Journal of Rock Mechanics and Geotechnical Engineering 9 (2017) 116e124

Contents lists available at ScienceDirect

Journal of Rock Mechanics and

Geotechnical Engineering
journal homepage: www.rockgeotech.org

Full Length Article

Dynamic rock tensile strengths of Laurentian granite: Experimental

observation and micromechanical model
Kaiwen Xia a, *, Wei Yao a, b, Bangbiao Wu b
a
Department of Civil Engineering, University of Toronto, Toronto, M5S 1A4, Canada
b
State Key Laboratory of Hydraulic Engineering Simulation and Safety, School of Civil Engineering, Tianjin University, Tianjin, 300072, China

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

Article history: Tensile strength is an important material property for rocks. In applications where rocks are subjected to
Received 24 March 2016 dynamic loads, the dynamic tensile strength is the controlling parameter. Similar to the study of static
Received in revised form tensile strength, there are various methods proposed to measure the dynamic tensile strength of rocks.
2 August 2016
Here we examine dynamic tensile strength values of Laurentian granite (LG) measured from three
Accepted 8 August 2016
Available online 8 December 2016
methods: dynamic direct tension, dynamic Brazilian disc (BD) test, and dynamic semi-circular bending
(SCB). We found that the dynamic tensile strength from direct tension has the lowest value, and the
dynamic SCB gives the highest strength at a given loading rate. Because the dynamic direct tension
Keywords:
Dynamic tensile strength
measures the intrinsic rock tensile strength, it is thus necessary to reconcile the differences in strength
Brazilian disc (BD) test values between the direct tension and the other two methods. We attribute the difference between the
Semi-circular bending (SCB) dynamic BD results and the direct tension results to the overload and internal friction in BD tests. The
Direct tension difference between the dynamic SCB results and the direct tension results can be understood by invoking
Split Hopkinson bar the non-local failure theory. It is shown that, after appropriate corrections, the dynamic tensile strengths
from the two other tests can be reduced to those from direct tension.
2017 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by
Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/
licenses/by-nc-nd/4.0/).

1. Introduction grips and bending effects due to instrumental misalignments can

Rocks are considerably weaker in tension than in compression,
and thus characterizing tensile parameters of rocks is of great
importance in many engineering and geophysical applications. For
instance, tensile failure is believed to be the main failure mode in
underground rock excavations. Tensile strength, which is dened as
the failure stress of a rock element under pure uniaxial tensile
loading, is thus an important material parameter of rocks.
Following the fundamental denition of tensile strength, direct
pull test appears to be best suited for tensile strength measure-
ment. The International Society for Rock Mechanics (ISRM) has
suggested a direct tension method to measure the static rock ten-
sile strength (Bieniawski and Hawkes, 1978). However, in practice,
the ideal uniform stress state in the specimen is very hard to be
achieved. Premature failure due to stress concentration around
* Corresponding author. Fax: 1 4169786813.
E-mail address: kaiwen.xia@utoronto.ca (K. Xia).
Peer review under responsibility of Institute of Rock and Soil Mechanics,
Chinese Academy of Sciences.
introduce signicant errors to the measurement results.
Because of the difculties associated with experimentation in
direct tensile tests, a variety of indirect methods have been pro-
posed as convenient alternatives to measure the tensile strength of
rocks, for example, Brazilian disc (BD) test (Mellor and Hawkes,
1971; Hudson et al., 1972; Bieniawski and Hawkes, 1978; Coviello
et al., 2005), ring test (Hudson, 1969; Hudson et al., 1972; Coviello
et al., 2005), and bending test (Hudson, 1969). The various indi-
rect tension testing methods aim at generating tensile stress in the
specimen by far-eld compression, which is much easier and
cheaper in instrumentation than direct pull tests. In addition, these
methods usually can give repeatable results. However, the inter-
pretation of these indirect tension results tends to rest on the
generally dubious assumption of the stress distribution prior to
fracture. Direct tension is thus still needed to verify the accuracy
and robustness of the indirect tests (Mellor and Hawkes, 1971).
Existing attempts to measure rock tensile strength are mostly
limited to quasi-static loading, primarily due to the difculties in
the dynamic experimentation and subsequent data interpretation.
However, in many mining and civil engineering applications, such
as quarrying, rock cutting, drilling, tunneling, rock blasts, and rock
bursts, rocks are stressed dynamically. Accurate characterizations

http://dx.doi.org/10.1016/j.jrmge.2016.08.007
1674-7755 2017 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by Elsevier B.V. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 K. Xia et al. / Journal of Rock Mechanics and Geotechnical Engineering 9 (2017) 116e124
of rock tensile strength over a wide range of loading rates are thus
crucial.
Due to the same reasons discussed above for static tension tests,
few dynamic direct tensile tests have been attempted (Goldsmith
et al., 1976), and research efforts have concentrated on extending
the indirect methods from quasi-static to dynamic loading. Zhao
and Li (2000) measured the dynamic tensile properties of granite
with the BD and three-point bend (TPB) techniques; the loading
was driven by air and oil and thus had a limited loading rate range.
To attain the tensile strength of rocks under high loading rates,
most researchers used the standard dynamic testing facility, split
Hopkinson pressure bar (SHPB), to apply the dynamic load (Xia and
Yao, 2015). For example, conventional SHPB tests were conducted
on BD and attened BD specimens of marble (Wang et al., 2006)
and on BD specimens of argillite (Cai et al., 2007). These attempts
followed the pioneer work on dynamic BD tests of concretes using
the SHPB (Ross et al., 1989, 1995).
The dynamic BD test using the SHPB was recently used to study
the loading rate dependence of rock tensile strength (Dai et al.,
2010a) and rock tensile strength anisotropy (Dai and Xia, 2010).
Furthermore, a dynamic semi-circular bending (SCB) method was
used in combination with the SHPB to measure the exural tensile
strength of rocks (Dai et al., 2008) and the anisotropy of the exural
tensile strength of rocks (Dai et al., 2013). Unlike earlier attempts on
dynamic indirect tests where quasi-static data regression was used
without sufcient validation, the conditions under which the
quasi-static stress analysis is valid were carefully addressed in these
recent studies. This concept was further adopted in the rst batch of
ISRM suggested methods for measuring dynamic properties of
rocks (Zhou et al., 2012). However, there is still a need to validate
the dynamic BD tests using the direct tension tests, due to the same
reason as in the static case (Mellor and Hawkes, 1971). Partially
motivated by the foregoing issues, we developed a split Hopkinson
tension bar (SHTB) system to measure the dynamic direct tensile
strength of Laurentian granite (LG) (Huang et al., 2010).
In the current study, we rst overview the three dynamic tensile
strength measurement methods: dynamic BD, dynamic SCB and
dynamic direct tension. The values of dynamic tensile strength for
the same rock (LG) obtained from these three methods are then
compiled and compared. It is found that the dynamic direct tensile
strength is consistently lower than the dynamic BD tensile strength
(Dai et al., 2010a), and the dynamic BD tensile strength is consis-
tently lower than the dynamic exural tensile strength obtained
using the dynamic SCB test (Dai et al., 2010b). It is thus the primary
objective of this work to rationale of these differences.
To understand the difference between the dynamic direct ten-
sile strength and the dynamic BD tensile strength, we propose two
mechanisms for the strength over-estimation in the dynamic BD
method: the overload effect and the internal friction effect. We
conduct dynamic BD tests using SHPB to illustrate the overload
effect, and the frictional effect is qualitatively derived based on the
micromechanical failure mechanism of rocks. After corrections
based on these mechanisms, the dynamic BD tensile strength can
be reduced to the dynamic direct tensile strength. The difference
between the dynamic exural tensile strength and the dynamic
direct tensile strength can be explained by invoking a non-local
failure theory as we used earlier (Dai et al., 2010b).
2. Overview of three dynamic tensile strength measurement
methods
2.1. Split Hopkinson pressure bar
The SHPB system is composed of three bars: a striker bar, an
incident bar, and a transmitted bar (Grag and Blumenthal, 2000). A
117
specimen is sandwiched between the incident bar and the trans-
mitted bar. The impact of the striker bar on the free end of the
incident bar induces a longitudinal compressive wave propagating
in both directions. The left-propagating wave is fully released at the
free end of the striker bar and forms the trailing end of the incident
compressive pulse i (Fig. 1). Upon reaching the barespecimen
interface, part of the incident wave is reected as the reected
wave r and the remainder passes through the specimen to the
transmitted bar as the transmitted wave t.
Based on the one-dimensional stress wave theory, the dynamic
forces on the incident end (P1) and the transmitted end (P2) of the
specimen are (Kolsky, 1949, 1953):
P1 AEi r ; P2 AEt (1)
where E is the Youngs modulus, and A is the cross-sectional area of
the bars.
2.2. Dynamic Brazilian disc method
A 25 mm diameter SHPB system is used in the study. A close-
view of the dynamic BD test in the SHPB system is schematically
shown in Fig. 2, where the disc specimen is sandwiched between
the incident bar and the transmitted bar. The principle of the BD
test comes from the fact that rocks are much weaker in tension than
in compression, and thus the diametrically loaded rock disc spec-
imen fails due to the tension along the loading diameter near the
center. The tensile stress at the central disc along the loading di-
rection is
2Pt
st (2)
pDB
where P(t) is the load; D and B are the diameter and the thickness of
the disc, respectively. It is usually believed that at the maximum
load, the corresponding tensile stress is the material tensile
strength st. In the dynamic case, the load is P1 (P2) obtained using
Eq. (1). The loading rate is the slope of the pre-peak linear portion
of the tensile stress curve (Zhou et al., 2012).
It is noteworthy that the prerequisite for using Eq. (2) for dy-
namic BD tests is the dynamic stress equilibrium in the BD spec-
imen (Dai et al., 2010a). With the pulse shaping technique (Zhou
et al., 2012; Xia and Yao, 2015), the dynamic force balance for a
typical BD test is achieved and shown in Fig. 3. The dynamic forces
P1 and P2 are calculated using Eq. (1). As shown in Fig. 3a, the dy-
namic forces on both ends of the BD specimen are almost identical
during the dynamic loading. In rock specimen, the force equilib-
rium state can be achieved when the stress wave propagates in the
rock specimen for about 3e4 times of the round-trip (Zhou et al.,
2012). Thus, the initial time for dynamic stress equilibrium in the
BD specimen can be estimated by the propagation distance and the
P-wave velocity of the rock specimen. Since the P-wave velocity of
LG is 5000 m/s (Yin et al., 2012), the stress equilibrium time for a
40 mm diameter BD specimen is theoretically about 48e64 ms. In
the typical BD test, the ratio of P1 to P2 is calculated (Fig. 3b) during
the dynamic loading period. It illustrates that the absolute value of
ratio of the forces on both ends of the BD specimen jP1/P2j has
drastic uctuations at the beginning, and then equals 1 at about
51 ms, after which the force balance is reached. The starting time (t0)
for the force balance in the typical BD test is in the range of the
theoretical force balance starting time. It is also noted that at the
peak load, the ratio is almost 1. Thus, the pulse shaping technique is
an efcient method to achieve the force balance in the rock BD
specimen and the dynamic force equilibrium is reached for all
dynamic BD tests. In addition, in our earlier work (Dai et al., 2010a),
118 K. Xia et al. / Journal of Rock Mechanics and Geotechnical Engineering 9 (2017) 116e124

P1 P2
t
Reflected wave
r
t

Transmitted wave
i

Incident wave

Gas gun o x
Sample

Striker bar Incident/Input bar Strain gauges Transmitted/Output bar

Fig. 1. Schematic of the SHPB system and the x-t diagram of stress waves propagation in SHPB.

Using a dimensional argument, the equation for calculating the

D
tensile stress at O is (Dai et al., 2008):
B
2Pt
st YS=2R (3)
pBR
Incident bar Transmitted bar
where the dimensionless stress Y(S/(2R)) is a function of dimen-
sionless distance S/(2R), and this function can be calibrated using
nite element analysis. The (exural) tensile strength is taken as
P1 P2 the maximum tensile stress in the history of s(t) and the corre-
sponding loading rate is measured from the slope of the pre-peak
linear portion of the s (t) curve (Dai et al., 2008).

Strain gauge

Fig. 2. Schematics of a BD specimen in a SHPB system. A strain gage is cemented on

the BD specimen for detecting failure onset.

the transient dynamic stress history at the disc center was calcu-
lated. It indicated that the states of the tensile stress component
(perpendicular to the loading direction) and compressive stress
component (parallel to the loading direction) at the disc center
calculated using dynamic stress analysis match with those calcu-
lated using static stress analysis. Thus dynamic stress equilibrium at
the center of the BD specimen can be guaranteed by the dynamic
force balance on the boundaries of the BD specimen (Dai et al.,
2010a). As a result, the inertia effect is eliminated in the BD spec-
imen and Eq. (2) can be used for calculating the dynamic tensile
strength of the BD rock specimen.

2.3. Dynamic semi-circular bend method

The SCB method can be viewed as an integration of the BD

method and the TPB method. Compared with the BD method, the
failure load needed for the SCB method is much smaller for a given
material. Consequently, the stress concentration at the contacts,
which may lead to premature failure, is less likely for the SCB
method. Compared with the TPB specimen, the SCB specimen has
the advantage of being core-based. In addition, the semi-circular
geometry of the SCB specimen facilitates sample alignment
(Fig. 4). As shown in the gure, R is the radius of the half disk, B is
the thickness of the rock disc, and the span between two sup-
porting pins is S. Upon loading, failure will be initiated at the failure Fig. 3. Dynamic force balance check for a typical dynamic BD test with pulse shaping:
spot O on the specimen due to the bending load. (a) Dynamic force balance; and (b) The ratio of forces on both ends of specimen.
 K. Xia et al. / Journal of Rock Mechanics and Geotechnical Engineering 9 (2017) 116e124 119

dynamic response of the material (i.e. stressestrain curve)

subsequently.
Using these three waves, the forces T1 and T2 on both ends of the
specimen can be calculated:

T1 t EAi t r t; T2 t EAt t (4)

If the specimen is in dynamic stress equilibrium, the stress of
specimen is given by

EA
st t t (5)
As

where As is the cross-sectional area of the specimen. The tensile

Fig. 4. Schematics of a semi-circular bend specimen in a SHPB system.
The tensile stress in the SCB specimen can be deduced from far-
eld measurement via quasi-static analysis Eq. (3) when the far-
eld dynamic force balance is achieved (Dai et al., 2008). The
pulse shaping technique was also employed in the SCB test to reach
the dynamic force balance on both ends of the SCB specimen. The
typical forces for both ends of the SCB specimen are almost iden-
tical and shown in Fig. 5a. Thus, the force balance is fully achieved
on both ends of the SCB specimen. The starting time of force bal-
ance in the SCB specimen can also be determined through the
method for obtaining the starting time of force balance for the BD
specimen in Section 2.2 (Zhou et al., 2012). The ratio of P1 to P2 for a
typical SCB specimen is shown in Fig. 5b. The stress equilibrium
initiation time for the SCB specimen with 20 mm radius is theo-
retically about 24e32 ms. The absolute value of jP1/P2j in the SCB
test varies dramatically rst and then almost equals 1 at 33 ms,
strength is then obtained as the maximum value of the dynamic
stress experienced by the specimen, and the loading rate can be
determined from the loading history (Huang et al., 2010).
To eliminate the inertial effect induced by the mismatch of
forces applied to both ends of tensile specimens during the dy-
namic loading, the pulse shaping technique was utilized in dynamic
direct tension tests (Huang et al., 2010). As shown in Fig. 7a, the
dynamic tensile forces T1 and T2 on both sides of the specimen are
almost identical during the dynamic loading period. The inertial
effect can thus be reduced since there is no global force difference.
In addition, since the force equilibrium state can be achieved at
about 3e4 times of the round-trip of stress wave in rock specimens,
the force balance initiation time for the direct tension specimen
with 38 mm length can be estimated theoretically to be about
45.6e60.8 ms (Zhou et al., 2012). In a typical direct tension test, the
absolute value of the ratio of T1/T2 (shown in Fig. 7b) equals 1 at
about 51 ms after drastic decrease. This illustrates that the force
25
In
20 Re
Tr (P2)
15 In+Re (P1)
10
Force (kN)

indicating the dynamic force balance being reached. The consis- 5

tency of the force balance initiation time (t0) in the typical SCB test 0
with that in theory is thus veried. It is noted that at the peak load,
the ratio is almost 1. Furthermore, with the force balance in the SCB -5
specimen, it has been proven that the peak of the transmitted force -10
on the SCB specimen is also synchronous with the rupture onset of
the specimen (Dai et al., 2008). Therefore, Eq. (3) can be utilized to -15
0 20 40 60 80 100 120 140 160 180 200 220
obtain the tensile stress in dynamic SCB tests and the dynamic force Time (s)
equilibrium is reached for all dynamic SCB tests.
(a)
12
2.4. Dynamic direct tension method using split Hopkinson tension
bar 10
t0=33 s
The SHTB system is composed of a striker tube, an incident bar, 8
and a transmitted bar (Fig. 6a). Two strain gages are mounted on
|P1/P2|

the incident bar and transmitted bar, respectively. The dumbbell 6

shaped specimen (Fig. 6b) is glued to the incident bar and the
transmitted bar with strong epoxy adhesive. The striker tube, 4
moving freely through the incident bar, is placed inside the gun
barrel and launched by a low-speed gas gun. The impact of the 2 |P1/P2|=1
striker tube on the ange attached to one end of the incident bar
generates a longitudinal tensile wave propagating in the incident 0
0 20 40 60 80 100 120 140 160 180 200 220
bar as incident wave i. When the incident wave reaches the bar-
Time (s)
specimen interface, part of the wave is reected back as reected
(b)
wave r, and the remainder passes through the specimen and then
enters the transmitted bar as transmitted wave t. These three Fig. 5. Dynamic force balance check for a typical dynamic SCB test with pulse shaping:
waves are measured using strain gages and used to infer the (a) Dynamic force balance; and (b) The ratio of forces on both ends of specimen.
120 K. Xia et al. / Journal of Rock Mechanics and Geotechnical Engineering 9 (2017) 116e124

Flange Barrel
i Strain gauges t

T1 T2

Sample
r Incident bar Transmitted bar
Striker

(a)
Epoxy glue

24 mm

22 mm
38 mm
Incident bar Transmitted bar
Sample
(b)

Fig. 6. (a) Schematic of the split Hopkinson tension bar (SHTB); and (b) The dumbbell shaped specimen.

In 3. Results and discussion

60
Re
40
Tr (T2) The tensile strength obtained from the three dynamic testing
In+Re (T1) methods is plotted against the loading rate in Fig. 8. It is clear that
20 for a given loading rate, the tensile strength obtained from the BD
test is higher than that obtained from direct tension, while the SCB
Force (kN)

0 test gives the highest tensile strength values. Because the tensile
strength obtained from direct tension is the true tensile strength, it
-20
is thus necessary to understand these differences.
-40
3.1. Corrections for dynamic BD tensile strength
-60

0 25 50 75 100 125 150 175 200 225 250 275 In their classic work on static BD tests, Mellor and Hawkes
Time (s) (1971) observed that in some cases, the primary crack occurred
(a) as loading continued, and a higher load was carried by the
12 cracked specimen before the nal collapse of the specimen. As a
result, the load at the failure (i.e. the failure load) is lower than
10 the peak load recorded. Theoretically, the failure load should be
t0=51 s used to calculate the tensile strength. However, in practice, the
8 nal load (i.e. the peak load) is used. This overload effect will
|T1/T2|

6
60
BD (Dai et al., 2010a)
4 SCB (Dai et al., 2010b)
50 Direct tension (Huang et al., 2010)
2 |T1/T2|=1
Tensile strength (MPa)

40
0
0 25 50 75 100 125
Time (s) 30
(b)

Fig. 7. Dynamic force balance check for a typical dynamic direct tension test with 20
pulse shaping: (a) Dynamic force balance; and (b) The ratio of forces on both ends of
specimen.
10

balance is reached at about 51 ms, which has a good agreement with 0

the theoretical starting time for the force balance. Thus, with the 0 200 400 600 800 1000
Loading rate (GPa/s)
pulse shaping technique, the dynamic force equilibrium was
reached for all dynamic direct tensile tests. Fig. 8. Comparison of dynamic tensile strength from three testing methods.
 K. Xia et al. / Journal of Rock Mechanics and Geotechnical Engineering 9 (2017) 116e124
thus lead to an over-estimation of the rock tensile strength. To
explain the reason that the dynamic BD results are consistently
higher than the dynamic direct tension results, we rst need to
check whether or not this overload phenomenon exists in dy-
namic tests.
Because the loading history can be determined from the
measured waves using Eq. (1), the failure load is readily to be
determined if the failure onset is detected. To determine the failure
onset of the primary fracture, a strain gage is glued on the disc
surface 5 mm away from the center of the disc (Fig. 2). The rock
specimen emits elastic release waves upon cracking, and this wave
causes a turning point in the recorded strain gage signal (Jiang et al.,
2004). This turning point thus indicates the crack onset. In addition,
the elastic release wave generated by failure initiation travels at the
sound speed of the rock material and bar material. Since the dis-
tance between the location of original strain gage and the failure
point is known, the time for the elastic wave to propagate from the
disc center to the original strain gage location can be accurately
determined (Xia et al., 2011). Thus, the original strain gage signal
can be corrected by considering the time for the elastic wave to
propagate from the disc center to the original strain gage location.
With the same method, the strain gage signal on specimen can also
be corrected and thus the original strain gage signal and that on
specimen are synchronized for determining the failure load at the
failure onset.
We conducted 8 dynamic BD tests of LG featuring loading rates
from 200 GPa/s to 1700 GPa/s (Fig. 9). The strain gage was used to
detect the failure onset in these tests. We rst simply used the peak
load to calculate the tensile strength. As shown in Fig. 9, the dy-
namic strength results obtained match well with the previous re-
sults for the same rock (Dai et al., 2010a). It is noted that the data
points when the loading rates are higher than 2500 GPa/s may not
be so reliable in earlier results because the dynamic force balance
condition may have been violated. That is exactly the reason that in
the new batch of tests, we did not go beyond the loading rate of
2000 GPa/s.
The tensile stress history from a typical test is compared with the
strain gage signal in Fig. 10. The overload is clearly identied: the
failure onset is observed at 75 ms as indicated by the turning point of
the strain gage signal while the peak load is reached at 88 ms. Here
the strain gage signal has already been shifted to consider the wave
propagation from the failure point (center of the disc) to the strain
gage location. In the correction, we used the elastic wave velocity of
the rock as 5000 m/s and the propagation distance from the center
of the disc to the gage as 5 mm, the time shift of the strain gage
Dai et al. (2010a)
50 BD without overload correction
BD with overload correction
40
Strength (MPa)
30
20
10
0 500 1000 1500 2000
Loading rate (GPa/s)
Fig. 9. Tensile strength measured from dynamic BD tests with and without overload
correction.
121
signal is thus 1 ms. If we simply use the peak load to calculate the
strength, the value is 40.9 MPa; the valid value calculated at the
failure onset is only 33 MPa. The overload thus leads to 20% over-
estimation of the tensile strength in this test. For the new batch of
BD results, all strength values were corrected in this way and the
corrected values are also shown in Fig. 9. It is noted that the per-
centage of overload is not a constant and thus the strain gage is
needed to detect the failure onset for each dynamic BD test. For the
old BD data set (Dai et al., 2010a), the overload correction is not
possible because the failure onset was not detected in those tests.
The BD tensile strength values after overload correction are
compared with the dynamic direct tension results as shown in
Fig. 11. It can be seen that the BD results with this correction are still
consistently higher than the direct tension results. To explain the
difference, we have to examine the microscopic failure congura-
tion of rocks in a direct tension test and a BD test (Fig. 12). The top
row shows the macroscopic loading conguration in both tests. If
we zoom in on a square element in the center for both cases, the
microscopic failure pattern is shown in the bottom row of the
gure. At the element level, the stress state for the direct tension
specimen is uniaxial while that for the BD specimen is biaxial, with
tension in the vertical direction and compression in the horizontal
direction (the far-eld loading direction). For real rocks, the nal
fracture surface is not mathematically at. The roughness of the
fracture surface is controlled by the rock grain size and nature of
the fracture. Because of the compression in the horizontal direction
and the roughness of the fracture surface, there will be frictional
resistance to the tensile failure for the BD specimen. This frictional
resistance is a result of relative motion between the two imaginary
rock failure surfaces that are under compression in the loading
direction.
Based on the microscopic failure conguration of the BD spec-
imen, the internal friction effect will lead to a nominal tensile
strength snt , which is larger than the intrinsic rock tensile strength
st. The nominal tensile strength is related to the intrinsic tensile
strength as
snt st hsnt 0snt st =1  h (6)
where the unknown parameter h describes the frictional effect as a
result of compression in the horizontal direction. This parameter
depends on the roughness of the fracture surface and the coef-
cient of friction of the rock material. It can be seen from the
equation that the nominal rock tensile strength from BD tests is
over-estimated due to the internal friction effect.
Although it is not possible to determine h directly from labo-
ratory measurements, we assume that it is an empirical material
1.0
40.9 MPa
40
33 MPa 0.8
Strain gauge signal (V)
30 Stress
Tensile stress (MPa)
0.6
Loading rate Signal
20
1689 GPa/s 0.4
10 0.2
2500 3000
0 0.0
0 50 100 150 200
Time (s)
Fig. 10. A typical dynamic BD test demonstrating the overload phenomenon.
122 K. Xia et al. / Journal of Rock Mechanics and Geotechnical Engineering 9 (2017) 116e124
35
Direct tension
BD with overload correction
BD with overload + friction corrections
30
Strength (MPa)
25
20
15
10
0 500 1000 1500
Loading rate (GPa/s)
Fig. 11. Dynamic tensile strengths from direct tension tests and dynamic BD tests.
3 3
(a) (b )
Fig. 12. Macroscopic and microscopic failure pictures in (a) a direct tension test and (b)
a Brazil test.
parameter. This parameter depends on the internal friction and the
roughness of the nal failure surface. The internal friction angle is a
material parameter and the roughness of the nal failure surface
depends on the grain structure of the material. As a result, we
postulate that h is an empirical material parameter and this
parameter would vary for different rocks. Based on this under-
standing, we applied a uniform correction to the dynamic BD
strength after the overload correction for the rock used in the study.
It is also noted that, to apply the empirical correction caused by the
friction effect, one also has to convert the loading rate. This can be
achieved by using stress instead of strength in Eq. (6) and differ-
entiating both sides of the function with respect to time, yielding
s_ n s_ =1  h.
As shown in Fig. 11, we obtained consistent tensile strength
values from the direct tension tests and the BD tests with correc-
tions (we assumed h 0.15 for LG to conduct the empirical
correction for the friction effect). The robustness of the mechanism
of frictional effect is veried by the fact that using single value of h,
we achieved consistent strength values over a wide range of
loading rate.
3.2. Correction for dynamic exural tensile strength
The difference between the SCB results (exural tensile
strength) and the direct tension results can be explained using
the non-local failure model (Lajtai, 1972; Carter, 1992; Van de
Steen and Vervoort, 2001). This theory states that the material
90 Direct tension
SCB
80 SCB with correction
70
Strength (MPa)
60
50
40
30
20
10
0 500 1000 1500 2000 2500 3000
Loading rate (GPa/s)
Fig. 13. Dynamic tensile strengths from direct tension and dynamic SCB tests.
fails when the local linear elastic stress averaged over a distance
d along the prospective fracture path reaches the tensile
strength st (Van de Steen and Vervoort, 2001). Integrating the
stress s normal to the prospective fracture path over that dis-
tance yields
Z
l0 d
st d sdl (7)
l0
Using testing results with different specimen size congura-
tions, a linear regression analysis or a least-square method was
adopted to determine d and st simultaneously. In our case, the
tensile stress along the potential fracture path of SCB specimen
was calculated numerically with nite element analysis. The
resulting stress was tted to a polynomial form and was substi-
tute to the right size of Eq. (7). The ratio of the exural strength
sf (stress at the failure spot when the specimen fails) to the
tensile strength st can thus be determined as a function of d. This
ratio can also be calculated from the experimental results. We
can thus nd one value of d from the strength pair of one loading
rate. Because we have results from a wide range of dynamic
loading rates, we can determine the optimal d inversely using Eq.
(7), and then nd the tensile strength as a function of the loading
rate (Dai et al., 2010b).
At a given loading rate, the SCB tensile strength is twice the
direct tension strength. To reconcile this, the characteristic length
d 6.5 mm is used. This length is different from what we estimated
earlier (d 6 mm) for the same rock (Dai et al., 2010b), which is
attributed to the reference value of the tensile strength: direct
tension in this work and BD tension in earlier work. It is noted that,
similar to early work, we only applied the non-local theory to
correct the strength; however, we did not convert the loading rate.
This is because the failure in the SCB test is at the center of the
original disc (i.e. point O in Fig. 4) (Dai et al., 2010b). It is thus
appropriate to use the stress history at this point to calculate the
loading rate.
As shown in Fig. 13, the corrected values from SCB tests are
consistent with the direct tension results. It is also noted that the
characteristic length is the basis of the no-local failure model
(Lajtai, 1972; Carter, 1992; Van de Steen and Vervoort, 2001).
Although this parameter cannot be directly measured, it should be
related to the characteristic length of material microscopic struc-
tures. Again because one cannot measure this length directly, the
correction should be treated somewhat empirical until a denite
relation between this length and the material microscopic structure
is established.
 K. Xia et al. / Journal of Rock Mechanics and Geotechnical Engineering 9 (2017) 116e124
4. Conclusions
In this work, we rst reviewed three dynamic tensile strength
measurement methods: dynamic BD, dynamic SCB and dynamic
direct tension. The values of dynamic tensile strength for the same
rock (LG) obtained from these three methods were then compared.
It is found that the dynamic direct tensile strength is consistently
lower than the dynamic BD tensile strength, and the dynamic BD
tensile strength is consistently lower than the dynamic exural
tensile strength obtained using the dynamic SCB test.
To understand the difference between the dynamic direct ten-
sile strength and the dynamic BD tensile strength, we proposed two
mechanisms for the strength over-estimation of the dynamic BD
method: the overload effect and the internal friction effect. We
conducted dynamic BD tests using SHPB to illustrate the overload
effect. The frictional effect was qualitatively derived from the
experimental results and the empirical correction for the friction
effect was obtained according to the micromechanical failure
mechanism of rocks. After corrections based on these mechanisms,
the dynamic BD tensile strength can be reduced to the dynamic
direct tensile strength. The differences between the dynamic ex-
ural tensile strength and the dynamic direct tensile strength were
explained by invoking a non-local failure theory.
From this study, we believe that the dynamic strength values
obtained using various dynamic indirect tensile tests are more
meaningful than we expected. By carrying out proper corrections,
we can obtain the intrinsic rock dynamic strengths from these in-
direct tensile results. In this way, we can measure the intrinsic
dynamic rock tensile strength while avoiding the difculties asso-
ciated with the direct tensile experiments.
Conict of interest
The authors wish to conrm that there are no known conicts of
interest associated with this publication and there has been no
signicant nancial support for this work that could have inu-
enced its outcome.
Acknowledgements
Funding was provided by the Innovative Research Groups of
Natural Science Foundation of China (NSFC) (Grant No. 51321065)
and NSFC (Grant No. 51479131). The research of Kaiwen Xia was
partially supported by the Natural Sciences and Engineering
Research Council of Canada (NSERC) through the Discovery (Grant
No. 72031326).
List of symbols
BD Brazilian disc
ISRM International Society for Rock Mechanics
LG Laurentian granite
SCB Semi-circular bending
SHPB Split Hopkinson pressure bar
SHTB Split Hopkinson tension bar
TPB Three-point bend
A The cross-sectional area of the bars (mm2)
As The cross-sectional area of the direct tension specimen
(mm2)
B The thickness of BD and SCB specimens (mm)
D The diameter of BD specimen (mm)
E The Youngs modulus of the bars (GPa)
P1 The dynamic forces on the incident end of the specimen
(N)
123
P2 The dynamic forces on the transmitted end of the
specimen (N)
P(t) The time-varying load recorded in the test (N)
R The radius of the half disk (mm)
S The span between two supporting pins in NSCB tests
(mm)
T1 The forces on incident end of the direct tension specimen
(N)
T2 The forces on transmitted end of the direct tensile
specimen (N)
Y The dimensionless stress
s The tensile stress of rock specimen (MPa)
st The intrinsic rock tensile strength (MPa)
snt The nominal tensile strength (MPa)
s_ n The loading rate for the nominal tensile strength (GPa/s)
s_ The tensile loading rate (GPa/s)
i The incident compressive pulse
r The reected compressive pulse
t The transmitted compressive pulse
h The parameter describes the frictional effect
d The length scale used in the non-local theory (mm)
References
Bieniawski ZT, Hawkes I. Suggested methods for determining tensile strength of
rock materials. International Journal of Rock Mechanics and Mining Sciences
1978;15(3):99e103.
Cai M, Kaiser PK, Suorineni F, Su K. A study on the dynamic behavior of the Meuse/
Haute-Marne argillite. Physics and Chemistry of the Earth, Parts A/B/C
2007;32(8e14):907e16.
Carter BJ. Size and stress gradient effects on fracture around cavities. Rock Me-
chanics and Rock Engineering 1992;25(3):167e86.
Coviello A, Lagioia R, Nova R. On the measurement of the tensile strength of soft
rocks. Rock Mechanics and Rock Engineering 2005;38(4):251e73.
Dai F, Xia KW. Loading rate dependence of tensile strength anisotropy of Barre
granite. Pure and Applied Geophysics 2010;167(11):1419e32.
Dai F, Xia KW, Luo SN. Semicircular bend testing with split Hopkinson pressure bar
for measuring dynamic tensile strength of brittle solids. Review of Scientic
Instruments 2008;79(12):123903. http://dx.doi.org/10.1063/1.3043420.
Dai F, Xia KW, Tang LZ. Rate dependence of the exural tensile strength of Lau-
rentian granite. International Journal of Rock Mechanics and Mining Sciences
2010b;47(3):469e75.
Dai F, Huang S, Xia KW, Tan ZY. Some fundamental issues in dynamic compression
and tension tests of rocks using split Hopkinson pressure bar. Rock Mechanics
and Rock Engineering 2010a;43(6):657e66.
Dai F, Xia K, Zuo JP, Zhang R, Xu NW. Static and dynamic exural strength anisotropy
of Barre granite. Rock Mechanics and Rock Engineering 2013;46(6):1589e602.
Goldsmith W, Sackman JL, Ewerts C. Static and dynamic fracture strength of Barre
granite. International Journal of Rock Mechanics and Mining Sciences & Geo-
mechanics Abstracts 1976;13(11):303e9.
Grag GT, Blumenthal WR. Split-Hopkinson pressure bar testing of soft materials. In:
ASM handbook: mechanical testing and evaluation. American Society for
Metals (ASM); 2000. p. 488e96.
Huang S, Chen R, Xia KW. Quantication of dynamic tensile parameters of rocks
using a modied Kolsky tension bar apparatus. Journal of Rock Mechanics and
Geotechnical Engineering 2010;2(2):162e8.
Hudson JA. Tensile strength and the ring test. International Journal of Rock Me-
chanics and Mining Sciences & Geomechanics Abstracts 1969;6(1):91e7.
Hudson JA, Brown ET, Rummel F. The controlled failure of rock discs and rings
loaded in diametral compression. International Journal of Rock Mechanics and
Mining Sciences & Geomechanics Abstracts 1972;9(2):241e8.
Jiang FC, Liu RT, Zhang XX, Vecchio KS, Rohatgi A. Evaluation of dynamic fracture
toughness KId by Hopkinson pressure bar loaded instrumented Charpy impact
test. Engineering Fracture Mechanics 2004;71(3):279e87.
Kolsky H. An investigation of the mechanical properties of materials at very high
rates of loading. Proceedings of the Physical Society, Section B 1949;62:676e
700.
Kolsky H. Stress waves in solids. Oxford: Clarendon; 1953.
Lajtai EZ. Effect of tensile stress gradient on brittle-fracture initiation. International
Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts
1972;9(5):569e78.
Mellor M, Hawkes I. Measurement of tensile strength by diametral compression of
discs and annuli. Engineering Geology 1971;5(3):173e225.
Ross CA, Thompson PY, Tedesco JW. Split-Hopkinson pressure bar tests on concrete
and mortar in tension and compression. ACI Materials Journal 1989;86(5):475e
81.
124 K. Xia et al. / Journal of Rock Mechanics and Geotechnical Engineering 9 (2017) 116e124
Ross CA, Tedesco JW, Kuennen ST. Effects of strain-rate on concrete strength. ACI
Materials Journal 1995;92(1):37e47.
Van de Steen B, Vervoort A. Non-local stress approach to fracture initiation in
laboratory experiments with a tensile stress gradient. Mechanics of Materials
2001;33(12):729e40.
Wang QZ, Li W, Song XL. A method for testing dynamic tensile strength and elastic
modulus of rock materials using SHPB. Pure and Applied Geophysics
2006;163(5):1091e100.
Xia KW, Yao W. Dynamic rock tests using split Hopkinson (Kolsky) bar systemeA re-
view. Journal of Rock Mechanics and Geotechnical Engineering 2015;7(1):27e59.
Xia KW, Dai F, Chen R. Advancements in Hopkinson pressure bar techniques and
applications to rock strength and fracture. In: Zhou Y, Zhao J, editors. Advances
in rock dynamics and applications. Boca Raton, USA: CRC Press/A.A.Balkema;
2011. p. 35e78.
Yin TB, Li XB, Xia KW, Huang S. Effect of thermal treatment on the dynamic fracture
toughness of Laurentian granite. Rock Mechanics and Rock Engineering
2012;45(6):1087e94.
Zhao J, Li HB. Experimental determination of dynamic tensile properties of a
granite. International Journal of Rock Mechanics and Mining Sciences
2000;37(5):861e6.
Zhou YX, Xia K, Li XB, Li HB, Ma GW, Zhao J, Zhou ZL, Dai F. Suggested methods for
determining the dynamic strength parameters and mode-I fracture toughness
of rock materials. International Journal of Rock Mechanics and Mining Sciences
2012;49:105e12.
Dr. Kaiwen Xia is currently an associate professor at the
Department of Civil Engineering of the University of Tor-
onto. He obtained both his B.S. and M.S. degrees from the
University of Science and Technology of China in 1994 and
1998, majored in Explosion Mechanics. Dr. Xia nished his
Ph.D. degree at the California Institute of Technology in
2005, with major in Mechanical Engineering and minor in
Geophysics. After a year working as a postdoctoral
research fellow at the Brown University, he joined the
University of Toronto in 2006 as an assistant professor and
was promoted and granted the tenure in 2012. Dr. Xias
research is focused on dynamic response of materials and
dynamic fractures. His academic contributions include the
discovery of supershear earthquakes in the laboratory,
systematic study of spontaneous fractures and the development of a series of dynamic
testing methods for rocks. He was the key member in the Commission on Rock Dy-
namics of the International Society for Rock Mechanics (ISRM-CRD) from 2007 to 2011,
and championed the drafting of the rst three dynamic testing methods of rocks. He
chaired the commission from 2011 to 2015 and is currently the chair of the commis-
sion. To date, Dr. Xia has published 65 SCI journal papers, 3 book chapters, and
numerous conference abstracts and other papers.