Tunnelling and Underground Space Technology 70 (2017) 375–387

Contents lists available at ScienceDirect

Tunnelling and Underground Space Technology

journal homepage: www.elsevier.com/locate/tust

Probabilistic analysis of tunnels considering uncertainty in peak and post- MARK

peak strength parameters
⁎
Gaurav Tiwari , Bhardwaj Pandit, Gali Madhavi Latha, G.L. Sivakumar Babu
Department of Civil Engineering, Indian Institute of Science, Bangalore, India

A R T I C L E I N F O A B S T R A C T

Keywords: Stability analysis of rock tunnels is a complex problem because of various types of uncertainties present in the
Rock tunnels rock mass properties and hence probabilistic approaches are used to systematically consider these uncertainties
Residual strength in the analysis. While the uncertainty in deformation modulus and peak strength parameters has been considered
Probabilistic analysis previously it has been observed that the uncertainty in post-peak strength parameters is generally neglected in
earlier studies. Post-peak strength parameters are among the most important factors which significantly influ-
ence the plastic zone development and displacements around the tunnel and hence neglecting uncertainty in
post-peak strength parameters is not appropriate. In the current study, a quantitative approach based on the
Geological Strength Index (GSI) has been used to estimate the uncertainty in peak strength, residual strength and
deformation parameters. Then the uncertainty in yield zone and displacement around the tunnel is estimated
using Hong’s point estimate method coupled with finite element method. The approach was used to estimate the
displacements around tunnels of different shapes in three case studies in average quality rock mass and the
predicted displacements were compared with the in-situ measurements. A comparison is provided with the
generally adopted conventional probabilistic approach in which uncertainty in peak strength parameters and
deformation modulus is considered and rock mass is assumed as elastic-perfectly plastic. It was observed that
predicted displacements were matching well with the measured displacements for all the tunnels by considering
uncertainty in residual strength parameters while displacements were underestimated when the conventional
approach was used. A parametric study was conducted to estimate the influence of Coefficient of Variation
(COV) of different parameters on the plastic zone development and displacements of rock tunnel by selecting one
of the case studies. It was observed that COV of yield zone depth and displacement was varying with the COV of
the intact rock properties and thus influencing the probability of failure.

1. Introduction
Precise determination of rock mass properties is very difficult be-
cause of the presence of different sources of uncertainties. Some
common types of uncertainties include spatial, statistical and sys-
tematic uncertainties emerging from in-situ testing, laboratory testing,
joint mapping, joint condition, etc. (Duzgun et al., 2002; Cai, 2011).
Though it is well established that probabilistic methods provide a ra-
tional and efficient means of considering these uncertainties in the
analysis systematically still the application of probabilistic methods in
the field of rock mechanics is not very common (El-Ramly et al., 2002;
Abdellah et al., 2014). One of the most important fields in rock me-
chanics for which probabilistic analysis has become an important tool is
rock tunnels stability analysis. It has been observed from earlier studies
that while uncertainty in deformation properties and peak strength
properties are considered in the analysis (Cai, 2011; Idris et al., 2015),
uncertainty in post-peak strength parameters are not properly con-
sidered in the analysis which can be an important factor.
Studies on post-peak strength behavior of intact rocks were started
in late 60s and early 70s after the development of stiff servo-controlled
testing machines (Rummel and Fairhurst, 1970; Wawersik and
Fairhurst, 1970; Hudson et al., 1971; Wawersik and Brace, 1971). After
these earlier studies, many studies have been conducted on intact rock
samples to investigate the effect of confining stress and rock mineralogy
on failure patterns, strength and deformational behavior of rock
(Santarelli and Brown, 1989; Yang et al., 2012; Arzua and Alejano,
2013; Walton et al., 2015). However, results of the intact rock behavior
could not be applicable in the in-situ conditions because the intact rock
samples tested in the laboratory are free from large scale defects which
might govern the behavior of the rock mass (Wawersik and Brace,

⁎
Corresponding author.
E-mail addresses: gaurav@civil.iisc.ernet.in (G. Tiwari), bpandit@civil.iisc.ernet.in (B. Pandit), madhavi@civil.iisc.ernet.in (G.M. Latha),
gls@civil.iisc.ernet.in (G.L. Sivakumar Babu).

http://dx.doi.org/10.1016/j.tust.2017.09.013
Received 10 February 2017; Received in revised form 6 September 2017; Accepted 7 September 2017
0886-7798/ © 2017 Elsevier Ltd. All rights reserved.
G. Tiwari et al.
1971). As compared to the intact rocks, studies on post-peak behavior
of jointed rocks are limited (Ribacchi, 2000; Crowder and Bawden,
2004; Tiwari and Rao, 2006a, 2006b). Because of the scarcity of the
studies and methods to quantify post-peak strength parameters of rock
mass, uncertainties in post-peak strength parameters were not properly
considered in the earlier studies (Hoek, 1998; Cai, 2011; Langford and
Diederichs, 2013). However, the literature suggests that the deforma-
tion and plastic radius around tunnels for average quality rock mass i.e.
rock mass with GSI ranging between 35 and 65 is highly dependent on
post-peak strength parameters (Cai et al., 2007; Alejano et al., 2009,
2012). Hence, ignoring uncertainty in post-peak strength parameters
may underestimate the coefficient of variations of displacements and
yield depth around the tunnel which may lead to underestimation of
probability of failure. Song et al. (2016) used analytical method with
First Order Reliability Method (FORM) to estimate the probability of
failure for a circular tunnel. However it is not easy to develop analytical
methods for all tunnel shapes, complicated stress and ground condi-
tions. Hence, there is a need for an approach based on numerical
method for the probabilistic analysis of tunnels which can consider
uncertainty in peak and post-peak strength parameters in average
quality rock mass.
In the current article, a Geological Strength Index (GSI) based
quantitative approach was used to carry out probabilistic analysis by
considering uncertainty in deformation modulus, peak and residual
strength parameters. Sensitivity analysis was carried out in-order to
show the importance of residual strength parameters along with peak
376
Tunnelling and Underground Space Technology 70 (2017) 375–387
Fig. 1. Description of different parameters required for estima-
tion of peak GSI (a) Rock joint roughness parameters (b) Rock
block volume demarcated by three joints.
strength parameters and deformation modulus on the response para-
meters (Yield zone depth and Displacements). The method suggested by
Cai et al. (2007) was included in the probabilistic framework to esti-
mate the uncertainty in the residual strength parameters, which were
later used to estimate the uncertainty in yield zone depth and dis-
placements around the tunnel. The approach was demonstrated using
three case studies of tunnels with different shapes and stress conditions
situated in average quality rock mass to verify the applicability of this
approach. A comparison is provided between estimated results using
current approach and conventional approach. In the conventional ap-
proach, uncertainty in peak strength parameters and deformation
modulus was considered and post-peak behavior of rock mass was as-
sumed as deterministic i.e. elastic-perfectly plastic. It was observed that
for the practical applications, the method can be used for determination
of distributions of yield zone depth and displacements for tunnels of
complex shapes and stress conditions with the determination of ap-
propriate range for GSI.
2. Calculation of variability in strength properties and
deformation modulus
Estimation of statistical parameters and Probability Density
Functions (PDFs) of peak and residual strength parameters and de-
formation modulus is discussed in this section. In the first step, statis-
tical parameters of peak and residual GSI were estimated which were
later used along with statistical parameters of intact rock properties to
G. Tiwari et al.
estimate statistical parameters of peak and residual strength parameters
and deformation modulus.
2.1. Determination of peak GSI and residual GSI
Because of the difficulties involved in conducting the in-situ tests to
determine the strength of rock mass, efforts were made over the years,
to relate the strength parameters of rock mass to the rock mass classi-
fication systems like Rock Mass Rating (RMR) or Geological Strength
Index (GSI). GSI is an established rock mass classification system, which
is used along with intact rock strength properties to obtain strength
parameters of the rock mass. Some researchers tried to provide a
quantitative way to estimate the GSI using parameters based on block
size and discontinuity condition for better classification of rock masses
for engineering purposes (Sonmez and Ulusay, 1999; Cai et al., 2004).
In this study, the method suggested by Cai et al. (2004) was used to
estimate the peak GSI value (GSIp ) i.e. GSI value used to estimate peak
strength parameters and deformation modulus. Peak condition of the
rock mass corresponds to rock mass with negligible disturbance/dis-
placement i.e. before tunnel excavation. Equations required to calculate
GSIp value for the rock mass are given below
26.5 + 8.79 ln JCp + 0.9 ln Vbp
GSIp (JCp,Vbp) =
1 + 0.0151 ln JCp−0.0253 ln Vbp (1)
JCp
where is the peak joint condition factor, Vbp
is the peak block volume.
Peak Joint condition factor ( JCp ) can be estimated using the equation
below
JWp JSp
JCp =
JAp (2)
where Jwp
is the peak large-scale waviness factor for joint length in
meters (varying between 1 and 10 m), JSp is the peak small-scale
roughness factor for joint length in centimeters (varying between 1 and
20 cm), JAp is the peak alteration factor. Fig. 1a shows the definition of
large scale waviness and small-scale roughness of a rock joint.
Peak block volume of a rock block determined by 3 joints can be
estimated using equation given below (Palmstrom, 2005; Cai et al.,
2004)
s1 × s2 × s3
Vbp =
Sin ϒ1 × Sin ϒ2 × Sin ϒ3 (3a)
where Vbp is the peak block volume, s1,s2,s3 are the joint spacings be-
tween individual joints for different joint sets, ϒ1,ϒ2,ϒ3 are the angles
between the joint sets. Fig. 1b shows the rock block demarcated by
different joint sets and definition of the parameters required for the
estimation of block volume.
However, compared to the joint spacing, the effect of the intersec-
tion angle between joint sets is relatively small. Hence, for practical
purpose, the block volume can be approximated as
Vbp = s1 × s2 × s3 (3b)
If the joint spacing of individual joint sets are not available,Vbp
can
be estimated using an empirical relation with volumetric joint count as
given below (Palmstrom, 1995)
Vbp = βJV−3 (4a)
1 1 1 JSp
JV = + +
s1 s2 s3 (4b)
where Vbp
is the peak block volume, β is the block shape factor and was
assumed to be 36 as suggested by Palmstrom (2005), Jv is volumetric
joint count (joints/m3), s1,s2,s3 are the joint spacings between individual
joints for different joint sets as shown in Fig. 1b.
Since, the joint length in a rock mass can be hundred meters; Joint
roughness can vary along a single joint at different locations which
introduces spatial variability in joint roughness mapping. Additional
377
Tunnelling and Underground Space Technology 70 (2017) 375–387
uncertainty can also be introduced by human errors in field mapping.
Joint set is defined as the collection of parallel oriented joints as shown
in Fig. 1b. Joint spacing is the perpendicular distance between two
joints within a joint set as shown in Fig. 1b. Spacing between the joints
along a particular joint set generally varies even if they are parallel
which results in the variation of volume of block.
Statistical parameters and Probability Density Functions (PDFs) of
JWp ,JSp and JAp and spacing of the joint sets can be estimated from careful
field investigations. Probability Density Functions (PDFs) of JCp and Vbp
can be estimated by Monte-Carlo simulation in MATLAB using Eqs. (2)
and (3b)/(4a) respectively. After obtaining statistical parameters and
PDFs of JCp and Vbp , statistical parameters and PDF of GSIp can again be
estimated using Eq. (1).
The residual strength in a stress-strain curve of rock mass is gen-
erally defined by the plateau after the peak strength, in a strain range of
about 5–10 times the strain corresponding to the peak strength at which
the stress remains constant even when the strain increases. Cai et al.
(2007) proposed that residual strength parameters of the rock mass can
be estimated using a similar method used for estimation of peak
strength parameters. It was proposed that as the displacement in the
rock mass increases towards residual stage, two phenomena occur, peak
rock block volume (Vbp) reduces due to shear and tensile fracturing in
intact rock; peak joint roughness (both large scale waviness and small
scale roughness) reduces due to shearing of the asperities. Fig. 2 shows
the summary of block volume reduction and joint roughness degrada-
tion at the residual stage of rock mass. It was proposed to estimate
residual GSI (GSIr) i.e. GSI at residual stage by estimating the reduced
block volume and joint roughness parameters at residual stage by the
method similar to the estimation of peak GSI. Guidelines were provided
to estimate residual block volume (Vbr ) i.e. block volume at residual
stage and residual joint roughness parameters i.e. residual large scale
waviness ( Jwr ), residual small scale roughness ( JSr ) and residual joint
alteration parameters ( JAr ) from their corresponding peak values Vbp ,
JWp ,JSp and JAp . It was observed that a constant value of 10 cm3 can be
taken for Vbr for most of the average quality rock masses as observed
along various tunnel case studies around the world and numerical
studies. Guidelines were provided to estimate the residual joint large-
scale waviness factor (Jwr ), residual small-scale smoothness factor ( JSr )
and residual alteration factor ( JAr ) from the values of peak joint large-
scale waviness factor ( Jwp ), peak small-scale smoothness factor ( JSp ) and
peak alteration factor ( JAp ) based on the experimental results of direct
shear tests of rock joints conducted by Barton et al. (1985) which were
used to estimate residual joint condition factor ( JCr ) as follows.
For the estimation of residual large scale waviness factor ( Jwr ) from
peak large scale waviness factor ( Jwp ) following relations were provided
Jwp
Jwr = 1 if Jwp < 2; and Jwr = if Jwp > 2
2 (5)
Eq. (5) was replaced by single polynomial equation using curve
fitting process, which gives smooth continuous transition over the en-
tire range of JWp as given below
JWr = 0.21(JWp )2−0.61(JWp ) + 1.41 for 1 < JWp < 3 (6)
For the estimation of residual small scale smoothness factor ( JSr )
from peak small scale roughness factor ( JSp ) following relations were
provided
JSr = 0.75 if JSp < 1.5; and JSr = if JSp > 1.5
2 (7)
Eq. (7) was replaced by a single polynomial equation, which gives
smooth continuous transition over the entire range of JSp as given
below:
JSr = 0.16(JSp)2−0.25(JSp) + 0.82 for 0.6 < JSp < 3 (8)
No reduction is recommended for joint alteration factor because
joint alteration is unlikely to occur in a short time period and hence,
G. Tiwari et al. Tunnelling and Underground Space Technology 70 (2017) 375–387

Fig. 2. Reduction of GSI from peak to residual value (a) Photograph of Frank slide showing small block volume after failure (b) Reduction in block volume in slip zone (Cai et al., 2007)
(c) Joint roughness degradation after shearing (Wang et al., 2016).

JAr = JAp (9) (Hoek et al., 2002)

Residual joint condition factor ( JCr ) can be computed using these mbp / r ⎛ ⎜
GSIp / r − 100
⎞
⎟
28
reduced parameters as per the equation given below: = e⎝ ⎠
mi (13)
Jr Jr
JCr = Wr S ⎛
GSIp / r − 100
⎞
JA (10) ⎜ ⎟

sbp / r = e⎝ 9
⎠ (14)
Now GSIr can be obtained by the following equation
1
26.5 + 8.79 ln JCr + 0.9 ln Vbr Em = Ei ⎛0.02 + ⎞
GSIr (JCr ,Vbr ) = ⎝ 1 + e (60 + 15D − GSIp)/11 ⎠ (15)
1 + 0.0151 ln JCr −0.0253 ln Vbr (11)
where mbp / r ,sbp / r
are peak/residual Hoek-Brown constants of rock mass
Since magnitude of Vbr is considered as a constant value of 10 cm3 respectively and mi is the frictional strength component of intact rock
for average quality rock masses Eq. (11) further reduces to Eq. (12) which can be estimated using triaxial testing in laboratory. Statistical
given below: parameters and PDFs of the rock mass strength parameters and de-
28.5 + 8.79 ln JCr formation modulus can be estimated using Monte-Carlo simulation by
GSIr (JCr ) = Eqs. (13)–(15) using statistical parameters and PDFs of mi and GSIp / r .
0.9417 + 0.0151 ln JCr (12)

Statistical parameters and PDFs of JWp ,JSp and JAp were used to esti-
mate the statistical parameters and PDFs of JWr ,JSr and JAr by
Monte-Carlo
simulation using Eqs. (6), (8) and (9). Statistical parameters and PDF of
JCr was estimated using statistical parameters and PDFs of JWr ,JSr and JAr
using Eq. (10). After obtaining statistical parameters and PDF of JCr
statistical parameters and PDF of GSIr was again estimated by Monte-
Carlo simulation using Eq. (12).
2.2. Calculation of strength parameters of rock mass and deformation
modulus
For the current analysis Hoek-Brown criterion was used as the yield
criterion. Parameters required for Hoek-Brown strength criterion for
rock mass are related to well-accepted rock mass classification system
GSI and Hoek-Brown intact rock strength parameter as given below
2.3. Estimation of statistical parameters of yield depth and displacement
around tunnel
After the estimation of statistical parameters and PDFs of rock mass
peak and residual strength parameters and deformation modulus, next
step is to estimate the statistical parameters of response parameters i.e.
yield depth and displacement around the tunnel. Because extensive
numerical analysis is needed to calculate the probability distributions of
output parameters (displacement and yield zone depth), the point es-
timate method coupled with any numerical stress analysis program
seems to be a better option. For the current article Hong’s Point
Estimate Method (PEM) (1998) was coupled with a finite element based
program Phase2 (Rocscience, 2014). Hong’s PEM was selected for the
current study because of the two advantages provided by it. Firstly the
number of evaluation points required is much lesser than the other PEM

378
G. Tiwari et al. Tunnelling and Underground Space Technology 70 (2017) 375–387

method like Rosenblueth (1975). For the current study the number of Table 2
random variables are six (mbp,sbp,mbr ,sbr , UCS and Em) and hence the Estimated statistical parameters and PDFs of peak rock mass properties for different
tunnels.
number of evaluation points required for Hong’s PEM is 12 as compared
to 64 in Rosenblueth’s PEM. Another advantage is the efficiency of Property Parameter Nathpa-Jakhri Kazunogawa Shimizu
Hong’s PEM to handle the asymmetrically distributed random variables
which is important since asymmetric distributions were observed for Peak joint condition Mean 1.51 4.02 0.44
some of the random variables as shown later. factor ( JCp ) COV 14 14.01 13.9
PDF Normal Normal Normal

Peak block volume (Vbp ) Mean 62008 12463 49837

(cm3) COV 38.80 54.21 48.80
3. Applicability of the method to some tunnel case studies
PDF Lognormal Lognormal Lognormal

Peak GSI (GSIp) Mean 54.77 59.88 47

3.1. Nathpa Jhakri tunnel
COV 3.53 3.51 4.78
PDF Normal Normal Normal
The Nathpa–Jhakri hydropower project in the state of Himachal
Peak Hoek-Brown Mean 3.11 4.54 2.57
Pradesh in India involved a major opening for powerhouse cavern of
constant (mbp ) COV 12.07 14.61 11.20
dimensions 216 m × 20 m × 49 m (length × width × height) at a PDF Normal Normal Normal
depth of 262.5 m below the ground level. The opening is located in the
Peak Hoek-Brown Mean 0.0067 0.0119 0.0029
left bank at about 500 m from the river Sutlej with longitudinal axis of constant (sbp ) COV 22.39 23.53 25.07
the openings oriented in the N-S direction. The measured in situ stresses PDF Lognormal Lognormal Lognormal
are 4.73 and 6.34 MPa in E–W and N-S and directions, respectively, and Deformation modulus Mean 10.62 16.5 2.62
5.89 MPa in vertical direction. The major lithology present at the site (Em) (GPa) COV 23.48 16.73 21.45
consists of quartz mica schist. It was observed that generally three joint PDF Lognormal Lognormal Lognormal
sets along with some random joints were present at the site. Joints
present at the site were generally undulated with smooth to rough
Table 3
surfaces having staining on the surfaces of clean joints and at some
Estimated statistical parameters and PDFs of residual rock mass properties for different
locations friction infill with non-softening material coating was ob- tunnels.
served. Shear strength parameters of soils generally consist of two
components i.e. cohesion and friction according to Mohr-Coulomb Property Parameter Nathpa-Jakhri Kazunogawa Shimizu
strength criterion. Frictional infill soil is referred to the soil with neg-
Residual joint condition Mean 0.42 1.00 0.24
ligible cohesion and high frictional strength like sand, silt and calcite factor ( JCr ) COV 9.60 11.04 8.25
while cohesive infill generally refers to the soil with high cohesion like PDF Normal Normal Normal
clay. More details regarding infill classifications can be seen in Cai et al. Residual GSI (GSIr) Mean 22.45 30.25 17.35
(2004). From the description, average values of Jwp , JSp and JAp were COV 3.85 3.19 4.65
taken as 2, 1.5 and 2 respectively with an assumed COV of 8%. Dis- PDF Normal Normal Normal
tributions of joint roughness parameters are normal as observed along Residual Hoek-Brown Mean 0.97 1.57 0.88
some locations for this site and as suggested in literature (Feng and constant (mbr ) COV 10.35 12.93 8.19
Lajtai, 1998; Cai, 2011). Similar distributions for joint roughness PDF Normal Normal Normal
parameters were suggested by Cai (2011) for Kazunogawa tunnel which Residual Hoek-Brown Mean 1.81e−04 4.33e−04 1.00e−04
is one of the case studies described in subsequent sections. Since the constant (sbr ) COV 9.57 10.76 8.42
details regarding spacing for individual joint sets were not available, PDF Lognormal Lognormal Lognormal

the variability in peak block volume (Vbp ) was calculated using varia-
bility in Jv and a constant value of block shape factor of 36 was used as
Statistical parameters and PDFs of peak and residual strength para-
suggested in the literature (Palmstrom, 1995, 1996, 2005). An average
meters were estimated using the method as suggested in the earlier
value of 8.7 was estimated for Jv from field investigation with a COV of
section. Table 2 summarizes the statistical parameters and PDFs of es-
approximately 13.40%, which was calculated from the range of Jv
timated peak rock mass properties. Table 3 summarizes the statistical
(3–10) using 3 sigma rule and a lognormal PDF was assumed for Jv.
parameters and PDFs of estimated residual rock mass properties. Fig. 3
Distributions of joint spacing and joint density are considered log-
shows the PDFs of peak and residual Hoek-Brown constants and de-
normal as observed by various researchers (Kulatilake et al., 2003; Park
formation modulus. Sensitivity analysis was carried out using the
et al., 2005). Similar distribution of joint spacing was suggested by Cai
method suggested by Hamby (1994). In this method the sensitivity
(2011) observed during field investigation for Kazunogawa tunnel
index (SI) is calculated by changing one random variable at a time over
which is one of the case studies described in subsequent sections.
three standard deviations from the mean and determining the percent
Table 1 summarizes the statistical parameters for intact rock properties.
change in the material response parameter of interest (in this case,
plastic zone depth and displacement) over that range. Fig. 4 shows the
Table 1
results of sensitivity analysis for the tunnel to show the relative influ-
Statistical parameters and PDFs of intact rock properties for different tunnels.
ence of various rock mass properties on yield zone depth and dis-
Property Parameter Nathpa-Jakhri Kazunogawa Shimizu placement at the middle location of the sidewall. It was observed from
the analysis that the wall displacement was highly influenced by de-
Hoek-Brown constant Mean 12 19 17
formation modulus, Uniaxial Compressive Strength (UCS) and residual
(mi) COV (%) 8.33 12.5 7.84
PDF Normal Normal Normal Hoek-Brown constants. Regarding yield depth it was observed that
yield zone depth was highly influenced by UCS, peak and residual
Elastic Modulus (Ei) Mean 26 31.87 10.2
(GPa) COV (%) 20.96 13.5 15.69 Hoek-Brown constants. Sensitivity analysis shows the importance of
PDF Lognormal Lognormal Lognormal residual strength parameters on the yield zone depth and displacement
Uniaxial compressive Mean 35 108 49
around tunnel which further shows that the variability in the residual
strength (UCS) COV (%) 14.28 15 19.72 strength parameters should be properly considered in the probabilistic
(MPa) PDF Normal Normal Normal analysis.

379
G. Tiwari et al. Tunnelling and Underground Space Technology 70 (2017) 375–387

Fig. 3. Probability density functions of various peak and residual rock mass properties (a) mbP (b) sbP (c) mbr (d) sbr (e) Em.

Fig. 4. Sensitivity analysis for Nathpa Jhakri tunnel to show the effect of different random variables on various response parameters a) displacement b) yield zone depth.

380
G. Tiwari et al.
Fig. 5. Comparisons of in-situ measured and estimated displacements at different locations of Nathpa-Jhakri tunnel (a) locations of measurement (b) location A (c) location C1 (d)
location C2 (e) location D1 (f) location D2.
The verification of the present method is made by comparing esti-
mated and in-situ measured displacements because displacements are
easy to measure and the field engineer will appreciate it more than the
approach based on strength computation (Singh et al., 2007). Since the
deformation at the tunnel wall and roofs are given for particular stages
(Varadarajan et al., 2001), it was decided to excavate the model in 12
stages (E1-E12 in Fig. 5a) for this tunnel (Varadarajan et al., 2001;
Sitharam and Latha, 2002), so that a comparison can be made between
predicted and measured deformations of the tunnel. Details of different
stages and MPBXs locations along the tunnel boundaries can be seen in
Sitharam and Latha (2002). Fig. 5a shows the locations of displacement
measurement (A, C1, C2, D1 and D2) around the tunnel and Fig. 5b–f
present comparisons between estimated and measured displacements at
these locations. The in-situ measured displacement value for the tunnel
is a single value i.e. deterministic and hence it cannot be represented by
a distribution. The in-situ measured value is also plotted in the figure
381
Tunnelling and Underground Space Technology 70 (2017) 375–387
in-order to compare the estimated and measured displacement value. It
was observed that the predicted ranges of the displacements for con-
ventional approach i.e. elastic-perfectly plastic model were smaller than
the displacement range predicted by considering current approach i.e.
strength drop in the analysis especially for the locations close to the
tunnel wall (A, C1, D1). Elastic-perfectly plastic nomenclature is given
to conventional approach since post-peak behavior in this approach is
assumed as elastic-perfectly plastic. Strength drop nomenclature for the
current approach is given to the current approach since the post-peak
behavior was not assumed to be elastic-perfectly plastic and residual
strength parameters were smaller than the peak strength parameters.
For the locations close to tunnel wall since yielding has occurred the
elasto-plastic stiffness matrix is governed by the residual strength
parameters and since the residual strength parameters for both cases
are different, hence there is a difference in the estimated displacement
for the locations close to the tunnel wall for both cases
G. Tiwari et al.
(Fig. 5b, c, and e). It was observed that the displacement ranges pre-
dicted by both approaches were almost same for locations far from the
tunnel wall (Fig. 5d and f). The reason is that the locations far from the
tunnel remain in elastic zone and hence for them the stiffness matrix is
governed by the elastic parameters which are same for both cases and
hence the difference in the calculated displacement range from both the
models was negligible. It was observed that when the current approach
i.e. strength-drop was considered in the analysis the estimated dis-
placement range was matching well with the in-situ measured dis-
placements at almost all the locations except location A. When the
conventional approach i.e. elastic-perfectly plastic was used it was
observed that the displacement range was underestimated when com-
pared with the in-situ measured displacements (Locations A, C1 and
D1). Results clearly show that conventional probabilistic approach can
lead to significant reductions in predicted displacements, especially for
locations close to excavations and hence assumptions regarding re-
sidual strength should be made carefully for the estimation of dis-
placements in the tunnel for average quality rock mass. Since there are
considerable differences in the displacement range estimated using two
assumptions, there could be significant underestimation in the prob-
ability of failure based on allowable displacement which can under-
estimate the support requirements for tunnel.
3.2. Kazunogawa tunnel
The second tunnel considered for this study is a part of Kazunogawa
power station (Koyama et al., 1997; Cai, 2011), located in Yamanashi
Prefecture, Japan which has a generating capacity of 1600 MW. The
cavern dimensions are width 34 m, height 54 m, and length 210 m with
an overburden of approximately 500 m. The principal stresses in the
plane perpendicular to the cavern axis were determined as 13.4 and
10.4 MPa, respectively. The maximum principal stress direction is in-
clined about 11° with respect to the vertical direction. The major li-
thology present at the site consists of sandstone and composite rock of
sandstone and mudstone. Three orthogonal joints were observed at the
site from field mapping. Average spacing for these joints are 10 cm,
25 cm and 50 cm with a standard deviation of 3 cm, 7.5 cm and 15 cm
respectively. Joints are fresh, have small undulations and rough. From
the conditions mentioned, average values of Jwp , JSp and JAp are taken as
2, 2 and 1 respectively with a COV of 8% and their distributions were
assumed to be normal as suggested for this tunnel by Cai (2011). Sta-
tistical parameters of intact rock properties for this case study were
estimated using laboratory testing and literature and are shown in
Table 1. Variability in peak and residual rock mass properties was es-
timated using the method suggested in the earlier section and the values
of statistical parameters and PDFs for peak and residual rock mass
properties are given in Tables 2 and 3 respectively.
Since the displacement data is available for different stages, tunnel
was excavated in 19 steps (E1-E19 in Fig. 6a) in the current analysis as
done in field (Kudoh et al., 1999; Cai, 2011), to make a comparison
between predicted and measured displacements. A comparison is pro-
vided between measured and predicted deformations of tunnel walls by
both approaches. Fig. 6a shows the locations of measurement of dis-
placements and Fig. 6b–d show the comparison between predicted and
observed displacements around the tunnel at these locations. The in-
situ measured displacement value for the tunnel is a single value i.e.
deterministic and hence it cannot be represented by a distribution. The
in-situ measured value is also plotted in the figure in-order to compare
the estimated and measured displacement value. It was observed that
displacement range predicted by current approach was in good agree-
ment with the field observed displacements while displacements are
underestimated when the rock mass was assumed to be elastic-perfectly
plastic for all the locations considered. The difference in measured and
predicted mean displacements is around 50% for the elastic-perfectly
plastic model, whereas it is less than 10% for the model considering the
post peak drop in strength. The reason is that near the tunnel wall,
382
Tunnelling and Underground Space Technology 70 (2017) 375–387
elements are yielded and hence, the elasto-plastic stiffness matrix is
governed by residual strength parameters which are completely dif-
ferent for the two different cases i.e. elastic-perfectly plastic and
strength-drop considered in the analysis. Hence it is very important to
consider the post peak strength variations properly in numerical ana-
lysis to closely match the yield behavior so as to predict the realistic
deformations in field.
3.3. Shimizu tunnel
Next case study considered in the current analysis is Shimizu Tunnel
No. 3 located in the Shimizu city of Japan and is a section of an im-
portant transport infrastructure project Tomei II Expressway. The
tunnel is of 12 m height and 18 m wide and runs through a length of
1.12 km. Depth of overburden was 83 m and the in-situ stresses re-
ported by Vardakos (2003) are 2 MPa and 1.73 MPa in vertical and
horizontal directions, respectively giving a ratio of horizontal to ver-
tical stress ratio (K) of 0.83. The major lithology present at the site is
weathered soft sandstone. Three major discontinuities were en-
countered during geological investigations. The first set J1 is formed by
bedding plane joints and the second set J2 consists of cross joints at a
random orientation. The third joint set J3 dips almost vertically.
Average joint spacing for J1, J2 and J3 are 55 cm, 55 cm and 165 cm
respectively with a standard deviation of 15 cm, 15 cm and 45 cm re-
spectively evaluated using 3 sigma rule. Joint conditions vary from
planar to undulating and from slickensided to rough with silty sandy
clayey coating. From the conditions mentioned, average values of Jwp , JSp
and JAp are taken as 1.25, 1.05 and 3 respectively with a COV of 8%.
Statistical parameters of intact rock properties for this case study were
estimated using laboratory testing and literature and are shown in
Table 1. Variability in peak and residual rock mass properties was es-
timated using the method suggested in the earlier section and the values
of statistical parameters and PDFs for peak and residual rock mass
properties are given in Tables 2 and 3 respectively.
For the current analysis, the tunnel is excavated in three stages in-
volving: (1) a pilot TBM stage E1, (2) a top heading E2 and (3) a lower
bench stage E3, as done in field (Fig. 7a) (Vardakos, 2003; Latha and
Garaga, 2012). However, measurement data are available mainly for
the top heading excavation, which is used for the comparison of pre-
dicted and measured displacement for the current analysis. Fig. 7 shows
the stages of excavation of the tunnel, locations of displacement mea-
surement (C, D and H) and comparison between predicted and observed
displacement for these locations. It was observed that the range of
displacement predicted by current and conventional approaches was
almost same for all the tunnel locations and the observed displacement
are within the displacement range predicted in the current study.
Considering post peak strength drop did not make any difference in
displacement predictions. The reason for this was that the yield zone
development around tunnel was negligible for the location where dis-
placements were measured and hence the stiffness matrix is governed
by elastic parameters for both the cases. Fig. 8 shows the yield zone
development around the tunnel, obtained from numerical analysis for a
typical evaluation point using strength drop model. Yield zone, which is
represented by the elements yielded in tension or shear is essentially
located in E3 portion of the tunnel. Since the displacement measure-
ments are available only for E2 zone, where no elements were yielded,
the deformation behavior can be considered as elastic in this zone. This
is the reason for both the models predicting similar displacements at
these locations.
4. Parametric study
After validation of the present method for three tunnel case studies,
the method was used to carry out a parametric study for analyzing the
effect of Coefficient of Variation (COV) of different intact rock prop-
erties on the mean values and COVs of yield zone and wall
G. Tiwari et al.
Fig. 6. Comparisons of in-situ measured and estimated displacements at different locations of Kazunogawa tunnel (a) Locations of measurement (b) location A (c) location B (d) location
D.
displacements and probability of failure of tunnel. Parametric study
was carried out by both approaches to investigate the effect on the
estimated statistical parameters of yield zone and wall displacements
and finally on the probability of failure. For this parametric study,
Kazunogawa tunnel was selected from the above three case studies. The
input parameters selected for the parametric study includes elastic
modulus of intact rock (Ei), intact rock Hoek-Brown constant (mi) and
uniaxial compressive strength of intact rock (UCS). The ranges of the
COVs of these parameters considered in the analysis are used as sug-
gested for most of the rocks in the literature. Probability of failure for
tunnel in the current analysis is defined by probability of exceeding the
allowable displacement. Allowable strains are calculated using critical
strain and from the allowable strains, allowable wall displacement was
approximately estimated for the tunnel by calculating equivalent dia-
meter of the tunnel. More details regarding critical strain calculation
and allowable strain calculation for rock tunnels can be seen in litera-
ture (Sakurai, 1981; Cai, 2011). Allowable strain and allowable wall
displacement for Kazunogawa tunnel are 0.24% and 53.1 mm (Cai,
2011) respectively. Results for displacements, yield zone and prob-
ability of failure are shown for only one location (D in Fig. 6a).
4.1. Influence of coefficient of variation of elastic modulus of intact rock
Importance of the deformation modulus in representing the me-
chanical behavior of a rock mass has been acknowledged by many re-
searchers in the past (Palmstrom and Singh, 2001; Hoek and
383
Tunnelling and Underground Space Technology 70 (2017) 375–387
Diederichs, 2006). One of the important factors in determining the
deformation modulus of rock mass (Em) is the elastic modulus of intact
rock (Ei) (Hoek and Diederichs, 2006) and since the analyses carried out
in earlier sections include only one COV value of Ei, it was decided to
carry out a parametric study to investigate the influence of COV of Ei on
the mean and COVs of wall displacement and yield zone. Coefficient of
variation of Ei suggested in the literature is 15–30% for most of the
rocks and hence this range is selected in the current study with an in-
terval of 5% (Ching et al., 2011). Fig. 9 shows typical results for mean
values and COV of yield zone and wall displacement and probability of
failure for both approaches. It was observed that mean values of wall
displacement and yield zone was almost constant for all COVs of Ei and
these mean values are underestimated when strength drop was ignored.
Mean value of yield zone and wall displacement was underestimated by
65% and 31% respectively. Since, the increase in the probability of
failure also indicates the increase in the COV of wall displacement, only
variation of probability of failure with COV of Ei has been shown. It was
observed that probability of failure is increasing with the increasing
COV of Ei for both approaches; however the probability of failure was
underestimated by 75–100% depending on the COV of Ei, when the
conventional approach was used. Compared to COV of wall displace-
ment, COV of yield zone was observed to be constant, which shows that
yield zone development is almost independent of Ei.
G. Tiwari et al. Tunnelling and Underground Space Technology 70 (2017) 375–387

Fig. 7. Comparisons of in-situ measured and estimated displacements at different locations of Shimizu tunnel (a) locations of measurement (b) location C (c) location D (d) location H.

4.2. Influence of coefficient of variation of UCS of intact rock
Uniaxial compressive strength of intact rock is an important para-
meter in Hoek-Brown strength criterion which can affect yield zone
development and displacement of tunnels as observed in sensitivity
analysis. It has been observed that the COV of intact rock UCS generally
ranges from 15% to 30% for most of the rocks (Ulusay et al., 1994; Li
et al., 2012; Langford, 2013) and hence, this range of COV was selected
with an interval of 5%. Fig. 10 shows the typical results for the effect of
COV of UCS on mean values and COVs of yield zone and wall dis-
placement and probability of failure by both approaches. It was
observed that mean values of wall displacement and yield zone were
almost constant with the COVs of UCS and these mean values of dis-
placement and yield zone depth are underestimated when conventional
approach was used. COV of yield zone depth was increasing with COV
of UCS in contrast to the variations with COV of Ei where COV of yield
depth was almost constant with the COV of Ei. Probability of failure was
increasing from 9% to 19% when current approach was used while
probability of failure was close to 0% for conventional approach when
the COV of UCS was increasing. It should be noted that COV of tunnel
wall displacement was increasing with COV of UCS for conventional
approach also; however, the probability of failure was 0% since mean

D H

Fig. 8. Total displacement contours with yielded elements after 2nd stage excavation of Shimizu tunnel and points for displacement measurement selected in the field (C, D, H).

384
G. Tiwari et al. Tunnelling and Underground Space Technology 70 (2017) 375–387

Fig. 9. Effect of COV of Ei on different response parameters of Kazunogawa tunnel (a) mean displacement (b) mean yield depth (c) COV of yield depth (d) probability of failure.

displacement value is small.
4.3. Influence of coefficient of variation of intact rock Hoek-Brown constant
(mi)
Hoek-Brown constant for intact rock (mi) is an important factor for
the estimation of peak and residual Hoek-Brown frictional constants of
rock mass (mbp,mbr ), which in turn can influence yield zone development
and wall displacement for tunnels. COV of mi for most of the rock
generally varies from 15% to 30% (Li et al., 2012; Langford, 2013) and
hence this range of COV was selected with an interval of 5%. It was
observed from Fig. 11a and b that mean wall displacement and mean
yield zone depth were almost constant with the COV of mi and mean
values of displacement and yield zone depth were underestimated by
31% and 65% respectively for conventional approach. It was also ob-
served from Fig. 11c that the COV of yield zone was increasing with the

Fig. 10. Effect of COV of UCS on different response parameters of Kazunogawa tunnel (a) mean displacement (b) mean yield depth (c) COV of yield depth (d) probability of failure.

385
G. Tiwari et al.
Fig. 11. Effect of COV of mi on different response parameters of Kazunogawa tunnel (a) mean displacement (b) mean yield depth (c) COV of yield depth (d) probability of failure.
COV of mi. Fig. 11d shows that the probability of failure was almost
constant with the COV of mi for both approaches. However, it was
observed that probability of failure was approximately 11% when
current approach was considered and probability of failure was 0%
when conventional approach was used. This shows that probability of
failure was highly underestimated by ignoring strength drop in the
probabilistic analysis.
4.4. Discussion on results from parametric study
Results from the parametric study shows that the two intact rock
properties i.e. Ei and UCS have significant influence on displacement
and yield zone depth around the tunnel. While the mean values of both
response parameters of tunnel i.e. yield zone depth and displacement
remains almost constant with the increasing COVs of intact rock
properties, the COVs of these response parameters increases with the
increasing COVs of intact rock properties. Since in most of the prob-
abilistic analyses of tunnels, the performance functions are defined in
terms of tunnel wall displacements and yield zone depth around tunnel,
increasing COVs of response parameters with the increasing COVs of
intact rock properties results in increasing probability of failure. This
effect of increasing variability of rock mass properties on the tunnel
stability in terms of probability of failure is impossible to define in
deterministic analysis, which deals with mean values of all parameters.
Instead considering the mean values of input and response parameters
to evaluate the stability and support requirement of tunnel, the stability
of tunnel should be evaluated in terms of probability of failure, since it
can take into account the variability in rock mass properties. Moreover,
it was observed that COVs of intact rock properties affect COVs of re-
sponse parameters irrespective of the assumption made regarding re-
sidual strength parameters. However it was also observed that there
was significant effect of residual strength parameters on the mean yield
depth and displacement, which results in significant underestimation of
probability of failure, when the rock mass is assumed to be elastic-
perfectly plastic. This shows that even if the COVs of response para-
meters increase with the increasing COVs on intact rock properties,
underestimation of mean values can also underestimate the probability
386
Tunnelling and Underground Space Technology 70 (2017) 375–387
of failure. The parametric study shows the importance of mean values
and COVs of response parameters on the estimated probability of
failure, which is defined in terms of allowable displacement in the
current article. The study conducted in the present article assumes the
drop modulus to be deterministic and infinite. Considering the un-
certainty in drop modulus and gradual softening of GSI with plastic
strain beyond peak will further improve the accuracy of the proposed
computational method.
5. Conclusions
This study shows the advantages of probabilistic method to evaluate
the tunnel stability in average quality rock mass. The study shows the
importance of considering uncertainty in residual strength parameters
on the yield zone depth and displacements around tunnel using a sen-
sitivity analysis. Since, it was observed that residual strength para-
meters have substantial effect on tunnel displacement and yield zone
depth, it is important to consider the uncertainty in residual strength
parameters along with peak strength parameters and deformation
modulus in probabilistic analysis. In this study a GSI based computa-
tional approach was used to compute the variability in peak and re-
sidual strength and deformation modulus properties for average quality
rock mass. Three case studies of tunnels are considered and the de-
veloped computational method is used to estimate the deformations
around these tunnels. It was observed that the displacements for all the
tunnels were matching reasonably well with the field observed dis-
placement values when the uncertainty in residual strength parameters
was considered using current approach. Displacements were under-
estimated especially in the plastic zone developed around the tunnels
when conventional approach was used as compared to current ap-
proach. It was also observed that for the elastic regions where no
yielding was observed, the estimated displacements are same for both
the approaches, as stiffness matrix is governed by the same elastic
parameters for the approaches.
A parametric study was conducted to estimate the effect of COVs of
various intact rock properties on the statistical parameters of yield zone
depth and displacements around the tunnel. This analysis was also
G. Tiwari et al.
carried out for both conventional approach and current approach. It
was observed that mean values of yield zone depth and displacement
were almost constant with the COVs of these parameters. However,
mean values of these parameters are highly underestimated in the
conventional approach. While the COV of yield zone was affected by
COVs of UCS and mi, effect of COV of Ei was negligible. Probability of
failure based on the allowable displacement was highly underestimated
for the conventional approach and it was increasing with the COVs of Ei
and UCS. Hence variability in residual strength should be properly
considered in the analysis of tunnels to obtain realistic estimates of
deformations and probability of failure, especially in average quality
rock mass.
References
Abdellah, W., Hani, S.M., Thibodeau, D., Lindsay, M.V., 2014. Stability of mine devel-
opment intersections – a probabilistic analysis approach. Can. Geotech. J. 51,
184–195.
Alejano, L.R., Dono, A.R., Alonso, E., Manin, G.F., 2009. Ground reaction curves for
tunnels excavated in different quality rock masses showing several types of post-
failure behavior. Tunn. Undergr. Space Technol. 24, 689–705.
Alejano, L.R., Dono, A.R., Veiga, M., 2012. Plastic radii and longitudinal deformation
profiles of tunnels excavated in strain-softening rock masses. Tunn. Undergr. Space
Technol. 30, 169–182.
Arzua, J., Alejano, L.R., 2013. Dilation in granite during servo-controlled triaxial strength
tests. Int. J. Rock Mech. Min. Sci. 61, 43–56.
Barton, N.R., Bandis, S.C., Bakhtar, K., 1985. Strength, deformation and conductivity
coupling of joints. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 22 (3), 121–140.
Cai, M., 2011. Rock mass characterization and rock property variability considerations for
tunnel and cavern design. Rock Mech. Rock Eng. 44, 379–399.
Cai, M., Kaiser, P.K., Tasaka, Y., Minami, M., 2007. Determination of residual strength
parameters of jointed rock masses using the GSI system. Int. J. Rock Mech. Min. Sci.
44 (2), 247–265.
Cai, M., Kaiser, P.K., Uno, H., Tasaka, Y., Minami, M., 2004. Estimation of rock mass
deformation modulus and strength of jointed hard rock masses using the GSI system.
Int. J. Rock Mech. Min. Sci. 41, 3–19.
Ching, J., Hu, G.Y., Yang, Z.Y., Shiau, J.Q., Chen, J.C., Li, S.Y., 2011. Reliability-based
design for allowable bearing capacity of footings on rock masses by considering angle
of distortion. Int. J. Rock Mech. Min. Sci. 48, 728–740.
Crowder, J.J., Bawden, W.F., 2004. Review of post-peak parameters and behaviour of
rock masses: current trends and research. RocNews 2004, Fall: 13.
Duzgun, H.S.B., Yucemen, M.S., Karpuz, C., 2002. A probabilistic model for the assess-
ment of uncertainties in the shear strength of rock discontinuities. Int. J. Rock Mech.
Min. Sci. 39, 743–754.
El-Ramly, H., Morgenstern, N.R., Cruden, D.M., 2002. Probabilistic slope stability analysis
for practice. Can. Geotech. J. 39 (3), 665–683.
Feng, P., Lajtai, E.Z., 1998. Probabilistic treatment of the sliding wedge with EZSlide.
Eng. Geol. 50, 153–163.
Hamby, D.M., 1994. A review of techniques for parameter sensitivity analysis of en-
vironmental models. Environ. Monitor. Assess. 32 (2), 135–154.
Hoek, E., 1998. Reliability of Hoek-Brown estimates of rock mass properties and their
impact on design. Int. J. Rock Mech. Min. Sci. 35 (1), 63–68.
Hoek, E., Carranza-Torres, C., Corkum, B., 2002. Hoek-Brown failure criterion—2002
edition. In: Proceedings of the 5th North American Rock Mechanics Symposium,
Toronto, Canada, pp. 267–273.
Hoek, E., Diederichs, M.S., 2006. Empirical estimation of rock mass modulus. Int. J. Rock
Mech. Min. Sci. 43, 203–215.
Hong, H.P., 1998. An efficient point estimate method for probabilistic analysis. Reliab.
Eng. Syst. Saf. 59, 261–267.
Hudson, J.A., Brown, E.T., Fairhurst, C., 1971. Optimizing the control of rock failure in
servo-controlled laboratory tests. Rock Mech. 3, 217–224.
Idris, M.A., Saiang, D., Nordlund, E., 2015. Stochastic assessment of pillar stability at
Laisvall mine using Artificial Neural Network. Tunn. Undergr. Space Technol. 49,
307–319.
Koyama, T., Nanbu, S., Komatsuzaki, Y., 1997. Large-scale cavern at a depth of 500 m.
Tunn. Undergr. 28 (1), 37–45 (in Japanese).
Kudoh, K., Koyama, T., Nambu, S., Suzuki, Y., Ishibashi, K., 1999. Support design of a
large underground cavern considering strain – softening of rock. In: Proc. of 9th ISRM
Congress, Paris, France.
387
Tunnelling and Underground Space Technology 70 (2017) 375–387
Kulatilake, P.H.S.W., Um, J., Wang, M., Escandon, R.F., Varvaiz, J., 2003. Stochastic
fracture geometry modeling in 3-D including validations for a part of Arrowhead East
Tunnel, California, USA. Eng. Geol. 70, 131–155.
Langford, C.J., 2013. Application of Reliability Methods to the Design of Underground
Structures (PhD thesis). Queen’s University, Ontario, Canada.
Langford, J.C., Diederichs, M.S., 2013. Reliability based approach to tunnel lining design
using a modified point estimate method. Int. J. Rock Mech. Min. Sci. 60, 263–276.
Latha, G.M., Garaga, A., 2012. Elasto-plastic analysis of jointed rocks using discrete
continuum and equivalent continuum approaches. Int. J. Rock Mech. Min. Sci. 53,
56–63.
Li, A.J., Cassidy, M.J., Wang, Y., Merifield, R.S., Lyamin, A.V., 2012. Parametric Monte-
Carlo studies of rock slopes based on the Hoek-Brown failure criterion. Comput.
Geotech. 45, 11–18.
Palmstrom, A., 1995. RMi – A Rock Mass Characterization System for Rock Engineering
Purposes (PhD thesis). University of Oslo, Denmark.
Palmstrom, A., 1996. The weighted joint density method leads to improved character-
ization of jointing. In: Int. Conf. on Recent Advances in Tunnelling Technology, New
Delhi, India.
Palmstrom, A., 2005. Measurements of and correlations between block size and rock
quality designation (RQD). Tunn. Undergr. Space Technol. 20, 362–377.
Palmstrom, A., Singh, R.B., 2001. The deformation modulus of rock masses-comparisons
between in situ tests and indirect estimates. Tunn. Undergr. Space Technol. 16 (3),
115–131.
Park, H.J., West, T.R., Woo, I., 2005. Probabilistic analysis of rock slope stability and
random properties of discontinuity parameters, Interstate Highway 40. Eng. Geol. 79,
230–250.
Ribacchi, R., 2000. Mechanical tests on pervasively jointed rock material: insight into
rock mass behaviour. Rock Mech. Rock Eng. 33 (4), 243–266.
Rosenblueth, E., 1975. Point estimates for probability moments. In: Conf Natl Acad Sci,
USA, pp. 3812–3814.
Rocscience, 2014. Phase2 Version 8.020, Finite Element Analysis for Excavations and
Slopes. Rocscience Inc., Ontario.
Rummel, F., Fairhurst, C., 1970. Determination of the post-failure behavior of brittle rock
using a servo-controlled testing machine. Rock Mech. 2, 189–204.
Sakurai, S., 1981. Direct strain evaluation technique in construction of underground
opening. In: Rock Mech, M.I.T. (Ed.), 22th US Symp. Rock Mech. MIT, Cambridge, pp.
278–282.
Santarelli, F.J., Brown, E.T., 1989. Failure of three sedimentary rocks in triaxial and
hollow cylinder compression tests. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 26
(5), 401–413.
Sitharam, T.G., Latha, G.M., 2002. Simulation of excavations in jointed rock masses using
a practical equivalent continuum approach. Int. J. Rock Mech. Min. Sci. 39, 517–525.
Singh, M., Singh, B., Choudhri, J., 2007. Critical strain and squeezing of rock mass in
tunnels. Tunn. Undergr. Space Technol. 22, 343–350.
Song, L., Li, H.Z., Chan, C.L., Low, B.K., 2016. Reliability analysis of underground ex-
cavation in elastic-strain-softening rock mass. Tunn. Undergr. Space Technol. 60,
66–79.
Sonmez, H., Ulusay, R., 1999. Modifications to the geological strength index (GSI) and
their applicability to the stability of slopes. Int. J. Rock Mech. Min. Sci. 36 (6),
743–760.
Tiwari, R.P., Rao, K.S., 2006a. Post failure behaviour of a rock mass under the influence of
triaxial and true triaxial confinement. Eng. Geol. 84, 112–129.
Tiwari, R.P., Rao, K.S., 2006b. Deformability characteristics of a rock mass undertrue-
triaxial stress compression. Geotech. Geol. Eng. 24, 1039–1063.
Ulusay, R., Türeli, K., Ider, M.H., 1994. Prediction of engineering properties of a selected
litharenite sandstone from its petrographic characteristics using correlation and
multivariate statistical techniques. Eng. Geol. 37, 135–157.
Varadarajan, A., Sharma, K.G., Desai, C.S., Hashemi, M., 2001. Analysis of a powerhouse
cavern in the Himalayas. Int. J. Geomech. 1 (1), 109–127.
Vardakos, S., 2003. Distinct Element Modelling of the Shimizu Tunnel No. 3 in Japan (MS
Thesis). Virginia Polytechnic Institute and State University, Blacksburg.
Walton, G., Arzua, J., Alejano, L.R., Diederichs, M.S., 2015. A laboratory-testing based
study on the strength, deformability, and dilatancy of carbonate rocks at low con-
finement. Rock Mech. Rock Eng. 48, 941–958.
Wang, G., Zhang, X., Jiang, Y., Wu, X., Wang, S., 2016. Rate-dependent mechanical be-
havior of rough rock joints. Int. J. Rock Mech. Min. Sci. 83, 231–240.
Wawersik, W.R., Brace, W.F., 1971. Post-failure behavior of a granite and diabase. Rock
Mech. 3, 61–85.
Wawersik, W.R., Fairhurst, C., 1970. A study of brittle rock fracture in laboratory com-
pression experiments. Int. J. Rock Mech. Min. Sci. 7, 561–575.
Yang, S.Q., Jing, H.W., Wang, S.Y., 2012. Experimental investigation on the strength,
deformability, failure behavior and acoustic emission locations of red sandstone
under triaxial compression. Rock Mech. Rock Eng. 45 (4), 583–606.