differential evolution (de) strategy for optimization of hydrogen production, cyclohexane...
TRANSCRIPT
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 0
Avai lab le at www.sc iencedi rect .com
journa l homepage : www.e lsev ie r . com/ loca te /he
Differential evolution (DE) strategy for optimization ofhydrogen production, cyclohexane dehydrogenation andmethanol synthesis in a hydrogen-permselective membranethermally coupled reactor
M.H. Khademi, M.R. Rahimpour*, A. Jahanmiri
Chemical Engineering Department, School of Chemical and Petroleum Engineering, Shiraz University, Shiraz 71345, Iran
a r t i c l e i n f o
Article history:
Received 26 June 2009
Received in revised form
15 December 2009
Accepted 17 December 2009
Available online 20 January 2010
Keywords:
Methanol synthesis
Dehydrogenation of cyclohexane
Recuperative coupling
Hydrogen-permselective membrane
Hydrogen production
Optimization
Differential evolution
* Corresponding author. Tel.: þ98 711 230307E-mail address: [email protected] (
0360-3199/$ – see front matter ª 2009 Publisdoi:10.1016/j.ijhydene.2009.12.080
a b s t r a c t
In this work a novel reactor configuration has been proposed for simultaneous methanol
synthesis, cyclohexane dehydrogenation and hydrogen production. This reactor configu-
ration is a membrane thermally coupled reactor which is composed of three sides for
methanol synthesis, cyclohexane dehydrogenation and hydrogen production. Methanol
synthesis takes place in the exothermic side that supplies the necessary heat for the
endothermic dehydrogenation of cyclohexane reaction. Selective permeation of hydrogen
through the Pd/Ag membrane is achieved by co-current flow of sweep gas through the
permeation side. A steady-state heterogeneous model of the two fixed beds predicts the
performance of this configuration. A theoretical investigation has been performed in order
to evaluate the optimal operating conditions and enhancement of methanol, benzene and
hydrogen production in a membrane thermally coupled reactor. The co-current mode is
investigated and the optimization results are compared with corresponding predictions for
a conventional (industrial) methanol fixed bed reactor operated at the same feed condi-
tions. The differential evolution (DE), an exceptionally simple evolution strategy, is applied
to optimize this reactor considering the mole fractions of methanol, benzene and hydrogen
in permeation side as the main objectives. The simulation results have been shown that
there are optimum values of initial molar flow rate of exothermic and endothermic stream,
inlet temperature of exothermic, endothermic and permeation sides, and inlet pressure of
exothermic side to maximize the objective function. The simulation results show that the
methanol mole fraction in output of reactor is increased by 16.3% and hydrogen recovery in
permeation side is 2.71 yields. The results suggest that optimal coupling of these reactions
could be feasible and beneficial. Experimental proof-of-concept is needed to establish the
validity and safe operation of the novel reactor.
ª 2009 Published by Elsevier Ltd on behalf of Professor T. Nejat Veziroglu.
1. Introduction resources, predominantly natural gas. It is produced from
Methanol is an important multipurpose base chemical,
a simple molecule which can be recovered from many
1; fax: þ98 711 6287294.M.R. Rahimpour).hed by Elsevier Ltd on be
synthesis gas on a large scale worldwide. Synthesis gas
consists of H2, CO2, CO and some inert components. Methanol
conversion in a conventional fixed bed methanol reactor is
half of Professor T. Nejat Veziroglu.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 0 1937
low due to the equilibrium nature of the reaction. Therefore,
most of the synthesis gas must be circulated around the loop
and this poses problems in operating costs. The methanol
synthesis reaction is exothermic and the total moles reduce as
the reaction proceeds. The temperature and pressure of
reaction are 495–535 K and 5–8 MPa, respectively [1].
Fig. 1(a) shows the schematic diagram of a conventional
methanol synthesis reactor. A conventional methanol reactor
is basically a vertical shell and tube heat exchanger. The
catalyst is packed in vertical tubes and surrounded by the
boiling water. The methanol synthesis reactions are carried
out over a commercial CuO/ZnO/Al2O3 catalyst. The heat of
exothermic reactions is transferred to the boiling water and
steam is produced. A two-dimensional steady-state simula-
tion of a single stage conventional type methanol reactor
shows that properties of reactor are not varying in the radius
of catalyst tube [2]. Recently, a dual-type reactor system
instead of a single-type reactor was developed for methanol
synthesis by Rahimpour et al. [3–5]. The dual-type methanol
reactor is an advanced technology for converting natural gas
to methanol at low cost and in large quantities.
Also several researches were performed on reactor modeling
and optimization of methanol synthesis. A dynamic simulation
and optimization for an auto-thermal dual-type methanol
synthesis reactor was developed by Askari et al. [6] in the
presence of catalyst deactivation using genetic algorithm. Also,
Rahimpour and Elekaei [7] present a study on optimization of
Product
Synthesis Gas from Reformer
Pro
Synthesis Gas from Reformer
Cyclohexane
Sweep Gas
H2 + Sweep Gas
a
b
Fig. 1 – A schematic diagram of (a) a conventional methanol synt
configuration.
a membrane dual-type methanol reactor in the presence of
catalyst deactivation using genetic algorithm. A theoretical
investigation has been performed in order to evaluate the
optimal operating conditions and enhancement of methanol
production in a membrane dual-type methanol reactor. In the
last few years, a Differential Evolution (DE) algorithm for dealing
with optimization problems has been proposed. DE algorithm is
a stochastic optimization method minimizing an objective
function that can model the problem’s objectives while incor-
porating constraints. The algorithm mainly has three advan-
tages; finding the true global minimum regardless of the initial
parameter values, fast convergence, and using a few control
parameters. Being simple, fast, easy to use, very easily adapt-
able for integrand discrete optimization, quite effective in non-
linear constraint optimization including penalty functions and
useful for optimizing multi-modal search spaces are the other
important features of DE algorithm. Recently, a dynamic opti-
mization of a novel methanol synthesis loop with hydrogen-
permselective membrane reactor is presented by Parvasi et al.
[8] using DE method. Also, Rahimpour et al. [9] studied
a dynamic optimization of a novel radial-flow, spherical-bed
methanol synthesis reactor in the presence of catalyst deacti-
vation using DE algorithm.
Multifunctional reactors integrate, in one vessel, one or
more transport processes and a reaction system (Agar [10];
Zanfir and Gavriilidis [11]) and are widely used in industries as
process intensification tools. These multifunctional reactors
BoilingWater
PureMethanol
Steam Drum
To Distillation Unit
duct
PureMethanol
Benzene
Hydrogen
Separator
To Distillation Unit
hesis reactor and (b) a membrane thermally coupled reactor
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 01938
make the process more efficient and compact and result in
large savings in the operational and capital costs (Dautzen-
berg and Mukherjee [12]). A multifunctional reactor can be
used, for example, for coupling exothermic and endothermic
reactions. In it, an exothermic reaction is used as the heat
producing source to drive the endothermic reaction(s). In the
last years promising concepts for the recuperative coupling of
exothermic and endothermic reactions have been published
(Itoh and Wu [13]; Kolios et al. [14]; Moustafa and Elnashaie
[15]; Fukuhara and Igarashi [16]; Ramachandran et al. [17,18];
Abo-Ghander et al. [19]). Also, Khademi et al. [20] optimized
the methanol synthesis reaction and cyclohexane dehydro-
genation in a thermally coupled reactor using differential
evolution (DE) method. From these previous studies, coupling
of endothermic and exothermic reactions may enable both
the concentration and temperature profiles along the reactor
to be manipulated, shifting the conversion of thermodynam-
ically limited reactions to higher values, and efficiently using
the heat liberated by an exothermic reaction to provide the
endothermic heat requirements of the other reaction [21].
Chemical reactants may be shifted to products due to
thermodynamic equilibrium by removing the reactants from
product gases [22]. By insertion of Pd/Ag membrane in
a packed-bed reactor, optimal concentration and temperature
profiles via controlled dosing of reactant along the reactor can
be created, so that large improvements in conversion and
selectivity could be achieved and also pure hydrogen can be
produced [3,23–26]. Recently, hydrogen energy has been
attracting much attention due to its potential to reduce envi-
ronmental burdens and the viewpoint of energy security [26].
At present, hydrogen is produced almost entirely from fossil
fuels such as natural gas, naphtha, and coal. In such cases,
however, the same amount of carbon dioxide is released during
the production of hydrogen as that formed by direct combus-
tion of those fuels. Dehydrogenation reactions are an attractive
alternative for hydrogen production because it has essentially
zero CO2 impact giving a positive environmental contribution.
In the present work, a catalytic dehydrogenation reaction in
the endothermic side is used instead of the cooler-water in the
methanol synthesis reactor. The dehydrogenation reaction
chosen is the catalytic dehydrogenation of cyclohexane to
benzene. Fig. 1(b) shows a schematic diagram for the co-
current mode of a membrane thermally coupled reactor
configuration with three sides. The first side is an exothermic
side, where methanol synthesis takes place on the CuO/ZnO/
Al2O3 catalyst. The second side is an endothermic side, where
dehydrogenation of cyclohexane to benzene takes place on the
Pt/Al2O3 catalyst. The sweep gas flows through the third side
(permeation side) which selectively removes the hydrogen by
permeation through the Pd/Ag membrane. Heat is transferred
continuously from the exothermic reaction to the endo-
thermic reaction. The clear advantages of this integrated
catalytic membrane reactor include: achieving a multiple
reactants multiple products configuration, production of pure
hydrogen and possibility of achieving higher degree of in-situ
energy integration between the coupled endothermic dehy-
drogenation reaction and the exothermic synthesis reactions.
One of the important key issues in methanol reactor
configurations is implementing a higher temperature at first
parts of reactor for higher kinetics constants and then
reducing temperature gradually at the end parts of reactor for
increasing thermodynamics equilibrium. In this novel inte-
grated catalytic membrane reactor, this important key issue
achieves and leads to increase the methanol production at the
same operating conditions.
In our previous work [27], a distributed mathematical
model for membrane thermally coupled reactor that is
composed of three sides is developed for methanol synthesis
and cyclohexane dehydrogenation. The effect of various key
operating variables on the performance of the reactor is
numerically investigated. While the optimal operating condi-
tions can increase the production and reduce the operating
cost so, the purpose of this study is optimization of membrane
thermally coupled reactor using DE method, as a strong opti-
mization method. The inlet temperature of all sides, initial
molar flow rate of exothermic and endothermic sides, and
inlet pressure of exothermic side have been considered as
decision variables to reach maximum mole fraction of meth-
anol, benzene and hydrogen in the permeation side.
Rigorous mathematical models are excellent tools for the
exploration of the basic characteristics of such novel
configurations. Such an exploration can achieve consider-
able savings in money and time during the expensive stage of
pilot plant development. The continuous development of the
model in conjunction with the pilot plant optimal utilization
can also achieve considerable benefits on the road towards
the successful commercialization of such efficient novel
configurations.
The paper is organized as follows: reactions scheme and
kinetics are shown in Section 2. Mathematical model and
concepts about DE are explained in Sections 3 and 4 respec-
tively, followed by optimization of membrane thermally
coupled reactor in Section 5. Numerical solution is presented
in Section 6 with results and discussion in Section 7 and
conclusions are drawn in Section 8.
2. Reaction scheme and kinetics
2.1. Methanol synthesis
In the methanol synthesis, three overall reactions are
possible: hydrogenation of carbon monoxide, hydrogenation
of carbon dioxide and reverse water–gas shift reaction, which
are as follows:
COþ 2H24CH3OH DH298 ¼ �90:55 kJ=mol (1)
CO2 þ 3H24CH3OHþH2O DH298 ¼ �49:43 kJ=mol (2)
CO2 þH24COþH2O DH298 ¼ þ41:12kJ=mol (3)
Reactions (1)–(3) are not independent so that one is a linear
combination of the other ones. In the current work, the rate
expressions have been selected from Graaf et al. [28]. The rate
equations combining with the equilibrium rate constants [29]
provide enough information about kinetics of methanol
synthesis. The corresponding rate expressions due to the
hydrogenation of CO, CO2 and reversed water–gas shift reac-
tions over commercial CuO/ZnO/Al2O3 catalysts are:
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 0 1939
r ¼k1KCO
hfCOf 3=2
H2� fCH3OH=f
1=2H2
KP1
i� �h � � i (4)
11þ KCOfCO þ KCO2fCO2
f 1=2H2þ KH2O=K
1=2H2
fH2O
r2 ¼k2KCO2
hfCO2
f 3=2H2� fCH3OHfH2O=f
3=2H2
KP2
i�
1þ KCOfCO þ KCO2fCO2
�hf 1=2H2þ�
KH2O=K1=2H2
�fH2O
i (5)
r3 ¼k3KCO2
hfCO2
fH2� fH2OfCO=KP3
i�
1þ KCOfCO þ KCO2fCO2
�hf 1=2H2þ�
KH2O=K1=2H2
�fH2O
i (6)
The reaction rate constants, adsorption equilibrium
constants and reaction equilibrium constants which occur in
the formulation of kinetic expressions are tabulated in Table 1,
respectively.
2.2. Dehydrogenation of cyclohexane
Hydrogen is an optimum large-scale fuel for the future,
although there remains some problems in transport and long-
term storage. One has to develop the use of alternative fuels
that are easily transformed into hydrogen and that can be
stored in liquid form, and, thus, more safely and economi-
cally. One of these fuels is cyclohexane.
The reaction scheme for the dehydrogenation of cyclo-
hexane to benzene is as follows.
C6H124C6H6 þ 3H2 DH298 ¼ þ206:2 kJ=mol (7)
The following reaction rate equation of cyclohexane, r4, is
used [30]:
r4 ¼�k�
KPPC=P3H2� PB
�
1þ�
KBKPPC=P3H2
� (8)
Table 1. – The reaction rate constants, the adsorptionequilibrium constants, and the reaction equilibriumconstants for methanol synthesis and dehydrogenationof cyclohexane reactions.
A B
Methanol synthesis reaction
k ¼ AexpðB=RTÞk1 (4.89 � 0.29)�107 �63,000 � 300
k2 (1.09 � 0.07)�105 �87,500 � 300
k3 (9.64 � 7.30)�106 �152,900 � 6800
K ¼ A expðB=RTÞKCO (2.16 � 0.44)�10�5 46,800 � 800
KCO2 (7.05 � 1.39)�10�7 61,700 � 800
ðKH2O=K1=2H2Þ (6.37 � 2.88)�10�9 84,000 � 1400
KP ¼ 10ðA=T�BÞ
KP1 5139 12.621
KP2 3066 10.592
KP3 �2073 �2.029
Cyclohexane dehydrogenation reaction
k ¼ AexpðB=TÞk 0.221 �4270
KB 2.03 � 10�10 6270
KP 4.89 � 1035 3190
where k, KB, and Kp are, respectively, the reaction rate
constant, the adsorption equilibrium constant for benzene,
and the reaction equilibrium constant that are tabulated in
Table 1. pi is the partial pressure of component i in Pa. The
reaction temperature is in the range of 423–523 K and the total
pressure in the reactor is maintained at 101.3 kPa. The catalyst
for this cyclohexane dehydrogenation reaction is Pt/Al2O3 [31].
3. Mathematical model
Fig. 2 shows a schematic diagram of the co-current mode for
a membrane heat-exchanger reactor configuration. A one-
dimensional heterogeneous model, which is a conventional
model for a catalytic reactor with heat and mass transfer resis-
tances, has been developed for this reactor in order to determine
the concentration and temperature distributions inside the
reactor. In this model the following assumptions are used:
� The gas mixture is an ideal gas in both catalytic reactor
sections.
� Both sections of the reactor are operated at steady-state
conditions.
� Radial variations in both beds are negligible (one-dimen-
sional model).
� With due attention to high gas velocity, axial diffusion of
mass and heat are negligible in both sections.
� Bed porosity in axial and radial directions is constant.
� Laminar plug flow is employed in both endothermic and
exothermic sides.
� The chemical reactions are assumed to take place only in the
catalyst particles and homogenous reactions are neglected.
� Heat loss to surrounding is neglected.
To obtain the mole balance equation and the energy balance
equation, a differential element along the axial direction inside
the reactor was considered. The balances typically account for
convection, transport to the solid phase, diffusion through the
membrane and reaction. The mass and energy balances for
solid and fluid phases in all sides, hydrogen permeation in Pd/
Ag membrane, pressure drop equation and boundary condi-
tions are summarized in Table 2. In equations (9) and (10), h is
effectiveness factor (the ratio of the reaction rate observed to
the real rate of reaction), which is obtained from a dusty gas
model calculations [28]. In equation (12), the positive sign is
used for the exothermic side and the negative sign for the
endothermic side. In equations (11) and (12), b is equal to 1 for
the endothermic side and 0 for the exothermic side. In equa-
tions (13) and (14), b is equal to 1 for hydrogen component and
0 for the sweep gas. In hydrogen-permeation equation, PH2 is
hydrogen partial pressure in Pa. The pre-exponential factor P0
above 200 �C is reported as 6.33 � 10�8 mol m�2 s�1 Pa�1/2 and
the activation energy Ep is 15.7 kJ mol�1 [32–34].
3.1. Auxiliary correlations
To complete the simulation, auxiliary correlations should be
added to the model. In the heterogeneous model, because of
transfer phenomena, the correlations of estimation of heat
and mass transfer between two phases and estimation of
Sweep gas
Permeation of hydrogen
Permeation side
Permeation sideC6H12
C6H6
H2
ArC6H12
Ar
CH3OHCO2
COH2OH2
N2
CH4
CH3OHCO2
COH2OH2
N2
CH4
Heat transfer
Endothermic side
Endothermic side
Exothermic side
Sweep gas H2
Fig. 2 – A schematic diagram of the co-current mode for a membrane heat-exchanger reactor configuration.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 01940
physical properties of chemical species and overall heat
transfer coefficient between two sides should be considered.
The correlations used for physical properties, mass and heat
transfer coefficient are summarized in Table 3. For the heat
Table 2. – Mass and energy balances for solid and fluid phasesbalances for permeation side, hydrogen permeation in Pd/Ag mconditions.
definition
Mass and energy balances for solid phase
Mass and energy balances for fluid phase
Mass and energy balances for permeation side
Hydrogen permeation in Pd/Ag membrane (Sievert’s law)
Pressure drop (Ergun momentum balance)
Boundary conditions
transfer coefficient between bulk gas phase and solid phase
(hf) in the exothermic and endothermic side, the heat
transfer coefficient between gas phase and reactor wall is
applicable.
in the exothermic and endothermic sides mass and energyembrane, pressure drop equation and boundary
Equation
avcjkgi;j
�yg
i;j � ysi;j
�þ hri;jrb ¼ 0 (9)
avhf
�Tg
j � Tsj
�þ rb
XN
i�1
hri;j
��DHf ;i
�¼ 0 (10)
�Fj
Ac
vygi;j
vzþ avcjkgi;j
�ys
i;j � ygi;j
�� b
JH2
Ac¼ 0 (11)
� Fj
AcCg
pj
vT gj
vzþ avhf
�T s
j � T gj
�� pDi
AcU1�2
�T g
2 � Tg1
�
� bJH2
Ac
Z T3
T2
CpdT� bpDi
AcU2�3
�T g
2 � T g3
�¼ 0 ð12Þ
�F3
vygi;3
vzþ bJH2
¼ 0 (13)
�F3Cgp3
vT g3
vzþ bJH2
Z T3
T2
CpdTþ pDiU2�3
�T g
2 � T g3
�¼ 0 (14)
JH2¼
2pLP0exp��Ep
RT
�
ln�
DoDi
� � ffiffiffiffiffiffiffiffiffiffiPH2 ;2
p�
ffiffiffiffiffiffiffiffiffiffiPH2 ;3
p �(15)
dPdz¼ 150
ð1� 3Þ2mug
33d2p
þ 1:75ð1� 3Þu2
gr
33dp(16)
z ¼ 0; ygi;j ¼ y g
i0;j; T gj ¼ Tg
j0; P gj ¼ P g
j0 (17)
Table 3. – Physical properties, mass and heat transfer correlations.
Parameter Equation Reference
Component heat capacity Cp ¼ aþ bTþ cT2 þ dT3
Mixture heat capacity Based on local compositions
Viscosity of reaction mixtures Based on local compositions
Mixture thermal conductivity Lindsay and Bromley [35]
Mass transfer coefficient between gas and solid phases kgi ¼ 1:17 Re�0:42 Sc�0:67i ug � 103 Cussler [36]
Re ¼ 2Rpug
m
Sci ¼ m
rDim�10�4
Dim ¼ 1�yiPi¼j
yiDij
[37]
Dij ¼10�7T3=2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi1=Miþ1=Mj
p
Pðv3=2ciþv2=3
cjÞ2
Reid et al. [38]
Overall heat transfer coefficient1U ¼ 1
hiþ Ai lnðDo=DiÞ
2pLKwþ Ai
Ao
1ho
Heat transfer coefficient between gas phase and reactor wall hCprmðCpm
K Þ2=3 ¼ 0:458
3Bðrudp
m�0:407 [39]
Heat transfer
coefficient between sweep
gas and reactor
wall in the permeation side
hDHK ¼ 0:023ðruDH
mÞ0:8ðCpm
K Þ0:3
[40]
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 0 1941
4. Differential evolution
Differential evolution (DE) method, a recent optimization
technique, is an exceptionally simple and easy to use evolu-
tion strategy, which is significantly fast and robust in
numerical optimization and is more likely to find a function’s
true global optimum [41]. A basic version of DE consists of
following steps [42,43]:
1. Input: dimensions of problem (D) and key parameters (the
population size NP, the scaling factor F, and the crossover
constant CR).
2. Generate initial population in upper (UB) and lower (LB)
bounds of each decision variable for NP times in an array.
3. Evaluate the objective function (performance index) for
each individual including penalty terms.
4. Perform mutation, crossover operation to obtain the trial
vector for each target vector in the population.
a) For each vector Xt (target vector), select three distinct
vectors Xa, Xb and Xc randomly from the population
array other than vector Xt. and create mutant vector as
mutant vector ¼ Xc þ F(Xa � Xb)
b) Perform crossover for each target vector with its mutant
vector to create a trial vector as below.
for k ¼ 1 to D
If ((random no. (0, 1)<CR) or k ¼ D)
trial vectork ¼mutant vectork
else
trial vectork ¼ target vectork
end for
5. Check whether the parameters of trial vector are within
the bounds. If an individual of this trial vector is outside
the bounds, then this parameter is assigned a value
randomly within the associated bounds.
6. Evaluate the cost for trial vector as in step 3.
7. Each trial or target vectors which has better performance
index is transferred as a member to the next generation.
8. Repeat steps 4–7 until the termination criterion is satisfied
(Section 4.2).
4.1. Different strategies of DE
Different strategies can be adopted in DE algorithm depending
upon the type of problem for which DE is applied. The strat-
egies can vary based on the vector to be perturbed, number of
difference vectors considered for perturbation, and finally the
type of crossover used. The following are the 10 different
working strategies proposed by [44]: (1) DE/best/1/exp, (2) DE/
rand/1/exp, (3) DE/rand-to-best/1/exp, (4) DE/best/2/exp, (5)
DE/rand/2/exp, (6) DE/best/1/bin, (7) DE/rand/1/bin, (8) DE/
rand-to-best/1/bin, (9) DE/best/2/bin, (10) DE/rand/2/bin.
The general convention used above is DE/x/y/z. DE stands
for differential evolution, x represents a string denoting the
vector to be perturbed, y is the number of difference vectors
considered for perturbation of x, and z stands for the type of
crossover being used (exp: exponential; bin: binomial). In this
study, the sixth strategy of DE, i.e., DE/best/1/bin is used.
Hence, the perturbation can be either in the best vector of the
previous generation or in any randomly chosen vector. Simi-
larly for perturbation either single or two vector differences
can be used. For perturbation with a single vector difference,
out of the three distinct randomly chosen vectors, the
weighted vector differential of any two vectors is added to the
third one. Similarly for perturbation with two vector differ-
ences, five distinct vectors, other than the target vector, are
chosen randomly from the current population. Out of these,
the weighted vector difference of each pair of any four vectors
is added to the fifth one for perturbation. In exponential
crossover, the crossover is performed on the D variables in one
loop until it is within the CR bound. The first time a randomly
picked number between 0 and 1 goes beyond the CR value, no
crossover is performed and the remaining D variables are left
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 01942
intact. In binomial crossover, the crossover is performed on
each of the D variables whenever a randomly picked number
between 0 and 1 is within the CR value. So for high values of CR,
the exponential and binomial crossovers yield similar results.
In binomial case, the last variable always comes from random
noisy vector to ensure that the random noisy vector is different
from the target vector, and hence the above procedure is
applied up to D � 1 variables. The strategy to be adopted for
each problem is to be determined separately by trial and error.
A strategy that works out to be the best for a given problem
may not work well when applied for a different problem.
4.2. Choosing NP, F, and CR
Choosing NP, F, and CR depends on the specific problem
applied, and is often difficult. But some general guidelines are
available. Normally, NP should be about 5–10 times the number
of parameters in a vector. As for F, it lies in the range 0.4–1.0.
Initially F ¼ 0.5 can be tried then F and/or NP is increased if the
population converges prematurely. A good first choice for CR is
0.1, but in general CR should be as large as possible [41]. The
stopping criteria may be of two kinds. One may be some
convergence criterion that states that the difference between
the best and the worst evaluations of objective function in
current generation should be less than some specified value
[42]. The other may be an upper bound on the number of
generations. The stopping criteria may be a combination of the
two as well. The best combination of these key parameters of
DE for each of the strategies mentioned earlier is again
different. Price and Storn [44] have mentioned some simple
rules for choosing the best strategy as well as the corresponding
key parameters. In this study, the population size is chosen 100,
the scaling factor 0.8 and the crossover constant 1.0. Among
DE’s advantages are its simple structure, ease of use, speed, and
robustness. Already, DE has been successfully applied for
solving several complex problems and is now being identified
as a potential source for accurate and faster optimization.
5. Optimization of membrane thermallycoupled reactor
In this study, maximization of the objective namely,
summation of output mole fractions of methanol, benzene
and hydrogen in permeation side is considered. The objective
is as follows:
J ¼ yCH3OH þ yC6H6þ yH2
(18)
Six decision variables namely, inlet temperature of
exothermic side T01, inlet temperature of endothermic side T02,
initial molar flow rate of exothermic side F01, initial molar flow
rate of endothermic side F02, inlet pressure of exothermic side
P01, and inlet temperature of permeation side T03 are consid-
ered for optimization. For tubular and exothermic reactors,
temperature and pressure are two important parameters that
change during the reactor length and they have a direct effect
on thermodynamics equilibrium and catalyst activity. The
main reason to develop an inlet optimal temperature in all
sides is energy saving and reducing in size of pre-heater. While
the transferred heat from the exothermic side to endothermic
side is dependent on the ratio of the exothermic-to-endo-
thermic side flow rates, these two parameters are selected as
a decision variable. Jeong et al. [31,45] reported that the
cyclohexane dehydrogenation reaction occurs in 101.3 kPa,
hence the inlet pressure of endothermic side is considered
equal to this value and is not selected as a decision variable.
The bounds of decision variables are:
495 < T01 < 535 K (19)
423 < T02 < 523 K (20)
0:05 < F01 < 1:0 mol=s (21)
0:05 < F02 < 1:0 mol=s (22)
50 < P01 < 80 bar (23)
298 < T03 < 535 K (24)
To ensure that the temperature of synthesis gas at the
reactor inlet, T01 is not too low for the methanol synthesis
reaction to occur, the lower bound on inlet temperature of
exothermic side is set at 495 K. At high temperatures catalyst
starts to deactivation [1], hence an upper bound of 535 K is
chosen for inlet temperature of exothermic side.
While the activity of the Pt/Al2O3 catalyst in the tempera-
ture range of 423–523 K was evaluated in a conventional
packed bed and no significant deactivation was found after at
least 3 days of use [31], the bounds for the inlet temperature of
endothermic side, T02 is chosen. The lower and upper bounds
for the initial molar flow rate of exothermic and endothermic
side have been selected with no prior intention. The pressure
of methanol synthesis reaction is 50–80 bar [1], therefore this
domain is considered for the inlet pressure of exothermic side.
The environment temperature (298 K) is selected as the lower
bond for inlet temperature of permeation side, and its upper
bound is the same as upper bound for inlet temperature of
exothermic side.
Three constraints are also considered for optimization:
T2 < T1 (25)
495 < T1 < 535 K (26)
423 < T2 < 523 K (27)
The constraints on temperature (Eq. (25)) are based on the
minimum temperature required in order to make a driving
force for heat transfer from the solid wall.
The optimization problem considered above is then refor-
mulated so as to include the constraints. Penalty function
method is employed for handling constraints. The constraints
in equations (25)–(27) are incorporated into the objective
function (Eq. (18)) using penalty functions. This method
involves penalizing the objective function in proportion to the
extent of constraint violation (i.e., the penalty function takes
a finite value when a constraint is violated and a value of zero
when constraint is satisfied). In the present case, we used 107
as our penalty parameter. But this value depends on order of
Table 4. – Comparison between simulation and plant datafor conventional methanol synthesis reactor.
Reactor inlet Reactor outlet
Exp. Calc.
Composition (mol %)
CO2 3.45 2.18 2.43
CO 4.66 1.44 1.52
H2 79.55 75.71 76.54
CH4 11.72 12.98 12.96
N2 0.032 0.16 0.035
H2O 0.08 1.74 1.47
CH3OH 0.032 5.49 5.05
Feed flow
rate (mol s�1)
0.565 0.510 0.511
Temperature (K) 503 528 524.1
Pressure (bar) 61.83 61.16 61.59
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 0 1943
magnitude of the variables in the problem; it may change from
problem to problem.
The objective function considered for minimization,
finally, is thus:
Minimize
f ¼ �Jþ 107X5
i¼1
G2i (28)
where
G1 ¼maxf0; ðT2 � T1Þg (29)
G2 ¼maxf0; ð495� T1Þg (30)
G3 ¼maxf0; ðT1 � 535Þg (31)
G4 ¼maxf0; ð423� T2Þg (32)
G5 ¼maxf0; ðT2 � 523Þg (33)
The resulting optimization problems are solved using the
proposed algorithm, DE.
6. Numerical solution
The formulated model composed of 11 ordinary differential
equations and the associated boundary conditions lends itself
to be an initial value problem. The algebraic equations in the
model incorporate the initial conditions, the reaction rates, the
ideal gas assumption, as well as aforementioned correlations
for the heat and mass transfer coefficients and the physical
properties of fluids. These equations along with the discretized
ordinary differential equations using backward finite differ-
ence form a set of non-linear algebraic equations. The reactor
length is then divided into 100 separate sections and the Gauss–
Newton method in MATLAB programming environment is used
to solve the non-linear algebraic equations in each section.
7. Results and discussion
The model of methanol synthesis side was validated against
conventional methanol synthesis reactor for a special case of
constant coolant temperature under the design specifications.
The comparison between simulation and plant data for
conventional methanol synthesis reactor is shown in Table 4.
It was observed that the model performed satisfactorily well
under special case of industrial conditions and the observed
plant data were in good agreement with simulation data.
In this section, optimal operating conditions in the co-
current membrane thermally coupled reactor are analyzed
and the predicted mole fraction, conversion and temperature
profiles are presented. The performance of the thermally
coupled membrane reactor is analyzed, using different oper-
ating variables, for methanol yield, cyclohexane conversion
and hydrogen recovery yield as follows:
Methanol yield ¼ FCH3OH; out
FCO; in þ FCO2 ; in(34)
Cyclohexane conversion ¼ FC6H12 ; in � FC6H12 ; out
FC6H12 ; in(35)
Hydrogen recovery yield ¼ FH2 ;3
FC6H12 ; in(36)
The reaction scheme for cyclohexane dehydrogenation
indicates that if all cyclohexane is converted to benzene and
hydrogen, then ideally, the hydrogen yield reaches the value of
three.
7.1. Base case
In order to establish a reference point, so that the influence of
various parameters can be evaluated, calculations are first
carried out for a ‘‘base case’’ and the operating conditions used
for all sides of the reactor are given in Table 5. Operating
conditions for the methanol synthesis side are similar to those
used by Rezaie et al. [46]. The inlet composition of the methanol
synthesis reaction is typical of industrial methanol synthesis
process. It corresponds to a hydrogen:carbon-dioxide ratio of 7
having small amount of CH3OH, CO and H2O together with inert
gases of CH4 and N2. On the endothermic side, the inlet mole
fraction of cyclohexane that is diluted with argon is the same as
that presented by Jeong et al. [31]. Thus, the base case aims to
investigate the situation when the cyclohexane dehydrogena-
tion is used to consumption the generated heat from methanol
synthesis and to cool down it, resulting in a higher temperature
atfirstpartsofexothermicsidefor higherkineticsconstantsand
then reducing temperature gradually at the end parts of reactor
for increasing thermodynamics equilibrium which is similar to
the temperature profile along a tube filled with catalyst within
a methanol conventional reactor. This allows comparison of the
methanol synthesis process in the optimized membrane ther-
mally coupled reactor (OMTCR) with conventional methanol
reactor (CMR) for similar thermal behavior. The simulation
results of the membrane reactor in the endothermic side are not
compared with any reference case.
7.2. Simulation and optimization
With due attention to subjects of Section 5, the optimization
approach is maximization of the mole fractions of methanol,
Table 5. – Operating conditions for methanol synthesisprocess (exothermic side), dehydrogenation ofcyclohexane to benzene (endothermic side) andpermeation side.
Parameter Value
Exothermic side
Gas phase
Feed composition (mole fraction)
CH3OH 0.005
CO2 0.094
CO 0.046
H2O 0.0004
H2 0.659
N2 0.093
CH4 0.1026
Catalyst particle
Density (kg m�3) 1770
Particle diameter (m) 5.47 � 10�3
Heat capacity (kJ kg�1 K�1) 5.0
Thermal conductivity (W m�1 K�1) 0.004
Specific surface area (m2 m�3) 626.98
Ratio of void fraction to tortuosity of catalyst particle 0.123
Length of reactor (m) 7.022
Bed void fraction 0.39
Density of catalyst bed (kg m�3) 1140
Tube inner diameter (m) 3.8 � 10�2
Tube outer diameter (m) 4.2 � 10�2
Wall thermal conductivity (W m�1 K�1) 48
Endothermic side
Gas phase
Feed compositiona (mole fraction)
C6H12 0.1
C6H6 0.0
H2 0.0
Ar 0.9
Inlet pressurea (Pa) 1.013 � 105
Particle diameterb (m) 3.55 � 10�3
Bed void fraction 0.39
Shell inner diameter (m) 6 � 1�2
Permeation side
Feed composition (mole fraction)
Ar (sweep gas) 1.0
H2 0.0
Total molar flow rate (mol s�1) 1.0
Inlet pressure (Pa) 1.013 � 105
Membrane thickness (m) 6 � 10�6
Thermal conductivity of membrane (W m�1 K�1) 153.95
Shell inner diameter (m) 8 � 10�2
a Obtained from Jeong et al. [31].
b Obtained from Markatos et al. [47].
Table 6. – The optimized parameters for membranethermally coupled reactor.
Parameter Value
Inlet temperature of exothermic side (T01), K 495.9
Inlet temperature of endothermic side (T02), K 476.4
Initial molar flow rate of exothermic side (F01), mol s�1 0.255
Initial molar flow rate of endothermic side (F02), mol s�1 0.082
Inlet temperature of permeation side (T03), K 535.0
Inlet pressure of exothermic side (P01), bar 80.00
Objective function 0.195
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 01944
benzene and hydrogen in permeation side through the optimal
initial molar flow rate of exothermic and endothermic stream,
inlet temperature of exothermic, endothermic and permeation
sides, and inlet pressure of exothermic side. Differential
evolution method is applied to determine the optimal reactor
operating conditions for hydrogen, methanol and benzene
production process in a membrane thermally coupled reactor.
The results of the optimization for membrane thermally
coupled reactor (MTCR) (using Differential evolution method
and MATLAB programming) are summarized in Table 6.
According to Le Chatelier’s principle, when an independent
variable of a system at equilibrium is changed, the equilibrium
shifts in the direction that tends to reduce the effect of the
change. When pressure of the methanol synthesis reaction is
increased, the equilibrium shifts in the direction to increase
the methanol production. Therefore, the optimal inlet pres-
sure of exothermic side is 80 bar (upper bound of pressure).
The simulation of membrane thermally coupled reactor is
carried out using optimization results in Table 6 and the results
of this simulation are shown in several figures. Fig. 3(a)–(e)
shows the comparison of mole fraction of components in
exothermic side of optimized membrane thermally coupled
reactor (OMTCR) with conventional methanol reactor (CMR).
Fig. 3(a) illustrates the mole fraction profile of methanol along
the reactor, at steady-state for exothermic side of OMTCR and
CMR. Fig. 3(b)–(e) presents similar results for other components.
The important point as illustrated in these figures is a reaction
kinetic controlling in the upper sections of reactor while in
other sections, the rate of reactions has decreased to its equi-
librium value and equilibrium is controlling.
As it can be seen in Fig. 3(a), the comparison of methanol
mole fraction in exothermic side of OMTCR with CMR shows
that the methanol mole fraction in output of OMTCR is
increased by 16.3%. This novel configuration leads to delay in
thermodynamic equilibrium, while CMR reaches to equilib-
rium in the first half of the reactor. Fig. 4 is simultaneous plot
of mole fraction for cyclohexane, benzene and hydrogen in
the endothermic side of OMTCR along the reactor axis.
Hydrogen permeation from endothermic side to separation
side results in shifting of the reaction to right and increases
the benzene mole fraction and its purity. Fig. 5 shows profile of
hydrogen mole fraction along the reactor axis in the perme-
ation side of OMTCR. While there is a difference between
hydrogen partial pressure in the endothermic and permeation
side, hydrogen can continuously pass from the endothermic
side into the permeation side. Therefore, the hydrogen mole
fraction in the permeation side should increase along the
reactor length.
Fig. 6(a)–(c) shows axial temperature profiles for CMR and
OMTCR in exothermic side, endothermic side and permeation
side, respectively. In addition, the highest temperature is
observed at the exothermic side, since this is where heat is
generated. Part of this heat is used to drive the endothermic
reaction and the rest to heat the reaction mixtures in both
sides. The temperature of the dehydrogenation side is always
lower than that of the exothermic side in order to make
a driving force for heat transfer from the solid wall. Along the
exothermic side of OMTCR, temperature increases smoothly
and a hot spot develop as demonstrated in Fig. 6(a) and then
0 0.2 0.4 0.6 0.8 1
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
HC
3n
oit
ca
rf
elo
m H
O
Dimensionless length
OMTCR
CMR
0 0.2 0.4 0.6 0.8 1
0.07
0.075
0.08
0.085
0.09
0.095
OC
2n
oit
ca
rf
elo
m
Dimensionless length
OMTCR
CMR
0 0.2 0.4 0.6 0.8 1
0.015
0.02
0.025
0.03
0.035
0.04
0.045
0.05
noi
tc
ar
f el
om
OC
Dimensionless length
OMTCR
CMR
0 0.2 0.4 0.6 0.8 1
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
H2
noi
tc
ar
f el
om
O
Dimensionless length
OMTCR
CMR
0 0.2 0.4 0.6 0.8 1
0.58
0.6
0.62
0.64
0.66
H2
noi
tc
ar
f el
om
Dimensionless length
OMTCR
CMR
a b
c d
e
Fig. 3 – Comparison of (a) CH3OH, (b) CO2, (c) CO, (d) H2O and (e) H2 mole fraction along the reactor axis between exothermic
side of OMTCR and CMR.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 0 1945
decreases to 495 K. Note that in the exothermic side, the outlet
temperature is the same as the inlet temperature. According to
Le Chatelier’s principle, temperature reduction in the second
half of the OMTCR leads to shift the reaction in the direction of
methanol production. The maximum temperature of OMTCR
moves towards the right with respect to CMR. This leads to
increase of kinetic duration of methanol synthesis reaction
and consequently delay in thermodynamic equilibrium.
At the entrance of dehydrogenation side in OMTCR, the
temperature increases and a hot spot form and then the
temperature decreases (see Fig. 6(b)). As can be seen in
Fig. 6(c), temperature profile in permeation side of OMTCR
decreases along the reactor length. Also, temperature differ-
ence between exothermic and endothermic sides is shown in
this figure.
Fig. 7(a) shows the variation of rate of reaction for both sides
of OMTCR. Comparing the values for the reaction rates present
in the exothermic side, it can be seen that the predominant
reactions are hydrogenation of CO and hydrogenation of CO2;
however water–gas shift reaction can be neglected, its contri-
bution being significant. At the reactor entrance, the rate of
cyclohexane dehydrogenation increases rapidly which is due
to increase in temperature of endothermic side (see Fig. 6(b)).
Fig. 7(b) illustrates the variation of the generated and
0 0.2 0.4 0.6 0.8 1
0
0.02
0.04
0.06
0.08
0.1
noi
tc
ar
f el
om
Dimensionless length
C6H
12
C6H
6
H2
Fig. 4 – Profiles of cyclohexane, benzene and hydrogen
mole fraction along the reactor axis in the endothermic
side of OMTCR.
0 0.2 0.4 0.6 0.8 1
490
500
510
520
530
540
K ,e
ru
ta
re
pm
eT
Dimensionless length
OMTCR
CMR
0 0.2 0.4 0.6 0.8 1
470
480
490
500
510
520
530
K ,e
ru
ta
re
pm
eT
Dimensionless length
OMTCR
CMR
514
521
528
535
K ,e
dis
noi
ta
em
re
p f
o e
ru
10
15
20
25
K ,e
cn
er
ef
fid
er
ut
ar
e
a
b
c
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 01946
consumed heat flux from the exothermic and the endothermic
reaction, respectively, and also transferred heat into endo-
thermic side, and transferred heat from permeation and
exothermic side along the reactor. In the first half of the
reactor, methanol reaction proceeds faster than dehydroge-
nation and as a result more heat is produced by the exothermic
reaction than consumed by the endothermic one. The excess
heat raises the temperature of the system in the first half of the
reactor as illustrated by the temperature profile in Fig. 6(a).
In this region, the generated heat flux is higher than the
consumed one. The system heats up and a peak in the
generated heat flux is observed. Afterward, the generated heat
flux decreases rapidly, mainly due to fuel depletion. The
opposite situation occurs when the consumed heat flux is
higher than the generated one. If the consumed heat flux is
higher than the generated one, the system starts to cool down
resulting to low temperature, which in turn decreases both
reaction rates. Thus, after a certain position along the reactor
(dimensionless length ¼ 0.25 in Fig. 7(b)) the generated heat
0 0.2 0.4 0.6 0.8 1
0
0.005
0.01
0.015
0.02
0.025
H2
noi
tc
ar
f el
om
Dimensionless length
Fig. 5 – Profile of hydrogen mole fraction along the reactor
axis in the permeation side of OMTCR.
0 0.2 0.4 0.6 0.8 1
500
507
Dimensionless length
ta
re
pm
eT
0 0.2 0.4 0.6 0.8 1
0
5
pm
eT
Fig. 6 – Variation of temperature for (a) exothermic side,
and (b) endothermic side of OMTCR and CMR and (c)
permeation side of OMTCR and temperature difference
between exothermic and endothermic sides along the
reactor axis.
flux becomes lower than the consumed one, which coincides
with a hot spot development (see Fig. 6(a) and (b)).
The transferred heat into the endothermic side is
summation of transferred heat from the permeation side by
convection and hydrogen diffusion, and transferred heat from
-1.5
-0.5
0.5
1.5
2.5
3.5
4.5
Dimensionless length
ml
om
,n
oit
ca
er
fo
et
aR
3-
s1
-
0 0.2 0.4 0.6 0.8 1
0.25
0.5
0.75
1
1.25
Rate o
f reactio
n fo
r cyclo
hexan
e
deh
yd
ro
gen
atio
n, m
ol m
-3s
-1
Hydrogenation of CO
Hydrogenation of CO2
Water-gas shift
(a)
0 0.2 0.4 0.6 0.8 1
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
Dimensionless length
W,
xul
ft
ae
H
Generated heat in exothermic side
Consumed heat in endothermic side
Transferred heat into endothermic side
Transferred heat from permeation side
Transferred heat from exothermic side
a
b
Fig. 7 – Variation of (a) rate of reaction for both sides and (b) generated and consumed heat flux in exothermic and
endothermic side respectively and also, transferred heat into endothermic side, and transferred heat from permeation and
exothermic side along the reactor for the OMTCR.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 0 1947
the exothermic side. Along the reactor length, the heat values
consumed by the endothermic side and transferred into the
endothermic side are close to each other. This demonstrates
the efficient thermal communication between the all sides,
and which is due to high solid wall and membrane thermal
conductivity. At the reactor entrance, the transferred heat into
the endothermic side is upper than the consumed heat by the
endothermic side, which is due to high temperature differ-
ence. A decrease in the transferred heat from exothermic side
is observed near the reactor entrance, and is associated to the
low temperature difference between exothermic and endo-
thermic sides in that region as shown in Fig. 6(c). This
reduction leads to decrease in the transferred heat into
endothermic side. A maximum in the reaction heat fluxes
consumed and transferred into the endothermic side are
located at the same axial position, namely 0.25. After this
position along the reactor, the consumed heat by the dehy-
drogenation side becomes larger than the transferred heat
into the endothermic side and the system starts to cool down
(see Fig. 6(b)). As can be seen in Fig. 6(c), in the permeation
side, the temperature profile decreases smoothly which is due
to variation of transferred heat from the permeation side (see
Fig. 7(b)).
In Fig. 8(a) and (b), respectively, methanol yield in the
exothermic side of OMTCR and CMR, cyclohexane conversion
in the endothermic side and hydrogen recovery yield in the
permeation side of OMTCR along the reactor are shown. In
OMTCR, cyclohexane reaches 99.9% conversion, and meth-
anol 0.4773 and 0.4182 yields in the exothermic side of OMTCR
and CMR, respectively. The comparison of methanol yield in
0 0.2 0.4 0.6 0.8 1
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
dlei
Yl
on
ah
te
M
Dimensionless length
OMTCR
CMR
0
0.2
0.4
0.6
0.8
1
Dimensionless length
noi
sr
ev
no
ce
na
xe
hol
cy
C
0 0.2 0.4 0.6 0.8 1
0
0.6
1.2
1.8
2.4
3
dlei
yy
re
vo
ce
rn
eg
or
dy
H
a
b
Fig. 8 – Variation of (a) methanol yield in exothermic side of
OMTCR and CMR, (b) cyclohexane conversion in
endothermic side and hydrogen recovery yield in
permeation side of OMTCR along the reactor length.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 01948
OMTCR with CMR shows that the methanol yield in output of
OMTCR is increased by 14.13%. Also, as it can be seen in
Fig. 8(b), hydrogen recovery of OMTCR is 2.71 yields. Overall,
the optimized operating parameters for this case lead to effi-
cient coupling of the two reactions.
8. Conclusion
Optimization of methanol synthesis reaction coupled with
dehydrogenation of cyclohexane to benzene by means of
indirect heat transfer and hydrogen production in a catalytic
membrane thermally coupled reactor was studied by a one-
dimensional model. The reactor consists of two separated side
for exothermic and endothermic reactions and a permeation
side for hydrogen recovery. A base case was generated
considering similar operating conditions to industrial meth-
anol reactor. Differential evolution (DE) method, an excep-
tionally simple evolution strategy, is applied to determine the
optimal reactor operating conditions. Maximization of outlet
mole fractions of methanol, benzene and hydrogen in the
permeation side is considered as the objective function. It is
shown that suitable amount of initial molar flow rate in the
exothermic and endothermic side, inlet temperature of all
sides and inlet pressure of exothermic side can provide the
necessary heat to heat up the mixtures and to drive the
endothermic process at the same time. The short distance
between the heat sink and transferred heat increases the
efficiency of heat transfer. This new configuration represents
some important improvement in comparison to conventional
methanol reactor as follows: reduces the size of the reactors;
lower outlet temperature of product stream in exothermic
side and so enhances the equilibrium conversion; methanol
mole fraction in output of reactor is increased by 16.3%;
produces pure hydrogen in the permeation side; increase in
kinetics duration of methanol synthesis reaction and delay in
thermodynamics equilibrium; benzene is also produced as an
additional valuable product; and autothermality is achieved
within the reactor. In the optimized reactor, cyclohexane
reaches 99.9% conversion in the endothermic side, methanol
0.4773 yields in the exothermic side and hydrogen recovery
2.71 yields in the permeation side. The comparison of meth-
anol yield in OMTCR with CMR shows that the methanol yield
in output of OMTCR is increased by 14.13%. The results indi-
cate that methanol synthesis reaction and cyclohexane
dehydrogenation in a membrane thermally coupled reactor is
feasible and beneficial.
Nomenclature
av specific surface area of catalyst pellet, m2 m�3
Ac cross section area of each tube, m2
Ai inside area of inner tube, m2
Ao outside area of inner tube, m2
c total concentration, mol m�3
Cp specific heat of the gas at constant pressure, J mol�1
dp particle diameter, m
Di inside diameter, m
Dij binary diffusion coefficient of component i in j,
m2 s�1
Dim diffusion coefficient of component i in the mixture,
m2 s�1
Do outside diameter, m
DH hydraulic diameter, m
Ep activation energy of permeability, kJ mol�1
fi partial fugacity of component i, bar
F total molar flow rate, mol s�1
G mass velocity, kg m�2 s�1
hf gas–solid heat transfer coefficient, W m�2 K�1
hi heat transfer coefficient between fluid phase and
reactor wall in exothermic side, W m�2 K�1
ho heat transfer coefficient between fluid phase and
reactor wall in endothermic side, W m�2 K�1
DHf,i enthalpy of formation of component i, J mol�1
JH2 hydrogen-permeation rate in Pd/Ag membrane,
mol m�1 s�1
k rate constant of dehydrogenation reaction,
mol m�3 Pa�1 s�1
k1 rate constant for the 1st rate equation of methanol
synthesis reaction, mol kg�1 s�1 bar�1/2
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 0 1949
k2 rate constant for the 2nd rate equation of methanol
synthesis reaction, mol kg�1 s�1 bar�1/2
k3 rate constant for the 3rd rate equation of methanol
synthesis reaction mol kg�1 s�1 bar�1/2
kg mass transfer coefficient for component i, m s�1
K conductivity of gas phase, W m�1 K�1
KB adsorption equilibrium constant for benzene, Pa�1
Ki adsorption equilibrium constant for component i in
methanol synthesis reaction, bar�1
Kp equilibrium constant for dehydrogenation reaction,
Pa3
Kpi equilibrium constant based on partial pressure for
component i in methanol synthesis reaction
Kw thermal conductivity of reactor wall, W m�1 K�1
L reactor length, m
Mi molecular weight of component i, g mol�1
N number of components (N ¼ 6 for methanol
synthesis reaction, N ¼ 3 for dehydrogenation
reaction)
P permeability of hydrogen through Pd/Ag membrane,
mol m�2 s�1 Pa�1/2
P0 pre-exponential factor of hydrogen permeability,
mol m�2 s�1 Pa�1/2
P total pressure (for exothermic side: bar; for
endothermic side, Pa
Pi partial pressure of component i, Pa
r1 rate of reaction for hydrogenation of CO, mol kg�1 s�1
r2 rate of reaction for hydrogenation of CO2,
mol kg�1 s�1
r3 rate of reversed water-gas shift reaction,
mol kg�1 s�1
r4 rate of reaction for dehydrogenation of cyclohexane,
mol m�3 s�1
ri Reaction rate of component i (for exothermic
reaction: mol kg�1 s�1; for endothermic reaction:
mol m�3 s�1)
R universal gas constant, J mol�1 K�1
Rp particle radius, m
Re Reynolds number
Sci Schmidt number of component i
T temperature, K
u superficial velocity of fluid phase, m s�1
ug linear velocity of fluid phase, m s�1
U overall heat transfer coefficient between exothermic
and endothermic sides, W m�2 K�1
vci critical volume of component i, cm3 mol�1
yi mole fraction of component i, mol mol�1
z axial reactor coordinate, m
Greek letters
aH hydrogen-permeation rate constant,
mol m�1 s�1 Pa�1/2
m viscosity of fluid phase, kg m�1 s�1
r density of fluid phase, kg m�3
rb density of catalytic bed, kg m�3
s tortuosity of catalyst
Superscripts
g in bulk gas phase
s at surface catalyst
Subscripts
0 inlet conditions
B benzene
C cyclohexane
i chemical species
j reactor side (1: exothermic side, 2: endothermic side,
3: permeation side)
k reaction number index
r e f e r e n c e s
[1] Rahimpour MR, Ghader S. Theoretical investigation of a Pd-membrane reactor for methanol synthesis. Chem Eng Tech2003;26:902–7.
[2] Jahanmiri A, Eslamloueyan R. Optimal temperature profile inmethanol synthesis reactor. Chem Eng Commun 2002;189:713–41.
[3] Rahimpour MR, Elekaei H. Enhancement of methanolproduction in a novel fluidized-bed hydrogen-permselectivemembrane reactor in the presence of catalyst deactivation.Int J Hydrogen Energy 2009;34:2208–23.
[4] Rahimpour MR, Alizadehhesari K. Enhancement of carbondioxide removal in a hydrogen-permselective methanolsynthesis reactor. Int J Hydrogen Energy 2009;34:1349–62.
[5] Rahimpour MR, Lotfinejad M. A comparison of co-currentand counter-current modes of operation for a dual-typeindustrial methanol reactor. Chem Eng Process 2008;47:1819–30.
[6] Askari F, Rahimpour MR, Jahanmiri A, KhosravanipourMostafazadeh A. Dynamic simulation and optimization ofa dual-type methanol reactor using genetic algorithms.Chem Eng Technol 2008;31:513–24.
[7] Rahimpour MR, Elekaei Behjati H. Dynamic optimization ofmembrane dual-type methanol reactor in the presence ofcatalyst deactivation using genetic algorithm. Fuel ProcessTechnol 2009;90:279–91.
[8] Parvasi P, Khosravanipour Mostafazadeh A, Rahimpour MR.Dynamic modeling and optimization of a novel methanolsynthesis loop with hydrogen-permselective membranereactor. Int J Hydrogen Energy 2009;34:3717–33.
[9] Rahimpour MR, Parvasi P, Setoodeh P. Dynamic optimizationof a novel radial-flow, spherical-bed methanol synthesisreactor in the presence of catalyst deactivation usingdifferential evolution (DE) algorithm. Int J Hydrogen Energy2009;34:6221–30.
[10] Agar DW. Multifunctional reactors: old preconceptions andnew dimensions. Chem Eng Sci 1999;54:1299–305.
[11] Zanfir M, Gavriilidis A. Catalytic combustion assistedmethane steam reforming in a catalytic plate reactor. ChemEng Sci 2003;58:3947–60.
[12] Dautzenberg FM, Mukherjee M. Process intensification usingmultifunctional reactors. Chem Eng Sci 2001;56:251–67.
[13] Itoh N, Wu TH. An adiabatic type of palladium membranereactor for coupling endothermic and exothermic reactions.J Memb Sci 1997;124:213–22.
[14] Kolios G, Frauhammer J, Eigenberger G. Efficient reactorconcepts for coupling of endothermic and exothermicreactions. Chem Eng Sci 2002;57:1505–10.
[15] Moustafa TM, Elnashaie SSEH. Simulation production ofstyrene and cyclohexane in an integrated membrane reactor.J Memb Sci 2000;178:171–84.
[16] Ch Fukuhara, Igarashi A. Performance simulation of a wall-type reactor in which exothermic and endothermic reactionsproceed simultaneously, comparing with that of a fixed-bedreactor. Chem Eng Sci 2005;60:6824–34.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 3 5 ( 2 0 1 0 ) 1 9 3 6 – 1 9 5 01950
[17] Ramaswamy RC, Ramachandran PA, Dudukovic MP.Recuperative coupling of exothermic and endothermicreactions. Chem Eng Sci 2006;61:459–72.
[18] Ramaswamy RC, Ramachandran PA, Dudukovic MP.Coupling exothermic and endothermic reactions in adiabaticreactors. Chem Eng Sci 2008;63:1654–67.
[19] Abo-Ghander NS, Grace JR, Elnashaie SSEH, Lim CJ. Modelingof a novel membrane reactor to integrate dehydrogenation ofethylbenzene to styrene with hydrogenation of nitrobenzeneto aniline. Chem Eng Sci 2008;63:1817–26.
[20] Khademi MH, Jahanmiri A, Rahimpour MR. Optimization ofmethanol synthesis and cyclohexane dehydrogenation ina thermally coupled reactor using differential evolution (DE)method. Int J Hydrogen Energy 2009;34:6930–44.
[21] Amon B, Redlingshofer H, Klemm E, Dieterich E, Emig G.Kinetic investigation of deactivation by coking of a noble metalcatalyst in the catalytic hydrogenation of nitrobenzene usinga catalytic wall reactor. Chem Eng Processing 1999;38:395–404.
[22] Wieland IS, Melin IT, Lamm IA. Membrane reactors forhydrogen production. Chem Eng Sci 2002;57:1571–6.
[23] Rahimpour MR, Asgari A. Production of hydrogen from purgegases of ammonia plants in a catalytic hydrogen-permselective membrane reactor. Int J Hydrogen Energy2009;34:5795–802.
[24] Iulianelli A, Longo T, Basile A. Methanol steam reformingreaction in a Pd–Ag membrane reactor for CO-free hydrogenproduction. Int J Hydrogen Energy 2008;33:5583–8.
[25] Forghani AA, Elekaei H, Rahimpour MR. Enhancement ofgasoline production in a novel hydrogen-permselectivemembrane reactor in Fischer–Tropsch synthesis of GTLtechnology. Int J Hydrogen Energy 2009;34:3965–76.
[26] Shirasaki Y, Tsuneki T, Ota Y, Yasuda I, Tachibana S,Nakajima H, et al. Development of membrane reformersystem for highly efficient hydrogen production from naturalgas. Int J Hydrogen Energy 2009;34:4482–7.
[27] Khademi MH, Jahanmiri A, Rahimpour MR. A novelconfiguration for hydrogen production from coupling ofmethanol and benzene synthesis in a hydrogen-permselective membrane reactor. Int J Hydrogen Energy2009;34:5091–107.
[28] Graaf GH, Scholtens H, Stamhuis EJ, Beenackers AACM.Intra-particle diffusion limitations in low-pressure methanolsynthesis. Chem Eng Sci 1990;45:773–83.
[29] Graaf GH, Sijtsema PJJM, Stamhuis EJ, Joosten GEH. Chemicalequilibrium in methanol synthesis. Chem Eng Sci 1986;41:2883–90.
[30] Itoh N. A membrane reactor using palladium. AIChE J 1987;33:1576–8.
[31] Jeong BH, Sotowa KI, Kusakabe K. Catalytic dehydrogenationof cyclohexane in an FAU-type zeolite membrane reactor.J Memb Sci 2003;224:151–8.
[32] Hara S, Xu WC, Sakaki K, Itoh N. Kinetics and hydrogenremoval effect for methanol decomposition. Ind Eng ChemRes 1999;38:488–92.
[33] Barbieri G, Maio FPD. Simulation of the methane steamre-forming process in a catalytic Pd-membrane reactor. IndEng Chem Res 1997;36:2121–212.
[34] Shu G, Grandjean BPA, Kaliaguine S. Methane steamreforming in a symmetric Pd– and Pd–Ag/porous ssmembrane reactor. Appl Catal 1994;119:305–25.
[35] Lindsay AL, Bromley LA. Thermal conductivity of gasmixture. Ind Eng Chem 1950;42:1508–10.
[36] Cussler EL. Diffusion, mass transfer in fluid systems.Cambridge University Press; 1984.
[37] Wilke CR. Estimation of liquid diffusion coefficients. ChemEng Prog 1949;45:218–24.
[38] Reid RC, Sherwood TK, Prausnitz J. In: The properties of gasesand liquids. 3rd ed. New York: McGraw-Hill; 1977.
[39] Smith JM. Chemical engineering kinetics. New York:McGraw-Hill; 1980.
[40] Holman JP. Heat transfer. New York: McGraw-Hill; 1989.[41] Price K, Storn R. Differential evolutiondA simple evolution
strategy for fast optimization. Dr Dobb’s J 1997;22:18–24. 78.[42] Babu BV, Angira R. Modified differential evolution (MDE) for
optimization of non-linear chemical processes. ComputChem Eng 2006;30:989–1002.
[43] Angira R, Babu BV. Optimization of process synthesis anddesign problems: a modified differential evolution approach.Chem Eng Sci 2006;61:4707–21.
[44] Price K, Storn R. Homepage of differential evolution as on,http://www.ICSI.Berkeley.edu/wstorn/code.html; May, 2006.
[45] Jeong BH, Sotowa KI, Kusakabe K. Modeling of an FAU-typezeolite membrane reactor for the catalytic dehydrogenationof cyclohexane. Chem Eng J 2004;103:69–75.
[46] Rezaie N, Jahanmiri A, Moghtaderi B, Rahimpour MR.A comparison of homogeneous and heterogeneous dynamicmodels for industrial methanol reactors in the presenceof catalyst deactivation. Chem Eng Process 2005;44:911–21.
[47] Koukou MK, Chaloulon G, Papayannakos N, Markatos NC.Mathematical modeling of the performance of non-isothermal membrane reactors. Int J Heat Mass Transfer1997;40:2407–17.