You are on page 1of 10

Int. J. Rock ,\f",h. Min. Sci. & G"'lnrech. Abslr. Vol. 20. No.5. pp. 227-236.

1983
Printed in Great Brilain. All rights reserved
0148-9062/83S3.00 +0.00
Copyright ':D 1983 Pergamon Press Lid
The Effect of Discontinuity Persistence
on Rock Slope Stability
H. H. EINSTEIN*
D. VENEZIANO*
G. B. BAECHERt
K. J. O'REILLY;
l
f
Discominuity persistence has a major effect on rock mass resistance (strength)
but, as direct mapping ofdiscominuities imemal to a rock mass is not possible,
persistence is a difficult parameter to measure. As a result, the conservative
approach of assuming full persistence is often taken. /n this paper a method
is developed for relating rock mass stability and hence persistence to the
geometry and spatia/variability of discontinuities. The method is applied to
slope stability ('alculations ill which the probability offailure is related to
discominuity data, as obtained in joim surveys. The complete method makes
use of dynamic programming and simulation, but a closed form expression
satisfactory for most purposes is also presemed.
0
01
- AREA OF INDIVIDUAL JOINT
AD - AREA OF JOWT PLANE
defined as the fraction of area that is actually discon-
tinuous. One can therefore express K as the limit
in which L
s
is the length of a straight line segment Sand
t
s
is the length of the i
lh
joint segment in S; or for a
joint (Fig. 2),
(I)
(2a)
I:aD
j
K= lim -'-,
..1
0
-,,,, AD
in which D is a region of the plane with area AD and aD;
is the area of the i
lh
joint in D (Fig. I). The summation
in equation (l) is over all joints in D. Equivalently, joint
persistence can be expressed as a limit length ratio along
a given line on a joint plane. In this case,
I: Is;
K= lim -'-
l.s,x L.
TRADITIONAL DEFINITION OF JOINT
PERSISTENCE AND ASSOCIATED
PROBLEMS
With reference to a joint plane (a plane through the
rock mass that contains a patchwork of discontinuities
and intact-rock regions), joint persistence K is usually
Discontinuity (hereafter referred to as joint) persistence
is among the parameters most significantly affecting rock
mass strength, and is a problematic one. While relatively
small bridges of intact rock between otherwise con-
tinouous joints substantially increase strength, the map-
ping of each joint is impossible on a practical basis. An
attractive alternative to separately considering specific
joints is offered by statistical techniques for sampling
and describing the geometry of discontinuities.
These techniques are at an early stage of development,
but offer a significant advancement of the state of the art:
they characterize persistence as a random variable and,
in conjunction with a mechanical model of rock failure,
produce the probability distribution of rock mass
strength.
A method is developed here for rock-slope reliability
analysis based on a probabilistic characterization of the
joint system. In doing so, it is found convenient to
modify the traditional definition of rock persistence by
accounting for the uncertain failure path. Preceding
work is briefly described, and is followed by a descrip-
tion of the present method.
INTRODUCTION
Professor, tAssociate Professor and :j:Formerly Research Assislanl
at: Department of Civil Engineering, Massachusseus Institute of
Technology. Cambridge. MA 02139. U.S.A.
K lim fODI <V Jointed Area
Ao+GO AD - Total Area
Fig. 1. Joint persistence.
227
228 EINSTEIN et at.: DISCONTINUITY PERSISTENCE AND SLOPE STABILITY
/
./ '--rock bridQ8 (RBR)
(JL)
K. IJL
IJL+IRBR
Fig. 2. Joint persistence as length ratio.
\
IN PLANE FAILURE
Fig. 4. En-echelon and in-plane failures.
Fig. J. Jennings' relations.
,3
1
.
in which N is the number of critical paths (failing and
not failing) and N
f
is the number of critical paths for
where tan cPa and c
a
, so-called Jennings' equivalent
friction and cohesion parameters, are given by
C
a
= (1- K)c
r
+KCj,
tan cPa = (I - K) tan cPn + K tan cPj' (7)
The use of equations (6) and (7) for shear resistance of
jointed rock masses has several shortcomings:
(I) Failure surfaces are restricted to joint planes. En-
echelon failures (Fig. 4), common in the field, are
neglected.
(2) Shearfailure does not typically occurfor the usually
low values of CT
a
For example, for slopes of 30 m (100 ft)
height, CT
a
is about 0.7 MPa (100 psi), whereas c
r
is
typically ten to hundred of MPa. If CT. is negligible, then
the major principal stress must exceed C
r
for shear failure
to occur (Fig. 5). This is unrealistic, as Lajtai [6] and
Stimpson [9] have pointed out. Also, peak shear re-
sistance in the intact rock and on the joint probably are
not mobilized simultaneously.
(3) Small variations of persistence produce large vari-
ations of resistance. Therefore, even modest uncertainty
about persistence forces the designer to the conservative
assumption of 100% persistence.
To surmount these difficulties a new definition of persis-
tence is required.
NEW CONCEPT OF PERSISTENCE
Any planar or non-planar surface (or "path") through
intact rock and joints in a rock mass (Fig. 4) constitutes
a potential failure surface (failure path) with associated
driving force L and resisting force R. For a given
configuration of the joint system and a given set of
strength parameters, there is a path of minimum safety
or "critical path" (Fig. 6). The critical path for a
particular joint configuration is that combination of
joint and intact-rock portions having the minimum
safety margin 8M = R - L. If the 8M for this path is
negative, the rock mass fails; otherwise it resists. Thus,
a critical path mayor may not be a failure path. The
probability of failure P
f
of a randomly-jointed rock mass
can be expressed as the limit of relative frequency of
failure across the spectrum of joint configurations,
n I' N
f
tm -, (8)
N.7. N
(3)
(4)
(6)
(2b)
E aj
I = lim _i_
V
l;JL
K = l;JL +ERBR'
Another useful index of rock mass discontinuity is joint
intensity I, defined as the area of joints per unit rock
volume,
R
r
= (CT
a
tan cPr + cr)A,
in the case of intact rock and
in which a
j
is the area of the i
1h
joint in a 3-D region of
volume V.
Joint persistence can be used to estimate the strength
of a rock mass against sliding along a given plane: if the
plane of sliding has area A, then shearing resistance can
be adequately expressed as
R
j
= (CT
a
tan cPj +cj)A, (5)
in the case ofcompletely jointed region. cPr and cPj are the
friction angles of intact rock and the joint respectively,
C
r
and Cj, the intact rock- and joint-cohesion. In both
cases, (1a is the average normal stress across the region
of sliding. If the sliding region is partitioned into an
intact-rock portion of area A
r
and a jointed portion of
area A
j
= A - A
r
(Fig. 3), then following Jennings [5] one
can evaluate the shear resistance to sliding, R, as a
weighted combination of R
r
and R
j
according to the
expression
r
..
EINSTEIN el oJ.: DISCONTINUITY PERSISTENCE AND SLOPE STABILITY
CT
r
a
Ta
-

Il-Vn
T
a
r::
Fig. 5. Mohr's circle at failure predicted by Jennings' relations at low stress levels.
229
which SM < 0 (the number of failure paths). Equation
(8) suggests a way to estimate P
r
: using statistical
information on joint length and spacing distributions
one can simulate a number of networks of joints such as
that in Fig. 7 and then determine the SM for all possible
paths in each network or configuration. The critical path
for a configuration of the type in Fig. 7 is obtained by
identifying the path of minimum SM among all or
among a reasonable number of in-plane and en-echelon
paths. In some configurations the critical path will be a
failure path (SM I) while in others it will not be
Fig. 6. Critical paths for different joint configurations.
(SM> I). Simulating many configurations (realizations)
represents various ways that joint populations with the
same spacing and length characterization may manifest.
and at the same time produces the parametrs Nand N
f
for use in equation (8). In anyone realization the SM for
the critical path can be used to calculate an apparent
persistence, and thus one obtains a relation between
resistance and apparent persistence for a rock mass
characterized by joint length and spacing distributions.
To date, these principles have been applied to 2-D
slope stability models in which the pattern of jointing,
and strength coefficients, are assumed similar for all
cross-sections: 3-D extensions for slope and tunnel appli-
cations have been limited [7].
Earlier probabilistic 2-D slope models
Call and Nicholas [I] and Glynn [3] have developed
methods of 2-D slope stability analysis that use statisti-
cal information on jointing and that allow for both
in-plane and en-echelon failures. The model of Call and
Nicholas considers two random joint sets. Given distri-
butions of joint length and separation and of spacing
Fig. 7. Join, configuration and its critical path in a portion of the rock
mass.
230 EINSTEIN ('/ Ill.: DISCONTINUITY PERSISTENCE AND SLOPE STABILITY
Step path" a
angle
% Tensile Failure" 100
Fig. 9. Rock slope with single set of parallel joints.
RANDOM VARIABLES
where
Fig. 8. Can and Nichol..s [1] model-gener..' Icatures.
J
(10)
Critical pathl
I
N
r
P
r
=--'
, N,'
Fig. 10. Elevation intervals H,.
Nell' probabilistic persistence model-SLOPESIM
SLOPESIM is a computer code for the analysis of
rock slopes that contain one set of parallel joints (Fig.
9). SLOPESIM use Monte Carlo simulation to generate
joint patterns ("realizations") in accordance with given
probability distributions of joint length and plane spac-
ing (corresponding to the statistical information on joint
length and spacing taken in surveys). For each exit point
the algorithm finds the path of minimum SM. Critical
paths may be planar or may involve transitions to
overlying joint planes. The distribution of SM and in
particular the probability of unstable paths (SM < 0)
depend on the elevation of the exit joint. SLOPESIM
estimates the distribution and probability by grouping
exit points according to elevation intervals (Fig. 10). For
example, the probability of unstable paths for the i
lll
elevation interval is calculated as:
Apparent persistence depends explicitly on strength
along the path and implicitly on configuration. Repeated
simulation of the joint pattern yields in a distribution of
K
a
to be used in a probabilistic version of Jennings'
approach. The result is probability of slope failure.
Both previous models consider stochastic jointing and
failures occuring in plane or an echelon. Results are
expressed as probability distributions. The limitations of
these models are that they (I) apply to a specific rock
mass geometry and (2) use unsatisfactory, mechanical
models (i.e. shearing is ignored or unrealistically treated)
[7.2).
(9)
Master Joint Set
Cross Joint Set
Dip }
Length
Spacing
Overlap
bclween joint planes for each set. the procedure simu-
lates critical "step-paths" as shown in
Fig. 8. Specifically. for each simulated realization of the
joint network. "exit points" arc identified (i.e. inter-
sections of the shallow joint planes with the slope face).
and the critical step-path through each exit point is
found. Critical paths are obtained by alternately follow-
ing jointed segments. which fail in shear, and tensile
fractures through the intact rock between joint planes.
Shear failure of intact rock bridges between joints is
considered improbable except for extremely short
bridges 6 cm). Faced with a choice among paths
through intact rock. the model chooses that with lowest
angle. Through simulation of the jointing pattern, the
model calculates the distribution of average step-path
angle and fraction of path containing jointed segments.
the latter taken as a measure of persistence. These
distributions are conditioned on slope geometry.
strength parameters, and spacing.
Glynn's [2,3] JOINTSIM model generates joint net-
works with exponential distributions of spacing and
length. The strength of intact rock bridges. in plane or
en echelon. is determined by superimposing a negative
increment of horizontal stresses Aall and a positive
increment of shear stress AT on the initial state of stress.
such that failure is caused. Using this calculated strength
of rock bridges and the resistance of the jointed seg-
ments. the critical path for a given joint pattern is
calculated. "Apparent persistence" K
a
is defined as the
value of K along a joint plane that would have the same
resistance as the failure path,
R - R
Ka=R'_R'
, J
R =resistance of failure path.
R
J
= resistance of joint plane if 100% persistence.
R, =resistance along joint path, if intact rock only.
R, =resistance along joint path, if intact rock only.
EINSTEIN el at.: DISCONTINUITY PERSISTENCE AND SLOPE STABILITY 231
Fig. II. SLOPESIM method of slices.
in which N
i
is the total number of simulated exit joints
(critical paths) in the ith elevation interval and N
r
is the
number of such exit joints associated with
critical paths (i.e. failure paths). The identification of the
critical path through each exit point is performed
through a dynamic programming method: the algorithm
starts with the exit points on the top of the slope and
progresses backwards towards the exit points on the face
of the slope. During this backwards progression, the
algorithm considers all the physically realizable paths
through a discrete set of points, including the end points
of each joint. For more details on this procedure, see [3].
An important feature of SLOPESIM is the realistic
modeling of failure mechanisms. Driving and resisting
force calculations are based on the method of slices
which is common to many deterministic slope stability
methods (the method as applied here is simplified by
neglecting interslice forces). The principle is illustrated in
simplified form in Fig. II: the slope overlying the failure
path is partitioned into a series of vertical slices,
bounded at their bottom end by joints or intact rock.
The total driving force L and the total resistance Rare
calculated by summing slice contributions, i.e.
A fundamental feature of Lajtai's model is in the
analogy of shear resistance of intact rock bridges to
resistance in direct shear tests. At least for short intact
rock bridges, this analogy is justified by the assumption
of rigid body motion of the overlying unstable wedge in
the direction of jointing. In direct shear tests, the re-
sistance of intact rock can be mobilized in one of two
ways:
At relatively low stress levels (0'. small), the applica-
tion of shear stress in the direction of movement leads
to a minimum principal stress 0'3 equal to the tensile
strength of intact rock. Hence in this case, failure
occurs by tensile fractures that develop at high angles
to the direction of sliding (Fig. 12a). Simultaneously
with the appearance of these fractures, peak shear
resistance fa in the sliding direction is attained. There-
after, shearing at residual stress values takes place in
the direction of sliding.
At higher normal stress levels, the minimum principal
stress does not exceed the tensile strength and failure
occurs when fa equals the shear resistance defined by
the Coulomb failure criterion. In this case, shear
fractures develop in the sliding direction at the time
when the applied shear stress is maximum (see Fig.
12b).
0'0
Primary tension
. fracture (high angle)
1(-I--/-- I Secondary (low angle)
t shear fracture
The two modes can be visualized by use of Mohr's
circle. In a direct sheart test, the center of Mohr's circle
remains at all times at O'a/2 as the shear stress varies from
zero to the value at failure. For small 0'. (Fig. 13),
Mohr's circle becomes tangent to the failure envelope at
0' = - T
s
' f = 0 and thus failure occurs in tension (Mode
I). For larger 0'., the center of Mohr's circle lies more to
the right and the point of first tangency is located on the
linear portion of the envelope. Thus, as shown in Fig. 14,
this mode of failure (mode 2) corresponds to shear
failure in the traditional formulation as used by Jennings
[5]. Both types of failure can occur, but Mode 2 probably
only in high slopes with weak intact rock. Thus, Mode
2 is neglected in the following discussion.
Failures may be in plane or out of plane (en echelon).
During in-plane failure, tension cracks develop first,
(10)
7 7
SM- 1: SMi-1:(Rj-Wisina)
i-I i-I

L = Wi sin (x,
I
01 FAILURE IN TENSION
where ct is the angle of jointing, Wi is the weight of the
i
1h
slice, and R
i
is the peak shear force mobilized by the
portion of path underlying that slice. The i
1h
portion of
the path may be jointed, in which case R
j
can be
calculated through equation (5), or it may consist of
intact rock. In the latter case R
i
is best calculated using
rock resistance criteria in Lajtai [6] and Einstein el al. [2].
R = I: R
i
, (11 )
0'0
TO_
1F4Low angle primary shear
b) FAILURE IN SHEAR
Fig. 12. Direct shear failure modes--after lajlai [6J.
232
EINSTEIN el al.: DISCONTINUITY PERSISTENCE AND SLOPE STABILITY
Fig. 13. Mohr's circle-failure by tensile fracturing (Mode I).
(15) Lj = Wjsincr
IN-PLANE FAILURE OF INTACT ROCK-
PRIMARY TENSION FRACTURES @ lB,) TO JOINT PLANE
Fig. IS. In-plane failure of intact rock-secondary shear fractures in
the joint plane.
in which X is the distance between the joint planes that
define the bridge (Fig. 17) and T, is the intact rock tensile
strength. The contributions to resistance from intact
rock bridges and from joint segments are added to
obtain the total resisting force associated with a given
path.
The driving force associated with the same path is
assumed to be due solely to the overburden weight; it is
therefore calculated as the sum of the driving force
contributions Lj from each slice above a path segment
(Fig. II). If cr denotes the angle of sliding (joint angle)
and Wj is the weight of the jlh slice, then
The SM of a given path can thus be calculated. (The
effect of cleft water pressure has not been included in the
(14) R=T,X.
Point of
tonllney
followed by secondary shear fractures (Fig. 15a and b).
The angle of the tension cracks 0., can be obtained,from
Mohr's circle. The same mechanism applies to out-of-
plane failures with low-angle transitions [with
(fJ - rx) < 0., see Fig. 16], whereas for high angle transi-
tions a continuous tension crack occurs directly between
joints, without secondary shear fractures (Fig. 17).
Therefore, Mode I failures with initial tension fractures
encompass the entire range of geometrical conditions in
slopes with a single joint set, including in-plane as well
as low- and high-angle out-of-plane transitions.
In summary, intact-rock resistance R can be calcu-
lated as follows: For in-plane or low-angle out-of-plane
transitions (fJ < 0. +(X),
R=l'ad, (12)
in which d is the "in-plane length" of the rock bridge
(Fig, 16) and T. is the peak shear stress mobilized in the
direction of jointing. In terms of the intact cohesion c,
and the ratio c = Talc" the peak shear stress is
T a ~ (13)
For high angle transition (fJ ~ Or +cr),
Fig. 16. Failure of "low angle" (P < (ll + xl) transitions through
Fig. 14. Mohr's circle-failure by shear fracturing (Mode 2). intact rock.
EINSTEIN et al.: DISCONTINUITY PERSISTENCE AND SLOPE STABILITY 233
o
Fig. 17. Failure of high angle transitions through intact rock.
analysis. Although in principle simple to do, such a
feature would only be consistent if the spatial variability
in pressure distributions were expressed, which has not
been done so far.) SLOPESIM uses dynamic pro-
gramming to scan the large number of potential failure
paths and to identify, for each "exitjoinC', the path with
minimum SM.
The program has been used to conduct a parametric
study of rock slope reliability, aimed at identifying
critical variables and at obtaining simple reliability
formulae. Results from the numerical analyses are sum-
marized in the following sections.
PARAMETRIC STUDY OF SLOPE
RELIABILITY
Slope safety depends on a variety of parameters,
which for the most part describe geometry and re-
sistance. In the course of the parametric study, these
parameters have been given values according to Table I.
Slope safety resulting from the parametric studies has
been expressed in several terms:
(I) Failure Probability Pr(z), is defined as the ratio of
critical paths having negative SMs to
all critical paths. as a function of the depth z (Fig. 18)
at which the paths daylight on the slope face. If a joint
plane daylights in height interval "i" on the slope face
(Fig. 10), the probability that at least one wedge be-
longing to this interval is unstable is Pr(z,) where Zi is the
vertical distance between the mid-point of the height
interval and the slope crest. At present, the only way to
calculate Pr(z) is through repeated Monte Carlo simu-
lation of the network of joints. However, it is possible
to obtain analytical lower bounds to Pr(z). One such
bound, in many cases close to the exact value, is derived
in Appendix A.
(2) Probability Distribution of Apparent Persistence
KQ K
a
is the average persistence along an existing joint
plane that produces a SM equal to that of the associated
critical path, the plane and its critical path daylighting
at the same point on the slope face (the critical path may
be the particular joint plane or it may have an en echelon
shape involving other joint planes). It follows from the
definition that K
a
is not smaller than the actual persis-
tence of the plane, K. The probability distribution of
apparent persistence depends on depth. Of special inter-
est is the variation with depth of the mean value m
K
and
the standard deviation (1K, which together with the
critical persistence defined 'below are used to define a
second moment reliability index.
(3) Critical Persistence K,.. The critical path is unsta-
ble and failure occurs if K. exceeds the critical value of
persistence K
c
which, in using the parameters in Table I,
is given by
Table I. Parameters and their ranges used in SLOPESIM parametric study
Parameter
Value or
Symbol Definition range
Geometric parameters (see Fig. 18)
H Slope height Fixed 30 m (100ft)
=
Depth below slope apex 0-30 m (0-100 ft)
(1
Slope angle 50-90
j[
Mean joint length (joint length 3-12m (lQ-40ft)
..
assumed exponentially distributed)
RBR Mean rock bridge length Not directly varied.
(rock bridge length assumed considered by
K
exponentially distributed) varying persistence
Mean joint plane persistence 10-73%
SP Mean joint plane spacing (joint 0.6-3 m (2-10 fll
plane spacing assumed exponentially
distributed)
I Mean joint intensity. a derived
_ K
variable I =-
. SP
I 3D-80
Cl Jomt plane ang e
Resistance parameters
c, Intact-rock cohesion 0.30-24.0 MPa
(assumed to be twice the rock (8-500 ksf)
tensile strength)
tP, Intact-rock friction angle (not Fixed 30
important at low stress levels)
c Joint cohesion Fixed 0
J Joint friction angle Q-40"
Other parameter
.,
Unit weight Fixed 2.2 g/cm
J
I,
(150Ib/ft
J
)
234
EINSTEIN e( al.: DISCONTINUITY PERSISTENCE AND SLOPE STABILITY
JL C
r
(4) Second-Movement Reliability Index p is the num-
ber of standard deviations separating the mean K
a
from
the critical value K
e
,
For FK (K) the cumulative distribution function of ap-
parent persistence at a given depth, the probability of
failure at that depth becomes
P
r
= I - F
K
(K
e
)
= I - F
K
(mK. +paK) (18)
The sensitivity of these four safety measures to the
parameters of Table I has been studied by varying one
parameter at a time within the specified ranges, while
holding all the other parameters fixed at given values. Of
special interest is the dependence of P
r
on depth z, which
is shown in Fig. 19. The same figure contains a plot of
the lower bound PI (see equation A2 in the Appendix)
which for slope heights up to 30 m (100 ft) and for
typical values of c, (c, 24 MPa (500 ksf) provides a
good approximation to P
r
(accuracy depends also on
other parameters, such as the mean joint length JL and
the friction angle for the joints, cPj)' The parametric
study also revealed that dependence of P
r
, m
K
and pon
intact-rock strength c, for depths up to 30 m (100 ft), is
small. This is a welcome result because it allows one to
calculate the reliability index {J in equation (17) after a
single use of SLOPESIM to obtain representative values
of mK. and aK.
Similar sensitivity analyses were made with respect to
the other parameters of Table I (see [7]), leading to the
following conclusions:
The influence of strength parameters c, and cP; usually
dominates other parameters. Slopes with high values of
C
r
and rpj (c
r
24 MPa (500 ksf), cPj ex) tend to be
reliable at all depths investigated (up to 30 m (100 fl,
regardless of the other parameters. When C
r
and cP; are
small, joint and slope geometry become important.
Among the joint geometry parameters (l<, JL, SP), mean
J
Cr =1.2-24 MPa
4>i =o
8 =60
0=40
y,= 2.211/cm
3
X=40
SP= 1.5m
K=50%
30
40
z_
Fig. 19. Effect of intact-rock cohesion (c,) on P
r
(=).
m
(0 25 50 100 ft:
50_---------------,
persistence K has the largest influence. Changing the
mean joint length JL may also substantially modify P
r
and {J at any given depth, whereas mean joint spacing SP
plays a less significant role. For small K (K < P
r
depends on K and SP almost exclusively through the
ratio K/SP, which to first-order accuracy equals the
mean intensity I.
Joint inclination, ex, has varying effect of reliability.
Values of ex for which reliability is smallest are typically
around 45. As a increases above this value, approaching
the slope angle, reliability increases due to the decrease
in driving force. This effect is especially significant in
slopes with weak intact rock (c
r
< 4.8 MPa (100 ksf)
and weak joints cPjex). Reliability increases also as ex
decreases from 45, especially as it approaches cPj'
Slope depth z is a very significant parameter. For this
reason, probability of failure is presented here as Pr(z)
curves. Figure 20 shows a few such curves for different
slope parameter combinations and leads to two obser-
vations:
(I) The shape of Pr(z) does not vary much with any
parameter and displays a minimum at a "characteristic
depth", Ze' This depth is in some cases outside the range
shown. That P
r
decreases with slope height before in-
creasing again for H > Ze seems at first glance to be
incorrect. However, the result is in fact correct: it is due
to the overriding part played by persistent portions of
joints for low slope heights (and thus small driving
forces). For fixed mean joint length JL the probability of
a 100% persistent joint increases as z decreases (at Z = 0
the probability of failure equals the average persistence
K!). Observations in nature are consistent with this
result. Natural slopes are often convex near the crest.
Although weathering effects playa role in this geometry,
(17)
(16)
Ke-m"
u _ ".
p - .
aKa
Fig. 18. Slope geometry for parametric study.
w cos ex
c=--==--
K
e
= 100 [I _2c(tanex - tan cPi) ,
Jk+l - 2c tan cPi
where
EINSTEIN el al.: DISCONTINUITY PERSISTENCE AND SLOPE STABILITY
235
K I ~ - - - - - -
0
z-
0 z ~ z-- z-
i
,
~
t t
\
f
\
P,
\
P,
\
\
/',
PI ....
...........
..........
---
0
z--
0
z--
z_ z_
Fig. 20. Probability of failure P
r
as a function of slope depth: and of the other slope paramctcrs.
1

the higher probability of a joint isolating a wedge near


the crest is significant.
(2) Pr<z) can often be approximated by p.(z) (see
Appendix A), which is the probability that the joint
plane at depth z is 100% persistent. This approximation
is good either when C
r
or c/>j is high (c/>j ::::::: oc or C
r
> 24 MPa
(500 ksf). Also when both C
r
and c/>j are small (c/>joc
and C
r
< 4.8 MPa (100 ksf), the approximation remains
good for
(a) Shallow depths [z ~ 6-9 m (20-30 ft)]
(b) Short joint length [JL ~ 3-6m (10-20 ft)]
(c) Large joint-plane spacing [SP > 3 m (10 ft)]
(d) Joint inclination angle oc close to c/>j or
to 8.
At very large depths, (on the order of 300 m (1000 ft), if
C
r
is very large) the approximation Pr(z)::::::: p)(z) loses
accuracy regardless of the other parameters.
The parametric analysis has been instrumental in
assessing the influence on reliability of strength and
geometry parameters. Critical parameter combinations
that lead to failure have been identified. These parame-
ters (e.g. JL and K if C
r
and c/>j are small) should be
accurately determined.
EXTENSION OF THE SLOPESIM APPROACH
In its present form, the model is limited to slopes with
a single parallel set of joints and neglects 3-D effects.
However, extensions are possible. For example, Shair [8]
has developed a version of SLOPESIM for two parallel
joint sets. As expected, reliability is smaller than in
otherwise comparable cases with a single joint set, the
plagnitude of the safety decrease depending on the
particular parameter combination. A first attempt at
including the third, along-slope dimension has been
made by O'Reilly [7], but more work is needed. Also, the
mechanical models of joint and intact-rock failures can
certainly be improved. Finally, procedures of the type
used here for slopes have potential application in rock-
mass stability problems in tunnelling and can be ex-
tended to problems of rock-mass deformation and flow.
CONCLUSIONS
Joint persistence has a major effect on rock-mass
resistance, and yet, it is difficult to define a persistence
parameter simply and directly related to resistance.
First, joint geometry internal to a rock mass is not
known with certainty, and second, failure involves a
combination of mechanisms. including shearing along
joints and failure through intact rock, either in plane or
en-echelon.
The proposed approach expressed probability of rock-
slope failure as a function of joint geometry and intact
rock and joint resistance. Spatial variability of joint
geometry is taken into account by making use of statis-
tical information obtained from standard joint surveys.
Probability of failure as derived with the SLOPESIM
approach or the related expression of apparent persis-
tence thus makes it possible to represent the effect of
joint persistence directly.
Parametric studies show the relations between rock-
slope reliability (I-P
r
) and various mechanical and geo-
metric parameters; graphs of probability of failure vs
slope height are particularly illustrative. An important
result is the indication of when strength parameters are
more important than geometry, and vice versa.
Although an initial step, the proposed approach
promises insights into a major problem in rock mechan-
ics.
Receit'ed 22 Decemher 1982; reri.red 20 May 1983.
"us lOIS ('
236 EINSTEIN ,'I al.: DISCONTINUITY PERSISTENCE AND SLOPE STABILITY
REFERENCES
I. Call R. D. and Nicholas D. E. Prediction of step path failure
geometry for slope stability analysis. Proc. 19Th U.S. Symp. on Rock
,\fee/ranics (1978).
2. Einstein H. H. el al. Risk analysis for rock slopes in open pit
mines. Parts I-V. USBM Technical Rept J0275015 (1980).
3. Glynn E. F. A probabilistic approach to the stability of rock
slopes. Ph.D. dissertation. M.LT. (February. 1979).
4. Hasofer A. M. and Lind N. C. Exact and invariant second moment
code format. ASC J. "KnK Mech. Dil'. 100. 111-121. No. EM I.
Proc. Paper 10376 (February. 1974).
5. Jennings J. E. A mathL'II1atical theory for the calculation of the
stability of open cast mines. Proc. Symp. on Iht' Theoretical Back-
ground 10 Ihe Planni"K Opt'n Pil Afines. pp. 87-102. Johannesburg
(1970).
6. Lajtai E. Z. Strength of discontinuous rocks in shear. Geolee/miqlll'
19(2). 218-233 (1969).
7. O'Reilly K. J. The elTL'Ct of joint phase persistence on slope
reliability. M.Sc. thesis. M.J.T. 553 pp (1980).
8. Shair A. K. The effect of two sets ofjoints on rock slope reliability.
M.Sc. thesis. M.LT. 307pp (1981).
9. Stimpson D. Failure of slopes containing discontinuous planar
joints. Pro(. 19Th U.S. Symp. (III Rock Afedwnics. pp. 246-300
(1978).
APPENDIX A
An 10ll'er bound 10 11r, prohahility (if slopt' failure
The probability of failure. P,(:) has been defined as the fraction of
unstable critical paths that daylight at depth :. Lower bounds to P,(:)
can be obtained by constraining the geometry of the critical path and
the pattern of jointing that can produce failure. One such bound is
obtainL'd here under the following conditions: with reference to Fig.
AI. failure of the joint plane exiting at : can occur only if:
(I) The joint plane AN is 100"" persistent. i.e. L, L. and failure
is by sliding along AN.
(2) The joint plane AA' is not completely jointed: however. the next
joint plane BB' is completely jointed persistent) and the distance
(0) between the joint planes is sulliciently small (smaller than a critical
distance D,). Failure occurs by sliding along the jointed segment of
AA'. fracturing through intact rock to connect to BB' and sliding
along 8B'.
(3) Only parts of AA' and BB' are jointed but the jointed parts
overlap or are equal to L (L
I
+ L! L). and the disllmce D is smaller
Fig. AI. Geometry for analytical lower bound to P
r
(:).
than Dc. Failure occurs by sliding along the two jointed portions and
a connecting fracture through intact rock.
Because these three failure events are mutually exclusive, the proba-
bility Pc that anyone of them occurs is the sum of their individual
probabilities (PI' P
2
and P
J
) and
P,(:) P
c
(:) =P,(:) +P
2
(:) +P
J
(:) (AI)
Let J[ be the mean joint length. RBR: the mean rock bridge length.
K = [[/(JL + RBR) the mean joint plane persistence and S]i the
average spacing between joint planes. Also denote by Dc the critical
joint separation that corresponds to unstable wedges in cases 2 and 3
(note that Dc is stress-dependent and thus dependent on its location in
the slope. Since the following approximation omits P
2
and PJ. no
further consideration of Dc is necessary). Thus. using Glynn's [3]
probabilistic model ofjoints. one finds the following expression for P"
Pl' and p):
P, =II. e-L'JL =II. e-:'lJI:... =K e-:'lL......
p! = (I - PI) p.(1 - e-D.SP,
(A2)
Figure 19 showed PI and Pc derived with SLOPESIM as a function of
depth and of intact-rock strength c, (P, does not depend on c,). while
all other parameters are kept constant. As c
r
increases. Pc becomes
closer to PI because in the limit, as C,-'CO, failure can occur only if a
joint plane is 100% persistent (Mode I). The probability PI is thus a
simple and often good approximation to Prl2.7).
I
I
.1
I

You might also like