You are on page 1of 13

Lee, K. K. et al. (2013). Geotechnique 63, No. 15, 12851297 [http://dx.doi.org/10.1680/geot.12.P.

176]
1285
Bearing capacity on sand overlying clay soils: a simplied conceptual
model
K. K. LEE

, M. F. RANDOLPH and M. J. CASSI DY


The derivation of a new conceptual model for predicting the peak penetration resistance of circular
footings installed vertically into sand overlying soft clay is presented in this paper. Based on
visualisation experiments and nite-element analyses, a failure mechanism of a frustum of sand being
forced into the underlying clay, with the outer angle reecting the dilation in the sand, is assumed.
The analytical basis of the new conceptual model follows the approach for silo analysis, and takes
into account the stress level and dilatant response of the sand. A comprehensive analysis of the
performance of the new model, and other existing analytical methods, in retrospectively predicting the
peak resistance of a database of 49 geotechnical centrifuge tests is provided. Signicant improvements
over existing punching shear and load spread models are shown, as a result of incorporating the
strength properties of the sand in a consistent manner. The new model has application to the offshore
mobile jack-up industry, where inaccurate predictions of peak capacity in layered soils continue to
cause damaging punch-through failures.
KEYWORDS: bearing capacity; footings/foundations; offshore engineering
INTRODUCTION
This paper describes the derivation of a new conceptual
model for predicting the peak penetration resistance of
circular foundations in stratied sediments with sand over-
lying clay. The motivation for this comes principally from
the offshore industry, where unexpected punch-through fail-
ures occur relatively frequently during the installation of
mobile jack-up platforms (Osborne & Paisley, 2002; Osborne
et al., 2006, 2009). Such failures involve one of the spudcan
footings of the jack-up platform uncontrollably punching
locally stronger sand into underlying softer clay material.
Accurate determination of the peak penetration resistance
before this occurs is a critical component in predicting
hazardous installation conditions. The model may also be
applied more broadly to the capacity of shallow circular
footings on layered sand over clay proles.
The traditional analytical methods used to calculate the
peak penetration resistance on sand overlying clay are the
punching shear method as shown in Fig. 1(a), which is
based on Hanna & Meyerhof (1980), and the projected area
method as shown in Fig. 1(b). Both methods have been
proposed as the basis for predicting the peak penetration
resistance, q
peak
, of spudcans in the jack-up industry guide-
lines (SNAME, 2002). In the primary approach, q
peak
is
calculated following the punching shear method, but with
the recommendation that the product of the punching shear
coefcient, K
s
, and tan 9 is related to the shear strength, s
u
,
of the underlying clay layer by
K
s
tan 9 3
s
u
9
s
D
(1)
where D is the equivalent spudcan diameter, and 9
s
is the
Manuscript received 19 November 2012; revised manuscript accepted
29 May 2013. Published online ahead of print 18 July 2013.
Discussion on this paper closes on 1 May 2014, for further details
see p. ii.

Advanced Geomechanics (former PhD student of the University of


Western Australia).
Centre for Offshore Foundation Systems, The University of Western
Australia, Perth, Australia.
Q q D
peak peak
2
( /4)
q
0
H
s
D
z
( ) tan q z s K
0 s s s

Sand: ,
s
Clay: s
u
C
L
s N s q
c c u 0

(a)
Q q D
peak peak
2
( /4)
q
0
H
s
D
Sand: ,
s
Clay: s
u
C
L
s N s q
c c u 0

(b)


Fig. 1. Prediction methods recommended in jack-up industry
guidelines (SNAME, 2002): (a) punching shear method;
(b) projected area method
effective unit weight of the sand. An alternative design
approach is provided in the commentary to the SNAME
guidelines using the projected area method, in which the
load is assumed to be projected through the upper sand layer
to an imaginary footing of increased bearing area at the
sand/clay boundary. There remains debate as to appropriate
values of the assumed angle of spread, with the range 1:5 to
1:3 (horizontal:vertical) recommended in SNAME. By ignor-
ing the properties of the upper sand layer (since even in the
punching shear approach the frictional resistance through the
sand is expressed in terms of the strength of the underlying
clay), both methods provide oversimplied approaches
whereby the ratio of q
peak
to the bearing capacity of the
underlying clay increases as a simple quadratic function of
the ratio of the upper sand thickness to the foundation
diameter, H
s
/D.
Two alternative prediction methods have been developed
more recently by Okamura et al. (1998) and Teh (2007).
Okamura et al. (1998) proposed a new limit equilibrium
method with a failure mechanism that combines the concept
of the projected area method and the punching shear method,
as shown in Fig. 2. To incorporate the inuence of stress
level on friction angle, Okamura et al. (1998) proposed
calculating the friction angle 9 through an iterative proce-
dure between 9 and the initial mean effective stress at the
mid-depth of the sand layer, with the assumption that the
normal stress at the slip surface is at a passive failure state.
Two empirical relationships in which 9 reduced with the
logarithm of mean effective stress were proposed to be used
in the iterative procedure, although these relationships were
developed for the sand used in their model tests. Using this
method, however, the stress-level dependence of friction
angle is not totally accounted for. This is because 9 is
related only to the initial stress level rather than the actual
stress level at failure, which is also inuenced by foundation
size and the strength of the underlying clay.
The method proposed by Teh (2007) is based on observa-
tions of spudcan failure mechanisms from visualisation ex-
periments (Teh et al., 2008), and was calibrated against
centrifuge test results (Teh et al., 2010). A simplied dia-
gram of the model is shown in Fig. 3; the peak penetration
resistance is calculated as the sum of regions of full and
reduced bearing capacity in the underlying clay and frictional
resistance along a shear surface at angle to the vertical
(noting that is not a dilation angle in this method, but a
geometrical parameter used to represent the shear surface
inclination). Teh (2007) provides semi-logarithmic design
charts to calculate and two other model parameters.
A series of 30 geotechnical centrifuge experiments on the
installation of circular at and spudcan foundations in sand
overlying clay was reported in Lee et al. (2013). This
database, obtained from tests conducted on two consistent
samples in a drum centrifuge, provides the most comprehen-
sive set of ratios of sand height to diameter for conceptual
model calibration. Retrospective simulation of this database
using nite-element analysis showed that larger values of
peak resistance gave lower operational friction and dilation
angles, which is consistent with the gradual suppression of
dilatancy under high conning stress. The nite-element
analysis gave failure mechanisms during peak resistance
similar to those observed in the centrifuge visualisation
experiments of Teh et al. (2008), where a frustum of sand
was forced into the underlying clay, with the outer angle
corresponding to the dilation angle of the sand (Lee et al.,
2013).
This mechanism and the database of experimental results
form the basis of the analytical model for bearing capacity
of foundations on sand overlying clay described here. The
model is based on silo analysis, applying force equilibrium
to an idealised frustum of sand rupturing along a conical
surface between the sand/clay interface and the periphery of
the bottom of the footing (Lee et al., 2009). As the key
motivation for this research was to develop a simple mathe-
matical model for use within the offshore jack-up industry,
the model is relatively simple, is straightforward to imple-
ment, and has well-dened input parameters that can be
derived from standard offshore site investigation.
CONCEPTUAL MODEL
Assumed failure mechanism
The model is consistent with the kinematically admissible
failure mechanism that was adopted by Vermeer & Sutjiadi
(1985) for the uplift resistance of anchor plates in sand.
They observed from nite-element computations and model
tests that the failure mechanism involved approximately
straight (in any vertical plane) rupture surfaces, at an
inclination corresponding to the dilation angle of the sand,
from the edge of the anchor to the soil surface. For a
circular anchor plate, the rupture surfaces formed an over-
turned truncated-cone sand block above the anchor.
The conceptual model of this paper assumes a similar
failure mechanism, except that the truncated-cone sand block
is upside down, compared with the anchor problem, and the
base of the conceptual sand frustum is resisted by the
underlying clay. It is a consistent but slightly idealised
Q q D
peak peak
2
( /4)
q
0
z
H
s
c
Sand: ,
s
Clay: s
u
q
clay
A
B

Footing
D
(a)
/s
u
Direction of
slip line
1
0
1
Pole for
circle A


c
A
B

mc u
/s
Pole for
circle B
/s
u
(b)
Fig. 2. Illustration of model of Okamura et al. (1998): (a) assumed
failure mechanism; (b) Mohr-circle of stress for soil elements A
and B (after Okamura et al., 1998)
1286 LEE, RANDOLPH AND CASSIDY
version of the failure mechanisms observed in centrifuge
tests using photogrammetry and particle image velocimetry
(PIV) techniques (Teh et al., 2008). These showed shear
surfaces in homogeneous sand that were trumpet-shaped at
peak penetration resistance, reducing gradually to vertical
shear surfaces during post-peak penetration on sand over-
lying clay. This indicates that the dilation angle is not
constant, even in homogeneous sand, and continues to
reduce with strain during the penetration process. The con-
ceptual model described here simplies this observed behav-
iour by adopting a single operative dilation angle and
friction angle. This is considered acceptable, since the
primary focus is on conditions at peak penetration resistance,
as required in the jack-up industry. The adopted operative
friction and dilation angles are consistent with a modied
set of stressdilatancy relationships from Bolton (1986),
which was obtained from back-analyses through nite-ele-
ment analyses using MohrCoulomb and Tresca models, as
discussed in Lee et al. (2013).
Figure 4 shows the kinematically admissible axisymmetric
failure mechanism for the conceptual model, where q
peak
is
the peak penetration resistance; q
0
is the surcharge at the
sand surface; q
b
is the bearing capacity of the underlying
clay; and 9
n
are the frictional resistance and normal
effective stress along the assumed slip surface; 9, and 9
s
are the friction angle, dilation angle and effective unit
weight of the upper sand layer respectively; s
u0
and r are
the undrained strength at the sand/clay interface and the
strength gradient of the underlying clay respectively; D is
the foundation diameter; and H
s
is the sand thickness.
COMPONENTS OF THE MODEL
Bearing capacity of the underlying clay
The resistance from the underlying clay is assumed to be
the full bearing capacity of the projected footing at the base
of the sand frustum. Classical bearing capacity theory is
used, with the behaviour assumed to be undrained under the
typical spudcan loading rate. The bearing capacity of the
underlying clay, q
b
, is calculated as
q
b
N
c0
s
u0
q
0
9
s
H
s
(2)
The bearing capacity factor, N
c0
, is given as a function of
the non-dimensional parameter k, dened as
k
r D 2H
s
tan
s
u0
(3)
The N
c0
values are adopted from Martin (2001) and
Houlsby & Martin (2003), who provided bearing capacity
factors for at-based rigid circular foundations on a weight-
less isotropic Tresca soil with strength increasing linearly
with depth. Houlsby & Martin (2003) also proposed a linear
curve-t for N
c0
, expressed as
D
H
diameter
sand thickness

s
Sand
Assumed straight conical
slip surface for simplicity
Trumpet-shaped slip
surface observed
from tests
Clay
R
r
Full
bearing
capacity
( ) q
clay
Linearly reduced bearing
capacity (from to 05 ) q q
clay clay

2

C
L
Q q D
peak peak
2
( /4)

d H 012
s
h H 088
s
Fig. 3. Schematic diagram of model of Teh (2007)
Q q D
peak peak
2
( /4)
q
0
r
R D/2
z
H
s
C
L
dz
R ztan
(Thin horizontal discs)
q
b
Clay: , s
u0

Sand: , ,
s
Fig. 4. Series of thin horizontal discs in the conceptual sand
frustum
BEARING CAPACITY ON SAND OVERLYING CLAY SOILS: A SIMPLIFIED CONCEPTUAL MODEL 1287
N
c0
6
:
34 0
:
56k (4)
This expression is accurate to within 5% of the tabulated
values in the range 0 , k , 10 (Houlsby & Martin, 2003),
and is adopted here.
Resistance of the slip surfaces
Analytical treatment of equilibrium within the sand fol-
lows the method of differential slices, as used for predict-
ing the forces on the walls of silos and hoppers (e.g.
Walters, 1973). This approach was also used by Randolph et
al. (1991) to analyse the soil plug capacity in open-ended
piles under various loading conditions.
In the punch-through problem of sand overlying clay, the
sand frustum is treated as a series of thin horizontal discs,
as shown in Fig. 4. Cylindrical coordinates are used, with
the origin at the centre base of the foundation, r the radial
coordinate, and z the vertical coordinate with positive direc-
tion downwards.
Normal and shear stresses at the slip surfaces. Drescher &
Detournay (1993) showed that for kinematic mechanisms
composed of rigid blocks and velocity discontinuities, the use
of a reduced friction angle,

, instead of 9 furnishes
reasonable limit load estimates for materials with coaxial
non-associated ow rules. They suggested that

be viewed
as a reduced friction angle that accounts for the softening that
may be induced under certain stress conditions by a non-
associated ow rule.
Davis (1968) and Drescher & Detournay (1993) gave the
equation for the reduced friction angle

as
tan


sin 9 cos
1 sin 9 sin
(5)
noting that this restricts

to sin 9 , tan

, tan 9,
depending on the dilatancy angle. The lower bound is
obtained for a non-dilatant material ( 0), and the upper
bound corresponds to an associated ow rule where 9.
Consequently, the shear stress at the slip surface for a
frictional material can be written as
9
n
tan

(6)
Equation (6) relates the normal and shear stresses along
the slip surface. However, just as for analogous problems
such as calculating shaft friction of piles, it is difcult to
estimate the normal effective stress along the slip surface.
The following section presents a simple method to overcome
this by relating 9
n
along the slip surface to the average
vertical effective stress within the sand frustum through an
empirical factor.
Vertical effective stress within the sand frustum. For a rigid
foundation on sand overlying clay, the vertical effective
stresses beneath the foundation vary with radial position
(even at the same depth). This can be conrmed by
inspecting a case of zero dilation in which the vertical
effective stress at the vertical slip surface is not a principal
stress, whereas the vertical effective stress along the centre-
line of the foundation is a principal stress from symmetry.
Evaluation of the distribution function of vertical effective
stress at a particular depth beneath the foundation is extre-
mely difcult. For convenience, a mean vertical effective
stress, 9
z
, is dened as
9
z

1
R z tan
2
_
9
z
dA (7)
where R is the radius and A is the area of the innitesimal
disc element at depth z. The mean vertical stress, 9
z
, is the
average vertical effective stress acting on the horizontal
surface of the disc element at that depth.
The normal effective stress at the slip surface, 9
n
, can
then be determined by introducing a distribution factor, D
F
,
dened as the ratio of 9
n
to the mean vertical effective
stress
D
F

9
n
9
z
(8)
Using the distribution factor D
F
, the normal stress 9
n
and
shear stress acting along the slip surfaces can be written in
terms of the mean vertical effective stress, 9
z
, in the sand
frustum. Hence equation (6) can be rewritten as
9
n
tan

D
F
9
z
tan

(9)
Vertical force equilibrium on an innitesimal element. Figure
5 shows the stresses on one of the innitesimal disc elements
within the conceptual sand frustum, where each element is
acted on by mean vertical stresses 9
z
at the top, 9
z
d 9
z
at
the bottom, and normal and shear stresses 9
n
and at the
side, which is inclined at the operative dilation angle. The
effective unit weight of the horizontal disc element is 9
s
:
The vertical force equilibrium equation on the innitesi-
mal element is
R z tan
2
9
z
R z tan
2
dz 9
s
R z dz tan
_
2
9
z
R z tan
2
d 9
z

2 R z tan cos 9
n
sin
dz
cos
(10)
Substituting for the normal stress 9
n
and shear stress ,
this can be simplied to the differential equation
d 9
z
dz

2 tan D
F
tan

tan 9
z
R z tan
9
s
0 (11)

z
ds

dz

s

z
d
z
Fig. 5. Stresses on an innitesimal disc element within the
conceptual sand frustum
1288 LEE, RANDOLPH AND CASSIDY
A new parameter E, dened below, is introduced, allowing
this differential equation to be simplied to
d 9
z
dz

E tan 9
z
R z tan
9
s
0 (12)
where
E 2 1 D
F
tan

tan
1
_ _ _ _
Integration of vertical force equilibrium. In reality, the
distribution factor D
F
is likely to vary with depth, but it
would be difcult to derive an accurate function for D
F
for
different combinations of relative soil strengths and ratios of
sand thickness over foundation diameter, H
s
/D. Therefore, in
order to integrate equation (12) and obtain a simple equation
for practical design, it is assumed that D
F
is constant. Since
there is no rigorous theoretical treatment to determine D
F
, its
values have been derived by calibrating the nal equation
with the experimental results detailed in Lee et al. (2013), as
described later.
Assuming D
F
is constant with depth, equation (12) can be
integrated to give
R z tan
E
9
z

9
s
R z tan
E1
E 1 tan
A (13)
where A is a constant to be determined from the boundary
condition.
Referring to Fig. 4, when z equals the sand thickness H
s
,
the mean vertical effective stress 9
z
equals the bearing
capacity of the clay (q
b
N
c0
s
u0
q
0
9
s
H
s
), and hence A
can be evaluated as
A R H
s
tan
E
3 N
c0
s
u0
q
0
9
s
H
s

9
s
R H
s
tan
E 1 tan
_ _
(14)
By re-substituting A into equation (13), the nal equation
for the mean vertical effective stress becomes
9
z

9
s
R z tan
E 1 tan

R H
s
tan
E
R z tan
E
3 N
c0
s
u0
q
0
9
s
H
s

9
s
R H
s
tan
E 1 tan
_ _
(15)
The centrifuge test results of Lee et al. (2013)
showed that the peak penetration resistance occurred
very near the sand surface, at penetrations less than a
prototype depth of 1 m (for prototype foundation dia-
meters ranging from 6 to 16 m). This slight indentation
of the foundation into the sand has an insignicant effect
on the calculated peak resistance. Therefore the peak
penetration resistance, q
peak
, can be obtained by setting
z 0. Replacing the radius, R, by D/2, q
peak
may be
expressed as
q
peak
N
c0
s
u0
q
0
1
2H
s
D
tan
_ _
E

9
s
D
2 E 1 tan
3 1 1
2H
s
D
Etan
_ _
1
2H
s
D
tan
_ _
E
_ _
< q
sand
(16)
where
E 2 1 D
F
tan

tan
1
_ _ _ _
The rst term for q
peak
consists of the contribution from
both the bearing capacity of the underlying clay and the
frictional resistance along the slip surfaces. The second term
is due to the effective weight of the sand frustum. The
inequality is to ensure that the value of q
peak
does not
exceed the bearing capacity of the circular foundation in the
sand alone, q
sand
: Methods for calculating the bearing cap-
acity of circular footings in a pure sand layer are described
in, among others, Brinch Hansen (1970), Cassidy & Houlsby
(1999, 2002), Randolph et al. (2004) and Lee (2009).
Design equation for zero dilation angle. Because the param-
eter E in equation (16) becomes innite when the dilation
angle equals zero, a separate expression is required for this
case. This is obtained either by taking the limit of equation
(16) as tends to zero, or by integrating equation (11), again
with tan 0 (Lee, 2009). Both approaches lead to
q
peak
N
c0
s
u0
q
0
e
E0
9
s
H
s
e
E0
1
1
E
0
_ _

1
E
0
_ _
< q
sand
(17)
where
E
0
4D
F
sin
cv
H
s
D
The rst term above indicates that, for zero dilation, q
peak
increases exponentially with the sand thickness H
s
(or the
ratio H
s
/D).
PARAMETER VALUES IN THE CONCEPTUAL MODEL
Input parameters
Table 1 summarises all input parameters for the new
model, and some remarks on the methods of determining
them. The input parameters are familiar quantities, and can
be determined from a proper site investigation scheme. The
conceptual model essentially involves one equation (equation
(16)) and requires only input of the necessary geometry and
soil strength parameters to calculate the peak penetration
resistance, q
peak
, of a foundation on sand overlying clay.
Strength parameters of sand
Although part of equation (16), the friction and dilation
angles cannot be used directly as input parameters to
calculate q
peak
: This is because they depend on the stress
level, which will vary for different combinations of sand
thickness and foundation diameter, as well as the strength
properties of the sand and clay. However, the stress-level
dependence of the friction angle can be incorporated easily
BEARING CAPACITY ON SAND OVERLYING CLAY SOILS: A SIMPLIFIED CONCEPTUAL MODEL 1289
through an iterative procedure. It was shown in the nite-
element results of Lee et al. (2013) that 9 and are
related to q
peak
through a modied set of strengthdilatancy
relationships from Bolton (1986), expressed as
I
R
I
D
Q ln q
peak
_ _
1, 0 , I
R
, 4 (18)
9
cv
2
:
65I
R
(19)
0
:
8 9
cv
, > 0 (20)
Starting from the in situ relative density I
D
, the critical
state friction angle
cv
and quantity Q (natural logarithm of
the grain-crushing strength expressed in kPa, with default
value 10), an iterative calculation procedure can be per-
formed to determine q
peak
: Convergence is rapid, starting
with an assumed value for , such as 18, evaluating 9 from
equation (20), then applying equations (16), (18), (19) and
(20) in sequence to nd new estimates of 9 and until
satisfactory convergence is achieved.
With this approach, the appropriate operative friction
angle and dilation angle (and also the peak penetration
resistance, q
peak
) can be determined using the fundamental
parameters of the sand. This iterative procedure is important,
because it allows the stress level and dilatant response of the
sand to be taken into account when calculating the value of
q
peak
: In practice, a curved trumpet shape may be antici-
pated for the frustum, with the outer angle (in a vertical
plane) increasing through the sand layer because of the
effects of: (a) slight strain-softening at shallow depths as
the rupture advances from the foundation base; and (b) the
gradual decrease in stress level with depth. The deduced
friction and dilation angles, taking the outer angle of the
frustum as constant, therefore represent average conditions,
allowing the silo calculation model described above to be
applied simply.
Determination of distribution factor D
F
Values of D
F
were determined from back-calculation of
the 25 at circular and ve spudcan centrifuge test results
presented in Lee et al. (2013), noting that this is the only
parameter in the new conceptual model to be determined
from experimental results. The D
F
values required to match
q
peak
are shown in Fig. 6 with the empirical relationships of
D
F
0
:
726 0
:
219
H
s
D
, H
s
=D , 1
:
12 (21)
for at circular footings and
D
F
1
:
333 0
:
889
H
s
D
, H
s
=D , 0
:
9 (22)
for spudcan-shaped footings. By adopting the distribution
factor D
F
from equations (21) and (22), an excellent match
between the calculated q
peak
values and the experimental
results is obtained. This is shown in Fig. 7 for the spudcan-
Table 1. Parameters for the new conceptual model equation
Category Symbol Parameter Method of determining parameter
Geometry D Footing diameter Known parameter
H
s
Sand thickness Measured from in situ tests
Sand I
D
and 9
s
Relative density and effective unit weight Measured from in situ or laboratory tests

cv
Critical state friction angle Measured from laboratory tests, or estimated based on mineralogy
(Randolph et al., 2004)
Q Parameter in Bolton (1986) equation Measured from laboratory tests, or estimated based on mineralogy
(Randolph et al., 2004)
I
R
Relative dilatancy index Obtained from iterative calculation between equation (16) (or
equation (17) for 0) and equations (18)(20)
9 and Friction angle and dilation angle Obtained from equations (19) and (20): 9
cv
2
:
65I
R
and
0
:
8 9
cv
(9,
cv
and in degrees) (Lee et al., 2013)

Reduced friction angle due to non-associated


ow rule
tan

sin 9 cos =(1 sin 9 sin ) (Davis, 1968, and Drescher


& Detournay, 1993)
Clay s
u0
Undrained shear strength at sand/clay interface Measured from in situ or laboratory tests
r and k Represent strength prole of the clay: r is
gradient of undrained strength with depth, and
k is non-dimensional parameter for
determining bearing capacity factor
r can be inferred from site investigation, and k is determined from
k
r D 2H
s
tan
s
u0
N
c0
Bearing capacity factor (taking account of
linear strength increment)
N
c0
6
:
34 0
:
56k (Martin, 2001, and Houlsby & Martin, 2003)
Empirical
factor
D
F
Distribution factor
Dened as ratio of normal effective stress at
slip surface to mean vertical effective stress
within sand layer
For at-based circular footing, use equation (21)
D
F
0
:
726 0
:
219
H
s
D
, H
s
/D , 1
.
12
For spudcans with conical base inclined at 138 to horizontal, use
equation (22)
D
F
1
:
333 0
:
889
H
s
D
, H
s
/D , 0
.
9
0
02
04
06
08
10
12
0 02 04 06 08 10 12
D
i
s
t
r
i
b
u
t
i
o
n

f
a
c
t
o
r
,
D
F
H D
s
/
13
Spudcan
Flat foundation
D H D
H D
F s
s
1333 0889 / ,
/ 09

D H D H D
F s s
0726 0219 / , / 112
Fig. 6. Back-calculated distribution factors for spudcan and at
foundations
1290 LEE, RANDOLPH AND CASSIDY
shaped and at-based circular foundations. The differences
between the calculated q
peak
values and the test results tend
to be less than 5% (at) and 8% (spudcan), except for two
outlying cases at around 15%.
It is observed that the spudcan, with a conical base
inclined at 138 and a small spigot, has higher empirical D
F
values than the at-based foundation at similar H
s
/D ratios.
The schematic diagram in Fig. 8 is provided to explain this
observation. Equations (18) and (19) show that the operative
dilation angles are less than 138, and hence the slip surface
of the sand frustum is close to vertical, with an inclination
of less than 1H:4
.
2V. This implies that the normal stress
approximates the horizontal effective stress at the slip sur-
face. For a similar H
s
/D ratio, a conical foundation invokes
higher lateral stress than does a at-based foundation, and
hence higher resistance along the slip surface. Therefore
higher D
F
values are required for conical foundations in the
new conceptual model.
It is also observed that the empirical D
F
function for
conical foundations reduces more rapidly with increasing
H
s
/D than for at foundations. This may be due to a higher
degree of progressive failure caused by the conical founda-
tion, since the conical base will pre-shear the sand before
peak resistance occurs. Although the simplied conceptual
model does not consider strain-softening and progressive
failure, the phenomenon is captured and reected in the
back-calculated values of D
F
:
It is hypothesised that all foundations with inclined bases
would have their empirical lines of D
F
above the line of at-
based foundations in Fig. 6. Specically, the D
F
lines for
conical foundations with inclined base angle less than 138 to
the horizontal are hypothesised to lie between the at-based
q
p
e
a
k
f
r
o
m

c
o
n
c
e
p
t
u
a
l

m
o
d
e
l
:

k
P
a
800 600 400 200
0
200
400
600
800
0
q
peak
from centrifuge tests: kPa
D1: Spudcans, 62 m sand
D2: Flat-based, 67 m sand
D1: Flat-based, 62 m sand
D2: Flat-based, 58 m sand
D2: Flat-based, 48 m sand
D1: Flat-based, 41 m sand
D2: Flat-based, 34 m sand
Line of equality
20% variation
10% variation
Fig. 7. Comparison of calculated and experimental q
peak
values for spudcan-shaped and at-
based circular foundations
Sand
Clay
(Invokes higher
lateral stress)
Fig. 8. Schematic diagram of at-based and conical foundations on sand overlying clay
BEARING CAPACITY ON SAND OVERLYING CLAY SOILS: A SIMPLIFIED CONCEPTUAL MODEL 1291
equation (21) and the spudcan equation (22), whereas those
with inclined base more than 138 to the horizontal would
have higher and steeper D
F
lines than equation (22).
The values of D
F
are less than unity, which indicates that
the normal effective stress at the slip surface, 9
n
, is lower
than the mean vertical effective stress, 9
z
: There is no
physical restriction that D
F
must stay below unity. Assuming
extrapolation of the experimental results is valid, the empiri-
cal D
F
functions in Fig. 6 suggest that D
F
may have values
larger than unity for conical foundations on sand overlying
clay with very low H
s
/D ratios. This could be possible, as
large lateral stresses could be incurred while the degree of
progressive failure is small.
The relationships of equations (21) and (22) are notion-
ally valid within the range of H
s
/D tested experimentally
and from which D
F
was back-calculated, with the at
foundations veried for 0
.
21 < H
s
/D < 1
.
12 and the spud-
can shape for 0
.
39 < H
s
/D < 0
.
78 (Lee et al., 2013).
However, it is considered appropriate to extrapolate the
empirical D
F
lines of equations (21) and (22) to cover all
values of H
s
/D < 1
.
12. Further, as both empirical lines
intersect at H
s
/D 0
.
9, equation (22) for a spudcan is
deemed valid for 0 , H
s
/D < 0
.
9. In the range 0
.
9 ,
H
s
/D < 1
.
12, it is suggested that equation (21) for a at-
based foundation may be taken as the most appropriate
for a spudcan.
NUMERICAL PREDICTION OF CENTRIFUGE RESULTS
The accuracy of the q
peak
values calculated from the new
method, as well as those of SNAME (2002), Teh (2007) and
Okamura et al. (1998), has been evaluated through
retrospective simulation of centrifuge experiments with
H
s
/D < 1
.
12. Centrifuge results additional to those used in
the model calibration were included in this comparative
study. These consisted of ve model spudcan penetration
tests carried out by the rst author using the beam centri-
fuges at the University of Western Australia, two centrifuge
model tests from Craig & Chua (1990), and 12 from Teh
(2007). The experimental details of these additional tests, as
interpreted by the authors, are provided in Table 2. Including
the 30 tests described in Lee et al. (2013), a total of 49
centrifuge tests are used.
The comparisons are shown in Figs 914, where they are
categorised as the industry methods of SNAME (three alter-
native approaches) and the recent methods of Okamura et
al., Teh and this model respectively. Figs 9(a)14(a) plot the
q
peak
values calculated from each method against the experi-
mental values, where the diagonal line is the line of equality.
Additionally, Figs 9(b)14(b) present the ratios of calculated
to measured q
peak
values against the H
s
/D ratio. A linear
regression line for the data points is also shown in each of
these graphs to highlight whether the average ratio of
calculated to measured q
peak
values exhibits any trend with
H
s
/D.
Some common attributes of the datasets are observed. In
general, all analytical methods consistently provide higher
predicted q
peak
values for the tests carried out by Craig &
Chua (1990) than for the other test data. Figs 9(b)13(b)
show that the ratios of calculated to measured q
peak
values
(q
peak,calculated
/q
peak,measured
) for both datasets from Teh (2007),
namely the NUS test series and the UWA test series, are
generally below the linear regression line. Two tests from
the NUS series are exceptions. The tests at H
s
/D 1 and
1
.
05 of the NUS series show particularly high predicted
values. In comparison, the new conceptual model shown in
Fig. 14(b) gives good average predictions for both sets of
tests.
The large testing area of the drum centrifuge allowed ve
pairs of tests to be conducted on the same sample and under
the same conditions (Lee et al., 2013). Each pair comprised
a circular at-based and a spudcan-shaped foundation of the
same diameter. None of the other published methods can
distinguish between these tests, as highlighted by consis-
Table 2. Experimental details of additional tests used in comparative study
Authors Footing Name of test

Geometry
y
Sand Clay Results
q
peak
:
H
s
: m D: m H
s
/D
ratio

cv
:
degrees
I
D
9
s
:
kN/m
3
s
u0
:
{
kPa
r: kPa/m kPa
Teh (2007) Spudcan NUS_F1 3 10 0
.
30 32 0
.
95 9
.
90 7
.
8 1
.
56 155
NUS_F2 5 10 0
.
50 32 0
.
88 9
.
80 12
.
7 1
.
56 300
NUS_F3 7 10 0
.
70 32 0
.
94 9
.
90 18
.
0 1
.
56 559
NUS_F4 7
.
7 10 0
.
77 32 0
.
935 9
.
90 19
.
8 1
.
56 626
NUS_F5 10 10 1
.
00 32 0
.
95 9
.
90 25
.
8 1
.
56 700
NUS_F6 10
.
5 10 1
.
05 32 0
.
96 9
.
90 27
.
2 1
.
56 758
NUS_F8 5 10 0
.
50 32 0
.
61 9
.
20 12
.
0 1
.
56 265
NUS_F9 7 10 0
.
70 32 0
.
58 9
.
20 16
.
7 1
.
56 522
Teh (2007) Spudcan UWA_F3 3
.
5 4 0
.
88 31 0
.
99 11
.
15 7
.
2 1
.
20 371
UWA_F4 3
.
5 6 0
.
58 31 0
.
99 11
.
15 7
.
2 1
.
20 270
UWA_F9 7
.
1 7 1
.
01 31 0
.
98 11
.
13 14
.
6 1
.
20 858
UWA_F10 7
.
1 8 0
.
89 31 0
.
98 11
.
13 14
.
6 1
.
20 608
Craig & Chua
(1990)
Flat-based
foundation
T6 7 14 0
.
50 32 0
.
89 10
.
30 42 0 543
T8 9
.
5 14 0
.
68 32 0
.
89 10
.
30 41 0 638
(approximate)
This study
(beam
Spudcan B1S7SP8a 7 8 0
.
88 31 0
.
99 11
.
15 13
.
2 1
.
85 622
B1S7SP8b 7 8 0
.
88 31 0
.
99 11
.
15 13
.
2 1
.
85 487
centrifuge) B1S7SP8c 7 8 0
.
88 31 0
.
99 11
.
15 13
.
2 1
.
85 559
B1S7SP14a 7 14 0
.
50 31 0
.
99 11
.
15 13
.
2 1
.
85 482
B1S7SP14b 7 14 0
.
50 31 0
.
99 11
.
15 13
.
2 1
.
85 421

NUS implies tests performed in the beam centrifuge at the National University of Singapore, whereas UWA are tests performed in the
centrifuge facilities at the University of Western Australia.
y
Prototype scale.
{
Undrained shear strength of clay at the sand/clay interface.
1292 LEE, RANDOLPH AND CASSIDY
Teh (2007) NUS tests
Craig & Chua (1990)
Lee . (2013) (spudcan) et al
Teh (2007) UWA tests
This study (beam centrifuge)
Lee . (2013) (footing) et al
1000 800 600 400 200
12 10 08 06 04 02
0
200
400
600
800
1000
0
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
:

k
P
a
q
peak,measured
: kPa
(a)
0
02
04
06
08
10
12
14
16
18
20
0
q
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
/
p
e
a
k
,
m
e
a
s
u
r
e
d
H D
s
/
(b)
Linear regression line
Fig. 9. Comparison of calculated and measured q
peak
with
SNAME (2002) using lower-bound value of K
s
tan 9
Teh (2007) NUS tests
Craig & Chua (1990)
Lee . (2013) (spudcan) et al
Teh (2007) UWA tests
This study (beam centrifuge)
Lee . (2013) (footing) et al
0
200
400
600
800
1000
0 200 400 600 800 1000
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
:

k
P
a
q
peak,measured
: kPa
(a)
0
02
04
06
08
10
12
14
16
18
20
0 02 04 06 08 10 12
q
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
p
e
a
k
,
m
e
a
s
u
r
e
d
/
H D
s
/
(b)
Linear regression line
Fig. 10. Comparison of calculated and measured q
peak
with
SNAME (2002) commentaries, based on spreading ratio 1H:5V
Teh (2007) NUS tests
Craig & Chua (1990)
Lee . (2013) (spudcan) et al
Teh (2007) UWA tests
This study (beam centrifuge)
Lee . (2013) (footing) et al
0
200
400
600
800
1000
0 200 400 600 800 1000
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
:

k
P
a
q
peak,measured
: kPa
(a)
0
02
04
06
08
10
12
14
16
18
20
0 02 04 06 08 10 12
q
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
p
e
a
k
,
m
e
a
s
u
r
e
d
/
H D
s
/
(b)
Linear regression line
Fig. 11. Comparison of calculated and measured q
peak
with
SNAME (2002) commentaries, based on spreading ratio 1H:3V
0
400
800
1200
1600
0 400 800 1200 1600
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
:

k
P
a
q
peak,measured
: kPa
(a)
0
02
04
06
08
10
12
14
16
18
20
0 02 04 06 08 10 12
q
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
p
e
a
k
,
m
e
a
s
u
r
e
d
/
H D
s
/
(b)
Teh (2007) NUS tests Teh (2007) UWA tests
Craig & Chua (1990) This study (beam centrifuge)
Lee . (2013) (spudcan) et al Lee . (2013) (footing) et al
Linear regression line
Fig. 12. Comparison of calculated and measured q
peak
with
method of Okamura et al. (1998)
BEARING CAPACITY ON SAND OVERLYING CLAY SOILS: A SIMPLIFIED CONCEPTUAL MODEL 1293
tently lower predictions of q
peak
for the spudcan-shaped
dataset than for the at foundations. The new approach can
so distinguish, as shown in Fig. 14, as a result of varying
the distribution factor D
F
with footing shape.
Following these observations on general trends of the test
data, the specic performance of each analytical method will
now be discussed. To assist this evaluation, the performance
of the various methods in predicting the q
peak
values is
summarised in Table 3, with statistical parameters for the
mean and standard deviation () provided. In addition, a
skew angle of the linear regression line, , is also provided
to represent the skewness of the linear regression line
of q
peak,calculated
/q
peak,measured
to the horizontal axis of H
s
/D,
taking as the arctangent of the gradient of the linear
regression line.
Numerical predictions using the SNAME-recommended
practice
As stated in SNAME (2002), the methods given in the
guidelines and commentary provide a conservative prediction
of q
peak
: This is evident in the comparison graphs of Figs
911, where the experimental values are signicantly under-
estimated. Some of the predicted q
peak
values are only about
a third of the test results using the lower-bound value of
K
s
tan 9 in the primary method from SNAME (2002), or
with the conservative load spreading of 1H:5V suggested in
the SNAME (2002) commentary. Even with the slightly
larger spreading ratio of 1H:3V, the calculated values are
only about half the experimental values. These results high-
light the oversimplied nature of these models, which do not
account explicitly for the strength of the sand.
Numerical predictions using the method of Okamura et al.
(1998)
The method of Okamura et al. (1998) generally over-
estimates the q
peak
values, as shown in Fig. 12. The method
relies on an angle,
c
, projected from the base edge of the
footing to the clay surface. For the retrospective predictions
this has been calculated to be between 158 and 188, values
causing too large a projected area at the sand/clay interface.
This causes additional bearing capacity of the clay to be
calculated, and therefore overestimation of q
peak
: The over-
estimation becomes more apparent for larger H
s
/D ratios.
This reinforces the consequence of excessive projection
angles being calculated using this method, because the error
of mobilising additional bearing capacity of the clay is
exacerbated in thicker sand layers.
The Okamura method was developed based on centrifuge
tests with prototype foundation diameters of 0
.
83 m
(Okamura et al., 1997), and with relatively thin sand layers.
As the projected area is a function of the square of founda-
tion diameter and sand thickness, the error due to excessive
projection angle is less signicant for the relatively small
foundation diameters and sand thicknesses tested by Oka-
mura et al. This method may prove suitable for onshore
footings, which generally have much smaller dimensions.
Numerical predictions using the method of Teh (2007)
Figure 13 indicates that the method of Teh (2007) gives a
relatively good average prediction of the peak penetration
resistance, although Fig. 13(b) and Table 3 indicate that the
general trend of calculated to measured q
peak
ratios is
skewed.
The method involves three empirical parameters that need
to be determined from separate, semi-logarithmic graphs.
Although the schematic diagram of the model is presented
0
200
400
600
800
1000
1200
0 200 400 600 800 1000 1200
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
:

k
P
a
q
peak,measured
: kPa
(a)
0
02
04
06
08
10
12
14
16
18
20
0 02 04 06 08 10 12
q
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
p
e
a
k
,
m
e
a
s
u
r
e
d
/
H D
s
/
(b)
Teh (2007) NUS tests Teh (2007) UWA tests
Craig & Chua (1990) This study (beam centrifuge)
Lee . (2013) (spudcan) et al Lee . (2013) (footing) et al
Linear regression line
Fig. 13. Comparison of calculated and measured q
peak
with
method of Teh (2007)
0
200
400
600
800
1000
0 200 400 600 800 1000
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
:

k
P
a
q
peak,measured
: kPa
(a)
0
02
04
06
08
10
12
14
16
18
20
0 02 04 06 08 10 12
q
q
p
e
a
k
,
c
a
l
c
u
l
a
t
e
d
p
e
a
k
,
m
e
a
s
u
r
e
d
/
H D
s
/
(b)
Teh (2007) NUS tests Teh (2007) UWA tests
Craig & Chua (1990) This study (beam centrifuge)
Lee . (2013) (spudcan) et al Lee . (2013) (footing) et al
Linear regression line
Fig. 14. Comparison of calculated and measured q
peak
with the
new conceptual model (this study)
1294 LEE, RANDOLPH AND CASSIDY
in Fig. 3, reference to Teh (2007) is required for full details
of the method. However, in brief, the inclination of the slip
surface is determined by parameter , whereas the extent of
the mobilisation of the underlying clay bearing capacity is
determined by parameter
a
: The values of the calculated
friction angle,
2
, and the empirical parameters and
a
obtained from this method are provided in Table 4.
Although it is difcult to evaluate the inuence of each
parameter in predicting the value of q
peak
, as all three
empirical parameters are related to the ratios of H
s
/D and
q
clay
/q
sand
, a brief discussion of the calculated model param-
eters and assumptions is provided here.
(a) The values of are generally much lower than the values
of
a
, producing a geometrical inconsistency between the
failure mechanisms in the sand layer and those in the
underlying clay.
(b) The operative friction angle
2
is taken as the average of
the critical state friction angle
cv
and the friction angle
calculated through an iterative procedure using Boltons
(1986) empirical relationships, with the assumptions that
the horizontal stress is at a passive failure state, and the
initial mean effective stress at the mid-depth of the sand
layer is taken as the representative stress. With this
approach, the effects of foundation size and undrained
shear strength are excluded when incorporating the effect
of stress level on friction angle.
(c) Table 4 shows that the values of
2
are higher for denser
sand layers, and this is as expected. It is generally
recognised that denser sand with a higher friction angle is
more dilative, and will cause larger dispersion angle of
the slip surface (e.g. Vermeer & Sutjiadi, 1985). How-
ever, the values of shown in Table 4 suggest a
contradictory trend. The lower limit of 28 for very
dense sand samples, with relative density of
0
.
88 , I
D
, 0
.
99 and calculated friction angle of
378 ,
2
, 398, is lower than the values 128 and
68 for tests NUS_F8 and NUS_F9, which had a looser
sand layer, with I
D
of around 0
.
6 and a calculated friction
angle between 358 and 368.
(d) Furthermore, the friction angles for tests on very dense
sand are in a relatively small range of 37398, but the
dispersion angles of the slip surface cover a much larger
range of 28 , , 268. This suggests that the inclination
of the slip surface is independent of the friction angle.
This contradicts the results of the numerical analysis
described in Lee et al. (2013), which indicates that the
slip surface inclination equals the dilation angle, which is
directly related to the operative friction angle.
In summary, it appears that the individual parameters are
not consistent with each other. However, compensating
effects among the empirical parameters have worked collec-
tively to produce reasonable estimated q
peak
values, albeit
with a signicant bias with H
s
/D.
Numerical predictions using the conceptual model of this
paper
Figure 14 shows that the new method gives a generally
good prediction of the experimental values of q
peak
:
Although it shows a slight bias with H
s
/D ratio, the linear
regression line in Fig. 14(b) falls close to unity.
The calculated values of friction angle 9 and dilation
angle are listed in Table 5, noting that the projection
angle of the conceptual sand frustum is equal to the
dilation angle (Fig. 4). The friction angles for the very
dense sand samples are in the range 37428, and the
dilation angles are between 68 and 128, following the
relationship proposed in equations (18)(20). Lower values
of 9 and are obtained for the looser sand tests NUS_F8
and NUS_F9 accordingly. This shows that the calculated
values of 9 and are consistent with the relative density
of the sand, and fall in a reasonable range.
EXAMPLE PARAMETRIC STUDY
One of the advantages of the proposed conceptual model
is the ease of calculation. To illustrate how the method may
be used to create parametric design charts, an example of a
at circular footing penetrating into sand overlying clay soils
is shown here. The sand layer is assumed to have properties

cv
318 and 9
s
8 kN/m
3
: The diameter of the footing is
taken as 15 m. The values of q
peak
were then calculated for
various sand thicknesses and relative densities, for two
different normally consolidated shear strength proles, of
Table 3. Summary of comparisons from various analytical methods
Method Value of q
peak,calculated
/q
peak,measured
Minimum Maximum Mean : degrees
SNAME (2002): punching shear method with lower-bound value of
K
s
tan 9
0
.
22 0
.
61 0
.
38 0
.
10 11
Commentaries to SNAME (2002): projected area method with ratio
1H:5V
0
.
23 0
.
71 0
.
45 0
.
12 17
Commentaries to SNAME (2002): projected area method with ratio
1H:3V
0
.
33 0
.
87 0
.
57 0
.
13 12
Okamura et al. (1998) 0
.
83 1
.
95 1
.
24 0
.
24 38
Teh (2007) 0
.
62 1
.
40 0
.
92 0
.
16 24
New conceptual model of this paper 0
.
67 1
.
33 0
.
99 0
.
12 2
Table 4. Values of
2
, and
a
from method of Teh (2007)
Test name I
D
Calculated
2
: degrees Empirical : degrees Empirical
a
: degrees
47 tests with very dense sand layer 0
.
880
.
99 36
.
839
.
1 2
.
026
.
4 19
.
624
.
8
NUS_F8 0
.
61 35
.
8 12
.
2 23
.
8
NUS_F9 0
.
58 35
.
3 5
.
7 22
.
4
BEARING CAPACITY ON SAND OVERLYING CLAY SOILS: A SIMPLIFIED CONCEPTUAL MODEL 1295
increasing strength with depth of 1
.
5 kPa/m and 2
.
5 kPa/m,
for the underlying clay. The results are summarised in Fig.
15. Both graphs present the normalised peak resistance,
q
peak
/N
c0
s
u0
, as a function of relative sand thickness, H
s
/D,
for ve different relative densities, from I
D
0
.
2 to I
D
1.
The graphs show that when the sand thickness becomes
zero (H
s
0; clay-only prole), the values of q
peak
converge
to the bearing capacity of the underlying clay. As H
s
/D
increases, the bearing capacity of the underlying clay is
enhanced signicantly, up to a factor of about 3
.
35
.
5
relative to the bearing capacity of the underlying clay at H
s
/
D 1, depending on the relative density of the sand layer
and the absolute strength of the clay.
Figure 15 also reects the effects of incorporating the
stress level and dilatant response of sand in the model. The
enhancement of bearing capacity for the lower increasing
strength with depth ratio is greater. This is because stronger
clay in the latter case invokes higher stresses in the sand
layer (and higher q
peak
values). This leads to greater suppres-
sion of dilation, and a lower operative friction angle at peak
resistance. The stress level will also be affected in a similar
manner by the absolute value of footing diameter. This
shows how the new model is an improvement over the
models adopted in the design guidelines of SNAME (2002),
which do not consider the strength properties of the sand.
CONCLUSIONS
The development of a new conceptual model and deriva-
tion of an equation to calculate the peak penetration resis-
tance of circular foundations on sand overlying clay were
presented in this paper. Only a homogeneous sand layer is
considered, whereas the undrained shear strength of the clay
may increase linearly with depth. In contrast to other
methods that relate the friction angle to an initial stress
within the sand layer, the operative friction angle in the new
model is related directly to the peak penetration resistance
using a modied strengthdilatancy equation. The method
involves an iterative approach to determine the stress-level-
dependent friction and dilation angles based on the peak
penetration resistance. This can be implemented numerically
using a simple spreadsheet approach. The iterative calcula-
tion process ensures that the stress-level-dependent friction
angle is correctly linked to the stress level mobilised at peak
penetration resistance. The inuence of foundation size and
undrained shear strength of the underlying clay on the
conning stress in the sand is thereby incorporated.
The approach relies on an empirically determined distribu-
tion factor (D
F
) relating the normal stress on the slip surface
to the average vertical stress beneath the foundation at that
depth. The peak penetration resistance calculated using the
deduced linear variation of D
F
with H
s
/D was shown to
provide excellent agreement not only with the 30 centrifuge
tests of Lee et al. (2013) used to calibrate D
F
(an expected
result), but also with another 19 geotechnical centrifuge tests
not used in the calibration.
The new conceptual model has been veried only for
ratios of sand thickness to footing diameter less than 1
.
12.
This covers the range of practical interest for the offshore
jack-up industry, where calculation of potential punch-
through capacities is critical. Further calibration of the
model for loose sand layers overlying soft clay, and for
different footing shapes and conical angles, is recommended.
Furthermore, verication with full-scale penetration events
would be highly benecial.
ACKNOWLEDGEMENTS
The research presented here was supported by the Austra-
lian Research Council through the ARC Linkage grant
1
2
3
4
5
6
0 02 04 06 08 10
C
a
l
c
u
l
a
t
e
d
/
q
N
p
e
a
k
c
0
s
u
0
H D
s
/
(a)
For both graphs:
15 m
31
8 kN/m
D

cv
s
3
1
2
3
4
5
6
0 02 04 06 08 10
H D
s
/
(b)
s
u
: kPa
D
e
p
t
h
:

m
H
s
1
1
25 (RHS graph)
15 (LHS graph)
I
D
10
I
D
08
I
D
06
I
D
04
I
D
02
Fig. 15. Normalised peak resistance as a function of relative sand thickness at various relative densities for
(a) s
u
1
.
5 kPa/m and (b) s
u
2
.
5 kPa/m, from surface
Table 5. Values of 9 and calculated using the new conceptual model for database of 49
centrifuge tests with H
s
/D < 1
.
12
Test name I
D
Calculated 9:
degrees
Calculated :
degrees
47 tests with very dense sand layer 0
.
880
.
99 36
.
741
.
7 6
.
312
.
2
NUS_F8 0
.
61 36
.
3 5
.
3
NUS_F9 0
.
58 35
.
3 4
.
1
1296 LEE, RANDOLPH AND CASSIDY
scheme (Project LP0561838), and by industry partner Keppel
Offshore and Maritime Limited. This work forms part of the
activities of the Centre for Offshore Foundation Systems and
the Australian Research Council Centre of Excellence for
Geotechnical Science and Engineering at UWA. The authors
acknowledge extensive support through the ARCs Federa-
tion and Future Fellowships, Discovery and Linkage pro-
grammes. The third author holds the Chair of Offshore
Foundations from the Lloyds Register Foundation. This
support is gratefully acknowledged.
NOTATION
A area of innitesimal disc element
D foundation diameter
D
F
distribution factor
E model parameter
E
0
model parameter when dilation angle is zero
H
s
sand thickness
I
D
in situ relative density
I
R
relative dilatancy index
K
s
punching shear coefcient
N
c0
bearing capacity factor
Q natural logarithm of grain-crushing strength
q
b
bearing capacity of underlying clay
q
peak
peak penetration resistance
q
sand
bearing capacity in sand
q
0
surcharge at sand surface
R radius of innitesimal disc element
r radial coordinate
s
u
shear strength
s
u0
undrained strength at sand/clay interface
z vertical coordinate

a
geometrical parameter
9
s
effective unit weight of sand
arctangent of gradient of linear regression line
k non-dimensional parameter
r strength gradient of underlying clay
standard deviation
9
n
normal effective stress
9
z
mean vertical effective stress
frictional resistance
9 friction angle

reduced friction angle

2
calculated friction angle

cv
critical state friction angle
dilation angle; empirical parameter
REFERENCES
Bolton, M. D. (1986). The strength and dilatancy of sands. Geo-
technique 36, No. 1, 6578, http://dx.doi.org/10.1680/geot.1986.
36.1.65.
Brinch Hansen, J. (1970). A revised and extended formula for
bearing capacity. Bull. Danish Geotech. Inst. No. 28, 511.
Cassidy, M. J. & Houlsby, G. T. (1999). On the modelling of
foundations for jack-up units on sand. Proceedings of the 31st
offshore technology conference, Houston, TX, USA, paper OTC
10995.
Cassidy, M. J. & Houlsby, G. T. (2002). Vertical bearing capacity
factors for conical footings on sand. Geotechnique 52, No. 9,
687692, http://dx.doi.org/10.1680/geot.2002.52.9.687.
Craig, W. H. & Chua, K. (1990). Deep penetration of spud-can
foundations on sand and clay. Geotechnique 40, No. 4, 541
556, http://dx.doi.org/10.1680/geot.1990.40.4.541.
Davis, E. H. (1968). Theories of plasticity and the failure of soil
masses. In Soil mechanics: Selected topics (ed. I. K. Lee), pp.
341380. London, UK: Butterworth.
Drescher, A. & Detournay, E. (1993). Limit load in translational
failure mechanisms for associative and non-associative materials.
Geotechnique 43, No. 3, 443456, http://dx.doi.org/10.1680/
geot.1993.43.3.443.
Hanna, A. M. & Meyerhof, G. G. (1980). Design charts for
ultimate bearing capacity of foundations on sand overlying soft
clay. Can. Geotech. J. 17, No. 2, 300303.
Houlsby, G. T. & Martin, C. M. (2003). Undrained bearing capacity
factors for conical footings on clay. Geotechnique 53, No. 5,
513520, http://dx.doi.org/10.1680/geot.2003.53.5.513.
Lee, K. K. (2009). Investigation of potential spudcan punch-through
failure on sand overlying clay soils. PhD thesis, University of
Western Australia, Perth, Australia.
Lee, K. K., Cassidy, M. J. & Randolph, M. F. (2013). Bearing
capacity on sand overlying clay soils: experimental and nite-
element investigation of potential punch-through failure. Geo-
technique, http://dx.doi.org/10.1680/geot.12.P.175.
Lee, K. K., Randolph, M. F. & Cassidy, M. J. (2009). New
simplied conceptual model for spudcan foundations on sand
overlying clay soils. Proceedings of the 41st offshore technology
conference, Houston, TX, USA, paper OTC-20012.
Martin, C. M. (2001). Vertical bearing capacity of skirted circular
foundations on Tresca soil. Istanbul. Proceedings of the 15th
international conference on soil mechanics and geotechnical
engineering, Istanbul, Turkey, vol. 1, pp. 743746.
Okamura, M., Takemura, J. & Kimura, T. (1997). Centrifuge model
tests on bearing capacity and deformation of sand layer over-
lying clay. Soils Found. 37, No. 1, 7388.
Okamura, M., Takemura, J. & Kimura, T. (1998). Bearing capacity
predictions of sand overlying clay based on limit equilibrium
methods. Soils Found. 38, No. 1, 181194.
Osborne, J. J. & Paisley, J. M. (2002). SE Asia jack-up punch-
throughs: the way forward. Proceedings of the international
conference on offshore site investigation and geotechnics
sustainability and diversity, London, UK, pp. 301306.
Osborne, J. J., Pelley, D., Nelson, C. & Hunt, R. (2006). Unpre-
dicted jack-up foundation performance. Proceedings of the 1st
jack-up Asia conference and exhibition, Singapore, vol. 1.
Osborne, J. J., Houlsby, G. T., Teh, K. L., Bienen, B., Cassidy,
M. J., Randolph, M. F. & Leung, C. F. (2009). Improved guide-
lines for the prediction of geotechnical performance of spudcan
foundations during installation and removal of jack-up units.
Proceedings of the 41st offshore technology conference,
Houston, TX, USA, paper OTC-20291.
Randolph, M. F., Leong, E. C. & Houlsby, G. T. (1991). One-
dimensional analysis of soil plugs in pipe piles. Geotechnique
41, No. 4, 587598, http://dx.doi.org/10.1680/geot.1991.41.4.
587.
Randolph, M. F., Jamiolkowski, M. B. & Zdravkovic, L. (2004).
Load carrying capacity of foundations. Proceedings of the
Skempton Memorial Conference, vol. 1, pp. 207240. London,
UK: Thomas Telford.
SNAME (2002). Guidelines for site specic assessment of mobile
jackup units, Technical and Research Bulletin 5-5A. Jersey, City,
NJ, USA: The Society of Naval Architects & Marine Engineers.
Teh, K. L. (2007). Punch-through of spudcan foundation in sand
overlying clay. PhD thesis, National University of Singapore.
Teh, K. L., Cassidy, M. J., Leung, C. F., Chow, Y. K., Randolph,
M. F. & Quah, C. K. (2008). Revealing the bearing failure
mechanisms of a penetrating spudcan through sand overlying
clay. Geotechnique 58, No. 10, 793804, http://dx.doi.org/
10.1680/geot.2008.58.10.793.
Teh, K. L., Leung, C. F., Chow, Y. K. & Cassidy, M. J. (2010).
Centrifuge model study of spudcan penetration in sand overlying
clay. Geotechnique 60, No. 11, 825842, http://dx.doi.org/
10.1680/geot.8.P.077.
Vermeer, P. A. & Sutjiadi, W. (1985). The uplift resistance of
shallow embedded anchors. Proceedings of the 11th interna-
tional conference on soil mechanics and foundation engineering,
San Francisco, CA, USA, vol. 3, pp. 1216.
Walters, J. K. (1973). A theoretical analysis of stresses in axially
symmetric hoppers and bunkers. Chem. Engng Sci. 28, No. 3,
779789.
BEARING CAPACITY ON SAND OVERLYING CLAY SOILS: A SIMPLIFIED CONCEPTUAL MODEL 1297

You might also like