You are on page 1of 14

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/245411315

A parametric study of the pull-out capacity of bucket foundations in soft clay

Article  in  Géotechnique · January 2001


DOI: 10.1680/geot.2001.51.1.55

CITATIONS READS

69 1,206

3 authors:

Lidija Zdravkovic Richard Jardine


Imperial College London Imperial College London
196 PUBLICATIONS   3,867 CITATIONS    166 PUBLICATIONS   7,454 CITATIONS   

SEE PROFILE SEE PROFILE

David Potts
Imperial College London
252 PUBLICATIONS   7,906 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

ICE-PICK: Installation effects on cyclic axial and lateral performance of displacement piles in chalk View project

Driven single piles and pile groups: static and cyclic field testing in Belfast soft clay View project

All content following this page was uploaded by Richard Jardine on 05 January 2015.

The user has requested enhancement of the downloaded file.


Zdravkovic, L., Potts, D. M. & Jardine, R. J. (2001). GeÂotechnique 51, No. 1, 55±67

A parametric study of the pull-out capacity of bucket foundations in


soft clay
L . Z D R AV KOV I C  , D. M . P OT T S  a n d R . J. JA R D I N E 

This paper investigates the effect of several parameters on Ce papier eÂtudie les effets de plusieurs parameÁtres sur la
the pull-out capacity of bucket foundations in soft clay. capacite de deÂcrochage de caissons baquets dans de l'argile
These parameters are load inclination, skirt length, founda- tendre. Ces parameÁtres sont : inclinaison de la charge, long-
tion diameter, soil adhesion and soil anisotropy. Initially the ueur de jupe, diameÁtre de fondation, adheÂrence du sol et
soil is assumed to be isotropic soft clay, and is represented anisotropie du sol. Initialement, nous supposons que le sol
using a form of the modi®ed Cam clay model. Some recent est une argile tendre isotrope, repreÂsenteÂe par une forme de
laboratory experiments performed in a hollow cylinder ap- modeÁle d'argile Cam modi®eÂ. Nous utilisons ensuite cer-
paratus, which were designed to investigate strength aniso- taines expeÂriences reÂcentes en laboratoire dans un appareil
tropy of a particular silt soil, are then used to represent the cylindrique creux, expeÂriences destineÂes aÁ eÂtudier l'anisotro-
behaviour of an anisotropic soft clay. The anisotropic beha- pie de reÂsistance d'un sol limoneux particulier, pour repreÂ-
viour is simulated using anisotropic soil model MIT-E3. Both senter le comportement d'une argile tendre anisotrope. Nous
models are then used in ®nite element analyses to predict simulons le comportement anisotrope en utilisant le modeÁle
the behaviour of bucket foundations on pull-out. All analyses de sol anisotrope MIT-E3. Les deux modeÁles sont alors
were performed using the three-dimensional Fourier series utiliseÂs dans des analyses d'eÂleÂments ®nis pour preÂdire le
aided ®nite element method (FSAFEM). comportement des caissons baquets au deÂcrochage. Toutes
les analyses ont eÂte faites en utilisant la meÂthode d'eÂleÂments
KEYWORDS: suction; foundations; numerical modelling and ®nis tridimensionnelle de Fourier.
analysis.

INTRODUCTION
Suction caissons (or `bucket' foundations) are becoming exten-
sively used in the offshore industry as deep water anchors for
¯oating structures or foundations for oil platforms. The caissons
are hollow cylindrical structures, which have a top cap and a
relatively thin wall, the so-called skirt. One example of ¯oating
tension leg oil platform (TLP) on bucket foundations is the 310 m depth
Snorre platform, illustrated in Fig. 1, which has a cluster of of water
three buckets connected to each of its four foundation legs
(Christophersen et al., 1992; Jonsrud & Finnesand, 1992).
Installing a bucket foundation involves initial penetration into
the seabed under self-weight. The pressure in the water trapped
inside the bucket, between the soil surface and the top cap, is D = 17 m
then lowered by pumping, to cause a positive differential water
pressure across the top of the bucket, thus forcing the bucket
further into the soil until its ®nal position is reached.
In the case of the Snorre structure, dead weights were added
to ensure that compressive loading was acting on the bucket
foundations. However, oil exploration is moving more into
progressively deeper waters, and there is considerable interest in
utilizing the tensile capacity of bucket foundations. Environ-
mental loading produces both vertical and horizontal compo-
nents of force, such that the resultant force is inclined to the
L = 12 m

vertical. One of the main design considerations for such founda-


tions becomes the assessment of their pull-out capacity.
This paper presents an extensive study of the short-term pull-
out capacity of bucket foundations in soft clay. An initial
parametric study, involving three-dimensional ®nite element
analyses of bucket foundations, in which the diameter, skirt
length, soil-structure adhesion and inclination of loading were Fig. 1. Bucket foundation for Snorre platform
varied, considered the soil to be isotropic soft clay, with an
undrained shear strength increasing linearly with depth. This
study produced a general picture of the in¯uence of each of to investigate the effects of soil anisotropy on the pull-out
these parameters on the pull-out capacity of bucket foundations. capacity of bucket foundations. For this purpose the soil was
Some preliminary results from this study have been reported by modelled using the MIT-E3 soil model (Whittle, 1987; Potts &
Zdravkovic et al. (1998). Further studies were then carried out Zdravkovic, 1999). Owing to the lack of anisotropic experimen-
tal data for soft clays, recent results from hollow cylinder
Manuscript received 11 November 1999; revised manuscript accepted
testing of silt material (Zdravkovic, 1996; Zdravkovic & Jardine,
3 July 2000. 2000) were used to obtain the model parameters. Results from
Discussion on this paper closes 2 July 2001; for further details see p. ii. such anisotropic analyses are compared with those of equivalent
 Department of Civil and Environmental Engineering, Imperial isotropic analyses, thus quantifying the effect of soil anisotropy
College, London. on the pull-out capacity of bucket foundations.

55
56 ZDRAVKOVIC, POTTS AND JARDINE
GEOMETRY
The reference bucket foundation geometry adopted for the
present study is shown in Fig. 2, and is based on the geometry
of the bucket foundations installed at the Snorre platform. A
single concrete cylinder, with a skirt 0´4 m thick and a top cap
1´0 m thick is analysed. The diameter of the cylinder, D, is
17´0 m, while the skirt penetrates the seabed to a depth L ˆ
12:0 m, giving a D=L ratio of 1´4. The depth of water is also
based on that at Snorre, and is assumed to be 310 m.
Undrained uplift loading is applied at the centre of the top
cap, at an inclination è to the vertical; see Fig. 2. Analyses
were performed varying the inclination è from 08 to 908. The
scenario considered is typical of that for a TLP. Other types of
¯oating structure may apply their loads in different ways.

FINITE ELEMENT ANALYSIS


Although the bucket foundation has an axisymmetric geome-
try, loading at a general inclination è produces a full three- Fig. 3. Geometry of bucket foundation considered in FSAFEM
dimensional state of stress and strain in the bucket and in the
soil. This implies that any realistic numerical analysis of this
problem must be three-dimensional. Only in the special case of formed, which further reduced the computational effort re-
vertical loading (i.e. è ˆ 08) can the problem be assumed to be quired.
axisymmetric. The bucket foundation was modelled as concrete material,
One way of analysing such problems is to perform a conven- assuming linear elastic behaviour (with Young's modulus
tional three-dimensional ®nite element analysis, which requires E ˆ 30 3 106 kN=m2 and Poisson's ratio ì ˆ 0:15). This im-
a three-dimensional mesh of solid elements. This puts substan- plies that the concrete has suf®cient strength and reinforcement
tial requirements on computer memory and time resources. to sustain the applied loads. The installation of the foundation
However, problems that have an axisymmetric geometry, but was not modelled in the analyses; the `bucket' was assumed to
non-axisymmetric loading and/or material properties, may be be `wished in place'.
analysed more ef®ciently by the Fourier series aided ®nite For the conventional axisymmetric analyses, when è ˆ 08,
element method (FSAFEM) (Potts & Zdravkovic, 1999). This the foundation was loaded to failure by applying increments of
method allows a two-dimensional ®nite element discretization vertical uplift displacement, v, at the centre of the top cap. The
of the problem geometry (in the r± z plane, see Fig. 3), while reaction force at this location represents the applied load. For
displacements, loads, etc. are allowed to vary as a Fourier series the foundation subjected to inclined loading, when è 6ˆ 08, the
in the third (out of plane) direction, â. The accuracy of an analyses had to be performed under load control, with incre-
analysis depends on both the element discretization and the ments of vertical and horizontal force being applied to the
number of Fourier series harmonics used to represent the centre of the top cap. The magnitudes of these increments were
variation of variables in the â direction. Potts & Zdravkovic kept in proportions required to give the desired loading inclina-
(1999) show that the method produces similar results to those tions. As failure was approached, it was necessary to reduce the
of conventional three-dimensional ®nite element analysis, but size of the force increments to obtain an accurate estimate of
can reduce computer runtimes by an order of magnitude. the ultimate loads.
The FSAFEM method was used in this study for the analyses The ®nite element mesh shown in Fig. 4 was used for both
of bucket foundations subjected to inclined loading, adopting axisymmetric and FSAFEM analyses. It consisted of 350 eight-
the ®nite element mesh presented in Fig. 4. As indicated in this noded isoparametric elements. All analyses were performed
®gure, prescribed displacements were imposed on both the using the Imperial College Finite Element Program (ICFEP).
bottom and right-hand side mesh boundary. No such displace- Reduced (2 3 2) integration was used, and undrained conditions
ment restrictions are imposed on the left-hand side boundary, in the soil were simulated by assigning a high bulk stiffness
because it represents only the geometrical axis of symmetry. (ˆ 1000 K s , where K s is the effective bulk stiffness of the soil
Ten harmonics were used to represent the variation of all skeleton) to the pore water (Naylor, 1974; Potts & Zdravkovic,
quantities in the â direction. A small parametric study, varying 1999). An accelerated modi®ed Newton±Raphson scheme, with
the number of harmonics, was performed to ensure that this was a sub-stepping stress point algorithm, was employed to solve
adequate. For a vertically loaded foundation (i.e. è ˆ 08) a the nonlinear ®nite element equations (Potts & Ganendra,
conventional axisymmetric ®nite element analysis was per- 1994).
The interface between the top cap and the soil underneath
was modelled by introducing a string of interface elements (Day
& Potts, 1994). The investigation of possible conditions that
T might exist at this interface and the adopted scenario for all the
analyses were discussed in Zdravkovic et al. (1998).
θ
D = 17 m

ISOTROPIC STUDY
1m
Soil conditions
As mentioned in the introduction, the initial study on the
pull-out capacity of bucket foundations in soft clay was per-
formed by assuming the soil to be isotropic. A lightly over-
L = 12 m

z
consolidated clay was assumed (OCR ˆ 1´1, Mair et al., 1992)
and a form of the modi®ed Cam clay (MCC) constitutive model
0·4 m
was used to simulate its behaviour (Roscoe & Burland, 1968).
The yield and plastic potential surfaces are given by a Mohr-
Coulomb hexagon and a circle respectively in the deviatoric
plane (Potts & Gens, 1984; Gens & Potts, 1988). This differs
Fig. 2. Idealized geometry of bucket foundation from many implementations of the model, which assume that
PULL-OUT CAPACITY OF BUCKET FOUNDATIONS IN SOFT CLAY 57
D /2 = 8·5 m

12 m

50 m
80 m

Fig. 4. Finite element mesh for the analysis of bucket foundations

both the yield and plastic potential surfaces are given by a means that the full undrained strength is mobilized between
circle in the deviatoric plane. The advantages of the present the soil and the skirt at failure (i.e. a ˆ 1). To investigate
implementation are described by Potts & Zdravkovic (1999). the effect of a reduced skirt adhesion, some of the analyses
Basic details of the model are given in Appendix 1, and the were repeated with a skirt adhesion equal to 50% of the
material properties used in the present investigation are sum- undrained shear strength of the adjacent soil.
marized in Table 1. The triaxial compression undrained shear
strength, which can be derived from the basic parameters as In each of these studies the inclination, è, of the pull-out
indicated in Appendix 1, varies linearly with depth, giving force, T, was varied between 08 and 908.
Su =ó v9 ˆ 0:33, which is typical of soft clays in triaxial compres-
sion (Hight et al., 1987).
Results
Results from a typical analysis are shown in Fig. 5. This
Parametric studies
analysis is for a bucket foundation with D ˆ 17:0 m and L ˆ
The reference geometry, identi®ed in Fig. 2, gives a D=L
12:0 m (i.e. D=L ˆ 1:4), subjected to a pull-out force, T,
ratio of 1´4. Three parametric studies were performed to inves-
inclined at 708 to the vertical (i.e. è ˆ 708). The vertical com-
tigate the pull-out capacity of bucket foundations:
ponent, V, of this force is plotted against the vertical displace-
(a) To investigate the effect of embedment, the skirt length was ment, and the horizontal component, H, of this force is plotted
varied, keeping the same diameter (D ˆ 17:0 m), to L ˆ against the horizontal displacement of the foundation. Fig. 5
8:0 m (D=L ˆ 2:1), L ˆ 12:0 m (D=L ˆ 1:4), L ˆ 17:0 m shows that clear limit values are predicted for both the horizon-
(D=L ˆ 1:0), and L ˆ 24:0 m (D=L ˆ 0:7). tal and vertical load components at failure.
(b) To investigate the effect of the bucket diameter, analyses Figure 6 shows, for this particular analysis, the vectors of
were performed with similar D/L ratios, but with a diameter incremental displacements at failure. While the absolute magni-
D ˆ 8:5 m. tudes of these vectors are not signi®cant, the directions and
(c) Most of the analyses were performed assuming full relative magnitudes indicate the nature of the failure mechan-
adhesion, cw (ˆ aSu ), between the soil and the skirt, which ism.

Table 1. Modi®ed Cam clay parameters for soft clay


Parameter Description Value
v1 Speci®c volume for a K o normally consolidated sample at p9 ˆ 1 kPa 3´1
ö9TC Critical state angle of shearing resistance in triaxial compression 328
ë Slope of the VCL in v±ln p9 space 0´2
k0 Slope of the swelling line in v±ln p9 space 0´03
ì Poisson's ratio 0´3

 While the use of an `a' approach was convenient, it is recognized that


skirt capacity is governed by effective stress processes, similar to those
that control the behaviour of driven piles (Jardine & Chow, 1996).
58 ZDRAVKOVIC, POTTS AND JARDINE

14 000
Hult Full skirt adhesion. The combinations of ultimate vertical, Vult ,
and horizontal, H ult , loads at which failure is predicted, consid-
ering a range of bucket skirt lengths and diameters and full skirt
Horizontal and vertical force: kN

12 000 H
adhesion, are plotted on the interaction diagram in Fig. 7. As
10 000 θ = atan(H/V ) = 70˚ expected, the magnitudes of the ultimate loads are dependent on
D /L = 17/12 the diameter and depth of embedment. However, for a particular
8000
foundation geometry the ®nite element predictions indicate that
there is an approximately elliptical relationship between the
6000 Vult
ultimate vertical and horizontal loads, developed under inclined
4000 V loading with è varying from 08 to 908. The ellipses ®tted to the
numerical predictions are plotted in Fig. 7. The maximum error
2000 between these curves and the numerical predictions is  (0´6 to
1´2)%. Three main conclusions can be derived from this ®gure:
0
0 0·2 0·4 0·6 0·8 1·0 1·2 1·4 (a) For a particular geometry, the largest pull-out capacity
Horizontal and vertical displacement: m occurs under vertical loading, and it continually reduces as
the inclination of loading increases (i.e. becomes more
Fig. 5. Typical load±displacement curves for inclined loading horizontal.

70˚ T

Fig. 6. Vectors of incremental displacements at the end of analysis

90˚ Fitted ellipses


Ultimate horizontal force, Hult: kN

D/L = 17/12
30 000
D/L = 17/8
70˚ 45˚
D/L = 17/17
20˚
20 000 D/L = 17/24
10˚ D/L = 8·5/8·5

10 000


0
0 30 000 60 000 90 000 120 000 150 000
Ultimate vertical tensile force, Vult: kN

Fig. 7. Envelopes of ultimate horizontal and vertical load (full skirt adhesion)
PULL-OUT CAPACITY OF BUCKET FOUNDATIONS IN SOFT CLAY 59
(b) The capacities of bucket foundations with constant diameter soil cannot sustain shear stress). Further analyses have been
are nearly directly proportional to their skirt length (e.g. performed to determine the limiting value of D=L over which
halving the length of the skirt, and hence its contact area equation (1) does not apply. These analyses indicate that equa-
with the soil, practically halves the limit loads). tion (1) is valid for D=L , 4.
(c) Changing the bucket diameter, while keeping the same skirt
length, has a larger impact than varying the skirt length Reduced skirt adhesion. In the analyses described above,
while keeping the same diameter (e.g. halving the diameter where the full adhesion between the foundation skirt and the
results in a 3´5 times reduction in ultimate loads, although adjacent soil was assumed, there was no need to incorporate
the contact area between the soil and skirt reduces by a interface elements in this locality. However, interface elements
factor of 2). were introduced between the skirt and the soil for the set of
analyses in which the skirt adhesion was assumed to be 50% of
The results plotted in Fig. 7 are replotted in a normalized the undrained strength of the adjacent soil (i.e. cw ˆ 0:5Su ).
form in Fig. 8. The data from a particular foundation geometry These elements were assigned normal and shear stiffnesses of
and for a particular load inclination have been normalized by 100G and 10G respectively, where G is the elastic shear
dividing the ultimate horizontal force, H ult , and the ultimate modulus of the adjacent soil, which varies with depth.
vertical force, Vult , by the maximum vertical force, Vmax . Vmax Two sets of analyses were repeated with the reduced skirt
is the maximum vertical load found from the analysis involving adhesion, considering D=L ˆ 17=12 and D=L ˆ 17=24. The
purely vertical loading for that particular foundation geometry. resulting interaction diagram is presented in Fig. 9. We note:
The values of H ult =Vmax and Vult =Vmax found for all the
(a) The ®nite element predictions for a particular geometry can
geometries and pull-out inclinations indicate a unique relation-
be ®tted with an elliptical curve, similar to the analyses
ship that can be represented by an ellipse of the following
performed with a full skirt adhesion, shown as full lines in
form:
Fig. 9.
H 2ult (b) The ultimate loads obtained for 50% reduced skirt adhesion
: ‡ V 2ult ˆ V2max (1) are approximately 10±15% smaller than those for the full
0 2352
skirt adhesion (marked as dashed lines in Fig. 9), indicating
All data ®t this ellipse with a maximum error of only  (1´5 to
that the reduction in ultimate loads is not directly
3´0)%.
proportional to the reduction in skirt adhesion.
Clearly, equation (1) cannot hold for large D=L ratios
because in the limit, when the embedment depth L approaches The results are also plotted in normalized form in Fig. 10,
zero, the ultimate horizontal force also becomes zero (i.e. the together with the ellipse from equation (1). They agree with this
water interface between the underside of the top cap and the curve with a maximum error of 6%.

θ = 90˚ 70˚ 45˚ 20˚ 10˚ 0˚

0·4
D/L = 17/12
D/L = 17/8

0·3 D/L = 17/17


D/L = 17/24
D/L = 8·5/8·5
Hult /Vmax

0·2

0·1
Ellipse from equation (1)

0
0 0·1 0·2 0·3 0·4 0·5 0·6 0·7 0·8 0·9 1·0
Vult /Vmax

Fig. 8. Normalized ultimate loads (full skirt adhesion)

Fitted ellipses
Ultimate horizontal force, Hult: kN

45˚
30 000 20˚ D/L = 17/12

D/L = 17/24
20 000
a = 1·0

10 000
a = 0·5

0
0 30 000 60 000 90 000 120 000 150 000
Ultimate vertical force, Vult: kN

Fig. 9. Envelopes of ultimate horizontal and vertical load (reduced skirt adhesion)
60 ZDRAVKOVIC, POTTS AND JARDINE
θ = 90˚ 45˚ 20˚ 0˚

0·4

D/L = 17/12
0·3

D/L = 17/24
Hult /Vmax

0·2

0·1
Ellipse from equation (1)

0
0 0·1 0·2 0·3 0·4 0·5 0·6 0·7 0·8 0·9 1·0
Vult /Vmax

Fig. 10. Normalized ultimate loads (50% skirt adhesion)

OBTAINING ANISOTROPIC SOIL PROPERTIES W MT


A further series of analyses was performed to investigate
how soil anisotropy might affect the pull-out capacity of bucket
foundations. po
θ pi

Testing equipment and procedures r σz


Anisotropic yielding and strength properties simulated in the
numerical study followed results obtained from experiments τzθ
performed in the Imperial College hollow cylinder apparatus τθz
(Zdravkovic, 1996; Zdravkovic & Jardine, 2000). This apparatus σθ
(Hight et al., 1983) allows control over four of the component
σr
stresses, ó z , ó r , ó è and ôzè, and hence control over the
directions (á) and magnitudes of the principal stresses (see Fig.
11). The con®guration also permits full control over the para- (b)
meter b(ˆ (ó 2 ÿ ó 3 )=(ó 1 ÿ ó 3 )). The tests on silt, on which the
analyses were based, involved controlling both b and á.
The silt, whose index properties are summarized in Table 2,
is an industrially produced rock ¯our, obtained by crushing z
σ1 α
quartz stone. Samples of HPF4 were prepared from a slurried (a)
silt±water mixture pluviated through water into membrane-lined
moulds, a procedure that is analogous to natural sedimentation σ3
through water. These samples had an initial void ratio in the α
range 0:65  0:03, giving an average initial relative density of
88%. εz σ2
HCA tests have also been used at Imperial College to
investigate the initial anisotropy of K o consolidated sand
(d)
(Porovic & Jardine, 1994; Porovic, 1995); clay±silt±sand and
clay±sand mixtures (Menkiti, 1995; Jardine & Menkiti, 1999). εθ
εr γzθ
Description of tests. Six HCA experiments (series M) have b = (σ2 – σ3)/(σ1 – σ3)
been performed on samples K o (Ko ˆ 0:5) consolidated to
OCR ˆ 1:0 (Zdravkovic, 1996; Zdravkovic & Jardine, 2000). A (c)
key diagram describing their effective stress paths is presented in
Fig. 12. Tests are denoted as M, followed by the á value at Fig. 11. Stress and strain components within the Imperial College
failure, áf , in degrees. Each test commenced with K o hollow cylinder apparatus: (a) hollow cylinder coordinates; (b)
consolidation to a mean effective stress p9 ˆ 200 kPa [point A; element component stresses; (c) element component strains; (d)
element principal stresses
p9 ˆ (ó 19 ‡ ó 29 ‡ ó 39 )=3, where ó 19 , ó 29 and ó 39 are major,

Table 2. Index properties of HPF4 silt


emin emax wL (%) wP (%) Speci®c gravity Particle size: mm Mineralogy
d 10 ˆ 0:005
0´563 1´332 27 0 2´65 d 50 ˆ 0:040 100% quartz
d 90 ˆ 0:100
PULL-OUT CAPACITY OF BUCKET FOUNDATIONS IN SOFT CLAY 61
strains), all samples showed a strong tendency for dilation
(generating negative pore water pressures), until the tests reached
Mαf
b-stage: their arbitrary ®nal strains. These ®nal strains were limited by
b = 0 to 0·5
α = 0˚
the load capacity of the apparatus, as well as by the proximity of
A local strain instrumentation to the deforming sample.
75
Phase transformation
Undrained From the beginning of shearing ± that is, point Cáf ± all
shearing: stress paths exhibited a steady increase in deviatoric stress,
b = 0·5 without having a distinct peak. Therefore the phase transforma-
t : kPa

Undrained
α = αf
unloading: tion points are chosen for comparison of different stress±strain
tion b = 0·5 conditions developed during the constant á-shearing phases of
lida the tests.
so Cαf Bαf α = 0˚
con Undrained In this respect, Fig. 13 indicates a constant decrease of the
Ko
α-rotation: deviatoric stress mobilized at PT, as á changes from 08 to 908,
b = 0·5
α = 0˚ to αf
reaching a maximum reduction of 65% when á ˆ 908. The
difference between the undrained strengths mobilized at á ˆ 08
p ′: kPa 200 and á ˆ 908 represents the amount of undrained strength aniso-
tropy for a particular material. A similar trend is observed for
Fig. 12. General scheme for M-series tests the angle of shearing resistance mobilized at the PT points,
which shows a maximum reduction of 178 when á changes
between 08 and 908. These results for HPF4 silt are summarized
intermediate and minor principal effective stress], followed by in Fig. 14: the thick solid line in Fig. 14(a) shows the change
drained creep, which allowed strains to stabilize. The value of b of deviatoric stress mobilized at PT, normalized by the mean
was then changed from 0 to 0´5 to simulate nominal plane strain effective stress at the end of K o consolidation, plotted against
conditions. The stress state (i.e. t± p9) remains at point A the direction á of the major principal stress. Similarly, the thick
(t ˆ (ó 19 ÿ ó 39 )=2 ˆ 75 kPa, p9 ˆ 200 kPa) during this drained solid line in Fig. 14(b) shows the change of ö9 mobilized at PT
change. Ideally, different degrees of principal stress rotation against á, for HPF4 silt.
should have been imposed directly from point A, as a common Also shown on these ®gures are results for three other
point for all tests. However, earlier work on anisotropically materials tested in the IC HCA: Ham River sand (HRS) a
consolidated Ham River sand (Porovic, 1995) and sand±clay mixture of HRS with 10% kaolin (HK) (Menkiti, 1995; Jardine
mixtures (Menkiti, 1995) showed that only a relatively small & Menkiti, 1999); and an arti®cial clay mixture of 50% kaolin
principal stress rotation (, 208) can be applied to normally with 25% silt and 25% ®ne sand (KSS) (Menkiti, 1995; Jardine
consolidated samples before they reach the local bounding & Menkiti, 1999). These results show that sand and silt have a
surface and collapse. In order to investigate behaviour at larger á similar overall strength anisotropy, with mobilized undrained
values it was necessary to bring the samples' stress states inside and drained strengths reducing dramatically as á increases from
the bounding surface by partial undrained unloading, while 08 to 908. Adding clay to sands reduces their anisotropic
keeping b ˆ 0:5 and á ˆ 08. The unloading stage with subse- potential. However, natural clays often possess a stronger degree
quent undrained creep is indicated as path A to Báf in Fig. 12. of anisotropy than that shown for KSS: Porovic (1995) reports
The undrained principal stress rotation, marked as `á-rotation', is 60% undrained strength anisotropy for a natural silty clay from
then applied from point Báf to point Cáf , involving rotation of Pentre (UK), whereas Leroueil (1977) shows a 65% difference
the ó 1 direction from 08 to áf and a subsequent undrained creep between triaxial compression and extension undrained strengths
phase. The ®nal stage, `constant á-shearing' involves undrained for soft Champlain clay from Canada. Because of the lack of
shearing from point Cáf to failure, keeping á and b constant at detailed experimental data on the anisotropy of natural clays,
áf and 0´5 respectively. the results obtained for the HPF4 silt material (which shows a
similar degree of anisotropy as the above natural soft clays)
Results. Figure 13 presents the undrained effective stress were used to derive the anisotropic model parameters for the
paths, in t± p9 space, followed by the M-series tests (see numerical study.
Zdravkovic & Jardine, 2000, for details of these tests). Shearing
from the respective Cáf points results in an initial contraction,
generating positive pore water pressures. After reaching a phase ANISOTROPIC STUDY
transformation ppoint
 (PT), usually at 1´5% to 2% deviatoric Modelling anisotropic soil behaviour
:
strain åd (ˆ 2= 6[(å1 ÿ å2 )2 ‡ (å2 ÿ å3 )2 ‡ (å1 ÿ å3 )2 ]0 5 , where Constitutive soil models developed in the last 30 years have
å1 , å2 and å3 are major, intermediate and minor principal mainly considered soil to be isotropic and therefore represen-
table with strength parameters that are independent of direction
(e.g. c9, ö9, Su ). It is usually assumed that these parameters
150
may be obtained from triaxial tests, although plane strain or
M0 simple shear experiments are sometimes used, depending on the
Phase transformation problem analysed. Anisotropic models are rarely used because
125 of the lack of experimental data on soil anisotropy, which is a
M15
direct result of the inability of conventional laboratory equip-
100 M30 ment to investigate such behaviour.
Well-designed hollow cylinder devices and directional shear
cells now enable soil anisotropy to be investigated more thor-
t : kPa

M45 A (also
75
B0, C0) oughly, allowing soil models capable of simulating the observed
M70
B15 anisotropic behaviour to be developed.
50 M90 C30 C15

C45
B30 MIT soil models. A comprehensive suite of anisotropic soil
25 B45 models has been developed at the Massachusetts Institute of
C70 B70 Technology: Kavvadas (1982) developed the MIT-E1 model for
B90, C90 normally consolidated clays, and Whittle (1987) extended it into
0
0 20 40 60 80 100 120 140 160 180 200 220
MIT-E3 for the behaviour of overconsolidated clays. Both
p ′: kPa models are loosely based on modi®ed Cam clay (Roscoe &
Burland, 1968). However, several extensions have been made to
Fig. 13. Effective stress paths for M-series HCA tests the basic critical state formulation to enable the representation of
62 ZDRAVKOVIC, POTTS AND JARDINE
0·7 60
KSS (ult), p′o = 400 kPa
55
0·6 HK (PTP), p′o = 300 kPa
HRS (PTP), p′o = 200 kPa 50

0·5 HPF4 (PTP), p o′ = 200 kPa


45

40

φ′: degrees
0·4
t /p′o

35

0·3
30

25
0·2

20
0·1
15

0 10
0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90
α˚ α˚
(a) (b)

Fig. 14. Initial strength anisotropy of four K o consolidated materials

certain features of soil behaviour not realized by modi®ed Cam that the observed change in undrained strength with á is well
clay. The key features of MIT-E1 are an anisotropic yield predicted, indicating that the MIT-E3 model is particularly good
surface, kinematic plasticity, and signi®cant strain-softening at simulating the anisotropic strength behaviour associated with
behaviour under undrained conditions. The two additional the change of principal stress direction orientation. However, it
features incorporated into MIT-E3 are small-strain non-linear is de®cient in one important aspect: while the silt soil showed
elasticity using a closed-loop hysteretic stress±strain formula- dilatant behaviour after reaching phase transformation, MIT-E3
tion, and bounding surface plasticity, wherein plastic straining indicates ductile failure. This was not thought to be a serious
occurs within the conventionally de®ned yield locus. Whittle
(1993) demonstrated the ability of MIT-E3 to represent
accurately the behaviour of three different clays when subjected 150
to a variety of loading paths. Both models can simulate initial
and induced anisotropic behaviour.
The MIT-E3 model has been incorporated at Imperial College M0
into ICFEP (Ganendra & Potts, 1995; Potts & Zdravkovic, 1999), M15
and a brief description of MIT-E3 is given in Appendix 2. 100 M30
M45
t : kPa

A
Soil conditions M70
M90
For this set of analyses the seabed clays were assumed to
have a similar pattern of strength anisotropy to that seen in the 50
B
silt experiments summarized in Fig. 13. Appropriate parameters lidatio
n
for the MIT-E3 model are given in Table 3. o nso
Ko c
The parameters listed in Table 3 were used in numerical
simulations of the HCA test data. The samples were, numeri-
0
cally, subjected to stress paths similar to those given in Fig. 13, 0 50 100 150 200
except that all samples were unloaded to the same point B p ′: kPa
before the á-rotation stage. The predicted effective stress paths
are presented in Fig. 15. Comparison of Figs 13 and 15 shows Fig. 15. Theoretical prediction of the HCA stress paths

Table 3. MIT-E3 parameters for HPF4 silt


Parameter Description Value
v100 Speci®c volume for a K o normally consolidated sample at p9 ˆ 100 kPa 1´668
K NC
o Normally consolidated coef®cient of earth pressure at rest 0´5
ö9TC Critical-state angle of shearing resistance in triaxial compression 358
ö9TE Critical-state angle of shearing resistance in triaxial extension 308
c Ratio of the semi-axes of the bounding surface ellipsoid 1´0
ø Parameter affecting rotation of bounding surface 100
ë Slope of the VCL in v±ln p9 space 0´014
k0 Initial slope of the swelling line in v±ln p9 space 0´0018
ì Poisson's ratio 0´3
St Parameter affecting the degree of strain softening 1´0
C Parameter affecting the hysteretic elasticity 1´0
n Parameter affecting the hysteretic elasticity 1´5
ù Parameter affecting the hysteretic elasticity 0´0
PULL-OUT CAPACITY OF BUCKET FOUNDATIONS IN SOFT CLAY 63
de®ciency, as the analyses were intended to model foundations trying to model the anisotropy experienced in the HCA. The
on a soft clay, which shows similar behaviour to the silt up to difference has no bearing on the conclusions that come.
phase transformation, without dilating at larger strains (Leroueil,
1977; Porovic, 1995).
The undrained triaxial compression strength pro®le for this Results
soft clay is shown in Fig. 16. It varies linearly with depth, In order to quantify the effect of soil anisotropy on the pull-
starting from a ®nite value at the soil surface, giving out capacity of bucket foundations two sets of analyses were
Su =ó v9 ˆ 0:36. This would be the pro®le normally adopted in performed, assuming the foundation to have a reference geom-
any analysis using an isotropic soil model. However, MIT-E3 etry (D=L ˆ 17=12) and full skirt adhesion:
allows the undrained strength pro®le to change depending on
(a) analyses where the bucket foundation was loaded with a
the direction of the major principal stress, as shown in Fig. 16.
tensile force whose inclination, è, varied from 08 to 908 and
This triaxial compression pro®le differs slightly from that used
in which the soil was assumed to be anisotropic ± that is, to
in the previous isotropic analyses, but is a consequence of
possess the variation of undrained strength presented in Figs
15 and 16
(b) analyses where the bucket foundation was loaded in the
same way as in (a), but the soil was assumed to be
0 isotropic. The soil properties for the MIT-E3 model were
chosen such that, for any inclination á of the major
principal stress, the undrained strength was the same as that
K NC
o = 0·5 obtained from the anisotropic soil when sheared with
σ′vo = 30 + 10z á ˆ 08. This means that the effective stress paths for any
b=0
–5 á match the effective stress path for test M0 in Fig. 15.
Again, an interaction diagram of ultimate vertical and hor-
izontal forces for each particular inclination of loading is
produced, as shown in Fig. 17. The upper envelope represents
the results from the isotropic analyses, while the lower envelope
–10
Triaxial compression is for the anisotropic analyses. Similar to the previous analyses,
both isotropic and anisotropic numerical predictions can be
®tted with a simple elliptical curve, with a maximum error of
 (0´6 to 1´2)%. However, the main conclusion from these
Depth, z: m

analyses is that taking account of soil anisotropy reduces the


–15
pull-out capacity of the bucket foundation. For this particular
soil the anisotropic analyses lead to capacities that are about
22% lower than those for isotropic cases. Clearly, the reduction
of pull-out capacity due to anisotropic soil behaviour will
depend on the amount of strength anisotropy that each particu-
–20 lar soil possesses.
The resulting normalized interaction diagram is presented in
Fig. 18, together with the ellipse from equation (1). The
α=

analytical data points conform to the curve with a maximum


error of 6%.
α=

–25
α=

45˚
90˚

CONCLUSIONS
A parametric ®nite element study has quanti®ed the effects
on the pull-out capacity of `bucket' foundations of (i) founda-
–30 tion geometry, (ii) skirt adhesion and (iii) load inclination. The
0 20 40 60 80 100 120
main conclusions are as follows:
Undrained strength, Su: kPa
(a) The largest pull-out capacity occurs under vertical loading.
Fig. 16. Variation of undrained strength with depth for anisotropic (b) Pull-out capacity continually reduces as the inclination of
MIT-E3 model loading rotates from the vertical.

θ = 90˚

Fitted ellipses
Ultimate horizontal force, Hult: kN

50 000
θ = 45˚ Isotropic MIT-E3 model

40 000 θ = 17˚
Anisotropic MIT-E3 model

30 000
θ = 6˚

20 000

10 000
θ = 0˚
0
0 50 000 100 000 150 000 200 000
Ultimate vertical force, Vult: kN

Fig. 17. Envelopes for ultimate horizontal and vertical load, using (d) isotropic and (m) anisotropic MIT-E3 model
64 ZDRAVKOVIC, POTTS AND JARDINE
θ = 90˚ 45˚ 17˚ 6˚ 0˚

0·4

0·3
Hult /Vmax

0·2
Isotropic MIT-E3 model

Anisotropic MIT-E3 model


0·1
Ellipse from equation (1)

0
0 0·1 0·2 0·3 0·4 0·5 0·6 0·7 0·8 0·9 1·0
Vult /Vmax

Fig. 18. Normalized ultimate loads from (d) isotropic and (m) anisotropic analyses using MIT-E3 model

(c) Foundation diameter affects pull-out capacity more sig- speci®c volume at unit pressure on the current swelling line, and is
ni®cantly than does embedment depth. therefore not a material property.
(d) Ultimate pull-out capacity is not directly proportional to the Behaviour under increasing deviator stress is assumed to be elastic
value of skirt adhesion. until reaching a yield curve of the form
 2  
J p9o
Finite element analyses using the MIT-E3 model have shown Fˆ ÿ ÿ1 ˆ0 (4)
p9 g y (è) p9
how account may be taken of the anisotropic soil strength
observed in the laboratory. Anisotropy in Su was seen to reduce where
the pull-out capacity by approximately 22%, compared with :
J ˆ 1=6[(ó 19 ÿ ó 29 )2 ‡ (ó 29 ÿ ó 39 )2 ‡ (ó 19 ÿ ó 39 )2 ]0 5 ˆ the deviatoric
equivalent analyses assuming isotropic behaviour. stress;
It has also been shown that normalizing the ultimate horizon- p9 ˆ 1=3(ó 19 ‡ ó 29 ‡ ó 39 ) ˆ the mean effective stress;
tal and vertical pull-out forces with respect to the ultimate ó 19 , ó 29 , ó 39 are the principal effective stresses;
vertical force, obtained from an analysis involving only vertical p9o ˆ exp[(v1 ÿ vs )=(ë ÿ k)]: the value of p9 at the intersection of the
pull-out, results in a unique interaction curve, at least for the current swelling line and the virgin consolidation line. It de®nes the
soil pro®les considered and for D=L , 4, that speci®es a non- position of the yield surface, and is therefore the hardening
dimensional failure envelope that is independent of the founda- parameter.
tion dimensions, skirt adhesion or soil anisotropy. This curve is The function g y (è) expresses the manner in which the yield surface
of a simple elliptical form with ratio of minor to major axis of varies in the deviatoric plane ( p9 ˆ constant) and takes the following
0´235, and once the vertical pull-out capacity has been evalu- form:
ated, designers may use such a curve to determine the capacity sin ö9
under any inclined loading conditions, without performing addi- g y (è) ˆ (5)
1
tional complex three-dimensional analyses. cos è ‡ p sin è sin ö9
3
Note that the above conclusions are valid for TLP-type
bucket foundations that are loaded at the centre of the top cap. where
p
If such foundations are used as suction anchors, for mooring è ˆ tanÿ1 ((2b ÿ 1)= 3), Lode's angle;
other vessel types, the load is often applied at a point some way b ˆ (ó 29 ÿ ó 19 )=(ó 19 ÿ ó 39 );
down the skirt. In such cases it is likely that some other ö9 is the angle of shearing resistance.
inclination of pull-out force (rather than vertical) will give the This function gives the yield surface the shape of a Mohr±Coulomb
maximum capacity. Such anchors are currently under investiga- hexagon in the deviatoric plane.
tion at Imperial College. Note also that the above conclusions Equation (4) plots as an ellipse, in terms of p9 and J, above each
are based purely on the results of a numerical study. Validation swelling line given by equation (3), and the major axis of the ellipse is a
against ®eld measurements and/or laboratory model tests would function of p9o , which varies with v in accordance with equation (2).
be bene®cial. The plastic strain increment vector is assumed to be perpendicular to a
plastic potential surface that is also given by equation (4), but with gy (è)
replaced by gp (è) ˆ X, where X is a parameter that varies such that
when the soil is yielding, the current plastic potential passes through the
APPENDIX 1. MODIFIED CAM CLAY CONSTITUTIVE MODEL current state of stress. The plastic potential is therefore not a function of
The soil model adopted for the ®nite element calculations is a form of the Lode's angle, è, and it has a circular shape in the deviatoric plane. By
the modi®ed Cam clay model of Roscoe & Burland (1968). In this model adopting such an expression it can be shown that plane strain failure will
it is assumed that the consolidation characteristics of the model may be occur with è ˆ 08 (i.e. b ˆ 0:5); see Potts & Gens (1984). The yield
adequately represented in v±ln p9 space by the virgin consolidation line function and plastic potential surfaces therefore differ only in their
v ˆ v1 ÿ ë ln( p9) (2) shapes in the deviatoric plane. In terms of the nomenclature of classical
plasticity, the yield surface and its plastic potential are non-associated.
and the family of swelling lines Soil behaviour under unloading and reloading of the mean effective
v ˆ vs ÿ k ln( p9) (3) stress p9 is assumed to obey equation (3), which gives the elastic bulk
modulus K B as
where v is the speci®c volume (ˆ 1 ‡ e, e being the void ratio), ë is the
v p9
slope of the virgin consolidation line, k is the slope of the swelling lines, KB ˆ (6)
and v1 is the virgin consolidation speci®c volume at unit pressure. k
Parameters ë, k and v1 are material properties. Parameter vs is the This expression indicates that the elastic bulk modulus depends on the
PULL-OUT CAPACITY OF BUCKET FOUNDATIONS IN SOFT CLAY 65
stress level p9 and the speci®c volume v. To complete the speci®cation
of elastic behaviour a second parameter is required. In the present study a
constant Poisson's ratio, ì, was assumed.
To completely specify the soil model, values of the following
parameters are required: ë, k, v1 , ö9 and ì. It is also necessary to

2by ÿ bx ÿ bz
establish the initial state of stresses in the soil by allocating values to the

2bz ÿ bx
overconsolidation ratio OCR ˆ ó v9 max =ó v9 , and the coef®cient of earth

b3 ˆ p2bxy
b4 ˆ p2 byz
b5 ˆ 2 bzx
Anisotropy

p
6
b2 ˆ p
p 2
pressure at rest, K o ˆ ó h9 =ó v9 , where ó v9 and ó h9 are the vertical and

(1, b)
1
horizontal effective stresses respectively.
Note that the above parameters are essentially effective stress

b1 ˆ
parameters. It is not possible to input directly the undrained strength
or its change with effective stress. However, the undrained strength can
be derived from the basic parameters, and is given by the following
equation:
 k=ë
Su 1 2(1 ‡ 2K oc
o )
ˆ g(è)cos è(1 ‡ 2K o )OCR(1 ‡ A2 ) nc
ó v9 6 (1 ‡ 2K o )OCR(1 ‡ A ) 2

(7)
where
sin ö9

Plastic ¯ow direction

2Py ÿ Px ÿ Pz
g(è) ˆ ;

P ˆ Px ‡ Py ‡ Pz
1

2Pz ÿ Px
cos è ‡ p sin è sin ö9

P3 ˆ p2 Pxy
 Pyz
P5 ˆ 2 Pzx
p
6
P2 ˆ p
p 2
3

(P, P)

P4 ˆ p2
p
3(1 ÿ K nc
o )

g(ÿ308)(1 ‡ 2K nco )

P1 ˆ
K nc
o ˆ 1 ÿ sin ö9;

K oc nc
o ˆ K o OCR
sinö9
;
and è ˆ the Lode's angle; OCR ˆ the overconsolidation ratio; ó v9 ˆ the
vertical effective stress; Su ˆ the undrained shear strength.
Note from the above equation that the undrained strength varies with
the value of the Lode's angle. It will therefore be different under triaxial
compression (è ˆ ÿ308), triaxial extension (è ˆ 308) and plane strain
(è ˆ 08) conditions. Yield surface gradient

2Q y ÿ Qx ÿ Qz
Q ˆ Qx ‡ Qy ‡ Qz

2Qz ÿ Qx

Q3 ˆ p2Qxy
 Qyz
Q5 ˆ 2 Qzx
p
6
p
p 2
APPENDIX 2. MIT-E3 CONSTITUTIVE MODEL
(Q, Q)

Q4 ˆ p2
All the MIT models are expressed in terms of so called `transformed
variables': see Table 4. The transformed variables for any quantity (e.g.
stresses, strains) comprise a linear combination of the tensorial measures Q2 ˆ
Q1 ˆ

of the quantity. This is necessary because the models reproduce


anisotropic soil behaviour, and in such cases all six components of the
stress tensor and six components of the strain tensor, or a meaningful
combination of these components, have to be used. The use of invariants
(e.g. q, p9, ås , åv , . . .) inevitably leads to an isotropic formulation, as
assumed by most of the currently existing models.

MIT-E3 model
Bounding surface. The bounding surface for the MIT-E3 model is an
2åy ÿ åx ÿ åz
å ˆ å x ‡ åy ‡ åz

anisotropic form of the elliptical modi®ed Cam clay yield surface, and is
2åz ÿ åx

E3 ˆ p2åxy
åyz
E5 ˆ 2åzx
p
6
E2 ˆ p
p 2

de®ned in generalized stress space using transformed variables (see Fig.


E4 ˆ p2
(å, E)
Strain

19):
X
5
E1 ˆ

Fˆ (s i ÿ p9b i )2 ÿ c2 p9(2È ÿ p9) ˆ 0 (8)


iˆ1

where c is the ratio of semi-axes of the ellipsoid, and is a material


Table 4. Transformed variables for MIT-E3 model

property; È is a scalar variable that de®nes the size of the bounding


surface; b ˆ fbi g, i ˆ 1, . . ., 5, is a vector that describes the orientation
of the bounding surface and is variable (see Table 4); and s ˆ fsi g,
i ˆ 1, . . ., 5, is the deviatoric stress vector expressed using transformed
variables (see Table 4).
The bounding surface is a distorted ellipsoid in six-dimensional stress
space, with its longer axis along the direction â (see Fig. 19). â is a
vector that indicates the principal direction of anisotropy, and ⠈ b ‡ I,
where fIgT ˆ f100 000g. â is initially orientated in the direction of
consolidation, and rotates as a function of any subsequent loading, in a
manner de®ned by the model's hardening laws.
A virgin normally consolidated soil element lies at the tip of the
bounding surface, at a stress state: p9 ˆ 2È and si ˆ 2Èbi .
2ó y9 ÿ ó x9 ÿ ó z9
ó x9 ‡ ó y9 ‡ ó z9
Effective stress

2ó z9 ÿ ó x9
p
6

Failure criterion. The failure criterion, at which critical state


3

s3 ˆ p2ó xy
ó yz
s5 ˆ 2ó zx
p 2
s2 ˆ p

behaviour is exhibited, is de®ned in generalized stress space by an


s4 ˆ p2

anisotropic conical surface with its apex at the origin and with its axis
( p9, s)

along the direction (I ‡ î). This surface is labelled the critical state
p9 ˆ

s1 ˆ

cone, and is illustrated in Fig. 20. î ˆ fî i g, i ˆ 1, . . ., 5 is a vector that


66 ZDRAVKOVIC, POTTS AND JARDINE
5 Bounding surface, F
F = Σ (s i –p ′bi)2 – c 2p ′(2Θ – p ′) = 0
s i=1 x

P s
P
(s – p′b) Load surface, f0 2Θ, 2Θb
cΘ C A
C I
C1 First yield b +I
I surface, f0i β=
b+ Θb
β= R
x O ial
O Rad ing
p′ Projection p
map
centre
p′
Θ Θ
s

Fig. 19. Geometry of the bounding surface for MIT-E3 model: P, 2Θ0i
current stress state; C, intersection of bounding surface axis and 2Θ0
current ð plane; A, tip of the boundary surface; C1 , centre of
bounding surface ellipsoid 2Θ

s Critical- Fig. 21. Illustration of the radial mapping rule


Critical-
state A state
cone cone variable that is a measure of the relative orientation of the yield surface
Rc to the critical state cone; hrx i ˆ rx if rx . 0; ˆ 0 if rx < 0.
There are three important properties of equation (11), which controls
P P the rate of rotation of the yield surface:
Rx C Bounding
C (a) When the axes of stress, s, and anisotropy, b, coincide, there is no
O′ surface
rotation of the yield surface.
O′ Bounding (b) In general, as plastic deformation occurs, the axes of anisotropy
surface
O rotate towards the stress axes.
O
p′ (c) The axes of anisotropy do not rotate outside the critical state cone.

Behaviour inside the bounding surface. A stress state within the


Critical-state bounding surface represents overconsolidation. The plastic behaviour at
s
cone any overconsolidated point R inside the bounding surface is linked to its
image point I on the bounding surface (see Fig. 21). This image point is
Fig. 20. Geometry of the bounding surface and the critical state described using a radial mapping rule. This rule enables the image point
cone: O9, intersection of axis of critical-state cone and current ð associated with any stress state within the bounding surface to be
plane; Rc , projection of vector CP onto the critical-state cone; Rx , obtained by constructing a radial line that passes through the stress point
projection of vector O9C onto the critical-state cone and the origin. The loading condition for plastic strains is then de®ned by
the gradients of the bounding surface at the image point:
X5 
> 0 for loading
de®nes the anisotropic nature of the failure criterion, and is a material KQI Äåv ‡ 2G QIi ÄE i (12)
, 0 for unloading
property. The critical state cone is described by the equation iˆ1

X
5 where QI is the spherical component of the bounding surface gradient at
hˆ (s i ÿ p9î i )2 ÿ k 2 p92 ˆ 0 (9) the image point; QI ˆ fQi g, i ˆ 1, . . ., 5, is the deviatoric component of
iˆ1 the bounding surface gradient at the image point; K is bulk modulus; G
where k is a scalar material property that de®nes the size of the cone. is shear modulus; åv is volumetric strain; and E ˆ fEi g, i ˆ 1, . . ., 5,
vector of deviatoric strains in transformed variables.
For further details about the model see Whittle (1987) and Potts &
Flow rule. The ¯ow rule satis®es two important criteria: Zdravkovic (1999).

(a) The model reaches a critical state at the critical state cone: that is,
it has plastic deviatoric strains, but neither any plastic volumetric
strains nor any strain hardening/softening. REFERENCES
(b) If initially virgin consolidated under K o conditions such that the Christophersen, H. P., Bysveen, S. & Stove, O. J. (1992). Innovative
stress state is at the tip of the bounding surface, any subsequent foundation systems selected for the Snorre ®eld development. Be-
loading under K o conditions does not change the normally haviour of Offshore Structures 1, 81±94.
consolidated value of K o . Kavvadas (1982) showed that to satisfy Day, R. A. & Potts, D. M. (1994). Zero thickness interface elements.
this criterion in general, a non-associated ¯ow rule is required. Int. J. Numer. Anal. Methods Geomech. 18, 689±708.
Ganendra, D. & Potts, D. M. (1995). Discussion on `Evaluation of
The ¯ow rule is expressed directly in terms of spherical, P, and constitutive model for overconsolidated clays' by A. J. Whittle.
deviatoric, P ˆ fPi g, i ˆ 1, . . ., 5, components of the ¯ow direction (the GeÂotechnique 45, 169±173.
derivatives of the plastic potential). No equation is given for the plastic Gens, A. & Potts, D. M. (1988). Critical state model in computational
potential P(fó9g, fmg). geomechanics. Engng Comput. 15, 178±197.
Hight, D. W., Gens, A. & Symes, M. J. (1983). The development of a
new hollow cylinder apparatus for investigating the effects of
Hardening laws. The hardening laws consist of two equations, one principal stress rotation in soils. GeÂotechnique 33, 355±384.
controlling the change of the size of the bounding surface, ÄÈ, and Hight, D. W., Jardine, R. J. & Gens, A. (1987). The behaviour of soft
another controlling the change of its orientation, Äbi : clays: Embankments on soft clay. Bull. Public Works Research
ÄÈ ˆ ÈæÄåvp (10) Centre, Athens, 33±158.
Jardine, R. J. & Chow, F. C. (1996). New design methods for offshore
1
Äb i ˆ ø0 hrx i (s i ÿ p9b i )Äåvp (11) piles. MTD publication 96/103, London.
È Jardine, R. J. & Menkiti, C. O. (1999). The undrained anisotropy of K 0
where æ is a variable that affects the rate at which the size of the yield consolidated sediments. Proc. 12th ECSMFE ± Geotechnical engin-
surface changes; ø0 is a dimensionless material property that controls the eering for transportation infrastructure, Amsterdam, 2, 1101±1108.
rate of rotation of the yield surface; and rx (ˆ CRx =O9Rx ) is a scalar Jonsrud, R. & Finnesand, G. (1992). Instrumentation for monitoring the
PULL-OUT CAPACITY OF BUCKET FOUNDATIONS IN SOFT CLAY 67
installation and performance of the concrete foundation templates implicit stress point algorithms. Comput. Methods Appl. Mech.
for the Snorre tension leg platform. Behaviour of Offshore Structures Engng 119, 341±354.
1, 690±703. Potts, D. M. & Gens, A. (1984). The effect of plastic potential in
Kavvadas, M. (1982). Nonlinear consolidation around driven piles in boundary value problems involving plane strain deformation. Int. J.
clays. PhD thesis, Massachusetts Institute of Technology. Numer. Anal. Meth. Geomech. 8, 259±286.
Leroueil, S. (1977). Quelques consideÂrations sur le comportement des Potts, D. M. & Zdravkovic, L. (1999). Finite element analysis in
argiles sensibles. PhD thesis, University of Laval, Canada. geotechnical engineering: Theory. London: Thomas Telford.
Mair, R. J., Hight, D. W. & Potts, D. M. (1992). Finite element analyses Roscoe, R. H. & Burland, J. B. (1968). On the generalised stress±strain
of settlements above a tunnel in soft ground, Contractors Report 265. behaviour of `wet' clay: Engineering plasticity. pp. 553±609. Cam-
Transport and Road Research Laboratory, Crowthorne. bridge: Cambridge University Press.
Menkiti, C. O. (1995). Behaviour of clay and clayey sand, with Whittle, A. J. (1987). A constitutive model for overconsolidated clays
particular reference to principal stress rotation. PhD thesis, Imperial with application to the cyclic loading of friction piles. PhD thesis,
College, University of London. Massachusetts Institute of Technology.
Naylor, D. J. (1974). Stresses in nearly incompressible materials by ®nite Whittle, A. J. (1993). Evaluation of a constitutive model for over-
elements with application to the calculation of excess pore water consolidated clays. GeÂotechnique 43, 289±313.
pressure. Int. J. Num. Methods Engng 8, 443±460. Zdravkovic, L. (1996). The stress±strain±strength anisotropy of a
Porovic, E. (1995). Investigation of soil behaviour using a resonant granular medium under general stress conditions. PhD thesis, Imper-
column torsional-shear hollow cylinder apparatus. PhD thesis, Im- ial College, University of London.
perial College, University of London. Zdravkovic, L. & Jardine, R. J. (2000). Undrained anisotropy of K o
Porovic, E. & Jardine, R. J. (1994). Some observations on the static and consolidated silt. Can. Geotech. J. 37, 178±200.
dynamic shear stiffness of Ham River sand. In Pre-failure deforma- Zdravkovic, L., Potts, D. M. & Jardine, R. J. (1998). Pull-out capacity
tion of geomaterials (eds S. Shibuya, T. Mitachi and S. Miura), of bucket foundations in soft clay. Proceedings of the international
Balkema, Rotterdam, 1, pp. 25±30. conference on offshore site investigation and foundation behaviour ±
Potts, D. M. & Ganendra, D. (1994). An evaluation of substepping and New Frontiers, Vol. SUT 1, pp. 301±324.

View publication stats

You might also like