Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Chemical Engineering Science xxx (xxxx) xxx

Contents lists available at ScienceDirect

Chemical Engineering Science


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

Analysis of grinding in a spiral jet mill. Part 1: Batch grinding


Mahesh M. Dhakate a, Jyeshtharaj B. Joshi b,c,d, D.V. Khakhar a,⇑
a
Indian Institute of Technology Bombay, Powai, Mumbai 400076, India
b
Institute of Chemical Technology, Mumbai 400019, India
c
Homi Bhabha National Institute, Mumbai 400094, India
d
UPL Ltd., Mumbai 400051, India

h i g h l i g h t s g r a p h i c a l a b s t r a c t

 Detailed study of breakage


unobscured by the complications of Air in
classification of fines. Particles
 Measurements of particle mass
distributions show that a unimodal Air in
feed results in a trimodal product. Air out
 Mean size correlates with the specific
energy input, independent of Air in
Mesh
operating conditions.
 A population balance model based on
three mechanisms of breakage is
gives a match with experiments.
Air in
Batch Air Jet Mill

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

Article history: Jet milling is widely used in several industrial applications to produce fine particles with a narrow size
Received 10 June 2020 distribution. We studied by means of experiments, the influence of grinding pressure, holdup and milling
Received in revised form 28 October 2020 time on the performance of a spiral air jet-mill operating in a batch mode, with respect to the particles.
Accepted 12 November 2020
The experimental results indicated that a uni-modal size distribution of feed particles yields trimodal
Available online xxxx
particle size distributions. The data for variation of the mean particle size and specific surface area with
energy supplied collapsed to a curve for all the cases studied, indicating that mean size is primarily
Keywords:
dependent on the energy supplied. A population balance model for the dynamics of the batch milling pro-
Air jet mill
Batch milling
cess, based on empirical breakage functions is developed, which describes the experimental results quite
Particle size distribution well. The form of the breakage functions provides an insight into the mechanisms of breakage of particles.
Population balance model Ó 2020 Elsevier Ltd. All rights reserved.

1. Introduction mill comprises a cylindrical grinding chamber with tangentially


directed nozzles for the air, as shown schematically in Fig. 1.
Grinding is an important process used for manufacturing fine Particle-particle and particle-wall collisions occur due to the accel-
particles in the chemical, pharmaceutical, ceramic, cosmetic, paint, eration of particles by high-velocity gas jets. The particles are sub-
paper, plastic, food, fine chemical and mining industries (Benz jected to centrifugal and drag forces inside the mill chamber. The
et al., 1996; Midoux et al., 1999; Chamayou and Dodds, 2007). centrifugal force drives the particles towards the periphery of the
The focus of the present study is grinding in a spiral air jet mill, also chamber, whereas the radial component of the gas velocity exerts
referred to as a fluid energy mill (Andrews, 1936). A spiral air jet a drag force on the particles, acting in the direction of the central
outlet. Particles exit from the outlet when the radial drag force
⇑ Corresponding author. overcomes the centrifugal force (Rodnianski et al., 2013). Finer par-
E-mail address: khakhar@iitb.ac.in (D.V. Khakhar). ticles have a greater tendency to exit since the ratio of the drag

https://doi.org/10.1016/j.ces.2020.116310
0009-2509/Ó 2020 Elsevier Ltd. All rights reserved.

Please cite this article as: M.M. Dhakate, J.B. Joshi and D.V. Khakhar, Analysis of grinding in a spiral jet mill. Part 1: Batch grinding, Chemical Engineering
Science, https://doi.org/10.1016/j.ces.2020.116310
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

it is found that the distribution narrows with an increase in grind-


ing pressure (Teng et al., 2009). The optimal nozzle angle, defined
as the nozzle angle at which the smallest particle mean size is
obtained compared to other nozzle angles, was reported to be
around 45° to a radial line by Tuunila and Nystrom (1998) and
Katz and Kalman (2007) and between 30°-38° by Midoux et al.
(1999). The height of the grinding chamber does not significantly
affect the process (Katz and Kalman, 2007; Rodnianski et al.,
2013) and adjusting the classifier tube height and including mod-
ifications can reduce the product particle mean size (Tuunila and
Nystrom, 1998; Kozawa et al., 2012; Rodnianski et al., 2013). The
product size is nearly independent of the feed size due to classifi-
cation, since only particles smaller than the cut size exit the mill
(Dobson and Rothwell, 1969; Muller et al., 1996; Rodnianski
et al., 2013; MacDonald et al., 2016). The air jet mill is suitable
for grinding materials with a wide range of hardness and the effect
of hardness and energy input is related to the size of fragments by
a simple scaling (Vogel and Peukert, 2003; Lecoq et al., 2011).
The efficiency of grinding has been considered primarily in
terms of the specific energy consumption. Midoux et al. (1999)
and MacDonald et al. (2016) found that the specific surface area
correlates well with the specific energy consumption and data
for different operating conditions collapse to a single curve com-
prising two linear segments. The segment at higher input energy
has a lower slope than the segment at lower energy, and is ascribed
to change in the mechanism of fragmentation by Midoux et al.
(1999). MacDonald et al. (2016) explained that the lower slope is
due to particles approaching the grinding limit at higher energies.
Fig. 1. Schematic view showing the geometry of the air jet mill used in the study.
Air enters from the 4 nozzles and exits from the central classifier tube. Dimensions Results are also reported for different gases and finer particles are
in mm are shown, b ¼ 45 . (a) Top view. (b) Side view. obtained for higher molecular weight gases (Zhao and Schurr,
2002).
Gas flow inside the grinding chamber has been studied in detail
force to centrifugal force increases with reducing particle size. using computational fluid dynamics, in the absence of particles
Thus, the air jet mill, simultaneously grinds as well as classifies (Teng et al., 2009; Kozawa et al., 2012; Rodnianski et al., 2013)
the particles. The spiral jet mill has several additional advantages and with particles (Han et al., 2002; Levy and Kalman, 2007;
over other grinding mills which include: production of particles Teng et al., 2011). Experimental characterization of the flow is also
smaller than 10 lm with a narrow size distribution, low tempera- reported (Muller et al., 1996; Katz and Kalman, 2007; Teng et al.,
ture rise due to high gas throughputs, low maintenance since the 2009; Kozawa et al., 2012; Luczak et al., 2018, 2019). These studies
mill has no mechanical moving parts, suitability for particles with provide useful insights into factors affecting the grinding and clas-
a wide range of material hardness and low contamination due to sification processes, for which models have been developed (Eskin
its autogenous action (Muller et al., 1996; Midoux et al., 1999; et al., 1999; Gommeren et al., 2000; Han et al., 2002; Rodnianski
Chamayou and Dodds, 2007). Further, jet mills have a relatively et al., 2013; Brosh et al., 2014,; MacDonald et al., 2016). Coupled
high energy efficiency of 2%-5% of the energy supplied being used CFD and DEM simulations, including particle breakage have been
to create new surfaces (Mebtoul et al., 1996; Prior, 2000; Lecoq carried out by Brosh et al. (2014). The predicted cumulative size
et al., 2003) compared to other mills, such as ball mills, which have distribution of particles is in reasonable agreement with experi-
an energy efficiency ranging from 0.1% to 1% (Brace and Walsh, mental data.
1962; Lowrison, 1974; Ocepec et al., 1986). Population balance models are widely used to predict the parti-
Several prior studies relate to the influence of operating and cle size distribution in grinding processes (Epstein, 1948; Austin,
design parameters on the performance of an air jet mill and a con- 1971; Kapur, 1982) and specifically for grinding in air jet mills
cise review is given by Midoux et al. (1999). We briefly highlight (Gommeren et al., 1996; Nair, 1999; Gommeren et al., 2000;
the major results pertaining to spiral air jet mills here. The primary Teng et al., 2010; Starkey et al., 2014; Bhonsale et al., 2017). The
operating parameters studied are the gas pressure at the inlet of models are based on empirical functions for the breakage rate (b)
the grinding nozzles (grinding pressure) and the mass flow rate and the breakage distribution function (p). Gommeren et al.
of the feed particles (feed rate); the primary design parameters (1996) considered a 3-zone model for an air jet mill using the form
that have been studied are the nozzle angle, height of the grinding of the b; p functions proposed by Epstein (1948). Predictions of the
chamber and the height of the classifier tube. Increase in the grind- model were found to agree with experimental cumulative distribu-
ing pressure results in a reduction in particle mean size (Teng et al., tions using assumed values of the model parameters. Teng et al.
2009; Djokc et al., 2014) and an increase in the specific surface area (2010) used a stochastic approach to model the evolution of the
(Ezerskii et al., 1972; Midoux et al., 1999). Increase in the feed rate particle size distribution in an air jet mill together with a sharp
leads to an increase in the holdup of particles in the mill and con- cut size to determine the size of the particles leaving the mill.
sequently an increase in the mean particle size (Dobson and The breakage functions were obtained from experiments on a sin-
Rothwell, 1969; Tuunila and Nystrom, 1998; Han et al., 2002; gle pass fluid energy mill, in which a stream of particles is
Zhao and Schurr, 2002; Katz and Kalman, 2007; Teng et al., impacted on a surface, using the method of Kapur (1982). Predic-
2009). Particle size distributions have been reported in a few works tions of the model were in good agreement with experimental
(Dobson and Rothwell, 1969; Han et al., 2002; Zhao and Schurr, results of the cumulative distribution of the product with one fit-
2002; Teng et al., 2009; Djokc et al., 2014; Brosh et al., 2014) and ting parameter, which was correlated to the grinding pressure
2
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

and the solid feed rate (Teng et al., 2010). Starkey et al. (2014) car- Analyzer (Horiba-LA 960). After mixing well, three samples are
ried out a similar analysis but the breakage distribution functions taken from the milled material and the size distribution for each
related feed size directly to the product size, hence no model for sample is measured three times yielding 9 measurements of the
the cut size was included. The breakage was assumed to be by size distribution. The measured variation of grinding pressure with
chipping or fragmentation and the breakage distribution functions flow rate (Q) is given in Fig. 2. The pressure has a linear relation
for each were assumed. A breakage rate function was obtained by with the flow rate because of the pressure drop across the mesh,
fitting to experimental data, using a single parameter, which was a which was measured to be DP ¼ 0:2, 0.5, 0.7 and 0.9 bar at
bilinear function of the grinding pressure, feed pressure, feed rate Pg ¼ 1, 2, 3 and 4 bar, respectively. The pressure and volumetric
and particle size. The conditional probability of chipping was also flow rate were monitored during each experiment and no change
obtained by fitting to experimental data in terms of another in pressure was observed during the experiment except at the
parameter, which was a linear function of the grinding pressure, holdup of h ¼ 15 g and at times beyond t ¼ 180 s. Assuming the
feed pressure and feed rate. Again, good agreement was obtained flow in the nozzle to be isentropic and air to be an ideal gas, we
between model predictions and experimental data in terms of a found the velocity in the nozzle to be subsonic at all the flow rates
size histogram with 6 size classes. considered. The velocities at the nozzle exit were estimated to
In this paper, we carry out a detailed analysis of an air jet mill v e ¼ 122 m/s (Ma ¼ 0:36), 166 m/s (Ma ¼ 0:50), 202 m/s
operated in a batch mode in which a specified mass of particles (Ma ¼ 0:62) and 233 m/s (Ma ¼ 0:72) at P g ¼ 1, 2, 3 and 4 bar,
is taken initially and the particles are prevented from exiting the respectively. The corresponding Mach numbers (Ma) are given in
grinding chamber by a fine mesh fixed on the outlet. In this mode, brackets. Similarly, the velocities at the outlet were estimated to
the mechanisms of breakage can be studied independently of the v o ¼ 6:03 m/s, 8.20 m/s, 9.98 m/s and 11.51 m/s at Pg ¼ 1, 2, 3
mechanisms of classification. Further, a small, fixed mass of mate- and 4 bar, respectively.
rial is used in the experiments and this avoids the complication of a Tamil Nadu river sand used in the study. The sand particles are
continuous injection of large particles of the feed, which results in primary particles (not aggregates) and are brittle in nature. Sand
a distribution of residence times of the particles. The effect of sys- particles of mean diameter 1.48 mm and density 2640 kg/m3 are
tem parameters (grinding pressure and particle holdup) on the used as the feed. The grains are angular in shape as shown in
time evolution of the particle size distributions are studied exper- Fig. 3(inset). The histogram of particle weight fractions (wi ; di ) for
imentally. A population balance model is developed to describe the the feed is shown in Fig. 3(a) where wi is the weight fraction of par-
experimental results and gain an insight into the mechanisms of ticles in the diameter range (di1 ; di ). The size class of d varies as
breakage of particles in the system. log ðdiþ1 =di Þ ¼ 0:059. Thus the interval size is constant in a log
scale. The error bars denote the standard error over the nine mea-
2. Experimental details surements of the size distribution. The particle mass distribution is
obtained as
2.1. Experimental setup nðsi Þ ¼ wi =ðdiþ1  di Þ; ð1Þ

Experiments are carried out in a spiral air jet mill shown where si ¼ ðdiþ1 þ di Þ=2 and nðsÞds is the weight fraction of particles
schematically in Fig. 1 along with dimensions. The jet mill com- in the size range (s; s þ ds). In dimensionless form, particle mass dis-
tribution is n ¼ n smax and particle mean size is si ¼ si =smax where
prises a cylindrical grinding chamber, grinding nozzles at the
periphery and a central tube as an outlet for air. Particles are pre- smax ¼ 5 mm is the largest size in the feed. The mass distribution
of the feed (n ðxÞ) is shown in Fig. 3(b) where x ¼ lnðsÞ.
vented from leaving the chamber by a fine mesh fitted on the out-
let tube (Fig. 1). The jet mill is constructed from stainless steel
except the top surface, which is a removable glass plate for visual- 3. Results and discussion
isation. Four grinding nozzles are positioned equidistant from each
other making an angle b ¼ 45° with the normal to the cylindrical 3.1. Particle size distribution
surface, as shown in Fig. 1. The diameter of each converging nozzle
varies from 5 mm to 2 mm. A compressor (7 bar) is used as a high- Fig. 4 shows the time evolution of the particle mass distribution
pressure air source and the inlet air pressure is maintained at a with time, for a grinding pressure (P g ) of 3 bar and a holdup (h) of
constant value in the experiments by using a pressure regulator 10 g, in three different forms. Fig. 4(a) shows the cumulative size
P
and a valve. The volumetric flow rate of air is measured using a distribution (W i ; di ), defined as W i ¼ ij¼1 wj , Fig. 4(b) shows the
rotameter. The outlet is covered with a fine steel mesh with an histogram of the size distribution (wi ; di ) and Fig. 4(c), the mass
opening size of about 100 lm. Two layers of cloth with a fine distribution. The error bars show the standard error for 9 measure-
weave are fixed above the mesh and a collection cloth bag is fixed
on the exit. We found the loss of mass of particles to be negligible
over the time of the experiment. We found the loss of material to
be negligible over the time of the experiment. The maximum loss
of material ( 5%) observed at Pg ¼ 3 bar and h = 15 g and mini-
mum loss (0.2%) at Pg ¼ 1 bar and h = 10 g, which includes han-
dling loss as well.

2.2. Materials and methods

The experimental procedure used is as follows. The chamber is


filled with a specific weight of sand (h) and air at the desired pres-
sure (P g ) is supplied to the nozzles. After a specified time of milling
the air supply is stopped, the top glass plate is removed and the
entire milled material in the chamber is collected for analysis.
The particle size distribution is obtained using a Laser Particle Size Fig. 2. Variation of grinding pressure (P g ) with flow rate (Q).

3
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

Fig. 3. Particle size distribution of sand used in the experiments (feed). (a)
histogram of the weight % of different sizes (wi ; di ). (b) dimensionless particle mass
distribution (nðxÞ), as defined in the text.

ments and are small in all cases. Most details of the size distribu-
tion are obscured in the cumulative size distribution (Fig. 4(a)),
which has been a common way of presenting results in a majority
of previous studies. The histogram gives more details (Fig. 4(b)) but
an accurate picture is provided only by the mass distribution (Fig. 4
(c)), since the variable widths of the size intervals in the histogram
are taken into account. The number distribution is directly related
Fig. 4. Experimental results of particle size distributions for different milling times
to the dimensionless mass distribution, used in the present study. (s) for grinding pressure, P g ¼ 3 bar and holdup, h ¼ 10 g. (a) cumulative weight
The behaviour of the two would be similar, except that the peaks at percent (W i ; di ). (b) weight percent (wi ; di ). (c) dimensionless particle mass
the smaller sizes would be even higher. The distribution indicates  ðxÞ).
distribution (n
that the largest mass density (weight fraction per unit size interval
width) of the particles corresponds to sizes less than 10 lm.
Although the cumulative weight fraction of these particles
(d < 10 lm) is small (less than 5%) at t = 30 s grinding time,
(Fig. 4(a)), this fraction is important since it corresponds to the pro-
duct which exits from the jet mill.
The reproducibility of the results was checked by repeating the
experiment for which results are shown in Fig. 4, using identical
operating conditions. Fig. 5 shows the size distributions for the
two experiments. The match between the two is excellent indicat-
ing a high degree of reproducibility. Error bars are not shown in
this case, for clarity.
Fig. 6 shows the evolution of the particle mass distribution with
time for 3 different grinding pressures (P g ). The evolution in the
three cases is qualitatively similar. Particles in the feed (peak at
d  1.5 mm, corresponding to x  1:2) break to form a second Fig. 5. Comparison of two experimental data sets of particle mass distributions for
different milling times (s) at P g = 3 bar, h ¼ 10 g. Solid line: data from Fig. 4(c).
peak at d  10 lm (x  6) and a third peak at d  2 lm
Dashed line: data from re.peat experiment.
(x  8). The peak heights for the peaks at 10 lm and 2 lm
increase with time, however, the peak positions remain nearly
the same with time and are nearly independent of operating con- results in significant increase in the peak heights at 10 lm and
ditions. In all three cases, the smallest size in the distribution (xt ) 2 lm, indicating the formation of higher fractions of fine particles.
reduces with time and the peak at 10 lm forms first followed by Fig. 7 shows the time evolution of the mass distributions for dif-
the peak at 2 lm. An exception is the case of P g ¼ 1 bar in which ferent holdups keeping other parameters constant. Increase in the
there is no peak formed at 2 lm. Increase in the grinding pressure holdup from 10 g (Fig. 7(a)) to 15 g (Fig. 7(c)), for a grinding pres-
from 1 bar (Fig. 6(a)) to 3 bar (Fig. 6(c)), for a hold-up of 10 g, sure of 3 bar results in a reduction in the peak heights, indicating a
4
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

Fig. 6. Experimental results for the variation of particle mass distributions with Fig. 7. Experimental results for the variation of particle mass distributions with
milling time (s) at different grinding pressures (P g ) and a fixed holdup, h ¼ 10 g. (a) milling time (s) at different holdups (h) and a fixed grinding pressure, P g ¼ 3 bar. (a)
P g ¼ 3 bar. (b) P g ¼ 2 bar. (c) P g ¼ 1 bar. h ¼ 10 g. (b) h ¼ 12:5 g. (c) h ¼ 15 g.

smaller fraction of fine particles. There are no data for h ¼ 15 g and the outlet when the mesh is removed. The fraction is collected
t ¼ 180 s because the fine particles block the mesh at the outlet for an interval of time and is used as the feed. The figure also shows
and continuation of the experiment beyond 120 s is not possible. the mass distribution of this fraction after regrinding for 240 s in
The results presented indicate that the mass distributions of parti- the batch mode. The two distributions are nearly identical indicat-
cles from air jet milling are trimodal except at low pressures. This ing that there is little further fragmentation of the fine particles on
is not apparent in the cumulative distribution (Fig. 4(a)) or the his- regrinding. The results indicate that there exists a limiting particle
togram (Fig. 4(b)). The formation of two peaks at such fine sizes size, sml  5 lm, defined as the mean size of the distributions
that are independent of time and operating conditions is surprising shown in Fig. 9.
and reveals the nature of fragmentation occurring in the system.
The effects of operating conditions on the particle mass distri- 3.2. Particle mean size
butions for a fixed time of grinding (60 s) are shown in Fig. 8.
Increase in the grinding pressure P g , for a holdup of h ¼ 10 g, Fig. 10 shows the influence of operating parameters (holdup
results in the formation of a larger fraction of fines and the smallest and grinding pressure) on the time evolution of the normalised
size obtained also reduces significantly. For an increase in pressure mean size (sm ¼ sm =smax ). The mean size is obtained from
from 1 bar to 4 bar the peak height at d ¼ 10 lm increases by a fac- Z 1
tor of 10 and the smallest particle size reduces from 6.8 lm to sm ¼ sn
 ðs; t Þds ð2Þ
0.75 lm. The grinding pressure clearly has a very significant 0

impact on particle size reduction. Similar trends are seen with and corresponds to the dimensionless weight average diameter. In
reduction in the holdup but the effects are smaller: reduction in all three cases, there is an initial rapid decrease in the mean size fol-
the holdup from 15 g to 7.5 g, for a grinding pressure P g ¼ 3 bar, lowed by a slower decrease at larger times. Increase in pressure
results in the increase in the 10 lm peak height by a factor of results in a significant reduction in the particle mean size but reduc-
1.5 and the smallest particle size reduces from 1 lm to 0.8 lm. tion in holdup has a smaller effect. This is seen more clearly in
As mentioned above, the fraction of particles smaller than Fig. 11 which shows the variation of the normalised mean size with
10 lm constitutes the product and would normally exit from the grinding pressure and holdup for a fixed time of grinding (60 s), cal-
mill. Using this fraction as the feed we carried out regrinding of culated from the particle size distributions shown in Fig. 8. The
this fraction in the mill to check if further size reduction occurs. increase in grinding pressure from 1 bar to 4 bar, at a holdup of
Fig. 9 shows the mass distribution of the fraction obtained from 10 g, yields a fourfold reduction in the particle mean size while
5
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

Fig. 10. Experimental data for the variation of dimensionless particle mean size (sm )
with milling time (t) at different operating conditions.

Fig. 8. Experimental results for the variation with operating parameters of particle
mass distribution at a fixed milling time, t ¼ 60 s. (a) varying grinding pressure (P g )
at constant holdup (h ¼ 10 g). (b) varying holdup (h) at constant grinding pressure
(P g ¼ 3 bar).

Fig. 9. Experimental results for particle mass distributions for grinding of feed with
size s < 10 lm at grinding pressure, P g ¼ 3 bar, holdup, h ¼ 10 g and a milling time
of t ¼ 240 s.

Fig. 11. Variation of dimensionless particle mean size with grinding pressure (P g )
the reduction in holdup from 15 g to 7.5 g, at a grinding pressure of
and holdup (h) for a constant milling time t ¼ 60 s. (a) Grinding pressure (b)
3 bar, yields about a twofold reduction in particle mean size. Holdup.
Fig. 12 shows the variation of the specific surface area (aw ) with
the mean size (sm ). The specific surface area is given by
3.3. Kinetics of grinding

Assuming the grinding process to follow first order kinetics, the


6
aw ¼ ð3Þ variation of the mean size with time is given by (Yang et al., 2007)
qp /s D32
 
smt  sml
ln ¼ kp t ð4Þ
sm0  sml
where qp is particle density, /s is sphericity and D32 (lm) is Sauter
mean diameter. We assume /s  1 here. The data collapse to a sin- where smt is the mean particle size after grinding for time t; kp is the
gle curve and specific surface area increases with reducing mean grinding rate constant and sml is the limiting size, which we assume
size, as expected. The relationship between aw and 1=sm is not lin- to be sml ¼ 103 for all the cases. A plot of Eq. (4) is shown in Fig. 13.
ear, indicating that the distributions are not self-similar, as is evi- Straight lines are obtained for t > 30 s in all cases, indicating that
dent from the distributions shown above. the process is first order beyond an initial phase of rapid breakage.
6
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

QPg t
Ew ¼ ð5Þ
h
where Q is the volumetric flow rate of air, Pg is the grinding pres-
sure, t is the time of grinding and h is the holdup. Fig. 14(a) shows
the variation of mean size of particles (sm ) with specific energy
input (Ew ) for all the experimental data for different pressures, hold-
ups and times of grinding (21 data points). The data collapse to a
single curve indicating that the size is determined primarily by
the energy supplied and not significantly by the specific operating
parameters. The solid line in the figure is a fit of the following
equation
 
sm ¼ sm0  smf exp ðEw =Ec Þ þ smf ; ð6Þ

where Ec and smf are fitting parameters and sm0 ¼ 0:30 is taken to be
the experimental value. We obtain a good fit for Ec ¼ 4:60  103 and
smf ¼ 0:03. smf is significantly larger than sml since all the particles in
the mill are considered when calculating the former as compared to
sml , for which only particles smaller than 10 lm are considered. The
results indicate that the mean particle size decreases exponentially
until a limiting value is reached at large Ew . The variation of specific
surface area with specific energy supplied is shown in Fig. 14(b).
Fig. 12. Variation of the specific surface area (aw ) with dimensionless particle mean
size (sm ) for all different operating conditions and times. The data again collapse to a single line and there is a linear increase
in the specific surface area with energy, independent of the individ-
ual operating parameter values. The solid line in the figure is a fit of
the equation
aw ¼ aw0 þ AEw ; ð7Þ
with aw0 ¼ 0:0017 taken to be the experimental value and
A ¼ 9:66  106 obtained from a least squares fit. The results indi-
cate that there is no optimal set of operating parameters for batch

Fig. 13. Fit of Eq. (4) (lines) to experimental data (symbols). For different operating
parameters, as indicated. The values of the correlation coefficient were greater than
0.97 in all the cases. Fitted values of the rate constant are given in Table 1.

Table 1
Breakage rate constant (kp ) at different operating conditions.

P g (bar) h (g) kp  103 (s1)

1 10.0 2.07
2 10.0 4.12
3 10.0 9.84
3 12.5 8.97
3 15.0 4.69

The rate constants, given in Table 1, increase with increasing grind-


ing pressure and reduce with increasing holdup.

3.4. Energy input


Fig. 14. Variation of (a) dimensionless particle mean size (sm ) and (b) specific
surface area (aw ) with energy supplied for all data spanning different operating
The energy, per unit mass of material being ground, that is sup- conditions and different times. Solid lines are fits to the data. The values of the
plied to the system is given by correlation coefficient were greater than 0.98 for both cases.

7
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

grinding and several combinations of parameters yield the same experimental finding above that fine particles smaller than about
result for the mean particle size and the specific surface area, pro- 10 lm do not break to any significant extent. For conservation of
vided the energy input is the same. A similar result was reported mass, pðs; s0 Þ must satisfy the following relation
by Midoux et al. (1999) as noted above, but for a continuous Z s0
process. pðs; s0 Þds ¼ 1: ð10Þ
In a continuous grinding process, particles smaller than the cut 0

size (10 lm) exit the mill as the product. We thus consider next Although several forms of pðs; s0 Þ have been proposed previ-
the amount of material produced that is less than 10 lm. Fig. 15 ously, none of them are able to describe the experimental distribu-
shows the weight of material less than 10 lm (w10 h) versus the tions obtained. We thus propose new forms for the breakage
total energy supplied (Et ¼ P g Qt), where w10 is the weight fraction function based on the experimental distributions.
of the material with size less than 10 lm. There is a linear increase We cast the population balance equation in dimensionless form
of this fraction with total energy supplied and most of the data col- using the following dimensionless variables: t ¼ t a; b ¼ b=a; s ¼
lapses to a single curve, except for the data Pg ¼ 2 bar, h ¼ 10 g, s=smax ; n ¼ nsmax ; p ¼ psmax . The governing equation (Eq. (8)) then
and Pg ¼ 3 bar, h ¼ 12:5 g. Fig. 15 indicates that the energy reduces to
required for producing a fixed mass of particles less than 10 lm Z
in size is the least for a grinding pressure of 2 bar and a holdup @n   1  
¼ bðsÞn s; t þ bðs0 Þn s0 ; t pðs; s0 Þds0 ð11Þ
of 10 g. This needs to be examined further. @t s

ðsÞ ¼ sc for s > s3


and the initial condition is nðs; 0Þ ¼ n0 ðsÞ, with b
4. Population balance model 
and bðsÞ ¼ 0 otherwise. Since the particle sizes span a wide range,
we define x ¼ lnðsÞ for computational convenience, and write Eq.
We consider here a population balance model to predict the (11) in terms of x as
evolution with time of the particle size distribution in the milling Z
chamber for a batch milling process, assuming that agglomeration @n   x   0
¼ bðxÞn x; t þ bðx0 Þn x0 ; t pðx; x0 Þ exp ðx0 Þdx ð12Þ
of particles is absent. The population balance equation to predict @t 0
the evolution of the mass fraction distribution, nðs; t Þ, is then given ðxÞ ¼ expðcxÞ for x > x3 and b ðxÞ ¼ 0 other-
with nðx; 0Þ ¼ n0 ðxÞ and b
by (Randolf and Larson, 1971; Ramkrishna, 2000)
wise. As formulated, the model has two parameters (a and c), in
Z
@n smax
0 addition to the parameters of the breakage function, pðx; x0 Þ consid-
¼ bðsÞnðs; t Þ þ bðs0 Þpðs; s0 Þnðs0 ; tÞds ð8Þ
@t s
ered below.
Breakage is a complex process, hence obtaining a theoretical
where bðsÞ is the rate of breakage of particles of size s and pðs; s0 Þ is expression for pðx; x0 Þ is not feasible and we take an empirical
the probability density of forming a particle of size s upon breakage approach. Recent studies on impact breakage show that in this case
of a particle of a larger size s0 , referred to here as the breakage func- the fragments formed on breakage of a particle are independent of
tion. We use the following expression for the breakage rate (Epstein, the size of the particle (Tavares, 2004; Portnikov and Kalman,
1948), 2019). Our preliminary measurements on single particle breakage
 c in an air jet mill indicate a two stage breakage process in which the
s
bðsÞ ¼ a ð9Þ particle first breaks into 2–3 large fragments and then into more
smax
large fragments and numerous very fine fragments
where a and c are fitting parameters, s is the particle size, and smax is (Venkatraman, 2020). Large particles also experience chipping
the largest size. We further assume bðsÞ ¼ 0 for s < s3 , based on the and erosion. The breakage function for the first process is clearly
dependent on the size of the particle that breaks, while the break-
age functions for the formation of fine particles due to impact and
erosion are independent of the parent particle size. Here we pro-
pose simple functions to broadly describe these processes.
Consider the first stage when the large particles break into a few
0
large fragments in the size range (fs ; s0 ), where s0 is the size of the
original particle and the factor f 1 < 1 defines the range. Assuming
the distribution of sizes of fragments formed on breakage in the
 0 
specified range to be uniform we have p1 ¼ c1 for s 2 fs ; s0 , where
s is the size of a fragment. Writing the above equation in terms of
x ¼ lnðsÞ, we have

c1 x 2 ðx0 þ ln f ; x0 Þ
p1 ðx; x0 Þ ¼ ð13Þ
0 otherwise:
The normalizing factor in this case is
Z x0
I1 ¼ c1 expðxÞdx ¼ c1 expðx0 Þ½f 2  f :
x0 þln f

We consider next the peak at x  6 (d ¼ 10 lm). The peak


does not shift significantly during process, hence we assume the
component of pðx; x0 Þ contributing to this peak to be independent
of the particle size (x0 ) and of the form

c2 ðx3  xÞðx  x2 Þ x 2 ðx2 ; x3 Þ
p2 ðxÞ ¼ ð14Þ
Fig. 15. Variation of the particle mass less than 10 lm (w10 h) with total energy 0 otherwise:
supplied (Et ) for all data.

8
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

The normalizing factor in this case is


I2 ¼ c2 ½expðx3 Þðx3  x2  2Þ þ expðx2 Þðx3  x2 þ 2Þ:
The mode of breakage resulting in fragments with a distribution
p2 is most likely due to erosion and chipping of larger particles.
The smallest particle size (xt ) corresponding to the peak at
x  8 reduces with time. We consider a breakage function of
the following form to take this into account

c3 expðrxÞ x 2 ðxt ; x0 Þ
p3 ðx; x0 Þ ¼ ð15Þ
0 otherwise;
where
xt ¼ x1 þ ðx2  x1 Þ expða1 tÞ: ð16Þ
The normalizing factor for this case is
c3
I3 ¼ ½exp fð1  rÞgx0  exp fð1  rÞgxt :
1r
The fragments corresponding to the distribution p3 are most
likely fines formed by impact breakage of larger particles. Multiple
impacts would result in the build up of cracks in each particle
(Brosh et al., 2014) and thus the size of smallest fragments (xt )
reduces with time. The distribution is independent of the parent
particle size, as observed in impact breakage (Tavares, 2004;
Portnikov and Kalman, 2019).
The composite breakage function is then
p1 þ p2 þ p3
pðx; x0 Þ ¼ ; ð17Þ
I1 þ I2 þ I 3
which satisfies Eq. (10) for all parameters values.
Eq. (12) is solved numerically along with the assumed breakage
rate and breakage function. The Euler method is used for the time
integration and the trapezoidal rule is used for calculating the inte-
gral in Eq. (12). The total mass fraction given by
Z smax Z 1 Fig. 16. Evolution of particle mass distribution for t = 1.5,3,6,9. The dashed line
N¼ nðsÞds ¼ nðxÞ expðxÞdx ð18Þ shows the initial distribution. (a) Breakage into large fragments
smin xmin (f ¼ 0:6; f 2 ¼ 1:0; c1 ¼ 1:0; c2 ¼ 0; c3 ¼ 0). (b) Breakage function p2 ðxÞ alone
(c1 ¼ 0; c2 ¼ 1:0; c3 ¼ 0). (c) Breakage function p3 ðxÞ alone (c1 ¼ 0; c2 ¼ 0; c3 ¼ 1:0).
is computed at specified time intervals, as a check of the numerical
accuracy of the method. For conservation of mass we must have
N ¼ 1. Small enough values of the size interval (dx) and the time
interval (dt) are taken to ensure that the error in the total mass frac- we make several assumptions to reduce the number of parameters
tion (N) is less than 1%. We first show the contribution of each com- that have to be fitted at a time. Initially, a number of simulations
ponent of the breakage function, individually, on the evolution of were carried out with different combinations of the parameter val-
the particle size distribution and then discuss how the parameters ues to establish reasonable ranges of each parameter that yielded
are obtained from the experimental data to enable comparison of results similar to the experimental distributions. Based on previous
the model predictions to experimental results. results (Gommeren et al., 1996; Bhonsale et al., 2017) and on the
initial simulations, we take the exponent in the breakage rate to
4.1. Contributions of components of pðx; x0 Þ be c ¼ 1. Initial fitting of the model to the data indicated that
  and the coefficient of the breakage rate (a) is related to the break-
Fig. 16(a) shows the variation of n x; t using only p1 ðx; x0 Þ cor- age rate constant as a ¼ Ckp with C ¼ 11:66 and we use these val-
responding to the initial breakage of large particles. The peak shifts ues in the fitting. x1 ; x2 , and x3 are directly obtained from the
downward and to the left, as in the case of experiments. The evo- experimental distributions: x1 corresponds to the smallest particle
lution of the distribution using the breakage function p2 ðxÞ is size and x2 and x3 are obtained based on the peak positions. a1 is
shown in Fig. 16(b). The shape of the distribution is parabolic sim- obtained from the variation of xt with time, shown in Fig. 17, by fit-
ilar to p2 ðxÞ, since there is no breakage for x < x3 . The evolution of ting Eq. (16) to the data. The fitted values of a1 along with x1 ; x2 and
the distribution using the breakage function p3 ðxÞ, the contribution x3 are given in Table 2.
due to impact breakage, is shown in Fig. 16(c). The edge of the dis- The procedure used to fit the remaining parameters is as fol-
tribution shifts to the left with time until the lowest value is lows. First f is adjusted to match the spreading of the peak at
achieved; the fragments span a large range of sizes. d ¼ 1:5 mm in the w  d graph. Using f ¼ 0:6 gives a reasonable
fit in all cases. The parameter r (Eq. (15)) relates to the rate of decay
4.2. Comparison between the experiments and population balance of the component of the breakage function p3 . In the initial fitting
model we found r  0:7 for all the cases, hence we fix r ¼ 0:7 in the fitting
procedure. The remaining parameters, c1 ; c2 and c3 , determine the
In addition to a and c, the model parameters arising from the relative contributions of each component to the breakage function.
choice of pðx; x0 Þ are: x1 ; x2 ; x3 ; c1 ; c2 ; c3 ; f ; r and a1 . A direct fitting Due to the normalization condition we can specify c1 ¼ 1, without
of all the parameters to experimental data is not feasible. Hence loss of generality. We then adjust c2 and c3 to match the heights of
9
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

c1 ; c2 and c3 give the contributions of the different breakage mech-


anisms. In all cases we see that p2 (erosion and chipping) has the
largest contribution (by weight) followed by p1 (initial breakage)
and then p3 (impact breakage). The relative contribution of p1
reduces with increase in pressure and reduction in holdup, while
those of p2 and p3 increase.

5. Conclusions

A detailed experimental study of the breakage of particles in a


batch spiral jet mill is presented. Operating in the batch mode, in
which a fixed mass of material remains in the mill, allows a study
of the breakage mechanisms unobscured by the complications of
fine particles exiting due to classification and large particles enter-
ing as the feed, as in a continuous system. Further, the batch mode
is easier to operate and uses only a small amount of material com-
pared to the continuous mode. The work demonstrates that such
experiments are repeatable to a high degree of accuracy and would
Fig. 17. Variation of smallest particle size (xt ) with milling time (t) for different be useful to generate data over a wide range of operating condi-
operating parameters.
tions for the selection of optimal parameters for the highest grind-
ing efficiency.
the peaks of the experimental data. The values of c2 and c3 that give Experimental particle mass distributions, which have small
the least square error are given in Table 2. standard errors and are shown to be reproducible based on repeat
Fig. 18 shows a comparison of the model predictions to exper- experiments, are reported for different holdups, grinding pressures
imental data for different grinding pressures and holdups. In addi- and times of grinding. We note that the particle mass distributions,
tion to the mass distributions, the histograms for the weight in the present case, are for all particles in the mill in contrast to
previous studies, in which the distributions are usually for the pro-
fraction (wi ; di ) are also shown, since the larger size fractions are
duct exiting the mill. The distributions reported in this work reveal
more clearly visible in the latter case. The agreement between
greater details than cumulative distributions or histograms, which
model predictions and experimental results is reasonably good in
are the usual tools for reporting particle size distributions. In par-
all cases, particularly for the mass densities at smaller sizes. The
ticular, a unimodal feed is shown to result in trimodal distribu-
largest deviations are seen for the largest sizes in the histograms.
tions, implying the existence several distinct mechanisms of
The deviations are due to the initial rapid decrease in size (see
breakage that occur in parallel.
Fig. 13), which is not included in the model. Fig. 19 shows a com-
The size of the smallest particles is found to reduce with time of
parison between model and experimental data for the mean parti-
grinding and with increasing grinding pressure. The fraction of fine
cle size. Again the match is quite good for all cases, except at short
particles are also found to increase with increasing grinding time
times, for the same reason given above.
and inlet pressure and with reducing holdup. The grinding pres-
The fitting of the model to the experimental data, using several
sure has a more significant effect compared to the holdup for the
simplifications, indicates that assumptions made with regard to
range of parameters studied. The weight averaged mean particle
the breakage rate (bðsÞ) and the simple functions chosen as compo-
size and the specific surface area follow similar trends. The reduc-
nents of the composite breakage function (pðs; s0 Þ), capture the
tion in the mean particle size with time is found to follow first
essential features of the experimental distributions and give a rea-
order kinetics, with the rate constant increasing with increasing
sonably good quantitative predictions, for the range of operating
grinding pressure and reducing holdup. The data for the mean par-
conditions studied. The breakage rate (Eq. (9)) gives a good
ticle size for all the cases studied collapse to a single curve when
description of the process when c ¼ 1 and a ¼ 11:66kp . The param-
plotted versus the specific energy for grinding. The data for specific
eters f and r, which control the size range of the functions p1 (initial
surface versus specific energy also collapse to a single curve, a
breakage) and p3 (impact breakage), are found to be nearly inde-
straight line. Correlations for the mean size and specific area in
pendent of the operating conditions. The parameters defining the
terms of the specific energy input are presented. These results
positions of the peaks (x1 ; x2 and x3 ) reduce with increasing pres-
imply that size reduction in the mill depends primarily on the
sure and saturate at the higher pressures. The parameter which
specific energy supplied to the mill and is nearly independent of
determines the rate of decrease of the size of the smallest particles
the operating parameters. However, when only the product (parti-
(a1 ) increases with pressure and reduces with holdup, except for
cles smaller than 10 lm) is considered, grinding is more energy
the case at Pg ¼ 1 bar. A similar trend with the variation of operat-
efficient for some combinations of operating parameters. The
ing parameters is seen for c2 and c3 , which increase with grinding
results indicate that the energy required for producing a fixed mass
pressure and reduce with increasing holdup. For a given set of
of particles less than 10 lm in size is the least for a grinding pres-
operating conditions, the relative magnitudes of the constants
sure of 2 bar and a holdup of 10 g.

Table 2
Values of the fitted parameters for the different operating conditions. In addition, c ¼ 1; a ¼ 11:66kp ; f ¼ 0:6 and r ¼ 0:7 for all cases.

P g (bar) h (g) x1 x2 x3 a1 (s1) c2 c3

1 10 7.1 6.3 4.7 0.017 3.6 0.030


2 10 8.9 7.1 4.9 0.012 5 0.045
3 10 8.9 7.3 5.3 0.024 10 0.050
3 12.5 8.9 7.3 5.3 0.021 9 0.050

10
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

Fig. 18. Evolution of particle mass distributions (left column) and weight fraction histograms (right column) for the composite breakage function. Continuous lines with error
bars show the experimental results at different pressure (P g ) and holdup (h). Dashed lines show the simulation results. (a,e) P g ¼ 1 bar, h ¼ 10 g; (b,f) P g ¼ 2 bar, h ¼ 10 g; (c,
g) P g ¼ 3 bar, h ¼ 10 g; (d,h) P g ¼ 3 bar, h ¼ 12:5 g.

Finally, a population balance model is presented based on


empirical models for the breakage function, considering three
mechanisms of breakage. The model gives reasonably good predic-
tions of the time evolution of the particle mass distributions (nðxÞ)
and histograms (wðdÞ) for different operating parameters. The
breakage rate is assumed to be proportional to particle size
(c ¼ 1) and the rate of breakage parameter (a) is found to be
directly proportional to the first order rate constant (kp ) for reduc-
tion in the mean particle size. The parameters specifying the size
ranges of the distributions (x1 ; x2 ; x3 and a1 ) are obtained from
the experimental data. Parameters which define the range of the
breakage function for initial breakage (f) and for impact breakage
(r) are found to be nearly independent of the operating conditions.
The magnitudes of the parameters for the breakage functions, c1 ; c2
and c3 , give an insight into the relative contributions of the differ-
ent mechanisms of breakage. The results indicate that the largest
contribution (by weight) is by erosion and chipping followed by
the initial breakage and then impact breakage. The relative contri-
bution of initial breakage reduces with increase in grinding pres-
sure and decrease in holdup.
The detailed characterization of particle breakage presented in
Fig. 19. Evolution of particle mean size for the composite breakage function. Dots this work forms the basis of the next stage in which the mill is
show the experimental results at different operating conditions. Dashed lines operating in the semi-batch mode. In this mode breakage and clas-
shows the simulation results.

11
M.M. Dhakate, J.B. Joshi and D.V. Khakhar Chemical Engineering Science xxx (xxxx) xxx

sification are active and results of the study will form Part 2 of this Han, T., Kalman, H., Levy, A., 2002. Dem simulation of particle comminutionin jet
milling. Particulate Sci. Technol. 20, 325–340. https://doi.org/10.1080/
work and will include the effect of different particle types. The
02726350216184.
overall objective of the work is to obtain a detailed quantitative Kapur, P., 1982. An improved method for estimating the feed-size breakage
understanding of the important processes in an air jet mill to distribution functions. Powder Technol. 33, 269–275. https://doi.org/10.1016/
enable selection of operating parameters to achieve the highest 0032-5910(82)85066-3.
Katz, A., Kalman, H., 2007. Preliminary experimental analysis of a spiral jet mill
efficiency of grinding. performance. Particle Particle Syst. Characteriz. 24, 332–338. https://doi.org/
10.1002/ppsc.200601114.
Kozawa, K., Seto, T., Otani, Y., 2012. Development of a spiral-flow jet mill with
CRediT authorship contribution statement improved classification performance. Adv. Powder Technol. 23, 601–606.
https://doi.org/10.1016/j.apt.2011.06.008.
Mahesh M. Dhakate: Conceptualization, Data curation, Formal Lecoq, O., Chamayou, A., Dodds, J.A., Guigon, P., 2011. Application of a simplifying
model to the breakage of different materials in an air jet mill. Int. J. Miner.
analysis, Investigation, Methodology, Validation, Visualization,
Process. 99, 11–16. https://doi.org/10.1016/j.minpro.2012.04.004.
Writing - original draft. J.B. Joshi: Conceptualization, Funding Lecoq, O., Chouteau, N., Mebtoul, M., Large, J.F., Guigon, P., 2003. Fragmentation by
acquisition, Supervision, Writing - review & editing. D.V. Khakhar: high velocity impact on a target: a material grindability test. Powder Technol.
133, 113–124. https://doi.org/10.1016/j.minpro.2011.01.003.
Conceptualization, Data curation, Formal analysis, Funding acquisi-
Levy, A., Kalman, H., 2007. Numerical study of particle motion in jet milling. Part.
tion, Investigation, Methodology, Project administration, Sci. Technol. 25, 197–204. https://doi.org/10.1080/02726350701257618.
Resources, Supervision, Writing - review & editing. Lowrison, G., 1974. Crushing and grinding, Butterworth’s. London, England.
Luczak, B., Müller, R., Kessel, C., Ulbricht, M., Schultz, H.J., 2019. Visualization of flow
conditions inside spiral jet mills with different nozzle numbers–analysis of
Declaration of Competing Interest unloaded and loaded mills and correlation with grinding performance. Powder
Technol. 342, 108–117. https://doi.org/10.1016/j.powtec.2018.09.078.
The authors declare that they have no known competing finan- Luczak, B., Müller, R., Ulbricht, M., Schultz, H.J., 2018. Experimental analysis of the
flow conditions in spiral jet mills via non-invasive optical methods. Powder
cial interests or personal relationships that could have appeared Technol. 325, 161–166. https://doi.org/10.1016/j.powtec.2017.10.048.
to influence the work reported in this paper. MacDonald, R., Rowe, D., Martin, E., Gorringe, L., 2016. The spiral jet mill cut size
equation. Powder Technol. 299, 26–40. https://doi.org/10.1016/j.
powtec.2016.05.016.
Acknowledgments Mebtoul, M., Large, J., Guigon, P., 1996. High velocity impact of particles on a
target—an experimental study. Int. J. Miner. Process. 44, 77–91. https://doi.org/
The authors express their gratitude to the Prime Minister’s Fel- 10.1016/0301-7516(95)00020-8.
Midoux, N., Hošek, P., Pailleres, L., Authelin, J., 1999. Micronization of
lowship Scheme for Doctoral Research, India, for financial support
pharmaceutical substances in a spiral jet mill. Powder Technol. 104, 113–120.
and to UPL Ltd. (India) for their financial support and active partic- https://doi.org/10.1016/S0032-5910(99)00052-2.
ipation in the Prime Minister’s Fellowship Scheme. Financial sup- Muller, F., Polke, R., Schadel, G., 1996. Spiral jet mills: hold up and scale up. Int. J.
Miner. Process. 44, 315–326. https://doi.org/10.1016/0301-7516(95)00042-9.
port of SERB, India (Grant No. SR/S2/JCB-34/2010) is gratefully
Nair, P.R., 1999. Breakage parameters and the operating variables of a circular fluid
acknowledged by DVK. energy mill: Part I. Breakage distribution parameter. Powder Technol. 106, 45–
53. https://doi.org/10.1016/S0032-5910(99)00067-4.
References Ocepec, D., Salatic, D., Grujic, M., 1986. Energy saving in the comminution
processes. In: World Congress Particle Technology, Part II. Comminution.
Portnikov, D., Kalman, H., 2019. Material comminution functions of wet particles.
Andrews, N.H., 1936. Method of and apparatus for providing material in finely Powder Technol. 343, 29–39. https://doi.org/10.1016/j.powtec.2018.11.009.
divided form. US Patent 2,032,827. Prior, M., 2000. Size reduction. Kirk-Othmer Encycl. Chem. Technol. https://doi.org/
Austin, L.G., 1971. Introduction to the mathematical description of grinding as a rate 10.1002/0471238961.1909260516180915.a01.
process. Powder Technol. 5, 1–17. https://doi.org/10.1016/0032-5910(71) Ramkrishna, D., 2000. Theory and applications to particulate systems in
80064-5. engineering. Population Balances. Academic Press, San Diego.
Benz, M., Herold, H., Ulfik, B., 1996. Performance of a fluidized bed jet mill as a Randolf, A.D., Larson, M.A., 1971. Theory of Particulate Processes. In: The
function of operating parameters. Int. J. Miner. Process. 44, 507–519. https:// Population Balance. Academic Press. https://doi.org/10.1016/B978-0-12-
doi.org/10.1016/0301-7516(95)00062-3. 579650-7.50008-7 (Chapter 3).
Bhonsale, S., Telen, D., Stokbroekx, B., Van Impe, J., 2017. Towards quality by design Rodnianski, V., Krakauer, N., Darwesh, K., Levy, A., Kalman, H., Peyron, I., Ricard, F.,
in pharmaceutical manufacturing: modelling and control of air jet mills. In: EPJ 2013. Aerodynamic classification in a spiral jet mill. Powder Technol. 243, 110–
Web of Conferences, EDP Sciences, p. 07003. https://doi.org/10.1051/epjconf/ 119. https://doi.org/10.1016/j.powtec.2013.03.018.
201714007003. Starkey, D., Taylor, C., Morgan, N., Winston, K., Svoronos, S., Mecholsky, J., Powers,
Brace, W., Walsh, J., 1962. Some direct measurements of the surface energy of K., Iacocca, R., 2014. Modeling of continuous self-classifying spiral jet mills part
quartz and orthoclase. Am. Mineral. J. Earth Planetary Mater. 47, 1111–1122. 1: Model structure and validation using mill experiments. AIChE J. 60, 4086–
Brosh, T., Kalman, H., Levy, A., 2014. Accelerating cfd–dem simulation of processes 4095. https://doi.org/10.1002/aic.14642.
with wide particle size distributions. Particuology 12, 113–121. https://doi.org/ Tavares, L., 2004. Optimum routes for particle breakage by impact. Powder Technol.
10.1016/j.partic.2013.04.008. 142, 81–91. https://doi.org/10.1016/j.powtec.2004.03.014.
Brosh, T., Kalman, H., Levy, A., Peyron, I., Ricard, F., 2014. Dem–cfd simulation of Teng, S., Wang, P., Zhang, Q., Gogos, C., 2011. Analysis of fluid energy mill by gas-
particle comminution in jet-mill. Powder Technol. 257, 104–112. https://doi. solid two-phase flow simulation. Powder Technol. 208, 684–693. https://doi.
org/10.1016/j.powtec.2014.02.043. org/10.1016/j.powtec.2010.12.033.
Chamayou, A., Dodds, J.A., 2007. Air jet milling. Handbook Powder Technol. 12, 421– Teng, S., Wang, P., Zhu, L., Young, M.W., Gogos, C.G., 2009. Experimental and
435. https://doi.org/10.1016/S0167-3785(07)12011-X. numerical analysis of a lab-scale fluid energy mill. Powder Technol. 195, 31–39.
Djokc, M., Djuris, J., Solomun, L., Kachrimanis, K., Djuric, Z., Ibric, S., 2014. The https://doi.org/10.1016/j.powtec.2009.05.013.
influence of spiral jet-milling on the physicochemical properties of Teng, S., Wang, P., Zhu, L., Young, M.W., Gogos, C.G., 2010. Mathematical modeling
carbamazepine form iii crystals: quality by design approach. Chem. Eng. Res. of fluid energy milling based on a stochastic approach. Chem. Eng. Sci. 65,
Des. 92, 500–508. https://doi.org/10.1016/j.cherd.2013.09.011. 4323–4331. https://doi.org/10.1016/j.ces.2010.03.017.
Dobson, B., Rothwell, E., 1969. Particle size reduction in a fluid energy mill. Powder Tuunila, R., Nystrom, L., 1998. Effects of grinding parameters on product fineness in
Technol. 3, 213–217. https://doi.org/10.1016/0032-5910(69)80080-X. jet mill grinding. Miner. Eng. 11, 1089–1094. https://doi.org/10.1016/S0892-
Epstein, B., 1948. Logarithmico-normal distribution in breakage of solids. Industr. 6875(98)00095-8.
Eng. Chem. 40, 2289–2291. https://doi.org/10.1021/ie50468a014. Venkatraman, A., 2020. Breakage of a single particle in an air jet mill. Bachelor of
Eskin, D., Voropayev, S., Vasilkov, O., 1999. Simulation of jet milling. Powder Technology project, Department of chemical engineering, Indian Institute of
Technol. 105, 257–265. https://doi.org/10.1016/S0032-5910(99)00146-1. Technology, Bombay.
Ezerskii, M., Travina, L., Eidelshtein, S., 1972. Use of a jet mill to obtain powders of Vogel, L., Peukert, W., 2003. Breakage behaviour of different materials—construction
pharmaceutical preparations. Pharm. Chem. J. 6, 681–684. https://doi.org/ of a mastercurve for the breakage probability. Powder Technol. 129, 101–110.
10.1007/bf00771308. https://doi.org/10.1016/S0032-5910(02)00217-6.
Gommeren, H., Heitzmann, D., Kramer, H., Heiskanen, K., Scarlett, B., 1996. Dynamic Yang, W., Kwan, C., Ding, Y., Ghadiri, M., Roberts, K., 2007. Milling of sucrose.
modeling of a closed loop jet mill. In: Comminution 1994. Elsevier, pp. 497–506. Powder Technol. 174, 14–17. https://doi.org/10.1016/j.powtec.2006.10.014.
https://doi.org/10.1016/B978-0-444-82440-0.50045-6. Zhao, Q.Q., Schurr, G., 2002. Effect of motive gases on fine grinding in a fluid energy
Gommeren, H., Heitzmann, D., Moolenaar, J., Scarlett, B., 2000. Modelling and mill. Powder Technol. 122, 129–135. https://doi.org/10.1016/S0032-5910(01)
control of a jet mill plant. Powder Technol. 108, 147–154. https://doi.org/ 00408-9.
10.1016/S0032-5910(99)00213-2.

12

You might also like