Professional Documents
Culture Documents
Abstract
This paper uses numerical limit analysis to produce stability charts for rock slopes. These charts have been produced using the most
recent version of the Hoek–Brown failure criterion. The applicability of this criterion is suited to isotropic and homogeneous intact rock,
or heavily jointed rock masses. The rigorous limit analysis results were found to bracket the true slope stability number to within 79%
or better, and the difference in safety factor between bound solutions and limit equilibrium analyses using the same Hoek–Brown failure
criterion is less than 4%. The accuracy of using equivalent Mohr–Coulomb parameters to estimate the stability number has also been
investigated. For steep slopes, it was found that using equivalent parameters produces poor estimates of safety factors and predictions of
failure surface shapes. The reason for this lies in how these equivalent parameters are estimated, which is largely to do with estimating a
suitable minor principal stress range. In order to obtain better equivalent parameter solutions, this paper proposes new equations for
estimating the minor principal stress for steep and gentle slopes, which can be used to determine equivalent Mohr–Coulomb parameters.
Crown Copyright r 2007 Published by Elsevier Ltd. All rights reserved.
Keywords: Safety factor; Limit analysis; Rock; Slope stability; Failure criterion
1365-1609/$ - see front matter Crown Copyright r 2007 Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijrmms.2007.08.010
ARTICLE IN PRESS
690 A.J. Li et al. / International Journal of Rock Mechanics & Mining Sciences 45 (2008) 689–700
[12,13] and Krabbenhoft et al. [14] are ideally suited to Currently, practising engineers typically use a number of
modelling jointed or fissured materials because discontinu- stability charts when attempting to predict the stability of
ities exist inherently throughout the mesh. This feature was rock slopes: (1) Hoek–Bray [1] charts can be used for rock
recently exploited by Sutcliffe et al. [15] and Merifield and rockfill slopes; (2) Zanbak [31] proposed a set of
et al. [16] for predicting the bearing capacity of jointed rocks. stability charts for rock slopes susceptible to toppling;
The purpose of this paper is to take advantage of the (3) stability charts were presented by Siad [32] based on the
limit theorems ability to bracket the actual stability upper bound approach that can be used for rock slopes
number of rock slopes. Both the upper and lower bounds with earthquake effects. However, these three sets of design
are employed to provide this set of stability charts. These charts require conventional Mohr–Coulomb soil para-
solutions are obtained from numerical techniques devel- meters, cohesion (c) and friction angle (f), as input. From
oped by Lyamin and Sloan [12,13] and Krabbenhoft et al. a review of the literature, the authors are not aware of any
[14] where the well-known Hoek–Brown yield criterion has slope stability chart solutions based on the native form of
been incorporated into limit analysis as presented by the Hoek–Brown failure criterion that requires Hoek–
Merifield et al. [16]. Brown material parameters as input. This paper is
As a means of comparison, the limit equilibrium method concerned with providing a set of stability charts for rock
will then be used in conjunction with equivalent Mohr– slopes based on the Hoek–Brown failure criterion that can
Coulomb parameters for the rock and compared with the be used by practising engineers to rapidly assess the
solutions obtained from the numerical limit analysis preliminary stability of rock slopes.
approaches. This will allow the validity of using equivalent
Mohr–Coulomb parameters for rock slope calculations to
be investigated. 3. The generalised Hoek–Brown failure criterion
The stability of rock slopes has attracted the attention Practitioners are often required to predict the strength of
of researchers for decades. In order to deal with the large-scale rock masses for design. Fortunately, Hoek and
complications of rock slope failure mechanisms, Goodman Brown [6] proposed an empirical failure criterion which
and Kieffer [17] and Jaeger [18] outlined several simple developed through curve fitting of triaxial test data suited
methods and their limitations for estimating strength and for intact rock and jointed rock masses. The criterion is
stability of rock slopes. Due to the advancement of various based on a classification system called the Geological
computational techniques, our ability to more accurately Strength Index (GSI). The Hoek–Brown criterion is one of
evaluate rock slope stability and interpret the likely failure the few non-linear criteria widely accepted and used by
mechanisms has improved [19]. Buhan et al. [20] found that engineers to estimate the strength of a rock mass. There-
the final results of a stability analysis may be influenced by fore, it is appropriate to use this criterion when assessing
scale effect of rock masses. Sonmez et al. [21] utilised back the stability of istopic rock slopes in this study.
analysis of slope failures to obtain rock slope strength The GSI classification system is based upon the
parameters. In their study, the applicability of rock mass assumption that the rock mass contains sufficient number
classification, and a practical procedure of estimating the of ‘‘randomly’’ oriented discontinuities such that it behaves
mobilised shear strength based on the Hoek–Brown yield as an isotropic mass. In other words, the behaviour of the
criterion were explained. Previous investigations [22–25] of rock mass is independent of the direction of the applied
progressive failures and/or safety factor assessment of rock loads. Therefore, it is clear that the GSI system should not
slopes have used a range of numerical methods. These be applied to those rock masses in which there is a clearly
include the continuum methods (finite element method and defined dominant structural orientation that will lead to
the finite difference method), the discontinuum methods highly anisotropic mechanical behaviour. In addition, it is
(distinct element and discontinuous deformation analysis), also inappropriate to assign GSI values to excavated faces
and finite-/discrete-element codes. In addition, the prob- in strong hard rock with a few discontinuities spaced at
abilistic analytical method is employed in [26,27] to find distances of similar magnitude to the dimensions of slope
the rock slope potential failure key-group and estimate the under consideration. In such cases the stability of the slope
probability of failure. It should be acknowledged that the will be controlled by the three-dimensional geometry of the
slope stability probability classification proposed by Hack intersecting discontinuities and the free faces created by the
et al. [27] does not require cohesion and friction as input. excavation.
Yang et al. [28–30] adopted tangential strength parameters In line with the above discussion, it is important to
(c and f) from the Hoek–Brown failure criterion in an realise the stability charts presented in this paper will be
upper bound analysis to obtain the optimised height of a subject to the same limitations that underpin the Hoek–
slope. As far as the authors are aware, these studies [28–30] Brown yield criterion itself. An excellent overview of the
represent the only attempt at providing slope stability applicability and limitations of the GSI system can be
factors for estimating rock slope stability. found in [33].
ARTICLE IN PRESS
A.J. Li et al. / International Journal of Rock Mechanics & Mining Sciences 45 (2008) 689–700 691
An explanation for the applicability of Hoek–Brown GSI 100
s ¼ exp , (3)
criterion when applied to rock slopes is displayed in Fig. 1. 9 3D
After Hoek [34], for the same rock properties throughout the
slope, rock masses can be classified into three structural 1 1 GSI=15
a¼ þ e e20=3 . (4)
groups, namely Group I, Group II and Group III. Fig. 1 2 6
shows the transition from an isotropic intact rock (Group I), The GSI was introduced because Bieniawski’s rock mass
through a highly anisotropic rock mass (Group II), to a rating (RMR) system [35] and the Q-system [36] were
heavily jointed rock mass (Group III). In this paper the rock deemed to be unsuitable for poor rock masses. The GSI
slope has been assumed to be either (1) intact or; (2) heavily ranges from about 10, for extremely poor rock masses, to
jointed so that, on the scale of the problem, it can be 100 for intact rock. The parameter D is a factor that
regarded as an isotropic assembly of interlocking particles. In depends on the degree of disturbance. The suggested value
the case of intact rock (Group I), it should be noted that the of disturbance factor is D ¼ 0 for undisturbed in situ rock
failure mechanism of intact rock may be brittle rather than masses and D ¼ 1 for disturbed rock mass properties. The
plastic, so the theories of plasticity may not be appropriate. magnitude of the disturbance factor is affected by blast
damage and stress relief due to overburden removal. For
3.2. Numerical implementation the analyses presented here, a value of D ¼ 0 has been
adopted.
The upper-bound and lower-bound methods developed The uniaxial compressive strength is obtained by setting
in [12–14] can deal with a wide range of yield criteria; s3 ¼ 0 in Eq. (1), giving
however, on deviatoric planes the surfaces of those criteria
sc ¼ sci sa , (5)
must be convex and smooth. The Hoek–Brown yield
surface has apex and corner singularities in stress space, and the tensile strength is
and therefore numerical smoothing is required to avoid ssci
singularities. Details of the implementation of the Hoek– st ¼ . (6)
mb
Brown criterion into the numerical limit analysis formula-
tions can be found in [16] and will not be repeated here.
In this study, the latest version of Hoek–Brown failure
criterion [11] is employed. 3.3. Equivalent Mohr–Coulomb parameters
a
s0 Since most geotechnical engineering software is still
s01 ¼ s03 þ sci mb 3 þ s , (1)
sci written in terms of the Mohr–Coulomb failure criterion, it
where is necessary for practising engineers to determine equiva-
lent friction angles and cohesive strengths for each rock
GSI 100 mass and stress range. In the context of this paper, the
mb ¼ mi exp , (2)
28 14D solutions obtained by using equivalent Mohr–Coulomb
Jointed Rock
σci, GSI, mi, γ
Fig. 1. Applicability of the Hoek–Brown failure criterion for slope stability problems.
ARTICLE IN PRESS
692 A.J. Li et al. / International Journal of Rock Mechanics & Mining Sciences 45 (2008) 689–700
p overestimated p overestimated
H
β d
Toe
Hoek-Brown
Mohr-Coulomb (best fit)
Jointed Rock
σci, GSI, mi , γ
β = 45 β = 30
100 10
Average
Lower bound Average
Upper bound SLIDE-Hoek-Brown Model
SLIDE-Hoek-Brown Model
10 1
GSI =10
N=σci / γHF
N= σci / γHF
1 GSI=10 0.1
GSI =50
GSI=50
0.1 0.01 GSI =100
β H GSI=100 H
β
0.01 1E-3
5 10 15 20 25 30 35 5 10 15 20 25 30 35
mi mi
Fig. 4. Average finite element limit analysis solutions of stability numbers Fig. 6. Average finite element limit analysis solutions of stability numbers
(b ¼ 451). (b ¼ 301).
β = 15 β = 60
10 100
Average
Average SLIDE-Hoek-Brown Model
SLIDE-Hoek-Brown Model
1 10
GSI = 10
N= σci /γHF
N=σci /γHF
0.1 1
GSI =10
GSI = 50
GSI =50
0.01 0.1
GSI =100
H GSI =100
β H
β
1E-3 0.01
5 10 15 20 25 30 35 5 10 15 20 25 30 35
mi mi
Fig. 5. Average finite element limit analysis solutions of stability numbers Fig. 7. Average finite element limit analysis solutions of stability numbers
(b ¼ 151). (b ¼ 601).
ARTICLE IN PRESS
694 A.J. Li et al. / International Journal of Rock Mechanics & Mining Sciences 45 (2008) 689–700
β = 75 GSI =100
1
Average
SLIDE-Hoek-Brown Model 0.1
N=σci /γ HF
100
0.01
β = 75
β = 60
GSI= 10 β = 45
β = 30 H
β
β = 15
1E-3
10
N= σci /γ HF
5 10 15 20 25 30 35
mi
GSI=80
1
GSI= 50
1
N=σci / γHF
0.1
GSI=100 β = 75
β H β = 60
0.1 β = 45
0.01
β = 30 H
β
β = 15
5 10 15 20 25 30 35
5 10 15 20 25 30 35 mi
mi
GSI=60
Fig. 8. Average finite element limit analysis solutions of stability numbers
(b ¼ 751).
1
average value from the bound solutions could be adopted
N=σci / γHF
for simplicity. In fact, it was found that, for all the analyses
performed, the range between upper- and lower-bound
0.1
stability numbers was always less than 75%. The only β = 75
exception to this observation occurs for the cases of β = 60
β = 45
b ¼ 451 and low GSI values, where the range is around β = 30
β H
79%. Therefore, average values of the stability number N 0.01
β = 15
GSI = 40
10
1
N=σci /γ HF
0.1
β = 75
= 15°
β = 60
β = 45
β = 30 H
β
β = 15
0.01
5 10 15 20 25 30 35
mi
GSI =20
10
= 30°
N=σci /γ HF
0.1 β = 75
β = 60
β = 45
β = 30 H
β
β = 15
0.01
5 10 15 20 25 30 35
mi
GSI =10
100
= 45°
10
Fig. 11. Upper bound plastic zones with the different slope angles
N= σci / γHF
β = 75
β = 60
limit equilibrium results using equivalent Mohr–Coulomb
0.1 β = 45 parameters compare with the limit analysis results using
β = 30 H
β
β = 15 the Hoek–Brown criterion. In order to make this compar-
5 10 15 20 25 30 35 ison, the commercial limit equilibrium software SLIDE [39]
mi and Bishop’s simplified method [40] have been used. The
software SLIDE can perform a slope analysis using the
Fig. 10. Average finite element limit analysis solutions of stability
numbers GSI ¼ 40, 20 and 10). Mohr–Coulomb yield or the generalised Hoek–Brown
criterion. When the Mohr–Coulomb criterion is used, the
cohesion (c) and friction angle (f) are constant along any
sci=20 MPa, GSI=30, intact rock yield parameter mi=8, given slip surface and are independent of the normal stress
and unit weight of rock mass g=23 kN/m3. With this as expected. However, when the Hoek–Brown criterion is
information, the safety factor (F) of this rock slope can be selected, the software will calculate a set of instantaneous
obtained as follows. First, from the values of sci, g and H, equivalent Mohr–Coulomb parameters when analysing the
we can calculate a dimensionless parameter sci/gH=20000/ slope based on the normal stress at the base of each
(23 25)=34.8. In Fig. 7, N ¼ sci/gHFE4. The factor of individual slice. More details on how the parameters are
safety can then be calculated as F ¼ 34.8/4 ¼ 8.7. actually calculated can be found in [37]. Therefore, the
cohesion (c) and the friction angle (f) will vary along any
6. Results and discussion of limit equilibrium analyses given slip surface. By calculating equivalent Mohr–
Coulomb parameters in this way, a more accurate
In general, rock slope stability is more often analysed representation of the curved nature of the Hoek–Brown
using the limit equilibrium method and equivalent Mohr– criterion in t–sn space is obtained. Referring to Figs. 4–8,
Coulomb parameters as determined by Eqs. (7) and (8). the triangular points shown represent the stability numbers
With this being the case, an obvious question is how do the obtained from the limit equilibrium method based on the
ARTICLE IN PRESS
696 A.J. Li et al. / International Journal of Rock Mechanics & Mining Sciences 45 (2008) 689–700
Hoek–Brown strength parameters. It can be found that lower-bound limit analyses. In contrast the results of
these points are remarkably close to the average lines of the SLIDE analyses using the Mohr–Coulomb model do not
limit analysis solutions and most of them locate between compare favourably with the lower-bound results. From
the upper and lower bound solutions. Table 1, it can be found that using the Mohr–Coulomb
For the given materials and geometrical properties of the model may lead to significant overestimations of safety
slope, the finite element lower bound analysis will provide factors, particularly for steep slopes. The average difference
the optimum unit weight (g) such that collapse has just between F and F2 for b ¼ 601 and 751 was found to be
occurred (i.e., F=1). A critical non-dimensional parameter 16.8% and 34.3%, respectively. For all cases, the average
(sci/gH)crit can then be defined for the subsequent SLIDE overestimation is 12.8%. It should be stressed that, a high
analyses. In Table 1, the safety factor (F1) and (F2) are estimation of safety factor will induce a non-conservative
obtained using the Hoek–Brown criterion and the Mohr– design. It was found that using the Hoek–Brown model in
Coulomb criterion in SLIDE, respectively. Both these SLIDE will produce a failure mechanism in good agree-
analyses are based on equivalent Mohr–Coulomb para- ment with the upper-bound mechanism. The same could
meters with the only difference being how these parameters not be said when using the Mohr–Coulomb model. For
are calculated (as discussed above). b ¼ 301, both of the above models achieve similar failure
The comparisons of the safety factors F, F1 and F2 are surfaces, which agree well with the upper-bound plastic
shown in Table 1 where the largest difference between F zone. In almost all cases, a toe-failure mode was observed,
and F1 and F and F2 are about 4 and 64%, respectively. the only exception being the case of b ¼ 151 (base failure).
This shows that the results of SLIDE analyses using the In order to determine the source of overestimations in
Hoek–Brown model compare well with the results of the factors of safety (F2) for steep slopes, the stress conditions
Table 1
Comparisons of safety factors between the Hoek–Brown strength parameters and the equivalent Mohr–Coulomb parameters
Nonlinear Nonlinear Eqs. (7), (8) and (9) Eqs. (7), (8) and (12) Eqs. (7), (8) and (13)
Hoek–Brown Hoek–Brown Linear Linear Linear
Mohr–Coulomb Mohr–Coulomb Mohr–Coulomb
Table 1 (continued )
Nonlinear Nonlinear Eqs. (7), (8) and (9) Eqs. (7), (8) and (12) Eqs. (7), (8) and (13)
Hoek–Brown Hoek–Brown Linear Linear Linear
Mohr–Coulomb Mohr–Coulomb Mohr–Coulomb
Table 1 (continued )
Nonlinear Nonlinear Eqs. (7), (8) and (9) Eqs. (7), (8) and (12) Eqs. (7), (8) and (13)
Hoek–Brown Hoek–Brown Linear Linear Linear
Mohr–Coulomb Mohr–Coulomb Mohr–Coulomb
10
'
to the limit analysis solutions that are based on the native [21] Sonmez H, Ulusay R, Gokceoglu C. A practical procedure for the
form of the Hoek–Brown criterion. back analysis of slope failures in closely jointed rock masses. Int J
Rock Mech Min Sci 1998;35(2):219–33.
[22] Hoek E, Read J, Karzulovic A, Chen ZY. Rock slopes in civil and
References mining engineering. In: Proceedings of the International Conference
on Geotech Geol Eng, 2000, Melbourne.
[1] Hoek E, Bray JW. Rock slope engineering. 3rd ed. London: Institute [23] Wang C, Tannant DD, Lilly PA. Numerical analysis of stability of
of Mining and Metallurgy; 1981. heavily jointed rock slopes using PFC2D. Int J Rock Mech Min Sci
[2] Goodman RE. Introduction to rock mechanics. 2nd ed. New York: 2003;40:415–24.
Wiley; 1989. [24] Eberhardt E, Stead D, Coggan JS. Numerical analysis of initiation
[3] Wyllie DC, Mah CW. Rock slope engineering. 4th ed. London: Spon and progressive failure in natural rock slopes—the 1991 Randa
Press; 2004. rockslide. Int J Rock Mech Min Sci 2004;41:69–87.
[4] Taylor DW. Stability of earth slopes. J. Boston Soc Civ Eng 1937;24: [25] Stead D, Eberhardt E, Coggan JS. Developments in the character-
197–246. ization of complex rock slope deformation and failure using
[5] Gens A, Hutchinson JN, Cavounidis S. Three-dimensional analysis numerical modelling techniques. Eng Geol 2006;83:217–35.
of slides in cohesive soils. Geotechnique 1988;38(1):1–23. [26] Yarahmadi Bafghi AR, Verdel T. Sarma-based key-group method for
[6] Hoek E, Brown ET. Empirical strength criterion for rock masses. J rock slope reliability analyses. Int J Numer Anal Meth Geomech
Geotech Eng Div ASCE 1980;106(9):1013–35. 2005;29:1019–43.
[7] Yu MH, Zan YW, Zhao J, Yoshimine M. A unified strength criterion [27] Hack R, Price D, Rengers N. A new approach to rock slope
for rock material. Int J Rock Mech Min Sci 2002;39:975–89. stability—a probability classification (SSPC). Bull Eng Geol Environ
[8] Grasselli G, Egger P. Constitutive law for the shear strength of rock 2003;62:167–84 and erratum, p. 185–185.
joints based on three-dimensional surface parameters. Int J Rock [28] Yang XL, Li L, Yin JH. Stability analysis of rock slopes with a
Mech Min Sci 2003;40:25–40. modified Hoek–Brown failure criterion. Int J Numer Anal Meth
[9] Sheorey PR. Empirical rock failure criteria. Rotterdam: Balkema; Geomech 2004;28:181–90.
1997. [29] Yang XL, Li L, Yin JH. Seismic and static stability analysis for
[10] Yudhbir, Lemanza W, Prinzl F. An empirical failure criterion for rock slopes by a kinematical approach. Geotechnique 2004;54(8):
rock masses. In: Proceedings of the fifth Congress ISRM, 1983. 543–9.
[11] Hoek E, Carranza-Torres C, Corkum B. Hoek-Brown Failure [30] Yang XL, Zou JF. Stability factors for rock slopes subjected to pore
criterion-2002 edition. In: Proceedings of the North American Rock water pressure based on the Hoek–Brown failure criterion. Int J Rock
Mechanics Symposium Toronto, 2002. Mech Min Sci 2006;43:1146–52.
[12] Lyamin AV, Sloan SW. Upper bound limit analysis using linear finite [31] Zanbak C. Design charts for rock slopes susceptible to toppling.
elements and non-linear programming. Int J Numer Anal Meth J Geotech Eng Div ASCE 1983;190(8):1039–62.
Geomech 2002;26:181–216. [32] Siad L. Seismic stability analysis of fractured rock slopes by yield
[13] Lyamin AV, Sloan SW. Lower bound limit analysis using non-linear design theory. Soil Dynam Earthquake Eng 2003;23:203–12.
programming. Int J Numer Meth Eng 2002;55:573–611. [33] Marinos V, Marinos P, Hoek E. The geological strength index:
[14] Krabbenhoft K, Lyamin AV, Hjiaj M, Sloan SW. A new applications and limitations. Bull Eng Geol Environ 2005;64:
discontinuous upper bound limit analysis formulation. Int J Numer 55–65.
Meth Eng 2005;63:1069–88. [34] Hoek E. Strength of jointed rock masses. Geotechnique 1983;33(3):
[15] Sutcliffe DJ, Yu HS, Sloan SW. Lower bound solutions for bearing 187–223.
capacity of jointed rock. Comput Geotech 2004;31:23–36. [35] Bieniaski ZT. Rock mass classification in rock engineering. In
[16] Merifield RS, Lyamin AV, Sloan SW. Limit analysis solutions for the Exploration for rock engineering, Proceedings of the rock mechanics
bearing capacity of rock masses using the generalised Hoek-Brown symposium. Balkema: Cape Town; 1976.
criterion. Int J Rock Mech Min Sci 2006;43:920–37. [36] Barton N. Some new Q-value correlations to assist in site
[17] Goodman RE, Kieffer DS. Behavior of rock in slope. J Geotech characterisation and tunnel design. Int J Rock Mech Min Sci 2002;
Geoenviron Eng 2000;126(8):675–84. 39(2):185–216.
[18] Jaeger JC. Friction of rocks and stability of rock slopes. Geotechni- [37] Hoek E, Rock Engineering. 2000 /http://www.rocscience.com/hoek/
que 1971;21(2):97–134. references/Published-Papers.htmS.
[19] Stead D, Coggan JS, Eberhardt E. Realistic simulation of rock slope [38] Hoek E, Brown ET. Practical estimates of rock mass strength. Int J
failure mechanisms: the need to incoporate principles of failure Rock Mech Min Sci 1997;34(8):1165–86.
mechanisms. Int J Rock Mech Min Sci 2004;41(32B):17–22. [39] Rocscience, 2D limit equilibrium analysis software, Slide 5.0.
[20] Buhan PD, Freard J, Garnier D, Maghous S. Failure properties of /www.rocscience.comS.
fractured rock masses as anisotropic homogenized media. J Eng [40] Bishop AW. The use of slip circle in stability analysis of slopes.
Mech 2002;128(8):869–75. Geotechnique 1955;5(1):7–17.