You are on page 1of 13

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

net/publication/238381379

Applications of the lower and upper bound theorems of plasticity to collapse


of circular foundations

Article · January 2001

CITATIONS READS

103 6,396

2 authors:

Chris Martin Mark Randolph


University of Oxford University of Western Australia
97 PUBLICATIONS   3,564 CITATIONS    383 PUBLICATIONS   14,289 CITATIONS   

SEE PROFILE SEE PROFILE

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

Tunnel-induced settlement damage to masonry buildings View project

Spudcan Foundations for Jackup Units View project

All content following this page was uploaded by Chris Martin on 30 June 2016.

The user has requested enhancement of the downloaded file.


Applications of the Lower and Upper Bound Theorems of Plasticity to
Collapse of Circular Foundations
C.M. Martin1 & M.F. Randolph
Centre for Offshore Foundation Systems, The University of Western Australia, Australia
1
Department of Engineering Science, Oxford University, UK (from October 2000)

ABSTRACT: The bound theorems of plasticity have proved a powerful tool in establishing design methods
for the collapse of shallow foundations, buried anchor plates and other geotechnical structures. Closely brack-
eted upper and lower bound solutions (and in some cases, exact solutions) are now available for a wide range
of bearing capacity problems in plane strain. These solutions have been extended to circular geometries in
certain cases, often through numerical techniques involving the method of characteristics. With the growth in
the use of shallow foundations, suction-emplaced caissons, and vertically loaded anchors for offshore struc-
tures, there has been strong practical interest in the capacity of embedded foundations, both under purely ver-
tical loading, and also under combined vertical, moment and horizontal (V, M, H) loading. In turn, this has
generated renewed interest in developing plasticity solutions for these problems, taking due account of the
variation of shear strength with depth.

1 INTRODUCTION of the early solutions were based on approximate ad-


justments of exact solutions, to allow for practical
The trend in the offshore industry towards develop- features such as the foundation geometry (circular or
ments in deep water, with consequential focus on square, rather than strip) and depth of embedment.
skirted foundations and anchoring systems in gener- Caisson foundations such as those used offshore
ally soft clays and silts, has revived interest in estab- may have skirts that extend up to a diameter below
lishing the ultimate capacity of embedded circular the seabed, penetrating through soil where the
foundations. Most practical applications involve strength increases with depth. The bearing capacity
combinations of vertical (V), moment (M) and hori- of such foundations must therefore consider not only
zontal (H) loading, and the construction of a three- the embedment ratio, d/D (where d is the skirt pene-
dimensional yield envelope in (V, M, H) space. tration and D the foundation diameter), but also the
However, collapse under purely vertical loading is a degree of strength non-homogeneity and the relative
key aspect of much offshore design, and represents roughness of the skirts.
the first step in establishing the full envelope. Referring to Figure 1, the strength profile can be
While finite element methods offer a general- idealised according to
purpose approach for evaluating the full load–
deformation response of foundation systems, much su = sum + kz (1)
design relies on a framework of bearing capacity
calculations, with limiting loads evaluated using
plasticity solutions derived from the classical lower
and upper bound theorems. In certain cases, exact D
sum suo su
analytical solutions are possible, but generally it is
necessary to resort to numerical approaches, for ex- d Caisson
ample using the method of characteristics or finite
element implementations of the bound theorems.
The basis of bearing capacity calculations for
Soil shear strength:
foundations in ‘cohesive’ soils (more accurately
su = sum + kz
soils of low permeability where the concept of
undrained shear strength may be applied) was estab-
z
lished some fifty years ago, with the work of Me-
herhof (1951), Skempton (1951) and others. Many Figure 1. Schematic of caisson and strength profile.
where sum is the strength at the seabed (mudline) and (a) (b) r, u
k is the strength gradient with depth, z. It is helpful
to normalise the bearing capacity, qu, by the strength σzz
at the effective base of the foundation: σ – su
θ α
q u = N c s uo = N c (s um + kd ) (2)
τrz
where Nc is a bearing capacity factor. σhh
σrr σ + su

2 THEORY β
characteristics
The bearing capacity analyses in this paper are based z, v at θ ± π/4
on the classical assumptions of rigid–perfectly plas-
tic soil behaviour, with yielding governed by the Figure 2. Axisymmetric stresses and characteristic directions.
Tresca criterion and an associated plastic flow rule.
All problems considered here involve axially sym-
metric geometry. Figure 2 defines the (r, z) coordi- ∂σ rr ∂τ rz σ rr − σ hh
nate system and the notation adopted for the various + + =0
non-zero components of stress (compression posi- ∂r ∂z r
(5)
tive) and velocity. The nature of the problems con- ∂τ rz ∂σ zz τ rz
+ + =γ
sidered is such that the soil self-weight does not af- ∂r ∂z r
fect the solutions, although for completeness it is
to yield two partial differential equations in σ and θ.
included in the governing equations below.
These are hyperbolic in type, with characteristic di-
rections given by
2.1 Lower bound stress fields
dr
The method of characteristics is a well-known tech- = tan (θ ± π 4 ) (6)
dz
nique for constructing statically admissible stress
fields in the solution of plasticity problems. Several The two families of characteristics are designated α
authors describe the development of the method for and β (Fig. 2b). In these directions σ and θ vary in
axially symmetric conditions (Shield 1955, Szcze- accordance with
pinski 1979, Houlsby & Wroth 1982a) and only a  ds s cos 2θ + ωsu 
brief summary is appropriate here. dσ ± 2su dθ =  ∓ u + u  dr +
Stresses in the meridional plane are assumed to  dz r 
(7)
satisfy the yield criterion (σmax – σmin = 2su). This al-  s sin 2θ 
lows the individual stress components σrr, σzz and τrz γ − u  dz
 r 
to be expressed in terms of two auxiliary variables,
σ and θ (see Fig. 2b): Equivalent versions of this equation are given by
Houlsby & Wroth (1982a) and Tani & Craig (1995),
σ rr = σ − su cos 2θ although the sign of dsu/dz was misprinted in the
σ zz = σ + s u cos 2θ (3) former. For the linearly increasing strength profile
τ rz = su sin 2θ considered here, dsu/dz is simply k.
Equations (6) and (7) can readily be implemented
The hoop stress σhh is usually equated (by assump- in finite difference form and used to solve the vari-
tion) with one or other of the principal stresses in the ous boundary value problems that arise in lower
meridional plane, i.e. bound bearing capacity analysis. Details of these
numerical methods and solution strategies are thor-
σ hh = σ + ωs u (4) oughly described elsewhere (see e.g. Shield 1955,
where ω = ±1 (this is the Haar/von Karman hypothe- Booker & Davis 1977, Szczepinski 1979).
sis). As discussed by Shield (1955), Houlsby &
Wroth (1982a) and others, the choice should be gov- 2.2 Upper bound velocity fields
erned by the expected direction of radial plastic
flow: ω = –1 if it is outward, ω = +1 if it is inward. A stress field constructed using the method of char-
Equations (3) and (4) may be substituted into the acteristics can also serve as the basis for a ‘consis-
equilibrium equations tent’ velocity field in which the principal strain rate
directions coincide with the principal stress direc-
tions, and the flow rule (here equivalent to the in-
compressibility condition) is satisfied. The deriva-
tion of the consistent velocity equations for axial
symmetry (Shield 1955, Szczepinski 1979, Houlsby 3 SURFACE FOUNDATIONS
& Wroth 1982a) closely follows that of the familiar
Geiringer equations in plane strain. A pair of hyper- The indentation of a semi-infinite Tresca medium by
bolic partial differential equations in u and v is ob- a flat-ended circular punch is a fundamental axially
tained, with characteristic directions coincident with symmetric bearing capacity problem, which has
those of the stress field. Along the characteristics, been widely investigated using both the method of
the velocity components vary according to characteristics and independent upper bound analy-
sis. Studies utilising the former technique have uni-
− udr
dusin (θ ± π 4) + dvcos(θ ± π 4) = (8) versally adopted the appropriate Haar/von Karman
2rsin (θ ± π 4) assumption (ω = −1 in the present notation).
Numerical finite difference techniques, similar to
those employed when constructing the stress field, 3.1 Homogeneous soil
are used to determine u and v throughout the plasti- For the simplest case of a perfectly smooth punch on
cally deforming region. Appropriate velocity bound- homogeneous material, Ishlinskii (1944) used a
ary conditions for smooth and rough foundations are graphical technique to construct the appropriate
discussed by Shield (1955) and Eason & Shield mesh of stress characteristics and obtained a lower
(1960) respectively. The extent of the plastically de- bound solution of Nc = 5.68 for the average bearing
forming region is identical to that of the ‘partial’ pressure. Shield (1955) obtained an almost identical
lower bound stress field (see below). result of 5.69, using an iterative finite difference ap-
The form of the consistent velocity field is often proach. More importantly, however, Shield showed
quite complex, and furthermore it does not always that the ‘partial’ stress field in the vicinity of the
lead to an optimal upper bound. A family of ‘inde- punch could be extended throughout the half-space
pendent’ collapse mechanisms can be developed as in a statically admissible manner, thus establishing
shown in Figure 3, and these are used extensively in the solution as a rigorous lower bound. He also
the examples that follow. The velocity fields in- demonstrated that the consistent velocity field gave
volved are very simple, consisting of a series of par- principal strain rates ε that were properly associated
allel streamlines along which the velocity decays with the stress field throughout the deforming re-
with 1/r (thus ensuring satisfaction of the incom- gion, thus proving that the bearing capacity factor of
pressibility requirement). Mechanisms of this type Nc = 5.69 was also an admissible upper bound (and
have been used to obtain upper bound collapse loads hence the exact solution).
for a wide range of axially symmetric bearing capac- This value has been confirmed subsequently by
ity problems (Levin 1955, Kusakabe et al. 1986, Hu several authors including Houlsby & Wroth (1983)
et al. 1999, Randolph et al. 2000). and Tani & Craig (1995). As part of the present
In the present study, the Hill-type mechanism on study, analyses with increasingly refined meshes of
the left of Figure 3 is always critical for smooth- characteristics have indicated that Nc = 5.689 to
based foundations, and is occasionally critical for three decimal places. The accuracy of Ishlinskii’s
rough-based foundations (notably when the non- original 5.68 is certainly remarkable. The value of
homogeneity ratio kD/suo is larger than 1 or 2). The 5.71 obtained by Meyerhof (1951) from numerical
Prandtl-type mechanism on the right usually gives a integration of an approximate analytical expression
better upper bound for rough-based foundations (unrelated to the method of characteristics) also
when kD/suo is small. The inclination of the (conical) compares favourably with the exact solution. Opti-
exit region BDEC can be optimised, and varies from misation of the independent upper bound mechanism
about 45º at shallow embedment of the caisson to of Figure 3 gives Nc = 5.82, some 2% above the
90º (annular flow adjacent to the caisson) at deeper method of characteristics solution. Levin (1955) ob-
embedments. The fan angle, ADB, is adjusted ac- tained an upper bound of 5.84 using a slightly sim-
cordingly. pler mechanism.
Indentation by a perfectly rough circular punch
was first investigated by Eason & Shield (1960). Us-
C E E C ing numerical stress and velocity calculations based
on those of Shield (1955), a rigorous lower bound
and a coincident upper bound were again obtained,
D D the exact bearing capacity factor in this instance be-
B ing Nc = 6.05. Other computations of this stress field
A B have generally given the same result (Salençon &
Matar 1982, Houlsby & Wroth 1983), although Tani
Smooth base Rough base & Craig (1995) have recently claimed a surprisingly
high lower bound of Nc = 6.34. This analysis appears
Figure 3. Independent kinematic collapse mechanisms. to have involved a faulty integration procedure, dis-
cussed below, and the present work has confirmed O A B
that the value of 6.05 is indeed exact (6.048 to three
decimal places). Meyerhof’s approximate result for a C (a) Nc (LB) = 7.56
perfectly rough punch was Nc = 6.18, about 2% D
above the exact solution. The best independent up-
per bound (Fig. 3) is also 6.18 for this problem.
Because Nc = 2 + π for a flat indenter in plane
strain, regardless of roughness, exact shape factors
for circular foundations on homogeneous, isotropic (b) Nc (UB) = 7.56
Tresca soil are 1.11 (smooth) and 1.18 (rough).
Clearly there is a sound theoretical basis for the
shape factors of about 1.2 that are commonly used in
bearing capacity formulae (e.g. Brinch Hansen 1970,
Vesic 1975). It should be noted, however, that most
real soils exhibit a degree of strength anisotropy, and (c) Nc (UB) = 8.08
this can have a surprisingly significant influence on
the bearing capacity in both plane strain and axial
Figure 4. Stress and velocity fields for smooth circular footing
symmetry (Booker & Davis 1972, Randolph 2000). on non-homogeneous soil (kD/sum = 4).
Further theoretical and numerical work in this area is
required.
O A B

3.2 Non-homogeneous soil Stress-


free
Davis & Booker (1973) studied the bearing capacity
of surface strip foundations on cohesive soil where
the strength increases linearly with depth. They used
the method of characteristics to obtain rigorous E
lower bound and coincident upper bound solutions
for both smooth and rough footings, for all possible F
values of the non-homogeneity ratio kB/sum (i.e. zero
to infinity). The method of characteristics has also
been used to investigate the corresponding problem Stress-
in axial symmetry, but no single study has been as free
comprehensive. Houlsby & Wroth (1983) listed
bearing capacity factors for both smooth and rough
σzz only
circular footings, but only for 0 ≤ kD/sum ≤ 10. G
Salençon & Matar (1982) covered the full range of
kD/sum, but only for a fully rough interface. Tani &
Craig (1995) considered both smooth and rough Figure 5. Extension of stress field of Figure 4(a).
foundations on soil with 0 ≤ kD/sum ≤ 30, but their
rough footing results are erroneous. In all three cases
the Nc factors obtained from the method of charac-
dW = 2 s u ε max (9)
teristics have been presented as lower bound col-
lapse loads, but they have not been formally estab- over each cell of the mesh. Velocity discontinuities
lished as rigorous solutions by demonstrating that and/or dissipation at the soil–footing interface, al-
the stress fields involved are extensible throughout though not applicable in this particular example, are
the soil mass. Nor have consistent velocity fields handled by introducing degenerate cells. With suffi-
been computed with a view to showing that the cient mesh refinement, it is found that the consistent
lower bound solutions are upper bounds as well (and upper bound also converges to Nc = 7.56. Presuma-
are therefore exact). bly the coincidence of lower and upper bounds could
An example of a more thorough approach is also be verified by examining principal strain rates
given in Figures 4 and 5, for a smooth circular foot- throughout the mesh and checking for proper asso-
ing on soil with kD/sum = 4. The ‘partial’ field of ciation with the stress field, as was done by Shield
stress characteristics (Fig. 4a) gives a lower bound (1955) for homogeneous material.
bearing capacity – at this stage non-rigorous – of Nc The best upper bound obtained from the inde-
= 7.56. Having found the consistent velocity field pendent collapse mechanism is Nc = 8.08 (Fig. 4c).
(Fig. 4b), the upper bound is computed by numerical This is 7% above the consistent upper bound, a
integration of the Tresca dissipation function somewhat greater differential than the 2% observed
for homogeneous soil. Interestingly, however, the
accuracy of the independent upper bound eventually O F A E B
improves again with increasing kD/sum, to the point
D (a) Nc (LB) = 8.74
where it gives the exact bearing capacity of qu = C
kD/6 (analogous to Davis & Booker’s qu = kB/4) as
sum approaches zero.
The final stage in establishing Nc = 7.56 as a rig-
orous exact solution is to demonstrate the extensibil-
ity of the stress field. As shown in Figure 5, the
mesh of characteristics can be extended using the (b) Nc (UB) = 8.74
strategy of Shield (1955), based on that developed
by Bishop (1953) for plane strain conditions. The
extension field consists of a region where the stress
state is fully plastic (governed by the same equations
as the partial stress field), bounded by a stress free
region to the right of BE and columns of uniaxial
vertical stress below EFG, extending to z = ∞. The
heavy lines around the plastic region denote stati-
cally admissible stress discontinuities. A stress dis- (c) Nc (UB) = 9.71
continuity also arises within the plastic region, just
above F. In this example the uniaxial stresses are
found to be acceptable (|σzz| ≤ 2su), but this is not the Figure 6. Stress and velocity fields for rough circular footing
on non-homogeneous soil (kD/sum = 4).
case when kD/sum becomes very large. It is then nec-
essary to use a different type of extension based on
that of Cox et al. (1961), which was also adapted by The upper bound obtained from the consistent ve-
Davis & Booker (1973) for their work in plane locity field (Fig. 6b) coincides with the lower bound
strain. It is always possible to perform a stress field at Nc = 8.74. By contrast the best independent upper
extension of one type or the other. bound (Fig. 6c) is Nc = 9.71, 11% above the exact
Figure 6 shows the analysis of a rough circular solution. Note that a Hill-type mechanism is critical
footing on soil with kD/sum = 4. To calculate the for this value of kD/sum, even though the foundation
lower bound, it is sufficient to extend the mesh of is fully rough. As with the smooth footing, it is
characteristics (Fig. 6a) as far as the final β charac- found that the independent upper bound mechanism
teristic FD. Note however that the major principal gives its best results for large kD/sum, converging to
stress is not vertical along FD, except of course at D the exact solution (again qu = kD/6) as the mudline
on the axis of symmetry. When integrating stresses strength sum → 0.
over AFD to obtain the collapse load, it is therefore Shape factors for circular foundations on non-
essential to use an expression of the form homogeneous cohesive soil have been given by
R Z various authors, based on either the method of char-
Qu = ∫ σ zz 2πrdr + ∫ τrz 2πrdz (10) acteristics (Salençon & Matar 1982, Houlsby &
0 0 Wroth 1983, Tani & Craig 1995) or independent up-
per bound analysis (Kusakabe et al. 1986). It can
where Z is the distance OD. Tani & Craig (1995) now be stated with confidence that shape factors for
appear to have omitted the term involving τrz, which Tresca soil derived from the method of characteris-
is negative along FD using the sign convention of tics are exact, but as discussed above no fully com-
Figure 2a. This explains why their Nc values for prehensive numerical study has been performed. An
rough footings are high, most noticeably when unexpected result is that the shape factors for
kD/sum is small and the ‘false head’ region OFD is smooth and rough footings drop below unity for
large. When using equation (10) to integrate along kD/sum values which exceed about 4 and 2 respec-
AFD it is tacitly assumed that the stress field could tively. They continue to drop as kD/sum increases,
be extended into OFD in a statically admissible slowly approaching the theoretical limit of 2/3 as sum
manner, if desired. In fact an alternative method of approaches zero, regardless of roughness. The tradi-
calculating Qu is to perform such an extension and tional shape factor of 1.2 is obviously inappropriate
truncate it along the footing boundary, allowing a when the non-homogeneity ratio kD/sum is large, and
straightforward integration of σzz along OFA; this this is a situation frequently encountered offshore.
was the method originally adopted by Eason &
Shield (1960) for homogeneous soil. Using either
technique a lower bound bearing capacity factor of
Nc = 8.74 is obtained. As above it can be shown that
the rough footing stress field is extensible, both for
kD/sum = 4 and for all possible values.
4 SKIRTED FOUNDATIONS O X

It is also possible to develop rigorous plasticity solu-


tions for the bearing capacity of caissons or skirted
circular foundations (Fig. 1). The soil enclosed by (a) Nc (LB) = 7.05
the caisson is normally assumed to be rigid for bear- Y
ing capacity analysis, and the base of the ‘equivalent
circular punch’ is taken to be fully rough. While
these assumptions are clearly appropriate for upper
bound analysis, they are questionable as far as a rig-
orous lower bound solution is concerned – there may
be difficulties in extending the stress field for a
rough ‘base’ into the skirt compartment, particularly
if the internal boundaries of the caisson are not fully
rough. In this paper, however, the usual assumption
of a fully rough interface at the skirt tip level will be
adopted, subject to a minor caveat regarding the rig-
(b) Nc (UB) = 11.86
our of the lower bounds so obtained.
Only axially symmetric geometry will be consid-
ered here, although it is of course possible to de-
velop lower and upper bound solutions for skirted
foundations in plane strain. Shape factors can then
be derived, but it is found that these are a function of
both the embedment ratio (d/D or d/B) and the non-
homogeneity ratio (kD/sum or kB/sum) such that a
compact presentation of results becomes difficult.

4.1 Homogeneous soil


Houlsby & Wroth (1982b) used the method of char-
acteristics to obtain lower bound solutions for the
bearing capacity of an embedded circular punch (0 ≤
d/D ≤ 1) with smooth sides and base. This work was
extended by Martin (1994) to include bases of full
and intermediate roughness (but smooth sides), for
embedment ratios of up to 2.5. Tani & Craig (1995)
developed lower bound solutions for the rough base
and rough sides case, but unfortunately they consid- (c) Nc (UB) = 8.90
ered only a very limited range of embedment (0 ≤
d/D ≤ 0.3). None of these analyses addressed the
question of stress field extensibility. Interestingly,
previous studies of the upper bound problem have Figure 7. Stress and velocity fields for circular caisson (d/D =
been focused on offshore applications, and have not 0.5, rough base, smooth sides) in homogeneous soil.
considered the case of homogeneous soil.
Figure 7a shows a typical mesh of stress charac-
teristics for a smooth-sided skirted foundation (d/D dation moves down, but there is no dissipation be-
= 0.5) in homogeneous soil. The lower bound bear- cause the interface is smooth.
ing capacity factor is Nc = 7.05, but the consistent The stress characteristic mesh (Fig. 7a) can be ex-
velocity field leads to an upper bound of Nc = 11.86 tended in the manner of Figure 5, and hence Nc =
(Fig. 7b). The discrepancy is surprisingly large, and 7.05 is a rigorous lower bound. However, the sig-
is essentially attributable to a conflict between the nificant separation between lower and upper bounds
direction of shearing stress and the direction of suggests that it may be possible to improve on the
shearing strain along the characteristic XY (which is ‘standard’ lower bound solution. This can indeed be
a velocity discontinuity). In this case, and for em- achieved, using an alternative stress field of the type
bedded foundations generally, the independent upper shown in Figure 8. The mesh of characteristics is
bound mechanism allows a much improved result, constructed by working from an artificial (straight)
Nc = 8.90 (Fig. 7c). Note that, in this example, soil free surface inclined at some angle to the horizontal
flows vertically upwards past the skirts as the foun- (cf. Meyerhof 1951). As this inclination is increased,
the calculated bearing capacity also increases, since
the α characteristics rotate through a larger angle violate yield. This alternative lower bound solution
(see equation (7)). There is a limit on the allowable gives Nc = 7.80, a 10% improvement on the 7.05
inclination, however, due to the need to extend the which was obtained earlier, bracketing the exact so-
partial stress field throughout the soil mass, an as- lution to within ±7%.
pect overlooked by Meyerhof. Figure 10a shows upper and lower bound bearing
In Figure 8 the artificial free surface is inclined at capacity factors calculated in a similar fashion for
47° since this is the maximum that allows an exten- skirt length ratios 0 ≤ d/D ≤ 2. Over most of this
sion of the characteristic mesh to be accommodated range the alternative lower bound gives a significant
beneath the real free surface. In this instance an ex- improvement on the standard lower bound. The con-
tension strategy based on that of Cox et al. (1961) is sistent upper bound has not been plotted because the
convenient, and this is shown in Figure 9. Heavy independent solution is better (i.e. lower) for all but
lines around the extended mesh denote statically the shallowest of embedments. There is reasonably
admissible discontinuities. Stresses to the right of EF close bracketing of the true solution for embedments
and FG extend to infinity as indicated, and do not up to about d/D = 1. The alternative lower bound
cuts off at Nc = 9.28 when d/D = 1.27, corresponding
to the point where the partial stress field for an arti-
O ficial free surface inclination of 90° first becomes
Stress- extensible. This cutoff Nc value can be marginally
free B 47°
improved by adopting more complex variations of
the Haar/von Karman parameter ω (for example as a
function of the principal stress inclination θ) rather
A C than fixing it at ω = –1 when constructing the partial
stress field and its extension. Also shown in Figure
10a are the semi-empirical recommendations of
Nc (LB) = 7.80 Skempton (1951) and Brinch Hansen (1970) for a
D smooth-sided circular foundation embedded in ho-

Figure 8. Alternative stress field for caisson of Figure 7. (a) Smooth sides
15
LB
O E 13 UB

B Alt. LB
σrr = σhh 11
only Nc
A 9

7
Brinch Hansen Skempton
5
F 0 0.5 1 1.5 2
d /D
(b) Rough sides
25
LB
UB
20
Alt. LB

N c 15

10
‘spokes’ of uniaxial stress
(cf. Cox et al. 1961)
5
G 0 0.5 1 1.5 2
d /D

Figure 10. Lower and upper bound solutions for rough-based


Figure 9. Extension of stress field of Figure 8. caissons embedded in homogeneous soil.
mogeneous cohesive soil. Although Brinch Hansen’s
Nc factors are much less conservative than those of
Skempton, they still plot below the alternative lower
bound.
Figure 10b shows bound solutions for a skirted (a) Nc (LB) = 9.34
foundation with rough sides in homogeneous soil. In
the alternative lower bound the soil adjacent to the
skirts is stress free (see Fig. 8), hence these Nc val-
ues are unchanged from the ones in Figure 10a. By
contrast the standard lower bound solution takes full
advantage of the rough skirts, mobilising an inter-
face shear stress of su over the entire depth d. Brack-
eting between the lower and upper bounds is again
fairly good up to around d/D = 1. Note that the bear-
ing capacity factors given in Figure 10b are inclu-
sive of interface shear stresses on the skirts, whereas (b) Nc (UB) = 15.95
Tani & Craig (1995) chose to report only the base
resistance component.
Hu et al. (1999) have reported results of finite
element analysis of embedded caissons, and noted
that an embedment ratio d/D = 1 marks the start of a
transition from surface failure, to a confined elasto-
plastic failure. Their results were for a rigidity index
of G/su = 167 (where G is the soil shear modulus)
and the transition depth will vary with this ratio. For
depths greater than this transition depth, the lower
and upper bounds are increasingly irrelevant, as the
bearing capacity is determined by a cavity expansion
limit pressure. At depths up to d/D = 1, the finite (c) Nc (UB) = 12.19
element results of Hu et al. (1999) lie close to mid-
way between the best lower and upper bounds
shown in Figure 10.

4.2 Non-homogeneous soil


Figure 11. Stress and velocity fields for circular caisson (d/D =
Relevant lower bound solutions for circular skirted 0.5, rough base, rough sides) in soil with su = kz.
foundations in non-homogeneous soil have been
published by Martin (1994) (smooth sides, 0 ≤ d/D ≤
2.5, 0 ≤ kD/sum ≤ 5) and Tani & Craig (1995) (rough there is no advantage from an ‘annular’ flow mecha-
sides, 0 ≤ d/D ≤ 0.3, 0 ≤ kD/sum ≤ 30). The only nism of the type shown in Figure 7c. The optimum
available upper bounds are those of Hu et al. (1998) mechanism does, however, involve a fairly steep exit
(smooth and rough sides, 0 ≤ d/D ≤ 0.5, su = kz) and angle, with a wedge of soil that remains stationary
Hu et al. (1999) (smooth and rough sides, 0 ≤ d/D ≤ with respect to the surrounding soil mass.
0.5, 2 ≤ kD/sum ≤ 30). A comprehensive numerical Corresponding solutions for other values of d/D
study of both lower and upper bounds for 0 ≤ d/D ≤ are shown in Figure 12 for smooth- and rough-sided
2 and the full range of kD/sum is currently being un- caissons. For the former, the alternative lower bound
dertaken by the authors. offers some improvement over the standard lower
The case chosen for illustration here involves a bound, and provides excellent bracketing up to about
foundation with rough skirts (d/D = 0.5) in normally d/D = 1.5. For rough-sided caissons, the alternative
consolidated soil (sum = 0, su = kz). As shown in Fig- lower bound is only optimal at very shallow em-
ure 11, the method of characteristics yields an (ex- bedments. Comparing Figures 12a and 12b, it is in-
tensible) lower bound of Nc = 9.34 and a consistent teresting to note that the assumed degree of skirt
upper bound of Nc = 15.95. The ‘alternative’ lower roughness is not particularly crucial for normally
bound gives Nc = 9.27, marginally below the stan- consolidated soil, in contrast to the situation for ho-
dard solution. Note that these are bearing capacity mogeneous soil (compare Figs 10a and 10b).
factors with respect to suo, the strength at skirt tip
level (see equation (2)). The independent kinematic
mechanism gives an improved upper bound of Nc =
12.19 (Fig. 11c). Since the skirts are fully rough,
(a) Smooth sides composed of two Haar/von Karman regions (ω = ±1)
15
separated by a buffer zone (Randolph et al. 2000) in
LB
which the meridional stress state is defined by
13 UB
σ = σo + γz + su z r
Alt. LB (11)
11 θ = 3π / 4
Nc
and the hoop stress by
9
σ hh = σo + γz (12)
7
Stresses in the buffer zone satisfy both equilibrium
and the yield criterion; there are statically admissible
5
jumps in σhh at each edge of the zone. The constant
0 0.5 1 1.5 2 σo defines an arbitrary reference value of σ on z = 0
d /D (cf. Randolph & Houlsby 1984). When building the
(b) Rough sides mesh, new α characteristics are initiated by enforc-
25 ing continuity of σ and θ (and thus of σrr, σzz and τrz)
LB between the buffer zone and the partially constructed
20
UB Haar/von Karman regions. The major principal
Alt. LB stress rotates from θ = 3π/4 in the buffer zone to θ =
0 underneath the plate and θ = 3π/2 above it. Inte-
N c 15 gration of the contact stresses leads to a lower bound
bearing capacity factor of Nc = 12.42. Extension of
10
the stress field to infinity (both below and above the
plate) can be achieved using the strategy developed
by Randolph et al. (2000) for their rigorous lower
5 bound analysis of a spherical penetrometer.
0 0.5 1 1.5 2 The consistent velocity field for the smooth plate
d /D is computed in the usual fashion (Fig. 13b), and with
sufficient mesh refinement gives a converged upper
Figure 12. Lower and upper bound solutions for rough-based bound of Nc = 12.42. This is coincident with the
caissons embedded in non-homogeneous soil (su = kz).

ω = +1

Buffer zone
5 BURIED PLATE FOUNDATIONS

This section presents some new plasticity solutions


relating to the collapse of a deeply buried circular
plate foundation in cohesive soil. The soil mass is (a) Nc (LB) = 12.42
assumed to be homogeneous (uniform su) and of in-
finite extent. It is also assumed that there is no re- ω = –1
striction on the development of tensile stress,
whether in the soil itself or on the rear face of the
plate. While these assumptions allow a relatively
simple analysis using the method of characteristics,
several complicating factors may be important in a (b) Nc (UB) = 12.42
practical design situation (particularly one involving
uplift loading), including proximity to the free soil
surface, a non-homogeneous profile of su, and the
possibility of ‘breakaway’ behind the plate. Indeed,
it is an interesting academic question as to whether
breakaway affects the bearing capacity of a deeply
embedded plate, either in rigid-plastic or elasto-
plastic soil.
Consider first the case of a vanishingly thin plate
with a fully smooth surface. As shown in Figure 13a,
(c) Nc (UB) = 12.64
the appropriate mesh of characteristics is symmetri-
cal above and below the plate. The stress field is Figure 13. Stress and velocity fields for smooth plate.
lower bound and therefore represents the exact solu- tended into the ‘false head’ region it is found that the
tion. Only half of the field is shown to allow a better maximum shear stress mobilised on the surface of
appreciation of the velocity profile across z = 0. As the plate is 0.76su, where the degenerate α character-
would be expected, the flow of soil is heavily con- istic leaves the fan zone. Hence the interface need
centrated in the region close to the edge of the plate. not be completely rough in order for this exact solu-
The independent upper bound mechanism (Fig. 13c) tion to be applicable. The optimum independent col-
yields a good result, Nc = 12.64, less than 2% above lapse mechanism for the rough plate (Fig. 14c) gives
the exact answer. This is despite the fact that the an upper bound of Nc = 13.31, again less than 2%
width of the deforming region is nearly twice that of above the exact solution.
the consistent velocity field. The solution to the analogous problem in plane
Figure 14 shows the corresponding lower and up- strain may be developed analytically using the
per bound solutions for a vanishingly thin, perfectly method of characteristics. As noted by Meyerhof
rough circular plate. The bounds obtained from the (1951) the collapse load is Nc = 3π + 2 = 11.42, in-
method of characteristics are again coincident, this dependent of the roughness of the plate. The fact
time giving an exact bearing capacity factor of Nc = that this solution is exact does not appear to have
13.11. The lower bound stress field can be extended been formally established, with workers such as
into the surrounding soil mass using the same tech- Rowe (1978) and Merifield et al. (1999a) acknowl-
nique as for the smooth case. If the stress field is ex- edging Nc = 11.42 as an upper bound but highlight-
ing the need for a rigorously extended lower bound
stress field (as opposed to Meyerhof's partial stress
field). In the course of the present work it has been
shown that the requisite extension can indeed be per-
formed, using either the Randolph & Houlsby
ω = +1 (1984) method or a plane strain version of the
Randolph et al. (2000) method. Details are given in a
Buffer zone forthcoming paper by the authors (Martin &
Randolph, 2000).
Exact shape factors for infinitely deep circular
plates in a homogeneous, isotropic Tresca material
are therefore 1.09 (smooth) and 1.15 (rough). It is
ω = –1 worth noting that on the basis of his approximate
stress field analysis, Meyerhof (1951) concluded that
the shape factor should be unity for a deeply buried
circular plate, independent of its roughness. While
(a) Nc (LB) = 13.11 this is clearly an erroneous result, at least within the
confines of the highly idealised plasticity problem
posed here, certain features of real soils (in particu-
lar strength anisotropy) may mean that a shape fac-
tor of 1.0 is actually quite appropriate in practice. As
noted by Randolph et al. (2000), the observed bear-
ing capacities of cylindrical and spherical penetro-
meters are in fact very similar in a wide range of soft
soils, despite theoretically predicted shape factors
(sphere ÷ cylinder) which average around 1.2 for
(b) Nc (UB) = 13.11
isotropic strength properties. A comparable experi-
mental investigation of plane strain and axisymmet-
ric ‘plate penetrometers’ would certainly be of inter-
est.

6 COMBINED LOADING

The bearing capacity of caissons under combined (V,


M, H) loading is of paramount importance for off-
shore applications. While rigorous bounds have been
developed for plane strain geometries (e.g. Salençon
& Pecker 1995, Houlsby & Puzrin 1999), corre-
(c) Nc (UB) = 13.31
sponding solutions for circular (or other) geometries
Figure 14. Stress and velocity fields for rough plate. remain a significant challenge.
1.2 7 CONCLUSIONS
1
This paper has endeavoured to summarise available
Horizontal, τ /s u

0.8 bound solutions for axially symmetric foundations in


FE sideswipe soil where the strength may vary linearly with depth.
0.6
Classical solutions stemming from the 1950s are put
Vertical UB in the context of more recent solutions, where im-
0.4
3-D (strips) provements have been made in accuracy (due to
0.2 modern computer capabilities) and rigour (with new
Factored 2-D techniques for lower and upper bound solutions).
0
Many previously published lower bounds have
0 2 4 6 8
not been rigorously proven by demonstrating exten-
Vertical, N c = q u /s u
sibility of the partial stress field. Methods of extend-
ing lower bound solutions using the method of char-
Figure 15. Yield envelopes in (V, H) space for rough circular
foundation on homogeneous soil. acteristics have been demonstrated here, for both
surface foundations and embedded caissons. The in-
troduction of an artificial free surface has been
While there is insufficient space to address this shown to yield significantly improved lower bounds
topic in detail in the present paper, it is interesting to for smooth-sided caissons.
consider how best to extend the plane strain solu- For surface foundations, an upper bound mecha-
tions to three dimensions, and the form of upper nism that is consistent with the lower bound stress
bound collapse mechanisms under general loading. characteristics allows exact bearing capacities to be
At present, finite element based solutions must be evaluated. However, for embedded caissons, the
relied on, either through the techniques of Sloan (apparently) consistent mechanism includes regions
(1988, 1989) for rigid plasticity, which have recently where the shear strain rate is of opposite sign to the
been extended to three-dimensional problems (Meri- lower bound shear stress, leading to very poor upper
field et al. 1999b), or using conventional elasto- bound estimates. Instead, a simple independent up-
plastic analysis. per bound mechanism, with only two parameters to
Figure 15 shows an example of a yield envelope be optimised, has been shown to furnish reasonably
in (V, H) space for a rough circular foundation rest- bracketed solutions, with uncertainty bands gener-
ing on the surface of a homogeneous soil. Two rig- ally better than ±10%.
orous upper bounds are indicated: a vertical cutoff at The challenge for the future is to derive rigorous
a bearing capacity of 6.05su; and a three-dimensional bound solutions for circular and square foundations
solution based on a series of parallel strips, each of under combined (V, M, H) loading. While the bear-
which deforms according to the classical Green ing capacity may be ascertained on a case-by-case
(1954) solution. The latter is a recent solution devel- basis from finite element analysis, an understanding
oped by Puzrin & Randolph (yet to be published). of the collapse mechanisms under combined loading
Also shown in Figure 15 is a result from an would be a major benefit in terms of permitting sim-
ABAQUS finite element analysis using the Fourier ple design calculations, particularly where the soil
mode for non-symmetric loading. A circular founda- stratigraphy is non-homogeneous.
tion resting on homogeneous soil has been subjected
first to bearing failure under purely vertical load, and
then a ‘sideswipe’ horizontal displacement at fixed ACKNOWLEDGEMENTS
vertical position. The computed vertical and hori-
zontal limits (normalised by su) are 5.91 and 1.03 re- The work described here forms part of the activities
spectively, compared with exact values of 6.05 and of the Special Research Centre for Offshore Founda-
1. Interestingly, although it would be expected that tion Systems, established and supported under the
the sideswipe load path would tend to lie marginally Australian Research Council's Research Centres
inside the true yield envelope (because of internal Program. This support is gratefully acknowledged.
plasticity within the soil mass), it actually crosses The authors would also like to thank Mr. A.
the upper bound solution. House for performing the ABAQUS analysis de-
The final curve on Figure 15 is a scaled version picted in Figure 15.
of the plane strain solution (Green 1954), where the
vertical load axis has been stretched by the shape
factor, 6.05/5.14 = 1.18. Does this represent a true
lower bound?
REFERENCES Randolph, M.F. 2000. Effect of strength anisotropy on capacity
of foundations. Proc. John Booker Memorial Symp., Sydney
Bishop, J.F.W. 1953. On the complete solution to problems of (in print).
deformation of a plastic-rigid material. J. Mech. Phys. Sol- Randolph, M.F. & G.T. Houlsby 1984. The limiting pressure
ids 2: 43-53. on a circular pile loaded laterally in cohesive soil. Géotech-
Booker, J.R. & E.H. Davis 1972. A general treatment of plastic nique 34(4): 613-623.
anisotropy under conditions of plane strain. J. Mech. Phys. Randolph, M.F., C.M. Martin & Y. Hu 2000. Limiting resis-
Solids 20: 239-250. tance of a spherical penetrometer in cohesive material.
Booker, J.R. & E.H. Davis 1977. Stability analysis by plasticity Géotechnique 50(5): 573-582.
theory. In C.S. Desai & J.T. Christian (eds), Numerical Rowe, R.K. 1978. Soil–structure interaction analysis and its
methods in geotechnical engineering: 719-748. New York: application to the prediction of anchor behaviour. PhD the-
McGraw-Hill. sis, University of Sydney.
Brinch Hansen, J. 1970. A revised and extended formula for Salençon, J. & M. Matar 1982. Capacité portante des fonda-
bearing capacity. Bulletin No. 28, Danish Geotechnical In- tions superficielles circulaires. Journal de Mécanique
stitute, Copenhagen: 5-11. théorique et appliquée 1(2): 237-267.
Cox, A.D., G. Eason & H.G. Hopkins 1961. Axially symmetric Salençon, J. & A. Pecker 1995. Ultimate bearing capacity of
plastic deformation in soils. Proc. R. Soc. London (Ser. A) shallow foundations under inclined and eccentric loads. I.
254: 1-45. Purely cohesive soil. Eur. J. Mech. (A) 14: 349-375.
Davis, E.H. & J.R. Booker 1973. The effect of increasing Shield, R.T. 1955. On the plastic flow of metals under condi-
strength with depth on the bearing capacity of clays. tions of axial symmetry. Proc. R. Soc. London (Ser. A) 233:
Géotechnique 23(4): 551-563. 267-287.
Eason, G. & R.T. Shield 1960. The plastic indentation of a Skempton, A.W. 1951. The bearing capacity of clays. Proc.
semi-infinite solid by a perfectly rough circular punch. J. Building Research Congress, London 1: 180-189.
Appl. Math. Phys. (ZAMP) 11: 33-43. Sloan, S.W. 1988. Lower bound limit analysis using finite ele-
Green, A.P. 1954. The plastic yielding of metal junctions due ments and linear programming. Int. J. Num. & Anal. Meth.
to combined shear and pressure. J. Mech. Phys. Solids 2: in Geomech. 12: 61-77.
197-211. Sloan, S.W. 1989. Upper bound limit analysis using finite ele-
Houlsby, G.T. & A.M. Puzrin 1999. The bearing capacity of a ments and linear programming. Int. J. Num. & Anal. Meth.
strip footing on clay under combined loading. Proc. R. Soc. in Geomech. 13: 263-282.
London (Ser. A) 455: 893-916. Szczepinski, W. 1979. Introduction to the mechanics of plastic
Houlsby, G.T. & C.P. Wroth 1982a. Direct solution of plastic- forming of metals. Warsaw: Polish Scientific.
ity problems in soils by the method of characteristics. Proc. Tani, K. & W.H. Craig 1995. Bearing capacity of circular
4th Int. Conf. on Num. Meth. in Geomech., Edmonton 3: foundations on soft clay of strength increasing with depth.
1059-1071. Soils and Foundations 35(4): 21-35.
Houlsby, G.T. & C.P. Wroth 1982b. Determination of Vesic, A.S. (1975). Bearing capacity of shallow foundations. In
undrained strengths by cone penetration tests. Proc. 2nd H.F. Winterkorn & H.Y. Fang (eds), Foundation engineer-
Eur. Symp. on Penetration Testing, Amsterdam 2: 585-590. ing handbook: 121-147. New York: Van Nostrand.
Houlsby, G.T. & C.P. Wroth 1983. Calculation of stresses on
shallow penetrometers and footings. Proc. IUTAM / IUGG
Symp. on Seabed Mech., Newcastle upon Tyne: 107-112.
Hu, Y. & M.F. Randolph 1998. H-adaptive FE analysis of
bearing capacity of skirted foundations. Proc. Int. Conf. on
Offshore and Polar Eng., Montreal 1: 549-556.
Hu, Y., M.F. Randolph & P.G. Watson 1999. Bearing response
of skirted foundations on non-homogeneous soil. J. Geo-
tech. Eng. Div. ASCE 125(11): 924-935.
Ishlinskii, A.I. 1944. Osesimmetrichnaia zadacha teorii plas-
tichnosti i proba Brinellia. Prikl. Mat. i Mekh. 8(3): 201-
208.
Kusakabe, O., H. Suzuki & A. Nakase 1986. An upper-bound
calculation on bearing capacity of a circular footing on a
non-homogeneous clay. Soils and Foundations 26(3): 143-
148.
Levin, A. 1955. Indentation pressure of a smooth circular
punch. Quart. Appl. Math. 13: 381-389.
Martin, C.M. 1994. Physical and numerical modelling of off-
shore foundations under combined loads. D.Phil. thesis,
University of Oxford.
Martin, C.M. & M.F. Randolph 2000. Limit analysis of flow-
round penetrometers in clay. Submitted for publication.
Merifield, R.S., S.W. Sloan & H.S. Yu 1999a. Stability of plate
anchors in undrained clay. Res. Rept No. 174.02.1999, Dept
of Civil Eng., University of Newcastle, Australia.
Merifield, R.S., A. Lyamin, S.W. Sloan & H.S. Yu 1999b.
Three dimensional lower bound solutions for the stability of
plate anchors in clay. Res. Rept No. 178.02.1999, Dept of
Civil Eng., University of Newcastle, Australia.
Meyerhof, G.G. 1951. The ultimate bearing capacity of founda-
tions. Géotechnique 2(4): 301-332.

View publication stats

You might also like