You are on page 1of 13

http://www.diva-portal.

org

This is the published version of a paper published in Powder Technology.

Citation for the original published paper (version of record):

Marchetti, L., Hulme-Smith, C. (2021)


Flowability of steel and tool steel powders: A comparison between testing methods
Powder Technology, 384: 402-413
https://doi.org/10.1016/j.powtec.2021.01.074

Access to the published version may require subscription.

N.B. When citing this work, cite the original published paper.

Permanent link to this version:


http://urn.kb.se/resolve?urn=urn:nbn:se:kth:diva-293861
Powder Technology 384 (2021) 402–413

Contents lists available at ScienceDirect

Powder Technology

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

Flowability of steel and tool steel powders: A comparison between


testing methods
Lorenzo Marchetti ⁎, Christopher Hulme-Smith
Department of Material Science and Engineering, KTH Royal Institute of Technology, Brinellvägen 23, 11324 Stockholm, Sweden

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

Article history: The flow behaviour of a powder is critical to its performance in many industrial applications and manufacturing
Received 24 September 2020 processes. Operations such as powder transfer, die filling and powder spreading all rely on powder flowability.
Received in revised form 29 January 2021 Multiple testing methods can help in assessing flowability, but it is not always clear which may better represent
Accepted 30 January 2021
specific flow conditions or how different metrics correlate. This study compares 8 different flowability testing
Available online 16 February 2021
methods using 11 steel powders varying in chemistries and size fractions. Regression analysis was used to test
Keywords:
the relationship between each flowability metric obtained. Some metrics, such as the conditioned bulk density, re-
Powder flowability late to many flowability indicators. Others, such as the basic flowability energy, show poor correlations to other
Steel powders variables, likely describing different aspects of the powder flow behaviour. When two metrics show a strong cor-
Comparative flowability testing relation, as between conditioned bulk density and Hausner ratio, a numerical relationship is derived: CBD =
Flowability correlation − (5.65 ± 0.86)HR g cm−3.
Powder rheometer © 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://
creativecommons.org/licenses/by/4.0/).

1. Introduction bed based additive manufacturing process, the metallic powder should
be spread in a thin, compact, and even layer [12–14]. Similarly, in press
The flow of a powder is a complex phenomenon to understand or and sinter and hot isostatic pressing a good powder flowability can
model. It is sensitive to properties of the powder itself and external fac- avoid cavities and improve the die filling [15], while in cold spray depo-
tors such as humidity [1,2], geometrical impediments [3], or applied sition the flow rate of powder and gas can directly affect the coating de-
stresses [4]. Within the powder, many factors affect flow properties, position rate [16]. There are many viable techniques to measure flow in
such as the size distribution and shape of the particles, their surface a powder, but no agreement as to which is the best testing method for a
roughness and oxidation state, the material density, and electrostatic given application [6,17–21]. Furthermore, there is no agreement on
and magnetic characteristics [5,6]. Powder handling also affects flow what flow behaviour is required for a powder to be considered suitable
by altering the packing of particles and the ratio of solid to other phases for any specific manufacturing process.
in the ensemble. This complexity has significant consequences for many The flow behaviour depends on the balance between forces that
fields that rely on powder and its propensity to flow or remain cohesive, drive the motion of particles and those that prevent it. Forces promot-
such as powder metallurgy, soil mechanics, civil engineering, the phar- ing flow are gravity and applied external forces, while the ones oppos-
maceutical industry, paint production and food processing. If the pow- ing it include frictional forces between the powder particles and the
ders are more cohesive than predicted, this could limit the materials external surfaces, as well as forces between neighbouring particles.
stream in a process or affect the final product quality [7]. Therefore, it Such inter-particular forces include mechanical friction, the mechani-
is crucial to provide an accurate description of the flow of a powder cal interlocking of particles, van deer Waals forces, capillary forces,
for a specific process. When the same powder is used in different appli- electrostatic force, and magnetic force [22–24]. The magnitude of
cations, it may be important to provide a description that is relevant for these resistive forces is determined by an array of properties arising
a general process [6]. from the powder, such as powder particle size and particle size distri-
The growing focus on metal powder-based additive manufacturing bution, particle shape, material density, particles roughness, surface
technologies increasingly highlights the need to deliver components chemistry, oxidation, and water content [18,25]. In addition, the ma-
that comply with industry standards and operate under demanding terial of the testing equipment, its surface conditions, geometry, and
conditions. The powder plays a key role in the final component quality operating conditions (e.g., if the test uses stationary or moving pow-
[8–11]. For example, to produce low defects components in a powder- der, stress state, temperature, gas flow) can also have a significant ef-
fect on the flow behaviour. For example, when a stress is applied to a
⁎ Corresponding author. powder column of irregular particles, its density increases, locking the
E-mail addresses: lormar@kth.se (L. Marchetti), chrihs@kth.se (C. Hulme-Smith). particles in position. Hence, mechanical forces will propagate across

https://doi.org/10.1016/j.powtec.2021.01.074
0032-5910/© 2021 The Author(s). Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

particles more easily, strongly affecting the flow behaviour and over- Table 1
coming the effect of the weaker cohesive forces. This web of proper- Overview of the powders used in the experimental trials.

ties and mechanisms that compete and act concurrently makes Material classification Particle size distribution
powder flow a complex multivariate problem [17], where is not fully Production process Alloy Designation D10 / μm D50 / μm D90 / μm
clear exactly how properties of the powder itself affects flowability
Water atomization 17–4 PH 1 7.4 19.9 44.6
[17,19].
316L 2 8.2 20.7 42.3
Due to the complexity of the powder flow, there is no single tech- Gas atomization 17–4 PH 3 5.7 10.7 15.4
nique to characterise flowability for a specific application [26] and stud- 4 7.5 17.0 37.2
ies do not always agree on the best one to use for any specific case 5 59.8 75.1 95.6
[18,21,27]. Furthermore, many common methods found in the literature 316L 6 16.5 35.4 50.4
7 14.5 26.4 46.2
are not standardised, such as the rotating drum angle of repose, powder 420 8 11.5 28.3 45.4
rheometry and spreading devices [18,21,28]. Others, such as tap density Tool steel 1 9 25.0 37.8 57.0
testing, static angle of repose and shear cell testing are standardised Tool steel 2 10 23.9 35.8 55.4
[29–32], but not included in the recommended standards when Tool steel 3 11 18.4 30.9 55.7
characterising the flowability for additive manufacturing metallic pow-
ders [26]. ASTM F3049-14 suggests instead to characterise powders for
additive manufacturing with a Hall funnel, Carney funnel or Arnold
2. Materials and method
meter [26]. However cohesive powders might not be suited for these
methods as they B. Spierings, M. Voegtlin, T. Bauer, K. Wegener, Powder
2.1. Sample preparation
flowability characterisation methodology for powder-bed-based and
give non-quantitative results [21]. When ranking the flowability of sev-
Flowability tests were performed on 11 steel powders with various
eral powders, it will not be possible to differentiate between the non-
chemistries, atomization routes and particle size ranges (Table 1). The
flowing ones, making a comparison difficult. Also, characterising
particle size distribution of each powder was measured using a
flowability with different methodologies may lead to contrasting results
Camsizer XT particle analyser with compressed air dispersion. Addi-
[33,34]. To avoid this, it is often recommended that testing should be
tional data regarding particle shape and particle size distribution are in-
performed under conditions that reproduce the specific case of interest
cluded in the supplementary files (Supplementary Fig. 1 and
as far as possible [17,26].
Supplementary Table 1). Each powder batch was sampled with a rotat-
A holistic evaluation of the flow behaviour of a powder based on
ing sampler divider (spinning riffler) to obtain 5 individual specimens,
multiple testing would solve the limitations of using a single
prepared according to the relevant industrial standard (ASTM B215)
flowability test for powders intended for applications with different
[38]. Three specimens with a mass of 50 g were prepared to test the
parameters or where it is not clear which test is the most suited [6].
flow rate and tap density [39–41]. Two further specimens were pre-
However, a comprehensive model describing flowability has proven
pared with a volume of 25 ml for testing in a powder rheometer
problematic to obtain. This is due to the complexity the high number
[29,42] and a fixed height funnel angle of repose meter [43]. The speci-
of inter-dependent variables that must be accounted for in powder
mens were dried for 30 min at 105 °C to remove moisture immediately
flow [10]. Some models are published in literature, mostly restricted
before testing. Adsorbed water layers and liquid bridging between par-
to specific flow conditions and without explicit physical variables
ticles are known to deteriorate the flow properties of metallic powders,
[21,35]. Comparative testing can complement existing models and
especially at higher relative humidity [44–46]. While it is reasonable
give a more complete description of the problem [6,8,17,19]. Numer-
that during testing powders adsorbs some moisture from the environ-
ical modelling simulations such as the direct element method (DEM)
ment, it is unlikely that such uptake within the short time between dry-
are also a powerful tool to link interactions between particles to
ing and testing will have a large impact on the results [47].
their flow and the behaviour of the ensemble. In the case of fine metal-
lic particles, such modelling techniques may have some limitations.
2.2. Flowability measurements
This is partly due to increased computational demands of modelling
micrometre-scale particles and the need to consider several factors,
Flowability was evaluated with multiple measuring techniques,
such as the particle size distribution, shape, roughness, van deer
which can be divided into separate categories (Fig. 1). The difference
Waals forces, liquid bridging and eventual electromagnetic forces
within these techniques can be important, as a testing method affects
that are known to affect the flow. It may be possible to simplify the
the stress state, velocity of the powder particles, packing density and
force models used in the simulations, but more research is required
other factors in a powder. The ambient conditions and the history of
before such solutions may be applied with confidence and be realistic
the powder are also well known to influence its flow behaviour
for industrial applications [36,37].
[42,48,49].
In the present work we compare a range of flowability measuring
techniques available using a range of metal powder size fractions and
2.2.1. Flow rate measurements
chemistries. The results are used to evaluate the relationships between
When in a funnel, a powder may flow out due to gravity. A way to
the metrics from the different flowability measuring techniques using
quantify the phenomenon is the time required for a fixed mass (or vol-
statistical analysis. By including a range of powder chemistries and
ume) of powder to flow entirely out of the funnel. This metric is com-
size distributions, it is intended that relationships should not be specific
monly referred to as the flow rate.1 A lower flow rate is an indicator of
to individual alloys or powder production conditions. The common
a better flowability. When frictional forces are low, the powder flows
flowability measurement techniques are assigned to categories, where
out of the funnel and the powder remaining in the funnel forms a sur-
one representative method is selected from each. Every powder is
face that is approximately horizontal and gradually descends. When
then measured using each selected technique and statistical analysis is
frictional forces dominate, powders may form an arch near to the outlet
performed on pairs of metrics to derive any correlations. From these
of the funnel, which leads to the presence of stable vertical cavities
evaluations, it may be understood which techniques are equivalent of
each other and which techniques do not correlate. These are likely to 1
There is a contradiction that a “rate” measures how much or how often something
be measuring different responses in the powder to flow. Multiple tech- happens per unit time, but the flow rate is the time required for a fixed mass (or volume)
niques would be needed to describe all aspects of the powder flow of powder to leave the funnel. More correctly, this is a specific (or volumetric) time, not a
behaviour. rate, but the term flow rate is common.

403
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

Fig. 1. Flowability testing methods overview. For every testing method class, some of the common testing methods applicable are listed below. The highlighted methods have been chosen
for this work [19,21,29,39,40,42,43,50–57].

(ratholes) or jamming, where the powder does not flow at all. These ef- can be overcome by shaking the powder and allowing particles to rear-
fects are well studied in the literature and can be observed when range through tapping. Similarly, in the first stages of the particle mo-
powders are cohesive, when the ratio of the orifice diameter to the par- tion, powders will locally vary in their densities when particles start to
ticle diameter is too low or when the particle size distribution is too move one past each other. In the current work, the apparent density is
wide [58]. replaced by the conditioned bulk density (CBD, ρc) [42], which is a similar
Such techniques have been common for a century and are currently quantity measured in the Freeman FT4 powder rheometer from Free-
recommended for additive manufacturing [21], with several funnel ge- man Technology Ltd. (Tewkesbury, U. K.) (Eq. (1)). Conditioned bulk
ometries described in the active standards. The best-known experimen- density is measured after the powder sample has been gently stirred
tal setup is the Hall flowmeter (ASTM B213–17) [39], which works well by an upwards-moving blade and reduced to a fixed volume.
for non-cohesive powders. Cohesive powders may be tested using the
Carney funnel (ASTM B964−16) [40], with a wider bottom opening, RH ¼ ρt=ρc ð1Þ
and the Gustavsson funnel [56], with a steeper conical surface. In the
current study, Hall, and Carney funnels from Qualtech Products Ltd.
(Manchester, U. K.). were used.
2.2.3. Angle of repose
The angle of repose is the angle of a pile of powder at which it stops
2.2.2. Density ratios flowing. It is the complementary property to avalanche angle, which is
The ability of a powder to pack can be taken as an indicator for its the angle at which a stationary powder will begin to flow. A well-
flowability, as particles must move past each other (i.e., flow must flowing powder will not stop flowing when the powder pile has high
occur) to increase packing density [54]. In general, the packing of a pow- angles. However, lowering the pile angle, the gravitational component
der is quantified through its bulk density (defined as the mass of a sam- moving the powder will diminish. At the point where the powder
ple divided by the total occupied volume, inclusive of particles and air stops flowing, the gravitational component driving the flow is balanced
gaps). Santomaso et al. [54] make a distinction between measuring by the attractive and frictional forces within the powder (Fig. 2). Previ-
the bulk density of a powder in a loose or a dense packing condition. ous research has used the angle of repose to qualitatively classify the
For example, a loose random packing condition is achieved through flowability powders [65, Table 1].This angle can be measured in several
test as aerated, poured, or apparent density [57,59–61]. Other tests pro- alternative methodologies: by allowing a powder to fall through a
mote a denser random packing through different methods, allowing
air to exit the bulk volume. The most well-known example is the tap
or tapped density, which is achieved after the powder is gently shaken
and allowed to settle to a constant volume [41]. In this study, 3 samples
of 50 g were tapped using an Autotap tap density tester from
Quantachrome (Boynton Beach, Florida, U. S. A) for twice as many
taps as was necessary for the volume to stop changing, in accordance
with ASTM B527-15.
When comparing different powders, the ratio between high and low
packing densities, or Hausner Ratio, RH is commonly used. This is often
expressed as the ratio between the tapped density ρt is and the apparent
density ρa. However, some authors replaced the apparent density with
other alternative metrics, such as aerated density [35,62], poured den-
sity [63], apparent density determined with the Scott volumeter [64].
A high value of Hausner ratio indicates a high variation between the
starting loose packing density and the final dense packing density. Fig. 2. Forces acting on a single particle in a powder pile. Fg is the gravitational force, Fa is
the sum of the frictional forces, Fc the sum of the cohesive forces, FN is the normal force. F is
This is usually due to weak forces, such as mechanical contact forces the gravitational component promoting the flow. (a) With a steeper pile, F > Fa and the
or cohesive interparticle forces, which prevent the powder from particle will move down the inclined surface; (b) when the powder pile is flatter, F < Fa
reaching an optimal loose packing density. The effect of these forces so the particle slows down, and eventually stop.

404
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

funnel onto a flat surface and form a pile (fixed funnel method), by filling 2.2.5. Rheometry
a hollow cone with powder and removing the cone, allowing the pow- Powder rheometry is becoming increasingly popular also for metal-
der to settle into a pile (hollow cylinder method), or by causing powder lic powders [6,8,67,68]. In addition to the shear cell unit, other test
to run along a planar surface tilted at some angle and finding the tilt methodologies are available to characterise the powder behaviour
angle at which the powder stops moving (tilting box method) [43]. under different conditions. The testing is mostly automated, while the
While this final method measures the avalanche angle, it is commonly interaction with the operator very limited. To homogenise the sample,
referred to as a technique to measure angle of repose. In all these a pre-mixing (conditioning) is performed prior testing. Two standard
cases the cone angle of the pile of powder is measured statically. It programs of the FT4 powder rheometer were used with a 25 ml sample
should be noted that there is evidence that the angle of repose depends and a vessel diameter of 25 mm. Each test was performed on two
strongly on the method chosen to measure it. For example, the proce- samples.
dure outlined in the standard ASTM C1444 (a variant of the fixed funnel The first program was the stability and variable flow rate test, in
method) gives a higher angle of repose than the cone lifting technique which an impeller moves in a set path through the powder (Fig. 3a).
[65]. In the current study, a fixed funnel method was used, with a pro- For each sample, the test automatically runs 11 times: seven tests run
tractor fitted to the funnel stand to measure the angle of the side of at a fixed vertical blade tip speed and then four runs at increasing verti-
the pile directly. The funnel has a cone angle of 60°, an orifice diameter cal blade speeds. The resistance met from the impeller is recorded and
of 5 mm and a distance between the orifice and the tray where the angle converted into synthetic indices. The basic flowability energy (BFE) is
forms of 80 mm (Qualtech Products Ltd., Manchester, U. K). When pos- the energy spent from the blades moving downwards, where the pow-
sible, two samples of 25 ml were measured three times for each powder, der is compressed and confined to a fixed volume by the blade and the
where every measurement consisted of three angles observations at an- walls of the test vessel. The specific energy (SE) represents the energy
gles from the same pile, as suggested in the instrument operating man- spent when the blades move upwards, when the powder is free to oc-
ual. The angle of repose was taken as the average angle of the pile. cupy more volume (unconfined flow) [42].
Measuring the angle with a protractor mounted on the funnel assembly, The second method performed was the compressibility test, where
it was possible to exclude the effect of possible distortions on the pile, the powder is confined in a vessel and slowly subjected to uniaxial com-
such as the presence of a small flat cone tip, or a wide base, by taking pression (Fig. 3b). At the same time, the variation in height is recorded.
the angle in the remaining part of the pile edge. Other methods, which The compressibility (CPS) is then calculated as change in volume before
calculate the angle based only on the diameter and the height of the and after the powder is compacted with a 15 kPa applied compressive
pile, could be more prone to these errors. stress [42].
These methods are inexpensive and traditionally used to character-
ise powder flow, but none of them are described in the current stan-
dards for the characterisation of metallic powders [30–32,51]. A 3. Calculation
related technique, which is mostly automated and not standardised, is
the dynamic angle of repose or rotating powder analyser, in which pow- 3.1. Parameters
ders are placed in a drum and rotated for a pre-programmed speed
cycle, during which pictures are taken and evaluated automatically. In this study a broad overview of the relationships between several
Powders rise the drum until some critical angle is reached (avalanche flowability measurements is presented. For every flow test performed,
angle), at which it begins to flow. It then stops at some lower angle one or more metric can be recorded to evaluate the behaviour of a pow-
(rest angle). Based on these angles and the irregularity of the powder der (Table 2). Each of these metrics was then compared with all the
profile, several parameters are automatically calculated and can be other, as well as conditioned bulk density. To focus on the relationship
used to indicate flow behaviour. between testing methods, no distinction was made between different
powders and all the eleven tests were taken as a dataset. The mean of
all repeats was reported for each powder. All results are supplied as a
2.2.4. Shear cell supplementary file in the online version of the article.
In a shear cell, a powder sample is confined between two flanges,
which subject the powder to a pre-set cycle of normal and shear
stresses. First, the powder sample is consolidated to a set pre-
compaction normal stress (σ1c). It is then sheared at different compres-
sive stresses below σ1c. Some powder particles move coherently with
one moving flange, while some other adopt the velocity of the second
flange, which is usually fixed. There is a region where particles will tran-
sition from following one flange to following the other. In this region,
particles move past to each other under a shear stress. The shear stress
necessary to initiate the flow is recorded at different compressive
stresses. The normal and shear stresses are then used in a Mohr's circle
to plot the stress state at which flow first began in at each compressive
stress, known as the locus of incipient flow. Several parameters can be
derived from this diagram, such as cohesion (C), unconfined yield
strength (UYS), major principal stress (MPS), angle of internal friction
(AIF) and the flow function coefficient (ffc or FF) [42,66]. In general, a
good flowability leads to lower shear values of these metrics, except
for the flow function coefficient, where a higher value should indicate
a better flow behaviour transitioning from static to dynamic friction.
In this work, a shear cell supplied with the FT4 powder rheometer
was used following ASTM D7891–15 [29]. Pre-compaction stresses of
3 kPa and 9 kPa were chosen. To increase the data reliability, two sam-
ples were tested for each powder. The pre-compacted volume of the Fig. 3. (a) Sketch of the FT4 vessel and impeller in the Stability and Variable flow rate,
tested sample was of 10 ml, while the vessel diameter 25 mm. (b) sketch of the vessel and the flange for the Compressibility test.

405
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

Table 2
Samples, devices, test performed and flowability indicators considered. N
∑ðxi −xÞðyi −yÞ
i¼1 σ xy
Sample Test Device Test Flowability Unit β11 ¼ ¼ ð4Þ
class indicator N σ 2x
∑ ðxi −xÞ2
50 g Flow Hall funnel Hall flow rate [31] Flow rate s (50 g)−1 i¼1
(3×) rate Carney Carney flow rate [32] Flow rate s (150 g)−1
funnel β10 ¼ y−β11 x ð5Þ
Packing Tap-meter, Tapped density [33], Hausner
density Rheometer Conditioned bulk ratio
density [34] β11 determines the steepness of the correlation between two vari-
25 ml Other Rheometer Flowability [34] BFE mJ ables (Fig. 4b). However, this does not measure the dispersion of the
(2×) SE mJ (g)−1 data. For this, the Pearson correlation coefficient, r, is used (Eq. (6)).
Other Rheometer Compressibility CPS (%)
For this reason, it can give an indication of the strength of the correlation
@ 15 kPa [34]
Shear Rheometer Shear test Cohesion kPa and its direction. As the Pearson correlation coefficient is scaled for each
cell with shear @ 3 kPa [23] UYS kPa variable, it oscillates between −1 (inverse perfect correlation) and +1
cell MPS kPa (direct perfect correlation), while 0 implies no correlation.
FF
AIF degrees N
Shear Rheometer Shear test Cohesion kPa ∑ðxi −xÞðyi −yÞ
cell with shear @ 9 kPa [23] UYS kPa i¼1 σ xy σx
r ¼ sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffisffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ ¼ β11 ð6Þ
cell MPS kPa N N σ xσ y σy
FF ∑ ðxi −xÞ2 ∑ ðyi −yÞ2
AIF degrees i¼1 i¼1
Angle of Fixed Fixed funnel angle of Angle of degrees
repose funnel repose repose
angle of
repose
3.3. Non-linear correlation

It is possible that pairs of metrics may be correlated in a non-linear


way [35]. To assess the non-linear data trend between flowability pa-
rameters, a multiple linear regression was performed. In addition to f1
3.2. Linear correlation
(xi), the data have been fitted to polynomial (Eqs. (7) and (8)) logarith-
mic (Eq. (9)) or mixed (Eq. (10)) functions. The polynomial expressions
As a first evaluation step, the correlation across each pair of param-
are a natural extension to the linear correlation. The logarithmic depen-
eters (X and Y) was investigated using the programming language R.
dence was tested in the eventuality that a correlation could gradually
To analyse the data, the generic equation for a straight line was used
decrease in strength.
(Eq. (2)). The intercept β10 and the slope coefficient β11 were deter-
mined performing a linear regression with the ordinary least square
f 2 ðxi Þ ¼ β20 þ β21 xi þ β22 x2i ð7Þ
method, where the sum of the squares of the residual ϵi (Eq. (3),
Fig. 4a) is minimized.
f 3 ðxi Þ ¼ β30 þ β31 xi þ β32 x2i þ β33 x3i ð8Þ
f 1 ðxi Þ ¼ β10 þ β11 xi ð2Þ
f 4 ðxi Þ ¼ β40 þ β41 log ðxi Þ ð9Þ

yi ¼ f ðxi Þ þ ϵi ð3Þ f 5 ðxi Þ ¼ β50 þ β51 xi þ β52 log ðxi Þ ð10Þ

The solution of a system between Eqs. (2) and (3) leads to the deter- Multiple linear regression is similar to the linear regression, where
mination of β10 and β11 (Eqs. (4) and (5)). linear coefficients (βk0, βk1, βk2, βk3) are found minimizing the matrix

Fig. 4. Graphical description of the correlation between the Angle of repose and the Hausner ratio. (a) shows the straight line fitted to the data, the predicted values of the Hausner ratio as a
function of angle of repose, f(xi), (red diamonds) and the actual Hausner ratio, yi, (black dots). The vertical lines highlight the residuals, ϵi; (b) the uncertainty σ11 around the slope
coefficient of the gradient β11.

406
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

of the square residuals [ϵ], obtained subtracting the fitted values [fk(x)] 4. Results and discussion
from the observed values [y]. The adjusted coefficient of determination
R2adj was used to test the suitability of the model. As R2, R2adj explains how 4.1. Linear correlation
much of the total variation from of the data from the average value is
expressed through the model (Eq. (11)). If the data had a similar disper- While most of the flowability parameters reveal a positive linear cor-
sion to the model, then R2 will tend to 1, meaning that the model fits the relation, the conditioned bulk density (CBD) and the flow function coeffi-
data well. Otherwise, if there are additional sources of variance in the cient (FF) clearly show an inverse one (Fig. 5a).
data, such as random scatter, R2 will fall to 0. In this case the model Hall and Carney flow rate exhibits a strong correlation with several
does not explain any variance of the data. However, R2 increases with variables (e.g., positive with Hausner ratio and negative with compress-
many coefficients, leading eventually to overfitting. R2adj adjusts for the ibility, CPS), while showing a weak dependence with most others. How-
number of coefficients in the model (k) relative to its data points (N) ever, a powder with a good flowability would be expected to have a
(Eq. (12)). lower flow rate, as well as a lower Hausner ratio and compressibility. It
is useful to note that, due to the cohesiveness of many of the powders
N
used in this study, few could give quantitative results in flow rate test-
2
∑ ð f k ðxi Þ−yÞ  2 ing: jamming or ratholing were observed on many samples, in agree-
σ xy
2
R ¼ i¼1
¼ ð11Þ ment with what was stated from Marnani et al. for fine and ultra-fine
N σ xσ y particles [71] and Muñiz-Lerma et al. for fine metallic powders [73].
∑ ðyi −yÞ2
i¼1 The low number of flow rate data available prevents us from drawing
conclusions on possible correlations (Figs. 2–4 in the Supplementary
  N−1 file).
R2adj ¼ 1− 1−R2 ð12Þ Conversely, Hausner ratio, angle of repose (AOR) compressibility and
N−k−1
specific energy (SE) all show a strong positive correlation between
them (Fig. 5 and Figs. 8–11 in the Supplementary file). These variables
also exhibit strong interactions with other shear test parameters. For in-
3.4. Statistical significance stance, Angle of repose and compressibility are strongly correlated with
cohesion, unconfined yield strength (UYS), major principal stress (MPS)
Although the linear coefficients βk (or the regression coefficient r) and flow function coefficient (FF), correlated with the conditioned bulk
are obtained minimizing the sum of residuals ϵi, there is always some fi- density and weakly correlated with the angle of internal friction (AIF).
nite probability that the fit is simply due to random scatter in the data Conversely, Hausner ratio and Specific energy are weakly correlated
(i.e., there is not enough evidence to confirm the relationship between with cohesion, unconfined yield strength and flow function coefficients,
variables). This probability is given from the p-value. When a p-value while strongly correlated with the major principal stress, angle of internal
is smaller than a pre-set significance level, there is enough evidence to friction and conditioned bulk density (Fig. 5). Many interactions between
consider the relationship statistically significant. In this work the re- shear test variables were similarly found in a study from Li et al. [73],
quired significance level for all the p-values was set at 0.05, thus mean- where 11 powders were tested with different Freeman FT4 setups. Co-
ing that there is 95% probability that the calculated fits are genuine hesion was found positively correlated with unconfined yield strength
trends. The p-values were calculated with a Student's t-test, based on and major principal stress, while negatively correlated with the flow
a Student t distribution. This is preferred in regression analysis when a function coefficient, indicating that a more cohesive has a lower
small data set with unknown variance is used [69,70]. flowability.

Fig. 5. Linear relationships across flowability indicators: (a) correlation matrix, where the size of the circles represents the modulus of the Pearson coefficients (from 0 to 1). The colour
describes the sign of the Pearson coefficients (white for a positive coefficient, black for a negative one); (b) map of the most important correlations between metrics.

407
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

Basic flowability energy (BFE) shows weak between most of the var- [35,54], while there is less information available for the other variables.
iables (Fig. 5 and Fig. 5 in the Supplementary file) and only some corre- These variables still exhibit some correlation with shear cell parame-
lation with the angle of internal friction and the flow function coefficients. ters: while all except for compressibility are correlated with the major
It is worth noting that basic flowability energy is one of the few parame- principal stress, only the angle of repose and compressibility are correlated
ters to be poorly related to the conditioned bulk density (Fig. 6 in the Sup- with the unconfined yield strength. Finally, only compressibility is corre-
plementary file). This disagrees from what was found by Li et al. [74], lated with both cohesion and the flow function coefficients. The correla-
where the basic flowability energy clearly correlates with the condi- tions can be summarised in the map shown in Fig. 6b, where only
tioned bulk density. In this study the powders tested had a significant variables correlating both to shear variables with pre-compaction at
difference in density. In our study instead, all powders used are iron 3 kPa and 9 kPa are included.
based steels, with a relatively similar size fraction and density, so that Across the shear cell results, the strong correlation between cohe-
the effect of the density on the basic flowability energy goes unseen. sion, the ultimate yield strength and the flow function coefficient can be
Other parameters poorly related to the conditioned bulk density are confirmed. The uncertain correlation the major principal stress with
shear cell metrics. Cohesion, unconfined yield strength and flow function other variables from shear cell testing is also confirmed, as well as its
coefficients show a strong correlation between each other, but a poor strong correlation with other variables (Hausner ratio, specific energy,
correlation with the conditioned bulk density. Conversely, the angle of in- angle of repose) and the conditioned bulk density (Fig. 6a). Finally, the
ternal friction is correlated with the conditioned bulk density. The Major Conditioned bulk density shows a strong negative correlation with the
principal stress always showed a strong correlation with the conditioned specific energy, Hausner ratio (Fig. 7 in the Supplementary file), Angle
bulk density (Fig. 5). of repose, major principal stress, while none with compressibility.
Fig. 5b further summarizes some of the correlations analysed. Pa- The diagram in Fig. 6b shows only the significant relationships pre-
rameters showing ambiguous relationships with other variables, such viously reported in Fig. 5b. Parameters showing ambiguity, such as
as Hall flow rate, Carney flow rate or the major principal stress (MPS) in Hall flow rate, Carney flow rate or major principal stress in relationship
relationship to some shear cell parameters, were not included. to some shear cell parameters (cohesion, ultimate yield strength, flow
All the Pearson correlation coefficients obtained (Fig. 5a) were sub- function coefficient, angle of internal friction) were not included.
sequently filtered according to their significance level. The matrix of
the statistically significant correlation coefficients (with a confidence 4.2. Non-linear correlation
of 95% or more) is shown in Fig. 6a. Despite the data filtering, 51 of
these relationships proved to be statistically significant, while the re- Along with the linear correlations, Fig. 7 shows the statistically sig-
maining 102 did not. For example, almost all the correlation coefficients nificant non-linear relationships obtained through the multiple linear
related to Hall and Carney flow rate are found to be not significant regression analysis (where all the individual coefficients were found
(Fig. 6a). The only exception is the correlation between Carney flow to be significant). Quadratic (Eq. (7)), cubic (Eq. (8)), logarithmic
rate and the flow function coefficient with pre-compaction at 3 kPa. Al- (Eq. (9)) and mixed (Eq. (10)) relationships were all evaluated, but
though this proved to be statistically significant, the data suggests that only quadratic or cubic were found to be significant in any case. All
the correlation is due to the lack of data available and not due to under- the non-linear relationships found are like the ones reported in Figs. 9
lying physical causes (Fig. 4 in the Supplementary info). and 12 in the Supplementary file. While it is clear there is a statistically
Angle of repose, specific energy, compressibility and Hausner ratio con- significant variation from linearity, it is also difficult to explain the data
firmed most of their mutual strong positive correlation (Fig. 6a, trend shown.
Figs. 8–11 in the Supplementary file). The correlation between the In some cases, non-linear relationships were found to fit the data
angle of repose and Hausner ratio is often found in the literature slightly better than some linear ones (for example the flow function

Fig. 6. Significant linear relationships across flowability indicators. a) Correlation matrix showing only the Pearson correlation coefficients with p a p-value < 0.05 (or with a 95% probability
of being representative). The size of the circles represents the modulus of the Pearson coefficients (from 0 to 1). The colour describes the sign of the Pearson coefficients (white for a pos-
itive coefficient, black for a negative one). b) Scheme of the most important and statistically significant (p<0.05) correlations.

408
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

Fig. 8. Scheme of the correlations statistically significant (p < 0.05) and representative
(R2adj > 0.7).

Fig. 7. Correlation matrix summarizing the significant relationships between flowability


indicators. The chart includes both linear and non-linear relationships. All the
In this study, only the correlations explaining the data to a good degree
correlations and correlation coefficients shown have p < 0.05.
should be considered. Table 4 in the Appendix reports all the statistically
significant regression coefficients βk along with their confidence inter-
val and the coefficient of determination R2 for the model considered,
coefficient with pre-compaction at 9 kPa and compressibility), with the which can be used to evaluate the goodness of fit for each correlation.
latter having already a significance level above the pre-set 95% value. Table 3 instead reports only the correlations coefficients having a
For this reason, considering a non-linear relationship between these R2adj > 0.7 (meaning that 70% of the variation observed in the data can
variables could lead to overfitting. explained through the proposed model). A visual explanation of these
The correlation between Carney flow rate and flow function coefficient results is shown in in Fig. 8. Out of the 59 statistically significant rela-
with pre-compaction at 3 kPa was instead excluded from Fig. 7 as there tionships, only 15 are shown to fit well the data. Many of the correla-
are few data available from the Carney flow rate trials (Fig. 4 in the Sup- tions previously found between specific energy, Hausner ratio, angle of
plementary file). Taking this into account, the total number of statisti- repose and compressibility are not found to be explanatory. However,
cally significant relationships increased from 51 to 59. The angle of these metrics are still found to be important, relating either to shear
repose was confirmed to correlate linearly to the Hausner ratio, as also cell parameters or to conditioned bulk density.
observed from Geldart et al. [35] and Santomaso et al. [54].

4.3. Significant and representative quantitative relationships between 4.4. Physical explanation
metrics
Many of the powders tested in this work showed a cohesive behav-
Although the relationships from in Fig. 7 are considered statistically iour during the flow rate trials. When the powders obstructed the Hall
significant, this does not explain how well the coefficients βk fit the data. and Carney funnels, no flow rate could be recorded. Table 4 compares
the flow rates and the powder size in relationship to the funnel outlet
diameter, showing how for powders of similar chemistry and produc-
Table 3 tion technique, the finer fractions have a greater tendency to jam
Statistically significant (p < 0.05) and representative (R2adj > 0.7) correlations across the funnels (sample 3 and 4, 7 and 11), while the coarsest powder
flowability indicators.

Variables Function β1 ± σβ1 R2adj

[β2] [± σβ2] Table 4


Measured Hall and Carney flow rates and ratio D50 (Doutlet)−1 for each sample.
MPS9 = ƒ(HR) Linear 10.1 ± 1.2 0.86
CBD = ƒ(HR) Linear −5.65 ± 0.86 0.81 Powder Hall funnel flow test Carney funnel flow test
SE = ƒ(AOR) Linear 0.083 ± 0.01 0.78 −1
D50 (Doutlet) Flow rate / s D50 (Doutlet)−1 Flow rate / s
MPS3 = ƒ(AOR) Linear 0.033 ± 0.006 0.74
MPS9 = ƒ(SE) Linear 0.337 ± 0.16 0.74 1 8.0 Jamming 4.0 Jamming
AIF9 = ƒ(SE) Polynomial,2 −34.9 ± 7.2 0.80 2 8.3 Rathole 4.1 Rathole
[6.52] [± 1.3] 3 4.3 Jamming 2.1 Jamming
FF3 = ƒ(CPS) Linear −0.895 ± 0.16 0.74 4 6.8 Jamming 3.4 Jamming
UYS3 = ƒ(C3) Linear 2.7 ± 0.1 0.98 5 30 21 15 4.7
FF3 = ƒ(C3) Linear −14 ±3 0.74 6 11 11 5.7 2.8
UYS9 = ƒ(UYS3) Linear 1.3 ± 0.3 0.70 7 11 Jamming 5.3 Jamming
UYS9 = ƒ(C9) Linear 2.5 ± 0.2 0.96 8 11 12 5.7 2.4
FF9 = ƒ(UYS9) Linear −6.8 ±1 0.82 9 15 12 7.6 4.4
FF9 = ƒ(C9) Linear −20 ±3 0.81 10 14 11 7.2 2.6
CBD = ƒ(MPS9) Linear −0.502 ± 0.096 0.72 11 12 11 6.2 Jamming

409
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

(sample 5) is always able to flow through the funnels (table). This is op- funnel and come to rest in a pile: the cohesive and frictional forces act
posite from what found from Janda et al. [74], where bulk solids jam to oppose a motion that is imposed on the powder, which are partly
when their particle size is too coarse in relationship to the funnel orifice. allowed to rearrange (Fig. 10 in the Supplementary file). Conversely,
While in this case the interlocking mechanisms, affecting coarser parti- in the compressibility test, powders are slowly compressed up to a pre-
cles, can explain the jamming, in our work the fractions that are finer in set stress, overcoming weak cohesive forces. However, the pre-set low
relationship to the funnel diameter are more prone to the jamming phe- level of stress and the single action movement prevent powders to rear-
nomenon due to their higher cohesive forces. This can be seen very range with a higher packing fraction. For this reason, this test does not
clearly with sample 11, where the same powder was able to flow correlate well with other metrics, such as Hausner ratio, for all the pow-
through the Hall funnel, but not through the Carney funnel. Our findings ders measured (Fig. 9 in the Supplementary file). Instead, compressibility
agree with other results in the literature [75]. With less data available, correlates with cohesion and ultimate yield strength, where the powder
any outlier can have a significant impact on any data trend observed resistance to flow is driven by weak applied forces and in static condi-
for the flow rate trials. This can explain why flow rate shows little corre- tions (Table 3 and Fig. 8). In these tests, the mechanism of particle rear-
lation with other variables (Figs. 10–12 in the Supplementary file). rangement is inhibited, allowing results to be compared to the
Basic flowability energy measures flowability dynamically in pow- compressibility test. Zegzulka et al. found that flow function coefficient,
ders confined under a compressive load. These conditions are unique angle of internal friction and angle of repose vary all similarly across dif-
amongst the techniques used in the current study and it explains why ferent metallic powders, and the particle shape has a prominent effect
basic flowability energy shows no correlations with other parameters on their flowability. This result is somewhat different from what we
(Figs. 5–6 and Figs. 5–6 in the Supplementary file). This suggests that found in our study, but it highlights the existence of a similar trend be-
powders subjected to a dynamic flow and confined under the blade tween the angle of repose and some rheometry variable [79].
stress behave very differently than other cases and highlights the need In the stability and variable flow rate test, specific energy is mostly de-
to test flowability under conditions that are representative of the pro- termined by interactions due to attractive forces between particles, such
cess for which the powders are intended. as electrostatic forces or polar attractions due to the presence of water.
In our study the conditioned bulk density was found to be clearly con- Since these forces generally weak, the inertia of the powder due to the
nected with multiple flowability metrics (Figs. 6,8 and Fig. 7 in the Sup- work done by the impeller strongly affects the results. For this reason,
plementary file), but a weak correlation between conditioned bulk specific energy is normalized by the sample mass (hence the use of the
density and basic flowability energy (Fig. 6 in the Supplementary file). word “specific” in the name of the metric). However, the current data
The correlation between flowability and powder density in generally still shown a strong and statistically significant correlation between
known in the literature: Santomaso et al. brings evidence of a correla- the specific energy and the conditioned bulk density. While normalizing
tion between the density ratios and the angle of repose [54], while Ab- the force recorded from the impeller for the sample mass removes the
dullah and Geldart find an important contribution of the particle shape effect of different powder densities, it is also true that the conditioned
distribution to both bulk density and flowability for fluid-cracking cata- bulk density, as well as the specific energy, can be an indicator of the at-
lyst and fire-retardant filler powders [76]. Hou and Sun note that higher tractive forces between particles in the powder (Fig. 6).
particle density corresponds to a better flow behaviour for pharmaceu-
tical powders tested with a powder shear cell [77]. The reason to this 4.5. Testing strategy
could be that in many flowability testing techniques, only mild forces
are applied to the powder, limiting the influence of mechanisms such Flow rate metrics proved unsuitable to evaluate flowability with our
as mechanical friction, or the propagation of forces through particle samples. For this reason, in disagreement with current standards [26],
interlocking. These mechanisms are more pronounced when stronger these metrics are not recommended to measure flowability for cohesive
forces are applied to the powder and with an increased particle packing. steel powders.
For this reason, in flow tests with cohesive powders, weaker forces be- Many single metrics proved to offer meaningful results, however not
tween neighbouring particles may predominate. Similarly, these always in agreement. For example, when evaluating flowability,
weaker forces are important when determining the conditioned bulk Hausner ratio, specific energy, and angle of repose showed an agreement.
density of these powders. Conversely, when a stress is applied to the However, a funnel flow device to measure angle of repose like the one
powder, the mechanical friction and interlocking are more important, used in this study is not indicated with cohesive powders as the results
and the flow is therefore determined by other powder properties. This reading heavily relies on the operator as the funnel could get blocked
could explain why the basic flowability energy does not correlated with with cohesive powders. Compressibility behaves similarly but
the conditioned bulk density (Fig. 6 in the Supplementary file). underestimating the effect due to particle rearrangement compared to
Specific energy, Hausner ratio, compressibility, and angle of repose Hausner ratio and specific energy. Conversely, basic flowability energy
(Fig. 6 and Figs. 7–11 in the Supplementary file) seem to be strongly re- gives unique results, strongly enhancing the effect of frictional mecha-
lated and are all correlated to the conditioned bulk density, suggesting nisms and particle interlocking. Flowability results that, similarly to
that the effect of weak cohesive forces is limiting in these flowability our case, are not in agreement were found also from Mellin et al.
tests. However, there is variation between these flowability indicators. when comparing the Basic flowability energy and the specific energy
From the current data, specific energy and Hausner ratio are more corre- with the Hall and Gustavsson flow rates [80]. These differences suggest
lated to each other than to other metrics (Fig. 8 in the Supplementary that when powder flowability is needed to be evaluated comprehen-
file) and with the major principal stress (Table 3 and Fig. 8). The correla- sively, a multiple tests strategy is needed, in agreement with similar
tion observed between Hausner ratio and angle of repose agrees with considerations made from Leturia et al. [6] and Prescott [17]. For exam-
what was found from Riley and Mann for glass particles of different ple, basic flowability energy, compressibility and one metric between
shapes [78]. Similarly, Geldart et al. found a clear correlation between Hausner ratio andspecific energy can measure the flow behaviour in dif-
the angle of repose and the Hausner ratio for different non-metallic ferent conditions. Instead, when the flow characterisation is needed for
powders [35]. The reason may be that Hausner ratio considers both a specific application, it is recommended to select a flow test that can
weak cohesive forces and the rearrangement of particles, as powders replicate adequately the powder condition. However, Ghadiri et al. [7]
are shaken in the tapped density measurement. It is reasonable that observe that this could not always be possible, as when evaluating the
both these phenomena are also affecting the specific energy and major spreading of powders for additive manufacturing.
principal stress measurements, where the powder resistance to an ap- Shear cell testing records accurately the powder stress state and can
plied stress is recorded under conditions of low stress. Similarly, during offers several metrics to evaluate flowability. The major principal stress
the angle of repose measurement powders are forced to fall from a was found correlated to other metrics (Hausner ratio, specific energy,

410
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

and angle of repose), while cohesion and ultimate yield strength were 6. Future work
found linked to the compressibility (Table 3 and Fig. 8). This means
that it is possible to reliably express all the shear cell metrics with In this work, we focused on the comparison between several
other metrics (and vice-versa) for the powders in our possession. How- flowability testing methods. However, this will be completed when
ever, as it was experienced that all shear cell metrics are sensitive to in- we will establish a procedure to characterise the spreadability of steel
terferences, including a correct sample preparation, these values can be powders. In addition, the influence of some relevant powder properties
reliably obtained from other metrics (Hausner ratio, specific energy, on flowability should be determined. For example, the effects of the par-
angle of repose, compressibility), as shown in Table 3 and Fig. 8. ticle size distribution, morphology, and environmental moisture. Fi-
nally, is could be useful to expand this work to a wider range of
5. Conclusions powder chemistries, such as titanium or aluminium alloys and non-
metal powders.
• A testing strategy using multiple techniques is necessary when
characterising the flow behaviour of a powder in a comprehensive
way: different metrics depend on different properties of the powder Declaration of Competing Interest
and, so, give different results. Basic flowability energy, compressibility
and at least one of Hausner ratio and specific energy revealed to repre- The authors declare that they have no known competing financial
sent different aspects of the flow behaviour of cohesive steel powders. interests or personal relationships that could have appeared to influ-
• Hall and Carney flow rate methods were found not suitable for ence the work reported in this paper.
assessing the flowability for cohesive steel powders. The funnel
method is not suitable for the cohesive powders because the jamming Acknowledgments
does not allow to record a flow rate, useful to compare different pow-
ders. For this reason, it is not possible to determine any significant and This work was made possible by financial support from the Swedish
representative correlations between either Hall or Carney flow rate Governmental Agency for Innovation Systems (Vinnova), project num-
and any other flowability metric. ber 2019-01087.
• The conditioned bulk density (CBD) showed a strong correlation with
most metrics. Many of these correlations were found to be statistically Appendix A. Supplementary data
significant, while 2 (CBD with Hausner ratio (HR) and major principal
stress with pre-compaction at 9 kPa (MPS9), Eq. (13) and (14) were Supplementary data to this article can be found online at https://doi.
also found to be representative of the data. org/10.1016/j.powtec.2021.01.074.

References


CBD ¼ −ð5:65  0:9Þ∙HR R2adj ¼ 0:81 ð13Þ [1] H. Lu, X. Guo, Y. Jin, X. Gong, Effect of moisture on flowability of pulverized coal,
Chem. Eng. Res. Des. 133 (2018) 326–334, https://doi.org/10.1016/j.cherd.2018.03.
 023.

CBD ¼ −ð0:502  0:22Þ∙MPS9 R2adj ¼ 0:72 ð14Þ [2] P. Mellin, M. Rashidi, M. Fischer, L. Nyborg, L. Marchetti, C. Hulme-Smith, M.
Uhlirsch, A. Strondl, Moisture in metal powder and its implication for processability
in L-PBF and elsewhere, Berg. Huettenmaenn Monatsh (2021)https://doi.org/10.
1007/s00501-020-01070-2.
[3] A. Janda, I. Zuriguel, A. Garcimartín, L.A. Pugnaloni, D. Maza, Jamming and critical
• The basic flowability energy (BFE) showed a weak correlation with outlet size in the discharge of a two-dimensional silo, Europhys. Lett. 84 (2008)
most flowability metrics, consistent with the fact that powders were 44002, https://doi.org/10.1209/0295-5075/84/44002.
confined and compressed during testing. It is reasonable that this [4] D. Macri, R. Chirone, H. Salehi, D. Sofia, M. Materazzi, D. Barletta, P. Lettieri, M.
Poletto, Characterization of the bulk flow properties of industrial powders from
stress state heavily promoted mechanical interlocking and inter- shear tests, Processes 8 (5) (2020) 540, https://doi.org/10.3390/pr8050540.
particle mechanical friction. [5] C.E. Capes, Chapter 2 - agglomerate bonding, Handbook of Powder Technology, 1,
• Specific energy (SE), Hausner ratio (HR), compressibility (CPS) and 1980, pp. 23–51, https://doi.org/10.1016/B978-1-4832-5666-5.50007-4.
[6] M. Leturia, M. Benali, S. Lagarde, I. Ronga, K. Saleh, Characterization of flow proper-
angle of repose (AOR) were found to be inter-correlated. Specific en- ties of cohesive powders: a comparative study of traditional and new testing
ergy, Hausner ratio and angle of repose were found to behave very sim- methods, Powder Technol. 253 (2014) 406–423, https://doi.org/10.1016/j.powtec.
ilarly and the results from Hausner ratio and angle of repose, being 2013.11.045.
[7] M. Ghadiri, M. Pasha, W. Nan, C. Hare, V. Vivacqua, U. Zafar, S. Nezamabadi, A. Lopez,
statistically significant and representative of the data, can be reliably
M. Pasha, S. Nadim, Cohesive powder flow: trends and challenges in characterisa-
converted between them (Eq. (15)). Compressibility was found to tion and analysis, KONA Powder Part. J. (2019)https://doi.org/10.14356/kona.
lead to some different results, probably because the test conditions 2020018.
lead to underestimate the effect of particle rearrangement compared [8] S. Vock, S. Jacobs, M. Härtel, B. Klöden, T. Weißgärber, B. Kieback, A systematic ap-
proach for understanding powder influence in powder bed based additive
to other techniques. manufacturing, Euro PM2018, EPMA, Bilbao (Spain), 2018.
[9] R. Dattani, K. Brockbank, J. Clayton, Optimising feedstock selection for additive
manufacturing, Euro PM2018, EPMA, Bilbao (Spain), 2018.
 [10] J. Clayton, Optimising metal powders for additive manufacturing, Metal Powder

SE ¼ ð0:083  0:03Þ∙AOR R2adj ¼ 0:78 ð15Þ Rep. 69 (2014) 14–17, https://doi.org/10.1016/s0026-0657(14)70223-1.
[11] A.B. Spierings, D. Bourell, N. Herres, G. Levy, Influence of the particle size distribution
on surface quality and mechanical properties in AM steel parts, Rapid Prototyp. J. 17
(2011) 195–202, https://doi.org/10.1108/13552541111124770.
[12] A. Strondl, O. Lyckfeldt, H. Brodin, U. Ackelid, Characterization and control of powder
• Other metrics were found strongly correlated. For example, the major properties for additive manufacturing, JOM 67 (2015) 549–554, https://doi.org/10.
principal stress (MPS) was found to be more correlated to Hausner 1007/s11837-015-1304-0.
[13] B. Liu, R. Wildman, C. Tuck, I. Ashcroft, R. Hague, Investigation of the effect of particle
ratio, specific energy, and angle of repose. Conversely, cohesion and ulti-
size distribution on processing parameters optimization in selective laser melting
mate yield strength were found to be more strongly correlated to com- process, International Solid Freeform Fabrication Symposium: An Additive
pressibility and angle of repose (Table 3). This means that is possible to Manufacturing Conference, SFF, Austin (US), 2011.
reliably express all the shear cell metrics with other metrics (and vice- [14] S. Ziegelmeier, P. Christou, F. Wöllecke, C. Tuck, R. Goodridge, R. Hague, E. Krampe, E.
Wintermantel, An experimental study into the effects of bulk and flow behaviour of
versa) for the powders in our possession, making the shear cell test laser sintering polymer powders on resulting part properties, J. Mater. Process.
flexible in representing different flow behaviours. Technol. 215 (2015) 239–250, https://doi.org/10.1016/j.jmatprotec.2014.07.029.

411
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

[15] C.Y. Wu, A.C.F. Cocks, Flow behaviour of powders during die filling, Powder Metall. [47] L. Marchetti, P. Mellin, C. Hulme-Smith, Effects of humidity on the flowability of
47 (2013) 127–136, https://doi.org/10.1179/003258904225015617. steel powders, Euro PM2020, EPMA, Virtual Congress, 2020.
[16] M. Meyer, R. Lupoi, An analysis of the particulate flow in cold spray nozzles, Mech. [48] J.A.H. de Jong, A.C. Hoffmann, H.J. Finkers, Properly determine powder flowability to
Sci. 6 (2015) 127–136, https://doi.org/10.5194/ms-6-127-2015. maximize plant output, Chem. Eng. Prog. 95 (1999) 25–34.
[17] J.K. Prescott, R.A. Barnum, On powder flowability, Pharm. Technol. 24 (10) (2000) [49] S. Vock, B. Klöden, A. Kirchner, T. Weißgärber, B. Kieback, Powders for powder bed
60–84. fusion: a review, Progr. Addit. Manuf. (2019)https://doi.org/10.1007/s40964-019-
[18] M.V.V. Antequera, A.M. Ruiz, M.C.M. Perales, N.M. Munoz, M.R.J.C. Ballesteros, Eval- 00078-6.
uation of an adequate method of estimating flowability according to powder char- [50] ASTM B855−17 Standard test method for volumetric flow rate of metal powders
acteristics, Int. J. Pharm. 103 (1994) 155–161, https://doi.org/10.1016/0378-5173 using the Arnold meter and hall flowmeter funnel, ASTM Int. doi:https://doi.org/
(94)90096-5. 10.1520/B0855-17.
[19] M. Krantz, H. Zhang, J. Zhu, Characterization of powder flow: static and dynamic [51] ASTM D6393-14 Standard test method for bulk solids characterization by Carr indi-
testing, Powder Technol. 194 (2009) 239–245, https://doi.org/10.1016/j.powtec. ces, ASTM Int. doi:https://doi.org/10.1520/d6393-14.
2009.05.001. [52] ASTM D6128-16 Standard test method for shear testing of bulk solids using the
[20] R. Engeli, T. Etter, S. Hovel, K. Wegener, Processability of different IN738LC powder Jenike shear tester, ASTM Int. doi:https://doi.org/10.1520/D6128-16.
batches by selective laser melting, J. Mater. Process. Technol. 229 (2016) 484–491, [53] Z. Snow, R. Martukanitz, S. Joshi, On the development of powder spreadability met-
https://doi.org/10.1016/j.jmatprotec.2015.09.046. rics and feedstock requirements for powder bed fusion additive manufacturing,
[21] A.B. Spierings, M. Voegtlin, T. Bauer, K. Wegener, Powder flowability characterisa- Addit. Manuf. 28 (2019) 78–86, https://doi.org/10.1016/j.addma.2019.04.017.
tion methodology for powder-bed-based metal additive manufacturing, Progr. [54] A. Santomaso, P. Lazzaro, P. Canu, Powder flowability and density ratios: the impact
Addit. Manuf. 1 (2015) 9–20, https://doi.org/10.1007/s40964-015-0001-4. of granules packing, Chem. Eng. Sci. 58 (2003) 2857–2874, https://doi.org/10.1016/
[22] Z. Berk, Fluid Flow, Food Process Engineering and Technology, Academic Press, US, S0009-2509(03)00137-4.
2018 56–65, https://doi.org/10.1016/C2016-0-03186-8. [55] ASTM D6773-16 Standard test method for bulk solids using Schulze ring shear tes-
[23] M.E. Fayed, L. Otten, Handbook of Powder Science & Technology, Chapman & Hall, ter, ASTM Int. doi:https://doi.org/10.1520/D6773-16.
1997https://doi.org/10.1007/978-1-4615-6373-0. [56] ISO 13517, Metallic Powders – Determination of Flowrate by Means of a Calibrated
[24] G.A. Turner, M. Balasubramanian, Investigations of the contributions to the tensile Funnel (Gustavsson Flowmeter), International Standard Organisation, 2013.
strength of weak particulate masses, Powder Technol. 10 (1974) 121–127, https:// [57] ASTM B329−14 Standard test method for apparent density of metal powders and
doi.org/10.1016/0032-5910(74)80037-9. compounds using the Scott Volumeter, ASTM Int. doi:https://doi.org/10.1520/
[25] M. Muthuswamy, A. Tordesillas, How do interparticle contact friction, packing den- B0329-14.
sity and degree of polydispersity affect force propagation in particulate assemblies? [58] Y. Zhao, R.A. Cocco, S. Yang, J.W. Chew, DEM study on the effect of particle-size dis-
J. Stat. Mech. (2006)https://doi.org/10.1088/1742-5468/2006/09/p09003. tribution on jamming in a 3D conical hopper, AIChE J. 65 (2018)https://doi.org/10.
[26] ASTM F3049−14 Standard guide for characterizing properties of metal powders 1002/aic.16483.
used for additive manufacturing processes, ASTM Int. doi:https://doi.org/10.1520/ [59] ASTM B212−17 Standard test method for apparent density of free-flowing metal
F3049-14. powders using the hall flowmeter funnel, ASTM Int. doi:https://doi.org/10.1520/
[27] D. Schulze, Powder and Bulk Solids, Springer-Verlag, Berlin Heidelberg, 2007https:// B0212-17.
doi.org/10.1007/978-3-540-73768-1. [60] ASTM B417−13 Standard test method for apparent density of non-free-flowing
[28] R. Freeman, X. Fu, Characterisation of powder bulk, dynamic flow and shear proper- metal powders using the carney funnel, ASTM Int. doi:https://doi.org/10.1520/
ties in relation to die filling, Powder Metall. 51 (2013) 196–201, https://doi.org/10. B0417-13.
1179/174329008x324115. [61] ASTM B703−17 Standard test method for apparent density of metal powders and
related compounds using the Arnold meter, ASTM Int. doi:https://doi.org/10.1520/
[29] ASTM D7891−15 Standard test method for shear testing of powders using the Free-
B0703-17.
man technology FT4 powder rheometer shear cell, ASTM Int. doi:https://doi.org/10.
[62] N.H. Harnby, D. Vandame, The use of bulk density determination as a mean of typ-
1520/D7891-15.
ifying the flow characteristics of loosely compacted powders under conditions of
[30] ISO 902, Aluminium Oxide Primarily Used for the Production of Aluminium - Mea-
variable relative humidity, Chem. Eng. Sci. 42 (1987) 879–888, https://doi.org/10.
surement of the Angle of Repose, International Standard Organisation, 1976.
1016/0009-2509(87)80046-5.
[31] ISO 4324, Surface Active Agents - Powders and Granules - Measurement of the
[63] A. Chi-Ying Wong, Characterisation of the fowability of glass beads by bulk densities
Angle of Repose, International Standard Organisation, 1977.
ratio, Chem. Eng. Sci. 55 (2000) 3855–3859, https://doi.org/10.1016/S0009-2509
[32] ASTM C1444-00 Standard test method for measuring the angle of repose of free-
(00)00048-8.
flowing mold powders, ASTM Int. doi:https://doi.org/10.1520/C1444-00.
[64] R.P.Y. Zou, A. B, The packing of spheres in a cylindrical container: the thickness ef-
[33] D.A. Ploof, J.W. Carson, Quality Control Tester to Measure Relative Flowability of
fect, Chem. Eng. Sci. 50 (1995) 1504–1507, https://doi.org/10.1016/0009-2509
Powders, Bulk Solids Handling, 14, 1994.
(94)00483-8.
[34] B.J. E, T.A. Bell, R.J. Grygo, W.J.F. Scholten, M.M. Schenkel, Practical Evaluation of the [65] P. Rousé, Comparison of methods for the measurement of the angle of repose of
Johanson Hang-Up lndicizer, Bulk Solids Handling, 14, 1994. granular materials, Geotech. Test. J. 37 (1) (2014) 164–168, https://doi.org/10.
[35] D. Geldart, E.C. Abdullah, A. Hassanpour, L.C. Nwoke, I. Wouters, Characterization of 1520/GTJ20120144.
powder flowability using measurement of angle of repose, China Particuol. 4 (2006) [66] J. Schwedes, Review on testers for measuring flow properties of bulk solids, Granul.
104–107, https://doi.org/10.1016/s1672-2515(07)60247-4. Matter 5 (2003) 1–43, https://doi.org/10.1007/s10035-002-0124-4.
[36] Y.M. Fouda, A.E. Bayly, A DEM study of powder spreading in additive layer [67] C. Hare, U. Zafar, M. Ghadiri, T. Freeman, J. Clayton, M.J. Murtagh, Analysis of the dy-
manufacturing, Granul. Matter 22 (2020) 10, https://doi.org/10.1007/s10035-019- namics of the FT4 powder rheometer, Powder Technol. 285 (2015) 123–127,
0971-x. https://doi.org/10.1016/j.powtec.2015.04.039.
[37] W. Zhong, A. Yu, X. Liu, Z. Tong, H. Zhang, DEM/CFD-DEM modelling of non- [68] G. Lumay, F. Boschini, K. Traina, S. Bontempi, J.C. Remy, R. Cloots, N. Vandewalle,
spherical particulate systems: theoretical developments and applications, Powder Measuring the flowing properties of powders and grains, Powder Technol. 224
Technol. 302 (2016) 108–152, https://doi.org/10.1016/j.powtec.2016.07.010. (2012) 19–27, https://doi.org/10.1016/j.powtec.2012.02.015.
[38] ASTM B215−15 Standard practices for sampling metal powders, ASTM Int. doi: [69] G.B. Arfken, H.J. Weber, F.E. Harris, Probability and Statistics, Mathematical Methods
https://doi.org/10.1520/B0215-15. for Physicists, 2013 1125–1179, https://doi.org/10.1016/b978-0-12-384654-9.
[39] ASTM B213−17 Standard test methods for flow rate of metal powders using the 00023-2.
hall flowmeter funnel, ASTM Int. doi:https://doi.org/10.1520/B0213-17. [70] K.L. Lange, R.J. A, J. Taylor, Robust statistical modeling using the t-distribution, J. Am.
[40] ASTM B964−16. Standard test methods for flow rate of metal powders using the Stat. Assoc. 84 (2019) 881–896, https://doi.org/10.2307/2290063.
carney funnel. ASTM Int. doi:https://doi.org/10.1520/b0964-16. [71] A.K. Marnani, A. Bück, S. Antonyuk, B. van Wachem, D. Thévenin, J. Tomas, The effect
[41] ASTM B527−15 Standard test method for tap density of metal powders and com- of very cohesive ultra-fine particles in mixtures on compression, consolidation, and
pounds, ASTM Int. doi:https://doi.org/10.1520/B0527-15. fluidization, Processes 7 (7) (2019) 439, https://doi.org/10.3390/pr7070439.
[42] R. Freeman, Measuring the flow properties of consolidated, conditioned and aerated [72] J.A. Muñiz-Lerma, A. Nommeots-Nomm, K.E. Waters, M. Brochu, A comprehensive
powders — a comparative study using a powder rheometer and a rotational shear approach to powder feedstock characterization for powder bed fusion additive
cell, Powder Technol. 174 (2007) 25–33, https://doi.org/10.1016/j.powtec.2006.10. manufacturing: a case study on AlSi7Mg, Materials (Basel) (2018)https://doi.org/
016. 10.3390/ma11122386.
[43] H.M. Beakawi Al-Hashemi, O.S. Baghabra Al-Amoudi, A review on the angle of re- [73] P. Li, Effect of Powder Flow Properties on the Process Performance of Loss-in-weight
pose of granular materials, Powder Technol. 330 (2018) 397–417, https://doi.org/ Feeders, Thesis from the State University of New Jersey (U.S.), Rutgers, 2015https://
10.1016/j.powtec.2018.02.003. doi.org/10.7282/T39025N7.
[44] M.K. Stanford, C. DellaCorte, Effects of humidity on the flow characteristics of a com- [74] A. Janda, I. Zuriguel, A. Garcimartín, L.A. Pugnaloni, D. Maza, Jamming and critical
posite plasma spray powder, J. Therm. Spray Technol. 15 (2006) 33–36, https://doi. outlet size in the discharge of a two-dimensional silo, Europhys. Lett. 84 (2008) 4,
org/10.1361/105996306x92550. https://doi.org/10.1209/0295-5075/84/44002.
[45] L.P.B. Lefebvre, N. Orsoni-Wiemer, C. Charbonneau, B. Alchikh-Sulaiman, S. Yue, [75] J.W. Carson, B.H. Pittenger, Bulk properties of powders, ASM Handbook, Volume 7:
Rheology of powders: assessing the robustness and impact of humidity, Powder Metal Technologies and Applications 1998, pp. 287–301, https://doi.org/
tribocharging, particle size and composition, Euro PM2019, EPMA, Maastricht 10.1361/asmhba0001530.
(NL), 2019. [76] E.C. Abdullah, D. Geldart, The use of bulk density measurements as flowability indi-
[46] L.C. Cordova, T. Tinga, Assessment of moisture content and its influence on laser cators, Powder Technol. 102 (1999) 151–165, https://doi.org/10.1016/S0032-5910
beam melting feedstock, Euro PM2019, EPMA, Maastricht (NL), 2019. (98)00208-3.

412
L. Marchetti and C. Hulme-Smith Powder Technology 384 (2021) 402–413

[77] H. Hou, C.C. Sun, Quantifying effects of particulate properties on powder flow prop- [79] J. Zegzulka, D. Gelnar, L. Jezerska, R. Prokes, J. Rozbroj, et al., Sci. Rep. 10 (2020)
erties using a ring shear tester, J. Pharm. Sci. 97 (9) (2008) 4030–4039, https://doi. https://doi.org/10.1038/s41598-020-77974-3.
org/10.1002/jps.21288. [80] P. Mellin, O. Lyckfeldt, P. Harlin, H. Brodin, H. Blom, A. Strondl, Evaluating flowability
[78] G.S. Riley, G.R. Mann, Effects of particle shape on angles of repose and bulk densities of additive manufacturing powders, using the Gustavsson flow meter, Metal Powder
of a granular solid, Mater. Res. Bull. 7 (1972) 163–169, https://doi.org/10.1016/ Rep. 72 (5) (2017) 322–336, https://doi.org/10.1016/j.mprp.2017.06.003.
0025-5408(72)90273-5.

413

You might also like