You are on page 1of 19

ARTICLE IN PRESS

International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265


www.elsevier.com/locate/ijrmms

Determination of residual strength parameters of jointed


rock masses using the GSI system
M. Caia,, P.K. Kaisera, Y. Tasakab, M. Minamic
a

Geomechanics Research Centre, MIRARCO, Laurentian University, Sudbury, Ont., Canada


Department of Advanced Engineering, Tokyo Electric Power Services Company Limited, Tokyo, Japan
c
Department of Construction, Tokyo Electric Power Company, Tokyo, Japan

Accepted 13 July 2006


Available online 26 September 2006

Abstract
The Geological Strength Index (GSI) system, proposed in 1995, is now widely used for the estimation of the rock mass strength and the
rock mass deformation parameters. The GSI system concentrates on the description of two factors, rock structure and block surface
conditions. The guidelines given by the GSI system are for the estimation of the peak strength parameters of jointed rock masses. There
are no guidelines given by the GSI, or by any other system, for the estimation of the rock mass residual strength that yield consistent
results. In this paper, a method is proposed to extend the GSI system for the estimation of a rock masss residual strength. It is proposed
to adjust the peak GSI to the residual GSIr value based on the two major controlling factors in the GSI systemthe residual block
volume V rb and the residual joint condition factor J rc . Methods to estimate the residual block volume and joint condition factor are
presented. The proposed method for the estimation of rock masss residual strength is validated using in-situ block shear test data from
three large-scale cavern construction sites and data from a back-analysis of rock slopes. The estimated residual strengths, calculated
using the reduced residual GSIr value, are found to be in good agreement with eld test or back-analyzed data.
r 2006 Elsevier Ltd. All rights reserved.
Keywords: Rock mass; Rock mass classication; Geological strength index; Rock failure

1. Introduction
Knowledge of the rock mass strength and deformation
behaviors is required for the design of many engineering
structures in or on rock, such as foundations, slopes,
tunnels, underground caverns, drifts, and mining stopes.
A better understanding of the rock mass strength behavior,
including the peak and residual strengths, will facilitate the
cost-effective design of such structures.
The determination of the global mechanical properties of
a jointed rock mass remains one of the most difcult tasks
in rock mechanics. Many researchers have developed
constitutive models to describe the strength and deformation behaviors of jointed rock masses e.g., [35]. Because
there are so many parameters that affect the deformability
and strength, it is generally impossible to develop a
Corresponding author. Tel.: +1 705 675 1151; fax: +1 705 675 4838.

E-mail address: mcai@mirarco.org (M. Cai).


1365-1609/$ - see front matter r 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijrmms.2006.07.005

universal model that can be used to a priori predict the


strength of the rock mass. Traditional methods to
determine these parameters include plate-loading tests for
deformation modulus and in-situ block shear tests for
strength parameters. These tests can only be performed
when the exploration adits are excavated and the cost of
conducting in-situ tests is high. Although back-analyses
based on eld measurement are helpful in determining the
strength and deformation parameters as a project proceeds,
they do not provide design parameters at the pre-feasibility
or feasibility study stages.
Few attempts have been made to develop methods to
characterize the jointed rock mass to estimate the deformability and strength indirectly. The Geological Strength
Index (GSI), developed by Hoek et al. [1], is one of them. It
uses properties of intact rock and conditions of jointing to
determine/estimate the rock mass deformability and
strength. GSI values can be estimated based on the
geological description of the rock mass and this is well

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

248

suited for rock mass characterization without direct access


to the rock mass from tunnels. The GSI system concentrates on the description of two factors, rock structure and
block surface conditions. It is a system that provides a
complete set of mechanical properties (HoekBrown
strength parameters mb and s, or the equivalent Mohr
Coulomb strength parameters c and f, as well as the elastic
modulus E) for design purposes. Recently, a means to
quantify this approach by use of eld data, which employs
the block volume (Vb) and a joint condition factor (Jc) as
quantitative characterization factors, was presented in [2].
Guidelines given by the GSI system are for the
estimation of the peak strength of jointed rock masses. In
general, rock masses, except when highly disturbed, exhibit
strain-softening post-peak behavior, so that the residual
strength parameters are lower than the peak parameters.
Both are required for design. Strain-softening behavior
describes the gradual loss of load-bearing capacity of a
material. For hard rocks, the term strength weakening
seems more appropriate than the term strain-softening
because softening refers to reduction of rock stiffness.
At lower connement levels such as near excavation walls,
most rock masses exhibit some post-peak strength loss, and
when strained sufciently reach the residual strength. The
peak and residual strengths are respectively the maximum
and minimum stresses of a rock mass that can be sustained
under a given connement condition. The residual strength
is generally only reached after considerable plastic deformation. There are some guidelines for the estimation of the
rock mass residual strength, given by some researchers
[6,7], but upon application of these guidelines, it is often
observed that there are signicant inconsistencies in the
residual strengths derived from them. Hence, a new method
has been developed and tested to extend the GSI system for
rock masss residual strength estimation. For this purpose,
we propose to adjust the peak GSI value based on the two
major controlling factors in the GSI system, the block
volume Vb and the joint condition factor Jc to arrive at a
residual GSIvalue (GSIr) based on a residual block
volume V rb and residual J rc . The residual GSIr value is
calculated from a relationship involving residual V rb and
J rc . The proposed method for the estimation of rock masss
residual strength is then validated using in-situ block shear

test data from three large-scale cavern construction sites


and data from a back-analysis of a rock slope stability
study.
The following denitions for the peak and residual
strength are illustrated in Fig. 1. The residual strength is
dened by the plateau after the peak, in a strain range of
about 510 times the strain corresponding to the peak
strength. This level of load bearing capacity is commonly
referred as the residual strength in most civil and mining
engineering applications. If straining is allowed to continue, then, the strength can further decreases and
eventually reaches a lower strength.
2. Inuence of the rocks residual strength on support design
for underground excavations
The post-peak behavior of rocks is important in the
design of underground excavations because it has a
signicant inuence upon the stability of the excavations.
Rock mechanics test data are available on the strength of
rock masses, especially for intact rocks. A brief review is
presented in the following sub-sections, with focus on the
post-peak behavior of rocks.
2.1. Laboratory tests
Pioneers in experimental study of the complete stress
strain relations of rocks include Paulding [8], Cook [9],
Hoek [10], Bieniawski [11], Wawersik [12], Wawersik and
Fairhurst [13], and many others. The post-peak behavior of
rocks was studied only after the development of stiff servocontrolled test machines in the middle of 1960s. In uniaxial
compression, two failure modes are observed [13]. One is
the local tensile or spalling fracture sub-parallel to the
applied load direction and the other is a local and
macroscopic shear fracture. In heterogeneous rocks and
under low connements, spalling-type failure dominates.
The post-failure behavior of the rocks can be divided into
two classes [13]. Class I behavior is characterized by a
stable fracture propagation. The rocks retain some strength
even when their maximum load-carrying capacity has been
exceeded. Unstable fracture propagation behavior is
characteristic of the Class II behavior.

Peak

Peak

Stress

Stress

5 to 10 Peak
Residual

Residual

(a)

Peak

Strain

(b)

Strain

Fig. 1. (a) Strain-softening of rocks; (b) perfectly brittle failure of rocks.

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

Axial stress, a (MPa)

300

48.3 MPa
34.5 MPa

200
27.6 MPa
20.7 MPa

100

13.8 MPa

6.9 MPa
3.45 MPa

0.10

0.20

0.30
0.40
0.50
Axial stress, a (%)

0.60

0.70

Fig. 2. Stressstrain curves for Tennessee Marble at different conning


stresses [13].

Triaxial test data on marbles by Wawersik and Fairhurst


[13] (Fig. 2) and Rummel and Fairhurst [14] revealed that
peak and residual strengths of rocks increase with
increasing conning pressures. At low connements, the
loss of the cohesive strength component around peak load
leads to strain localization with signicant stress-drop
which is traditionally called strain-softening behavior.
Seeber [15] noticed that if the conning pressure was
greater than one-fth of the axial stress at failure, strainsoftening was unlikely to occur. For reference, the brittleductile transition limit given by Mogi [16] is s3/s1 1/3.4.
Clearly, strain-softening behavior must be expected to
dominate near underground excavations where connement is reduced.
The post-peak behavior of rocks tested in the laboratory
is dependent on the specimen geometry. This is because
that the post-failure curve is altered depending on the
relative stiffness of the machine and the specimen, as well
as the internal connement prole. With the development
of fractures in the post-peak region, the effective area at the
center of the specimen slowly decreases. The relative
decrease in cross-sectional area is greater for the specimens
of greater height (h) to diameter (d) ratio (h/d). Consequently, the post-peak stress strain curves of specimens
with higher h/d ratios are steeper [17].
Besides the uniaxial and triaxial tests, double shear
testing method are utilized by some researchers to study the
complete shear stressshear displacement relations of rocks
[18,19]. The residual shear strength typically depends on
the applied axial pressure (frictional materials).
The behavior of joints affects the strength and deformation properties of jointed rock masses signicantly. The
shear strength of joints is a major factor in controlling the
strength of jointed rock masses. Early experimental studies
on rock joints were carried out by Patton [20], followed by
Goodman [21] Barton and Choubey [22], Bandis et al. [23],
Barton et al. [24], and others. Conceptually, there are
three modes of failure conrmed from these tests, i.e.,
(a) asperity (roughness) override at low normal stresses;

249

(b) failure through asperities at elevated normal stresses;


and (c) combinations of asperity override and failure at
intermediate normal stresses.
Barton and Choubey [22] proposed the concept of joint
roughness coefcient (JRC) to describe the peak strength
of a joint. This concept was further developed incorporating the mobilized JRC to account the joint surface
evolution at different deformation stages [23]. The joint
shear strength is known to be dependent on three
components, i.e., the residual or basic frictional component, the geometrical component, and the asperity component. The asperity and the geometrical components
constitute the roughness strength that has to be mobilized
during shearing of the joint.
One important observation from Bartons joint model is
that within limited displacements, only an ultimate
mobilized joint roughness coefcient (JRCmob) can be
reached. JRCmob is roughly half of JRCpeak when the
displacement is about 10 times of the joint peak strength
displacement (Fig. 3). According to this gure, the residual
strength is only reached when the shearing displacement is
extremely large. In most engineering applications, such
large straining cannot be tolerated. Therefore, the joint
strength at dE10dpeak can be considered as corresponding
to the rocks residual strength dened in Fig. 1.
2.2. In-situ tests
Strength and deformation properties determined from
the laboratory tests are seldom applicable to eld conditions. To overcome this problem, large-scale in-situ tests
have been conducted in some engineering projects. The
tests include in-situ uniaxial compressive tests, triaxial
tests, block shear tests, etc. Uniaxial compressive tests have
been conducted mostly in coalmines to study the stability
of pillars [25]. In-situ block shear tests are often executed in
large civil projects to obtain the shear strength of rock
masses and strengths of bedding or other weakness planes.
The block shear test is often conducted in an underground
gallery or adit. The roof and sidewalls are used to carry the
reaction of the applied normal and shear loads. The rock
blocks are of square base of suitable dimensions in width
and height. The shear stress vs. shear displacement relation
is recorded to identify the peak and residual strengths. It is
observed that the block shear test has a major deciency
that it provides residual strength of a single shear plane
where in the rock mass, some degree of interlocking is
retained even at its residual state. Thus, in-situ block
shear tests tend to underestimate the residual values. In
addition, the loading system can generally be viewed as a
soft system so that the post-peak stressdisplacement
curves may not characterize the strain-softening process
properly for conned states. Most large-scale cavern
designs in Japan, however, employ the residual strength
parameters obtained from these in-situ block shear tests
and utilize post-peak brittle failure models for precautious
design (Fig. 1(b)).

ARTICLE IN PRESS
250

M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

15

: JRCm=

Example

1275

105

PEAK

1.0

ROUGHNESS
MOBILIZED

45

1.0
DILATION
BEGINS
AT
JRCM = 0

-1.0

-1.5

-2.0

N OF FRICTIO
N

-0.5

0
0.3
0.6
1.0
2.0
4.0
10.0
100.0

MOBILIZATIO

JRCmob / JRCpeak

0.5

ULTIMATE

ROUGHNESS
DESTROYED

RESIDUAL

2.0

3.0

30

4.0

(/ PEAK)
i = p r

JRCm

JCS
i = JRC log ( )
n

JRCp
r/i
0
0.75
1.0
0.85
0.70
0.50
0

START
15
peak
ultimate
residual

EXAMPLE:
r = 30, i = 15
JRC = 15, n = 10.0 MPa.
JCS = 100 MPa

( )

r/i

Fig. 3. Normalized joint roughnessshear displacement relationship [24].

2.3. Need for accurate determination of the residual strength


of rock masses
It is observed that following the strain-softening
behavior of rocks under loading, the residual strength
represents more or less the shear strength along a surface or
shear zone of the fractured rock. In most cases, the residual
strength can be described by the MohrCoulomb criterion
with near zero cohesive strength. The post-peak strength
depends on the resistance developed on the failure plane
against further straining. Initially, the fracture orientation,
degree of interlocking, surface irregularity or roughness
will affect the level of resistance. However, as strain
increases, the residual strength will be less. In the eld, the
post-peak strength level will be inuenced by the boundary
conditions as well. If further straining is constrained, then,
the residual strength level cannot be reached and the rock
mass can thus support a higher load than the residual
strength would suggest.
It is a very challenging and difcult task to correctly
represent the strain-softening behavior of rock masses, due
to a lack of large-scale test data. Most numerical tools
designed for rock engineering application, however,
provide strain-softening constitutive models of varying
sophistication to describe the behavior of jointed rock
masses [2628]. In these models, the residual strength of the
rock mass and the rate of post-peak strength degradation
play an important role in the determination of the size of

the plastic zones and the associated rock mass deformation, affecting the nal rock support system design. For
example, the current version of Phase2D [29], an FEM
program developed by Rocscience, allows the user to dene
both peak and residual strength parameters of rock masses.
When the stress of an element has exceeded its peak
strength, it fails in a perfectly brittle manner, switching
directly from peak to post-peak residual parameter values,
with no strain-dependent softening mechanism (Fig. 1(b)).
Although extremely important for these numerical models,
only limited suggestions are given in the users manual on
how to determine the residual strength parameters.
If the residual strengths are not determined appropriately, an optimal rock support design can never be
achieved. The inuence of the residual strength on the
yielding zone around a 6 m wide tunnel is illustrated in
Fig. 4. The tunnel is located at a depth of about 500 m and
the maximum and minimum in-situ stresses are 12.5 and
4.8 MPa, respectively. The angle between the maximum
principal stress direction and the vertical is 261. Rock mass
peak cohesion, friction angle, dilation angle, and tensile
strength are 3 MPa, 551, 51, and 0.6 MPa, respectively. It is
assumed that after peak strength, the rock mass reaches the
residual strength in a brittle manner. The residual tensile
strength is assumed to be zero and the dilation angle
unchanged from the peak dilation. It should be noted that
constant dilation is an approximation that is clearly not
physically correct. This assumption is made largely because

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

251

the strength reduction mechanism and hence provide a


method for residual strength estimation [6].
Russo et al. [7] proposed to set the residual GSIr value at
36% of the peak GSI value. This empirical relation may
underestimate the residual GSI values for poor-quality
rock masses. On the other hand, for very good-quality rock
masses, it may overestimate the residual GSIr values. Based
on laboratory triaxial test on limestone, Ribacchi [31]
suggested to use the following relations to estimate the
residual strength of jointed rock masses:
(a)

(c)

(b)

(d)

Fig. 4. Inuence of residual strength on the yielding zone around a tunnel.

little is known about how the dilation of a rock mass


changes past peak. Even if the peak strength is the same for
all cases, for different residual friction angles and cohesions
shown in Fig. 4, the yielding zones are drastically different.
The underlying implication is that the residual strength of
rock masses has to be properly determined in order to
design appropriate rock support systems.
2.4. Review of existing methods to determine the residual
strength of rock masses
To design underground structures properly, both the
peak and residual strengths of the rock mass are needed.
Much research has been focused on the determination of
peak strengths, and limited attempts have been made to
estimate the residual strength of jointed rock masses.
The existing GSI system only provides guidance for rock
mass peak strength estimation. To address the issue of rock
mass residual strength, Hoek [6,30] suggested elasticbrittle, strain-softening, and elastic-perfect plastic postpeak rock mass behavior for very good, average, and very
poor quality rock masses, respectively. Hoek also suggested
that in the case of an average quality rock mass, it is
reasonable to assume that the post failure characteristics
can be estimated by reducing the GSI value from the in-situ
value to a lower value which characterizes the broken rock
mass. The reduction of the rock mass strength from an
undisturbed to a broken state corresponds to the strainsoftening behavior. However, the validity of this assumption is unknown, and new study is needed to understand

mr 0:65mb ;

sr 0:04s

or sc r 0:2sc ,

(1)

where mb and s are the HoekBrown peak strength


parameters, the subscript r indicates residual values,
and sc is the uniaxial compressive strength of the intact
rock. Taking into account the structure of the tested rock,
these relations may be valid only for rock masses in which
joints are characterized by a thin inlling or slightly
weathered to unweathered joint walls. The corresponding
GSI reduction that would t such parameters is approximately GSIr 0.7GSI.
The opinions of several rock mechanics experts on the
post-peak strength parameters were summarized in [32]. It
is generally agreed that the reduction of sc would be
physically and conceptually incorrect because this is a
xed index parameter that is determined from intact
rock specimens.
In summary, several attempts have been made to
estimate the residual strength of jointed rock masses. The
reduction of GSI to its residual value is a logical choice,
because the failure of rock masses is associated with the
crushing of intact rock and the wearing of the joint surface
roughness. Current reduction methods, however, lack
generality and lead to inconsistent results for different
rock masses. Here, a new method is proposed based on the
observation of actual rock mass failure process from
laboratory and in-situ tests, as well as on the understanding
of the rock fracturing process from numerical simulation.
3. Determination of the strength parameters using the GSI
system
3.1. Estimation of peak strength of rock masses using the
GSI system
Two types of strength criteria, i.e., the MohrCoulomb
and HoekBrown failure criteria, are widely used in rock
engineering. The equivalent MohrCoulomb parameters
can be obtained based on the HoekBrown envelope and a
chosen range of connement (s3). In terms of major and
minor principal stresses, s1 and s3, the MohrCoulomb
failure criterion can be expressed as
s1

2ccosf
1 sinf

s3 ,
1  sinf 1  sinf

(2)

where c and f are the cohesive strength and angle of


friction of the rock mass, respectively.

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

The generalized HoekBrown criterion for jointed rock


masses [33] is

a
s3
(3)
s1 s3 sc mb s ,
sc

jointing. To facilitate the use of the system, Cai et al. [2]


presented a quantitative approach that employed the block
volume Vb and a joint surface condition factor Jc as
quantitative characterization factors. The quantitative
approach was validated using eld test data and applied
to the estimation of the rock mass properties at two cavern
sites in Japan. The quantied GSI chart is presented in
Fig. 5. It provides a means for consistent rock mass
characterization and thus improves the utility of the GSI
system.

where mb, s, a are constants for the rock mass, and sc is the
uniaxial compressive strength of the intact rock. The GSI
system was developed to determine the HoekBrown
strength parameters, using the rock structure and joint
surface condition description to describe the rock mass

Very poor
Slickensided, highly weathered surfaces with
soft clay coatings or fillings

Poor
Slickensided, highly weathered surfaces with
compact coating or fillings of angular fragments

on
ez
lur
fai

ritt

le

1E+6
(1 m3)

tia

75

ten

80

65

100E+3

70

30 cm

Very Blocky - interlocked, partially


disturbed rock mass with multifaceted
20
angular blocks formed by four or more
discoutinuity sets
Joint spacing 10 - 30 cm

Fair
Smooth, moderately weathered or
altered surfaces

Good
Rough, slightly weathered,
iron stained surfaces

ne
zo
ure
ail

ef

85

55
45

10E+3

60
50

35

10 cm

1000
(1 dm3)

40
Blocky/disturbed - folded and/or
faulted with angular blocks formed by
many intersecting discontinuity sets
5
Joint spacing 3 - 10 cm

25
30

100
15

Disintegrated - poorly interlocked,


heavily broken rock mass with a
mixture or angular and rounded
rock pieces
Joint spacing < 3 cm

20

10

10
1 cm

Foliated/laminated/sheared - thinly
laminated or foliated, tectonically sheared
N/A
N/A
weak rock; closely spaced schistosity
prevails over any other discontinuity set,
resulting in complete lack of blockiness
12
4.5
1.7
Joint spacing < 1 cm

0.67
0.25
Joint Condition Factor Jc

Fig. 5. Quantication of GSI chart [2].

1
5

0.1
0.09

Block Volume Vb (cm3)

40

90

10E+6

Po

Blocky - very well interlocked


undisturbed rock mass consisting
of cubical blocks formed by three
orthogonal discontinuity sets
Joint spacing 30 - 100 cm

90
80
70
60
50

ittl

100 cm

95

Br

Block Size
Massive - very well interlocked
undisturbed rock mass blocks formed
by three or less discontinuity sets
with very wide joint spacing
150
Joint spacing > 100 cm

Very good
Very rough, fresh unweathered surfaces

Joint or Block Wall Condition

lb

252

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

Fig. 6. Visualization of the GSI system [34].

In numerical model implementation, it is sometimes


troublesome to refer to a chart for the determination of the
GSI values. Recently, based on the proposed quantitative
chart, and using surface tting techniques, the following
equation for the calculation of GSI from Jc and Vb was
developed [34]:
GSI V b ; J c

26:5 8:79 ln J c 0:9 ln V b


,
1 0:0151 ln J c  0:0253 ln V b

(4)

where Jc is a dimensionless factor, and Vb is in cm3. A


graphic representation of Eq. (4) is presented in Fig. 6. In
other words, the HoekBrown strength parameters and
deformation modulus can be directly expressed as a
function of Vb and Jc:


GSI  100
mb mi exp
,
(5)
28  14D

s exp


GSI  100
,
9  3D

(6)


1  GSI=15
e
 e20=3 ,
(7)
6
where D is a factor that depends on the degree of
disturbance to which the rock mass has been subjected by
blast damage and stress relaxation. The D factor was
introduced in the latest update [35] of the HoekBrown
failure criterion.

a 0:5

3.2. Estimation of residual strength of rock masses using the


GSI system
As is demonstrated by the identication and visualization of inuencing parameters in major rock mass
classication systems, the block volume and the joint
surface condition are the two most important factors that
control the quality and hence the strength and deformability of jointed rock masses [34]. Block volume is affected

253

by the joint set spacing and persistence. Joint condition is


controlled by joint roughness, weathering, and inlling
material. These are important factors that need to be
characterized for rock masss residual strength estimation.
Sjoberg [36] reported that when using the GSI system to
estimate the rock mass strength at the Aznalcollar open pit
mine in Spain, it was found that by assuming the disturbed
rock mass category, good agreement was found between
estimated strength values and back-calculated strengths
from observed slope failures in the footwall. The same
author reckoned that these strength values were probably
conservative and representative of the residual strength of
the rock mass. Although this hypothesis needs to be further
veried by additional data from other slope failures, it
suggests that the reduction of GSI for residual strength
estimation is logical.
To extend the GSI system for rock mass residual
strength estimation, we propose to adjust the original
GSI value based on the two major controlling factors in the
GSI system, i.e., block volume Vb and joint condition
factor Jc, to reach their residual values.
3.2.1. Residual block volume
Block size, which is determined from the joint spacing,
joint orientation, number of joint sets and joint persistence,
is an extremely important indicator of rock mass quality.
Block size is a volumetric expression of joint density. The
block volume spectrum from massive to very blocky
rock masses ranges from 103 to 107 cm3, and for
disturbed to sheared rock from 0.1 to 103 cm3.
Joints are often of limited length, even in a larger scale
[37]. If the joints are not persistent, i.e., with rock bridges,
the rock mass strength is higher and the global rock
stability is enhanced. This effect can be considered using
the concept of equivalent block volume as suggested in [2].
The difference between the peak and residual strength of
a rock mass with non-persistent joints is larger than that
of a rock mass with persistent joints. The implication is
that a drop of GSI from peak to residual values is larger for
rock masses with non-persistent joints. Besides rock
bridges, rock asperity interlocking also contributes to the
difference between peak and residual strengths.
If a rock experiences post-peak deformation, the rock in
the broken zone is fractured and consequently turned into
a poor and eventually very poor rock. Hence, the rock
mass properties of a rock mass after extensive straining
should be derived from the rock class of very poor rock
mass in the RMR system [38] or disintegrated in the
GSI system.
For the residual block volume, it is observed that the
post-peak block volumes are small because the rock mass
has experienced tensile and shear fracturing. After the peak
load, the rock mass becomes less interlocked, and is heavily
broken with a mixture of angular and partly rounded rock
pieces. Numerical simulation using ELFEN [39] and DIGS
[40] revealed that the rock masses in the fracture zone
around underground openings are broken to small blocks.

ARTICLE IN PRESS
254

M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

Intact

Fig. 7. Example of sheared Flysches in Greece (photo courtesy of Evert


Hoek).

ELFEN was also used to simulate the rock failure process


in uniaxial and biaxial loading conditions [41]. The results
indicate that the rock will gradually disintegrate into small
blocks, mostly along the localized shear or kink band zone,
before the residual strength in reached.
Detailed examination of the rock mass damage state
before and after the in-situ block shear tests at some
underground cavern sites in Japan revealed that in areas
that were not covered by concrete, the failed rock mass
blocks are 15 cm in size. The rock mass is disintegrated
along a shear zone in these tests.
It can be seen from Fig. 5 that the block volume sizes of
the disintegrated rock masses are in the range of 127 cm3,
with an average of about 10 cm3. This is supported by fault
outcrop observations. The strength of a fault can be
regarded as the lower-bound strength of the rock mass.
Shearing disintegrates and damages the rock mass and
weathering further weaken the fault strength. Another
example of residual block size is presented in Fig. 7, in
which a sheared Flysches in the middle of the picture are
totally disintegrated with a block volume of about 10 cm3.
In summary, the residual block volumes can be
considered independent of the original (peak) block
volumes for most strain-softening rock masses. This is
illustrated in Fig. 8, showing the fractured residual rock
mass will have more or less the same residual block volume
in the shear band for intact rocks, moderately jointed and
highly jointed rock masses. As an estimate, if the peak
block volume Vb is greater than 10 cm3, then, the residual
block volume V rb in the disintegrated category can be taken
to be 10 cm3. If Vb is smaller than 10 cm3, then, no
reduction to the residual block volume is recommended,
i.e., V rb V b .
3.2.2. Residual joint condition factor
In the GSI system, the joint surface condition is dened
by the roughness, weathering and inlling condition [1,2].
The combination of these factors denes the strength
of a joint or block surface. The joint condition factor is

Moderately
jointed

Highly
jointed

Initial state

Residual state

Fig. 8. Illustration of the residual block volume.

dened as
Jc

J WJS
,
JA

(8)

where JW, JS, and JA are the joint large-scale waviness


factor, small-scale smoothness factor, and alteration factor,
respectively. The tables for peak JW, JS, and JA are given
in [2].
The failure process affects the joint surface condition,
especially the joint roughness. According to [24], the
difference between peak and residual JRC is large if the
peak JRC value is high. The underlying implication is that
the drop of GSI from peak to residual values should be
larger for rock masses with fresh and rough joints.
The major factor that alters the joint surface condition in
the post-peak region is the reduction of joint surface
roughness, as shown in Fig. 3 for the gradual degradation
of JRC. Peak mobilized roughness angle is given as JRC
log(JCS/sn), where JCS is the joint wall compressive
strength, and sn is the normal stress acting on the joint.
The mobilized joint residual roughness is zero according to
the same gure, which can only be achieved when the joint
experiences a very large shearing displacement. On the
other hand, the concept of ultimate mobilized joint
roughness was suggested by Barton et al. [24]. According

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

to Fig. 3, the joint surface roughness is gradually destroyed


during the shearing process and the ultimate mobilized
roughness is about half of the peak roughness (JRCmob/
JRCpeak 0.5). As stated before, the strain levels in most
civil and mining underground excavation structures are not
large so that the shearing of joints at the displacement level
around 10dp would correspond to the straining of jointed
rock masses at the residual strength level dened in Fig. 1.
It is therefore proposed here that the large-scale waviness
and the small-scale smoothness of joints be calculated by
reducing its peak value by half to calculate the residual GSI
value. In a short time period, joint alteration is unlikely to
occur so that the joint alteration factor JA will be
unchanged in most circumstances. However, when water
and clay inll material is involved, the fractured rock
surface can have a lower residual JA.
The residual joint surface condition factor J rc is
calculated from
J rc

J rW J rS
,
J rA

(9)

where J rW , J rS , and J rA are residual values for large-scale


waviness, small-scale smoothness, and joint alteration
factor, respectively. The residual values are obtained based
on the corresponding peak values. The reduction of J rW and
J rS are based on the concept of mobilized joint roughness,
and the equations are given as
If

JW
o1;
2

J rW 1;

Else

J rW

JW
,
2

JS
JS
o0:75; J rS 0:75; Else J rS ,
2
2
There is no reduction for JA in the present study.

If

(10)
(11)

3.3. Residual GSI value and strength parameters


According to the logic of the original GSI system, the
strength of a rock mass is controlled by its block size and
joint surface condition. The same concept is valid for failed
rock masses at the residual strength state. In other words,
the residual GSIr is a function of residual J rc and V rb , i.e.,
GSI r

f J rc ; V rb ,

255

where mr , sr, ar are the residual HoekBrown constants for


the rock mass. It is postulated that these constants can be
determined from a residual GSIr value using the same
equations for peak strength parameters (Eqs. (57)). This
simply means that the equations for peak strength
parameter calculation hold true to the residual strength
parameter calculation. This statement is supported by the
fact that the rock mass in its residual state represents one
particular kind of rock mass in the spectrum in the GSI
chart (Fig. 5). The rock mass spectrum is dened by the
combination of the block volume spectrum and the joint
surface condition factor spectrum. In fact, the GSI chart
had been expanded from its original spectrum [1] to
account for weak or fractured rocks [42,43].
Once the reduced GSIr is obtained, the residual
HoekBrown strength parameters or the equivalent residual MohrCoulomb strength parameters can be calculated, assuming that other parameters such as sc and mi are
unchanged, i.e.,


GSI r  100
mr mi exp
,
(15)
28



GSI r  100
sr exp
,
9

(16)


1  GSIr =15
e
 e20=3 .
(17)
6
Because the rock masses are in a damaged, residual state,
D 0 is used for the residual strength parameter calculation.
ar 0:5

3.3.1. Discussion
When GSI is reduced in the post-peak yielding, the
frictional and cohesive strength components will reduce at
different rates. This can be clearly seen in Figs. 9 and 10.
The frictional strength component, mb, decreases gradually

25

(12)

mi=25
mi=20
mi=15
mi=10

20

25

or, applying the explicitly Eq. (4) to rewrite Eq. (12) as


26:5 8:79 ln J rc 0:9 ln V rb
.
1 0:0151 ln J rc  0:0253 ln V rb

15

(13)

As for the intact rock properties, fracturing and shearing


do not weaken the intact rocks (even if they are broken into
smaller pieces) so that the mechanical parameters (sc and
mi) should be unchanged. What has changed are the block
size and joint surface condition (especially the roughness).
Therefore, the generalized HoekBrown criterion for the
residual strength of jointed rock masses can be written as

ar
s3
s1 s 3 s c m r s r ,
(14)
sc

20

mb

GSI r V rb ; J rc

10
mi=10

0
0

20

40

60

80

GSI
Fig. 9. Relationship between mb and GSI.

100

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

256

with decreasing GSI value. The relationships between GSI


and a and s are shown in Fig. 10. For GSIo40, s, the
cohesive strength component, becomes very small and can
be ignored in dening the residual rock mass strength. For
GSI440, a is approximately 0.5, whereas a is slightly larger
than 0.5 for 20oGSIo40.
Table 1 presents several GSI estimations on the residual
strength parameters of some typical rock masses. For
example, for a strong rock mass whose Vb is 12 500 cm3,
JW 2, JS 2, JA 1, the peak HoekBrown strength
parameters are mb 4.845 and s 0.012. According to the
proposed method, V rb 10 cm3 and J rc 1 for the residual
rock mass. The residual GSIr is 30.3, and the corresponding
residual HoekBrown strength parameters are: mr 1.659,
sr 0. For other rock types, similar strength parameters
can be obtained following our proposed approach. A plot
of the peak and residual HoekBrown strength envelops
given by our approach is presented in Fig. 11 for massive
brittle rocks, jointed strong rocks and jointed intermediate

rocks. The methodology described here provides consistent


results for different rock types considered.
The average residual block size of 10 cm3 is suggested for
the determination of the residual GSI value. To evaluate
the inuence of the residual block size on the residual
strength, maximum (V rb 27 cm3) and minimum (V rb
1 cm3) residual block volumes in the disintegrated category
are used to calculate the equivalent residual MohrCoulomb strength parameters. It is seen from the results
presented in Table 2 that the maximum difference in the
friction angle is about 21 and the difference of the residual
cohesion is small if the maximum and the minimum block

140

Massive brittle rocks (Peak)


Residual
Jointed strong rock (Peak)
Residual
Jointed intermediate rocks (Peak)
Residual

120

100

0.9

0.9

0.8

0.8

0.7

0.7

0.6

0.6

a 0.5

0.5 S

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0
0

20

40

60

80

1 (MPa)

80

60

40

20

0
0

0
100

10

12

3 (MPa)

GSI
Fig. 11. Peak and residual HoekBrown strength envelops for three
typical rock masses.

Fig. 10. Relationship between GSI and a and s.

Table 1
Examples of rock mass residual strength parameters of typical rock masses

JW
JS
JA
Jc
Vb (cm3)
GSI
m
s

Massive brittle rocks


(70oGSIo90)

Jointed strong rock


(50oGSIo65)

Jointed intermediate rocks


(40oGSIo50)

Very weak rock (GSIo30)

Peak

Residual

Peak

Residual

Peak

Residual

Peak

Residual

3
3
1
9
500,000
82.2
10.591
0.138

1.5
1.5
1
2.25
10
37.4
2.138
0.001

2
2
1
4
12,500
60.3
4.845
0.012

1
1
1
1
10
30.3
1.659
0.000

1.5
1.5
2
1.125
6000
45.2
2.805
0.002

1
0.75
2
0.375
10
21.5
1.212
0.000

1
1
4
0.25
100
21.4
1.208
0.000

1
0.75
4
0.1875
10
15.1
0.964
0.000

Note: Peak and residual strength parameters are calculated based on sc 100 MPa and mi 20. We only recommend use of these residual values for
GSIo75. The brittle HoekBrown criterion [44,45] is recommended for GSI475.

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

volumes in the disintegrated category are used. Therefore,


in most applications, it is reasonable to use V rb 10 cm3 to
calculate the residual GSI value and hence the corresponding strength parameters.
For very weak and sheared rock masses such as the
Athens Schist Formation [42] and ysch [43], the peak and
residual block volumes are roughly the same, with an
average block volume of about 1 cm3 and very poor joint
surface condition. The estimated GSI values are in the
range of 515 for this type of rock masses. The volume and
joint surface condition degradation methodology presented
above is able to consistently consider the residual strength
even for the weak rock masses. The validation of the
proposed method using in-situ test data and back analysis
data is presented in the next section.

4. Verication of GSI reduction approach


4.1. Verification from in-situ block shear tests at three
cavern sites in Japan
4.1.1. Kannagawa site
The Kannagawa pumped hydropower project [46] in
Gumma Prefecture in Japan is now under construction
with a maximum output of 2820 MW. The powerhouse
cavern at 500 m depth has a width of 33 m, a height of
52 m, and a length of 216 m. The cavern excavation was
started in 1998 and the last bench was completed in 2000.
The rock mass at the site consists of conglomerate,
sandstone, and mudstone. The rock masses are classied
into ve major groups or domains. Sixty-four uniaxial
compressive tests were conducted to determine the average
strength and standard deviation of each rock type. The
parameter mi for each rock types was obtained from a
limited number of tri-axial tests. A total of 21 block shear
tests were conducted at six test locations. The peak and
residual strength parameters estimated from the GSI
system are given in Table 3, along with the data obtained
from the in-situ block shear tests, for domains CG1, CG2,
FS1 and M1. A residual block volume of 10 cm3 is used in
the calculation. The residual joint surface condition factor
is obtained by degradation of the joint roughness. For
example, J rW JW =2 1:25, J rS JS=2 1 are obtained
for rock CG1. For rock FS1, J rW 1 instead of JW/2
0.75 is used because of the minimum constraint on JW is
that it cannot be smaller than 1 according to the rating [2].
GSIr is calculated using Eq. (13), and cr and fr ( fb+i)
are equivalent residual MohrCoulomb strength parameters calculated from the HoekBrown strength parameters for a s3 range of 05 MPa. The predicted residual
strength in terms of cohesion and friction angle for CG1,
CG2 and FS1 are comparable to the results obtained from
the in-situ block shear tests. For M1 rock mass, the GSI
estimation underestimates the eld residual friction angle.

Table 2
Comparison of residual strength parameters for different residual block
sizes
Residual Jc

Residual block
volume Vb(cm3)

GSIr

Residual strength
parameters
fr

cr (MPa)

2.25

Max. 27
Average 10
Min. 1

39.4
37.4
33.2

51.4
50.9
49.8

1.10
1.04
0.92

Max. 27
Average 10
Min. 1

32.1
30.3
26.5

49.4
48.9
47.6

0.90
0.85
0.77

0.375

Max. 27
Average 10
Min. 1

23.1
21.5
18.1

47.1
46.3
44.7

0.57
0.55
0.51

0.1875

Max. 27
Average 10
Min. 1

16.6
15.1
12.1

43.9
43.1
41.3

0.48
0.46
0.41

257

Note: The calculation of the residual strength parameters is based on


sc 100 MPa and mi 20.

Table 3
Characterization of the rock mass peak and residual strengths at the Kannagawa site using the GSI system
Rock zone

CG1

CG2

FS1

M1

Peak

Residual

Peak

Residual

Peak

Residual

Peak

Residual

GSI system

JW
JS
JA
Jc
Vb (cm3)
GSI
sc (MPa)
mi
c (MPa)
f fb+i (degree)

2.5
2
1
5
309,000
73.8
111
22
4.1
58

1.25
1
1
1.25
10
32.3
111
22
1.1
51.8

1.5
1.5
1
2.25
303,000
64.9
162
19
3.7
57.8

1
0.75
1
0.75
10
27.8
162
19
0.96
51.0

1.5
1.5
1
2.25
295,000
64.8
126
19
3
56.6

1
0.75
1
0.75
10
27.8
126
19
0.96
49.3

1.5
1.5
2
1.125
110,000
53.6
48
9
1.1
42

1
0.75
2
0.38
10
21.5
48
9
0.35
33.2

Block shear test

c (Mpa)
f fb+i (degree)

5.2
57

1.3
52.8

3.4
57

1.3
52.8

3.4
57

0.5
49

1.9
40

0.5
40

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

258

25

Table 4
Characterization of the rock mass peak and residual strengths at the
Kazunogawa site using the GSI system

GSI (peak)
Test data (peak)
GSI (residual)
Test data (residual)

CH rock mass
Peak

Residual

GSI system
JW
JS
JA
Jc
Vb (cm3)
GSI
sc (MPa)
mi
c (MPa)
f fb+i (degree)

2
2
1
4
12,500
60.3
108
19
2.29
54.7

1
1
1
1
10
30.3
108
19
0.87
49

Block shear test


c (MPa)
f fb+i (degree)

1.5
58

0.47
50.3

Shear stress (MPa)

20

15

10

0
0

10

Normal stress (MPa)


Fig. 12. Comparison of peak and residual strength calculated from the
GSI system and eld test data (FS1).

The residual strength estimated from the GSI system


roughly represents the lower bound of the eld test data for
M1 rock mass. Note that the peak and residual strength
parameters determined from the in-situ block shear test
have been used in the cavern design. The displacement and
yielding zone predicted by the FEM analysis agree well
with the eld monitoring data [46].
A comparison of the GSI estimate and the eld test data
for FS1 rock mass is presented in Fig. 12. The average
residual strength estimated from the GSI system is slightly
lower than the eld data average, but is well within the data
variability shown in the eld test data [41].
4.1.2. Kazunogawa site
Kazunogawa power station [47], located in Yamanashi
Prefecture, Japan, at about 500 m depth, has a generating
capacity of 1600 MW. The cavern dimensions are: width
34 m, height 54 m, and length 210 m. The cavern excavation
was started in 1994 and the last bench was excavated in
1996.
The rock mass consists of sandstone and composite
rocks of sandstone and mudstone, described as two groups
(CH and CM) of rock mass types based on the Denken rock
mass classication system [48]. Three joint sets are
observed at this site. The joint spacing of the major joint

set (JEW-h) is in the range of 120 cm. The average joint


spacings of the other two joint sets are 25 and 50 cm,
respectively. Joints are fresh, have small undulation and
are rough. Rough joint surface assessment can also be
indirectly obtained from joint proles in previous laboratory joint test. The block sizes are basically controlled by
the joint frequency of the major joint set. From the joint
density distribution graph, it is seen that the average joint
spacing is about 10 cm for CH rock mass.
Seventy-ve uniaxial compressive tests were conducted
to determine the strength parameters of the intact rocks.
The peak and residual shear strengths of the rock mass
were obtained from 12 in-situ block shear tests. The peak
and residual strength parameters of CH rock mass
estimated from the GSI system are given in Table 4,
along with the data obtained from the in-situ block shear
tests. A method similar to the Kannagawa case is
employed to determine the residual block volume and
joint surface condition factor. The residual GSIr is about
half of the peak GSI value. The predicted residual
strength in terms of cohesion and friction angle is
comparable to the results obtained from the in-situ
block shear tests. A comparison the GSI estimate to the
eld test data is presented in Fig. 13. The GSI system
approach slightly overestimates the cohesion of both
peak and residual strengths.

4.1.3. Okawachi site


Okawachi powerhouse, which is about 280 m deep
underground, has a generating capacity of 4 
320,000 KW. The cavern dimensions are: width 24 m,
height 46.6 m, and length 134.5 m. The cavern excavation
was started in 1988 and the last bench excavation was
completed in 1991. Detailed information about the cavern
construction can be found in Harada et al. [49].

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

20

Table 5
Characterization of the rock mass peak and residual strengths at the
Okawachi site using the GSI system

GSI (peak)
Test data (peak)
GSI (residual)
Test data(residual)

18

CH rock mass
Peak

Residual

GSI system
JW
JS
JA
Jc
Vb (cm3)
GSI
sc (MPa)
mi
c (MPa)
f fb+i (degree)

2.5
2
1
5
13,352.9
62.8
236.7
19
4.45
59.2

1.25
1
1
1.25
10
32.3
236.7
19
1.32
54.8

Block shear test


c (MPa)
f fb+i (degree)

4.53
60.9

1.23
55.1

16

14
Shear stress (MPa)

259

12

10

4
20
GSI (peak)
Test data (peak)
GSI (residual)
Test data (residual)

2
18
0
0

10
16

Normal stress (MPa)


Fig. 13. Comparison of peak and residual strength calculated from the
GSI system and eld test data at the Kazunogawa site (CH).

Shear stress (MPa)

Rock mass around the cavern is porphrite with an


average uniaxial compressive strength of 237 MPa. Three
sets of joints exist at the site with an RQD value that varies
in the range of 6070. Joints are fresh and rough. In-situ
block shear tests were conducted to obtain the peak and
residual shear strength of the jointed rock masses. Plate
loading tests were also conducted to determine the in-situ
deformation modulus of the rock masses. The average
deformation modulus obtained from the eld test is
24.1 GPa, which roughly corresponds to a peak GSI value
of 63.
The peak block volume shown in Table 5 was calculated
using the relationship between the Vb and RQD [50], i.e.,
Vb b ((115-RQD)/3.3)3, where RQD 70 and b 35.
The large-scale roughness (JW 2.5) is determined based
on the data tting by matching the peak strength
parameters of the GSI estimate to the peak strength
parameters from the in-situ tests. This matching excise is
also supported by the good agreement between the
deformation moduli obtained from the GSI system
(21.1 GPa) and the eld test (24.1 GPa). The peak and
residual GSI values are about 63 and 32, respectively. As
can be seen from Table 5 and Fig. 14, the predicted residual
strength in terms of cohesion and friction angle is
comparable to the result obtained from the in-situ block
shear tests.

14

12

10

0
0

10

Normal stress (MPa)


Fig. 14. Comparison of peak and residual strength calculated from the
GSI system and eld test data at the Okawachi site (CH). Note that the
measured and predicted residual strength envelopes are overlapping.

4.2. Verification from a slope stability back-analysis


Back-analysis of the strength and deformation parameters of the rock mass has been applied to many

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

Fig. 15 presents the relationship between the friction


angles and cohesive strengths mobilized at failure for some
slopes [52]. Cohesion is generally small (o 0.2 MPa) and
the friction angle varies between 201 and 451 for most
cases. It is interesting to note that for undisturbed hard
rock masses, (f 40451 in Fig. 15) the cohesion is in the
range of 0.30.5 MPa. Based on our experience, we
consider the back-analyzed c and f values representative
of the residual strength parameters.
Sjoberg [36] used the HoekBrown strength criterion to
estimate the strength of the rock mass at the Aznalcollar
open pit mine located in southern Spain. The dominant
footwall rock types are slates and schist with welldeveloped cleavage. At the end of mining, the pit was
approximately 1300  700 m in area and 270 m deep with
an overall slope angles varied from 301 to 381. Despite the
relative moderate slope, the mine has suffered several largescale failures of the footwall slope. Failure was not
structurally controlled but rather stress controlled. The
failure surfaces were identied from the slope monitoring
using techniques such as surfaces displacement stations and
inclinometers.

Undisturbed soil and


jointed rock masses
with relatively low
clay mineral content

Residual strength of
slickensided surfaces
coated with high clay
mineral content materials

Disturbed material with


rounded weakly cemented
particles and appreciable
clay mineral content

engineering projects [51]. It is especially useful when failure


has occurred and reached to the residual state as is in the
case of slope instability. It is tempting to consider the
possibility of back analyzing existing slope failures in order
to determine the shear strengths that must have been
mobilized in the full-scale rock mass at the time of the
failure. In fact, back-analysis of slope failures can be a very
important source of shear strength data [52].
In back-analysis of slope stability, the shear strength
parameters, c and f are adjusted till the factor of safety is
unity (1.0) as a prerequisite for failure in a limit equilibrium
analytical model. This pair of parameters can be considered
as the residual strength parameters. This is so because it is
generally required that the rock mass must experience a large
deformation in excess of that required to mobilize the peak
strength. Thus, the resistance mobilized by reactivated
landslides is equal to the residual strength of the material
within the slip zone. This is, however, only valid for rotational
or sliding failure involving the entire failure volume, not for
progressive failures. Thus, a back-analysis using limit
equilibrium method or FEM/DEM employing strength
reduction method can obtain residual strength parameters.

Rockmasses or dumps
containing hard clean
angular interlocking
particles and blocks

260

Cohesion - MPa

0.5

0.4

Undisturbed hard rock


masses with no major
structrural patterns
dipping towards slope

0.3

Undisturbed hard rock


masses with no throughgoing structures dipping
towards slope
Undisturbed rock masses
with a few structures
dipping towards slope

0.2
Soft rock masses or
jointed hard rock
disturbed by blasting
or excess loading
0.1
Weatherd soft rock or
discontinuities in hard rock
Clay
Soil
Sand

0.0
0

10

15

20
25
30
35
Friction angle - degrees

40

45

50

Fig. 15. Relationship between the friction angles and cohesive strengths mobilized at failure for the some slopes [52].

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

Table 6
Characterization of the rock mass peak and residual strengths at the
Aznalcollar open pit mine footwall using the GSI system

GSI system
JW
JS
JA
Jc
Vb (cm3)
GSI
sc (MPa)
mi
c (MPa)
f fb+i (degree)

GSI (peak)
GSI (residual)

Back analysis (lower bound residual)


Back analysis (higher bound residual)

Schist-foliation

Peak

Residual

Peak

Residual

1.5
1.5
1
2.25
100,000
61.0
25
9
0.97
38.8

1
0.75
1
0.75
10
27.8

2
1.5
2
1.5
150,000
57.9
25
8
0.87
37

1
0.75
2
0.375
10
21.5

0.39
30.1

Back-analysis using limit equilibrium method [36]


c (MPa)
00.3
2535
f fb+i (degree)

0.27
27.2
00.12
2230

6
Shear stress (MPa)

Slate

261

GSI (peak)

5
GSI (residual)

Back analysis (average)


3

0
0

Normal stress (MPa)

Because failures were not structurally controlled, the


continuum numerical tool FLAC was used to simulate the
slope failure by using a perfectly plastic material model
[36]. It was found that by assuming disturbed rock mass
parameters, good agreement could be achieved between
estimated strength values and back-calculated strengths
from observed slope failures in the footwall. The presence
of the stiff and strong pyrite prevented the failure to initiate
at the toe. The failure was developed rather inside the
slope. At the stage when the toe buttress zone reached
critical state, the post failure state was probably reached
and hence the calibrated strength values were representative of the residual rock mass strength.
No direct joint spacing and surface condition were
available in the report by Sjoberg [36]. However, peak GSI
values, inferred from the RMR values, were given. The
representative GSI values for slate and schist-foliation are
61 and 58, respectively. As listed from Table 6, the GSI
values can be estimated by using the good and fair joint
surface condition for slate and schist and their corresponding block volume (back-tted from known GSI value),
respectively. Using the method developed in this study, we
can estimate the peak and residual strength of the rock
masses. From the back-analysis, the residual cohesion for
the slate is found in the range of 00.3 MPa and the
residual friction angle in the range of 25351. The estimated
cohesion and friction angle for the same rock mass, using
the GSI reduction approach, are 0.39 MPa and 301,
respectively. As can be seen from Fig. 16, the estimated
residual strength of the slate is well within the lower and
upper bounds indicated from the back-analysis.
Back-analysis data from slope stability provides excellent in-situ data for method validation. As more data
becomes available, the proposed method for rock mass
residual strength estimation can be further validated.

Fig. 16. Comparison of the residual strength calculated from the GSI
system and back calculated data at the Aznalcollar open pit mine footwall
(slate).

4.3. Discussion of results


Traditionally, the determination of mechanical properties of jointed rock masses in Japan and other countries is
achieved through well planned and executed in-situ block
shear test and plate-loading test. Such tests are expensive
and time consuming. Most importantly, results only
become available once underground access has been
established. An alternative to the test approach is the use
of a rock mass classication system such as the GSI system
to provide design parameters early in the design phase and
reduce the need for extensive in-situ testing. Nevertheless,
in-situ tests can be used to verify the GSI prediction or the
observational (back-analysis) method [53] will be required
to conrm the GSI predictions.
The quantitative approach uses the block volume and
joint surface condition factor to determine both the peak
and residual GSI values. These input parameters in the
validation examples were obtained from eld mapping and
from borehole logging data. The strength and deformation
parameters estimated from the GSI system are very close to
those obtained from in-situ tests or back analysis,
indicating that the GSI system can be effectively applied
to the design of underground caverns and rock slopes.
The degradation of the block volume and the joint
surface condition for CG2 rock mass at the Kannagawa
site is graphically presented in Fig. 17. GSI is reduced from
a peak value of 64.9 to a residual value of 27.8. The gradual
decrease of the GSI value can be linked to the post-peak
strain softening of the rock mass (see Fig. 1(a)). Future

ARTICLE IN PRESS
262

M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

Blocky - very well interlocked


undisturbed rock mass consisting
of cubical blocks formed by three
orthogonal discontinuity sets
Joint spacing 30 - 100 cm

90
80
70
60
50

90

Very poor
Slickensided, highly weathered surfaces with
soft clay coatings or fillings

Poor
Slickensided, highly weathered surfaces with
compact coating or fillings of angular fragments

Fair
Smooth, moderately weathered or
altered surfaces

85

1E+6
(1m3)
Degradation of joint surface condition

75
80

Peak
65

70

10 cm

55

10E+3

45
50

Residual

35

dation

1000
(1 dm3)

40
25
30

100

of GS

Blocky/disturbed - folded and/or


faulted with angular blocks formed by
many intersecting discontinuity sets
5
Joint spacing 3 - 10 cm

60

100E+3

Peak

Degra

Very Blocky - interlocked, partially


disturbed rock mass with multifaceted
20
angular blocks formed by four or more
discoutinuity sets
Joint spacing 10 - 30 cm

Degradation of block volume

40
30 cm

10E+6
95

15

Disintegrated - poorly interlocked,


heavily broken rock mass with a
mixture or angular and rounded
rock pieces
Joint spacing < 3 cm

Block Volume Vb (cm3)

100 cm

Good
Rough, slightly weathered,
iron stained surfaces

Block Size
Massive - very well interlocked
undisturbed rock mass blocks formed
by three or less discontinuity sets
with very wide joint spacing
150
Joint spacing > 100 cm

Very good
Very rough, fresh unweathered surfaces

Joint or Block Wall Condition

20

10

Residual
10
1 cm

Foliated/laminated/sheared - thinly
laminated or foliated, tectonically sheared
N/A
N/A
weak rock; closely spaced schistosity
prevails over any other discontinuity set,
resulting in complete lack of blockiness
12
4.5
1.7
Joint spacing < 1 cm

0.67
0.25
Joint Condition Factor Jc

0.1
0.09

Fig. 17. Degradation of the block volume and joint surface condition of CG2 rock mass from peak to residual state.

research will address the issue of the rate of GSI value


decrease associated with the plastic strain.
The ratios of residual GSIr to peak GSI depend on the
peak GSI values, as shown in Fig. 18. The investigated case
histories have peak GSI values between 40 and 80 and the
GSIr/GSI ratios vary from 0.37 to 0.51. The point with
a low GSI value of 21 is adopted from Table 1 with the
GSIr/GSI ratio obtained by our proposed method. For
very weak rock masses, the residual GSIr is equal to the
peak GSI. If a trend line is drawn, it should pass through

the point (0,1). A trend line by forcing it to pass through


point (0,1) is hence obtained (as shown in Fig. 18). The
residual GSIr value can then be empirically expressed as a
function of the peak GSI value as
GSI r GSI e0:134GSI .

(18)

Russo et al. [7] suggested that the residual GSIr value is


36% of the peak GSI value. This is represented as a
horizontal line in Fig. 18. It is observed that their
suggestion may underestimate the residual GSIr values

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

[44] should be used and proper test program be planned for


the determination of the residual strengths. Furthermore, if
the rock mass fails by block rotation and local crushing,
probably a different analysis approach such as UDEC or
3DEC should be considered instead of a continuum
analysis. The users must be aware of the limitations when
applying the GSI system and the methodology for
determining the peak and residual strength parameters
using this quantitative approach.

1
0.9
0.8
GSIr / GSI

0.7
0.6
0.5
0.4

GSIr = 0.36GSI

0.3
-0.0134GSI

0.2

263

GSIr / GSI = e

0.1

5. Conclusions

0
0

20

40

60

80

100

GSI
Fig. 18. Relationship between GSIr/GSI ratio and GSI.

for poor quality rock masses (e.g., GSIo40). For very


good quality rock masses (GSI480), their suggestion may
overestimate the residual GSIr values.
To obtain reliable results of the residual GSIr value, the
method proposed in Section 3.2 should be followed, i.e., by
obtaining the residual block volume and joint surface
condition factor and using the GSI chart or Eq. (13) to
calculate the residual GSIr. For quick estimates, Eq. (18)
can also be utilized if the peak GSI value is known.
Because very large straining is needed to reach the true
residual state, the residual GSIr value discussed here refers
to the post-peak strength in a limited straining range. In
the design of underground structures, most residual
strength parameters utilized are in fact the residual postpeak strength parameters representing limited post-peak
deformation.
The residual strength of intact rocks, as interpreted from
the triaxial test, is at the same level as the residual strength
of the jointed rock mass. In the low connement range, the
residual cohesion and friction angle of the Tennessee
Marble are 2.4 MPa and 51.61, respectively. As can be seen
from Table 3 and Table 5, the residual strengths of some of
the hard jointed rock masses (CG1 conglomerate and CH
porphrite) are roughly at the same level of the residual
strength of intact rocks (Tennessee Marble), suggesting
that our assumption of the independence of the residual
block volume on the original jointing state is valid.
As stated previously, if the peak block volume is small
(o10 cm3), the residual block volume is equal to the peak
block volume and the same approach outlined above can
be applied to the estimation of the strength parameters. In
this fashion, consistent estimation of the both peak and
residual strength parameters can be obtained.
The proposed method is applicable to most rock types
when failure is dominated by shear failure. Care must be
given for brittle failure of massive rocks involving spalling
failure and very weak rocks that have been over
consolidated or re-bonded. In such a case, special
failure criteria such as brittle HoekBrown failure criterion

It is observed from laboratory and eld test data that


following the strain-softening behavior of rocks under
loading, the residual strength represents more or less the
mobilized shear strength along a surface or shear zone of
the fractured rock. The post-peak strength depends on the
resistance developed on the failure plane (zone) against
further straining. Initially, the fracture orientation, degree
of interlocking, surface irregularity or roughness will affect
the post-peak load level. However, as the straining
continues, the residual strength is less dependent on these
factors.
The Geological Strength Index (GSI) system is a
universal rock mass classication system. It is a rock mass
classication system that is directly linked to engineering
parameters such as MohrCoulomb or HoekBrown
strength parameters or rock mass modulus. The current
GSI system guidelines, however, are for the estimation of
the peak strength and do not include guidelines for the
estimation of the rock mass residual strength that yield
consistent results. A new method is proposed here to
extend the GSI system for the estimation of rock masss
residual strength. The peak GSI value is reduced based on
the reduction of the two major controlling factors in the
GSI system, i.e., residual block volume V rb and residual
joint condition factor J rc , to obtain the residual GSIr value.
The residual block volume is found to be in the category of
the disintegrated rocks in the GSI chart, characterized
by the facts that the failed rock masses at the residual
strength level are poorly interlocked, heavily broken with a
mixture of angular and rounded rock pieces. The average
block size of 10 cm3 is suggested for the residual GSIr value
estimation. For joint surface condition, the major factor
that alters the condition in post-peak region is the
reduction of joint surface roughness. The actual degradation of the joint surface is based on the concept of
mobilized residual joint roughness suggested by Barton
et al. [24]. The large-scale waviness and the small-scale
smoothness of joints can be calculated by reducing their
peak values by half with conditions to meet the minimum
values. The joint alteration factor JA is assumed unchanged. The residual GSIr value is calculated from the
relationship between GSIr and V rb and J rc .
It has also been assumed that the intact rock properties
such as sc and mi remain unchanged as the rock mass
changes from its peak to residual state. Hence, the residual

ARTICLE IN PRESS
264

M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265

strength parameters are calculated using the same form of


the generalized HoekBrown strength criterion. The
equivalent MohrCoulomb strength parameters are calculated based on the HoekBrown strength parameters.
The proposed method for the estimation of rock mass
residual strength is validated using in-situ block shear test
data from three large-scale cavern construction sites and
the data from the back-analysis of a rock slope stability.
The estimated residual strengths, calculated using the
residual GSIr value, are in good agreement with eld test
data or back analyzed data. The proposed method for
residual strength estimation extends the GSI system and
adds quantitative means to determine the complete set of
rock mass properties needed for design.
When applying the GSI system to a numerical simulation, the users must be aware of the limitation of the
approach related to quantifying a discontinuous rock mass
in a continuum-modeling framework. In certain circumstances, a discontinuous analysis tool, rather than continuum models with parameters obtained by the GSI
system, should be used. In addition, one needs to be aware
of the mechanical instability problem associated with
strain-softening materials in continuum elasto-plastic
analyses. The simulation results could be highly dependent
on the mesh size and slight change of material parameters;
hence the uniqueness of a solution can often not be
guaranteed. Although the paper provides a contemporary
method for rock masss peak and residual strength
parameter determination, its successful application relies
heavily on the professional judgment, as is typically the
case in rock mechanics and rock engineering.
Acknowledgements
This study was funded by Tokyo Electric Power Services
Co. Ltd (TEPSCO). The authors wish to thank Tokyo
Electric Power Company (TEPCO) for providing access to
test sites and test data and permitting to publish the results.
The authors also thank Evert Hoek for his valuable
comments and suggestions during the preparation of the
manuscript.
References
[1] Hoek E, Kaiser PK, Bawden WF, Hoek E, Kaiser PK, et al. Support
of underground excavations in hard rock. Rotterdam: Balkema;
1995.
[2] Cai M, Kaiser PK, Uno H, Tasaka Y, Minami M. Estimation of rock
mass strength and deformation modulus of jointed hard rock masses
using the GSI system. Int J Rock Mech Min Sci 2004;41(1):319.
[3] Oda M. A method for evaluating the effect of crack geometry on the
mechanical behavior of cracked rock masses. Mech Maters
1983;2:16371.
[4] Amadei B. Strength of a regularly jointed rock mass under biaxial
and axisymmetric loading. Int J Rock Mech Min Sci Geomech Abstr
1988;25(1):313.
[5] Cai M, Horii H. A constitutive model of highly jointed rock masses.
Mech Maters 1992;13:21746.
[6] Hoek E, Practical rock engineering. www.rocscience.com, 2000.

[7] Russo G, Kalamaras GS, Grasso P. A discussion on the concepts of


geomechanical classes behavior categories and technical classes for an
underground project. Gallerie e Grandi Opere Sotterranee 1998;54.
[8] Paulding BW, Crack growth during brittle fracture in compression.
PhD. thesis, MIT, Cambridge, MA, 1965.
[9] Cook NGW. The failure of rock. Int J Rock Mech Min Sci 1965;
2(4):389403.
[10] Hoek E. Rock fracture under static stress conditions. Pretoria, South
Africa: National Mechanical Engineering Research Institute, CSIRO;
1965.
[11] Bieniawski ZT. Mechanism of brittle fracture of rock, Parts I, II and
III. Int J Rock Mech Min Sci Geomech Abstr 1967;4(4):395430.
[12] Wawersik WR. Detailed analysis of rock failure in laboratory
compression tests. PhD. thesis, University of Minnesota, Minneapolis, MN, 1968.
[13] Wawersik WR, Fairhurst C. A study of brittle rock fracture in
laboratory compression experiments. Int J. Rock Mech Min Sci
Geomech Abstr 1970;7:56175.
[14] Rummel F, Fairhurst C. Determination of the post failure behaviour
of brittle rock using a servo-controlled testing machine. Rock Mech
1970;2:189204.
[15] Seeber G. Druckstollen und Druckschachte. Stuttgart: Enke; 1999.
[16] Mogi K. Pressure dependence of rock strength and transition from
brittle fracture to ductile ow. Bull Earthquake Res Inst Univ Tokyo
1966;44:21532.
[17] Hudson JA, Brown ET, Fairhurst C. Shape of the complete
stressstrain curve for rock. In: Proceedings of the 13th symposium
on rock mechanics, 1971. p. 77395.
[18] Maurer WC. Shear failure of rock under axial and hydrostatic
pressure. In: Proceedings of the rst international congress ISRM,
vol. 1, 1966, p. 33741.
[19] Lundborg N. Triaxial shear strength of some Swedish rocks and ores.
In: Proceedings of the rst International Congress ISRM, vol. 1,
1966, p. 2515.
[20] Patton FD. Multiple modes of shear failure in rock and related
materials. PhD. thesis, University of Illinois, Urbana-Champaign,
1966.
[21] Goodman RE. Methods of geological engineering in discontinuous
rocks. St Paul: West Publication; 1976.
[22] Barton N, Choubey V. The shear strength of rock joints in theory and
practice. Rock Mech 1977;10:154.
[23] Bandis SC, Lumsden AC, Barton NR. Fundamentals of rock joint
deformation. Int J Rock Mech Min Sci Geomech Abstr
1983;20(6):24968.
[24] Barton NR, Bandis SC, Bakhtar K. Strength, deformation and
conductivity coupling of joints. Int J Rock Mech Min Sci Geomech
Abstr 1985;22(3):12140.
[25] Van Heerden WL. In situ complete stressstrain characteristics of
large coal specimens. J South Afr Inst Min Metall 1975;75(8):20717.
[26] Curran JH, Corkum BT. Phase2-2D nite element program for
calculating stresses and estimating support around underground
excavations. Rock Engineering Group, University of Toronto, 1997.
[27] Tasaka Y, Uno H, Omori T, Kudoh K. Numerical analysis of
underground powerhouse excavation considering strain softening and
failure of joints. In: Proceedings of the 10th Japan Symposium on
Rock Mechanics, 1998, p. 57580.
[28] FLACfast Lagrangian analysis of continua. Minneapolis: Itasca
Consulting Group; 2000.
[29] Phase2. Rocscience Inc. Toronto: Rocscience; 2004.
[30] Hoek E. Rock mass properties for underground mines. In: Underground mining methods: engineering fundamentals and international
case studies. Soc Min Metall Explor (SME) 2001.
[31] Ribacchi R. Mechanical tests on pervasively jointed rock material:
insight into rock mass behaviour. Rock Mech Rock Eng 2000;
33(4):24366.
[32] Crowder JJ, Bawden WF. Review of post-peak parameters and
behaviour of rock masses: current trends and research. RocNews
2004;Fall:13.

ARTICLE IN PRESS
M. Cai et al. / International Journal of Rock Mechanics & Mining Sciences 44 (2007) 247265
[33] Hoek E, Brown ET. The HoekBrown failure criteriona 1988
update. In: Curran JC, editor. Rock engineering for underground
excavations. Proceedings of the 15th Canadian rock mechanics
symposium. Toronto: University of Toronto; 1988. p. 318.
[34] Cai M, Kaiser PK. Visualization of rock mass classication systems.
Geotech Geolog Eng 2006;24(4):1089102.
[35] Hoek E, Carranza-Torres C, Corkum B. HoekBrown failure
criterion2002 edition. In: Proceedings of the fth North American
rock mechanics symposium, vol. 1, 2002, p. 26773.
[36] Sjoberg J. Estimating rock mass strength using the HoekBrown
failure criterion and rock mass classicationa review and application to the Aznalcollar open pit. Division of Rock Mechanics,
Department of Civil and Mining Engineering, Lulea University of
Technology, 1997.
[37] Pollard DD, Aydin A. Progress in understanding jointing over the
past century. Geol Soc Amer Bull 1988;100:1181204.
[38] Bieniawski ZT. Rock mass classication in rock engineering. In:
Bieniawski ZT, editor. Proc. Symp Explor Rock Eng, vol. 1.
Rotterdam: Balkema; 1976. p. 97106.
[39] Roberts DP, Sellers EJ, Sevume C. Numerical modelling of fracture
zone development and support interaction for a deep level tunnel in a
stratied rockmass. In: Hagan TO, editor. SARES 99. SANIRE;
1999. p. 26472.
[40] Sellers EJ, Berlenbach J, Schweitzer J. Fracturing around deep level
stopes: comparison of numerical simulation with underground
observations. In: Rossmanith H, editor. Mechanics of jointed and
faulted rock. Rotterdam: Balkema; 1998. p. 42530.
[41] Cai M, Kaiser PK. Determination of the residual strength of jointed
rock masses using the GSI system. In: Report to TEPSCO. Sudbury,
Ont.: Geomechanics Research center, Laurentian University; 2005.
[42] Hoek E, Marinos P, Benissi M. Applicability of the geological
strength index (GSI) classication for very weak and sheared rock

[43]
[44]

[45]

[46]

[47]
[48]
[49]

[50]
[51]
[52]
[53]

265

masses. The case of Athens Schist Formation. Bull Eng Geol Enviorn
1998;57:15160.
Hoek E, Marinos P. Predicting tunnel squeezing problems in weak
heterogeneous rock masses. Tunnels Tunnel 2000.
Martin CD, Kaiser PK, McCreath DR. HoekBrown parameters for
predicting the depth of brittle failure around tunnels. Can Geotech J
1999;36(1):13651.
Kaiser P.K., Diederichs M.S., Martin C.D., Sharp J., Steiner W.,
Underground works in hard rock tunnelling and mining.
In: Proceedings of the GeoEng 2000, vol. 1. Technomic Publication;
2000. p. 841926.
Maejima T, Morioka H, Mori T, Aoki K. Evaluation of the loosened
zone on excavation of the large underground rock cavern. In: Adachi
T, et al., editors. Modern tunnel science and technology. Rotterdam:
Balkema; 2001. p. 10338.
Koyama T, Nanbu S, Komatsuzaki Y. Large-scale cavern at a depth
of 500 m. Tunnel Underground 1997;28(1):3745.
Tanaka H. Introduction to geology for civil engineers. Tokyo:
Sankaidou; 1964.
Harada M, Katayama T, Yada A. Design and construction of the
underground cavern of Okawachi pumped-storage powerhouse. Elec
Power Civ Eng 1991;230:4657.
Palmstrm A. RMia rock mass characterization system for rock
engineering purposes. PhD. thesis, University of Oslo, Norway, 1995.
Sakurai S, Takeuchi K. Back analysis of measured displacement of
tunnels. Rock Mech Rock Eng 1983;16:17380.
Hoek E, Bray JW. Rock slope engineering. London: Inst Min Metall;
1981.
Kaiser PK. Observational modeling approach for design of underground excavations. In: Proceedings of the international workshop
on observational method of construction of large underground
caverns in difcult ground conditions, 1995. p. 117.

You might also like