Chem. Eng. Technol.

19 (1996) 391 -397

391

Simulation and Design of Membrane Plants with AspenPlus

Robert Rautenbach, Riidiger Knauf, Alexander Struck, and Jens Vier * In this paper the simulation of hybrid processes containing membrane units will be discussed. For this purpose a user-defined module for simulation and design of membrane processes was implemented into the simulation program AspenPlus. The advantages can be summarized as follows: - any combination of membrane processes with all other units already implemented in AspenPlus is possible, including internal recycle streams, - utilization of the physical property models and data bases of AspenPlus is possible, - cost and sensitivity analysis can be performed. These benefits are demonstrated in detail for a membrane vapor recovery unit for the treatment of tank farm off-gas, for a two-stage reverse osmosis plant for organic/-organic separations and for a combination of distillation and pervaporation for the separation of a dimethylcarbonate/methanol mixture.

1 Introduction
Membrane processes are well established in desalination and wastewater treatment, dialysis, ethanol dehydration and nitrogen enrichment. Presently, many applications of membrane systems are end-of-pipe installations. But due to the development of chemically and mechanically resistant membranes and modules and due to interesting separation characteristics, the chances of gadvapor permeation, pervaporation or reverse osmosis as a part of a separation train including other unit operations like distillation, adsorption or absorption are growing. In almost all cases, such hybrid processes are not a simple straight-forward combination of unit operations but include recycles. Decisive for the economics of such processes are the concentrations and mass flow rates between the process stages and usually, these parameters must be varied in a wide range in order to determine the optimal combination. In principle, AspenPlus is a design tool developed just for such purposes. Modules for membrane processes, however, are not included and, consequently, process combinations including a membrane stage are not covered by the standard AspenPlus program. We developed such modules for the membrane processes gas permeation, pervaporation and reverse osmosis. The benefits of the implementation of these modules are demonstrated in detail for three examples
- a membrane vapor recovery unit for the treatment of

- a combination of distillation and pervaporation for the

separation of dimethylcarbonate/methanol mixtures.

2 Implementation of User-Defined Subroutines

AspenPlus offers the possibility of simulation of any combination of various unit operations (distillations, mixers, reactors, condensers etc.). Additionally, so called usermodels can be integrated into the process flow scheme. The characteristics of the user-model are defined by a user-supplied FORTRAN subroutine. A list of flow sheet variables (flow rate, pressure, temperature of inlet streams, etc.) is transferred to the subroutine. These variables can be manipulated in order to describe a given unit operation model (the solution/diffusion model for gas permeation or reverse osmosis or the empirical model for pervaporation). The results (flow rates, pressure, temperature of the outlet streams, etc.) are retransferred to AspenPlus and will be used in the next block. The user-block has one inlet (feed) and two outlet streams (retentate and permeate, s. Fig. 1). In gas permeation, the local material transport (molar partial fluxes) can be almost always calculated with sufficient accuracy by )

tank farm off-gas, - a two-stage reverse osmosis plant for organidorganic separations,

In reverse osmosis, the situation is different. The use of one, generally valid transport model is not advisable. Accordingly, the user has the choice among several (in total five) models, for example, in case of diluted aqueous salt solutions between the classic two-parameter solution diffusion model

Prof. Dr.-Ing. R. Rautenbach, Dip].-Ing. R. Knauf, Dip1.-Ing. A. Struck, Dipl.-Ing. J. Vier, Institut fur Verfahrenstechnik, RWTH Aachen, Turmstr. 46. D-52056 Aachen. Germany. 0 VCH Verlagsgesellschaft mbH, D-69451 Weinheim, 1996

1) List of

symbols a t the end of the paper

0930-75 16/96/05 10-0391 $10.00+ .25/0

392

Chem. Eng. Technol. 19 (1996) 391 -397 vapors. Depending on the vapor composition and temperature, this corresponds to a hydrocarbon concentration of 500 to 900g/m3. According to the German clean air act (TA-Luft) these off-gases must be treated. The concentration of organic compounds must be reduced to 150 mg/m3 when the total emitted flow rate of organics exceeds 3 kg/h. Carcinogenic substances, such as benzene or toluene, are restricted to 5 mg/m3. In the past years, in Germany, more than 25 vapor recovery plants have been installed, where the first separation stage consists of a hybrid process of gaspermeation and condensation according to Fig. 1. The flow scheme for a simple vapor recovery unit consists of a compressor, a condensor or scrubber and a membrane stage with permeate recycle to the compressor (s. Fig. 1). There are two possibilities of designing this separation stage: either so that the flowrate of organics in the retentate does not exceed the maximum flowrate of 3 kg/h or by addition of a second stage (membrane stage, gas engine or pressure swing adsorption) in order to meet the requirements of the TA-Luft. The saturated off-gases are compressed (3 to 9 bar absolute pressure) and fed into a condenser. One part of the vapor stream is liquified and recovered (recycled to the gasoline storage). The gaseous stream is fed into the membrane stage, where it is separated into an hydrocarbon enriched permeate and a hydrocarbon depleted retentate. The membranes used for the separation of organic vapors from air are composite membranes with a 1 pm polydimethylsiloxane (PDMS) selective layer. In PDMS the permeability of hydrocarbons is significantly higher than of permanent gases like nitrogen or oxygen. The permeate is recycled to the compressor and the condenser. More detailed information on this process is given in [2 - 41. In principle, the flow scheme of Fig. 1 consisting of compression, condensation, and membrane unit can be modified: Instead of a condensation unit an absorption stage could be installed if liquid gasoline from the storage tanks is available, since this is a very suitable absorption liquid. The performance of the membrane stage could be optimized by applying a vacuum on the permeate side. The condenser for organics recovery could be arranged in the permeate stream, because this stream contains the maximum hydrocarbon concentration (dotted line in Fig. 1 : this flow scheme often is especially advantageous in case of low concentrated vapor streams). Further important design variables are the compressor capacity (pressure ratio and flow rate), the condensor (scrubber) temperature, and the membrane area. These process variations must be studied in order to define the oDtimum with respect to volatile organic carbon (VOC) recovery, power consumption, investment costs. membrane replacement, and operating costs. Furthermore, the design
and optimization of such aplant has
to consider the

tank-farm off-gas

compressor to post treatment

A : .frI
0 ,

condensate

permeate

y--.
: 1

/-1 -

I i
5 I
- 7

r
condensate

-.--

vacuum pump

Figure 1. Simplified flow scheme of a vapor recovery unit.

or the very simple model assuming a constant rejection coefficient R

The transport coefficients must be known or determined from experiments with the actual material/membrane system [I]. For pervaporation a regression analysis empirical model has been incorporated in the program, since the available semiempirical models are not generally valid. Permeate flux and concentration are calculated according to experimentally determined functions of feed composition and temperature. Material transport Eqs. (I), (2), and (3) or, in case of pervaporation, a set of experimental data must be solved in combination with differential balances for mass, momentum and energy. Since the influence of flow pattern on module performance is small in all membrane processes, the simple so called cross-flow model, assuming plug flow at the feed side and unhindered permeate flow, has been chosen for the calculations. Furthermore, additional mass transfer resistances (friction losses, concentration polarization, etc.) are neglected. The comparison with field data shows that these simplifications are reasonable for a basic plant design. The simulation of hybrid processes and complex process flow sheets has to concentrate on the analysis of recycling effects between the process stages rather than on a detailed description of single units. Hence, in AspenPlus most of the conventional unit operations are described by rather simple models.
3 Example - Gasoline Vapor Recovery

The off-gases from storage and handling of gasoline (on tank farms Or gasoline stations) are saturated with gasoline

varia-

tion of membrane area, condensation temperature, feed

Chem. Eng. Technol. 19 (1996) 391 -397 pressure, and permeate pressure (by applying vacuum on the permeate side). A typical tank farm off-gas composition and the corresponding permeabilities of the components are listed in Tab. 1. This composition corresponds to a concentration of about 612 g/m3. The simulation will be performed for an off-gas flow rate of 200 m3/h at 1 bar absolute pressure and 20C. The AspenPlus physical property models (here: ideal behavior) supply the required thermodynamic data of the material streams (enthalpy, entropy, density, etc.). In Figs. 2 and 3 the simulation results are shown for the simple pressurized mode (flow scheme s. Fig. 1). The absolute permeate pressure is 1 bar, the membrane area is kept constant. The condenser is modelled as ideal vapor/liquid equilibrium stage with total phase separation. Instead of this model any other AspenPlus unit operation or physical property model can be used. This is a tremendous advantage compared to taylor-made membrane simulation programs. In Fig. 2 the hydrocarbon retentate concentration is plotted against the condensation temperature for different feed pressures. As can be expected, the concentration is reduced for increasing feed pressure and decreasing condensation temperature (it should be noted that the condensation stage is operated at about ambient temperature since in most applications there are no cooling utilities available). Very low retentate hydrocarbon concentrations (< 10 mg/m3) are not advisable since these low concentrations correspond to a low driving force at the exit of the membrane unit resulting in lower permeate concentrations and, accordingly, lower concentrations in the condenser. Hence, an increase of the feed pressure will not reduce the retentate concentration below a certain limit. The limit may be defined by setting the permeate concentration equal to the offgas concentration. This explains why a membrane unit can meet the strict TA-Luft requirements in principle but only for unrealistically large membrane areas. This is demonstrated in Fig. 3a. An increase of membrane area results in a decrease of retentate concentration, an increase of the permeate flowrate, and, consequently, an increase of the compressor power consumption. In Fig. 3 b the influence of membrane area and feed pressure on compressor power consumption is shown. High
Table 1. Off-gas composition and permeabilities of the components (according to [2]).
component ethane propane i-butane n-butane n-pentane n-hexane oxygen nitrogen mole fraction [mol-%I
0.5 1.5 5.0
200

393

+a

Feed Rowrate 200 ~n',~p/h Feed concentrabon 612 g H , J m 3 , 150

(0

k
Y

2
0
0

u )

0: c

100

a , a ,

m E

50

2 !

0
10 20
30

Temperature ["C]
Figure 2. Influence of condensation temperature and operating pressure on the separation efficiency (membrane area: 40 m2, absolute permeate pressure 1 bar).

(0

2250 -

Condensabon temp 20 "C Feed flowrate 200 m p ,'h / Feed concentrabon 612 gH$naSTp

2
Y

s c
c

0)

.
0.6-

Condensabon temperature 20C Feed Rowrate 200 m J , h ' Feed concentrabon 612 g J ,m 3 ,

: :

,
02
10 20

,
40

,
50

1
60

30

Membrane area [m']

Figure 3. Hydrocarbon retentate concentration and specific energy consumption for variation of membrane area (condensation temperature: 20C).

permeability [m3/m2h bar]

2.5 4.6 6.0 9.0 11.5 26.0 0.81 0.4

8.0
5.0

pressure and high membrane area result in a high power consumption. Accordingly, very low hydrocarbon concentrations in the retentate can only be achieved with high energy consumption. Since the intention of this paper is a demonstration of the possibilities of AspenPlus combined with a membrane subroutine, gasoline vapor recovery units will not be discussed

2.0 16.0 62.0

394 in more detail in this paper. A detailed optimization must be based on the individual situation (capacity, utility costs, location, etc.).

Chem. Eng. Technol. 1 9 (1996) 391 -397 A two-stage process designed as a stripping section increases the product recovery significantly. This process can be realized by subdividing the entire membrane area of a single stage into two parts. No additional pumps are required. The retentate stream of the first stage is fed into the second stage. Whereas the permeate of the first stage is almost product-free, the permeate of the second stage contains, due to the moderate selectivity of the membrane, product and, consequently, is recycled to the feed (s. Fig. 4 ) . The simulation results for this two-stage RO process using an AspenPlus user-model for reverse osmosis are shown in Fig. 5 for several sets of total membrane areas and membrane area ratios A2/AtOt. By a simple splitting of the membrane area as indicated by the membrane area ratio A2/AtOt, and recycling the permeate of the second stage, the recovery rate can be increased between 15 and 30%. These considerable higher product recoveries can be useful, especially for concentrating high value products or components, which are thermally or mechanically sensitive. Naturally, the permeate recycling results in an increased feed flow rate to the first stage and, consequently, in higher energy consumption. Furthermore, the two-stage process requires a slightly higher membrane area for achieving retentate concentrations equal to the single-stage process. According to Fig. 5, the additional energy requirements of a two-stage process are lower than 50%. The additional energy consumption and the larger membrane area must be economically weighted against the higher profit due to the increased product recovery.

4 Example - Separation of Binary Organic Mixtures by a Two-Stage Reverse Osmosis Process

Reverse osmosis (RO) is a pressure-driven membrane process for the separation of homogeneous liquid mixtures. Established as an efficient process for sea- and brackish water desalination, it has been successfully employed for the separation of aqueous organic mixtures and for the treatment of industrial wastewaters containing a multitude of organic and inorganic components. With now available solvent-stable RO membranes and modules operating at pressure differences up to 200 bar, the application of reverse osmosis to the separation of nonaqueous organic solutions is intriguing. Experiments confirmed the possibility of a separation of polar components, such as methanol, from less polar components. As an example, the separation of a binary methanol/ tertiary butylalcohol mixture is considered (feedstream m , = 4.6 t/h, feed mole fraction xtert,but,= 0.01 ( = 2.28 wt.-Yo)). Due to polarity, molecular weight, and steric hindrance, methanol is the preferentially permeating component of the mixture. The separation characteristics for this mixture and the commercially available Desal 3 SC membrane were determined at ambient temperature and transmembrane pressure differences of Ap = 60, 120, and 180 bar. In the user-model, the solution/diffusion model for non-diluted binary solutions was used for describing the separation characteristics. This model is one of the five RO material transport models installed in the FORTRAN routine and has been proposed by Meares [5]. According to Meares, the molar flux of a permeating component i is:

5 Example - Processing of Azeotropic Mixtures with a Combination of Distillation and Pervaporation or Vapor Permeation
The dehydration of organic solvents by pervaporation can be considered as state-of-the-art. As a result of continuous further development, suitable membranes are now available for the separation of methanol from non-aqueous organic mixtures. With these membranes, a whole category of processes in the chemical and petrochemical industry is a possible candidate for the implementation of pervaporation or vapor permeation units.
Retentate
xR, tart But

(4)

with Oi = viAp/(R T ) for non-diluted organic solutions. Depending on the selectivity of the membrane, the required product quality, and product recovery, the process must be designed as a single- or a two-stage process. In principle, it is possible to concentrate the feed mixture in a single stage to any retentate concentration within the limits set by the pressure difference (up to 200 bar) and the osmotic pressure difference. Unfortunately, high retentate (product: tertiary butylalcohol) concentrations, when achieved in a single stage, are accompanied by serious product losses due to the limited selectivity of the membrane.

Reverse osmosis

Feed

Permeate Recycle

Figure 4. Two-stage reverse osmosis process designed as a stripping

section.

Chem. Eng. Technol. 19 (1996) 391 -397

395
DMC I MeOH,

c
DMC I MeOH,

- wAz
1

< wAz

DMC

MeoH

A1:ernalive a I b depending on reqbired MeOH quality

MeOH

MTBE I C4-HC I MeOH

"TRIM" process Debutanizer sidedraw

1 , 0 . . . . . . 4
U

single stage
' I .

.-.-.
50 mz
.

0
MTBE

Figure 6. Separation of azeotropic mixtures by combination of distillation and pervaporation/vapor permeation.

For obvious reasons, azeotrope-breaking is the most important category of separation problems for pervaporation and vapor permeation [6]. Conventional separation techniques require either the addition (and subsequent separation) of a suitable solvent or a pressure variation. In both cases, at least one additional column is required for the separation problem. In addition, the specific (primary) energy consumption is significantly higher compared to simple distillation.
A combination of distillation/pervaporation (or vapor permeation) is often superior. As shown in Fig. 6, two different cases of azeotrope-breaking must be distinguished: a combination of distillation/pervaporation where the pervaporation stage is designed for azeotrope-breaking and for guaranteeing the product specifications and combinations, where the task of the pervaporation stage is limited to azeotrope-breaking.

azeotrope-breaking alone. These basic considerations emphasize the necessity of detailed optimization calculations for every individual separation problem - for the hybrid process as well as for the in most cases possible conventional alternative. With respect to the overall optimum of the hybrid process, the following parameters are of primary interest: the interstage (transfer) concentrations, i.e., the concentrations between distillation and pervaporation/vapor permeation and vice versa. In case of treating column side streams by a membrane process, the point of extraction (tray number) and the extracted mass flow rate are important. Last but not least, one has to choose between pervaporation and vapor permeation. For these optimization problems AspenPlus is a very helpful tool since it allows a simultaneous simulation and optimization of all process stages. As already discussed in section 2, the quantitative description of material transport in pervaporation incorporated in the program is strictly empirical: The partial fluxes are determined in the range of interesting feed side concentrations from pervaporation experiments

Especially in cases where the azeotrope is located at medium concentrations only detailed design studies can answer the question, whether pervaporation should be designed for azeotrope-breaking and product purification or for

396 with the system membrane/mixture for an appropriate permeate pressure (for example 20 mbar) and an appropriate temperature (one additional experiment at a different temperature is required for the determination of the Arrhenius parameter). The temperature dependance is assumed to follow an Arrhenius-type relationship, the permeate pressure has to be identical in technical unit and in experiment. This implies, naturally, that permeate side module friction losses can be neglected (which is a reasonable assumption for optimized pervaporation modules). These data are loaded into the program and constitute the material transport model. Essential for pervaporation is the phase change of the permeate from liquid to vapor. This requires heat energy, which in technical pervaporation units is supplied by the sensible heat of the feed, resulting in a temperature decrease of the liquid feed along the process axis. Since permeate flux is a function of temperature, the simulation of pervaporation process requires the simultaneous solution of the differential material and energy balances. A second consequence is that the pervaporation process has to be designed as an alternating series of membrane units and feed reheaters. The permeate of all membrane units is collected and condensed at a specified pressure as discussed before. The necessary VLE-data and thermodynamic models can be obtained from the AspenPlus data bases which is a significant advantage compared to stand-alone simulation routines. The flow diagram of the separation sequence distillation/pervaporation for the separation of methanol (MeOH) from dimethylcarbonate (DMC) is shown in Fig.7. Depending on composition, the feed is either fed to the distillation. column or - in case of azeotropic composition - to the pervaporation unit. With the plasma polymerized membranes, permeate concentrations of about 95 wt.-Yo MeOH are obtained. This quality allows a recycling of the permeate to the DMC synthesis without further treatment [7].
DMC column ~ Perva Doration Feed, azeotropic

Chem. Eng. Technol. 19 (1996) 391 -397

75

70

g
c

65
0 VI
D

60

8
$

e a
55
m
a

50

10

20

30

40

50

60

Retentate concentration wR [wt -% MeOH]

Figure 8. DMC/MeOH hybrid process: influence of distillate and retentate concentration on processing costs (mFeed = 1000 kg/h, wFeed= 20 wt.-Yo MeOH).

The optimal separation sequence is determined by variation of the interstage concentrations and mass flow rates. Distillation (variation of reflux rate) and pervaporation (variation of the number of membrane/heat exchanger sections and permeate pressure) must be optimized with respect to specific processing costs for each set of interstage concentrations. In principle, a large number of combinations is feasible. The ultimate criterion for assessing their potential is the return of investment. In early design stages, this criterion is usually replaced by the criterion of minimal overall specific processing costs:
-

The calculations clearly indicate that the optimal configuration does not coincide with near-azeotropic compositions of the top product of the distillation column: The distillation column of the optimal hybrid process is in our case characterized by low reflux ratios and a low number of theoretical plates. Furthermore, the optimal process is characterized by a relatively small concentration difference between feed and retentate of the pervaporation unit.

Pat,

Heating steam,

The hybrid process was compared with a conventional and of course optimized - two-pressure distillation. The combination of pervaporation and distillation seems to be economically superior due to its low (primary) energy consumption.
6 Conclusions

99 wt.-% DMC

95 wt.-% MeOH

Figure 7. Separation of DMC/MeOH by distillation and pervaporation.

In this paper a tool is presented for the design, simulation, and optimization of hybrid processes containing internal recycles. The tool consists of the well-known simulation program Aspenplus, combined with a user-defined subroutine for calculation of the membrane processes reverse osmosis, pervaporation, and gas permeation. The advantages can be summarized as follows:

Chem. Eng. Technol. 19 (1996) 391 -397

397
D DMC distillate dimethylcarbonate feed hydrocarbon component indices methanol methyltertiarbutylether permeate retentate salt total water

of membrane processes with all other unit operations already implemented in AspenPlus is possible, - utilization of the physical property models and data bases of AspenPlus is possible, and - cost and sensitivity analyses can be performed.
- any combination
Received: August 29, 1995 [CET 7691

F
HC i ,j MeOH MTBE P R S tot W

Symbols used
a A A* B*
activity area membrane constant membrane constant mass flow mass flux molar flux pressure power permeability or flux density rejection coefficient gas constant molar volume mass concentration molar feed concentration molar permeate concentration ideal separation factor temperature osmotic pressure

References
[l] Rautenbach, R., Albrecht, R., Membrane Processes, John Wiley and Sons, New York 1989. [2] Peinemann, K. V., Ohlrogge, K., Wind, J., Industrial Applications of Membranes to Control VOC Emissions, in: Characterization and Control of Odours and VOC in the Process Industries (Vigneron, S., Hermia, J., Chaouki, J., Eds.), Elsevier, Amsterdam 1994. [3] Baker, R. W., et al., J. Membr. Sci. 31 (1987) pp. 259-271. [4] Ahlers, R., Rautenbach, R., The Recovery of Organic Vapors from Process Air with Gas Permeation Membranes, in: Characterization and Control of Odours and VOC in the Process Industries (Vigneron, S., Hermia, J . , Chaouki, J. Eds.), Elsevier, Amsterdam 1994. (51 Adam, W.J., Luke, B., Meares, P., J . Membr. Sci. 13 (1983) pp. 127- 149. [6] Rautenbach, R., Klatt, S., Vier, J., Proc. 6th Int. Conf. Pervaporation Processes in the Chemical Industry, OttawaIKanada (Bakish, R., Ed.), Bakish Materials Corp., Englewood 1992, pp. 2 - 15. [7] Shah, V.M., Bartels, C.R., Proc. 5th Int. Conf. Pervaporation Processes in the Chemical Industry, Heidelberg, (Bakish, R., Ed.), Bakish Materials Corp., Englewood 1991, pp. 331 -337.

m
ri" P

P R R
W

Y a

L9

Indices and Abbreviations

atm Az atmosphere azeotrope