You are on page 1of 10

Journal of Industrial and Engineering Chemistry 20 (2014) 19421951

Contents lists available at ScienceDirect

Journal of Industrial and Engineering Chemistry


journal homepage: www.elsevier.com/locate/jiec

Fixed-bed reactor modeling for methanol to dimethyl ether (DME)


reaction over g-Alumina using a new practical reaction rate model
Mohammad Ghavipour *, Reza Mosayebi Behbahani
Gas Engineering Department, Petroleum University of Technology, 63431 Ahwaz, Iran

A R T I C L E I N F O A B S T R A C T

Article history: Dimethyl ether (DME) synthesis reaction rate was studied over a commercial sample of g-Alumina to
Received 11 June 2013 investigate the accuracy of the most applicable rate models. Due to the deviation of the former proposed
Accepted 12 September 2013 models especially at temperatures below 593 K, a new simple empirical rate model was proposed.
Available online 19 September 2013
Besides, previous proposed correlations for methanol dehydration equilibrium constant were examined
experimentally and a new equation was developed. Subsequently, one dimensional unsteady state
Keywords: heterogeneous model was applied to simulate adiabatic and non-adiabatic xed-bed reactors.
Methanol dehydration
Temperature prole and methanol conversion along the reactors were predicted while varying feed
DME
g-Alumina rate and feed temperature.
Fixed-bed reactor modeling 2013 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights
Reaction rate reserved.

1. Introduction aluminum) and Brnsted acid sites. The reaction takes place at
the temperature range of 423523 K. The HZSM-5 catalyst is
Considering environmental pollution, energy security and deactivated sooner than the g-Alumina because a layer of coke
future oil supplies, the global community is seeking new covers its strong acid sites and it has lower DME selectivity
alternative fuels. A promising alternative fuel could be dimethyl especially at higher temperatures. Modication of the strong
ether (DME) due to low NOx and SOx emission and because it acidic sites by Na, Si or P inhibits hydrocarbon formation (such
does not have large issues with toxicity, production, infrastruc- as methane) and thereby enhances the catalyst stability [1014].
ture, and transportation [13]. It also can be used as hydrogen- By taking all of the above facts into account, it seems that the g-
rich feed of fuel-cells [4,5] or as a substitute fuel in domestic Alumina is a better choice and its commercial use in large extent
appliances [6]. DME synthesis can be performed through two for this reaction proves this opinion.
routes: direct synthesis (i.e. syngas to DME over hybrid The mechanism of methanol adsorption, dehydration and
catalysts) and indirect synthesis (i.e. methanol dehydration to specially, formation of the rst C-C bond and the nature of the
DME) [7]. g-Alumina and HZSM-5 are the most common intermediates involved, are still not fully understood, but they
catalysts for DME indirect synthesis reaction. g-Alumina tends have been studied extensively and different rate equations have
to adsorb water on its surface and thereby loses its activity been developed till now (see Table 1) [15]. Most of these rate
because of its hydrophilic nature. When pure methanol is used equations have been derived from experiments conducted at
as the process feed, the catalyst deactivation occurs very slowly conditions far away from an industrial reactor. Almost all of these
and this hydrophilic nature is not a considerable problem [8]. experiments were performed with diluted methanol by nitrogen
The g-Alumina modied by silica or phosphorous has shown and/or water, as the reactor feed. In spite of that, to meet the
better performance compared to the untreated one. The best highest production in the industrial reactors, pure methanol is
operating temperatures are from 523 to 673 K and the amounts consumed as the feed. However, water in the reactor feed will
of coking and by-products are very low [911]. The activity of deactivate the g-Alumina by covering the catalyst active sites [8]
HZSM-5 is higher than the g-Alumina because of more strong and also will retard the DME production reaction (i.e. forward
acidic sites. Unlike the g-Alumina, which exhibits only Lewis reaction as indicated below) and speed up the DME consumption
acidity, HZSM-5 has both Lewis (due to the extra-framework reaction (i.e. backward reaction) according to Le Chateliers
principle in equilibrium reactions

* Corresponding author. Tel.: +98 6115550868; fax: +98 6115550868.


E-mail addresses: Ghavipour@put.ac.ir, M.Ghavipour@gmail.com, lghavipour@-
2CH3 OH ? CH3 OCH3 H2 O
put.ac.ir (M. Ghavipour).

1226-086X/$ see front matter 2013 The Korean Society of Industrial and Engineering Chemistry. Published by Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jiec.2013.09.015
M. Ghavipour, R.M. Behbahani / Journal of Industrial and Engineering Chemistry 20 (2014) 19421951 1943

Table 1
Nomenclature Methanol dehydration reaction rate models.
 
Bercic and Levec (1992) [15] ks k2
M
2 C D C W
CM K
I
rM
av specic catalyst surface area (m2/m3) 12kM C M 0:5 kW C W
4

Ci molar concentration of component i (mol/m3) Gates and Johanson (1971) [37]


rM
ks k2 C2
M M II
1kM C M kW C W 2
Cp heat capacity of gas mixture at constant pressure
Figueras et al. (1971) [38] 0:5
k s k M CM III
(J/mol K) rM k 1k 0:5
M CM kW C W

dp catalyst particle diameter (m) Kallo and Knozinger (1967) [39] C 0:5 IV
rM k C 0:5 k
M
dt reactor diameter (m) M 2 CW

hf overall heat transfer coefcient between gas and Rubio et al. (1980) [40] 0:5
rM k1  CM 0:5
 k 2  CW V
solid (J/m2 K s) Schmitz (1978) [41] rM k1 k2  C M VI
 
kgi mass transfer coefcient for component i (m/s) This work C
rM k0 expEa =RT C MeOH  KW VII
eq

K thermal conductivity (J/m K)


Keq equilibrium constant for methanol dehydration
reaction based on the realistic experimental data (i.e. more similar to the
L reactor length (m) industrial conditions).
Q liquid methanol ow rate (mL/min) Among the proposed kinetic equations for DME direct synthesis
r rate of methanol dehydration reaction (mol/kgcat s) from syngas over a bifunctional catalyst of CuOZnOAl2O3 and g-
R reactor diameter (m) Alumina in a xed-bed reactor [16,17], one of the most common
Re Reynolds number kinetic models [1820] is the combination of the methanol
synthesis model proposed by Graaf and Stamhuis [21] and the
Sci Schmidt number of component i
methanol dehydration model proposed by Bercic and Levec [15].
T temperature (K)
Therefore, the new proposed model at present work can also be
Tin temperature of the reactor inlet feed (K)
used in DME direct synthesis investigations as the methanol
Tr temperature of the reactor outside wall (K) dehydration model.
t time (s) Some studies have been performed on the modeling of xed-
us supercial velocity of gas phase (m/s) bed reactors for the methanol dehydration. Nasehi et al. [22]
U overall heat transfer coefcient (J/m2 K) simulated an industrial adiabatic xed-bed reactor for DME
W catalyst weight (g) production from methanol dehydration at steady state conditions
WHSV weight hourly space velocity (gMeOH/gcat h) and found that the difference between one and two dimensional
X methanol conversion (dimensionless) modeling for adiabatic xed bed reactor is negligible. Farsi et al.
z axial reactor coordinates (m) [23] simulated an industrial adiabatic reactor of DME synthesis
with accompanying feed pre heater and controlled it in dynamic
conditions. Fazlollahnejad et al. [24] investigated methanol
Greek letters
dehydration in a bench scale adiabatic packed bed reactor. They
m viscosity of uid phase (kg/m s)
assessed the effects of weight hourly space velocity and tempera-
r density (kg/m3) ture on the methanol conversion. Bercic and Levec [25] employed
e catalyst bed void fraction one-dimensional heterogeneous and pseudo-homogeneous plug
h effectiveness factor (dimensionless) ow models to assess an adiabatic xed bed reactor for the
DHr heat of reaction (J/molMeOH) catalytic dehydration of methanol to dimethyl ether. They found
that intraparticle mass transfer was the rate-controlling step while
Subscripts using 3 mm g-Alumina pellets as the catalyst. Farsi et al. [26]
B bulk of solid phase (catalyst bed) modeled a shell and tube xed-bed reactor and optimized it for
g in bulk of gas phase maximum DME production via adjusting the optimal temperature
distribution along the reactor using genetic algorithm. None of
i gaseous chemical species involved in the reaction
these studies has performed an unsteady state modeling to show
M methanol
the progress of the reaction along the reactor from the start up to
s solid phase
the steady state conditions and none of them has discussed in
W water detail the temperature prole through the reactor at different
weight hourly space velocities (WHSVs) and feed temperatures.
Superscripts Moreover, a clear comparison between adiabatic and non-
s at the catalyst surface adiabatic xed bed reactors for this reaction has not been
accomplished yet.
In the present study, the methanol conversion to DME over a
From another point of view, it should be mentioned that in the commercial sample of g-Alumina at different temperatures and
previous works, to nd the rate parameters, differential reactors WHSVs was measured experimentally. Due to the considerable
were used and because they did not use a partially converted feed difference between the equilibrium constant models proposed
(i.e. a feed that contains DME and water as well as methanol) and recently, a new reliable correlation was proposed that matched our
the reactor outlet conversion could not exceed 10% due to the experimental data satisfyingly. The Bercic and Levec rate model
restriction of differential reactor assumption; the reaction rates which is the most applicable rate model in the literature, shows
were measured at average conversions from zero to less than 5%, lower conversions than our experimental data. Therefore, various
causing low accuracy in the prediction of the rate parameters reaction rate models were examined via non-linear regression to
especially at higher conversions where the industrial reactors nd a new reliable model and a new empirical model based on the
operate. During the present study, integral reactor has been used experimental data was proposed. Finally, a one dimensional
and the methanol dehydration rate model has been investigated heterogeneous model was used to study the behavior of xed bed
1944 M. Ghavipour, R.M. Behbahani / Journal of Industrial and Engineering Chemistry 20 (2014) 19421951

Table 2 conductivity detectors. The column was TRB-5 (Teknokroma Co.-


Catalyst characterizations.
(95%) Dimethyl-(5%) diphenylpolysiloxane bonded and cross
Particle size Bulk BET surface NH3-TPD linked phase-length 30 m-inside diameter 0.53 mm) with helium
distribution density area (m2/gcat) as the carrier gas. The oven temperature was kept constant at
(g/cm3)
313 K and run time was 3 min. The 6-port gas sampling valve with
Temp (K) mmolNH3 =g cat a 0.05 mL sample loop was used to inject the gas samples into the
Dp > 0.060 mm 98% 402 0.295 capillary column. A calibration curve was used to determine the
Dp > 0.20 mm 4% 0.869 183.2 666 0.452 gas samples composition. The compositions of the reactor outlet
gases were monitored via the Gas Chromatograph until no change
observed by time and steady state conditions were obtained.
reactor for the DME synthesis via methanol dehydration. The
experimental data that had been acquired of our biggest size xed 3. Equilibrium constant estimation
bed reactor was employed to validate the simulation results of
non-adiabatic xed bed reactor. Two types of xed bed reactors Equilibrium constant of the methanol to dimethyl ether
(adiabatic and non-adiabatic) were simulated and the methanol reaction has been studied previously and some correlations have
conversion and the temperature prole of the reactor were been derived based on the experimental works or thermodynamic
described in detail. calculations. Hayashi and Moffat (1982) developed the equilibrium
constant using thermodynamic properties at 298 K and heat
2. Experiments capacity data at higher temperatures as following [27]:

In our experiments, a commercial sample of g-Alumina from ln K P 3440=T  1:67 ln T 2:39  104 T 0:055
Engelhard Corporation in powder form was used. Several catalyst  106 T 2 5:496 (1)
characterization tests have been performed such as BET, NH3-TPD,
Bulk density and particle size distribution that are listed in Table 2. Diep and Walnwright (1987) determined thermodynamic
A multipurpose micro reactor and catalyst test setup was equilibrium constant of the methanoldimethyl etherwater
employed for this work. Pure liquid methanol was injected to a pre system experimentally at temperatures from 498 to 623 K over
heater by means of a HPLC metering pump (working range of 0.1 a commercial g-Alumina catalyst in a ow reactor maintained at a
99.9 mL/min), then evaporated in the pre heater at the constant constant pressure of 200 kPa [27]
temperature of 453 K. The reactor consisted of two heating zones.
First zone was to raise the feed temperature to a desirable level and ln K P 2835:2=T 1:675 ln T  2:39  104 T  0:21
the other one was to maintain the reactor surrounding at a proper  106 T 2  13:360 (2)
temperature to minimize the heat losses. The thermocouples at the
second zone were arranged in a way that the operator could x the They also calculated the enthalpy of formation (DHf 298) and the
outer surface of the stainless steel reactor at a constant free Gibbs energy of formation (DGf 298) of dimethyl ether by use of
temperature (non-adiabatic reactor). There were other thermo- published thermochemical data together with the experimental
couples to report the temperatures of the pre heater, the feed, the data to be 180.22 and 109.66 kJ/mol, respectively.
reactor center and other critical sections. Finally, the equilibrium constant which is coupled to Bercic and
The experimental data were obtained from three different Levec model and has been used extensively in xed-bed reactor
sizes of reactors (i.e. outside diameter of 3/8, 1/2 and 3/4 inches modeling papers [2226].
with heights of 0.4, 1 and 1.5 inches respectively). Catalyst
weights of 1.05, 6.32, 12.64 g with the constant liquid methanol ln K P 3138=T 0:86 log T 1:33  103 T  1:23
ow rate of 2 mL/min were used (weight hourly space velocities
of 90, 15 and 7.5, respectively). The temperatures of the reactor  105 T 2 3:5  1010 T 3 (3)
zones were adjusted in a way that the reactor center temperature
A model based on the thermodynamic calculations has been
varied from 523 to 623 K. All of these reactors were assumed to
developed in the present work using the thermodynamic proper-
be isothermal at their center temperatures which were
ties of the pure substances which are listed in Table 3 [28]. It
measured through the submerged thermocouple. For the
should be mentioned that the pressure is not an effective
assessment of our xed bed reactor modeling, the biggest size
parameter in this reaction due to the conversion of two moles
reactor was used (i.e. outside diameter of 1/2 with height of 8
of methanol to one mole of water and one mole of DME (i.e. the
inches), 40 g catalyst was loaded and with different liquid
reaction occurs with no changes in gaseous moles).
methanol ow rates, various WHSVs were adjusted. Reactor
If the variation of the reaction heat with temperature is taken
pressure was maintained constant at 3 bar via a back pressure
into account, temperature dependency of the equilibrium constant
regulator installed at the reactor exit. Activity of the commercial
(Keq) will be found by using the integrated form of Vant Hoff
catalyst decreased during the rst day (activity depletion of 1
equation:
3% that was attributed to the strong unstable acidic sites) and the
conversion remained almost constant during 120 h. Therefore, ZT ZT
K 1 DHr
the experimental data were recorded during 3060 h after ln dT & DHr DHr0 rC P dT (4)
reactor startup. For every specic temperature or WHSV, fresh
K0 R T2
T0 T0
catalyst was used and data gathering started after 30 h of
reaction. The effect of coke formation on the catalyst activity was
rC P rC 1 rC 2  T rC 3  T 2 rC 4  T 3 rC 5
neglected, because the experiments were not long enough that
the deactivation of the catalyst could be monitored. Conse-  T4 & rC n
quently, the proposed model of the present work is applicable to Cn Cn  2C n (5)
DME H2 O MeOH
predicting the activity of the fresh catalyst.
Gas sample analysis was performed by means of Agilent 6890
Gas Chromatograph equipped with ame ionization and thermal K 0 expDG0r =R T 0 (6)
M. Ghavipour, R.M. Behbahani / Journal of Industrial and Engineering Chemistry 20 (2014) 19421951 1945

Table 3
Thermodynamic properties of pure substances.

Substance Heat capacity (J/kmol K) Enthalpy of formation (J/kmol) Gibbs energy of formation (J/kmol)

C1 C2 C3 C4 C5

MeOH 1.0580e5 3.6223e2 9.3790e1 0 0 20.0940e7 16.2320e7


DME 1.1010e5 1.5747e2 5.1853e1 0 0 18.4100e7 11.2800e7
H2O 2.7637e5 2.0901e3 8.125 1.4116e2 9.3701e6 24.1814e7 22.8590e7

After integration we will have:


@T s
 rB C ps 1  e hi rB DHr r i  h f av gTss  T (13)
K 1 T rC 2 rC 3 2 @t
ln rC 1 ln T  T 0 T  T02
K0 R T0 2 6 For adiabatic reactor modeling, last term of the gas phase
rC 4 3 rC 5 4 energy balance should be omitted.
T  T03 T  T04
12 20 Boundary conditions: at t > 0 and z 0; C C 0 & T T 0
 
rC 2 2 rC 3 3 rC 4 4 rC 5 5 Initial conditions: at t 0 and all z; C C i & T T i
DH0r rC 1 T 0 T0 T0 T0 T0
2 3 4 5 Previous experiments showed that within the particle size of

0.17 mm, the intra-particle resistances are negligible [15,29]. In
 1=T  1=T 0
our experiments, powder catalyst has been used and when the
catalyst particle size decreases, effectiveness factor tends to 100%
Finally, the following expression is obtained: and can be neglected [30]. Effectiveness factor is dened as bellow:
  R
1 T 1
r M dV Actual reaction rate
K eq 863:2 exp 174870 ln  761:555T  298 hV (14)
8314 298 r M jS Rate predicted from intrinsic kinetics
1:127955T 2  2982  0:00117633T 3  2983 4:68550 Overall mass and heat transfer coefcients between solid and
 gas phases were estimated from the proposed correlations by
 107 T 4  2984 44789470:261=T  1=298 Cussler and Smith respectively [31,32].
Overall mass transfer coefcient:
To nd the best correlation of equilibrium constant, the
correlations mentioned above have been used to calculate kgi 103 1:17Re0:42 Sci 0:67 ug (15)
methanol conversion at different temperatures as bellow: Overall heat transfer coefcient:
h i2    0:407
X
C DME  C H2 O
eq 0
2 CMeOH
2
Xeq hf C P m 2=3 0:458 rudP
K eq (9) (16)
2
CMeOH 0 2
41  X eq 2
C P rm K eB m
1  X eq CMeOH 
The pressure drop in the bed is calculated by Ergun equation [33]:
0
Xeq is the equilibrium conversion of methanol and CMeOH is the
initial concentration of methanol. dP 1  e3 mus 1  erus 2
 150 1:75 (17)
l d2p e3 d p e3
4. Fixed-bed reactor modeling for methanol to DME reaction
The overall heat capacity:
The mathematical model was developed based on the following
C Pi c1 c2  T c3  T 2 c4  T 3 c5  T 4 & CP
assumptions: X
yi  C Pi (18)
 Negligible concentration and temperature variations in radial
direction (plug ow). The differential equations were converted to algebraic equa-
 Heat losses to the surrounding are neglected (for adiabatic tions by means of Finite Difference method. The reactor length was
reactors). divided into equal discrete intervals, and the algebraic equations
 There are interfacial gradients of temperature and concentration were solved by 4th order RungeKutta method for every segment
between solid and gas phases (heterogeneous model). at each time step [34]. Calculations were repeated at the next time
 Pore diffusion resistance is negligible due to the catalyst powder step with the initial values which were the results of the previous
shape (h = 1). time step and the iteration continued until no changes observed by
time (i.e. steady state condition). Because, the reactor length was
The mass and energy balances for the gas and solid phases in divided into separate segments that conversion through each one
non-adiabatic reactors are expressed by the following equations. was small enough, differential reactor assumption could be applied
Gas phase: for each segment. For each run in a differential reactor, the plug
ow performance equation is used as the following form:
@C i @C
e us i  kgi av C i  Ciss (10) X AZ out X AZ out
@t @z W dX A 1 X A out  X A in
dX A (19)
F A0 r A r A ave r A ave
@T @T U X A in XA in
rg C pg e us rg C p h f av Tss  T  4 T  T r (11)
@t @z dt

Solid phase: 5. Results and discussion

@C is As shown in Fig. 1, the methanol conversion increases at higher


1  e hi rB r i  kgi av C i  Ciss (12)
@t temperatures and lower WHSVs. Higher temperature speeds up
1946 M. Ghavipour, R.M. Behbahani / Journal of Industrial and Engineering Chemistry 20 (2014) 19421951

1 0.9

0.9 experiment modeling


0.8
0.8
0.7
0.7
0.6
0.6 Exp. Data WHSV=90
0.5
X

0.5 Exp. Data WHSV=15

X
0.4 Exp. Data WHSV=7.5 0.4

0.3 Equi. Conv.(This work) 0.3


Equi. Conv. (Hayashi & Moat)
0.2 0.2
Equi. Conv. (Diep & Walnwright)
0.1
0.1
Equi. Conv. (Bercic & Levec)
0
498 523 548 573 598 623 648 0
T(K) 498 523 548 573 598 623 648
T(K)
Fig. 1. Methanol conversion (experimental data) and methanol equilibrium
conversion versus temperature. Fig. 2. Methanol conversion versus reactor temperature. Comparison between
experimental data and modeling results based on Bercic rate model (WHSV = 15 g/
h gcat).
the reaction and lower WHSV means more retention time to let the
reaction to proceed. For exothermic reactions, temperature rising
has a negative effect too. It shifts the reaction to the backward
direction and methanol equilibrium conversion decreases. This were calculated. Every slope equals reaction rate at that special
effect defeats the positive effect (i.e. speeding up the forward point (i.e. at that conversion). Then, the rates of the reaction at
reaction rate) near equilibrium points. Conversion drops at 598 different temperatures and conversions were measured. (See
623 K in lower WHSVs (i.e. WHSV = 15 or 7.5) conrm this fact. Table 4).
DME selectivity was almost 100% at the temperature range of 523 Non-linear regression toolbox of PASW Statistics 18 was
623 K. The optimum reaction temperature is a function of WHSV employed to try the most common rate expressions suggested
and as the WHSV decreases, the best operating temperature earlier. Among all rate expressions that are listed at Table 1,
decreases too. Roughly, 570620 K can be known as the optimum equations number IV, V and VII were better, but simple reversible
operating temperature range. rst order equation (i.e. number VII) had the lowest deviation from
The equilibrium conversions which are calculated from the the experimental data. But, why a rst order rate equation was
correlations presented earlier, and our experimental data have examined? There are numerous mechanisms and reaction rate
been depicted together to nd the best equilibrium constant models which are proposed for this reaction and every one claims
model. From Fig. 1, it is clear that our correlation and the to t the experimental data well. If a number of alternative
correlation proposed via Diep and Walnwright match the mechanisms t the data equally well, it is recognized that the
experimental data better than other correlations. selected equation can only be considered to be one of the good ts,
The reactor modeling program was conducted at the experi- not one that represents reality. With this admitted, there is no
mental conditions as the same as experiments (i.e. the same reason why we should not use the simplest and easiest-to-handle
temperature, WHSV and feed composition) while Bercic and Levec equation of satisfactory t. Also it is mentioned in literature that
rate model was used. In Fig. 2, both of the experimental and most catalytic conversion data can be tted adequately by
modeling results have been depicted. From this Figure, it can be relatively simple rst- or nth-order rate expressions [35].
concluded that Bercic and Levec model is not valid for tempera- Therefore, it seems possible to replace the multi constant rate
tures below 593 K, because the parameters in this model were equations by an equivalent rst-order expression. Although this
evaluated from the experimental data at three temperatures of model does not follow any mechanism and is empirical, but it
593, 613 and 633 K with an impure methanol as the feed.
Therefore, another rate equation should be developed, especially
Table 4
for lower temperatures. At rst, reaction rate as a function of Calculated rate values at different conditions based on experimental data.
temperature and methanol conversion should be calculated.
Integral reactor method was used for this goal. Differential reactor Temperature Methanol Methanol Liquid Catalyst Rate
(K) concentration conversion methanol weight (g) (mol/gcat h)
assumption is used when the rate is considered to be constant at all
(mol/L) (%) ow rate
points through the reactor. Since the rate is concentration- (mL/min)
dependent, this assumption is reasonable only for small conver-
523 0.0709 0.0000 2.00 0.00 0.1460
sions or small reactors. But, for pure methanol as used in this study, 523 0.0653 0.0793 2.00 1.05 0.1361
conversion was so large that differential reactor assumption was 523 0.0549 0.2264 2.00 6.32 0.0862
not acceptable. Therefore, an integral reactor was assumed and the 523 0.0486 0.3160 2.00 12.6 0.0268
differential analysis was used to nd the rate equation: 548 0.0677 0.0000 2.00 0.00 0.2710
548 0.0522 0.2295 2.00 1.05 0.2483
dX 548 0.0361 0.4668 2.00 6.32 0.1343
r A  A (20) 548 0.0250 0.6309 2.00 12.6 0.0016
d FWA0 573 0.0647 0.0000 2.00 0.00 0.3750
573 0.0340 0.4742 2.00 1.05 0.3381
573 0.0223 0.6555 2.00 6.32 0.1528
At rst, methanol conversion versus (W/FA0) was plotted. After
573 0.0117 0.8203 2.00 12.6 0.0679
tting a smooth curve to the available points, slopes at each point
M. Ghavipour, R.M. Behbahani / Journal of Industrial and Engineering Chemistry 20 (2014) 19421951 1947

2 0.9
R-squared = 0.997
y = -8038x + 15.966 0.8
1.5 R = 0.9868
0.7
Ln(K)

Experimental Conversion
0.6

0.5 0.5

0.4
0
0.0017 0.00175 0.0018 0.00185 0.0019 0.00195
0.3
1/T
0.2
Fig. 3. Arrhenius plot for constants obtained by non-linear regression for reversible
rst order equation rate.
0.1
simplies reactor modeling and increases the accuracy of the
results. Fig. 3 indicates that the rate constants obtained by non- 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9
linear regression for equation number VII at different tempera-
Modeling Conversion
tures, satisfy reasonably the Arrhenius theory and its temperature
dependency, with pre exponential coefcient of 8,537,681 L/gcat h Fig. 4. Accuracy assessment of the modeling results with the experimental data.
and activation energy of 66,828 J/mol and therefore, this equation
can be considered as the most favorable rate expression.
Two types of reactors were simulated (i.e. adiabatic and non- zones were at the same temperature). Validity of the simulated
adiabatic reactors). In adiabatic reactors, it was assumed that there non-adiabatic reactor by mathematical modeling using the new
was no heat loss in the radial direction and the reactor wall was proposed rate model, was tested by varying the methanol ow rate
insulated completely. In non-adiabatic reactors, the outside and reactor temperatures (i.e. feed temperature and temperature
surface of the reactor wall was maintained at the particular of the outside wall of the reactor) and the experimental data of the
temperature equaled the feed temperature (i.e. rst and second biggest size reactor (i.e. outside diameter of 3/8 with height of 8

Fig. 5. Reactor modeling results. Temperature prole along the non-adiabatic reactor at different feed ow rates (mcat = 40 g and Tr = Tin = 523 K). (a) Q = 19 mL/min;
WHSV = 22.5. (b) Q = 38 mL/min; WHSV = 45. (c) Q = 76 mL/min; WHSV = 90.
1948 M. Ghavipour, R.M. Behbahani / Journal of Industrial and Engineering Chemistry 20 (2014) 19421951

Table 5
Comparing the modeling and experimental results of non-adiabatic reactor at different conditions.

Tin = Tr (K) Liquid methanol Catalyst WHSV (gMeOH/gcat h) Modeling Experimental Relative error (%)
ow rate (mL/min) weight (g) conversion (%) conversion (%)

523 19.0 40.00 22.50 0.2135 0.2211 3.44


523 38.0 40.00 45.00 0.1341 0.1299 3.23
523 76.0 40.00 90.00 0.0701 0.0734 4.49
498 25.4 40.00 30.00 0.0745 0.0769 3.12
523 25.4 40.00 30.00 0.1771 0.1655 7.01
548 25.4 40.00 30.00 0.3212 0.3109 3.31
573 25.5 40.00 30.00 0.5073 0.5298 4.25
598 25.4 40.00 30.00 0.7318 0.7474 2.09
623 25.4 40.00 30.00 0.8144 0.7903 3.05
498 12.7 40.00 15.00 0.1003 0.1103 9.07
523 12.7 40.00 15.00 0.3038 0.2853 6.48
548 12.7 40.00 15.00 0.5439 0.5603 2.93
573 12.7 40.00 15.00 0.8210 0.8067 1.77
598 12.7 40.00 15.00 0.8336 0.8198 1.68

inches) were compared to the simulation results. Table 5 and Fig. 4 rates, the reaction proceeds in forward direction at the reactor
show the results of both modeling and experiments. The average beginning. As methanol converts to DME, at the end of the reactor,
relative error of the modeling predictions was 4 percent. Fig. 4 and methanol concentration decreases and the forward reaction is
Table 5 conrm the accuracy of the proposed model. Also, the retarded, therefore, much more heat is produced at the reactor
temperature prole along the non-adiabatic reactor was investi- inlet than the reactor outlet. When the feed ow rate increases,
gated. Fig. 5 indicates the modeling results and it shows that as the there is not enough retention time for methanol molecules to react
feed ow rate increases, the hot spot (i.e. maximum temperature and to be consumed at the reactor inlet. So, there are still enough
location) moves to the end of the reactor and the conversion methanol molecules to maintain the reaction rate high along the
decreases. When methanol concentration is high, at low feed ow reactor and the hot spot moves to the end of the reactor. In Fig. 6

Fig. 6. Reactor modeling results. Temperature prole along the adiabatic reactor at different feed ow rates (mcat = 40 g and Tin = 523 K). (a) Q = 19 mL/min; WHSV = 22.5. (b)
Q = 38 mL/min; WHSV = 45. (c) Q = 76 mL/min; WHSV = 90.
M. Ghavipour, R.M. Behbahani / Journal of Industrial and Engineering Chemistry 20 (2014) 19421951 1949

a 527 the temperature prole of the adiabatic reactor is presented. As


526.5
shown in Figs. 5 and 6, in non-adiabatic reactors, due to high rate of
heat transfer from the reactor wall, after that the reaction reaches
526
to its maximum rate and the hot spot forms, the reaction rate
Temperature(K)

525.5 decreases and less heat is produced and the temperature drops
525
down. But, in adiabatic reactors, there is no heat loss from the wall
and temperature rises along the reactor, then the conversion is
524.5 more than that of non-adiabatic reactors at the same initial
Q=19 ml/min
524 temperature. The problem will rise at industrial reactors where it is
Q=38 ml/min attempted to maximize the feed ow and conversion. In this
523.5
Q=76 ml/min situation, adiabatic reactors are more sensitive to control [36] and
523 the reactor may run away due to accumulation of the reaction heat.
0 5 10 15 20
Another point is that it takes more time for adiabatic reactors to
Reactor length(number of segments)
reach to the steady sate conditions in comparison with non-
0.25
adiabatic reactors, because in non-adiabatic reactors, temperature
b Q=19 ml/min
of the catalyst bed rises until the temperature difference can cause
Q=38 ml/min enough heat transfer from the reactor wall and also the convective
0.2
Q=76 ml/min heat transfer between catalyst bed and gas bulk ow that
Methanol Conversion

compensate the heat which is produced from the reaction and


0.15
no more temperature rising occurs by time. In spite of that, in
adiabatic reactors, the only way to release the heat of the reaction
0.1
is the convective heat transfer from catalyst bed to the gas bulk
ow; hence temperature of the catalyst bed rises much more until
0.05 the thermal equilibrium takes place, so it takes more time to reach
the steady state condition. Figs. 7 and 8 indicate the methanol
0 conversion and the temperature prole for both kinds of reactors.
0 5 10 15 20 These gures are the steady state equivalents of Figs. 5 and 6.
Reactor length(number of segments)

Fig. 7. Reactor modeling results. (a) Temperature prole along the non-adiabatic
reactor. (b) Conversion changes along the non-adiabatic reactor (Q = 19, 38, 76 mL/
min (mcat = 40 g and Tr = Tin = 523 K)). a
700

a 553
650
Temperature(K)

Q=19 ml/min Tin=498K


548
Q=38 ml/min Tin=523K
600 Tin=548K
Temperature(K)

543 Q=76 ml/min


Tin=573K
538 Tin=598K
550
Tin=623K
533

500
528
0 5 10 15 20
Reactor length(number of segments)
523
0 5 10 15 20
Reactor length(number of segments)
b 0.9

b 0.35 0.8
Q=19 ml/min
0.7
0.3
Q=38 ml/min
Methanol conversion

0.6
Q=76 ml/min Tin=498K
Methanol conversion

0.25
0.5 Tin=523K
0.2
0.4 Tin=548K
0.15 Tin=573K
0.3
Tin=598K
0.1
0.2 Tin=623K
0.05 0.1

0 0
0 5 10 15 20 0 5 10 15 20
Reactor length(number of segments) Reactor length(number of segments)

Fig. 8. Reactor modeling results. (a) Temperature prole along the adiabatic reactor. Fig. 9. Reactor modeling results. (a) Temperature prole along the non-adiabatic
(b) Conversion changes along the adiabatic reactor (Q = 19, 38, 76 mL/min reactor. (b) Conversion changes along the non-adiabatic reactor (Q = 25.4 mL/min,
(mcat = 40 g and Tin = 523 K)). mcat = 40 g, WHSV = 30, Tr = Tin).
1950 M. Ghavipour, R.M. Behbahani / Journal of Industrial and Engineering Chemistry 20 (2014) 19421951

a 700 experimental data. The constants obtained by non-linear regres-


680 sion for this simple reversible rst order equation agreed
reasonably well with Arrhenius equation. Also, one dimensional
660
unsteady state heterogeneous model was applied to predict the
640 temperature prole and methanol conversion along the xed-bed
Temperature(K)

620 Tin=498K reactor. Adiabatic and non-adiabatic xed-bed reactors were


600 Tin=523K simulated and methanol conversion and thermal behavior of these
Tin=548K
reactors were discussed and the following results were concluded:
580
Tin=573K
560  The correlations for equilibrium constant of the present work
Tin=598K
540 and that of Diep and Walnwright match the experimental data
520
better than other correlations.
 In non-adiabatic reactors, as the feed ow rate increases, the hot
500
spot (i.e. maximum temperature) moves to the end of the reactor
0 5 10 15 20
Reactor length(number of segments)
and conversion decreases.
 In non-adiabatic reactors, due to high rate of heat transfer from
the reactor wall, after that the reaction reaches to its maximum
b 0.9 rate and the hot spot forms, when the reaction rate decreases and
less heat is produced, the temperature will drop. But in adiabatic
0.8
reactors, there is no heat loss from the wall and temperature will
0.7 rise continuously along the reactor and the conversion is more
than the conversion of non-adiabatic reactors at the same initial
Methanol conversion

0.6
temperature.
Tin=498K
0.5  It takes more time for adiabatic reactors to reach to the steady
Tin=523K
sate conditions in comparison with non-adiabatic reactors.
0.4 Tin=548K  Conversion and temperature prole of the non-adiabatic reactor
0.3 Tin=573K with different temperatures of feed and heating medium at the
Tin=598K range of 498623 K were calculated and the minimum required
0.2
temperature of heating medium and feed temperature to reach
0.1 the equilibrium conversion at WHSV of 30 and 15, were
estimated.
0  Another point is that in non-adiabatic reactors at a constant feed
0 5 10 15 20
ow, as the heating medium temperature increases, the hot spot
Reactor length(number of segments)
location shifts to the reactor inlet.
Fig. 10. Reactor modeling results. (a) Temperature prole along the non-adiabatic
reactor. (b) Conversion changes along the non-adiabatic reactor (Q = 12.7 mL/min,
mcat = 40 g, WHSV = 15, Tr = Tin). Acknowledgements

The authors acknowledge the Iranian Petrochemical Research


and Technology Company for their technical and nancial supports
Finally, conversion and temperature proles of the non-
(Project ID: 0860249001).
adiabatic reactor with different reactor temperatures, in two
WHSVs are depicted in Figs. 9 and 10. The results of these
modelings have been compared to the experimental data at Table References
6. It is concluded from Figs. 9 and 10 that in our experimental
[1] M.Ch. Lee, S.B. Seo, J.H. Chung, Y.J. Joo, D.H. Ahn, Fuel 87 (2008) 21622167.
reactor with 40 grams of catalyst, at WHSV of 30, the minimum
[2] R.J. Crookes, K.D.H. Bob-Manuel, Energy Convers. Manage. 48 (2007) 29712977.
required temperature of reactor outside wall and feed to reach the [3] T.A. Semelsberger, R.L. Borup, H.L. Greene, J. Power Sources 156 (2005) 497511.
equilibrium conversion is 623 K and at WHSV of 15 this value is [4] T.A. Semelsberger, K.C. Ott, R.L. Borup, H.L. Greene, Appl. Catal., B 65 (2006) 291
573 K. Another point is that in non-adiabatic reactors at a constant 300.
[5] J.H. Yoo, H.G. Choi, Ch.H. Chung, S.M. Cho, J. Power Sources 163 (2006) 103106.
WHSV, as the temperature of the reactor surrounding increases, [6] M. Marchionna, R. Patrini, D. Sanlippo, G. Migliavacca, Fuel Process. Technol. 89
the hot spot location shifts to the reactor inlet and this (2008) 12551261.
phenomenon is reasonable because as the medium temperature [7] J.H. Kima, M.J. Park, S.J. Kim, O.Sh. Joo, K.D. Jung, Appl. Catal., A 264 (2004) 3741.
[8] F. Raoof, M. Taghizadeh, A. Eliassi, F. Yaripour, Fuel 87 (2008) 29672971.
increases, the reaction speeds up and most of methanol molecules [9] D. Liu, Ch. Yao, J. Zhang, D. Fang, D. Chen, Fuel 90 (2011) 17381742.
will react at the reactor inlet and a bulk of heat will be released as [10] M. Xu, J.H. Lunsford, D.W. Goodman, A. Bhattacharyya, Appl. Catal., A 149 (1997)
the reaction heat. 289301.
[11] F. Yaripour, F. Baghaei, I. Schmidt, J. Perregaard, Catal. Commun. 6 (2005)
147152.
6. Conclusion [12] Y.J. Lee, J.M. Kima, J.W. Bae, Ch.H. Shin, K.W. Jun, Fuel 88 (2009) 19151921.
[13] V. Vishwanathan, K.W. Jun, J.W. Kim, H.S. Roh, Appl. Catal., A 276 (2004) 251255.
[14] S. Hassanpour, F. Yaripour, M. Taghizadeh, Fuel Process. Technol. 91 (2010) 1212
The aim of this work was to examine one of the most applicable
1221.
rate equations (i.e. Bercic and Levec rate model) by comparing [15] G. Bercic, J. Levec, Ind. Eng. Chem. Res. 31 (1992) 10351040.
experimental data with the simulation results. Needless to say that [16] Z.G. Nie, H.W. Liu, D.H. Liu, W.Y. Ying, D.Y. Fang, Chem. Reac. Eng. Tech. 20 (2004)
17.
it was attempted to make the experimental conditions and the
[17] A.T. Aguayo, J. Eren, D. Mier, J.M. Arandes, M. Olazar, J. Bilbao, Ind. Eng. Chem. Res.
commercial ones close together. Due to the deviation of the 46 (2007) 55225530.
relevant model from the experimental data especially at tem- [18] K.L. Ng, D. Chadwick, B.A. Toseland, Chem. Eng. Sci. 54 (1999) 35873592.
peratures below 593 K, non-linear regression was employed to try [19] X.D. Peng, B.A. Toseland, P.J.A. Tijim, Chem. Eng. Sci. 54 (1999) 27872792.
[20] G.R. Moradi, J. Ahmadpour, F. Yaripour, Chem. Eng. J. 144 (2008) 8895.
the most common rate expressions suggested earlier. Simple [21] G.H. Graaf, E.J. Stamhuis, A.A.C.M. Beenackers, Chem. Eng. Sci. 43 (1988) 3185
reversible rst order equation had the lowest deviation from the 3195.
M. Ghavipour, R.M. Behbahani / Journal of Industrial and Engineering Chemistry 20 (2014) 19421951 1951

[22] S.M. Nasehi, R. Eslamlueyan, A. Jahanmiri, Proc. 11th Chem. Eng. Conf. Iran, Kish [33] J.F. Richardson, J.H. Harker, J.R. Backhurst, Chemical Engineering, fth ed.,
Island., 2006. ButterworthHeinemann, 1999.
[23] M. Farsi, R. Eslamloueyan, A. Jahanmiri, Chem. Eng. Process. 50 (2011) 8594. [34] J.D. Hoffman, Numerical Methods for Engineers and Scientists, McGraw-Hill,
[24] M. Fazlollahnejad, M. Taghizadeh, A. Eliassi, G. Bakeri, Chin. J. Chem. Eng. 17 New York, 2001 .
(2009) 630634. [35] C.D. Prater, R.M. Lago, Advances in Catalysis, Academic Press, New York, 1956.
[25] G. Bercic, J. Levec, Ind. Eng. Chem. Res. 32 (1993) 24782484. [36] F.G. Froment, K.B. Bischoff, Chemical Reactor Analysis and Design, John Wiley &
[26] M. Farsi, A. Jahanmiri, R. Eslamloueyan, Int. J. Chem. React. Eng 8 (2010), Article A79. Sons, New York, 1979.
[27] B.T. Diep, M.S. Walnwright, J. Chem. Eng. Data 32 (1987) 330333. [37] B.C. Gates, L.N. Johanson, LungmuirHinshelwood kinetics of the dehydration of
[28] R.H. Perry, D.W. Green, J.O. Maloney, Perrys Chemical Engineers Handbook, 70th methanol catalyzed by cation exchange resin, AICHE J. 17 (1971) 981983.
ed., McGraw-Hill, New York, 1997. [38] F. Figueras, A. Nohl, L. Mourgues, Y. Trambouze, Dehydration of methanol and
[29] S.J. Royaee, C. Falamaki, M. Sohrabi, S.S. Ashraf Talesh, Appl. Catal., A 338 (2008) tert-butyl alcohol on silicaalumina, Trans. Faraday SOC. 67 (1971) 11551163.
114120. [39] D. Kallo, H. Knozinger, Zur Dehydratisierung von Alkoholen an Aliminiumoksid,
[30] E.B. Nauman, Chemical Reactor Design, Optimization and Scale Up, McGraw-Hill, Chem. Ing. Tech. 39 (1967) 676680.
New York, 2002. [40] F.C. Rubio, S.D. Diaz, D.D. Castillo, J.D. Trujillo, R.A. Alvarez, Deshidratacion
[31] E.L. Cussler, Diffusion Mass Transfer in Fluid Systems, Cambridge University Catalitica de Metanol en Fase vapor, Ing. Quim. (Madrid) 12 (1980) 113119.
Press, United Kingdom, 1984. [41] G. Schmitz, Deshydration du Methanol Sur Silice-Alumine, Chim. Phys. (1978)
[32] J.M. Smith, Chemical Engineering Kinetics, McGraw Hill, New York, 1980. 746504355.

You might also like