Desalination 338 (2014) 26–32

Contents lists available at ScienceDirect

Desalination

j ourna l homepage: www.e lsev ie r .com/ locate /desa l

Performance improvement on distillate flux of countercurrent-flowdirect contact membrane distillation systems

Chii-Dong Ho ⁎, Cheng-Hao Huang, Feng-Chi Tsai, Wei-Ting ChenEnergy and Opto-Electronic Materials Research Center, Department of Chemical and Materials Engineering, Tamkang University, Tamsui, New Taipei 251, Taiwan

H I G H L I G H T S

• A roughened-surface device of countercurrent-flow DCMD was developed theoretically.• Experimental study indicated its feasibility with 42% of performance enhancement.• The pure water productivity with the expense of energy consumption is discussed.• A heat-transfer coefficient correlation of roughened-surface channels is obtained.

⁎ Corresponding author. Tel.: +886 2 2621 5656; fax:E-mail address: cdho@mail.tku.edu.tw (C.-D. Ho).

0011-9164/$ – see front matter © 2014 Elsevier B.V. All rihttp://dx.doi.org/10.1016/j.desal.2014.01.023

a b s t r a c t

a r t i c l e i n f o

Article history:Received 5 November 2013Received in revised form 23 January 2014Accepted 26 January 2014Available online 15 February 2014

Keywords:Countercurrent flowDirect contact membrane distillationTemperature polarizationMass flux incrementEddy promoter

The theoretical predictions of pure water productivity in a parallel-plate direct contact membrane distillation(DCMD)module using roughened-surface flow channel for enhancing heat transfer enhancementwere obtainedunder countercurrent-flow operations. The device performance improvements with increasing the pure waterproductivity in saline water desalination were achieved as compared to the concurrent-flow operation. Theroughened surface was fabricated using siphonic-blasting with aluminum oxide (Al2O3) sand grains and arcspraying for Ni film coating, and the experimental data were correlated in a simplified expression to predictthe heat transfer coefficient for the DCMD device. The pure water productivity and temperature distributionsof both hot and cold feed streams are represented graphically with the fluid flow rate and inlet saline tempera-ture as parameters. Both flow-pattern and roughened-surface effects have demonstrated the technical feasibilityin the roughened-surface channel device and up to 42.11% of the device performance enhancementwas achievedfor the countercurrent-flowDCMD system. The influences of operation and design parameters on the pure waterproductivity with the expense of energy consumption are also discussed.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

Membrane distillation (MD) has been recognized as an economicallyfeasible technology for desalination processes [1,2] in its simplicity andthe low energy demand. The rejection of dissolved solids is nearly 100%[3]. The direct contact membrane distillation (DCMD) device in thisstudy is a MD system for which hot saline and cold liquids directly con-tact both membrane surfaces with a small temperature-driving force inproviding a phase-change process, which results in a vapor pressure dif-ference in between to allowonly the vapor transport across a hydropho-bic porous membrane where water is the permeating flux. Otherapplication of membrane-based separation processes includes juiceconcentration and waste water treatment [4–7].

The membrane distillation process analysis of countercurrent-flowoperations is to develop a mathematical model considering both heat

+886 2 2620 9887.

ghts reserved.

and mass transfer mechanisms for evolving a heat transfer coefficientcorrelation interpolated by experimental data. The permeation rate ofpure water in direct contact membrane distillation DCMD is governedby the heat transfer resistances among the hot liquid, membrane, andcold liquid, called temperature polarization [8,9], as well as the masstransfer resistance in the membrane. Attempts to reduce the effect oftemperature polarization were made implementing eddy promoters[10,11] to improve the heat andmass transfer rate by inserting channelspacers [12,13]. The new design of roughened-surface channels [14]was fabricated using siphonic-blasting with aluminum oxide (Al2O3)sand grains and arc spraying for Ni film coating by arc spraying processin aiming to promote the eddy turbulence of the hot saline feed stream.The arc spraying for Ni layer on aluminum oxide (Al2O3) has gained keyimportance in structural applications because of corrosion resistance[15].

This study investigates the heat andmass transfer of the countercur-rent flow in DCMD processes with the eddy promoter to achieve theheat-transfer correlation equation incorporated with the experimentalruns, and the results show that a good agreement is obtained between

27C.-D. Ho et al. / Desalination 338 (2014) 26–32

the experimental results and theoretical predictions. Once the tempera-ture distributions and the amount of vapor flux across the membraneare calculated, the correlated equation is expressed as a function of rel-ative roughness and can be used for predicting the heat transfer coeffi-cient under operating the device roughened-surface channels. Themassflux enhancement accompanyingwith the penalty of the friction loss in-crement due to employing roughened-surface channels was correlatedexperimentally [16] and the extra power consumption was calculatedin terms of the relative surface roughness. The improvements of deviceperformance were considerably achieved under the countercurrent-flow operation roughened-surface channels as compared. The influ-ences of operation and design parameters on the pure water productiv-ity improvement are also discussed.

2. Theoretical model

Fig. 1 shows a DCMDmodulewith inserting a hydrophobicmicropo-rous membrane of thickness δm into a parallel conduit of width B andlength L, and with the same thickness d for both hot and cold feedstreams to conduct a double-flow countercurrent operation. The distil-late flux of pure water production is collected by an overflow tankinto a beaker as the distillation process proceeds and measured usinga timer and weighted on an electronic balance. The thermal boundarylayers on both liquid streams build up a temperature differences be-tween bulkfluid andmembrane surfaces. Thewater vaporization occurson themembrane surface in the hot saline stream and then the vapor istransferred through themembrane pores with the condensation of per-meate on the other side of themembrane surface in the cold feed streamthereafter.

The effective thermal conductivity of the membrane can be deter-mined by taking account of the contributions on the gas inside themembrane and solid part of the membrane [17]. The energy balanceequations among the three heat fluxes and for the bulk fluids in Fig. 1give

q ¼ qh ¼ hh Th−T1ð Þ ð1Þ

¼ qc ¼ hc T2−Tcð Þ ð2Þ

¼ qm ¼ N}λþ kmδm

T1−T2ð Þ ð3Þ

Membrane

Feed side

Permeate side

Th

Tc

T1

T2Boundary Polarization

Layers

Hot stream

Cold stream

qh

qm

qc

N”

Fig. 1. Heat and mass transfer in countercurrent-flow DCMD systems.

dTh

dz¼ −Wq

Qρh Cphð4Þ

dTc

dz¼ −Wq

QρcCpc: ð5Þ

The heat loss associated with the vaporization process due to con-ductive heat transport across the membrane has been considered asthe second term in the right-hand side of Eq. (3) and km is the effectivethermal conductivity of microporous membrane, and was estimated bythe combination of the gas and solid conductivities [18]

km ¼ εkg þ 1−εð Þks ð6Þ

In general, themass flux of the condensate water was expressed using amembrane permeation coefficient (Cm) and the across-membrane satu-ration vapor pressure difference (ΔP)

N″ ¼ cmΔP ¼ cm Psat1 T1ð Þ−Psat

2 T2ð Þh i

ð7Þ

where P1sat(T1) and P2

sat(T2) are the saturated pressure of wateron the membrane surfaces in hot and cold streams, respectively.The saturated pressure of water on the membrane surface inthe hot stream was correlated with water activity coefficient aw =1 − 0.5xNaCl − 10xNaCl2 as follows:

Psat1 ¼ ywP ¼ xwawP

satw : ð8Þ

There are three essentialmembrane coefficientmodels, the Knudsendiffusion model, Poiseuille flow model, and molecular diffusion modelthat can be used to describe the mass flux across the hydrophobic po-rous membrane. Many researchers used the expressions of interfacialtemperature in terms of bulk temperature with specified empirical cor-relations of heat-transfer coefficients [19,20] due to the uncertainty ofmicroporous membrane morphology in the molecular diffusion model(say the effective gas diffusivity) leading to inaccuracy calculation ofthemass transfer [21]. Moreover, the trans-membrane temperature dif-ference creates the pressure difference across membrane owing to theexistence of saturated pressure difference across the membrane,resulting in Poiseuille flow occurrence if the mean free path is muchsmaller than the pore size. The membrane coefficient including the tor-tuosity (τ) of the porous hydrophobic PTFEmembranewas proposed bySchofield et al. [21–23] by inspection of the Knudsen diffusion model(due to the larger mean free path of vapor molecules than the mem-brane pore size) and Poiseuille flow model to describe the watervapor flux through a deaerated microporous membrane in a semi-empirical equation, this is

cm ¼ ck þ cp ¼ 1:064ε rpτδm

Mw

RTm

� �1=2þ 0:125

ε r2pτδm

MwPm

ηvRTm: ð9Þ

Therefore, the combination of Knudsen diffusion and Poiseuille flowmodelswas proposed in the present study and validated by the theoret-ical predictions as compared to experimental runs.

The mass flux and the temperature distributions of hot stream, coldstream, and membrane interfaces along the flow direction wereachieved using the finite difference techniques of the Runge–Kuttamethod in solving Eqs. (4) and (5), as illustrated in Fig. 2.

The value of the standard deviation calculated for the previous study[24] indicates that the best agreement between the experimental per-meate flux and calculated permeate flux was achieved with four timeshigher in comparisonwith other correlations includingGrashof number[25] when Eq. (10) was used for the determination of the convectiveheat-transfer coefficients in the countercurrent MD model. The heatfluxes transferred across the thermal boundary layers to themembrane

Tc,in

Th,in

Tc,out

Th,out

z

Th,z

Tc,z

Th,z+dz

Tc,z+dz

z z+dz

hh

hc km N”

T1

T2

Hot stream

Cold stream

ε

Fig. 2.Modeling countercurrent DCMD systems.

28 C.-D. Ho et al. / Desalination 338 (2014) 26–32

surfaces were estimated using the following correlation for the Nusseltnumber [26] with laminar flow in calculating the convective heat trans-fer coefficients as follows:

Nulam ¼ 4:36þ 0:036RePr De=Lð Þ1þ 0:011 RePr De=Lð Þð Þ0:8 : ð10Þ

A modified factor (ar) for finding the Nusselt number in turbulentflow is proposed for correlating the turbulence effect with the aid ofthe experimental data of countercurrent-flow DCMD modules withroughened-surface channels [12]

Nur ¼ αrNulam ð11Þ

where ar is the modified factor depending on relative roughness.Eq. (11) was used to correlate the heat transfer correlation forroughened-surface channels from experimental results for the modulewith roughened-surface channels. On the other hand, a pure water pro-ductivity increment, IN, is defined by calculating the mass fluxes withroughened-surface channels (Nr

"(counter)) under countercurrent-flowoperations based on the smooth-surface channel (Ns

"(con)) underconcurrent-flow operations. Noted that the smooth surface channelused in this study has a relative surface roughness (εr/d = 0.0035).

IN counterð Þ ¼ N″r counterð Þ−N″

s conð ÞN″

s conð Þ : ð12Þ

(A)

(B) (C)

(D)

(E)

(F)(G)

(H)

(I)

(D)

(E)

Fig. 3. Experimental setup of the c

Similarly, a pure water productivity increment, IN(con), inconcurrent-flow operations is defined in the previous work [14] as

IN conð Þ ¼ N″r conð Þ−N″

s conð ÞN″

s conð Þ : ð13Þ

3. Experiments

An experimental module of the parallel-plate countercurrent-flowDCMD system was shown in Fig. 3. Two channels were fabricated hori-zontally and well insulated outside with inserting a hydrophilic com-posite membrane PTFE (polytetrafluoroethylene) from ADVANTECwith a nominal pore size of 0.1 μm, a porosity of 0.72 and a thicknessof 130 μm into the parallel-plate module. A replaceable aluminumplatewas implemented in the hot fluid channel to conduct experimentsfor variations of relative surface roughness. The spacers made of silicondioxide (silicon gel) were inserted to create a 2 mm thickness in fluidflow channels. Nets made by threads were utilized to support the bothsides of membrane surface for preventing from vibration. The workingdimensions of each channel are L = 0.29 m, W = 0.21 m, and d =0.2 cm. The aluminum plate was fabricated by siphonic-blasting tospray a nickel film coating on the alumina plate. The saline water of3.5 wt.% NaCl was prepared using distilled water and monitored theconductance to be less than 2 μs/cm during the whole experimentalruns. The experimental runswere carried outwith three relative surface

(D) Pump

(E) Flow meter

(B) Hot fluid Thermostat

(I) Temperature indicator

(F) Overflow barrel

(G) Beaker

(H) Electronic balance

(A) Membrane distillation module

(C) Cold fluid Thermostat

ountercurrent DCMD system.

29C.-D. Ho et al. / Desalination 338 (2014) 26–32

roughness of the aluminumplates, i.e. εr=14 μm (smooth surface), 240and 440 μm, four inlet hot fluid temperatures (30, 40, 50, 60 °C),and four fluid flow rates (0.3, 0.5, 0.7, 0.9 L/min). The inlet cold fluidtemperature was kept at 25 °C. The relative surface roughness of the ar-tificially roughened aluminum plate was measured using Elcometer224S for its maximum roughness and Mitutoyo Surf-Test 301 for aver-age roughness. A quantitative agreement is achieved between experi-mental data and theoretical predictions in countercurrent-flow DCMDsystems.

4. Results and discussion

The roughened-surface DCMD module is a physical process involv-ing both heat and mass transfers. The flow sheet for simply expressingin calculating the convective heat transfer coefficients aswell as Nusseltnumbers Nuexp. is shown in Fig. 4. A simplified relationship betweenNur(the Nusselt number for roughened-surface channel) and Nulam (theNusselt number for laminar flow in smooth channel) with the aid ofthe experimental data from Eq. (10) was expressed in Eq. (11). There-fore, The correlated equation to express the heat transfer coefficientfactor (ar) in terms of the relative surface roughness (εr/d), as shownin Eq. (11), which is obtained from curve-fitting with the deviation is

Fig. 4. Flowdiagramof the algorithm for prediction of convective heat-transfer coefficientsof hot fluid.

within ±10% and indicated in Fig. 5 with the correlation equation inEq. (14) to find the theoretical prediction of Nusselt numbers Nutheo

αr ¼Nur

Nulam¼ f

εrd

� �¼ a exp

εrd

� �b ¼ 0:9394 expεrd

� �6:82: ð14Þ

The theoretical hydraulic frictional loss increment due to theroughened-surface channel [16] was estimated using friction factor byBernoulli equation for both hot and cold channels as follows:

ℓwf ;i ¼2 f F;ivi

2LDei

; i ¼ h; c ð15Þ

inwhich, the hydraulicmean diameter for Reynolds number calculationwith surface roughness in hot feed stream

Deh ¼ 4 d−εð ÞW½ �2 d−εð Þ þW½ � : ð16Þ

The total energy consumption and friction loss [27] are defined as

Pr ¼ Ph þ Pc ¼ Qρhℓwf ;h þ Qρcℓwf ;c: ð17Þ

The friction factor has been given by Kakac et al. [28] for laminarflow in smooth rectangular channels, that is

f F ¼ CRe

ð18Þ

C ¼ 24 1−1:3553αþ1:9467α2−1:7012α3 þ0:9564α4−0:2537α5� �

;α

¼ d=W:

ð19Þ

4 6 8 10 12

4

6

8

10

12

0%

-10%

0.1413

0.0686

0.0035

r /dSymbol

Nuexp.

Nu ca

l.

+10%

ε

Fig. 5. Comparison of estimated and experimental Nusselt numbers.

0.3 0.4 0.5 0.6 0.7 0.8 0.90.0

0.4

0.8

1.2

1.6

2.0

2.4

2.8

3.2

N"×

103

(kg/

m2 s

)

Qh and Qc(L/min)

Th,in

= 40 oC

Th,in

= 60 oC

Tc,in

=25 oC, NaCl solution 3.5 wt%

Hot stream

Cold stream

εr /d Theo. Exp.

0.06860.1413

0.0035

Fig. 6. Effects of hot fluid inlet temperature, fluid flow rate, and relative surface roughnesson pure water flux.

0.3 0.4 0.5 0.6 0.7 0.8 0.90.0

0.4

0.8

1.2

1.6

2.0

2.4

2.8

3.2

ε r /d=0.1413

N"×

103

(kg/

m2 s

)

Qh and Qc(L/min)

Th,in

= 40 oC

Th,in

= 60 oC

Tc,in

=25 oC, NaCl solution 3.5 wt%

concurrent countercurrent

Fig. 7. Pure water flux for saline water versus as the fluid flow rate for both operations.

30 C.-D. Ho et al. / Desalination 338 (2014) 26–32

Therefore, the power consumption increment, IP, of roughened-surface channel Pr is calculated based on the power consumption insmooth surface channel Ps as follows:

Ip ¼ Pr−Ps

Ps: ð20Þ

The deviation analysis of the theoretical predictions of pure waterproductivity is defined and determined comparing to the experimentaldata by

E ¼ 1N

XNi¼1

ω̂ j−ω j

��� ���ω̂ j

ð21Þ

the deviation E ≅ 1.08 × 10−1 for all the experimental data, where ω̂ j

and ωj are the theoretical prediction and the experimental measure-ment, andN is thenumber of experimental data. Fairly good agreementsare obtained between the theoretical predictions of the pure water fluxwith the aid of the heat transfer coefficient factor in Eq. (14) and the ex-perimental data, as confirmed by Fig. 6. Figure 6 presents the graphicalrepresentations of the theoretical predictions and experimental runsof pure water fluxes by using the device with roughened-surfacechannels in countercurrent-flow operations. The fluid flow rates in theroughened-surface channel enhance the fluid-side heat transfer coeffi-cients, and thus, yield the saturation vapor pressure enlargement, espe-cially for operating at a higher inlet hot fluid temperature. The purewater flux increases with increasing the fluid flow rates owing to

Table 1Effects of operation conditions and roughness on productivity increase and temperature polari

Th,in (°C) Q (L/min) 0.0035 0.0686

Ntheo.″ × 10−3

(kg/(m2 s))TPCa Ntheo.

″ × 10(kg/(m2 s)

Con Counter Con Counter Con

40 0.3 0.349 0.384 0.278 0.302 0.3820.5 0.445 0.464 0.285 0.305 0.5050.7 0.503 0.517 0.291 0.306 0.5800.9 0.542 0.553 0.296 0.310 0.633

60 0.3 1.07 1.18 0.238 0.272 1.190.5 1.38 1.45 0.243 0.276 1.600.7 1.57 1.62 0.251 0.280 1.870.9 1.71 1.74 0.257 0.284 2.06

a TPC data is the average value of the entire DCMDmodule.

strengthening the convective heat-transfer coefficient and inlet hotstream temperature under a given cold feed stream temperature incountercurrent-flow DCMD systems, as inferred from Fig. 6. Restated,the reduction of temperature polarization with increasing the fluidflow rates is accomplished in thinning both the velocity boundary andthermal boundary layers, resulting in a larger temperature drivingforce, and hence, a higher trans-membrane pressure difference acrossthe membrane surfaces.

The pure water productivity obtained in both concurrent-flow andcountercurrent-flow devices was presented in Fig. 7 for comparisons.It is seen from Fig. 7 that the pure water productivity in operating thecountercurrent-flow device is higher than that in the concurrent-flowdevice. The higher permeation flux of pure water was obtained whenthe countercurrent-flow arrangement was used, as confirmed by Fig. 6in this study and the same conclusion was also found by the previousstudy [29]. The extent of temperature polarization coefficient improve-ment (TPC=(T1− T2)/(Th− Tc)) in the countercurrent-flow operationis higher than that of concurrent-flow one due to relative surface rough-ness in reducing temperature gradients between the bulk stream andmembrane surface, resulting in the temperature driving-force incre-ment on both cold and hot streams, and thus, pure water flux is in-creased accordingly. The pure water production and temperaturepolarization coefficient improvement for both concurrent- andcountercurrent-flow operations due to relative surface roughness wasillustrated with respect to the smooth channel (εr/d = 0.0035) arelisted in Table 1, while the pure water productivity increments wereshown in Table 2. The pure water productivity increment (IN(counter))in countercurrent-flow operations can be increased up to 42.11%, andTPC still increases with increasing relative surface roughness and inlet

zation.

0.1413

−3

)TPCa Ntheo.

″ × 10−3

(kg/(m2 s))TPCa

Counter Con Counter Con Counter Con Counter

0.431 0.300 0.344 0.405 0.466 0.361 0.3700.536 0.303 0.347 0.549 0.589 0.368 0.3820.602 0.308 0.348 0.640 0.670 0.370 0.3860.650 0.312 0.350 0.704 0.729 0.376 0.3901.35 0.272 0.306 1.27 1.48 0.297 0.3271.71 0.276 0.309 1.77 1.92 0.303 0.3471.95 0.281 0.310 2.10 2.21 0.310 0.3522.11 0.286 0.317 2.34 2.43 0.315 0.353

Table 2The improvements of device performance.

εr/d Q (L/min) IN (con) (%) IN (counter) (%) IE (%)

Th,in (°C)

40 60 40 60 40 60

0.0686 0.3 9.46 11.21 22.05 26.17 11.50 13.450.5 13.48 15.94 25.50 26.60 10.59 9.190.7 15.31 19.11 25.60 28.50 8.92 7.880.9 16.79 20.47 26.50 29.50 8.83 7.45

0.1413 0.3 16.05 18.69 31.72 38.32 15.06 16.540.5 23.37 28.26 32.36 39.13 7.28 8.470.7 27.24 33.76 33.20 40.76 4.69 5.240.9 29.89 36.84 34.50 42.11 3.55 3.85

0.3 0.4 0.5 0.6 0.7 0.8 0.90.3

0.6

0.9

1.2

concurrent countercurrent

ε r /d=0.1413

Tc,in

=25 oC, NaCl solution 3.5 wt%

Th,in

= 40 oC

Th,in

= 60 oC

I N/I

P

Qh and Qc(L/min)

Fig. 9. The effect hot fluid inlet temperatures on IN/IP for both flow operations.

31C.-D. Ho et al. / Desalination 338 (2014) 26–32

hot stream temperature as well. Moreover, the effects of inlet hot fluidtemperatures and relative surface roughness on pure water flux en-hancement are also depicted in Fig. 8. Considering both the purewater productivity increment and frictional energy consumption IN/IPin making the economic viewpoint are presented in Fig. 9. The valueof IN/IP increaseswith increasing thefluidflow rate, and there exists eco-nomic feasibility in operating countercurrent-flowoperations, as shownin Fig. 9. Summarizing the effects of relative surface roughness, one canconclude that increasing the relative surface roughness enhances thepurewater productivity but at the expense of a higher energy consump-tion. Consequently, some optimal selection of relative roughness shouldbe suitably adjusted.

The enhancement in pure water flux IE by the effect of using thecountercurrent-flow operation may be illustrated, based on a device ofthe same working dimensions but in concurrent-flow operations, as:

IE ¼ N″r counterð Þ−N″

r conð ÞN″

r conð Þ ð22Þ

Eq. (22) may be rewritten using Eqs. (12) and (13) as

IE ¼N″

r counterð Þ−N″s conð Þ

� �− N″

r conð Þ−N″s conð Þ

� �N″

s conð Þ

24

35 N″

s conð ÞN″

r conð Þ

!

¼ IN counterð Þ−IN conð Þ½ � N″s conð Þ

N″r conð Þ

!¼ IN counterð Þ−IN conð Þ

1þ IN conð Þ :

ð24Þ

Some values for IE were calculated from Table 1, with the calculatedresults listed in Table 2. The purewater productivity increments in bothflow operations IN(con) and IN(counter) increase with increasing the

0.3 0.4 0.5 0.6 0.7 0.8 0.915

20

25

30

35

40εr /d=0.1413

Th,in

= 40 oC

Th,in

= 60 oC

Tc,in

=25 oC, NaCl solution 3.5 wt%

concurrent countercurrent

I N(%

)

Qh and Qc(L/min)

Fig. 8. The pure water productivity increment for salinewater versus as the fluid flow ratefor both flow operations.

fluidflow rate, inlet hot stream temperature, and relative surface rough-ness. As shown in Table 2 further pure water productivity increment IEin the roughened-surface device under countercurrent-flow operationsas compared to the concurrent-flow operation increases with the inlethot stream temperature and relative surface roughness but decreaseswith fluid flow rate.

5. Conclusions

Investigation of the feasibility and the performance of artificialroughened surface for enhancing the productivity of DCMD module incountercurrent-flow operations was obtained in both theoreticalmodeling and experimental results. There are many operation and de-sign parameters such as the fluid flow rate and inlet hot stream temper-ature under the countercurrent-flow operation that may affect thedevice performance in the direct contact membrane distillation mod-ules. It is also found in Figs. 6 and 7, the pure water flux increaseswith increasing the fluid flow rate and inlet hot stream temperature.Furthermore, the pure water flux obtained in the countercurrent-flowdevice is larger than that in the concurrent-flow device, as confirmedin Figs. 7 and 8 as well as in Tables 1 and 2. The roughened surface fab-ricated by siphonic-blasting with aluminum oxide (Al2O3) sand grainshas been demonstrated to be capable of providing up to 42.11% ofwater production increment under the countercurrent-flow DCMD op-eration. The roughened-surface channel of the present devicewas oper-ated under countercurrent-flow operation for further improvedperformance. Further enhancement in pure water flux IE based on theconcurrent-flow operation reaches 16.54% for εr/d = 0.1413, Th,in =60 °C, and Q = 0.3 L/min as indicated from Table 2. Further enhance-ment in pure water flux IE in operating the countercurrent-flow opera-tion decreases when the fluid flow rate and/or the inlet hot streamtemperature and relative surface roughness increase. Therefore, operat-ing countercurrent-flow device is ineffective when the performance isoperated under the larger fluid flow rate.

Nomenclatureaw water activity in NaCl solutionck membrane coefficient based on the Knudsen diffusion model

(kg/(m2 Pa s))cm membrane permeation coefficient (kg/(m2 Pa s))cp membrane coefficient based on the Poiseuille flowmodel

(kg/(m2 Pa s))CP heat capacity (J/(kg K))d channel height (m)De equivalent hydraulic diameter of channel (m)

m

32 C.-D. Ho et al. / Desalination 338 (2014) 26–32

E deviation between theoretical prediction and experimentalmeasurement

fF friction factorh convective heat-transfer coefficient (W/m2 K)IE pure water productivity increment based on concurrent

device, defined by Eq. (22)IN(con) pure water productivity increment of concurrent device,

defined by Eq. (13)IN(counter) pure water productivity increment of countercurrent

device, defined by Eq. (12)IP energy consumption increase factor, defined by Eq. (20)kf thermal conductivity of fluid (W/m K)kg thermal conductivity of gas (W/m K)km thermal conductivity of membrane (W/m K)ks thermal conductivity of solid membrane (W/m K)L channel length (m)ℓwf friction loss (J/kg)Mw molecular weight of water (kg/mol)m molality of NaCl in NaCl solution˙ mass flow rate (kg/s)N″ pure water flux (kg/(m2 s))Nu Nusselt numberP hydraulic dissipate energy (W)Psat Saturation vapor pressure (Pa)Q volumetric flow rate (m3/s)R gas constant (J/(mol K))Re Reynolds numberrp membrane pore radius (m)T temperature (°C)TPC temperature polarization coefficientv average velocity (m/s)W width of channel (m)xw liquid mole fraction of waterxNaCl mole fraction of NaCl in saline solutionyw vapor mole fraction of waterz axial coordinate along the flow direction (m).

Greek lettersα dimensionless thickness, defined in Eq. (18)αr heat transfer coefficient correction factorδm thickness of membrane (m)ε membrane porosityεr relative roughnessηv gas viscosity ((Ns)/m2)λ latent heat of water (J/kg)μ viscosity ((Ns/m2))ρ density (kg/m3)τ 1/ε, membrane tortuosity factor

Subscripts1 membrane surface on hot fluid side2 membrane surface on cold fluid sideh hot fluidc cold fluidcal. calculated value during iteration in Fig. 4exp. experimental runslam laminarin inletm mean temperature

theo. theoretical predictionsr rough surfaces smooth surface.

Acknowledgment

The authors thank the National Science Council of the Republic ofChina and Tamkang University for their financial support.

References

[1] K.W. Lawson, D.R. Lloyd, Membrane distillation, J. Membr. Sci. 124 (1997) 1–25.[2] M. Gryta, M. Tomaszewska, J. Grzechulska, A.W. Morawski, Membrane distillation of

NaCl solution containing natural organic matter, J. Membr. Sci. 181 (2001) 279–287.[3] K.W. Lawson, D.R. Lloyd, Membrane distillation. II. Direct contact MD, J. Membr. Sci.

120 (1996) 123–133.[4] J.M. Ortiz De Zárate, C. Rincón, J.I. Mengual, Concentration of bovine serum albumin

aqueous solutions by membrane distillation, Sep. Sci. Technol. 33 (1998) 283–296.[5] V. Calabrò, E. Drioli, F. Matera, Membrane distillation in the textile wastewater treat-

ment, Desalination 83 (1991) 209–224.[6] V. Calabrò, B.L. Jiao, E. Drioli, Theoretical and experimental study on membrane dis-

tillation in the concentration of orange juice, Ind. Eng. Chem. Res. 33 (1994)1803–1808.

[7] Z. Ding, L. Liu, J. Yu, R.Ma, Z. Yang, Concentrating the extract of traditional Chinesemed-icine by direct contact membrane distillation, J. Membr. Sci. 310 (2008) 539–549.

[8] R.W. Schofield, A.G. Fane, C.J.D. Fell, Heat and mass transfer in membrane distilla-tion, J. Membr. Sci. 33 (1987) 299–313.

[9] L. Martı́nez-Dı́ez, M.I. Vázquez-González, Effects of polarization on mass transportthrough hydrophobic porous membranes, Ind. Eng. Chem. Res. 37 (1998)4128–4135.

[10] J.L.C. Santos, V. Geralds, S. Velizarov, J.G. Crespo, Investigation of flow patterns andmass transfer in membrane module channels filled with flow-aligned spacersusing computational fluid dynamics (CFD), J. Membr. Sci. 305 (2007) 103–117.

[11] M. Shakaib, S.M.F. Hasani, M. Mahmood, CFDmodeling for flow andmass transfer inspacer-obstructed membrane feed channels, J. Membr. Sci. 326 (2009) 270–284.

[12] J. Phattaranawik, R. Jiraratananon, A.G. Fane, Effects of net-type spacers on heat andmass transfer in direct contact membrane distillation and comparison with ultrafil-tration studies, J. Membr. Sci. 217 (2003) 193–206.

[13] L. Martinez, M.I. Vazquez-Gonzalez, F.J. Florido-Diaz, Study of membrane distillationusing channel spacers, J. Membr. Sci. 144 (1998) 45–56.

[14] C.D. Ho, H. Chang, C.L. Chang, C.H. Huang, Theoretical and experimental studies offlux enhancement with roughened surface in direct contact membrane distillationdesalination, J. Membr. Sci. 433 (2013) 160–166.

[15] A. Laik, D.P. Chakravarthy, G.B. Kale, On characterization ofwire-are-plasma-sprayedNion alumina substrate, Mater. Charact. 55 (2005) 118–126.

[16] S.G. Kandlikar, D. Schmitt, Characterization of surface roughness effects on pressuredrop in single-phase flow in minichannels, Phys. Fluids 17 (2005)(Article number100606).

[17] S.B. Iversen, V.K. Bhatia, K. Dam-Jphasen, G. Jonsson, Characterization of microporousmembranes for use in membrane contactors, J. Membr. Sci. 130 (1997) 205–217.

[18] G.C. Satri, C. Gostoli, S. Matulli, Low energy desalination processes using hydropho-bic membranes, Desalination 56 (1985) 277–287.

[19] S. Srisurichan, R. Jiraratananon, A.G. Fane, Mass transfer mechanisms and transportresistances in direct contact membrane distillation process, J. Membr. Sci. 277(2006) 186–194.

[20] E. Curcio, E. Drioli, Membrane distillation and related operations—a review, Sep. Sci.Technol. 34 (2005) 35–86.

[21] R.W. Schofield, A.G. Fane, C.J.D. Fell, Gas and vapour transport through microporousmembranes. I. Knudsen-Poiseuille transition, J. Membr. Sci. 53 (1990) 159–171.

[22] R.W. Schofield, A.G. Fane, C.J.D. Fell, Gas and vapour transport through microporousmembranes. II. Membrane Distillation, J. Membr. Sci. 53 (1990) 173–185.

[23] T.C. Chen, C.D. Ho, H.M. Yeh, Theoretical modeling and experimental analysis of di-rect contact membrane distillation, J. Membr. Sci. 330 (2009) 279–287.

[24] M. Gryta, M. Tomaszewska, Heat transport in the membrane distillation process,J. Membr. Sci. 144 (1998) 211–222.

[25] S. Kakac, R.K. Shah, A.E. Bergles, Low Reynolds Number Flow Heat Exchangers,Hemisphere, Washington, DC, 1983.

[26] J. Phattaranawik, R. Jiraratananon, A.G. Fane, Heat transport andmembrane distillationcoefficient in direct contact membrane distillation, J. Membr. Sci. 212 (2003) 177–193.

[27] J.R. Welty, C.E. Wick, R.E. Wilson, Fundamentals of Momentum, Heat, and MassTransfer, Third ed. John Wiley & Sons, New York, 1984.

[28] S. Kakac, R.K. Shah, W. Aung, Handbook of Single-Phase Convective Heat Transfer,Wiley, New York, 1987.

[29] 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.