Vous êtes sur la page 1sur 19

Bransby, M. F. & Randolph, M. F. (1998). GeÂotechnique 48, No.

5, 637±655

Combined loading of skirted foundations

M . F. B R A N S B Y  a n d M . F. R A N D O L P H 

The response of skirted offshore foundations to Nous eÂtudions la reÂponse de fondations offshore
combined vertical (V), moment (M) and hori- aÁ jupe aÁ une combinaison de charge verticale
zontal (H) loading has been studied using two- (V), instantaneÂe (M) et horizontale (H) en utili-
dimensional ®nite-element analysis and upper- sant des analyses d'eÂleÂments ®nis en deux di-
bound plasticity analysis assuming the soil to be mensions et une analyse de plasticite de limite
undrained. New information has been gained supeÂrieure en supposant un sol non draineÂ. Nous
about the shape of the yield locus and the soil tirons de l'analyse d'eÂleÂments ®nis de nouveaux
deformation mechanisms occurring at yield from renseignements sur la forme du lieu d'eÂlasticiteÂ
the ®nite-element analysis. The shape of the limite et sur les meÂcanismes de deÂformation du
yield locus was found to be similar to that sol qui se produisent aÁ la limite eÂlastique. La
predicted by previous workers in V±M and V± forme du lieu se reÂveÁle comme similaire aÁ celle
H space but differed signi®cantly in M±H space. preÂsageÂe par les travaux preÂceÂdents dans les
This behaviour is explained using upper-bound espaces V±M et V±H mais consideÂrablement
plasticity mechanisms suggested by the soil de- diffeÂrente dans l'espace M±H. Nous expliquons
formation mechanisms calculated in the ®nite ce comportement en utilisant les meÂcanismes de
element analysis. This procedure is then used to plasticite de limite supeÂrieure qui sont suggeÂreÂs
give a good approximation to the shape of the par les meÂcanismes de deÂformation du sol cal-
yield locus and thus may form the basis for culeÂs dans l'analyse d'eÂleÂments ®nis. Cette pro-
future design methods. Additionally, a simplify- ceÂdure est alors utiliseÂe pour donner une bonne
ing transformation is suggested for the yield approximation de la forme du lieu qui pourra
locus in H±M space based on plasticity analysis, ainsi former la base des futures meÂthodes de
which allows use of simple mathematical expres- design. De plus, nous suggeÂrons une transforma-
sions to form a design envelope. tion simpli®catrice pour le lieu d'eÂlasticite limite
dans l'espace H±M sur la base de l'analyse de
plasticite qui permet l'utilisation d'expressions
KEYWORDS: bearing capacity; footings/foundations; matheÂmatiques simples pour former une envel-
numerical modelling and analysis; plasticity. oppe de design.

INTRODUCTION undergoing combined vertical, horizontal and mo-


Offshore structures may consist of multiple shallow ment loadings has long been a fundamental soil
foundations which act together in consort (e.g. mechanics research problem. Solutions have gener-
jack-up rigs and tension leg platforms) or single ally been based on plasticity (e.g. Hill, 1950;
shallow foundations (e.g. gravity based platforms). Prandtl, 1921) or use approximate/empirical meth-
Both of these types of foundations undergo com- ods (e.g. Meyerhof, 1953; Hansen, 1970). Recent
bined vertical (V), moment (M) and horizontal (H) research workers working on the problem in refer-
loading due to environmental loadings on the ence to offshore foundation behaviour (e.g. Tan,
superstructure. In order to predict the behaviour of 1990; Murff, 1994; Butter®eld & Gottardi, 1995;
an entire offshore structure the response of these Houlsby & Martin, 1992; Dean et al., 1992;
individual foundations to this combined loading Salencon & Pecker, 1995; Ngo-Tran, 1996) have
must be understood. analysed the combined loading response of shallow
The bearing capacity of shallow foundations foundations using plasticity methods. Yield loci in
V±M±H space have been deduced for different
foundation systems in different soil types. A yield
surface is introduced whereby
Manuscript received 14 April 1997; revised manuscript
 
V M H
accepted 4 February 1998. f , , ˆ0 (1)
Discussion on this paper closes 1 January 1999; for As AsD As
further details see p. ii.
 University of Western Australia, Nedlands. where A is the area of the foundation, D its

637
638 BRANSBY AND RANDOLPH

diameter or breadth, s is a representative soil size linked to the vertical bearing capacity. With-
strength or effective stress, and V, M and H are in the yield surface, behaviour is usually assumed
the total vertical, moment and horizontal forces to be linear elastic (Bell et al., 1991; Martin,
applied to the soil by the footing. Associated 1994).
plasticity is often assumed so that this yield surface Previous workers have investigated either ¯at
also describes a plastic potential de®ning the rela- footing or spudcan behaviour in drained or par-
tive magnitudes of the incremental plastic displace- tially drained sand (e.g. Butter®eld & Gottardi,
ments during elastoplastic yielding (e.g. Murff, 1995; Tan, 1990) or spudcan behaviour in un-
1994) although Tan (1990) required a separate drained clay (Houlsby & Martin, 1992). In both
plastic potential to describe spudcan behaviour in cases there is no resistance to uplift loading, and
sand. To use the normality relationships the load so the yield envelope appears as plotted in Figs
and displacement de®nitions must form work con- 1(a) and 1(b) with M max and H max occurring
jugate pairs so that normalized total system work, when V =V0  0:4ÿ0:5: Because there is detach-
W, is written as ment of the footing from the soil during moment
      loading, normality is not observed. Tan (1990)
W V äv M H äh studied this, and devised three elastoplastic mod-
ˆ ‡ äè ‡ (2)
ADs As D ADs As D els, each with different forms of plastic potential,
which were used successfully to predict footing
where äv=D is the normalized incremental verti- response.
cal footing displacement, äh=D is the normalized Recently, various empirical curve ®ts have been
incremental horizontal displacement and äè is the suggested for the shape of the yield locus based on
incremental footing rotation, and these displace- centrifuge or 1 g model test data (Martin, 1994;
ments are measured at the same point at which Murff, 1994; Dean et al., 1992). Murff (1994)
the loads are taken to act. The yield locus suggested a form of associated yield surface of
describes an isotropic work hardening surface of size given by its intercept on the M ˆ 0, H ˆ 0

H
M /D Moment, M /D M0 Normality:
Yield envelope
δθ
δv/D

V/V0 5 0.5 V0

V0
V Murff (1994)

(a) (b)

M Murff (1994) M0
Butterfield
& Martin (1994)
Gottardi
(1995)

2Vt V0 V
H

(c) (d)

Fig. 1. Yield envelope approach to combined loading response of footings. (a) 3D view of the combined loading
failure envelope (after Butter®eld and Ticof, 1979). (b) Cross-section of the yield locus in V±M space ( H ˆ 0). (c)
Cross-section of the yield locus in V±H space (V ˆ V0 =2). (d) Cross-section of the yield locus in V±M space
( H ˆ 0); skirted foundation (V0 ˆ ÿVt )
COMBINED LOADING OF SKIRTED FOUNDATIONS 639
axis designated as the ultimate vertical load V0 . fuge test results (Watson & Randolph, 1997a) sug-
The yield locus is given as gest that the ultimate uplift load, Vt  ÿV0 and so
 2   Murff's (1994) relationship (equation (3)) gives a
V Vt V Vt yield locus symmetrical about the V ˆ 0 axis (Fig.
f ˆ ÿ 1‡ ‡
V0 V0 V0 V 0 1(d)). Existence of this symmetry is supported by
"    #0:5 results of yield locus probes in V±H space by
H 2 M 2 Watson & Randolph (1997b), but the detailed
‡ ‡ ˆ0 (3) shape of the yield locus has not been investigated.
H0 M0
One of the main motivations behind the present
where Vt is the normalized bearing capacity under work is the behaviour of skirted foundations in the
purely vertical tension and H 0 and M 0 are the normally consolidated calcareous sediments typical
ultimate horizontal load and moment, respectively. of the North-West Shelf in Western Australia.
This is plotted in Fig. 1(b) for Vt ˆ 0. However, due to the consolidation characteristics
Skirted shallow foundations (or bucket founda- of these ®ne grained sediments, environmental
tions) are being used increasingly in areas such as loading periods provoke undrained soil behaviour
the North Sea (Tjelta & Haaland, 1993), and there and so the system behaviour will be applicable to
is increasing interest in their use on the North-West more general undrained soil conditions.
Shelf in Western Australia as they provide uplift
resistance and transfer loads below the surface
without the expense of pile driving. They consist GEOMETRY OF SKIRTED FOUNDATIONS
of shallow foundations with thin skirts at the Investigation into seabed properties of the outer
circumference which penetrate the soil vertically to North-West Shelf suggests that the soil is normally
a depth d, and therefore the foundation appears consolidated, with undrained shear strength su that
like an upturned bucket of diameter D and height varies proportionally with depth z. This is ex-
d. They may form groups of buckets as in the pressed as
corners of a jacket structure or a single large su ˆ kz (4)
footing for a gravity base platform. The presence
of the skirt alters the behaviour of the foundation where k is the undrained shear strength gradient.
signi®cantly compared to ¯at or conical founda- A shallow foundation resting on the surface of
tions. Suctions develop when uplift loadings are this soil would have a small bearing capacity, and
applied so that there is an uplift capacity, and large footing settlement would occur before signi®-
bearing capacity in compression is increased be- cant bearing capacity could be achieved. Alterna-
cause the soil within the skirts is con®ned, forcing tively, deep foundations would be required. Tani &
the failure down to the level of the skirt tips. Craig (1995) and Watson & Randolph (1997a)
Because skirted foundations are able to resist suggest that foundations with skirts of length d
uplift loading, they may resist moment loading transfer the footing load to the skirt tips when the
even when V , 0, and there will be no detachment soil undrained shear strength su0 ˆ kd, and so the
of the footing from the soil surface. Initial centri- bearing capacity will be related to su0 (Fig. 2).

Environmental
loading H
Loads
V
V Displacements
Undrained shear strength
M v
Su0 Su
Skirt length, d θ
h

Footing H
breadth/diameter, D kz

Moment reference point, 'O'

Depth, z
(a) (b)

Relative homogeneity: kD /su0 (=D /d for skirted foundations)

Fig. 2. Offshore skirted foundation geometry. (a) Prototype foundation. (b) Equivalent surface foundation
640 BRANSBY AND RANDOLPH

Tani & Craig (1995) proposed that the behaviour ized yield loci produced from 2D and 3D ®nite-
of the skirted foundation in undrained soil with element analyses in elastic, perfectly plastic soil.
shear strength proportional to depth can be ap- The normalized shapes are very similar, although
proximated by analysing a foundation on the soil the actual limit loads, V0 , M 0 and H 0, will depend
surface but with the soil shear strength pro®le on the geometry of the foundation. As the same
described by su ˆ su0 ‡ kz (Fig. 2). This ®nding types of behaviour of circular footings may be
was a result of comparison of different lower- observed by analysing a strip footing, detailed
bound plasticity analyses of the vertical bearing analysis of the strip footing will lead to insight
capacity of embedded foundations in soil with applicable to the understanding of circular footing
undrained shear strength proportional to depth. The behaviour but requiring signi®cantly less comput-
conclusion suggests that soil above the level of the ing capacity and processing time. In addition,
footing base does not increase the bearing capacity upper-bound plasticity mechanisms are much sim-
of the footing signi®cantly. Centrifuge tests suggest pler to develop for the strip footing case. There-
that this assumption is reasonable for the initial fore, ®ndings from a ®nite-element study of a strip
bearing capacity for skirts less than about 30% of footing are reported in this paper. Naturally, exten-
the foundation width (Tani & Craig, 1995; Watson sion of the results from the strip footing to more
& Randolph, 1997a), but there is increasing bene®t general 3D geometries will require allowance for
from the overlying material for deeper skirts or the different geometry, in that the soil deformation
after signi®cant penetration. mechanisms will be changed, but it is likely that
Beneath the skirt tips the undrained shear the design methods for each geometry will differ
strength of the soil is increasing at a rate k, and only slightly.
this will affect the bearing capacity of the founda-
tion. A dimensionless soil non-homogeneity factor
kD=su0 is introduced to describe this variation with FINITE-ELEMENT ANALYSIS
depth. For a skirted foundation when su0 ˆ kd then Geometry and soil model
kD=su0  D=d. The strip footing geometry is shown in Fig. 2b.
Skirted foundations may be approximated as The footing was assumed to be rigid, rough and of
circular in plan. To understand the behaviour of a breadth D, resting on the soil surface. It was
circular foundation undergoing vertical, moment considered to be in®nitely long to enable plane
and horizontal loading, three-dimensional (3D) strain analysis. The soil was modelled as an elas-
analysis is required. Initial ®nite-element analysis tic-plastic Tresca material with an undrained shear
of both circular footings and in®nitely long strip strength pro®le described using su ˆ su0 ‡ kz,
footings suggests that their behaviour is similar. where the ratio kD=su0 was varied to investigate
This is illustrated in Fig. 3, which shows normal- footing behaviour in soils with different undrained

1.2

Circular footing
1

0.8
Horizontal load, H /H0

Strip footing

0.6

0.4

0.2

0 0 .2 0.4 0.6 0.8 1 1.2


Vertical load, V /V0
kD /su0 5 6

Fig. 3. Combined V±H loading response of footings: strip footing and circular footing
COMBINED LOADING OF SKIRTED FOUNDATIONS 641
shear strength non-homogeneity. However, most ratio of v=D:è:h=D (e.g. (h=D)=(v=D) ˆ 3, è ˆ 0
analyses were carried out with kD=su0 ˆ 6 unless as shown later in Fig. 6) both the shape of the
the effect of kD=su0 variation was being studied yield locus and direction of plastic incremental
speci®cally. Hence, unless otherwise stated in the displacement vector could be ascertained at failure.
text, kD=su0 ˆ 6. This particular value of non- The second type of displacement probe follows
homogeneity factor was chosen as it led to me- that used by Tan (1990) during his centrifuge
chanistic behaviour typical for a wide range of tests, and is known as a `side-swipe' test. It
kD=su0 values. The rigidity index for the soil was consisted of two stages. In the ®rst stage (e.g. OA
taken as G=su ˆ 200, but the ®ndings of the analy- in Fig. 6) a given displacement (typically vertical)
sis will be independent of soil stiffness as only was prescribed to the footing from the zero load
ultimate behaviour is investigated. This simple soil state until ultimate load was reached. Then a
model approximated the soil behaviour appropri- second displacement probe (AB in Fig. 6) was
ately for investigation of the system failure (yield) applied from this load state, during which the
locus but would have to be re®ned for investigation vertical displacement increment äv ˆ 0. The pre-
of prefailure behaviour. mise of the test is that because elastic stiffnesses
Footing loads and displacements are as de®ned are much greater than plastic stiffnesses, the stress
in Fig. 2b. The reference point for moment (and path in the second stage will almost exactly
vertical displacement) is in the centre of the foot- follow the shape of the yield locus as negligible
ing at the soil±footing interface. The position of expansion of the yield locus will be required to
this reference point will affect the results (Martin, balance the small elastic deformation. This means
1994; Butter®eld & Gottardi, 1995). that the shape of the yield locus (on a certain
Finite-element analysis was carried out using plane in V±M±H space) may be determined in
CRISP90 (Britto & Gunn, 1990) on a 5DX133 one test. In small-strain ®nite-element analysis,
personal computer, and the ®nite-element mesh once ultimate vertical load is reached there is no
consisted of 472 cubic strain triangles with ten further increase in vertical load with increased
elements across the width of the footing. Rigid footing penetration. Hence, the plastic stiffness at
vertical boundaries were placed 2´5D from the yield is zero and so there is no increase in yield
edge of the footing while a rigid base was also locus size as the side-swipe test progresses. How-
2´5D below the footing. Although the position of ever, because elastoplastic yielding occurs within
these rigid boundaries would affect the stiffness of the yield locus, the side-swipe test tracks a load
the foundation during initial (elastic) loading, the path marginally inside the yield locus. This is
failure loads will be unaffected as the failure soil because plastic displacements occurring within the
deformation mechanisms occur close to the foot- global yield locus balance the elastic displace-
ing. A consolidation analysis was carried out with ments along an internal load path to give äv ˆ 0
small-time increments and low-permeability soil so (Houlsby, 1997).
that the soil behaviour was undrained.

UPPER-BOUND PLASTICITY ANALYSIS


Aims Methodology
The aim of the series of ®nite-element analyses Upper-bound plasticity calculations were carried
was to investigate the yield locus for the strip out in order to provide independent estimates of
footing in V±M±H space. Of special interest was the yield locus. The collapse mechanisms were
the shape of the yield locus, the soil deformation based on those calculated in the ®nite-element
mechanisms occurring at yield and whether nor- analysis at different combinations of load. How-
mality applied. The shape of the yield locus was ever, by the nature of the upper bound approach
also to be compared with those presently recom- the calculated yield locus will lie on or outside the
mended. true locus.
For any given two load components, e.g. V and
M, a kinematic mechanism with associated work
Probing the yield locus dissipation was used to give a prediction of the
Displacement controlled events were found to be unknown load, e.g. H, at failure. The procedure
more suitable for probing the yield locus than was repeated with different mechanisms until a
stress controlled probes as discussed by Bransby & mimimum upper bound value of the unknown load
Randolph (1997a). was obtained for the speci®ed loading. This was
Two types of displacement probes were carried then repeated for a range of loads in order to
out. In the ®rst, a single probe was carried out describe the upper-bound yield locus. Plasticity
from a state of zero load at a constant ratio of theorems state that normality will be obeyed on
v=D:è:h=D until the footing loading did not vary the deduced yield locus and the yield locus must
with increased displacement. By choosing a given be convex (Chen, 1975).
642 BRANSBY AND RANDOLPH

Mechanisms
Speci®c forms of mechanisms were suggested δv
by results from the ®nite-element analysis. How- R δθ
ever, all the mechanisms used were based on two L
2f δh
different component mechanisms: the `scoop' me- e

chanism (Fig. 4(a)) and the `wedge' mechanism M D


α β
(Fig. 4(b)). The equations for work dissipation in K
J
A C B G
the soil for these mechanisms are given in a more I
E
F
detailed report (Bransby & Randolph, 1997b).
D
The global mechanism comprised combinations Wedge Scoop Wedge
of the scoop and wedge mechanisms, as shown in
Fig. 5(a). The wedge angles á and â, eccentricities (a)
e and f , and scoop rotation centre height L could
all be varied in order to ®nd the lowest collapse
load. In addition, the footing displacement (in
particular, the horizontal displacement, äh) could δv
R
be varied, while retaining compatibility of vertical δθ
displacements.
L δh
The general mechanism encompasses a number f e
of simpler mechanisms. For example, as the eccen- D
tricities e and f decrease ( f becoming more nega- J A C B
β G
tive) the central scoop mechanism may be E F
`eliminated'. The resulting two wedges may both
move downwards (for example, for purely vertical Scoop
D
Wedge
load) or one wedge may move upwards as the
footing rotates. A particular degenerate form of the
mechanism is shown in Fig. 5(b), where one wedge (b)
is eliminated, leaving just the scoop and a single
wedge. The scoop may extend beyond the edge of Fig. 5. Combined `wedge±scoop' mechanisms. (a)
`Wedge±scoop±wedge' mechanism. (b) `Scoop±wedge'
mechanism
O

the footing, so that the eccentricity f becomes


positive.
η R
L
δθ
V±H LOAD SPACE (INCLINED LOADING)
δh Finite-element results
The yield envelope was probed using both hor-
A C B izontal side-swipe tests and constant displacement
D ratio tests (probes). In the side-swipe test the initial
w probe was a vertical displacement probe with no
rotation or horizontal displacement, v=D ˆ 0:083,
(a) h=D ˆ 0, è ˆ 0 (OA in Fig. 6), which allowed
calculation of V0 . The horizontal side-swipe was
carried out with displacement increments äv=D ˆ
δθ 0, äh=D ˆ 0:083, äè ˆ 0 (AB in Fig. 6), and the
w stress path approximated the shape of the yield
V1 locus in V±H space.
δh
A The ultimate vertical load, V0 , was compared
B α
E
with solutions from plasticity analyses. Lower-
D bound analyses from Davis & Booker (1973) and
C
Tani & Craig (1995) predict that V0 =Dsu0 ˆ 10:49
(b)
when kD=su0 ˆ 6, whereas Houlsby & Wroth
(1983) found V0 =Dsu0 ˆ 10:37. These compare
Fig. 4. Component upper bound plasticity mechan- well with V0 =Dsu0 ˆ 10:7 obtained from the ®nite-
isms. (a) The `scoop' mechanism. (b) The `wedge' element analysis. Equally, the measured ultimate
mechanism horizontal load H 0 =Asu0 ˆ 1:034 compares well
COMBINED LOADING OF SKIRTED FOUNDATIONS 643

1.2 δv /D 50

C B
1

0.8
Horizontal load, H /Dsu0

h /v 5 3

0.6 Probe, Murff (1994)


h /v 5 3 D

0.4
‘Side swipe’ , Equation (11),
δv 5 0 n 5 5, m 5 2.5,
0.2 p 5 2,

‘Side swipe’ (h, θ 50)


O
0

0 2 4 6 8 10 12
kD /su0 56 Vertical load, V /Dsu0

Fig. 6. Yield locus under V±H loading (M  0)

with the exact solution H=Asu0 ˆ 1. Even closer supporting previous experimental work (Tan, 1990;
agreement with the maximum horizontal load Martin, 1994; Watson & Randolph, 1997b), but
could have been achieved using interface elements suggests that probe tests are required for veri®ca-
or more elements beneath the surface of the foot- tion of the shape of a yield locus determined soley
ing, but the degree of accuracy was considered by side-swipe tests.
acceptable. Knowledge of the direction of the plastic dis-
During the displacement probe from the origin placement increment vector at different points on
with h=v ˆ 3, the stress path reaches the yield the yield locus, obtained by the different probe
locus near C in Fig. 6, and follows the yield locus tests, allows investigation of normality. Fig. 7 com-
until stresses no longer change at `D'. When on pares the gradient of the yield locus with plastic
the yield locus, the incremental horizontal displace- displacement increment ratio obtained from the
ment äh=D ˆ ähe =D ‡ ähp =D and äv=D ˆ äve = end-points of different probes. Normality holds.
D ‡ ävp =D, where the superscripts `e' and `p' As the elastic displacements become negligible
represent elastic and plastic deformations, respec- at the stabilization point on the yield locus during
tively. Hence, the elastic deformation balances the displacement probe tests, the incremental soil dis-
plastic deformations, ähp =ävp , dictated by the placements will be those for the plastic failure
plastic potential so that äh=äv is as prescribed by mechanism, and the soil deformation mechanism at
the external displacements. When stabilized (at D) that point on the yield locus can be found. Soil
on the yield locus, elastic displacement increments deformation mechanisms at different points on the
are zero as there is no load change (ÄV , yield locus are shown in Fig. 8, and show that
Ä H, ÄM ˆ 0) and so ähp =ävp ˆ äh=äv. Hence, there are clearly different mechanisms on different
the direction of the plastic displacement increment sections. On the ¯at section of the yield locus
at the ®nal failure point is known. when H= H 0  1 the deformation mechanism is
Displacement probes with a range of different that of simple sliding of the footing over the soil
displacement ratios (h=v) were carried out to in- (as it must be if normality is obeyed as äv  0.)
vestigate the plastic displacement increment vector However, when H  0, a type of double-wedge
at different points on the yield locus and to exam- mechanism is seen, with soil being pushed out in
ine the accuracy of the side-swipe test (Fig. 7). two Prandtl-type wedges so that the soil deforma-
The probe tests suggest that the side-swipe test tion mechanism remains in the softer upper layers
follows a load path just within the yield locus. For immediately beneath the foundation. At intermedi-
a given vertical load, the side swipe underestimates ate points along the yield locus an asymmetrical
the amount of horizontal load sustained by ap- double-wedge mechanism prevails, with the footing
proximately 5% of H 0 . For most conditions this moving with the soil in the larger wedge zone
may be considered to be a negligible difference, (Fig. 8). This mechanism is more symmetrical (like
644 BRANSBY AND RANDOLPH

1.2 Final point of displacement


probe with h /v 5 6

1
h /v 5 4

Note: gradients of yield locus if normality


20.167
Horizontal load, H /Dsu0

0.8 holds are shown. Normals would not appear


as such due to the distorted V–H scale
20.25
h /v 5 3
0.6

20.33
0.4 ‘Side swipe’ h /v 5 2
δv 5 0
20.5
0.2

kD /su0 5 6 ‘Side swipe’ (h,θ 5 0)


0

0 2 4 6 8 10 12
Vertical load, H /Dsu0

Fig. 7. Investigation of normality under V±H loading (M  0)

δh /δv 56 δh /δv 5 3

3.85 m 3.2 m

148
448 98 408
δh /δv 56 δh /δv 5 3.9

1.2
Yield locus: double–wedge mechanism FE probe
1 h /v 5 6
Horizontal load, H /Dsu0

0.8 8.7

6
0.6 h /v 5 3 δh /δv 5 0
3.9
0.4
3.03 V
0.2 h /v 5 2
FE yield locus
0 298
0 2 4 6 8 10 12 δh /δv 5 0
Vertical load, V /Dsu0
kD /su0 5 6

Fig. 8. Comparison of ®nite-element and upper-bound displacement mechanisms under V±H loading (M  0;
D ˆ 12 m)
COMBINED LOADING OF SKIRTED FOUNDATIONS 645
the H ˆ 0 mechanism) when H=V reduces, and soil deformation mechanisms and the upper-bound
more asymmetrical (approaching the H= H 0 ˆ 1 deformation mechanisms at three points along the
mechanism) as H=V increases. yield locus. The plastic deformation mechanisms
re¯ect the changing soil deformation mechanisms
along the yield locus, con®rming the suitability of
The shape of the yield locus the mechanism chosen.
The shape of the yield locus is compared to As a check to the upper-bound methodology,
Murff's (1994) equation (3). If Vt ˆ ÿV0 (as sug- and to assess its suitability to predict the shape of
gested by results from Watson & Randolph (1997a); the yield locus in more general soil conditions, the
and implicit in the ®nite-element boundary condi- double-wedge mechanism was used to predict the
tions) Murff's (1994) equation simpli®es to shape of the yield locus in V±H space for homo-
 2 "    #0:5 geneous soil (i.e. kD=su0 ˆ 0), for which a theo-
V H 2 M 2 rectically exact solution exists. The deduced yield
f ˆ ÿ1 ‡ ‡ ˆ 0 (5) locus is shown in Fig. 9 alongside Green's (1954)
V0 H0 M0
solution. There is almost exact agreement between
where H 0 is the ultimate horizontal force when the upper-bound analysis and the theoretical solu-
M ˆ 0 and V ˆ 0 and M 0 is the ultimate moment tion.
when V ˆ 0 and H ˆ 0. This expression is plotted A side-swipe ®nite-element analysis was also car-
in Fig. 6 (for M ˆ 0), and using the V0 and H 0 ried out for homogeneous soil, and is shown as
values from the ®nite-element analysis. This gives `AB' in Fig. 9. It suggests that, for this soil pro®le,
a very conservative approximation to the observed the side-swipe test underestimated the size of the
yield locus. As shown later, an improved ®t can be yield locus only slightly.
obtained by increasing the power of H= H 0 to 5.

(V±M) LOAD SPACE (ECCENTRIC LOADING)


Upper-bound plasticity analysis Finite-element results
The mechanism used in V±H space was the Similar probes were carried out in V±M load
double-wedge mechanism: a form of the mechan- space as for the investigation of V±H space except
ism shown in Fig. 5(a) but with a scoop of zero that the rotational side-swipe tests consisted of
width. The computed yield locus agrees well with a vertical penetration, V =D ˆ 0:083(è, h=D ˆ 0)
that obtained from the ®nite-element analyses (Fig. followed by a rotation of è ˆ 0:0167 rad with äv=
8). Fig. 8 compares the ®nite-element calculated D ˆ 0 and äh unconstrained ( H ˆ 0). The main

1.2

1 B
kD /su0 5 0

0.8 Using Green's (1954) solution


Horizontal load, H /Dsu0

0.6
Upper-bound yield locus
using double-wedge mechanism

0.4

0.2
Upper bound Green’s solution Side-swipe test

A
0
0 1 2 3 4 5 6
Vertical load, V /Dsu0

Fig. 9. Yield locus under V±H loading (M  0)


646 BRANSBY AND RANDOLPH

®ndings are summarized in Fig. 10, and show the mechanism gives a very good ®t to the yield locus.
deduced yield locus, displacement probes and soil A detailed comparison of the soil deformation
displacement mechanisms on the yield surface. mechanisms in the ®nite-element analysis and the
Again, normality holds on the yield locus. As for optimum plastic mechanism at different points on
V±H space, the side-swipe load path test remains the yield locus reveals that the deformation me-
slightly within the yield locus as deduced from the chanisms are similar.
displacement probes, but the error is much smaller
than previously.
At the ultimate moment when H ˆ 0, M= HORIZONTAL±MOMENT (H±M) LOAD SPACE
D2 su0 ˆ 1:449. This compares well with the upper- Finite-element results (V ˆ 0)
bound value of M=D2 su ˆ 1:491 predicted using Probes similar to those carried out for V±H and
the plastic scoop mechanism shown later in Fig. 13. V±M space revealed that the maximum moment
The soil deformation mechanisms along the that the footing could sustain did not occur when
yield locus showed the same `family' of different H ˆ 0 but when there was a combination of posi-
mechanisms on different parts of the yield locus. tive horizontal load and moment. To reach the true
When V  0 the simple scoop-type mechanism ultimate moment a probe from the origin with
prevailed, when M  0 the simple symmetrical è . 0, h ˆ 0 was required (OAB in Fig. 12) during
double-wedge mechanism occurred, while at inter- which H increased. Once the yield locus was
mediate points along the yield locus an intermedi- reached (at A) the stress path followed the yield
ate combined scoop and wedge mechanism locus (AB in Fig. 12). At the true ultimate moment
prevailed (Fig. 10). M u (B in Fig. 12) normality applies as äh ˆ 0,
and the soil deformation mechanism is shown in
Fig. 13. Following this probe, two different side-
Shape of the yield locus swipe tests were carried out with äè ˆ 0. When
Murff's (1994) equation with H ˆ 0 gives a äh . 0 the side swipe probed the yield locus for
good, though slightly conservative, ®t to the shape increasing values of H (BD in Fig. 12), whereas
of the yield envelope when the observed V0 and when äh , 0 the yield locus was probed decreas-
M 0 values were used (Fig. 10). ing H (BC in Fig. 12). These two side-swipe tests
allowed approximation of the entire yield locus in
M±H space with V ˆ 0.
Upper-bound plasticity analysis To con®rm the shape of the yield locus and to
The combined scoop-wedge deformation me- investigate normality and the soil deformation me-
chanism (Fig. 5(b)) was used to predict the shape chanisms occurring on the yield surface, constant
of the yield locus in V±M space (Fig. 11). The displacement ratio probes were also carried out.

δθ
5 10
δv /D
δv /D 5 0

1.5
C
B
Moment load, M /D 2su0

θ
5 10
v /D δθ
55
1 δv /D

Murff (1994)
0.5 θ
55
v /D Side swipe
δv 5 0

δθ
50
δv /D
0 A
0 2 4 6 8 10 12
Vertical load, V /Dsu0

Fig. 10. Yield locus and displacement mechanisms under V±M loading ( H ˆ 0)
COMBINED LOADING OF SKIRTED FOUNDATIONS 647

f e
L
2

α
θ
5 4/(1 2 2e /D 1 2f /D)
v /D
1.5
Scoop–wedge mechanism
Moment, M /D 2Su

Final point of
displacement
1 probe with θ
5 10
v /D
θ
53
θ v /D
5 00
v /D
0.5

θ
55
v /D
Side swipe
0

0 2 4 6 8 10 12
Vertical load, V /DSu kD /su0 5 6

Fig. 11. Yield locus an plastic displacement mechanism under V±M loading (M  0)

δh 5 0

Likely yield Mo B
A θ
locus 51
h /D Side swipe,
1.5 E
δθ 5 0,
θ δh . 0
5 22
Moment load, M /D 2su0

h /D Side swipe
from E
F
1 side swipe,
δθ 5 0,
δh , 0 H50
δh 5 0 C

0.5
D

θ θ
5 22 51
h /D h /D
O
0
21 20.5 0 0.5 1
Horizontal load, H /Dsu0 kD /su0 5 6

Fig. 12. Yield locus under H±M loading (V  0)

Fig. 12 shows the results from one probe with ing the horizontal displacement while rotation was
è=(h=D) ˆ 1 which investigates a section of the prescribed) in order to obtain M 0 . The results
yield locus with H . 0 and another probe with con®rmed the approximate shape of the yield locus
è=(h=D) ˆ ÿ2 which investigates the yield locus found during the side-swipe tests but showed that
with H , 0. A seperate displacement probe (OE) the side-swipe test slightly underestimated the size
was carried out with H ˆ 0 (set by not constrain- of the true yield locus especially for small magni-
648 BRANSBY AND RANDOLPH

tudes of H. However, the probes suggested that assumed that the mechanism was symmetrical
normality held as before. about the centre line of the footing, and so there
Preliminary testing with H ˆ 0, è . 0 and then were only four variables to ®nd the minimum
a side swipe with äè ˆ 0, äh=D . 0 failed to show moment for any given horizontal loading: half-
the existence of the true peak moment. Instead, the width of the scoop, e, height of the scoop centre
maximum moment, M 0 with H ˆ 0, V ˆ 0 was of rotation, L, angle of the wedges, â, and horizon-
reached in the initial probe (E in Fig. 12) and the tal footing displacement, h. Work equations for this
side swipe then missed the corner of the yield deformation mechanism are given in Bransby &
locus as elastic behaviour was seen with little Randolph (1997b). This mechanism was used to
change in M until the yield locus was reached calculate a second approximation to the yield locus
(dotted line `EF' in Fig. 12). This suggests that the (Fig. 13). As a further cut-off, when j H=Asu0 j > 1
true moment peak could be missed in an experi- a simple slip mechanism would prevail, and this
mental investigation so that the yield locus would was used as an additional bound to the yield locus.
appear almost symmetrical. Using the three different plastic deformation
mechanisms allowed the shape of the yield locus
to be approximated very closely (Fig. 13), and the
Upper-bound plasticity analysis soil deformation mechanisms calculated on the
The soil deformation mechanisms calculated on yield locus in the ®nite-element analysis show
the yield locus in the ®nite-element analyses sug- good agreement.
gest two different upper bound plasticity mechan-
isms. The ®rst is the scoop mechanism (Fig. 13)
previously used to calculate ultimate moment and a The shape of the yield locus
particular form of the scoop±wedge mechanism The eccentricity of the yield locus in M±H
used in V±M space with e ˆ D=2 and f ˆ 0. For space implies that, if symmetric relationships of
any given horizontal load, L was varied to give the the form suggested by Murff (1994) are to be used,
minimum moment, and so a yield locus was ob- the reference point for moments must be chosen
tained (Fig. 13). The second mechanism consists of with some care. Eccentricities of the yield locus in
a scoop and two wedges, as suggested by the soil H±M space have been previously observed by
deformation mechanisms calculated close to ulti- Butter®eld & Gottardi (1995) and Martin (1994)
mate moment during ®nite-element analysis. It was for a range of different V =V0 sections of the yield

δθ
5 22
δh /D
δh 5 0
O
L 5 8.1 m L 5 3.25m e 5 5.53m

δθ β 5 678
δθ 5 24
5 21.5 δh /D
δh /D Scoop mechanism
Wedge–scoop–wedge mechanism

2
kD/su0 5 6 θ
5 0.5
h /D
1.5
Moment, M /D 2 su0

θ
1 δθ 51
5` h /D
δh/D δθ
Probe end-point 51 L 5 1.71m
0.5 δh /D δθ 5 0 e 5 3.65m
with δθ
5 22
δh/D
0 . θ β 5 288
21 20 5 0 0.5 1 5 1.25
Horizontal load, H /Dsu0 h /D

Fig. 13. Comparison between ®nite-element and upper-bound deformation mechanisms under H±M loading
(D ˆ 12 m)
COMBINED LOADING OF SKIRTED FOUNDATIONS 649
locus. Martin analysed spudcan foundations in soft mechanism shown in Fig. 13. For a given kD=su0
clays, and found an eccentricity in the same direc- value an optimum scoop mechanism can be found
tion (Fig. 1(c)), which he suggested was due to the using this plastic mechanism by varying the height
kinematics of the system, shape of the footing and of the point of rotation above the base of the
the normality property, although exact normality footing (distance L in Fig. 13) until a mimimum
may well not have occurred due to detachment of moment, M 0 , is found (with H, V ˆ 0). Values of
the footing from the soil at low vertical loads. M 0 and L=D from this analysis are plotted against
Butter®eld & Gottardi (1995) studied strip footings kD=su0 in Fig. 15 together with the ®nite-element
on drained sand, and found that the eccentricity of measured M 0 . Agreement is good between theory
the yield locus was different from that observed and ®nite-element analysis, given that the side-
here with maximum moment sustained with small swipe tests generally underestimate the size of the
negative horizontal loads (Fig. 1(c)). They sug- yield locus and thus give a lower estimate of M 0
gested rede®ning the position of the moment refer- (e.g. Fig. 12).
ence point to a point below the footing so that the The gradient of the yield locus at H ˆ 0 can be
yield locus was symmetrical about the H ˆ 0 axis calculated using the scoop mechanism, which gave
although no attempt was made to justify this trans- the optimum moment (M 0 ) even without the as-
formation theoretically. sumption of normality. The work dissipated, W , in
The yield locus found in this investigation dif- the slip plane of the scoop mechanism is equated
fers from both the investigations previously men- to the work input by external loading. For the
tioned in that there is an almost straight section general loading case then
when H  0. This section was investigated further V äv ‡ M äè ‡ H äh ˆ äW (7)
by varying the soil non-homogeneity factor kD=su0
and carrying out side-swipe tests for each soil For a single-scoop mechanism the work dissipation
condition. The effect of this variation is shown in per unit rotation, äW =äè(ˆ M 0 ), remains constant,
Fig. 14, and it can be seen that as the soil non- and when the footing remains attached to the soil
homogeneity increases (i.e. kD=su0 increases) so in the scoop then äh ˆ ÿäè L and äv ˆ 0. Hence
too does the gradient of the straight section of the equation (7) simpli®es to
yield locus in H±M space. M ˆ M 0 ‡ HL (8)
Normality implies that a single soil deformation
mechanism prevails on any straight section of the and so the yield locus in M±H space will have
yield locus. For the straight section with H  0 in gradient L for this single deformation mechanism.
H±M space the mechanism occurring is the scoop In normalized M±H space then

2 kD /su0 5 10

kD /su0 5 6

1.5

kD /su0 5 3
Moment, M /D 2 su0

1
kD /su0 5 1.2

kD /su0 5 0

0.5

0
21 20.5 0 0.5 1
Horizontal load, H /Dsu0

Fig. 14. Yield loci for varying kD=Su values under H±M loading (V  0)
650 BRANSBY AND RANDOLPH

M /D 2su0 Max.
moment, Mu

M0
L /D

H /Dsu0

3 0.6

Gradient M /DH (FE)


2.5 0.5
Mu (theory)
L /D (theory) M0 (theory)

L/D, gradient M/HD


2 0.4
M /D 2 su0

O Scoop
mechanism
1.5 0.3
L
M0
Moment, M0 (FE)
1 0.2

Max. moment, Mu (FE)


B
0.5 0.1
0 5 10 15 20
kD /su0

Fig. 15. Scoop mechanism under H±M loading (V  0): comparison with ®nite-element
results

M M0 L H on the superstructure above the footing will


ˆ ‡ (9)
ADsu ADsu D Asu produce positive horizontal and moment loadings
on the footing so that it is only this quadrant of
and so the straight-line section of the yield loci in the yield locus that is critical for design.
normalized M±H space (Figs 12 and 14) should Side-swipe tests were carried out in M±H space
have gradient (d(M=D)=d H ˆ)L=D. These gradi- for varying V =V0 values for the foundation with
ents are plotted for a range of kD=su0 values in kD=su0 ˆ 6. In order to remain at approximately
Fig. 15 and compare well with the theoretically constant V =V0 during the moment side swipe, a
derived L=D values. vertical displacement increment was applied during
Use of the normality property suggests that the the test. For the side-swipe OABC (Fig. 16) the
yield locus will have the same gradient because the initial displacement probe (OA) had v=D ˆ
kinematics of the scoop mechanism gives äh ˆ 0:00125, h=D ˆ è ˆ 0, which was followed by
ÿL äè, which is equivalent to äè=(äh=D) ˆ probe from state A, AB with äè ˆ 0:00119,
ÿD=L. Therefore, the gradient of the plastic poten- äv=D ˆ 0:00238, äh ˆ 0 and then äh=D ˆ
tial, and hence yield locus, must be L=D as ÿ0:0167, äv=D ˆ 0:00139, äè ˆ 0 (BC). A sec-
determined above. ond side swipe but with äv ˆ 0 during the last two
stages of the test is shown in Fig. 16 (OAB9C9).
The line plots above the previous side-swipe test
H±M load space (varying V =V0 ) because V reduces during the test. Due to the
For foundations comprising a single footing, dif®culty of conducting accurate displacement
such as in gravity-based platforms, the vertical probes for ®xed vertical loads, stress-controlled
loading on the footing is almost constant tests were also carried out. Fig. 16 shows results
throughout environmental loading periods and will from tests with initial vertical stress probes (with
equal the (generally large) weight of the platform V =V0 ˆ 0:75) followed by M±H stress probes with
above. Hence, environmental conditions will result ÄV ˆ 0. Due to prefailure yield, the position of
in loading in one plane of ®xed V =V0 in V±M± the yield locus is dif®cult to ascertain, and so
H space. Therefore, for design, yield loci will be bounds to its probable position are given to allow
required in H±M space for different values of for this uncertainty.
V =V0 . Indeed, environmental horizontal loadings The yield locus at V =V0 ˆ 0:75 displays a simi-
COMBINED LOADING OF SKIRTED FOUNDATIONS 651

V50
Wedge–scoop–wedge
1.5 mechanism
Scoop–wedge
mechanism
Moment load, M /D 2 su0

1 V /V0 5 0.75 Wedge–scoop–wedge


Scoop–wedge mechanism
mechanism
B′
B

Side–swipe
tests
0.5 Double–wedge
mechanism

FE stress probes
C′ (regions of failure)

O, A
0
21 20.5 0 0.5 1
Horizontal load, H /Dsu0

Fig. 16. Yield loci under H±M loading (V ˆ 0, V =V0 ˆ 0:75)

lar eccentricity to the yield locus at V ˆ 0, and so agreement at H ˆ 0 is due to the side-swipe tests
similar upper-bound mechanisms were used to remaining within the yield loci.
estimate the position of the yield locus (Fig. 16). The shape of the M  ÿ H plot is almost symme-
The mechanisms capture the shape of the yield trical about H ˆ 0 (Fig. 17), suggesting that a
locus but slightly overpredict its magnitude. This curve ®t similar to the expression suggested by
suggests that the mechanisms may be used to Murff would be appropriate for description of the
predict fully the shape of the yield locus in V±M± yield locus in V±M±H space:
H space. 2 3
!2  ! m ! n 1= p
V M H
f ˆ ÿ1 ‡ 4 ‡ 5
V 0 M H
0 0
SIMPLIFYING TRANSFORMATION AND DESIGN
CURVE RECOMMENDATION (11)
The scoop mechanism observed on the yield
locus suggests that the type of simplifying transfor- When H= H 0 , 0 the M  ÿ H lines diverge with
mation suggested by Butter®eld & Gottardi (1995) reducing kD=su0 , suggesting that n and m increas-
would allow simpler expression of the yield locus. ing with kD=su0 may approximate closely the yield
Indeed, if the reference point for moment loading locus for design use. Equation (11) has been used
is changed to the rotation centre for the scoop (i.e. with n ˆ 5, m ˆ 2:5 and p ˆ 2 and L=D ˆ 0:545
at a height L above the base of the footing), then and M 0 ˆ 1:426 as calculated using the plastic
the new moment is given by scoop analysis to produce the approximation to the
yield surface for kD=su0 ˆ 6 (Fig. 18). A good ®t
M M L H results, with the poorest agreement in the M ˆ 0
ˆ ÿ (10) plane.
ADsu0 ADsu D Asu
An alternative expression for the yield locus in
and a plot of M  against H should have a hori- V±M±H space is suggested below:
zontal section at H ˆ 0. M  =M 0 is plotted against !2:5 !1=3  !
H= H 0 when V ˆ 0 for a range of kD=su0 values V H M
f ˆ ÿ 1ÿ 1 ÿ
in Fig. 17, where the values of L=D and M 0 were V0 H0 M0
calculated using the plastic scoop mechanism (Fig.
! !5
13). As expected, the yield loci plot close to a 1 M  H

unique line of zero gradient when H  0 but ‡ (12)
diverge as j H= H 0 j increases. The lack of exact 2 M0 H0
652 BRANSBY AND RANDOLPH

1
kD /su0 5 10

kD /su0 5 0
0.8
Moment, M */M0 5 (M2LH )/ M0

kD /su0 5 1.2

0.6
kD /su0 5 6 kD /su0 5 3

kD /su0 5 6

0.4

0.2

NB: theoretical values of L/D and M0 from


plastic scoop mechanism

0
21 20.5 0 0.5 1
Horizontal load, H /H0

Fig. 17. Normalized yield loci for varying kD=Su under H±M loading (V  0)

1.8
1.6 1.6 V50
1.4 1.4
1.2 1.2
H50
M /ADsu0

1 1
M /ADsu0

0.8 0.8
V /V0 5 0.75
0.6 0.6

0.4 0.4
0.2
0.2 kD /su0 5 6
0
0 21.5 21 20.5 0 0.5 1 1.5
0 2 4 6 8 10 12
V /Dsu0 H /Dsu0

1.2

1 FE analysis
l/p
2  m n
0.8  V   M * H  
f5  211    1  
 M0 H0 
H /Dsu0

Curve fit V0
0.6 M50
0.4
From plasticity analysis: Curve fit:
0.2 L /D 5 0.454 n55
M0 5 1.426 m 5 2.5
0 p52
0 2 4 6 8 10 12
V /Dsu0

Fig. 18. Curve ®t to yield locus in V±M±H space (modi®ed Murff equation)
COMBINED LOADING OF SKIRTED FOUNDATIONS 653
The expression allows separation of the curve ing moment transformation based on plasticity
®tting in the V±H and V±M planes as well as analysis together with a design curve modi®ed
including a separate expression for the curvature of from the form suggested by Murff (1994) allows
the yield locus in the M±H plane. The ®t to the good prediction of yield locus shape for skirted
yield locus in V±M±H space (Fig. 19) is good and foundations. Allowing for the increase of moment
better than that using the modi®ed Murff (1994) capacity as horizontal load increases should result
equation in V±H space. in less conservative and therefore less costly foun-
Both equations (11) and (12) are simple arbi- dation design.
trary curve ®ts to a complex yield surface. Many Although normality has been shown to hold for
other equations may ®t the yield loci equally well. the system and soil considered, there will be
However, the accuracy of the above two curve ®ts separate yield loci and plastic potential required
suggests that curve ®tting the data with the simpli- for any system with partial drainage or if there is
fying transformation may allow simple approxima- detachment of the footing from the soil surface.
tion of the yield locus. This will be the case with foundations resting on
the surface of clay or sand and explains the dis-
parity between the results presented in this paper
IMPLICATIONS FOR DESIGN and, for example, the ®ndings of Tan (1990). The
Use of the upper-bound plastic deformation lack of detachment with skirted foundations due to
mechanisms to calculate yield loci for soils with the suctions developed within the skirts leads to
different kD=su0 pro®les should allow close ap- the normality property and easier modelling of the
proximation of the yield locus for different un- foundation response.
drained soil shear strength non-homogeneity Inevitably, approximation to the 3D yield loci
pro®les. Furthermore, this method coupled with such as the expressions in equations (11) and (12)
further ®nite element analysis to identify important fail to capture abrupt changes in collapse mechan-
deformation mechanisms for different foundation ism that occur for particular loading combinations.
geometries and stratigraphies should allow estima- However, at conceptual stages of design, such
tion of yield loci for many different skirted founda- approximations offer a powerful framework for
tion systems. sizing alternative foundations with minimal compu-
It has been demonstrated that use of a simplify- tational effort.

1.8
1.6 1.6
1.4 1.4
1.2 H50 1.2
M /ADsu0

1 1
0.8 0.8
M /ADsu0

0.6 0.6
0.4 0.4 V50
0.2 kD /su0 5 6 0.2
0 0
0 2 4 6 8 10 12 21.5 21 20.5 0 0.5 1 1.5
V/Dsu0 H /Dsu0

1.2
FE analysis
1

0.8 2.5 5
Curve fit V   H  1/ 3  M*  1  M*   H 
H /Dsu0

f 5   2 12 12 12
0.6 V0 H 0   M0  2  M0   H 0 

0.4
From plasticity analysis:
0.2 M50 L /D 5 0.454
M0 5 1.426
0
0 2 4 6 8 10 12
V/Dsu0

Fig. 19. Curve ®t to yield locus in V±M±H space (equation (12))


654 BRANSBY AND RANDOLPH

CONCLUSIONS REFERENCES
To gain insight into the response of individual Bell, R. W., Houlsby, G. T. & Burd, H. J. (1991). Finite
skirted foundations to combined vertical, horizontal element analysis of axisymmetric footings subjected to
and moment loadings, a ®nite-element study was combined loads. Proceedings of the International con-
undertaken for a strip footing in undrained soil ference on computer methods and advances in geome-
chanics, Cairns, Australia, vol. 3. pp. 1765±1770.
with shear strength increasing linearly with depth. Bransby, M. F. & Randolph, M. F. (1997a). Shallow
Plasticity mechanisms were also developed for dif- foundations subject to combined loadings. Proc. 9th
ferent combinations of vertical, horizontal and mo- Int. Conf. Comput. Methods Adv. Geomech. Wuhan 3,
ment loading. 1947±1952.
Displacement probes and `side-swipe' tests simi- Bransby, M. F. & Randolph, M. F. (1997b). Combined
lar to those conducted when model testing enabled loading of skirted foundations, Geomechanics Group
investigation of the shape of the failure locus in Report G1269. Department of Civil Engineering, Uni-
V±M±H space and the soil deformation mechan- versity of Western Australia.
isms occurring at different points on the yield Britto, A. M. & Gunn, M. J. (1990). CRISP90. User's and
programmers's guide.
locus. Butter®eld, R. & Gottardi, G. (1995). Simplifying Trans-
Footing displacements were found to obey nor- formations for the Analysis of Shallow Foundations
mality on the yield locus. This was a result of the on Sand. Proc. 5th Int. Offshore Polar Engng Conf.,
lack of detachment between the footing and the The Hague, 534±538.
soil and the plastic associative material law (Tres- Butter®eld, R. & Ticof, J. (1979). The use of physical
ca) used to describe the soil. models in design. Discussion. Proc. 7th Eur. Conf.
Soil deformation mechanisms on the yield locus Soil Mech, Brighton 4.
were calculated by ®nite-element analysis, and Chen, W. F. (1975). Limit analysis and soil plasticity.
upper-bound plasticity analysis using similar me- New York: Elsevier.
Davis, E. H. & Booker, J. R. (1973). The effect of
chanisms gave good approximations to the yield increasing strength with depth on the bearing capacity
locus obtained from ®nite element analyses. Such of clays. Geotechnique 23, No. 4, 551±563.
upper-bound calculations allow prediction of yield Dean, E. T. R., James, R. G., Schol®eld, A. N., Tan,
loci for different soil undrained shear strength pro- F. S. C. & Tsukamoto, Y. (1992). The bearing capa-
®les and suggest a similar procedure should be city of conical footings on sand in relation to the
carried out with different geometries and extreme behaviour of spudcan footings of jack-ups. Proceed-
soil strength pro®les. ings of the Wroth Memorial Symposium `Predictive
The yield locus was found to be eccentric in H± soil mechanics', Oxford, pp. 230±253.
M space with the maximum moment sustained Green, A. P. (1954). Plastic yielding of metal junctions
due to combined shear and pressure. J. Mech. Phys.
with a signi®cant amount of horizontal load. Solids, 2, No. 3, 197±211.
Hence, previously recommended expressions for Hansen, J. B. (1970). A revised and extended formula for
the yield loci and plastic potentials of shallow bearing capacity, Bulletin 28, pp. 5±11. Copenhagen:
foundations may not be suitable for use with Danish Geotechnical Institute.
skirted foundations. The eccentricity of the yield Hill, R. (1950). The mathematical theory of plasticity.
locus was found to be due to the kinematics of the Oxford: Clarenden Press.
soil deformation mechanism when V , H ˆ 0. Houlsby, G. T. (1997). Private communication.
The plastic analysis of the behaviour in H±M Houlsby, G. T. & Martin, C. M. (1992). Modelling of the
space suggested a simplifying transformation of the behaviour of foundations of jack-up units on clay.
Proceedings of the Wroth Memorial Symposium `Pre-
type proposed by Butter®eld & Gottardi (1995) to dictive soil mechanics', Oxford, pp. 339±358.
correct the eccentricity of the yield locus. The Houlsby, G. T. & Wroth, C. P. (1983). Calculation of
correction is linked to the kinematics of the plastic stresses on shallow penetrometers and footings. Pro-
soil deformation mechanism when H ˆ 0. Preli- ceedings of the IUTAM/IUGG symposium on seabed
minary expressions are suggested to describe the mechanics, Newcastle, Sept. pp. 107±112.
shape of the eccentric yield locus in H±M space Martin, C. M. (1994). Physical and numerical modelling
based on plasticity analysis. of offshore foundations under combined loads. D.Phil
In conclusion, the study of this simple founda- thesis, University of Oxford.
tion led to knowledge about the yield locus and Meyerhof, G. G. (1953). The bearing capacity of footings
under eccentric and inclined loads, Proc. 3rd
soil deformation mechanisms which suggest im- ICSMFE, Zurich, 440±445.
proved design methods for shallow foundations. Murff, J. D. (1994). Limit analysis of multi-footing foun-
Later, similar analyses of more realistic (3D) geo- dation systems. Proc. 8th Int. Conf. Comput. Methods.
metries with complex soil models and foundation Adv. Geomech. Morgantown 1, 223±244.
loadings will allow further advances in the under- Ngo-Tran, C. L. (1996). The analysis of offshore founda-
standing of the combined loading response of tions subjected to combined loading. D.Phil thesis,
skirted foundations to be made, but the behaviour University of Oxford.
is expected to be mechanistically similar to that Prandtl, G. (1921). Eindringungsfestigkeit und festigkeit
observed with the simple strip footing. von schneiden, Angew. Math. U. Mech. 1, 15.
COMBINED LOADING OF SKIRTED FOUNDATIONS 655
Salencon, J. & Pecker, A. (1995). Ultimate bearing capa- concept for a jacket ®nding its place. Offshore Site
city of shallow foundations under inclined and ec- Investigation and foundation behaviour, vol. 28, pp.
centric loads. Part 1: purely cohesive soil. Eur. J. 717±728. Society for Underwater Technology.
Mech. A/Solids, 14, No. 3, 349±375. Watson, P. G. & Randolph, M. F. (1997a). Vertical
Tan, F. S. (1990). Centrifuge and theoretical modelling of capacity of caisson foundations in calcareous sedi-
conical footings on sand. Ph.D. thesis, University of ments. Proc. 7th Int. Offshore Polar Engng Conf.,
Cambridge. Honolulu 1, 784±790.
Tani, K & Craig, W. H. (1995). Bearing capacity of Watson, P. G. & Randolph, M. F. (1997b). A yield
circular foundations on soft clay of strength increasing envelope design approach for caisson foundations in
with depth. Soils Found. 35, No. 4, 21±35. calcareous sediments. Proceedings of BOSS `97, The
Tjelta, T. I. & Haaland, G. (1993). Novel foundation Hague, vol. 1, pp. 259±273.

Vous aimerez peut-être aussi