Minerals Engineering 93 (2016) 50–56

Contents lists available at ScienceDirect

Minerals Engineering
journal homepage: www.elsevier.com/locate/mineng

Multi component modelling of an air classifier

Okay Altun a,⇑, Alper Toprak a, Hakan Benzer a, Ozgun Darilmaz b
a
Hacettepe University, Mining Engineering Department, 06800 Beytepe, Ankara, Turkey
b
First Quantum Minerals, Cayeli Copper Corporate, 53200 Madeni, Çayeli, Rize, Turkey

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

Article history: This paper focusses on investigating the classification behaviours of the components having different
Received 24 November 2015 densities and flow characteristics then developing preliminary model structure where these properties
Revised 1 March 2016 are considered. Such a study can improve the prediction accuracy of the existing models since material
Accepted 21 April 2016
characteristics are of crucial importance. Within the scope of this study, laboratory scale experimental
tests were undertaken on clinker, copper ore, magnetite and coal samples, at different operating condi-
tions. The results concluded that, increasing the density decreased the cut size in the meantime increased
Keywords:
the bypass of the classification operation. In addition, the sharpness and the fish hook parameters were
Air classifier
Classification
found to be correlated with the flow characteristics of the material e.g., the higher the fluidity the higher
Modelling the sharpness and the lower the fish-hook. As a conclusion of the study, the correlations presented in the
Comminution paper were integrated into an existing air classifier model and preliminary multi component model struc-
Efficiency curve ture for air classifiers was developed.
Ó 2016 Elsevier Ltd. All rights reserved.

1. Introduction introduced and improved performances have been reported

(Duda, 1985; Yardi, 2005). Improved efficiency of HES can be
Air classification is a process of separating particles into two or attributed to mounting of the fan outside of the classifier body,
more groups according to their shapes, sizes and specific gravities. using the cyclones to collect the fines and using the rotor cage
The process has been utilized by different industries i.e., cement, structure that enabled the forces participating in the classification
food, coal, where water interaction is avoided. Up to date, different well defined (Klumpar et al., 1986). Table 1 compares the perfor-
types of air classifiers e.g., static and dynamic classifiers, have been mances of the above-mentioned generations regarding to their
developed. Static air classifiers adjust the target product size only sharpness of separation parameters. As can be understood, the
by changing the magnitude and the direction of the airflow. On third generation classifier has sharper separation means the tech-
the other hand, dynamic classifiers have a rotating cage that is nology has improved classification efficiency.
used both to disperse the feed and adjust the product size distribu- During the classification operation of the HES, the particles are
tion. Dynamic air classification technology has been evolving since under the influences mainly of centrifugal (Fc), drag (Fd) and grav-
1885 when the first generation air classifiers were introduced. ity (Fg) forces. The centrifugal force is generated by rotor, which
Briefly, first-generation classifier has a dispersing plate on which accelerates the particles towards the outside edge of the distribu-
the material is poured then thrown towards the separator wall tion plate. Air enters the classification zone tangentially and cre-
where the final classification is performed with the introduction ates a drag force that performs the final separation. The
of the air. Within the technology, air is generated inside the sepa- mathematical definitions of the forces (Klumpar et al., 1986;
rator body. Following the first-generation technology, second gen- Duda, 1985) are given in Eqs. (1)–(3).
eration air classifiers were developed. The second-generation
4 V2
classifiers are operated with cyclones in order to increase the fines Fc ¼  p  r 3p  qp  ð1Þ
collection efficiency. Additionally, fan that is generating air for the 3 r
2
V
classification is mounted outside of the separator body that also F d ¼ cD  q  p  r 2p  a ð2Þ
improved the overall efficiency of the machine. Finally, the third a
generation air classifiers or high efficiency classifiers (HES) were F g ¼ m  ðq  qair Þ  g ð3Þ
where;
⇑ Corresponding author. rp: particle radius
E-mail address: okyaltun@hacettepe.edu.tr (O. Altun). qp: particle density

http://dx.doi.org/10.1016/j.mineng.2016.04.014
0892-6875/Ó 2016 Elsevier Ltd. All rights reserved.
 O. Altun et al. / Minerals Engineering 93 (2016) 50–56 51

Table 1
The variation of the sharpness of separation parameter with the classifier design
(Yardi, 2005).

Sharpness (d25/d75) Classifier type

0.25 Static
0.30–0.35 Static to generation 1
0.40–0.45 Generation 1–2
0.50–0.55 Generation 2–3
0.60–0.70 Generation 3 (HES)

V: peripheral velocity of the rotor

r: rotor radius
cD: drag coefficient
q: gas density
Va: air velocity
m: mass of particle
g: gravitational constant.

As can be understood from Eqs. (1)–(3), the magnitudes of the

forces depend both on the diameters and the densities of the par-
ticles. The coarser and the denser particles will be affected most by
the gravitational and the centrifugal forces. Consequently it can be
concluded that, as the feed gets denser, the cut size of the classifi-
cation is to decrease and vice versa.
Fig. 1. Alpine 100 MZR air classifier.
The literature reports that the material properties have influ-
ence on the performance of the classification operation. It is a
well-known fact that, different components exhibit different beha- Table 2
viours and the models should be developed accordingly so as to The features of the Alpine 100 MZR classifier.
improve the predicting capabilities of the models. Within the scope
Wheel speed (rpm) 1000–15,000
of the study, density and agglomeration tendencies of the bulk Air volume (m3/h) 5–50
material were considered in mathematical modelling of a labora-
tory scale air classifier. Initially, experimental studies were under-
taken with different samples. Afterwards, mass balancing studies Once the material is fed, the particles entering the classification
were performed and the size-by-size efficiencies were calculated zone are under the influence of drag, centrifugal and gravity forces
then inputted to the Whiten’s efficiency curve equation (Napier- thus either subject to the coarse or the fine stream. Fig. 2a–c illus-
Munn et al., 1996; Benzer et al., 2001; Altun and Benzer, 2014). trates the cross sectional views of the classifying chamber, rotor
Finally, the parameters exist in the Whiten’s equation (Eq. (4)) structure and the influencing forces in the classification operation.
were correlated with the operating conditions of the air classifier The operational range of the classifier is reported as between 2
as well as the material characteristics. The study contributes to and 80 lm and the feed rate is between 2 and 6 kg based on the
the literature regarding explaining the behaviour of the compo- density of the material. As can be seen from Fig. 2, feeding is per-
nents in the air classification operation that is then used in devel- formed from the screw feeder. The air enters the classifier and
oping the multi-component modelling structure. reaches to the rotor. It flows through the rotor from the outside
2
 3 to the inside and leaves the classifier, taking the fine particles with
1 þ b  b  d50c
d
 ðexpðaÞ  1Þ
Eoa ¼C4
 5 ð4Þ it. Coarse material that is flung by the centrifugal force is taken
exp a  b  d50c
d
þ expðaÞ  2 from one point on the circumference of the housing and collected
for weighing. The classifier has a rotor with zigzag, radially
where; arranged channels (Fig. 2b). The influencing forces in classification
Eoa: The actual efficiency to overflow operation are illustrated in Fig. 2c.
C: Fraction subjected to real classification; (100-Bypass)
b: Parameter that controls the initial rise of the curve in fine 2.2. Experimental and mass balancing studies
sizes (fish-hook)
b⁄: Parameter that preserves the definition of d50c; d = d50c Experimental studies comprise the air classification tests that
when E = (1/2)C are followed by characterization works and the weighing of the
a: Sharpness of separation fine and coarse products of the classifier. Within the context of
d: Size the study, coal, magnetite, clinker and copper ore samples were
d50c: Corrected cut size. subjected to the classification tests of which the test plan is given
in Table 3.
2. Materials and methods Regarding to the characterization, the density and the particle
size measurements were undertaken. Within the study, the densi-
2.1. Description of the experimental apparatus ties of the feed samples were determined by pycnometer method
(TS EN 1097-7) where mass and volume measurements were
Within the study, experimental tests were undertaken with undertaken with a glass container having the specified volume
Alpine 100 MZR Classifier (Fig. 1) having the features given in (Table 4). It should be emphasized that the densities determined
Table 2. In this machine, the target size is adjusted by changing within the study represent the overall density of the samples for
the wheel speed and the air flow rate used in the classification. a given size distribution. The particle size measurements were
52 O. Altun et al. / Minerals Engineering 93 (2016) 50–56

(a) (b) (c)

particle
Fd Fc

Fg

Fig. 2. Cross sectional view of the classifying chamber (a), wheel or rotor structure (b), and the influencing forces in the classification (c).

Table 3
The test plan of the classification tests arranged for each sample.

Wheel speed (rpm)

Air flow rate (m3/h) 3000 5000 7000
10 T1 T2 T3
20 T4 T5 T6
30 T7 T8 T9
40 T10 T11 T12

Table 4
The measured densities of the feed samples.

Material Density (g/cm3)

Coal 1.4
Clinker 3.05
Copper ore 4.1
Magnetite 5.1 Fig. 4. Measured particle size distribution curves around the classifier at 5000 rpm;
20 m3/h air flow.

undertaken via laser diffractometry method and the whole distri- this regard, JK-SimMet mass balance module (Napier-Munn et al.,
bution down to 1.8 lm was determined. Fig. 3 illustrates the mea- 1996) was used and the agreement between the measured and
sured size distribution curves of the feed samples and Fig. 4 the calculated passing percentages was assessed as illustrated in
illustrates the size distribution curves obtained as a consequence Fig. 5.
of the classification test for a given operating condition. As seen
in Fig. 3 the feed samples having similar size distributions were
2.3. Performance evaluation of an air classifier
prepared hence the effects of feed fineness on classification perfor-
mance were not taken into consideration. The results of the entire
Performance of an air classifier is evaluated by drawing the ‘‘ac-
measurements were used in developing the correlations with the
tual efficiency curve” or ‘‘Tromp curve”, which indicates the per-
parameters of the efficiency curve that will be presented in the fol-
centage reporting to either the underflow or overflow streams for
lowing sections.
a given feed size (Pastala, 1975). Fig. 6 depicts the actual and the
The measured size distributions together with the weight of the
corrected efficiency curves plotted for underflow stream. The x-
samples were used to perform mass balancing studies hence the
axis denotes to the particle size and y-axis denotes to the probabil-
errors arising from the experimental studies were distributed. In

Fig. 3. Feed size distributions of the materials tested. Fig. 5. Measured and calculated cumulative passing values.
 O. Altun et al. / Minerals Engineering 93 (2016) 50–56 53

Fig. 6. Actual and corrected efficiency curves (Tromp curve) of an air classifier.
Fig. 7. Actual efficiency curves of the classifier for a given operating condition.

ity for being separated as fines and coarse. For most of the cases,
the y-axis does not reach to 0% and this portion is called as bypass
fraction. The literature reports that this fraction is influenced by
the dust load of the classifier feed (Strasser et al., 1997; Altun
and Benzer, 2014) as well as the fines agglomeration owing to
the poor feed dispersion (Nakach et al., 2004). In general, when
the performances of different classifiers are to be compared, this
fraction is eliminated and the corrected efficiency curve is plotted
then the evaluations are carried out based on the corrected cut size
parameter. The studies have shown that this parameter correlated
with the operating conditions of air classifier i.e., wheel speed, air
amount, feed fineness (Altun and Benzer, 2014). The sharpness of
the curve is also the performance indicator for the classification
process. The steeper the curve, the sharper the separation. Finally,
initial rise of the curve after the minimum efficiency is defined as
the fish hook behaviour, which can be attributed to the agglomer-
ation of particles or the particles rebounding from the classifier
wall (Eswaraiah et al., 2012).
Fig. 8. Reduced efficiency curves of the classifier for a given operating condition.
So far, several mathematical expressions defining the Tromp
curve have been developed. Within the study focus was given on
the Whiten’s model (Napier-Munn et al., 1996) since Altun material properties i.e., fluidity of the bulk material, which is to be
(2007) compared the accuracy of different equations and found discussed in the following sections.
out that the Whiten’s approach fitted the curve with the least
sum of squares of deviations, in particular at fine sizes. Within
3.1. Modelling of the cut size and bypass parameters
the scope, correlations were developed between the operating
parameters of the laboratory scale air classifier and the parameters
The cut size of the operation is mainly influenced by the wheel
of Whiten’s equation, then integrated into an existing model struc-
speed and the air flow rate of the classification process (Benzer
ture proposed by Altun and Benzer (2014).
et al., 2001; Yu et al., 2014). It is also reported that the feed fine-
ness has effects on the performance of the air classifiers (Altun
3. Results and discussion and Benzer, 2014). Within the scope, it was not taken into consid-
eration since similar feed size distributions were arranged
Prior to performing the modelling studies, the efficiency curves throughout the tests (Fig. 3). Fig. 9 illustrates the variation of the
of the classifier (Nageswararao, 1999a,b) were plotted and the vari- corrected cut size (d50c) parameter that is directly proportional to
ations in the curves were observed. Figs. 7 and 8 depict the actual the ratio of air flow rate and wheel speed. Furthermore, it is seen
and the reduced efficiency curves obtained at the same operating that the trends vary systematically with the densities of the sam-
conditions. ples. The higher the density, the lower the corrected cut size owing
Fig. 7 illustrates the actual curves of the classifier obtained for to the increased effect of both the gravitational and the centrifugal
each of the sample. As can be understood, the main differences forces. Consequently, a density dependent mathematical model for
between the curves are in their bypass and cut size values. In brief, d50c parameter can be developed as given in Eq. (5).
the magnetite sample, which has the highest density, has the high-
d50c ¼ 2727  x  3:05  density þ 7:72 ð5Þ
est bypass value on the contrary the coal sample has the lowest
value. Similar conclusions can be drawn for the cut size parameter, The literature reports that the bypass of the classification pro-
which is also density dependent. When the shapes of the reduced cess is affected mainly by the dust loading of the classifier feed
efficiency curves are considered (Fig. 8), it is seen that the sharp- stream since the higher the dust loading, the higher the bypass
ness of the separation and fish hook of the curves change depend- fraction (Onuma and Ito, 1994; Altun and Benzer, 2014). Within
ing on the sample processed. This change can be attributed to the the study, the feed rate was constant and only the air flow rate
54 O. Altun et al. / Minerals Engineering 93 (2016) 50–56

Fig. 9. The variation of d50c with operating parameters and densities of the samples.
Fig. 12. The fluidity test results of the samples.

and wheel speed parameters were changed thus the effects of Table 5
these parameters on C parameter (Eq. (4)) were investigated. As Calculated fluidity indices (f) of the samples.
illustrated in Fig. 10, C parameter (100-bypass) is directly propor-
Sample f
tional to the ratio of air flow rate and wheel speed. In addition, it is
found to be density dependent since it decreases gradually from Clinker 0.030
Magnetite 0.008
the lightest to the heaviest sample. Coal 0.005
Fig. 10 also indicates that the curves are fitted to the logarith- Copper ore 0.002
mic equation (Eq. (6)) and parallel to each other where the slopes

Fig. 10. The variation of C with the operating parameters of the classifier and the
densities of the samples.
Fig. 13. The variation of alpha parameter for different materials.

Fig. 11. Reproducibility of the fluidity test results of the clinker sample. Fig. 14. The variation of beta parameter for different materials.
 O. Altun et al. / Minerals Engineering 93 (2016) 50–56 55

Fig. 15. Density dependent bypass and corrected cut size models.

Fig. 16. Fluidity dependent sharpness of separation (a) and the fish hook (b) models.

are the same and the constants of the equation change with the U: Cumulative passing (%) at the maximum sieving time.
densities of the samples.
Fig. 12 depicts the results of the fluidity tests of the samples
C ¼ 23:8  lnðxÞ þ 209:84  e0:016density ð6Þ
where the time corresponds to 50% passing is also pointed out.
In addition, the fluidity indices of each sample were calculated
3.2. The studies on the shape of the efficiency curve from Eq. (7) and are given in Table 5. The results imply that the
order of fluidity from the highest to the lowest is as clinker, mag-
The shape of the efficiency curve is expressed by two parame- netite, coal, and copper ore, hence copper ore sample is said to
ters which are; the sharpness of separation (a) and the fish-hook have higher tendency of forming agglomerates.
(b). The sharpness parameter was correlated with the operating The modelling studies indicated that the variations in a and b
parameters and the design features of the classifiers (Altun and parameters could be correlated with the fluidity indices. Fig. 13
Benzer, 2014). The fish-hook behaviour was attributed to particles’ illustrates the variation of a parameter with the operating param-
tendencies of forming agglomerates as a result of the surface eters of the classifier adjusted for each of the sample. It is seen that,
charges of the particles (Napier-Munn et al., 1996; Guizani et al., increasing the ratio of air flow rate to wheel speed results in having
2014). In fine grinding applications, grinding aids are utilized to sharper classification. In addition, it is understood that it decreases
neutralize the surface charges of particles hence both the grinding gradually from the sample having the highest fluidity index (clin-
and the classification efficiencies are improved (Klimpel and ker) to the sample having the lowest value (copper ore). The study
Manfroy, 1978). Toprak et al. (2014) reported that the fish hook concludes that as the fluidity characteristics of the samples
was reduced when the grinding aids were utilized. Consequently, improve, sharper classification is achieved since the particles are
a testing approach expressing the agglomeration tendency of the more dispersed.
samples can be used in defining the fish hook of the classification. As seen in Fig. 13, all of the trends are fitted to Eq. (8) where a
Within the context of the study, a method proposed by Jolicoeur parameters are the same and the constants of the equations
et al. (2007), who developed a correlation between the fluidity (parameter, b) can be correlated with the fluidity indices of the
characteristics of the bulk material and its agglomeration ten- samples.
dency, was applied. In this approach, the material was sieved on
0:093
a vibrating sieve and the retained percentage on this sieve is Alpha ¼ 274:95  x0:649fluidity ð8Þ
weighed at different time intervals. Then it is plotted and a fluidity
The variations of beta parameter with operating conditions are
index is calculated (Eq. (7)). In this study, the reproducibility of this
methodology was tested with 106 lm sieve and illustrated in illustrated in Fig. 14. As can be understood, the trends are parallel
to each other and, the higher the ratio of air flow rate to wheel
Fig. 11.
  speed, the lower the beta value. It is due to the utilization of higher
1 amount of air that is thought to be effective on dispersing the par-
m¼U 1 ð7Þ
expðf  tÞ ticles and having reduced agglomeration. Conversely, increasing
the wheel speed increased the tendency of forming agglomerates,
where;
owing to the increased number of particle–particle collisions
t: Sieving time
around the rotor structure; hence higher value of beta was
m: Cumulative passing (%) at time t
obtained. Fig. 14 also implies that the beta parameter is material
f: Fluidity index
dependent and the change can be attributed to the variation in
56 O. Altun et al. / Minerals Engineering 93 (2016) 50–56

the fluidity indices as it decreases from the sample having the low- showed that the variation in the densities of the samples had an
est fluidity index (copper ore) to the highest (clinker). As a conclu- influence on cut size parameter since it was concluded that
sion of the study, Eq. (9) was derived to express the beta increase in the density decreased the cut size owing to the
parameter. improved effect of both gravitational and centrifugal forces. At
0:131 the same operating conditions, d50c of the magnetite sample
Beta ¼ 136:5  x þ 0:873  fluidity ð9Þ (5.1 g/cm3) was the lowest and it increased gradually for the cop-
per ore (4.1 g/cm3), clinker (3.05 g/cm3) and coal (1.4 g/cm3) sam-
3.3. Integrating the multi-component model structure to an existing ples respectively. Similar conclusions were also drawn for the
model bypass parameter. It is expected that, the density and the bypass
parameters are directly proportional to each other due to the
Within the scope of the study, the developed multi-component improved influence of gravitational force in the classification zone.
correlations were integrated into an existing model structure pro- Within the study, at the same operating conditions, the bypass of
posed by Altun and Benzer (2014) who studied on clinker grinding the classification was the highest for the magnetite sample and it
circuits. In this context, the parameters that are corrected cut size, decreased gradually for the copper ore, clinker and coal samples
bypass, alpha and beta of the Whiten’s equation were considered. respectively. As a result, a density dependent model structure
Since this study showed that the corrected cut size (d50c) and was developed for both the corrected cut size and the bypass
bypass parameters (100-C) were density dependent, the trends parameters.
obtained by Altun and Benzer (2014) were adjusted accordingly The shape of the efficiency curve is expressed mainly by its
as illustrated in Fig. 15. As a summary, the current state of the sharpness (a) and the fish hook (b) parameters. These two param-
model is capable of predicting the two parameters depending on eters were found to be dependent on the operating conditions as
the change in the densities of the samples. well as the fluidity characteristics of the tested samples. In order
This study showed that the parameters expressing the shape of to determine the fluidities of the samples an approach proposed
the efficiency curve, which are a (sharpness of separation) and b by Jolicoeur et al. (2007) was used and found out that the clinker
(fish hook), could be correlated with the fluidity indices of the sam- sample has the highest fluidity that is followed by the magnetite,
ples. Fig. 16 depicts the adjusted model structure of the air classi- coal and copper ore samples. The study concluded that the higher
fier (Altun and Benzer, 2014) when the fluidity results are taken the fluidity, the lower the fish hook (b) owing to having reduced
into consideration. agglomeration. On the other hand, increase in fluidity resulted in
Where; having sharper classification (increased a).

D: Classifier chamber diameter (m) References

DL: Dust loading of classifier feed (kg/m3)
Altun, O., 2007. Comparison of different efficiency curve approaches in modelling of
AF: Air flow used in the classification process (m3/h) air classifiers MSc. Thesis. Hacettepe University, Ankara, Turkey.
RS: Rotor speed of the classifier (m/s) Altun, O., Benzer, H., 2014. Selection and mathematical modelling of high efficiency
F: The amount of (36 + 3) lm size fraction in separator feed (t/ air classifiers. Powder Technol. 264, 1–8.
Benzer, H., Ergun, L., Lynch, A.J., Oner, M., Gunlu, A., Celik, I.B., Aydogan, N., 2001.
h). Modelling cement grinding circuits. Miner. Eng. 14 (11), 1469–1482.
Duda, W.H., 1985. Cement-Data-Book. International Process Engineering in the
It should be emphasized that, the multi-component air classifier Cement Industry. Bauverlag GmbH, Wiesbaden and Berlin..
Eswaraiah, C., Angadi, S.I., Mishra, B.K., 2012. Mechanism of particle separation and
model structure proposed within the study needs to be verified
analysis of fish-hook phenomenon in a circulating air classifier. Powder
with industrial data where different components are processed. Technol. 218, 57–63.
Guizani, R., Mokni, I., Mhiri, H., Bournot, P., 2014. CFD modelling and analysis of the
fish-hook effect on the rotor separator’s efficiency. Powder Technol. 264, 149–
4. Conclusions 157.
Jolicoeur, C., Morasse, S., Sharman, J., Tagnit-Hamou, A., Slim, F., Page, M., 2007.
This study initially aimed to show the influences of the material Polyol-type compounds as clinker grinding aids: influence on powder fluidity
and on cement hydration. In: 12th International Congress on the Chemistry of
characteristics e.g. overall density, fluidization, on performance of Cement.
the laboratory scale air classifier, then developing a model struc- Klimpel, R.R., Manfroy, W., 1978. Chemical grinding aids for increasing throughput
ture including these properties as well as the operating parameters in the wet grinding of ores. Ind. Eng. Chem. Process Des. Dev. 17 (4), 518–523.
Klumpar, I.V., Currier, F.N., Ring, T.A., 1986. Air classifiers. Chem. Eng. 93 (5), 77–92.
of the classifier. Throughout the evaluations, the efficiency curve
Nageswararao, K., 1999a. Normalisation of the efficiency curves of hydrocyclone
parameters i.e., bypass, corrected cut size, alpha and beta, were classifiers. Miner. Eng. 12 (1), 107–118.
taken into consideration and correlated with the material proper- Nageswararao, K., 1999b. Reduced efficiency curves of industrial hydrocyclones-an
analysis for plant practice. Miner. Eng. 12 (5), 517–544.
ties. Within the context of the study, a laboratory scale air classifier
Nakach, M., Authelin, J.-R., Chamayou, A., Dodds, J., 2004. Comparison of various
was utilized and magnetite, copper ore, clinker, coal samples hav- milling technologies for grinding pharmaceutical powders. Int. J. Miner. Process.
ing similar size distributions were processed at different operating 74 (10), 173–181.
conditions. Napier-Munn, T.J., Morrell, S., Morrison, R.D., Kojovic, T., 1996. Mineral
comminution circuits-their operation and optimization. JKMRC Monograph
The study implied that depending on the change in the operat- Series in Mining and Mineral Processing, Brisbane, Australia.
ing conditions of the air classifier, the efficiency curves of the clas- Onuma, E., Ito, M., 1994. Separators in grinding circuits. ZKG 47 (9), 535–542.
sification operation varied hence correlations could be developed Pastala, A.L., 1975. Contribution to the design of air separation equipment for closed
circuit grinding systems. Cem. Technol., 119–1249
and the classifier could be modelled. In addition, the study showed Strasser, S., Somani, R.A., Dembla, A.K., 1997. Improvements in the production of
that the material characteristics were also of crucial importance raw meal and cement by the combined use of roller press and V-separator. ZKG
since it was proved that the variation in the density and the fluidity 50 (3), 140–146.
Toprak, N.A., Altun, O., Aydogan, N., Benzer, H., 2014. The influences and selection of
properties of the samples resulted in the change in the efficiency grinding chemicals in cement grinding circuits. Constr. Build. Mater. 68 (15),
curves of the classification. The study indicated that the density 199–205.
had major effects on the efficiency curve since it changed the TS EN 1097-7. Tests for mechanical and physical properties of aggregates-Part 7:
Determination of the particle density. Pycnometer method.
bypass and the corrected cut size parameters of the classification.
Yardi, J., 2005. Separator technology. World Cem. 36 (10), 87–104.
It is a well-known fact that the cut size of the classification is influ- Yu, Y., Liu, J., Zhang, K., 2014. Establishment of a prediction model for the cut size of
enced by both the rotor speed and the air flow rate. This study also turbo air classifiers. Powder Technol. 254, 274–280.