You are on page 1of 11

Available online at www.sciencedirect.

com

Computers and Geotechnics 35 (2008) 144–154


www.elsevier.com/locate/compgeo

Bearing capacity of shallow foundations on rocks obeying a modified


Hoek–Brown failure criterion
Z. Saada a, S. Maghous b,*
, D. Garnier c

a
Laboratoire de Génie Civil, ENIT, Tunis, Tunisia
b
Centro de Mecânica Aplicada e Computacional, UFRGS, Porto Alegre, Brazil
c
Laboratoire des Matériaux et Structures du Génie Civil, ENPC, France

Received 27 November 2006; received in revised form 1 June 2007; accepted 5 June 2007
Available online 17 July 2007

Abstract

The bearing capacity of shallow foundations resting on a modified Hoek–Brown rock mass is investigated within the framework of the
kinematic approach of limit analysis theory. At the rock material nivel, closed-form expressions are obtained for the support functions of
the rock failure criterion. These expressions of the P-functions are then applied to derive upper bounds estimates of the load-bearing
capacity by implementing two classes of failure mechanisms. The numerical results are first compared to existing limit analysis solutions,
showing the efficiency of the present approach. The effects of the loading parameters are then discussed at the light of these preliminary
results. The results are compared with finite element limit analysis solutions, emphasizing the efficiency of the present approach. Design
tables are finally presented for practical use in rock engineering.
Ó 2007 Elsevier Ltd. All rights reserved.

Keywords: Bearing capacity; Modified Hoek–Brown criterion; Limit analysis; Failure mechanism; Rock

1. Introduction text, a limit equilibrium expression for the ultimate bearing


capacity of a strip footing is classically written as [3]:
Assessment of bearing capacity of shallow foundations qu ¼ cN c þ q0 N q þ 0:5cB0 N c ð1Þ
is one of the most common problem in civil engineering,
and numerous works have been devoted to this purpose. In the above equation, c denotes the cohesive strength, c
The methods generally used mainly fall within one of the the unit weight of the geomaterial, q0 the equivalent sur-
four categories: (a) the limit equilibrium method [1–4], charge load, and B0 is the footing width. The capacity fac-
(b) the slip-line method [5–10], (c) the limit analysis method tors Nc, Nq and Nc are only functions of the internal
[11–24], and (d) the numerical methods implementing finite friction angle.
element techniques or finite difference techniques [25,26]. It is however, well-known that failure criteria of most
Most conventional bearing capacity calculations are rocks are non-linear, as it has been emphasized by several
based on the assumption that rock strength is governed experimental works [27–34]. Among the non-linear failure
by the linear Mohr–Coulomb failure criterion. In this con- criteria proposed in the literature, the Hoek–Brown failure
criteria are considered to reasonably well model the
strength properties of isotropic rock. Nevertheless, few
studies were devoted to the assessment of bearing capacity
*
Corresponding author. Present address: Samir Maghous, CEMA- of foundations over rocks modeled by the latter strength
COM/DECIV, Universidade Federal do Rio Grande do Sul, Av. Osvaldo
Aranha, 99, 3 andar, CEP. 90035-190, Porto Alegre, RS, Brazil. Tel.: +55
criteria. In this context, the works of Serrano, Olalla and
51 33083588; fax: +55 51 33083999. co-authors seem to be among the pioneer contributions
E-mail address: maghous@ppgec.ufrgs.br (S. Maghous). in the domain [8–10,35–37].

0266-352X/$ - see front matter Ó 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2007.06.003
Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154 145

Disregarding the rock weight, Serrano and Olalla [8] Hoek–Brown strength criterion are provided in Section 3.
proposed a methodology for the determination of the bear- Section 4 describes two failure mechanisms implemented
ing capacity of a shallow foundation resting on Hoek– for the upper bound estimates of the bearing capacity.
Brown rock type, implementing the characteristics method Numerical results together with some comparisons with
in order to solve the differential equation systems govern- existing solutions are finally given in Section 5.
ing the stress field. The problem of bearing capacity at
the tip of a pile embedded in rock was also analyzed in Ser- 2. Problem description
rano and Olalla [36,37]. The related results proved to be in
accordance with previous experimental and theoretical The plane strains bearing capacity problem under con-
results in the case of soft rocks (uniaxial compressive sideration is described in Fig. 1. A strip footing of width
strength of the rock <20–30 MPa), whereas the ultimate B0 rests upon a homogeneous rock mass of unit weight c.
bearing capacity of hard rocks was overestimated. In Ser- Let Q be the force per unit transversal length exerted
rano and Olalla [9,10], the bearing capacity problem of a through the footing. The loading mode in the considered
strip foundation resting on a weightless rock crossed by a material system is defined by three parameters: the specific
family of joints was studied. The strength properties of weight c, the surface surcharge q0, and the equivalent foot-
the intact rock were modeled by a Hoek–Brown criterion, ing load q = Q/B0. The ultimate value of q will be denoted
while a Mohr–Coulomb strength condition was adopted by qu.
for the joints. In this study, different failure mechanisms As regard the strength capacities of the constitutive rock
were identified depending on the loading conditions and material, isotropy is assumed in the sequel. At the macro-
joints characteristics. scopic scale (i.e. the scale of the structure), this assumption
In a recent paper, Yang and Yin [11] developed an upper may reasonably be adopted for intact rocks or heavily
bound solution for ultimate bearing capacity of a strip jointed rock masses (i.e. sufficiently dense and randomly
footing adopting a modified Hoek–Brown failure criterion. distributed joints). This means in particular that the case
In some extent, the authors extend to the case of non-linear of rocks with few discontinuities cannot be considered in
failure criterion, previous works using the linear Mohr– this framework.
Coulomb failure criterion (Chen [13], Michalowski [16], Moreover, the strength capacity of the rock mass is
Soubra [12], to cite a few). Actually, the idea implemented assumed to be described by a modified Hoek–Brown con-
in [11] consisted in replacing the original Hoek–Brown dition [39,40]:
strength criterion by an ‘optimal’ tangential Mohr–Cou-  n
r1
lomb domain. Upper bound estimates for the bearing F ðrÞ ¼ r1  r3  rc m þ s 6 0 ð2Þ
capacity are derived implementing the kinematic approach rc
of limit analysis. The generalized tangential technique is where r1 and r3 denote respectively the major and minor
thus used to evaluate the optimal tangential Mohr–Cou- principal stresses (stresses are counted positive in tensile),
lomb domain and the corresponding upper bound bearing and rc P 0 is the uniaxial compressive strength. The
capacity values. parameters m, s and n depend on the geological strength in-
Unlike the approach developed in [11], the main objec- dex (GSI) [40]:
tive of this paper is to derive upper bound estimate for  
the bearing capacity preserving exactly the original modi- m GSI  100
¼ exp ð3Þ
fied Hoek–Brown failure criterion. The closed-form expres- mi 28  14D
 
sion of the maximum resistant work rate developed in any GSI  100
virtual strain field is derived. It is found in particular that s ¼ exp ð4Þ
9  3D
the obtained results improve those provided by Yang and     
1 1 GSI 20
Yin [11]. n ¼ þ exp   exp  ð5Þ
More recently, Merifield et al. [38] developed a finite ele- 2 6 15 3
ment approach to compute lower and upper estimates of where D is a disturbance coefficient that varies from 0.0 for
the ultimate bearing capacity, based on the implementation the undisturbed in situ rock masses to 1.0 for very dis-
of linear and non-linear programming techniques and turbed rock masses. mi is the value of parameter m for in-
adopting a smooth approximation for the modified tact rock and can be obtained from experiment measures.
Hoek–Brown yield surface. With respect to the numerical One may refer to Hoek [41] where approximate values of
approach developed in [38], the main difference of the pres- parameter mi are given for some typical rocks.
ent work lies on the fact that upper bound solutions are In the rn  s stress plane, where rn and s are the normal
derived herein from the direct analysis of failure and shear stresses, the modified Hoek–Brown failure crite-
mechanisms. rion (2) can be drawn as a curve (Fig. 2). The equation of
The paper is organized as follows. Description of the the latter may be derived as the convex envelope of ‘Mohr–
considered problem is detailed in Section 2. After a brief Coulomb’ lines
recall of the kinematic approach of limit analysis, explicit
expressions for the support functions of the modified s ¼ ct  rn tan ut ð6Þ
146 Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154

Q = q B0

q0 q0

B0

γ homogeneous rock

Fig. 1. Problem geometry and loading mode.

τ W e ðc; q0 ; qu ; U Þ 6 W mr ðU Þ 8U ð8Þ

In the above fundamental inequality, We(c,q0,qu, U) de-


ϕt
notes the work done by the external forces, whereas
Wmr(U) represents a positive quantity, named the maxi-
ct mum resisting work. Its general expression is
Z Z h i
W_ mr ðU Þ ¼ P½dðxÞdX þ P mðxÞ; ½U ðxÞ dR ð9Þ
X R

σn In this formula, d is the strain rate field associated with U,


σn
[U] is the jump of U at a point x when crossing a possible
Fig. 2. Representation of the modified Hoek–Brown criterion in the Mohr velocity discontinuity surface R following its normal m(x).
plane.
The P-functions, called support functions are defined by
duality from the strength condition F(r) 6 0:
ut = arctan(os/orn) being the tangential friction angle P½d ¼ supfr : djF ðrÞ 6 0g ð10Þ
and ct the intercept of the straight line to s axis. ut and ct r

are functions of the normal stress rn and their expression P½m; ½U  ¼ supf½U   r  mjF ðrÞ 6 0g ð11Þ
may be found in the literature. For instance, the relation- r

ship between these parameters writes [11]


 n=1n 3.2. Support functions
cos ut mnð1  sin ut Þ
ct =rc ¼
2 2 sin ut
  1=1n The practical implementation of the kinematic approach
tan ut sin ut mnð1  sin ut Þ of limit analysis requires to preliminary derive the closed-
 1þ
m n 2 sin ut form expressions of the P-functions for the modified
s Hoek–Brown strength criterion.
þ tan ut ð7Þ
m It is first observed that:

^ with 1
3. Limit analysis kinematic approach and support functions P½m; ½U  ¼ P½d d^ ¼ ðm  ½U  þ ½U   mÞ ð12Þ
2
of the modified Hoek–Brown criterion
which means that the determination of the P-functions re-
After a brief recall of the upper bound theorem of limit duce to that of P[d], P[m; [U]] being therefore deduced from
analysis, the closed-form expressions for the support func- the latter through (12).
tions of the modified Hoek–Brown convex of admissible Classical results of convex analysis indicate that:
stresses are provided.
 P[d] takes a finite value if and only if d belongs to the cone
of outward normals to the yield surface. This condition
3.1. Limit analysis kinematic approach
means that d is necessary of the form d ¼ k oF or
with
Given any kinematically admissible velocity field U k P 0. If a strain rate field d complies with this condition,
(termed failure mechanism in the sequel), the kinematic the- the corresponding velocity U is termed as relevant [42,43].
orem of limit analysis states that a necessary condition for  Given any relevant velocity field U, P[d] actually writes
the system under consideration to remain stable under the
external loading (c, q0, qu) reads [42,43] P½d ¼ rH : d ð13Þ
Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154 147

where rw represents the stress state located on the Since the closed-form expressions of the support function
boundary of the yield surface and where the outward of the modified Hoek–Brown criterion have been deter-
normal is parallel to d (Fig. 3). mined, the maximum resisting work Wmr(U) can be com-
We now apply the above considerations to the case puted in any virtual failure mechanism U. The kinematic
where the strength properties are defined by the modified approach may therefore be implemented to derive upper
Hoek–Brown failure condition (2). As regards the condi- bound estimates for the ultimate bearing capacity.
tion of relevancy, it may be shown that a velocity field U It is worth noting that formulas (15), (16), (18) and (19)
is relevant if the associated strain rate satisfies the following extend the results already obtained by Garnier and Mag-
inequality hous [44] in the particular situation n ¼ 12.
trd > 0 ð14Þ
4. Application to the bearing capacity problem
Furthermore, it is found after mathematical developments
that for such velocity fields The purpose of the following section is to derive upper
!n bound estimates for the bearing capacity of the foundation
src  n 1
 mMðdÞ 1n by making use of the kinematic approach implemented on
P½d ¼ trd þ rc n1n  n1n ð15Þ
m trd the problem sketched in Fig. 1. Two classes of failure
mechanisms will be considered in the subsequent analysis.
where function MðdÞ is defined by the relation
1 4.1. Generalized Prandtl-type failure mechanism
MðdÞ ¼ ½maxð0; d 1 Þ þ maxð0; d 2 Þ þ maxð0; d 3 Þn
ð16Þ Generalizing to the situation of modified Hoek–Brown
criterion, the classical Prandtl’s failure mechanism usually
in which d1, d2 and d3 represent the eigenvalues of d.
considered for soil or rock material with a Mohr–Coulomb
Likewise function P[m;[U]] relative to a velocity discon-
failure criterion, the class of failure mechanisms depicted in
tinuity (Eq. (11)) might be directly computed from (12)
Fig. 4 is analyzed herein. It will be referred to as mecha-
and (14)–(16). The traduction of the relevancy condition
nism (M1).
(14) in terms of velocity jump simply reads
Such a mechanism involves four angular parameters: a
½U   m > 0 ð17Þ and a 0 defining the triangular wedge AA 0 B located beneath
If the field U complies with the above inequality every- the foundation, d which is the aperture of the fan zone
where along its discontinuity surface, one gets ABC bounded by the arc of log-spiral BC of focus A,
and u defining the position of the point C
src
P½m; ½U  ¼ ½U   m
m AC ¼ ABed tan u ð20Þ
 n mMðm; ½U Þ1n
n

þ rc n  n
1n
1
1n ð18Þ The velocity field is then defined in each zone as follows:
½U   m
 The triangular wedge AA 0 B is given a uniform transla-
where
tion motion U = u = ct, whose orientation is character-
1 1
ized by angle u with respect to the side A 0 B.
Mðm; ½U Þ ¼ ðj½U j  ½U   mÞn ð19Þ
21=n  The velocity at any point M of the fan ABC is given by
U ðMÞ ¼ vðhÞ ¼ v0 eh tan u eh ð21Þ
where h is the polar orientation of M with respect to AB,
σkl and eh represents the unit orthoradial vector defined
accordingly, so that the velocity profile along any radius
d Π ( d= ) = σ∗
=
: d
= AM is constant;
=
 The triangular wedge ACD is also given a uniform
σ∗ translation motion U = w perpendicular to AC and
=
whose magnitude is equal to v0 ed tan u .
 The remaining part of the structure is kept motionless.

σij Such a velocity field involves discontinuities inclined at


angle u with the segments A 0 B, AB and CD as well as
F=0 along the arc BC, but it remains continuous across the seg-
ment AC. It can be readily seen that U is relevant in the
d
=
Π ( =d ) = + ∝
sense of (14) within the fan ABC, and in the sense of (17)
along the discontinuities lines A 0 B, AB, CD and BC. The
Fig. 3. Graphical interpretation of the P-function in the stress space. corresponding velocity hodograph is sketched in Fig. 5,
148 Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154

Q = q B0

q0
A’ A
α’ α δ D
u
θ
ϕ

w
B M v (θ ) ϕ
C
Fig. 4. Generalized Prandtl-type failure mechanism.

p
0 < a; 0 < d; 0<u< ; u < a0 ; a þ a0 < p;
v ( δ) = w 2
p
δ þu<aþd<p ð25Þ
2
θ
ϕ v (0 ) 4.2. Multi-wedge translation failure mechanism
u ϕ
// AC Referred to as mechanism (M2), this failure mechanism
is directly adapted from that originally presented by Sou-
// AB // A’B bra [12] for the calculation of ultimate bearing capacity
Fig. 5. Velocity hodograph for the generalized Prandtl failure mechanism of foundation resting on a Mohr–Coulomb soil. Since this
(M1). failure mechanism is symmetric, only half of the problem
domain is represented in Fig. 6.
Geometrically, the zone in motion is defined by k + 1
triangular wedges. The wedge AA 0 B (wedge (0) in Fig. 6)
which allows to calculate v0 = v(h = 0) and w in function of beneath the strip footing is characterized by the angle h.
u, a, a 0 , d and u. Each wedge (i), i = 1, . . . , k, is characterized by the length
The work performed by the external forces (surcharge of the base di, the angles ai and bi, and the length Li of
q0, rock unit weight c and foundation load q) in such a fail- the interface (i  1)/(i).
ure mechanism write The velocity field U, depicted in Fig. 7, is defined in each
B20 wedge as follows:
W e ðU Þ ¼ qu sinða0  uÞ þ ucF1 þ B0 uq0 F2 ð22Þ
2
 The wedge (0) is defined by a downward vertical trans-
where F1 and F2 are non-dimensional functions of the lation v0.
angular parameters a, a 0 , d and u. Their expressions are  Each wedge (i), i = 1, . . . , k, moves with a uniform trans-
provided in Appendix A. lation vi inclined at angle ui with respect to the base.
On the other hand, the maximum resisting work writes
by virtue of (15) and (18) Such a velocity field is relevant in the sense of (17) and
W mr ðU Þ ¼ rc B0 uF3 ð23Þ involves discontinuities [U] = vi  vi1 across the interface
Li between wedges (i) and (i  1), and [U] = vi along the
where expression of the non-dimensional function F3 is
base di. Furthermore, [U] is inclined at angle ui > 0 with
also given in Appendix A.
respect to the interface Li as well as to the base di. Taking
Finally, one obtains from the fundamental inequality (8)
into account these kinematics considerations, the velocity vj
rc F3  0:5B0 cF1  q0 F2 of each wedge can be expressed as a function of the velocity
qu 6 B0 min ¼ qu1 ð24Þ
0
a;a ;d;u sin ða0  uÞ v0 and the set of angular parameters h, ai, bi, ui
where qu1 denotes the best upper-bound estimate of qu (i = 1, . . . , k).
which can be obtained from exploring the above class of The work done by the external forces writes
failure mechanisms (M1) through a minimization proce- B20
dure with respect to the angular parameters a, a 0 , d and W e ðU Þ ¼ qv0 þ v0 cG1 þ B0 v0 q0 G2 ð26Þ
2
u, which is carried out numerically. It should be pointed
out that these parameters are subjected to the following where expressions of the non-dimensional functions G1 and
constraints: G2 are provided in Appendix B.
Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154 149

A’ A C
αk
θ
(0) (k)
L1 α1 dk
αi Lk βk

(1)
β1 Li
B
(i)
d1
βi
di

Fig. 6. Geometry of the symmetrical failure mechanism (M2) [12].

A’ A C

(0) (k)
vk
v0 (1) ϕk

v1 [U]
B ϕi
(i)
ϕ1 vi

ϕi

Fig. 7. Kinematic description of the failure mechanism (M2).

8
Since the considered velocity field is piecewise uniform, >
> 0 < ai ;bi ; ui < p; 0 < h < p2 ;0 < ai þ bi < p;ui < b2i
the maximum resisting work reduces to the contribution >
<
b1  u1  p2  h < 0; biþ1  ai  bi  uiþ1 þ ui < 0; h
P[m; [U]] of the velocity jump [U] ð29Þ
>
> Pk
>
: þ ai ¼ p
W mr ðU Þ ¼ rc B0 v0 G3 ð27Þ
i¼1
(expression of G3 is also given in Appendix B).
It comes from the necessary condition We(U) 6 Wmr(U)
5. Numerical results
that
qu 6 B0 min frc G3  0:5B0 cG1  q0 G2 g ¼ qu2 ð28Þ This section provides some numerical results for the
h;ai ;bi ;ui
bearing capacity derived from the failure mechanisms
where qu2 represents the best upper-bound estimate of qu (M1) and (M2). It must be recalled that the upper bounds
which can be obtained from exploring the above class of qu1 and qu2 defined respectively in (24) and (28) are
failure mechanisms (M2) through a minimization proce- obtained through constrained minimization procedures
dure with respect to the angular parameters h, ai, bi and with respect to corresponding failure mechanisms parame-
ui. These parameters must comply with the following ters. In mechanism (M1), they correspond to the four vari-
constraints: ables a, a 0 , d, u associated to the constraints (25). The
150 Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154

minimization parameters in mechanism (M2) are the angles [11]


h, ai, bi and ui (i = 1, . . . , k), which corresponds to 3k + 1 Mechanism (M1)
variables with the constraints (29). Theoretically, increas- Mechanism (M2)
9
ing the number of triangular wedges will improve the upper
bound estimate qu2. Actually, Soubra [12] showed in the
8
context of a Mohr–Coulomb failure criterion that the num-
ber of triangular wedges k = 14 proves to be sufficient. This
value has been adopted throughout the present analysis. 7

For illustrative purpose, Fig. 8 shows an example of the

qu (MPa)
optimal failure mechanism geometry obtained when 6
exploring the family of mechanisms (M2).
Combining the two kinematical approaches described in 5
Sections 4.1 and 4.2, the optimal upper bound estimate is
defined as the minimum of qu1 and qu2 4

qu 6 qþ
u ¼ minðqu1 ; qu2 Þ ð30Þ
3
0 10 20 30 40 50
q0 (kPa)
5.1. Effects of the loading parameters
Fig. 9. Ultimate bearing capacity upper bound estimate qu versus
In the approach proposed in [11], the original modified surcharge load q0 (B0 = 1 m, D = 0, mi = 10, GSI = 30, rc = 10 MPa
Hoek–Brown strength criterion is replaced by an ‘optimal’ and c = 0).
tangential Mohr–Coulomb failure condition. The same
type of mechanism (M2) has been considered by the 10
authors for the calculations of upper bound estimates for
q 0 =40 kPa
the bearing capacity.
Adopting the following data: B0 = 1 m, D = 0, mi = 10,
9.5
GSI = 30, rc = 10 MPa and c = 0, Fig. 9 displays the vari- q 0 =30 kPa
ations of the upper bound estimates qu versus the surcharge
qu [MPa]

load q0. This figure shows the results derived from mecha-
9 q 0 =20 kPa
nisms (M1) and (M2) as well as those given in [11].
For the specific data characterized above, mechanism
(M2) leads to estimates lower than mechanism (M1) for q 0 =10 kPa
the whole values of q0, which means that in this case qþ
u ¼ qu2 .
8.5

On the other hand, it rises from the results depicted in


Fig. 9 that the estimates derived in this work significantly
improve those obtained in [11]. Since both the approaches 8
20 21 22 23 24
implement the same failure mechanism (M2), this improve-
γ [kN/m3]
ment is clearly attributed to the fact that we deal in this
contribution with the original modified Hoek–Brown Fig. 10. Ultimate bearing capacity upper bound estimate qu obtained for
strength criterion, whereas in [11] the latter is replaced by the failure mechanism (M2) versus rock unit weight c (B0 = 1 m, D = 0,
a unique ‘optimal’ tangential one for the whole rock zone mi = 17, GSI = 30 and rc = 10 MPa).
in motion within the failure mechanism (M2). Such proce-
dure actually amounts to overestimating the local strength estimates than (M1) (i.e. qþ u ¼ qu2 ), and in turn than the
properties of the rock material. upper bound obtained in [11].
The effect of the rock unit weight c on the ultimate bear- Besides, it turns out that for the whole values considered
ing capacity upper bound estimates obtained for the failure for c and q0, qu2 is lower than qu1, which is in turn lower
mechanism (M2) is illustrated in Fig. 10, adopting than the upper bound obtained in [11]. For example, if
B0 = 1 m, D = 0, mi = 17, GSI = 30 and rc = 10 MPa. It c = 20 kN/m3 and q0 = 20 kPa, the upper bound estimate
is worth noting that, once again (i.e. for the considered of qu given in [11] is equal to 14.540 MPa, whereas
data), the failure mechanism (M2) always leads to lower qu1 = 13.942 MPa and qu2 = 8.932 MPa.

Fig. 8. Geometry of the optimal failure mechanism (M2) obtained for D = 0, mi = 17, GSI = 60, rc = 10 MPa, c = 0 and q0 = 0.
Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154 151

Instead of (1), the ultimate bearing capacity qu are some- The latter ones were found to be very close to the numerical
times put in the form [11] average values provided in [38] for almost values of
pffiffi GSI 2 [10, 80] and mi 2 [5, 25] (D has been fixed at zero).
qu ¼ src N r þ q0 N q þ 0:5cB0 N c ð31Þ
This comparison, which is illustrated in Table 1 in the par-
where the non-dimensional parameters Nr, Nq and Nc are ticular case mi = 10, emphasizes the high efficiency of the
bearing capacity factors related to the uniaxial compressive approach described in the present paper.
strength, the surcharge load and the unit weight of the For practical use in rock engineering, several computa-
rock, respectively. The above preliminary results suggest tions of this bearing capacity factor have been performed,
the following comments:

 The linear dependence of qu with respect to c, postulated Table 1


in (31), seems to be confirmed by the present analysis Bearing capacity factor Nr for weightless rock: D = 0 and mi = 10
and by that developed in [11]. GSI Nc Present work Nc [38] Relative difference (%)
 The rock unit weight (within the range of its usual val- 10 11.561 11.427 1.1
ues) has a small effect on the bearing capacity when 20 17.848 17.796 0.3
compared to that of the surcharge load. 30 19.513 19.396 0.6
 The upper bound qu2 shown in Fig. 9 exhibits (at least 40 18.582 18.472 0.6
50 16.746 16.678 0.4
for the considered data) a non-linear dependence on
60 14.784 14.736 0.3
the surcharge load q0, which is not the case if the a priori 70 12.977 12.939 0.3
form (31) is adopted. 80 11.402 11.376 0.2

Keeping in mind that these comments are based upon


partial results, in the sense they are only upper bound esti- Table 2
Bearing capacity factor Nr for five types of rocks (D = 0.0)
mates of the bearing capacity, they must thus be considered
carefully. However, they indicate that the dependence nat- GSI mi = 7 mi = 10 mi = 15 mi = 17 mi = 25
ure of qu on the loading parameters remains to be more 5 5.016 7.054 10.718 12.285 19.045
investigated. 10 8.341 11.561 17.201 19.546 29.430
15 11.231 15.334 22.360 25.227 37.064
20 13.241 17.848 25.588 26.713 41.376
5.2. Further comparisons and design tables 25 14.358 19.150 27.077 30.244 42.951
30 14.760 19.513 27.286 30.369 42.637
We deal in this section with the situation of weightless 35 14.670 19.240 26.660 29.587 41.167
rock and absence of surcharge load: c = 0 and q0 = 0. 40 14.263 18.582 25.551 28.287 39.065
45 13.676 17.713 24.189 26.726 36.678
A classical dimensional analysis argument indicates that
50 13.002 16.746 22.730 25.067 34.212
the bearing capacity qu may be put in the form 55 12.298 15.756 21.265 23.411 31.793
er 60 11.598 14.784 19.845 21.814 29.484
qu ¼ r c N ð32Þ
65 10.923 13.854 18.499 20.303 27.319
e r is a function of D, GSI and mi defining
where the factor N 70 10.283 12.977 17.239 18.891 25.309
75 9.683 12.160 16.070 17.583 23.454
the strength parameters of the rock material, i.e.
er ¼ Ne r ðD; GSI;mi Þ. 80 9.125 11.402 14.989 16.376 21.748
N
In order to meet form (31) in the particular case c = 0
and q0 = 0, the above equation is rewritten as Table 3
pffiffi pffiffi Bearing capacity factor Nr for five types of rocks (D = 0.1)
qu ¼ src N r with N r ¼ N e r= s ð33Þ
GSI mi = 7 mi = 10 mi = 15 mi = 17 mi = 25
The results derived from the present analysis are first 5 4.630 6.487 9.848 11.259 17.417
compared to finite element solutions obtained by Merifield 10 7.918 10.951 16.259 18.480 27.774
et al. [38] which implemented linear and non-linear pro- 15 10.848 14.804 21.565 24.330 35.720
gramming techniques in the framework of limit analysis. 20 12.953 17.459 25.013 28.050 40.417
25 14.156 18.885 26.689 29.809 42.319
The same modified Hoek–Brown failure criterion (2) was 30 14.632 19.341 27.046 30.096 42.253
used by these authors. The computed upper and lower 35 14.590 19.138 26.519 29.432 40.946
bound estimates for the bearing capacity factor 40 14.220 18.528 25.475 28.204 38.949
Ne r ¼ qu =rc obtained by these authors were found to be 45 13.655 17.685 24.154 26.685 36.622
within 5% of each other. As a consequence, average values 50 12.994 16.735 22.718 25.054 34.195
55 12.297 15.755 21.266 23.411 31.795
of the upper and lower bound bearing capacity factor can 60 11.601 14.788 19.851 21.821 29.497
be considered as highly accurate estimates for the true col- 65 10.927 13.859 18.508 20.313 27.333
lapse e r have been divided
pffiffiload. These
e paverage
ffiffi values of N 70 10.288 12.983 17.248 18.901 25.323
by s (i.e. N r = s ¼ N r ) and compared to the results 75 9.687 12.165 16.078 17.592 23.466
derived from the analysis of mechanisms (M1) and (M2). 80 9.128 11.406 14.995 16.383 21.758
152 Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154

Table 4 considering the following ranges for the strength parame-


Bearing capacity factor Nr for five types of rocks (D = 0.2) ters : D 2 [0, 0.6], GSI 2 [5, 80], mi 2 [7, 25]. The obtained
GSI mi = 7 mi = 10 mi = 15 mi = 17 mi = 25 numerical results are given in Tables 2–6 of Appendix C.
5 4.228 5.884 8.912 10.169 15.682
10 7.445 10.267 15.209 17.266 25.921 6. Conclusion
15 10.396 14.176 20.635 23.265 34.102
20 12.586 16.949 24.271 27.219 39.184
25 13.889 18.510 26.153 29.202 41.440
The classical problem of assessing the ultimate bearing
30 14.442 19.079 26.664 29.676 41.648 capacity of shallow foundations resting on a rock mass
35 14.460 18.962 26.262 29.145 40.540 has been investigated within the framework of the kine-
40 14.131 18.406 25.303 28.009 38.676 matical approach of limit analysis. The paper focused
45 13.594 17.603 24.037 26.557 36.440 on the situation where the strength properties of the rock
50 12.952 16.679 22.639 24.965 34.070
55 12.267 15.716 21.211 23.351 31.709
material can be described by a modified Hoek–Brown fail-
60 11.581 14.761 19.815 21.779 29.437 ure criterion. At the material level, closed-form expres-
65 10.914 13.841 18.481 20.284 27.292 sions of the support functions (i.e. P-functions) for such
70 10.278 12.971 17.230 18.881 25.294 a criterion have been derived. The latter ones are neces-
75 9.681 12.156 16.064 17.578 23.446 sary for the practical implementation of the upper bound
80 9.124 11.400 14.986 16.373 21.745
approach of limit analysis. It is reminded that since the P-
functions have been determined, they can readily be
applied to the stability analysis of any structure involving
such a failure criterion, like slope stability or tunnel face
Table 5 stability.
Bearing capacity factor Nr for five types of rocks (D = 0.3)
As a structural application, the kinematic approach has
GSI mi = 7 mi = 10 mi = 15 mi = 17 mi = 25 been implemented for the problem of bearing capacity
5 3.804 5.283 7.924 9.052 13.877 making use of two failure mechanisms: a generalized Pra-
10 6.907 9.519 14.046 15.934 23.849 ndtl-type failure mechanism and a multi-wedge translation
15 9.879 13.451 19.530 22.000 32.201
20 12.139 16.325 23.343 26.176 37.631
mechanism. The numerical results have been first com-
25 13.527 18.015 25.435 28.386 40.263 pared to previous results. In particular, it has been shown
30 14.172 18.710 26.131 29.067 40.770 that they lead to a significant improvement of the ultimate
35 14.256 18.683 25.864 28.700 39.898 bearing capacity estimates provided in the recent work of
40 13.978 18.200 25.008 27.679 38.206 Yang and Yin [11].
45 13.480 17.449 23.817 26.310 36.093
50 12.868 16.564 22.474 24.783 33.811
The analysis of the effects of the loading parameters on
55 12.204 15.630 21.087 23.213 31.514 the bearing capacity confirms, as already pointed out in
60 11.532 14.694 19.721 21.674 29.288 [11], the linear dependence of qu with respect to the specific
65 10.876 13.790 18.410 20.203 27.180 weight c of the rock material. As regards the dependence of
70 10.250 12.932 17.176 18.821 25.210 qu with respect to the surcharge load q0, the preliminary
75 9.660 12.129 16.025 17.533 23.383
80 9.109 11.380 14.958 16.341 21.700
results given in this paper suggested that this aspect
remains to be clarified and a linear-type dependence could
be questionable. More investigations should be therefore
done to assess the nature of the dependence qu = qu(q0).
The comparison of the results with finite element limit
Table 6 analysis solutions given in [38] showed the efficiency of
Bearing capacity factor Nr for five types of rocks (D = 0.6) the proposed approach. For practical use in rock engineer-
GSI mi = 7 mi = 10 mi = 15 mi = 17 mi = 25 ing, design tables for the bearing capacity factor Nr are
5 2.492 3.372 4.911 5.571 8.357 finally given in the last part of this paper.
10 5.025 6.838 9.946 11.189 16.525
15 7.797 10.506 15.081 16.948 24.599 Acknowledgements
20 10.166 13.554 19.219 21.496 30.704
25 11.790 15.592 21.862 24.366 34.364
30 12.697 16.676 23.173 25.743 35.947 The preparation of the paper had received financial sup-
35 13.043 17.023 23.458 25.995 36.008 port from the Ecole Nationale des Ponts et Chaussées
40 12.990 16.854 23.065 25.503 35.088 (ENPC-Paris) and from the Commission Universitaire au
45 12.685 16.363 22.256 24.562 33.601 Développement (UCL-Brussels) through project
50 12.223 15.689 21.219 23.377 31.815
PIC2004-UCL/ENIT. The authors have greatly appreci-
55 11.687 14.928 20.081 22.088 29.920
60 11.118 14.134 18.919 20.780 28.022 ated the financial support of both institutions.
65 10.549 13.347 17.779 19.498 26.184
70 9.995 12.588 16.686 18.274 24.439 Appendix A
75 9.466 11.866 15.653 17.119 22.801
80 8.967 11.189 14.687 16.039 21.276
 Parameter F1
Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154 153

F1 ¼ f11 þ f12 þ f13 ð34Þ Appendix B


with
 Parameter G1
sinðaÞ sinða0 Þ
f11 ¼ sinða0  uÞ ð35Þ tan ðhÞ cos ðh  u1 Þ
sinða þ a0 Þ G1 ¼ þ
2 2 cos2 ðhÞ sinðb1  2u1 Þ
sin2 ða0 Þ sinða þ dÞ cosða þ dÞ !
f12 ¼  X k
sinðai Þ sinðbi Þ X
i1
sin2 ða þ a0 Þ cosða þ d  uÞ  sin bi  h  aj  ui
 sinðai þ bi Þ
p  i¼1 j¼1
 cos a þ a0  u   u e3d tan u ð36Þ Y
2 i1
sinðaj þ bj  uj  ujþ1 Þ sin2 ðbj Þ

 ð45Þ
sin ða Þ cos a þ a0  u  p2  u
2 0
j¼1 sinðbjþ1  2ujþ1 Þ sin2 ðaj þ bj Þ
f13 ¼ 2
sin ða þ a0 Þ cosðuÞ
  Parameter G2
3 tanðuÞ cosða þ dÞ þ sinða þ dÞ 3d tan u P
 e cosðh  u1 Þ sinðbk  h  k1
9 tan2 ðuÞ þ 1 j¼1 aj  uk Þ sinðbk Þ
 G2 ¼
3 tanðuÞ cosðaÞ þ sinðaÞ cosðhÞ sinðbk  2uk Þ sinðak þ bk Þ
 ð37Þ Y
k 1
9 tan2 ðuÞ þ 1 sinðbj Þ sinðaj þ bj  uj  ujþ1 Þ
 ð46Þ
 Parameter F2 j¼1
sinðbj  2uj Þ sinðaj þ bj Þ

sinða0 Þ cosða þ dÞ  Parameter G3


F2 ¼ 
sinða þ a0 Þ cosða þ d  uÞ G3 ¼ g31 þ g32 þ g33 ð47Þ
 p 

 cos a þ a0  u   u e2d tan u ð38Þ
2 Denoting by Vi ¼ viþ1  vi , i = 0, . . . , k  1, the velocity
 Parameter F3 jump at the interface Li+1 between triangles (i + 1) and
(i), Vi its modulus and Vin the corresponding normal com-
F3 ¼ f31 þ f32 þ f33 þ f34 þ f35 ð39Þ ponent when crossing the interface Li+1, the above quanti-
ties are defined as
with

0  1 1
1n
sinða0 Þ cos aþa0 u p2 u   V0 V0n
f31 ¼ 1 B s V0n n 1 n C 2v0
sinðaþa0 Þ 2cosðuÞ g31 ¼ @ þ n1n  n1n m1n   n A
"  1# cosðhÞ m v0 V0n 1n
s  n 1
 n 1sinðuÞ 1n 2dtanu v0
 þ n1n n1n m1n ðe 1Þ
m 2sinðuÞ ð48Þ
ð40Þ 2  13
Vi Vin 1n
sinðaÞ X
k1  
f32 ¼ sinðuÞ 6 s Vin n 1 n 2v0 7
sinðaþa0 Þ g32 ¼ 4 þ n1n  n1n m1n   n 5
" # i¼1
m v 0 Vin 1n

s  n 1
 n 1sinðuÞ1n 1 v0
 þ n1n n1n m1n ð41Þ

m 2sinðuÞ 1 Y i
sin bj
 ð49Þ
sinða0 ÞtanðuÞ 0 p cos ðhÞ j¼1 sinðaj þ bj Þ
f33 ¼ sin aþa u
sinðaþa0 Þ 2
" # Xk
s  n 1
 n 1sin ðuÞ1n 1
g33 ¼
vi
2 sinðui Þ
di
 þ n1n n1n m1n ð42Þ v 0 B 0
m 2sin ðuÞ i¼1

" #
sinða0 Þ cos aþa0 u p2 u s  n 1
 n 1  sinðu Þ1n
1
i
f34 ¼  þ n1n  n1n m1n ð50Þ
sinðaþa0 Þ 2 m 2 sinðui Þ
"  1#
s   1sin ðuÞ 1n 2dtanu
 þ n n
n
1n
1
1n m
n
1n ðe 1Þ where
m 2sin ðuÞ
B0 sinðai Þ Y i1
sinðbj Þ
ð43Þ di ¼ ð51Þ
2 cosðhÞ sinðai þ bi Þ j¼1 sinðaj þ bj Þ
sinða0 ÞsinðaþdÞ  p 
0
f35 ¼ cos aþa u u tan ðuÞ
sinðaþa0 ÞcosðaþduÞ 2
" # Appendix C
s  n 1
 n 1sin ðuÞ1n 1

 þ n1n n1n m1n e2dtanu ð44Þ


m 2sin ðuÞ See Tables 2–6.
154 Z. Saada et al. / Computers and Geotechnics 35 (2008) 144–154

References [24] Garnier D. Analyse par la théorie du calcul à la rupture des facteurs
de réduction de la capacité portante de fondations superficielles. PhD
[1] Meyerhof GG. The ultimate bearing capacity of foundations. thesis, Ecole Nationale des Ponts et Chaussées, France; 1995.
Geotechnique 1951;2(4):301–32. [25] Griffiths DV. Computation of bearing capacity factors using finite
[2] Silvestri V. A limit equilibrium solution for bearing capacity of strip elements. Geotechnique 1982;32(3):195–202.
foundations on sand. Can Geotech J 2003;40(2):351–61. [26] Burd HJ, Frydman S. Bearing capacity of plane-strain footings on
[3] Terzaghi K. Theoretical soil mechanics. New York: Wiley; 1943. layered soils. Can Geotech J 1997;34(2):241–53.
[4] Zhu DY, Lee CF, Jiang HD. A numerical study of the bearing [27] Agar JG, Morgenstern NR, Scott J. Shear strength and stress–strain
capacity factor Nr. Can Geotech J 2001;38:1090–6. behaviour of Athabasca oil sand at elevated temperatures and
[5] Bolton MD, Lau CK. Vertical bearing capacity factors for circular pressure. Can Geotech J 1985;24(1):1–10.
and footing footings on Mohr–Coulomb. Can Geotech J [28] Hoek E, Brown ET. Empirical strength criterion for rock masses.
1993;30(3):1024–33. ASCE J Geotech Eng 1980;106(GT9):1013–36.
[6] Hansen BJ. A general formula for bearing capacity. Bull Dan [29] Hoek E. Strength of jointed rock masses. Geotechnique 1983;33:
Geotech Inst 1961;11:38–46. 187–223.
[7] Sokolovskii VV. Statics of soil media. London: Butterworth Science; [30] Goodman RE. Introduction to rock mechanics. 2nd ed. New York:
1965 (Translator R. Jones and A. Schofield). Wiley; 1989.
[8] Serrano A, Olalla C. Ultimate bearing capacity of rock masses. Int J [31] Baker R. Inter-relations between experimental and computational
Rock Mech Mining Sci Geomech 1994;31(2):93–106. aspects of slope stability analysis. Int J Numer Anal Meth Geomech
[9] Serrano A, Olalla C. Ultimate bearing capacity of an anisotropic 2003;27:379–401.
discontinuous rock mass. Part I: basic modes of failure. Int J Rock [32] Hoek E, Brown ET. Practical estimates of rock mass strength. Int J
Mech Mining Sci 1998;35(3):301–24. Rock Mech Mining Sci 1997;34(8):1165–86.
[10] Serrano A, Olalla C. Ultimate bearing capacity of an anisotropic [33] Santarelli F. Theoretical and experimental investigation of the
discontinuous rock mass. Part II: determination procedure. Int J stability of the axisymmetric borehole. PhD thesis, University of
Rock Mech Mining Sci 1998;35(3):325–48. London, London; 1987.
[11] Yang XL, Yin JH. Upper bound solution for ultimate bearing [34] Jiang JC, Baker R, Yamagami T. The effect of strength envelope
capacity with a modified Hoek–Brown failure criterion. Int J Rock nonlinearity on slope stability computations. Can Geotech J
Mech Mining Sci 2005;42:550–60. 2003;40:308–25.
[12] Soubra AH. Upper-bound solutions for bearing capacity of founda- [35] Serrano A, Olalla C, Gonzalez J. Ultimate bearing capacity of rock
tions. ASCE J Geotech Geoenviron Eng 1999;125(1):59–68. masses based on the modified Hoek–Brown criterion. Int J Rock
[13] Chen WF. Limit analysis and soil plasticity. Amsterdam: Elsevier; Mech Mining Sci 2000;37:1013–8.
1975. [36] Serrano A, Olalla C. Ultimate bearing capacity at the tip of a pile
[14] Drescher A, Detournay C. Limit load in translational failure in rock-part 1: theory. Int J Rock Mech Mining Sci 2002(39):
mechanisms for associative and non-associative materials. Geotech- 833–46.
nique 1993;43(3):443–56. [37] Serrano A, Olalla C. Ultimate bearing capacity at the tip of a pile
[15] Michalowski RL, Shi L. Bearing capacity of footings over two-layer in rock-part 2: application. Int J Rock Mech Mining Sci
foundation soils. ASCE J Geotech Eng 1995;121(5):59–68. 2002:847–66.
[16] Michalowski RL. An estimate of the influence of soil weight on [38] Merifield RS, Lyamin AV, Sloan SW. Limit analysis solutions for the
bearing capacity using limit analysis. Soil Found 1997;37(4):421–8. bearing capacity of rock masses using the generalised Hoek–Brown
[17] Sarama SK, Iossifelis IS. Seismic bearing capacity factors of shallow criterion. Int J Rock Mech Mining Sci 2006;43:920–37.
strip footings. Geotechnique 1990;40(2):265–73. [39] Hoek E, Brown ET. The Hoek–Brown failure criterion update. In:
[18] Sloan SW. Lower bound limit analysis using finite elements and linear Rock engineering for underground excavation: proceedings of 15th
programming. Int J Numer Anal Meth Geomech 1988;12:61–77. Canadian symposium, Toronto; 1988. p. 31–8.
[19] Sloan SW. Upper bound limit analysis using finite elements and linear [40] Hoek E, Carranze-Torres C, Corkum B. Hoek–Brown failure
programming. Int J Numer Anal Meth Geomech 1989;13:263–82. criterion-2002 edition. In: Proceedings of the North American rock
[20] Sloan SW, Kleeman PW. Upper bound limit analysis using discon- mechanics society meeting, Toronto, July 2002. p. 267–73.
tinuous velocity fields. Comp Meth Appl Mech Eng 1995;127: [41] Hoek E. Estimating Mohr–Coulomb friction and cohesion values
293–314. from the Hoek–Brown failure criterion. Int J Rock Mech Mining Sci
[21] Ukritchon B, Whittle J, Sloan SW. Undrained limit analysis for 1990:227–9.
combined loading of strip footings on clay. ASCE J Geotech [42] Salençon J. Yield design: a general survey of the theory. CISM
Geoenviron Eng 1998;124(1):1–11. courses and lectures No. 332, Berlin: Springer; 1992.
[22] Wang YJ, Yin JH, Lee CF. The influence of a non-associated flow [43] Salençon J. An introduction to the yield theory and its applications to
rule on the calculation of the factor of safety of the soil slopes. Int J soil mechanics. Eur J Mech – A/Solid 1990;9(5):477–500.
Numer Anal Meth Geomech 2001;25:1351–9. [44] Garnier D, Maghous S. Design of high pressure Gas storage in rock
[23] Yin JH, Wang YJ, Selvadurai APS. Influence of non-association on caverns. In: Design and construction in mining, petroleum and civil
bearing capacity of a strip footing. ASCE J Geotech Geoenviron Eng engineering: proceeding of the 5th SARocks’98, São Paulo; 1998. p.
2001;127(11):985–9. 415–19.

You might also like