evaluation of heat utilization in membrane distillation desalination system.pdf

8/18/2019 Evaluation of heat utilization in membrane distillation desalination system.pdf

1/14

Evaluation of heat utilization in membrane distillation desalination

system integrated with heat recovery

Guoqiang Guan a, Xing Yang b,⁎, Rong Wang c,d,⁎⁎, Anthony G. Fane c,d

a School of Chemistry and Chemical Engineering, South China University of Technology, Guangzhou 510640, PR Chinab Institute for Sustainability and Innovation, College of Engineering and Science, Victoria University, PO Box 14428, Melbourne, Victoria 8001, Australiac Singapore Membrane Technology Centre, Nanyang Technological University, 639798 Singapore, Singapored School of Civil and Environmental Engineering, Nanyang Technological University, 639798 Singapore, Singapore

H I G H L I G H T S

• An implicit expression of GOR was derived to quickly evaluate the heat utilization of desalination system.

• Low equivalent owrates in both sides of hollow-ber membranes are necessary for high GORs.

• High GOR is accompanied by the low water productivity in integrated DCMD system.

• Membranes with large heat resistances promote GOR.

• Non-linearly scale-up effect reveals a higher GOR of industrial DCMD system than lab-scale one.

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

Article history:

Received 20 November 2014

Received in revised form 8 January 2015

Accepted 10 January 2015

Available online 21 January 2015

Keywords:

Desalination

Direct contact membrane distillation

Heat recovery

Gain output ratio

Scale-up effect

Aiming to optimize the system-level heat utilization, a pilot-scale direct contact membrane distillation desalina-

tion system integrated with heat recovery (DCMD–HX) was studied using Aspen Plus. An implicit expression of

gain output ratio (GOR) was derived to reveal the interplay of heat utilization and process parameters including

operating conditions, module specications as well as membrane properties in the DCMD–HX desalination sys-

tem. Compared to operating temperatures, the feed/permeate recirculating owrates were identiedas the mostinuentialoperational factors affecting the GOR.In the current settings, the maximal GOR of 6.0 was observed at

low and equivalent feed- and permeate-side owrates regardless of module specications. Low owrates, how-

ever, resulted in undesirable low water productivity, which was consistent with the trade-off relationship

observed between the heat utilization ef ciency and water recovery rate in MD. Employing membranes with

high heat-transfer resistance (low conductivity and thicker membrane wall) helped to improve the GOR up to

32%. Simulated results also showed that the GOR value increased by 1.3-fold with the preheater parameter

ΔT HX varying from 5 to 0 °C. The non-linear scale-up relationship existed between the membrane area and

heat utilization (i.e., GOR) was also observed, indicating the possible uncertainty in accurately predicting the

GOR value for industrial-scale desalination systems based on lab-scale module testing.

© 2015 Elsevier B.V. All rights reserved.

1. Introduction

Due to the rising fresh water crisis worldwide in recent decades,

desalination technologies have drawn much attention. As a promis-

ing alternative for seawater desalination, membrane distillation

(MD) is operated at mild temperature and ambient pressure [1,2],

in which water vapor generated from the hot brine diffuses through

a hydrophobic porous membrane and condensates by the cold distil-

late stream in direct contact MD (DCMD) mode. Compared to con-

ventional desalination processes such as multi-stage ash distillation

(MSF), multi-effect distillation (MED) or reverse osmosis (RO) [1,3],

MD has many inherent benets: low sensitivity to salinity and high

salt rejection; low vulnerability to membrane fouling and good perfor-

mance under mild operating conditions; feasibility to utilize low-grade

heat and renewable energy (e.g., geothermal heat or solar power) [4,

5]. In recent years, several pilot-scale MD desalination systems have

been developed to utilize solar energy for fresh water supply in arid re-

gions [3,6–9]. Thus, such desalination technology serves dual roles in re-

lieving global water shortage as well as energy crisis and enabling more

and more arid areas/countries to access safe desalted water [10].

Desalination 366 (2015) 80–93

⁎ Correspondence to: X. Yang, Institute for Sustainability and Innovation, College of

Engineering and Science, Victoria University, PO Box 14428, Melbourne, Victoria 8001,

Australia.

⁎⁎ Correspondence to:R. Wang,School of Civil and Environmental Engineering,Nanyang

Technological University, 639798 Singapore, Singapore.

E-mail addresses: [email protected] (X. Yang), [email protected] (R. Wang).

http://dx.doi.org/10.1016/j.desal.2015.01.013

0011-9164/© 2015 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Desalination

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / d e s a l

http://dx.doi.org/10.1016/j.desal.2015.01.013http://dx.doi.org/10.1016/j.desal.2015.01.013http://dx.doi.org/10.1016/j.desal.2015.01.013mailto:[email protected]:[email protected]://dx.doi.org/10.1016/j.desal.2015.01.013http://www.sciencedirect.com/science/journal/00119164http://www.sciencedirect.com/science/journal/00119164http://dx.doi.org/10.1016/j.desal.2015.01.013mailto:[email protected]:[email protected]://dx.doi.org/10.1016/j.desal.2015.01.013http://crossmark.crossref.org/dialog/?doi=10.1016/j.desal.2015.01.013&domain=pdf


2/14

In recentdecades the resurgence of research interest in MD is mainly

attributed to the advancement in polymer material developments and

breakthroughs in membrane fabrication technologies [11–19], and

novel module designs [20–28] as well asow enhancement techniques

to alleviate the temperature polarization phenomenon and enhance

permeationux [29–32]. However, the main challenge for the commer-

cialization of large-scale MD desalination systems still remains due to

the uncertainty in energy requirement. Fortunately, it was widely

reported that MD can be quite economically competitive when low-grade heat such as industry waste heat, geothermal energy or solar

power is available [2]. Nevertheless, even with no waste-heat or limited

thermal energy available, system optimization by incorporating heat re-

covery units can extend the applicability of MD to more rural regions.

Optimal heat recovery is also essential in reducing operational costs.

Yet, thus faronly limited studies are available in the literature on energy

analysis in terms of heat utilization and the interplay of various operat-

ing parameters. Also, no standardized/universal correlations have been

developed to evaluate system-level energy ef ciency in MD [2].

In general, the energy consumption in DCMD systems includes the

necessary thermal energy for heating the feed solution and cooling the

permeate stream, as well as the electricity needed for the pumps and

auxiliary devices. In most lab-scale or pilot plant studies, the MD energy

consumption is evaluated via three thermally related metrics namely

the thermal ef ciency η , gain output ratio (GOR) and water production.

As a common measure of the process ef ciency for thermal desalination

systems such as MD [33], the GOR is associated with useful heat and

reects how well the energy input is utilized for water production in a

system, indicative of the maximum amount of heat recoverable with

certain heat transferred across the membrane. Many attempts have

been made to increase the GOR by incorporating heat recovery devices

[34–36], improved module designs, effective insulation, optimized pip-

ing system and multi-staged operation [37–39]. However, a trade-offre-

lationship is found between the GOR and permeation rate [40,41]

i.e., high GORcould be achieved by designing a system with large mem-

brane area, low ow velocity and more recovery stages; while the ux

decreased due to either the decreased temperature driving force or se-

vere temperature polarization effect. A module-scale thermodynamic

analysis of DCMD modules suggested that high GOR could be achievedat a cost of extremely low water recovery rate in a single-pass DCMD

system [42]. A well-designed MD system is expected to have a GOR

higher than unity. For instance, a cascade of cross-ow hollow ber

MD devices integrated with a heat exchanger was reported to achieve

a GOR as high as 12 at carefully optimized operating conditions [39].

Among a handful of GOR studies in this eld, however, most MD pilot

plants exhibited GOR values far below expectations [43]. To our best

knowledge, only three out of the nine MD systems reported in the liter-

ature were found to have a GOR exceeding 3 while the rest less than

unity [43]. Overall, a wide dispersion on the GOR values from 0.3 to 12

is found in reported MD systems with similar owsheet structures indi-

cating that the prediction of GOR could be effected by various complex

factors such asow conditions, operating temperatures, and even mem-

brane properties. A full factorial analysis on operational factors affectingthe GOR is yet to be comprehensively explored.

To achieve a system-level optimization in a predictive manner, pro-

cess modeling for large-scale MD applications can provide valuable

guidance. However, thus far there are limited process modeling studies

focused on membrane module design to facilitate the overall MD per-

formance and reduce energy consumption [36,40,44–47]. For process

design purposes, owsheet simulation tools such as Aspen Plus have

become more convenient and powerful in revealing the interplay of

key process parameters and system performance to guide practical

applications. Due to the process complexity of combined heat and

mass transfer, the establishment of MD operation units associated

with transport mechanism using Aspen Plus is sparsely reported [48].

Recently, the process development of membrane distillation crystalliza-

tion system for high salinity brine treatment with zero discharge [49]

hasshown thefeasibility of theuser unit operation model forsimulating

the module performance and evaluating process ef ciency in MD brine

process. Later on, further improvement was reported to establish a

more accurate transport model (user customized operation unit in

Aspen plus) in MD modeling incorporated with boundary correction

[48].

With the improved one dimensional (1-D) transport MD model re-

ported in [48], this current work aims to explore a direct contact mem-

brane distillation desalination system integrated with heat recovery(DCMD–HX) for leveraging the advantages of MD practicability in the

context of limited heat resource. An implicit expression of GOR was de-

rived to conveniently correlate the DCMD–HX system ef ciency in

terms of heat utilization with single-unit hollowber module modeling.

A full factorial analysis was conducted to identify the operational factors

that are most inuential in system-level heat utilization in terms of

GOR. Necessary mathematical conditions were proposed for achieving

maximal GOR in a given DCMD–HX desalination system. The newly-

developed implicit GOR correlation was testied through a series of in-

vestigations such as the interplay between GOR and various process

variables (dependent or independent), including owrates, inuent

temperatures of feed and permeate streams, thermal ef ciency of MD

module that is strongly affected by membrane properties, as well as

water recovery rate. Theconcept of “non-linear scale-up” was proposed

for large-scale MD systems integrated with heat recovery in terms of

thermal energy evaluation.

2. Theory and methodology

2.1. DCMD hollow ber module modeling

In this study, an improved 1-D transport model was used to simulate

the heat- and mass-transfer process of DCMD modules [48], in which a

certain number of N hydrophobic PVDF hollow ber membranes with

an effective length of L are regularly packed into a cylindrical housing.

The current transport equations with boundary correction, which

showed higher accuracy in predicting the MD module performance

[48], are summarized in Table 1. In both lumen and shell sides of

DCMD module, the governing equations for mass, momentum andenergy conservation together with the wall correlation equations and

boundary conditions were solved simultaneously. Although this model

is applicable to MD module with either shell or lumen-side feeding

modes, only the latter was investigated in this study. Also, in this

model both the effects of feed concentration on the change of vapor

pressure and concentration polarization are considered negligible [50].

Thecurrent transport model hasbeen veried previously [48], based

on an established DCMD system for a series of experimental settings,in-

cluding various feed inlet temperatures, ber lengths and ow veloci-

ties. Also, the membrane properties were the same as that in previous

verication experiments. Hence, the model verication was not repeat-

ed here and the veried MD model was used as a customized unit for

Aspen owsheet simulation in the following sections.

2.2. DCMD–HX desalination system

In this simulation study, an ideal heat exchanger (HX), in which the

heat transfer takes place through innitely large area and hence is not

limited by heat exchanging kinetics [42], was used as the heat recovery

unit and integrated into the DCMD desalination system to recover heat

from the returning permeate stream, namely DCMD–HX. Therecovered

heat could be utilized to preheat thebrine feed inuent before entering

the membrane module.

A series of pilot-scale hollowber modules were integrated into the

MD owsheet in Aspen Plus. The rst set of module specications is

given in Table 2, while three pilot-scale hollow ber modules with

various packing densities and ber lengths were used in the owsheet

simulations to correlate module performance with the GOR in the

81G. Guan et al. / Desalination 366 (2015) 80–93


3/14

DCMD–HX system. Packed with polyvinylideneuoride (PVDF) hollow

bers tested in previous experiments [27], with inner diameter Di,1 of

0.98 mm and averaged wall thickness of δ of 0.24 mm, module #1

was used as the benchmark for comparison (N = 15,089, L = 2.153 m,

polypropylene housing Di,2 = 0.216 m). The PVDF hollow ber mem-

brane used in this series of simulations has an averaged thermal con-

ductivity (kM ) and MD coef cient (C M ) of 0.066 kJ m−1 K−1 and

3.8 × 10−7 kg m−2 s−1 Pa−1 [48], respectively.

Using the same hollowber membrane, another set of modules was

designed to investigate the scale-up effect of the DCMD–HX desalina-

tion system. Ten DCMD modules with scale-up factor from 1 to 10 and

membrane areas varying from 0.044 to 44.33 m2 at constant packing

density and L/ Di,2 were simulated. In a word, these modules were

designed to remain both hydrodynamic (i.e., Reynolds number, Re)

and geometric (i.e., L/ Di,2) similarities. The module specications and

corresponding process parameters for each case are listed in Table 3.

Considering the very low single-pass water recovery rate in MD,

which is usually below 6.4% [42], in this study the feed ef uent wasrecycled continuously for further treatment to reduce the volume of

brine discharge. The recirculation of process streams can also help to

alleviate polarization effects [50] in MD. Both the DCMD module and

HX were operated in counter-current ow mode to maximize the

mass- and heat-transfer ef ciency.

The conceptual schematic of the DCMD–HX system designed for

seawater desalination is presented in Fig. 1, in which the red and blue

lines represent the feed and permeate cycles, respectively. The fresh

feedstock (3% w/w NaCl solution as synthetic seawater, T 1,0 = 25 °C)

joins the returning brine stream (feed ef uent) pumped back from the

DCMD module and forms a new stream S 1,0 entering the preheater

(HX), in which S 1,0 recovers heat from the returning permeate stream

S 2,2 and becomes stream S 1,1 with elevated temperature. The feed

stream S 1,1 then ows through a heater and becomes the inlet streamS 1,2 of the MD module with a specied inlet temperature T 1,2. In the

MD module, water vapor is generated from the hot feed driven by tem-

perature difference between the feed and permeate, and transports

through the membrane wall and condenses at the cold permeate side.

As a result, heat and mass transfer in the DCMD module take place be-

tween hot feed stream S 1,2 and cold permeate stream S 2,1. Subsequently

the feed temperature decreases from T 1,2 (stream S 1,2) to T 1,3 (stream

S 1,3) and the permeate temperature rises from T 2,1 (stream S 2,1) to T 2,2(stream S 2,2) alongthe module length. The heat gained by the permeate

stream is further utilized through the heat recovery unit HX to pre-

heat thefeedstock. In this DCMD–HX process most feed concentrates

(ef uent) arerecycled continuouslyat the hot side with fresh feedstock

to maintain a given owrate W 1,2; while minimal brine is discharged to

reduce environmental impacts. Similarly, at the cold side of the DCMD

module, the permeate (distillate water) is recycled to maintain a xed

permeate owrate W 2,1 with continuous production of distillate. It is

assumed that no wetting occurs during operation.

2.3. Evaluation of heat utilization in MD

2.3.1. Thermal ef ciency η of DCMD module

In the DCMD–HX system, the heat transfer occurs only in the DCMD

module and HX unit respectively. The knowledge of heat transfer in the

heat exchanger has been well established to study the heat recovery in

the HX [51]. In MD, the vapor pressure difference between the feeding

and permeating sides drives the vapor to transfer across the membrane.

Theoverall heat ux q includinglatentheat (qv) of evaporation and con-

duction heat (qc ) is accompanied with mass transfer [48]. The latent

heat is considered as the effective heat used for MD water production;

while the conductive heat through the membrane matrix caused by

transmembrane temperature difference is taken as heat loss in MD.

In DCMD, the thermal ef ciency η , whichis dened astheratioof la-

tent heat to the totalheatinput, is widely used to evaluate the effective-ness of heat utilization associated with water production [35]. Hence,

the universal expression of η is given as:

η ≡ qv

qv þ qc ¼

J M Δhv

J M Δhv þ κ M δ

T 1−T 2ð Þτ ¼

C M dp

dT Δhv

C M dp

dT Δhv þ

κ M δ

ð1Þ

where J M is the permeationux indicatingMD performance, kg/(m2 h);

Δhv is the specic latent heat of evaporation, kJ kg−1; operating param-

eters such as (T 1− T 2) is the bulk temperature difference between the

feed and permeate, and τ is dened as temperature polarization coef -

cient, which characterizes the actual transmembrane driving force of

theheat- andmass-transfer andis strongly inuenced byow conditions

Table 1

Equations of 1-D transport model with boundary correlation for hollow ber DCMD module [48].

Lumen side Shell side

Mass d ρ1v1ð Þ

dz ¼

4

Di;1 J M

d ρ2v2ð Þ

dz ¼ −

4N Do;1

D2i;2 J M

Momentum d

dz p1−2 μ 1

dv1dz

þ ρ1v21

¼ 0

d

dz p2−2 μ

dv2dz

þ ρv22

¼ 0

Energy ρ1v1c p;1

dT 1dz

þ v1dp1dz

¼ 4

Di;1 J H ;1 ρ2 v2c p;2

dT 2dz

þ v2dp2dz

¼ −4N Do;1

D2

i;2

J H ;2

B.C. v1 z ¼0 ¼ v1;0

p1 z ¼0 ¼ p1;0

T 1 z ¼0 ¼ T 1;0

v2 z ¼L ¼ v2;0

p2 z ¼L ¼ p2;0

T 2 z ¼L ¼ T 2;0

Heat transfer coef cientNu ¼ 1:86Gz 1=3

μ

μ w

0:14Nu ¼ 0:4Re1=2 þ 0:06Re2=3

Pr 0:4

μ

μ w

1=4Transmembrane heat ux J H ,1 = J M Δhv|T W ,1 + J H ,c J H ,2 = J M Δhv|T W ,2 + J H ,c Film heat ux J H ,1 = h1(T 1− T W ,1) J H ,2 = h2(T W ,2− T W ,2)

Conductive heat ux J H ;c ¼ κ M δ

T W ;1−T W ;2

Permeation ux J M ¼ C M pW ;1 T W ;1− pW ;2

T W ;2

Table 2

Specications for DCMD modules used in Aspen Plus owsheet simulations (membrane

properties: D i,1 = 0.98 mm, δ = 0.24 mm, kM = 0.066 kJ m−1 K−1 and C M = 3.8 ×

10−7 kg m−2·s−1·Pa−1).

Module

#1 #2 #3

Membrane area, Am m2 100.0 100.0 50.2

Packing density, φ 0.689 0.502 0.689

Housing Inner diameter. Di,2 mm 216 253 216

Fiber length, L mm 2153 2153 1080

Ratio of module length to housing

diameter, L/ Di,2

10.0 8.5 5.0

82 G. Guan et al. / Desalination 366 (2015) 80–93


4/14

(e.g., Reynolds number, Re) [48], the kM andδ are respectively the mem-

brane conductivity and thickness, and the C M is the membrane distilla-

tion coef cient. In this current study, the averaged η was determined

and the averaged temperature was used to calculate d p/dT and latent

heat in Eq. (1). When the MD system is operated at constant inuent

temperatures, the η mainly depends on both the MD coef cient (C M )

and characteristics (i.e., kM and δ) instead of ow conditions.

In an ideal situation, the DCMD module is treated as an adiabatic

system. The latent heat required for evaporation is provided through

the enthalpy change of the feed. As depicted in Fig. 2, a suf ciently

thin element of the cross section of the DCMD module can be used to

correlate the stream property changes associated with transmembrane

water production.

The energy balance through the element can be written as:

W 1 z ð Þh1 z ð Þ ¼ W 1 z þ Δ z ð Þh1 z þ Δ z ð Þ þ J M hvπ Di;1Δ z þ qc π Di;1Δ z : ð2Þ

When the limit taken as Δ z approaches zero, Eq. (2) can be simpli-

ed and rearranged as:

−d W 1h1ð Þ ¼ J M hv þ qc ð Þπ Di;1dz : ð3Þ

The mass balance through the element gives:

−dW 1 ¼ J M π Di;1dz : ð4Þ

Also, the specic enthalpy of vapor can be expressed as the sum of

the specic enthalpy of liquid (hl) and latent heat (Δhv):

hv ¼ hl þ Δhv: ð5Þ

Substituting Eqs. (4) and (5) into Eq. (3) gives

−W 1dh1 ¼ J M Δhv þ qc ð Þπ Di;1dz : ð6Þ

Assuming with constant density, the enthalpy change of a uid can

be derived based on the thermodynamic relation as:

dh ¼ c P dT þ1

ρ−T

∂vm∂T

dp ¼ c P dT þ

dp

ρ : ð7Þ

Substituting Eqs. (1) and (7) into Eq. (6) when the pressure drop is

negligible, the thermal ef ciency in the DCMD module given in Fig. 1,

can be expressed in terms of feed-side temperature change as:

η ¼ W P Δhv

W 1;2c P ;1 T 12−T 13ð Þ ¼

2369W P W 1;2c P ;1 T 12−T 13ð Þ

ð8Þ

where latentheat of 2369 kJ/kg [51] at theaveraged temperature of feed

and permeate (55 °C) is used in this work.

2.3.2. Calculation of gain output ratio (GOR) in DCMD–HX process

simulations

As one of the most useful measures, the GOR is often used to evalu-

ate the MD performance in terms of the specic energy required for per

kg distillate output. The benets of the DCMD–HX system is to possiblyrecover the thermal energy from the returning permeate stream for

raising the specic enthalpy of the feed, which is the combined stream

of fresh feedstock and brine reux, and hence signicantly reducing

Table 3

Module specications and operating owrates for scale-up effect study of DCMD–HX system (T 1,2 = 80 °C and T 2,1 = 30 °C, membrane properties: Di,1 = 0.98 mm, δ = 0.24 mm, kM =

0.066 kJ m−1·K−1 and C M = 3.8 × 10−7 kg m−2·s−1·Pa−1).

Module

#4 #5 #6 #7 #8 #9 #10 #11 #12 #13

Scale-up factor 1 2 3 4 5 6 7 8 9 10

Number of bers 72 288 648 1152 1800 2592 3528 4608 5832 7200

Module shell diameter mm 15 30 45 60 75 90 105 120 135 150

Module Length mm 200 400 600 800 1000 1200 1400 1600 1800 2000Lumen-side owrate kg/h 10 40 90 160 250 360 490 640 810 1000

Shell-side owrate kg/h 10 40 90 160 250 360 490 640 810 1000

Membrane area m2 0.044 0.355 1.197 2.837 5.542 9.576 15.21 22.70 32.32 44.33

Fig. 1. Schematic diagram of direct contact membrane distillation desalination system

with heat recovery unit.

Fig. 2. Heat and mass prolesacross a suf ciently thin cross-sectional elementof a DCMD

module.



5/14

the total heat input to the system. Subsequently, the process ef ciency

is greatlyimproved. TheGOR in theDCMD–HXsystem (Fig.1)isdened

as the ratio of latent heat for evaporation associated with water produc-

tion W P to total heat input to the heater:

GOR ¼ W P Δhv

W 1;2 h1;2−h1;1

¼ 2369W P W 1;2c P ;1 T 1;2−T 1;1

ð9Þ

Also, Eq. (9) can be rewritten and rearranged by multiplying three

dimensionless groups:

GOR ¼ 4L

Di;1

! Δhv

c P ;1 T 1;2−T 1;1

24

35 J M

G1

ð10Þ

where the rst term (4L/ Di,1) indicates the geometric characteristics of

an MD module; the second term is related to the latent heat and system

heat input, namely relative heat input. Correlated with the mass

owrateper unit cross-sectionalarea of thestream in theber channels,

namely mass rate G1 = 4 W 1,2 / (N π Di,12 ), theratio J M / G1 is often referred

to as the water recovery rate γ

, which is a key metric for desalination

systems [42]. Based on Eq. (10), the magnitudes of each term can be es-

timated forperforming a factorialanalysisin Section 2.3.4 to identify the

most inuential parameters affecting the heat utilization in the DCMD–

HX system. Commonly in hollow ber DCMD modules, the module

geometric parameter 4L/ Di,1 has a magnitude of 10+3. The relative

heat input in the DCMD–HX system is in the magnitude range of

10+1–10+2. For most of MD membrane, the magnitude of permeation

ux is in the range of 10−1–10+1 kg m−2 h−1. Thus, the mass rate G1 is

estimated to be from 10+3 to 10+6 kg m−2 h−1 when taking GOR as 1.

2.3.3. Correlation of GOR and thermal ef ciency (Implicit expression of

GOR)

Since GOR and thermal ef ciency η are both related to the effective

latent heat of evaporation, a correlation of GOR and η can be derived

for a system-level evaluation of MD. In an integrated DCMD–HX system,

the performance of DCMD module and HX is determined by the bulk

temperature difference of the MD operation (ΔT MD) and the tempera-

ture difference in the HX (ΔT HX), respectively. Based on the process

schematic in Fig. 1, the ΔT MD is expressed in terms of inlet temperature

difference of the feed and permeate and is xed at 50 °C in this study:

ΔT MD ¼ T 1;2−T 2;1 ð11Þ

And the ΔT HX, reects the extent of heat recovery in the HX unit and

is also considered as an input in the given system:

ΔT HX ¼ T 2;2−T 1;1: ð12Þ

For an ideal heat recovery unit, an innite heat-exchanging area re-

sults in a complete recovery of sensible heat from the permeate stream.

Two scenarios were considered: 1) when the feed owrate (W 1,0) isnot

greater than that of the permeate stream (W 2,2), the temperature of the

cold-side ef uent approaches the hot-side inuent and the HICO

mode (conguration for simulating heat exchanger in Aspen Plus

[52]) was used to simulate the HX unit in the owsheet shown in

Fig. 1; 2) when W 1,0 N W 2,2, the temperature of the hot-side ef uent

approaches the cold-side inuent in theHX unit, namely HOCI mode.

This study mainly focuses on the HICO mode for heat exchanger sim-

ulation (W 1,0 = W 2,2) and hence assumes a constant ΔT HX = 0 in the

following discussions, except the investigation of GOR vs. ΔT HX in

Section 3.6.

Substituting Eqs. (11) and (12) into Eq. (9) yields:

GOR ¼ W P Δhv

W 1;2c P ;1 ΔT MD þ ΔT HX þ T 2;1−T 2;2

ð13Þ

where W P / W 1,2 can also be written as J M / G1, i.e., water recovery rate γ .

Similar to the derivation of Eq. (8), the thermal ef ciency can be

expressed through the permeate-side temperature change as:

η ¼ W P Δhv

W 2;1c P ;2 T 2;1−T 2;2

¼ 2369W P W 2;1c P ;2 T 2;1−T 2;2

: ð14Þ

Substituting Eq. (14) into Eq. (13), yields an implicit expression of

GOR comprising of dimensionless groups as:

Δhvc P ;1 ΔT MD þ ΔT HXð Þ

W P W 1;2

1

GOR þ

Δhvc P ;2 ΔT MD þ ΔT HX ð Þ

W pW 2;1

1

η ¼ 1: ð15Þ

The above implicit equation clearly shows a general relationship

between the GOR and thermal ef ciency η in the DCMD–HX system, it

has an advantage to conveniently correlate the module performance

with the desalination system ef ciency when heat recovery capacity

ΔT HX is provided. Compared to Eq. (9), which is used to calculate

the GOR upon the acquisition of the complete set of simulation results

of a DCMD–HX system, Eq. (15) requires only the outputs of the

DCMD module to predict the GOR of the whole system. Thus, time-

consuming and complicated process simulations can be avoided.

2.3.4. Factorial analysis of operational factors affecting GOR

In the DCMD–HX system, it is essential to analyze the inuence of

operating conditions on the heat utilization in terms of GOR. As

shown in Fig. 1, for a specied feedstock with xed uid properties,

four key operational factors, i.e., inlet temperatures (T 1,2 and T 2,1) and

recirculating owrates (W 1,2 and W 2,1) through both sides of the

DCMD module, are closely related to the module performance and

hence GOR values, as discussed in Eq. (15). To screen the statistically

signicant factors affecting GOR in DCMD–HX, commercial softwareMinitab® 16 was used to conduct the factorial analysis.

According to the analysis in Section 2.3.2, inthis study the massrate G

of feed and permeate streams is varied from 10+3 to 10+6 kg m−2 h−1

andhence the magnitudeof the owrates W 1,2and W 2,1can be calculated

as 10+1–10+4 kg m−2 h−1. The upper temperature limit for feed inlet is

set as 80 °C and thelower temperature limit forcold permeate inlet T 2,1is set as 30 °C allowing minimal refrigeration requirements and low in-

vestment cost. The four factors including the mass owrates and inlet

temperatures i.e., W 1,2, W 2,1, T 1,2, and T 2,1, are denoted as A, B, C, and

D, respectively and the corresponding low and high levels for each fac-

tor are indicated in Table 4.

As shown in Table 4, this full factorial design includes 2 × 2^4 sets of

combinations. Each combination was used as the inputs for DCMD–HX

owsheet simulations, which will be analyzed in Section 2.4 to identifytheir impacts on the GOR.

2.4. Flowsheet simulation of DCMD–HX system in Aspen Plus

2.4.1. User unit operation model for DCMD module

With the 1-D transport equations presented in Table 1, a user unit

operation model coded in software Intel© Visual FORTRAN v11.1 was

developed to simulate the heat and mass transfer in the DCMD process.

The module dimensions and membrane properties were specied as

process parameters(simulation inputs), the physicochemicalproperties

of the uids (feed/permeate) were assigned into the interface routines,

as well as the temperature differences for heat exchangers were set as

design parameters. Thesolved proles of uid temperature, permeation

ux, pressure and owrate served as the unit outputs for module



6/14

evaluation. The detailed computational algorithm for MD module

modeling can be found in the literature [48].

2.4.2. DCMD–HX owsheet simulations

In the current studythe ow chart of an integrated DCMD–HX desa-

lination system was established to recover fresh water with heat regen-

eration using the commercial software Aspen Plus (Version 7.3). Based

on the conceptual schematic in Fig. 1, the detailed owsheet consisting

of interconnecting material streams and unit blocks is developed and

presented in Appendix A.1. Each material stream and unit block was

named consistently as that in Fig. 1. Except the user-dened DCMD

unit described in Section 2.4.1, all process units including heat

exchangers, pumps and splitters are built-in models in Aspen Plus.

The feed/permeate conditions and operation details were consistentwith that of the conceptual DCMD–HX described in Fig. 1, Section 2.2.

DCMD modules presented in Tables 2 and 3 were employed in the

owsheet simulations for investigating various aspects. An example of

initial input settings for owsheet simulation of the DCMD–HX system

using module #1 is provided in Appendix A.1, Table A.1.

3. Results and discussion

3.1. Comparison of correlated and simulated GOR

As analyzed in Section 2.3, there are two approaches to obtain the

GOR values of a DCMD desalination system with heat recovery. Based

on the implicit expression of GOR, Eq. (15), single-unit simulations of

the user-dened DCMD module can be conducted to convenientlycorrelate the DCMD–HX system ef ciency with module performance

by specifying operating conditions, i.e., recirculating owrates W 1,2 &

W 2,1, xed uid properties Δhv & c P , and heat exchanger settings

ΔT MD = 50 °C & ΔT HX = 0 °C. Thus the simulation outputs were used

to correlate the GOR using Eq. (15). Alternatively, DCMD–HX system

simulations (Fig. A.1 in Appendix A.1) are performed to obtain the

GOR using the process-related expression Eq. (9), as the simulation

outputs including heat inputs and water productivity are accessible

variables in Aspen. The difference of GOR values obtained from these

two approaches is that the former is only based on DCMD module sim-

ulations; while the latter relies on comprehensive owsheet simula-

tions of the DCMD–HX system. The results are compared in Table 5.

Clearly, the module correlated GOR values by Eq. (15) agree well

with the simulation results obtained by Eq. (9) with a small relative

error less than 0.17%. Therefore, using Eq. (15), the GOR of the DCMD–

HX system can be accurately predicted based on the module perfor-

mance under similar outputs such as water productivity W P and

thermal ef ciency η at specied uid properties and heat exchanger

settings (ΔT MD & ΔT HX). Thus, it is possible to evaluate the heat utiliza-

tion of such system in a simpler manner to avoid performing time-

consuming owsheet simulations.

3.2. Effects of DCMD operating conditions in DCMD–HX system

According to Eq. (15), the GOR in the DCMD–HX system is closely

related to the operating conditions and module performance, which

largely depends on membrane characteristics and operating parame-

ters. This section focuses on the factorial analysis to investigate the

effects of four operational factors on the heat utilization in DCMD–HX

with a constant ΔT HX of 0 °C.

3.2.1. Factorial analysis of factors affecting GOR

As discussed previously, in the DCMD–HX system four operational

factors, i.e., T 1,2 and T 2,1, and W 1,2 and W 2,1, are the key variables affect-

ing the total heat input and distillate output. Based on the GOR valuesobtained at varying operating conditions (Table 4), factorial screening

analysis was conducted to identify the most signicant factors affecting

the GOR using statistical software Minitab®.

The inuence of the four factors and their interactions (combina-

tions) are illustrated in the Pareto diagram shown as Fig. 3, in which

the impact of each factor and interactions of factors is illustrated as

Table 5

Comparison ofthe GORs correlatedby Eqs. (9)and (15) (HX unitand module temperature

differences ΔT HX = 0 °C and ΔT MD = 50 °C, i.e., T 1,2 = 80 °C & T 2,1 = 30 °C, simulated

module #1).

W 1,1 = W 2,1 GOR Relative error

( kg/h) Co rr elat ed in Eq. (15) Correlated in Eq. (9)

10 5.857 5.854 0.05%

20 5.637 5.635 0.04%

50 5.059 5.064 −0.11%

100 6.183 6.169 0.23%

200 4.481 4.500 −0.43%

500 3.520 3.521 −0.01%

1000 2.943 2.941 0.09%

2000 2.029 2.028 0.03%

5000 1.206 1.205 0.06%

10000 0.775 0.773 0.18%

RMS 0.17%

CD

BCD

ABCD

ACD

C

ABC

BC

AC

BDABD

AD

D

B

AB

A

0 1 2 3 4

Absolute effect

T e r m s

Factor Name

A W 1,2

B W 2,1

C T 1,2

D T 2,1

1.519

Fig. 3. Pareto diagram of full factorial analysis for factors affecting GOR in DCMD –HX

system.

Table 4

Full factorial analysis for screening operation factors affecting GOR in DCMD–HX system.

Notation Factor Low level High level

A W 1,2 kg/h 10 10000

B W 2,1 kg/h 10 10000

C T 1,2 °C 60 80

D T 2,1 °C 30 50

Run # A B C D GOR

1 10 10 60 30 6.598

2 10000 10 60 30 0.341

3 10 10000 60 30 0.708

4 10000 10000 60 30 0.659

5 10 10 80 30 5.856

6 10000 10 80 30 0.411

7 10 10000 80 30 0.726

8 10000 10000 80 30 0.775

9 10 10 60 50 10.19

10 10000 10 60 50 0.189

11 10 10000 60 50 1.042

12 10000 10000 60 50 0.838

13 10 10 80 50 7.817

14 10000 10 80 50 0.276

15 10 10000 80 50 1.057

16 10000 10000 80 50 0.960



7/14

horizontal bars. The length of each bar is proportional to the effect of in-

dividual factor on GOR divided by its standard deviation. The bars are

also sorted by their standardized effects. Therefore, factors with the

most signicant effect on GOR can be identied. In addition, on the

Pareto chart a vertical line at 1.519 serves as a critical point to identify

factors exhibiting strong dependence (signicant effects) on GOR,

i.e., any factors with bars over the line showedstatistically signicant in-

uence within a condence level of 95% [19].

As shown in Fig. 3, factor A (W 1,2) is found tobe the mostin

uentialparameteraffecting GOR, followed by factorB (W 2,1) and then AB;while

the combined factor of C (T 1,2) & D (T 2,1) has the least effect on GOR.

Among allfactors, only therecirculatingowratesA and B, and thecom-

bined AB show an absolute effect exceeding the critical line of 1.519.

Other than the ow parameters W 1,2 and W 2,1, the GOR values seem

to be statistically irrelevant to the operating temperatures T 1,2 (factor

C) and T 2,1 (factorD), which have been conrmed through the statistical

analysis in Fig. 3 provided their lower absolute effect bars than the

vertical line. This can also be explained through the GOR correlation in

Eq. (15) where a simultaneous change of ΔT MD related terms in both

denominator and numerator (i.e., W P ∝ΔT MD) that weakens the impact

of temperature on GOR.

Similarly,the signicant effect of the feed-siderecirculatingowrate

on the GOR can be explained through Eq. (15), in which the ratio of

water productivity to recirculating owrate W P / W 1,2 determines the

GOR. Dened as water recovery rate γ , the ratio J M / G1 is proportional

to W P / W 1,2 and is critical for designing a DCMD module. Thus, the

relationship of J M and G1 was rst studied to verify the effect of the

recirculating owrate on the GOR. The simulation results are shown in

Fig. 4. Clearly, Fig. 4(A) shows an increasing trend of J M with increasing

G1. This is due to the increase of MD driving force with reduced bound-

ary layer thickness and hence lower heat-transfer resistance at higher

owrates. Subsequently, the mass/heat transfer across the membrane is

greatly enhanced. However, the rise of permeation ux J M is not propor-

tional to the increase of mass rate G1. To further investigate the reasoning,

another graph of water recovery rate vs. G1 isgivenin Fig. 4(B),in which an

initial steep decline of J M / G1 at low G1 (b103 kg m−2 h−1) and then a

slow decrease is observed until G1 reaches 107 kg m−2 h−1. This is be-

cause of the extremelylow permeate rate J M associated with high trans-membrane resistance at low G1; as G1 further increases, the improved

transmembrane mass and heat transfer promotes a signicant increase

of J M leading to a mild decreasing trend of water recovery rate J M / G1.

This has veried the strong impact of J M / G1 on the GOR in the DCMD–

HX system.

3.2.2. Necessary conditions for achieving maximal GOR

As the recirculating owrates have signicant effects on GOR in the

DCMD–HX system, the GOR was quantied at simultaneously varied

feed- and permeate-side recirculating owrates (W 1,2 and W 2,1) at

constant heat exchanger settings ΔT HX = 0 °C and ΔT MD = 50 °C with

T 1,2 = 80 °C & T 2,1 = 30 °C. Fig. 5 shows the simulation results of GOR

as colored contour in terms of the feed and permeate recirculating

owrates (W 1,2 and W 2,1) ranging from 101 to 104 kg h−1 in a DCMD–

HX system employing module #1, i.e., a warmer color indicates a higher

GOR value, for instance, orange and red colors.

Apparently, the GOR of a DCMD–HX system varies signicantly at

varying operatingowrates. With a maximum of 6.0 achieved at a com-

bination of equally low W 1,2 and W 2,1, the GOR generally decreases with

increasing owrates at either feed or permeate side — extremely low

GOR less than 1 is obtained at a combination of low W 1,2 and high

W 2,1, or vice versa, indicating poor system performance. Interestingly,

it is observed that a ridge of warmer color regions, indicative of high

GOR values, is located along the diagonal of Fig. 5. This reveals that bet-

ter energyutilization of theDCMD–HX systemcan be achieved at equiv-

alent feed- and permeate-side recirculating owrates W 1,2 and W 2,1,

preferably in the lower owrate range. This is similar to the ndings

in the literature [42,49] that in DCMD the owrate of feeding inuentneeds to match with the permeate side for achieving higherpermeation

ux and hence better module performance. In this study this phenome-

noncan be conveniently explainedvia Eq.(15) — assuming insignicant

-5

0

5

10

15

20

25

30

35

(B)

J M ( k g m

- 2 h - 1 )

(A)

0.0 5.0x106

1.0x107

1.5x107

0

1

2

3

4

5

( J M

/ G 1

) x 1 0 5

G1 (kg m

-2 h

-1)

Fig. 4. Effect of feed-side mass rate G1 on the (A) permeation ux J M ; (B) water recovery

rate J M / G1 in DCMD–HX system (simulated module #1, W 1,2 = W 2,1, T 1,2 = 80 °C and

T 2,1 = 30 °C).

1.0 1.5 2.0 2.5 3.0 3.5 4.0

1.0

1.5

2.0

2.5

3.0

3.5

4.0

GOR for module 1

log W 1,2

/(kg h-1)

l o g W

2 , 1 /

( k g h - 1 )

0.000

0.7325

1.465

2.197

2.930

3.662

4.395

5.127

5.860

Fig. 5. The GOR colored contours at varied recirculating owrates (log scales) in DCMD–

HX system (simulated module #1, ΔT MD = 50 °C with T 1,2 = 80 °C and T 2,1 = 30 °C;

HICO mode (ΔTHX = 0 °C) used for HX simulations of owrate range W 1,2/ W 2,1 ≤ 1

(top-left contour) and HOCI mode for W 1,2/ W 2,1 N

1 (bottom-right contour)).



8/14

change of thermal ef ciency for varying owrates as shown in Eq. (1),

the increase of either W 1,2 or W 2,1 will lead to the decrease of W P / W 1,2or W P / W 2,1 in the rst or second terms on the left-hand side of the equa-

tion and hence a reduced GOR. In the case of equal W 1,2 and W 2,1, the

decrease of owrates results in the increase of W P / W 1,2 (= W P / W 2,1)

and eventually signicant improvement of GOR in Eq. (15) to bal-

ance the equation. Therefore, equivalent feed- and permeate-side

owrates in the low range (b101.5 kg h−1) in the DCMD–HX system

are necessary for achieving high GOR values (N

5.0). This has further ver-ied the strong impact of operating owrates on GOR in the DCMD–HX

process.

To generalize the effects of owrates, similar simulations were

conducted to analyzeGORs in DCMD–HX with three MD modules of dif-

ferent congurationspacked with thesame numberof bers N = 15089

listed in Table 2. Benchmarked against module #1 (φ = 0.689,

L = 2153 mm), module #2 has a lower packing density of 0.502 and

module #3 has a shorter membrane length of 1080 mm. Fig. 6 shows

the GOR as a function of owrate ratio W 1,2/ W 2,1, which would further

highlight the inuence of equivalent owrates on heat utilization in

different DCMD–HX systems. It is noted that W 2,1 was adjusted from

10 to 1000 kg/h at a xed W 1,2 of 50 kg h−1.

In Fig. 6 a single GOR peak is observed for all DCMD–HX systems

operated at equivalent owrates, i.e., W 1,2/ W 2,1 = 1, regardless of the

variations in module geometries (e.g., housing size, ber length, and

packing density). Compared to module #1 at the same membrane

area and packing density (Fig. 5), the system with module #2 exhibits

a slightly lower peak of GOR. This is due to its lower packing density

with larger housing diameter Di,2 and hence a lower water production

W P is anticipated because of the deterioration of shell-side (feed) heat

transfer and lower permeation ux J M (W P = J M AM ) at a the same

owrate but lower Reynolds number, Re ( 4W 2;1

πμ ffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffi

D2i;2−nD2o;1ð Þ

p ). Subsequently,a lower GORvalueis obtained based on Eq. (15). Similar explanation can

be used for the slightly lower curve of the system with module #3, in

which a shorter membrane length leads to a greater permeation ux

but slightly lower water productivity (W P ) due to the greatly reduced

membrane area to 50 m2. Nevertheless, the GOR peaks of DCMD–HX

systems with modules #1 to #3 are still considered quite similar inshapes and heights.

Based on the above discussions for Figs. 5 and 6, the unity of owrate

ratio (W 1,2 = W 2,1) seems to be necessary for achieving maximum

GORs, regardless of module specications. Compared to the previously

reported theoretical value of 0.918 by Lin et al. [42], the slightly higher

owrate ratio obtained here is mainly due to theassumption of negligi-

ble inuence of concentration on vapor pressure (Section 2.1). In addi-

tion, recirculating owrates in the lower range (b101.5 kg/h) are

preferred. In summary, low and equivalent feed- and permeate-side

owrates are considered as the necessary conditions for optimal heat

recovery ef ciency (GOR) in the DCMD–HX system. Besides the operat-

ing

ow and temperature conditions, the GOR in the DCMD–

HX processalso greatly depends on the module performance and membrane char-

acteristics, which are often evaluated in terms of thermal ef ciency ( η )

and water productivity (W P ). Thus the relationship of GOR and thermal

ef ciency is further explored in the following sections.

3.3. Relationship of GOR and thermal ef ciency η

3.3.1. Effects of membrane characteristics on GOR

The inevitable conductive heat loss plays an important role in deter-

mining the heat-transfer resistance and thermal ef ciency of the MD

process. As dened in Eq. (1) as kM (T 1− T 2)τ / δ, the conductive heat

loss largely depends on membrane characteristics such as wall thick-

ness (δ) and thermal conductivity (kM ). High kM indicates a highly con-

ductive membrane and hence low thermal (heat-transfer) resistance

across the membrane in MD heat transfer, which leads to more conduc-

tive heat loss. Otherwise, the reduction of conductive heat loss across

the membrane matrix results in a higher thermal ef ciency and im-

proved permeation ux J M , which are key variables in determining the

GOR values. Therefore, as shown in Eq. (1) with other parameters

(e.g., C M , d p/dT , and Δhv) that remainconstants the effects of membrane

characteristics kM and δ on GOR are investigated and presented in Fig. 7.

With the currently used PVDF bers with a thermal conductivity

(kM ) of 0.066 W/m-K as benchmark, a series of membranes with com-

monly studied kM values ranging from 0.03 to 0.09 W m−1·K−1 were

used to investigate the effect of membrane properties on GOR. The

relationship of GOR and kM at three sets of equivalent owrates of

W 1,2/ W 2,1 = 1 are presented in Fig. 7(A). Obviously, the GOR de-

creases with increasing kM regardless of the owrates range. This is

mainly because that a high kM indicates a small thermal resistanceacross the membrane in MD heat transfer and hence greater conductive

heat loss, which is undesired and tends to cause steeper temperature

decline at the feed side and more signicant rise of the permeate tem-

perature. Both leads to a decreased MD driving force and reduced

water production W P [53]. Asa result, the GOR oftheDCMD–HX system

decreases.

The membrane thickness δ, which indicates the distance of water

vapor permeation through the membrane matrix, is another important

membrane property in MD for correlating the thermal resistance and

hence process ef ciency. Fig. 7(B) shows theGOR as a function of mem-

brane wall thickness δ with other membrane properties that remain

constant (e.g., kM = 0.066 W m−1·K−1) in DCMD–HX — an increasing

trend of GOR with increasing δ was observed. This is due to the lower

conductive heat loss with thicker membranes and hence higher heat-transfer resistance, and subsequently a desirable high GOR. However,

in Fig. 7 the improvement of GOR by varying membrane properties kM and δ is fairly insignicant, corresponding to only approximately 10%

decrease in GOR with kM varying from 0.03 to 0.09 W m−1·K−1 and

22% increase with membrane thickness δ rising from 0.15 to 0.35 mm,

respectively. This is due to the non-dominant role of the conductive

heat resistance in MD heat transfer with a thermal ef ciency η greater

than 0.5, which was calculated based on simulation results.

Overall, it is observed that the maximal GOR of a DCMD–HX system

tends to increase with decreasing owrates from W 1,2 = W 2,1 =

5000 kg h−1 to W 1,2 = W 2,1 = 50 kg h−1, indicative of the negative

effect of owrate on promoting GOR. This is consistent with the previ-

ous discussions for Figs. 5 and 6 on system-level evaluation of heat

utilization.

0.1 1

0

1

2

3

4

5

G O

R

W 1,2

/W 2,1

Module #1

Module #2

Module #3

Fig. 6. GORpeaksof DCMD–HXsystems with modules ofvarious geometriccongurations

at varying owrate ratio W 1,2/ W 2,1 (ΔT MD = 50 °C,simulatedmodules #1–#3;HICO mode

(ΔT HX = 0 °C)usedfor HXsimulations of owrate range W 1,2/ W 2,1≤1 and HOCImode for

W 1,2/ W 2,1 N

1).



9/14

3.3.2. Interplay of GOR and thermal ef ciency η

As indicated by Eq. (1), the thermal ef ciency η of MD is mainly af-

fected through the variations of membrane properties. As a summary

of the simulation results presented in Fig. 7, a direct relationship be-

tween GOR and thermal ef ciency η at various owrates is depicted in

Fig. 8. In this series of simulations the x-axis variable η was adjusted

through simultaneously varying membrane conductivity kM and wall

thickness δ , respectively; while the operating conditions W 1,2, W 2,1,

T 1,2 and T 2,1 remain constant. Therefore, two sets of GOR curves are pro-

duced to indicate the combined effectof membrane properties, i.e., solid

and hollow markers representing the changes caused by kM and δ, re-

spectively. Interestingly, all GOR values at the same owrate settings

fall on the same line regardless of the effects incurred by decreasing

kM or increasing δ, indicating that either kM or δ contributes to a com-

mon factor affecting the GOR, i.e., heat-transfer resistance. This linear

relationship between GOR and η exhibits an increasing trend at anygiven owrate. This is due to the improved water production (W P ) at

a higher thermal ef ciency (thicker or less conductive membrane),

which results in up to 32% increase in GOR. This is also consistent with

the relationship explained through Eq. (15).

With regard to the recirculating owrates, however, GOR decreases

with increasingW 1,2 (=W 2,1) at a certain η — this is agreeswell with the

conclusion drawn from Section 3.2 that low recirculating owrates are

preferred for achieving high GOR value. Also, Fig. 8 shows that high

recirculating owrates tend to result in low thermal ef ciency. For

instance, the lowest η at W 1,2 = W 2,1 = 5000 kg h−1 is about 0.48;

while it is greater than 0.60 at much lower owrates of W 1,2 =

W 2,1 = 5 0 kg h−1 for the same thermal resistance δ/ kM . This is attribut-

ed to the trade-off relationship between η and W 2,1, as shown in

Eq. (14).

3.4. Relationship of GOR and water production W P

For a given DCMD–HX system, the concepts of energy utilization and

water productivity are critical in evaluating the process performance.

Although both GORand JM are dependent variables in MD, it is essentialto understand the interplay of two metrics for selecting appropriate op-

erating parameters and membrane characteristics in different situa-

tions, i.e., limited thermal energy supply or abundant resources.

With the commercial PVDF membrane (kM =0.066Wm−1·K−1) as

benchmark, two other scenarios in DCMD–HX were created with two

membranes of different thermal conductivities packed in the same

module conguration as module #1 in Table 2 with constant AM of

100 m2. The simulation results revealing the relationship of GOR and

permeation ux ( J M = W P / AM ) for three membrane types are shown

in Fig. 9, in which the increase of J M was realized through adjusting

the recirculating owrates from 10 kg/h to 10000 kg/h. Clearly, a

trade-off relationship is observed — the GOR declines with rising J M re-

gardless of the thermal conductivity. The curve exhibits an initial sharp

decline and subsequent slow decrease. This can be explained throughthe slower increase of J M as a result of more rapidly increasing

recirculating owrates (Fig. 4(A)), which leads to the decrease of ratio

W P / W 1,2 (= J M / G 4L/ Di,1, Fig. 4(B)) and increase of thermal ef ciency η .

Therefore, the GOR decreases accordingly (Eq. (15)). However, the im-

provement of permeation ux slows down with further increase of

theowrate leading to aattening curve of GOR. This is well understood

that the relatively steady ux at high Re range indicates a shift to the

heat and mass transfer being controlled by the membrane and/or the

lumen-side boundary layer. For instance, the GORs show insignicant

changes as J M exceeds 3 kg/(m2 h). Nevertheless, the GOR values on

all three curves drop below 1.0 beyond this point. Thus, for a given

DCMD–HX system, high GOR often comes at the cost of low J M , or vice

versa. In addition, it is observed in Fig. 9 that the GOR generally de-

creases with increasing kM . For instance, at J M = 1.0 kg/m2

·h, the GOR

0.03 0.04 0.05 0.06 0.07 0.08 0.09

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

6.0

G O R

k M (W m

-1 K

-1)

W 12

=W 21

=50 (kg h-1)

W 12

=W 21

=500 (kg h-1

)

W 12

=W 21

=5000 (kg h-1)

(A)

0.00015 0.00020 0.00025 0.00030 0.00035

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

6.0

(B)

(m)

Fig. 7. Effects of membrane characteristics on GOR in DCMD–HX system in terms of membrane thermal resistance (A) thermal conductivity kM ; (B) wall thickness δ (constant settings in

HX unit and module temperature differences ΔT HX = 0 °C and ΔT MD = 50 °C, simulated module #1 with N = 15089, φ = 0.689 and L = 2153 mm).

0.5 0.6 0.7 0.8 0.9 1.0

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

5.5

G O R

W 12

=W 21

=50 (kg h-1

)

W 12

=W 21

=500 (kg h-1)

W 12

=W 21

=5000 (kg h-1

)

Fig. 8. The relationship of GOR and thermal ef ciency in DCMD–HX system by varying

thermal conductivity (solid markers) and membrane thickness (hollow markers)

(ΔT HX = 0 °C and ΔT MD = 50 °C, simulated module #1 with n = 15089, φ = 0.689 and

L = 2153 mm, kM = 0.03–

0.09 W m

−1

·K

−1

, δ = 0.15–

0.35 mm).



10/14

for system with kM = 0. 033 W m−1·K−1 is 2.5, followed by that with

kM = 0.066 W m−1·K−1; while system with kM of 0.099 W m

−1·K−1

drops below 1.5. The results are corroborating with the conclusion

drawn from Fig. 7(A). This makes substantial difference when sustain-

able solutions are sought for designing an MD system in arid/rural

areas under extreme energy constraints.

In a word, due to the trade-off relationship between the heat recov-

ery ef ciency and water production, multiple factors can be manipulat-

ed for optimizing a DCMD–HX process, e.g., high recirculating owrates

can be employed to achieve high water productivity at the cost of low

GOR. Thus, considerations must be given when limited thermal energy

resourcesare available andhencea high GOR is desired. Otherwise,high

water production is the priority with cheap and abundant waste heat.

3.5. Effect of module scale-up in DCMD–HX desalination system

As the industrialimplementation of MD hasbeen longdiscussed, the

scale-up of MD desalination systemfor higher water recovery is of great

interest. To investigate the scale-up effect, bigger membrane modules

with large membrane area and capacity should be designed. With the

aid of Aspen Plusowsheet simulations, the DCMD–HX system installed

with a series of membrane modules, which have proportionally in-

creased membrane areas with geometrical similarities such as constant

packing density and length-to-diameter ratio L/ Di,2, were studied and

compared. Similar hydrodynamic conditions were maintained at con-

stant mass rate G (i.e., constant Re) regardless of the module size.

Fig. 10 presents the module scale-up effects in terms of system-level

heat utilization ef ciency (GOR and W P , Fig. 10(A)) and specic moduleperformance (γ and J M , Fig. 10(B)).

As shown in Fig. 10(A), both GOR and W P increase with increasing

membrane area AM . Particularly, the curve of GOR vs. AM exhibits an ini-

tial rapid growth as AM increases in the lower range (b10 m2); while a

slower rise is observed with further increase of AM . According to

Eq. (15), the GOR in a specied DCMD–HX system with constant ΔT MDand ΔT HX largely depends on the feed- and permeate-side water recov-

ery rates, which are identical when operated at equivalent owrates,

i.e., γ ∝ (W P / W 1,2) = (W P / W 2,1). Thus, a higher water production W P and hence a higher water recovery rate is achieved at a larger AM ,

which leads to a higher GOR value — the increase of γ mathematically

offsets the increase of GOR in the equation. Compared to the steeper

shape of the GOR curve, the water productivity (W P ) behaves fairly lin-

early with increasing AM .

Correspondingly, Fig. 10(B) shows that the water recovery rateγ has

a similar increasing trend to the GOR curve — a steep increment with

initial increase of membrane area AM , and the curve tends to at out

as module size increases further. Compared to the steady increase of

W P as AM increases, the owrate was varied more signicantly to main-

tain a constant G and hence leading to a slow increase of the recovery

rate for bigger modules. In contrast, in Fig. 10(B) the J M decreases signif-

icantly withincreasingmembranearea.It waswell-understood that this

is mainly due to thedecline of averaged transmembrane driving force as

module length and size increase simultaneously [32].

Overall, coupling the system ef ciency (i.e., GOR, W P in Fig. 10(A))

with module performance (i.e., J M and γ in Fig. 10(B)) reveals thescale-up effect in designing a DCMD–HX system. Clearly, the increase

on module size and membrane area AM greatly facilitates the improve-

ment of system ef ciency as a result of higher GOR and W P . However,

the performance of the membrane module, i.e., the permeation ux, de-

teriorates in a larger scale module. This is consistent with the previous

experimental ndings on the effect of module size on permeation ux

[27]. In a word, different to the misconception of “linear scale-up” of

membrane systems, the non-linear relationship of membrane area and

heat utilization (i.e., GOR) has indicated the possible uncertainty in

accurately predicting the GOR value for industrial-scale desalination

systems based on lab-scale module testing, which usually employs

small membrane area less than 1 m2. Thus, it is anticipated that the

pilot-scale DCMD–HX systems would have higher GORs than the

laboratorial ones.

1.5

2.0

2.5

3.0

3.5

(B)

(A)

G O R

0 5 10 15 20 25 30 35 40 45

0.0

2.5

5.0

7.5

10.0

12.5

J M

( k g h - 1 )

AM (m

2)

0

20

40

60

80

W P ( k

g h - 1 )

5.0%

5.5%

6.0%

6.5%

Fig. 10. Effect of module scale-up (AM) in DCMD–HX system ef ciency in terms of

(A) heat utilization (GOR) and water production (W P ); and module performance in

terms of (B) permeation

ux ( J M ) and water recovery rate (γ ) (Δ

T HX = 0 °C andΔT MD = 50 °C, simulated modules #4–#13 in Table 3, kM = 0.066 W m

−1·K−1).

0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

0

1

2

3

4

5

6

7

G O R

J M (kg m

-2 h

-1)

k M = 0.033 (W m

-1 K

-1)

k M = 0.066 (W m

-1 K

-1)

k M = 0.099 (W m

-1 K

-1)

Fig. 9. Relationship of energy utilization (GOR) with permeation ux water production

( J M ) in DCMD–HX system at varied thermal conductivities (ΔT HX = 0 °C and ΔT MD =

50 °C, simulated module #1, AM = 100 m2).



11/14

3.6. Effect of heat recovery

In the DCMD–HX system, heat recovery unit (HX) is critical for re-

covering reusable heat to reduce the total heat input. In previous sec-

tions, an ideal situation with full heat recovery was considered,

i.e., the additional heat contained in the returning permeate stream

(S 2,2 and T 2,2, Fig. 1) was fully extracted to the feed stream from the

HX unit (S 1,1 and T 1,1, Fig. 1) resulting in ΔT HX = T 2,2 − T 1,1 = 0. A

threshold of Δ

T HX, which relates to the characteristics of the heat ex-changer and ow conditions, practically exists due to the limited heat

exchanger area andnite heat-transfer rate. In reality, theΔT HX is great-

erthan0. A greater ΔT HX indicatespoorerheat recovery in analogy to in-

suf cient area for fully exchanging heat throughout the HX unit. For

instant, at ΔT HX = 5 °C, the temperature of the feed-side ef uent T 1,1will have to be further elevated to approach that of the permeate-side

T 2,2with additional heat-transfer area. Hence, the effectof heat recovery

ef ciency (ΔT HX) on heat utilization (GOR) is essential in the process

design. The simulation results are presented in Fig. 11, in which the

GOR curve shows a decreasing trend as the ΔT HX rises from 0 to 5

°C—a total decline of 30% indicating a strong effect of heat recovery

has on the overall heat utilization of the DCMD–HX system. To achieve

more ef cient heat utilization and high GOR when limited heat supply

is available, the design of highly effective heat exchangers is strongly

recommended. Also, Fig. 11 shows that the GOR is greatly improved at

relatively lower recirculating owrates for both feed and permeate. It

is consistent with previous simulation results obtained based on the

ow conditions in Fig. 5 and has again veried the controlling effect of

ow parameters in designing a DCMD–HX desalination system, as ex-

plained via the factorial analysis in Section 3.3.1.

4. Conclusions

With the aid of Aspen Plus simulations, an integrated direct contact

membrane distillation desalination system with heat recovery (DCMD–

HX)was studied in thecontext of limited heat resources. An implicit ex-

pression of GOR was derived to conveniently correlate the DCMD–HX

system ef ciency in terms of heat utilization using module (unit)

modeling and hence avoid complicated owsheet simulations. Factorialanalysis was conducted to identify the controlling factors for achieving

high GOR in the DCMD–HX desalination system. The following conclu-

sions can be drawn:

(1) Based on the full factorial analysis in terms of absolute effect on

the GOR, the most inuential operational factors were identied

as the recirculating owrates of the feed and permeate streams

(W 1,2 and W 2,1). The rapidly declining trend of water recovery

rate in terms of mass rate G has veried its controlling effect.

(2) The colored contours of GOR in terms of recirculating owrates

have veried the controlling effect of ow conditions in a

DCMD system with heat recovery unit (HX). A maximal GOR

up to 6.0 was obtained in the given MD system. It was found

that the necessary conditions for achieving maximal GORs in a

DCMD–

HX system, indicative of high heat recovery ef

ciency, isthe low equivalent recirculating owrates.

(3) Less conductive (i.e., low membrane thermal conductivity) and

thicker membranes (i.e., large wall thickness) with high heat-

transfer resistance were preferred in achieving a higher GOR.

An increase in GOR was observed by varying either the thermal

conductivity or the thickness in the given ranges. Furthermore,

a linearly increasing relationship between the GOR and thermal

ef ciency was revealed the enhanced MD water production at a

higher thermal ef ciency leading to higher GOR. The strong cor-

relation of GOR and thermal ef ciency in the implicit expression

Eq. (15) were testied.

(4) The interplay of heat recovery ef ciency and water production

in MD was investigated. A trade-off relationship existed be-

tween the GOR and water production, i.e., high water produc-

tivity was achieved at high recirculating owrates at the cost

of low GOR. Multiple factors including operating owrates,

temperatures, membrane characteristics as well as heat recov-

ery unit (HX) could be manipulated for optimizing a DCMD–

HX system. Thus, a compromise must be made and consider-

ations must be given when limited thermal energy resources

are available and hence a high GOR is desired. Otherwise,

high water production is the priority with cheap and abundant

waste heat.

(5) The scale-up effect of was studied by coupling the DCMD–HX

system ef ciency in terms of GOR with module performance in

terms of water recovery rate. The increase of module size and

membrane area greatly promoted the system-level heat utiliza-

tion as a result of high GOR and water production, but deteriorat-

ed the membrane capacity leading to low MD driving force andlow water recovery rate.

(6) The concept of “non-linear scale-up” was proposed for large-

scale MD systems integrated with heat recovery in terms of

thermal energy evaluation. Uncertainties are anticipated with

the attempts to accurately predict the GOR for industrial scale

desalination system based on lab-scale outcomes. Fortunately,

comprehensive process simulations are benecial in designing

anef cient desalination system and improving the practicability

of MD technology.

Nomenclature

AM Membrane area m2

C M Membrane distillation coef cient kg m−2 s−1 Pa−1

c P Specic heat capacity kJ kg−1 K−1

Di Inner diameter m

Do Outer diameter m

G Mass rate in the bulk ow kg m−2 s−1

GOR Gain output ratio Dimensionless

Gz Graetz number Dimensionless

h Specic enthalpy of materials kJ kg−1

kM Thermal conductivi ty of membrane W m−1·K−1

J H Heat ux kJ m−2 h−1

J M Permeation ux kg m−2 h−1

L Length m

N Hollow ber numbers Dimensionless

Nu Nusselt number Dimensionless

p Pressure Pa

Pr Prandtl number Dimensionless

0 1 2 3 4 5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

5.0

G O R

T HX

W 12

=W 21

=50 (kg h-1

)

W 12

=W 21

=500 (kg h-1)

W 12

=W 21

=5000 (kg h-1)

Fig. 11.Effectof heatrecoveryef ciency(ΔT HX) onheat utilization(GOR inDCMD–HXsys-

tem) (Δ

T MD = 50 °C, DCMD module #1).



12/14

qv Heat ow ux of evaporation kJ m−2 h−1

qc Heat ow ux of conduction kJ m−2 h−1

Q c Heat ow of conduction kJ h−1

Re Reynolds number Dimensionless

T Temperature K

v Velocity m s−1

vm Specic volume m3 mol−1

W Mass owrate kg h−1

z Axial location m

Greek letters

δ Membrane thickness mφ Packing density Dimensionless

γ Mass recovery rate Dimensionless

η Averaged thermal ef ciency of DCMD module Dimensionless

μ Viscosity Pa s

π Constant of circumference ratio

ρ Density kg m−3

τ Temperature polarization coef cient Dimensionless

Δhv Latent heat kJ kg−1

ΔT Temperature difference K

Subscripts

0 Inuent of the DCMD module

1 lumen sid e in th e D CMD mod ule

2 Shell sid e in t he DCMD mo du le

HX Heat r ecovery u nit

MD DCMD module

P Production of permeatel Liquid phase

v Vapor phase

W Wall

Acknowledgments

The following funding agencies are gracefully acknowledged for

funding this work: Fundamental Research Funds for the Central Uni-

versities, China (2014ZZ0061); Environmental and Water Industry

Programme Of ce (EWI), Singapore project (#0901-IRIS-0203);

and Industry Postdoctoral Fellowship Scheme, Victoria University,

Australia.

Appendix A.1. Simulation settings for owsheet modeling in Aspen

Plus

Based on the concept proposed in Fig. 1, the detailed steady-state

owsheet simulation for the DCMD–HX desalination system was built

in Aspen Plus V7.3 as shown in Fig. A.1. The following settings of reux

streams were adopted in the simulation to greatly shorten the comput-

ing time:

1. For the quick convergence in computation, the permeate-side recir-

culation was articially broken as the single-pass stream, so the

permeate-side recirculating owrate can be easily specied as the

initial settings of the stream named as S20.

2. The S11 stream was set as TEAR stream [52], so the initial parameters

such as temperature, pressure and mass ow are required in

advance.

3. TheCALCULATIONblock [52] wasused to determine themassowof

feed-side reux assigned by the FSLIT1 block [52].

Fig. A.1. Simulation

owsheet of DCMD–

HX system in Aspen Plus.



13/14

The feedstock seawater is simplied as synthetic seawater, i.e., NaCl

solution and the physical properties of all streams were calculated in

Aspen Properties Engine with using the ELEC-NRTL thermodynamic

model [52]. The initial settings for conducting the Aspen simulations

using module #1 in this study are listed as an example in Table A.1.

References

[1] K.W. Lawson, D.R. Lloyd, Membrane distillation, J. Membr. Sci. 124 (1997) 1–25.[2] X. Yang, A.G. Fane, R. Wang, Membrane Distillation: Now and Future, Scrivener/

Wiley, 2013. 2013.[3] M.S. El-Bourawi, Z. Ding, R. Ma, M. Khayet, A framework for better understanding

membrane distillation separation process, J. Membr. Sci. 285 (2006) 4–29.[4] C. Charcosset, A review of membrane processes and renewable energies for desali-

nation, Desalination 245 (2009) 214–231.[5] E. Curcio, E. Drioli, Membrane distillation and related operations — a review, Sep.

Purif. Rev. 34 (2005) 35–86.[6] Z.W. Ding, L.Y. Liu, M.S. El-Bourawi, R.Y. Ma, Analysis of a solar-powered membrane

distillation system, Desalination 172 (2005) 27–40.[7] P.A. Hogan, Sudjito, A.G. Fane, G.L. Morrison, Desalination by solar heated mem-

brane distillation, Desalination 81 (1991) 81–90.[8] J. Koschikowski, M. Wieghaus, M. Rommel, Solar thermal-driven desalination plants

based on membrane distillation, Desalination 156 (2003) 295–304.[9] Z.S. Wang, Z.L. Gu, S.Y. Feng, Y. Li, Applications of membrane distillation technology

in energy transformation process-basis and prospect, Chin. Sci. Bull. 54 (2009)2766–2780.

[10] M. Hightower, S.A. Pierce, The energy challenge, Nature 452 (2008) 285–286.[11] L. Song, B. Li, K. Sirkar, J. Gilron, Direct contact membrane distillation-based desali-

nation: novel membranes, devices, larger-scale studies, and a model, Ind. Eng.Chem. Res. 46 (2007) 2307–2323.

[12] P. Wang, M.M. Teoh, T.-S. Chung, Morphological architecture of dual-layer hollowber for membrane distillation with higher desalination performance, Water Res.45 (2011) 5489–5500.

[13] M. Khayet, T. Matsuura, J.I. Mengual, Porous hydrophobic/hydrophilic compositemembranes: estimation of the hydrophobic layer thickness, J. Membr. Sci. 266(2005) 68–79.

[14] B. Li, K.K. Sirkar, Novel membrane and device for vacuum membrane distillation-based desalination process, J. Membr. Sci. 257 (2005) 60–75.

[15] M. Khayet, T. Matsuura, J.I. Mengual, M. Qtaishat, Design of novel direct contactmembrane distillation membranes, Desalination 192 (2006) 105–111.

[16] M. Qtaishat, D. Rana, M. Khayet, T. Matsuura, Preparation and characterization of novel hydrophobic/hydrophilic polyetherimide composite membranes for desalina-tion by direct contact membrane distillation, J. Membr. Sci. 327 (2009) 264–273.

[17] M. Su, M.M. Teoh, K.Y. Wang, J. Su, T.-S. Chung,Effectof inner-layer thermal conduc-tivity on ux enhancement of dual-layer hollow ber membranes in direct contactmembrane distillation, J. Membr. Sci. 364 (2010) 278–289.

[18] Y. Liao, R. Wang, A.G. Fane, Engineering superhydrophobic surface on poly(vinylideneuoride) nanober membranes for direct contact membrane distillation, J. Membr.Sci. 440 (2013) 77–87.

[19] Minitab, Meet Minitab 16, Minitab Inc., 2010[20] B. Li, K.K. Sirkar, Novel membrane and device for direct contact membrane

distillation-based desalination process, Ind. Eng. Chem. Res. 43 (2004)5300–5309.

[21] T.Y. Cath, V.D. Adams, A.E. Childress, Experimental study of desalination using directcontact membrane distillation: a new approach to ux enhancement, J. Membr. Sci.228 (2004) 5–16.

[22] L. Martinez, J.M. Rodriquez-Maroto, Effects of membrane and module design im-provements on ux in direct contact membrane distillation, Desalination 205(2007) 97–103.

[23] M.M. Teoh, S. Bonyadi, T.-S. Chung, Investigation of different hollow ber module

designs for ux enhancement in the membrane distillation process, J. Membr. Sci.311 (2008) 371–379.[24] X. Yang, R. Wang, L. Shi,A.G. Fane, M. Debowski, Performance improvementof PVDF

hollow ber-based membrane distillation process, J. Membr. Sci. 369 (2011)437–447.

[25] K.W. Lawson, D.R. Lloyd, Membrane distillation. II. Direct contact MD, J. Membr. Sci.120 (1996) 123–133.

[26] K. He, H.J. Hwang, M.W. Woo, I.S. Moon, Production of drinking water from salinewater by direct contact membrane distillation (DCMD), J. Ind. Eng. Chem. 17(2011) 41–48.

[27] X. Yang, R. Wang, A.G.Fane, Novel designs for improving the performance of hollowber membrane distillation modules, J. Membr. Sci. 384 (2011) 52–62.

[28] X. Yang, E.O. Fridjonsson, M.L. Johns, R. Wang, A.G. Fane, A non-invasive studyof ow dynamics in membrane distillation hollow ber modules using low-eld nuclear magnetic resonance imaging (MRI), J. Membr. Sci. 451 (2014)46–54.

[29] G. Chen, X. Yang, R. Wang, A.G. Fane, Performance enhancement and scaling controlwith gas bubbling in direct contact membrane distillation, Desalination 308 (2013)47–55.

[30] G. Chen, X. Yang, Y. Lu, R. Wang, A.G. Fane, Heat transfer intensication and scalingmitigation i n bubbling-enhanced membrane distillation for brine concentration, J.Membr. Sci. 470 (2014) 60–69.

[31] G. Chen, Y. Lu, X. Yang, R. Wang, A.G. Fane, Quantitative study on crystallization-induced scaling in high-concentration direct-contact membrane distillation, Ind.Eng. Chem. Res. 53 (2014) 15656–15666.

[32] X. Yang, R. Wang, A.G.Fane,C. Tang, I.G.Wenten, Membrane moduledesignand dy-namic shear-induced techniques to enhance liquid separation by hollowber mod-ules: a review, Desalin. Water Treat. 51 (2013) 3604–3627.

[33] A.G. Fane, R.W. Schoeld, C.J.D. Fell, The ef cient use of energy in membrane distil-lation, Desalination 64 (1987) 231–243.

[34] S. Al-Obaidani, E. Curcio, F. Macedonio,G. Di Proo, H. Al-Hinai, E. Drioli, Potential of membrane distillation in seawater desalination:thermal ef ciency, sensitivity studyand cost estimation, J. Membr. Sci. 323 (2008) 85–98.

[35] H.Y. Lee, F. He, L.M. Song, J. Gilron, K.K. Sirkar, Desalination with a cascade of cross-ow hollow ber membrane distillation devices integrated with a heat exchanger,AICHE J. 57 (2011) 1780–1795.

[36] E.K. Summers, H.A. Arafat, J.H. Lienhard, Energy ef ciency comparison of single-stage membrane distillation (MD) desalination cycles in different congurations,Desalination 290 (2012) 54–66.

[37] R.W. Schoeld, A.G. Fane, C.J.D. Fell, The ef cient use of energy in membrane distil-lation, Desalination 64 (1987) 231–243.

[38] S. Bouguecha, B. Hamrouni, M. Dhahbi, Small scale desalination pilots powered byrenewable energy sources: case studies, Desalination 183 (2005) 151–165.

[39] J. Gilron, L. Song, K.K. Sirkar, Design for cascade of crossow direct contact mem-brane distillation, Ind. Eng. Chem. Res. 46 (2007) 2324–2334.

[40] G. Zuo, R. Wang, R. Field, A.G.Fane, Energy ef ciency evaluation and economic anal-yses of direct contact membrane distillation system using Aspen Plus, Desalination283 (2011) 237–244.

[41] M. Khayet, Solar desalination by membrane distillation: Dispersion in energy con-sumption analysis and water production costs (a review), Desalination 308(2013) 89–101.

[42] S. Lin, N.Y. Yip, M. Elimelech, Direct contact membrane distillation with heat recov-ery: thermodynamic insights from module scale modeling, J. Membr. Sci. 453(2014) 498–515.

[43] R.B. Sa ffarini, E.K. Summers, H.A. Arafat, J.H. Lienhard V, Technical evaluation of stand-alone solar powered membrane distillation systems, Desalination 286

(2012) 332–

341.[44] M. Khayet, Membranes and theoretical modeling of membrane distillation: a re-

view, Adv. Colloid Interf. Sci. 164 (2011) 56–88.[45] X. Yang, H. Yu, R. Wang, A.G. Fane, Analysis of the effect of turbulence promoters in

hollow ber membrane distillation modules by computational uid dynamic (CFD)simulations, J. Membr. Sci. 415 (2012) 758–769.

[46] H. Yu,X. Yang, R. Wang, A.G.Fane,Numerical simulationof heatand masstransferindirect membrane distillation in a hollow ber module with laminar ow, J. Membr.Sci. 384 (2011) 107–116.

[47] H. Yu, X.Yang, R.Wang, A.G. Fane, Analysis ofheatandmasstransfer by CFD for per-formance enhancement in direct contact membrane distillation, J. Membr. Sci. 405(2012) 38–47.

[48] G. Guan, X. Yang, R. Wang, R. Field, A.G. Fane, Evaluati

evaluation of heat utilization in membrane distillation desalination system.pdf

Documents