Desalination 424 (2017) 122–130

Contents lists available at ScienceDirect

Desalination
journal homepage: www.elsevier.com/locate/desal

Numerical investigation of air gap membrane distillation (AGMD): Seeking MARK

optimal performance
Isam Janajreha,⁎, Khadije El Kadia, Raed Hashaikeha, Rizwan Ahmedb
a
Mechanical and Materials Engineering Department, Khalifa University of Science and Technology, Masdar Institute, UAE
b
Takreer Research Center, ADNOC, Abu Dhabi, UAE

A R T I C L E I N F O A B S T R A C T

Keywords: Membrane distillation (MD) is used in desalination, wastewater treatment, and medicinal application. Direct
Air gap membrane distillation contact (DCMD) and air-gap (AGMD) membrane distillation are the most common configurations. The simplicity
CFD and high flux marks the advantages of the former while the low fouling is attributed to the latter. The air-gap
DCMD integration used between the bottom surface of membrane and the permeate although adds thermal resistance it
Temperature polarization coefficient
reduces membrane wetting and fouling. Researchers continue to investigate these configurations to optimize
Heat transfer
their performance. In this work, high fidelity numerical analysis is carried out to assess and quantify the per-
formance of the AGMD and compare it with the DCMD. Different geometric and operating parameters are
considered. Results are demonstrated in terms of temperature distributions, polarization-coefficient (TPC), mass-
flux, heat-flux, surface heat coefficients, and thermal efficiency (η). Results reveal that the integration of a thin
air-gap reduces the TPC by 38%, the total heat-flux by 37%, and nearly 22% for each of the mass-flux and the
thermal efficiency. Furthermore, increasing AGMD feed temperature from 50 °C to 75 °C cause increase in the
mass flux from 3.34 to 15.3 g/m2·s that corresponds to increase in the thermal efficiency from 11.5% to 52.7%.
Higher temperature show much larger effect on the performance than flow velocity.

1. Introduction
Membrane distillation (MD) is a low energy and effective method
for water treatment and desalination. It is a separation process driven
by temperature gradient, where hot-salt water is circulated in one side
of a hydrophobic porous membrane, and cold-fresh stream is circulated
in the other side. Thus, creating a difference of partial pressure between
the two sides of the membrane that constitutes the main driving force of
the process. The MD system can be configured in four different ways.
These are the conventional direct contact membrane distillation
(DCMD), air gap membrane distillation (AGMD), sweeping gas mem-
brane distillation (SGMD), and vacuum membrane distillation (VMD) as
depicted in Fig. 1. The DCMD and AGMD configurations are the most
researched and used technology in desalination processes due to their
simple configuration [1–5]. However, the main obstruction of DCMD
and AGMD is the heat loss by conduction and additional thermal re-
sistance, respectively [1]. The addition of air gap in the AGMD makes
the temperature and partial pressure gradients less pronounced than the
conventional DCMD, thus causing lower AGMD fluxes than in the other
MD configurations as DCMD [6]. However, the indirect contact of
permeate with the lower membrane surface in the AGMD gives the
advantage to halt the wetting at the permeate side when volatile
components are present. As such components are likely to wet the lower
membrane surface due to their lower surface tension [6,7]. Introducing
lower pressure at the permeate side by mean of VMD can also improve
wetting problem. In AGMD, as shown in Fig. 1 the feed is directly in
contact with membrane upper surface, while the permeate is condensed
on a cold collection plate which is separated from the membrane sur-
face by an air gap. This method ensures a reduction in energy loss due
heat conduction through the membrane but at the same time it adds an
insulation layer at the gap side [2]. This penalizes the yield production
in AGMD. Air gap MD is suitable for all DCMD applications. Moreover,
it can also extract other volatile substances from solutions such as al-
cohols [8,9]. Those substances cause the permeate side to be wet be-
cause the lower side of the membrane is considered as hydrophilic. In
other words, permeate side of AGMD unit is not directly touching the
membrane. This fact reduces the risk of wetting the bottom side of the
membrane marking AGMD as better choice than DCMD in some ap-
plication. As a quantitative comparison between the performances of
these two-setups is limited in literature, this work intends to quantify
this difference targeting system performance [10]. In line of past studies
[11–14], this can be presented in terms of the attained temperature

⁎
Corresponding author.
E-mail address: ijanajreh@masdar.ac.ae (I. Janajreh).

http://dx.doi.org/10.1016/j.desal.2017.10.001
Received 29 April 2017; Received in revised form 27 September 2017; Accepted 1 October 2017
0011-9164/ © 2017 Published by Elsevier B.V.
I. Janajreh et al.
distribution, the distribution and average of TCP, mass flux as well as
the attained thermal efficiency corresponding to parametrical con-
sideration. It is worth mentioning that MD modules (i.e. flat sheet
membranes) require mechanical support to prevent rupture of the thin
membrane. Such strong support (i.e. spacers) is capable also to promote
turbulence which as a result offers a great potential for system opti-
mization [15–16]. The role of spacers in the MD performance is not
covered in this work, however, a full study was well-illustrated in a
previous work of the authors [17].
Numerical simulation of DCMD and AGMD has been sought by
several researchers to gain insight of their operation [18–28]. Yu et al.
[18] used Navier-Stokes flow analysis while accounted to heat and mass
fluxes in a hollow fiber tube. They however used a fixed mass transfer
coefficient independent of temperature or membrane properties. Em-
pirical and coarse model that based on semi-empirical correlation or
constant mass flux coefficient, one side flow, or thermal resistances
analogy stack that evaluate temperature distribution are appear in
several literature works [19–23]. Zhang et al. [24–25] used conjugate
heat flow model while Charfi et al. [26] used Navier-stokes equations in
a single chamber modeling associated with pressure drop Ergun model
and membrane flux of Knudson-diffusion. Asghari et al. [27] developed
conjugated heat and mass transfer numerical model for AGMD as-
suming a static air gap between the bottom side of the membrane and
the condensing plate. Exact quantification of the system yield and ex-
tended system parameters (temperature and pressure distribution) as
well as performance metrics (TPC, mass and thermal yield and effi-
ciency) were incomplete in these models. Swaminathan et al. [28] have
considered different AGMD including conductive and permeate filled
they reported the thermal efficiency of the model. They however im-
plemented lower fidelity one-dimensional model that ignores Navier-
stokes based flow formulation. Nevertheless, it was successfully used in
evaluating the effect of pure water flow direction, integrate the influ-
ence of conductive gap materials side to the membrane material con-
ductivity [28]. More up to date work appears in the dissertation of
Winer [15] who carried out extensive thermodynamic and economic
analysis of DCMD. He coupled the heat and mass transfer in the nu-
merical model and assessed it experimentally. In this work, conjugated
heat transfer Navier-Stokes flow model for counter flow of DCMD and
AGMD is developed and assessed simultaneously under range of
123
Desalination 424 (2017) 122–130
Fig. 1. Membrane distillation (MD) process con-
figurations.
velocity and temperature conditions. These models are validated fol-
lowing the previous work of the authors [14]. Strong validation of these
models is incepted by the accounted conjugated heat thermal coupling,
appropriate boundary conditions and accurate modeling of the con-
sidered temperature-dependent Poiseuille and Knudson-diffusion mass
transfer. Parametric study is done to explore and quantify the role of
velocity and temperature in the performance of the AGMD. Re-
commendation on the conditions that provide best system performance
is also given. The developed model accounts for numerous DCMD
parameters, gap thickness, membrane properties and its composite,
porosity, toruisity, system geometry including membrane support and
various operational conditions. Thus, it can be used in the process de-
velopment of MD that pushes the technology closure for full scale de-
ployment.
2. Methodology
2.1. Model setup and governing equations
In this work, a numerical study will be performed on air gap
membrane distillation (AGMD) using non-isothermal computational
fluid dynamics (CFD) and thermally conjugated/coupled with the solid
porous membrane. The computational domain comprises of two chan-
nels (210 mm length by 1 mm height) sandwiching the 0.13 mm
thickness membrane as shown in the baseline setup in Fig. 2. Another
conjugated heat is also considered at the air gap wall to keep the pro-
blem coupled. At this wall, however a super thermal conductive layer is
utilized to minimize the thermal resistance created by the gap. It should
be noted that the height of the gap is directly proportional to thermal
resistance and hence causing higher temperature of the bottom wall and
leading to reduction in the overall performance. In this work, several
gap heights are considered, i.e. starting from the no-gap baseline then
sweeping from the smallest feasible air gap (0.05 mm) to 0.4 mm.
Appropriate flow conditions will be utilized for the feed and
permeate flow channels with a prescribed inlet velocity/mass flow rate
and temperature and insulated outer walls as well as the outlet condi-
tions including pressures. These conditions depend on the application,
for example in water desalination it appears that channel Reynolds
number near 100 and 75 °C temperature lead to high thermal efficiency
I. Janajreh et al. Desalination 424 (2017) 122–130

the thermal conductivity, and Sh indicated the sink/source heat that is

Fig. 2. Schematic of the flow configuration for the AGMD.
near 50% and reasonable pressure drop within the flow channel [8].
Because the membrane is thermally conjugated the temperature dis-
tribution of the surfaces are the main solicited flow variable. As the
temperature and velocity field are computed, performance metrics in-
cluding polarization temperature coefficient, mass flux, and thermal
efficiency will be evaluated to assess AGMD system performance. Fur-
thermore, geometrical and operational sensitivity study will be con-
ducted to arrive to the optimal AGMD process metrics. The loss of mass
flux will be compared with the baseline DCMD performance and sug-
gested method to compensate this loss will be stipulated.
A planer 2D flow system is discretized as shown in Fig. 3. This
numerical system is governed by the conjugated steady incompressible
Navier-Stokes and energy equations. These equations are written as:
Continuity:
∂ρU ∂ρV
+ =0
∂x ∂y
attributed to the latent heat of evaporation at both the feed and
permeate membrane surface. A structured quadrilateral mesh type is
used for the geometry refined to capture the wall boundary layer. The
baseline mesh size is 2,100 × 64 per channel and 2,100 × 8 to
21,100 × 24 for the membrane and the gap, respectively (see Fig. 3).
Boundary conditions are flow inlet and wall Dirichlet conditions
(prescribed velocity and temperature value) and outlet zero-velocity
gradient and constant atmospheric pressure Neumann conditions (zero
gradients). Walls are subjected to no slip and no penetrating velocity.
Thermally, membrane wall is coupled while outer channel is insulated
or Neumann of zero gradient. A cutaway of the established baseline
mesh and the computational domain boundary conditions are shown in
Fig. 3. Reynolds numbers of 10, 50, and 100 are nominally set for the
flow conditions for a brine/feed and fresh/permeate are set at. This is
equivalent to prescribed velocities of 0.01, 0.05, and 0.1 m/s respec-
tively. The feed temperature alternatively set at 50 °C and 75 °C,
whereas permeate is fixed at 25 °C. The solution is carried using the
commercial CFD code FLUENT that based on finite volume approach
and segregated solver. The semi-implicit method for pressure-linked
equations (SIMPLE) algorithm is used for pressure-velocity coupling
and 2nd order upwind spatial derivatives. Convergence residuals is set
very tight of 1E10− 12 for all scalar equations, i.e. continuity, and x- & y-
momentums and energy. Summary of the boundary conditions are
listed in Table 1, whereas all membrane, fresh feed, and saline permeate
properties are listed in Table 2.
2.2. System metrics evaluation
Temperature polarization (θ): It defines the ratio of boundary layer
(1) resistance to the total/bulk heat transfer resistance and is expressed as:

x-Momentum: ∆Tm Tm, f − Tm, p

θ= =
∂ρU ∂ρU ∂ρ ∂ 2U ∂ 2U ⎞ ∆Tb Tb, f − Tb, p (12)
U +V =− + μ⎛ 2 + ⎜ ⎟
∂x ∂y ∂x ⎝ ∂ x ∂y 2 ⎠ (2) where the subscripts m, b, f, p signify the membrane, bulk, and feed and
y-Momentum: permeate flow, respectively. A smaller value of θ( ≤ 0.2) puts the
DCMD system in the limit of heat transfer and reflects a poor design,
∂ρV ∂ρV ∂ρ ∂ 2V ∂ 2V ⎞ whereas a larger value of θ(≥ 0.6) can hinder the system to enter the
U +V =− + μ⎛ 2 + + ρgy
⎜ ⎟
∂x ∂y ∂y ⎝ ∂x ∂y 2 ⎠ (3) mass transfer limitation which is restricted to lower membrane per-
meability [1,31–32].
Energy:
Mass Flux ( J"): The transported flux within the membrane follows
∂ρCp T ∂ρCp T ∂ 2T ∂ 2T ⎞ the molecular diffusion theory as described by Khayet [2] and depends
U +V = k⎛ 2 +
⎜ ⎟ + Sh on the membrane properties and established driving force, and has
∂x ∂y ⎝ ∂ x ∂y 2 ⎠ (4)
shown to increase with the increase of the pore size and porosity as well
where U and V are the velocity components, ρ is the density, μ is the as by reducing the tortuosity and thickness of the membrane [33]. This
viscosity, gy is the gravitational acceleration, Cp is the specific heat, k is flux is written as [20,34–36]:

Fig. 3. AGMD numerical model setup and discretized mesh.

124
I. Janajreh et al. Desalination 424 (2017) 122–130

Table 1 pores radius, ε is porosity of the membrane, τ is tortuosity factor.

Summary of boundary conditions for DCMD and AGMD. Heat flux: The heat transfer in DCMD can be described following
three steps, transfers through the feed boundary layer, then across the
Boundary conditions Feed channel Permeate Airgap channel
channel membrane, and finally through the permeate boundary layer [2]. The
total heat flux across the membrane (Qm) accounts for the conduction
Inlet channel Kinetic 0.01/0.05/ 0.01/0.05/ 0.01/0.05/ (Qc) and latten heat of evaporation (Qv). Considering Δ Hm is the change
0.1 m/s 0.1 m/s 0.1 m/s
of enthalpy due to the latent heat of the transmembrane mass flux, it
Thermal 50/75 °C 25 °C 25 °C
Outlet channel Zero pressure, Zero pressure, Zero pressure, can be written as described by Termpiayakul as [41].
dV/dx = 0 dV/dx = 0 dV/dx = 0 Q v = J "∆H = J " (Hm, f − Hm, p) (7)
Upper wall Kinetic Stationary/no- Stationary/no- Stationary/no-
slip slip slip The conduction is part due to the bulk membrane material (Qc) and
Thermal Zero heat flux Coupled Coupled
(insulated)
the vapor-filled pores (Qv). The total membrane heat flux due to con-
Lower wall Kinetic Stationary/no- Stationary/no- Stationary/no- ductive and latent heat flux is expressed as:
slip slip slip
Qm = Qc + Q v = hA∆T (8)
Thermal Coupled Zero heat flux Coupled
(insulated) where h is the heat transfer coefficient, and Qc is the conductive heat
flux fraction and is described by Fourier heat equation as:
Table 2 km
Qc = − (Tm, f − Tm, p)
Summary of PVDF, vapor, and total membrane thermo-physical properties. δm (9)
Material Density Specific heat Conductivity (W/ Viscosity where km is the equivalent conductivity which is due to the bulk con-
(kg/m3) (J/kg·K) m·K) (pas) ductivity kb and the vapor conductivity kg and is estimated from as
weighted volume average. The subscripts f and p signify the feed and
PVDF [1] 1175 1325 0.2622 –
Vapor 0.554 2014 0.0261 – permeate, respectively.
Membrane 302.2 1896.9 0.0662 – Thermal efficiency (η): It represents the fraction of the heat used as
Saline sea 1013.2 4064.8 0.642 5.86E − 4 latent heat of evaporation to the total heat flux of Eq. (6) and is de-
watera [29]
scribed as:
Pure waterb 995.2 4182.1 0.613 8.38E − 4
[30] ∆Hm
η = J"
a
Qm (10)
At 4% salinity and 323 K.
b
At 303 K. where the denominator accounts for the total heat transfer that include
both latent heat flux and conduction and is written as:
J " = cm (P fsat + Ppsat ) (5) km (Tm, f − Tm, p )
Qm = J "∆Hm +
where cm is the distillation mass coefficient and Pfsat and Ppsat are the δm (11)
water vapor pressures at the feed and permeate membrane surfaces, Additionally, one can estimate the average heat transfer coefficient
respectively. The vapor pressure of pure water and its temperature re- which also can be used to assess the amount of heat transverse through
lationship is tabulated in steam tables and it is adjusted for saline and the membrane in form of latent and conductive type as depicted in
multiple specie solutions or waste-water as per the work of [11–13]. Fig. 4. A higher coefficient is typically indicative of larger amount of
The mass coefficient is obtained from the simulation following either heat and for the same membrane material this implies a larger latent
Knudson, molecular, Poiseuille or Monte Carlo as reported by Ding heat and better system.
et al. [37], Bui et al. [38] and Imdakum and Mussarra in [39]. The
common combination of Knudson and Poiseuille models is used here
3. Results and discussion
following the work of Chen et al. [20,40] and is described by:

εrp Mw εr 2 Mw Pm 3.1. Mesh sensitivity

cm = ck + cp = 1.064α (T ) + 0.125β (T )
τδm RTmt τδm RTmt ηv (6)
The planer flow configuration of Fig. 2 is discretized using struc-
where α(T), and β(T) are Knudsen diffusion model and Poiseuille flow tured quadratic mesh that capture the kinetic and thermal boundary
model contributions, respectively. Mw is molar mass of the water in (kg/ layers. Four levels of meshes have been implemented to assess the
mol), Tmt is mean membrane temperature (°C), R is gas constant, Pm is discretized mesh independency on the solution. Results of the weighted
mean pressure, δm thickness of the membrane, ηv is gas viscosity, r is area average pressure difference between the inlet and outlet for the

Fig. 4. Thermal energy path way cross the membrane of the base-
line DCMD (left) and the AGMD (right) configurations.

125
I. Janajreh et al. Desalination 424 (2017) 122–130

Table 3
Mesh sensitivity results on the evaluated pressure gradient of air gap, feed and permeate
channels.

Mesh level Cells number Air gap dP Error (%) Feed/permeate Error (%)
(pas) dP (pas)

Fine 1,142,400 25.310 – 450.4254 –

Baseline 285,600 25.272 0.1521 449.2267 0.2661
Coarse I 142,800 25.203 0.4243 444.5925 1.2950
Coarse II 7,1400 25.183 0.5022 444.5483 1.3048

feed and air gap is evaluated along with its absolute errors (see
Table 3). These meshes are denoted as fine, baseline, and coarse I and
coarse II. The fine mesh seems to be the best choice as far as accuracy
but at longer computation time. The deployed baseline mesh is ap-
propriately accurate with an absolute error in the evaluated pressure
drop of a few digits' percentage in the air gap channel and the feed and
permeate channels.

3.2. Effect of air gap integration

Fig. 5 shows the velocity vector field and temperature colored

contours of the baseline DCMD and AGMD models respectively. The
figure is stretched (50 ×) in the y-direction to better display of the
results. It shows the conventional parabolic profile for the velocity and
descending and ascending trend of the temperature in the feed and
permeate channels, respectively. These values are well behaved as they
are bounded by the imposed boundary conditions. The two channels
appear to be semi-isolated, and more so for the AGMD, due to the low
conductive membrane, and thermal resistance air gap. These results
serve as validation of the deployed high fidelity numerical model.
The membrane surface and bulk temperatures distribution are
captured and compared at three different velocity values of each
channel and the air in the air gap, i.e. Re = 10, 50, and 100 as depicted
in Fig. 6. The presence of air gap modifies temperature field noticeably
in the feed and permeate sides. It is also clear the AGMD configuration
adds another thermal resistance for the bottom membrane surface
leading to lower temperature polarization and thereby lower driving
pressure and mass flux through the membrane. As can be seen in Fig. 6, Fig. 6. AGMD and DCMD membrane and bulk surfaces temperature distribution at dif-
ferent Re numbers, Re = 10(a), Re 50(b), and Re 100(c).
the integration of air gap adds another insulation layer but its effect at
the two channels are quite different. At the feed side, air gap integration
results in capturing less heat and hence higher membrane surface Results of the mass flux distribution as per (5) are depicted in Fig. 8.
temperature is maintained which helps the mass flux and yield. On the The air gap clearly penalizes the mass flux significantly of 12–22% from
contrary, temperature at the bottom membrane surface is higher in the low to high Re values. That is a quite acceptable mass flux of 7.20 g/
AGMD configuration than that in the baseline DCMD. This attributed to m2·s for DCMD and lower value of 5.55 g/m2·s when using air gap in
lack of cooling potential of the as air gap stream. Overall, the difference similar boundary conditions. Results of the thermal efficiency dis-
in temperature across the upper and lower membrane surface in the tribution as per Eq. (11) are depicted in Fig. 9. At the implemented
AGMD is much less than in the DCMD and that leads to lower polar- conditions (Re = 100, Feed temperatures 50 °C and permeate tem-
ization and driving pressure that eventually translate into lower mass perature 25 °C) which can be naturally implemented during summer
flux. The performance of the AGMD can be more illustrated by using the period in the Gulf states thermal efficiencies near 12% and 9% is at-
TPC plot as seen in Fig. 7. Air gap show a significant drop in the TPC tained in the DCMD and AGMD, respectively. The trends at different Re
value down to thermally limited level (~ 0.35) compared to the base- follow the same of the mass flux trend. Summary of the attained
line DCMD (~ 0.55). This drop is below the desirable TCP range average TCP, mass flux, and thermal efficiency and relative loss due to
(0.5–0.75) observed across all the values of imposed values Re. gap integration are tabulated in Table 4.

Fig. 5. Velocity and temperature fields of the DCMD (left) and AGMD (right) at Re 10 (y axis is stretched for visualization purpose).

126
I. Janajreh et al. Desalination 424 (2017) 122–130

Fig. 7. TCP distribution of AGMD and the baseline DCMD.

Fig. 10. Bottom membrane surface heat transfer coefficient distribution of AGMD and the
baseline DCMD at different Re number.

Fig. 8. Mass flux distribution of AGMD and the baseline DCMD.

Fig. 11. Top membrane surface heat transfer coefficient distribution of AGMD and the
baseline DCMD at different Re number.

Table 5
AGMD and DCMD total heat flux and total average surface heat transfer coefficients.

Configuration Re Total surface Avg. surface heat transfer coef. (W/m2·K)

heat rate (W)
Top membrane Bottom membrane
surface surface

DCMD 10 646.15 124.39 167.07

50 1270.95 213.91 372.12
100 1474.54 239.56 455.62
Fig. 9. Thermal efficiency distribution of AGMD and the baseline DCMD. AGMD 10 510.64 90.33 110.33
50 847.18 132.02 178.42
100 934.16 141.56 196.20
Table 4
AGMD and DCMD quantitative metrics and comparison of TPC, mass flux, and thermal
efficiency.
the top, this can be explained from energy conservation that total heat
Setup TPC loss (%) TPC J′ loss (%) J′ (g/m2s) ὴ loss (%) ὴ (%) has to be the same. Moreover, heat transfer coefficients of AGMD at
both top and bottom membrane surfaces are lower when compared to
DCMD Re10 – 0.54 – 3.16 – 5.18
their corresponding values for DCMD configuration. Also, one can ob-
AGMD Re10 −35.8 0.35 − 12.1 2.78 − 12.1 4.56
DCMD Re50 – 0.56 – 6.19 – 10.14 serve that increasing Re leads to an increase in the surface heat transfer
AGMD Re50 −36.7 0.35 − 20.2 4.94 − 20.2 8.09 coefficient. Table 5 summarizes and compares the heat transfer coeffi-
DCMD Re100 – 0.58 – 7.20 – 11.79 cient and the total surface heat flux at the membrane surface for both
AGMD Re100 −37.7 0.36 − 22.3 5.55 − 22.3 9.09 AGMD and baseline DCMD. It is noticeable that integrating an air gap
(operated at Re = 100) can cause a reduction in DCMD total heat flux,
top membrane and bottom membrane surface heat transfer coefficients
Additionally, results of heat transfer coefficient distribution on both
by 36.65%, 40.93%, and 56.94%, respectively.
top and bottom membrane surfaces are decipited in Figs. 10 and 11,
respectively. In general, it is noticeable that bottom membrane surface
heat transfer coefficient is much higher than the top membrane sur-
3.3. Inlet temperature sensitivity study
faces. As per Eq.(8), temperature difference at the membrane lower
surface is much lower than its value at the top membrane surface which
Thermal efficiency of AGMD is strongly dependent on feed inlet
will lead to a heat transfer coefficient at the bottom surface higher than
temperature conditions as over 50% can be achieved at higher feed

127
I. Janajreh et al. Desalination 424 (2017) 122–130

Fig. 12. AGMD membrane and bulk surfaces temperature distribution at different feed inlet temperatures and at (a) Re = 10, (b) Re = 50, and (c) Re = 100.

Table 6
AGMD quantitative metrics and comparison at two different feed temperature.

Setup Avg. TPC Avg. J″ (g/ ΔJ″ gain Avg. η Δη gain

m2s) (g/m2s) (%) (%)
Re Feed temp
(°C)

10 50 0.3466 2.78 4.56

75 5.17 2.39 17.79 13.23
50 50 0.3537 4.94 8.09
75 9.60 4.56 33.04 24.95
100 50 0.3623 5.55 9.09
75 10.95 5.40 37.71 28.62

temperature of 75 °C [14] at the cost of heat source integration or heat

Fig. 13. AGMD mass flux distribution at different operating velocities and temperatures.
Fig. 14. AGMD thermal efficiency distribution at different operating velocity and tem-
perature.
regeneration. Similar approach as in the previous section was followed
that capture bulk and membrane temperatures (see Fig. 12) at three
different velocities (i.e. Re = 10 (Fig. 12a), Re = 50 (Fig. 12b), and
Re = 100 (Fig. 12c)) but using 75 °C feed inlet temperature instead of
50 °C. Temperature polarization coefficient between both bulk and
membrane size will remain the same as the previous case (Fig. 7) as
both temperature differences (Δ Tm and Δ Tb) were increased by the
same ratio. Fig. 13 represents the different mass flux distributions for
both 50 °C and 75 °C. It is clearly shown that increasing the feed inlet
temperature will cause a significant increase in the mass transfer
through the membrane from 5.55 g/m2·s to reach around 16 g/m2·s.
Moreover, results of thermal efficiency have been enhanced sig-
nificantly at higher inlet temperature. Fig. 14 depicts thermal efficiency
profiles for AGMD at different inlet temperature and velocities. Around
55% of thermal efficiency can be reached at 75 °C, whereas only 11% as
maximum thermal efficiency can be reached at 50 °C inlet feed tem-
perature. Similarly, the summary of the attained average TCP, mass
flux, and thermal efficiency and relative gain due higher feed tem-
perature are tabulated in Table 6.
Additionally, both bottom and top membrane surface heat transfer
coefficient for AGMD at different operating temperature and Reynolds

128
I. Janajreh et al. Desalination 424 (2017) 122–130

400 400
Surface Heat Transfer Coef. (W/ m2.K)

Surface Heat Transfer Coef. (W/ m2.K)

75˚C, Re10 75˚C, Re50 75˚C, Re100 75˚C, Re10 75˚C, Re50 75˚C, Re100
350 350 50˚C, Re10 50˚C, Re50 50˚C, Re100
50˚C, Re10 50˚C, Re50 50˚C, Re100
300 300
250 250

200 200

150 150

100 100

50 50

0 0
0 0.05 0.1 0.15 0.2 0 0.05 0.1 0.15 0.2
Spatial (m) Spatial (m)

Fig. 15. Bottom (left) and top (right) membrane surface heat transfer coefficient distribution of AGMD at different feed inlet temperature and Re number.

Table 7 cause a thermal and mass resistance, and thus a decrease in mass flux
AGMD total heat flux and total average surface heat transfer coefficients at 50 °C and and thermal efficiency of the AGMD. In other words, doubling the air-
75 °C feed inlet temperatures.
gap thickness (C to D) can reduce the TPC, mass flux, total heat flux,
Feed inlet Reynolds Total Surface heat transfer coefficient and thermal efficiency by 41.17%, 34.0%, 41.04%, 34.01% respec-
temperature (°C) number surface heat (W/m2·K) tively. Fig. 16 summarizes the most important metrics for different air
rate (W) gap thicknesses, in general, it is noticeable that increasing the air gap
Top Bottom thickness causes mass and thermal losses to the system. Adding vacuum
membrane membrane
surface surface
aspiration as in the case of VMD can recover this loss but at the cost of
additional setup and aspiration energy. One also can think to of re-
75 10 1021.28 111.46 144.14 duction of the thermal boundary layer resistance or “concentration
50 1694.37 157.89 229.33 polarization” by mean of static turbulence promoters in the form of
100 1868.32 168.35 251.81
conductive spacer and conductive gap [28,42].
50 10 510.64 90.33 110.33
50 847.18 132.02 178.42
100 934.16 141.56 196.20
4. Conclusion

number were plotted in Figs. 14 and 15. It is noticeable that at higher In this work, high fidelity validated numerical analysis using com-
inlet feed temperature, membrane surface heat transfer coefficient in- putational fluid dynamics (CFD) modeling is carried out to assess the
creases. However, bottom membrane surface has higher rates when performance of the AGMD. Different flow velocity is considered and
compared to the top surface of the membrane. This is due to the lower results are compared to the baseline DCMD. The results are demon-
surface temperature under the conservation of the same heat flux per strated in terms of temperature profiles, temperature polarization
the heat flux network presented in Fig. 4. Summary of total membrane coefficient (TPC), mass flux, heat flux, surface heat transfer coefficient,
heat flux and surface heat transfer coefficient, which is also transfer and thermal efficiency (η). The integration of a thin air gap led to re-
equal amount of heat but at a lower convective temperature gradients duction on the average TCP of 38%, in total membrane heat flux of
between the membrane and the gap, are represented in Table 7. 37%, in mass flux and thermal efficiency of 22%. Furthermore, in-
creasing the air gap thickness lead to much higher and drastic reduc-
tions in these metrics, i.e. doubling the Air-gap thickness can reduce the
3.4. Air gap thickness sensitivity study TPC, mass flux, total heat flux, and thermal efficiency by 61.5%, 50.2%,
61.6%, 50.2%, respectively Moreover, it was found that thermal effi-
Air gap thickness sensitivity study was done to observe the effect of ciency and mass flux are strongly dependent on feed inlet temperature
air gap thickness on the mass and heat transfer in the AGMD. Four conditions as mass flux at inlet feed temperature of 75 °C is about two
different thicknesses were implemented: 0.05 mm, 0.1 mm, 0.2 mm, times higher than that at 50 °C, this causes the thermal efficiency to
and 0.4 mm to quantify this influence on the AGMD performance. The reach 53%. Future work may target other gap parameters to compen-
setup is attempted at similar operational conditions, i.e. Re = 100, and sate for this temperature/polarization loss and driving pressure. Also
feed and permeate temperatures of 75 °C and 25 °C, respectively. inclusion of static turbulence promoters in the form of highly con-
Table 8 below summarized the different quantitative metrics, i.e. TPC, ductive spacer that reduces thermal boundary layer resistance or con-
mass flux, heat flux, surface heat transfer coefficient, and thermal ef- centrated polarization. Additionally, inducing gap vacuum pressure and
ficiency. Generally, it is noticed that increasing air gap thickness will interplay with the operational conditions or modifying in the gap

Table 8
Summary of AGMD quantitative metrics at different air-gap thickness, Re = 100 and T = 75 °C feed and 25 °C permeate.

Configuration Air gap thickness Avg. TPC Avg. mass flux (g/ Total membrane heat flux Avg. surface heat transfer coefficient (w/ Avg. thermal efficiency
(mm) m2·s) (W) m2·K) (%)

Top memb. Bottom memb.

No airgap 0.00 0.58 12.33 2949.09 288.96 680.46 42.46

AGMD-A 0.05 0.36 10.95 1868.32 168.35 251.81 37.71
AGMD-B 0.10 0.26 9.36 1366.30 118.77 154.96 32.22
AGMD-C 0.20 0.17 7.06 889.17 74.80 87.70 24.31
AGMD-D 0.40 0.10 4.66 524.29 43.05 47.05 16.04

129
I. Janajreh et al.
material that compensate for the quantified losses while continue to tap
on the advantages of the AGMD performance both qualitatively and
quantitatively. In summary, all evaluated metrics are reduced in the
AGMD compared to the conventional DCMD. Therefore a judicial
choice of the flow conditions, membrane properties, and setup is re-
quired to limit the anticipated performance reduction.
Acknowledgement
The authors would like to acknowledge the support of the Takreer
Research Center (TRC) and Masdar Institute of Abu Dhabi UAE under
grant number X2016-000022.
References
[1] A. Alkhudhiri, N. Darwish, N. Hilal, Membrane distillation: a comprehensive review,
Desalination 287 (2012) 2–18, http://dx.doi.org/10.1016/j.desal.2011.08.027.
[2] M. Khayet, Membranes and theoretical modeling of membrane distillation: a review, Adv.
Colloid Interf. Sci. 164 (2011) 56–88.
[3] K. Edward, H. John, V. Lienhard, Experimental study of thermal performance in air gap
membrane distillation systems, including the direct solar heating of membranes,
Desalination (2013) 100–111.
[4] A. El Amali, S. Bouguecha, M. Maalej, Experimental study of air gap and direct contact
membrane distillation configurations: application to geothermal and seawater desalina-
tion, Desalination 168 (2004) 357.
[5] A.S. Alsaadi, N. Ghaffour, J.-D. Li, S. Gray, L. Francis, H. Maab, G.L. Amy, Modeling of air-
gap membrane distillation process: a theoretical and experimental study, J. Membr. Sci.
445 (2013).
[6] A. Basile, Handbook of Membrane Reactors. Reactor Types and Industrial Applications,
Vol. 2 Woodhead Publishing Series in Energy, PA, 2013.
[7] M. Khayet, T. Matsuura, Air Gap Membrane Distillation, Membrane Distillation Principles
and Applications, Elsevier, 2011, pp. 361–398.
[8] F. Banat, A, J. Simandl, Membrane distillation for dilute ethanol, J. Membr. Sci. 163 (2)
(1999) 333–348, http://dx.doi.org/10.1016/s0376-7388(99)00178-7.
[9] M. Garcı́a-Payo, M. Izquierdo-Gil, C. Fernández-Pineda, Air gap membrane distillation of
aqueous alcohol solutions, J. Membr. Sci. 169 (1) (2000) 61–80, http://dx.doi.org/10.
1016/s0376-7388(99)00326-9.
[10] A. Alklaibi, N. Lior, Heat and mass transfer resistance analysis of membrane distillation, J.
Membr. Sci. 282 (1–2) (2006) 362–369, http://dx.doi.org/10.1016/j.
[11] I. Janajreh, D. Suwwan, Numerical simulation of direct contact membrane desalination
(DCMD): II, Int. J. Eng. Res. Innov. 6 (2) (2014) 21–33.
[12] I. Janajreh, D. Suwwan, Numerical simulation of direct contact membrane desalination in
conjugate heat transfer configuration: role of membrane conductivity, Int. J. Sustain.
Water Environ. Syst. 6 (2) (2014) 81–87.
[13] I. Janajreh, D. Suwwan, F. Hassan, Flow analysis of low energy direct contact membrane
desalination, Int. J. Therm. Environ. Eng. 8 (2) (2014) 133–138.
[14] I. Janajreh, D. Suwwan, R. Hashaikeh, Assessment of direct contact membrane distillation
under different configurations, velocities, and membrane properties, Appl. Energy 185
(2017) 2058–2073.
[15] D. Winter, Membrane Distillation – A Thermodynamic, Technological and Economic
Analysis, Dissertation University of Kaiserslautern, Germany, 2014 (ISBN 978-3-8440-
3706-7).
[16] M. Khayet, T. Matsuura, MD membrane modules, Membrane Distillation Principles and
Applications, Elsevier, 2011, pp. 361–398.
[17] I. Janajreh, R. Hashaikeh, D. Suwwan, Numerical Simulation of Membrane Desalination
in a Conjugated Heat Transfer Configuration: Role of Spacers, International Renewable
and Sustainable Energy Conference (IRSEC), Morocco, 2014.
[18] H. Yu, X. Yang, R. Wang, A.G. Fane, Numerical simulation of heat and mass transfer in
direct membrane distillation in a hollow fiber module with laminar flow, J. Membr. Sci.
130
Desalination 424 (2017) 122–130
Fig. 16. Bar chart summarizes quantitative me-
trics for different air gap thicknesses.
384 (1–2) (2011) 107–116.
[19] Y.M. Manawi, M. Khraisheh, A.K. Fard, F. Benyahia, S. Adham, Effect of operational
parameters on distillate flux in direct contact membrane distillation (DCMD): comparison
between experimental and model predicted performance, Desalination 336 (2014)
110–120.
[20] T.-C. Chen, C.-D. Ho, H.-M. Yeh, Theoretical modeling and experimental analysis of direct
contact membrane distillation, J. Membr. Sci. 330 (1–2) (2009) 279–287.
[21] A. Bahmanyar, M. Asghari, N. Khoobi, Numerical simulation and theoretical study on
simultaneously effects of operating parameters in direct contact membrane distillation,
Chem. Eng. Process. Process Intensif. 61 (2012) 42–50.
[22] L. Martınez-Dıez, M.I. Vázquez-González, Temperature and concentration polarization in
membrane distillation of aqueous salt solutions, J. Membr. Sci. 156 (2) (1999) 265–273.
[23] R. Saffarini, E. Summers, H. Arafat, V. Lienhard, Technical evaluation of stand-alone solar
powered membrane distillation systems, Desalination 286 (2012) 332–341.
[24] L.-Z. Zhang, C.-H. Liang, L.-X. Pei, Conjugate heat and mass transfer in membrane formed
channels in all entry regions, Int. J. Heat Mass Transf. 53 (5–6) (2010) 815–824.
[25] L.-Z. Zhang, Heat and mass transfer in a quasi-counter flow membrane-based total heat
exchanger, Int. J. Heat Mass Transf. 53 (23–24) (2010) 5478–5486.
[26] K. Charfi, M. Khayet, M. Safi, Numerical simulation and experimental studies on heat and
mass transfer using sweeping gas membrane distillation, Desalination 259 (1–3) (2010)
84–96.
[27] M. Asghari, A. Harandizadeh, M. Dehghani, H. Harami, Persian Gulf desalination using
air gap membrane distillation: numerical simulation and theoretical study, Desalination
374 (2015) 92–100.
[28] J. Swaminathan, H.W. Chung, D.M. Warsinge, F.A. AlMarzooqi, H. Arafat, J.H. Lienhard,
Energy efficiency of permeate gap and novel conductive gap membrane distillation, J.
Membr. Sci. 502 (2016) 171–178.
[29] B.S. Sparrow, Empirical equations for the thermodynamic properties of aqueous sodium
chloride, Desalination 159 (2) (Oct. 2003) 161–170.
[30] Carl L. Yaws, Chemical Properties Handbook: Physical, Thermodynamic, Environmental,
Transport, Safety and Health Related Properties for Organic and Inorganic Chemicals, 1st
ed, (1999).
[31] J.M.O.D. Záarate, A. Velázquez, L. Peña, J.I. Mengual, Influence of temperature polar-
ization on separation by membrane distillation, Sep. Sci. Technol. 28 (7) (May 1993)
1421–1436.
[32] K.W. Lawson, D.R. Lloyd, Membrane distillation: II direct contact MD, J. Membr. Sci. 120
(1) (1996) 123–133.
[33] G. Meindersma, C. Guijt, A. Haan, Desalination and water recycling by air gap membrane
distillation, Desalination 187 (1–3) (2006) 291–301, http://dx.doi.org/10.1016/j.desal.
2005.04.088.
[34] L.F. Greenlee, D.F. Lawler, B.D. Freeman, B. Marrot, P. Moulin, Reverse osmosis desali-
nation: water sources, technology, and today's challenges, Water Res. 43 (9) (May 2009)
2317–2348.
[35] B.L. Pangarkar, M.G. Sane, Heat and mass transfer analysis in air gap membrane dis-
tillation process for desalination, Membr. Water Treat. 2 (3) (2011) 159–173.
[36] Z. Jin, D. Yang, S. Zhang, X. Jian, Hydrophobic modification of poly(phthalazinone ether
sulfone ketone) hollow fiber membrane for vacuum membrane distillation, J. Membr. Sci.
310 (2008) 20–27.
[37] Z. Ding, L. Liu, M.S. El-Bourawi, R. Ma, Analysis of a solar-powered membrane distillation
system, Desalination 172 (1) (Feb. 2005) 27–40.
[38] V.A. Bui, L.T.T. Vu, M.H. Nguyen, Modelling the simultaneous heat and mass transfer of
direct contact membrane distillation in hollow fibre modules, J. Membr. Sci. 353 (1–2)
(May 2010) 85–93.
[39] A.O. Imdakmm, T. Matsuura, Simulation of heat and mass transfer in direct contact
membrane distillation (MD): the effect of membrane physical properties, J. Membr. Sci.
262 (1–2) (Oct. 2005) 117–128.
[40] T.-C. Chen, C.-D. Ho, Immediate assisted solar direct contact membrane distillation in
saline water desalination, J. Membr. Sci. 358 (1–2) (Aug. 2010) 122–130.
[41] P. Termpiyakul, R. Jiraratananon, S. Srisurichan, Heat and mass transfer characteristics of
a direct contact membrane distillation process for desalination, Desalination 177 (1–3)
(Jun. 2005) 133–141.
[42] S. Popović, D. Jovičević, M. Muhadinović, S. Milanović, M.N. Tekić, Intensification of
microfiltration using a blade-type turbulence promoter, J. Membr. Sci. 425–426 (0)
(2013) 113–120.