Professional Documents
Culture Documents
Abstract
In this paper we develop a performance model for impact crushers. The product size distribution is obtained as a function of the
crusherÕs rotor radius and angular velocity, the feed rate and the feed size distribution. The model is based on the standard matrix
formulation that includes classification and breakage matrices. It can be applied to both hammer and vertical-axis impact crushers
with the help of the corresponding estimations for the impact energy per unit mass.
Here we propose classification and breakage functions for impact crushers taking into account the dynamic character of the
impact breakage. The classification function has the form of a cumulative Weibull distribution and incorporates a minimum
breakable size of the particles depending on the impact energy and the feed rate. The breakage function is modelled as the sum of
two Broadbent–Callcott distributions. It is assumed to depend on the impact energy and the feed rate through the proposed ex-
pression for the proportion of the fine fraction in the product.
The model predictions are compared with experimental data for limestone treated in a pilot-plant hammer crusher. The vari-
ations of the product size distribution resulting from changes in the rotor velocity and the feed rate are investigated.
Ó 2002 Elsevier Science Ltd. All rights reserved.
tant than the particle velocity, the kinetic energy asso- lated to any movement of the crusher parts and the
ciated with a single particle is negligible compared with particles. With their experiments on impact fracture of
that of the rotor. Considering the conservation of linear single particles, King and Bourgeois (1993) demon-
momentum of the system particle-crushing bar before strated that the probability of impact breakage for a
and after impact, Attou et al. (1999) derived the fol- given particle depends mainly on its size and the impact
lowing expression for the impact energy per unit mass: kinetic energy, which is clearly a dynamic variable that
includes the velocity of impact. Therefore, we have to
E ¼ 0:5 ðR þ 0:5 Hb Þ2 x2 ð3Þ
define another, more suitable classification function re-
where R (m) is the rotor radius, Hb (m) is the height of flecting the dynamic character of the impact breakage.
the impact surface of the crushing bars and x (s1 ) is the The experiments performed by King and Bourgeois
rotor angular velocity. (1993) as well as by Kapur and Fuerstenau (1995) show
In vertical-axis crushers, the particles are fed to a that the probability of impact breakage for single par-
horizontal turning table (rotor) with radially oriented ticles in function of the impact energy is well described
guides and are projected towards the crusherÕs walls by by the Weibull distribution proposed by Weichert
the centrifugal forces. Unlike in hammer crushers, here (1988). Attou et al. (1999) adapted this distribution as a
most of the fragmentation takes place at the crusherÕs classification function for impact crushers in the fol-
walls rather than at the rotorÕs periphery. With the as- lowing form:
sumption that the particle energy does not change dur-
a k n
ing its flight from the rotor periphery to the crushing Ci ðdi Þ ¼ 1 exp s di E ð6Þ
Q
walls, i.e., the particle–particle interactions are neglected
in a first approximation, Nikolov and Lucion (2002) where E (J/kg) is the average impact energy per unit
derived the following expression for the impact energy mass, Q (t/h) is the feed rate; a and s depend on the
per unit mass: specific crusher design and size; k and n are related to
the material properties of the granulate. In Eq. (6) the
E ¼ R2v x2 ð4Þ
breakage probability for particles with size of several
1
where Rv (m) and x (s ) are the rotor radius and an- dozens of microns is not exactly zero at given feed rate
gular velocity respectively. The notation Rv is used to and rotor velocity, which is in contradiction with the
distinguish between the impact energy for hammer and experimental evidence. Moreover, Eq. (6) is an almost
vertical-axis crushers given with Eqs. (3) and (4) re- step-like function of the particle size when applied to
spectively. typical size distributions. Actually, Eq. (6) has been fit-
It is interesting to note that for the same rotor ra- ted to experimental data for breakage of identical par-
dius, the impact energy per unit mass provided by ticles with different impact energies while the inverse
hammer crushers is lower than that provided by vertical- problem (breakage of particles with different size with
axis crushers. This could explain the fact that vertical- the same impact energy) has not been investigated. Ac-
axis crushers produce more fines and perform better cording to Eq. (6), a very limited number of particles
when finer granulate must be reduced in size, which is should have breakage probabilities different from one or
most probably due to the higher level of impact energy zero, which is unrealistic.
reached in these machines. In order to deal with these problems, Nikolov and
Lucion (2002) recently proposed a new classification
2.3. Classification function function in the form:
" k #
In their model for cone and jaw crushers, Whiten and di dmin
White (1979) compute the components of the classifi- Ci ðdi Þ ¼ 1 exp ð7Þ
dmin
cation matrix C with the help of the following classifi-
cation function: where dmin (mm) is the minimum size of the particles that
m undergo breakage for the given operating conditions
di k2
Ci ðdi Þ ¼ 1 ð5Þ and k controls the shape of the classification function.
k1 k2
The probability of breakage for particles smaller than
where Ci ðdi Þ is the probability of breakage for a particle dmin is Ci ðdi Þ ¼ 0 by definition.
of size di (mm), k1 (mm) is the minimum size of particles Eq. (7) is actually a cumulative Weibull distribution
that undergo breakage, k2 (mm) represents the maxi- that can be derived from physical considerations related
mum particle size found in the product (i.e., all particles to brittle fracture of solids. Similarly to Eq. (5), the
in the feed with size di > k2 are broken); m is a shape classification function (7) contains a minimum break-
parameter. able size but does not contain a predefined upper size of
In Eq. (5), both k1 and k2 depend explicitly on the the particles in the product. This is in accord with the
crusher setting, which is a static variable and is not re- fact that the product issued from impact crushing has
718 S. Nikolov / Minerals Engineering 15 (2002) 715–721
rffiffiffiffiffi
broader size distribution compared with that issued E
from cone or jaw crushing. k ¼ k0 k1 ð9Þ
E0
Our basic assumption is that the minimum breakable
size dmin for impact crushing is a function of both the where k0 and k1 are material coefficients.
impact energy and the feed rate. Higher rotor velocity or,
equivalently, higher impact energy per unit mass E, leads 2.4. Breakage function
to a finer product size for the same feed rate and size.
Therefore, dmin should decrease with increasing of E. According to Karra (1982), the breakage distribution
As for the influence of the feed rate on dmin , an in- function bij represents the fraction of the debris created
crease in the feed rate, keeping the feed size unchanged, from breakage of identical parent particles of size dj and
results in higher frequency of the particle–particle col- passing through a screen with mesh size di . It is assumed
lisions. As each collision dissipates energy, the average that the shape of the size distribution of the debris is
impact energy will decrease at higher feed rates and that independent on the size of the parent particles. As
would result in a coarser product, i.e., a greater value for mentioned in a review paper by Kelly and Spottiswood
dmin . Taking into account these considerations, we can (1990), this is true for almost all of the experimentally
write: studied product size distributions obtained by crushing
n or grinding of a large number of different minerals.
Q0 E The breakage distribution function for crushers de-
dmin ¼ dmax exp c0 þ c1 ln
Q E0 rived by Whiten and White (1979) reads:
ð8Þ m l
di di
bij ðdi ; dj Þ ¼ / þ ð1 /Þ ð10Þ
where Q and E are the feed rate and the average impact dj dj
energy per unit mass respectively; Q0 and E0 are refer- where / denotes the mass fraction of the fine product; m
ence feed rate and reference impact energy respectively; and l are material coefficients accounting for the shape
dmax (mm) is the maximum particle dimension in the of the fine and the coarse product size distributions re-
feed; n is a material parameter; c0 is a rate constant and spectively.
c1 accounts for the intensity of the particle–particle in- The form of Eq. (10) reflects the fact that the
teractions. breakage function for crushers is better represented as
The dependence of dmin on the feed rate is modelled the sum of two probabilistic distributions. According to
with a logarithmic function because of the fact that rel- Kelly and Spottiswood (1990), this is due to the presence
atively important variations in the feed rate do not of two different breakage mechanisms at a microscopic
change significantly the product size distribution ob- level: shatter and cleavage. An important remark is that
tained with impact crushers. In addition, at reference the proportion of the product fine fraction / in Eq. (10)
feed rate Q ¼ Q0 , lnðQ0 =QÞ ¼ 0 so that all other model is assumed to be independent on the operating condi-
parameters can be identified independently on c1 . It is tions. The size distributions of the fine and the coarse
noted that c1 should depend, among the other factors, on products are modelled with a power-law function that
the volume enclosed between the rotor and the crusherÕs cannot be derived from a known statistical distribution.
internal walls. A larger volume of the fragmentation In a recent work, Nikolov and Lucion (2002) used the
chamber leads to a lower volume fraction of solids for classification function given with Eq. (7) together with
the same feed rate and particle size and therefore, the the breakage distribution function (10) to simulate the
particle–particle interactions are less frequent. performance of hammer crushers. The results of the
The parameter 0 < n < 1 specifies the relationship simulations suggest that the assumption of a constant
between the impact energy per unit mass and the mini- fine product fraction / in Eq. (10) may be too restrictive
mum breakable size dmin and controls the saturation to correctly describe the changes in the product size
value of dmin for given granulate properties and crusher distribution obtained from impact crushing with differ-
design. ent rotor velocities and feed rates. Here we propose a
The parameter k in Eq. (7) controls the shape of the breakage function consisting of the sum of two Broad-
classification function and hence, the shape of the bent–Calcott distributions:
product size distribution. Greater values for k corre-
( m
spond to a higher breakage probability for the large di
particles. Therefore, k can be expected to decrease with bij ðdi ; dj Þ ¼ 1:582 / 1 exp
dj
increasing the impact energy. In this work, we assume "
that k depends directly on the rotor velocity. As the l !#)
di
impact energy is proportional to the second power of the þ ð1 /Þ 1 exp
dj
rotor velocity, the dependence of k on the impact energy
can be written as: ð11Þ
S. Nikolov / Minerals Engineering 15 (2002) 715–721 719
where /, m and l have the same meaning as in Eq. to 0.74, 2.6 and 0.55 respectively. The values of the
(10). The normalisation factor is computed as 1=½1 parameters necessary to compute the shape of the clas-
exp tð1Þ ¼ 1:582. sification function are fixed to k0 ¼ 1:35 and k1 ¼ 0:1.
The cumulative distribution function derived by We have performed simulations for two sets of ex-
Broadbent and Callcott (1956) is actually a normalised perimental data taken at different feed rates, namely
Weibull distribution starting from a particle size d ¼ 0 Q ¼ 2 and 7 t/h. At fixed feed rate, the products ob-
and, as already explained, not only provides a better fit tained with three different rotor velocities (x ¼ 540; 720;
to the experimental data but can also be obtained from 900 rpm) have been analysed. It is important to note
physical considerations related to brittle fracture. that except for the rotor velocity and the feed rate, all
As for the proportion of the fine product /, Na- other model parameters have been kept unchanged for
rayanan and Whiten (1988) found that during impact all performed simulations.
breakage of single particles, / increases with increasing The experimental and simulated size distributions of
the energy intensity while m and l remain virtually un- the products obtained with different rotor velocities at
changed. In the case of impact crushing, this means that feed rate of 2 t/h are compared in Fig. 2. The corre-
/ increases with increasing the rotor velocity. On the sponding size distributions obtained at feed rate of 7 t/h
other hand, / should decrease with increasing the feed are depicted in Fig. 3.
rate because higher frequency of the particle–particle It is seen that the model is able to predict the product
interactions involves lower average impact energy in the size distribution with a reasonable accuracy even when
crushing chamber. Taking into account that 0 < / < 1 important variations in both the rotor velocity and the
by definition, we can express the above arguments in a feed rate are imposed. At fixed feed rate, higher rotor
mathematical form as follows:
rffiffiffiffiffi
Q0 E
/ ¼ 1 exp c2 þ c1 ln ð12Þ
Q E0
where c2 is a rate constant; for simplicity, c1 is assumed
to be the same as in Eq. (8).
The breakage matrix B is constructed with the
method outlined in King (2000). Let the size distribution
of the granulate is obtained with a series of N screens
with mesh sizes Di (i ¼ 1; N 1); DN ¼ 0 by definition.
Then di is the representative size of particles with di-
mensions Di > di > Diþ1 . The non-zero components Bij
of the breakage matrix B can be computed using Eq.
(11) as shown in King (2000):
Bij ¼ bði1Þj ðDi1 ; dj Þ bij ðDi ; dj Þ
ð13Þ
Bjj ¼ 1 bjj ðDj ; dj Þ
Fig. 2. Measured (symbols) and predicted (curves) product size dis-
tributions for different rotor velocities at feed rate Q ¼ 2 t/h.
3. Results
velocity produces a finer product size distribution. Al- ing contains more fines than that obtained with cone or
ternatively, an increase in the feed rate results in a jaw crushers.
coarser product at fixed rotor velocity. The results also
confirm that the product size distribution issued from
impact crushing is broader than that obtained with cone 4. Conclusions
or jaw crushers.
The influences of the rotor velocity and the feed rate In conclusion, we have developed a performance
on the minimum size of the particles that undergo model for impact crushers that is able to predict the
breakage dmin are shown in Fig. 4. It is seen that dmin product size distribution at steady-state operating con-
strongly depends on both the rotor velocity and the feed ditions and contains a reasonable number of para-
rate and ranges from 3.8 to 7.8 mm for the given op- meters. The specific behaviour of impact crushers is
erating conditions. modelled through classification and breakage functions
The evolution of the proportion of the fine fraction / that both depend on the rotor radius and angular ve-
in the product with rotor velocity and feed rate is shown locity as well as on the feed rate. The introduction of
in Fig. 5. It increases with increasing the rotor velocity variable minimum size of the breakable particles and
and decreases with increasing the feed rate as expected proportion of fine fraction in the product seems to be
and ranges from 0.35 to 0.59 for the given operating very important for successful modelling of the impact
conditions. crushing.
For comparison, when the model of Whiten and The simulation results are in a good agreement with
White (1979) is used for simulation of the behaviour of the experiment and show that at higher rotor velocity
short-head cone crushers, the values for /, m and l are the product size distribution becomes finer, provided
often fixed to 0.2, 0.5 and 2.5 respectively. A greater that the feed rate and size are kept unchanged. Alter-
value for the fine fraction / in our case reflects the well- natively, higher feed rate at constant rotor velocity re-
known fact that the product issued from impact crush- sults in a coarser product.
The model can be easily implemented in the existing
commercial codes for mineral processing simulations. It
can be used for prediction of the steady-state perfor-
mance of hammer and/or vertical-axis impact crushers
integrated in complex flowsheets. Further work is re-
quired to adapt the model for unsteady and transient
operating regimes.
Acknowledgements
References
Kapur, P.C., Fuerstenau, D.W., 1995. A model analysis of the impact tion to ball-mill scale-up. In: Transactions (C) IMM, 97. pp. C115–
grinding of single particles. In: Proceedings of the 19th Interna- C124.
tional Mineral Processing Congress, Colorado, USA, 1. pp. 125– Nikolov, S., Lucion, Chr., 2002. Modelling and simulation of particle
130. breakage in impact crushers. In: Proceedings of the 10th European
Karra, V.K., 1982. A process performance model for cone crushers. In: Symposium on Comminution, Heidelberg, Germany, C3.2, pp. 1–
Proceedings of the 14th International Mineral Processing Congress, 10.
Ontario, Canada, III. pp. 6.1–6.14. Oka, Y., Majima, W., 1970. A theory of size reduction involving
Kelly, E.G., Spottiswood, D.J., 1990. The breakage function; What is fracture mechanics. Canadian Metallurgical Quarterly 9, 429–439.
it really? Minerals Engineering 3 (5), 405–414. Weichert, R., 1988. Correlation between probability of breakage and
King, R.P., 2000. Technical notes 5, Crushers. In: Continuing fragment size distribution of mineral particles. International
education course on simulation and modeling of mineral processing Journal of Mineral Processing 22, 1–8.
plants, University of Utah Internet course. p. 5. Whiten, W.J., 1972. A model for simulating crushing plants. Journal of
King, R.P., Bourgeois, F., 1993. Measurement of fracture energy South African Institute of Mining and Metallurgy 72 (10), 257.
during single particle fracture. Minerals Engineering 6 (4), 353–367. Whiten, W.J., White, M.E., 1979. Modeling and simulation of high
Narayanan, S.S., Whiten, W.J., 1988. Determination of comminution tonnage crushing plants. In: Proceedings of the 12th International
characteristics from single-particle breakage tests and its applica- Mineral Processing Congress, Sao Paulo, Brasil, 2. pp. 148–158.