Martin, C. M. & Houlsby, G. T. (2001). GeÂotechnique 51, No.
8, 687±699
Combined loading of spudcan foundations on clay: numerical modelling
C . M . M A RT I N  a n d G . T. H O U L S B Y 
The load±displacement response of a spudcan foundation on
clay is described by means of an incremental plasticity
model with three degrees of freedom (vertical, rotational,
horizontal). The model is termed `Model B' and employs a
yield surface and ¯ow rule that are derived from a pro-
gramme of carefully controlled small-scale laboratory tests.
Behaviour inside the yield surface is de®ned by a set of
elastic stiffness factors for embedded conical footings, deter-
mined from three-dimensional ®nite element analysis. The
hardening law, which de®nes the vertical bearing capacity as
a function of plastic spudcan penetration, is based on a set
of theoretical lower bound bearing capacity factors for
embedded conical footings. Care is needed to ensure consis-
tent treatment of partially and fully penetrated spudcans
under combined loading. Full details of the elastic and
elasto-plastic stiffness matrices needed to predict incremen-
tal load±displacement behaviour are provided, and the per-
formance of the model is demonstrated by simulating some
of the laboratory calibration tests numerically. While Model
B succeeds in capturing many important features of the
observed spudcan behaviour, its prediction of sudden rather
than gradual yielding is a noticeable weakness, particularly
in the simulation of cyclic loading events. Practical applica-
tions of the foundation model are discussed with reference to
the soil±structure interaction analysis of independent leg
jack-up units.
KEYWORDS: bearing capacity; footings/foundations; numerical
modelling and analysis; offshore engineering; plasticity; soil±
structure interaction.
INTRODUCTION
An offshore jack-up drilling platform typically has three or four
latticework legs, which are supported by individual `spudcan'
foundations (Poulos, 1988). The spudcans are primarily sub-
jected to vertical loading, but during storms they also experi-
ence signi®cant overturning moments and horizontal shearing
loads. Numerical modelling of foundation behaviour under these
conditions is a challenging three-dimensional problem, and one
that is not readily amenable to detailed ®nite element analysis
(Martin & Houlsby, 2000). In the interests of simplicity and
computational ef®ciency, researchers have tended to concentrate
on models that idealise the spudcan footing as a rigid body
attached to vertical, rotational and horizontal springs. The more
advanced models expressed in terms of stress resultants, such as
the one developed in this paper, incorporate both yielding of the
foundation and coupling between the various degrees of free-
dom. This allows a realistic characterisation of spudcan behav-
iour under combined loading, but one that remains easy to
incorporate into the analysis of a complete jack-up unit. In
practice, environmental loading is usually assumed to act along
one of the structural axes of symmetry, thus allowing the jack-
up to be idealised as an equivalent plane frame. The problem of
modelling spudcan response then becomes one of relating three
co-planar loads (V , M, H) and their corresponding displace-
ments (w, è, u). Fig. 1 de®nes these variables and illustrates the
sign convention adopted here.
Manuscript received 16 June 2000; revised manuscript accepted 8 May
2001.
Discussion on this paper closes 1 April 2002, for further details see
inside back cover.
 Department of Engineering Science, University of Oxford, UK.
687
Nous deÂcrivons la reÂponse au deÂplacement de charge d'une
fondation `spudcan' sur de l'argile au moyen d'un modeÁle de
plasticite increÂmentiel ayant trois degreÂs de liberte (verticale,
rotationnelle et horizontale). Ce modeÁle, appele `ModeÁle B',
emploie une surface de limite eÂlastique et une reÁgle d'eÂcoule-
ment deÂriveÂes d'un programme d'essais en laboratoire aÁ petite
eÂchelle soigneusement controÃleÂs (Martin & Houlsby, 2000).
Nous eÂtablissons le comportement aÁ l'inteÂrieur de cette surface
par une seÂrie de facteurs de rigidite eÂlastique pour des pieds
coniques enterreÂs, facteurs deÂtermineÂs apreÁs des analyses
d'eÂleÂments ®nis en trois dimensions. La loi de durcissement qui
de®nit la capacite porteuse verticale comme fonction de la
peÂneÂtration plastique des spudcans, est baseÂe sur un ensemble
de facteurs theÂoriques de capacite porteuse de limite infeÂrieure
pour les pieds coniques enterreÂs. Nous avons veille aÁ maintenir
un traitement coheÂrent des spudcans peÂneÂtreÂs partiellement et
des spudcans peÂneÂtreÂs entieÁrement sous charge combineÂe.
Nous donnons tous les deÂtails des matrices de rigidite eÂlas-
tiques et eÂlastoplastiques neÂcessaires pour preÂvoir le comporte-
ment increÂmentiel au deÂplacement de charge et nous
deÂmontrons la performance du modeÁle en simulant de manieÁre
numeÂrique quelques-uns des essais de calibrage en laboratoire.
Tandis que le ModeÁle B reÂussit aÁ saisir un grand nombre de
caracteÂristiques importantes du comportement observe des
spudcans, sa preÂdiction d'un ¯eÂchissement soudain plutoÃt que
graduel constitue un eÂchec notable, particulieÁrement dans la
simulation des eÂveÂnements de charge cyclique. Nous discutons
des applications pratiques de ce modeÁle de fondations en
faisant reÂfeÂrence aÁ l'analyse de l'interaction sol-structure des
veÂrins eÂleÂvateurs aÁ pieds indeÂpendants.
2R
Undisturbed soil
surface level
O
Reference
position
w
M
H
O
Current θ
position
V
u
Fig. 1. Notation and sign convention for spudcan loads and
displacements
The most rudimentary version of the spring-based spudcan
footing model is the pin joint (zero rotational stiffness, in®nite
vertical and horizontal stiffnesses). This is still extensively used
in jack-up analyses in the manner described by Reardon (1986),
but its shortcomings are well understood and it is now usual to
include at least a ®nite rotational stiffness, mainly because
critical bending moments at the leg±hull interface of the struc-
ture are reduced. This has important economic implications,
688 MARTIN AND HOULSBY
since a given jack-up unit can operate safely in deeper water (or
harsher environmental conditions) if a certain level of rotational
spudcan ®xity can be relied upon. For jack-up analyses such as
fatigue and natural period studies it was formerly regarded as
acceptable to model the foundations using linear elastic springs
(Hambly & Nicholson, 1989; Carlsen et al., 1986). However, it is
now widely acknowledged that predictions of structural perform-
ance, especially for extreme storm loading, should take some
account of non-linear spudcan behaviour. This can be achieved
using either an iterative analysis, with secant foundation stiff-
nesses (Noble Denton & Associates, 1987; Arnesen et al., 1988;
SNAME, 1997), or an incremental analysis, with tangent founda-
tion stiffnesses (Schotman, 1989; Hambly et al., 1990; Martin,
1994; Thompson, 1996; Cassidy, 1999; Martin & Houlsby,
1999). An important result in dynamic analysis is that the
rotational stiffness at the spudcan has a major in¯uence on the
natural period of the unit (Cassidy, 1999).
The jack-up unit analyses of Schotman (1989) marked a
signi®cant development because the tangent stiffness matrix for
each spudcan was derived from a work-hardening plasticity
formulation. In the simplest form of this type of model, footing
behaviour is taken to be linear elastic as long as the (V , M, H)
load combination lies within a three-dimensional yield surface.
Only when the load state reaches and remains on the yield
surface does elasto-plastic deformation occur, with incremental
plastic displacements determined by a ¯ow rule (associated or
otherwise) and a hardening law (typically linking the size of the
yield surface to the current value of plastic vertical displace-
ment). This idea of applying `macroscopic plasticity' to shallow
foundations under combined loading was originally suggested
by Roscoe & Scho®eld (1957), although the more recent renew-
al of interest can largely be attributed to the papers by
Butter®eld (1980, 1981). There have since been many experi-
mental, theoretical and numerical investigations of the shape of
the (V , M, H) yield surface for various footing geometries and
soil types, but relatively few authors have gone on to propose
complete incremental plasticity models capable of predicting the
load±displacement behaviour beyond initial yielding. Of those
who have, Schotman (1989) stressed that the parameters used in
his models for spudcans on sand and clay were only prelimin-
ary, and had not been calibrated or validated experimentally.
Bransby & Martin (1999) added linear elasticity and a simple
hardening law to Murff's (1994) rigid/perfectly plastic model
for suction caissons on clay, but again this was primarily for
demonstration purposes and there was no detailed experimental
veri®cation. The conical footing model developed by Tan
(1990) and the strip footing models proposed by Nova &
Montrasio (1991) and Gottardi & Butter®eld (1995) did include
certain features derived from laboratory or centrifuge tests, but
the testing in each case involved surface foundations on sand.
The authors are not aware of any published work of a similar
nature involving foundations on cohesive soil.
This paper describes Model B, a plasticity-based method for
predicting the incremental load±displacement response of a
spudcan footing on clay under combined (V , M, H) loading (its
predecessor, Model A, was a simpler, hypothetical model).
Some components of Model B are based on theoretical solu-
tions, while others are derived from the experimental calibration
tests reported by Martin & Houlsby (2000). The model is
primarily intended for application to monotonic, undrained
loading, for example when a jack-up unit is struck by a single
extreme wave. Model B footings can easily be incorporated into
the structural analysis of a complete jack-up when simulating
loading events of this type (Thompson, 1996; Martin & Houls-
by, 1999). Here, however, attention is con®ned to the develop-
ment of the numerical model and an assessment of its
predictive capabilities.
FEATURES OF MODEL B
Preliminary remarks
Although Model B is intended for use with spudcan-shaped
footings, two of its components (elasticity and hardening law)
are based on solutions for an embedded conical footing. It is
therefore convenient to treat the penetrated portion of a spud-
can, or more precisely the penetrated portion of its underside,
as an `equivalent conical footing' (see Fig. 2). This footing has
the same contact radius as its parent, r or R as appropriate, and
an apex angle âequiv chosen to give equality of the two
embedded volumes. Note that both the geometry of the equiva-
lent conical footing and its embedment, D, are determined from
the plastic component of vertical displacement, wp , not the total
vertical displacement, wp ‡ we . As indicated in Fig. 2, these
calculations of geometry and embedment take no account of the
soil displaced during spudcan penetration. This assumption is
regarded as conservative. In this section the features of Model
B are ®rst described with reference to a fully penetrated
spudcan (Fig. 2(b)). Because the loads (V , M, H) and displace-
ments (w, è, u) are always referred to the footing origin O,
some minor modi®cations are needed to cope with the partially
penetrated case (Fig. 2(a)). Details of these modi®cations are
given at the end of the section. Finally it should be pointed out
that the version of Model B used by Martin & Houlsby (1999)
featured a simpli®ed three-parameter yield function, rather than
the full six-parameter version that is incorporated here.
Elastic response
Jack-up unit foundations frequently undergo very deep pene-
trations on soft clay soils (Endley et al., 1981) and in these
circumstances it is no longer appropriate to use the familiar
closed-form elastic solutions for (V , M, H) loading of a rigid
circular footing at the soil surface. The de®nition of elasticity
in Model B is based on the work of Bell (1991), who used
three-dimensional ®nite element analysis to study the behaviour
of a rough, rigid circular footing subjected to combined
(V , M, H) loading. He considered footing embedments, D, of
up to two diameters and a range of Poisson's ratios for the soil,
V, w
M, θ
H, u
O
h O′
= –wp
2r
(a)
V, w
M, θ
wp
H, u
O
2R
(b)
Fig. 2. Partially and fully penetrated spudcan footings (equivalent
conical footings shown dotted): (a) partial penetration (wp , 0;
D 0 for equivalent conical footing); (b) full penetration (wp > 0;
D wp for equivalent conical footing)
 COMBINED LOADING OF SPUDCAN FOUNDATIONS ON CLAY: NUMERICAL MODELLING 689
although only the undrained situation (í ˆ 0:5) is relevant here. spudcan on samples of soft Speswhite kaolin. A full account of
Bell's numerical investigations highlighted the existence of an this work is given by Martin & Houlsby (2000), who proposed
elastic cross-coupling effect between the rotational and horizon- the following empirical equation for the yield surface:
tal degrees of freedom, so an appropriate expression for the  2  2     2â1  
incremental elastic response in Model B is M H M H V V 2â2
‡ ÿ2e ÿ â2 1ÿ
8 9 2 38 9 M0 H0 M0 H0 V0 V0
< äV = K 1 GR 0 0 < äwe =
äM ˆ 4 0 K 2 GR3 K 4 GR2 5 äèe (1) ˆ 0 (2)
: ; : ;
äH 0 K 4 GR2 K 3 GR äue
where
where K1 to K4 are dimensionless stiffness factors and G is a M 0 ˆ m0 :2RV0 , H 0 ˆ h0 : V0 ,
representative shear modulus of the clay. Values of K 1 to K4   
for various embedment ratios D=2R are listed in Table 1. Note V V ( â1 ‡ â2 )â1 ‡â2
that these factors are applicable to the situation shown in Fig. e ˆ e1 ‡ e2 ÿ1 , âˆ
V0 V0 ( â 1 )â 1 ( â 2 )â 2
2(b), with the hole above the embedded footing assumed to
remain open. In practice there is usually partial or complete and V0 is the bearing capacity under pure vertical load. The six
in®lling of the hole created by spudcan penetration (Endley parameter values obtained from regression analysis of the
et al., 1981), but the elastic footing response is still likely to be experimental data were m0 ˆ 0:083, h0 ˆ 0:127, e1 ˆ 0:518,
dominated by the soil that remains relatively undisturbed. e2 ˆ 1:180, â1 ˆ 0:764 and â2 ˆ 0:882. The ®tted yield surface
Bell (1991, private communication) extended his original is illustrated in Fig. 3.
®nite element work by investigating the elastic response of For various reasons it proves advantageous to manipulate
rough, rigid conical footings located at the soil surface (D ˆ 0). equation (2) and de®ne the Model B yield function as
Table 2 gives dimensionless stiffness factors K 1 to K 4 for "   
several cone apex angles, â, over the range relevant to spudcan    #1=2â2
M 2 H 2 M H
footings. Although factors for embedded cones were not ob- f (V , M, H, wp ) ˆ ‡ ÿ2e
tained directly by Bell (1991, private communication), they may M0 H0 M0 H0
be estimated by assuming that conical footings are subject to  â1 =â2  
the same changes ÄK i ˆ Ki (embedded) ÿ Ki (surface) as the V V
ÿ â 1=â2 1ÿ (3)
¯at circular footing of Table 1. More recent research by Ngo- V0 V0
Tran (1996) has con®rmed that this is an acceptable approxima-
tion.
In Model B the shear modulus, G, is taken to be a constant Note that this gives a yield surface ( f ˆ 0) that is identical to
multiple of the undrained shear strength su0 at the current that of equation (2). The plastic component of vertical displace-
(plastic) footing embedment, D. In cases where the undrained ment, wp , is listed as an argument of the function because, as
strength varies with depth it is recognised that there is some described below, it directly determines the reference vertical
minor inconsistency in using Bell's elastic stiffness factors, all bearing capacity, V0 .
of which pertain to a homogeneous soil with uniform G. It should be reiterated that this yield function is based on a
detailed programme of physical modelling performed at unit
gravity. An important difference in the full-scale situation is that
the overburden pressure, ã9D, (taking into account buoyancy
Yield function effects) makes a signi®cant contribution to the vertical bearing
The combined load yield surface used in Model B is derived capacity of an embedded footing. In these circumstances it
from a series of laboratory tests conducted with a small model would seem appropriate to divide V0 into components due to

Table 1. Elastic stiffness factors for a rough, rigid, ¯at circular footing embedded in
incompressible soil (í 0:5) (after Bell, 1991)
Cone apex angle, Embedment, Dimensionless elastic stiffness factors{
â D=2R
K1 K2 K3 K4
1808 0´0 8´00 5´33 5´33 0´00
1808 0´25 8´58 6´30 7´00 ÿ0´32
1808 0´5 9´21 7´03 7´63 ÿ0´52
1808 1´0 9´98 7´50 8´03 ÿ0´69
1808 2´0 11´18 8´14 8´46 ÿ0´93
{ Note that Bell (1991, 1991, private communication) adopted a (V , H, M) load ordering, so the
K 2 values appearing here correspond to his original K 3 values, and vice versa.

Table 2. Elastic stiffness factors for a rough, rigid, conical footing on incompressible soil
(í 0´5) (after Bell, 1991, private communication)
Cone apex angle, Embedment, Dimensionless elastic stiffness factors{
â D=2R
K1 K2 K3 K4
1808 0´0 8´00 5´33 5´33 0´00
1508 0´0 8´00 5´35 5´74 ÿ0´06
1208 0´0 8´05 5´71 6´45 ÿ0´52
908 0´0 8´22 7´47 7´68 ÿ1´84
{ Note that Bell (1991, 1991, private communication) adopted a (V , H, M) load ordering, so the
K 2 values appearing here correspond to his original K 3 values, and vice versa.
690 MARTIN AND HOULSBY
Associated flow
vectors, equation (5),
shown dotted for
comparison
H, δup
O V0
Fig. 3. Model B yield surface and ¯ow rule
cohesion and due to overburden, and make H 0 proportional to
just the cohesive component, thus:
H 0 ˆ h0 :(V0 ÿ ã9D:ðR2 ) (4)
There should, however, be no need to modify the de®nition of
M 0 when applying Model B to real spudcan footings.
Flow rule
At any point on the (V , M, H) yield surface, an outward-
pointing normal may be found by evaluating the gradient vector,
, f . With an associated ¯ow rule the vector of incremental
plastic displacements would be parallel to this vector,
8 9 8 9
< äwp = < @ f =@V =
äèp ˆ Ë @ f =@ M (5)
: ; : ;
äup @ f =@ H
with Ë being a non-negative scalar determining the magnitude
of the plastic increment. The validity of assuming such a ¯ow
rule for spudcan footings on clay was investigated in a second
set of laboratory tests reported by Martin & Houlsby (2000).
The observed rotational and horizontal displacements con®rmed
the hypothesis of normality to the yield surface in the (M, H)
plane, but the vertical displacements, äwp , were found to be
consistently smaller than those predicted by an associated ¯ow
rule. This trend can be captured by including an empirical
`association parameter' æ in the ¯ow rule for Model B:
8 9 8 9
< äwp = < æ:@ f =@V =
äèp ˆ Ë @ f =@ M (6)
: ; : ;
äup @ f =@ H
From regression analysis of their experimental results, Martin &
Houlsby (2000) obtained a best-®t value of 0´645 for æ. There
was, however, considerable scatter in the data, and only two
vertical load levels (V =V0 ˆ 0:75 and V =V0 ˆ 0:2) were em-
ployed in the testing programme. Until further data are available
for other soils and footing geometries it is not clear how much
this parameter may vary, but it would be expected to be less
than or equal to unity. Fig. 3 shows the pattern of incremental
plastic displacement vectors obtained from equation (6). The
selective adjustment of the vertical component, äwp , is closely
analogous to Burd's (1986) treatment of plastic volumetric
strains when formulating a constitutive law for sand.
H, δup
M/2R,
V, δwp 2Rδθup
Hardening law
The reference vertical bearing capacity, V0 , is calculated
from the standard expression
V0 ˆ (Nc su0 ‡ ã9D):ðR2 (7)
where Nc is a dimensionless bearing capacity factor and su0 is
the undrained shear strength of the clay at depth D (ˆ wp for a
fully penetrated spudcan). An additional contribution from skin
friction is included if the spudcan has a vertical perimeter wall
of signi®cant height (Martin & Houlsby, 1999). Model B uses a
set of theoretical Nc factors that are particularly suitable for
estimating spudcan bearing capacity in offshore clay deposits.
These factors are based on the work of Houlsby & Wroth
(1983), who used the method of characteristics to obtain lower
bound collapse loads for axisymmetric footings on clay with a
linearly increasing undrained strength pro®le. Houlsby & Martin
(1992) extended this study to examine the effect of the footing
embedment ratio, D=2R, in combination with Houlsby &
Wroth's original parameters of cone apex angle, â, interface
friction coef®cient, á, and dimensionless undrained strength
gradient, r2R=sum (here sum denotes the mudline strength and r
the rate of increase of strength with depth). The parameter
ranges considered were as follows: ⠈ 308 to 1808, D=2R ˆ 0
to 2:5, á ˆ 0 to 1, r2R=sum ˆ 0 to 5. The clay was idealised as
a weightless, rigid-plastic Tresca material, and the space above
the footing was assumed to be occupied by a smooth-sided
shaft. Results of the numerical study, in the form of tables of
Nc factors, can be found in Martin (1994) and in SNAME
(1997). For the special case of a ¯at circular footing embedded
in uniform soil, Martin (1994) gives a detailed comparison bet-
ween the theoretical bearing capacity factors and those cal-
culated from various semi-empirical formulae (Skempton, 1951;
Brinch Hansen, 1970; Vesic, 1975).
Figure 4 shows a typical ®eld of stress characteristics for an
embedded conical footing. It can be shown that stress ®elds of
this type are extensible throughout the semi-in®nite soil mass,
and hence the bearing capacity factors used in Model B
represent rigorous lower bound collapse loads. The validity of
introducing an auxiliary smooth-sided shaft does, however,
remain open to question. Lower bound calculations of the type
shown in Fig. 4 are felt to be appropriate for the relatively
shallow embedments (0 < D=2R < 2:5) considered in the para-
metric study, but extrapolation of the tabulated N c factors to
greater depths is not recommended. This is because deep failure
is more likely to involve localised ¯ow around the spudcan, and
 COMBINED LOADING OF SPUDCAN FOUNDATIONS ON CLAY: NUMERICAL MODELLING 691

–1·5 –1·25 –1·0 –0·75 –0·5 –0·25 0·25 0·5 0·75 1·0 1·25 1·5
x

0·25

0·5

0·75

1·0

Fig. 4. Typical ®eld of stress characteristics for embedded conical footing ( â 1508, D=2R ˆ 0:5, á 0:8, r2R=sum 5). Bearing capacity
factor, Nc 7:93

the assumption of rigid-plastic soil behaviour is also likely to äó ˆ Ke äå or äó ˆ Kep äå (11)

become increasingly inadequate (Martin, 2001).
as appropriate. The elastic stiffness matrix, Ke , can be obtained
directly from equation (1) for a fully penetrated spudcan or
Handling of partial penetration equation (8) during partial penetration. Construction of the
Consider now the partially penetrated spudcan of Fig. 2(a). elasto-plastic stiffness matrix, Kep , is more involved. The load
The equivalent conical footing has a radius r, and its origin, O9, derivatives of the yield function f are required:
lies a distance h ˆ ÿwp below the overall footing origin, O. 8 9
Assuming that the rotational and horizontal displacements are @f < @ f =@V =
small, it can be shown that the incremental elastic response at ˆ @ f =@ M (12)
@ó : ;
O is given by a modi®ed version of equation (1): @ f =@ H
8 9 2 38 9
< äV = K91 Gr 0 0 < äwe = A notional plastic potential function, g, is also introduced, even
äM ˆ 4 0 K92 Gr 3 K94 Gr 2 5 äèe (8) though it is only de®ned in terms of its derivatives (cf. equation
: ; : ;
äH 0 K94 Gr 2 K93 Gr äue (6)):
8 9 8 9
@ g=@V = < æ:@ f =@V =
where the effective stiffness factors, K9i , are related to the basic @g <
stiffness factors for D ˆ 0 (given in Table 2) by the transforma- ˆ @ g=@ M ˆ @ f =@ M (13)
@ó : ; : ;
tions @ g=@ H @ f =@ H
K91 ˆ K1 , K92 ˆ K2 ‡ (h=r)2 K3 ÿ 2(h=r)K4 , K93 ˆ K3
A feature of the present problem is that the elastic stiffness
and K94 ˆ K 4 ÿ (h=r)K3 matrix, Ke , is not constant, but depends on the current value of
plastic vertical penetration, wp . This phenomenon is called
The yield function, equation (3), must also be modi®ed during elastic-plastic coupling, and requires careful treatment of the
partial penetration because the moment acting on the equivalent components of the displacements. In particular, it is essential
conical footing is now M ‡ hH (rather than M) and the that the elasto-plastic stiffness matrix, Kep , accounts for the
reference moment capacity, M 0 , is now m0 :2rV0 (rather than `coupled' elastic displacements that arise from the change of Ke
m0 :2RV0 ). These are the only changes required in equation (3), during an elasto-plastic increment (Maier & Hueckel, 1979).
so the revised version will not be given in full. Note that as The appropriate expression for Kep, derived in full by Martin &
long as the ¯ow rule of equation (6) is applied to the modi®ed Houlsby (1999), is
yield function, incremental plastic displacements at O will be !
calculated correctly (Dean et al., 1992). Finally a revised @ g @ g dKÿ1 e @f T
Ke ‡ ó Ke
version of the hardening law, equation (7), is needed during @ó @V dwp @ó
partial penetration. The required expression is Kep ˆ Ke ÿ ! (14)
@f @g @f T @ g @ g dKÿ1 e
V0 ˆ Nc sum :ðr 2 (9) ÿ ‡ Ke ‡ ó
@wp @V @ó @ó @V dwp
where the Nc factor relates to a conical footing with embed-
ment D ˆ 0 and a dimensionless undrained strength gradient This formulation of the tangent elasto-plastic stiffness is suitable
r2r=sum (rather than r2R=sum ). for use in an explicit (forward Euler) integration scheme, as
discussed below. Note that in general the matrix Kep is fully
populated and unsymmetric.
INCREMENTAL LOAD±DISPLACEMENT FORMULATION
It is convenient to introduce the notation
8 9 8 9 Computational aspects
<V= <w= Some features of an in-house FORTRAN implementation of
óˆ M and å ˆ è (10) Model B will now be discussed. Input to the model can take
: ; : ;
H u the form of a load increment, äó, a displacement increment, äå,
or a mixed increment (e.g. one involving combined rotation äè
The incremental load±displacement response during elastic or and horizontal displacement äu at constant vertical load
elasto-plastic deformation is de®ned by the tangent stiffness äV ˆ 0). Regardless of the chosen control method, equation
relationship (11) can readily be solved for the three conjugate unknowns
692 MARTIN AND HOULSBY
once the appropriate stiffness matrix Ke or Kep has been
formed. If the current state of the system is elastic ( f , 0) it is
assumed that the next increment will also be elastic, so a trial
solution is made using the matrix Ke . If the trial load state
proves to be inadmissible ( f . 0) a simple bisection algorithm
is used to isolate the elastic portion of the increment, and the
remainder is applied elasto-plastically using the matrix Kep .
Likewise if the current state of the system is elasto-plastic
( f ˆ 0) it is assumed that the next increment will also be
elasto-plastic, and a trial solution is made using the matrix Kep .
The scalar plastic multiplier Ë is then evaluated (see Martin &
Houlsby, 1999), and its sign is examined. If the trial solution
indicates attempted unloading from the yield surface (Ë , 0),
the increment is reanalysed elastically using the matrix Ke .
During elasto-plastic deformation, all terms in equation (14)
are evaluated with reference to the load state, ó, and the plastic
vertical displacement, wp , at the start of a new increment.
While closed form expressions for @ f =@ó and @ g=@ó can
easily be worked out from the de®nitions above, the derivatives
of f and the elastic ¯exibility matrix Kÿ1 e with respect to wp
must be evaluated numerically. This is because: (a) calculations
of f and Kÿ1e both involve interpolation from tables; (b) during
partial penetration the geometry of the equivalent conical foot-
ing does not evolve in a straightforward manner. Drift of the
solution during elasto-plastic deformation is controlled by cor-
recting the ®nal loads, ó, back onto the yield surface (in a
direction normal to the surface) after each increment. Any
residual differences between the corrected and externally ap-
plied loads are then removed using the Newton±Raphson tech-
nique, with an updated elasto-plastic matrix Kep for every
iteration. As with all incremental schemes there is a trade-off
between solution accuracy and the number of increments em-
ployed. The solutions presented all involve suf®cient increments
to obtain a converged solution. For a given sequence of loading
steps, analyses with different numbers of increments were used
until there was no discernible change in the ®nal results.
It should be pointed out that certain details of the incremen-
tal formulation and solution procedure described above differ
from those used in the original implementation of Model B
(Martin, 1994). The changes are purely for reasons of improved
ef®ciency and compactness of expression, and have no effect on
the numerical predictions obtained.
MODEL B SIMULATIONS OF SELECTED TESTS
A total of 24 model spudcan tests were used to de®ne the
yield surface and ¯ow rule of Model B. As a ®rst step towards
evaluating the performance of the model, retrospective simula-
tions have been performed for ®ve of these tests (the same ones
described in detail by Martin & Houlsby, 2000).
The spudcan tests were performed on heavily overconsoli-
dated samples of kaolin (ó v9max ˆ 200 kPa followed by full
unloading to ó v9 ˆ ã9D, where ã 9  6:85 kN=m3 ). The follow-
ing power-law expression (Martin & Houlsby, 2000) was found
to give a good description of the variation of su with depth:
su =ó v9 ˆ 0:25 OCR0:75 (15)
This equation is plotted in Fig. 5, together with undrained
strength measurements from the ®ve samples of interest (each
data point represents the average of four miniature vane tests).
Because these individual strengths all fall close to the `average'
strength pro®le, equation (15) was used to de®ne the increase of
su with depth in all of the Model B analyses reported below. At
any given footing embedment, values of sum , r and su0 appro-
priate to the local variation of undrained strength were chosen
by ®tting a least-squares straight line to the su pro®le between
D and D ‡ R, or between the surface and a depth r in the case
of a partially penetrated spudcan (cf. Young et al., 1984). This
was necessary because the bearing capacity factors used to
calculate V0 have only been tabulated for linearly increasing
strength pro®les. Full soil±footing adhesion (á ˆ 1) was as-
sumed when determining the Nc factors, and the rigidity index,
G=su , was arbitrarily taken to be 100. The geometry of the
Vane strength, su: kPa
0 5 10 15 20
0
100
Depth, D: mm
200
Test AH1
Test AH3
300
Test AL2
Test BH1
Test BL2
Equation (15) ± 10%
400
Fig. 5. Measured values of undrained shear strength in selected
model spudcan tests
footing used in the Model B simulations was identical to that of
the spudcan used in the tests, with a shape as shown in Fig. 1
and a diameter 2R ˆ 125 mm.
Each numerical simulation commenced from a plastic pene-
tration wp,init chosen such that the tip of the spudcan was just
embedded in the soil, thus allowing an initial yield surface to
be established. After initialising all loads and displacements to
zero (with the exception of w ˆ wp,init ), each Model B analysis
proceeded using exactly the same sequence of control steps as
the test being simulated. The basic pattern was the same in all
tests, with displacement-controlled vertical penetration being
interrupted for a series of `swipe' or constant-V events at
nominal embedments of 0, 20, 50, 100 and 200 mm, as well as
vertical unload±reload events at depths of 35, 75 and 150 mm.
The swipe events were fully displacement-controlled, with rota-
tion or horizontal displacement (or a combination of the two)
being applied at constant embedment w, commencing from
either maximum vertical load (V =V0 ˆ 1) or very low vertical
load (V =V0 ˆ 0:01). Constant-V events involved the application
of rotation and/or horizontal displacement at constant vertical
load, and were performed after unloading to an intermediate
level of V =V0 (0:75 or 0:2). Because the experimental proce-
dures and results have already been described by Martin &
Houlsby (2000), the focus here is on the comparison between
observed and predicted behaviour in the selected tests.
Vertical load±penetration
Figure 6 illustrates the performance of Model B in predicting
the vertical load±displacement response from test AH1. This
test contained swipe events involving pure rotation (AB to IJ),
and unload±reload events involving full unloading to V ˆ 0
(UR1 to UR3). Fig. 6(a) shows how each of these events was
preceded by a short period of load-controlled penetration at
constant V, designed to minimise the effects of soil creep. Note,
however, that Model B has no facility for modelling time-
dependent behaviour, so these load holding stages were ignored
in the simulation and treated as straightforward vertical displa-
cement increments.
 COMBINED LOADING OF SPUDCAN FOUNDATIONS ON CLAY: NUMERICAL MODELLING 693
Vertical load, V: N Vertical load, V: N
0 400 800 1200 1600 0 400 800 1200 1600
–50

Start of test Start of test

B B
A
0 A

D D
C C
UR1 UR1

F F
50 E E
Vertical penetration, w: mm

UR2 UR2

H H
100 G G

150 UR3 UR3

200 J J
I I
Retraction Retraction

250
(a) (b)

Fig. 6. Test AH1: vertical load±penetration plot: (a) measured; (b) calculated

Comparing Figs 6(a) and 6(b), there is good agreement
during the initial period of penetration (w , 0) as full contact is
established between the soil and the underside of the spudcan.
The theoretical bearing capacity factors then give an excellent
prediction of the observed increase in V with depth. The use of
a rigidity index G=su ˆ 100 leads to elastic unload±reload
stiffnesses that agree fairly well with the chord slopes of the
experimental hysteresis loops, but this is largely fortuitous.
Unload±reload loops of a smaller amplitude, say V =V0 ˆ 0:5,
would have shown a much stiffer response (Martin, 1994). The
most serious difference between the two plots in Fig. 6 is the
manner in which the virgin load±penetration curve is rejoined
after a rotation or unload±reload event. Rather than a sudden
transition from elastic to elasto-plastic behaviour, gradual re-
yielding (particularly after the rotation events) is a feature of
the experimental curve.
Another obvious shortcoming of the Model B prediction is
the restriction to vertical loads V > 0. In fact the ¯ow of soil
around the penetrating spudcan created a seal that led to the
development of signi®cant tensile loads during the ®nal retrac-
tion. Note, however, that because of the 1 g nature of the
laboratory tests, in®lling above the model spudcan did not have
a signi®cant in¯uence on the vertical loads measured during the
main body of the test (Martin & Houlsby, 2000). For this reason
the predicted values of V in Fig. 6(b)Ðand in the other Model
B analyses reported belowÐhave not been adjusted to account
for the weight of any in®ll soil.
Swipe events
Figure 7 shows the observed and predicted (V , M) and
(V , H) load paths from the swipe events in test AH3. These
events commenced from V =V0 ˆ 1 and involved combined
rotation and horizontal displacement. For clarity only the `out-
ward' or monotonic part of each swipe event is shown, though
it should be noted that the entire test was simulated in a single
Model B analysis. In general there is good agreement between
the predicted load paths and the experimental results. Model B
correctly predicts that the early stages of each event produce
the greatest changes in load, as shown by the reduction in data
marker spacing along each load path. With continued displace-
ment, however, the numerical simulation gives much more
sharply de®ned `critical states' in which all three loads
(V , M, H) approach constant values. Unfortunately the latter
stages of the spudcan test at wnom ˆ 200 mm were affected by
spurious load cell readings, indicated by dashed lines in Figs
7(a) and 7(b). Martin & Houlsby (2000) give a more detailed
discussion of this problem, which arises when the waterproof
shaft surround attached to the spudcan makes contact with the
side of the hole formed by spudcan penetration.
Figure 8 shows the (V , M) and (V , H) load paths from test
AL2, in which the swipe events involved unloading to
V =V0 ˆ 0:01 followed by horizontal displacement at constant w.
In the Model B simulation there is a short period of elastic
response at the start of each event, with the vertical load
remaining constant until the yield surface is reached. Subse-
quently the increase in V during each swipe event is substan-
tially overpredicted, possibly suggesting that there is a need to
re®ne the Model B ¯ow rule at very low vertical loads. While
the numerical simulation of test AL2 captures the general
character of the observed results, including the increase in
moment and horizontal load capacities with additional penetra-
tion, the prediction is not as accurate as that of test AH3.
The measured and predicted bearing capacities V0 listed in
Figs 7 and 8 generally agree well, with the exception of the
events at wnom ˆ 0 mm. At this level the soil±footing contact
area changes very rapidly, and both experimental and numerical
results are very sensitive to the exact depth at which penetration
is interrupted. It will also be recalled that the baseline for
vertical displacements is the undisturbed soil surface level, so
694 MARTIN AND HOULSBY

Data markers denote 2·5 mm intervals of [(2Rθ)2 + u2]½ Data markers denote 2·5 mm intervals of [(2Rθ)2 + u2]½

wnom = 0 mm (V0 = 595 N) wnom = 0 mm (V0 = 428 N)

wnom = 20 mm (V0 = 814 N) wnom = 20 mm (V0 = 782 N)
wnom = 50 mm (V0 = 989 N) wnom = 50 mm (V0 = 951 N)
wnom = 100 mm (V0 = 1281 N) wnom = 100 mm (V0 = 1181 N)
wnom = 200 mm (V0 = 1557 N) wnom = 200 mm (V0 = 1534 N)

200

100
Moment load, M/2R: N

–100

–200
(a) (c)
200

100
Horizontal load, H: N

–100

–200
0 400 800 1200 1600 0 400 800 1200 1600
Vertical load, V: N Vertical load, V: N
(b) (d)

Fig. 7. Test AH3: load paths observed in swipe events (combined rotation and horizontal displacement, initial V =V0 1): (a), (b)
measured; (c), (d) calculated

when the real spudcan reaches a penetration of (say)
w ˆ ÿ2 mm it may already have made full contact with the
clay. By contrast Model B does not account for the soil
displaced during earlier penetration, so full contact is achieved
only when wp ˆ 0 mm. The effects of this discrepancy can also
be seen in the early stages of penetration in Figs 6(a) and 6(b).
Constant vertical load events
Figure 9 shows the key experimental data from test BH1
(footing rotations with V held constant at 0:75V0 ) together with
corresponding excerpts from the Model B simulation of the test.
Comparison between Figs 9(a) and 9(d) shows that a rigidity
index G=su ˆ 100 is much too low for modelling the very high
initial rotational stiffness observed in the experiments. A better
prediction was obtained with G=su  400, but the quality of the
test data did not warrant a detailed investigation of spudcan
behaviour at very small rotations, where apparatus ¯exibility
was a source of considerable uncertainty. In the experimental
moment±rotation curves it is dif®cult to distinguish post-yield
hardening from the gradual process of yielding itself, but in
most events there does appear to be a signi®cant `plastic
rotational stiffness', which is underestimated by Model B. This
implies that a hardening law based purely on the plastic vertical
penetration, wp , may be overly simplistic. In Fig. 9(e) the
predicted responses show the effect of using a negative cross-
coupling stiffness factor K 4 in equation (1). Within the elastic
domain, a positive applied rotation results in a positive moment
but a negative horizontal load. While there is no experimental
evidence of this phenomenon in Fig. 9(b), this does not mean
that the negative values of K 4 obtained by Bell (1991; 1991,
private communication) should be rejected. The most likely
explanation for the observed behaviour is that signi®cant plastic
deformations are occurring right from the start of each rotation
event. There are similar differences between Figs 9(c) and 9(f),
with vertical settlement beginning immediately in each experi-
mental event but only after yielding in the Model B predictions.
Improvements allowing Model B to capture this gradual onset
of full plasticity (discussed in more detail below) would
certainly be more worthwhile than empirical adjustment of the
elastic stiffness factors. The calculated settlements in Fig. 9(f)
are signi®cantly smaller in the ®rst rotation (wnom ˆ 0 mm) than
in subsequent events, a consequence of the steep slope, dV =dw,
of the virgin load±penetration curve as the footing makes full
contact with the soil (see Fig. 6).
Largely analogous comments can be made with respect to
 COMBINED LOADING OF SPUDCAN FOUNDATIONS ON CLAY: NUMERICAL MODELLING
Data markers denote 2·5 mm intervals of u
wnom = 0 mm (V0 = 632 N)
wnom = 20 mm (V0 = 845 N)
wnom = 50 mm (V0 = 1038 N)
wnom = 100 mm (V0 = 1302 N)
wnom = 200 mm (V0 = 1470 N)
200
100
Moment load, M/2R: N
–100
–200
(a)
200
100
Horizontal load, H: N
–100
–200
0 400 800 1200
Vertical load, V: N
(b)
Fig. 8. Test AL2: load paths observed in swipe events (horizontal displacement, initial V =V0
calculated
Fig. 10, which shows measured and predicted behaviour for the
constant-V events in test BL2 (horizontal footing displacement
with V held constant at 0:2V0 ). Fig. 10(d) resembles Fig. 9(e),
with the negative stiffness factor K4 predicting that a positive
horizontal displacement should result in a negative footing
moment prior to yielding. This behaviour is certainly not
manifested in the experimental curves of Fig. 10(a), and clearly
there is a need for further experimental study of the footing
response at small displacements (using a stiffer testing rig and
more sensitive instrumentation) in conjunction with appropriate
re®nement of this aspect of the numerical model. At larger
(post-yield) horizontal displacements the Model B predictions in
Figs 10(d) and 10(e) generally show good agreement with the
test results. The calculated vertical displacements in Fig. 10(f)
agree quite well with the observed ones until about u ˆ 2 mm,
but thereafter too much heave is predicted. This re¯ects the fact
that the empirical association parameter, æ, in equation (6) was
determined on the basis of incremental plastic displacements at
®rst yield. A more complicated ¯ow rule would be needed to
improve this aspect of the model. Although the Model B
prediction of plastic vertical heave is associated with post-yield
softening of the load±displacement responses in Figs 10(d) and
10(e), this is only discernible in the event at wnom ˆ 0 mm.
695
Data markers denote 2·5 mm intervals of u
wnom = 0 mm (V0 = 531 N)
wnom = 20 mm (V0 = 786 N)
wnom = 50 mm (V0 = 964 N)
wnom = 100 mm (V0 = 1188 N)
wnom = 200 mm (V0 = 1526 N)
(c)
1600 0 400 800 1200 1600
Vertical load, V: N
(d)
0:01): (a), (b) measured; (c), (d)
LIMITATIONS OF MODEL B
Applicability
A number of speci®c shortcomings of Model B have already
been highlighted. One concern of a general nature is that the
component features of the model (whether based on theory or
on small-scale experiment) all pertain to conditions where there
is negligible in®lling behind the penetrating spudcan. By con-
trast, at a typical offshore soft clay site, deep spudcan installa-
tion is usually associated with a substantial amount of in®lling
(Endley et al., 1981; Le Tirant & PeÂrol, 1993). While it is easy
to account for the additional vertical load acting on the footing
(Martin & Houlsby, 1999), in®lling may have other effects on
spudcan behaviour such as increased moment and horizontal
load capacities, increased elastic stiffnesses and increased capa-
city under tensile vertical loading. A further consequence of
in®lling is that signi®cant passive resistance can be mobilised
by the embedded portion of a jack-up leg under combined
(V , M, H) loading (Springman & Scho®eld, 1998). Under these
circumstances it may no longer be suf®cient to consider the
behaviour of the spudcan footing alone. In passing it should be
noted that the empirical yield surface and ¯ow rule used in
Model B are speci®c to the representative spudcan shape (Noble
Denton & Associates, 1987) used in the experiments. This
696 MARTIN AND HOULSBY

wnom = 0 mm (V0 = 458 N) wnom = 0 mm (V0 = 420 N)

wnom = 20 mm (V0 = 746 N) wnom = 20 mm (V0 = 770 N)
wnom = 50 mm (V0 = 981 N) wnom = 50 mm (V0 = 952 N)
wnom = 100 mm (V0 = 1310 N) wnom = 100 mm (V0 = 1181 N)
wnom = 200 mm (V0 = 1757 N) wnom = 200 mm (V0 = 1525 N)

200

100
Moment load, M/2R: N

–100

–200
(a) (d)
200

100
Horizontal load, H: N

–100

–200
(b) (e)
–0·5
Vertical displacement, ∆w : mm

Heave Heave
0
settlement settlement

0·5

1·0

1·5
–10 –5 0 5 10 –10 –5 0 5 10
Rotational displacement, 2Rθ: mm Rotational displacement, 2Rθ: mm
(c) (f)

Fig. 9. Test BH1: results of constant V events (rotation at V =V0 0:75): (a)±(c) measured; (d)±(f) calculated

evidently limits the applicability of the model to footings having
a similar geometry.
Non-monotonic loading
The comparisons in the previous section concentrated on the
performance of Model B in simulating monotonic loading
events. The only instances of cyclic behaviour considered were
the unload±reload events in Fig. 6, where the calculated
response was linear and elastic but the experiments gave
hysteretic loops. It is not surprising that similar limitations
become apparent when Model B is used to simulate reversals of
moment or horizontal loading. Fig. 11, for example, shows a
complete rotation cycle from test BH1 (only the monotonic part
of this event appeared in Fig. 9). As discussed above, the Model
B analysis with G=su ˆ 100 underpredicts the initial moment±
 COMBINED LOADING OF SPUDCAN FOUNDATIONS ON CLAY: NUMERICAL MODELLING 697

wnom = 0 mm (V0 = 461 N) wnom = 0 mm (V0 = 497 N)

wnom = 20 mm (V0 = 651 N) wnom = 20 mm (V0 = 780 N)
wnom = 50 mm (V0 = 861 N) wnom = 50 mm (V0 = 963 N)
wnom = 100 mm (V0 = 1190 N) wnom = 100 mm (V0 = 1191 N)
wnom = 200 mm (V0 = 1414 N) wnom = 200 mm (V0 = 1533 N)

200

100
Moment load, M/2R: N

–100

–200
(a) (d)
200

100
Horizontal load, H: N

–100

–200
(b) (e)
–1·5
Vertical displacement, ∆w : mm

–1·0

–0·5

Heave Heave
0
settlement settlement

0·5
–10 –5 0 5 10 –10 –5 0 5 10
Horizontal displacement, u: mm Horizontal displacement, u: mm
(c) (f)

Fig. 10. Test BL2: results of constant V events (horizontal displacement at V =V0 0:2): (a)±(c) measured; (d)±(f) calculated

rotation stiffness during primary loading, but more seriously it
fails to capture the gradual process of yielding. Identical com-
ments apply to the reverse cycle, where the observed yielding
process is even more gradual. Model B does at least give a
good prediction of the ultimate moment loads attained during
the outward and inward rotations, but more serious discrepan-
cies between the experimental and calculated results would be
expected to develop if further cycles were performed.
Figure 12 illustrates another situation involving moment and
horizontal load reversals. The spudcan tests featuring `loop
events' of the type depicted here are not described by Martin &
Houlsby (2000) as they were not used to calibrate any features
of Model B. It is nevertheless interesting to compare the results
of a typical event, performed under constant vertical load at
V =V0 ˆ 0:4, with the corresponding numerical prediction. In
Fig. 12 the imposed (è, u) displacement path commences with
pure horizontal displacement in a positive direction. The elastic
stiffness factors used in Model B predict an initial load path
698 MARTIN AND HOULSBY
Test (V0 = 981 N) latter stages of the loop event as the footing returns to pre-
Model B (V0 = 952 N) viously disturbed soil, possibly suffering some loss of contact as
well. Martin (1994) gives further details of three complete
200 spudcan tests incorporating loop events of various types.
wnom = 50 mm Byrne & Houlsby (2000) discuss some recent theoretical
V/V0 = 0·75 developments that allow much improved simulations of grad-
100
ual yielding and load±displacement behaviour under cyclic
(V , M, H) loading, while still retaining the overall framework
Moment load, M/2R: N

of a macroscopic plasticity model for the foundation response.

These new continuous hyperplasticity methods appear to offer
0 many advantages over the ad hoc interpolations between elastic
and elasto-plastic stiffness adopted by authors such as Schotman
(1989) and Hambly et al. (1990). The new approach is based
on the concept of a continuous ®eld of yield surfaces, and
–100 therefore results in smooth transitions between elastic and
plastic behaviour.

–200
0 5 10 15 CONCLUSIONS
Rotational displacement, 2Rθ: mm A complete incremental plasticity model for describing the
Fig. 11. Measured and calculated results for typical rotation cycle at
behaviour of spudcan footings on clay under combined
constant V (V , M, H) loading has been described. Model B uses analyti-
cally based de®nitions of elasticity and the variation of vertical
bearing capacity with embedment, with an empirically deter-
mined yield surface and a ¯ow rule based on normality (but
Test (V0 = 1632 N) with one empirical adjustment). Several of the small-scale
Model B (V0 = 1527 N) spudcan tests used to calibrate the latter two features have been
simulated a posteriori with the numerical model. The theor-
250
etical hardening law gives good predictions of the virgin load±
wnom = 200 mm
V/V0 = 0·4
penetration behaviour, but vertical unload±reload loops cannot
be replicated in detail because of the assumption of linear
elasticity. Similarly, while the model is very successful in
125 capturing the general patterns of behaviour in combined loading
events, its `hard' yield surface does not allow realistic predic-
Horizontal load, H: N

tion of the gradual yielding observed experimentally. This is

perhaps the major weakness of Model B, and one that is also
responsible for its relatively poor performance in tests involving
0 cyclic reversals of moment and horizontal load.
The main reason for developing a compact plasticity-based
model of spudcan behaviour is to allow more rational analysis
of the performance of complete jack-up units on clay. Martin
–125
(1994) and Martin & Houlsby (1999) have incorporated Model
B footings into some simpli®ed structural analyses of three-
legged jack-up units subjected to monotonic, quasi-static lateral
loading. Thompson (1996) and Cassidy (1999) have performed
similar work, but with more detailed modelling of wave loading
–250 and the inclusion of structural dynamics. These analyses have
–250 –125 0 125 250
afforded many interesting insights into the complex processes of
Moment load, M/2R: N
soil±structure interaction (e.g. post-yield load redistribution be-
tween spudcans) that determine factors such as the critical stress
resultants in the structure and the (V , M, H) load paths
followed at individual footings. Model B footings offer an
additional bene®t over traditional pinned or linear spring ideali-
2·5 sations in that the ultimate lateral capacity of the soil±structure
system emerges directly from the requisite incremental loading
u: mm

0 2Rθ: mm analysis. In an effort to check the validity of such predictions,

5 10 experimental research involving pushover testing of a model
–2·5 three-legged jack-up unit on clay is currently in progress.
Displacement path

Fig. 12. Measured and calculated results for typical loop event at
ACKNOWLEDGEMENT
constant V
The ®rst author gratefully acknowledges the support of the
Rhodes Trust.
direction that is dominated by H but with a slight negative
component of M. With the switch to elasto-plastic deformation
NOTATION
the calculated (M, H) load path begins to track around the D depth below soil surface; footing embedment
yield surface (there is a very slight degree of softening, or yield e1 , e2 yield surface ®tting parameters
surface contraction, but this is not discernible in Fig. 12). Apart e (subscript) elastic component
from the expected `rounding off' of the transition from elastic ep (subscript) sum of elastic and plastic components
to elasto-plastic behaviour, the early part of the experimental f yield function
curve shows reasonable qualitative and quantitative agreement g plastic potential function
with the Model B prediction. This agreement deteriorates in the G shear modulus
 COMBINED LOADING OF SPUDCAN FOUNDATIONS ON CLAY: NUMERICAL MODELLING
h height of footing origin O above effective origin O9
(partial penetration)
h0 dimensionless maximum horizontal load parameter
H horizontal load
H0 maximum horizontal load capacity when M ˆ 0
K stiffness matrix
K 1 ±K 4 elastic stiffness factors (full penetration)
K91 ±K94 elastic stiffness factors (partial penetration)
m0 dimensionless maximum moment parameter
M moment load
M0 maximum moment capacity when H ˆ 0
nom (subscript) nominal
Nc bearing capacity factor with respect to su0
p (subscript) plastic component
r soil±footing contact radius (partial penetration)
R overall footing radius
su undrained shear strength
sum su at mudline level
su0 su at footing level
u horizontal displacement
V vertical load
V0 maximum vertical load capacity when H ˆ M ˆ 0
w vertical displacement
á soil±footing interface friction coef®cient
â conical footing apex angle
â1 , â2 yield surface ®tting parameters
ä (pre®x) incremental value
Ä (pre®x) change from initial or reference value
å vector of displacements
è rotational displacement
Ë scalar plastic multiplier
í Poisson's ratio
r rate of increase of su per unit depth
ó vector of loads
æ ¯ow rule association parameter
REFERENCES
Arnesen, K., Dahlberg, R., Kjeùy, H. & Carlsen, C. A. (1988). Soil±
structure interaction aspects for jack-up platforms. Proc. 5th Int.
Conf. Behaviour of Offshore Struct., Trondheim, 259±277.
Bell, R. W. (1991). The analysis of offshore foundations subjected to
combined loading. MSc thesis, University of Oxford.
Bell, R. W. (1991). Private communication.
Bransby, M. F. & Martin, C. M. (1999). Elasto-plastic modelling of
bucket foundations. Proc. 7th Int. Symp. Numer. Models Geomech.,
Graz, 425±430. Rotterdam: Balkema.
Brinch Hansen, J. (1970). A revised and extended formula for bearing
capacity. Bulletin No. 28, Danish Geotechnical Institute, Copenha-
gen, 5±11.
Burd, H. J. (1986). A large displacement ®nite element analysis of a
reinforced unpaved road. DPhil thesis, University of Oxford.
Butter®eld, R. (1980). A simple analysis of the load capacity of rigid
footings on granular materials. JourneÂe de GeÂotechnique, ENTPE,
Lyon, France, 128±137.
Butter®eld, R. (1981). Another look at gravity platform foundations.
Presented at CISM course Soil Mech. Found. Engng in Offshore
Technology, Udine.
Byrne, B. W. & Houlsby, G. T. (2000). Experimental investigations of
the cyclic response of suction caissons in sand. Proc. 32nd Offshore
Technology Conf., Houston, paper no. OTC 12194.
Carlsen, C. A., Kjeùy, H. & Eriksson, K. (1986). Structural behaviour of
harsh environment jack-ups. Proc. 1st Int. Conf. Jack-up Drilling
Platform, London, 90±136. London: Collins.
Cassidy, M. P. (1999). Non-linear analysis of jack-up structures sub-
jected to random waves. DPhil thesis, University of Oxford.
Dean, E. T. R., James, R. G., Scho®eld, A. N. & Tsukamoto, Y. (1992).
Combined vertical, horizontal and moment loading of circular spuds
on dense sand foundations, Report No. CUED/D-SOILS/TR244.
Cambridge University Engineering Department.
Endley, S. N., Rapoport, V., Thompson, P. J. & Baglioni, V. P. (1981).
699
Prediction of jack-up rig footing penetration. Proc. 13th Offshore
Technology Conf., Houston, paper no. OTC 4144.
Gottardi, G. & Butter®eld, R. (1995). The displacements of a model
rigid surface footing on dense sand under general planar loading.
Soils Found. 35, No. 3, 71±82.
Hambly, E. C. & Nicholson, B. A. (1989). Fatigue of jack-ups: simplify-
ing calculations. Proc. 2nd Int. Conf. Jack-up Drilling Platform,
London, 53±73. London: Elsevier Applied Science.
Hambly, E. C., Imm, G. R., & Stahl, B. (1990). Jackup performance
and foundation ®xity under developing storm conditions. Proc. 22nd
Offshore Technology Conf., Houston, paper no. OTC 6466.
Houlsby, G. T. & Martin, C. M. (1992). Modelling of the behaviour of
foundations of jack-up units on clay. Proceedings of the Wroth
Memorial Symposium `Predictive Soil Mechanics', Oxford, pp.
339±358. London: Thomas Telford.
Houlsby, G. T. & Wroth, C. P. (1983). Calculation of stresses on shallow
penetrometers and footings. Proceeedings of the IUTAM/IUGG
Symposium on Seabed Mechanics, Newcastle upon Tyne, pp.
107±112.
Le Tirant, P. & PeÂrol, C. (1993). Stability and operation of jackups.
Paris: EÂditions Technip.
Maier, G. & Hueckel, T. (1979). Nonassociated and coupled ¯ow rules
of elasto-plasticity for rock-like materials. Int. J. Rock Mech. Min.
Sci. 16, 77±92.
Martin, C. M. (1994). Physical and numerical modelling of offshore
foundations under combined loads. DPhil thesis, University of
Oxford.
Martin, C. M. (2001). Vertical bearing capacity of skirted circular
foundations on Tresca soil. Proc. 15th Int. Conf. Soil Mech. Geo-
tech. Engng, Istanbul (in press).
Martin, C. M. & Houlsby, G. T. (1999). Jackup units on clay: structural
analysis with realistic modelling of spudcan behaviour. Proc. 31st
Offshore Technology Conf., Houston, paper no. OTC 10996.
Martin, C. M. & Houlsby, G. T. (2000). Combined loading of spud-
can foundations on clay: laboratory tests. GeÂotechnique 50, No. 4,
325±338.
Murff, J. D. (1994). Limit analysis of multi-footing foundation systems.
Proc. 8th Int. Conf. Comp. Methods and Advances in Geomech.,
Morgantown 1, 223±244.
Ngo-Tran, C. L. (1996). The analysis of offshore foundations subjected
to combined loading. DPhil thesis, University of Oxford.
Noble Denton & Associates (1987). Foundation ®xity of jack-up units: a
joint industry study. London: Noble Denton & Associates.
Nova, R. & Montrasio, L. (1991). Settlements of shallow foundations on
sand. GeÂotechnique 41, No. 2, 243±256.
Poulos, H. G. (1988). Marine geotechnics. London: Unwin Hyman.
Reardon, M. J. (1986). Review of the geotechnical aspects of jack-up
unit operations. Ground Engng 19, No. 7, 21±26.
Roscoe, K. H. & Scho®eld, A. N. (1957). The stability of short pier
foundations in sand: discussion. Br. Weld. J., January, 12±18.
Schotman, G. J. M. (1989). The effects of displacements on the stability
of jack-up spudcan foundations. Proc. 21st Offshore Technology
Conf., Houston, paper no. OTC 6026.
Skempton, A. W. (1951). The bearing capacity of clays. Proceedings of
the Building Research Congress, London, Vol. 1, pp. 180±189.
SNAME (1997). Recommended practice for site speci®c assessment of
mobile jackup units, Rev. 1. Jersey City, NJ: Society of Naval
Architects and Marine Engineers.
Springman, S. M. & Scho®eld, A. S. (1998). Monotonic lateral load
transfer from a jack-up platform lattice leg to a soft clay deposit.
Proc. Int. Conf. Centrifuge 98, Tokyo. Rotterdam: Balkema,
563±568.
Tan, F. S. C. (1990). Centrifuge and theoretical modelling of conical
footings on sand. PhD thesis, University of Cambridge.
Thompson, R. S. G. (1996). Development of non-linear numerical
models appropriate for the analysis of jackup units. DPhil thesis,
University of Oxford.
Vesic, A. S. (1975). Bearing capacity of shallow foundations. In
Foundation engineering handbook (eds H. F. Winterkorn and H. Y.
Fang), pp. 121±147. New York: Van Nostrand.
Young, A. G., Remmes, B. D. & Meyer, B. J. (1984). Foundation
performance of offshore jack-up drilling rigs. J. Geotech. Engng
Div., ASCE 110, No. 7, 841±859.