Sei sulla pagina 1di 10

International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

Contents lists available at ScienceDirect

International Journal of
Rock Mechanics and Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms

Non-local failure theory and two-parameter tensile strength model for semi- T
circular bending tests of granitic rocks

Wei Yao, Kaiwen Xia , Xing Li
Department of Civil & Mineral Engineering, University of Toronto, Toronto, ON, Canada M5S 1A4

A R T I C LE I N FO A B S T R A C T

Keywords: Several indirect tension tests have been proposed and widely used to quantify rock tensile strength due to
Tensile strength difficulties associated with experimentation in direct tension tests. However, discrepancies exist between rock
Flexural tensile strength tensile strengths obtained from the direct tension tests and those from indirect tension tests. The non-local theory
Semi-circular bend has been used to reconcile such differences. However, the physical meaning of the characteristic material length
Non-local theory
δ in the non-local theory is still unclear. We investigate the physical meaning of the characteristic length δ by
Characteristic material length
Rocks
testing three typical granitic rocks featuring different grain sizes. An indirect tension method - semi-circular
bending (SCB) test is applied to obtain the flexural tensile strength for these rocks. The specimens are perfectly
split into two quadrants by a main crack through the center of the specimen. The flexural tensile strengths of
three rocks generally decrease with the span distance. From flexural tensile strengths under different testing
conditions, the optimized values of the characteristic material length δ and the rock intrinsic tensile strength are
determined for these three rocks. It is found the tensile strengths of three rocks predicted by the non-local theory
match well with those obtained from independent experiments using the Brazilian disc method. The non-local
failure theory is used to explain the observation that the flexural tensile strength of three rocks decreases with
the span distance. Moreover, the material characteristic length δ correlates well with the rock grain size, and the
physical meaning of material characteristic length can be considered as the scaled average grain size of rocks.
Based on the successful application of the non-local theory, a two-parameter tensile strength model for rocks is
proposed.

1. Introduction strength of rocks.3,4 Brazilian disc (BD) test is the most popular indirect
tension test for rocks due to ease in alignment and low-cost in experi-
Underground opening is common in many important geological and mentation.3,5,6 Under uniform far-field compression, the bi-axial stress
geophysical applications including mining, petroleum, defense infra- state is induced at the failure spot (i.e., center of the disc) in the BD
structure and hydropower. Tensile stress and tensile stress gradients are specimen: the compression along the loading direction and the tensile
often encountered in underground opening although the general in-situ stress perpendicular to the loading direction. In this case, the tensile
stress state for underground rocks is normally compressive. Due to the stress near the failure spot in the BD rock specimen is uniform.7 The
much smaller tensile strength of rocks as compared to their compressive tensile strength measured using the BD method is thought to be the
strength, tensile failure is the main and significant failure mode of closest one to the rock intrinsic tensile strength and the small difference
rocks. Consequently, tensile strength is an important material para- between these two has been explained.2
meter for rocks. However instead of uniform loading condition, bending load is
Tensile strength is defined as the failure stress under pure uniaxial common near underground openings, e.g., flexural tensile stresses are
tensile loading, the direct tension method (e.g., direct pull test) is thus induced at the roof of an underground opening and the failure mode is
the most suitable method for the determination of rock tensile consequently dominated by the flexural tensile failure. In such cases,
strength.1 However in practice, significant error may be introduced to stress gradients of varying magnitude and in-situ stresses always ac-
the direct tension measurement results from stress concentration due to company the flexural tensile stresses in rocks, depending on both the
gripping of the rock specimens and misalignment.2 Hence, several al- configuration of the excavation and on the specific rock breaking
ternative indirect tension tests were proposed to measure the tensile technique.8 High tensile stress gradients also exist on a microscale in


Corresponding author.
E-mail addresses: wei.yao@mail.utoronto.ca (W. Yao), kaiwen.xia@utoronto.ca (K. Xia), xingmail.li@utoronto.ca (X. Li).

https://doi.org/10.1016/j.ijrmms.2018.07.002
Received 18 April 2018; Received in revised form 28 June 2018; Accepted 26 July 2018
1365-1609/ © 2018 Elsevier Ltd. All rights reserved.
W. Yao et al. International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

Table 1 stress concentrators. Some generalized forms of non-local failure cri-


The comparison between the flexural tensile strength and tensile strength of teria were presented in the literature,14,19,24 including a general form of
rocks. body non-local strength conditions, a general form of point non-local
Rock Flexural tensile strength Tensile strength (MPa) strength conditions, and several functional forms of non-local strength
(MPa) conditions. For non-local failure criteria based on strength evaluation,
the generalized form can be expressed as follows:
Barre granite (BG) 16.510 9.45
Laurentian granite (LG) 26.216 12.816 f (sij (y)) = σc (1)
Bowral Trachyte 7.817 3.718
Gosford Sandstone 11.617 8.618 where f is a function of stress tensor sij , y is an analyzed point, σc is a
Carrara Marble 25.117 12.018 material constant that is called the intrinsic tensile strength in this
work. For two-dimensional cases, there are three most popular non-
local failure criteria based on strength evaluation:14 fracture criterion
the vicinity of flaws, at grain contacts and at pores in rocks. Van de
based on average stress over a characteristic length or average stress
Steen et al.8 experimentally and theoretically investigated the influence
fracture criterion (ASFC),20,21 fracture criterion based on a minimum
of stress gradients on the flexural tensile failure of rocks. It has been
stress over a characteristic length or minimum stress fracture criterion
proven that the fracture initiation mainly depends on both the stress
(MSFC)21 and fracture criterion based on a model of fictitious crack
gradient and the magnitude of the bending stress. Therefore, both the
with a characteristic length or fictitious crack fracture criterion
rock flexural tensile strength and the stress gradient in rocks con-
(FCFC).22,23 We assume that (ρ , θ) is a local polar coordinate system
siderably affect the tensile failure of rocks.
with the center at the analyzed point y of a body; η(θ) is a unit vector
Semi-circular bending (SCB) test is an effective method to study the
with angle θ to the coordinate axis; and σρρ , σθθ and σρθ are the stress
rock flexural tensile failure in the laboratory. The tensile failure of rock
components in this coordinate system.
specimen in the SCB method is induced by far-field compression, and
The ASFC states that the material fails when the tensile strength of
the tensile stress at the failure point of the SCB tests is produced by
material is equal to the average local stress over a specific distance (i.e.
bending load. Along the fracture path in the SCB specimen, there exists
a characteristic length) along the prospective fracture direction. It can
a tensile stress gradient of various magnitudes, depending on the con-
be written in the following generalized form:
figuration of the SCB specimen.8,9 This usually leads to a quite higher
flexural tensile strength of rocks as compared to the tensile strength 1 δ1
max
δ1 −π < θ < π
∫0 σθθ (y + ρ η (θ)) dρ = σc
(2)
obtained from the BD tests (as shown in Table 1).2,3,6,8,10–13 This dis-
crepancy has been reconciled by non-local failure theories.2,8,10 In these where σc is the intrinsic tensile strength of a body under uniform trac-
theories, the strength of a body at a point is generally a function of the tion, δ1 is a characteristic material length. If the maximum direction θ0
stress state from a certain representative volume, area or line around in Eq. (2) is known and the integration is performed along this direc-
that point.14 It is clear that non-local failure theories always involve the tion, the ASFC equation can be simplified as follows:
introduction of a characteristic material length δ. Using an appropriate
1 δ1
characteristic length value, the intrinsic tensile strength of a material in
δ1
∫0 σθθ (y + ρ η (θ0)) dρ = σc
(3)
the non-local theory as obtained from the apparent strength (i.e.,
flexural tensile strength) values is similar to the tensile strength ob- The MSCF implies that the material fails when the tensile strength of
tained from the BD tests with a negligible stress gradient.2,8,10 It has material is equal to the minimum local stress over a characteristic
been approved that the characteristic length can be considered as a length along the prospective fracture direction. It may be written as:
material parameter, being related to the aggregate or grain size.15
Nevertheless, the physical meaning of the characteristic material length max ⎡0min σθθ (y + ρ η (θ)) ⎤ = σc
scale in the non-local theories is still unclear. Hence, the primary ob- −π ≤ θ ≤ π ⎣ ≤ ρ ≤ δ2 ⎦ (4)
jective of this work is to physically determine the meaning of the
where σc and δ2 are material constants with the similar meaning as
characteristic material length for rocks.
those in Eq. (2). With a known maximum direction θ0 in Eq. (4), MSFC
To achieve the objective of determining the physical meaning of the
can be rewritten in a simpler form:
characteristic material length, we employed the SCB method to obtain
the flexural tensile strength for three typical granitic rocks featuring min σθθ (y + ρ η (θ0)) = σc
0 ≤ ρ ≤ δ2 (5)
different grain sizes. Based on the different loading conditions, the
optimized values of the characteristic material length and the rock in- The FCFC indicates that the material fails when the stress intensity
trinsic tensile strength are determined for these three rocks. In addition, factor of material is equal to the critical local stress intensity factor over
the physical meaning of the material characteristic length δ is given by a characteristic length along the prospective fracture direction. In the
using the rock grain size. Based on the successful application of the non- FCFC, it is assumed that there exists a fictitious crack with a char-
local theory, a two-parameter tensile strength model for general tensile acteristic length δ3 originating from the considered point y of the body.
rock failure is proposed, which provides good and consistent guidelines The FCFC may be represented in the following form:
for rock structural design involving a stress gradient. max min K1i (y, θ , δ3 ) = K1c
This paper is organized as follows. After the introduction, Section 2 −π ≤ θ ≤ π i (6)
reviews the non-local failure criteria. The experimental methodology is Here K1c and δ3 are material constants, K11 = K1 (y) and
introduced in Section 3 and Section 4 describes the rock characteristics. K12 = K1 (y + δ3 η (θ)) are the stress intensity factors at the ends of the
The results along with discussions are presented in Section 5. Section 6 fictitious crack along the η(θ) direction. If the direction θ0 of fracture is
summarizes the entire paper. known, Eq. (6) yields:
min(K1 (y), K1 (y + δ3 η (θ0))) = K1c (7)
2. Non-local failure criteria
For an edge crack beginning from the body boundary, there exists
There are a number of non-local failure criteria proposed by re- only one stress intensity factor K1 (y + δ3 η (θ0)) and the minimum dis-
searchers.14,19–23 Some of these criteria not only give a good description appears in Eq. (7).
of experimental data when the stress distribution is close to a uniform Comparing these three non-local failure criteria, one can see that
stress state, but are also applicable to bodies with cracks and singular each of the criterion mentioned above contains two material

10
W. Yao et al. International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

parameters: a characteristic length and the tensile strength (ASFC and P (t ) S


σ (t ) = Y⎛ ⎞
MSFC) or the critical stress intensity factor (FCFC). The tensile strength πBR ⎝ 2R ⎠ (8)
or the critical stress intensity factor can be determined from traditional
where P(t) is the loading history during the test, S is the span length of
tensile or fracture experiments. The characteristic length is related to
two pins, Y is a function of the dimensionless geometry parameter αs
the structural characteristics of materials (microstructures for rocks),
= S/(2 R) and is calibrated using static finite element analysis.31 Dai
and can be indirectly determined from either a tension experiment of a
et al.31 gave the Y-calibration process and reported Y function as the
plate containing a central straight crack,14 or a linear regression ana-
polynomial fitting Y= 2.22 + 2.87αs + 4.54αs2, which is also used in
lysis/a least-square method of testing results with different rock spe-
this study. For the different configurations, the value of Y can be ob-
cimen configurations.2 In addition, Isupov et al.14 found that ASFC is
tained from the fitting equation and the flexural tensile strength σf is
preferable to predict the experimental data in most cases. Indeed, ASFC
then determined with the peak value of the tensile stress history at the
has been successfully used to reconcile the difference between the
failure spot.
flexural tensile strength and the tensile strength under both static and
A series of flexural tensile strengths are required to accurately ob-
dynamic loading conditions.2,5,8,12,25 Consequently, ASFC is used in this
tain the value of the characteristic material length for a specific rock. In
work to determine the physical meaning of the characteristic material
this study, the SCB tests with the various span distances of two pins
length.
were applied to attain a series of flexural tensile strengths of the same
rock because the studies have shown that the flexural tensile strength of
3. Experimental methodology rocks strongly depends on the SCB specimen geometry.3 Hence, a spe-
cial holder was designed to achieve the variation of the span distance in
It is well known that three points bending (TPB) tests are adopted as the SCB tests. As shown in Fig. 2, two pins are fixed on the top of the
a standard for determining the flexural tensile strength of materials holder. Three shafts are used to guarantee that two parts of the holder
such as natural and artificial building stones, rocks, cement and con- can relatively move in the desired direction, which ensures that two
cretes.9,26 As a modified TPB method, the semi-circular bending (SCB) pins are always parallel when the span length is changing. Three
method is one of the famous indirect methods for determining the springs and one threaded bolt at the center work together to fix two
flexural tensile strength of rocks (Fig. 1).2,27–30 The core-based SCB parts of the holder during the SCB tests. This guarantees that the span
specimen has advantages of convenient sample preparation and easy distance of two pins is fixed during the SCB test. The span distance can
specimen alignment.3,10 Thereby, the SCB method is cost-effective and be easily changed by turning the nut on threaded bolt. Due to the force
reliable, and is chosen in this study to measure the flexural tensile balance between the springs and threaded bolt, two pins can be fixed at
strength of rocks. different desire span distances and thus various span distances are ob-
According to the geometry of the SCB specimen in Fig. 1, we first tained in the SCB tests.
drilled 56 mm diameter rock cores at the same direction in the rock
blocks and then sliced and polished the cores into discs with 20 mm
4. Rock characteristics
thickness (B in Fig. 1) and a surface roughness variation of less than
0.5% of the specimen thickness. The SCB samples were subsequently
4.1. Properties of granitic rock with three different grain sizes
made from the full discs by diametrically cutting into half discs with
radius R = 28 mm (Fig. 1).
Three granites with different grain sizes and similar mineral com-
Based on dimensional analysis, the tensile stress history at the
positions are chosen in this study: Barre granite (BG), Laurentian
failure spot O (Fig. 1) of the SCB sample can be calculated by31
granite (LG) and Westerly granite (WG). These three granitic rocks have
been extensively studied in rock mechanics laboratories because they
can provide a high level of repeatability under carefully controlled test
conditions. Barre granite (BG), from Burlington, Vermont, USA, is a
medium-coarse grained rock with the mineral grain size ranging from
0.25 to 3 mm.32 The dominant mineral is feldspar (65%) with an
average grain size of 0.83 mm. The remainder minerals are quartz
(25%) with an average grain size of 0.9 mm and biotite (6%) with an
average grain size of 0.43 mm. Laurentian granite (LG) is a fine-medium
grained rock from the Laurentian region of the Precambrian Canadian
Shield. The mineral grain size of LG is from 0.2 to 2 mm. The average
grain size of feldspar, the dominant mineral (60%), is 0.4 mm.32 Quartz
constitutes 33% in LG and its average grain size is 0.5 mm. The grain
size of Biotite (3–5%) is of the order of 0.3 mm. Westerly granite (WG)
found mainly in the southeast corner of Rhode Island is a fine grained
rock and is composed of 66% feldspar (i.e., 36% microcline and 30%
plagioclase), 27% quartz, 6% biotite and 1% accessory minerals.33 The
grain size of feldspar is 0.1–0.8 mm and the grain size of quartz is
0.2–0.6 mm.34 The mean grain size is 0.3 mm.35,36 Some of physical and
mechanical properties of these three granitic rocks are summarized in
Table 2.

4.2. The microscopic characteristic of three rocks

Thin section was produced for these three rocks to observe the
minerals and obtain the average grain size for each rock in this study.
Microphotographs of BG, LG and WG derived from optical microscope
are shown in Figs. 3–5, respectively. The minerals of these three rocks
Fig. 1. The semi-circular bending (SCB) specimen. are illustrated in microphotographs with both transmitted plane

11
W. Yao et al. International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

Fig. 2. (a) Schematics and (b) photo of the semi-circular bending (SCB) testing system.

Table 2 feldspar and biotite.


Summary of the mechanical properties of three granitic rocks. In the thin section, the grain shape was digitized and the grain sizes
Properties Materials
were counted and measured with respect to the longest diameter of
grains. However, the grains showed in the thin section are along a
BG LG WG random plane. It is noteworthy that the grain size derived from the
3 37 16
randomly oriented grains in thin sections is generally biased. This bias
Density (g/cm ) 2.65 1.63 2.6238
Young's Modulus (GPa) 5510 9216 7039
of the grain size from thin section is caused by the so-called corpuscle
Poisson's Ratio 0.2432 0.2116 0.2539 effect.41 Generally, the grain size obtained from thin section is finer
Tensile Strength (MPa) 9.45 12.816 1040 than the nominal grain size of minerals since the two-dimensional
Flexural Tensile Strength (MPa) 16.510 26.216 – sections of mineral grains in most cases are not proportional to the
nominal cross-section of mineral grains in three dimensions. For this
reason, the average mineral grain sizes estimated from thin sections in
polarizers and crossed-polarizers. It can be seen that the quartz in mi-
this study are a little smaller than the nominal average grain sizes of
crophotographs are much clearer under transmitted plane polarized
rocks,42 but this little difference is acceptable in this study as we focus
condition. These three rocks are composed of similar minerals: quartz,

Fig. 3. Microphotographs of two typical areas in BG sample with (a) & (b) transmitted plane polarizers and (c) & (d) crossed-polarizers (A is quartz, B is feldspar and
C is biotite).

12
W. Yao et al. International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

Fig. 4. Microphotographs of two typical areas in LG sample with (a) & (b) transmitted plane polarizers and (c) & (d) crossed-polarizers (A is quartz, B is feldspar and
C is biotite).

on the trend of the data. The average grain sizes of all the minerals in from the middle of flat edge of the SCB specimens and then extended
these three rocks are estimated based on the microphotographs. The towards the top loading point. In this study, all SCB specimens with αs
average grain sizes of BG, LG and WG are about 0.98 mm, 0.74 mm and in the range of 0.40 and 0.85 exhibit the same failure pattern as shown
0.48 mm, respectively. The average grain sizes for all rocks are in the in Fig. 7, i.e., the fracture paths for these tested samples were generally
ranges of the grain size as reported in the literature.32,33 straight without significant curving. The fracture patterns for three
rocks are also identical. Therefore, the SCB tests with αs between 0.40
and 0.85 are valid for these three rocks.
5. Results and discussions For each type of rock, the average value of flexural tensile strength
under each span distance was measured through three repeated tests.
5.1. Testing procedure and experimental results By using the peak load for each SCB specimen, the corresponding value
of flexural tensile strength was calculated from Eq. (8) for three rocks.
The SCB tests were conducted on a material test system (MTS) hy- Fig. 8 presents the average value of flexural tensile strengths of three
draulic servo-control machine. A constant loading speed of 0.001 mm/s rocks in SCB tests with different span distances. It can be seen that the
was applied to all tests. The samples were symmetrically supported by flexural tensile strengths of three rocks generally decrease with the span
two fixed rollers (pins) with different span distances. Ten different span distance.
distances S were adopted for three rocks: αs = S/(2 R) = 0.40, 0.45,
0.50, 0.55, 0.60, 0.67, 0.70, 0.75, 0.80 and 0.85. Teflon tape was used 5.2. The physical meaning of the characteristic material length
to reduce the friction between the rock sample and the two rollers. The
SCB specimens were loaded by the loading plate until the final fracture. According to the non-local approach of ASFC,8,12,25 the maximum
All these SCB tests were performed with the same test procedure as direction in Eq. (2) is x direction of the SCB specimen in Fig. 6. In the
mentioned above. The complete load-displacement data were recorded SCB test, the tensile strength and the characteristic length in the ASFC
during the tests using a computerized data logger. Fig. 6 shows typical are the intrinsic tensile strength σt and the characteristic material length
load-displacement curves obtained from three repeated tests for WB δ of rocks, respectively. Hence, Eq. (3) can be rewritten in the following
when αs = 0.67. These curves have similar initial compliances and peak specific form for the SCB specimen:
loads, which indicates that the repeatability of the SCB tests is rea-
l0 + δ
sonable. The load-displacement curves were almost linear, showing the σt δ = ∫l
0
σdl
(9)
brittle failure behavior of the tested rocks.
The intact and recovered specimens for three rocks with typical where σ is the tensile stress over δ, l0 is the initial length of the fracture
span distances (αs = 0.40, 0.67 and 0.85) are shown in Fig. 7, where the path. For the SCB specimen, l0 is 0.
specimens were perfectly split into two quadrants by a main crack in The tensile stress gradient in the SCB specimen varies with the
the center due to the maximum tensile stress taken place along the different span distances of two pins. Finite element analysis was thus
symmetry line. It is observed in Fig. 7 that a tensile crack was initiated used to numerically calculate the tensile stress gradient in the SCB

13
W. Yao et al. International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

Fig. 5. Microphotographs of two typical areas in WG sample with (a) & (b) transmitted plane polarizers and (c) & (d) crossed-polarizers (A is quartz, B is feldspar and
C is biotite).

to build this model, and the finite element model totally consists of
4601 elements and 14,082 nodes. The static loading forces are acted on
the corresponding points, which is the same as the experimental geo-
metry. Because the prospective facture path of the SCB specimen is
predictable (i.e., the failure commences at point O and along x direction
in the insert of Fig. 9), the stress gradient along the fracture path of the
SCB specimen can be numerically determined. Fig. 9 shows the nor-
malized stress distribution along the fracture path from failure spot to
8 mm since this range of fracture path is sufficient to determine the
characteristic material length scale. Polynomial fits of the normalized
stress gradients of the SCB specimen with different span distances are
expressed as

σ
= Z1 x 2 + Z2 x + Z3
σm (10)

where σ is also the tensile stress along the prospective fracture path,
σm is the tensile stress at the failure spot (i.e., the maximum tensile
Fig. 6. Typical load-displacement curves of tested WB SCB specimens for three stress in the SCB specimen and it is the flexural tensile strength σf at
replicates (αs = 0.67). rock fracture), x is the distance of a point to the failure spot O (see the
insert in Fig. 9), Z1, Z2 and Z3 are the fitting parameters, the values of
which for the SCB specimens with different span distances are given in
specimen. A commercial software ANSYS was employed to perform the
Table 3. As a result, with a certain SCB geometry (i.e., a SCB test with a
finite element analysis and a symmetry boundary is applied to the two-
certain span distance in this study), the relationship between the
dimensional finite element model as shown in the insert of Fig. 9. The
characteristic material length scale δ and the intrinsic tensile strength σt
quadrilateral PLAIN82 (eight-node) element in ANSYS is implemented
is obtained by substituting Eq. (10) to Eq. (9):

14
W. Yao et al. International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

Fig. 7. Typical original and recovered SCB specimens for three granitic rocks (the scale in the picture is 10 mm).

different values to redo the calculation and comparison until the dif-
ference satisfies a reasonable tolerance.
There are several methods to perform this optimization process.
Genetic algorithm (GA) and Particle swarm optimization (PSO) are
most popular and widely used.43–47 PSO and GA are both population-
based search approaches and depend on information sharing among
their population members to enhance their search processes using a
combination of deterministic and probabilistic rules. PSO is a relatively
recent heuristic global optimization method, proposed originally by
Kennedy et al.48, attempting to simulate the choreographed, graceful
motion of swarms of birds as part of a sociocognitive study in-
vestigating the notion of “collective intelligence” in biological popula-
tions. It has been approved that the PSO is more computationally effi-
cient (uses less number of function evaluations) than the GA although
the PSO and GA on average yield the same effectiveness (solution
quality).43–47 In addition, the PSO has been developed and utilized in
various geotechnical engineering problems.49–52 Hence, the PSO algo-
Fig. 8. The flexural tensile strength of three granitic rocks derived from ex-
rithm is adopted here to optimize the values of σt and δ for three rocks.
periments and model.
In order to save calculation time and to avoid endless iterations in
the PSO, the tolerance between the predicted and experimental flexural
δ
σt δ = ∫0 [(Z1 x 2 + Z2 x + Z3) σf ] dx tensile strength is set as 5%. All flexural tensile strengths under dif-
(11)
ferent span distances for the same rock are simultaneously used to
Because the flexural tensile strengths and the stress gradients are conduct the optimization process. With the optimized values of σt and δ
known for all the SCB geometries here, the intrinsic tensile strength σt in Table 4, the corresponding predicted flexural tensile strength is ob-
and the characteristic material length δ can be determined by using Eq. tained and shown in Fig. 8, indicating that the flexural tensile strength
(11). For each granite, we need to find the most suitable value for each predicted from the non-local failure theory exhibits good agreement
parameter (e.g., σt and δ) to provide a good prediction for all the with the data derived from the SCB experiments. Meanwhile, the in-
flexural tensile strengths under different span distances. It is an opti- trinsic tensile strength of three rocks predicted by the non-local failure
mization process, in which we initially give a reasonable guess to each theory is close to the data derived from independent experiments using
parameter, and then calculate the maximum difference between the the BD method. This further proves that the non-local failure theory is
flexural tensile strengths predicted from Eq. (11) and the flexural ten- applicable to predict both the intrinsic tensile strength and the material
sile strengths derived from all the SCB experiments with different span characteristic length of rocks. In addition, the non-local failure theory
distances; if this difference is not in a satisfactory tolerance, we use

15
W. Yao et al. International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

Fig. 9. The distribution of normalized tensile stress along the prospective fac-
Fig. 10. Average grain size vs. the characteristic material length for three rocks.
ture path in the SCB specimen (x is the distance between a point on the fracture
path and the failure spot in the SCB sample, shown as an insert).
parameter is related to the grain size.15 Thereby, in this study, we as-
Table 3
sume that the different values of the characteristic length result from
The parameters of polynomial fits of the normalized stress gradient. the different grain sizes of three granites because these three granites
have almost the same mineral compositions and mineral proportions.
αs Z1 Z2 Z3
The relationship between the characteristic material length and the
0.40 0.01346 − 0.21249 0.99998 average grain size is shown in Fig. 10, in which the characteristic ma-
0.45 0.01011 − 0.18093 1.00021 terial length is linearly proportional to the average grain size. As a re-
0.50 0.00788 − 0.15850 0.99959 sult, the physical meaning of characteristic material length can be
0.55 0.00634 − 0.14239 0.99893
considered as the scaled average grain size of rocks:
0.60 0.00524 − 0.13064 0.99847
0.67 0.00417 − 0.11903 0.99819 δ = kdg (12)
0.70 0.00381 − 0.11520 0.99818
0.75 0.00330 − 0.11004 0.99829
where dg is the average grain size of rocks, k is the fitting constant
0.80 0.00288 − 0.10585 0.99853
0.85 0.00250 − 0.10236 0.99884
and k = 4.68 as determined from data (Table 4 and Fig. 10).

5.3. Two-parameter tensile strength model


Table 4
The values of σt and δ for three granitic rocks. One can see that the flexural tensile strengths of rocks with different
BG LG WG bending configurations are governed by two parameters: the char-
acteristic material length δ and the rock intrinsic tensile strength σt.
δ (mm) 7.284 5.572 4.928 Based on the successful application of the non-local theory, we pro-
σt (MPa) 9.80 14.31 13.56
posed a two-parameter tensile strength model for general tensile fail-
ures. In this model, the flexural tensile strength σf for different specimen
can be used to explain the phenomenon that the flexural tensile configurations can be determined through the intrinsic tensile strength
strength of three rocks decreases with the span distance in Fig. 8. As σt and the characteristic material length scale δ. This model can be
shown in Fig. 9, the stress gradient along the perspective fracture path expressed as:
in the SCB specimen decreases with the span distance. In other words, σf = f (δ , σt ) (13)
the smaller the span distance, the lower the average local stress in the
same specific distance (the length to the failure spot O along the x di- where the function f has to be determined by stress analysis using
rection). In this case, if both the characteristic material length scale and analytical or numerical methods. In the study, the expression of the
intrinsic tensile strength are the same (i.e., for the same rock), the function f can be obtained by integrating Eq. (11). In addition, the
flexural tensile strength under the smaller span distance is higher than material characteristic length δ has a positive correlation with the rock
that under the larger span distance based on the non-local theory (Eq. grain size, Eq. (13) can thus be rewritten as
(11)). Therefore the flexural tensile strength of a granite increases with
σf = f (kdg , σt ) (14)
the decrease of the span distance in the SCB test.
As discussed above, the tensile stress gradient exists in the SCB The two-parameter tensile failure model proposed here establishes
configuration, leading to the variation of the flexural tensile strength of the relationship between the flexural tensile strength and the intrinsic
rocks under different span distances. This behavior is successfully ex- tensile strength of rocks through the average rock grain size. With this
plained by the non-local failure theory. Besides, the non-local failure relationship, the flexural tensile strength of rocks can be obtained
theory gives a remarkable interpretation about the discrepancy be- through the intrinsic tensile strength σt and the average rock grain size
tween the BD results and the SCB results of three granites in this study dg, which are two material constants for rocks. This model gives an
and other rocks in the literature.2,3,5,6,11,12 Furthermore, the different effective approach to determine the flexural tensile strength of rocks
values of the characteristic length for various rocks have been proposed with different bending configurations. It can also be widely applied to
by many researchers.2,3,5,6,11,12 However, the physical meaning of the assess the rock structure safety when subjected to bending load without
characteristic material length δ in the non-local failure theory is still conducting a series of bending tests, contributing to the efficient and
vague. economical design for rock constructions, where the tensile strength is
It is well known that the characteristic length as a material commonly used instead of the flexural tensile strength.

16
W. Yao et al. International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

6. Conclusions Int J Rock Mech Min Sci Geomech Abstr. 2000;37(5):861–866.


14. Isupov LP, Mikhailov SE. A comparative analysis of several non-local fracture cri-
teria. Arch Appl Mech. 1998;68:597–612.
In this work, we first reviewed three most popular non-local failure 15. Bažant ZP, Li Z. Modulus of rupture: size effect due to fracture initiation in boundary
criteria based on strength evaluation. The fracture criterion based on Layer. J Struct Eng. 1995;121(4):739–746.
average stress over a characteristic length (ASFC) was then used in this 16. Iqbal M, Mohanty B, Xia K. Dynamic tensile strength and mode-I fracture toughness
in granitic rocks. In: Proceedings 11th International Congress. Exposition, Orlando;
work to determine the physical meaning of the characteristic material Society for Experimental Mechanics; 2008:2-5.
length. The flexural tensile strengths of three typical granitic rocks 17. Jaeger JC. Failure of rocks under tensile conditions. Int J Rock Mech Min Sci Geomech
featuring different grain sizes were measured using the semi-circular Abstr. 1967;4(2):219–227.
18. Jaeger JC, Hoskins ER. Rock failure under the confined Brazilian test. J Geophys Res.
bending (SCB) test with different configurations. A special holder was 1966;71(10):2651–2659.
designed to achieve the different loading configurations in the SCB 19. Guo F, Xu J, Mutoh Y. Fracture theory based on the concept of characteristic fracture
tests. A constant loading speed of 0.001 mm/s was applied to all SCB length of materials. Sci China Ser G-Phys Mech Astron. 2009;52(12):2041–2046.
20. Novozhilov VV. On a necessary and sufficient criterion for brittle strength. J Appl
tests conducted on a MTS machine.
Math Mech. 1969;33(2):201–210.
The specimens were perfectly split into two quadrants by a main 21. Whitney JM, Nuismer RJ. Stress fracture criteria for laminated composites containing
crack in the center. The flexural tensile strengths of three rocks gen- stress concentrations. J Compos Mater. 1974;8(3):253–265.
erally decrease with the span distance. Based on the different loading 22. Cruse TA. Tensile strength of notched composites. J Compos Mater.
1973;7(2):218–229.
conditions, the optimized values of the characteristic material length 23. Caprino G, Halpin JC, Nicolais L. Fracture mechanics in composite materials.
scale δ and the rock intrinsic tensile strength σt were determined for Composites. 1979;10(4):223–227.
these three rocks. It is found the intrinsic tensile strength of three rocks 24. Mikhailov SE. A functional approach to non-local strength conditions and fracture
criteria—I. Body and point fracture. Eng Fract Mech. 1995;52(4):731–743.
predicted by the non-local failure theory has good agreement with the 25. Carter BJ. Size and stress gradient effects on fracture around cavities. Rock Mech Rock
data derived from independent experiments using Brazilian disc Eng. 1992;25(3):167–186.
method, theoretically reconciling the difference between the flexural 26. Huang B, Shu X, Tang Y. Comparison of semi-circular bending and indirect tensile
strength tests for HMA mixtures. Advances in Pavement Engineering (GSP 130). Austin,
tensile strength due to bending and the tensile strength. Further, the Texas, United States: ASCE; 2005:1–12.
non-local failure theory was used to explain the phenomenon that the 27. Chong KP, Kuruppu MD. New specimen for fracture toughness determination for rock
flexural tensile strength of three rocks decreases with the span distance. and other materials. Int J Fract. 1984;26(2):R59–R62.
28. Kuruppu MD, Obara Y, Ayatollahi MR, Chong KP, Funatsu T. ISRM-suggested method
In addition, the material characteristic length δ has a positive correla- for determining the mode i static fracture toughness using semi-circular bend spe-
tion with the rock grain size. Hence, the physical meaning of char- cimen. Rock Mech Rock Eng. 2014;47(1):267–274.
acteristic material length can be considered as of the scaled average 29. Zhou YX, Xia K, Li XB, et al. Suggested methods for determining the dynamic
strength parameters and mode-i fracture toughness of rock materials. In: Ulusay R,
grain size of rocks.
ed. The ISRM Suggested Methods for Rock Characterization, Testing and Monitoring:
Based on the successful application of the non-local theory, a two- 2007-2014. Cham: Springer International Publishing; 2015:35–44.
parameter tensile failure model for general tensile failure was proposed. 30. Dai F, Chen R, Xia K. A semi-circular bend technique for determining dynamic
This model provides an alternative and effective approach to determine fracture toughness. Exp Mech. 2010;50(6):783–791.
31. Dai F, Xia K, Luo SN. Semicircular bend testing with split Hopkinson pressure bar for
the flexural tensile strength of rocks with different bending configura- measuring dynamic tensile strength of brittle solids. Rev Sci Instrum.
tions. It can also be widely applied to assess the rock structure subjected 2008;79(12):123903–123906.
to bending load without conducting a series of bending tests. 32. Nasseri MHB, Mohanty B. Fracture toughness anisotropy in granitic rocks. Int J Rock
Mech Min. 2008;45(2):167–193.
33. Nasseri MHB, Tatone BSA, Grasselli G, Young RP. Fracture toughness and fracture
Acknowledgments roughness interrelationship in thermally treated westerly granite. Pure Appl Geophys.
2009;166(5):801–822.
34. Cohn SN, Ahrens TJ. Dynamic tensile strength of lunar rock types. J Geophys Res:
This research was supported by the Natural Sciences and Solid Earth. 1981;86(B3):1794–1802.
Engineering Research Council of Canada (NSERC) through the 35. Nasseri MHB, Schubnel A, Benson PM, Young RP. Common evolution of mechanical
Discovery Grant # 72031326. This work was partially supported by and transport properties in thermally cracked westerly granite at elevated hydro-
static pressure. Pure Appl Geophys. 2009;166(5):927–948.
Mitacs through the Mitacs Accelerate program.
36. Rutter EH, Neumann DHK. Experimental deformation of partially molten Westerly
granite under fluid-absent conditions, with implications for the extraction of granitic
References magmas. J Geophys Res: Solid Earth. 1995;100(B8):15697–15715.
37. Sano O, Kudo Y, Mizuta Y. Experimental determination of elastic constants of Oshima
granite, Barre granite, and Chelmsford granite. J Geophys Res: Solid Earth.
1. Bieniawski Z, Hawkes I. Suggested methods for determining tensile strength of rock 1992;97(B3):3367–3379.
materials. Int J Rock Mech Min Sci Geomech Abstr. 1978;15(3):99–103. 38. Knopoff L. Seismic wave velocities in westerly granite. Eos, Trans Am Geophys Union.
2. Xia K, Yao W, Wu B. Dynamic rock tensile strengths of Laurentian granite: experi- 1954;35(6):969–973.
mental observation and micromechanical model. J Rock Mech Geotech Eng. 39. Bhat HS, Sammis CG, Rosakis AJ. The micromechanics of Westerley granite at large
2017;9(1):116–124. compressive loads. Pure Appl Geophys. 2011;168(12):2181–2198.
3. Coviello A, Lagioia R, Nova R. On the measurement of the tensile strength of soft 40. Yong C, Wang Cy. Thermally induced acoustic emission in Westerly granite. Geophys
rocks. Rock Mech Rock Eng. 2005;38(4):251–273. Res Lett. 1980;7(12):1089–1092.
4. Hudson JA. Tensile strength and ring test. Int J Rock Mech Min Sci Geomech Abstr. 41. Burger H, Skala W. Comparison of sieve and thin-section technique by a Monte-Carlo
1969;6(1):91–97. model. Comput Geosci. 1976;2(2):123–139.
5. Dai F, Xia KW, Tang LZ. Rate dependence of the flexural tensile strength of 42. Schäfer A, Teyssen T. Size, shape and orientation of grains in sands and sand-
Laurentian granite. Int J Rock Mech Min Sci Geomech Abstr. 2010;47(3):469–475. stones—image analysis applied to rock thin-sections. Sediment Geol.
6. Hudson J, Brown E, Rummel F. The controlled failure of rock discs and rings loaded 1987;52(3):251–271.
in diametral compression. Int J Rock Mech Min Sci Geomech Abstr. 43. Hassan R, Cohanim B, De Weck O, Venter G. A comparison of particle swarm opti-
1972;9(2):241–248. mization and the genetic algorithm. In: Proceedings of the 1st AIAA Multidisciplinary
7. Zhu WC, Tang CA. Numerical simulation of Brazilian disk rock failure under static Design Optimization Specialist Conference; 2005: 18–21.
and dynamic loading. Int J Rock Mech Min Sci Geomech Abstr. 2006;43(2):236–252. 44. Chiu C-C, Cheng Y-T, Chang C-W. Comparison of particle swarm optimization and
8. Van de Steen B, Vervoort A. Non-local stress approach to fracture initiation in la- genetic algorithm for the path loss reduction in an urban area. J Appl Sci Eng.
boratory experiments with a tensile stress gradient. Mech Mater. 2012;15(4):371–380.
2001;33(12):729–740. 45. Panda S, Padhy NP. Comparison of particle swarm optimization and genetic algo-
9. ASTM-C293/C293M-10. Standard Test Method for Flexural Strength of Concrete (Using rithm for FACTS-based controller design. Appl Soft Comput. 2008;8(4):1418–1427.
Simple Beam With Center-Point Loading). ASTM International; 2010:1–3. 46. Roberge V, Tarbouchi M, Labonte G. Comparison of parallel genetic algorithm and
10. Dai F, Xia K, Zuo JP, Zhang R, Xu NW. Static and dynamic flexural strength aniso- particle swarm optimization for real-time UAV path planning. IEEE Trans Ind Inform.
tropy of Barre granite. Rock Mech Rock Eng. 2013;46(6):1589–1602. 2013;9(1):132–141.
11. Mellor M, Hawkes I. Measurement of tensile strength by diametral compression of 47. Villarroel RD, Garcia DF, Davila MA, Caicedo EF. Particle swarm optimization vs
discs and annuli. Eng Geol. 1971;5(3):173–225. genetic algorithm, application and comparison to determine the moisture diffusion
12. Lajtai EZ. Effect of tensile stress gradient on brittle-fracture initiation. Int J Rock Mech coefficients of pressboard transformer insulation. IEEE Trans Dielectr Electr. Insul.
Min Sci Geomech Abstr. 1972;9(5):569–578. 2015;22(6):3574–3581.
13. Zhao J, Li HB. Experimental determination of dynamic tensile properties of a granite. 48. Kennedy J, Eberhart R Particle swarm optimization. In: Proceedings of the IEEE

17
W. Yao et al. International Journal of Rock Mechanics and Mining Sciences 110 (2018) 9–18

International Conference on Neural Networks, Vols 1-6. 1995: 1942–1948. layer perceptron algorithm for bivariate fractal analysis of rock fractures roughness.
49. Zhang YL, Gallipoli D, Augarde C. Parallel hybrid particle swarm optimization and Int J Rock Mech Min Sci Geomech Abstr. 2013;60:66–74.
applications in geotechnical engineering. Lect Notes Comput Sc. 2009;5821:466–475. 52. Momeni E, Armaghani DJ, Hajihassani M, Amin MFM. Prediction of uniaxial com-
50. Yagiz S, Karahan H. Prediction of hard rock TBM penetration rate using particle pressive strength of rock samples using hybrid particle swarm optimization-based
swarm optimization. Int J Rock Mech Min Sci Geomech Abstr. 2011;48(3):427–433. artificial neural networks. Measurement. 2015;60:50–63.
51. Babanouri N, Nasab SK, Sarafrazi S. A hybrid particle swarm optimization and multi-

18

Potrebbero piacerti anche