You are on page 1of 11

Studies on simulation and experiments of ethanolwater mixture

separation by VMD using a PTFE at membrane module


Jiao-Yu Shi, Zhi-Ping Zhao

, Chun-Yan Zhu
School of Chemical Engineering and the Environment, Beijing Institute of Technology, Beijing 100081, China
a r t i c l e i n f o
Article history:
Received 16 August 2013
Received in revised form 23 November 2013
Accepted 21 December 2013
Available online 28 December 2013
Keywords:
Vacuum membrane distillation
Knudsenviscous transition model
Numerical simulation
Model parameters
Ethanolwater mixture
a b s t r a c t
Model parameters modication and numerical simulation based on the Knudsenviscous transition
model to predict the vacuum membrane distillation (VMD) performance of ethanolwater mixture and
experiments were studied in a polytetrauoroethylene at-sheet membrane module. A novel approach
to determine the model parameters K and B (modied Knudsen diffusion and viscous ow membrane
characteristics, respectively) consisted of membrane morphological parameters was especially developed
by tting the water VMD experimental results. Furthermore, linear relations between the modied model
parameter K or B and feed temperature were rst established, and then successfully employed to predict
by simulation the separation performances of 5 wt% ethanolwater mixture. The experiments of ethanol
water VMD demonstrated that the uxes of ethanol and water both increased with temperature or vac-
uum degree, but the separation factor decreased. The simulation values agree quite well with experimen-
tal ones, nding the minimum and maximum discrepancies are 0.5% and 9.6% for ux and separation
factor, respectively.
2013 Elsevier B.V. All rights reserved.
1. Introduction
Membrane distillation (MD) technology has been reported as a
cost-effective and energy-saving separation process, it has been
widely applied to a range of conventionally difcult separations.
According to the different patterns of condensation for volatile
components in the permeate side, MD can be divided into four con-
gurations named as direct contact membrane distillation (DCMD)
[13], air gap membrane distillation (AGMD) [4,5], sweeping gas
membrane distillation (SGMD) [6,7], and vacuummembrane distil-
lation (VMD) [813].
In MD process, temperature polarization phenomenon plays vi-
tal role in the separation performance which is mainly caused by:
(1) the vaporization of species on the membrane surface and (2)
the heat loss conducted by both membrane matrix and the vapor
in membrane pores [13]. The temperature polarization caused by
the latter factor in VMD was considerably reduced compared with
that in DCMD conguration. On the other hand, the applied vac-
uum pressure is lower than the saturation pressure of volatile com-
ponent to be separated from the feed solution, consequently, the
VMD process can obtain a larger permeate ux. As one of the most
promising separation technologies, the VMD was widely
researched for seawater desalination [1416], drinking water
purication [17,18], separation of volatile component [13,19],
non-volatile solutions concentration [8,20], wastewater treatment
[21,22], separation of azeotropic aqueous mixture [23], and plant
extracts concentration [9,24]. Especially, in the face of worldwide
energy shortage [13], production of ethanol from biomass has be-
come a popular alternative approach to relief global energy crisis
[25,26]. Due to its possibility in separation of ethanol aqueous
solutions, the membrane technology was proposed as an applica-
tion to in situ separation of ethanol from fermentation system
[13]. The in situ separation can reduce the natural inhibitions of
cell growth that caused by high concentrations of ethanol [13]
and then promote the biomass conversion efciency. Compared
with the pervaporation (PV) separation for ethanolwater mix-
tures, the VMD process has the same selection level. For example,
a series of composite silicone rubber membranes with PVDF as
supports were investigated in the PV separation of 5 wt% etha-
nolwater mixtures, and the experimental separation factors range
from 3.2 to 11.7 [27]. Other silanol-based silicones, such as PTMSP-
silica membranes, have also been utilized in PV separation of eth-
anolwater mixtures, and the similar separation factor results
were obtained [28]. In fact, MD process exhibited a comparable
or even better performance to PV with polymer membranes. Lew-
andowicz et al. [29] presented an excellent selectivity in separation
of ethanolwater solutions and greatly increased the bioreactor
productivity by removing the ethanol by DCMD. The selectivity
factor is about 20 when transmembrane temperature difference
1383-5866/$ - see front matter 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.seppur.2013.12.015

Corresponding author. Tel./fax: +86 10 68911032.


E-mail address: zhaozp@bit.edu.cn (Z.-P. Zhao).
Separation and Purication Technology 123 (2014) 5363
Contents lists available at ScienceDirect
Separation and Purication Technology
j our nal homepage: www. el sevi er. com/ l ocat e/ seppur
is 15 C. Also Gryta et al. [30] studied ethanol production by batch
fermentation integrated with MD system. The ethanol concentra-
tion in the permeate side can reach 65 g/dm
3
when the initial eth-
anol concentration of broth was 12 g/dm
3
at 60 C (corresponding
separation factor was 9.4). Actually, VMD is considered as an avail-
able conguration which can be applied in the fermentation pro-
cess for its good performance in separation of volatile solution
[8,20,31].
The heat and mass transfer in VMD have received increasing
attention. Some transfer mechanisms have been proposed for va-
por through micropores in the hydrophobic membranes, and corre-
sponding mathematical models describing heat and mass transfer
in VMD have been presented [10,11,13,32]. In various models,
the dusty gas model was frequently employed to describe the mass
transfer in membrane pores in VMD process [13,33]. Meanwhile,
the reliability of model equation largely depends on the model
parameters K
0
and B
0
(the Knudsen diffusion and viscous ow
membrane characteristics, respectively) which consist of mem-
brane morphological parameters, such as pore radius (r), porosity
(e), and tortuosity (s). Some researchers made efforts to develop
an accurate method for the determination of the above model
parameters. For instance, Guijt et al. [34] reported one method to
measure the K
0
and B
0
by employing the gases He, N
2
, and CO
2
to permeate through three kinds of polypropylene membranes,
Nomenclature
Symbols
A area of membrane in one element cell (m
2
)
B viscous ow model parameter (m)
B
0
viscous ow coefcient (m
2
)
c
fb
concentration of ethanol in the feed bulk liquid (mol/
m
3
)
c
fm
concentration of ethanol at the feed-membrane inter-
face (mol/m
3
)
c
p
concentration of ethanol in the permeate (mol/m
3
)
C
p,i
specic heat (J/kg K)
d equivalent diameter of ow channel (m)
D diffusion coefcient (m
2
/s)
h
f
mass transfer coefcient (W/m
2
K)
DH
i
evaporation enthalpy (J/mol)
J trans-membrane mass ux (kg/(m
2
h))
k themal conductivity (W/m K)
k
L
mass transfer coefcient (m/s)
K Knudsen diffusion model parameter
k
B
Boltzmann constant
K
0
Knudsen diffusion coefcient (m)
l ow channel length (m)
M molecular mass (kg/kmol)
m mass ow rate (kg/h)
N trans-membrane molar ux (mol/m
2
s)
N
i
molar ux of the component i (mol/m
2
s)
N
0
1
initial value of water ux in simulation program (mol/
m
2
s)
N
0
2
initial value of ethanol ux in simulation program (mol/
m
2
s)
Nu Nusselt number
P total vapor pressure (Pa)
P
i
vapor partial pressure of component i (Pa)
P
0
i
saturation vapor pressure of component i (Pa)
P
o
total pressure within the membrane pores (Pa)
DP trans-membrane vapor pressure difference (Pa)
DP
i
trans-membrane vapor partial pressure difference of
component i (Pa)
P
i,avg
average partial pressure of component i (Pa)
P
fm
vapor pressure at feed-membrane interface surface (Pa)
P
p
vapor pressure in the permeate side (Pa)
P
v
vacuum degree in the permeate side (kPa)
Pr Prandtl number
Q
f
convective heat transfer (W/m
2
)
Q
H
latent heat (W/m
2
)
Q
C
heat by convection (W/m
2
)
Q
m
heat transfer across the membrane (W/m
2
)
r pore radius (m)
R gas constant (8.314J/mol K)
Re Reynolds number
Dl step length in simulation
Sc Schmidts number
Sh Sherwood number
t feed inlet temperature (C)
T feed inlet temperature (K)
T
avg
average temperature across the membrane (K)
T
fb
temperature of the feed bulk (K)
T
fm
temperature at membrane surface on the feed side (K)
T
0
fm
initial value of T
fm
in simulation program (K)
T
m
mean temperature of thermal boundary layer (K)
T
o
temperature within membrane pores (K)
T
P
temperature in the permeate side (K)
u ow velocity of feed liquid (m/s)
x liquid mole fraction
y vapor mole fraction
z pure component properties (viscosity in Pa s or specic
heat in J/(kg K))
Z
mix
ethanolwater mixture properties (viscosity in Pa s or
specic heat in J/(kg K))
Greek symbols
a separation factor
c activity coefcient (m)
d membrane thickness (m)
e membrane porosity
k
i
the mean free path of component i (m)
l
i
viscosity of specie i (Pa s)
q density of ethanolwater mixture (kg/m
3
)
r
i
collision diameter (m)
q
i
density of specie i (kg/m
3
)
s membrane tortuosity factor
x mass fraction
Superscripts
n in the nth divided unit
n+1 in the (n+1)th divided unit
Subscripts
1 pure water
2 ethanol
avg average value
f feed liquid
fb feed bulk
fm feed-membrane interface
i species i
j loop counter
o membrane pores
P permeate side
mix ethanolwater mixture
54 J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363
respectively. To ensure that the pressure drop inside the permeable
part of the membrane ber can be neglected, the length of the
membrane ber was reduced into 7 mm, under these conditions,
the K
0
and B
0
were regarded as constants. And the results revealed
that the Heliumgives less accurate results due to its lowmolecular,
whereas, the N
2
and CO
2
lead to consistent values of K
0
and B
0
.
Lawson et al. [35] presented another method to determine the val-
ues of K
0
and B
0
, and took the effects of the membrane compaction
on gas permeability of microporous membranes into account. The
compaction can result in the changes of porosity (e), thickness (d),
pore radius (r), and tortuosity (s) of membrane. Therefore, experi-
ments were carried out with 16 layers membrane, and both N
2
and H
2
were permeated through the membrane at 25 C over a
range of feed and permeate pressures (0200 kPa), respectively.
Besides, Beuscher and Gooding [36] adopted similar single gas
permeation experiments to determine morphological parameters.
They suggested that K
0
and B
0
should be constants for a given med-
ium and independent of the gas used. Single gas permeation exper-
iments were performed for six different gases, including H
2
, He, N
2
,
O
2
, C
3
H
8
, and CO
2
. The calculated model parameters varied slightly
with six different gases used. They ascribed those differences to the
different asymmetry of the porous structure.
However, the operating conditions in a realistic MD process
were very different from those in the experimental methods men-
tioned above. In fact, not only the membrane morphological
parameters which constitute K
0
and B
0
but also the membrane
thickness (d) varies with the feed temperature and trans-mem-
brane pressure. That is to say that the effects of operating condi-
tions in a real VMD process, such as feed temperature and
vacuum degree in the permeate side, on the model parameters,
and the membrane thickness (d) should be taken into account to
determine the model parameters.
One aim of the present work was to provide a new approach to
determine the modied model parameters consisted of membrane
morphological parameters by pure water VMD experiments in a
polytetrauoroethylene (PTFE) at-sheet membrane module, tak-
ing the inuence of operating temperature change into account.
Then, a simulation program was developed and numerical simula-
tion of ethanolwater VMD in a PTFE at membrane module was
carried out using the obtained model parameters. The simulated
uxes and separation factors were compared with the experimen-
tal ones, nding a good agreement between them.
2. Experimental procedures and methods
2.1. Materials and module
PTFE hydrophobic microporous at membranes used for the
VMD process in this work, with average pore size of 0.22 lm,
thickness of 50 lm and membrane porosity of 6070%, was
purchased from Beijing Institute of Plastics. Scanning electron
microscopy (SEM) analysis of membrane surface was done on a
JSM-7401F scanning electron microscope. Fig. 1 shows the SEM
image of PTFE microporous at membrane. This membrane was
prepared by uniaxial stretching; it comprises brils completely
oriented in the stretching direction, remaining island-likely long
and narrow fractures of the rolled PTFE sheet.
A at-sheet membrane module with helical diversion groove
was designed by our lab. Membrane with effective ltration area
of 20 cm
2
was supported using a porous plate covered with stain-
less steel net, and sealed with a PTFE ring. The rectangular feed
channel in the module is 1 0.5 cm.
Pure water and 5 wt% ethanolwater mixtures were used for
the VMD experiments and simulation. De-ionized water was used
as feeds in order to avoid the interference of various ions in water.
Anhydrous ethanol (99.7%) was purchased from Beijing Chemical
Plant.
2.2. Set-up and methods
A schematic view of the VMD apparatus used in this work was
shown in previous work [24]. All the pipes and feed reservoir were
insulated to minimize the heat losses. A spacer was placed at the
entrance of the module to reduce the feed direct impact on the
membrane surface. Within the whole ow passage of membrane,
no spacer was set. The temperatures of the feed at the inlet of
the module was measured and controlled by an electric heater cou-
pled with a temperature controller.
The feed liquid in the reservoir was heated to the desired tem-
perature and then continuously pumped through a rotormeter to
the membrane module, a condenser and a vacuum pump were em-
ployed as a part of the permeate side apparatus. After transferred
through membrane pores to the vacuum side under the pressure
difference across the membrane as driving force, the vapor could
be soon condensed in a glass condenser jacket and was sampled
every a certain time by an electronic balance, and the feed owed
back to the circulation reservoir to be heated again. In order to
maintain feed concentration stability, the samples were also put
back into the reservoir after tested. The obtained distillate in the
condensate collector using running water or liquid nitrogen as
the coolant in condenser jacket were compared, and these two
kinds of coolants did not make any difference in the membrane
performances (ux and separation factor). Therefore, running
water was used as the coolant for permeate condensation in our
experimental process.
A gas chromatograph (GC7890 II) with a thermal conductivity
detector was used to determine the concentration of ethanol in
distillation.
3. Theoretic model and simulation algorithm
The transfer mechanisms in the VMD process involve heat and
mass transfer simultaneously. In the separation process of etha-
nolwater mixture, the trans-membrane water and ethanol per-
meations lead to the appearance of polarization phenomenon at
the surface of membrane and the consequent reduction in driving
force as can be showed in Fig. 2. As the volatile component in the
feed side, the concentration of ethanol at membrane surface is low-
er than that in the feed bulk, it can be considered as water polari-
zation which is quite different from brine systems. The mass
transfer process in VMD is driven by vapor pressure gradient across
Fig. 1. SEM image of PTFE microporous at membrane.
J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363 55
the membrane. When the Poiseuille ow mechanism governs the
mass transfer in membrane pores, the permeate ux is propor-
tional to the trans-membrane vapor pressure difference (TMVP),
therefore, the ethanol concentration of the permeating vapor in
the pores is the same as that of the permeated vapor in the vacuum
side [13,31]. However, it is generally accepted that both the
Knudsen diffusion and Poiseuille/viscous ow contribute to the
total vapor ux through the membrane pores. Then the ethanol
concentration of vapor decreases from C
fm
at the feed-vapor
interface to C
p
in the permeated vapor due to the fact that the small
water molecules have greater diffusion rate than the large ethanol
molecules.
Correspondingly, the vaporization of solution at liquidvapor
interface requires heat, which results in a temperature gradient,
the thermal boundary layer, from the bulk feed temperature T
fb
to the membrane surface temperature T
fm
. For a well thermal
retardation system, it is acknowledged that there is no tempera-
ture gradient between the two sides of the membrane in the
VMD. Therefore, the temperature and concentration polarizations
mainly occurred in the feed side during a VMD process.
3.1. Heat transfer models
In a MD process with good thermal insulation conditions, the
heat Q
m
that transfers across the membrane mainly consists of
two parts [10]: (1) the latent heat Q
H
accompanying the trans-
membrane mass transfer, (2) and the heat Q
C
conducted by the
membrane matrix. The thermal resistance outside the membrane
mainly concentrates within the thermal boundary layer. The fol-
lowing equation is usually used to describe the convective heat
transfer Q
f
in the boundary layer:
Q
f
h
f
T
fb
T
fm
1
where h
f
is the convective heat transfer coefcient in boundary
layer; T
fb
is the bulk feed temperature, T
fm
is the membrane surface
temperature; and h
f
was calculated by Nusselt equation: h
f
= Nuk/d,
where k is thermal conductivity, d is the ow channel equivalent
diameter, Nusselt number Nu is related to the Reynolds number
Re and can be estimated from the simplied semiempirical correla-
tions [37]:
Nu 1:86 RePr
d
l

0:33
Re < 2100 2
Nu0:116Re
2=3
125Pr
0:33
1
d
l

2=3
!
2100<Re <10000
3
Nu 0:023Re
0:8
Pr
0:33
Re > 10000 4
where Pr is Prandtl number, and l is the ow channel length.
For a VMD process, the steam is drawn out of the system after
passing through the membrane. The temperature difference across
the membrane is negligible. Therefore, the heat loss Q
C
by conduc-
tion through membrane matrix and the gas lled in pores can be
considered negligible [10,11,13], the convective heat transfer Q
f
is used to vaporize the liquid at the liquidvapor interface, and
the following correlation can be achieved [8]:
Q
f
Q
m
Q
H
Q
C

X
n
i1
N
i
DH
i
5
where N
i
is the molar ux of the component i, DH
i
is the evapora-
tion enthalpy per unit mole of the component i.
Considering Eqs. (1) and (5), the temperature T
fm
at feed-mem-
brane interface can be calculated [8]:
T
fm
T
fb

P
N
i
DH
i
h
f
6
3.2. Mass transfer models
For the VMD system, the mass transfer process consists of the
convective transfer in the boundary layer and the vapor perme-
ation through the membrane pores, and the mass transfer resis-
tance in the vacuum side is ignored. The mass transfer ux
across the membrane can be described as the lm theory model
[12,13]:
J k
L
q ln
x
fm
x
p
x
fb
x
p

7
where J is the total mass ux, k
L
is the mass transfer coefcient, q is
the feed liquid density, and the xs represent the weight fraction of
ethanol in the bulk feed (x
fb
), at the membrane surface (x
fm
) and in
the permeate (x
p
). The weight fraction of ethanol at the feed-mem-
brane interface can be obtained from Eq. (7).k
L
can be estimated
from the semiempirical correlation [23]:
k
L

D Sh
d
8
where D is the diffusion coefcient which can be estimated by Wil-
keChang empirical formula [38], and Sh is Sherwood number.
While,
Re < 2100, Laminar regime
Sh 1:86 ReSc
d
l

0:33
9
2100 < Re < 10000, Transitional regime
Sh 0:116Re
2=3
125Sc
0:33
1
d
l

2=3
!
10
Re > 10000, Turbulent regime
Sh 0:023Re
0:8
Sc
0:33
11
where Sc is Schmidt number.
Due to the lowconcentration of ethanolwater mixtures used in
this work, it can be assumed that (1) there is no wetting occurred
during the experimental process; (2) there is no species accumu-
lated within the membrane pore or detained on the membrane
Fig. 2. Schematic view of VMD process.
56 J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363
surface. The analysis of heat and mass transfer in this process
above is based on dividing the ow channel into elements as
shown in Fig. 3.
Thus, species material balance:
m
n
f
x
n
i;f
J
n1
x
n1
i;p
A m
n1
f
x
n1
i;f
12
Total mass balance in the feed side:
m
n
f
m
n1
f
J
n1
A 13
where m
n
f
and m
n1
f
are the mass ow rate on the feed side in ele-
ment cell n and n + 1, respectively; x
n
i;f
and x
n1
i;f
are the mass frac-
tions of component i on the feed side in element cell n and n + 1,
respectively. J
n+1
is the mass ux in element cell n + 1, x
n1
i;p
is the
mass fraction of component i in the permeate side and A is the area
of membrane in one element cell.
Numerical models were based on the assumption that vapor
permeates through a porous membrane including three parts of
contribution, namely, Knudsen diffusion, Poiseuille/viscous ow,
and molecular diffusion [29]. In the VMD conguration, the ordin-
ary molecular diffusion resistance is neglected because it is propor-
tional to the partial pressure of air in the membrane pores and only
traces of air exist within the membrane pores in VMD [39]. Admit-
tedly, which trans-membrane mass transfer mechanismdominates
in VMD mainly depends on the mean membrane pore size and on
the mean free path of the diffusing component. Generally, when
the ratio of the pore radius to the mean free path (i.e. r/k) is lower
than 0.05, the molecule-pore wall collisions is conspicuous, in this
case, Knudsen diffusion is dominate. When 0.05k < r < 50k, both
Knudsen and transition mechanism are applicable; if r > 50k,
molecularmolecular collisions cannot be omitted, and all mecha-
nisms (molecular diffusion, Knudsen diffusion and viscous ow)
are serving simultaneously [39]. Microporous and hydrophobic
membranes are used in MD process, and their mean pore size gen-
erally ranges from 0.01 lm to 1 lm [29]. The molecular diameter is
2.641 for water [40], and 4.3 for ethanol [41]. And the mean
free path of a given specie i can be calculated by the following
expression [42]:
k
i

k
B
T
o

2
p
pP
o
r
2
i
14
where k
i
is the mean free path of component i, k
B
is Boltzmann con-
stant, T
o
is temperature within the membrane pores (K), P
o
is total
pressure within the membrane pores in Pa, and r
i
is the collision
diameter of component i.
Taking into account that, VMD process is generally conducted
under 4070 C, and operates at permeate pressure from 7 kPa to
31 kPa, which is below the saturated vapor pressure of the diffus-
ing species. The response surface methodology was carried out in
Fig. 4, in which the effects of two independent variables (temper-
ature and pressure) on the mean free molecular path of water or
ethanol can be presented clearly as to explain mass transfer mech-
anism better.
As can be seen obviously, the mean free path declines rapidly
with increase of the pressure and ascends gradually with the in-
crease of temperature, the mean free path of water is larger than
that of ethanol under the same operating conditions. Also, the
mean free paths of water and ethanol range from 0.491.99 lm
and 0.180.72 lm, respectively.
In respect that the mean free path of water or ethanol is on the
same order as the mean pore size of the membranes used in the
VMD process, and there is trace amount of gas in membrane pores
due to the low operating pressure, the moleculemolecule colli-
sions are ignored, accordingly, the molecular diffusion resistance
is neglected [31,39]. Therefore, both the Knudsen diffusion and
the viscous ow contribute to the total vapor ux through the
membrane pores. The Knudsenviscous transition equation based
on dusty-gas model is applicable as [32]:
Fig. 3. Schematic diagram of the materials balance in the divided units.
310
320
330
340
350
0
10
20
30
40
0
0.5
1
1.5
2
2.5
M
e
a
n

f
r
e
e

p
a
t
h

o
f

w
a
t
e
r
/

m
)
(a)
310
320
330
340
350
0
10
20
30
40
0
0.2
0.4
0.6
0.8
T
e
m
p
a
tu
re
/T
(K
) A
b
so
lu
te
p
re
ssu
re
/P
(
k
P
a)
M
e
a
n

f
r
e
e

p
a
t
h

o
f

e
t
h
a
n
o
l
/

m
)
(b)
A
b
so
lu
te
p
re
ssu
re
/P
(
k
P
a)
T
e
m
p
a
tu
re
/T
(K
)
Fig. 4. The mean free path variations of water (a) and ethanol (b) with temperature
and pressure.
J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363 57
N
i

1
RT
avg
d
K
0
8RT
pM
i
1
2
DP
i
B
0
P
i;avg
l
DP
" #
15
where N
i
is the molar ux of the component i, R is gas constant. T
avg
is the average of the membrane surface temperatures on the feed
and permeate side. DP is the TMVP, M
i
is the molecular weight of
component i and DP
i
is the partial vapor pressure difference of com-
ponent i across the membrane. P
i,avg
is the average partial pressure
of component i across the membrane.
For non-ideal binary mixtures the vapor partial pressures can be
determined by modied Raoults law [38]:
P
i
y
i
P x
i
c
i
P
0
i
16
where y
i
and x
i
are the vapor and liquid mole fractions of compo-
nent i respectively, P is the total vapor pressure, P
0
i
is the saturation
pressure of pure component i which can be estimated by Antoine
equation, and c
i
is the activity coefcient of component i in the
solution. The value of c
i
can be calculated from van Laar equation.
In ethanolwater binary system, the separation factor (a) was
used to describe the selectivity of VMD separation process which
is dened as:
a
x
p
1 x
f

x
f
1 x
p

17
In the Knudsenviscous equation, both of the Knudsen diffusion
and viscous ow model characteristics K
0
and B
0
depend on the
membrane specic geometric properties like average pore radius
(r), porosity (e), tortuosity (s), according to Eq. (17) [34,36].
K
0

2er
3s
; B
0

er
2
8s
18
Because the membrane thickness (d) may vary with the operat-
ing conditions in VMD, we integrated the d into K
0
and B
0
, and ob-
tained the modied model parameters K and B respectively, as
given in formula (19),
K
K
0
d

2er
3sd
; B
B
0
d

er
2
8sd
19
By the way, the temperature in formula (15) refers the mem-
brane surface temperature T
fm
. The trans-membrane temperature
gradient in VMD process can be ignored [13,24]. Therefore, the
Knudsenviscous Eq. (15) can be modied as follows:
N
i

1
RT
fm
K
8RT
fm
pM
i
1
2
DP
i
B
P
i;avg
l
DP
" #
20
For the pure water VMD, the DP
i
is the same as DP, and P
i,avg
equals to (P
fm
+ P
p
)/2, correspondingly, for easily tting experimen-
tal data, the mass transfer equation can be expressed as below:
N
1
RT
fm

B
2
1
l
1
P
2
P
K
8RT
fm
pM
1
1
2
P
P

B
2l
1
P
2
fm
K
8RT
fm
pM
1
1
2
P
fm
" #
21
where P
P
is the vapor pressure in permeate side, l
1
is the viscosity
of water vapor, M
1
is the molar weight of water, P
fm
is the vapor
pressure at feed-membrane interface. Poiseuille/viscous ow varies
as the difference between the 2nd power of the vapor pressures
across the membrane, i.e. P
2
fm
P
2
P
, and Knudsen diffusion is di-
rectly proportional to the TMVP (P
fm
P
P
).
In the models reported above, there are a number of physical
properties of water and ethanol that need to be evaluated as a
function of changing temperature of along the ow channel in
membrane module. These properties of pure component at the
average temperature T
m
of the bulk feed temperature and the
membrane surface temperature were obtained from the physical
property database and listed as follows:
The density can be calculated by [43]:
q
i
a bT
m
cT
2
m
eT
3
m
fT
4
m
22
The viscosity was obtained by [43]:
l
i
a
0
b
0
T
m
c
0
T
2
m
e
0
T
3
m
23
The specic heat was calculated by [38]:
C
p;i
a
00
b
00
T
m
c
00
T
2
m
e
00
T
3
m
f
00
T
4
m
24
The regressed values of coefcients for pure component proper-
ties can be obtained in chemical database, and the density of eth-
anolwater mixture can be calculated by [44]:
1
q

w
1
q
1

w
2
q
2
25
The viscosity and specic heat of ethanolwater mixture were
calculated using the following mixture rule [45]:
Z
mix

X
n
i1
x
i
z
i
26
where Z
mix
represents mixture property, z
i
represents pure compo-
nent properties.
3.3. Simulation algorithm
Mathematical model of VMD is necessary for understanding the
process and fundamental basis for membrane module design and
industrial scale. Based on the analysis of heat and mass transfer
mechanisms in VMD process in theory section, the Knudsenvis-
cous transition model was employed (i.e. Eqs. (20) and (21)).
Simultaneous mass and heat transfer across the membrane in the
separation process causes the changes in concentration and tem-
perature at the vaporliquid interface along the ow channel. In
other words, both the transfer characteristics and separation per-
formances vary along with the ow direction. The main purpose
of applying the Knudsenviscous transition model is to predict
the values of the permeate ux, separation performance, and their
variations along the ow channel. The derivation of heat and mass
transfer equations is based on dividing the ow channel into a
number of elements in the module as shown in Fig. 4, which illus-
trates the mass and heat balance on each liquid element. Thus, a
trail of nonlinear equations with a number of unknown variables
should be solved using the known operation conditions including
the inlet temperature, inlet ow rate, ethanol concentration of
the feed solution and the downstream pressure. These operation
conditions and model parameters determined by experiments are
used as input data to the developed program written by C code uti-
lizing linear scanning coupled with dichotomy to solve the set of
nonlinear equations (Eqs. (1)(13) and (20) and (21)). The outputs
of the developed program for each element are: the permeate
uxes of water and ethanol, separation factor, the temperature
on the membrane surface on the feed side, the concentration of
ethanol, the physic properties of each liquid element.
To balance computational speed and simulation precision, the
error of the simulated total ux is smaller than 1%, the step length
(Dl) of 0.1 cm (200 elements) was selected. The program owchart
was presented in Fig. 5.
58 J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363
4. Results and discussion
4.1. Validation on models and program
The numerical simulation program developed in present work
was rst validated with some results reported by Lawson and
Lloyd [32] for pure water VMD using three kinds of membranes.
The model parameters calculated by the given membrane struc-
tural parameters presented in the Lawson and Lloyds work were
listed in Table 1, which were employed in the calculation based
on our simulation program. The total membrane area of the mem-
brane used in their work is 9.7 cm
2
and the square channel in their
module is 0.63 0.63 cm. The experiments were conducted by
Lawson and Lloyd at a constant feed ow rate of 1gpm and a con-
stant permeate pressure of 3 kPa, and the temperature was varied
between 30 and 70 C.
Our simulation results compared with the experimental and the
simulated data from the above-mentioned literature were pre-
sented in Fig. 6. There is an exponential increase in simulated per-
meate ux with the increase of the feed temperature which is due
to the exponential increase of vapor pressure of the feed solution
with temperature. These results are quite consistency with the lit-
erature experimental data for 3MA membrane and 3MB membrane
except for the ones at 74 C, even though the simulated values are
Fig. 5. Flowchart of simulation program.
Table 1
Values of membrane structure parameters in literature [30].
Membrane K
0
/(10
7
m) B
0
/(10
14
m
2
) d/(lm)
3MA 2.61 0.934 91
3MB 4.08 1.55 81
3MC 5.93 2.83 76
J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363 59
slightly greater than the experimental ones for 3MC, and the mag-
nitude of the deviation gradually increases with the increasing of
feed temperature. The discrepancies between the present simula-
tion and the experimental results reaches a maximum of 13.4%
at 74 C for 3MB membrane and a minimum of 1.4% at 42 C for
3MA membrane. The maximum discrepancy is mainly ascribed to
the accumulation of some experimental error as explained by Law-
son and Lloyd and simulation error. But it is noted here that the
present simulation values are in better accordance with the results
simulated by Lawson and Lloyd.
4.2. Water VMD experiments and determination of model parameters
Pure water VMD experiments were performed to determine the
parameters of K and B in model (20). The experimental data were
shown in Fig. 7. The VMD is a vapor pressure difference driven
membrane process. Increasing the feed temperature and/or
decreasing the vapor pressure in permeate side can enhance the
permeate ux. As can be seen, the permeate uxes increase rapidly
with the increase in the feed inlet temperature. In contrary, the
ux declines with the increase in the absolute pressure in perme-
ate side.
The pure water VMD experimental data were tted according to
Eq. (21). Quadratic relationships between ux (J) and vapor pres-
sure (P
P
) in the permeate side at different feed temperatures were
obtained and listed in Table 2. By comparing the tting equations
with Eq. (21), the values of the modied model parameters K and
B for the PTFE at-sheet membrane at different temperatures were
obtained.
Fig. 8 shows the relations between the model parameter (K or B)
and operating temperature T. An interesting result is that both the
values of K and B decrease linearly with the feed temperatures. The
primary reason is that the thermal expansion of membrane mate-
rial can cause decreases in pore size and porosity, but an increase
in thickness. If a membrane module is fully lled with a hot liquid
feed, membrane thermal expansion will take place. Araki [46] mea-
sured the linear thermal expansion of molded PTFE samples, the
results revealed that the change of length exhibited a large jump
when temperature increased from 25 C to 60 C. Furthermore,
the relationship between K or B and feed temperature was ob-
tained by linear tting, providing a method to estimate the model
parameters under different feed temperatures. Although the tting
results of parameters K and B as shown in Fig. 8 overlooked the
inuence of pressure difference across the membrane, we believe
that the trans-membrane hydrostatic pressure difference (TMHP)
also has some inuence on the membrane morphological parame-
ters. The change in TMHP can cause elastic deformation of mem-
brane supported by a porous plate covered with stainless steel
net. So the microstructures, such as pore size and porosity, will in-
crease with the TMHP. It should be noted that the TMHP decreased
with the assayed vapor pressure in permeate side at different feed
temperatures as shown in Fig. 7, for example, the TMHP decreased
about 8 kPa with the feed temperature from 45 C to 60 C. That is
to say that the obtained changes in modied model parameters K
and B mainly resulted from the coupling effect of feed temperature
and TMHP.
30 40 50 60 70 80
0
2
4
6
8
10
Feed inlet temperature/ t (
o
C)
W
a
t
e
r

f
l
u
x
/
N

(
m
o
l
e
/
(
m
2
s
)
)
Simulated water flux in this work
Experimental water flux by Lawson et al. 1996
Simulated water flux by Lawson et al. 1996
Membrane: 3MA
0
2
4
6
8
10
W
a
t
e
r

f
l
u
x

/
N
(
m
o
l
e
/
(
m
2
s
)
)
Membrane:3MB
simulated water flux in this work
experimenral water flux by Lawson et al. 1996
simulated water flux by Lawson et al. 1996
0
2
4
6
8
10
Membrane: 3MC
W
a
t
e
r

f
l
u
x
/
N

(
m
o
l
e
/
(
m
2
s
)
)
Simulated water flux in this work
Experimental water flux by Lawson et al. 1996
Simulated water flux by Lawson et al. 1996
30 40 50 60 70 80
Feed inlet temperature/ t (
o
C)
30 40 50 60 70 80
Feed inlet temperature/ t (
o
C)
(a)
(b)
(c)
Fig. 6. Water uxes of pure water VMD for validating the simulation program
membrane: (a) 3MA; (b) 3 MB; and (c) 3MC.
6 7 8 9 10 11 12 13 14 15
0
4
8
12
16
20
24
28 Experimental J Fitting J
60
o
C 60
o
C
55
o
C 55
o
C
50
o
C 50
o
C
45
o
C 45
o
C
W
a
t
e
r

f
l
u
x
/
J

(
k
g
/
(
m
2

h
)
)
Absolute pressure in the permeate side/P
p
(kPa)
Feed flow rate: 50 L/h
Fig. 7. Water ux variations of pure water VMD with vapor pressure in permeate
side at different feed temperatures.
60 J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363
However, it should be pointed out that the thermal expansion of
membrane matrix in water will be different from that in ethanol,
because PTFE is a highly hydrophobic material, yet can be wettable
by ethanol. Since only 5 wt% ethanol aqueous solution was em-
ployed, the changes of K and B with the used dilute ethanolwater
solutions were overlooked in present work.
4.3. Ethanolwater VMD experiments and simulation
4.3.1. Ethanolwater VMD experiments
VMD experiments for ethanolwater mixture separation at dif-
ferent temperatures and vacuum degrees were carried out. The rel-
evant evaluations describing the separation performance and
transfer characteristic are given by uxes of water and ethanol, to-
tal permeate ux and separation factor which are greatly affected
by changes in operating conditions, particularly, the feed tempera-
ture, and the vacuum degree as shown in Figs. 912. For all the
cases investigated here, increasing the feed temperature typically
results in an increase in the total permeate ux, correspondingly,
a remarkable decrease in separation factor occurs. Analogous con-
clusions are drawn for change the vacuum degree. Several factors
may account for this. Increasing the feed temperature and/or the
vacuum degree in permeate side can increase the TMVP, i.e. mass
transfer driving force. And the reduction of liquid viscosity caused
by the temperature raising can improve uidity of the feed, and en-
hanced turbulent movement, which decreases the mass transfer
resistance in the boundary layer of feed side. Consequently, the
permeate uxes were improved.
Table 2
Nonlinear tting results of water ux (J) vs. pressure in permeate side (P
P
) at different feed temperatures.
Temperature/C J/(kg/(m
2
h)) Correlation coefcient
45
J 11:1 10
8
P
2
P
9:4 10
4
P
P
24:540
0.9944
50
J 8:814 10
8
P
2
P
7:94 10
4
P
P
25:308
0.9926
55
J 7:12 10
8
P
2
P
7:26 10
4
P
P
30:839
0.9938
60
J 5:4 10
8
P
2
P
6:438 10
4
P
P
36:360
0.9899
318 320 322 324 326 328 330 332 334
0
2
4
6
8
10
12
B=-0.29148T + 101.65946
B
(

1
0
-
1
1
m
)
value of B
value of K
Linear Fit of B
Linear Fit of K
Feed inlet temperature/ T (K)
K=-0.04006T + 14.72279
1.2
1.4
1.6
1.8
2.0
2.2
2.4
K

(

1
0
-
3
)
Fig. 8. The variations of model parameters K and B with feed temperature.
89.0 89.5 90.0 90.5 91.0 91.5 92.0 92.5 93.0 93.5 94.0 94.5
0
4
8
12
16
20
24
28
Simulated J
45
o
C
50
o
C
55
o
C
60
o
C
Experimental J
45
o
C
50
o
C
55
o
C
60
o
C
T
o
t
a
l

f
l
u
x
/
J

(
k
g
/
(
m
2
h
)
)
Vacuum degree / P
v
(kPa)
Feed flow rate: 50 L/h
Fig. 9. Experimental and simulated total uxes of ethanolwater VMD at different
temperatures and vacuum degrees.
89 90 91 92 93 94
0
1
2
3
4
5
6
7
8
Feed flow rate: 50 L/h
Vacuum degree/P
v
(kPa)
F
l
u
x

o
f

e
t
h
a
n
o
l
/
J
2
(
k
g
/
(
m
2

h
)
)
Experimental J
2
45C
50C
55C
60C
Simulated J
2
45C
50C
55C
60C
Fig. 10. Experimental and simulated ethanol uxes of ethanolwater VMD at
different temperatures and vacuum degrees.
89 90 91 92 93 94
0
2
4
6
8
10
12
14
16
18
20
Feed flow rate: 50 L/h
Vacuum degree/P
v
(kPa)
F
l
u
x

o
f

w
a
t
e
r
/
J
1
(
k
g
/
(
m
2

h
)
)
Experimental J
1
45C
50C
55C
60C
Simulated J
1
45C
50C
55C
60C
Fig. 11. Experimental and simulated water uxes of ethanolwater VMD at
different temperatures and vacuum degrees.
J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363 61
In terms of separation factor, a monotonic decrease with the in-
crease of vacuum degree can be observed in Fig. 12. Moreover, the
separation factors increase to greater levels with the decline of
feed temperature. These trends are ascribed to that the feed tem-
perature and vacuum degree inuenced the mass transfer behav-
iors within membrane pores and the vaporliquid equilibrium of
ethanolwater system at interface. At the same feed temperature,
increasing vacuum degree resulted in the increase of mean free
path as shown in Fig. 4, which enhanced the Knudsen diffusion
within membrane pores. Imdakm and Khayet et al. [47] also found
that viscous ow is lower than the Knudsen type of ow by simu-
lating water VMD. Moreover, the ethanol molecules are bigger than
the water molecules, and have smaller mean free path. According
to the mass transfer Eq. (20) or Eq. (21), the effect of vacuum de-
gree on water in Knudsen diffusion is different from that on etha-
nol. The transport resistance of ethanol across membrane is larger
than that of water, which was also revealed by comparing the
slopes of ux vs. vacuum degree in Figs. 10 and 11. On the other
hand, as reported by Bandini and Sarti [8,13], in case of aqueous
mixtures containing VOC separation process for a given feed-ow
rate and feed temperature, the effects of overall polarization on
water ux is unlike that on ethanol ux as downstream pressure
decreases; Higher-pressure/lower-vacuum values result in a signif-
icant reduction of driving force for water ux, whereas the driving
force ethanol undergo only a smaller decrease. As a consequence,
the higher vacuum reduced the separation factor.
Under the same vacuum degree, any given variation in feed
temperature leads to different changes in the species vapor pres-
sure, and different effects on water ux and ethanol ux. Generally,
the mole fraction of ethanol vapor at the feed-vapor interface ther-
modynamically increased with the decreasing of feed temperature
for that the driving force of the ethanol decreased less than that of
water, which leads to an obvious increase in separation factor. By
contrast, this kind of decrease of feed temperature probably ex-
erted inuence so that the mean free path decreased slightly, and
weakened the percentage of Knudsen diffusion within membrane
pores. As the molar mass of the ethanol is larger with respect to
that of water, the Knudsen is certainly always unfavorable to the
process [13]. As a consequence, the separation factor was increas-
ing with temperature decrease.
These experimental results demonstrated that the VMD process
should be operated at relatively low temperature and/or vacuum
degree in order to obtain higher ethanol composition in the
distillate.
4.3.2. Ethanolwater VMD simulation
The modied model parameters (K and B) that were obtained by
tting the pure water VMD results were further applied to predict
the performances of the PTFE VMD module for separation of etha-
nolwater mixture. And the accuracy of model parameters was
evaluated by the VMD experimental results of 5 wt% ethanol aque-
ous solution. The comparisons between the simulated values and
VMD experimental data of ethanol aqueous solutions were also
shown in Figs. 912. The results show that all the calculated total
uxes and most partial ones are in good agreement with the mea-
sured ones. Among them, the theoretical ethanol uxes at 60 C
slightly overestimated the experimental ones, but the correspond-
ing uxes of water under the vacuum degrees of 91 kPa and 92 kPa
underestimated, perhaps caused by experimental errors. It is also
found that the separation factors experimentally measured de-
crease more rapidly with the increase of vacuum degree than the
simulated values. Meanwhile, the maximum and minimum dis-
crepancies are 9.0% and 0.5% for separation factors and total uxes,
respectively. From a practical point of view, these simulation re-
sults predicted the separation performances of the VMD process
quite well. VMD process of binary vaporizable component mixture
mainly concerns with improving the selectivity, therefore, it
should be operated at a low temperature and a relatively lower
trans-membrane pressure difference, as shown in Fig. 12.
The following factors are probably responsible for that the
experimental separation factors declined faster than the calculated
ones with the descent of permeate pressure. First, when the vac-
uum degree increases, the difference in the mean free paths of
water and of ethanol also increases, which caused the difference
in transfer behaviors between ethanol and water molecules as
interpreted above. Second, the ethanol molecules are bigger than
water, and diffuse at a slower speed in the membrane pores. How-
ever, these factors were not embodied in the obtained model
parameters of K and B.
5. Conclusions
The model parameters K
0
and B
0
in Knudsenviscous transition
model were modied. A novel reliable method using at-sheet
membrane module VMD experiments to determine the modied
model parameters K and B was proposed with the aim of making
the model equation more precisely. A numerical simulation de-
pended on Knudsenviscous transition model was presented to
predict the performances of the VMD process which is feasible
for not only at membrane module but also for hollow ber mem-
brane module. The simulation program was completely tested for
pure water VMD with the experimental results obtained in a liter-
ature. The accuracy of model parameters was further evaluated by
predicting the separation of ethanolwater mixture and comparing
with the experimental results. It can be concluded that the model
with the obtained parameters K and B described the VMD separa-
tion performances of ethanolwater mixture reasonably well. The
consistency between the simulated results and the experimental
data illustrated that the dependence of K or B on the feed solution
can be neglected in the assayed range, and proves that the pro-
posed method to determine the values of model parameters (K
and B) is reliable.
The relationships associating K or B and the feed temperature
provide a good way to predict the model parameters at different
temperatures with which the effect of TMHP on model parameters
coupled in the experiments assayed. The VMD process of binary
vaporizable component mixture, such as ethanolwater, mainly
concerns with improving the selectivity, consequently, it is recom-
mended to operate at a low temperature and a relatively lower
TMVP.
89.0 89.5 90.0 90.5 91.0 91.5 92.0 92.5 93.0 93.5 94.0 94.5
5
6
7
8
9
10
11
12
13
Simulated experimental
45
o
C 45
o
C
50
o
C 50
o
C
55
o
C 55
o
C
60
o
C 60
o
C
S
e
p
e
r
a
t
i
o
n

f
a
c
t
o
r
/

Vacuum degree/ P
v
(kPa)
Feed flow rate: 50 L/h
Fig. 12. Experimental and simulated separation factors of ethanolwater VMD at
different temperatures and vacuum degrees.
62 J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363
Acknowledgment
The authors would like to thank the support of the National
Natural Science Foundation of China (Nos. 21276024 and
20976012).
References
[1] H. Yu, X. Yang, R. Wang, A.G. Fane, Numerical simulation of heat and mass
transfer in direct membrane distillation in a hollow ber module with laminar
ow, J. Membr. Sci. 384 (2011) 107116.
[2] M.M. Teoh, T.S. Chung, Y.S. Yeo, Dual-layer PVDF/PTFE composite hollow bers
with a thin macrovoid-free selective layer for water production via membrane
distillation, Chem. Eng. J. 171 (2011) 684691.
[3] M. Gryta, M. Barancewicz, Inuence of morphology of PVDF capillary
membranes on the performance of direct contact membrane distillation, J.
Membr. Sci. 358 (2010) 158167.
[4] F.A. Banat, F. Abu Al-Rub, R. Jumah, M. Al-Shannag, Application of Stefan
Maxwell approach to azeotropic separation by membrane distillation, Chem.
Eng. J. 73 (1999) 7175.
[5] K. Yao, Y.J. Qin, Y.J. Yuan, L.Q. Liu, F. He, Y. Wu, A continuous-effect membrane
distillation process based on hollow ber AGMD module with internal latent-
heat recovery, AIChE J. 59 (2012) 12781297.
[6] C. Cojocaru, M. Khayet, Sweeping gas membrane distillation of sucrose
aqueous solutions: response surface modeling and optimization, Sep. Purif.
Technol. 81 (2011) 1224.
[7] C.A. Rivier, M.C. Garca-Payo, I.W. Marison, V. von Stockar, Separation of binary
mixtures by thermostatic sweeping gas membrane distillation: I. Theory and
simulations, J. Membr. Sci. 201 (2002) 116.
[8] S. Bandini, G.C. Sarti, Heat and mass transport resistances in vacuum
membrane distillation per drop, AIChE J. 45 (1999) 14221433.
[9] Z.P. Zhao, C.Y. Zhu, D.Z. Liu, W.F. Liu, Concentration of ginseng extracts aqueous
solution by vacuum membrane distillation 2. Theory analysis of critical
operating conditions and experimental conrmation, Desalination 267 (2011)
147153.
[10] S. Bandini, C. Gostli, G.C. Sarti, Separation efciency in vacuum membrane
distillation, J. Membr. Sci. 73 (1992) 217229.
[11] G.C. Sarti, C. Gostoli, S. Bandini, Extraction of organic components from
aqueous streams by vacuum membrane distillation, J. Membr. Sci. 80 (1993)
2133.
[12] M.C. Porter, Concentration polarization in membrane ultraltration, I&EC Prod.
Res. Dev. 11 (1972) 234248.
[13] S. Bandini, A. Saavedra, G.C. Sarti, Vacuum membrane distillation: experiments
and modeling, AIChE J. 43 (1997) 398408.
[14] F. Banat, F. Abu Al-Rub, K. Bani-Melhem, Desalination by vacuum membrane
distillation: sensitivity analysis, Sep. Purif. Technol. 33 (2003) 7587.
[15] H.W. Fan, Y.L. Peng, Application of PVDF membranes in desalination and
comparison of the VMD and DCMD processes, Chem. Eng. J. 79 (2012) 94102.
[16] J.P. Mericq, S. Laborie, C. Cabassud, Evaluation of systems coupling vacuum
membrane distillation and solar energy for seawater desalination, Chem. Eng.
J. 166 (2011) 596606.
[17] X.Y. Wang, L. Zhang, H.J. Yang, H.L. Chen, Feasibility research of potable water
production via solar-heated hollow ber membrane distillation system,
Desalination 247 (2009) 403411.
[18] N. Coufn, C. Cabassud, V. Lahoussine-Turcaud, A new process to remove
halogenated VOCs for drinking water production: vacuum membrane
distillation, Desalination 117 (1998) 233245.
[19] Z.W. Ding, L.Y. Liu, Z.M. Li, R.Y. Ma, Z.R. Yang, Experimental study of ammonia
removal from water by membrane distillation (MD): the comparison of three
congurations, J. Membr. Sci. 286 (2006) 93103.
[20] S. Al-Asheh, F. Banat, M. Qtaishat, M. Al-Khatee, Concentration of sucrose
solutions via vacuum membrane distillation, Desalination 195 (2006) 6068.
[21] M. Gryta, M. Tomaszewska, K. Karakulski, Wastewater treatment by
membrane distillation, Desalination 198 (2006) 6773.
[22] F. Banat, S. A1-Asheh, M. Qtaishat, Treatment of waters colored with
methylene blue dye by vacuum membrane distillation, Desalination 174
(2005) 8796.
[23] M.A. Izquierdo-Gil, G. Jonsson, Factors affecting ux and ethanol separation
performance in vacuum membrane distillation (VMD), J. Membr. Sci. 214
(2003) 113130.
[24] Z.P. Zhao, F.W. Ma, W.F. Liu, D.Z. Liu, Concentration of ginseng extracts
aqueous solution by vacuum membrane distillation. 1. Effects of operating
conditions, Desalination 234 (2008) 152157.
[25] D.A. Salvi, G.M. Aita, D. Robert, V. Bazan, Ethanol production from sorghum by
a dilute ammonia pretreatment, J. Ind. Microbiol. Biot. 37 (2010) 2734.
[26] T. Zhang, Z. Chi, C.H. Zhao, Z.H. Chi, F. Gong, Bioethanol production from
hydrolysates of inulin and the tuber meal of Jerusalem artichoke by
Saccharomyces sp. W0, Bioresour. Technol. 101 (2010) 81668170.
[27] X.L. Han, L. Wang, J.D. Li, X. Zhan, J. Chen, J.C. Yan, Separation of ethanol from
ethanol/water mixtures by pervaporation with silicone rubber membranes:
effect of silicone rubbers, J. Appl. Polym. Sci. 119 (2011) 34133421.
[28] S. Ulutan, T. Nakagawa, Separability of ethanol and water mixtures through
PTMSP-silica membrane in pervaporation, J. Membr. Sci. 143 (1998) 275284.
[29] G. Lewandowicz, W. Biaas, B. Marczewski, D. Szymanowska, Application of
membrane distillation for ethanol recovery during fuel ethanol production, J.
Membr. Sci. 375 (2011) 212219.
[30] M. Gryta, A.W. Morawski, M. Tomaszewska, Ethanol production in membrane
distillation bioreactor, Catal. Today. 56 (2000) 159165.
[31] M. Khayet, T. Matsuura, Pervaporation and vacuum membrane distillation
processes: modeling and experiments, AIChE J. 50 (2004) 16971712.
[32] K.W. Lawson, D.R. Lloyd, Membrane distillation. I. Module design and
performance evaluation using vacuum membrane distillation, J. Membr. Sci.
120 (1996) 111121.
[33] F.M. Allan, N. Qatanani, I. Barghouthi, K.M. Takatka, Dusty gas model of ow
through naturally occurring porous media, Appl. Math. Comput. 148 (2004)
809821.
[34] C.M. Guijt, I.G. Rcz, T. Reith, A.B. de Haan, Determination of membrane
properties for use in the modeling of a membrane distillation module,
Desalination 132 (2000) 255261.
[35] K.W. Lawson, M.S. Hall, D.R. Lloyd, Compaction of microporous membranes
used in membrane distillation. I. Effect on gas permeability, J. Membr. Sci. 101
(1995) 99108.
[36] U. Beuscher, C.H. Gooding, Characterization of the porous support layer of
composite gas permeation membranes, J. Membr. Sci. 132 (1997) 213227.
[37] J.I. Mengual, M. Khayet, M.P. Godino, Heat and mass transfer in vacuum
membrane distillation, Int. J. Heat. Mass. Tran. 47 (2004) 865875.
[38] B.E. Poling, J.M. Prausnitz, J.P. O
0
Connell, The properties of gases and liquid, 5th
ed., McGraw-Hill Inc., New York, 2001.
[39] M. Khayet, K.C. Khulbe, T. Matsuura, Characterization of membranes for
membrane distillation by atomic force microscopy and estimation of their
water vapor transfer coefcients in vacuum membrane distillation process, J.
Membr. Sci. 238 (2004) 199211.
[40] E.L. Cussler, Diffusion, mass transfer in uid system, 2nd ed., Cambridge
University Press, New York, 1997.
[41] Y.G. Yan, T. Bein, Molecular sieve sensors for selective ethanol detection,
Chem. Mater. 4 (1992) 975977.
[42] T. Matsuura, Synthetic membranes and membrane separation processes, CRC
Press, FL, Boca Raton, 1993.
[43] P.S. Ma, Chemical engineering databook, China Petrochemical Press, Beijing,
2003.
[44] M. Celere, C. Gosto, The heat and mass transfer phenomena in osmotic
membrane distillation, Desalination 147 (2002) 133138.
[45] V. Soni, J. Abildskov, G. Jonsson, R. Gani, Modeling and analysis of vacuum
membrane distillation for the recovery of volatile aroma compounds from
black currant juice, J. Membr. Sci. 320 (2008) 442455.
[46] Y. Araki, Thermal expansion coefcient of polytetrauoroethylene in the
vicinity of its glass transition at about 400K, J. Appl. Polym. Sci. 9 (1965) 421
427.
[47] A.O. Imdakm, M. Khayet, T. Matsuura, A Monte Carlo simulation model for
vacuum membrane distillation process, J. Membr. Sci. 306 (2007) 341348.
J.-Y. Shi et al. / Separation and Purication Technology 123 (2014) 5363 63

You might also like