Anda di halaman 1dari 14

Lee, K. K. et al. (2013). Géotechnique 63, No. 15, 1271–1284 [http://dx.doi.org/10.1680/geot.12.P.

175]

Bearing capacity on sand overlying clay soils: experimental and finite-


element investigation of potential punch-through failure
K . K . L E E  , M . J. C A S S I DY † a n d M . F. R A N D O L P H †

When a jack-up spudcan foundation is installed on seabeds consisting of a sand layer overlying soft
clay, the potential for ‘punch-through’ failure exists. This happens as a result of an abrupt reduction
in bearing resistance when the foundation punches a block of sand into the underlying soft clay in an
uncontrolled manner. This paper details an extensive series of 30 tests of flat circular and spudcan
foundations continuously penetrated through samples of sand overlying clay, and performed under
relevant stress conditions using a drum centrifuge. The large testing area of the drum centrifuge was
used advantageously to produce test results that could be compared directly with tests covering a sand
thickness over foundation diameter of 0.21 to 1.12. Results from retrospective finite-element analysis
of the experiments are also described, with back-calculated values of the stress-level-dependent friction
and dilation angles in the sand during peak penetration resistance shown to fit correlations in the
literature. The back-analysis showed that larger values of peak resistance gave lower friction and
dilation angles, which is consistent with gradual suppression of dilatancy under high confining stress.
When compared with published results from visualisation experiments, the finite-element analysis
showed a similar failure mechanism during peak resistance, with a frustum of sand forced into the
underlying clay at an angle reflecting the dilation in the sand.

KEYWORDS: bearing capacity; centrifuge modelling; finite-element modelling; footings/foundations; offshore


engineering

INTRODUCTION spudcans, which are of the order of 10–20 m in diameter


The potential for unexpected punch-through failure of a (Menzies & Roper, 2008). The geotechnical centrifuge test-
jack-up exists during installation in layered soils (Young et ing programmes of Craig & Chua (1990) and Teh et al.
al., 1984; Osborne & Paisley, 2002; Brennan et al., 2006). (2010) have already provided evidence that they are inade-
This occurs when the circular ‘spudcan’ foundation uncon- quate for the size of spudcan foundations used offshore.
trollably pushes a locally strong zone of soil into underlying The centrifuge experiments described in this paper follow
softer material. Such failures can lead to buckling of the these testing programmes. They are of value to the offshore
jack-up leg, effectively temporarily decommissioning the jack-up community, because they
platform, and even toppling the unit. It has been reported
(a) are performed at stress levels similar to those caused by
that spudcan punch-through failures occur at an average rate
modern spudcan sizes
of one incident per year, costing the industry between US$1
(b) comprehensively cover the range of sand layer height (Hs )
million and US$10 million per incident, owing to rig
to foundation diameter (D) ratio applicable to spudcan
damage and loss of drilling time (Osborne & Paisley, 2002).
punch-through failures
Soil conditions of a thin layer of sand overlying a weaker
(c) are directly comparable with each other, as they were
stratum of clay are particularly hazardous.
performed on the same samples.
The bearing capacity of sand is found to reduce with
increasing footing size (e.g. de Beer, 1963; Kimura et al., To interpret the test results, finite-element analysis was
1985; Ueno et al., 1998), and this is attributed largely to the undertaken to back-calculate the stress-dependent friction
stress-level dependence of the friction angle (Hettler & and dilation angles of the upper sand layer at peak resis-
Gudehus, 1988; Ueno et al., 2001; Zhu et al., 2001). tance. As will be demonstrated, the foundation size effect
Although the size effect on the bearing capacity of sand is can be captured by modifying the strength–dilatancy rela-
well known, there is no established analytical method to tionships proposed by Bolton (1986), and with only simple
quantify it. It is also not considered within the industry Mohr–Coulomb and Tresca models. The extensive experi-
SNAME (2002) standard for predicting peak load on sand mental results, together with back-calculated ‘operative’ fric-
overlying clay soils. The design approaches in SNAME tion and dilation angles, provide a valuable and
(2002) are based on punching shear and load-spreading comprehensive database for developing design approaches
methods that were developed using model tests that do not for the jack-up industry.
share stress levels similar to those caused by modern-day

EXPERIMENTAL DETAILS
Manuscript received 19 November 2012; revised manuscript accepted UWA drum centrifuge
29 May 2013. Published online ahead of print 16 July 2013.
The experiments were conducted in the drum centrifuge at
Discussion on this paper closes on 1 May 2014, for further details
see p. ii. the University of Western Australia (UWA). The drum
 Advanced Geomechanics, Nedlands, Western Australia (former centrifuge has twin concentric shafts that are coupled with a
PhD student of the University of Western Australia). precision servo motor. This allows the central tool table to
† Centre for Offshore Foundation Systems, the University of Western be rotated differentially from the outer sample containment
Australia, Crawley, Western Australia. channel, which is 300 mm high and 200 mm in radial depth.

1271

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


1272 LEE, CASSIDY AND RANDOLPH
The radius to the base of the channel is 600 mm. The tool Table 2. Engineering properties of UWA kaolin clay (after
table can also be locked with the channel, allowing synchro- Stewart, 1992)
nous rotation during testing, or stopped independently of the
channel and lifted vertically. This allows the model founda- Property Value
tions to be changed without affecting the soil in the channel.
A detailed description of the drum centrifuge facility at Liquid limit, LL 0.61
Plastic limit, PL 0.27
UWA was reported by Stewart et al. (1998). Plasticity index, IP 0.34
A major advantage of the drum centrifuge is the large Specific gravity, Gs 2.6
plan area of the test sample. This allows parametric studies Angle of internal friction, 9: degrees 23.5
of foundation response, as a relatively large number of Critical state frictional constant, M 0.92
model tests within a consistent sample can be accommo- Slope of normal consolidation line, º 0.207
dated. This was the case in the research reported here. Two Slope of swelling line, k 0.044
drum centrifuge tests were conducted with 14 model tests in Coefficient of consolidation, cv : m2 /year (at 20 kPa) 2
the first sample and 16 in the second.  After Acosta-Martinez & Gourvenec (2006).

Preparation of soil samples


The two samples of sand overlying clay, named D1 and Once the clay had reached full consolidation, the channel
D2, were prepared in the drum centrifuge channel using a was stopped to prepare for surcharging using sand. First, a
consistent procedure. Commercially available superfine silica fabric blanket was laid on top of the normally consolidated
sand and kaolin clay were used as the upper sand and lower clay. This facilitated the removal of the sand layer after the
clay layers respectively. Both have been used extensively in surcharging process. The channel was then accelerated to
previous model studies at UWA, with their behaviour well 20g before spraying sand onto the clay surface using a hose
documented. The engineering properties of the sand are connecting a sand hopper to a sinusoidal rotating actuator, as
provided in Table 1. The clay has been characterised by described by Gaudin et al. (2005). After the sand-spraying
Stewart (1992), and its properties are summarised in Table 2. process had been completed, the channel speed was increased
further to an acceleration of 300g, and the clay was left to
consolidate under the self-weight of the sand. The clay
Preparation of lower clay layer. The following procedure was settled about another 10 mm with the surcharge until full
followed to achieve an underlying clay with an over- consolidation. This sand had to be removed, because it had
consolidation ratio (OCR) of around 1.5–3.0 depending on been disturbed when the clay settled under the surcharge
the sand thickness during testing. All introduction of soil into load, and hence its homogeneity could not be ensured.
the drum was undertaken with the centrifuge spinning at a Before stopping the channel completely, the solenoid valves
low acceleration level of 20g. at the back of the channel were opened when the acceleration
A sand drainage layer approximately 7 mm thick was first had been reduced to around 10g to drain away the free water
placed at the base of the drum channel to facilitate two-way at the surface of the sand. This created a suction to hold up
drainage of the underlying clay. The pore water at the the upper sands when the channel was stopped completely
bottom sand drainage layer was connected hydraulically to (see Fig. 1(a) and Fig. 5). The thickness of the sand used for
the free water at the surface of the channel through filter surcharging was measured with a ruler, as shown in Fig.
fabrics attached to the wall of the channel at several 1(a), and its relative density was calculated based on the dry
locations. weight of all sand collected during the removal process,
A clay slurry with 120% water content (twice the liquid divided by its volume calculated from the measured sand
limit) was prepared by mixing kaolin clay powder with thickness. The average relative density was 92%, correspond-
water under vacuum. The well-mixed clay slurry was then ing to an effective unit weight of 11.0 kN/m3 : Fig. 1(b)
poured into the drum channel, which had been prefilled with shows the removal process of the fabric blanket after sur-
water by way of a hose fitted on a specialised clay place- charging. Unlike the centrifugal acceleration of 300g used
ment actuator. The nozzle of the hose was always kept during surcharging, all tests were carried out at an accelera-
below the water, to avoid entrapping air during the clay tion of 200g. Therefore the underlying clay had an OCR of
placement process. Once the channel was full, the sample at least 1.5, with an effective unit weight at the sand/clay
was left to consolidate at a centrifugal acceleration of 300g. interface of approximately 7.5 kN/m3 :
Water was supplied continuously to compensate any evapora-
tion during consolidation, and hence ensure that the soil
sample was submerged under water. As the clay settled, Preparation of upper sand layer. After the surcharging stage,
more clay slurry was added in order to achieve a final a new layer of sand was sprayed directly on the preloaded
sample height of around 160 mm after consolidation. The clay. This was performed at a centrifugal acceleration of 20g.
degree of consolidation was estimated to be around 99% Sand was laid to an initial sand height (Hs ) of 31 mm and
from time considerations (with a minimum of 1.5 days’ 33.5 mm in samples D1 and D2 respectively. Both centrifuge
consolidation allowed, following the last addition of slurry), sand samples had a relative density of 92%, and, based on the
and also from the monitored settlements. miniature ball penetrometer tests performed at various

Table 1. Engineering properties of superfine silica sand

Property Value Remark

Specific gravity, Gs 2.65


Average effective particle size, d50 : mm 0.19
Maximum void ratio, emax 0.7472 Corresponds to effective unit weight, ª9 ¼ 9.3 kN/m3
Minimum void ratio, emin 0.4485 Corresponds to effective unit weight, ª9 ¼ 11.2 kN/m3
Critical state friction angle, cv : degrees 31 See White et al. (2008)

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


BEARING CAPACITY ON SAND OVERLYING CLAY SOILS 1273

Measuring sand thickness

Preloaded
clay

(a) (b)

Fig. 1. (a) Removing sand used for surcharging and measuring its thickness with ruler; (b) removing fabric blanket after all sand has
been removed

Undrained shear strength, su: kPa


locations around the drum channel (Lee et al., 2012), the
dense sand was found to be relatively homogeneous. 0 40 80 120
0

Soil strength profiles NT-bar ⫽ 10·5


The undrained shear strength profile of the underlying Nball ⫽ 13·5
clay was estimated by interpretation of T-bar and miniature
ball penetrometer tests conducted at the conclusion of the 4
testing programme, and in the sample after the entire upper
sand layer had been scraped away. This avoided interpreta-
tion of a two-layered penetration test, including the inter-
ference of sand trapped under the penetrometer and brought
down into the clay layer. As an example, the undrained shear 8
strength profiles inferred from both penetrometers for sample
D2 are shown in Fig. 2. A factor of NT-bar ¼ 10.5 was used
in the interpretation of the T-bar resistance profiles, follow-
ing the recommendations of Stewart & Randolph (1994) and
Randolph & Hope (2004), whereas Nball ¼ 13.5 was used for
Depth: m

the ball penetrometer. The larger value reflects a higher 12


interface friction in the ball due to the use of epoxy in its
manufacture, as described in detail in Lee et al. (2012). The
normally consolidated strength ratio of su = v9 ¼ 0:185 de-
rived is consistent for kaolin clay samples in the UWA
centrifuge (Stewart, 1992). For interpretation, however, a 16
linearly increasing strength with depth profile was estimated
for the section of the drum sample representing tests on a
particular thickness of sand. These are provided in Table 3.

20
Model spudcans and flat-based foundations
Tests were conducted on a series of flat-based circular
foundations and generically shaped spudcans (noting that
spudcan shapes and sizes vary among rig builders and
operators).
The shape of the model spudcans was consistent, although 24
the diameters ranged from 40 mm to 80 mm. All had a
conical base inclined at 138 to the horizontal, with a small Fig. 2. Shear strength profile of underlying clay of sample D2
spigot protruding at the centre of the conical base. The tip (after Lee et al., 2012)

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


1274 LEE, CASSIDY AND RANDOLPH
Table 3. Undrained shear strength of underlying clay at different sand thicknesses

Centrifuge test Sand thickness Shear strength at Strength gradient


(equivalent surcharge sand/clay of clay,
pressure on clay):y mm interface, sum : kPa r:{ :kPa/m

Surcharge for D1 32.5 (106 kPa) – –


D1 31 (66 kPa) 17.7 2.1
D1 20.5 (44 kPa) 16.3 2.1
Surcharge for D2 35 (113 kPa) – –
D2 33.5 (73 kPa) 19.1 2.1
D2 29 (63 kPa) 18.6 2.1
D2 24 (53 kPa) 17.9 2.1
D2 17 (38 kPa) 16.6 2.1
 Centrifugal acceleration of 300g was used during surcharging, whereas all tests were carried
out at an acceleration of 200g.
y
Model scale.
{
Prototype scale.

of the spigot had an angle of 768. Fig. 3(a) shows the model behaviour, as related to the properties of the soils and sand
spudcans, and Fig. 3(b) the dimensions of the largest spud- thickness (or ratio of sand thickness to foundation diameter),
can. to be investigated. As the surface of the sand in the circular
The majority of tests were, however, on circular flat-based drum channel was slightly curved, in order to simulate a
foundations. This eliminated the influence of the spigot and flat-based foundation penetrating a flat sand surface using
the conical base, allowing the key aspects of punch-through the drum centrifuge, the model foundations were fabricated
with cylindrically curved bases. These are shown in Fig.
4(a). The schematic drawing shown in Fig. 4(b) presents the
dimensions of the model foundations. The diameters ranged
from 30 mm to 80 mm, and the cylindrical curve at the base
followed the curvature of 420 mm radius. This matched the
curvature of the sand surface.
During testing, a flat-based or spudcan foundation was
connected to a steel rod loading arm on the tool table, which
had a load cell fixed at the other end. The penetration force
was measured by the load cell during penetration, which
was then divided by the area of the foundation to obtain the
(average) penetration resistance, q. The measured loads were
corrected for increasing g level and buoyancy with penetra-
tion, as consistent with centrifuge testing.

EXPERIMENTAL PROGRAMME AND METHOD


Experimental programme
Nine flat-based foundation and five spudcan penetration
tests were performed on sample D1. In the second sample
(D2), 16 flat-based foundation penetration tests were per-
formed. Details of these tests are provided in Tables 4 and 5
for the flat-based and spudcan foundation penetration tests
respectively.
(a) Different sand thicknesses were prepared within each
sample to optimise the testing parameters. To achieve this,
14·6 the thickest sand layer, of 31 mm and 33.5 mm for D1 and
D2 respectively, was initially sprayed onto the underlying
clay. A series of penetration tests were then carried out
within a discrete section of the drum channel. After the
thickest sand layer had been tested, the channel was brought
to a standstill so that thin layers of sand could be scraped
13°

away until the next sand thickness was reached for the next
8
14·8

76° series of tests. These were performed at another section of


6·8

the channel. Particular care was taken to minimise any


10·7 disturbance to the remaining sand. Using the sand-scraping
80 procedure, two different sand thicknesses were created in the
first sample D1 (31 mm and 20.5 mm), whereas sample D2
(b)
had four thicknesses (33.5 mm, 29 mm, 24 mm and 17 mm).
Fig. 3. (a) Model spudcans used in drum centrifuge experiments; On each occasion the sample was left to consolidate under
(b) dimensions of 80 mm diameter spudcan (all measurements in the new sand thickness for 1.5 days before commencement
mm; other diameter foundations have consistent dimensions with of the penetration tests. Fig. 5 shows the sand-scraping
138 and 768 conical angles) process.

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


BEARING CAPACITY ON SAND OVERLYING CLAY SOILS 1275

80 mm diameter
foundation

Series of `flat-based’ foundations

Side view
Plan view

(a)

C
L

M9 thread to fit rod


Cylindrical curvature
of 420 mm radius

(b)

Fig. 4. (a) Model ‘flat-based’ foundations used in drum centrifuge experiments;


(b) dimensions

All flat-based and spudcan foundation penetration tests 0.095 mm/s for the largest, 80 mm diameter foundation to
were located at the centreline of the drum channel to 0.254 mm/s for the smallest, 30 mm foundation. This facili-
minimise boundary effects, with a minimum centre-to-centre tated interpretation, as the effects of strain rate (with a
distance of three times the diameter between the adjacent difference by a factor of 8) and partial consolidation could
foundations. largely be ignored.
The penetration rates also ensured a drained response in
the sand layer. The coefficient of consolidation, cv , for sand
Penetration rate must only exceed 24 000 m2 /year for V to be less than 0.01,
The penetration rates for all tests were based on providing and therefore ensure fully drained conditions. In these
a drained penetration in sand and undrained penetration in experiments the thinnest sand layer had a prototype thick-
clay, following the field observation under typical spudcan ness of 3.4 m. Hence the effective Young’s modulus, E, at
installation rates (SNAME, 2002; Teh, 2007; Osborne et al., the mid depth of the dense sand layer may be estimated as
2011). The normalised penetration rate (Finnie, 1993; Finnie around 20 MPa (Schanz & Vermeer, 1998). Assuming a
& Randolph, 1994) conservative estimate of permeability of more than 106 m/s
vD (Leroueil & Hight, 2003), the superfine silica sand used in
V¼ (1) the centrifuge tests had an estimated cv of at least
cv 60 000 m2 /year. Higher cv values would be expected for
where v is the absolute penetration velocity, D is the thicker sand layers, owing to higher average E values. There
foundation diameter and cv is the coefficient of consolida- is confidence, therefore, that the penetration within the sand
tion, was used to determine the penetration rate required to layer was fully drained for the conditions tested.
achieve these drainage conditions.
Experiments have shown fully undrained behaviour for
normalised velocities greater than about 30, drained for less EXPERIMENTAL RESULTS
than about 0.01, and partially drained in between (Chung et Typical penetration profile
al., 2006). Penetration resistance increases with higher velo- A typical penetration profile from the centrifuge tests is
cities within the undrained region, owing to a strain-rate shown in Fig. 6, in this case for test D1F50a (i.e. prototype
effect (Barbosa-Cruz, 2007; Low et al., 2008). It was ob- scale 6.2 m sand thickness and 10 m diameter, flat-based
served by Low et al. (2008) for penetrometers of 40 mm foundation). In general, the penetration profile consisted of a
and 113 mm diameter that the lowest undrained penetration peak penetration resistance, qpeak , near to the sand surface,
resistance occurred in a transition zone of normalised velo- at a depth of 0.74 m prototype scale in this case, followed
cities of 30 , V , ,300, after which the rate effects became by an abrupt post-peak softening and then a gradual increase
noticeable. The exact transition point will, however, depend of penetration resistance as the foundation continued to
on diameter and consolidation coefficient. penetrate into the deeper clay. The post-peak (locally mini-
For all penetration tests in this drum centrifuge pro- mum) penetration resistance, qpost-peak as labelled in Fig. 6,
gramme the penetration velocity in the underlying clay was was observed to occur well before the sand/clay interface,
kept at V ¼ 120. Therefore the penetration rates varied from and for this case at a depth of around 2.1 m in prototype

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


1276
Table 4. Summary of flat-based penetration tests and experimental results

Experimental details Results

Test Sample Sand Sand Diameter, Diameter, Hs /D ratio Peak penetration Post-peak penetration Punch- Penetration resistance factor
thickness, thickness, D:y mm D:{ m resistance, qpeak resistance, qpost-peak through in clay, Nc ,k at depth below
Hs :y mm Hs :{ m depth:{,} sand/clay interface of
m
kPa Depth:{,} m Depth/Hs kPa Depth:{,} m 0.5D 1.0D

D1F30a 1 31 6. 2 30 6 1.03 712 0.66 0.11 529 2.10 10.7 25.5 25.4
D1F40a 40 8 0.78 521 0.70 0.11 371 1.97 9.3 20.8 19.9
D1F50a 50 10 0.62 447 0.74 0.12 336 2.10 9.0 17.5 15.9
D1F60a 60 12 0.52 385 0.81 0.13 307 1.97 7.2 16.1 16.3
D1F70a 70 14 0.44 342 0.83 0.13 298 1.80 4.3 15.5 15.0yy
D1F80a 80 16 0.39 332 0.78 0.13 300 1.61 3.1 14.8 14.8yy
20.5 4. 1 0.51 0.62 0.15 1.40 3.5 17.4 16.3

LEE, CASSIDY AND RANDOLPH


D1F40b 1 40 8 318 259
D1F50b 50 10 0.41 289 0.69 0.17 255 1.52 2.7 14.9 15.7
D1F60b 60 12 0.34 262 0.72 0.18 238 1.08 1.8 14.0 14.3
D2F30a 2 33.5 6. 7 30 6 1.12 704 0.74 0.11 N.A. N.A. 12.2 22.2 22.1
D2F40a 40 8 0.84 579 0.83 0.12 503 1.66 9.5 22.1 21.5
D2F60a 60 12 0.56 430 0.90 0.13 368 1.89 5.0 21.0 19.7
D2F80a 80 16 0.42 361 0.94 0.14 327 1.74 3.5 14.4 15.8yy
D2F30b 2 29 5. 8 30 6 0.97 707 0.70 0.12 557 1.68 9.4 26.4 25.6
D2F40b 40 8 0.73 490 0.80 0.14 407 1.74 8.7 18.8 18.4
D2F60b 60 12 0.48 373 0.91 0.16 334 1.57 3.0 18.1 17.6
D2F80b 80 16 0.36 314 0.85 0.15 299 1.35 1.1 15.1 16.4yy
D2F30c 2 24 4. 8 30 6 0.80 490 0.67 0.14 420 1.33 7.1 20.6 20.7
D2F40c 40 8 0.60 395 0.74 0.15 345 1.36 6.0 17.4 17.6
D2F60c 60 12 0.40 311 0.70 0.15 294 1.18 1.8 15.3 15.3
D2F80c 80 16 0.30 278 0.83 0.17 273 1.07 0.5 13.8 15.1
D2F30d 2 17 3. 4 30 6 0.57 333 0.84 0.25 287 1.41 5.0 15.5 16.0
D2F40d 40 8 0.43 294 0.67 0.20 272 1.18 3.6 15.0 15.3
D2F60d 60 12 0.28 236 0.59 0.17 225 0.88 0.8 11.6 12.9
D2F80d 80 16 0.21 219 0.73 0.21 218 0.87 0.2 12.8 14.1
 The test naming system is sample–foundation type (F for flat-based foundation and SP for spudcan)–model diameter (in mm)–alphabetic character signifying a particular sand thickness.
y
Model units.
{
Prototype scale at centrifugal acceleration of 200g.
}
Depth at which qpeak or qpost-peak occurred.
}
Distance between qpeak and the next depth with bearing pressure equivalent to qpeak (see Fig. 6 for an example).
k
The Nc factor has been determined by dividing q measured at a normalised depth of 0.5D and 1.0D below the sand/clay interface by the undrained shear strength of the same depth derived from the T-bar tests
(Table 3).
 Test D2F30a did not produce a conclusive post-peak value.
yy
Value reported is the Nc factor at depth ¼ 0.9D to avoid base effect from the centrifuge test.

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


BEARING CAPACITY ON SAND OVERLYING CLAY SOILS 1277

 The N factor has been determined by dividing q measured at a normalised depth of 0.5D and 1.0D below the sand/clay interface by the undrained shear strength of the same depth derived from the T-bar tests
through in clay (Nc ) at depth below
Punch- Penetration resistance factor
Drum

1.0D
sand/clay interface of
channel

14.4yy
18.2
18.0
14.0
14.9

 The test naming system is sample–foundation type (F for flat-based foundation and SP for spudcan)–model diameter (in mm)–alphabetic character signifying a particular sand thickness.
0.5D

18.6
19.2
15.4
14.2
14.5
Craters from a
test series on Scraping to
thicker sand thinner sand layer
resistance, qpost-peak depth:{,k

Hand-dug pocket for


11.6
9.3
11.6
9.4
9.3
m

measuring sand thickness


kPa Depth:{,} m

Fig. 5. Scraping sand after completing a series of tests on thicker


3.70
4.23
4.75
4.48
4.36
penetration

sand
Post-peak

Penetration resistance: kPa


0 200 400 600 800 1000 1200
493
445
422
364
401

0
Results

Peak penetration
2 resistance, qpeak
dpeak /Hs {,}

0.04
0.00
0.05
0.03
0.04

4 Post-peak penetration
Sand resistance, qpost-peak
6
Clay
8
Depth: m
Peak penetration resistance, qpeak

dpeak :{,} m

0.25
0.03
0.32
0.16
0.23

10

12

14 Test D1F50a
Depth:{,} m Depth/Hs

0.31
0.38
0.46
0.47
0.51

Prototype scale:
16
Hs ⫽ 6·2 m
Distance between qpeak and the next depth with bearing pressure equivalent to qpeak (see Fig. 6 for an example).

18 D ⫽ 10 m

20
1.95
2.33
2.84
2.90
3.19

Fig. 6. Typical foundation penetration profile on sand overlying


clay from centrifuge test
Value reported is the Nc factor at depth ¼ 0.9D to avoid base effect from the centrifuge test.
kPa

605
545
513
430
461

scale. After reaching the clay layer, the bearing capacity


increased steadily. In this case a reduction of 111 kPa was
recorded between a qpeak of 447 kPa and a qpost-peak of
Hs /D ratio

Depth at which qpeak occurred, measured from maximum diameter of the spudcan.

336 kPa. Further, not until a depth of 9.7 m did the vertical
0.78
0.62
0.52
0.44
0.39

bearing regain a value equal to qpeak : For a jack-up founda-


Table 5. Summary of spudcan penetration tests and experimental results

tion being load controlled in the field this would represents


a substantial 9.0 m punch-through (equivalent to 0.9D).
Depth at which qpeak or qpost-peak occurred, measured from tip of spigot.
Diameter,
D:{ m

8
10
12
14
16

Summarised results from all tests


The complete penetration profiles of all 25 flat-based and
Diameter,

five spudcan foundations are presented in Figs 7–13 using


D:y mm

prototype scale. Each graph presents the penetration profiles


40
50
60
70
80
Experimental details

of a series of flat-based or spudcan foundations on a


Prototype scale at centrifugal acceleration of 200g.

particular sand thickness. A depth of zero corresponds to the


touching of the foundation base on to the sand surface,
thickness,
Hs :{ m
Sand

whereas for the case of the spudcan it corresponds to the


6.2

instance when the tip of the spigot touches the surface. The
legend of the graph shows the test names, with their
corresponding test configurations tabulated in Tables 4 and
thickness,
Hs :y mm

5. Selected experimental values, such as qpeak and qpost-peak


Sand

31

and their respective penetration depths, are provided in


Tables 4 and 5 for the flat plate and spudcan tests respec-
tively.
Sample

The experimental values of the peak penetration pressure,


1

qpeak , from both centrifuge samples are shown in Fig. 14.


Model units.

They are presented against the prototype areas, grouped in


accordance with their respective sand thicknesses. Symbols
c
(Table 3).
D1SP40a
D1SP50a
D1SP60a
D1SP70a
D1SP80a

in the figure are the actual experimental values, whereas the


Test

lines are hyperbolic functions obtained by best-fit linear


regression between the peak penetration force (qpeak 3 area)
yy
}
{

k
y

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


1278 LEE, CASSIDY AND RANDOLPH
Penetration resistance: kPa Penetration resistance: kPa
0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
0 0
6 m diameter 6 m diameter
16 m flat foundation 2 flat foundation
2
diameter 8 m diameter 8 m diameter
4 10 m diameter 4 12 m diameter
12 m diameter Sand 16 m diameter
Sand 14 m diameter
6 6
Clay 5·8 m sand thickness
Clay 6·2 m sand thickness
8 8

Depth: m
Depth: m

10 10

12 12

14 14

16 16

18 18
Series of D2F30b to D2F80b
Series of D1F30a to D1F80a
20 20

Fig. 7. Penetration profiles for flat-based foundations on 6.2 m Fig. 10. Penetration profiles for flat-based foundations on 5.8 m
prototype sand thickness (sample D1) prototype sand thickness (sample D2)

Penetration resistance: kPa Penetration resistance: kPa


0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
0 0
8 m diameter flat foundation 6 m diameter flat foundation
2 10 m diameter 2 8 m diameter
Sand 12 m diameter 4·1 m sand thickness 12 m diameter
4 4 Sand 16 m diameter 4·8 m sand thickness
Clay
6 6 Clay

8 8
Depth: m

Depth: m

10 10

12 12

14 14

16 16

18 18
Series of D1F40b to D1F60b Series of D2F30c to D2F80c
20 20

Fig. 8. Penetration profiles for flat-based foundations on 4.1 m Fig. 11. Penetration profiles for flat-based foundations on 4.8 m
prototype sand thickness (sample D1) prototype sand thickness (sample D2)

Penetration resistance: kPa Penetration resistance: kPa


0 200 400 600 800 1000 1200 0 200 400 600 800 1000 1200
0 0
6 m diameter 6 m diameter flat foundation
2 flat foundation 2 8 m diameter
8 m diameter Sand 12 m diameter 3·4 m sand thickness
4 12 m diameter 4 Clay
Sand 16 m diameter
6 6
16 m diameter
Clay 6·7 m sand thickness
8 8
Depth: m

Depth: m

10 10

12 12

14 14

16 16

18 18
Series of D2F30a to D2F80a Series of D2F30d to D2F80d
20 20

Fig. 9. Penetration profiles for flat-based foundations on 6.7 m Fig. 12. Penetration profiles for flat-based foundations on 3.4 m
prototype sand thickness (sample D2) prototype sand thickness (sample D2)

and the foundation area. These are not design lines, but that the stress level of the centrifuge tests covers the range
merely a visual aid to demonstrate the consistent trend of of the jack-up industry very well.
the experimental results. As spudcan bearing pressures for At a similar sand thickness, a consistent trend of larger
modern jack-ups are reported to be in the range foundations having lower qpeak values can be observed. This
200–600 kPa (Randolph et al., 2005; Osborne et al., 2009, indicates that although the sand samples prepared in the
2011), the experimental results shown in Fig. 14 indicate centrifuge tests had similar relative density, different values

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


BEARING CAPACITY ON SAND OVERLYING CLAY SOILS 1279
Penetration resistance: kPa
Based on these results, operative friction and dilation angles
0 200 400 600 800 1000 1200
0 for each experiment were back-calculated, as discussed later
8 m diameter spudcan in the paper. It is also shown in Figs 7–13 that the values of
2
10 m diameter qpost-peak generally follow the same trend as the qpeak values.
4 14 m diameter 12 m diameter
Sand 16 m diameter
6
Clay 6·2 m sand thickness
8 Penetration depth at peak resistance. The depths at which
Depth: m

10
qpeak was measured are provided in Tables 4 and 5. These
values are also shown in Fig. 15, where dpeak is defined as the
12 depth of the peak punch-through load. Fig. 15 indicates that
14
the foundation diameter D appears not to play a significant
role in dpeak : This may be due to the relatively low values of
16 Hs /D investigated in this paper, where the foundation is
18 Series of D1SP40a sufficiently close to invoking the punch-through mechanism.
to D1SP80a However, for very high Hs /D ratios, the depth of the peak
20 punch-through load may eventually be related to a critical
Fig. 13. Penetration profiles for spudcans on 6.2 m prototype sand distance between the foundation base and the sand/clay
thickness (sample D1) interface, which would be a function of D. However, such
cases of high Hs /D are beyond the scope of this paper. For
flat-based foundations, the test results show that dpeak falls in
of friction and dilation angles would have been mobilised at a relatively narrow range of 0.11Hs to 0.25Hs : Therefore, for
peak resistance, since they are stress-level dependent; they simplicity and practical purposes for cases with Hs /D < 1.12,
will therefore vary with the magnitude of the failure load dpeak may be assumed to be the average of all the tests, and
arising for various foundation sizes and sand thicknesses. be expressed as

900
D1: Spudcans on 6·2 m sand
D2: Flat foundations on 6·7 m sand
800 D1: Flat foundations on 6·2 m sand
Peak penetration pressure, qpeak: kPa

D2: Flat foundations on 5·8 m sand


700 D2: Flat foundations on 4·8 m sand
D1: Flat foundations on 4·1 m sand
D2: Flat foundations on 3·4 m sand
600
Hyperbola curve fitting

500

400

300

200
0 20 40 60 80 100 120 140 160 180 200 220
Prototype area: m2

Fig. 14. Peak penetration resistance, qpeak , against prototype area

1·0
*D1: Spudcans on 6·2 m sand (*dpeak measured from maximum
diameter of spudcan)
0·9 D2: Flat foundations on 6·7 m sand
D1: Flat foundations on 6·2 m sand
0·8
D2: Flat foundations on 5·8 m sand
0·7 D2: Flat foundations on 4·8 m sand
0·6 D1: Flat foundations on 4·1 m sand
dpeak /Hs

D2: Flat foundations on 3·4 m sand


0·5

0·4

0·3
Flat-based foundations:
0·2 Average dpeak /Hs ⫽ 0·15

0·1
*Spudcans: dpeak /Hs ⬍ 0·05
0
0 1 2 3 4 5 6 7
Sand thickness, Hs: m

Fig. 15. Distance of peak resistance to sand/clay interface (normalised by sand thickness) against
sand thickness

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


1280 LEE, CASSIDY AND RANDOLPH
30
d peak ¼ 0:15H s (2) Nc ⫹ 2·5
Linear trendline:
25 Nc ⫽ 14Hs /D ⫹ 9·5
This is also consistent with recommendations from the
centrifuge test results of Teh (2007) and Teh et al. (2010), 20
although they found a slightly shallower average of 0.12Hs : Nc 15 Nc ⫺ 2·5
Despite the fact that values of dpeak /Hs fall in a relatively
narrow range, Fig. 15 does exhibit some subtle trends of 10
dpeak with varying Hs , in which the normalised dpeak /Hs At depth below sand/clay interface ⫽ 0·5D
5
increases with thinner sand layers (lower Hs ). This is most At depth below sand/clay interface ⫽ 1·0D
distinctly shown for Hs ¼ 3.4 m, where the average 0
dpeak  0.21Hs : This may be because a level of displacement 0 0·2 0·4 0·6 0·8 1·0 1·2
is required to first mobilise the strength of the sand and then Hs /D
shear off the sand block into the underlying clay. This
Fig. 16. Trend of bearing capacity factor of flat-based foundation
mobilisation distance may govern the dpeak for the thinner
and spudcan in underlying clay with normalised sand height
layers. This trend was also observed in the data of Teh
(2007) and Teh et al. (2010).
For the spudcans, dpeak is simplistically defined with its depths of 0.5D and 1.0D below the sand/clay interface is
datum at maximum diameter, as the influence of the spigot provided in Fig. 16. This line should be considered as
and the conical base in mobilising the peak resistance is not indicative behaviour for the two-layered system tested, rather
studied in this paper, owing to the limited number of than definitive bearing capacity factors. In order to indicate
spudcan tests. Using this definition, the test results show that the level of scatter in the dataset, two lines, at Nc  2.5, are
the mobilisation distances to achieve qpeak are less than 5% shown.
of Hs , as shown in Fig. 15. These values are considerably The bearing profiles in the clay are not as consistent as
lower than those from the flat-based foundations. This is the qpeak values. This may be due to the meta-stable state of
because, when the maximum diameter of the spudcan arrives the trapped sand below the foundation. As the foundation
at the sand surface, a portion of soil strength has already penetrates into the underlying clay, some trapped sand may
been mobilised by the progressive penetration of the spigot flow around from below the foundation during the penetra-
and the conical underside of the spudcan foundation. tion process. This will cause the trapped sand to change
It is also noted that mobilising displacements depend on shape and thickness. This is reflected in the penetration
the strength and compression characteristics of both the sand profiles by irregular reductions in the trend of increasing
and the clay, and more tests are recommended to investigate penetration resistance with depth.
this aspect. However, and in recognising the subtle trends in
the datasets, these experimental results provide further vali-
dation and confidence that for engineering purposes, and for INTERPRETATION AND DISCUSSION
the Hs /D ratios tested, the depth of penetration required to Back-analysis of the operative friction and dilation angles at
mobilise the peak resistance, dpeak, can be expressed simply qpeak
as a function of Hs , as proposed by Teh et al. (2010). The friction and dilation angles within the soil will
gradually reduce with increasing level of stress mobilised
beneath the foundation. Small-deformation finite-element
Post-peak response. After the initial peak, a minor ‘second modelling has been used to back-analyse a single ‘operative’
peak’ is observed for some tests, especially those with high friction angle, 9, and dilation angle, ł, at qpeak for each
Hs /D ratios (see the 8 mm diameter test in Fig. 9, for flat-based foundation test.
instance). However, these minor second peaks are normally The commercially available finite-element software
not a critical feature for offshore punch-through problems, PLAXIS (2006) was used with an axisymmetry model of a
since they are smaller than the initial qpeak values. Therefore circular, flat-based foundation of the prototype geometry
no further analysis of trends was performed. simulated in each centrifuge test. The finite-element meshes
Below the sand layer, the penetration resistance in the contained around 500 15-node triangular elements. The
underlying clay generally increases with depth. Undrained vertical boundaries along the centre of the axisymmetry
bearing capacity factors (Nc ¼ q/su ) at depths of 0.5D and model and at the far field located about three foundation
1.0D below the sand/clay interface have been back- diameters from the centreline were prescribed conditions of
calculated from the experimental results, and are presented zero horizontal displacement. Both vertical and horizontal
in Tables 4 and 5. These nominal Nc values are higher than fixities were applied at the bottom boundary, situated at
the range of 9–14 quoted typically for buried flat plates or 15 m below the sand/clay interface. Examination of the
spudcans in clays with strength increasing with depth stress field of the finite-element results showed that this was
(Skempton, 1951; Martin & Randolph, 2001; Houlsby & sufficient to avoid interference from the boundary. The sur-
Martin, 2003; Hossain & Randolph, 2009). This reflects the face of the sand was prescribed as stress free. The rigid, flat-
fact that a significant volume of sand was trapped beneath based foundation was modelled by means of prescribed
the foundations when it penetrated from the sand layer into vertical displacements, indicating that the foundation base
the underlying clay (a phenomenon also observed in post- was assumed to be fully rough.
test dissected samples by Craig & Chua (1990) and Lee The upper sand layer was modelled as a frictional material
(2009)), as well as in the particle image velocimetry (PIV) using the Mohr–Coulomb model, and the underlying clay as
analysis of Teh et al. (2008)). This effectively increases the a Tresca material. This is consistent with the centrifuge test
size of the foundation, enlarging the failure mechanism and conditions of drained penetration in the upper sand layer and
mobilising stronger soil at greater depth. undrained penetration in the underlying clay layer (see
The Nc values tabulated in Tables 4 and 5 are plotted in ‘Penetration rate’ above). Initial soil stress conditions were
Fig. 16 against Hs /D ratio. As expected, the trend is for established in the model according to the effective unit
nominal Nc to increase with increasing Hs /D. This is due to weights of the sand and clay layers. Although the elastic
the increasing size of soil plug becoming trapped under the stiffness of the numerical models did not affect the failure
foundation. A linear trend line for the Nc data recorded at load, typical values were used for all analyses. For the

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


BEARING CAPACITY ON SAND OVERLYING CLAY SOILS 1281
Mohr–Coulomb model, values of E from 50 MPa to depend on the magnitudes of the failure bearing stress
65 MPa, with  ¼ 0.2, were adopted to represent dense sand caused by various foundation sizes and sand thicknesses.
(Schanz & Vermeer, 1998). For the Tresca model, the value As an example, a typical load–settlement curve from the
of Eu was set as a constant value equal to 350 times the finite-element analysis is shown in Fig. 17. This case,
undrained shear strength of the clay at the sand/clay inter- D2F60b, is for a 12 m diameter, flat-based foundation on
face, as the rigidity index (G/su ) ranges typically from 100 5.8 m of sand overlying soft clay. It was observed that the
to 150. A high Poisson’s ratio of  ¼ 0.495 was used to load–settlement curve reached a plateau, which has been
simulate undrained conditions in the underlying clay. The defined as the failure load. The initial elastic region of the
undrained shear strength of the Tresca model was assigned curve was smooth, but it gradually became jagged when
with a profile that increased linearly with depth, following approaching the plastic region. This is an indication of the
the experimental values tabulated in Table 3. numerical instability of the cohesionless sand. In spite of
The actual values for the friction angle, 9, and dilation that, all the finite-element models in this study reached a
angle, ł, of the upper sand layer at peak resistance were not plateau before termination. In this case, values of 9 ¼ 38.38
known. An attempt was made here to determine ‘operative’ and ł¼9.18, obtained from equations (6) and (7) with the
values of 9 and ł for the Mohr–Coulomb model that are experimental qpeak of 373 kPa, gave a numerical qpeak value
consistent with Bolton’s empirical strength–dilatancy rela- of 395 kPa; this matched the experimental value to 5.9%. Of
tionships (Bolton, 1986) through finite-element back-analy- course, 9 and ł could be further adjusted until an exact
sis. match to the experimental value was obtained. However, this
Bolton (1986) studied the effects of density and confining would match only the one test, would require adjustment of
stress on angles of shearing and dilation in sands, and the parameters 2.65 and 3.3 in equations (6) and (7), and
proposed the following empirical relationships. would not necessarily represent an optimal fit to the entire
database of tests.
I R ¼ I D ðQ  ln p9Þ  1, 0 < I R < 4 (3)
Over the entire experimental database, by using the fric-
9  cv ¼ mI R (4) tion angles and dilation angles calculated from equations (6)
0:8ł ¼ 9  cv (5) and (7), the qpeak values obtained from the finite-element
analyses matched very well with the experimental values.
where cv is the critical state friction angle (here 318); p9 is Fig. 18 shows a comparison between the values of qpeak
the mean effective stress at failure (in kPa); IR is a dilation
450
indicator; ID is the relative density index; and Q (established
for this work as Q ¼ 10 for silica sands) may be considered Failure load
400
qpeak ⫽ 395 kPa
as the natural logarithm of the grain-crushing strength (in
Penetration resistance, q: kPa

350
kPa). The value of m determines the enhancement 9 above
FE simulation (PLAXIS) of
the critical state friction angle due to dilatancy, with sug- 300 centrifuge test D2F60b
gested values for m of 3 for triaxial (and general) stress
250 Hs ⫽ 5·8 m, D ⫽ 12 m
conditions and 5 for plane-strain conditions.
Properties of sand layer
To incorporate the empirical strength–dilatancy relation- 200
φ⬘ ⫽ 38·3°, ψ ⫽ 9·1°, γ⬘ ⫽ 11 kN/m3
ships in the foundation problem, it was considered more c ⫽ 0·1 kPa, E ⫽ 50 MPa, v ⫽ 0·2
150
rational to replace the mean effective stress at failure, p9, in Properties of underlying clay
the dilation indicator IR with the peak penetration resistance, 100 φ⬘ ⫽ ψ ⫽ 0, γ⬘ ⫽ 8 kN/m3
qpeak , since an ‘operative’ mean stress is not easily assess- 50 su ⫽ 18·6 ⫹ 2·1 (z ⫺ 5·8 m) kPa
able, and furthermore the (average) mean effective stress in E ⫽ 6·5 MPa, v ⫽ 0·495
the region of interest is directly related to qpeak : Conse- 0
quently, a new value for m must be determined to convert 0 0·02 0·04 0·06 0·08 0·10 0·12
Normalised settlement, δ/D
the contribution of the ‘modified’ IR to the operative friction
angle. This was achieved by iteration. Each flat-based foun-
Fig. 17. Example of a load–settlement curve from PLAXIS (FE,
dation test was simulated in the finite-element analysis for finite element)
different values of m, with the values of 9 and ł for the
Mohr–Coulomb model (for the upper sand layer) calculated 1000
according to equations (3) to (5) for ID ¼ 0.92 (similar for
all tests), but substituting the experimental value of qpeak for
the given test instead of p9. The finite-element values of
qpeak from finite-element analysis: kPa

800
qpeak were then compared with the corresponding experimen-
⫾20%
tal values, adjusting m to arrive at a best overall fit for the
entire set of tests. A value of m ¼ 2.65 was established from
the finite-element back-analysis in this way. The resulting 600
deduced ‘operative’ values of 9 and ł are then given by ⫾10%

9 ¼ cv þ 2:65[I D (Q  ln qpeak )  1] (6)


400 D2: 6·7 m sand
ł ¼ 3:3[I D (Q  ln qpeak )  1] (7) D1: 6·2 m sand
D2: 5·8 m sand
with the square-bracketed terms restricted to the range 0–4. 200 D2: 4·8 m sand
In addition, the cohesion in the Mohr–Coulomb model
Line of equality D1: 4·1 m sand
was set as 0.1 kPa to avoid numerical instability. The above
D2: 3·4 m sand
relationships indicate that larger qpeak values give lower
0
values of 9 and ł, which is consistent with gradual 0 200 400 600 800 1000
suppression of dilatancy under high confining stress. There- qpeak from centrifuge tests: kPa
fore, although the sand samples prepared in the centrifuge
tests had similar relative density of 92%, different values of Fig. 18. Comparison of peak penetration resistance, qpeak , from
9 and ł were mobilised at peak resistance, since they finite-element and centrifuge tests

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


1282 LEE, CASSIDY AND RANDOLPH
from the finite-element analyses and from the centrifuge foundation mechanism, with a small conical wedge near the
tests for all 25 flat-based foundation cases. In all but two centre, and slip surfaces extended to the upper sand layer.
cases, differences of less than 10% were achieved. The results are consistent with the PIV deformation fields
observed during the half spudcan penetration tests on sand
overlying soft clay of Teh et al. (2008), except that the
Failure mechanism at qpeak rupture surface from the visualisation experiments appears
One of the finite-element results will be discussed in more curved. This may be due to the strain-softening behav-
detail in order to examine the failure mechanism of a iour of real sand. That is, the sand near to the spudcan has
foundation on sand overlying clay. Fig. 19 shows the been subjected to larger strains; consequently, the dilation
deformed mesh and load–settlement curve from PLAXIS. angle has been reduced, and the rupture surface becomes
These results simulated experimental test D2F40d, which more vertical with minimal dispersion. On the other hand,
consisted of an 8 m prototype diameter flat-based foundation the lower portions of the rupture surface have larger disper-
on 3.4 m sand thickness. The friction angle, 9 ¼ 38.98, and sion angles due to smaller strains, and hence a larger
dilation angle, ł ¼ 9.88, were back-calculated from the dilation angle. By contrast, the sand was modelled with a
experimental qpeak value from equations (6) and (7), and the single operative friction angle and dilation angle in the
numerical failure load (taken as the plateau or local peak in finite-element analysis presented here.
the load–settlement curve) matched the experimental value
to within 10%.
Figure 20 shows the incremental displacement and shear Development of a new conceptual model
strain shadings during the failure stage. A block of sand is With equations (6) and (7) able to predict the operational
pushed down to the underlying clay, with rupture lines of friction for the entire experimental database, and all the
the sand block at a dispersion angle close to the prescribed finite-element results showing a consistent and definable
dilation angle. Examination of the failure stage for all the failure mechanism, there is potential to use these to develop
other cases showed that only occasionally the dispersion a new conceptual model for predicting qpeak : It will allow an
angle deviated slightly from the dilation angle, but this could iterative calculation approach to determine qpeak and the
be due to some numerical inaccuracy. On the other hand, the stress-dependent 9 and ł simultaneously. A new conceptual
underlying clay failed according to a classical shallow model has been derived, based on this, and is detailed in

300

Sand
250
Penetration resistance, q: kPa

Clay
200

150 qpeak (PLAXIS) ⫽ 269 kPa


qpeak (experiment) ⫽ 294 kPa
100 Error ⫽ ⫺8·7%

Sand: φ⬘ ⫽ 38·9°, ψ ⫽ 9·8°, γ⬘ ⫽ 11 kN/m3


50
Clay: su ⫽ 16·6 ⫹ 2·1 (z ⫺ 3·4) kPa, γ⬘ ⫽ 8 kN/m3
Sand thickness, Hs ⫽ 3·4 m
Foundation diameter, D ⫽ 8 m 0
0 0·02 0·04 0·06 0·08 0·10 0·12
Normalised settlement, δ/D

Fig. 19. Deformed mesh and load–settlement curve of PLAXIS analysis using 9 and ł defined by equations (6) and (7)

Foundation Foundation
radius radius

Sand Sand

Clay Clay

Incremental Incremental
9·8° ⫽ ψ displacement 9·8° ⫽ ψ shear strain
(shading) (shading)

Fig. 20. Incremental displacement and shear strain shadings at failure

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


BEARING CAPACITY ON SAND OVERLYING CLAY SOILS 1283
Lee et al. (2013). This paper also discusses the effect of qpeak peak penetration resistance (average over entire
having a surcharge above the sand layer. foundation)
qpost-peak post-peak penetration resistance (average over entire
foundation)
su undrained shear strength
CONCLUSIONS sum undrained shear strength at sand/clay interface
This paper has detailed an extensive database of centri- V normalised penetration rate
fuge test results for the penetration of flat-based circular and v penetration velocity
spudcan foundations on sand overlying clay. The large test- ª9s effective unit weight of sand
ing area of a drum centrifuge was used advantageously to k slope of swelling line
produce test results that could be compared directly. The º slope of normal consolidation line
experiments included a wide range of ratios of sand thick-  Poisson’s ratio
ness to diameter (Hs /D), varying from 0.21 to 1.12. The r gradient of linear increment of undrained shear strength
with depth
measured peak penetration resistance fell in the range 220–  v9 vertical effective stress
710 kPa. These experimental limits essentially bracket the 9 friction angle
typical range relevant to offshore, with spudcan preloads of cv critical state friction angle
around 200–600 kPa reported by Randolph et al. (2005) and ł dilation angle
Osborne et al. (2009, 2011).
It was found that the measured values of the peak
penetration resistance followed very consistent trends. This
reflects the good quality of the test results, which have REFERENCES
provided reliable experimental data for the development of a Acosta-Martinez, H. E. & Gourvenec, S. (2006). One-dimensional
new conceptual model. consolidation tests on kaolin clay, Research Report GEO:
The experimental results show consistent trends, and 06385. Perth, Australia: University of Western Australia, Geo-
mechanics Group.
provide a high-quality database for development and verifi-
Barbosa-Cruz, E. R. (2007). Partial consolidation and breakthrough
cation of numerical models. Through simple finite-element of shallow foundations in soft soil. PhD thesis, University of
analysis, a modified version of Bolton’s stress–dilatancy was Western Australia, Perth, Australia.
shown to predict the operative friction angle and dilation Bolton, M. D. (1986). The strength and dilatancy of sands. Géo-
angle at the peak resistance of the centrifuge tests well. The technique 36, No. 1, 65–78, http://dx.doi.org/10.1680/geot.1986.
failure mechanism was a sand frustum being sheared off and 36.1.65.
pushed into the underlying clay at a dispersion angle equal Brennan, R., Diana, H., Stonor, R. W. P., Hoyle, M. J. R., Cheng,
to the operative dilation angle. C.-P., Martin, D. & Roper, R. (2006). Installing jackups in
punch-through-sensitive clays. Proceedings of the offshore tech-
nology conference, Houston, TX, USA, paper OTC18268.
Chung, S. F., Randolph, M. F. & Schneider, J. A. (2006). Effect of
ACKNOWLEDGEMENTS penetration rate on penetrometer resistance in clay. J. Geotech.
The research presented here was supported by the Aus- Geoenviron. Engng, ASCE 132, No. 9, 1188–1196.
tralian Research Council through the ARC Linkage grant Craig, W. H. & Chua, K. (1990). Deep penetration of spud-can
scheme (Project LP0561838) and by industry partner Keppel foundations on sand and clay. Géotechnique 40, No. 4, 541–
Offshore and Maritime Limited. This work forms part of 556, http://dx.doi.org/10.1680/geot.1990.40.4.541.
the activities of the Centre for Offshore Foundation Systems de Beer, E. E. (1963). The scale effect in the transposition of the
and the Australian Research Council Centre of Excellence results of deep-sounding tests on the ultimate bearing capacity
of piles and caisson foundations. Géotechnique 13, No. 1,
for Geotechnical Science and Engineering at UWA. The 39–75, http://dx.doi.org/10.1680/geot.1963.13.1.39.
authors acknowledge extensive support through the ARC’s Finnie, I. M. S. (1993). Performance of shallow foundations in
Federation and Future Fellowships, Discovery and Linkage calcareous soil. PhD thesis, University of Western Australia,
programmes. The second author holds the Chair of Offshore Perth, Australia.
Foundations from the Lloyd’s Register Foundation. The Finnie, I. M. S. & Randolph, M. F. (1994). Punch-through and
Lloyd’s Register Foundation supports the advancement of liquefaction induced failure of shallow foundations on calcar-
engineering-related education, and funds research and eous sediments. Proceedings of the 7th international conference
development that enhances the safety of life at sea, on land on the behaviour of offshore structures, BOSS ’94, Boston, MA,
and in the air. All this support is gratefully acknowledged. USA, pp. 217–230.
Gaudin, C., Lehane, B. M. & Wallis, P. F. (2005). A centrifuge
study of the monotonic and cyclic resistance of piles and pile
groups in sand. Proceedings of the international symposium on
NOTATION frontiers in offshore geotechnics (ISFOG 2005) (eds M. Cassidy
cv coefficient of consolidation and S. Gourvenec), pp. 749–755. London, UK: Taylor &
D diameter Francis.
dpeak depth of peak penetration resistance Hettler, A. & Gudehus, G. (1988). Influence of the foundation
d50 average effective particle size width on the bearing capacity factor. Soils Found. 28, No. 4,
E elastic modulus in Mohr–Coulomb model 81–92.
Eu elastic modulus in Tresca model Hossain, M. S. & Randolph, M. F. (2009). New mechanism-based
emax maximum void ratio design approach for spudcan foundations on single layer clay.
emin minimum void ratio J. Geotech. Geoenviron. Engng 135, No. 9, 1264–1274.
G shear modulus Houlsby, G. T. & Martin, C. M. (2003). Undrained bearing capacity
GS specific gravity factors for conical footings on clay. Géotechnique 53, No. 5,
Hs thickness of sand layer 513–520, http://dx.doi.org/10.1680/geot.2003.53.5.513.
ID relative density Kimura, T., Kusakabe, O. & Saitoh, K. (1985). Geotechnical model
IR relative dilatancy index tests of bearing capacity problems in a centrifuge. Géotechnique
Nc bearing capacity factor due to cohesion 35, No. 1, 33–45, http://dx.doi.org/10.1680/geot.1985.35.1.33.
Nball bearing capacity factor for ball penetrometer Lee, K. K. (2009). Investigation of potential spudcan punch-through
NT-bar bearing capacity factor for T-bar penetrometer failure on sand overlying clay soils. PhD thesis, University of
Q natural logarithm of grain-crushing strength (in kPa) Western Australia, Perth, Australia.
q penetration resistance (average over entire foundation) Lee, K. K., Cassidy, M. J. & Randolph, M. F. (2012). Use of epoxy

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.


1284 LEE, CASSIDY AND RANDOLPH
in developing miniature ball penetrometers for application in a ceedings of the 16th international conference on soil mechanics
geotechnical centrifuge. Int. J. Phys. Model. Geotech. 12, No. 3, and geotechnical engineering, Osaka, Japan, pp. 123–176.
119–128. Schanz, T. & Vermeer, P. A. (1998). On the stiffness of sands. In
Lee, K. K., Randolph, M. F. & Cassidy, M. J. (2013). Bearing Pre-failure deformation behaviour of geomaterials (eds R. J.
capacity on sand overlying clay soils: a simplified conceptual Jardine, M. C. R. Davies, D. W. Hight, A. K. C. Smith and
model. Géotechnique http://dx.doi.org/10.1680/geot.12.P.176. S. E. Stallebrass), pp. 383–387. London, UK: Thomas Telford.
Leroueil, S. & Hight, D. W. (2003). Behaviour and properties of Skempton, A. W. (1951). The bearing capacity of clays. Proceed-
natural soils and soft rocks. In Characterisation and engineering ings of the building research congress, London, UK.
properties of natural soils (eds T. S. Tan, K. K. Phoon, D. W. SNAME (2002). Guidelines for site specific assessment of mobile
Hight and S. Leroueil), vol. 1, pp. 29–254. Lisse, the Nether- jackup units, Technical and Research Bulletin 5-5A. Jersey, City,
lands: Swets & Zeitlinger. NJ, USA: The Society of Naval Architects & Marine Engineers.
Low, H. E., Randolph, M. F., DeJong, J. T. & Yafrate, N. J. (2008). Stewart, D. P. (1992). Lateral loading of piled bridge abutments due
Variable rate full-flow penetration tests in intact and remoulded to embankment construction. PhD thesis, University of Western
soil. In Proceedings of the 3rd international conference on Australia, Perth, Australia.
geotechnical and geophysical site characterization (ISC ’3) (eds Stewart, D. P. & Randolph, M. F. (1994). T-bar penetration testing
A.-B. Huang and P. W. Mayne), vol. 1, pp. 1087–1092. London, in soft clay. J. Geotech. Engng Div., ASCE 120, No. 12, 2230–
UK: Taylor & Francis. 2235.
Martin, C. M. & Randolph, M. (2001). Applications of lower and Stewart, D. P., Boyle, R. S. & Randolph, M. F. (1998). Experience
upper bound theorems of plasticity to collapse of circular with a new drum centrifuge. Proceedings of the international
foundations. Proceedings of the 10th international conference on conference Centrifuge ’98 (eds T. Kimura, O. Kusakabe and
computer methods and advances in geomechanics, Tucson, AZ, J. Takemura), vol. 1, pp. 35–40. Rotterdam, the Netherlands:
USA, vol. 2, pp. 1417–1428. A. A. Balkema.
Menzies, D. & Roper, R. (2008). Comparison of jackup rig Teh, K. L. (2007). Punch-through of spudcan foundation in sand
spudcan penetration methods in clay. Proceedings of the off- overlying clay. PhD thesis, National University of Singapore.
shore technology conference, Houston, TX, USA, paper OTC Teh, K. L., Cassidy, M. J., Leung, C. F., Chow, Y. K., Randolph,
19545. M. F. & Quah, C. K. (2008). Revealing the bearing failure
Osborne, J. J. & Paisley, J. M. (2002). SE Asia jack-up punch- mechanisms of a penetrating spudcan through sand overlying
throughs: the way forward. Proceedings of the international clay. Géotechnique 58, No. 10, 793–804, http://dx.doi.org/
conference on offshore site investigation and geotechnics – 10.1680/geot.2008.58.10.793.
sustainability and diversity, London, UK, pp. 301–306. Teh, K. L., Leung, C. F., Chow, Y. K. & Cassidy, M. J. (2010).
Osborne, J. J., Houlsby, G. T., Teh, K. L., Bienen, B., Cassidy, M. Centrifuge model study of spudcan penetration in sand overlying
J., Randolph, M. F. & Leung, C. F. (2009). Improved guidelines clay. Géotechnique 60, No. 11, 825–842, http://dx.doi.org/
for the prediction of geotechnical performance of spudcan 10.1680/geot.8.P.077.
foundations during installation and removal of jack-up units. Ueno, K., Miura, K. & Maeda, Y. (1998). Prediction of ultimate
Proceedings of the 41st offshore technology conference, Hous- bearing capacity of surface footings with regard to size effects.
ton, TX, USA, paper OTC-20291 Soils Found. 38, No. 3, 165–178.
Osborne, J. J., Teh, K. L., Houlsby, G. T., Cassidy, M. J., Bienen, B. Ueno, K., Miura, K., Kusakabe, O. & Nishimura, M. (2001).
& Leung, C. F. (2011). ‘InSafeJIP’ Improved guidelines for the Reappraisal of size effect of bearing capacity from plastic
prediction of geotechnical performance of spudcan foundations solution. J. Geotech. Geoenviron. Engng, ASCE 127, No. 3,
during installation and removal of jack-up units. Final 275–281.
Guidelines of the InSafe Joint Industry Project, RPS Energy White, D. J., Teh, K. L., Leung, C. F. & Chow, Y. K. (2008). A
Report Number EOG0574-Rev1. Woking, UK: RPS Energy. See comparison of the bearing capacity of flat and conical circular
http://insafe.woking.rpsplc.co.uk/ (accessed 04/07/2013). foundations on sand. Géotechnique, 58, No. 10, 781–792, http://
PLAXIS (2006). PLAXIS 2D Version 8.4. http://www.plaxis.nl/. dx.doi.org/10.1680/geot.2008.58.10.781.
Randolph, M. F. & Hope, S. (2004). Effect of cone velocity on Young, A. G., Remmes, B. D. & Meyer, B. J. (1984). Foundation
cone resistance and excess pore pressures. Proceedings of the performance of offshore jack-up drilling rigs. J. Geotech. Engng
international symposium on engineering practice and perform- Div., ASCE 110, No. 7, 841–859.
ance of soft deposits, Osaka, Japan, pp. 147–152. Zhu, F., Clark, J. I. & Phillips, R. (2001). Scale effect of strip and
Randolph, M. F., Cassidy, M. J., Gourvenec, S. & Erbrich, C. T. circular footings resting on dense sand. J. Geotech. Geoenviron.
(2005). Challenges of offshore geotechnical engineering. Pro- Engng, ASCE 127, No. 7, 613–621.

Downloaded by [] on [10/05/19]. Copyright © ICE Publishing, all rights reserved.

Anda mungkin juga menyukai