Full Paper

Comminution of Polymers ± An Example of Product

Engineering*
By Wolfgang Peukert and Lutz Vogel**

Investigations into the processes of grinding and classification are presented. It is shown how the properties of filled polymer
products like powder coatings or toner can be tailored through grinding and classification. A systematic approach to describe the
process by a machine and a material function is introduced. Single particle comminution tests allow the determination of the
material function. The breakage probability curves of different organic and inorganic materials can be transferred to one single
master curve. The transfer of single particle tests to production mills seems to be possible, at least for the given example that of
powder coatings. This comparison shows that the width of the final size distribution is determined in this example by the material
function and not by the operation of the mill. Significant improvements were achieved for both powder quality, i.e. the amount of
fines below 10 lm, and the yield by employing two new classifiers for the removal of fines. The classifiers were operated inline,
directly after a classifying mill.

1 Introduction

The objective of many processes is the production of

substances with well-defined properties with the considera-
tion of economical and ecological boundary conditions. For
particulate materials the properties of the product depend
amongst other things on the dispersity of the material. The
dispersity is characterized by the particle size, particle size
distribution, particle shape and morphology and the inter-
facial properties. Rumpf [1] called this relation the ªproperty
functionº. The control of the property function is known as
product engineering or product design.
Powder coatings or toner are examples of filled polymers in
a disperse form. Their production often includes comminution Figure 1. Product engineering for the grinding process.
or classifying steps. It is crucial that the particle size is tuned to
a well-defined, usually very narrow size distribution within the
particles higher stresses can be achieved with increasing stress
given minimum and maximum particle size. The maximum
velocity and decreasing temperature. On the other hand the
particle size for powder coatings generally lies between
operating conditions of the mill determine the number of
40±100 lm, whereas for toner it is less than 20 lm. Because
stress events, their intensity and the distribution of both.
fine particles show high adhesive forces they tend to produce
Conditions of operation can be characterized e.g. by the
unwanted agglomerates, therefore a minimum amount of
revolution speed, the airflow rate and the particle concentra-
fines below 5 to 10 lm is striven for.
tion. For impact mills the speed of the hammers or the particle
The question underlying this work is how to produce a
velocity, respectively, correlates with the stress intensity. The
narrow size distribution for this group of materials by
number of stress events depends on the residence time
comminution and if necessary an additional classifying step.
distribution of the particles in the mill. Schwedes et al. have
The basic approach is shown in Fig. 1.
carried out investigations on stress intensity and number for
We start by assuming that a machine and a material function
stirred media mills [2, 3], and sieve hammer mills [4]
can formally describe the comminution process. The machine
respectively.
function comprises the type of mill as well as all the conditions
The design of mills and comminution processes is presently
of operation. By choosing the mill the kind of particle stressing
based on empirical scale-up rules. Costly experiments and
is defined. Choosing the right type of mill depends on the
personal experience are necessary to achieve a suitable mill
material function (amongst other things). Impact mills are
design. A reason for this unsatisfying procedure is a lack of
used for the grinding of viscoelastic materials. Within stressed
understanding of how different materials behave in the
± comminution process, i.e. the material function is unknown.
[*] Presented at the annual GVC meeting at Karlsruhe (Germany), So far no systematic differentiation between the various
22.09.2000.
materials and their grinding performance can be made based
[**] Prof. Dr.-Ing. W. Peukert, Dipl.-Ing. L. Vogel, Institute of Particle
Technology, Technische Universität München, Boltzmannstraûe 15, on reproducible measurements of relevant material proper-
D-85748 Garching, Germany. ties. It would be a considerable improvement if a basic

Chem. Eng. Technol. 24 (2001) 9, Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0930-7516/01/0909-0945
0930-7516/01/0909-0945$$17.50+.50/0
17.50+.50/0 945
 Full Paper

understanding of the influence of material properties on the test, which is industrially feasible. Fig. 2 shows the single
comminution behavior could be achieved. Furthermore, this particle impact device used for the experiments, developed by
would allow a systematic optimization of existing processes. Schönert [5]. A vibration feeder transports single particles
The design of new grinding steps would become more reliable from the feed chamber a) to the disk shaped rotor b). There
and due to a reduction of the milling tests more efficient. the particles enter the rotor centrally and are accelerated
Comminution experiments at well defined and reproducible radially in one of the four radial channels by centrifugal force.
stressing conditions allow the characterization of the grinding Reaching the outer diameter, da, of the rotor they are ejected
performance of different materials. It is possible to determine with a final velocity consisting of a radial and a tangential
a material function independent of apparatus and machine component, both equal to the circumferential speed of the
parameters. This material function will be correlated to bulk rotor. As both the feed and the grinding chamber are
material properties in future studies. evacuated in order to avoid any disturbing influence of air
flows, the impact velocity, vP, of the particles on the target ring
c) is given by the ejecting velocity from the rotor as a function
2 Experimental Set-up of the rotational speed n1):
p
Single particle comminution experiments show some vP ˆ 2da n (1)
advantages, which allow the determination of the unknown
material function. The concentration in single particle tests is For the single particle experiments shown here, impact
small enough to ensure that particle-particle interactions can velocities of 60±140 m/s were realized. At the end of a single
be excluded. The stress intensity (here impact velocity) is well test the size distribution of the product particles was
defined and because of the shape of the target ring, Fig. 2, the determined by sieve size analysis. After the analysis the
impact takes place at the ideal angle of 90. product particles were stressed again under the same
conditions to simulate multiple impacts. This procedure was
repeated up to five times.
For the experiments presented here two different polymers
(polymethyl methacrylate (PMMA G55 and G88) and
polystyrene (PS 144C), both products of BASFAG, Germany)
with different mechanical properties were used. Sieving
produced narrow size fractions of the feed material. In each
experiment approximately 2500 particles were stressed in
order to obtain statistically reliable results.

3 Results and Discussion of the Single Particle

Experiments

As expected, the particle size distribution of the product for

all three materials exhibited an increasing fineness with
increasing impact velocity and increasing impact number. The
total mass specific impact energy, Wm,kin, can be calculated
2
from the mass specific energy of the single impact, 1=2vP , and
the impact number, k, as follows:
v2P
Wm;kin ˆ k
(2)
Figure 2. Single particle impact device, developed by Schönert. 2
For experiments with approximately the same total mass
The results obtained allow the description of the materials
specific impact energy the comminution product showed
grinding properties, being independent of mill type and
approximately the same particle size distribution. Differences
operation conditions. Furthermore the comparison of grind-
only occurred for low impact velocities (below 80 m/s). Here,
ing experiments in production mills with single particle tests as
fewer impacts with higher energy resulted in a finer product
reference leads to the efficiency of the process under
for the same total impact energy. Obviously, in this regime,
investigation. Single particle tests can be used as basic grinding
viscoelastic energy dissipation occurred whereas the commi-
tests for new products. This is of great benefit if only small
nution result for high velocities is strongly determined by the
amounts of sample are available. A certain disadvantage is the
time consuming procedure necessary to perform these tests. ±
Therefore, it is planned to develop a particle comminution 1) List of Symbols at the end of the paper.

946 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0930-7516/01/0909-0946 $ 17.50+.50/0 Chem. Eng. Technol. 24 (2001) 9
 Full Paper

total energy input and only to a lesser extent by the intensity of activated and induce particle breakage (contrary to Wei-
the single impacts. chert's assumption of beginning breakage on the whole
Fig. 3 shows the plot of the breakage portion, PB, versus the surface) for the breakage probability of spheres with diameter
total mass specific impact energy. x one gets:
( !m )
v2P
PB ˆ 1 ÿ exp ÿfMat:
x
(4)
2
The parameter fMat. comprises of the material properties of
the impacting bodies and their geometry. The mass specific
2
energy of the single impact, 1=2vP , in Eq. (4) can be
interpreted as a measure of the external load, which induces
the stress field inside the particle. An extension of Eq. (4) with
the total mass specific impact energy reduced by the energy
threshold, Wm,min, leads to:
  m 
PB ˆ 1 ÿ exp ÿfMat:
x
Wm;kin ÿ Wm;min (5)

The lines in Fig. 3 show the fit of Eq. (5) to the experimental
data. One can see good agreement between experiment and
Figure 3. Breakage portion as a function of total mass specific impact energy. the modified model. The extension of Eq. (4) with the total
impact energy and the energy threshold shows that Weibull
The breakage portion is defined as the mass fraction of statistics not only provide a good description of single impact
material which is smaller than the feed material. events as introduced by Weichert [6] but also of multiple
For each material there exists a characteristic energy input, impacts. This result is not self-evident as it was implicitly
where the breakage portion reaches 100 %. For a lower energy assumed for the derivation of Eq. (5) that the material
input it was realized that a certain minimum energy, Wm,min, properties which determine the material function do not differ
has to be provided in order to achieve noticeable comminu- before and after the impact events. For the material strength
tion. Below this energy threshold no particle breakage occurs. this assumption is obviously not fulfilled, as prestressed
This energy threshold can be seen as the energy that can be particles compared to unstressed ones are broken at lower
stored within the particle elastically without extending impact velocities or give a higher breakage portion at the same
existing cracks or inducing new flaws. The energy threshold impact speed respectively. This relates to a reduced material
is characteristic for the material. For the materials under strength and as a consequence should lead to a different value
investigation here it was between 0.5 and 2 kJ/kg. for the material function. The reasons why the breakage
Weichert [6] introduced Weibull statistics [7] to the field of probability for multiple impacts can be described by a uniform
comminution in order to describe the probability of particle material function cannot be answered completely at the
breakage. The following slightly modified approach is based moment and is the subject of further studies.
on his method. The breakage probability relates to the amount The exponent m of the Weibull distribution is in the range of
of particles which were destroyed in an experiment relative to 1±1.5 for best fit. But as a variation of the exponent in this
the original amount of particles. For regular shapes such as range does not change the distribution significantly we chose
spheres or cubes the breakage can be determined easily by to proceed with an exponent of unity for all materials.
fragments missing from the original shape. For irregular The influence of the particle size, x, can be verified by using
shaped particles or large numbers this cannot reasonably be data from the literature. The breakage probabilities of glass
realized anymore. As an approximation for the breakage spheres in the range of 95 lm up to 8 mm were available from
probability the breakage portion will be used in the following. two authors [9,10]. The significantly different breakage
The Weibull statistic is based on the principle of the weakest probabilities of particles covering nearly two decades in size
link in a chain. It gives the probability, PB, for the fracture of a can be shifted onto one uniform curve if the data are plotted
chain which consists of z links of strength rS when a load, r, is versus the product of mass specific impact energy and particle
applied. size in accordance with Eq. (5). Differences only appear for
( !m ) probabilities close to unity. They are likely to be due to

PB ˆ 1 ÿ exp ÿz
(3) differences in the material used by the two authors.
S
The dependence of the breakage probability on the product
m is a material dependent parameter of the probability of the specific impact energy and the particle size can also be
distribution which can be adjusted to experimental results. derived from an approach first suggested by Rumpf [11]. From
Using Hertz theory for the contact of elastic bodies [8] and dimensional reasoning he concluded that similar fracture
the assumption that only flaws at locations of highest tensile patterns occur for geometrically similar particles of different
stress, i.e. at the circumference of the contact circle, are size if the product of particle size and volume specific impact

Chem. Eng. Technol. 24 (2001) 9, Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0930-7516/01/0909-0947 $ 17.50+.50/0 947
 Full Paper

energy is the same. The assumption that the destruction of distributions were (at least in this example) surprisingly
particles is caused by similar fracture patterns (e.g. a single similar. The amount of fines smaller than 10 lm was equal for
crack dividing the particle in two fragments) is consistent with both experiments within the range of experimental error,
Eqs. (4) and (5), respectively. Fig. 5.
Besides the influence of impact energy and feed particle 1
size Eq. (5) accounts for the grinding properties of impact velocity
different materials in the form of the material parameter v = 100 m/s
or material function, fMat. Young's modulus, Poisson ratio air classifier mill ACM 10
and density of the contacting bodies are accounted for by epoxy polyester
0,75 300 kg/h
the implementation of Hertz's equations and the material
strength and feed particle size due to the Weibull statistics.
Therefore the material function allows for quantitative
differentiation between diverse materials and their grind-

Q 3 [-]
ing performance. 0,5
Fig. 4 shows the breakage probability for the polymers
PMMA G55, G88 and PS 144C, glass spheres [9,10] and
limestone particles [10] versus the product of specific impact
energy, feed particle size and material function. 0,25
single imp act grinding
feed material
1250 µm - 2000 µm
5 grinding steps

0
8 10 100 400
x [µm]
Figure 5. Comparison between classifier mill and single particle test.

The comparison between single particle tests and grinding

experiments under realistic conditions and the results
obtained indicate that single particle tests may be a reliable
tool for the characterization of the quality and efficiency of
real grinding processes.
Due to the idealized impact conditions for single particle
experiments the particle size distribution obtained is the
Figure 4. Master curve for the breakage probability of different materials. narrowest physically possible distribution that can be
achieved by impact comminution. Under process conditions
By varying the material function the plots of these very in a mill narrower distributions cannot be produced. As a
different types of materials can be shifted until they are consequence, the optimization of the mill (at least for the
overlapping. The master curve generated describes the example shown here) does not yield an improvement
breakage probability for all materials and particle sizes. The concerning the width of the size distribution. Narrower
material function is fitted to the results of the comminution distributions can only be achieved if either the mechanical
experiments, and not yet predicted from material bulk material properties of the product are changed or the
properties. For a single material and for the particle size range unwanted amount of fines is removed by an additional
investigated the material function is constant. classifying step.
The latter way was chosen in order to optimize the hardware
side of the process. In standard powder production systems
4 Example: Application to Powder Coatings either cyclones or cyclone classifiers are used after an air
classifier mill. Cyclone classifiers have a similar shape to a
As described above the production of powder coatings aims cyclone. However, the vortex finder is exchanged for an
at the generation of narrow particle size distributions with impeller wheel and additional air is introduced from the
well-defined maximum particle size and small amounts of bottom to enhance the dispersion of the fine material. With
fines. Galk and Peukert [12] showed for one material that the growing tendency towards finer particles, it became
results of single particle experiments could be transferred to evident that the existing cyclone classifier had to be improved
technically relevant mills. The particle size distribution of in order to achieve finer cut sizes and a steeper sharpness of
repeated single particle impact tests was compared with the cut. As a result, Hosokawa Micron developed two new
one obtained for a classifier mill at a throughput of 300 kg/h impeller wheel classifiers, based on particle separation by the
(solids loading: 280 g powder coating per kg air). Both radially opposed centrifugal and drag forces.

948 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0930-7516/01/0909-0948 $ 17.50+.50/0 Chem. Eng. Technol. 24 (2001) 9
 Full Paper

The first classifier (MikroClassifier) is based on the existing 40

cyclone classifier. The MikroClassifier features a shorter main
body design and an additional internal tube for better 35
dispersion of the fine material. This improves the transport
of the fines to the impeller wheel [12]. In the new machine, the
fine particles adhering to the coarse fraction, which moves 30
down the cyclone wall, are redispersed by secondary air close
to the coarse outlet. Separated by the inner tube they are again 25
transported towards the impeller wheel in a defined manner.
The optimum cut size for this classifier is around 10 lm and

Q 3 [-]
above. 20

The range of cut sizes for the second classifier (MikroCut)

reaches 1 lm and below [13]. This high-performance 15
classifier consists of a flat, cylindrical housing in which the
product is led towards the impeller wheel in a well-
10
controlled way by an improved airflow. Feed material and
air enter the classifier tangentially. Whereas the fines leave
the classifier centrally through the impeller wheel, the 5
coarse fraction is deflected by the rotor and removed at the
outer wall. Additionally, secondary air may be led into the
0
classifier at the outer periphery of the housing in order to 1 10
enhance dispersion and to transport fine particles back to x [µm]
the impeller wheel. Figure 7. Reduction of fines in the product by improved classifying.
Both classifiers can be utilized directly after an impact mill,
i.e. inline. The feasibility of both for the production of powder
coatings has been shown experimentally. The new classifiers
exhibit significantly steeper classification curves than pre-
5 Conclusions
viously possible, Fig. 6. The fundamental principles for the product engineering in
the field of comminution have been introduced and discussed.
The process of size reduction was divided into a machine and a
material function. The choice of the mill and its operating
conditions determine the kind of stress, the number of stress
events and the stress intensity. These values are summed up in
the machine function. Up to now no systematic investigation
into the material function was carried out. It was shown that by
using Weibull statistics and the Hertz equation for colliding
bodies that the breakage probability which was determined in
single particle experiments for different materials and
different sizes can be described well by a single master curve.
Furthermore single particle experiments allow the estima-
tionof thenarrowestsizedistribution possible. Thisdistribution
can be interpreted as a physical limit of the grinding process and
the quality of a mill in producing narrow size distributions can
be evaluated by comparison. A narrower size distribution can
only be achieved by variation of the feed materials mechanical
properties or by an additional classifying step. It was shown that
the product quality and the yield could be improved signifi-
Figure 6. Comparison of the separation curves of different classifiers. cantly with the help of an efficient classifying technique.

Fig. 6 shows an increase in the course fraction, v2, (i.e. the

yield) for a feed of 100 kg from 73 wt.-% to 89 and 94 wt.-%, Acknowledgements
respectively. The product lost (27 kg) was reduced by 59 % and
78 %. The significantly improved classification also led to an This project is partly sponsored by the German Federal
improved product quality. In one application the amount of Ministry of Economics and Technology (AiF-FV-Nr. 12680N).
fines below 10 lm in the final product could be reduced from The authors greatly appreciate the support.
19 wt.-% to 10 and 6 wt.-%, respectively, Fig. 7. Received: December 20, 2000 [CET 1336]

Chem. Eng. Technol. 24 (2001) 9, Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0930-7516/01/0909-0949 $ 17.50+.50/0 949
 Full Paper

Symbols used [2] Schwedes, J., Rührwerkskugelmühlen, Preprints GVC-Dezembertagung

ªFeinmahl- und Klassiertechnikº, Köln 1993.
[3] Kwade, A.; Schwedes, J., Wet Comminution in Stirred Ball Mills, KONA
da [m] rotor diameter 15 (1997) pp. 91±102.
[4] Schallnus, H., Zerkleinerungs- und Klassierverhalten in Siebhammer-
fMat. [kJ±1 kg m±1] material parameter mühlen, Dissertation, TU Braunschweig 1986.
k [±] number of stress events [5] Schönert K.; Marktscheffel. M., Liberation of Composite Particles by
m [±] Weibull exponent Single Particle Compression, Shear and Impact Loading, 6. European
Symposium Comminution, Nürnberg 1986.
n [s±1] revolution speed [6] Weichert, R., Anwendung von Fehlstellenstatistik und Bruchmechanik
PB [±] breakage portion/probability zur Beschreibung von Zerkleinerungsvorgängen, Zement-Kalk-Gips 45
vP [m s±1] impact velocity (1992) 1, pp. 1±8.
[7] Weibull, W., A Statistical Distribution Function of Wide Applicability,
Wm,kin [kJ kg±1] total mass specific impact energy J. Appl. Mechanics (1951) 9, pp. 293±297.
Wm,min [kJ kg±1] mass specific energy threshold [8] Hertz, H., Über die Berührung Fester Elastischer Körper, J. Reine und
Angewandte Mathematik 92 (1882) pp. 156±171.
x [m] particle size [9] Behrens, D., Über die Prallzerkleinerung von Glaskugeln und Unregel-
z [±] number of chain links mäûig Geformten Teilchen aus Schwerspat, Kalkstein, Quarzsand im
r [Pa] stress Korngröûenbereich Zwischen 0,1 und 1,5 mm, Dissertation, Universität
Karlsruhe 1964.
rS [Pa] material strength [10] Priemer, J., Untersuchungen zur Prallzerkleinerung von Einzelteilchen,
Dissertation, Universität Karlsruhe 1964.
[11] Rumpf, H., Physical Aspects of Comminution and a New Formulation of
a Law of Comminution, Powder Technol. 7 (1973) pp. 145±159.
[12] Galk, J.; Peukert, W., Production of Narrow Size Distribution in Air
References Classifier Mills (ACM), 9th European Symposium on Comminution,
Albi 1998, pp. 567±576.
[1] Rumpf, H., Über die Eigenschaften von Nutzstäuben, Staub±Reinhal- [13] Galk, J.; Peukert, W.; Krahnen, J., Industrial Classification in a New
tung der Luft 27 (1967) 1, pp. 3±13. Impeller Wheel Classifier, Powder Technol. 105 (1999) pp. 186±189.

_______________________

950 Ó WILEY-VCH Verlag GmbH, D-69469 Weinheim, 2001 0930-7516/01/0909-0950 $ 17.50+.50/0 Chem. Eng. Technol. 24 (2001) 9