3484 Ind. Eng. Chem. Res.

2009, 48, 3484–3495

Control of a Column/Pervaporation Process for Separating the Ethanol/Water

Azeotrope
William L. Luyben*
Department of Chemical Engineering, Lehigh UniVersity, Bethlehem, PennsylVania 18015

The pervaporation process features a liquid feed and retentate and a vapor permeate. The phase change produces
a temperature decrease as the retentate flows through the unit. Since flux rates decrease with decreasing
temperature, the conventional pervaporation unit consists of several membrane modules in series with interstage
heating. The vapor permeate must be condensed for recovery and recycle, and refrigeration is usually required.
Hybrid systems of distillation columns and pervaporation units are frequently used in situations where distillation
alone is impossible or very expensive. Despite the many papers dealing with pervaporation, the issue of
dynamic control seems to be almost completely unexplored. That is the purpose of this paper. A hybrid
column/pervaporation process is studied that is designed to produce 99.77 wt % ethanol from a feed stream
of ethanol/water mixture with composition near the azeotrope. The control objective is to maintain the purity
of the ethanol product retentate stream in the face of disturbances in feed flowrate and feed composition.
There are two possible manipulated variables: permeate pressure and retentate temperature. Permeate flux is
increased by decreasing permeate pressure or increasing retentate temperature. A simplified dynamic
pervaporation model is developed that captures the essential features of the process using energy and component
balances along with overall pervaporation performance relationships. Dynamic simulations are used to
demonstrate the effectiveness of a control structure that uses a cascade composition/temperature structure. A
simple process modification is shown to improve controllability.

1. Introduction alcohol composite membranes. The process is clearly discussed

by Sander and Soukup.3 They provide engineering performance
Membrane processes are becoming more important in this
curves for removing water from an ethanol/water mixture. The
era of escalating energy costs because they typically require
much less energy than alternative separation methods such as curves show how the temperature and ethanol concentration of
distillation. Gas permeation using membranes is widely used the feed affect the flux of water through the membrane and the
for separating mixtures such as oxygen/nitrogen, hydrogen/ composition of the permeate. These performance curves are used
carbon dioxide, methane/nitrogen, etc. These gaseous membrane in this paper to provide a simple pervaporation model that
systems operate essentially isothermally because there are no captures the behavior of the process with sufficient fidelity for
phase changes. use in control studies. Sander and Soukup make the statement
There are several types of liquid membrane systems. One of that “The product concentration is simply controlled by adjusting
the most important is pervaporation in which the feed and the alcohol feed rate”. In most chemical plants, throughput is
retentate are liquid on the high-pressure side of the membrane, set by product demand and therefore cannot be used as a
but the permeate is removed as a vapor on the low-pressure manipulated variable in a process control structure.
side of the membrane. The phase change produces a drop in Many papers point out the economic advantages of pervapo-
temperature in adiabatic operation due to the heat of vaporiza-
ration over conventional processes. A quantitative economic
tion. Since flux rates through the membrane decrease with
comparison between azeotropic distillation and hybrid column/
decreasing temperature, the conventional pervaporation unit
pervaporation is presented by Guerreri4 whose results show
consists of several membrane stages in series with interstage
heating. The vapor permeate must be condensed, which often higher capital cost but much lower energy cost for the ethanol/
requires expensive refrigeration. water separation. A review of industrial applications of per-
Pervaporation has been the subject of many papers for several vaporation coupled with either distillation columns or chemical
decades. The literature is quite extensive with many papers and reactors is given by Lipnizki et al.5 They show a variety of
several books on the subject. The 1991 book edited by Huang1 flowsheets, list 22 chemical systems, and provide 72 references.
gives 262 citations in the first chapter. The book by Rautenbach Another review has been given by Feng and Huang6 who cite
and Albrect,2 written in 1989, has a chapter on pervaporation. over 100 references. Daviou et al.7 study the dehydration of
There many papers that discuss the fundamentals of diffusion, isopropanol using a hybrid column/pervaporator process. They
adsorption, evaporation, and membrane characteristics. Some develop a process design using the HYSYS commercial
papers develop detailed mass-transfer models. A number of simulator. The model of the pervaporator was developed in
papers discuss design considerations and present comparative gProms and integrated into HYSYS using a user extension
economic studies of pervaporation versus alternatives such as module. They found the optimum economic design using two
distillation. design optimization parameters: number of modules and amount
Industrial applications began in the 1970s. This was made of interstage heating. Van Hoof et al.8 presented an economic
possible by the development of highly selective polyvinyl comparison between azeotropic distillation and different hybrid
* To whom correspondence should be addressed. E-mail: WLL0@ systems combining distillation with pervaporation for the
Lehigh.edu. Tel.: 610-758-4256. Fax: 610-758-5057. dehydration of isopropanol.
10.1021/ie801428s CCC: $40.75  2009 American Chemical Society
Published on Web 03/09/2009
 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009 3485

Figure 1. Column/pervaporation flowsheet.

No papers have been found that specifically deal with the component j in retentate liquid (kmol/m3), and CPj ) concentra-
issue of dynamic control in any quantitative way. As mentioned tion of component j in permeate vapor (kmol/m3).
above, Sander and Soukup3 speculate that feed rate could be The density of the incompressible liquid permeate (731 kg/
used. The purpose of this paper is to study the dynamic control m3) is assumed to be independent of temperature and composi-
of a pervaporator system coupled with a distillation column. tion over the small ranges present in the pervaporator. Therefore
The example system is the important ethanol/water separation. the liquid molar concentrations can be calculated from the
The commercial simulation tools of Aspen Technology are used weight fractions zwRj of the retentate liquid.
in this study. Figure 1 gives the process flowsheet studied.
731 kg/m3
2. Pervaporation Model CR j ) (zwR j) (2)
MWj
The study of the dynamic control of any process requires a
dynamic model that is capable of capturing the important where MWj ) molecular weight of component j (kg/kmol) and
dynamic features of the process. The macroscopic effects of zwRj ) weight fraction of component j in the liquid retentate.
changing manipulated variables on the variables to be controlled The molar concentrations in the permeate vapor can be
must be capture with sufficient fidelity to give realistic dynamic calculated assuming ideal gases.
and steady-state responses. The model used for control does
not need to be as detailed as one used for gaining fundamental (Pperm)(zPj)
understanding of the microscopic details of mass and energy CPj ) (3)
(0.08206)(TP)
transfer and phase equilibria in these complex, multiphase
systems. where Pperm ) pressure on the permeate side of membrane (atm),
2.1. Diffusivities. In this spirit, an approximate dynamic zPj ) mole fraction of component j in the permeate, and TP )
model was developed that is based on pervaporation perfor- temperature of the permeate (K).
mance characteristic curves. Sander and Soukup3 give curves The permeate pressure for the data given in Figure 2 is
for the ethanol/water system showing how the flux of water assumed to be 0.015 atm. The temperatures of the retentate and
through a pervaporation membrane and the composition of the permeate are assumed to be the same and decrease as mass
permeate vapor vary with the concentration of ethanol in the transfer occurs in the cell because of the phase change.
liquid feed. Figure 2 panels a and b reproduce these curves. The data given in Figure 2 panels a and b are used to calculate
Note that these curves show concentrations in wt % and mass the diffusivities of both ethanol and water. A point on the
flowrates. abscissa (wt % ethanol in the feed, which is assumed to be the
The model assumes that the diffusion through the membrane retentate in each cell) is selected. The corresponding composition
depends on the concentration difference between the retentate of the permeate is read from Figure 2a and the corresponding
and permeate sides of the membrane for each component. flux of water is read from Figure 2b. The water compositions
fluxj ) Dj(CR j - CPj) (1) of both the retentate and the permeate are converted to mole
fractions, so zP,water and zR.water are known. Since the system is
where fluxj ) molar flux of component j (kmol h-1 m-2), Dj ) binary, the mole fractions of ethanol in both the retentate and
diffusivity of component j (m/h), CRj ) concentration of permeate can be calculated (zP,ethanol and zR,ethanol). The liquid
3486 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009

temperature dependent diffusivities used in the model are given

as

Dwater ) 1.158×109e-9385/T
Dethanol ) 2.028×106e-9385/T (5)

2.2. Dynamic Component and Energy Balances. A lumped

model of the pervaporation unit is used to describe the dynamic
behavior. Three cells are assumed in each pervaporation module.
Since the liquid retentate provides most of the capacitance of
the pervaporation unit, the ordinary differential equations
describing the dynamic changes compositions and temperature
of the retentate in each cell are used. The molar holdup in each
cell MR is assumed constant, so the total molar balance is
algebraic.

dMR
) 0 ) FR,n-1 - FR,n - FP,n (6)
dt

dhR,n
MR ) FR,n-1hR,n-1 - FR,nhR,n - FP,nHP,n (7)
dt

dzR,n,j
MR ) FR,n-1zR,n-1,j - FR,nzR,n,j - FP,nzP,n,j (8)
dt

where FR,n ) molar flowrate of liquid retentate from cell n

(kmol/h), FP,n ) molar flowrate of vapor permeate from cell n
(kmol/h), hR,n ) molar enthalpy of liquid retentate in cell n (GJ/
kmol), and HP,n ) molar enthalpy of vapor permeate leaving
cell n (GJ/kmol).
The flowrate of permeate is the sum of the two component
fluxes multiplied by the area of the membrane (Amem), and the
composition of the permeate is determined by these two fluxes.

FP,n ) Amem(fluxn,ethanol + fluxn,water)

zP,n,j ) Amemfluxn,j /FP,n (9)

The flux of each component in each cell is calculated using

eq 1 and eq 5 since all compositions and temperatures are
known. The temperature of the retentate is calculated from the
Figure 2. (a) Permeate composition; (b) permeate water flux at 373 K. known liquid enthalpy hR,n and the known retentate composition
zR,n using physical property relationships. The UNIQUAC
retentate concentrations CR,water and CR,ethanol can then be relationships are used in both the pervaporation unit and the
calculated using the liquid density of 731 kg/m3 and eq 2. distillation column.
The flux of water is converted to molar units, so fluxwater is The differential equations (eq 7 and eq 8) for each cell and
known. The molar flux of ethanol is calculated by multiplying each module are incorporated in an Aspen Custom Modeler
the molar water flux by the ratio of the molar composition of program, which is given in Table 1. The invaluable help of Jay
ethanol to water in the permeate. Yun of Aspentech is gratefully acknowledged in overcoming
the significant complexities of translating the model into a
fluxethanol ) fluxwater ( )
zP,ethanol
zP,water (
) fluxwater
zP,water)
1 - zP,water
(4)
workable Aspen Custom Modeling file.

3. Pervaporation/Column System
The concentrations of both components in the vapor permeate The system studied is based on a process discussed by Ryan
are calculated using eq 3. Finally, the diffusivities of both and Doherty9 that produces an 85 mol % ethanol distillate from
components are calculated using eq 1 since all fluxes and a 22-stage distillation column with a feed flowrate of 700 kmol/h
concentrations are known. and feed composition of 7 mol % ethanol and 93 mol % water.
Diffusivities were calculated over a range of feed composi- The bottoms product from the column has a composition of
tions given in Figure 2a,b. The calculated diffusivities are almost 99.9 mol % water. Figure 1 gives stream conditions, heat duties,
constant over the range from 92 to 99.5 wt % ethanol. Average and equipment sizes. The column operates at 1.1 atm, giving a
values of Dwater ) 0.0137 m/h and Dethanol ) 2.4 × 10-5 m/h reflux drum temperature of 354 K. The column feed is
are used. introduced on stage 19, which is the location that minimizes
These values are at 373 K. The temperature dependence of reboiler heat input.
the diffusivities on temperature is reported3 to be of the The distillate is heated to 375 K and fed to the first of five
Arrhenius type, with a doubling for every 10 K. Therefore the pervaporation modules in series. Each module has three cells.
 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009 3487
Table 1. ACM Pervaporator Model
3488 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009

Table 1. Continued

The retentate decreases in temperature from 375 to 358 K as it attractive to modify the design of the process to improve
flows through the first module due to the phase change. The dynamic controllability by using reheating on all modules.
temperature of the permeate from each cell is assumed to be The permeate pressure is 0.155 atm, and the vapor permeate
equal to the temperature of the retentate in that cell. The distillate streams from each module flow to a refrigerated condenser (with
is pumped up to a pressure of 7 atm to ensure that the reten- area 2.74 m2 and overall heat transfer coefficient of 730 kcal
tate remains liquid. h-1 m-2 K-1), which uses 62.5 kmol/h of 273 K refrigerant to
The retentate streams leaving the first and second modules condense the permeate vapor. Condenser heat duty is 108 kW.
are heated back up to 375 K. The drop in retentate temperature The total permeate flowrate is 8.69 kmol/h with a composition
in the last three modules is quite small (7 K in module 3 and 3 of 3.84 mol % ethanol, and this stream is fed back into the
K in module 4) because the flux rates are small due to the distillation column along with the fresh feed on stage 14
decrease in retentate water composition. Therefore, it seems (numbering from the top with the reflux drum as stage 1).
logical to save the capital investment in additional heat Figure 3 illustrates what happens inside the first 3-cell module.
exchanger area and use no heating on these retentate streams. Temperatures decrease from cell to cell. Permeate ethanol
We will return to this question of how many modules should composition increases and permeate flowrate decreases from
use reheating later in this paper and show that it may be cell to cell. The fluxes from each of the three cells (with different
 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009 3489

Figure 3. -Module 1.

Figure 4. Open-loop responses.

temperatures and compositions) are mixed to give the total vapor (328 m2/m3). Thus the 300 m2 module has a volume of 0.9 m3
permeate from the module. Each module has a total area of that is split among the three cells in the module.
300 m2 or 100 m2 per cell. There are two potential variables that can be manipulated to
control the product composition: retentate temperature or
4. Dynamics and Control permeate pressure. To see what the dynamic response of product
impurity is to changes in these two input variables, openloop
The control objective is to maintain the purity of the ethanol step changes were made. Results are shown in Figure 4. The
product stream leaving the last stage of the pervaporation unit. solid line is for a step change in the setpoint of the permeate
The desired purity is 99.42 mol % ethanol (99.77 wt %). Process pressure controller from 0.15 to 0.05 atm made at time equal
control wisdom says that it is more effective to control the 0.2 h. The response of product impurity xproduct,water is slow, but
impurity level than the purity level in a high-purity stream. more importantly there is little change. Thus the process gain
Therefore, we choose to control the water impurity at 0.58 mol between product impurity and permeate pressure is very small,
% water. A 5-min deadtime is assumed for the composition so using this manipulated variable would only be able to handle
measurement. very small disturbances.
An important dynamic parameter is the holdup in the The dash line in Figure 4 is for a step change in the setpoints
pervaporator unit. Since the mass of the liquid retentate phase of the three temperature controllers that manipulate heat inputs
is much larger than that of the vapor permeate phase, it will to the three heaters on the retentate feed streams to the first,
dominate the dynamic response. The holdup depends on the second, and third pervaporation modules. At time equal 0.2 h,
volume. In this study, the data given in Geankoplis10 is used these three setpoints are raised from 375 to 380 K. There is a
3490 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009

Figure 5. (a) Aspen Dynamics PFD; control structure using Tret. (b) Controller faceplates using Tret.

very significant change in product impurity. Therefore it appears 3. Base level is controlled by manipulating bottoms flowrate.
that a composition/temperature cascade control structure should 4. Reflux drum level is controlled by the valve on the retentate
provide effective control. stream leaving the last module.
The Aspen Dynamics process flow diagram is shown in 5. Retentate pressure is controlled by manipulating the speed
Figure 5a. The control structure is summarized below. All of the distillate pump.
controllers have conventional PI action except level controllers, 6. The temperature on stage 19 is controlled by manipulating
which are proportional only. reboiler heat input. A 1-min deadtime is inserted in this loop.
1. Feed to the column is flow controlled. 7. Column pressure is controlled by manipulating condenser
2. Reflux is flow controlled. heat removal.
 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009 3491
Table 2. Controller Tuning Parameters
manipulated variable retentate temperature permeate pressure refrigerant flowrate column reboiler heat input
controlled variable retentate impurity retentate impurity permeate pressure (atm) column stage 19 temperature (K)
SP 0.0058 (mf water) 0.0058 (mf water) 0.155 atm 375 K
transmitter span (mf water) 0-0.01 0-0.01 0-0.5 350-450 K
OP range 370-390 K 0-0.5 atm 0-100% 0 to 6.63 × 106 cal/sec
ultimate gain 0.76 31 35 1.3
ultimate period (min) 71 28 3 13
KC 0.34 (ZN) 23 (ZN) 11 (TL) 0.4 (TL)
τI (min) 60 (ZN) 44 (ZN) 6..6 (TL) 6.6 (TL)
controller action direct reverse direct reverse

8. Inlet feed temperature to the pervaporation unit is controlled The lower graphs in Figure 6 give results when retentate
by manipulating heat input to the heater HX0 (TC1 controller). temperature is used. Product impurity is well controlled with
9. Inlet retentate temperature to the second pervaporation the steady-state retentate temperature increased from 375 to
module is controlled by manipulating heat input to the heater 378 K.
HX1 (TC2 controller). It should be remembered that the performance of this loop
10. Inlet retentate temperature to the third pervaporation could be improved if the composition deadtime can be reduced
module is controlled by manipulating heat input to the heater or if a feedforward structure were used to increase retentate
HX2 (TC3 controller). temperature if feed flowrate increases.
11. The water impurity in the retentate product stream is Figure 7 gives the responses using the retentate temperature
controlled by manipulating the setpoints of the three inlet as the manipulated variable for changes in either feed flowrate
retentate temperature controllers (CC controller). or feed composition to the distillation column. The feed flowrate
12. Pressure in the permeate drum is controlled by manipulat- changes are 20% increases (solid lines) and 20% decreases
ing the flow of refrigerant to the permeate condenser. (dashed lines) at 0.2 h. The feed composition disturbances are
13. Liquid level in the permeate drum is controlled by changes from 7 to 8 mol % ethanol (solid lines) and changes
manipulating the flow of recycle permeate back to the column. from 7 to 6 mol % ethanol in the column feed at 0.2 h. Good
regulatory performance is demonstrated. However, the closed-
The product composition, the permeate pressure, and the
loop dynamics are fairly slow, and it takes about 3 h to return
column temperature loops are tuned by running relay-feedback to the desired purity. In the following section, we discuss a
tests and applying either Ziegler-Nichols or Tyreus-Luyben modification of the process to improve dynamics.
tuning rules. Table 2 gives controller parameters. Figure 5B Figures 6 and 7 show that there is a small amount of what
shows the controller faceplates. Notice that the three retentate looks like inverse response in the permeate composition.
temperature controllers are on cascade, receiving their setpoint Remember that these are the responses to disturbances in feed
signals from the output signal from the composition controller. flowrate and feed composition to the middle of the distillation
Figure 6 shows the responses of the system for a 20% increase column. The dynamic effects have to work their way to the top
in feed flowrate to the column. The upper two graphs show of the column and then through five pervaporation modules,
what happens when product composition is controlled by each of which has several cells. So the system is a very
permeate pressure. The refrigerant control valve goes wide open distributed one. It is not surprising that it can exhibit dynamic
and product impurity cannot be controlled at the desired 0.58 features similar to those seen in tubular reactors. The “wrong
mol % water. This poor result is due to the very small process way” effect is sometimes seen in plug flow reactors when the
gain between product impurity and permeate pressure. capacitance contributed by the solid catalyst can produce

Figure 6. 20% increase in feed flowrate using Pperm or Tret.

3492 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009

Figure 7. Feed disturbances using Tret.

Figure 8. Flowsheet with five heat exchangers.

unexpected responses. The similarity with the column/pervapo- heat exchangers on the inlet streams to the first three modules.
ration complex system is obvious. Thus there is a significant lag between the variable to be
controlled (product impurity) and the variable to be manipulated
5. Process Modification (heat input).
The results shown in Figure 7 reveal fairly large closed-loop This suggests that it may be possible to improve control by
time constants. This occurs because of the large holdup of modifying the process. Installing heat exchangers on the feed
retentate in the five modules relative to the small flowrate of streams to the last two modules would reduce the lag in the
retentate. The total retentate holdup in the five modules is 4.5 loop.
m3. The flowrate of distillate to the pervaporation unit is 57.4 Figure 8 shows the resulting revised flowsheet. The composi-
kmol/h, which corresponds to 54 L/min. Thus the residence time tion of the product stream (retentate leaving module 5) is kept
of retentate is about 80 min. the same as in the previous design. The retentate feed temper-
In addition, the corrective action is being taken in the front atures to the five modules are reduced from 375 K in the
end of the pervaporator unit, using the heat input to the three previous design with three heat exchangers to 371 K with five
 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009 3493

Figure 9. -Comparison of three- and five-heat-exchanger processes; 20% feed flowrate disturbance.

Figure 10. Feed flowrate disturbances with five-heat-exchanger process.

heat exchangers. The required temperature is lower because the temperature controllers on the feeds to the first two modules
fluxes in the last two modules have been increased due to their have a fixed setpoint of 371 K.
higher inlet temperatures. In the three heat exchanger design, With this new control structure and new process, the
these temperatures are 368 and 365 K. In the five heat exchanger composition controller was retuned. Table 3 compares the three-
design, they are 371 K. The heat duties in the two new heat heat-exchanger process with the five-heat-exchanger process.
exchangers are smaller than those in the first three. Note that the ultimate period is smaller in the new design (49
From a control point of view, the best control would be to versus 71 min). Figure 9 compares the responses of the two
use only the heat input to the last module. However, it is quite designs for a 20% increase in feed flowrate. The solid lines are
small (9.69 kW) compared to the heat inputs in the upstream for the original three-heat-exchanger process. The dashed lines
heaters (37.5, 34.9 25.9, and 16.8 kW) and would not be able are for the modified five-heat-exchanger process. The control
to handle large disturbances. A decrease in throughput would performance is improved in terms of both settling time and
drive the heat input in this last heat exchanger to zero, but the maximum transient deviation. Notice that the initial steady-state
composition setpoint would not be maintained because the temperature is lower in the five-heat-exchanger process.
temperature would be too high (control would be lost). To Figures 10 and 11 show how a number of variables change
handle a 20% decrease in feed and not encounter a zero heat for disturbances in feed flowrate and feed composition. The
input constraint, the heat inputs in the last three heaters must output signal from the composition controller (OPCC) changes
be adjusted. Thus, with the modified process, the output signal the setpoints of the temperature controllers on the last three
from the composition controller changes the setpoints of the heaters (TC3, TC4, and TC5) and the corresponding heat duties
feed temperature controllers to the last three modules. The change (Q2, Q3, and Q4). The setpoints of the temperature
3494 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009

Figure 11. Feed composition disturbances with five-heat-exchanger process.

Figure 12. Product purity setpoint changes.

controllers on the first two heaters (TC1 and TC2) are constant fifth heat exchanger changes very little after going through
at 371 K, but their heat duties change as the load changes. Figure some transient responses.
10 shows that none of the last three heat exchangers saturate at It is interesting to explore what would happen if a pervapo-
zero for the 20% decrease in feed. If only the last two heat ration unit with less area, and therefore less holdup, were used.
exchangers were used, saturation would occur for this size drop Would the dynamic control improve because of the smaller
in feed. Figure 11 shows that feed composition disturbances residence time of the retentate? Cutting the area in half (300 to
are well handled. 150 m2 in each module) and also cutting the volume in half
Another type of disturbance would be to change the (0.45 m3 in each module) and maintaining the same product
setpoint of the composition controller. Figure 12 shows that composition required that the feed temperatures of a five-heat-
increasing the desired product purity (reducing the impurity exchanger process be increased from 371 to 381 K. Tuning the
setpoint) from the original 0.58 mol % water. At time equal composition controller (using all five heat exchangers) gave an
0.2 h, the composition setpoint is reduced to 0.2 mol % water. ultimate period of 53 min. This should be compared to the 75
The setpoints of the last three heaters are increased from 371 min ultimate period for the larger area unit with the same control
to 374 K to achieve the higher purity. The setpoints of the structure. Figure 13 shows that dynamic control improves in
first two heaters remain at 371 K. At time equal 5 h, the terms of settling time. The disturbance is a 20% increase in
composition setpoint is further reduced to 0.1 mol % water. feed flowrate. The maximum deviation is not reduced, but is
The setpoints of the last three heaters are increased again up actually slightly larger. This is due to the disturbance affecting
to 378 K to achieve the still higher purity. The heat duties the process more quickly when there is less capacitance to
in all the heat exchangers increase, but the heat duty in the attenuate disturbances.
 Ind. Eng. Chem. Res., Vol. 48, No. 7, 2009 3495

Figure 13. Smaller membrane; +20% feed flowrate.

Table 3. Controller Tuning Parameters for Three- and The dynamic controllabilities of two different process designs
Five-Heat-Exchanger Processes are compared. As expected, moving the manipulated variable
retentate temperature retentate temperature (heat input to feed preheaters) closer to the controlled variable
manipulated into modules into modules (retentate product composition from the last module) improved
variable 1, 2 and 3 3, 4 and 5
control. Smaller pervaporation units give faster dynamics.
controlled variable retentate impurity retentate impurity
SP 0.0058 (mf water) 0.0058 (mf water) Literature Cited
transmitter span (mf water) 0 to 0.01 0 to 0.01
OP Range 370-390 K 370-390 K (1) Huang, R. Y. M., Ed. PerVaporation Membrane Separation Pro-
ultimate gain 0.76 0.77 cesses; Elsevier: New York, 1991.
ultimate period (min) 71 49 (2) Rautenbach, R., Albrecht, R. Membrane Processes; Wiley: New
KC 0.34 (ZN) 0.35 (ZN)
York, 1989.
τI (min) 60 (ZN) 40 (ZN)
(3) Sander, U.; Soukup, P. Design and operation of a pervaporation plant
controller action direct reverse
for ethanol dehydration. J. Membr. Sci. 1988, 36, 463–475.
(4) Guerreri, G. Membrane alcohol separation processsintegrated per-
6. Conclusion vaporation and fractional distillation. Trans. IChemE, Part A 1992, 70, 501–
508.
This paper has studied the control of a column/pervaporation (5) Lipnizki, F.; Field, R. W.; Ten, P. Pervaporation-based hybrid
process for producing high-purity ethanol to overcome the process: a review of process design, applications and economics. J. Membr.
azeotropic limitation encountered in distillation. A conventional Sci. 1999, 153, 183–210.
(6) Feng, X.; Huang, R. Y. M. Liquid separation by membrane
control structure is developed that provides effective disturbance pervaporation. Ind. Eng. Chem. Res. 1997, 36, 1048–1066.
rejection for both production rate and feed composition changes. (7) Daviou, M. C.; Hoch, P. M.; Eliceche, A. M. Design of membrane
A simple pervaporation model is developed in Aspen Custom modules used in hybrid distillation/pervaporation systems. Ind. Eng. Chem.
Modeler that captures the important dynamic features of the Res. 2004, 43, 3403–3412.
process. The model uses pervaporator characteristic performance (8) Van Hoof, V.; Van den Abeele, L.; Buekenhoudt, A.; Dotremont,
C.; Leysen, R. Economic comparison between azeotropic distillation and
curves to determine diffusivities. Component fluxes depend upon different hybrid systems combining distillation with pervaporation for the
composition driving forces between the retentate and permeate dehydration of isopropanol. Sep. Purif. Technol. 2004, 37, 33–49.
sides of the membrane. The dynamics of the pervaporation cells (9) Ryan, P. J.; Doherty, M. F. Design/Optimization of ternary hetero-
are assumed to be dominated by composition and energy geneous azeotropic distillation systems. AIChE J. 1989, 35, 1592–1601.
(10) Geankoplis, C. J. Transport Processes and Unit Operations, 3rd
capacitance of the liquid retentate. ed.; Prentice-Hall, New York, 1993; p 761.
A composition cascade to retentate temperature control
structure is demonstrated to provide stable regulatory control. ReceiVed for reView September 23, 2008
The use of permeate pressure as a manipulated variable is shown ReVised manuscript receiVed January 31, 2009
Accepted February 6, 2009
to provide poor control because of the very small process gain
between product purity and permeate pressure. IE801428S