Powder Technology 364 (2020) 98–114

Contents lists available at ScienceDirect

Powder Technology

journal homepage: www.elsevier.com/locate/powtec

Investigation of powder flowability at low stresses: Influence of particle

size and size distribution
Alexandros Georgios Stavrou a, Colin Hare a,⁎, Ali Hassanpour b, Chuan-Yu Wu a
a
Department of Chemical and Process Engineering, Faculty of Engineering and Physical Sciences, University of Surrey, Guildford GU2 7XH, UK
b
School of Chemical and Process Engineering, Faculty of Engineering, University of Leeds, Leeds LS2 9JT, UK

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

Article history: At moderate stresses, shear cells are the preferred method of powder flow measurement. However, several in-
Received 28 September 2019 dustrial processes operate at low stresses, where the determination of unconfined yield strength by the shear
Received in revised form 30 December 2019 cell technique may be inconsistent, or found not to correlate with observed behaviour. Alternatively, ball inden-
Accepted 22 January 2020
tation can be used, which directly measures hardness; related to unconfined yield strength by the constraint fac-
Available online 23 January 2020
tor. However, it is not known how constraint factor is influenced by particle properties. Here, ball indentation and
Keywords:
shear cell methods are applied for glass beads of various size distributions, and the influence of particle size dis-
Powder flowability tribution on the constraint factor is explored. The constraint factor is shown to be independent of the pre-consol-
Low consolidation stresses idation stress, though reduces as the d10, d50 or d90 are increased. Unconfined yield strength inferred from
Ball indentation indentation measurements suggest that extrapolation of shear cell data to low stresses overestimates the uncon-
Shear cell fined yield strength.
Particle size © 2020 Elsevier B.V. All rights reserved.
Particle size distribution

1. Introduction becomes increasingly important as the particle size decreases, especially

for powders that are very fine e.g. b50 μm, since the interparticle forces
Numerous industries such as pharmaceuticals, food and fast-moving are significant in comparison with the weight of the particles [14].
consumer goods often handle materials in the form of powders. Reliable Though it should be noted that this threshold depends on other particle
and consistent prediction of the flow behaviour of powders can be very properties such as density, shape and roughness.
challenging, especially when the powders are cohesive. Cohesive mate- The effect of the particle size distribution within a mixture is more
rials can lead to the formation of stagnant regions or flow stoppages in complex. Lumay et al. [14] tested five flour powders, finding that as
process equipment, resulting in uncontrolled or erratic flow rates from the size distribution becomes narrower, and at the same time the d10 be-
industrial equipment, and potentially causing segregation problems comes larger, flowability improves. Abdullah and Geldart [15] examined
[1]. Powder flow is not an inherent material property, being dependent the packing of binary mixtures of coarse and fine particles in aerated
on material physical properties, process conditions and environmental and tapped states, with their findings being easily translatable to pow-
conditions [2]. der flow behaviour. In the case of aerated mixtures, the poured density
Particle size and its distribution is one of the most influential proper- initially increased with a reduction in fines content, eventually reaching
ties on powder flow. For a given powder, reducing particle size tends to a plateau. On the other hand, for the tapped system the tapped density
reduce flowability [3–9], because the particle surface area per unit mass initially increased with a reduction in fines content, and then decreased
increases as particle size decreases, providing a greater surface area for since insufficient small particles were available to fill the voids in be-
surface cohesive forces to interact, and therefore resulting in a more co- tween the larger particles. The Hausner ratio exhibited a continuous de-
hesive flow behaviour [10]. However, powders with similar size can ex- crease with a reduction in fines content, indicating an improvement in
hibit different flow behaviours due to differences in other properties flowability. Gold et al. [16] showed that when fine particles are added
such as particle morphology and surface roughness [11,12]. Larger par- to lactose granules, the flowrate of the mixture increases with an in-
ticles pack more efficiently due to the ease with which they flow past crease in the amount of fine particles until a given maximum flowrate
one another to fill voids in the bed, while as the particle size decreases is achieved. However, when this maximum is reached, any increase in
flowability deteriorates and particles pack more loosely [13]. Cohesion the amount of fine particles results in a decrease in the flowability of
the mixture. In addition, their finding shows that the quantity of fine
⁎ Corresponding author. particles required to reach the maximum flowrate for a given material
E-mail address: [email protected] (C. Hare). decreases with a reduction of the size of the fines. Liu et al. [17] showed

https://doi.org/10.1016/j.powtec.2020.01.068
0032-5910/© 2020 Elsevier B.V. All rights reserved.
 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114 99

Fig. 1. Indentation step [29].

that when the finest particles of a needle-shaped ibuprofen powder are stress, and subsequently allowing the unconfined yield strength to be
separated from the bulk, the fine powders flow better than the bulk estimated from the measured yield locus. Shear cells typically operate
powder. This is attributed to the narrower size distribution. under moderate to high stresses, and like the majority of powder testers
The flowability of a powder is generally improved by the inclusion of often fail to reliably assess powder flowability at low consolidation
larger particles and worsened by the inclusion of finer particles. This stresses (≤1 kPa). At such stresses shear cells are normally unable to
means that the influence of widening size distribution on the powder generate steady-state shear, or the reproducibility of the measurement
flowability can be difficult to predict [18]. Molerus and Nwylt [19] of unconfined yield strength is greatly reduced, or does not correlate
found that for binary mixtures of coarse and fine limestone particles, with observed process behaviour [23,24]. Flow behaviour at low
an increase in the fines content results in an increase in the unconfined stresses may be estimated from shear testing at higher stresses by as-
yield strength, eventually becoming equal to the strength of the fines suming linearity of the yield loci, and extrapolating towards zero nor-
alone once a fines content of 30% w/w is reached. At fines contents mal stress. However, this leads to an overestimation of unconfined
equal or greater than this it is expected that the coarse particles are yield strength and cohesion, since yield loci tend to deviate sharply
completely embedded by the fines, and so the flow behaviour is from the linear regression in the region of low stresses [23].
governed by the interparticle forces between the fines. In a binary mix- There are many industrial processes during which granular mate-
ture of coarse and fine particles, contacts between coarse particles dom- rials are subjected to low stresses, such as die filling and dosing of pow-
inate flow behaviour when the fines content is small, while contacts ders in capsules. Under such stresses, small contact areas exist between
between fine particles dominate when the fines content is large, with constituent particles, and very little particle deformation occurs, leading
the coarse-fine contacts not seen to dominate at any fines content [20]. to a low structural strength [25]. In order to address the need for reliable
Reliable prediction of powder flow based solely on particle proper- methods to measure flow resistance of weakly consolidated powders,
ties is not yet possible, due to the complexity of powder systems. How- ball indentation (BI) was introduced by Hassanpour and Ghadiri [26],
ever, a large number of techniques have been developed for evaluating with its operational window being thoroughly established experimen-
powder flowability, thus enabling the decoupling of the contribution of tally by Zafar et al. [27] and computationally by Pasha et al. [28]. In
particle size, among other parameters, on granular flow. None of these this technique a die, made of low friction material, is filled with powder
techniques are applicable across a full range of applied stresses and and pre-consolidated by uniaxial compression with a piston, which is
strain rates though, and therefore consideration needs to be given for then unloaded. Following this, a spherical indenter is driven into the
the measurement technique to be used for each circumstance. However, sample, whilst its penetration depth and the resulting vertical force
shear cells are the most well-established flowability measurement are measured until a desired depth is reached, and then the indenter
method, and are readily used for silo and hopper design [21,22]. Shear is unloaded (Fig. 1).
cells determine the onset of powder flow in a quasi-static manner, mea- From the force-displacement response of the powder bed, the hard-
suring the shear stress required to initiate flow under a given normal ness of the material is directly measured via Eq. (1), which corresponds

Fig. 2. Size distributions of glass bead single sieve cuts. Fig. 3. Size distributions of glass bead wider distribution mixtures.
100 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114

the penetration depth at maximum indentation load can be used in

place of hc [26].
Ball indentation can be applied as long as the powder compact has a
relatively flat surface, which is typically achieved at pre-consolidation
stresses as low as 100 Pa. It therefore offers the capability of obtaining
hardness measurements at any stress level above this. However, it is
commonly of interest to measure the unconfined yield strength, as mea-
sured by uniaxial compression tests, or determined in a shear cell. Tabor
[30] demonstrated for continuum materials that hardness is directly
linked to the unconfined yield strength, σc, via Eq. (3):

H ¼ C σc ð3Þ

where C is the constraint factor. The constraint factor represents the ad-
Fig. 4. Size distributions of glass bead mixtures with coarse particles.
ditional resistance caused by an elastically deforming region around the
plastically deforming indentation zone. This leads to an increase in the
local yield strength, represented by the hardness [31]. This has also
been observed in particulate systems [26,32]. In the case of continuum
solids, the constraint factor has been stated to have a value of 3 for
rigid-perfectly plastic materials [33], while according to Tabor [30] this
value is applicable only for ductile metals. Furthermore, for continuum
materials C is known to depend on material properties [34]. Johnson
[35] introduced a relationship between indentation hardness and
yield strength for elastic-perfectly plastic materials, based on Young's
modulus, radius of the impression, and the indenter radius. For particle
systems the constraint factor doesn't have a fixed value, with different
values determined for a variety of powders [26,32,36]. Currently the
constraint factor of a powder is not known a priori, nor is it known
which particle properties influence C, and to what extent. Shedding
light on all of the above is of particular importance because it will render
it possible for Eq. (3) to be utilised to infer unconfined yield strength
from ball indentation measurements at low stresses, which otherwise
cannot be easily determined [32].
Fig. 5. Size distributions of glass bead mixtures with fines. The aim of this study is to measure powder flowability using shear
cell testing and ball indentation in order to determine how powder
flowability at low stresses (≤ 1 kPa) differs from that at high stresses,
and to investigate the influence of particle size and size distribution
to the resistance of the bed to plastic deformation.
on powder flowability and the constraint factor. Furthermore, the reli-
ability of both techniques at low stress levels is evaluated.
H ¼ F max =A ð1Þ
2. Materials and methods
where Fmax is the maximum indentation load and A is the projected area
of the impression of the indenter, calculated from Eq. (2): In this study, glass beads supplied by Sigmund Lindner GmbH (Ger-
many) are tested as a model material, due to their high sphericity and
 
2 availability in a wide range of sizes. For each experimental series, a set
A ¼ π db hc −hc ð2Þ
of samples is prepared which vary by one parameter: median particle
size, width of size distribution, d10, or d90. Glass beads were sieved
where db is the indenter diameter and hc is the indent depth after using British Standard sieves to produce five consecutive single sieve
unloading. If unloading has negligible effect on the material's recovery, cuts of 45–53, 53–63, 63–75, 75–90 and 90–106 μm, for the study of

Table 1
Material characterisation overview.

Material d10 d50 d90 Span [(d90-d10)/d50] Sphericity AR (width/length)

a
45–53 μm glass beads 48.9 58.4 72.1 0.40 0.91 0.85
53–63 μm glass beadsa 58.8 67 80.5 0.32 0.92 0.88
63–75 μm glass beadsa 68.3 79.9 98 0.37 0.92 0.89
75–90 μm glass beadsa 84 92.9 112.1 0.30 0.93 0.91
90–106 μm glass beadsa 96.7 108 128.3 0.29 0.93 0.92
53–90 μm glass beadsa 65.1 81.2 103.5 0.47 0.92 0.89
45–106 μm glass beadsa 61.9 81.5 106.8 0.55 0.92 0.88
63–75 μm + 10% coarse glass beadsa 67 80.1 159.3 1.15 0.93 0.9
63–75 μm + 20% coarse glass beadsa 68.7 84.4 183.5 1.36 0.93 0.91
63–75 μm + 10% fine glass beadsb 41 70.1 99.7 0.84 – –
63–75 μm + 20% fine glass beadsb 9.8 67.5 97.5 1.30 – –
a
Size measurement by QICPIC (GRADIS).
b
Size measurement by Mastersizer.
 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114 101

Fig. 6. Yield locus automatically generated by the FT4 software at σpre = 8 kPa.

the influence of particle size on the flow behaviour. Furthermore, by mixtures of 90% w/w 63–75 μm with 10% w/w fines and 80% w/w 63–
mixing the above single sieve cuts, wider size distributions were created 75 μm with 20% w/w fines that were analysed by laser diffraction
to study the influence of the width of the size distribution on flowability. using the Mastersizer 2000 (Malvern Panalytical, UK), since the fines
A 53–90 μm mixture was created by mixing 50% w/w of the 63–75 μm were too small to be analysed using the QICPIC. The size distributions
single sieve cut and 25% w/w of each of the 53–63 and 75–90 μm single of the consecutive single sieve cuts, the wider size distribution mixtures,
sieve cuts, while a 45–106 μm mixture was created by mixing 40% w/w the mixtures with coarse particles and the mixtures with fines are pre-
of the 63–75 μm sieve cut, 20% w/w of each of the 53–63 and 75–90 μm sented in Figs. 2, 3, 4 and 5, respectively. Shape characterisation of the
sieve cuts, and 10% w/w of each of the 45–53 and 90–106 μm sieve cuts. samples was carried out with the QICPIC and is shown in Table 1,
Moreover, the median single sieve cut of 63–75 μm was mixed with along with the size data for all samples.
fractions of fine and coarse particles to investigate the influence of the All samples of glass beads were then silanised in order to make them
shift of d10 and d90, respectively, on powder flowability. In this case, cohesive. The commercially available Sigmacote® silane solution sup-
the coarse particles are from a 150–180 μm sieve cut and the fine parti- plied by Merck (Germany) was used for the surface treatment.
cles are from an as received ‘0–20 μm’ batch. Mixtures consisting of 90% Sigmacote® is 1,7-Dichloro-1,1,3,3,5,5,7,7-octamethyltetrasiloxane in
w/w 63–75 μm and 10% w/w coarse/fine particles and others having Heptane solution. A 50–75 g sample of powder (depending on particle
80% w/w 63–75 μm and 20% w/w coarse/fines were created. All mix- size) was submerged in Sigmacote® for 30 min and the excess solution
tures were created by mixing in a TURBULA T2C Shaker-Mixer at was removed by vacuum filtration for reuse. This step was repeated
49 rpm for 45 min. three times in total. Then, the solids were washed with de-ionised
Particle characterisation of all aforementioned samples was con- water in order to remove the hydrochloric acid by-product of the reac-
ducted by dynamic image analysis using the QICPIC (Sympatec, Ger- tion. The water was then removed and the coated samples were left in
many) system with the GRADIS dry dispersion mode, except for the the oven overnight to dry at 50 °C.

Fig. 7. Yield locus generated using the Warren Spring model at σpre = 8 kPa.
102 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114

than, the tangency point of the yield locus to the failure Mohr circle.
This approach is followed to minimise extrapolation of the yield locus
in order for the failure Mohr circle to be constructed, which would
lead to increased uncertainty when determining the unconfined yield
strength [23]. The desired range of normal stresses for shear to failure
is covered extensively by Schulze [23]. As a result of this approach, the
target normal stresses that were chosen to be applied in the shear
tests varied for each of the materials tested.
At each pre-shear normal stress the shear cell software takes the
measured shear stress at each normal stress to generate the yield
locus for this packing state. By default the FT4 software applies a linear
fit to the measured points, followed by application of Mohr circle anal-
ysis to allow the major principal stress, σ1, the unconfined yield
strength, σc, and subsequently the flow function coefficient, ffc, to be de-
termined for each pre-shear normal stress. However, it was found that
in many cases the measured yield locus was not tangent to the con-
structed failure Mohr circle, but cut through the circle, as shown in
Fig. 6, therefore the unconfined yield strength was overestimated. In
order to address this issue, the Warren Spring model [37] was employed
for the characterisation of the yield locus using a MATLAB code provided
by Dr. Massih Pasha (The Chemours Company, USA). A representative
example of the Warren Spring fit to the same experimental data as in
Fig. 6 is shown in Fig. 7. The pre-shear point was not considered for
the fitting, to avoid the extra curvature to the yield locus and a reduction
of the estimated major principal stress that its inclusion would cause.
For each material, three repeats are made at each pre-shear normal
stress, and the average results are reported with the error bars indicat-
ing the standard deviation of the measurement.
Following the determination of σ1 and σc at each pre-shear normal
stress, the hardness values for all samples were measured by ball inden-
tation at the major principal stresses derived from the FT4 shear cell
tests, to allow comparison with shear cell measurements and the con-
straint factor to be computed using the approach outlined in the intro-
duction. Additionally, indentation tests are conducted at low
consolidation stresses, namely 0.1, 0.2, 0.4, 0.6, 0.8 and 1 kPa. For the
ball indentation experiments, the criteria for sample, die and indenter
Fig. 8. Ball indentation setup (a: consolidation, b: penetration). dimensions established by Zafar et al. [27] are adhered to for this
work. A 20 mm diameter stainless steel die, which is attached to a
metal plate extending beyond the outer wall of the die, is filled by pass-
ing the powder through a sieve with an aperture approximately five
The flowability of all samples was measured both by shear cell test- times greater than d50. The sieve is placed directly above a funnel,
ing and ball indentation. First, in order to determine the unconfined above the die. The die height is 20 mm, with a bed height of 15–
yield strength of the materials, shear testing was carried out using the 20 mm generated in all cases, and the powder mass is weighed. The
cylindrical shear cell attachment of the FT4 Powder Rheometer (Free- die is placed below a stainless steel piston of 19.8 mm diameter attached
man Technology, UK). The principles of shear testing are covered exten- to an Instron 1175 mechanical testing machine (Instron, USA) by a 1 N
sively in literature, and can be found elsewhere [23]. For each sample, load cell, which has a resolution of 0.25 mN. Before each test is started,
tests at 2, 4, 6 and 8 kPa pre-shear normal stress (σpre) were carried the metal plate to which the die is attached is driven towards the piston
out. In addition to this, in the case of the 45–53 μm sample, additional while the force is recorded (with the die offset to prevent contact with
shear tests were performed at low pre-shear normal stresses of 0.06, the die walls) until contact is made, in order to determine the distance
0.1, 0.25, 0.5 and 1 kPa. A pre-shear normal stress of 0.06 kPa represents between the base of the die and the piston. After that, the plate is
the lowest possible stress in the FT4 that will allow five unique, lower returned to its starting position and the die is centred below the piston.
values of target applied normal stress to be given. In each case, the tar- At the start of the actual test, the die is driven upwards, towards the pis-
get normal stresses were chosen by trial and error, so that they are dis- ton, at a vertical speed of 1 mm/min, therefore testing in the quasi-static
tributed approximately equidistantly and the point of incipient flow regime, until the desired consolidation stress is reached. The final dis-
with the lowest normal stress is located close to, but at a higher stress placement of consolidation, zf, is recorded and used along with the

Table 2
FT4 shear test data for five consecutive single sieve cuts of silanised glass beads.

σpre 45–53 μm 53–63 μm 63–75 μm 75–90 μm 90–106 μm

σ1 σc ffc σ1 σc ffc σ1 σc ffc σ1 σc ffc σ1 σc ffc

2 3.2 2.0 1.6 3.2 1.8 1.8 3.1 1.5 2.1 3.1 1.5 2.1 3.0 1.3 2.4
4 5.7 2.2 2.6 5.8 2.0 2.9 5.7 1.8 3.1 5.7 1.6 3.7 5.7 1.5 3.8
6 8.2 2.3 3.5 8.3 2.1 3.9 8.3 1.9 4.3 8.4 1.8 4.6 8.3 1.7 5.0
8 10.8 2.6 4.2 10.9 2.4 4.6 10.9 2.1 5.2 10.9 1.9 5.9 10.8 1.7 6.3
 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114 103

ffc = 1 ffc = 2 ffc = 4

ffc = 10

Fig. 9. Unconfined yield strength as a function of major principal stress for five consecutive single sieve cuts of silanised glass beads.

distance between the base of the die and the piston at the starting point, penetration depth is reached, and the sample is then unloaded. The ball
zo, to determine the bed height, and consequently determine the pack- indentation setup is shown in Fig. 8.
ing fraction, χ, using Eq. (4): The bed hardness is calculated using Eq. (1), with the projected area
of the impression of the indenter determined using Eq. (2). Hardness is
typically overestimated at shallow depths, whilst at large penetration
4M depths further consolidation may occur, which also leads to an overes-

2
ρb M=V π z0 −z f Dd timation of hardness. It is necessary for the measured hardness to be in-
χ¼ ¼ ¼ ð4Þ
ρt ρt ρt dependent of the penetration depth in order to represent plastic yield
stress [32]. The range of penetration depths that provide a stable hard-
ness measurement is therefore determined. The dimensionless penetra-
where ρb and ρt are the bulk and true densities, respectively, M and V are tion depth, hd, is determined using Eq. (5), with values in the range of
the mass and volume of the powder, respectively, and Dd is the die 0.1–0.7 being applied for each powder at consolidation stresses of 0.1
diameter. and 1 kPa.
The sample is then unloaded at the same velocity, and the piston is
replaced by a 4 mm diameter, spherical, stainless steel indenter aligned hd ¼ 2hc =db ð5Þ
centrally above the powder bed. The die is then driven upwards, to-
wards the indenter, at the same speed as the consolidation step, until As such, a dimensionless penetration depth determined to be in the
contact is detected, which is considered to be when a force of 3 mN is stable hardness range is then applied in all experiments for the remain-
registered. Following that, the penetration is continued until the desired ing consolidation stresses for a given powder. The ball indentation

Fig. 10. Hardness as a function of dimensionless penetration depth for the 63–75 μm single sieve cut of silanised glass beads.
104 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114

Fig. 11. Force as a function of dimensionless penetration depth during indentation of a 63–75 μm glass beads bed at 10.9 kPa.

technique is applied at the major principal stresses determined in the Table 2. For all sizes, the unconfined yield strength is found to increase
shear cell experiments, where it is assumed that the normal stress in approximately linearly with major principal stress, whilst decreasing
the indentation process is equal to the major principal stress. The con- with increasing particle size, as shown in Fig. 9. For a given pre-shear
straint factor is then determined at these major principal stresses. In ad- normal stress, there is a clear increase in flowability with increasing par-
dition, the ball indentation method is applied at low consolidation ticle size, as evidenced from the ffc values in Table 2. In addition to this,
stresses of 0.1, 0.2, 0.4, 0.6, 0.8 and 1 kPa. The unconfined yield strength the flow function coefficient increases with pre-shear normal stress, and
is then inferred at these stresses using Eq. (3) and the established con- ranges from 1.6–4.2 for the 45–53 μm glass beads to 2.4–6.3 for the 90–
straint factor for the powder. For each material, five repeats are made 106 μm glass beads.
at each pre-shear normal stress, and the average results are reported Hardness measurements at dimensionless penetration depths of 0.1,
with the error bars indicating the standard deviation of the measure- 0.3, 0.5 and 0.7 were performed on separate powder beds of 63–75 μm
ment. For all the experiments carried out here the temperature was glass beads consolidated to 0.1 and 1 kPa, as shown in Fig. 10. The di-
20–25 °C and the relative humidity (RH) was 30–65%. mensionless penetration depths are calculated using Eq. (5). It can be
seen that for both tested consolidation stresses hardness is
3. Results and discussion overestimated at a dimensionless penetration depth of 0.1. This phe-
nomenon is observed for all materials at shallow indentation depths.
3.1. Effect of particle size on constraint factor and flow behaviour However, beyond this point hardness reaches a plateau, becoming con-
stant for dimensionless penetration depths greater than 0.3. These find-
The measurements of unconfined yield strength at the correspond- ings are in agreement with the DEM simulations of Pasha et al. [28] and
ing major principal stresses determined for five consecutive single the experiments of Zafar et al. [27]. Here the hardness was determined
sieve cuts of glass beads silanised by Sigmacote® are shown in by considering the penetration depth at maximum indentation load

Fig. 12. Force against displacement during indentation of a 63–75 μm glass beads bed.
 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114 105

Fig. 13. Hardness as a function of major principal stress for five consecutive single sieve cuts of silanised glass beads.

rather than the elastically-recovered depth. Fig. 11 shows the force-dis- result, the use of the penetration depth at maximum indentation load
placement profile for indentation up to a dimensionless penetration in hardness calculations is justified for this material, it was therefore
depth of 0.5 for a bed of 63–75 μm glass beads consolidated to considered for all the hardness calculations. Furthermore, the behaviour
10.9 kPa. The hardness calculated at the highest major principal stress of hardness as a function of penetration depth is independent of the ap-
of 10.9 kPa for the 63–75 μm sample, by considering the projected plied stress, therefore the obtained trends at higher stresses are ex-
area of the impression of the indenter after unloading, was found to pected to be qualitatively the same, with the same depth range
be almost identical (b 0.2% difference) to the value computed by using providing valid measurements. This behaviour is consistent for all
the penetration depth at maximum indentation load to determine hard- other sizes of glass beads investigated in this work. With the reliable
ness, i.e. by ignoring unloading. This is expected given the almost verti- range of hardness measurements now established, a dimensionless
cal slope of the unloading curve. If the unloading effect is ignored, then penetration depth of 0.5 is applied for all following ball indentation
hardness can be estimated at any penetration depth up to the depth measurements on silanised glass beads.
tested. In Fig. 10 this is referred to as ‘continuous hardness’, and is calcu- Although a 3 mN force is taken to indicate contact between the in-
lated from indentation tests at a dimensionless penetration depth of 0.7. denter and the powder bed, the indenter penetrates the specimen
This estimate is shown to be almost identical to the direct measurement slightly before this target force is reached. This can be seen from the
made at lower depths for both stresses, since the unloading is negligible slope of the force-displacement curve during the contact detection
for this material. The effect of unloading on hardness was also investi- step in Fig. 12. The penetration depth is controlled from the point that
gated at the highest major principal stresses for all other samples of the 3 mN force is detected, however the actual penetration depth is de-
glass beads tested in this work, and was found to be negligible. As a termined based on the point where contact is made. Consequently, even

Fig. 14. Packing fraction against major principal stress for five consecutive single sieve cuts of silanised glass beads.
106 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114

Fig. 15. Hardness as a function of packing fraction for five consecutive single sieve cuts of silanised glass beads.

though the target dimensionless penetration depth for all hardness decreasing with an increase in particle size. At low stresses (≤1 kPa)
measurements was 0.5, the true dimensionless penetration depth was the increase of hardness with stress is observed to be much steeper
larger and varied between tests. However, this does not affect the valid- than at high stresses, a phenomenon also observed by Zafar [32]. Fur-
ity of most of the hardness measurements (particularly at moderate to thermore, in the case of weakly consolidated powder beds, hardness
high stresses), since the true dimensionless penetration depth of 0.7 values are not distinctively different among the different particle sizes.
was not exceeded, and the measurements have already been shown In this range the error bars are somewhat larger, due to the difficulty
to be independent of penetration depth for dimensionless penetration of reproducing a uniformly flat powder bed surface, as well as creating
depths of 0.3–0.7. In some repeats at very low stresses though, the di- a consistent packing structure. As the applied consolidation stress in-
mensionless penetration depth of 0.7 was exceeded, and hardness was creases, the influence of bed surface asperities becomes less important.
overestimated, however in most cases this overestimation was within At all stress levels the standard deviation of five measurements (indi-
test error. cated by the error bars) is low; with the coefficient of variation being
The ball indentation method is applied for all five consecutive single less than 10% in most cases.
sieve cuts of silanised glass beads using the average major principal In order to investigate the cause of the discrepancy of the hardness
stress determined from the three shear cell tests for each sample at increase against stress between high and low stresses, the packing frac-
each pre-shear normal stress, with the results presented in Fig. 13. At tion for all five sieve cuts is calculated using Eq. (4), and is shown against
high stresses (N1 kPa) the hardness values of the five sieve cuts are dis- major principal stress in Fig. 14.
tinctively different, following the same trend as the shear cell results; As can be seen from Fig. 14, all glass bead samples exhibit a dramatic
increasing approximately linearly with major principal stress, and increase of packing fraction with the increase of consolidation stress in

Fig. 16. Constraint factor as a function of major principal stress for five consecutive single sieve cuts of silanised glass beads.
 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114 107

ffc = 1 ffc = 2 ffc = 4

ff c = 10

Fig. 17. Unconfined yield strength shear cell measurements and inferred values from ball indentation for five consecutive single sieve cuts of silanised glass beads.

the low stress region. A small increase of the applied stress leads to a low stresses. This assumption is validated by DEM simulations of
much more compacted powder bed, which in turn provides a great in- Stavrou et al. [29], which show C to remain constant down to the lowest
crease of resistance to plastic deformation. On the other hand, in the investigated stress of 0.1 kPa. For all samples, the average constraint fac-
high stress region, the packing state of the powder beds does not change tor across all major principal stresses is used along with the ball inden-
considerably with applied stress. The aforementioned behaviour leads tation measurements at low consolidation stresses to determine the
to an approximately linear increase of hardness with packing fraction unconfined yield strength at such stresses via Eq. (3). Fig. 17 shows
in the range 0.45–0.55, as shown in Fig. 15. Generally, hardness is the inferred unconfined yield strength values from the ball indentation
greater for smaller particles at a given packing fraction, although this be- measurements at major principal stresses of 0.1, 0.2, 0.4, 0.6, 0.8 and
haviour is clear only at high stresses. At low stresses the error bars of 1.0 kPa, along with the unconfined yield strength measurements carried
packing fraction are significant, and the difference in hardness among out in the shear cell at pre-shear normal stresses of 2, 4, 6 and 8 kPa
the different particle sizes is not clear. shown in Fig. 9. The indentation technique suggests a significant reduc-
The constraint factor is determined at each stress level for all sieve tion in unconfined yield strength at lower consolidation levels in com-
cuts of silanised glass beads using Eq. (3) and the measurements of un- parison to values that would be linearly extrapolated from the shear
confined yield strength and hardness, and presented against major tests.
principal stress in Fig. 16. The constraint factor is shown to be approxi- Measurement of unconfined yield strength at low stresses is often
mately constant for a given sieve cut across all tested major principal not reliable, or even possible, using a shear cell, however the shear cell
stresses. Moreover, C is found to generally decrease with an increase is more likely to achieve steady-state failure, and therefore generate a
in particle size. These findings agree with the work of Zafar [32]. result, for more cohesive powders. Since the 45–53 μm are the most co-
Since the constraint factor was found to be virtually independent of hesive glass beads used here, shear cell measurements are made at pre-
the major principal stress applied, it is assumed to remain constant at shear normal stresses of 0.06, 0.1, 0.25, 0.5 and 1 kPa for this sample, and

ffc = 1 ffc = 2 ffc = 4

ff c = 10

Fig. 18. Unconfined yield strength at low stresses determined from the shear cell and ball indentation for the 45–53 μm single sieve cut of silanised glass beads.
108 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114

Fig. 19. Measured shear and applied normal stresses for 45–53 μm silanised glass beads at pre-shear normal stresses of a) 0.06 kPa, b) 0.25 kPa, c) 1.00 kPa, d) 4.00 kPa and e) 8.00 kPa.

are shown compared to the indentation measurements in Fig. 18. The

Table 3
trends of unconfined yield strength against major principal stress are re-
FT4 shear test data for three samples of silanised glass beads with varying width of size
markably similar in this low stress range for both techniques. distribution.
In order to further investigate the shear cell measurements, the mea-
σpre 63–75 μm 53–90 μm 45–106 μm
sured shear stresses and applied normal stresses of three repeats of the
FT4 shear cell measurements at the pre-shear normal stresses of 0.06, σ1 σc ffc σ1 σc ffc σ1 σc ffc
0.25, 1, 4, and 8 kPa, are shown in Fig. 19. At pre-shear normal stresses 2 3.1 1.5 2.1 3.2 1.6 1.9 3.2 1.8 1.8
of 0.06 and 0.25 kPa the data show great variation among repeats, with 4 5.7 1.8 3.1 5.8 1.9 3.1 5.8 2.0 2.9
the generated yield loci not consistently showing a monotonic increase 6 8.3 1.9 4.3 8.4 2.1 4.1 8.4 2.2 3.8
8 10.9 2.1 5.2 10.9 2.2 5.0 11.0 2.4 4.6
in shear stress with normal stress. In addition to this, in both cases there
 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114 109

ffc = 4
ffc = 1 ffc = 2

ff c = 10

Fig. 20. Unconfined yield strength as a function of major principal stress for three samples of silanised glass beads with varying width of size distribution.

is notable discrepancy between the target normal stresses and the ac- the pre-shear normal stress is increased. This highlights the need for de-
tual applied stresses. The nature of the FT4 shear cell protocol, which tailed analysis of shear cell data to assess the validity of the measured
leads to shearing from the highest to the lowest chosen target stress, yield locus, particularly at low pre-shear normal stresses.
can lead to data points for which the shear stress exceeds the pre-
shear stress when the applied stress is greater than the target stress, 3.2. Flowability of silanised glass bead mixtures of varying particle size
which invalidates the measurement. This phenomenon is observed in distribution
the case of tests at 0.06 kPa, where stresses beyond 0.06 kPa have
been applied during the shear test, and leads to an overestimation of 3.2.1. Effect of width of particle size distribution on constraint factor and
the unconfined yield strength, however this is not observed at higher flow behaviour
stresses. At a pre-shear normal stress of 1 kPa the generated test data The FT4 shear cell measurements of unconfined yield strength at the
are highly reproducible, although the applied stresses are still not equi- corresponding major principal stresses are shown in Table 3 for the me-
distant, and deviate notably from the target normal stresses, but less so dium 63–75 μm single sieve cut, along with two mixtures of 53–90 μm
than at lower pre-shear normal stresses. At a pre-shear normal stress of and 45–106 μm silanised glass beads. The two mixtures have essentially
4 kPa, the shear tests are highly reproducible and the achieved stresses the same d50 as the single sieve cut, but wider size distributions, as re-
are equidistant, though deviate slightly from the target normal stresses. ported in Table 1. As the size distribution is widened, the d10 and d90 re-
At 8 kPa, not only are the tests highly reproducible, but the target duce and increase, respectively, by about 3–4 μm with each additional
stresses have been virtually achieved. Therefore, a general trend of in- sieve cut. The unconfined yield strength against major principal stress
creasing reliability and reproducibility of shear testing is observed as is shown in Fig. 20. Unconfined yield strength increases approximately

Fig. 21. Hardness as a function of major principal stress for three samples of silanised glass beads with varying width of size distribution.
110 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114

Fig. 22. Packing fraction as a function of major principal stress for three samples of silanised glass beads with varying width of size distribution.

Fig. 23. Constraint factor as a function of major principal stress for three samples of silanised glass beads with varying width of size distribution.

ffc = 1 ffc = 2 ffc = 4

ffc = 10

Fig. 24. Unconfined yield strength shear cell measurements and inferred values from ball indentation for three samples of silanised glass beads with varying width of size distribution.
 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114 111

Table 4
FT4 shear test data for 63–75 μm silanised glass beads mixed with varying amounts of coarse and fine particles, along with the 63–75 μm single sieve cut.

σpre 63–75 μm 63–75 μm 63–75 μm 63–75 μm 63–75 μm

+ 20% coarse + 10% coarse + 10% fines + 20% fines

σ1 σc ffc σ1 σc ffc σ1 σc ffc σ1 σc ffc σ1 σc ffc

2 3.2 1.6 2.0 1.7 1.8 1.9 3.1 1.5 2.1 3.3 2.0 1.7 3.5 2.5 1.4
4 5.8 1.6 3.8 5.9 1.9 3.1 5.7 1.8 3.1 6.2 2.4 2.6 6.5 3.5 1.8
6 8.4 2.0 4.2 8.5 2.1 4.0 8.3 1.9 4.3 8.9 2.4 3.8 9.4 4.4 2.2
8 10.9 2.1 5.1 11.0 2.2 5.0 10.9 2.1 5.2 11.6 2.9 4.0 12.2 4.7 2.6

ffc = 1 ffc = 2

ffc = 4

ffc = 10

Fig. 25. Unconfined yield strength as a function of major principal stress for 63–75 μm silanised glass beads mixed with varying amounts of coarse and fine particles, along with the 63–75
μm single sieve cut.

Fig. 26. Hardness as a function of major principal stress for 63–75 μm silanised glass beads mixed with varying amounts of coarse and fine particles, along with the 63–75 μm single sieve cut.

linearly with major principal stress, as with the single sieve cuts in Fig. 21 shows the hardness measurements made at all major princi-
Section 3.1. It can be seen from both Table 3 and Fig. 20 that at a given pal stresses for all three size distributions of silanised glass beads. As
pre-shear normal stress there is a slight increase in unconfined yield with the unconfined yield strength from the shear tests above, hardness
strength as the size distribution is widened. Moreover, ffc is found to in- is observed to marginally increase when widening the size distribution
crease with the pre-shear normal stress applied and decrease as the size at higher stresses, whereas no notable difference can be seen at low
distribution is widened. stresses. As with Fig. 13, indentation tests indicate a more rapid increase
112 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114

Fig. 27. Packing fraction as a function of major principal stress for 63–75 μm silanised glass beads mixed with varying amounts of coarse and fine particles, along with the 63–75 μm single
sieve cut.

of hardness with major principal stress for weakly consolidated powder 3.2.2. Effect of d10 and d90 on constraint factor and flow behaviour
beds, which is explained by the packing fraction trend presented in The FT4 shear testing data of 63–75 μm silanised glass beads mixed
Fig. 22. with 10% and 20% w/w coarse and fine particles are reported in Table
The constraint factor determined from the ball indentation and 4, along with the data for the 63–75 μm single sieve cut for comparative
shear cell measurements for all size distributions is shown in Fig. 23. purposes. The unconfined yield strength against major principal stress is
Once again, constraint factor is found to remain constant throughout shown in Fig. 25. The addition of coarse particles is found to have negli-
the range of consolidation stresses applied. In addition to this, a slight gible effect on the unconfined yield strength regardless of the quantity
reduction in constraint factor is observed with an increase in the span added, whilst the addition of fines substantially increases the uncon-
of the sample, however this effect may not be statistically significant. fined yield strength, with a further increase observed as the quantity
Fig. 24 shows the unconfined yield strength values inferred from the added increases. This is in agreement with the finding of Molerus and
ball indentation method at low stresses, along with the measurements Nwylt [19] that unconfined yield strength increases with fines content
made in the shear cell at higher stresses. As in the case of the consecu- up to 30% w/w, beyond which it becomes equal to the strength of the
tive single sieve cuts, the increase of the unconfined yield strength is es- fines alone. The addition of 20 and 10% w/w coarse particles can be
timated to be sharper with increasing major principal stress at lower viewed as 80 and 90% w/w 63–75 μm added, respectively, to the 150–
stresses. At lower stresses the inferred values of unconfined yield 180 μm sample, and so the yield strength of the mixtures is essentially
strength are not distinctively different among the samples. equal to the yield strength of the 63–75 μm sample in this case. Adding

Fig. 28. Constraint factor as a function of major principal stress for 63–75 μm silanised glass beads mixed with varying amounts of coarse and fine particles, along with the 63–75 μm single
sieve cut.
 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114 113

Fig. 29. Average constraint factor as a function of d10.

10% w/w coarse particles leads in most cases to a slight reduction in ffc, the case of the mixtures with coarse particles, the large error bars
with an increase in coarse content to 20% w/w coarse particles causing mean that firm conclusions cannot be drawn.
the flow function coefficient to slightly increase again. The flow function The constraint factor values for the same samples with added coarse
coefficient decreases with the addition of fines, as shown in Table 4. particles or fines are plotted in Fig. 28, and are shown to be relatively con-
Fig. 26 shows the hardness against major principal stress for the stant regardless of stress. Moreover, as a result of the trends observed in
same mixtures of glass beads. A reduction of d10 (addition of fines) Figs. 25 and 26, C is found to decrease as the quantity of coarse particles
leads to an increase in hardness, as in the case of unconfined yield in the mixture is increased, while it increases as the quantity of fines in
strength. In contrast to the shear tests though, an increase in the the mixture increases. Regarding the glass bead mixtures studied in
d90 (addition of coarse particles) leads to a reduction in hardness. Section 3.2.1, as size distribution is widened the d10 is reduced and d90 is in-
Though the addition of coarse particles has a less significant influ- creased by similar amounts, so the two competing effects seen in Fig. 28
ence on the hardness of the mixture than the addition of fine parti- cancel each other out, hence leading to only slight differences in the con-
cles. The packing fraction data, shown in Fig. 27, partially explain straint factor between the narrow and wide size distributions (see Fig. 23).
the increased resistance to plastic deformation with the addition In order to clearly illustrate the effect of the addition of coarse and
of fines. However, when the quantity of fines added increases the fine particles on the constraint factor, the average constraint factor
packing fraction is not greatly increased, whereas hardness is signif- values are shown for samples with added fines against d10 in Fig. 29,
icantly affected. The mixture that has 20% w/w fines has more con- and with added coarse particles against d90 in Fig. 30. It can be seen
tacts between fine particles, hence being more resistant to flow. In that C decreases as either d10 or d90 are increased.

Fig. 30. Average constraint factor as a function of d90.

114 A.G. Stavrou et al. / Powder Technology 364 (2020) 98–114

4. Conclusions [6] I.M. Wouters, D. Geldart, Characterising semi-cohesive powders using angle of re-
pose, Part. Part. Syst. Charact. 13 (4) (1996) 254–259.
[7] D. Geldart, E.C. Abdullah, A. Hassanpour, L.C. Nwoke, I. Wouters, Characterization of
The ball indentation technique was employed along with shear test- powder flowability using measurement of angle of repose, China Particuology 4 (3–
ing on a wide variety of glass bead samples in order to investigate the dif- 4) (2006) 104–107.
[8] H. Hou, C.C. Sun, Quantifying effects of particulate properties on powder flow prop-
ference between powder flow behaviour at low and high stresses, and the erties using a ring shear tester, J. Pharm. Sci. 97 (9) (2008) 4030–4039.
influence of a number of size parameters (median particle size, width of [9] M. Krantz, H. Zhang, J. Zhu, Characterization of powder flow: static and dynamic
size distribution, d10 and d90) on the constraint factor and flowability. testing, Powder Technol. 194 (3) (2009) 239–245.
[10] J.J. Fitzpatrick, S.A. Barringer, T. Iqbal, Flow property measurement of food powders
Both unconfined yield strength and hardness were found to increase and sensitivity of Jenike's hopper design methodology to the measured values, J.
with an increase in the major principal stress applied, due to an increased Food Eng. 61 (3) (2004) 399–405.
packing fraction and interparticle contact area. At low stresses the in- [11] X. Fu, D. Huck, L. Makein, B. Armstrong, U. Willen, T. Freeman, Effect of particle
shape and size on flow properties of lactose powders, Particuology 10 (2) (2012)
crease in packing fraction with stress was more pronounced than at
203–208.
higher stresses, so were the hardness and unconfined yield strength [12] C. Hare, M. Ghadiri, The influence of aspect ratio and roughness on flowability, AIP
measurements. Hardness was shown to be independent of penetration Conf. Proc. 1542 (1) (2013, June) 887–890.
depth for dimensionless penetration depths between 0.3 and 0.7. The [13] J.M. Valverde, A. Castellanos, P.K. Watson, The effect of particle size on interparticle
adhesive forces for small loads, Powder Technol. 118 (3) (2001) 236–241.
constraint factor determined from indentation and shear cell tests was [14] G. Lumay, F. Boschini, K. Traina, S. Bontempi, J.C. Remy, R. Cloots, N. Vandewalle,
virtually independent of the stress applied for all samples. As a result, Measuring the flowing properties of powders and grains, Powder Technol. 224
the inferred unconfined yield strength from ball indentation at low (2012) 19–27.
[15] E.C. Abdullah, D. Geldart, The use of bulk density measurements as flowability indi-
stresses followed a similarly steep trend as hardness. This sharp change cators, Powder Technol. 102 (2) (1999) 151–165.
in behaviour at low stresses suggests that an extrapolation of shear cell [16] G. Gold, R.N. Duvall, B.T. Palermo, J.G. Slater, Powder flow studies III. Factors affect-
results from higher stresses would overestimate the yield strength. ing the flow of lactose granules, J. Pharm. Sci. 57 (4) (1968) 667–671.
[17] L.X. Liu, I. Marziano, A.C. Bentham, J.D. Litster, E.T. White, T. Howes, Effect of particle
An increase of median particle size led to an increase in powder properties on the flowability of ibuprofen powders, Int. J. Pharm. 362 (1–2) (2008)
flowability and decrease of the constraint factor. In addition to this, wid- 109–117.
ening the size distribution, while maintaining the same median size, re- [18] A. de Ryck, C. Hare, Chapter 2. Flow related properties of bulk particulate systems,
in: C. Hare, A. Hassanpour, M. Pasha (Eds.), Powder Flow: Theory Measurement
sulted in a slight decrease of both flowability and constraint factor. The and Application, Royal Society of Chemistry, Cambridge, UK 2019, pp. 4–38.
addition of fines caused a great decrease of powder flowability and an [19] O. Molerus, M. Nywlt, The influence of the fine particle content of the flow behav-
increase of constraint factor, while the addition of coarse particles ap- iour of bulk materials, Powder Technol. 37 (1) (1984) 145–154.
[20] T. Kojima, J.A. Elliott, Incipient flow properties of two-component fine powder sys-
peared to only decrease the material's resistance to plastic deformation,
tems and their relationships with bulk density and particle contacts, Powder
with the unconfined yield strength being unchanged. As a result, the in- Technol. 228 (2012) 359–370.
crease of coarse particle content led to a decrease of constraint factor. [21] A.W. Jenike, Gravity Flow of Bulk Solids, Bulletin No. 108, University of Utah, USA,
Overall, ball indentation shows good reproducibility down to consol- 1961.
[22] A.W. Jenike, Storage and Flow of Solids, Bulletin No. 123, University of Utah, USA,
idation stresses of 0.1 kPa. Whereas shear cell measurements in this low 1964.
stress region produce inconsistent results for this material. [23] D. Schulze, Powders and bulk solids, Behavior, Characterization, Storage and Flow,
Springer, NY, USA, 2008.
[24] S.V. Søgaard, T. Pedersen, M. Allesø, J. Garnaes, J. Rantanen, Evaluation of ring shear
Declaration of Competing Interest testing as a characterization method for powder flow in small-scale powder pro-
cessing equipment, Int. J. Pharm. 475 (1–2) (2014) 315–323.
[25] N. Harnby, A.E. Hawkins, D. Vandame, The use of bulk density determination as a
The authors declare that they have no known competing financial means of typifying the flow characteristics of loosely compacted powders under
interests or personal relationships that could have appeared to influ- conditions of variable relative humidity, Chem. Eng. Sci. 42 (4) (1987) 879–888.
ence the work reported in this paper. [26] A. Hassanpour, M. Ghadiri, Characterisation of flowability of loosely compacted co-
hesive powders by indentation, Part. Part. Syst. Charact. 24 (2) (2007) 117–123.
[27] U. Zafar, C. Hare, A. Hassanpour, M. Ghadiri, Ball indentation on powder beds for
Acknowledgements assessing powder flowability: analysis of operation window, Powder Technol. 310
(2017) 300–306.
[28] M. Pasha, C. Hare, A. Hassanpour, M. Ghadiri, Analysis of ball indentation on cohe-
The financial support of the International Fine Particle Research Insti- sive powder beds using distinct element modelling, Powder Technol. 233 (2013)
tute (IFPRI) is gratefully acknowledged. Furthermore, the authors greatly 80–90.
appreciate the MATLAB code provided by Dr. Massih Pasha (The [29] A.G. Stavrou, C. Hare, A. Hassanpour, C.Y. Wu, Investigation of powder flowability at
low stresses by DEM modelling, Chem. Eng. Sci. 211 (2020) 115307.
Chemours Company) for the Warren Spring fitting of the shear cell data.
[30] D. Tabor, The Hardness of Metals, Oxford, UK, Oxford University Press (Clarendon
The authors are also thankful to Mojtaba Ghadiri and his research group Press), 1951.
at the University of Leeds for the opportunity to conduct particle size [31] G.V. Kozlov, V.D. Serdyuk, V.A. Beloshenko, The plastic constraint factor and me-
measurements using the Mastersizer 2000 laser diffraction instrument. chanical properties of a high-density polyethylene on impact loading, Mech.
Compos. Mater. 30 (5) (1995) 506–509.
[32] U. Zafar, Assessing Flowability of Cohesive Powders by Ball Indentation, PhD Thesis
References University of Leeds, UK, 2013.
[33] R. Hill, The Mathematical Theory of Plasticity, Oxford, UK, Oxford University Press
[1] K. Johanson, Effect of particle shape on unconfined yield strength, Powder Technol. (Clarendon Press), 1950.
194 (3) (2009) 246–251. [34] D. Tabor, Indentation hardness: fifty years on a personal view, Philos. Mag. A 74 (5)
[2] J.K. Prescott, R.A. Barnum, On powder flowability, Pharm. Technol. 24 (10) (2000) (1996) 1207–1212.
60–85. [35] K.L. Johnson, Contact Mechanics, Cambridge University Press, Cambridge, UK, 1985.
[3] R.L. Carr, Classifying flow properties of solids, Chem. Eng. 1 (1965) 69–72. [36] C. Wang, A. Hassanpour, M. Ghadiri, Characterisation of flowability of cohesive pow-
[4] H.H. Hausner, Friction Conditions in a Mass of Metal Powder, Polytechnic Inst. of ders by testing small quantities of weak compacts, Particuology 6 (4) (2008)
Brooklyn. Univ. of California, Los Angeles, 1967. 282–285.
[5] R. Farley, F.H.H. Valentin, Effect of particle size upon the strength of powders, Pow- [37] M.D. Ashton, D.H. Cheng, R. Farley, F.H.H. Valentin, Some investigations into the
der Technol. 1 (6) (1968) 344–354. strength and flow properties of powders, Rheol. Acta 4 (3) (1965) 206–218.