1

Final paper22 June 2018

The Hoek-Brown failure criterion and GSI – 2018 edition

E. Hoek a*, E.T. Brown b
a North Vancouver, British Columbia, V7R 4H7, Canada
b Brisbane, Queensland, 4169, Australia

ARTICLE INFO ABSTRACT

_______________________________ ___________________________________________________________________________
Article history: The Hoek and Brown criterion was introduced in 1980 to provide input for the
design of underground excavations in rock. The criterion now incorporates both
intact rock and discontinuities, such as joints, characterized by the Geological
Strength Index (GSI), into a system designed to estimate the mechanical
behaviour of typical rock masses encountered in tunnels, slopes and foundations.
____________________________ The strength and deformation properties of intact rock, derived from laboratory
Keywords: tests, are reduced based on the properties of discontinuities in the rock mass. The
Hoek-Brown criterion non-linear Hoek-Brown criterion for rock masses is widely accepted and has
Geological Strength Index been applied in many projects around the world. While, in general, it has been
Rock mass strength found to provide satisfactory estimates, there are several questions on the limits
Uniaxial Compressive Strength of its applicability and on inaccuracies related on the quality of the input data.
Tension cutoff
This paper introduces relatively few fundamental changes, but it does discuss
Rock mass deformation modulus
many of the issues of utilization and presents case histories to demonstrate
practical applications of the criterion and the GSI system.
c 2018 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and
hosting by Elsevier B.V. All rights reserved.
____________________________________________________________________________________________________________________________________________________

1. Introduction
The Geological Strength Index (GSI) is a system of rock
The Hoek-Brown criterion was derived from the results of mass characterization that was developed to link the failure
research into the brittle failure of intact rock by Hoek (1965) criterion to engineering geology observations in the field by
and on model studies of jointed rock mass behaviour by Hoek (1994) and Hoek et al. (1995). The most complete
Brown (1970). description of the current use of the GSI and the Hoek-Brown
The brittle fracture theory published by Griffith (1924), criterion is given in a chapter entitled Rock Mass Properties
modified by McClintock and Walsh (1962) to account for in an eBook by Hoek, called Practical Rock Engineering, which
friction on sliding surfaces, formed the basis for the non- can be downloaded from [Link].
linear failure criterion for intact rock published by Hoek and The failure criterion and the associated Geological
Brown (1980a, b). This 2018 version of the criterion Strength Index have gained wide acceptance as tools for
incorporates all the modifications that have been estimating the strength and deformation characteristics of
implemented in the past 38 years, based on experience gained heavily jointed rock masses. Because of the lack of suitable
in applying this criterion to practical problems. alternatives, the criterion was adopted by the rock mechanics
_____________________________ community and its use quickly spread beyond the original
* Corresponding author
assumptions based on interlocking joint-defined blocks in
E-mail addresses: ehoek@[Link] (E. Hoek),
et_brown@[Link] (E.T. Brown) hard rock. Consequently, it became necessary to re-examine
these assumptions and to introduce new elements from time
to time to account for the wide range of practical problems to
which the criterion was being applied.
One of the early difficulties arose because many
geotechnical problems, particularly slope stability issues, are
more conveniently dealt with in terms of shear and normal
 2
stresses rather than the principal stresses used in the where 𝜎1 , 𝜎3 are the major and minor principal stresses, 𝜎𝑐𝑖
definition of the original Hoek-Brown criterion. At that time, is the unconfined compressive strength, and 𝑚𝑖 is a material
geotechnical software did not allow the incorporation of the constant for the intact rock.
constitutive relationships, including flow rules that describe Zuo et al. (2008, 2015) showed that a very similar equation
the behaviour of the rock after reaching the peak strength could be derived from an analysis of failure propagation from
predicted by the Hoek-Brown criterion. Hence, it was a penny-shaped crack in a triaxial stress field. Their equation
necessary to find equivalent Mohr-Coulomb parameters for can be written:
use with existing software. In 2018, most geotechnical
𝜎
software for stress and slope stability analysis allows the 𝜎1 = 𝜎3 + 𝜎𝑐𝑖 √ (µ𝜎𝑐𝑖 ⁄𝜅|𝜎𝑡 |) 𝜎 3 + 1 (2)
Hoek-Brown criterion to be used directly. Consequently, in 𝑐𝑖

this presentation, only the Hoek-Brown criterion is discussed where µ =tan ϕ (ϕ is the crack surface friction angle); κ is a
in detail. coefficient used for mixed mode fracture which can be
For readers who require equivalent Mohr-Coulomb derived from various approximations, such as 𝜅 = √3⁄2 for a
friction angles and cohesive strengths, a detailed discussion maximum stress criterion, with 𝜅 =1 for a maximum energy
on how these can be obtained is given in Hoek et al. (2002). It release criterion; and |𝜎𝑡 | is the absolute value of the uniaxial
is recommended that these friction angles and cohesive tensile strength.
strengths, derived from the Hoek-Brown criterion, should not Substitution of 𝑚𝑖 = (µ𝜎𝑐𝑖 ⁄𝜅|𝜎𝑡 |) in Eq. (2) results in the
be used without a tension cutoff. Hoek-Brown Eq. (1) for intact rock. Hence, the constant 𝑚𝑖
The Geological Strength Index (GSI) was extended to cover has physical meaning. As will be shown later in this paper, the
folded and tectonically sheared rock masses in a series of relationship between 𝑚𝑖 and 𝜎𝑐𝑖 ⁄|𝜎𝑡 | is important in the
papers by Hoek et al. (1998), Hoek and Marinos (2000) and application of the Hoek-Brown criterion to rock and rock
Marinos and Hoek (2000). The GSI will be discussed in detail mass failure.
in Sections 6 and 11.
For clarity, the equations given and discussed here are 3. Generalized Hoek-Brown criterion
given in total stress terms. However, as discussed by Hoek
and Brown (1997), the solution of some rock engineering The Generalized Hoek-Brown criterion for the estimation of
problems requires an effective stress approach. In this case, rock mass strength, introduced by Hoek (1990), is expressed
effective stress equivalents of the equations given here may as
be used. 𝜎 𝑎
𝜎1 = 𝜎3 + 𝜎𝑐𝑖 (𝑚𝑏 𝜎 3 + 𝑠) (3)
𝑐𝑖
2. The origin of the Hoek-Brown criterion
where 𝜎1 and 𝜎3 are the major and minor principal stresses
There is abundant evidence to show that the failure in at failure, 𝜎𝑐𝑖 is the unconfined compressive strength of the
brittle materials such as rock, concrete, ceramic and glass intact rock material and mb, s and a are rock mass material
originates from micro-cracks or flaws in the intact material. constants, given by
In rock, these flaws are typically grain boundaries or inter-
granular cracks and tensile cracks that propagate from their 𝑚𝑏 = 𝑚𝑖 . exp((𝐺𝑆𝐼 − 100)/(28 − 14𝐷)) (4)
tips when frictional sliding occurs along the flaw. 𝑠 = exp((𝐺𝑆𝐼 − 100)/(9 − 3𝐷)) (5)
Griffith (1921) proposed that tensile failure in brittle
materials such as glass initiates at the tips of defects which he 𝑎 = 1⁄2 + 1⁄6 (𝑒 −𝐺𝑆𝐼⁄15 − 𝑒 −20⁄3 ) (6)
represented by flat elliptical cracks. His original work dealt
with fracture in material subjected to tensile stress, but later where, for intact rock, the material constants are denoted by
he extended this concept to include biaxial compression 𝑚𝑖 , s = 1 and a = 0.5. GSI is the Geological Strength Index, and
loading (Griffith, 1924), obtaining a non-linear compressive D is a factor which depends upon the degree of disturbance to
failure envelope for brittle materials. which the rock mass has been subjected by blast damage and
Murrell (1958) proposed the application of the Griffith stress relaxation. Guidelines for the selection of D are
theory to rock. This suggestion was immediately discussed in Section 8.
implemented by researchers such as Brace (1964), Cook
Eqs. (4), (5) and (6) were developed to deal with rock
(1965), Hoek (1964), McClintock and Walsh (1962) and many
masses, such as that illustrated in Figure 1, comprised of
others. The early findings of this research were summarized
interlocking angular blocks in which the failure process is
by Jaeger and Cook (1969). More recent research has been
dominated by block sliding and rotation without a great deal
summarized by Andriev (1995).
of intact rock failure, under low to moderate confining
Based on this research on the non-linear Griffith failure
stresses.
criterion, Hoek and Brown (1980a, b) proposed the following
Originally the GSI term in these equations was estimated
empirical equation to fit the results of a wide range of triaxial
directly from Bieniawski’s Rock Mass Rating (RMR)
tests on intact rock samples:
classification (Brown and Hoek, 1988). The Geological
𝜎3 Strength Index (GSI) was introduced by Hoek, Kaiser and
𝜎1 = 𝜎3 + 𝜎𝑐𝑖 √𝑚𝑖 +1 (1)
𝜎𝑐𝑖 Bawden (1995) as a direct replacement for RMR.
 3

Fig. 2: Normal distributions and Uniaxial Compressive

Strength values determined from tests on cores from seven
rock types recovered during the site investigation and design
phase for the Ingula Pumped Storage Project in South Africa
(Keyter et al., 2008).

The Hoek-Brown criterion was developed to deal with

shear failure in rock. Fig. 3, which plots the results of triaxial
compression tests on Indiana Limestone by Schwartz (1964),
shows that the range of applicability of the criterion is
determined by the transition from shear to ductile failure at
Fig. 1: Interlocking blocks of very strong Panguna andesite approximately 1 = 4.0 3. Mogi (1967) investigated the
and granodiorite in the Bougainville open pit mine in Papua transition from shear to ductile failure in a wide range of rock
New Guinea for which the original Hoek-Brown criterion for types and found that the average transition is defined by 1 =
rock mass strength estimation was developed (Hoek and 3.4 3. This is a useful guide for the maximum confining
Brown, 1980a, b). pressure for triaxial testing of intact rock specimens.

4. Strength of intact rock

In Eq. (3) the unconfined compressive strength 𝜎𝑐𝑖 is the

dominant parameter which sets the scale of the rock
mass strength failure curve on a 𝜎1 versus 𝜎3 plot. The
constants mb, s and a define the shape of the curvilinear
failure plot.
At this point it is important to explain the difference
between the unconfined compressive strength, 𝜎𝑐𝑖, and the
uniaxial compressive strength (UCS) of intact rock. The UCS is
generally determined by testing several specimens without
applying a confining stress. Fig. 2 shows the distribution
curves obtained from high quality laboratory UCS tests on a
range of rock types encountered on a typical construction
project.
In developing the Hoek-Brown criterion it was recognized
that including a collection of UCS test results in a series of
triaxial test data would result in a significant bias in the curve
fitting process required to determine the constants of the
equation. Consequently, it was decided to use only the
average value for a UCS data set to represent the value of the
principal stress at zero confining stress. The triaxial data set,
including this average value, was then used in a regression Fig. 3: Limit of applicability of the Hoek-Brown criterion and
analysis to determine the unconfined compressive strength, for maximum confining pressure for triaxial tests on Indiana
𝜎𝑐𝑖, and the constant, 𝑚𝑖. Limestone.
 4
In some laboratories, triaxial tests are carried out by The combination of two failure criteria on one plot can
applying a constant confining stress and increasing the axial result in significant complications in programming for
load until the onset of shear failure is detected in the stress- numerical analyses. Therefore, it would be preferable to
strain plot. The confining stress in then increased and the simplify the resulting combination as far as possible. Hoek
axial load is again increased until the onset of the next failure and Martin (2014) proposed that, for practical rock
is detected. This stage testing process is repeated several engineering purposes, a Hoek-Brown failure envelope with a
times to arrive at a complete failure plot from a single tensile cutoff, based on the generalized Griffith failure
specimen. Since the specimen has been damaged in the first criterion theory proposed by Fairhurst (1964), can provide
loading cycle and all subsequent test stages involve damaged an effective solution. This is illustrated in the plot presented
rock, this method does not produce an acceptable peak in Fig. 4.
strength plot for intact rock. Therefore, it is recommended The tests by Ramsey and Chester (2004) and Bobich
that this type of triaxial test should not be used for (2005) are among the very few reliable triaxial data sets
determining the Hoek-Brown parameters ci and mi. which include direct tensile tests. Some suggestions on
Tensile failure (𝜎3 < 0) is not dealt with by the Hoek- testing procedures required to provide reliable data are given
Brown criterion. However, tensile failure is an important in the Appendix. As an interim measure, the following
factor in some rock engineering problems. In the context of approximate relationship between the compressive to tensile
this discussion, the most effective solution to this problem is strength ratio 𝜎𝑐𝑖 ⁄|𝜎𝑡 | and the Hoek-Brown parameter mi is
the Griffith theory which, as proposed by Fairhurst (1964), proposed:
can be generalized in terms of the ratio of compressive to
tensile strength as 𝜎𝑐𝑖 ⁄|𝜎𝑡 | as follows: 𝜎𝑐𝑖 ⁄|𝜎𝑡 |= 0.81mi + 7 (8)
If 𝑤(𝑤 − 2)𝜎3 + 𝜎1 ≤ 0 failure occurs when 𝜎3 = 𝜎𝑡 Eq. (8) is based on triaxial test data and curve fitting
If 𝑤(𝑤 − 2)𝜎3 + 𝜎1 ≥ 0 failure occurs when: estimates, listed in Table 1, and plotted in Fig. 5.

(2𝜎3 −𝐴𝜎𝑡 )+√(𝐴𝜎𝑡 −2𝜎3 )2 −4(𝜎32 +𝐴𝜎𝑡 𝜎3 +2𝐴𝐵𝜎𝑡2 )

𝜎1 = (7) Table 1: Analysis of data containing tensile values.
2
𝑤−1 2 𝜎 MPa mi 𝜎𝑐𝑖 ⁄|𝜎𝑡 | Data set
where 𝐴 = 2(𝑤 − 1)2 , 𝐵 = ( ) − 1, w = √|𝜎𝑐𝑖| + 1
2 𝑡 224 32.4 32 Granite: Lau and Gorski (1992)
600.4 18.8 22.2 Granite Aplite: Hoek (1965)
95.5 9.65 14.9 Berea sandstone: Bobich (2005)
125.5 10.6 14.4 Webtuck dolomite: Brace (1964)
516.5 8.45 13.9 Blair Dolomite: Brace (1964)
128.5 8.25 16.6 Marble: Ramsey & Chester (2004)
228 14.1 18.6 Quartzite: Hoek (1965)
1 5 10
1 7.2 12 Estimated by matching Hoek-Brown and
1 10 14 Fairhurst Generalized Griffith curves as
1 15 20 illustrated in Fig. 2.
1 20 24
1 30 32

Fig. 5: Relationship between c/|t| and mi

Fig. 4: Dimensionless plot of triaxial test data for Carrara
Marble showing the use of the Generalized Griffith theory for An example of plotting the Hoek-Brown failure curve with
tensile failure and the Hoek-Brown criterion for shear failure. a tension cut-off, is presented in Fig 6. The data for this plot
 5
were obtained from triaxial tests on specimens of Granite the Brazilian test can be regarded as an index test which must
Aplite, a uniformly fine grained intrusive igneous rock from be calibrated against direct tensile tests for each rock type.
South Africa. These tests were carried out by Brace at
Massachusetts Institute of Technology in the USA and Hoek at 5. Limits of applicability of the Hoek-Brown criterion
the Council for Scientific and Industrial Research in South
Africa. The average uniaxial compressive strength of 588 MPa Fig. 3 shows that the Hoek-Brown criterion is only
was used, with the triaxial test results, to fit the peak strength applicable for confining stresses within the range defined by
curve. The tension cutoff was calculated using Eq. (8). σ3 = 0 and the transition from shear to ductile failure.
It will be noted that, for intact rock, only two variables are A case in which the Hoek-Brown criterion does not apply
needed to define the Hoek-Brown failure envelope with a may arise when massive rock is in a state of relatively high
tension cut-off. These are the uniaxial compressive strength confinement. Kaiser et al. (2010) discuss this case in the
of the intact rock, ci, and the material parameter, mi. For context of highly stressed pillars in hard, brittle rock at depth.
intact rock, the parameter s is always equal to 1 and the In this case, it was found that the amount of rock mass
constant a = 0.5. strength degradation given by Eqs (4) and (5) for mb and s
was reduced by replacing the constants 29, in Eq (4), and 9, in
Eq (5) with higher values that Kaiser et al. (2010) related to
GSI and the confining pressure. Substituting the resulting
confinement-dependent parameters mb and s into Eq. (3) gave
a sigmoidal peak strength curve in the transition from low to
higher confinement. Importantly, in this case, higher
confinement produced rock mass strengths that were greater
than those given by the conventional application of Eqs. (3) –
(6).
A more general case in which Eqs. (3) - (6) may not apply
is in massive to moderately jointed hard rock having high
values of GSI, for example, GSI > ̴ 65. Bewick et al. (2019)
show how carefully distinguishing the failure modes of
heterogeneous hard rock specimens in laboratory uniaxial
and compression strength tests can allow the conventional
Hoek-Brown and GSI approach to strength estimation to be
modified to provide good fits to test data for massive to
moderately jointed rock. In this study, Bewick et al. (2019)
use a tri-linear envelope in preference to the standard
Hoek-Brown rock mass strength envelope to represent the
observed peak strength of massive to moderately jointed
rock at both low and high confinement.

6. The Geological Strength Index – GSI

Hoek (1994) and Hoek et al. (1995) introduced the

Geological Strength Index as a tool for collecting field
information for incorporation in Eqs. (4) - (6). This is used to
estimate the constants mi, s and a in the Hoek-Brown criterion
defined by Eq. (3). The GSI classification was set up to
address the two principal factors considered important
influences on the mechanical properties of a rock mass, i.e.,
the structure (or blockiness) and the condition of the joints.
The latest, major, revision of the index and its use in Eqs. (4)
Fig. 6: Plot of Hoek-Brown failure plot for triaxial tests on – (6) was made by Hoek et al. (2002). The basic version of the
Granite Aplite carried out by Hoek (1965) and Brace (1964). GSI chart, for use with jointed rocks, is reproduced in Fig. 7,
from Hoek and Marinos (2000).
Note that the Brazilian test, in which tensile failure is induced The Hoek-Brown failure was originally developed on the
as a centre of a diametrically loaded disc specimen, is not an assumption that intact rock is free from microcracks and
acceptable direct tensile test for inclusion in the analysis flaws. The GSI system was developed to deal with rock masses
described above. Due to the complex stress distribution and comprised of interlocking angular blocks in which the failure
the influence of the stress concentrations at the loading process is dominated by block sliding and rotation without a
points, the calculation of the tensile strength requires great deal of intact rock failure.
significant correction (Perras and Diederichs, 2014). At best, Figs. 8 to 12 show typical applications of the GSI chart to
exposed faces in a range of rock formations. The original
 6
purpose of the GSI chart was to provide a guide for the initial
estimation of rock mass properties. It was always assumed
that the user would improve the initial estimates with more
detailed site investigations, numerical analyses and back
analysis of the tunnel or slope performance to validate or
modify these estimates.

Fig. 9: Orthogonal jointing in

granitic rock on a dam site. GSI
is not applicable on this scale
since the stability of the
exposed face is controlled by
the geometry of intersecting
joints. It can be applied to
larger scale excavations.

Fig. 10: Interlocking angular

Andesite blocks defined by
several joint sets, exposed in
an open pit mine bench. GSI is
fully applicable in this
situation and on this scale.
Fig. 7: Basic GSI chart, Hoek and Marinos (2000).

Fig. 8: Spalling in the

sidewalls of a raise-bored Fig. 11: Complex folding in a
shaft in intact hard rock bedded sedimentary deposit.
subjected to anisotropic GSI is applicable with care
horizontal stresses. GSI is not since averaging of the intact
applicable in the analysis of properties is required to
these stress-induced spalls calculate rock mass properties.
but it can be used for other
applications.
 7
the rock mass modulus 𝐸𝑟𝑚 (MPa) was proposed by Hoek and
Diederichs (2006):
𝐸𝑟𝑚 = 105 (1 − 𝐷⁄2)/(1 + exp(60 + 15𝐷 − 𝐺𝑆𝐼 ⁄11)) (10)

Fig. 12: Tectonically deformed

sediments with almost
complete loss of structural
patterns. Care is required in
using GSI in this type of rock
mass. Use the GSI charts by
Marinos et al (2005) and
Marinos (2017) .

In dealing with the tectonically disturbed rock masses,

illustrated in Figs. 10 and 11, the original GSI chart is
adequate for estimates during the site investigation stage.
However, during later design stages, it becomes more difficult
to apply this chart effectively unless observations and Fig. 13: Plot of normalized in situ rock mass deformation
measurements of the rock mass behaviour in response to modulus from China and Taiwan against Hoek and Diederichs
excavation are available to provide a basis for calibration. Eq. (13). Each data point represents the average of multiple
To simplify this problem, Marinos and Hoek (2001) tests at the same site in the same rock mass.
published a GSI chart for heterogeneous and tectonically
deformed sedimentary rocks. An extended version of this Fig. 14 gives a comparison between the deformation
chart was published by Marinos (2017). Additional charts for modulus estimated from Eq. (10) and a number of field
ophiolites (Marinos et al., 2005) and tectonically undisturbed measurements and predictions by Read et al. (1999), Serafim
molassic rocks (Hoek et al., 2006) were also developed to and Pereira (1983), Bieniawski (1978), Stephens and Banks
cover tunnelling projects in northern Greece. (1989) and Barton (2002). The general agreement between
these results suggests that all these predictions, including
7. Estimating rock mass Deformation Modulus those of Hoek and Diederichs (2006), can be used with
confidence for estimating field values.
In addition to the estimate of the strength of intact rock and
rock masses, the analysis of the behaviour of a slope,
foundation or tunnel also requires an estimate of the
deformation modulus of the rock mass in which these
structures are excavated. This is a significant challenge and
numerous authors have presented suggestions on how these
estimates can be made.
Hoek and Diederichs (2006), using a database of rock mass
deformation modulus measurements from projects in China
and Taiwan, provided by Chern (2005), proposed the
following equation for estimating rock mass modulus:

𝐸𝑟𝑚 = 𝐸𝑖 (0.02+ (1- 𝐷⁄2)/(1+ exp(60+15D -𝐺𝑆𝐼 ⁄11)) (9)

where 𝐸𝑖 is the intact rock deformation modulus (MPa),

D is the disturbance factor introduced in Section 3, and
GSI is the Geological Strength Index.

Hoek and Diederichs (2006) recommend that, when

laboratory measured values for 𝐸𝑖 are not available, the rock
mass reduction values (MR) proposed by Deere (1968) can be
used for estimating the intact rock modulus. When no
information on the intact rock deformation modulus is
available, the following alternative equation for estimating Fig. 14: Comparison between field measurements and
deformation modulus values predicted by several authors.
 8
Cai et al. (2004) carried out a detailed review of the mass failure. In this case, a damage factor of D = 0.5 is
application of the GSI system for the estimation of rock mass considered appropriate for the rock mass in which the lower
strength and deformation properties in two underground half of the tunnel is excavated. Note that this damage factor
powerhouse projects in Japan. In their conclusion they state: should only be applied to a zone of about 2 m width around
“The GSI system was applied to characterize the jointed the bottom half of this 12 m span tunnel.
rock masses at Kannagawa and Kazunogawa underground An example of a very poorly designed and executed tunnel
powerhouses in Japan. Based on the estimated GSI values and blast is shown in the third illustration in Table 2. Poor
intact rock strength properties, equivalent Mohr–Coulomb drillhole alignment control and lack of attention to the blast
strength parameters and elastic modulus of the jointed rock design and detonation sequence has resulted in damage to the
mass were calculated and compared to in situ test results. The rock walls. A damage factor D = 1, with a linear decrease to
Point Estimate Method was applied to approximate variance zero, has been assigned to the first 3 m of the rock mass
of the mechanical properties of the jointed rock masses. It is surrounding this 8 m span tunnel.
found that both the means and variances of c, , and E The fourth illustration in Table 2 shows a 15 m high slope
predicted from the quantified GSI approach are generally in in a dam spillway in which pre-split blasting has been used to
good agreement with field data. Hence, the quantitative create the face on the left. Relaxation of the face can still occur
approach added to the GSI system provides a means for and a blast damage factor D = 0.5 has been assigned to 1 to 2
consistent rock mass characterization and thus improves the m of rock behind this face. The rock mass on the right has
utility of the GSI system.” been mass blasted with little control of the drillhole spacing
and alignment of the charges and detonation sequence. The
8. Disturbance Factor D most severe damage factor of D = 1.0 has been assigned to 2
to 3 m of the rock mass behind this slope.
When tunnels, slopes or foundations are excavated in rock The final illustration in Table 2 shows a very large open pit
masses, the removal of the rock results in stress relief which with slopes approaching 1000 m in total height. Several
allows the surrounding rock mass to relax and dilate. The aim different damage factors must be considered in this example.
of any good design is to control this dilation, and the It is important to differentiate between slopes created during
consequent displacements, to minimize failure. This can be active mining and the final design slopes which are required
achieved by a careful selection of excavation shape, method of to remain stable for many years. During active mining, the
excavation and, if necessary, the installation of reinforcement blasting is required to produce large volumes of uniformly
and support. In many cases, drainage of the rock mass is also fragmented ore to meet the requirements of ore processing
an important factor in maintaining the stability of the for mineral extraction. On the other hand, the final slopes are
excavation. required to remain stable to ensure access to the ore and safe
Table 2 sets out several examples in which the method of and efficient disposal of the waste.
excavation and the control of blasting are of great importance. Individual 18 m high benches will generally have suffered
In the case of tunnels, this is particularly important since the significant damage because of their proximity to the
limited amount of space available in a tunnel means that any production blasts required for removal and fragmentation of
failure can have a serious impact on the excavation schedule the ore. A damage factor of D = 1.00 is assigned to the rock
and cost and even on the performance of the final tunnel. immediately behind these benches. This damage factor can be
Careful excavation by a well-chosen tunnel boring machine or graded downward, to a final value of D = 0, as the distance
road-header can reduce many of these problems. However, in behind the face increases to about 30% of the slope height.
a drill and blast tunnel, the blasting design and execution is of The inter-ramp and final slopes will have suffered blasting
critical importance. and stress relaxation damage, but it is not appropriate to
A common error is to assume that the disturbance factor apply the overall damage factor D = 1.0 to a layer parallel to
D should be applied to the entire rock mass in which the slope. Rose et al (2018) state that “Selection of an appropriate
excavation is created. This will result in an extremely range of depth or stress defining the disturbance transition
conservative and inappropriate design. requires consideration of whether slope stability conditions
The first illustration in Table 2 shows a tunnel in which the are dominated by geologic structure, rock mass conditions,
blast-hole pattern, explosive charges and the detonation groundwater, in situ stresses, slope geometry, poor blasting,
sequence have all been carefully designed and executed. Of or a combination of these factors.” They have developed a
importance is the careful control of the drillhole alignment for Disturbance Rating for open pit mine slopes which can
the ultimate smooth blast used to create the tunnel walls. In provide guidelines for the decrease in the damage factor D
this case, the disturbance factor D = 0 can be used with over a range of slope heights.
confidence since there is minimal damage to the surrounding While much smaller blasts are used for slopes for
rock mass. roadcuts, dam spillways and foundation excavations, the
A more complex situation is illustrated in the second application of the damage factor should be similar to that
photograph in Table 2, showing a tunnel excavated by the top applied in open pit mining. However, the overall factor of
heading and bench method. Unless the displacements safety of the design may be higher than that for open pit mine
induced by the excavation of the lower bench are controlled slopes to accommodate the longer life expectancy.
by the placement of an invert strut, excessive displacements
in the lower part of the tunnel can result in significant rock
 9

Table 2: Guidelines for estimating disturbance factor D due to stress relaxation and blasting damage

The damage factor D should never be applied to the entire rock mass surrounding an excavation

Appearance of rock mass Description of rock mass Suggested value of D

Excellent quality-controlled blasting or excavation by a

road-header or tunnel boring machine results in
minimal disturbance to the confined rock mass
surrounding a tunnel. D=0
The blasting design for this tunnel is discussed in
[Link]/assets/resources/learning/hoek
/Practical-Rock-Engineering-Chapter-16-Blasting-
[Link]

Mechanical or manual excavation in poor quality rock D=0

masses gives minimal disturbance to the surrounding
rock mass.
Where squeezing problems result in significant floor D = 0.5 with no invert
heave, disturbance can be severe unless a temporary
invert, as shown in the photograph, is placed.

Poor control of drilling alignment, charge design and D = 1.0 at surface with a
detonation sequencing results in very poor blasting in linear decrease to
a hard rock tunnel with severe damage, extending 2 or D = 0 at ± 2 m into the
3 m, in the surrounding rock mass. surrounding rock mass

D = 0.5 for controlled

Small scale blasting in civil engineering slopes results presplit or smooth wall
in modest rock mass damage when controlled blasting blasting with D = 1.0 for
is used, as shown on the left-hand side of the production blasting
photograph. Uncontrolled production blasting can
result in significant damage to the rock face.

In some weak rock masses, excavation can be carried D = 0.7 for mechanical
out by ripping and dozing. Damage to the slopes is due excavation
primarily to stress relief. D = 1.0 for production
Very large open pit mine slopes suffer significant blasting, decreasing to
disturbance due to heavy production blasting and zero over a distance of up
stress relief from overburden removal. to 45% of the slope
*A Disturbance Rating for open pit slopes has been height, depending upon
published by Rose et al (2018). the Disturbance Rating*
 10

9. The overall design process 10. Determination of intact rock strength properties

Having set out all the input data required for a full The starting point for the procedure outlined in the
analysis, using the Hoek-Brown and GSI system, it is flow chart in Fig. 15, is the determination of the intact
useful to consider the full sequence of data acquisition, rock properties. This involves laboratory uniaxial and
interpretation, utilization and back analysis. Fig. 15 is a triaxial tests on carefully collected and prepared rock
flow chart in which the sequence of data acquisition from core samples. Generally, care is taken to ensure that the
laboratory tests and field observations are combined to core is recovered from homogeneous rock in which
calculate the principal stress relationship for a rock mass. failure will occur through intact rock material. These
This is followed using analytical or numerical models to samples are tested using current standard and suggested
produce an excavation design which is then methods outlined in the ISRM Suggested Methods
implemented, and its performance monitored by (Ulusay and Hudson, 2007).
convergence measurements. When the Hoek-Brown criterion was introduced it
A final step is the back analysis of the monitoring was recommended that triaxial test results should be
results and the feed-back of the results of this analysis analysed by linear regression of the following version of
into the early stages of the flow chart. This step is critical Eq. (1) (Hoek, 1983):
since it is the only means whereby the design method and
(𝜎1 − 𝜎3 )3 = 𝑚𝑖 𝜎𝑐𝑖 𝜎3 + 𝜎𝑐𝑖2 (11)
the input parameters used in the calculations can be
validated. Back analysis should be an ongoing process This approach was used for several years until it was
throughout and even after the construction process so realized that the method was inadequate for the analysis
that adjustments and corrections can be made at all of data other than closely spaced points with very little
stages. This provides not only confidence in the design scatter about a general trend line. A variety of methods is
but also information which can be used to improve on the available for fitting curves through non-uniform
determination of input parameters and the design distribution of triaxial test data. One of these, known as
methodology. the Modified Cuckoo search (Walton et al., 2011), is
included in the Rocscience program RocData which can
be used for the interpretation of laboratory test data.
Bozorgzadeh et al. (2018) and Contreras et al. (2018)
used Bayesian statistics to quantify the uncertainty of
intact rock strength. This approach provides an
alternative to conventional probabilistic or frequentist
methods such as those described above. To deal with the
problem of outliers in sets of test data for rock, Contreras
et al (2018) use Student’s t distribution in place of the
commonly assumed normal distribution as a starting
point in the analysis. The difference between these two
distributions, for a hypothetical, but not unrealistic, data
set is illustrated in Fig. 16 in which the impact of a single
outlier is evident.

Fig. 15: Flow chart for the application of the Hoek-

Brown and GSI system to an excavation design. Fig. 16: Comparison between a normal distribution and
Student’s t distribution for the analysis a small data set
with an outlier (after Kruschke, 2015).
 11

specimens large enough to include representative

sections containing these defects, but collection and
preparation of such specimens can be challenging.
Bewick et al. (2015, 2019) and Kaiser et al. (2015)
have examined the issue of veins and microfractures in
intact rock core or blocks in considerable detail. Their
emphasis is on the effect of these veins and micro-
fractures on rock mass classification and rock block
strength. They also have a significant influence of the
determination of the intact rock unconfined compressive
strength. Fig. 18, reproduced from the paper by Kaiser et
al. (2015) shows a good example of a block of rock
containing veins and other defects.

Fig. 18: A rock mass with joint bounded blocks that

contain veins and other defects. Reproduced from Kaiser
et al. (2015).

Following Kaiser et al. (2019), a specimen of intact

rock will be taken to be a sound piece of rock material,
Fig. 17: Analysis of the triaxial tests on Coburg Limestone typically obtained by coring without breaking during
using a Bayesian analysis incorporating Student’s t drilling. It may be fissured, veined and/or damaged at the
distributions, compared with an analysis using the micro-scale but the specimen remains intact. Kaiser et al.
RocData Modified Cuckoo method. (2019) further distinguish between intact homogeneous
and intact heterogeneous rock.
Fig. 17 is a plot of the results obtained from a Bayesean In discussing rock mass classifications, such as GSI,
analysis of a triaxial data set, giving an unconfined Day et al. (2012) define the joints forming blocks as
compressive strength of 𝜎𝑐𝑖 = 114.5 MPa and 𝑚𝑖 = 11.5. interblock structures. They define the veins, stockwork
For comparison, the result given by the RocScience and other defects as intrablock structures and point out
RocData Modified Cuckoo non-linear regression analysis, that these should also be considered in the rock mass
with absolute residuals, is 𝜎𝑐𝑖 = 116.2 MPa and 𝑚𝑖 = 10.6, characterization since they have a significant influence
which has also been plotted in Fig. 17. In this case, the on intact rock strength. They suggest that the defects in
differences between the Bayesian analysis and the non- both the interblock and intrablock structures can be
linear regression analysis are not large. As Bozorgzadeh incorporated into the GSI classification on a similar basis.
et al (2018) demonstrate, the advantages of their novel Bewick et al. (2019) point out that the intrablock
Bayesian regression analysis technique become more structures can serve to reduce the block strength below
apparent for sparser data sets. the intact rock strength.
In the preceding discussion it has been assumed that Day et al’s (2012) suggestion is illustrated in Fig. 19 in
the rock specimens are homogeneous and isotropic and which the influence of size is considered in determining
that the values of the unconfined compressive strength the use of GSI. The starting point for this chart is a typical
𝜎𝑐𝑖 and the constant 𝑚𝑖 are representative of the intact intact rock core, but there is no reason why this starting
rock in the blocks of the rock mass. In fact, this point should not be the intrablock structure within the
assumption is not always valid since in many rock core as suggested by Day et al. (2012). They emphasize
masses, defects such as veins, micro-fractures and that the reduction of the intact rock strength by this
weathered or altered components can reduce the intact method must be carried out with care to avoid over-
rock strength. Ideally, tests should be carried out on penalization of the rock mass strength.
 12

Fig. 20: Core of sandstone, siltstone and mudstone

immediately after recovery (left) and after several
months of storage in a core shed (right).

The GSI characterization scheme was devised for

engineering geologists and geologists who can utilize all
the information contained in the chart presented in Fig. 7
Fig. 19: Size effects in rock mass characterization. to arrive at a range of probable GSI numbers for each rock
Modified after Hoek and Brown (1980a). unit. As for the case of triaxial test data obtained from
laboratory testing, discussed in Section 10, the range of
Weathering, alteration and the deterioration of the GSI values should also be treated as a distribution.
core in storage are factors that need to be considered Langford and Diederichs (2015) and Contreras and
during collection and preparation of specimens. An Brown (2018) advocate that the use of the same
example of the deterioration of mudstones and siltstones statistical processes should be applied to both the intact
due to changes in moisture content during storage is rock properties and the GSI estimates to provide ranges
illustrated in Fig. 20. In such cases, care needs to be taken of the final rock mass properties chosen for design.
to seal the core during transportation and storage or, in Many projects have been completed successfully
extreme cases, to carry out the strength tests on site as using a deterministic approach in which mean values for
soon as possible after core recovery. In this project, intact rock properties and GSI are chosen and applied to
immediate sealing of excavated surfaces with shotcrete the design process outlined in Fig. 15. This approach is
was necessary to preserve the rock mass strength. acceptable when it is associated with a well-planned
The triaxial cell, illustrated in Fig. A1 in the appendix, rigorous back analysis program and where the contract
was originally designed to permit triaxial testing of rock can accommodate the changes which are necessary to
specimens, such as those illustrated in Fig. 20, to be utilize the information from this back analysis. Examples
tested on drilling sites. of this type of approach are presented in the next
sections.
11. Practical application of the GSI characterization The GSI system assumes that, because the rock mass
is made up of a sufficiently large number of joint sets and
The starting point for any site investigation program is a randomly oriented discontinuities, it can be treated as a
good geological model of the site. Ideally, this model homogeneous and isotropic mass of interlocking blocks.
should be constructed by local geologists who have Failure of this rock mass is the result of sliding along
familiarity with the regional geology and experience in discontinuities or rotation of blocks, with relatively little
working with the rock types encountered on the site. failure of the intact rock blocks. The ideal rock mass for
Without such a model, the application of GSI can become which GSI was originally developed is a heavily jointed
a confusing array of numbers being manipulated by rock mass with high intact rock strength, such as that
engineers anxious to obtain input for analytical or illustrated in Fig. 1.
numerical models. Fig. 19 shows that the ratio of the size of the blocks to
the size of the structure in which they exist is a very
important factor to be considered when deciding
 13

whether GSI should be used. For example, in the face of a available for calculating the factors of safety of sliding
10 m span tunnel, an average joint spacing of 0.5 m would blocks or wedges.
result in about 400 blocks being exposed in a square mine Many of the applications and limitations of the GSI
tunnel or about 315 blocks in a circular tunnel. This were discussed by Marinos and Hoek (2000) and Marinos
would be considered a reasonable scale for the et al. (2005). Users who are not already familiar with the
application of GSI. The same GSI rating would be applied GSI system are advised to read these papers before
to smaller blocks with similar geometry. It is the shape of embarking on applications in the field. The following
the blocks and the characteristics of the discontinuities three case histories have been chosen to illustrate the
which separate them, rather than their size, that controls practical application of the Hoek-Brown criterion and the
their interlocking behaviour. In this example, joint GSI system in a variety of geological environments and
spacings of 2 m or more would result in fewer than 25 project settings.
blocks which, as shown in Fig. 9, would result in the
failure of individual blocks rather than the overall failure 12. The Driskos Tunnel on the Egnatia Highway
of a jointed rock mass. GSI should not be used in this case.
In a 100 m high rock slope, a blocky rock mass with The 670 km long Egnatia Highway across northern
an average joint spacing of 3 m would expose about 1000 Greece has a total of 77 twin tunnels of almost 100 km
blocks in a 100 m length of the slope. This would qualify total length. These 12 m span tunnels pass through
for a condition in which GSI could be applied. On the other complex geological conditions in a converging rim
hand, 15 m high benches in the same rock mass would not between the European and African plates. Many
qualify since only about 25 blocks would occur in a 15 m unfavourable geotechnical environments occur along the
length of the slope. highway route leading to difficult tunnelling conditions.
In cases where GSI is not applicable, the failures will One of the tunnels on this route is the Driskos tunnel,
be controlled by the three-dimensional geometry of the which will be discussed in this example.
intersecting features in the rock mass. Stability analysis
in these cases should be carried out using tools that are

Fig. 21: Longitudinal profile along the Driskos tunnel depicting the geological formations encountered and the
predicted percentage closure strain in these sections. Adapted from Vlachopoulos et al. (2012).
 14

Between 1998 and 2006, Dr Evert Hoek and Professor The relationship, proposed by Hoek et al (2002), is
Paul Marinos formed a Panel of Experts to advise Egnatia used to calculate the strain for different ratios to rock
Odos A.E., the company set up to manage the construction mass strength to in situ stress shown in Fig. 23.
of the project, on geotechnical issues related to tunnel
design and construction. In 2000, they reviewed the
design of the 4.6 km long Driskos tunnel. A longitudinal
profile along the tunnel, depicting the geological
formations, is presented in the upper half of Fig. 21.
Based on their knowledge of the regional geology of
the area and the site investigations that had been carried
out, they estimated the GSI values along the tunnel route
and calculated the percentage strain which could be
anticipated. These percentage strains are plotted along
the tunnel in the lower graph in Fig. 21.
The largest strains were anticipated in a section of
very poor-quality flysch at the deepest central section of
the tunnel. A typical outcrop of this flysch, a tectonically
deformed sequence of sandstones, siltstones and
mudstones, is illustrated in Fig. 12.
Hoek and Marinos (2000) developed a method for
estimating the strain, defined as the ratio of tunnel
closure to tunnel diameter x 100, for a tunnel subjected
to in situ stresses sufficiently high to cause squeezing. Fig. 23: Percentage strain as a function of the ratio of rock
To carry out the calculations of strain, an estimate of mass strength to in situ stress, Hoek and Marinos (2000).
the rock mass strength is required, and this can be made
using the approximation given for line 4 in Fig. 22. A In the case of the Driskos tunnel, the in-situ stress 𝑝0 is
comparison between this estimate and estimates made assumed to equal the product of the depth of the tunnel
by other authors and in situ test results shows acceptable and the unit weight of the rock mass. The calculated
agreement for values of GSI up to 65. percentage strains, for the lowest and highest GSI
estimates, are plotted along the tunnel alignment in Fig.
21, which shows that strains of the order of 10% were
anticipated, for the lowest GSI values, for a section of the
Driskos tunnel between approximate chainages of 8300
to 9000. During the tunnel construction, significant
strains occurred in the tunnel in this zone and the
installed steel sets, rockbolts and shotcrete proved to be
inadequate to prevent the deformation from encroaching
on the space required to accommodate the final lining.
Additional tensioned cables had to be installed to provide
the support required to stabilize the tunnel.
A comprehensive retrospective analysis of the Driskos
tunnel design and construction issues is given by
Vlachopoulos et al. (2012).

13. The Ingula underground powerhouse project

The Ingula Pumped Storage Project in South Africa

comprises two reservoirs interconnected by a tunnel
system, with reversible pump/turbine units with a total
rated generation capacity of 1332 MW located in an
underground powerhouse complex. This complex
consists of a 26 m span machine hall, a transformer hall
Fig. 22: Approximate relationship between the ratio of with a 19 m span, 11 m diameter busbar tunnels, 5 m
rock mass to laboratory unconfined compressive diameter high pressure penstocks, a 9 m diameter main
strength for a range of RMR or GSI values. access tunnel and a series of smaller adits and shafts. It is
 15

located at a depth of almost 400 m below ground level. implemented in the construction of the underground
The 184 m long machine hall has a double curvature complex.
profile roof with a relatively low span: height ratio of 2.5 Determinations of the deformation moduli of the in-
and is up to 50 m deep in the turbine pits. situ rock masses, in the Drakensberg project, were
The Ingula power caverns were constructed under a carried out by means of plate bearing tests and by back
prominent mountain ridge off the Drakensberg analysis of the deformations measured in the full-scale
escarpment between the Free State and KwaZulu Natal tests described above. A surprising result was that the in-
provinces, South Africa, in the Volksrust Formation of the situ deformation moduli were close to the values
Ecca Group, Karoo Supergroup which comprises determined from laboratory tests on intact samples. This
horizontally bedded siltstones, mudstones and suggested that, under the confined stress conditions
carbonaceous mudstones. The intact rock uniaxial around the excavations, the rock masses were behaving
compressive strength properties derived from field and as very tightly interlocked blocky structures which, in
laboratory testing are presented as normal distributions today’s terms, would have to be assigned a very high GSI
in Fig. 2. value. Note that the tendency of the mudstones and
In situ stresses were measured in hydro fracture tests siltstones, illustrated in Fig. 20, to disintegrate upon
in boreholes and in a small number of overcoring tests. prolonged exposure to air, was remedied by the
The major horizontal stress is greater, and the minor immediate application of shotcrete to all excavated faces.
horizontal stress slightly lower, than the estimated This sealed the rock masses from exposure and
vertical overburden stress. Hydro fracture tests at cavern preserved the intact properties very effectively.
level gave a horizontal/vertical stress ratio of between In designing the underground caverns for the Ingula
0.5 and 0.9 while overcoring tests indicated a ratio of project, the rock mass behaviour in the Drakensberg
approximately 1.0 in the powerhouse area. project was considered, and the interbedded mudstones
In the design of the Ingula underground powerhouse and siltstones were treated as intact rock with weak
complex, the conventional deterministic method of horizontal bedding planes. Since this was an unusual
combining the Hoek-Brown 𝜎𝑐𝑖 and 𝑚𝑖 parameters was design assumption, the designers were reluctant to
used, with GSI values determined in the field to estimate assign a GSI value of 100 to the siltstone and mudstone
rock mass strengths for different rock units. The final units. Therefore, it was decided to use GSI = 70 in the
successful design and construction of the excavations is design process.
well documented in papers by Keyter et al. (2008) and Two-dimensional finite element models, set up during
Kellaway et al. (2010) which provide an excellent record the detailed design of the cavern excavation and support,
of this case history. were revised towards the end of the main power cavern
Initial geological and geotechnical investigations for to account for the actual geology encountered, the
the Ingula project commenced with borehole drilling in excavation sequence and support installation and the
1999. The main phase of the surface geotechnical convergence information collected during construction.
investigation and the drilling of additional boreholes and The results of these analyses confirmed that, in fact, the
an exploration tunnel was completed in 2005. in-situ mudstones and siltstones should have been
The preliminary design of the powerhouse complex assigned a GSI value of 100. Table 3 lists some of the
was based on the experience gained during the design details in the comparison between the design
and construction of the Drakensberg Pumped Storage assumptions and the results of the back-analysis,
Project, a sister scheme commissioned in 1981 (Bowcock reproduced from Kellaway et al. (2010).
et al., 1976). At the time of the design of this project, This example illustrates the fact that, in many cases,
worldwide experience in the construction of engineers tend to under-estimate the capacity of rock
underground powerhouse complexes was very limited masses when tightly confined by the stress field
and, after a careful review of three published case surrounding underground excavations. Kaiser et al.
histories, the trapezoidal roof arch in the Poatina Project (2015) have examined this issue in detail for highly
in Tasmania, commissioned in 1964 (Endersbee and stressed brittle rocks. They conclude that:
Hofto, 1963) was adopted for the Drakensberg project. “Common use of currently available rock mass
Since the Drakensberg powerhouse cavern was the first characterization systems tend to underestimate the
major underground civil engineering cavern to be strength of highly stressed brittle and often defected
constructed in South Africa, full scale tests of the cavern rock. It is demonstrated that this is primarily related to
arch and the concrete lined pressure tunnel were carried flawed interpretation of rock mass characteristics
out to confirm the support design assumptions. These derived from boreholes and laboratory tests without
tests were very successful and provided the information proper consideration of, for example, GSI applicability,
required to complete the design which was successfully laboratory test results failure mode sorting, and failure
modes of rock in underground settings.”
 16

Table 3: Parameter calibration of Units 1 and 2. Similar comments can be made for weaker rocks, such
as the mudstones and siltstones discussed in the example
Model parameter Original design Calibrated
value model value
of the Ingula project. In particular, the tendency for these
rocks to slake when removed from the in-situ
Geological Strength Index GSI 70 100
environment, can lead to significant under-estimation of
Ratio of horizontal to vertical in situ stress 1.0 0.7
the rock mass properties (see Flores and Catalan, 2019).
Intact rock modulus, Ei mudstone: 19.0 GPa 21.0 GPa
siltstone/mudstone: 22.5 GPa 25.0 GPa
In the case of the Ingula project, the back analysis of a
Joint shear stiffness, Ks bedding: 1 330 MPa/m 2 500 MPa/m
carefully investigated and well-designed project,
jointing in Karoo strata: 1 490 MPa/m 2 500 MPa/m illustrated in Fig. 24, provides a valuable example of the
Joint normal stiffness, Kn bedding: 9 750 MPa/m 15 000 MPa/m additional information that be can be gained on
jointing in Karoo strata: 6 490 MPa/m 15 000 MPa/m completion of the project. As emphasized by Sakurai
Bedding strengths: peak strength: ϕ 28.5⁰ 28.5⁰ (2017) in his book, Back Analysis in Rock Engineering:
c 0.12 MPa 1.45 MPa “Field measurement data are only numbers unless they
𝜎𝑡 0 MPa -0.64 MPa are properly interpreted. Therefore, the most important
residual strength: ϕ 24.2⁰ 28.5⁰
c 0 MPa 0.12 MPa
aspect of field measurements is the quantitative
𝜎𝑡 0 MPa 0 MPa interpretation of measurement results”.
Joint strengths peak strength: ϕ 23.6⁰ 23.6⁰
(for jointing in Karoo strata) c 0.1 MPa 4.5MPa 14. Chuquicamata mine slope stability analysis and
𝜎𝑡 0 MPa -1.93 MPa conveyor transfer chamber design
residual strength: ϕ 22.2⁰ 23.6⁰
c 0 MPa 0.1 MPa
𝜎𝑡 0 MPa 0 MPa The Chuquicamata mine in northern Chile has one of
the largest open pits in the world, measuring
approximately 4 km long, 3 km wide and 1 km deep.
Removing ore and waste from the mine on conveyors or
by truck, using the haul roads such as that illustrated in
Fig. 25, is a complex and expensive process. Hence,
planning started more than 10 years ago for a transition
from open pit to block caving underground as the mining
method (Olavarría et al., 2006). The transition is currently
scheduled to occur in 2019.

Fig. 24: The 26 m span, 50 m high Ingula powerhouse Fig. 25: The east slope of the Chuquicamata mine in 2013.
cavern nearing completion. Photograph provided by G.
Keyter and reproduced with permission of ESKOM South
Africa.
 17

For many years the ore has been transported to the The second important component of the analysis was
surface by means of a conveyor installed in a tunnel the existence of a very sophisticated slope displacement
behind the East Wall slope. The conveyor has been monitoring program based on more than 1000 prisms
extended downwards as the depth of the pit increased located in sensitive areas of the pit, measured
and, due to limits in conveyor belt lengths, a transfer automatically at frequent intervals by electro-optical
station was installed in the conveyor tunnel in 2005. measuring devices. Information is telemetered to a
Progressive deepening of the open pit has resulted in central monitoring station for interpretation. The
ongoing displacements in the East Wall and in the rock locations of the most important prisms around the
mass surrounding the conveyor transfer chamber. This entrance of the access tunnel to the conveyor transfer
resulted in the need for detailed monitoring of the cavern station are shown in Fig. 26.
deformations and periodic adjustment of reinforcement In addition, several radar displacement monitoring
cable tensions and, in some cases, installation of units, such as that illustrated in Fig. 27, are available on
replacement cables. It is important that this chamber the mine. One of these was deployed to monitor the
remains stable until it is decommissioned when the open displacements of the slope in which the transfer chamber
pit mining is completed. is located. A radar image of displacements in the east face
In 2012, a review of the conveyor transfer chamber of the mine is reproduced in Fig. 28.
was set up by the mine management. This review was
monitored by E. Hoek, a member of the mine’s Technical
Advisory Board. The detailed analysis was carried out by
P. Varona of Itasca and F. Duran of the Chuquicamata
Geotechnical Department.
An important component of this analysis was the
establishment of the rock mass model to be used in
numerical models of the slope and chamber. This was
based on the results of a geotechnical characterization
program initiated by E. Hoek and J. Read, members of the
first Technical Advisory Board established in 1992. This
program involved laboratory testing of intact samples
and joints in the 7 major rock types surrounding the open
pit, as well as 185 km of borehole core logging and 195
km of bench mapping. The results of this geotechnical
characterization program, agreed upon by the mine’s
geotechnical department and approved by the Technical
Advisory Board, are summarized in Table 4.
Fig. 26: View of the east face of the Chuquicamata mine
showing the location of the conveyor transfer chamber,
Table 4: Rock mass and discontinuity properties
major structural features and the location of critical
optical distance measurement targets (circled in red).

Fig. 27: Radar equipment for monitoring slope

displacements.
 18

Fig. 28: Radar image showing displacements in the east

face.
Analysis of the slope displacements, measured by both
the electro-optical system and the radar unit,
Fig. 30: 3DEC model of the East Wall including the
demonstrated that the displacements in a zone in the
Transfer Chamber.
rock mass surrounding the transfer chamber were
significantly larger than those in the remainder of the
East Wall. This suggested that a wedge, bounded by major
structural features, shown by blue lines in Fig. 26, had
formed in the rock mass and was moving more than the
surrounding rock.
During the 2012 review, a three-dimensional discrete
element model, using the Itasca 3DEC program, was
created to study the displacements and rock mass
behaviour in both the face of the East Slope and that
surrounding the Transfer Chamber. This model
incorporated the joint-defined rock blocks with rock
mass strength and deformation properties defined by the
Hoek-Brown and GSI parameters given in Table 4. The
major structural features were assigned strength
properties defined by the discontinuity property values
listed in Table 4. The cable reinforcement installed from
the transfer chamber, shown in Fig. 29, was included in
the model illustrated in Fig. 30.

Fig. 31: Displacement contours in the East Wall,

generated in a progressively mined 3DEC model.

The outcome of this analysis, shown in Figure 31, was

that the deformation results, for both the slope face and
the transfer chamber, were in acceptable agreement with
the monitored values. This provided the Geotechnical
Department with a sound basis on which to plan cable
reinforcement tension adjustments and cable
replacement installations to ensure that the chamber
Fig 29: Cable reinforcement installed from the conveyor remained stable for the remainder of the open pit
transfer chamber. operation.
 19

15. Conclusions Felipe Duran, Davide Elmo, Esteban Hormazabal, Jean

Hutchinson, Peter Kaiser, Gerhard Keyter, Tom Lam,
In the almost 40 years since its introduction, the Hoek- Loren Lorig, Derek Martin, Paul Marinos, Vassilis
Brown criterion for intact rock and rock masses has Marinos, Dougal McCreath, Bruno Marrai, Luis Olivares,
gained wide-spread international use for a wide range of John Read, Laurie Richards, Nick Rose, Peter Stacey,
engineering applications. The criterion for intact rock Nicholas Vlachopoulos, David Wood.
was based on brittle fracture concepts and so it should
only be used in the brittle behaviour range. Since its Appendix: Triaxial and tensile testing
introduction, several revisions and updates have been
made to the criterion, but its basic form has remained
Tensile testing can be introduced into a triaxial test
unchanged. A major revision made in the early 1990s
program by means of a modification used by Ramsey and
accompanied the development of the Geological Strength
Chester (2004). Instead of inserting a cylindrical
Index (GSI) to quantify engineering geological
specimen into a triaxial cell they substituted a dogbone
observations of the structure and condition of rock
specimen as illustrated in Fig. A1. The reduced section of
masses.
the specimen was wrapped in a Plasticine modelling clay.
The present update emphasizes the intended
When subjected to triaxial confinement, this modelling
application of the criterion to the brittle fracture of intact
clay fails plastically and transmits a uniform pressure
rock; expands on the previously existing methods of
onto the curved section of the dogbone specimen. Since
evaluating test results for the mechanical properties of
the ends of the specimen are larger in diameter than the
intact rock by applying the Bayesian approach to assess
central test section, a tensile stress is induced in the test
uncertainty; discusses the use of the GSI to describe the
section.
structure and condition of a wide range of types of rock
If D is the diameter of the core and d the diameter of the
masses; expands on the previously published guidelines
for the selection of the important disturbance factor, D; reduced test section, the tensile stress 3 induced along
sets out a recommended sequence of calculations for use the axis of the specimen, for a confining pressure of P, is:
in applying the criterion; and illustrates its application in
practical rock engineering through three different 𝜎3 = 𝑃(𝐷2 − 𝑑 2 )/𝑑 2 (A1)
examples.
Despite the revisions that have been made to the By adjusting the ratio of the diameters D and d, the
criterion and the experience gained in its use over almost confining pressure P and the axial load applied to the
40 years, care must be exercised in seeking to apply the specimen, a range of 3 and 1 stresses can be generated
criterion to some rock masses, particularly those at the in the tensile zone, as shown in Fig. A2. The preparation
higher and lower ranges of GSI. It is essential that the of dogbone specimen on a lathe is illustrated in Fig. A3.
geology of the site be well understood, and that the
mechanics of the engineering problem involved be
evaluated critically. When this has been done, the
methods discussed in this paper may be used to evaluate
the parameters in the criterion for the intact rock and the
rock mass.

16: Acknowledgements

The authors wish to express their acknowledgements to

organizations and individuals who have provided
permission to publish material in case histories and who
have contributed to discussions on the development of
the Hoek Brown creation and the Geological Strength
Index. It is impossible to name everyone, but a partial list
of organizations includes Egnatia Odos S.A., the South
African Electricity Supply Commission (ESKOM),
Braamhoek Consultants Joint Venture (BCJV), the
Chuquicamata Division of Codelco, Chile. Individuals who
have contributed are Rob Bewick, Ming Cai, Trevor
Carter, Carlos Carranza-Torres, Joe Carvalho, Luis- Fig. A1: Triaxial cell with dogbone specimen for triaxial
Fernando Contreras, Brent Corkum, Mark Diederichs, tensile testing
 20

References

Aydan O, Dalgic S. Prediction of deformation

behaviour of 3-lane Bolu tunnels through squeezing
rocks of North Anatolian fault zone (NAFZ). In:
Proceedings of the Regional Symposium on
Sedimentary Rock Engineering, Taipei; 1998;228-
233.
Andriev GE. Brittle failure of rock materials.
Rotterdam: Balkema; 1995.
Balmer G. A general analytical solution for Mohr’s
envelope. Proceedings of the American Society for
Testing and Materials 1952;52:1260-71.
Barton N. Some new Q value correlations to assist in
site characterization and tunnel design.
International Journal of Rock Mechanics and
Mining Sciences 2002;39(2):185-216.
Bieniawski ZT. Determining rock mass deformability
– experience from case histories. International
Fig. A2: Plot of triaxial tension data points obtained Journal of Rock Mechanics and Mining Sciences
by Ramsey and Chester (2004). and Geomechanics Abstracts 1978;15(5):237-247.
Bewick RP, Amann F, Kaiser PK, Martin CD.
Interpretation of UCS test results for engineering
design. In: Proceedings of the 13th International
Congress on Rock Mechanics: ISRM Congress
2015 – advances in applied & theoretical rock
mechanics, Montréal, Canada: International
Society for Rock Mechanics; 2015. Paper 521.
Bewick RP, Kaiser PK, Amann F. Strength of massive
to moderately jointed hard rock masses. Journal of
Rock Mechanics and Geotechnical Engineering
2019, 11 (this issue).
Bobich JK. Experimental analysis of the extension to
shear fracture transition in Berea sandstone. MS
thesis, Texas A & M University; 2005.
Bowcock JB, Boyd LM, Hoek E, Sharp JC.
Drakensberg pumped storage scheme: rock
Fig. A3: Preparation of a dogbone specimen using engineering aspects. In Bieniawski ZT, editor.
a toolpost grinder attached to a profile follower on Exploration in Rock Engineering. Proceedings of
a lathe. the Symposium on Exploration for Rock
Engineering 1976;2:121-139. Balkema: Rotterdam.
There is strong justification for using the dogbone Bozorgzadeh N, Escobar MD, Harrison JP.
specimen, illustrated in Fig. A2, for triaxial Comprehensive statistical analysis of intact rock
compression testing as well as the confined tensile strength for reliability-based design. International
testing described. This is because the smooth Journal of Rock Mechanics and Mining Sciences
transition of the stresses from the enlarged 2018;106:374-387.
specimen ends to the central test section reduced Brace WF. Brittle fracture of rocks. In: Judd WR,
the potential for axial splitting, particularly at low editor. State of stress in the Earth’s crust. New
confinement, which is common when testing York: Elsevier; 1964. p. 111-74.
cylindrical specimens loaded by steel platens. Brown ET. Strength of models of rock with
intermittent joints. Journal of the Soil Mechanics
and Foundations Division ASCE
1970;96(SM6):1935-1949.
 21

Brown, E.T and Hoek, E. Discussion on paper 20431 International Congress on Rock Mechanics,
by R. Ucar entitled Determination of shear Beijing, China. Boca Raton: CRC Press/AA
failure envelope in rock masses. Journal Balkema; 2011. p. 81-9.
Geotechnical Engineering. Division, American Griffith AA. The phenomena of rupture and flow in
Society of Civil Engineers, 1988;114(3):37l-373. solids. Philosophical Transactions of the Royal
Cai M, Kaiser PK, Uno H, Tasaka Y, Minami M. Society of London, Series A 1921; 221:163-98.
Estimation of rock mass deformation modulus and Griffith AA. Theory of rupture. In: Proceedings of the
strength of jointed hard rock masses using the GSI 1st International Congress on Applied Mechanics,
system. International Journal of Rock Mechanics Delft, The Netherlands; 1924. p. 55-63.
and Mining Sciences 2004;41(1):3-19. Hoek E. Fracture of anisotropic rock. Journal of the
Chern JC. Database of deformation modulus South African Institute of Mining and Metallurgy
measurements on projects in China and Taiwan. 1964;64(10):501-18.
Personal communication 2005. Hoek E. Rock fracture under static stress conditions.
Contreras LF, Brown ET. Bayesian inference of CSIR Report MEG 383, Pretoria, South Africa;
geotechnical parameters for slope reliability 1965.
analysis. In: Proceedings Slope Stability 2018, Hoek E. Brittle failure of rock. In: Stagg, KG,
Seville, Spain (in press). Zienkiewicz OC, editors. Rock mechanics in
Contreras, LF, Brown ET, Ruest M. Bayesian data engineering practice. London: Wiley; 1968. p. 99-
analysis to quantify the uncertainty of intact rock 124.
strength. Journal of Rock Mechanics and Hoek E. Strength of jointed rock masses.
Geotechnical Engineering 2018;10(1):11-31. Géotechnique 1983;33(3);187-223.
Cook NGW. The failure of rock. International Journal Hoek E. Estimating Mohr-Coulomb friction and
of Rock Mechanics and Mining Sciences cohesion values from the Hoek-Brown failure
1965;2(4):389-403. criterion. International Journal of Rock Mechanics
Day JJ, Hutchinson DJ, Diederichs MS. A critical look and Mining Sciences and Geomechanics Abstracts
at geotechnical classification for rock strength 1990;12(3);227-9.
estimation. In: Proceedings of the 46th US Rock Hoek E. Strength of rock and rock masses. ISRM
Mechanics/Geomechanics Symposium, Chicago, News Journal 1994;2(2):4-16.
USA; American Rock Mechanics Association Hoek E, Brown ET. Underground excavations in rock.
(ARMA); 2012. Paper 12-563. London: Institution of Mining and Metallurgy;
Deere DU. Geological considerations. In: Stagg, KG, 1980a.
Zienkiewicz OC, editors. Rock mechanics in Hoek E, Brown ET. Empirical strength criterion for
engineering practice. London: Wiley; 1968. p. 1-20. rock masses. Journal of the Geotechnical
Endersbee LA, Hofto EO. Civil engineering design Engineering Division ASCE 1980b;106(GT9):
and studies in rock mechanics for Poatina 1013-35.
underground power, Tasmania. Journal of the Hoek E, Brown ET. The Hoek-Brown failure criterion
Institution of Engineers, Australia, 1963;35:187- – 1988 update. In: Curran JH, editor. Proceedings
206. of the 15th Canadian Rock Mechanics Symposium,
Fairhurst C. On the validity of the “Brazilian” test for Toronto, Canada. Toronto: Department of Civil
brittle materials. International Journal of Rock Engineering, University of Toronto; 1988. p. 31-8.
Mechanics and Mining Sciences 1964;1(4):535-46. Hoek E, Brown ET. Practical estimates of rock mass
Flores G, Catalan A. A transition from a large open pit strength. International Journal of Rock Mechanics
to a novel “macroblock variant” block caving and Mining Sciences and Geomechanics Abstracts
geometry at Chuquicamata Mine, Codelco Chile. 1997;34(8):1165-86.
Journal of Rock Mechanics and Geotechnical Hoek E, Carranza-Torres C, Corkum B. Hoek-Brown
Engineering 2019,11 (this issue). criterion – 2002 edition. In: Hammah R, Bawden
Franklin JA, Hoek E. Developments in triaxial testing W, Curran J, Telesnicki M, editors. Mining and
technique. Rock Mechanics 1970;2(2):223-8. tunnelling innovation and opportunity, Proceedings
Gerogiannopoulos NG, Brown ET. The critical state of the 5th North American Rock Mechanics
concept applied to rock. International Journal of Symposium and 17th Tunnelling Association of
Rock Mechanics and Mining Sciences and Canada Conference, Toronto, Canada. Toronto:
Geomechanics Abstracts 1978;15(1):1-10. University of Toronto 2002; 1:267-273.
Gonzalez-Garcia AJ. Rock strength and failure: some Hoek E, Diederichs MS. Empirical estimates of rock
common and uncommon issues. In: Qian Q, Zhou mass modulus. International Journal of Rock
Y, editors. Harmonising rock engineering and the Mechanics and Mining Sciences 2006; 43(2): 203-
environment, Proceedings of the 12th ISRM 15.
 22

Hoek E, Kaiser PK, Bawden WF. Support of Johannesburg: Southern African Institute of Mining
underground excavations in hard rock. Rotterdam: and Metallurgy; 2010. p.1-23
AA Balkema; 1995. Keyter GJ, Ridgeway M, Varley PM. Rock
Hoek E, Marinos PG. Predicting tunnel squeezing engineering aspects of the Ingula powerhouse
problems in weak heterogeneous rock masses. caverns. In: Proceedings of the 6th International
Tunnels and Tunnelling International Symposium on Ground Support in Mining and
2000;132(11):45-51. Civil Engineering Construction, Cape Town, South
Hoek E, Marinos P, Benissi M. Applicability of the Africa. Johannesburg: Southern African Institute of
Geological Strength Index (GSI) classification for Mining and Metallurgy; 2008. p. 409-445.
very weak and sheared rock masses. The case of the Kovari K, Tisa A. Multiple failure state and strain
Athens schist formation. Bulletin of Engineering controlled triaxial tests. Rock Mechanics
Geology and the Environment 1998;57(2):151-60. 1974;7(1):17-33.
Hoek E, Marinos P, Marinos V. Characterization and Kruschke JK. Doing Bayesian data analysis: a tutorial
engineering properties of tectonically undisturbed with R, JAGS and Stan. 2nd ed. Amsterdam, New
but lithologically varied sedimentary rock masses. York: Academic Press; 2015.
International Journal of Rock Mechanics and Langford JC, Diederichs MS. Quantifying uncertainty
Mining Sciences 2005;42(2):277-85. in Hoek-Brown intact strength envelopes.
Hoek E, Martin CD. Fracture initiation and International Journal of Rock Mechanics and
propagation in intact rock – a review. Journal of Mining Sciences 2015; 74:91-104.
Rock Mechanics and Geotechnical Engineering Lau JSO, Gorski B. Uniaxial and triaxial compression
2014;6(4):278-300. tests on URL rock samples from boreholes 207-
Hoek E, Wood D, Shah S. A modified Hoek-Brown 045-GC3 and 209-069-PH3. Divisional Report
criterion for jointed rock masses. In: Hudson JA, (Mining Research Laboratories (Canada)), MRL
editor. Rock Characterisation, Proceedings of the 92-025(TR). Ottawa: Mining Research
ISRM Symposium Eurock ’92, Chester, UK. Laboratories; 1992.
London: British Geotechnical Society; 1992. p. Marinos P, Hoek E. GSI – a geologically friendly tool
209-14. for rock mass strength. In: Proceedings GeoEng
Jaeger JC, Cook NGW. Fundamentals of rock 2000, International Conference on Geotechnical
mechanics, 3rd edition. London: Chapman and Hall; and Geological Engineering, Melbourne, Australia.
[Link] PK, Amann F, Bewick RP. Lancaster, PA: Technomic Publishing Co; 2000b.
Overcoming challenges of rock mass p. 1422-40.
characterisation for underground construction in Marinos P, Hoek E. Estimating the geotechnical
deep mines. In: Proceedings of the 13th properties of heterogeneous rock masses such as
International Congress on Rock Mechanics: ISRM flysch. Bulletin of Engineering Geology and the
Congress 2015 – advances in applied & theoretical Environment 2001;60(2):85-92.
rock mechanics, Montréal, Canada: International Marinos V, Marinos P, Hoek E. The geological
Society for Rock Mechanics; 2015. Paper 241. strength index: applications and limitations.
Kaiser PK, Kim B, Bewick RP, Valley B. Rock mass Bulletin of Engineering Geology and the
strength at depth and implications for pillar design. Environment 2005;64(1):55-65.
In: Van Sint Jan M, Potvin Y, editors. Deep Mining Marinos V. A revised geotechnical classification GSI
2010, Proceedings of the Fifth International system for tectonically disturbed rock masses, such
Seminar on Deep and High Stress Mining, as flysch. Bulletin of Engineering Geology and the
Santiago, Chile. Perth, Australia: Australian Centre Environment 2017; published on line.
for Geomechanics, 2010. p. 463-76. McClintock FA, Walsh JB. Friction on Griffith cracks
Kalamaris GS, Bieniawski ZT. A rock mass strength in rocks under pressure. In: Proceedings of the 4 th
concept for coal incorporating the effect of time. In: US National Congress of Applied Mechanics,
Fuji T, editor. Proceedings of the 8th Congress on Berkeley, USA. New York: American Society of
Rock Mechanics, ISRM, Tokyo, Japan. Rotterdam: Mechanical Engineers; 1962. p. 1015-21.
Balkema: AA Balkema. 1995;1:295-302. Mogi K. Pressure dependence of rock strength and
Kellaway M, Taylor D, Keyter GJ. The use of transition from brittle fracture to ductile flow.
geotechnical instrumentation to monitor ground Bulletin Earthquake Research Institute, University
displacement during excavation of the Ingula power of Tokyo 1966; 44: 215-232.
cavern, for model verification and design Murrell SAF. The strength of coal under triaxial
verification purposes. In: Proceedings of South compression. In: Walton WH, editor. Mechanical
African Tunnelling 2012 – Lessons Learned on properties of non-metallic brittle materials.
Major Projects, Ladysmith, South Africa.
 23

London: Butterworths Scientific Publications; Stephens RE, Banks DC. Moduli for deformation
1958. p. 123-45. studies of the foundation and abutments of the
Olavarria S, Adriasola P, Karzulovic A. Transition Portugues Dam – Puerto Rico. In: Khair AW,
from open pit to underground mining at editor. Rock mechanics as a guide for efficient
Chuquicamata, Antofagasta, Chile. In: Proceedings utilization of natural resources, Proceedings of the
of the International Symposium on Stability of 30th US Symposium, Morgantown, USA.
Rock Slopes in Open Pit Mining and Civil Rotterdam: AA Balkema; 1989:31-8.
Engineering, Cape Town: South Africa. Ulusay R, Hudson JA, editors. (eds), The complete
Johannesburg: Institute of Mining and Metallurgy; ISRM suggested methods for rock characterization,
2006. p.421-434. testing and monitoring: 1974-2006. Ankara: ISRM
Perras MA, Diederichs MS. 2014. A review of the Turkish National Group; 2007.
tensile strength of rock: concepts and testing. Von Kármán T. Festigkeitsversuche unter allseitigem
Geotechnical and Geological Engineering Druck. Zeitschrift Verein Deutscher Ingenieure
2014;32(2):525-46. 1911; 55:1749-57.
Ramsey JM, Chester FM. Hybrid fracture and the Vlachopoulos N, Diederichs MS, Marinos V, Marinos
transition from extension fracture to shear fracture. P. Tunnel behaviour associated with the weak
Nature 2004;428(4 Mar):63-66. Alpine rock masses of the Driskos Twin Tunnel
Read SAL, Richards LR, Perrin ND. Applicability of system, Egnatia Odos Highway. Canadian
the Hoek-Brown failure criterion to New Zealand Geotechnical Journal 2012; 50:91-120.
greywacke rocks. In: Vouille G, Berest P, editors. Walton S, Hasan O, Morgan K, Brown MR. Modified
Proceedings of the 9th International Congress on cuckoo search; a new gradient free optimization
Rock Mechanics, Paris, France. Lisse: AA algorithm. Chaos, Solutions and Fractals 2011;
Balkema;1999. p. 655-60. 44(9):710-18.
Ros M, Eichinger A. Experimental study of theories of Zuo JP, Li HT, Xie HP, Ju Y, Peng SP. A nonlinear
rupture. Non-metallic materials. Eidgenoss, strength criterion for rocklike materials based on
Materialprufungsanstalt, E.T.H Zurich, No 28;1928 fracture mechanics. International Journal of Rock
(in German). Mechanics and Mining Sciences 2008;45(4):594-9.
Rosengren KJ, Jaeger JC. The mechanical properties Zuo JP, Liu H, Li H. A theoretical derivation of the
of an interlocked low-porosity aggregate. Hoek-Brown failure criterion for rock materials.
Géotechnique 1968;18(3):317-26. Journal of Rock Mechanics and Geotechnical
Rose N D, Scholz M, Burden J, King M, Maggs C and Engineering 2015; 7(4):361-6.
Havaej M. 2018, Quantifying Transitional Rock
Mass Disturbance in Open Pit Slopes Related to
Mining Excavation, in Slope Stability 2018 – XIV
International Congress on Energy and Mineral
Resources: Asociación Nacional de Ingenieros de
Minas, Seville, Spain, p. 1273–1288.
Sakurai S. Back analysis in rock engineering. ISRM
Book Series, London: Taylor & Francis Group;
2017.
Schwartz AE. Failure of rock in the triaxial shear test.
In: Proceedings of the 6th Rock Mechanics
Symposium, Rolla, USA, 1964. p. 215-32.
Serafim JL, Pereira JP. Consideration of the
geomechanical classification of Bieniawski. In:
Proceedings of the International Symposium on
Engineering Geology and Underground
Construction, Lisbon, Portugal. Lisbon:
SPG/LNEC; 1983. 1(II):33-44.
Shah S, Hoek E. Simplex reflection analysis of
laboratory rock strength data to obtain Hoek-Brown
parameters. Canadian Geotechnical Journal
1992;29(2):278-87.
Sheorey PR. Empirical rock failure criteria.
Rotterdam: AA Balkema. 1997.