SEPARATIONS: MATERIALS, DEVICES AND PROCESSES

Simulation of Gas Separation Using Partial Element Stage

Cut Modeling of Hollow Fiber Membrane Modules
Sina Gilassi , Seyed Mohammad Taghavi, Denis Rodrigue, and Serge Kaliaguine
Dept. of Chemical Engineering, Universite Laval, Quebec, QC, Canada G1V 0A6

DOI 10.1002/aic.16044
Published online in Wiley Online Library (wileyonlinelibrary.com)

A mathematical model is developed to simulate a gas separation process using a hollow fiber membrane module. In par-
ticular, a new numerical technique is introduced based on flash calculation. Such analysis allows identifying the
required membrane properties needed to reach module performance of interest. This model is validated for six different
gas separation cases taken from literature. Then, the validated model is used to investigate the effect of O2 and N2 per-
meances on O2 recovery and O2 mole fraction in the permeate stream. A realistic two-stage air enrichment process is
also proposed for O2 production using an industrial module with different fibers numbers. Moreover, this model is used
to simulate a natural gas purification process using a single unit to determine the required membrane separation area
and CH4 loss. Finally, a two-stage process is proposed to equally enhance CH4 retentate mole fraction and decrease
CH4 loss. VC 2017 American Institute of Chemical Engineers AIChE J, 00: 000–000, 2017

Keywords: gas separation, hollow fiber membrane, heat- and mass-transfer analysis, natural gas purification, O2
production

Introduction H2), and bio-methane separation from syngas are other exam-
ples of industrial gas treatment processes. The current methods
Based on current data, a substantial growth in global energy
for gas separation are absorption, adsorption, and cryogenic dis-
consumption is predicted for the coming years.1 The world
tillation.4,5 Over the last 30 years, numerous attempts have been
population is about 7.5 billion in 2017, and this number is
made to replace these conventional methods suffering from
expected to increase to around 9.2 billion by 2050.2 This popu- operation instability, high energy consumption, as well as high
lation growth inevitably poses serious challenges to the global capital and maintenance costs. Today, the main option is mem-
energy needs. At present, a large amount of energy is supplied brane technology.6–10
from fossil sources which unwillingly forces the governments Gas separation using membrane processes requires a large
to deal with controversial challenges such as greenhouse gas membrane surface area for high gas capacity. Presently, three
(GHG) emission. The reduction of GHG emissions demands types of membrane contactors including hollow fiber, spiral
strong and global collaboration. Kyoto protocol (1992) and wound, and envelope are commonly used. The hollow fiber
Paris agreement (2016) based on United Nations Framework and spiral wound membranes provide larger surface area than
Convention on Climate Change hopefully determined essential the envelope. Efficient gas separation processes require, how-
plans to mitigate and control these GHG emissions. The Cana- ever, an active area from hundreds to thousands of square
dian government has also undertaken CO2 mitigation plans to meters. Hollow fiber membrane modules (HFMM) are more
honor the Canadian’s conditional commitment. In this way, economical as they have the highest effective surface area per
the GHG emissions’ intensity will decline to 17% below the unit volume of the membrane module. HFMM consists of
2005 levels by 2020 that is equivalent to 622 Mt of CO2.3 large numbers of thin tubular fibers bundled and sealed
Gas separation is one of the main processes in chemical and together on each end with epoxy in a housing.11 HFMM oper-
industrial plants. For example, in the field of natural gas trans- ating at low-pressure, are commonly used for gas separation
portation and treatment, gas composition needs to meet some compared to other module types. In the gas separation process,
criteria to avoid instability in operational conditions and decline the gas mixture is fed to the HFMM, the penetrants pass
in process performance. In terms of pipeline system, the content through the polymer membrane depending on their permselec-
of acidic gases needs to be carefully controlled because at high tivity and desorb on the permeate side whereas nonpenetrants
concentration they can corrode the network. Flue gas separation remain on the residue side and finally leave the module. The
(CO2/N2), air enrichment (O2/N2), hydrogen separation (CO2/ gas mixture can enter tube-side or shell-side of the module
depending on the separation objectives. In terms of flow con-
Additional Supporting Information may be found in the online version of this
article. figuration, three HFMM including countercurrent, cocurrent,
and cross-flow can also be used for gas separation. Other
Correspondence concerning this article should be addressed to S. Kaliaguine at parameters, such as number of fibers, membrane material,
[email protected].
module size, as well as feed pressure, temperature, and flow
C 2017 American Institute of Chemical Engineers
V rate can directly affect the separation efficiency.

AIChE Journal 2017 Vol. 00, No. 00 1

 The modeling of membrane gas separation systems is of algebraic equations for the permeate stream was solved with a
importance in engineering and design thereby both process Powell hybrid algorithm and finite-difference approximation
efficiency and cost analysis can accurately be estimated. A for the derivatives. The bulk mole fraction and flow rate of the
robust model developed on the basis of heat and mass transfer permeate and retentate streams were later calculated from the
can be used as a reliable tool to initiate a feasibility study of a known permeate local mole fraction, and this procedure was
separation project. In this work, a new approach based on flash repeated until the end of the module. Khalilpour et al.25 also
equilibrium calculation, which is used to calculate liquid- presented a new solution technique for the model developed
vapor concentration in a distillation system, is modified to find by Pan.22 They converted all mass balance equations (ordinary
a solution for heat- and mass-transfer equations inside a hol- differential equation, ODE, system) for cocurrent and counter-
low fiber membrane system. A mathematical model is pro- current flows to derive algebraic equations using backward
posed to determine the flow rate and composition of both finite differential equations over all elements and then solved
permeate and retentate streams for different module configura- them by Gauss-Seidel algorithms. This method had a major
tions (countercurrent, cocurrent, and cross-flow). Furthermore, difficulty in setting initial values for the residue flow rate and
a highly efficient new numerical technique is introduced for permeate pressure. The iteration method was slow because
the membrane gas separation systems. The modeling results this numerical algorithm had an extra loop for pressure calcu-
are validated with various experimental data obtained from lit- lations. More details are also available in reference.26 Kundu
erature and to determine the model’s performance. The model et al.27 introduced a new solution technique to solve the bal-
is also allowing to be useful in establishing target performance ance equations. They transformed the model equations to
in the development of new membrane materials. Finally, the ODE systems as an initial-value problem in two successive
validated model is used to simulate air enrichment (O2/N2 sep- steps using Gear’s Backward Differentiation Formula (BDF)
aration) and natural gas purification (CO2/CH4) processes method. This technique required minimum computational
using an asymmetric polymer membrane in single- and two- time and effort, and presented better solution stability for a
stage units. multicomponent gas membrane separation system.
Thundyil and Koros28 presented and analyzed a new model-
Modeling Background ing approach to solve the mass-transfer equations in the
Numerous models with various limitations were proposed HFMM. The numerical algorithm was improved by the suc-
for HFMM simulation under different operating condi- cession of states method to separate the module into small size
tions.12–17 For instance, Bansal et al.18 introduced a numerical elements. This model was suitable for gas separation under
method for the separation of multicomponent gas mixtures in isothermal conditions. The results revealed that the cross-flow
a single membrane unit. The model equations for a cocurrent pattern was more effective than the other patterns due to a bet-
system constituted an initial boundary problem solved by a ter feed distribution inside the module whereas the countercur-
Runge-Kutta method. The flow rates and mole fractions of the rent pattern was more suitable for the module with larger
permeate and retentate streams were also calculated by an iter- bundle size. Later, Ahmad et al.29 modified the model devel-
ation method. The results showed that this numerical algo- oped by Thundyil and Koros28 to show the effect of tempera-
rithm was slow for multicomponent gas systems. More ture and pressure on membrane permeation. The results
theoretical studies of the performance of single-stage perme- showed that increasing the feed CO2 concentration for CO2/
ation showing the effects of pressure, membrane area, and CH4 separation, increased the temperature difference between
flow patterns are also available in reference.19,20 Shindo the feed and residue end, but they did not report any data about
et al.21 developed a model based on Fick’s law for the single- permeate temperature. Under nonisothermal conditions, the
stage permeation of a multicomponent gas mixture with differ- temperature of the feed and permeate streams noticeably
ent flow patterns. In their study, they did not consider the changes because of the Joule-Thomson (JT) effect and mem-
effect of temperature on gas components’ permeability and brane conductivity. The JT effect is also attributed to the high
pressure variation in the feed and permeate sides. Later, Pan22 ratio of transmembrane pressure between the permeate and
developed a model of multicomponent permeation systems for retentate sides. More details about the numerical procedure to
high-flux hollow fiber membranes in which the pressure varia- calculate the JT coefficient are available elsewhere.30,31
tion along the fiber was also taken into consideration. The Coker et al.32 presented a new model of multicomponent
solution method consists in the calculation of local mole frac- gas separation using a hollow fiber membrane for the various
tions of the permeate stream by an iteration method. The bulk types of flow pattern. The model was compatible with any
mole fraction was calculated from the mass balance using the change in pressure sweep, permeability, and pressure gradient
compositions and flow rates of the feed and residue streams. on both sides of the membrane. The module was divided into
The pressure drop inside the fibers was calculated using the a number of segments and the mass balance was applied to
Hagen-Poiseuille equation with a trial-and-error shooting each one. The solution method to solve a system of the alge-
method. In terms of pressure variation, Lim et al.23 developed braic equation was, however, complicated as initial guesses
a model to accurately estimate the pressure drop inside the needed to be made for the flow rate and pressure on each seg-
fibers. The advantage of their improved Hagen-Poiseuille ment. Later, Coker et al.33 modified the previous model to
model was computation simplicity when the fiber permeability exhibit the effect of gas expansion on heating or cooling inside
and gas compressibility changed. Murad Chowdhury et al.24 the membrane permeator. The results showed that increasing
presented a new numerical method for the model developed either CO2 concentration in the feed gas or stage cut increased
by Pan,22 which could be incorporated in process simulators the temperature difference between the feed and residue sides.
such as Aspen Plus. They reported that the equations were They reported that the numerical solution based on nested suc-
solved faster by their numerical technique without initial guess cessive substitution method could also be unstable due to the
on the pressure, flow rate, and mole fraction inside the fiber. In limited radius of convergence. Lock et al.34 developed a new
terms of the numerical solution, a system of nonlinear solution technique for the model developed by Thundyil and

2 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
Koros.28 The calculation method started with an initial guess
for the more permeable gas component to calculate an initial
stage cut. The flow rate and composition of the permeate and
retentate streams were later calculated for this initial stage cut.
The proposed method was slow because an iteration method
was needed to numerically calculate the permeate local mole
fraction at each node. Moreover, this method was only suitable
for isothermal conditions.
Scholz et al.35 studied the effect of concentration polariza-
tion, JT effect, pressure losses, and real gas behavior for the
simulation of membrane separation systems. They coded the
modeling algorithm in Aspen Custom Modeler and reported
acceptable performance and high potential of their model for
gas separation. Later, Hosseini et al.36 modified the model pre-
sented by Kundu et al.27 They also presented a comprehensive Figure 1. Schematic diagram of a cocurrent flow mem-
model for nonideal conditions in which real gas behavior, tem- brane separation module, and the feed and
perature, pressure, and concentration polarization effects were permeate flows in the first and two succes-
taken into account. sive elements.
The HFMM can also be used for another type of gas separa-
tion system when an absorbent flows inside the fiber. In this Modeling algorithm
case, the membrane is seen as a physical barrier between the
liquid and gas phases and prevents liquid penetration to the The numerical solution is based on the “succession of
shell side. In the case of CO2, penetrants which pass through states” method. The module is discretized into a large number
the membrane react with an absorbent such as an alkanol- of independent finite elements in which the mass-transfer driv-
amine solution. Later, the permeate stream rich in CO2 is sent ing force is constant. For the first element, the computation
to a regeneration unit and then recirculated to the membrane starts from the inlet with initial feed conditions (flow rate,
separation unit. A mathematical model can also be developed mole fraction, pressure, and temperature) to the outlet (residue
based on continuity, mass, and momentum equations for the end). The outlet conditions of the first element are selected as
shell, membrane, and fiber sides, and a numerical software the known inlet variables for the next element to compute the
such as COMSOL is allowing to find accurate solutions for mass and energy balances, and pressure variation along the
this system. More details about the equations and model devel- module. The same solution procedure is repeated until the resi-
opment are also available in references.37–46 due end. The packing fraction (u), defined as the ratio of the
fibers cross-sectional area to the cross-sectional area of the
Model Development fiber module, is defined as
A new numerical technique is introduced here for the do 2
u512E5nf (1)
modeling of HFMM for gas separation to calculate the flow dm 2
rate and mole fraction in permeate and retentate streams. In
this modeling approach, a system of ODE for a binary feed where E and nf are the void fraction and fiber number in the
gas contains six dependent equations to be solved simulta- module, respectively. do and dm are the outer fiber diameter and
neously using ODE solution methods like Runge-Kutta or inner module diameter. A multicomponent feed gas with a flow
finite difference method. Increasing the number of feed gas rate of Vf and mole fraction of xf is fed to the HFMM. Based on
components not only increases the computation time but also the numerical method used, the permeate and retentate flow
decreases the efficiency of the numerical technique for a mul- rates (Vp, Vs) and mole fractions (yp, xs) are calculated with the
tistage separation system. The new numerical technique pre- known variables of each element. The driving force is the pres-
sented in this work introduces the concept of partial element sure difference between the shell (Ps) and fiber (Pt) sides.
stage cut which has similarity with K-value used in flash cal-
Cocurrent flow (shell side feed) configuration
culation of distillation column. This approach greatly reduces
the computation time and improves the modeling result In the cocurrent flow pattern, the feed gas flows in the axial
efficiency. direction and parallel to the fiber bundle while the penetrant
gases diffuse through the membrane and finally leave the mod-
Model assumption ule. Figure 1 shows a schematic diagram of a cocurrent flow
The assumptions for the model development are as follows: membrane separation module and, the feed and permeate
1. The HFMM operates at steady state under isothermal flows in the first and two successive elements.
or nonisothermal conditions. The volume and surface area for each element are calcu-
2. The permeability is independent of pressure and gas lated as
composition.
3. The fibers do not deform during operation at high Dv5prm2 DL (2)
pressure. 2
4prm ð12EÞDL
4. Polarization at the membrane surface is negligible. DAf 5 (3)
do
5. Pressure drop inside the fibers is calculated using the
Hagen-Poiseuille equation. where rm and DL stand for the module inner radius and ele-
6. The gas flow is laminar, and ideal gas behavior is ment length, respectively. Based on the solution method, the
considered. mass balance equations for the first element can be given by

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 3
 X
m wt ð1Þ5f ðxf ;c ð0Þ; Ps ð1Þ; Pt ð1Þt ; Qc ; DAf ; Vp ð1ÞÞ (18)
DVt ðiÞ5 DVc ðiÞ (4)
c51 In terms of the numerical solution, Eq. 17 can be solved by
DVc ðiÞ5DAf Qc ½Ps xs;c ði21Þ2Pt yp;c ðiÞ iteration via Bisection, Newton-Raphson, and Brent’s meth-
ods. Consequently, the stage cut, permeate composition, and
c51; 2; . . . ; m (5) retentate flow rate of the first element can be calculated by the
i51; 2; . . . ; n numerical algorithm. The retentate composition is then
obtained from a mass balance calculation over the element via
where m and n are the numbers of components and elements, xf ;c ð0ÞVf ð0Þ5xs;c ð1ÞVs ð1Þ2DVt ð1Þ (19)
respectively
Vs ðiÞ5Vs ði21Þ2DVt ðiÞ (6) For the element beside the first element, the permeate compo-
sition and flow rate are also dependent on the permeate flow
Vp ðiÞ5Vp ði21Þ1DVt ðiÞ (7)
rate of the previous element. Similarly, a new equation can be
where DVt and DVc represent the total and individual local per- derived for the other elements as
meate molar flow rates through the i element, respectively. wði; cÞVs ði21Þ2yp;c ði21ÞVp ði21Þ
The exact number of elements is determined after the mesh (20)
5Qc DAf ðPs xs;c ði21Þ2Pt yp;c ðiÞÞ
size analysis. For the first element, index (i – 1) represents
y ði21ÞV ði21Þ
the feed conditions (Vf, xf) whereas the permeate flow rate, Ps xs;c ði21Þ1 p;c Qc DApf
Vp(i – 1), is 0. The stage cut, which is the ratio of permeate to wði; cÞ5   (21)
1 Pt ðiÞ
feed flow rate in a single membrane module, is defined as Vs ði21Þ Qc DA f
1 Vp ðiÞ

Vp Thus, Eq. 21 can be written for all components and then

h5 (8)
Vf substituted into Eq. 14 to give
Analogous to the vapor-liquid separation in a distillation col- yp;1 ði21ÞVp ði21Þ
umn, a new parameter called the partial element stage cut, Ps xs;1 ði21Þ1
Q1 DAf
which determines the permeation efficiency of an individual gas  
1 Pt ðiÞ
component in each element, is introduced in a manner similar 1
Q1 DAf Vp ðiÞ
to a separation equilibrium constant similar to a K-value as
yp;c ðiÞVp ðiÞ yp;2 ði21ÞVp ði21Þ
wði; cÞ5 (9) Ps xs;2 ði21Þ1
Vs ði21Þ Q2 DAf
1   1... (22)
1 Pt ðiÞ
where Vp(i) stands for the net flow rate of permeate stream in 1
Q2 DAf Vp ðiÞ
the fiber leaving element i. The sum of partial element stage
cuts for all components in each element is equivalent to a local yp;m ði21ÞVp ði21Þ
Ps xs;m ði21Þ1
element stage cut (wt) and defined for element i as Qm DAf
1   5Vp ðiÞ
1 Pt ðiÞ
wt ðiÞ5wði; 1Þ1wði; 2Þ1 . . . 1wði; mÞ (10) 1
Qm DAf Vp ðiÞ
yp;1 Vp ðiÞ yp;2 Vp ðiÞ yp;m Vp ðiÞ
wt ðiÞ5 1 1...1 (11)
Vs ði21Þ Vs ði21Þ Vs ði21Þ This new MFEE equation is also a function of the previous
element conditions and simply written as
Vp ðiÞ
wt ðiÞ5ðyp;1 1yp;2 1 . . . 1yp;m Þ (12) wt ðiÞ5f ðxx;c ði21Þ; Ps ðiÞ; Pt ðiÞ; Qc ; DAf ;
Vs ði21Þ (23)
Vp ðiÞ; yp;c ði21Þ; Vp;c ði21ÞÞ
yp;t ðiÞ5yp;1 ðiÞ1yp;2 ðiÞ1 . . . 1yp;m ðiÞ51 (13)
Vp ðiÞ The same numerical approach is used to solve Eq. 22, to calcu-
wt ðiÞ5 (14) late Vp(i) and other unknown variables.
Vs ði21Þ
The pressure variation in the shell and fiber is calculated
The partial element stage cut in Eq. 9 is then substituted into using the Hagen-Poiseuille equation as
Eq. 5 yielding a new equation for the c components of the first  
dðPt 2 Þ 25:6Rg Tgmix Vp
element as 5 (24)
dL pdi4 nf
wð1; cÞVs ð0Þ5DAf Qc ðPs xf ;c ð0Þ2Pt yp;c ð1ÞÞ (15)
dPs q
52k c (25)
P x ð0Þ dL 2dhyd
wð1; cÞ5  s f ;c  (16)
1
Vs ð0Þ Qc DA f
1 VPpt ð1Þ
ð1Þ where Pt, R, T, and gmix are the fiber pressure, universal gas
Similarly, Eq. 16 can be written for all components and later constant, gas temperature, and mixture viscosity, respectively.
substituted into Eq. 14 to give Ps, dhyd, k, q, and c represent the shell pressure, hydraulic
diameter, friction factor, density, and gas velocity, respec-
Ps xf ;1 ð0Þ Ps xf ;2 ð0Þ Ps xf ;m ð0Þ tively. Since a laminar flow is assumed in both shell and fiber
 1 1...1  5Vp ð1Þ
1
Q1 DAf 1 VPpt ð1Þ
ð1Þ
1
Q2 DAf 1 VPpt ð1Þ
ð1Þ
1
Qm DAf 1 VPpt ð1Þ
ð1Þ
sides, the friction factor is inversely proportional to the Reyn-
olds number and expressed as
(17)
2Gs do
Equation 17 is called the membrane flash equilibrium equation Re5 (26)
gmix
(MFEE) which can be written as

4 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
 Table 1. The Properties of the Experimental Hollow Fiber Modules and Feed Conditions
Parameters Unit Case I48 Case II22 Case III47 Case IV51 Case V49 Case VI22
Number of fiber (nf) – 70 20 100 270 368 20
Fiber outer diameter (do) lm 230 200 735 156 160 200
Fiber inner diameter (di) lm 84 80 389 63 80 80
Active length (L) cm 63.8 15 15 26 25 15
Number of elements – 128 30 30 52 50 30
Module inner diameter (dm) cm 0.48 2.1 1 0.5 0.95 0.21
Packing fraction (u) – 0.16 0.18 0.54 0.26 0.10 0.18
Temperature (T) K 298 298 301 303 301 298
Feed pressure (Ps) kPa 404 6964 405.3 1570 690 3520
Permeate pressure (Pt) kPa 101 1123 101.3 101.3 100 92.80
Feed composition (xf) % CO2 40 H2 51.78 CO2 60 CO2 50 O2 20.5 CO2 48.5
N2 60 N2 24.69 CH4 40 O2 10.5 N2 79.5 CH4 27.9
CH4 19.57 N2 39.5 C2H6 16.26
Ar 3.96 C3H8 7.34
Permeance (Qn) GPU CO2 63.6 H2 284 CO2 31.6 CO2 204.20 O2 9.30 CO2 40.04
N2 3.05 N2 2.95 CH4 8.81 O2 60.20 N2 1.80 CH4 1.11
CH4 2.84 N2 13.10 C2H6 0.30
Ar 7.70 C3H8 0.05

64 Pan22 conducted an experiment to separate a gas mixture

k5 (27)
Re (H2, N2, CH4, and Ar) using a high-flux asymmetric CA mem-
brane. Figure 2b shows a comparison between the modeling
where Gs is the mass flow rate per unit cross-sectional area of results of Case II with the experimental data. The high value
the module. According to the process design data, the gas of H2 permeance (284 GPU) compared to the other gases
velocity is set between 1.5 and 1.7 m/s on the shell side and makes this membrane very selective for the separation of H2.
between 0.01 and 0.38 m/s on the fiber side.35 The fiber pres- As seen, increasing the stage cut up to 40% does not have a
sure in the first element is unknown and an iteration method is very significant effect on H2 mole fraction in the permeate
needed to calculate the pressure profile along the module stream. A moderate stage cut of 50% still provides high H2
length. Wike’s equation is also chosen to calculate the gas- mole fraction (0.92%) which is appropriate for a single-
mixture viscosity with an average error of 2%28 stage separation unit.
X
m
x a ga Tranchino et al.47 used a composite membrane with a support
gmix 5 X (28) polymer (polysulfone) and an aliphatic copolymer coating to
a51 xa 1 x r
b b ab separate CO2 and CH4 under different operating conditions.
2 !0:5   32 Figure 3a shows a comparison between the modeling of Case
 
1 Ma 20:5 4 ga Mb 0:25 5 III with experimental data at two different temperatures. As
rab 5 pffiffiffi 11 11 (29) shown, a higher feed temperature produces lower CO2 mole
8 Mb gb Ma
fraction in the permeate stream while the selectivity of CO2/
CH4 decreases with increasing temperature. The nonisothermal
where a and b are the gas component indices. r and M are
Wike’s coefficient and molecular weight, respectively.
The mass balance equations for cross-flow and countercur-
rent flow configurations are also available in the Supporting
Information.

Model Validation
The model combined with the proposed numerical proce-
dure was used to predict the separation performance of six dif-
ferent gas mixtures under different operating conditions and
the results are compared with experimental data reported in
the literature.22,47–50 Table 1 presents the module characteris-
tics and feed conditions for the modeling cases.
An asymmetric cellulose acetate (CA) membrane was used
in Case I to separate a binary gas mixture of CO2 and N2. Fig-
ure 2a shows a comparison between the modeling results and
experimental data.48 As shown, the highest value of CO2 mole
fraction in the permeate (0.85%) is observed for the lowest
value of stage cut (0.1). As a general rule, increasing the stage
cut results in a reduction of the CO2 mole fraction in the per-
meate stream. In terms of modeling, the stage cut value can be Figure 2. Comparison of modeling results with the
increased by two different scenarios: (a) increasing fiber’s experimental data of: (a) Sidhoum et al.48
length and (b) reducing feed flow rate. and (b) Pan.22

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 5
 Figure 5. Comparison between Case VI modeling
results and the experimental data of Pan.22
Figure 3. Comparison of modeling results with the
experimental data of Tranchino et al.47: (a) at
slowly decreases from 0.85 to 0.70 up to a stage cut as high as
two different feed temperatures (25 and
0.7 for a single-stage separation unit. This indicates the high
658C) and (b) at two different feed pressures
potential use of the selected membrane in a multistage separa-
(200 and 600 kPa).
tion system.
Feng et al.49 conducted an experiment to separate O2/N2
option of the model is used to calculate the permeability of CO2 mixtures with asymmetric hollow fiber membranes. Figure 4b
and CH4 along the fiber. In this case, a good agreement is also shows the Case V modeling results which fitted the experi-
seen between the modeling result and experimental data at mental data for both cocurrent and countercurrent configura-
658C. However, a deviation of 5% between both curves can be tions. The N2 mole fraction in the retentate side shows no
considered negligible when the CO2 mole fraction and stage cut significant difference between the two modes over the lowest
values are high and moderate, respectively. The model’s perfor- range of stage cut values. When the stage cut increases, how-
mance is also investigated for the cases in which the transmem- ever, from 0.4 to 0.9, the conditions become more interesting
brane pressure changes moderately. Figure 3b shows the effect for production of nitrogen at low oxygen content. Similarly,
of pressure change from 200 to 600 kPa on CO2 mole fraction. the modeling results show that the separation performance for
The modeling results show that CO2 mole fraction increases the countercurrent configuration is better than the one of the
with increasing feed pressure due to the higher separation driv- cocurrent configuration, especially when the separation pro-
ing force. cess is designed for high stage cut ranges.
Sada et al.51 also investigated the performance of an asym- Pan22 also carried out another experiment to illustrate the
metric cellulose triacetate membrane for the separation of good potential of asymmetric CA membranes for CO2 separa-
CO2-air mixtures at 308C using a HFMM. Figure 4a shows a tion from a hydrocarbon mixture. Figure 5 shows a compari-
comparison between the Case IV modeling results and experi- son between the Case VI modeling results and experimental
mental data. The CO2 mole fraction on the permeate side data. High CO2 mole fraction (>90%) can be achieved in the
permeate side for stage cut values of 30–50%. The modeling
results also show that increasing the stage cut value results in
a reduction of the quality of the permeated stream. More inter-
estingly, CH4 mole fraction lost in the permeate side is in an
acceptable range (2–6%) when the stage cut value is around
40–50%. According to these results, CA membranes have
good potential for natural gas separation.
As shown above, the new modeling approach was used to
predict the separation performance of six different experimen-
tal systems. In all cases, the modeling results were in good
agreement with the experimental data. It is expected that the
model proposed combined with the new numerical technique
is robust and reliable for the modeling and simulation of gas
separation modules and two typical examples are presented
next.

Case Study: Oxygen Enrichment

Air separation can be carried out by several processes such as
Figure 4. Comparison between modeling results and cryogenic distillation, pressure swing adsorption, and mem-
the experimental data of: (a) Sada et al.51 brane separation in which the main target is to enrich air in
and (b) Feng et al.49 either nitrogen or oxygen.52–55 Oxygen enriched air production

6 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
 Table 2. Parameters Used for the Modeling of the O2 per module is set to 3 3 105 and 6 3 105, respectively. Refer-
Production Unit ring to the aforementioned ranges of gas velocity, the feed
flow rates needed for the O2 production of 50,000 ft3/day are
Parameters Unit Value
different for both cases so that an intrinsic stage cut can be
Feed pressure kPa 600 defined for a primary analysis of separation performance. The
Feed temperature K 296
Fiber length cm 183
modeling results show that an increase in the number of fibers
Shell diameter cm 20.5 results in a reduction in O2 mole fraction in the permeate side
Fiber outer diameter mm 160 and an increase in the stage cut. A comparison between Figure
Fiber inner diameter mm 90 6 shows that the average permeated flow rates are 2600 and
Fibers number – 3 3 105, 6 3 105 5000 cm3/s for 3 3 105 and 6 3 105 fibers, respectively.
Packing density % 48
Packing fraction – 0.19 The modeling results show that the average O2 mole frac-
tion on the permeate side is about 40% for both cases under
the same operating conditions that is below the separation tar-
with different purities is used in various chemical applications get (60%). As an alternative, an extra membrane gas separa-
such as combustion enhancement of natural gas and coal gasifi- tion unit can be added to improve O2 purity. In this case, the
cation. A membrane separation system may be used to produce same membrane modules with 3 3 105 and 6 3 105 fibers are
oxygen with lower energy consumption compared to conven- used in the second unit. In the case of designing a two-stage
tional methods. Membranes with O2/N2 selectivities above 2 gas separation system, the inlet flow rate of the module is cal-
are sufficient to produce 99% pure N2, but N2 recovery is low culated based on the gas velocity inside the module, number
which consequently, would impose an extra cost for gas com- of fibers, and number of modules in each unit. In the case of
pression.56 On the contrary, O2 production strongly depends on N2 production, as air already has 79% N2, the residue stream
the membrane O2/N2 selectivity. This separation is more diffi- usually contains N2 in higher purity. Thus, the production of
cult as the feed gas (air) only contains 21% O2 and a large pure N2 using a single-stage separation unit is easier. On the
amount of N2 must be removed. Therefore, the number of mem- contrary, the design of an O2 production unit is more difficult
brane modules needed is high to produce a high volume of oxy- as N2 can also permeate along with O2, limiting O2 purity in
gen with a desired purity in the permeate stream. Future the permeate stream. Referring to the selected membrane
improvement in membrane selectivity may result in a solution properties, high O2/N2 selectivity results in an increase in the
for this problem. permeate flow rate. In the techno-economic analysis, it is
Here, the proposed model is used to design single-stage and expected that the number of modules in the first separation
multistage separation processes to predict the required number unit increases to supply the required inlet flow for the second
unit. Referring to the single-stage separation results, the per-
of commercial modules for a given oxygen production. The
meate flow (40% O2 purity) is relatively low to feed the sec-
model and numerical technique are used to simulate a mem-
ond unit. Thus, the number of modules must be increased in
brane gas separation system for the production of 16,387 cm3/s
the first separation unit. In this case, 25 (for 3 3 105 fibers) or
(50,000 ft3/day) O2 with a purity of 60% as the permeate which
13 (for 6 3 105 fibers) modules are needed to supply enough
can be consumed in other industrial units. Referring to the
flow for the second unit.
above validation section (Case V), Feng et al.49 used an asym-
Based on the estimated number of modules, the permeate
metric hollow fiber membrane with a selectivity of 5.7 and con-
flow leaving the first separation unit is compressed and there-
ducted experimental work to separate a O2/N2 gas mixture. The
after fed to the second separation unit. Figure 7 shows the sep-
aim is to compare the separation performance and show the
aration performance of the second unit while the number of
potential of these membranes for O2 production using a mem-
fibers is set to 6 3 105. In these conditions, the mole fraction
brane gas separation unit. It is important to define a real case
unit to compare with experimental results. Hence, the same
feed composition, permeability, and fiber size as in Case V are
chosen as input. The gas mixture (O2/N2) is first fed to a single-
stage unit and then distributed into a number of parallel mem-
brane modules to produce O2 with the desired quality. Then, the
exact number of required modules is calculated based on the
membrane O2/N2 selectivity as reported by Feng et al.49 and on
typical ranges of gas velocity in the retentate and permeate sides
which are 1.5–1.7 and 0.01–0.38 m/s, respectively.35 Table 2
presents the parameters used for the modeling of the O2 produc-
tion unit.
In all modeling cases, the element refinement approach is
taken to determine the most accurate modeling results under
different operating conditions. Hence, the number of elements
along the module active length used in the calculation may be
varied to find the optimal value above which no changes in
stage cut and product concentration of both permeate and
retentate streams are observed. Figure 6. Separation performance in terms of stage cut
For the modeling, the feed and permeate pressures are cho- and O2 mole fraction as a function of module
sen to be 600 and 100 kPa, respectively. Figure 6 presents the feed flow rate: (a) Pf 5 600 kPa, Pp 5 100 kPa,
separation performance of the membrane used for O2 produc- number of fibers 5 3 3 105, (b) Pf 5 600 kPa,
tion using a single-stage system, while the number of fibers Pp 5 100 kPa, number of fibers 5 6 3 105.

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 7
Figure 7. Separation performance in terms of stage cut
and O2 mole fraction as a function of module
feed flow rate (Pf 5 600 kPa, Pp 5 100 kPa,
number of fibers 5 6 3 105).

of O2 in the permeate reaches 62% at a flow rate of

22,000 cm3/s. Figure 8. Effect of different CO2/CH4 selectivities on the
Table 3 presents a summary of the current case study of O2 CH4 retentate mole fraction and total mem-
production using a two-stage separation unit. With respect to brane area: (a) CO2 content in feed 5 10 mol %
the modeling results, 13 and 3 membrane modules are required and (b) CO2 content in feed 5 25 mol %.
for the first and second separation stages to produce
16,387 cm3/s (50,000 ft3/day) O2 with a purity higher than
60%. The permeate product could also be fed to a third separa- 80%. As CO2 reduces the natural gas heating value and is
tion unit to further enhance O2 purity. This configuration highly corrosive, it needs to be removed and thereafter sent to
would, however, require more equipment and pipelines to the CO2 Sequestration Unit or used for Enhanced Oil Recov-
recompress the permeate flow of the second unit resulting in a ery projects. The current conventional separation methods
higher energy consumption. The optimization of the mem- including absorption, adsorption, and cryogenic distillation
brane separation system for O2 production is beyond the scope can be employed to capture CO2. Despite the CO2 content,
of this study but the further analysis can also be carried out to these methods are even capable of removing 99% of CO2 from
optimize the operational costs by finding the optimum values the feed stream. Referring to the gas pipeline standards, the
of required separation area, transmembrane pressures, and effi- product enriched to methane >97% which has a CO2 content
cient ranges of O2/N2 selectivities. as low as 2–3%, can be injected into the natural gas grid.
Moreover, the current natural gas sweetening market is domi-
nated by amine absorption process. However, this process is
Case Study: Natural Gas Processing
highly energy-intensive and suffers from serious operational
Raw natural gas collected at a wellhead as a gaseous mix- problems. In this case, membrane technology can be regarded
ture contains a wide range of compounds in addition to meth- as an alternative method for the natural gas purification due to
ane, for example, heavier alkanes and aromatics, water, its simplicity, ease of installation, and low operation cost.
hydrogen sulphide, mercury and silicon-containing com- The model and new numerical technique were used to simu-
pounds, carbon dioxide, nitrogen, and helium. The composi- late a natural gas purification process using incorporation of
tion highly depends on the geological area and the Aspen Plus (version 8.8) and MATLAB (2015b) software. As
underground deposit type, depth, and location of reservoirs. there is no unit component in Aspen Plus toolbox for hollow
Some of these components might cause serious operational fiber membrane, Excel software is deployed to link Aspen
problems and therefore need to be removed through a series of Plus to MATLAB software. In this case, Aspen Plus indirectly
separation processes. In terms of acid gases, H2S can be calls the model’s function through Excel interface to use for
removed by a desulphurization process and its product can its internal calculation. In the first part, raw natural gas with
then be utilized for other applications such as Sulphur Recov- different CO2 contents (10, 25, 50, and 75%) was chosen as
ery Unit. The CO2 content in raw natural gas varies from 2 to feed to a single membrane unit. The gas flow rate is set to
10 mol/s, and the feed and permeate pressures are kept at 35
Table 3. Summary of the Modeling Results of a Two-Stage and 1 bar, respectively. The initial CO2 and CH4 permeance
Membrane Separation Unit for O2 Production are also experimental data57 of 17.7 and 0.73 GPU, respec-
tively. The aim of this simulation is to find the proper CO2/
O2 Mole Fraction Number Total CH4 selectivity while a single membrane unit is used for the
Number of of Fiber/ Separation natural gas purification. Figure 8a shows the relation among
Unit Permeate Retentate Modules Module Area (m2)
stage cut, retentate CH4 mole fraction, and membrane area
First 0.390 0.149 13 6 3 1025 7135 when the feed gas contains CO2 10%. As shown, increasing
Second 0.622 0.295 3 6 3 1025 1646 the CO2/CH4 selectivity results in increasing the retentate CH4

8 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
 of permeate gas is recycled. The higher the retentate CH4
mole fraction, the higher the membrane area is required which
results in an uneconomical and unrealistic process.
In the second part, a two-stage separation unit was designed
to simultaneously reduce the CH4 loss in the permeate stream
and the total required membrane area. Figure 10 shows the
process flow diagram of the natural gas purification process
using two membrane units. The feed (CO2 50%) has a flow
rate of 10 mol/s and a membrane with a CO2/CH4 selectivity
of 72 was chosen for this simulation case. As shown, the raw
natural gas is initially fed to the first membrane unit and the
retentate product is then sent to the second one. The permeate
product of the second unit is recompressed and recirculated to
the mixer. The simulation result shows that the retentate and
permeate CH4 mole fraction can reach 97 and 2% while the
required membrane areas of the first and second units are 200
and 300 m2, respectively. The CH4 loss in the permeate stream
also decreases by 2.6%. The current simulation result also
unveils the high potential of membrane technology for natural
gas separation area.

Use of the Model in Guiding Experimental

Figure 9. Effect of different CO2/CH4 selectivities on Membrane Development
the CH4 retentate mole fraction and total The new modeling approach which results in significant
membrane area: (a) CO2 content in reduction in computation time allows a rapid estimation of the
feed 5 50 mol % and (b) CO2 content in effects of membrane properties on large scale separation per-
feed 5 75 mol %. formance. The problem is made especially significant by
recent developments of new membrane materials with excep-
mole fraction. Using CO2/CH4 selectivity of 48 and 72 is tional properties, either extremely large selectivities58 or
appropriate to reach CH4 mole fraction of 97% for the stage extremely large permeabilities.59 How will such developments
cut of 10%. The required membrane area is 500 m2 which is affect the dimensions of commercial hollow fiber modules?
much lower than in the other cases reported in Figure 8a. This Figures 11–13 show the effect of O2/N2 selectivity (a)
also ensures a lower CH4 loss in the permeate stream. The changes on O2 permeate mole fraction and O2 recovery in a
simulation findings reveal that no significant difference in the single-stage process. In these simulation cases, the total mem-
retentate CH4 mole fraction is seen between CO2/CH4 selec- brane separation area was varied from 500 to 9000 m2 which
tivities of 48 and 72. The retentate CH4 mole fraction reaches can be fitted in approximately from 3 to 15 typical industrial
97% by using CO2/CH4 selectivities of 12 and 24 only as the modules with a module length of 180 cm. The initial O2 and
required membrane area is set to 2500 m2. In this case, stage N2 permeances were experimental data49 of 9.3 and 1.8 GPU,
cut increases approximately to 20% resulting in a higher CH4 respectively. In addition, the feed gas (air) with a flow rate of
loss in the permeate stream. Figures 8b and 9 show the effect 10 mol/s is initially compressed up to 6 bar and the permeate
of CO2/CH4 selectivities on the retentate CH4 mole fraction side is kept at ambient pressure. As shown, in the case of var-
and membrane area for the different CO2 (25, 50, and 75%) ied O2 permeance and constant N2 permeance (Figure 11),
content in the feed gas. As expected, using low CO2/CH4 increasing the O2/N2 selectivity from 5 to 40 increases the O2
selectivities results in a dramatic increase in the required mole fraction in the permeate stream from 0.42 to 0.65 for a
membrane area particularly when the selectivities are 12 and single unit having 500 m2 membrane area. This figure also
24. This causes an increase in CH4 loss even if some amount indicates that the enhancement of O2 mole fraction in the

Figure 10. Process flow diagram of the natural gas purification using two membrane units.
[Color figure can be viewed at wileyonlinelibrary.com]

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 9
Figure 11. Effect of O2/N2 selectivity on the O2 perme- Figure 13. Effect of O2/N2 selectivity on the O2 perme-
ate mole fraction and O2 recovery in a sin- ate mole fraction and O2 recovery in a sin-
gle membrane unit (N2 permeance 5 1.8 gle membrane unit (N2 permeance 5 3.6
GPU, O2 permeance variable). GPU, O2 permeance variable).
[Color figure can be viewed at wileyonlinelibrary.com] [Color figure can be viewed at wileyonlinelibrary.com]

permeate stream for a given membrane selectivity entails the compared to the simulation case of Figure 11. It is obvious
reduction of the O2 recovery. Similarly, at a constant O2 per- that a membrane with lower N2 permeance and higher O2 per-
meance (Figure 12), increasing the O2/N2 selectivities from 5 meance is more effective for a realistic air enrichment process.
to 40 by decreasing N2 permeance improves the O2 mole frac- Most importantly, the simulation results show that using high
tion in the permeate stream. The major difference between the selectivities cannot allow enriching to 99% O2 through a
two simulation cases is associated with the O2 recovery. The single-stage process and it is therefore essential to use a multi-
simulation results show that a lower O2 permeance leads to an stage process to separate O2 using a moderate O2/N2 selectiv-
increase in the required membrane area to compensate for the ity. The calculations reported in Figures 11–13 indicate how
O2 loss in the retentate stream. Figure 13 also shows the effect the model would allow setting realistic membrane perfor-
of O2/N2 selectivities on the O2 permeate mole fraction and mance targets in the development of new membrane materials.
O2 recovery using a single-stage process. In this case, N2 per-
meance is doubled reported in Figure 11 whereas O2 perme- Conclusion
ance increases to make O2/N2 selectivities from 5 to 40. This
change in membrane permselectivities results in a reduction in In this study, the simulation of a gas separation process was
O2 permeate mole fraction and an increase in O2 recovery carried out by developing a model for hollow fiber membrane
modules. A new numerical technique was introduced on the
basis of the flash calculation method. To validate the model,
the results for cocurrent and countercurrent gas separation pro-
cesses were compared with experimental data taken from liter-
ature and good agreements were observed for six different
cases. The results showed that the current model, using the
new numerical technique, was also useful for the simulation of
separation units using multicomponent gas mixtures under dif-
ferent operating conditions. In this numerical technique, the
MFEE was introduced to calculate the mole fractions and gas
flow rates on the permeate and retentate sides. After valida-
tion, the model was also used to simulate a real case air-
enrichment process using an asymmetric hollow fiber mem-
brane in single- and two-stage separation units. The results
showed that low O2 mole fraction (40%) could be achieved
in the permeate side using a single-stage unit with two differ-
ent numbers of fibers (3 3 105 and 6 3 105). Then, a two-
stage membrane unit was selected to reach the target O2 purity
(60%) and flow rate (16,387 cm3/s). The results also showed
Figure 12. Effect of O2/N2 selectivity on the O2 perme- that 13 and 3 modules with a total number of fibers of 6 3 105
ate mole fraction and O2 recovery in a sin- for the first and second membrane separation units are required
gle membrane unit (O2 permeance 5 9.3 to meet these targets. Either improvement of O2/N2 selectivity
GPU, N2 permeance variable). or adding more membrane separation units is required to get
[Color figure can be viewed at wileyonlinelibrary.com] higher purity (above 60%). For oxygen production, different

10 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal
economic parameters such as membrane fabrication cost, Q= permeance
membrane life, and transmembrane pressure ranges should be r= module inner radius
Rg = universal gas constant
considered to confirm the feasibility of these membrane sepa-
Re = Reynolds number
ration processes. Furthermore, the model with the numerical T= temperature
technique was used to simulate the natural gas purification Dt = element volume
process using the incorporation of Aspen Plus and MATLAB V= molar flow rate
software. Four different CO2/CH4 selectivities (12, 24, 48, and DVt = permeate flow rate
x= shell side mole fraction
72) were chosen to show the separation performance of a sin- y= fiber side mole fraction
gle membrane unit. The simulation result showed that CO2/
Greek letters
CH4 selectivities of 48 and 72 were more effective to reach
97% CH4 and the required membrane area significantly a= component index
b= component index
decreased. Lower CO2/CH4 selectivity resulted in an increase c= gas velocity
in the required membrane area and CH4 loss in the permeate E= void fraction
stream dramatically went up. A two-stage separation unit was g= viscosity
suggested to decrease both the required membrane area and h= stage cut
k= friction factor
the CH4 loss in the permeate stream. The simulation finding q= density
revealed that the feed gas (CO2 50%) was enriched to 97% r= Wike’s coefficient
CH4 in the retentate stream while the CH4 loss was also u= packing fraction
declined by 2.6%. Finally, the model was used to estimate the w= element stage cut
effect of membrane properties on the separation performance. Subscripts
The model was capable of showing the effect of intrinsic and c= gas component index
modified O2 and N2 permeances on O2 recovery and O2 mole f= feed side
fraction in the permeate stream using a single membrane unit. g= gas
hyd = hydraulic
The results showed that increasing O2/N2 selectivity calcu- i= element number
lated based on constant and varied O2 permeances was more m= membrane
in favor of increasing the product quality. The lower O2 per- mix = mixture
meance, however, led to increase the membrane separation o= outer
p= permeate
area to compensate for the O2 loss in the retentate stream. By r= retentate
doubling N2 permeance and increasing O2 permeance, the con- s= shell side
sequent change in the membrane permselectivities also t= fiber side
resulted in a reduction in O2 permeate mole fraction and an Literature Cited
increase in O2 recovery. Hence, using the membrane with the 1. Enerdata. 2016. Total Energy Consumption. Available at: https://
lower N2 permeance and the higher O2 permeance was the yearbook.enerdata.net/. Accessed on November 1, 2017.
proper modification target resulting in the more economical 2. Balzani V, Armaroli N. Energy for a Sustainable World: From the
air enrichment process. Thus, using this model, it is now possi- Oil Age to a Sun-Powered Future. New York: Wiley, 2010.
ble to assess how improvements in membrane properties will 3. Government of Canada. 2016. Canada’s Second Biennial Report on
Climate Change. Available at: https://www.ec.gc.ca/GES-GHG/
affect the size and configuration of membrane modules for a default.asp?lang5En&n502D095CB-1. Accessed on November 1,
given set of production conditions (design parameters). It is 2017.
therefore shown that the model is an especially useful tool in 4. Pires J, Martins F, Alvim-Ferraz M, Sim~ oes M. Recent developments
setting membrane properties’ targets for experimentalists on carbon capture and storage: an overview. Chem Eng Res Des.
when developing advanced new membrane materials. Refer- 2011;89(9):1446–1460.
5. Rubin ES, Mantripragada H, Marks A, Versteeg P, Kitchin J. The
ring to the model performances and simulation results, it can outlook for improved carbon capture technology. Prog Energy Com-
be concluded that the proposed methodology based on the bust Sci. 2012;38(5):630–671.
flash calculation method can be utilized to design single-stage 6. Harasimowicz M, Orluk P, Zakrzewska-Trznadel G, Chmielewski A.
and multistage gas separation processes using polymeric hol- Application of polyimide membranes for biogas purification and
enrichment. J Hazard Mater. 2007;144(3):698–702.
low fiber membranes under different operating conditions. In
7. Markos J. Mass Transfer in Chemical Engineering Processes. Croa-
future works, a techno-economic analysis based on these simu- tia: InTech, 2011.
lations will be made to optimize a membrane separation pro- 8. Mansourizadeh A, Ismail AF. Hollow fiber gas–liquid membrane
cess for O2 production and natural gas purification processes. contactors for acid gas capture: a review. J Hazard Mater. 2009;
171(1):38–53.
9. Chen XY, Vinh-Thang H, Ramirez AA, Rodrigue D, Kaliaguine S.
Acknowledgment Membrane gas separation technologies for biogas upgrading. RSC
The authors would like to thank the Natural Sciences and Adv. 2015;5(31):24399–24448.
10. Zhang Y, Sunarso J, Liu S, Wang R. Current status and development
Engineering Research Council of Canada (NSERC) for of membranes for CO2/CH4 separation: a review. Int J Greenhouse
financial support. Gas Control. 2013;12:84–107.
11. Cui Z, Muralidhara H. Membrane Technology: A Practical Guide to
Notation Membrane Technology and Applications in Food and Bioprocessing.
Burlington, VT: Elsevier, 2010.
DAf = membrane separation area 12. Blaisdell CT, Kammermeyer K. Counter-current and co-current gas
d= diameter separation. Chem Eng Sci. 1973;28(6):1249–1255.
G= mass flow rate 13. Oishi J, Matsumura Y, Higashi K, Ike C. An analysis of gaseous dif-
L= fiber active length fusion separating unit. Nippon Genshiryoku Gakkai-Shi. 1961;3(12):
M= molecular weight 923–928.
nf = number of fibers 14. Naylor RW, Backer PO. Enrichment calculations in gaseous diffu-
P= pressure sion: large separation factor. AIChE J. 1955;1(1):95–99.

AIChE Journal 2017 Vol. 00, No. 00 Published on behalf of the AIChE DOI 10.1002/aic 11
15. Weller S, Steiner WA. Separation of gases by fractional permeation 38. Ghasem N, Al-Marzouqi M, Rahim NA. Modeling of CO2 absorp-
through membranes. J Appl Phys. 1950;21(4):279–283. tion in a membrane contactor considering solvent evaporation. Sep
16. Brubaker DW, Kammermeyer K. Separation of gases by plastic Purif Technol. 2013;110:1–10.
membranes-permeation rates and extent of separation. Ind Eng 39. Luis P, Garea A, Irabien A. Modelling of a hollow fibre ceramic
Chem. 1954;46(4):733–739. contactor for SO2 absorption. Sep Purif Technol. 2010;72(2):174–
17. Pan CY, Habgood H. Gas separation by permeation part I. Calcula- 179.
tion methods and parametric analysis. Can J Chem Eng. 1978;56(2): 40. Wang R, Li D, Liang D. Modeling of CO2 capture by three typical
197–209. amine solutions in hollow fiber membrane contactors. Chem Eng
18. Bansal R, Jain V, Gupta SK. Analysis of separation of multicompo- Process Process Intensif. 2004;43(7):849–856.
nent mixtures across membranes in a single permeation unit. Sep Sci 41. Rezakazemi M, Niazi Z, Mirfendereski M, Shirazian S, Mohammadi
Technol. 1995;30(14):2891–2916. T, Pak A. CFD simulation of natural gas sweetening in a gas–liquid
19. Walawender W, Stern S. Analysis of membrane separation parame- hollow-fiber membrane contactor. Chem Eng J. 2011;168(3):1217–
ters. II. Counter-current and cocurrent flow in a single permeation 1226.
stage. Sep Sci. 1972;7(5):553–584. 42. Sohrabi MR, Marjani A, Moradi S, Davallo M, Shirazian S. Mathe-
20. Stern S, Walawender W, Jr. Analysis of membrane separation matical modeling and numerical simulation of CO2 transport
parameters. Sep Sci Technol. 1969;4(2):129–159. through hollow-fiber membranes. Appl Math Model. 2011;35(1):
21. Shindo Y, Hakuta T, Yoshitome H, Inoue H. Calculation methods 174–188.
for multicomponent gas separation by permeation. Sep Sci Technol. 43. Eslami S, Mousavi SM, Danesh S, Banazadeh H. Modeling and sim-
1985;20(5–6):445–459. ulation of CO2 removal from power plant flue gas by PG solution in
22. Pan C. Gas separation by high-flux, asymmetric hollow-fiber mem- a hollow fiber membrane contactor. Adv Eng Softw. 2011;42(8):612–
brane. AIChE J. 1986;32(12):2020–2027. 620.
23. Lim S, Tan X, Li K. Gas/vapour separation using membranes: effect 44. Gilassi S, Rahmanian N. CFD modelling of a hollow fibre membrane
of pressure drop in lumen of hollow fibres. Chem Eng Sci. 2000; for CO2 removal by aqueous amine solutions of MEA, DEA and
55(14):2641–2652. MDEA. Int J Chem Reactor Eng. 2016;14(1):53–61.
24. Murad Chowdhury MH, Feng X, Douglas P, Croiset E. A new 45. Wang R, Zhang H, Feron P, Liang D. Influence of membrane wet-
numerical approach for a detailed multicomponent gas separation ting on CO2 capture in microporous hollow fiber membrane contac-
membrane model and AspenPlus simulation. Chem Eng Technol. tors. Sep Purif Technol. 2005;46(1):33–40.
2005;28(7):773–782. 46. Zhang H-Y, Wang R, Liang DT, Tay JH. Theoretical and experi-
25. Khalilpour R, Abbas A, Lai Z, Pinnau I. Analysis of hollow fibre mental studies of membrane wetting in the membrane gas–liquid
membrane systems for multicomponent gas separation. Chem Eng contacting process for CO2 absorption. J Memb Sci. 2008;308(1):
Res Des. 2013;91(2):332–347. 162–170.
26. Khalilpour R, Abbas A, Lai Z, Pinnau I. Modeling and parametric 47. Tranchino L, Santarossa R, Carta F, Fabiani C, Bimbi L. Gas separa-
analysis of hollow fiber membrane system for carbon capture from tion in a membrane unit: experimental results and theoretical predic-
multicomponent flue gas. AIChE J. 2012;58(5):1550–1561. tions. Sep Sci Technol. 1989;24(14):1207–1226.
27. Kundu PK, Chakma A, Feng X. Simulation of binary gas separation 48. Sidhoum M, Sengupta A, Sirkar K. Asymmetric cellulose acetate hol-
with asymmetric hollow fibre membranes and case studies of air sep- low fibers: studies in gas permeation. AIChE J. 1988;34(3):417–425.
aration. Can J Chem Eng. 2012;90(5):1253–1268. 49. Feng X, Ivory J, Rajan VS. Air separation by integrally asymmetric
28. Thundyil MJ, Koros WJ. Mathematical modeling of gas separation hollow-fiber membranes. AIChE J. 1999;45(10):2142–2152.
permeators—for radial crossflow, countercurrent, and cocurrent 50. Rautenbach R, Knauf R, Struck A, Vier J. Simulation and design of
hollow fiber membrane modules. J Memb Sci. 1997;125(2):275– membrane plants with AspenPlus. Chem Eng Technol. 1996;19(5):
291. 391–397.
29. Ahmad F, Lau K, Shariff A, Yeong YF. Temperature and pressure 51. Sada E, Kumazawa H, Wang JS, Koizumi M. Separation of carbon
dependence of membrane permeance and its effect on process eco- dioxide by asymmetric hollow fiber membrane of cellulose triacetate.
nomics of hollow fiber gas separation system. J Memb Sci. 2013; J Appl Polym Sci. 1992;45(12):2181–2186.
430:44–55. 52. Sircar S, Golden T. Purification of hydrogen by pressure swing
30. Marić I. The Joule–Thomson effect in natural gas flow-rate measure- adsorption. Sep Sci Technol. 2000;35(5):667–687.
ments. Flow Meas Instrum. 2005;16(6):387–395. 53. Prasad R, Shaner R, Doshi K. Comparison of membranes with other
31. Marić I. A procedure for the calculation of the natural gas molar gas separation technologies. In: Paul DR, Yampol’skii YP, editors.
heat capacity, the isentropic exponent, and the Joule–Thomson coef- Polymeric Gas Separation Membranes. Boca Raton: CRC Press,
ficient. Flow Meas Instrum. 2007;18(1):18–26. 1994:531–614.
32. Coker D, Freeman B, Fleming G. Modeling multicomponent gas 54. Gottschlich DE, Roberts DL. Energy Minimization of Separation
separation using hollow-fiber membrane contactors. AIChE J. 1998; Processes Using Conventional/Membrane Hybrid Systems. Idaho
44(6):1289–1302. Falls, ID: EG & G Idaho, Inc., 1990.
33. Coker D, Allen T, Freeman B, Fleming G. Nonisothermal model for 55. Smith A, Klosek J. A review of air separation technologies and their
gas separation hollow-fiber membranes. AIChE J. 1999;45(7):1451– integration with energy conversion processes. Fuel Process Technol.
1468. 2001;70(2):115–134.
34. Lock S, Lau K, Shariff A. Effect of recycle ratio on the cost of natu- 56. Baker RW. Future directions of membrane gas separation technol-
ral gas processing in countercurrent hollow fiber membrane system. ogy. Ind Eng Chem Res. 2002;41(6):1393–1411.
J Ind Eng Chem. 2015;21:542–551. 57. Chen XY, Kaliaguine S, Rodrigue D. A comparison between several
35. Scholz M, Harlacher T, Melin T, Wessling M. Modeling gas perme- commercial polymer hollow fiber membranes for gas separation.
ation by linking nonideal effects. Ind Eng Chem Res. 2013;52(3): J Memb Sep Technol. 2017;6(1):1–15.
1079–1088. 58. Zhang C, Koros WJ. Ultraselective carbon molecular sieve mem-
36. Hosseini SS, Roodashti SM, Kundu PK, Tan NR. Transport proper- branes with tailored synergistic sorption selective properties. Adv
ties of asymmetric hollow fiber membrane permeators for practical Mater. 2017;29(33):1701631.
applications: mathematical modelling for binary gas mixtures. Can J 59. Tien-Binh N, Vinh-Thang H, Chen XY, Rodrigue D, Kaliaguine S.
Chem Eng. 2015;93(7):1275–1287. Crosslinked MOF-polymer to enhance gas separation of mixed
37. Faiz R, Al-Marzouqi M. Mathematical modeling for the simulta- matrix membranes. J Memb Sci. 2016;520:941–950.
neous absorption of CO2 and H2S using MEA in hollow fiber mem-
brane contactors. J Memb Sci. 2009;342(1):269–278. Manuscript received Oct. 21, 2017, and revision received Nov. 23, 2017.

12 DOI 10.1002/aic Published on behalf of the AIChE 2017 Vol. 00, No. 00 AIChE Journal