Undrained Bearing Capacity of Strip Footings on Slopes

K. Georgiadis1

Abstract: The undrained bearing capacity of foundations on or near slopes is commonly calculated using empirical equations or from
design charts which have been produced based on limit equilibrium or upper bound plasticity calculations. Many of the available methods
do not take account of important parameters that affect the undrained bearing capacity factor, such as the distance of the footing from the
slope, the slope height, or the soil properties. This paper presents finite element analyses of strip footings on or near undrained soil slopes
Downloaded from ascelibrary.org by University of Brighton on 10/18/13. Copyright ASCE. For personal use only; all rights reserved.

performed in order to investigate the influence of the various parameters that affect undrained bearing capacity. The results of the analyses
are compared to available methods. It is found that while some of these methods compare well with the finite element results for certain
combinations of geometrical parameters and soil properties, they cannot produce sufficiently accurate results as they either do not take
account of all of the affecting parameters or are generally not conservative. Based on the finite element results, design charts, equations,
and a design procedure are proposed for the calculation of the undrained bearing capacity factor Nc as a function of the undrained shear
strength and the bulk unit weight of the soil, the footing width, the distance of the footing from the slope, the slope angle and the slope
height.
DOI: 10.1061/共ASCE兲GT.1943-5606.0000269
CE Database subject headings: Load bearing capacity; Slope stability; Finite element method; Footings.
Author keywords: Bearing capacity; Slope stability; Finite-element analysis.

Introduction

The undrained bearing capacity qu of surface foundations is con-

冉
Nc = 5.14 1 −
2␤
␲+2
冊
−
␥B
cu
sin ␤ 共3兲

ventionally calculated from the bearing capacity equation pre- Hansen’s expression is in fact the exact solution 共lower and
sented by Meyerhof 共1963兲. The solution for vertical only loading upper bound solution兲 for weightless soil 共␥ = 0兲 and is identical
is identical to Prandtl’s solution and is theoretically exact. For a to the stress field solution proposed by Kumar and Mohan Rao
strip surface footing, the undrained bearing capacity is calculated 共2003兲. Since the assumption of ␥ = 0 is not conservative, Eq. 共2兲
as follows 关Georgiadis 共2009兲兴: cannot be considered a lower bound solution. The applicability of
the above Eqs. 共2兲 and 共3兲 is limited to foundations in a slope or
Vu at the crest of a slope. Moreover, neither of them considers the
qu = = c uN c 共1兲 influence of the slope height and Hansen’s solution 关Eq. 共2兲兴 does
B
not take account of the effect of the soil properties 共cu and ␥兲. An
where Vu = ultimate vertical load; B = footing width; cu empirical solution has also been presented by Bowles 共1996兲.
= undrained shear strength; and Nc = ␲ + 2共⬇5.14兲 = undrained This solution takes account of the distance of the footing from the
bearing capacity factor. slope but not the slope height nor the soil properties.
In contrast to the horizontal ground surface case, no exact Meyerhof 共1957兲 proposed design charts which are currently
solution exists for the bearing capacity of foundations on slopes. adopted by many design manuals such as Unified Facilities Cri-
The bearing capacity equations of Hansen 共1961兲 and Vesic teria 共UFC兲 and WSDOT 共2005兲 Geotechnical Design Manuals.
共1975兲 include empirical factors for sloping ground. For a strip An upper bound solution to the same problem of vertical loading
surface footing under central vertical loading the following equa- of footings on slopes was presented by Kusakabe et al. 共1981兲
tions are obtained for the undrained bearing capacity factor: who also provided design charts based on their analyses. Upper
• Hansen 共1961兲: bound bearing capacity solutions for locally loaded slopes have
been proposed by Michalowski 共1989兲 and de Buhan and Garnier
冉
Nc = 5.14 1 −
2␤
␲+2
冊 共2兲
共1998兲. The limit equilibrium method has also been applied by
several investigators to provide solutions and design charts for
strip and square footings. Among them are Azzouz and Baligh
• Vesic 共1975兲:
共1983兲 who performed circular arc limit equilibrium calculations
1
to provide charts for both strip and square footings on clay slopes,
Lecturer, Dept. of Civil Engineering, Aristotle Univ. of Thessaloniki, Narita and Yamaguchi 共1990兲 who presented a log-spiral solution,
541 24 Thessaloniki, Greece. E-mail: [email protected] and recently, Castelli and Motta 共2008兲 who used the method of
Note. This manuscript was submitted on August 4, 2008; approved on
slices to propose design charts for drained and undrained slopes.
October 13, 2009; published online on October 15, 2009. Discussion
period open until October 1, 2010; separate discussions must be submit- Georgiadis et al. 共2008兲 used the finite element 共FE兲 method to
ted for individual papers. This paper is part of the Journal of Geotech- investigate the undrained bearing capacity of strip footings at the
nical and Geoenvironmental Engineering, Vol. 136, No. 5, May 1, 2010. crest of slopes under inclined and eccentric loading.
©ASCE, ISSN 1090-0241/2010/5-677–685/$25.00. This paper presents FE analyses for the calculation of the bear-

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 677

J. Geotech. Geoenviron. Eng. 2010.136:677-685.

 V
10B B 10B

λB B

H
Undrained clay 7.5B
cu, Eu, γ
β

Fig. 1. Problem definition

ing capacity of surface strip foundations on undrained clay slopes. Fig. 3. FE mesh for strip footing on horizontal ground surface
The solutions are first compared to those proposed by other writ-
ers. The influence of slope height, footing distance, and soil prop-
共Brinkgreve and Broere 2006兲. Fifteen-noded triangular elements
Downloaded from ascelibrary.org by University of Brighton on 10/18/13. Copyright ASCE. For personal use only; all rights reserved.

erties on the undrained bearing capacity factor is then

were used to model the soil while the footing was modeled with
investigated. Based on the analyses design charts and an equation
beam elements. Interface elements were placed between the foot-
useful for design purposes are proposed.
ing and the soil. Fig. 3 shows the FE mesh used in the analyses of
a strip footing on horizontal ground. The mesh consists of 2,073
elements and has dimensions 42 m horizontal by 15 m vertical
Details of Analyses 共for footing width B = 2 m兲. The calculated vertical undrained
bearing capacity factor for horizontal ground surface was Nc
The geometry of the problem analyzed is shown in Fig. 1. Three
= 5.16, only 0.4% higher than the exact value of 5.14. Use of a
footing widths were considered B = 1, 2, and 4 m and three slope
coarser mesh leading to a slightly higher value of Nc would have
angles ␤ = 15°, 30°, and 45°. Several slope heights H and normal-
been acceptable but was not considered necessary as the compu-
ized footing distances ␭ 共=footing distance/footing width兲 were
tational time for each analysis was relatively small. The dimen-
used in order to investigate the influence of these two parameters.
sions of the mesh are larger than those required for the calculation
The undrained Young’s modulus and the bulk unit weight of the
of failure loads on a strip footing on horizontal ground, but were
soil were kept constant 共Eu = 30 MPa and ␥ = 20 kN/ m3兲 while
chosen to be consistent with the analyses of a strip footing at the
the undrained shear strength was varied. The FE analyses per-
crest of a slope. Fig. 4共a兲 shows a typical such FE mesh for the
formed to investigate the problem are presented in the following
case of a 7.5-m-high 45° soil slope at a distance of 2 m from the
paragraphs.
footing. The mesh consists of 2,670 elements and has the same
Depending on the combination of the above parameters three
dimensions as the mesh discussed above for horizontal ground
distinct failure modes are observed, as discussed in more detail in
surface. Zero horizontal displacements were prescribed to the ver-
the presentation of the results of the analyses. The first two failure
tical boundaries and full 共vertical and horizontal兲 fixities to the
modes shown in Figs. 2共a and b兲 are similar to Prandtl’s failure
base of the mesh. The slope was constructed by excavation of the
mechanism for the case of horizontal ground surface and will be
appropriate soil layers for each analysis.
referred to as “bearing capacity” failure modes, while the second
Various distances of the mesh boundaries from the footing
failure mode 关Fig. 2共c兲兴 involves the whole of the slope and will
were examined but only proved to have an influence on the cal-
be referred to as “overall slope failure.”
culated ultimate load for combinations of geometrical and mate-

FE Analyses 12B B 8B

All FE analyses presented in this study are plane strain analyses

and were performed with the program Plaxis Version 8.6

7.5B
λB B λB B
V
V
(a)

(a)
B B 1.25B
β (b)

B
V B

β
(b)
(c)

Fig. 4. FE mesh for strip footing at a distance of 2 m from the crest

Fig. 2. Failure modes: 关共a兲 and 共b兲兴 bearing capacity failure and 共c兲 of 7.5-m-high 45° angle slope: 共a兲 whole mesh with boundary condi-
overall slope failure tions; 共b兲 detail of mesh under footing

678 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010

J. Geotech. Geoenviron. Eng. 2010.136:677-685.

 5.5 fore modeled as a Tresca material with no restriction in the de-
velopment of tensile stresses.
5 The properties of the interface elements in Plaxis V8.6 are
related to the properties of the adjacent soil elements. The footing
4.5 was modeled as rough with the same shear strength and shear
modulus for the interface and adjacent soil elements. In order to
4 B allow an effective gap to form between part of the footing and the
Nc

Hansen (H)
Bowles (B) H
soil in the case of footing rotation, a very thin 2.5-cm zone of zero
Vesic cu/γΒ = 1 (V1) FE5/K5 tensile strength soil elements was modeled beneath the footing.
3.5 V5
Vesic cu/γΒ = 5 (V5)
Kusakabe cu/γΒ = 1 (K1) K1 As a consequence, the interface elements were also not allowed to
FE1 sustain tension. This however, did not prove to be critical for any
3 Kusakabe cu/γB = 5 (K5)
FE cu/γΒ = 1 (FE1) V1 of the analyses performed because footing rotation was never
FE cu/γΒ = 5 (FE5)
2.5
followed by the development of tensile stresses in the zone im-
mediately beneath the footing. The footing was assumed linear
Downloaded from ascelibrary.org by University of Brighton on 10/18/13. Copyright ASCE. For personal use only; all rights reserved.

0 10 20 o 30 40 50
β( ) elastic with 1-m thickness and Young’s modulus for concrete
Ec = 2.9⫻ 107 kPa giving a bending stiffness of EcI = 2.4
Fig. 5. Variation of Nc with slope angle ␤ for footing at crest of slope
⫻ 106 kNm2 / m and axial stiffness of EcA = 2.9⫻ 107 kN/ m.
共␭ = 0兲

rial parameters resulting in overall slope failure 关Fig. 2共c兲兴. The Comparison of Various Methods
mesh dimensions shown in Fig. 5共a兲 were found to be adequate
for the majority of the cases analyzed. However, it was necessary The various available solutions for the bearing capacity of a foot-
to increase both the vertical and horizontal dimension for greater ing on a slope can be divided into two categories. The first cat-
slope heights. In these cases, the boundary position was selected egory includes the methods by Hansen 共1961兲 and Vesic 共1975兲
to be as close as possible to the slope without affecting the cal- which only consider footings at the crest of a slope 共␭ = 0兲. The
culated ultimate load. For this purpose, different mesh sizes were second category includes methods that also consider the influence
examined for representative geometrical and material properties of the distance of the footing from the slope 共␭ ⱖ 0兲. These are the
combinations. Finally, the horizontal dimension was also in- methods proposed by Meyerhof 共1957兲, Kusakabe et al. 共1981兲,
creased for all analyses involving the smaller slope angle of 15° Azzouz and Baligh 共1983兲, Narita and Yamaguchi 共1990兲, and
without however changing the overall mesh density. Bowles 共1996兲. The comparison of the FE results with these
The mesh density in the area beneath the footing is of most methods is presented separately, first for ␭ = 0 共footing at the crest
importance for the cases involving bearing capacity failure. Fig. of a slope兲 and then for ␭ ⱖ 0. It is noted that most of the above
4共b兲 shows the detail of the mesh refinement in this area for the methods consider only bearing capacity failure 关Figs. 2共a and b兲兴
mesh illustrated in Fig. 4共a兲. Mesh refinement exercises were per- and are therefore compared to the FE results corresponding to this
formed for the two limiting slope angles considered in this study type of failure. Overall slope stability failure 关Fig. 2共c兲兴 is only
of 0° 共as discussed above兲 and 45° in which both the overall and considered in this section in the comparison of the FE results to
local 共beneath the footing兲 element densities were gradually in- the solutions by Meyerhof and Azzouz and Baligh.
creased. For the latter case a footing distance of 2 m from the
slope, slope height of 7.5 m 共Fig. 4兲, and cu / 共B␥兲 ratio of 2.5
Footing at the Crest of a Slope
were considered. The number of elements was varied from 526 to
3,052. For the coarse mesh consisting of 526 elements, the calcu- The results presented in this study are first compared to the solu-
lated undrained bearing capacity factor was 4.48. This value de- tions by Hansen, Vesic, Bowles, and Kusakabe et al. for the case
creased only slightly with the increase of the number of elements of a footing at the crest of a slope. Fig. 5 shows the variation of
to 4.45 for 2,093 elements and remained practically constant for the undrained bearing capacity factor with the slope angle. FE
further increase of mesh density. The adopted mesh 共2,670 ele- analyses were performed for two ratios cu / 共B␥兲 = 1 and 5. As
ments兲 is clearly denser than necessary but was preferred because seen, all solutions give a linear decrease of the undrained bearing
it was used as the basis on which all other meshes were con- capacity factor with the increase of the slope angle. The FE re-
structed while still requiring reasonable computational time. The sults and the upper bound solution by Kusakabe et al. 共K兲 are in
number of elements slightly varied for the other slope angles excellent agreement and show a decrease of the value of Nc with
considered. As noted above in the case of large slope heights the the decrease of the ratio cu / 共B␥兲. Vesic’s equation also gives a
dimensions of the mesh were larger and consequently the number decrease of Nc with the decrease of cu / 共B␥兲, is in good agreement
of elements also increased, however, it was not possible to with the 共FE兲 results for cu / 共B␥兲 = 5 but gives up to 10% lower
achieve the same mesh density in the area beneath the footing. undrained bearing capacity factor for cu / 共B␥兲 = 1. Bowles’ solu-
This did not significantly affect the accuracy of the results be- tion overestimates the value of Nc, while Hansen’s equation is in
cause in this case overall slope failure is the dominant failure better agreement with the FE results and Kusakabe’s upper bound
mode. solution, but does not take account of the effect of cu / 共B␥兲 and
The Mohr-Coulomb elastic-perfectly plastic constitutive overestimates Nc by up to 12.5% for the low cu / 共B␥兲 = 1 case.
model was used for the soil with total stress undrained param- For the specific case of a footing at the crest of a slope 共␭
eters: shear strength cu, angle of shearing resistance ␾u = 0, = 0兲 a further comparison of the methods considered above is
Young’s modulus Eu = 30 MPa, and Poisson’s ratio ␮ = 0.49. made in Fig. 6 which also includes results by Narita and Yamagu-
Seven different values of undrained shear strength were consid- chi 共1990兲. All results shown in this figure are for ␤ = 30°. For the
ered cu = 20, 40, 100, 150, 200, 300, and 1,000 kPa. A bulk unit low cu / 共B␥兲 value of 0.5 it was not possible to obtain a bearing
weight of 20 kN/ m3 was selected for the soil. The soil was there- capacity type of failure 共Failure Mode 2兲 with the FE analyses for

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 679

J. Geotech. Geoenviron. Eng. 2010.136:677-685.

 4.6 5.5
Meyerhof (λ=0)
4.4 B
Meyerhof (λ=1)
NY 5
4.2 FE (λ=0)
H
FE FE (λ=1)
4
4.5
3.8 K
Nc

Nc
Hansen (H)
3.6 Bowles (B) 4
V Vesic (V)
3.4 Kusakabe (K)
Narita & Yamaguchi (NY)
3.5
3.2
FE analyses (FE)
3 3
Downloaded from ascelibrary.org by University of Brighton on 10/18/13. Copyright ASCE. For personal use only; all rights reserved.

0 5 10 15 20 25 30 0 1 2 3 4 5 6
cu/(Bγ) H/B

Fig. 6. Variation of Nc with cu / 共B␥兲 for footing at crest of slope Fig. 7. Variation of Nc with normalized slope height for ␤ = 30° and
共␭ = 0 , ␤ = 30°兲 cu / 共B␥兲 = 1—comparison with Meyerhof’s solution

any slope height, since overall slope failure 共Failure Mode 3兲 was cases of ␤ = 30° and 45° and cu / 共B␥兲 = 2.5. For low values of ␭,
always observed. As seen, the FE results and the Kusakabe et al. the two solutions are in excellent agreement. As ␭ increases, how-
upper bound solution 共K兲 are in excellent agreement, Vesic’s 共V兲 ever, the curves start to diverge with the FE analysis always giv-
solution gives lower undrained bearing capacity factor, especially ing lower values of Nc. The reason for this divergence 共maximum
at low cu / 共B␥兲 ratios, while the values proposed by Narita and overestimation of the upper bound solution compared to the FE
Yamaguchi 共NY兲 generally overestimate Nc by approximately results of 4.4% for ␤ = 30° and 10% for ␤ = 45°兲 can be attributed
10%, compared to the FE results and Kusakabe’s 共K兲 solution. to the fact that the assumed upper bound failure mechanism be-
For the same problem the values obtained from Hansen 共H兲 and comes less accurate for ␭ ⫽ 0.
Bowles 共B兲 are 4.09 and 4.38, respectively. In order to investigate this point the improved kinematic
mechanism shown in Fig. 9 which consists of two rigid blocks 共1兲
and 共3兲 and a slip fan 共2兲 is considered 共Georgiadis 2009兲. This
Footing at a Distance from the Slope mechanism is similar to the one assumed by Kusakabe et al.
共1981兲. However, their solution, although more general as it ap-
Comparison to Meyerhof’s Method plies to both drained and undrained conditions, assumed a value
Meyerhof 共1957兲 presented a chart for the calculation of the un-
drained bearing capacity factor which is adopted by design manu-
als such as the Unified Facilities Criteria 共UFC兲 共2005兲 and 5.5
WSDOT 共2005兲. For slope heights H smaller than the foundation
width B, overall slope failure is assumed and Nc is calculated as a
5
function of the slope angle ␤, the stability number Ns = ␥H / cu and β=30ο

the ratio ␭B / H. For H ⱖ B, bearing capacity failure is assumed

and Nc is calculated from the slope angle ␤ and the normalized 4.5
footing distance ␭. In this way, an unrealistic jump of the un-
Nc

drained bearing capacity factor is predicted at H / B = 1, as shown 4

in Fig. 7, for the case of a B = 2-m-wide footing at distances of 0 FEA
β=45ο
and 2 m from the crest of a ␤ = 30° slope with ␥ = 20 kN/ m3 and Kusakabe
3.5 Meyerhof
cu = 40 kPa. The overall variation of Nc with H / B is unrealistic
UB
and does not compare well with the FE results also shown in the
same figure. The FE results give the expected initial reduction of 3
Nc starting from the horizontal ground value, which is followed 0 0.5 1 1.5 2 2.5
λ
by a horizontal section of constant Nc corresponding to bearing
capacity failure 共Failure Mode 2兲 and a final drop corresponding Fig. 8. Variation of Nc with ␭ for ␤ = 30° and 45° and
to overall slope failure 共Failure Mode 3兲 as the slope height ap- cu / 共B␥兲 = 2.5—comparison with the results of Meyerhof and Kusak-
proaches an ultimate value Ho. Meyerhof’s method for 共H / B abe et al.
⬎ 1兲 and the FE results are also compared in Fig. 8, which shows
the variation of Nc with ␭ for the case of a ␤ = 30° slope and λB B
cu / 共B␥兲 = 2.5. As seen Meyerhof’s solution overestimates the V δ3
value of Nc by up to 6.5% compared to the FE results. α
κ 1 κ
3
2
Comparison to Upper Bound Solutions δ1
β κ-α
As discussed above, the FE results are in excellent agreement α δv
with the upper bound solution proposed by Kusakabe et al. 共1981兲 δh
for the case of a footing at the crest of a slope 共␭ = 0兲. The two
methods are compared for varying values of ␭ in Fig. 8 for the Fig. 9. Kinematic mechanism and associated hodograph

680 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010

J. Geotech. Geoenviron. Eng. 2010.136:677-685.

 6

(a) (b) 5

4
∆(γΗ/cu) = 3

Nc
∆(γΗ/cu) = 1

(c) (d) Azzouz & Baligh

1 FE

0
Downloaded from ascelibrary.org by University of Brighton on 10/18/13. Copyright ASCE. For personal use only; all rights reserved.

0 1 2 3 4
λ
Fig. 10. Comparison of FE analysis with optimum upper bound ki-
nematic mechanism 共solid gray lines兲 for ␥ = 0: 关共a兲 and 共b兲兴 incre- Fig. 11. Variation of Nc with ␭ for ␤ = 45°—comparison with Azzouz
mental displacements and principal stress directions for ␭ = 0 and 关共c兲 and Baligh results
and 共d兲兴 incremental displacements and principal stress directions for
␭=1

equilibrium and compatibility requirements within numerical ap-

of 45° for the angle ␣ of the kinematic mechanism. As seen in proximation but obviously cannot be considered as theoretically
Fig. 8 the improved kinematic mechanism yields lower values of exact due to this approximation. However, in the light of the
Nc which, however, are still greater than those obtained from the above discussion it is reasonable to assume that they are more
FE results by up to 3% for ␤ = 30° and 7% for ␤ = 45°. Part of this accurate than the upper bound solutions for ␭ ⬎ 0.
difference can be attributed to the effect of self weight, which is
included in the calculation of the work done by external forces Comparison to Azzouz and Baligh (1983) Method
but not on the shape of the kinematic mechanism. The overesti- A final comparison of the results of this study is made with the
mation of Nc by the upper bound solution for the ␤ = 45° and ␭ circular arc analysis results presented by Azzouz and Baligh
= 1 case, reduces from 3.8 to 2.7% when self weight is taken as 0 共1983兲. They defined a margin of safety of the slope with respect
共in both the upper bound and FE analysis兲. In order to explain the to gravity alone ⌬共␥H / cu兲 = Ns − ␥H / cu where Ns is the stability
remainder of the difference, the velocity and stress fields need to number of the slope. Their results were presented for two values
be examined. of the margin of safety ⌬共␥H / cu兲 = 1 and 3, ratios of B / H = 0.25,
The case of ␭ = 0 共footing at the crest of the slope兲 is first 0.5, and 1, and slope angles ␤ = 15°, 45°, and 90°. In order to
examined for which the FE and upper bound results are in excel- compare these results with the FE results, the case of B / H
lent agreement. Figs. 10共a and b兲 show the incremental displace- = 0.25 and ␤ = 45° was selected, with B = 2 m and H = 8 m. For a
ments and principal stress directions, respectively, at failure 45° slope the stability number is 5.86 共Michalowski 1989; Chen
obtained from the FE analysis for 0 self weight and ␭ = 0 and the 1975兲 which corresponds to cu values of 32.9 and 55.9 kPa for
optimum upper bound kinematic mechanism for the same case. margins of safety ⌬共␥H / cu兲 = 1 and 3, respectively. The specific
As seen, the failure mechanism resulting from the FE analysis low B / H ratio of 0.25 was selected for this comparison so that
matches the optimum mechanism of the upper bound analysis. both bearing capacity and overall slope stability failure mecha-
Moreover, as seen in Fig. 10共b兲 the two rigid blocks 共1 and 3兲 of nisms resulting from the FE results could be compared with the
the upper bound mechanism correspond to homogeneous stress general limit equilibrium solution of Azzouz and Baligh. For the
regions 共constant principal stress directions兲 with the plastic low margin of safety ⌬共␥H / cu兲 = 1, the slope is already close to
zones separating these blocks from the rigid soil mass intersecting failure prior to the application of the footing load. In this case
the principal stress directions at angles of 45°. This indicates that overall slope failure takes place at relatively low foundation
the assumed kinematic mechanism also satisfies equilibrium so loads. For ⌬共␥H / cu兲 = 3 the type of failure depends on the nor-
that the solution is theoretically exact. In fact it is relatively malized distance of the footing from the slope. At low values of ␭
straightforward to prove this by performing a Mohr circle calcu- the FE analysis predicts bearing capacity failure while for higher
lation to obtain the identical lower bound solution given by Eq. values overall slope failure is predicted. The results are compared
共2兲 共Hansen’s solution兲. in Fig. 11 which shows the variation of Nc with ␭. The circular
In contrast, in the case of a footing at a normalized distance arc calculations by Azzouz and Baligh generally overestimate the
␭ = 1 from the slope, the FE and upper bound failure mechanisms undrained bearing capacity factor by up to 30% compared to the
shown in Fig. 10共c兲 are different resulting in different calculated FE analysis.
Nc values. In this case the change of stress boundary conditions at
the crest of the slope results in varying principal stress directions
within Rigid Block 3 of the upper bound mechanism. As seen in Numerical Results
Fig. 10共d兲 the principal stress directions also vary along the plas-
tic zone which separates this block from the rigid soil mass. As a The upper bound methods discussed above and also those by
consequence, the mechanism no longer satisfies equilibrium and Vesic 共1975兲 and Narita and Yamaguchi 共1990兲 relate the un-
yields a greater value of Nc than the theoretically exact value. The drained bearing capacity factor Nc to the slope angle ␤, the nor-
FE results, which are generally more conservative than the results malized distance ␭ of the footing from the slope, and the ratio
of both upper bound solutions 共as seen in Fig. 8兲, satisfy both cu / 共B␥兲. FE results for various values of ␤, ␭, cu, ␥, and B con-

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 681

J. Geotech. Geoenviron. Eng. 2010.136:677-685.

 6 共second failure mode兲. A typical mechanism of this kind is shown
Failure mode 1 in Fig. 13共b兲. The third failure mode involves overall slope failure
5 as illustrated in Fig. 13共c兲 and takes place when the slope height
Failure mode 2 approaches its ultimate value for gravitational only loading. For
4
the low cu / 共B · ␥兲 = 0.5 case Failure Mode 2 is never observed.
Although an intermediate second mode can also be identified in
Nc

3
Fig. 12 for this case, the failure mechanism always extends to the
Failure mode 3
toe of the slope. As expected, increase of the ratio cu / 共B␥兲 leads
2
to increased Nc values and also increased ultimate slope heights.
0.5 1 1.5 2.5 cu/(γB) = 5 In addition, Failure Mode 2 becomes increasingly dominant with
1
the increase of cu / 共B␥兲.
0
For smaller slope angles of 15° and 30° the calculated Nc
0 5 10 15 20 25 30 values show the same trend as that illustrated in Fig. 12. The
transition from Failure Mode 1 to Failure Mode 2 in these cases
Downloaded from ascelibrary.org by University of Brighton on 10/18/13. Copyright ASCE. For personal use only; all rights reserved.

H/B
takes place at slightly lower H / B values of 0.25 and 0.5, respec-
Fig. 12. Variation of Nc with H/B for ␭ = 0 and ␤ = 45° tively, instead of 0.75 for ␤ = 45°. However, as discussed in more
detail below, the ultimate slope heights are the same. Similar
trends were obtained for ␭ ⬎ 0, although as shown in the follow-
firmed this observation. The same Nc values were calculated from
ing paragraphs overall slope failure 共Failure Mode 3兲 was ob-
analyses with the same ratio cu / 共B␥兲 but different combinations
served for more combinations of the affecting parameters.
of cu, B, and ␥. This is a particularly important finding since the
This limiting slope height is commonly calculated from the
number of parameters necessary to define the problem is reduced
stability number of the slope Ns = ␥H / cu. For slope angles lower
from five to three. The parametric analyses performed in this
than 53°, the stability number depends on the existence of a hard
study considered cu / 共B␥兲 ratios of 0.5, 1, 1.5, 2.5 3.75, 5, and 7.5
stratum at some depth from the initial horizontal ground surface
to investigate the influence of slope height and footing distance
共Taylor 1948兲 and on the slope angle 共Chen 1975; Yu et al. 1998兲.
from the slope on Nc. The influence of slope height is first exam-
In the present study, FE analyses were performed for uniform
ined in this section, while the discussion on the influence of
soil, using large mesh sizes for all three slope angles and the
cu / 共B␥兲 and the normalized footing/slope distance ␭ is presented
cu / 共B␥兲 mentioned above. The size of the mesh for each case was
in the following section with the help of the proposed design
increased gradually in consecutive runs until it did not signifi-
charts.
cantly affect the obtained ultimate slope height. A unique stability
Fig. 12 illustrates the variation of the undrained bearing capac-
number of approximately 5.33 was calculated for all cases con-
ity factor Nc with the normalized slope height for various ratios
sidered, independent of slope angle and cu / 共B␥兲 ratio 共with a
cu / 共B␥兲 for the case of slope angle ␤ = 45° and normalized foot-
small deviation of ⫾ 0.5 due to numerical accuracy兲 which is
ing distance ␭ = 0. With the exception of the low cu / 共B␥兲 = 0.5
consistent with the results by Yu et al. 共1998兲 which show that the
case, three distinct failure modes can be observed. An initial
effect of the slope angle on Ns becomes insignificant for large
mode, in which a steep reduction of the value of Nc from the
depths of the horizontal hard stratum. The obtained stability num-
horizontal ground value is observed, a second mode of constant
ber is in very good agreement with the value 5.52 共4% lower兲
Nc and a final mode where the undrained bearing capacity factor
proposed by Taylor 共1948兲 which is adopted in most soil mechan-
drops rapidly to 0. These failure modes were briefly discussed in
ics text books. It lies also between and very close to the lower and
the introduction of the paper 共schematically represented in Fig. 2兲
upper bound solutions by Yu et al. 共1998兲.
and referred to in the previous paragraphs as “bearing capacity
failure” and “overall slope failure” modes for the first two and the
third failure modes, respectively. In the initial mode the failure
mechanism is restricted by the slope height, as shown in Fig. Design Charts
13共a兲. Beyond a certain slope height ratio H / B, which in this
particular case is approximately 0.75 the failure mechanism ex- Figs. 14–17 show the variation of Nc with ␭ for ratios cu / 共B␥兲
tends to the slope and is not affected by the height of the slope 0.5, 1, 1.5, 2, 2.5, and 5. Each of these charts provides curves for
slope angles 0°, 15°, 30°, and 45° and different H / B ratios.
As expected, the undrained bearing capacity factor increases
with the increase of cu / 共B␥兲. The value of cu / 共B␥兲 also affects
the dominant failure mode. It is noticed that for low values of
cu / 共B␥兲 and high values of ␭ and slope angle ␤, the undrained
bearing capacity factor depends on the ratio H / B, as overall slope
failure 共Failure Mode 3兲 occurs, irrespective of the slope height
and foundation width. For this reason, curves are provided for
different H / B values. For large values of cu / 共B␥兲 共2.5 and 5 in
Fig. 17兲, the undrained bearing capacity factor is independent of
H / B and corresponds to bearing capacity failure 共Failure Mode
2兲.
The undrained bearing capacity factor increases with the in-
crease of the footing distance from the slope. This increase con-
tinues until a critical value of the normalized footing distance ␭o
Fig. 13. Failure mechanisms: 共a兲 Mode 1; 共b兲 Mode 2; and 共c兲 Mode is reached beyond which the undrained bearing capacity factor for
3 horizontal ground surface is obtained. An exception to this gen-

682 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010

J. Geotech. Geoenviron. Eng. 2010.136:677-685.

 5.5 5.5
β = 0ο H/B=2

5 H/B=1.5 β = 0ο
H/B≤1.5 H/B=1
5 H/B≤6
4.5 β = 15ο
H/B=0.5 H/B=8

4
H/B=2 4.5 β = 30ο

3.5
Nc

Nc
3 4 β = 45ο

2.5 c u /(Bγ) = 2
o
β = 15 3.5
o
c u /(Bγ) = 2.5
2 β = 30
o c u /(Bγ) = 5
β = 45
Downloaded from ascelibrary.org by University of Brighton on 10/18/13. Copyright ASCE. For personal use only; all rights reserved.

1.5 3
0 0.5 1 1.5 2 2.5 3 3.5 4 0 0.5 1 1.5 2 2.5 3 3.5 4
λ λ

Fig. 14. Variation of Nc with ␭ for cu / 共B␥兲 = 0.5 Fig. 17. Variation of Nc with ␭ for cu / 共B␥兲 = 2, 2.5, and 5

erally observed behavior occurs for certain combinations of low

due to the change of failure mode with the increase of ␭, from
values of cu / 共B␥兲, large slope angles ␤ and high H / B ratios 共e.g.,
Failure Mode 2 at ␭ = 0 to Failure Mode 3 at higher values of ␭. It
cu / 共B␥兲 = 1, ␤ = 45°, and H / B = 4 in Fig. 15兲, in which the value of
is noted that this reduction is not observed for the very low
Nc reduces initially with the increase of ␭ before increasing to-
cu / 共B␥兲 ratio of 0.5 since overall slope failure 共Mode 3兲 is ob-
ward the value for horizontal ground surface. This reduction is
served even for ␭ = 0, as shown above in Fig. 12.
For low H / B ratios the undrained bearing capacity factor is
greater than that obtained from the design charts 共Figs. 14–17兲
5.5
H/B ≤ 2.5 H/B=3 and approaches the horizontal ground surface value of 5.14 as
β = 0ο H / B tends to 0. Although use of Figs. 14–17 is on the safe side in
5 this case, a linear increase can be assumed for Nc from the value
H/B=4
obtained from these charts at the limiting H / B value 共of 0.25, 0.5,
4.5 and 0.75 for slope angles 15°, 30°, and 45°, respectively兲 to Nc
H/B=1
H/B=2
= 5.14 at H / B = 0.
4
Nc

H/B=3

Design Equations
3.5

H/B=4
o Fig. 18 shows the variation of the calculated values of the un-
β=15 (all H/B)
3 drained bearing capacity factor for a footing at the crest of a slope
β=30 o
β=45
o Nco 共Nc for ␭ = 0兲 with the slope angle ␤ for different cu / 共B␥兲
2.5 ratios. It can be seen that Nco reduces linearly with increasing ␤.
0 0.5 1 1.5 2 2.5 3 3.5 4 It also increases slightly with increasing cu / 共B␥兲. Based on these
λ
results the following equation is proposed for the calculation of
Fig. 15. Variation of Nc with ␭ for cu / 共B␥兲 = 1 Nco:

5.5

β = 0ο 5.5
5 H/B=1 H/B=2
H/B=4 cu /(γΒ) = 1
5 cu /(γΒ) = 1.5
4.5
cu /(γΒ) = 2.5
cu /(γΒ) = 5
H/B=6 4.5
4 cu /(γΒ) = 7.5
Nc

Nco

β = 15o (all H/B) 4

3.5
β = 30 o (all H/B)
β= 45o (H/B=1)
β = 45o (H/B=2) 3.5
3
β = 45 o (H/B=4)
β= 45o (H/B=6)
2.5 3
0 0.5 1 1.5 2 2.5 3 3.5 4 0 10 20 30 40 50
λ β (o)

Fig. 16. Variation of Nc with ␭ for cu / 共B␥兲 = 1.5 Fig. 18. Variation of Nco with ␤

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 683

J. Geotech. Geoenviron. Eng. 2010.136:677-685.

 2.5 • For H / B ratios lower than the limiting value of 0.25, 0.5, or
0.75 for slope angles 15°, 30°, and 45°, respectively, the cal-
2 culated value of Nc may be increased using linear interpolation
between this value for the corresponding limiting H / B ratio
and 5.14 for H / B = 0; and
1.5
• For H / B ratios higher than 0.8Ho / B the calculated value of Nc
λο

should be conservatively reduced by linear interpolation be-

1 c u /(γΒ) = 2 tween this value at H / B = 0.8Ho / B and 0 at H / B = Ho / B. The
c u /(γΒ) = 2.5 slope height Ho corresponds to the slope stability number 5.33
c u /(γΒ) = 3.75 and is equal to Ho = 5.33cu / ␥.
0.5
c u /(γΒ) = 5 Example: Calculate the undrained bearing capacity of two strip
c u /(γΒ) = 7.5 footings, 1.5 and 4 m wide, placed at a distance of 2 and 5 m,
0 respectively, from the crest of a 35° inclination and 14-m-high
15 20 25 30 35 40 45 50
Downloaded from ascelibrary.org by University of Brighton on 10/18/13. Copyright ASCE. For personal use only; all rights reserved.

β (o) slope. The soil has an undrained shear strength of cu = 120 kPa
and a bulk unit weight of 20 kN/ m3.
Fig. 19. Variation of ␭o with ␤ 1. B = 1.5 m: the ratio cu / 共B␥兲 = 4 is calculated. Since cu / 共B␥兲
⬎ 2.5, Eq. 共6兲 can be applied. From Eqs. 共4兲 and 共5兲, Nco
= 3.85 and ␭o = 1.78⬎ ␭ = 2 / 1.5= 1.33. From Eq. 共6兲, Nc
2␤ = 4.9 is calculated 共the same result is obtained from the chart
Nco = 5.14 − 共where ␤ in rads兲 共4兲
␥␤ of Fig. 18兲 and therefore: qu = 4.9⫻ 120= 588 kPa; and
1− 2. B = 4 m: the ratio cu / 共B␥兲 = 1.5 is calculated. Since cu / 共B␥兲
5.14 ⫻ cu
⬍ 2.5 and ␤ ⬎ 30° the design charts have to be used. For
Fig. 19, derived from the results of the FE analyses, shows the cu / 共B␥兲 = 1.5 共Fig. 16兲 the value of Nc = 4.7 is obtained for
variation of the critical value of the normalized distance ␭o 共be- ␭ = 5 / 4 = 1.25 through linear interpolation between the ␤
yond which Nc = 5.14兲 with the slope angle ␤ for ratios cu / 共B␥兲 = 30° 共Nc = 4.9兲 curve and the ␤ = 45° 共Nc = 4.4 for H / B = 3.5兲
= 2, 2.5, 3.75, 5, and 7.5. As seen, the value of ␭o increases curves. For ␤ = 35° the lower limiting value of H / B is 0.58
nonlinearly with the angle ␤ while the effect of cu / 共B␥兲 is only 共interpolation between 0.5 for ␤ = 30° and 0.75 for ␤ = 45°兲.
minor. The following equation is proposed for the calculation of For cu = 120 kPa and ␥ = 20 kN/ m3 Ho = 5.33⫻ 120/ 20
␭␱ = 32 m and therefore, the upper limiting ratio of Ho / B is

␭␱ = 冉 冊
5.14
2
␤
共where ␤ in rads兲 共5兲
calculated as 0.8Ho / B = 6.4. Since 0.58⬍ H / B共=3.5兲 ⬍ 6.4 no
increase or reduction of the calculated Nc is required. The
undrained bearing capacity is equal to qu = 4.7⫻ 120
The undrained bearing capacity factors of footings at a normal- = 564 kPa.
ized distance ␭ from the slope can be approximated, for bearing
capacity failure 共Failure Mode 2兲, through the following equa-
tions: Conclusions
• For ␭ ⬍ ␭␱:

Nc = Nco + 共5.14 − Nco兲

␭
␭␱
1+冋 冉 冊册
␤
2
1−
␭
␭␱
共6兲
The FE method was used to investigate the undrained bearing
capacity of strip footings on or near slopes. Various geometries
and soil properties were considered. The results of the analyses
• For ␭ ⱖ ␭␱: were compared to other available solutions. It was found that
most of the available methods for the calculation of the undrained
Nc = 5.14 bearing capacity factor Nc do not take account of parameters that
The above equations approximate the FE results with very good affect the value of Nc such as the distance of the footing from the
accuracy 共error of less than 2%兲 but apply only for bearing ca- slope, the slope height, and the soil properties. Solutions that do
pacity failure 共Failure Mode 2兲 and not for overall slope stability take account of all or most of the affecting parameters, and which
共Failure Mode 3兲. are based on either the limit equilibrium or upper bound methods,
were shown to be less conservative than the FE results presented
here.
Design Procedure The FE results showed that depending on the height of the
slope normalized by the footing width, three failure modes can be
Based on the design charts and equations presented above, a de- identified. At low H / B ratios the failure surface is restricted by
sign procedure for the calculation of Nc is proposed consisting of the slope height and the undrained bearing capacity factor Nc
the following steps: reduces steeply with the increase of H / B. Beyond a critical value
• The ratio cu / 共B␥兲 is calculated; of H / B, which ranges between 0.25 and 0.75 depending on the
• If cu / 共B␥兲 ⱖ 2.5, bearing capacity failure occurs and Nc can be slope angle, bearing capacity failure takes place with the failure
calculated from Eq. 共6兲; surface extending to the slope. The value of Nc in this case is
• If 1.5ⱕ cu / 共B␥兲 ⬍ 2.5 and ␤ ⱕ 30° bearing capacity failure oc- unaffected by the ratio H / B. At high values of H / B, as the ulti-
curs again and Nc can be calculated from Eq. 共6兲; mate slope height is approached, overall slope failure is observed.
• For all other combinations of cu / 共B␥兲 and ␤ linear interpola- The undrained bearing capacity factor reduces in this failure
tion between the values obtained from the design charts of mode to 0 at the ultimate slope height. For low cu / 共B␥兲 ratios and
Figs. 14–17 is required; large slope angles, this failure mode is the most dominant. In all

684 / JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010

J. Geotech. Geoenviron. Eng. 2010.136:677-685.

 other cases only the bearing capacity failure mode needs to be References
considered for the practical range of H / B ratios.
In most cases an increase of the distance of the footing from Azzouz, A. S., and Baligh, M. M. 共1983兲. “Loaded areas on cohesive
the slope increases the value of Nc. However, for low cu / 共B␥兲 slopes.” J. Geotech. Engrg., 109共5兲, 724–729.
ratios, high H / B ratios and large slope angles the undrained bear- Bowles, J. E. 共1996兲. Foundation analysis and design, 5th Ed., McGraw-
Hill, New York.
ing capacity factor may be initially reduced before increasing
Brinkgreve, R. B. J., and Broere, W. 共2006兲. Plaxis user’s manual, Plaxis
toward the value for horizontal ground surface. B.V., Delft, Netherlands.
Based on the results of the FE analyses, design charts are Castelli, F., and Motta, E. 共2008兲. “Bearing capacity of shallow founda-
proposed from which the undrained bearing capacity factor can be tions near slopes: Static analysis.” Proc., 2nd BGA Int. Conf. on Foun-
obtained for any given cu / 共B␥兲 ratio, slope angle, normalized dations, ICOF 2008, HIS BRE Press, Watford, U.K., 1651–1660.
footing distance ␭, and normalized slope height H / B. In addition, Chen, W. F. 共1975兲. Limit analysis and soil plasticity, Elsevier Science,
Amsterdam.
design equations are provided which give Nc for any combination
de Buhan, P., and Garnier, D. 共1998兲. “Three dimensional bearing capac-
Downloaded from ascelibrary.org by University of Brighton on 10/18/13. Copyright ASCE. For personal use only; all rights reserved.

of the above parameters. For the particular combination of low ity of a foundation near a slope.” Soils Found., 38共3兲, 153–163.
cu / 共B␥兲 ratio, large slope angle, high normalized footing distance Georgiadis, K. 共2009兲. “The influence of load inclination on the un-
␭, and high normalized slope height H / B, the applicability of drained bearing capacity of strip footings on slopes.” Comput. Geo-
these equations is limited as they only apply to bearing capacity tech., 37共3兲, 311–322.
Georgiadis, K., Karatzetzou, A., and Lazari, M. 共2008兲. “Undrained bear-
failure and the design charts should be used instead. Finally, a
ing capacity interaction diagrams for strip footings on slopes.” Proc.,
design procedure is proposed which is demonstrated through an
2nd BGA Int. Conf. on Foundations, HIS BRE Press, Watford, U.K.,
example. 1673–1682.
Hansen, J. B. 共1961兲. “A general formula for bearing capacity.” Bulletin
11, Danish Geotechnical Institute, Copenhagen, Denmark, 38–46.
Kumar, J., and Mohan Rao, V. B. K. 共2003兲. “Seismic bearing capacity of
Notation foundations on slopes.” Geotechnique, 53共3兲, 347–361.
Kusakabe, O., Kimura, T., and Yamaguchi, H. 共1981兲. “Bearing capacity
The following symbols are used in this paper: of slopes under strip loads on the top surfaces.” Soils Found., 21共4兲,
29–40.
B ⫽ footing width; Meyerhof, G. G. 共1957兲. “The ultimate bearing capacity of foundations
cu ⫽ undrained shear strength; on slopes.” Proc., 4th ICSMFE, 384–386.
Ec ⫽ concrete Young’s modulus; Meyerhof, G. G. 共1963兲. “Some recent research on the bearing capacity
EcA ⫽ axial stiffness of foundation; of foundations.” Can. Geotech. J., 1共1兲, 16–26.
EcI ⫽ bending stiffness of foundation; Michalowski, R. L. 共1989兲. “Three-dimensional analysis of locally loaded
Eu ⫽ undrained Young’s modulus; slopes.” Geotechnique, 39共1兲, 27–38.
H ⫽ slope height; Narita, K., and Yamaguchi, H. 共1990兲. “Bearing capacity analysis of
Nc ⫽ undrained bearing capacity factor; foundations on slopes by use of log-spiral sliding surfaces.” Soils
Nco ⫽ undrained bearing capacity factor for footing Found., 30共3兲, 144–152.
at crest of slope; Taylor, D. W. 共1948兲. Fundamentals of soil mechanics, Wiley, New York.
Ns ⫽ stability number of slope; Unified Facilities Criteria 共UFC兲. 共2005兲. “Geotechnical engineering pro-
cedures for foundation design of buildings and structures.” Rep. No.
qu ⫽ undrained bearing capacity;
UFC 3-220-01N, UFC.
Vu ⫽ ultimate vertical load;
Vesic, A. S. 共1975兲. “Bearing capacity of shallow foundations.” Founda-
␣ ⫽ angle of kinematic mechanism;
tion engineering handbook, H. F. Winterkorn and H. Y. Fang, eds.,
␤ ⫽ slope angle; Van Nostrand Reinhold, New York.
␥ ⫽ bulk unit weight; WSDOT. 共2005兲. Geotechnical design manual M46-03, Washington State
⌬共␥H / cu兲 ⫽ margin of safety of slope; Department of Transportation, Wash.
␭ ⫽ normalized footing distance; Yu, H. S., Salgado, R., Sloan, S. W., and Kim, J. M. 共1998兲. “Limit
␭o ⫽ critical normalized footing distance; and analysis versus limit equilibrium for slope stability.” J. Geotech.
␮ ⫽ Poisson’s ratio. Geoenviron. Eng., 124共1兲, 1–11.

JOURNAL OF GEOTECHNICAL AND GEOENVIRONMENTAL ENGINEERING © ASCE / MAY 2010 / 685

J. Geotech. Geoenviron. Eng. 2010.136:677-685.