Download as pdf or txt
Download as pdf or txt
You are on page 1of 11

Undrained Bearing Capacity of Square

and Rectangular Footings


Susan Gourvenec1; Mark Randolph2; and Oliver Kingsnorth3

Abstract: The uniaxial vertical bearing capacity of square and rectangular footings resting on homogeneous undrained clay is investi-
gated with finite element analyses, using both Tresca and von Mises soil models. Results are compared with predictions from conventional
bearing capacity theory and available analytical and numerical solutions. By calibrating the finite element results against known exact
solutions, best estimates of bearing capacity for rough-based rectangular footings are derived, with the shape factor fitted by a simple
quadratic function of the footing aspect ratio. For a square footing, the bearing capacity is approximately 5% lower than that based on
Skempton’s shape factor of 1.2.
DOI: 10.1061/共ASCE兲1532-3641共2006兲6:3共147兲
CE Database subject headings: Shallow foundations; Numerical analysis; Analytical techniques; Bearing capacity; Vertical loads;
Footings.

Introduction and is independent of footing roughness. For circular geometry


the corresponding bearing capacity factor, Nc = Vult / Asu, lies be-
Traditional bearing capacity theory, developed based on condi- tween 5.69 and 6.05 for smooth and rough footings respectively
tions of plane strain 共Terzaghi 1943兲, is appropriate for strip foot- 共Cox et al. 1961兲.
ings where the length of the footing is “long” relative to its No exact solution exists for the bearing capacity of a square
breadth. Many shallow foundations, particularly those used off- footing, even for the simple case of a surface footing resting on an
shore, have a lower aspect ratio with a rectangular or square foot- isotropic, homogeneous deposit obeying an undrained failure cri-
print. Edge effects improve bearing capacity 共per unit area兲 of a terion. However, the bearing capacity factors for square and rect-
three-dimensional footing compared to a strip footing as slip angular footings are expected to be bracketed by those for strip
planes must develop around the perimeter of the three- and circular geometries 共Levin 1955兲.
dimensional footing as opposed to only adjacent to the “long” Few analytical or numerical studies of the bearing capacity of
edge under conditions of plane strain. square and rectangular footings have been carried out. Shield and
Edge effects are typically accounted for in bearing capacity Drucker 共1953兲 proposed an upper bound solution for the ultimate
calculations by an empirical shape factor applied to the basic limit state under uniaxial vertical loading for smooth rectangular
equation for conditions of plane strain 共Terzaghi 1943兲. The most footings, which gave a bearing capacity factor for a square foot-
widely adopted shape factor is that proposed by Skempton ing of Nc = 5.71. Some 5 decades later, Michalowski and Dawson
共1951兲, which adjusts bearing capacity linearly with the aspect 共2002兲 proposed a lower bearing capacity factor of Nc = 5.43,
ratio; for a square footing the ultimate limit state is taken as 20% based on finite-difference analyses using FLAC 共Itasca 1997兲.
higher than for a strip footing. Skempton’s expression was re- This is in line with the suggestion by Levin 共1955兲, that the bear-
tained for the undrained case in later work by De Beer 共1970兲 and ing capacity of a square footing should be slightly less than that
Hansen 共1970兲. of a circular footing of equivalent area.
The vertical bearing capacity of a strip footing resting on a Michalowski 共2001兲 proposed an upper bound solution for a
homogeneous Tresca material is given by Vult = 5.14Asu 共Prandtl rough square footing, giving a bearing capacity factor of
1921兲, where A⫽area of the footing and su the soil shear strength, Nc = 6.56; this is approximately 10% higher than the bearing ca-
pacity factor for a rough circular footing, contrary to Levin’s sup-
1
Senior Lecturer, Centre for Offshore Foundation Systems, Univ. of position. Salgado et al. 共2004兲 presented results of numerical limit
Western Australia, Crawley, Perth, WA 6009, Australia 共corresponding analyses, which indicate lower and upper bounds to the bearing
author兲. E-mail: susan@civil.uwa.edu.au capacity factor Nc of 5.52 and 6.22 for a rough square footing.
2
Professor, Centre for Offshore Foundation Systems, Univ. of Western The corresponding bound solutions given by Salgado et al. for the
Australia, Crawley, Perth, WA 6009, Australia. E-mail: randolph@ circular case were 5.86 and 6.23, which is a narrower range than
civil.uwa.edu.au for the more difficult square geometry and spans the exact solu-
3
Student, Centre for Offshore Foundation Systems, Univ. of Western tion of Nc = 6.05 for a rough circular footing. Comparing the
Australia, Crawley, Perth, WA 6009. E-mail: plopdiact@hotmail.com lower and upper bounds for the two cases suggests that the true
Note. Discussion open until October 1, 2006. Separate discussions
capacity for a square foundation is lower than for the equivalent
must be submitted for individual papers. To extend the closing date by
one month, a written request must be filed with the ASCE Managing circular case, as supposed by Levin 共1955兲.
Editor. The manuscript for this paper was submitted for review and pos- This paper presents finite element results for smooth and rough
sible publication on January 31, 2005; approved on April 28, 2005. This square footings, and proposes best estimates of the bearing capac-
paper is part of the International Journal of Geomechanics, Vol. 6, No. ity factors for rough rectangular footings of varying aspect ratio.
3, May 1, 2006. ©ASCE, ISSN 1532-3641/2006/3-147–157/$25.00. The kinematic mechanisms accompanying failure of the

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006 / 147


Fig. 1. Finite element meshes: 共a兲 square footing L = B; 共b兲 rectangular footing with L = 5B

square and rectangular footings observed in the finite element rection to maintain uniform element size across the models and
analyses are also presented. The mechanism for a square footing the rear mesh boundary was positioned sufficiently remote not to
indicates fourfold symmetry, with displacements orthogonal to affect footing behavior. For example the mesh for the L = 5B foot-
the edges but also symmetric about the diagonals; this contrasts ing 关shown in Fig. 1共b兲兴 comprises 18,000 elements, and the mesh
with the optimal upper bound solutions of Shield and Drucker and for the L = 10B footing comprises 25,000 elements 关more than
Michalowski, where the deformation fields along parallel edges of twice as many as the square footing mesh shown in Fig. 1共a兲兴.
the footing differ from those along the orthogonal edges. It is A plane strain mesh was also constructed using first order fully
hoped that the mechanisms observed in the finite element analy- integrated quadrilateral hybrid elements. The same geometry and
ses may provide a useful starting point for an alternative upper discretization as the front face of the three-dimensional mesh was
bound solution. used and equivalent boundary conditions, soil conditions, and
analysis procedures were modeled.
A vertical bearing capacity factor of Nc = 5.31 was obtained
Finite-Element Model from the plane strain finite element analysis, over predicting the
exact solution by just 3%. A much finer plane strain mesh was
All the finite element analyses were carried out using the software
also constructed using 1,500 elements, two and a half times as
ABAQUS 共HKS 2002兲.
many elements as the coarser mesh. The finer mesh gave a verti-
cal bearing capacity factor Nc = 5.27, only a 0.5% increase in ac-
Mesh curacy, indicating that the coarser mesh was sufficiently fine.
The three-dimensional finite-element mesh used for analysis of a
square footing of width, B, and equal length, L, is shown in Fig.
Material Properties
1共a兲. It represents a half-footing cut through one of the orthogonal
planes of symmetry. The mesh extends 3B from the edges of the Simple soil conditions were modeled representing an isotropic
footing and 2.5B beneath the footing. Zero-displacement bound- linear elastic-perfectly plastic material failing according to a
ary conditions prevent out-of-plane displacements of the vertical Tresca flow rule with uniform undrained shear strength with
boundaries, and the base of the mesh is fixed in all three coordi- depth, su. The conditions considered are intended to represent a
nate directions. fine grained material subjected to a period of loading sufficiently
A number of different mesh densities were investigated to short that no drainage will take place. Constant stiffness index
achieve a time-efficient model without compromising accuracy.
Eu / su = 1,000, Poisson’s ratio ␯ = 0.49, and buoyant unit weight
The mesh shown in Fig. 1共a兲 comprises approximately 12,000
␥⬘ = 8 kN/ m3 were prescribed. Although soil self-weight was
first order fully integrated hexahedral hybrid elements.
used in the analysis, the nature of the problem considered, i.e., a
The hybrid element formulation uses a mixture of displace-
ment and stress variables 共as opposed to solely displacement兲 to surface footing resting on an isotropic homogeneous soil without
approximate the equilibrium equations and compatibility condi- drainage, leads to the calculated bearing capacity being unaf-
tions. Hybrid elements are recommended for modeling the re- fected by the value of ␥⬘.
sponse of near incompressible materials 共such as is appropriate A Tresca soil model was adopted in order to allow direct com-
for undrained soil conditions兲. parison with other published solutions based on the same failure
Six three-dimensional models were created to investigate the criterion. However, for completeness, solutions are also provided
effect of varying footing length to breadth aspect ratio on bearing for rough rectangular footings resting on soil modeled using a von
capacity 关L / B = 1 共square兲, 1.5, 2, 3, 5, and 10兴. Each mesh rep- Mises failure criterion.
resents a half-footing, measured along the longitudinal axis, and The footing was represented as a discrete rigid body resting on
maintains the same geometry and discretization on the front face the surface of the soil. Most of the analyses considered a fully
of the mesh as for the square footing shown in Fig. 1共a兲. For the rough interface on the underside of the footing, but a frictionless
longer footings more elements were used in the longitudinal di- interface was also modeled for a square footing for comparison.

148 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006


Load Paths Square Footings
Analyses were carried out to investigate the ultimate limit state Bearing capacity factors 共Nc兲 and corresponding shape factors
under uniaxial vertical loading. The foundation was subjected to 共sc = Nc / Nc-strip兲 from Skempton’s empirical expression, other
controlled displacements, as opposed to directly applied loads, to available analytical and numerical solutions for square footings,
enable postfailure conditions to be observed. The reference point and those from this study are compared in Table 1. The shape
for the applied displacement was taken as the midpoint of the factors are also presented graphically in Fig. 2. Bearing capacity
foundation base. and shape factors for circular geometry are presented for
completeness.
For plane strain conditions the bearing capacity is not affected
by the interface condition between the foundation and the soil,
Results
with Nc = 5.14 an exact solution for any roughness. For three-
dimensional footing geometries interface roughness does affect
Accuracy of Finite Element Analyses the bearing capacity. Exact solutions for circular geometry 共Cox
et al. 1961兲 give bearing capacity shape factors sc 共=Nc / Nc-strip兲
Before presenting results for rectangular footings, the accuracy of
= 1.11 and 1.18 for smooth and rough footings, respectively,
the finite element analyses is explored with reference to capacities
marking the upper limit to the expected bearing capacity factor
computed for smooth- and rough-based circular footings. Finite
for equivalent square footings 共Levin 1955兲.
element analyses of a rough-based circular footing on a Tresca
Skempton’s empirical shape factor does not distinguish be-
material were reported by Gourvenec and Randolph 共2003兲 and
tween a smooth or rough footing interface, suggesting a bearing
Taiebat and Carter 共2000兲. In both cases the bearing capacity
capacity factor of Nc = 6.17 共corresponding to a shape factor
factor Nc was underpredicted compared to the exact solution of
sc = 1.2兲 for either case. Skempton’s factor is therefore non-
Nc = 6.05 共Cox et al. 1961兲, by 2.3% by Gourvenec and Randolph,
conservative by at least 2% for a rough square footing 共assuming
and by 6% by Taiebat and Carter.
the true bearing capacity is at most equal to that for a circular
Gourvenec and Randolph 共2003兲 note that the slight underpre- foundation兲, and by more than 9% if applied, incorrectly, to a
diction in the finite element calculation of the bearing capacity of smooth-based footing.
the rough circular footing may have been due, at least in part, to
larger distortion of the elements local to the footing in the three-
dimensional analysis. Element distortion in the three-dimensional Smooth Footing Interface
mesh is greater than in the two-dimensional mesh because of the For a smooth square footing on a Tresca soil Shield and Drucker
different kinematic mechanisms under vertical loading: in-plane 共1953兲 proposed an upper bound solution suggesting a bearing
for the strip footing and axisymmetric for the circular footing. capacity factor Nc = 5.71. Michalowski and Dawson 共2002兲 report
There is also evidence that underprediction of vertical bearing finite difference results obtained with FLAC 共Itasca 1997兲 of a
capacity in three-dimensional analyses may be due to rounding smooth square footing on a Tresca material suggesting a bearing
the vertices of the hexagonal Tresca yield surface in finite- capacity factor Nc = 5.43, approximately 5% lower than the Shield
element software 共Taiebat 2005兲. Rounding at the vertices is nec- and Drucker solution. No bearing capacity for the strip footing
essary as the point of the vertex causes a discontinuity of the yield case, or for a circular foundation, is presented so the accuracy of
surface, causing a singularity and resulting in computational dif- the FLAC model cannot be assessed. The shape factor relative to
ficulties. Conventionally the vertices of the surface are smoothed a strip footing is sc = 1.06.
by an arc taken 5° either side of the point. The vertices of the The lower bearing capacity reported by Michalowski and
hexagonal Tresca surface represent conditions where two of the Dawson is attributed to a failure mechanism that takes advantage
principal stresses are equal 共such as in triaxial compression and of diagonal symmetry, i.e., planes of symmetry from corner to
extension兲 and rounding of the vertex causes an underprediction corner as well as orthogonal to the edges of the footing. This is in
of vertical load capacity as the modified yield surface falls inside contrast to Shield and Drucker’s mechanism, which involves only
the true Tresca surface. The larger the arc angle the further within two planes of symmetry, running orthogonal to the footing edges,
the vertex the modified surface lies and the greater the underpre- and constraining all soil displacements to be normal to the rel-
diction of vertical capacity. Taiebat 共2005兲 has observed a 1.5% evant edge.
increase in bearing capacity by decreasing the arc angle from 5 The finite element analysis for a smooth-based square footing
to 0.5°. carried out in this study also indicates fourfold symmetry of the
For the element type and mesh density adopted in this study, deformation mechanism 共that is, including diagonal symmetry兲. A
the bearing capacity factors computed using the ABAQUS imple- slightly higher bearing capacity factor of Nc = 5.56 was obtained
mentation of the Tresca failure criterion are 5.31 for a strip foot- compared with the value of 5.43 from Michalowski and Dawson,
ing, and 5.58 and 5.96, respectively, for smooth- and rough-based giving a shape factor of 1.08. A corresponding analysis for a
circular footings. These values represent overprediction for the smooth-based circular foundation gave Nc = 5.58, suggesting that
strip footing by 3.3%, and underprediction by 1.9 and 1.5% for square and circular bearing capacities are extremely close, and in
the smooth- and rough-based circular footings. These results sug- line with Levin’s suggestion that the bearing capacity of a square
gest that the finite element results of bearing capacity of square footing would lie slightly below that for an equivalent circular
footings would be similarly underpredicted, but that the reverse footing 共Levin 1955兲.
may be true for rectangular footings of high aspect ratio. In arriv-
ing at a best estimate of shape factors for rectangular footings, it Rough Footing Interface
is necessary to adjust the finite element results to reflect the vary-
ing error, which has been estimated here as varying linearly with Michalowski 共2001兲 proposed an upper bound solution for a
the aspect ratio, L / B, of the footing. rough square footing on Tresca soil, which gives a bearing capac-

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006 / 149


Table 1. Comparison of Undrained Bearing Capacity and Shape Factors; Square and Circular Footings 共Tresca Soil兲
Reference Analysis type Smooth/rough Nc s ca
Strip Prandtl 共1921兲 Exact 5.14 1.0
Circle Cox et al. 共1961兲 Exact Smooth 5.69 1.11
Rough 6.05 1.18
Salgado et al. 共2004兲 Numerical limit analysis Rough 5.86 LB 1.14
6.23 UB 1.21
Plane strain 5.13 LB
5.20 UB
This study Finite element analysis Smooth 5.58 1.09
Rough 5.96 1.16
Plane strain 5.27 1.03
Square Skempton 共1951兲 Empirical — 6.17 1.2
Shield and Drucker 共1953兲 Upper bound Smooth 5.71 1.11
Michalowski and Dawson 共2002兲 Finite difference Smooth 5.43 1.06
This study Upper bound Roughb 6.41 1.25
Michalowski 共2001兲 Upper bound Rough 6.56 1.28
Salgado et al. 共2004兲 Numerical limit analysis Rough 5.52 LB 1.07
6.22 UB 1.21
Plane strain 5.13 LB
5.20 UB
This study Finite element analysis Smooth 5.56 1.08
Rough 5.91 1.15
Plane strain 5.31 1.03
a
Indicates shape factor calculated assuming exact plane strain Nc = 5.14.
b
See Appendix.

ity factor of Nc = 6.56. The solution is based on a different mecha- eled with similar mesh density and element type 共Nc = 5.96兲, but
nism to that proposed by Shield and Drucker for a smooth square lower than the exact solution for a circle of 6.05. The finite-
punch, with a 共Prandtl-type兲 central pyramid moving downward element results for the rough footings are in line with Levin’s
as opposed to a Hill-type double-wedge mechanism with soil di- suggestion, as were the results from the smooth footing analyses,
rectly beneath the footing sliding laterally outwards. Additionally that the bearing capacity of a square footing would lie slightly
the central downward moving block in Michalowski’s mechanism below that for an equivalent circular footing 共Levin 1955兲. The
is flanked by multiple sliding blocks with shearing taking place shape factor relative to the plane strain solution is sc = 1.15, which
between each of the blocks. is lower than Skempton’s proposed factor of 1.2. However, it is
It is possible to modify Shield and Drucker’s upper bound not possible to gauge the extent of potential underprediction by
mechanism to a Prandtl-type with a central pyramid, but retaining
the other features of their mechanism. The details are given in the
Appendix, and the modified mechanism leads to an upper bound
for a rough-based square footing of Nc = 6.41, slightly below
Michalowski’s value.
Numerical limit solutions for a rough square footing on a
Tresca soil were obtained by Salgado et al. 共2004兲, identifying
lower and upper bounds for the bearing capacity factor in the
range 5.52艋 Nc 艋 6.22, an envelope of 12%. Lower and upper
bounds for plane strain conditions from the same limit solutions
gave a range of bearing capacity factor of 5.13艋 Nc 艋 5.20. Al-
though the Salgado et al. upper bound gives a lower bearing ca-
pacity than the Michalowski or 共equivalent兲 Shield and Drucker
upper bound solutions, it is still on the high side compared with
the bearing capacity for a rough circular footing 共Nc = 6.05兲 and
the range is relatively wide.
Salgado et al.’s investigation addressed a broader range of
conditions than this study, considering effects of footing embed-
ment as well as plan aspect ratio. Interestingly they found that, as
the depth of embedment increased, coupling of conventional
shape and depth factors often led to conservative estimates of
bearing capacity compared to their limit solutions.
The finite element results from this study gave a bearing ca- Fig. 2. Comparison of shape factors for square footings from finite-
pacity factor for a rough square footing Nc = 5.91, slightly less element analyses and available published data 共all Tresca soil兲,
than the finite element solution for a rough circular footing mod- compared with exact solutions for circular footings

150 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006


Table 3. Adjusted Shape and Bearing Capacity Factors for Undrained
Vertical Loading of Rectangular Footings
L/B 1 1.5 2 3 5 10 ⬁
sc 1.15 1.11 1.09 1.06 1.04 1.02 1.00
Nc 5.91 5.72 5.61 5.47 5.36 5.26 5.14

overprediction for the square footing 共at least, relative to the


finite- element result兲. The upper bound results from Michalowski
共2001兲 are somewhat irregular in their trends, and mostly lie
above the upper bound based on Shield and Drucker 共Appendix兲.
An exception is the point for L / B = 10, where the Michalowski
共2001兲 upper bound is surprisingly low, with Nc = 5.20.
Discrete values of the calculated bearing capacity factors and
shape factors are presented in Table 2, together with the average
of the numerical lower and upper bounds from Salgado et al.
Fig. 3. Bearing capacity factor as function of footing aspect ratio.
共2004兲. Tentatively, taking the finite element result of 5.91 for a
Finite element results compared with available published data 共all
square footing as “correct” 共being consistent with Levin’s hypoth-
Tresca soil兲.
esis that it should be just less than the value for a circular foot-
ing兲, the finite element results can be adjusted so as to eliminate
the finite element analysis, as occurred for the circular footing. A the gradual overprediction as the footing tends toward a strip,
possible indication is to take the average of the Salgado et al. reducing them by a factor of 0.032共1 − B / L兲. This leads to the
bounds, which leads to an excellent estimate of 6.04 for the bear- corrected shape and bearing capacity factors given in Table 3. A
ing capacity factor of a circular footing 共bounds of 5.86⬍ Nc consistent adjustment of the upper bound results, ranging from
⬍ 6.23兲. This approach would give 5.87 for a square footing, zero adjustment for a strip footing, up to 8% adjustment for a
which is within 1% of the finite element value of 5.91, supporting square footing, leads to virtually identical shape factors, as shown
a bearing capacity factor close to 5.9 for a rough-based square in Fig. 4.
foundation. The relationship between shape factor and footing aspect ratio
obtained from the finite element and upper bound results adjusted
in this way can be accurately expressed by the simple quadratic
Rectangular Footings function
In addition to the square footing analyses, six analyses were car-
sc = 1 + 0.214B/L − 0.067共B/L兲2 共1兲
ried out on rectangular footings to investigate the effect of aspect
ratio 共length to breadth, L / B兲 on uniaxial vertical bearing capac- This gives marginally higher shape factors for very slender foot-
ity. All the rectangular footing analyses modeled a rough interface ings than Skempton’s original recommendation 共a linear variation
on the underside of the footing. with coefficient of 0.2兲, but with reduced shape functions for
Fig. 3 compares bearing capacities as a function of aspect ratio square and rectangular footings of low aspect ratio.
for square and rectangular footings with a rough interface from
the finite element results from this study with Skempton’s empiri-
Comparison of Bearing Capacity Predictions
cal expression and other available bound solutions. As would be
with Tresca and von Mises
expected, there is a clear trend in each case of increasing bearing
capacity as the aspect ratio of the footing increases towards unity. All the previous studies discussed in this paper 共Terzaghi 1943;
For aspect ratios 共L / B兲 less than 5, the finite element results ap- Shield and Drucker 1953; Levin 1955; Cox et al. 1961; Taiebat
pear approximately midway between the bounds of Salgado et al. and Carter 2000; Michalowski 2001; Michalowski and Dawson
共2004兲, but the trend of the results gives increasing overprediction 2002; Gourvenec and Randolph 2003; Salgado et al. 2004兲 as
as the footing tends toward a strip. By contrast, the upper bound well as the present study, are based on a Tresca failure criterion,
results give the correct solution for a strip footing, but up to 8% therefore neglecting the intermediate principal stress component.

Table 2. Comparison of Undrained Bearing Capacity and Shape Factors for Rough-Based Rectangular Footings 共Tresca Soil兲
Skempton 共1951兲 Michalowski 共2001兲 Salgado et al. 共2004兲 This study

Nc Nc Nc Nc Nc s ca
L/B Nc sc Nc 共LB兲 共UB兲 共average兲 共UB兲 共FE兲 共FE兲
1 6.17 1.20 6.56 5.52 6.22 5.87 6.41 5.91 1.15
1.5 5.83 1.13 6.19 — — — 6.03 5.78 1.12
2 5.65 1.10 6.06 5.36 6.02 5.57 5.83 5.69 1.11
3 5.48 1.07 5.64 5.25 5.89 5.51 5.62 5.58 1.09
5 5.35 1.04 5.49 5.17 5.78 5.47 5.44 5.49 1.07
10 5.24 1.02 5.20 — — — 5.30 5.41 1.05
⬁ 5.14 1.00 5.14 5.13 5.20 5.17 5.14 5.31 1.03
a
sc calculated assuming plane strain Nc of 5.14.

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006 / 151


Fig. 5. Comparison of calculated bearing capacity factors; Tresca
and von Mises
Fig. 4. Best-estimate shape factors for square and rectangular
footings on Tresca soil. Comparison with uncorrected finite element
results and Skempton’s empirical shape factor expression. Kinematic Mechanisms at Failure
The upper bound solutions based on the mechanisms proposed by
Shield and Drucker 共1953兲 and Michalowski 共2001兲 lead to rela-
tively high bearing capacities. Intriguingly, the optimum mecha-
Additional finite element analyses were carried out with a von
nisms for a square footing turn out to have only twofold symme-
Mises failure criterion to assess the effect on bearing capacity of
try, with four plane strain regions where movement is orthogonal
including the intermediate principal stress. Table 4 compares the
to the edges of the footing, but lacking diagonal symmetry. By
finite element bearing capacity and shape factors calculated using
contrast, the finite element results for square footings indicate a
both Tresca and von Mises failure criteria. The von Mises bearing
failure mechanism that has fourfold symmetry, including being
capacity factors have been calculated based on the shear strength
symmetric about the diagonals of the footing. The kinematic fail-
in plane strain 共where su is equal to the von Mises parameter, k兲,
ure mechanisms observed in the finite element analyses from this
rather than the shear strength under triaxial conditions which is
study are discussed in more detail below, in the hope that they
13% lower. It can be seen that at the limit when B / L = 1 the
may form the basis of improved upper bound solutions.
Tresca analysis predicts a bearing capacity 3% higher than using
von Mises 共based on plane strain shear strength兲, the disparity
diminishing as the footing length increases until the solutions Rough Footing Interface
converge for plane strain conditions. Fig. 6 shows contours of resultant soil displacement at failure for
Randolph 共2000兲 suggested that bearing capacity factors for each of the footings modeled. Under plane strain conditions a
three-dimensional problems should be expressed in terms of an distinct Prandtl-type mechanism is observed in the finite element
average of triaxial compression 共suC兲, triaxial extension 共suE兲, and analyses 关Fig. 6共a兲兴, of the form illustrated schematically in the
simple shear 共suSS兲 strengths; for the von Mises failure criterion left side of Fig. 7 共Prandtl 1921兲. Angles ABC, EBD, and BED
the average strength may be expressed as are 45°, wedge ABC moves downward as a rigid block, equalized
by lateral and upward movement of rigid wedge BDE, separated
suC + suE + suSS 冑3 + 1 by a 90° slip fan zone DBC, where surface CD is a circular arc.
suAV = = suSS = 0.91suSS 共2兲 A plane strain Prandtl-type mechanism is also observed at the
3 3
midline cross section of the footings with aspect ratios L / B of 10
While this might be an appropriate choice for square 共or circular兲 and 5 关Figs. 6共b and c兲兴. As the footing becomes shorter the soil
footing geometry the applicability breaks down for gradually wedge underneath the footing becomes shallower and an addi-
more slender rectangles as they tend to plane strain. A comparison tional sliding wedge develops 关Fig. 6共d兲兴. The observed mecha-
of the various bearing capacity factors calculated using the Tresca nism is illustrated schematically in the right side of Fig. 7 in the
failure criterion and the von Mises criterion, normalizing the bear- form proposed by Kusakabe et al. 共1986兲. Angle AB⬘C⬘ decreases
ing capacity by either the plane strain, triaxial compression 共and as wedge AB⬘C⬘ becomes shallower. A new rigid wedge B⬘C⬘F,
extension兲 or the average shear strength, is shown in Fig. 5. It is traveling at an angle ␣ to the horizontal, develops between the
clear that use of the plane strain, or simple shear, strength with the central downward moving wedge of soil and the 90° slip fan zone
von Mises failure criterion gives the closest agreement with the D⬘B⬘F. The displacement of the sliding rigid wedges beneath the
Tresca results, and even for a square footing is better than using footing are equalized by lateral and upward movement of rigid
the average shear strength. This implies that only a small propor- wedge B⬘D⬘E⬘ similar to the plane strain mechanism. Note that
tion of the soil is loaded under 共quasi-兲 triaxial conditions for the for ␤ = ␲ / 4 the general mechanism reverts to a Prandtl-type
vertical bearing capacity problem. mechanism, while for ␤ = 0, a Hill-type mechanism is obtained. In

Table 4. Undrained Bearing Capacity Factors for Rectangular Footings from Finite Element Analyses Using Tresca and von Mises Failure Criteria
L/B 1 1.5 2 3 5 10 ⬁
Tresca Nc 5.91 5.78 5.69 5.58 5.49 5.41 5.31
von Mises 共plane strain su兲 Nc 5.74 5.68 5.60 5.51 5.44 5.38 5.31

152 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006


Fig. 8. Extent of failure mechanisms as function of footing aspect
ratio

Fig. 6. Kinematic failure mechanisms of footings of various aspect


ratio 共rough interface兲

the present case, for the square footing, a wedge angle ␤, 共i.e.,
AB⬘C⬘兲 of approximately 23° is observed 共compared with 45° in
the classical Prandtl plane strain mechanism兲.
Fig. 8 presents graphically the zone of influence of the failure
mechanisms at the midpoint cross section, in terms of lateral ex-
tent, i.e., to either side of the footing, and depth beneath the
footing. A linear relationship is evident between the extent of the
failure zone and the aspect ratio of the footing.
Consideration of the regions of displaced soil, viewed in plan
as illustrated in Figs. 9共a–f兲, shows the extent of the failure
mechanism in the longitudinal direction 共i.e., along the long axis兲
is independent of footing aspect ratio, and equal to approximately
0.6B. Diagonal symmetry of the failure mechanism for the square
footing is also clearly illustrated by the displacement contours
shown in Fig. 9.
It could also be surmised from the displacement contours in
Fig. 9 that plane strain conditions prevail approximately 1.5B
from the end of the footing, measured along the long axis.

Smooth Footing Interface


Fig. 9. Plan view of failure mechanisms 共rough footings兲
The kinematic mechanism at failure observed in the finite-
element analysis for the smooth square footing is illustrated in
Fig. 10, indicating a double wedge Hill-type mechanism. No cen-
tral wedge is evident 共i.e., ␤ = 0, Fig. 7兲, with slip taking place on
the underside of the footing with the soil beneath the footing

Fig. 7. Plane strain and 3D failure mechanisms Fig. 10. Kinematic failure mechanisms of smooth square footing

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006 / 153


Fig. 11. Displacement vectors at failure for square footings: 共a兲 rough 共Prandtl-type兲; 共b兲 smooth 共Hill-type兲

moving out toward the edge of the footing 共at an angle ␣, Fig. 7兲. 1. Bearing capacity of a square footing on a Tresca soil is ap-
Some soil movement directly below the center of the footing, to a proximately 3% less than the bearing capacity of an equiva-
depth of approximately 0.18B, is evident. Displacement vectors lent circular footing. This study suggests a bearing capacity
from the finite element analyses for rough and smooth square factor of Nc = 5.9, equivalent to a shape factor sc = 1.15,
footings are shown in Fig. 11 and clearly indicate the different would be appropriate for a rough square footing and
direction of soil movement directly below the footing. Nc = 5.56, sc = 1.08 for a smooth square footing.
The mechanism accompanying failure of the smooth footing, 2. Bearing capacity of a rough square footing on a von Mises
seen in Fig. 10, is much shallower than the Prandtl-type mecha- soil assuming plane strain shear strength calculated in the
nism observed with the rough footing, as seen in Fig. 7共a兲. The finite element analyses is 3% less than a rough square footing
deepest part of the failure mechanism observed for the smooth on a Tresca soil. The choice of shear strength to use in con-
square footing taking place at 0.28B, compared with a maximum junction with a von Mises model can lead to a difference in
depth of 0.53B for the rough square footing failure mechanism. bearing capacity of up to 15%.
Contours of resultant displacement viewed in plan over the 3. If a “rounded” Tresca flow potential were used in the devia-
rough and smooth square footings are shown in Fig. 12. These toric plane 共as opposed to the von Mises circle used in
also illustrate the variation in nature of the soil displacement de- ABAQUS兲 a slightly higher value of bearing capacity would
pending on interface roughness. The footings are subjected to a be expected. It would be interesting to see the results pre-
uniform vertical displacement; the contours of resultant displace- sented in this study compared with results from numerical
ment beneath the rough footing show that the soil moves uni- software that employs a rounded Tresca flow potential.
formly vertically downwards with no relative soil movement on 4. Comparison of the conventional linear expression to account
the underside of the footing. The contours beneath the smooth for three-dimensional footing geometry sc = 1 + 0.2B / L
footing show a variation of soil displacement across the footing 共Skempton 1951兲 with the finite element results from this
area, the maximum soil movement occurring at the footing pe- study suggest that the former might overpredict bearing ca-
riphery with less displacement beneath the center. Fig. 12 also pacity for a square footing on a Tresca soil by up to 5% or
shows the mechanism accompanying failure of the smooth square 12% depending on interface roughness.
footing exhibited diagonal symmetry similar to the rough square 5. The finite element results from this study suggest the rela-
footing 关as seen in Fig. 9共a兲兴. tionship between shape factor and footing aspect ratio for
rough rectangular footings can be accurately predicted by a
quadratic polynomial
Conclusions sc = 1 + 0.214B/L − 0.067共B/L兲2
The results presented in this paper address the uniaxial vertical 共Rectangular footings with a smooth interface were not con-
bearing capacity of square and rectangular footings under sidered in this study.兲
undrained conditions on a uniform soil. Best-estimate bearing ca- 6. Available upper bounds of rough square and rectangular
pacity factors for square and rectangular footings are presented. footings by Michalowski 共2001兲 and Salgado et al. 共2004兲,
Key findings of this study are: and with the upper bound solution presented in this paper

Fig. 12. Plan view of displacement contours at failure for square footings: 共a兲 rough; 共b兲 smooth

154 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006


based on Shield and Drucker’s mechanism, overpredict bear- nism comprises a plane strain Hill-type mechanism over the cen-
ing capacity. Taking the bearing capacity of a circular footing tral part of the longer sides. At the ends of the longer sides, and
as an upper limit to the expected bearing capacity of a square over the shorter side, the size of the mechanism gradually reduces
footing 共after Levin 1955兲, the available upper bound solu- as the corner is approached although the soil motion remains
tions overpredict capacity by at least 3–10% 共depending on normal to each corresponding side. The sections MTUV and
the study兲. Alternatively, comparison with the bearing capac- MQRX are parts of right circular cones, with M the apex and
ity factors recommended by this study suggests an overpre- TUV 共or QRX兲 the base of the cone, perpendicular to the axes MT
diction of 6–13% 共depending on the study兲. or MX.
7. The numerical results from this study, and those presented by Shield and Drucker state the upper bound bearing factor, Nc as


Michalowski and Dawson 共2002兲, show failure mechanisms
for square footings exhibit fourfold symmetry 共i.e., deforma- c
Nc = 2共␣2 + ␤2兲 + cot ␣2 + cot ␤2 + ␣1 + ␤1 + 共␣1 + ␤1
tion about planes orthogonal to the edges and corner to cor- 2a
ner兲. In contrast, existing upper bound solutions are based on
mechanisms with only two planes of symmetry 共orthogonal
to the edges兲.
+ cot ␣1 + cot ␤1兲 冑
1+
c2
b2 sin2 ␤1
+ 共␣2 + ␤2 + cot ␣2

8. Considering the geometry of the kinematic mechanisms ac-


companying failure in the finite element analyses of rough
rectangular footings various observations were made: 共1兲 A
+ cot ␤2兲 冑
1+
b2
c2 sin2 ␤2
− 3共␣2 + ␤2兲 − 2 cot ␣2 − 2 cot ␤2 册
classic Prandtl-type mechanism, characteristic of plane strain which may be minimized in terms of the defining angles 共␣1, ␤1,
conditions, prevailed approximately 1.5B from the ends of ␣2, and ␤2兲 and the dimension c. For a square foundation, the
the footings. 共2兲 The zone of influence of the failure mecha-
nism beyond the end of the footing was independent of the
footing aspect ratio and equal to approximately 0.6B. 共3兲 The
zone of influence of the failure mechanism to either side of
and beneath the footing varied linearly with footing aspect
ratio. 共4兲 Beyond the region where the plane strain mecha-
nism prevailed the lateral extent of the soil affected by fail-
ure reduced, i.e., tending towards the mechanism observed at
the cross section of the square footing.
9. A comparative analysis of a square footing modeling a
smooth interface showed a distinct Hill-type mechanism, as
opposed to the Prandtl-type observed for the footing with the
rough interface.
10. The mechanisms associated with failure of the rough and
smooth square footings and the rough rectangular footings,
of all aspect ratios, suggested diagonal symmetry.
Extension of this work to address the effect of varying interface
roughness as a function of footing aspect ratio, as well as the
effect of other variables such as nonuniform soil strength profile,
footing embedment 共to complement Salgado et al. 2004兲 and con-
sideration of footing response to more complex loading condi-
tions would be useful. Some of these issues are currently under
investigation at the Centre for Offshore Foundations Systems.

Acknowledgments

The work described here forms part of the activities of the Special
Research Centre for Offshore Foundation Systems, established
and supported under the Australian Research Council’s Research
Centres Program. This support is gratefully acknowledged.

Appendix. Upper Bound Solution for Rectangular


Foundations

An upper bound solution for smooth-based rectangular founda-


tions was described by Shield and Drucker 共1953兲. This Appendix
documents that solution and also a corresponding one for rough-
based foundations.
The mechanism for smooth-based foundations is shown in Fig. Fig. 13. Mechanism for smooth-based rectangular foundation 共after
13, following the notation of Shield and Drucker. The rectangular Shield and Drucker 1953兲: 共a兲 overall schematic; 共b兲 plan and
foundation has width 2b, and length 2a 共with b 艋 a兲. The mecha- elevation views

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006 / 155


Table 5. Optimum Upper Bound Solution for Smooth-Based Square
Foundation
Smooth-based Rough-based

Parameter Shield and Drucker 共1953兲 Optimum Optimum


c/a 0.707 0.686 0.585
␣1 共°兲 46.3 46.2 48.3
␤1 共°兲 39 38.0 49.7
␣2 共°兲 47.1 47.0 50.6
␤2 共°兲 34 34.0 63.6
Nc 5.711 5.710 6.411

Fig. 15. Upper bound bearing capacity factors for smooth- and
rough-based rectangular foundations

minimum Nc is 5.71, which is obtained with the parameters tabu-


lated in Table 5. Note that the optimum solution is marginally
lower than for the parameters suggested by Shield and Drucker
共1953兲—before the modern computer age! The optimum upper
bound for a square foundation based on this mechanism has c / a
less than unity, hence giving a mechanism that does not have
diagonal symmetry, as might have been assumed intuitively.
For the rough-based foundation, the mechanism switches to a
Prandtl-style, with a central wedge of soil moving down with the
foundation, as indicated in Fig. 14. In other respects, the mecha-
nism is similar to that for the smooth-based foundation, except
that there is a constraint on one of the angles, since the depth of
point T below the foundation is equal to c cot ␤1 and also
b cot ␤2.
It may be shown that the upper bound bearing factor is given
by

Nc = 2共␣2 + ␤2兲 + cot ␣2 + cot ␤2 +


c
2a

␣1 + ␤1 + 冋冉 冊
c
b
共␣1 + ␤1

+ cot ␣1兲 + ␣2 + ␤2 + cot ␣2 册冑 1+


b2
c sin2 ␤2
2 + cot ␤1

− 3共␣2 + ␤2兲 − 2 cot ␣2 − cot ␤2 册


The optimal parameters for a square foundation are given in Table
5, resulting in a bearing factor of Nc = 6.41. Upper bound values
of Nc for rectangular foundations of aspect ratio, b / a, are shown
in Fig. 15 for smooth- and rough-based conditions.

References

Cox, A. D., Eason, G., and Hopkins, H. G. 共1961兲. “Axially symmetric


plastic deformation in soils.” Proc. R. Soc. London, Ser. A, 254, 1–45.
De Beer, E. E. 共1970兲. “Experimental determination of the shape factor
and the bearing capacity factors for sand.” Geotechnique, 20共4兲,
387–411.
Gourvenec, S. M., and Randolph, M. R. 共2003兲. “Effect of strength non-
homogeneity on the shape and failure envelopes for combined loading
of strip and circular foundations on clay.” Geotechnique, 53共6兲, 575–
Fig. 14. Corresponding mechanism for rough-based rectangular
586.
foundation: 共a兲 overall schematic; 共b兲 plan and elevation views Hansen, B. J. 共1970兲. “A revised and extended formula for bearing ca-

156 / INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006


pacity.” Bulletin 28, Danish Geotechnical Institute Copenhagen, Angew. Math. U. Mech, 1:15–20.
Denmark. Randolph, M. F. 共2000兲. “Effect of strength anisotropy on capacity of
Hibbit, Karlsson and Sorensen, Inc. 共HKS兲. 共2002兲. ABAQUS users’ foundations.” Proc., Booker Memorial Symp., Sydney, Australia,
manual, version 6.1. 313–328.
Itasca Consulting Group 共1997兲. FLAC3D users’ manual, Minneapolis. Salgado, R., Lyamin, A. V., Sloan, S. W., and Yu, H. S. 共2004兲. “Two-
Kusakabe, O., Susuki, H., and Nakase, A. 共1986兲. “An upper bound cal- and three-dimensional bearing capacity of foundations in clay.” Geo-
culation on bearing capacity of a circular footing on a nonhomoge- technique, 54共5兲, 297–306.
neous clay.” Soils Found., 26共3兲, 143–148. Shield, R. T., and Drucker, D. C. 共1953兲. “The application of limit analy-
Levin, E. 共1955兲. “Indentation pressure of a smooth circular punch.” Q. sis to punch indentation problems.” J. Appl. Mech., 20, 453–460.
Appl. Math., 13共2兲, 133–137. Skempton, A. W. 共1951兲. “The bearing capacity of clays.” Proc., Building
Michalowski, R. L. 共2001兲. “Upper bound load estimates on square and and Research Congress, Vol. 1, London, 180–189.
rectangular footings.” Geotechnique, 51共9兲, 787–798. Taiebat, H. A. 共2005兲. “Effects of flow rules on the prediction of Tresca
Micahlowski, R. L., and Dawson, E. M. 共2002兲. “Three-dimensional model.” submitted to Geotechnique.
analysis of limit loads on Mohr–Coulomb soil.” Foundations of civil Taiebat, H. A., and Carter, J. P. 共2000兲. “Numerical studies of the bearing
and environmental engineering, Vol. 1, Poznan University of Technol- capacity of shallow foundations on cohesive soil subjected to com-
ogy Press, Poland, 37–147. bined loading.” Geotechnique, 50共4兲, 409–418.
Prandtl, L. 共1921兲. “Eindringungsfestigkeit und festigkeit von schneiden.” Terzaghi, K. 共1943兲. Theoretical soil mechanics, Wiley, New York.

INTERNATIONAL JOURNAL OF GEOMECHANICS © ASCE / MAY/JUNE 2006 / 157

You might also like