You are on page 1of 25

Rock Mech Rock Eng

DOI 10.1007/s00603-012-0365-4

ORIGINAL PAPER

A Constitutive Model for Shear Behavior of Rock Joints Based


on Three-Dimensional Quantification of Joint Roughness
Jung-Wook Park Yong-Ki Lee
Jae-Joon Song Byung-Hee Choi

Received: 15 May 2012 / Accepted: 29 December 2012


Springer-Verlag Wien 2013

Abstract A new constitutive model to describe the shear 1 Introduction


behavior of rock joints under constant normal stiffness
(CNS) and constant normal load (CNL) conditions is pro- Joint roughness is the most important factor influencing the
posed. The model was developed using an empirical shear behavior of rock joints. Due to the undulating surface
approach based on the results of a total of 362 direct shear of such joints, their shear behavior is accompanied by
tests on tensile fractured rock joints and replicas of tensile complicated phenomena such as normal dilation, asperity
joints and on a new quantitative roughness parameter. This failure, and changes in contact area. Depending on the
parameter, the active roughness coefficient Cr, is derived circumstances, normal dilation, especially, can have an
from the features of the effective roughness mobilized at effect on the normal load conditions exerted on the joint. If
the contact areas during shearing. The model involves a a joint is adjacent to a free face, as in a rock slope,
shear strength criterion and the relations between stresses movement in the normal direction is permitted during
and displacements in the normal and shear directions, sliding, and the normal force acting on the joint is not
where the effects of the boundary conditions and joint changed by the normal displacement. On the other hand,
properties are considered by the shape indices Cd and Cf. joints proximal to underground excavations are confined by
The model can be used to predict the shear behavior under the surrounding rock mass. In such cases, the normal
CNS as well as CNL conditions. The shear behavior dilation of the joint leads to an increase in normal stress.
obtained from the experimental results is generally in good These different boundary conditions can be simulated in
agreement with that estimated by the proposed model, and the laboratory by maintaining constant normal load (CNL)
the effects of joint roughness, initial normal stress, and or constant normal stiffness (CNS), respectively.
normal stiffness are reasonably reflected in the model. In the last few decades, considerable effort has been
devoted to explaining the shear strength and behavior of
Keywords Rock joint  Constitutive model for shear joints in terms of CNL conditions. Following the intro-
behavior  Contact area  Joint roughness quantification  duction of Pattons bilinear model based on sawtooth joints
Constant normal load condition  Constant normal stiffness (Patton 1966), peak shear strength criteria were developed
condition by Ladany and Archambault (1970), Barton and Choubey
(1977), Kulatilake et al. (1995), and Grasselli and Egger
(2003). Several empirical and theoretical models have also
J.-W. Park  B.-H. Choi been proposed to explain the postpeak response and
Underground Space Research Team, Korea Institute
asperity degradation (Bandis et al. 1983; Crotty and
of Geoscience and Mineral Resources (KIGAM),
92 Gwahang-no, Yuseong-gu, Daejeon 305-350, Korea Wardle 1985; Plesha 1987; Lee et al. 2001).
On the other hand, investigations that use models to
Y.-K. Lee  J.-J. Song (&) explain the shear behavior of rock joints under CNS con-
Department of Energy Resources Engineering, Research
ditions are very limited; this may be due to the difficulty of
Institute of Energy and Resources, Seoul National University,
599 Gwanak-ro, Gwanak-gu, Seoul 151-742, Korea building a suitable apparatus to control normal loads. In
e-mail: songjj@snu.ac.kr earlier studies, the stiffness of the surrounding rock mass

123
J.-W. Park et al.

was simulated by inserting a spring assembly with a known coefficient (JRC, Barton and Choubey 1977), statistical
stiffness between a rock specimen and a normal loading parameters (Tse and Cruden 1979; Maerz et al. 1990), and
plate. Despite some successful applications reported in the the fractal dimension (Huang et al. 1992; Xie et al. 2001).
literature (Johnston et al. 1987; Indraratna and Haque Despite these efforts, description of joint roughness is a
1997), this method typically encounters problems related to problem of considerable interest that is still awaiting a
the replacement of the spring and to possible breakage of solution. In most analyses of the shear strength and
the test sample due to the relatively high normal stiffness behavior of a joint, its roughness is characterized by a
used. More recently, with the development of the servo- single value that remains constant regardless of the shear-
controlled shear machine, experimental studies on shear ing area or shearing stage. In fact, only certain fractions of
behavior under CNS conditions have increased. Johnston the joint are in contact during a shearing process, and every
et al. (1987) presented the idealized shear behavior of part of the joint surface makes a different contribution to
joints under CNS conditions using experiments on saw- the shear behavior. Therefore, from a mechanical point of
tooth rockconcrete joints. Since then, numerous experi- view, the surface roughness of a rock joint should be
mental studies of shear behavior under CNS conditions characterized on the basis of these aspects.
have been reported (Van Sint Jan 1990; Kodikara and In this paper, a constitutive model for the shear behavior
Johnston 1994; Ohnishi and Dharmaratne 1990; Indraratna of rock joints under CNL and CNS conditions is proposed.
et al. 1998; Seidel and Haberfield 2002a; Jiang et al. 2006). The model includes a criterion for peak shear strength and
Most such experimental studies used an artificially gener- describes the relationships between stresses and displace-
ated sawtooth-shaped joint, and the applied loading con- ments in the normal and shear directions. The developed
ditions were insufficient to fully understand the shear model is based on a newly proposed roughness parameter
behavior under CNS conditions. and on a series of direct shear tests of tensile fractured rock
Saeb and Amadei (1990) proposed a graphical method joints and replicas of tensile joints under CNL and CNS
for prediction of joint responses to different normal stiff- conditions. It is capable of representing the combined
nesses and to normal stress, and Skinas et al. (1990) pre- effects of the influencing parameters, such as the initial
sented a numerical method for calculation of dilation and normal stress, normal stiffness, joint strength, and joint
the corresponding normal stress using the BartonBandis roughness. Details of the procedures used to derive the
model (Barton et al. 1985). These two methods, however, roughness parameter from the characteristics of the contact
were developed based on behavior under CNL conditions. areas during the shearing process are presented, and the
Son et al. (2006) observed that the stress path (the varia- experimental results are summarized and discussed.
tions of shear stress against normal stress) of CNS condi-
tions lay under the peak failure envelopes determined from
CNL conditions. Similar observations have already been 2 Description of Experiments
reported by Indraratna et al. (1998). From the comparison
between the friction coefficient curve measured in CNS Direct shear tests of tensile fractured rock joints and rep-
tests and that predicted by the BartonBandis model, Son licas of tensile joints were carried out under constant nor-
et al. concluded that the BartonBandis model tended to mal load (CNL) and constant normal stiffness (CNS)
overestimate the friction coefficient of CNS tests; this conditions, to investigate the effects of normal load con-
suggests the necessity for a reliable model to describe the ditions, joint properties, and roughness mobilization
shear behavior of rock joints under CNS conditions. Seidel according to shear direction.
and Haberfield (2002b) proposed a model for shear Rock blocks of Hwangdeung granite, Geumsan lime-
behavior under CNS conditions; however, in this model, the stone, and Onyang gneiss from Korea were prepared, and a
process used to determine the parameters was very com- tensile fracture (150 mm 9 118 mm) was generated at the
plex. Two problems associated with all the models cited center of each block. To make replicas of tensile fractured
above are that insufficient experimental evidence was pre- joints, rectangular aluminum molds (143 mm 9 116 mm)
sented and that the complicated behavior observed under were manufactured from five granite joints, one limestone
CNS conditions was immoderately simplified from a theo- joint, and four gneiss joints. Two molds of gneiss joints
retical viewpoint. In addition, there have been few studies to were then cut into circles with 105 mm diameter, as shown
date that predict postpeak behavior under CNS conditions. in Fig. 1. The replicas of the rectangular molds were used
For a constitutive model of the shear behavior of rock to examine the influences of normal load conditions and
joints to be realistic, proper quantification of joint rough- joint properties, while the replicas of the circular molds
ness is a priority, regardless of the boundary conditions. were used to investigate the effects of shear direction. In
Several methods have been proposed to quantify joint the case of gneiss, there was well-developed schistosity
roughness; these use indices such as the joint roughness structure. The tensile fracture was induced by applying a

123
A Constitutive Model for Shear Behavior of Rock

line load on each block in a direction perpendicular to the by visual comparison with standard profiles was not con-
schistosity structure such that the anisotropy of the joint sidered due to its subjective nature.
roughness with respect to shear direction could be maxi- Rearrangement of the shear strength criteria suggested
mized. In Fig. 1, line AB indicates the direction parallel to by Barton (1973) results in the following equation for the
the schistosity, and the shear direction angle a refers to the back-calculated JRC:
clockwise angle of line AB from the shear direction indi- tan1 spk =rn  /b
cated by the arrow. JRC ; 1
log10 JCS=rn
Each type of circular replica was tested in 12 different
shear directions at 30 intervals. The specimens are des- where spk is the peak shear strength, rn is the normal stress,
ignated according to rock type and shape: GR, LM, and GN JRC is the joint roughness coefficient, JCS is the joint wall
indicate granite, limestone, and gneiss tensile joints, compressive strength, and b is the basic friction angle. In
respectively, and Rc and Rr indicate circular and rectan- this procedure, the JCS was assumed to be equal to the
gular replicas, respectively. uniaxial compressive strength of the specimen, because all
To minimize possible errors and to enhance the repro- the tensile fractures were fresh without weathering (Barton
ducibility of the specimens, a complete product of high- and Choubey 1977).
strength grout consisting of cement, sand, plaster, and The method to relate the coefficient Z2 to the JRC has
additives mixed in weight ratio of 32:50:15:3 was used as been widely used in the literature to overcome the sub-
the material for all replicas. To diversify the properties of jectivity of standard profiles, although this method was not
the replica specimens, four different weight ratios of the suggested by Barton and his coauthors. We also evaluated
product to water (100:14, 100:15, 100:17, and 100:19) the JRC of each specimen using the correlations between
were used. All the specimens were cured for 4 days at the JRC and the coefficient Z2 (Eq. 2), which had been
room temperature. suggested by Yu and Vayssade (1991), where the coeffi-
Table 1 lists the physical and mechanical properties of cient Z2 is expressed as Eq. 3.
the specimens used in this study. The basic friction angle JRC 61:79Z2  3:47; 2
was determined from direct shear tests on perfectly flat " #1=2
surfaces in which five different normal stresses in the range 1XN 1
zi1  zi 2
of 0.22.0 MPa were applied. In the table, the ranges of Z2 ; 3
L i1 xi1  xi
JRC according to specimen type are also included. In this
study, the JRC value of each specimen was obtained by two where L is the length of the profile, (xi, zi) is the ith datum,
methods: back-calculation via the results of direct shear and N is the total number of data. In this procedure, the
testing, and calculation from correlations between the JRC values of the coefficient Z2 were calculated using data
of ten standard profiles (Barton and Choubey 1977) and the at 0.5-mm intervals drawn from the raw data measured at
statistical parameter Z2. The method of evaluating the JRC 0.1-mm intervals, because Eq. 2 was derived based on
profiles obtained at 0.5-mm intervals. The ranges of JRC
provided in Table 1 are the length-weighted average values
of all the roughness.
For the direct shear tests, a servo-controlled apparatus
that was designed and built to test the shear behavior of
joints under constant normal load (CNL) and constant
normal stiffness (CNS) conditions was used. The body of
the apparatus comprises a main frame, a control program
for data acquisition and operation of the instrumentation
system, a hydraulic power unit, and an analog-to-digital
converter. A number of shear tests have been conducted
using this apparatus and have indicated good performance
(Rim et al. 2005; Son et al. 2006).
In this study, a total of 362 laboratory tests were carried
out on 52 granite joints, 28 limestone joints, 25 gneiss
joints, and 115 replicas. Shear and normal forces were
continuously applied and monitored until the total shear
Fig. 1 Circular aluminum molds for specimens Rc-01 and Rc-02;
displacement reached approximately 10 % of the specimen
the shear direction angle a refers to the angle between line AB
parallel to the schistosity and the shear direction indicated by the length: 10 mm for circular specimens and 15 mm for
arrow rectangular specimens.

123
J.-W. Park et al.

Table 1 Physical and mechanical properties of the specimens


Property Granite Limestone Gneiss Replica (Rr & Rc)
(GR) (LM) (GN)
100:14 100:15 100:17 100:19

Uniaxial compressive strength (MPa) 184.08 79.16 134.30 83.91 74.03 65.33 54.19
Youngs modulus (GPa) 55.43 51.28 54.37 31.10
Poissons ratio 0.28 0.30 0.28 0.24
Brazilian tensile strength (MPa) 8.38 5.24 11.26 5.41 5.04 3.92 3.39
P-wave velocity (m/s) 3,530 5,113 5,209 3,210
S-wave velocity (m/s) 1,949 2,522 2,067 1,800
Porosity (%) 0.63 0.34 0.69 9.15
Dry specific gravity 2.65 2.75 2.74 2.20
Basic friction angle () 34.29 31.41 34.19 31.03 31.94 32.21 31.45
JRC back-calculated from test results 5.9715.49 5.5817.82 5.0613.59 10.0315.79 3.5815.79 10.0315.79 10.0315.79
JRC calculated using Z2 10.1614.71 12.1216.61 6.1210.14 10.17 5.2715.2 10.17 10.17

Normal stresses in the range of 0.12.0 MPa were were damaged. To minimize possible errors in data anal-
applied in the CNL tests; stresses of this magnitude cor- ysis, the surface of the specimen was measured and the
respond roughly to load conditions at overburden depth roughness was evaluated repeatedly before and after each
of 1080 m. It is generally considered that, in most rock shear test. The number of CNS tests conducted in this study
engineering (rock slope) problems, the normal stress acting was 117. Most of the CNS tests were designed to examine
on rock joints lies within this range (Barton and Choubey the effects of initial normal stress and normal stiffness on
1977). The normal stiffness applied in the CNS tests ranged the joints shear behavior and were therefore carried out on
from 0.1 to 2.5 MPa/mm. The normal stiffness Kn of the the replicas, allowing us to perform several tests on joints
CNS test depicts the increment in the normal stresses on with the same morphology and properties.
the interface between the rock mass and the joint caused by
normal dilation. It can be calculated by expanding the
infinite cylinder theory, as shown in the following equation 3 Quantification of Joint Roughness
(Johnston et al. 1987; Jiang et al. 2004):
E 3.1 Micro-Slope Angle and Contact Areas
Kn ; 4
1 mr
During the shearing process, only certain areas of the joint
where E and m are the deformation modulus and Poissons surface come into contact and contribute to the shear
ratio of the surrounding rock mass, respectively, and r is behavior. If we imagine a sawtooth-shaped joint with dif-
the influenced radius. ferent slopes, we can easily infer that some of the faces
The number of CNL tests conducted in this study was leaning towards the shear direction are in contact, and that
245; some of these were repeated shear tests on identical of these, the faces with the steeper slopes are more likely to
rock specimens with three or four different normal stresses; come into contact. These facts have been reported in pre-
25 of 52 granite specimens, 8 of 28 limestone specimens, vious experimental observations (Barton 1973; Bandis
and 8 of 21 gneiss specimens were used in the repeated et al. 1981; Haberfield and Johnston 1994; Yeo et al. 1998;
loading tests. Each rock specimen was tested under a low Yang and Chiang 2000). Based on these considerations,
normal stress until its residual state was reached; it was joint roughness should be qualified based on the charac-
then repositioned to the initial position and tested again teristics of the contact areas rather than the whole surface.
under a higher stress. In terms of the MohrCoulomb Park and Song (2013) studied the relationship between
parameters, such repeated testing may cause a decrease in the location of the contact areas and the steepness of
the friction angle and increase in the apparent cohesion. localized asperities. In their procedure, a joint surface was
However, it was assumed that the asperity damage was of modeled as a group of triangular planes, and the contact
relatively minor significance because the stress levels condition of each plane was examined by calculating the
applied in this test were much lower than the joint wall relative displacements of both surfaces from their initial
strength. Observations of the surface conditions after the locations. The steepness of a joint surface, including its
shear test showed that only small fractions of the surface contact areas, was represented using the concept of micro-

123
A Constitutive Model for Shear Behavior of Rock

slope angle, an extension of the apparent dip angle angle bE between ~ eE and ~
S is less than 90 for an active
proposed by Grasselli and Egger (2003). element and more than 90 for an inactive one, ranging
In the present study, the characteristics of the roughness between 0 and 180. Then, the apparent slope angle of
mobilization at contact areas were examined by adopting element E (hE) with respect to the shear direction can be
Park and Songs approach, and a three-dimensional calculated as follows:
parameter intended to qualify the joint roughness was
derived based on the investigations. Joint morphology was hE j90  bE j: 5
measured at 0.1-mm intervals using a three-dimensional To distinguish between active and inactive elements, the
laser profiler with precision of 5 lm in the z direction. micro-slope angle h*E is defined as follows:
The joint surface was reconstructed from the data as a
hE 90  bE : 6
group of triangular elements. In this procedure, four
neighboring points defined two triangular planes. Figure 2 Note that the micro-slope angle h*E lies within the range
shows a schematic diagram that explains the definition of from -90 to ?90. The plus and minus signs denote
the micro-slope angle of each element. Each element has a active and inactive elements, respectively. The magnitude
single-valued true dip angle in space, but the apparent dip (absolute value) and sign of h*E indicate the steepness and
varies with the shear direction; For example, element E in orientation of a surface element E with respect to the shear
Fig. 2a has a true dip angle of h when it is sheared in a direction.
direction parallel to the x-axis, while it is apparently not For the circular specimen Rc-01, the micro-slope angles
inclined when sheared in a direction parallel to the y-axis. of the joint surface were calculated according to the
For the same reason, we obtain a horizontal line with zero shearing direction; the contact areas at the peak stages were
inclination when we measure the profile of element E in estimated using the method suggested by Park and Song
the y direction. (2013). As mentioned in Sect. 2, the circular specimens
If the relative shear direction vector (S ~) points in the were tested in 12 different shear directions (Fig. 1). Fig-
direction of movement of one block toward the fixed ure 3 shows contour plots of the micro-slope angles when
opposite block, and the shear-direction plane indicates the the shear-direction angle a is 30, 90, and 270. The
plane that includes the vector and the z-axis, the apparent arrows and dotted lines in the circles beside each plot
dip angle of an element can be obtained from the line of indicate the shear direction (direction of movement) and
intersection between the element and the shear-direction the direction parallel to the schistosity. Active and inactive
plane; For example, on the xz shear-direction plane in elements are displayed in warm and cool colors, respec-
Fig. 2b, elements A and B have apparent angles of hA and tively, with steeper elements indicated by brighter colors.
hB, respectively. The elements dipping toward the shear The regions near the borders of warm and cool colors
direction (e.g., A) may come into contact and actively indicate elements for which the magnitude of the micro-
contribute to the shear behavior of the joint at the initial slope angle is very low; such low angles create hills and
stage, and the elements with the opposite inclination (e.g., valleys in the spatial shear-direction plane. In spite of
B) may be detached, causing voids in space. To reflect this the identical surface, the distribution of the micro-slope
directionality, elements are classified into two groups: the angle varied with the shear direction angle a. The portion
first group contains the active elements, and the second of elements with large apparent slope angle increased as
group contains the inactive elements. If the vector ~ eE the angle between the shearing and schistosity directions
indicates the projection of the outward normal vector of increased. As observed in the reverse images in Fig. 3b, c,
triangular element E onto the shear-direction plane, the in the opposite shear direction the sign of the micro-slope

Fig. 2 Definition of active and


inactive micro-slope angles
(after Park and Song 2013):
a apparent dip angle and
b micro-slope angle

123
J.-W. Park et al.

angle is reversed, while the magnitude remains the same. respect to the shear direction, which may offer a viable
Figure 4 shows the contact areas estimated at peak shear alternative for quantification of the effective roughness of
strength; white depicts noncontact areas. Regardless of the rock joints.
shear direction, the contact areas were almost entirely Figure 7a shows the areal distribution of the active ele-
made up of active elements, and the locations of the contact ments of specimen Rc-01 according to their shear direc-
areas were remarkably consistent with those of the active tions. The active elements were sorted according to their
zones. micro-slope angle at 2 intervals, and the areas of the ele-
Figure 5 shows the areal distributions of the micro-slope ments within each interval were summed. In the figure, the
angle on specimen Rc-01 when the shear direction angle sum of the columns heights represents the total area of the
was 0, 30, 60, and 90. The total elements covering the active elements, 5,438.2 mm2. Most of the active elements
surface were sorted according to their micro-slope angles at have a micro-slope angle that is \50. As the micro-slope
2 intervals, and the areas of the elements within each increases, the portion of the corresponding area decreases.
interval were summed. The total surface area was If we proportionally adjust the height of each column in
9,048 mm2, and the projected area on the xy plane was Fig. 7a so that the total area is 1.0, Fig. 7b is obtained.
8,495 mm2 (radius = 52 mm). The distribution of the Assuming the active micro-slope angle as a random vari-
micro-slope angle shows common characteristics regard- able x, Fig. 7b indicates the probability density function
less of the shear direction and joint sample; the micro-slope (PDF) p(x) that describes the relative likelihood of the
angles exhibited bell-shaped areal distributions which were active elements exhibiting a particular active micro-slope
nearly symmetric around approximately 0, although in the angle. In other words, the area of each column in Fig. 7b is
examples of Fig. 5 the active elements occupied slightly the geometrical probability P(x) that an active element has
more area than the inactive ones. With increasing a, the a micro-slope angle within the corresponding range:
area of the elements with steeper apparent slopes increased P(x) = p(x)dx, where dx = 2.
and that of those with low apparent slope decreased. To model the probability distribution of the active
Figure 6 presents the areal distribution of micro-slope micro-slope angle as a continuous function, the following
angle for the element determined as contact areas at the PDF, which uses the parameter Cr, was established by trial
peak stage when a is 30. The areal ratio of contact surface and error:
to total surface was 57.61 %. As Fig. 4 illustrates, the 2 2 2
active elements made a major contribution to the contact px p ex =Cr x  0; Cr [ 0; 7
Cr p
areas at the peak stage.
where
3.2 Active Roughness Coefficient Cr Z1
2 2 2
p ex =Cr dx 1:0: 8
As discussed above, contact areas according to shear Cr p
0
direction are closely correlated with micro-slope angle. In
particular, the distribution of the active micro-slope angle Most of the characteristics of the proposed probability
can capture the features of roughness mobilization with distribution can be explained using only the parameter Cr.

Fig. 3 Contour plots of the micro-slope angles of specimen Rc-01 cool colors, respectively. The arrow and dotted line in the circle
according to shear direction angle a: a a = 30, b a = 90, and beside each plot indicate the relative shear direction and schistosity
c a = 270; active and inactive elements are displayed in warm and direction

123
A Constitutive Model for Shear Behavior of Rock

Fig. 4 Estimated contact areas and their micro-slope angles at the dotted line in the circle beside each plot indicate the relative shear
peak stage according to the shear direction: a a = 30, b a = 90, direction and schistosity direction
and c a = 270; white denotes noncontact areas. The arrow and

The variable x can be any real number C0, while the


active micro-slope angle ranges from 0 to 90. Thus, the
applicability of this distribution model to the analysis of
the active micro-slope angle needs to be examined. If xc is
the x value when the cumulative probability is 99.99 %, it
can be expressed as Eq. 12 by resolving the condition that
the integration of Eq. 7 from 0 to xc is 0.9999.
s
 
1 1
xc C r ln  2:604Cr : 12
2 1  0:99992

In practice, xc can be regarded as the maximum of the


variable x. According to Eq. 12, if Cr is less than approx-
Fig. 5 Areal distributions of micro-slope angle according to shear imately 43.6, the probability P(x [ 90) is nearly zero, and
direction angle a: results obtained with specimen Rc-01 consequently the proposed PDF is applicable to the active
micro-slope angle.
The expectation E(x) and standard deviation r(x) of the All of the joint surfaces used in our experiments were
probability distribution are expressed as Eqs. 9 and 10, analyzed for active micro-slope angle distribution. In this
respectively. procedure, the parameter Cr giving the best fit was calcu-
Z1 lated through nonlinear regression of the joint data using
Cr Eq. 7. The maximum value of the parameter Cr was found
Ex px xdx p  0:564Cr ; 9
p to be 30.64, which suggests that the proposed distribution
0
model can be applied to general rock joints.
v
uZ1 Figure 8 shows the regression results for specimen Rc-02
u
u according to shear direction. In each plot, the best-fit curve is
rx t pxx2 dx  Ex2
presented with the parameter Cr and the coefficient of
0
q determination R2. The proposed PDF corresponds well to the
1=2  1=pCr2  0:426Cr 10 areal distribution of the active micro-slope angle of the rock
joint. As the shear direction angle a varies from 0 to 90, the
The x value at the inflection point of the PDF curve (x0) parameter Cr increases consistently from 13.34 to 24.85; this
can be calculated from the condition that the second suggests that the parameter Cr has the capability to charac-
derivative is zero. terize the effective roughness. Thus, we propose the param-
Cr eter Cr as a new coefficient for quantification of rock joint
x0 p  0:717Cr : 11 roughness and term it the active roughness coefficient.
2

123
J.-W. Park et al.

In all cases considered in this study, the measured PDFs the active roughness coefficient of a completely planar
of the active micro-slope angles fit well with Eq. 7, joint. However, we assume the active roughness coefficient
showing high values of R2 [ 0.95. Although the calculated of a planar joint to be zero in the following discussion.
Cr within the same rock type did not cover a wide range
due to the nature of the tensile fractured joint, the gneiss
joints showed markedly lower Cr than the granite and 4 Experimental Results and Discussion
limestone joints: 14.0419.62 for the gneiss, 20.9727.79
for the granite, 21.9230.64 for the limestone, and 4.1 Shear Behavior of Rock Joints Under CNL
12.5427.20 for the replicas. and CNS Conditions
In fact, the active roughness coefficient is only appli-
cable to a rough joint, and its value is always positive. Based on the experimental results obtained in this study and
Because this concept is based on statistical analysis of the those reported in the literature (Goodman 1974; Barton
active slope angle, it is technically impossible to evaluate et al. 1985; Johnston et al. 1987; Son et al. 2006), the shear
behavior of rough rock joints under CNL and CNS condi-
tions can be simplified as shown in Fig. 9. In the figure,
shear stress, normal stress, and the friction coefficient are
indicated by s, rn, and l, respectively, Ks is the shear
stiffness, spk and srs are the peak and residual shear stresses,
lpk and lrs are the peak and residual friction coefficients, dN
and dT are the normal and shear displacements, and dpk is
the shear displacement at the peak stage. The friction
coefficient equals the ratio of shear stress to normal stress.
Under CNL conditions, the shear stress increases line-
arly until the peak strength is passed, and it reaches a
constant value (ultimate or residual shear resistance).
Because the normal stress does not change during the test,
the friction coefficient curve resembles the shear stress
curve. Under CNS conditions, the shear behaviors exhib-
Fig. 6 Areal distributions of micro-slope angle of the whole surface
and the contact areas at peak stage: results obtained with specimen ited prior to the occurrence of the first peak stress are
Rc-01 when a = 30 nearly identical to those that occur under CNL conditions.

Fig. 7 Areal distribution of active micro-slope angle: results obtained with specimen Rc-01 when a = 30

123
A Constitutive Model for Shear Behavior of Rock

Fig. 8 Determination of the active roughness coefficient Cr from the probability distribution of active micro-slope angles according to shear
direction angle a: results obtained with specimen Rc-02 when a a = 0, b a = 30, c a = 60, and d a = 90

Under conditions of constant normal stiffness, once the displacement, the shear stress sometimes shows a second
normal dilation is initiated around the first peak, an peak stress due to high normal load, and the stress some-
increase in normal stress occurs (rn = rn0 ? KndN). The times increases persistently to the end of the test. In other
normal stress path resembles the normal dilation path cases, the peak shear stress cannot be observed due to great
except that the initial point is rn0, not zero. Because the increase in normal stress at the peak stage. Son et al.
increase in normal stress leads to successive increases in (2006) categorized the shear behavior of joints under CNS
shear stress, the shear stress exhibits a variety of curves conditions into three types (I, II, and III) based upon the
depending on theinitial normal stress (rn0) and normal shear stressshear displacement curves they observed. In
stiffness (Kn). Under conditions of continued shear their experiments, the second peak stress was not observed;

123
J.-W. Park et al.

Fig. 9 Simplified shear behavior of rough rock joints under a CNL and b CNS conditions

however, Johnston et al. (1987) included the second peak Therefore, it should be noted that the residual friction
stress in their idealized model of shear behavior under CNS coefficient lrs in the following discussion indicates precisely
conditions. Although the shear stress curve plotted in the ultimate friction coefficient.
Fig. 9b cannot represent all possible behaviors under CNS Table 2 lists a summary of the shear properties of the
conditions, it is possible to generalize the features of the joints tested in this study. The peak friction coefficient lpk
friction coefficient curves. In all the tests conducted in this ranged from 0.849 to 4.775; in 228 of 362 cases, it was
study, the curves showed obvious peak and residual stages, between 1.0 and 2.0. The effect of the initial normal stress
and in spite of the existence of different initial degrees of was clearly observed. When the initial normal stress was
normal stress and normal stiffness, the behavior of the less than 1.0 MPa, the rock specimens showed higher peak
friction coefficient after the peak was nearly consistent friction coefficients overall than replica specimens. The
with the curve plotted in Fig. 9b. residual friction coefficient (ultimate friction coefficient)
In this section, the experimental results obtained are lrs was determined by averaging the data with values
briefly presented and discussed. Because of the considerable between 80 and 100 % of the shear displacement. In most
volume of data, only a summary can be included here. The of the results we obtained, the friction coefficient over the
CNL and CNS conditions are not treated separately in the considered range showed little variation over this range of
discussion below. Here, the CNL condition is assumed to shear displacement. In 303 of 362 cases, the residual fric-
represent a CNS condition in which rn0 = rn and Kn = 0.0. tion coefficient was higher than the basic friction coeffi-
The peak and residual stages were determined based cient lb, where the basic friction coefficient was calculated
upon the friction coefficient curve because, as mentioned from the basic friction angle b (lb = tan b); the basic
above, the friction coefficient curves exhibited similar friction coefficients were 0.682 for GR, 0.611 for LM,
behaviors under CNL and CNS conditions. According to the 0.796 for GN, and 0.623 for replica specimens. The shear
International Society for Rock Mechanics (ISRM) suggested stiffness Ks was calculated by dividing the shear stress by
methods for determining shear strength (Brown 1981), it is the shear displacement over the linear portion of the shear
recommended to determine the residual shear strength when stress curve. In performing this calculation, the range of
the variation in shear stress is not more than 5 % over a shear values between 30 and 70 % of the peak strength (or the
displacement of 1.0 cm. However, it is difficult to attain the first peak) in the prepeak stage was taken. The replicas
true residual friction coefficient in laboratory shear testing showed a lower average stiffness with higher variations
unless the normal stress is very high, or unless repeated tests than the rock specimens, but it was difficult to demonstrate
are made (Barton 1973), even though the friction coefficient the effects of rock type on the shear stiffness. The shear
reaches a constant value (ultimate friction coefficient). displacement at the peak stage dpk tended to be affected by

123
A Constitutive Model for Shear Behavior of Rock

the initial normal stress. Under the same normal load When the same rn0 was applied, the shear stress paths in
conditions, the rock specimens showed peak coefficients at the prepeak stages were indistinguishable from one
larger displacements than the replica specimens. another. This is because the normal load conditions were
not affected by normal stiffness due to the restricted dila-
4.2 Effect of Normal Load Conditions on the Shear tion at this stage. However, after the first peak shear stress
Behavior of Rock Joints occurred, the shear stress exhibited many different paths
according to Kn. At low Kn, the stress showed a tendency to
For replica specimens Rr-01 to Rr-07, direct shear tests decrease after the first peak point, while it showed a ten-
under 723 different kinds of normal load conditions were dency to increase at high Kn. Although the parameters Kn
carried out to investigate the effects of initial normal stress and rn0 are not directly comparable due to the use of dif-
(rn0) and normal stiffness (Kn) on the shear behavior of ferent physical units, when rn0 was relatively larger than
rock joints. Because of the considerable volume of data Kn, the first peak and the stress softening tendency in the
obtained, only representative examples are presented here. postpeak stage were clearly observed. In Fig. 11a, the first
Figure 10 shows the experimental results obtained using peak shear stresses were within a similar range when
specimen Rr-06; it presents the shear stress and friction Kn = 0.0, 0.2, and 1.0 MPa/mm; when Kn = 2.0 MPa/
coefficient curves when rn0 = 0.1, 0.5, 1.0, and 1.5 MPa and mm, the shear stress increased dramatically and showed a
Kn = 0.1 MPa/mm. In all the shear stress curves, the first second peak stress. Similar tendencies related to the effect
peak stress was clearly observed; this may be ascribed to a of Kn on shear stress curves are illustrated in Fig. 11b.
small increment in normal stress around the peak stage due to Figure 12 shows the effect of the normal stiffness Kn on
a relatively low normal stiffness. At high rn0, a stress the friction coefficient. Figure 12a and b present the fric-
oscillation around the peak stage was observed; under these tion coefficient curves at the normal stiffnesses shown in
conditions, the shear stress increased to a certain value, and a Fig. 11a and b, respectively. For both specimens, the
sudden drop then repeatedly occurred. For rough rock joints, friction coefficient was considerably lower at higher Kn
this jerky behavior is known as stickslip; it is thought to than at low Kn, and the shear displacement at the peak was
result from instability of sliding due to interlocking and shortened. The reason for this is suggested by the shear
sudden failure of asperities (Byerlee 1970; Scholz and Eng- stressnormal stress history of the specimens.
elder 1976) and the difference between static and dynamic Figure 13 presents the shear and normal stress histories
friction (Awrejcewicz and Olejnik 2005). of the friction coefficient paths shown in Fig. 12a. The
With increasing rn0, the peak friction coefficient lpk small circles indicate the states of stresses at the peak stage
decreased, and the first peak stress was shifted to a larger and at shear displacements of 5, 10, and 15 mm. Consid-
shear displacement. The shear stiffness showed slight erably higher normal stress on the joint occurred at
variation with rn0. In choosing between the constant every shear stage when Kn = 2.0 MPa/mm than when
stiffness and the constant shear displacement models sug- Kn = 0.2 MPa/mm. When Kn = 2.0 MPa/mm, the normal
gested by Goodman (1976), the former seems more con- stress at the peak stage was already increased by the normal
sistent with the results we obtained regarding the linear dilation, in spite of the very small shear displacement
behavior of rock joints in the prepeak stage. As the shear (0.263 mm). This suggests that Kn can affect the peak
displacement reached approximately 9 mm, all the curves friction coefficient.
approached the residual state and showed similar behavior. Evidence for a decrease in lpk with increasing Kn was
Figure 11 shows the effect of the normal stiffness Kn on found in many of our experiments, as shown in Fig. 14, in
the shear stress. Panel (a) presents the experimental results which the values of the applied rn0 are included in
obtained with specimen Rr-01 at rn0 = 0.1 MPa and parentheses. The decrease in lpk was more pronounced at
Kn = 0.0, 0.2, 1.0, and 2.0 MPa/mm, while panel lower rn0 and higher Kn. Under this boundary condition,
(b) presents the results obtained with specimen Rr-06 at normal dilation begins to occur with a very small shear
rn0 = 0.5 MPa and Kn = 0.0, 0.1, 0.2, and 0.5 MPa/mm. displacement. This may cause an increment in normal

Table 2 Summary of the shear


Parameter Granite Limestone Gneiss Replica
properties of the specimens
(GR) (LM) (GN) (Rc & Rr)

Peak friction coefficient, lpk 1.0604.775 0.8494.583 0.9643.491 0.8313.301


Residual friction coefficient, lrs 0.4541.182 0.4441.283 0.6401.123 0.4221.074
Shear stiffness, Ks (MPa/mm) 0.2362.307 0.7472.400 0.2632.400 0.2683.126
Shear displacement at the peak stage, dpk (mm) 0.2693.687 0.3862.731 0.4042.099 0.1842.102

123
J.-W. Park et al.

Fig. 10 Effect of initial normal stress (rn0) on shear behavior: results obtained with specimen Rr-06

Fig. 11 Effect of normal stiffness (Kn) on shear stress: results obtained with specimens (a) Rr-01 and (b) Rr-06

stress and significant damage in asperities before devel- friction coefficient or peak friction angle provides a more
opment of the peak resistance. appropriate description of the shear resistance. In terms of
A decrease in lpk with increasing Kn can be observed in the peak friction coefficient, it can be concluded that the
the experimental results reported in the literature (Ohnishi ratio of shear resistance to normal load decreases with
and Dharmaratne 1990; Van Sint Jan 1990; Indraratna et al. increasing Kn as well as with increasing rn0.
1998; Olsson and Barton 2001; Son et al. 2006). Most Figure 15 shows the effects of initial normal stress rn0
studies, however, have paid little attention to the decrease in and normal stiffness Kn. In most of the experimental
peak friction coefficient; instead they assume that the first results, normal dilation was initiated around the first peak
peak shear stress represents the shear strength and conclude shear stress. With increasing rn0 and Kn, the maximum
that the mechanical behavior of rock joints under CNL and dilation angle and total amount of dilation were restricted.
CNS conditions differs only after the peak stage. However, The joints initially underwent little dilation or even con-
the shear resistance of rock joints under CNS conditions tracted with increasing normal loads. The point at which
cannot be fully described using only the first peak shear dilation was initiated was delayed by an increase in initial
stress. The first peak shear stress does not always represent normal stress, while it occurred earlier with an increase in
the maximum shear resistance. In such cases, the peak normal stiffness.

123
A Constitutive Model for Shear Behavior of Rock

Fig. 12 Effect of normal stiffness (Kn) on friction coefficient: results obtained with specimens (a) Rr-01 and (b) Rr-06

Fig. 13 Histories of normal and shear stresses under different normal Fig. 14 Effect of normal stiffness on peak friction coefficient; the
stiffnesses (Kn): results obtained with specimen Rr-01 values in parentheses are the initial normal stresses applied to the
respective experiments

4.3 Effect of Joint Wall Strength on the Shear Behavior


of Rock Joints Figure 16 shows the normal displacement and friction
coefficient as a function of the applied weight ratio when
To examine the effect of the joint wall strength on shear CNS tests were conducted at rn0 of 0.1 MPa and Kn of
behavior, a set of joint replicas with the same roughness 0.2 MPa/mm. With increasing joint strength (decreasing
and different properties were generated. The aluminum water ratio), the shear displacement at which normal dila-
mold for specimen Rr-07 was used, and four different tion was initiated and the peak friction coefficient was
weight ratios of the grout product to water (100:14, 100:15, developed was delayed. In addition, the maximum dilation
100:17, and 100:19) were applied. With increasing water angle, the final normal displacement, and the peak friction
ratio, the uniaxial compressive strength and tensile strength coefficient tended to increase. Observation of the sheared
decreased, while there were no noticeable variations in the surface of the joint specimen with weight ratio of 100:19
basic friction angle of the planar surface, as shown by the showed that its surface was significantly crushed and
data in Table 1. damaged.

123
J.-W. Park et al.

Fig. 15 Effects of initial normal stress (rn0) and normal stiffness (Kn) on normal displacement: results obtained with specimens Rr-06 and Rr-01

Fig. 16 Effect of joint wall strength on shear behavior: results obtained with specimen Rr-07

4.4 Effect of Joint Roughness on the Shear Behavior coefficient Cr increases with the angle a. With increasing
of Rock Joints Cr, the dilation angle at the peak stage and the final normal
displacement increased consistently, although the cases
As mentioned in Sects. 2 and 3, the circular replicas were with a = 60 and a = 90 showed similar dilation paths.
generated from gneiss joints to maximize the anisotropy of The peak friction coefficient and residual friction coeffi-
the roughness mobilization and the corresponding shear cient also tended to increase with Cr. It was difficult to find
behaviors according to the shear direction (Fig. 1). The the relationship between the joint roughness and the shear
effective roughness in a direction perpendicular to the stiffness Ks. The shear displacement at which normal
schistosity (a = 90 or a = 270) would be expected to be dilation was initiated and the peak friction coefficient was
much higher than that in a parallel direction (a = 0 or obtained was delayed as the roughness increased. Barton
a = 180). and Bandis (1990) also reported a positive correlation
Figure 17 shows the normal displacement and friction between JRC and the shear displacement at peak strength
coefficient curves depending on shear direction angle a in through fitting of 650 experimental results.
the range from 0 to 90. In the figure, the active roughness Overall, our experimental results indicated that, as the
coefficient Cr of the surface of specimen Rc-02 in each roughness of the joint increased, the friction coefficient
direction is also presented. Note that the active roughness tended to show a dramatic decrease after the peak stage.

123
A Constitutive Model for Shear Behavior of Rock

The joint roughness seemed to also affect the residual predicts the peak friction coefficient of freshly mated joints
friction coefficient; in most of the results we obtained, the under CNL and CNS conditions, according to the following
residual friction coefficient was higher than the contribu- equation:
tion by the basic friction angle.  C
B rt
lpk ACr eDKn =rn0 lb ; 13
rn0
5 Constitutive Model for the Shear Behavior in which A, B, C, and D are dimensionless parameters used
of Rock Joints to fit the experimental data.
In the model, four terms are used: the ratio of tensile
In this section, a new constitutive model for the shear strength to initial normal stress (rt/rn0), the active rough-
behavior of rock joints is presented. The model defines a ness coefficient (Cr), the friction coefficient mobilized by
criterion for the peak friction coefficient, and uses the the basic friction angle (lb = tan b), and the ratio of
relationships between stresses and displacements in normal normal stiffness to initial normal stress (Kn/rn0). The first
and shear directions. It can be used to predict the shear three terms are dimensionless to avoid the scale effect and
behavior under CNS as well as CNL conditions; for this therefore enhance the applicability of the suggested model,
reason, CNL and CNS conditions are not discussed sepa- while the last term has units of reciprocal length. If Cr is
rately in the material that follows. Here, the CNL condition zero, Eq. 13 represents the behavior of a planar joint; the
is assumed to represent a CNS condition in which rn0 = rn peak friction coefficient is only attributed to the basic
and Kn = 0.0. In developing the model, an empirical friction angle.
approach based on our experimental observations was The results of the laboratory tests indicate that asperity
taken. The calculations used in developing the model were failure may occur at stress levels that are even lower than
performed using the NCSS 2007 software package (Hintze the compressive strength; in such cases, the damaged
2007). thickness can be as small as a fraction of a millimeter up to
a few millimeters. These failures may be ascribed to the
5.1 Peak Friction Coefficient development of tensile forces due to interlocking of the
surfaces rather than to compressive forces. Thus, the tensile
The parameters that influence the peak friction coefficient strength (rt) of the specimen was chosen as the represen-
(lpk) include normal load conditions such as the initial tative asperity strength. The shear strength criteria sug-
normal stress (rn0) and normal stiffness (Kn) and joint gested by Ladany and Archambault (1970), Schneider
properties such as the asperity strength, roughness, and (1976), and Grasselli and Egger (2003) also adopted tensile
basic friction angle (b). In consideration of the effects of strength as a parameter for the joint wall strength. As rn0
the parameters discussed in Sect. 4, the regression model increases, the peak friction coefficient decreases, even
for the peak friction coefficient (lpk) was established as though the maximum shear stress may increase. Most
Eq. 13 by trial and error. Note that the proposed model experimental studies conducted under CNS conditions have

Fig. 17 Effect of joint roughness on shear behavior: results obtained with specimen Rc-02

123
J.-W. Park et al.

 
reported that Kn had no influence on the shear behavior spk JCS
before the first peak stress. The observations from our lpk tan JRC log10 /b ; 17
rn rn
laboratory, however, suggest that the peak friction coeffi-
cient decreases with normal stiffness due to the increment where spk is the peak shear strength, rn is the normal stress,
of normal stress around the peak stage. Thus, the ratio of JRC is the joint roughness coefficient, JCS is the joint wall
normal stiffness to initial normal stress (Kn/rn0) is used to compressive strength, and b is the basic friction angle. In
model the decrease in the peak coefficient with the normal the estimations made using Bartons model, the JCS was
stiffness. Note that Kn/rn0 has units of mm-1 because Kn assumed to be equal to the uniaxial compressive strength of
and rn0 have dimensions of MPa/mm and MPa, respec- the specimen, because all the tensile fractures were fresh
tively. The ratio Kn/rn0 is only required for the estimation without weathering (Barton and Choubey 1977).
of the peak friction coefficient under CNS conditions. In fact, reliable estimates of the JRC can be determined
The results of multiple regression analysis using all 362 through the results of tilt/pull test or direst shear test. How-
experimental data (GR: 112, LM: 52, GN: 47, Rr and Rc: ever, in the case of circular specimens, it does not make sense
151) for the model in Eq. 13 yield the following equation, to estimate the peak shear strength using the back-calculated
where the correlation coefficient (R2) is 0.782: JRC and then to compare it with the measured shear strength
 0:60 because the specimens were tested only once in each shear
3 1:43 rt direction; there were no additional shear tests for obtaining
lpk 3:1 10 Cr e0:02Kn =rn0 lb 14
rn0 the JRCs; in such cases, the peak shear strength estimated
using the back-calculated JRC naturally equals the measured
where CNL conditions apply, Eq. 14 can written as Eq. 15 peak shear strength. Therefore, as an alternative, the JRC
because Kn = 0.0. values calculated using the equation suggested by Yu and
 0:60 Vayssade (1991) were used for the estimations.
rt
lpk 3:1 103 Cr1:43 lb : 15 Figure 19 shows the results of the comparisons. The
rn0
measured strength of the circular specimens Rc-01 and Rc-02
The experimental results obtained in this study were showed good agreement with those estimated from the two
compared with their estimations by the proposed model. criteria. In both specimens, the shear strengths were con-
Figure 18 shows the results of the comparisons. In the figure, siderably higher when a = 270 than when a = 90. The
the peak shear stress spk is presented instead of the peak proposed criterion can reflect this tendency; it demonstrates
friction coefficient lpk to show the results more effectively, the ability of the active roughness parameter Cr to capture the
because more than 62 % of the peak friction coefficients anisotropy due to the effective roughness mobilization.
obtained in the experiments ranged from 1.0 to 2.0. Here, the However, Bartons model presents identical strengths in the
peak shear stress under CNL conditions indicates the peak reverse shear directions, because the JRCs were evaluated by
shear strength, while under CNS conditions it indicates the Yu and Vayssades equation, not by performing tilt/pull test
first peak shear stress or the shear stress at the peak friction or direct shear test according to shear directions.
coefficient. The experimental results agree well with those
estimated using the proposed model, even though the peak 5.2 Constitutive Model for the Shear Behavior
friction coefficients of the replicas tended to be of Rock Joints
underestimated when the initial normal stress rn0 was
higher than 1.0 MPa. The root-mean-square relative error As shown in Sect. 4, the shear behavior of rock joints
calculated from Eq. 16 was 18.7 %. under CNL and CNS can be simplified using the relation-
v
!2 ship between normal displacement and shear displacement
u
u 1 X 362 lipk; measured  lipk; estimated and that between friction coefficient and shear displace-
RMS of relative error t :
362 i1 lipk; measured ment. In this study, these relationships were formulated as
functions of the shear displacement on the basis of the
16
experimental results. In each function, the behavior was
divided into a prepeak stage and a postpeak stage. From the
The measured peak shear strength of the circular spec- relationships, the variations in the normal and shear stres-
imens Rc-01 and Rc-02 were compared with those esti- ses with the shear displacement were also derived.
mated from the proposed model and from Bartons
empirical model (Barton 1973). Bartons model is the most 5.2.1 Modeling of the Normal Dilation Curve
widely used criterion for estimating the shear strength of
rock joints; it can be written as follows in terms of the In terms of normal displacement, it was assumed that there
friction coefficient: was no contraction or dilation before the peak stages and

123
A Constitutive Model for Shear Behavior of Rock

Fig. 18 Comparison of shear strengths measured in the laboratory tests with those estimated by the proposed model: the results are displayed
according to (a) rock type and (b) applied initial normal stress

Fig. 19 Comparison of peak shear strengths measured in laboratory the estimations made using Bartons model, the JRC was calculated
tests with those estimated by the proposed model and Bartons using the equation suggested by Yu and Vayssade (1991)
empirical model: results obtained with samples Rc-01 and Rc-02. In

that the dilation curve followed a logarithmically increas- Here, dN, dT, and dpk are the normal displacement, the
ing function in the postpeak stage, as expressed in Eq. 18. shear displacement, and the shear displacement at the peak
dN 0 dT
dpk ; 18a stage, respectively, and Cd is the parameter used to
determine the shape of the normal displacement curve.
Cd
dN ln0:5dT  dpk 1:0 dT [ dpk : 18b Note that the unit of displacement is the millimeter.

123
J.-W. Park et al.

From the derivative of dN with respect to dT, the


dilation angle dN can be written as Eq. 19. The dilation
angle shows a maximum value of arctan(0.5Cd) when
dT = dpk and gradually decreases with increasing shear
displacement.
   
ddN 0:5Cd
dN arctan arctan :
ddT 0:5dT  dpk 1 19
If dT ! 1; then dN ! 0:

Cd is termed the dilation shape index; it can be a real


number that is positive or zero. Increasing Cd affects the
normal displacement and the dilation angle at a certain
shear displacement. If Cd is zero, normal dilation does not
occur; this indicates the shear behavior of a completely
planar joint. Figure 20 shows normal displacement curves
according to the dilation shape index Cd.
For all dilation curves obtained in the experiments, the
measured data (dT, dN) in the postpeak stage were fitted Fig. 20 Modeling of the relationship between normal displacement
and shear displacement; the shape of each curve is determined by the
using Eq. 18b, and the Cd of the best fit was obtained. In shape index Cd
this procedure, dpk was substituted by the measured value
in each experiment. The calculated Cd of the curves 5.2.2 Modeling of the Friction Coefficient
measured in our study ranged from 0.194 to 2.285. We
attempted to derive the relationship between Cd and the It was assumed that the friction coefficient increases line-
influencing parameters. The effect of each parameter on arly to a maximum value in the prepeak stage, and dra-
Cd was similar to that on lpk except for basic friction matically decreases in the residual stages, following a
angle. Thus, we set the regression model of Cd as negative exponential function in the postpeak stage, as
follows: expressed in Eq. 22.
 C
rt l kl dT dT
dpk ; 22a
Cd ACrB eDKn =rn0 ; 20
rn0
l lpk  lrs eCf dT dpk lrs dT [ dpk : 22b
where A, B, C, and D are dimensionless parameters to fit
the experimental data. In Eq. 22, l, lpk, and lrs are the friction coefficient, the
In comparison with Eq. 13, the model of Cd includes the peak friction coefficient, and the residual friction
same components as the model of lpk, except for basic coefficient, respectively; dT and dpk are the shear
friction angle. The result of the multiple regression analysis displacement and the shear displacement at the peak
using all experimental data for the model of Eq. 20 is as stage, and kl is the slope of the linear portion of the friction
follows, where the correlation coefficient (R2) is 0.748: coefficient curve in the prepeak stage. Note that the unit of
 0:18 displacement is the millimeter.
2 0:96 rt Cf is termed the friction shape index; it can be a real
Cd 4:32 10 Cr e0:1Kn =rn0 : 21
rn0 number that is positive or zero. With increasing Cf, the
In Eq. 21, Cr is the active roughness coefficient, rt is the friction coefficient curve shows a dramatic decrease after
tensile strength, rn0 is the initial normal stress, and Kn is the peak stage and reaches the residual state more quickly.
the normal stiffness. Note that the ratio of normal stiffness If Cf is zero, the friction coefficient does not change in the
to initial normal stress (Kn/rn0) has units of mm-1. postpeak stage (lpk = lrs); this represents the behavior of a
Figure 21 shows a comparison of the dilation index Cd completely planar joint. Figure 22 shows the friction
obtained through curve-fitting of the experimental results coefficient curves according to the friction shape index Cf.
to Eq. 18b with that estimated from the influencing The slope kl has following relationship to the shear
parameters using Eq. 21. Substituting the Cd of Eq. 21 into stiffness, Ks:
Eq. 18, we can predict the normal dilation curve from the Ks l s s
kl *kl ; Ks and l
boundary conditions (Kn and rn0) and the joint properties rn dT dT rn 23
(Cr and rt). dT
dpk :

123
A Constitutive Model for Shear Behavior of Rock

Fig. 22 Modeling of the relationship between friction coefficient and


shear displacement; the shape of the curve is determined by the shape
index Cf
Fig. 21 Comparisons of the index Cd obtained through curve-fitting
(Eq. 18b) with the value of Cd estimated from the influencing
parameters (Eq. 21)
As discussed in Sect. 4, the residual friction coefficient
In the relationship, the normal stress rn can be substi- was generally higher than the basic friction coefficient.
tuted by the initial normal stress rn0 in the case of CNS Note that the residual friction coefficient was determined
conditions because no dilation is assumed in the prepeak by averaging the data with values between 80 and 100 % of
stage. total shear displacement: 1215 mm for GR, LM, GN, and
For all the friction coefficient curves, the measured data Rr (rectangular replica) specimens and 810 mm for Rc
(dT, l) in the postpeak stage were fitted to Eq. 22b, and the Cf (circular replica) specimens. In the strict sense, the residual
of the best fit was obtained. In this procedure, dpk, lpk, and lrs friction coefficient lrs in this study indicates the ultimate
are substituted by the measured values in each experiment. friction coefficient. When we substituted the basic friction
The calculated Cf of the curves measured in our study ranged coefficient lb (=tan b) instead of the residual friction
from 0.13 to 4.67. Unlike the case for the dilation shape index coefficient lrs into Eq. 22b, the model showed a poor fit to
Cd, it was difficult to clearly define the relationship between the experimental data in many cases. Therefore, it was
Cf and the influencing parameters because the friction coef- necessary to model the residual friction coefficient.
ficient curves showed many different shapes in the peak According to the experimental results, the effects of the
stages. However, Cf tended to increase with decreasing initial boundary condition and joint properties on lrs were similar
normal stress and with increasing joint roughness. Thus, from to their effects on lpk except for the basic friction angle. In
the multiple regression analysis of the experimental results, the model of the peak friction coefficient expressed as
we suggest that the relationship between Cf and influencing Eq. 14, lpk - lb indicates the shear resistance contributed
factors is as follows: by the combination of the initial normal stress, normal
 0:2 stiffness, joint roughness, and joint strength. Thus, from the
0:3 rt multiple regression analysis of the experimental results, we
Cf 0:19Cr : 24
rn0 propose Eq. 25 as a model of the residual friction
Although the friction shape index Cf obtained by fitting coefficient:
the experimental results to Eq. 22b was not always lrs 0:16lpk  lb 0:36 ls : 25
consistent with its value estimated by Eq. 24, this
approach may be more reasonable than the use of other Substituting Cf of Eq. 24, lpk of Eq. 14, and lrs of
methods (Goodman 1974; Grasselli and Egger 2003) in Eq. 25 into Eq. 22, we can predict the friction coefficient
which the shape of the friction coefficient curve is curve from the boundary conditions (Kn and rn0) and the
relatively simplified. joint properties (Cr, rt, and lb).

123
J.-W. Park et al.

5.3 Application of the Proposed Model Figures 2325 show the results of CNL tests, and
Figs. 26 and 27 show the results of CNS tests. In CNL
In Sect. 5.2, models for the normal dilation and peak tests, the normal stress remains constant; accordingly, the
friction coefficient were proposed based on the experi- friction coefficient curve resembles that of the shear stress
mental results. Using these relationships, the normal stress curve. Thus, only the shear stress and normal dilation of
and shear stress can also be formulated as functions of the each case are presented. On the other hand, for the CNS
shear displacement. tests, the friction coefficient, normal displacement, shear
The relationship between normal stress (rn) and normal stress, and normal stress curves are all presented.
dilation (dN) is written as Eq. 26. The comparisons suggest that there is good agreement
rn rn0 Kn dN : 26 between the constitutive model and the experimental
results. Under both CNL and CNS conditions, the proposed
Substituting dN of Eq. 18 into Eq. 26, we obtain the model predicts the shear behaviors of rock joints with
relationship between normal stress (rn) and shear different joint properties and different boundary conditions
displacement (dT) as follows: well (Figs. 23, 26). The effects of joint roughness
rn rn0 dT
dpk ; 27a (Fig. 24), initial normal stress (Fig. 25), and normal stiff-
 C ness (Fig. 27) are reasonably reflected in the model. In
rn rn0 Kn ln 0:5dT  dpk 1 d dT [ dpk : 27b addition, the various characteristics of shear stress curves
under CNS conditions, including stress softening and
On the other hand, the relationship between shear stress
hardening tendencies, are well presented.
(s) and normal stress (rn) is written in Eq. 28.
s lrn : 28
Substituting l of Eq. 22 and rn of Eq. 27 into Eq. 28, 6 Summary and Conclusions
we obtain the relationship between shear stress (s) and
shear displacement (dT) as follows: In this paper, a new constitutive model relating stress and
deformation is developed to describe the shear behavior of
s rn0 kl dT dT
dpk ; 29a rock joints under constant normal load (CNL) and constant
h i
s lrs lpk  lrs eCf dT dpk normal stiffness (CNS) conditions. This model is derived
h i on the basis of experimental results and a newly proposed
rn0 Kn ln0:5dT  dpk 1Cd dT [ dpk : 29b roughness parameter, which is termed the active roughness
coefficient, Cr.
In summary, variations in the normal dilation, friction Adopting the approach taken by Park and Song (2013),
coefficient, normal stress, and shear stress with the shear the features of roughness mobilization at the contact areas
displacement can be expressed as Eqs. 18, 22, 27, and 29, were investigated. In this procedure, a joint surface is
respectively; in developing these expressions, the effects of represented as a group of triangular planes. The contact
the boundary conditions and joint properties were condition of each plane was examined, and the surface
considered by adopting the indices Cd and Cf. roughness was characterized using the active/inactive
Figures 2327 present comparisons between the shear micro-slope angle with respect to the shear direction. From
behaviors obtained in laboratory tests and those estimated the observations that the contact areas were almost entirely
by the proposed models. Due to the considerable volume made up of active elements and that the locations of the
of test data obtained, only representative cases are contact areas were remarkably consistent with those of the
included here. In the estimation, the experimental data active zones, we characterized the surface roughness using
were used only for dpk. In practice, it was quite difficult the active roughness coefficient, Cr, which determined the
to find a systematic relationship between dpk and the characteristics of the probability distribution of the active
influencing parameters in our experiments. In some cases, micro-slope angle.
smooth and weak joints exhibited nonlinear behaviors in To investigate the shear behavior of rock joints under
the prepeak stages or showed a ductile tendency around CNL and CNS conditions, a series of direct shear tests were
the peak stage, which caused problems in the determi- carried out on tensile fractured rock joints and on replicas
nation of the point at dpk and in the observation of the of tensile joints that had various strength and roughness
effects of various parameters. According to Eq. 23, the values. The effects of the initial normal stress, normal
shear displacement dpk may be calculated from the shear stiffness, joint strength, and joint roughness on the shear
stiffness Ks because we can estimate spk. However, a behavior were discussed. According to the experimental
model to estimate Ks has not been reported in the liter- results, it was notable that the relative magnitudes of nor-
ature to date. mal stiffness and initial normal stress affected both the

123
A Constitutive Model for Shear Behavior of Rock

Fig. 23 Comparison of the results obtained from CNL tests with those estimated by the proposed model: results obtained with specimens a GR-
10, b LM-12, c GN-08, and d Rr-07

shear stress curves after the first peak stage and the peak normal stiffness to initial normal stress (Kn/rn0) were
friction coefficient. In particular, the peak friction coeffi- selected. Empirical models for the peak friction coefficient,
cient decreased with increasing normal stiffness. the normal displacementshear displacement relationship,
The constitutive model proposed in this paper involves a and the normal dilationshear displacement relationship
criterion for peak friction coefficient, and the relations were developed from the multiple regression analysis of
between stresses and displacements in normal and shear 362 experimental data. The latter two models are expressed
directions. In our discussion, we focused on the variations as functions of the shear displacement, and they imply the
in normal dilation and friction coefficient with shear dis- effects of the boundary conditions and joint properties by
placement because it was possible to generalize these the shape indices Cd and Cf. From these relationships, the
behaviors regardless of the boundary condition, contrary to variations in the normal stress and shear stress with the
the normal and shear stresses, which showed complicated shear displacement can also be formulated. The shear
behaviors under CNS conditions. As factors that influence behaviors obtained from the experimental results were in
the shear behavior of rock joints, the ratio of tensile general in good agreement with those estimated by the
strength to initial normal stress (rt/rn0), the active rough- proposed model, and the effects of the joint roughness,
ness coefficient (Cr), the friction coefficient mobilized by initial normal stress, and normal stiffness were reasonably
the basic friction angle (lb = tan b), and the ratio of reflected in the model. Moreover, the active roughness

123
J.-W. Park et al.

Fig. 24 Comparison of the results obtained from CNL tests on specimen Rc-02 with those estimated by the proposed model: effect of joint
roughness

Fig. 25 Comparison of the results obtained from CNL tests on specimen GN-07 with those estimated by the proposed model: effect of normal
stress

coefficient Cr could capture the features of the roughness comprehend the features of the effective roughness mobi-
mobilization and the corresponding shear behavior with lized at the contact areas during shearing. Further study is
respect to the shear direction. needed to ascertain if the characteristics of the distribution
Nevertheless, in the present study, no attempt was made of active micro-slope angle are invariant to joint size. The
to demonstrate the scale effect on the shear behavior of effects of the sampling interval at which the joint surface is
rock joints; the proposed model has validity only for rock described on the active roughness coefficient also should
joints at laboratory scale. The scale effect of the shear be thoroughly examined, because reducing the size of the
behavior may be mainly ascribed to the description of joint sampling interval would provide more detailed descriptions
roughness. Therefore, to obtain large-scale surface rough- about the measured surface, but the sampling interval may
ness of rock joints, a reliable method is needed for influence the micro-slope angle and the active roughness
extrapolating laboratory data to field scale. A new approach coefficient. In addition, the procedure to obtain the active
is proposed here for considering the joint roughness in roughness coefficient needs to be more simplified for the
engineering problems, and it would be the first step to proposed constitutive model of shear behavior of rock

123
A Constitutive Model for Shear Behavior of Rock

Fig. 26 Comparison of the results obtained from CNS tests with those estimated by the proposed model: results obtained with specimens a GN-
01, b GR-28, and c Rr-02

123
J.-W. Park et al.

Fig. 27 Comparison of the results obtained from CNS tests on specimen Rr-07 with those estimated by the proposed model: effect of normal
stiffness

joints to be more practical. To overcome this deficiency, Byerlee JD (1970) The mechanics of stick-slip. Tectonophysics
use of two-dimensional profiles parallel to the shear 9:475486
Crotty JM, Wardle LJ (1985) Boundary integral analysis of piecewise
direction when establishing the distribution of micro-slope homogeneous media with structural discontinuities. Int J Rock
angle could offer a viable alternative to constructing a Mech Min Sci Geomech Abstr 22:419427
three-dimensional joint model with triangular elements; Goodman RE (1974) The mechanical properties of joints. In:
this will be one of the subjects in our future research. Proceedings of the 3rd congress of international society of rock
mechanics, Denver, pp 127140
Goodman RE (1976) Methods of geological engineering in discon-
Acknowledgments This research was supported by a National tinuous rocks. West, New York
Research Foundation of Korea (NRF) grant funded by the Ministry of Grasselli G, Egger P (2003) Constitutive law for the shear strength of
Education, Science, and Technology of Korea (No. 2012-0005326), rock joints based on three-dimensional surface parameters. Int J
and supported by a Basic Research Project of the Korea Institute of Rock Mech Min Sci 40:2540
Geoscience and Mineral Resources (KIGAM, GP2012-001) funded Haberfield CM, Johnston IW (1994) A mechanistically-based model
by the Ministry of Knowledge and Economy of Korea. We appreciate for rough rock joint. Int J Rock Mech Min Sci Geomech Abstr
anonymous reviewers for their valuable comments and suggestions to 31:279292
improve this manuscript. Hintze JL (2007) NCSS 2007: Statistical analysis and graphics.
Users Guide, Number Cruncher Statistical Systems, Kaysville,
Utah
References Huang SL, Oelfke SM, Speck RC (1992) Applicability of fractal
characterization and modeling to rock joint profiles. Int J Rock
Awrejcewicz J, Olejnik P (2005) Analysis of dynamic systems with Mech Min Sci Geomech Abstr 29:8998
various friction laws. Appl Mech Rev 58:389411 Indraratna B, Haque A (1997) Experimental study of shear behavior
Bandis SC, Lumsden AC, Barton NR (1981) Experimental studies of of rock joint under constant normal stiffness conditions. Int J
scale effects on the shear behaviour of rock joints. Int J Rock Rock Mech Min Sci 34(3-4):141.e1141.e14
Mech Min Sci Geomech Abstr 18:121 Indraratna B, Haque A, Aziz N (1998) Laboratory modelling of shear
Bandis SC, Lumsden AC, Barton NR (1983) Fundamentals of rock behaviour of soft joints under constant normal stiffness condi-
joint deformation. Int J Rock Mech Min Sci Geomech Abstr tions. Geotech Geol Eng 16:1744
20:249268 Jiang Y, Xiao J, Tanabashi Y, Mizokami T (2004) Development of an
Barton N (1973) Review of a new shear-strength criterion for rock automated servo-controlled direct shear apparatus applying a
joints. Eng Geol 7:287332 constant normal stiffness condition. Int J Rock Mech Min Sci
Barton N, Bandis S (1990) Review of predictive capabilities of JRC- 41:275286
JCS model in engineering practice. In: Barton N, Stephansson O Jiang Y, Li B, Tanabashi Y (2006) Estimating the relation between
(eds) Rock joints. Balkema, Rotterdam, pp 603610 surface roughness and mechanical properties of rock joints. Int J
Barton N, Choubey V (1977) The shear strength of rock joints in Rock Mech Min Sci 43:837846
theory and practice. Rock Mech Rock Eng 10:154 Johnston IW, Lam TSK, Williams AF (1987) Constant normal
Barton N, Bandis S, Bakhtar K (1985) Strength, deformation and stiffness direct shear testing for socketed pile design in weak
conductivity coupling of rock joints. Int J Rock Mech Min Sci rock. Geotechnique 37:8389
Geomech Abstr 22:121140 Kodikara JK, Johnston IW (1994) Shear behaviour of irregular
Brown ET (ed) (1981) ISRM suggested methods: rock characteriza- triangular rock-concrete joints. Int J Rock Mech Min Sci
tion testing and monitoring. Pergamon, Oxford Geomech Abstr 31:313322

123
A Constitutive Model for Shear Behavior of Rock

Kulatilake PHSW, Shou G, Huang TH, Morgan RM (1995) New peak Schneider H (1976) The friction and deformation behaviour of rock
shear strength criteria for anisotropic rock joints. Int J Rock joints. Rock Mech Rock Eng 8:169184
Mech Min Sci Geomech Abstr 32:673697 Scholz CH, Engelder JT (1976) The role of asperity indentation and
Ladany B, Archambault G (1970) Simulation of shear behavior of a ploughing in rock friction-I: asperity creep and stick-slip. Int J
jointed rock mass. In: Proceedings of the 11th US symposium on Rock Mech Min Sci Geomech Abstr 13:149154
rock mechanics, New York, pp 105125 Seidel JP, Haberfield CM (2002a) Laboratory testing of concrete-rock
Lee HS, Park YJ, Cho TF, You KH (2001) Influence of asperity joints in constant normal stiffness direct shear. Geotech Test J
degradation on the mechanical behavior of rough rock joints 25:391404
under cyclic shear loading. Int J Rock Mech Min Sci 38:967980 Seidel JP, Haberfield CM (2002b) A theoretical model for rock joints
Maerz NH, Franklin JA, Bennett CP (1990) Joint roughness subjected to constant normal stiffness direct shear. Int J Rock
measurement using shadow profilometry. Int J Rock Mech Min Mech Min Sci 39:539553
Sci Geomech Abstr 27:329343 Skinas CA, Bandis SC, Demiris CA (1990) Experimental investiga-
Ohnishi Y, Dharmaratne PGR (1990) Shear behaviour of physical tions and modelling of rock joint behavior under constant
models of rock joints under constant normal stiffness condition. stiffness. In: Barton N, Stephansson O (eds) Rock joints.
In: Barton N, Stephansson O (eds) Rock joints. Balkema, Balkema, Rotterdam, pp 301308
Rotterdam, pp 413422 Son BK, Lee CI, Park YJ, Lee YK (2006) Effect of boundary
Olsson R, Barton N (2001) An improved model for hydromechanical conditions on shear behaviour of rock joints around tunnel. Tunn
coupling during shearing of rock joints. Int J Rock Mech Min Sci Undergr Sp Tech 21:347348
38:317329 Tse R, Cruden DM (1979) Estimating joint roughness coefficients. Int
Park JW, Song JJ (2013) Numerical method for determining contact J Rock Mech Min Sci Geomech Abstr 16:303307
areas of a rock joint under normal and shear loads. Int J Rock Van Sint Jan ML (1990) Shear tests of model rock joints under stiff
Mech Min Sci 58:822 normal loading. In: Barton N, Stephansson O (eds) Rock joints.
Patton FD (1966) Multiple modes of shear failure in rock. In: Balkema, Rotterdam, pp 323327
Proceedings of the 1st congress of international society of rock Xie H, Sun H, Ju Y, Feng Z (2001) Study on generation of rock
mechanics, vol 1. Lisbon, pp 509513 fracture surfaces by using fractal interpolation. Int J Solids Struct
Plesha ME (1987) Constitutive models for rock discontinuities with 38:57655787
dilatancy and surface degradation. Int J Numer Anal Meth Yang ZY, Chiang DY (2000) An experimental study on the
Geomech 11:345362 progressive shear behavior of rock joints with tooth-shaped
Rim HR, Choi HJ, Son BK, Lee CI, Song JJ (2005) Experimental asperities. Int J Rock Mech Min Sci 37:12471259
study for shear behaviour of pseudo rock joint under constant Yeo IW, de Freitas MH, Zimmerman RW (1998) Effect of shear
normal stiffness condition. In: Proceedings of the 31st ITA- displacement on the aperture and permeability of a rock fracture.
AITES world tunnel congress on underground space use, Int J Rock Mech Min Sci 35:10511070
Istanbul, pp 175181 Yu X, Vayssade B (1991) Joint profiles and their roughness
Saeb S, Amadei B (1990) Modelling joint response under constant or parameters. Int J Rock Mech Min Sci Geomech Abstr
variable normal stiffness boundary conditions. Int J Rock Mech 28:333336
Min Sci Geomech Abstr 27:213217

123

You might also like