Desalination 430 (2018) 98–106

Contents lists available at ScienceDirect

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

Modeling of semibatch air gap membrane distillation T

⁎
Martijn Bindels, Niels Brand, Bart Nelemans
Aqua | Still, Nusterweg 69, 6136 KT Sittard, The Netherlands

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

Keywords: Membrane distillation is used to separate non-volatile components from an aqueous solution. This is particularly
Membrane distillation useful in the production of potable water from seawater. A major challenge to the latter is the high thermal
Energy consumption energy consumption. By using a semibatch process (salinity rises during the process) or multistage process
Theoretical model (several steady states after each other) instead of a steady state process (salinity remains constant during the
Process evaluation
process) thermal energy can be saved and more distillate can be produced. Furthermore, using semibatch
processes reduce the risk of bio fouling and scaling compared to steady state or multistage processes due to
salinity shocking and rinsing.
In this report, the thermal energy consumption of a steady state, multistage and semibatch membrane dis-
tillation process are measured using a AGDM membrane module from Aqua| Still. In addition a theoretical model
is developed and validated using the experimental results.
The theoretical model shows good correlation with the test results and can be used to compute a steady state,
multistage or semibatch process. Furthermore, experimental results confirmed that by using multistage or
semibatch membrane distillation processes instead of steady state processes, the thermal energy consumption
can be reduced.

1. Introduction the hot solution is in direct contact with the membrane. The water
vapor condenses on the other side of the membrane. DCMD is the
Membrane distillation relies on a temperature difference between simplest MD configuration. However, the main drawback is heat loss
two fluids separated by a hydrophobic membrane to realize mass due to conduction [8]. The AGMD configuration does not have this
transfer of the vapor phase. The separation is driven by the vapor drawback as there is an air gap between the membrane and the con-
pressure difference between the two surfaces of the hydrophobic, densation surface. The main disadvantage of AGMD system is the added
micro-porous membrane and only vapor molecules are allowed to pass resistance to the mass transfer [8]. In SGMD, the water vapor is swept
the membrane [1]. by an inert gas to condense outside the MD module. This results in an
Membrane distillation gained attention in the recent years because enhanced mass transfer coefficient due to the non-stationary gas barrier
it is a promising solution to reduce the water scarcity. There are around [8]. On the other hand, in VMD there is negligible heat loss by con-
1.2 billion people living in areas of physical water scarcity [2]. More- duction due to the vacuum at the permeate membrane side. In VMD and
over, there are 500 million people approaching the situation of physical SGMD, the condensation takes place outside the MD module and results
scarcity and another 1.6 billion people are facing economic water in a more difficult setup.
shortage. The situation is worsened by the high population growth rate The thermal efficiency of MD is often described by the Gained
of 80 million people a year [2]. An example of the growing water Output Ratio [9], GOR:
shortage is the Arabian Peninsula, where the fresh water demand in- hf
creases at a rate of at least 3% annually [3]. Another benefit of mem- GOR =
qin (1)
brane distillation is that is can be used with alternative energy sources
such as solar energy to reduce water scarcity in remote areas [4,5]. MD processes can be steady state, multistage with elevating salinity
There are four major membrane distillation configurations which brine or semibatch. Summers et al. [7] made a comparative study that
are: direct contact membrane distillation (DCMD), air gap membrane showed that steady state MD system typically have a low GOR. Lu and
distillation (AGMD), sweep gas membrane distillation (SGMD) and Chen [39] modeled a multistage AGMD system for analyzing the op-
vacuum membrane distillation (VMD) [6,7]. In the DCMD configuration timal path for the cold stream. However, there was no conclusion on the

⁎
Corresponding author.
E-mail addresses: [email protected] (M. Bindels), [email protected] (N. Brand), [email protected] (B. Nelemans).

https://doi.org/10.1016/j.desal.2017.12.036
Received 4 April 2017; Received in revised form 1 September 2017; Accepted 19 December 2017
Available online 04 January 2018
0011-9164/ © 2017 Elsevier B.V. All rights reserved.
M. Bindels et al. Desalination 430 (2018) 98–106

and DCMD has therefore the lowest thermal efficiency. SGMD and VMD
are more complicated to test and the practical applications are limited.
Besides, if a semibatch process is more efficient compared to a steady
state process on an AGMD module, it should also be more efficient on a
DCMD, SGMD or VMD configuration.
Generally, after the water treatment the feedwater is usually col-
lected in a tank before entering the modules. When concentrating it in a
steady state or multistage manner, the feed water will be added gra-
dually to keep the salinity constant at each stage. However, when op-
erating in a semibatch process a large amount of the feedwater will be
added to the tank, and the latter will be refilled once the desirable
salinity has been reached. This means that all three processes are fed
feedwater from a tank at the same salinity; however the salinity of the
water at the input of the module is different. Examples of brine con-
centrations can be found in [11,12]. This report will use salt (NaCl)
Fig. 1. Schematic graph of the salinity during the steady state, multistage and semibatch water as a concentrating medium. The reasons for this are that scaling
process. and biofouling is low and will therefore not influence the test results.
In this work, a mathematical model is being presented to calculate
GOR or other efficiency analyses. Chung et al. [9] concluded that in- the flux and energy consumption of a spiral wound AGMD module with
creasing the number of stages increased the GOR with VMD. a semibatch process. Furthermore, an experimental study is employed
An infinite multistage MD can be described as a semibatch process to evaluate the model and to investigate the saved thermal energy of
where the salinity of the brine increases during the process. The salinity semibatch AGMD with a salinity between 4% and 10%. 4% is chosen as
during the process in steady state, multistage and semibatch is shown standard seawater has a salinity of 3.5% [13], more seawater con-
schematically in Fig. 1. A schematic graph of the total thermal energy centrations can be found in [14]. To prevent the risk of scaling the
needed to produce the same amount of distillate for the steady state and maximum salinity for the semibatch test is 10% to 12%. A commercially
semibatch processes is shown in Fig. 2. available AS-7 AGMD module from Aqua| Still is used in the experi-
A possible advantage of using semibatch instead of steady state or ments.
multistage is the reduction of bio-fouling and scaling during the pro-
cess. Bio-fouling can occur at low feed temperatures and low salinity
and can be a problem at the cooling channel of an AGMD module or at 2. Theoretical background
the cold heat exchanger. However, at higher salinity scaling may occur.
Furthermore, in commercial MD-systems bio fouling is likely to be a The performance of a MD-module depends on the heat and mass
concern as it has been a critical issue in RO membranes [10]. Bacterial transfer. With the equations that were found in literature a theoretical
growth can be hindered by using higher salinity, elevated temperatures model is developed to theoretically compute the distillate flux of a
[3] and low pH values [10]. However, the risk of scaling increases with spiral wound MD-module and to evaluate the thermal energy con-
higher salinity and higher temperatures [10]. In a semibatch process sumption. The model will calculate a steady state process. The semi-
the periodically lower salinity is used to reduce scaling and the peri- batch process can be calculated by using several steady states with in-
odically higher salinity is used to reduce bacterial growth. This results creasing salinity. The following assumptions were made:
in less scaling at the beginning of the cycle and less bacterial growth at
the end of the cycle compared to steady state processes. 1. The system operates at steady state.
Other advantages are the simplicity of the semibatch setup, which 2. The module operates in a counter current flow direction.
means it is a cheaper option compared to steady state or multistage 3. The distillate is trapped inside the mesh of the spacer and no liquid
processes and that less feed water is required for the same amount of film is present. Furthermore, the amount of stagnant distillate inside
distillate compared to one pass through system thanks to a higher the module does not change.
concentration factor. 4. The water vapor has to travel from the membrane to the condenser
For several reasons, the AGMD configuration is tested in this report. and cannot condense on the spacer or on the trapped water in the air
As explained earlier, DCMD loses more heat by conduction than AGMD gap spacer.
5. No heat is exchanged with the surroundings as it is a self-isolating
system.
6. Membrane compaction effects as reported in [15] are ignored.
7. The temperature that is measured in front of the module is the same
as that at the entrance of the envelope.

The envelope is divided into 25 sections of equal temperatures

difference. The sections are discussed further down in the paper. In the
following two sections the heat and mass transfer is discussed. The
thermophysical properties can be computed from [16,17] and the latent
heat can be computed from [18–20]. The velocity of the spacer filled
channel can be computed from [21]. The formula from [21] is adapted
to include the removal of distillate, the number of channels and the unit
conversion which results in the following two equations:

Fic,in
v ich,c =
3600n ch Hch h sp ϵsp (2)
Fig. 2. Schematic graph of the total thermal energy needed to produce the same amount
of distillate.

99
M. Bindels et al. Desalination 430 (2018) 98–106

where κs, m is the conductivity of the membrane polymer and κg is the

thermal conductivity of air that can be calculated by [28]:

k g = 2.72 × 10−3 + 7.77 × 10−5Tm (10)

The thermal conductivity of the spacer filled distillate channel can

be calculated using the following equation:
k sp = X spκ s,sp + X g κ g + Xw kw (11)

The condensation layer of the distillate channel can be calculated by

Nusselt's film condensation theory. However, Nusselt film condensation
does not take the interfacial interactions of the air gap into account.
There exists to the authors’ knowledge no formulation to calculate the
water layer in a spacer filled air gap of a full-scale module. Therefore,
the condensation layer in the distillate channel, Xw, is used as a cali-
bration factor. The heat transfer of the membrane and the air gap can be
calculated by:
−1 −1 −1
κ κ sp ⎞ ⎞
α m + a = ⎜⎛ ⎛ m ⎞
⎜ ⎟ +⎛⎜
⎟ ⎟

⎝ δm ⎠
⎝ ⎝ δa ⎠ ⎠ (12)

The following formula will later be used to compute the driving

temperature difference of one slice:
Fig. 3. Schematic view of a spacer in a module. Hch is the channel height, ϵsp is the spacer −1
porosity, hsp is the thickness of the spacer and nch is the amount of channels in the
module.
⎛ 1
α fc = ⎜ +
(
1 + αch,c
con
α
) ⎞⎟
⎜ α ch,h α ch,c ⎟
⎝ ⎠ (13)
Di − 1
Fih,in − ρ
n ch
v ich,h =
3600n ch Hch h sp ϵsp (3) 2.2. Mass transfer

A schematic view of one of the envelopes of the MD module is

The mass transfer across the membrane can be calculated by [29]:
shown in Fig. 3 with the symbols of Eqs. (2) and (3).
J = Cm (Pm1 − Pm2) (14)
2.1. Heat transfer
where the pressure difference across the membrane can be calculated
with the Clausius-Clapeyron Equation [30].
The heat transfer through the brine and the cooling channel can be
calculated by: dP
J = Cm [Tm1 − Tm2]
dT (15)
k sw Nuf,p
α ch = dP
dh,sp (4) can be calculated by [31]:
dT

Hitsov et al. [22] reported that the Nusselt's number of the spacer dP P∗hf
=
filled channel used in this report can be calculated from: dT RT2m (16)
0.25 Combining Eqs. (13) and (14) and adding the water activity the
Pr b,f,p ⎞
Nu f,p = 0.223(Ref,p )0.69 (Pr b,f,p )0.13 ⎜⎛ ⎟ following equation is obtained:
⎝ Prwf,p ⎠ (5)
Cm + a P∗hf
Other Nusselt's numbers for spacers can be found in [21] or should J= a w ΔTest
RT2m (17)
be determined experimentally according to the formulas of [23–25].
For an empty channel the equation from [26] can be used. According to where ΔTest is the estimated temperature difference across the mem-
Alsaadi et al. [27] the Reynolds number and hydraulic diameter can be brane. The factor Cm + a is a combined factor for the permeability of the
calculated from: membrane and the diffusion resistance in the air gap that can be cal-
culated with the following equation:
i
ρsw dh,sp v ich
Rech =
μ sw (6) Cm + a = Hp (C−m1 + A−a 1)−1 (18)

4ϵsp df h sp where Hp will be used as a calibration factor.

dh,sp =
2df + 4(1 − ϵsp )h sp (7) According to Martínez et al. [32] the membrane permeability can be
computed with the following equation:
The Prandtl number can be computed from [22]:
−1
ϵm M ⎛ 1 pa ⎞
μ sw Cp,sw Cm = ⎜+ ⎟

Pr = τm δm RT ⎝ Dk PDwa ⎠log mean (19)

κ sw (8)
The thermal conductivity of the membrane matrix can be calculated As different values of pa exists in both membrane surfaces the
−1
by [22]: logarithmic mean of ( 1
Dk
+
pa
PD wa ) should be used to calculate the
−1 permeability [32].
ϵm 1 − ϵm ⎤
km = ⎡
⎢ κ g + κ s,m ⎥
Dk is the resistance that comes from collisions from molecules with
⎣ ⎦ (9) the membrane [8]. The Knudsen diffusion can be calculated by [32]:

100
M. Bindels et al. Desalination 430 (2018) 98–106

2 8RT 0,5 Dmd = n ch ∑ Di (31)

Dk = r⎛ ⎞
3 ⎝ πM ⎠ (20)
The diffusion coefficient, Dwa, comes from collision of molecules
L ch = ∑ Li (32)
with other molecules [8] which can be calculated by [18,19]: The flux of each slice can be calculated with Eq. (17). The distillate
1.895 × 10−5T 2.072 production can be calculated with the following equation.
Dwa =
P (21) i i
⎛ −J ih if + α fc ln(t d) ⎞
B
The mass transfer also depends on Poiseuille flow (viscous flow) but ΔTical = ⎜ 1000 i
− t av +C− i
+ 3ΔTiest ⎟ 0.25
α im + a + α ifc A − ln(P sw )
can be neglected as the net flux of air is small compared to the flux of ⎜ ⎟
⎝ 1000 ⎠
water [33].
(33)
The mass transfer coefficient of the air gap can be calculated with
[27]: m B
vH dh,sp Cip,sw (Tih,in − Tih,out) ⎛ln(t id) − ΔTiest − t av
i
− i )
⎞
⎝ A − ln(Psw ⎠
D M Li =
A a = wa −1
δa RTav (22) ⎛ 1 +
⎜
1
+
1 ⎞
⎟
i α ich,c α con
⎝ α ch,h ⎠
According to Hitsov et al. [34] the vapor pressure of an aqueous salt
(34)
solution, Psw, can be calculated by:
Psw = a w P∗ (23)
2.4. Thermal energy and efficiency
where the vapor pressure of pure water can be computed with the
Antoine equation [20,35]: The thermal energy consumption of the module can be calculated
by:
B
P∗ = exp ⎛A − ⎞
⎝ T − C⎠ (24) FΔTtopC p,sw ρsw
Q=
3600 (35)
The Antoine constants A, B and C are 23.328, 3841 and 45 re-
spectively [36]. The activity of water of an aqueous NaCl solution can where ΔTtop is the temperature difference between the entry tempera-
be computed with the following equation [34]: ture of the hot brine channel and the exit temperature of the cooling
channel.
a w = 1 − 0.03112mNaCl − 0.001482m2NaCl (25) The GOR can be determined by:
Am Jhf
2.3. Calculation algorithm GOR =
Cp,sw Fρsw ΔTtop (36)

The calculation algorithm is shown in Fig. 5. First, the temperatures

of the entry and exit of the brine and cooling channel should be known 3. Experimental
or guessed. Secondly the length of the membrane envelope (which
contains one brine channel, cooling channel and air gap) is divided into 3.1. Experimental setup
25 sections with equal temperature differences. It is found that dividing
into 25 sections gave accurate results. The temperature of the brine and A commercially available AS-7 counter flow AGMD module from
cooling channel of each slice can be calculated with the following Aqua |Still was used. The spiral wound AMGD module consist of 6
equations: evaporator channels, 6 condenser channels, and 12 distillate channels.
The membrane area is 7.2 m2. The distillate channel is created by a
Tm m
h,in − Th,out
−1
Tih,in = Tih,in − mesh spacer which is inserted between the evaporator and condenser
25 (26)
channel [6]. The hot and cold channels have the same geometry.
Tih,out = Tih,in
+1
(27) The brine flow is measured by a flow meter at the entrance of the
cooling channel.
Tm m
c,out − Tc,in The temperatures are measured by 4 resistance thermometers with
−1
Tic,out = Tic,out −
25 (28) an accuracy of 0.7 K.
The salinity of the brine is measured with a density meter. The feed
+1
Tic,in = Tic,out (29) solution consists of tap water where NaCl is added to raise the salinity.
m m
where Th, in and Th, out are the temperatures of the brine channel at
the beginning and the end of the slice. Tc, outm and Tc, inm are the 3.2. Steady state and multistage
temperatures of the cooling channel at the end and at the beginning.
The temperatures of a slice can be seen in Fig. 4. The produced distillate will be transmitted to a tank to measure the
After the temperatures of the channels are known, a temperature amount of yielded distillate. To keep the salinity constant the same
difference over the membrane and air gap, Test, should be estimated. If amount of produced distillate was added to the feed tank. This was
calculated and estimated temperatures are not equal, the calculated done by using a level switch. To prevent big changes in salinity the
temperature becomes the estimated temperature. This is repeated until range of the level switch was small.
ΔTest and ΔTcal are equal after calculating the temperature difference. The brine flow for evaluating the theoretical model is between the
The last step is to calculate the distillate flux and the length of the slice. 0.81 and 0.87 m3/h. The brine flow for evaluating the energy con-
The length of the slice is determined by calculating the length that is sumption will be held at 0.83 m3/h. The cross flow velocity at 0.83 m3/
needed to get the temperature difference according to the calculated h is around 0.048 m/s.
heat transfers coefficients of the channels and the evaporation energy. The multistage evaluation will consist of 2 stages. The first and
The distillate flow of a slice, calculated module length and distillate second stage consist of a salinity ranging from 42 to 82 and 84–105 g/
flow of the module are computed by: kg respectively. To evaluate the saved thermal energy 50% of the dis-
tillate will be produced at the first stage and 50% will be produced at
Di = J iAm = J iLH
i
ch (30) the second stage.

101
M. Bindels et al. Desalination 430 (2018) 98–106

Fig. 4. Magnification of a slice. Red, white and blue are the brine
channel, membrane + air gap and the cooling channel respec-
tively. (For interpretation of the references to colour in this
figure legend, the reader is referred to the web version of this
article.)

Fig. 6. Steady state test setup.

Fig. 5. Calculation algorithm for each slice.

The distillate flux and energy consumption of a steady state process

will be measured at salinities and temperatures of Table 1. Those of the
multistage process will be according to two steady state test, which is
shown in Table 2. The steady state test setup is shown in Fig. 6. The
amount of produced distillate will be measured using a 6-liter tank and
a weighing scale. According to the measured weight of the distillate the
flux will be calculated. Fig. 7. Semibatch test setup.

3.3. Semibatch semibatch test setup is shown in Fig. 7.

In semibatch, the salinity of the brine tank rises during the test. The
salinity of the brine at the begin and at the end of the test is shown in 4. Results and discussion
Table 3. The produced distillate will be stored in a bigger tank than the
steady state test. Every 10 min, the produced distillate will be quanti- 4.1. Model validation
fied using a scale with a higher measuring range than the steady state
scale and the salinity will be measured with a density meter. The As can be seen in Figs. 8 and 9 the theoretical model correlates well

102
M. Bindels et al. Desalination 430 (2018) 98–106

Fig. 11. Semibatch test results – GOR. With error bars.

Fig. 8. Predicted versus measured distillate flux of test 1 SS to 23 SS, steady state. With
error bars.

Fig. 12. Distillate flux during the steady state, multistage and semibatch process. With
Fig. 9. Calculated versus measured flux, semibatch. With error bars. error bars for the semibatch process.

Fig. 10. Semibatch test results – flux. With error bars.

Fig. 13. GOR during the steady state, multistage and semibatch process. With error bars
for the Semibatch test.

103
M. Bindels et al. Desalination 430 (2018) 98–106

Table 1
Steady state test parameters.

Test number Salinity [g/kg] Flow [m3/h] Hot brine in [C°] Cold brine in [C°] Test number Salinity [g/kg] Flow [m3/h] Hot brine in [C°] Cold brine in [C°]

1 SS 41 0.83 50.3 19.1 15 SS 86 0.86 69.7 18.7

2 SS 39 0.83 50.4 19.1 16 SS 87 0.85 70.3 18.4
3 SS 41 0.84 60.0 18.7 17 SS 117 0.82 50.3 19.0
4 SS 41 0.85 60.0 19.2 18 SS 118 0.81 50.4 19.1
5 SS 40 0.83 66.1 19.2 19 SS 119 0.82 59.1 18.3
6 SS 41 0.79 70.1 19.1 20 SS 118 0.82 60.9 18.7
7 SS 43 0.79 70.1 19.0 21 SS 119 0.82 66.1 19.3
8 SS 40 0.82 70.2 18.9 22 SS 120 0.83 70.4 18.1
9 SS 41 0.82 70.2 19.1 23 SS 119 0.82 70.2 18.6
10 SS 85 0.87 51.8 19.5 24 SS 42 0.83 69.5 25.2
11 SS 85 0.86 49.8 19.6 25 SS 84 0.83 70.1 26.2
12 SS 84 0.87 60.4 19.7 26 SS 105 0.83 70.2 25.9
13 SS 85 0.86 60.4 19.6 27 SS 124 0.83 70.1 25.8
14 SS 86 0.87 65.7 19.7

Table 2 Table 4
Multistage test 28 MS parameters. Practical evaluation of semibatch versus one stage steady state.

First stage Second stage Semibatch Steady state Salinity Salinity Steady state Total time
semibatch steady versus needed for
Input condition Output condition Input condition Output condition [g/kg] state [g/ semibatch steady
kg] thermal state versus
Test 24 SS Test 25 SS Test 25 SS Test 26 SS energy semibatch
consumption

Table 3 Test 29 MS Test 25 SS 41 → 82 84 2.1% 4.8%

Semibatch test parameters. Test 30 MS Test 25 SS 42 → 83 84 4.5% 4.3%
Test 31 MS Test 26 SS 45 → 107 105 8.5% 9.3%
Test Start salinity End salinity Flow Brine in Cold in Test 32 MS Test 26 SS 44 → 107 105 9.7% 9.9%
number [g/kg] [g/kg] [m3/h] [C°] [C°] Test 33 MS Test 27 SS 84 → 122 124 3.2% 10.7%
Test 34 MS Test 27 SS 84 → 122 124 3.5% 10.2%
29 SB 41 82 0.83 70.0 25.5
30 SB 42 83 0.83 70.0 26.2
31 SB 45 107 0.83 70.4 25.6 state process. Furthermore, 2.4% less time is needed to produce the
32 SB 44 107 0.83 71.3 26.2 same amount of water. The steady state and semibatch test result are
33 SB 84 122 0.83 69.8 25.9
compared in Table 4. The deviation of the test results between cases
34 SB 84 122 0.83 69.6 35.8
with similar input and output conditions comes from the accuracy of
the weight scale and the slight temperature difference. Semibatch has
with the distillate flux. The calibration factor, Hp and Xw should be 0.4 the lowest amount of thermal energy consumption, followed by mul-
and 0.2 respectively. The deviation of the steady state test comes from tistage.
the test setup, especially the temperature sensors. The deviation of the The result of modeling the multistage test results in a 2.7% less
semibatch test comes from the weighing scale. The change in distillate thermal energy consumption and 1.7% less time needed to produce the
flux for the semibatch test was in some cases too low to measure due to same amount of distillate. In Table 5 the thermal energy consumption
the precision of the scale. This is the reason that the flux stayed constant and total time needed of the semibatch process is compared to the
in between 40 and 80 min in Fig. 9. Hence the results of the semibatch steady state process. The difference between the theoretical model and
test are indicative, more tests should be done for more precise results. practical evaluation comes from inaccuracies in the test setup. As the
theoretical results have a good correlation with the test results the
model can be used to estimate the performance and energy consump-
4.2. Process evaluation tion of the semibatch process.

In Fig. 10 the distillate flux is plotted against the salinity. As the

salinity raises during the process the distillate production drops, which
corresponds to the results of [37,38]. Furthermore, the GOR is lower at
higher concentrations, as can be seen in Fig. 11. Table 5
In Fig. 12 the flux is shown of the three processes. Steady state has Theoretical evaluation of semibatch compared to one stage steady state.
the lowest flux, multistage has a higher flux at the first stage than at the
Semibatch Steady state Salinity Salinity Steady state Total time
second stage. The semibatch process start with a high flux but slowly semibatch steady versus needed for
drops to the same level as steady state. The GOR during a semibatch test [g/kg] state [g/ semibatch steady state
is shown in Fig. 13 together with the tested values for steady state and kg] thermal versus
the second stage of multistage. The GOR of the semibatch process de- energy semibatch
consumption
clines during the process. The GOR of steady state is constant and the
GOR of multistage is high at first, but is lower at the second stage. The Test 29 SB Test 25 SS 41 → 82 84 3.1% 5.9%
total time needed to produce the same amount of distillate is longer for Test 30 SB Test 25 SS 42 → 83 84 5.5% 5.3%
steady state than multistage or semibatch. The error bars for the steady Test 31 SB Test 26 SS 45 → 107 105 7.9% 8.7%
Test 32 SB Test 26 SS 44 → 107 105 10.0% 10.1%
state and multistage are too small to see and therefore not shown in
Test 33 SB Test 27 SS 84 → 122 124 3.4% 10.9%
Figs. 12 and 13. Test 34 SB Test 27 SS 84 → 122 124 4.0% 10.7%
The multistage test used 3.3% less thermal energy than the steady

104
M. Bindels et al. Desalination 430 (2018) 98–106

5. Conclusion Greek letters

A theoretical model for spiral wound air gap membrane distillation α Heat transfer coefficient (W m− 2 K− 1)
(AGMD) has been developed and its validation against experimental ΔTtop Temperature difference between the top of the module (K)
data shows a good correlation with the test results. δ Thickness (m)
Comparison of the semibatch process, multistage and steady state ϵ Porosity
processes showed that the semibatch process results in between 4.8% κ Thermal conductivity (W m− 1 K− 1)
and 10.7% less time and 2.1% to 9.7% less thermal energy consumption μ Dynamic viscosity (kg m− 1 s− 1)
than the steady state process. The multistage process results in 2.4% ρ Density (kg m− 3)
less time and 3.3% less thermal energy consumption than the steady τ Tortuosity
state process. Hence the semibatch process is more efficient than the
multistage process and the multistage process is more efficient than the Antoine coefficients
steady state process.
The theoretical model can be used to compute the semibatch process A Antoine coefficient
but this will result in a small deviation due to the uncertainty of the B Antoine coefficient (K)
model. The thermal energy consumption calculated by the theoretical C Antoine coefficient (K)
model is between 3.1% and 10% lower when using semibatch instead of
steady state. The time needed to produce the same amount of distillate Dimensionless groups
is 5.9% to 10.9% lower when using semibatch instead of steady state.
By using multistage 2.7% less thermal energy is consumed and 1.8% Nu Nusselt's number
less time is needed theoretically compared to steady state. There is a Pr Prandtl number
good correlation between the theoretical model and the experimental Re Reynolds number
results. Therefore, the model can be used to compute the performance
of a spiral wound AGMD module with a steady state, multistage or Superscripts & subscripts
semibatch process.
a Air gap
Nomenclature av Average temperature
b Bulk
Aa Mass transfer coefficient of the air gap (kg m− 2 s− 1 Pa− 1) c Cooling channel
Am Membrane area (m2) ch Channel
aw Activity factor cal Calculated
Cm Mass transfer coefficient of the membrane (kg m− 2 s− 1 con Condenser
Pa− 1) est Estimated
Cp Specific heat (J kg− 1 K− 1) f Feed
D Distillate flux (kg h− 1) fc For calculation
df Filament diameter (m) g Air
dh Hydraulic diameter (m) h Hot brine channel
Di − 1 Amount of produced distillate in the channel before slice i i Slice
(kg) in Beginning of the slice
Dk Knudsen diffusion (m2 s− 1) m Membrane
F Brine flow (m3 h− 1) md Module
Hch Height of the channel (m) m+a Combined membrane and air gap
hf Latent heat of evaporation (kJ kg− 1) NaCl Sodium chloride
Hp Calibration factor out End of the slice
hsp Spacer thickness (m) p Permeate
J Distillate flow (kg m− 2 h− 1) s Property of solid material
ln(tdi) Logarithmic temperature difference (°C) sp Spacer
L Length (m) sw Seawater
m Mole fraction w Water
M Molecular weight of water (kg mol− 1) 1 One side of the membrane
mNaCl Molality of NaCl (mol kg− 1) 2 Other side of the membrane
nch Number of envelopes in module ∗ Vapor
P Pressure (Pa)
pa Partial pressure of air entrapped in membrane pores (Pa) Abbreviations
pDwa Molecular diffusion (Pa m2 s− 1)
qin Specific heat input (kJ kg− 1) AGMD Air gap membrane distillation
r Radius of membrane pores (m) DCMD Direct contact membrane distillation
R Universal gas constant (J mol− 1 K− 1) VMD Vacuum membrane distillation
S Salinity (g kg− 1) SGMD Sweep gas membrane distillation
T Temperature (K) GOR Gained output ratio
t Temperature (°C) SS Steady state
v Velocity in spacer filled channel (m s− 1) MS Multistage
W Width (m) SB Semibatch
X Fraction of area

105
M. Bindels et al. Desalination 430 (2018) 98–106

References Elsevier, 2011.

[20] J. Xu, Y.B. Singh, G.L. Amy, N. Ghaffour, Effect of operating parameters and
membrane characteristics on air gap membrane distillation performance for the
[1] K. Zhao, W. Heinzl, M. Wenzel, S. Büttner, F. Bollen, G. Lange, S. Heinzl, N. Sarda, treatment of highly saline water, J. Membr. Sci. 512 (2016) 73–82, http://dx.doi.
Experimental study of the memsys vacuum-multi-effect-membrane-distillation (V- org/10.1016/j.memsci.2016.04.010.
MEMD) module, Desalination 323 (2013) 150–160, http://dx.doi.org/10.1016/j. [21] A. Hagedorn, G. Fieg, D. Winter, J. Koschikowski, A. Grabowski, T. Mann,
desal.2012.12.003. Membrane and spacer evaluation with respect to future module design in mem-
[2] W.L. Ang, A.W. Mohammad, N. Hilal, C.P. Leo, A review on the applicability of brane distillation, Desalination 413 (2017) 154–167, http://dx.doi.org/10.1016/j.
integrated/hybrid membrane processes in water treatment and desalination plants, desal.2017.03.016.
Desalination 363 (2015) 2–18, http://dx.doi.org/10.1016/j.desal.2014.03.008. [22] I. Hitsov, L. Eykens, K. De Sitter, C. Dotremont, L. Pinoy, B. Van der Bruggen,
[3] L.D. Tijing, Y.C. Woo, J.S. Choi, S. Lee, S.H. Kim, H.K. Shon, Fouling and its control I. Nopens, Calibration and analysis of a direct contact membrane distillation model
in membrane distillation-a review, J. Membr. Sci. 475 (2015) 215–244, http://dx. using Monte Carlo filtering, J. Membr. Sci. 515 (2016) 63–78, http://dx.doi.org/10.
doi.org/10.1016/j.memsci.2014.09.042. 1016/j.memsci.2016.05.041.
[4] A. Ruiz-aguirre, D.-C. Alarcón-Padilla, G. Zaragoza, Productivity analysis of two [23] M. Gryta, M. Tomaszewska, A.W. Morawski, Membrane distillation with laminar
spiral-wound membrane distillation prototypes coupled with solar energy, Desalin. flow, 5866 (1997) 2–6.
Water Treat. 55 (2015) 2777–2785, http://dx.doi.org/10.1080/19443994.2014. [24] J. Phattaranawik, R. Jiraratananon, A. Fane, Heat transport and membrane dis-
946711. tillation coefficients in direct contact membrane distillation, J. Membr. Sci. 212
[5] A. Ruiz-Aguirre, M.I. Polo-López, P. Fernández-Ibáñez, G. Zaragoza, Assessing the (2003) 177–193, http://dx.doi.org/10.1016/S0376-7388(02)00498-2.
validity of solar membrane distillation for disinfection of contaminated water, [25] C.C. Wang, On the heat transfer correlation for membrane distillation, Energy
Desalin. Water Treat. 55 (2015) 2792–2799, http://dx.doi.org/10.1080/19443994. Convers. Manag. 52 (2011) 1968–1973, http://dx.doi.org/10.1016/j.enconman.
2014.946717. 2010.11.014.
[6] H.C. Duong, P. Cooper, B. Nelemans, T.Y. Cath, L.D. Nghiem, Evaluating energy [26] Y. Taamneh, K. Bataineh, Improving the performance of direct contact membrane
consumption of air gap membrane distillation for seawater desalination at pilot distillation utilizing spacer-filled channel, Desalination 408 (2017) 25–35, http://
scale level, Sep. Purif. Technol. 166 (2016) 55–62, http://dx.doi.org/10.1016/j. dx.doi.org/10.1016/j.desal.2017.01.004.
seppur.2016.04.014. [27] A.S. Alsaadi, N. Ghaffour, J.D. Li, S. Gray, L. Francis, H. Maab, G.L. Amy, Modeling
[7] E.K. Summers, H.A. Arafat, J.H. Lienhard V, Energy efficiency comparison of single- of air-gap membrane distillation process: a theoretical and experimental study, J.
stage membrane distillation (MD) desalination cycles in different configurations, Membr. Sci. 445 (2013) 53–65, http://dx.doi.org/10.1016/j.memsci.2013.05.049.
Desalination 290 (2012) 54–66, http://dx.doi.org/10.1016/j.desal.2012.01.004. [28] A. Bahmanyar, M. Asghari, N. Khoobi, Numerical simulation and theoretical study
[8] A. Alkhudhiri, N. Darwish, N. Hilal, Membrane distillation: a comprehensive re- on simultaneously effects of operating parameters in direct contact membrane
view, Desalination 287 (2012) 2–18, http://dx.doi.org/10.1016/j.desal.2011.08. distillation, Chem. Eng. Process. Process Intensif. 61 (2012) 42–50, http://dx.doi.
027. org/10.1016/j.cep.2012.06.012.
[9] H.W. Chung, J. Swaminathan, D.M. Warsinger, J.H. Lienhard V, Multistage vacuum [29] L. Martínez-Díez, M. Vázquez-González, Temperature and concentration polariza-
membrane distillation (MSVMD) systems for high salinity applications, J. Membr. tion in membrane distillation of aqueous salt solutions, J. Membr. Sci. 156 (1999)
Sci. 497 (2016) 128–141, http://dx.doi.org/10.1016/j.memsci.2015.09.009. 265–273, http://dx.doi.org/10.1016/S0376-7388(98)00349-4.
[10] D.M. Warsinger, J. Swaminathan, E. Guillen-Burrieza, H.A. Arafat, J.H. Lienhard V, [30] A.S. Kim, A two-interface transport model with pore-size distribution for predicting
Scaling and fouling in membrane distillation for desalination applications: a review, the performance of direct contact membrane distillation (DCMD), J. Membr. Sci.
Desalination 356 (2015) 294–313, http://dx.doi.org/10.1016/j.desal.2014.06.031. 428 (2013) 410–424, http://dx.doi.org/10.1016/j.memsci.2012.10.054.
[11] C.A. Quist-Jensen, F. Macedonio, C. Conidi, A. Cassano, S. Aljlil, O.A. Alharbi, [31] R.W. Schofield, A.G. Fane, C.J.D. Fell, Heat and mass transfer in membrane dis-
E. Drioli, Direct contact membrane distillation for the concentration of clarified tillation, J. Membr. Sci. 33 (1987) 299–313, http://dx.doi.org/10.1016/S0376-
orange juice, J. Food Eng. 187 (2016) 37–43, http://dx.doi.org/10.1016/j. 7388(00)80287-2.
jfoodeng.2016.04.021. [32] L. Martínez, J.M. Rodríguez-Maroto, Characterization of membrane distillation
[12] M. Gryta, The concentration of geothermal brines with iodine content by membrane modules and analysis of mass flux enhancement by channel spacers, J. Membr. Sci.
distillation, Desalination 325 (2013) 16–24, http://dx.doi.org/10.1016/j.desal. 274 (2006) 123–137, http://dx.doi.org/10.1016/j.memsci.2005.07.045.
2013.06.019. [33] K.W. Lawson, D.R. Lloyd, Membrane distillation. II. Direct contact MD, J. Membr.
[13] C. Hanshik, H. Jeong, K.W. Jeong, S.H. Choi, Improved productivity of the MSF Sci. 120 (1996) 123–133, http://dx.doi.org/10.1016/0376-7388(96)00141-X.
(multi-stage flashing) desalination plant by increasing the TBT (top brine tem- [34] I. Hitsov, T. Maere, K. De Sitter, C. Dotremont, I. Nopens, Modelling approaches in
perature), Energy 107 (2016) 683–692, http://dx.doi.org/10.1016/j.energy.2016. membrane distillation: a critical review, Sep. Purif. Technol. 142 (2015) 48–64,
04.028. http://dx.doi.org/10.1016/j.seppur.2014.12.026.
[14] M.W. Shahzad, M. Burhan, K.C. Ng, Pushing desalination recovery to the maximum [35] A. Alkhudhiri, N. Darwish, N. Hilal, Treatment of high salinity solutions: applica-
limit: membrane and thermal processes integration, Desalination 416 (2017) tion of air gap membrane distillation, Desalination 287 (2012) 55–60, http://dx.
54–64, http://dx.doi.org/10.1016/j.desal.2017.04.024. doi.org/10.1016/j.desal.2011.08.056.
[15] I. Hitsov, L. Eykens, W. De Schepper, K. De Sitter, C. Dotremont, I. Nopens, Full- [36] M. Qtaishat, T. Matsuura, B. Kruczek, M. Khayet, Heat and mass transfer analysis in
scale direct contact membrane distillation (DCMD) model including membrane direct contact membrane distillation, Desalination 219 (2008) 272–292, http://dx.
compaction effects, J. Membr. Sci. 524 (2017) 245–256, http://dx.doi.org/10. doi.org/10.1016/j.desal.2007.05.019.
1016/j.memsci.2016.11.044. [37] A. Alkhudhiri, N. Hilal, Air gap membrane distillation: a detailed study of high
[16] M.H. Sharqawy, J.H. Lienhard, S.M. Zubair, Thermophysical properties of seawater: saline solution, Desalination 403 (2016) 179–186, http://dx.doi.org/10.1016/j.
a review of existing correlations and data, Desalin. Water Treat. 16 (2010) 354–380, desal.2016.07.046.
http://dx.doi.org/10.5004/dwt.2010.1079. [38] J.A. Bush, J. Vanneste, T.Y. Cath, Membrane distillation for concentration of hy-
[17] K.G. Nayar, M.H. Sharqawy, L.D. Banchik, J.H. Lienhard, Thermophysical proper- persaline brines from the great salt Lake: effects of scaling and fouling on perfor-
ties of seawater: a review and new correlations that include pressure dependence, mance, efficiency, and salt rejection, Sep. Purif. Technol. 170 (2016) 78–91, http://
Desalination 390 (2016) 1–24, http://dx.doi.org/10.1016/j.desal.2016.02.024. dx.doi.org/10.1016/j.seppur.2016.06.028.
[18] A. Khalifa, D. Lawal, M. Antar, M. Khayet, Experimental and theoretical in- [39] Y. Lu, J. Chen, Optimal design of multistage membrane distillation systems for
vestigation on water desalination using air gap membrane distillation, Desalination water purification, Ind. Eng. Chem. Res. 50 (2011) 7345–7354.
376 (2015) 94–108, http://dx.doi.org/10.1016/j.desal.2015.08.016.
[19] M.S. Khayet, T. Matsuura, Membrane Distillation: Principles and Applications,

106