Desalination 366 (2015) 8093

Contents lists available at ScienceDirect

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

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 China
Institute for Sustainability and Innovation, College of Engineering and Science, Victoria University, PO Box 14428, Melbourne, Victoria 8001, Australia
Singapore Membrane Technology Centre, Nanyang Technological University, 639798 Singapore, Singapore
d
School of Civil and Environmental Engineering, Nanyang Technological University, 639798 Singapore, Singapore
b
c

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 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

a b s t r a c t
Aiming to optimize the system-level heat utilization, a pilot-scale direct contact membrane distillation desalination system integrated with heat recovery (DCMDHX) 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 DCMDHX desalination system. Compared to operating temperatures, the feed/permeate recirculating owrates were identied as the most
inuential operational 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, however, resulted in undesirable low water productivity, which was consistent with the trade-off relationship
observed between the heat utilization efciency 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
THX 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 promising 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
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: xing.yang@vu.edu.au (X. Yang), rwang@ntu.edu.sg (R. Wang).

http://dx.doi.org/10.1016/j.desal.2015.01.013
0011-9164/ 2015 Elsevier B.V. All rights reserved.

a hydrophobic porous membrane and condensates by the cold distillate stream in direct contact MD (DCMD) mode. Compared to conventional 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 performance 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 regions [3,69]. Thus, such desalination technology serves dual roles in relieving global water shortage as well as energy crisis and enabling more
and more arid areas/countries to access safe desalted water [10].

G. Guan et al. / Desalination 366 (2015) 8093

In recent decades the resurgence of research interest in MD is mainly

attributed to the advancement in polymer material developments and
breakthroughs in membrane fabrication technologies [1119], and
novel module designs [2028] as well as ow enhancement techniques
to alleviate the temperature polarization phenomenon and enhance
permeation ux [2932]. However, the main challenge for the commercialization 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 lowgrade 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 recovery units can extend the applicability of MD to more rural regions.
Optimal heat recovery is also essential in reducing operational costs.
Yet, thus far only limited studies are available in the literature on energy
analysis in terms of heat utilization and the interplay of various operating parameters. Also, no standardized/universal correlations have been
developed to evaluate system-level energy efciency 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 efciency , gain output ratio (GOR) and water production.
As a common measure of the process efciency 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
[3436], improved module designs, effective insulation, optimized piping system and multi-staged operation [3739]. However, a trade-off relationship is found between the GOR and permeation rate [40,41]
i.e., high GOR could be achieved by designing a system with large membrane area, low ow velocity and more recovery stages; while the ux
decreased due to either the decreased temperature driving force or severe temperature polarization effect. A module-scale thermodynamic
analysis of DCMD modules suggested that high GOR could be achieved
at 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 literature 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 indicating that the prediction of GOR could be effected by various complex
factors such as ow conditions, operating temperatures, and even membrane properties. A full factorial analysis on operational factors affecting
the GOR is yet to be comprehensively explored.
To achieve a system-level optimization in a predictive manner, process 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 performance and reduce energy consumption [36,40,4447]. 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 crystallization system for high salinity brine treatment with zero discharge [49]

81

has shown the feasibility of the user unit operation model for simulating
the module performance and evaluating process efciency 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 reported in [48], this current work aims to explore a direct contact membrane distillation desalination system integrated with heat recovery
(DCMDHX) for leveraging the advantages of MD practicability in the
context of limited heat resource. An implicit expression of GOR was derived to conveniently correlate the DCMDHX system efciency in
terms of heat utilization with single-unit hollow ber 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 DCMDHX desalination system. The newlydeveloped implicit GOR correlation was testied through a series of investigations such as the interplay between GOR and various process
variables (dependent or independent), including owrates, inuent
temperatures of feed and permeate streams, thermal efciency of MD
module that is strongly affected by membrane properties, as well as
water recovery rate. The concept 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 and
energy 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].
The current transport model has been veried previously [48], based
on an established DCMD system for a series of experimental settings, including various feed inlet temperatures, ber lengths and ow velocities. Also, the membrane properties were the same as that in previous
verication experiments. Hence, the model verication was not repeated here and the veried MD model was used as a customized unit for
Aspen owsheet simulation in the following sections.
2.2. DCMDHX 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 DCMDHX. The recovered
heat could be utilized to preheat the brine feed inuent before entering
the membrane module.
A series of pilot-scale hollow ber 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

82

G. Guan et al. / Desalination 366 (2015) 8093

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

Mass
Momentum
Energy
B.C.

Heat transfer coefcient

Transmembrane heat ux
Film heat ux
Conductive heat ux
Permeation ux

Lumen side

Shell side

d1 v1
4

J
dz
Di;1 M



d
dv1
p 2 1
1 v21 0
dz 1
dz
dT 1
dp
4
v1 1
J
1 v1 cp;1
Di;1 H;1
dz
dz


v1 z0 v1;0
p1 z0 p1;0
T 1 z0 T 1;0



0:14
Nu 1:86Gz1=3
w
JH,1 = JMhv|TW,1 + JH,c
JH,1 = h1(T1 TW,1)

M 
J H;c
T W;1 T W;2




J C M p T W;1 p 

4N Do;1
d2 v2

JM
dz
D2i;2



d
dv2
p 2
v22 0
dz 2
dz
4N Do;1
dT 2
dp
v2 2
2 v2 cp;2
J H;2
dz
dz
D2i;2


v2 zL v2;0
p2 zL p2;0
T 2 zL T 2;0





1=4
Nu 0:4Re1=2 0:06Re2=3 Pr 0:4
w
JH,2 = JMhv|TW,2 + JH,c
JH,2 = h2(TW,2 TW,2)

W;1

DCMDHX 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 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 membrane used in this series of simulations has an averaged thermal conductivity (kM) and MD coefcient (CM) of 0.066 kJ m 1 K 1 and
3.8 107 kg m2 s1 Pa1 [48], respectively.
Using the same hollow ber membrane, another set of modules was
designed to investigate the scale-up effect of the DCMDHX desalination 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 efuent was
recycled 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 efciency.
The conceptual schematic of the DCMDHX 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, T1,0 = 25 C)
joins the returning brine stream (feed efuent) pumped back from the
DCMD module and forms a new stream S1,0 entering the preheater
(HX), in which S1,0 recovers heat from the returning permeate stream
S2,2 and becomes stream S1,1 with elevated temperature. The feed
stream S1,1 then ows through a heater and becomes the inlet stream
S1,2 of the MD module with a specied inlet temperature T1,2. In the
Table 2
Specications for DCMD modules used in Aspen Plus owsheet simulations (membrane
properties: Di,1 = 0.98 mm, = 0.24 mm, kM = 0.066 kJ m1 K1 and CM = 3.8
107 kg m2s1Pa1).
Module

Membrane area, Am
Packing density,
Housing Inner diameter. Di,2
Fiber length, L
Ratio of module length to housing
diameter, L/Di,2

mm
mm

#1

#2

#3

100.0
0.689
216
2153
10.0

100.0
0.502
253
2153
8.5

50.2
0.689
216
1080
5.0

W;2 T W;2

MD module, water vapor is generated from the hot feed driven by temperature 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 between hot feed stream S1,2 and cold permeate stream S2,1. Subsequently
the feed temperature decreases from T1,2 (stream S1,2) to T1,3 (stream
S1,3) and the permeate temperature rises from T2,1 (stream S2,1) to T2,2
(stream S2,2) along the module length. The heat gained by the permeate
stream is further utilized through the heat recovery unit HX to preheat the feedstock. In this DCMDHX process most feed concentrates
(efuent) are recycled continuously at the hot side with fresh feedstock
to maintain a given owrate W1,2; 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 W2,1 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 efciency of DCMD module
In the DCMDHX 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.
The overall heat ux q including latent heat (qv) of evaporation and conduction 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 efciency , which is dened as the ratio of latent heat to the total heat input, is widely used to evaluate the effectiveness of heat utilization associated with water production [35]. Hence,
the universal expression of is given as:
dp
hv
CM
qv
J M hv
dT

dp

qv qc J h M T T
hv M
CM
M
v
1
2

dT

where JM is the permeation ux indicating MD performance, kg/(m2 h);

hv is the specic latent heat of evaporation, kJ kg1; operating parameters such as (T1 T2) is the bulk temperature difference between the
feed and permeate, and is dened as temperature polarization coefcient, which characterizes the actual transmembrane driving force of
the heat- and mass-transfer and is strongly inuenced by ow conditions

G. Guan et al. / Desalination 366 (2015) 8093

83

Table 3
Module specications and operating owrates for scale-up effect study of DCMDHX system (T1,2 = 80 C and T2,1 = 30 C, membrane properties: Di,1 = 0.98 mm, = 0.24 mm, kM =
0.066 kJ m1K1 and CM = 3.8 107 kg m2s1Pa1).
Module

Scale-up factor
Number of bers
Module shell diameter
Module Length
Lumen-side owrate
Shell-side owrate
Membrane area

mm
mm
kg/h
kg/h
m2

#4

#5

#6

#7

#8

#9

#10

#11

#12

#13

1
72
15
200
10
10
0.044

2
288
30
400
40
40
0.355

3
648
45
600
90
90
1.197

4
1152
60
800
160
160
2.837

5
1800
75
1000
250
250
5.542

6
2592
90
1200
360
360
9.576

7
3528
105
1400
490
490
15.21

8
4608
120
1600
640
640
22.70

9
5832
135
1800
810
810
32.32

10
7200
150
2000
1000
1000
44.33

(e.g., Reynolds number, Re) [48], the kM and are respectively the membrane conductivity and thickness, and the CM is the membrane distillation coefcient. In this current study, the averaged was determined
and the averaged temperature was used to calculate dp/dT and latent
heat in Eq. (1). When the MD system is operated at constant inuent
temperatures, the mainly depends on both the MD coefcient (CM)
and characteristics (i.e., kM and ) 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 sufciently
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 zh1 z W 1 z zh1 z z J M hv Di;1 z qc Di;1 z:

When the limit taken as z approaches zero, Eq. (2) can be simplied and rearranged as:
dW 1 h1 J M hv qc Di;1 dz:

The mass balance through the element gives:

dW 1 J M Di;1 dz:

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

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

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

W 1 dh1 J M hv qc Di;1 dz:

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

be derived based on the thermodynamic relation as:



1
vm
dp
dh cP dT T
dp cP dT :

Substituting Eqs. (1) and (7) into Eq. (6) when the pressure drop is
negligible, the thermal efciency in the DCMD module given in Fig. 1,
can be expressed in terms of feed-side temperature change as:

W P hv
2369W P

W 1;2 cP;1 T 12 T 13 W 1;2 cP;1 T 12 T 13

where latent heat of 2369 kJ/kg [51] at the averaged temperature of feed
and permeate (55 C) is used in this work.
2.3.2. Calculation of gain output ratio (GOR) in DCMDHX process
simulations
As one of the most useful measures, the GOR is often used to evaluate the MD performance in terms of the specic energy required for per
kg distillate output. The benets of the DCMDHX system is to possibly
recover 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

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

with heat recovery unit.

Fig. 2. Heat and mass proles across a sufciently thin cross-sectional element of a DCMD
module.

84

G. Guan et al. / Desalination 366 (2015) 8093

the total heat input to the system. Subsequently, the process efciency
is greatly improved. The GOR in the DCMDHX system (Fig. 1) is dened
as the ratio of latent heat for evaporation associated with water production WP to total heat input to the heater:
GOR

W h
2369W P
 P v



W 1;2 h1;2 h1;1
W 1;2 cP;1 T 1;2 T 1;1

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

dimensionless groups:

GOR

3
!2


4L 4
hv
J

5 M
Di;1 c
G1
T
T
P;1
1;2
1;1

10

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
owrate per unit cross-sectional area of the stream in the ber channels,
namely mass rate G1 = 4 W1,2 / (ND2i,1), the ratio JM/G1 is often referred
to as the water recovery rate , which is a key metric for desalination
systems [42]. Based on Eq. (10), the magnitudes of each term can be estimated for performing a factorial analysis in 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 DCMDHX system is in the magnitude range of
10+110+2. For most of MD membrane, the magnitude of permeation
ux is in the range of 10110+1 kg m2 h1. Thus, the mass rate G1 is
estimated to be from 10+3 to 10+6 kg m2 h1 when taking GOR as 1.
2.3.3. Correlation of GOR and thermal efciency (Implicit expression of
GOR)
Since GOR and thermal efciency are both related to the effective
latent heat of evaporation, a correlation of GOR and can be derived
for a system-level evaluation of MD. In an integrated DCMDHX system,
the performance of DCMD module and HX is determined by the bulk
temperature difference of the MD operation (TMD) and the temperature difference in the HX (THX), respectively. Based on the process
schematic in Fig. 1, the TMD is expressed in terms of inlet temperature
difference of the feed and permeate and is xed at 50 C in this study:
T MD T 1;2 T 2;1

11

And the THX, reects the extent of heat recovery in the HX unit and
is also considered as an input in the given system:
T HX T 2;2 T 1;1 :

12

For an ideal heat recovery unit, an innite heat-exchanging area results in a complete recovery of sensible heat from the permeate stream.
Two scenarios were considered: 1) when the feed owrate (W1,0) is not
greater than that of the permeate stream (W2,2), the temperature of the
cold-side efuent 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 W1,0 N W2,2, the temperature of the hot-side efuent
approaches the cold-side inuent in the HX unit, namely HOCI mode.
This study mainly focuses on the HICO mode for heat exchanger simulation (W1,0 = W2,2) and hence assumes a constant THX = 0 in the
following discussions, except the investigation of GOR vs. THX in
Section 3.6.

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

GOR

W P hv


W 1;2 cP;1 T MD T HX T 2;1 T 2;2

13

where WP/W1,2 can also be written as JM/G1, i.e., water recovery rate .
Similar to the derivation of Eq. (8), the thermal efciency can be
expressed through the permeate-side temperature change as:

W P hv
2369W P



:
W 2;1 cP;2 T 2;1 T 2;2
W 2;1 cP;2 T 2;1 T 2;2

14

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

GOR comprising of dimensionless groups as:
Wp 1
hv
WP 1
hv

1:
cP;1 T MD T HX W 1;2 GOR cP;2 T MD T HX W 2;1

15

The above implicit equation clearly shows a general relationship

between the GOR and thermal efciency in the DCMDHX system, it
has an advantage to conveniently correlate the module performance
with the desalination system efciency when heat recovery capacity
THX 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 DCMDHX system, Eq. (15) requires only the outputs of the
DCMD module to predict the GOR of the whole system. Thus, timeconsuming and complicated process simulations can be avoided.
2.3.4. Factorial analysis of operational factors affecting GOR
In the DCMDHX 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 (T1,2 and T2,1) and
recirculating owrates (W1,2 and W2,1) 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 DCMDHX, commercial software
Minitab 16 was used to conduct the factorial analysis.
According to the analysis in Section 2.3.2, in this study the mass rate G
of feed and permeate streams is varied from 10+3 to 10+6 kg m2 h1
and hence the magnitude of the owrates W1,2 and W2,1 can be calculated
as 10+110+4 kg m2 h1. The upper temperature limit for feed inlet is
set as 80 C and the lower temperature limit for cold permeate inlet T2,1
is set as 30 C allowing minimal refrigeration requirements and low investment cost. The four factors including the mass owrates and inlet
temperatures i.e., W1,2, W2,1, T1,2, and T2,1, are denoted as A, B, C, and
D, respectively and the corresponding low and high levels for each factor 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 DCMDHX
owsheet simulations, which will be analyzed in Section 2.4 to identify
their impacts on the GOR.
2.4. Flowsheet simulation of DCMDHX 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 physicochemical properties
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. The solved proles of uid temperature, permeation
ux, pressure and owrate served as the unit outputs for module

G. Guan et al. / Desalination 366 (2015) 8093

Table 4
Full factorial analysis for screening operation factors affecting GOR in DCMDHX system.
Notation

Factor

A
B
C
D

W1,2
W2,1
T1,2
T2,1

kg/h
kg/h
C
C

Low level

High level

10
10
60
30

10000
10000
80
50

Run #

GOR

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

10
10000
10
10000
10
10000
10
10000
10
10000
10
10000
10
10000
10
10000

10
10
10000
10000
10
10
10000
10000
10
10
10000
10000
10
10
10000
10000

60
60
60
60
80
80
80
80
60
60
60
60
80
80
80
80

30
30
30
30
30
30
30
30
50
50
50
50
50
50
50
50

6.598
0.341
0.708
0.659
5.856
0.411
0.726
0.775
10.19
0.189
1.042
0.838
7.817
0.276
1.057
0.960

evaluation. The detailed computational algorithm for MD module

modeling can be found in the literature [48].
2.4.2. DCMDHX owsheet simulations
In the current study the ow chart of an integrated DCMDHX desalination system was established to recover fresh water with heat regeneration 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 consistent
with that of the conceptual DCMDHX 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 DCMDHX system
using module #1 is provided in Appendix A.1, Table A.1.

85

Table 5
Comparison of the GORs correlated by Eqs. (9) and (15) (HX unit and module temperature
differences THX = 0 C and TMD = 50 C, i.e., T1,2 = 80 C & T2,1 = 30 C, simulated
module #1).
W1,1 = W2,1

GOR

Relative error

(kg/h)

Correlated in Eq. (15)

Correlated in Eq. (9)

10
20
50
100
200
500
1000
2000
5000
10000

5.857
5.637
5.059
6.183
4.481
3.520
2.943
2.029
1.206
0.775

5.854
5.635
5.064
6.169
4.500
3.521
2.941
2.028
1.205
0.773
RMS

0.05%
0.04%
0.11%
0.23%
0.43%
0.01%
0.09%
0.03%
0.06%
0.18%
0.17%

error less than 0.17%. Therefore, using Eq. (15), the GOR of the DCMD
HX system can be accurately predicted based on the module performance under similar outputs such as water productivity WP and
thermal efciency at specied uid properties and heat exchanger
settings (TMD & THX). Thus, it is possible to evaluate the heat utilization of such system in a simpler manner to avoid performing timeconsuming owsheet simulations.
3.2. Effects of DCMD operating conditions in DCMDHX system
According to Eq. (15), the GOR in the DCMDHX system is closely
related to the operating conditions and module performance, which
largely depends on membrane characteristics and operating parameters. This section focuses on the factorial analysis to investigate the
effects of four operational factors on the heat utilization in DCMDHX
with a constant THX of 0 C.
3.2.1. Factorial analysis of factors affecting GOR
As discussed previously, in the DCMDHX system four operational
factors, i.e., T1,2 and T2,1, and W1,2 and W2,1, are the key variables affecting the total heat input and distillate output. Based on the GOR values
obtained 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 (combinations) are illustrated in the Pareto diagram shown as Fig. 3, in which
the impact of each factor and interactions of factors is illustrated as

3. Results and discussion

1.519
A

3.1. Comparison of correlated and simulated GOR

AB
B
D
AD
ABD
BD

Terms

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 conveniently
correlate the DCMDHX system efciency with module performance
by specifying operating conditions, i.e., recirculating owrates W1,2 &
W2,1, xed uid properties hv & cP, and heat exchanger settings
TMD = 50 C & THX = 0 C. Thus the simulation outputs were used
to correlate the GOR using Eq. (15). Alternatively, DCMDHX 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 simulations; while the latter relies on comprehensive owsheet simulations of the DCMDHX 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

AC
BC

Factor Name
A
W1,2

ABC
C
ACD

W2,1

ABCD

T1,2

BCD

T2,1

CD
0

Absolute effect
Fig. 3. Pareto diagram of full factorial analysis for factors affecting GOR in DCMDHX
system.

86

G. Guan et al. / Desalination 366 (2015) 8093

horizontal bars. The length of each bar is proportional to the effect of individual 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 showed statistically signicant inuence within a condence level of 95% [19].
As shown in Fig. 3, factor A (W1,2) is found to be the most inuential
parameter affecting GOR, followed by factor B (W2,1) and then AB; while
the combined factor of C (T1,2) & D (T2,1) has the least effect on GOR.
Among all factors, only the recirculating owrates A and B, and the combined AB show an absolute effect exceeding the critical line of 1.519.
Other than the ow parameters W1,2 and W2,1, the GOR values seem
to be statistically irrelevant to the operating temperatures T1,2 (factor
C) and T2,1 (factor D), 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 TMD related terms in both
denominator and numerator (i.e., WP TMD) that weakens the impact
of temperature on GOR.
Similarly, the signicant effect of the feed-side recirculating owrate
on the GOR can be explained through Eq. (15), in which the ratio of
water productivity to recirculating owrate WP/W1,2 determines the
GOR. Dened as water recovery rate , the ratio JM/G1 is proportional
to WP/W1,2 and is critical for designing a DCMD module. Thus, the
relationship of JM 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 JM with increasing

35
30

(A)

-1

20

-2

JM (kg m h )

25

15
10

G1. This is due to the increase of MD driving force with reduced boundary 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 JM is not proportional to the increase of mass rate G1. To further investigate the reasoning,
another graph of water recovery rate vs. G1 is given in Fig. 4(B), in which an
initial steep decline of JM/G1 at low G1 (b 103 kg m2 h1) and then a
slow decrease is observed until G1 reaches 107 kg m2 h1. This is because of the extremely low permeate rate JM associated with high transmembrane resistance at low G1; as G1 further increases, the improved
transmembrane mass and heat transfer promotes a signicant increase
of JM leading to a mild decreasing trend of water recovery rate JM/G1.
This has veried the strong impact of JM/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
DCMDHX system, the GOR was quantied at simultaneously varied
feed- and permeate-side recirculating owrates (W1,2 and W2,1) at
constant heat exchanger settings THX = 0 C and TMD = 50 C with
T1,2 = 80 C & T2,1 = 30 C. Fig. 5 shows the simulation results of GOR
as colored contour in terms of the feed and permeate recirculating
owrates (W1,2 and W2,1) ranging from 101 to 104 kg h1 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 DCMDHX system varies signicantly at
varying operating owrates. With a maximum of 6.0 achieved at a combination of equally low W1,2 and W2,1, 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 W1,2 and high
W2,1, 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 better energy utilization of the DCMDHX system can be achieved at equivalent feed- and permeate-side recirculating owrates W1,2 and W2,1,
preferably in the lower owrate range. This is similar to the ndings
in the literature [42,49] that in DCMD the owrate of feeding inuent
needs to match with the permeate side for achieving higher permeation
ux and hence better module performance. In this study this phenomenon can be conveniently explained via Eq. (15) assuming insignicant

5
0

GOR for module 1

-55

4.0

3.5

0.000
0.7325
1.465

-1

log W2,1/(kg h )

(JM/G1) x 10

2.197

3
2

2.930

3.0

3.662
4.395

2.5

5.127
5.860

2.0

1
1.5

(B)
1.0

0.0

5.0x10

1.0x10

1.5x10

1.0

1.5

2.0

2.5

3.0

3.5

4.0

-1

-2

-1

G1 (kg m h )
Fig. 4. Effect of feed-side mass rate G1 on the (A) permeation ux JM; (B) water recovery
rate JM/G1 in DCMDHX system (simulated module #1, W1,2 = W2,1, T1,2 = 80 C and
T2,1 = 30 C).

log W1,2/(kg h )
Fig. 5. The GOR colored contours at varied recirculating owrates (log scales) in DCMD
HX system (simulated module #1, TMD = 50 C with T1,2 = 80 C and T2,1 = 30 C;
HICO mode (THX = 0 C) used for HX simulations of owrate range W1,2/W2,1 1
(top-left contour) and HOCI mode for W1,2/W2,1 N 1 (bottom-right contour)).

G. Guan et al. / Desalination 366 (2015) 8093

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

the increase of either W1,2 or W2,1 will lead to the decrease of WP/W1,2
or WP/W2,1 in the rst or second terms on the left-hand side of the equation and hence a reduced GOR. In the case of equal W1,2 and W2,1, the
decrease of owrates results in the increase of WP/W1,2 (= WP/W2,1)
and eventually signicant improvement of GOR in Eq. (15) to balance the equation. Therefore, equivalent feed- and permeate-side
owrates in the low range (b101.5 kg h 1) in the DCMDHX system
are necessary for achieving high GOR values (N 5.0). This has further veried the strong impact of operating owrates on GOR in the DCMDHX
process.
To generalize the effects of owrates, similar simulations were
conducted to analyze GORs in DCMDHX with three MD modules of different congurations packed with the same number of bers N = 15089
listed in Table 2. Benchmarked against module #1 ( = 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 W1,2/W2,1, which would further
highlight the inuence of equivalent owrates on heat utilization in
different DCMDHX systems. It is noted that W2,1 was adjusted from
10 to 1000 kg/h at a xed W1,2 of 50 kg h1.
In Fig. 6 a single GOR peak is observed for all DCMDHX systems
operated at equivalent owrates, i.e., W1,2/W2,1 = 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
WP is anticipated because of the deterioration of shell-side (feed) heat
transfer and lower permeation ux JM (WP = JMAM) at a the same
4W 2;1
). Subsequently,
owrate but lower Reynolds number, Re ( p

D2i;2 nD2o;1
a lower GOR value is 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 (WP) due to the greatly reduced
membrane area to 50 m2. Nevertheless, the GOR peaks of DCMDHX
systems with modules #1 to #3 are still considered quite similar in
shapes and heights.
Based on the above discussions for Figs. 5 and 6, the unity of owrate
ratio (W1,2 = W2,1) seems to be necessary for achieving maximum

Module #1
Module #2
Module #3

GOR

0
0.1

W1,2/W2,1
Fig. 6. GOR peaks of DCMDHX systems with modules of various geometric congurations
at varying owrate ratio W1,2/W2,1 (TMD = 50 C, simulated modules #1#3; HICO mode
(THX = 0 C) used for HX simulations of owrate range W1,2/W2,1 1 and HOCI mode for
W1,2/W2,1 N 1).

87

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 the assumption of negligible inuence of concentration on vapor pressure (Section 2.1). In addition, 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 efciency (GOR) in the DCMDHX system. Besides the operating ow and temperature conditions, the GOR in the DCMDHX process
also greatly depends on the module performance and membrane characteristics, which are often evaluated in terms of thermal efciency ()
and water productivity (WP). Thus the relationship of GOR and thermal
efciency is further explored in the following sections.
3.3. Relationship of GOR and thermal efciency
3.3.1. Effects of membrane characteristics on GOR
The inevitable conductive heat loss plays an important role in determining the heat-transfer resistance and thermal efciency of the MD
process. As dened in Eq. (1) as kM(T1 T2) / , the conductive heat
loss largely depends on membrane characteristics such as wall thickness () and thermal conductivity (kM). High kM indicates a highly conductive membrane and hence low thermal (heat-transfer) resistance
across the membrane in MD heat transfer, which leads to more conductive heat loss. Otherwise, the reduction of conductive heat loss across
the membrane matrix results in a higher thermal efciency and improved permeation ux JM, which are key variables in determining the
GOR values. Therefore, as shown in Eq. (1) with other parameters
(e.g., CM, dp/dT, and hv) that remain constants the effects of membrane
characteristics kM and 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 commonly studied kM values ranging from 0.03 to 0.09 W m1K1 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 decreases with increasing kM regardless of the owrates range. This is
mainly because that a high kM indicates a small thermal resistance
across 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 temperature. Both leads to a decreased MD driving force and reduced
water production WP [53]. As a result, the GOR of the DCMDHX system
decreases.
The membrane thickness , 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 efciency. Fig. 7(B) shows the GOR as a function of membrane wall thickness with other membrane properties that remain
constant (e.g., kM = 0.066 W m1K1) in DCMDHX an increasing
trend of GOR with increasing was observed. This is due to the lower
conductive heat loss with thicker membranes and hence higher heattransfer resistance, and subsequently a desirable high GOR. However,
in Fig. 7 the improvement of GOR by varying membrane properties kM
and is fairly insignicant, corresponding to only approximately 10%
decrease in GOR with kM varying from 0.03 to 0.09 W m 1K 1 and
22% increase with membrane thickness 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 efciency greater
than 0.5, which was calculated based on simulation results.
Overall, it is observed that the maximal GOR of a DCMDHX system
tends to increase with decreasing owrates from W1,2 = W2,1 =
5000 kg h 1 to W1,2 = W2,1 = 50 kg h1, indicative of the negative
effect of owrate on promoting GOR. This is consistent with the previous discussions for Figs. 5 and 6 on system-level evaluation of heat
utilization.

88

G. Guan et al. / Desalination 366 (2015) 8093

6.0

6.0
-1

(A)

5.5

(B)

5.5

-1

W12=W21=500 (kg h )
-1

5.0

GOR

W12=W21=50 (kg h )

5.0

W12=W21=5000 (kg h )

4.5

4.5

4.0

4.0

3.5

3.5

3.0

3.0

2.5

2.5

2.0

2.0

1.5

1.5

1.0

1.0

0.5

0.5
0.03

0.04

0.05

0.06
-1

0.07

0.08

0.09

0.00015

0.00020

0.00025

0.00030

0.00035

(m)

-1

kM (W m K )

Fig. 7. Effects of membrane characteristics on GOR in DCMDHX system in terms of membrane thermal resistance (A) thermal conductivity kM; (B) wall thickness (constant settings in
HX unit and module temperature differences THX = 0 C and TMD = 50 C, simulated module #1 with N = 15089, = 0.689 and L = 2153 mm).

3.3.2. Interplay of GOR and thermal efciency

As indicated by Eq. (1), the thermal efciency of MD is mainly affected through the variations of membrane properties. As a summary
of the simulation results presented in Fig. 7, a direct relationship between GOR and thermal efciency at various owrates is depicted in
Fig. 8. In this series of simulations the x-axis variable was adjusted
through simultaneously varying membrane conductivity kM and wall
thickness , respectively; while the operating conditions W1,2, W2,1,
T1,2 and T2,1 remain constant. Therefore, two sets of GOR curves are produced to indicate the combined effect of membrane properties, i.e., solid
and hollow markers representing the changes caused by kM and , respectively. Interestingly, all GOR values at the same owrate settings
fall on the same line regardless of the effects incurred by decreasing
kM or increasing , indicating that either kM or contributes to a common factor affecting the GOR, i.e., heat-transfer resistance. This linear
relationship between GOR and exhibits an increasing trend at any
given owrate. This is due to the improved water production (WP) at
a higher thermal efciency (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).

5.5
-1

5.0
4.5

W12=W21=50 (kg h )
-1

W12=W21=500 (kg h )
-1

W12=W21=5000 (kg h )

GOR

4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5

0.6

0.7

0.8

0.9

1.0

Fig. 8. The relationship of GOR and thermal efciency in DCMDHX system by varying
thermal conductivity (solid markers) and membrane thickness (hollow markers)
(THX = 0 C and TMD = 50 C, simulated module #1 with n = 15089, = 0.689 and
L = 2153 mm, kM = 0.030.09 W m1K1, = 0.150.35 mm).

With regard to the recirculating owrates, however, GOR decreases

with increasing W1,2 (=W2,1) at a certain this is agrees well 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 efciency. For
instance, the lowest at W1,2 = W2,1 = 5000 kg h 1 is about 0.48;
while it is greater than 0.60 at much lower owrates of W1,2 =
W2,1 = 50 kg h1 for the same thermal resistance /kM. This is attributed to the trade-off relationship between and W2,1, as shown in
Eq. (14).
3.4. Relationship of GOR and water production WP
For a given DCMDHX system, the concepts of energy utilization and
water productivity are critical in evaluating the process performance.
Although both GOR and JM are dependent variables in MD, it is essential
to understand the interplay of two metrics for selecting appropriate operating parameters and membrane characteristics in different situations, i.e., limited thermal energy supply or abundant resources.
With the commercial PVDF membrane (kM = 0.066 W m1K1) as
benchmark, two other scenarios in DCMDHX 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 (JM = WP/AM) for three membrane types are shown
in Fig. 9, in which the increase of JM 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 JM regardless of the thermal conductivity. The curve exhibits an initial sharp
decline and subsequent slow decrease. This can be explained through
the slower increase of JM as a result of more rapidly increasing
recirculating owrates (Fig. 4(A)), which leads to the decrease of ratio
WP/W1,2 (=JM/G 4L/Di,1, Fig. 4(B)) and increase of thermal efciency .
Therefore, the GOR decreases accordingly (Eq. (15)). However, the improvement of permeation ux slows down with further increase of
the owrate leading to a attening 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 JM exceeds 3 kg/(m2 h). Nevertheless, the GOR values on
all three curves drop below 1.0 beyond this point. Thus, for a given
DCMDHX system, high GOR often comes at the cost of low JM, or vice
versa. In addition, it is observed in Fig. 9 that the GOR generally decreases with increasing kM. For instance, at JM = 1.0 kg/m2h, the GOR

G. Guan et al. / Desalination 366 (2015) 8093

89

3.5

80

(A)
6

-1

-1

-1

-1

-1

-1

kM = 0.033 (W m K )

3.0

kM = 0.066 (W m K )

GOR

40

2.0
20

-1

2.5

WP (kg h )

kM = 0.099 (W m K )

GOR

60

1.5

0
0.0

0.5

1.0

1.5

2.0
-2

2.5

3.0

3.5

4.0

(B)

-1

JM (kg m h )

12.5
6.5%

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

( JM) in DCMDHX system at varied thermal conductivities (THX = 0 C and TMD =
50 C, simulated module #1, AM = 100 m2).

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

kM = 0.066 W m1K1; while system with kM of 0.099 W m1K1
drops below 1.5. The results are corroborating with the conclusion
drawn from Fig. 7(A). This makes substantial difference when sustainable 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 recovery efciency and water production, multiple factors can be manipulated for optimizing a DCMDHX 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
resources are available and hence a 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 DCMDHX desalination system
As the industrial implementation of MD has been long discussed, the
scale-up of MD desalination system for 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 Plus owsheet simulations, the DCMDHX system installed
with a series of membrane modules, which have proportionally increased 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 constant 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 efciency (GOR and WP, Fig. 10(A)) and specic module
performance ( and JM, Fig. 10(B)).
As shown in Fig. 10(A), both GOR and WP increase with increasing
membrane area AM. Particularly, the curve of GOR vs. AM exhibits an initial 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 DCMDHX system with constant TMD
and THX largely depends on the feed- and permeate-side water recovery rates, which are identical when operated at equivalent owrates,
i.e., (WP/W1,2) = (WP/W2,1). Thus, a higher water production WP
and hence a higher water recovery rate is achieved at a larger AM,
which leads to a higher GOR value the increase of mathematically
offsets the increase of GOR in the equation. Compared to the steeper
shape of the GOR curve, the water productivity (WP) behaves fairly linearly with increasing AM.

7.5

5.0

5.5%

-1

6.0%

JM (kg h )

10.0

2.5
5.0%

0.0

10

15

20

25

30

35

40

45

AM (m )
Fig. 10. Effect of module scale-up (AM) in DCMDHX system efciency in terms of
(A) heat utilization (GOR) and water production (WP); and module performance in
terms of (B) permeation ux (JM) and water recovery rate () (THX = 0 C and
TMD = 50 C, simulated modules #4#13 in Table 3, kM = 0.066 W m1K1).

Correspondingly, Fig. 10(B) shows that the water recovery rate 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
WP as AM increases, the owrate was varied more signicantly to maintain a constant G and hence leading to a slow increase of the recovery
rate for bigger modules. In contrast, in Fig. 10(B) the JM decreases significantly with increasing membrane area. It was well-understood that this
is mainly due to the decline of averaged transmembrane driving force as
module length and size increase simultaneously [32].
Overall, coupling the system efciency (i.e., GOR, WP in Fig. 10(A))
with module performance (i.e., JM and in Fig. 10(B)) reveals the
scale-up effect in designing a DCMDHX system. Clearly, the increase
on module size and membrane area AM greatly facilitates the improvement of system efciency as a result of higher GOR and WP. However,
the performance of the membrane module, i.e., the permeation ux, deteriorates 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 DCMDHX systems would have higher GORs than the
laboratorial ones.

90

G. Guan et al. / Desalination 366 (2015) 8093

3.6. Effect of heat recovery

In the DCMDHX system, heat recovery unit (HX) is critical for recovering reusable heat to reduce the total heat input. In previous sections, an ideal situation with full heat recovery was considered,
i.e., the additional heat contained in the returning permeate stream
(S2,2 and T2,2, Fig. 1) was fully extracted to the feed stream from the
HX unit (S1,1 and T1,1, Fig. 1) resulting in THX = T2,2 T1,1 = 0. A
threshold of THX, which relates to the characteristics of the heat exchanger and ow conditions, practically exists due to the limited heat
exchanger area and nite heat-transfer rate. In reality, the THX is greater than 0. A greater THX indicates poorer heat recovery in analogy to insufcient area for fully exchanging heat throughout the HX unit. For
instant, at THX = 5 C, the temperature of the feed-side efuent T1,1
will have to be further elevated to approach that of the permeate-side
T2,2 with additional heat-transfer area. Hence, the effect of heat recovery
efciency (THX) 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 THX rises from 0 to 5
Ca total decline of 30% indicating a strong effect of heat recovery
has on the overall heat utilization of the DCMDHX system. To achieve
more efcient 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 DCMDHX desalination system, as explained via the factorial analysis in Section 3.3.1.

(2)

(3)

(4)

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 the context of limited heat resources. An implicit expression of GOR was derived to conveniently correlate the DCMDHX
system efciency in terms of heat utilization using module (unit)
modeling and hence avoid complicated owsheet simulations. Factorial
analysis was conducted to identify the controlling factors for achieving
high GOR in the DCMDHX desalination system. The following conclusions can be drawn:

(5)

(6)

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

the GOR, the most inuential operational factors were identied

5.0

-1

W12=W21=50 (kg h )

as the recirculating owrates of the feed and permeate streams

(W1,2 and W2,1). The rapidly declining trend of water recovery
rate in terms of mass rate G has veried its controlling effect.
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
DCMDHX system, indicative of high heat recovery efciency, is
the low equivalent recirculating owrates.
Less conductive (i.e., low membrane thermal conductivity) and
thicker membranes (i.e., large wall thickness) with high heattransfer 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
efciency was revealed the enhanced MD water production at a
higher thermal efciency leading to higher GOR. The strong correlation of GOR and thermal efciency in the implicit expression
Eq. (15) were testied.
The interplay of heat recovery efciency and water production
in MD was investigated. A trade-off relationship existed between the GOR and water production, i.e., high water productivity was achieved at high recirculating owrates at the cost
of low GOR. Multiple factors including operating owrates,
temperatures, membrane characteristics as well as heat recovery unit (HX) could be manipulated for optimizing a DCMD
HX system. Thus, a compromise must be made and considerations 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.
The scale-up effect of was studied by coupling the DCMDHX
system efciency 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 utilization as a result of high GOR and water production, but deteriorated the membrane capacity leading to low MD driving force and
low water recovery rate.
The concept of non-linear scale-up was proposed for largescale 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
an efcient desalination system and improving the practicability
of MD technology.

-1

W12=W21=500 (kg h )

4.5

-1

W12=W21=5000 (kg h )

4.0

GOR

3.5
3.0
2.5
2.0
1.5
1.0
0

THX

Fig. 11. Effect of heat recovery efciency (THX) on heat utilization (GOR in DCMDHX system) (TMD = 50 C, DCMD module #1).

Nomenclature
AM
CM
cP
Di
Do
G
GOR
Gz
h
kM
JH
JM
L
N
Nu
p
Pr

Membrane area
Membrane distillation coefcient
Specic heat capacity
Inner diameter
Outer diameter
Mass rate in the bulk ow
Gain output ratio
Graetz number
Specic enthalpy of materials
Thermal conductivity of membrane
Heat ux
Permeation ux
Length
Hollow ber numbers
Nusselt number
Pressure
Prandtl number

m2
kg m2 s1 Pa1
kJ kg1 K1
m
m
kg m2 s1
Dimensionless
Dimensionless
kJ kg1
W m1K1
kJ m2 h1
kg m2 h1
m
Dimensionless
Dimensionless
Pa
Dimensionless

G. Guan et al. / Desalination 366 (2015) 8093

qv
qc
Qc
Re
T
v
vm
W
z

Heat ow ux of evaporation
Heat ow ux of conduction
Heat ow of conduction
Reynolds number
Temperature
Velocity
Specic volume
Mass owrate
Axial location

Greek letters

Membrane thickness

Packing density

Mass recovery rate

Averaged thermal efciency of DCMD module

Viscosity

Constant of circumference ratio

Density

Temperature polarization coefcient

hv
Latent heat
T
Temperature difference
Subscripts
0
1
2
HX
MD
P
l
v
W

Inuent of the DCMD module

lumen side in the DCMD module
Shell side in the DCMD module
Heat recovery unit
DCMD module
Production of permeate
Liquid phase
Vapor phase
Wall

91

kJ m2 h1
kJ m2 h1
kJ h1
Dimensionless
K
m s1
m3 mol1
kg h1
m

Acknowledgments

m
Dimensionless
Dimensionless
Dimensionless
Pa s

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

Plus

kg m
Dimensionless
kJ kg1
K

The following funding agencies are gracefully acknowledged for

funding this work: Fundamental Research Funds for the Central Universities, China (2014ZZ0061); Environmental and Water Industry
Programme Ofce (EWI), Singapore project (#0901-IRIS-0203);
and Industry Postdoctoral Fellowship Scheme, Victoria University,
Australia.

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

owsheet simulation for the DCMDHX 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 computing time:
1. For the quick convergence in computation, the permeate-side recirculation 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. The CALCULATION block [52] was used to determine the mass ow of
feed-side reux assigned by the FSLIT1 block [52].

Fig. A.1. Simulation owsheet of DCMDHX system in Aspen Plus.

92

G. Guan et al. / Desalination 366 (2015) 8093

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.

Table A.1
Initial input settings for streams and unit blocks in DCMDHX system with module #1 in
Table 1.
Unit type

Settings

FEED

STREAM

S11

STREAM

S20

STREAM

FSLIT1
FSLIT2
CHILLER
HEATER
HX

FSLIT
FSLIT
HEATER
HEATER
HEATX

PUMP1
DCMD

PUMP
USER2

TEMP = 25. bCN PRES = 200. bkPaN MASS-FLOW = 0.05

MASS-FRAC H2O 0.97/NACL 0.03
TEMP = 30. bCN PRES = 200. bkPaN MASS-FLOW = 0.1
MASS-FRAC H2O 0.97/NACL 0.03
TEMP = 30. bCN PRES = 200. bkPaN MASS-FLOW = 0.1
MASS-FRAC H2O 1.
MASS-FLOW S14 0.9
MASS-FLOW S20 0.9
PARAM TEMP = 30. bCN PRES = 0 bKPAN
PARAM TEMP = 80. bCN PRES = 0 bKPAN
PARAM DELT-COLD = 0.01 CALC-TYPE = DESIGN
AREA-TOL = 1E-005 & MIN-TAPP = 0.0001
U-OPTION = CONSTANT
FEEDS HOT = S22 COLD = S10
PRODUCTS HOT = S23 COLD = S11
PARAM PRES = 200. bkPaN
SUBROUTINE DCMD
PARAM NINT = 2 NREAL = 14 NCHAR = 14
INT VALUE-LIST = & 2 &;OPT 15089;NTUBES
REAL VALUE-LIST = & 2.153 &;LEN 0.00098 &;IDT
0.00146 &;ODT 0.216 &;IDS 0.25 &;ODS
0.066 &;KM 3.8E-007 &;CM

References
[1] K.W. Lawson, D.R. Lloyd, Membrane distillation, J. Membr. Sci. 124 (1997) 125.
[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) 429.
[4] C. Charcosset, A review of membrane processes and renewable energies for desalination, Desalination 245 (2009) 214231.
[5] E. Curcio, E. Drioli, Membrane distillation and related operations a review, Sep.
Purif. Rev. 34 (2005) 3586.
[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) 2740.
[7] P.A. Hogan, Sudjito, A.G. Fane, G.L. Morrison, Desalination by solar heated membrane distillation, Desalination 81 (1991) 8190.
[8] J. Koschikowski, M. Wieghaus, M. Rommel, Solar thermal-driven desalination plants
based on membrane distillation, Desalination 156 (2003) 295304.
[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)
27662780.
[10] M. Hightower, S.A. Pierce, The energy challenge, Nature 452 (2008) 285286.
[11] L. Song, B. Li, K. Sirkar, J. Gilron, Direct contact membrane distillation-based desalination: novel membranes, devices, larger-scale studies, and a model, Ind. Eng.
Chem. Res. 46 (2007) 23072323.
[12] P. Wang, M.M. Teoh, T.-S. Chung, Morphological architecture of dual-layer hollow
ber for membrane distillation with higher desalination performance, Water Res.
45 (2011) 54895500.
[13] M. Khayet, T. Matsuura, J.I. Mengual, Porous hydrophobic/hydrophilic composite
membranes: estimation of the hydrophobic layer thickness, J. Membr. Sci. 266
(2005) 6879.
[14] B. Li, K.K. Sirkar, Novel membrane and device for vacuum membrane distillationbased desalination process, J. Membr. Sci. 257 (2005) 6075.
[15] M. Khayet, T. Matsuura, J.I. Mengual, M. Qtaishat, Design of novel direct contact
membrane distillation membranes, Desalination 192 (2006) 105111.
[16] M. Qtaishat, D. Rana, M. Khayet, T. Matsuura, Preparation and characterization of
novel hydrophobic/hydrophilic polyetherimide composite membranes for desalination by direct contact membrane distillation, J. Membr. Sci. 327 (2009) 264273.
[17] M. Su, M.M. Teoh, K.Y. Wang, J. Su, T.-S. Chung, Effect of inner-layer thermal conductivity on ux enhancement of dual-layer hollow ber membranes in direct contact
membrane distillation, J. Membr. Sci. 364 (2010) 278289.
[18] Y. Liao, R. Wang, A.G. Fane, Engineering superhydrophobic surface on poly(vinylidene
uoride) nanober membranes for direct contact membrane distillation, J. Membr.
Sci. 440 (2013) 7787.

[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)
53005309.
[21] T.Y. Cath, V.D. Adams, A.E. Childress, Experimental study of desalination using direct
contact membrane distillation: a new approach to ux enhancement, J. Membr. Sci.
228 (2004) 516.
[22] L. Martinez, J.M. Rodriquez-Maroto, Effects of membrane and module design improvements on ux in direct contact membrane distillation, Desalination 205
(2007) 97103.
[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) 371379.
[24] X. Yang, R. Wang, L. Shi, A.G. Fane, M. Debowski, Performance improvement of PVDF
hollow ber-based membrane distillation process, J. Membr. Sci. 369 (2011)
437447.
[25] K.W. Lawson, D.R. Lloyd, Membrane distillation. II. Direct contact MD, J. Membr. Sci.
120 (1996) 123133.
[26] K. He, H.J. Hwang, M.W. Woo, I.S. Moon, Production of drinking water from saline
water by direct contact membrane distillation (DCMD), J. Ind. Eng. Chem. 17
(2011) 4148.
[27] X. Yang, R. Wang, A.G. Fane, Novel designs for improving the performance of hollow
ber membrane distillation modules, J. Membr. Sci. 384 (2011) 5262.
[28] X. Yang, E.O. Fridjonsson, M.L. Johns, R. Wang, A.G. Fane, A non-invasive study
of ow dynamics in membrane distillation hollow ber modules using loweld nuclear magnetic resonance imaging (MRI), J. Membr. Sci. 451 (2014)
4654.
[29] G. Chen, X. Yang, R. Wang, A.G. Fane, Performance enhancement and scaling control
with gas bubbling in direct contact membrane distillation, Desalination 308 (2013)
4755.
[30] G. Chen, X. Yang, Y. Lu, R. Wang, A.G. Fane, Heat transfer intensication and scaling
mitigation in bubbling-enhanced membrane distillation for brine concentration, J.
Membr. Sci. 470 (2014) 6069.
[31] G. Chen, Y. Lu, X. Yang, R. Wang, A.G. Fane, Quantitative study on crystallizationinduced scaling in high-concentration direct-contact membrane distillation, Ind.
Eng. Chem. Res. 53 (2014) 1565615666.
[32] X. Yang, R. Wang, A.G. Fane, C. Tang, I.G. Wenten, Membrane module design and dynamic shear-induced techniques to enhance liquid separation by hollow ber modules: a review, Desalin. Water Treat. 51 (2013) 36043627.
[33] A.G. Fane, R.W. Schoeld, C.J.D. Fell, The efcient use of energy in membrane distillation, Desalination 64 (1987) 231243.
[34] S. Al-Obaidani, E. Curcio, F. Macedonio, G. Di Proo, H. Al-Hinai, E. Drioli, Potential of
membrane distillation in seawater desalination: thermal efciency, sensitivity study
and cost estimation, J. Membr. Sci. 323 (2008) 8598.
[35] H.Y. Lee, F. He, L.M. Song, J. Gilron, K.K. Sirkar, Desalination with a cascade of crossow hollow ber membrane distillation devices integrated with a heat exchanger,
AICHE J. 57 (2011) 17801795.
[36] E.K. Summers, H.A. Arafat, J.H. Lienhard, Energy efciency comparison of singlestage membrane distillation (MD) desalination cycles in different congurations,
Desalination 290 (2012) 5466.
[37] R.W. Schoeld, A.G. Fane, C.J.D. Fell, The efcient use of energy in membrane distillation, Desalination 64 (1987) 231243.
[38] S. Bouguecha, B. Hamrouni, M. Dhahbi, Small scale desalination pilots powered by
renewable energy sources: case studies, Desalination 183 (2005) 151165.
[39] J. Gilron, L. Song, K.K. Sirkar, Design for cascade of crossow direct contact membrane distillation, Ind. Eng. Chem. Res. 46 (2007) 23242334.
[40] G. Zuo, R. Wang, R. Field, A.G. Fane, Energy efciency evaluation and economic analyses of direct contact membrane distillation system using Aspen Plus, Desalination
283 (2011) 237244.
[41] M. Khayet, Solar desalination by membrane distillation: Dispersion in energy consumption analysis and water production costs (a review), Desalination 308
(2013) 89101.
[42] S. Lin, N.Y. Yip, M. Elimelech, Direct contact membrane distillation with heat recovery: thermodynamic insights from module scale modeling, J. Membr. Sci. 453
(2014) 498515.
[43] R.B. Saffarini, E.K. Summers, H.A. Arafat, J.H. Lienhard V, Technical evaluation of
stand-alone solar powered membrane distillation systems, Desalination 286
(2012) 332341.
[44] M. Khayet, Membranes and theoretical modeling of membrane distillation: a review, Adv. Colloid Interf. Sci. 164 (2011) 5688.
[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) 758769.
[46] H. Yu, X. Yang, R. Wang, A.G. Fane, Numerical simulation of heat and mass transfer in
direct membrane distillation in a hollow ber module with laminar ow, J. Membr.
Sci. 384 (2011) 107116.
[47] H. Yu, X. Yang, R. Wang, A.G. Fane, Analysis of heat and mass transfer by CFD for performance enhancement in direct contact membrane distillation, J. Membr. Sci. 405
(2012) 3847.
[48] G. Guan, X. Yang, R. Wang, R. Field, A.G. Fane, Evaluation of hollow ber-based direct
contact and vacuum membrane distillation systems using aspen process simulation,
J. Membr. Sci. 464 (2014) 127139.
[49] G. Guan, R. Wang, F. Wicaksana, X. Yang, A.G. Fane, Analysis of membrane distillation crystallization system for high salinity brine treatment with zero discharge using Aspen owsheet simulation, Ind. Eng. Chem. Res. 51 (2012)
1340513413.

G. Guan et al. / Desalination 366 (2015) 8093

[50] L. Martinez-Diez, M.I. Vazquez-Gonzalez, Temperature and concentration polarization in membrane distillation of aqueous salt solutions, J. Membr. Sci. 156 (1999)
265273.
[51] R.H. Perry, D.W. Green, J.O. Maloney, Perry's Chemical Engineers' Handbook,
McGraw-Hill Companies, Inc., New York, 1997.

93

[52] AspenTech, Aspen Plus V7.3 User Guide, Aspen Technology, Inc., Cambridge, MA.
USA, 2011.
[53] A.M. Alklaibi, N. Lior, Heat and mass transfer resistance analysis of membrane distillation, J. Membr. Sci. 282 (2006) 362369.