You are on page 1of 8

Computers and Chemical Engineering 25 (2001) 693 700

www.elsevier.com/locate/compchemeng

Detailed mathematical modelling of membrane modules


J.I. Marriott, E. Srensen *, I.D.L. Bogle
Department of Chemical Engineering, Uni6ersity College London, Torrington Place, London, WC1E 7JE, UK
Received 10 May 2000; accepted 5 January 2001

Abstract
Membrane technology is used for a wide range of separations from particle-liquid separations to gaseous and liquidliquid
separations. In this paper, we introduce a detailed model that describes a general membrane separation. The model disregards
many common assumptions such as plug flow; constant temperature; constant pressure; binary mixture; steady-state conditions;
and constant physical properties. Our approach is applicable to any membrane separation and in this paper we demonstrate its
application to both liquid mixture separation (pervaporation) and gas separation in hollow-fibre modules. Both cases are seen to
exemplify the need for a detailed model. 2001 Elsevier Science Ltd. All rights reserved.
Keywords: Liquid liquid separation; Reynolds number; Hollow fibre

1. Nomenclature
A
D
E
F
fv
H
J
K
k
L
P
Q
R
r
T
t
U
V
6

Membrane area
Dispersion coefficient
Energy flux
Viscous and diffusive flux
Friction factor
Molar enthalpy
Molar mass flux through membrane
Molar kinetic energy
Thermal conductivity
Fibre length
Pressure
Energy flux through membrane
Fibre radius
Radial dimension
Temperature
Time
Molar internal energy
Volume
Velocity

* Corresponding author. Tel.: +44-20-76793802; fax: + 44-2073832348.


E-mail address: e.sorensen@ucl.ac.uk (E. Srensen).

z
z
zm
Subscripts
i
r
z

Axial direction
Molar density
Mass density
Component i
r-direction
z-direction

2. Introduction
During the last 30 years, the search for alternatives to
traditional energy intensive separation methods such as
distillation has led to the introduction of processes
based on membranes. Membrane processes often offer
cheaper capital and utility costs and can be used for a
wide range of separations: from particle-liquid separation, such as reverse osmosis, to gaseous and liquid
liquid separation. To achieve a desired separation, a
large membrane area is often required. In an industrial
plant, this area is supplied in a modular configuration
for which there are a number of variations. Historically,
plate-and-frame modules have been used, but more
recently, spiral-wound and hollow-fibre modules have
attracted a lot of interest as these offer much higher
packing densities (Krovvidi, Krovvali, Vemury, &
Khan, 1992).

0098-1354/01/$ - see front matter 2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 9 8 - 1 3 5 4 ( 0 1 ) 0 0 6 7 0 - 6

J.I. Marriott et al. / Computers and Chemical Engineering 25 (2001) 693700

694

In many areas of chemical process design and operation the use of mathematical models for process synthesis, optimisation and for control studies has shown
significant benefit. The application of these methods to
the design of membrane processes is now beginning to
be considered (El-Halwagi, 1992; Qi & Henson, 1998;
Marriott, Srensen, & Bogle, 1999). However, to minimise the technical risk that is inherent in the design of
any new process, it is essential that detailed unit models are used. Unfortunately, to the best of our knowledge, a detailed model of a general membrane
separation process is not currently available from published literature. Although a number of models do exist
(Krovvidi et al., 1992; Qi & Henson, 1997; Coker,
Freemann, & Fleming, 1998), due to the complex nature of flow through membrane modules, these usually
rely on a number of fixed assumptions. These include:
plug flow; constant temperature; constant pressure; bi-

Fig. 1. Flow pattern in a hollow-fibre module.


Table 1
Characteristics of the 1-D and 2-D flow sub-models
Model characteristics

One dimensional plug flow is assumed


Fibre side flow
Shell side flow
Non-isothermal flow
Viscous and dispersive flow mechanisms
Multicomponent separation
Physical properties provided externally
Steady-state and dynamic simulation

Sub-model
1-D

2-D

Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes

No
Yes
No
Yes
Yes
Yes
Yes
Yes

nary mixture; steady-state conditions; and constant


physical properties. As a result of these assumptions,
existing models are typically process specific and are
only valid within a limited operating range. However,
with better solution algorithms and the improved computer power that is now available, such assumptions
are no longer necessary.
The objectives of this paper are to describe a detailed approach to modelling a general membrane separation process and to illustrate its use for different
systems, in particular, pervaporation and gas separation.

3. Mathematical model
We have developed a consistent modelling approach
that can consider spiral-wound, plate-and-frame, tubular, and hollow-fibre membrane modules. However, in
this paper, we will demonstrate the approach on the
latter. Hollow-fibre modules contain a large number of
membrane fibres housed in a module shell. Feed can be
introduced on either the fibre or the shell side and
permeate is withdrawn in a co-current, counter-current,
or cross-current manner. The flow pattern in a hollowfibre module with fibre side feed and a counter-current
permeate stream is illustrated in Fig. 1.
To simulate a membrane module three sub-models
are required; two to describe the flow on either side of
the membrane and a third model which characterises
the separative properties of the membrane and any
porous support material. Two alternate flow sub-models have been developed: a one dimensional (1-D) flow
model and a two dimensional (2-D) flow model. Both
models are developed from rigorous dynamic mass,
momentum and energy balances, and thus are applicable to any membrane separation. The model characteristics are summarised in Table 1 and the main model
equations are given in the Appendix.
The 2-D sub-model formally describes axial variations in concentration, velocity, temperature and pressure and radial variations in velocity and concentration (concentration polarisation). However, whilst it
provides the most rigorous approach, the 2-D
sub-model is limited to flow inside the fibres. In
contrast, the 1-D (plug flow) sub-model is more versatile but an additional empirical parameter is required
to describe concentration polarisation (this is discussed
in more detail in Section 4.1). Both models disregard
common assumptions such as isothermal flow, binary
mixture, negligible axial diffusion, steady-state conditions, and constant physical properties (Table 1).
Therefore, it is apparent that model accuracy is
now mainly constrained by uncertainty in parameter
values, such as molecular diffusivity, and uncertainty in

J.I. Marriott et al. / Computers and Chemical Engineering 25 (2001) 693700

the module design, e.g. the non-uniformity of the


fibres.

3.1. Solution method


To form the complete model, the 1-D flow submodel which describes shell side conditions is coupled
(using an appropriate membrane sub-model) with either the 2-D sub-model or a second 1-D sub-model
describing fibre side conditions. The system of partial
differential and algebraic equations (PDAEs) that
make up the mathematical model is solved in this work
on an IBM RS6000 using the gPROMS simulation
software (Process Systems Enterprise Ltd, 1999). Fluid
properties are provided by Multiflash (Infochem Computer Services Ltd, 1996).
A spatial discretrisation technique is employed for
the solution of the distributed model. This converts the
system of PDAEs into a set of ordinary DAEs, which
can be solved by gPROMS using implicit numerical
integration techniques. A variety of spatial discretisation techniques are available, we have investigated the
use of both finite difference and orthogonal collocation
(on finite elements) methods. The latter approach is
preferred, as it uses the least number of equations and
thus minimises solution times. Typically, solution times
are 0.4 CPUs using the 1-D sub-model and 2 3 CPUs
using the 2-D sub-model.

3.2. Summary
A detailed model of a hollow fibre membrane module has been described in this section. The model is
applicable to any membrane separation and can be
used for a wide range of applications, from steadystate simulation and process synthesis, to investigating
control issues and dynamic optimisation. In future
work, we will address many of these issues using the
detailed model. However, in this work we have restricted our studies to the steady-state simulation1 of
two different membrane systems: a pervaporation separation and a gas separation.

4. Model application
We now investigate some of the features of the
detailed model described in the previous section. To
demonstrate the generality of the model, both liquid
mixture (pervaporation) and gas mixture separation are
considered.

1
Steady-state simulation is easily achieved using a dynamic model
by setting all the time derivatives to zero.

695

4.1. Per6aporation
Many of the liquid mixtures which need to be separated in the fine and speciality chemical industries
are either close-boiling or azeotropic in nature. Extractive distillation is a common technique for separating
such mixtures, however, it is energy intensive and requires the addition of a solvent to change the relative
volatilities of the various components. The solvent is
often toxic (e.g. benzene) and later must itself be separated from the mixture. These drawbacks fuel continuous research looking for new and better separation
techniques. The use of pervaporation as an alternative
to distillation for difficult separations has been suggested for a number of cases. In particular, the use of
pervaporation to separate organics from water is becoming increasingly important (Fleming and Slater,
1992).
In this process, the membrane forms a semi-permeable barrier between a liquid feed and a low pressure
gaseous product. Consequently, heat must be supplied
to evaporate the permeating material-typically resulting
in a feed stream temperature drop. However, in contrast to distillation, only the heat of evaporation for a
small fraction of the mixture, the permeating material,
must be provided. Thus, the energy requirements of a
pervaporation plant are much lower. Furthermore,
when compared to distillation, pervaporation offers a
number of additional advantages (Fleming and Slater,
1992), such as: lower capital costs, modular design
(easier retrofitting and de-bottlenecking), and often superior separations as it is not constrained by thermodynamic azeotropes.
In this section, we consider two important pervaporation systems: the removal of organics from wastewater and the dehydration of azeotropic ethanol/water
mixtures.

4.1.1. Remo6al of organics from wastewater


In pervaporation, like in many other membrane processes such as reverse osmosis and pervaporation, a
concentration gradient, or boundary layer, is formed
due to the slow rate of molecular diffusion to the
membrane surface. This effect, commonly termed concentration polarisation, is particularly apparent in the
flux of dilute organic material though pervaporation
membranes. In such cases, the resistance of the
boundary layer to the flux of organic material can be
significantly larger than that of the membrane, and
thus often controls the overall mass transfer rate (Cote
and Lipski, 1988). Clearly, therefore, the effect of concentration polarisation cannot be neglected.
Concentration polarisation is usually modelled using
a 1-D flow model coupled with an empirical mass
transfer coefficient (Brouchaert and Buckley, 1992).

696

J.I. Marriott et al. / Computers and Chemical Engineering 25 (2001) 693700

Fig. 2. Radial concentration profile (2-D model). Radial concentration profile assuming a stagnant boundary layer (1-D model).

Fig. 3. Mass transfer coefficient as a function of Reynolds number,


calculated and experimental results.

The development of a 2-D model enables this effect to


be described more rigorously, as it does not assume a
boundary layer and enables full radial variations in
both concentration and velocity. The difference between the two approaches is shown in Fig. 2a and b.
In separate works, Cote and Lipski (1988) and
Psaume, Aptel, Aurelle, Mora, and Bersillon (1988)
have described the removal of trichloroethylene from
wastewater using lab-scale silicone rubber hollow-fibre
modules. The purpose of their experimental work was

to investigate the importance of concentration polarisation in wastewater treatment by pervaporation. The


authors calculated the experimental mass transfer coefficient for a range of feed flows from the overall
trichloroethylene flux rate.
Here, we use the 1-D and 2-D flow models developed in this paper to simulate the liquid feed flow
through the fibre bore, and compare the results with
the experimental data from Cote and Lipski (1988) and
Psaume et al. (1988). The Wilke correlation (Bird,
Stewart, & Lightfoot, 1960) is used to estimate diffusion coefficients for both models. For the 1-D model,
the empirical mass transfer coefficient is estimated as a
function of Reynolds number using the Leveque correlation (Cote and Lipski, 1988).
The accuracy of the simulation results is assessed in
Fig. 3, which shows the mass transfer coefficient as a
function of Reynolds number. The average mass transfer coefficient is calculated from the ratio of permeate
flux to the mean feed-side concentration. Unfortunately, uncertainty in the value of the diffusion coefficient and in the experimental results, which show a
wide degree of scatter (Fig. 3), limits the scope of this
analysis. However, both models are seen to predict the
experimental results reasonably well and the predictions of the two models show a close correlation with
each other (Fig. 3). It can be concluded that, though
the 2-D flow model is computationally more expensive,
it provides a rigorous and more general approach to
describing concentration polarisation than the 1-D
model. Furthermore, an empirical mass transfer coefficient does not need to be estimated for the 2-D flow
model.

4.1.2. Ethanol dehydration


The separation of azeotropic ethanol water mixtures
is perhaps the most important example of pervaporation. We now consider the use of our general model to
describe this system.
A process for ethanol dehydration has previously
been detailed by Tsuyumoto, Teramoto, and Meares
(1997). The plant contains nine hollow fibre membrane
(a polyion complex) modules as well as a number of
heaters and vacuum compressors-further details can be
found by reference to their paper. Here, the detailed
model is applied to this case study using the membrane
characterisation supplied by Tsuyumoto et al. (1997).
The performance of a single module at two different
feed rates, as well as the performance of the whole
plant has been calculated. The results, which are given
in Table 2, show excellent agreement with the experimental data. The small error is most likely due to
uncertainty in the membrane characterisation
parameters.

J.I. Marriott et al. / Computers and Chemical Engineering 25 (2001) 693700

This case also highlights the importance of using a


detailed (non-isothermal) model. For example, the calculated temperature change for the single module (Case
A) is 26C. If an isothermal model were used, product
purity would be significantly overestimated to be
98.4wt% ethanol. Thus, isothermal flow should not be
assumed for pervaporation systems at the design stage
as it could lead to a significantly under-sized plant.

4.2. Gas separation


Membrane technology is commonly used to separate
gas mixtures; important examples of this include: hydrogen and helium recovery, air separation, and the
Table 2
Simulation and experimental results (ethanol dehydration)
Simulation

Feed flowrate (kg/h)


Feed concentration,
(wt% ethanol)
Feed temperature (C)
Permeate pressure (Pa)
Product concentrations
(wt% ethanol)
Calculated in this study
Experimental (Tsuyumoto
et al., 1997)

Single module

Plant

(nine modules)

248.5
96.8

100
94

60
400

60
13.3400

44.8
94
60
400

97.3
97.2

97.4
97.4

99.7
99.8

Table 3
Multicomponent gas separation system (Pan, 1986)a
Feed temperature (K)
Feed pressure (bar)
Permeate pressure (bar)
H2 permeability (1010 mol/m2 s
N2 permeability (1010 mol/m2 s
Ar permeability (1010 mol/m2 s
CH4 permeability (1010 mol/m2
a

Pa)
Pa)
Pa)
s Pa)

298
69.64
11.23
284
2.95
2.84
7.70

Hollow fibres: 0.33 m length, 80 mm i.d., 200 mm o.d.

Fig. 4. Permeate purity as a function of stage-cut, calculated and


experimental results.

697

removal of acid gases from light hydrocarbons. In this


isothermal process, the membrane forms a barrier between a high pressure feed gas and a low pressure
permeate gas. To aid in the design of gas separation
systems, Smith, Hall, Freemann, and Rautenbach
(1996) and Krovvidi et al. (1992) have recently proposed two approximate models that describe the separation of a binary gas mixture in a hollow-fibre module.
In this section, we evaluate the accuracy of these approximate models using the detailed model presented in
the previous chapter. First, however, we verify the
detailed model using experimental results for a well
established multicomponent system that has previously
been simulated by Pan (1986).

4.2.1. Multicomponent gas separation


For this study, the detailed model is used to describe
the removal of hydrogen from a multicomponent gas
mixture containing 51.78% hydrogen, 24.69% nitrogen,
19.57% methane and 3.96% argon. Module and operating details are given in Table 3. The calculated results
for this study are in excellent agreement with the experimental data (Pan, 1986) over the whole operation
range. This is seen in Fig. 4 which shows the hydrogen
purity in the permeate stream as a function of the
module stage-cut (the fraction of the feed stream passing through the membrane).
4.2.2. Comparison with approximate permeator models
Having verified the detailed model against experimental data we will now use this case study to illustrate
the potential problems with using approximate models
to describe membrane systems. Over recent years, a
number of approximate models have been proposed for
a range of different membrane systems. We will consider two approximate design models that have been
developed for gas separation by Smith et al. (1996) and
Krovvidi et al. (1992).
To accurately describe gas separation, mass and momentum balances must be calculated on both sides of
the membrane. The momentum balances are solved to
yield the pressure profile along the length of the module, and the mass balances to yield concentration profiles. Approximate models generally disregard the
momentum balance, and simplify the mass balances.
Smith et al. (1996) assume that the permeate concentration is constant along the entire length of the module,
whereas Krovvidi et al. (1992) assume a linear relationship between the feed-side and permeate-side concentrations-the operating line method 2.
2
Krovvidi, Krovvali, Vemury, and Khan (1992) have also described a second model. The authors report that though it does not
perform well for co-current operation, it is a better model for
counter-current operation. Unfortunately, there appear to be a number of mistakes in the equations given in their paper, and it was
impossible to reproduce the authors results.

J.I. Marriott et al. / Computers and Chemical Engineering 25 (2001) 693700

698

Table 4
Binary gas separation system (Smith et al., 1996)a
Case

Flow pattern
Feed composition (H2 fraction)
Feed temperature (K)
Feed pressure (bar)
Permeate pressure (bar)
H2 permeability (1010 mol/m2 s
Pa)
N2 permeability (1010 mol/m2 s
Pa)

Co-current
20%
308
137.9
1.379
109.3

Counter-current
85%
308
137.9
27.58
109.3

10.93

0.607

Hollow fibres: 0.76 m length, 100 mm i.d., 200 mm o.d.

Fig. 5. Calculated retentate purity (Case A).

Fig. 6. Calculated retentate purity (Case B).

Smith et al. (1996) use their model to describe the


separation of hydrogen and nitrogen in a shell fed
hollow-fibre module. This system will also be used in
this work. Two cases with different operating condi-

tions, membrane selectivities and feed compositions are


considered. For each study, the retentate purity and
the stage-cut are calculated for several feed flows. System
details are given in Table 4. Binary gas separation
system (Smith et al., 1996)
Simulation results from the two approximate models
for Case A are compared with the results from the
detailed model in Fig. 5. This shows the retentate
product purity as a function of the stage-cut. The
approximate model results are generally poor, showing
a significant deviation from the detailed model results
over the entire operating range. This error is primarily
due to the effect of neglecting the pressure profile,
when in fact, there is a significant pressure build up on
the fibre side.
The simulation results for Case B are given in Fig. 6.
In this case, retentate product purity is plotted as a
function of mass feed rate. The permeate pressure is
higher than in Case A, and as a result the relative
pressure drop is much lower. Consequently, the approximate model results show a good agreement (9 1
2%) with those from the detailed model, particularly at
high feed rates (i.e. low stage-cuts). This is because the
concentration profiles along the length of the module
are relatively constant. However, at low flowrates, the
Krovvidi et al. (1992) model predicts an infeasible
permeate concentration (i.e. greater than 100%), and
the results from the Smith et al. (1996) model diverge
significantly from the detailed model results (Fig. 6).
The reason for this is that the Smith et al. (1996)
model assumes constant permeate concentration along
the entire length of the module. This assumption is
particularly invalid at low flowrates, when there are
significant concentration changes along the length of
the module.
Even for this simple system (ideal gas and constant
membrane permeability), the approximate model results are unreliable, showing relatively poor agreement
with the detailed model. It is seen that only at high
permeate pressures and low stage-cuts do the approximate models provide a reasonable level of accuracy. In
contrast, hollow-fibre modules are usually connected in
parallel (Zolandz and Fleming, 1992), with high stagecuts, and low pressure operation will often be optimal.
However, perhaps the biggest disadvantage of the approximate models presented in the literature is that,
unlike the detailed model, they are limited to binary
separations.

5. Conclusions
In this paper, a detailed model that describes a
general membrane separation has been introduced. The

J.I. Marriott et al. / Computers and Chemical Engineering 25 (2001) 693700


Table 5
1-D flow sub-model

699

Appendix A. Appendix

Equations

Boundary
condition

Mass
balance

#zi #(F zi ) AJi


=
+
#t
V
#z
#z
F zi = zi 6zDi i
#z

Momentum
balance

#(z mwz )
#(z mw 2z ) #P 2fw
=

#t
#z
#z R

z (0, L]

Energy
balance

#(z(U+K))
#(Ez ) AQ
=
+
#t
#z
V

z (0, L)

z (0, L)

Ez = zwz (H+K)k

#(F zi )
=0
#z z = L

#T
=0
#z z = L

#T
#z

The main equations in the 1-D and 2-D flow submodels are presented in Table 5 and Table 6. For a full
derivation of the model equations please refer to Marriott (2001).The 2-D sub-model is derived assuming
radially symmetric laminar flow through a circular conduit. In addition, radial variations in temperature and
pressure are neglected. The 1-D sub-model assumes
plug flow through a conduit, for which pressure losses
are characterised using the frictional parameter f6.
For both flow sub-models the membrane flux variables, J and Q, are calculated by a coupled membrane
sub-model. This model characterises the separative
properties of the membrane. However, the model used
depends on the type of separation (several were used in
this work) and hence is not given here.

References
model is developed from rigorous mass, momentum
and energy balances and disregards many common
assumptions. The use of the model to investigate complex modelling issues in a liquid separation system has
been demonstrated. The model has also been used to
assess the performance of approximate models in a gas
separation system. In both these cases it is seen that a
detailed model is essential. In future work, we will
investigate the use of the model for the optimal design
of different membrane systems.

Table 6
2-D flow sub-model

Mass
balance

Momentum
balance

Equations

Boundary
condition

#zi #(F zi ) 1 #(rF ri )


=

#t
#z
r #r
#z
z
F i = zi wzDi i
#z
#zi
r
F i = zi wrDi
#r

Z  (0, L)
r (0, R)

#(z mwz )
#(z mw 2z ) #P
=

#t
#z
#z

z (0, L]
r (0, R)

Energy
balance

  

F ri r = R = JI

1 #(rz mwrwz )
#2w
1 # #wz
r
+ 2z
+v
r
#z
#r
r #r #r

#(z(U+K))
#(Ez ) AQ
=
+
#t
#z
V
Ez = zwz (H+K)k

#T
#z

#(F zi )
=0
#z z = L

z (0, L)

#wr
=0
#r r = R
wz r = R = 0

#T
=0
#z z = L

Bird, R. B., Stewart, W. E., & Lightfoot, E. N. (1960). Transport


Phenomena. New York: John Wiley.
Brouchaert, C. J., & Buckley, C. A (1992). Simulation of tubular
reverse osmosis. Water SA, 18 (3), 215 224.
Coker, D. T., Freemann, B. T., & Fleming, G. K. (1998). Modeling
multicomponent gas separation using hollow-fiber membrane contactors. American Institute of Chemical Engineering Journal, 44,
1289 1302.
Cote P. and Lipski C. (1988). Mass transfer limitations in pervaporation for water and wastewater treatment. In: R.Baksih. Proceedings of the 3rd international conference on pervaporation
processes in the chemical industry.
El-Halwagi, M. M. (1992). Synthesis of reverse osmosis networks for
waste reduction. American Institute of Chemical Engineering Journal, 38, 1185 1198.
Fleming, H. L., & Slater, C. S. (1992). Applications and economics of
pervaporation systems. In W. S. W. Ho, & K. K. Sirkar, Membrane Handbook (pp. 132 159). New York: Van Nostrand
Reinhold.
Infochem Computer Services Ltd (1996). Mulitflash, London.
Krovvidi, K. R., Krovvali, A. S., Vemury, S., & Khan, A. A. (1992).
Approximate solutions for gas permeators separating binary mixtures. Journal of Membrane Science, 66, 103 118.
Marriott J. I. (2001). Detailed modelling and optimal design of
membrane separation systems, PhD Thesis, University of London.
Marriott J. I., Srensen E. and Bogle I.D.L. (1999). Rigorous optimal
design of a pervaporation plant. FOCAPD Proceedings, Brekenridge, Colorado.
Pan, C. Y. (1986). Gas separation by high-flux, asymmetric hollow
fibre membrane. American Institute of Chemical Engineering Journal, 32 (12), 2020 2027.
Process Systems Enterprise Ltd. (1999). gPROMS Advanced Users
Guide, London.
Psaume, R., Aptel, P., Aurelle, Y., Mora, J. C., & Bersillon, J. I.
(1988). Pervaporation: Importance of concentration polarisation
in the extraction of trace organic from water. Journal of Membrane Science, 36, 373 384.
Qi, R., & Henson, M. A. (1997). Modelling of spiral-wound permeators for multicomponent gas separations. Industrial and Engineering Chemistry Research, 36, 2320 2331.
Qi, R., & Henson, M. A. (1998). Optimal design of spiral-wound
membrane networks for gas separations. Journal of Membrane
Science, 148, 71 89.

700

J.I. Marriott et al. / Computers and Chemical Engineering 25 (2001) 693700

Smith, S. W., Hall, C. K., Freemann, B. D., & Rautenbach, R.


(1996). Corrections for analytical gas-permeation models for separation of binary gas mixtures using membrane modules. Journal of
Membrane Science, 118, 289294.
Tsuyumoto, M., Teramoto, A., & Meares, P. (1997). Dehydration of

ethanol on a pilot plant scale using a new type of hollow-fiber


membrane. Journal of Membrane Science, 133, 83 94.
Zolandz, R. R., & Fleming, G. K (1992). Design of gas permeation
systems. In W. S. W. Ho, & K. K. Sirkar, Membrane Handbook
(pp. 54 77). New York: Van Nostrand Reinhold.

You might also like