Professional Documents
Culture Documents
Destilacion Por Membrana
Destilacion Por Membrana
www.elsevier.com/locate/compchemeng
Abstract
The main objective of this work is the analysis and optimisation of azeotropic distillation columns, when a liquid side stream
with the distributing non-key component is treated in a pervaporation membrane and the retentate is recycled to the column. The
objective is to separate the pure distributing non-key component using a pervaporation membrane, thus helping to improve the
purity in the top and/or bottom products. The operating conditions of the column such as reflux ratio, product and side draw
flowrates and pressure are selected optimally to minimise the operating cost of the hybrid system. The simulation of the
pervaporation membrane units is implemented solving the differential– algebraic equation system of mass transport and energy
balances. The optimisation of the operating conditions of the azeotropic distillation column in the hybrid distillation/pervapora-
tion system for Methyl tert-butyl ether production is presented. Numerical results with a significant reduction in the operating cost
are reported. © 2002 Elsevier Science Ltd. All rights reserved.
Nomenclature
* Corresponding author.
E-mail address: meliceche@plapiqui.edu.ar (A.M. Eliceche).
0098-1354/02/$ - see front matter © 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 9 8 - 1 3 5 4 ( 0 1 ) 0 0 7 7 5 - X
564 A.M. Eliceche et al. / Computers and Chemical Engineering 26 (2002) 563–573
ration with a graph showing how the number of papers performing the simulation of the process at different
on pervaporation increased over the last 20 years. operating conditions. As previously stated, the optimi-
Lipnizki, Field, and Ten (1999b) present an extense sation of the hybrid distillation–pervaporation process
review of pervaporation-based hybrid processes, focus- has not been attempted in former work. The formal
ing on industrial applications and pointing out the need optimisation of the hybrid system that includes an
of optimisation of some of the analysed processes. algebraic and differential equation system is described
One of the fields where there is need of optimisation in this work. Numerical results of the optimisation are
is the distillation/pervaporation hybrid process. Höm- shown for the separation of a mixture of butanes,
merich and Rautenbach (1998) studied the integration methanol and MTBE. This work was motivated by the
of pervaporation and vapour permeation into the Hüls evaluation of the feasibility of revamping the MTBE
process. They analysed the influence of the operating separation sector of a refinery.
conditions in a hybrid distillation–pervaporation– va-
pour permeation system for the MTBE production and 2. Distillation –pervaporation process for MTBE
implemented the simulation of the process in ASPEN. production
González and Ortiz Uribe (2001a,b) carried out experi-
mental work to find a semi-empirical model for the MTBE has been the chemical with the highest in-
pervaporation membrane to separate methanol and crease in its rate of production, a notorious 20%/year
MTBE, and simulated the hybrid distillation/pervapo- during the years 1985–1995 (Trevisan, 1995). MTBE is
ration process using gPROMS (PSEnterprises, 2000) used as a high octane fuel additive. It provides the fuel
performing a cost analysis. with additional oxygen, decreasing the amount of CO
However, the formal optimisation of hybrid distilla- produced as it benefits combustion completeness.
tion/pervaporation process has not been reported previ- The production process of MTBE consists of a reac-
ously. In particular, the optimisation of the debutanizer tion sector, where i-C4H10 is combined with methanol
column with pervaporation membrane units to treat the to form the ether, and a separation sector where all the
side stream is reported in this work. Previous work on MTBE must be separated from unreacted methanol
the optimisation of non-conventional distillation and C4’s. Unreacted methanol forms azeotropic mix-
columns (Hoch & Eliceche, 1991) neither contemplate tures with MTBE and butanes. In the conventional
the azeotropic distillation case nor include simulta- Hüls process, a sequence of azeotropic distillation
neously the rigorous simulation of pervaporation mem- columns was used to break the azeotropes, thus recov-
branes. Thus the main objective of this work is the ering MTBE and methanol to be recycled to the reactor
optimisation of the azeotropic distillation column in a (Fig. 1). A hybrid distillation–pervaporation process
hybrid distillation/pervaporation system using a rigor- seems as an attractive alternative, because it combines
ous simulation of the azeotropic distillation column and the advantages of both methods. The use of hybrid
the pervaporation membrane units. Recent work (Höm- systems can improve the cost of the traditional Hüls
merich & Rautenbach, 1998; González & Ortiz, 2001b) separation sequence as reported recently (Hömmerich
explored the influence of the optimisation variables & Rautenbach, 1998).
566 A.M. Eliceche et al. / Computers and Chemical Engineering 26 (2002) 563–573
Different configurations for the hybrid distillation/ For a binary system, where the permeation of one of
pervaporation process can be used, locating the perva- the components Eq. (1) is favoured over the other
poration membrane to treat the feed, products or side component Eq. (2), the equation system representing
stream. In this work, the pervaporation membrane is the mass transport and energy balances can be synthe-
located to treat the side stream and remove the dis- sized as (González, 2000):
tributing component as a permeate, helping to improve
Total flux: J(c, T)= Jtotal(c, T)= J1(c, T)+ J2(c, T),
the top and bottom purity.
(1)
The separation of pure MTBE as the bottom product
of an azeotropic distillation column from a mixture of where 1 refers to methanol and 2 refers to MTBE.
C4’s, methanol and MTBE is performed by means of a Recently, a mathematical model based on the general-
combined distillation column and pervaporation mem- ized Fick%s law and the assumption that transport
brane process. The process being studied is shown in through the membrane is the rate-limiting step was
Fig. 3, where the pervaporation process is synthesized developed in order to describe the pervaporation flux of
as a single unit. both components (González & Ortiz Uribe, 2001a). The
The azeotrope formation of methanol with both prediction of the flux of methanol needed a concentra-
MTBE and C4 limits the purity of the products. A high tion-dependent diffusion coefficient whereas a simple
purity of MTBE is required in the bottom product (B) model with concentration-independent diffusivity was
of the column that separates a multicomponent mixture sufficient for the description of the MTBE flux. Pre-
(F1) including MTBE, C4’s and methanol. A side dicted results with the mathematical model and
stream (E) of the column, rich in methanol, is pro- parameters agreed with experimental data obtained in a
cessed through the membrane units. The membrane wide range of variables working with Sulzer’s Per-
selectivity allows the methanol in the column side- vap2256 membranes. Further details of the pervapora-
stream to be permeated and then condensed and recy- tion modelling can be found elsewhere (González &
cled to the reactor (Permeate liquid stream), thus Ortiz Uribe, 2001a). Thus, the resulting equations for
helping to improve the MTBE bottom product purity. the prediction of pervaporative fluxes are,
The retentate is recycled to the column (F2). The
butanes do not permeate through the membrane, and J1(c, T)
this is confirmed by some patents of the hybrid process = [− 1Do exp(~1)k2,1z1/(ln(k2,11 + 1)(k2,11 +1))]
(Chen et al., 1988) and experimental results carried out
by González (2000). × d1/dzp (2)
J2(c, T)= − Dz2d2/dzp. (3)
At zp =t c%= zF (13)
1 =[exp(P2x1,2/(P v1(T)k1,1)) − 1]/k2,1 (6) T= TF (14)
2 =(P2x2,2/(P (T)k1,2)),
v
2 (7) Values for fluxes and enriching factor are obtained
from experimental data. The parameters were found for
where ki is the activity coefficient for component i concentrations of methanol between 5 and 40% in a
obtained with a gE model, and i the composition of binary mixture methanol–MTBE. For different values
component i. of methanol composition in a binary mixture of
dm; methanol– MTBE, the values for fluxes were found at
Global mass balance: − =WJ(c, T) (8) different temperatures ranging between 30 and 50 °C.
dzA
The experimental results can be found in González and
Component mass balance: Ortiz Uribe (2001a). This allows to find the necessary
dci J(c, T) parameters for modelling the mass transport in the
− =[ii (c, T) −1]ci W, (9) membrane. For a certain composition, the flux of com-
dzA m;
ponent i varies with the temperature according to the
where W is the characteristic membrane width, ci is the expression:
mole fraction of component i in the permeate, T the
temperature along zA and ii is the enriching factor, Ji,T 1(c, T)= Ji,T 2(c, T)exp
Ea,i 1
−
1 , (15)
defined as: R T2 T1
As seen from the equations, the pervaporation mem- The pervaporation takes place when a vacuum of
brane is modelled in terms of a differential/algebraic 15–20 torr is maintained at the permeate side. For the
system involving lumped and distributed parameters. start-up a vacuum pump is needed, but when steady
The differential equations through the membrane must state is reached, vacuum is maintained by the permeate
be discretized in the direction of zp, and an integration condensation. The distillation column has 24 theoretical
package for Initial Value Problems (IVPs) is used to stages plus a condenser (stage 0) and a reboiler (stage
solve the equations along zA. As a simplification, it is 25). The membrane can be purchased in modules of 30
assumed that only two components permeate through m2, so several modules are needed to perform the
the membrane, while the rest of the components remain separation task. The number of modules required is
at the retentate side and do not permeate. This simplifi- automatically selected by fixing the required composi-
cation was experimentally tested, and holds with the tion of the permeant in the retentate, thus allowing the
pervaporation membrane Pervap2256 used to find the introduction of as many intermediate heaters as re-
experimental data. quired. For the integration, it was assumed a character-
The membrane favoured the permeation of methanol istic width of the module (W) of 50 m, and the
over MTBE so that Eqs. (1)– (15) could be used. The integration proceeds up to zA = 0.6 m for each module.
remaining components do not permeate through the Intermediate heating is provided, and the retentate
membrane, so the assumption of binary fluxes is hold. stream is heated until the temperature reaches 83 °C.
Experimental data were available for the system being The purpose of the intermediate heating is to avoid the
studied, so there was information regarding the values decrease of the flux across the membrane.
of the flux and retentate concentration (González,
2000), which allowed the implementation of this model.
5. Formulation of the optimisation problem
Table 1 Table 3
Feed to the column Cost results for the example shown
min Co(x)
x configurations and combinations of units can be imple-
s.t. g(x)50 mented adapting the software developed. Once the rig-
orous simulation of the pervaporation membrane
b L 5x5 b U, (18)
module is generated and interfaced automatically in
where the optimisation variables x are the reflux ratio HYSYS, the optimisation of different systems is feasi-
(r), a product flow rate (B), the side draw flow rate (E) ble by modifying the membrane models and implement-
and the column pressure (P), b L and b U are lower and ing a solution strategy for the system of differential
upper bounds on the variables x. equations.
The separation objectives like minimum composition
or recovery in the products and operating constraints
related with the maximum sidedraw flowrate are formu-
lated as inequality constraints g(x). 6. Numerical results
The simulation of the hybrid system and the solution
of problem Eq. (18) is implemented in HYSYS. The A mixture of iso-butene, MTBE and methanol has to
membrane modules are modelled in a user extension be separated to obtain MTBE as a bottom product of
unit in HYSYS. For each set of x variables provided by the column. A hybrid distillation/pervaporation process
the HYSYS optimisation code, a rigorous simulation of is analysed to perform this task. The liquid side draw is
the hybrid system is carried out including the solution located above the fresh feed and the retentate from the
of the differential equations to model the pervaporation membrane is recycled to the column in the stripping
membranes. HYSYS is a sequential modular simulator, section. The distillation column has 26 stages, the liquid
therefore the equality constraints modelling the pro- side draw is located in stage 4, the fresh feed in stage 9
cesses are not explicitly formulated as equality con- and the permeate from the membrane modules is recy-
straints in problem Eq. (18). This includes the cled in stage 20. The composition, temperature and
differential equations used to model the membrane pressure of the fresh feed (F1) are shown in Table 1.
units. For the sake of clarity, the mixture of butanes of the
The software developed allows the incorporation of original process is replaced by one component, iso-
different membrane models as user extension units in butene, which forms azeotrope with methanol, so a
HYSYS. Thus the optimisation of similar technologies ternary mixture is considered instead of the original
like the production of ethyl tert-butyl ether (ETBE) or multicomponent mixture. MTBE also forms an
other hybrid separation processes combining distillation azeotropic mixture with methanol at the working
columns with pervaporation membranes in different pressure.
Table 2
Optimisation results for the example shown
In this work, the objective is to obtain grade gasoline assumed to perform the pervaporation process, while
MTBE 98 wt.% with a recovery of 99% in the column. González and Ortiz Uribe (2001b), Hömmerich and
The inequality constraints g(x) of problem Eq. (18) are: Rautenbach (1998) have used 5 wt.% The number of
pervaporation modules required to perform this separa-
Minimum MTBE recovery in the recMTBE,B tion is calculated.
bottom product The HYSYS selected optimisation option to solve
] 0.99
problem Eq. (18) is the Mixed method. It starts the
Minimum MTBE composition in the xMTBE,B
optimisation with the Box method using a very loose
bottom product
] 0.98 convergence. After convergence, an SQP method is then
Maximum value of sidedraw E50.41 * used to locate the final solution using the desired toler-
Lj−1 ance. The optimisation variables selected were the side
draw flow rate, the reflux ratio and the pressure of the
being j the stage from where the side draw is extracted column.
and Lj − 1 the liquid flow rate of stage ( j −1). It is assumed that cooling water is used to condense
Optimisation results with the initial and final solution the distillate of the column and low pressure steam is
points, upper and lower bounds used in problem Eq. used to heat the bottom of the column and to provide
(18) are shown in Table 2. An important reduction in interstage heating between the membrane modules.
the reflux ratio of 22% and a significant reduction of For this example, 22 modules are required to obtain
9.7% in the operating cost have been achieved by the specified composition of methanol in the retentate.
selecting the optimum operating conditions. The temperature profile for the pervaporation modules
The operating cost of the distillation column is nearly is shown in Fig. 5.
90% of the total cost and the membrane cost is nearly The retentate methanol molar composition decreases
10% as reported in Table 3. Thus the optimisation of from 40 to 15% (C4-free basis), while the permeate
the operating conditions of the distillation column has composition remains above 99% throughout the mem-
a great impact in reducing the energy consumption and branes. Composition profiles in the membrane are
operating cost. shown in Fig. 6a and b. It is considered that only
The minimum composition and recovery are active MTBE and methanol permeate through the membrane,
constraints. The constraints related to the minimum while the iso-butene remains at the retentate side (Chen
composition and recovery of MTBE in the bottom et al., 1988).
product of the column are active because the cost of the Liquid composition profiles for the distillation
column increases with higher composition and recov- column, when a side stream is withdrawn and the
ery, requiring a higher reflux ratio. retentate is recycled to the column, are shown in Fig. 7.
A fixed composition of methanol in the retentate of Fig. 8 shows in detail the methanol composition
6% wt C4-free basis (equivalent to 15% mole-basis) is profile considering the ternary system, with a clear
Fig. 6. (a) Composition profiles of the permeate. (b) Composition profiles of the retentate.
reference to the stages where the fresh feed, recycle and objective of this work is to improve the operating cost
side stream are located. There is no formal exploration without changes in these variables, although the place-
of other alternatives for the location of these streams, ment of the fresh feed, side stream extraction and
as the purpose of this work is to optimise the operating retentate recycle locations can be modified to improve
conditions only, with no structural changes involved. the objective function, while maintaining the total num-
The extraction stream is optimally located, in the max- ber of stages fixed. The procedure developed by Hoch
imum value of the methanol composition profile in the and Eliceche (1991) treat the number of stages as
rectification section. continuous optimisation variables, allowing the simul-
The location of fresh feed, recycle and side draw taneous selection of feed and side stream locations with
have not been modified during the optimisation, as the the flowrates in a non-linear programming problem
572 A.M. Eliceche et al. / Computers and Chemical Engineering 26 (2002) 563–573
formulation. The implementation of this option is cur- 10% on the total cost of the process. It was not the
rently being studied, as it is not possible to do it within purpose of this work to compare the hybrid system cost
the HYSYS environment in an algorithmic fashion. with the cost of the original Hüls process, but from the
Instead, a trial-and-error procedure can be performed data in Table 3 it can be seen that the operating cost of
and was initially used to locate the side draw, fresh feed the reboiler of the column represents almost 90% of the
and recycle from the membranes reported in Table 2. total cost, thus replacing a distillation column with the
The cost of the process has been improved from 156 pervaporation membrane seems to be economically
to 148 $/h, representing a reduction of approximately attractive.