Computers and Geotechnics 74 (2016) 36–44

Contents lists available at ScienceDirect

Computers and Geotechnics

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

Research Paper

Physical and numerical study of strain burst of mine pillars

Ali Fakhimi ⇑, Omid Hosseini, Roosevelt Theodore
Department of Mineral Engineering, New Mexico Institute of Mining and Technology, Socorro, NM 87801, USA

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

Article history: Strain burst is a spontaneous dynamic failure of rock that can cause serious injury to the miners and dam-
Received 11 September 2015 age to the underground excavations. To simulate the strain burst in the lab, a steel beam was designed
Received in revised form 21 November 2015 and connected to the compression loading machine. The beam acts as an energy absorber and is in direct
Accepted 29 December 2015
contact with the rock specimen which is under uniaxial compression loading. Upon failure of the speci-
Available online 11 January 2016
men, the absorbed energy in the beam is transferred to the rock specimen to simulate the strain burst in
underground pillars. Based on the physical tests, rock fragment velocities of more than 4 m/s were mea-
Keywords:
sured using a high speed camera. The interaction between the steel beam and the rock was modeled
Strain burst
Kinetic energy
using a hybrid discrete-finite element computer program. The effect of different parameters such as pil-
Physical testing lar’s length and diameter, friction coefficient between pillar and roof, compressive strength of pillar, rock
Scaling post-peak behavior, roof stiffness, and pillar and roof rock densities on the intensity of the strain burst
Numerical modeling were studied. The strain burst intensity was defined as the kinetic energy of the simulated rock.
Rock burst Dimensional analysis was applied to find relationships between the dimensionless parameters in the
numerical simulation. The proposed scaling model together with the numerical analysis appears to be
able to show the significance of different parameters involved in the strain burst. In particular, it is shown
that the pillar diameter and its uniaxial compressive strength have significant impacts on the induced
kinetic energy during a strain burst.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction
Rock burst is the dynamic failure of rock that poses serious
threat to the underground activities. This is particularly the case
in deep underground mining in which high in situ stresses and
brittle rocks are involved. Ortlepp and Stacey [1] based on source
mechanisms, categorized the field rock bursts as: strain bursting,
buckling, face crushing, virgin shear in rock mass and reactivated
shear on existing faults or discontinuities. They described the
strain burst that is involved with violent ejection of sharp rock
fragments as the most common damage mechanism observed in
underground excavations. During a rock burst, rock particles can
be ejected with a velocity of 8–50 m/s [2] which can cause fatal
injuries and damages to the underground equipment. Stacey
et al. [3] reported that during a rock burst, the thickness of the
ejected rock can be in the order of 1 m and hence supports for
the rock must be capable of absorbing the rock kinetic energy. Rock
burst had been known in mining since the 18th century; however
it remained essentially a subject of qualitative study [4]. Cook [5,6]
discovered the fundamental requirement for violent fracture of
⇑ Corresponding author.
E-mail address: hamed@nmt.edu (A. Fakhimi).
rock (rock burst); energy release rate plays a critical role in rock
burst. Based on Cook’s studies, rock burst occurs when an excess
energy becomes available during the post peak deformation stage
of rock. One of the first attempts to model rock burst in a room
and pillar mining system was proposed by Salamon [7] who used
the stiffness matrix (K) of the mining layout together with the
slope matrix (A) of the complete load convergence relations of pil-
lars to predict the stability of the rock structure. He showed that
the stable situation is achieved if the system matrix K + A is posi-
tive definite. Petukhov and Linkov [8] considered the interaction
between a linear elastic rock mass with a softening material and
used some energy equations to introduce a criterion for the stabil-
ity of the system. Burgert and Lippmann [9] studied models of
translatory rock bursting in an idealized coal seam. A model mate-
rial and a model technique were suggested. They divided the coal
ahead of tunnel face to three regions: (1) Passive elastic zone, (2)
active elastic zone and (3) active plastic zone. Their model was able
to explain some of the observations in the field. Zubelewicz and
Mroz [10] considered rock burst phenomenon as a dynamic
instability problem. In their approach, a dynamic perturbation
was superimposed into the static solution of the problem and then
the possibility of kinetic energy growth of the system as an
indication of rock burst was investigated.

http://dx.doi.org/10.1016/j.compgeo.2015.12.018
0266-352X/Ó 2016 Elsevier Ltd. All rights reserved.
 A. Fakhimi et al. / Computers and Geotechnics 74 (2016) 36–44 37

Rock burst has been considered as a problem of surface instabil- approach was used to define some dimensionless parameters that
ity by some researchers. Biot [11] performed the pioneering work play important roles in strain bursting of the rock.
on surface instability of a half space. Vardoulakis [12] used the
bifurcation theory to analyze the rock burst as a surface instability
2. Experimental study
phenomenon. Bardet [13] used a finite element formulation to
study rock burst as a surface instability problem. The bifurcation
A soft steel loading frame was designed and attached to the MTS
of the solution was detected by evaluating the eigenvalues of the
loading machine to absorb and store some strain energy that can
tangential stiffness matrix. Whyatt and Loken [14] utilized the
be released when the post-peak regime of the rock is approached.
boundary element method to simulate sudden dislocation along
The steel frame which is made of two W5
16 steel profiles with
a fault plane. Their model was able to explain some of the odd
the yield strength of about 345 MPa is shown in Fig. 1. The frame is
dynamic phenomena that are observed during the rock bursts.
to represent the roof structure in a room and pillar mining system.
Some researchers have investigated the phenomenon of rock
The top of the frame is bolted to the loading machine cross-head
burst experimentally. A double rock sample model was studied
and the rock specimen is accommodated between the platen con-
both physically and numerically by Chen et al. [15]. The physical
nected to the bottom of the frame structure and the machine actu-
sample was made of granite and marble specimens of cylindrical
ator. Experimental tests were conducted on the frame structure to
shape and was tested in uniaxial compression. From the results
obtain its stiffness. Fig. 2 shows the load–deflection curve for the
of the tests, they concluded that it may be possible to predict the
frame structure which suggests that the stiffness of the structure
occurrence of rock burst when a sudden decrease of micro-
is about 14.1 kN/mm.
seismic rate occurs in one zone while the micro-seismic rate
The experimental tests were conducted on the Pennsylvania
continues to increase in an adjacent zone. A true triaxial rock test
blue sandstone which has the following average mechanical
system was used by He et al. [16] to study the rock burst. The lime-
properties: Elastic modulus = 26.3 GPa, Poisson’s ratio = 0.15,
stone specimen, 15
6
3 cm in dimension, was loaded initially
uniaxial compressive strength = 110.9 MPa, and Brazilian tensile
in three mutually perpendicular directions and then abrupt
strength = 9.9 MPa. Rock specimens 25 mm in diameter and
unloading of the minimum principal stress in one loading face
68 mm in length were used for strain burst testing. The tests were
was performed, creating a stress state and boundary conditions
conducted under stroke control. Each test took about 20–30 min to
in the rock sample relatively similar to those that exist at a tunnel
finish; the applied displacement rate was 0.0025 mm/s. The results
face. The physical tests by these authors showed the ejection of
of the uniaxial compressive tests using the frame structure are
rock fragments from the unloaded surface of rock that was inter-
reported in Table 1. All tests were finished by violent failure of rock
preted as rock burst. Kuch et al. [17] simulated coal mine bumps
specimens. A high speed camera was used to record the bursting
using a model material. Their experimental investigation showed
event (Fig. 3a). During the strain burst of the rock specimens, the
that the scatter in the critical rock stress, necessary for bump initi-
velocities of some rock fragments flying in the camera plane were
ation, is due to variations of the post-peak stiffness of the material.
In this paper, strain burst in pillars was studied by using a soft
loading system. Strain burst is a type of rock burst that is associ- 120
ated with gradual accumulation of strain in rock. Uniaxial com- Loading y = 14.08x
R² = 0.9955
pression tests were conducted on sandstone specimens using this 100 Unloading
loading system to allow for accumulation of the strain energy
80
and sudden release of this energy when the rock approached to
Load (kN)

its post-peak regime. A high speed camera was used to measure 60

the peak particle velocity. In addition, a numerical model was
utilized for simulation of the rock burst. The rock was represented 40
by a bonded discrete element domain while the loading system
20
was modeled as a linear elastic body. Different parameters that
are involved in the induced dynamic rock fracture and strain burst- 0
ing such as the loading system stiffness, the rock strength, the 0 2 4 6 8
pillar dimensions, and the rock-loading system interface friction Deformation (mm)
coefficient were investigated in this study. Furthermore, a scaling
Fig. 2. Load–deflection data for the frame structure.

Fig. 1. (a) The frame structure and (b) its dimensions (in mm).
38 A. Fakhimi et al. / Computers and Geotechnics 74 (2016) 36–44

Table 1 5.0
The uniaxial compressive strength of sandstone specimens using the soft loading
frame.

Particle Velocity (m/s)

Specimen number Uniaxial compressive strength (MPa)
4.0

1 110.7
2 121.1 3.0
3 114.2
4 116.0
5 97.6
2.0
6 117.4
7 104.6
8 105.5
1.0
Average 110.9 0.0 0.1 0.2
Standard deviation 7.8
Particle Mass (g)

Fig. 4. Measured rock fragments velocities vs. their masses.

calculated with the use of Adobe Photoshop and MATLAB. An Exi-
lim Casio camera capable of video recording at a rate of 1000
frames per second was used in the recording and once the video
14.1 kN/mm, which is consistent with the stiffness of the physical
of the test was complete, each frame was extracted and individu-
frame structure (Fig. 2).
ally saved in a file. The time difference between two consecutive
The bonded particle model for the rock was calibrated using the
frames was 2.083 ms; the video was recorded at a rate of 480
technics introduced by Fakhimi in [18] and [19]. As a result, the
frames per second to obtain a better resolution. The sandstone
micromechanical parameters for the discrete system were
fragments identified to be approximately within the camera plane
obtained: normal spring contact stiffness (Kn) = 2.2
107 N/m,
were singled out in the Adobe Photoshop by whiting out the pixel
shear spring contact stiffness (Ks) = 8.0
106 N/m, normal bond
area of the fragment area and then the background was erased
(nb) = 5.1 N, shear bond (sb) = 27.1 N, friction coefficient (l) = 0.5
with a black eraser (Fig. 3b). This made it easier for the MATLAB
and genesis pressure = 5.9 GPa. As explained in [19], the genesis
program to find the area and the x–y location of the center of mass
pressure is the initial isotropic stress applied to the surrounding
of that particular fragment. Hence by knowing the x–y values of a
walls of the specimen to induce small overlap of the particles in
fragment center on two different frames, and the elapsed time
contact. As a consequence of this small overlap, a more realistic
between the two, the fragment velocity was found. The mass of
bonded particle model can be generated for rock. The spherical
each particle was estimated assuming a spherical shape for the
particles were assumed to have a radius within the range 0.4–
particle and its known density. Fig. 4 shows the measured velocity
0.8 mm with the average radius Rave = 0.6 mm. Using these
of some rock fragments versus the mass of the particles. As Fig. 4
micromechanical parameters, a uniaxial compression test was con-
indicates, high particles velocities of 0.5–4.3 m/s were measured
ducted on a cylindrical specimen 25 mm in diameter and 68 mm in
in our experiments which are typical particle velocities during an
height. The simulated model resulted in the elastic and strength
actual field rock burst.
parameters that are in good agreement with those for the sand-
stone (Table 2).
3. Numerical modeling of strain burst The calibrated bonded particle model for the sandstone which is
25 mm in diameter and 68 mm in height was loaded through the
The computer program CA3 [18] which is a hybrid discrete- frame structure (Fig. 5a). 51,800 spherical particles were used in
finite element code for 3D simulation of geo-materials was used the specimen. To expedite the computational process, the speci-
for the numerical modeling of the rock burst. The sandstone was men was loaded initially by applying a uniform vertical traction
simulated as a bonded particle system which is made of spherical to the lower loading platen. The applied forces versus the numer-
particles that are glued at the contact points to withstand the ical cycles for specimens with and without end friction are shown
applied stresses. The loading frame was modeled as a linear elastic in Fig. 6. The intensity of the traction was such that upon equilib-
material using finite elements. The discretized frame together with rium of the structure, the specimen was subjected to about 70% of
the simulated rock specimen is shown in Fig. 5a. A numerical test its compressive strength (path AB in Fig. 6). Subsequently, a verti-
conducted on the simulated beam confirmed that its stiffness is cal velocity of 0.03 m/s (with a time step of 4.2
108 s or

(a) (b)

RF1
RF1

RF2
RF2

Fig. 3. (a) Photo of a rock specimen during the strain bursting together with two studied rock fragments, (b) the singled out rock fragments introduced to MATLAB to measure
their velocities.
 A. Fakhimi et al. / Computers and Geotechnics 74 (2016) 36–44 39

(a) (b) (c)

Simulated bonded particle specimen μ' = 0 μ' = 0.18

Fig. 5. (a) Finite element model of the frame structure together with the bonded particle rock specimen, and the damaged specimen after the strain burst assuming the
interfaces between loading platens and specimen ends to have a friction coefficient (l0 ) of (b) zero, (c) 0.18.

Table 2
Comparison between the elastic and strength parameters of the sandstone and the simulated material.

Elastic modulus (E) GPa Poisson’s ratio (m) Uniaxial compressive strength (qu) (MPa) Tensile strength (rt) (MPa)
Sandstone 26.3 0.15 110.9 9.9
Simulated model 26.3 0.14 111 10.7

70.0
μ´=0.0

60.0
μ´=0.18 C'
C
50.0
B
Axial Force (kN)

40.0

30.0

20.0

10.0 D D'
A
0.0
0.03 0.23 0.43 0.63 0.83 1.03 1.23 1.43 1.63 1.83
Numerical Cycles (× 106)

Fig. 6. Axial force vs. numerical cycles for two specimens 25
68 mm2 in dimension with and without end friction. The friction coefficient is shown with l0 .

0.12
108 m/cycle) was applied to the bottom of loading platen
(paths BC and BC0 ). During these two stages of loading, a quasi-
static solution was achieved by using a damping coefficient of
0.7. The damping force in CA3 program is obtained by multiplying
the unbalanced force for each finite element node or each discrete
particle by the damping coefficient [20]. When the applied stress
was about 90% of the rock compressive strength, the damping coef-
ficient was reduced to 10% to allow a dynamic simulation of the
rock-loading frame structure at failure point of the specimen.
Fig. 5b and c shows two damaged specimens following the strain
burst. In Fig. 5b, no friction was assumed for the interface between
the loading platens and the specimen ends. On the other hand, a
friction coefficient of 0.18 [21] was used in Fig. 5c. Note how the
introduced friction coefficient and the end frictional forces have
protected the specimen ends; only the middle part of the specimen
is subjected to the bursting. The maximum particle velocities
reported by the CA3 program for Fig. 5b and c are 11.4 and
9.8 m/s, respectively. These maximum particle velocities are
greater than those reported in Fig. 4 based on the actual physical
measurement. This discrepancy between the numerically calcu-
lated peak particle velocity and the physical values should be
expected; the numerical model does not incorporates a rate depen-
dent micro-mechanical model and in the physical measurement,
the particle with greatest velocity is hard to find and it may not
be exactly within the plane of the camera. Nevertheless, the phys-
ical and numerical particle velocities are within the range observed
in actual rock bursts.
The velocity of the platen of the steel beam which is in direct
contact with the top of the specimen is shown in Fig. 7a. The veloc-
ities for situations with and without specimen end friction are
shown. Note that the specimen with end friction fails at an axial
compression of qu = 121 MPa compared to qu = 109 MPa when the
end friction is zero. This increase in rock strength results in greater
stored strain energy in the beam and greater beam velocity. As a
consequence, the maximum kinetic energy of the rock burst is
expected to be greater in the beam with end friction (Fig. 7b). This
is not the case as the friction at the specimen ends absorbs part of
the beam energy. Furthermore, the ends frictional forces which
support the specimen ends put restriction on the kinematics of
the damaged specimen (Fig. 5). Consequently, the numerical model
suggests that the rock burst intensity is reduced if the friction coef-
ficient between the specimen ends and the loading platens is not
zero.
4. Dimensional analysis
The kinetic energy of a failing pillar (Q) due to strain burst can
be considered as a function of the following parameters:
40 A. Fakhimi et al. / Computers and Geotechnics 74 (2016) 36–44

2.0
μ' = 0 0.2
1.5 μ' = 0.18 (a) μ' = 0.18 (b)
0.16 μ' = 0
1.0
Beam Velocity (m/s)

Kinetic Energy (J)

0.5
0.12
0.0
8 10 12 14 16 18 0.08
-0.5

-1.0
0.04
-1.5
0
-2.0
5 7.5 10 12.5 15 17.5
Numerical Cycle × 105 Numerical Cycle × 105

Fig. 7. (a) Top platen vertical velocity and (b) kinetic energy of the bonded particle specimen vs. numerical cycles for specimen ends friction coefficient of zero and 0.18.

In which a, b, h are the dimensions of the roof (Fig. 8) and Er is the

elastic modulus of the roof material. Therefore, the last term in Eq.
(3) can be written as:
D
Tributary area of
pillar qu D4 qu D4
h ¼ 4
ð5Þ
Kr Vr Er h
b Eqs. (4) and (5) indicate that the thickness of the roof has a great
a impact on the roof stiffness and consequently the intensity of the
rock burst. Note that the roof thickness in Fig. 8 is not necessarily
L the whole thickness of the overburden material in a mine; a joint
that separates the immediate roof from the overburden material
Pillar can dictate and define the effective roof thickness in our analysis.
Based on Eq. (3), parametric studies were conducted using the
beam-specimen model shown in Fig. 5a and the results are dis-
cussed in the next sections of the paper.
Fig. 8. A pillar and its tributary roof volume.
5. Rock pillar brittleness
Q ¼ f 1 ðqp ; qr ; K r ; V r ; b; l0 ; L; D; qu Þ ð1Þ

In Eq. (1), qp, qr, Kr, Vr, b, l0 , L, D, qu are pillar density, roof density, The rock pillar brittleness (b) in Eq. (3) is defined in the dimen-
roof stiffness, tributary roof volume (equal to abh in Fig. 8), pillar sionless form as the ratio of the Kp to Kr. The Kp parameter is the
brittleness, friction coefficient between pillar and roof or floor, pillar post-peak slope of the force–displacement curve of the pillar.
length, pillar diameter, and uniaxial compressive strength of pillar, To vary the rock pillar brittleness, the micro-mechanical param-
respectively. In Eq. (1), the elastic properties of the pillar are not eters (such as normal and shear bonds and genesis pressure) in the
explicitly considered. In particular, the explicit effect of pillar elastic bonded particle system were modified. However, for the simple
modulus was ignored assuming that this parameter is approxi- contact bond model used in this paper, it was realized that the only
mately proportional to the uniaxial compressive strength of the pil- parameter that has noticeable impact on the Kp value is the friction
lar; in general as the elastic modulus of rock is increased, its coefficient between the particles (l). The effect of l on the post-
uniaxial compressive strength is increased as well. Considering peak behavior of the simulated rock specimen with D = 25 mm
the 3 independent dimensions of force, mass and length, and the and L = 68 mm is shown in Fig. 9. Notice that as expected by
10 involved parameters, the following 6 dimensionless parameters increasing the internal friction coefficient, more ductile behavior
control the dimensionless kinetic energy of the pillar upon bursting: is realized and more energy is needed for complete failure of the
! specimen. It is interesting to note that the aspect ratio (L/D) of
Q L qp Dq D3
¼ f2 ; ; b; l0 ; u ; ð2Þ the pillar has a negligible effect on the Kp value; Kp/Kr (for
qu LD2 D qr Kr V r Kr = 14.08 kN/mm) varies from 117 to 135 when L/D changes
from 0.5 to 8 for l = 0.5. This finding is consistent with the obser-
In the next sections of the paper and in [22], it has been shown
vations in the previous work [24,25].
that for the situation considered in this paper in which the mass of
the beam in Fig. 5a (or the mass of the roof) is much larger than the
mass of the specimen (or the mass of the pillar), the last two 6. Numerical tests results
parameters in Eq. (2) can be combined such that:
! To verify the appropriateness of Eq. (3) in describing the
Q L qp q D4 induced kinetic energy in the specimen due to the strain burst, sev-
¼ f2 ; ; b; l0 ; u ð3Þ
qu LD2 D qr Kr V r eral numerical tests were conducted. Fig. 10a shows the variation
of dimensionless kinetic energy (the expression on the left side
Note that based on the theory of linear elastic plates [23], the of Eq. (3)) vs. the L/D of the specimen. In these tests, only the spec-
roof stiffness in Fig. 8 can be written as: imen length was modified and a friction coefficient of l0 = 0.18 was
3 used for the interface of the specimen ends and the loading pla-
Er h
Kr  ð4Þ tens. Note that the dimensionless kinetic energy remained almost
ab
unaffected by changes in the specimen length (within L/D = 1–8)
 A. Fakhimi et al. / Computers and Geotechnics 74 (2016) 36–44 41

(a) -120
(b)
60.0

50.0 -100
Axial Force (kN)
40.0 -80

Kp/Kr
30.0 -60
y = 89.98x - 148.21
20.0 -40 R² = 0.99

10.0 -20

0.0 0
0.00 0.10 0.20 0.30 0.40 0.50 0 0.5 1 1.5 2
Axial Displacement (mm) μ

Fig. 9. The effect of micro-mechanical internal friction coefficient (l) on the post-peak slope (Kp) of the simulated rock specimen. (a) Force–displacement curves, (b)
normalized post-peak slope vs internal friction coefficient (l). The results are shown for D = 25 mm, L = 68 mm, and Kr = 14.08 kN/mm.

200
(a) ρp is variable, ρr is 7800 kg/m3 (b)
100 ρp is variable, ρr is 2600 kg/m3
160 ρp is 2600 kg/m3, ρr is variable

/(qu.L.D2) (×10-6)
/(qu.L.D2) (×10-6)

80
120
60 y = 36.06x + 14.56
R² = 0.98
80
40

40
20

0 0
0 2 4 6 8 10 0 1 2 3 4 5
L/D ρp/ρr

Fig. 10. Dimensionless kinetic energy vs. (a) pillar (specimen) aspect ratio for qp = 2600 kg/m3, and qr = 7800 kg/m3, (b) ratio of pillar (specimen) density to roof (beam)
density with L = 68 mm. For these tests, D = 25 mm, Kr = 14.08 kN/mm, Kp = 1520 kN/mm and l0 = 0.18 were used.

and only when the specimen aspect ratio is less than 1, it showed in the relatively identical qu values for these specimens. This strat-
some sensitivity to the L/D parameter. This should be expected as egy helped to single out the effect the Kp on the kinetic energy.
for low L/D values, the uniaxial compressive strength of the speci- In Fig. 11b, the variation of the dimensionless kinetic energy
men increases and that causes more energy to be stored and deliv- with the friction coefficient between the specimen ends and the
ered to the specimen at the time of strain burst. The important loading platens is shown. Note that the dimensionless kinetic
suggestion by Fig. 10a is that the kinetic energy of the pillar is energy is not affected by the friction coefficient unless the friction
approximately a linear function of the pillar length for the L/D val- coefficient is close to zero. The role of the specimen ends friction
ues within the range of 1–8 studied in this paper; longer pillars are coefficient is to confine the specimen ends which results in greater
subjected to more violent strain bursting. Note that the increase in material confinement at its ends and less induced kinetic energy.
the intensity of the strain burst due to increase in the specimen Fig. 12 shows how the dimensionless kinetic energy is affected
length is due to the freedom of a longer section (far away from by the 5th dimensionless parameter in Eq. (3). Notice that different
the loading platens) of the specimen to the lateral movement symbols have been used in the figure to indicate the variation in
and that a longer pillar can absorb greater elastic energy before different parameters. From Fig. 12, it appears that a linear equation
failure. Nevertheless, for very large pillar lengths, the dimension- can approximately fit the data points.
less energy is expected to approach zero as the supplied energy
from the roof is a finite value. 7. A simplified analytical model and discussion of the results
Fig. 10b shows the variation of dimensionless kinetic energy
with the ratio of pillar density to roof density. Note that irrespec- Consider the frame structure in Fig. 5a which is under the load P
tive of the change in the pillar or roof density, the data points fol- applied at its mid span through the rock specimen. Sudden
low a relatively linear trend. From this figure, it is concluded that removal of the applied load due to spontaneous failure of the rock
reduction in the roof mass, causes a more violent bursting of the causes the strain energy in the linear elastic beam to transform
pillar as the stored energy in the roof or beam can result in greater into kinetic energy, i.e.
initial velocity when a lighter material is involved; this observation
1 1 P2
supports the simplified energy approach in predicting the kinetic mr v 2r ¼ ð6Þ
energy of the pillar burst as discussed in the next section. 2 2 Kr
In Fig. 11a, the effect of the variation of Kp/Kr (for a fixed In Eq. (6), Kr is the beam (or roof) stiffness, vr is the beam velocity at
Kr = 14.08 kN/mm) on the dimensionless kinetic energy is shown. its mid span, and mr is part of the beam mass that can be considered
It appears that there is a linear relationship between the dimen- as a lumped mass in Eq. (6) which is about one half of the whole
sionless kinetic energy and Kp/Kr parameter. Note that as discussed beam mass. By substitution for P with respect to the uniaxial com-
in the Section 5, in playing with the Kp parameter, the microme- pressive strength of rock (qu), we have
chanical parameter l was varied which resulted in small changes
in the uniaxial compressive strength of the specimens. The normal p2 q2u D4
v 2r ¼ ð7Þ
and shear bond between particles were slightly adjusted to result 16K r mr
42 A. Fakhimi et al. / Computers and Geotechnics 74 (2016) 36–44

30 50
(a) (b)
25

/(qu.L.D2) (×10-6)
40
/(qu.L.D2) (×10-6)
y = -0.16x + 5.68
20 R² = 0.96
30
15
20
10
10
5

0 0
0 -20 -40 -60 -80 -100 -120 0 0.2 0.4 0.6 0.8 1 1.2
Kp/Kr μ'

Fig. 11. Dimensionless kinetic energy vs. (a) the ratio of post-peak pillar stiffness to the roof stiffness with l0 = 0.18., (b) the friction coefficient between the pillar and the roof
with Kp = 1520 kN/mm. In these tests, L = 68 mm, D = 25 mm, Kr = 14.08 kN/mm, qp = 2600 kg/m3, and qr = 7800 kg/m3 were used.

350
qu is variable
D is variable
300 L is variable y = 5E+06x + 10.749
R² = 0.9937

250
/(qu.L.D2) (×10 6)

200

150

100

50

0
0 0.00001 0.00002 0.00003 0.00004 0.00005 0.00006
(qu.D4)/(Kr.Vr)

Fig. 12. Dimensionless kinetic energy vs. the 5th parameter in Eq. (3). The involved parameters including D (within the range of 12.5–50 mm), L (within the range of 17–
204 mm), qu (within the range of 28–169 MPa) were allowed to vary. In these tests, Kr = 14.08 kN/mm, Kp = 1520 kN/mm, Vr = 1 m3, qp = 2600 kg/m3, qr = 7800 kg/m3 and
l0 = 0.18 were used. L/D is within the range of 1.36–8.16 in this figure.

If the mass of the specimen (or pillar) is assumed to be small
compared to the mass of the frame structure (mass of the tributary
volume of the roof), the specimen shows negligible resistance in the
impact of the beam to the specimen top and hence the velocity vr is
approximately equal to the velocity of the top of the specimen
(pillar). Therefore, an estimate for the kinetic energy (Q) of the
damaged specimen or pillar is as follows:
mp q2u 4
Q D ð8Þ
mr K r
In Eq. (8), mp is the mass (or part of the mass involved in the burst)
of the specimen (or pillar) and  is to show that Q is proportional to
the right hand side expression. For a pillar and the attached roof in
Fig. 8, the kinetic energy becomes:
Lqp q2u D6
Q
abhqr K r
Based on Eq. (9), it is easy to see that the dimensionless kinetic
energy has a linear relationship with the second and fifth parame-
ters in Eq. (3), which is consistent with the numerical results shown
in Fig. 10b and 12. The shortcoming of Eq. (9) is that it does not
include the other involved parameters which were identified
through dimensional analysis and are reported in Eq. (3). Notice
that in Eqs. (3) and (9), the two parameters that have the greatest
impact on the rock burst energy are the specimen diameter and
the uniaxial compressive strength of rock (pillar). In particular, dou-
bling the specimen or pillar diameter can increase the rock burst
energy by about two orders of magnitude. This suggests that in
underground activities that are prone to the strain bursting, large
diameter pillars should be avoided. Numerical results in Fig. 13 con-
firm the linear relationship between the kinetic energy of the pillar
with the qu2 and D6.
To reduce the possibility of rock burst in the mining engineer-
ing, a technics called de-stressing is used. De-stressing is to move
highly stressed zones some distance from the mining activities.
The idea of de-stressing is not new, and this technique has been
reported in the 1950s [26,27]. In general, three de-stressing tech-
niques are available: blasting, drilling, and water infusion/hydrau-
lic fracturing. Blast de-stressing or preconditioning has been
successfully used in the South African gold mines. Grodner [28]
reports that rock preconditioning by detonating explosives in the
confined rock mass ahead of the mining face transfers the stresses
further away and effectively unloads the immediate face area.
To study the effect of preconditioning on the induced kinetic
ð9Þ energy in the numerical simulation, drilling approach was utilized;
in the circumferential area of the bonded particle specimen 2.5 (di-
ameter)
6.5 (height) cm2 in size, circular holes within the range
of 0–0.3 cm in diameter and 0.5 cm in depth in a square pattern
of 1.57
1.57 cm2 were generated (Fig. 14). Two sets of numerical
tests were conducted. In the first set, the uniaxial compressive
strength of the simulated specimen was allowed to reduce due to
the presence of the holes. As Fig. 15a shows, in this case, the
induced kinetic energy reduced linearly as the diameter of the
holes is increased. In the second set of the numerical tests, the
normal bond and shear bond at the contact points of the particles
were increased proportionally such that the uniaxial compressive
strength was not affected by the presence of the holes. The
 A. Fakhimi et al. / Computers and Geotechnics 74 (2016) 36–44 43

0.3 7

6
0.25 y = 0.0004x + 0.026
R² = 1
Kinetic Energy (J)

Kinetic Energy (J)

y = 1E-05x + 0.0077 5
0.2 R² = 0.998
4
0.15 (b)
(a) 3
0.1
2

0.05 1

0 0
0 5000 10000 15000 20000 25000 30000 0 5000 10000 15000 20000

qu2 (MPa)2 D6 (cm6)

Fig. 13. The effect of pillar uniaxial compressive strength (qu) and diameter (D) on rock burst kinetic energy. (a) Linear relationship between kinetic energy and q2u. (b) Linear
Relationship between kinetic energy and D6. In these tests L = 68 mm, Kr = 14.08 kN/mm, l0 = 0.18, qp = 2600 kg/m3, and qr = 7800 kg/m3 were used.

due to the reduction in uniaxial strength of the specimen; field pre-

conditioning mostly targets the rock strength in the strain burst
region through which the potential for sever rock bursts is
diminished.

1.57 cm
8. Conclusion

A steel frame structure was designed to act as a soft loading sys-

tem in uniaxial compression testing of sandstone specimens. The
physical uniaxial tests showed that violent failure can be captured
by using this testing configuration. High speed camera photogra-
phy showed particle velocities of the failed specimen in the order
of 4–5 m/s which should be expected in an actual field rock burst.
The physical tests were mimicked by using a linear elastic finite
element model for the frame structure while the rock was simu-
1.57 cm lated by a bonded particle discrete element system. Dimensional
analysis was implemented to predict the relationship between
the induced kinetic energy and the involved parameters. It was
shown that the pillar (specimen) diameter, pillar length, roof stiff-
1.04 cm ness, post-peak pillar behavior, roof-pillar interface friction coeffi-
cient, pillar and roof rock densities, tributary volume of the roof,
and uniaxial compressive strength of pillar are contributing factors
Fig. 14. The hole (3.0 mm diameter) pattern in the simulated rock specimen to
into the induced kinetic energy of the bursting pillar. In particular,
reduce the induced kinetic energy of strain burst.
the pillar diameter and its uniaxial compressive strength have
greatest impact on the severity of the strain burst. The numerical
variation of the kinetic energy with the hole diameter is shown in model confirmed that preconditioning of the pillar by drilling can
Fig. 15b. Note that in this latter case, even though the data points be used as a tool to diminish the intensity of the strain bursting
show some small scatter, the change in kinetic energy with the in a pillar and that the reduction in the kinetic energy of the rock
hole diameter seems to be negligible. This observation suggests burst in this situation is mostly due to the reduction of uniaxial
that the reduction in the kinetic energy in the first case is mostly compressive strength of pillar through the generated holes.

0.2
0.15
(a) (b)
0.16
Kinetic Energy (J)

0.12
Kinetic Energy (J)

0.09 y = -0.03x + 0.13 0.12

R² = 0.98

0.06 0.08

0.03 0.04

0 0
0 1 2 3 4 0 1 2 3 4
Holes Diameter (mm) Holes Diameter (mm)

Fig. 15. The effect of hole diameter in the circumferential area of rock specimen on the induced kinetic energy, (a) the uniaxial strength of the rock was allowed to reduce due
to the presence of the holes, (b) the normal and shear bonds were increased in proportion to keep the uniaxial compressive strength of the specimen unchanged. In these
tests, L = 68 mm, D = 25 mm, qp = 2600 kg/m3, qr = 7800 kg/m3, Kr = 14.08 kN/mm and l0 = 0.18 were used.
44 A. Fakhimi et al. / Computers and Geotechnics 74 (2016) 36–44

References [16] He MC, Miao JL, Feng JL. Rock burst process of limestone and its acoustic
emission characteristics under true-triaxial unloading conditions. Int J Rock
Mech Min Sci 2010;47:286–98.
[1] Ortlepp WD, Stacey TR. Rockburst mechanisms in tunnels and shafts.
[17] Kuch R, Lippmann H, Zhang J. Simulating coal mine bumps with model
Tunneling Underground Space Technol 1994;1(9):59–65.
material, rockbursts and seismicity in mines. In: Gibowicz SJ, Lasocki S,
[2] Ortlepp WD. High ground displacement velocities associated with rockburst
editors. Proceedings of the 4th international symposium on rockbursts and
damage. In: Young RP, editor. Int symp on rockbursts and seismicity in
seismicity in mines, Krakow, Poland, 11–14 August; 1997.
mines. Rotterdam: Balkema; 1993. p. 101–6.
[18] Fakhimi A. A hybrid discrete-finite element model for numerical simulation of
[3] Stacey TR, Ortlepp WD, Kirsten HAD. Energy-absorbing capacity of reinforced
geomaterials. Comput Geotech 2009;36:386–96.
shotcrete, with reference to the containment of rockburst damage. J S Afr Inst
[19] Fakhimi A. Application of slightly overlapped circular particles assembly in
Min Metall 1996;May/June:137–40.
numerical simulation of rocks with high friction angles. Eng Geol
[4] Linkov AM. Rockbursts and instability of rock masses. Int J Rock Mech Mining
2004;74:129–38.
Sci Geomech Abstr 1996;33:727–32.
[20] Itasca Consulting Group, Inc., Particle flow code in 2 dimensions (PFC2D).
[5] Cook NGW. The failure of rock. Int J Rock Mech Min Sci 1965;2:389–403.
Theory and background, Minneapolis, MN 55415, USA; 1999.
[6] Cook NGW. A note on rockbursts considered as a problem of stability. J S Afr
[21] Hemami B. Rock-loading machine interaction in uniaxial compression test [MS
Inst Min Metall 1965;65(10):437–46.
thesis]. Department of Mineral Engineering, New Mexico Institute of Mining
[7] Salamon MDG. Stability, instability and design of pillar workings. Int J Rock
and Technology, Socorro, NM 87801, USA; 2015.
Mech Min Sci 1970;7:613–31.
[22] Hosseini O, Theodore R, Razavi M, Fakhimi A. Rock burst of underground
[8] Petukhov IM, Linkov AM. The theory of post-failure deformations and the
pillars. 49th US rock mechanics/geomechanics symposium, ARMA 15–0650
problem of stability in rock mechanics. Int J Rock Mech Min Sci Geomech Abstr
San Francisco, California; 2015.
1979;16:57–76.
[23] Timoshenko S, Woinowsky-Krieger S. Theory of plates and shells. 2nd
[9] Burgert W, Lippmann H. Models of translatory rock bursting in coal. Int J Rock
ed. McGraw-Hill; 1959.
Mech Min Sci Geomech Abstr 1981;18:285–94.
[24] van Vliet MRA, van Mier JGM. Experimental investigation of concrete fracture
[10] Zubelewicz A, Mroz Z. Numerical simulation of rock burst processes treated as
under uniaxial compression. Mech Cohes-Frict Mater 1996;1:115–27.
problems of dynamic instability. Rock Mech Rock Eng 1983;16:253–74.
[25] Vardoulakis I, Labuz JF, Papamichos E, Tronvoll J. Continuum fracture
[11] Biot MA. Mechanics of incremental deformations. New York (US): Wiley; 1965.
mechanics of uniaxial compression on brittle materials. Int J Sol Struct
[12] Vardoulakis I. Rockbursting as a surface instability phenomenon. Int J Rock
1998;35(31–32):4313–35.
Mech Min Sci Geomech Abstr 1984;21:137–44.
[26] Holland CT. Cause and occurrence of coal mine bumps. Min Eng 1958;
[13] Bardet JP. Finite element analysis of rock burst as surface instability. Comput
September:994B–1004B.
Geotech 1989;8:177–93.
[27] Peperakis J. Mountain bumps at the sunnyside mines. Min Eng 1958;
[14] Whyatt JK, Loken MC. Coal bumps and odd dynamic phenomena – a numerical
September:982–6.
investigation. In: Peng SS, Barczak T, Mark C, Tadolini S, Finfinger G, Heasley K,
[28] Grodner M. Delineation of rock burst fractures with ground penetrating radar
et al. editors. Proceedings of 28th international conference on ground control
in Witwatersrand Basin, South Africa. Int J Rock Mech Min Sci
in mining, Morgantown, WV, USA; 2009.
2001;38:885–91.
[15] Chen ZH, Tang CA, Huang RQ. A double rock sample model for rock bursts. Int J
Rock Mech Min Sci 1997;34(6):991–1000.