Vous êtes sur la page 1sur 13

A plasticity model for the behaviour of footings on sand under combined

loading
G. T. HOULSBY
+
and M. J. CASSI DY{
A complete theoretical model is described for the behaviour
of rigid circular footings on sand, when subjected to com-
bined vertical, horizontal and moment loading. The model,
which is expressed in terms of work-hardening plasticity
theory, is based on a series of tests specically designed to
allow evaluation of the various components of the theory.
The model makes use of the force resultants and the
corresponding displacements of the footing, and allows pre-
dictions of response to be made for any load or displacement
combination. It is veried by comparison with the database
of tests. The use of the model is then illustrated by some
demonstration calculations for the response of a jack-up unit
on sand. This example illustrates the principal purpose of
the development, which is to allow a realistic modelling of
foundation behaviour to be included as an integral part of a
structural analysis.
KEYWORDS: footings/foundations; model tests; numerical model-
ling and analysis; offshore engineering; plasticity; sands
Nous decrivons un modele theorique complet pour le com-
portement d'assises circulaires rigides sur du sable quand
ces assises sont soumises a une charge combinee verticale,
horizontale et de moment. Le modele, qui est exprime en
terme de theorie de plasticite de durcissement a froid, est
base sur une serie d'essais (publies par Gottardi et al, 1999)
speciquement concus pour permettre l'evaluation des divers
composants de la theorie. Le modele utilise les resultantes
des forces et le deplacement correspondant de l'assise et
permet de predire la reponse a faire pour toute combinaison
de charge ou de deplacement. Il est verie par des compar-
aisons avec la base de donnees de tests. Nous illustrons
ensuite l'utilisation du modele par quelques calculs de de-
monstration pour la reponse d'une unite auto-elevatrice sur
du sable. Cet exemple illustre le but principal du developpe-
ment, qui est de permettre une modelisation realiste du
comportement des fondations, modelisation qui fera partie
integrante d'une analyse structurale.
INTRODUCTION
The purpose of this paper is to describe a theoretical model,
based on strain-hardening plasticity theory, which is capable of
describing the behaviour of a circular footing on sand when it
is subjected to all possible combinations of drained vertical,
horizontal and moment loading. The motivation for this work
comes principally from the offshore industry, specically arising
from the problem of assessment of jack-up units under extreme
loading. The applications are, however, much broader, since the
model could be applied to many instances of combined loading
of a footing on sand.
Structural engineers carry out detailed analyses of jack-up
units, and ask geotechnical engineers to provide them with the
values of spring stiffnesses to model the foundations. Geotech-
nical engineers tend to take the view that such a simplistic view
of foundation behaviour is unrealistic. Unfortunately, however,
they often describe the complexities and non-linearities of
foundation behaviour by a series of ad hoc procedures, which a
structural engineer cannot implement within a standard analysis.
The purpose of the model described here is to provide a means
by which the structural and geotechnical engineers can commu-
nicate. Geotechnical engineers must be prepared to re-cast their
knowledge of foundation behaviour within a terminology (plas-
ticity theory) that is amenable to numerical analysis. Structural
engineers must accept that soil behaviour cannot be described
merely by `springs', but can be accommodated if they are
prepared to use strain-hardening plasticity theory within their
analyses.
The ad hoc procedures for describing foundation behaviour
under combined loading have their roots in the work on bearing
capacity by Meyerhof (1953), and are typied by the procedures
described by Brinch Hansen (1970) and Vesic (1973). These
methods are adequate for predicting failure under combined
loads, but they are unsuitable for numerical analysis, principally
because they formulate the problem using a series of factors
applied to the bearing capacity formula for vertical loading,
modifying it to account for horizontal and moment loading.
This renders the analysis unsuitable for direct inclusion in
numerical analysis programs. Furthermore the conventional ana-
lyses pay no attention to the issue of plastic strains pre-failure,
since they treat only the failure problem.
An alternative is to address the problem directly as one of
loading within a three-dimensional (V, M, H) load space, and
to explore, for instance, the shape of the yield surface in this
space. This approach was pioneered by Roscoe & Schoeld
(1956), who were also concerned with a problem of soil
structure interaction: that of calculating the fully plastic moment
resistance of a short pier foundation for a steel framework. The
general framework of plotting load paths in (V, M, H) space has
been adopted by the offshore industry, but the formulae used to
derive the failure surfaces are often based on the shape and
inclination factor approach (see e.g. Hambly & Nicholson,
1991).
Recently there has been considerable interest in the develop-
ment of models based on plasticity theory, and on the experi-
mental work necessary to support this approach (e.g.
Schotmann, 1989; Nova & Montrasio, 1991: Gottardi & Butter-
eld, 1993, 1995; Houlsby & Martin, 1992; Martin, 1994). The
model described here is intended for the description of drained
loading of a circular foundation on dense sand, subjected to an
arbitrary combination of vertical, horizontal and moment loads.
It is complete in the sense that any load or deformation path
can be applied to the footing and the corresponding unknowns
(deformations or loads) calculated. The model is based on
experimental data by Gottardi & Houlsby (1995) and Gottardi
et al. (1999).
The loading of a footing clearly results in a complex state of
stresses in the soil. In the approach used here the response of
the foundation is, however, expressed purely in terms of force
resultants (V, M, H) on the footing. This simplication is very
convenient, especially as it allows the model to be coupled
directly to a numerical analysis of a structure. It is directly
analogous to the use of force resultants (tension, bending
moment and shear force) in the analysis of beams and columns.
117
Houlsby, G. T. & Cassidy, M. J. (2002). Geotechnique 52, No. 2, 117129
Manuscript received 27 February 2001; revised manuscript accepted 1
November 2001.
Discussion on this paper closes 1 September 2002, for further details
see p. ii.
+
Department of Offshore Engineering, Oxford University, UK.
{ Centre for Offshore Foundation Systems, University of Western
Australia (formerly at Oxford University).
However, it obscures some of the detailed response of the
footingfor instance the fact that a real footing probably does
not exhibit a truly `elastic' response of the sort employed within
the model for certain load combinations. Nevertheless, it proves
to be a useful idealisation.
OUTLINE OF THE MODEL
Before giving the detailed mathematical form of the expres-
sions used (see the next section), it is worth describing the
model in outline.
The principal concept adopted is that at any penetration of a
foundation into the soil, a yield surface in (V, M, H) space will
be established. Any changes of load within this surface will
result only in elastic deformation. Load points that touch the
surface can also result in plastic deformation. Although the
shape of this surface is assumed constant, the size may vary,
with the yield surface expanding as the footing is pushed further
into the soil. For simplicity the expansion of the yield surface is
taken solely as a function of the plastic component of the
vertical deformation.
The model is thus one of the strain-hardening plasticity type.
The precise form of the hardening law is specied by a
relationship between the size of the yield surface and the plastic
vertical deformation.
Within the yield surface, where the deformation is assumed
as elastic, the behaviour is specied by a set of elastic con-
stants.
Finally a statement must be made about the ow rule, which
determines the ratio between the plastic strains. The simplest
type of ow rule is `associated ow', in which the plastic
potential is the same as the yield surface. In this model a slight
variation is used in that the shape of the yield surface and
plastic potential are described by similar mathematical expres-
sions but with different parameter values. It is necessary to
introduce these parameters if the modelling of plastic vertical
deformations is to be at all reasonable.
There is a striking analogy between the structure of the
proposed model and that of constitutive models based on
critical-state concepts. In the analogy the vertical load plays the
same role as the mean normal stress, p9, the horizontal load or
the moment are equivalent to deviator stress, q, and the vertical
penetration plays the same role (with a change of sign) as the
voids ratio or specic volume. The analogy is pursued in more
detail by Houlsby & Martin (1992) and Martin (1994).
DETAILS OF THE MODEL
The model described here is known as Model C (Models A
and B were developed by Martin (1994) for footings on clay).
The sign conventions and nomenclature used in the following
are those suggested by Buttereld et al. (1997) and are shown
in Fig. 1. Typical parameter values for Model C are presented
in Table 1.
Reference point
Current position
w
u
V
M
H

Fig. 1. Sign conventions for load and displacement T


a
b
l
e
1
.
P
r
o
p
e
r
t
i
e
s
u
s
e
d
i
n
M
o
d
e
l
C
C
o
n
s
t
a
n
t
D
i
m
e
n
s
i
o
n
E
x
p
l
a
n
a
t
i
o
n
C
o
n
s
t
r
a
i
n
t
s
T
y
p
i
c
a
l
v
a
l
u
e
N
o
t
e
s
R
L
F
o
o
t
i
n
g
r
a
d
i
u
s
V
a
r
i
o
u
s

F
/
L
3
U
n
i
t
w
e
i
g
h
t
o
f
s
o
i
l
2
0
k
N
/
m
3
g

S
h
e
a
r
m
o
d
u
l
u
s
f
a
c
t
o
r
4
0
0
F
o
r
e
q
u
a
t
i
o
n
(
2
)
k
v

E
l
a
s
t
i
c
s
t
i
f
f
n
e
s
s
f
a
c
t
o
r
(
v
e
r
t
i
c
a
l
)
2

6
5
k
h

E
l
a
s
t
i
c
s
t
i
f
f
n
e
s
s
f
a
c
t
o
r
(
h
o
r
i
z
o
n
t
a
l
)
2

3
k
m

E
l
a
s
t
i
c
s
t
i
f
f
n
e
s
s
f
a
c
t
o
r
(
m
o
m
e
n
t
)
0

4
6
k
c

E
l
a
s
t
i
c
s
t
i
f
f
n
e
s
s
f
a
c
t
o
r
(
h
o
r
i
z
o
n
t
a
l
/
m
o
m
e
n
t
c
o
u
p
l
i
n
g
)

1
4
h
0

D
i
m
e
n
s
i
o
n
o
f
y
i
e
l
d
s
u
r
f
a
c
e
(
h
o
r
i
z
o
n
t
a
l
)
0

1
1
6
M
a
x
i
m
u
m
v
a
l
u
e
o
f
H
a
V
0
o
n
M
=
0
m
0

D
i
m
e
n
s
i
o
n
o
f
y
i
e
l
d
s
u
r
f
a
c
e
(
m
o
m
e
n
t
)
0

0
8
6
M
a
x
i
m
u
m
v
a
l
u
e
o
f
M
a
2
R
V
0
o
n
H
=
0
a

E
c
c
e
n
t
r
i
c
i
t
y
o
f
y
i
e
l
d
s
u
r
f
a
c
e

1
X
0
,
a
,
1
X
0

C
u
r
v
a
t
u
r
e
f
a
c
t
o
r
f
o
r
y
i
e
l
d
s
u
r
f
a
c
e
(
l
o
w
s
t
r
e
s
s
)

1
<
1
X
0
0

9
0

1
=

2
=
1
g
i
v
e
s
p
a
r
a
b
o
l
i
c
s
e
c
t
i
o
n

C
u
r
v
a
t
u
r
e
f
a
c
t
o
r
f
o
r
y
i
e
l
d
s
u
r
f
a
c
e
(
h
i
g
h
s
t
r
e
s
s
)

2
<
1
X
0
0

9
9

1
=

2
=
1
g
i
v
e
s
p
a
r
a
b
o
l
i
c
s
e
c
t
i
o
n

C
u
r
v
a
t
u
r
e
f
a
c
t
o
r
f
o
r
p
l
a
s
t
i
c
p
o
t
e
n
t
i
a
l
(
l
o
w
s
t
r
e
s
s
)

3
<
1
X
0
0

5
5

C
u
r
v
a
t
u
r
e
f
a
c
t
o
r
f
o
r
p
l
a
s
t
i
c
p
o
t
e
n
t
i
a
l
(
h
i
g
h
s
t
r
e
s
s
)

4
<
1
X
0
0

6
5

A
s
s
o
c
i
a
t
i
o
n
f
a
c
t
o
r
(
h
o
r
i
z
o
n
t
a
l
)
1

5
V
a
r
i
a
t
i
o
n
a
c
c
o
r
d
i
n
g
t
o
e
q
u
a
t
i
o
n
(
9
)
a
n
d

=
2
X
5

A
s
s
o
c
i
a
t
i
o
n
f
a
c
t
o
r
(
m
o
m
e
n
t
)
1

1
5
V
a
r
i
a
t
i
o
n
a
c
c
o
r
d
i
n
g
t
o
e
q
u
a
t
i
o
n
(
1
0
)
a
n
d

=
2
X
1
5
k
9

R
a
t
e
o
f
c
h
a
n
g
e
i
n
a
s
s
o
c
i
a
t
i
o
n
f
a
c
t
o
r
s
0

1
2
5
f

I
n
i
t
i
a
l
p
l
a
s
t
i
c
s
t
i
f
f
n
e
s
s
f
a
c
t
o
r
0

1
4
4
N

B
e
a
r
i
n
g
c
a
p
a
c
i
t
y
f
a
c
t
o
r
(
p
e
a
k
)
1
5
0

3
0
0

D
i
m
e
n
s
i
o
n
l
e
s
s
p
l
a
s
t
i
c
p
e
n
e
t
r
a
t
i
o
n
a
t
p
e
a
k
0

0
3
1
6
118 HOULSBY AND CASSIDY
Elastic behaviour
The elastic relationship between the increments of load (dV,
dM, dH) and the corresponding elastic displacements (dw
e
,
d
e
, du
e
) is
dV
dMa2R
dH
P
R
Q
S
= 2RG
k
v
0 0
0 k
m
k
c
0 k
c
k
h
P
R
Q
S
dw
e
2Rd
e
du
e
P
R
Q
S
(1)
where R is the radius of the footing, G is a representative shear
modulus, and k
v
, k
m
, k
h
, k
c
are dimensionless constants. The
values of these constants may be derived using, for instance,
nite element analysis of a footing (Bell, 1991; Ngo Tran,
1996), and typical values are given in Table 1. The values of
the dimensionless constants depend on the geometry of the
footing (e.g. cone angle and depth of embedment) as well as
the Poisson's ratio for the sand.
An appropriate value of G is one of the most difcult
parameters to establish for the model. Recognising that the
mobilised shear stiffness is strongly dependent on the shear
strain, the value has to be a compromise one that is representa-
tive of typical strains in the soil. It has been determined here by
tting of overall curves to experimental data. The shear mod-
ulus also depends on stress level, and is typically proportional
to approximately the square root of the mean effective stress. It
is convenient therefore to estimate the shear modulus through
use of a formula such as
G
p
a
= g

V
Ap
a
s
(2)
where p
a
is atmospheric pressure, V is a representative vertical
load on the foundation, A = r
2
is the plan area of the
foundation, and g is a dimensionless constant. A typical value
of g is approximately 400 for medium dense sand, but would
be expected to depend mildly on the relative density. Note that
equation (2) represents a different scaling relationship than was
used in Cassidy (1999), and is suggested on the basis of more
recent work.
Yield surface
The yield surface is most conveniently expressed in dimen-
sionless terms, using the variables v = VaV
0
, m = Ma2RV
0
,
h = HaV
0
, where V
0
is the parameter that denes the size of
the yield surface. The chosen form of the surface that ts the
observed behaviour of footings well is that used by Martin
(1994):
f =
h
h
0

2

m
m
0

2
2a
h
h
0
m
m
0

12
()
21
(1 )
22
= 0
(3)
where the factor

12
=
(
1

2
)
12
(
1
)
1
(
2
)
2
2 3
2
is introduced so that h
0
and m
0
have simple physical interpreta-
tions. This surface may seem unnecessarily complicated, and it
is perhaps useful to consider a simplied form in which a = 0
and
1
=
2
= 1:
f =
h
h
0

2

m
m
0

2
16
2
(1 )
2
= 0 (4)
It is straightforward to show that this is a `rugby ball' shaped
surface that is elliptical in section on planes at constant V, and
parabolic on any section including the V-axis: see Fig. 2.
Although there is some theoretical justication for this choice
of shape (particularly in the (V, M) plane), it is largely chosen
empirically. The size of the surface is determined by the point
on the surface at maximum V value, which is given by
(V, M, H) = (V
0
, 0, 0). The shape of the surface is determined
by the two parameters h
0
and m
0
, which determine the ratios of
HaV and Ma2RV at the widest section of the surface, which
occurs at V = V
0
a2.
The factor a in equation (3) allows the ellipse to become
eccentric (that is, the principal axes are no longer aligned with
the H- and M-axes). This is necessary for accurate modelling
of the experimental data, and accounts for the fact that if, for
instance, the footing is subjected to a horizontal load from left
to right, a clockwise moment will produce a different response
from an anticlockwise moment. The factors
1
and
2
are
introduced following Nova & Montrasio (1991). They have two
advantages: (a) the position of the maximum size of the
elliptical section can be moved from V = V
0
a2 to
V =
2
V
0
a(
1

2
), thus tting experimental data better; and
(b) by choosing
1
, 1 and
2
, 1 the sharp points on the
surface at V = 0 and V = V
0
can be eliminated, which has
advantages in the numerical implementation of the model. If

1
=
2
= 0
X
5 the yield surface becomes an ellipsoid. The
factor
12
in equation (2) is simply so that h
0
and m
0
retain
their original meanings.
Strain hardening
The form of the strain-hardening expression can be deter-
mined from a vertical loadpenetration curve, since for pure
vertical loading V
0
= V. Typical loadpenetration curves are
shown in Fig. 3, showing a peak in the loadpenetration curve
for the dense sand tested by Gottardi & Houlsby (1995). An
expression that ts the data well, and which is shown in Fig. 3,
is
V
0
=
kw
p
1
kw
pm
V
0m
2

w
p
w
pm

w
p
w
pm

2
(5)
where k is an initial plastic stiffness, w
p
is the plastic compo-
nent of the vertical penetration, V
0m
is the peak value of V
0
,
and w
pm
is the value of w
p
at this peak. No special signicance
is attached to this particular form of the t to the vertical load
penetration response, and alternative expressions that tted other
experimental data could also be appropriate.
A formula that models post-peak work softening as well as
pre-peak performance was essential. However, equation (5)
unrealistically implies V
0
0 as w
p
. Therefore it can be
used only for a limited range of penetrations. It is assumed that
for most properly designed foundations on dense sand, loading
post-peak would not be expected; however, for a complete
model capable of tting post-peak behaviour more realistically,
equation (3) can be altered to
V
0
=
kw
p

f
p
1 f
p

w
p
w
pm

2
V
0m
1
kw
pm
V
0m
2

w
p
w
pm

1
1 f
p

w
p
w
pm

2
(6)
M/2R
H
V
Fig. 2. Shape of yield surface
PLASTICITY MODEL FOR FOOTINGS ON SAND 119
where f
p
is a dimensionless constant that describes the limiting
magnitude of vertical load as a proportion of V
0m
(that is,
V
0
f
p
V
0m
as w
p
). It is possible to use the same
parametric values of k, V
0m
and w
pm
as in equation (5). For
realistic footing designs in which it was not required to describe
softening, a much simpler equation than equation (6) could be
used. The precise form of this equation is not in fact central to
the model; all that is required is a convenient expression that
ts observed data and denes V
0
as a function of w
p
.
Plastic potential
In the (Ma2R, H) plane an associated ow rule is found to
model the ratios between the plastic displacements well, but this
is not the case in the (V, Ma2R) or (V, H) planes, for which
an associated ow rule is found to predict unrealistically large
vertical displacements. A plastic potential different from the
yield surface must therefore be specied. A convenient expres-
sion is, however, very similar to that used for the yield surface:
g =
h9
h
0

2

m9
m
0

2
2a
h9
h
0
m9
m
0

2
v

34
9 ( )
23
1 9 ( )
24
= 0
(7)
where

34
=
(
3

4
)
( 34)
(
3
)
3
(
4
)
4
2 3
2
and
v
is an association parameter (associated ow is given by

v
= 1
X
0). Note that the condition g = 0 is used to dene a
dummy parameter V9
0
which gives the intersection of the plastic
potential with the V-axis. The primed parameters are dened by
v9 = VaV
0
9, m9 = Ma2RV
0
9 and h9 = HaV
0
9. Factors
3
and

4
have been introduced, which can be chosen independently
from
1
and
2
. The association parameter
v
allows for
variation of the vertical displacement magnitude, with values
greater than 10 resulting in the increase of the vertical displa-
cements. It also controls the position of the `parallel point' as
dened by Tan (1990), which is the point on the yield locus at
which the footing could rotate (or move sideways) at constant
vertical load and with no further vertical deformation. Accurate
prediction of this point is important as it describes the transition
between settlement and heave of the footing and where sliding
failures will occur. In the analogy with critical-state models, this
point plays the same role as the critical state. When associated
ow is used (
v
= 1,
3
=
1
,
4
=
2
) the parallel point
occurs at =
2
a(
1

2
): that is, the largest constant vertical
load section of the yield surface. As
v
is decreased the
position of the parallel point moves to a lower value of vertical
load, but the exact expression for the value of becomes very
complex. The modelling of realistic vertical displacements and
of the position of the parallel point are linked, and with only
one parameter it is difcult to model both adequately.
Increasing h
0
or m
0
with two association factors, rather than
scaling the vertical component, enables the plastic potential's
shape to change in the radial plane. This consequently changes
radial plastic displacements. This method has the advantage of
more exibility in modelling subtle differences between hori-
zontal and moment loading results. Using two association
factors the plastic potential may be dened as
g =
h9

h
h
0

2

m9

m
m
0

2
2a
h9m9

m
h
0
m
0

34
(9)
23
(1 9)
24
= 0 (8)
If
h
and
m
are constant and equal, equation (7) is equiva-
lent to equation (8) for the same value of
v
.
In fact it was found that experimental data can be tted well
only if the
h
and
m
factors are themselves taken as variable.
The values of
h
and
m
that best t both the radial displace-
ment and constant V tests of Gottardi & Houlsby (1995) were
found to be hyperbolic functions of plastic displacement his-
tories:

h
=
k9
h
(u
p
aw
p
)
k9 (u
p
aw
p
)
(9)

m
=
k9
m
(2R
p
aw
p
)
k9 (2R
p
aw
p
)
(10)
where k9 determines the rate of change of the association
factors. For no previous radial displacements,
h
and
m
equate
to 1 and associated ow is assumed. The rates at which
h
and

m
vary in Model C are depicted in Fig. 4. With the plastic
potential dened as in equation (6), the following values were
evaluated:

3
= 0
X
55;
4
= 0
X
65;
h
= 2
X
5;
m
= 2
X
15; k9 = 0
X
125
Further details of the development of the plastic potential in
equation (8) and comparisons between the theory and experi-
mental data can be found in Cassidy (1999).
Partially drained behaviour
The model described above is based on data from tests on
dry sand, and thus describes fully drained behaviour. For
2500
2000
1500
1000
500
0
1 0 1 2 3 4 5 6 7 8 9 10
Experiments
Theory
w
p
: mm
V
:

N
Fig. 3. Theoretical t of the vertical load tests
120 HOULSBY AND CASSIDY
realistic loading times of large offshore foundations, partially
drained behaviour is expected, and the above model would need
to be modied to take into account the transient pore pressures
beneath the foundation. Both Mangal (1999) and Byrne (2000)
have carried out model tests equivalent to those used here, but
on saturated sand and at loading rates where partially drained
behaviour occurs. They record that loading rate has remarkably
little effect on the loaddeformation response, so that the
current model provides a reasonable starting point for descrip-
tion of partially drained behaviour. Some caution is of course
necessary if there is any possibility that the magnitude of the
transient pore pressures might be sufcient to induce liquefac-
tion phenomena.
RETROSPECTIVE MODELLING OF EXPERIMENTS
To investigate the capabilities of Model C to model footing
behaviour, numerical simulations were carried out for a number
of representative experiments. In each of these simulations the
measured values of three of the measured quantities (e.g. the
displacements) were taken as input, and the other three quanti-
ties (e.g. the loads) were calculated as output for comparison
with the experiments. No idealisation of the experimental input
data was carried out, so that the input values contain all the
minor uctuations associated with experimental measurements.
The program used to implement Model C is able to handle such
perturbations.
The simulations are carried out for the tests reported by
Gottardi & Houlsby (1995), using a 100 mm diameter footing
on medium dense Leighton Buzzard sand. These are the same
tests that were used for the development of Model C, so that
the quality of the t is of course expected to be good. The
purpose of this exercise is, however, twofold: (a) to demonstrate
that Model C can be implemented numerically, and used to
simulate footing behaviour; and (b) to assess the overall cap-
ability of the model to capture the salient features of the
original data.
Vertical penetration test
Figure 5(a) shows the experimental results for a vertical
penetration test. Fig. 5(b) is a simulation of this same test in
which the measured displacement is taken as input, and the
vertical load calculated. Model C gives loads that accurately
represent the original test, and this is principally a test of the
chosen strain-hardening law. The three vertical unload/reload
loops pre-peak are modelled well, although Model C does not
reect the hysteresis that occurs in the experimental results.
This does make a slight, but not too signicant, reduction in
the displacements compared with their corresponding loads.
The Model C program predicts the location of the existing
yield surface when being reloaded in an unloadreload loop. It
does not overshoot the yield surface because of a bisection
algorithm used to determine the proportion of the increment
that is elastic, with the remaining proportion allocated as
elastoplastic.
In each of Figs 6Fig. 11 following, (a) and (b) represent the
measured experimental data, and (c) and (d) represent the
Model C simulation.
Moment and horizontal swipe tests from V ~ 1600 N
In a swipe test the footing is load-controlled in the vertical
direction until it reaches a prescribed load, in this case
V ~ 1600 N. Rotation or horizontal displacement is then ap-
plied to the footing with the trace corresponding to a track
along the yield surface, appropriate for that embedment.
Figure 6 represents a moment swipe starting at V ~ 1600 N.
Prior to the swipe the footing is loaded in the purely vertical
direction with only small amounts of horizontal and moment
load being developed. However, for clarity, only the swipe has
been plotted. Model C simulates the magnitude of peak moment
adequately, reaching a value just over Ma2R = 150 N. The
numerical peak moment in Fig. 6(d) and the experimental peak
moment in Fig. 6(b) occurred at the same vertical load.
Additionally, Figs 6(a) and (c) show that the amount of rotation
before the peak was modelled accurately. However, in this test
Model C locates the `parallel point' slightly lower than the
experiment (point A in Fig. 6(d)). In the Model C simulation in
Fig. 6(d) movement back along the yield surface can be seen to
occur, for instance at V ~ 800 N and again at V ~ 600 N.
Figure 7 represents an equivalent swipe, but in the horizontal
direction, with Model C load-controlled to V ~ 1600 N and
then displacement-controlled for the swipe. The program models
the track along the yield surface very well, with the peak
horizontal load almost exactly matching that of the experiment
at just over 200 N. Fig. 7(c) shows Model C predicting a very
similar displacement path to the experiments (Fig. 7(a)), verify-
ing the ow rule for this case. The simulation stops tracking at
around the same horizontal and vertical load levels, indicating
accurate prediction of the `parallel point' in the horizontal
plane. Further justication of the use of two independent
association factors (
h
and
m
) in the ow rule is given by the
more accurate prediction of the `parallel point' for both the
moment and horizontal swipes than would be possible if there
were only one.
Moment and horizontal swipe tests from V ~ 200 N
Figure 8 represents moment and horizontal swipes starting at
V ~ 200 N, highlighting the yield surface at low vertical loads.
In order to depict the experiments, Model C is load-controlled
to V ~ 1600 N and then unloaded to V ~ 200 N, before being
displacement-controlled throughout the swipe. Figs 8(a) and (c)
show that at low vertical loads both the experimental results
and Model C depict work-hardening, with Model C simulating

h
30
25
20
15
10
05
0
0 1 2 3
Horizontal
Moment
4

h

o
r

m
u
p
/w
p
or 2R
p
/w
p
Fig. 4. Rates of variation of
h
and
m
in Model C
2500
2000
1500
1000
500
0
V
:

N
2500
2000
1500
1000
500
0
V
:

N
0 1 2 3 4 5 6 7 8 9 10 11
(b)
(a)
w: mm
Fig. 5. Retrospective simulation of vertical penetration test
PLASTICITY MODEL FOR FOOTINGS ON SAND 121
the experiment well. This was not the case horizontally. Fig.
8(b) shows that the experiment elastically loads in the horizon-
tal direction before yielding occurs at H ~ 80 N (A B), and
then tracks along the yield surface with a reduction in vertical
and horizontal load (B C). This implies work-softening of
the sample. Model C simulates the elastic horizontal loading
very well, predicting the yield surface at the same position (line
segment A B on Fig. 8(d)). However, it then predicts that
0
50
50
100
150
200
1 0 1 2 3
(c)
0 200 400 600 800 1000 1200 1400 1600
(d)
(a) (b)
4 5 6 7
2R: mm V: N
M
/
2
R
:

N
0
50
50
100
150
200
M
/
2
R
:

N
A
Fig. 6. Retrospective simulation of moment swipe test
0
50
50
100
150
250
200
1 0 1 2 3
(c)
0 200 400 600 800 1000 1200 1400 1600
(d)
(a) (b)
u: mm V: N
H
:

N
0
50
50
100
150
250
200
H
:

N
Fig. 7. Retrospective simulation of horizontal swipe test
122 HOULSBY AND CASSIDY
work-hardening will occur, with increasing vertical and horizon-
tal load tracking up around the yield surface. This is consistent
with, and entirely related to, Model C's prediction of the
parallel point from a swipe at 1600 N. Nevertheless, it does
indicate that Model C's ow rule will not always follow the
experimental performance.
Constant vertical load tests
The constant V tests as shown in Figs 9 and 10 are simulated
with full load control to V ~ 1600 N, before the vertical load is
held constant at around that value (with slight uctuations
according to the experimental data), while horizontal and mo-
ment displacement control models an excursion. The constant V
tests involve the expansion and then later contraction of the
yield surface. Figs 9(c) and Fig 10(c) show that Model C
models expanding yield surfaces reasonably well, reaching a
similar peak for horizontal and moment load as the experimen-
tal values. Once the peak value has been reached, the Model C
surface then contracts back as predicted by the post-peak per-
formance of the hardening law. Fig. 9(c) shows that for the
predominately moment case this post-peak performance is ade-
quately modelled, although the experimental data did not con-
tinue until Ma2R = 0. However, in the horizontal constant V
test the experiment did continue until H = 0, but this was not
reproduced by the simulation program, with the surprising result
of increasing H occurring at the end of the test (du . 4 mm at
point A on Fig. 10(c)). The cause of this rise in H in the
numerical simulation is a rapid decrease in the experimentally
recorded vertical load, which was used as input. Between point
A and the end of the test, V falls from approximately 1600 N
to 1400 N. If V were held constant at 1600 N, Model C would
simulate the horizontal load decreasing back to zero, as would
be theoretically expected during a constant V test. This is a
good example showing that prediction in Model C is very
sensitive to the value of V near the peak value of capacity.
Figures 9(d) and Fig. 10(d) show that the ow rule satisfacto-
rily predicts the vertical displacements when compared with the
horizontal or rotational displacements, which are part of the
input. Fig. 9(d) shows a slight over-prediction in vertical
displacements, indicating that for this case the plastic potential's
surface is too steep, or too normal when compared with the
V9 axis. However, with Fig. 10(d) showing a slight under-
prediction, the ow rule is predicting balanced results.
Radial displacement tests
Constant gradients of horizontal to vertical and moment to
vertical displacement were used as inputs to simulate horizontal
and moment radial displacement tests. The resultant experimen-
tal moment and horizontal loads and Model C predictions are
shown in Figs 11 and 12 respectively, noting that (a) represents
the measured experimental data and (b) the Model C simulation.
The simulations are of similar gradient, implying that the Model
C ow rule is performing well. The noise that can be seen in
Figs 11(b) and Fig. 12(b) is due to the uctuations that occur in
the real experimental input data.
APPLICATION OF MODEL C TO A JACK-UP UNIT
The advantage of Model C is that it uses strain-hardening
plasticity theory, which can be easily implemented within a
conventional structural analysis program. To show this an
example analysis of a jack-up unit, a typical application in the
offshore industry, is outlined. Conventionally the foundations
for a jack-up platform are represented as pinned, or at best by
linear, springs. This is an over-simplication of foundation
behaviour, and can therefore lead to over-conservative design.
Model C represents an important advance from this type of
analysis, as it calculates a non-linear stiffness matrix based on
measured soil behaviour under combined loads. In the following
example a three-legged jack-up has been idealised as a plane
frame with Model C footings making up one component of a
realistic representation of the structure, foundations and environ-
mental loading. The model representing the idealised jack-up is
shown in Fig. 13.
0
50
50
100
150
200
(c)
0 200 400 600 800 1000 1200 1400 1600
(d)
(a) (b)
V: N
0 200 400 600 800 1000 1200 1400 1600
V: N
M
/
2
R
:

N
0
50
50
100
150
200
M
/
2
R
:

N
C
C
B
B
A
A
Fig. 8. Retrospective simulation of swipe test from V 200 N
PLASTICITY MODEL FOR FOOTINGS ON SAND 123
For the modelling of jack-up response, structural non-linear-
ities must be considered if reasonable accuracy is to be
achieved. In this analysis P and Euler effects are both
accounted for by using Oran's (1973) formulation of beam
column theory to specify the stiffness matrix, with additional
modications to produce the additional end rotations on the
beam due to the presence of shear (see Martin, 1994). The
stiffness matrix is derived in incremental form, as the structural
response is path dependent. The dynamic equations of motion
are solved in the time domain using the Newmark = 0
X
25,
= 0
X
5 numerical step-by-step direct integration technique.
Two examples of environmental wave loading were used; the
rst being NewWave theory (Tromans et al., 1991) and the
second a constrained NewWave in a background random sea
state (Taylor et al., 1995). These theories are a signicant
advance on conventional deterministic wave theories as they
account for the spectral composition and randomness found in
the ocean. In the derivation of the wave kinematics Wheeler
stretching was used. Hydrodynamic loading was evaluated using
the extended Morison's equation, including relative motion
effects, integrated to the moving free surface. In the examples
the environmental force was purely wave loading with no
current or wind.
Figure 14 shows the surface elevation of a NewWave in the
time domain for both jack-up legs with the wave focused on the
upwave leg at the reference time (t = 0 s). The sea-state can be
described by the Pierson Moskowitz wave energy spectrum,
with a signicant wave height ( H
s
) of 12 m and a mean zero
crossing period (T
Z
) of 10 s.
The corresponding horizontal deck displacements due to this
NewWave are shown in Fig. 15 for three foundation cases:
pinned, Model C, and linear springs. Pinned footings represent
innite horizontal and vertical stiffness, but no rotational
stiffness. Model C is the strain-hardening plasticity model for
sand described in this paper, and linear springs use nite
stiffness values as in the elastic region of the Model C case
(equation (1)). Though only deck displacements have been
shown, any other measure of structural response could be
determined. After the NewWave passes, the rig can be seen to
be vibrating in its natural mode. With increased rotational
xity the natural periods decrease, with approximate values of
9 s, 5 s and 5 s for the pinned, Model C and linear springs
respectively. In this example the load combinations on the
Model C footings were contained entirely within the yield
surface, thus giving a response identical to the linear spring
case. By increasing the NewWave crest amplitude to = 15 m
or = 18 m, as shown in Fig. 16, the increased loading
caused plastic displacements in the Model C footings, shifting
the entire foundations and leaving a permanent offset in the
displacement of the deck. This yielding of the sand footings
occurred during the peak of the NewWave. This direct indica-
tion of yielding is a major benet in using elastoplastic
formulations for the spud-can footings. The natural period after
this event may also be modied by the plastic behaviour.
7
8
6
5
4
3
2
1
0
1 0 1 2 3
(c)
4 5 6
2R: mm
1 0 1 2 3
(d)
(a) (b)
4 5 6
2R: mm
M
/
2
R
:

N
M
/
2
R
:

N
0
50
50
100
150
200
250
Fig. 9. Retrospective simulation of constant V test
124 HOULSBY AND CASSIDY
Figure 17 illustrates the surface elevation of a NewWave with a
crest elevation of 15 m embedded in a random sea characterised
by H
S
= 12 m and T
Z
= 10 s. The wave has been constrained,
such that at about 5934 s its peak collides with the upwave leg
of the jack-up; the surface elevation for the downwave is also
displayed. The corresponding deck displacements with time are
shown in Fig. 18 for the pinned, Model C and linear spring
foundation assumptions. For this example, the peak displace-
ments have been increased compared with just the equivalent
NewWave (Fig. 16). This is due to the random background and
the structural memory it causes; displaying that for dynamically
sensitive structures, such as jack-ups, the response is conditional
not only on the present applied load, but also on the load
history. As was the case for a jack-up loaded exclusively by a
7
8
6
5
4
3
2
1
0
1 0 1 2 3
(c)
4 5 6
u: mm u: mm
1 0 1 2 3
(d)
(a) (b)
4 5 6
H
:

N
M
/
2
R
:

N
0
50
50
100
150
200
250
Fig. 10. Retrospective simulation of constant V test
0
50
100
150
250
200
0 200 400 600 800 1000 1200 1400 1600
(a)
V: N
0 200 400 600 800 1000 1200 1400 1600
(b)
V: N
M
/
2
R
:

N
Fig. 11. Retrospective simulation of radial displacement test
PLASTICITY MODEL FOR FOOTINGS ON SAND 125
NewWave, the assumption of pinned footings is clearly illu-
strated in Fig. 17 as overly conservative. The linear springs can
be seen to yield lower displacements than the Model C footing
owing to the greater stiffness exhibited. In addition, Model C
indicates a permanent horizontal displacement of the jack-up
due to the passage of this extreme wave.
These two examples show that Model C offers a method for
structural analysis with realistic soilstructure interaction. Mod-
el C has the exibility to account for complex loading schemes
and dynamic response, and represents realistic non-linear stiff-
ness of circular footings on dense sand. In a recent project
Model C has been successfully used in tting a number of case
records of the response of jack-up units in the North Sea.
CONCLUSIONS
Based on a series of experiments performed at the University
of Oxford by Gottardi & Houlsby (1995), this paper details a
work-hardening plasticity model entitled Model C that has been
developed to represent circular footings on sand. The yield
surface, ow rule and hardening law of Model C are all
empirically determined to t the experimental data. Stiffness
factors derived from three-dimensional nite element analyses
are used to describe elastic behaviour within the yield surface.
In order to investigate the predictive capabilities of Model C,
a number of experimental tests were chosen for numerical
simulation, with the experimental data resulting from the model
tests incorporated as input for the numerical simulation. These
0
50
100
150
250
200
0 200 400 600 800 1000 1200 1400 1600
(a)
V: N
0 200 400 600 800 1000 1200 1400 1600
(b)
V: N
H
:

N
Fig. 12. Retrospective simulation of radial displacement test
Upwave Downwave
5196 m
352 m
80 m
Single leg
Two legs
M
e
a
n

w
a
t
e
r

d
e
p
t
h

9
0

m
Values:
For a single leg:
E = 200 GPa
I = 15 m
4
A = 06 m
2
M = 193 10
6
kg
A
s
= 004 m
2
G = 80 GPa
D
E
= 844 m
A
h
= 394 m
2
C
d
= 11
C
m
= 20
For hull:
I = 150 m
4
A
s
= 02 m
2
M = 161 10
6
kg
For spud-cans:
R = 10 m
Structural damping 2%
of crucial
Fig. 13. General layout of idealised jack-up used in the analyses
126 HOULSBY AND CASSIDY
tests represented the varying nature of the load and displace-
ment paths imposed upon the model footing during the testing
regime. It has been shown that Model C adequately provides all
the requirements of a work-hardening plasticity theory to model
the available experimental data.
The experimental evidence of Gottardi & Houlsby (1995) did
not support the application of associated ow in Model C, and
a plastic potential function, g, was dened. Model C's plastic
potential is dened by ve parameters, arguably an overly
complex arrangement to explain the ow rule of shallow
12
8
4
0
4
8
12
60 40 20 0 20 40 60
Time: s
S
u
r
f
a
c
e

e
l
e
v
a
t
i
o
n
:

m
Upwave leg
Downwave leg
Downwave
leg
Upwave
leg
x = 0 m x = 5196 m
H
s
= 12 m
T
z
= 10 s
= 12 m
Fig. 14. NewWave surface elevation at the upwave and downwave legs
08
06
04
02
02
04
06
08
10
0
60 40 20 0 20 40 60
Time: s
H
o
r
i
z
o
n
t
a
l

d
e
c
k

d
i
s
p
l
a
c
e
m
e
n
t
s
:

m
Pinned
Model C and linear springs
Fig. 15. Horizontal deck displacements due to the NewWave loading
08
06
04
02
02
04
06
08
10
0
60 40 20 0 20 40 60
Time: s
H
o
r
i
z
o
n
t
a
l

d
e
c
k

d
i
s
p
l
a
c
e
m
e
n
t
s
:

m
= 18 m
= 15 m
= 12 m
Fig. 16. Horizontal deck displacements due to increasing amplitude NewWaves
PLASTICITY MODEL FOR FOOTINGS ON SAND 127
circular footings. However, systematic variation of the associa-
tion factors makes it possible to model the differences caused
by the loading direction, resulting in greater condence in the
ability to model a real load path. Furthermore, with uncoupled
horizontal and moment association factors, greater exibility in
the modelling of the location of the parallel point is possible.
As a less complicated alternative, with only three parameters, a
compromise solution with constant association factors could be
used. This is outlined in Cassidy (1999) and has the values

3
= 0
X
55,
4
= 0
X
65 and
h
=
m
= 2
X
05. Further physical
experimentation, especially at low radial to vertical load levels,
could lead to a more accurately dened, and perhaps a less
complex, ow rule.
Model C has been developed from monotonic loading tests
on sand. However, for the analysis of offshore structures,
loading rates and cyclic loading are both important. The
strength of the sand foundation is related to the rate of the
applied load and the degree of drainage of developed pore
pressures. Furthermore, reversal of load paths and cyclic behav-
iour (as can be expected in an ocean environment) can cause
reduction of the strength in the soil. The next step in formulat-
ing a more advanced plasticity model is to account for these
effects. However, Model C has reached a stage in its develop-
ment where it can be condently implemented as a realistic
soilstructure model for dense sand.
Within the offshore industry, Model C represents a signicant
advance in the response analysis of jack-up units. When com-
pared with techniques widely used in the jack-up industry, a
signicantly different response is found, as was shown in this
paper. An extension to Model C to account for the conical
shape of spud-can footings has been suggested in Cassidy
(1999).
ACKNOWLEDGEMENTS
The experimental work on which this paper is based was
carried out in cooperation with Dr G. Gottardi of the University
of Bologna. Support from the Rhodes Trust for the second
author is gratefully acknowledged.
NOTATION
a eccentricity factor for yield surface
f yield surface
f
p
softening parameter
g shear modulus factor
g plastic potential
G shear modulus
h
0
dimensionless size factor for yield surface
15
10
5
0
5
10
15
0 20 40 60 80 100 120
Time: s
S
u
r
f
a
c
e

e
l
e
v
a
t
i
o
n
:

m
Upwave leg
Downwave leg
Downwave
leg
Upwave
leg
x = 0 m x = 5196 m
H
s
= 12 m
T
z
= 10 s
= 15 m
Fig. 17. Surface elevations at the upwave and downwave legs for a constrained NewWave
12
16
08
04
0
04
08
12
0 20 40 60 80 100 120
Time: s
Pinned
Model C
Linear springs
H
o
r
i
z
o
n
t
a
l

d
e
c
k

d
i
s
p
l
a
c
e
m
e
n
t
s
:

m
Fig. 18. Horizontal deck displacements due to the constrained NewWave
128 HOULSBY AND CASSIDY
h, h9 dimensionless horizontal load, modied dimensionless
horizontal load in plastic potential
H horizontal load
H
S
signicant wave height
k hardening parameter
k9 constant in expression for association factors
k
v
, k
m
, k
h
, k
c
elastic stiffness factors
v, v9 dimensionless vertical load, modied dimensionless
vertical load in plastic potential
m
0
dimensionless size factor for yield surface
M moment
p
a
atmospheric pressure
R foundation radius
u horizontal displacement
m, m9 dimensionless moment, modied dimensionless
moment in plastic potential
T
Z
mean zero crossing period
V hertical load
V
0
dimension of yield surface in V-direction (maximum
past vertical load for pure vertical loading)
V
0
9 dummy size parameter for plastic potential
V
0m
peak value of V
0
w vertical displacement
w
pm
value of w
p
at V
0
= V
0m
wave crest amplitude

v
,
m
,
h
association factors

m
,
h
limiting values of association factors

1
,
2
,
12
shaping factors for yield surface

3
,
4
,
34
shaping factors for plastic potential
rotation
Subscripts
e elastic
p plastic
REFERENCES
Bell, R. W. (1991). The analysis of offshore foundations subjected to
combined loading. MSc thesis, Oxford University.
Brinch Hansen, J. (1970). A revised and extended formula for bearing
capacity. Bulletin No. 98, pp. 511. Copenhagen: Danish Geotechni-
cal Institute.
Buttereld, R., Houlsby, G. T. and Gottardi, G. (1997). Standardized
sign conventions and notation for generally loaded foundations.
Geotechnique 47, No. 5, 10511054.
Byrne, B. W. (2000). Investigations of suction caissons in dense sand.
DPhil thesis, Oxford University.
Cassidy, M. J. (1999). Non-linear analysis of jack-up structures sub-
jected to random waves. DPhil thesis, Oxford University.
Gottardi, G. & Buttereld, R. (1993). On the bearing capacity of surface
footings on sand under general planar loads. Soils Found. 33, No. 3,
6879.
Gottardi, G. & Buttereld, R. (1995). The displacement of a model rigid
surface footing on dense sand under general planar loading. Soils
Found. 35, No. 3, 7182.
Gottardi, G. & Houlsby, G. T. (1995). Model tests of circular footings
on sand subjected to combined loads, OUEL Report No. 2071/95.
Department of Engineering Science, Oxford University.
Gottardi, G., Houlsby, G. T. & Buttereld, R. (1999). The plastic
response of circular footings on sand under general planar loading.
Geotechnique 49, No. 4, 453470.
Hambly, E. C. & Nicholson, B. A. (1991). Jackup dynamic stability
under extreme storm conditions, Proc. 23
rd
Offshore Technology
Conference, Houston, 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, July,
pp. 339358.
Mangal, J. K (1999). Partially drained loading of shallow foundations
on sand. DPhil thesis, Oxford University.
Martin, C. M. (1994). Physical and numerical modelling of off-
shore foundations under combined loads. DPhil thesis, Oxford
University.
Meyerhof, G. G. (1953). The bearing capacity of foundations under
eccentric and inclined loads. Proc. 3rd Int. Conf. Soil Mech. Found.
Engng, Zurich 1, 440445.
Ngo-Tran, C. L. (1996). The analysis of offshore foundations subjected
to combined loading. DPhil thesis, Oxford University.
Nova, R. & Montrasio, L. (1991). Settlements of shallow foundations on
sand. Geotechnique 41, No. 2, 243256.
Oran, C. (1973). Tangent stiffness in plane frames. J. Struct. Engng
Div., ASCE 99, No. ST6, 973985.
Roscoe, K. H. & Schoeld, A. N.(1956). The stability of short pier
foundations in sand. Br. Weld. J., August, 343354.
Schotmann, G. J. M. (1989). The effects of displacements on the
stability of jackup spudcan foundations. Proc. 21st Offshore Technol-
ogy Conf., Houston, OTC 6026
Tan, F. S .C. (1990). Centrifuge and theoretical modelling of conical
footings on sand. PhD thesis, University of Cambridge.
Taylor, P. H., Jonathon, P. & Harland, L. A. (1995). Time domain
simulation of jack-up dynamics with the extremes of a Gaussian
process. Proc.14th Conf. Offshore Mechanics and Arctic Structures
(OMAE), Copenhagen 1-A, 313319.
Tromans, P. S., Anaturk, A. R. & Hagemeijer, P. (1991). A new model
for the kinematics of large ocean waves: applications as a design
wave. Proc. ISOPE-91 Conf. Edinburgh, 3, 6471.
Vesic, A. S. (1973). Analysis of ultimate loads of shallow foundations.
J. Soil Mech. Found. Engng Div., ASCE 99, No. SM1, 4573.
PLASTICITY MODEL FOR FOOTINGS ON SAND 129

Vous aimerez peut-être aussi