Computers and Geotechnics 51 (2013) 101–115

Contents lists available at SciVerse ScienceDirect

Computers and Geotechnics

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

Numerical direct shear tests to model the shear behaviour of rock joints
M. Bahaaddini a,b,⇑, G. Sharrock a, B.K. Hebblewhite a
a
School of Mining Engineering, The University of New South Wales, Sydney, Australia
b
Shahid Bahonar University of Kerman, Kerman, Iran

a r t i c l e i n f o a b s t r a c t

Article history: In this paper, the shear behaviour of rock joints are numerically simulated using the discrete element
Received 19 December 2012 code PFC2D. In PFC, the intact rock is represented by an assembly of separate particles bonded together
Received in revised form 18 February 2013 where the damage process is represented by the breakage of these bonds. Traditionally, joints have been
Accepted 19 February 2013
modelled in PFC by removing the bonds between particles. This approach however is not able to
Available online 26 March 2013
reproduce the sliding behaviour of joints and also results in an unrealistic increase of shear strength
and dilation angle due to the inherent micro-scale roughness of the joint surface. Modelling of joints
Keywords:
in PFC was improved by the emergence of the smooth joint model. In this model, slip surfaces are applied
Direct shear test
Rock joint
to contacts between particles lying on the opposite sides of a joint plane. Results from the current study
Particle flow code show that this method suffers from particle interlocking which takes place at shear displacements greater
Bonded particle model than the minimum diameter of the particles. To overcome this problem, a new shear box genesis
Smooth joint model approach is proposed. The ability of the new method in reproducing the shear behaviour of rock joints
is investigated by undertaking direct shear tests on saw-tooth triangular joints with base angles of
15°, 25° and 35° and the standard joint roughness coefficient profiles. A good agreement is found between
the results of the numerical models and the Patton, Ladanyi and Archambault and Barton and Choubey
models. The proposed model also has the ability to track the damage evolution during the shearing
process in the form of tensile and shear fracturing of rock asperities.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction /l. Above the transition stress, the shear strength is governed
by shearing of asperities with cohesion cj and residual friction
The mechanical behaviour of jointed rock masses is strongly angle /r.
dependent on the mechanical and geometric properties of discon-
tinuities. At surface and shallow depth, the rock mass behaviour is s ¼ rn tanð/l þ iÞ for rn < rT ð1Þ
predominately controlled by shear displacement along discontinu- s ¼ cj þ rn tan /r for rn P rT ð2Þ
ities rather than failure of intact rock [1]. Surface roughness is
known to have a significant effect on the shear behaviour of rock where rT ¼ ðtan /lctan
j
/r Þ
.
discontinuities. Much has been done to develop empirical and ana- Although this analytical model captures two key mechanisms of
lytical shear strength criteria to account the effect of roughness. asperity sliding and shearing, it has several drawbacks such as the
Patton [2] and Ladanyi and Archambault [3] were among the first lack of quantitative description of joint cohesion cj which is depen-
to develop a shear strength criterion for rough rock joints. Patton dent on intact rock properties and asperity geometry. Furthermore,
[2] by undertaking experiments on saw-tooth triangular asperities for real rock joints shearing and sliding take place simultaneously
proposed a bilinear shear strength criterion. In this model, the [4,5]. Based on the energy principles, Ladanyi and Archambault
failure envelope consists of two linear segments intersecting at a [3,6] developed a failure model which recognises the simultaneous
normal stress rT, called the transition stress. For normal stresses sliding and shearing mechanisms.
rn less than rT, the shear strength s is governed by sliding along
the joint asperities with an inclination angle of i and friction angle
rn ð1  as Þðv_ þ tan /l Þ þ as SR
s¼ ð3Þ
1  ð1  as Þv_ tan /l

⇑ Corresponding author at: School of Mining Engineering, The University of New where as is the shear area ratio, v_ is dilation rate and SR is the intact
South Wales, Sydney, Australia. Tel.: +61 2 93856118; fax: +61 2 9313 7269. rock strength. Ladanyi and Archambault [3,6] suggested that the in-
E-mail addresses: [email protected] (M. Bahaaddini), g.sharrock@ tact rock strength SR can be obtained from the Fairhurst intact
unsw.edu.au (G. Sharrock), [email protected] (B.K. Hebblewhite). strength criterion [7]:

0266-352X/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compgeo.2013.02.003
102 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115

Patton’s model for sliding Fairhurst’s model results of tilt, push or pull tests. They also proposed that the peak
for intact material dilation angle can be estimated as:
 
Shear Strength (τ)
1 JCS
dn ¼ JRC log ð8Þ
M rn
Ladanyi and Archambault’s where M is a damage coefficient, takes values of 1 or 2 for shearing
model
under low or high normal stress, respectively [11]. This parameter
Residual strength of planar joints
can also be approximated from the following relation [10]:

JRC
M¼ þ 0:70 ð9Þ
12 logðJCS=rn Þ
Several empirical and analytical methods have been proposed
in the last few decades to estimate the shear properties of rock
joints in direct shear tests. However, a robust method that can con-
sider the effect of joint roughness on the shear behaviour of rock
Normal Stress ( ) joints is not still available. One part of this problem relates to the
lack of suitable method for quantitative representation of the joint
Fig. 1. Ladanyi and Archambault model (modified from [8]). roughness. Another part is related to unknown mechanism of
asperities degradation during the shearing process.
The rapid progress of computer technology has provided the
Table 1 opportunity to numerically simulate the shear behaviour of rock
Micro-properties of balls and bonds. joints. Several researchers have tried to simulate the shear behav-
Particle micro-mechanical Parallel bonds micro-mechanical iour of rock joints using the constitutive models [12–14]. However,
properties properties these approaches are unable to trace the asperity degradation and
Ball density (kg/m3) 2205 Ec (GPa) 2.8 development of cracks inside the intact material. Furthermore,
Ec (GPa) 2.8 Normal strength (MPa) 20 ± 4.5 special difficulties exist in modelling the geometries of real rock
Coefficient of friction 0.6 Shear strength (MPa) 20 ± 4.5 joints. Karami and Stead [15] by using a hybrid FEM/DEM code, EL-
kn/ks 1.45 k s
 n =k 1.45
FEN software, tried to investigate the process of joint surface deg-
radation in the direct shear test. They carried out direct shear test
on three standard JRC profiles of 0–2, 10–12 and 18–20 and studied
pffiffiffiffiffiffiffiffiffiffiffiffi  
nþ11 rn 0:5 the joint surface damage degradation in terms of asperity wear and
SR ¼ r c 1þn ð4Þ propagation of tensile cracks in intact material along the joint
n rc
plane. However, the ability of their method in reproducing the
where rc is the uniaxial compressive strength of intact rock and n is shear strength and dilation angle was not studied quantitatively.
the ratio of tensile to uniaxial compressive strength of the intact Particle flow code (PFC) is a promising numerical approach to
rock. This model is shown in Fig. 1. study the mechanical behaviour of discontinuities explicitly. PFC
Ladanyi and Archambault [6] based on extensive experimental is a discrete element software in which the intact material is
study on saw-tooth triangular joints, proposed the following represented by a composite of spherical (in 3D) or circular (in
empirical power laws for estimation of as and v_ : 2D) particles bonded together, called bonded particle model
(BPM). The damage process in the BPM is represented by breakage
  of bonds and propagation and coalescence of these micro-cracks
rn k1
as ¼ 1  1  ð5Þ control the macroscopic behaviour of the BPM. Therefore, it pro-
rT
  vides a scientific tool to track the asperities degradation during
rn k2 the shearing process and investigate the effect of joint roughness
v_ ¼ 1 tan i ð6Þ
rT on the shear behaviour of rock joints. The BPM also has the ability
to reproduce many features of intact rock behaviour [16–21].
The parameter rT is the transition stress at which the strength
The main aim of this paper is to investigate the shear behaviour
of joint is equal to the strength of intact rock and k1 and k2 are
of rock joints in a direct shear test using PFC2D [22]. The most
empirical constants with values of 1.5 and 4.0, respectively.
commonly used method for modelling joints in PFC is the bond re-
Although this method is based on sound theoretical concepts,
moval approach. In this method particles lying on the joint surface
there are several shortcomings, such as the difficulty to determine
are left unbonded. Shortcomings of this method are presented in
rT and i for irregular joint geometries and incorrect kinematic con-
Section 3. A new approach to model the joints in PFC is the smooth
sideration of the shearing of irregular joints [4,9].
joint (SJ) model, in which slip surfaces are applied at contacts
Barton and Choubey [10], based on extensive experimental
between particles that lie on the opposite sides of the joint plane.
studies on real rock joints, proposed the following empirical shear
A detailed analysis of the SJ model in reproducing the shear
strength model:
  
JCS
s ¼ rn tan /r þ JRC log ð7Þ Table 2
rn Model output for calibrated intact rock (macro-properties).

UCS (MPa) E (GPa) m

where /r is residual friction angle and JRC and JCS are joint rough-
Experimental Average 27.4 4.2 0.2
ness coefficient and joint compressive strength, respectively. For
unweathered rock joints, the residual friction angle /r is equal to Numerical Average 27.45 4.25 0.198
SD 0.59 0.01 0.007
the basic friction angle /b and JCS is equal to the uniaxial compres-
CV (%) 2.15 0.23 0.38
sive of intact rock rc. JRC can be determined either from visual com-
parison with 10 standard JRC profiles or by back calculation of the SD: Standard Deviation, CV (%) = (Standard Deviation/Average)  100.
 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115 103

Wall 1
Loading wall Normal Force
xvel = 0.1m/s Wall 2
xvel = 0.1m/s

40 mm
Box gap
1.2 mm

Wall 3
Fixed wall
100 mm
Fig. 2. Simulation of a direct shear test under constant normal load.

Fig. 3. Generation of a planar joint by bond removal method (joint particles are shown in green colour). (For interpretation of colors in this figure legend, the reader is
referred to the web version of this article.)

behaviour of rock joints is presented in Section 4. Investigations [22,23]. In previous numerical modelling of direct shear tests in
presented in this paper show that this approach suffers from par- PFC, the CB model has been predominately used due to its reduced
ticle interlocking at large shear displacement. To overcome this number of micro-parameters compared to the PB model [24–26].
problem a new method is presented and the ability of proposed However, it is well understood that the PB model produces a more
method in reproducing the shear behaviour of rock joints is inves- realistic representation of rock-like materials [23,27]. Therefore,
tigated by undertaking direct shear tests on saw-tooth triangular the PB model is used in the current study.
joints and standard JRC profiles. The proper selection of the micro-properties for the particles
and bonds is an important step in bonded particle modelling. In or-
der to generate a numerical model to simulate the real rock, the
2. Numerical model set up micro-mechanical properties of the models are calibrated against
physical experiments on Hawkesbury sandstone. The Hawkesbury
2.1. Bonded particle model calibration sandstone dominates the Sydney region in Australia and hosts
many significant civil and mining excavations. The strength of
Two basic bond models are provided in PFC; namely the contact the Hawkesbury sandstone is known to vary considerably depend-
bond (CB) and parallel bond (PB) models. Both CB and PB models ing on deposition type, the degree of cementation and weathering.
approximate the physical behaviour of cement like substance lying An extensive dataset of intact rock properties of Hawkesbury sand-
between particles and joining them together. The PB is active over stone exists at the University of New South Wales [28].
a finite rectangular (2D) or circular (3D) cross-section lying on the Based on the procedure developed by Potyondy and Cundall
contact plane and centred at the contact point and can transmit [29], BPM samples having a width of 42 mm and height of
both force and moment between particles while the CB is active 84 mm are generated. The size of the particles are controlled by
only at the contact point and can only transmit force acting at predefined minimum and maximum diameters (Dmin = 0.28 mm
the contact point. In the PB model, bond breakage results in imme- and Dmax = 0.42 mm) which satisfies a uniform particle-size distri-
diate reduction of contact stiffness while in the CB model, the con- bution. Inverse calibration is used to calibrate the micro-properties
tact stiffness is active as long as the particles stay in contact of the bonds and the particles. This is involved an iterative process
104 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115

are compared with physical experiments in Table 2. The numerical

tests results are in good agreement with the physical experiments
and the low values of coefficients of variation (CV) of the numerical
experiments indicate the minor difference between the results of
numerical tests.

2.2. Numerical modelling of direct shear test under constant normal

load

In direct shear tests, shear box specimens with width 100 mm

and height 40 mm are generated. Samples consist of around
38,000 particles and their micro-properties are presented in
Table 1. A schematic of the direct shear test under constant normal
load (CNL) is presented in Fig. 2. The normal load is applied verti-
cally to the upper block and this load is kept constant during the
test by using a servo-control mechanism [22]. The lower block is
kept stationary during the test and a horizontal velocity of 0.1 m/
s is applied to the upper block. Calculations in PFC are based on
an explicit time stepping algorithm. The time step (Dt) in each
calculation has small value of Dt  4.139  108 s. This means that
the upper block moves horizontally at the rate of 4.139  1010 m/
time step. Sensitivity studies conducted as a part of this study have
shown that this displacement rate is slow enough to ensure that
the sample remains in quasi-static equilibrium. The shear displace-
ment is measured by tracing the horizontal displacement of the
upper block and the shear stress is calculated by dividing the
recorded the reaction force on wall 3 by the joint length. Normal
displacement is measured by the vertical displacement of the
upper wall.

Fig. 4. Effect of particle coefficient of friction on (a) shear stress and (b) dilation of
3. Bond removal method
planar joint.
One of the first attempts to model the shear behaviour of a rock
joint using PFC2D was carried out by Cundall [30]. In this model,
to select the micro-parameters to reproduce the desired macro- for generation of the joint, the particle contacts within a specific
properties measured in the laboratory experiments. It is common distance (i.e. Dmax) on either side of the joint track were left un-
to calibrate micro-parameters against uniaxial compressive bounded. Subsequently, this method has been used by a number
strength UCS, elastic modulus E and Poisson’s ratio v. The first step of researchers for simulation of the direct shear test in PFC [24–
involves the calibration of E which is controlled by the particle con- 26,31–34]. A generated planar joint using this method is shown
tact modulus Ec,particle normal/shear stiffness ratio kn/ks, parallel in Fig. 3. Unbonded particles are shown in green colour. Previous
bond modulus Ec and parallel bond normal/shear stiffness ratio studies have shown that the coefficient of friction of joint particles
n =k
k s . The second step is matching v which is influenced by k /k
n s has considerable effect on the shear behaviour of joints and there-
and k n =k
s and these micro-parameters are calibrated in an iterative
fore need to be set close to zero to reproduce a realistic shear
process with the first step. Finally, the UCS is calibrated by altering strength [24,26]. Cundall [30] suggested that the coefficient of fric-
the normal and shear strengths of parallel bonds [22]. The micro- tion of joint particles should be calibrated against the friction angle
parameters of the calibrated BPM model are listed in Table 1. Ten of the joint plane.
uniaxial compression tests on are carried out on BPM models with In order to investigate the effect of the coefficient of friction of
different randomly generated particle assemblies and mechanical joint particles on the shear strength and dilation, a sensitivity anal-
properties listed in Table 1. The results of numerical experiments ysis is undertaken. Planar joints are generated and the coefficient

Fig. 5. Stress concentration at micro-scale asperities generated the by bond removal method (black line: contact force, red line: tensile crack, blue line: shear crack). (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115 105

4.5 either side of the joint track. Previous studies have shown that the
τ = 0.58σn + 0.93 increase of band width results in a reduction in the peak shear
4 Triangular R2 = 0.996 strength and normal displacement, but does not lead to realistic
3.5 shear behaviour of planar joints [35]. Furthermore, free movement
Shear Stress (MPa)

Planar of several particles across the joint track reduce the efficiency of
3
this method, especially for modelling rough joints.
2.5 In order to investigate the ability of the bond removal method
τ = 0.43σn + 0.89 to reproduce the shear strength of rock joints, direct shear tests
2 τ = 0.96σn + 0.33 R2 = 0.995
R2 = 0.997 on a saw-tooth triangular joint with a base angle of h = 15° and a
1.5 planar joint are carried out. The coefficient of friction of joint par-
ticles is assumed 0.005. The peak shear strengths of these joints
1
under different normal stresses are presented in Fig. 6. Bilinear
0.5
τ = 0.82σn + 0.28
peak strength envelopes are found for both saw-tooth triangular
R2 = 0.992 and planar joints. It is well understood that the shear strength of
0
0 1 2 3 4 5 planar joints obeys the Coulomb criterion (s = rntan /l). Therefore,
Normal Stress (MPa) its bilinear peak shear strength envelope is related to the existing
micro-scale roughness of the joint surface. There are also unrealis-
Fig. 6. Peak shear strength envelopes of planar and saw-tooth triangular joint with tic cohesions of 0.28 MPa and 0.33 MPa at low normal stresses for
the base angle of 15° using the bond removal method. planar and saw-tooth triangular joints, respectively. Furthermore,
there is a significant difference in the slope of the peak shear
of friction of joint particles l is varied from 0.005 to 0.15. Direct strength envelopes at high normal stresses.
shear tests under constant normal stress of 1 MPa are carried out The results clearly reveal that the bond removal method is un-
and the results are presented in Fig. 4. Results show unrealistic able to simulate the shear behaviour of rock joints. Generation of
shear behaviour for planar joints. Reduction of the coefficient of joints using this approach results in the inherent micro-scale
friction results in a decrease in the peak shear strength and the roughness and recognition of the shear strength is caused by the
normal displacement but generation of the joint by the bond re- interlocking particles or the joint asperities becomes complicated,
moval approach leads to a micro-scale roughness at the joint sur- especially for rough joints.
face (Fig. 3). Although the average size of particles is very small
compare to the length of shear box (0.35 mm to 100 mm), due to
4. Smooth joint model
the circular shape of the particles, their non-uniform distribution
and unequal size, the joint surface is not smooth at the initial stage
The smooth joint model is a newly developed method for sim-
of the shear test. Therefore, the local shear resistance occurs for
ulation of rock discontinuities in PFC [36]. Limited efforts have
riding up of the interlocking particles. Due to stress concentration
been carried out to apply this approach for simulation of direct
at these interlocking particles, some bonds break and when the
shear test [37,38]. In this model, rectangular slip surfaces (i.e.
particle jumps over another particle, the stored energy converts
smooth joint) are assigned to all contacts between particles lying
to kinetic energy and energy dissipation takes place. As a result,
on the opposite sides of the joint plane. Particle pairs intersected
the shear stress drops after reaching the peak shear stress, unlike
by a smooth joint may overlap and pass through each other rather
real planar joints where the shear stress is approximately constant
than forced to move around one another [39,40].
at post peak. The reduction of the coefficient of friction only results
A typical example of the smooth joint model is shown in Fig. 7.
in a decrease of the local shear resistance for riding up of interlock-
The joint plane consists of two coincident planar surfaces (indi-
ing particles.
cated as surface 1 and surface 2). The orientation of these surfaces
There is also significant dilation in the initial stages of shearing
is defined by the joint unit normal vector n ^ j , where n
^ j points into
which is due to the riding up of interlocking particles over each
surface 2. The unit normal vector is defined by the joint dip angle
other. Fig. 5 shows the contact force distribution during the shear-
hp:
ing process. Due to the micro-scale roughness of the joint plane,
contact forces are not distributed uniformly over the joint plane
^ j ¼ ðsin hp ; cos hp Þ
n ð10Þ
and there are unrealistic stress concentrations at interlocking par-
ticles. As the coefficient of friction increases, stress concentration The smooth joint is assigned at a contact between particles their
at these micro-scale asperities increases and results in increased centres lying on opposite sides of the joint plane. In order to deter-
breakage of adjacent joint particles bonds. mine in which surface each particle lies, the dot product of n ^ j and
Cundall [30] noted that the unbonded band (green particles in contact unit normal vector n ^ c is used. The n
^ c of each ball is directed
Fig. 3) should have a width of several particle diameters; otherwise from its centre to the centre of connected ball. If n ^j  n
^ c P 0, the
large resisting forces may occur between the opposing particles on ball lies on surface 1. For example in Fig. 7, ball 1 lies on surface

x
θp

Surface 2

Surface 1 Joint Plane

Fig. 7. Application of smooth joint (modified from [22]).

106 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115

| |

| |

Unload/reload

| |
(a) (b)
Fig. 8. Smooth joint force–displacement law (a) Normal force versus normal displacement and (b) Shear force versus shear displacement [22].

2. The smooth joint is applied in the direction parallel to the joint particle model and a constant normal stress of 1 MPa is applied
plane and remains active while there is a nonzero overlap between to the upper block. The results of a representative direct shear test
particles [22]. are shown in Fig. 9. It is found that this model reproduces the shear
By creation of the smooth joint, the bond between particles (if behaviour of planar joints when the shear displacement is less than
exist) is removed and a set of elastic spring is applied uniformly the minimum particle diameter (Dmin = 0.28 mm). However, when
over a rectangular cross section. The area of smooth joint cross sec- the shear displacement exceeds Dmin, there is an unrealistic in-
tion A is: crease in the shear stress and normal displacement. The contact
A ¼ 2Rt ð11Þ force distribution at the initial stage of the test and after 0.3 mm
shear displacement is presented in Fig. 10. At the initial stage of
where t is the thickness (t = 1.0) and R is the radius of the smooth testing, stresses are distributed uniformly along the joint plane
joint cross section. while after 0.3 mm shear displacement, stresses are concentrated
at lock up points. These stress concentrations lead to an increase
R ¼ k minðRA ; RB Þ ð12Þ
in the shear stress and associated initiation and propagation of
where  k is radius multiplier, which is usually taken 1.0, and RA, RB cracks at the lock up points. In order to find the reason of these
are particles radii. The properties of smooth joints are defined by
the SJ normal stiffness k  , shear stiffness k and coefficient of fric-
nj sj
tion lj.
Let force F and displacement U be expressed in the local coordi-
nate system of the joint plane. ^tj is the tangential unit vector of
joint plane where the cross product of n ^ j and ^tj points in the posi-
^ Therefore,
^ j  ^tj ¼ þk).
tive z direction (i.e. n
^ j þ Fs
F ¼ Fnn ð13Þ
^ j þ Us
U ¼ Un n ð14Þ

where Fn and Un are the normal force and displacement and Fs and
Us are the shear force and relative shear displacement vectors,
respectively. Positive values of Fn and Un denotes compression
and overlap, respectively. The strength model of the smooth joint
contact is that of the Coulomb sliding model. In each time step,
the relative displacement increment between two particles inter-
sected by the smooth joint is decomposed to normal and tangential
components to the joint surface. The normal and shear forces are
updated via (Fig. 8):
 ADU e
F n :¼ F n þ k ð15Þ
nj n

F0 :¼ Fs  k ADUe ð16Þ
s sj s

where D denotes an increment. An updated shear force value F0s is

computed at the end of each time step. The greatest possible value
of shear force F s is dependent on the smooth joint coefficient of fric-
 
tion lj F s ¼ F n lj . If jF0s j 6 Fs , then jFs j ¼ jF0s j ; otherwise sliding
takes place and jFs j ¼ F s (Fig. 8b).

4.1. Simulation of direct shear test by the smooth joint model

In order to study the ability of the SJ model to reproduce the

shear behaviour of joints, a direct shear test on a planar joint is
undertaken. A planar joint plane is inserted into the bonded Fig. 9. Results of direct shear test on planar joint using the smooth joint model.
 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115 107

Particle
Surface 2
interlocking
Surface 1

(a) (b)
Fig. 10. Particle interlocking in the smooth joint model (a) initial stage and (b) after 0.3 mm shear displacement.

Fig. 11. Procedure of shear box genesis (a) vessel generation, (b) assembly generation, (c) isotropic stress installation, (d) elimination of floaters, (e) parallel bond installation
and (f) application of smooth joints.

stress concentrations, the process of smooth joint application is 5803) and ball 3068 is checked and both balls lie on the same side
investigated at the micro-scale (Fig. 10). At the initial stage, all of the joint plane and hence a smooth joint is not assigned at this
ball-ball contacts are checked and smooth joints are assigned at contact. This results in the interlocking of ball 5803 and the stress
contacts where the balls lie on opposite sides of the joint plane. concentration around this particle. This particle interlocking leads
Centre of ball 5803 is above the joint plane (lie on surface 2) and to the unrealistic increase of shear stress. This result shows that
the smooth joints are assigned to its contacts with balls 6910 the SJ model cannot reproduce the shear behaviour of joints at
and 7191 that lie on surface 1. By shear movement of the upper shear displacements larger than Dmin.
block to the right side, ball id 5803 passes through ball id 7219
and a new contact is generated with ball 3068. The newly gener-
ated contacts are checked each cycling step. Due to the applied 5. Shear box genesis approach
normal stress to the upper block, ball 5803 moves downward dur-
ing the shearing process and its centre lies below the joint plane As discussed in Section 4.1, the inability of the smooth joint
(on surface 1). The newly generated contact of this ball (ball model to reproduce the shear behaviour of joints is related to the
108 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115

detection of the top and bottom particles of the shear box blocks lower blocks (green lines in Fig. 11f). The orientations of smooth
and the proper application of the smooth joint. To overcome this joints are parallel to the predefined joint profile which is set to
problem, a shear box genesis method is proposed. In this approach the stationary lower block. This approach resolves the problem
the procedures of sample preparation and smooth joint application of particle interlocking and can be easily employed for direct
are modified. Upper and lower blocks of the shear box are gener- shear tests of irregular joint profiles.
ated separately and the smooth joint is only applied at contacts be-
tween the particles of these blocks. The following six steps are 6. Validation study
employed in the shear box genesis approach (Fig. 11).
In order to investigate the ability of the shear box genesis ap-
1. Material vessels generation: Each upper and lower block consists proach in reproducing the shear behaviour of rock joints, a valida-
of four frictionless walls (Fig. 11a). In order to generate a rough tion study is carried out on saw-tooth triangular joints and
joint in the numerical model, the rock joint profile is digitised standard JRC profiles where the results of the numerical models
and its nodes are used to generate the joint surface walls. In are compared against well-known shear strength criteria. In order
PFC, only one side of the standard wall is active and a wall is to generate an objective model, smooth joint parameters are cali-
defined by the order of its nodes [22]. Therefore, two walls brated before undertaking the validation study.
are required to produce a joint profile.
2. Initial assembly generation: Assemblies of randomly placed par-
6.1. Calibration of smooth joint parameters
ticles are generated and the upper and lower vessels filled sep-
arately (Fig. 11b). The particle diameters are controlled by the
To calibrate the smooth joint (SJ) parameters, experimental nor-
predefined Dmin = 0.14 mm and Dmax = 0.21 mm and satisfy uni-
mal deformability and direct shear tests are carried out on sawn
form particle-size distribution. The number of particles for each
planar joints of Hawkesbury sandstone. The SJ normal stiffness is
block is determined by their volumes and the overall porosity.
calibrated against experimental normal deformability test and
The normal stiffness of walls is made 10% higher than the aver-
the SJ shear stiffness and SJ coefficient of friction are calibrated
age normal stiffness of particles to ensure the ball-wall overlap
against direct shear tests.
remains small. Then, the system is allowed to rearrange and
Normal deformability experiments involve loading of single-
reach the static equilibrium under zero friction of particles
jointed rectangular blocks of Hawkesbury sandstone with length
[22,29].
100 mm, width 100 mm and height 40 mm and recording the nor-
3. Applying a specified isotropic stress: In order to reduce the mag-
mal force and normal displacement (based on the procedure devel-
nitude of locked-in stresses, the radii of all particles are reduced
oped by Bandis et al. [41]). A typical result of an experimental test
uniformly to reach a specified isotropic stress (rt0 ¼ 1% of the
on intact rock and rock with planar joint is shown in Fig. 12. As the
uniaxial compressive strength) [29]. Five measurement circles
applied normal stress increases, joint normal stiffness increases.
are
placed
in each block (Fig. 11c). Isotropic stress
The numerical direct shear tests are carried out under the constant
r0 ¼ r11 þ2 r22 inside these measurement circles are calculated normal stress of 0.5–5 MPa. Therefore, the secant normal stiffness
for each step. If the normalised difference between the specified
isotropic stress and the measured t isotropic stress is less than
r r
the isotropic stress tolerance 0rt 0 6 0:5 , iteration stops and 12
0
floating particles are checked, otherwise the iteration continues Intact rock
[22,29]. 10
Rock with planar joint
Normal Stress (MPa)

4. Elimination of floating particles: The assembly of randomly dis-

tributed and non-uniform size particles may have a large num- 8
ber of particles with less than three contacts (red1 particles in
Fig. 11d). Before installing bond between particles, it is desirable 6
5 MPa
to reduce the number of these floating particles and produce a
dense pack of highly interlocked grains [22,29]. Therefore, in this 4
step floaters of upper and lower blocks are eliminated and assem-
Kn
blies are generated in which all particles away from the bound- 2
aries have at least three contacts. 0.5 MPa
5. Parallel bond installation: Parallel bonds are installed at contacts 0
between particles that have a separation less than 106 times of 0 0.1 0.2 0.3 0.4
average radius of two particles (black lines in Fig. 11e). Normal Displacement (mm)
6. Applying smooth joints: By separate generation of the upper and
Fig. 12. Experimental normal deformability test (Kn: normal stiffness of the
lower blocks, particles of these blocks can be recognised easily system).
in each stage of test which solves the problem of detection of
upper and lower particles of shear box blocks. After generation
Table 3
of upper and lower blocks, the interface walls are removed. A Calibrated smooth joint parameters and result of experimental and numerical tests.
specified constant normal stress is applied to the upper block
and allowing the assembly to relax. This is done by stepping Calibrated SJ parameters
Normal stiffness (MPa/mm) 200
the model until the static equilibrium is achieved. During the Shear stiffness (MPa/mm) 50
relaxation process (and also shearing process), new contacts Friction coefficient, lj (tan(/j)) 0.78
are generated between the particles of the upper and lower Results of experiments
blocks. Smooth joint is applied at newly generated contact if
Parameter Physical Numerical
the contact lies between the particles defining the upper and
experiments experiments
System normal stiffness (MPa/mm) 28.771 28.766
System shear stiffness (MPa/mm) 6.42 6.417
1
For interpretation of color in Fig. 11, the reader is referred to the web version of Friction angle (°) 37.62 37.81
this article.
 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115 109

Normal stress = 1 MPa

Normal stress = 2 MPa

Fig. 13. Numerical direct shear tests on planar sawn joints under different normal
stresses.

Normal stress = 3 MPa

Normal stress = 5 MPa

Fig. 15. Effect of normal stress on asperity degradation of saw-tooth triangular joint
with the base angle of 15 degrees after 3 mm shear displacement (red: tensile crack,
blue: shear crack). (For interpretation of the references to colour in this figure
legend, the reader is referred to the web version of this article.)

are the assembly vessel which is used for sample generation. The
normal stiffness of walls is set equal to the average normal stiffness
of contacting particles [22]. The vertical load is applied at a con-
stant displacement rate of 0.01 m/s. The loading continues to half
of the uniaxial compressive strength of the intact material. The
normal stress and normal displacement are recorded during the
test and the secant normal stiffness of system in the range of
0.5–5 MPa is calculated. The inverse calibration method is used
to calibrate the SJ normal stiffness, in which SJ normal stiffness is
varied until the normal stiffness of experimental tests is achieved.
Result of this calibration is presented in Table 3.
To calibrate the SJ shear stiffness and SJ coefficient of friction,
experimental direct shear tests on planar joints of Hawkesbury
sandstone are undertaken. First, numerical direct shear test under
constant normal stress of 1 MPa is carried out and the SJ shear stiff-
Fig. 14. Results of numerical direct shear tests on saw-tooth triangular joint with ness is varied to reproduce the system shear stiffness measured in
the base angle of 15 degrees, (a) shear stress versus shear displacement and (b) laboratory experiments. Then, numerical direct shear tests under
normal displacement versus shear displacement.
different normal stresses are undertaken and SJ coefficient of fric-
tion is calibrated. Calibrated SJ parameters and results of numerical
of the system in that range is calculated (Kn = 28.77 ± 2.42 MPa/ and experimental tests are presented in Table 3. Results of direct
mm), as shown in Fig. 12. shear test under different normal stresses are depicted in Fig. 13.
Based on the procedure explained in Section 5, numerical sam- This results show that the proposed model can reproduce the shear
ples with width 100 mm and height 40 mm (the same dimensions behaviour of planar joints.
as the physical specimens) are generated with planar interface. In
the initial step, side walls and interface walls are removed and the 6.2. Direct shear test on saw tooth triangular joints
assembly is allowed to reach the static equilibrium. The specimen
is loaded by a pair of opposing frictionless walls. The top wall acts Three saw-tooth triangular joint profiles with the base angle
as a loading platen and the bottom wall is stationary. These walls h = 15°, 25° and 35° are generated and direct shear test under
110 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115

7
Patton 12

Peak Shear Strength (MPa)

Peak Shear Strength (MPa)

Patton
6 Ladanyi & Archambault
Numerical model 10 Ladanyi & Archambault
5 Numerical model
8
4
6
3

2 4

1 2

0 0
0 1 2 3 4 5 0 1 2 3 4 5
Normal Stress (MPa) Normal Stress (MPa)
(a) (b)
12
Peak Shear Strength (MPa)
Patton
10 Ladanyi & Archambault
Numerical model
8

0
0 1 2 3 4 5
Normal Stress (MPa)
(c)
Fig. 16. Investigation of the suggested model’s ability to reproduce the peak shear strength of saw-tooth triangular joints, (a) h = 15°, (b) h = 25° and (c) h = 35°.

different normal stresses are carried out. Results of the direct shear with Patton model. When the normal stress exceeds 3 MPa, shear-
tests on the saw-tooth triangular joint with the base angle of 15° ing of asperities occurs and the peak shear strength of numerical
are shown in Figs. 14 and 15. As the normal stress increases, the
peak shear strength increases. When the normal stress is less than
3 MPa, shearing takes place by sliding along the joint. At a normal θ = 15°
stress of 1 MPa, when the shear stress reaches the peak value,
plastic behaviour starts and sliding takes place along the joint
(Fig. 14a). Only limited number of cracks is observed during shear-
ing (Fig. 15). As the normal stress increases, the stress concentra-
tions at asperities increase and more tensile cracks are observed.
At a normal stress of 3 MPa, asperity wearing is observed in the
model. For normal stresses higher than 3 MPa, asperities are
sheared-off and the shear stress after reaching to the peak stress θ = 25°
drops to residual value. As the normal stress increases, the peak
shear displacement dpeak (shear displacement at peak shear
strength) and shear stiffness increase. The increase of normal stress
also results in a decrease in normal displacement.
Results of direct shear tests on saw-tooth triangular joints un-
der different normal stresses are presented in Fig. 16. These results
are compared with Patton [2] and Ladanyi and Archambault [3,6]
models which were developed on triangular joint profiles. Due to
lack of data for the joint cohesion cj, comparison of numerical mod- θ = 35°
els against Paton model is carried out for only the sliding mecha-
nism (Eq. (1)). The Ladanyi and Archambault model is derived
using the mechanical properties of Hawkesbury sandstone. The
transition pressure, rT, is obtained 26.3 MPa using Mogi [42] ap-
proach. Goodman [43] suggested that in the absence of sufficient
data, this transition pressure can be selected as the uniaxial com-
pressive of intact rock (rc = 27.4 MPa). The selected transition
pressure is close to the Goodman proposed value.
At the asperity base angle of 15°, when the normal stress is less
than 3 MPa, shearing takes place in sliding mode and the peak Fig. 17. Asperity degradation of saw-tooth triangular joints under the constant
shear strength of the numerical models are in good agreement normal stress of 2 MPa (shear displacement = 3 mm).
 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115 111

profiles corresponding to different JRC values. They suggested that

the JRC value can be estimated by visual comparison to these stan-
dard profiles.
In order to investigate the ability of the suggested model in
reproducing the shear strength and dilation angle of real rock joint
profiles, numerical direct shear tests on standard JRC profiles are
carried out and the results are compared against Barton and Chou-
bey [10] model. The standard JRC profiles were originally obtained
by using a profile comb which can measure profile points every 0.5
to 1 mm. Therefore, these standard JRC profiles are digitised with a
horizontal spacing of 0.5 mm. Ten series of shear box samples are
generated and direct shear tests under different normal stresses
are carried out. Results of the direct shear test on JRC 16–18 are
presented in Figs. 18 and 19. As the normal stress increases, the
shear stiffness increases and the maximum shear strength is
achieved at a higher shear displacement. At low normal stresses
the second order asperities have an influence and limited number
of fracturing occurs on steep-sided roughness features (Fig. 19). As
the normal stress increases, shearing of first order and second or-
der asperities occurs and first order asperities take over the shear-
ing behaviour. Therefore, the peak shear strength increases and
dilation rate decreases (Fig. 18). Stress concentration at some
asperities results in penetration of tensile cracks inside the intact
rock material (Fig. 19, normal stress = 5 MPa).
Results of the comparison between the peak shear strength of
numerical models and Barton and Choubey model for 10 standard
JRC profiles are presented in Fig. 20. The JRC values of joint profiles
are assumed to be the experimental back-calculated value of
Barton and Choubey for each standard profile and the JCS is the

Normal stress = 0.5 MPa

Fig. 18. Results of numerical direct shear tests on JRC 16–18 (a) shear stress versus
shear displacement and (b) normal displacement versus shear displacement.

models falls below the Patton sliding model. At the asperity base
angle of 25°, transition from sliding to shearing mode takes place
Normal stress = 1 MPa
at normal stress of 2 MPa while at the asperity base angle of 35°,
the transition stress is reduced to 1 MPa. Good agreement is found
between the results of numerical models and Paton and Ladanyi
and Archambault models in reproducing the peak shear strength.
Effect of the asperity angle on the joint surface degradation
under the constant normal stress of 2 MPa is shown in Fig. 17. As
the asperity angle increases, stress concentration at triangular
asperities increases and results in greater damage to the asperities.
Normal stress = 3 MPa
At triangular base angle h = 15°, shearing takes place by sliding
along the asperities and only limited number of cracks occurs in
model. At h = 25°, the first asperity is worn out but the shearing
is controlled by sliding along other asperities. At h = 35°, the high
stress concentration at the asperities results in the asperity shear-
ing-off. This shows that at low asperity base angles, sliding is the
dominant shear mechanism while for steeper asperities crushing
and asperity shearing is the controlling mechanism. It is found that Normal stress = 5 MPa
asperity degradation occurs by initiation of tensile cracks. Propaga-
tion of these tensile cracks results in the asperities shearing.

6.3. Standard JRC profiles

The most widely used criterion for the estimation of the shear
strength of rock discontinuities is Barton and Choubey [10] model.
They carried out extensive series of experiments on various rock Fig. 19. Effect of normal stress on asperity degradation of JRC 16–18 after 3 mm
joint profiles and published a set of 10 standard joint roughness shear displacement.
112 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115

Fig. 20. Investigation of the suggested model’s ability to reproduce the peak shear strength of standard JRC profiles.

uniaxial compressive strength of intact rock (rc = 27.4 MPa). Re- strength for other JRC profiles are in good agreement with Barton
sults of the numerical models shows that the peak shear strength and Choubey model. The results show that this method can repro-
is slightly overestimated for JRC 6–8 and 8–10 while the peak shear duce the peak shear strength.
 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115 113

Fig. 21. Investigation of the suggested model’s ability to reproduce the peak dilation angle of standard JRC profiles.

Normal displacement is recorded during the shearing process the pre-yield stage (before shear displacement of 0.5 mm) only
and the peak dilation angle is calculated at the peak shear displace- minor cracks occur in the model and shear displacement takes
ment dpeak. The peak dilation angles of numerical models are com- place by internal deformation of the intact rock and joint surface.
pared against Barton and Choubey model, as shown in Fig. 21. For When the shear stress exceeds 80% of the peak shear strength,
JRC profiles of 0–2 and 2–4, the peak dilation angle is very small in signs of yielding in the shear stress curve are observed by a depar-
both the numerical and Barton and Choubey models. For other ture from linear elastic behaviour. The surface damage is observed
joint profiles, a good agreement is found between the numerical to commence when the yield stress is passed. From the yield stress,
models and Barton and Choubey model. The peak dilation angle reduction of effective area results in stress concentration at asper-
is underestimated at JRC 4–6 and JRC 6–8 and overestimated at ities. The stress at some asperities exceeds the bonds normal
JRC 12–14 and JRC 18–20 under low normal stress, while for other strength of particles and tensile cracks occur at these asperities.
JRC profiles satisfactorily estimated. These results show that the From the peak stress, tensile cracks develop at these asperities
shear box genesis approach has the ability to reproduce the shear which lead to shear strength reduction. From the peak shear stress
strength and the peak dilation angle of real rock joint profiles. to residual shear stress the number of cracks increases rapidly. As
Joint surface degradation during the shearing process for JRC the shear displacement continues, further crushing of the joint
18–20 under the normal stress of 4 MPa is shown in Fig. 22. In asperities takes place.
114 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115

Fig. 22. Asperity degradation during shearing process for standard JRC 18–20 under normal stress of 4 MPa (a) shear stress, normal displacement and cumulative crack
number during the shearing process and (b) corresponding model state (green: smooth joint, red: tensile crack and blue: shear crack). (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

7. Conclusions of these tensile cracks and crushing of the asperities leads to reduc-
tion of shear stress to residual stress and rapid increase in the
In this paper, the shear behaviour of rock joints in direct shear number of cracks.
tests are numerically simulated using PFC2D. Previously, joints
were modelled in PFC using two approaches, namely the bond re-
moval method and the smooth joint model. It is found that simu- Acknowledgements
lation of joints using the bond removal method results in an
unrealistic increase of shear strength and normal displacement The authors thank Dr. Xavier Garcia, Dr. David Potyondy and Dr.
due to inherent micro-scale roughness of the joint surface. The Matthew Pierce from Itasca Consulting group and Dr. Dion Weath-
smooth joint model also suffers from particle interlocking which erley from the University of Queensland for their invaluable tech-
occurs at shear displacement greater than the minimum diameter nical help and comments.
of the particles. In this study, the shear box genesis approach is
presented to resolve this problem. The ability of this method in
reproducing the shear behaviour of saw-tooth triangular joints References
and standard JRC profiles is demonstrated. Results of the numerical
models are compared against well-known empirical models and [1] Hoek E. Shear strength of discontinuities. Practical rock engineering; 2007. p.
14.
good agreement is found between them.
[2] Patton FD. Multiple modes of shear failure in rock. In: 1st ISRM congress.
The asperity degradation during the shearing process is studied Lisbon, Portugal; 1966. p. 509–15.
numerically. It is found that in the pre-yield stage only minor [3] Ladanyi B, Archambault G. Simulation of shear behavior of a jointed rock mass.
cracking takes place and when the shear stress reaches the yield In: The 11th US rock mechanics symposium (USRMS). Berkeley, CA; 1969. p.
105–25.
shear stress, reduction of the effective area results in stress concen- [4] Kodikara JK. Shear behaviour of rock-concrete joints and side resistance of
tration and occurrence of tensile cracks at asperities. Propagation piles in weak rock. PhD thesis, Melbourne, Australia: Monash University; 1989.
 M. Bahaaddini et al. / Computers and Geotechnics 51 (2013) 101–115 115

[5] Seidel JP, Haberfield CM. The application of energy principles to the [26] Asadi MS. Experimental and PFC2D numerical study of progressive shear
determination of the sliding resistance of rock joints. Rock Mech Rock Eng behaviour of single rough rock fractures. PhD thesis, Perth, Australia: Curtin
1995;28(4):211–26. University; 2011.
[6] Ladanyi B, Archambault G. Direct and indirect determination of shear strength [27] Cho N. Discrete element modeling of rock pre-peak fracturing and dilation.
of rock mass. In: Preprint number 80-25 AIME annual meeting. Las Vegas, PhD thesis, Edmonton, Alberta: University of Alberta; 2008.
Nevada; 1980. [28] Sharrock GB, Akram MS, Mitra R. Application of synthetic rock mass modeling
[7] Fairhurst C. On the validity of the ‘Brazilian’ test for brittle materials. Int J Rock to estimate the strength of jointed sandstone. In: 43rd US Rock mechanics
Mech Min Sci Geomech Abstr 1964;1(4):535–46. symposium and 4th US – Canada rock mechanics symposium. Asheville, North
[8] Hoek E, Bray J. Rock slope engineering. London: The Institution of Mining and Carolina; 2009.
Metallurgy; 1977. [29] Potyondy DO, Cundall PA. A bonded-particle model for rock. Int J Rock Mech
[9] Seidel JP. The analysis and design of pile shafts in weak rock. PhD thesis, Min Sci 2004;41(8):1329–64.
Melbourne, Australia: Monash University; 1993. [30] Cundall PA. Numerical experiments on rough joints in shear using a bonded
[10] Barton N, Choubey V. The shear strength of rock joints in theory and practice. particle model. In: Aspects of tectonic faulting. Berlin; 2000. p. 1–9.
Rock Mech 1977;10(1–2):1–54. [31] Kusumi H, Matsuoka T, Ashida Y, Tatsumi S. Simulation analysis of shear
[11] Olsson R, Barton N. An improved model for hydromechanical coupling during behavior of rock joint by distinct element method. In: Eurock 2005 – impact of
shearing of rock joints. Int J Rock Mech Mining Sci 2001;38(3):317–29. human activity on geological environment. London; 2005. p. 281–6.
[12] Indraratna B, Haque A. Experimental and numerical modeling of shear [32] Jing L, Stephansson O. Fundamentals of discrete element methods for rock
behaviour of rock joints. In: GeoEng 2000, An international conference on engineering: theory and applications. Amsterdam: Elsevier; 2007.
geotechnical and geological engineering. Pennsylvania, USA; 2000. [33] Lai ZHF, Douglas KJ, Mostyn G. The strength of rock defects – numerical
[13] Vosniakos K. Physical and numerical modelling of shear behaviour of saw- analysis of scale effects. In: 11th congress of the international society for rock
toothed filled rock joint. PhD thesis, Manchester: University of Manchester; mechanics. Lisbon, Portugal; 2007. p. 481–4.
2007. [34] Rasouli V, Harrison JP. Assessment of rock fracture surface roughness using
[14] Oh JM. Three dimensional numerical modeling of excavation in rock with Riemannian statistics of linear profiles. Int J Rock Mech Min Sci
dilatant joints. PhD thesis, Urbana, Illinois: University of Illinois; 2005. 2010;47(6):940–8.
[15] Karami A, Stead D. Asperity degradation and damage in the direct shear test: a [35] Bahaaddini M, Sharrock G, Hebbelwhite B, Mitra R. Direct shear tests to model
hybrid FEM/DEM approach. Rock Mech Rock Eng 2008;41(2):229–66. the shear behaviour of rock joints by PFC2D. In: 46th US rock mechanics/
[16] Hunt SP, Meyers AG, Louchnikov V. Modelling the Kaiser effect and geomechanics symposium. Chicago, USA; 2012.
deformation rate analysis in sandstone using the discrete element method. [36] Pierce M, Cundall P, Potyondy D, Mas Ivars D. A synthetic rock mass model for
Comput Geotech 2003;30(7):611–21. jointed rock. In: Rock mechanics: meeting society’s challenges and demands.
[17] Fakhimi A, Villegas T. Application of dimensional analysis in calibration of a 1st Canada–US rock mechanics symposium. London: Vancouver; 2007.
discrete element model for rock deformation and fracture. Rock Mech Rock [37] Lambert C, Buzzi O, Giacomini A. Influence of calcium leaching on the
Eng 2007;40(2):193–211. mechanical behavior of a rock-mortar interface: a DEM analysis. Comput
[18] Tran TH, Vénier R, Cambou B. Discrete modelling of rock-ageing in rockfill Geotech 2010;37(3):258–66.
dams. Comput Geotech 2009;36(1–2):264–75. [38] Lambert C, Coll C. A DEM approach to rock joint strength estimate. In: Rock
[19] Hazzard JF, Young RP, Maxwell SC. Micromechanical modeling of cracking and slope stability 2009. Santiago, Chile; 2009.
failure in brittle rocks. J Geophys Res 2000;105(B7):16683–97. [39] Mas Ivars D, Pierce ME, Darcel C, Reyes-Montes J, Potyondy DO, Young RP, et al.
[20] Zhang ZX, Hu XY, Scott KD. A discrete numerical approach for modeling face The synthetic rock mass approach for jointed rock mass modelling. Int J Rock
stability in slurry shield tunnelling in soft soils. Comput Geotech Mech Min Sci 2011;48(2):219–44.
2011;38(1):94–104. [40] Bahaaddini M, Sharrock G, Hebblewhite BK. Numerical investigation of the
[21] Akram MS, Sharrock GB. Physical and numerical investigation of a cemented effect of joint geometrical parameters on the mechanical properties of a non-
granular assembly of steel spheres. Int J Numer Anal Methods Geomech persistent jointed rock mass under uniaxial compression. Comput Geotech
2010;34(18):1896–934. 2013;49:206–25.
[22] Itasca Consulting Group Inc. PFC2D manual, version 4.0 Minneapolis, [41] Bandis SC, Lumsden AC, Barton NR. Fundamentals of rock joint deformation.
Minnesota; 2008. Int J Rock Mech Min Sci Geomech Abstr 1983;20(6):249–68.
[23] Cho N, Martin CD, Sego DC. A clumped particle model for rock. Int J Rock Mech [42] Mogi K. Pressure dependence of rock strength and transition from brittle
Min Sci 2007;44(7):997–1010. fracture to ductile flow. Bull Earthq Res Inst 1966;44:215–32.
[24] Park JW, Song JJ. Numerical simulation of a direct shear test on a rock joint [43] Goodman RE. Methods of geological engineering in discontinuous rocks. West
using a bonded-particle model. Int J Rock Mech Min Sci 2009;46(8):1315–28. Publishing Co.; 1976.
[25] Asadi M, Rasouli V, Barla G. A bonded particle model simulation of shear
strength and asperity degradation for rough rock fractures. Rock Mech Rock
Eng 2012;45(5):649–75.