8/9/2019 El Kanzi 1996
1/9
ELSEVIER
Chemical Engineering and Processing 35 (1996) 131-139
Optimization of MTBE synthesis in a fixed-bed reactor system
E.M. Elkanzi
Department of Chemical and Petroleum Engineering, Faculty o f Engineering, UAE University, P.O. Box 17555, Al-Ah, United Arab Emirates
Received 12 April 1995; accepted 23 August 1995
Abstract
A fundamental fixed-bed catalytic reactor model has beendeveloped or use n selectingalternative operating strategies n a
commercialmethyl-t-butyl ether (MTBE) unit. The model is basedon generalchemicalengineering rinciplesand is tuned to
represent he operation of the reactor’s systemof a given MTBE unit. Constrainedoptimization techniques re used o determine
the optimum operating conditions of the reactor’s system hat give the maximum net profit. The model will enable he user to
predict the required bed temperature rises, the required recycle rate for specificsingle-pass onversion and the required heat
removal rate in the coolers of an existing unit. For optimization purposes, he model is used o investigate he effect on unit
profitability when variablessuchas he conversionper passor the total conversionweremodified. A computer program,MTBEC,
wasdeveloped or solving the model equations.Conversionper passwasvaried from 62.3% o 87.3% n increments f 5%. Based
on thesecases, onversionper passhigher than the original design alue of 72.3% s suggested. he optimum conversionper pass
value appears o be about 75.5% and represents n improvement of about 1.5%on daily net profit for the pricesquoted. These
resultsdemonstrate he useof the model n selectingmore economicallyattractive operating targets.The modelcan alsobe used
to investigate he effect of other variablesamenable o optimization, e.g. fresh feed stock quality, feed costsand bed temperature
profiles.
Keywords: MTBE synthesis;Fixed-bed reactor system; Catalytic reactor model; Constrained optimization techniques;Model
equations
1. Introduction
The use of MTBE as a gasoline octane blending
component is growing rapidly due to the phase out of
tetraethyl lead as an octane boaster. According to most
estimates, demand for MTBE could reach 24 million
tonne y-i in 1995 [l]. This follows the second phase of
reformulated gasoline in 1997 in the USA, plus the
potential later in Europe and elsewhere. Moreover,
there are some viable alternatives to MTBE, including
ethanol, ETBE, TAME and DIPE, and new technolo-
gies such as reactive distillation using a solid catalyst
where both chemical reaction and fractionation of
products can proceed simultaneously are already run-
ning. New MTBE plants wil l come on stream and the
existing plants will have to adjust themselves to the
current economic situation, The availability and cost of
raw materials may become a problem and the existing
plants have to improve their eff iciency and increase
their profitability rather than expand the plant. One
0255-2701/96/ 15.001996 ElsevierScience .A. All rights eserved
way of improving productivity per given reactor vol-
ume may be accomplished by optimizing the plant
operating conditions [2]. However, the optimization of
an existing producing unit is a different type of opti-
mization than that made by the original designer of the
unit. The original design is based on a given set of
product prices and capital equipment costs with given
feed and product specifications. All process equipment
are sized and rated to meet the design operation. After
a certain period of operation, the economics may
change significantly from the original design basis.
Certain pieces of equipment, feed and/or product
qualtities may become bottlenecks or constraints to the
most economic operation. In addition, new catalysts are
being developed whose activities and selectivities are
better than that used in the original design [3]. As such,
a process model is needed to predict optimum opera-
tions more economically. In its simplest form this
model may be a heuristic model developed by an oper-
ator through first-hand observation of the unit. How-
8/9/2019 El Kanzi 1996
2/9
132 EM. Elkanzi / Chemical Engineering and Processing 35 (1996) 131-139
ever such models cannot be relied upon for accurate
quantitative predictions and more rigorous models
based on basic chemical engineering principles - ma-
terial and energy balances,equilibrium thermodynamics
and kinetics - are needed to evaluate the subtle eco-
nomic trade-offs involved in process optimization.
The aim of this study is to maximize the profitability
of an existing MTBE reactor’s systemby maximizing
an objective function which includes the value of
MTBE produced, and the operating and feed costs.The
technical basis of a fundamental fixed-bed catalytic
reactor model to search for more economically desir-
able operation of the reactor’s system is described.
Constrained optimization techniques are used to deter-
mine the key variables so as to satisfy the objective
function while maintaining the reactor’s temperatures
within safe limits.
2. Process description
2.1. Process chemistry
Methyl-t-butyl ether, or MTBE, is the equilibrium
product of the reaction of isobutylene (iC,) with
methanol in the presenceof an acid catalyst. The main
reaction is [4].
H
CI-4
H
CH,
I I I
H-C-OH + C=CH, H-C-0-C-CH,
(1)
I I I I
H CH, H CH,
This reaction is exothermic with 37.7 kJ of heat released
for each mole of isobutylene converted, the maximum
conversion being determined by a thermodynamic equi-
librium value [4-61. The reaction is also extremely
selective to isobutylene, i.e. other butenes are not con-
verted [7] other than under conditions of hot spots
when linear butenes can react [8]. Under design condi-
tions and using a sulfonated polystyrene resin (e.g.
polystyrene/divinylbenzene copolymer), essentially the
only hydrocarbon molecule to react is isobutylene. The
remaining hydrocarbons in the feed stream are almost
entirely inert. However, undesirable side-reactionsmay
take place side by side with the main reaction of Eq.
(1). The most important side-reaction is the dimeriza-
tion of isobutylene to a mixture of the isomers 2,4,4-
trimethyl-1-pentene (TMP-1) and 2,4,4-trimethyl-
2-pentene (TMP-2). Interaction of two molecules of
isobutylene produces the dimer [4],
CH, CH3 CH, CH,
I I I I
C=CH2 + C=CH, Z+ C=C-C-CH,
(2)
I
I
I I I
CH, CH,
CH, H CH,
At design conditions the isobutylene selectivity to form
these isomers is less than 0.5% [9]. Higher reaction
temperatures tend to produce more and should be
avoided.
Dimethyl ether (DME) can be formed by the interac-
tion of two methanol molecules in the presence of an
acid catalyst [9].
H H
H H
I
I I I
H-C-OH + H-C-OH +H-C-0-C-H+H1O
I I I I
H H H H
(3)
Reaction (3) is not only harmful by itself but the water
produced may also react with isobutylene to form
t-butyl alcohol (TBA) [9].
CK CH,
H,O + C=CH2 + HO-C-CH3 (4)
I I
CH, CH,
Production of TBA is not totally undesirable because
its blending octane number is lower than that of MTBE
but higher than that of base gasoline. Besides the
formation of TBA, water has another harmful effect in
that it reduces the acidity of the catalyst and thus
lowers its activity and a higher reaction temperature is
therefore required. This effect disappears when water is
converted into TBA and in fact only a few centimeters
at the reactor are affected by water [lo]. At design
conditions, the isobutylene selectivity to form TBA is
less hen 0.9% [ll].
2.2. Reaction kinetics
The direct addition of olefins catalyzed by ion-ex-
change resins to give ethers was investigated by Ancil-
lotti et al. [12]. They confirmed the high reactivity of
isobutylene with methanol compared to other olefins.
Several investigators [4,5,12- 171have studied the kinet-
ics of the reaction of Eq. (1) both in the liquid and
vapor phases, and generally two approaches were fol-
lowed for the analysis of the kinetic data. In the first
approach the reaction is considered as a homogeneous
catalytic reaction where the reactants are confined
within the gel-like resin. In the second approach, the
resin is treated like a solid catalyst and the reaction is
considered as a heterogeneouscatalytic reaction. In one
study [5], a first-order dependence with respect to
isobutene was found together with a -0.25 order de-
pendence with respect to methanol. In another study
[12], a power law rate expression with a zero-order
dependenceof rate on methanol concentrations of more
than 4 mall-’ was reported. In the same study, a
8/9/2019 El Kanzi 1996
3/9
E.M. Elkauzi 1 Chemical Engineering and Processing 3.5 (1996) 131-139
133
first-order dependence of the rate on the isobutylene
concentration with an activation energy of 17 kcal
mol-’ was also reported. In another study 1131, re-
versible rate expressions with a second-order depen-
dence of the forward reaction and a first-order
dependence of the reversible step have been sug-
gested.
In the heterogeneous catalytic reaction approach,
the kinetic data analysis was based on classical mod-
els such as the Langmuir-Hinshelwood-Hougen-
Watson (LHHW) models [4,5,13-171. A typical
LHHW model of the reaction is of the form:
Pa)
Under industrial conditions the forward reaction goes
almost to completion (96% conversion of iso-
butylene). Moreover, the reported [4] ratio of k,/kl at
60 “C is about 0.03 and the corresponding activation
energies ratio is about 1.6. Thus the reversible step
may be neglected and Eq. (4a) reduces to a first-or-
der dependence on the isobutylene concentration. In
fact, the analysis of the kinetic data of Subramanian
and Bhatia [13] based on first-order kinetics gives a
better fit of the data than the assumed second-order
forward step and the reversible first-order one. Based
on the above argument and on the results of Ancil-
lotti et al. [12] and on the analysis of industrial data,
a heterogeneous rate expression for the first-order de-
pendence on isobutylene concentration with an acti-
vation energy of 17.4 kcal mol-’ will be used in this
study.
2.3. The seactor system
In the idealized process involving the flow diagram
shown schematically in Fig. 1, the reactor system in-
cludes the catalytic reactors, coolers and static mix-
ers. The fresh feed enters a static mixer where
methanol and isobutylene liquids are blended. The
combined fresh feed (FF) then mixes with the first
reactors effluent recycle (R*) and passes through a
Fig. 1. Schematic flow diagram for the reactor system.
Table 1
Normal reactor system operation
Parameter
LHSV, s-’
Inlet temp., K
AT, K
iC, conversion, % based on
feed to reactor
Reactor I
2.94 x 1O-3
321
14.2(q)
72.3(X,)
Reactor II
7.5 x 10-4
314.7
3.2(%)
55.5(X,)
iC, conversion, % based on
fresh feed
MTBE selectiv ity, %
Recycle ratio
Pressure, bar
Catalyst volume, m3
QR> W
QE>
W
Products:
MTBE, kg/s
TMP, kg/s
MeOH, kg/s
ic,, kg/s
91(X,) 96(X,)
87.3
91.4
2.88 -
25 25
29.3
29.3
1.3x103 -
8.5 x 10’
4.16
4.9
0.0072 0.01
0.25 0.83
0.64 0.2
second static mixer where it is thoroughly blended to
ensure a homogeneous mixture - in the stoichiomet-
ric proportions of both reactants - before entering
the firs t reactor. In industrial practice, a 10% excess
to methanol is used in order to reduce mainly
isobutylene dimers by product fermentation. The re-
actors are operated near adiabatic conditions and the
reaction produces a significant amount of heat, caus-
ing a temperature rise across the catalytic bed. This
temperature is controlled by circulating a portion of
the first reactor’s effluent (E*) to the reactor inlet.
The circulating pumps take suction from the first re-
actor’s outlet and develop enough discharge head to
circulate the effluent through the recycle cooler and
combine it with the fresh feed. The remaining effluent
(EE) is cooled before entering the second reactor
where the remaining unconverted isobutylene reacts
to form MTBE, as in the firs t reactor, so that 96%
of the isobutylene contained in the fresh feed to the
reactor’s system is converted.
Table 2
Price assumptions
Feeds:
iC,(mixed butenes)
MeOH
Products:
MTBE
Utilities
Water cooling (11.1 K rise), m3
Electricity, kWh
Power, pumping
17.5 per BBL
23.9 per BBL
45.2 per BBL
14 per kg MTBE
2.2 per kg MTBE
SO.07 per BHP h-’
8/9/2019 El Kanzi 1996
4/9
134
EM. Elkanzi / Chemical Engineering and Processing 35 (1996) 131-139
Table 3
Base case economics
Product Quantity Unit value S per day
MTBE
3780.7 BBL d-’ S45.2 per BBL 170 888
Feeds:
iC4 2993.4 BBL d-’ 17.5 per BBL 52 385
MeOH 1290 BBL d-’ 23.9 per BBL 30 828
Operating costs:
Pumping
2532
Cooling 7597
Other? 34 445
Total cost
127 787
Net profit per day
43 100
Net profit per BBL 11.4
MTBE
a Others include: operating costs other than reactor sys -
tem + freight to USA + sales/delivery + fixed costs + import duty +
depreciation.
3. Optimization model
3.1. Reactor equations
Commercial MTBE synthesis reactors are operated
near adiabatic conditions and are approximately plug
flow in nature, resulting in temperature and concentra-
tion gradients through the catalytic bed. Since tempera-
ture and concentration are the key driving forces in the
reaction rate expressions, a model which represents the
reactor by individual catalyst bed must integrate the
reaction rate and its related heat balance along the
catalyst bed. Only a model with such detail will enable
the user to predict the required bed inlet temperature,
the bed temperature rises, the required recycle rate for
a specific single pass and overall conversions and the
required heat removal rate in the coolers.
The detailed reactor model consists of differential
material and energy balances over adiabatic fixed-bed
reactors. These balances are coupled with physical
property and heat-transfer correlations. For the opti-
mal operation of the reactor’s system, optimization
variables such as the recycle ratio and the reactor’s inlet
temperatures are related to the objective function. The
details of the balances around the firs t reactor with
reference to the notation in Fig. 1 and to the nomencla-
ture are:
s
s
t; = Fi
dXi
0 Cmyi)
(5)
where
.
(-r,)=k,k,exp FF *Cy
01
Xl
I
‘=l+RR(l-XI)
I
>
(6)
Heat balance yields (datum = 298 K):
CpidT- t ET
s
T
C,i dT- FiA.H~X’= 0
i=l 298
(7)
and
QR= i RT
“CpidT
i-l
s
T
For the second reactor:
s
2
VI, = EEi
dXi
0 CBri>
and
x
2
= 0.96 -X,
1 -x,
Heat balance yields:
n
OTT- , I
(8)
(9)
(10)
C,,dT- 2 Pi
i=l
s
T T
X
Cpi dT- EEiAHgX2/= 0
(11)
298
and
QE = i EEi
‘TO Cpi dT
i=l
s
T
w
The variables to be determined from these balances
are the recycle ratio, RR, second reactor inlet tempera-
ture TT,, and heat removal rates, QR and QE, for given
values of X,, X0 and To.
3.2. Objective function and constraints
Optimization of the MTBE reactor’s system opera-
tion implies the maximization of unit profit. Within the
optimization model this is stated as an objective func-
tion defined by [2].
Profit = (product value) - (feed costs) -(operating costs)
(13)
The model is to be used to investigate the effect on
unit profitability when variables such as conversion per
pass, total conversion or feed quality are modified. This
leads to optimization of heat removal from the coolers
and recycle rates. These, in turn, determine the external
heat flux and recycle ratio which maximize the net
profit while maintaining the reactor’s temperature
within safe limits. Assuming that the product value,
feed costs and those components of the operating cost
which are not related to the reactor sysem are con-
stants, the net profit will then only be a function of the
reactor’s system operating costs. These operating costs
are effectively determined by the recycle pumping cost
8/9/2019 El Kanzi 1996
5/9
E.M. Elkanzi / Chemical Engineering and Processin g 3.5 (1996) 131-139
135
0.2 -
/
O.l- d
I I I
I
I I I
2 2.2
2.4 2.6 2.8 3.0 3.2 3.4 3.6
RECYCLE RATIO
Fig. 2. Variation of conversions with recycle ratio.
and external heat removal cost. Both recycle and heat
removal costs are functions of conversion. These rates
are used, therefore, to represent the reactor system
operating costs in the objective function. Thus, there is
an optimal reactor’s system operating conditions for
which the optimal operating cost gives a globally maxi-
mum profit. The operating cost itself is thus an opti-
mization variable. In general, the optimization of the
reactor’s system operating conditions is therefore a
continuous variable optimization within a single-
parameter (operating cost) optimization.
The objective function of Eq. (13) is subject to the
following implicit constraints:
ATI < E,
(14)
ATzG%
(15)
where E, and % are limits based on the consideration of
the reaction kinetics and MTBE selectivity (typical safe
values of these limits are shown in Table 1). It is further
assumed that the reactor system is operating against no
mechanical constraints, i.e. there is additional capacity
on the recycle and cooling water pumps.
3.3. Solution methodology
Each set of equations [(5)-(7) and (9)-( 1 )] must be
solved simultaneously in order to determine the re-
quired recycle ratio and cooler heat loads for various
values of a single-pass conversion. These equations
were written in forms that were appropriate for solu-
tion using a finite-difference approach. It was found
convenient to consider the reactor as consisting of a
series of elements of volume and employing these equa-
tions by incrementing the known volume of the reactor.
The difference equations were solved with iterative con-
vergence on the conversion at each increment. Integra-
tion is continued until the reactor volume slightly
exceeds the design value and the conversion simulta-
neously matches with the current value.
The objective function [Eq. (13)] and constraints
[Eqs. (14) and (15)] were evaluated between each evalu-
ation of the recycle ratio and cooler heat loads in order
to determine a new search direction. The optimization
search was continued until a maximum profit (or mini-
mum operating cost) was determined. A computer pro-
gram, MTBEC, was developed for solving the model
equations. The algorithm for solving MTBEC was cou-
pled with the method for finding a maximum profit.
The model has two main modes of operation; it updates
the initial design operating conditions and predicts new
operating conditions imposed by the current economic
situation. In the update mode, commercial unit test run
data are input. These data include observed catalyst
bed temperature, liquid feed and recycle rates, cooler
heat loads, liquid feed, and product qualities and yields.
The model calculates both the conversion per pass and
total conversion, and updates all other design parame-
ters. In the predict mode the updated parameters are
utilized to solve cases after the user defines the problem.
Typically the user specifies the feed rates and qualities,
total and per pass conversion levels, current feed costs
and product value. The model firs t iterates to find the
required bed temperature rises, the recycle rates, the
cooler heat loads and the product yields. Additionally,
the predict mode calculations are performed until a
maximum profit is achieved.
8/9/2019 El Kanzi 1996
6/9
136
EM. Elkanzi / Chemical Engineering and Processing 35 (1996) 131-139
I I
I I
I
I
I
I
I
)
2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 d
RECYCLE RATIO
Fig. 3. Variationof heat emoval ateswith recycle atio.
0
3.4. Model use example
To provide a specific example of the model’s use in
optimizing an actual unit, an MTBE reactor system was
modelled and a predict case was made to investigate
changes in some independence process parameters. The
schematic of the reactor system was as shown in Fig. 1.
In updating, the model was tuned using a test run
representing normal (or base case) operation of the
reactor system. This operation is shown in Table 1. The
unit produces 444 tonne d-l MTBE and the prices to
be used in selecting more optimum operations are as
shown in Table 2 [ 1, 18,191. The profit for the base case
(assuming a Middle East location for the unit) is shown
in Table 3.
The conversion per pass was varied from 62.3% to
87.3% and Fig. 2 shows the variation of the conversions
(X,, X, and X,) with recycle ratio. The model was run
to investigate the effect on unit profitability when this
variable was modified. Fig. 3 shows the variation of the
heat removal rates with recycle ratio. As part of the
reactor’s operating cost, it can be seen that the mini-
mum heat removal cost would occur at a recycle ratio
of about 2.33. However, at this value of the recycle
ratio the constraint of Eq. (14) is not satisfied and the
calculated temperature rise of 16.7 “C exceeds the safe
value. When coupled with the cost of the pumping
recycle rate, the variation of the reactor’s operating cost
with recycle ratio is as displayed in Fig, 4. From this
figure it can be seen that the minimum cost occurs at a
recycle ratio o f about 2.7, corresponding to a conver-
sion per pass of 75.4% (see Fig. 2). At this value the
constraints of Eqs. (14) and (15) are both satisfied as
displayed in Fig. 5 which shows the temperature profi-
les along the reactor length under optimal conditions.
At these optimal operating conditions the improvement
in daily net profit, for the prices quoted, is about 1.5%
higher than the base case conditions.
8/9/2019 El Kanzi 1996
7/9
8/9/2019 El Kanzi 1996
8/9
138
EM. Elkanzi 1 Chemical Engineering and Processing 35 (1996) 131-139
300
u
0.2 0.4 0.6 0.8
DIMENSIONLESS AXIAL DISTANCE
Fig. 5. Temperature profiles under optimum conditions.
LO
P
reactor II effluent stream, kmol s-l
Q
rate of heat removal, kW
R* recycle rate, kmol s-l
p-“l,
recycle ratio
rate of reaction, kmol s-l mm3
T, +T temperature of first and second reactor, K
V reactor volume, m3
x conversion
& temperature difference limit, K
Subscripts
E effluent
i
component
ii
inlet
recycle
S
per pass
References
[l] Methanol, MTBE suppliers will likely keep up with rising de-
mand, Oil Gas J. , (March 19, 1993) 48.
[2] T.F. Edgar and D.M. Himmelblau, OptimCation of Chemical
Processes, McGraw-Hill, New York , 1989.
[3] M.R. Ladisch, R.L. Hendrickson, M.A. Brewer and P.J. West-
gate, Catalyst-induced yield enhancement in a tubular reactor,
Ind. Eng. Chem. Res., 32 (1993) 1888.
[4] A. Ali and S. Bhatia, MTBE formation in a cataly tic bed
reactor-kinetic and modelling study, Chettr. Etrg. J., 44 (1990)
97.
[5] F. Colombo, I. Carl, I. Dalloro and P. Delgu, Equilibrium
constant fo r the methyl tert-buty l ether liquid phase synthesis by
use of UNIFAC, Ind. Etzg. Chetn. Amdam., 22 (1983) 219.
[6] J.F . Izquierdo, F. Cunill, M. Vila, M. Iborra and J. Tejero,
Equilibrium constants for methyl tert-butyl ether and ethyl
tert-butyl ether liquid phase syntheses using C,olefinic cut, Ind,
Eng. Chetn. Res., 33 (1994) 2830.
[7] S.A. Nijhuis, F.P.J.A. Kerkhof and A.N.S. Mak, Multiple
steady states during reactive distillation of MTBE, Itzd. Eng.
Chetn. Res., 32 (1993) 2767.
[8] M. Vila, F. Cunill, J.F . Izquierdo, J. Gonzalez and A. Hernan-
dez, The role of by-products formation in methyl tert-buty l ether
synthesis catalysed by a macroporous acidic resin, Appl. Cutal,,
117 (1994) L99.
[9] F. Lewis and S. Matar, From Hydrocarbons to Petrochemicals,
Gulf Publishing Co., 1981, p. 128.
8/9/2019 El Kanzi 1996
9/9
EM. Elkanzi / Chemical Engineering and Process ing 35 (1996) 131-139 139
[lo] F. Cunill, M. Vila, J.F. Izquierdo, M. Iborra and J. Tejero,
Eff ect of water presence on methyl tert-butyl ether and ethyltert-
butyl ether liquid-phase syntheses, 6zd. Eng. Chem. Res., 32
(1993) 561.
[I
I] Hydrocarbon Process., 69
(1990) 128.
[12] F. Ancillotti, M.M. Marcelle and E. Pescarollo, J. Catal., 46
(1977) 49.
[13] C. Subramanian and S. Bhatia,
Can. J. Chem. Eng., 65
(1987)
613.
[14] A. Gicquel and B. Torck, J . Catal., 83 (1983) 9.
[15] Kyung-Ho Chang, Geon-Joong Kim and Wha-Seurg Ahn, Ind.
Eng. Chem. Res., 31 (1992)
125.
[16] A. Rehfinger and V. Hoffmann, Kinetics of methyl tertiary butyl
ether liquid phase synthesis catalysed by ion exchange resin. 1.
Intrinsic rate expression in liquid phase activit ies,
Chem, Eng.
Sci., 45
(1990) 1605.
[17] D. Parra, J. Tejero, F. Cunill, M. Iborra and J.F. Izquierdo,
Kinetic study of MTBE liquid-phase synthesis using C,olefinic
cut, Chem. Etzg. Sci., 99 (1994) 4563.
[18] C.H. Vervalin, Economically feasible MTBE plants high-
light period of new growth,
Hydrocarbon Process., 69
(1990)
19.
[19] W. Ludlow, Methyl tertiary butyl ether production grows dra-
matically, Oil Gas J., (June 8, 1987) 54.