You are on page 1of 6

7th South American Congress on Rock Mechanics, December 2 to 4, 2010.

Comparacin del ndice de fiabilidad de un talud de roca obtenidas


por equilibrio lmite y mtodos de modelacin numrica

Comparison of reliability index of a rock slope derived by limit
equilibrium and numerical modeling methods

Gheibie S., S.H.B Duzgun

Mining Engineering Department, Middle East Technical University 06531, Ankara, Turkey





Brief
This paper deals with probabilistic analysis of rock slopes through numerical approaches as well as comparing
the results with probabilistic limit sate formulation of the same slope. It is found that these two probabilistic
approaches yield different safety conditions where the probabilistic limit equilibrium approach provides safe
slope condition, whereas the probabilistic numerical analysis approach gives unstable slope condition. The
main reason for such a difference would be different parameters used in the two probabilistic stability analyses
Abstract
Rock slope stability analysis is considered one of the important issues in rock engineering. The engineers always
try to have the best design in terms of economical and safety aspects. Nowadays, it is tried to consider the
uncertainty as influential concept in rock structure designs. In most of the cases the limit equilibrium method is
used to model the rock slopes probabilistically to include the uncertainty effects. However, it was known that this
classical method does not consider some main facts in rock slope stability. Thus, it is offered to use advanced
numerical modeling methods in analyzing the rock slopes. In this paper, the probabilistic slope stability analysis
was combined by Distinct Element Numerical method to compare the probability of failure of a sliding rock
slope. Results of this study shows that there is a considerable difference between numerical and classical
approaches, the slope modeled by classical limit equilibrium method does not fail even in one pair of friction
angle and cohesion and the reliability index is 3.57 which shows the safety level of the slope. However, in
numerical method the reliability index is about 0.235 which shows that the slope is highly unsafe. This is
because, the classical method does not include the joints mechanical factors in detail, however, distinct element
method also considers Joint Shear Stiffness which is believed to have the most important effect on joints shear
behavior.

1. INTRODUCTION
Numerous numbers of lives and properties were
lost all over the world due to slope failures
although stability analyses are carried out. Most
of these analyses are based on the deterministic
methods which do not consider the effect of
uncertainty associated with parameters like
ground water pressure, rock mass, and
discontinuities shear strength. Such
uncertainties cause variation in failure
probability of slopes that have the same factor
of safety. As a result, the use of probabilistic
analysis techniques that take into account such
kind of uncertainties became more common in
recent years. According to Newtons basic
mechanics, it is clear that there are two load and
resistance concepts that will lead to stability or
instability of a block on a slope. If the
resistance is lower than the exerting load the
block will slide. Blocks mass, earthquake and
water pressure due to existing water in joints
are considered as load parameters and shear
strength (cohesion and friction angle) of the
interface between block and the slope are called
resisting parameters. To analyze the stability of
the rock blocks on a slope, it is needed to
determine both the load and resistance
parameters. However, most of the parameters
are varying due to time and position. For an
example, earthquake and rain fall are time
varying parameters or cohesion and friction
angle are varying spatially. Due to varying
nature of the rock mass, most of the time,
engineers collect data from different parts of a
rock mass. Usually, engineers use the mean
value of parameters or the most critical data to
be sure of their designs reliability. However,
these days it is clear that this type of design will
over/under estimate the most economical
design. Fortunately, in reliability engineering
there are methods that quantifies the
uncertainties and can determine the reliability
of a design in more rational ways. Therefore,
the application of reliability methods has
considerably increased in recent decades.
However, in rock mechanics, especially in rock
slope stability, due the complexity of media in
comparison with other engineering media, they
are rarely applied in practice. In recent decades,
considerable works have been done in
probabilistic modeling of rock slope; however,
they have just used classical mechanical
modeling such as Dzgn et al. (1995), Low
(1997), Park and West (2000), Dzgn and
Bhasin (2008), Rodriguez et al. (2006),
Jimenez-Rodriguez and Sitar (2007), Dianqing
et al. (2009) and Tatone and Grasselli (2010)
Some of the most widely used probabilistic
methods are Monte Carlo simulation technique,
Rosenblueth point estimate method, and
reliability index methods. Among these the
First Order Reliability Method (FORM)
proposed by Hasofer and Lind (1974) is most
widely used one as it considers the uncertainty
and variability of the parameters involved as
well as their correlation structure.
On the other hand, numerical methods are
considered the most efficient methods of
analyzing the stability of rock slope. This
method is extensively used in research and
industrial approaches. However, it has not used
well in probabilistic methods. It is estimated
that if the powerful numerical modeling method
can be applied probabilistically, the more
realistic designs can be performed. In this study
probabilistic analysis of rock slopes through
numerical approaches as well as comparing the
results with probabilistic limit sate formulation
of the same slope is presented.

2. FORM APPROACH
2.1 THE PERFORMANCE FUNCTION
The reliability assessment of an engineering
structure usually involves the consideration of
many variables. In particular, the supply and the
demand generally depend on several other
variables. In the FORM approach the supply
and demand concepts should be generalized.
The level of performance of a structure
obviously depends on the properties of supply
and demand formulation. For the purpose of
generalized formulation, it is necessary to
define a performance function or a state
function as shown below (Duzgun, 1994):
( )
( , , ,..., )
1 2 3
~
g x g x x x x
n
=
(1)
Where,
( )
( )
1 2
~
, ,...,
n
x x x x = is a vector of basic
variables. The function ( ) g x determines the
performance or the state of the structure.
Accordingly, the limiting performance is
defined as
~
( ) 0 g x = which is the limit-state of
the system. When
~
( ) g x >0, system is in safe
state and
~
( ) g x <0, the system is in failure state.
Geometrically, the limit state equation,
~
( ) g x =0,
forms an n-dimensional surface that may be
called failure surface.
The performance function is usually a non-
linear function. Unlike the linear case, there is
no unique distance from the failure surface to
the origin of the reduced variates. However,
Ang and Tang (1984) refer to Shinozuka
(1983), as the person who proposed the
point
( )
'* '*
1
,...,
n
x x as the most probable point on
the failure surface with the minimum distance
to the origin of the reduced variates. Hence, the
tangent plane to the failure surface at
( )
'* '*
1
,...,
n
x x can be used to the actual failure
surface to evaluate the reliability index.
The tangent at the
( )
'*
'* '*
, ...,
1
~
x x x
n
= is

( )
' *'
0
1 '
1
*
n
g
x x
i
x
i
i

=
| |
|
|
\
(2)
In which the partial derivatives
'
*
g
x
i

| |
|
|
\
are
evaluated at the
( )
'* '*
, ...,
1
x x
n
.
Thus the minimum distance from the tangent
plane to the origin of the reduced variates is the
reliability index . This minimum distance to
the tangent plane on the failure surface is
determined through the Lagrange multiplier.
The computation procedure is summarized in
(Ang and Tang 1984). The other important
factor in this procedure is the computation of
direction cosines
i
as
1
2
1
_
2
i
i
n
k
g
z
i
g
z

=
| |

\
=







| |
| |

|
|

\
\

By considering that
*
i i
x = .
Clearly, Figure (1) illustrate that the point with
the minimum distance to the origin of the
reduced variates is the most probable failure
point.

Figure 1 Tangent Plane to g(x) = 0 at x*
(After Ang and Tang, 1984)
3. PROBABILISTIC MODELING OF
SLOPE BY UDEC
For the cases where rock mass is discontinuous
media, Distinct Element Modeling is one of the
most suitable approch. The Universal Distinct
Element Code (UDEC) which is a two-
dimensional numerical program based on the
distinct element method for discontinuum
modeling was used in this study.
A rock slope shown in figure (2) was
constructed in UDEC media, the slope has a
joint inclined 30
0
toward the out of slope, and
that the angel of slope itself is about 54
0
. It was
assumed that the rock material is in elastic stage
and will not yield; this assumption was done to
assess the reliability of rock slope easily, since,
if rock material falls in plastic it is also needed
to assess its reliability which will be more
complicated. Moreover, usually rock slope
stability problems involve low normal loads,
where rock mass usually behaves elastically.
The constitutive law was Mohr-Coulomb in the
numerical analysis.
















The advantage of numerical modeling against
limit state modeling is that, it is possible to
formulate the process much realistic than limit
state modeling. For example, the Joint Shear
and Normal Stiffness are not considered in limit
state modeling, which were proved to have the
most controlling effect in joint behavior (Bandis
1990).
The first step in assessing the reliability is to
model the performance function, which was
assumed that when the sample reaches its peak
strength it is called failure. Thus it was tried to
consider the failure function a state in which the
rock joint had reached its peak value.
Barton (1972) described joint shear stiffness
(Ks) as the average gradient of the
shear stress-shear displacement curve for the
section of the curve below peak strength. Shear
stiffness can be estimated from direct shear
(3)
Figure 2 the geometry of rock slope
1
54
0
30 m
30
0
30 m
testing results, and its value depends on the size
of a sample tested and generally increases with
an increased in normal stress. Barton and
Choubey (1977) suggested the following
equation for the estimation of the peak shear
stiffness (MPa/m):
[ ]
R
n
n s
JCS
Log JRC
L
K

+
|

\
|
=
10
. tan . .
100
(4)
Where L is the joint length in meters,
n
is
normal stress acting on joint, JRC is joint
roughness coefficient, JCS is joint compressive
strength and
r
is the residual friction angle of a
joint. Barton and Bakhtar (1983) revealed that
the peak shear displacement is reached when
the joint has displaced 0.98% of its length. The
crack was assumed to have smooth surface, this
means the JRC value is zero, thus the above
equation can be rewritten as:
[ ]
R n s
L
K tan . .
100
= (5)
Joint length is 60 m,
r
is a random variable
and
n
was calculated by writing codes in
UDEC. The other joint parameter is Joint
Normal Stiffness. However, as the joint surface
is smooth, according to the applied joint normal
stiffness, it is estimated that there will not be
considerable movement in normal direction.
Then, it is considered almost fix. Thus, in pairs
of friction angle and cohesion the block
assumed to fail when it has displacement of
0.98% of its length. The random variables in
this study were joint friction angle and cohesion
of discontinuity, which were assumed to have
the normal distribution. Table (1) shows the
mean and standard deviation for both of
variables.

Table 1 the mean and STD for random variables
Standard Deviation Mean Min Max c.o.v
Cohesion (Pa) 3e4 12e4 2e4 18e4 0.25
Friction Angle 8.5
0
35
0
20
0
40
0
0.24

After a wide range of model running, figure
(3) shows the failure surface and also the safe
and non-safe regions.
Failure Surface
30
32
34
36
38
1.00E+04 6.00E+04 1.10E+05 1.60E+05 2.10E+05
Cohesion
F
r
i
c
t
i
o
n

A
n
g
l
e
Mean of cohesion
and friction angle
Safe Region
Unsafe Region

Figure 3 the failure surface

As can be seen in the figure (4) and figure (5)
the failure surfaces can be derived by curve
fitting through regression analysis. To have a
precise regression, the curve was divided into
two different domains then the relation of friction
angle and cohesion was derived for each domain.
The equations 6.1 and 6.2 show the failure
function derived by UDEC:
For 4 10 4 18 . 2 e x e (Figure 4)
( ) + =

95 . 36 0001 . 0 10 6
2 10
C C x g (6.1)
And for 4 18 4 10 e x e (Figure 5)
( ) + + =

59 . 32 10 5 10
6 2 11
C C x g (6.2)

Performance Function 1
y = 6E-10x
2
- 0.0001x + 36.946
R
2
= 0.9852
32
32.5
33
33.5
34
34.5
35
35.5
0.00E+00 5.00E+04 1.00E+05 1.50E+05 2.00E+05
Cohesion
F
r
i
c
t
i
o
n

a
n
g
l
e

Figure 4 the failure function for domain 1

Performance Function 2
y = -1E-11x
2
+ 5E-06x + 32.594
R
2
= 0.9841
32.95
33
33.05
33.1
33.15
33.2
0.00E+00 5.00E+04 1.00E+05 1.50E+05 2.00E+05
Cohesion
F
r
i
c
t
i
o
n

a
n
g
l
e

Figure 5 the failure function for domain 2

By applying the FORM to the derived equations
the design point or the closest point on the failure
surface from mean of random variable is
(C=12e4, = 33.18) and the reliability index is
0.235.
The variability of pairs of random variables on
the failure surface was calculated, where c.o.v.
for is 0.02 and for C is 0.51.
4. PROBABILISTIC MODELING OF THE
SLOPE BY LIMIT EQUILIBRIUM METHOD
Figure (2) shows the geometry of a slope which
was studied. The, first step in probabilistic limit
equilibrium analysis is to define the performance
function. The usual method to model the
performance of a rock slope is to use the limit
state equation. Most of the researches have used
this method to define the performance function.
It is generally defined as given in Eq. 4. (Duzgun,
1994):

( ) g x R D
f f
= (7)

Then, for a rock slope lying on a crack:

( ) ( )
p p
p p
Cos V Sin W
Sin V U Cos W cA x g



+ =
tan
(8)
Where,
= W Weight of the sliding block (ton/m)
= V Force due to water pressure in the
tension crack (ton/m)
= U Uplift force due to pressure on the
sliding surface (ton/m)
=
f
Dip of slope face (radians)
=
p
Dip of discontinuity plane (radians)
= Base area of the sliding block (m
2
/m)
= c Cohesion (ton/m
2
)
= Friction angle

The same random variables in table (1) were
also used in this formulation. For computation
of the RocPlane software and spreadsheet by
Fadlemula (2007) were utilized. The computed
probability of failure is zero, since there is no
pair of friction angle and cohesion in domain by
which the slope fails and the corresponding is
3.57.
5. CONCLUSIONS
The stability of a rock slope potential to slide
was studied in this article. Joints are playing
important roles in controlling the slide of a rock
blocks. Joints shear stiffness is an important
parameter which is not considered in limit
equilibrium method, thus a factor which affects
the shearing of a rock joint is not taken into
account in computations. Thus, it seems that
analyzing the rock slopes stability without
considering this parameter may produce errors
in designs.
In this article, a rock block sliding on a joint
plane was modeled by both limit equilibrium
and distinct element methods. Comparison of
the results between the two methods revealed
the importance of the way a rock slope is
modeled. For limit equilibrium method
RocPlane software and a spreadsheet provided
by Fadlemula (2007) were used, the analysis
show that for the given random variables in
table (1), the slope in figure (1) never will fail
and the probability of failure is zero and the
corresponding reliability index is 3.57.
However, results obtained form UDEC proves
that the rock slope has a critical circumstance
which may be considered as a failing structure.
The reliability index has been calculated as
0.235 which is showing a low safety.
This difference mainly comes from the fact
that the limit equilibrium modeling does not pay
attention to the facts of rock joints shear
behavior like joint shear stiffness. Also, there
are some differences between the calculated
normal stress by numerical modeling and limit
equilibrium method which affects the shearing
phenomena.
By considering the fact that numerical methods
are more reliable than limit equilibrium ones, it
can be concluded that for more reliable designs
the engineers should combine the probabilistic
method by numerical modeling.


REFERENCES
Ang, A.H.S., and Tang, W.H., (1984). "Probability
Concepts in Engineering Planning and Design" Vol.2.
Decision, risk, and reliability. John Wiley and Sons.
Bandis, S.C. (1990). "Mechanical properties of rock
joints" in Proceeding of the international symposium
on rock joints, Loen, Norway, 4-6 June: 125-140.
Barton, N: (1972). "A model study of rock-joint
deformation" Int. J Rock Mech. Min Sci. Geomech.
Abstr. (9): 579602.
Barton N.and V. Choubey, (1977). "The shear
strength of rock joints in theory and practice" Rock
Mech. (10): 154.
Duzgun HSB, Bhasin RK. (2008). "Probabilistic
stability evaluation of Oppstadhornet rock slope"
Norway, Rock Mech. Rock Eng doi: 10.1007/s00603-
008-0011-3.
Duzgun, H. S. B., Bozdag, T. and
Pasamehmetoglu, A.G., (1995). "A Reliability
Approach to Wedge Stability Analysis" Proc. of 8th
ISRM Congress: 389-392.
Duzgun, H. S. B., (1994). "Plane failure analysis of
rock slopes: A Reliability Approach" M.Sc. Thesis,
Middle East Technical University, Turkey.
Fadlelmula, M. M (2007). Probabilistic Modeling of
Failure in Rock Slopes M.Sc. Thesis, Middle East
Technical University, Turkey.
Hasofer, A.M. and Lind, N.C., (1974). "Exact and
invariant second-moment code format" J.Engrg.
Mech., ASCE, 100(1):111-121.
Jimenez-Rodriguez, R., Sitar, N., and Chacon, J.,
(2006). "System reliability approach to rock slope
stability" Int. J. Rock Mech. Min. Sci. 43 (6):847859.
Jimenez-Rodriguez R, Sitar N. (2007). "Rock wedge
stability analysis using system reliability methods"
Rock Mech. Rock Eng. 40(4):41927.
Low, B. K., (1997). "Reliability analysis of rock
wedges J. Geotech. Geoenviron. Engng. 123
(6):498505.
Park, H.J., West, T.R., (2001). "Development of a
probabilistic approach for rock wedge failure Eng.
Geol. (59): 233 251.
Shinozuka, M., (1983). "Basic Analysis of Structural
Safety" J. of Structural Division, ASCE (3): 109.

You might also like