of 3-5 meters appears clay
layer (low and high plasticity, medium
to hard consistency). Clays contain
sharp-edged fragments of Triassic and
Palaeozoic rocks and rare dolomite
blocks. Tested soil samples have
cohesion c = 6-30 kN/m2 and friction
angle ij = 13.6-34.3 °.
x Below the clay layer, clay schist is
present, which is in the upper part of
layer weathered. With increasing
depth, shear strength of clay schist is
also increasing. Weathered
sandstones and siltstones are also
appearing with significantly higher
parameters of shear resistance than
schist.
It was concluded, that sliding is occurring
up to a depth of weathered clay shale
appearance. The level of ground water
according to the measurements in
boreholes is near the ground surface. It
can be approximated for stability analysis
at 1 m from the ground surface.
4. DESCRIPTION OF LANDSLIDE
With detailed engineering and geological
terrain mapping, all visible damage to
stone facing is registered along the tracks
as well as reinforced concrete channel in
the slope bottom. The basic contour of the
sliding body in the left and right sides and
the top is assumed, and nowhere is clearly
visible due to a small displacement in the
bottom (a maximum of about 1 m
displacement of reinforced concrete
channel in the bottom of the slope).
According to the established lithological
sequence and geological characteristics, it
is logical that the lower contour of the
possible formation of sliding surface is
placed on the bottom of the clay layer. The
impact of groundwater can be
approximated with line of groundwater
level at 1 m from the ground surface.
Figure 3. Sliding body with implemented
solution
5. LANDSLIDE STABILIZATION
METHOD
Implementation of pile retaining wall and
dug drains is planned in the instable slope
of the railway line from km 517+498,914 to
km 517+637,673 section; length of the
section is 139 m. The pile retaining wall is
planned to be built over the stone facing
on the right side of the cut. Dug drains are
planning to be built on both sides of the
cut. The total length of all drains is 120 m.
Pile wall is retaining structure which takes
lateral loads and permanently prevents
excessive horizontal displacements or the
possible instability of the existing slope.
305
Dug drains are placed perpendicular to
the pile retaining wall, to lower ground
water level in the slope. Surface waters
are gathered with concrete gutter near the
pile retaining wall.
Console pile retaining wall structure
consists:
x Piles with nominal diameter of ĭ 88 cm
on the axial spacing of 140 cm. Piles
are performed by bored drilling using
casing. Concrete piles, C 30/37,
reinforcement bars B 500B. The
lengths of the piles are 6 and 10 m.
x Piles were capped by a concrete beam
(dimensions B x H = 100 x 60 cm),
concrete strength class C 30/37,
reinforcement bars B 500B.
Dug drains are build out from a
concrete drainage pipe (ĭ 20 cm) and set
in the trenches at the appropriate depth. In
total, 10 drains will be provided, length 12
m. Depth of a dug drains is maximum up
to 3 m at the beginning (the highest point),
and the outlet is in the existing concrete
channels along the railway line.
Figure 4. Implemented pile retaining wall
6. GEOTECHNICAL CALCULATIONS
6.1. INPUT DATA AND CALCULATION
DESCRIPTION
Calculations were performed for the slope
stability of the existing instable slope and
for the improved slope with pile retaining
wall and drainage system. For the
calculation was used computer program
"GeoStudio-Slope/W", in which the sliding
body is divided into segments and is
calculated the ratio of shear resistance in
the ground and actions on the sliding
mass that is displayed as a safety factor.
Besides that, computer program "Plaxis
2D" was used for finite element analysis.
Using "Plaxis 2D" additionally, it was
calculated safety factor of existing The calculation model in software
landslides and improved slope with "Plaxis 2D" for the existing instable
obtaining stress-deformation image of a slope
computational model. The resulting factor of safety - 1.05, which
Static system of retaining structure represents the limit state of equilibrium:
consist console pile retaining wall with
piles (ĭ 88 cm) on 140 cm axial spacing
detained with passive resistance on the
buried part of the structure.
With "Plaxis 2D" calculation, next
values were obtained: value of stress and
displacements in the soil, and also
displacements, bending moments,
transverse and longitudinal forces in piles,
activated passive resistance to the buried
part of the pile and factor of global stability
with ij-c reduction.
Figure 6. Plaxis 2D calculation result
The soil is modelled using nonlinear
hardening soil model.
Soil parameters used in calculation: The calculation model in software
x Up to the depth of 3-5 meters is present "GeoStudio-Slope/W" for the improved
clay layer, low and high plasticity, slope:
medium to hard consistency, cohesion The calculation is conducted according to
c = 8 kN/m2 and friction angle ij = 17°, EC7, part 1, Design Approach 3.
Eoed = 4000 kN/m2.
x Below that, clay schists is present -
hard consistency Title : Stabilnost klizista_UKM _571+560
1.264

o Weathered Schist - cohesion c=20 Name : Osnovna sti je na

1.400

Model: Mohr-Coulom b
Uni t Weig ht: 20 kN/m ³
1.600

kN/m2 and friction angle ij=17°, Eoed

Cohesion : 16 kPa
Phi : 1 3.74 °
1.800

Name : Pokrov/Gl ina

540 Name: Pilo tska stij ena Model: Mohr-Coulom b
Model: M ohr-Coulomb Uni t Weig ht: 20 kN/m ³

= 8000 kN/m2.
Uni t Weight: 25 kN/m ³ Cohesion : 6.4 kP a
Cohesion: 2800 kPa Phi : 1 3.74 °
Name: Betonska zastita
Phi : 24 .79 °
M odel: Mo hr-Coulomb
Uni t Wei ght: 25 kN/m ³

o Schist - cohesion c=100 kN/m2 and

Cohesion: 2800 kPa
530 Phi : 24.7 9 °

friction angle ij=20°, Eoed = 20000 520

kN/m2.
6.2. RESULTS OF THE CALCULATION 510

500
The calculation model in software 0 10 20 30 40 50 60 70 80

"GeoStudio-Slope/W" for the existing Figure 7. Slope/W calculation result

instable slope
The resulting factor of safety - 1.0, which The calculation model in software
represents the limit state of equilibrium: "Plaxis 2D" for the improved slope:
Version 2010.1.0.6019

Nam e: Osnovna stij ena

M odel: M ohr-Coulom b 10,00 20,00 30,00 40,00 50,00 60,00 70,00 80,00
Unit Wei ght: 20 kN/m ³
Cohesion: 20 kPa
550,00
Phi: 17 ° 1.000
Nam e: Pokrov/Gl ina
540 M odel: M ohr-Coulom b
Unit Weight: 20 kN/m ³
Cohesion: 8 kPa
Phi: 17 °
Nam e: Betonska zastita
540,00
M odel : Mohr-Coulom b
Uni t Weight: 25 kN/m ³
Cohesion: 3500 kPa
530 Phi : 30 °

530,00

520
520,00

510 510,00

500,00
500
0 10 20 30 40 50 60 70 80

Figure 8. Plaxis 2D calculation model

Figure 5. Slope/W calculation result
306
Maximum horizontal displacements of pile
retaining wall are up to 1 cm:

Figure 10. Photo of renewed stone facing with

pile retaining wall

Figure 9. Total displacement of pile

7. CONSTRUCTION OF PILE
RETAINING WALL
Reconstruction was made during 2014
and 2015.

Figure 11. Photo of performed pile retaining

wall

8. DISPLACEMENT MONITORING
Because of the possible landslide activity,
the project planned the monitoring
program development for retaining
structure. The purpose of monitoring is to
confirm design assumptions and the
possibility of right time intervention in case
of larger displacement than anticipated.

Figure 10. Photo of performed piles during

drainage installation Figure 12. Display of inclinometer
measurements
307
 Observation included horizontal Mihalinec, Z. (1995) Detailed design,
displacement monitoring of pile retaining Landslide “Lokvice”, Zagreb, Croatia.
wall, using the built-in inclinometer tube. Eurocode 7: Geotechnical design – Part 1:
Inclinometer tube is built into the pile. The General rules (EN 1997-
measurement shows a peak horizontal 1:2004+AC:2009)
displacement of pile for about 2 mm in the
period since the end of reconstruction to
today.

9. CONCLUSIONS
The slope stability problem of Lokvice cut
appears since 1958. Over time were given
temporary solutions which proved to be
more expensive solution then solving the
problem in the end. By analysing previous
studies and situation on the site, it was
decided to propose a permanent solution.
For permanent solution was chosen a pile
retaining wall, resulted in minimal
measured displacement that
approximately coincides with the design
solution. The design approach was based
on calculation software: GeoSudio-
Slope/W (slope global stability) and
PLAXIS 2D (stress and strain analysis).
First, existing situation on the site
(instability) was modelled. Slope/W and
Plaxis 2D software showed the same
results regarding the global stability
(Global Safety Factor – 1,0). Further,
modelling in Slope/W and PLAXIS 2D
defined were pile retaining wall elements
and further monitoring program. According
to all above, it is concluded that it is very
important to bring the right decision about
how to mitigate the landslide, since in this
case the previous simple solutions have
proved insufficient to prevent further
displacement development.

ACKNOWLEDGEMENTS
The authors are grateful to colleagues in
geotechnical design department of
Institute IGH for the assistance and
gathering of data from older projects.

REFERENCES
Mihalinec, Z. (1992) Conceptual design of
remediation and geotechnical
investigation report, Landslide “Lokvice”
km 571+5/7, Zagreb, Croatia.
308
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Methodology for preparation of thematic geotechnical maps for

urbanization purposes

Natasha Nedelkovska1
1
Geohydroconsulting, Skopje, R.MACEDONIA

ABSTRACT
This paper treats very popular and complex engineering field related to the definition of
geotechnical conditions for urbanization purposes. Main goal is to present an approach for
preparation of thematic geotechnical maps, which should serve as basis for planining activities. In
order to prepare such maps, apropriate zoning methodology is proposed, where, the terrain
suitability for urbanization depends on following basic factors: engineering-geological properties of
the present materials, slope angle, groundwater level, seismicity and excavation conditions.
According to the proposed methodology, ratings are assigned to the selected factors, depending
on their importance for successful urbanization. Based on the assigned ratings, rating map for
each factor is prepared, and then the final map is created, representing the sum of influences of
each analyzed factor on the urbanization suitability. On all prepared maps, four terrain categories
according the suitability for construction - urbanization, are separated. The zonation methodology
is practically applied for the territory of City of Skopje. All mentioned maps are prepared with
contemporary GIS tools.

Keywords: urbanization, zoning, suitability, maps, Skopje, GIS-technology

Depending on the location of each

1. INTRODUCTION
With the long development of civil
engineering, it became obvious that for
rational and successful urban planning,
design and construction of structures, an
excellent understanding of the ground
conditions is essential. Different factors
govern the behavior of the natural rock
masses during the construction and
exploitation phase and these should be all
well understood in order to have
successful projects.
* presenting author
particular structure, the influence of these
factors can have different meaning. As
most important in most cases we can
consider the morphological, geological,
seismic, hydrological, hydrogeological and
geotechnical factors. Complete
understanding of these factors, will enable
definition and allocation of the different
geo-hazards. In many instances, the
combination of the geo-hazards and the
engineering activities has been reason for
enormous socio-economic damages, and
unfortunately even human losses.
In order to prevent and avoid these socio-
economic losses, worldwide practice and
trend is the preparation of appropiriate
thematic maps that serve as basis for
urbanization purposes.
The preparation of these maps is
according appropriate methodologies,
based on detailed analysis of available
literature, specific principles for site
zoning, and right selection of various
qualitative and quantitative parameters.
Terrain zoning on these maps is
presented, from which urbanization
suitability categories can be recognized.
The use of such maps before the start of
the processes of planing and construction,
helps engineers to avoid terrains with
natural unfavorable geological – geotech-
nical conditions, leading to more efficient
construction and safer structures.
Methodology for preparation of
geotechnical maps as bases for
urbanization purposes is presented in this
paper, applied for the territory of Skopje,
the capital city of R. Macedonia. The maps
are prepared with GIS technology, more
exactly using the software ArcGIS. In the
present state of art, using GIS is the most
Figure 1. Geographical position of the analysed area: Skopje, Republic of Macedonia
(after MapQuest maps)
310
suitable approach for preparation of such
technical documentation.
2. BASIC INFORMATION FOR CITY
OF SKOPJE
Skopje is located in the north of the
Republic of Macedonia, in the center of
the Balkan Peninsula (Figure 1). The city
is built in Skopje valley, which is oriented
on a west-east axis, along the course of
Vardar River which flows into Aegean Sea
in Greece.
The valley of Skopje represents a
depression surrounded by mountainous
and hilly formations, built of rocks of
different age, starting from Precambrian
up to Paleogene, while the valley itself is
filled with Neogene-Quaternary and recent
sediments (alluvium, proluvium, soil
debris, Pliocene sediments).
From tectonic point of view, Skopje valley
is a mosaic of differently uplifted and
down-thrown blocks, separated by faults.
This kind of tectonic setting of the valley
exerts its high seismicity in the central
area as well as in the peripheral zones.
3. METHODOLOGY FOR The details for ratings for each factor are
URBANIZATION SUITABILITY ZONING given in Tables 2 to 6. For each of these
factors four groups of possible values are
The process of defining a methodology for
defined, related to four terrain categories:
urbanization suitability zoning requires an
understanding of all factors affecting the 1. Optimally favourable terrain
urbanization of a given area. The first step 2. Favourable terrain
is selection of factors. Then, for each 3. Conditionally favourable terrain
factor a rating system is being defined, so 4. Unfavourable terrain
that each factor has several classes. The
Table 2. Range of values and ratings for
analyzed factors are related to the corres-
Slope Angle (SA-R)
ponding ratings using polynomial inter-
polation method. The main idea in Slope angle (°) Rating
devoloping this method is to find a way to
1–5 0
establish analythical correlations between
any value of factor and its rating. < 1 and 5 – 10 0.3
3.1. Selection of factors the 10 – 20 1
urbanization suitability
> 20 2
Urbanization suitability of a given area
may depend on many factors, related to Table 3. Range of values and ratings for
the morphological, geological, Ground Water Level (GWL-R)
hydrogeological and geotechnical GWL (m) Rating
properties of the terrain. Here as most
important ones are considered: the > 12 0
Lithological type (LT), Slope Angle (SA), 5 – 12 0.5
Ground Water Level (GWL), Seismic
Intensity (SI) and Excavation Conditions 1.5 – 5 1.5
(EC). < 1.5 2
For each of these factors, a maximal Table 4. Range of values and ratings for
rating has been assigned, depending on maximal Seismic Intensity (SI-R)
their influence on urbanization suitability
(Table 1). In the proposed methodology Seismic intensity (°) Rating
for zonation, all of the factors have same
<5 0.3
value for the maximal rating, i.e. the
author considers that all factors included 5–6 0.9
are equally important when determining 7–8 1.4
the urbanization suitability of the terrain.
Table 1. Maximal values of the ratings >9 2
for the factors Table 5. Range of values and ratings for
Classification Excavation Conditions (ȿɋ-R)
Maximal rating
parameter ERMR
Rating
LT 2 (excavation method)

SA 2 4 – 25
0.3
(easy excavator digging)
GWL 2
26 – 40
0.8
SI 2 (hard excavator digging)
EC 2 41 – 60
1.4
(ripping)
Total (TUSR) 10
61 – 100
2
(blasting)
311
 Table 6. Range of values and ratings for
Lithological Type (LT-R)
Lithological composition defined
Rating
with Lithological Type
1 0
2 0.4
3 1
4 2
Analyzing the data from Table 6, it can be
seen that Lithological composition of the
bedrock is defined with adequate
Lithological Type (LT). Having in mind that
a good part of the rock masses are
characterized with similar physical-
mechanical properties although their
lithological composition and age are
different, engineering geological grouping
of the rock masses is done. Arbitrary
values were used from 1 to 4 as a basis
for correlation with ratings for Lithological
Types. So, the following types are
differentiated:
1. Group of hard magmatic, metamorphic
or sedimentary rocks as granites,
marbles, massive limestone and others,
with range of values for internal friction
angle M>45° and cohesion c>100 kPa.
In this group also belong the conso-
lidated rocks: soft to semi hard rocks
(sandstones, calcare-ous marls, schist
with favorable dip of foliation, with
range of values for internal friction
angle M=36–45° and cohe-sion c=50–
100 kPa.
2. Group of rocks with a relatively low
degree of lithification: marly clays,
poorly cemented sandstones, marls,
argillaceous shale, weathered schists,
with range of values for internal friction
angle M=26–35° and cohesion c=30–50
kPa.
3. Group of rocks with a low degree of
lithification; soft rocks to hard soils as
hard clays, compacted sands, claylike
gravels, with range of values for internal
friction angle M=21–25° and cohesion
c=15–30 kPa.
312
4. Loose rocks with low shear strength
and high deformability, and loose
detritial rocks or rocks with a reduced
degree of lithification (clays and silts,
sandy clays, sandy silts), with range of
values for angle of friction M=10–20°
and cohesion c=0–15 kPa.
Groundwater level is considered as
zonation factor because of the possible
water flows in the construction pits, the
aggressiveness of the groundwater,
conditions for suffosion development,
bearing capacity reduction etc. So, the
most unfavorable case is when the aquifer
zone is shallow below the terrain surface
and then serious problems occur with
dewatering of the construction pit, capillary
effects on the footings and the
construction, liquefaction development
and so on.
The dip of the terrain is dictated by its
morphology, which is important because
the dimensions of the excavation and the
stability of the terrain depend on the dip
and height of the slope. That means that,
terrains with 1 – 5% dip require small
volume of earth works i.e. low cuttings, cut
and fillings and embankments. That dip is
suitable for easy dewatering of the
atmosphere water and sewage systems
construction as well. Furthermore,
dewatering of the terrains with dip lower
than 1% is quite difficult. On the other
hand the excavation is problematic when
the dip of the terrain is steep and also that
terrain is susceptible to instabilities.
The degree of the maximal seismic
intensity is very crucial factor considering
that our country is seismically active area.
Before the Skopje earthquake in 1963,
poor attention has been paid for aseismic
design. The earthquake pointed out that
consideration of the expected seismic
intensity is of great importance in
designing and construction of structures in
seismically active areas as Skopje.
Terrain categorization according to the
factor Excavation Conditions is expressed
through the excavation categories defined
in the ERMR system (ERMR - Excavation
Rock Mass Rating by M. Jovanovski
2001). This factor is important in planning
and designing of structures, considering
that the cost of the excavation per m3
depends on the applied excavation
method (digging, ripping and blasting).
All factors can be mapped, calculated,
measured or assumed using different
direct or indirect investigation methods.
The possible combinations of the factors’
ratings define the suitability for
urbanization of a given area.
3.2. Application of polynomial
interpolation method
The polynomial interpolation method, used
for solving many problems in geotechnics,
here is applied for Urbanization Suitability
Zonation.
The method is applied using the data from
Table 2 to Table 5. For all evaluation
factors, correlative curves are defined,
with main goal to have possibility to assign
an adequate rating for all parameters.
Figure 2. Correlation between ratings for Slope Angle, Ground Water Level, Seismic Intensity and
Excavation Conditions with factor values using polynomial interpolation
313
The established correlations between
values for evaluation factors and ratings
are presented with the following
equations:
SA-R=0.0017SA2+0.0145SA–0.005
GWL-R=0.0101GWL2–0.3073GWL+2.3114
SI-R=0.0131SI2+0.0701SI+0.0077
EC-R=–9E-5ERMR2+0.0286ERMR+0.0117
Where: SA-R, GWL-R, SI-R ɢ EC-R are
adequate calculated ratings for any value
of individual factors.
Graphical presentation of the defined poly-
noms in a form of interpolation charts is
given in Figure 2.
In the presented equation the
determination coefficient for all cases has
very high values (R2 = 0.9995 to R2 =
0.9998) which refers to very strong
connection between analyzed parameters.
3.3. Definition of terrain’s categories
according to urbanization suitability
In order to obtain the final map, so called
urbanization suitability map, a sum of the
ratings from each factor is required. The
sum of the ratings represents total rating,
TUSR (Total Urbanization Suitability
Rating), based on which the zonation is
conducted.
In the next step, 4 (four) terrain’s
categories, according to the suitability for Figure 3. Correlation between USC and TUSR
urbanization – construction were defined,
The correlative equation is:
presented in Table 7 together with the
appropriate total ratings. USC=0.0062TUSR2+0.2396TUSR+0.9939
Table 7. Terrain’s suitability categories with
appropriate TUSR 4. THEMATIC MAPS AND FINAL
URBANIZATION SUITABILITY MAP
Suitability Category ɌUSR
According to the previously defined
Optimally favourable terrain 0–3 methodology for zonation of the terrain
Favourable terrain 3–5 from an aspect of urbanization suitability,
five thematic maps, for each factor, were
Conditionally favourable terrain 5–7 prepared. All of these maps were
Unfavourable terrain 7 – 10 prepared with GIS technique, which
means ArcGIS software was used.
The defined interpolation chart and The results obtained in the thematic maps
correlation for analytical connection for each analyzed factor are presented in
between the Urbanization Suitability Table 8, as a statistical output.
Category (USC) and the Total
Urbanization Suitability Rating (TUSR) are
presented in Figure 3.

Table 8. Statistic data from the thematic maps for each factor

Suitability categories in percent per each factor in the analysed area (%)
Total
Factor
Optimally Favourable Conditionally Unfavourable (%)
favourable terrain terrain favourable terrain terrain

LT 11.66 17.66 45.27 25.42 100

SA 24.36 56.27 11.79 7.58 100
GWL 35.10 56.43 8.47 0.00 100
SI 0.00 0.00 4.31 55.69 100
EC 57.24 31.55 7.55 3.66 100

With intersection of these thematic maps In Table 9 a statistical overview of the data
in ArcGIS, the final map was obtained for obtained in the map is given. Exactly, the
urbanization suitability of Skopje. presence of each suitability category in
percentage is shown.

314
 Table 9. Percentage presence of each Furthermore, in Figure 3 the obtained
suitability category in the analyzed area urbanization suitability map for the territory
of Skopje is presented.
Suitability category Percent (%)

Optimally favourable terrain 10.24

Favourable terrain 86.34
Conditionally favourable terrain 3.39
Unfavourable terrain 0.03
Total: 100

Figure 4. Urbanization suitability map for the territory of Skopje

geohazard, which means that the safety of

5. CONCLUSIONS structures in such zones can become
Morphological, geological and questionable over time. Therefore, with
geotechnical factors throughout an area- the presented approach and its further
region proposed for new development of development, some type of standardized
further urbanization of existing cities have method for urbanization purposes can be
great impact on the civil engineering. It is established. If applied in fight time, such
important to understand their nature in all method can present strong tool,
phases, starting from the design, contributing not only for the improvement
construction, and exploitation of the of civil engineering, but the society in
structures. Furthermore, their general.
unfavourable combination can make some
terrains very susceptible to some type of
315
REFERENCES

Devi, S., S., & Stalin, V., K. (2011),

“Development of soil suitability map for
geotechnical applications using GIS
approach”, Proceedings of Indian
Geotechnical Conference, Paper No.
M-253.
Djuric, U., et all. (2013), “Land-use
suitability analysis of Belgrade city
suburbs using machine learning
algorithm”, GIS Ostrava.
Dragicevic, et al. (2011) “Natural Hazard
Assessment for Land-use Planning in
Serbia”, Int. J. Environ. Res., 5(2): 371-
380.
Effat, A., H. & Hegazy, N., M. (2013) “A
Multidisciplinary Approach to Mapping
Potential Urban Development Zones in
Sinai Peninsula, Egypt Using Remote
Sensing and GIS”, Journal of
Geographic Information System, 567-
583.
Jovanovski, M. (2012) “Engineering
geology” Faculty of Civil Engineering in
Skopje.
Malczewski, J. (2004) “GIS-based land-
use suitability analysis: a critical
overview”, Progress in Planning 62, 3-
65.
MapQuest maps available from:
http://www.mapquest.com/ [Access on
14.04.2016]
McGill, T., J. (1964) “Growing importance
of urban geology”, Geological survey
circular 487, Washington.
Meisina, C., (2006) “Engineering
geological mapping for urban areas of
the Oltrepo Pavese plain (Northern
Italy)”, IAEG2006 No. 188.
Peshevski, I. (2015) “Landslide
susceptibility modeling using GIS
technology”, PhD thesis, Faculty of Civil
Engineering in Skopje.
Tudes. S. “Correlation Between Geology,
Earthquake and Urban Planning”,
University of Gazi, Faculty of
Architecture, Department of Urban and
Regional Planning, Maltepe, Ankara,
Turkey.
316
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Experimental and numerical models in landslide behaviour

Ferran PARERA1*
1
Universitat Politècnica de Catalunya, Barcelona, SPAIN

ABSTRACT
The contributing factors and mechanisms involved in landslides can be addressed from different
perspectives, which include real cases, testing scale models and numerical modelling. The work
presented in this paper focuses on the last two mentioned ways. Instabilities observed in scaled
slopes will be described and analysed. Tests were performed in a transparent tank in which
boundary conditions in terms of stresses, displacements, water flow and pore water pressure were
controlled. The landslide motion was recorded with a digital camera and the images were
processed using the Particle Image Velocimetry (PIV) technique. The experiments performed,
supplemented by laboratory tests which characterize the soil behaviour, are analysed to evaluate
the landslide failure development and the post-failure behaviour.
The small-scaled experiments are modelled using the material Point Method (MPM). This
numerical technique combines a discreet system of material points, representing the moving
continuum and a fixed computational mesh. Because of this duality, the method is able to simulate
automatically large displacements without mesh tangling and, therefore, it is especially useful for
landslide modelling.
The aim of this work is to advance in the knowledge of the behaviour of landslides and their
interactions with protecting structures. Combining the analysis of experimental scale models and
the numerical modelling is the chosen way to understand the relevant mechanisms and controlling
factors.
In this paper, the tank and its features to control boundary conditions are described. An experiment
in which the instability of a dry sandy landslide is induced is later presented and simulated by using
MPM. Numerical and experimental results, derived from PIV, are compared with the aim of
validating the numerical technique.

Keywords: Landslides, Material Point Method, Particle Image Velocimetry.

physical (1g) models and numerical

1. INTRODUCTION
simulations using and developing
This work focuses on the study of numerical tools that allow interpreting the
landslides from the initiation of the motion phenomena observed in the laboratory.
to the post-failure behaviour. It involves
large displacement in saturated and
Because of the uncertainties associated
unsaturated soils. Two complementary
with real cases and the difficulties in taking
lines of research are followed: reduced
* presenting author
field measurements, physical scaled
models have been performed to
investigate the behaviour of landslides
(Montrasio et al. 2015). The main
advantage of scale models is the
possibility of controlling the boundary
conditions of the experiment as well as the
properties of the materials involved. These
facts simplify the numerical modelling and
offer the possibility of defining interesting
benchmark exercise to validate numerical
codes. The main drawback of physical
models lies in the scale effects. However,
if the observed response is well
understood at the reduced scale and the
interpretation of measurements is based
on well established concepts, the
conclusions of the study can be
extrapolated to larger scales.
Scaled experiments performed were
recorded by means of a digital camera
which provides correlative images
describing the motion. The images are
later processed by means of a Particle
Image Velocimetry (PIV) technique with
the aim of knowing the displacements
occurred between two consecutive images
taken in a given time interval. PIV was
firstly developed by Adrian (1991) to
evaluate the movement in fluids. Later,
PIV was implemented in the field of
geotechnics as a non-invasive tool for
measuring displacement in the laboratory
(White et al., 2003; Take and Bolton,
2004). PIV is a useful complementary
technique to traditional laboratory
instruments. In particular, PIV is a
powerful technique to analyse the soil
behaviour and the mechanisms that occur
in the whole domain of the landslide
(Baba, 2002).
New methods of numerical modelling are
currently being developed for the
interpretation and prediction of
phenomena that involve large
displacements. In this work, the Material
Point Method (MPM) is selected as a
convenient tool to simulate the
deformation processes involved in
318
landslides. The MPM discretizes the
continuous medium in a set of lagrangian
material points. Each of these points
contains properties of the subdomain that
represents. The material points move and
transport the properties information
through a fixed computational grid. The
governing equations are solved in the
nodes of the computation mesh. This
duality makes this method particularly
interesting for the study of landslides
because it is able to analyse the initial
static conditions, such as the initiation of
the failure mechanism, and post-failure
behaviour.
The work presented here is part of an
ongoing research project developed in the
frame of a PhD thesis carried out in the
Civil Engineering and Enviormental
Departement of Universitat Politècnica de
Catalunya (Spain).
2. METHODOLOGY
2.1. Scaled experiments
The small-scaled models are performed in
a tank specially designed to reproduce
different kinds of slope failures and
boundary conditions including
displacement constraints, water flow, pore
water pressures and relative humidity in
the case of unsaturated soil states
The tank is made of transparent
methacrylate and glass. Its dimensions
are: 1000mm long, 200mm width and
412mm height. These dimensions allow
reproducing scaled landslides in an
essentially 2D plane strain configuration
(Figure 1).
Water can be injected from the bottom and
the lateral part of the tank. In addition,
rainfall can be simulated from the upper
part.
The initial slope angle can be defined by
the glass guillotine which can be moved
manually (Figure 1). Landslide instabilities
can be induced by removing the guillotine
or by wetting, imposing water flow or
changes in relative humidity.

Figure 1. Tank showing the guillotine and
water injection system to perform the
scaled landslides.
A test programme was defined to analyse
the landslide response of sand under
different atmospheres equilibrated at
different values of relative humidity. The
effect of saturation condition on the stress-
strain-strength behaviour of sands has
been extensively discussed in the
literature (Fraysse et al. 1999; Chen et al
2013; Fern et al. 2015). In general,
increments in suction within the range
investigated in this work induce
increments in strength and stiffness. The
work presented here will focus on the
effect of non-saturation condition in fine
uniform sand, and in particular, on its
effect on the onset of landslide instability
and the subsequent motion.
Analysis of the results by PIV
The full process of the scaled landslide,
from failure initiation to a new equilibrated
state, is recorded in HD at 50 frames per
second. The frames are extracted and
correlated using the Particle Image
Velocimetry technique.
The technique involves comparison of two
digital images of the same element
captured in an interval time (ǻt). The
changes between the first image and the
second one are analyzed to obtain a set of
relative displacement vectors. In this work,
the free-commercial software PIVLab
(Thielicke et al. 2014) developed at the
Groningen University (Netherlands) was
selected. The images are processed and
encoded in patterns to be compared
319
(Figure 2). The area of interest of the
image t1 is divided in a mesh of
interrogation sub-areas. The code
analyzes geometrical and colour patterns
of each sub-area of the image t1, and
searches its corresponding counterpart in
the image t2. The search-area extends a
distance Smax beyond the area of initial
interrogation. This procedure is performed
for all sub-areas of interrogation image t1.
This technique can be applied to series of
images to analyze the evolution of the
displacement vector field during a time
interval. Notice that this methodology
provides displacements vectors computed
from two correlated images in points fixed
in the space instead of giving the motion
of a given point, which is the outcome of a
numerical code based on finite element
techniques. In order to compare the
results obtained from PIV with numerical
results an auxiliary code was written to
analyze the sequence of displacement
vector fields and to calculate the
accumulated displacement of materials
points and the associated volumetric and
shear strains.
Figure 2. Ideal example of a codification
in patterns in the PIV technique. Left:
digital picture; center: simplificated
gray scaled picture; right: binary code.
2.2. Numerical model
MPM methodology
The MPM method, originally called
Particle-In-Cell method, was developed by
Harlow et al. (1964) to be used in fluid
dynamics. Later Sulsky and Schreyer
extended the method to be used in soil
mechanics (Sulsky et al., 1994; Sulsky et
al., 1995).
In standard MPM, two levels of
discretization of the media are defined: Step 1. From the mass information
contained in the Material Points (mp), the
- Materials Points: The continuum is nodal lumped mass matrix at time tk (M) is
discretized by means of material points calculated using the shape function matrix
that represent a portion of domain, called (Np),
subdomain. The mass of the sub-domain (1.1)
is assumed to be concentrated in the
material point and it remains constant
during calculation to ensure solid mass
conservation. Besides the mass, the
material point contains also information of
variables that change in each time step:
The internal forces (Fint) are evaluated in
position, velocity, strain and stress.
the nodes,
Materials points move attached with the
solid skeleton and provide the lagrangian (1.2)
description of the media.

- Computational mesh: This grid is

similar to the mesh used in the Finite
Element that is generated throughout the where Bp is the nodal matrix particularized
whole domain, including the space for a material point which includes the
invaded by materials points during spatial derivatives of the nodal shape
calculation even when initially are empty. functions, Vp is the volume associated with
The governing equations are solved in the each material point, and ıp is the material
grid nodes. The variables required for point stress tensor.
calculation and the results obtained after
solving equations are transferred from The external forces (Fext) are evaluated on
nodes to materials points using mapping nodes following equation (1.3),
functions. After each time step calculation,
(1.3)
all data is stored in the materials points
and the data associated with the mesh is
discarded. As a consequence, the grid
remains constant throughout the
calculation.

The momentum balance equation is

solved and nodal accelerations ( are
Figure 3. Squeme of the computational determined (1.4)
sequence in MPM. Material points (1.4)
(brown), grid nodes (red).

Computational cycle: The algorithm is Step 2. The velocity at the material points
based on the work presented by Slusky et is updated.
al. (1995). A calculation cycle is divided in
the following steps: (Figure 3) (i and j are (2.1)
associated with nodes, p is associated
with materials points)

320

Nodal velocities are calculated
with the velocity of the material points and
the nodal mass (mp):
(2.2)
Step 3. Material points positions
are updated,
(3.1)
The strain increment of a material points
can be expressed in terms of the
nodal velocity
(3.2)
The stresses ( are updated using a
material constitutive model (D).
(3.3)
Step 4. The material properties are
updated in the material points.
The nodal values are discarded because
all the updated information is already
transferred to material points. The
computational grid is initialized for the next
step.
321
3. RESULTS
The comparison between the experimental
and numerical results is a direct way to
evaluate the performance of the numerical
model. It is also an interesting way to get
more information about the evolution of
some parameters which are difficult to
measure experimentally. Landslide
failures have been induced and analysed
using PIV and then modelled by MPM.
Consider a first case (A). Failure of 60°
sand slope was induced by removing the
guillotine. The slope dimensions are
250mm height, 200mm width and 330mm
long.
The granular soil is a calcic-silicic dune
sand from Castelldefels beach. The
properties of the material and the slope
are summarised in the Table 1.
Table 1. Sand proprieties
Sand density 1540 kg/m³
Grain density 2665 kg/m³
Porosity 0.42
Friction angle 30o
Dilatancy angle 8o
Poisson coefficient 0.3
Skeleton elastic 30 MPa
modulus
Cohesion 0
The humidity of Case A was the laboratory
ambient humidity (RH = 34.4%). The slope
is not stable without the support provided
by the guillotine. The failure starts when
the guillotine is removed and the slope
recover a new equilibrium after 0.9
seconds.
Figure 4 shows the evolution of the
landslide. The sliding surface developing
at the beginning of the experiment
remains in the same place during the
remaining failure stages. During the first
stages of the landslide, all the points in the
mobilised part have a similar velocity.

Figure 4. Velocity maps derived from the
digital imaging of the failure induced in
a 60º sand slope using a PIV
technique. The color indicates the
magnitude of the displacement velocity
vector of each point.
However, as the landslide evolves, the
velocity distribution became less
homogeneous. This is a consequence of
the complex granular flow established.
Between times 0.2s and 0.3s, after
landslide initiation, the maximum speed is
reached. The average velocity in this time
322
interval ranges between 0.2 and 0.4 m/s.
The acceleration during this period is low.
This observation is interpreted as being
the result of the similar values reached by
the gravitational forces and the frictional
and impact forces between particles which
oppose the motion. In the final stage of
the landslide, the effect of the
accumulated flatter slope on the bottom
part of the slope becomes dominant. In
the image, at t=0.4s the velocity in the
lower part is close 0 m/s. The stabilized
slope surface becomes a curved profile
having minimum inclination at the lower
part and a steepest angle at the top of the
slope.
3.1. Comparison of experimental and
numerical results.
Figures 5 to 8 present a comparison
between experimental and numerical
results. To be able to compare both
results the images have been printed with
the same colour scale (each time has its
own velocity scale, depending on the
maximum attained speed).
The landslide behaviour predicted by the
MPM model is similar to the experimental
observations. The main difference lies in
the velocity reached by the slope surface
which is higher in the numerical model.
This difference can be attributed, in part,
to the difficulties of PIV techniques to
calculate the velocity vectors on the model
boundaries. However, the MPM simulation
can also be improved. It is felt that the
contact between the mobilized part of the
slope and the thin stationary sand base is
not properly accounted for. This is
probably associated with the contact
algorithm introduced in the numerical
model. A final source of discrepancy may
derive from the constitutive model.
Granular flows imply a loss of energy by
impact among particles, a phenomenon
not included in the Mohr-Coulomb model
adopted in the simulations performed. An
additional internal dissipation would
probably result in a decrease of the
observed velocity.
Figure 5. Comparison between experimental (a) and Figure 7. Comparison between experimental (a) and
numerical (b) results of the velocity magnitude numerical (b) results of the velocity magnitude
distribution at t = 0.1s. distribution at t = 0.22s.

Figure 6. Comparison between experimental (a) and Figure 8. Comparison between experimental (a) and
numerical (b) results of the velocity magnitude numerical (b) results of the velocity magnitude
distribution at t = 0.16s. distribution at t = 0.28s.

323

4. CONCLUSIONS
The comparison between the experimental
and the numerical results confirms that the
Material Point Method is a powerful
numerical tool to analyse the behaviour of
landslides.
The combination of the PIV analysis of
experimental results and the MPM
numerical simulation was found to be a
useful procedure to test the numerical
method and to suggest further
developments of the code.
ACKNOWLEDGEMENTS
This thesis is being possible due the a FI
grant from the Agency for Management of
University and Research Grants (AGAUR)
of Generalitat de Catalunya (Catalan
government). The guidance received from
the Thesis supervisors, Dr. N. Pinyol and
Prof. E. Alonso is acknowledged.
REFERENCES
Adrian, R. J. (1991). Particle-imaging
techniques for experimental fluid
mechanics. Annual review of fluid
mechanics, 23(1), 261-304.
Baba, H. O., & Peth, S. (2012). Large
scale soil box test to investigate soil
deformation and creep movement on
slopes by Particle Image Velocimetry
(PIV). Soil and Tillage Research, 125,
38-43.
Chen, S., Chen, L., Zhou, M., & Huang, J.
(2013, December). Experimental
Investigation on Factors Influencing
Stable Slope Angle of Granular
Accumulation. In Applied Mechanics
and Materials (Vol. 438, pp. 1238-
1243).
Fern, J., Soga, K., & Robert, D. (2015,
January). Shear strength and dilatancy
of partially saturated sand in direct
shear tests. In TC105 ISSMGE
International Symposium on
Geomechanics from Micro to Macro, IS-
324
Cambridge 2014 (pp. 1391-1396).
Taylor & Francis Group.
Fraysse, N., Thomé, H., & Petit, L. (1999).
Humidity effects on the stability of a
sandpile. The European Physical
Journal B-Condensed Matter and
Complex Systems, 11(4), 615-619.
Khalili, N., Geiser, F., & Blight, G. E.
(2004). Effective stress in unsaturated
soils: review with new evidence.
International Journal of Geomechanics,
4(2), 115-126.
Montrasio, L., Schilirò, L., & Terrone, A.
(2015). Physical and numerical
modelling of shallow landslides.
Landslides, 1-11.
Sulsky, D., S.-J. Zhou, and H. L. Schreyer
(1995). Application of a particle-in-cell
method to solid mefchanics. Computer
Physics Communications, 87(1-2):236–
252.
Sulsky, D., Z. Chen, and H. Schreyer
1994. A particle method for history-
dependent materials. Computer
Methods in Applied Mechanics and
Engineering, 118(1-2):179–196.
Take, W. A., Bolton, M. D., Wong, P. C.
P., & Yeung, F. J. (2004). Evaluation of
landslide triggering mechanisms in
model fill slopes. Landslides, 1(3), 173-
184.
Thielicke, W., & Stamhuis, E. J. (2014).
PIVlab–Towards user-friendly,
affordable and accurate digital particle
image velocimetry in MATLAB. Journal
of Open Research Software, 2(1), e30.
White, D. J., Take, W. A., & Bolton, M. D.
(2003). Soil deformation measurement
using particle image velocimetry (PIV)
and photogrammetry. Geotechnique,
53(7), 619-632.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Combined geotechnical and structural numerical analysis for

cut slope stabilization

Katerina LAKOýEVIû1, Mario STOJANOVSKI1*

1
Civil Engineering Institute “Macedonia” JsC – Skopje, MACEDONIA

ABSTRACT
For providing stability of embankment formed for the purpose of obtaining geometry planned for
the E-75 Highway, reinforced concrete retaining wall with counterforts with total height of
H=13.20m has been designed. For the sections in side cut, slope stability analysis has been made
and appropriate measures are designed.
The static analysis of the retaining structure is made in the software GEO 5 – module Cantilever
Wall, for calculation of the strengths of the active earth pressure and the external loads (vehicle
and seismic load), maximal edge stresses, safety factors against sliding and overturning and in
software Radimpex Tower 6.0 for dimensioning of the reinforced concrete cross sections, as well
as control of the cracks at the structure elements, deflection and settlement on the structure.
Slope stability analysis has been made using the software Roscience Slide according to
Spencer’s Method. For achieving the necessary safety factors with and without seismic impact,
decreasing of slope inclination is made and geocomposite for erosion control have been designed.
As finial result of the dimensioned reinforced concrete retaining wall with counterforts and slope
stability, technical drawings are made and next phase is construction of planned embankment and
side cut implementing the designed measures for providing stability.

Keywords: retaining wall, GEO 5 - module Cantilever Wall, Radimpex Tower 6.0,
slope stability, Roscience Slide

physical and mechanical characteristics of

1. INTRODUCTION
On the new design Highway E-75 for a
part of the subsection Cariþina Dolina -
tunnel Manajle embankment formed of
reinforced soil and slope with high
inclination grade are designed. The main
problem has appeared after the start of
the terrain construction work and
examination of the prevalent material
parameters. The terrain investigation and
testing has shown the need of designing a
new technical solution. Namely, the
* presenting author
the excavated materials from this section
on the highway are lower than the material
parameter used in the made analysis for
reinforced soil embankment. As a result of
this they cannot be used for forming on
the designed embankment as it was
planned. Also the mechanical parameters
on the material in cut and side cut are
lower from the material parameter used in
the made analysis, so redesigning of the
slope for providing stability should be
done. Bringing the material with necessary
parameters from other place make the
embankment formed of reinforced soil 1:1
more expensive, so the Investor and PPRDR
.5

Constructor decided for comparison new

1:
12.0
technical solution to be designed. For

1
redesigning on this subsection new 12.0 N
geotechnical and structural numerical al/PC

analyses are carried out for reinforced Dq

concrete wall with counterforts and slope
stability provided with decrease on the
N - embankment
slope grade. PPRDR - sandy-silty debris
For the two designed solutions PC - sandy gravel
PPRDR D q - rock (dacite)
(reinforced soil and reinforced concrete
retaining wall) bill of quantity was prepared
and as a final decision on Investor and N
al/PC
Constructor, the total cost for constructing
on this part from the Highway E-75 will be Dq
taken into consideration.
Figure no.1.
2. STRUCTURAL AND SLOPE Characteristic engineering-geological profiles
STABILITY ANALYSIS in side cut and embankment

For the purpose of preparing the new Table No.1. Materials parameters
technical solution, as part of performed Material J>kN/m3@ M>ஈ@ c [kPa]
investigations made for construction on
Embankment
the highway in this section, appropriate (N)
20.00 32.0 0.0
geotechnical investigations and testing of
locations have been performed, for needs Sandy silty
20.00 30.0 5.0
of designing retaining structure for making debris (PPRDR)
of the planned embankment and cut slope Sandy gravel
21.00 35.0 0.0
stabilization. The results acquired from the (Pý)
geotechnical terrain investigations and Compact dacite
laboratory tests have been presented in 26.00 45.0 750.0
(Įq)
the Report prepared by the Civil
Engineering Institute “Macedonia” which For providing stability of embankment
are using as input data for the analysis. formed for the purpose of obtaining
The designed highway on this section geometry planned for the E-75 Highway, a
should be constructed on the existing reinforced concrete retaining wall with
terrain, present with non - coherent and counterforts found in sandy silty debris
poorly compacted materials consist of and/or sandy gravel is designed.
sandy silty debris (dilluvium-proluvium) Stabilization and arrangement of the slope
and poorly compacted materials consist of formed in a layer of sandy silty debris at
sandy gravel (alluvium) that cover the the side cut is necessary to be with slope
basic rock presented with compact, very grade of the berms 1:1 and 1:1.5 with
solid and fresh dacite (Įq) (figure no.1). 8.0m high and 3.0m berms width. In the
The parameters for prevalent material first technical solution, the inclination of
are presented in table no.1, they are the slope berms is 2.5:3.0. In this design
according to the prepared Report and are the inclination of the upper berms are
using for the structural and slope analysis. decreased, but the regulation line is taken
into consideration.
The new design reinforced concrete
retaining wall with counterforts is with total

326
length (measured along the axis of the
wall stem) of 203.08m and it is formed by
22 sections with different height. The
highest section is with total height of
H=13.20m, the dimensions of foundations
are B/d=7.20m/1.20m and the thickness of
the reinforced concrete stem and
counterforts amounts to 60cm. The total
height of the other section that formed the
retaining wall varies depending on the
position of the sections and it amounts to
h=7.0÷12.0m. Dimensions of the wall
foundation slab and the height of the
counterforts are also changeable
Figure no.2. Layout on the retaining wall and the stabilized slope parts
Figure no.3. Front view on the retaining wall and the stabilized slope parts
Due to the big height of the retaining
wall, counterforts positioned on the side of
the embankment are planned. By
designing reinforced concrete retaining
wall with counterforts, the horizontal
displacements of the reinforced concrete
stem are decreases (Yp,max=2.5cm on the
top of the wall stem, from permanent and
variable loads), and in this case they are
within allowed boundaries. Also the static
system for the calculation of the structure
is changing, thus it decreases the internal
static forces, as well as the necessary
327
appropriate with the total height of the
section. The geometry of the retaining wall
in its base follows the right edge of the
highway and the crown of the retaining
wall follows the longitudinal gradient of the
road (figure no.2 and figure no.3). For
optimal solution, different variants are
made with changes on section lengths
and disposition on the counterforts. This
one has shown as optimal, taking into
consideration the dimensions on
constructive elements and amount of
necessary reinforcement.
reinforcement for reinforcing on the
retaining wall structural elements. The
design of a reinforced concrete retaining
wall with counterforts for wall with such
height makes this solution economical and
provides the necessary stability.
The static analysis of the retaining
structure i.e. the calculation of the
strengths of the active earth pressure and
the external loads (vehicle V600 load and
seismic), the maximal edge stresses, as
well as the safety factors against sliding
and overturning, are calculated by the Fs,o = Mres /Movr > 1.50 – safety factor
software GEO 5, module – Cantilever against overturning
Wall. The obtained static forces from the Fs,s = Hres /Hact > 1.30 – safety factor
different load cases are used as input
parameters in the software Radimpex against sliding
Tower 6.0 for a loading on reinforcement The calculated safety factors for the
concrete stem and foundation slab. With highest section are Fs,o=1.66 and
them, the dimensioning of the reinforced Fs,s=1.36, and they are greater that the
concrete cross sections are carried out, as allowed ones in condition of earthquake
well as control of the cracks at the influence.
structure elements, deflection and On the figure no.4 all forces that act on
settlement on the structure have been the reinforced concrete steam and
made. foundation are shown. Presented forces
The calculation of the allowable bearing are values obtained from: wall weight,
capacity of the foundation sub-base has earth wedge weigh, active pressure,
been made according to the Terzaghi’s vehicle, passive pressure and forces given
theory for strip foundation and according from earthquake influence on construction
to the Technical regulations for and earthquake influence on earth as
foundation. The following correlations are additional active pressure. These forces
used for calculation: presented as surface triangles or trapeze
According to Terzaghi’s theory: loads are used for loading on the
qf c˜Nc  J1 ˜Df ˜Nq  0.5˜ J2 ˜B˜N J reinforced concrete retaining wall and
dimension on the structural elements.
According to the Technical regulations for
foundation:
qa (J ˜B˜N J ˜s J ˜i J)/2  (cm  q˜tgMm)
˜Nc ˜sc ˜dc ˜ic  q
The allowed bearing given as an
average value of the results obtained
according the two methods is 600-750kPa,
depending of the foundations strip width.
As a result of the allowed settlements, all
values are limited to 350kPa, which when
the seismic influence is taken into
consideration are increased for 50%.
The forces that act on the retaining wall
are calculated using GEO 5, module –
Cantilever Wall, the horizontal (T) and the
vertical (N) resultant components from
different load types, its eccentricity, the Figure no.4. Forces acting on construction
maximal stresses of the sub-base, as well As next step in the computer software
as the safety factors against sliding and Radimpex Tower 6 3D model on the
overturning for safety control on the wall. highest section is made (figure no.5) and
The performed structural analysis of the according to the output data for the
retaining wall gives us the conclusion that necessary reinforcement, dimensioning of
the maximum edge stresses the retaining wall structural elements is
(ımax=505.97kPa) are lower than the carried out.
allowable bearing capacity, taking the
seismic influence into account. The safety
factors against overturning and sliding are
calculated using the following correlation:
328
 contact foundation - ground (s,tla). Output
results for these parameters are lower
than the allowed ones, so the ultimate limit
state on the retaining wall is satisfied. With
the dimensioned reinforcement, crack
control is made (ak2/ak1(t’)) and the
cracks are lower than the allowed ones.
With that the serviceability limit state is
also satisfied.
The construction on the retaining wall
needs to be in sections (two at the most),
starting from the section no.1 which has
deepest foundation. All sections are
interchangeably dilated with expansion
joints, made of Styrofoam with 2cm
Figure no.5. 3D model of highest section thickness and bitumen bands on the inner
Under influence of all loads (as input side.
values from GEO 5) the necessary Behind the foundations of the retaining
reinforcement in each element separately wall, a drainage system has been
(foundation, stem and counterfort) is designed for the purpose of reception and
obtained. This information is used for draining the infiltrated ground waters, i.e.
reinforcing on RC stem - top and bottom on the side toward the embankment. For
zone, top and bottom zone of the the drainage, PVC drainage pipe with
foundation slab and counterforts with the diameter Ø250mm is designed that should
ribbed reinforcement B500B that covers be perforated. In order to protect the
the needed reinforcement and the minimal granular material from getting sludgy, the
percentage of reinforcement. All perforated openings should be wrapped
calculations are made with concrete with non-woven geotextile. In the retaining
parameter for MB30 (C25/30). wall, stem weepholes made of PVC pipes
Reinforcement used for reinforcing on the with diameter Ø80mm have been planned.
concrete structural elements is given in
For the purpose of achieving slope
table no. 2.
stabilization on sections in side cut, a
Table no.2. Max. necessary and dimensioned decrease of the slope grade and
reinforcement
geocomposite net for landslide protection
Constructive Apot Adim is designed. Stability analysis has been
element [cm ]2
[cm2] performed using the software Roscience
b. zone 48.81 Ø25/10 Slide according to Spencer’s Method.
foundation
t. zone 14.89 Ø16/10 Circular-cylindrical sliding plains with a
b. zone 4.17 Ø12/15 minimal security factor have been
Ø16/20 obtained using this software. Geotechnical
stem models that were analyzed are in fact
t. zone 31.07 +
Ø25/20 modelled using terrain lithology with
76.78 strength parameters of the existent
counterfort 16Ø25 formations according prepared
(hidden pillar)
Geotechnical Report.
3D model is also used for control on
the maximal horizontal displacements (Yp) A stability analysis has been carried out
which appear at the top of the reinforced on the cross sections with slope grade 1:1
concrete retaining walls and the vertical on the first slope platform and 1:1.5 on the
settlements on the structure under the remaining two platforms, whereas the
influence of the seismic action on the height of the platforms amounts to 8.0m

329
and berms width being 3.0m. The stability
analysis of the slopes have been made for
two cases, with and without seismic
impact, and security factors bigger than
the allowed ones have been obtained
(Fs>1.3 without seismic impact and
Fs>1.1 taking in consideration the seismic
impact). On figure no.6 the slope stability
analysis is shown for the characteristic
section on the designed subsection
without seismic impact. The lowest safety
factors are Fs=1.522 without seismic and
Fs=1.158 with earthquake influence. That
means that the necessary slope stability
with these inclinations is secured.
Figure no.6. Slope stability model
For erosion control on the slopes, a
geocomposite formed of double turned
galvanized wire mesh and plastic
polypropylene mash for erosion protection
have been planned. A geocomposite
mesh should be well fastened and
anchored on the top of the slope using
steel anchors. A geocomposite mesh
should be anchored to the base along the
entire area using galvanized steel fixing
pins defined by the geocomposite mesh
manufacturer. Before the process of
setting the geocomposite mesh starts,
cleaning of the slope surface should be
performed and all unstable parts to be
removed.
3. CONCLUSIONS
For achieving optimal retaining wall with
such height (13.20m), the positioning and
thickness on the counterforts is very
important in order to obtain minimal
dimensions on constructive elements and
330
minimal amount of reinforcement that
satisfy needed parameter for stability and
serviceability on the wall.
For achieving optimal stably slope,
berms with different inclination grade are
analyzed and stabilized slope with the
lowest amount of excavated material is
chosen.
As finial result of the dimensioned
reinforced concrete retaining wall with
counterforts and slope stability, technical
drawings are made and bill of quantity is
prepared. With comparison on the new
technical solution cost and the technical
solution from the main design
(embankment formed of reinforced soil), it
is decided that their costs are almost
equal. So, as final decision, a new
designed technical solution is chosen and
next phase is construction of planned
embankment and side cut slope
implementing the designed measures for
provided stability.
REFERENCES
[1] Technical norms and regulations for
concrete and reinforced concrete -
Nacional regulation (PBAB 1987)
[2] Pro. Dr. NabeelAbdulrazzaqJasim
(2012), “Optimal design of reinforced
concrete counterfort retaining walls”,
Basrah Journal for Engineering Science
[3] Prof. Dr. Spasen Gjorgjevski, “Shallow
foundation”, Faculty of Civil Engineering,
Chair for Geotechnics, University of
Skopje, Macedonia
[4] Prof. Dr. Maksimovic Milan (1995), “Soil
mechanics”, Faculty of Civil Engineering,
University of Beograd, Serbia
[5] Prof. Nikola Najdanoviü Civil Eng.,
Dr. Radmilo Obranoviü, Geo. Eng. (1979),
“Soil mechanics in engineering practice”
[6] Matasovic, N., (1991), “Selection of
Method for Seismic Slope Stability
Analysis”, 2nd International Conference
St. Louis, Missouri
[7] Hugh Brooks,Civil & Structural Eng,
John P. Nielsen,Civil and Geotech. Eng.
Basics of retaining wall design, 10th
Edition (1992), Newport Beach, California
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Erosion of cuts in flysch: field monitoring by TLS and

modelling of slope geometry development

Goran VLASTELICA

University of Split, Faculty of Civil Engineering, Architecture and Geodesy

ABSTRACT
The amount of eroded material is vital information for understanding weathering processes in rock
masses that contain soft rock members, especially for the purpose of forecasting and retrograde
analysis of slope geometry development. On steep slopes and cuts it is often not possible to easily
and precisely determine slope geometry without using modern surveying technology such as TLS.
Series of observations on man-made cuts at 14 pilot sites in the area of city of Split were made and
selected results of that monitoring are presented in this paper.

Keywords: flysch, erosion, cut, TLS, monitoring

costs of the objects situated at the foot of

1. INTRODUCTION
Along the east Adriatic coast the
occurrence of soft rock is associated with
sedimentary deposits known as flysch. It
represents a complex geological and
sedimentary environment made
different lithological components, whereas
only few, depending on their mineral
composition, are extremely susceptible to
weathering and erosion (Mišþeviü &
Vlastelica, 2014). Precisely, the diversity
of components in the flysch rock mass
emphasises the issue of differential
weathering and small to medium scale
rock falls on cuts in this strata (Figure 1).
The paper focuses on the existent cuts
made in the wider area of the town Split.
The exogenous processes that take place
(weathering, surface layer erosion,
localised slides, rock falls, etc.) have
causal effect on the security, maintenance
the slope, and the stability of objects at
the top (Figure 2).
Observing the cuts in soft rock by the
use of terrestrial laser scanner (TLS) is the
basis for understanding the erosion
of process in this rock type. Although it has
been available for some time, the
relatively high price has made the TLS
miss the wider practical usage in geology
and other disciplines. According to the
available literature there are only few
similar studies in this field (e.g. Petrie &
Toth, 2008).
The results of this study, weather the
methodology approach or definite findings
for similar/same rock mass type, could be
used to analyse stability and slope
geometry development in existent and
future cuts.

Figure 1. Rockfall caused by long term
differential weathering, Town of Omiš.
Figure 2. Cut in flysch, location Žnjan, City
of Split, Croatia.
2. METHODOLOGY
2.1. Terrestrial Laser Scanning
Terrestrial Laser Scanning (TLS) is a term
for surveying by which it is possible to
obtain large quantity of data (coordinates
of points), unlike any other conventional
surveying methods (i.e. total stations). In
addition to the term TLS, the term
frequently used is LiDAR (Light Detection
and Ranging), which is usually associated
with obtaining data from the air (eg. using
an aircraft as a platform) or the definition
of the technology itself.
With TLS a large amount of processed
data is obtained, which is called cloud of
points (Kordiü, 2014). When checkpoints
are referenced in the known coordinate
system, then the whole cloud of points can
be oriented in the same system. The
332
points can be further specified with colour,
i.e. RGB component can be defined and
associated, when scanners have an
integrated and calibrated digital camera.
2.2. Comparison of point clouds from
different epochs
The change in morphology of cuts is
carried out by comparing the point clouds
from different epochs using the following
methodology (Abellan et al 2011, Lim et
all 2009):
x Obtaining a reference point cloud
(Combine multiple point clouds in case of
more scanning positions. Preferably
georeferencing is done to determine
spatial orientation of a cut).
x Creating a Triangle Irregular Network
(TIN) model of the surface of the cut -
reference surface (S0).
x Obtaining new point cloud after a
certain period (PC1, PC2, ..., PCn)
x Preparing the data for alignment
between the multiple epochs (defining
alignment matrix).
x Comparing of the PCi with the S0.
x Calculating the difference for each
different epoch.
x Creating cross-section for comparison
or directly comparing point cloud with S0 in
3D.
The alignment matrix is defined by
using a fixed object in the environment
(buildings near of cuts or geological
members which are not subject to
weathering in engineering time scale), in
following four steps:
x Identification of the stable part of the
cut or an object in vicinity of the cut.
x Removal of the part of the point cloud
where the changes take place
(detachment and deposition) and any
unwanted measurements (vegetation,
moving objects, etc.).
x Alignment of the chosen stable part
using Iterative Closest Point (ICP)
algorithm, which defines the alignment
matrix of the point cloud.
x Using the alignment matrix on the and it is a measure of permanent increase
original point cloud. of volume.
Distance of each point in PCi and S0 is Change in slope angle E can be described
calculated using the "Data vs. reference by introducing the following:
comparison", (Polyworks v12), wherein ­h ˜ ctgȕ t 0
the direction of comparison is determined °

y t ®h ˜ ctgȕ  R y,s 0  t  y ult  h ˜ ctgȕ  R y,s (3)
by the direction of the vector perpendicular
to the reference plane P0.
2.3. Fisher-Lehmann mathematical
model of erosion of vertical cuts
This model (Figure 3) is a combination
of Fishers (1866) mathematical description
of the degradation of abandoned, initially
vertical, cuts in layers of chalk without
accumulation of eroded material at the
base of the cut, and Lehmann (1933)
generalized model, which introduces the
possibility of initial inclination of the slope
and the accumulation of eroded material
at the foot of the cut.
°y ult t t t ult y ult  h ˜ ctgȕ  R y,s
¯
where Ry,s is average annual erosion.
According to observations by Vlastelica
(2015), it is possible to notice a very quick
initial change in morphology of the cut;
however it can equally be attributed to the
initial stress redistribution in rock mass
and local instabilities, which depend on
the stratification and the quality of
excavation.
Taking into account this observation,
instead of the parameter Ry,s, a non-linear
function Ry,s=Ry,s(t) can be introduced, or
the linear criterion described with
expression (3) can be replaced with the bi-
linear criteria:
­h ˜ ctgȕ t 0
°°h ˜ ctgȕ  n ˜ R y,s if 0  t  t1

y t ®h ˜ ctgȕ  R  
if t1  t  y ult  h ˜ ctgȕ R y,s
(4)
convex boundary ° y,s
of intact rock °¯ y ult if t t t ult  
y ult  h ˜ ctgȕ R y,s
mass
where: n = coefficient of the initial
Initial cut
relaxation of the cut; t1 = period within the
Geometry at
time step t effects of initial relaxation of the cut
(Mišþeviü & Vlastelica, 2012).

3. PILOT SITE LOCATIONS

Figure 3. Fisher-Lehmann mathematical
In order to determine the rates of erosion
model of erosion of a vertical cut.
of the cuts, 14 locations in the area of Split
Expression for determining convex core are selected (Figure 4). Locations have
of intact rock mass takes the following different geological settings, as well as
form (Hutchinson 1998): orientations of cuttings and layering. Basic
ª m º information about the pilot locations are
y k ˜ l  m  ˜ ln « » k˜z (1)
¬ m  z  ¼
shown in Table 1.
The main aim of observations that are
where: m h/c ; k a  ac  b  c ; carried out is to determine the quantity of
l b ˜ h a  ac  b  ; a ctgĮ ; b ctgȕ ; h - eroded material (i.e. the rate of erosion
height of the cut, D- talus angle and E - progression). However, equally important
angle of the initial slope cut. The objective is to define the parameter of
parameter c is a constant needed for permanent increase of volume, since it is
exact derivation: the main parameter for all known
rock volume 1 c mathematical models that helps to define
(2) the profile of cuts due to erosion during
talus volume 1
different time periods.
333
 8-12

3 2
7 6 1 4

47
Duüe: 13-14
Figure 4. The position of the pilot locations (cuts) on the geological map of the city of Split

Table 1. Basic information about the pilot locations

Orientation
Cut Monitoring period
of bedding Height
Pilot location orientation (mm.yyyy. -
planes H (m)
Dz/Ez mm.yyyy.)
Dmp/Emp
7.7
1 Žnjan 1 255q/79q 180q/58q-74q 03.2012. – 09.2014.
(5.6-9.8)
2 Žnjan 2 270q/73q 170q/80q 5.0 03.2014. – 09.2014.
3 Trstenik 1 161q/90q 170q/90q 4.0 03.2014. – 09.2014.
4 Duilovo 1 190q/55-60q 5q/25q 25.0 03.2012. – 09.2014.
5 Gripe 1 95q/74-80q 180q/85q 5.5 03.2014. – 09.2014.

6 Kaštelet 1 165q/68q 30q/45q 16.0 03.2014. – 09.2014.

6.3
7 Kaštelet 2 142q/79q 30q/45q 03.2014. – 09.2014.
(3.5-7.5)
8 10. kolovoz 1 130q/72q 30q/40q 12.0 03.2014. – 09.2014.
9 10. kolovoz 2 230q/65q 25q/30q 7.5 03.2014. – 09.2014.
10 10. kolovoz 3 285q/77q 10q/40q 7.0 03.2014. – 09.2014.
11 10. kolovoz 4 210q/78q 25q/40q 13.0 03.2014. – 09.2014.
12 10. kolovoz 5 210q/78q 30q/35q 22.5 03.2014. – 09.2014.
13 Duüe 1 190q/69q 140q/60q 4.0 04.2012. – 09.2013.
90q270q/ Up to
14 Duüe 2 140q/60q 04.2012. – 09.2014.
45q-65q 100.0

334

4. RESULTS
4.1. Presentation of results
The paper shows results of comparisons
of point clouds for two selected pilot sites
are presented. These comparisons can be
displayed three-dimensionally, as a field of
differences in the face of the cut, or two-
dimensionally through selected
representative cross-sections.
In 3D the distances are shown
throughthe field of values. Negative values
indicate a lack of material in a given epoch
(erosion), or separation of larger rock
fragments. Positive values indicate the
accumulation of eroded material in the
form of talus at the base of the cut or
larger blocks due to rockfall. The positive
shift towards the instrument, if it is in the
2m
(a) 7.3.2012. - 21.3.2013.
(c) 7.3.2012. - 10.3.2014.
Detachment
Figure 5. Comparison of point clouds at the pilot site "Žnjan 1" for selected epochs of
measurement (a - d) with the reference surface S0
335
open face of cuts, may indicate a
displacement that preceded the rockfall.
4.2. Pilot location “Žnjan 1”
The change in the morphology of the cut
at pilot location “Žnjan 1” for the selected
time period of observation (7.3.2012.-
10.03.2014.) is shown in Figure 5. Rockfall
zones with higher amount of detachment
due to secondary cracks are shown by
white arrows in Figure 5.a, while an area
that is exclusively under the influence of
erosion is indicated by a dotted line.
The change is shown with values
ranging from -1.0 to 1.0 m, therefore in the
central part of the cut it is not possible to
discern a change after another epoch
observations. For that reason the central
part of the cut is shown in a larger scale (-
0.4 to 0 m) in Figure 6.
(b) 7.3.2012. - 27.9.2013.
(d) 7.3.2012. - 3.10.2014.
Accumulation
 2m

(a) 7.3.2012. - 21.3.2013. (b) 7.3.2012. - 27.9.2013.

(c) 7.3.2012. - 10.3.2014. (d) 7.3.2012. - 3.10.2014.

Detachment

Figure 6. Comparison of point clouds for the middle part of the cut at the pilot site "Žnjan 1" for
selected epochs of measurement (a - d) with the reference surface S0

It can be noticed that the erosion of the
surface is not uniform, although it could be
assumed it is so, solely by visual
observation. Also, it should be noted that
the erosion of cuts in flysch sequence
influence in this case is the amount and
frequency of rainfall.
Table 2. Average erosion in a typical cross-
section at the pilot site "Žnjan 1"
03.-09.2012.

03.-09.2013.

03.-09.2014.

could be also affected by methodology of

09.2012. -

09.2013. -
03.2013.

03.2014.

excavation, susceptibility to material Time period

fragmentation, the presence of secondary
cracks and incidence thin-layered resistant
members. Erosion
2.3 1.9 2.6 2.4 2.8 -
Average amount of eroded material in (cm)
the central part of the studied cut is shown
Annual
in Table 2. It is possible to notice a trend
erosion 4.2 5.0 5.5*
of increase in the amount of eroded
material for this location, with a fluctuation Ry (cm/yr)
between "winter" and "summer" six-month Indeed, by examining the documents of
period. Since this cut has a west the State Meteorological Survey a trend of
orientation, on the micro-location that is increased rainfall in the area of Split
protected from the dominant wind in this (measuring station Marjan) was noted
area (bora and south wind), it is during the TLS scanning period. At the
reasonable to assume that a dominant beginning of observation in summer of
336
2012 there was a long dry season with no Field observations on this and other
recorded rainfall, while the summer of pilot sites reveal that, after the initial
2014 had twice the amount of rainfall than relaxation of cuts, Fisher-Lehmann model
average (more than 300 mm per square correctly describes the change in the
meter). profile of cuts in soft rocks that occur
Another way of presenting results is by alone or in the flysch sequence in the area
cross sections. This way 2D mathematical of Dalmatia.
model can be used as a tool for forecast
or retrograde analysis of the geometry of
4.3. Pilot location “Trstenik 1”
the cut. As a result of a two year period
observation (7.3.2012.-10.3.2014.), Pilot site "Trstenik 1" is located 300m
representative cross-sections are created northwest of the location "Žnjan 1", with
(Figure 7). Result of a numerical similar geological settings. In this case
calculation based on the Fisher-Lehmann thicker layers of sandstone are also
model of erosion of the cut is shown in absent, however the plane of the cut is
Figure 8. directed to the south, almost parallel to the
bedding layers. Although initial slope
angle of the cut is not known, since the
excavation is probably carried out for the
construction of the building it can be
assumed to be carried out vertically.
In this case erosion is further enhanced
by toppling of thin bedded marls. The
change of the morphology of the cut for a
six month period (31.03.2014.-
04.10.2014.) is shown in Figure 10 a-d.
In this case the average change is not
possible to estimate because it can be
confused with toppling of layers on some
parts of the slope. For example, in
selected typical cross section total change
is 8 cm, which corresponds to the
thickness of layers, plus the possible shift
Figure 8. Results of observation carried out on of the layer prior to collapse.
pilot location "Žnjan 1".
yult (m) When the results of the difference
8 between the first two epochs are shown in
z (m)
the range of value from -0.03 to 0.03 m
7
(Figure 11.e), shift toward the instrument
6 on the order of 1 to 2 cm (yellow to red
areas indicated by the arrows) on the free
5
face of the cut can be noticed in some
Fisher-Lehmann
4
Fisher-Lehmann
areas. Later on, in the following epoch of
Translated Fisher–Lehmann
Fisher-Lehmann
Translatirani
measurement, in the same positions
3 7.3.2012.
rockfall or toppling is recorded.
10.3.2014.
2
Poþetni With this data it is conclusively proven
1 that the above events are preceded by a
y (m) shift on the free part of the excavation.
-2 -1
0
0 1 2 3 4 5 6 7
Therefore, with timely and frequent
Figure 9. Fisher-Lehmann model based on the observation with TLS these events can be
results of observations carried out on pilot observed and predicted.
location "Žnjan 1".

337
 (a) Referent surface 31.3.2014. (b) 31.3.2014. - 23.6.2014.
(scale -0.4 to 0.4 m)

(c) 31.3.2014. - 4.10.2014. (d) Characteristic cross sections

(scale -0.4 to 0.4 m)

(e) 31.3.2014. - 23.6.2014. (f) 31.3.2014. - 4.10.2014.

(scale -0.03 to 0.03 m) (scale -0.4 to 0.4 m)

Detachment Accumulation (displacement)

Detachment Accumulation

Figure 10. Comparison of point clouds at the pilot site "Trstenik 1" for selected epochs of
measurement.

338
4.4. Result summary deviations through reinterpretation of
Selected results of observations are characteristic profiles or influence of
presented in Table 3. On each of the 14 stronger members in the flysch sequence.
pilot sites there is a possibility of local

Table 3. Summary of the most important results of observations by TLS on selected pilot locations
Angle of
Height
Pilot location the free Talus angle Average erosion
H (m)
face

7,7
1 Žnjan 1 81q 37q 5,0 cm/year
(5,6-9,8)
2 Žnjan 2 5,0 73q 37q 2,6 / 6 months

3 Trstenik 1 4,0 90q 37q Not applicable

4 Duilovo 1 25,0 55q-70q - 20 cm/2,5 years

5 Gripe 1 5,5 74q 36q 1,5 cm/6 months

6 Kaštelet 1 16,0 68q 37q 4,5 cm/6 months

6,3
7 Kaštelet 2 77q 35q N/A
(3,5-7,5)
8 10. kolovoz 1 12,0 72q 38q 3,4 cm/6 months

9 10. kolovoz 2 7,5 65q 36q 3,0 cm/6 months

10 10. kolovoz 3 7,0 77q 36q 3,0 cm/6 months

11 10. kolovoz 4 13,0 78q 38q 3,0 cm/6 months

12 10. kolovoz 5 22,5 78q 38q 3,0 cm/6 months

13 Duüe 1 4,0 65q 35q N/A

14 Duüe 2 do 100,0 - - 2-5cm/10 months

Remark: For complete data please refer to Vlastelica (2015)

5. CONCLUSIONS
Except for the results of direct geometrical
comparisons, this paper proposes a
modified setup of long term monitoring for
this type of rock mass. After the first scan
is georeferenced, by using software with
ICP algorithm and detecting fixed objects
on or near the slope, it is possible to use
only the TLS for all other data acquisitions
on the monitored cut.
The most prominent results are
recorded through slope angles of the cut
339
and scree deposit at the foot of the slope,
as well as average erosion rate over each
time interval, depending on the mass rock
or soft/hard layer ratio.
Angle of the free face of the cut, in this
work, ranges from 70q to 80q except when
a milder angle of the initial slope was
used. The angle of the talus ranges from
35q to 38q, regardless of the type of
material or the ratio of the softer and
harder layers of the flysch sequence. Even
after many years of observations (e.g. for
location "Gripe 1" after 40 years) the basic
relations of the free face and talus
coincide with similar examples with shorter
intervals. The average annual erosion at
all locations varies from 3 to 7 cm/year
(average of 5 cm/year), up to 10 cm/year
for coastal cliffs ("Duilovo 1" and "Kaštelet
1"). These values are a rough guideline,
set out on more than one profile for each
selected location as the average value for
a short period of observation. Locally
higher values can occur.
For a full analysis of the time depended
stability, it is always necessary to consider
the possibility of local instability
phenomena such as: sliding on primary or
secondary fracture systems (e.g. "Žnjan
1"), toppling (e.g. "Trstenik 1"), landslides
caused by differential weathering and
erosion (e.g. "10. Kolovoz 4") or large
scale landslides (e.g. "Duilovo 1").
Pertaining to the observations made on
“Žnjan 1” (Figure 8), Fisher-Lehmann
model (Figure 9) as one of the most
common mathematical models for erosion
was tested. It can be concluded that this
model has full applicable value for this
type of materials.
The results of this study, weather the
methodology approach or definite findings
for similar rock mass type, could be used
to analyse stability of existent and future
cuts.
ACKNOWLEDGEMENTS
The presented results derive from the
Japanese-Croatian bilateral research
project: Risk Identification and Land-Use
Planning for Disaster Mitigation of
Landslides, sponsored by Japan Science
and Technology Agency - JST, Japan
International Co-operation Agency - JICA
and Ministry of Science, Education and
Sport of Republic of Croatia.
Special thanks to my mentor prof.
Predrag Mišþeviü for support and advices
at all stages of research and to prof.
Hiroshi Fukuoka from Niigata University,
Research Institute for Natural Hazards
and Disaster Recovery.
340
REFERENCES
Abellan, A., Vilaplana, J. M., Calvet, J.,
Garcia-Selles, D., Asensio, E. (2011)
“Rockfall monitoring by Terrestrial
Laser Scanning – case study of the
basaltic rock face at Castellfollit de la
Roca (Catalonia, Spain)”. Nat. Hazards
Earth Syst. Sci., 11, 829–841.
Fisher, O. (1866) “On the disintegration of
a chalk cliff”, Geo-logical Magazine, 3,
354–356.
Hutchinson, J.N. (1998). “A small-scale
field check on the Fisher–Lehmann and
Bakker–Le Heux cliff degradation
models”. Earth Surface Processes and
Landforms 23, 913–926.
Kordiü, B. (2014) “Development of 3D TLS
method for determining and analysing
of landslide surface movements”,
Zagreb: University of Zagreb, Faculty of
Geodesy. PhD thesis.
Lehmann, O. (1933) “Morphologische
Theorie der Verwitterung von
Steinschlagwänden”. Zurich:
Vierteljahrsschrift der Naturforschende
Gesellschaft, 78, 83–126.
Lim, M., Rosser, N. J., Allison, R. J.,
Petley, D. N. (2009). “Erosional
processes in the hard rock coastal cliffs
at Staithes, North Yorkshire”.
Geomorphology, 114(1–2), 12–21.
Mišþeviü, P. & Vlastelica, G. (2012) “Time-
dependant stability of slopes excavated
in marl”. Graÿevinar 64, 6; 451-461
Mišþeviü, P., Vlastelica, G. (2014) „Impact
of weathering on slope stability in soft
rock mass”, Journal of Rock Mechanics
and Geotechnical Engineering, Volume
6, Issue 3, Pages 240–250.
Petrie, G., Toth, C.K. (2008) “Introduction
to laser ranging, profiling and scanning”
Topographic Laser Ranging and
Scanning: Principles and Processing.
Eds.: Shan, J., Toth, C.K., CRC Press /
Taylor & Francis, London, pp. 1–28.
Vlastelica, G. (2015): “The Influence of
Weathering on Durability of Cuts in Soft
Rock Mass”. Split: University of Split,
Faculty of Civil Engineering,
Architecture and Geodesy. PhD thesis.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Geotechnics in Belarus: problems and solutions

Sergey IGNATOV1, Igor BOIKO2

1
Institut Belzeldorproekt, Minsk Belarus, Member of the Board Belorussian Geotechnical
Society
2
Belarusian national technical university, Minsk Belarus, Member of the Board Belorussian
Geotechnical Society

ABSTRACT
The Republic of Belarus construction industry is faced daily with the need to solve geotechnical
problems such as great loads on the base foundation, the cramped conditions construction,
buildings and structures reconstruction in the conditions of modern development, technical and
technological progress. It is clear that the material structural characteristics are determined in
advance, ensured their strength and reliability during the erection process, building, expluatation.
However, this can't be applied to base foundation soils, which are "dynamic" system, and physical,
mechanical and deformational characteristics change in the process of the sediments formation,
the construction and buildings and structures operations.
The existing buildings and structures ground physical, mechanical and deformation
characteristics are affected by the environment cyclical changes (winter, summer), anthropogenic
factors (vibration, dynamics, the overhead structures loading methods, topography and landscape
changes and so on). These influences often degrade the ground characteristics, reduce the
grounds and structures reliability reduce. In the article the Belarusian geological and engineering
conditions, the zonation are presented, information about the deposits and landforms formation is
also given.
This article describes the facilities of the new building, reconstruction, carried out in different
geological conditions and in the city building constrained conditions. The significant Belarusion
construction projects, information about geotechnical solutions adopted for their construction is
also provided.

Keywords: pressure, foundation, settlement, monitoring, construction

anthropogenic period. The relief final

1. BELARUS GROUND PROPERTIES
Belarus modern ground properties were
developed over a long period of time.
Belarusian territory was subjected to five
glaciations, which were distributed almost
in the whole territory In the modern
formation was occurred in the period of
Moscow glacier formation, motion and
melting (figure 1).
Up to 52 % of the total anthropogenic
territory cover accounts for glacial
deposits. Moraine was formed from debris
resulting from the movement of the
glacier. Moraine formation is often
represented by a class of cohesive soils:
sandy loam, loam and clay. Glacial clays
and loams can be characterized by the
presence of stratification, non-uniform
addition, and chemical and mineralogical
compositions variability. Moraine clay
deposits humidity, the modulus of total
deformation and the resistance to uniaxial
compression increase with the depth of
the. Glacial sands are often deposited in
the form of layers and interlayers in
moraine clay soils massifs. Sands are
often heterogeneous in granulometric
composition and sometimes contain
pressurized water.
Figure 1. Belarusian geological conditions
Fluvioglacial (glacial–fluvial) deposits
are represented by the glacier meltwater
deposits? which passed through the
cracks in the body of the glacier and
moved under the ice toward its edge.
Fluvioglacial deposits accupy one-third of
soils, which are exposed to the
construction industry. The internal texture
can be inclined (15 – 30 degrees) and
horizontal. The soils formation,
transformation and deposition history data
has influenced their physical and
mechanical characteristics. In the sand
deposits massifs can take place changes
of the density (from soft to dense).
Alluvial deposits (deposits of river
flows) are structure and heterogeneity
complex soils, which were formed in
various physical-geographical conditions
342
of water flows. These deposits are
characterized by high porosity, high
content of organic matter and a large
compressibility, and the anisotropy of
properties in the natural condition. The
average power of these deposits occurring
is in the range of 3 – 7 meters in Belarus.
2. DIFFICULT GEOLOGICAL
CONDITIONS CONSTRUCTION
EXAMPLES
One of the examples of technical
solutions of the ground properties
improvement is the construction of the
plant "SOOO Alumintechno" in Minsk.
The existing topography was changed
during mans activity by the bulk soils,
landfills, construction debris. The pond
was located in the south-eastern part of
the site, territory was heavily pitted, low-
lying areas were flooded. Soil conditions
are characterized by the heterogeneity,
space variability and shown in the figure 2.
Figure 2. The geological conditions of
horizontal reinforcing
It was decided to perform a horizontal
reinforcement of the ground, and to create
a reinforced strips, which should be used
as base for the foundation slabs (figure 3,
4).
Figure 3. The construction of horizontal
reinforcing
 - fluvioglacial sediments (fIIsž) are
distributed throughout the site. They are
disposed from the surface and under -
technogenic deposits. Their power is up to
25.0 m. Represented by yellow-gray, light-
gray, brown sands from shallow to
gravelly.

Figure 4. The realization of horizontal

reinforcing

The graphics sediment from pressures

under the foot of the foundations erected
on reinforced base are shown in figure 5.

Figure 6. The geological conditions of vertical

reinforcing

Figure 7. The geological conditions of vertical

Figure 5. The graph S=f(p) for the plate reinforcing
foundation on the horizontal reinforced base

The vertical reinforcement example is

the technical solution for the 19-storey
residential building ground vertical
reinforcement (st. F. Skorina, Minsk).
The building place is located on the
slope of the fluvioglacial hill.
The place can be characterize by
following stratigraphic-genetic complexes:
- technogenic deposits of the modern
age (tIV) with the capacity of 0.4 – 1.9 m,
Figure 8. The graph S=f(p) for the plate
represented by a mixture of soil and foundation on the vertical reinforced base
various debris:

343
3. FOUNDATION RECONSTRUCTION
EXAMPLES
State archive reconstruction is the
symbolic object, where soil bases
strengthening by the grout injection
became the important and only possible
method of the strengthening. This building
dates from the mid-second half of XVII
century. The building was made of brick,
with longitudinal outer and inner bearing
walls and pillars. The thickness of the
masonry exterior walls exceeds 1.05 m, a
Figure 10. Strengthening design project
internal bearing walls – brick, with the
thickness of 250, 380 and 510 mm.
The foundations reconstruction 4. CRAMPED CONDITIONS
technology violation was the urgent CONSTRUCTION EXAMPLES
execution reinforce works reason: under
the foundation base soil watering, the soil The object "XVII–XVIII architecture
excavation under the foundation base. monument century Reconstruction (Snct.
These facts cause to critical fallout of Cyril and Methodius center, Minsk) is the
foundations, cracks in the brick arches, example of a device deep pits in cramped
opening width up to 85mm (figure 9) conditions with application of anchors for
fastening the retaining walls.
The anchors usage was the necessity
to perform excavation for the construction
of a new building in the cramped
conditions of the old buildings of the
historic spots of Minsk (figure 11).

Figure 9. The walls destruction

The foundation base ground is medium

strength gravelly sand with the following
characteristics: Y=17,1 kN/m3; C=1 kPa M
=38, E=42 MPa.
Figure11. The deep pit example
The fastest and most reliable method of
base soil consolidating, the stabilizing of
the sediment of the building was the Geomorphologically the building site is
drillenjected technology usage. A project dedicated to the Minsk regional glacial
was designed to enhance ground hills. That fact determines the nature of
properties and with a partial transfer of an the soils occurrence, their strength and
existing shallow foundation on gantry deformation characteristics.
system of piles (figure 10). The place can be characterize by
following stratigraphic-genetic complexes:
344
 - a) modern technogenic sediments (th
IV), which are disposed from the surface,
and represented by different grain size
sands, with lenses of clayey soils, tuff soil
layers;
- b) alluvial deposits are represented by
fine, medium, large, gravelly and gravelly
sands, with a total capacity of alluvial
deposits 1,5-8,0 m;
- c ) moraine deposits are represented
red-brown loams and clay loams, with thin
layers and lenses of water-saturated
sands;
- d) fluvioglacial (glacial–fluvial)
deposits are presented by heterogeneous
sands with lenses of sandy loam, with a
thickness up to 15 cm.
Figure 12. The pile and anchor position plan
The anchors length has 11.0 m., the
length of the root is 3.0 m (for anchors of
the design load of 289 kN) and 6.0 m (for
anchors of the design load 358,8 kN). All
construction and installation works were
performed and today the building is put
into operation.
5. TRANSPORT FLYOVER
CROSSING
Minsk is growing city, so a great
number of new and reconstructed roads,
bridges are constructed and built.
The flyover crossing at the intersection
of Independence Avenue with the
Filimonov str. construction is the example
of such kind of transport building. The
345
project task was to insure the stability of
present exploited metro tunnels and other
adjoined buildings of Minsk.
The project involved construction over
existing subway tunnels of the overpass
bridge superstructure with precast
concrete structures for highway bridges
with a length of 24 m and 21 m.
Reinforced concrete lining of the
existing tunnels was carried out using
rings consisting of ribbed reinforced
concrete blocks
The results of the calculation of the
phased development of the soil under the
flyover with a length of 200 m and 45 it
found that the existing tunnel deformation
had significant size from the extra project
influence (figure 13).
Figure 13. The transport flyover crossing
calculation schema
It was decided to execute on-site
maintenance of construction works with
the continuous monitoring of the stress-
strain state of structures tunnels and other
structures within the area of influence in
order to ensure the required construction
quickness and safe operation of the tunnel
The monitoring included overhead and
underground geodesy monitoring and
visual monitoring, structures instrumental
monitoring and electronic remote
monitoring.
Ongoing monitoring allowed us to
assess underground structures stress and
deformations during construction of the
flyover crossing, as well as to exploit the
tunnels as planned, to avoid defects that
reduce the bearing capacity, unacceptable
inclinations and other deformations of
tunnels. The figure shows the project of
the transport flyover crossing.

Figure 14. Transport flyover crossing design

project

6. CONCLUSIONS
Thus this paper reviews the current
situation of the Belarusian
geotechniks,examples of new construction
and renovation allow the reader to review
the status of geotechnical engineering in
our country.

346
25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Session 7

Soil dynamics
www.KellerGeotehnica.ro
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

4D seismic and velocity evolution in monitoring heated heavy

oil sand reservoirs

Hong DOAN*

FUGRO GeoConsulting SAS, Nanterre, France

ABSTRACT
Thermally enhanced recovery processes involving continuous steam injection entail relatively
complicated changes in pore fluid, pore pressure, stress and temperature in the rock reservoir.
Time-lapse (or 4D) seismic surveys are currently developed to delineate the steam-affected areas,
but the data interpretation is often a highly difficult exercise.
Numerical simulations were carried out on a case study so as to provide an estimation of the
evolution of pressure, temperature, pore fluid saturation, stress and strain in any zone located
around the injector and producer wells. These outputs then were used as inputs for the velocity
assessment in the oil sand rock reservoir through the Ciz and Shapiro (2007)’s approach.
Results are helpful for a better understanding and interpretation of 4D seismic data. The
significant changes of the wave velocities with the steam arrival can be used as an indicator for the
monitoring of steam chamber growth around the wells, which is crucial for optimization of reservoir
development.

Keywords: heavy oil sand, coupled thermo-hydro-mechanical, 4D seismic

and seismic properties of the rock layers.

1. INTRODUCTION
Huge quantities of heavy oils (heavy oil,
extra heavy oil and bitumen) are mainly
trapped in unconsolidated sand and
sandstone reservoirs in Western Canada
and Eastern Venezuela basins. In
thermally enhanced recovery processes
like cyclic steam stimulation (CSS) or
steam assisted gravity drainage (SAGD),
the injection of steam in oil sand deposits
produces changes in temperature,
pressures, stress and pore fluid content.
These changes obviously affect the elastic
* presenting author
Time-lapse 3D seismic (4D seismic) and
other seismic technologies can be used to
monitor the impact of these changes in
unconsolidated sands or weakly
consolidated sand reservoirs (Zhang et al.,
2005; Nakayama et al., 2008; Tanaka et
al., 2009). By comparing maps of seismic
attributes (velocity, amplitude, attenuation,
etc) at different times, the continuous
spatial distribution of the heated reservoir
zones can be approximately located. The
feasibility of seismic monitoring methods is
based on the visible changes of seismic
attributes during thermal process that
could be understood from a combination
of rock physics modelling and direct
laboratory measurements.
Several attempts to model the elastic
properties of rock saturated with heavy oil
were previously conducted (Das and
Batzle, 2008; Ciz et al., 2009). Difficulties
are largely related to the viscoelastic
behaviour of heavy oils. They behave like
a quasi-solid at high frequencies and/or
low temperature and like a viscous fluid at
low frequencies and/or high temperature
(Batzle et al., 2006; Hinkle, 2008). The
physical property of rock saturated with
heavy oil is different from that of rock
saturated with conventional fluids with no
shear modulus. It makes the standard
poroelastic Biot-Gassmann approach
(Biot, 1941; Gassmann, 1951)
inapplicable (Ciz and Shapiro, 2007;
Gurevich et al., 2008; Ciz et al., 2009).
This paper describes an approach for
the evolution of seismic velocities of heavy
oil sands during the steam injection on the
basis of a coupled thermo-hydro-
mechanical modelling of SAGD and of the
generalization of Biot-Gassmann theory.
2. EVOLUTION OF PRESSURE,
TEMPERATURE AND PORE FLUID
SATURATION DURING SAGD
The coupled thermo-hydro-mechanical
modelling of SAGD (Lerat et al., 2010;
Zandi, 2011) showed how the different
fields (stress, pressure, temperature and
steam) develop with respect to time. The
simulation is based on an iterative
coupling between a reservoir simulator
(PumaFlowTM) and a geomechanical
simulator (ABAQUSTM) in order to take
into account the geomechanical effects on
the porous volume and the permeability
changes of the reservoir.
A 20 meters thick (in the Z vertical
direction) reservoir at a depth of 250
meters is modelled. The domain
considered for the reservoir simulator is
rectangular with a width of 147 meters (in
the X direction) and a length of 500
meters along the Y-axis of the well pair.
350
The mesh used by the reservoir simulator
in the X-Z plane is shown in Figure 1. It
contains only one grid block in the Y
direction. The pair of wells is located at the
centre of the domain and the injection well
is 3.5 meters upper than the production
well (see Figure 1 where points I and P
represent the position of the injection and
production wells, respectively). The
domain considered in the geomechanical
model is built from the surface to a depth
of 280 meters, including an underburden
layer 10 meters thick. It should be noted
that, in the geomechanical model, the
displacements along the Y axis are
blocked, corresponding to a plane strain
condition. For simplicity, an elastic and
isotropic behaviour is considered for the
rock. Further details can be found in Zandi
(2011) and Nauroy et al. (2012).
Simulations were performed over 1500
days. The first 120 days correspond to the
pre-heating of the regions surrounding
both wells by circulating steam in order to
reduce the oil viscosity and create a
hydraulic link between them. Once pre-
heating is finished, steam injection starts
with a temperature of about 260°C. The
steam injection and production rates are
automatically controlled with a maximal
pressure sets to 5 MPa in the injection
well and a minimum pressure of 0.5 MPa
for the production well. Rates are
automatically controlled based on the
analysis of the temperature of each well.
The rates are adjusted in order to keep
the production well temperature between
20°C and 35°C lower than the injection
well temperature.
In the model associated with the
reservoir simulator, fluids cannot flow
through the boundaries, but heat losses
by conduction through the upper and
lower boundaries are taken into account
by a simplified one-dimensional modelling
of the overburden and underburden that is
oriented in the vertical direction. In the
geomechanical model, the horizontal
displacements of lateral boundaries, as
well as all the displacements of the lower
boundary, are blocked.
 The iterative coupling between the permeability and porous volume are
reservoir simulator and the geomechanical corrected 300 times during the
simulator is performed for periods of 5 simulations.
days. It means that the reservoir

a) Geometry modelling and definition of grid cells

b) Pressure field

c) Temperature field
Figure 1. Reservoir simulator spatial discretization (a) and pressure (b) and temperature (c) fields
at the end of simulation (1500 days) in reservoir

The evolution of the main variables in water and 8% oil. Temperature is

grid cell B is presented in Figure 2. The
arrival of different fronts can be
successively observed. The stress front is
followed first by the pore pressure front
followed by the temperature front and, at
the end, the change in saturation.
The invasion of steam (consisting of at
least 70% vapour water with the
remainder in the liquid phase) occurs
when temperature exceeds 70°C, firstly by
replacing oil with water, then with steam
vapour when temperature reaches 260°C.
Accordingly, at grid cell B, after 1500 days
the pore space is filled of steam injection
with a mixture of 72% steam vapour, 20%
estimated of about 260°C.
3. VISCOELASTIC MODEL FOR HEAVY
OIL SAND
The elastic properties of the oil sands
depend on the properties of the sand
matrix, on the nature of the pore fluid
(heavy oil, water or steam) and on in-situ
stresses and pore pressure. These
properties also depend on the frequency
of the elastic waves since oil sands are
viscoelastic materials.
The approach of Ciz and Shapiro
(2007), originally developed for porous
rocks saturated with an elastic material,

351
can be used to compute the effective 4. APPLICATION FOR TIME-LAPSE
elastic properties of porous rocks filled MONITORING
with heavy oil, a viscoelastic material with
The probable evolution of P- and S-wave
a non negligible shear modulus. This
velocities in an oil sand reservoir at
approach reduces to the classical Biot-
seismic frequency bandwidth (cf. 100 Hz)
Gassmann one if the pore fluid is non-
at grid cell B is illustrated in Figure 2 with
viscous in the case of a rock skeleton
respect to the arrival of the different fronts
made up of a single homogeneous
associated to steam injection. The seismic
mineral.
velocities were calculated by Ciz and
According to Ciz and Shapiro, the two Shapiro approach by using pore pressure,
following equations describe the effective temperature and fluid saturation changes
moduli of a sand saturated with a viscous deduced from SAGD coupled thermo-
fluid: hydro-mechanical modelling.
1 ( K dr1  K s1 ) 2
K sat K dr1 
I ( K f 1  K s1 )  ( K dr1  K s1 )
1 (Gdr1  Gs1 ) 2
Gsat Gdr1 
I (G f 1  Gs1 )  (Gdr1  Gs1 )
where I is an uniformly distributed
porosity.
K dr , Gdr are the drained bulk and shear
moduli of the clean sand (with no oil),
respectively.
K sat , Gsat are the effective bulk and shear
modulus of the undrained saturated
system.
K f , G f are the bulk and shear moduli of
saturating fluid.
K s , Gs are the moduli of the solid phase.
A comprehensive calibration of this Ciz
and Shapiro model on the laboratory
results (in terms of P- and S-waves
velocities) was presented in Nauroy et al.
(2012). Laboratory measurements were
performed on natural oil sand samples
coming from fluvial-estuarine McMurray
sand cores extracted at 75m below
surface from a shallow Athabasca deposit.
Velocities data were obtained for a range
of pressure, of temperature and of
frequency. Further details can be found in
Doan et al. (2010) and Doan (2011).
Results suggested a good promise of
Ciz and Shapiro approach in assessing
such physic properties of rock saturated
with heavy oil. Figure 2. Evolution of the different fields at a
typical grid cell (B)

352
 The arrival of the first front called stress
is related to a structural effect. As the
mean total stress remains constant, the
velocities do not reflect any change.
When the pore pressure front arrives,
velocities Vp and Vs gently decrease, as
the mean effective stress decreases.
As the temperature front arrives, the
velocities decrease of 10% for the
compressional waves and of 30% for
shear waves. The decrease appears more
pronounced below 60°C, levelling off
beyond this temperature level. These
modelling results are consistent with
experimental studies reporting
noticeable change in both velocities at low
temperature (below 60°C) (Nauroy et al.
2012). In addition, because the amount of
decrease of S-wave velocity is relatively
larger than that of P-wave velocity, the
Vp/Vs ratio significantly increases.
The substitution of heavy oil and water
by steam (at around 260°C) drastically
reduces the compressional velocities Vp
with little effect on the shear velocities Vs.
The introduction of even a small
percentage of gas bubbles into the fluid
will have a significant effect on the bulk
modulus of the mixture but very little effect
on its density. The sudden appearance of
gas has little effect on Vs due to
decreasing density. The Vp/Vs ratio thus
drops significantly.
In summary, it is observed that
velocities decrease with the steam arrival.
These successive changes could be
identified by 4D seismic. The sudden
appearance of gas occurring as a result of
the oil production is the most influent
factor leading to the velocity changes.
With respect to the temperature, it is
appreciated that the evolution of S-wave
velocities are relatively more pronounced
than that of P-wave velocities, suggesting
to use the S-wave velocities as an
indicator of temperature. As pointed out
Nauroy et al. (2012), at seismic
frequencies (about 100 Hz), most changes
in velocities occur at low temperature
(below 60°C) with only small changes in
velocities above this temperature. In other
words, one could effectively locate the
heated zones in the reservoirs but one
probably could not estimate the exact
temperature of these zones.
5. CONCLUSIONS
A coupled thermo-hydro-mechanical
modelling of SAGD showed how the
different fronts (dilation, pressure,
temperature and steam) develop and
propagate with respect to time. The
invasion of these fronts impact
consequently 4D seismic monitoring
because of the changes in seismic
a attributes (velocities, attenuations, etc.).
Ciz and Shapiro approach showed a
good promise in assessing the physic
properties of rock saturated with heavy oil.
This approach has been employed to
estimate the evolution of the P and S-
wave velocities in the reservoir at seismic
frequency during a SAGD process.
Prior to the arrival of the steam,
temperature appears to be the dominant
factor affecting the wave velocity
response. The sudden appearance or
disappearance of gas is probably the
strongest factor in producing the velocity
changes. Future work would need to
confirm these findings.
ACKNOWLEDGEMENTS
The work has been supervised by J.F.
Nauroy (IFPEN), P. Delage (ENPC), and
M. Mainguy (TOTAL).
The author would like also to thank P.
Rasolofosaon, N Guy and G. Renard
(IFPEN) for useful recommendations and
helpful comments. This work is part of the
first author’s PhD thesis funding by
IFPEN.
REFERENCES
Batzle M.L, Hofmann R. and Han D.H.
(2006) Heavy oils-seismic properties,
The Leading Edge 25, 750–756.
Biot M.A. (1941) General theory of three-
dimensional consolidation, J. Applied
Phys. 12, 155-164.
353
Ciz R. and Shapiro S.A. (2007)
Generalization of Gassmann equations
for porous media saturated with a solid
material, Geophysics 72, A75-A79.
Ciz R., Saenger E.H., Gurevich B. and
Shapiro S.A. (2009) Temperature-
dependent poroelastic and viscoelastic
effects on microscale—modelling of
seismic reflections in heavy oil
reservoirs, Geophysical Journal
International 176, 3, 822–832.
Das A. and Batzle M. (2008) Modeling
studies of heavy oil in between solid
and fluid properties, The Leading Edge
27, 1116–1123.
Doan, D.H., Nauroy J.F., Delage P.,
Baroni A., and Mainguy M. (2010)
Effect of temperature on ultrasonic
velocities of unconsolidated sandstones
reservoirs during the SAGD recovery
process. Paper in 44th US Rock
Mechanics Symposium and 5th U.S.-
Canada Rock Mechanics Symposium,
Salt Lake City, June 27-30.
Doan D.H. (2011) Impact de la
température sur les propriétés
mécaniques et acoustiques des roches
concernées par la production en SAGD,
lors de l'injection de vapeur dans les
réservoirs d'huile lourde, Thèse, Ecole
des Ponts ParisTech.
Gassmann F. (1951) Elastic waves
through a packing of spheres,
Geophysics 16, 673-685.
Gurevich B., Osypov K., Ciz R. and
Makarynska D. (2008) Modeling elastic
wave velocities and attenuation in rocks
saturated with heavy oil, Geophysics
73, E115–E122.
Hinkle A. (2008) Relating chemical and
physical properties of heavy oils,
Master Thesis, Colorado School of
Mines.
Lerat O., Adjemian F., Baroni A., Etienne
G., Renard G., Bathellier E., Forgues
E., Aubin F. and Euzen T. (2010)
Modelling of 4D Seismic Data for the
Monitoring of Steam Chamber Growth
During the SAGD Process, Journal of
354
Canadian Petroleum Technology 49, 6,
21-30.
Nakayama T., Takahashi A., Skinner L.
and Kato A. (2008) Monitoring an oil-
sands reservoir in northwest Alberta
using time-lapse 3D seismic and 3D P-
SV converted-wave data, The Leading
Edge 27, 9, 1158-1175.
Nauroy J.F., Doan, D.H., Guy, N., Baroni,
A., Delage P., and Mainguy M. (2012).
Evolution of seismic velocities in heavy
oil sand reservoirs during thermal
recovery process. Oil & Gas Science
and Technology,
doi.org/10.2516/ogst/2012027.
Tanaka M., Endo K. and Onozulka S.
(2009) Estimation of Steam-Chamber
Extent Using 4D Seismic, Canadian
International Petroleum Conference,
Calgary, Alberta, Jun 16 – 18.
Zandi S. (2011) Modélisation des effets
géomécaniques de l’injection de vapeur
dans les réservoirs de bruts lourds,
Thèse, Ecole des Mines ParisTech.
Zhang W., Youn S. and Doan, Q. (2005)
Understanding reservoir architectures
and steam chamber growth at Christina
Lake, Alberta, by using 4D seismic and
crosswell seismic imaging. Paper SPE
97808 presented at the SPE/PS-
CIM/CHOA International Thermal
Operations and Heavy Oil Symposium,
Calgary, Alberta, 1 - 3 November.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Cohesive soil laboratory investigation using resonant column

and torsional shear device

Alexandra Alisa GĂINĂ1*

1
“Gheorghe Asachi” Technical University of Iasi, Department of Transportation
Infrastructure and Foundations, ROMANIA

ABSTRACT
The two key parameters used in computers software specific for the seismic ground response
analysis are: the shear modulus and damping ratio. The paper presents the results obtained using
as equipment a fixed-free resonant column, adapted to perform both conventional resonant column
tests (RCT) and cyclic torsional shear tests (TST). The samples were carried out by sampling silty
clay soil retrieved from a site of Iasi city, Romania.
The first part of the paper describes the tested material and the second part deals with the
dynamic testing and discussions of the results. The soil sample was subjected to an isotropic and
undrained consolidation at a strain range between 0.0003 % and 0.1 %.
The results have been graphically expressed, mainly the maximum and normalized shear
moduli and damping ratios versus cyclic shear strain amplitudes. A comparison of the normalized
shear modulus and damping ratio curves computed from an empirical model with those obtained in
laboratory tests it is also presented.

Keywords: shear modulus, damping ratio, cohesive soil, resonant column

know the behaviour of the soil subjected

1. INTRODUCTION
The South-Eastern areas of Europe are
known as having the highest seismic
activity. One of them includes Romanian
country, where the majority of the
earthquakes occurred had the epicentre in
Vrancea region. The seismic activity of
this source is characterized as subcrustal
with a hypocentre between 70 and 170 km
and it is situated at the curvature of the
Carpathian Mountains. That is why it is
very important for Romanian territory to
to dynamic loads.
It is very important to know the dynamic
parameters of soil layers for the evaluation
of the ground response analysis because
they can amplify or attenuate the seismic
waves. They can be determined via
laboratory tests such as resonant column
tests (RCT), torsional shear test (TST),
cyclic triaxial test (CTT), cyclic direct shear
test (CDST), etc. or in situ tests by using
reflection and refraction seismic methods,
down-hole, cross-hole, up-hole tests, etc.

* presenting author
 The paper aim is to evaluate the shear
modulus and damping ratio obtained in
laboratory on dynamic resonant column
and static cyclic torsional shear apparatus.
The tested soil is a clayey sandy silt
sampled from a site of Ia‫܈‬i city, Romania.
2. TESTED MATERIAL
The tested material is a clayey sandy silt
found on alluvial complex of Bahlui River,
which crosses the city over a length of 14
km. The site from where the clay was
retrieved is located near the center of the
city, at 350 m distance from the riverbed.
The undisturbed specimens were
sampled from a depth of 8.0 m. In Table 1
the average physical properties of the
tested soil are presented. The material
has a clay fraction of 25% and a low
plasticity PI=16.77%, according to
Casagrande classification.
Table 1. Physical characteristics of tested
material
ȡS w0 wL e SR PI
[g/cm3] [%] [%] [-] [-] [%]
2.65 19.72 31.6 0.581 0.89 16.77
3. LABORATORY TESTING
APPARATUS AND PROCEDURES
3.1. Laboratory testing apparatus
The shear modulus and damping ratio
characteristics of the tested soil were
obtained from a fixed-free resonant
column torsional shear (RCTS) apparatus
adapted to perform both conventional RCT
and TST (Cavallaro, 2003).
The experimental studies have been
made at the geotechnical laboratory of the
University of Rome “Sapienza”, Italy.
RCTS provides measurements of dynamic
soil properties at low strain. Two types of
tests can be performed on RCTS using
either filled cylinder soil specimen, where
only an isotropic confinement can be done
3.2. Testing program
Two cylindrical soil samples have been
made and subjected one of them to RCT
and the other one to TST. The soil sample
was subjected to an isotropic and
356
or hollow cylinder subjected only to an
anisotropic confinement. In the present
paper a cylindrical soil sample has been
tested with a 35.7 mm in diameter and
72.0 mm in height. The soil specimen is
fixed at the bottom and at the top is
connected an electrical motor consists of
four permanent magnets placed at the end
of each arm of a four-arm driven-plate and
inside of a magnetic field generated by
four pairs of coils (Mosallamy, 2014).
Through this driving system periodically
torsional loads (sinusoidal, triangular or
trapezoidal) can be applied
(Subramaniam, 2015). The maximum
torque moment for the device used in this
study is up to 0.43 Nm (D’Elia, 2001).
An accelerometer mounted on the
driven-plate is used to determine the
resonant frequency of the sample, by
controlling the amplitude and frequency of
the applied voltage (Mosallamy, 2014).
For the cyclic condition a proximity
transducer placed on the driven-plate, in
order to monitor the rotation of the sample
is used. Through a linear variable
differential transformer LVDT the vertical
displacements can be measured (D’Elia,
2001). All the components mentioned
above are schematically presented in
Figure 1.
Figure 1. Schematic diagram of the RCTS
undrained consolidation at a strain range
between 0.0003 % and 0.1 %. Both tests
were carried out in four stages.
In the first stage the sample has been
subjected to a cell pressure of 200 kPa
and the in situ vertical effective stress ı’0 This section deals with the effects of
= 102 kPa was estimate. After the stage a various factors that influence the shear
RC test was performed up to a strain level modulus and damping ratio.
of 0.004 %. Then with opening the
4.1. Shear modulus and damping ratio
drainage a consolidation phase was
performed for minimum 24 hours. In this The reduction of the shear modulus for a
second stage the effective stress was ı’0 high strain level, obtained in RCTs, is
= 100 kPa and with closed drainage a RC presented in Figure 2. It can be noticed
test was done until a strain level of 0.07%. that the G values of the tested material are
The third and fourth stages follow the decreasing continuously with increasing
same procedure as the one described in the strain amplitude. In respect with the
second stage, but at increasing the consolidation stress, the initial shear
effective stress at ı’0 = 200 kPa and 300 modulus G0 values increase from 60 MPa
kPa, respectively. up to 133 MPa.
All the tests were performed at a back
pressure of 200 kPa. The frequencies
range was between 10 to 65 Hz.
The TSTs have been carried out in the
same manner as RCTs, but using a
sinusoidal time history at a frequency of
0.1 Hz and 5 cycles have been applied for
each strain level.
In Table 2 the laboratory test conditions
are presented.
Table 2. RCTS testing conditions Figure 2. The G dependency to Ȗ, by RCT
Stage ı’0 TS RC
(kPa) Ȗ N Ȗ f In Figure 3 the damping ratio values are
plotted versus Ȗ for RC tests. The
(%) (%) (Hz)
damping ratio curves have, in general, a
I 102 0.0003 5 0.0004 10-65 continuously ascending in respect with
-0.003 -0.004 strain level. Another aspect can be
II 100 0.0003 5 0.0004 10-65 highlighted, namely that a dependency on
-0.005 -0.07 the effective stress could not be found.
Thus, the influence of the consolidation
III 200 0.0002 5 0.0004 10-65
-0.002 -0.07 pressure can be neglected.

IV 300 0.0003 5 0.0003 10-65

-0.004 -0.1

4. RESULTS AND DISCUSSION

Shear wave velocity, Vs, initial shear
modulus, G0, initial damping ratio, D0,
shear modulus, G(Ȗ), corresponding to
particular shear strain, Ȗ and damping
ratio, D(Ȗ) are the main parameters
obtained in RCTS test (Diaz-Rodriguez,
1992). The determination of Vs and G is Figure 3. The D dependency to Ȗ, by RCT
based on the elasticity theory applied for In Figure 4 the normalized maximum
an isotropic material and for D the shear modulus, G/G0, versus shear strain,
viscoelasticity theory is used (Kramer, Ȗ, is presented. It can be noticed that
1996; Stanciu, 2006).
357
consolidation pressure has only a slight
influence on the G/G0.

Figure 6. The G dependency to Ȗ, by TST

Figure 4. Influence of the effective stress on

G/G0 – Ȗ reduction, by RCT
During the fourth stage an excessive pore
pressure of 63 kPa was generated during
the RCT. In Figure 5 the normalized
excess pore pressure ratio, ǻu/p’, which
means the ratio between the measured
excess pore pressure and the confining
pressure at which the test has been
performed (Mohtar, 2014) in respect with Ȗ Figure 7. The D dependency to Ȗ, by RCT
is plotted. The excess pore pressure
It is also remarked (Figure 8) that there is
generation represents an important issue
no influence of the effective stress on
in liquefaction prediction phenomenon
G/G0 values from TST.
(Mohtar, 2014).

Figure 8. G/G0 – Ȗ reduction, by TST

Figure 5. Normalized pore pressure, ǻu/p’,
against Ȗ measured in RCT
The results obtained in TST are
presented in Figures 6 and 7, in terms of
G and D versus Ȗ. Compared to the RCT
data the initial shear modulus G0 have
approximately the same values, but in
TST the strain level do not exceed 0.004
%. Thus, at small strain levels the data
obtained on both RCT and TST are
approximately equal, at the same isotropic
confining pressure.
4.2. Empirical models
Two empirical models have been used in
this study, as developed by Ishibashi and
Zhang (1993) and Darandeli (2001). Both
models take into consideration the mean
effective confining stress ı’0m and the
plasticity index PI, over consolidation ratio
(OCR) only in Darandeli model is taken
into account. The calculus was done for
an OCR=1.
Figures 8 and 9 highlight the
comparison between the normalized shear
modulus and damping ratio against strain
level obtained in the laboratory conditions

358
and by empirical prediction, for the same
isotropic confining stress.
It was noticed that Ishibashi and Zhang
(1993) curve is situated above the
laboratory curve, which means that G/G0
given by this model is overestimated. On
the contrary, a good agreement between
the two curves was provided by using the
Darandeli (2001) model.
Figure 8. Comparison between laboratory
results and values given by the empirical
model (after Ishibashi and Zhang, 1993) -
G/G0 reduction curves
Figure 9. Comparison between laboratory
results and values given by the empirical
model (after Darandeli, 2001) - G/G0 reduction
curves
359
5. CONCLUSIONS
In the present paper the static
characteristics by performing TST and
dynamic characteristics by performing
RCT for a clayey sandy silt have been
analyzed. The influence of the isotropic
confining stress and dependency on the
shear strain has been analyzed on the
shear modulus, damping ratio and
normalized shear modulus. It was
remarked a good agreement between the
data obtained on both RCT and TST at
small strain level, which means that the
testing condition do not affect the
measured dynamic parameters. It was
noticed that the shear modulus is
influenced by the confining stress, but less
influence on damping ratio was observed,
as well as on the normalize shear
modulus.
An empirical prediction of the G/G0 by
using the Ishibashi and Zhang (1993) and
Darandeli (2001) models was also
performed and compared with the
measured values from RCTS tests. The
following comparison led to these
conclusions:
x the two prediction models do not
completely agree with each other;
x a good agreement between the
measured values and the values
given by the empirical model
proposed by Darandeli (2001) has
been noticed.
Therefore, a close attention should be
given in determining the soil dynamic
parameters, because it represents a key
tool in ground response analysis. More
accuracy is needed in their determination
providing to obtain a site response
analysis.
ACKNOWLEDGEMENTS
The author would like to thank to Luigi
Callisto, associate professor at the
University of Rome “La Sapienza” for the
help, guidance and advices given during
research activity and to Irina Lungu,
professor at “Gheorghe Asachi” Technical
University of Ia‫܈‬i for the review of the Tests on Cement Treated Marine Clay”,
paper. Indian Geotechnical Journal.
Stanciu A., Lungu I. (2006) “Funda‫܊‬ii”,
REFERENCES Editura Tehnică, Bucure‫܈‬ti.
Cavallaro A., Lanzo G., Pagliaroli A.,
Maugeri M., Lo Presti D.C.F. (2003) “A
comparative study on shear modulus
and damping ratio of cohesive soil from
laboratory tests”, Proceedings of the
third international symposium on
deformation characteristics of
geomaterials (Eds: H. Di Benedetto, T.
Doanh, H. Geoffroy, C. Sauzéat), Lyon,
France.
Darendeli, M. B. (2001) “Development of a
new family of normalized modulus
reduction and material damping
curves”, Doctor of Philosophy,
University of Texas at Austin
D’Elia M. (2001) “Comportamento
meccanico in condizioni cicliche e
dinamiche di un’argilla natural
cementata”, tesi di Dottorato on
Ingegneria Geotecnica, Univerisitá di
Roma “La Sapienza”.
Diaz-Rodriguez I. A. (1992) “Dynamic
properties of Mexico City clay for wide
strain range”, Proceedings of the tenth
world conference on earthquake
engineering, Madrid, Spain.
Ishibashi I, Zhang X. (1993) “Unified
dynamic shear moduli and damping
ratios of sand and clay”, Soils and
Foundations, 33, 1, 182-191.
Kramer S. (1996) “Geotechnical
Earthquake Engineering”, Pentice Hall,
New Jersey.
Mohtar E., Bobet A., Drenvich V.P.,
Johnston C.T., Santagata M.C. (2014)
“Pore pressure generation in sand with
bentonite: from small strains to
liquefaction”, Géotechnique
Mosallamy M., Fattah T., Khoily M. (2014)
“Experimental study on the
determination of small strain-shear
modulus of loess soil”, Housing and
Building National Research Center
Journal.
Subramaniam P., Banerjee S. (2015)
“Torsional Shear and Resonant Column
360
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Combined analysis of single station ambient vibration

measurements and MASW method

Fátima GOUVEIA*

CEris, Instituto Superior Técnico, Lisbon, PORTUGAL

ABSTRACT

The fundamental frequency of the soil and the shear wave velocity soil profile until bedrock are key
parameters for an appropriate seismic site response estimation. Both can be identified in an easy
and cost-effective way by using surface seismic methods, such as the HVSR technique, the
MASW method and the microtremor Rayleigh wave curve inversion. The possibility to identify the
shear wave velocity soil profile until a considerable depth without the implementation of large
receiver spreads is now possible through the combination of the latter two techniques (joint
inversion of Rayleigh wave dispersion and ellipticity curves). In this study, we test the accuracy and
capabilities of the latter methodology for the identification of the shear wave velocity soil profile of a
site whose soil structure is known. The reliability of the results obtained through the HVSR method,
for the estimation of the fundamental frequency of the soil, is also analysed.

Keywords: shear wave velocity profile, joint inversion, MASW method, Rayleigh
wave ellipticity

seismic amplification. According to several

1. INTRODUCTION
The current EC8 soil classification
scheme has shown to be inadequate for
the estimation of the seismic response of
certain sites, such as thick soil deposits
(for instance: Park & Hashash, 2004). This
classification system, considers the
average shear-wave velocity soil profile
down to 30m depth (Vs,30) as a proxy of
seismic amplification. As a consequence,
the method is limited to superficial effects
and does not take into account the real
impedance contrast between soil and
bedrock, characteristic that highly affects
* presenting author
authors (Pitilakis, 2004; Castellaro et al,
2008; Pitilakis et al, 2013), the current soil
classification system could be optimized
by including the fundamental frequency of
the soil deposit as a primary parameter of
the classification scheme, because it
carries implicit information about soil
stiffness and thickness of the whole soil
deposit. By these means, the identification
of the fundamental frequency of the soil, in
addition to the shear wave velocity profile
until bedrock, is becoming more important
for seismic design purposes.
 Surface seismic methods, such as the
HVSR method (Nakamura, 1989, 2000,
2008) and the Multichannel Acquisition of
Surface Waves (MASW) method (Gabriels
et al, 1987; Miller et al, 1999; Park et al,
1999; Foti, 2000), are competitive
solutions for the identification of the
fundamental frequency of the soil deposit
and estimation of the shear wave velocity
soil profile, respectively.
The HVSR method, also known as
Nakamura’s technique (Nakamura, 1989,
2000, 2010), is a passive seismic method
that is applied to three-component single-
station ambient vibration recordings,
carried out at the surface, in order to
determine the fundamental frequency of
the soil deposit. It consists on the
determination of the spectral ratio between
both horizontal components and vertical
component, operation that aims to remove
the effect of the Rayleigh waves. This
technique has been widely used in many
site effect studies (Field, 1996; Lombardo
et al, 2001; Guéguen et al, 2006; Herak et
al, 2010) and showed consistency
between the HVSR peak frequency and
the fundamental frequency of the soil
deposit. The efficiency of the method for
identifying soils fundamental frequency
mainly depends on the impedance
contrast between the soil and the bedrock.
In case of high impedance contrast the
curve exhibits a clear and reliable peak.
Although its exhaustive application,
there is still a lack of consensus between
researchers concerning its fundamentals.
The main source of disagreement is
associated to the wavefield composition
around the HVSR peak. According to the
author of the technique (Nakamura, 2000),
the HVSR peak frequency results from
multiple reflections of SH waves in the
sedimentary layer. However, many
authors have recently proved, through
numerical studies (Lachet & Bard, 1994;
Bonnefoy-Claudet et al, 2006; Konno &
Ohmachi, 1998; Tuan, 2009; Fäh et al,
2001), that in general this peak is mainly
controlled by the ellipticity of Rayleigh
waves.
362
The MASW method (Gabriels et al,
1987; Miller et al, 1999; Park et al, 1999;
Foti, 2000; Lopes et al, 2008), is a surface
multi station seismic method that draws on
Rayleigh-wave dispersion characteristics
to determine the shear wave velocity soil
profile. It has been gaining popularity due
to its efficiency and ability to identify soil
profiles with velocity inversion or low
velocity zones, i.e., profiles with softer soil
layers bellow higher stiffness soil layers,
situation that is not detected by other
seismic methods (Lopes et al, 2008).
The shear wave velocity soil profiles
obtained through surface seismic methods
result from an inversion process that aims
to identify the soil profiles whose
theoretical dispersion curve presents a
good adjustment to the experimental one.
The inversion process has a non-
uniqueness solution problem which is
reflected into a given uncertainty in the
results. The reliability of the results could
be increased by combining or jointly
inverting different data types which
provide information about soil structure, as
for example, borehole data or even
seismic data like Love and Rayleigh
dispersion curves, HVSR peak or Rayleigh
wave ellipticity curve, respectively (Arai &
Tokimatsu, 2004; Fäh et al, 2008;
Boxberger et al, 2011).
The maximum depth of investigation of
active methods is very site dependent but
also depends on the receivers spread and
on the energy of the sources that generate
the recorded signal. The higher its
potential to generate low-frequency
surface waves (longer wavelengths), the
higher is the depth of investigation (Lopes,
2005). Due to the difficulty to generate
low-frequency surface waves, in a cost-
effective way, using an active-source, it is
usual to combine both active and passive
source methods (Park et al, 2005; Humire
et al, 2014). The latter is able to get low
frequency seismic data from ambient
noise sources.
The inversion of the microtremor
Rayleigh wave ellipticity curve (passive)
jointly with active-source array data has
revealed to be a good solution to increase
the depth of investigation (Hobiger et al,
2012), as an alternative to the association
of other passive methods (ex.: spatial
autocorrelation method (SPAC) (Aki,
1957)) which imply the implementation of
a wide receiver spread. The ellipticity
curve is tightly linked to the soil structure
and corresponds to the ratio between
horizontal and vertical components of the
ellipse that is described by particle motion
of Rayleigh waves, as a function of
frequency. It can be extracted from single-
station three component ambient-vibration
recordings performed at the surface, using
the RayDec method (Hobiger et al, 2009;
Hobiger, 2011). This method is based on
the Random Decrement Technique
(Asmussen, 1997) and is able to identify
Rayleigh waves by summing a large
number of specially tuned signal windows.
The effect of this type of waves is
highlighted by the high correlation
between the horizontal and vertical
components, considering a 90º phase shift
between them.
The aim of this study is to test the
accuracy and analyse the capabilities of
the latter methodology, i.e., joint inversion
of Rayleigh wave ellipticity and dispersion
Figure 1. Location map of the study area (left) and spatial distribution of the microtremor
measurements, location of the boreholes and a schematic representation of active MASW
acquisition lines (right).
363
curves, in determining the shear wave
velocity soil profile of a site located over
Tagus alluvial basin. For this purpose,
several ambient vibration recordings were
carried out in several points. One of those
recordings was made at the middle of an
active-source acquisition line in order to
perform the joint inversion process. The
HVSR curve associated to each recording
was determined in order to evaluate the
continuity of the soil layering through the
identification of its fundamental frequency.
Afterwards, Rayleigh wave ellipticity
curves were extracted using the RayDec
method. This information was used to
determine the soil profile by performing
the joint inversion with the dispersion
curve. The results were validated using
the available borehole data and compared
with the soil profiles obtained only with the
inversion of the Rayleigh wave dispersion
curve (active MASW method).
2. GEOLOGICAL SETTING AND DATA
ACQUISITION
The study area is located in Forte da
Casa, more precisely, in Vila Franca de
Xira, central Portugal. The area is flat and
located at the edge of Tagus alluvial basin
(Fig. 1).
 The available borehole data (Fig. 2)
indicates a geological sequence of landfill
materials, recent muddy alluvial soils and
Quaternary river terraces laying over a
Miocene carbonated formation. The landfill
materials are very heterogeneous,
composed by sand, sandy clay, clay and
gravel. The alluvial soils are composed by
muddy materials, sand, clay and some
shells. The river terraces are composed by
fine to coarse sand, gravel and some clay
deposits. The Miocene formation
corresponds to sandstone, marly
limestone, micaceous sandy clay and
fossil remains. A detailed revision of the
site geology as well as 3D geological
models can be found in Lopes (2005).
According to the latter study, the alluvium
thickness increases in the NW-SE
direction, towards the river.
The site is located next to population
clusters, sharing urban environmental
characteristics. It is limited at northwest
(NW) by a hillside composed by Miocene
material and at southeast (SE) by a
railway viaduct, located approximately
25m from measurement point 5. The main
identifiable sources of transient noise are
associated to road traffic at a national
Figure 2 - Soil profiles estimated from 3D geological model (Lopes, 2005) for all measurement
points and SPT results obtained at the boreholes that are close to the measurement points.
component velocity sensor (MS2003+,
SYSCOM) that presents the same
364
road (N10) located on the top of the hill
and to the passage of trains.
Two types of seismic data were used in
this study, namely single-station ambient
vibration measurements and active-source
linear array measurements. The location
of the single-station measurements is
presented in Figure 1 (right), as well as a
schematic representation of the data
acquisition lines (active MASW) used by
Lopes (2005) to determine the shear wave
velocity soil profile and whose data was
re-used in this work. In the same figure it
is also presented the location of the
boreholes whose information was used to
validate the results obtained through the
inversion of Rayleigh wave data.
The active-source array measurements,
MASW measurements, were made using
a 24bits seismograph (RAS-24,
SEISTRONIX), connected to vertical
geophones with 4.5Hz (GEOSPACE). The
data were retrieved from 24 geophones
linear spread with 2 m spacing. The active
source was located within 2 m and
generated using a 5 kg sledge hammer.
The ambient noise vibrations were
recorded using a recorder unit MR2002-
CE (SYSCOM) and an external three
characteristics for the three components
and a flat response between 1 and 350Hz.
Time series were recorded during 35 to
40min, with a sample frequency of 400Hz
and under favourable weather conditions,
i.e., weak wind and no rain. The data was
detrended, baseline corrected and
bandpass filtered between 0.5Hz and
200Hz in order to avoid excessive
distortion of the signal and aliasing,
respectively.
3. METHODOLOGY
The fundamental frequency of the soil
deposit, at each measurement point, was
estimated by applying the HVSR method
to each three-component signal. The
Horizontal-to-Vertical Spectral Ratios were
computed using the software GEOPSY
(http://www.geopsy.org). The average
spectral ratios and respective standard
deviation were determined based on the
most stationary time windows, tapered
with a 5% cosine function. The length and
number of time windows used for the
determination of the H/V spectral ratios
was defined to be high enough to
adequately capture the frequencies of
interest and ensure a considerable
number of significant cycles, respectively.
The presented HVSR curves were
smoothed using the Konno & Ohmachi
algorithm (Konno & Ohmachi, 1998) with
smoothing constant of 40 and a 5% cosine
taper. Finally, the reliability of the results
was evaluated according to the criteria
presented in SESAME guidelines
(SESAME, 2004). According to this
document, the evaluation of the reliability
of the estimated fundamental frequency
should be made in terms of stability and
clarity or resolution of the peak.
Rayleigh wave ellipticity curves were
computed for all seismic noise
measurements using the RayDec method.
Each seismic signal was splitted into 7 or
8 time windows with 5 min length (total
length of the recorded signals varies from
35 to 40 min). The method was applied to
each time window and the average
ellipticity curve was computed as the
geometric mean of all ellipticity curves
associated to each part of the signal. The
computation of the referred curves has
365
two free parameters, which were defined
according to the recommendations of the
author (Hobiger, 2011), i.e., considering a
length of the buffered signal (¨) and a
width of the frequency filter (df) fixed as a
function of frequency, corresponding
respectively to 10/f and to 0.2f.
The inversion of Rayleigh wave
dispersion curve (SWM-FC1), as well as
the joint inversion of Rayleigh wave
ellipticity (FC 04) and dispersion curves
(SWM-FC1), was performed using Dinver,
a tool from the open source software
package GEOPSY. In this program, the
inversion procedure is made by using the
Conditional Neighbourhood algorithm
(Wathelet, 2008).
The inversion process was made
considering an initial model with four
homogeneous layers over half-space. A
wide variation of wave velocities of the
layers was allowed, but considering the P-
wave velocity (VP) linked to the S-wave
velocity of the same layer (VS). The
Poisson ratio was allowed to vary between
0.2 and 0.5 and a fixed value was
considered for the soil density (2000
kg/m3), since the effect of this parameter
during this process is small. A velocity
inversion was allowed at the second layer,
as indicated by the available borehole
data. It should be noted that although low
velocity zones are not identified through
Rayleigh wave ellipticity curves (Hobiger
et al, 2012), they are clearly detected
through surface wave dispersion curves.
An equal misfit weight was initially
considered during the joint inversion of
Rayleigh wave ellipticity and dispersion
curves. Several tests were made,
considering different misfit weights, in
order to achieve a better adjustment of
both curves. This was obtained by
increasing the weight of the worst fitted
data. In this case, the imposition of a
minimum misfit of 0.05 (5% of the absolute
value of slowness) was necessary, for the
inversion of the dispersion curve, in order
to avoid the perfect adjustment of a part of
the theoretical curve at the expense of a
complete mismatch at certain frequencies.
4. RESULTS AND DISCUSSION between 3-4 m, increasing from FC 1 to
FC 3). The microtremor HVSR curve
The experimental H/V spectral ratios and
obtained at FC 04 presents a peak around
Rayleigh wave ellipticity curve are
3.4 Hz, which might correspond to the
presented in Figure 3. In this case, it was
fundamental frequency. Since the peak is
only possible to identify, with a
not clear, not respecting the criteria
considerable dispersion, the fundamental
presented in SESAME, this result cannot
frequency of the soil deposit at
be considered as reliable. In this case, the
measurement point FC 05. The first three
low impedance contrast, suggested by the
points (FC 01 03) present a typical shape
SPT results as a gradual increase of
of rock sites, which is in agreement with
resistance with depth, could be
the available borehole data. According to
responsible for the low HVSR peak
the latter, those points are located over a
amplitude.
very thin soil layer (thickness varying

FCͲ01 FCͲ02

FCͲ03 FCͲ04

FCͲ05
HVSR
fpeak =4.20±0.48Hz

Figure 3 - Mean Rayleigh wave ellipticity and standard deviation (continuous red line and dashed
red lines, respectively), average Microtremor HVSR curve and standard deviation (continuous
black line and dashed black lines, respectively).
Concerning Rayleigh wave ellipticity, it wavefield. It should be noted that the
was verified a high contribution of these bedrock depth was not identified in
type of waves on the FC 05 measured borehole B05 (near FC 05) and thus the
366
thickness of the soil deposit is not known.
According to the HVSR results, in this
point, the impedance contrast between the
soil and the bedrock must be
considerable. However, due to the
significant error associated to the
estimated fundamental frequency value,
no qualitative analysis could be made
concerning soil’s thickness. It should be
noted that in case of horizontal layering,
an increase of soil’s frequency is
associated to a reduction of its thickness.
The same is not verified in case of FC
04. It should be noted that the relative
proportion between surface and body
waves on the measured wavefield is not
only controlled by soil properties, but also
by noise source distribution (Bonnefoy-
Claudet et al, 2006). According to the
latter study, a higher contribution of body
waves is verified in case the source is
located within the bedrock.
The results obtained with the MASW
method are presented in Figure 4. The
dispersion curve SWM FC1showed a
typical behavior of a site with velocity
inversion, exhibiting the energy
dominance of higher modes at higher
frequencies and mode jumping. In order to
cover a wider range of frequencies, the
first higher mode was inverted in addition
to the fundamental mode, which allowed
Figure 4 – Local 01: Results obtained through the inversion of Rayleigh wave dispersion
SWM-FC1 (minimum misfit: 0.0119, maximum misfit presented 1.5min misfit).
This is not clear in case of ellipticity curves
without singularities, like in case of FC 04.
According to Hobiger et al. (2012), the
367
to increase the depth of investigation from
8 to 11 m. The dispersion curves were
adjusted with and approximate error of
1.19%. The best soil model, indicated in
Figure 4 as a continuous black line, is in
accordance with the borehole data,
presenting a velocity inversion at the
second layer which corresponds to the
alluvial soil (between 1 and 4 m depth),
followed by a layer with higher velocity
that might correspond to the river terrace
layer. The location of the interface
between the soil and the bedrock was not
identified. For depths higher than 11 m,
the variability of the results is too high to
consider its velocity values as reliable.
The results obtained through the joint
inversion of both Rayleigh wave
dispersion (SWM FC1) and ellipticity (FC
04) curves, are presented in Figure 5.
Several tests were made in order to
identify the segment of the ellipticity curve
to invert and the misfit weight to assign for
each data.
It should be noted that in case of
ellipticity curves with singularities (peak
and/or trough), the segment between the
peak and the subsequent minimum should
be inverted because is the part that
contains the most important information
about the soil structure.
whole plateau of the curve should be
inverted. In this study, the same was
verified and better results were obtained
while inverting the segment between 3.5
and 8.5 Hz.
Better results were obtained while
inverting the fundamental mode dispersion
and ellipticity curves. In this case, no
higher modes were considered during the
inversion process. The inversion of the
first higher mode didn’t lead to a
significant reduction of the dispersion at
higher depths and contributed to a higher
mismatch of the fundamental mode
dispersion curve at certain frequencies.
The curves were adjusted with a
minimum misfit of 10.82%, considering a
misfit weight (parameter that is used in
joint inversions as way to sum the
individual misfits, in this case, the misfits
associated to the dispersion curve and to
the ellipticity curve) of 70% and 30% for
Figure 5 – Local 01: Results obtained for the inversion of Rayleigh wave ellipticity FC-04 and
dispersion curve SWM-FC1 (min misfit: 0.1082) until misfit 1.5times.
5. CONCLUSIONS
In this study, the fundamental
frequency of the soil deposit and shear
wave velocity soil profile was determined
by the microtremor HVSR method and
through the joint inversion of Rayleigh
wave dispersion and ellipticity curves,
respectively.
Since the site geology favours the
energy dominance of higher modes, the
modal delineation of active data required
significant attention. A sensibility analysis
was made in order to identify the
segments of the ellipticity and dispersion
curves (fundamental and higher modes),
the appropriate initial model and the
368
the dispersion and ellipticity curves,
respectively. The inversion of the ellipticity
curve, in addition to the dispersion curve,
allowed to increase the depth of
investigation from 11 m to almost 60 m,
with much less variability of results. The
initial part of the soil profile is similar to the
one obtained by the MASW method
(Figure 4), with the advantage that the
interface between the soil deposit and the
bedrock is much better constrained
(between 17 and 21 m depth). Although
borehole data (borehole B16, near FC 04)
indicates that the Miocene formation is at
10 m depth, the SPT results show a
smooth transition at the interface,
consistent with a gradual increase of soil
stiffness.
model parametrization that should be
considered on the inversion process.
The use of the microtremor Rayleigh
wave ellipticity curve, in association with
the active array data (MASW), revealed to
be an effective method to increase the
depth of investigation without implying the
implementation of a wider receiver spread.
Furthermore, in geological environments
such as the one here presented (alluvial
valleys), where the variability of soil’s
thickness is big, the use of large arrays
will introduce even more uncertainty to
data interpretation. In this case, the use of
a single-station three component
microtremor recording allowed to increase
the resolution of the deeper layers and
obtain a better constrain of the bedrock
depth, when compared to the active
MASW method.
ACKNOWLEDGEMENTS
An acknowledgement to the Portuguese
funding agency, Fundação para a Ciência
e Tecnologia, for funding the PhD
(SFRH/BD/86189/2012) in which this
research is based. A special thanks to
Isabel Lopes and Rui Gomes, my
supervisors, for the continuous guidance
and great contributions.
REFERENCES
Aki K (1957). Space and Time Spectra of
Stationary Stochastic Waves, with
Special Reference to Microtremors.
Bull. Earthq. Res. Inst. Tokyo Univ. Vol.
35, pp. 415-457.
Arai H & Tokimatsu K (2004) S-Wave
Velocity Profiling by Inversion of
Microtremor H / V Spectrum. Bulletin of
the Seismological Society of America,
Vol. 95(5), pp. 1766-1778.
Asmussen JC (1997). Modal analysis
based on the random decrement
technique: application to civil
engineering structures. PhD Thesis,
University of Aalborg, pp. 227.
Bonnefoy-Claudet S, Cornou C, Bard P-Y,
Cotton F, Moczo P, Kristek J & Fäh D
(2006). H/V ratio: a tool for site effects
evaluation. Results from 1-D noise
simulations. Geophys. J. Int. Vol. 167,
pp. 827–837.
Boxberger T, Picozzi M & Parolai S (2011)
Shallow geology characterization using
Rayleigh and Love wave dispersion
curves derived from seismic noise array
measurements. J. Appl. Geophys. Vol.
75, pp. 345–354.
Castellaro S, Mulargia F & Rossi PL
(2008). Vs30: Proxy for Seismic
Amplification? Seismol. Res. Lett. Vol.
79, pp. 540–543.
Fäh D, Kind F & Giardini D (2001). A
theoretical investigation of average H/V
ratios. Geophys. J. Int. Vol-145, pp.
535–549.
369
Fäh D, Stamm G & Havenith H-B (2008).
Analysis of three-component ambient
vibration array measurements.
Geophys. J. Int. Vol. 172, pp. 199–213.
Field EH (1996). Spectral Amplification in
a Sediment-Filled Valley Exhibiting
Clear Base-Edge-Induced Waves. Bull.
Seismol. Soc. Am. Vol. 86, pp. 991–
1005.
Foti S (2000). Multistation Methods for
Geotechnical Characterization using
Surface Waves. PhD Thesis, pp. 229.
Gabriels P, Snieder R & Nolet G (1987). In
situ measurements of shear-wave
velocity in sediments with higher-mode
Rayleigh waves. Geophys. Prospect.
Vol. 35, pp. 187–196.
Guéguen P, Cournou C, Garambois S &
Banton J (2006). On the Limitation of
the H/V Spectral Ratio Using Noise as
an Exploration Tool: Application to the
Grenoble Valley (France), a Small Apex
Ratio Basin. Pure Appl. Geophys.
Vol.164, pp. 1–20.
Herak M, Allegretti I, Herak D, Kuk K, Kuk
V, Mariü K, Markušiü S & Stipþeviü J
(2010). HVSR of ambient noise in Ston
(Croatia): comparison with theoretical
spectra and with the damage
distribution after the 1996 Ston-Slano
earthquake. Bull. Earth. Eng. Vol. 8, pp.
483–499.
Hobiger M (2011). Polarisation des ondes
de surface: Caractérisation, inversion et
application à l’ étude de l’ aléa
sismique. PhD Thesis, Université de
Grenoble, Grenoble.
Hobiger M, Bard P-Y, Cornou C & Le
Bihan N (2009). Single station
determination of Rayleigh wave
ellipticity by using the random
decrement technique (RayDec).
Geophys. Res. Lett. Vol. 36, L14303.
Hobiger M, Cornou C, Wathelet M, Giulio
GD, Knapmeyer-Endrun B, Renalier F,
Bard P-Y, Savvaidis a., Hailemikael S,
Le BN, Ohrnberger M & Theodoulidis N
(2012). Ground structure imaging by
inversions of Rayleigh wave ellipticity:
sensitivity analysis and application to
 European strong-motion sites.
Geophys. J. Int. Vol. 192, pp. 207–229.
Humire F, Sáez E & Leyton F (2014).
Combining active and passive multi-
channel analysis of surface waves to
improve reliability of VS30 estimation
using standard equipment. Bull Earthq.
Eng.
Konno K & Ohmachi T (1998). Ground-
Motion Characteristics Estimated from
Spectral Ratio between Horizontal and
Vertical Components of Microtremor.
Bull. Seismol. Soc. Am. Vol. 88, pp.
228–241.
Lachet C & Bard P-Y (1994). Numerical
Possibilities and Theoretical
Investigations on the Possibilities and
Limitations of Nakamura’s Technique.
J. Phys. Earth. Vol. 42, pp. 377 397.
Lombardo G, Coco G, Corrao M, Imposa
S, Azzara R, Cara F & Rovelli A (2001).
Results of microtremor measurements
in the urban area of Catania, Italy.
Bolletino di Geofis. Teor. ed Appl. Vol.
42, pp. 317–334.
Lopes I, Santos JA & Almeida IM De
(2008). The surface wave method:
acquisition, processing and inversion.
Revista Geotecnia. Revista Luso-
Brasileira Geotecnia. Vol. 112, pp 79
109 (in Portuguese).
Lopes IF (2005). Small Strain
Characterization of Soils. Application of
the Surface Wave Method. PhD Thesis,
Universidade de Lisboa, pp. 369 (in
Portuguese).
Miller RD, Xia J, Park CB & Ivanov JM
(1999). Multichannel analysis of surface
waves to map bedrock. Leading Edge.
Vol. 18, pp. 1392–1396.
Nakamura Y (1989). A Method for
Dynamic Characteristics Estimation of
Subsurface Using Microtremor on the
Ground Surface. Q. Rep. Railw. Tech.
Res. Institute, Tokyo. Vol: 30, pp. 25–
33.
Nakamura Y (2000). Clear identification of
fundamental idea of Nakamura’s
technique and its applications.
Proceedings of the 12th World
370
Conference on Earthquake
Engineering, Auckland, New Zealand.
Nakamura Y (2008). On the H/V spectrum.
14th World Conference on Earthquake
Engineering, Beijing, China.
Nakamura Y (2010). Comment on
“Microtremor Measurements in the Nile
Delta Basin, Egypt: Response of the
Topmost Sedimentary Layer” by E. A.
Fergany and S. Bonnefoy-Claudet.
Seismol. Res. Lett. Vol. 81, pp. 241–
243.
Park CB, Miller RD, Ryden N, Xia J &
Ivanov J (2005). Combined Use of
Active and Passive Surface Waves. J.
Environ. Eng. Geophys. Vol. 10, pp.
323–334.
Park CB, Miller RD & Xia J (1999).
Multichannel analysis of surface waves.
Geophysics. Vol. 64, pp. 800–808.
Park D & Hashash YMA (2004).
Probabilistic seismic hazard analysis
with nonlinear site effects in the
Mississippi embayment. 13th World
Conference on Earthquake
Engineering. Vancouver, Canada.
Pitilakis K (2004). Site effects. Recent
advances in earthquake geotechnical
engineering and microzonation, Ansal A
(ed), Kluwer, Dordrecht. pp 139–197.
Pitilakis K, Riga E & Anastasiadis A
(2013). New code site classification,
amplification factors and normalized
response spectra based on a worldwide
ground-motion database. Bull. Earthq.
Eng. Vol. 11, pp. 925–966.
SESAME Team (2004). Guidelines for the
implementation of the H/V spectral ratio
technique on ambient vibrations -
Measurements, processing and
interpretation. SESAME Eur. Res. Proj.
WP12 – Deliverable D23.12.
Tuan TT (2009). The ellipticity (H/V-ratio)
of Rayleigh surface waves. PhD Thesis.
Friedrich-Schiller-Universität Jena. pp.
119.
Wathelet M (2008). An improved
neighborhood algorithm: Parameter
conditions and dynamic scaling.
Geophys. Res. Lett. Vol. 35, L09301.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

The influence of dynamic loads on postglacial lacustrine fine-

grained deposits

Michael HAVINGA1*
1
Graz University of Technology, Institute of Soil Mechanics and Foundation Engineering,
Graz, AUSTRIA

ABSTRACT
In earthquake engineering it is well known that different types of soil, especially sand, have the
tendency to liquefy due to dynamic loading. In this research, the dynamic loads are caused by
construction methods like jet grouting, vibro compaction or vibro replacement and vibrated sheet
pile walls and the soil types investigated are postglacial lacustrine fine-grained deposits, in general
silt with different amounts of clay and/or fine sand. The most common liquefaction criteria the
“Chinese criteria”, Wang et al. (1979) does not fit for this type of soil. It shows in some cases
similar liquefying or softening behaviour. Results using the liquefaction susceptibility criteria,
Boulanger and Idriss (2006), are presented in this paper. Furthermore a case study to determine
dynamic loads like acceleration and changes of pore water pressure during jet grouting is
presented.

Keywords: dynamic loads, jet grouting, softening

required to cut and mix the soil is between

1. INTRODUCTION
100 and 400 bars. It can be applied in
Bringing dynamic loads due to soils with a grain size from gravel to clay
construction methods like vibro for improving soil strength and reducing
replacement, vibrated sheet pile walls or soil permeability. While this has worked
jet grouting into saturated soft to very soft successfully in numerous cases, it has to
fine grained soils often result in settlement be considered that jet grouting can induce
effects without additional loads. In this settlements and ground movement as well
paper the influence of dynamic loads - in (Wong et al. 1997). In several regions of
particular jet grouting - on postglacial Austria, jet grouting also caused ground
lacustrine fine-grained deposits is surface deformation. Especially in
discussed. Jet grouting is a ground relatively young (10.000 to 5.000 years
modification system whereby the soil gets old) soft to very soft fine-grained lacustrine
eroded by a high speed water and/or a deposits, great care seems to be required
cement slurry jet (depending on whether while using methods such as jet grouting.
single, double or triple tube jet grouting The change of pore water pressure and/or
takes place). The range of pressure the vibrations have a significant influence
* presenting author
on certain particle arrangements. Possible
rearrangements of particles can lead
either to the liquefaction of soils with a

10
sand-like behaviour, or to a cyclic
softening failure. Both terms describe
similar phenomena in fine grained soils

10
exhibiting clay-like behaviour (Boulanger &
Idriss 2006). Another known term, cyclic
mobility, which was introduced by
Robertson & Fear (1995) is, in this case,
not suitable because no settlements Figure 1. Lamination of lacustrine sediment
should occur after the pore water pressure The entry values for the “Chinese criteria”
decreases to its initial value. In two cases (Wang et al. 1979) are the mass
researched here, settlements were percentage of grains finer than 0,005 mm
measured even months after using jet and the liquid limit (LL). For this research,
grouting. the grain size distribution was determined
Regarding the effects during earthquakes, by sieving and hydrometer analysis and,
a great deal of research has been for the liquid limit as well as for the plastic
conducted into the liquefaction of sandy limit (PL), the fall cone test (cone
soils and there is an increasing research penetrometer test) and the linear
focus on silty and clayey soils. The shrinkage test according to
influence of construction methods on soils, ÖN B 4411:2009 were performed.
which are susceptible to liquefaction,
however, has hardly been studied.
Questions that must still be addressed
here are how large the dynamic loads
(vibrations and changes in pore water
pressure) induced by e.g. jet grouting are
and whether they are big enough to cause
liquefaction, cyclic softening, or even
trigger further consolidation.

2. SOIL PARAMETERS FOR LIQUE-

FACTION CRITERIA
To produce a comparison of the two
different liquefaction criteria, soil samples
from five different locations in Austria were
tested in the laboratory and necessary
parameters like grain size distribution,
liquid limit (LL) and the plasticity index (PI)
were determined. The main similarity of
these samples is that they are all
postglacial lacustrine still water sediments.
Due to the process of sedimentation, this
type of soil is laminated (Figure 1). The
thickness of these layers varies between
millimetres and centimetres, and the grain
size of the composition can be between
fine sand and clay. The main grain-size
fraction, however, is usually silt (Figure 2).
372
Figure 2. Grain size distribution
3. DETERMINING ACCELERATION
AND PORE WATER PRESSURE
DURING JET GROUTING
To determine the acceleration and the
change in pore water pressure during jet
grouting, two tests were conducted. The
first test was performed in a 2.70 x 2.70 m
chamber in the laboratory concurrently to
a different research project dealing with jet
grouting. The second series was a field
test in St. Kanzian/Carinthia.
3.1. Measurement instrumentation
The measurement of the vibration was
performed by two 3D 2g acceleration
sensors (ACS) for high frequencies of up
to 400 Hz. They are highly shock resistant
and the casing (32 x 32 x 25 mm) is
nitrogen dampened and hermetically
sealed.The measurement of the changes
of the pore water pressure was performed
with two Glötzl push-in pore water
pressure sensors (PWS). The
measurement range of the sensors is
between 0 and 10 bars.For data logging, a
universal data logger and a laptop were
used. The sampling rate was set to 400
Hz.
3.2. Laboratory test jet grouting
The laboratory test was performed by
Keller Grundbau GmbH in an approx.
2.70 x 2.70 x 2.45m chamber (Figure 3).
As filling, material sand was used and the
jet parameters were 150 bar, 16 rpm with
a pull speed of z = 3.5 cm/sec. The double
tube system was used. The diameter of
the jet grout column was projected with
1.0 m. After the installation of the sand
including the sensors on two levels, the
chamber was flooded and the sand
saturated. The sensors had a distance to
the centre of the jet grout column of 0.75
m and 0.25 m to the outline.
Sand
2.45 m PWS ACS
ACS PWS
Cross section Jet grouting column
Test chamber
0.75 m
2.70 m
0.50 m
2.70 m
Figure 3. Sketch test chamber
373
3.3. Field test
The field test was located in St. Kanzian in
the southern part of Austria. In addition to
some jet parameter tests for a huge
infrastructure project, it was possible to
perform some in situ vibration and pore
water pressure measurements. Therefore,
the equipment used in the laboratory test
was transferred to the testing field with the
exception that only one PWS was used.
The sensors were placed between the
drilling points as shown in the following
graphic at a depth of approx. 3.0 m by pre-
drilling and using push-in rods that were
withdrawn after the installation of the
sensors. A double tube system was used
with a maximum pressure of 400 bars, 48
rpm and a variation of pull speed z. The
starting point for jet grouting was -8.5 m
and the end point -2.5 m below ground
level. All columns were pre-cut so the
whole process included drilling, pre-cutting
and jet grouting.
KPSU-2 PSU-4a
KPSU-4 KPSU-5
Figure 4. Layout field test St. Kanzian
4. RESULTS
4.1. Comparison of liquefaction
susceptibility criteria
The results of the laboratory tests
compared against the “Chinese criteria”
chart show that only a few soils are
susceptible to liquefaction. The range of
the liquid limits values is between 26 %
and 43.5 %. In detail, several samples
from the two locations (Schüttdorf and St.
Kanzian) using this criterion have the
potential to liquefy. Some results are
close, but most of the soil specimens
(Dant and LBH Salzburg) are in the “not
susceptible” range. Based on the results
of St. Kanzian, the influence of the grain
size distribution can be shown. The
difference in the grain size distribution
(Figure 2) leads to a significant shift to the
right, the not susceptible part of the
Chinese criteria chart.
Not susceptible
Susceptible to
liquefaction
Figure 5. Graph Chinese criteria (Wang, 1979)
The same specimens are presented in
the following chart using the liquefaction
screening criteria (Figure 6). Most of the
specimens are in the region of clay-like
behaviour. Only two samples are in the
transition zone. The samples that would
potentially liquefy according to the
previous criteria are in the sand-like
behaviour area. The graph, extended with
Zones A and B (shaded part) taken from
the liquefaction screening criteria of Seed
et al. (2003), shows that 40 % of the
investigated specimens are potentially
liquefiable and 60 % should undergo
further testing if the water content wc •
0.85 of the liquid limit.
Clay-like behaviour
Zone B
Zone A Transition zone
Sand-like behaviour
Figure 6. Graph liquefaction susceptibility
criteria (Boulanger & Idriss, 2006)
374
4.2. Laboratory test jet grouting
The experiment in the chamber showed a
change in the pore water pressure ¨p of
up to 0.044 bars (Figure 7). The drilling
process caused this increase from a
starting level (0.0 bars) to 0.01 bars. At the
2300 second mark, the jet grouting began
and induced an increase of the pore water
pressure. PWS 1 has a sharper increase
and reaches the highest value of 0.043
bars after 2370 seconds, and decreases
slightly until the end of the jet grouting
procedure after 2425 seconds. The graph
of PWS 2 shows a slightly more moderate
but steady increase to a maximum of
0.044 bars. Both graphs decrease after
the jet grouting has stopped. They also
show a wave-like course during jet
grouting. The waves have a frequency of
0.27 Hz and show the passing of the
nozzle exactly (as seen in Figure 7 below).
Figure 7. Lab test Pore water pressure [bar]
The following graph (Figure 8) presents a
section (2325 to 2405 sec.) of the PWS 1
values. The difference between maximum
and minimum of each wave is approx.
0.0015 bars and gets smaller towards the
end.
Figure 8. Pore water pressure PWS 1, detail
The measurement of the acceleration also
showed the passing of the nozzle, but the
background noise was high and the
values (± 0.008 g) very low.
4.3. Field test jet grouting
In this paper, the results of acceleration
and pore water pressure measurements of
two jet grouting columns (KPSU-5 and
KPSU-2) are presented. The results of the
jet grouting column KPSU-5 show the two
process steps, the pre-cutting approx. 990
to 1350 seconds and the jet grouting from
approx. 1520 to 1860 seconds. The pore
water pressure increases in the first step
from 0.232 to 0.244 bars and only
increases slightly until the next step. The
maximum was 0.250 bars. The final
difference of pore water pressure is 0.018
bars as seen in Figure 9.

Figure 10. Acceleration KPSU-5, X-, Y- and Z-

axes
The pore water pressure increases during
the production of the column KPSU-2.
This is partially significant as it coincides
with the start of pre-cutting at the 1200
Figure 9. Pore water pressure KPSU-5
sec. mark. The pore water pressure rises
from approx. 0.224 to 0.248 bars. After a
In the following charts (Figure 10), the decrease to a constant level of 0.242 bars,
accelerations of ACS 1 and 2 in all three the jet grouting phase increases the pore
axes are presented. The maximum values water pressure to a maximum value of
of ACS 1 are 0.020, 0.011 and 0.017 g (X- 0.251 bars. The final difference of pore
, Y- and Z-direction). The values are water pressure is 0.027 bars (Figure 11).
slightly higher during jet grouting
compared to the pre-cutting phase. The
smaller distance of ACS 2 to the column
results in higher accelerations. The
maximum values of ACS 2 are 0.200,
0.211 and 0.080 g. It is obvious that the
maximum acceleration during jet grouting
was reached when the nozzle was at the
level of the senor. The vertical
acceleration (X-direction) is almost as high
as the acceleration parallel to the jet. The Figure 11. Pore water pressure KPSU-2
acceleration orthogonal (Z-direction) had The accelerations in all directions were
the lowest value. lower than those measured for KPSU-5.
There is only a slight difference in the X-
and the Z-axis of ACS 1 and ACS 2. The
375
maximum values for ACS 1 are 0.090,
0.030 and 0.049 g and for ACS 2 0.041,
0.019 and 0.020 g.
Figure 12. Acceleration KPSU-2, X-, Y- and Z-
axes
5. CONCLUSION
All of the soils tested here have shown
their potential to liquefy or undergo cyclic
softening. Due to the lamination of these
soils, it is also possible that only certain
layers liquefy or particles rearrange. The
input of vibrations and/or excess pore
water pressure during jet grouting is not
significant, as is shown in the presented
results. In some cases, it was still enough
to induce settlements not only in an area
next to the jet grouting column, but also up
to more than 10 meters away. The trigger
for these settlements still has to be
identified, especially when high clay
contents (up to 30 %) contradict the
theses of liquefaction. It also has to be
considered that the mineralogical
composition may have a major influence
on the soil behaviour.
376
ACKNOWLEDGEMENTS
The author wishes to thank Keller
Grundbau GmbH, in particular Dr.
Clemens Kummerer for the support and
the given opportunity to perform these
measurements.
REFERENCES
Boulanger, R.W. & Idriss, I.M. (2006)
“Liquefaction susceptibility criteria for
Silts and Clays”, Journal of
Geotechnical and Geoenvironmental
Engineering, 1413 – 1426
Seed, R.B., Cetin, K.O. et al. (2003)
“Recent Advances in Soil Liquefaction
Engineering: A Unified and Consistent
Framework”, Earthquake engineering
research centre, Berkeley, USA
Robertson, P.K., Fear, C.E. (1995)
“Liquefaction of sands and its
evaluation”, Proceedings of the 1st
International Conference on
Earthquake Geotechnical Engineering,
Tokyo, Japan
Wang, W.S. (1979) “Some Findings in Soil
Liquefaction. Water Conservancy and
Hydroelectric Power”, Scientific
Research Institute, Beijing, China
Wong, L.W., Ju, D.H., Wu, P.J. (1997)
“Control of ground movements caused
by jet grouting”, Proceedings of
International Conference on Ground
Improvement Techniques, 649 -656,
Macau, China
 25th European Young Geotechnical Engineers Conference
21st – 24th of June 2016, Sibiu, Romania

Prevention of soil liquefaction using stone columns

Jamal HLEIBIEH1*
1
Technische Universität Dresden, Dresden, GERMANY

ABSTRACT
In order to prevent soil liquefaction there are various ground improvement measures such as the
soil cementation, soil compaction or the use of drainage. Although stone columns were frequently
used for the prevention of soil liquefaction in the last few years, there is not much research that
explains how it works. Both numerical calculations with conventional constitutive models and small
scale laboratory experiments cannot represent the improving mechanisms. The interaction
between the soil and the columns and between the pore water and the soil skeleton under seismic
effects makes the understanding of this method difficult. In this work, the application of stone
columns as a prevention to soil liquefaction is numerically investigated with a hypoplastic
constitutive model. The focus of the study is on the influence of stiffness and permeability of the
columns. Furthermore, the effects of columns group was studied. Finally, it was proved that the
columns installation should be considered in the initial state of the numerical model.

Keywords: Soil liquefaction; Earthquake; Stone column; Hypoplasticity

compaction or the use of drainage. Soil

1. INTRODUCTION
Earthquakes cause enormous economic
and humanitarian damage, especially if
the earthquake cause soil liquefaction.
Soil liquefaction is the loss of soil stiffness
and shear resistance due to the build-up
of pore water pressure and the decrease
of effective stress. This loss of soil
stiffness and strength appears as ground
failures in the form of sand boils,
differential settlements, flow slides, lateral
spreading, and loss of bearing capacity
beneath buildings [Adalier et al. 2004]. In
order to prevent soil liquefaction there are
various ground improvement measures
such as the soil cementation, soil
* presenting author
compaction using deep vibrator has a high
effectiveness in cohesionless soils.
However, this effectiveness decreases
rapidly with an increase of the fine
fraction. In this case, the soil improvement
can be achieved using stone columns. As
an effective method for reducing the risk of
liquefaction the vibro replacement was
used frequently in the last years. The
advantage of this method is the
combination of a drainage effect, soil
compaction and a rigid support which
prevent the liquefaction.
The high permeability of the stone
columns prevents the build-up of pore
water pressure (pwp) in the columns. This
377
causes a pressure difference between the
columns and the surrounding soil. The
pressure difference produces water flow in
the direction of columns, and thus a
decrease in the pwp in the soil. Since the
permeability of sand and gravel is greater
in the horizontal direction than in the
vertical one, the benefit of the stone
columns is not only the shortening of the
drainage path, but also the change in the
direction of water flow to the more
permeable direction. Furthermore, the
installation of stone columns causes a
compaction of the surrounding soil, which
also reduces the risk of liquefaction
[Madhav et al. 2008]. In addition, the
stone columns are much stiffer than the
surrounded soil. Therefore, the
earthquake-induced or already existing
loads are redistributed towards the
columns. This means that the earthquake
produces smaller stress changes in the
soil [Priebe 1990].
Although stone columns were often
used for the prevention of soil liquefaction
in recent years, there is not much
research explaining how they work. Both
numerical calculations with conventional
constitutive models as well as small-scale
laboratory tests cannot reproduce the
mechanisms of preventing soil liquefaction
using stone columns. In the numerical
models, it is difficult to simulate the
column preparation and the resulting soil
compaction and stress redistribution.
Furthermore, the interaction between the
soil and the columns cannot be simulated
with simple constitutive models. In
addition, the dynamic coupled calculations
are extremely costly in terms of
computation time and computer capacity.
In small-scale laboratory experiments, it is
again impossible to reproduce realistic
stresses and drainage path.
In this paper numerical calculations
using a hypoplastic constitutive model are
performed to investigate the effects of
stone columns as a countermeasure
against soil liquefaction. The applied
hypoplastic constitutive equation
corresponds to the model by [Von
Wolffersdorff 1996] with the intergranular
378
strain extension according to [Niemunis
and Herle 1997].
2. NUMERICAL MODEL
The finite element program TOCHNOG
PROFESSIONAL [Roddeman 2013] was
applied for the numerical simulation (2D
model). The plane strain mesh was
generated with constant strain triangular
elements using the software GiD [GiD
2013]. The boundary conditions were
defined so that an unrealistic wave
reflection was prevented at the model
boundaries. The left edge was connected
to the right edge in order to obtain the
same velocity at the same height of the
vertical boundaries. The bottom boundary
was fixed in the vertical direction, while no
kinematic constraints were prescribed at
the top boundary.
The ground model consists of a loosely
deposited sand layer with two soil
improvement columns at the lateral edges.
The model size is 10 * 10 m and the water
table is 0.5 m below the ground surface
(Figure 1). In the initial state, geostatic
stresses were prescribed and activated
with start of the calculation. In addition, a
constant initial void ratio was considered
for the stone columns regardless of the
depth. After that, an idealized seismic load
was applied as a sinusoidal horizontal
velocity at the bottom of the model. The
velocity signal has a frequency of 2 Hz
and an amplitude up to 0.1 m/s. The
acceleration generated by the velocity
signal can be determined by differentiating
the velocity time history. For example the
differentiation of the velocity history with
an amplitude 0.1 m/s and a frequency of 2
Hz yields an acceleration with an
amplitude of approximately 1.25 m/s2 and
an unchanged frequency.
The numerical calculations have been
conducted using the hypoplastic model
from [Von Wolffersdorff 1996], linked with
the intergranular strain approach
according to [Niemunis and Herle, 1997].
The hypoplastic parameters of Leighton
Buzzard sand have been used,
determined from cyclic and dynamic
 laboratory tests. Detailed description of
the determination of the parameters is
presented in [Hleibieh et al. 2014]. The
parameters for the hypoplastic model
according to [Von Wolffersdorff 1996] and
the parameters for the description of the
intergranular strain model are listed in
Tables 1 and 2.
Figure 1. Geometry of the numerical model
Table 1. Parameters for the constitutive
model [Von Wolffersdorff 1996].
ijc hs0 n0 ed0 ec0 ei0 Į ȕ
[º] [MPa]
33 100 0.6 0.513 1.014 1.17 0.1 1.0
Table 2. Additional parameters for the
intergranular strains after [Niemunis and
Herle, 1997].
R mR mT ȕr Ȥ were simulated having the same
0.0003 10.0 4.0 0.3 1.0
3. CALCULATIONS OF A SAND LAYER
WITHOUT COLUMNS
Fig. 2 shows the development of the pore
water pressure within the sand in
dependence on the seismic events
(different amplitudes of the velocity).
These results refer to a point in the middle
of the model in 5 m depth. In these
379
calculations, the permeability of the sand
amounts 5*10-4 m/s. The larger the
amplitude of the velocity, the greater the
generated pwp. If the amplitude of the
velocity achieves the value 0.08 m/s or
more, the generated pore water pressure
rises strongly and finally produces a non-
convergency of the calculation. On the
contrary, as long as the amplitude of the
velocity remains less than 0.08 m/s, the
pwp reaches an asymptotic value after a
slight peak.
Figure 2. Pore water pressures generated at
different velocity amplitudes
4. CALCULATIONS OF A SAND LAYER
WITH STONE COLUMNS
The influence of the soil improvement
columns for the prevention of soil
liquefaction was investigated using two
improvement columns at the lateral edges
of the model (Figure 1). The influences of
the column stiffness and of the column
permeability, respectively, were
investigated separately. First, the columns
permeability as the sand but a higher
stiffness (the shear modulus G0 for the
columns was 7 times higher than G0 for
the sand). Afterwards, the columns were
simulated having the same stiffness as the
sand but a higher permeability.
Permeability of the columns was 100
times greater than the permeability of the
sand.
The development of the pore water
pressure in the sand with and without
columns is presented in Figure 3. The
permeability of the sand is 5*10-4 m/s and
the velocity amplitude is 0.1 m/s. The
results show, that the liquefaction is
prevented by both an increase in stiffness
as well as an increase in permeability.
Figure 3. Pore water pressures in sand for the
configuration with and without columns
The drain effect of the columns with a
high permeability is illustrated in Figure 4.
Within the sand the water flows
horizontally in the direction of the columns
and vertically towards the ground surface.
The horizontal flow dominates in the
deeper areas of the sand, while the
vertical flow is observed in the near-
surface areas. In the columns the water
flows only towards the ground surface.
The velocity of the water flow in the
columns is much larger than in the sand
because of the increased permeability of
the columns.
Figure 4. Water flow within the sand and the
columns for the calculation with permeable
columns
380
The horizontal flow of water in the
direction of the columns prevents a strong
build-up of excess pore water pressure,
preventing thus liquefaction. The stiff
columns increase the stiffness of the
entire system, consisting of the sand and
the columns. Due to the high column
stiffness, a seismic event produces less
shear strain in the sand than in case of an
unimproved soil. In turn, the smaller shear
strains lead to lower excess pore water
pressures and thus reduce the liquefaction
risk, too.
5. GROUP OF COLUMNS WITH A
REALISTIC GEOMETRY
In the already introduced calculations the
column diameter and the distance
between the columns was selected to be
very large to illustrate the drainage effect
and the stiffness effect. In order to
examine the effects of a group of stone
columns, a modified model with a realistic
geometry was used. The geometry of the
modified model is shown in Figure 5. On
the lateral edges only half columns were
simulated due to the symmetry (periodicity
of the area). The model corresponds to
sand with stone columns, each having a
diameter of 0.6 m and a distance of 2 m.
The seismic loading is applied as a
sinusoidal acceleration of a frequency of 1
Hz and a maximum amplitude of 3 m/s2.
The permeability is 1*10-4 m/s for the sand
and 1*10-2 m/s for the stone columns.
Figure 6 presents the evolution of
excess pwp in the sand with and without
columns for different acceleration
amplitudes (point A in Figure 5). In sand
without columns the pwp increases with
increased number of cycles up to 95% of
the geostatistical vertical stress already
after 6 cycles with the acceleration
amplitude of 1.5 m/s2. In the calculations
with stone columns, the pwp in the sand
reaches about 50% of the geostatic
vertical stress after several cycles (3 to 4)
and then remains constant, regardless of
the acceleration amplitude. Nevertheless,
Figure 7 shows that the effective vertical
stress at the point A decreases for various
accelerations down to the value zero.
Thus the effective stress decreases,
although the PWD is not increasing
analogously, which results in a decrease
of the total
stress.
Figure 5. Geometry of the model with a
columns group
Figure 6. Pore water pressure in the sand
(point A) at different acceleration amplitudes
for the model with and without columns
A decrease in the total vertical stress in
the sand is also observed in Figure 8. At
the same time, the total stress in the
columns increases (Figure 9).
Due to the higher stiffness of the stone
columns in comparison to the surrounding
sand a redistribution of the vertical
stresses occurs in the direction of the
columns. A portion of the own weight of
381
the sand is then carried by the columns
and and can be interpreted as a Siloeffekt.
This leads in
Figure 7. Effective vertical stress in the sand
(point A) for different acceleration amplitudes
Figure 8. Total vertical stress in the sand
(point A) at an acceleration amplitude of
2 m/s2
Figure 9. Total vertical stress in the columns
(point B) at an acceleration amplitude of
2 m/s2
deeper layers to a decrease of the total
stresses in the sand and to an increase of
the total stresses in the columns. This
stress redistribution is realistic. However,
one can assume that it takes place International Conference GEDMAR08,
already during the installation of the stone Nanjing, China. pp 66-79.
columns [Priebe 1995]. Therefore, the Niemunis, A. & Herle, I., (1997)
consideration of the columns installation “Hypoplastic model for cohesionless
and the resulting stress redistribution in soils with elastic strain range”,
the initial state of the numerical model play Mechanics of Cohesive-Frictional
a decisive role in the numerical Materials Vol. 2, Nr. 4, S. 279–299.
investigation of the stone columns for the Priebe H.J. (1990) “The prevention of
prevention of soil liquefaction. liquefaction by vibro replacement”,
Proceedings of Eathquake Resistance
6. CONCLUSIONS
Construction and Design, Berlin,
In this paper the build-up of excess pore Germany.
water pressures due to seismic events in a Priebe H.J. (1995) “The design of vibro
liquefiable soil was investigated using replacement”, Ground Engineering 28
numerical simulations by means of a (10). pp 31-37.
hypoplastic constitutive model with
Roddeman D. (2013) “TOCHNOG
intergranular strains. The liquefaction
PROFESSIONAL User's manual”,
hazard increases with an increased
Version 14. http://www.feat.nl/
magnitude of the seismic loading. The risk
of liquefaction can be reduced using a soil Von Wolffersdorff, A. (1996) “A
improvement by stone columns. The hypoplastic relation for granular
columns act against the liquefaction by materials with a predefined limit state
both their high stiffness as well as their surface”, Mechanics of cohesive-
high permeability. The numerical frictional materials, Vol. 1, Issue 3,
calculations with advanced constitutive pp.251-271.
models can be used successfully to
investigate the effects of stone columns.
However, the installation of the columns
should be considered in the generation of
the initial state of the model in order to get
realistic stress distribution.

REFERENCES
Adalier, K. & Elgamal, A (2004) “Mitigation
of liquefaction and associated ground
deformations by stone columns”,
Engineering Geology 72, 275 – 291.
GiD. (2013): “Pre and post processing
system for Numerical Simulations”,
International Center For Numerical
Methods In Engineering (CIMNE).
http://www.gidhome.com/
Hleibieh, J., Wegener, D. & Herle, Ivo.
(2014) Numerical simulation of RRTT
with a hypoplastic model, Acta
Geotechnica Vol. 9, Issue 4, pp 631-
640.
Madhav M. & R., Murali A. (2008)
“Liquefaction Mitigation of Sand
Deposits by Granular Piles- an
Overview”, Proceedings of the 2nd
382
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Application of Offshore Seismic CPT data for Soil Stiffness

Interpretation

Olsi KORETA1*, Lindita KELLEZI1

1
Geo, Copenhagen,DENMARK

ABSTRACT
One of the most important dynamic soil parameters is the shear wave velocity (Vs), used mostly for
calculating the shear modulus (G), based on a simple elastic relationship with the mass density.
The methods to derive the shear wave velocity include direct measurement by use of in-situ
geophysical tests (seismic CPT, PS Logging, etc.), laboratory tests, or by using empirical and site-
specific correlations from cone penetration tests (CPT). For the detailed design of an Offshore
Wind Farm (OWF), a site investigation has been conducted using static and seismic CPT, the later
for the measurement of the shear wave velocity. In addition, PS logging (down-the-hole acoustic
probe) and several laboratory bender element (BE) tests have been carried out, enabling an
evaluation and application of the site-specific CPT-Vs correlation. The application of the in-situ
methods will be discussed with regards to the limitations, and important recommendations will be
given to overcome the challenges during offshore soil investigation.

Keywords: seismic CPT, shear modulus (G), shear wave velocity (Vs)

Poisson’s ratio (ȝ), Young’s modulus (E),

1. INTRODUCTION
Estimation of the shear wave velocity (Vs)
through different measurements or
correlations, is an important component of
various site response analyses and soil-
structure interaction. The compression
wave velocity (Vp) and Vs, have been used
to describe the elasticity of the soils, to
predict the soil dynamic response, due to
earthquake or other vibrations. The
importance in the accuracy of estimating
the Vs, is related directly with the
equations from elasticity theory, applied
for calculating the shear modulus (Gmax),
* presenting author
etc.
In this paper, several aspects for the
application of offshore seismic CPTs for
soil stiffness interpretation, are discussed.
For the detailed design of an Offshore
Wind Farm (OWF), full interpretation of the
seismic CPT, including estimation of shear
wave velocity and derivation of the shear
modulus have been carried out. A detailed
assessment of the existing site-specific
correlations, based on the measured Vs
from seismic CPT and other in-situ or
laboratory tests, has been performed.
2. SITE DATA (IN-SITU AND
LABORATORY MEASUREMENTS)
2.1. In-Situ Data
As part of an offshore site investigation
campaign, different in-situ and laboratory
tests have been carried out. For the
considered location, down-the-hole CPTs
(DTH-CPTs), seismic CPTs and several
laboratory tests (classification and
advanced) have been performed.
The chosen CPT (CPT-1) and the
available soil density measurements are
shown in Figure 1.
Figure 1. Representative CPT (CPT-1) data
and soil density measurements
Filtering of the CPT data was
considered necessary in order to remove
pick measurements.
The shallow geology of the area is
interpreted to comprise of three main
units. The first Holocene unit has been
seen with a thickness of maximum 4.5 m
below seabed and comprises of very soft,
low strength silty and gravelly organic
clay. The second unit comprises
Pleistocene sediments and consist of
glacial deposits (glacial clay, sand, gravel
and till). Both Holocene and Pleistocene
are Quaternary sediments. A geological
384
description of the units, together with the
depths where Vs measurements have
been performed, is shown in Figure 2.
Figure 2. Geological descriptions of the units
2.2. Normalized Soil Behaviour Type
(SBTn) chart
Due to increasingeffective overburden
stress with depth, the CPT data requires
normalization. The normalized chart
(Robertson, 2010) provides a more
accurate identification of the soil type,
than the non-normalized one(Robertson,
2009). However, a slight differenceis
expected for the presented CPT, with
maximum in-situ vertical effective stress of
about 150 kPa.
Soil classification based on the
interpreted SBT according to (Robertson,
2010)has been performed and presented
in Figure 3 and Figure 4. The charts have
been created by using a self-developed
visual basic form implemented in excel.
The chart proposed by (Robertson,
2010)is able to identify general trends in
ground response, such as, over
consolidation ratio (OCR), age and
cementation for sandy soils, increasing
stress history and soil sensitivity (St) for
cohesive soils (Robertson & Cabal, also an important role in the transmission
2015).All the above-mentioned of shear wave velocity (Vs)
parameters play

Figure 3. Normalized SBTnIndex (IC) – CPT-1

3. DETERMINATION OF SMALL
STRAIN SHEAR MODULUS (GMAX)
USING VS
The determination of the soil stiffness
parameters needs a very careful
consideration of the associated level of
strains. Various geophysical methods and
laboratory testing could be used to
measure the shear wave velocity. In all the
cases, it is important to highlight the fact
that G is highly dependent upon the strain
level.
All the geophysical methods are testing
techniques, which provide low strain in-
situ compression and shear wave velocity
measurements. At low strain levels (less
than about 10-4 %) the shear modulus in
soil is constant at its maximum value
Gmax(Andrus, et al., 2007).The shear
response for different strain levels could
be estimated by using laboratory tests or
Figure 4. Normalized SBTn chart for CPT-1
different published theoretical degradation
curves of the ratio G/Gmax.The shear
modulus (Gmax) is a parameter calculated

385
based on the Vs, using a simple elastic wave module (Geo SBF-Hammer) has
relationship as given in Equation (1). been used. Geo’s Vs module for seismic
down-the-hole SCPT (DTH-SCPT) is an
integrated part of the seabed frame (see
(1) Figure 5). The seismic “hammer” (Figure
6) is controlled and operated by the CPT
where soil density ȡ is the total unit operator on-board of the drilling vessel.
weight of soil divided by gravity (9.81
m/sec2). Gmax has units of force per length
squared andcan also be measured in the
laboratory by using resonant column or
bender element tests.

4. MEASUREMENT OF VS
Vscould be measured by using different in-
situ geophysical methods as well as by
laboratory tests. The in-situ methods are
divided into invasive and non-
invasive.(Wair, et al., 2012). Invasive
methods require drilling on the ground and Figure 5. The Seabed Frame (the seismic
include different downhole and cross-hole hammer fits in the cavity to the right)
logging, suspension logging, etc. A very
rapid and cost effective invasive method is
the seismic CPT (SCPT). Non-invasive
methods do not require drilling or
penetration into the ground. These
methods include different seismic
refraction and spectral analysis of surface
waves (SASW).
Same as for the other types of
geotechnical laboratory tests, also for the
measurement of the Vs, high quality
undisturbed samples are required. This
quality of the testing is very often not
possible to be achieved, especially for
sandy soils.
In this paper, only the measurement of
the Vs by use of SCPT is discussed. The Figure 6. Geo’s seismic hammer unit
interpretation and analysis of the datais In combination with Geo’s fast
basedon an offshore SCPT measurement. operating DTH-CPT system, the Geo SBF
In addition, some results of other invasive – Hammer provides essential data for
methods such as PS suspension log different offshore site investigations for
probe and bender element test have been wind farms,jack-up rig installations,
used. platforms, subsea structures and bridges.
4.1. Seismic CPT (SCPT) 4.2. Data acquisition system
The SCPT measurements have been The data acquisition system used is A.P.
carried out in conjunction with an offshore Van den Bergs Icone and the Vs
drilling campaign. For the measurements measurements were recorded using a
of the Vs , Geo’s state of-the art shear seismic piezocone. With the Icone seismic
386
module, the digital cone is turned into a
seismic or SCPT cone.
The standard Icone seismic module
contains three accelerometers to receive
shear waves(left, right and compression
waves). The main principle of the SCPT
testing consists of recording shear waves
at a known depth, below the source
(seismic hammer). First, the shear waves
are generated by a driven spring hammer
mounted on the seabed frame. These
shear waves are later received by a
geophone incorporated in the
conventional piezocone penetrometer.
The SCPT is pushed down to the
required depth and then the hammer is
activated to generate shear waves. All the
signals received on the geophone are
monitored by a seismograph. For the
required testing depth, the distance
between the source and the geophone is
a known parameter. The difference in
distance between the geophone and the
source is calculated for different testing
depths. This distance, divided by the
difference in travel time for each depth
gives the shear wave velocity at that depth
interval.
4.3. Processing shear wave velocity
signals
The commercial software SPAS 2009
v.2.0.2.69 (Seismic Processing and
Analysis of Signals)(SPAS 2009 v.2.0 -
GeoLogismiki, 2009) has been used to
process and analyse the measured
signals.
After importing the signals, the first step
in the processing of the data is to check
for signals repeatability. The program
allows to select multiple signals from the
same depth. In this way is possible to
identify “bad” signals and easily remove
these by unchecking.
The quality of the signal is primarily
affected by the efficiency of the source
hammer, the energy level, and also from
the noise generated by the seabed drilling
system. The ambient noise that comes
from the drilling system should be
387
minimized in order to increase the signal-
to-noise ratio (SNR). Considering the
different challenges that come during an
offshore site investigation, it is impossible
to eliminate all the sources of background
noise. In order to increase the quality of
the signal data, digital filtering techniques
have to be used. However, it is to be
mentioned that the filtering technique,
compared with other ways of improving
signal quality, is the last means of
improvement. The filtering technique is
based on cutting-off the frequency filters
(Figure 7 shows the filtering technique
applied to one SCPT measurement at 3.0
m depth).
Figure 7. Signal filtering using SPAS 2009,
SCPT measurement at 3.0 m depth.
The most important aspect in applying
the digital filteringtechnique, is to identify
the major dominant noise frequencies. A
wide band-pass filter covering the range of
predominant frequencyis
recommended(e.g. 10 to 50 Hz)(Nguyen,
et al., 2015).
The measured data have been filtered
and processed in order to estimate the
shear wave velocity. The depths where
SCPT measurements have been
successfully performed, are shown in
Figure 2.
4.4. Bender Element Test
In addition to SCPT, bender element tests
(BE) have been carried out in laboratory in
order to measure Gmax. The depths of the data such as: geological age, deposits,
tested samples are shown in Figure 2. OCR, test locations, etc.
4.5. PS Logging The main CPT parameters used for
The PS logger is a high energy, low correlations include:depth (D), tip
frequency acoustic probe, designed to resistance (qc), sleeve friction (fs), soil
measure compression and shear wave behaviour type index (IC), effective stress
velocities in fluid filled boreholes. PS (ı’v), etc. The published CPT-Vs
logging have been carried out at the correlations are generally developed for
investigated borehole at depths starting specific soils types (i.e., “Sand” or “Clay”),
from 14 m below seabed. For this study, but also for “All soils”. According to (Wair,
only the measurement of PS logging from et al., 2012), when comparing the
correlations developed for specific soils
14 to 15 m will be considered.
with the correlationsdeveloped for “All
4.6. Vs measurements soils” it results that statistically both
The results of Vs measurements from methods perform similarly. However, the
SCPT, BE and PS logging, together with soil type-specific method under-predicted
the corresponding depths are given in Vs by 8% and “All soils” method under-
Table 1. predicted by 3%. In addition, the soil type-
specific method produced spikes in the Vs
profile, at transition soils between layers.
Table 1 Vs measurements
As a result, the “All soils” method is
Method Method Method considered more applicable(Wair, et al.,
SCPT BE PS 2012).
D Vs D Vs D Vs This study focuses ononly three most
[m] [m/s] [m] [m/s] [m] [m/s] recent CPT-Vs correlations, developed for
“All soils” and based on a relatively
2.0 173 - - - - independent and large numbers of data
3.0 173 - - - - pairs.Also, only the correlations developed
for Holocene and Pleistocene age soils
5.4 321 5.5 293 - -
have been considered.The correlations
6.4 321 - - - - used have been presented
7.0 294 - - - - from(Robertson, 2009), (Mayne, 2007)
and (Andrus, et al., 2007).
8.0 294 8.5 306 - -
5.2. (Robertson, 2009)
9.0 296 - - - -
(Robertson, 2009) developed a general
10.0 313 10.5 333 - - correlation CPT-Vs based on 1035 data
- - - - 14.0 284 pairs, from Holocene and Pleistocene soil
sites. The general expression of Vsis given
- - - - 15.0 285
in Equation (2).

5. REVIEW OF VS CORRELATIONS (2)

WITH CPT
5.1. General
Several authors and researchers have
studied and published relationships
between CPT and Vs. The correlations
have been developed based on different
Vs measurements and considering other
whereqt is the corrected cone tip
resistance,ıvo is the in-situ vertical stress,
pa is the atmospheric pressure and Icis the
SBTindex. The SBTn zones have been
calculated for each layer, as presented in
Figure 4.

388
5.3. (Mayne, 2007)
Initially, (Mayne, 2006), proposed an “All
soils” correlation between Vs, CPT and fs
based on regression of a large dataset
from different sites.
(Mayne, 2007) updated the correlation
by considering the logarithm of fs, rather
than the natural logarithm that was
proposed originally in (Mayne, 2006). The
general form of the correlation is given in
Equation (3).
(3)
to 0.9 and for the Pleistocene layer was
chosen equal to 1.1.
6. INTERPRETATION OF THE
RESULTS
The considered correlations have been
applied to the available CPT data in order
to estimate the Vs. The calculated Vs for
each correlation was compared to the
measured Vs values, determined from the
SCPT, BE and PS logging. A general
overview, showing the comparison
between the measured and the correlated
Vs, is given inFigure 8.

The correlationwas derived from a

database that included sands, silts, clays,
as well as mixed soil types.
5.4. (Andrus, et al., 2007)
(Andrus, et al., 2007) proposed a
correlation applicable to all soils and
based on a dataset of 229 CPTs and Vs
measurements. The dataset included 72
data soils of Holocene geologic age, 113
data of Pleistocene and 44 data of Tertiary
age. The majority of the Vs measurements
were performed using the SCPT.
For the regression analysis, (Andrus, et
al., 2007) has taken into account several
previous publications and only the ones
with lower standard deviation of the
residuals have been used. Based on
thecombined Holocene and Pleistocene
dataset, the best-fit regression equation
for predicting Vs in m/s is given from
Equation (4).

(4)
Figure 8. Comparison between measured and
estimated Vs profiles.
whereSF is a scaling factor that takes
into account the reference age for the As seen from Figure 8, the general
combined Holocene and Pleistocene data. trend of the Vs, suggested by the
measured values, has been captured by
For Holocene soils the value of SF the considered correlations. However,
varies from (0.88-0.92) and for differences in predicting the Vs are evident
Pleistocene soils from (1.11-1.12). In our andcan be seen through the depth.
calculations, for the Holocene layer (0.0 - Generally, (Robertson, 2009) and (Andrus,
4.5) the SF factor has been chosen equal et al., 2007) show a more similar
prediction, compared with (Mayne, 2007).
389
An under-estimation of the Vscould be
seen for the first layer from (Robertson,
2009) and (Mayne, 2007). This under-
estimation is expected and mentioned
also by (Robertson, 2009).However, at the
end of the CPT, these methods over-
estimate the predicted Vs.(Mayne,
2007)correlation looks visually closer to
the measured values and it has captured
almost all the trends suggested by the
measured data. A tentative explanation of
this could be the fact that (Mayne, 2007)
correlation follows the trend of the fs,
which in this case, comply well with the
measured data.
Figure 9. Performance of (Robertson, 2009)
In order to have a better quantification
correlation (SD=0.20, ȝ = 0.85)
of each correlation method,the Vsis plotted
in terms of the ratio between the estimated
Vs to the measured one. This ratio is
considered as Vs bias. Presented in this
way, a ratio equal to 1, would assume an
ideal correlation, where the estimated is
equal to the measured. Moreover, a ratio
less than 1,represents an under-
estimation of the Vs and a ratio larger than
1 represents an over-estimation.
The results for each method have been
plotted in charts and are shown in Figure
9for (Robertson, 2009), inFigure 10
for(Mayne, 2007) and inFigure 11 for
(Andrus, et al., 2007). For the available
dataset,(Robertson, 2009) correlation Figure 10. Performance of (Mayne, 2007)
under-estimates the Vswith a mean bias of correlation (SD=0.06, ȝ = 1.03)
ȝ=0.85 and corresponding coefficient of
variation COV=23%. (Andrus, et al., 2007)
correlation follows the same trend as
(Robertson, 2009), and has shown almost
the same performance (ȝ=0.89,
COV=24%).(Mayne, 2007) correlation,
presented in Figure 10, appears to be
most applicable for the available Vs
measurements. The performance of this
correlation show plotted data close to 1,
with a mean value of ȝ=1.03 and
COV=6%. However, the relatively low
number of the available data, does not
allow for a more definitive conclusion
regarding which method could be
considered more accurate for these Figure 11 Performance of (Andrus, et al.,
specific soil conditions. 2007) correlation (SD=0.21, ȝ = 0.89)

390
 Recommendations given from (Wair, et
al., 2012) suggest that for Quaternary soils
(Holocene and Pleistocene) the Vs could
be estimated by taking the averaged value
derived from (Robertson, 2009), (Andrus,
et al., 2007) and (Mayne, 2007). In all the
cases, where is possible to know the
ageing of the soil and the geology, it is
important to apply the ageing factors on
Holocene (0.88 – 0.92) and on
Pleistocene (1.11 – 1.12).For same soil
conditions with same CPT penetration
resistance, the Vsin Holocene could be
(22-26)% smaller than inPleistocene
deposits (Andrus, et al., 2007).
In the cases where it is not possible to
get samples or identify the subsoil
geology, the charts developed by
(Robertson & Cabal, 2015) could help to
evaluate the CPT data with regards of
estimating the soils deposits age.
In Figure 12is given an evaluation of
the normalized Vs for Holocene and
Pleistocene age soils, according to
(Robertson & Cabal, 2015).
Figure 12 Evaluation of Vsfor Holocene and
Pleistocene soils (Robertson & Cabal, 2015)
391
According to (Robertson & Cabal,
2015), younger deposits, such as
Holocene age soils, tend to plot towards
the centre and lower left of the SBTn
charts given in Figure 12 (first Holocene
layer is plotted in red triangles). Older
soils, such as Pleistocene, tend to plot
toward the upper, right part of the chart
(second Pleistocene layer is marked with
blue squares). These results highlight the
known fact that the shear wave velocity is
sensitive to age and cementation and
older deposits have higher shear wave
velocity than younger deposits.
The calculation of the small strain shear
modulus (Gmax)have been carried out
according to Equation (2). The correlation
of Vs given from (Mayne, 2007) and the
soil density measurements given in Figure
1,have been used. The measurement of
the Gmax from BE tests using the
measured Vs are also plotted.
Figure 13 Measured and estimated Gmax
7. CONCLUSIONS
Important aspects in measuring the Vs
from the SCPT tests, processing of the
data and discussing the application of
existing site specific Vs-CPT correlations,
are given in this paper.
A particular care must be given to the
SCPT measurements during the offshore
site investigation. The frequency of the
seismic hammer is an important
parameter that helps to filter out the
dominant noise frequencies.
An engineering judgement is needed
before the application of the available CPT
– Vscorrelations. The Vs is also very much
dependent on the deposits age of the soils
and the developed correlations in general
have been derived for soil data of a
specific soil deposits age, such as
Holocene or Pleistocene. For this reason,
the ageing scale factors must be applied
to the corresponding soil.
Measurement of the Vs by using SCPT
or other geophysical testing, including BE
tests in the laboratory, provides low strain
in-situ compression and shear wave
velocity measurements. The shear
modulus (G) is highly dependent upon the
strain level, and the determination of the
dynamic soil stiffness properties, needs a
very careful consideration of the
associated level of strains.
ACKNOWLEDGEMENTS
The authors are grateful to Albanian
Geotechnical Society for choosing this
paper and Geo for sponsoring the
participation in the conference.
REFERENCES
Andrus, R., Mohanan, N., Piratheepan, P.,
Ellis, B., & Holzer, T. (2007). Predicting
shear-wave velocity from cone
penetration resistance. Proceedings of
the 4th International Conference on
Earthquake Geotechnical Engineering.
Thessaloniki, Greece.
Mayne, P. (2006). In situ test calibrations
for evaluating soil parameters. Proc.,
392
Characterization and Engineering
Properties of Natural Soils II.
Singapore.
Mayne, P. (2007). Cone penetration
testing state-of-practice. NCHRP
Project 20-05.
Nguyen, H., Meidani, M., DeGroot, D.,
Lunne, T., & Vanneste, m. (2015).
Measurement and interpretation of
downhole seismic probe data for
estimating shear wave velocity in deep-
water environments. Frontiers in
Offshore Geotechnics III (ISFOG 2015).
Oslo: Taylor & Francis Group.
Robertson, P. (2009). Interpretation of
cone penetration tests - a unified
approach. Canadian Geotechnical
Journal, 46(11), 1337-1355.
Robertson, P. (2010). Soil Behaviour type
from the CPT: an update. 2nd
International Symposium on Cone
Penetration Testing, Vol.2, pp. 575-583.
Huntington Beach, CA.
Robertson, P., & Cabal, K. (2015). Guide
to Cone Penetration Testing for
Geotechnical Engineering, Prepared for
Gregg Drilling & Testing Inc. (6th
Edition ed.).
SPAS 2009 v.2.0 - GeoLogismiki. (2009).
SPAS 2009 User's Software Manual.
(Version 2.0). (GeoLogismiki, Ed.)
Greece.
Wair, B., DeJong, J., & Shantz, T. (2012).
Guidelines for Estimation of Shear
Wave Velocity Profiles. University of
California, California Department of
Transportation, Pacific Earthquake
Engineering Research Center.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Numerical analyses of the performance of seismically isolating

buried barriers

Valeria NAPPA1*
1
University of Napoli Federico II, Department of Civil, Architectural and Environmental
Engineering, Naples, ITALY

ABSTRACT
The installation of passive structural systems to protect existing buildings from seismic risk is very
common, yet it could be expensive and not always feasible in the case of valuable buildings. They
may affect the integrity of structures of artistic or historical relevance in a way that cannot be
accepted in the light of the most recent rules of conservation and retrofitting. As an alternative,
ground improvement techniques can be used to modify ground properties, in order to mitigate the
intensity of shaking at ground level. This paper focuses on the case of a V-shaped barrier that may
be formed by assembling inclined and partially overlapped columns of artificially softened soil to
isolate a mass in the ground. The model was calibrated using the results of centrifuge testing on a
reduced scale model. The response of a SDOF founded within the isolated ground mass is then
studied via FE analyses and the results are commented.

Keywords: seismic isolation, centrifuge tests, soft columns, deep mixing,

geotechnical earthquake engineering

propagation, the volume of soil to be

1. INTRODUCTION
Technological interventions into the
ground to mitigate the effects of vibrations
have been employed in the past, mostly
with reference to the effects of anthropic
actions (e.g. surface vibrations induced by
vehicles and rail-bound traffic). To this
aim, completely different solutions have
been studied, all with the aim to modify the
impedance ratio Į, defined as the
dimensionless ratio between the dynamic
impedance of the natural and of the
treated soil (D=Șs/Șg). Lombardi (2014)
have shown that in order to have an
effective modification of shear waves
* presenting author
protected has to be completely bounded
by the soft and weak layer, i.e. the barrier
must be continuous to ground level.
Based on this finding, numerical
parametric analyses have been carried out
to identify the geometrical and mechanical
properties of a soft barrier that optimize
the results under both static and dynamic
conditions (Lombardi et al., 2014; Flora et
al., 2016). The results have shown that
there is a range of values for the
parameters in which the barrier can be
effective as seismic isolation without
inducing excessive static settlements. A
suitable material to create such a low
impedance barrier is for instance a to capture the hysteretic behaviour of
hydrophilic polymer (Super Absorbent sands and the associated hysteretic
Polymer SAP) that can absorb and retain damping in unloading-reloading cycles.
extremely large amounts of a liquid with The model parameters for HN31 Hostun
respect to its own mass (Flora et al., sand were extracted from literature (Benz,
2015). Given the possible uncertainty on 2007) and they are reported in Table 1
the polymer properties measured in the Table 1. Hardening Soil Small Strain
laboratory, centrifuge experiments on parameters adopted for the sand
physical models of similar barriers at
Symbol Value Unit
reduced scale were used to further ref
support the numerical evidence of their G0 202000 kN/m2
effectiveness. The tests were carried out Ȗ0.7 0.0002 -
at the Schofield Centre of Cambridge Ȟ 0.25 -
University. The experimental results ref
E50 30000 kN/m2
confirm the effectiveness of such soft
barriers to reduce amplification in the Eoedref 30000 kN/m2
isolated volume during seismic events Eurref 90000 kN/m2
(Nappa et al. 2016). This paper will focus pref 100 kN/m2
on the case of a V shaped barrier. Starting
m 0.55 -
from the back-analysis of the results of the
free-field centrifuge tests, a series of c 0 kN/m2
numerical analyses have been carried out. ij 42 °
A series of different input signals were ȥ 16 °
applied to the base of the numerical nc
K0 0.4 -
model.
The soft barrier was modelled as an
2. CALIBRATION OF THE NUMERICAL elastic-perfectly plastic Mohr-Coulomb
MODEL material. Its shear strength was
determined through shear tests (Flora et
Centrifuge tests were carried out in the 10-
al., 2015) while the elastic shear wave
m diameter Turner beam centrifuge
velocity was obtained by means of air-
(Madabhushi, 2014). A sketch of the
hammer pulse tests during the centrifuge
centrifuge model is shown in Figure 1. The
flight. The adopted values of parameters
ground layer was made of dry Hostun
are summarized in Table 2.
sand and the V-shaped barrier was
created therein through latex balloons Table 2. Mohr-Coulomb parameters adopted
for the soft barrier
filled with a hydrogel with low shear
stiffness and strength (Flora et al., 2015). ȡ G0 ij Ȟ Vs
Numerical simulations of the two (kg/m ) (kN/m2)
3
(-) (°) (m/s)
centrifuge tests were performed by the FE
1020 109 5 0.4 12
code Plaxis 2D (Brinkgreve et al. 2011).
Additionally the same sand layer without To check the reliability of the numerical
the soft barrier was modelled, to have a model, the centrifuge tests with V shaped
reference free-field model for comparison. barrier were first simulated at model scale
The soil was characterized by a and after at prototype scale. Figure 2
constitutive model implemented in the shows, as an example, the results of such
Plaxis code, Hardening Soil with small calculations for one signal, compared with
strain overlay that accounts for strain the experimental result obtained from
hardening plasticity and small-strain centrifuge test in the control point at
behaviour of soils (Schanz et al., 1999; prototype scale. It can be observed that
Benz et al., 2009). The model is also able calculated and recorded acceleration time

394
 24m (80g)
300 mm
L1 L2

M1

A14 A1 A13 A7

12m (80g)
A2 A8
150 mm
A3 A9
22.8m (80g)

45°
285 mm
M3 A4 A10

A5 A11

A6 A12

AH
A15
500 mm
40m (80g)
A : Piezoelectric Accelerometers
M: Microelectromechanical System Accelerometer
L: Linear Variable Differential Transformer
AH: Air Hammer Piezoelectric Accelerometer

Figure 1. Benchmark centrifuge model (adapted after Nappa et al. 2016)

histories and the corresponding Fourier Once the model was calibrated, numerical
spectra are in good agreement. simulations of the model with V-shaped
barrier were performed with different input.
The time histories of acceleration used as
base input motions at the bottom
boundary of the FE mesh are reported in
Table 3 with the main characteristics of
each record. It is worth noting that the
input signals were chosen to cover the
broadest possible range of frequency
contents (approximately 0.5 to 10 Hz). In
such a way, the potential beneficial or
detrimental effects of the isolating barrier
can be better investigated, and its
limitations considered. All signals were
scaled at 0.3 g. To investigate the
potential influence of the barrier in the
earthquake performance of buildings, a
simplified SDOF models were developed.
The SDOF structure was composed of a
lumped mass on the top and two thin
supporting plates representing the lateral
stiffness of the structure. The stiffness of
the column supporting the mass in the
Figure 2. Recorded and calculated time SDOF model was calculated using an
history of acceleration and Fourier spectra at assumed fixed base natural period (Ts)
prototype scale varying from 0.1s to 1.5 s.

395
 Table 3. Time histories of acceleration used as input motions in FEA
Arias Predominant Significant
PGA Predominant
Database ID Name Date 2 Intensity frequency Duration
(m/s ) period (s)
(m/s) (Hz) (s)
146 Campano Lucano 23/11/1980 3.17 1.35 0.2 5.0 38.5
ITACA
175 Lazio Abruzzo 07/05/1984 1.23 1.30 0.4 2.5 12
1635x South Iceland 17/06/2000 1.53 0.87 0.24 4.2 4.5
1635y South Iceland 17/06/2000 1.29 0.73 0.16 6.3 5.8
ESD 1885 Kalamata 13/10/1997 1.15 1.93 0.3 3.3 17.7

South Iceland
2142x 21/06/2000 1.13 1.23 0.3 3.3 5.3
(aftershock)

Japanese SZO002 Japan 1997 1.67 0.89 0.3 3.3 4.2

Network
database EW 83 Kobe
3. THE ISOLATED MASS
The isolated soil mass can be created in
different shapes. As long as the barrier is
soft enough to prevent significant
deformations within the isolated soil mass,
it is reasonable to assume that such a
mass (m) will undergo an almost rigid-
body motion, as shown by Flora et al.
(2016) via numerical analyses. Then, the
isolated mass itself can be considered as
a SDOF system, whose stiffness k is
mostly ruled by the soft barrier geometrical
and mechanical properties. Based on
these considerations, Flora et al. (2016)
have proposed a simplified equation to
calculate the natural frequency fIM of the
isolated mass, assumed as a Single
Degree Of Freedom system. For the case
of a V-shape caisson (of maximum depth
HIM, base LIM and thickness of the soft
barrier S) the expressions of the natural
frequency is:
­ ª 2 º½
1 ° 4 «§ 2 ˜ E · § G · »°
2
fIM ® ˜¨ ¸ ¨ ¸ ¾
2S °̄ U 2 «¬© S ˜ LIM ¹ © S ˜ H IM ¹ »¼°¿
17/01/1995 3.02 1.35 0.08 12.5 15
and S=0.8m. In order to check the
expression (1), the frequency of the
isolated mass was also evaluated by
means of the software Plaxis2D. A free
vibration analysis has been carried out to
identify a natural frequency of vibration of
the isolated mass. A plastic analysis
thereby has been performed with a static
force acting laterally at the top left corner
of the isolated mass. The soil outside the
V-shaped mass was assumed as rigid in
order to avoid an influence on the
frequency evaluated. The next phase was
chosen to carry out the free vibration
analysis and the calculation for this phase
has been allowed to take into account the
displacements obtained from the previous
phase. The time history of displacements
is obtained at a point situated in
correspondence to the static force. The
corresponding Fourier transform (Fig. 3)
clearly shows a peak at frequency 1.4 Hz,
confirming the value obtained with
expression (1).
0.25
0.16
0.14
Fourier Amplitude

(1) 0.12
0.10
where E and G are the Young’s and the 0.08
shear modulus of the soft soil isolating the 0.06
mass and U is the isolated soil density. 0.04
The frequency evalueted for the model 0.02
0.00
reported in Figure 1 was equal to 1.4 Hz. It 0.00 0.50 1.00 1.50 2.00 2.50
was calculated according Eq. (1), where Frequency (Hz)
the mechanical parameter of the soft
barrier (E, G) and its density, U, can be Figure 3 Fourier spectrum of the
displacements at the top of the isolated mass
obtained from Table 2, HIM=12 m, LIM=24
396
 10
4. RESULTS Average
The numerical analyses confirmed that the

amax_SDOF_IM/amax_SDOF
insertion of a soft barrier completely
modifies the motion of the soil within the
protected volume. In order to understand 1
0 1 2 3 4 5 6 7 8
the filtering effect of the soft barrier, Fig.
4a shows the maximum acceleration of
the SDOF system founded on untreated
soil plotted against its predominant
0.1
frequency, while Fig 4b shows the fSDOF/fIM
acceleration of the same SDOF founded Figure 5. Ratio between maximum
on the untreated ground. acceleration of a SDOF founded within the
10.00
isolated mass and the same SDOF founded
on the untreated ground plotted against the
ratio between the frequency of the SDOF
amax_SDOF (g)

1.00
0 1 2 3 4 5 6 7 8 9 10 11
system and the natural frequency of the
isolated mass.
0.10
In fact, by plotting the ratio between the
maximum acceleration of the SDOF
system founded on untreated and founded
0.01 on treated soil against the ratio between
fSDOF
the frequency of the SDOF and the
(a) frequency of the isolated mass, is evident
10 the peak around 1, since confirming the
filtering effect. The results was also
reported in terms of Arias Intensity (Fig. 6)
amax_SDOF_IM (g)

1 calculated as
0 1 2 3 4 5 6 7 8 9 10 11
S T

2g ³
IA a(t ) 2 dt (m / s ) (2)
0.1 0

where T is the duration of the signal.

0.01
The ratio IASDOF_IM/IASDOF is plotted along
fSDOF
fSDOF/fIM. The average function (shown
(b) with bold dashed line) clearly confirms the
Figure 4 Maximum acceleration of a SDOF results discussed above in terms of
founded within the isolated mass (a) and the maximum acceleration, since the peaks of
same SDOF founded on the untreated ground Arias Intensity are attracted around the
(b). natural frequency of the isolated mass.
Comparing Figure 4a to Figure 4b the 100
“filtering” effect of the isolated mass of soil Average
at the SDOF foundation can be clearly 10
observed. The energy content of the
IASDOF_IM/IASDOF

SDOF system motion can be largely

attenuated by the isolated foundation, with 1
0 1 2 3 4 5 6 7 8
the exception of SDOF systems with a
natural frequency around that of the 0.1
isolated soil mass, fIM, also shown in the
figure 5.
0.01 fSDOF /fIM

Figure 6 Ratio between Arias intensity of the

time history of acceleration of a SDOF
397
5. CONCLUSIONS
Reducing seismic demand can
conveniently protect structures with high
natural frequency, namely squat buildings.
Soft barriers that isolate a mass of ground
well beneath the foundation level can be
created to this aim by modifying the soil
properties using the same techniques
usually adopted for ground improvement.
Consequently, seismic site amplification
can be locally modified. The isolated mass
of ground tends to shift the amplification of
the input motion at the base of the
structure towards its own frequency, thus
reducing the seismic demand on those
structures having larger natural
frequencies. As a consequence, it can be
concluded that the unusual idea of
inserting a soft barrier within the soil mass,
far from the building, is potentially an
innovative and alternative means to
protect existing buildings from seismic
excitations. Hence, whatever the
frequency content of the seismic event, a
structure founded within the isolated mass
will be efficiently protected if its
fundamental frequency (or frequencies, in
the case of a multiple degree of freedom
system) is (are) far enough from the
fundamental frequency of the isolated
mass of ground. . From a design point of
view, this means that the geometry of the
isolated soil mass, as well as the
properties of the soft grout to be adopted,
must be carefully chosen to maximize this
difference.
ACKNOWLEDGEMENTS
The research activity was carried out at
the University of Napoli Federico II as part
of the national research project PON_03
METRICS and the regional research and
innovation project TICISI. The financial
support of the consortium Stress s.c.a r.l.
is kindly acknowledged.
REFERENCES
Benz T. 2007. “Small Strain Stiffness of
Soil and its Numerical Consequences”.
Ph.d. thesis, Universitat Stuttgart
398
Brinkgreve R.B.J., Swolfs W.M., and
Engine E. 2011 “PLAXIS user’s
manual”, PLAXIS bv the Netherlands.
Flora A., Bilotta E., Lirer S., Lombardi D.,
Nappa V. 2015. “Behavior of a sand-
polyacrilate mixture”. Proc. of Sixth Int.
Symp. on Deformation Characteristics
of Geomaterials, 15-18 Nov, Buenos
Aires (Argentina).
Flora, A., Lombardi, D., Nappa V, Bilotta
E. 2016 “Numerical analyses of the
effectiveness of soft barriers into the
soil for the mitigation of seismic risk”,
submitted to Journal of Earthquake
Engineering.
Lombardi, D. 2014. “An innovative ground
treatment approach for seismic risk
mitigation of existing structures”. Ph.D.
Thesis, University of Napoli Federico II
(Italy).
Lombardi, D., Flora, A., Lirer, S., Silvestri,
F. 2014. “Soil grouting for the seismic
protection of existing buildings”. Ground
Improvement, 167, GI3: 206-218.
Madabhushi S.P.G. 2014. “Centrifuge
modelling for Civil Engineers”, Spon
Press, ISBN-10: 567 0415668247
Nappa V., Bilotta E., Flora A., Madabhushi
S.P.G. 2016 “Centrifuge modelling of
the seismic performance of soft buried
barriers”. Bulletin of Earthquake
Engineering. Accepted for pubblication
Schanz T., Vermeer P.A., and Bonnier
P.G. 1999. “The hardening soil model:
formulation and verification”. In Beyond
2000 in computation geotechnics (ed.
R. B. J. Brinkgreve), pp. 281–290.
Rotterdam: Balkema.
25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Session 8

Geosynthetics
Your success depends on
excellent results. That’s why
you can rely on our innovative
solutions. Customised to your
requirements, our tried and
tested products provide the Your project in safe hands.
basis for any earthworks or
ground engineering project.
Discover the world of geo-
synthetics – discover HUESKER.

www.HUESKER.com
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Swelling ability, volume of cracks and hydraulic conductivity of

a polymer modified bentonite subjected to wet and dry cycles

Michela DE CAMILLIS1
1
Ghent University, Department of Geotechnics, Ghent, BELGIUM

ABSTRACT
Geosynthetic Clay Liners are widely used to isolate waste disposal facilities. However, long
exposure to electrolyte solutions combined with temperature changes may impair their
performance as barrier liners. Wet and dry cycles lead to desiccation and cracking of the bentonite.
This study investigates the influence of wet and dry cycles with seawater on swelling ability, crack
formation and permeability of a polymer modified clay, HYPER clay, and untreated bentonite.
Untreated bentonite, bentonite treated with 2% and 8% polymer were evaluated through swelling
tests, μCT scanning and hydraulic conductivity tests.
HYPER clay 8% presented the best performance. It swelled the most and its thickness was
considerably larger compared to untreated clay. μCT analysis demonstrated the smaller volume of
cracks of HYPER clays compared to untreated bentonite. In addition, the hydraulic conductivity of
untreated bentonite increased within three cycles with seawater, while HYPER clay preserved low
permeability.

Keywords: bentonite, polymer, wet/dry cycles, crack formation

two geotextiles or glued to a

1. INTRODUCTION
Geosynthetic clay liners (GCLs) are
bentonite-based liners that are gaining
acceptance as hydraulic barriers in
containment and sealing applications
(Petrov & Rowe, 1997). One important
field of application is landfill capping
systems. The aim of clay liners is to limit
the infiltration of moisture, due to rainfall or
water migration, through the barrier into
the waste and to limit the release of
leachate and gasses from the waste.
GCLs are factory-manufactured clay liners
containing a thin uniform layer of sodium
or calcium bentonite sandwiched between
* presenting author
geomembrane.
The major significant component of the
bentonite in GCLs is sodium
montmorillonite. Sodium cations are able
to bond with water molecules, increasing
the interlayer space and forming tortuous
flow paths. However, valence,
concentration and dielectric constant of
the hydrating solution influence the
expansion of the diffuse double layer
(DDL) of negatively charged clays
(McBride, 1994). Accordingly, hydraulic
conductivity and swelling of bentonite are
related to the thickness of the DDL. A
decrease of the thickness leads to an
increase of hydraulic conductivity resulting
in particle attraction, shrinkage and
cracking of clay (Shackelford et al., 2000).
Therefore, bentonite is sensitive to
chemical interactions with the hydrating
liquid, and ion exchange can alter its
physical properties (Meer & Benson,
2007). Several laboratory studies have
investigated barrier performance
deterioration of GCLs in contact with
electrolyte solutions (Petrov & Rowe,
1997; Bouazza, 2002; Kolstad et al., 2004;
Jo et al. 2005).
Nevertheless, the service life of a GCL
cover can also be damaged by
degradation due to seasonal temperature
changes. Moreover, temperature as high
as 70°C may occur due to daily thermal
cycles when a geomembrane overlies a
GCLs (Take et al., 2014). As a result, the
hydraulic conductivity increases and loss
of self-healing capacity occurs due to the
combination of ion exchange and
desiccation (Egloffstein, 2001). In
particular, desiccation and contact with
high electrolyte solutions are responsible
for the collapse of the diffuse double layer
and crack formation. Therefore, crack
formation might not heal during rewetting
due to the low swelling ability of the
bentonite caused by the compression of
the DDL. Several studies have been
conducted to assess the effect of wet and
dry cycles (Lin and Benson, 2000;
Bouazza et al., 2006; Thiel et al., 2006;
Bouazza et al., 2007; Rowe et al., 2011;
Tang et al., 2011; Take et al., 2012; Hoor
and Rowe, 2013; Mukunoki et al., 2014).
Lin and Benson (2000) investigated the
impact of wet and dry cycles on the
swelling ability and hydraulic conductivity.
The initial exposure to deionized water or
tap water temporally delayed the reduction
in swelling. On the contrary, the swelling
capacity was reduced when the bentonite
was wetted directly with 0.0125 M CaCl2.
The GCLs permeability increased of
around two orders of magnitude after 4-6
wet and dry cycles due to crack formation
and loss of self-healing capacity of the
bentonite.
402
The GCL shrinkage caused by wet-dry
cycles was investigated by Rowe et al.
(2011). They found that the magnitude of
shrinkage in the field is correlated to the
water-retention of the GCLs, subsoil,
number and duration of cycles.
Modified bentonites have been developed
to improve bentonite performance in
aggressive environments (Katsumi et al.,
2008; Di Emidio, 2010; Malusis &
McKeehan, 2013; Scalia et al., 2013). In
this research, the HYPER clay technology
has been investigated. HYPER clay is a
polymer-treated bentonite created by
combining natural bentonite with
carboxymethyl cellulose (CMC). Once the
CMC intercalates the clay platelets, the
diffuse double layer is maintained open
even in presence of factors that generally
produce the collapse of the interlayer (Di
Emidio, 2010).
The purpose of this investigation is to
examine the impact of wet and dry cycles
of HYPER clay and untreated bentonite on
swelling ability, self-healing capacity,
crack formation and hydraulic conductivity.
The bentonites are hydrated with highly
concentrated electrolyte solution such as
seawater.
2. MATERIALS
Sodium bentonite (NaB) was used in this
study and treated according to the HYPER
clay procedure. Various physical and
chemical properties of the bentonite are
listed in Table 1.
Table 1. Material characterization
Property NaB
Specific Gravity [-] 2.66
Liquid Limit [%] 660.55
Plastic Limit [%] 48.92
Plasticity Index [-] 611.62
CEC [meq/100g] 73
Smectites [%] 91
Quartz [%] 4
Feldspars [%] 2
The HYPER clay was prepared according
to the procedure proposed by Di Emidio
(2010). The principle of HYPER clay is to
combine powdered Na-bentonite (NaB), 3. METHODS
referred to as base bentonite, with an
3.1. One-dimensional swell tests
anionic polymer (sodium Carboxymethyl
Cellulose, Na-CMC). The treatment One-dimensional swell tests were
consists of dissolving the polymer in water performed according to the procedure
and then adding the base clay. The used by De Camillis et al. (2016). Samples
material is mixed with a mechanical stirrer of dry bentonite of untreated clay, HYPER
for 30 minutes to increase the specific clay 2% and HYPER clay 8% were poured
surface area available for polymer in 70 mm diameter cells with initial
adsorption. This slurry is then oven dried porosity of 0.718 and 7.5 kg/m2 dry mass.
at 105°C for 16 hours to adsorb A sitting pressure of 1 kPa was used and
irreversibly the polymer. The HYPER clay the vertical swells were continuously
is then ground first manually, using a recorded during hydration. The samples
mortar and pestle, and then mechanically. were subjected to six wet and dry cycles.
In this study, the NaB was combined with Deionized water was the hydrating
2% and 8% of CMC, by dry mass of clay solution during the first cycle and seawater
(HYPER clay 2% and HYPER clay 8% in the consecutive cycles. The specimens
respectively). The dry mass per unit area were allowed to swell for about 400 hours
of the bentonite was 7.5 kg/m2 and the (16 days). After wetting, the samples were
initial porosity was 0.718. oven dried at 40°C until constant weight or
the water content was between 10%-15%.
Deionized water was used as reference
solution during the first wet cycle, for 3.2. μCT scanning
oedometer and hydraulic conductivity μCT scanning is a non-destructive three
tests. Natural seawater, collected in the dimensional (3D) imaging and analysis
North Sea (Ostend, Belgium), was filtered technique (Cnudde h Boone, 2013). With
through Grade 4 Whatman filter paper to this method, the samples are fully
remove coarse particles and stored in a reconstructed in 3D, based on a set of two
tank. Table 2 shows the chemical dimensional (2D) projections or
characteristics of the solutions used in this radiographs. μCT scans are carried out on
study. the HECTOR system (High Energy μCT
scanner, Optimized for Research), at the
Table 2. Chemical properties of electrolyte Centre for X-ray Tomography at Ghent
solutions University (UGCT), Belgium (Masschaele
et al., 2013). Samples are placed on a
Parameter Deionized Seawater
rotating stage between the X-ray source
water
and detector. After acquisition of the
EC (mS/cm) 0.002 44.8 radiographs, the 2D projections are
Salinity (-) 0.0 28.6 reconstructed into a stack of 2D slices
pH 7.57 7.42 through the object, building up the 3D
Na+ (mg/L) - 11517.9 image. The reconstruction step, as well as
K+ (mg/L) - 469.2 the subsequent image analysis on the
images, are carried out using the in-house
Mg2+ (mg/L) - 1281
developed software tools Octopus
Ca2+ (mg/L) - 478.5 Reconstruction (Inside Matters,
Cl- (mg/L) - 19897 Vlassenbroeck et al., 2007) and Octopus
SO42- (mg/L) - 2352 Analysis, formerly Morpho+ (Brabant et
HCO3- (mg/L) - 183.1 al., 2011), respectively. With VG Studio
CO32- (mg/L) - 18.0 MAX®, a software tool of Volume
NO32- (mg/L) - 43.4 Graphics, the analysed data could then be
visualized in 3D.
In this research, μCT scans were
performed on untreated sodium bentonite,
403
HYPER clay 2% and HYPER clay 8%,
after the third dry cycle and during the
fourth wetting stage in order to assess the
self-healing capacity and volume of
cracks. In order to exclude side effects at
the borders of the samples, the focus of
this comparison lies within the central
parts of the samples.
3.3. Hydraulic conductivity tests
Hydraulic conductivity tests were
performed on untreated sodium bentonite,
HYPER clay 2% and HYPER clay 8%
according to the procedure of De Camillis
et al. (2016). Tests were carried out in
rigid wall permeameters with 71 mm
diameter for NaB and HYPER clay 2%
and 70 mm for HYPER clay 8%. The initial
porosity of the samples was 0.718 and 7.5
kg/m2 dry density. HYPER clay 8%, due to
later shrinkage, was moved to a flexible
wall 70 mm diameter permeameter during
the second wet cycle to be able to
permeate the sample. The specimen of
HYPER clay 8% was confined with an
effective stress of 15 kPa, in order to
simulate in situ condition of a cover
(Bouazza, 2002; Mazzieri and Pasqualini,
2006; Scalia and Benson, 2011). NaB and
HYPER clay 2% recovered the initial
diameter once rewetted, for this reason
sidewall leakage did not occur.
The permeant liquids used were deionized
water, during the first cycle, and seawater
in the next cycles (as for swell tests).
Termination criteria from ASTM D6766
were followed. The dry cycles were
performed in a 40°C oven until constant
mass was reached, or the water content
was between 10%-15%.
4. RESULTS
4.1. One-dimensional swell tests
The first set of analysis investigated the
impact of wet and dry cycles on swelling
ability and self-healing capacity on
untreated sodium bentonite, HYPER clay
2% and 8% through free one-dimensional
swell tests in oedometer cells. Figure 1
outlines the swell related to the temporal
behaviour of the tested samples.
404
The first cycle in deionized water last 20
days. NaB and HYPER clay 2% were able
to reach equilibrium while HYPER clay 8%
did not yet achieve its maximum swelling.
On average, the swelling capacity
decreased with increasing the number of
wet-dry cycles.
Figure 1. Temporal behavior of 1-D free swell
tests at various wet-dry cycles of untreated
sodium bentonite (NaB), HYPER clay 2%
(HC+2%) and HYPER clay 8% (HC+8%)
Untreated sodium bentonite swelled the
least compared to HYPER clays, reaching
its maximum swelling of 7.3 mm in
deionized water. However, the swell of
NaB sharply decreased in contact with
seawater. Therefore, these results are
likely to be associated to the compression
of the DDL which causes loss of self-
healing and swelling capacities. As a
consequence, the cracks formed during
dry cycles might not be healed after
rewetting.
HYPER clay 2% showed a clear trend of
decreasing swelling ability with cycles. It
swelled around 11.45 mm during the first
cycle and it decreased up to 1.63 mm at
the end of the sixth cycle.
As shown in Figure 1, HYPER clay 8%
reported the best performance among the
analysed samples. It swelled the most in
deionized water (18.3 mm) and then the
thickness decreased with the consecutive
cycles. The most striking result to emerge
from Figure 1 is that HYPER clay 8%
thickness remained higher compared to
untreated bentonite and HYPER clay 2%
during further cycles with seawater. This
behaviour may be linked to the presence
of the polymer, which helps to keep the
interlayer open (Di Emidio et al., 2015)
and to enhanced water adsorption (Qui
and Yu, 2007). The more surprising
correlation is with sodium bentonite, the
final swelling of HYPER clay 8% (7.22
mm) after six cycles is comparable to the
maximum swelling in deionized water of
NaB.
Swelling and adsorption capacity are
strongly dependent on the electrolyte
concentration of the hydrating liquid, such
as seawater. Moreover, the combination
with wet and dry cycles can further
deteriorate the barrier performance of the
bentonite. The diffuse double layer
thickness is so compressed when it comes
in contact with multi-valence and
concentrated solutions. For these reasons,
collapse of swelling ability was detected
for NaB and the cracks, formed during
desiccation, were not healed. On the
contrary, the polymer treatment improves
the swelling and healing capacity likely
maintaining the DDL open in the long-
term. The sample of HYPER clay 8%
formed a rigid disk after the first dry cycle
with seawater, which was maintained
during the next cycles. Crack formation
was detected on the edge where the
sample was more in contact with the heat.
However, HYPER clay 8% presented the
first cracks during the fourth cycle.
4.2. μCT scanning
In order to quantitatively compare the
amount of cracks and check for self-
healing capacity of NaB, HYPER clay 2%
and HYPER clay 8%, μCT scanning were
performed in cooperation with Van
Stappen J. and Cnudde V. at
PProGress/UGCT - Department of
Geology – SHE, Faculty of Sciences,
Ghent University.
Figure 2. Specimens at the end of the third dry
cycle and at the end of the fourth wet cycle
405
The analysis of the images was focused
on the internal part of the samples to
avoid edge effects. A cylindrical volume
with a diameter of 35 mm and a height of
6.6 mm was chosen as a subsection to
perform image analysis on. The samples
were scanned at the end of the third cycle
and during the fourth wet cycle (Figure 2).
Figure 3 presents the amount of cracks
detected at the dry (end third dry cycle)
and wet (fourth wet cycle) conditions. By
keeping the dimensions of the cylindrical
subsection identical, and by ensuring it is
positioned at similar locations in each of
the samples, the analysis of the crack
opening can be used as a measure for the
swelling of the clays.
It can be seen from figure 2 that NaB
shows wider cracks throughout the full
volume after the third dry stage. On the
contrary, the cracks in the HYPER clays
are mainly due to edge effects.
Figure 3. Amount of cracks versus dry and
wet condition of untreated bentonite (NaB),
HYPER clay 2% (HC+2%) and HYPER clay
8% (HC+8%)
The amount of cracks developed by NaB
after the third dry was about 6% of total
initial volume. Whereas the amount of
cracks of HYPER clays were 0.94% and
0.45% for HYPER clay 2% and HYPER
clay 8% respectively. The amount of
cracks decreased for all the samples on
wet conditions due to the swelling and
healing of the bentonites. In particular,
untreated bentonite presented an amount
of cracks around 1.2% of the analysed
subsection, comparable to the amount of
cracks of HYPER clay in dry condition.
4.3. Hydraulic conductivity tests
Hydraulic conductivity tests were
performed in order to assess the influence
of wet and dry cycles on the self-healing
capacity and permeability of samples of
untreated sodium bentonite, HYPER clay
2% and HYPER clay 8%. Measured
hydraulic conductivity (k) values are
plotted as a function of wet-dry cycles in
Figure 4.
Figure 4. Hydraulic conductivities (k) of
sodium untreated bentonite (NaB), HYPER
clay 2% (HC+2%) and HYPER clay 8%
(HC+8%) at each wetting cycles
The samples presented similar and low
permeability in deionized water, which
increased in the next cycles in seawater.
The greatest increase was observed for
untreated sodium bentonite. NaB
significantly increased its permeability up
to 2.93 × 10-7 m/s during the fourth cycle.
These results are in agreement with the
DDL theory. Untreated bentonite forms
aggregate structure once in contact with
strong electrolyte solution due to the
contraction of the DDL. As a water. Untreated sodium bentonite has
consequence, self-healing and swelling
capacity are weakened and the barrier
performance of the bentonite is impaired.
On the other hand, HYPER clays showed
lower permeability compared to untreated
clay. The k of HYPER clay 2% was lower
than the limit value of 10-9 m/s until the
fourth cycle. The sudden increase in k
during the second cycle for NaB and
HYPER clay 2% might be related to the
406
strong exposure of the samples to the
heat in the oven during the first dry cycle.
For this reason, cracks may not be
completely healed. The exposure was
then confined to better represent the
gradual dehydration expected in the field.
HYPER clay 8% presented a gradual and
constant increase of permeability. At the
end of the third cycle k was 9.11×10-11
m/s. The difference between untreated
clay and HYPER clays might be explained
by the presence of the polymer
intercalated in the clay particles, which
remains over wet and dry cycles.
Moreover, it is likely that the presence of
the polymer helped to keep the diffuse
double layer open during hydration.
However, wet-dry cycles likely showed
lower impact on the polymer treated clays.
The samples of HYPER clay presented
few cracks during desiccation which were
healed after rehydration, as can be noted
from the low permeability.
5. CONCLUSIONS
The potential influence of wet and dry
cycles in combination with cation
exchange was investigated by means of
swell test, μCT scanning and hydraulic
conductivity tests. Samples of untreated
sodium bentonite were compared to
samples of HYPER clay 2% and 8%. The
performance of these bentonites
subjected to wet and dry cycles in contact
with seawater was studied. Seawater is a
highly concentrated electrolyte solution
and represents an aggressive
environment in the field.
The swelling ability increased with
increasing polymer dosage in deionized
swollen the least and its swelling ability
was strongly affected from the consecutive
wet and dry cycles. μCT analysis of
untreated bentonite, HYPER clays 2% and
8% have shown the effect of the polymers
on the healing capacity. HYPER clays
presented lower amount of cracks both on
dry and wet conditions and better healing
capacity upon rewetting.
The effect of wet and dry cycles in
seawater had a negative impact also on
the hydraulic conductivity of the untreated
bentonite. On the other hand, the
hydraulic conductivity of HYPER clay 2%
was 3.5×10-10 m/s, during the fourth cycle,
9.11×10-11 m/s for HYPER clay 8%, during
the third cycle.
HYPER clay presented enhanced
performance after being subjected to wet
and dry cycles in seawater. The presence
of the anionic polymer likely helps to keep
the interlayer open, allowing greater water
adsorption and swelling capacity. For this
reason might be that the cracks were
healed after rewetting and the
permeability was maintained low.
REFERENCES
ASTM D6766, 2002. Standard Test
Method for Evaluation of Hydraulic
Properties of Geosynthetic Clay Liners
Permeated with Potentially
Incompatible Liquids. American Society
for Testing and Materials Internationals,
West Conshohocken, Pennsylvania,
USA.
Bouazza, A. (2002) “Geosynthetic clay
liners”, Geotext. Geomembr. 20 (1), 3–
17.
Bouazza, A., Vangpaisal, T. & Jefferis, S.
(2006) “Effect of Wet – Dry Cycles and
Cation Exchange on Gas Permeability
of Geosynthetic Clay Liners”, J.
Geotech. Geoenviron. Eng. 132 (8),
1011–1018.
Bouazza, A., Jefferis, S. & Vangpaisal, T.
(2007) “Investigation of the effects and
degree of calcium exchange on the
Atterberg limits and swelling of
geosynthetic clay liners when subjected
to wet–dry cycles”, Geotext.
Geomembr. 25 (3), 170–185.
Brabant, L., Vlassenbroeck, J., De Witte,
Y., Cnudde, V., Boone, M.N.,
Dewanckele, J. & Van Hoorebeke, L.
(2011) “Three-Dimensional Analysis of
High-Resolution X-Ray Computed
Tomography Data with Morpho+”,
Microscopy and Microanalysis 17(2):
252-263.
407
Cnudde, V. & Boone, M.N. (2013) “High-
resolution X-ray computed tomography
in geosciences: a review of the current
technology and applications”, Earth
Science Reviews 123, 1-17.
De Camillis, M., Di Emidio, G., Flores
Verastegui, D. & Bezuijen, A. (2016)
“Hydraulic conductivity and swelling
ability of a polymer modified bentonite
subjected to wet-dry cycles in
seawater”, Geotext. Geomembr., in
Press.
Di Emidio, G. (2010) “Hydraulic and
Chemico-Osmotic Performance of
Polymer Treated Clays”, PhD thesis,
Ghent University.
Egloffstein, T. A. (2001) “Natural
bentonites - influence of the ion
exchange and partial desiccation on
permeability and self-healing capacity
of bentonites used in GCLs”, Geotext.
Geomembr. 19, 427–444.
Hoor, A. & Rowe, R. K. (2013) “Potential
for Desiccation of Geosynthetic Clay
Liners Used in Barrier Systems”,
Journal of Geotechnical and
Geoenvironmental Engineering,
139(10), 1648–1664.
Jo, H. Y., Benson, C. H., Shackelford, C.
D., Lee, J.M. & Edil, T. B. (2005) “Long-
Term Hydraulic Conductivity of a
Geosynthetic Clay Liner Permeated
with Inorganic Salt Solutions”, J.
Geotech. Geoenviron. Eng. 131 (4),
405–417.
Katsumi, T., Ishimori, H., Onikata, M. &
Fukagawa, R. (2008) “Long-term barrier
performance of modified bentonite
materials against sodium and calcium
permeant solutions”, Geotext.
Geomembr. 26 (1), 14–30.
Kolstad, D. C., Benson, C. H. & Edil, T. B.
(2004) “Hydraulic Conductivity and
Swell of Nonprehydrated Geosynthetic
Clay Liners Permeated with
Multispecies Inorganic Solutions”, J.
Geotech. Geoenviron. Eng. 130 (12),
1236–1249.
Lin, L. C. & Benson, C. H. (2000) “Effect of
Wet-Dry Cycling on Swelling and
 Hydraulic Conductivity of GCLs”, J.
Geotech. Geoenviron. Eng. 126 (1),
40–49.
Malusis, A. M. & McKeehan, D. M. (2013)
“Chemical Compatibility of Model Soil-
Bentonite Back fill Containing
Multiswellable Bentonite”, J. Geotech.
Geoenviron. Eng. 139 (2), 189–198.
Masschaele, B., Dierick, M., Van Loo, D.,
Boone, M., Brabant, L., Pauwels, E.,
Cnudde, V. & Van Hoorebeke, L.
(2013) “HECTOR: a 240kV micro-CT
Setup Optimized for Research”, Journal
of Physics Conference Series. Vol. 463
(012012). Bristol, UK: IOP.
Mazzieri, F. & Pasqualini, E. (2006)
“Evaluating the Permeability of an
Organically Modified Bentonite to
Natural Seawater”, Proceedings of the
5th ICEG Environmental Geotechnics ,
Thomas Telford, London. pp. 749–756.
McBride M. B. (1994) “Environmental
Chemistry of Soils”. Oxford University
Press, New York.
Meer, S. R. & Benson, C. H. (2007)
“Hydraulic Conductivity of Geosynthetic
Clay Liners Exhumed from Landfill Final
Covers”, J. Geotech. Geoenviron. Eng.
133 (5), 550–563.
Mukunoki, T., Nakano, T., Otani, J. &
Gourc, J. P. (2014) “Study of cracking
process of clay cap barrier in landfill
using X-ray CT”, Appl. Clay Sci. 101
(10), 558–566.
Petrov, R. J., & Rowe, R. K. (1997)
“Geosynthetic clay liner (GCL) -
chemical compatibility by hydraulic
conductivity testing and factors
impacting its performance”, Canadian
Geotechnical Journal, 34(6), 863–885.
Rowe, R. K., Bostwick, L. E. & Take, W. A.
(2011) “Effect of GCL Properties on
Shrinkage When Subjected to Wet-Dry
Cycles”, J. Geotech. Geoenviron. Eng.
137 (11), 1019–1027.
Scalia, J. & Benson, C. H. (2011)
“Hydraulic Conductivity of Geosynthetic
Clay Liners Exhumed from Landfill Final
Covers with Composite Barriers”, J.
408
Geotech. Geoenviron. Eng. 137 (1), 1–
13.
Scalia J., Benson, C. H., Bohnhoff, G. L.,
Edil, T. B. & Shackelford, C. D. (2013)
“Long-Term Hydraulic Conductivity of a
Bentonite-Polymer Composite
Permeated with Aggressive Inorganic
Solutions”, J. Geotech. Geoenviron.
Eng. 13 (9).
Shackelford, C. D., Benson, C. H.,
Katsumi, T., Edil, T. B. & Lin, L. (2000)
“Evaluating the hydraulic conductivity of
GCLs permeated with non-standard
liquids”, Geotext. Geomembr. 18, 133–
161.
Take, W. A., Brachman, R. W. I., Rowe, R.
K. & Rentz, A. (2014) “Temperature
measurements of exposed GMB / GCL
composite liners”, Proceedings 10th
International Conference on
Geosynthetics. Berlin, Germany.
Take, W. A., Potvin, J., Siemens, G. A., &
Kerr, A. (2012) “Comparison of
desiccation cracking patterns between
air- and oven-dried GCL specimens”.
Proceedings 5th European
Geosynthetic Congress. Vol. 3, pp. 71–
76. Valencia.
Tang, C.S., Shi, B., Liu, C., Suo, W.B. &
Gao, L. (2011) “Experimental
characterization of shrinkage and
desiccation cracking in thin clay layer”,
Appl. Clay Sci. 52 (2), 69–77.
Thiel, R., Giroud, J. P., Erickson, R.,
Criley, K. & Bryk, J. (2006) “Laboratory
measurements of GCL shrinkage under
cyclic changes in temperature and
hydration conditions”, Proceedings of
the Eighth International Conference on
Geosynthetics, 18/09/2006 to
22/09/2006. Yokohama, Japan.
Vlassenbroeck, J., Dierick, M.,
Masschaele, B., Cnudde, V., Van
Hoorebeke, L. & Jacobs, P. (2007)
“Software tools for quantification of X-
ray microtomography”, Nucl. Instr.
Meth. Phys. Res., Sect. A 580 (1), 442–
445.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Physical and numerical modeling of an unpaved road structure

liable to void forming. The effect of a biaxial geosynthetic
reinforcement.

Andor-Csongor NAGY1*, Dorin Vasile MOLDOVAN1

1
Technical University of Cluj-Napoca, Department of Structures, Cluj-Napoca, ROMANIA

ABSTRACT
Earth structures are commonly designed using numerical calculation programs. This kind of
modeling relies on predefined mathematical equations and boundary conditions for approximating
deformations, strains and interactions. In most cases, numerical modeling is sufficient for drawing
long-term conclusions on earth structure behavior, but there are special situations which require
physical modeling in order to verify the results of virtual modeling. For example, a cavern forming
in a road embankment requires a physical model, because simulating the arch effect in the earth
structure represents an important challenge for most of the finite element based calculation
programs. The comparison between the scale model settlements and the deformations from the
modeling program can show how suited the applied mathematical model is, and if the results are
close, then the numerical calculation is reliable in displaying the behavior of the granular structure,
which is impossible to measure on the scale model.

Keywords: biaxial geogrid, finite element modeling, scale model testing

pipeline intervention conducted in the

1. INTRODUCTION
Cavern forming in road embankments
is relatively common, due to a multitude of
favorable factors, like hydrological factors
leading to formation of sinkholes, or
different construction activities, such as
pipeline leaking, or poor compaction of the
base layer which causes differentiated
settlements. The following pictures were
taken on the DJ108A road located in Sălaj
county, Romania. Figure 1.a. shows a
* presenting author
summer of 2012, while Figure 1.b. shows
roughly the same location six months
later, in autumn 2012. Clearly the road
safety was affected by the resulting
deformations. Eventually, the road
structure was subjected to costly
interventions, that implied scraping and
auxiliary compaction.
 geogrid with 40 mm apertures was
inserted. [Moldovan D., Nagy A., Farcas
V., Muntean L., CoĠ, R. 2014]

Figure 1.a. Regular pipeline interventions on

DJ108A Figure 2. The configuration of the scale model
subjected to laboratory testing
A void of 50 cm wide and 25 cm high was
created at the bottom-center of the span
on both series of models.

Pressure
Equivalent
transmitted
Equivalent pressure in
Load to the
concentrated tone-
stage surface of
force force/square
the testing
meter
plate
[kPa] [kN] [tf]/m2 [tf]
50 3.55 5.100 0.362
Figure 1.b. Differentiated settlements resulted
from poor compaction works 100 7.10 10.200 0.724
Although biaxial geogrids are designed for 150 10.65 15.300 1.086
a combined effect of dissipating the
200 14.20 20.400 1.448
pressure cone due to traffic loads through
surface friction, the question of using 250 17.76 25.500 1.811
geosynthetic reinforcement in the present 300 21.31 30.600 2.173
case in order to stabilize the road structure
and decrease settlement is open to 350 24.86 35.700 2.535
debate. 400 28.41 40.800 2.897
450 31.96 45.900 3.259
2. SCALE MODELING
500 35.51 51.000 3.621
Trying to replicate the presented case,
our study consisted of building two 550 39.06 56.100 3.983
structures at natural scale (1:1) in the 600 42.61 61.200 4.345
laboratory using a box, according to the
detail given in Figure 2. The road bed was 650 46.16 66.300 4.707
made of 40 cm thick cohesive earth/clay 700 49.71 71.400 5.069
where a void was created. Over the earth, 750 53.27 76.500 5.432
a layer of 40 cm crushed stone was laid
and compacted. For the second model, we Table 1. Load stages and their equivalent
had the same structure, only that, between concentrated force and pressure transmitted
the two material layers, a layer of biaxial on the surface of the Lucas plate

410
 The void was formed by using an
inverted drawer, having the specified
dimensions, which was retracted after
ending the compaction works for both of
the earth layers used in the experimental
stage. The final configuration of the test
model was influenced by similar studies
on geosynthetic reinforced earth models
combined with void forming, found in the
works of Agaiby S., Jones J. F. P.,
Asakereh A., Ghazavin M., and Tafreshi
S.N. Moghaddas, quoted in the reference
part of the study. For a preliminary
calculation we used a design pattern from
a study conducted by Giroud J.P.,
Bonaparte R., Beech J.F., and Gross B.A.,
also quoted in the references.
Loading stages were applied according
to the regulations in force in Romania,
specified in STAS 2914/4-89 for road and
railway works. Testing consisted in placing
a 300 mm diameter plate on the top center
of the span, and applying consecutive
load stages, starting from 50 kPa (with a
50 kPa step) until the structure fails. Load
stages were only increased when the
surface settlement was considered
stabilized (when settlement does not
increase by more than 0.1 mm for 30
minutes under the applied load stage).
Figure 3. Lucas plate testing on the laboratory
model
Behaviour differences between the
reinforced and unreinforced structures
became clear under the 350 kPa stage.
Under this load, the geogrid provided
sufficient confinement for the superior
layer, in order to stabilize the deformations
411
in a time interval of just 20 minutes, as
against the other structure, of which
consolidation under the same stage lasted
45 minutes. The tendency emphasized on
the following phases of loading,
respectively at 400 kPa we recorded a
consolidation time of 45 minutes on the
reinforced structure, compared to 65
minutes on the unreinforced one, whereas
the surface deformations were noticeably
larger in the case of the unreinforced
structure. The 450 kPa load stage caused
the falling of medium sized earth lumps
from the gap ceiling on the unreinforced
complex, projecting the nearby failure,
while the gap on the reinforced structure
suffered only slight geometrical
modifications, because the grid acted by
friction on the foundation layers. During
this experiment, we also monitored the
interior deformations of the void, and
concluded that in the case of the
unreinforced structure, the surface
deformation coincides with the vertical
deformation of the gap ceiling, while on
the reinforced structure the settlement was
significantly larger on the surface, due to
the rearrangement of particles in the grill
meshing, which seems to “filter”
deformations. Under the 550 kPa stage
the unreinforced structure collapses
around the gap, the failure is sudden and
unexpected. The maximum surface
settlement on the unreinforced laboratory
model was 2.76 cm (measured before the
structure collapsed). The reinforced
structure loses balance at 700 kPa, but
unlike the previous model, the latter fails in
10 minutes, with constant deformations of
cavity, until its ceiling touches down on the
bottom of the testing box.
Comparing these results with those for
the unreinforced structure we found that
the presence of the geogrid grants an
important increase in bearing capacity.
Also the maximum settlement measured
on the reinforced scale model was 5.20
cm. This highlights the difference between
the failure types of the two structures.
While the unreinforced structure collapses
suddenly, with a fairly small surface
settlement, the reinforced structure fails
slowly and gradually, making it less
hazardous for traffic.
Figure 4.a. Failure surface on the unreinforced
scale model
Figure 4.b. Failure surface on the reinforced
the simulated structure consisted of both
cohesive (initial layer of clay) and non
cohesive (base layer of crushed stone)
soil types.
Failure criteria define the linear – elastic
behavior limit of materials. The natural
humidity of the clay used in the scale
model was 40%, determined in laboratory
conditions. Cohesion and internal friction
angle values were considered under
drained conditions, and were determined
in laboratory shear tests. The crushed
stone base layer would have been more
accurately defined by the Mohr – Coulomb
failure criterion, but it was stored outdoors,
in rainy weather, and as a consequence, it
accumulated a high amount of water. As
the primary objective of the study was to
correlate scale and numerical modeling
results, von Mises criterion was chosen for
this layer too. The von Mises perfectly
plastic models approach is based on the
assumption that plastic deformation
begins when the potential energy required
for changing the shape of the finite
elements (noted with Wd) reaches a
critical value specific to each type of
material introduced in the numerical
model.
Wd
1X
6E
> @
V1 V2 2  V2 V32  V3 V12 (1)
scale model
where E is the linear deformation
The failure surface is more extended in
modulus, ȣ is the Poisson ratio,
the case of the reinforced structure, as
respectively ı1, ı2 and ı3 are the normal
seen in Figure 4 a. and b.
stresses on the main directions. A
graphical representation of the von Mises
3. VIRTUAL MODELING
failure criterion can be seen in Figure 5.
A suited method for creating a virtual
model for the present experiment must
allow interaction between the earth layers
and reinforcement material. Finite element
modeling determines the stress domain
corresponding to the applied load stage,
and geometry of the given structure. Using
a predefined failure criterion, the software
indicates if flowing occurs in any given
point of the model. Also it can model
failure through conditions dependent or
independent of hydrostatic pressure
(inside the soil pores), which is an
important aspect, considering the fact that Figure 5. Graphical representation of the von
Mises failure criterion [Chiorean C. 2010]
412
 The two foundation layers were corresponding settlement values. The
introduced with their geological boundary condition on the FEM model
characteristics, while the geogrid was was set as fixed, as no interaction was
introduced as a linear element, defined by determined between the box walls and the
its physical and mechanical properties. A earth layers on the scale model.
cooperation coefficient was also
introduced, for simulating the confinement
effect.
The two layers were defined as (after
being tested in the laboratory):
Layer 1 – clay
module of elasticity: E=4200 kN/m2
specific weight: Ȗ=18 kN/m3
cohesion (drained): cd=79.92 kN/m2
cohesion (undrained): cu= 45.81 kN/m2 Figure 6. Boundary conditions and load
internal friction angle (drained): ijd=9.01° application on the joints of the FEM model
Layer 2 – crushed stone The amount of iterations set by the user
requires attention. Usually, the number of
linear deformation modulus: E=130000
iterations ranges from 200-1000, but total
kN/m2
displacement resulting from the program
specific weight: Ȗ=20 kN/m3 should be followed at every step. If after
cohesion: c=0 kN/m2 two different iteration values, the
internal friction angle: ij=35° difference between the two surface
Layer of geosynthetical reinforcement – settlement values (total maximum
biaxial geogrid with 40 mm apertures displacement) is greater than 20%, the
number of optimal iterations is exceeded,
linear deformation modulus: E=185000
and the program gives altered values.
kN/m2
[Chiorean C. 2009]
density: ȡ= 1kN/m3
traction resistance: T=30 kN/m [Tensar
UK Catalogue] Number Settlement
Iteration
Iteration of value
Load stages on the scale model were force value
value F iterations
applied statically. On the virtual model, a [kN] [cm]
performed
number of seven joints were selected (on
0.1 0.6 2 0.02
a surface roughly the size of the 300 mm
Lucas plate), applying concentrated forces 0.2 1.2 23 0.049
onto the joints, equivalent to the failure 0.3 1.8 57 0.126
value obtained from the laboratory tests.
This way, on the unreinforced model a 0.4 2.4 103 0.271
value of 6 kN was chosen (with a total 0.5 3 139 0.485
value of 7x6 kN = 42 kN, slightly superior
0.6 3.6 217 0.85
to the failure value of 39 kN), while on the
reinforced model the 7 kN value was 0.7 4.2 371 1.636
found appropriate, as seen in Figure 6. 0.8 4.8 500 3.1
Then an iteration value was introduced, in
order to apply the force gradually. Tables Table 3. Performed iterations and
corresponding settlement values for the
2 and 3 show the number of iterations
unreinforced FEM model
performed on the reinforced and
unreinforced models, also displaying the

413
 Iteration Number Settlement
Iteration force of value
value F value iterations
[kN] performed [cm]

0.1 0.7 2 0.02

0.2 1.4 4 0.045
0.3 2.1 39 0.09
0.4 2.8 73 0.221
0.5 3.5 93 0.397
0.6 4.2 153 0.689 Figure 7.b. Vertical stress distribution on the
elements of the reinforced FEM model
0.7 4.9 239 1.299
0.8 5.6 367 2.562 4. COMPARRISON BETWEEN THE
0.9 6.3 500 5.17 RESULTS OBTAINED BY SCALE
MODELING AND VIRTUAL MODELING

Table 4. Performed iterations and

corresponding settlement values for the
reinforced FEM model
FEM model displays internal stress
on the meshed elements. The average
value on the elements of the reinforced
model is between 90-110 kPa. Values on
the unreinforced model are sensibly higher
averaging in between 130-210 kPa, as
shown in Figures 7.a. and b. Comparing
the resulted graphics we can conclude
that in the case of the reinforced model
the vertical stress values on the elements
above the void is considerably lower due
to the presence of the geosynthetical
reinforcement, and only a small amount of
pressure is transmitted to the side of the
void.
Figure 7.a. Vertical stress distribution on the
elements of the unreinforced FEM model
The results of virtual modeling were
largely confirmed by the measurements
obtained from the 1:1 scale model testing.
On the numerical model without
geosynthetical reinforcement failure
occurred at a force equivalent of 34 kN.
This value corresponds to the 500 kPa
load stage. The scale model collapsed at
a uniformly applied pressure (under the
loading plate) of 550 kPa. The maximum
settlement of the computer generated
model was 3.10 cm (at 500 kPa), while at
the same pressure on the laboratory
model a 2.76 cm value was measured,
resulting a 12% difference. Figure 8.a.
displays a chart based on the settlement –
equivalent applied force function for both
the FEM (highlighted in blue) and scale
model (highlighted in red). The behavior of
the two models is remarkably similar.
In the following step of the study a
rectangular meshed (biaxial) geogrid was
used for the road structure reinforcement.
The nonlinear analysis performed on this
configuration showed that failure occurs at
650 kPa load stage, meanwhile the
laboratory model collapsed at 700 kPa.
Comparing these results with those for the
unreinforced structure we found that the
presence of the geogrid grants an
important increase of bearing capacity.

414

Figure 8.a. Applied force – surface settlement
comparison chart between FEM (blue) and
scale modelling (red) on the structure without
geosynthetic reinforcement
Figure 8.b. Applied force – surface settlement
comparison chart between FEM (blue) and
scale modelling (red) on the structure with
geosynthetic reinforcement
Also, the difference between the maximum
settlements of the virtual and real model
was very small, 5.17 cm in the finite
element method, respectively 5.20 cm on
the scale model, with a difference of 1%.
Comparing the scale model and the FEM
model behavior wise, as displayed in
Figure 8.b., shows, as in the first case, a
highly similar curve. Reinforcing the road
structure with biaxial geogrid grants an
increase of 22% in load capacity. The
presence of the geosynthetic
reinforcement also allows higher
415
settlement values to be reached, before
the collapsing of the structure.
5. CONCLUSIONS
The resulting 1% difference between
surface settlements in numerical and
physical modeling, in the case of the
reinforced structure proves that virtual
modeling by the finite element method can
get very close to the behavior of a 1:1
scale model. The 50 kPa difference
between failure values, obtained in both
cases, is due to the constant particle
rearrangement in the soil structure, which
results in additional bearing capacity. This
aspect is impossible to take in
consideration in the numerical modeling,
as the finite element method only allows
deformation of the subdivisions, and does
not permit rotations or translations of the
divided parts.
Other noteworthy conclusions of the study
are:
x if the failure criterion is chosen
respecting the initial condition and
type of material used, the behavior
of the numerical model comes very
close to the behavior of the scale
model;
x the unreinforced scale model fails
suddenly, while reinforced models
fail slowly making them less
hazardous for traffic;
x biaxial geogrid increases the
bearing capacity by 22% as against
the results obtained on the
unreinforced structure;
x while the surface settlement
coincides with the gap deformation
in the case of the unreinforced
model, using geosynthetic
reinforcement leads to a smaller
deformation of the void, due to the
arrangement of the earth particles
from the superior layer;
x in the FEM model, stress
distribution inside the structure is
influenced by the presence of the
reinforcement, as the values
obtained on the elements adjacent
to the void are considerably lower;
ACKNOWLEDGEMENTS polypropylene rectangular mesh
geogrids”, 14th International
This research was supported by IRIDEX
Group Plastic, VIAROM CONSTRUCT Multidisciplinary Scientific
and Geostru Romania, who provided raw Geoconference SGEM2014,
material for the laboratory models, and Conference Proceedings Volume II
software license for the numerical (Eds: STEF92 Technology Ltd.), 11 –
17, Albena, Bulgaria
calculation. We are also thankful to our
colleagues Vasile Fărcaú and Cosmin Nagy A., Ursu I., Moldovan D. (2016)
Chiorean who provided expertise that “Összehasonlító tanulmány egy
greatly assisted the research. háromszög rácsos geoszintetikus
anyagokkal megerĘsített útszakasz
REFERENCES valós és virtuális modellezése
kapcsán”, Proceedings of the XXI-th
Agaiby S., Jones J. F. P. (1996) „Design
International Conference of Young
of reinforced fill systems to support
Engineers (Eds: Erdélyi Múzeum
footings overlying cavities”, Geotextiles
Egyesület, F&F International Kft. Kiadó
and Geomembranes 14, 57-72.
és Nyomda), 289 – 293, Cluj-Napoca,
Asakereh A., Ghazavin M., Tafreshi S.N. Romania.
Moghaddas (2013) „Cyclic response of
STAS 2914/4-89 Road and railway works.
footing on geogrid-reinforced sand with
Determination of linear deformation
void”, Soils and Foundations 53(3),
modulus.
363–374.
Tensar UK Catalogue, www.tensar.co.uk
Briançon L., Villard P. (2008) „Design of
geosynthetic-reinforced platforms
spanning localized sinkholes”,
Geotextiles and Geomembranes 26,
416–428.
Chiorean C. (2009) Calculul neliniar al
structurilor vol. 1, U.T. Press, Cluj-
Napoca.
Chiorean C. (2010) Geostru software,
GFAS Finite Element Systems for
Geotechnical applications, Theoretical
and User Manual 2010,
www.geostru.com
Găzdaru A., Manea S., Fedorov V., Batali,
L. (1999) Geosinteticele în construc‫܊‬ii,
Ed. Academiei Române, Bucure‫܈‬ti.
Giroud J.P., Bonaparte R., Beech J.F.,
Gross B.A. (1990), „Design of soil layer-
geosynthetic systems overlying voids”,
Geotextiles and Geomembranes 9, 11–
50.
Koerner, R. M. (2005) Designing with
geosynthetics. Fifth Edition. Upper
Saddle River, N.J. 07458.
Moldovan D., Nagy A., Farcas V.,
Muntean L., CoĠ, R. (2014) “A
comparative study of the failure mode
of conventional road structures and of
road structures reinforced with
416
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Influence of prehydration condition on the hydraulic

conductivity of GCLs permeated with landfill leachates

Tu÷çe ÖZDAMAR KUL1*, Havva DEMøRKIRAN2

1
Dokuz Eylül University, Graduate School of Natural and Applied Sciences, Department of
Civil Engineering, TURKEY
2
Purdue University, Lyles School of Civil Engineering, USA

ABSTRACT
In this study, the effect of prehydration on the hydraulic conductivity of geosynthetic clay liners
(GCLs) was evaluated using landfill leachates as the permeants (LL-A and LL-B). For this purpose,
two local GCLs (GCL-1 and GCL-2) were conducted to hydraulic conductivity tests using flexible-
wall permeameters. The influence of prehydration conditions on the hydraulic conductivity was
assessed by considering two situations of GCLs: (i) non-prehydrated and (ii) prehydrated-
prepermeated. The hydraulic conductivities of GCL-1 and GCL-2 with deionised water (DIW) were
1.2×10-9 cm/s and 1.4×10-9 cm/s, respectively. Up to four orders of magnitude higher hydraulic
conductivities were obtained with respect to DIW when the GCLs were permeated directly with
landfill leachates (represents the non-prehydrated conditions). However, when the GCLs were
prehydrated and prepermeated with DIW, the final hydraulic conductivities of GCL-1 and GCL-2 to
LL-1 were 6.8×10-10 cm/s and 3.3×10-10 cm/s; whereas to LL-2 were 1.5×10-9 cm/s and 1.7×10-10
cm/s, respectively.

Keywords: flexible-wall permeameter, GCL, hydraulic conductivity, landfill leachate,

prehydration

1. INTRODUCTION geotextiles or bonded to a geomembrane

(Koerner, 2005). It is known that the
Geosynthetic clay liners (GCLs) have
hydraulic conductivities of GCLs depend
been used as hydraulic barriers in waste
mostly on the hydraulic performance of
containment facilities. Due to their low
bentonite layer. Some of the most
thickness, easy handling and installation,
important factors that affect the hydraulic
low hydraulic conductivity to water
properties of GCLs are pore fluid
(<2.0×10-9 cm/s) and self-healing
chemistry and prehydration condition
properties, GCLs became an alternative
applied during the test. Today, it is well
material when compared to compacted
known that the hydraulic conductivities of
clay liners.
GCLs increase depending on the chemical
GCLs are consisted of thin bentonite composition of the solutions used in the
layer between a woven and non-woven tests. Petrov and Rowe (1997); Ruhl and
* presenting author
Daniel (1997); Joe et al (2001); Kolstad et
al. (2004); Joe et al. (2005); Katsumi et al.
(2007) and Benson et al. (2010)
investigated the factors affecting the GCL
hydraulic conductivity such as cation
valence, pH and concentration of the
permeant liquid. In these studies; the
effect of pore fluid chemistry is more
pronounced when GCLs are non-
prehydrated. On the other side, the
hydraulic conductivities of GCLs are
affected less from pore medium properties
when GCLs are prehydrated with water.
This is due to swelling of bentonite
particles while permeating with water
which reduce the flow paths across the
GCL even pore fluid chemistry is changed.
The combined influence of permeating
liquid and prehydration conditions have
been widely investigated using standard
or synthetic solutions (Petrov and Rowe,
1997; Shackelford et al., 2000; Kolstad et
al. 2004; Lee and Shackelford, 2005;
Katsumi et al., 2007). However, there is
little information about the influence of
prehydration condition on hydraulic
conductivity of GCLs when landfill
leachates are used (Ashmawy et al., 2002;
Shan and Lai, 2002). Thus, the aim of this
study is to investigate and discuss the
influence of prehydration conditions on the
hydraulic performance of GCLs which
were permeated with real landfill
leachates.
2. MATERIALS AND METHODS
2.1. Materials
In the content of this study, two local
needle-punched GCLs (GCL-1 and GCL-
2) were used. Bentonites in the GCLs
were both in granular form. The initial
water contents of bentonites were 11% for
GCL-1 and 22% for GCL-2. In order to
determine the particle size distributions of
bentonites, wet-sieving method was
performed in accordance with ASTM
D422-63. Based on the particle size
analysis, bentonites contain 17% and 7%
sand size grains for GCL-1 and GCL-2,
respectively. The fine contents and clay
contents were 83% and 57% for GCL-1
418
and 93% and 67% for GCL-2. The liquid
limits of bentonites were determined as
108% for GCL-1 and 310% for GCL-2
which were determined by using
Casagrande device (ASTM D4318-10).
Hydraulic conductivity tests were
conducted using deionized water (DIW)
and two landfill leachates (LL-A and LL-B)
as permeant liquid. Deionized water was
collected from Milli-Q Gradient water
purification system. Landfill leachates
were taken from two different landfills in
Turkey. Some chemical properties such as
cation concentrations (i.e. Na+, K+, Mg2+,
Ca2+), pH and electrical conductivities
(EC) of landfill leachates are given in
Table 1.
Table 1. Landfill leachate properties
LL-A LL-B
Na+ (mg/L) 1011 2903
K+ (mg/L) 892 1135
Mg2+ (mg/L) 92 265
Ca2+ (mg/L) 82 101
pH 8.0 7.6
EC (mS/cm) 13.3 20.8
2.2. Methods
The hydraulic conductivity tests were
performed with flexible wall permeameters
in accordance with the ASTM D 6766-12.
The falling head constant tail water
method was employed. During the tests,
the average effective stress was 90 kPa
and the hydraulic gradient was around
200. The flow was from top to bottom and
the outflow was open to the atmosphere.
No backpressure was applied.
Hydraulic conductivity tests were
performed for both GCLs in non-
prehydrated and prehydrated-
prepermeated conditions. GCLs in the
non-prehydrated condition were directly
exposed to landfill leachates whereas
GCLs in the case of prehydrated
conditions were soaked in (or hydrated
with) DIW for 48 hours before the
permeation. At the end of the hydration,
outflow valve was opened and GCLs were
permeated with DIW until at least 1 pore
volume of flow (PVF) passed across the overnight even in the case of non-
sample. After that DIW in the inflow and prehydrated condition. However,
outflow tubings were removed and the crystalline swelling is expected for
hydraulic conductivity test was continued bentonite particles, because particles
using landfill leachates as permeant liquid. swell negligible amount. Thus, hydraulic
Henceforth, prehydration term will be used conductivity increased in the subsequent
to simulate the prehydrated-prepermeated permeation during the day. Figure 1a
condition with DIW. reveals that stabilized hydraulic
Total of 10 hydraulic conductivity tests conductivity values were obtained after
were conducted throughout the study. The each interruption.
tests were terminated when outflow and On the other hand, the hydraulic
inflow rate (Qout/Qin) and four consecutive conductivity of GCL-2 with LL-A initially
hydraulic conductivity measurements were decreased and then became stable during
within 1±0.25. the test (Figure 1b). The final hydraulic
conductivities of GCL-1 and GCL-2 with
3. RESULTS AND DISCUSSIONS LL-A were recorded as 9.5×10-6 cm/s and
The tests were initially run with deionized 2.4×10-6 cm/s, respectively.
-3 4.0
water (DIW) for the reference. The Hydraulic Conductivity (cm/s) 10
GCL-1 HC
hydraulic conductivities of GCL-1 and 10
-4
Q /Q
out in
Valves closed 3.0
GCL-2 with DIW were 1.2×10-9 cm/s and -5
1.4×10-9 cm/s, respectively. To determine

Q /Q
10

out
2.0
the influence of prehydration conditions on 10
-6 Restart the test

in
the GCL hydraulic conductivity, GCLs -7
1.0
10
were tested for non-prehydrated and (a)
prehydrated conditions with landfill 10
-8
0 5 10 15 20 25 30
0.0
35
leachates. The test results were Pore Volume of Flow (PVF)
expressed as a function of pore volume of 10
-3 4.0
Hydraulic Conductivity (cm/s)

flows (PVFs). GCL-2 HC

-4 Q /Q
10 out in
3.0
3.1. Influence of landfill leachates on

Q /Q
-5
the non-prehydrated GCLs 10

out
2.0
The hydraulic conductivity behavior of 10
-6

in
non-prehydrated GCLs permeated with 10
-7
1.0

LL-A is shown in Figure 1. In this figure, (b)

-8 0.0
closed markers indicate the hydraulic 10 0 4 8 12 16 20 24 28 32
conductivities and open markers shows Pore Volume of Flow (PVF)

the volumetric flow ratio (Qout/Qin). Dashed
lines are the upper and lower limits of
volumetric flow ratio suggested in ASTM
D6766-12.
Figure 1a demonstrates the hydraulic
conductivity of GCL-1 permeated with LL-
A. Although the flow was rapid during the
test, constant hydraulic conductivity was
not achieved at the end of the day.
Eventually, inflow valve was closed at the
end of the day and the test was restarted
in the following morning. When the valve
was open in the following day, a sudden
decrease was observed in the hydraulic
conductivity. This decrease may be due to
swelling of bentonite particles during
419
Figure 1. Hydraulic conductivity of non-
prehydrated GCLs to landfill leachate-A
(LL-A): a) GCL-1; b) GCL-2
The hydraulic conductivity of GCLs
which were directly exposed to LL-B are
shown in Figure 2. As seen in the Figure
2a, the hydraulic conductivity of GCL-1
with LL-B was initially 7.5×10-6 cm/s and
then decreased to 3.5×10-7 cm/s at the
end of the test. However, almost stable
hydraulic conductivities (no significant
changes in the hydraulic conductivities)
were measured for GCL-2 (Figure 2b).
Thus, the final hydraulic conductivity of
GCL-2 with LL-B was 2.7×10-5 cm/s.
 -4
10 4.0 beginning of the test (Figure 3a). After 11
Hydraulic Conductivity (cm/s)

GCL-1 HC
-5
10
Q /Q
out in
PVF, the flow amount decreased and
3.0
became stable. Then, the permeant liquid
-6
was switched to LL-A. When GCL-1 was

Q /Q
10

out
2.0
-7
10 started to permeate with LL-A, the

out
-8 1.0 hydraulic conductivity did not change until
10
(a)
the end of the experiment.
-9
10
0 2 4 6 8 10 12 14 16 18
0.0
20 The test duration for GCL-2 to DIW was
Pore Volume of Flow (PVF) almost 1 PVF. The final hydraulic
10
-2 4.0 conductivity was around 1.6×10-10 cm/s
Hydraulic Conductivity (cm/s)

-3
GCL-2 HC
Q /Q
(Figure 3b). However, when the permeant
10
liquid was changed to LL-A, the hydraulic
out in
3.0
-4
conductivity of GCL-2 suddenly increased

Q /Q
10

out
-5
2.0
to 1.0×10-9 cm/s. Then, it decreased
10
in slightly along the test duration and
1.0
10
-6
reached a final value of 3.3×10-10 cm/s.
(b)
-7 0.0 10
-3 4.0

Hydraulic Conductivity (cm/s)

10
0 2 4 6 8 10 12 14 16 18 -4 GCL-1 HC
10
Pore Volume of Flow (PVF) Q /Q
10
-5 out in
3.0
DI
Figure 2. Hydraulic conductivity of non- 10
-6
LL-A

Q /Q
prehydrated GCLs to landfill leachate-B -7 2.0

out
10
(LL-B): a) GCL-1; b) GCL-2 10
-8

in
-9 1.0
When the landfill leachates were used 10
-10
10
as the first wetting liquid, up to 4 orders of -11
(a)
0.0
10
magnitude higher hydraulic conductivies 0 2 4 6 8 10 12 14 16 18 20
Pore Volume of Flow (PVF)
were obtained for GCLs with respect to
-8
DIW. The high hydraulic conductivities 10 4.0
Hydraulic Conductivity (cm/s)

GCL-2 HC
(~10-6-10-5 cm/s) were attributed to the Q /Q
out in
3.0
prehydration condition. Direct exposure of 10
-9

landfill leachates on GCLs resulted in

Q /Q
DI
2.0

out
LL-A
cation exchange between the GCL -10
10

in
bentonite and the permeant. The cations 1.0
in the pore fluid bombarded the bentonite (b)
surfaces and the particles exhibited only 10
-11
0 1 2 3 4 5 6
0.0
crystalline swelling which is not enough to Pore Volume of Flow (PVF)
close the flow paths. Figure 3. Hydraulic conductivity of
3.2. Influence of landfill leachates on prehydrated GCLs to landfill leachate-A
the prehydrated GCLs (LL-A): a) GCL-1; b) GCL-2
Due to rapid flow for non-prehydrated The hydraulic conductivity behaviors of
GCLs, GCL-1 and GCL-2 were prehydrated GCL-1 and GCL-2 with LL-B
prehydrated and then permeated with DIW are shown in Figure 4a-b. As seen from
until stabilized hydraulic conductivities Figure 4a, greater hydraulic conductivity
were achieved. After that, DIW was values were obtained for GCL-1 until 6.5
removed from the system and GCLs were PVF. This behaviour is consistent with the
permeated with landfill leachates. The hydraulic behavior of GCL-1 with LL-A
tests were generally lasted 5-9 months. (Figure 3a). In order to obtain stabilized
Figure 3 shows the hydraulic behaviors hydraulic conductivities for GCL-1,
of prehydrated GCLs with LL-A. Since permeation with DIW was prolonged up to
bentonite particles in GCL-1 were 8 PVF. When the permeant solution was
granular, high hydraulic conductivity switched to LL-B at 9 PVF, there was no
significant change on the GCL hydraulic
values (~10-5 cm/s) were observed at the
conductivity.
420
 Figure 4b shows the hydraulic behavior hydraulic conductivities of GCL-1 and
of GCL-2 with LL-B. As seen in the Figure GCL-2 to LL-A were 6.8×10-10 cm/s and
4b, GCL-2 was permeated with DIW up to 3.3×10-10 cm/s; whereas to LL-B were
2.5 PVF. The hydraulic conductivity in the 1.5×10-9 cm/s and 1.7×10-10 cm/s,
initial state of permeation is comparable respectively.
with the reference value that was The comparison of the final hydraulic
measured with DIW as well. The hydraulic conductivities of GCLs with respect to
conductivity slightly decreased when the prehydration condition is also shown in
permeant liquid was changed from DIW to Figure 5. As seen from Figure 5, hydraulic
LL-B. Note that, some discontinuities were conductivities of prehydrated GCLs were
observed on the hydraulic conductivity significantly lower than that of non-
(Figure 4b). These discontinuities were prehydrated GCLs.
possibly due to bacteria induced gas -4 DI Non-prehydrated Prehydrated
generation in the tubings. After a while, 10

Hydraulic Conductivity (cm/s)

Landfill Leachate-A (LL-A) Landfill Leachate-B (LL-B)
many air bubbles were seen in the tubings 10
-5

which obstruct the flow across the GCL. -6

10
Once this was observed, the tubings on
the flexible wall permeameters were 10
-7

flushed to remove air bubbles. Then, the -8

10
tests were initiated after adding fresh
landfill leachate to the tubings. 10
-9

-3 4.0
Hydraulic Conductivity (cm/s)

10 -10
10
GCL-1
10
-4 GCL-1 GCL-2 GCL-1 GCL-2
10
-5 DI 3.0
Figure 5. Comparison of the hydraulic
-6 LL-B
10 conductivities of GCLs as a function of
Q /Q

-7 2.0
out

10 prehydration condition
-8
10
in

10
-9 1.0 The prehydration with DIW allowed
10
-10 bentonite particles to swell and obstruct
(a)
10
-11 0.0 the flow paths across GCLs. Further
0 2 4 6 8 10 12
Pore Volume of Flow (PVF) permeation with landfill leachates did not
10
-7 4.0 change the hydraulic conductivity,
Hydraulic Conductivity (cm/s)

GCL-2 HC
Flush Q /Q because the cation exchange between the
DI out in

10
-8 LL-B Flush 3.0 pore fluid and the bentonite was not
Flush completed. Thus, the crystalline and
-9 2.0
Q /Q

10 osmotic swelling of bentonite particles

out

governed the hydraulic behavior. In

inn

10
-10
10
-11
0 1 2 3 4 5
Pore Volume of Flow (PVF)
Figure 4. Hydraulic conductivity of
prehydrated GCLs to landfill leachate-B
(LL-B): a) GCL-1; b) GCL-2
Based on the findings above, the
similar hydraulic conductivities were
achieved as with the reference values of
the GCLs in the first phases of the
permeations (i.e. DIW). In the second
phases (i.e. landfill leachates), the
hydraulic conductivities of GCLs were not
affected from the detrimental effects of the
landfill leachates. That is, the final
1.0
contrast, bentonite particles were not able
(b)
0.0 to swell when GCLs were directly
6 7 8 9
subjected to landfill leachates. Since only
crystalline swelling appears in non-
prehydrated GCLs, generally greater
hydraulic conductivities were obtained.
4. CONCLUSIONS
The findings of this study are summarized
below:
x The prehydration condition has a deep
impact on the GCL hydraulic conductivity.
Up to 4 orders of magnitude greater
hydraulic conductivies were obtained for
non-prehydrated GCLs when compared to

421
the hydraulic conductivities to DIW. It is
showed that the chemical composition of
the landfill leachates have detrimental
effect on the GCL hydraulic conductivity.
x The hydraulic conductivities of
prehydrated GCLs were low (~10-10 cm/s)
even they were further permeated with
landfill leachates. Pre-hydrating the GCLs
with DIW resulted in swelling of bentonite
particles and restricting the flow channels
across the GCL. Thus, the hydraulic
conductivity decreased even the GCLs
subsequently permeated with landfill
leachates.
ACKNOWLEDGEMENTS
This study was supported by the Scientific
and Technical Research Council of
Turkey, TUBITAK (Grant No: 111M718).
The authors appreciate TUBITAK for this
support. The authors also thanks their
supervisor, Assoc. Prof. Ali Hakan Ören,
for his valuable contributions and
recommendations.
REFERENCES
Ashmawy, A. K., El-hajji, D., Sotelo, N., &
Muhammad, N. (2002) “Hydraulic
performance of untreated and polymer-
treated bentonite in inorganic landfill
leachates”, Clays and Clay Minerals, 50
(5), 546-552.
ASTM (2007-2012) The American Society
for Testing and Materials, West
Conshohocken, United States.
Benson, C. H., Ören, A. H., & Gates, W.
P. (2010) “Hydraulic conductivity of two
geosynthetic clay liners permeated with
a hyperalkaline solution”, Geotextiles
and Geomembranes, 28, 206-218.
Jo, H. Y., Katsumi, T., Benson, C. H., &
Edil, T. B. (2001) “Hydraulic
conductivity and swelling of non-
prehydrated GCLs permeated with
single-species salt solutions”, Journal of
Geotechnical and Geoenvironmental
Engineering,127(7),557-567.
Jo, H. Y., Benson, C. H., Shackelford, C.
D., Lee, J. M., & Edil, T. B. (2005)
“Long term hydraulic conductivity of a
422
geosynthetic clay liner permeated with
inorganic salt solutions”, Journal of
Geotechnical and Geoenvironmental
Engineering,131(4), 405-417.
Katsumi, T., Ishimori, H., Ogawa, A.,
Yoshikawa, K., Hanamoto, K., &
Fukagawa, R. (2007) “Hydraulic
conductivity of non-prehydrated
geosynthetic clay liners permeated with
inorganic solutions and waste
leachates”, Soils and Foundations, 47
(1), 79-96.
Koerner, R. M. (Ed). (2005) “Designing
with Geosynthetics (5th ed.)”, Prentice
Hall, Upper Saddle River, NJ.
Kolstad, D. C., Benson, C. H., & Edil, T. B.
(2004) “Hydraulic conductivity and swell
of nonprehydrated geosynthetic clay
liners permeated with multispecies
inorganic solutions”, Journal of
Geotechnical and Geoenvironmental
Engineering, 130 (12), 1236-1249.
Lee, J. M., & Shackelford, C. D. (2005)
“Impact of bentonite quality on hydraulic
conductivity of geosynthetic clay liners”,
Journal of Geotechnical and
Geoenvironmental Engineering, 131(1),
64-77.
Petrov, R. J., Rowe, R. K. & Quigley, R.
M. (1997) “Selected factors influencing
GCL hydraulic conductivity”, Journal of
Geotechnical and Geoenvironmental
Engineering, 123 (8), 683-695.
Ruhl, J. & Daniel, D. E. (1997)
“Geosynthetic clay liners permeated
with chemical solutions and leachates”,
Journal of Geotechnical and
Geoenvironmental Engineering, 123(4),
369-381.
Shackelford, C. D., Benson, C. H.,
Katsumi, T., Edil, T. B., & Lin, L. (2000)
“Evaluating the hydraulic conductivity of
GCLs permeated with non-standard
liquids”, Geotextiles and
Geomembranes, 18, 133-161.
Shan, H. Y., & Lai, Y. J. (2002) “Effect of
hydrating liquid on the hydraulic
properties of geosynthetic clay liners”
Geotextiles and Geomembranes, 20,
19-38.
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Geotextile encased columns – verification of the analytical

design method

Katarina SIRK1*
1
University of Ljubljana, Faculty of Civil and Geodetic Engineering, Chair of Soil
Mechanics, Ljubljana, SLOVENIA

ABSTRACT
There exist numerous methods for the design of geotextile encased columns and for the prediction
of settlements of improved soil. Numerical ones may be more accurate, but at the same time far
less convenient for use. Thus, a derivation of an analytical method, which would be in good
agreement with the numerical solution, would be more appropriate. In this paper the analytical
method for calculating settlement reduction and stresses in soil and column for one soft soil layer
proposed by Pulko et al. (2011) was modified to calculate settlements of multiple soft soil layers
and respective stresses in soil and column. The method is based on the unit cell assumption and
considers column as an elasto-plastic material, while soil and geotextile are treated as elastic
materials. Documented case studies from literature were investigated and used for the validation of
the analytical method. Measured settlements for each case were compared to settlements
calculated using the analytical method. The method verification presented in the paper is helpful in
predicting the right values for the critical design parameters that are difficult to measure in situ or in
the laboratory.

Keywords: GEC, geotextile encased columns, stone columns, soil improvement,

settlement reduction, analytical solution, elasto-plastic model

geotextile encased stone columns

1. INTRODUCTION
In the past 20 years ground improvement
technique of installing geotextile encased
stone columns to reduce settlements has
become an established practice in a
variety of projects all over the world.
Experiences prove them to be very useful,
even when post construction settlements
pose considerably stringent limitations.
This paper focuses on an analytical
method for calculating settlements of
multilayered ground improved with
* presenting author
(GECs). The method is derived from the
analytical method proposed by Pulko et al.
in 2011, which was in its final form
developed for a single layer of improved
soil. A detailed description of the proposed
method’s principal together with the most
important equations is presented.
Next, the paper focuses on
documented case history projects
presented by Alexiew & Raithel (2015).
Four projects are described and soil and
column parameters, which were included
in the calculations, are presented. All
cases are analyzed according to the newly
proposed method. Results and
conclusions are presented individually,
compared to each other and discussed.
2. ANALYTICAL METHOD FOR
CALCULATING SETTLEMENTS OF
GROUND WITH GEC
2.1. General description
The analytical method for calculating
settlements of multilayered ground
improved by GEC is based on a
commonly known “Unit cell” concept
(Figure 1) introduced by Priebe (1976).
Soil is considered as an elastic material
and column as an elasto-plastic material
(Balaam and Booker, 1985). The method
assumes drained condition, since the
columns assure rather quick consolidation
of the surrounding soil.
Figure 1: Unit Cell (Pulko et al., 2011)
For the purposes of this analysis the
original method proposed by Pulko et al.
(2011) was extended to take into account
multilayered ground. The proposed
method is user friendly, since new layers
of soil can be easily added and all the
424
respective parameters can be taken into
account.
Beside previously mentioned
assumptions, the method also takes into
account:
x The load is assumed rigid, therefore
vertical settlements of soil and column are
supposed to be equal.
x Settlements of the bearing ground are
neglected.
x The soil remains in elastic state.
x Stone column is modeled as an
elastoplastic material, satisfying Mohr-
Coulomb’s yield criterion with constant
dilation angle ȥ.
x Geotextile encasement is modeled as
an elastic material with constant
deformation modulus (Pulko et al., 2011)
2.2. Basic equations for one layer
ground
In order to describe the response of stone
column and the surrounding soil under the
applied load qA the basic principles are
presented below. The method proposed
by Pulko et al. (2011) is based on
equations for stress and strain increments
for the elastic and elasto-plastic response
of the slice of “Unit cell”, as given in Table
1.
2.3. Extension of the method for
multilayered ground
Extension of the method enables us to
calculate total settlements of improved
multilayered ground. For this purpose
ground is divided in layers of different
types of soil. each having its individual soil
properties. Each soil layer can further be
divided into sublayers, if necessary. For
each sublayer stresses and strains are
calculated by using equations from
Chapter 2.2 along with equations
presented below.
The stone column yield criteria is given by:
1  sin Mc V zc ,ini  V zce
K pc (1)
1  sin Mc V rc ,ini  V rce
ızc,ini and ırc,ini are initial vertical and
radial stresses in the stone column after
installation at a selected depth.
 Table 1: Basic equations for calculating stresses and strains (Pulko et al., 2011)
elastic response elasto-plastic response

­ V zc ½ ª Oc  2Gc  2Oc F º e D ª K pc ª¬ 2k0  K\  C2  T  º¼ º

® ¾
¯ V rc ¿ «O  2 O  G F » H z
 c c ¼ ^ ` «
C5 « 2k0  K\  C2  T  »
» ^q A `
¬ c ¬ ¼

ª FAr º
­ V zs ½ « Os  2Gs  2Os (1  A ) »
1 ª D  C1 K\  2   Eoed  C3  T  º
® ¾
¯ V rs ¿
«
« 2 Ar F §
r

Gs ·»
^ `
» He
z «
C5 « DC2 K\  2 Dk0  Eoed Tk0 »
» ^q A `
« Os  ¨ Os  Gs  ¸» ¬ ¼
«¬ (1  Ar ) © Ar ¹ »¼

q A ª¬ 2 D  Eoed  C2  T  º¼
qA
Hz
 Oc  2Gc  Ar   Os  2Gs 1  Ar   2 Ar  Oc  Os  F C5 Eoed

q A  DK\  k0 Eoed 
Hr F H ze
C5 Eoed

FR J H re J H rp

Table 2: Equations for abbreviations used in Table 1

 Oc  Os 1  Ar 
2 ª¬ Ar  Os  Gs  Oc  Gc   Oc  Gc  Gs º¼  1  Ar  2Gs  Os  T
F

J
T
Eoed rc
Ec
D 2  K\ K pc  2Xc 1  K pc  K\ 

1  2Xs  Ar
C2
1  Ar 1  Xs 
C5 Eoed 1  Ar  C3  T   D ª¬1  Ar   C1 K\  2   Ar K pc  K\  C2  T   2k0  º¼

2k0 Ar
C1
1  Ar

C3 C2  k0C1

When the applied load qA exceeds the
yield load, Eq. (1) becomes:
V zc ,ini  V zc J c z  V zc
K pc
V rc ,ini  V rc K iniJ s ' z  V rc
where stress increments represent the
sum of elastic and plastic response at the
depth of interest. Until the criteria (Eq. (1))
(2) is met, the stress state of the column will
remain in the elastic state. When the
criteria (Eq. (1)) is violated, the stress and
425
strain state can be determined as a linear Finally, the total settlement of improved
combination of elastic and elasto-plastic multilayered ground, which can in general
solutions (Table 1) with regard to the yield be expressed as
criteria given by Eq. (2). In the latter case H
the applied load can be presented as a uz ³ H dz
z (9)
linear combination of elastic (e) and plastic 0

(p) load (Eq. (3)), where the plastic part of can be obtained numerically by summing
the load induces plastic strains inside the up the strain contributions at various
stone column. preselected depths (usually at 0.5 or 1 m).
qA qe  q p 1  G  q A  G q A (3)
3. MODEL VERIFICATION BASED
0 < į < 1 represents the proportion of ON CASE HISTORY DESCRIPTION
the plastic load. Following this principle
the following relations for stresses can be Designing GEC is a great challenge due
obtained: to many material properties of the soil and
column which can be difficult to determine.
'V ij  
'V ije q e  'V ijp q p   (4)
Back calculation of well documented case
1  G  'V  q A   G'V ijp  q A 
e histories is useful and sometimes a
ij
necessity in the prediction of reinforced
where i = {z, r} and j = {c, s}. ground behaviour.
By using equations for elastic and In this chapter measured settlements of
elasto-plastic stress increments from four cases described by Alexiew & Raithel
Table 1, Eq. (1) and Eq. (2) can be (2015) were back calculated with the
rewritten: proposed analytical method. Parameters
1  G  'V zce  G'V zcp  V zc ,ini  for soil were mainly given in the original
(5) paper, while stone column parameters Ec
 K pc ª¬1  G  'V rce  G'V rcp  V rc ,ini º¼ 0 and ijc’ were assumed and varied to
approach the settlements measured in the
and solved for į:
field.
­ 0 ½ The stone column data (Ec and ij’)
° °
G ® V rc ,ini K pc  V rc K pc  V zc  V zc ,ini ¾ (6)
e e
found in the literature are often back
° Ve K  V pK  Ve  V p ° calculated based on the past case
¯ rc pc rc pc zc zc ¿
histories. Estimations for stone column
V zce  V zc ,ini modulus Ec range from 7 up to 58 MPa,
When G 0 for d K pc and else with stone shear angle ijc’ ranging
V rce  V rc ,ini
between 30° for sand columns and up to
V zce  V zc ,ini 50° for stone columns (Barksdale &
when ! K pc .
V rce  V rc ,ini Bachus, 1983). Another restriction
concerning GEC elastic modulus is
Once į is determined, stresses ѐıij in mentioned in literature as a ratio Ec/Es.
the soil/column can be calculated as a Authors normally use a value between 10
linear combination of solutions, presented and 40. (Pulko et al., 2011), (Sexton et al.,
in Table 1 by using Eq. (4). Total stresses 2014).
are calculated as:
3.1. Bastions West, Netherlands
V ij V ij ,ini  'V ij (7)
3.1.1. General description
Vertical strain H z can be obtained similarly The landscape embankment was built on
for any chosen depth: very soft soil in a new residential area in
Hi 1  G  H ie  qA   GH ip  qA  (8) Houten-Zuid. The predicted settlements of
non-treated ground were between 1.6 and
1.9 meters, which was unacceptable due
to lack of time for such an extended

426
consolidation and also because the and was assumed to be infinitely wide.
adjacent building’s foundation would be Then Eoed was calculated by the following
endangered (Alexiew & Raithel, 2015). equation (Hardening Soil Model):
m
§ c ' ctgM s ' V ' ·
Eoed Eoed , ref ¨ (10)
¨ c ' ctgM s ' pref ¸¸
© ¹
Eoed,ref represents a reference value of
oedometer modulus of soil at a reference
value of stress pref = 100 kPa (Brinkgreve
R.B.J. et al, 2011).
Properties of columns are shown in
Table 4. Stone columns at Bastions West
were installed using displacement
Figure 2. Embankment at Bastions West one technique and filled with sand, which led
year after construction (Huesker, 2013) to significantly lower values of the friction
3.1.2. Soil and column properties angle in comparison to the ones filled with
gravel. Alexiew & Raithel (2015) proposed
The soft ground on which the a value of ij’ = 32.5°.
embankment was built consists of 7.5
meters of organic clay and peat. The Two values were given for tensile
parameters given by Alexiew & Raithel stiffness of geotextile. Jd represents a
(2015) are listed in Table 3. value for long term stiffness and Jk a value
for short term stiffness.
The load of 93.5 kPa was assigned
from the embankment height of 5.5 meters
Table 3: Soil properties at Bastions West project
Depth Ȗs’ Eoed,ref pref c’ ijs’ m Ȟ
Soil layer [m] [kN/m3] [kPa] [kPa] [kN/m2] [°] [-] [-]
Organic 7.5 4 2000 100 20 2 1 0.4
clay&peat

Table 4: Properties of geotextile encased stone columns at Bastions West project

Depth [m] d [m] Ȗc’ [kN/m3] Ȟ [-] Ar [%] Jd / Jk [kN/m]
GEC 7.5 0.8 9 0.3 15 2100/3500

3.1.3. Results and discussion combinations of stone column stiffness

Settlements of improved ground were (Ec) and shear strength (ijc’) and various
calculated according to the analytical geotextile stiffness Jd and Jk (values in
method described in Chapter 2 and are brackets).
presented in Table 5. Values outside the
brackets were calculated for different
Table 5: Settlements of treated ground [cm] at Bastions West calculated with the proposed
analytical method
ijc’ [°] \ Ec [MPa] 8 10 12 15 20
32.5 31 (29) 30 (27) 29 (26) 28 (25) 27 (23)
35 30 (28) 28 (26) 27 (25) 26 (23) 25 (21)
40 28 (27) 26 (24) 24 (22) 23 (21) 21 (19)

427
Measured settlements at Bastions West
were 30 cm on top of GEC and 32 cm in
between GEC. It can be seen that by
using modest values of Ec and ijc’ good
agreement between calculated and
measured settlements can be achieved.
The modest values of Ec and ijc’ are also
in good agreement with the values
proposed by Alexiew & Raithel (2015),
hence with values expected to mobilize
inside a sand column.
3.2. Railroad embankment Bothnia line,
Sweden
3.2.1. General description
The next case study is from Sweden,
where a 190 km long high-speed railway
line runs along Bothnia Bay. The route
was opened in 2010 and it allows trains to
travel as fast as 250km/h.
Figure 3: A typical cross section at Bothnia
line GEC project (Alexiew & Raithel, 2015)
On a part of the route where the railway
line crosses a valley of very soft soils GEC
foundation was used to reduce great
settlements. A typical cross section of the
embankment is shown in Figure 3
(Alexiew & Raithel, 2015).
3.2.2. Soil and column properties
Soil properties used in the calculations are
given in Table 6. In the case of Bothnia
line, soil’s Eoed of each layer was
calculated using Hardening Soil Model
following Eq. (10) in Chapter 3.1.3.
The load was assumed to be infinitely
wide and was calculated from the
embankment height. The value of 230 kPa
was used for the analysis of settlements.
GECs were installed using the
displacement method and filled with
crushed rock basalt, very common for the
area. Thus, a high internal friction angle of
the column material was to be expected.
3.2.3. Results and discussion
Expected settlements of treated ground at
Bothnia project were a combination of
actual measurements and a prognosis
based on them which resulted in
approximately 44 cm. Based on the given
number and the type of rock that was
installed in stone columns Alexiew &
Raithel (2015) proposed an internal friction
coefficient of GEC ijc’ = 45°.

Table 6: Soil properties at Bothnia line project

Depth Ȗs’ Eoed,ref pref c’ ijs’ m Ȟ
Soil layer [m] [kN/m3] [kPa] [kPa] [kN/m2] [°] [-] [-]
Clay, silty 2.75 6 1300 100 3.5 30 0.5 0.4
Silt, clayey 1.25 6.5 1100 100 2.5 30 1 0.4
Clay, silty 1.00 5 700 100 2 30 1 0.4
Clay, silty 0.85 7 1200 100 1.75 30 0.9 0.4
Clay, silty 1.65 7 800 100 1.75 30 1 0.4

Table 7: Properties of geotextile encased stone columns at Bothnia Line project

Depth [m] d [m] Ȗc’ [kN/m3] Ȟ [-] Ar [%] Jd / Jk [kN/m]
GEC 7.5 0.8 9 0.3 15 2100/3500

428
Table 8: Settlements of treated ground [cm] at Bothnia line calculated with the proposed analytical
method
ijc’ [°] \ Ec [MPa] 10 15 20 25 30
35 62 54 50 47 45
40 57 48 43 40 38
45 56 44 38 35 32

In the calculations shear angle ijc’ and
column modulus Ec were varied to see
which combination gives calculated
settlements similar to those measured on
the site. The results are gathered in Table
8.
By looking at the results from a GEC
designer’s perspective, the best choice for
Ec according to the proposed value of ijc’
= 45° would be 15 MPa, which gives the
ratio of Ec/Es between 35 and 21 for the
stiffest and the softest soil layer,
respectively. These ratios are in
agreement with the common values
presented in the beginning of Chapter 3,
hence confirming our choice.
When designing GECs a bit more
conservatively, a combinations of 40° –
42.5° for ijc’ and 15 – 20 MPa for Ec would
also predict settlements in agreement with
the measured ones.
3.3. Hamburg, Germany
3.3.1. General description
The extension of Airbus site at
Mühlenberger Loch in Hamburg was the
biggest GEC project ever accomplished in
Germany until the end of the year 2015.
By enclosing an area of extremely soft
soils and building a 2.4 km long dyke to
protect the site form tide, around 60,000
Table 9: Soil properties at Hamburg Airbus site
Soil layer Depth [m]
Sludge 2.1
Clay 1.3
Peat 2.8
Clay 1.8
GECs were installed in a total length of
about 650 km. (Alexiew & Raithel, 2015)
Figure 4: Airbus site at Mühlenberger Loch in
Hamburg (Fit Fuer Innovation)
3.3.2. Soil and column properties
Properties for soil and GEC used in the
calculations are presented in Table 9 and
Table 10. In a typical cross section layers
of sludge, clay and peat exchange depth-
wise.
3.3.3. Results and discussion
Measured settlements at the end of
primary consolidation were approximately
105cm. Calculated settlements are shown
in Table 11. They are calculated for
different combinations of stone column
friction angle ijc’ and elastic modulus Ec.
Results close to the measured ones are
colored blue.
Ȗs’ [kN/m3] Eoed,s [kPa] Ȟ [-]
4 450 0.4
6 600 0.4
1 550 0.4
6 600 0.4

Table 10: Column properties at Hamburg Airbus site

Depth [m] d [m] Ȗc’ [kN/m3] Ȟ [-] Ar [%] J [kN/m]
GEC 8 0.8 10 0.3 15 2800
429
Table 11: Settlements [cm] of treated ground at Hamburg Airbus site calculated with the proposed
analytical method
ijc’ [°] \ Ec [MPa] 6 8 10 12 14
35 128 113 104 97 92
40 120 103 92 85 79
45 120 99 85 77 71

The results direct us into choosing
lower values of Ec (8 – 10 MPa) and
combining them to a friction angle
between 35° and 40°. Since the soil’s Es is
very low, ranging from 200 – 300 kPa,
lower values of Ec inside a stone column
are more realistically expected. In order to
effectively asses the material parameters
which ought to be chosen during project
planning, the creep should also be closely
investigated. Measured settlements at
Hamburg airbus site progressed with time
for further 35 cm due to creep behavior of
these extremely soft soils.
3.4. Jordanovo, Poland
3.4.1. General description
During construction works at a section of a
highway in Poland the first “State of the
art” project of GEC installation was
successfully executed. Maximum depths
of soft soil encountered were up to 28 m.
The upper 5 meters of the soil consisted of
Table 12: Soil properties at Jordanovo site
Soil layer Depth [m] Ȗs’ [kN/m3]
Peat 5 1
peat and bellow it a layer of sensitive soil
named gyttja was found (Alexiew &
Raithel, 2015).
Coupling a quite flat final embankment
geometry with the proposed highway
speed limit of 130 km, a serviceability limit
state was very strict and thus maximum
allowed post-construction settlements
were extremely low. Due to the mentioned
reasons a temporary preload was applied
to accelerate the consolidation process
(Alexiew & Raithel, 2015).
3.4.2. Soil and column properties
The parameters for soil and for columns
used in the model are gathered in the
tables below. The total load of 114 kPa
was calculated from the embankment
height. Since the value of the geotextile
modulus was not given in the original
source, two different values were adopted
for the analysis, i.e. 2500 kN/m and 3000
kN/m.
Eoed,s [kPa] Ȟ [-]
500 0.3

Gyttja 23 4 750 0.3

Table 13: Properties of geotextile encased stone columns at Jordanovo site

Depth [m] d [m] Ȗc’ [kN/m3] Ȟ [-] Ar [%] Jd / Jk [kN/m]
GEC 28 0.8 10 0.3 15 2100/3500

3.4.3. Results and discussion settlements are worth mentioning, since

In comparison to expected settlements of
230 cm, the maximal settlements
measured at Jordanovo site were only 105
cm. Besides a probable compression
modulus underestimation it was assumed
that the difference also occurred because
the settlements caused during the GEC
installation were not considered. These
due to the length of columns which led to
a time consuming installation process,
heavy equipment was stationed in one
position for a considerable amount of time
(Alexiew & Raithel, 2015). For this reason
the calculated highlighted settlements in
Table 14 include values that deviate up to
25% from those measured on site.

430
 Table 14: Settlements of treated ground [cm] at Jordanovo site calculated with the proposed
analytical method
ijc’ [°] \ Ec [MPa] 10 15 20 25 30
35 172 (164) 151 (143) 140 (131) 134 (124) 129 (119)
40 151 (146) 128 (122) 116 (109) 108 (100) 102 (95)
45 141 (139) 112 (108) 98 (93) 89 (84) 8 (77)

Due to many uncertainties in the material columns (40° – 45°, i.e. Bothnia
description of this project, it is difficult to Line), as suggested by Alexiew & Raithel
give a trustworthy conclusion. By (2015) and confirmed with our back
assessing calculated settlements and analysis.
simultaneously keeping in mind the For the future development of GEC it
suggested values for Ec and ijc’ from would be essential to monitor new GEC
Chapter 3, only rough assumptions can be projects carefully and to share the
drawn as in keeping the value of Ec under collected data with the scientific
15MPa and a value of ijc’ around 40°. community, since case history project
analysis offers an excellent opportunity to
4. CONCLUSIONS verify analytical methods and input
Based on analyzed case history projects it parameters.
can be concluded that the method for
ground improved with GEC is capable of NOTATION
yielding good settlement predictions. Like The following symbols are used in this
in all similar geotechnical situations it is paper:
clear that the selection of the input Subscripts / superscripts
parameters is of key importance for the
credibility of results. Therefore, their c, s column, soil
validation in the laboratory and on the field r, z radial, vertical coordinate
is crucial for effective and safe design of e, p elastic, plastic
GEC. ini initial value
From the analyzed cases we can Symbols
conclude that in extremely soft soils (like
in the Hamburg case) their very low Ar replacement ratio
stiffness modulus Es leads to relatively low c cohesion
stone stiffness Ec (between 6 and 8 MPa). C1, C2, C3, C5 material/geometrical
When the elastic modulus of soil constants
increases, so does the one in GEC, like in D material constant
the case of Bothnia line (Ec = 15 MPa).
Eoed eodometer modulus of soil
When analyzing stiffness modulus ratio
E elastic modulus of soil / column
Ec/Es, all results stay inside boundaries
found in literature (Ec/Es = 10 – 40) which F material/geometrical constant
indicates our choices of Ec are valid. The FR encasement hoop force
only exception is ratio values in the G shear modulus of soil / column
Jordanovo case. This could be explained J geosynthetic encasement stiffness
with already mentioned and very probable
underestimation of the soils’ modulus. k0 coefficient of earth pressure at rest
The expected shear strength of the Kini initial lateral pressure coefficient
column is much lower for sands (32°, i.e. after columns installation
Bastions West) than for stone (gravel) Kpc passive earth pressure coefficient

431
K\ dilation constant
T dimensionless encasement
stiffness
qA applied load
uz total settlement of improved ground
G proportion of the plastic load
J unit weight
O Lame’s parameter
H strain
V stress
V stress increment
X Poisson’s ratio
ACKNOWLEDGEMENTS
I am very grateful to my mentor Dr.
Boštjan Pulko for all his assistance and
invaluable guidance during my research.
I would also like to thank two University
colleagues, Žiga Šebenik and Dr. Matej
Maþek for their help while writing this
paper.
REFERENCES
Alexiew D., Raithel M. (2015) “Geotextile-
Encased Columns: Case Studies over
Twenty Years”. Ground improvement
case histories: Embankments with
special reference to consolidation and
other physical methods (Eds:
Indraratna, B., Chu, J., Rujikiatkamjorn,
C.), Elsevier, pp. 451-478.
Balaam, N.P., Booker, J.R. (1985) “Effects
of stone columns yield on settlement of
rigid foundations in stabilized clay”,
International Journal for Numerical and
Analytical Methods in Geomechanics
vol. 9, no. 4, pp. 331-351.
Barksdale, R.D., Bachus, R.C. (1983)
“Design and construction of stone
columns”, Report FHWA/RD-83/026,
National Information Service,
Springfield, Virginia.
Brinkgreve, R.B.J., Swolfs, W.M., Engin,
E. (2011) “PLAXIS 2D 2010 Material
Models Manual”, PLAXIS B.V.
Fit Für Innovation:
http://www.fitfuerinnovation.de/?page_id=
531 (photo downloaded: 20. 1. 2016).
Huesker (2013) “Geotextile-Encased
Columns, Earthworks and foundation
embankments: Design and construction
of geotextile encased columns
supporting geogrid reinforced
landscape embankment at Bastions
Vijfwal Houten in The Netherlands,
http://www.huesker.com/uploads/tx_itm
download/JR-Earthworks-and-
Foundations-Embankments-Geotextile-
Encased-Columns-Ringtrac-Houten-
NL_01.pdf (downloaded: 19. 2. 2016).
Mitchell, J. K. (1993) “Fundamentals of
Soil Behavior”, Second Edition, John
Wiley & Sons, Inc., New York, 437 p.
Pulko, B., Majes, B., Logar, J. (2011)
“Geosynthetic-encased stone columns:
Analytical calculation model”,
Geotextiles and Geomembranes, vol.
29, no. 1, pp. 29-39.
Priebe, H.J. (1976) “Evaluation of the
settlement reduction of a foundation
improved by Vibro-replacement”,
Bautechnik 2, pp. 160-162.
Raithel M., Kuester V., Lindmark A. (2004)
“Geotextile-Encased Columns - a
foundation system for earth structures,
illustrated by a dyke project for a works
extension in Hamburg”, Nordic
Geotechnical Meeting NGM 2004,
Ystadt, Sweden, http://www.kup-
geotechnik.de/files/2004_raithel_ystad.
pdf (downloaded: 15. 1. 2016).
Rowe, P.W. (1962) “The Stress-Dilatancy
Relation for Static Equilibrium of an
Assembly of Particles in Contact”,
Proc., Royal Soc., pp 500-527.
Sexton B., McCabe B.A., Castro J. (2013)
“Appraising stone column settlement
prediction methods using finite element
analyses”, Acta Geotechnica, vol. 9, pp.
993–1011.

432
 25th European Young Geotechnical Engineers Conference
st th
21 – 24 of June 2016, Sibiu, Romania

Author index

BARBIR Olja 237 GOMES Rui Carrilho 361

BISCAN Marko 303 GÖRÖG Péter 119

BOFFA Giacomo 21 GOUVEIA Fátima 361

BOIKO Igor 341 HAVINGA Michael 371

BRANDT Madeleine 99 HLEIBIEH Jamal 377

ýÁPOVÁ Kristýna 247 IGNATOV Sergey 341

CAROW Christian 257 JAKIMOVSKI Bojan 87

CHARLET Francois 61 KELLEZI Lindita 275, 383

DAM Asbjørn M. 191 KNAPIK Karolina 105

DE CAMILLIS Michela 401 KNUUTI Mika 201

DEMøRKIRAN Havva 417 KOPPMANN David 147

DI DONNA Alice 133 KORETA Olsi 383

DIMITRIEVSKI Teodor 87 LAKOýEVIû Katerina 325

DIZDAR Goran 303 LIETAERT Bert 61

DOAN Hong 349 LÖFMAN Monica 73

DOMINIJANNI Andrea 21 LOPES Isabel 361

DUNNE Helen P. 267 MANASSERO Mario 21

FRANK Roger 3 MATHEWS John 237

GĂINĂ Alexandra Alisa 355 MIESHKOVA Iryna 113

GILLERON Nicolas 139 MOLDOVAN Dorin Vasile 409

GàUCHOWSKI Andrzej 53 MOLINA Carlos 275

GÖKALP Alp 123 MONFORTE Lluís 285

433
MONSEREZ Sahin 153

NAGY Andor-Csongor 409

NAPPA Valeria 393

NEDELKOVSKA Natasha 309

OLINIC Ernest 37

ÖZDAMAR KUL Tu÷çe 417

PARERA Ferran 317

POKLONSKYI Sergii 83

PULKO Boštjan 423

RADU Cristian 163

RIPPI Katerina 209

ROSE Alexis 171

RUMENOVSKI Borche 87

SCHADEE Leon 217

SEXTON Brian G. 225

SIRK Katarina 423

STAELENS Peter 61

STOJANOVSKI Mario 325

STRØMME Terje 99

SZEPESHÁZI Attila 177

TASEVSKA Katerina 87

TEIXEIRA Ana 209

VINCZE Ádám 119

VLASTELICA Goran 331

WORREN Aleksander 293

YILDIZ Nejla 123

434
ISBN 978-973-100-421-1