You are on page 1of 9

Computers and Geotechnics 36 (2009) 871–879

Contents lists available at ScienceDirect

Computers and Geotechnics


journal homepage: www.elsevier.com/locate/compgeo

Bearing capacity of strip and circular footings in sand using finite elements
D. Loukidis *, R. Salgado 1
School of Civil Engineering, Purdue University, 550 Stadium Mall Drive, Civil Engineering Building, West, Lafayette, IN 47907-1284, USA

a r t i c l e i n f o a b s t r a c t

Article history: Design of shallow foundations relies on bearing capacity values calculated using procedures that are
Received 1 July 2007 based in part on solutions obtained using the method of characteristics, which assumes a soil following
Received in revised form 3 December 2008 an associated flow rule. In this paper, we use the finite element method to determine the vertical bearing
Accepted 26 January 2009
capacity of strip and circular footings resting on a sand layer. Analyses were performed using an elastic–
Available online 28 March 2009
perfectly plastic Mohr–Coulomb constitutive model. To investigate the effect of dilatancy angle on the
footing bearing capacity, two series of analyses were performed, one using an associated flow rule and
Keywords:
one using a non-associated flow rule. The study focuses on the values of the bearing capacity factors
Bearing capacity
Finite elements
Nq and Nc and of the shape factors sq and sc for circular footings. Relationships for these factors that
Footings are valid for realistic pairs of friction angle and dilatancy angle values are also proposed.
Plasticity Ó 2009 Elsevier Ltd. All rights reserved.
Sand

1. Introduction ships for Nc have been established by fitting to the values obtained
numerically from the method of characteristics for footings with no
The bearing capacity qbL of footings is traditionally calculated surcharge. The most popular expressions for Nc are those by Meyer-
using the bearing capacity equation, which takes into account hof [22], Brinch Hansen [5] and Vesić [34]. Among the better known
the effects of soil cohesion, surcharge and soil unit weight in an equations for Nc, the Brinch Hansen [5] expression
uncoupled manner by simply superposing the three terms that
Nc ¼ 1:5ðN q  1Þ tan / ð3Þ
are related to each of these factors. In the case of a footing in an
uncemented sand deposit (in which case the soil cohesion c is is fairly accurate for / < 40° [21,13,17]. Regarding the shape factors
zero), the bearing capacity equation reduces to sq and sc, the most popular formulas are those by Meyerhof [22],
Brinch Hansen [5] and Vesić [34]. Meyerhof [22], based partly on
1
qbL ¼ q0 Nq sq dq þ cBNc sc dc ð1Þ theory (MOC) and partly on experimental observations, gives the
2
same formula for both shape factors:
where c is the soil unit weight, B is the footing width, Nq and Nc are
bearing capacity factors, sq and sc are the shape factors that intro-
1 þ sin / B
sq ¼ sc ¼ 1 þ 0:1 ð4Þ
duce the effect of footing geometry in the case of a footing other 1  sin / L
than strip footing, and dq and dc are depth factors. The subscripts where L is the longest dimension (length) of the footing. For sc, both
‘q’ and ‘c’ indicate terms associated with the contribution of sur- Brinch Hansen [5] and Vesić [34] propose the following equation:
charge and soil unit weight, respectively. The values for the bearing
B
capacity factors Nq and Nc can be derived based on the method of sc ¼ 1  0:4 ð5Þ
characteristics (MOC). For Nq, there exists an exact solution in L
closed form [25]: For sq, Brinch Hansen [5] proposed
1 þ sin / p tan / B
Nq ¼ e ð2Þ sq ¼ 1 þ sin / ð6Þ
1  sin / L
where / is the soil friction angle. The factor Nc cannot be obtained while Vesić [34] proposed
using MOC as a closed-form solution. Therefore, several relation- B
sq ¼ 1 þ tan / ð7Þ
L
* Corresponding author. Tel.: +357 22892192; fax: +357 22892295.
E-mail addresses: loukidisdim@alumni.purdue.edu (D. Loukidis), rodrigo@ecn.
Eqs. (5)–(7) are based on experimental work by De Beer [6].
purdue.edu (R. Salgado). The method of characteristics assumes that the soil follows an
1
Tel.: +1 765 494 5030; fax: +1 765 494 0395. associated flow rule. Rigorous limit analysis (LA) makes the same

0266-352X/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2009.01.012
872 D. Loukidis, R. Salgado / Computers and Geotechnics 36 (2009) 871–879

assumption. This means that both methods apply to soils with a horizontal and vertical degrees of freedom are fixed. At the lateral
dilatancy angle w equal to the friction angle /, although it is well boundaries, only the horizontal degree of freedom is fixed (Fig. 1).
known that w is significantly lower than / for frictional soils in Preliminary analyses showed that fixing both degrees of freedom
reality. There are a number of other important aspects of the actual or just the horizontal one at the lateral boundary nodes does not
bearing capacity problem that these traditional approaches do not affect the resulting collapse load as long as the lateral boundaries
account for. These stem mainly from the fact that / depends on a are placed far enough from the developing collapse mechanism.
number of variables that evolve continuously during the footing The distances of the bottom and lateral boundaries from the foot-
loading process: the relative density DR, the level of mean effective ing varied from analysis to analysis in order to ensure that the
stress p0 , the relative magnitude of the intermediate principal boundaries did not interfere with the development of the collapse
effective stress (stress-induced anisotropy), and the direction of mechanism while maintaining adequate mesh density close to the
the principal effective stresses relative to the axis of sand deposi- footing.
tion (fabric-induced anisotropy). There are no interface elements placed between the soil and the
Many researchers have investigated numerically the vertical footing, so any slippage between footing and soil occurs within
bearing capacity of shallow foundations on cohesionless soil that the soil. This is realistic because concrete footings poured against
follows the Mohr–Coulomb failure criterion using finite elements the ground form a very rough interface. There are strong indica-
(FE) or finite differences (FD) [11,7,18,10,24,35,9,37]. In this paper, tions from the application of MOC that the exact value of the
we will use finite element analysis to determine values of the bear- soil-footing interface friction angle d has minimal effect on the col-
ing capacity factors Nq, Nc from strip footing analyses and of the lapse load as long as / P 30° and d// P 0.8 [12,15]. Although no
shape factors sq, sc for circular footings by combining results from slippage is allowed between soil and footing right at the footing
analyses of circular and strip footings. A series of analyses are per- base in the FE analyses, this condition does not correspond to the
formed for associated flow rule w = /) and non-associated flow so-called ‘‘narrow rough” interface condition, in which case the
rule, including realistic pairs of values for w and / applicable to width of the trapped rigid (or ‘‘active”) wedge that forms under-
sands. The constitutive model used for the soil is an elastic–per- neath the footing is equal to the footing width B, and trapped
fectly plastic model following a Mohr–Coulomb (M–C) failure cri- wedge and footing move as a single body. This is because, in the
terion. The effect of soil lying above footing base level is FE analyses, plastic yielding can occur inside the soil elements that
represented only as a surcharge load. Any contribution from the are in contact with the footing base. To capture the intense concen-
portion of the slip surfaces extending above the footing base level tration of plastic shear strains in those elements due to potential
(which is traditionally taken into account through the depth fac- relative horizontal movement between footing and soil (especially,
tors dq and dc) is not considered. Results are compared with predic- close to the footing edges), the element size was limited to 0.035–
tions from the formulas used widely in current practice, as well as 0.06B close to the footing edge and 0.07–0.11B at the vicinity of the
with results from recent studies employing accurate MOC, rigorous footing center. The number of soil elements with an edge at the
LA, and finite element (or finite difference) analysis. The paper footing base across the footing half-width was 6–10 (25–41 nodes
demonstrates how accurate numerical algorithms and fine mesh- along the footing half-width).
ing enable good predictions of footing bearing capacity that are
in close agreement with the rigorous results from MOC and LA 2.2. Material model
and helps in assessing the effect of non-associativity on the col-
lapse load. Modified/improved expressions for the factors Nq, Nc, The constitutive model used is an elastic–perfectly plastic mod-
sq, and sc are also proposed. el following the Mohr–Coulomb failure criterion. The corners that
the M–C failure criterion exhibits in the deviatoric plane were
rounded according to Sloan & Booker [29] using formulations al-
2. Finite element analyses ready available in SNAC [1]. The rounding takes place in a window
of hT = ±1°, in Lode’s angle h terms (Fig. 2a), around the corners of
2.1. Mesh discretization and boundary conditions the M–C yield surface in the deviatoric plane. In addition to the
rounding of the corners in the deviatoric plane, the apex of the
The finite element analyses were performed using the code M–C pyramid on the hydrostatic axis was also rounded using a
SNAC [2]. The analyses used unstructured meshes consisting of hyperbolic approximation (a feature also already available in
15-noded triangular elements. Unstructured meshing allows effi- SNAC) following Zienkiewicz & Pande [38], in order to the elimi-
cient element arrangement and refinement of the elements in nate the gradient singularity there. The values assigned to the
the vicinity of the corners of the footing, which is crucial for the meridional eccentricity em (Fig. 2b) were very small
accurate prediction of the collapse load [23]. The 15-noded ele- (0.004–0.021 kPa) while a small cohesion in the range of 0.002–
ments converge more rapidly and perform better numerically than 0.021 kPa was also assigned to the soil. The combinations of merid-
6-noded elements when a strongly non-associated flow rule is ional eccentricity values and cohesion values in these ranges imply
used, allowing use of a smaller number of nodes to achieve the that the tip of the yield surface passes through the origin of the
same level of accuracy as a mesh consisting of 6-noded elements. stress space (Fig. 2b). The rounding was deemed necessary to avoid
The plane-strain elements used in the strip footing simulations ill-conditioning due to non-uniqueness of the stress gradients and,
and the axisymmetric elements used in the circular footing simu- consequently, divergence of the stress–stain integration algorithm.
lations possessed 12 and 16 Gauss-quadrature points, respectively, The Young’s modulus and Poisson’s ratio were set to 80 MPa
following Sloan and Randolph [28]. and 0.35, respectively. As in previous studies [24,16], it was ob-
Loading of the footing is accomplished through prescription of served that the values of Young’s modulus, the Poisson’s ratio
uniform incremental vertical displacements at the nodes located and the earth pressure coefficient at rest K0 affect the evolution
at the base of the footing (displacement control), while the hori- of the footing settlement but have no effect in the value of the col-
zontal degrees of freedom are fixed. This way, the footing is consid- lapse load. The analyses with a non-associated flow rule were done
ered to be perfectly rigid and rough. The footing load is obtained as considering three sets of pairs of values for w and /, which are gi-
the sum of the reactions at all the nodes where the footing nodal ven in Table 1. Set 2 and Set 3 can be considered as representative
displacements are prescribed (i.e., across the entire footing width of sands with low and high critical-state friction angle values,
B). At the bottom boundary of the finite element mesh, both the respectively, and are consistent with prediction using the Bolton
D. Loukidis, R. Salgado / Computers and Geotechnics 36 (2009) 871–879 873

B/2 prescribed nodal


displacements

fixed in horizontal direction

fixed in horizontal
direction
fixed in horizontal
direction
CL

fixed in both directions

Fig. 1. Typical mesh and boundary conditions for footing simulations.

a θT
σ1 b
θT 2J 2

θT
θT

p'
em=c cotφ

σ3 σ2

π -plane cross-section θ = 0 o cross-section

Fig. 2. Elimination of the gradient singularities of the Mohr–Coulomb surface through rounding: (a) in the deviatoric plane [29] and (b) at the hydrostatic axis [38].

Table 1 one order of magnitude compared to the rest of the analyses). This
Pairs of friction angle and dilatancy angle used in the analyses with a non-associated is because the explicit modified Euler stress-point algorithm is
flow rule.
forced to divide the applied strain increments in fine sub-incre-
Friction angle / (o) Dilatancy angle w (°) ments in order to satisfy the strict stress-error tolerance
Set 1 Set 2 Set 3 (0.000001), especially in the case of Gauss-quadrature points that
lie close to the ground surface and close the footing corners, where
30 10 2 –
35 15 6 2 the strain rate is high and the mean effective stress is very small.
40 20 12 6 Therefore, the choice of setting w = 2° when / = 30° was made in
45 25 18 12 order to limit the time required to complete an analysis.

2.3. Solution scheme


[4] equation and data from Tatsuoka [31]. Results from all three
sets will allow us later to establish equations for the bearing capac- In the analyses with an associated flow rule, the original New-
ity factors in terms of both the friction angle and dilatancy angle. ton–Raphson scheme was used for the solution of the global
Since we assume that the soil is perfectly plastic, the values re- load–displacement equations. The modified Newton–Raphson
ported in Table 1 represent peak friction angle and peak dilatancy scheme was used in the case of non-associated flow, with the glo-
angle values. bal stiffness matrix being always the elastic stiffness matrix. This is
Most of the sands are expected to exhibit w = 0 when / = 30o. because the global elasto-plastic stiffness matrix is subject to se-
However, setting w = 0 results in very long runtimes (more than vere ill-conditioning when a flow rule with a high degree of non-
874 D. Loukidis, R. Salgado / Computers and Geotechnics 36 (2009) 871–879

associativity is used, thus preventing the analysis to proceed and tion iterations. Using this configuration, the ratio of the maximum
reach collapse load conditions. The integration of the stress–strain unbalanced force to the maximum external nodal force (a ratio
equations of the constitutive model was done using the modified usually referred to as the unbalanced force norm) did not exceed
Euler (explicit) sub-incrementation scheme [27] with stress rela- 0.001.
tive error tolerance equal to 0.0001%.
We observed through a number of trial runs that, by increasing 3. Results and discussion
mesh refinement, decreasing relative error tolerances in the solu-
tion algorithms, and decreasing load increment size, the resulting All analyses were performed for a footing with B = 1 m. The
limit load decreases. This means that, as the accuracy of the simu- analyses for the determination of Nc and sc were done with
lation increases, the exact limit load value is approached from c = 20 kN/m3 and no surcharge. In analyses for the determination
above (i.e., relatively coarse limit load estimates would tend to of Nq and sq, q0 was set equal to 15 kPa and the soil was assumed
be too high and thus unconservative). All analyses started from to be weightless. Although the collapse load QbL depends on the
an initial stage in which the geostatic stress field is established values considered for B, c or q0, the values of the resulting bearing
in the finite element mesh (which corresponds to application of capacity equation factors are independent of these parameters
gravity to the model). Each subsequent footing loading stage con- when the soil friction angle and dilatancy angle are always con-
sisted of a large number (of the order of 100,000) of constant-size stant (as in this study). Figs. 3 and 4 show examples of the evolu-
increments and a maximum number of 10–20 equilibrium-correc- tion of normalized footing load with footing settlement w,

a 40
b 250
φ=35 , ψ=35
o o
strip footing φ=45o , ψ=45o
strip footing

200
normalized footing load 2Q/(γB 2) = N γ

normalized footing load 2Q/(γB 2) = Nγ

30

φ=35o , ψ=6o φ=45o , ψ=18o


150

20

φ=30o , ψ=30o 100


φ=40o , ψ=40o
φ=30 , ψ=2
o o

10 φ=40o , ψ=12o
50

0 0
0 0.005 0.01 0.015 0 0.04 0.08 0.12
normalized settlement w/B normalized settlement w/B

Fig. 3. Load–settlement response from the analyses of strip footings for the determination of Nc.

a 600 b 600

circular footing
φ=45o , ψ=45o
500
normalized footing load Q/(q0πB2/4) = Nqsq

500
normalized footing load Q/(q0πB2/4) = Nqsq

φ=45o , ψ=18o
400 400

circular footing

300 300

200 φ=40o , ψ=40o 200


φ=40o , ψ=12o

100 100
φ=35o , ψ=35o φ=35o , ψ=6o
φ=30o , ψ=30o φ=30 , ψ=2
o o

0 0
0 0.1 0.2 0.3 0 0.04 0.08 0.12 0.16 0.2
normalized settlement w/B normalized settlement w/B

Fig. 4. Load–settlement response from the analyses of circular footings for the determination of Nqsq.
D. Loukidis, R. Salgado / Computers and Geotechnics 36 (2009) 871–879 875

illustrating that the analyses with a non-associated flow rule pro-


CL
duce load–displacement curves that oscillate as the collapse load
(m)
(peak load) is approached and after its attainment. Such oscilla- 0.0
tions have been observed in other studies involving an M–C consti-
φ=40o
tutive model with w < / [7,18,35,9]. The intensity of the
ψ =40o
oscillations increases with increasing / and increasing mesh
1.0 γ=0
refinement. These oscillations are due to the apparent softening circular
exhibited in shear bands by materials following a M–C model with q0>0
a non-associated flow rule, even with the strength parameters
remaining always constant. As explained in [33,8], the apparent (m)
strain softening is related to the rotation of the principal stress 0.0
axes as strains are intensely localized inside the shear bands. In
0 φ=40o
the case of a non-associated flow rule, the material ceases to obey
ψ =1
=12o
Drucker’s postulate, and energy is released under certain condi- 1.00
γ=0
tions and stress paths. In essence, the observed oscillations are a circular
direct consequence of the energy release during bifurcation (in- q0>0
tense shear banding) and the energy transfer to regions where
localization has not propagated yet. It should be pointed out that 0..0 1.0 2.0 3.0 4.0 5.0 6.0 (m)
the oscillations present in the load–settlement response do not
undermine the validity of the present FE simulations. The only
problem posed by the numerical oscillations is the choice of the Fig. 6. Collapse mechanisms as depicted by contours of the plastic maximum shear
strain increment compared against the mechanism yielded by Martin’s ABC
collapse load value. In this study, the collapse load is taken as
program (shown with dashed lines) for circular footings on weightless soil with
the maximum footing load value observed during the analysis. surcharge.
Fig. 5 shows examples of the collapse mechanisms for footings
on ponderable soil (soil with non-zero unit weight) and no sur-
charge (analyses for Nc and Ncsc). Figs. 6 and 7 show examples of
mechanisms on weightless soil with a surcharge (analyses for Nq C
L
and Nqsq). The mechanisms are depicted by contours of incremen- (m)
0.0
tal plastic maximum shear strain Dcpl max . The mechanisms devel-
oped in the FE simulations with w = / agree well with the
mechanism predicted by MOC (using the ABC computer program 1.0
[19]) both in shape and in size. As observed in previous studies φ=40o
[7,18,24,35,9], the collapse mechanism for the w = / case is larger 2.0 ψ =4
=40o
than the one for w < /. Besides the mechanism size, another impor- strip γ=0
tant difference between the simulations for associated and non- q0>0
associated flow rules is that the deformation in the non-associated
flow cases is highly localized in thin shear bands, while, in the (m)
0.0
associated flow cases, the plastic strains appear to be more diffused
inside the mechanism. The intense shear banding when w < / is a
φ=40o
direct consequence of the apparent softening and energy release 1.0
exhibited by a material with a non-associated flow rule. ψ =12o
strip γ=0
2.0
q0>0

CL
0.0 1.0 2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 (m)
(m)
0.0
Fig. 7. Collapse mechanisms as depicted by contours of the plastic maximum shear
φ=40o strain increment compared against the mechanism yielded by Martin’s ABC
ψ=40o program (shown with dashed lines) for strip footings on weightless soil with
1.0 surcharge.
γ>0
strip
q0=0
3.1. Bearing capacity factors Nc and Nq
(m)
0.0 Table 2 and Fig. 8a summarize the Nc (=2QbL/(cB2)) values
resulting from various methods of analysis, including the present
φ=40o
FE analyses, for soil friction angles from 30° to 45°. The main rea-
ψ=12o
1.0 son for the various expressions for Nc found in the literature is the
γ>0 different assumptions in MOC regarding the geometry of the part
strip
q0=0 of the collapse mechanism located immediately below the footing
base (referred to as the ‘‘active” or ‘‘rigid” wedge) [36,14]. Accurate
bearing capacity solutions based on MOC are obtained here using
0.0 1.0 2.0 3.0 4.0 5.0 6.0
(m) the program ABC [19]. Nc values from ABC are also cited in
[20,21]. Recently, Martin [21] and Smith [30] proved that these
Fig. 5. Collapse mechanisms as depicted by contours of the plastic maximum shear
Nc values, which are in close agreement with those by Hansen
strain increment compared against the mechanism yielded by Martin’s ABC
program (shown with dashed lines) for strip footings on ponderable soil and no and Christensen [12], Booker [3] and Kumar [14], are exact by
surcharge. extending the stress field to the entire half–space. Rigorous lower
876 D. Loukidis, R. Salgado / Computers and Geotechnics 36 (2009) 871–879

Table 2
Bearing capacity factor Nc values resulting from FE and FD analyses, limit analysis, and the method of characteristics.

/ Lower Upper This This paper This paper This paper MOC Manoharan and Manoharan and Frydman and Frydman and
(°) bound – bound – paper FEM (w < / FEM (w < / FEM (w < / (Martin’s Dasgupta [18] Dasgupta [18] Burd [10] Burd [10]
Hjiaj et al. Hjiaj et al. FEM , Set 1) , Set 2) , Set 3) ABC) (FEM, w = /) (FEM, w = 0) (FLAC, w = /) (FLAC, w = 0)
[13] [13] (w = /)
30 14.6 15.2 15.2 14.1 12.9 – 14.8 23 18 21.7 16.7
35 34.0 35.7 35.5 33.6 29.0 25.0 34.5 53 36 54.2 35.6
40 83.3 88.4 87.7 76.1 67.9 57.6 85.6 – – 147.0 73.0
45 225.0 240.9 239.9 211.7 169.5 139.1 234.2 – – 422.0 144.0

a 250 a 1.1
Lower bound - Hjiaj et al. [13]
Upper bound - Hjiaj et al. [13]
MOC (Martin's ABC)
200 Present study (ψ=φ) 1.0
Present study (ψ<φ, Set 1)
Present study (ψ<φ, Set 2)
Present study (ψ<φ, Set 3)
0.9
150

Nγ / Nγ(MOC)

0.8
100

0.7
Lower bound - Hjiaj et al. [13]
50 Upper bound - Hjiaj et al. [13]
MOC (Martin's ABC)
0.6 Present study (ψ=φ)
Present study (ψ<φ, Set 1)
Present study (ψ<φ, Set 2)
0 Present study (ψ<φ, Set 3)
30 35 40 45 0.5
φ( ) o 30 35 40 45
φ (o)
b 160
MOC exact - Prandtl [25] b 1.1
Present study (ψ=φ)
Present study (ψ<φ, Set 1)
Present study (ψ<φ, Set 2)
Present study (ψ<φ, Set 2) 1.0
120

0.9
Nq/Nq(MOC)
Nq

80
0.8

40 0.7

MOC exact - Prandtl [25]


Present study (ψ=φ)
0.6
Present study (ψ<φ, Set 1)
Present study (ψ<φ, Set 2)
0
Present study (ψ<φ, Set 3)
30 35 40 45
φ ( o) 0.5
30 35 40 45
Fig. 8. Comparison of results from various methods of analysis: (a) bearing capacity φ (o)
factor Nc and (b) bearing capacity factor Nq.
Fig. 9. Bearing capacity factor values normalized with respect to the corresponding
value from MOC: (a) for Nc and (b) for Nq.
and upper bounds using limit analysis (LA) by Ukritchon et al. [32],
Hjiaj et al. [13] and Lyamin et al. [17] provide extra corroboration
on the validity of the MOC solutions. as closely as desired. Previous efforts based on the finite element
The present Nc values for w = / lie inside the LA bounds by Hjiaj method (FEM) or finite difference method (FDM) yielded results
et al. [13] (Fig. 9a). This fact establishes the validity of the present (Table 2) that lie further away from the exact solution mainly be-
FE simulations. The FE results for an associated flow rule are cause of the coarseness of the meshes and the order of the ele-
roughly 2.5% higher than the ABC results. With further mesh ments used. The present study agrees with previous ones on the
refinement, the FE results for w = / can approach the MOC ones fact that the values of the bearing capacity factors Nq and Nc for
D. Loukidis, R. Salgado / Computers and Geotechnics 36 (2009) 871–879 877

a soil with realistic w values are clearly smaller than the Nq and Nc hand. Despite this fundamental difference between the two analy-
values for a material with an associated flow rule [7,18,10,35]. ses, FE results for w = / from this study and results from MOC differ
According to the present FE results, the Nc for non-associated flow at most by 2% (Fig. 10a). The results for realistic w values are 11–
rule with realistic w values (Sets 2 and 3) is 15–42% smaller than 35% smaller than the Ncsc values for w = /.
the Nc for w = /, with the differences increasing with increasing Results for Nqsq for circular footings are compared in Table 5 and
/. Fig. 10b. In this case, the MOC suffers from an additional drawback:
Results for the bearing capacity factor Nq (=QbL/(q0B)) are shown the solution contains overlapping stress characteristics. This may
in Table 3 and Fig. 8b. The exact solution for Nq is well known [25]. be the reason why, in the case of Nqsq, the agreement between
Our results for w = / lie only 1.0–1.6% above the exact solution our FE results and the MOC results is not as strong as in the previ-
(Fig. 9b). The Nq values for realistic w values are 12–42% smaller ous cases, with differences ranging from 1% for / = 30° to 9% /
than the exact Nq values for w = /. The Nq values in Table 3 from = 45°. The results for realistic w values are 10–41% smaller than
our FE analyses can be fitted by the following equation: the Nqsq values for w = /.
The results given in Tables 4 and 5 can be used to establish val-
1 þ sin / Fð/;wÞp tan /
Nq ¼ e ð8Þ ues for the shape factors scirc:
c and scirc:
q . In foundation engineering, it
1  sin /
has been traditionally assumed that the solution for a square foun-
where dation (or square loaded area) can be approximated by the solution
for a circular one, which is much easier to obtain because of the ax-
Fð/; wÞ ¼ 1  tan /½tanð0:8ð/  wÞÞ2:5 ð9Þ ial symmetry of the problem. The resulting shape factor values are
It can be seen that Eq. (8) is based on the formula for the exact Nq for presented in Figs. 11 and 12. In Fig. 11, the shape factor values
w = /, with the only difference that a function F (Eq. (9)) of the dif- based on limit analysis [17] are derived from the average between
ference (/  w) has been introduced in the exponential term. In the corresponding lower and upper bounds. All methods produce
turn, the Nc values from the present FE analyses can be approxi- higher shape factor values with increasing /. Moreover, the rate
mated by of increase of scirc:
c or scirc:
q with / is nearly the same for FE and
MOC. From the formulas used for sc in foundation design practice,
Nc ¼ ðN q  1Þ tanð1:34/Þ the only one that predicts an increase in sc with / is that by Mey-
 
1 þ sin / Fð/;wÞp tan / erhof [22] Eq. (4). However, the rate of increase of sc with / pre-
¼ e  1 tanð1:34/Þ ð10Þ
1  sin / dicted by Eq. (4) is small compared with those obtained from
MOC, LA or FE analysis. Following the form of Meyerhof’s equation,
which resembles the form of the Nc equation by Meyerhof [22].
we propose the following equation for the shape factor scirc: c for the
case of an associated flow rule:
3.2. Shape factors sc and sq
 
1 þ sin /
The products Ncsc and Nqsq in Eq. (1) can be considered as equiv- scirc:
c ¼ 1 þ 0:26  0:73 ð11Þ
1  sin /
alent bearing capacity factors for footings other than strip footings.
A comparison between Ncsc for circular footings resulting from Eq. (5), endorsed by Brinch Hansen [5] and Vesić [34] and originally
limit analysis [17], the method of characteristics (Martin’s ABC), proposed by De Beer [6] based on footing model tests, suggests that
and FE simulations is given in Table 4 and Fig. 10a. As in the case sc is independent of the friction angle. Moreover, it predicts shape
of strip footings, the results from MOC are inside the rigorous LA factor values that are always below unity, with sc = 0.6 for square
bounds, despite the fact that the MOC solution is no longer rigor- or circular footings. It is true that, in reality, the bearing capacity
ous for circular footings since it relies on an arbitrary assumption of square and circular footings with zero embedment is smaller
regarding the magnitude of the intermediate principal effective than the bearing capacity of strip footings, although theoretical
stress. In FE analyses, the intermediate principal stress is a product methods assuming a simple M–C failure criterion suggest the oppo-
of the analysis itself through the satisfaction of the three-dimen- site. As pointed out by Meyerhof [22] and discussed in Lyamin et al.
sional M–C failure criterion and the kinematics of the problem at [17] and Salgado [26], this is because of the fact that the peak fric-

Table 3
Bearing capacity factor Nq values resulting from FE analyses and the method of characteristics.

/ Exact – This paper FEM This paper FEM This paper FEM This paper FEM Manoharan and Dasgupta Manoharan and Dasgupta
(°) Prandtl [25] (w = /) (w < /, Set 1) (w < /, Set 2) (w < /, Set 3) [18] (FEM, w = /) [18] (FEM, w = 0)
30 18.4 18.6 17.7 16.3 – 19.5 17
35 33.3 33.6 30.5 26.5 24.6 35.0 29
40 64.2 65.0 57.6 48.2 42.0 – –
45 134.9 137.0 118 95.8 79.3 – –

Table 4
Equivalent bearing capacity factor Ncsc values (circular footings) resulting from FE and FD analyses, the method of characteristics, and limit analysis.

/ Lower Upper This This paper This paper This paper MOC Manoharan and Manoharan and Erickson and Erickson and
(o) bound – bound – paper FEM FEM FEM (Martin’s Dasgupta [18] Dasgupta [18] Drescher [9] Drescher [9]
Lyamin Lyamin FEM (w < /, Set (w < /, Set (w < /, Set ABC) (FEM, w = /) (FEM, w = 0) (FLAC, w = /) (FLAC, w = 0)
et al. [17] et al. [17] (w = /) 1) 2) 3)
30 14.1 19.8 15.8 14.9 13.7 – 15.5 28 22.5 – –
35 37.2 52.5 42.0 39.3 37.5 34.9 41.9 76 52.0 45 33
40 106.6 157.2 122.2 113.1 105.9 84.6 123.7 – – 130 73
45 338.0 539.2 408.5 378.0 320.9 266.7 417.7 – – 456 198
878 D. Loukidis, R. Salgado / Computers and Geotechnics 36 (2009) 871–879

a 1.4 2.2
Present study (ψ=φ)
Present study (ψ<φ)
MOC (Martin's ABC)
2.0
LA bound average for ciruclar - Lyamin et al. [17]
1.2 LA bound average for square - Lyamin et al. [17]
Eq. (4) - Meyerhof [22]
Eq. (11)
1.8
Nγsγ/Nγsγ(MOC)

1.0

sγcirc.
1.6

0.8
1.4

Lower bound - Lyamin et al. [17]


Upper bound - Lyamin et al. [17]
0.6 MOC (Martin's ABC) 1.2
Present study (ψ=φ)
Present study (ψ<φ, Set 1)
Present study (ψ<φ, Set 2)
Present study (ψ<φ, Set 3) 1.0
0.4 30 35 40 45
30 35 40 45
φ (o )
φ (o )
Fig. 11. Comparison of shape factor sc values for circular footings obtained from

b FEM, from MOC, and from rigorous LA.

1.0
5.0
Present study (ψ=φ)
Present study (ψ<φ)
MOC (Martin's ABC)
Eq. (4) - Meyerhof [22]
Nqsq/Nqsq(MOC)

Eq. (6) - Brinch Hansen [5]


0.8 4.0 Eq. (7) - Vesic [34]
Eq. (12)
sqcirc.

3.0
0.6
MOC (Martin's ABC)
Present study (ψ=φ)
Present study (ψ<φ, Set 1)
Present study (ψ<φ, Set 2)
Present study (ψ<φ, Set 3) 2.0

0.4
30 35 40 45
φ (o )
1.0
Fig. 10. Equivalent bearing capacity factor values for circular footing normalized 30 35 40 45
with respect to the corresponding value from MOC: (a) for Ncsc and (b) for Nqsq.
φ (o)
Fig. 12. Comparison of shape factor sq values for circular footings obtained from
FEM, from MOC, and from equations used in standard practice.
tion angle of the sand is lower in nearly triaxial (TX) conditions
(square/circular footings) than in plane strain (PS) conditions (strip
footings). Meyerhof [22] suggested that, for the determination of
the bearing capacity of strip footings, the peak friction angle value operative / values for strip footings are distinctively greater than
used in the calculations should be 10% higher than the peak friction for circular or square footings.
angle for TX conditions. Current knowledge allows more accurate In the case of sq, Eqs. (4), (6), and (7) predict an increase of the
computation of appropriate values of /, but it remains true that shape factor with /. However, both the shape factor values and

Table 5
Equivalent bearing capacity factor Nqsq values (circular footings) resulting from FE analyses and the method of characteristics.

/ This paper FEM This paper FEM This paper FEM This paper FEM MOC Manoharan and Dasgupta Manoharan and Dasgupta
(°) (w = /) (w < /, Set 1) (w < /, Set 2) (w < /, Set 3) (Martin’s [18] (FEM, w = /) [18] (FEM, w = 0)
ABC)
30 36.9 35.0 33.5 – 37.2 42 33
35 79.4 74.6 69.8 63.6 80.8 90 62
40 188.8 175.4 160.8 142.3 192.7 – –
45 503.1 454.1 396.5 296.2 550.6 – –
D. Loukidis, R. Salgado / Computers and Geotechnics 36 (2009) 871–879 879

their rate of increase with / are significantly smaller than those [3] Booker JR. Applications of theories of plasticity to cohesive frictional soils. PhD
Thesis, University of Sydney, Australia; 1969.
resulting from the present FE analyses and from MOC calculations
[4] Bolton MD. Strength and dilatancy of sands. Géotechnique 1986;36(1):65–78.
using ABC. Following the form of Zhu and Michalowski [37], we [5] Brinch Hansen J. A revised and extended formula for bearing capacity. Danish
may propose the following equation for the shape factor scirc:
q :
Geotechn Inst Bull 1970;28:5–11.
[6] De Beer E. Bearing capacity and settlement of shallow foundations on sand. In:
scirc:
q ¼ 1 þ 2:9 tan2 / ð12Þ Proceedings of the symposium on bearing capacity and settlement of
foundations, Duke University, Durham, NC; 1965. p. 15–33.
[7] de Borst R, Vermeer PA. Possibilities and limitations of finite elements for limit
analysis. Géotechnique 1984;34(2):199–210.
4. Conclusions [8] Drescher A, Detournay E. Limit load in translational failure mechanisms for
associative and non-associative materials. Géotechnique 1993;43(3):443–56.
[9] Erickson HL, Drescher A. Bearing capacity of circular footings. J Geotech
We performed finite element simulations of strip and circular
Geoenviron Eng ASCE 2002;128(1):38–43.
footings on sand. The sand layer was assumed to be uniform, and [10] Frydman S, Burd HJ. Numerical studies of bearing-capacity factor Nc. J
there was zero footing embedment. The analyses employed an Geotechn Geoenviron Eng ASCE 1997;123(1):20–9.
elastic–perfectly plastic constitutive model following the Mohr– [11] Griffiths DV. Computation of bearing capacity factors using finite elements.
Géotechnique 1982;32(3):195–202.
Coulomb failure criterion. Two series of runs were performed: [12] Hansen B, Christensen NH. Discussion of ‘‘Theoretical bearing capacity of very
one with an associated flow rule and one with realistic pairs of fric- shallow footings” by Larkins AL. J Soil Mech Found Div ASCE
tion angle and dilatancy angle values. Based on the results of this 1969;95(6):1568–72.
[13] Hjiaj M, Lyamin AV, Sloan SW. Numerical limit analysis solutions for the
numerical study, we can draw the following conclusions: bearing capacity factor Nc. Int J Solids Struct 2005;42(5):1681–704.
[14] Kumar J. Nc for rough strip footing using the method of characteristics. Can
(1) Finite element simulation can be in very close agreement Geotech J 2003;40(3):669–74.
[15] Kumar J, Kouzer KM. Effect of footing roughness on bearing capacity factor Nc. J
with bearing capacity results from rigorous analytical or Geotech Geoenviron Eng ASCE 2007;133(5):502–11.
semi-analytical methods such as the method of characteris- [16] Lee J, Salgado R. Estimation of bearing capacity of circular footings on sands
tics and limit analysis. For this, we must construct ade- based on cone penetration test. J Geotech Geoenviron Eng ASCE
2005;131(4):442–52.
quately fine meshes and employ accurate and efficient [17] Lyamin A, Salgado R, Sloan SW, Prezzi M. Two- and three-dimensional bearing
algorithms. The present FE results for the case of an associ- capacity of footings in sand. Géotechnique 2007;57(8):647–62.
ated flow rule were always inside rigorous bounds from LA [18] Manoharan N, Dasgupta SP. Bearing capacity of surface footings by finite
elements. Comput Struct 1995;54(4):563–86.
and did not differ by more than 2.5% from the values
[19] Martin CM. User guide for ABC – Analysis of Bearing Capacity, Version 1.0.
obtained from rigorous MOC solutions. OUEL Report No. 2261/03. Department of Engineering Science, University of
(2) The effect of flow rule non-associativity on bearing capacity Oxford; 2003.
is not negligible when considering realistic pairs of / and w [20] Martin CM. Discussion of ‘‘Calculations of bearing capacity factor Nc using
numerical limit analyses” by Ukritchon B, Whittle AJ, Klangvijit C. J Geotech
values exhibited by sands. The differences between the non- Geoenviron Eng ASCE 2004;130(10):1106–7.
associated and associated flow rule cases were in the 5–45% [21] Martin CM. Exact bearing capacity calculations using the method of
range with the larger values corresponding to the case of characteristics. In: Barla G, Barla M, editors. Proceedings of the 11th
IACMAG, vol. 4, Turin; 2005. p. 441–50.
strip footings and high friction angle values. Therefore, the [22] Meyerhof GG. Some recent research on the bearing capacity of foundations.
current bearing capacity solutions for sands based on meth- Can Geotech J 1963;1(1):16–26.
ods that assume that w = / produce unconservative results. [23] Potts DM, Zdravkovic L. Finite element analysis in geotechnical engineering:
theory. London: Thomas Telford Ltd; 1999.
This is true for both strip and circular footings. [24] Potts DM, Zdravkovic L. Finite element analysis in geotechnical engineering:
(3) Although MOC solutions for circular footings contain applications. London: Thomas Telford Ltd; 2001.
assumptions regarding the intermediate principal effective [25] Prandtl L. Über die eindringungsfestigkeit (härte) plastischer baustoffe und die
festigkeit von schneiden. Zeitschrift für Angewandte Mathematik und
stress at failure and, occasionally, overlapping stress charac- Mechanik 1921;1:15–20.
teristic lines, they are still in close agreement with the FE [26] Salgado R. The engineering of foundations. New York: McGraw-Hill Inc.; 2008.
analysis with an associated flow rule. [27] Sloan SW. Substepping schemes for the numerical integration of elastoplastic
stress–strain relations. Int J Numer Meth Eng 1987;24(5):893–911.
(4) There is agreement between the FE results, rigorous LA
[28] Sloan SW, Randolph MF. Numerical prediction of collapse loads using finite
bounds and MOC solutions regarding the shape factors for element methods. Int J Numer Anal Meth Geomech 1982;6(1):47–76.
circular footings. The factors sq and sc always increase with [29] Sloan SW, Booker JR. Removal of singularities in Tresca and Mohr–Coulomb
increasing friction angle. However, equations for sq and sc yield functions. Commun Appl Numer Meth 1986;2(2):173–9.
[30] Smith CC. Complete limiting stress solutions for the bearing capacity of strip
used in current practice seem to underestimate the rate of footings on a Mohr–Coulomb soil. Géotechnique 2005;55(8):607–12.
increase of these factors with /. [31] Tatsuoka F. Discussion of ‘The strength and dilatancy of sands’ by M.D. Bolton.
Géotechnique 1987;37(2):219–26.
[32] Ukritchon B, Whittle AJ, Klangvijit C. Calculations of bearing capacity factor Nc
Finally, modified/improved equations are proposed for the using numerical limit analyses. J Geotech Geoenviron Eng
bearing capacity factors Nq and Nc as functions of both / and w, 2003;129(5):468–74.
considering realistic pairs of w and u values suitable for sands. [33] Vermeer PA. The orientation of shear bands in biaxial tests. Géotechnique
1990;40(2):223–36.
New equations are also proposed for the shape factors sq and sc [34] Vesić AS. Analysis of ultimate loads of shallow foundations. J Soil Mech Found
for circular footings based on the FE analysis results. Div ASCE 1973;99(1):45–73.
[35] Yin J-H, Wang Y-J, Selvadurai APS. Influence of nonassociativity on the bearing
capacity of a strip footing. J Geotech Geoenviron Eng 2001;127(11):985–9.
References [36] Zhu DY, Lee CF, Jiang HD. A numerical study of the bearing capacity factor Nc.
Can Geotech J 2001;38(5):1090–6.
[1] Abbo AJ. Finite element algorithms for elastoplasticity and consolidation. PhD [37] Zhu M, Michalowski RL. Shape factors for limit loads on square and rectangular
Thesis, University of Newcastle, Australia; 1997. footings. J Geotech Geoenviron Eng 2005;131(2):223–31.
[2] Abbo AJ, Sloan SW. SNAC, User manual, Version 2.0. Department of Civil, [38] Zienkiewicz OC, Pande GN. Some useful forms of isotropic yield surfaces for
Surveying and Environmental Engineering, University of Newcastle, Callaghan, soil and rock mechanics. In: Gudehus G, editor. Finite elements in
Australia; 2000. geomechanics. Wiley; 1977. p. 179–90.

You might also like