See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/309692989

X-ray Computed Tomography Imaging of the

Microstructure of Sand Particles Subjected to
High Pressure...

Article in Materials · November 2016

DOI: 10.3390/ma9110890

CITATIONS READS

2 43

2 authors:

Asheque Mahbub Asadul Haque

Monash University (Australia) Monash University (Australia)
2 PUBLICATIONS 2 CITATIONS 56 PUBLICATIONS 497 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Microstructural analysis of uncemented and cemented sand using 3D X-Ray CT imaging technique
View project

Geological Sequestration of Carbon Dioxide in Deep Saline Aquifers: Coupled Flow-Mechanical

Considerations View project

All content following this page was uploaded by Asheque Mahbub on 16 March 2017.

The user has requested enhancement of the downloaded file.

 materials
Article
X-ray Computed Tomography Imaging of the
Microstructure of Sand Particles Subjected to High
Pressure One-Dimensional Compression
Asheque al Mahbub and Asadul Haque *
Department of Civil Engineering, Monash University, Victoria 3800, Australia; asheque.mahbub@monash.edu
* Correspondence: asadul.haque@monash.edu; Tel.: +61-3-9905-4974

Academic Editor: Jung Ho Je

Received: 5 August 2016; Accepted: 27 October 2016; Published: 3 November 2016

Abstract: This paper presents the results of X-ray CT imaging of the microstructure of sand particles
subjected to high pressure one-dimensional compression leading to particle crushing. A high
resolution X-ray CT machine capable of in situ imaging was employed to capture images of the whole
volume of a sand sample subjected to compressive stresses up to 79.3 MPa. Images of the whole
sample obtained at different load stages were analysed using a commercial image processing software
(Avizo) to reveal various microstructural properties, such as pore and particle volume distributions,
spatial distribution of void ratios, relative breakage, and anisotropy of particles.

Keywords: 3D X-ray CT imaging; 1-D compression; particle crushing; microstructure; sands

1. Introduction
Practical application of engineering mechanics on soils requires an in-depth understanding
of their mechanical behaviour. In modelling soil behaviour, a soil mass has often been treated as
a continuum, however, in reality it is composed of individual particles. The mechanical behaviour of
soils (strength and deformation behaviour) is governed by the arrangement of soil particles (known as
fabric) and the interparticle bonds, together they are known as soil structure [1]. In the case of
reconstituted soils or granular soils, where the bond is negligible or even absent, its microstructure
becomes basically the arrangements of particles and interparticle voids.
Although geotechnical problems deal with low stresses, there are cases where soils can be
subjected to significantly high stresses, such as large earth dams, deep driven piles, and deep well
shafts [2]. Under high stresses in one-dimensional compression, crushing of granular materials occurs,
a phenomenon which was first described by Terzaghi and Peck [3]. Later, De Souza [4] linked particle
crushing with the yield stress of granular soils. Development of a constitutive model for granular
materials demands an in-depth understanding of the micro-structure and its evolution. Under low
vertical stresses, grain-scale frictional slip, rotation and sliding accompanied by an insignificant particle
crushing, contribute to the compaction of granular materials [5]. However, at high vertical stresses,
a significant crushing of particles is known to take place which causes large deformations associated
with tight grain packing and reduction of void ratios. Particle crushing is strongly associated with the
tensile strength of individual grains [6]. However, some other factors such as the particle morphology
and minerology, particle distribution, contacts, orientations, and void ratio which contribute to
the fabric of granular materials, significantly influence the complex micro-mechanical behaviour
of granular materials. Many researchers have studied the crushing and yielding behaviour of granular
soils subjected to one-dimensional compression loading [4,7–15].
De Souza [4] and Hendron [7] reported that the yield stress increases with the increase of initial
density, and decreases with the increase of particle angularity. Hendron [7] observed that the effect of

Materials 2016, 9, 890; doi:10.3390/ma9110890 www.mdpi.com/journal/materials

Materials 2016, 9, 890 2 of 17

initial void ratio, which is significant at low stresses, diminishes at higher stresses. Analysing initial and
crushed particle size distributions, Hardin [8] suggested valuable equations for estimating the degree
of particle crushing. Fractal particle distribution from successive crushing of soil particles gives a linear
distribution on a double logarithmic graph [9–11]. Results of investigation by Hagerty et al. [12] agreed
well with the findings of earlier researchers, in particular, a significant increase of particle crushing
was observed with the increase of median particle size. The change of particle sizes with stresses
affects many important properties of sands, such as the yield stress, compressibility, susceptibility
to erosion, shear strength, and hydraulic conductivity. The effect of particle size distributions on the
grain crushing under 1D compression was studied by Nakata et al. [13] and Altuhafi and Coop [14].
They concluded that a uniformly graded sand shows marked yielding compared to a well graded
sand and they related such differences with the nature of the microscopic crushing of the particles.
Nakata et al. [13] pointed out that after reaching the yield stress, a sudden increase in particle crushing
was obvious and it took place mostly within the smaller size particles. In order to study crushing
behaviour under 1D compression, Nakata et al. [13] conducted a series of tests, each of which was
terminated at a defined stress level, and sieve analysis was performed for each case to determine
particle size distributions. Altuhafi and Coop [14] concluded more breakage occurs in samples with
higher initial void ratios.
Past studies on the 1D compression of soils heavily relied on invasive tests, which tend to
disturb the fabric of soils. More importantly, it was not possible to examine non-destructively the
micro-structural changes of soils subjected to an incremental stress regime. In addition, particles for
observation had to be chosen randomly and in limited number which greatly influences the accuracy
of the outcome. Due to these limitations, researchers have been relying on numerical tools (such as
the Discrete Element Method (DEM)) to investigate the material behaviour under different loading
conditions [15–19]. However, the DEM uses idealized digital particles which do not truly represent the
actual particle size distributions in many cases. Therefore, to make the DEM more close to the reality,
efforts have to be made to precisely characterize the micro-structure of soils.
High resolutions X-ray computed tomography (X-ray CT) is a robust and non-destructive
imaging technique, which can be used to overcome many difficulties through 3D visualization
and quantification of many parameters closely describing the fabric of granular soils. Some
researchers [20–22] used this technology for the quantification of fundamental particle characteristics,
such as particle morphology, contact etc., while others [23–26] used it for the quantification of
the microstructural evolution of granular soils subjected to biaxial [25] and triaxial loading [26],
where the focus was mainly on the development of shear bands. However, with the advancement of
laboratory-scale X-ray CT technology and image segmentation techniques [27,28], it is now possible to
conduct investigation at grain-scale level. For example, Fonseca et al. [23] used X-ray CT to capture
images of intact sample at a very high resolution (voxel size = 5 microns) and compared them with the
gradations of reconstituted samples. However, in this study, only sub-volumes instead of the whole
sample were scanned at the end of triaxial tests. Recently, Zhao et al. [29] conducted in situ imaging
of a single sand grain under very high compression loading with the objective of understanding the
particle fracture mechanisms.
This paper investigates the 1D compression behaviour of uniformly graded sands using a high
resolution 3D X-ray CT machine. In situ imaging of the whole sample was performed at the end of
each load, and images were post-processed and analysed to study the evolution of microstructure of
a sand sample subjected to a very high 1D compression loading leading to sand particle crushing.

2. Laboratory Investigation

2.1. Materials
In this study, a commercially available washed sand sample was sourced from a natural deposit
site located at Gippsland in Victoria, Australia. The sand particles are brown in colour and sub-angular
Materials 2016, 9, 890 3 of 17
Materials 2016, 9, 890 3 of 17

in shape. First,
sub-angular in the sand
shape. particles
First, wereparticles
the sand dried inwere
an oven
driedforin24anh oven
at 105for 0.5h◦ C.
± 24 Second,
at 105 ± 0.5the
°C.particles
Second,
were sieved through 250, 212, and 150 microns sieves. In this study, sand particles
the particles were sieved through 250, 212, and 150 microns sieves. In this study, sand particles passed through the
250 microns
passed and the
through were250retained
micronson and
the 212 and
were 150 microns
retained on the sieves used.150
212 and Themicrons
particlesieves
size distribution
used. The
plot
particle size distribution plot is shown in Figure 1. The values of the coefficient of uniformity (Cof
is shown in Figure 1. The values of the coefficient of uniformity (Cu ) and the coefficient u)
curvature (Cc ) wereof
and the coefficient 1.24 and 1.06(C
curvature respectively,
c) were 1.24which classified
and 1.06 the sandwhich
respectively, to be classified
uniformlythe graded
sandastoper
be
the unified graded
uniformly soil classification
as per the system.
unified The mean diameter
soil classification (D50 ) of
system. themean
The sand diameter
particles was
(D50)230 microns.
of the sand
The selection of this size is important to achieve an acceptable representation
particles was 230 microns. The selection of this size is important to achieve an acceptable of sand particles within
the volume [30]of
representation including the minimum
sand particles within the height to diameter
volume ratio of the
[30] including the minimum
sample to be tested
height to under 1D
diameter
compression with a full field of view imaging. X-ray diffraction test showed
ratio of the sample to be tested under 1D compression with a full field of view imaging. X-ray that the sand particles
were comprised
diffraction of quartzthat
test showed minerals.
the sand Theparticles
specific gravity of sand particles
were comprised wasminerals.
of quartz measuredThe three times
specific
using
gravity a Multipycnometer
of sand particles(Quantachrome
was measuredInstruments,
three timesBoynton
using aBeach, FL, USA), which
Multipycnometer produced an
(Quantachrome
average value of 2.68.
Instruments, Boynton Beach, FL, USA), which produced an average value of 2.68.

100

80 Mechanical sieve analysis

60
% Finer

40

20

0
0.1 Particle size (mm) 1.0

Figure
Figure 1. Initial particle
1. Initial particle size
size distributions
distributions from
from mechanical
mechanical sieve
sieve analysis.
analysis.

2.2. Experimental Setup

A new 1D compression apparatus was designed in this study. study. It comprises a 2-mm thick walled
aluminium
aluminium cylinder
cylinderofof8.58.5mmmminternal
internaldiameter
diameter and 10 10
and mm mmheight, twotwo
height, 2-mm 2-mmthick bronze
thick filters
bronze and
filters
aand
6-mm height
a 6-mm stainless
height steelsteel
stainless plunger withwith
plunger a 1-mm diameter
a 1-mm stainless
diameter steelsteel
stainless ball.ball.
The The
aluminium
aluminium cell
was selected
cell was due due
selected to itstolow
its X-ray absorption
low X-ray absorptioncapacity and the
capacity andbronze
the bronzefiltersfilters
for the
fordissipation
the dissipationof poreof
pressure underunder
pore pressure loading. A typical
loading. setup setup
A typical of theofapparatus is shown
the apparatus in Figure
is shown 2a. A sliding
in Figure fit of filters
2a. A sliding fit of
and plunger
filters was ensured.
and plunger It can hold
was ensured. samples
It can of heightof
hold samples upheight
to 4 mmup thus
to 4 ensuring
mm thusspecimen
ensuring minimum
specimen
diameter
minimumtodiameter
height ratio of 2.5 [31].
to height ratioTheof sand particles
2.5 [31]. were particles
The sand placed inwere the compression
placed in the apparatus
compression from
aapparatus
height of from
abouta 10 mm.of
height The mass
about 10ofmm.the The
particles
masswas recorded
of the after
particles wasitrecorded
reached the target
after heightthe
it reached of
3.5 mm, which corresponded to an initial bulk density of 1.62 g/cm 3 (20,882 particles).
target height of 3.5 mm, which corresponded to an initial bulk density of 1.62 g/cm3
Theparticles).
(20,882 5 kN compressive load capacity load-stage (CT5000, Deben, Suffolk, IP30 9QS, UK) was used
in this study.
The 5 kNFigure 2b shows
compressive the capacity
load image ofload-stage
the interior(CT5000,
of the X-ray
Deben, CT machine with 9QS,
Suffolk, IP30 the load-stage
UK) was
and
used1D in compression apparatus.
this study. Figure It is tothe
2b shows be image
noted that theinterior
of the clearanceof between
the X-raythe CTtop and thewith
machine bottom
the
platens
load-stageof the
andload-stage
1D compression is 15 mm, which It
apparatus. restricted the height
is to be noted of the
that the samplebetween
clearance to be testedthe topin situ.
and
The bottom platens
the bottom platen ofofthe
theload-stage
load-stage moves upward
is 15 mm, andrestricted
which compresses thethe sample
height against
of the samplethe top
to be reaction
tested
platen.
in situ. The
Theload-stage
bottom platen and data
of theacquisition
load-stage system
movesare controlled
upward andby the MICROTEST
compresses the samplesoftware (V6.13)
against the
developed
top reactionbyplaten.
DebenThe (Suffolk, UK) [32].
load-stage and dataThe system has asystem
acquisition wide range of control
are controlled byfunctions
the MICROTEST(such as
data acquisition
software (V6.13) interval,
developed loading rate) and
by Deben a continuous
(Suffolk, UK) [32].display window
The system hasfor plotting
a wide rangevariations
of control of
afunctions
selected group
(such asof variables (such as interval,
data acquisition load vs. deformation,
loading rate)deformation
and a continuous vs. time).
display window for
plotting variations of a selected group of variables (such as load vs. deformation, deformation vs.
time).
Materials 2016, 9, 890 4 of 17
Materials 2016, 9, 890 4 of 17

Figure
Figure 2. (a)
2. (a) Sample
Sample holder
holder setup;
setup; (b) setup
(b) Test Test setup
withinwithin
X-ray X-ray computed
computed tomography
tomography (X-ray (X-ray CT)
CT) machine.
machine.
In the test, an initial load of 100 N was applied. Thereafter, the loads were doubled for the next
In the test, an
five consecutive initial load
increments of 100
until theyNreached
was applied.
3200 N. Thereafter,
The final loadthe loads
after were
3200 N doubled
was 4500 forNtheduenext
to
fiveload
the consecutive
capacityincrements
of the stage until
(5000 theyN).reached 3200 N.
The vertical The final load
compressive after corresponding
stresses 3200 N was 4500 toNthedue to
load
the load
cases were capacity
1.8, 3.5,of7.0,
the14.1,
stage (5000
28.2, 56.4,N).
andThe vertical
79.3 MPa. A compressive stresses load
target compressive corresponding
was achieved to the load
through
cases
the were 1.8,
upward 3.5, 7.0,of14.1,
movement the 28.2,
bottom 56.4, and at
platen 79.3 MPa.
a rate of A
0.1target
mm/min. compressive
Once theload loadwashadachieved
reached
through the upward movement of the bottom platen at a rate of
a target value, the sample was allowed to undergo complete deformation under the constant 0.1 mm/min. Once the loadload,
had
reached a target value, the sample was allowed to undergo complete
which was monitored by reading axial deformation with time. It was found that about 30 to 60 min deformation under the
constant
time wasload, which
required to was monitored
complete by reading
the immediate axial deformation
compression with time.
of the sample (i.e.,It the
wasdeformation-time
found that about
30 to 60 min time was required to complete the immediate
plot reached an asymptotic value), where a higher time corresponded to a higher load. compression of the sample (i.e., the
During the
deformation-time plot reached an asymptotic value), where a higher time
imaging, the movement of the bottom platen was paused and the final stress value at the end of the corresponded to a higher
load. During
imaging was the imaging,
recorded, the movement
which showed less ofthan
the bottom
5% stress platen was paused
relaxation. At theand the final stress
completion value
of imaging,
at the end of the imaging was recorded, which
the next level of load was applied and the whole process was repeated.showed less than 5% stress relaxation. At the
completion of imaging, the next level of load was applied and the whole process was repeated.
2.3. Image Acquisition and Processing
2.3. Image Acquisition and Processing
An ultra-high resolution (0.7 microns, Zeiss Xradia XRM520Versa) X-ray Microscopy Facility
An Pleasanton,
(Xradia, ultra-high resolution
CA, USA) (0.7 microns,Geo-materials
for Imaging Zeiss Xradia (XMFIG)
XRM520Versa)was usedX-ray Microscopy
for image Facility
acquisition in
(Xradia, Pleasanton, CA, USA) for Imaging Geo-materials (XMFIG) was
this study. The XMFIG was established through an Australian Research Council Linkage Infrastructureused for image acquisition
in this
and study. The
Equipment FundXMFIG was established
(LE130100006) and equipped through an Australian
with various Research
in situ imaging Council Linkage
capabilities (such as
Infrastructure and Equipment Fund (LE130100006)
unconfined, triaxial and 1D compression load stages). and equipped with various in situ imaging
capabilities (such as unconfined,
Image projections were acquired triaxial and 1Dthe
by rotating compression
load-stage load 360◦ stages).
around its vertical axis. The frame
Image
size of projections
the projections waswere
1024acquired by rotating
× 1024 pixels. the load-stage
The scanning parameters 360°used
around
for allitsthe
vertical
scans axis. The
are given
frame size
in Table 1. of the projections was 1024 × 1024 pixels. The scanning parameters used for all the scans
are given in Table 1.
Table 1. Scanning parameters and their values.
Table 1. Scanning parameters and their values.
Parameter Value/Description
SourceParameter
to sample distance Value/Description
85 mm
Source totosample
Detector sampledistance
distance 85mm
322 mm
Detector toVoltage
sample distance 140
322keV
mm
Power
Voltage 10 WkeV
140
Exposure time 2.5 s
Power
Camera binning 10
2 W
Exposure
Lens time Macro2.5(0.4s×)
Camera binning 2
Lens Macro (0.4×)
In this study, a total of 801 projections with a pixel size of 14.28 microns (≈0.062D50 ) was taken,
which took about half an hour. The 2D projections were reconstructed to 3D volumes using XRM
In this study, a total of 801 projections with a pixel size of 14.28 microns (≈0.062D50) was taken,
which took about half an hour. The 2D projections were reconstructed to 3D volumes using XRM
Materials 2016, 9, 890 5 of 17

Reconstructor software (Cone Beam-10, Xradia, Pleasanton, CA, USA) [33]. The full volume of the
Materials 2016, 9, 890 5 of 17
bounding box that contained the initial cylindrical sample was 600 × 600 × 225 voxels.
The reconstructed
Reconstructor softwareimages
(Conewere post-processed
Beam-10, using a commercially
Xradia, Pleasanton, CA, USA) [33].available image processing
The full volume of the
software Avizo (V9.1.1, FEI, Hillsboro, OR, USA) [34]. Images of
bounding box that contained the initial cylindrical sample was 600 × 600 × 225 voxels.sand particles were cropped from
the wholeThe apparatus assembly
reconstructed followed
images werebypost-processed
removal of noises usingusing appropriate filters.
a commercially availableSubsequently,
image
the solid and void
processing phasesAvizo
software of the(V9.1.1,
image were segmented
FEI, Hillsboro, OR, and sand[34].
USA) particles
Imageswere separated
of sand and
particles labelled
were
croppedquantitative
for various from the whole apparatus assembly followed by removal of noises using appropriate filters.
analyses.
Subsequently,
For the solid
noise reduction, theand void phases
non-local means of filter
the image
(FEI, were segmented
Hillsboro, and sand
OR, USA) was particles
used in thiswerestudy.
separated and labelled for various quantitative analyses.
Despite being GPU accelerated, the runtime of this filter was very high compared to other filters such as
For noise reduction, the non-local means filter (FEI, Hillsboro, OR, USA) was used in this study.
the 3D median filter. However, the non-local filter was found to be highly effective in reducing noises
Despite being GPU accelerated, the runtime of this filter was very high compared to other filters
while preserving the edges of particles which is critical for particle separation. After filtering, the greyscale
such as the 3D median filter. However, the non-local filter was found to be highly effective in
imagereducing
was transformed
noises while to preserving
binary image the by applying
edges the interactive
of particles thresholding
which is critical module
for particle whichAfter
separation. prompts
the user to set the grey level intervals manually with a visual feedback.
filtering, the greyscale image was transformed to binary image by applying the interactive As a preliminary identification
of intensity rangesmodule
thresholding separatingwhich solids fromthe
prompts voids,
user an intensity
to set the greyrange
levelpartitioning tool was
intervals manually used
with which can
a visual
automatically
feedback. guessAs a thresholds
preliminaryseparating different
identification densities
of intensity of materials.
ranges separatingThen by from
solids manuallyvoids,adjusting
an
intensity range partitioning tool was used which can automatically guess thresholds
this preliminary threshold range in the interactive thresholding, voxels were defined as either solid (sand separating
different
particle) or void densities
by trialofandmaterials. Then on
error based by visual
manually adjusting(human
assessment this preliminary
judgement) threshold
[35] ofrange
their in the
intensities.
interactive thresholding, voxels were defined as either solid (sand particle) or
However, to perform grain scale analysis, further segmentation was required to separate sand particles void by trial and error
based on visual assessment (human judgement) [35] of their intensities. However, to perform grain
from each other. The morphological watershed algorithm module was applied to perform this task of
scale analysis, further segmentation was required to separate sand particles from each other. The
separating individual sand particles. The procedure consisted of: (a) calculating the Euclidian distance
morphological watershed algorithm module was applied to perform this task of separating
map individual
(EDM) on sand the solid phaseThe
particles. of the binary consisted
procedure image byof: applying fast yetthe
(a) calculating accurate
Euclidianapproximation
distance mapbased
Chamfer
(EDM) metric
on theconsidering
solid phase26-neighbourhood
of the binary imagewhen propagating
by applying fast yetthe distance
accurate value; (b) identifying
approximation based
the local
Chamfermaximametricofconsidering
the EDM 26-neighbourhood
with contrast value whenofpropagating
1; (c) labelling the local
the distance value;maxima as markers;
(b) identifying
theapplying
and (d) local maxima markerof the EDMwatershed
based with contrast value
with 3Dofinterpretation
1; (c) labelling and
the local maxima as markers;
26-neighbourhood and
connectivity.
(d) applying
In every load case, marker
threebased
to fourwatershed
iterationswith
of 3D interpretation
watershed, and 26-neighbourhood
depending on the extent of connectivity.
connectedIn voxels
everyfrom
resulting load increased
case, three loading,
to four iterations of watershed,
were found necessary depending
to achieve on the
the extent
desired of level
connected
(>90% voxels
particles
resulting from increased loading, were found necessary to achieve the desired level (>90% particles
separation) of segmentation. To avoid over segmentation, a subsequent run of the algorithm was
separation) of segmentation. To avoid over segmentation, a subsequent run of the algorithm was
conducted for only the particles which could not be separated in the previous instance. The whole image
conducted for only the particles which could not be separated in the previous instance. The whole
processing method applied in this study is depicted briefly in the flowchart shown in Figure 3.
image processing method applied in this study is depicted briefly in the flowchart shown in Figure 3.

Figure 3. Flow chart depicting all operations of image processing in this study.
Figure 3. Flow chart depicting all operations of image processing in this study.
 Materials 2016, 9, 890 6 of 17
Materials 2016, 9, 890 6 of 17

3.3.Results
Resultsand Discussions
and Discussions

3.1.Void
3.1. VoidRatio
Ratiovs.
vs.Logarithm
LogarithmofofVertical
VerticalStress
Stress(e-logσ’
(e-logσ’v)vPlot
) Plot
Theexperimental
The experimentaland andphysical
physicalmeasurements
measurementsofofthe thesand
sandsample
samplewere
wereutilised
utilisedtotocalculate
calculatethethe
void ratio
void ratio at
at each
eachload
loadcases. The
cases. variations
The of void
variations ratio ratio
of void with the logarithm
with of vertical
the logarithm of stress (e-logσ’
vertical stressv )
plot forv)the
(e-logσ’ sand
plot sample
for the sandtested in tested
sample this study is presented
in this in Figure in
study is presented 4. Figure
The initial void
4. The ratiovoid
initial (eo ) of the
ratio
(esample
o) of thewas 0.60. was 0.60.
sample

0.8

0.6 e0

Crushing stress = 14 MPa

Void ratio e

0.4

0.2 Experimental data

Image data
Nakata et al.[13]
0.0
1 10 100
Vertical stress v (MPa)

Figure
Figure4.4.e-logσ’
e-logσ’v vplot
plotfor
forthe
theuniformly
uniformlygraded
gradedsand
sandsample.
sample.

The e-logσ’v plot shows a gradual change of slope up to 7 MPa of vertical stress followed by a
The e-logσ’v plot shows a gradual change of slope up to 7 MPa of vertical stress followed by
significant change of slope, which could be due to the crushing of particles. In this study, the
a significant change of slope, which could be due to the crushing of particles. In this study, the crushing
crushing stress or yield stress is defined as the stress corresponding to the intersection of two
stress or yield stress is defined as the stress corresponding to the intersection of two straight parts of the
straight parts of the e-logσ’v curve [36,37]. The yield stress for the uniformly graded sands tested
e-logσ’v curve [36,37]. The yield stress for the uniformly graded sands tested under 1D compression
under 1D compression was found to be 14 MPa which is denoted by the arrow sign in the above
was found to be 14 MPa which is denoted by the arrow sign in the above figure. The observed value
figure. The observed value is similar to that reported by Nakata et al. [13] for uniformly graded
is similar to that reported by Nakata et al. [13] for uniformly graded quartz sand of relatively larger
quartz sand of relatively larger particle sizes but of almost similar uniformity coefficient (1.1 vs. 1.24
particle sizes but of almost similar uniformity coefficient (1.1 vs. 1.24 of this study) and initial void ratio
of this study) and initial void ratio of 0.6 ± 0.03. This indicates that the small compression apparatus
of 0.6 ± 0.03. This indicates that the small compression apparatus designed in this study is capable
designed in this study is capable of producing an acceptable result when compared with the result
of producing an acceptable result when compared with the result obtained from the relatively large
obtained from the relatively large apparatus (50 mm diameter and 10 mm height) of Nakata et al.
apparatus (50 mm diameter and 10 mm height) of Nakata et al. [13].
[13].
The e-logσ’ plot also shows that the decrease of void ratio in the post-yield stresses is higher,
The e-logσ’v vplot also shows that the decrease of void ratio in the post-yield stresses is higher,
indicating a higher degree of particle crushing [9]. The slope of the e-logσ’v curve in the post-yield
indicating a higher degree of particle crushing [9]. The slope of the e-logσ’v curve in the post-yield
region is reasonably in agreement with that of Nakata et al. [13]. In order to explore the capability
region is reasonably in agreement with that of Nakata et al. [13]. In order to explore the capability of
of 3D X-ray CT imaging, only the greyscale images of vertical and horizontal sections through the
3D X-ray CT imaging, only the greyscale images of vertical and horizontal sections through the
centre of the whole sample under six load cases including the initial state are presented in Figure 5.
centre of the whole sample under six load cases including the initial state are presented in Figure 5. It
It is evident from these images that with increased stresses, especially after the yield stress (14 MPa),
is evident from these images that with increased stresses, especially after the yield stress (14 MPa), a
a significant collapse of voids and crushing of particles were encountered. At the maximum vertical
significant collapse of voids and crushing of particles were encountered. At the maximum vertical
stress of 79.3 MPa, the void ratio was observed to decrease to 0.21, which is one-third of the initial
stress of 79.3 MPa, the void ratio was observed to decrease to 0.21, which is one-third of the initial
value (0.60). A void ratio close to 0.19 was reported by Nakata et al. [13] for the uniformly graded sand
value (0.60). A void ratio close to 0.19 was reported by Nakata et al. [13] for the uniformly graded
tested under a similar vertical stress.
sand tested under a similar vertical stress.
Materials 2016, 9, 890 7 of 17
Materials 2016, 9, 890 7 of 17

500 microns

XY section XY section
No load e = 0.60 28.2 MPa e = 0.39

XZ section XZ section

XY section XY section
7 MPa e = 0.55 56.4 MPa e = 0.29

XZ section
XZ section

XY section XY section
14.1 MPa e = 0.47 79.3 MPa e = 0.23

XZ section
XZ section

Figure 5.
Figure 5. Horizontal
Horizontaland vertical
and image
vertical sections
image through
sections the centre
through of theof
the centre sample at different
the sample vertical
at different
stresses.stresses.
vertical

3.2. Evolution of Microstructure

3.2.1. Void Size Distributions with Vertical Stresses

The images acquired for each load case were thresholded to obtain the volume of solid and void
phases. Subsequently, the void ratios were calculated for each load cases and plotted together with
Materials 2016, 9, 890 8 of 17

3.2. Evolution of Microstructure

3.2.1. Void Size Distributions with Vertical Stresses

Materials 2016, 9, 890 8 of 17
The2016,
Materials images
9, 890acquired for each load case were thresholded to obtain the volume of solid and 8 of void
17
the experimentally
phases. Subsequently, obtained
the void values
ratios(Figure 4). It can be
were calculated forseen that
each thecases
load voidand ratios calculated
plotted fromwith
together the
theexperimentally
image
the experimentally
analysis usingobtained values (Figure
the threshold
obtained 4).
4). Itas
intensities
(Figure Itcan be
beseen
shown
can that
thatthe
in Figure
seen 6 void
the are
void ratios
in goodcalculated
ratios agreementfrom
calculated with
fromthethe
imageanalysis
image analysis
experimental using
data.
using An the
the thresholdtrend
increasing
threshold intensities as
asshown
of threshold
intensities ininFigure
intensity
shown 6 6are
values
Figure inin
as
are good
shown
good agreement
in Figure with
agreement thethe
6 could
with be
experimental
associated with
experimental data.
theAn
data. An increasing
increase trend
of thetrend
increasing of
fraction threshold
of fines and
of threshold intensity values
the bulkvalues
intensity as
density shown
asofshown in
the sand Figure 6 could be
sample6subjected
in Figure could be
associated
to increasedwith the increase
compressive of
loads.the fraction of fines and the bulk density of the sand
associated with the increase of the fraction of fines and the bulk density of the sand sample subjected sample subjected
totoincreased
increasedcompressive
compressiveloads. loads.
16000
16000
14000
14000
12000
intensity

12000
10000
intensity

10000
8000
Threshold

8000
Threshold

6000
6000
4000
4000
2000
2000
0
0 0 20 40 60 80 100
0 20 40 60 80 100
Vertical stress'v (MPa)
Vertical stress'v (MPa)

Figure6.6. Threshold
Figure Threshold intensity
intensity values
values of
of grey
grey scale
scale images
images for
for different
different vertical
vertical stresses.
stresses.
Figure 6. Threshold intensity values of grey scale images for different vertical stresses.

The
The void
Thevoid volume
voidvolume obtained by
volume obtained
obtained by the
by the thresholding
the thresholdingwas
thresholding wasfurther
was furtherprocessed
further processedusing
processed usingthethewatershed
watershed
algorithm
algorithm
algorithmtoto to obtain
obtain
obtain the
thethe
pore pore volume
volume
pore volume distributions.
distributions. Figure
distributions. Figure 7
7 presents
Figure presents the distribution
the distribution
7 presents of pore
of poreofvolumes
the distribution pore
volumes
with
volumes with
increase
withinincrease instresses.
verticalin
increase vertical stresses.
vertical stresses. Unlikeparticles,
Unlike Unlike
particles,particles, pores
pores pores
are areinterconnected
interconnected
interconnected
are and and and
often often
form
often form
large
form
large
volumes volumes
based on
large volumes based on 26-neighbourhood
26-neighbourhood
based on 26-neighbourhood connectivity.
connectivity. As
As anticipated,
connectivity. anticipated,
large pores
As anticipated, large
largewerepores
pores were
observed observed
to reduce
were observed
toto reduce
tosmall
reduce to
size small
to pores size pores
withpores
small size with the
the increase
with theofincrease
verticalofof
increase verticalstresses.
stresses.
vertical stresses.Interestingly,
Interestingly, Interestingly,
not much notnot much
difference
much difference
between
difference
between
the pore size
between the pore size distribution
the distribution
pore size distribution plots
plots for plots
the forthe
final
for thefinal
two final two
loadtwo
cases load
loadwas cases wasobserved,
observed,
cases was observed,
which is which is is
in-line
which in-line
with
in-linethe
with the relatively
with the small
relatively relatively small change
change ofchange of void
of void
void ratios ratios experienced
ratios experienced
experienced under
under
under these these stresses
these stresses
stresses (Figure
(Figure (Figure
4). 4).4).

100
100
no load
no load
7.0 MPa
7.0 MPa
14.1 MPa
14.1 MPa
28.2 MPa
28.2 MPa
80
56.4
56.4 MPa
MPa
79.3
79.3 MPa
MPa
volume)
(byvolume)

60
60
Finer(by
%%Finer

40
40

20
20

0
0 5 50 500
5 50
Pore volume × 106 (cu. micron) 500
Pore volume × 106 (cu. micron)

Figure7.7. Pore
Figure Pore volume distribution of
volume distribution ofsamples
samplestested
testedunder
underdifferent
differentvertical
vertical stresses.
stresses.
Figure 7. Pore volume distribution of samples tested under different vertical stresses.

3.2.2. Void Ratio Distributions

3.2.2. Void Ratio Distributions
The 3D reconstructed slices of 8.5 mm × 8.5 mm × 0.14 mm (600 × 600 × 1 voxels) size were
The 3D reconstructed slices of 8.5 mm × 8.5 mm × 0.14 mm (600 × 600 × 1 voxels) size were
analysed to determine the change of void ratios with the height of the sample for all the load cases
analysed to determine the change of void ratios with the height of the sample for all the load cases
(Figure 8a). A shift of the plot to the left indicates a reduction of void ratio under the effect of vertical
Materials 2016, 9, 890 9 of 17

3.2.2. Void Ratio Distributions

The 3D reconstructed slices of 8.5 mm × 8.5 mm × 0.14 mm (600 × 600 × 1 voxels) size were
Materials 2016, 9, 890 9 of 17
analysed to determine the change of void ratios with the height of the sample for all the load cases
(Figure 8a). A shift of the plot to the left indicates a reduction of void ratio under the effect of vertical
stress. It can be seen that the void ratios almost remained unchanged within the height of the sample
stress. It can be seen that the void ratios almost remained unchanged within the height of the sample
for all the load cases except for the initial case where the top 100 microns of the sample had a higher
for all the load cases except for the initial case where the top 100 microns of the sample had a higher
void ratio due to the uneven surface characteristics. Moreover, the change of void ratios with load
void ratio due to the uneven surface characteristics. Moreover, the change of void ratios with load
cases up to the yield stress of 14 MPa were insignificant (0.6 at no load to 0.57 at 1.8 MPa to 0.55 at 7.0
cases up to the yield stress of 14 MPa were insignificant (0.6 at no load to 0.57 at 1.8 MPa to 0.55 at
MPa) compared to the values observed for higher stresses exceeding 14.1 MPa where crushing of
7.0 MPa) compared to the values observed for higher stresses exceeding 14.1 MPa where crushing of
particles was encountered. The crushing of particles resulted in a more uniform distribution of void
particles was encountered. The crushing of particles resulted in a more uniform distribution of void
ratios along the height of the sample (Figure 8a).
ratios along the height of the sample (Figure 8a).
The spatial distribution of void ratios with the increase of vertical stresses was further
The spatial distribution of void ratios with the increase of vertical stresses was further
investigated by selecting eight sub-volumes or representative elementary volume (REV) of
investigated by selecting eight sub-volumes or representative elementary volume (REV) of
1.7 mm × 1.7 mm × 3.2 mm (side length >7D50) [35] with a total voxels count of 3,175,873 (Figure 8b,c).
1.7 mm × 1.7 mm × 3.2 mm (side length > 7D50 ) [35] with a total voxels count of 3,175,873 (Figure 8b,c).
It can be seen that the initial value of the void ratios for the REVs compared reasonably well with the
It can be seen that the initial value of the void ratios for the REVs compared reasonably well with the
initial void ratio of the whole volume (eo = 0.60). A wagon wheel plot of the void ratios calculated
initial void ratio of the whole volume (eo = 0.60). A wagon wheel plot of the void ratios calculated
from the image analysis for the REVs under different load cases is shown in Figure 8d. As expected
from the image analysis for the REVs under different load cases is shown in Figure 8d. As expected
the void ratio of the REVs decreases with the increase in pressure. Interestingly, the void ratio of the
the void ratio of the REVs decreases with the increase in pressure. Interestingly, the void ratio of the
REVs under a given pressure is observed to be almost the same. Moreover, these values were very
REVs under a given pressure is observed to be almost the same. Moreover, these values were very
close to the values obtained along the full height of the sample (Figure 8a), indicating image analysis
close to the values obtained along the full height of the sample (Figure 8a), indicating image analysis of
of a properly selected REV could produce meaningful outcomes of pressure-void ratio variations
a properly selected REV could produce meaningful outcomes of pressure-void ratio variations under
under very high pressure.
very high pressure.

(a)Change
Figure 8. (a) Changeofofvoid
voidratios
ratiosalong
alongthethe height
height of of
thethe sample
sample withwith increased
increased vertical
vertical stresses;
stresses; (b)
(b) Locations
Locations of eight
of eight sub-volumes
sub-volumes selected
selected forfor spatialanalysis;
spatial analysis;(c)
(c)3D
3Dimage
imageofofaasub-volume;
sub-volume; (d) Spatial
distribution of void ratios
ratios of
of sub-volumes
sub-volumes withwith increased
increased vertical
vertical stresses.
stresses.
Materials 2016, 9, 890 10 of 17
Materials2016,
Materials 2016,9,9,890
890 10of
10 of17
17

3.2.3. Particle Size Distributions

3.2.3. Particle Size Distributions
3.2.3. Particle Size Distributions
Initial Distributions
Initial Distributions
Initial Distributions
3D X-ray CT image of the initial sample was post-processed and separated using the procedure
3D X-ray CT image of the initial sample was post-processed and separated using the procedure
mentioned
3D X-ray in Figure
CT image 3. Two
of theexamples, one for
initial sample was nopost-processed
load and another andfor 14.1 MPa,
separated of the
using the process of
procedure
mentioned in Figure 3. Two examples, one for no load and another for 14.1 MPa, of the process of
separation
mentionedare presented
in Figure in Figure
3. Two examples,9. Theonevolume
for no of particles
load obtained
and another forfrom the label
14.1 MPa, analysis
of the was
process of
separation are presented in Figure 9. The volume of particles obtained from the label analysis was
used to calculate
separation the equivalent
are presented in Figuresphere diameter
9. The volume of of particles,
particles whichfrom
obtained wasthe then compared
label withused
analysis was the
used to calculate the equivalent sphere diameter of particles, which was then compared with the
particle size the
to calculate distributions
equivalentobtained from the of
sphere diameter sieve analysis
particles, (Figure
which was10).
thenIt is understood
compared withthat
thedifferent
particle
particle size distributions obtained from the sieve analysis (Figure 10). It is understood that different
size
size descriptors
distributions(e.g., Feret from
obtained diameter, equivalent
the sieve analysissphere
(Figure diameter)
10). It iswill have different
understood degreessize
that different of
size descriptors (e.g., Feret diameter, equivalent sphere diameter) will have different degrees of
success [38] when correlating with the particle size distribution curve obtained
descriptors (e.g., Feret diameter, equivalent sphere diameter) will have different degrees of success [38] from the mechanical
success [38] when correlating with the particle size distribution curve obtained from the mechanical
sieve
when analysis.
correlating Inwith
this the
study, the size
particle equivalent sphere
distribution curvediameter
obtainedoffrom
the 3D volume of particles
the mechanical was
sieve analysis.
sieve analysis. In this study, the equivalent sphere diameter of the 3D volume of particles was
considered
In this study, duethetoequivalent
its wide application
sphere diameterin laser
of particle size analysis.
the 3D volume It canwas
of particles be seen that thedue
considered particle
to its
considered due to its wide application in laser particle size analysis. It can be seen that the particle
size
widedistribution
applicationcurvein laserobtained
particlefrom
size the imageItanalysis
analysis. compares
can be seen well
that the with the
particle sizesieve analysis curve
distribution
size distribution curve obtained from the image analysis compares well with the sieve analysis curve
for the uniformly
obtained graded
from the image sand particles
analysis comparestested in this
well with thestudy. Existence
sieve analysis curveof insignificant
for the uniformlypercentage
graded
for the uniformly graded sand particles tested in this study. Existence of insignificant percentage
(<4%) of particles
sand particles smaller
tested in thisthan
study. 150Existence
microns of (Figure 10) could
insignificant be due to
percentage the tolerances
(<4%) of particlespermitted in
smaller than
(<4%) of particles smaller than 150 microns (Figure 10) could be due to the tolerances permitted in
the
150 average
microns opening
(Figure 10) of the testing
could be duesieves and
to the abrasionpermitted
tolerances of particles
in during
the averagesieving, whichofare
opening theusually
testing
the average opening of the testing sieves and abrasion of particles during sieving, which are usually
less
sievesthan
and 5% of the total
abrasion sampleduring
of particles [39]. sieving, which are usually less than 5% of the total sample [39].
less than 5% of the total sample [39].

Figure 9. Examples ofof imageprocessing

processingdepicting
depicting particle separation
andand identification forno(a)load;
no
Figure9.9.Examples
Figure Examples ofimage
image processing particle
depicting separation
particle separation identification
and for (a)
identification for (a) no
load;
and and (b) 14.1 MPa.
load;(b) 14.1
and (b)MPa.
14.1 MPa.
100
100

80
80

60
% Finer

60
% Finer

40
40
Mechanical sieve analysis
Mechanical
Image sieve
analysis analysis sphere dia)
(equivalent
20
20 Image analysis (equivalent sphere dia)

0
00.1 Particle size (mm) 1.0
0.1 Particle size (mm) 1.0

Figure 10. Initial grading obtained from mechanical sieving and image data.
Figure 10. Initial
Figure 10. Initial grading
grading obtained
obtained from
frommechanical
mechanicalsieving
sievingand
andimage
imagedata.
data.

Above 80% finer, the distribution obtained from the image analysis shows presence of particles
Above 80%
Above 80% finer,
finer,the
thedistribution
distributionobtained
obtainedfrom
fromthe theimage
imageanalysis
analysisshows
shows presenceof ofparticles
particles
of equivalent sphere diameters as large as 300 microns or even more. This valuepresence
is some 20% higher
of equivalent
of equivalent sphere
sphere diameters
diameters asas large
large as
as 300
300 microns
micronsor oreven
evenmore.
more. This
This value
value is
is some
some 20%
20% higher
higher
than the opening size of the largest mesh (250 microns) used in this study. The only possible reason
than the opening
thanrelatively
the opening size of
sizevolumethe largest
of the largest mesh
mesh (250 microns) used in this study. The only possible reason
for larger particles to (250
pass microns)
through used in this study.
the smaller The could
sieve size only possible reason
be related for
to the
for relatively
relatively larger
larger volumevolume particles
particles to to pass
pass through
through the the smaller
smaller sieve sieve
size size could
could be be related
related to the to the
large
large value of particle size anisotropy (which is discussed in a later section of this paper), and
large value of
value of particle particle size
size anisotropyanisotropy (which is discussed in a later section of this paper), and
orientation. A further verification(which is discussed
using the in a later
sand particles section
volume of thisfrom
obtained paper),
theand orientation.
image analysis
orientation. A further verification using the sand particles volume obtained from the image analysis
Materials 2016, 9, 890 11 of 17

Materials 2016, 9, 890 11 of 17

A further verification using the sand particles volume obtained from the image analysis confirms that
confirms that theofbulk
the bulk density the density ofinthe
sample is sample
good is in good
agreement with agreement with
the measured the measured
initial initial
bulk density bulk
(Table 2).
density (Table 2).
Table 2. Calibration of image data with physical measurement.
Table 2. Calibration of image data with physical measurement.
Mass-Volume-Density Relationships Image Analysis
Mass-Volume-Density Relationships Image Analysis
TotalTotal
volume of sand particles (Vs(V) =s) 0.09 3
volume of sand particles = 0.09cm
cm3
Specific gravity of sand particles
Specific gravity of sand particles (G(G ) = 2.68
s s) = 2.68
3 MassMass
of sand particles (Mc(M) =c)V=s V
GssGρswρw==0.24
Initial
Initial bulk
bulk density (ρii))==1.62
density (ρ 1.62g/cm
g/cm
3
of sand particles 0.24gm
gm
BulkBulk
volume (V ) = 0.15 cm 3 3
volumeb(Vb) = 0.15 cm
3
Bulk density = Ms /V
Bulk density = Mbs/V =1.60≈≈1.62
= b1.60 1.62 g/cm
g/cm3

Load-Dependent
Load-DependentDistributions
Distributions
Figure showsthe
Figure 11 shows theevolution
evolution of the
of the particle
particle size distribution
size distribution with vertical
with vertical stressesstresses
up to 79.3 upMPa.
to
79.3
For MPa. For case,
each load each the
load case, were
images the images
analysed were analysed
to obtain to obtain
the particle thedistributions
size particle size(PSD).
distributions
As there
(PSD). As little
was very therechange
was very little change
of particle of particle size
size distributions up distributions
to a yield stress upof to14a yield
MPa, stress
for theofsake
14 MPa, for
of clarity
the sake
of the of clarity ofonly
presentation the the
presentation
evolution onlyof PSDthefor
evolution
stresses of PSD than
higher for stresses
the yield higher
stressthan the yield
are presented.
stress are 3D
Both the presented.
volume of Both the 3Dand
particles volume of particlessphere
their equivalent and their equivalent
diameters weresphere
considereddiameters were
for plotting
considered
the PSD. for plotting the PSD.
The
The nature of ofthe
thegradual
gradual upward
upward shifting
shifting of PSD
of the the curves
PSD curvesin Figurein Figure
11 from11 no from no load
load condition
condition to subsequent
to subsequent higher loads higher loads indicates
indicates generation generation
of smallerofsize smaller size particles
particles due tocrushing.
due to particle particle
crushing. The development
The development of a pivotofpoint
a pivot point250
around around 250 microns
microns equivalent equivalent sphere diameter
sphere diameter and an
and an upward
upward
sifting ofsifting of PSD
PSD below below
this this size particle
size indicate indicatecrushing
particle predominantly
crushing predominantly at the pivot at point
the pivot point
and below.
and below.
Similar Similar observations
observations were reported were by reported
Nakata etby al.Nakata et al. [13].
[13]. However, thisHowever,
study finds thisrelatively
study finds less
relatively
breakage of less breakage
particles of particles
in terms in terms
of increase of The
of fines. increase
reason ofmight
fines. beThedue reason
to the might
smallerbemean due particle
to the
smaller
size (D50mean
= 230particle
microns)size in
(D50 = 230
this microns)
study in this
compared tostudy
a much compared to a much
higher value (D50 higher
= 1550value
microns)(D50 of
=
1550 microns)
Nakata of Nakata
et al. [13], whichetposes
al. [13], whichpotential
a higher poses a higher potential
to breakage [8]. to
On breakage
the other [8]. On the
hand, other hand,
particles above
particles above the
the 250 microns size250 microns
showed size showeddownward
an unexpected an unexpected shift ofdownward
the PSD curve shiftcompared
of the PSD curve
to the no
compared to the no load curve. This could be related to the large equivalent
load curve. This could be related to the large equivalent diameter of the small number of unseparated diameter of the small
number
particlesofinunseparated
the processedparticles in the processed
image. However, image. on
their influence However,
explaining their
theinfluence on explaining
particle crushing the
behaviour
particle crushing
of the sand samplebehaviour of the sandrelatively
can be considered sample cansmallbe (Figure
considered12). relatively small (Figure 12).

100
no load Initial parameters:
90 14.1 MPa
D50 = 230 micron
28.2 MPa Cu = 1.24
80
56.4 MPa
70 79.3 MPa
% Finer (by volume)

60

50

40

30

20

10

0
1.5 Particle volume × 106 (cu. micron) 15

142 179 205 225 243 258 271 284 295 306
Equivalent sphere dia (micron)

Figure11.
Figure Evolutionofofparticle
11.Evolution particlesize
sizedistribution
distributionfrom
fromcrushing
crushingatatdifferent
differentloads.
loads.

In
Inorder
orderto
tounderstand
understandthetheevolution
evolutionof
ofPSD
PSDwith
withincreased
increasedloading,
loading,the
thefrequency
frequencyof ofaadefined
defined
range
range of particles and their corresponding volume fraction with respect to the total volumeof
of particles and their corresponding volume fraction with respect to the total volume ofthe
the
sample were determined. A weighted frequency is calculated by multiplying the volume fraction of
a range of particle size with its frequency i.e., weighted frequency = number of particles (%) ×
volume fraction. Figure 12 shows the change of weighted frequency of various particle ranges with
Materials 2016, 9, 890 12 of 17

sample
Materialswere
2016, determined.
9, 890 A weighted frequency is calculated by multiplying the volume fraction of17
12 of
a range of particle size with its frequency i.e., weighted frequency = number of particles (%) × volume
increased
fraction. stresses.
Figure It is the
12 shows evident
change from the plot frequency
of weighted that crushing of particles
of various occurred
particle ranges withpredominantly
increased
within the
stresses. It is205 to 258
evident microns
from size.
the plot In crushing
that particular,of particles
particles of size range
occurred between 225within
predominantly and 243 themicrons
205 to
showed
258 microns a significant drop ofparticles
size. In particular, weighted frequency
of size with loading
range between 225 and(5.5
243for no load
microns to 3 for
showed 79.3 MPa),
a significant
indicating
drop a relatively
of weighted frequencyhighwith
crushing
loading of (5.5
particles
for nowith
loadsizes close
to 3 for toMPa),
79.3 the mean particle
indicating diameter high
a relatively (D50 =
230 microns).
crushing Interestingly,
of particles with sizes particle
close tosizes close to
the mean 0.93D50
particle diameter
diameter (D(i.e., 205–225 microns) showed a
50 = 230 microns). Interestingly,
very small change of the weighted frequency (5.5 for no
particle sizes close to 0.93D50 diameter (i.e., 205–225 microns) showed a very load vs. 5 for 56.4 MPa) except
small changefor the 79.3
of the
weighted frequency (5.5 for no load vs. 5 for 56.4 MPa) except for the 79.3 MPa, where a weightedof
MPa, where a weighted frequency value of approximately 4.5 associated with a higher degree
crushing value
frequency was observed. As expected,
of approximately the weighted
4.5 associated withfrequency
a higher of particles
degree below 205
of crushing wasmicrons size
observed.
As expected, the weighted frequency of particles below 205 microns size was observed to increase withof
was observed to increase with the increase of stress. This indicates the formation of new particles
smaller
the increase sizes at the expense
of stress. of crushing
This indicates of largerofparticles
the formation (>D50 size)
new particles with sizes
of smaller increased
at theloading.
expenseThis of
was alsoofreflected
crushing in the (>D
larger particles rapid50
reduction
size) with of the
increased effective
loading. particle
This diameter
was also (D 10) with
reflected in increased
the rapid
loading (Figure
reduction 11).
of the effective particle diameter (D10 ) with increased loading (Figure 11).

no load
14.1 MPa
5
28.2 MPa
56.4 MPa
79.3 MPa
4
Weighted frequency

0
<1.5 1.5-3.0 3.0-4.5 4.5-6.0 6.0-7.5 7.5-9.0 9.0-10.5 10.5-12.0 12.0-13.5 13.5-15.0
Particle volume × 106 (cu. micron)
<142 142-179 179-205 205-225 225-243 243-258 258-271 271-284 284-295 295-306

Equivalent sphere dia (micron)

Figure12.
Figure 12.Particle
Particlesize
sizefrequency
frequencydistribution.
distribution.

3.2.4.Particle
3.2.4. ParticleBreakage
Breakage
Therelative
The relativebreakage
breakageparameter
parameter(B(B r), which is defined by Hardin [8] as Bt/Bp where Bt = total
r ), which is defined by Hardin [8] as Bt /Bp where Bt = total
breakageand
breakage andBpBp==breakage
breakagepotential,
potential,waswascalculated
calculatedfor forall
allload
loadcases
cases(Figure
(Figure13a).
13a).ItItcan
canbebeseenseen
that the values of B for stresses below the crushing stress (14 MPa) are insignificant
that the values of Br for stresses below the crushing stress (14 MPa) are insignificant (Br ≈ 0). In the
r (B r ≈ 0). In the
post-crushing stress range, a linear variation between B
post-crushing stress range, a linear variation between Br and stress can be approximated, which is inin
r and stress can be approximated, which is
goodagreement
good agreementwithwithCoopCoopand andLee
Lee[40].
[40].This
Thislinear
linearvariation
variationofofBrBcan
r can bebeexplained
explainedwith
withthe thehelp
helpofof
the gradual change of the slope of the e-logσ’ plot in the post-crush region (Figure
the gradual change of the slope of the e-logσ’v plot in the post-crush region (Figure 4). It is believed
v 4). It is believed
thatififthe
that theapplied
appliedstress
stressisissignificantly
significantlyhigher
higherthan thanthe
thepresently
presentlyused
usedmaximum
maximumstressstress(79.3
(79.3MPa),
MPa),
thevalue
the valueofofBrBmay
r mayreach
reachan anasymptotic
asymptoticvaluevalueindicating
indicatingno nofurther
furtherbreakage
breakageofofparticles.
particles.Figure
Figure13b 13b
illustrates the breakage of particles at high stresses relative to no load condition. The breakage ofof
illustrates the breakage of particles at high stresses relative to no load condition. The breakage
particlesunder
particles underhigh
highstresses
stressesisismarked
markedby bycircles.
circles.TheThedifferent
differentcolours
coloursofofparticles
particlesininFigure
Figure13b 13b
indicate different intensity
indicate different intensity values. values.
Materials 2016, 9, 890 13 of 17
Materials 2016, 9, 890 13 of 17

Materials 2016, 9, 890 13 of 17

Figure 13. Particle breakage under different vertical stresses: (a) relative breakage; (b) images
Figure 13.Particle
Figure13. breakage
Particle under
breakage different
under vertical
different stresses:
vertical (a) relative
stresses: breakage;
(a) relative (b) images
breakage; showing
(b) images
showing breakage of particles.
breakage of particles.
showing breakage of particles.

3.2.5. Particle Size Anisotropy

3.2.5.Particle
3.2.5. ParticleSize
SizeAnisotropy
Anisotropy
Particle size anisotropy is defined as 1 minus the ratio of the smallest to the largest eigenvalue
Particlesize
Particle sizeanisotropy
anisotropy is defined
defined asas11minus
minusthe theratio
ratioofofthe
thesmallest
smallest to to
thethelargest
largesteigenvalue
eigenvalue
of the covariance matrix [34]. Anisotropy
Anisotropy measuresa aparticle’sparticle’s deviation from a spherical shape,
ofofthe
thecovariance
covariance matrix
matrix [34].
[34]. Anisotropy measures
measures a particle’sdeviation deviation fromfrom a spherical
a spherical shape,shape,
with a value
witha avalue of 1 indicating
valueofof11 indicating highly
indicating highly non-spherical
highly non-spherical particle and a value of zero indicating a fully
with non-sphericalparticle
particleandanda avaluevalueof ofzerozeroindicating
indicating a fully
a fully
spherical particle.
sphericalparticle.
particle.
spherical
Size anisotropy
Sizeanisotropy
anisotropy for for all
for all the particles
all the
the particles underno no loadcondition
condition was determined from the labelled
Size particles under
under noloadload conditionwas was determined
determined fromfromthethe
labelled
labelled
images
images and their distributions are plotted in Figure 14. It is clear that more than 80% of particles hadhad
and their distributions are plotted in Figure 14. It is clear that more than 80% of particles
images and their distributions are plotted in Figure 14. It is clear that more than 80% of particles had
anisotropy
anisotropyvalues
values greater than 0.6.
greater than 0.6. Therefore,
Therefore,ititisislikely,
likely,during
during mechanical
mechanical sieving,
sieving, thatthat a particle
a particle
anisotropy values greater than 0.6. Therefore, it is likely, during mechanical sieving, that a particle with
with
withaagiven
given volume
volume may pass pass aa square
squaremesh
meshsizesizewhich
whichis issmaller
smaller than
than thethe equivalent
equivalent sphere sphere
a given volume may pass a square mesh size which is smaller than the equivalent sphere diameter at
diameter
diameterat ataa suitable
suitable orientation
orientation as as opposed
opposedtotoa aparticle
particlewith
with thethe same
same volume
volume butbut withwith a lower
a lower
a suitable orientation as opposed to a particle with the same volume but with a lower anisotropy value.
anisotropy value.
anisotropy value. This
This phenomenon
phenomenon isis illustrated
illustratedininFigure
Figure15,15,where
where thirteen
thirteen particles
particlesof equal
of equal
This phenomenon is illustrated in Figure 15, where thirteen particles of equal volume (6.3 × 106 cubic
volume(6.3
volume (6.3××10
1066 cubic
cubic microns)
microns) areareshown
shownwithwithvarying
varyinganisotropy
anisotropy values
values (0.44 to 0.90).
(0.44 DueDue
to 0.90). to theto the
microns)
relativelyarelarge
shown with varying
long-axis anisotropy
dimension of valuesassociated
particles (0.44 to 0.90).
withDue to the
high relatively
anisotropy large long-axis
values, these
relatively large long-axis dimension of particles associated with high anisotropy values, these
dimension of particles
particles will
will have aaassociated
greater with high anisotropy values, these sizesparticles will have atheir
greater
particles have greater chance
chance ofof passing
passing through
throughsieve sieve sizessmaller smaller thanthan their
chance of passing
equivalent through sieve sizes smaller than their equivalent diameters.
diameters.
equivalent diameters.
With
Withthetheincrease
increase of of load,
load, particles undergo
undergo moregrindinggrindingand and breakage leading to particles
With the increase of load, particles
particles undergomore more grinding andbreakage
breakage leading
leading to particles
to particles
more
more spherical in shape. This results in decrease of anisotropy (particles become more spherical)with
spherical in shape. This results in decrease of anisotropy (particles become more spherical)
more spherical in shape. This results in decrease of anisotropy (particles become more spherical)
anwith
increase of loading
an increase which has
of loading beenhas
which depicted from thefrom
been depicted inwardthe shift
inwardof theshiftanisotropy distribution
of the anisotropy
with an increase of loading which has been depicted from the inward shift of the anisotropy
curves (Figurecurves
distribution 14). (Figure 14).
distribution curves (Figure 14).
100
100
90
no load
90
no load
80 14.1 MPa
80 14.1 MPa
28.2 MPa
70
28.2 MPa
Cumulative % of particles

70 79.3 MPa
60
Cumulative % of particles

79.3 MPa
60
50

50
40

40
30

30
20

20
10

100
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Anisotropy
0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Anisotropy
Figure 14. Anisotropy distributions of particles with vertical stresses.
Figure 14. Anisotropy distributions of particles with vertical stresses.

Figure 14. Anisotropy distributions of particles with vertical stresses.

Materials 2016, 9, 890 14 of 17
Materials 2016, 9, 890 14 of 17
Z
Z

500 microns
500 microns
Anisotropy = 0.86
0.60
0.60 Anisotropy = 0.86

0.70
0.70 0.57
0.57
0.47
0.47 0.67
0.80 0.56 0.67
0.55 0.80 0.56
0.55
0.81
0.44 0.90 0.81
0.44 0.90 Y
0.70 Y
0.70
X X

Figure 15. 15.

Figure Variation of of
Variation anisotropy
anisotropyofofparticles
particleswith
withequal
equal volumes (6.3 ×× 10
volumes (6.3 cubicmicrons).
1066cubic microns).
Figure 15. Variation of anisotropy of particles with equal volumes (6.3 × 106 cubic microns).
Figure 16 16
Figure shows
showsa bubble
a bubbleplot
plotrepresenting
representingthe the frequency
frequency distribution
distribution ofof particles
particlesofofvarious
various
sizes Figure
andand
sizes 16
their shows a
anisotropy
their bubble
anisotropywith plot representing
different
with differentvertical the frequency
verticalstresses.
stresses.ItIt is distribution
is observed of
that with
observed that particles
withthe of
theincreasing various
increasingload,
load,
sizes and their
anisotropy,
anisotropy, in anisotropy
general,
in with different
general,decreases
decreases andvertical
and the stresses.
the bubble
bubble sizeItrepresenting
size is observed that
representing thewith
the the increasing
frequency
frequency load,
ofof particles
particles
anisotropy,
increases
increases inindicating
general,
indicating decreases
formation
formation ofand
of the bubble
smaller
smaller size representing
sizeparticles
size particles due to
due the frequency of particles increases
to crushing.
crushing.
indicating formation of smaller size particles due to crushing.

1 1

0.9
0.9

0.8
0.8

0.7
0.7
Anisotropy

0.6
Anisotropy

0.6
0.5
0.5
0.4
0.4 no load
0.3 no load
14.1 MPa
0.3 14.1MPa
28.2 MPa
0.2 Note: Bubble indicates the number of particles at a given size and anisotropy 28.2MPa
56.4 MPa
0.2 value, the larger the size of the bubble the higher is the number of particles
Note: Bubble indicates the number of particles at a given size and anisotropy 79.3
56.4MPa
MPa
0.1 value, the larger the size of the bubble the higher is the number of particles
79.3 MPa
0.1
0
1.5 15 150
0 Particle volume × 106 (cu micron)
1.5 15 150
65 83 95 105 113 Particle
120 volume
126 132 × 106 (cu micron)178
137 142 205 225 243 258 272 284 295306
Equivalent sphere dia (micron)
65 83 95 105 113 120 126 132 137 142 178 205 225 243 258 272 284 295306
Equivalent sphere dia (micron)
Figure 16. Particle anisotropy and frequency distributions with vertical stresses.
Figure
Figure 16.
16. Particle
Particle anisotropy
anisotropy and
and frequency distributions with
frequency distributions with vertical
vertical stresses.
stresses.
4. Conclusions
4. Conclusions
4. Conclusions
High resolution X-ray CT in situ imaging of microstructure of sand particles subjected to high
pressure one dimensional
High resolution
resolution X-ray CTCTcompression
in situ leading
situ imaging
imaging to particle crushing
of microstructure
microstructure of sand
sandwas conducted.
particles The images
subjected to high
high
High X-ray in of of particles subjected to
taken
pressure at
onedifferent load
dimensional stages were
compression analysed to
leading capture
to the
particlemicro-structural
crushing was characteristics
conducted. including
The images
pressure one dimensional compression leading to particle crushing was conducted. The images taken
the void and particles volume distributions, change of void ratios, range of particles undergoing
taken
at at different
different load stages
load stages were were
analysedanalysed to capture
to capture the micro-structural
the micro-structural characteristics
characteristics including
including the
crushing, and distribution of particle size anisotropy. The outcomes of this study are summarised
the void
voidbelow. and particles volume distributions, change of void ratios, range of particles
and particles volume distributions, change of void ratios, range of particles undergoing crushing, undergoing
crushing,
and and distribution
distribution of particle
of particle size size anisotropy.
anisotropy. The outcomes Theofoutcomes
this studyof arethis study are below.
summarised summarised

below. The small-scale 1D compression apparatus setup designed for in situ X-ray CT imaging was
• Thefound
small-scale 1D compression
to produce comparable apparatus
results forsetup designed
e-logσ’ for in situ X-ray CT imaging was
v variations including the yield strength
 Theobtained
found small-scale
to produce 1D compression
comparable apparatus
results for setup
e-logσ’
from the lab-scale 1D compression test designed
variations for in situ
including
v performed by Nakata
X-ray
theetyield CTon imaging
strength
al. [13] was
obtained
uniformly
found
fromgradedto produce
the lab-scale comparable
1D compression
sand particles results for
test performed
of similar uniformity e-logσ’ variations
by Nakata
v
coefficients. Theet including
al. [13]
void ratios the
oncalculated
uniformlyforyield strength
graded sand
each load
obtained
particles from
of
cases from theimage
similar
the lab-scale
uniformity 1Dcoefficients.
analysis compression
were in good test
The performed
void
agreementratios bytheNakata
calculated
with et each
for
experimentalal. [13] oncases
load
data, uniformly
from
including
graded
the image
the sand particles
analysis
initial bulk wereofin
density similar
ofgood uniformity
agreement
the sand coefficients.
with
particles The void data,
the experimental
tested. ratiosincluding
calculatedthe forinitial
each bulk
load
cases from the image analysis
density of the sand particles tested. were in good agreement with the experimental data, including
the initial bulk density of the sand particles tested.
Materials 2016, 9, 890 15 of 17

• The increase of loads resulted in decreased global void ratios and pore volume sizes. Moreover,
the void ratio variations along the height of the sample showed a gradual decrease until the yield
stress and thereafter, a significant decrease instigated by marked particles crushing. With the
increase of stresses, the sinusoidal variations of void ratios encountered along the height of the
sample under low stresses were observed to diminish. A more uniform change of void ratio of
sub-volumes (REV) located at peripheral positions was also observed.
• The initial particle size distributions of sand particles obtained from mechanical sieve analysis and
that obtained from the image analysis using an equivalent sphere diameter were in reasonably
good agreement. The evolution of particle size distributions resulting from the crushing of
particles under incremental stresses could be captured using the non-destructive X-ray CT in situ
imaging. Moreover, the specific size group of particles predominantly subjected to crushing or
forming under different stresses could be identified. With increased stresses, the creation of more
fine particles associated with reduced values of size anisotropy was evident.
• The crushing of sand particles in the pre-yield stress region was insignificant as supported by the
near zero values of the relative breakage parameter. The value of the relative breakage parameter
was found to increase at the onset of yielding and thereafter, a linear variation with logarithm of
vertical pressure could be approximated.

The non-destructive X-ray CT imaging of micro-structure and analysis of image data for sand
particles subjected to high pressure one dimensional compression are believed to add significant insight
into the development of robust soil models using the discrete element modelling technique (DEM),
where the particles can be modelled as crushable with the capacity of accommodating microstructural
evolution. The outcomes will serve as the basis for future X-ray CT investigation, which are currently
underway, on cemented sand particles subjected to one dimensional compression.

Acknowledgments: The authors acknowledge the financial support provided by the Department of Civil
Engineering Research Fund of Monash University, Australia in preparation of the publication. The Australian
Research Council Linkage Infrastructure and Equipment Fund (LE130100006) for establishing the X-ray
Microscopy Facility for Imaging Geo-materials (XMFIG) at the Department of Civil Engineering of Monash
University is acknowledged.
Author Contributions: Asadul Haque and Asheque al Mahbub designed the experiments; Asadul Haque acquired
the X-ray CT images; Asheque al Mahbub post-processed the images and analysed the data; Asheque al Mahbub
and Asadul Haque wrote the manuscript.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Mitchell, J.K. Fundamentals of Soil Behaviour; Wiley: New York, NY, USA, 1976.
2. Yamamuro, J.; Bopp, P.; Lade, P. One-dimensional compression of sands at high pressures. J. Geotech. Eng.
1996, 122, 147–154. [CrossRef]
3. Terzaghi, K.; Peck, R.B. Soil Mechanics in Engineering Practice, 1st ed.; John Wiley & Sons, Inc.: New York, NY,
USA, 1948.
4. De Souza, J.M. Compressibility of Sand at High Pressure. Master’s Thesis, Massachusetts Institute of
Technology, Cambridge, MA, USA, 1958.
5. Chuhan, F.A.; Kjeldstad, A.; Bjorlykke, K.; Hoeg, K. Experimental compression of loose sands: Relevance to
porosity reduction during burial in sedimentary basins. Can. Geotech. J. 2003, 40, 995–1011. [CrossRef]
6. McDowell, G.R. On the yielding and plastic compression of sand. Soils Found. 2002, 42, 139–145. [CrossRef]
7. Hendron, A.J. The Behavior of Sand in One-Dimensional Compression. Ph.D. Thesis, University of Illinois,
Urbana, IL, USA, 1963.
8. Hardin, B.O. Crushing of soil particles. J. Geotech. Eng. 1985, 111, 1177–1192. [CrossRef]
9. Hagerty, M.M.; Hite, D.R.; Ullrich, C.R.; Hagerty, D.J. One-dimensional high pressure compression of
granular material. J. Geotech. Geoenviron. Eng. 1993, 119, 1–18. [CrossRef]
10. McDowell, G.R.; Bolton, M.D.; Robertson, D. The fractal crushing of granular materials. J. Mech. Phys. Solids
1996, 44, 2079–2102. [CrossRef]
Materials 2016, 9, 890 16 of 17

11. McDowell, G.; Bolton, M. On the micromechanics of crushable aggregates. Géotechnique 1998, 48, 667–679.
[CrossRef]
12. McDowell, G.R. A physical justification for e-logσ based on fractal crushing and particle kinematics.
Géotechnique 2005, 55, 697–698. [CrossRef]
13. Nakata, Y.; Hyodo, M.; Hyde, A.F.L.; Kato, Y.; Murata, H. Microscopic particle crushing of sand subjected to
high pressure one-dimensional compression. Soils Found. 2001, 41, 69–82. [CrossRef]
14. Altuhafi, F.N.; Coop, M.R. Changes to particle characteristics associated with the compression of sands.
Géotechnique 2011, 61, 459–471. [CrossRef]
15. Tengattini, A.; Das, A.; Einav, I. A constitutive modelling framework predicting critical state in sand
undergoing crushing and dilation. Géotechnique 2016, 66, 695–710. [CrossRef]
16. Muir Wood, D. Geomaterials with changing grading: A route towards modelling. In Geomechanics and
Geotechnics of Particulate Media; Hyodo, M., Murata, H., Nakata, Y., Eds.; Taylor & Francis: London, UK, 2006;
pp. 313–325.
17. Muir Wood, D. Critical state and soil modelling. In Deformational Characteristics of Geomaterial; Burns, S.E.,
Mayne, P.W., Santamarina, J.C., Eds.; IOS Press: Amsterdam, The Netherlands, 2008; pp. 51–72.
18. Bolton, M.D.; Nakata, Y.; Cheng, Y.P. Micro-and macro-mechanical behaviour of DEM crushable materials.
Géotechnique 2008, 58, 471–480. [CrossRef]
19. Cil, M.B.; Alshibli, K.A. 3D evolution of sand fracture under 1D compression. Géotechnique 2014, 64, 351–364.
[CrossRef]
20. Al-Raoush, R. Microstructure characterization of granular materials. Phys. A Stat. Mech. Appl. 2007, 377,
545–558. [CrossRef]
21. Alshibli, K.; Druckrey, A.; Al-Raoush, R.; Weiskittel, T.; Lavrik, N. Quantifying morphology of sands using
3D imaging. J. Mater. Civ. Eng. 2014, 27, 04014275. [CrossRef]
22. Druckrey, A.M.; Alshibli, K.A.; Al-Raoush, R.I. 3D characterization of sand particle-to-particle contact and
morphology. Comput. Geotech. 2016, 74, 26–35. [CrossRef]
23. Fonseca, J.; O’Sullivan, C.; Coop, M.R.; Lee, P.D. Non-invasive characterization of particle morphology of
natural sands. Soils Found. 2012, 52, 712–722. [CrossRef]
24. Karatza, Z.; Andò, E.; Papanicolopulos, S.A.; Ooia, J.Y.; Viggiani, G. Observing breakage in sand under
triaxial and oedometric loading in 3D. In Deformation Characteristics of Geomaterials, Proceedings of
the 6th International Symposium on Deformation Characteristics of Geomaterials, IS-Buenos Aires
2015, Buenos Aires, Argentina, 15–18 November 2015; IOS Press: Amsterdam, The Netherlands, 2015;
Volume 6, p. 431.
25. Hall, S.A.; Muir Wood, D.; Ibraim, E.; Viggiani, G. Localised deformation patterning in 2D granular materials
revealed by digital image correlation. Granul. Matter. 2010, 12, 1–14. [CrossRef]
26. Higo, Y.; Oka, F.; Sato, T.; Matsushima, Y.; Kimoto, S. Investigation of localized deformation in partially
saturated sand under triaxial compression using microfocus X-ray CT with digital image correlation.
Soils Found. 2013, 53, 181–198. [CrossRef]
27. Hashemi, M.A.; Khaddour, G.; François, B.; Massart, T.J.; Salager, S. A tomographic imagery segmentation
methodology for three-phase geomaterials based on simultaneous region growing. Acta Geotech. 2014, 9,
831–846. [CrossRef]
28. Faessel, M.; Jeulin, D. Segmentation of 3D microtomographic images of granular materials with the stochastic
watershed. J. Microsc. 2009, 239, 17–31. [CrossRef] [PubMed]
29. Zhao, B.; Wang, J.; Coop, M.R.; Viggiani, G.; Jiang, M. An investigation of single sand particle fracture using
X-ray micro-tomography. Géotechnique 2015, 65, 625–641. [CrossRef]
30. Razavi, M.R.; Muhunthan, B.; Al Hattamleh, O. Representative elementary volume analysis of sands using
X-ray computed tomography. Geotech. Test. J. 2007, 30, 1–8. [CrossRef]
31. Standard Test Methods for One-Dimensional Consolidation Properties of Soils Using Incremental Loading;
ASTM D2435/D2435M-11; ASTM International: West Conshohocken, PA, USA, 2011. Available online:
http://www.astm.org (accessed on 2 November 2016).
32. Microtest, version 6.13; Microtest materials testing modules; Deben UK Ltd.: Suffolk, UK, 2013.
33. Xradia, XMReconstructor—Cone Beam-10; Xradia, Inc.: Pleasanton, CA, USA.
34. Avizo 9.1.1; FEI Visualizations Sciences Group: Hillsboro, OR, USA, 2015.
 Materials 2016, 9, 890 17 of 17

35. Hasan, A.; Alshibli, K.A. Experimental assessment of 3D particle-to-particle interaction within sheared sand
using synchrotron microtomography. Géotechnique 2010, 60, 369–379. [CrossRef]
36. Chowdhury, B.; Haque, A.; Muhunthan, B. New pressure-void ratio relationship for structured soils in the
virgin compression range. J. Geotech. Geoenviron. Eng. 2014, 140, 1–5. [CrossRef]
37. Islam, S.; Haque, A.; Wilson, S.; Ranjith, P. Time-dependent strength and mineralogy of Lime-GGBS treated
naturally occurring acid sulfate soils. J. Mater. Civ. Eng. 2015, 28, 04015077. [CrossRef]
38. Fonseca, J.; Sim, W.W.; Shire, T.; O’sullivan, C. Microstructural analysis of sands with varying degrees of
internal stability. Géotechnique 2014, 64, 405–411. [CrossRef]
39. Carpenter, F.G.; Deitz, V.R. Methods of sieve analysis with particular reference to bone char. J. Res. Natl.
Bur. Stand. 1950, 45, 328–346. [CrossRef]
40. Coop, M.R.; Lee, I.K. The behaviour of granular soils at elevated stresses. In Proceedings of the Wroth
Memorial Symposium Predictive Soil Mechanics, Oxford, UK, 27–29 July 1992.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).

View publication stats