HOSTED BY Available online at www.sciencedirect.

com

ScienceDirect
Soils and Foundations 60 (2020) 1299–1311
www.elsevier.com/locate/sandf

Technical Paper

Role of particle shape in determining tensile strength and energy

release in diametrical compression of natural silica grains
Aashish Sharma a, Dayakar Penumadu b,⇑
a
CEE Department, The University of Tennessee, 851 Neyland Drive, Knoxville, TN 37996, USA
b
CEE Department, The University of Tennessee, 851 Neyland Drive, Knoxville, TN 37966, USA

Received 28 March 2019; received in revised form 25 August 2020; accepted 26 August 2020
Available online 15 September 2020

Abstract

Extensive particle fracture has been noticed in projectile penetration tests. However current models for predicting projectile penetra-
tion depths do not consider particle fracture. Therefore, in order to understand the role of particle shape on single grain strength and the
subsequent comminution process, single grain crushing tests on sand grains of different shapes and sizes were performed. High-
resolution, three dimensional images of grain surface, were created using confocal microscope and fracture of the grains were captured
with high-speed imaging. Acoustic emissions during single grain crushing was used to estimate the energy released during grain fracture.
It was found that local stress fields influence particle strength and in natural granular material surface flaws maybe the critical flaw caus-
ing fracture. The estimated critical flaw size causing fracture were submicron in size. The mass specific fracture energy increased with
increasing failure stress and decreasing particle size and is significantly greater than that computed from pure diametrical breaking.
Except for large angular grains, mass specific energies for different shapes were similar. Angular sands required multiple events to break.
The outcome of this work can be utilized in geomechanics to understand the comminution process and in power industry to estimate the
energy required for comminution.
Ó 2020 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society. This is an open access article under the CC BY-
NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

Keywords: Single grain crushing; Sand; Particle morphology; Acoustic emission; Fracture energy; High speed imaging

1. Introduction successful in estimating the penetration depth, however

they do not account for energy dissipation in particle frac-
Laboratory projectile penetration tests have shown ture. Also, simulations employing Discreet Element Model
extensive comminution along the path of the projectile, (DEM) generally do not consider particle breakage
and the formation of false tip composed of crushed grains (Takeda et al., 2018). With increasing sophistication of
at the tip of projectiles (Allen et al., 1957; Glößner et al., numerical models and cheaper hardware, multi-scale mod-
2017). Energy is therefore dissipated in fracturing particles eling is being undertaken in science and engineering to
in addition to mass movement, in friction and in the form understand the effect of micro and meso scale features on
of heat. Projectile penetration models based on cavity the macro scale behavior. Understanding the meso scale
expansion theory (Forrestal and Luk, 1992) are quite feature of particle fracture with some insights of micro
scale granular properties may contribute to the develop-
ment of models which can exhibit realistic energy dissipa-
Peer review under responsibility of The Japanese Geotechnical Society. tion in particle crushing in high stress applications.
⇑ Corresponding author. When the force on a particle exceeds its tensile strength,
E-mail addresses: asharma6@vols.utk.edu (A. Sharma), dpenuma- a grain fractures creating new surfaces and progenies of
d@utk.edu (D. Penumadu).

https://doi.org/10.1016/j.sandf.2020.08.004
0038-0806/Ó 2020 Production and hosting by Elsevier B.V. on behalf of The Japanese Geotechnical Society.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

different sizes. Fracture in brittle material is understood to momentum of the falling ball and deformation of the
occur when the stress intensity factor at one of the flaws, rod. The fracture energy was computed as the area under
either volume or surface, exceeds the critical value. The the force displacement curve. The authors state that the
crack then propagates instantly and uninhibited until it method is best suited for elastic spheres but could be
emerges at a boundary. This may lead to the formation applied to irregular particles if a large number of particles
of either two or many progenies of smaller sizes and differ- with similar shape were tested. Once again, the method
ent shapes. Energy is dissipated in the creation of new sur- may not be suitable for very angular particles as the initial
faces, sound, and motion of the progenies. The knowledge breaking and crushing of the asperities and surface features
of the ratio of energy dissipated in creating new surfaces to will be significantly different from grain to grain. In this
the overall energy required for comminution will help in work no assumption of particle shape is made for deter-
the design of efficient systems for creating fines. mining the fracture energy of natural sand grains. The
Fracture of single particles are governed by stress field at energy is directly measured using the non-destructive tech-
the contacts (Zhang et al., 1990). Stress field at the contacts nique of acoustic emission. The measured energy is then
are in turn influenced by the loading geometry and surface scaled by computing the elastic energy at fracture of a
features. These geometrical and loading features may man- smooth spherical manufactured fused silica grain.
ifest as decrease in strength with angularity (Nakata et al., In this work, single particle crushing data on rounded
2001b). The crushing strength then influences the stress and angular sands are presented. In order to better under-
strain behavior and the comminution process in high stress stand the cause of the variability of strength, in addition to
isotropic compression, one-dimensional (1D) compression angularity, thin sections were prepared to facilitate obser-
(McDowell, 2002; Nakata et al., 2001a), and triaxial load- vation of defects in the grains. Thin sections can also reveal
ing (Nakata et al., 1999). The effect of surface features on if a sand grain is single crystal or multi-crystalline. Further,
the strength of single grain is explained using high resolu- high-resolution three-dimensional (3D) images of grain
tion three-dimensional (3D) surface images captured with surfaces were created to gain further insights on the effect
confocal microscope. Also, the role of particle morphology of local stress field and grain-platen contact, on single par-
on fracture is shown with the help of high-speed images. ticle strength. High speed images of particle crushing were
It is generally agreed that the coordination number captured to observe fracture initiation and if possible,
influences particle fracture (Einav, 2007; McDowell et al., crack propagation during particle crushing. Energy
1996; Sammis et al., 1987; Tsoungui et al., 1999) and is released during crushing was measured using acoustic emis-
the dominant factor in the comminution process. Smaller sion (AE) for different particle shapes and sizes. These
particles with fewer neighbors are more likely to fracture observations have been synthesized to present a probable
before larger particles with higher coordination number. framework for comminution process and the observation
The dominance of the coordination number leads to a frac- of fractal nature of comminuted remains.
tal particle size distribution of the crushed particles which
has been observed in high stress (Zhong et al., 2018) and 2. Materials
large strain tests (Coop et al., 2004). Zhang et al. (1990)
on the other hand observed grain crushing occurring ini- Three different sands, Ottawa sand, Q-Rok and Euro-
tially in the larger particles with smaller particles remaining quarz Siligran were used for single grain crushing tests.
intact. Also, Kanda et al. (1985), and Tavares and King These sands have very similar mineralogy, more than
(1998) have shown increasingly more energy is required 99.5% quartz, but very different particle morphology.
to fracture smaller particles. The fracture energy measured Ottawa sand is unground silica with sub-rounded grains,
for grains of different shapes in this work may help to shed Q-Rok is unground silica with angular grains, and Siligran
more light on the interplay of size dependence of strength grains are sub-angular in shape. The maximum grain size
and the coordination number at various stages of the high of Ottawa sand and Q-Rok was 850 lm, and 710 lm for
stress compression. Siligran. Particle shape and size play major roles in deter-
Particle fracture energy has generally been computed mining the fracture strength of single grains. The sands
with the assumption of elastic spherical grains. Kanda were first washed and then divided into three size fractions
et al. (1985) integrated the force displacement curve to using US standard sieves; retained on #30 sieve (R30),
compute elastic energy at fracture assuming Hertzian con- passing #30 sieve and retained on #35 sieve (P30R35),
tact between a sphere and a flat plate. Though this and passing #35 and retained on #40 sieve (P35R40).
approach may be useful for spherical grains, it is inade- Shape parameters for each fraction were computed from
quate to compute displacements in irregular particles. two dimensional (2D) images using computational geome-
Tavares and King (1998) placed single particles on the try (Zheng and Hryciw, 2016) and are presented in Table 1.
end of a long rod and applied impact load by dropping a The values obtained from the computational geometry
ball on the particles. The elastic waves produced in the algorithm are comparable to those presented in
rod from impact were recorded by strain gage. Forces Krumbein and Sloss (1963). Roundness indicates the
and displacements were computed assuming one- degree of angularity; roundness values close to unity indi-
dimensional wave propagation and conservation of cate smooth rounded grains. Sphericity indicates the close-
1300
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

Table 1 the load cell frame enabling very precise and zero compli-
Shape parameters for different size fractions for Ottawa sand, Q-Rok and ance displacement measurements. The base plate was
Siligran.
5 mm thick and made of Silicon Carbide (SiC). The bound-
Fraction Sand Median d [lm] Roundness Sphericity ary conditions in single grain crushing can have significant
R30 Ottawa 644 0.73 0.85 influence on the fracture force (Shipway and Hutchings,
Q-Rok 524 0.49 0.76 1993a) as it affects the contact area with the grain, and con-
Siligran 555 0.66 0.82
P30R35 Ottawa 486 0.73 0.82
sequently the stress distribution within the grain. Hard pla-
Q-Rok 461 0.45 0.74 tens were chosen to reduce the contact area between the
Siligran 479 0.55 0.77 grain and the platen and to determine the smallest critical
P35R40 Ottawa 404 0.68 0.85 fracture force. Single grains were placed on SiC base plate
Q-Rok 387 0.45 0.74 and compressed with the diamond surface of the flat punch
Siligran 411 0.59 0.78
indenter at 0.1 mm per minute. A seating load of 0.2 N was
used to ensure seating of the indenter on the sand grains.
ness of the shape of the 2D image of the grain to a circle The initial height of the specimen was measured as the dif-
and is computed as the ratio of the length and width of ference in the displacement values at the base and the top
the grain. Sphericities of all the sands are similar, and of the grain at seating load. Approximately 30 specimens
though the roundness values are different for different were tested for each size fraction.
sands, it does not vary significantly for the different size
fractions. Thin section images of sands observed under 3.2. Single particle crushing with acoustic emission (AE)
cross polarized light indicated that Ottawa sand grains measurement
were single crystal. Q-Rok grains may contain polycrys-
talline grains, and Siligran grains may contain other min- A separate testing system, Psylotech l-TS was used for
eral inclusions or bulging recrystallization regions as single grain crushing test with AE measurement. The sys-
shown in Fig. 1. In addition to the sand grains, four tem was equipped with 1.6 kN load cell and 9 mm linear
350 lm diameter and three 800 lm diameter fused silica variable displacement transformer (LVDT). A proprietary
spheres were also tested. technology incorporated in the system enabled a floating
window scale which could resolve displacements to sub-
micron resolution and force to sub newton resolution. This
3. Experimental methods floating window keeps moving as the limits of the window
is reached enabling the record of both full-scale measure-
3.1. Single particle crushing ments and high-resolution windowed measurements
throughout the duration of the test. This presents a unique
Most of the single grain diametrical compression tests opportunity to measure small variations in force due to
were performed with Zwick Z2.5 hardness testing machine. surface features and at the same time the capability to frac-
The hardness measuring head was equipped with a 200 N ture particles of reasonable size. The base plate and flat
loadcell, a 0.02 lm resolution depth measuring device, punch diamond indenter used in the Psylotech system were
and a 3 mm diameter flat punch diamond indenter. The the same as those used with Zwick system. Mistras Group
transducer that measured displacements was separate from AEWIN was used to record and analyze AE signals. A
general-purpose AE sensor, R15a, was attached to the
loading head as shown in Fig. 2. The sampling rate was
1 MHz, and the threshold amplitude was set at 40 dB to
cancel testing system vibrations. A 120 kHz high pass filter
was applied to cancel audible noise. Approximately 25
specimens were tested from the R30 size fraction and 15
specimens from P35R40 size fraction.
AE measures transient elastic waves that is generated in
a medium when energy is suddenly released. In engineering
material, generation of elastic waves generally coincides
with the formation and propagation of cracks. AE can cap-
ture this dynamic phenomenon. The five parameters that
are generally used to characterize AE signals are amplitude
(A), counts (C), duration (D), rise time (RT) and the mea-
sured area under rectified signal envelope (MARSE), which
is also known as energy (E), with arbitrary units. These
parameters are shown in Fig. 3. An event is triggered when
Fig. 1. Thin section images of multigrain Q-Rok (a) - (b), mineral the rising signal crosses the threshold and lasts until the sig-
inclusion in Siligran (c), and bulging recrystallization in Siligran (d). nal is smaller than the threshold. Though the sensor was
1301
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

AE has diverse applications, spanning from the study of

damage and damage propagation in brittle materials (Dai
and Labuz, 1997) to health monitoring of bridges (Yapar
et al., 2015). Particularly in the study of material damage,
AE has been used to investigate damage in concrete
(Ohtsu and Watanabe, 2001), relate acoustic energy to
fracture energy in concrete beams (Vidya Sagar and
Raghu Prasad, 2011), monitor failure in fiber composites
(Bohse, 2000; De Rosa et al., 2009), monitor tests on coarse
and fine grained soils (Koerner et al., 1981), and crushing
of coarse grained soils (Brzesowsky et al., 2014; Li et al.,
2018; Muñoz-Ibáñez et al., 2019). Recently, the technique
has also been used to monitor single grain crushing tests
(Ibraim et al., 2017; Mao and Towhata, 2015). However,
determination of fracture energy from these measurements
for different particle shapes and sizes have not been
attempted in literature.

3.3. Surface imaging

High resolution KEYENCE VK-X200 series 3D laser

scanning confocal microscope was used for very fine reso-
lution 3D images of few grains at 10X magnification. Con-
focal microscopes block out of focus-plane light using a
spatial pinhole, allowing light only from the focal-plane.
A 3D image of the object is then constructed from 2D
images at various heights by moving the focal plane from
the bottom to the top of the object. Single sand grains were

Fig. 2. Psylotech testing system for particle crushing with AE

measurement.

Fig. 3. Explanation of typical AE parameters.

attached to the loading head, energy released at fracture is

quickly dissipated via the lower base and free surfaces in
the loading head with unknown proportion of energy
released reaching the sensor. Therefore, manufactured
spherical fused silica spheres were tested in the same config-
uration for scaling purpose by equating the measured Fig. 4. High resolution 3D surface images of (a - b) Ottawa sand, (c - d)
energy with the elastic energy at facture. Q-Rok, (e - f) Siligran. All images have same scale as the shown scalebar.
1302
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

placed on a dark surface and the start and end focus eleva-
tions were set below the base surface and above the top of
the grain respectively. Three dimensional images for the
three different sands displaying typical surface features of
the sands are shown in Fig. 4. In addition, six grains were
imaged at higher resolution of 150X to determine the sur-
face roughness of the grains. For each grain the roughness
was determined from approximately 50 lm square area as
the average sum of the differences of the surface points
from the mean surface height. The roughness values of
the sands were similar; 0.35, 0.52 and 0.20 lm for Ottawa
sand, Q-Rok and Siligran respectively. These values are
higher than those reported by (Altuhafi et al., 2016).

3.4. High speed imaging

High speed images enable visualizing the formation and

propagation of cracks in engineering materials. High speed
images of particle crushing were captured with Photron
Fastcam APX RS high speed camera. A low magnification
lens was attached to the camera and sand grains were illu-
minated using two focused light beams. Due to explosive
nature of fracture, the frame rates for Ottawa sand and
Siligran were 3000 frames per second (fps), whereas the
frame rate for Q-Rok was 150 fps for longer imaging time
to capture asperities breaking and the final breakage.

4. Results and discussion

4.1. Effect of particle morphology on crushing behavior

Single grain crushing tests assumes that fracture occurs

from tensile stresses (r) along the axis of loading and is
computed using Eq. (1) (McDowell, 2001; Nakata et al.,
2001b).
F
r¼ ð1Þ
d2
where, F is the force at failure and d is the height of the par-
ticle between the two flat surfaces. Example stress–strain
curves for five grains are shown in Fig. 5. The stress is com-
Fig. 5. Response of the sand grains to diametrical compression, (a)
puted using Eq. (1), and engineering strain is computed by Ottawa sand, (b) Q-Rok, (c) Siligran. The length values in the parenthesis
dividing the displacement by the original height of the par- are particle sizes in the direction of loading.
ticle. During diametrical crushing between two flat platens,
the particles are in contact at their extremity points (highest
and lowest) with the platens. For grains devoid of sharp increases the likelihood of encountering a critical flaw lead-
asperities, with few contact points, the stress–strain rela- ing to failure (Shipway and Hutchings, 1993b). There are
tionship for Ottawa sand (O) and Siligran (S) are typical similar observations with Siligran specimens; the strength
as shown in Fig. 5. The influence of surface features at of S2 (Fig. 4(e)) is significantly lower than that of S5
the contacts on the strength of the grains are evident when (Fig. 4(f)), which has a large flat contact surface. With
viewed in conjunction with the surface images shown in highly angular grains there will be multiple contact points
Fig. 4. The small contact area due to the pointed top sur- of the grains with the bottom surface for stability. During
face of O2 (Fig. 4(a)) coincides with lower strength. The the application of force, the extremity points which may be
O5 (Fig. 4(b)) with the flatter contact surface has higher asperities will break before the particle fractures. Also, on
strength. With smaller contact area, maximum tensile stress loading, the particles may rotate for greater stability
occurs nearer to the surface, and the volume of the grain (Cavarretta and O’Sullivan, 2012). These asperities break-
under tensile stress along the loading axis increases. This ing and readjustments appear as numerous reductions in
1303
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

force as observed in the curves for Q-Rok (Q). The break- High speed images captured during these tests further
ing of the small asperities in Q2 (Fig. 4(c)) and the ultimate illustrate the role of particle shape and surface features
failure is seen in the response of Q2 in Fig. 5. Furthermore, on particle strength. Ottawa sand and Siligran generally
if the grains are polycrystalline, the bond between the exhibited diametrical fracture as seen in Fig. 6. Surface fea-
grains which are weaker, will determine the strength of tures, however, may profoundly influence the type of fail-
the grain. The polycrystalline nature of Q4 can be seen in ure; the large surface depression (Fig. 4(e)) is responsible
Fig. 4(d), which fractured at a very low stress with three for failure in S2, as the particle fails by chipping at the flaw
large progenies. (Fig. 6(m)- (o)). The asperities breaking in Q2, flattens the

Fig. 6. Highspeed images of particle fracture, showing the effect of particle shape and surfaces features on the type of fracture.
1304
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

contact surface during the initial stages of loading, it is fol-

lowed by some rotation of the grain, and failure occurs by
crack opening at the top. In Q4, the fracture occurs along
the grain boundaries of the polycrystalline grain. In most
of the tests, probably due to the harder diamond surface
at the top, failure was generally seen to initiate from top
of the grains.

4.2. Single particle crushing

The results of single grain crushing are modelled using

the two parameter Weibull distribution (Weibull, 1951)
with a scale and the shape parameter as shown in Eq. (2)
(McDowell, 2001; McDowell and Amon, 2000; Nakata Fig. 7. Weibull plot for the largest size cluster for the three sands.
et al., 1999):

m 
r ent sands indicate different flaw populations. For similar
P s ðV o Þ ¼ exp  ð2Þ size range silica sand particles, Nakata et al. (2001a,b) have
ro
reported values of 72.9 MPa and 2.17 for r0 and m respec-
where, P s ðV o Þ is the survival probability of a particle of vol- tively and for Ottawa sand Cil and Alshibli (2012) have
ume V o , r is the nominal stress as given in Eq. (1), ro is the reported values of 137.9 MPa and 3.26. The value of m
characteristic strength at which 37% of the particles sur- for rounded quartz of similar size is in the range of 2.57
vive, also known as the scale parameter, and m is the Wei- (Kanda et al., 1985). For engineering ceramics, m can be
bull modulus or shape parameter, indicating the variability as high as 10. In general, low values of m for Q-Rok sug-
of the test data. The value of m decreases as variability in gests high variability of the measured strength. Even for
the data increases. The survival probabilities were mean very low m values ro increases with decreasing size. As m
rank survivability (Davidge, 1979). does not vary significantly over different size clusters, a
Size fractions based on sieve analysis results in a very Weibull analysis with the assumption of single flaw popu-
non-uniform particle size. As the grains are bouncing in lation across different sizes may be used to determine size
the mesh, particles with their smallest dimension smaller dependent strength for these sands in the size range pre-
than the mesh size can pass through the openings. Particles sented. However, Kschinka et al. (1986), based on tests per-
were then chosen randomly based on visual estimation of formed on ten different sizes of glass spheres caution that
equal size which may result in particles of varying height Weibull approach may hide multiple flaw population and
in the direction of loading. Therefore, the height of the par- data consistent with single flaw population across different
ticles, immediately before the application of the load was size range may just be accidental. Additionally, Nakata
used to group the particles into three groups based on k- et al. (2001b) report decreasing m with decreasing particle
means clustering algorithm. When the desired number of size and increasing angularity in quartz sand, and Kanda
groups is specified, the algorithm begins by randomly gen- et al. (1985) have shown decreasing m with decreasing size
erating the size centers for the desired number of groups in quartz spheres.
and computing the distances to each particle from these The relationship between crushing strength and the size
centers. The data is then clustered into groups by minimiz- of the grains is shown in Fig. 8 and is a power law of the
ing within group variance. A new centroid for each group is form r / da (McDowell, 2001; McDowell and Amon,
then computed and the clustering is repeated until the 2000; Nakata et al., 2001b). The slope (a) of the regression
change in the centroid is less than a specified tolerance. line in a log–log plot were 0.75, 1.78, 0.92, and 0.63
The Python module scikit-learn (Pedregosa et al., 2011) for Ottawa sand, Q-Rok, Siligran and fused silica. With
was used for clustering single particle crushing data based increasing angularity there is a steeper decrease in strength.
on grain heights into three clusters. The values for quartz sand may range from 0.79 (Nakata
Weibull plots for the largest size clusters for the three et al., 2001b) to 0.92 (McDowell, 2001). The angular nat-
sands are shown in Fig. 7 and the parameters for all size ure of Q-Rok coupled with increasing likelihood of multi-
clusters are presented in Table 2. With increasing angular- crystallinity in larger grains may have contributed to the
ity both ro and m decreases, the contributing factor, shaper reduction in strength with particle size. Assuming
though, is not only angularity but also surface features volume flaws, the values of m determined as m ¼ 3=a
and volume flaws present in Q-Rok and Siligran as shown are 4, 1.69, 3.26, and 4.76 respectively for Ottawa sand,
in Fig. 4 and Fig. 1. Despite all three sands consisting of Q-Rok, Siligran and fused silica. These values for the sands
more than 99.5% quartz, there are significant variations are larger than those presented in Table 2 for all size frac-
in strength due to shape, surface features and volumetric tions. Assuming surface flaws the m values computed as
flaws. Therefore, the different m values for the three differ- m ¼ 2=a are 2.67, 1.12, and 2.17 which are closer to the
1305
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

(Cavarretta and O’Sullivan, 2012) resulting in lower stres-

ses at the bottom than at the top (Turner et al., 2016).
Analysis of high-speed images also showed crack propagat-
ing from the top of the specimens in most specimens. Wang
and Coop (2016) have reported similar observation from
high speed images during fracture for Leighton Buzzard
sand. Therefore, with the assumption that higher stresses
occur in the vicinity of contact with the diamond surface,
cf was computed using mg ¼ 0:077, Eg ¼ 95:68GPa,
pffiffiffiffi
mp ¼ 0:20, Ep ¼ 1100 GPa and K IC ¼ 1 MPa m
(Brzesowsky et al., 2011; Ferguson et al., 1987). Character-
istic flaw size (cfo ) computed assuming a Weibull distribu-
tion of critical surface flaws are also shown in Table 2.
For Ottawa sand with larger m, the flaw size becomes a
Fig. 8. Size dependence of strength for Ottawa sand, Q-Rok, Siligran, and
fused silica spheres.
greater proportion of the grain with decreasing particle size
(McDowell and De Bono, 2013). The flaw size in Q-Rok,
with smaller m, increases almost an order of magnitude fas-
values reported in Table 2. Given the highly uneven sur- ter with increasing grain size. The values of cfo for Ottawa
faces seen in Fig. 4, it is likely that surface flaws may have sand and Siligran are smaller than cf = 0.115l m reported
dominated the fracture. by Brzesowsky et al. (2011) for quartz sand in The
Surface flaws can significantly reduce the strength of pffiffiffiffi
Neatherlands. Using K IC ¼ 0:31 MPa m as Zhang et al.
grain; Shipway and Hutchings (1993a) report that the frac- (1990) the values of cfo are within the range of 0.004–
ture force reduced from 167 N for smooth glass spheres to 0.07 lm reported by them. Flaw sizes in Q-Rok are three
70 N for abraded glass spheres. The size of Griffith type times as large as those in Ottawa sand for the largest cluster
flaw on grain surface at the edge of the circular Hertzian size and twice as large for the smaller clusters. The poly-
contact area is computed employing linear elastic fracture crystalline nature of the larger grains of Q-Rok may have
mechanics as shown in Eq. (3) (Brzesowsky et al., 2011; contributed to this difference. Though the values of cfo
Zhang et al., 1990). are smaller than the surface roughness values presented

4=3 earlier, they are in the same range or marginally smaller
K2IC 4p 3rg
cf ¼  2  ð3Þ than roughness values from interferometer measurements
Y2 1  2mg F2=3 4E presented by Altuhafi et al. (2016).
Assuming surface flaws, and linear scaling of flaw size
where, cf is the size of the critical flaw size on the surface with particle size (cf =rg ¼ constant) Zhang et al. (1990)
and radial in direction, equivalent to edge crack in a plate, arrived at excellent agreement between predicted crushing
which causes failure under diametrical force F , Y is a pressure in one dimensional (1D) compression and experi-
dimensionless factor equal to 1.12 for single edge crack, mental data. For linear scaling, the values of cf =rg are
K IC is the critical stress intensity factor, rg is the grain
  6.93  105, 4.57  104, and 1.17  104 respectively
 for Ottawa sand, Q-Rok, and Siligran. These values are
radius and 1=E ¼ 1  m2g =Eg þ 1  m2p =Ep where mg
smaller than those reported by Zhang et al. (1990) for
and Eg , mp and Ep are the Poisson’s ratio and elastic mod- quartz sand. A power law scaling (cf / d a ) is shown in
ulus of the grain and platen respectively. In compression Fig. 9. The values of the exponent are 0.50, 1.19, and
tests of irregular natural particles, there will be greater 0.61 for Ottawa sand, Q-Rok, and Siligran respectively.
number of contact points at the bottom for stability

Table 2
Weibull analysis for fracture stress and flaw size for different size clusters of Ottawa sand, Q-Rok and Siligran.
Stress Flaw size,cf
Centroid[lm] Sand Number of grains ro [MPa] m cfo [lm] m
663 Ottawa 36 144 2.89 0.05 4.10
699 Q-Rok 22 49 1.12 0.14 1.63
737 Siligran 59 93 1.84 0.07 2.66
532 Ottawa 42 157 2.87 0.04 4.36
513 Q-Rok 64 87 1.31 0.09 1.80
526 Siligran 50 96 2.49 0.06 3.71
395 Ottawa 49 210 2.61 0.04 4.02
372 Q-Rok 38 130 1.27 0.07 1.64
389 Siligran 19 138 1.93 0.05 2.81

1306
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

typical for Ottawa sand and Siligran. In majority of the

cases, failure coincides with the maximum force and maxi-
mum energy. Q-Rok on the other hand, exhibits signifi-
cantly greater number of events, that are also called hits.
The median number of events leading to failure were 4,
23, 10 for Ottawa sand, Q-Rok and Siligran respectively.
The energy released in each breakage is also significantly
lower in Q-Rok than in Ottawa sand and Siligran. How-
ever, in some cases, the cumulative energy released in all
the hits leading up to the fracture may exceed that of
Ottawa sand. Also, the energy released at the peak force
may not always coincide with maximum energy in Q-
Rok. The median values of E at maximum force were
Fig. 9. Flaw size scaling with particle size for sands with different particle 1625, 606, 913 and the median value of the cumulative E
shape, Ottawa sand, Q-Rok, Siligran, and fused silica spheres. at break were 1820, 1172, 1342 for Ottawa sand, Q-Rok
and Siligran respectively. Median values for AE parame-
The increase in flaw size with particle size is twice as much ters defined in Fig. 3 are shown in Table 3. The values
in Q-Rok than in Ottawa sand and Siligran. for Ottawa sand and Siligran are similar for most parame-
The values of m for Griffith type surface flaw analysis ters, except for E. In tests in which the grain did not com-
(Table 2) are higher than those from Weibull analysis for pletely break at maximum force, the definition of break is
all sands. The difference is larger for Ottawa sand than shown in Fig. 10(b). The cumulative energy released at
angular Q-Rok. The computed values of cfo for Ottawa the breakpoint is considered as the fracture energy. The
sand do not vary as much as those for Q-Rok for the dif- relationship between maximum force on the particle and
ferent size fractions. In highly angular Q-Rok, the size of the energy released at fracture, E, is shown in Fig. 11.
the surface features which may contribute to failure also For the range of size tested, E is a linear function of max-
increases with particle size. Zhang et al. (2016) have shown, imum force on the particle, irrespective of particle shape.
adopting Hertzian contact and linear fracture mechanics, Higher energy is released with increasing force. One of
that the assumed location of critical flaw, either surface the 800 lm fused silica spheres failed at a very high load
or volume, influences the size dependence of strength. of 317 N, though the energy associated with the fracture
The two assumptions, surface, and volume flaws, bound is smaller than other 800 lm fused silica spheres which
the Weibull analysis, with surface flaws presenting a lower fractured at a smaller force. The Q-Rok and Siligran grains
bound and center crack (volume flaw) the upper bound for with very high energy release were associated with greater
the m value. However, in the current analysis m values pre- number of events leading up to the failure.
dicted assuming critical surface flaw (m associated with cfo ) The relationship between elastic strain energy at fracture
are larger than those determined from random distribution and particle strength have been shown to be a power law
of flaws in Weibull analysis (m associated with r0 ). relationship (Kanda et al., 1985; Yashima et al., 1987). In
the above discussion the energy E was in arbitrary units,
which is reasonable for comparison purpose. AE systems
4.3. AE emissions also compute absolute energy by integrating the squared
of the voltage with time and dividing by a reference resis-
Example force history curves along with the AE energy tance. Though, it must be noted that the absolute energy
releases are shown in Fig. 10. Single to few AE events are computed in this manner is not the energy released during
fracture. Due to small sensor area compared to the free sur-
face of the base platen and loading head, energy recorded
by the sensor is only a fraction of the energy released.
Energy is also dissipated in the audible range and lost
due to wave attenuation in elastic medium. The energy,
thus, computed from the recorded signal is significantly
lower than the actual energy released. Mass specific abso-
lute energy was determined by dividing the measured AE
absolute energy (ABS-E) by the mass of the particle height
equivalent sphere. This measured mass specific energies for
all grains were scaled based on the elastic energy at fracture
of 350l m spheres. Hertzian contact was assumed between
flat platens and the sphere, and the mass specific elastic
Fig. 10. Example force and energy time histories for Ottawa sand, Q-Rok, energy (ESE/M) at one contact was computed as shown
and Siligran. in Eq. (4):
1307
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

Table 3
Median values of characteristic AE parameters at maximum force.
Sand Amplitude, A[dB] Rise time, RT[ls] Duration, D[ls] Counts, C MARSE, E
Ottawa sand 99 41 2080 225 1625
Q-Rok 91.5 43.5 1699 178 606
Siligran 95 41 1926 204 913

Fig. 11. Maximum force required to fail particles of different sizes and
Fig. 12. Energy released at grain fracture expressed as mass specific
energy released in fracture. The color bar scale has been limited to the
energy for sand grains and spheres as a function of strength.
median energy released in Ottawa sand.

5=3 mass specific energies for the grains are lower than the
1 F
ESE=M ¼ 2:4961 ð4Þ upper bound line. The mass specific energy decreases at a
pqE2=3 d 2 faster rate with decreasing failure stress for sand particles
where q is the density of the grain (2.65 and 2.22 g/cm3 for than for the spheres. The mass specific energies computed
sand and fused silica respectively) and other terms have by Kanda et al. (1985) were similar to those computed
been previously defined for Eq. (3). The values for m and from Eq. (4).
E for fused silica sphere were 72 GPa and 0.17 and those The mass specific energy required to crush particles
used for silicon carbide base were 410 GPa and 0.16 respec- increases with decreasing particle size as shown in
tively. The values for sand grains and diamond platen were Fig. 13. The values computed by Kanda et al. (1985) and
the same as used in Eq. (2). The above equation was Yashima et al. (1987) are similar for larger particle sizes
applied separately to each contact of the sphere with the but significantly underestimate energies released by smaller
top diamond plate and the bottom silicon carbide plate, particles. They computed elastic energy assuming Hertzian
and the total specific energy was computed as the sum of contact and employed Weibull relationship for size depen-
the two. Similar approach was followed by Kanda et al. dence of strength. King and Bourgeois (1993) have
(1985) and Yashima et al. (1987), they computed compres- reported similar values for 0.5–0.7 mm quartz and the val-
sive strength of spheres using r ¼ 0:9F =d 2 as proposed by
Hiramatsu and Oka (1966). Here Eq. (2), which is generally
used in geomechanics is consistently used for both sand
grains and silica spheres. The average mass specific elastic
energy for two 350l m spheres from AE measurement
was 0.0022 J/kg and that obtained from Eq. (4) was
20.278 kJ/kg, a scaling factor of 9.221  106. All specific
energy values were scaled using this scaling factor and
are shown in Fig. 12. Also shown in Fig. 12 are mass speci-
fic energies for spherical fused silica and quartz at different
failure stresses computed using Eq. (4). The mass specific
energies for the spheres appear as the upper bound limit.
At higher failure stresses which generally correspond to
smaller particle sizes the energies released at fracture are
close to this upper bound. The Q-Rok grains which are
close to the upper bound are associated with multiple Fig. 13. Energy released at fracture for different grain sizes expressed as
events leading up to fracture. At lower failure stress the mass specific energy.
1308
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

ues reported by Tavares and King (1998) for quartz for size specific energy with increasing grain size. Also, the com-
range tested in this study are also similar to those com- pressive strength decreases with increasing size and the par-
puted in this study. As particle size decreases greater energy ticles with higher specific energies are associated with high
is required for fracture. Also shown in Fig. 13 is the mass fracture stress. Hence both strength and specific energy
specific energy computed assuming a single diametrical show similar trend with particle size, and King and
break and surface energy. The average value of surface Bourgeois (1993) have noted that the distribution of mass
energy can range from 0.675 J/m2 (Brace and Walsh, specific fracture energy is identical to the probability of
1962) to 2.678 J/m2 (Tromans and Meech, 2004), the line fracture.
shown in the figure was drawn using 2.678 J/m2. The mass
specific energies of smaller particles are significantly larger 5. Conclusions
than that computed assuming diametrical breaking. With
the high stresses required to fracture smaller particles there In this research work, results of single grain crushing
is a greater possibility of larger number of progenies than tests have been presented along with some detailed imaging
that from a diametrical break. With increasing particle size of grain surfaces, high speed imaging of fracture and mea-
and angularity, the mass specific energy based on surface surement of acoustic emissions of grain fracture. Local
energy provides a lower limit estimation. Luo et al. stress fields at contacts play a major role in fracture
(2011) estimated the mass specific energy of Eglin (quartz) strength of single grains. Some differences in fracture
sand to be in the range of 5–25 kJ/kg based on one- strength may be attributed to grain shape and features at
dimensional dynamic compression tests in Split Hopkins contacts. For the size group presented, Weibull distribution
Pressure Bar (SHPB). Based on median particle size in their is found to be acceptable for modeling size dependent
tests (d50 = 350 lm), the average mass specific energy from strength of the grains. Weibull modulus is affected by par-
Fig. 13 is 3 kJ/kg. The mass specific energy from grain ticle shape and decreases with increasing angularity. The
crushing alone will not account for the momentum of the computed Weibull moduli are consistent with the assump-
particles and the energy lost in friction between particles tion of critical surface flaws for all sands. With decreasing
and along the sides of the container in SHPB tests. Thus, grain size, the flaw size is a greater proportion of the grain
energy estimates from such tests may always provide a size in rounded and high Weibull modulus grains, the flaw
higher value than that from single grain crushing. size increases at a faster rate for angular sands with smaller
If coordination number is the dominant phenomenon Weibull modulus. Acoustic emissions reveal that angular
that controls comminution process, then the effect of smal- sands release lower energy at break and may require
ler grains surrounding a larger grain is such that it offsets, greater number of events leading up to the failure. The
in some cases, orders of magnitude of higher energy energy released at fracture increases with fracture stress.
required to crush smaller particles than larger particles. Low strength angular particles may at times release larger
This may not be possible until enough fines have been cre- energy due to multiple small asperities breaking leading
ated. For example, if an 800 lm grain is surrounded by up to the main fracture. The mass specific energy required
300 lm grains, the energy required to crush the smaller to cause fracture increases with decreasing particle size.
grains is almost two orders of magnitude greater than that This coupled with stress reduction in grains with coordina-
required to crush the larger grain. Therefore, initially the tion number may explain the fractal distribution of com-
size dependence of strength maybe the dominant process minuted remains.
with a shift to coordination number dominating the com-
minution as more fines are produced. A stage may then Acknowledgements
be reached at which for a given input energy no more
crushing is possible. Therefore, for a starting particle size This research was funded in its entirety through Defense
distribution and relative density it is possible to arrive at Threat Reduction Agency (DTRA) Grant # HDTRA1-12-
a unique particle size distribution (McDowell and Bolton, 1-0045. We would like to acknowledge the support of our
1998; Sammis et al., 1987). Program Managers Dr. Suhithi Peiris, Dr. Douglas A. Dal-
The power law relationship between mass specific energy ton (Allen), and Dr. Fidel Ruz-Nuglo. Also acknowledged
and particle size (MSE=M / d a ) is characterized by 4.2, are support of Dr. Andrew A. Wereszczak and Dr. Ben-
6.9, 4.2 and 2.4 for the exponent a for Ottawa sand, jamin L. Hackett for the use of Zwick hardness testing sys-
Q-Rok, Siligran and fused silica respectively. Though, tem, Dr, Colin D. Sumrall and Dr. Cameron Hughes for
there were significant differences in the strengths of Ottawa help with sand thin sections, and Dr. J. Michael Starbuck
sand and Siligran, the similar value for a could be a conse- and Dr. Srdjan Simunovic for the high-speed camera.
quence of similar flaw size and flaw size scaling with parti-
cle size. Q-Rok, on the other hand, is highly angular and References
with increasing grain size there is a greater possibility of
encountering polycrystalline grains and larger asperities, Allen, W.A., Mayfield, E.B., Morrison, H.L., 1957. Dynamics of a
which may have contributed to rapid decrease in mass projectile penetrating sand. J. Appl. Phys. 28, 370–376.

1309
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

Altuhafi, F.N., Coop, M.R., Georgiannou, V.N., 2016. Effect of particle McDowell, G.R., 2002. On the yielding and plastic compression of sand.
shape on the mechanical behavior of natural sands. J. Geotech. Soils Found. 42, 139–145.
Geoenviron. Eng. 142, 04016071. McDowell, G.R., 2001. Statistics of soil particle strength. Géotechnique
Bohse, J., 2000. Acoustic emission characteristics of micro-failure 51, 897–900.
processes in polymer blends and composites. Compos. Sci. Technol. McDowell, G.R., Amon, A., 2000. The application of Weibull statistics to
60, 1213–1226. the fracture of soil particles. Soils Found. 40, 133–141.
Brace, W.F., Walsh, J.B., 1962. Some direct measurements of the surface McDowell, G.R., Bolton, M.D., 1998. On the micromechanics of
energy of quartz and orthoclase. Am. Mineral. 47, 1111–1122. crushable aggregates. Géotechnique 48, 667–679.
Brzesowsky, R.H., Spiers, C.J., Peach, C.J., Hangx, S.J.T., 2014. Time- McDowell, G.R., Bolton, M.D., Robertson, D., 1996. The fractal
independent compaction behavior of quartz sands. J. Geophys. Res. crushing of granular materials. J. Mech. Phys. Solids 44, 2079–2101.
Solid Earth 119, 936–956. McDowell, G.R., De Bono, J.P., 2013. On the micro mechanics of one-
Brzesowsky, R.H., Spiers, C.J., Peach, C.J., Hangx, S.J.T., 2011. Failure dimensional normal compression. Géotechnique 63, 895–908.
behavior of single sand grains: theory versus experiment. J. Geophys. Muñoz-Ibáñez, A., Delgado-Martı́n, J., Grande-Garcı́a, E., 2019. Acous-
Res. Solid Earth 116, B06205. tic emission processes occurring during high-pressure sand com-
Cavarretta, I., O’Sullivan, C., 2012. The mechanics of rigid irregular paction. Geophys. Prospect. 67, 761–783.
particles subject to uniaxial compression. Géotechnique 62, 681– Nakata, A.F.L., Hyde, M., Hyodo, H., Murata, H., 1999. A probabilistic
692. approach to sand particle crushing in the triaxial test. Géotechnique
Coop, M.R., Sorensen, K.K., Bodas Freitas, T., Georgoutsos, G., 2004. 49, 567–583.
Particle breakage during shearing of a carbonate sand. Géotechnique Nakata, Y., Hyodo, M., Hyde, A.F.L., Kato, Y., Murata, H., 2001a.
54, 157–163. Microscopic particle crushing of sand subjected to high pressure one-
Dai, S.T., Labuz, J.F., 1997. Damage and failure analysis of brittle dimensional compression. Soils Found. 41, 69–82.
materials by acoustic emission. J. Mater. Civ. Eng. 9, 200–205. Nakata, Y., Kato, Y., Hyodo, M., Hyde, A.F.L., Murata, H., 2001b. One-
Davidge, R.W., 1979. Mechanical Behaviour of Ceramics. Cambridge dimensional compression behaviour of uniformly graded sand related
University Press. to single particle crushing strength. Soils Found. 41, 39–51.
De Rosa, I.M., Santulli, C., Sarasini, F., 2009. Acoustic emission for Ohtsu, M., Watanabe, H., 2001. Quantitative damage estimation of
monitoring the mechanical behaviour of natural fibre composites: a concrete by acoustic emission. Constr. Build. Mater. 15, 217–224.
literature review. Composites Part A, Special Issue: Repair 40, 1456– Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B.,
1469. Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V.,
Einav, I., 2007. Breakage mechanics—part I: theory. J. Mech. Phys. Solids Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M.,
55, 1274–1297. Duchesnay, E., 2011. Scikit-learn: machine learning in Python. J.
Ferguson, C.C., Lloyd, G.E., Knipe, R.J., 1987. Fracture mechanics and Mach. Learn. Res. 12, 2825–2830.
deformation processes in natural quartz: a combined Vickers inden- Sammis, C., King, G., Biegel, R., 1987. The kinematics of gouge
tation, SEM, and TEM study. Can. J. Earth Sci. 24, 544–555. deformation. Pure Appl. Geophys. 125, 777–812.
Forrestal, M.J., Luk, V.K., 1992. Penetration into soil targets. Int. J. Shipway, P.H., Hutchings, I.M., 1993a. Fracture of brittle spheres under
Impact Eng. 12, 427–444. compression and impact loading. II. Results for lead-glass and
Glößner, C., Moser, S., Külls, R., Heß, S., Nau, S., Salk, M., Penumadu, sapphire spheres. Philos. Mag. A 67, 1405–1421.
D., Petrinic, N., 2017. Instrumented projectile penetration testing of Shipway, P.H., Hutchings, I.M., 1993b. Fracture of brittle spheres under
granular materials. Exp. Mech. 57, 261–272. compression and impact loading. I. Elastic stress distributions. Philos.
Hiramatsu, Y., Oka, Y., 1966. Determination of the tensile strength of Mag. A 67, 1389–1404.
rock by a compression test of an irregular test piece. Int. J. Rock Takeda, S., Ogawa, K., Tanigaki, K., Horikawa, K., Kobayashi, H., 2018.
Mech. Min. Sci. 3, 89–90. DEM/FEM simulation for impact response of binary granular target
Ibraim, E., Luo, S., Diambra, A., 2017. Particle soil crushing: passive and projectile. Eur. Phys. J. Spec. Top. 227, 73–83.
detection and interpretation, in: ISSMGE. Presented at the 19th Tavares, L.M., King, R.P., 1998. Single-particle fracture under impact
International Conference on Soil Mechanics and Geotechnical Engi- loading. Int. J. Miner. Process. 54, 1–28.
neering, Republic of Korea, pp. 389–392. Tromans, D., Meech, J.A., 2004. Fracture toughness and surface energies
Kanda, Y., Sano, S., Saito, F., Yashima, S., 1985. Relationships between of covalent minerals: theoretical estimates. Miner. Eng. 17, 1–15.
particle size and fracture energy for single particle crushing. Kona Tsoungui, O., Vallet, D., Charmet, J.C., 1999. Numerical model of
Powder Part. J. 3, 26–31. crushing of grains inside two-dimensional granular materials. Powder
King, R.P., Bourgeois, F., 1993. Measurement of fracture energy during Technol. 105, 190–198.
single-particle fracture. Miner. Eng. 6, 353–367. Turner, A.K., Kim, F.H., Penumadu, D., Herbold, E.B., 2016. Meso-scale
Koerner, R.M., McCabe, W.M., Lord, A.E., 1981. Acoustic emission framework for modeling granular material using computed tomogra-
behavior and monitoring of soils. In: Drnevich, V.P., Gray, R.E. phy. Comput. Geotech. 76, 140–146.
(Eds.), Acoustic Emissions in Geotechnical Engineering Practice, Vidya Sagar, R., Raghu Prasad, B.K., 2011. An experimental study on
ASTM STP 750. ASTM International, Conshohocken, PA, pp. 93– acoustic emission energy as a quantitative measure of size independent
93–49. specific fracture energy of concrete beams. Constr. Build. Mater. 25,
Krumbein, W.C., Sloss, L.L., 1963. Stratigraphy and Sedimentation, 2349–2357.
second ed. W. H. Freeman. Wang, W., Coop, M.R., 2016. An investigation of breakage behaviour of
Kschinka, B.A., Perrella, S., Nguyen, H., Bradt, R.C., 1986. Strengths of single sand particles using a high-speed microscope camera. Géotech-
glass spheres in compression. J. Am. Ceram. Soc. 69, 467–472. nique 66, 984–998.
Li, J., Huang, Y., Chen, Z., Li, M., Qiao, M., Kizil, M., 2018. Particle- Weibull, W., 1951. A statistical distribution function of wide applicability.
crushing characteristics and acoustic-emission patterns of crushing J. Appl. Mech. 18, 293–297.
gangue backfilling material under cyclic loading. Minerals 8, 244. Yapar, O., Basu, P.K., Volgyesi, P., Ledeczi, A., 2015. Structural health
Luo, H., Lu, H., Cooper, W.L., Komanduri, R., 2011. Effect of mass monitoring of bridges with piezoelectric AE sensors. Eng. Fail. Anal.
density on the compressive behavior of dry sand under confinement at The Sixth International Conference on Engineering Failure Analysis
high strain rates. Exp. Mech. 51, 1499–1510. 56, 150–169.
Mao, W., Towhata, I., 2015. Monitoring of single-particle fragmentation Yashima, S., Kanda, Y., Sano, S., 1987. Relationships between particle
process under static loading using acoustic emission. Appl. Acoust. 94, size and fracture energy or impact velocity required to fracture as
39–45. estimated from single particle crushing. Powder Technol. 51, 277–282.

1310
 A. Sharma, D. Penumadu Soils and Foundations 60 (2020) 1299–1311

Zhang, J., Wong, T., Davis, D.M., 1990. Micromechanics of pressure- Zheng, J., Hryciw, R.D., 2016. Roundness and sphericity of soil particles
induced grain crushing in porous rocks. J. Geophys. Res. 95, 341–352. in assemblies by computational geometry. J. Comput. Civil Eng. 30,
Zhang, Y.D., Buscarnera, G., Einav, I., 2016. Grain size depen- 04016021.
dence of yielding in granular soils interpreted using fracture Zhong, W., Yue, F., Ciancio, A., 2018. Fractal behavior of particle size
mechanics, breakage mechanics and Weibull statistics. Géotech- distribution in the rare earth tailings crushing process under high stress
nique 66, 149–160. condition. Appl. Sci. 8, 1058.

1311