You are on page 1of 7

Part. Part. Syst. Charact.

19 (2002) 5 ± 11 5

Measurement of Granule Attrition and Fatigue in a Vibrating Box

Willem J. Beekman 1, Gabrie M. H. Meesters1,2, Brian Scarlett 3, Todd Becker 4


(Received: 7 June 2001; resubmitted: 1 September 2001)

Abstract

This paper describes a new test machine that has been particle-particle interaction negligible but is sufficiently
designed to measure the strength of single particles in large that the distribution of these characteristics can
the size range of 102 ± 103 mm. The device is a vibrating be determined.
box that subjects each particle in the sample to a large The instrument has wide application to particle systems
number of impacts of known but variable strength. By where breakage occurs, whether desired or not. It was
tracking the size and shape of the particles as a function particularly developed to test granules of different and
of the number of impacts, their strength characteristics complex structure. Some experimental results are
against the mechanisms of fracture, fatigue and attri- presented in this paper to illustrate the test×s effective-
tion can be differentiated. The number of particles ness in this regard.
tested in one sample is restricted in order to make any

Keywords: attrition test, impact, orientation, reproducibility, fatigue, enzyme granules

1 Introduction particular material as a function of the stress exerted,


independent of the particular equipment which gener-
When a particle is subjected to a mechanical stress, it is ates that stress. These tests have been frequently
liable to breakage. It may be that the breakage is reviewed, for example by Bembrose and Bridgewater
required, in which case we call the process comminution, [1]. They always divide conveniently into bulk tests and
but if it is not required, we will probably call it attrition. single particle tests. A bulk test measures the average
Whether desired or not, the breakage behavior is strength of a sample of particles; the single particle tests
dependent upon both the stresses exerted and the measure each particle individually and thus may present
physical properties of the material. In the early literature, a distribution of that strength. It may sometimes be that a
there was a preoccupation with the energy used to cause bulk test is sufficient for quality control purposes but an
breakage and some of the tests used were literally a small average value usually cannot be interpreted and used to
scale-grinding machine operated at prescribed condi- develop a better process or product. Single particle tests
tions. However, any general model requires the operat- are more informative but require a statistically mean-
ing, equipment and material variables to be separated ingful number of particles to be tested. The development
and later tests try to determine the behavior of a of such a test is the concern of this paper.
Any single particle testing method aims to first subject
1
Dr. W. J. Beekman (corresponding author), Dr. G. M. H. the particle to a controlled but variable stress and then to
Meesters, DSM Food Specialties, P.O. Box 1, 2600, AP Delft, determine the damage caused. It is possible, even usual,
(The Netherlands). to separate the methods on the basis of the stressing
E-mail: Joost.Beekman@dsm.com
2
Dr. G. M. H. Meesters, Delft University of Technology, Particle
mechanism, for example normal and tangential or
Technology Group, Department of Chemical Engineering and compression and impact. However, the stress field is
Materials Science, P.O. Box 5054, 2600 GA Delft (The Nether- not a material variable, it is the primary operating
lands). variable. On the other hand, the breakage that occurs
3
Prof. B. Scarlett, Engineering Research Center for Particle does result in different particle sizes and shapes and these
Science and Technology, PO Box 116135, Gainesville, Florida
32611 (USA). are material properties. In order to quantify the damage,
4
Dr. T. Becker, Genecor International, 925 Page Mill Road, Palo it may be best to consider the effect produced as being
Alto, California 94304 (USA). due to different mechanisms. The corollary to this

¹ WILEY-VCH Verlag GmbH, 69469 Weinheim, Germany, 2002 0934-0866/02/1901/0005 $ 17.50+.50/0


6 Part. Part. Syst. Charact. 19 (2002) 5 ± 11

classification of the damage is that the particle may later event. This is the reason that the fracture strength of
possess more than one strength parameter, and it further a particle cannot be calculated from the first principles.
follows that the test must be designed to measure and to Each small event that a particle has undergone in its
differentiate these parameters. In this paper, we describe history has contributed to weakening it. The only way to
an instrument, the Repeated Impact Machine (RIM), quantify this history is to consider it to be a breakage
which applies the stress to each particle in a controlled mechanism, fatigue, and to measure it. The second
and repeated fashion and which, thus, allows the gradual mechanism which the Repeated Impact Machine can
degradation which a particle experiences to be differ- simulate is the local damage which occurs when a particle
entiated from the abrupt failures. The gradual degrada- impacts and does not fail. This attrition occurs in the
tion will be called attrition and abrupt failure fracture. region of the impact and the machine can apply
The generic term is breakage. numerous such events randomly on the surface of a
Single particle breakage tests exert the stress on the particle. There will be a distribution of the stress
particle either by compression or by impact. In a single generated at each impact, even if the velocity is
particle-crushing test, one particle is compressed be- ostensibly the same. This is because the orientation of
tween two hard platens and the stress-strain behavior is the particle is also random3. Nevertheless, as a first
recorded until failure. What is recorded is a force. It is approximation, the attrition rate of smooth particles can
common to divide this force by the cross-section of the be normalized with respect to the total kinetic energy per
particle to yield a parameter with the dimensions of unit surface of the particle.
stress. However, this connection does not normalize the The Repeated Impact Machine is thus suitable for
influence of particle size and, therefore, appears to be applying a stress history to the particle whereby three
spurious. If a single stress is applied to the surface of a different mechanisms can be differentiated:
perfectly spherical, elastic particle, then the stress
Fracture is the massive failure of the particle into several
pattern within the particle is not normalized by the
smaller fragments. The failure is due to the tensile forces,
second power of the particle size, neither theoretically
generated by the compaction of the particle, and results
nor experimentally [2]. There is, therefore, no reason to
in longitudinal cracks propagating through the whole
think that the force that will break it can be normalized in
body of the particle. The fragments produced are usually
this way.
sharper and more angular than the parent particle as well
An impact test does not usually record the force that
as being significantly smaller.
causes failure; rather it records the chance that breakage
Attrition is due to local damage at the points on the
will occur at a particular velocity. In fact, the impact does
surface where the force is applied. The result is the
cause a stress-strain deformation to be applied to the
production of fine debris with little change in the size of
point of impact, the total work being equal to the kinetic
the parent particle. This particle, however, becomes
energy of the particle. The rate of strain is probably
rounder and smoother, particularly when the impacts are
higher than that in a typical compression test. However,
oblique and when the particle slides.
for small particles at typical velocities, that rate is low
Fatigue occurs in the body of the particle and is not
compared to the relaxation time of the particle. Thus, all
immediately obvious. However, each small event creates
of the kinetic energy is absorbed as elastic or plastic
weaknesses that make the particle more liable to
deformation before the particle starts either to break or
eventual fracture. The degree of fatigue that a particle
to rebound. With large particles or at high velocities this
has acquired dictates not only the force that will
may not be so; the particle may start to break at its point
eventually break it but also the way in which the cracks
of contact before it is completely decelerated. For the
will propagate during fracture. This in turn, determines
conditions considered in this paper, the damage is
the breakage function of the particles, the size, and
commensurate with that caused by the peak stress
number of fragments that are produced by the fracture.
generated by the impact, regardless of the strain rate
[3]. The major advantages of the impact test over the This machine has already been applied to diverse
compression test are two facts; (1) there is only one point processes, for example: milling, where the fracture is
of contact and (2) the particle is not necessarily brought required; and crystallization, where the attrition is
to the point of destruction. It is these two factors that the responsible for the secondary nucleation of the process.
Repeated Impact Machine (RIM) is designed to exploit In this paper, we describe its use to measure the fracture
[4]. In the first instance, a particle can be subjected to a attrition and fatigue behavior of complex agglomerate
large number of impacts that do not cause fracture but particles.
which do cause fatigue. As in all materials, so also in a
particle, repeated stress causes fatigue that increases the
chance that a particle will suffer massive fracture in a
Part. Part. Syst. Charact. 19 (2002) 5 ± 11 7

total weight of the aluminum arms and the collision


chambers has been tuned to resonate with frequency
close to 50 Hertz. For fine-tuning of the resonant
frequency, small masses are mounted that can be
positioned anywhere along the aluminum arms. When
operating without mechanical amplitude control, this
system results in a chamber vibration with an amplitude
of 3 centimeters, amplifying the vibration amplitude of
the platform approximately 20 times. By mechanical
restriction of the amplitude, smooth control is obtained,
provided that the resonance frequency of the system is
slightly larger than the frequency of the plate. In the
instrument described here, the frequency of the vibration
is 50 cycles per second and each particle experiences 100
collisions per second at a velocity that can be varied
Fig. 1: Particle breakage mechanisms. between 4 and 12 m/s.

3 Number of Granules and Tests

A series of tests were made to determine the maximum


number of particles that can be tested without effects of
particle-particle interactions. At each condition, the tests
were repeated with incremental additions to the number
of particles. For this purpose, simple sieving assessed the
damage caused to the particles. The samples were
Fig. 2: Schematic drawing of the symmetrical repeated impact initially prepared into three particle size fractions,
machine set-up. 510 ± 640 mm, 640 ± 710 mm and 710 ± 840 mm. After
each test, a sieve of 325 mm separated the particles
remaining in the box. Since this sieve is appreciably
2 Experimental Setup smaller than the initial size, it was assumed that all of the
particles that passed, the debris, were the result of both
A schematic drawing of the Repeated Impact Machine is fracture and attrition. The particles that were retained in
given in Figure 2. the sieve of 325 mm included the initial parent particles
A heavy platform is mounted on a laboratory sieve and the major fragments. They were returned to the box
shaker that is connected to two long metal beams by for the next time cycle. Figure 3 shows a typical plot of
means of a thin blade spring. Plastic boxes are mounted the debris produced after 200 collisions as a function of
on both ends of the spring. Due to the movement of the the relative collision area. This concept was used to
platform, the two metal beams resonate and achieve normalize the number of particles in the box. It is defined
vibration amplitudes of up to 2 centimeters. The sample
of particles is placed in one of the boxes and each particle
collides with each wall during each half cycle of the
vibration. The amplitude of the vibration controls the
velocity of the impact and the frequency controls the
number of impacts. The height of the box must be greater
than the amplitude of the vibration and the coefficient of
restitution should be low. In the experiments described
here, a plastic box was used which is flexible but this can
be coated internally with different surfaces if a hard
impact is required. The most essential part of the new
instrument is the steel spring blade. Only 1 centimeter of
the spring blade on each side of the platform is allowed to
deform and this results in a very stiff spring blade Fig. 3: Determination of the effect of the number of particles in
connecting the aluminum arms to the vibrating plate. The the R.I.M. as a function of available wall area per particle.
8 Part. Part. Syst. Charact. 19 (2002) 5 ± 11

Fig. 6: Collisions are attributed to the top wall when this wall is
moving downwards and to the bottom wall when this wall is
Fig. 4: Maximum number of particles and maximum box load as moving upwards. A simple computer algorithm can perform the
a function of particle diameter for reliable R.I.M.-operation. reconstruction of the particle movement.

samples of about 300 particles, the maximum variation


on the total damage is about 10%. By using the necessary
number of tests, the error on the mean can be reduced to
any required value. In order to differentiate the different
mechanisms, it is necessary to make a more elaborate size
and shape analysis.

4 Validation of the Collision Pattern

The assumption underlying the Repeated Impact Ma-


chine is that each particle impacts with the top and
Fig. 5: Determination of reproducibility of the damage develop-
ment as a function of number of collisions. bottom of the box during each cycle. How well this ideal
behavior is achieved can, of course, be determined by
direct filming. However, a simpler technique that can, if
as the area of the top and bottom collision plates required, be automated is to listen to the sound
(280 mm2) compared to the projected area of all the generated by the impacts. In this case, two microphones
particles. were mounted, one close to each end of the maximum
The Figure illustrates that the available surface should traverse of the box. The frequency response of the
be at least three or four times the projected area of the microphones was linear up to a value of 16 KHz. Both
particles. Thus, the number of particles that can be tested microphones recorded both sets of impacts but, because
is inversely proportional to the square of their diameter of the way they were mounted, could be differentiated by
and varies between 200 particles of 600 mm diameter and their amplitude. The position of the box was recorded by
450 of 400 mm. This is illustrated in Figure 4. the intersection of a light beam and this motion was used
The machine was also evaluated for reproducibility. to deduce the velocity of the particles at the time of the
Figure 5 shows the results of tests made at identical impact. Such a reconstruction is shown in Figure 6, which
conditions on four samples of the same granules. In this shows the recorded impacts and the reconstructed
case, the percentage by mass of particles that were motion of the particle. In this case, ideal glass beads
retained on the sieve of 325 mm is plotted as a function of were used so that no breakage was occurring during the
the number of impacts that the particles have experi- demonstration.
enced. The sound generated by the glass beads was recorded for
This fraction includes the particles that have suffered twenty seconds, thus 4000 separate collisions, at several
some attrition but not massive fracture as well as some of different amplitudes of the vibration. The resulting
the largest fragments produced. The greatest variation distributions of impact velocity are shown in Figure 7.
thus occurs at the point where approximately 50% of the In this case, the smallest distribution of impact velocities
particles are in each fraction, greater and less than was obtained at amplitude of 1.3 centimeters. This is
325 mm. This is the point at which the large fragments are primarily due to the size of the amplitude in comparison
also contributing considerably to the attrition but, with to the height of the box, in this case 2.6 centimeters.
Part. Part. Syst. Charact. 19 (2002) 5 ± 11 9

Fig. 7: (Glass bead): particle-wall impact velocity profiles as a


function of vibration amplitude.

Fig. 9: Simulated trajectories of the particle, both as seen by an


observer moving with the box (a) and from a fixed observer (b),
showing position of the particle and position of the top and
bottom walls of the box every millisecond of half a vibration cycle.

possible to simulate the motion of the particles in the


machine without, of course, breakage.
Such a simulation is shown in Figure 9, which shows the
same simulated motion of the glass beads relative to the
box and relative to a fixed axis. With this figure, it is
possible to visualize better the importance of the
dimensions of the box and of the amplitude of the
displacement as well as the value of the coefficient of
Fig. 8: (Non-pareil): particle-wall impact velocity profiles as a
function of vibration amplitude.
restitution. With an increase in both amplitude and
coefficient of restitution the velocity of impact becomes
larger. The particle may then travel to the next impact
When the experiment was repeated with granules, the before that wall has reversed direction and accelerated
results were similar but the optimum amplitude had back to its required velocity. On the other hand, if the
increased to 1.7 centimeters as shown in Figure 8. amplitudes and velocities are too low, the particle may
The difference in the two cases was the coefficient of not reach the wall before it has reversed direction; the
restitution relevant to the impacts. This coefficient can particle should each time hit the wall when they are
also be deduced from the trajectory traces of Figure 6 moving in opposite directions. With low velocities, the
and results in a value of 0.45 for the glass beads and 0.35 motion of the particle relative to the box is not periodic; it
for the granules. These results demonstrate that the is erratic. The best results are obtained when the particles
model value of the impact velocity can be varied, over a remain in a relatively small zone relative to the fixed axis
restricted range, by varying the amplitude of the and impact each wall when it is moving at a velocity close
vibration. However, an optimum amplitude exists for to its maximum. This simulation makes it clear that the
any particular box at which the distribution of impact velocity range of the prototype instrument described can
velocities is very small. The acoustic technique is a very be extended by using two other variables. They are the
convenient method to detect the optimum automatically height of the box and the frequency of the vibration.
and to measure the coefficient of restitution. By using the Figure 10 shows the simulated impact velocities for the
measured values of the coefficient of restitution, it is glass beads when the frequency has been increased to 60
10 Part. Part. Syst. Charact. 19 (2002) 5 ± 11

different, it is necessary to perform the tests on a


representative sample of individual particles.
It is convenient to generate the repeated stresses either
by compression or by impact. Machines utilizing both
techniques have been improvised and used to demon-
strate that a full characterization of the particle proper-
ties requires the two to be used in conjunction [5]. The
compression test has the advantage that the stress is more
accurately measured and controlled. The impact test has
two advantages, that the stress is applied randomly over
the surface of the particle and that a large number of
particles can be measured relatively quickly. For these
reasons the Repeated Impact Machine can be claimed to
Fig. 10: Impact velocity distribution for different vibration be the primary tester which can be supplemented by
amplitudes (60 Hertz simulations).
others when necessary.
In the RIM the intensity and number of stress events can
be varied at will. In a single box, the amplitude and
frequency of the vibration can be utilized to vary the
frequency. In order to create a wider range, the size and
surfaces of the box must also be varied. It is foreseen that
this machine is the prototype for a Combinatorial
Particle Tester in which a number of different boxes
will be mounted on the same shaker that has variable
frequency. Boxes with optical windows will allow in-situ
measurements and the motion can be monitored. The
RIM can be used to distinguish at least the mechanisms
of fatigue, fracture and attrition. The machine can be
used initially with impacts at low velocity to create
Fig. 11: Relations between vibration amplitude and 20% fastest samples of particles that have a different fatigue history.
impact velocities, determined from sound record analysis and by By subsequently observing the size and shape of the
simulations.
particles when subjected to larger impacts, the particles
that have not yet suffered fracture can be easily recorded.
The attrition rate of the particles is very dependent upon
Hz. The smallest distribution of velocities is now their shape and is best determined from the initial rate at
obtained with amplitude of only 0.6 centimeters, but which debris is produced. Figure 12 shows results where
the model impact velocities are increased by 20%. the attrition rate of three different particle sizes of the
The experimental and simulated data for this particular same material are normalized.
box is collected together in Figure 11, which shows the The results fit approximately a relationship of the
range of velocities that can be obtained. If higher following form:
velocities are required, it is necessary to increase the
frequency and the height of the box. R ˆ a12d

* R is the fractional rate at which the mass of the


5 Conclusions particle sample is attrited.
* (1v2d)/12 is the Kinetic Energy per impact per unit
Particle strength is a parameter that cannot be calculated surface area of the particle.
from first principles. The inherent reason for this is that
each small mechanical event which the particle experi- This relationship has the same form as that proposed by
ences causes some damage and change to the particle Cleaver and Ghadiri [6].
which contributes to its response to future events. It is, The coefficient a may be considered to be a material
therefore, necessary to devise tests that measure the constant for round particles but it does change with
particle strength as a function of its history. This require- particle shape. When significant fracture has occurred
ment implies tests must be able to subject the particle to and there are sharp fragments present, then the rate of
repeated stresses. Furthermore, since each particle is attrition significantly increases. The fragments can be
Part. Part. Syst. Charact. 19 (2002) 5 ± 11 11

Fig. 12: Attrition rate as a function of


particle size and impact velocity for
three different size classes of the fluid
bed and high shear granule; these fig-
ures show that the attrition is propor-
tional to the particle diameter.

mechanism that could be called peeling. The conclusion


is clear, that the measurement and interpretation of the
particle damage may vary from the simple to the complex
and from the empirical to the mechanistic. Whatever
degree of sophistication is required, the Repeated
Impact Machine can provide the stress history that
varies the number of events and their intensity in a
controlled and flexible manner.

6 Acknowledgements
Fig. 13: Impact damage development for a HPMC coated
This research would not have been possible without the
particle, showing the performance of the core particle, the enzyme
layer and of the HPMC-coating and its thickness. generous support of Genencor International.

tested separately. It is significant that there is a minimum 7 References


velocity at which damage occurs, at smaller velocities, the
collision is elastic. [1] C. R. Bembrose, J. Bridgewater: Powder Technol. 49 (1987)
The RIM can be used to test complex particles. Figure 13 97 ± 126.
shows the results of a series of tests on a complex granule. [2] I. M. J. Hutchings: Phys. D. Applied Phys. 10 (1977).
[3] P. M. M. Vervoorn, L. G. Austin: Powder Technol. 64 (1990)
The purpose of the tests was to optimize the thickness of 141 ± 147.
a coating of hydroxy propyl methylcellulose. [4] W. J. Beekman, G. M. H. Meesters, B. Scarlett: US Pat. No 08/
The simple sieve analysis was sufficient to evaluate and 681.250 (1996), Genencor International; Method and device
control the thickness of this outer layer. More detailed for characterizing granule strength.
[5] W. J. Beekman: Measurement of the Mechanical Strength of
interpretation of the total breakage behavior requires
Granules and Agglomerates. Thesis, Delft University of
the different components of the granule to be differ- Technology, December 11, 2000.
entiated within the debris. A complex particle with [6] J. A. S. Cleaver, M. Ghardiri, N. Rolfe: Powder Technol. 76
several layers also requires the postulation of another (1993) 15 ± 22.

You might also like