Professional Documents
Culture Documents
Desalination
journal homepage: www.elsevier.com/locate/desal
Comparative study of air gap and permeate gap membrane distillation using MARK
internal heat recovery hollow fiber membrane module
Lan Chenga,b, Yajing Zhaoa,b, Pingli Lia,b,⁎, Wenlong Lia,b, Fang Wanga,b
a
Chemical Engineering Research Center, School of Chemical Engineering and Technology, Tianjin University, Tianjin 300350, PR China
b
Tianjin Key Laboratory of Membrane Science and Desalination Technology, Tianjin University, Tianjin 300350, PR China
A R T I C L E I N F O A B S T R A C T
Keywords: Air gap membrane distillation (AGMD) and permeate gap membrane distillation (PGMD) were investigated and
Hollow fiber membrane module compared through self-developed hollow fiber membrane modules with internal energy recovery. The effects of
Air gap membrane distillation the feed temperature, the coolant temperature, the flow rate, the gap thermal conductivity, and the gap width on
Permeate gap membrane distillation the flux and the gained output ratio (GOR) were evaluated experimentally. The gap between the membrane and
Gap thermal conductivity
the condensing surface was a key factor affecting the flux and the GOR in both AGMD and PGMD. It was found
Heat and mass transfer
that the improvement of gap thermal conductivity using brass net was useless in improving flux and GOR. The
permeate water filled in the gap changed the process of mass and heat transfer in the gap which gave rise to the
increase in the flux and thermal recovery. The flux and GOR in PGMD increased by 7.9% and 59.82% in the gap
of 0.5 mm compared to the AGMD. The salt rejection of all experiments was greater than 99.8%.
⁎
Corresponding author.
E-mail address: lipingli@tju.edu.cn (P. Li).
http://dx.doi.org/10.1016/j.desal.2017.10.039
Received 30 June 2017; Received in revised form 17 October 2017; Accepted 23 October 2017
Available online 28 October 2017
0011-9164/ © 2017 Elsevier B.V. All rights reserved.
L. Cheng et al. Desalination 426 (2018) 42–49
with internal heat recovery were self-developed and tested in AGMD Table 1
and PGMD. PP net and brass net between the membrane and the heat Parameters of iPP hollow fiber membrane.
exchange hollow fibers were used to fix the gap width. A brief analysis
Outer Inner Porosity Average Contact Tortuosity
and comparison of the mechanism of mass and heat transfer in AGMD diameter diameter pore size angle (°) factor
and PGMD was carried. Based on it, the effect of the feed temperature, (mm) (mm) (μm)
the coolant temperature, the flow rate, the gap width and the gap
0.66 0.44 68% 0.2 110 2.56
thermal conductivity on the flux and GOR were studied and compared
experimentally under the same operating conditions and geometric
parameters during AGMD and PGMD process. 2.2. Membrane materials
2. Theory analysis and experimental methods Polypropylene (PP) hollow fiber membrane provided by Tianjin
Chemical Separation Technologies Co. Ltd., China, was used. The TIPS-
2.1. Comparison of AGMD and PGMD mechanism iPP hollow fiber membranes had a mean pore size of 0.2 μm with a
porosity of 68% and contact angle of 110°. The inner and outer dia-
2.1.1. Mass transfer meters of the hollow fiber membrane measured by the high resolution
Regardless of concentration polarization, in AGMD, mass transfer is optical microscope were 0.44 mm and 0.66 mm respectively. All these
divided into two sub-sections: mass transport across the porous mem- membrane parameters were shown in Table 1. The iPP heat exchange
brane and mass transport across the gap. hollow fibers with inter diameter of 0.40 mm and outer diameter of
Vapor transport across the membrane can be described by the 0.63 mm were also provided by Tianjin Chemical Separation Technol-
Knudsen-molecular diffusion [15]: ogies Co. Ltd.
1 ΔPm ΔPm
Jmem = × = kmem ×
2.3. Module fabrication
( 1
kk
+
1
kmol ) δm δm
(1)
The structure of membrane module was shown in Fig.1. Firstly, four
2 1 rpε 8RT layer sheets were assembled: net, membrane, net and heat exchange
kk = ⋅ ⋅ ⋅
3 RT τ πM (2) hollow fibers layered sheet in a large flat of table; secondly, the sheets
were sealed at one edge, and were tightly packed where four layer
1 Dε P sheets were wrapped around the edge in a spiral fashion. The net be-
kmol = ⋅ ⋅
RT τ Pml (3) tween the membranes and the exchange heat fibers was used to fix the
where Jmem is the flux across the membrane, kk is the overall Knudsen gap width and change the thermal conductivity in the gap. Thirdly,
diffusion coefficient, kmol is the overall molecular diffusion coefficient, ends of the fiber bundle together with Acrylonitrile Butadiene Styrene
Δ Pm is the trans-membrane partial pressure difference of the diffusing (ABS) plastic tube were sealed with solidified epoxy resin to compose a
molecular, δm is the thickness of the membrane, R and T are gas con- membrane module. Then the hollow fibers were cut open at both ends
stant and absolute temperature respectively. rp, ɛ and τ are the mem- to accommodate the feed channel and the condensation channel.
brane mean pore radius, membrane porosity and tortuosity respec- Four hollow fiber membrane modules with internal energy recovery
tively, M is the molecular weight of the diffusing molecular, D is were assembled and the major parameters of the modules were shown
diffusion coefficient of the species into air in the gap, P is the gas in Table 2. Each membrane module contained 240 hollow fiber mem-
pressure, Pml is the logarithmic mean of air partial pressure between branes and 480 heat exchange hollow fibers. The number ratio of
membrane sides. hollow fiber membranes to heat exchange hollow fibers decided by
Mass transport across the gap which is similar to the molecular previous experiment [16] was 1:2 (the surface ratio of hollow fiber
diffusion Eq. (3) is the main mass transfer resistance: membrane taken by inner surface area to heat exchange hollow fibers
taken by outer surface area was 0.35).
D P ΔPa ΔPa
Ja = ⋅ ⋅ = ka ×
RT Pml δa δa (4)
2.4. Experimental procedure and equation
Δ Pa and δa are the pressure difference between the cold side of the
membrane and the condensation surface, air gap thickness respectively. The schematic diagram of this experimental set-up of AGMD was
The AGMD flux will be: shown in Fig. 2. While in PGMD the module was inverted by collecting
water from the top, hydrostatically forcing the gap region to be flooded
1 1 ⎞ with permeate water as seen in the Fig. 3. And Fig. 3 showed how the
J = ΔP/ ⎛ ⎜ + ⎟
⎝ ka kmem ⎠ (5) flows pass through the module for AGMD and PGMD. During the ex-
periment, the preheating NaCl solution (20 °C, 30 °C, 40 °C) as con-
The PGMD flux Jp differs from AGMD for no air gap resistance: densate water was pumped into the bottom of heat exchange hollow
Jp = Jmem > J. fibers channel by the magnetic pump A for internal heat recovery. Then
the feed from the top of heat exchange hollow fibers flowed into
2.1.2. Heat transfer thermostatic water bath B and was heated up to the pre-set temperature
The heat transfer in the AGMD is generally divided into six parts: (a) (60 °C, 70 °C, 80 °C) and would be pumped into the top of membrane
Heat transfer through the thermal feed side boundary layer by con- channel by the magnetic pump B. The concentrated feed out from the
vection; (b) Heat transfer across the membrane by conduction and la- bottom of membrane channel flowed into the concentrated tank. All
tent heat of evaporation; (c) Heat transfer through the air gap by con- experiments were operated in a counter-current flow.
duction and latent heat of evaporation; (d) Heat transfer through the The flow rate in the system was regulated with the rotameter that
flux by conduction; (e) Heat transfer through the condensation film by has a measuring range of 0–60 L/h. In all experiments, the feed flow
conduction; (f) Heat transfer through the cold side boundary layer by rate was equal to the coolant flow rate. The water produced was col-
convection. lected on a measuring cylinder and recording the value in each 10-
While in PGMD, the process c and d were replaced by heat transfer minute period for three times after the system ran stably. The tem-
through permeate water by latent heat of evaporation and conduction. peratures of inlets and outlets of feed and coolant (T1, T2, T3, and T4)
43
L. Cheng et al. Desalination 426 (2018) 42–49
Table 2 where, Qin is the heat input from the external heat source (J), Δ H is the
The major parameters of the modules (used in both AGMD and PGMD). enthalpy of water evaporation (J/kg), ρ is the density of produced water
(kg/L), Cp is the specific heat of feed (J/kg·°C), and Qf is the volume
Parameter Module 1 Module 2 Module 3 Module 4
(M1) (M2) (M3) (M4) flow of feed (L/h).
44
L. Cheng et al. Desalination 426 (2018) 42–49
Fig. 4. Effect of feed and coolant temperature on the flux for AGMD and PGMD.
Fig. 5. Effect of feed and coolant temperature on the GOR for AGMD and PGMD.
that the feed in the membrane and the permeate water in the gap had increase of the temperature of coolant outlet T2. So, the decrease of the
higher thermal conduction resulting in decrease in the temperature energy needed to heat the feed gave rise to the improvement of the
difference between both sides of the membrane, then leading to the GOR, while for flux, the AGMD with small air gap could obtain a re-
reduced water vapor through the membrane. That is to say, in the lower latively large flux for small mass transfer resistance, resulting in a re-
feed temperature, the temperature difference declined with the increase latively smaller effect from the permeate water on the flux.
of thermal conduction in the PGMD. However, the temperature differ-
ence between the membrane sides had a dominant influence on the
flux. So the flux of PGMD was lower than that of AGMD. Another reason 3.2. Effect of flow rate on flux and GOR
was that there existed a permeate gap between the hollow fiber bundle
and the ABS tube shell, which would cause the effective temperature The effect of flow rate on the flux and GOR during AGMD and
difference decreasing and heat losing. Such phenomenon was more PGMD was tested under the flow rates set at 12 L/h (0.0914 m/s) or
likely to occur in the low feed temperature range (as to say low grade 24 L/h (0.1828 m/s) at different feed temperature T3. During the ex-
heat), which may be similarly explained by the theory that a given periment, the concentration of the feed was 30 g/L (pure NaCl solu-
temperature difference resulted in a larger flux with increasing feed tion), the coolant temperature was 20 °C and M1 was used.
temperature [17]. As shown in Fig. 6, one reason the flux increased was that the
Fig. 5 shows the higher GOR was obtained at a higher feed tem- thickness of the boundary layer reduced with the higher flow rate,
perature for AGMD and PGMD. The PGMD got a higher GOR than the which reduced the temperature polarization on the hot side. Another
AGMD in our testing range whose increasing percentage reason was that the temperature along the flow direction of the feed
(0.66%–59.82%) was obvious when compared to that of flux declined more slowly due to the shorter residence time for the higher
(− 21.08%–7.9%). It could be explained that the heat loss caused by flow rates than the lower flow rates. This kept higher temperature
the smaller gap width in the AGMD was larger [18], and then the en- difference between both sides of membrane. From Fig. 7, the higher
hancement which permeate water in the gap promoted the energy re- flow rates resulted in the decrease of GOR which could also be proved
covery into the coolant was stronger. Such explanation could be con- by the decrease of the temperature of the coolant outlet T2 and the
firmed by the decrease of the temperature of hot feed outlet T4 and the increase of the temperature of the hot feed outlet T4. The short re-
sidence time was the main reason that the heat recovery by the coolant
45
L. Cheng et al. Desalination 426 (2018) 42–49
To investigate the effect of gap width (δg) on the flux and GOR, two
Fig. 6. The effect of flow rate on flux for AGMD and PGMD. sets of comparative experiments (0.5 mm–0 and 0.5 mm–0.25 mm)
were studied at different feed temperature T3 and coolant temperature
T1 through M1 (gap width = 0.5 mm, PP net), M4 (gap width = 0) and
M2 (gap width = 0.5 mm, brass net), M3 (gap width = 0.25 mm, brass
net). The concentration of the feed was 30 g/L; the flow rate was 12 L/
h.
As shown in Fig. 9 and Fig. 10 which was the first comparative
result, it was almost impossible for the permeate water in M4 with gap
width of 0 to promote the flux and the GOR, which can be seen clearly
in Fig. 11. Here came to an important conclusion that the PGMD with
relatively larger gap width would display a larger increasing percentage
in not only the flux but also the GOR when compared to AGMD. Such
conclusion was also confirmed by the second group of modules latter. It
was potentially interpreted that a larger gap width resulted in a higher
mass transfer resistance in AGMD. The appearance of the permeate
water in the gap reduced the mass transfer resistance caused by the air
gap in AGMD. So, the performance difference between PGMD and
AGMD increased with the gap width. As for GOR, the permeate water
filled in the gap changed the process of mass and heat transfer which
promoted the energy into the coolant water. The heat capacity for the
permeate water is higher than the air, leading to a higher adsorption of
Fig. 7. The effect of flow rate on GOR for AGMD and PGMD.
heat.
From Fig. 9 and Fig. 10, it was demonstrated that the gap between
was insufficient in the higher flow rate. the membrane and the condensing surface was a key factor affecting the
From Fig. 8, the flux and GOR of the PGMD were larger than that of flux and the GOR in both AGMD and PGMD. The increase of the gap
the AGMD at any rates in testing temperature except for the case when width from 0 to 0.5 mm led to 1.62%–23.07% increase in the flux and
flow rate = 12 L/h, T1 = 30 °C or 40 °C, T3 = 60 °C. The maximum − 7.63%–33.17% increase in the GOR for AGMD, while
increases in the flux and GOR were 13.25% and 59.82% under 10.72%–32.90% increase in the flux and 62.04%–84.50% increase in
T3 = 70 °C, T1 = 20 °C, flow rate = 24 L/h and T3 = 60 °C, the GOR for PGMD. For gap width = 0, the resistance in the AGMD was
T1 = 20 °C, flow rate = 12 L/h respectively within testing range. disappeared, then the flux should be larger than the gap
Conclusion according to the major points in Fig.8 could be drawn that width = 0.5 mm theoretically. While in fact the flux and GOR with the
flow rate = 24 L/h displayed a higher increasing percentage in flux gap width of 0 were smaller than that of gap width of 0.5 mm for
than flow rate = 12 L/h, while flow rate = 12 L/h displayed a higher AGMD. It could be attributed to that the thermal conduction was con-
increasing percentage in GOR than flow rate = 24 L/h. For the flux, sidered as the main process of heat transfer between the hot feed in the
Fig. 8. The increasing percentage in flux (a) and GOR (b) for
PGMD as AGMD acted as a reference.
46
L. Cheng et al. Desalination 426 (2018) 42–49
Fig. 9. The effect of gap width on flux for AGMD and PGMD at different feed temperature Fig. 12. The effect of gap width (gap = 0.5 mm and 0.25 mm) on flux for AGMD and
T3 at T1 = 20 °C. PGMD at different feed temperature T3 at T1 = 20 °C.
Fig. 10. The effect of gap width on the GOR for AGMD and PGMD at different feed Fig. 13. The effect of gap width (gap = 0.5 mm and 0.25 mm) on the GOR for AGMD and
temperature T3 at T1 = 20 °C. PGMD at different feed temperature T3 at T1 = 20 °C.
hollow fiber membrane and coolant water in the heat exchange hollow between the membrane and the heat exchange hollow fiber did not
fibers, which reduced the temperature difference between both sides of exist for AGMD and PGMD. In either AGMD or PGMD there existed no
membrane leading to lower flux in the gap width of 0. An opposite mass transfer resistance in the gap, so the permeate water just existed
behaviors between the GOR of δg = 0 or 0.5 mm during PGMD and between the hollow fiber bundle and the ABS tube shell in the gap
AGMD mode were observed. Because for δg = 0.5 mm, the permeate width of 0 in the PGMD. The heat loss caused by such permeate water in
water reduced the mass transfer resistance, promoting the increase of the PGMD was larger as compared to the AGMD. So when the δg = 0,
the flux, and the heat capacity for the permeate water was higher than the permeate water has negative effects on both flux and GOR.
the air, leading to a higher adsorption of heat. While for δg = 0, the gap Fig. 12 and Fig. 13 showed the effect of second group gap width on
Fig. 11. The increasing percentage in flux (a) and GOR (b)
for PGMD (gap width = 0, 0.5 mm) compared to AGMD
(gap width = 0, 0.5 mm) at different feed temperature T3
and different coolant temperature T1.
47
L. Cheng et al. Desalination 426 (2018) 42–49
4. Conclusions
48
L. Cheng et al. Desalination 426 (2018) 42–49
experimental studies on full scale spiral wound modules, J. Membr. Sci. 375 (2011) fiber AGMD modules with energy recovery for high saline water desalination,
104–112. Desalination 344 (2014) 55–63.
[15] A. Cipollina, M.G. Di Sparti, A. Tamburini, G. Micale, Development of a membrane [17] G.W. Meindersma, C.M. Guijt, A.B. de Haan, Desalination and water recycling by air
distillation module for solar energy seawater desalination, Chem. Eng. Res. Des. 90 gap membrane distillation, Desalination 187 (2006) 291–301.
(2012) 2101–2121. [18] A. Alklaibi, N. Lior, Transport analysis of air-gap membrane distillation, J. Membr.
[16] H. Geng, Q. He, H. Wu, P. Li, C. Zhang, H. Chang, Experimental study of hollow Sci. 255 (2005) 239–253.
49