Scientia Iranica C (2017) 24(3), 1253{1263

Sharif University of Technology

Scientia Iranica
Transactions C: Chemistry and Chemical Engineering
www.scientiairanica.com

Optimization of ethylene oxychlorination uidized-bed

reactor using Dierential Evolution (DE) method
M.H. Khademia; and S. Angooraj Taghavib;c
a. Department of Chemical Engineering, Faculty of Engineering, University of Isfahan, Isfahan, P.O. Box 81746-73441, Iran.
b. Chemical Engineering Department Fars Science and Research Branch, Islamic Azad University, Shiraz, Iran.
c. Department of Chemical Engineering, Shiraz Branch, Islamic Azad University, Shiraz, P.O. Box 71987-74731, Iran.
Received 22 December 2015; received in revised form 29 October 2016; accepted 7 January 2017

KEYWORDS Abstract. The present work aims to employ Dierential Evolution (DE) algorithm to
Ethylene dichloride; optimize ethylene oxychlorination process to produce 1,2-dichloroethane in a uidized bed
Ethylene reactor as a feedstock of PVC production. A steady-state reactor model, based on two-
oxychlorination; phase theory of uidization, was developed to investigate the eects of various parameters
Fluidized-bed reactor; on C2 H4 and HCl conversions. The model's results were compared favorably with the
Optimization; industrial data obtained from a pilot plant working in Italy. The feed temperature, pressure,
Dierential evolution. HCl and O2 molar ow rates, and cooling medium temperature were selected as decision
variables to minimize the objective function subject to the environmental constraints. The
highest performance was found at HCl/C2 H4 and O2 /C2 H4 molar ratios of 2 and 0.55,
respectively; feed and cooling medium temperatures of 440 and 360 K, respectively; pressure
of 367.6 kPa. The results show a decrease of 20 C in the feed temperature, which leads to
saving energy and reducing the size of the pre-heater.

© 2017 Sharif University of Technology. All rights reserved.

1. Introduction EDC is a major chemical intermediate which

belongs to a group of chemicals with the highest
Oxychlorination of ethylene is a chemical reaction in annual production rate. EDC is the precursor of Vinyl
which oxygen and hydrogen chloride react with ethy- Chloride Monomer (VCM), which is polymerized to
lene in vapor phase over a supported copper chloride polyvinyl chloride (PVC). PVC polymer is utilized
catalyst to produce ethylene dichloride (EDC or 1,2- commercially in a wide variety of useful vinyl plastic
dichloroethane) and water: products and appears in both rigid and exible (plas-
C2 H4 + 2HCl + 1=2O2 ! C2 H4 Cl2 + H2 O; ticized) forms. These products include plastic pipe,
vinyl house siding, ooring and wall covering, window

Hr0 = 240 kJ/mol: (1) frames, wire insulation, vinyl seat covers, recording
discs, and many others.
This reaction occurs at a temperature range of 473- Only a few papers in the literature are concerned
573 K and a pressure of 0.4-0.6 MPa [1,2]. Since with mathematical modeling of ethylene oxychlorina-
this reaction is highly exothermic, heat is removed by tion uidized-bed reactor. Al-Zahrani et al. [3] investi-
internal cooling coils submerged in the uidized-bed. gated the eects of dierent parameters on the behavior
of ethylene oxychlorination for EDC production in a
*. Corresponding author. Tel.: +98 31 37934075; uidized-bed reactor. Prasad et al. [4] estimated the
Fax: +98 31 37934031 rate parameters of ethylene oxychlorination reaction
E-mail address: [email protected] (M.H. Khademi) by developing an unsteady-state model in a xed-bed
1254 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263

reactor while taking into account the change in catalyst separate governing equation for each phase and mass
composition, variation of temperature proles, and exit interchange between them.
gas-phase conversions with time. A two-step ethylene
oxychlorination process of oxidation and reduction 2.1. Assumptions
reactions was carried out in the uidized-bed reactor The following assumptions are incorporated in the
by Go et al. [5]. They studied the eects of supercial development of the mathematical model:
gas velocity and gas composition on the reactivity i The bubble gas is devoid of solids and is in plug
and uidization stability using an experimental set-up. ow;
Keshavarz Moraveji et al. [6] modeled a uidized bed
reactor for 1,2-dichloroethane production from ethy- ii The extent of reaction in the bubble-cloud phase
lene oxychlorination by computational uid dynamics. is negligible. This assumption is justied by
The eect of catalyst amount was considered carefully. the experimental evidence reported by Toor and
They showed that the enhancement of catalyst up to 6 Calderbank [9] for small-sized particles and high
m (height of the bed) increases the products eciency, ow rates;
especially 1,2-dichloroethane. Recently, Montebelli et iii The dense phase is assumed to be perfectly mixed
al. [7] have presented a development and validation of and uniform in temperature;
a new kinetic model for the ethylene oxychlorination
reaction, relying on 9 chemical reactions to describe iv An average value of the bubble size, and hence,
the evolution of 12 species (including 6 major byprod- an average value of the inter-phase exchange pa-
ucts). They also reported two mathematical models rameter are used for the whole bed. The eective
of oxychlorination uidized-bed reactors, namely the bubble size is taken as that which exists at 40% of
simple two-phase and the axially dispersed plug- ow the expanded bed height [10];
models, as well as their comparative validation against v The mass and heat transfer resistances between the
industrial data. particles and the dense-phase gas are negligible;
To the best of our knowledge, the mathematical vi The ideal gas law applies to the gas phase in both
modeling of ethylene oxychlorination reactor has been phases;
carried out by a few researchers in order to achieve
two goals: investigating the eects of operational vii The volumetric ow rate through the bubble phase
parameters on the reactor performance and estimating is constant.
the reaction kinetic parameters. Furthermore, no study
has been found about the optimization of the ethylene 2.2. Mass and energy balances
oxychlorination reactor. To improve the reactor perfor- The material balance of the ith component and the
mance, economic feasibility, and safety, it is required to energy balance equations are developed for the bubble
operate the reactor under optimal conditions. and dense phases at steady state, allowing for a
In this work, a steady-state distributed mathe- change in volumetric ow rate in the dense phase gas.
matical model is developed for ethylene oxychlorination These equations followed by boundary conditions and
to produce EDC in a uidized-bed reactor. A sensi- ideal gas law equations are summarized in Table 1.
tivity analysis is carried out to indicate numerically Rearranging Eqs. (2) and (3), integrating them by
the eects of various key operating variables on the using boundary conditions, and then substituting them
performance of the reactor. Finally, the purpose of into Eqs. (4) and (5), respectively, as well as ideal gas
this study is to optimize the ethylene oxychlorination law equations result in the following equations:
uidized-bed reactor using a Dierential Evolution  
Q P Ab ub xif xid
(DE) algorithm as an exceptionally simple evolution Nid =xif Nf df + (1 e i H )
strategy. The feed temperature, pressure, HCl and O2 Qf R Tf Td
molar ow rates, and cooling medium temperature are
considered as decision variables to minimize the dened + V (1 )(1 ") i r; (9)
objective function.
g Cpg Qdf (Tf 298) g Cpg Qd (Td 298)
2. Mathematical modeling + ub g Cpg Ab (Tf Td )(1 e H )
The two-phase model, bubbling bed model proposed by
Kunii and Levenspiel [8], which considers a uidized + hw Aw (Tw Td ) + V (1  )(1 ")
bed as consisting of a bubble phase and a dense (or
emulsion) phase, has been proven to be the most (
Hr )r = 0; (10)
suitable model for incorporating recent ndings in the
where i = (Kbd )ib and  = Hbd
hydrodynamics of uidization. This model proposes a ub ub g Cpg .
 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263 1255

Table 1. Mass and energy balance equations and boundary conditions for the bubble and dense phases as well as ideal gas
law equations [11].
Equation type Equation Equation no.
Mass and  
dNib = Ab (Kbd )ib Nid Nib (2)
energy balances dz Qd Qb
in bubble phase

ub g Cpg dT
dz =R(Hbd )b (Td Tb )
b (3)
Mass and Nid = Nidf + 0H (Kbd )ib Ab ( NQibb NQidd )dz + V (1  )(1 ") i r (4)
R
energy balances g Cpg Qdf (Tf 298) g Cpg Qd (Td 298) + 0H (Hbd )b Ab (Tb Td )dz
in dense phase +hw Aw (Tw Td ) + V (1  )(1 ")(
Hr )r = 0 (5)

xid P Qd = Nid RTd (6)

Ideal gas law xif P Qf = Nif RTf (7)
Nidf = xif Nf QQdff , (8)

Boundary conditions at z = 0; Nib = Nibf Tb = Tf

2.3. Hydrodynamic and transport property Eq. (9) consists of ve simultaneous nonlinear algebraic
correlations equations involving ve unknowns (the number of
Table 2 shows the hydrodynamic and transport prop- moles of the ve components), and is thus amenable
erty correlations used in this study. to numerical solution techniques. The equations are
solved by a combination of the Newton-Raphson and
2.4. Reaction kinetics the bisection methods, as described by Wgialla et
Several mechanism and kinetic studies have been al. [11].
reported in the literature for ethylene oxychlorina- In the calculation, it is supposed that the heat of
tion [18-20]. In this paper, a rate equation of the reaction and heat capacity are dependent on temper-
Hougen-Watson type, reported by Carrubba and ature. The thermal conductivity, heat capacity, and
Spencer [18], is used. This reaction rate model was density of gas mixture are predicted through linear
based on the surface reaction of ethylene and oxygen, combination of properties of the individual component
on the CuCl2 -Al2 O3 based catalyst, to form an ethylene gases.
oxide intermediate as the rate-determining step. The
reaction rate increases with increasing partial pressures
of ethylene and oxygen, but it is independent of 3. Optimization
hydrogen chloride partial pressure. The reaction rate 3.1. Dierential Evolution (DE)
of EDC production has the following form: Over the last few years, Dierential Evolution (DE)
algorithm has been proposed for dealing with op-
kr PC2 H4 PO1=22 timization problems. DE algorithm is a stochastic
r=  2 optimization method minimizing an objective func-
1 + 0:0017PC2 H4 + 0:0355PO1=22 + 0:0165PH2 O tion that can model the problem's objectives while
incorporating constraints. This algorithm mainly has
kr = 6:83  10 7 exp( E=RT ); (11) three advantages: nding the true global minimum
regardless of the initial parameter values, fast con-
where kr is the reaction rate constant; PC2 H4 , PO2 , vergence, and using a few control parameters. Other
and PH2 O are partial pressures of ethylene, oxygen, and important features of DE algorithm include: (a) being
water in millimeters of Hg, respectively; E is activation simple, fast and easy to use; (b) very easily adaptable
energy, 24.5 kcal/mol; the rate is expressed as moles of to integrand discrete optimization; (c) being quite
EDC per liter-second. eective in non-linear constraint optimization including
penalty functions; and (d) being useful for optimizing
2.5. Numerical solution multi-modal search spaces.
The solution of the mass balance equations is fol- Dierent strategies can be adopted in DE algo-
lowed by the calculation of the heat balance equation. rithm depending upon the type of problem to which DE
1256 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263

Table 2. Hydrodynamic and transport property correlations.

Parameter Theoretical or empirical expressions Ref.

Bed voidage at minimum uidization "mf = 0:586's 0:72 Ar 0:029 ( pg )0:021 Broadhurst and Becker [12]
d3
Ar = g (p g )g p2

2 "3
Supercial velocity at minimum uidization umf = ('s dp )150g(p g ) 1 mf
"mf Kunii and Levenspiel [8]
Bubble diameter dB = dBM (dBM dB o)e 0:3z=D Mori and Wen [13]
dBM = 0:652[A(uf umf )]0:4
dBo = 0:00376(uf umf )2

Bubble rising velocity ub = uf umf + 0:711(gdB )0:5 Davidson and Harrison [14]
Fraction of bubble phase  = uf uub mf Kunii and Levenspiel [8]
Total feed volumetric ow rate Qf = u f A Kunii and Levenspiel [8]
Bubble-phase volumetric ow rate Qb = (uf um )A Kunii and Levenspiel [8]
Dense-phase feed volumetric ow rate Qdf = Qf Qb Kunii and Levenspiel [8]
Expanded bed height H = H1 mf Kunii and Levenspiel [8]

Overall mass transfer 1 1 1

(Kbd )ib = (Kbc )ib + (Kcd )ib Kunii and Levenspiel [8]
umf 0:5 g0:25
coecient (bubble phase/dense (Kbc )ib = 4:5( dB ) + 5:85( Dim 1:25 )
dB
phase) based on bubble volume "mf Dim ub 0:5
(Kcd )ib = 6:78( d3 )
B

Overall heat transfer coecient 1 1 1

(Hbd )b = (Hbc )b + (Hcd )b Kunii and Levenspiel [8]
umf g Cpg )0:5 g0:25
(bubble phase/dense phase) based (Hbc )b = 4:5( dB ) + 5:85 (kg g Cdpg1:25
B
on bubble volume (Hcd )b = 6:78(kg g Cpg )0:5 ( "mf ub 0:5
d3B )

Heat transfer coecient (bed-vertical tubes) hw = 0:88 kdpg Ar0:213 Renz et al. [15]
3 1:75 i +1=Mj )0:5
Binary diusivity Dij = 10 PT (V 0:(133=M
+V 0:33 )2 Fuller et al. [16]
i j
Diusivity of component i in gas mixture Dim = 1 xi Reid et al. [17]
Pn xi
j =1 Dij
j 6=1

is applied. The strategies can vary based on the vector 10. DE/rand/2/bin.
to be perturbed, number of dierence vectors consid-
ered for perturbation, and nally the type of crossover The general convention used above is DE/x=y=z ,
used. Ten dierent working strategies, proposed by in which DE stands for dierential evolution, x rep-
Price and Storn [21], are as follows: resents a string denoting the vector to be perturbed,
y is the number of dierence vectors considered for
1. DE/best/1/exp; perturbation of x, and z stands for the type of crossover
2. DE/rand/1/exp; being used (exp: exponential; bin: binomial).
3. DE/rand-to-best/1/exp; Choosing population size (NP), scaling factor (F),
4. DE/best/2/exp; and crossover constant (CR) depends on the specic
problem applied, which is often dicult. But, some
5. DE/rand/2/exp; general guidelines are available. Normally, NP should
6. DE/best/1/bin; be about 5-10 times the number of parameters in a
7. DE/rand/1/bin; vector. As for F , it lies in the range 0.4-1.0. Initially,
F = 0:5 can be tried, then F and/or NP are increased
8. DE/rand-to-best/1/bin; if the population converges prematurely. A good rst
9. DE/best/2/bin; choice for CR is 0.1, but in general, CR should be as
 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263 1257

Table 3. Strategy and parameters used for DE. Penalty function method is employed for handling
Strategy DE/1/best/bin constraints. Inequality constraint, Eq. (13), is incor-
Population size (NP) 100 porated into Eq. (12) using penalty function. This
method involves penalizing the objective function in
Scaling factor (F) 0.8 proportion to the extent of constraint violation (i.e.,
Crossover constant (CR) 1.0 the penalty function takes a nite value when a con-
straint is violated and a value of zero when constraint
large as possible [22]. More details of basic version is satised). In the present case, we used 107 as our
of DE (pseudo code), its strategies, and choosing penalty parameter, but this value depends on the order
of operating parameters were reported by Babu and of magnitude of the variables in the problem; it may
Angira [23,24] and Babu and Munawar [22]. In this change from problem to problem.
optimization, the strategies and parameters used for Thus, the objective function considered for mini-
DE are presented in Table 3. mization, nally, is:
3.2. Optimization problem formulation Minimize
In this work, the goal of optimization is to nd out 2
X
the operating conditions, containing feed temperature, OF = J + 107 G2i ; (14)
pressure, HCl/C2 H4 and O2 /C2 H4 molar ratios, and i=1
cooling medium temperature, that lead to the best
performance of the reactor, i.e. objective function. For where:
the best performance, four sub-functions are dened as G1 = maxf0; (473 T )g; (15)
follows:
G2 = maxf0; (T 573)g: (16)
ˆ The rst term: Maximize the ethylene conversion;
ˆ The second term: Maximize the HCl conversion;
The postulated optimization problem is solved with
DE.
ˆ The third term: Minimize the feed temperature.
The main reason to develop this term is energy 4. Results and discussion
saving, reducing the size of the pre-heater, and
also obtaining the highest conversion at the lowest 4.1. Base case
possible temperature; In order to establish a reference point, simulation is
ˆ The fourth term: Build up a logical relation between carried out for a \base case". The reactor design data
input ow rate and bed height; the residence time as well as operating conditions for the base case are
must be nearly 25 seconds. given in Table 4. The parametric sensitivity analysis
is carried out to assess the impact of the variation of
Therefore, a suitable objective function is: the following parameters on C2 H4 and HCl conversions:
  height at minimum uidization; feed temperature;
1
J = jXC2 H4 1j + jXHC L 1j + j 25j : (12) pressure; cooling medium temperature; residence time;
Tf
HCl/C2 H4 and HCl/O2 molar ratios.
In the commercial ethylene oxychlorination reactors,
the reaction occurs at a temperature range of 473- 4.2. Model validation
573 K [1,2]. Therefore, the reactor temperature The uidized-bed model of ethylene oxychlorination
must not exceed 573 K. However, local hot spots is validated against the pilot plant data, reported
may still remain, creating some problems. For ex- by Cavaterra [25], for pilot plant working in Italy
ample, copper chloride (CuCl2 ) as a catalyst can under the design specications and input data listed
be evaporated above 673 K, and it has been re- in Table 4. It is observed in Table 5 that the model
ported that the product distribution can change performed satisfactorily well under special case of
from high yields of 1,2-dichloroethane to a mix- industrial conditions, and simulation data are in good
ture of more highly chlorinated hydrocarbons such agreement with the observed plant data.
as 1,1,2-trichloroethane, trichloroethylene, vinyl chlo- 4.3. Eect of residence time and bed height at
ride, and 1,2-dichloroethylene with increasing temper- minimum uidization
ature [1,20]. Therefore, one constraint is considered as Figure 1(a) and (b) show the eect of residence time
follows: and bed height at minimum uidization on C2 H4 and
473 < T < 573 K: (13) HCl conversions, respectively. As the residence time
increases, the supercial velocity and the bubble size
The optimization problem considered above is decrease. When the bubble diameter decreases, the
then reformulated so as to include the constraints. overall mass transfer area per unit bubble volume
1258 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263

Table 4. Design and operating conditions of pilot plant.

Parameter Value
Bed height at minimum uidization 7.0 m
Bed diameter 3.4 m
Density of solid particles 3075 kg/m3
Catalyst particle diameter 80 m
Bed pressure 400 kPa
Feed temperature 460 K
Cooling medium temperature 360 K
Molar ow rate of HCl/C2 H4 /O2 64/31.85/21.5 mol/s
Residence time 25 s

Table 5. Comparison between model predictions and pilot plant data at the base case.
Plant Model RE (%)
Outlet temperature (K) 498 497.79 0.04
C2 H4 conversion (%) 99.3 99.55 0.25
HCl conversion (%) 98.3 98.80 0.50

Figure 1. Eect of (a) residence time and (b) bed height at minimum uidization on C2 H4 and HCl conversions.

between bubble and dense phase increases; conse- supercial velocity leads to an increase in residence
quently, C2 H4 and HCl conversions increase. It is time. This gure shows that small change of height
very important to take care of the bubble diameter by at minimum uidization, less than 5 m, has signicant
selecting an appropriate reactant distributor in order eect on C2 H4 and HCl conversions, but increasing the
to generate small bubbles. Furthermore, the reactor bed height at minimum uidization means increasing
must be operated at suciently low supercial gas the amount of catalyst.
velocity. It is clear from Figure 1(a) that C2 H4 and HCl
conversions increase asymptotically until they reach a 4.4. Eect of reactor pressure
residence time of 20 s with C2 H4 and HCl conversions The eects of total pressure on C2 H4 and HCl conver-
of 99.4% and 98.6%, respectively; subsequently, the sions are illustrated in Figure 2. Increasing the reactor
conversions increase slowly till the residence time of pressure from 250 to 500 kPa leads to an increase
35 s. At residence times in excess of 35 s, the conver- of C2 H4 and HCl conversions from 96.9% to 99.8%
sions approach an equilibrium conversion. Therefore, and from 96.2% to 99.2%, respectively. As shown in
j 25j is considered as a sub-function in the objective this gure, increase of pressure above 400 kPa has
function. slight eect on C2 H4 and HCl conversions. Increasing
In Figure 1(b), increasing the bed height at the reactor pressure results in increasing the rate of
minimum uidization increases C2 H4 and HCl conver- reaction, and consequently, increasing C2 H4 and HCl
sions, because an increase in bed height at a constant conversions.
 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263 1259

when the feed temperature increases from 360 to 440

K, the bed temperature increases from 377 to 498 K
while C2 H4 conversion increases from 0.09 to 0.99 as
the rate of reaction increases. Also, the diusivity of
each component in the gas mixture increases as the feed
temperature increases. Furthermore, the diusivity
of gas component aects the mass transfer coecient.
The variation of feed temperature above 440 K has very
little eect on bed temperature and conversion. The
feed temperature is selected as a decision variable for
optimization, because the feed temperature has large
eect on conversions.
4.6. Eect of cooling medium temperature
Figure 2. Eect of total pressure on C2 H4 and HCl The eect of cooling medium temperature on C2 H4
conversions. and HCl conversions and bed temperature is shown in
Figure 4(a) and (b), respectively. When the cooling
4.5. Eect of feed temperature medium temperature increases from 360 to 440 K,
Figure 3(a) and (b), respectively, show the in uence C2 H4 and HCl conversions and bed temperature in-
of variation of feed temperature on C2 H4 conversion crease 0.47%, 0.55%, and 14.5%, respectively. There-
and bed temperature. It is clear from this gure that fore, the variation of cooling medium temperature

Figure 3. Eect of feed temperature on (a) ethylene conversion and (b) bed temperature.

Figure 4. Eect of cooling medium temperature on (a) C2 H4 and HCl conversions and (b) bed temperature.
1260 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263

Table 6. Minimal and maximal values as well as optimal values of decision variables.
Optimization variables Minimal values Maximal values % Optimal values
Tf (K) 400 510 12% 440
Tw (K) 300 360 | 360
P (kPa) 200 600 50% 367.6
HCl molar ow rate (mol/s) 40 80 30% 63.57
O2 molar ow rate (mol/s) 5 35 75% 17.42

has little eect on the conversions, but large eect ables to minimize the objective function (Eq. (14)).
on the bed temperature. An increase in cooling C2 H4 molar ow rate is considered equal to C2 H4 molar
medium temperature will decrease the heat transfer ow rate of pilot plant working in Italy (31.85 mol/s),
rate and increase the bed temperature. The cooling so that the results with the pilot plant data can be
medium temperature must be maintained below 440 K compared; it is not selected as a decision variable.
in order for the temperature of uidized-bed ethylene Minimal and maximal values of the decision vari-
oxychlorination reactor not to exceed 573 K. This, in ables are set, as shown in Table 6. The minimal
turn, prevents catalyst deactivation. Therefore, the and maximal allowed values of the decision variables
cooling medium temperature plays an important role were determined empirically based on reactor feed
in this uidized-bed reactor. conditions, \base case", presented in Table 4. As shown
in Table 6, the minimal and maximal values of most
4.7. Eect of feed molar ratio decision variables are disturbed % around \base
Figure 5 illustrates the variation of ethylene conversion case". To ensure that the cooling water is in the liquid
versus HCl/C2 H4 and HCl/O2 molar ratios at constant phase, the upper bound of cooling medium temperature
HCl molar ow rate, 64 mol/s. Decreasing HCl/C2 H4 is set at 360 K.
molar ratio from 4 to 2 at constant HCl/O2 molar In order to optimize the reactor, the DE code
ratio and increasing HCl/O2 molar ratio from 2 to 4 is coupled with the non-linear mathematical model
at constant HCl/C2 H4 molar ratio result in increasing (Eqs. (2)-(10)). Figure 6(a)-(e) show variations of ob-
C2 H4 conversion. Totally, when the feed molar ratio jective function in terms of (a) O2 molar ow rate, (b)
shifts to the stoicheiometry coecient of ethylene HCl molar ow rate, (c) pressure, (d) feed temperature,
oxychlorination reaction, the ethylene conversion will and (e) cooling medium temperature, respectively.
increase. It is clear from Figure 5 that at HCl/C2 H4 As shown in these gures, the objective function is
= 2, C2 H4 conversion at HCl/O2 = 3.5 is higher than minimized in one point. These points, tabulated in
that at HCl/O2 = 4. This is due to ethylene as the
limiting reactant and the excess availability of oxygen.
4.8. Results of the optimization with DE
With due attention to parametric sensitivity analysis
carried out in Sections 4.3-4.7, feed temperature, pres-
sure, HCl and O2 molar ow rates, and cooling medium
temperature are selected as optimization decision vari-

Figure 6. Variations of objective function versus (a) O2

Figure 5. Eect of HCl/C2 H4 and HCl/O2 molar ratio molar ow rate, (b) HCl molar ow rate, (c) pressure, (d)
on ethylene conversion. feed temperature, and (e) cooling medium temperature.
 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263 1261

the last column of Table 6, are optimal operating ˆ The optimum ethylene and HCl conversions (i.e.,
conditions of the ethylene oxychlorination reactor. Un- 99.5 and 99.7%) can be achieved by working at
der these conditions, HCl/C2 H4 and O2 /C2 H4 molar HCl/C2 H4 and O2 /C2 H4 molar ratios of 2 and 0.55,
ratios are 2 and 0.55, respectively, which are in very feed and cooling medium temperatures of 440 and
good agreement with the stoichiometric coecient of 360 K, respectively, and pressure 367.6 kPa.
ethylene oxychlorination reaction (Eq. (1)). Residence
time 25 s, bed temperature 510.2 K, HCl and C2 H4 Nomenclature
conversions 99.5% and 99.7%, respectively, are asso-
ciated with the set of optimal operating conditions A Cross-sectional area of bed (m2 )
presented in Table 6. The results obtained, according
to optimization, show an increase of 1.1% in the HCl Ab Cross-sectional area of bubble phase
conversion and a decrease of 8.1% and 18% in the (m2 )
pressure and O2 /C2 H4 molar ratio, respectively. Also, Aw Area of heat transfer between coolant
decreasing feed temperature from 460 to 440 K leads to and bed (m2)
saving energy and reducing the size of the pre-heater. Cpg Gas heat capacity (J/kg K)
D Bed diameter (m)
5. Conclusion Dij Diusivity of component i in
Optimization of uidized-bed ethylene oxychlorination component j (m2 /s)
reactor is studied by a steady-state mathematical Dim Diusivity of component i in gas
model based on the two-phase theory of uidization. mixture (m2 /s)
A base case is generated considering similar operat- dB Bubble diameter (m)
ing conditions to pilot plant working in Italy. The
model is validated against this pilot plant data, and dp Particle diameter (m)
good agreement is obtained. The eects of dierent g Gravitational acceleration (m/s2 )
operating variables on C2 H4 and HCl conversions are H Bed height (m)
numerically investigated. DE is used as a powerful op-
timization technique to determine the optimal reactor (Hbd )b Inter-phase heat transfer coecient
operating conditions. The feed temperature, pressure, between bubble and dense phases
HCl/C2 H4 and O2 /C2 H4 molar ratios, and cooling based on bubble volume (J/m3 s K)
medium temperature are selected as decision variables (Hbc )b Inter-phase heat transfer coecient
to minimize the objective function. The results show between bubble and cloud phases
that the obtained optimal operating conditions can based on bubble volume (J/m3 s K)
increase HCl conversion by 1.1% as well as a decrease (Hcd )b Inter-phase heat transfer coecient
of 8.1% and 18% in the pressure and O2 /C2 H4 molar between cloud and dense phases based
ratio, respectively. Also, a decrease of 20 C is observed on bubble volume (J/m3 s K)
in the optimal feed temperature which results in energy
saving. Also, the following key results are achieved: hw Bed to cooling surface heat transfer
coecient (J/m2 s K)
ˆ Increasing the height at minimum uidization more
(Kbd )ib Inter-phase mass transfer coecient
than 5 m has no signicant eect on C2 H4 and between bubble and dense phases
HCl conversions. Therefore, above 5 m height, based on bubble volume (1/s)
there is no economic justication for the extra
capital investments on reactor volume and catalyst (Kbc )ib Interphase mass transfer coecient
quantity; between bubble and cloud phases
ˆ The residence time must be more than 20 s to
based on bubble volume (1/s)
achieve high conversion. This might occur by choos- (Kcd )ib Interphase mass transfer coecient
ing a proper reactant distributor and regulating the between cloud and dense phases based
supercial gas velocity; on bubble volume (1/s)
ˆ Increasing pressure and feed temperature above 400 kg Thermal conductivity of gas (J/K m s)
kPa and 440 K, respectively, will have no signicant M Molecular weight
eect on the conversion. An increase in temperature N Molar ow rate (mol/s)
(above 440 K) causes loss of energy and enlargement
of the size of the preheater; OF Objective Function
ˆ The cooling medium temperature should be less P Reactor pressure (Pa)
than 440 K to avoid catalyst deactivation; Q Volumetric ow rate (m3 /s)
1262 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263

R Gas constant (m3 Pa/g mol K) 7. Montebelli, A., Tronconi, E., Orsenigo, C. and Ballar-
r Kinetic rate of reaction (kg mol/m3 s) ini, N. \Kinetic and modeling study of the ethylene
oxychlorination to 1,2-dichloroethane in uidized-bed
T Temperature (K) reactors", Ind. Eng. Chem. Res., 54, pp. 9513-9524
u Supercial gas velocity (m/s) (2015).
V Volume of overall reactor (m3 ) 8. Kunii, D. and Levenspiel, O., Fluidization Engineering,
x Mole fraction Wiley, New York (1991).
z Distance along bed height (m) 9. Toor, F.D. and Calderbank, P.H. Proceedings of the
Tripartite Chemical Engineering Conference, London
Greek letters Inst. of Chemical Engrs., Montreal, Quebec (1968).
i Stoichiometric coecient of component 10. Fryer, C. and Potter, O.E. \Bubble size variation in
i in reaction two-phase models of uidized bed reactors", Powder
 Fraction of bed consisting of bubbles Technology, 6, pp. 317-322 (1972).

Hi Heat of reaction (J/kg mol) 11. Wgialla, K.M., Helal, A.M. and Elnashaie, S.S.E.H.
" Bed void fraction \The use of mathematical and computer models to
 Viscosity (kg/m s) explore the applicability of uidized bed technology
for highly exothermic catalytic reactions. I. Oxidative
g Density of gas (kg/m3 ) dehydrogenation of butane", Math. Comput. Modeling,
p Density of solid particles (kg/m3 ) 15, pp. 17-31 (1991).
 Residence time (s) 12. Broadhurst, T.E. and Becker, H.A. \Onset of uidiza-
tion in slugging beds of uniform particles", AIChEJ,
's Sphericity of particles 21, pp. 238-247 (1975).
Subscripts 13. Mori, S. and Wen, C.Y. \Estimation of bubble diame-
b Bubble phase gas ter in gaseous uidized beds", AIChEJ, 21, pp. 109-115
(1975).
d Dense phase gas
df Inlet of dense phase gas 14. Davidson, J.F. and Harrison, D. Fluidized Particles,
Cambridge University Press, New York (1963).
f Fresh feed gas
i ith component 15. Renz, U., von Wedel, G. and Reinartz, A. \Heat
transfer characteristics of the FBC at Acchen Technical
mf Minimum uidization University", Proceedings of the 1987 International
w Cooling medium Conference on Fluidized Bed Combustion, American
Society of Mechanical Engineers, New York (1987).
References 16. Fuller, E.N., Schettler, P.D. and Giddings, J.C. \A
1. Magistro, A.J. and Cowfer, J.A. \Oxychlorination of new method for prediction of binary gas phase diusion
ethylene", J. Chem. Educ., 63, pp. 1056-1058 (1986). coecients", Ind. Eng. Chem., 58, pp. 19-27 (1966).
2. Naworski, J.S. and Velez, E.S. \Oxychlorination of 17. Reid, R.C., Prausnitz, J.M. and Sherwood, T.K.,
ethylene", Appl. Ind. Catal., 1, p. 239 (1983). Properties of Gases and Liquids, McGraw-Hill, New
York (1966).
3. Al-Zahrani, S.M., Aljodai, A.M. and Wagialla, K.M.
\Modeling and simulation of 1,2-dichloroethane pro- 18. Carrubba, R.V. and Spencer, J.L. \Kinetics of the
duction by ethylene oxychlorination in uidized-bed oxychlorination of ethylene", Ind. Eng. Chem. Process
reactor", Chem. Eng. Sci., 56, pp. 621-626 (2001). Des. Develop., 9, pp. 414-419 (1970).
4. Prasad, P.S.S., Prasad, K.B.S. and Ananth, M.S. 19. Fengqiu, C., Yongrong, Y., Shunni, R. and Gatang,
\Parameter estimation in a xed-bed reactor operating C. \Studies on ethylene oxychlorination, II. Reaction
under unsteady state: Oxychlorination of ethylene", mechanism and kinetics", Petrochemical Technology,
Ind. Eng. Chem. Res., 40, pp. 5487-5495 (2001). 23, pp. 421-425 (1994).
5. Go, K.S., Kim, Y., Son, S.R. and Kim, S.D. \1,2-
dichloroethane production by two-step oxychlorination 20. Wachi, W. and Asai, Y. \Kinetics of 1,2 dichloroethane
reactions in a uidized bed reactor", Chem. Eng. Sci., formation from ethylene and cupric chloride", Ind.
65, pp. 499-503 (2010). Eng. Chem. Res., 33, pp. 259-264 (1994).
6. Keshavarz Moraveji, M., Davarnejad, R., Filsoof, A. 21. Price, K. and Storn, R. \Dierential Evolution (DE)
and Asgari, M. \Reactor modeling of ethylene oxy- for continuous function optimization", Homepage of
chlorination for 1,2-dichloroethane production", World dierential evolution as on http://www.ICSI.Berke-
Applied Sciences Journal, 21(8), pp. 1113-1115 (2013). ley.edu/wstorn/code.html (May 2006).
 M.H. Khademi and S. Angooraj Taghavi/Scientia Iranica, Transactions C: Chemistry and ... 24 (2017) 1253{1263 1263

22. Babu, B.V. and Munawar, S.A. \Dierential evolution Biographies
strategies for optimal design of shell-and-tube heat
exchangers", Chem. Eng. Sci., 62, pp. 3720-3739 Mohammad Hasan Khademi is an Associate Pro-
(2007). fessor of Chemical Engineering at Isfahan University,
Isfahan, Iran. His main research interests include pro-
23. Babu, B.V. and Angira, R. \Modied dierential evo- cess modeling, simulation and optimization, especially,
lution (MDE) for optimization of non-linear chemical
processes", Comput. Chem. Eng., 30, pp. 989-1002 in the eld of chemical reactor.
(2006).
Sajjad Angooraj Taghavi obtained his BS degree
24. Angira, R. and Babu, B.V. \Optimization of process in Chemical Engineering in 2006 from Islamic Azad
synthesis and design problems: A modied dierential University of Gachsaran, Iran, and his MS degree in
evolution approach", Chem. Eng. Sci., 61, pp. 4707- 2013 from the Department of Chemical Engineering
4721 (2006).
at Fars Science and Research Branch, Islamic Azad
25. Cavaterra, E. \Catalysts to make dichloroethane", University, Shiraz, Iran. His main research interests in-
Hydrocarbon Processing, 67, pp. 63-67 (1988). clude modeling and optimization of chemical reactors.