A Practical Procedure For The Back Analysis of Slope Failures in Closely Jointed Rock Masses

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Int. J. Rock Mech. Min. Sci. Vol. 35, No. 2, pp.

219±233, 1998
# 1998 Elsevier Science Ltd. All rights reserved
Printed in Great Britain
PII: S0148-9062(97)00335-5 0148-9062/98 $19.00 + 0.00

A Practical Procedure for the Back Analysis


of Slope Failures in Closely Jointed Rock
Masses
H. SONMEZ
R. ULUSAY
C. GOKCEOGLU
Where closely jointed rock masses are encountered in slopes, failure can
occur both through the rock mass, as a result of combination of macro and
micro jointing, and through the rock substance. Determination of the
strength of this category of rock mass is extraordinarily dicult since the
size of representative specimens is too large for laboratory testing. This di-
culty can be overcome by using a non-linear rock mass failure criterion or
by back analysis of such slopes to estimate the rock mass strength. In this
paper, a practical procedure and a computer program are presented for the
back determination of shear strength parameters mobilized in slopes cut in
closely jointed rock masses which obey a non-linear failure criterion rather
than a linear one. The procedure shows that the constants to derive normal
stress dependent shear strength parameters of the failed rock masses can be
determined by utilizing a main cross-section and without a pre-determined
value of rock mass rating (RMR). Trials are made for di€erent RMRm and
RMRs values corresponding to various possible combinations of the constant
m and s, which are used in the Hoek±Brown failure criterion, satisfying the
limit equilibrium condition. It is also noted that the procedure provides a
quick check for the rock mass rating obtained from the site investigations.
The method is used in conjunction with the Bishop's method of analysis
based on circular slip surfaces. The procedure outlined in this paper has also
been satisfactorily applied to documented slope failure case histories in three
open pit mines in Turkey. # 1998 Elsevier Science Ltd.

INTRODUCTION (a) Methods suitable for slopes in soils or soil like


In a rock mass with clearly de®ned discontinuity sets, materials where the strength of the material can be
failure mechanisms related to discontinuities can be determined from testing small specimens of the ma-
analyzed and the stability of slopes excavated in that terial in the laboratory.
rock mass can be calculated providing the shear (b) Methods suitable for slopes in hard jointed rocks
strength along the discontinuities is known. However, where slope stability is controlled by the discontinuities
such an analytical approach might not be feasible for in the rock material. The potential for failure is depen-
slopes containing multiple discontinuity sets with large dent on the presence and orientation of discontinuities,
variations in mechanical characteristics. Continuum and shear strength along them.
calculations for engineering structures in or on a rock (c) Methods suitable for closely jointed rock masses
mass, whether analytical or numerical, cannot be where failure can occur both through the rock mass,
appropriate, since over-simpli®cations result from pre- as a result of a combination of macro and micro joint-
senting the rock mass as a continuum. ing, and through the rock substance. Determination of
In general, the slope stability determination methods the strength of this category of rock mass is a much
depending on the material involved may be divided more dicult task. There are formidable diculties in
into three broad categories: the sampling and testing of undisturbed samples that
are suciently large to represent the combined e€ects
of rock material and discontinuities. The possibility for
Hacettepe University, Faculty of Engineering, Geological
Engineering Department, Applied Geology Division, 06532 the measurement of the shear strength of such rock
Beytepe, Ankara, Turkey. masses is usually based on some form of classi®cation
219
220 SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES

Fig. 1. E€ect of scale on rock strength and possible mechanisms of failure in rock slopes.

techniques [1±3] in conjunction with a non-linear fail- when applied to near-surface applications such as rock
ure criterion [4±8]. slopes. This is due to the restrictions of these systems
A rock mass is described as closely jointed when the which are not well considered.
joint spacing is small in relation to the scale of the Recently, an empirical failure criterion developed by
project in question. In closely jointed media it seems Hoek and Brown [5±8] has been adopted to the RMR
appropriate to assume that the material is approxi- rock mass classi®cation scheme [1] to assess the shear
mately isotropic and homogeneous, i.e. there are no strength of the jointed rock masses in surface and
clearly de®ned joint planes or joint sets which control underground excavations. This approach has been also
the form of the failure mode. In these rocks, the joint employed in slope stability analyses by several
spacing is a fraction of meter, the individual particles investigators [11±14]. The slope mass rating (SMR)
of rock mass are very small compared to the dimen- classi®cation scheme proposed by Romana [10] also
sion of slope and these particles are not interlocked involves the input parameters used by the RMR-sys-
due to their shape. Depending on the number and tem, but generally provides assessments on structurally
nature of the discontinuities, the intact rock pieces will controlled slope failures.
translate, rotate or crush in response to stresses The main input parameters used in various classi®-
imposed on the rock mass. The behavior of the mass is cation systems are more or less the same. Namely,
thus a consequence of the combined action of a large these systems consider intact rock strength, RQD, dis-
number of individual joints. When the rock mass con- continuity spacing, condition and orientation of dis-
tains a number of discontinuity sets, having relatively continuities and groundwater conditions. Although a
small spacings in relation to the slope size, failure can number of additional input parameters and some
occur along a shear surface similar to those observed modi®cations are required in the RMR classi®cation
in soil slopes. Therefore, the required conditions for a scheme, the advantage of the system is that it provides
circular failure are mostly satis®ed in heavily jointed an easy connection to the Hoek±Brown failure cri-
rock masses as illustrated in Fig. 1. terion for jointed rock masses. The intact rock strength
The standard method for assessing the strength of a is one of the input parameters involved in the RMR-
geotechnical material is to recover a sample and test it System and is only of limited interest with regard to
in laboratory. In the case of a closely jointed rock the stability of rock slopes in which failure is most
mass it is clearly not possible to recover a sample that often associated with the shear strength of discontinu-
is large enough to represent the joint system. ities. Sometimes a rock mass having low intact rock
Therefore, an empirical approach such as rock mass strength is a consequence of the failed rock containing
classi®cation can be attractive alternative, provided a large number of discontinuities. In addition to this,
that the appropriate parameters are included in the the purpose of including intact rock strength in the
classi®cation system. In order to overcome the dicul- classi®cation system for slopes is to give an assessment
ties in laboratory determination of the shear strength of wall rock strength of the discontinuities. As stated
of jointed rock masses; the Hoek±Brown failure cri- by Hoek [15], the Hoek±Brown failure criterion is only
terion in conjunction with geomechanics classi®cation applicable to intact rock or to closely jointed rock
system [1] is commonly used. masses which can be considered homogeneous and iso-
Rock mass classi®cation has been applied success- tropic. The rock mass parameters RQD and disconti-
fully in tunnelling and underground mining [1±3, 9]. A nuity spacing de®ne the block size and block form and
number of systems, introduced by Bieniawski [1] and are also very useful in analyzing stability of slopes.
by Romana [10], has also been suggested for rock Therefore, these two parameters are considered by the
slopes. It should be noted, however, that the use of authors to be the parameters of meaningful value in
rock mass classi®cations developed particularly for rock mass classi®cation, particularly for slopes exca-
underground works may lead to unsatisfactory results vated in closely jointed rock masses.
SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES 221

The condition of discontinuities includes the items sent paper think that the above mentioned rating sys-
related to roughness, continuity, in®ll material, aper- tem is still questionable. First of all, Singh and
ture and degree of weathering. Laubscher [9] takes Gahrooee [16] did not change the values of ratings
into account in his ®nal RMR rating only the con- which can reach up to ÿ60 points out of 100. As dis-
dition factor of the most prominent discontinuity set cussed before, such an adjustment is not applicable in
or the discontinuity set with the most adverse in¯uence practice. Secondly, in a closely jointed rock mass, the
on the stability of an underground excavation. This is most probable mode of failure occurs in the form of a
too simple for slopes where the failure is often not circular shape regardless of discontinuity orientation.
determined by one main discontinuity set. Particularly Consequently, only one de®nition namely ``one poss-
for the slopes in a closely jointed rock mass, the con- ible mode of failure'' is considered to be more logical,
dition rating becomes more important and it is taken and a single adjustment of ÿ5 for discontinuity orien-
as the mean value of the condition ratings of the tation is more realistic for slope failures in closely
di€erent discontinuity sets. For the rock slopes, the jointed rock masses.
persistence has a considerable in¯uence on the stability Some factors such as method of excavation, major
and the RMR-System takes into account the persist- planes of weakness or change in stress are treated as
ence as a quantitative factor. Weathering a€ects the local features which have in¯uenced the rock mass at a
condition of discontinuities and discontinuity spacing. particular location and are not rock mass constants.
It is also noted that the state of weathering is con- These have been discussed by Laubscher [9],
sidered to be a local feature which has changed the Romana [10] and Kendorski et al. [17]. The greatest
rock mass at a particular location. Within the lifespan in¯uence of the method of excavation will be on the
of a cut slope, future weathering might lead to instabil- spacing of discontinuities. Depending of the blasting
ity. Therefore, the weathering parameter included in damage, blasted slopes may have closer discontinuity
the RMR-System is a very important factor in slope spacing than natural slopes. Therefore, in order to
stability. compensate for the in¯uence of such local factors,
The main problem of water in slopes is the pressure necessary adjustments [1, 9, 17] are taken into consider-
of the water in discontinuities. The presence of water ation in rock mass classi®cation for the slope failures
in discontinuities reduces the stability of slopes by in closely jointed rock masses investigated in this
reducing the strength of discontinuity surfaces or of study.
any in®ll material. The water pressure is taken into On the other hand, during a classi®cation process,
account in the slope stability analysis by estimating the serious diculties are encountered in determining or
pressure or the position of groundwater table in slope. describing some of the rock mass parameters, particu-
But the softening or weakening e€ect of water on dis- larly in poor quality rock masses [18±20]. Due to such
continuity surfaces becomes more important for slopes. uncertainties, the calculated rock mass rating may
Consequently, the groundwater rating is an integral erroneously a€ect the constants and shear strength
part of the rock mass classi®cation and should be parameters derived from the non-linear rock mass fail-
assigned for each particular outcrop for slopes. ure criterion. The most reliable way to obtain a mean
In closely jointed or crushed rock masses it is very value of the constants m and s employed by the
dicult or impossible to determine the orientation of Hoek±Brown failure criterion in an extended slope is
discontinuities. In such cases, the orientation is not by back-calculation and by comparison of the results
meaningful, because part of the rock mass will fall into of back-calculation with the available data derived
the underground opening and require immediate sup- from the Hoek±Brown criterion [21]. However, in
port regardless of discontinuity orientation. In the case some cases it is unlikely that an accurate assessment of
of slopes excavated in such rocks, the situation is not the true strength parameters for a given rock mass will
di€erent. Bieniawski [1] in his RMR classi®cation ever be available due to limitations, so RMR values
scheme, suggests rating adjustments for discontinuity cannot be precisely determined. Because the results of
orientations, relative to proposed slope orientation, back-analysis provide a range of combinations of
ranging between 0 and ÿ60. No guidelines have been apparent friction angle and cohesion, the problem of
published for the de®nition of each adjustment values, parameter selection becomes dicult in such cases.
and no reference is given by Bieniawski to use of the The procedure presented herein is to perform a
RMR classi®cation in slopes. The reason for this lack back-analysis of failed slopes cut in jointed rock
of use is probably the extremely high values of the masses to estimate the rock mass rating and shear
adjustment rating values which may sometime result in strength parameters mobilized at the time of failure.
negative RMR values. Therefore, the ratings assigned The main philosophy of the method recognizes that it
for discontinuity orientation adjustments suggested by is unlikely that an accurate assessment of the value of
Bieniawski [1] is unrealistic. Singh and Gahrooee [16] RMR and shear strength parameters for a given rock
proposed better and clearer descriptions for disconti- mass will ever be available. A detailed description of
nuity orientation in slopes. This approach was quanti- the procedure which can be readily incorporated into
®ed on the basis of rating with regard to the number the conventional back analysis of a slope failure in a
of possible modes of failure. The authors of the pre- jointed rock mass, where only a single cross-section is
222 SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES

available, is presented with a computer solution devel- work for the redesign of failed slopes and for new pro-
oped for the purpose. The proposed method is also jects in similar types of material. Therefore, it is con-
applied to failure case histories in jointed rock masses sidered that back analyses are an integral part of the
at three open pit mines located in Turkey to check its slope design.
performance. The shear strength parameters of a failed slope have
been back calculated by geotechnical engineers and en-
gineering geologists in the following procedures:
METHOD OF ANALYSIS (a) Assuming the value of the angle of internal fric-
Theoretical background-basic procedure tion f or of the cohesion c to calculate another [22]
One of the most dicult tasks in slope stability (Fig. 2(a)).
analysis is the determination of the shear strength par- (b) Utilizing a main cross-section of a failed slope
ameters (c, f) along the sliding surfaces. In geotechni- and another cross-section near the main one in the
cal engineering practice, failure of a slope can be same failed slope or utilizing two cross-sections in two
regarded as a full scale ®eld test and an assessment of failed slopes which have similar geological and hydro-
any failure is, therefore, of considerable value. geological conditions to establish two equations and
Appropriate geomechanics models can be used to esti- then evaluate the values of c and f (single solution;
mate the values of shear strength parameters on the Fig. 2(b)).
basis of certain assumptions. These back calculated (c) Because of the variations in the mechanical prop-
values may then be used for preventative and remedial erties of the same material in di€erent places, utilizing

Fig. 2. Basic back analysis approaches applied for the slope forming materials obeying linear failure envelopes: (a) derived
range of c and f and determination of c from an assumed f; (b) single solution for two slides with di€erent geometry; (c)
multiple solutions for four slides with di€erent geometry; and (d) multiple solutions with a comparison with laboratory de-
rived strength test results.
SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES 223

more than two slope cross-sections to obtain as many ing can be very dicult because of tendency of these
as n(n ÿ 1)/2 points of intersections (solutions) for n materials to slake and de-laminate. In addition, as
curves c(f) (multiple solutions; [23]; Fig. 2(c)). The set reported by Unal et al. [18], Ulusay et al. [19] and
of continuous curves represents the range of back cal- Unal [20], serious diculties are encountered in deter-
culation solutions from which the most realistic sol- mining or describing some of the rock mass par-
ution can be obtained based on engineering judgement, ameters, particularly in weak, strati®ed and clay-
experience and veri®ed with shear test results if these bearing rocks. In such circumstances overestimated
are available (Fig. 2(d)). rock mass ratings might be obtained and they result in
The above procedures, however, are based on the deriving di€erent m and s values than those in real
back calculation of the shear strength parameters of situation. On the other hand, in some areas where
the materials obeying linear Mohr±Coulomb failure slope failures have occurred, because of the limited
criterion which are characterized by c and f values number of outcrops or no borehole data, the rock
independent from the normal stress. But a consensus mass rating can not be precisely determined.
has gradually emerged among the rock mechanics
Therefore, a back analysis based on such limited or
community that the failure envelope for a closely
questionable data may yield unrealistic results.
jointed rock mass is curved rather than linear. The
The strategy of this study is aimed at overcoming
authors believe that the Hoek±Brown non-linear fail-
the diculties associated with the limitations discussed
ure criterion [4±7], which has gained an increasing
above. In this strategy, a procedure is suggested to
popularity in stability analyses made in conjunction
with rock mass classi®cation systems, provides a mean- identify the most reasonable and a common rock mass
ingful estimate of rock mass behavior. Due to the non- rating (RMR) value which corresponds to the pair of
linear nature of this failure criterion, the above men- m and s satisfying the limit equilibrium condition. In
tioned methods are unrealistic for use with closely jointed rock masses obeying the Hoek±Brown failure
jointed rock slopes, i.e. the shear strength parameters criterion, a function F, the conventional factor of
of a failure surface in closely jointed rock masses can safety commonly speci®ed in the limit equilibrium
be calculated for any speci®c normal stress value using methods of slope stability analysis, depends on several
the material constants (m and s) as a function of rock variables and for any particular sliding surface may be
mass rating (RMR) from the following equation [24]; written in the following form:
for disturbed rock masses: F ˆ FfRMR…m, s†, GW, G g …3†
 
m RMR ÿ 100
ˆ exp …1a† where RMR: rock mass rating (m and s are the ma-
mi 14
terial constants), GW: groundwater conditions prevail-
  ing in the slope, G: geometry of the slope and the
RMR ÿ 100
s ˆ exp …1b† failure surface.
6
for undisturbed or interlocking rock masses:
 
m RMR ÿ 100
ˆ exp …2a†
mi 28
 
RMR ÿ 100
s ˆ exp …2b†
9
where mi is the material constant of intact rock sample
and can either be calculated form laboratory triaxial
test on intact samples or taken from the tables pro-
posed by Hoek [24], and Hoek et al. [8].
In the case of a slope instability with accurately
speci®ed failure geometry in a closely jointed rock
mass, if the value of RMR is precisely determined and
the triaxial test data are available, back analysis of the
failure provides a realistic comparison between the
rock mass strength obtained from the failure surface
yielding a safety factor of unity and the failure envel-
ope derived with the updated Hoek±Brown failure cri-
terion as reported by Ulusay and Aksoy [21] (Fig. 3).
However, in weak sedimentary rocks, such as shales, Fig. 3. Comparison between the rock mass shear strength obtained
from the failure surfaces yielding safety factors of unity and the fail-
marls and siltstones, and in heavily fractured schistose ure envelope with the updated Hoek±Brown criterion for coal-bear-
rock masses, preparation of specimens for triaxial test- ing rocks (after Ref. [21]).
224 SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES

The real factor of safety F is considered to be given by the updated Hoek±Brown failure
known and equal to one for a case study concerned criterion [24].
with a slope that has failed. The value of the geometry Step 3. Trials are made for di€erent values of
data G in Equation (3) can be delineated from the RMRs(s) to obtain various possible combinations of
results of ®eld inspection or by surveying the actual RMRs and RMRm satisfying the limit equilibrium con-
failed slope. The values of the constants m and s at the dition.
time of failure are unknowns and groundwater con- The results of the back analysis are best presented in
dition, GW, may be either known or unknown. a RMRs±RMRm function forms, i.e. RMRs plotted
The suggested approach involves the determination against RMRm considering each combination to lead
of various possible combinations of m and s satisfying to a value of the factor of safety F = 1 (Fig. 4). All
the following equation: the points (or RMR pairs) located on the curve indi-
cate a safety factor of unity. Because the closely
1 ˆ FfRMR…m, s†, GW, G g …4†
jointed rock mass is an approximately homogeneous
where G and GW are considered as known in the pro- material, it is logical to consider that the rock mass
cedure. must have a unique RMR value from which a pair of
The back analysis method presented herein is based m and s representing a given rock mass can be derived
on the following assumptions: using Equations (1a)±(b) and (2a)±(b). Thus, if a
(1) The geometry of the slope before and after fail- straight line passing from the origin of the graph (see
ure, the position of the sliding surface, and the Fig. 4) with an inclination of 458 is drawn, it intersects
groundwater conditions are known. the RMRs±RMRm curve at a certain point which indi-
(2) The mechanism of the movement is known. cates a common RMR (RMRRM:the actual RMR for
(3) A condition of static equilibrium at the point of the rock mass) value for both constants at the time of
failure (limit equilibrium) exists at the time of failure. failure and utilization of this back analyzed RMRRM
(4) In closely jointed media, it seems appropriate to value will yield the right combination of the two con-
assume that material is approximately homogeneous. stants, m and s, of the rock mass.
(5) What is obtained by back calculation is a
weighed mean value of RMR and corresponding m Software description
and s values along the failure surface at the time of The method described above has been used to
failure. develop a computer program for conventional determi-
(6) A set of relations between the RMR from the nistic slope stability analysis and back calculation. The
Bieniawski's rock mass classi®cation [1] and the con- computer program was written in QBasic and can run
stants given in Equations (1a)±(b) and (2a)±(b) are on any type of IBM PC or compatible equipped with
used in conjunction with the equations given by the a graphics card and monitor. The program
updated Hoek±Brown failure criterion [24]. HOBRSLP, which has routines that search the more
(7) Uniaxial compressive strength (sc) and the ma- critical failure surface in a grid system or automati-
terial constant mi are the input parameters. cally, can handle slope stability analysis of circular slip
The back analysis procedure starts with the fact that surfaces for slopes involving many benches with di€er-
the constants m and s of a given rock mass depend ent geometries, various materials and di€erent ground-
upon an RMR value (Equations (1a)±(b) and (2a)± water conditions, and includes simpli®ed Bishop's
(b)), and therefore, various possible combinations of method of analysis [25].
(m, s) pairs at the time of failure (F = 1) can be de- Two options are included in the program: (a) con-
rived from di€erent RMR values. The procedure ventional stability analysis for searching the most criti-
which performs back calculations for three unknown cal failure surface and corresponding lowest factor of
parameters can be carried out using the following al- safety; (b) back analysis of a failed slope with known
gorithm. failure geometry. Input data for the program includes
Step 1. One variable, RMR, out of three unknown the coordinates of the points specifying slope geome-
geomechanical parameters (RMR, m, s) is selected and try, water conditions prevailing in the slope, and ma-
the second unknown, the constant s, is calculated by terial properties. It will also prompt users to enter the
the utilization of Equation (1b) or Equation (2b) tension crack position. Output consists of a table of
depending on the condition of disturbance (blasted input data, safety factor, a cross-section of the slope
and/or excavated rock, or none) of the rock mass. The showing all strata, water table, the failure surface, and
RMR value selected to calculate the parameter s is a list of ci, fi, sn, t for each slice base if the case con-
denoted by RMRs. sists of materials having non-linear failure envelopes.
Step 2. By utilizing the position of the sliding sur- Three di€erent methods of shear strength data input
face, normal stress acting on each slice base is calcu- are incorporated in the program with keyboard selec-
lated. Keeping the previously chosen RMRs value and tion of the input mode for conventional analysis.
the corresponding RMR (RMRm) which lead a value These three modes are as follows:
of safety factor of unity are calculated by trial and (1) Input of the known shear strength parameters
error technique in conjunction with the equations derived from linear Coulomb equation.
SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES 225

allel to the normal faults crossing the Tertiary depos-


its. Excepting local deviations, three dominant joint
sets dipping 758±858 NE and SW were identi®ed. Their
persistence is high and reaches up to 8 m in some
places. The presence of cross joints, faults and ¯at
lying bedding planes result in a closely jointed rock
mass. The groundwater level rises above the coal seam
into the compact marls and where seepage occurs it
tends to decline toward the compact marl±coal seam
boundary. Thus, the failed part of the investigated
slope was dry.
In the strip coal mine, the overburden rocks com-
posed of the compact marls were evaluated based on
Fig. 4. Basic concept of the proposed back analysis technique. Bieniawski's 1989 classi®cation [1]. The data required
for rock mass rating determinations were obtained
(2) Calculating the shear strength parameters from from the geotechnical logs recorded and the scanline
input data for rock types, RMR value, sc, and ma- surveys carried out in accordance with the procedure
terial constant mi. suggested by ISRM [26]. Values of RMR for the rock
(3) Calculating the shear strength parameters from mass were determined for a number of individual sec-
normal stress (sn) acting on each slice base, and the tions from seven fully cored geotechnical boreholes
constants A and B for the materials ®tted to power considering drill-run lengths ranging between 1 to 3 m.
curve strength equation (t = AsB). In addition, a total of seven scanline sections were also
The back calculation option provides the use of the evaluated. Joint systems show negative exponential dis-
®rst two modes mentioned above. In the back analysis tribution. Mean joint spacing (x) and the average num-
option, mi and sc are given as material properties with ber of joints per meter (l) of the rock mass were
the condition of rock mass (disturbed or undisturbed). calculated as 0.386 m and 2.59 mÿ1, respectively. In the
The existing program can analyze slopes with up to compact marls overlying the coal, excepting occasional
150 slices. The steps to be followed during the ex- laminated levels, spacing between bedding planes ran-
ecution of the program are shown diagrammatically in ged 0.3 to 1 m. Discontinuity surfaces observed on the
the ¯ow chart illustrated in Fig. 5. faces of the benches were normally dry. However,
moisture appeared in some places when the surfaces
were scraped by a geologist hammer. The ranges of
EXAMINATION OF THE PROCEDURE ON ACTUAL the ®ve main parameters employed in the determi-
EXAMPLES nation of RMR values are tabulated in Table 1. As
The procedure outlined above has been applied to explained in the ®rst section, the adjustment rating for
failed slopes in three open pit mines located in the wes- discontinuity orientation was quanti®ed on the basis of
tern and central parts of Turkey (Fig. 6). All the slopes rating with regard to the number of possible modes of
presented in the particular and well documented case failure [16]. In this study, only one mode of failure, cir-
histories were cut in jointed rock masses where the cular failure through the rock mass, was considered
joint spacing is a fraction of a meter. It is, therefore, for discontinuity orientation adjustment. Mining appli-
very much smaller than the scale of the cut slopes cations include dynamic processes. In the studied pit a
which are tens of meters high. controlled blasting with a slight damage to loosen the
overburden, compact marls, is made. For this con-
An externally loaded highwall slope failure (Case 1) dition, a blasting damage adjustment of 0.94 [17] to
The particular case history presented and described the RMR values of the compact marls was assigned.
below is concerned with the instability of a highwall in Using the statistical methods, individual RMR values
Eskihisar strip coal mine (Yatagan±Mugla) in south- were assessed and then RMR values ranging between
western Turkey. No sign of instability in highwalls was 50 and 62 with a mean value of 53 were obtained. Due
observed until 1989. During a comprehensive slope to light blasting carried out in the compact marls to
stability research project by Ulusay [14], the highwall loosen the overburden, disturbed rock mass condition
of the ninth slice was found to be unstable after load- is considered and the value of mi (9.87) for intact rock
ing the slope by a temporary spoil pile (Fig. 7). was calculated by linear regression analysis on the
The failed highwall, located at the southern end of measured triaxial data pairs from the intact rock, and
the ninth slice, was excavated in the compact marls the constants m and s were found to be 0.344 and
which lie above the coal seam with a thickness of 15± 0.0004, respectively [14]. To assess the various controls
20 m. In the failed highwall and in the entire pit, con- on slope movements, the development of mobilized
tinuous cross joints are well developed within the com- shear strength and the failure mechanism under the in-
pact marl. Except local deviations, there are three ¯uence of the loads exerted by the spoil pile were
dominant joint sets developed parallel and/or subpar- investigated by Ulusay and Aksoy [21] using determi-
226 SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES

Fig. 5. The ¯ow chart for the proposed method of analysis code HOBRSLP.
SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES 227

back analyses. A mean uniaxial compressive strength


of 4.15 MPa determined from 40 test specimens for the
compact marl, and average values of unit weight of
13 kN/m3 and 16 kN/m3 were utilized for the spoil ma-
terial (in-situ) and the compact marl, respectively.
For each cross-section, starting from an arbitrarily
chosen initial RMR value of 18 for the calculation of
the constant s, the values of the constant m and corre-
sponding RMRm which satisfy a factor of safety of
unity for the given failure surfaces are calculated. The
results of the analyses are plotted in the form of
Fig. 6. Location map of the back analyzed case study sites.
RMRm±RMRs graphs (Fig. 10). It is evident from
nistic and numerical (FEM) methods. For this pur- Fig. 10(a)±(d) that common values of RMR for the
pose, available monitoring record, structural data and constants m and s along the failure surfaces in section
groundwater information were examined, and a rock 1-1' is 51, along section 2-2' is 52 and along section 3-
mass shear strength envelope was derived from the 3' is 53. The RMR values back calculated for four fail-
Hoek±Brown criterion in conjunction with rock mass ure surfaces are equal to or nearly identical to 53 and,
classi®cation for the highwall material. Ulusay and thus they con®rm the average value of RMR (53)
Aksoy [21] back analyzed the failure utilizing four obtained from the comprehensive geotechnical logging
cross-sections and indicated that the updated Hoek± and scanline surveys performed by Ulusay [14]. Shear
Brown failure criterion used with rock mass classi®- strength values calculated for the base of each slice
cation gives strength values equal to those obtained by involved by the four failure surfaces con®rmed by the
the mobilized strength curve, and results of the back predicted surfaces (at F = 1 condition) were plotted
analyses con®rm the applicability of the loaded slope against normal stresses acting at the slice bases onto
model proposed for the case. the original failure envelope of the rock mass derived
The procedure presented herein was applied to the from the Hoek±Brown failure criterion by utilizing an
case summarized above. Taking into consideration the average RMR value of 53 (Fig. 11(a)). This compari-
loaded slope model (symmetrical vertical triangular son indicates that the mobilized strength plots match
external loading condition), the program the original failure envelope of the investigated rock
HOBRSLP [14] was modi®ed by the authors to incor- and the method proposed gives identical results to
porate external loading conditions (Fig. 8). those obtained in a previous study by Ulusay and
Considering that the predicted (based on the site ob- Aksoy [21]. The resulting curvilinear failure envelopes
servations and monitoring data) surfaces were con- with RMR values of 51±53 given in Fig. 11(b). Figure
®rmed by the calculated failure surfaces [21], four 11(b) suggest that failure envelopes for the range of
failure surfaces given in Fig. 9 were employed in the calculated normal stress levels (sn) in the back ana-

Fig. 7. Initiation of the slide in the highwall externally loaded by a spoil pile (Case 1).
228 SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES

Table 1. Range of parameters employed in rock mass classi®cations for three cases considered in the study
Parameter Range (mean)/description
case 1 case 2 case 3
Uniaxial compressive strength 1.14±6.41 (4.15) 4.20±6.15 (5.2) 35.4±44.3 (40.2)
(MPa)
RQD (%) 37±98 0 90±95
joints: 250±410 (386) bedding: 300± 30±40 310±390 (370)
Spacing of discontinuities (mm) 1000
Condition of discontinuities aperture 0±1 mm; very thin soft aperture 1±3 mm; soft in®lling; apertures <1 mm and 1±5 mm
coating; planar-smooth surfaces; slickensided surfaces; highly between bedding and joint planes,
fresh/slightly weathered; high weatered; high persistence respectively; soft coating <1 mm;
persistence smooth-slightly rough surfaces;
fresh to slightly weathered; high
persistence
Groundwater dry-damp dry dry
Adjustment for discontinuity one mode of failure ÿ5 one mode of failure ÿ5 one mode of failure ÿ5
orientation
Blasting damage adjustment smooth blasting 0.94 ÿ fair blasting 0.90
Adjustment for major plane of ÿ very close to discrete fault zones ÿ
weakness 0.7
Adjusted RMR'76 not determined 21 not determined
Adjusted RMR'89 50±62 (53) 20.6 40±47 (43)
The values given in the parentheses indicate mean value.

lyzed slope show negligible and/or slight di€erences middle and lower benches. The unit weight of the
which result from probably due to small variations in schists ranges between 17.2 kN/m3 and 28.5 kN/m3
the mechanical properties of the same rock in di€erent with a mean value of 22.2 kN/m3. The uniaxial com-
places. pressive strength of the intact rock determined on a
limited number of specimens due to the diculties in
Slope failure in a closely jointed schist rock mass at a sample preparation was 5.2 MPa. Slope failures cover-
barite open pit mine (Case 2) ing a single bench or two benches were observed at
The Baskoyak mine at the central part of Turkey is three locations in the pit. The failures were circular
an open pit mine operated for the extraction of barite. and one of them occurred in the closely jointed rock
A comprehensive slope stability project was carried mass. Back analysis of the failures indicated that the
out to determine the engineering properties of the rock calculated sliding surfaces con®rm the actual failure
mass, and to assess the failure mechanism and the surfaces delineated from the site measurements [27].
alternatives for improving the overall stability between No any sign of groundwater was encountered through
1987 and 1988, and the investigation was published by the geotechnical and previously drilled boreholes and
Ulusay and Yucel [27]. on the benches. Thus, the pit slopes was considered as
Based on the scanline surveys consisting of 90 schist- dry for stability assessments. The overburden material
osity and 160 joint measurements and geotechnical and the ore are removed by the excavators without
logging of a borehole of 75 m deep, Ulusay and any blasting.
Yucel [27] reported that the schists should be regarded The rock mass parameters of the heavily broken
as comprising two rock mass types. The ®rst type con- part of the rock mass are given in Table 1. Ulusay and
sists of a schist rock mass heavily broken by closely YuÈcel [27] declared an RMR value of 21 in their
spaced discontinuities (Fig. 12), and the second type is report based on Bieniawski's 1976 classi®cation [28].
a weathered schist in di€erent degrees both in the However, the authors of this recent study also calcu-
hangingwall and footwall, particularly observed at the lated the RMR value of the rock mass based on 1989
version of the RMR classi®cation [1] using the par-
ameters given in Table 1 for this case. In this calcu-
lation a discontinuity adjustment of ÿ5 considering
one mode of failure, mass failure, was assigned.
Because the presence of discrete fault zones running
very close to the failed slope, a major structure adjust-
ment of 0.7 [17] was also considered to obtain ®nal
RMR value. An RMR value of 20.6 which is identical
to that derived from Bieniawski's 1976 classi®cation
was obtained.
Utilizing the well delineated circular slip surface il-
lustrated in Fig. 13 and the geomechanical parameters
given above, the proposed method was applied to the
Fig. 8. The model with the parameters for the slope under the in¯u-
ence of a symmetrical vertical triangular spoil loading used in the failure occurred in closely jointed part of the schists.
back analysis (Case 1). Choosing an initial RMR value of 10 for the calcu-
SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES 229

Fig. 9. Slope pro®les, and the predicted and calculated failure surfaces employed in the back analyses for the loaded high-
wall case (Case 1).

lation of the constant s, the analysis was started. The rock mass shear strength values obtained from the fail-
pairs of RMRm and RMRs which lead a value of ure surface yielding F = 1 are plotted on the original
safety factor of unity are plotted and then the s±t curve derived from the updated Hoek±Brown cri-
RMRRM value which satis®es limit equilibrium con- terion utilizing an RMR value of 21 (Fig. 14(b)).
dition for the constants of m and s is found as 21 These results indicate that the back calculated RMR
(Fig. 14(a)). Besides, on the basis of normal stresses value and the mobilized shear strength plots match the
acting at the bottom of 10 slices in the failed mass, the RMR derived from site investigations, and the original

Fig. 10. Back analysis plots illustrating the derivation of RMRs±RMRm pairs satisfying the limit equilibrium condition for
the slope pro®les examined (Case 1).
230 SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES

Fig. 13. Slope geometry before and after failure and circular slip sur-
Fig. 11. (a) Comparison between the rock mass shear strength face in closely jointed schist rock mass (Case 2).
obtained from the back analysis and the failure envelope derived
with the Hoek±Brown criterion considering the average RMR value drilled holes in the vicinity of the investigated slope
(53) for the rock mass; (b) failure envelopes based on empirical fail-
ure criterion for mean and lower bound RMR values derived from indicated that the groundwater table lies below the
the proposed method (Case 1). failed marly rock mass. As being in the ®rst case, the
coal seam acts an aquifer, and therefore, the failed
failure envelope of the investigated rock mass. Thus, it
slope is dry. Bedding planes dip into opposite direction
is concluded that the procedure outlined above also
of the slope. The marly rock which forms the majority
yielded realistic results for this case.
of the sequence has a carbonate content considerably
higher than its clay content. The actual slip surface
A slope instability in a coal mine (Case 3)
was in circular shape, which was evident from the ®eld
As an example of the proposed method, back analy- inspection and topographical measurements carried
sis on a typical instability was carried out in out along the failure surface, and passed through the
Kisrakdere open pit mine which is located at Soma lig-
compact marl rock mass and the clay, above the coal
nite basin (see Fig. 6). The necessary geotechnical data
seam. Because the thickness of the coal seam reduces
were collected by the authors from this pit. The coal
in this part of the pit, highly steep slopes were cut to
seam is generally 20 m thick, but becomes thinner
towards the basin margins where the failed slope is extract the coal. In addition to this application, it is
located. Figure 15(a) shows the geometry of the slope concluded that the presence of a weak and soft clay,
in which a single thin coal seam with a thickness of and the jointed nature of the marly rock in the
4.5 m is overlain by a sequence consisting of compact sequence made the failure easier. Scanline surveys were
marl, and soft clay beds about 10 m thick. The obser- carried out in the close vicinity of the failed slope to
vations on the slope surfaces, measurements through collect data for the discontinuities and to assess rock
the blast-holes, and the records of the previously mass conditions. Three main joints moderately and

Fig. 12. A view from the schist rock mass heavily broken by closely spaced joints and schistosity planes at a barite open pit
mine (Case 2).
SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES 231

Fig. 14. (a) Back analysis plots illustrating the derivation of RMRs±
RMRm pairs satisfying the limit equilibrium condition for the failure
in the schist; (b) comparison between the rock mass shear strength
obtained from the back analysis and the failure envelope derived
with Hoek±Brown criterion utilizing the RMR value (21) determined
from the site investigation (Case 2).

closely spaced, and bedding planes in the marly


sequence resulted in a jointed rock mass.
In this study, Bieniawski's 1989 [1] classi®cation
scheme was used and the data for the rock mass rating Fig. 15. (a) Cross-section illustrating the geometry of the failed slope
determinations were obtained from the scanline sur- and the position of the strata; (b) RMR histogram for the marly
rock mass (Case 3).
veys carried out at twenty®ve locations in the studied
pit. The range of the rock mass parameters determined
in this study is given in Table 1. It is also noted that a shear strength parameters of the clay given in Table 1
discontinuity adjustment of ÿ5 for the case of one were assumed to be used for back analysis. The pro-
mode of failure and a blasting damage adjustment of cedure presented was applied to the failed slope by uti-
0.90 for fair blasting carried out in the compact marls, lizing data available for the site for the assessment of
which have considerable higher strength when com- shear strength parameters of the jointed marly rock
pared to those mentioned in Case 1, were considered. mass.
A histogram of RMR values based on the assess- The results are presented as a plot of RMRs vs
ment of the line survey data (Fig. 15(b)) has a normal RMRm (Fig. 16(a)). The method suggests that the
form which indicates that RMR values are concen- RMRRM value satisfying limit equilibrium condition is
trated between 42 and 44 with a mean value of 43. 42.5. The back calculated RMRRM value (42.5) con-
The geotechnical properties of the marl and the clay ®rms the actual RMR (43) previously determined by
determined by an experimental program are listed in the authors through the site investigations. It is also
Table 2. evident from Fig. 16(b) that there is a good agreement
Considering the similarities between engineering between the back calculated shear strengths at the base
behavior of the clays in this site and the clays in a of slices and the failure envelope derived from the
transition zone at Yatagan coal mine, which were back Hoek±Brown failure criterion utilizing the actual
analyzed by Ulusay and Doyuran [29], the residual RMR value of 43.
232 SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES

Table 2. Material properties employed in the black analysis of Kisrakdere open pit mine (Case 3)
Unit weight (kN/ U.C.S. (MPa) mi cp (kPa) cr (kPa) fp (8) fr (8)
Material m3)
Marl rock mass 23.7 40.2 9.04 ÿ ÿ ÿ ÿ
Soft clay 18.0 ÿ ÿ 17.7 14.9 21 18
cp, cr: Peak and residual cohesion, respectively.
fp, fr: Peak and residual internal friction angle, respectively.
U.C.S.: Uniaxial compressive strength.

CONCLUSIONS estimating the mobilized shear strength required to


explain existing states of stability.
Conventional back analysis of slope failures can
This study is based on the conventional deterministic
provide functional relations between shear strength
analysis framework. However, the procedure outlined,
parameters c and f for slopes of homogeneous ma- which is based on the Hoek±Brown failure criterion, is
terials with linear failure envelopes provided all the suitable for back calculations with a maximum of
other parameters are known. But in closely jointed three unknown parameters (RMR and the constants m
rock masses shear strength determination, particularly and s), and requires iterations. The main emphasis in
due to the scale e€ect, is very dicult. In addition, this paper is the application of the method where no
such back analyses have limited practical application procedure of direct strength or RMR measurement is
because these rocks obey a non-linear failure criterion. possible. The rock mass rating (RMR) of the rock and
In this study, the diculty of determining the shear the corresponding constants, m and s, satisfying limit
strength of such rocks and applicability of rock mass equilibrium condition can be readily obtained from a
classi®cation to rock slopes are explained and a practi- graphic representation of the possible range of sol-
utions.
cal procedure with a computer solution for the back
Three examples have been given to illustrate the ap-
analysis of failed slopes is put forward as a means of
plication of the method in practical geotechnical engin-
eering. In the application of this approach, it was
found that the back calculated and predetermined
values of RMR with the constants m and s were iden-
tical. However, it should be kept in mind that the
classi®cation systems which have been mainly devel-
oped for underground works may give unrealistic
results when applied to rock slopes if their limitations
are not well considered. Adjustment for the discontinu-
ity orientation is one of the most important question-
able parameter in the RMR system when it is applied
to rock slopes. Particularly in closely jointed rock
masses, which obey the non-linear Hoek±Brown failure
criterion, slope failures occur only in the form of a cir-
cular shape regardless of discontinuity orientation.
Therefore, in such rock masses expecting of one poss-
ible mode of failure and assignment and adjustment
value of ÿ5 for the discontinuity orientation seems to
be more realistic. This approach was also con®rmed by
the results of the stability analysis. On the other hand,
consideration of the factors such as method of exca-
vation, major planes of weakness and change in stress
which in¯uence the rock mass at a particular location
and thus an adjustment for these factors become
necessary in rock mass classi®cation applied to rock
slopes. It is also noted that the Hoek±Brown failure
criterion in conjunction with the RMR classi®cation
system is only applicable to intact rock or to closely
jointed rock masses, otherwise unrealistic results may
be obtained.
Therefore, it is reasonable to conclude that the
method seems to be a practical tool for back analyzing
Fig. 16. (a) Back analysis plots; (b) comparison between the back
analyzed shear strength and the failure envelope derived with the of slopes in jointed rock masses and to check the rock
Hoek±Brown criterion for an RMR value of 43 (Case 3). mass rating obtained from site and laboratory investi-
SONMEZ et al.: BACK ANALYSIS OF SLOPE FAILURES IN ROCK MASSES 233

gations. In other words, the method may lead to the 11. Priest, S. D. and Brown, E. T., Probablistic stability analysis of
variable rock slopes. Trans. Inst. Miner. Metall. A Miner. Ind.,
development of possible modi®cations in describing 1993, 92(10), A1±A12.
the rock mass parameters particularly for the slopes, if 12. Pender, M. J. and Free, M. W., Stability assessment of slopes in
necessary. closely jointed rock masses. Proc. Eurock'93, ed. A. Sousa and
P. Grossmann. A. A. Balkema, 1993, pp. 863±870.
A better understanding of the mechanics of jointed 13. Singh, R. N. and Gahrooee, D. R., Deterministic stability and
rock mass behavior is a problem of major signi®cance sensitivity analyses of slopes in jointed rock masses. Mining Sci.
in geotechnical engineering. The authors believe that Technol. , 1990, 10(10), 265±286.
14. Ulusay, R., Geotechnical considerations and deterministic design
the Hoek±Brown failure criterion provides a good esti- considerations for pitwall slopes at Eskihisar (Yatagan±Mugla)
mate for the shear strength of jointed rock masses. strip coal mine. Unpublished Ph.D. dissertation. Middle East
However, the authors hope that the application of the Technical University, Ankara, Turkey, 1991.
15. Hoek, E., Strength of rock and rock masses. News J. ISRM,
proposed method on various failure case histories in 1995, 2(2), 4±16.
the future may lead to provide a better tool for more 16. Singh, R. N. and Gahrooee, D. R., Application of rock mass
precise input data and to check the equations weakening coecient for stability assessment of slopes in heavily
jointed rock masses. Int. J. Surf. Mining, 1989, 3(2), 207±219.
employed by the non-linear failure criterion. 17. Kendorski, F. S., Cummings, R. A., Bieniawski, Z. T. and
Skinner, E. H., Rock mass classi®cation for block caving mine
drift support. Proc. 5th Int. Cong. Rock Mech. ISRM.
AcknowledgementsÐThe authors express their gratitude to Professor Melbourne, 1983, pp. B51±B63.
Evert Hoek of Canada, and to Professor Hasan Gercek of 18. Unal, E., Ozkan, I. and Ulusay, R., Characterization of weak,
Karaelmas University, Turkey for their valuable comments and sug- strati®ed and clay Ð bearing rock masses. ISRM Symposium:
gestions in preparing the manuscript. Eurock' 92 Ð Rock Characterization, Chester, UK, 14±17
September, 1992, ed. J. A. Hudson. British Geotechnical Society,
London, 1992, pp. 330±335.
Accepted for publication 26 November 1997
19. Ulusay, R., Ozkan, I. and Unal, E., Characterization of weak,
strati®ed and clay Ð bearing rock masses for engineering appli-
cations. Fractured and Jointed Rock Masses Conference, ed. L.
REFERENCES R., Myer, N. G. W., Cook, R. E., Goodman, and C. F., Tsang,
3±5 June, 1992, CA. A. A. Balkema, Rotterdam, 1995, pp. 229±
1. Bieniawski, Z. T., Engineering Rock Mass Classi®cation. John 235.
Wiley, 1989, 237 pp. 20. Unal, E., Modi®ed rock mass classi®cation: M-RMR System.
2. Barton, N. R., Lien, R. and Lunde, J., Engineering classi®cation The Bieniawski Jubilee Collection Ð Milestones in Rock
of rock masses for the design of tunnel support. Rock Mech., Enginering. A. A. Balkema, 1996, pp. 203±222.
1974, 6, 189±239. 21. Ulusay, R. and Aksoy, H., Assessment of the failure mechanism
3. Grimstad, E. and Barton, N. R., Updating the Q-System for of a highwall slope under spoil pile loadings at a coal mine. Eng.
NMT. Proc. Int. Symp. on Sprayed Concrete Ð Modern use of Geol. , 1994, 38(2), 117±134.
wet mix sprayed concrete for underground support, Fagernes, ed. 22. Fookes, P. G., Reeves, B. J. and Dearman, W. R., The design
Kompen, Opsahl and Berg. Oslo, Norwegian Concrete Assoc., and construction of a rock slope in weathered slate at Fowey,
1993. South-West England . Geotechnique, 1977, 27(2), 533±556.
4. Hoek, E. and Brown, T., Underground Excavations in Rock. Inst. 23. Sancio, R. T., The use of back-calculations to obtain shear and
Min. Metall. Stephen Austin and Sons, London, 1980. tensile strength of weathered rocks. Proc. Intnl. Symp. on Weak
5. Hoek, E. and Brown, E. T., Empirical strength criterion of rock Rock, 21±24 September, 1981, Tokyo, Vol. 2. 1981, pp. 647±652.
masses. J. Geotech. Eng. Div. Am. Soc. Civil Eng. , 1980, 106, 24. Hoek, E., Estimating Mohr±Coulomb friction and cohesion
1013±1035. values from Hoek±Brown failure criterion. Int. J. Rock Mech.
6. Hoek, E. and Brown, T., The Hoek±Brown failure criterion Ð a Miner. Sci. Geomech. Abstr., 1990, 27(2), 227±229.
1988 update. Proc. 15th Canadian Rock Mech. Symp. Univ. of 25. Bishop, A. W., The use of slip circle in the stability analysis of
Toronto, 1988, pp. 31±38. earth slopes. Geotechnique, 1955, 5(2), 7±17.
7. Hoek, E., Wood, D. and Shah, S., A modi®ed Hoek±Brown cri- 26. ISRM (International Society for Rock Mechanics), ISRM
terion for jointed rock masses. Proc. Eurock'92, ed. J. A. Suggested Methods: Rock Characterization, Testing and
Hudson. Thomas Telford, 1992, pp. 209±213. Monitoring, ed. E. T. Brown. Pergamon, London, 1981, 211 pp.
8. Hoek, E., Kaiser, P. K. and Bawden, W. F., Support of 27. Ulusay, R. and Yucel, Z., An example for the stability of slopes
Underground Excavations in Hard Rock. A. A. Balkema, excavated in weak rocks: Baskoyak Barite Open Pit.
Rotterdam, 1995. Earthsciences (Bull. of Earth Sciences Application and Research
9. Laubscher, D. H., A geomechanics classi®cation system for the Center of Hacettepe University), 1989, 15(2), 15±27in Turkish.
rating of rock mass in mine design. J. South Afr. Inst. Miner. 28. Bieniawski, Z. T., Rock mass classi®cation in rock engineering.
Metall., 1990, 90(10), 257±273. Exploration for Rock Engineering, Proc. of the Symp., Vol. 1, ed.
10. Romana, M., A geomechanical classi®cation for slopes: Slope Z. T. Bieniawski. Balkema, CapeTown, 1976, pp. 97±106.
Mass Rating. in Comprehensive Rock Engineering, Vol. 3, Ch. 29. Ulusay, R. and Doyuran, V., Characteristics of a multiple retro-
22, ed. J. A. Hudson. Pergamon Press, London, 1993, pp. 575± gressive failure in a coal mine in southwest Turkey. Eng. Geol. ,
599. 1993, 36(2), 79±89.

You might also like