Engineering Geology 204 (2016) 4152

Contents lists available at ScienceDirect

Engineering Geology
journal homepage: www.elsevier.com/locate/enggeo

Experimental and numerical study of asperity degradation in the direct

shear test
M. Bahaaddini a,b,, P.C. Hagan b, R. Mitra b, M.H. Khosravi c
a
b
c

Zarand Higher Education Complex, Shahid Bahonar University of Kerman, Kerman, Iran
School of Mining Engineering, UNSW Australia, Sydney, Australia
School of Mining Engineering, College of Engineering, University of Tehran, Tehran, Iran

a r t i c l e

i n f o

Article history:
Received 23 September 2015
Received in revised form 11 December 2015
Accepted 31 January 2016
Available online 3 February 2016
Keywords:
Direct shear test
Shearing mechanism
Asperity degradation
PFC
Smooth joint model

a b s t r a c t
In this paper, the shear behaviour and mechanisms of asperity degradation of rock joints under direct shear tests
were studied using numerical and experimental approaches. PFC2D was used for numerical simulations, in which
the intact material is simulated by a dense packing of circular particles bonded together at their contact points
and by breakage of these bonds under loading regimes, the damage process is simulated. The joint interfaces
were simulated by a newly developed modied smooth joint model in which micro-scale slip surfaces are
applied at contacts between particles of upper and lower blocks of the shear box. In order to study the ability
of this numerical approach in reproducing the shearing mechanisms and asperity degradation of rock joints in
direct shear tests, a comparative study was carried out against the physical experiments. Experimental and
numerical direct shear tests were carried out on saw-tooth triangular joints with the base angles of 20 and
30 under different normal stresses. Three shearing mechanisms of sliding, surface wear and asperity shearing
off were observed in these experiments. The comparison of the shear behaviour and mechanisms of asperity
degradation of physical and numerical experiments showed that the results of numerical models are in good
agreement with physical experiments and this numerical approach can reproduce the shear behaviour of rock
joints under different loading conditions.
2016 Elsevier B.V. All rights reserved.

1. Introduction
It is well understood that the mechanical behaviour of jointed rock
masses is greatly inuenced by the mechanical properties of joints,
especially at shallow depths (Singh and Rao, 2005). Joint surface roughness has been recognised as one of the parameters having a signicant
impact on the mechanical behaviour of joints and numerous researchers
have investigated its effect on the shear behaviour of rock joints
(Asadollahi and Tonon, 2010; Barton, 1971; Barton and Choubey,
1977; Grasselli, 2001; Grasselli and Egger, 2003; Grasselli et al.,
2002; Kulatilake et al., 1995; Ladanyi and Archambault, 1969, 1980;
Patton, 1966).
Patton (1966) was among the rst who developed a bilinear shear
strength model for estimation of the shear strength of rock joints.
In this bilinear model, it is assumed that when the applied normal stress
n is less than a specic stress T, shear strength is controlled by

sliding along the joint but when n exceeds T, the shear behaviour is
controlled by shearing the asperities. However, in reality sliding and
shearing take place simultaneously. Difculty in determination of the
joint cohesion is another shortcoming of this approach (Seidel and
Habereld, 1995). Ladanyi and Archambault (1969) developed a shear
strength model, based on the work and energy principles, as follows:


n 1as v_ tan as SR

http://dx.doi.org/10.1016/j.enggeo.2016.01.018
0013-7952/ 2016 Elsevier B.V. All rights reserved.

where v_ is the dilation rate, is the joint friction angle and as is the
sheared area ratio. SR is the intact rock strength which was suggested
to be estimated by the Fairhurst (1964) intact rock strength criterion,
as follows:
SR c

Corresponding author at: Zarand Higher Education Complex, Shahid Bahonar

University of Kerman, Kerman, Iran.
E-mail addresses: m.bahaaddini@unsw.edu.au, m_bahaaddini@uk.ac.ir
(M. Bahaaddini), p.hagan@unsw.edu.au (P.C. Hagan), r.mitra@unsw.edu.au (R. Mitra),
mh.khosravi@ut.ac.ir (M.H. Khosravi).

11as v_ tan

p


n 11
n 0:5
1n
n
c

where c and n are the uniaxial compressive strength and the ratio of
tensile to uniaxial compressive strength of the intact rock, respectively.
_
The particular problem in this model is the estimation of as and v.
Ladanyi and Archambault (1980), by undertaking a large number of

42

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

Fig. 1. Numerical simulation of rock joints in PFC: (a) Bond removal method and (b) Smooth joint model.

experimental tests on triangular joints, proposed the following empirical equations for estimation of these parameters.


n k1
as 1 1
T

k2
n
v_ 1
tan i
T

3
4

where T is the transition stress at which the strength of the rock joint is
equal to that of intact rock and i is the asperity angle. They proposed the
empirical values of 1.5 and 4.0 for the constants of k1 and k2,

respectively. However, determination of T and i are challenging, especially for real irregular rock joints (Kodikara, 1989; Seidel and
Habereld, 1995).
The Barton model (Barton, 1973; Barton and Choubey, 1977) is the
most widely used empirical model for estimation of the shear behaviour
of rock joints:

 

JCS
n tan JRC log
r
n

where joint roughness coefcient JRC, joint compressive strength JCS and
residual friction angle r are the parameters of Barton model. The main

Fig. 2. Procedure of simulation of direct shear test using shear box genesis (a) Vessel generation, (b) Filling up the vessel by randomly placed particles, (c) Application of isotropic stress,
(d) Elimination of oaters, (e) Installation of bonds between particles and (f) Application of smooth joint contacts (Bahaaddini et al., 2013a).

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

43

Fig. 3. Numerical simulation of real rock joint proles using the modied smooth joint model and ability of this approach in reproducing asperity degradation during the shearing process:
(a) At initial stage of test (b) after 3 mm shear displacement (dark blue: shear crack, red: tensile crack).

difculty in this model is the determination of the input parameters,

especially for JRC. A large number of empirical, statistical and fractal
methods have been proposed to measure the joint surface roughness.
However, a universally accepted method for estimation of the joint
roughness is still not available.
A large number of studies can be found in literature that have tried to
investigate the effect of joint roughness on the shear behaviour of rock
joints by means of experimental, practical and analytical approaches.
However, it is still not well understood which is related to difculty in
quantitative representation of the joint roughness and unknown mechanisms of asperity degradation in the shearing process (Bahaaddini
et al., 2013b).
Some attempts have been carried out to estimate the shear behaviour of rock joints using numerical approaches. Most of these studies
have employed constitutive models (Indraratna and Haque, 2000; Oh,
2005; Oliveira and Indraratna, 2010; Vosniakos, 2007). In these constitutive models, the deformation behaviour of the system is controlled
by parameters which are predened by the user. These models cannot
explicitly simulate the real geometry of rock joints. Furthermore, the
intact material consists of angular blocks which cannot break under
loading during the test. Therefore, they are not able to track the asperity
degradation and propagation of cracks inside the intact material.
Particle ow code (PFC) is a discrete element code which has been
used by several researchers to study the shear behaviour of rock joints.
In PFC, the intact rock is simulated by a dense packing of spherical
(PFC3D) or circular (PFC2D) particles bonded together at their contact
points. The generated material is called bonded particle model (BPM)
and the mechanical behaviour of BPM is governed by the microproperties of its constituents; particles and bonds. The deformation
behaviour of the BPM is controlled by two sets of stiffness; normal
and shear stiffness of particles, kn and ks, and normal and shear stiffness
n
s
of bonds, k and k , respectively. The strength of the material is controlled
 c and c , and particle coefcient of
by bond normal and shear strength,
friction, . The bond between particles breaks if the applied stress
exceeds the bond strength. When the bond breaks under shear
stress, the shear strength at that contact reduces to residual value
which is dependent on and the applied normal stress. Failure of
bonds under normal or shear stresses results in the generation of
tensile or shear cracks between particles which enable this numerical approach to study the crack initiation and propagation under different loading regimes (Bahaaddini et al., 2016; Itasca Consulting
Group Inc., 2008; Potyondy and Cundall, 2004).
This paper aims to study the shearing mechanism and asperity degradation of rock joints in the direct shear tests using PFC2D. Traditionally, joint interfaces have been modelled in PFC by the bond removal or
smooth joint models. In the following section, shortcomings of these approaches have been investigated. To overcome the deciencies of these
methods, recently Bahaaddini (2014) developed the modied smooth
joint model that has shown promising results in estimation of the
peak shear strength and peak dilation angle of rock joints. However,
the ability of this numerical approach in reproducing the shearing
mechanisms and asperity degradation of rock joints has not been studied. To this end, different synthetic joint proles were prepared in the

laboratory and direct shear tests under different normal stresses were carried out. Numerical models with the same geometries were prepared
and results of numerical models and experimental tests were then
compared.

2. Numerical simulation of direct shear test in PFC

The most commonly used method for generation of rock joints in
PFC is the bond removal method. In this approach, the positions of
bonds between particles are checked relative to the joint plane and
bonds of particles which lie along the joint plane are removed. A generated joint plane using the bond removal approach is shown in Fig. 1(a).
This method has been used by several researchers for simulation of the
direct shear test (Asadi et al., 2012, 2013; Cundall, 2000; Kusumi et al.,
2005; Park and Song, 2009, 2013). However, this method suffers from
an inherent micro-scale roughness of the joint surface, as shown in
Fig. 1(a). This deciency is related to the circular shape, uneven size
and random distribution of particles (Bahaaddini et al., 2015, 2013b)
which results in an unrealistic increase of shear strength and dilation
along the joint (Bahaaddini et al., 2013c; Mas Ivars et al., 2011).
Smooth joint (SJ) model is a recently developed approach to overcome the shortcomings of the bond removal method (Pierce et al.,
2007). In the SJ model, the joint plane is dened and micro-scale slip
surfaces (smooth joint contacts) are applied at contacts between particles that lie on the opposite sides of the joint plane, as shown in
Fig. 1(b). By application of SJ contact between particles, the bond
between particles (if it exists) is removed and particles are allowed to
overlap or pass through each other. The orientation of SJ contact is parallel to the dened joint plane. The mechanical behaviour of each microscale SJ contact obeys the Coulomb sliding model and is controlled by SJ
normal stiffness knj , SJ shear stiffness ksj and SJ coefcient of friction j
(Bahaaddini et al., 2015; Itasca Consulting Group Inc., 2008).
Recently, Bahaaddini et al. (2013b) investigated the ability of the
SJ model in reproducing the shear behaviour of rock joints in the direct
shear test. Results of their study show that at large shear displacements,
particle interlocking occurs which is related to the difculty in
recognising of upper and lower blocks' particles during the shearing
process and also proper application of SJ contacts between these
particles. The particle interlocking leads to an unrealistic increase
of shear strength and dilation of rock joints.

Table 1
Mechanical properties of Hydro-Stone TB.
Test

Uniaxial

Brazilian

E
v
c
(MPa) (GPa)
Average
55.93
Standard deviation 4.58
Coefcient of
8.18
variation (%)

9.26
0.50
5.40

Triaxial

t
c

mi
(MPa) (MPa) (Deg.)

0.245 5.06
0.021 0.44
8.73
8.66

12.85

40.86

Coefcient of variation = (Average/Standard Deviation) 100

ci

9.15 55.65

44

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

Table 2
Calibrated micro-properties of particles and bonds.

Fig. 4. Servo-control hydraulic testing machine RDS-300 (GCTS Testing Systems, 2016).

To overcome this issue, Bahaaddini et al. (2013b) proposed the shear

box genesis approach in which the process of sample generation and
application of smooth joint contacts has been modied. The procedure
of shear box genesis method is shown in Fig. 2. In this approach, the
upper and lower vessels of the shear box, consisting of frictionless
walls, are generated separately (Fig. 2a) and a dense packing of nonuniform sized particles ll each vessel (Fig. 2b). A specic isotropic
stress is applied between particles and by using measurement spheres
the magnitude of stress is controlled (Fig. 2c). The particles with less
than three contacts (oaters) are eliminated (Fig. 2d) and then parallel
bonds are installed between particles at their contact points (Fig. 2e). In
the nal stage, the interface walls between the upper and lower blocks
are eliminated and the normal stress is exerted to the upper block. Then,
the assembly is allowed to reach static equilibrium. New contacts are
created between the particles of upper and lower blocks, and smooth
joint contacts are applied at these contacts (Fig. 2f) (Bahaaddini et al.,
2013b). This method resolves the deciency of particle interlocking in
the SJ model by recognition of upper and lower blocks' particles and
proper application of SJ model at these contacts through all stages of
numerical experiments.
The validity of this approach was investigated by undertaking direct
shear tests on ten standard JRC proles. The peak shear strength and the

Particle micro-scale properties

Parallel bonds micro-scale properties

Ball density (kg/m3)

1716

Ec (GPa)
Coefcient of friction
kn/ks

6.7
0.7
1.8

Ec (GPa)
Normal strength (MPa)
Shear strength (MPa)
n

k =k

6.7
40 9
40 9
1.8

peak dilation angle of these numerical experiments were compared

against the Barton model (Barton, 1973; Barton and Choubey, 1977)
and good agreement was found (Bahaaddini et al., 2013b). In this
approach, the real prole of rock joints can be explicitly simulated and
the asperity degradation of rock joints during the shearing process can
be investigated in terms of tensile and shear cracks, as shown in Fig. 3.
Therefore, it provides a scientic tool to study the shearing mechanisms
of rock joints. However, the ability of this approach in reproducing the
shearing mechanisms and asperity degradation of rock joints in the
direct shear tests has not yet been studied.
This approach can also be employed in PFC3D for 3D simulation of
rock joints. The minimum size of asperities that can be simulated in
PFC is controlled by the size of particles. Size of particles should be
small enough to simulate the micro-scale asperities. However, selection
of a small particle size leads to an increase in the number of particles and
consequently an increase in the computation time. This problem can be
very restrictive for large scale joints and numerical simulation of 3D
joint surfaces. Application of pbricks and parallel processing or Adaptive
Continuum/DisContinuum logic may resolve this problem (Bahaaddini
et al., 2014).
As discussed, the intact material in PFC is represented as an assembly
of particles bonded together and the micro-properties of these particles
and bonds control the macro-scale behaviour. The joint surface is also
dened by micro-scale SJ contacts and the properties of these contacts
need to be determined before undertaking the numerical experiments.
There is no direct method or experiment to obtain these micro-scale parameters in the laboratory and the only conventional method to estimate
these parameters is by trial and error through a direct comparison of the
macro-scale behaviour of the material by undertaking conventional laboratory experiments (Cho, 2008; Hsieh et al., 2008; Potyondy and Cundall,
2004). Therefore, an essential step before undertaking numerical experiments is the proper selection of these micro-scale parameters through
the calibration process.
3. Experimental study
The synthetic material used for the experimental study of this
research was the gypsum cement, Hydro-Stone TB,. Hydro-Stone TB is
a mixture of plaster of Paris, Portland cement and crystalline silica. For
sample preparation, the Hydro-Stone TB was mixed with water in the
ratio of 3.125:1 by the weight (32 parts of water per 100 parts of
Hydro-Stone TB). The mixture was poured into moulds and after 24 h
they were removed from the moulds and placed in an oven to cure at
a temperature of 55 C for 28 days.
To measure the mechanical properties of the Hydro-Stone TB, the
uniaxial compressive test, triaxial and Brazilian tests were undertaken.
These experiments were carried out using an MTS test machine. For
uniaxial and triaxial tests, cylindrical samples with diameters of 41 mm
and heights of 86 mm were prepared and experiments were undertaken
by following the suggested methods outlined by the International Society
Table 3
Results of numerical and physical experimental uniaxial compressive tests.

Fig. 5. Numerical modelling of direct shear test under constant normal stress.

Physical experiment
Numerical experiment

UCS (MPa)

E (GPa)

55.93
55.74

9.26
9.42

0.245
0.238

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

45

Fig. 6. Comparison between the results of experimental and numerical normal deformability
tests on planar joints.

Fig. 7. Experimental and numerical direct shear tests on the planar joint under the normal
stress of 1 MPa.

of Rock Mechanics (ISRM) (Ulusay and Hudson, 2007). For measurement

of the tensile strength, Brazilian disc samples with the thickness to diameter ratio of 0.2 to 0.75 were prepared (ASTM, 2008b) and experiments
followed the procedure outlined by the ISRM (Ulusay and Hudson,
2007). The results of these experiments are summarised in Table 1.
Based on the results of uniaxial and triaxial compressive experiments,
the Couloumb parameters (c, ) and HoekBrown (mi, ci) parameters
were determined. In these experiments, the conning pressure (3)
was in the range of 06 MPa.

Normal and shear loads are measured by load cells, and the normal
and shear displacements are measured by Linear Variable Differential
Transducers (LVDTs). The normal displacement is measured by four
LVDTs positioned in a square pattern around the shear box. Each of
these LVDTs has a measurement range of 12 mm and sensitivity of
0.0025 mm. The average value of these four LVDTs is used as a normal
displacement of the upper block. The shear displacement is measured
by relative displacement of the lower box to the upper box in the horizontal direction using one LVDT. This LVDT has a measurement range of
100 mm and sensitivity of 0.025 mm.

3.1. Experimental direct shear test

4. Numerical modelling of direct shear test

Three joint types of planar, triangular joints with the base angles of
20 and 30 were prepared and direct shear tests under different normal
stresses were carried out. To prepare the test samples, metal moulds having a length of 100 mm, width of 100 mm and height of 50 mm were fabricated. To ensure the atness of the bottom surface of sample blocks prior
to testing, the bottom surfaces were ground using a grinder.
A servo-control hydraulic testing machine (RDS-300 manufactured
by Geotechnical Consulting and Testing Systems) was employed for direct shear tests. Different sections of this machine are shown in Fig. 4.
The shear box is made up of upper and lower parts. In this machine,
the upper part moves vertically and the lower part moves horizontally.
Two actuators are used to apply the forces in vertical and horizontal directions. The servo-control hydraulic system is composed of an electrohydraulic normal actuator with 500 kN load capacity, 100 mm stroke
and a 19 LPM electro-hydraulic servo valve for closed-loop control of
normal load or normal displacement. The electro-hydraulic shear actuator of this system has the load capacity of 300 KN with 50 mm stroke
and a 19 LPM electro-hydraulic servo valve for closed-loop control of
shear load or shear displacement. Two linear rail bearings are employed
for guidance of the lower box in a controlled linear displacement.

Based on the procedure explained in Section 2, numerical specimens

with lengths of 100 mm and heights of 100 mm were generated in
PFC2D. Each specimen consisted of around 30,000 particles. A schematic
representation of the numerical simulation of direct shear test under
constant normal stress is presented in Fig. 5. The normal stress was applied vertically to the upper block, and by using a servo-control mechanism (Itasca Consulting Group Inc., 2008) this normal stress was kept
constant during the shearing process. The lower block was restrained
in the vertical direction and only allowed to move in the horizontal
direction. A horizontal velocity of 0.1 m/sec was applied to the lower
block. A preliminary study showed that this displacement rate was sufciently slow to ensure that the system remained in the quasi-static

Table 4
Calibrated smooth joint parameters and results of numerical and physical normal
deformability and direct shear tests.
Calibrated SJ
parameters

Results of experiments

 ; k (MPa/mm)
SJ normal stiffness k
n ji
njf

 (MPa/mm)
SJ shear stiffness ks ji ; k
sjf
SJ coefcient of friction j (tan(j))
Parameter
System normal stiffness
(MPa/mm)
System shear stiffness (MPa/mm)
Friction angle (deg.)

25, 225
5, 55
0.73

Physical Numerical
14.2
13.9
2.24
36.4

2.03
36.2

Fig. 8. The peak shear strength of numerical and physical direct shear tests on planar joints
under different normal stresses.

46

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

Fig. 9. Results of experimental direct shear tests on triangular joint with the base angle of 20: (a) Shear stress versus shear displacement and (b) Normal displacement versus shear
displacement.

condition. The upper block was restrained against the horizontal displacement but it could move vertically during the shearing process.
The shear displacement was measured by recording the horizontal
displacement of the lower block and the normal displacement was
measured by tracing the vertical displacement of the upper wall. The
shear stress was measured by dividing the reaction force on wall id 3
(Fig. 5) over the joint length.
5. Calibration of numerical model
5.1. Calibration of bonded particle model
In order to calibrate the micro-scale properties of BPM, numerical
samples having a width of 41 mm and height of 86 mm were generated.
The generation of a BPM sample involves the production of a dense
packing of well-connected particles with low locked-in forces and installation of parallel bonds at contacts between particles (Potyondy
and Cundall, 2004). The size of particles satises the uniform particlesize distribution and is controlled by pre-dened minimum particle radius Rmin and particle size ratio Rmax/Rmin (in this study Rmin = 0.25 mm
and Rmax/Rmin = 1.5). The calibration process involves selection of the
micro-scale properties by trial and error to reproduce the macro-scale

mechanical behaviour observed in the laboratory scale. The laboratory

scale mechanical properties which are commonly used for the calibration of BPM are uniaxial compressive strength UCS, elastic modulus E
and Poisson's ratio (Itasca Consulting Group Inc., 2008; Potyondy
and Cundall, 2004). The calibration involves an iterative process to
determine the micro-properties. First, E is calibrated. E is controlled by
particle and parallel bond modulus (Ec and Ec ) and particle and parallel
n
s
bond normal/shear stiffness ratio (kn/ks and k =k ). Then, is calibrated
n
s
which is affected by kn/ks and k =k . In the last step, UCS is calibrated
which is inuenced by the parallel bond normal and shear strength
 c and c ) (Itasca Consulting Group Inc., 2008; Potyondy and Cundall,
(
2004). The calibrated micro-scale properties of particles and bonds are
presented in Table 2. The results of the uniaxial compressive test on
the calibrated BPM model are compared against experimental tests in
Table 3. The results of numerical experiments are in good agreement
with physical experiments.

5.2. Calibration of smooth joint parameters

The SJ micro-properties were calibrated against the experimental
normal deformability and direct shear tests on planar joints. The SJ

Fig. 10. Results of experimental direct shear tests on triangular joint with the base angle of 30: (a) Shear stress versus shear displacement and (b) Normal displacement versus shear
displacement.

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

47

Fig. 11. Results of numerical direct shear tests on triangular joint with the base angle of 20: (a) Shear stress versus shear displacement and (b) Normal displacement versus shear
displacement.

normal stiffness knj was calibrated against the normal deformability test
and SJ shear stiffness ksj and SJ coefcient of friction j were calibrated
against the direct shear test.
In the physical normal deformability tests, the planar jointed block
of synthetic material having a length of 100 mm, width of 100 mm
and height of 100 mm was axially loaded and the applied normal stress
and normal displacement were recorded during the test. The result
from the physical normal deformability test is shown in Fig. 6. This
graph shows a distinct hyperbolic shape. Normal stress is observed
to increase at an increasing rate with displacement, hence the system
normal stiffness, which is the slope of this curve, increases with
displacement. This curve displays an initial low normal stiffness and as
the normal load increases, the contact between the upper and lower
blocks increases and the load distributed more uniformly over the
joint surface which results in the increase of normal stiffness (Bandis
et al., 1983; Resende et al., 2010).
In the numerical normal deformability tests, samples with the same
geometry of experimental samples were generated. The side and interface walls (which were used for sample generation) were removed and
the sample allowed reaching static equilibrium. A constant normal
velocity was applied to the upper wall and the normal stress and normal
displacement were recorded during the test.
The SJ normal stiffness was calibrated by trial and error to reproduce
the normal deformation behaviour of experimental normal deformability
tests. In order to reproduce the hyperbolic shape of normal stress-normal

displacement curve, two values of initial and nal SJ normal stiffness ( knji
and knjf ) were selected. The initial value was employed for SJ contacts
that were generated at the initial stage of the test. During the loading process, new SJ contacts were generated between the particles of the upper
and lower blocks. The nal SJ normal stiffness was applied at these new
generated contacts. Results of numerical and experimental normal
deformability tests are illustrated in Fig. 6, which shows that there is a
good agreement between the results of experimental and numerical
tests and this approach can reproduce the hyperbolic shape of the normal
stress-normal displacement curve.
In order to calibrate the SJ shear stiffness and SJ coefcient of friction,
numerical direct shear tests on planar joints were carried out. To
calibrate the SJ shear stiffness, the numerical direct shear test under a
constant normal stress of 1 MPa was undertaken and the SJ shear stiffness was varied to reproduce the system shear stiffness recorded in
the experimental tests. Then, direct shear tests under different normal
stresses were carried out and the SJ coefcient of friction was calibrated.
The calibrated smooth joint parameters are presented in Table 4. Results
of numerical and experimental direct shear tests on a planar joint under
the constant normal stress of 1 MPa are depicted in Fig. 7, which shows a
good agreement between the results of experimental and numerical
model. In the physical experiment, a very slight drop in the shear stress
after it reaches the peak value can be observed. This slight strainsoftening can be related to microscopic roughness of the planar surfaces,
as pointed out by Terzaghi (1920). However, in the numerical model the

Fig. 12. Results of numerical direct shear tests on triangular joint with the base angle of 30: (a) Shear stress versus shear displacement and (b) Normal displacement versus shear
displacement.

48

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

Fig. 13. Shearing mechanisms of experimental direct shear tests under different normal stresses on triangular joints with the base angle of 20 (direction of shearing is indicated by arrow).

Fig. 14. Shearing mechanisms of numerical direct shear tests under different normal stresses on triangular joints with the base angle of 20 (red: tensile crack and blue: shear crack).

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

Fig. 15. Comparison between the peak shear strengths of experimental tests, numerical
and analytical models for triangular joint with the base angle of 20.

joint surface is simulated by at SJ contacts in which when the shear

stress reaches the peak value, sliding takes place at these contacts and
the shear stress remains constant (Bahaaddini et al., 2015). The peak
shear strengths of numerical and experimental tests under different
normal stresses are shown in Fig. 8 which shows that the results of numerical models are in good agreement with the physical experiments.

49

triangular joint with the base angle of 20 and 0.5 to 3 MPa was applied
for the triangular joint with the base angle of 30.
To undertake the direct shear test, the specimen was placed in the
shear box and a small seating normal load was applied to the sample.
Then, the shear actuator and normal LVDTs were set to the zero position
and the normal load was increased continuously at a constant rate until
the desired normal stress was attained. Following the stabilisation of the
normal load on upper block, the shear load was applied (ASTM, 2008a;
Ulusay and Hudson, 2007). The shear load was applied continuously at a
constant displacement rate of 0.5 mm/min. After the shear stress
reached the peak shear strength, loading continued until the residual
shear strength was achieved.
Results of experimental direct shear tests on triangular joints with
base angles of 20 and 30 are presented in Figs. 9 and 10. For triangular
joints with the base angle of 20, at a normal stress of 1 MPa, when the
shear stress reached the peak strength, sliding took place along the joint
and the sample dilated with a constant dilation rate. At a normal stress
of 5 MPa, when the shear stress was exceeded around 70% of the peak
strength, a sign of yielding in the shear stress curve was observed
by departing from linear elastic behaviour. The rate of shear stress increment decreased and was followed by the asperity shearing off. As a consequence of the asperity shearing off, the shear stress reduced to
residual strength. For the triangular joints with the base angle of 30,
at the normal stress of 0.5 MPa, the failure mechanism was sliding
along the joint. At the normal stress of 3 MPa, the asperity was sheared
off. Due to the steep base angle of this asperity, signs of yielding were
observed very close to the peak shear strength and the asperity failed
in a very brittle mode. Results of the experiments show that the shear
failure is signicantly controlled by the asperity base angle and the applied normal stress, as also observed in the experiments of Guo and Qi
(2015).

6. Experimental direct shear tests on triangular joints

7. Numerical direct shear tests on triangular joints

Experimental direct shear tests were undertaken on triangular joints

with the base angles of 20 and 30. Direct shear tests were carried out
under different normal stresses to investigate the shearing mechanisms
and asperity degradation of these two sets of joints at different normal
stresses. In order to observe different modes of asperity failure for
both joint proles, the normal stress of 1 to 5 MPa was applied for the

Numerical specimens with similar geometry to the experimental

specimens were generated and direct shear tests under different normal
stresses were carried out. Results of numerical direct shear tests on
triangular joints with the base angle of 20 are illustrated in Fig. 11. At
the normal stress of 1 MPa, sliding was the failure mechanism and the
sample slid along the joint with a constant dilation rate. At the normal

Fig. 16. Shearing mechanisms of experimental direct shear tests under different normal stresses on triangular joints with the base angle of 30.

50

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

Fig. 17. Shearing mechanisms of experimental direct shear tests under different normal stresses on triangular joints with the base angle of 30.

stress of 5 MPa, the asperity sheared off. In the numerical model, similar
to the experimental test, sign of yielding was observed close to the shear
stress of 5 MPa and followed by a gradual decrease of shear stress

Fig. 18. Comparison between the peak shear strengths of experimental tests, numerical
and analytical models for triangular joint with the base angle of 30.

increment and brittle failure of the asperity. After the peak shear
strength, the shear stress reduced to residual strength. Unlike the experimental test, in the numerical tests at the residual stage the shear stress
uctuated. In the experimental tests after the asperity is sheared off, the
sheared surface is degraded during the shearing displacement whereas
in the numerical model when the stress at the asperity exceeds the
strength of the asperity, the bonds between particles break and shearing
takes place along the broken surface between the particles. Broken particles are stiff and these particles cannot be degraded during the shear
displacement. This leads to uctuation of the shear stress at the residual
stage. Results of numerical direct shear tests on the triangular joint with
the base angle of 30 are shown in Fig. 12. The shearing mechanisms of
sliding and asperity shearing off were observed at the normal stresses of
0.5 MPa and 3 MPa, respectively. At the normal stress of 3 MPa, sign of
yielding in the shear stress was observed close to the peak strength
and followed by a brittle failure of the asperity.
Most of the studies on the shear behaviour of rock joints have been
focused on the peak shear strength and the pre-peak behaviour of
rock joints. However, the residual shear strength and the post-peak behaviour of rock joints can be of great importance for long term stability
of rock structures. Results of this study show that PFC has the intrinsic
ability to reproduce the post peak behaviour. The coefcient of friction
of broken particles has a controlling effect on the residual strength of
the joint. However, as the broken particles are stiff and cannot be degraded during the shearing process at the residual stage, more studies
are required to solve the problem of shear stress uctuation at the residual stage.

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

8. Comparison of numerical and experimental shear tests on

triangular joints
In order to investigate the ability of the suggested numerical model to
reproduce the shear behaviour of rock joints, results of numerical and experimental methods were compared. The shearing failure modes of experimental and numerical shear tests under different normal stresses
for triangular joint with the base angle of 20 are shown in Figs. 13 and
14, respectively. As previously explained, in the experimental and numerical tests the upper block of the shear box was horizontally restrained and
the lower block moved in the horizontal direction. The direction of the
horizontal movement of the lower blocks is shown by red arrows. Three
shearing mechanisms of sliding, surface wear and asperity shearing off
were observed in these experiments.
At the normal stress of 1 MPa, the shearing mechanism for both experimental and numerical tests was sliding. As shown in Fig. 13 (n =
1MPa), only minor asperity degradation took place at the joint surface
but the shear behaviour was controlled by sliding along the asperity.
Similar behaviour was observed in the numerical test and only a few
bonds along the surface of asperity were broken (Fig. 14 (n =
1MPa)). In the numerical models, tensile and shear cracks are shown
in red and blue colours, respectively. At the normal stress of 3 MPa,
the observed shearing mechanism in experimental test was asperity
surface wear. In the numerical model, similar to the experimental test,
the surface of the joint was degraded and most of the bonds at the surface of the asperity were broken under tension, but due to low normal
stress, these tensile cracks were not propagated inside the intact material and the shearing mode was the asperity surface wear (Figs. 13 and
14 (n = 3MPa)). At the normal stress of 5 MPa in both numerical and
experimental tests, the asperity was sheared off. The propagation of
tensile cracks inside the asperity was followed by shearing off and
crushing the asperity (Figs. 13 and 14 (n = 5MPa)).
A comparison between the peak shear strength of numerical and experimental tests for the triangular joints with the base angle of 20 is
presented in Fig. 15. These results are also compared against the sliding
model of the Patton (Patton, 1966), and Ladanyi and Archambault
(Ladanyi and Archambault, 1969, 1980) models which are the most
well-known analytical models for triangular joints and these models
were originally developed on triangular joint proles. Due to difculty
in determination of the joint cohesion in the Patton model, as explained
in Section 1, the shearing part of the model cannot be captured and only
sliding part of this model was presented in Fig. 15. The results show that
the peak shear strengths of numerical models are in a good agreement
with the physical experiments. At the normal stresses of 1 MPa and
3 MPa, there is a good agreement between the peak shear strength of
numerical and experimental tests with the Patton sliding model. This
shows that in spite of surface wear of the asperity at the normal stress
of 3 MPa, the peak shear stress was controlled by the slope of asperity
and sliding along the joint. At the normal stress of 5 MPa, the asperity
was sheared off and the Patton sliding model cannot predict the peak
shear strength. The Ladanyi and Archambault model also overestimated
the peak shear strength at this normal stress.
The shearing mechanisms of triangular joints with the base angle of
30 for experimental and numerical shear tests are shown in Figs. 16
and 17, respectively. Normal stress was varied from 0.5 MPa to 3 MPa
and three shearing mechanisms of sliding, asperity surface wear and
shearing off were observed in both numerical and experimental tests.
At the normal stress of 0.5 MPa, the shearing mechanism was sliding.
At the normal stress of 1 MPa, in both experimental and numerical
tests, it was observed that tensile cracks were generated and propagated inside the asperity but these cracks terminated inside the intact material and the shearing mechanism was asperity surface
wear. At the normal stress of 3 MPa in both experimental and numerical tests, the asperity shearing off took place in which tensile
cracks propagated in the asperity followed by brittle failure and
shearing off the asperity.

51

A comparison between the peak shear strength of experimental, numerical and analytical models is presented in Fig. 18. The peak shear
strengths of numerical models are in good agreement with experimental
tests. At the normal stresses of 0.5 and 1 MPa, the shear behaviour was
controlled by sliding along the joint and there is a good agreement between the shear strength of experimental and numerical tests with the
Patton sliding model and Ladanyi and Archambault model. At the normal
stress of 3 MPa, the Ladanyi and Archambault model overestimated the
peak shear strength. This overestimation can be related to brittle failure
mode of the asperity in experimental and numerical tests. In the Ladanyi
and Archambault model, it is assumed that as the normal stress increases,
the asperity degrades gradually and parameters of dilation rate v_ and the
shear area ratio as change gradually which can result in overestimation of
the peak shear strength for this type of material.
The results presented in this study clearly show that the shear box
genesis approach can reproduce the shear behaviour of rock joints
well and also has the ability to investigate the asperity degradation
and shearing mechanisms of rock joints. This approach provides a useful
tool to predict the shear behaviour of rock joints without a need to a
quantitative value for estimation of the joint roughness and also study
the shear behaviour of large scale joints in which the current knowledge
is very limited (Bahaaddini et al., 2014).
9. Conclusion
One of the main difculties in the estimation of the shear behaviour
of rough rock joints is related to unknown mechanism of asperity degradation during the shearing process. Recent development in numerical
modelling using Particle ow code has shown that this code has the
ability to trace the crack initiation and propagation inside the rock asperities and intact materials. In this paper, the validity of this numerical
approach in reproducing the shear mechanisms and asperity degradation of rock joints was investigated by undertaking experimental and
numerical direct shear tests under different loading regimes. Synthetic
triangular joints with base angles of 20 and 30 were prepared in the
laboratory and experimental direct shear tests under different normal
stresses were carried out. Three shearing mechanisms of sliding, asperity surface wear and asperity shearing off were observed in both experimental and numerical tests. The shearing mechanism is controlled by
the applied normal stress and the asperity base angle. For gentle slope
asperities or under low normal stress, when the shear stress reaches
the peak value, sliding takes place along the asperities while for steep
asperities or under high normal stress, tensile cracks propagate inside
the asperities which led to shearing off and crushing the asperities. It
was found that the initiation and propagation of tensile cracks inside
the asperities controls the asperity shearing off process.
The results of this study clearly show that PFC has the ability to reproduce the shear behaviour and shearing mechanisms of rock joints
under different loading regimes. Unlike the empirical, analytical and
constitutive numerical models, in this numerical approach the real geometry can be explicitly simulated and there is no need for a quantitative value for description of the joint roughness. Furthermore, as the
previous studies have been mostly focused on the pre-peak and the
peak strength, this approach can provide a useful and scientic tool to
study the post peak behaviour of rock joints. Results of this study
show that this numerical approach has the intrinsic ability to reproduce
the post-peak behaviour of rock joints. However, further studies are required to investigate the process of broken asperities degradation at the
residual stage and considering the relation between asperity degradation and the residual strength which is recommended for future studies.
Acknowledgement
The authors would like to express their gratitude to Dr. Glenn
Sharrock, Dr. David Potyondy, Dr. Mathew Pierce and Dr. Xavier Garcia

52

M. Bahaaddini et al. / Engineering Geology 204 (2016) 4152

from Itasca Consulting group for their useful suggestions and technical
help.
References
Asadi, M., Rasouli, V., Barla, G., 2012. A bonded particle model simulation of shear strength
and asperity degradation for rough rock fractures. Rock Mechanics and Rock Engineering 45, 649675. http://dx.doi.org/10.1007/s00603-012-0231-4.
Asadi, M.S., Rasouli, V., Barla, G., 2013. A laboratory shear cell used for simulation of shear
strength and asperity degradation of rough rock fractures. Rock Mechanics and Rock
Engineering 46, 683699. http://dx.doi.org/10.1007/s00603-012-0322-2.
Asadollahi, P., Tonon, F., 2010. Constitutive model for rock fractures: revisiting Barton's empirical model. Eng. Geol. 113, 1132. http://dx.doi.org/10.1016/j.enggeo.2010.01.007.
ASTM, 2008a. Standard Test Method for Performing Laboratory Direct Shear Strength
Tests of Rock Specimens Under Constant Normal Force. ASTM D 3967ASTM International, United States.
ASTM, 2008b. Standard Test Method for Splitting Tensile Strength of Intact Rock Core
Specimens. ASTM D 3967ASTM International, United States.
Bahaaddini, M., 2014. Numerical Study of the Mechanical Behaviour of Rock Joints and
Non-persistent Jointed Rock Masses (PhD) UNSW Australia.
Bahaaddini, M., Hagan, P., Mitra, R., Hebblewhite, B.K., 2013a. Numerical investigation of
asperity degradation in the direct shear test of rock joints. Eurock 2013 Conference,
Wroclaw, Poland.
Bahaaddini, M., Hagan, P.C., Mitra, R., Hebblewhite, B.K., 2014. Scale effect on the shear
behaviour of rock joints based on a numerical study. Eng. Geol. 181, 212223.
http://dx.doi.org/10.1016/j.enggeo.2014.07.018.
Bahaaddini, M., Hagan, P.C., Mitra, R., Hebblewhite, B.K., 2015. Parametric study of smooth
joint parameters on the shear behaviour of rock joints. Rock Mechanics and Rock Engineering 48, 923940. http://dx.doi.org/10.1007/s00603-014-0641-6.
Bahaaddini, M., Hagan, P.C., Mitra, R., Hebblewhite, B.K., 2016. Numerical study of the mechanical behaviour of non-persistent jointed rock masses. Int. J. Geomech. 16,
04015035.
Bahaaddini, M., Sharrock, G., Hebblewhite, B.K., 2013b. Numerical direct shear tests to
model the shear behaviour of rock joints. Comput. Geotech. 51, 101115. http://dx.
doi.org/10.1016/j.compgeo.2013.02.003.
Bahaaddini, M., Sharrock, G., Hebblewhite, B.K., 2013c. Numerical investigation of the effect of joint geometrical parameters on the mechanical properties of a non-persistent
jointed rock mass under uniaxial compression. Comput. Geotech. 49, 206225. http://
dx.doi.org/10.1016/j.compgeo.2012.10.012.
Bandis, S.C., Lumsden, A.C., Barton, N.R., 1983. Fundamentals of rock joint deformation.
Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 20, 249268. http://dx.doi.org/10.1016/
0148-9062(83)90595-8.
Barton, N., 1971. A relationship between joint roughness and joint shear strength. International Symposium on Rock Mechanics and Rock Fracture, Nancy, France, pp. 18.
Barton, N., 1973. Review of a new shear-strength criterion for rock joints. Eng. Geol. 7,
287332. http://dx.doi.org/10.1016/0013-7952(73)900136.
Barton, N., Choubey, V., 1977. The shear strength of rock joints in theory and practice.
Rock Mech. 10, 154. http://dx.doi.org/10.1007/bf01261801.
Cho, N., 2008. Discrete Element Modeling of Rock Pre-Peak Fracturing and Dilation (PhD)
University of Alberta.
Cundall, P.A., 2000. Numerical Experiments on Rough Joints in Shear Using a Bonded
Particle Model. Aspects of tectonic faultingSpringer, Berlin, pp. 19.
Fairhurst, C., 1964. On the validity of the Brazilian test for brittle materials. Int.
J. Rock Mech. Min. Sci. Geomech. Abstr. 1, 535546. http://dx.doi.org/10.1016/
0148-9062(64)90060-9.
GCTS Testing Systems. RDS-300 direct shear testing systems brochure. World Wide Web
Address: http://www.gcts.com/?s=prod_ver&p=products&ID=89, 2016. (accessed
03.02.2016).
Grasselli, G., 2001. Shear Strength of Rock Joints Based on Quantied Surface Description
(PhD) Ecole Polytechnique Fdrale de Lausanne.
Grasselli, G., Egger, P., 2003. Constitutive law for the shear strength of rock joints based on
three-dimensional surface parameters. Int. J. Rock Mech. Min. Sci. 40, 2540. http://
dx.doi.org/10.1016/S1365-1609(02)001016.

Grasselli, G., Wirth, J., Egger, P., 2002. Quantitative three-dimensional description of a
rough surface and parameter evolution with shearing. Int. J. Rock Mech. Min. Sci.
39, 789800. http://dx.doi.org/10.1016/S1365-1609(02)000709.
Guo, S., Qi, S., 2015. Numerical study on progressive failure of hard rock samples with an
unlled undulate joint. Eng. Geol. 193, 173182. http://dx.doi.org/10.1016/j.enggeo.
2015.04.023.
Hsieh, Y.-M., Li, H.-H., Huang, T.-H., Jeng, F.-S., 2008. Interpretations on how the macroscopic
mechanical behavior of sandstone affected by microscopic propertiesrevealed by
bonded-particle model. Eng. Geol. 99, 110. http://dx.doi.org/10.1016/j.enggeo.2008.
01.017.
Indraratna, B., Haque, A., 2000. Experimental and numerical modeling of shear behaviour
of rock joints. GeoEng 2000, An International Conference on Geotechnical & Geological
Engineering, Pennsylvania, USA.
Itasca Consulting Group Inc., 2008. PFC2D manual, version 4.0, Minneapolis, Minnesota.
Kodikara, J.K., 1989. Shear Behaviour of Rock-concrete Joints and Side Resistance of Piles
in Weak Rock (PhD) Monash University.
Kulatilake, P.H.S.W., Shou, G., Huang, T.H., Morgan, R.M., 1995. New peak shear strength
criteria for anisotropic rock joints. Int. J. Rock Mech. Min. Sci. Geomech. Abstr. 32,
673697. http://dx.doi.org/10.1016/0148-9062(95)000229.
Kusumi, H., Matsuoka, T., Ashida, Y., Tatsumi, S., 2005. Simulation analysis of shear behavior of rock joint by distinct element method. Eurock 2005 - Impact of Human Activity
on Geological Environment. Taylor and Francis, London, pp. 281286.
Ladanyi, B., Archambault, G., 1969. Simulation of shear behavior of a jointed rock mass.
The 11th US Rock Mechanics Symposium (USRMS), Berkeley, CA, pp. 105125.
Ladanyi, B., Archambault, G., 1980. Direct and indirect determination of shear
strength of rock mass. Preprint number 8025 AIME Annual Meeting, Las Vegas, Nevada,
pp. 116.
Mas Ivars, D., Pierce, M.E., Darcel, C., Reyes-Montes, J., Potyondy, D.O., Young, R.P., Cundall,
P.A., 2011. The synthetic rock mass approach for jointed rock mass modelling. Int.
J. Rock Mech. Min. Sci. 48, 219244. http://dx.doi.org/10.1016/j.ijrmms.2010.11.014.
Oh, J.M., 2005. Three Dimensional Numerical Modeling of Excavation in Rock with Dilatant
Joints (PhD) University of Illinois.
Oliveira, D., Indraratna, B., 2010. Comparison between models of rock discontinuity
strength and deformation. J. Geotech. Geoenviron. 136, 864874. http://dx.doi.org/
10.1061/(ASCE)GT.1943-5606.0000284.
Park, J.W., Song, J.J., 2009. Numerical simulation of a direct shear test on a rock joint using
a bonded-particle model. Int. J. Rock Mech. Min. Sci. 46, 13151328. http://dx.doi.org/
10.1016/j.ijrmms.2009.03.007.
Park, J.W., Song, J.J., 2013. Numerical method for the determination of contact areas of a
rock joint under normal and shear loads. Int. J. Rock Mech. Min. Sci. 58, 822.
http://dx.doi.org/10.1016/j.ijrmms.2012.10.001.
Patton, F.D., 1966. Multiple modes of shear failure in rock. 1st ISRM Congress, Lisbon,
Portugal, pp. 509515.
Pierce, M., Cundall, P., Potyondy, D., Mas, Ivars D., 2007. A synthetic rock mass model for
jointed rock. Rock Mechanics: Meeting Society's Challenges and Demands, 1st
CanadaU.S. Rock Mechanics Symposium. Taylor & Francis Group, Vancouver,
London.
Potyondy, D.O., Cundall, P.A., 2004. A bonded-particle model for rock. Int. J. Rock Mech.
Min. Sci. 41, 13291364. http://dx.doi.org/10.1016/j.ijrmms.2004.09.011.
Resende, R., Lamas, L.N., Lemos, J.V., Calada, R., 2010. Micromechanical modelling of
stress waves in rock and rock fractures. Rock Mech. Rock. Eng. 43, 741761. http://
dx.doi.org/10.1007/s00603-010-0098-1.
Seidel, J.P., Habereld, C.M., 1995. The application of energy principles to the determination of the sliding resistance of rock joints. Rock Mech. Rock. Eng. 28, 211226. http://
dx.doi.org/10.1007/bf01020227.
Singh, M., Rao, K.S., 2005. Empirical methods to estimate the strength of jointed rock
masses. Eng. Geol. 77, 127137.
Terzaghi, C., 1920. New facts about surface-friction. Phys. Rev. 16, 5461.
Ulusay, R., Hudson, J.A., 2007. The Complete ISRM Suggested Methods for Rock Characterization, Testing and Monitoring: 19742006. International Society of Rock Mechanics
(Compilation arranged by the ISRM Turkish National Group).
Vosniakos, K., 2007. Physical and Numerical Modelling of Shear Behaviour of Saw-toothed
Filled Rock Joint (PhD) University of Manchester.