Scale Effect on Shear Behavior of Loose Sand in

Triaxial Testing and its Implication in Engineering

Design and Analysis
Tarek Omar and Abouzar Sadrekarimi
Department of Civil and Environmental Engineering, Western University,
London, Ontario, Canada

ABSTRACT
The influence of sample size on the shear behavior of loose sand is presented. Several drained and undrained static
triaxial compression shear tests are performed on three different specimen sizes of the same sand. The test results
indicate that the behavior of loose sand is strongly influenced by the specimen size, with larger specimens exhibiting a
stiffer behavior during isotropic compression, and mobilizing smaller shear strengths and effective friction angles. The
measured critical state parameters and shear strengths are employed to investigate scale effects in engineering design
and analysis. The results show that all analyses are significantly affected by the strength parameters of the same soil
determined from different specimen sizes. Accordingly, the choice of a representative sample size could have a profound
impact on the safety or cost effectiveness of engineering analysis and design.

RÉSUMÉ
L'influence de la taille de l'échantillon sur le comportement en cisaillement d’un sable lâche est présentée. Réalisés en
compression triaxiale statique, plusieurs essais en cisaillement drainé et non drainé ont été effectués sur trois différentes
tailles de spécimens du même sable. Les résultats des essais indiquent que le comportement du sable lâche est
fortement influencé par la taille du spécimen, les plus gros de ceux-ci présentant un comportement plus raide durant la
compression isotrope et mobilisant de plus petites forces de cisaillement et des angles de friction effectifs moindres. Les
paramètres d'état critique et les résistances au cisaillement mesurés permettent d’évaluer les effets d'échelle liés à la
conception et à l'analyse. Les résultats montrent que toutes les analyses sont influencées de manière significative par
les paramètres de résistance déterminés sur le même sol à partir de différentes tailles de spécimens. En conséquence,
le choix de la taille d'un échantillon représentatif pourrait avoir de profondes répercussions sur la sécurité ou sur
l’efficience de l'analyse d’ingénierie et de la conception.

1 INTRODUCTION Several researchers have studied the effect of sample

Different studies employ different specimen diameters (D) size on the behavior of cohesionless soils using triaxial
and heights (H) in triaxial compression tests. The behavior shear tests (Marsal 1967; Scott 1987; Been et al.1991; Hu
of a particular soil from different studies are often et al. 2010). For example, Scott (1987) performed drained
compared without due attention to the differences in triaxial compression tests on dense Leighton Buzzard
specimen size and its effect on soil shear behavior. For sand specimens of diameters 38 mm and 100 mm and
example, Table 1 presents a summary of the different found a higher peak strength and initial shear modulus in
specimen sizes used in different triaxial testing studies of the larger specimen whereas smaller post peak shear
sand behavior. strength was mobilized in the larger specimen. Jefferies et
al. (1990) investigated the influence of sample size on the
Table 1. Summary of specimen sizes used in past studies. drained shearing behavior of Ticino 9 sand specimens of
diameters 35, 75, 150, and 300 mm. They found that the
size (mm) smaller specimen exhibits the greatest volumetric strain
Sand Researcher despite its smallest peak deviator stress. Hu et al. (2010)
D H
1
38 NP Leighton Buzzard Scott (1987) developed a set of drained triaxial compression tests on
50 100 Unimin Garga (1988) various sizes of Loire River sand and found that the pre-
51 102 Ottawa/Mississippi Sadrekarimi (2011) peak behavior was not affected by the specimen size,
71 150 Ottawa Frost (2000) whereas the post-peak behavior mobilized smaller friction
74 150 Monterey Ladd (1978)
1 angle.
75 NP Banding Castro (1969)
76 150 Erkask Been et al. (1991) Similar results have been documented by many other
100 NP1 Leighton Buzzard Scott (1987) studies on the effect of shear box size on the results of
100 200 Loire river Hu et al. (2010) direct shear tests (Cerato and Lutenegger 2006; Wu et al.
102 200 Silty sand Yamamuro (1997) 2007; Bareither et al. 2008). Some other investigators
250 500 Loire river Hu et al. (2010) have compared the behavior of granular materials with
300 675 Ticino 9 Jefferies et al. (1990) different ranges of particle sizes as an alternative
300 600 Granular material Seif el Dine. (2009) approach for studying specimen scale effect (Tatsuoka
1000 1500 Loire river Hu et al. (2010)
1 1997; Oie et al. 2003; Farbodfar 2013). For example,
NP: not provided
Fabodfar (2013) performed series of direct shear tests on
three sands with different mean particles size (D50 = 0.19,
0.49, and 2.77 mm) and found that the peak shear
strength and the mobilized friction angle increased with
increasing D50. Based on the analogy between D50 and
specimen size, his results imply that the peak shear
strength and the mobilized friction angle decrease with
increasing specimen size.
Previous investigations of specimen-size effect in
triaxial tests have mainly focused on testing dense sands
or rockfill which also require the construction of larger
triaxial devices. In the presented study, a comprehensive
experimental program is conducted to investigate and
compare the sample size effect on the consolidation,
drained and undrained shear behavior of very loose
Ottawa sand specimens.
2 EXPERIMENTAL PROGRAM
In the triaxial tests of this study, a cylindrical specimen
was sealed in a water-tight rubber membrane and
confined in a water-filled acrylic cell. The details of the
monotonic triaxial compression shear tests are briefly
described below.
2.1 Properties of the tested sand
Clean, uniformly-graded Ottawa sand (with a commercial
name of “Barco 71”) with round to sub-round particle
shapes is used in the tests of this study. The sand
classified as a SP according to the Unified Soil
Classification System (ASTM D2487–11). Average mean
particle size (D50), coefficient of uniformity (CU), and
coefficient of curvature (CC) of 0.22 mm, 1.71, and 1.07
were determined from sieving analysis, respectively.
Specific gravity of the sand particles (GS), and maximum
and minimum void ratios of respectively 2.65, 0.821, and
0.487 were measured following the ASTM standard
procedures D854 -10 and D4253-00, respectively.
2.2 Specimen preparation
Non-uniform deformations in triaxial testing often result
from the mechanical boundary restraints as a result of the
friction between the soil specimen and the end platens
(Bishop and Green 1965; Lee 1978). This end restraint
could affect the pore pressure or volume change
measurements and lead to incorrect shear strength
parameters. In order to reduce specimen non-uniform
deformation and bulging at large strains (Rowe and
Barden 1964; Bishop and Green 1965), cylindrical
specimens were prepared to 38, 50, and 70 mm
diameters with equal height and diameter (H/D = 1) and
the end restraints were nearly eliminated by employing
enlarged and lubricated end platens covered with two
layers of lubricated latex sheets.
All specimens were prepared at an initial void ratio (ei)
of 0.821, corresponding to an initial relative density (D ri) of
0% by moist tamping in order to produce sufficiently loose
specimens of quartz sands which would exhibit entirely
strain-softening (in undrained shear) or contractive (in
drained shear) behavior (Sadrekarimi and Olson 2011). In
order to minimize the density variations and void ratio
non-uniformities within the specimens, the under
compaction technique introduced by Ladd (1978) was
employed to achieve a relatively uniform density
throughout the specimen height. A small vacuum (about 4
to 5 kPa) was applied by the pore pressure pump in order
to provide confinement and hold the specimen in place
during dismantling of the mold. The vacuum pressure was
replaced by an external cell pressure of 10 kPa following
the removal of the specimen mold.
2.3 Triaxial testing system
The triaxial tests of this study were conducted using an
automated stress path triaxial compression testing
system. The main components of this system include a
triaxial cell, a loading frame, two electromechanical
pressure pumps, and a data acquisition and control
system. The electromechanical pumps (i.e. cell and pore
pressure pumps) are used to control and measure the
volume and pressure of the cell fluid and specimen’s pore
water. The system also includes an external load cell, a
deformation sensor, and three fluid pressure sensors. All
experiments of this study were conducted at a constant
axial strain of 5%/hour. The axial deformation of the
sample during shear was measured externally by a linear
variable displacement transducer (LVDT) on the loading
piston of the triaxial cell.
2.4 Triaxial testing procedure
Static triaxial compression tests were performed in this
study. Saturation started by flushing the specimen with
carbon dioxide then with deaired water. The saturation
procedure was continuing with a backpressure saturation
phase as recommended by Black and Lee (1973) by
applying a back pressure of 200 kPa to the specimen pore
water in order to drive any remaining air into solution.
Saturation was verified by ensuring that a Skempton’s
pore water pressure parameter B of at least 0.98 was
achieved in all specimens. Isotropic consolidation
commenced subsequent to the completion of specimen
saturation by increasing the effective confining pressure
(p'c). The volume of water driven out of the sample was
measured from the change in the pore pump volume
before and after consolidation and hence the post-
consolidation void ratio (ec) was readily obtained. The
specimens were sheared following isotropic consolidation
up to an axial strain of 30% at an axial strain rate of
5%/hour to ensure full pore pressure equalization during
undrained shearing and complete excess pore pressure
dissipation during drained shearing. Specimen drainage
was not permitted during the undrained tests while the
specimen pore pressure was kept at the value of the
backpressure during drained shear tests so all the shear-
induced pore water pressure was dissipated. The void
ratio at each strain level was calculated from the volume
change measurements during the drained shear tests
while a constant specimen volume was maintained in the
undrained shear tests. Table 2 summarizes the
specifications of triaxial tests of this study.
Table 2. Specifications of the triaxial compression tests smaller compressibility during isotropic compression,
conducted in this study which could be associated with the larger number of sand
particles and thus the greater number of particle contacts
Test No.a Ds (mm) p'c (kPa) ec Drc (%) in a larger specimen. As a result of the less compressible
MT-1D 500 0.771 15 behavior of the larger specimens (70 mm), a slightly lower
MT-2D 300 0.779 13 relative density was produced at the end of consolidation
MT-3D 200 0.785 11 just before shear in the 70 mm specimens compared to
MT-4D 100 0.797 7 the 50 mm and 38 mm specimens. We note that the
70
MT-5UD 500 0.775 14 specimens prepared at even looser initial void ratios (ei >
MT-6UD 300 0.782 12 0.821) significantly deformed and collapsed after
MT-7UD 200 0.791 9
preparation and specimen saturation. Hence, it was not
MT-8UD 100 0.797 7
MT-9D 500 0.761 18
possible to prepare the 50 and 38 mm diameter
MT-10D 300 0.769 16 specimens at looser ei in order to produce exactly the
MT-11D 200 0.773 14 same ec in all specimen sizes. Furthermore, the slight
MT-12D 100 0.786 10 differences in ec is within the void ratio variations in most
50
MT-13UD 500 0.769 16 triaxial tests (Jefferies et al. 1990).
MT-14UD 300 0.775 14
MT-15UD 200 0.785 11
MT-16UD 100 0.795 8
MT-17D 500 0.76 18
MT-18D 300 0.766 16
MT-19D 200 0.775 14
MT-20D 100 0.786 10
38
MT-21UD 500 0.766 16
MT-22UD 300 0.773 14
MT-23UD 200 0.784 11
MT-24UD 100 0.794 8

3 TRIAXIAL COMPRESSION TEST RESULTS

Triaxial shear tests involve several sources of errors that

could significantly affect test results if not corrected.
Corrections were made to account for the volume change
due to backpressure saturation (ASTM D4767-11),
membrane penetration (Baldi and Nova 1984), axial
deformation due to bedding error (Sarsby et al. 1980),
stress correction due to membrane resistance (ASTM
D7181-11), and the change of specimen cross-sectional
area during shear (Garga and Zhang 1997). The Figure 1. Effect of specimen size on the isotropic
corrected results are presented in the following compression behavior of loose Ottawa sand at (a) p' c =
paragraphs and compared for different specimen sizes 100 kPa, (b) p'c = 200 kPa, (c) p'c = 300 kPa, and (d) p'c =
during isotropic compression and shear. 500 kPa.

3.1 Isotropic compression response 3.2 Undrained shear behavior

As discussed earlier, all the triaxial specimens were
prepared at a very low relative density (Dri = 0%). All
specimens were slighted compressed to a relative density
of about 2.4% following saturation as a result of the
volume changes occurring during flushing and the
application of back-pressure (corresponding to about
0.45% volumetric strain). Figure 1 presents the isotropic
compression lines following the consolidation phase for
the different specimen sizes tested in this study.
According to the plots of Figure 1, the 38 mm specimens
display the most compressive response during isotropic
compression (i.e. steepest compression line) followed by
the 50 mm and 70 mm specimen sizes. In other words,
the 70 mm specimens experienced less volumetric strains
than the 38 and 50 mm specimens at the same p'c in all
tests. This indicates that larger sand specimens exhibit
markedly stiffer isotropic compression behavior and
All specimens exhibited strain-softening behaviors and the
deviator stress mobilized throughout the tests was
consistently larger in the smaller specimens and the axial
strain (a) corresponding to the peak deviator stress
increases with decreasing specimen size. As the
mechanical behavior and shear strength mobilization in
granular soils essentially depends on the interaction
among soil particles and the amount of particle
movement, rearrangement, reorientation and possible
particle crushing, the shearing behavior of a soil, which
reflects this fundamental particle interaction, depends on
the amount of relative displacement among soil particles
and therefore soil shearing behavior should be essentially
unrelated to the dimensions of the specimen. Whereas
the axial strain (a) is calculated by normalizing specimen
axial deformation (a) with respect to specimen height (H)
and therefore, the stress-strain behaviors of specimens
with different sizes would depend on H. Accordingly, in
order to remove the effect of specimen height from soil
shearing behavior, Figure 2 compares the undrained
deviator stress versus the axial deformation behavior of
different specimen sizes. Note that the final a = 2.1, 1.5,
and 1.2 cm correspond to the axial displacements
reached at a = 30% in the 70, 50, and 38 mm specimen
sizes, respectively. According to Figure 2, the peak
deviator stresses occur at the same a ≈ 0.7 mm for all
specimen sizes and the differences in deviator stresses
among different specimen sizes at the same a is obvious.
Note that although the lower deviator stresses mobilized
in the 70 mm specimens could be partly due to their
slightly looser void ratios, the 50 mm and 38 mm
specimens were sheared from the same ec and exhibit
clear specimen size effects.

Figure 3. Excess pore water pressure developed during

undrained triaxial compression tests on different specimen
sizes at (a) p'c = 100 kPa, (b) p'c = 200 kPa, (c) p'c = 300
kPa, and (d) p'c = 500 kPa.

3.3 Drained shear behavior

Figure 4 compare the drained shearing behaviors of the

sand specimens where larger deviator stresses are
mobilized in the smaller specimens. Note that similar to
the undrained shear tests, although the lower deviator
stress of the 70 mm specimens could have been partly
associated with their slightly looser ec, the 50 mm and 38
mm diameter specimens were sheared from the same e c
and exhibit obvious specimen size effects.

Figure 2. Undrained stress – axial displacement behaviors

of different specimen sizes in triaxial compression shear
tests at (a) p'c = 100 kPa, (b) p'c = 200 kPa, (c) p'c = 300
kPa, and (d) p'c = 500 kPa.

On the other hand, Figure 3 shows similar excess pore

water pressures (u) developed during undrained shear in
the different specimen sizes consolidated to the same p'c.
Although the minor differences in the initial rates of u
generation could be related to the slight differences in the
specimen consolidation relative density (Drc), equal u are
developed after an axial strain of 10% where a critical
state is reached. This implies that although the contraction
tendency of the sand fabric was the same for all specimen
sizes, the mobilized strength, deviator stress, and the
strain-softening behaviors include an inherent specimen
size dependency irrespective of the differences in the
amount of a experienced by each specimen size at
similar a. This could be attributed to more intense strain-
softening resulting from shearing along longer and larger Figure 4. Drained deviator stress – axial displacement
number of shear bands formed in larger specimens and behaviors of different specimen sizes in triaxial
the differences in the available space or freedom for compression shear tests at (a) p'c = 100 kPa, (b) p'c = 200
particle rearrangement in the specimens of different sizes. kPa, (c) p'c = 300 kPa, and (d) p'c = 500 kPa.
In order to investigate the influence of specimen size on
the volumetric contraction occurring during the drained
shear tests, the rate of volumetric contraction (C = v/a)
versus a obtained from specimens of different sizes are
compared in Figure 5. These plots demonstrate that the
rate of volumetric contraction increases with decreasing
specimen size with the maximum volumetric contraction
occurring at a small a of about 1% (i.e. the same a of the
peak undrained strength). This could be again attributed
to the larger number of particles, and thus particle
contacts in the larger specimen which produced a less
compressible particle fabric and reduced v.
related to more intense strain-softening resulting from
shearing along longer and larger number of shear bands
formed in larger specimens and the differences in the
available space or freedom for particle rearrangement in
the specimens of different sizes.
3.4 Influence of Specimen Size on Effective Stress
Paths
According to the stress - displacement and volumetric
contraction plots in Figures 2 and 5, both the undrained
and the drained shear tests exhibit strain-softening or
contractive behavior moving towards a critical state. At the
critical state, a unique relationship is established between
the effective stress, void ratio, and shear stress, which is
often described by the critical stress ratio line in the stress
path (q versus p') diagram and the critical state line in the
void ratio versus p' plane (Been et al.1991). The effective
stress paths obtained from the testing of different
specimen sizes are compared in Figures 6 for undrained
shearing condition. The undrained effective stress paths
of all specimen sizes reach an instability line (IL) which
defines the lower bound of all the possible unstable
conditions (Lade 1992). The undrained stress paths then
move into the region of instability, and approach a
constant stress ratio at different p'c values, corresponding
to the CSL of each specimen size. The drained stress
paths also reach the CSL, but without exhibiting unstable
behavior. Unique CSLs are obtained from both the
undrained and drained stress paths for each specimen
size, which become steeper with decreasing specimen
size indicating the influence of specimen size on the
effective stress paths of loose sand in triaxial compression
shear tests.

Figure 5. Rate of volumetric contraction during drained

triaxial compression shear tests on different specimen
sizes at (a) p'c = 100 kPa, (b) p'c = 200 kPa, (c) p'c = 300
kPa, and (d) p'c = 500 kPa.

Failure of granular materials is often characterized by

localization of non-uniform deformations commonly
referred to as shear bands (Roscoe 1970). Loose sand
specimens exhibit a complex internal failure pattern
involving multiple shear bands that may not be visible
without using advanced visualization techniques
(Tatsuoka 1997; Desrues et al. 1996; Alshibli and Sture
2000). Accordingly, although no shear bands were
observed at the end of our triaxial tests on the exterior
surface of the specimens, it is quite possible that a
complex and diffuse mode of failure and shear localization
(involving multiple shear bands) developed in the
specimens. In fact, multiple shear band formation and
complex localization is an inevitable response of sand to
the stress state and shear path imposed in a shear test,
irrespective of the test boundary condition and specimen
imperfections (Desrues et al. 1996). Techniques such as
enlarged and lubricated end platens used to induce more
uniform shear deformations in triaxial and biaxial tests Figure 6. Effect of specimen size on undrained effective
merely delay and conceal shear localization. Again, the stress paths of Ottawa sand at (a) p'c = 100 kPa, (b) p'c =
lower deviator stresses of the larger specimens could be 200 kPa, (c) p'c = 300 kPa, and (d) p'c = 500 kPa.
3.5 Influence of Specimen Size on Effective Friction
Angle

Granular soils mobilize shear resistance through their

internal friction angle, and therefore friction angle is
considered as the most important mechanical property to
analyze granular soils' response to loading. The friction
angle (ϕ'mob) mobilized in a triaxial compression shear test
is calculated from the effective stress ratio (M = q/p').
ϕ'mob in both drained and undrained shear tests
steadily increased with increasing a and approached a
constant value at the critical state (ϕ'cs). The mobilized
friction angle corresponding to the slope of the instability
line (in Fig. 6) is referred to as the yield friction angle
(ϕ'yield) which is obtained from undrained shear tests. The
effect of specimen size on ϕ'yield and ϕ'cs is summarized in
Figure 7 which shows that the size of a specimen can
have a profound effect on the ϕ'yield and ϕ'cs measured Figure 8. Effect of specimen size on the CSLs from both
from triaxial compression shear tests. This could further drained and undrained triaxial compression shear tests.
produce significant impact in geotechnical engineering
practice and designs which utilize ϕ'yield and ϕ'cs (e.g.,
stability of slopes, liquefaction analysis, and bearing Table 3. Effect of specimen size on the critical state
capacity of foundations). parameters.

Specimen Diameter M = q/p cs degrees  

70 mm 1.12 28 0.023 0.887
50 mm 1.25 31 0.026 0.881
38 mm 1.29 32 0.028 0.878

3.7 Influence of specimen size on undrained shear

strength

Based on the Mohr-Coulomb failure theory, the undrained

Figure 7. Effect of specimen size on (a) ϕ'yield, and (b) ϕ'cs shear strength (su) is obtained from the deviator stress
from undrained triaxial compression shear tests. (q). Accordingly, as presented in the stress-displacement
plot of Figure 2, a peak undrained shear strength, su(yield)
is attained at a ≈ 1%. This is followed by significant drop
3.6 Influence of specimen size on critical state of the undrained shear resistance as a result of rapid
parameters increase in the excess pore pressure at constant volume.
Strain-softening continues until a more-or-less constant
Figure 8 presents the projection of the CSL in the e – log undrained strength is mobilized at the critical state, s u(cs).
(p'c) plane from the undrained and drained triaxial Figure 9 describes the increasing of su(yield) and su(cs)
compression tests on different specimen sizes. According with increasing p'c for each specimen size.
to this figure, although unique CSLs are established from
drained and undrained shear tests for each specimen
size, the CSLs become steeper and largely shift to denser
void ratios with decreasing specimen diameter from
70 mm to 38 mm, reflecting the less compressible
response of the larger specimens as also observed in
Figure 5. The slope (cs) and intercept (cs) of the semi-
logarithmic CSLs are summarized in Table 3. The
significant change of the CSL and the critical state
parameters (cs, cs, and 'cs) would have large effects in
estimating soil state parameter for evaluating the
liquefaction susceptibility and strain-softening behavior of
cohesionless soils and the predictions of critical state soil
constitutive models. Figure 9. Effect of specimen size on (a) su(yield), and (b)
su(cs) in undrained triaxial compression shear tests.
According to this figure, su(yield) and su(cs) increase not
only with increasing p'c and Drc, but also with decreasing
specimen size. Note that although the lower undrained
shear strengths of the 70 mm specimens partly result from
their slightly looser Drc besides the effect of specimen
size, the effect of specimen size on the undrained
strengths of the 50 mm and 38 mm specimens is
undeniable as they were sheared from the same
consolidation values (p'c and Drc)
4 DISCUSSION
Although the micromechanical and particle-scale
mechanisms associated with the observed specimen size
effects were not directly investigated in this study, it is
likely that the larger number of particles and particle
contacts was the primary mechanism for the less
compressible and smaller volumetric contraction of the
larger specimens during isotropic compression (in Fig. 1)
and drained shear (in Fig. 4), while their drained and
undrained shear strengths reduced as a result of shearing
along longer and a larger number of shear bands.
Specimen size effect on the shear strength parameters
can affect the design and analysis of many geotechnical
applications. For example, differences in the instability
lines (in Fig. 6) can affect the evaluation of liquefaction
triggering resistance of soils. The predication of sand
behavior using critical state constitutive models depends
on the critical state parameters (cs, cs, ϕ'cs) and the
different critical state parameters obtained from the testing
of different specimen sizes can affect the predictions of
such soil models. In addition, static and seismic slope
stability analysis would be directly affected by the choice
of the specimen size due to the differences in the shear
strength parameters. Moreover, the differences in the
measured effective friction angles can affect soil bearing
capacity calculations and hence, the design of shallow
footings on granular soil as well as the design and stability
analysis of soil retaining structures. This phenomenon
imposes further scaling-limitations for physical model tests
and their application to devise prototype behavior.
Besides the fact that a larger specimen provides a
better representation of field soil behavior, the smaller
friction angles and undrained shear strengths mobilized in
a larger specimen are more critical and should be used for
analysis and design. However, the associated cost and
the practical size of triaxial testing devices limit specimen
size. Accordingly, several studies have proposed
minimum sizes for direct shear and triaxial specimens in
order to reduce or avoid specimen size and scale effects.
For example, the ASTM D3080-11 standard testing
method for direct shear testing of soils requires a
minimum specimen thickness of six times the maximum
soil particle diameter (Dmax) and a minimum specimen
width of 10×Dmax. Similarly, ASTM D4767-11and D7181-
11 standard methods for triaxial undrained and drained
compression tests require cylindrical specimens with a
minimum diameter of the largest of 6×Dmax or 33 mm.
Furthermore, Scarpelli and Wood (1982) and Bareither et
al. (2008) respectively suggest using direct shear box
lengths of at least 100 and 93 times the mean particle
diameter (D50) for direct shear testing while other studies
suggest shear box widths or diameters no less than
50×Dmax (Cerato and Lutenegger 2006) or 60×Dmax (Wang
and Gutierrez 2010). The 38, 50, and 70 mm specimen
sizes used in the triaxial tests of this study correspond to
respectively 173×D50, 227×D50, and 318×D50 and
63×Dmax, 83×Dmax, and 117×Dmax of the tested sand.
While these ratios are larger than the specimen size
criteria proposed by the aforementioned studies and the
ASTM standards, it was shown here that the effect of
specimen size did not disappear or reduce with increasing
specimen diameter. Although a minimum specimen size
or sand particle diameter (D50 or Dmax) at which specimen
size effect would disappear was not identified in this
study, the results indicate that the triaxial compression
test results were still sensitive to the size of the specimen
for specimen diameters of up to 117×Dmax or 318×D50.
Figure 10 shows the average increases of su(yield), su(cs),
ϕ'yield, and ϕ'cs with percentage reduction of a triaxial
specimen’s diameter (D) from the largest of 318×D50 or
117×Dmax specimen diameter and followed by correlations
of curve-fitted to these data.
Figure 10. Changes in (a) su, and (b) ' with changes of
specimen diameter (D) with reference to those mobilized
in a 70 mm specimen diameter
su %  0.53D%  0.23 [1]
 deg   0.08D%  0.15
[2]
These equations indicate that with every 50% decrease in
specimen size (from a specimen diameter of the
maximum of 318×D50 or 117×Dmax), the undrained
strengths and mobilized friction angles increase on
o
average about 27% and 4 , respectively. In the absence
of a larger triaxial cell, the impact of testing a specimen
diameter of less than the maximum of 117×D max or
318×D50 could be eliminated or reduced by employing the
above equations to decrease the undrained shear
strength and mobilized friction angle of the smaller
specimen diameter to a larger specimen diameter of
117×Dmax or 318×D50 (whichever is greater). Note Figure
10 and Equations 3 and 4 are derived from the triaxial
compression shear tests on a single sand with only one
value of D50 (= 0.22 mm) and Dmax ( = 0.60 mm).
However, since an increase (or decrease) in D50 would
replicate decreasing (or increasing) specimen diameter,
the effect of D50 is accounted for by expressing D relative
to a specimen diameter of 318×D50. Therefore, by
increasing D50 the maximum specimen diameter at which
the specimen size effect could be reduced increases
accordingly. Indeed, further testing on different sands with
different values of D50, particle size distributions and
particle shapes is necessary to explicitly study the effect
of these parameters on specimen size effect and
developing more precise corrections.
5 CONCLUSIONS
The main objective of this study was to investigate the
specimen size effect on sand behavior in triaxial
compression tests. The triaxial test results indicated that
sand behavior was affected by the size of the specimen
during isotropic compression as well as during drained
and undrained shear. The larger specimens presented
less compressible isotropic compression behavior than
the smaller specimens. During shear, the smaller
specimens exhibited steeper CSL, indicating a higher
compressibility and mobilized larger friction angles and
shear strengths. The greater number of particles as well
as the longer and larger number of shear bands and
failure planes in the larger specimens could have
produced the observed effects of specimen size. The
influence of sample size on sand behavior, particularly on
the internal friction angle is very important and directly
affects many aspects of geotechnical engineering practice
(design of shallow footings on granular soils, slope
stability analysis, design of retaining structures). In order
to reduce specimen size effects, triaxial testing of larger
specimens is recommended as it also provides a better
representation of field soil behavior. Based on the results
of this study, it is recommended to use specimen
diameters of no less than 318×D50 or 117×Dmax in triaxial
shear testing of granular soils. Two equations are
proposed to correct the undrained shear strength and
mobilized friction angle of smaller diameter specimens to
a maximum specimen diameter of 318×D50 or 117×Dmax
(whichever is greater).
REFERENCES
Alshibli, K. A., and Sture, S. (2000). “Shear band
formation in plane strain experiments of sand” J.
Geotech. & Geoenv. Eng., Vol. 126 (6), pp: 495–503.
Baldi, G., and Nova, R. (1984). "Membrane penetration
effects in triaxial testing" J. Geotechnique. Vol. 110
(3), pp: 403-420.
Bareither, C. A., Benson, C. H., and Edil, T.B. (2008).
“Comparison of shear strength of sand backfills
measured in small scale and large scale direct shear
tests” Can. Geotech. J. Vol. 45, pp: 1224-1236.
Been, K., Jefferies, M. G., and Hachey, J. E. (1991). “The
critical state of sands” J. Geotechnique. Vol. 41(3), pp:
365–381.
Bishop, A. W., and Green, G. E. (1965). “The influence of
end restraint on the compression strength of a
cohesionless soil” J. Geotechnique. Vol. 15 (3), pp:
243- 266.
Cerato, A.B., and Lutenegger, A.J. (2006). “Specimen
size and scale effects of direct shear box tests of
sands’ J. Geotech. Testing. Vol. 29 (6), pp: 1-10.
Desrues, J., Chambon, R., Mokni, M., and Mazerolle, F.
(1996). “Void ratio evolution inside shear bands in
triaxial sand specimens studied by computed
tomography” J. Geotechnique. Vol. 46 (3), pp: 529–
546.
Farbodfar, H. (2013). “Studying sand size and scale
effects on its shear behavior” M.Eng. Project.
University of Western Ontario.
Garga, V., and Zhang, H. (1997). “Volume changes in
undrained triaxial tests on sands” Can. Geotech. J.
Vol. 34, pp: 762-773.
Hu, W., Dano, C., Hicher, P., Touzo, J., and Derkx, F.
(2010). “Effect of Sample Size on the Behavior of
Granular Materials” J. of Geotech. Testing, Vol. 34 (3),
pp: 1-12.
Jefferies, M. G., Been, K., and Hachey, J. E. (1990).
“Influence of scale on the constitutive behavior of
sand” Cand. Geotech. Conferernce. pp: 263-273
Ladd, C. C. (1978). "Preparing test specimen using
undercompaction" J. Geotech. Testing. Vol.1, 16-23.
Lade, P. V. (1992). “Static instability and liquefaction of
loose fine sandy slopes” .J. Geotech. Eng., Vol. 118
(1), pp: 51–72.
Lee, K. L. (1978). “End restraint effects on undrained
static triaxial strength of sand”. J. Geotech. Eng.,
Vol.104 (6), pp: 687–704.
Marsal, R. J. (1967). “Large-scale testing of rockfill
materials” J. Soil Mech. and Found. Div. Vol. 93
(SM2), pp: 27–44.
Oie, M., Sato, N., Okuyama, Y., Yoshida, T., Yoshida, T.,
Yamada, S., and Tatsuoka, F. (2003). “Shear banding
characteristics in plane strain compression of granular
materials” J. soil mech. and found. Eng., Vol. 4, pp:
597-606.
Rowe, P.W., and Barden, L. (1964). “Importance of free
ends in triaxial testing” J. soil mech. and found., Vol.
90 (1), pp: 1–27.
Sadrekarimi, A., and Olson, S.M. (2011). “Critical state
friction angle of sands” J. Geotechnique, Vol. 61 (9),
pp: 771–783.
Scarpelli, G. and Wood, D. M., (1982). “Experimental
observations of shear band patterns in direct shear
tests.” Con. on Deformation and Failure of Granular
M., pp: 22-35.
Scott, R. F. (1987). “Failure” J. Geotechnique. Vol. 37 (4),
pp: 423-466.
Tatsuoka, F. (1997) “Deformation and strength
characteristics of granular materials” Research center,
Tokyo University of Science. pp: 1-100.
Wang, J., and Gutierrez, M. (2010). “Discrete element
simulations of direct shear specimen scale effects” J.
Geotechnique, 60 (5): 395 – 409.
Wu P, Matsushima K. and Tatsuoka F. (2007). “Effects of
specimen size on the shear strength and deformation
of granular soil” J. of Geotech. Eng., Vol. 31(1), 45-64.