Anda di halaman 1dari 38

212

CHAPTER 5 ONE-G SCALE MODELING


5.1 Introduction
The use of scale models in geotechnical engineering offers the advantage of simulating
complex systems under controlled conditions, and the opportunity to gain insight into the
fundamental mechanisms operating in these systems. In many circumstances (e.g., a static
lateral pile load test), the scale model may afford a more economical option than the
corresponding full-scale test. For other investigations (e.g., seismic soil-pile interaction),
scale model tests allow the possibility of simulating phenomena that cannot be achieved
at-will in the prototype. The practice of conducting parameter studies with scale models
can be used to augment areas where case histories and/or prototype tests provide only
sparse data. In addition to qualitative interpretation, scale model test results are often
used as calibration benchmarks for analytical methods, or to make quantitative predictions
of the prototype response. For such applications it is necessary to have a set of scaling
relations that relate the observed model and predicted prototype behavior.
This chapter will first describe theories of scale model similitude, and elucidate the
development of scale modeling criteria for the shaking table test program. The application
of these criteria to and design of the model soil and model piles used in the test program
will then be described.
213
5.1.1 Theories of Scale Model Similitude
The relationship between a scale model and the corresponding prototype behavior
is described by a theory of scale model similitude. Kline (1965) defines three methods of
increasing complexity and power for scale modeling applications. They are dimensional
analysis, similitude theory, and the method of governing equations. Dimensional analysis
consists of converting a dimensionally homogenous equation, containing physical
quantities and describing a physical phenomenon, into an equivalent equation consisting of
dimensionless products of powers of the physical quantities. Dimensional analysis may be
used exclusively to understand the form of the problem solution without application to
scale modeling. Similitude theory identifies the forces operating in the system and uses
dimensional analysis to construct and equate dimensionless terms for the model and
prototype. The scaling relations between model and prototype are also known as
prediction equations. The method of governing equations involves the transformation of
the differential equation describing the process to nondimensional form, and the formation
of similarity variables that relate model to prototype. Similarity variables must also be
determined for both initial and boundary conditions operating on the system.
Scale models can be defined as having geometric, kinematic, or dynamic similarity
to the prototype (Langhaar, 1951). Geometric similarity defines a model and prototype
with homologous physical dimensions. Kinematic similarity refers to a model and
prototype with homologous particles at homologous points at homologous times.
Dynamic similarity describes a condition where homologous parts of the model and
prototype experience homologous net forces. Scale models meet the requirements of
similitude to the prototype to differing degrees, and researchers apply nomenclature such
214
as true, adequate, or distorted to the model (Moncarz and Krawinkler, 1981). A
true model fulfills all similitude requirements. An adequate model correctly scales the
primary features of the problem, with secondary influences allowed to deviate; the
prediction equation is not significantly affected. Distorted models refer to those cases in
which deviation from similitude requirements distorts the prediction equation, or where
compensating distortions in other dimensionless products are introduced to preserve the
prediction equation.
Dimensional analysis in its simplest form proposes to reduce an engineering
parameter to its fundamental Mass-Length-Time measures of nature while developing
scale factors for each of the three quantities. For example, modulus of elasticity is a
measure of stress with units of force/area and dimensions ML
-1
T
-2
, so scale factors for
mass , length , and time are combined to form a scaling relation
1

2
that relates
model to prototype stress response. Following the same line of reasoning, strains map 1:1
between the model and prototype, as strain is a dimensionless quantity. Model and
prototype material densities are commonly used as a basis for determining the relation
between the and scale factors. If
model
/
prototype
= 1, then can be computed as equal
to
-1/3
.
The time scaling factor can then be derived by equating the inertial force ratio (m
subscripts refer to model quantities, and p subscripts refer to the prototype)

,
_

,
_

,
_

A
A
A M
A M
p
m
p
m
p p
m m

3
(5.1)
with the weight ratio
215

,
_

p
m
(5.2)
with the result that the model accelerations must equal the prototype accelerations.
Therefore

,
_

,
_

,
_

,
_

T
m
T
p
T
L
T
L
A
A
p
p
m
m
p
m
2
2
2
1 (5.3)
from which

,
_

T
p
T
m
2
(5.4) and (5.5)
With the mass , length , and time scale factors all determined in terms of , a
complete set of dimensionally correct scaling relations can be derived for all variables
being studied. This is the methodology followed by Clough and Pirtz (1956) and Seed
and Clough (1963), who used scale models to study the earthquake resistance of dams. A
drawback to this method is that each variable is treated independently without particular
regard to its function in the system.
A more sophisticated type of dimensional analysis involves application of the
Buckingham Pi Theorem, which states that any dimensionally homogeneous equation
involving certain physical quantities can be reduced to an equivalent equation involving
a complete set of dimensionless products. Thus the solution equation for some physical
quantity of interest, i.e.,
F(X
1
, X
2
,, X
n
) = 0 (5.6)
can be expressed in the form
G(
1
,
2
,,
m
) = 0 (5.7)
216
where the Pi terms are independent dimensionless products of the physical quantities X
1
,
X
2
,, X
n
. The number of dimensionless products (m) is equal to the number of physical
variables (n) minus the number of fundamental measures that are involved. The individual
Pi terms are formed by grouping the physical variables into dimensionless terms; all
variables must be included and the m terms must be independent. There is theoretically no
unique set of Pi terms for a given problem, but for scale modeling problems it is essential
that the correct variables be identified and the Pi terms be formed appropriately. Scaling
relations may then be determined by equating model and corresponding prototype Pi
terms, i.e.
i,m
must equal
i,p
. As previously noted, similitude theory attempts to describe
the problem more rationally by basing the formation of the Pi terms on the forces
prevailing in the system.
Moncarz and Krawinkler (1981) consider the formation of Pi terms for the
determination of the time history of stress components
ij
(r,t) in a scale model resultant
from an imposed acceleration time history a(t). They point out two requirements to meet
true scale modeling criteria: the Froude and Cauchy conditions. Given stress as a
function of
= F (r, t, , E, a, g, l,
o
, r
o
) (5.8)
where r = position vector, t = time, = density, E = modulus of elasticity, a =
acceleration, g = gravitational acceleration, l = length,
o
= stress, and r
o
= initial position
vector, the following Pi terms can be formed:

,
_

l
r
E E
gl
g
a E
l
t
l
r
E
o o
, , , , , ,

(5.9)
217
In 1-g scale modeling, the dimensionless product a/g (Froudes number, commonly
expressed as v
2
/lg) must equal unity, which implies that the ratio of model to prototype
specific stiffness (E/) equals the geometric scaling factor . This is known as the Cauchy
condition and can also be stated in terms of shear wave velocity:
( )
( )

V
s
V
s
m
p
(5.10)
Moncarz and Krawinkler also show that the Cauchy condition is a necessary
requirement for simultaneous replication of restoring forces, inertial forces, and
gravitational forces in a dynamic model system. The difficulty in designing true scale
models lies in selecting model materials that have a combination of both small modulus
and large mass density to meet the Cauchy condition. Two alternatives are to conduct
what the authors call artificial mass simulation and gravity effects ignored scale model
tests.
5.1.2 Scale Model Similitude As Applied to Soil Mechanics
Rocha (1957) was the first to systematically describe scale modeling for problems
in soil mechanics. He differentiated between total stress and effective stress conditions,
deriving separate similitude relations for each case. To account for the different stress
regime present in a 1-g scale model from the prototype, Rocha proposed that the soil
constitutive behavior be scaled, and therefore assumed that both the stress and strain held
a linear relationship between the model and prototype. This concept is illustrated in
Figure 5.1, where is the stress scaling factor and is the strain scaling factor (Note that
scaling strain is contrary to the dimensional analysis approach). He limited his derivations
218
to elastic deformations, and opined that the analysis becomes insuperably complicated
when nonlinear response is considered.
Figure 5.1 - Scale Model Constitutive Behavior Described by Stress and Strain Scaling
Factors (after Rocha, 1957)
Figure 5.2 - Critical State Soil Mechanics Concept of Geometrically Similar Stress Paths
for Prototype A
1
Z
1
and Model A
2
Z
2
(after Roscoe, 1968)
Roscoe (1968) investigated the difficulty of scale models replicating prototype
constitutive behavior for soils whose response is dependent on self-weight, i.e., confining
pressure. He extended Rocha's assumptions and recast them in the form of critical state
soil mechanics, asserting that the strain behaviours of two elements of soil will only be
identical when the elements are subjected to two geometrically similar stress paths if
their initial states on an e-ln ' plot are equidistant from the critical state line. This
219
theory is illustrated in Figure 5.2, and was substantiated by limited laboratory testing. He
also observed that centrifuge testing presented a potentially viable alternative to such an
approach.
Figure 5.3 - Tangent Modulus Formulation for Scale Modeling of Soil Constitutive
Behavior (after Iai, 1989)
Kana et al. (1986) describe the application of the Buckingham Pi theorem to the
problem of scale modeling the dynamic interaction of a pile in clay. They formed the
following nondimensional equation

,
_

g
D
T
EI
D M
EI
D F
D E
EI
D E
EI
D
M
M D
J
DM
M
D
y
D
x
o
l r
c c

2 2 4 2
4 4 2 3
, , , , , , , , , (5.11)
where x = lateral deflection, D = pile diameter, M
c
= pile cap mass, M = pile mass per unit
length, J
c
= pile cap moment of inertia, = soil density, E = pile Youngs modulus, I = pile
section moment of inertia, E
r
= soil storage modulus, E
l
= soil loss modulus, F = applied
lateral load, = frequency of oscillation, T
o
= linear freuency sweep duration, and g =
acceleration due to gravity, which implies the requirement for an elevated gravity field in
the last term. But the researchers surmised that gravity effects were negligible for lateral
pile response in overconsolidated clay, and therefore conducted their tests under this
220
scaling regime but in a 1-g environment. Their findings demonstrated the noneffect of
gravity for the particular conditions being tested, and suggested that frequency response
was primarily dependent on soil and pile properties.
Gohl (1991) also used dimensional analysis to derive the following functional
relationship for scale model similitude as applied to shaking table tests of model piles:

,
_

u
m
g
u
u G
EI
u
l
K
b
y
o s
o o
o
s s
p
o
3
2
4
, , , ,

(5.12)
where y = dynamic lateral deflection of the pile, b = pile diameter, l = pile length, u
o
=
amplitude of input base motion,
p
= pile mass density,
s
= soil mass density, EI = pile
flexural rigidity, G
s
= depth and strain level dependant shear stiffness of the soil, =
frequency of input base motion, g = gravitational constant, and m
o
= superstructure mass.
Gohl points out that it is very difficult to simultaneously satisfy the second scaling law,
which implies the same model and prototype material densities, and the third scaling law,
which derives the ratio of prototype to model pile flexural rigidity as equal to
5
. He
comments that a way of accepting imperfect model similitude is by viewing the tests as
prototype events themselves, against which analytical models can be verified. He also
suggests that the test results may be expressed in terms of dimensionless variables to allow
comparison with full scale results.
Iai (1989) built on Rocha's work by considering a tangent modulus approach to
scale model constitutive behavior for saturated soils (see Figure 5.3). He derived a
comprehensive set of scaling relations for a soil-structure-fluid system under dynamic
loading and defined the entire problem in terms of geometric, density, and strain scaling
221
factors. His method proscribes the geometric () and density (
p
) scaling factors, and then
derives the strain scaling factor (

) from shear wave velocity tests in both the model and


prototype soil:
( ) ( ) [ ]
V
s
m
V
s
p
2

(5.13)
The non-intuitive result is that certain model quantities with the same dimensions may
have different scaling factors, such as length and deformation. Again, the validity of his
technique was demonstrated by laboratory testing. Iai qualifies his method by stating that
it only applies for small strains where the soil particles do not lose contact, but does make
some application to liquefaction problems involving medium to dense sands. Iais
complete set of scaling factors is listed in Appendix A.
Scott (1989) applied the method of governing equations to dynamic equilibrium
for constructing model soil scaling relations for centrifuge testing. Gibson (1996) refined
this derivation and generalized it for a granular saturated soil subject to 1-g or centrifuge
testing:
X X
x
t
u
x t
u
t
u
t
x x
im
m
im
jm m
jm
m
im
m

,
_


1
]
1

,
_

,
_

*
*
*
2
2
*
*
*
*
1
*
1
2

(5.14)
where x represents length, is stress, t is time, is density, X is the body force, and
starred (*) quantities refer to the prototype to model ratio. Gibson also considered the
problem of scaling soil constitutive behavior for 1-g testing, and proposed modifying the
model material so that under 1-g stress conditions it would still exhibit strain behavior
similar to the prototype. This approach utilizes the steady state line and is depicted in
Figure 5.4; it differs from Rochas and Iais methods which both elected to modify the
222
constitutive relation rather than the soil properties. Gibson also showed that the dynamic
time scale and the diffusion time scale (related to pore pressure response and liquefaction)
were incompatible for 1-g testing, unless provisions were made to use a finer grained soil
or a more viscous pore fluid.
Figure 5.4 - Definition of Model Soil Properties Based on Steady-State Line
(after Gibson, 1996)
An important fact should become apparent in the work of the soil mechanics
researchers described in this section. They have added the important feature of
constitutive similarity to the set of scale modeling requirements for soil response
problems. This reflects the fact that it is not adequate to simply model a discrete elastic
parameter of the system, but rather to consider its full range of nonlinear behavior.
Constitutive similarity will be discussed with reference to the model soil and model pile
designs in sections 5.2 and 5.3, respectively.
223
5.1.3 Scale Modeling Methodology and Implied Prototypes
Consider the SSPSI problem at hand, previously depicted in Figure 1.1, with its
interdependent processes. This figure constitutes a crucial component of the scale
modeling method in that it delineates the variables and modes of the system, and the scale
model program must be designed to adequately capture the behavior(s) of principal
interest. It is clear that no governing equation can be written that describes this entire
system, nor can dimensional analysis or similitude theory be directly applied to this
complex system to achieve true model similarity. The viable scale modeling approach
for this application therefore consists of identifying and successfully modeling the primary
forces and processes in the system, while suppressing secondary effects, thereby yielding
an adequate model. This scale modeling design procedure is implemented as an iterative
process that can be described by the flowchart shown in Figure 5.5.
Figure 5.5 describes a scale modeling approach in which the primary modes of
system response are first identified and prototype values for the variables contributing to
these modes are established. Scaling relations are derived and used to compute scale
model parameters for the variables of interest. Scale model components are fabricated and
tested to verify their actual behavior. Scaling relations are then used to determine whether
the measured model behavior implies a reasonable prototype response. This method of
implied prototypes provides a suitable modeling approach for the wide range of potential
prototype soil, pile, and superstructure conditions. Caution must be exercised when
interpreting scale model test results in terms of the prototype, however. The most
accurate use of numerical analysis applied to the modeling process is analysis of the scale
model, not to predict the behavior of the (implied) prototype. Full extension of model
224
Identify primary modes of system response and
variables contributing to these modes
Use dimensional analysis to develop scaling
relations for primary modes of system
response
Assign target range of prototype
parameters to primary system
variables
Use scaling relations to compute trial scale
model parameters for primary system variables
Select and fabricate model
materials to meet trial model
parameters
Establish actual model parameters
and model response by component
testing
Use scaling relations to compute prototype
parameters implied by actual model
parameters
Are implied prototype parameters
reasonable and within target range ?
No
Figure 5.5 - Flowchart Describing Scale Modeling Methodology of Implied Prototypes
225
scale results to quantitative prototype behavior implies a high confidence in all aspects of
the modeling process (see Section 6.4.5). One technique employed to evaluate the
accuracy of the scale modeling technique is known as modeling of models. In this
approach, independent tests of the same prototype are conducted at different scaling
factors; if the results uniformly meet similitude requirements, then the modeling technique
can be judged sound.
5.1.4 Scale Modeling Factors for Shaking Table Testing
Examining Figure 1.1, the relevant modes of SSPSI system response can be
identified. They are the free-field soil site response, soil-pile lateral kinematic interaction,
soil-pile lateral inertial interaction, soil-pile axial response, and pile and pile cap radiation
damping. Table 5-1 lists the variables associated with each interaction mode.
Table 5-1 Identification of SSPSI Primary System Modes and Associated Variables
SSPSI Interaction Mode Variables
1. Free-field Site Response (V
s
(z), (z), Modulus Degradation & Damping(z))
soil
2. Soil-Pile Lateral Kinematic
Interaction
1 + (EI, l, d, fixity)
pile
, (-, S
u
(z))
soil
3. Soil-Pile Lateral Inertial
Interaction
2 + (M, K)
superstructure
4. Soil-Pile Axial Response 1 + (E, l, d)
pile
, (-, S
u
(z))
soil
, (M, K)
superstructure
5. Radiation Damping 1 + (l, d, M, E)
pile
The objective of the scale modeling procedure for this test program is to achieve
what has been previously defined as dynamic similarity, where model and prototype
experience homologous forces. Dimensional analysis is the framework for scale model
226
similitude in this test program. Three principal test conditions establish many of the
scaling parameters. The first is that testing is conducted in a 1-g environment, which
defines model and prototype accelerations to be equal. Secondly, a model soil with similar
density to the prototype soil is desired, which fixes another component of the scaling
relations. Thirdly, the test medium is primarily composed of saturated clay, whose
undrained stress-strain response is independent of confining pressure, thereby simplifying
the constitutive scaling requirements.
As previously shown in section 5.1.1, by defining scaling conditions for density and
acceleration, the mass, length, and time scale factors can all be expressed in terms of the
geometric scaling factor , and a complete set of dimensionally correct scaling relations
(ratio of prototype: model) can be derived for all variables being studied. The scaling
relations for the variables contributing to the primary modes of system response are shown
in Table 5-2.
Table 5-2 Scaling Relations for Primary System Variables Expressed in Terms of the
Geometric Scaling Factor
Mass Density 1 Acceleration 1 Length
Force
3
Shear Wave Velocity
1/2
Stress
Stiffness
2 2
Time
1/2
Strain 1
Modulus Frequency
-1/2
EI
5
227
With the shear wave velocity scaling factor =
1/2
, the Cauchy condition is met,
and Iais strain scaling factor can be calculated equal to one, with the result that his set of
scaling relations falls in absolute agreement with the values derived for this study.
The application of the scaling relations and development of model soil, pile, and
superstructure system components will be discussed in the following sections. But before
doing so, it is important to recognize the following problem conditions that can propagate
into the scale modeling process. They must be accounted for in the design of the model
components and/or testing procedures:
Initial Conditions - referring to both the soil and pile initial stress states,
Boundary conditions - incorporating not only model boundary conditions but interface
conditions between soil and pile, and soil and cap,
Constitutive behavior - remains a soil scale modeling criterion ,
Ductility - for pile and superstructure,
Material damping - applicable to soil, pile, cap, and superstructure,
Strain rate effects - applies to both soil and pile,
Long term effects - such as consolidation or creep for the soil, and
Group effects - reflecting the configuration of piles in groups.
5.2 Model Soil Design
Model soil properties are reflected in all five primary modes of SSPSI described in
Table 5-1, but can be segmented into the general categories of free-field response and soil-
pile interaction. Free-field site response is primarily a function of the small strain soil
properties, while soil-pile interaction is ultimately a large strain phenomenon. The
228
parameters describing these soil properties are shear wave velocity, density, modulus
degradation and damping, stress-strain response, and undrained shear strength. These
parameters are both discrete and nonlinear, and are a function of the loading rate, the
number of cycles and strain reversals. The method of implied prototypes is therefore
especially well suited to this complex scale modeling problem.
5.2.1 Identification of Soil Modeling Criteria
With the model soil density equal to the prototype soil density, one scaling
condition is determined. The nonlinear stress-strain and modulus degradation and
damping curves are not directly modeled from a prototype case, but rather the method of
implied prototypes is used to consider whether the scale model properties for these
parameters are reasonable. This leaves undrained shear strength and shear modulus (or
shear wave velocity) as the principal soil modeling criteria. If the elastic response of both
the free-field soil and the soil-pile system is desired, the soil shear modulus should be
properly modeled. If the inelastic response of the soil-pile system is desired, then the
undrained shear strength should be emphasized. If the full nonlinear system response is
desired, then both criteria must be satisfied simultaneously. Unfortunately these
parameters have different scaling factors, for undrained shear strength and
1/2
for shear
wave velocity, which greatly complicates the scale modeling effort. An additional soil
modeling criteria not reflected in Table 5-1 is plasticity index. Both static and dynamic
soil behavior are known to be strongly influenced by plasticity index (PI), and it is
therefore desirable to use a model soil with a similar PI to the prototype (PI is
dimensionless and therefore scales 1:1 from model to prototype).
229
5.2.2 Definition of Prototype Soil Parameters
The target prototype soil selected for this study is San Francisco Bay Mud, a
marine clay whose index properties have ranges of values, and is therefore well-suited to
the method of implied prototypes. It is also a well-characterized soil that was the subject
of a study by Bonaparte and Mitchell (1979), who conducted tests on samples of Bay Mud
retrieved from Hamilton Air Force Base in Novato, California. Their findings are shown
in Table 5-3, which reflects prototype parameters adopted for this research.
Table 5-3 Selected Properties of San Francisco Bay Mud
Property Value
Saturated Unit Weight (pcf) 94
Natural Water Content (%) 90
Liquid Limit (%) 88
Plastic Limit (%) 48
Plasticity Index (%) 40
Undrained Strength Ratio S
U
/P 0.32
Coefficient of Consolidation C
v
(ft
2
/yr) 8-10
Dickenson (1994) investigated the seismic response of Bay Mud during the 1989
Loma Prieta earthquake, and proposed the following empirical relationship between
undrained shear strength and shear wave velocity:
V
s
= 18 (S
u
)
0.475
(5.15)
where V
s
is in feet per second and S
u
is in pounds per square foot. This relationship,
shown in Figure 5.6, was used to establish the target shear wave velocities for the
prototype soil. For prototype soil undrained shear strengths from 600 psf to 1200 psf,
230
appropriate shear wave velocities from 375 ft/sec to 525 ft/sec are computed.
Figure 5.6 - Variation of Shear Wave Velocity with the Undrained Shear Strength (Static)
of Shallower Cohesive Soils (after Dickenson, 1994)
5.2.3 Model Soil History
The option of using a reconstituted soil as the model soil was considered, a
technique commonly used in centrifuge testing. In this approach, a natural soil is mined
and then mixed with water to form a slurry that can easily be placed in the test container.
The soil is then consolidated to achieve the desired strength/stiffness profile, a process
greatly facilitated during spin-up in the centrifuge. The consolidation process offers the
advantage of fixing anisotropy and a stress history into the soil. But this method was
deemed impractical due to the large size of the test container and the very long time that
would be required for consolidation in a 1-g environment. More importantly, a
reconstituted soil would not be able to satisfy the competing scale modeling criteria of
undrained shear strength and dynamic shear modulus.
Therefore a synthetic model soil was chosen as the soil medium for the testing
231
program. A synthetic soil was recognized to sacrifice actual in-situ soil properties such as
heterogeneity, anisotropy, fabric, and stress history, but without serious detriment to the
performance of a well-designed model soil. Tavenas et al. (1973) describe the
development of an artificial model soil using kaolinite, Portland Cement, and bentonite to
replicate a brittle Lake Champlain clay. Blaney and Mallow (1987) tested numerous
stiffening agents used in conjunction with bentonite to fabricate a synthetic
overconsolidated clay for dynamic soil-pile interaction tests. Their final model soil design
consisted of fumed silica and bentonite, and was extruded in blocks that were bonded
together in the test container. Tables 4-3 to 4-6 described model pile testing programs
using a variety of model soil materials including kaolinite, bentonite, supersil, plastellina,
aerosil, veegum, silicon gum, plasticine, polyacrylimide, sands, and reconstituted clays and
clayey silts.
But the most extensive research with model clay soil has taken place at U.C.
Berkeley. This work originated with Seed and Clough (1963), who developed a mix of
kaolinite and bentonite for shaking table modeling of the seismic response of earth
embankments. The mix was proportioned as 3 kaolinite:1 bentonite, at approximately 200
% water content, as that fraction of bentonite was found to arrest the consolidation
process over the testing time frame; this mix was also noted to exhibit pronounced
thixotropy. In addition, the scaled stress-strain curve of the model soil favorably
compared to that of typical dam core materials. Seed and Clough defined inelastic
deformations and therefore undrained shear strength as their primary modeling criterion,
which was found to be a function of water content. But they also recognized that it was
the soil shear strength under dynamic loading which was of interest, and that model and
232
prototype soils experience different degrees of dynamic strength gain from the reference
static strength. They proposed that the prototype soil static shear strength be multiplied
by an additional scaling factor of 0.65 to account for the unequal dynamic strength
increases between model and prototype.
Sultan and Seed (1967) continued using this model soil for shaking table tests of
sloping core earth dams, also at water contents in the range of 200 %. Kovacs (1968)
investigated the dynamic response of clay embankments and conducted cyclic simple shear
tests to evaluate dynamic moduli and damping ratios for the model clay soil. His model
soil was batched at water contents ranging from 88 - 125 %. Arango-Greiffenstein (1971)
studied the seismic stability of slopes in saturated clay by shake table testing of
embankments constructed from the model soil, again at water contents approaching 200
%. Bray (1990) used the same model soil mixed at 130 - 136 % water content to examine
fault rupture propagation through scale model clay deposits. This stiffer mix was required
to properly reproduce a fault rupture failure mode, as higher water contents resulted in too
viscous behavior. Lazarte (1996) also studied fault rupture propagation with tests on
model soil mixes at water contents ranging from 97 - 120 %, and characterized the time-
dependent strength and stress-strain properties of the model soil. He recognized that the
relatively high tensile capacity of the model soil may have suppressed the formation of
cracks and unrealistically redistributed the strain field in zones under tension. Figure 5.7 is
a summary plot of the model soil undrained shear strength as a function of water content
as determined by these researchers.
233
0
50
100
150
200
250
300
1 10 100
Undrained Shear Strength (psf)
W
a
t
e
r

C
o
n
t
e
n
t

(
%
)
Seed and Clough (1963) - Direct Shear
Sultan and Seed (1967) - Direct Shear
Kovacs (1968) -Vane Shear
Arango (1971) - Model Tests
Bray (1990) -UUTX
Lazarte (1996) -UUTX
Figure 5.7 - Model Soil Undrained Shear Strength Versus Water Content As Determined
by Various Researchers (after Lazarte, 1996)
5.2.4 Development of Model Soil
The development of a model clay soil for this research project commenced in 1995
with initial mix designs following the traditional 3 kaolinite: 1 bentonite Berkeley recipe.
The kaolinite used was a Huber 35 hydrated aluminum silicate, and the bentonite an
American Colloid Volclay KWK montmorillonite (these products are similar in
composition but not identical to other bentonites and kaolinites used by previous
researchers). Small trial batches were prepared in 5 gallon plastic buckets by initially
blending together the dry materials and then mixing in the water by hand. The buckets
were sealed with an airtight lid and stored in a constant-temperature high humidity wet
room. Test specimens were prepared by retrieving material from the storage buckets and
remolding the clay into 1.4 and 2 in diameter cylindrical molds for unconsolidated-
234
undrained triaxial (UUTX) and bender element shear wave velocity tests, respectively.
The time from mixing to testing is referred to as mix age and the time from specimen
preparation to testing is termed cure age.
UUTX compression tests and bender element shear wave velocity tests were
performed on specimens with a range of water contents from 50 - 130 % to assess the
model soil performance. Details of using bender elements for the determination of soil
shear wave velocity are provided by Viggiani and Atkinson (1995) and Riemer et al.
(1998). Initial tests indicated that for a given undrained shear strength, the corresponding
shear wave velocity was too low to meet the criteria of the implied prototype. It was
therefore desired to find an admixture that would increase the small strain dynamic
stiffness without appreciably affecting the undrained shear strength. The effect of several
admixtures, including fine sand, silt, and fly ash, in varying proportions and over a range
of water contents, was tested; only fly ash was found to have the desired effect. Fly ash is
a silt-sized calcium-rich byproduct from the combustion of coal at electric power plants,
and was considered as an economical admixture for the model soil. Results from a series
of UUTX compression tests on specimens with 20 % fly ash (all fly ash contents are stated
as a percentage of dry weight) are shown in Figure 5.8, illustrating the dependence of
undrained shear strength on water content. Figure 5.9 plots the undrained shear strength
versus the water content of the clay fraction of the model soil for test specimens with fly
ash contents ranging from 0 - 60 %; the clear trend in the data indicated that the fly ash
had little influence on the shear strength of the model soil.
235
0
50
100
150
200
250
300
350
400
450
500
0.00 0.05 0.10 0.15 0.20 0.25
Axial Strain
D
e
v
i
a
t
o
r
i
c

S
t
r
e
s
s

(
p
s
f
)
Wc = 61%, Su = 207 psf
Wc = 71%, Su = 132 psf
Wc = 79%, Su = 92 psf
Wc = 94%, Su = 57 psf
Figure 5.8 - Unconsolidated-Undrained Triaxial Compression Test Results For Model
Soil Mixture with 20% Fly Ash at Four Water Contents
0
50
100
150
200
250
60 70 80 90 100 110 120 130 140
Water Content Clay Fraction (%)
S
u

(
p
s
f
)
0% Fly Ash
20% Fly Ash
28.5% Fly Ash
40% Fly Ash
60% Fly Ash
Figure 5.9 - Model Soil Undrained Shear Strength Versus
Water Content of Clay Fraction
236
Gruber (1996) carried out a series of 66 UUTX tests on model soil specimens and
samples of Bay Mud retrieved from Hamilton Air Force Base in Novato, California to
ascertain whether the nonlinear stress-strain response of the model soil implied a
reasonable prototype behavior. The model soil specimens were prepared at water
contents ranging from 60 - 110 % and all contained 10 % class C fly ash. Testing was
performed under confined (1 kilogram per square centimeter) and unconfined conditions,
and at normal (0.045 in/min) and fast (4.5 in/min) strain rates. Figures 5.10 and 5.11
illustrate representative UUTX test results for the model soil and Bay Mud. These model
soil specimens had a cure age of 4 and 5 days and water contents ranging from 98 - 100
%, while the Bay Mud was sampled from a depth of 4.8 m and had water contents ranging
from 89 - 94 %. It is evident that both materials exhibited higher peak strengths under
fast loading, and decreased sensitivity under confining pressure. Bay Mud displayed
higher failure strains under fast loading, while the model soil failure strains remained
relatively constant for both loading rates. In fact, the model soil behaved as a strain
hardening material when tested under confining pressure at both normal and fast
loading rates. Under normal unconfined loading, the model soil had a similar failure
strain to Bay Mud, but was less sensitive. This raises the question as to whether peak
strength or residual strength is most appropriate to model. The observation that lateral
and axial pile capacities are formulated in terms of peak strength supports that modeling
approach. In summary, the model soil did not precisely replicate the prototype stress-
strain behavior, but it did exhibit a reasonable response and was therefore found to
constitute an adequate scale model of a higher plasticity soft to medium stiff clay such as
San Francisco Bay Mud.
237
0
50
100
150
200
250
300
350
400
0 0.05 0.1 0.15 0.2 0.25
Axial Strain (%)
D
e
v
i
a
t
o
r
i
c

S
t
r
e
s
s

(
p
s
f
)
unconfined, strain rate = 0.045 in/min
unconfined, strain rate = 4.5 in/min
confined (1 ksc), strain rate = 0.045 in/min
confined (1 ksc), strain rate = 4.5 in/min
Figure 5.10 - Model Soil Unconsolidated-Undrained Triaxial Compression Test Results
Showing Effects of Strain Rate and Confining Pressure (after Gruber, 1996)
0
100
200
300
400
500
600
700
800
900
0 0.05 0.1 0.15 0.2 0.25
Axial Strain (%)
D
e
v
i
a
t
o
r
i
c

S
t
r
e
s
s

(
p
s
f
)
unconfined, strain rate = 0.045 in/min
unconfined, strain rate = 4.5 in/min
confined (1 ksc), strain rate = 0.045 in/min
confined (1 ksc), strain rate = 4.5 in/min
Figure 5.11 - Bay Mud Unconsolidated-Undrained Triaxial Compression Test Results
Showing Effects of Strain Rate and Confining Pressure (after Gruber, 1996)
0.0
0.5
1.0
1.5
2.0
2.5
3.0
60 70 80 90 100 110 120 130 140
Water Content Clay Fraction (%)
S
u

d
y
n
a
m
i
c

/
S
u

s
t
a
t
i
c

(
p
e
a
k
)
Figure 5.12 - Ratio of Undrained Shear Strength at Dynamic (4.5 in./min.) and Static
(0.045 in./min.) Strain Rates for Model Soil in Unconsolidated-Undrained Triaxial
Compression Tests (after Gruber, 1996)
238
The purpose of studying higher strain rates was to explore Seed and Cloughs
(1963) observation that model soil and Bay Mud have different shear strength increases
under dynamic loading, and to quantify what correction must be made to equilibrate the
dynamic shear strengths. The Bay Mud dynamic strength increase was on the order of 70
% for both confined and unconfined conditions, while the model soil dynamic strength
increase was 25 % and 10 % for confined and unconfined conditions, respectively. Figure
5.12 summarizes the model soil dynamic strength increase for all specimens tested; this
data suggests that an average dynamic strength increase on the order of 25 % may be
adopted. Following Seed and Cloughs methodology, the dynamic strength correction
factor to be applied to the scaled prototype undrained shear strength is therefore 0.75.
Wartman (1996) conducted a detailed study of the effects of fly ash on the
geotechnical properties of the model clay soil. Both class F and class C fly ash materials
were tested; the class F fly ash was generated at the Jim Bridger power plant in Point of
Rocks, Wyoming, and the class C material was obtained from the Laramie River power
plant in Wheatland, Wyoming. The chemical composition of the two fly ash materials is
summarized in Table 5-4. He concluded that class F fly ash acted as an inert filler material
with only a marginal effect on the model soils undrained shear strength and shear wave
velocity, but that class C fly ash had appreciable effects:
The class C fly ash used in this study acted as a chemically reactive
material when mixed with the kaolinite-bentonite clay. The chemical
reactivity is attributed to the high calcium oxide content of the fly ash.
When mixed with the clay, the class C fly ash caused rapid cation exchange
to occur on the clay minerals leading to a substantial reduction in plasticity.
The cation exchange caused the double layer around the clay mineral to
shrink resulting in an increase in stiffness and by association, shear wave
velocity. The class C fly ash also caused cementacious pozzolanic reaction
products to form in the specimens. Increases in undrained strength were
239
not always realized from these cementacious bonds, however, because
many of the fly ash specimens were remolded as part of the preparation
procedures for the unconfined compression tests.... Cure age had little to
no effect on the undrained strength of the specimens over the period of
interest in this study (days to weeks). Mix age had no effect on the
undrained strength of the fly ash-clay mixtures.
Table 5-4 Chemical Composition of Class F and Class C Fly Ashes
Chemical Composition Class F Fly Ash Class C Fly Ash
Calcium Oxide (%) 5.97 30.13
Silicon Oxide (%) 61.71 31.53
Aluminum Oxide (%) 19.67 16.87
Iron Oxide (%) 4.47 5.82
Wartman performed bender element tests to determine the shear wave velocity of
the model soil mixtures, following recommendations for discerning pulse travel times by
Riemer, et al. (1998). Soil shear wave velocity versus cure time for an assortment of
model soil mixtures is plotted as Figure 5.13. The strong influences of fly ash content and
cure time on shear wave velocity are apparent, with the most dramatic increases in
stiffness occurring in the first few days after mixing.
A constant rate of strain (0.00012 in/min) consolidation test was performed on a
specimen of model soil with 10 % class C fly ash at an initial water content of 100 %.
The e-log p curve is shown in Figure 5.14, and the coefficient of consolidation C
v
was
calculated as 10 in
2
/year. This slow rate of consolidation implied relatively stable soil
properties throughout the shaking table testing time window.
240
0
20
40
60
80
100
120
140
160
180
200
0 5 10 15 20 25
Cure Age (days)
S
h
e
a
r

W
a
v
e

V
e
l
o
c
i
t
y

(
f
t
.
/
s
e
c
.
)
20% class F - 100% m.c.
20% class C - 100% m.c.
10 % class C - 110% mc
10% class C(#2) - 100% m.c.
10% class F - 100% m.c.
10% class C(#2) - 100% m.c.
20% class F - 80% m.c.
10% class C(dyn) - 100% m.c.
5% class F - 100% m.c.
0% fly ash - 100% m.c.
Figure 5.13 - Shear Wave Velocity Versus Cure Age for Model Soil Specimens with
Varying Fly Ash Contents (after Wartman, 1996)
2.00
2.20
2.40
2.60
2.80
3.00
3.20
V
o
i
d

R
a
t
i
o
100 1000 10000 100000
Stress (psf)
Figure 5.14 - Void Ratio Versus log Pressure for Constant Rate of Strain Consolidation
Test of Model Soil Specimen
241
Cyclic triaxial testing of the model soil was performed to determine modulus
reduction and damping curves, but testing inaccuracies initially yielded incomplete data to
construct these curves. An advanced triaxial testing device was later used to establish
modulus reduction and damping curves from samples of the insitu model soil, which were
then compared with reference curves for Bay Mud and medium plasticity clays. This work
is described in Chapter 7, and the model soil specific curves were used for site response
and SSPSI analyses.
Finally a model soil composed of 67.5 % kaolinite, 22.5 % bentonite, and 10 %
class C fly ash, at 100 % water content was selected as the design mix. This water content
was expected to provide the key elements of mixability and pumpability for shake table
model testing. The model soil had a unit weight of 94 pcf, a liquid limit of 115, a plastic
limit of 40, and a plasticity index of 75. The undrained shear strength of the mix at a cure
age of 5 days was determined to be 85 psf, and the shear wave velocity at the same cure
age was 130 ft/sec. This benchmark cure age was used as it was expected that the time
from soil placement to time of testing and the time between tests would be approximately
five days. The prototype values implied by these model properties with a geometric
scaling factor of 8 are a static undrained shear strength of 520 psf (with 0.75 dynamic
strength correction factor) and a shear wave velocity of 365 fps. These values are in
agreement with Dickensons relation and the model soil was therefore found to constitute
an adequate scale model of a higher plasticity soft to medium stiff clay such as San
Francisco Bay Mud.
242
5.3 Model Pile Design
As for the model soil, the model pile was subject to competing scale modeling criteria.
By addressing the principal governing factors of pile response, a successful model pile
design was attained. The four primary modes of pile response recognized in Table 5-1 are
soil-pile lateral kinematic interaction, soil-pile lateral inertial interaction, soil-pile axial
response, and pile radiation damping.
5.3.1 Identification of Pile Modeling Criteria
Numerous pile properties contributing to the principal modes of pile response were
identified, including slenderness ratio L/d, flexural rigidity EI, yield behavior/mechanism,
ductility, moment-curvature relationship, buckling properties P
cr
and d/t, natural period of
vibration, and relative soil/pile stiffness. Geometric similarity was adopted as a strict
modeling constraint, so that overall pile slenderness and relative contact surface area
would be preserved in the model. This also ensured that pile group relative spacing and
consequent group interaction would be replicated at the model scale.
The pile moment-curvature relation was selected as a primary modeling criterion
as it encompasses both flexural rigidity and yield behavior to describe the full range of
nonlinear pile response to lateral loading. In this regard it is important to consider the
current state-of-practice seismic design for bridge pile foundations. The prevailing
philosophy is to generally design piles to respond in their elastic range without yielding,
concentrating ductile behavior in the bridge columns. The rationale is that damage to
above ground structures is much easier to detect and repair than damage to subsurface
elements. The implication for this modeling program is that by scaling pile EI and
243
ensuring a yield point equal to or greater than that of the scaled prototype, the working
range of pile lateral dynamic response is correctly modeled. With soil resistance properly
scaled, both soil-pile lateral kinematic and inertial interaction should then be accurately
reproduced in the model. Both nonlinear and cyclically degrading lateral pile response are
therefore captured by the soil behavior.
Soil-pile axial response for end-bearing piles is primarily a function of soil
properties in the bearing strata. Soil-pile interface friction/cohesion along the shaft and
pile elastic deformation constitute secondary factors for end-bearing pile axial response,
especially in soft soils. While the axial loading is dynamic in nature, the static axial
capacity of the pile is a crucial factor as it determines the inertial load the pile carries.
Pile radiation damping can be considered to have two components. First, the
inherent vibration characteristics of the pile determine its ability to generate energy to
radiate into the soil. Second, the propagation of energy away from the pile depends on the
relative soil-pile stiffness. With soil and pile elastic properties consistently scaled, the
relative soil-pile stiffness automatically scales from the prototype to the model. But the
inherent pile vibration characteristics are a more difficult modeling criterion to meet. The
frequency of vibration of an end-bearing pile can be idealized by the equation describing
the frequency of vibration of a cantilever rod (Clough and Penzien, 1996), which is seen to
be a function of the rods mass:

L m
EI
4
516 . 3 (5.16)
With pile geometry and EI scaled as previously described, the mass per unit length of the
model pile must be scaled by a factor of 1/
2
from the prototype. With conventional
244
materials and other modeling constraints, this criterion could not be satisfied. But
recognizing that radiation damping is most pronounced at smaller levels of shaking, it
could be expected to have a diminished influence at the strong levels of shaking planned in
this test program. Secondly, piles are only a component of the soil-pile-superstructure
system, and altering to some degree the vibration characteristics of a lesser component of
that system would not be expected to as significantly affect the vibration characteristics of
the entire system.
5.3.2 Definition of Prototype Pile Parameters
From standard Caltrans design, a 16 in diameter x 0.5 in wall concrete-filled steel
pipe pile was selected as the target prototype. Scaling constraints dictated a maximum
prototype pile length of 44 ft, which provided a L/d ratio of 33, acceptable for a slender
pile. The fixity conditions of the pile, known to be significant in lateral response, were
established as fixed against rotation at the head, and fixed against (relative) translation at
the tip. This corresponds to a pile driven into a firm strata at the base, and cast into the
pile cap with a connecting reinforcing cage. The flexural rigidity EI of a composite
concrete/steel pile is nonlinear due to concrete cracking; therefore the concrete
contribution to the composite EI was degraded by 50 % to yield a composite EI of
181,920 k-ft
2
. The first mode period of vibration of a cantilever rod with the prototype
pile properties was computed as 0.74 seconds.
245
5.3.3 Development of Model Pile
An iterative spreadsheet solution was employed to satisfy the principal pile design
criteria of flexural rigidity EI and first mode vibration period of an equivalent cantilever
rod (see Appendix B). A geometric scaling factor of 8 was found to optimize both soil
and pile scaling requirements. With a target model EI of 5.55 k-ft
2
and the model pile
outer diameter fixed at 2 in, the moment of inertia was computed for cases of solid and
thin wall tubes, and the corresponding lower and upper bound elastic moduli were
calculated to be 1000 and 10000 ksi, respectively. At these two bounding values, the pile
material density was calculated that would impart the scaled vibration modes.
Unfortunately, these density values ranged from 1880 to 3650 pcf for the solid and thin
wall sections, respectively. Obviously such materials are not feasible, and this requirement
was relaxed as previously explained in section 5.3.1.
Many materials were investigated to ascertain their suitability as a model pile
material; Table 5-5 summarizes the moduli and yield stresses of these prospective
materials. From this table it can be seen that aluminum 6061 T-6 alloy is the only
candidate that falls in the range of acceptable modulus, and must be configured as a thin
wall section to meet the EI criterion. A wall thickness of 0.0265 in was calculated as
producing the correctly scaled model pile flexural rigidity. The minimum wall thickness of
commercially available aluminum tubing is 0.028 in, which results in an EI of 5.86 k-ft
2
, a
5 % deviation from the target value. Although a thin wall tube provides a potential local
buckling mechanism not present in the solid prototype cross-section, this geometry proved
favorable for the external mounting of foil strain gages and internal routing of the gage
lead wires. Importantly, aluminum is a relatively economical material, is readily available
246
in thin wall tube sections, and is suitable for mounting foil strain gages. With respect to
axial performance, calculations of driving stresses, static loads, and stresses under dynamic
loading were found not to exceed the aluminum tube buckling load P
cr
.
To ensure elastic response of the thin-walled aluminum tube, the theoretical
moment-curvature relation of the trial model pile was compared to that of the prototype.
The moment-curvature relation describes the pile nonlinear response to applied loading
and is analogous to a soil stress-strain curve. Lower and upper bound prototype cases
were established using yield stresses in the steel pipe pile of 50 and 70 ksi, with
contributions of 0 and 100 % of concrete EI representing intact and fully cracked sections.
The pile analysis code COM624P was used to define the lower and upper
Table 5-5 Mechanical Properties of Candidate Model Pile Materials
Material Elastic Modulus (ksi) Yield Stress (ksi)
Steel 29000 60
copper 17000 30
aluminum 6061 T-6 10000 40
nylon 420 17
PVC 420 14.5
polyamide 410 12
polyacetal 410 10 14
acrylic PMMA 390 - 480 12 17
polycarbonate 300 - 350 13.5
ABS 250 4 14
polypropylene 170 - 250 7
PVDF 213 9.7
ryton 160 2.9
CPE 150 6
vinylester fiberglass 140 15.5
epoxy fiberglass 135 10
high density polyethylene 60 - 180 7
flurocarbons 70 3
teflon TFE 58 2 5
teflon FEP 50 2.7 - 3.1
polybutylene 35 4.2
low density polyethylene 14 - 38 2.3
bound moment-curvature relations shown in prototype scale in Figure 5.15. This method
247
was calibrated against the results of a four-point loading test conducted by Caltrans on a
24 in diameter concrete-filled steel pipe pile (Brittsan, 1995). To compute the moment-
curvature of the model pile, the method of Langhaar (1951) was followed, where the
equation of statics describing bending of a ductile beam of circular cross-section is
expressed as

1
0
3
2 d c M (5.17)
in which c is the radius, is the ordinate from the neutral axis divided by the radius, is
the width of the cross-section at the ordinate divided by the radius, and is the stress at
the ordinate of the cross-section. This derivation assumes an elastic-perfectly plastic
stress-strain relation for the beam. The moment-curvature relation for the trial aluminum
model pile is superimposed on Figure 5-15 at prototype scale, and can be seen to exceed
the yield behavior in the target prototype range. As previously explained, this is an
acceptable result as the pile is intended to respond in its elastic range. An ideal model
pile material that would precisely fit the scaled moment-curvature relation was calculated
to have a wall thickness of 0.2 in, an elastic modulus of 3000 ksi, and a yield stress of 2
ksi. Extensive research indicated that such a material could not be obtained nor easily
manufactured, and the thin-walled aluminum tube was proof tested to confirm its
performance.
248
0
2000
4000
6000
8000
10000
12000
14000
16000
18000
20000
0 0.0005 0.001 0.0015 0.002
Curvature (/in)
M
o
m
e
n
t

(
k
-
i
n
)
Aluminum Tube 2" dia. X 0.028" wall scaled up
Steel Yield = 80 ksi, Concrete = 7 ksi
Steel Yield = 50 ksi, Concrete = 0
Figure 5.15 - Theoretical Lower and Upper Bound Moment-Curvature Relations for
Prototype Pile as Determined by COM624P
P/2
strain gages
Bending Moment Diagram
Loading Diagram
0.875 P
P/2
Figure 5.16 - Diagram of Four-Point Loading Test of Model Pile
0
500
1000
1500
2000
2500
3000
3500
4000
4500
5000
0 0.005 0.01 0.015 0.02
Curvature (/in)
M
o
m
e
n
t

(
l
b
-
i
n
)
Experimental Data
Theoretical
FAILURE
Figure 5.17 - Theoretical and Experimental Moment-Curvature Relations for 2 Diameter
x 0.028 Wall Aluminum Tube Model Pile
249
5.3.4 Four Point Loading Test
A 6 ft long section of the 2 in diameter aluminum tube with a wall thickness of
0.028 in was load tested to verify its moment-curvature relation. A four point loading test
was performed, with the tube simply supported near its ends, and equal loads were applied
at two points as diagrammed in Figure 5.16. This loading pattern results in a zone of
constant moment being applied across the central section of the member. Foil strain gages
mounted to the compression and tension faces of the tube were read at each loading
increment, and moment and curvature calculated from the strain data. The experimental
moment-curvature relation for the model pile is shown in Figure 5.17, superimposed on
the theoretical plot, both at model scale. The agreement in the range of elastic response is
excellent, and the test pile failed instantaneously with a buckling mechanism very near the
theoretical yield point. The failure load was 163 lbs, imposing a bending moment of 3420
lb-in. These results proved the aluminum tube to be an adequate model pile for the scale
model testing program.
With scale model similitude relations defined and a model soil and model pile
designed, the following chapter will describe the development of the shaking table test
program for the study of SSPSI.

Anda mungkin juga menyukai