arma-04-483_modelling dilation in brittle rocks
TRANSCRIPT
Copyright 2004, ARMA, American Rock Mechanics Association
This paper was prepared for presentation at Gulf Rocks 2004, the 6th North America Rock Mechanics Symposium (NARMS): Rock Mechanics Across Borders and Disciplines, held in Houston, Texas,June 5 - 9, 2004.
This paper was selected for presentation by a NARMS Program Committee following review of information contained in an abstract submitted earlier by the author(s). Contents of the paper, as presented,have not been reviewed by ARMA/NARMS and are subject to correction by the author(s).The material, as presented, does not necessarily reflect any position of NARMS, ARMA, CARMA, SMMR, theirofficers, or members. Electronic reproduction, distribution, or storage of any part of this paper for commercial purposes without the written consent of ARMA is prohibited. Permission to reproduce inprint is restricted to an abstract of not more than 300 words; illustrations may not be copied. The abstract must contain conspicuous acknowledgement of where and by whom the paper was presented.
ARMA/NARMS 04-483
Modelling Dilation in Brittle Rocks
N. Cho, C.D. Martin & D.C. Sego
Dept. Civil & Environmental Engineering, University of Alberta, Edmonton, Canada, T6G 2G7
R. Christiansson
Swedish Nuclear Fuel and Waste Management Co., Stockholm, Sweden
ABSTRACT: In laboratory tests, the onset of dilation occurs at stress levels far below the peak strength but yielding of
the laboratory specimen is not synonymous with the onset of dilation, and is seldom measured or reported in traditional
laboratory testing. In field tests, the on-set of dilation is often associated with stress-induced extension fracturing. The
displacements associated with these stress-induced fractures, cannot be replicated using traditional constitutive modelling
and associated or non-associated flow rules. In this paper a methodology is developed for modeling dilation using the
Particle Flow Code (PFC) that captures many of the observations reported in conventional laboratory test results. The
findings from this research show that clumped-particle geometry provides the best agreement with laboratory test results
for both tensile and compressive loading paths.
1 INTRODUCTION
Experience with underground excavations at depth in-
dicates that one of the most significant phenomena
observed in brittle rocks is extensile fracturing. This
fracturing occurs as a result of tangential stress con-
centrations. Direct observation of brittle rock failure
around underground openings reveals that this exten-
sile fracturing exhibits significant dilation (Fig. 1). A
detailed description of the spalling process observed
around a circular test tunnel was given by Martin et al.
[1] and Lajtai [3] showed that in laboratory samples
the brittle failure process resulted in the opening of
fractures.
In materials such as metals and clays, yielding can oc-
cur without significant volume change. However, in
brittle rocks on the boundary of underground openings
overstressing results in the development of micro- and
macro-cracks. In the mining industry, the process is
often referred to as ‘spalling’ or ‘dog-earing’. In the
petroleum industry, the problem is often cast as ‘well-
bore breakouts’. One of the early descriptions in civil
engineering was given by Terzaghi [2] and referred
to as ‘popping rock’. Modeling of this process has
always been challenging and has received a lot of at-
tention in the mining, nuclear waste and petroleum
industries since the 1950’s.
With the advent of modern computers, both contin-
uum mechanics and traditional fracture mechanics
approaches have been used to model this fracturing
process [3–6]. The use of continuum mechanics to
Fig. 1: Dilation associated with stress-induced fracturing
observed in a 600-mm-diameter borehole.
simulate a fracturing process that results in an open
rough fracture, as described by Lajtai [3] and shown in
Fig. 2, is extremely problematic as the displacement
field across the fracture in a continuum must be con-
tinuous. But if the fracture is open, this requirement
cannot be satisfied. In traditional fracture mechanics,
the fracture has zero width, again suggesting that this
approach is not applicable for representing a process
that results in open fractures. In all these approaches
specific flow rules are required to capture the displace-
ment field. In continuum mechanics an associated or
non-associated flow rule is assumed.
For the fracture mechanics approach the control of
the fracture growth is related to the fracture toughness
(KIC) [4, 6–8]. In both approaches there are funda-
mental shortcomings. Using continuum mechanics,
there is no easy way to generate tension, ie., an exten-
sile crack, when the stress state is all round compres-
sion. The same restriction also applies to fracture me-
chanics codes. In fracture mechanics this shortcoming
is overcome by inclining the crack to the stress field to
generate ’wing-cracks’. By taking this approach there
is a fundamental assumption that the rock is weaker
in shear than in tension since the only way tensile
cracks can be generated is by shearing first. This is
counter to one of the most important characteristics
of rocks as noted by Neville Cook in his Müller Lec-
ture: [9], rocks are fundamentally weaker in tension
than in shear. Hence any modeling process that vi-
olates this characteristic would appear to have little
chance of modeling the spalling process.
Fig. 2: Example of the fracture surface observed in core
discing (left) and direct tension test (right). Note the rough-
ness along the fractures at the grain scale.
Since the mid 1990’s there has been growing interest
in discrete element modeling. As noted by Cundall
[10], one of the major advantages of this numerical
method is that one does not specify a flow rule. How-
ever, there are other issues associated with model cal-
ibration and micro-scale properties that need to be
resolved. In this paper the discrete element program
PFC is used to explore if it can overcome some of the
difficulties noted previously when modeling fracture
initiation and growth and the associated dialtion.
2 A BONDED PARTICLE MODEL
All rocks are heterogeneous at the micro and macro
scale. In rocks such as granite the heterogeneous na-
ture can be readily observed with the naked eye. Lab-
oratory testing of such rocks shows that when sub-
jected to sufficient deviatoric loads these rocks dilate
laterally indicating that axially aligned fractures are
forming [11]. The deviatoric load required to generate
these opening, i.e, dilatant fractures, is considerably
less than the peak strength of the rock and for uniaxial
compressive tests is often reported as 1/3 to 1/2 the
uniaxial strength. If unloaded permanent straining is
recorded in the lateral direction, while no permanent
straining is recorded in the axial direction indicating
that the dilatancy process is not elastic and not uni-
formly distributed. One of the mechanisms that can
explain this observation is illustrated in Fig. 3.
σ1
σ3
Tension Crack
(a) (b)
(c) (d)
Distortioncausedby deviatoricstress
Initiation of permanent dilation
Fracture propagation& associated dilation
Fig. 3: An illustration of the dilation process in brittle het-
erogeneous rocks.
As rock is subjected to deviatoric stresses, local ten-
sile stress will be generated because of the heteroge-
neous nature of the rock. A likely location for a crack
resulting from these tensile stresses is at the grain
boundary (Fig. 3a). Because of the irregularities at
grain boundaries, even a small amount of elastic dis-
tortion, once the tensile crack forms, could result in
dilation. This dilation may be sufficient to induce ten-
sile stresses at the crack tip and cause the crack to open
(Fig. 3b). Moreover, this elastic distortion will result
in additional tensile stresses and when distributed to
neighboring cracks could result in additional cracking
and permanent dilation (Fig. 3c). The coalescence or
alignment of individual cracks may ultimately prop-
agate in an unstable manner, i.e., because the region
of tensile stresses grows, and form a macro fracture
surface with associated dilation (Fig. 3d). The essen-
tial elements of this process are the deviatoric stresses
required to cause the distortion once the crack forms
and the nonplanar surface of the microcrack. Core
disking is an extreme example of the self propagating
fracture resulting from this process.
2.1 PFC
The bonded particle analogue used in PFC repre-
sents a rock mass as an assemblage of circular disks
connected with cohesive and frictional bonds. In this
model, the breakages of bonds between particles by
applied local stresses are expressed as cracks. PFC
does not require any flow rule for describing the post
peak region or fracture toughness to control fracture
behavior, but only requires the law of motion of parti-
cles and laws for particle bond rupture and deforma-
tion.
Diederichs [12] successfully used this approach to
simulate brittle behavior of rock under compressive
loading and noted that a bond rupture in PFC does not
create the same singular stress concentration present
at the tip of an extending micro crack within a con-
tinuum. The rupture in PFC results in stress redistri-
bution in neighboring bonds, but this redistribution
may not be adequate to rupture the adjacent contacts.
As a result the crack generating process in PFC is a
stable process, i.e., to generate new cracks the devi-
atoric stress must be increased. Therefore, PFC will
not, without modification model the dilation and crack
generation process described above.
In direct tensile loading in PFC, once cracks are ini-
tiated, they are immediately propagated in an unsta-
ble fashion because the loading boundary conditions
cause stress concentrations at the crack tips which lead
to unstable fracturing, i. e., the load does not need to be
increased once fracturing begins to get the sample to
fail. This means that once a crack is initiated, it cannot
be stabilize until the boundary stress is removed or re-
duced. Whereas in compression, crack accumulation
must occur and individual cracks must interact to cre-
ate a macroscopic rupture surfaces.This is very impor-
tant when simulating extensile fracturing around un-
derground openings. Because the stress-induced ex-
tensile fracturing which occurs at the boundary of the
underground opening where the confining stress is ap-
proximately zero, and therefore if large scale tensile
stresses can occur, extensile fracturing can propagate
in an unstable fashion similar to the tensile loading
example.
One of the disconcerting results when using PFC
to represent rock is that the uniaxial compressive
strength obtained in PFC is approximately 4 times
greater than the tensile strength. This is dispropor-
tionately low, compared to granite and most other
rocks where the ratio of σt/σci is typically reported as
0.04 to 0.03 (σt/σci = 1/24 to 1/30) [13]. Diederichs
[12] showed that calibrating PFC to the compressive
strength resulted in significantly higher relative ten-
sile strength, and a linear strength envelope without a
tension cut-off as illustrated in Fig. 4. This is clearly
not in keeping with measured laboratory results.
s1
s3
Contact
Force chain
Rupture by
tensile bond
breakage
Cracks caused by
tensile bond breakage
doesn't cross
compressive bridge
Fig. 4: Contact force chain structure in PFC and the failure
envelope when compression is used for calibration tension
[12].
Diederichs [12] concluded that the linear tensile en-
velope in PFC is attributed to the intrinsic contact
force fabric structure in PFC. Diederich’s argument
is very important when modeling the dilation process
discussed above because extensile fracturing induces
dilation and vice versa. In other words, modeling of
extensile fracturing, dilation and high tensile strength
problem in PFC are not separate issues. In the fol-
lowing section the factors that could control dilation
in PFC are explored and quantified.
3 FACTORS CONTROLLING DILATION IN
PFC
PFC is a discrete element code that represents a rock
mass as an assemblage of circular disks confined by
planar walls. In this system, the particles can move
independently of one another and interact only at con-
tacts. They are assumed to be rigid but can be over-
lapped at the contacts under compression. The parti-
cles can be bonded together be specifying the shear
and tensile bond strength. The values assigned to these
strengths influence the macro strength of the sample
and the nature of cracking and failure that occurs dur-
ing loading. The shear strength can only be mobilized
once the tensile bond strength is broken or set to zero.
Similarly, the contact shear forces are a function of
compressive normal force at the contact and coeffi-
cient of friction. After a bond breaks in PFC, stress is
redistributed and this may then cause adjacent bonds
to break. Thus, PFC only requires basic parameters
to describe contact stiffness (kn, ks), bond strength
(bn, bs), friction (µ) and does not require any plas-
tic flow rule formulation. This implies that only those
parameters and geometrical factors that related to par-
ticle structure such as particle size or particle shape
can control dilation in a discrete element model. It
is therefore, most important to understand how those
parameters may control dilation in PFC.
3.1 Bond Models
Two basic bond models are provided in PFC: (1) Con-
tact Bond and (2) Parallel bond model. The contact
bonds can be envisaged as a pair of elastic springs (or
point of glue) with constant normal and shear stiffness
acting at a point. A parallel bond approximates the
physical behavior of a cement-like substance joining
the two particles. Parallel bonds establish an elastic
interaction between particles that acts in parallel with
the slip or contact-bond constitutive models. It also
t = 1
M
V
T
L
A B
L L
2R
σ = Fn
AR+
M
I
Fn Fn
M M
Fig. 5: Illustration of parallel bond model.
can be envisioned as a set of elastic springs uniformly
distributed over a rectangular cross section with con-
stant normal bond stiffness shear bond stiffness lying
on the contact plane and centered at the contact point.
Particles in PFC are free to move in the normal and
shear direction and can also rotate between particles.
This rotation may induce a moment between particles
but the Contact Bond model cannot resist this mo-
ment. With the Parallel Bond model however, bond-
ing is activated on finite area, this bonding can resist
the moment as shown Fig. 5. The Parallel Bond model
is therefore a more realistic bond model for rock like
materials whereby the bonds break by tension rather
than by particle rotations.
3.2 Parametric Study
The PFC bond parameters were examined using uni-
axial compressive test and brazilian test simulation.
The procedure for the sample generation and load-
ing methods are illustrated elsewhere and in the PFC
manual.
To avoid issues associate with loading rate, the plate
velocity was set to 0.2 m/s for uniaxial compressive
test and 0.05 m/s for the Brazilian test. Those load-
ing rates are slow enough to ensure that the speci-
men remains in quasi-static equilibrium throughout
the test, such that no inertial effects occur (e.g., pre-
mature bond failure resulting from a stress wave pass-
ing through the specimen or platen loads in excess of
their quasi-static equilibrium values). Because of the
heterogeneity introduced in PFC (e.g. standard devi-
ation applied to micro strength and arbitrarily gener-
ated disk assembly with different ball radius), no two
PFC runs produces identical results. During this para-
metric study, each simulation was run 10 times with
same parameter and the results averaged. While this is
somewhat laborious it approximates the real variabil-
ity encountered when testing rocks and avoids getting
unique answers that may not be relevant.
The lateral strains for the compressive tests were
recorded at 80% of peak strength for the purposes
of this study. The lateral strain is used to gauge
the amount of dilation in each sample. All compres-
sive samples measured 30 mm by 60 mm and con-
sisted of 963 particles. Tensile strength was obtained
from Brazilian test simulation with the PFC specimen
taken from the uniaxial compressive test sample us-
ing the same diameter of the Brazilian sample as the
width of the compressive sample. The ratio of tensile
strength to compressive strength was determined for
each simulation. The micro parameters used during
the parametric study are were based on the Lac Du
Bonnet Granite and given in Table 1.
Table 1: Micro-parameters in PFC to represent Lac du Bon-
net granite.
Ec 49 GPa Ec 49 GPa
kn/ks 1 kn/ks 1
µ 0 σc 200 ± 50 MPa
σc/τc 1 λ 1
Rmin 0.55 mm Rmin/Rmax 1.65
3.3 Effect of Particle Friction
As parallel bond breaks either by rotation or shear in
PFC, the force acting on the particle contact are set
to a residual value that depends on the compressive
normal force at the contact and coefficient of fric-
tion. The friction between particles, when mobilized
after bond breakage, can suppress independent par-
ticle movement such as rotations and slippage. This
process could increase the shear resistance resulting
in high compressive macro strength. With increasing
friction, dilation may be suppressed provided the di-
lation process is dominated by the micro friction pa-
rameters.
In this study, the coefficient of friction was varied
from 0.0 to 0.9. Fig. 6 illustrates that as friction in-
creases dilation is suppressed. Also shown in Fig. 6 is
the ratio of tensile strength to compressive strength.
The strength ratio is not influenced by friction at the
contacts. This implies that friction between particles
does not effect the force chain structure discussed by
[12].
3.4 Effect of Bond Stiffness
Diederichs [12] showed that the normal to shear stiff-
ness ratio (kn/ks) involves Poisson’s ratio. If the stiff-
ness ratio is high, Poisson’s ratio increases and the
sample becomes brittle. For brittle material therefore,
this stiffness ratio must be set to a lower value around
1.0. In this study, the stiffness ratio was varied from
1.0 to 10.0. Fig. 7 indicates that as stiffness ratio is
increased, the amount of dilation is also increased im-
plying that more tension cracks develop.
Potyondy and Cundall [14] also found similar results
and suggested that when the stiffness ratio is relatively
high, the Poisson’s effect attracts more tensile cracks.
When the number of tensile cracks increases, the lat-
eral dilation also increases. However, the strength ra-
tio in Fig. 7 indicates that this increase in dilation does
not impact the strength ratio.
3.5 Effect of particle size
Potyondy and Cundall [15] suggested that particle
size could control tensile strength. They found a rela-
tionship between particle size and facture toughness
by using Brazilian test simulations (Fig. 8). They as-
sumed that wedge fractures occurring at the edge of
the sample could initiate a single micro fracture that
forms the rupture surface. For their example the stress
intensity factor was calculated as:
KI = α1σ√
πa (1)
At peak load, KI = KIc, σ = σt , a = βD and β is a
constant defined by the ratio of initial crack length a
and sample diameter D. Thus, the Brazilian strength
can be related to fracture toughness by:
KIc = α1σt
√
πβD (2)
0.00
0.01
0.015
0.02
0.025
0.03
0.035
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Coefficient of Friction (µ)
La
tera
l D
ilata
nt
Str
ain
(%
)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
Lateral dilatant strain
Strength Ratio
σt /σ
ci
Fig. 6: Effect of friction on dilation and the ratio of tensile
strength to compressive strength.
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
0.16
0.18
0 2 4 6 8 10 12
Bond Stiffness Ratio (kn / ks)
La
tera
l D
ilata
nt
Str
ain
(%
)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
Lat. dilatant strain
Strength Ratio
σt / σ
ci
Fig. 7: Effect of stiffness ratio on dilation and the ratio of
tensile strength to compressive strength.
and fracture toughness can be related to micro prop-
erties by:
KIc = φ√
α2πR (3)
where φ is mean bond strength, R is mean particle
radius and α1, α2 are constants.
From Equations 2and 3, the Brazilian strength can be
related to material micro properties by: and fracture
toughness can be related to micro properties, where,
σt = α3φ
√
R
D(4)
where, α3 =√
α2/α21β and indicates all constants.
Equation 4 implies that particle size is proportional to
magnitude of tensile strength. Micro properties used
for the particle size simulation analysis are shown in
Table 2. In this case the same properties are used for
both the Brazilian and compressive test simulation.
Fig. 9 shows that as particle size decreases, the tensile
to compressive strength ratio also decreases. The ten-
sile strength to compressive strength ratio for Lac du
Bonnet granite is 10/200 = 0.05. Thus to achieve this
strength ratio the particle size would have to approach
unrealistic values as the grain size of Lac du Bonnet
F
F
a
a
σσ D
F
F
a
(a) (b)
Fig. 8: Effect of friction on dilation and the ratio of tensile
strength to compressive strength.
Table 2: Micro-parameters in PFC used in particle size
effect.
Ec 62 GPa Ec 62 GPa
kn/ks 2.5 kn/ks 2.5
µ 0.5 σc 157 ± 36 MPa
σc/τc 1 λ 1
Rmin varies Rmin/Rmax 1.66
granite is approximately 2 mm. Also as particle size
gets smaller the required calculation time exponen-
tially increases. In terms of dilation, the smaller the
particle size the lower the amount of dilation (Fig. 9).
Thus reducing the tensile to compressive strength ra-
tio in this case, does not contribute to the dilation
process.
Potyondy and Cundall [15] also revealed that particle
assemblies PFC circular particles result in macroscale
failure envelope, with unrealistically low friction an-
gles, when compared to measured values. Hence there
are two deficiencies in PFC that cannot be corrected
by simply changing the microscale contact parame-
ters: the tensile strength to compressive strength ratio
and the failure envelope.
3.6 Effect of grain shape
In brittle rock modeling using the discrete element
method (DEM) analogue, researchers often use the
circular disk elements. This circular disk element has
advantages related to reducing the calculation times,
but most microscale particles that the DEM represents
are seldom circular. Jensen et al. [16] and Thomas and
Bray [17] indicated that circular disk elements are not
adequate to model geometry dependent particles such
as irregular shaped grains. One of the solutions to
overcome this limitation is to introduce a cluster con-
0.00
0.01
0.01
0.02
0.02
0.03
0.03
0.04
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Mean particle size (mm)
La
tera
l d
ilata
nt
str
ain
(%
)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
Lateral dilatant strain
Strength Ratio
σt / σ
ci
Fig. 9: Effect of particle size on dilation and the ratio of
tensile strength to compressive strength.
Crystalline rock Clustered particles
Fig. 10: Clustered particles mimicking crystalline rock.
cept. In clustered assemblies, grains are modeled as
a group of individual circular disks. The intra-cluster
bond strength can be set to different value from the
bond strength of boundary particles neighboring with
other clusters. In this study, intra-cluster bond strength
was set to infinite, thus, bond breakage can only occur
at cluster boundaries. The clustered particles with this
setting are a better representation of naturally irregu-
lar grain (Fig. 10). Moreover, this irregular boundary
modeling method may change the force chains devel-
oped in PFC.
Fig. 11 illustrates the effect of a clustered assemblage
on dilation and the strength ratio. The cluster size in
Fig. 11 represents the maximum number of circular
disks in one cluster. Slaved particle size of cluster
was determined approximately matching the cluster
diameter into previous non-clumped case. Coefficient
of friction (µ) was set to 0.0, thus friction can only be
mobilized by roughness of cluster boundaries. Based
on the previous analysis, σc/τc strength ratio was set
to 20 thus only tensile cracking is allowed. Fig. 11
clearly shows that as cluster size increases, the amount
of dilation is significantly increased and hence asso-
ciated with the irregular cluster boundary, i.e, rough-
ness. It is worth noting that this dilation process is
much more closer to the mechanical dilation describe
by [3].
0.00
0.05
0.10
0.15
0.20
0.25
0 2 4 6 8 10 12 14 16
Maximum Cluster Size
La
tera
l D
ilata
nt
Str
ain
(%
)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
Lateral dilatant strain
Strength Ratio
σt / σ
ci
Fig. 11: Effect of grain (cluster) size on dilation and the
ratio of tensile strength to compressive strength.
Also shown in Fig. 11 is reduction in the strength
ratio as cluster size and dilation increase. However,
the strength ratio shown in Fig. 11 is still high com-
pared with the 0.05 for Lac du Bonnet granite and
even seems to converge to 0.1. One possible expla-
nation for this is that rotation between particles are
not suppressed enough even in the clustered particles.
Although bond strength between intra-cluster parti-
cles has been set to infinitely high (e.g. an order of
magnitude) but particle rotation cannot be completely
suppressed and this may induce unnecessary cracking
and may reduce strength. One alternative approach to
avoid this problem is to use clumped particles instead
of clustered particles.
3.7 Clustered particles and clumped particles
The clump logic provides a means to create a group
of slaved particles and they behave as a rigid body
in a clump. Particles within a clump may overlap to
any extent as a deformable body that will not break
apart, regardless of the forces acting upon it. In this
sense, a clump differs from clustered particles. The
other big difference between clustered and clumped
particle is the particle rotation mechanism. As shown
in Fig. 12, intra-cluster particles in clustered material
have rotational velocities. Whereas rotational veloc-
ities of particles in clumped material are fixed. Only
the clumped body itself can have rotational velocities.
This mechanism is much more realistic to crystalline
rock grain behavior when fracturing occurs.
Introducing clump logic to the assemblage, dila-
tion and strength were investigated again. The re-
sults are given in Fig. 13 and clearly show as clump
size increased, dilation is significantly increased and
strength ratio is close to 0.05. It is worth noting that
the amount of dilation is almost an order of magni-
tude greater than the other cases discussed above. The
strength ratio is also closer to the values measure for
+
Clustered Particles Clumped Particles
Fig. 12: Particle rotation in cluster and clump particles.
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
2 3 4 5 6 7 8 9 10
Maximum clump size
La
tera
l d
ilata
nt
dtr
ain
(%
)
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
Lateral dilatant strain
Strength Ratio
σt / σ
ci
Fig. 13: Effect of clump on dilation and the ratio of tensile
strength to compressive strength.
Lac du Bonnet granite and many other brittle rocks.
Fig. 14 shows failed sample of clumped material.
Unlike conventional PFC biaxial test, cracks oc-
curred everywhere whereas in traditional PFC mod-
els, macro-scale shear zones always develop. In this
model large scale dilation leads to axial splitting
which is also in better agreement with laboratory ob-
servations.
3.8 Controlling factors of dilation and strength ratio
From the above discussion, it is clear that the most im-
portant factor in controlling dilation and the strength
ratio is the geometrical factors rather than micro-
contact parameters. The clustered and clumped mate-
rial all showed an order of magnitude larger amount
of dilation and lower strength ratio compared with all
other cases. Thus, it is evident that geometrical fac-
tors such as cluster or clump cases are more realistic
and effective for modeling dilation and that dilation
actually controls the strength.
PFC2D 2.00
U of A Geotechnical Engineering
Job Title: Parallel-bonded, uniaxial compression test com_cl_size11
Step 226642 03:33:02 Wed Feb 25 2004
View Size: X: -5.024e-002 <=> 5.024e-002 Y: -4.950e-002 <=> 4.950e-002
Ball
Wall
FISH function draw_sample
FISH function crk_item
Fig. 14: Sample failure in clumped assemblage.
The other interesting result is that rotation of parti-
cles in assemblage has a significant effect on mate-
rial strength. For instance, the clumped material has
more than 20 times larger compressive strength than
non-clumped material with the same micro proper-
ties. Whereas, tensile strength of the clumped mate-
rial only increased 2 to 3 times more than the non-
clumped material. Hence this kept the strength ratio
low in clumped material. In conventional PFC mod-
eling, if one adjusts the micro-parameters to match
the macro compressive strength the tensile strength
is overestimated. The clumped logic overcomes this
deficiency.
4 CALIBRATION WITH SYNTHETIC ROCK
The parametric study described above was carried out
on Lac Du Bonnet granite because it has been exten-
sively studied both in the laboratory and by other re-
searchers using PFC. In this section, we use the same
PFC logic and apply it to a brittle material that has a
peak strength that is an order of magnitude lower than
Lac du Bonnet granite.
4.1 Laboratory test properties
For the purpose of comparison and verification, con-
ventional triaxial tests and Brazilian tests were per-
formed using a synthetic rock. This synthetic rock
has brittle characteristics but a much lower compres-
sive strength. The synthetic rock chosen for this test
is ‘sulfaset’ that is generally used for setting anchor
bolt.
The strength and stiffness of our synthetic rock is
highly dependent on its initial moisture content at
mixing, and for the test reported here the initial mois-
ture content has been fixed to 50% and cured for 3
days in a constant temperature and moisture room.
To induce random heterogeneity in the sample, 10%
sand was added to all the samples. The measured sam-
ple properties are: UCS = 11.6 ± 1.0 MPa, Ten-
sile strength = 1.4 ± 0.4 MPa, and Young’s Mod-
ulus = 2.5 GPa.
Uniaxial test results for this material (75 x 150mm,
cylindrical samples) clearly showed that the material
has brittle characteristics (Fig. 15). Note that initial
nonlinearity of stress-strain curve in Fig. 15 results
from closure of pores during seating.
0
2
4
6
8
10
12
14
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6Axial Strain(%)
PFC data
Lab data
σ3= 0.0 MPa
σ1-σ
3 (
MP
a)
(a)
0
2
4
6
8
10
12
14
16
18
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0Axial Strain(%)
PFC data
Lab data
σ3= 1 MPa
σ1-σ
3 (
MP
a)
(b)
0
2
4
6
8
10
12
14
16
18
20
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5Axial strain (%)
PFC data
Lab data
σ3= 2 MPa
σ1-σ
3 (
MP
a)
(c)
0
5
10
15
20
25
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0Axial Strain(%)
PFC data
Lab data
σ3= 3 MPa
σ1-σ
3 (
MP
a)
(d)
Fig. 15: Comparison of PFC and laboratory test results for each confinement.
4.2 PFC calibration to laboratory data
The PFC micro parameters were determined from the
macro-scale laboratory results. The clump size was
set to 2 and the bond strength andYoung’s modulus is
reduced to half of the macro strength. When the ten-
sile strength bonds are broken, cohesion is lost and
the strength mobilized is composed of clump friction
and particle contact friction. The coefficient of friction
has been calibrated to 0.7 and this friction can be mo-
bilized when shear cracking occurs and as confining
stress is increased.
Fig. 15 compares the biaxial test simulation results
from PFC with laboratory test results by using the
micro-scale parameters in Table 3. When compared
with the laboratory results, the initial nonlinearity is
not present which is expected because in PFC no
cracks exist prior to loading. If a random distribution
of pores or cracks were used the nonlinearity mea-
sured would also be recorded in PFC.
Fig. 15 also shows in all cases, the agreement be-
tween PFC and laboratory results is excellent. It is
interesting to note that no flow rule is required for
modeling the stress-strain behavior in post peak re-
gion in PFC. The failure envelope in Fig. 16 shows
that PFC has a slightly higher slope compared with
the measured laboratory envelope. This might be be-
cause of the two dimensional assumption defined in
PFC. Nonetheless, the agreement with the laboratory
envelope is very good. More importantly the agree-
ment with the tensile and compressive strength is also
excellent, e.g. compare the results from [12] in Fig. 4.
The shape of the failure of zone in PFC shows the typ-
ical macro shear fractures observed in the laboratory
test (Fig. 14).
Table 3: Micro parameters in PFC used to represent the
Sulfaset synthetic rock.
Ec ( GPa) 1.3 Ec (GPa) 1.3
kn/ks 2.7 kn/ks 2.5
µ 0.7 σc (MPa) 4.5 ± 0.1
σc/τc 2 λ 1
Rmin (mm) 0.25 Rmin/Rmax 1.2
Density (kg/m3) 2500 Max clump 2
5 DILATION IN LABORATORY TESTS
Numerous researchers [18–21] have reported that the
onset of dilation in uniaxial tests initiates at about 1/4
to 1/2 of the peak strength. In most cases the dila-
tion is associated with the initiation of axially aligned
microcracks. Scholz [22], Holcomb and Martin [23],
Pestman and Van Munster [24] studied the crack initi-
ation of granite, sandstone, and marble using acoustic
emission and they showed that the onset of dilation
could be approximated by:
σ1 = 0.4σci + 1.5 to 2σ3 (5)
A series of PFC simulations were carried out to deter-
mine this dilational boundary in PFC. Fig. 17 shows
the onset of dilation in a typical PFC model for 3 MPa
confining stress. Fig. 18 shows the dilational boundary
for all the calibrated PFC models and this boundary
can be approximated by:
σ1 = 0.38σci + 2.1σ3 (6)
Note that the measured dilational boundary measured
in the laboratory specimens for various rocks and
given by Equation 5 is in good agreement with the
dilational boundary determine from PFC for the syn-
thetic rock and given by Equation 6.
Currently the calibration between real materials and
PFC is usually made via the traditional laboratory
compression tests. In order to know if PFC can be
used to model engineering problems such as slopes
and tunnels more complex loading paths are required.
Fig. 19 shows bending of a rock beam confined at the
5
10
15
20
25
30
-2.0 -1.0 0.0 1.0 2.0 3.0 4.0σ3 (MPa)
σ1 (MPa)
PFC best fitLab data best fitPFC dataLab data
c = 2.7 MPa, φ=40o
c = 2.7 MPa, φ=37o
Fig. 16: Strength envelope from PFC compared with en-
velope from laboratory test results.
σ3 = 3.0MPa-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0 0.2 0.4 0.6 0.8 1 1.2
Axial Strain(%)
V /
V (
%)
3
8
13
18
23
28
Dilation Volumetric Strain
Total Volumetric Strain
Axial Stress Strain
Onset of Dilation
σ1 (M
Pa)
Fig. 17: Onset of dilation in PFC for a sample confined by
3 MPa.
ends. This complex loading path also results in ax-
ial fractures at the top of the beam. Proof testing has
been completed and samples are being prepared for
the final testing. During the testing extensive monitor-
ing will be used to measure the dilation. The dilation
obtained from these tests will then be compared with
PFC results.
6 CONCLUSIONS
Modeling of brittle failure has been attempted by vari-
ous approaches, i.e., continuum and fracture mechan-
ics, over the past 50 years. In nearly all cases these
approaches have provided poor agreement with ob-
servations of brittle failure in the field. In laboratory
tests calibration is often made with the peak strength
yet the onset of dilation in uniaxial compression test
is observed far below the peak strength.
During the past ten years discrete elements, particu-
5
10
15
20
25
-2 -1 0 1 2 3 4σ3 (MPa)
σ1 (MPa)
Peak Strength
Dilational boundaryσ1= 0.38 σci + 2.1 σ3
Fig. 18: The dilational boundary in the PFC models of
synthetic rock.
Frame
Frame
Ball
Ball
Beam
Wiremesh
Roller
Notch
Loading plateHydraulic ram
Fig. 19: Test equipment for measuring dilation induced by
extensile fracture.
larly the software PFC, have been used to model brit-
tle failure. Three significant deficiencies have been
identified when doing conventional PFC modeling:
(1) the tensile strength to compressive strength ratio is
considerably greater than that measured in laboratory
tests, (2) the failure envelop gives very low friction
angles compared to measured laboratory values, and
(3) the onset of dilation occurs much closer to the peak
strength compared to measured values. Adjusting the
micro parameters appears to have little effect on these
deficiencies. The findings from this research indicate
that by introducing clumped-particle geometry these
deficiencies are minimized or eliminated.
Using the clumped logic excellent agreement is found
between laboratory tests results and PFC for the Sul-
faset synthetic rock. Testing of more complex stress
paths and different rock types is continuing before
applying this logic to engineering problems such as
tunnels and slopes.
ACKNOWLEDGEMENTS
This work was supported by the Natural Sciences
and Engineering Research Council of Canada and the
Swedish Nuclear Fuel and Waste Management Co.
REFERENCES
[1] Martin, C. D., R. S. Read, and J. B. Martino, 1997. Obser-
vations of brittle failure around a circular test tunnel. In-
ternational Journal Rock Mechanics And Mining Science
34(7):1065–1073.
[2] Terzaghi, K., 1946. Rock defects and loads on tunnel sup-
ports. In Rock Tunneling with Steel Supports, R. V. Proctor
and T. L. White, eds., chap. 1, 5–99. Youngstown, Ohio:
Commercial Shearing, Inc.
[3] Lajtai, E. Z., 1998. Microscopic fracture processes
in a granite. Rock Mechanics and Rock Engineering
31(4):237–250.
[4] Ashby, M. F. and S. D. Hallam, 1986. The failure of brittle
solids containing small cracks under compressive stress.
Acta. Metall 34:497–510.
[5] Detournay, E. and C. M. St. John, 1988. Design charts for
a deep circular tunnel under non-uniform loading. Rock
Mechanics and Rock Engineering 21(2):119–137.
[6] Germanovich, L. N. and A. V. Dyskin, 1988. A model of
brittle failure for material with cracks in uniaxial loading.
Mechanics of Solids 23(2):111–123.
[7] Kemeny, J. M. and N. G. W. Cook, 1987. Crack models
for the failure of rock under compression. In Proc. 2nd Int.
Conference Constitutive Laws for Engineering Materials,
Theory and Applications, Tucson, Arizona, C. S. Desai,
E. Krempl, P. D. Kiousis, and T. Kundu, eds., vol. 1, 879–
887. Elsevier Science Publishing Co.
[8] Sammis, C. G. and M. F.Ashby, 1986. The failure of brittle
porous solids under compressive stress states. Acta metall.
34:511–526.
[9] Cook, N. G. W., 1995. Müller Lecture: Why Rock Me-
chanics. In Proc. 8th, ISRM Congress on Rock Mechan-
ics, Tokyo, T. Fujii, ed., vol. 3, 975–994. Rotterdam: A.A.
Balkema.
[10] Cundall, P. A., 2001. A discontinuous future for numerical
modelling in geomechanics. Geotechnical Engineering
149(1):41–47.
[11] Martin, C. D. and N. A. Chandler, 1994. The progressive
fracture of Lac du Bonnet granite. International Jour-
nal Rock Mechanics Mining Science & Geomechanics Ab-
stracts 31(6):643–659.
[12] Diederichs, M. S., 1999. Instability of Hard Rockmasses:
The Role of Tensile Damage and Relaxation. Ph.D. thesis,
Dept. of Civil Engineering, University of Waterloo, Water-
loo, Canada.
[13] Hoek, E. and E. T. Brown, 1997. Practical estimates of rock
mass strength. International Journal Rock Mechanics And
Mining Science 34(8):1165–1186.
[14] Potyondy, D. O. and P. A. Cundall, 1998. Modeling notch-
formation mechanisms in the URL Mine-by Test Tunnel
using bonded assemblies of circular particles. Interna-
tional Journal Rock Mechanics Mining Science & Geome-
chanics Abstracts 35(4-5):510–511. Paper:067.
[15] Potyondy, D. and P. Cundall, 2002. Particle Flow Code
(PFC2D/3D) training course, Gelsenkirchen, Germany.
[16] Jensen, R. P., R. P. Jensen, P. J. Bosscher, M. E. Plesha,
and T. B. Edil, 1999. Dem simulation of granular me-
dia - structure interface: effects of surface roughness and
particle shape. International Journal for Numerical and
Analytical Methods in Geomechanics 23(6):531–547.
[17] Thomas, P. A. and J. D. Bray, 1999. Capturing nonspher-
ical shape of granular media with disk clusters. Jour-
nal of Geotechnical and Geoenvironmental Engineering
125:169–178.
[18] Fairhurst, C. and N. G. W. Cook, 1966. The phenomenon
of rock splitting parallel to the direction of maximum com-
pression in the neighbourhood of a surface. In Proc. of the
1st Congress of the International Society of Rock Mechan-
ics, Lisbon, 687–692.
[19] Hallbauer, D. K., H. Wagner, and N. G. W. Cook, 1973.
Some observations concerning the microscopic and me-
chanical behaviour of quartzite specimens in stiff, triaxial
compression tests. International Journal Rock Mechanics
Mining Science & Geomechanics Abstracts 10:713–726.
[20] Tapponnier, P. and W. F. Brace, 1976. Development of
stress-induced microcracks in Westerly granite. Interna-
tional Journal Rock Mechanics Mining Science & Geome-
chanics Abstracts 13:103–112.
[21] Martin, C. D., 1997. Seventeenth Canadian Geotechnical
Colloquium: The effect of cohesion loss and stress path
on brittle rock strength. Canadian Geotechnical Journal
34(5):698–725.
[22] Scholz, C. H., 1968. Microfracturing and the inelastic
deformation of rock in compression. Journal Geophysical
Research 73(4):1417–1432.
[23] Holcomb, D. J. and R. J. Martin, 1985. Determining peak
stress history using acoustic emissions. In Proc. 26th U.S.
Symp. on Rock Mechanics, Rapid City, E. Ashworth, ed.,
vol. 1, 715–722. A.A. Balkema, Rotterdam.
[24] Pestman, B. J. and J. G. Van Munster, 1996. An acoustic
emission study of damage development and stress-memory
effects in sandstone. International Journal Rock Mechan-
ics Mining Science & Geomechanics Abstracts 33(6):585–
593.