Professional Documents
Culture Documents
Revisedmanuscript
Revisedmanuscript
net/publication/322289513
CFD simulation of the preheater cyclone of a cement plant and the optimization
of its performance using a combination of the design of experiment and multi-
gene genetic programming
CITATIONS READS
17 6,969
3 authors, including:
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Ali Mohebbi on 11 January 2018.
Abstract
Hurriclon cyclone is a specially designed preheater cyclone with two outlet connector pipes of
cleaned gas in the cement industry. In Kerman cement plant, Iran, the initial structure of this
cyclone was changed. This caused a decrease in the cyclone efficiency. In this study, to optimize
the changed cyclone performance, one of the twin cyclones in the first-stage of the preheater
tower, which had the most significant effect on particle separation from gas was simulated and
validated by computational fluid dynamics. Using the design of experiment based on the
simulation results, the effects of three dimensions (vortex-finder length, cylinder height, and
cone tip diameter) were investigated on cyclone performance. The turbulent gas flow inside the
cyclone was modelled using the Reynolds stress model due to the swirling flow inside the
cyclones. The discrete phase model was used to calculate the trajectory of particles. It was
observed that because of high gas inlet velocity and particle density as well as the geometry of
the preheater cyclone, particles larger than the critical diameter continue spinning in the cyclone.
Corresponding author: Tel & Fax: +983432118298
E-mail addresses: amohebbi2002@yahoo.com, amohebbi@uk.ac.ir
The Multi-Gene Genetic Programming (MGGP) was used to obtain two equations for efficiency
and pressure drop in order to optimize the preheater cyclone performance. For this purpose, two-
objective optimization using the Genetic Algorithm (GA) was performed. The optimization
results showed that by using the optimized dimensions for the preheater cyclone, the pressure
The application of cyclone separators in separating particles from gas flow began hundreds of
years ago. Owing to the wide range and simplicity of the cyclone structure that leads to a low
cost of investment and maintenance, they are still widely used in the chemical and petrochemical
industries. The main parameters of cyclone performance are separation efficiency and pressure
drop [1]. Many efforts have been made to derive an appropriate mathematical model for the
study and prediction of flow behaviour as well as to find suitable cyclone geometries. However,
due to the hydrodynamic complexity of cyclones, no exact mathematical model for cyclones has
been offered yet. For numerical solutions, CFD is a proper tool [2, 3]. The key to the successful
use of CFD for finding turbulent flow pattern depends on the selection of the right turbulence
model. Selecting a suitable turbulence model has been studied by many researchers [2, 4]. The
result of the Reynolds stress model (RSM) among the turbulence models has a very good
agreement with the experimental data [5–8] since the isotropy of the flow assumption is ignored.
In addition, large-eddy simulation methodology can predict turbulent flow inside the cyclone [4,
9, 10].
The pressure drop and the cut-diameter in cyclones are important objective functions, to be
researches have been conducted to investigate the effect of geometrical parameters on cyclone
performance [11–13]. Safikhani et al. [14] obtained the pressure drop and cut-diameter by CFD
and then optimized the cyclone performance with the assistance of multi-objective optimization;
the artificial neural network was then used to calculate the objective function. Elsayed and Lacor
performed many studies in order to optimize the cyclone [15–17]. They used multi-objective
optimization for the performance of the cyclone [18]. Two radial-basis function neural networks
were used to model the pressure drop and the cut-diameter. Their results showed a noticeable
influence of the vortex finder diameter and its length, the inlet width, and the total height on the
efficiency of the cyclone. Fathizadeh et al. [19] used Leith and Licht theory to prepare some
algorithms to obtain the dimensions of the Stairmand cyclone. Vortex finder length was
optimized using genetic algorithm to prevent the direct escape of particles from vortex finder.
Owing to environmental concerns-and as the cement industry is one of the major producers of
greenhouse gases-strict regulations have been enforced in this industry. In recent years, different
researches on the various techniques that improve the cement production process and reduce
resultant pollution [20] have been carried out. Mikulcic et al. [21] showed the application of
CFD in the implementation of the calcination reaction, which is carried out in calciner in order to
reduce carbon dioxide emissions in the cement industry. Mikulcic et al. [22] numerically studied
gas-solid flows in a cement cyclone using the Eulerian-Lagrangian approach. The heat exchange
phenomenon in the cyclone, arising from the heat exchange between the gas-solid phase and the
endothermic calcination reaction, was investigated. Cristea and Conti [23] studied the heat
exchange between the gas and the solid phase in a cement plant preheater. They concluded that
the most of the heat distribution occurs within the duct, and the temperatures of the gas and
particles in the cyclones remain approximately constant and equal. Ferrite and Ari [9] optimized
the preheater cyclone by means of the Taguchi method. They optimized the cyclone performance
by changing the four parameters of diameter, vortex-finder length, velocity inlet, and the
saving can be achieved. The role of the first-stage cyclones is significantly important as
specifying the separation efficiency. Wasilewski and Duda [24] for optimizing the first-stage
cyclones proposed guidelines depending on the objective function. In another study [25] they
optimized the geometry of cyclone. Mariani et al. [26] optimized the length and angle of vortex
finder to increase the separation efficiency and decrease the exit gas temperature.
2. Background
In traditional cement manufacturing technology, the preheater system is positioned before the
rotary kiln in order to increase the heat transfer between the raw materials and hot gases. Raw
materials in the preheater are gradually dried, heated, and calcined by moving against the flow of
the hot gases coming from the rotary kiln. In Fig. 1, a rotary kiln with a suspension preheater
similar to that one from Kerman cement plant is shown. Preheaters are built in several stages in
Fig. 1
The preheater of the Kerman cement plant, Iran, was made by the PMT–Zyklontechnik Gmbh
Company (the company name was changed to 'A TEC' in 2005). The cyclone used by this
company is a special type named Hurriclon. Hurriclon has been patented for its technically
superior characteristic of the double immersion tube principle and its integrated vortex-finder
vanes. By adding vortex finder vanes, this company has reduced the cyclone pressure drop by
30%. The reason for the improvement of its performance is a result of collision prevention of the
cyclone outlet gases with the swirling mixture of gas and particles inside the cyclone [28].
However, after two years of preheater operation, both the vortex-finder vanes broke down, which
resulted in the plant staff having to make changes in the twin cyclones of the first-stage. They
removed the two vortex-finder vanes and the immersion tube that had been driven out from the
bottom. Fig. 2 (a) and (b) illustrate the Hurriclon cyclone before and after modernisation
respectively.
Since the structure of the preheater cyclones of the Kerman cement plant was significantly
varied, the purpose of this study was to simulate one of the twin cyclones in the first-stage of the
preheater by CFD. In order to increase the performance, a two-objective optimization study was
performed. The vortex-finder length, the cylindrical height, and the cone-tip diameter of this
cyclone were optimized by using the Genetic Algorithm (GA). In order to model the
relationships between the performance parameters (i.e., the pressure drop and the efficiency) and
the geometrical dimensions, multi-gene genetic programming (MGGP) was employed. The
effect of these dimensions on the performance was obtained using the design of experiments.
Fig. 2
For an incompressible and steady-state fluid flow, the continuity equation for the mean flow
is as follows [29]:
ui
0 (1)
xi
p ui u j 2 uk
x j
ui u j
xi x j x j
xi
ij
uiu j
(2)
3 xk
x j
The last term is defined as the Reynolds stress tensor, which reflects the effects of turbulent
fluctuations in the fluid flow. The RSM used in this study is required for solving equations for
each of the Reynolds stresses. In the Reynolds stress model, the transport equation is written as
follows [29]:
uk uiu j Dij Pij ij ij S (3)
xk
The expression on the left-hand side of the equation represents the convective transport term.
Five terms on the right-hand side of the equation are respectively diffusion, production, pressure
strain, dissipation, and the source terms. The final transport equations of RSM can be written as
follows [19]:
t u j u u u u u
uk uiu j ui u j uiuk uj uk i p( i j ) 2 i j (4)
xk xk k xk xk xk x j xi xk xk
k2
t C (5)
where, C is 0.09.
4. Simulation procedure
The cyclone was a scroll type in which the radius of the cylindrical part (R) was 1.26 m and
the radius of the scroll part (Rs) was 1.722 m, which was located at a distance of r 0.64 m from
the cylinder centre. Cyclone dimensions are provided in Table 1; it consists of cyclone diameter
(D), inlet height (a) and width (b), vortex-finder diameter (De) and length (S), the height of outlet
pipe (l) (it was selected greater than usual, in order to be sure of accessing fully developed flow
in the outlet and preventing recirculation flow), cylinder height (h), part of the scroll (hs) and
conical (Hc) height, the cyclone total height (H) and the cone-tip diameter (B). Fig. 3 shows a
Table 1
Hexahedral and tetrahedral meshes were generated by using the GAMBIT commercial
software. Fig. 4 shows simulated cyclone; Fig. 4(b) shows part of the cyclone, which is a gridded
Fig. 3
Fig. 4
The velocity-inlet boundary condition for the inlet gas, the outflow for outlet gas, and the wall
condition for the outlet particles were considered. The atmospheric pressure was 81395 Pa, the
inlet gas volumetric flow rate was 28.22 m3/s, and the gas inlet velocity was 14.7 m/s. The
density and the viscosity of the gas were 0.43 kg/m3 and 2.1 × 10-5 kg/m.s respectively. The
density of particles was 2710 kg/m3. The turbulent intensity was 2.3% and the hydraulic diameter
was 1.3 m. The gas inlet temperature was 420 ᵒC; heat discussion had not been performed in this
research because the inlet gas temperature was lower than the decomposition temperature [22,
30]. This study was based on the Eulerian-Lagrangian approach using the discrete phase model
(DPM), so that while solving the time-averaged Navier-Stokes equations, the fluid phase was as
a continuous phase and the solid phase was as a discrete phase. More details about the equations
of motion of particle can be found elsewhere [19]. If we plot CD versus Re for Schiller and
Nauman [31], Schuh et al. [32] and Morsi and Alexander [33] on one figure, it is found that the
results are close to each other. Therefore, in this study, the drag coefficient was calculated
according to Morsi and Alexander [33] method. The second phase in DPM has to occupy a low
volume fraction (less than 10–12%). It was assumed that there was no particle-particle
interaction and the particles did not affect the flow field, which is known as one-way coupling
[29, 34]. To consider the effect of turbulence fluctuations on particles, the Discrete Random
Since the problem convergence with the RSM is difficult and time-consuming, at first the
flow field had been derived with the k–ε standard model and then changed to the RSM [36]. The
The governing equations were solved numerically by the finite-volume method using the
commercial CFD code FLUENT 6.3.26. For pressure-velocity coupling, the SIMPLEC algorithm
was used. Because of highly swirling flows in the cyclone, PRESTO was used to interpolate the
pressure. To measure the flow in the cyclone, the momentum equations’ QUICK scheme was
recommended. The Reynolds stress equations were solved based on the second-order upwind
scheme discretization. In this study, for all the simulations, these discretization schemes were
used.
solutions for optimization. Genetic algorithms apply the principle of Darwin’s theory.
(According to Darwin’s theory, the generations that have better characteristics than other
generations will have a better chance of survival and reproduction; their superior characteristics
are also transmitted to the next generation). Genetic algorithm with random initial population
generation checks the problem space stochastically. Then, the fitness of each individual in a
population is investigated and the algorithm selects the fittest individuals, and then, genetic
operators affect some of the population and the next generation is generated. These operators are
This study is performed to optimize the preheater cyclone of the Kerman cement plant since
this cyclone is a special case and there is no empirical relationship between its geometries and
performance (i.e., pressure drop and collection efficiency). Thus, a relationship was derived. In
order to achieve this purpose, Multi-Gene Genetic Programming (MGGP) was used. MGGP is a
modified version for model structure selection, which is combined with the least squares
technique for parameter estimations. MGGP is of the symbolic regression class aimed at finding
an explicit relationship between one or more inputs and one output by using mathematical
symbols, variables, and functions. To find equations in symbolic regression, a large and
representation, despite common optimization methods such as the genetic algorithm, where
potential solutions are presented with numbers. Each gene consists of a set of F functions and T
Terminals. F includes the basic arithmetic operators (such as −, +, exp, etc.) while the T
terminals involve input variables and constants. Potential solutions can be labelled as branches in
a labelled tree with ordered branches that function and are used in a function set with arguments
in the terminal set. The GP gives a platform with genetic operators such as random generation,
mutation, and crossover that produce, modify, and select individuals in a given population. Any
symbolic model of MGGP is a weighted linear combination of the output of some GP trees. Each
of these trees can be considered as a gene, which is a representative GP of a tree. More details
The computational results of the three different meshes for the preheater cyclone pressure
drop and efficiency are provided in Tables 2 and 3 respectively. Since the pressure drop and
cyclone efficiency difference between both the sequential meshes were minimum (less than 5%),
the CFD results are independent of the mesh size. To reduce computational times, the mesh
number of 2208864 was used in this study. The last row in Tables 2 and 3 represent experimental
pressure drop and efficiency of the cyclone respectively. The experimental pressure drop was
measured by a portable device (testo 512) and of course, the monitoring of operating parameters
was available. The amount of dust returned from the cyclones (11 ton/h) could be weighed by
truck. Therefore, considering the amount of kiln feed (13.1 ton/h), the experimental efficiency
could be calculated. Since there was a little difference between simulation results and
Table 2
Table 3
Fig. 5 (a) and (b) show the contours of static pressure and velocity magnitude in the sections
z=9.5 m and x=0 respectively. Pressure decreases from the wall to the centre and the area with
the lowest pressure and the highest velocity occurs below the vortex finder, which causes upward
suction at the centre that makes the outer vortex in the cyclone change its direction and move
Fig. 5
In order to compare Stairmand and preheater cyclones, Stairmand cyclone was designed using
the algorithm written by Fathizadeh et al. [19]. It was found that the total height of the Stairmand
cyclone was greater than that of the old structure while the efficiency will increase 2.5%.
However, due to technical and economic reasons, preheaters tower should be designed with a
limited height.
It was observed that when the inlet cyclone particle diameter becomes greater than a certain
value, the particles spin initially downward and then keep spinning at a certain height on a fixed
path near the wall. This value is called the critical diameter. This phenomenon was also reported
by Wang et al. [41]. A critical diameter of 55 μm was obtained for the simulated cyclone. Fig. 6
demonstrates particle spin with a diameter larger than the critical diameter (a) in operation [41]
Fig. 6
In one-way coupling models, particles do not interact with each other, but due to particle
collisions, particles larger than the critical diameter are removed from the cyclone. At high
temperatures, if the particles stagnate on the wall, the permanent deposition, which decreases the
performance of the cyclone, may be formed. Furthermore, the avalanche phenomenon occurs due
parameters were investigated. In each case, one parameter was examined while the remaining
parameters were taken at the same value, which were the parameters for preheater cyclone,
according to Table. 1 and section 4.2. In general, it can be concluded that by increasing inlet
velocity (U) and particle density (ρp), a reduction in critical diameter is expected, and with a
decrease in cylinder height (at constant total height) and an increase in the cone-tip diameter, the
critical diameter increases. Thus, by increasing the slope of the cyclone conical part (i.e.,
reducing the angle with the vertical axis, which is the Z-axis here), the critical diameter tends to
higher values.
Table 4
Since the diameter of the cylindrical part, the vortex-finder diameter, the inlet dimensions and
the overall height of the preheater cyclone cannot be changed economically (changes of these
values will cause more changes in the entire system), the selected variables for the design of
experiments (DOE) were vortex-finder length, cylindrical height, and cone tip diameter. The
DOE method was used to investigate the effect of the mentioned dimensions on the performance
of the cyclone. DOE is a data collection process while one or more variables are changed. In
DOE, the designer is usually interested in investigating the effect of some parameters on the
output of a process. The analysis of variance determines the impact of each parameter on the
behaviour of the system; the appropriate method is selected depending on the limitations of a
problem [43]. In this study, the full factorial design method was used: all three mentioned
dimensions were studied in three levels, then, in accordance with the full factorial design
method, 3 × 3 × 3 simulation runs with at least two repetitions in the results were performed.
Table 5
The averaged pressure drop and efficiency for each simulation run are listed in Table 6. In the
table, the cyclone number is such that the first number represents the vortex-finder length, the
second number is the cylindrical height, and the third number is the cone-tip diameter; for
example, cyclone number 132 means: the vortex-finder length, cylindrical height, and cone-tip
diameter are 1.625, 4.553, and 0.628 m respectively. The rest of the main cyclone dimensions
remain constant, as given in Table 1. Owing to lack of experimental data, the Grid Convergence
Index (GCI) method [44] was used to verify the results of 27 cyclones. All the calculations were
Table 6
The design of experiments was performed in the Minitab 16 software; the analysis of variance
(ANOVA) for pressure drop and efficiency are provided in Tables 7 and 8 respectively. A
significant level of P>0.05 (95% confidence) was used in all the tests. If the reported p-value for
any term is smaller than the confidence level, then it is said that the term has a significant effect
on the response [16]. According to two tables, all the three factors (dimensions) and their
Table 7
Table 8
The impacts of factors (dimensions) on pressure drop and efficiency with the assistance of the
and the efficiency increased. By increasing the vortex-finder length, the probability of the input
particles exiting from the bottom of the vortex finder is lowered, and this leads to greater
efficiency.
b) when the factor h was changed from Level 1 to 3 (with increasing h), both the pressure drop
and efficiency decreased. By increasing cylindrical height (at constant total height), the conic
height decreased. Particles separation from the gas in the cyclone occurs more in the conical part,
where the radius is reduced. Therefore, by decreasing the height of the cone, the efficiency also
decreased.
c) when factor B was changed from 1 to 3 (by increasing B), both the pressure drop and
efficiency decreased. In Fig. 8, the velocity contours for three cyclones-121, 122, and 123-at
section x=0 are compared (the difference among these three cyclones is only in the cone-tip
diameter). As one can see from this figure, by increasing the diameter, the velocity of the
cyclone decreases. This causes a decrease in the efficiency. This finding is in accordance with
Ref. 45.
Fig. 7
Fig. 8
In order to model the relationship between the performance of the cyclone and the studied
dimensions, MGGP was used from the toolbox of the computer program GPTIPS, which has
been developed under the MATLAB software. Euler number and efficiency as a function of
Ef
f S / D, h / D, B / D
(7)
Ef xii
where ΔP is the pressure drop and v is the velocity characteristic, xi is the mass fraction of the
particles with the same diameter and i is the collection efficiency for that diameter. The inlet
particle size distribution is given in Appendix A. According to Eqs. 6 and 7, the three above
mentioned variables and the two dependent variables (i.e., the Euler number and the efficiency)
respectively were considered as inputs and outputs of the models. A total of 22 data from 27 data
were used as the training set, while five data were used as the testing set. In this study, the
Table 9
The best results on predicting the Euler number and efficiency (based on MGGP) with
Ef 2.28X13 (0.4903X 2 16.49) X12 (0.4903X 2 0.4903X 3 3.938) X1 4.108( X 3 X 2 ) 81.72 (9)
where X1 = S/D, X2 = h/D and X3 = B/D. Figs. 9 and 10 show MGGP prediction versus the
calculated values by CFD for the Euler number and efficiency respectively in the training and the
testing data set. As one can see from these figures, the linear correlation coefficients for the Euler
number and efficiency are 99 and 99.98%, respectively, in the training data set. These figures
also show that these coefficients for the aforementioned parameters are 96.67 and 99.75%
respectively, in the testing data set. This indicates that the equations obtained have favourable
accuracy.
Fig .9
Fig. 10
A two-objective optimization was used by the genetic algorithm to find the lowest pressure
drop and the highest efficiency of the preheater cyclone. Two objective functions were defined:
Eq. 8 (Euler number) for minimizing and Eq. 9 (efficiency) for maximizing or the reciprocal of
Eq. 9 for minimizing. The Pareto front (non-dominated points) is presented in Fig. 11. Pareto
front represents the set of responses in the search space that are better than the other responses.
In other words, in the Pareto front, one solution cannot be dominated by any other member of
this set [18]. The final selection of the optimal response from the Pareto front results depends on
the awareness of the problem. The best answer choice obtained for dimensions, pressure drop
and efficiency is given in Table 10, which is the point c in Fig. 11. Table 11 lists the dimension
Fig. 11
Table 10
Table 11
The optimized cyclone was simulated again, and the pressure drop and the efficiency were
obtained as 548.42 Pa and 94.2%, respectively. Fig. 12 shows the comparison between the
tangential velocity of the optimized cyclone and the preheater cyclone at section z=7 m versus
the radius of the cyclones. Tangential velocity is the dominant component of gas flow in the
cyclone, which affects the centrifugal force; with an increase in velocity, the force increases.
Thus, by increasing tangential velocity, the efficiency increases. It is observed from Fig. 12 that
the maximum optimized cyclone tangential velocity is greater than the preheater cyclone;
therefore, greater efficiency for the optimized cyclone is expected. By using the dimensions of
the Kerman cement plant preheater cyclone obtained from the two-objective optimization in
CFD simulation, the pressure drop reduces by about 2.2% and collection efficiency increases by
about 13.4%. The pressure drop and the efficiency of the current cyclone are respectively 560 Pa
and 84%.
Fig. 12
7. Conclusions
The first-stage cyclones' efficiency in the preheater system of a cement plant is very
important. In this study, one of the twin cyclones in the first-stage of the preheater in the Kerman
cement plant was investigated using a combination of CFD, the design of the experiment, and
multi-gene genetic programming. Since its structure had been changed, a decrease in its
efficiency had taken place. In the same pressure drop and volumetric flow rate, the Stairmand
cyclone efficiency was 2.5% greater than the current preheater cyclone while the Stairmand
cyclone was relatively larger. The effects of the three dimensions of the preheater cyclone were
investigated using the design of experiments. By increasing the vortex-finder height, the pressure
drop and the efficiency increase. On the other hand, by decreasing the cylindrical height and the
cone-tip diameter, the pressure drop and the efficiency increase. The MGGP was applied to
obtain two equations for calculating the Euler number and efficiency. These equations are able to
predict results with less error. Finally, these equations were used as objective functions for
preheater cyclone optimization. The optimized cyclone had 2.2% less pressure drop and 13.4 %
greater efficiency than the Kerman cement plant preheater cyclone. It could be concluded that
the modernisation of Hurriclon cyclone performed by the cement plant was rather impractical.
Because it had a negative effect on the cyclone performance. The new proposed structure
obtained by optimization technique was more efficient. Also, it is observed that particles which
are larger than the critical diameter spin initially downward, and then keep spinning at a certain
height on a fixed path near the wall. The critical diameter increases by decreasing the inlet
velocity, particle density, and by increasing the slope of the conical part of the cyclone.
Acknowledgements
The authors acknowledge the Kerman cement plant, Iran in allowing them to use their
experimental data.
Funding
This research did not receive any specific grant from funding agencies in the public,
Ef efficiency
Eu Euler number
P pressure (Pa)
X1 S/D
X2 h/D
X3 B/D
Greek symbols
ij Kronecker delta
ij source term
ij pressure-strain
μ viscosity (kg/m.s)
Abbreviations
GA Genetic Algorithm
of particle
(μm)
0-4.5 20.36
4.5-5.5 4.06
5.5-6.5 3.49
6.5-7.5 2.99
7.5-9 3.73
9-11 3.87
11-13 2.92
13-15.5 2.71
15.5-18.5 2.35
18.5-21.5 1.8
21.5-25 1.75
25-30 3.23
30-37.5 3.22
37.5-45 3.22
45-52.5 3.18
52.55-62.5 4.08
62.5-75 4.86
75-90 5.28
90-105 4.48
105-125 4.75
125-150 4.38
150-180 3.8
180-215 3.08
215-255 2.12
255-305 1.28
References
[2] W.D. Griffiths, F. Boysan, Computational fluid dynamics (CFD) and empirical modelling
of the performance of a number of cyclone samplers. Aerosol Science 27 (2) (1996) 281–
304.
[3] G. Wan, G. Sun, X. Xue, M. Shi, Solids concentration simulation of different size particles
[4] S.K. Shukla, P. Shukla, P. Ghosh, The effect of modeling of velocity fluctuations on
[5] T.G. Chuah, J. Gimbun, S.Y. Choong, A CFD study of the effect of cone dimensions on
(2006) 126–132.
[6] X. Gao, J. Chen, J. Feng, Numerical investigation of the effects of the central channel on the
flow field in an oil–gas cyclone separator, Computers & Fluids 92 (2014) 45-55.
[7] Y. Su, A. Zheng, B. Zhao, Numerical simulation of effect of inlet configuration on square
[8] A. Kępa, The efficiency improvement of a large-diameter cyclone – The CFD calculations.
[9] F. Ferit, V. Ari, Optimization of the preheater cyclone separators used in the cement
[11] Kh. Elsayed, Ch. Lacor, The effect of cyclone inlet dimensions on the flow pattern and
[12] R. Xiang, S.H. Park, K.W. Lee, Effects of cone dimension on cyclone performance, Aerosol
[13] K.S. Lim, H.S. Kim, K.W. Lee, Characteristics of the collection efficiency for a cyclone
with different vortex finder shapes, Aerosol Science 35 (6) (2004) 743–754.
cyclone separators using CFD and genetic algorithms, Computers & Chemical Engineering
[15] Kh. Elsayed, Ch. Lacor, Optimization of the cyclone separator geometry for minimum
pressure drop using mathematical models and CFD simulations, Chemical Engineering
[16] Kh. Elsayed, Ch. Lacor, Modeling, analysis and optimization of aircyclones using artificial
neural network, response surface methodology and CFD simulation approaches, Powder
[17] Kh. Elsayed, Ch. Lacor, Multi-objective optimization of gas cyclone based on CFD
simulation, in: ECCOMAS Thematic Conference, CFD & Optimization, Antalya, Turkey
(2011).
[18] Kh. Elsayed, Ch. Lacor, Modeling and Pareto optimization of gas cyclone separator
performance using RBF type artificial neural networks genetic algorithms, Powder
pressure cyclones for a gas city gate station using semi-empirical models, genetic algorithm
and computational fluid dynamics, Natural Gas Science and Engineering 26 (2015) 313-
329.
Stefanovic, The application of CFD modelling to support the reduction of CO2 emissions in
[22] H. Miculcic, M. Vujanovic, M. Sami Ashha, N. Duic, Large eddy simulation of a two-phase
[23] E.D. Cristea, P. Conti, CFD simulation of large dust collection cyclones positioned
[25] M. Wasilewski, L.S. Brar, Optimization of the geometry of cyclone separators used in
clinker burning process: A case study, Powder Technology 313 (2017) 293-302.
[26] F. Mariani, F. Risi, C.N. Grimaldi, Separation efficiency and heat exchange optimization in
[27] J.I. Bhatty, Innovations in Portland Cement Manufacturing, Technology & Engineering,
[30] B.R. Stanmore, P. Gilot, Review—calcination and carbonation of limestone during thermal
cycling for CO2 sequestration. Fuel Processing Technology 86 (16) (2005) 1707–1743.
[31] L. Schiller, and A. Naumann, Z. Ver Deut. In g., 77, 318 (1933).
[34] Oh, S. Choi, J. Kim, Numerical simulation of an internal flow field in a uniflow cyclone
[35] M. Azadi, M. Azadi, A. Mohebbi, CFD study of the effect of cyclone size on its
[36] G. Gong, Z. Yang, Sh. Zhu, Numerical investigation of the effect of helix angle and leaf
margin on the flow pattern and the performance of the axial flow cyclone separator, Applied
[37] Kh. Elsayed, Ch. Lacor, CFD modeling and multi-objective optimization of cyclone
geometry using desirability function, artificial neural networks and genetic algorithms,
Genetic Programming for Predicting the Sweet Natural Gas Compressibility Factor. Gas
[40] D.P. Searson, D.E. Leahy, M.J. Willis, GPTIPS: an open source genetic programming
toolbox for multi-gene symbolic regression. In: Proceedings of the international multi
[41] B. Wang, K. Chu, A. Yu, Numerical study of gas–solid flow in a cyclone separator, Applied
[42] H.F. Meier, K. Ropelato, M. Mori, Computational fluid dynamics CFD for cyclone
[43] D. Montgomery, design and analysis of experiments, sixth edition, GGs Book Services,
Atlantic Highlands.
[44] I. Celik, U. Ghia, P.J. Roache, p. Raad, Procedure of Estimation and Reporting of
Uncertainty Due to Discretization in CFD Applications, Fluids Engineering 130 (7) (2008).
[45] M. Azadi, M. Azadi, an analytical study of the effect of inlet velocity on the cyclone
b (m) a (m) Hc (m) H (m) h (m) hs (m) l (m) S (m) B (m) De (m) D (m)
Mesh Pressure
Relative error(%)
number drop(Pa)
2746807 525.9
Experimental
560 6.25
data
Table 3 Comparison of cyclone efficiency for three different meshes.
1722590 81 0.8
2746807 81.7
Experimental
84 2.78
data
Table 4 The effect of parameters on the critical diameter.
(μm)
U (m/s) 7 -
12 160
14.7 55
20 20
1500 80
2710 55
3300 48
a (m) 1.5 55
1.92 55
2.1 54
b (m) 0.8 54
1 55
1.2 54
De (m) 0.9 53
1.193 55
1.4 55
S (m) 1.625 55
2 55
2.345 56
h (m) 4.553 55
4.44 60
3.255 75
B (m) 0.314 42
0.428 55
0.5 93
Table 5 Selected dimensions for the design of experiments method.
Dimension (m) S h B
Level
2 2 4.44 0.628
3 2.345 4.553 1
Table 6 The averaged pressure drop and efficiency for 27 simulated cyclones.
( %)
311 88 554
Error 0.14
Table 8 Analysis of variance for efficiency.
Error 0.000
Table 9 The parameter settings applied in MGGP.
Parameter Value
Function set ×, ÷, +, -
Tournament size 4
Table 10 Optimized pressure drop, efficiency and dimensions.
(%)
gridded tetrahedral.
Fig. 5 Contour of (a) static pressure (Pa) (b) velocity magnitude (m/s) in sections z=9.5 m and x=0 for the
cyclone.
Fig. 6 Particle spinning in operation [41] (a) and a simulated cyclone (b).
Fig. 7 Analysis of design of experiment (a) main effect plot for pressure drop (b) main effect plot for
efficiency.
Fig. 8 Velocity contours for cyclones 121, 122 and 123 at section x = 0.
Fig. 9 MGGP model prediction versus calculated Euler number by CFD in (a) training data set (b) testing
data set.
Fig. 10 MGGP model prediction versus calculated efficiency by CFD in (a) training data set (b) testing
data set.
Fig. 11 Pareto chart for minimizing Euler number and maximizing efficiency for two-objective
optimization.
Fig. 12 Comparison of tangential velocity for optimized cyclone and Kerman cement plant preheater