Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377

Journal of Rock Mechanics and Geotechnical Engineering

Journal of Rock Mechanics and Geotechnical

Engineering
journal homepage: www.rockgeotech.org

Behaviour of layered sandstone under Brazilian test conditions:

Layer orientation and shape effects
Abbass Tavallali ∗ , André Vervoort
Department of Civil Engineering, University of Leuven, Kasteelpark Arenberg 40, 3001 Leuven, Belgium

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

Article history: The experimental study in this paper focuses on the effects of the layer orientation and sample shape on
Received 3 July 2012 failure strength and fracture pattern of samples tested under Brazilian test conditions (i.e. diametrical
Received in revised form 6 November 2012 loading of cylindrical discs) for one particular layered sandstone which is from Modave in the south of
Accepted 2 January 2013
Belgium. The variations of the strength in combination with the failure patterns are examined as a function
of the inclination angle between the layer plane and the loading direction. The experimental, results clearly
Keywords:
show that the induced fracture patterns are a combination of tensile and/or shear fractures. In shape effect
Layered sandstone
experiments the layer thickness and the number of layer boundaries are investigated. Different blocks
Brazilian test
Fracture pattern
of Modave sandstone are used to prepare samples. The layer thickness is different among the various
Layer activation blocks, but the layer thickness in each studied rock block can be considered to be constant; hence, the
Shape effect number of layer boundaries changes according to the sample diameter for samples of the same block. The
experimental study shows that the layer thickness plays a more important role than the number of layer
boundaries per sample.
© 2013 Institute of Rock and Soil Mechanics, Chinese Academy of Sciences. Production and hosting by
Elsevier B.V. All rights reserved.

1. Introduction Experimental methods are the most important and reliable way
to study rock properties and rock failure mechanisms. The failure
The knowledge of the anisotropic properties of rock material mechanism of rock materials due to their mineralogical texture is
is required in the fields of civil, mining, geoenvironmental and complex and it is more problematic for anisotropic rock.
petroleum engineering, as in general, rock material near the earth’s In rock mechanics, the uniaxial (unconfined) compressive
surface is anisotropic. Material properties of anisotropic rock are strength (UCS) test and Brazilian tensile strength (BTS) test are
directionally dependent, as opposed to isotropic rock that implies the most widely used methods to obtain rock strength properties.
identical properties in all directions. An additional characteristic of Change in rock strength properties (e.g. UCS and BTS) with sample
rock material is the large spatial variation of its properties, even at size is referred to as scale effect. In this context, the so-called scale
small distance, which makes a rock mechanical study even more effect is divided into two categories: size effect and shape effect
difficult. More and more this anisotropy is taken into account when (Thuro et al., 2001; Pan et al., 2009). The size effect can be stud-
testing the rock material, as input for design and for numerical ied by comparing the rock strength values from specimens with
modelling. The engineering applications without considering the the same shape but different sizes (the same length/diameter ratio
rock anisotropic behaviour produce errors of different magnitudes, but with different diameters). The shape effect can be studied by
depending on the extent of rock anisotropy (Amadei, 1996; Cho comparing the rock strength values from specimens with the same
et al., 2012). diameter (or length) but a variation in length/diameter ratio (see
further, Thuro et al., 2001; Pan et al., 2009).
In this study, the focus is on stratified rock material as one spe-
cific form of anisotropy that is often present in geological material.
∗ Corresponding author. Tel.: +32 16321171. Transverse isotropy (as a form of anisotropy) is a characteristic
E-mail address: [email protected] (A. Tavallali). of intact foliated metamorphic rocks (slates, gneisses, phyllites,
Peer review under responsibility of Institute of Rock and Soil Mechanics, Chinese
schists) and intact laminated, stratified or bedded sedimentary
Academy of Sciences.
rocks (shales, sandstones, siltstones, limestones, coal, etc.) (Chen
et al., 1998).
The tests are conducted on hard and brittle stratified sand-
stone (psammite) and the Brazilian method (diametrical loading of
cylindrical discs) is applied. The stratified (layered or banded) rock
1674-7755 © 2013 Institute of Rock and Soil Mechanics, Chinese Academy of
Sciences. Production and hosting by Elsevier B.V. All rights reserved.
material is sometimes called transversely isotropic rock material
http://dx.doi.org/10.1016/j.jrmge.2013.01.004 (Ye et al., 2009), i.e. rocks with one dominant direction of planar
 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377 367

anisotropy. The effects of the layer orientation and sample shape

are investigated in detail for 123 samples from three blocks taken
from the same quarry.
The experimental research in this paper focuses on the effect of
the layer orientation on both the failure strength and the failure
pattern of this layered sandstone under Brazilian test conditions.
The research shows clearly that the induced fracture patterns are
a combination of tensile and/or shear fractures. Or in other words,
the failure in transversally isotropic rock material is not necessar-
ily a pure tensile failure; even in some cases it can be pure shear
failure, but often it is a combination of tensile and shear failure
(Debecker and Vervoort, 2009; Cho et al., 2012; Vervoort et al.,
2012). Hence, one should rather talk about failure strength or peak
load than about tensile strength. The effect of the layer orientation
is investigated in detail for 89 samples from three blocks taken from
the same quarry. This investigation concentrates on the strength Fig. 1. Disc-shaped sample and configuration of layers in Brazilian tests.  varies
and fracture pattern of the tested samples. between 0◦ and 90◦ . Set of inclined parallel lines symbolises average layer direction
In Section 4, the effect of sample shape on the failure behaviour (bedding).
of (disc-shaped) samples after conducting a Brazilian test is inves-
tigated for the same material. In shape effect experiments, the using a loading machine with a loading capacity of 100 kN, at a
layer thickness (the distance between the layer boundaries) and the constant displacement rate of 1 mm/min. The disc-shaped samples
number of layer boundaries are taken into account. Similar to Sec- are positioned as in Fig. 1 and compressed vertically till failure.
tion 3, shape effect on the strength and fracture pattern of the tested The indirect tensile strength,  t , perpendicular to the loaded
samples is investigated. It should be noted that in the literature one diameter, based on linear elastic calculations for homogeneous and
mostly focuses on the sample strength (Rocco et al., 1999; Thuro isotropic rock can be written as
et al., 2001; Yoshinaka et al., 2008; Pan et al., 2009). Recently, effort
2F
has been made to shed light on correlation of strength and frac- t = (1)
␲Dt
ture pattern (Debecker and Vervoort, 2009; Tavallali and Vervoort,
2010a,b). where F is the applied load when the sample fails; and D and t are
Three blocks of layered sandstone are taken from a quarry in the diameter and thickness of the disc-shaped sample. Of course
Modave in the south of Belgium. These blocks are taken at very this formula is considered for the loaded diameter in the condition
short distance (from 1 m to 10 m) from each other. Visual inspec- of a typical vertical splitting or extension type of fracture occurring,
tion (hand samples) would normally lead to the conclusion that which is often not the case for transversally isotropic rock material
they are all similar material. This sandstone is also known as “Psam- (see further, Figs. 4 and 7). However, it is even not straightforward
mite of Condroz” (Poty and Chevalier, 2004). This sandstone is to judge that the failure through an inclined fracture or through a
characterised by numerous thin and parallel layers. The plane of layer boundary is in tension or shear (Vervoort et al., 2012). There-
anisotropy (transverse isotropy) is assumed to be parallel to the fore, this formula is used just for comparison purposes and the
apparent direction of rock layer. results from Eq. (1) do not necessarily present the tensile strength
The sandstone studied is stratified and massive, and belongs to of the samples (further explanations are presented in Section 3).
the Famennian stage. The Famennian stage is one of the two stages
in the Late Devonian epoch. It lasted from 374.5 ± 2.6 million years 2. Material properties
to 359.2 ± 2.5 million years ago (Gradstein et al., 2004). The name of
Famennian stage originated in the 19th century from the Famenne 2.1. Physical properties
region in the southern part of Belgium.
The number of layer boundaries on hand samples from each The dry density and effective porosity of the Modave sandstone
block is counted. The number of layer boundaries over 1 cm of blocks are determined by saturation and buoyancy techniques. For
three blocks is not the same; therefore, the three blocks are fur- each block (sub-type) of sandstone, five samples are prepared (see
ther called the three sandstone sub-types a, b and c. The number Table 1). The variation in dry density is relatively small, i.e. from
of layer boundaries over 1 cm is 5.4, 2.6 and 1.4 for the three sub- 2.57 g/cm3 for sub-type b to 2.63 g/cm3 for sub-type a. The variation
types. Graded bedding of shallow-water environments is generally in the effective porosity is from 2.0% for sub-type a to 3.2% for sub-
composed of thin layers from a few millimetres to 1 cm or 2 cm, and type b.
seldom makes sequences more than 10 cm or 20 cm thick (Reineck Also two disc-shaped samples from each sub-type are prepared
and Singh, 1986). It can be concluded that all the blocks are from for P-wave velocity measurements. P-wave velocity is measured
shallow-water environments as their layer thicknesses are 1.8 mm, parallel to layer direction and in the direction perpendicular to
3.5 mm and 6.4 mm for sandstone sub-types a, b and c, respectively. the layers. In different configurations, slight differences of wave
In sub-type c, ripples are observed (Tavallali and Vervoort, velocity are recorded. However, the minimum and the maximum
2008). It should be noted that ripples are not considered as the
layer boundary. The reason is that ripples are not in the main direc- Table 1
tion of sedimentation and that they are dispersed between layer Physical properties of the three sub-types of layered Modave sandstone.
boundaries. Sub-type Dry density (g/cm3 ) Porosity (%) P-wave velocity (km/s)
The rock is cored in the laboratory by using a 50 mm diameter
drill bit for layer orientation effect. For shape effect investigation, Perpendicular Parallel

drill bits with different diameters from 30 mm to 115 mm are used. a 2.62–2.63 2.0–3.0 4.5–4.7 4.9–5.0
The direction of coring is parallel to the layers. For all samples a b 2.57–2.58 2.8–3.2 4.3–4.4 4.5–4.6
c 2.58–2.61 2.1–3.0 4.5–4.6 5.0
constant thickness of 25 mm is taken. All tests are carried out by
368 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377

wave velocities are always measured perpendicular and parallel to groups (Reineck and Singh, 1986). The grains of the three sub-types
the layers, respectively (see Table 1). The minimum difference of are in the class of subangular.
wave velocities is measured in samples of sub-type b. In this sub- For quartz grain size measurements, the point-counting method
type, wave velocity perpendicular to the layers is from 4.3 km/s to is applied too. Three hundred points in each thin section are
4.4 km/s while in direction parallel to the layers it is from 4.5 km/s counted in a similar pattern to the composition quantification.
to 4.6 km/s. Similar to sieving analysis the smallest dimension of the grain is
considered as the grain size. The average (and minimum, maxi-
2.2. Petrographical study of untested samples mum) values (in ␮m) of the quartz grains size in sub-types a, b and
c are 69 (22, 141), 107 (42, 218), and 82 (29, 146), respectively. The
For petrographical purpose, thin sections with a thickness of grain size calculation is done by arithmetic method. By considering
30 ␮m from an untested sandstone sample from each sub-type the British Standards (Craig, 1990), the grains of studied sub-types
are prepared and analysed. Thin section observation is done by are in the size range of fine sand. The measurements are in agree-
petrographical microscope with transmitted polarised light. In all ment with the study of Thorez et al. (2006) of the sandstone in the
samples, quartz, carbonate and mica are observed. Pores, organic same area. They have mentioned that the sandstone of Famennian
material, clay minerals and feldspar are also observed, although in stage in the south of Belgium is fine grained with the average grain
very small quantities. Clays are found usually mixed with micro- size of 45–120 ␮m.
scopic crystals of carbonates, feldspars, micas and quartz. These The nature of minerals suggests that the presence of strong ele-
minerals are very common in fine grain sedimentary rocks such as ments such as quartz increases the rock strength and, on the other
the studied sandstone. Determining precisely the clay mineral by hand, the abundance of weak minerals such as mica and carbonate
means of X-ray analysis and some spectroscopic methods is pos- decreases the rock strength. It is widely accepted that for higher
sible. By referring to some studies, it is understood that the clay quartz content, rock strength increases. Contrary to quartz, the
minerals in the studied sandstone are dominated by illite (Thorez abundance of easily cleavable minerals would result in lowering the
et al., 1988; Han et al., 2000). strength values (Tavallali, 2010). It should be noted that sub-type
To quantify the relative amount of minerals, the point-counting a has the highest weak minerals percentage (23.5%) and sub-type
method is applied (Chayes, 1956; Přikryl, 2001). The thin section is b has the lowest weak minerals percentage (7.9%).
put in the point-counting apparatus which is attached to the petro-
graphical microscope. Above-mentioned apparatus moves the thin 3. Brazilian test results
section with a given interval. After each movement, the mineral in
the centre of the microscopic view is observed. For example, if in The rock layers are inclined at different angles (see Fig. 1) ran-
100 steps of movement 75 times a quartz grain is observed in the ging between 0◦ (perpendicular to the loading direction) and 90◦
centre of microscopic view, it is concluded that quartz content in (parallel to the loading direction). It should be highlighted that
that rock material is 75%. the orientation of the layers with respect to loading direction is
In this study, 500 points in each thin section are counted in a evaluated. As the loading direction in this study is always vertical,
regular pattern with a fixed interval of 400 ␮m between the points. changing the layer orientation implies that the angle between the
Direction of counting is perpendicular to the layers. The distance of layer direction and the loading direction changes. Seven different
800 ␮m is taken between the lines of point-counting. Fig. 2 presents values are considered: 0◦ , 20◦ , 45◦ , 60◦ , 70◦ , 80◦ and 90◦ . Since one
the percentage of different minerals in each sub-type. sample is not necessarily representative for the failure behaviour
In the three Modave sandstone sub-types, the main constitutive corresponding to a specific inclination angle, four to six samples
mineral is quartz (from 72% to 88%). The shape and roundness of are tested per inclination angle value per sub-type. The number
the quartz grains are the same in all the sub-types. For roundness of tested samples for the sub-types a, b and c is 29, 30 and 30,
of sand grains, Powers (1953) and Shepard (1963) distinguished six respectively (89 samples in total).
The variation in the average failure strength with the inclination
angle  for the sub-types is presented in Fig. 3. All the curves show a

Fig. 2. Observed mineral percentage in different sandstone sub-types by 500

counted points. Category ‘Others’ corresponds to pores, organic material and clay Fig. 3. Variation in average failure strength as a function of the inclination angle 
minerals. for sandstone sub-types a, b and c.
 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377 369

Fig. 4. Schematic representation of different fracture types in Brazilian test. (1)

Layer activation, (2) central fracture, and (3) non-central fracture.

small downwards trend. The largest average failure strength corre-

sponds mostly to the samples of sub-type b which varies between
15.1 MPa and 11.3 MPa, while the lowest corresponds mostly to the
samples of sub-type a which varies between 14.2 MPa and 5.3 MPa.
However, it is important to note that when  = 0◦ the average
failure strength corresponds to the tensile strength of the “intact”
material. For  = 90◦ , the average failure strength should rather cor-
respond to the tensile strength of the “layers”. For other angles, one
can probably not talk about a pure tensile or extension type of fail-
ure, and hence about a tensile strength. One should rather consider
it as failure strength.
Similar to the idea introduced by Szwedzicki (2007) for UCS
tests, different failure modes are suggested. By considering the
samples after failure, different types of fractures are observed (see
Fig. 4):

(1) Some fractures are parallel to the isotropic layers which are
further called “layer activation”.
(2) Some fractures are roughly parallel to the loading direction and
they are located in the central part of the sample between the
two loading lines. The central part is arbitrarily defined as 10%
of the diameter on both sides of the central line. These fractures
are further called “central fractures”.
(3) Fractures outside the central part are also observed. If they do
not correspond to layer activation, they are further called “non-
central fractures”. The latter are often curved lines, starting at
or around the loading platens.

In most cases, two or three different fracture types occur in the

same experiment (Tavallali et al., 2007; Vervoort et al., 2012). It is
good to note that in case of  = 90◦ and when a straight fracture
parallel to the layers between both loading platens is induced, this
is classified as layer activation and not as central fracture. However,
if in that case ( = 90◦ ) a fracture is situated in the central part but
does not follow the layer direction, it is classified as central fracture.
The fracture length within the various rock samples is esti-
mated and used as a tool to distinguish between the different failure
modes. First, the lengths of all fractures in the sample are measured.
Second, the length corresponding to the various fracture types is
considered. When a fracture is classified as layer activation, the
entire length including the part within the central zone is put into
the total sum of layer activation. In a similar way, if the major part
of a curved shape fracture (e.g. the non-central fracture in Fig. 4)
is outside the central zone, the entire length including the small
portion in the central zone is put into the total sum of non-central
fractures. Fig. 5. Variation in average fracture length as a function of the inclination angle ,
The fracture pattern in most of the samples is similar on both for Modave sandstone sub-types a, b and c.
sides (i.e. front and back; however they are not exactly the same).
It means that the fracture pattern can mostly be considered as two-
dimensional. Therefore, all the fracture lengths in these cases are For all samples corresponding to a single angle value, the aver-
measured on one side. The fracture pattern in a few samples is age fracture length is calculated. For the samples of the three
truely three-dimensional and on both sides of the sample, the frac- sub-types, variation of the three types of fractures, as well as the
ture pattern is significantly different. In such cases, averages of the total fracture length is plotted as a function of the inclination angle
fracture lengths on both sides are considered.  in Fig. 5.
370 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377

The graph corresponding to sub-type a (Fig. 5a) shows that, for

an inclination angle less than 45◦ , most of the fractures are cen-
tral fractures, however, for an inclination angle of 0–70◦ , central
fracture(s) is the main failure mode. Although, for an inclination
angle of 0–90◦ , there is an increasing trend for layer activation,
only predominant for 80◦ and 90◦ .
The graph corresponding to sub-type b (Fig. 5b) shows that,
for inclination angles smaller than 50◦ , the fracture length cor-
responding to layer activation is negligible. When the inclination
angle exceeds 50◦ , fracture length corresponding to layer activation
increases and only for 90◦ becomes predominant. Layer activation
for the inclination angle of 0◦ and 20◦ for sub-type c is negligible.
For larger inclination angles, layer activation increases, but only for
90◦ it becomes the main failure mode in sub-type c.
Fig. 5 also illustrates that the non-central fracture length for the
three sub-types is small and independent of .
Recently, detailed investigation on fracture-strength correla-
tion for layered sandstone has been done (Tavallali and Vervoort,
2010a). They concluded that larger fracture length correlates with
higher strength. They also concluded that strength is smaller for
larger inclination angles and for smaller total fracture lengths.
Therefore, it is interesting to correlate Fig. 3 with Fig. 5. The varia-
tion on failure strength values as a function of total fracture length
for all the samples of the three Modave sandstone sub-types is
plotted in Fig. 6. Fig. 6 indicates that the samples showing higher
strength are more fractured (especially for sub-type a).
It can be concluded that, for all sandstone sub-types, as
inclination angle increases from 0◦ to 90◦ , the fracture length cor-
responding to central fracture(s) decreases while layer activation
increases. However, the decreasing and the increasing rates of
fractures length for the three sandstone sub-types are different.
Consequently, the transition angle, which indicates the change of
failure mode from central fracture to layer activation, is not the
same for all sub-types of studied layered sandstone. It is important
to note that the number of layer boundaries over one centimetre of
the three sub-types is not the same. Up to now, it is understood that
the failure strength and fracture pattern are considerably affected
by the layer orientation. Therefore, by focusing on one inclination
angle (i.e. 70◦ ) for the three sub-types, their failure behaviour for
the same layer orientation could be compared to each other in more
detail. For the inclination angle of 70◦ , fracture length mainly cor-
responds to both central fracture and layer activation. It can be
useful to consider also the percentage of central fracture(s) and
layer activation compared to the total fracture length. The largest
percentage of either fracture in the form of layer activation or as
central fracture(s) indicates the predominant failure mode.
For the inclination angle of 70◦ , the average failure strength val-
ues of sandstone sub-types a, b and c are 7.9 MPa, 14.5 MPa and
12.8 MPa, respectively. At a macroscopic scale, the blocks look sim-
ilar, but there is about 6.6 MPa difference between the smallest and
Fig. 6. Variation in failure strength as a function of total fracture length for all the
largest average values of failure strength. In other words, it can samples of the three layered Modave sandstone sub-types.
be concluded that considerable difference exists among the failure
strength values, as the average value of failure strength for sub-type
b (14.5 MPa) is about two times that for sub-type a (7.9 MPa). in Fig. 7, it is understood that the fracture patterns of samples from
The failure pattern (one side) of all the samples (that are tested the three sub-types of layered sandstone are different.
under an angle of 70◦ ) is shown in Fig. 7 for the three sandstone In Fig. 8, the average percentage of layer activation, central
sub-types. Layer activation in three of the four samples of sub-type fracture(s) and non-central fracture(s) compared to the total frac-
a is observed (the third sample from left of Fig. 7a: minor layer ture length is given for the three sandstone sub-types. Central
activation in the central part). In two of them, layer activation is fracture(s) is predominant for all three sub-types of Modave sand-
the predominant failure mode. On the other hand, central fracture stone. However, layer activation for sub-type a (41%) is more than
is the predominant failure mode for all the sandstone samples of two times that for sub-type b (19%). Minimum layer activation is
sub-type b and only in one sample layer activation is observed as observed in sub-type b. It should be noted that this sub-type shows
the secondary failure mode. The samples from sub-type c show a the smallest difference in P-wave velocities measured between the
failure by either central fracture(s) only or a combination of central direction perpendicular and parallel to the layers. In other words,
fracture(s) and layer activation. By observing the fracture patterns sub-type b has the lowest degree of transverse isotropy.
 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377 371

Fig. 7. Observed failure patterns of samples from different layered Modave sandstone sub-types in Brazilian tests. Inclination angle of 70◦ for all samples. Set of inclined
parallel lines symbolises average layer direction (bedding). The predominant and secondary failure mode are given in parentheses, CF: central fracture, LA: layer activation.

The conducted Brazilian tests illustrate the effect of the layer

orientation on the fracture patterns and failure strength on macro-
scale. However, from the above explanation, it can be concluded
that samples from different blocks (which are relatively similar, but
not completely similar) do not behave similarly on macro-scale.

4. Shape effect

In this part of the study, the effect of sample shape on the failure
behaviour of disc-shaped samples after conducting Brazilian tests
is investigated for Modave sandstone. To be sure that the same
material is tested, the samples coming from each block are consid-
ered separately. So, in one way, with this condition, the amount of
rock material is limited. Therefore, size effect experiments are not
possible, as for the same length/diameter (or thickness/diameter)
ratio a large amount of rock material is needed for large diameters.
Therefore, only the shape effect is investigated.
Fig. 8. Variation in fracture length percentage corresponding to layer activation,
central fracture(s) and non-central fracture(s) for different sandstone sub-types.
It is observed that the number of layer boundaries (e.g. over
1 cm) of the three sandstone sub-types is not the same, and also
their failure patterns are different. It is essential to know whether
the number of layer boundaries per sample is important or the
372 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377

Fig. 9. Observed failure patterns (one side) in Brazilian test for samples in different diameters from Modave sandstone sub-types a, b and c (one typical sample for each
diameter). Inclination angle for all samples is 70◦ . Sample number and the predominant failure mode(s) are put under the samples. The first part of the numbers is the sample
diameter in mm. CF: central fracture, NCF: non-central fracture, LA: layer activation, #LB: number of layer boundaries.

distance between the layers plays the main role in failure sandstone sub-type b (2.6). It means that a sample from sub-type
behaviour. By preparing larger samples, the number of layer a with “D” in diameter has the same number of layer boundaries as
boundaries per samples increases; however, the distance between a sample from sub-type b with “2D” in diameter. However, the dis-
them does not change. For example, the number of layer bound- tance between the layer boundaries in the sample from sub-type
aries over 1 cm for sub-type a (5.4) is about two times that for b is two times larger than that for the sample from sub-type a.
 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377 373

Table 2
Quantity of sandstone samples for shape effect analyses.

Sandstone sub-type Samples quantity

D = 30 mm D = 40 mm D = 50 mm D = 80 mm D = 100 mm D = 115 mm Total

a 4 4 4 2 2 – 16
b 4 – 4 3 – 4 15
c 4 – 4 – 3 4 15

The aim of the following experiments can be summarised by the An effort has been made to answer these questions. For this
following two questions: purpose, several samples with different diameters from different
sub-types of sandstone should be tested.
The blocks of sandstone sub-types a, b and c are cored in the
(1) What is the effect of sample diameter on failure behaviour of laboratory by using drill bits with different diameters from 30 mm
disc-shaped samples from a particular sandstone sub-type? to 115 mm. The direction of coring is parallel to the layering. For
(2) Is there any similarity in failure pattern of samples from differ- all samples a constant thickness of 25 mm is taken. Therefore,
ent sub-types which have (roughly) the same number of layer the thickness/diameter ratio varies from 0.2 to 0.9. Variation in
boundaries? mentioned ratio in some studies is executed by changes in the

Fig. 10. Shape effect on failure behaviour of disc-shaped samples after conducting Brazilian test. Samples with different diameters but constant thickness (25 mm) from
sandstone sub-types a, b and c are tested. The inclination angle for all the samples is 70◦ .
374 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377

thicknesses, while keeping the diameter as a constant value. How-

ever, for the purposes of this study, changes in diameter are needed
(in order to have different numbers of layer boundaries per sample),
while the samples thickness is kept constant.
Some important notes should be considered. In Brazilian tests,
the minor and major principal stresses are tensile and compres-
sive, respectively. The intermediate principal stress depends on
length/diameter ratio. Lower length/diameter ratio makes the state
of the rock sample (disc) closer to plane stress rather than plane
strain (Lavrov et al., 2002). The intermediate principal stress for
the low length/diameter ratio can be assumed zero; while, it is not
the case when this ratio is larger.
From the three mentioned sandstone sub-types, 46 samples are
prepared for conducting Brazilian tests in order to investigate the
shape effect and also to find the answer to the two mentioned ques-
tions. All the samples are tested under an inclination angle of 70◦ .
Sample quantities with different diameters from sandstone sub-
types a, b and c are presented in Table 2. It should be noted that all
samples with 50 mm are already tested, presented and described
in Section 3. These samples are integrated in this part of the study.
The failure pattern (one side) of the samples after failure from
the three sandstone sub-types is shown in Fig. 9 (one typical sample
for each diameter). Apart from the predominant failure mode in
some samples, such as in sandstone sub-type a, some secondary
modes are also observed (see Fig. 9). Although in samples 80-2 and
115-3 of sandstone sub-types b some parts of layer activation are
also observed, the predominant failure mode in these samples is
central fracture. In these samples, layer activation is not considered
as a secondary failure mode, because fracture length corresponding
Fig. 11. Shape effect on failure behaviour of disc-shaped samples after conducting
to layer activation is very small in comparison to fracture length
Brazilian test. Average values are calculated from 2 to 4 samples for each diameter.
corresponding to central fracture. Samples with different diameters but constant thickness (25 mm) from the three
After conducting the tests the fracture analysis is done for all sandstone sub-types are tested. The inclination angle for all the samples is 70◦ .
the samples. To observe the effect of sample diameter on failure
behaviour of disc-shaped samples, relative layer activation fracture
length and also the failure strength are considered. sub-types a, b and c. As stated earlier, Fig. 11 shows that the shape
Fig. 10 presents the variation in the mentioned two factors (rel- effect on failure behaviour of disc-shaped samples is limited,
ative layer activation fracture length and failure strength) as a particularly on failure strength in the tested diameter range.
function of sample diameter for sandstone sub-types a, b and c. Thuro et al. (2001) studied the shape effect of limestone samples
For sandstone sub-types b and c, a clear difference in layer activa- by conducting Brazilian tests. For Brazilian tests in their study, the
tion and also in failure strength of samples with different diameters diameter of 70 mm is taken and length/diameter ratio varies from
is not distinguished (see Fig. 10). 0.5 to 2. It can be concluded that the stress state of samples in their
Variation in failure strength values as a function of sample diam- study varies from plane stress to plane strain condition by increas-
eter for sandstone sub-type a is also not significant, especially when ing the length/diameter ratio. Consequently, the tensile strengths
sample diameter varies between 50 mm and 100 mm. The condi- of all mentioned samples are not comparable to each other, as they
tion for relative layer activation fracture length of samples from are not in the same stress state condition.
sandstone sub-type a is slightly different. The samples with diam- However, in the results by Thuro et al. (2001), if one compares
eter of 30 mm, 40 mm and 50 mm have similar layer activation; the samples with low length/diameter ratio (plane stress condition)
however, for larger diameter (80 mm and 100 mm), layer activa- with each other, a significant shape effect cannot be observed. This
tion is slightly increased. For each of the diameters 80 mm and is also the case for the high length/diameter ratio (plane strain con-
100 mm from sandstone sub-type a, two samples are tested (see dition). Therefore, from above discussion it can be concluded that
Table 2). Relative layer activation fracture length of one sample in shape effect in Brazilian tests for the rock samples with the same
each of these two diameters is in the range of layer activation of degree of homogeneity is not dominant.
the other samples with diameters of 30 mm, 40 mm and 50 mm. In general, it can be stated that dominant scale effect in Brazil-
In other words, only one sample with diameter of 80 mm and one ian tests for the rock samples with a limited size variation is not
sample with diameter of 100 mm show higher relative layer acti- observed, because they have the same degree of homogeneity.
vation fracture length. Therefore, it is difficult to judge the layer However, for rock samples with a large size variation, the degree of
activation trend in sandstone sub-type a as sample diameter is heterogeneity is different and consequently scale effect is observed.
increased. From the above explanation it can be concluded that the By observing the results shown in Fig. 11, another interesting
shape effect is limited in the tested diameter range for the studied conclusion is observed. Failure strength is inversely correlated with
sandstone sub-types. relative layer activation fracture length for the tested diameter
Mean values of relative layer activation fracture length and of range. Samples of sandstone sub-type a have the highest relative
failure strength are calculated from 2 to 4 samples for each diam- layer activation fracture length and the lowest failure strength in
eter in sandstone sub-types a, b and c. Fig. 11 shows the variation the tested diameter range. Samples of sandstone sub-type b have
in mean values of relative layer activation fracture length and also the lowest relative layer activation fracture length and the high-
failure strength as a function of sample diameter for sandstone est failure strength. The inverse correlation of failure strength and
 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377 375

Table 3
Number of layer boundaries per sample in the tested diameter range for the three sandstone sub-types.

Sandstone sub-type #LB/cm Number of layer boundaries per sample

D = 30 mm D = 40 mm D = 50 mm D = 80 mm D = 100 mm D = 115 mm

a 5.4 15.7 21.1 27.0 43.2 54.0 –

b 2.6 7.5 – 13.0 20.8 – 29.6
c 1.4 4.1 – 7.0 – 14.0 16.0

Fig. 12. Correlation of average failure strength and average relative layer activation
fracture length for the three sandstone sub-types.

Fig. 13. Variation in relative layer activation fracture length as a function of number
relative layer activation fracture length for the tested samples is
of layer boundaries per sample. Items 1 to 5 show the samples with (roughly) the
clearly presented by Fig. 12. This figure shows that higher strength same number of layer boundaries (see Table 4).
corresponds to the samples with lower layer activation fracture
length.
As can be seen in Fig. 11, although the number of total layer sandstone sub-type b and the samples with 4 cm diameter from
boundaries is larger in samples of sandstone sub-types b and c sandstone sub-type a are the other example (see Table 3 and also
by increasing the samples diameter, their relative layer activation item 4 in Table 4). Five pairs with relatively similar numbers of layer
fracture length does not change considerably. It means that the boundaries can be found in Table 3. These five pairs are presented
failure patterns of samples from different sub-types of sandstone in Table 4 as items 1–5 with their sandstone sub-types, diameters
which have (roughly) the same number of layer boundaries per and individual numbers of layer boundaries (#LB/sample).
sample are not similar. Mentioned conclusion is verified by com- Variation in relative layer activation fracture length for samples
paring the failure patterns of samples with similar number of layer with (roughly) the same number of layer boundaries (items 1–5
boundaries. in Table 4) is presented in Fig. 13. Fig. 13a shows that the effect
By knowing the number of layer boundaries per centimetre for of the number of layer boundaries on layer activation cannot be
each sub-type of sandstone, the number of layer boundaries in observed. Furthermore, this figure shows that by increasing the
each sample with different diameters is calculated. Table 3 presents number of layer boundaries per sample (by increasing the sample
the number of layer boundaries per sample in the tested diameter diameter), any positive trend of relative layer activation fracture
range for the three sandstone sub-types. From Table 3 the samples length does not exist. Fig. 13b presents the variation in average
with similar numbers of layer boundaries per sample are grouped. relative layer activation fracture length as a function of number
For example, the numbers of layer boundaries in the samples with of layer boundaries per sample in every item. In this figure, only
3 cm diameter from sandstone sub-type a is 15.7, close to the num- item 2 shows the same average relative layer activation fracture
bers of layer boundaries in the samples with 11.5 cm diameter from length for the samples with roughly the same number of layer
sandstone sub-type c (16.0). The samples with 8 cm diameter from boundaries and for the other four items no correlation can be

Table 4
Sandstone samples with (roughly) the same number of layer boundaries per sample.

Item Sandstone sub-type D (mm) #LB/sample D (mm) Sandstone sub-type Item

1 c 50 7.0 7.5 30 b 1
2 b 50 13.0 14.0 100 c 2
3 a 30 15.7 16.0 115 c 3
4 b 80 20.8 21.1 40 a 4
5 a 50 27.0 29.6 115 b 5
376 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377

sub-types a and b. In Fig. 14b, both sub-types present relatively

identical rate of relative layer activation as a function of number of
layer boundaries per sample.

5. Discussions and conclusions

The experimental study on the sandstones from Modave in

the South of Belgium, which is a layered rock, shows that the
failure stress and fracture pattern are considerably affected by
the layer orientation. However, measurements of P-wave veloc-
ity show small differences in wave velocity parallel (4.5–5.0 km/s)
and perpendicular to the layers (4.4–4.6 km/s).
The results show that larger fracture length correlates with
higher strength. Also the samples with low inclination angles which
are more fractured have the higher values of strength.
The influence of layer orientation is investigated by three sub-
types of the layered sandstone of Modave. The samples are taken
from three blocks in the same quarry, collected at close distance
(1–10 m). The global behaviour is similar; however, they do not
behave exactly the same. For all sandstone sub-types, as inclination
angle increases from 0◦ to 90◦ , the fracture length corresponding to
central fracture(s) decreases while layer activation increases. How-
ever, the decreasing and the increasing rates of fractures length for
the three studied sub-types are different.
Brazilian test results for one inclination angle (i.e. 70◦ ) for a few
samples from these three sub-types are noted. The average value
of the failure strength for the three sub-types varies from 7.9 MPa
to 14.5 MPa. Although at a macroscopic scale the blocks look sim-
ilar, there is about 6.6 MPa difference between the smallest and
largest average values of the failure strength. Central fracture(s) is
predominant for all three sub-types of Modave sandstone; how-
ever, layer activation is observed as the secondary failure mode in
sub-type a. The percentage of layer activation for sub-type a (41%)
Fig. 14. Variation in relative layer activation fracture length as a function of number is more than two times that for sub-type b (19%). From the above
of layer boundaries per sample for Modave sandstone sub-types a and b.
explanation it can be concluded that different failure behaviour of
two blocks of layered sandstone from the same quarry is observed.
observed. Comparison of Fig. 13b with Fig. 13a leads to the same In this study, sandstone sub-type b has the highest average
conclusion, i.e. increasing the number of layer boundaries by quartz size (107 ␮m) and sub-type a the lowest average (69 ␮m). It
enlarging the samples does not affect the layer activation. Of is interesting to note that the average value of the failure strength
course parameters like mineral composition are not significantly and the total fracture length are the highest for sub-type b of
different for all samples of a sub-type. Hence, it is concluded that Modave sandstone, while these two parameters are the lowest for
these parameters mainly affect the failure process. sandstone sub-type a. The latter has the lowest average grain size
Observed weak mineral percentages in sandstone sub-types a, (Tavallali, 2010; Tavallali and Vervoort, 2010b).
b and c are 23.5%, 7.9% and 12.4%, respectively. Fig. 13 shows that The shape effect in the tested diameter range (from 30 mm to
most of the samples from sandstone sub-type a are in the highest 115 mm) with a constant thickness of 25 mm is limited and the
range of layer activation while most of the samples from sub-type failure behaviour of samples is not influenced by their diameter.
b are in the lowest range. This observation is compatible with their Also no obvious similarity of failure behaviour of sandstone sam-
weak mineral percentage. Among mentioned Modave sandstone ples from different sub-types which have the same number of layer
sub-types, sub-type a has the largest amount of weak minerals boundaries per sample is observed. It means that increasing the
(23.5%) and sub-type b has the lowest (7.9%). Therefore, it is logi- number of layer boundaries per sample by enlarging the samples
cal to conclude that layer activation depends first on weak mineral does not influence failure process. By enlarging the samples (in
percentage and, hence, on the number of layer boundaries per cen- diameter range of 30–115 mm), weak mineral percentage does not
timetre. change and this is probably the main reason of having relatively
It is interesting to redraw the graphs of Fig. 13 with only the the same layer activation in each sandstone sub-type. Above expla-
samples of sandstone sub-types a and b (see Fig. 14), as in these nation can be summarised as follows: layer activation depends on
sub-types layer boundaries are the only planes of weakness. In weak mineral percentage and the number of layer boundaries plays
sandstone sub-type c apart from the layer boundaries, ripples exist only a role in the case of identical diameter. Or, in other words, the
as the secondary plane of weakness. Fig. 14a shows the variation number of layer boundaries per centimetre is more important than
in relative layer activation fracture length as a function of num- the number per sample.
ber of layer boundaries per samples for sandstone sub-types a and It is also observed that the failure strength is inversely corre-
b. As expected, Fig. 14a shows that similarity in number of layer lated with relative layer activation fracture length for the tested
boundaries of samples with different diameters does not affect the diameter range. Samples of sandstone sub-type a with the highest
relative layer activation. Fig. 14b presents the variation in average weak minerals percentage (23.5%) have the highest relative layer
relative layer activation fracture length for samples of sandstone activation fracture length and the lowest failure strength in the
 A. Tavallali, A. Vervoort / Journal of Rock Mechanics and Geotechnical Engineering 5 (2013) 366–377 377

tested diameter range. Samples of sandstone sub-type b with the Rocco C, Guinea GV, Planas J, Elices M. Size effect and boundary conditions
lowest weak minerals percentage (7.9%) have the lowest relative in the Brazilian test: experimental verification. Materials and Structures
1999;32(3):210–7.
layer activation fracture length and the highest failure strength. Shepard FP. Submarine geology. New York: Harper & Row; 1963.
Szwedzicki T. A hypothesis on modes of failure of rock samples tested in uniaxial
Acknowledgements compression. Rock Mechanics and Rock Engineering 2007;40(1):97–104.
Tavallali A. Failure of layered Modave sandstone: effect of layer orientation and
material properties. Leuven: KU Leuven; 2010 [Ph.D. thesis].
The financial support of the Research Council of the Katholieke Tavallali A, Vervoort A. Effect of layer orientation on the failure of layered sandstone
Universiteit Leuven (OT-project OT/03/35) is gratefully appreci- under Brazilian test conditions. International Journal of Rock Mechanics and
Mining Sciences 2010a;47(2):313–22.
ated. The work is part of the Ph.D. study by A. Tavallali. Authors Tavallali A, Vervoort A. Failure of layered sandstone under Brazilian test conditions:
acknowledge their sincere thanks to Prof. Rudy Swennen from the effect of micro-scale parameters on macro-scale behaviour. Rock Mechanics and
Geology Section of KU Leuven for helpful comments and also per- Rock Engineering 2010b;43(5):641–53.
Tavallali A, Vervoort A. Failure of transversely isotropic rock material: effect of layer
mitting the use of their microscopic instruments. The authors also
orientation and material properties. In: Proceedings of the 6th international
thank An De Cleyn for her helpful comments in the geological part symposium on ground support in mining and civil engineering construction.
of the study. Johannesburg: The Southern African Institute of Mining and Metallurgy; 2008.
p. 317–28.
Tavallali A, Debecker B, Vervoort A. Evaluation of Brazilian tensile strength in
References transversely isotropic sandstone. In: Proceeding of the 11th congress of the
international society for rock mechanics (ISRM 2007). London: Tailor and Francis
Amadei B. Importance of anisotropy when estimating and measuring in situ stresses Group; 2007. p. 269–72.
in rock. International Journal of Rock Mechanics and Mining Sciences & Geome- Thorez J, Dreesen R, Streel M. Famennian. Geologica Belgica 2006;9(1–2):27–45.
chanics Abstracts 1996;33(3):293–325. Thorez J, Goemare E, Dreesen R. Tide- and wave-influenced depositional envi-
Chayes F. Petrographic modal analysis: an elementary statistical appraisal. New ronments in the Psammites du Condroz (Upper Famennian) in Belgium. In:
York: Wiley; 1956. Tide-influenced sedimentary environments and facies, sedimentology and
Chen CS, Pan E, Amadei B. Determination of deformability and tensile strength of petroleum geology. Dordrecht: D. Reidel Publishing Company; 1988. p. 389–415.
anisotropic rock using Brazilian tests. International Journal of Rock Mechanics Thuro K, Plinninger RJ, Zäh S, Schütz S. Scale effects in rock strength properties,
and Mining Sciences 1998;35(1):43–61. part 1: unconfined compressive test and Brazilian test. In: Proceedings of ISRM
Cho JW, Kim H, Jeon S, Min KB. Deformation, strength anisotropy of Asan gneiss, regional symposium EUROCK 2001: rock mechanics—a challenge for society.
Boryeong shale, and Yeoncheon schist. International Journal of Rock Mechanics Espoo: A.A. Balkema; 2001. p. 169–74.
and Mining Sciences 2012;50:158–69. Vervoort A, Debecker B, Tavallali A. Fracturing of transversely isotropic rock: an
Craig RF. Soil mechanics. London: Chapman and Hall; 1990. experimental study. In: Proceedings of the 2nd southern hemisphere interna-
Debecker B, Vervoort A. Experimental observation of fracture patterns in layered tional rock mechanics symposium; 2012. p. 17–32.
slate. International Journal of Fracture 2009;159:51–62. Ye JH, Wu FQ, Sun JZ. Estimation of the tensile elastic modulus using Brazilian disc
Gradstein FM, Ogg JG, Bleeker W, Lourens LJ. A new Geologic Time Scale with special by applying diametrically opposed concentrated loads. International Journal of
reference to Precambrian and Neogene. Episodes 2004;27:83–100. Rock Mechanics and Mining Sciences 2009;46(3):568–76.
Han G, Preat A, Chamley H, Deconinck JF, Mansy JL. Palaeozoic clay mineral sedi- Yoshinaka R, Osada M, Park H, Sasaki T, Sasaki K. Practical determination of mechan-
mentation and diagenesis in the Dinant and Avesnes Basins (Belgium, France): ical design parameters of intact rock considering scale effect. Engineering
relationships with Variscan tectonism. Sedimentary Geology 2000;136:217–38. Geology 2008;96:173–86.
Lavrov A, Vervoort A, Wevers M, Napier JAL. Experimental and numerical study of
the Kaiser effect in cyclic Brazilian tests with disk rotation. International Journal
Abbass Tavallali is an Advisor Engineer at International
of Rock Mechanics and Mining Sciences 2002;39(3):287–302.
Marine and Dredging Consultants (IMDC) in Belgium. He
Pan PZ, Feng XT, Hudson JA. Study of failure and scale effects in rocks under uni-
graduated as a Civil Engineer in 1995 from Shiraz Uni-
axial compression using 3D cellular automata. International Journal of Rock
versity in Iran. He had an opportunity to participate in
Mechanics and Mining Sciences 2009;46(4):674–85.
the huge geotechnical and structural projects in Iran.
Poty E, Chevalier E. L’activité extractive en Wallonie, situation actuelle et perspec-
He gained nine years of professional experience in the
tives. Belgium: Ouvrage publié par le Ministère de la Region Wallonne; 2004.
field of dam construction and tunnelling projects prior to
Powers MC. A new roundness scale for sedimentary particles. Journal of Sedimentary
continuing his studies for Ph.D. He got his Ph.D. degree
Petrology 1953;23:117–9.
in geotechnical engineering from KU Leuven (Katholieke
Přikryl R. Some microstructural aspects of strength variation in rocks. International
Universiteit Leuven) in Belgium in 2010. Then (in 2011) he
Journal of Rock Mechanics and Mining Sciences 2001;38(5):671–82.
joined IMDC and he carries out a diverse range of geotech-
Reineck HE, Singh IB. Depositional sedimentary environments. Berlin, Germany:
nical and dredging subjects.
Springer-Verlag; 1986.