Sunteți pe pagina 1din 11

International Journal of Heat and Mass Transfer 53 (2010) 1249–1259

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


journal homepage: www.elsevier.com/locate/ijhmt

Monte Carlo simulation and experimental heat and mass transfer in direct
contact membrane distillation
M. Khayet a,*, A.O. Imdakm b, T. Matsuura b
a
Department of Applied Physics I, Faculty of Physics, University Complutense of Madrid, Avda. Complutense s/n, 28040 Madrid, Spain
b
Industrial Membrane Research Center, Department of Chemical Engineering, University of Ottawa, Ont., Canada K1N 6N5

a r t i c l e i n f o a b s t r a c t

Article history: A Monte Carlo (MC) simulation model is developed to study heat and mass transfer through hydrophobic
Received 9 October 2008 membranes applying direct contact membrane distillation (DCMD) process. In this study, the membrane
Accepted 5 December 2009 pore space is described by a three-dimensional network model of inter-connected cylindrical pores with
Available online 12 January 2010
distributive pore size. Vapor flux through membrane pores is described by gas transport mechanism(s)
based on the kinetic theory of gases. The present MC model can take into consideration the influence
Keywords: of temperature polarization phenomenon, membrane physical properties including pores interconnectiv-
Monte Carlo
ity and the DCMD fluid dynamic conditions. The developed model can simultaneously predict the DCMD
Direct contact membrane distillation
Network model
process vapor flux and membrane surface temperatures, contrary to other models in which one of them
Heat and mass transfer has to be given in order to solve the other. The model is comprehensive in its approach and does not
Temperature polarization involve any adjustable parameter. The simulated results were compared with the experimental ones of
different membranes and the comparisons were found to be in excellent qualitative and quantitative
agreement.
Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction The correct description of fluid dynamics (laminar or turbulent)


in the MD modules and its influence on the energy transport at the
Membrane distillation (MD) is a thermally driven separation membrane boundaries will affect heat transfer coefficients and the
process in which liquid feed evaporates at feed/membrane inter- resultant interfacial surface temperatures. These heat transfer
face and is transported through membrane pores following differ- coefficients are often calculated by means of dimensionless Nusselt
ent transport mechanisms before condensation step, which can number (Nu), which is most of the time estimated from experi-
occur at permeate/membrane interface or outside the membrane mentally determined empirical correlations depending on process
module [1–13]. In this process, hydrophobic membranes must be flow conditions and MD module design [6–11]. Hence, it is very
used, and therefore, only water vapor and volatile components in essential to be particularly cautious in using the empirical heat
the feed are transported through membrane pores and are con- transfer correlations to describe the heat flux in the MD module.
densed as cooled distillate stream at the permeate side of the The proper selection of these correlations will facilitate signifi-
membrane. The non-wettability criterion of the hydrophobic cantly the accurate calculation of the interfacial surface tempera-
membrane will prevent the penetration of liquid feed into mem- tures as well as the heat and mass fluxes at the membrane
brane pores. boundaries.
MD process operates based on the principle of vapor–liquid In direct contact membrane distillation (DCMD) liquid feed and
equilibrium (VLE) conditions and both heat and mass transfer oc- liquid permeate are always brought in direct contact with mem-
cur simultaneously. The driving force for the mass transport brane surfaces as shown in Fig. 1. The temperature gradient caused
through membrane pores is the vapor pressure difference across by the resistance to the heat flux across the membrane boundary
the membrane, which results from the difference in composition layers means that temperatures, ts,f and ts,p at the membrane–li-
and temperature of the solutions in the layers adjoining mem- quid interfaces, are different from bulk solution temperatures, tb,f
brane–liquid interfaces. The temperatures at the membrane–liquid and tb,p known as temperature polarization, it causes losses of
interfaces cannot be measured directly and their values are af- thermal energy available for vapor flux in DCMD process. The tem-
fected significantly by membrane physical properties and flow perature polarization coefficient, h, is defined in the literature as
conditions in the MD module (process dynamics). [1–13]:

t s;f  ts;p
* Corresponding author. Tel.: +34 91 3945185; fax: +34 91 3945191. h¼ ð1Þ
E-mail address: khayetm@fis.ucm.es (M. Khayet). t b;f  tb;p

0017-9310/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijheatmasstransfer.2009.12.043
1250 M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259

Nomenclature

As membrane area (m2) qm


s conduction heat transfer rate across membrane solid
C membrane distillation coefficient (kg m2 s1 Pa1) phase (polymer) at Z = 0 (Wm2)
Dpore pore diameter (m) qpconv : convective heat transfer rate across membrane-perme-
hf and hp heat transfer coefficient at feed and permeate side ate boundary layer (Wm2)
(Wm2 K1) qpHl heat transfer rate due water flux across membrane-per-
Hl{T} liquid water enthalpy at temperature T (kJ kg1) meate boundary layer (Wm2)
Hv{T} saturated vapor water enthalpy at temperature T Qf total heat transfer rate across membrane-feed boundary
(kJ kg1) layer (Wm2)
J vapor flux (kg m2h1) Qm total heat transfer rate across the membrane (Wm2)
l pore length (lm) Qp total heat transfer rate across membrane-permeate
L network size (number of nodes ‘‘sites”) boundary (Wm2)
km membrane thermal conductivity (Wm1 K1) rnk pore radius (attached to the nth node) in direction of
ks membrane solid phase (polymer) thermal conductivity flow, Z, (m)
(Wm1 K1) Sp membrane surface porosity
kv membrane vapor phase (water vapor) thermal conduc- tb,f and tb,p bulk temperatures at feed and permeate sides, respec-
tivity (Wm1 K1) tively (°C)
Pb,f and Pb,p vapor pressure at feed and permeate side, respec- ts,f and ts,p interfacial membrane-surface temperatures at feed
tively (Pa) and permeate sides, respectively (°C)
Ps,f and Ps,p vapor pressure at membrane-feed and membrane- T absolute temperature (K)
permeate interface, respectively (Pa) X, Y, and Z axial coordinates
qmcond: conduction heat transfer through the membrane
(Wm2) Greek letters
qpconv : convection heat transfer rate across membrane-feed lp mean pore size of the pores (nm)
interface (Wm2). rp geometric standard deviation of pore size distribution
qpHl heat transfer due to liquid flux across membrane-feed d membrane thickness (m)
boundary layer (Wm2) e membrane void volume (volume porosity)
qmHv heat transfer due to vapor flow through membrane h temperature polarization coefficient
pores (Wm2) np percolation correlation length (dimensionless)
qmp conduction heat transfer rate across vapor phase at the
entrance of the membrane pores, Z = 0 (Wm2)

which heat transfer at the membrane boundaries will be limited.


(a) Membrane
Permeate Therefore, the effect of temperature polarization on DCMD trans-
port process can be reduced to a certain degree by enhancing heat
transfer coefficients hf and hp, which reduces the temperature gra-
Mass flux, J dients in the boundary layers leading to an increase in the vapor
(tb,f ; Pb,f)
Heat flux, Q flux. This can be accomplished by increasing liquid turbulence
hf employing net-shaped turbulence promoters, high stirring rate,
(ts,p ; Ps, p) spacers, etc. In fact experimental studies show that the presence
of spacers enhances the vapor flux [10,11].
(ts,f ; Ps,f) Heat transfer analysis of the DCMD process cited in the litera-
(tb,p ; Pb,p,)
ture shows that the feed and permeate hydrodynamic conditions
hp (laminar or turbulence) can significantly affect the energy trans-
Feed ferred flux through the boundary layers [1–13]. The heat transfer
Z=0 1 2 ….. L L+1 coefficients of the feed (hf) and permeate (hp) are obtained from
the knowledge of the hydrodynamic conditions of the fluids at
Permeate
(b) Feed region Membrane
region region
the membrane boundaries. The heat transfer rate across the mem-
q f
q m p
q conv brane boundary layers, Qf and Qp, are calculated by means of the
conv . cond . .
following equations [10–13]:

tb,f tb, p Q f ¼ hf ðtb;f  t s;f Þ; ð2Þ

and
q Hf l ts f q Hmv ts, p q Hp l
Q p ¼ hp ðts;p  t b;p Þ: ð3Þ
Fig. 1. (a) Heat and mass transfer through hydrophobic membrane in DCMD
process and (b) heat transfer model electrical analogy in DCMD process [8,9].
The total heat flux across the membrane, Qm., is the sum of the heat
conducted through the membrane solid phase (i.e. polymer) and
vapor occupying membrane pores and the heat carried out by the
This equation represents the ratio of the actual driving force that is vapor that flows through membrane pores (latent heat of vaporiza-
contributing to the heat transfer (ts,fts,p) to the applied driving tion). Accordingly, the heat flux across the membrane is given by
force (tb,f tb,p) (see Fig. 1a and b). [1,4,5,7,11–13]:
For a well designed system h approaches unity, which would
happen when ts,f ? tb,f as hf ? 1 and ts,p ? tb,p as hp ? 1. On the km
Q m: ¼ Jk þ ðt s;f  t s;p Þ; ð4Þ
other hand, h approaches zero for a poorly designed system in d
M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259 1251

where km is the effective thermal conductivity of the membrane, d is 0.01


the membrane thickness, J is the vapor flux, and k is the latent heat

Probability density function


of vaporization.
0.008 Membrane A
Heat transfer analysis based on the assumption of non-linear

df(dp)/d(d p) , (nm )
-1
Membrane B
temperature profile across the membrane and non-isoenthalpy of
0.006 Membrane C
flow vapor was reported [6,8,9]. In this analysis the heat flux across
the membrane was written as [6,8,9]:
dT 0.004
Q mem: ¼ JHv fTg  km ; ð5Þ
dZ
0.002
where Z is the distance to the direction of vapor flux and Hv{T} is the
enthalpy per unit mass of vapor at temperature T.In both cases dis-
cussed above Eqs. (4) and (5), the vapor flux, J, is either measured 0
0 200 400 600 800 1000
experimentally [7–9] or calculated by different models developed
Pore size, D p , (nm)
for the DCMD process [1–3,10,11]. However, in many studies, pro-
cess vapor fluxes were calculated by the following expression Fig. 2. Probability density function curves for membranes A (lp = 200 nm), B
[1,4,9–11]: (lp = 250 nm) and C (lp = 300 nm), with a constant geometric standard deviation
(rp = 1.25).
J ¼ CðP f  Pp Þ; ð6Þ
where Pf and Pp is the vapor pressure of transporting fluid at the
where H{T} is the transporting fluid enthalpy per unit mass at tem-
membrane feed and permeate side, respectively and C is the
perature T, h is the heat transfer coefficient and DT is the tempera-
membrane distillation coefficient, which can be determined
ture difference across each region of interest. The above equation
experimentally, and it is a function of membrane physical proper-
describes heat fluxes in all three regions (feed, membrane and per-
ties (pore size, porosity, thickness, and membrane pore tortuosity),
meate). The energy fluxes across the boundary layers (feed and per-
transporting fluid physical properties (molecular weight and
meate) become:
diffusivity) and operating temperature. Once the vapor flux, J, is ob-
tained, it is substituted into the respective heat transfer equations
Q f ¼ qfHl þ hf ðt b;f  ts;f Þ; ð8Þ
to evaluate the membrane effective surface temperatures. The cal-
culation of membrane surface temperatures in many cases is car- and
ried out iteratively [7–9].
In the present study, a Monte Carlo (MC) simulation model is Q p ¼ qpHl þ hp ðt s;p  tb;p Þ; ð9Þ
developed to describe MD transporting process and to predict
MD vapor flux (permeability). The model can take into consider- where qfHl and qpHl are the heat fluxes associated to the liquid water
ation the effects on DCMD process performance and the resultant transport across the boundary layers at the feed and permeate side,
vapor flux calculation of the temperature polarization phenome- respectively. These heat fluxes are equal to JHl{T}, where J is the
non, the membrane physical properties, including membrane pores DCMD vapor flux and Hl{T} is the enthalpy per unit mass of liquid
interconnectivity, the DCMD operating conditions and the hydro- water at temperature T and is written in the temperature range
dynamic conditions of liquids in both sides of the membrane (mod- 273 K–373 K as follows [19]:
ule design). Comparisons between the MC simulated DCMD fluxes
and the experimental ones are carried out to demonstrate the Hl fTg ¼ 1117:8 þ 4:0312T þ 2:  104 T 2 ; ð10Þ
models’ consistency.
where T is assumed to be equal to an average value of the temper-
atures of bulk liquid and the membrane surface of interest (perme-
2. Theory and calculation procedure
ate or feed).
In Fig. 1b, hf(tb,f – ts,f) and hp(ts,p – tb,p) are, respectively, desig-
The membrane pore space is described by a three-dimensional
nated as qfconv : and qpconv : . The total heat flux across the membrane,
network of inter-connected pores (bonds) and nodes (sites). It is a
Qm, is the sum of qm Hv and the total heat conducted through the
simple cubic network model (L  L  L), where L is the network
membrane (solid and vapor phase), qm cond: , as shown in Fig. 1b and
size, and it is selected such that L >> np, where np is network perco-
therefore the total heat flux across the membrane is:
lation correlation length [15,16]. The network boundaries are L  L
and represent membrane boundaries (feed and permeate). The 1st Q m ¼ qm m m
H v þ qp þ qs ð11Þ
network section (layer) at Z = 0 represents membrane feed side and
is assumed to be exposed to feed boundary layer conditions (Ps,f In the above equation, qm Hv is the heat flux associated to water vapor
and ts,f), and the last network section (layer) at Z = L + 1 represents transport across the membrane pores, and it can be written as:
the permeate side and is assumed to be exposed to permeate qm
Hv ¼ JHv fTg, where Hv{T} is the enthalpy per unit mass of saturated
boundary layer conditions (Ps,p and ts,p). More detailed description water vapor at temperature T, where T is assumed to be equal to the
of network model formulation, including membrane pore size dis- average of the temperatures at the bulk feed and at membrane sur-
tribution and vapor flux transport mechanism(s) governing vapor T þT
face on the feed side b;f 2 s;f . For water saturated vapor Hv{T} is writ-
flux through membrane pores were discussed elsewhere [14]. ten as follows [19]:
The geometrical pore sizes, standard deviations and the corre-
sponding probability density curves for three hypothetical mem- Hv fTg ¼ 1850:7 þ 2:8273T  1:6  103 T 2 ð12Þ
branes used in this study (A, B, and C) are given in Fig. 2.
Fig. 1a shows schematically mass and heat fluxes in DCMD pro- The second term of Eq. (11), qm p ; means the total heat conducted
cess. The total heat flux can be written in general for each region through water vapor occupying all the entrance pores at Z = 0 (see
(feed, membrane and permeate) as follows [8,9]: Fig. 3). Assuming creeping flow across these pores (i.e., ignoring
possible contribution of convection heat transfer, although it can
Q ¼ JHfTg þ hDT; ð7Þ be included if necessary), qm
p can be written as:
1252 M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259

<t >
ts,f q sm ks
Pore

m
Pore qH nth node (site)
v kv
entrance q mpore
Pore
( t n , pn )

Feed
ts,f 1 μm Pore
<t >
1st Section (layer) 2nd Section (layer)

Z=0 Z=1

Fig. 3. Heat transfer through membrane solid surface and membrane pores in direct contact with the feed (feed side).

L2
kv X 3. Results of monte carlo simulation
qm
p ¼ ðt s;f  tn Þpr 2n ð13Þ
As l n¼1
The MC simulation model was applied for a DCMD process in
which water vapor is assumed to be the only permeate, thus phys-
where kv is the thermal conductivity of vapor, l is the average mem-
ical properties of the transporting media are assumed to be those
brane pore length, As is membrane surface area, rn is the pore radius
of water vapor. The influence of the following factors was investi-
(in the direction of flow, z) attached to the nth node on the plane
gated because of its importance in the DCMD process: temperature
Z = 0, and tn is the temperature of the nth node at the plane Z = 1
polarization, feed temperature, membrane pore size distribution
as shown in Fig. 3. The summation is carried out for all L2 number
and pore level transport mechanism(s). The effect of flow condi-
of pores at Z = 0. The third term, qm s , which means the heat con-
tions in the DCMD module (module design) on the process perfor-
ducted through membrane solid phase (polymer) in direct contact
mance and the resultant vapor flux was simulated by applying
with the liquid feed (at Z = 0), can be calculated by:
various values of heat transfer coefficients, hf and hp. The range
ks in which these transfer coefficients were tested was based on
qm
s ¼ ðt s;f  htiÞð1  Sp Þ ð14Þ experimentally reported heat transfer coefficients for different
l
flow conditions in DCMD modules [2,9].
where ks is the membrane solid phase (polymer) thermal conduc- Accordingly, the MC simulation model is applied for the follow-
tivity, Sp is the membrane surface porosity and hti is the average no- ing three cases: (a) hf and hp, are assumed to be equal and they are
PL2 in a range of 5000–25,000 Wm2 K1. (b) hf is varied in a range of
tn
dal temperatures at Z = 1, hti ¼ Ln¼1
2 . 2500–50,000 Wm2 K1, while hp is held constant at a given value
It should be noted as schematized in Fig. 1b that qm m m
cond: ¼ qp þ qs
of 25,000 Wm2 K1. (c) hp is varied in a range of 2500–
and the thermal conductivity of the membrane, km, can be evalu- 50,000 Wm2 K1 and hf is held constant at 25,000 Wm2 K1.
ated from the vapor and the solid phase thermal conductivities The bulk temperature of the feed is varied in a range of 40–90 oC,
as [1–11]: km = ekv + (1e)ks, where e is the membrane void volume while permeate temperature is maintained at 20 oC. Membrane
(volume porosity). surface porosity, Sp, membrane solid phase thermal conductivity,
Once the geometrical configuration of the porous membrane, ks, water vapor thermal conductivity, kv, are taken as 3.5%,
DCMD process operating conditions (i.e. the temperature of the 0.04 Wm1 K1 and 0.02 Wm1 K1, respectively. The network
bulk liquid and the heat transfer coefficient on both sides of the size, L = 12 so that each network contains over 5200 pores, and
membrane) and membrane physical properties are given, the heat the results averaged over 20 realizations.
flux across membrane boundary layers at steady state, Qf and Qp, Fig. 4a and b shows, respectively, the change in ts,f and ts,p, with
the membrane surface temperatures, ts,f and ts,p, and the energy in- the change of heat transfer coefficients, hf and hp, for membrane A
put into the membrane, Qm, can be computed. The calculation pro- and C. Different feed temperatures were used in this calculation. As
cedure applied in this study is iterative, using proper iterative can be seen, surface temperatures approach their corresponding
calculation method such as the method of successive substitutions bulk temperatures with an increase in heat transfer coefficient
[18]. An initial guess for membrane surface temperatures, ts,f and (i.e., ts,f ? tb,f when hf increases and ts,p ? tb,p when hp increases).
ts,p, is given to calculate vapor pressure at the plane Z = 0 and There is a little effect of membrane pore size on ts,f and ts,p when
Z = L + 1 (feed, and permeate side), using Antoine equation only Knudsen diffusion is taken into consideration. However, when
[13,14]. Afterward, network nodal pressure and temperature dis- viscous flow is considered, the effect of membrane pore size on ts,f
tribution and the corresponding vapor flux, J, can be calculated increase and ts,p decrease is more obvious for membrane C having
as discussed elsewhere [14]. Once the vapor flux is calculated, greater average pore size. This is due to the increase in the fraction
the energy fluxes, Qm, Qf, and Qp can be determined as discussed of membrane pores governed by viscous flow as the membrane
above. As a result, the relative energy error (REE) [17], which is de- pore size increases from A to C. This fact will be discussed later on.
fined as REE = (Qf  Qm)/Qf, can be determined assuming no energy Fig. 5 shows vapor flux versus feed temperature for membrane
losses to the surrounding. If REE exceeds or becomes equal to 2%, A and C. When only Knudsen diffusion is taken into consideration,
membrane surface temperatures, ts,f and ts,p, are re-calculated the vapor flux increases with the increase of the heat transfer coef-
using membrane surface temperature, (ts,f)m, obtained in the pre- ficients and approaches its maximum when hf and hp approach
ceding calculation such that: (ts,f)m+1 = (ts,f)m(1 + REE). Correspond- infinity [4]. This is expected since the membrane surface tempera-
ingly, ts,p can be calculated using Eqs. (8) and (9). Note that REE ? 0 tures approach their respective bulk values when the heat transfer
as ts,f and ts,p approach their steady state values. This calculation coefficients increase. However when viscous flow is taken into con-
procedure is repeated until REE becomes below 2%. sideration, a significant reduction in the vapor flux is observed at
M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259 1253

(a) 120

Feed temperature (ºC)


♦50 ºC Δ 70 ºC 80 ºC o 90 ºC
100

80
Temperature (ºC)

t s,f

60

40

} t s,p
20

0
0 5 10 15 20 25 30
Heat transfer coefficient, h f and h p , (103 W/m2.K)

120
(b)
Feed temperature (ºC)
♦50 ºC Δ 70 ºC 80 ºC o 90 ºC
100

80
Temperature (ºC)

t s,f

60

40

} t s,p
20

0
0 5 10 15 20 25 30
3 2
Heat transfer coefficient, h f and h p , (10 W/m .K)

Fig. 4. Membrane surface temperatures versus heat transfer coefficients (hf and hp) at different feed bulk temperatures and a bulk permeate temperature of 20 oC. (—) process
controlled by Knudsen diffusion only and (- - - -) process controlled by Knudsen diffusion and viscous flow: (a) membrane A and (b) membrane C.

higher feed temperatures. This is again due to an increase in the tb,f, and the transition from the Knudsen mechanism to the com-
fraction of membrane pores governed by the viscous flow with bined transport of Knudsen and viscous flow mechanism clearly
an increase in the membrane surface temperature of the feed side. coincide with the increase in the fraction of membrane pores gov-
Fig. 6 shows the temperature polarization coefficient, h, vs. the erned by viscous flow as shown in Fig. 7.
heat transfer coefficients for membrane A (Fig. 6a) and for mem- When the flow condition in the DCMD module can be varied
brane C (Fig. 6b). As can be seen for both membranes, when only only in one side of the membrane, the variation in the heat transfer
Knudsen diffusion is taken into consideration, h increases with coefficient is assumed to be as well occurring only in one side of
the increase of the heat transfer coefficients. It will approach unity the membrane. Fig. 8a shows the surface temperature vs. the feed
with hf and hp ? 1. When viscous flow is taken into consideration, side heat transfer coefficient for different feed temperatures (i.e., hf
h also increases with the increase of the heat transfer coefficient, increases while hp is maintained at 25,000 Wm2 K1). On the
but at relatively slower rate. These reductions of h with the in- other hand, in Fig. 8b, hp is changed while hf was maintained at
crease in membrane pore size (membrane C), feed temperature, 25,000 Wm2 K1. In Fig. 8a when only Knudsen diffusion is taken
1254 M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259

25 0.04
(a)

Viscous fraction of network pores


Feed bulk temperature (ºC)
Heat transfer coefficients (h f and h p ), (W/m2.K) ♦70 ºC 80 ºC o 90 ºC
20
Δ 5000 10000 o 25000 ♦ → ∞
J (kg/h.m )
2

15

0.02
10

0
20 40 60 80 100
Feed bulk temperature (ºC) 0
0 5 10 15 20 25 30
Heat transfer coefficient, h f and h p , 103 W/m2.K
35
(b)
Fig. 7. Fraction of network pores governed by viscous flow versus heat transfer
30 Heat transfer coefficients (h f and h p ) (W/m2.K) coefficients (hf and hp) at different bulk feed temperatures and a bulk permeate
Δ 5000 10000 o 25000 ♦ → ∞ temperature 20 oC. (—) membrane A and (- - - -) membrane C.
25
J (kg/h.m2)

20
(a) 120
Bulk feed temperature (ºC)
15
100 ♦50 ºC Δ 70 ºC 80 ºC o 90 ºC

10

Temperature (ºC)
80
t s,f
5
60
0
20 40 60 80 100 40
Feed bulk temperature (ºC)
} t s,p
20
Fig. 5. DCMD vapor flux versus feed bulk temperature for different values of
heat transfer coefficients (hf and hp) and a bulk permeate temperature 20 oC.
0
(—) process controlled by Knudsen diffusion only and (- - - -) process 0 10 20 30 40 50 60
controlled by Knudsen diffusion and viscous flow: (a) membrane A and (b) 3 2
Heat transfer coefficient, h f , (10 W/m .K)
membrane C.

120
(b) Bulk feed temperature (ºC)
100 ♦50 ºC Δ 70 ºC 80 ºC o 90 ºC
1
Temperature (ºC)

80
t s,f
0.8
60

0.6
Feed bulk temperature (ºC) 40
θ

♦ 50 ºC Δ 70 ºC 80 ºC o 90 ºC
0.4 20 } t s,p

0.2 0
0 10 20 30 40 50 60
Heat transfer coefficient, h p , (103 W/m2.K)
0
0 5 10 15 20 25 30 Fig. 8. Interfacial surface temperature versus heat transfer coefficients for mem-
Heat transfer coefficient, h f and h p , (103 W/m2.K) brane B at different bulk feed temperatures and a bulk permeate temperature 20 oC.
(—) process controlled by Knudsen diffusion only and (- - - -) process controlled by
1 Knudsen diffusion and viscous flow. (a) hf was varied while hp was maintained at
25,000 Wm2 K1, and (b) hp was varied while hf was maintained at
25,000 Wm2 K1.
0.8

0.6
into consideration, increasing hf increases both ts,f and ts,p, although
θ

Feed bulk temperature (ºC) the increase of ts,f is insignificant compared to the increase of ts,f.
0.4 50 ºC Δ 70 ºC 80 ºC o 90 ºC
When viscous flow is taken into consideration the increase of ts,p
is more pronounced. This occurs because increasing hf will enhance
0.2
energy input across the feed/membrane interface at Z = 0. The
enhancement of energy input requires an increase of ts,p in order
0
0 5 10 15 20 25 30 for this energy to be dissipated at the permeate side since hp is held
3
Heat transfer coefficient, h f and h p , (10 W/m2 .K) constant. Incorporating the viscous flow enhances this increase in
energy input. These increases of ts,f and ts,p cause the increase in no-
Fig. 6. Temperature polarization coefficient, h, versus heat transfer coefficients (hf dal vapor pressure and temperature across the entire membrane.
and hp) at different bulk feed temperatures and a bulk permeate temperature
This will further lead to a significant increase in the fraction of net-
20 oC. (—) process controlled by Knudsen diffusion only and (- - - -) process
controlled by Knudsen diffusion and viscous flow: (a) membrane A and (b) work pores governed by viscous flow (with an increase in hf) as
membrane C. shown in Fig. 9a.
M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259 1255

(a) 0.04 (a) 35


Viscous fraction of network pores

Bulk feed temperature (ºC)


Bulk feed temperature (ºC) 30
♦50 ºC Δ 70 ºC 80 ºC o 90 ºC
♦70 ºC Δ 80 ºC o 90 ºC
25

J (kg/h.m )
2
20
0.02
15

10

0
0 0 10 20 30 40 50 60
0 10 20 30 40 50 60 3 2
Feed heat transfer coefficient, h f , (10 W/m .K)
Feed heat transfer coefficient, h f , (103 W/m2.K)

(b) 0.04 (b) 35


Viscous fraction of network pores

Bulk feed temperature (ºC)


Bul feed temperature (ºC) 30
♦ 50 ºC Δ 70 ºC 80 ºC o 90 ºC
◊ 70 ºC Δ 80 ºC o 90 ºC
25

J (kg/h.m )
2
20
0.02 15

10

0
0 0 10 20 30 40 50 60
0 10 20 30 40 50 60
3
Permeate heat transfer coefficient, h p , (10 W/m .K)
2 Permeate heat transfer coefficient, h p , (103 W/m2.K)

Fig. 10. DCMD vapor flux versus heat transfer coefficients for membrane B at
Fig. 9. Fraction of network pores governed by viscous flow versus heat transfer
different bulk feed temperatures and a bulk permeate temperature 20 oC. (—)
coefficients for membrane B. (a) hf was varied while hp was maintained at
process controlled by Knudsen diffusion only and (- - - -) process controlled by
25,000 Wm2 K1, and (b) hp was varied while hf was maintained at
Knudsen diffusion and viscous mechanisms. (a) hf was varied while hp was kept
25,000 Wm2 K1.
constant at 25,000 Wm2 K1, and (b) hp was varied while hf was kept constant at
25,000 Wm2 K1.

Fig. 8b shows ts,f and ts,p versus hp when hf is kept constant at


25,000 Wm2 K1 for different values of the bulk feed tempera- ture at 20 oC. When the transporting process is controlled by Knud-
tures. As can be seen, when only Knudsen diffusion is taken into sen diffusion only, the increase of the vapor flux occurs only insig-
account, ts,p decreases with an increase in hp, approaching tb,f. But nificantly with the increase of hp despite the reduction of ts,p. When
there is no significant effect on ts,f. When both Knudsen diffusion viscous flow is taken into consideration, the increase in vapor flux
and viscous flow are taken into account, the hp increase causes is insignificant but it is the most pronounced at lower range of hp
no significant effect on the ts,f. On the other hand, the decrease in and at the highest feed temperature (90 oC). Again, in Fig. 9b, the
ts,p is more pronounced, particularly at high feed temperatures fraction of network pores governed by the viscous flow decreases
(above 70 oC) and low hp values. This will eventually reduces the significantly at the lower hp values and at the highest feed temper-
fraction of the network pores governed by viscous flow, as depicted ature (90 oC).
in Fig. 9b. The increase in the fraction of the network pores gov- Fig. 11a and b shows temperature polarization coefficient, h, vs.
erned by viscous flow occurs most strongly at a feed temperature feed and permeate heat transfer coefficients, respectively. In
of 90 oC and low hp values, which corresponds to the highest value Fig. 11a it can be seen that h increases approaching an asymptotic
of ts,p (Fig. 8b). The pores through which the mechanism of trans- value as hf increases depending on the bulk feed temperature.
port changes from viscous flow to Knudsen diffusion with the in- When viscous mechanism is considered (broken line in Fig. 11a),
crease of hp are most probably those located in close proximity the observed asymptotic value is smaller than the one obtained
to the permeate side of the membrane and/or those which have when Knudsen diffusion is considered alone. The difference de-
small sizes. In both cases no significant effect will be expected on pends on the fraction of membrane pores governed by viscous
the DCMD process vapor flux [14] as shown in Fig. 10b. flow. This insignificant change of h with the increase of hf, regard-
Fig. 10a shows the vapor flux vs. the feed side heat transfer coef- less of the considered mass transport mechanism, can be clearly
ficient, hf, for different values of feed temperatures. When only attributed to the insignificant change of ts,f and ts,p, with the in-
Knudsen diffusion is taken into account, vapor flux increases crease of hf as can be seen Fig. 8a. Fig. 11b also shows the increase
monotonically with the increase of hf and seems to approach of h with the increase of hp due to the reduction of ts,p and the insig-
asymptotic values at high hf. This is due to the increase of ts,f shown nificant change of ts,f, when the mass transfer is described by Knud-
previously in Fig. 8a. However, when viscous flow is taken into sen diffusion only, as presented previously in Fig. 8b. When both
consideration the increase of the vapor flux at low hf values is less Knudsen and viscous mass transport mechanisms are considered,
significant. At higher hf values than 20,000 Wm2 K1, the vapor the temperature polarization coefficients are reduced to smaller
flux is maintained practically the same. It even reduces the process values, depending on the fraction of membrane pores governed
vapor flux (see Fig. 10a). This is attributed to the increase of the by viscous flow, as shown earlier in Fig. 9b.
fraction of membrane pores governed by viscous flow transport The observed insignificant effect of hp on the DCMD transport
mechanism (see Fig. 9a). On the other hand, Fig. 10b shows the va- properties and on the vapor flux is particularly obvious as the
por flux vs. the permeate heat transfer coefficient, hp, for different fraction of membrane pores governed by viscous flow (Fig. 9b)
bulk feed temperatures, maintaining the bulk permeate tempera- decreases approaching an asymptotic value with the increase of
1256 M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259

1 using the laboratory system described elsewhere [22]. The central


(a)
part of the system is a stainless steel cell composed of two cylindri-
cal chambers. One of the chambers is connected to a heating sys-
tem through its jacket to control the temperature of the liquid
0.8
feed. The other chamber is connected to a cooling system to control
the temperature of the permeate. The membrane was placed be-
θ

Bulk feed temperature (ºC)


♦50 ºC Δ 70 ºC 80 ºC o 90 ºC tween the two chambers. The effective membrane area of the
0.6 membrane system is 2.75  103 m2. The bulk feed and permeate
temperatures were measured inside each chamber by a pair of sen-
sors connected to a digital meter with an accuracy of ±0.1 oC and
0.4 both the feed and permeate liquids were stirred inside the cell
0 10 20 30 40 50 60 by graduated magnetic stirrers. The conditions of the experiments
3 2
Feed heat transfer coefficient, h f , (10 W/m .K) were a stirring rate of 500 rpm and 670 rpm in both feed and per-
meate chambers, mean temperatures varying from 20 oC to 55 °C
(b) 1 with steps of 5 °C and the bulk temperature difference between
the feed and permeate was varied from 10 °C to about 70 oC. All
experiments were carried out under atmospheric pressure in both
0.8 the feed and permeate chambers. Similar sets of experiments were
Bulk feed temperature (ºC) conducted for each membrane and a new membrane was used for
θ

♦50 ºC Δ 70 ºC 80 ºC o 90 ºC each experiment. The DCMD flux was calculated in every case by
measuring the condensate collected in the permeate chamber for
0.6
a predetermined time.

4.2. Results and discussion


0.4
0 10 20 30 40 50 60
4.2.1. Morphological characterization of MD membranes
Permeate heat transfer coefficient, h p , (103 W/m2.K) Table 1 lists the characteristics of the membranes employed in
this study. It can be observed that the membrane TF200 exhibits
Fig. 11. Temperature polarization coefficient versus heat transfer coefficients for
membrane B. (—) process controlled by Knudsen diffusion only and (- - - -) process higher liquid entry pressure of water (LEPw) than the membrane
controlled by Knudsen diffusion and viscous flow. (a) hf was varied while hp was GVHP. This is attributed to the lower maximum pore size and to
kept constant at 25,000 Wm2 K1, and (b) hp was varied while hf was kept constant the higher hydrophobicity of TF200. In fact, the measured advanc-
at 25,000 Wm2 K1. ing and receding water contact angles of TF200 are higher than
those of the membrane GVHP. The advancing contact angles are
110.2° and 113.6°, while the receding contact angles are 95.9°
hp. The only significant change of the DCMD flux occurs at low hp
and 100.7° for the membranes GVHP and TF200 membranes,
values, which may correspond to the upper limit of laminar flow
respectively.
condition (transition zone) in the membrane permeate side. When
The volume porosity (void volume) of the membrane GVHP is
the fluid condition in the permeate side goes into turbulent region,
specified as 75% by the manufacturer, while that of TF200 is 80%.
the increase of the cooling flow rate at the permeate side does not
However, the measured values of the porosity given in Table 1
show any significant effect on the DCMD interfacial surface tem-
are lower than these values (i.e. 70.1 ± 2.7% for GVHP and
peratures, ts,f and ts,p, and on the vapor flux [14].
68.7 ± 5.4% for TF200). This may be attributed to the different ap-
plied measurement techniques used.
4. Experimental The average pore size and the pore size distribution of the
membranes were determined by the dry/wet flow method as sta-
4.1. Materials and direct contact membrane distillation (DCMD) ted elsewhere [20]. Fig. 12 shows the pore size distribution of each
membrane. It can be observed that the pore size distribution curve
Two commercial microporous hydrophobic flat-sheet mem- of the membrane TF200 is lower and narrower around the mean
branes were used in this study: TF200 (Gelman) made of polytetra- pore size and is shifted to the left compared to the pore size distri-
fluoroethylene (PTFE) polymer supported by a polypropylene net bution of the membrane GVHP. In addition, the obtained values of
and GVHP (Millipore) made of polyvinylidene fluoride (PVDF) poly- the geometrical standard deviation, rp, of both membranes were
mer. Their principal characteristics are given in Table 1. The meth- close to unity although the Millipore membrane (GVHP) and the
ods used to obtain these membrane parameters were detailed Gelman one (TF200) were prepared with different polymers and
elsewhere [20,21]. different techniques.
DCMD experiments were performed using pure water as feed at Pore density, N, and surface porosity, es, were calculated using
different mean temperatures and bulk temperature differences the pore size distributions as stated in [20]. The results are also

Table 1
Membrane characteristics: membrane thickness, d; liquid entry pressure of water, LEPw; void volume (volume porosity), ev; surface porosity, es; mean pore size, lp; geometric
standard deviation, rp; pore density, N.

Membrane d (lm) LEPw (bar) ev (%) esc (%) lpc (nm) rpc N (lm2)
TF200a 55 ± 6 2.76 ± 0.09 69 ± 5 43.18 233.38 1.07 9.87
GVHPb 118 ± 4 2.04 ± 0.03 70 ± 3 32.74 265.53 1.12 5.73
a
Membrane supplied by Gelman. Measured total thickness: 165 ± 8 lm.
b
Membranes supplied by Millipore.
c
lp, rp, es were determined from the wet/dry flow method.
M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259 1257

0.05 For both membranes, an exponential increase of the DCMD flux


with the average temperature was observed (Table 2, Fig. 13a). This
Probability density function,

TF200
result is attributed to the exponential increase of the vapor pres-
0.04
sure with temperature according to Antoine’s equation. Comparing
df(dp)/d(d p), (nm )
-1

GVHP Tables 3 and 4, the DCMD flux is higher when the stirring rate is
0.03 higher. This is expected since the enhanced mixing in both the feed
and permeate chamber decreases the temperature polarization ef-
0.02 fect. This corresponds to the effect of heat transfer coefficient on
the flux discussed previously in Section 4.
Fig. 14a illustrates, as an example, the effect of the bulk temper-
0.01
ature difference on the DCMD flux for the membrane TF200. In
these experiments the stirring rate in both the feed and the perme-
0 ate chamber was maintained at 500 rpm and the bulk permeate
0 100 200 300 400 500 600 temperature at 20 °C. As the bulk feed temperature increases the
Pore size (nm) temperature at the feed membrane surface increases, leading to
Fig. 12. Probability density curves of the membranes TF200 and GVHP [20]. an increase of the DCMD flux. Similar trends were observed in
Fig. 5a and b and discussed previously in Section 4.

summarized in Table 1. In comparison to the membrane TF200, it


can be seen that the pore density and surface porosity of the mem- 4.2.3. Heat transfer coefficients
brane GVHP are lower; while the void volume of both membranes The feed and the permeate heat transfer coefficients were calcu-
is similar. This may be attributed to the larger thickness the Milli- lated using the semi-empirical model reported in [20] and the
pore membrane (GVHP). results are summarized in Tables 2–4. As can be seen, the heat trans-
fer coefficients of both feed and permeate vary with temperature
4.2.2. Results of DCMD experiments and stirring rate for both membranes. For example, in Table 2, the
The effects of the involved operating parameters such as feed heat transfer coefficients in the feed side of both membranes are
and permeate temperatures and circulation velocities on the mass higher than those in the permeate side. This is due to the fact that
transfer coefficients in DCMD applications have been investigated the feed temperature is higher than the permeate temperature
thoroughly in previous papers [1,4–13,17,20–22]. It was observed affecting the hydrodynamics parameters of the feed and permeate
experimentally that the DCMD flux increased exponentially with liquid. Moreover, by comparing the heat transfer coefficients of
the average temperature between the bulk feed and permeate, the membrane TF200 summarized in Tables 2 and 4, the heat
T þT
i.e. T m ¼ b;f 2 b;p , and increased with the stirring rate (circulation transfer coefficients increase with increasing stirring rate. This is
velocity) approaching asymptotic values. attributed to the decrease of the temperature polarization effect
The objective of the experiments in this work is to test the pre- as it was observed in previous studies [7,10,13,22]. Furthermore,
dictability of the MC simulation model. The experimental results from Table 3, the heat transfer coefficient in the feed membrane side
are given in Tables 2–4 and in Fig. 13a (solid lines) for GVHP mem- decreases while that in the permeate side increases with increasing
brane and in Fig. 14a (solid lines) for TF200. Under similar operat- the bulk feed temperature maintaining the bulk permeate temper-
ing conditions, it was observed that the DCMD flux of the ature around 20 °C. This is due to the decrease of the permeate
membrane TF200 is higher than that of the GVHP membrane. This temperature at the membrane surface, which is affected by the
is attributed mainly to the larger thickness of the GVHP membrane. increase of the bulk feed temperature.

Table 2
Heat transfer coefficients and permeate fluxes of the membranes GVHP and TF200 at a stirring rate 500 rpm, a transmembrane bulk temperature of about 10 °C and different
mean temperatures.

Membrane GVHP Membrane TF200


tb,f (°C) tb,p (°C) J (g/m2 s) hf (W/m2 K) hp (W/m2 K) tb,f (°C) tb,p (°C) J (g/m2 s) hf (W/m2 K) hp (W/m2 K)
25.5 15.4 0.37 3374 2674 25 15 0.89 3401 2618
30 20 0.46 3493 2814 30 20 1.13 3524 2773
35 25 0.60 3593 2959 35 25 1.44 3608 2925
40.1 30.1 0.76 3639 3080 40 30 1.79 3632 3056
45 35 0.95 3612 3151 45 35 2.17 3583 3146
50.1 40.1 1.17 3500 3160 50 40 2.57 3457 3174
54.8 44.8 1.38 3326 3102
60.1 50.1 1.63 3063 2961

Table 3
Heat transfer coefficients and permeate fluxes of the membranes GVHP and TF200 at a stirring rate 500 rpm and different bulk feed temperatures.

Membrane GVHP Membrane TF200


2 2 2
tb,f (°C) tb,p (°C) J (g/m s) hf (W/m K) hp (W/m K) tb,f (°C) tb,p (°C) J (g/m2 s) hf (W/m2 K) hp (W/m2 K)
40.1 19.9 1.59 3628 2794 39.8 20 3.23 3612 2838
50.1 19.9 2.91 3464 2841 50.1 20.1 5.92 3471 2914
60.3 19.9 4.58 3058 2888 60 19.9 9.82 3136 2991
70.1 20.0 6.49 2578 2939 70.1 19.9 13.90 2708 3068
80.3 20.0 9.53 2128 2999 80.1 20.1 18.69 2310 3148
90.7 19.7 13.52 1768 3060
1258 M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259

Table 4
Heat transfer coefficients and permeate fluxes of the membrane TF200 at a stirring (a) 20

rate 670 rpm, a transmembrane bulk temperature 10 °C and different mean ------- Simulation Experimental
temperatures. 16
Δ k s = 0.126 W/m.K
X k s = 0.100 W/m.K
tb,f (°C) tb,p (°C) J (g/m2 s) hf (W/m2 K) hp (W/m2 K)

J (g/m .s)
12

2
25 15 1.05 4305 3302
30 20 1.35 4459 3497
35 25 1.64 4564 3689
8
40 30 1.99 4594 3855
45 35 2.37 4530 3969 4
50 40 2.80 4366 4007

0
0 10 20 30 40 50 60 70
Δ T (ºC)
4.2.4. Comparison of Monte Carlo simulation and DCMD experimental
data 0.8
The MC simulation was carried out by the method described in
(b)
Section 2. It should be noted that the membrane pore structure gi-
ven in Table 1 and the heat transfer coefficients given in Tables 2–4 0.6
were independent from the MD experiments. Therefore, the flux
values obtained by MC simulation are truly predicted values. 0.4

θ
Fig. 13a, which refers to the membrane GVHP, shows both the
Δ k s = 0.126 W/m.K
experimental and the MC simulated DCMD flux as a function of X k s = 0.100 W/m K
0.2
the average temperature for two different thermal conductivity
values of the membrane material (0.126 and 0.12 W/m K). The bulk
temperature difference was maintained at 10 °C and the stirring 0
rate of both the feed and permeate chambers was maintained at 0 10 20 30 40 50 60 70
500 rpm. It must be pointed out that the thermal conductivity of
Δ T (ºC)
polyvinylidene fluoride (PVDF) polymer (material for GVHP mem- Fig. 14. (a) Simulated and experimental DCMD vapor flux, J, of the TF200
brane) is 0.126 W/m K [20,23,24]. Fig. 13a presents excellent membrane, and (b) temperature polarization coefficient, h, vs. the bulk temperature
agreement between the experimental and predicted values. The difference, DT, at a stirring rate 500 rpm. The MC model predictions were carried
figure also shows that the effect of the thermal conductivity is out for different membrane material thermal conductivity (ks) values.

insignificant in the studied thermal conductivity range. The tem-


perature polarization coefficients were calculated and the results
are plotted in Fig. 13b. The temperature polarization coefficient de- creases slightly with the average temperature. This is in accor-
dance with Fig. 6a and b, where the polarization coefficients de-
crease with the feed temperature (as well as the average
(a) 1.6
temperature). Similar results were also observed in DCMD when
------- Simulation Experimental using other theoretical approaches [9,13]. Referring to Fig. 13b, it
Δ k s = 0.126 W/m K
X k s = 0.120 W/m K is worth noting that a slight decrease of the temperature polariza-
1.2 tion coefficient occurs with the increase of the polymer
J (g/m s)
2

conductivity.
0.8 Fig. 14a shows the effect of the bulk feed temperature when the
permeate temperature is fixed at 20 oC. The stirring rate was
500 rpm. The predicted value represents the experimental data
0.4
very well when the thermal conductivity is 0.1 W/m K. The slight
deviations may be attributed to the estimated values of the feed
0 and permeate heat transfer coefficients from the semi-empirical
20 30 40 50 60
T m (ºC) model and to the support of the TF200 membrane, which also af-
fects the permeate heat transfer coefficient.
In the entire range of the feed temperature the simulated DCMD
(b) 1
flux is slightly lower than the experimental one. The difference be-
tween the experimental and the simulated value increases as the
0.8
thermal conductivity increases. It must be stated that the reported
thermal conductivity of polytetrafluoroethylene (PTFE), which is
0.6 the material used to prepare the membrane TF200, is 0.22 W/m K
θ

Δ k s = 0.126 W/m.K [23,24]. However, lower thermal conductivities coefficients were


0.4 X k s = 0.120 W/m.K found to produce far better agreement. This may be attributed to
the fact that this membrane is supported by a polypropylene net,
0.2 which may affect both the heat transfer coefficient on the perme-
ate side and the DCMD flux.
0
20 30 40 50 60
T m (ºC) 5. Summary and conclusions

Fig. 13. (a) Simulated and experimental DCMD flux, J, of the GVHP membrane, and
(b) temperature polarization coefficient, h, vs. average temperature, Tm, at a stirring
A Monte Carlo (MC) simulation model was developed for direct
rate 500 rpm and a bulk temperature difference 10 oC. The MC model predictions contact membrane distillation (DCMD) process. This model is
were carried out for different membrane material thermal conductivity (ks) values. designed to be comprehensive in its approach. It can predict
M. Khayet et al. / International Journal of Heat and Mass Transfer 53 (2010) 1249–1259 1259

simultaneously the DMCD process vapor flux (permeability) and [7] L. Martínez-Díez, M.I. Vázquez-González, Temperature and concentration
polarization in membrane distillation of aqueous salt solutions, J. Membr.
membrane surface temperatures, taking into consideration the ef-
Sci. 156 (1999) 265–273.
fects of the applied membrane physical properties, which include [8] J. Phattaranawik, R. Jiraratananon, Direct contact membrane distillation: effect
membrane pore space inter-connectivity, transport mechanisms, of mass transfer on heat transfer, J. Membr. Sci. 188 (2001) 137–143.
and fluid flow conditions through the DCMD process module (pro- [9] J. Phattaranawik, R. Jiraratananon, A.G. Fane, Heat transport and membrane
distillation coefficients in direct contact membrane distillation, J. Membr. Sci.
cess dynamics). In principle, this model is a significant advance- 212 (2003) 177–193.
ment in MD simulation compared to other models in which [10] J. Phattaranawik, R. Jiraratananon, A.G. Fane, C. Halim, Mass flux enhancement
either process vapor flux or membrane surface temperatures has using spacer filed channels in direct contact membrane distillation, J. Membr.
Sci. 187 (2001) 193–201.
to be given in order to predict the other. The results obtained in [11] L. Martínez-Díez, M.I. Vázquez-González, F.J. Florido-Díaz, Study of membrane
this study show an excellent qualitative and quantitative agree- distillation using channel spacer, J. Membr. Sci. 144 (1998) 45–56.
ment with available experimental data without the need of using [12] K.W. Lawson, D.R. Lloyd, Membrane distillation. II. Direct contact MD, J.
Membr. Sci. 120 (1996) 123–133.
any adjustable parameter(s). Therefore, the MC simulation model [13] K.W. Lawson, D.R. Lloyd, Review membrane distillation, J. Membr. Sci. 124
developed in this study may play a supportive role to experimental (1997) 1–25.
work, optimizing membrane structuring and MD process module [14] A.O. Imdakm, T. Matsuura, A Monte Carlo simulation model for
membrane distillation processes: direct contact (MD), J. Membr. Sci.
design. 237 (2004) 51–59.
[15] A.O. Imdakm, M. Sahimi, Computer simulation of particles transport
Acknowledgements processes in flow through porous media, Chem. Eng. Sci. 46 (1991)
1977–1993.
[16] M. Sahimi, A.O. Imdakm, The effect of morphological disorder on
The authors gratefully acknowledge the support of the Spanish hydrodynamic dispersion in flow through porous media, J. Phys. A: Math.
Ministry of Science and Education (MEC) through its project No. Gen 21 (1988) 3833–3870.
FIS2006-05323 and the UCM-BSCH (Project GR58/08, UCM group [17] A.O. Imdakm, T. Matsuura, Simulation of heat and mass transfer in direct
contact membrane distillation (MD): the effect of membrane physical
910336). properties, J. Membr. Sci. 262 (2005) 117–128.
[18] R.G. Rice, D.D. Do, Applied Mathematics and Modeling for Chemical Engineers,
References John Wiley & Sons, Inc., 1995.
[19] J.M. Smith, H.C. Van Ness, Introduction to Chemical Engineering
Thermodynamics, 4th ed., McGraw-Hill, New York, 1987.
[1] R.W. Schofield, A.G. Fane, C.J.D. Fell, R. Macoun, Factors affecting flux in
[20] M. Khayet, A. Velázquez, J.I. Mengual, Modeling mass transport through a
membrane distillation, Desalination 77 (1990) 279–294.
porous partition: effect of pore size distribution, J. Non-Equilib. Thermodyn. 29
[2] J.I. Mengual, M. Khayet, M.P. Godino, Heat and mass transfer in vacuum
(2004) 279–299.
membrane distillation, Int. J. Heat Mass Transfer 47 (2004) 865–875.
[21] M. Khayet, K.C. Khulbe, T. Matsuura, Characterization of membranes for
[3] F.A. Banat, F.A. Al-Rub, R. Jumah, M. Shannag, Theoretical investigation of
membrane distillation by atomic force microscopy and estimation of their
membrane distillation role in breaking the formic acid–water azeotropic point:
water vapor transfer coefficients in vacuum membrane distillation, J. Membr.
comparison between Fickian and Stefan–Maxwell-based models, Int. Commun.
Sci. 238 (2004) 199–211.
Heat Mass Transfer 26 (1999) 879–888.
[22] M. Khayet, M.P. Godino, J.I. Mengual, Modelling transport mechanism through
[4] R.W. Schofield, A.G. Fane, C.J.D. Fell, Heat and mass transfer in membrane
a porous partition, J. Non-Equilib. Thermodyn. 26 (2001) 1–14.
distillation, J. Membr. Sci. 33 (1987) 299–313.
[23] J.H. Perry, Chemical Engineers Handbook, 4th ed., McGraw Hill, New York,
[5] R.W. Schofield, A.G. Fane, C.J.D. Fell, Gas and vapor transport through
1963.
microporous membranes. II. Membrane distillation, J. Membr. Sci. 5 (1990)
[24] C.A. Speraty, Physical Constants of Fluoropolymers, 3rd ed., Polymer
73–185.
Handbook, Wiley, New York, 1989.
[6] M. Gryta, M. Tomaszewska, Heat transport in the membrane distillation
process, J. Membr. Sci. 144 (1998) 211–222.

S-ar putea să vă placă și