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

Computers and Geotechnics 37 (2010) 288298

Contents lists available at ScienceDirect

Computers and Geotechnics


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

Non-linear shear strength reduction technique in slope stability calculation


Wenxi Fu *, Yi Liao
State Key Laboratory of Hydraulic and Mountain River Engineering, School of Water Resource and Hydropower, Sichuan University, Chengdu 610065, China

a r t i c l e

i n f o

Article history:
Received 9 June 2009
Received in revised form 28 October 2009
Accepted 1 November 2009
Available online 2 December 2009
Keywords:
HoekBrown criterion
MohrCoulomb criterion
Non-linear shear strength reduction
Slope stability

a b s t r a c t
To exactly implement the non-linear HoekBrown shear strength reduction in slope stability calculations,
three aspects of the problem are considered. Firstly, the normal and shear stress relationship of the generalized HoekBrown criterion (2002 edition) is derived by introducing the MohrCoulomb instantaneous friction angle as a variable. Secondly, the instantaneous MohrCoulomb friction angle and
cohesive strength are derived for a micro-unit at a given stress state and they can be used to describe
the shear strength of the rock mass under different stress states. Finally, the implementation of the
non-linear shear strength reduction is described and a slope example is selected to verify our method.
This technique can be also applied to other non-linear failure criteria.
2009 Elsevier Ltd. All rights reserved.

1. Introduction
The shear strength reduction method is currently very popular
for evaluating slope stability. This method was rst proposed by
Zienkiewicz et al. in 1975 [1]. Its denition of the factor of safety
for a slope is often considered to be the ratio of the actual shear
strength to the lowest shear strength of a rock or soil material that
is required to maintain the slope in equilibrium [2]. Most of the
existing shear strength reduction techniques, discussed by Ugai
[3], Dawson et al. [4], Grifths and Lane [5], Luan et al. [6], Duncan
[7], Zheng et al. [8] and many others, are based on the linear Mohr
Coulomb criterion. However, little has been reported about the
non-linear shear strength reduction technique.
The existing shear strength technique, based on the MohrCoulomb criterion, is implemented by reducing the values of friction
angle and cohesive strength. For a MohrCoulomb material, its
shear strength is a linear function of stress level. When the
strength parameters of the original MohrCoulomb failure envelope have been reduced, the corresponding shear strength of all
points in the medium failing in shearing, e.g., in the case of a slope
stability problem, can be described with a single shear failure
envelope that still satisfy the linear MohrCoulomb criterion. But
for a non-linear failure criterion, e.g., the HoekBrown criterion
discussed in this paper, after the shear strength of each point corresponding to the original HoekBrown failure envelope has been
reduced, the stress state of all points in the medium failing in
shearing will not align to a single shear failure envelope that still
obeys the HoekBrown criterion anymore. In other words, because

* Corresponding author. Tel.: +86 28 80865292; fax: +86 28 85405604.


E-mail addresses: wxf_lee@scu.edu.cn, wxf_lee@sina.com (W. Fu).
0266-352X/$ - see front matter 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.compgeo.2009.11.002

the shear strength described with a non-linear failure criterion is


signicantly non-linear, the reduced shear strength cannot be directly characterized by a same-type non-linear failure criterion
with reduced mechanical parameter values with respect to the original failure criterion. Therefore, for a non-linear failure criterion,
e.g., the HoekBrown criterion, it is not possible to make direct
use of the existing shear strength techniques implemented by
reducing the strength parameters of a linear failure criterion, e.g.,
the MohrCoulomb criterion.
Currently, it is widely accepted that the shear strength of the
rock mass is a non-linear function of stress level. The non-linear
HoekBrown criterion accounts for this observation. This criterion
was originally proposed by Hoek and Brown in 1980 [9]. Its most
recent version is the generalized HoekBrown criterion (2002 edition) [10], which is the version discussed in this paper. The generalized HoekBrown criterion (2002 edition) is expressed by [10]:

r1 r3 rci mb

a

r3
s
rci

where r1 and r3 are the effective major and minor principal stresses of the rock mass at failure respectively (compressive stress is
taken to be positive), rci is the uniaxial compressive strength of
GSI100
GSI100
1
1
the
 GSI intact20  rock material, mb mi  e 2814D , s e 93D , a 2 6
 15
3
e
e
, mi is a material constant of the intact rock which
can be assessed from laboratory triaxial compression test results
and the method introduced by Hoek [11], GSI is the Geological
Strength Index, which can be estimated according to rock mass
structures and surface features of discontinuities in the rock mass,
and D is the disturbance factor reecting the degree of inuence
to which the rock mass has been subjected by blast damage and
stress relaxation due to excavation.

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

The generalized HoekBrown criterion has been widely accepted in the eld of geotechnics and is also often used for estimating the shear strength of rock mass, although it is an empirical
relationship that characterizes the stress conditions of the rock
mass at failure. In application to rock engineering projects, it
should be noted that the HoekBrown criterion is suited only to
homogeneous, isotropic massive rock with few discontinuities or
a heavily jointed rock mass [914].
As for the generalized HoekBrown failure criterion, Hoek et al.
[10] gave the exact equations for estimating the equivalent Mohr
Coulomb parameters over an articial interval [rt, r3max], where rt
is the biaxial tensile strength (rt = srci/mb, which can be derived
by setting r1 = r3 = rt in Eq. (1)), and r3max is the upper bound of
the minor principal stress. They also gave the corresponding equations for estimating r3max for simple slopes and circular tunnels
[10]. However, because the r1r3 relationship in Eq. (1) is non-linear, the articial interval [rt, r3max] does not represent the actual
failure stress conditions of rock mass in a number of situations:
for example, where a slope bears reinforcement loads at slope surface and/or distributed loads at slope shoulder. In particular, when
the equivalent MohrCoulomb criterion is selected, the shear
strength for the low and high stress levels will be overestimated,
and that for the intermediate stress level will be underestimated.
So in a rigorous mathematical sense, the equivalent MohrCoulomb criterion over the interval [rt, r3max] suggested by Hoek
et al. [10] is only an approximation to the HoekBrown criterion
over that interval.
As mentioned above, it is not possible to exactly implement the
non-linear HoekBrown shear strength reduction by directly
reducing the values of mechanical parameters, e.g., rci, mb, s, a,
etc. It is true, of course, when instantaneous MohrCoulomb
parameters are selected to describe any point on the HoekBrown
shear strength envelope, the widely used MohrCoulomb criterion
based shear strength reduction method can be also employed to
exactly implement the non-linear HoekBrown shear strength
reduction. However, it is not easy to describe the HoekBrown
shear failure envelope by direct use of the MohrCoulomb friction
angle and cohesive strength.
At present, the shear strength reduction technique, related to
the HoekBrown criterion, often make direct use of the equations
of equivalent MohrCoulomb friction angle and cohesive strength
derived by Hoek et al. [10]. In addition, the papers by Hammah
et al. [15] and Benz et al. [16] are among the few papers, known
to the present authors, that have discussed the non-linear shear
reduction technique for slope stability calculations by making direct use of the HoekBrown criterion in an elasto-plastic analysis.
The methods introduced by Hammah et al. [15] are: (1) to directly reduce the HoekBrown shear strength envelope by a factor
and generate a new envelope in the normal and shear stress plane;
(2) to obtain the basic parameters of the new HoekBrown model
by best-tting the lowered shear strength envelope; (3) to input
new parameters of the HoekBrown model and implement a new
elasto-plastic analysis. But the normal stress interval, selected for
tting the reduced shear strength envelope, is also determined
according to the minor principal stress interval [rt, r3max], suggested by Hoek et al. [10].
The methods introduced by Benz et al. [16] are: (1) to use the
Roscoe invariants p and q, which are functions of the rst and second invariants of the stress tensor respectively, to describe the
HoekBrown and MohrCoulomb criteria; (2) to connect the
HoekBrown criterion with the MohrCoulomb criterion by setting
the instantaneous slope of the HoekBrown criterion to be equal to
that of the MohrCoulomb criterion in the pq plane; (3) to introduce a material strength reduction factor g into the HoekBrown
yield function, and derive its relationship to the MohrCoulomb
shear strength reduction factor c; (4) to vary the MohrCoulomb

289

shear strength reduction factor c and use the reduced HoekBrown


criterion determined through the gc relationship in a new elastoplastic analysis.
However, the HoekBrown model is not embedded in most
commercially available software, so it is not easy to make direct
use of the methods proposed by Hammah et al. [15] and Benz et
al. [16]. In particular, the method proposed by Hammah et al.
[15] still involves the above-mentioned problem that the normal
stress interval need to be determined in terms of the minor
principal stress interval suggested by Hoek et al. [10]. As a
result, the reduced shear strength envelope obtained by best-tting is also an approximation to the true reduced envelope. So
the method adopted by Hammah et al. [15] will inevitably result
in errors between the reduced and tted HoekBrown shear
envelopes.
The non-linear shear strength reduction technique proposed in
this paper is also based on the generalized HoekBrown criterion
(2002 edition). It is known that, when the stress state of any point
in the rock mass is determined, its corresponding lowest shear
strength should be unique and can be exactly described with the
instantaneous MohrCoulomb friction angle and cohesive
strength. Therefore, when the instantaneous friction angle and
cohesive strength are obtained according to the HoekBrown criterion, the non-linear shear strength reduction can be also achieved
with the existing shear strength reduction techniques based on the
MohrCoulomb criterion.
This paper considers three aspects of the problem, summarized
as follows: (1) it introduces the instantaneous MohrCoulomb friction angle as a variable and derives the normal and shear stress
relationship of the HoekBrown criterion; (2) for a micro-unit at
a given stress condition, it derives the corresponding instantaneous
friction angle and cohesive strength, which can be used to describe
the lowest shear strength of the rock mass under different stress
states; (3) it describes in detail the implementation of the non-linear shear strength reduction and uses a slope example to verify the
proposed method.
Compared with the methods introduced by Hammah et al. [15]
and Benz et al. [16], the method proposed in this paper has the following merits: (1) it avoids errors between the reduced and tted
shear strength envelopes; (2) it still selects the widely used Mohr
Coulomb criterion based shear strength technique; (3) it can easily
be embedded in the users program code.

2. The normal and shear stress relationship of the HoekBrown


criterion
As Brown [13] has said Deriving exact analytical expressions
for the equivalent MohrCoulomb parameters at a given effective
normal stress has proven to be a challenging task. For the relationship between the normal and shear stresses of the generalized
HoekBrown criterion, Hoek et al. [10] have used Balmers equations [17]. Priest [18] has adopted the difference formula. Considering the minor principal stress as a variable, Carranza-Torres
[14] has derived the explicit expressions of the normal and shear
stresses, and in particular, he also discussed in detail the dimensionless expressions of the generalized HoekBrown criterion.
Considering the instantaneous MohrCoulomb friction angle as
a variable, the present authors have derived the expressions of the
normal stress, shear stress and their relationship for the Hoek
Brown criterion in another way. According to the derived expressions, the instantaneous MohrCoulomb friction angle and cohesive strength at a given effective normal stress can be obtained
iteratively. The detailed derivations will be described in Appendix
A. The general expressions related to our derivations are introduced briey below.

290

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

The normal and shear stresses of the failure plane of a microunit at critical failure in Fig. 1a are given by:

rn r1 2 r3  r1 2 r3 cos 2h
ss r1 2 r3 sin 2h

where rn and ss are the normal and shear stresses of the failure
plane, respectively, r1 and r3 are the effective major and minor
principal stresses at failure, respectively (compressive stress is taken to be positive), h is the angle between the failure plane and
the plane on which the minor principal stress acts.
The dip angle of the straight line, which is commonly tangential
to the Mohrs stress circle and HoekBrown shear envelope in
Fig. 1b, is ui, and its intercept at the shear stress axis is ci. The
instantaneous MohrCoulomb friction angle and cohesive strength
are ui and ci, respectively. According to Fig. 1b, Eq. (2) can be
rewritten as:

rn r1 2 r3  r1 2 r3 sin ui
ss r1 2 r3 cos ui

In the r1r3 plane, the MohrCoulomb criterion can be written


as:

r1

1 sin ui

r3 2c
1  sin ui

s
1 sin ui
1  sin ui

We differentiate Eq. (4) with respect to r3 and obtain:

dr1 1 sin ui

dr3 1  sin ui

Eq. (5) can be rewritten as:

sin ui




dr1
dr1
1
1
dr3
dr3

According to the trigonometric functions, we can also obtain:

0 s1


dr1 A
dr1
cos ui @2
1
dr3
dr3

Substituting Eqs. (6) and (7) into Eq. (3), we obtain:

8
r r1 r3 r3
>
>
< n 1ddrr13
q
r1 r3
dr1
>
>
: ss 1dr1 dr3

Eq. (8) can be also obtained in terms of the method introduced


by Balmer [17]. That is, we differentiate the Mohrs stress circle

2

2
equation rn  r1 2 r3 s2s r1 2 r3 with respect to r3.
For the HoekBrown criterion described with the r1r3 relationship of Eq. (1), its expressions of the normal stress, shear stress

Shear stress

HoekBrown envelope

i
s
i

ci

2 2

Normal stress

(a)

ss

9
10
11

It should be noted that Kumar [19] has derived similar equations for the normal and shear stresses.
As for the HoekBrown rock mass, its friction angle ui ranges
from 0 to 90, that is 0 < sin ui < 1. Under the condition 0 < sin
ui < 1, it is obvious that the normal stress rn in Eq. (9) and shear
stress ss in Eq. (10) are both monotonically decreasing functions
with respect to the friction angle ui. When the friction angle ui is
regarded as a variable, the HoekBrown shear failure envelope in
the normal and shear stress plane can be obtained through the following steps:
(1) It is assumed that the basic material parameters rci, mi, GSI
and D are known. And the material parameters mb, s and a
GSI100
can be obtained from the
 equations
 [10]: mb mi  e 2814D ,
GSI100
GSI
20


1
1
s e 93D , and a 2 6 e 15  e 3 . Consequently, for a
given value of friction angle ui, the corresponding normal
stress rn can be obtained by solving Eq. (9).
(2) The corresponding shear stress ss can be obtained by solving
either Eqs. (10) and (11).
(3) After a serial of pairs (rn, ss) are obtained by repeating steps
(1) and (2) above, the HoekBrown shear strength envelope
can be plotted.
The basic parameters are taken to be rci = 30.0 MPa, mi = 2.0,
GSI = 5.0 and D = 0.0, which are selected from the rst slope example in Hammah et al. [15]. The ui values are selected from 0 to 90
by Dui = 0.1. And the HoekBrown shear failure envelope, obtained by using the above method, is plotted in Fig. 2.

3. The instantaneous MohrCoulomb friction angle and


cohesive strength
Theoretically, when the stress state of any point in the rock
mass is known, the corresponding lowest shear strength should
be unique. We will now analyze the instantaneous MohrCoulomb

3.0

ss

rci

dr3

1 
a1

2 sin ui
sin ui
srci
1 
mb mb a1  sin ui
a
mb
a

a1
rci cos ui
2 sin ui

2
mb a1  sin ui
!a
rci cos ui mb rrcin s

sin ui
2
1
a

rn

(b)

Fig. 1. (a) Stress state of a micro-unit at failure and (b) relationships between the
Mohr stress circle and HoekBrown envelope in the normal and shear stress plane.

Shear stress s (MPa)

and their relationship have been derived from the above equations
and trigonometric functions. The step-by-step derivations have
been included in Appendix A. The normal stress, shear stress and
their relationship of the generalized HoekBrown criterion are given below:

ci = 30.0MPa
mi = 20.0
GSI = 5.0
D = 0.0

2.0

1.0

0.0
0.0

5.0

10.0
15.0
20.0
Normal stress n (MPa)

25.0

30.0

Fig. 2. The HoekBrown shear envelope in the normal and shear stress plane.

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

friction angle and cohesive strength, which can be used to describe


the shear strength of the rock mass at a specied stress state.
Any point in the rock mass is regarded as a micro-unit, as illustrated in Fig. 3a. The major and minor principal stresses of this micro-unit are denoted as re1 and re3, respectively, and the angle
between an arbitrary section and the plane on which the minor
principal stress acts is denoted as b. The thickness of the microunit is assumed to be one meter. The factor of safety Fs of an arbitrary section of the micro-unit may then be dened as the ratio of
the resisting shear force Ts to the mobilized shear force Tm, and is
given by:

Fs

Ts
ss  dA ss

T m sm  dA sm

12

where Fs is the factor of safety, Ts is the resisting shear force, Tm is


the mobilized shear force, ss is the shear strength, sm is the mobilized shear stress, and dA is the area of an arbitrary section of the
micro-unit.
In a physical sense, the above denition of the factor of safety is
rigorous, as both Ts and Tm are the force vectors acting on the same
section of the micro-unit. When the minimum of Fs is solved, the
corresponding shear strength can be regarded as a minimum. For
the micro-unit at a given stress state, however, the instantaneous
friction ui corresponding to the minimum of Fs needs to be solved
rst through a Newton iteration formula. The derivations of the
Newton iteration formula used for solving sin ui or ui are described
in Appendix B.
If sin ui is set to be a variable x, the Newton iteration formula to
solve x can be written as:

xk1 xk 

f xk
f 0 xk

k 0; 1; . . .

13

The numerator of the fraction expression in Eq. (13) is given by:

f x px qx r x  wx C 21  p2x

14

x

1

 x a1
where px qx a 1  C 3  C 2 , qx C 4 1x
, r 1  ax3 x2
re1 re3
2
2
2
2
2ax a; wx a  ax  a x  a ;C 1 2 , C 2 re1 2 re3 ; C 3 rmci s, and
b
 1
rci 2 a1
.
C4 m
m a
b

The denominator of the fraction expression in Eq. (13) is the


rst-order derivative of the numerator, and is given by:

f 0 x p0x qx rx 2wx px p0x px qx ux 


rx
 v x C 21  p2x
1  ax  x2
15

1ax2 axa

where p0x qx a2 axx2 , ux 31  ax2 2x 2a, and v x 2a


a  1x  a2 .
For the HoekBrown rock mass, because its friction angle ui is
larger than 0 and less than 90, x is larger than 0 and less than

m n
e3

Shear stress

e1
HoekBrown envelope
Fs =

s
m

n
Normal stress

(a)

(b)

Fig. 3. (a) Stress state of a micro-unit and (b) denition of the factor of safety for an
arbitrary section.

291

1. When x has been solved by the Newton iteration formula given


above, the instantaneous friction angle ui can be also obtained by
solving sin ui = x. The normal stress rn is obtained by substituting
x into Eq. (9), and the shear stress ss is obtained by substituting x
into either Eqs. (10) and (11).
The MohrCoulomb criterion in the normal and shear stress
plane is written as:

ss rn tan ui ci

16

Substituting the values of ui, rn, and ss into Eq. (16), the instantaneous cohesive strength ci is then obtained by:

ci rn tan ui  ss

17

4. Implementation of the non-linear shear strength reduction


It is assumed that the basic parameters of the HoekBrown rock
mass are known. We will now describe our calculation procedure
for the non-linear HoekBrown shear strength reduction. And the
corresponding ow chart is shown in Fig. 4.
(1) First, the grid elements should be generated for a computational model of the slope. After the basic parameters, such
as the deformation modulus E0, Poisson ratio l and bulk
density cb, are inputted, and the boundary and loading conditions are set correctly, the elastic stresses of elements can
be calculated by the nite element method. The plane
strain analyzes were performed and the four-noded quadrilateral elements were selected as in the example given in
Section 5.
(2) The elastic stress solution in step (1) is considered as the
starting step during an elasto-plastic iteration calculation.
In an elasto-plastic solution, each element is regarded as
a MohrCoulomb material with different friction angle
and cohesive strength, which can be determined from the
corresponding stress state by our methods introduced in
Section 3. Prior to an elasto-plastic analysis, the friction
angle and cohesive strength of each element are calculated
according to the elastic stress solution of an analytical
model. Here it is assumed that the total number of element
is n. When the major and minor principal stresses, calculated in step (1), of an arbitrary element j (j = 1, 2, . . ., n),
are substituted into Eq. (13), the instantaneous friction
angle uij of element j is obtained by solving the Newton
iterative formula given above. The corresponding normal
stress rnj of element j is calculated by Eq. (9), and the shear
stress ssj is then calculated by either Eqs. (10) and (11).
When uij, rnj and ssj are determined, the instantaneous
cohesive strength cij of element j is then calculated by Eq.
(17). For the HoekBrown rock mass, it should be noted
that, when the instantaneous friction angle of each element
is calculated by the iterative method given above, the
selected initial value ui0 of each element must satisfy the
condition 0 < ui0 < 90 or the condition 0 < sin ui0 < 1.
Because the Newton iteration formula is adopted, even if
a high precision is selected (convergence tolerance is set
to be 1.0E-8 in the slope example in Section 5), the instantaneous friction angle of each element can be solved in
only a few iteration steps.
(3) Once the instantaneous friction angles and cohesive
strengths of all the elements in an analytical model have
been obtained, the instantaneous MohrCoulomb strength
parameters of all elements are simultaneously reduced


tan u
s
c
by a factor ks, that is ksjs rnj ks ij kijs . The reduced
MohrCoulomb shear strength parameters of all elements

292

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

Loading and
boundary
conditions

Element grid
of slope

Input material
parameters

Computation mode
of slope stability

Input cklow, ckup


(cklow < ckup)

cks = (ckup+cklow)/2

Elastic stress
solution

ckup = cks
cklow = cklow

Solving for and c


of each element by
Eq.s 13, 910, 17

cklow = cks
ckup = ckup

Reducing , c and
tm of all elements
by cks

Yes

No

Elasto-plastic
analysis

fpls < 0.001 ?

nplas > 100 ?

No

Yes
ckdif = ckupcklow

No

ckdif < 0.01 ?


Yes
cks = (ckup+cklow)/2

end
Fig. 4. Flow chart of the non-linear shear strength reduction technique. (Note: (1) u, c, and rtm are friction angle, cohesive strength and uniaxial tensile strength of each
element; (2) cklow, ckup are the lower and upper bracket values given by user; (3) cks is the reduction factor; (4) nplas is the cumulative number of elasto-plastic analysis; (5)
fpls is the convergence tolerance of elasto-plastic analysis; (6) ckdif is the difference between the lower and upper bracket values.)

are then used in an elasto-plastic analysis. In order to satisfy


the non-linear HoekBrown shear strength envelope, during
the elasto-plastic analysis subject to the condition that the
reduced factor ks is maintained at a constant, the values of


tan u
s
c
uij and cij in the equation ksjs rnj ks ij kijs should be

adjusted according to the principal stresses of element j in


each iteration. In other words, for a given ks value, the uij
and cij values of each element, based on the elastic stress
solution in step (1), are employed in the rst step of
elasto-plastic analysis. The new uij and cij values of each

293

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

element, calculated according to the current step elastoplastic stress solution, are then used in the next step of elasto-plastic analysis. When the unbalance force convergence
value is less than a specied value or the iteration number
of elasto-plastic analysis is greater than a specied value, a
new given ks value is selected and the above calculation procedure including steps (1) and (2) is repeated. When the
analyzed slope model is in a critical failure state, the corresponding ks value is regarded as the slope safety factor.
The ks value is adjusted through the bracketing approach
similar to that proposed by Dawson et al. [4]. In the stability
analysis of the slope example in Section 5, the convergence
tolerance of unbalance force, maximum iteration number
of elasto-plastic analysis, and tolerance value of the difference between the upper and lower bracket ks values were
set to be 0.001, 100, and 0.01, respectively.
(4) The tensile strength rtm of the rock mass can be also contained
in the strength reduction given above, and it is suggested that
the same reduction factor be also selected for the tensile
strength as the MohrCoulomb shear strength given above.
The uniaxial tensile strength rtm of the rock mass can be
derived according to the HoekBrown criterion. Setting
r1 = 0 and r3 = rtm in Eq. (1), we can obtain Eq. (18). The uniaxial tensile strength rtm can be also solved iteratively.


a
r
rtm rci mb tm s 0

18

rci

The above calculation procedure has been embedded in the FEM


program written in Fortran language and commonly developed by
several researchers of State Key Laboratory of Hydraulic and Mountain River Engineering, Sichuan University. And the basic program
code, used in the stability analysis of the slope example in Section
5, is written in Appendix C.
5. A slope example
5.1. Basic conditions and methods
To verify our proposed non-linear shear strength reduction
method, the method presented by Benz et al. [16] and the simplied Bishop method [20], a well-known limit equilibrium method,
are also used. The uniaxial tensile strength is also considered in
the MohrCoulomb elasto-plastic analysis. Its reduction factor is
assumed to be the same as the shear reduction factor, and its estimation has been discussed in Section 4. The detailed calculation
procedure used by Benz et al. has been given in their paper [16].
During the calculation of the factor of safety of the slope by the
simplied Bishop method [20], according to the normal stress at

10m

(a)

(b)

32.0

30.0

34.0
28.0

10m

28.0

34.0
30.0

32.0

26.0

24.0

(c)

(d)

13.00
16.25

6.0e-4
9.0e-4
1.2e-3
1.5e-3
1.8e-3
2.1e-3
2.4e-3
2.7e-3

13.00
19.50

(e)
4.0e-4
8.0e-4
1.2e-3
1.6e-3
2.4e-3
2.8e-3
3.2e-3
3.4e-3

Fs = 1.145

Fs = 1.145

(f)
Fs = 1.152

Fig. 5. The slope example: (a) geometry sizes and element grids of the slope; (b) contours of the instantaneous friction angle () calculated according to the elastic stresses; (c)
contours of the instantaneous cohesive strength (kPa) calculated according to the elastic stresses; (d) contours of the plastic shear strain at failure and potential deformation
trend by the method proposed in this paper; (e) contours of the plastic shear strain at failure and potential deformation trend by the method of Benz et al.; (f) slip circle by the
simplied Bishop method.

294

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

the base of a slice, the corresponding instantaneous friction angle


is calculated iteratively by Eq. (9), and the convergence tolerance
is set to be 1.0E-8. The shear strength at the base of a slice is then
directly calculated by either Eqs. (10) and (11). The slice number,
convergence tolerance, and maximum iteration number are set to
be 100, 0.001 and 100, respectively. The slope stability analysis
program, based on the limit equilibrium theory, is developed on
the platform of Fortran PowerStation by the rst author.
The example analyzed in this paper was selected from the paper
by Hammah et al. [15]. Its basic material parameters: rci = 25.0 MPa, mi = 2.0, GSI = 5.0, D = 0.0, E0 = 5000.0 MPa, l = 0.3, and
cb = 25.0 kN/m3 were also selected. The geometry sizes and element grids of the slope is shown in Fig. 5a.
During the calculations of the slope failure, according to the
input parameters rci, mi, GSI and D, the HoekBrown material
parameters mb, s and a were calculated by the equations [10]
 GSI

GSI100
GSI100
20
mb mi  e 2814D , s e 93D and a 12 16 e 15  e 3 , respectively.
The slope model was considered to be homogeneous and isotropic.
The elasto-plastic analysis was assumed to be the plain strain
model. Only the gravitational stress eld was considered. Its horizontal displacements were xed for nodes along the left and right
boundaries, while both horizontal and vertical displacements were
xed along the bottom boundary. During the solution of the factor
of safety, the tolerance value of unbalance force convergence was
0.001, the iteration number of elasto-plastic analysis was 100,
and the difference between the upper and lower bracket ks values
was 0.01.
5.2. Results and discussion
The contours of the instantaneous friction angles and cohesive
strengths of the slope example analyzed in this paper, which were
obtained by our method and were calculated according to the elastic stresses of the elements, are shown in Fig. 5b and c, respectively. Its failure mechanisms and potential deformation trends,
predicted by our method and the method of Benz et al. [16], are
plotted in Fig. 5d and e, respectively. The slip circle, obtained by
the simplied Bishop method [20], is also shown in Fig. 5f.
The computational results in Fig. 5b show that the instantaneous friction angles near the slope surface are larger than those
of other parts of the slope, but the instantaneous cohesive
strengths are opposite each other in Fig. 5c. This occurs because
the stress levels near the slope surface are smaller than those of
other parts of the slope. The variations of the instantaneous friction
angles and cohesive strengths reect well the non-linear shear
strengths of the HoekBrown rock mass under different stress
states.
The geometrical shapes of the localized plastic shear strains and
values of the factor of safety, calculated by our method and the
method of Benz et al. [16], respectively, are identical. At the same
time they are also in reasonably good agreement with the circular
failure surface assumed in the simplied Bishop analysis. In addition, the factor of safety, calculated by Hamma et al. [15] with their
best-tting non-linear strength reduction method, was 1.15. Our
method also gives the results that are almost identical to those given by Hamma et al. [15].
6. Conclusions
For the generalized HoekBrown criterion (2002 edition), the
instantaneous MohrCoulomb friction angle and cohesive strength
are used to reect the non-linear shear strengths of the rock mass
under different stress states. Moreover, the widely used Mohr
Coulomb criterion based shear strength technique can still be used
to exactly implement the non-linear shear strength reduction. The

method proposed in this paper can be easily embedded in the


users program code, and it can be also applied to other non-linear
failure criteria.
Acknowledgements
The authors wish to thank Sichuan University and State Key
Laboratory of Hydraulic and Mountain River Engineering of China
for nancial support provided by the youth science fund. They also
thank International Science Editing for editing and polishing this
paper. Finally, they also thank the two peer reviewers for their
good suggestions.
Appendix A. The derivations of the normal and shear stress
relationship of the HoekBrown criterion
Using the general expressions in Section 2 and the trigonometric functions, and considering the instantaneous friction angle as a
variable, the authors have derived the normal stress, shear stress
and their relationship of the generalized HoekBrown criterion in
another way. The detailed derivations are described below.
According to Fig. 1b, we can derive tan a rnssr3 . Substituting
Eq. (8) into it, we obtain:

s
dr1
tan a
dr3

A:1

Also, tan 2a tanui p2  cot ui can be directly derived


according to Fig. 1b. Through transformation of the trigonometric
2 tan a
functions, tan 2a 1tan
2 a is obtained. We then obtain:

2 tan a
 cot ui
1  tan2 a

A:2

Solving tana in Eq. (A.2), we obtain:

tan a

1 sin ui
cos ui

A:3

Comparing Eq. (A.1) with Eq. (A.3), we obtain:

s
dr1 1 sin ui

dr3
cos ui

A:4

Again, the generalized HoekBrown criterion is written below:

r1 r3 rci mb

a

r3
s
rci

A:5

Eq. (A.5) can be rewritten as:

mb

r3
r1  r3
s
rci
rci

1a
A:6

We differentiate Eq. (A.6) with respect to r3 and obtain:


a1
dr1
r1  r3 a
1 mb a
dr3
rci

A:7

We can rewrite the expression of ss in Eq. (8) as:

r1  r3 ss 4
dr1

1
rci
rci
dr3

3
,s
dr1 5
dr3

A:8


.q
dr1
at the right-hand side of
According to Eq. (A.4), 1 ddrr13
dr3
Eq. (A.8) is expressed as a function related to ui:


,s
dr1
dr1
2

1
dr3
dr3 cos ui
Substituting Eq. (A.9) into Eq. (A.8), we obtain:

A:9

295

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

r1  r3
2ss

rci
rci cos ui

A:10

Substituting Eq. (A.10) into Eq. (A.7), we obtain:


a1
a
dr1
2ss
1 mb a
dr3
rci cos ui

A:11

Substituting Eq. (A.4) into Eq. (A.11), we obtain:


2

a1
a
1 sin ui
2ss
 1 mb a
cos ui
rci cos ui

A:12

Simplifying the left-hand side of Eq. (A.12), we obtain:


a1
a
2 sin ui
2ss
mb a
1  sin ui
rci cos ui

A:13

ss

2 sin ui
mb a1  sin ui

rn

a
a1

A:14

rci

mb

2 sin ui
mb a1  sin ui

0 ,s1
dr1 A
rn  @ss
dr3

1 
a1

sin ui
srci
1 
a
mb

A:22

Eq. (A.22) can be rewritten as:

!a
mb rrcin s

We can rewrite the relationship between rn and ss in Eq. (8) as:

r3

We thus obtain the expression of rn:

Eq. (A.13) can be rewritten as:

rci cos ui

a1
a1
rn
cos2 ui
2 sin ui
1
2 sin ui
s


mb mb a1  sin ui
mb
rci 21 sin ui mb a1  sin ui
a
1

a1

a1
1  sin ui
2 sin ui
1
2 sin ui
s


2
mb mb a1  sin ui
mb
mb a1  sin ui
1

a11

a1
a1
1  sin ui
2 sin ui
1
2 sin ui
s


2
mb mb a1  sin ui
mb
mb a1  sin ui
1

a1
1  sin ui
2 sin ui
2 sin ui

2
mb a1  sin ui mb a1  sin ui
1

a1
1
2 sin ui
s


mb mb a1  sin ui
mb
1
1

a1

a1
sin ui
2 sin ui
1
2 sin ui
s


mb mb a1  sin ui
mb
mb a mb a1  sin ui
1 

a1

1
2 sin ui
sin ui
s

A:21
1 
mb mb a1  sin ui
mb
a

a
a1
2 sin ui

mb a1  sin ui

sin ui
a

A:23

Substituting Eq. (A.23) into Eq. (A.14), we then obtain:

A:15

ss

rci cos ui mb rrcin s


sin ui
a

!a
A:24

Eq. (A.5) can be also rewritten as:

r1  r3
r3
mb
s
rci
rci

a
A:16

Substituting Eqs. (A.8) and (A.15) into Eq. (A.16), we then


obtain:

2


ss 4
dr1
1
rci
dr3

3 2 0
3a
,s1
,s
dr1 5 4mb @
dr1 A

s5
r s
dr3
rci n s
dr3
A:17

Substituting Eqs. (A.4) and (A.9) into Eq. (A.17), we obtain:



a
2ss
mb
s cos ui

rn  s
s
rci cos ui
rci
1 sin ui

A:18

2sin ui
mb a1  sin ui

a
a1

"

For the micro-unit at a given stress state, as illustrated in Fig. 3a,


the instantaneous friction ui corresponding to the minimum of Fs
needs to be solved iteratively. So the expression of the factor of
safety in Eq. (12) needs to be rst written as a function related to
the instantaneous friction ui. In this appendix, we introduce our
derivations in detail.
The stress circle equation in Fig. 3b is written as:

rn 

re1 re3 2
2

#
a
a1

rn
cos2 ui
2sin ui
mb

rci 21 sin ui mb a1  sin ui

mb

rn
cos2 ui
2 sin ui

rci 21 sin ui mb a1  sin ui

B:2

)a

Obviously, the minima of Fs and F 2s are obtained at the same


point. Therefore, the minimum of F 2s in Eq. (B.3) is solved rst.

A:19

F 2s 
re1 re3 2

s2s
2
 rn  re1 2 re3

B:3

Substituting Eqs. (9) and (10) into Eq. (B.3) and setting sin ui = x, we
can rewrite Eq. (B.3) as:

1
a1

B:1

ss

Eq. (A.19) can be rewritten as:

2 sin ui
mb a1  sin ui
(

r  r 2
e1
e3
2

Substituting the shear stress sm of Eq. (B.1) into Eq. (12), we can
rewrite Eq. (12) as:

s2m

F s q
re1 re3 2 
2
 rn  re1 2 re3
2

Substituting Eq. (A.14) into Eq. (A.18), we obtain:

Appendix B. The derivations of the equations to solve for sin ui


or ui

)
a
a1

r2ci 1x2

A:20

According to Eq. (A.20), the expression of rrcin is derived as


follows:

F 2s

r

e1  e3

h
rci
 m
m
b

4
2x
b a1x

2x
mb a1x

1
ia1


x
a

Eq. (B.4) can be rewritten as:

2a
ia1


1  smrci  re1 2 re3
b

B:4

296

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

r2ci

F 2s

r

e1  e3


r
2
ci

rci

mb

2a
a1

2a
 x a1
1  x2 1x

2
mb a

2
mb a

1
a1


x
1x

1 
a1
x

1 

Simplifying Eq. (B.9) and substituting it into Eq. (B.8), we obtain:


rci s
mb

re1 re3

B:5

2a
a1

Because 4 m2 a
in Eq. (B.5) is a constant, the solution of the
b
minimum of F 2s is transformed to that of the minimum of B in Eq.
(B.6).
2a
 x a1
1  x2 1x
B

 1

rci s re1 re3 2
re1 re3 2
1
a1 
rci
2
x a1 x


1


m a
1x
a
2
m
2
m
b

2a
 x a1
1  x2 1x
i2
1 

 x a1
x
C 21  C 4 1x
1  C3  C2
a

2
mb a

1
a1

, we

B:7

Differentiating Eq. (B.7) with respect to x, we then set the rstorder derivative of B to be zero:

dB
0
dx
dB
dx

in Eq. (B.8) is derived as follows:

dB

dx

B:10

f x px qx rx  wx C 21  p2x

B:11

x

1

 x a1
where px qx a 1  C 3  C 2 , qx C 4 1x
, r 1  ax3
x2 2ax a, and wx a2  ax2  a2 x  a2 .
Setting the left-hand side of Eq. (B.10) to be f(x), we obtain:

We can also write the rst-order derivative of f(x) as:

f x p0x qx r x 2wx px p0x px qx ux 


0

rx
1  ax  x2

 v x C 21  p2x

rci
Setting C 1 re1 2 re3 , C 2 re1 2 re3 ; C 3 rmci s and C 4 m
b
b
can rewrite Eq. (B.6) as:

B:6

px qx r x  wx C 21  p2x 0

B:8

2 axa
qx 1ax
,
a2 axx2

B:12

where

ux 31  ax 2x 2a, and v x 2a
a  1x  a2 .
There upon, we can solve x by the Newton iteration formula
below:
p0x

xk1 xk 

f xk
f 0 xk

k 0; 1; . . .

B:13

Appendix C. The program code written in fortran language


The program code, related to the non-linear HoekBrown shear
strength reduction, is written in Fortran language and is shown
below.

n  2a h
io
2a
 x a1
2a
x a11 1
x
1  x2 a1
1x
2
2x 1x
1x
1x
h  1
i2
h  1
i2


x a1 x
x a1 x
C 21  C 4 1x
C 21  C 4 1x
1  C3  C2
1  C3  C2
a
a
n h  1
ion  1
h
i
o
2a
1
 x a1
 x a1


C4
x a1 x
x a1
1
x
x
1  x2 1x
2 C 4 1x
1  C3  C2
 1 1x
1x
1 Ca4 1x
2
a1 1x
a
a


h  1
i2 2

x a1 x

1

C

C
C 21  C 4 1x
3
2
a

!
! The variables related to the non-linear H-B SSR method
! hbsigc: uniaxial compressive strength of intact rock;
! hbgsi: Geological Strength Index;
! hbmi: material constant of intact rock;
! hbd: disturbance factor;
! hbmb, hbs, and hba are material parameters in Eq. (1);
! hbst0: rock mass biaxial tensile strength;
! hbst1: rock mass uniaxial tensile strength, calculated by Eq. (18);
! nmax: total element number;
! cklow: lower bracket of cks;
! ckup: upper bracket of cks;
! cks=(ckup+cklow)/2.0;
! ckdif: difference between cklow and ckup;
! ni: sine value of initial instantaneous friction angle;
! fcon: convergence tolerance in calculating friction angle;
! fval: a variable in calculating friction angle;
! fpls: a variable in justifying convergence of elasto-plastic analysis;
! pi: ratio of the circumference of a circle to its diameter;
! nplas: cumulative number of elasto-plastic analysis.
!
! The arrays related to the non-linear H-B SSR method
! hbsign(nmax): normal stress of element;
! hbtao(nmax): shear stress of element;
! hbphi(nmax): instantaneous friction angle of element;

B:9

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

Appendix C (continued)
! hbcoh(nmax): instantaneous cohesive strength of element;
! hbst(nmax): uniaxial tensile strength of element;
! sig1(nmax): major principal stress of element;
! sig3(nmax): minor principal stress of element.
!
! Assume that all conditions related to analysis have been prepared.
! Only a H-B rock mass is considered in the program code.
! But two or more H-B rock masses can be easily expanded.
! The values of cklow and ckup used in the slope example.
cklow=1.00
ckup=1.30
ckdif=ckup-cklow
cks=(ckup+cklow)/2.0
loop_1: do while (ckdif .ge. 0.01)
! Set that hbcoh and hbst of each element is large enough to
! guarantee the implementation of elastic stress analysis.
! Here implement the elastic stress solution rst.
loop_2: do i=1,nmax,1
! Calculate sig1 and sig3 from element stress components.
! Compressive stress is taken to be positive for sig1 and sig3.
ni=0.05
fcon=1.0e-8
fval=1.0
s1=sig1(i)
s3=sig3(i)
loop_3: do while (fval .ge. fcon)
! Eq. (13) is used to calculate the sine value of hbphi(i) here.
! Absolute differece of ni between the current and last steps is
! transmitted to fval.
enddo loop_3
! Here sine value of hbphi(i) is transmitted to ni.
! Here hbphi(i) is calculated by hbphi(i)=(asin(ni))*180.0/pi.
! Here calculate hbsign(i) and hbtao(i) by Eqs. (9) and (10).
! Here calculate hbcoh(i) by Eq. (17).
! Here the iteratively calculated hbst1 is transmitted to hbst(i).
enddo loop_2
nplas=0
! Here implement the elasto-plastic iteration analysis.
loop_4: do while (nplas .le. 100)
! Here nplas is increased by nplas=nplas+1.
loop_5: do i=1,nmax,1
! Here reduce hbphi(i), hbcoh(i), and hbst(i) by cks.
enddo loop_5
! Here implement the elasto-plastic solution.
loop_6: do i=1,nmax,1
! Here implement the loop similar to loop_2 and loop_3.
! Then hbphi(i) and hbcoh(i) are calculated from current step.
enddo loop_6
! Here unbalance force percentage is transmitted to fpls.
if (fpls .le. 0.001) then
nplas=101
! Here nplas is greater than 100, so as to exit loop_4.
endif
enddo loop_4
! Adjust cks below through Dawsons bracketing approach.
if (fpls .le. 0.001) then
cklow=cks
ckup=ckup
else
ckup=cks
cklow=cklow
endif
cks=(cklow+ckup)/2.0
ckdif=ckup-cklow
enddo loop_1
!

297

298

W. Fu, Y. Liao / Computers and Geotechnics 37 (2010) 288298

References
[1] Zienkiewicz OC, Humpheson C, Lewis RW. Associated and nonassociated
viscoplasticity in soil mechanics. Gotechnique 1975;25(4):67189.
[2] Chowdhury RN. Slope analysis development in geotechnical engineering, vol.
22. AmsterdamOxford, New York: Elsevier Scientic Publishing Company;
1978.
[3] Ugai K. A method of calculation of total factor of safety of slopes by elastoplastic FEM. Soils and Foundations 1989;29(2):1905.
[4] Dawson EM, Roth WH, Drescher A. Slope stability analysis by strength
reduction. Gotechnique 1999;49(6):83540.
[5] Grifths DV, Lane PA. Slope stability analysis by nite elements. Gotechnique
1999;49(3):387403.
[6] Luan, MT, Wu, YJ, Nian, TK. An alternating criterion based on development of
plastic zone for evaluating slope stability by shear strength reduction FEM. In:
Proceedings of the SinaJapanese symposium on geotechnical engineering,
Beijing, China; 2003. p. 1818.
[7] Duncan JM. State of the art: limit equilibrium and nite-element analysis of
slopes. J Geotech Eng ASCE 1996;122(7):57796.
[8] Zheng H, Sun GH, Liu DF. A practical procedure for searching critical slip
surfaces of slopes based on the strength reduction technique. Comput Geotech
2009;36(1-2):15.
[9] Hoek E, Brown ET. Empirical strength criterion for rock masses. J Geotech
Geoenviron Eng Div ASCE 1980;106(GT9):101335.
[10] Hoek E, Carranza-Torres C, Corkum B. HoekBrown failure criterion 2002
edition. In: Proceedings of the 5th North American symposium NARMSTAC,
Toronto, <http://www.rockscience.com/roc/Hoek/references/Published-Papers.
htm>; 2002.

[11] Hoek E. Practical rock engineering. <http://www.rockscience.com/roc/Hoek/


PracticalRockEngineering.asp>; 2007.
[12] Hoek E, Marinos P. A brief history of the development of the HoekBrown
failure criterion. Soils Rocks 2007;30(2):8592.
[13] Brown ET. Estimating the mechanical properties of rock masses. In: Potvin Y,
Carter J, Dyskin A, Jeffrey R, editors. Proceedings of the 1st southern
hemisphere international rock mechanics symposium: SHIRMS 2008, Perth,
Western Australia, vol. 1; 2008. p. 321.
[14] Carranza-Torres C. Some comments on the application of the HoekBrown
failure criterion for intact rock and rock masses to the solution of tunnel and
slope problems. In: Barla G, Barla M, editors. Proceedings of the X conference
on rock and engineering mechanics: MIR 2004, Torino, Italy; 2004. p. 285326
[chapter 10].
[15] Hammah RE, Yacoub TE, Corkum B, Curran JH. The shear strength reduction
method for the generalized HoekBrown criterion. In Proceedings of the 40th
US symposium on rock mechanics, Alaska Rocks 2005, Anchorage, Alaska.
<http://www.rocscience.com/library/pdf/SSR-Method-for-GHB.pdf>.
[16] Benz T, Schwab R, Kauther RA, Vermeera PA. A HoekBrown criterion with
intrinsic material strength factorization. Int J Rock Mech Min Sci
2008;45(2):21022.
[17] Balmer G. A general analytical solution for Mohrs envelope. Am Soc Test Mater
1952;52:126071.
[18] Priest SD. Determination of shear strength and three-dimensional yield
strength for the HoekBrown criterion. Rock Mech Rock Eng
2005;38(4):299327.
[19] Kumar P. Shear failure envelope of HoekBrown criterion for rockmass. Tunn
Undergr Space Technol 1998;13(4):4538.
[20] Bishop AW. The use of the slip circle in the stability analysis of slopes.
Gotechnique 1955;5(1):717.

You might also like