Experimental and theoretical research on N-methyl-2-pyrrolidone concentration by vacuum membrane distillation using polypropylene hollow fiber membrane

Journal of Membrane Science 452 (2014) 157–164

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Experimental and theoretical research on N-methyl-2-pyrrolidone concentration by vacuum membrane distillation using polypropylene hollow fiber membrane Feifei Shao, Changqing Hao, Lei Ni, Yufeng Zhang n, Runhong Du, Jianqiang Meng, Zhen Liu, Changfa Xiao State Key Laboratory of Hollow Fiber Membrane Materials and Processes, Tianjin Polytechnic University, Tianjin 300387, PR China

art ic l e i nf o

a b s t r a c t

Article history: Received 18 June 2013 Received in revised form 18 September 2013 Accepted 21 September 2013 Available online 15 October 2013

In this work, N-methyl-2-pyrrolidone (NMP) aqueous solution was concentrated by means of vacuum membrane distillation (VMD) using polypropylene (PP) hollow fiber hydrophobic membrane. The effects of operating variables including feed concentration, feed temperature and vacuum degree on VMD performance were studied experimentally and theoretically. The results indicated that higher feed temperature and/or higher vacuum pressure could lead to an enhancement in H2O removal and no change in NMP rejection. The permeation flux of 9.5 L/m2 h and the rejection 98% can be achieved at the feed temperature of 80 1C and vacuum degree of 0.09 MPa. Higher feed concentration leads to a decrease in H2O flux and an increase in NMP flux and NMP molar fraction in permeate side. It proved that the Knudsen-viscous flow diffusion dominated the mass transform of H2O/NMP vapor in our experiment condition. Furthermore, “Dusty-Gas” model was used to predict membrane properties and Wilson equation method was used for predicting γ of the NMP aqueous solution. The theoretical data fitted well with the experimental data in our experiment conditions. & 2013 Elsevier B.V. All rights reserved.

Keywords: Vacuum membrane distillation Hydrophobic membrane Permeation flux Knudsen-viscous flow Wilson equation

1. Introduction Membrane distillation (MD) uses hydrophobic membrane such as polypropylene (PP), poly-tetrafluoroethylene (PTFE), polyethylene (PE) and polyvinylidene fluoride (PVDF) membranes as a physical barrier to prevent the feed solution passing through. In MD process components, gas evaporates in the heat side of the membrane and transports through the membrane pores under pressure gradient which caused by temperature gradient or by vacuum degree in permeate side, and then condenses in the cold side. Compared with many other membrane separations, the membrane itself does not interfere with MD performance but only provide a vapor–liquid interface. In contrast with the conventional distillation, MD has three advantages: firstly it takes place at low operating temperatures enabling separation of temperaturesensitive substances [1]; secondly the large membrane surface area increases the vapor–liquid contact area; thirdly it can reduce the heat loss and improve separation performance. Although most of MD can only be carried out in laboratory stage; MD is regarded as a promising membrane separation direction in the near future.

n

Corresponding author. Tel.: þ 86 22 83955078; fax: þ 86 22 83955055. E-mail address: [email protected] (Y. Zhang).

0376-7388/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.memsci.2013.09.041

MD can be divided into four types in accordance with the condensation methods: direct contact membrane distillation (DCMD), air gap membrane distillation (AGMD), gas membrane distillation (SGMD), and vacuum membrane distillation (VMD). DCMD is usually used in desalination, crystallization, concentrating fruit juices and treatment of waste water; SGMD has been applied at laboratory scale for the treatment of aqueous solutions containing non-volatile solutes such as salts (NaCl) as well as volatile solutes such as ammonia, alcohols (ethanol, isopropanol) and acetone; AGMD is used in desalination, solar units, food processing, treatment of aqueous alcohol solutions, breaking of azeotropic mixtures and extraction of volatile organic compounds [2]. VMD has been found to be a suitable process for the removal of trace amount of contaminants form dilute aqueous solutions such as chloroform removal from aqueous solutions [3], removal of 1,1,1-trichloroethane from water using a polyvinylidene fluoride hollow fiber membrane module [4] and recovery of volatile aroma compounds from black currant juice by vacuum membrane distillation [5]. NMP is organic solvent that is widely used in preparation of membranes. Polymer was dissolved in the solvent at the first time to make a dope, then the dope was extruded into a coagulation bath of water though a spinneret to form a hollow fiber membrane or through a split die to a plan one. In the bath, bilateral diffusion

158

F. Shao et al. / Journal of Membrane Science 452 (2014) 157–164

occurred. Organic solvent form dope (or membrane) into water and water into membrane while polymer precipitated to form membrane. At the same time, a large amount of dilute organic aqueous solution formed in the bath. It would pollute the environment if the waste solution were discharged directly, and NMP was identified as a reproductive toxicant by California in 2001 and then by the European Commission in 2003. For environmental friendliness, the waste water could be concentrated to produce pure NMP to reuse. Generally speaking, the aqueous solution is concentrated to a certain concentration by MD and then reproduces pure NMP after common distillation, which will reduce energy consumption. Concentration of organic solvents usually uses traditional distillation method [6,7] which commonly needs a high temperature and large energy consumption. Recently, membrane pervaporation is used [8–12]. The concentration and dehydration of NMP aqueous solution using membrane pervaporation with PAN and up-scaled high-silica CHA-type zeolite membranes were reported [13,14]. In this paper, NMP aqueous solution was concentrated by VMD technology with PP hollow fiber membrane prepared by melt-spinning and stretching method. PP membrane is a cheap membrane with good solvent resistance, high packing density and good mechanical strength. The effects of the operating variables of feed concentration, feed temperature and vacuum degree on VMD performance were studied both theoretically and experimentally. The mathematical model Knudsen-viscous Flow Diffusion together with Wilson equation method for γ of mixed solution was used to predict membrane properties.

2. Theoretical principle and correlation equations The mechanism types of gas mass transfer mainly include Knudsen flow diffusion, viscous flow diffusion and Knudsenviscous flow diffusion. There are two basic models for all possible transport mechanisms, “Dusty-Gas” model and “Schofield's” model. The main advantage of using “Dusty-Gas” model is the fact that they are parameters dependent only on membrane geometry and interactions of the membrane and the gas particle, yet independent of the temperature, which can be much briefer, whereas “Schofield's” model depends on the temperature and the gas used. The other disadvantage of “Schofield’s” model is that the experimental error of some parameters can be fairly large [15]. In order to select a suitable mechanism, the relationship between the mean free path of the molecules λ and the size of the micropores r should be concerned. When r r0.05λ, Knudsen flow diffusion should be selected, and when 0.05λ rr r50λ, Knudsenviscous flow diffusion dominates the mass transfer through the membrane [16]. Eq. (1) indicates that the mass flux N of component i in Knudsen flow diffusion, is linearly related to the pressure gradient (ΔP) across the membrane as follows [17]: pffiffiffiffiffiffi N i ¼ K m M i ΔP i ð1Þ Ni is the mass flux, Mi is the molar mass. Km is the transmembrane mass coefficient which only depends on membrane geometry, structure and temperature and has nothing to do with the types of the gas going through the membrane. When viscous contribution is taken into account, “Dusty-Gas” model [18] provides the mass calculation shown in Eq. (2).   P avg B0 ΔP M N¼ K 0V þ ð2Þ RT δ μ V¼

rffiffiffiffiffiffiffiffiffi 8RT πM

ð3Þ

where μ is the gas viscosity, K0 and B0 the parameters characteristic of the Knudsen and viscous flows, respectively, R the gas constant, T the thermodynamic temperature, Pavg the average of the pressure at both membrane sides, δ the membrane thickness, V the gas's mean molecular speed calculated by Eq. (3). B0 is a constant characteristic of the medium alone, and K0 is related in first approximation to a geometric constant characteristic of the dust particles and also a quantity. The main advantage of using K0 and B0 is the fact that they are parameters dependent only on membrane geometry and interactions of the membrane, but independent of the temperature. Eq. (2) may be rearranged as follows: NRTδ ΔPMV

¼ K 0 þ B0

P avg μV

i.e. Z ¼ K 0 þ B0 X

ð4Þ

where Z ¼ NRT=ΔPV and X ¼ P avg =μV, which is more suitable in order to obtain the parameters characteristic of the transport mechanisms (K0 and B0) from the linear fit of the experimental data (Z versus X). Eq. (5) is used to calculate the transmembrane flux for mixed solution. P1Yi1 and P2Yi2 are respectively the partial pressure of component i in the heat side and in the vacuum side, and P1Yi1 ¼Pi ¼PinγiXi. Yi2 is measured by the experimental test, γi is the activity coefficient of i. pffiffiffiffiffiffi Ni ¼ K m M i ðP 1 Y i1  P 2 Y i2 Þ ð5Þ There are three methods to calculate activity coefficient γ: Wilson equation, NRTL equation and UNIFAC group contribution method. Wilson equation is relatively simple, and its parameters are not appreciably affected by temperature, so we chose Wilson equation to estimate γi. Wu et al. [19] studied the NMP concentration using distillation, and obtained Antoine equation to calculate the saturation vapor pressure of pure NMP, shown in Eq. (6). lg PV ¼ 6:882  2196:6434=ð243:7 þtÞ

ð6Þ

Eqs. (7)–(12) are Wilson equation expressions, Wu et al. also obtained binary interaction parameters of the Wilson equation, and proved that the test results of thermodynamic consistency meet the Herington thermodynamic consistency requirements.   Λ12 Λ21 ln γ 1 ¼  lnðx1 þ Λ12 x2 Þ þ x2  ð7Þ x1 þ Λ12 x2 x2 þ Λ21 x1  ln γ 2 ¼  lnðx2 þ Λ21 x1 Þ þ x1

Λ21 Λ12  x2 þ Λ21 x1 x1 þ Λ12 x2

 ð8Þ

Λ12 ¼

V2 exp½  ðg 12  g 11 Þ=RT V1

ð9Þ

Λ21 ¼

V1 exp½  ðg 21  g 22 Þ=RT V2

ð10Þ

ðg 12  g 11 Þ ¼ 2447:16 J=mol

ð11Þ

ðg 21  g 22 Þ ¼ 1417:85 J=mol

ð12Þ

3. Experimental 3.1. Membranes and chemicals In this work, different mass fraction aqueous solutions of NMP (0–40 wt%) were employed. NMP purchased from Tianjin Kermel Chemical Reagent Co., Ltd was of AR grade and water content

F. Shao et al. / Journal of Membrane Science 452 (2014) 157–164

159

Table 1 The parameters of different membrane modules. Types

n

do (mm)

di (mm)

δ (mm)

Porosity (%)

S (m2)

PP-1 PP-2 PP-3

60 60 60

0.455 0.325 0.370

0.356 0.245 0.273

0.050 0.040 0.049

62 53 48

1.543  10  2 0.870  10  2 0.718  10  2

Fig. 2. Schematic representation of vacuum membrane distillation apparatus.

Fig. 1. Gas permeation device.

r0.1%. There are three types of PP hollow fiber membrane used in our VMD test. PP-1, PP-2 and PP-3 were fabricated using the meltspinning and stretching method [20] by us, and their properties are listed in Table 1.

3.4.2. Porosity test Membrane porosity was calculated by Eq. (13). Where ε is the membrane porosity, ρs the density of PP entity which has no pores, ρs ¼ 0.9 g/cm3, ρm the density of the membrane. ρm can be calculated by ρm ¼ m=V, in which m is the weight of the membrane (g), V the volume of the membrane (cm3).   ρ ε ¼ 1  m  100% ð13Þ ρs 3.4.3. Membrane performance evaluation The permeation rate can be calculated by Eq. (14), where N, V, S, and Ttime respectively represent the permeation rate (L/m2 h), volume of the solution in permeate side (L), effective hollow fiber area (m2), and operation time (h).

3.2. Gas permeation N¼ Gas permeation test was used to estimate the parameters K0 and B0 in Knudsen-viscous flow diffusion. Nitrogen was used as the permeating gas and its temperature was 25 1C. The permeation device is shown in Fig. 1 [21]. A regulator valve was used to control gas pressure and the pressure difference ΔP varied from 30 to 200 kPa. The pressure was measured at the inlet and outlet of the membrane cell using mercury manometers, gas flow capacity (Q) was measured with a calibrated gas rotameter.

V S T time

ð14Þ

The NMP rejection Rj is defined by Eq. (15), in which Cp and Cf are the weight concentration of the solution in the permeate side and the feed side (kg/L). TOC-VCPH (Total organic carbon analyzer, SHIMADZU) was used to measure the concentration of NMP in the aqueous solution.   Cp Rj ¼ 1   100% ð15Þ Cf

3.3. Equipment of VMD and the parameters of membrane module The schematic representation of VMD set-up is shown in Fig. 2. NMP aqueous solution was put into a feed tank and transported through a heat exchanger for a stable temperature then circulated through the shell side while the permeate vapor was removed from the lumen side of the hollow fibers in a closed circuit using a vacuum pump and condensed in a cold serpentine condenser. The natures of membrane module using three different PP membranes including outside diameter (do) and inside diameter (di), porosity, and the total outer surface area of the fibers (S) are listed in Table 1.

4. Results and discussion 4.1. Surface morphology The pore size of the three different PP membranes was examined by SEM, and is shown in Fig. 3. The size of membrane pores can be seen clearly with SEM, which was used in comparing the relationship between the mean free path (λi) and pore size (r). The maximum pore size of PP-1 is about 0.5  0.1 μm2, PP-2 is about 0.3  0.08 μm2 and PP-3 is about 0.6  0.1 μm2. 4.2. Gas permeation

3.4. Characterization 3.4.1. Morphology FESEM (FESEM, HItachi-s-4800) was used in observing microstructure of the membrane.

Ten different groups of ΔP and Q were tested with changing the pressure valve in Fig. 1, and with deformed Eq. (4) ten different groups of X and Z were obtained. From the linear fit of the experimental data (Z versus X), distinguishable trends were

160

F. Shao et al. / Journal of Membrane Science 452 (2014) 157–164

Fig. 4. Plot of the experimental gas permeation data obtained with different membranes by dusty gas model using Eq. (2).

Table 2 Values of K0 and B0 with their standard errors obtained from the fit of the experimental fluxes to Eq. (2) with nitrogen. Types PP-1 PP-2 PP-3

K0

B0 –11

R –16

2.070  10 1.133  10–16 1.077  10–16

8.091  10 5.332  10–11 4.162  10–11

0.9973 0.9935 0.9920

Table 3 Values of Km of different types of PP membrane with pure water changing temperature of the feed solution. Km  10  7 (s mol1/2/m kg½)

PP-1 PP-2 PP-3

Fig. 3. Surface morphology of different membranes (a. PP-1, b. PP-2, and c. PP-3).

yielded, and are shown in Fig. 4. K0 and B0 were obtained, and are shown in Table 2.

T (1C) 55

60

65

70

75

80

3.19 2.14 1.73

3.64 2.36 1.84

3.84 2.61 2.04

4.29 2.93 2.27

5.05 3.42 2.62

5.41 3.68 2.84

bulk was considered as the approximation for the interfacial temperature which was proved by Izquierdo-Gil and Jonsson [22]. From that different interfacial temperature from 55 1C to 80 1C, and using Antoine equation for water vapor [15] and Eq. (6) for NMP vapor, six different pressures at the interface were determined. From those interfacial pressures, the value of Km at each temperature was determined using Eq. (1) for water vapor, and are listed in Table 3. Eq. (16) [2] for the formula of Km in Knudsen flow diffusion is as follows, which indicates that Km decreases with temperature increasing. The calculated results in Table 3 were not in contradiction to Eq. (16). It proved that Knudsen flow diffusion was not suitable for the mass transform of water vapor in our experiment.   2 εr 8RT 1=2 K m ¼ Bki ¼ ð16Þ 3RT τδ πM i

4.3. VMD test of pure water

Furthermore, the relationship between the mean free path (λi) and pore size (r) was used to study the mechanism type of gas mass transfer. Eq. (17) [2] was the expression of λi for a species i.

Pure water was used in VMD test to predict Km, K0 and B0 of different PP membranes. In our test, the feed temperature in the

kB T λi ¼ pffiffiffi 2πP avg si 2

ð17Þ

F. Shao et al. / Journal of Membrane Science 452 (2014) 157–164

respectively used to calculate the saturation vapor pressure of pure NMP and water. The saturation vapor pressure of NMP and water in mixed solution were obtained using P1Yi1 ¼Pi ¼PiγiXi. The partial pressure in the vacuum side P2Yi2 was calculated by experiment data. The theoretical data of flux was calculated by Eq. (2). The theoretical and experimental variations of the membrane performances include H2O and NMP flux and NMP rejection with feed temperature are respectively shown in Figs. 5–7. The curves

10

PP-1 PP-2 PP-3

8

NH2O( L/m2h)

where si is the collision diameter (2.641 Å for water vapor), kB is the Boltzmann constant, Pavg the mean pressure of membrane on both sides and T the absolute temperature. From the experimental temperature 55–80 1C and the vacuum degree at 0.09 MPa, six different groups of T and Pavg were determined, and λi in different temperatures was estimated by Eq. (17). The results are listed in Table 4. From the results, we can see that 0.05λr r r50λ, which indicate that Knudsen-viscous flow diffusion is fit for our experiment. Pure water was used in VMD to predict K0 and B0 in Knudsenviscous diffusion. Test different ΔP and N with changing temperature from 55 1C to 80 1C. The viscosity of water vapor μ can be calculated from μ  109 ¼8.15T1.23 [18]. From the linear fit of the experimental data (Z versus X) with Eq. (4), K0 and B0 were obtained for water. The results are listed in Table 5. In our VMD experiments, values of K0 and B0 calculated by water vapor are very close to our GP experiment. These values can be compared with others in previous literature. The experiments were made through a moderately low-permeability graphite at 25 1C in [23], obtaining from the fit to Eq. (2) the following parameter values K0 ¼3.12  10  11 m and B0 ¼ 2.13  10  18 m2. The value K0 was approach to us, and B0 was smaller than us. In references [15,18], it is K0 about four times greater than ours, and B0 about two times higher than ours.

161

6 4 2 0 325

330

335

340

345

350

355

T(K)

4.4. Effect of feed temperature on the VMD performance for NMP aqueous solution

Fig. 5. Effect of feed temperature on theoretical and experimental H2O flux (XNMP ¼ 10 wt%, Pvacuum ¼0.09 MPa).

Eqs. (7)–(12) were used to calculate the γ of H2O and NMP in NMP aqueous solution at different temperatures. The results are shown in Table 6. Eq. (6) and Antoine equation of H2O [17] were

T (K)

Saturation vapor pressure of H2O (kPa)

Pavg (kPa)

λi (μm)

328.15 333.15 338.15 343.15 348.15 353.15

15.752 19.932 25.022 31.176 38.563 47.373

13.376 15.466 18.011 21.088 24.782 29.187

1.093 0.960 0.837 0.725 0.626 0.539

NNMP( Kg/m2h)

Table 4 λi at different temperatures.

0.020 PP-1 PP-2 PP-3

0.015 0.010 0.005 0.000 325

330

335

340

345

350

355

T(K)

Types

K0

B0

R

PP-1 PP-2 PP-3

8.211  10–11 5.322  10–11 4.211  10–11

2.067  0–16 1.132  0–16 1.087  10–16

0.9983 0.9963 0.9965

Table 6 Values of γ in different temperatures of NMP aqueous solution (XNMP ¼10 wt%, Pvacuum ¼0.09 MPa). T (1C)

P n H2 O (kPa)

PnNMP (kPa)

V1

V2

λH2 O

λNMP

P H2 O (kPa)

PNMP (kPa)

55 60 65 70 75 80

15.752 19.932 25.022 31.176 38.563 47.373

0.337 0.446 0.583 0.759 0.975 1.225

18.261 18.306 18.356 18.407 18.463 18.52

96.205 96.167 96.188 96.015 96.236 96.273

1.002 1.002 1.002 1.0019 1.0018 1.0018

2.273 2.206 2.144 2.085 2.028 1.975

15.467 19.570 24.568 30.611 38.560 46.509

0.015 0.020 0.025 0.032 0.040 0.049

Fig. 6. Effect of feed temperature on theoretical and experimental NMP flux (XNMP ¼ 10 wt%, Pvacuum ¼0.09 MPa).

1.00 0.98 0.96

Rj

Table 5 Values of K0 and B0 with their standard errors obtained from the fit of the experimental fluxes to Eq. (2) for the different membranes, with pure water changing the feed temperature.

PP-1 PP-2 PP-3

0.94 0.92 0.90 0.88 325

330

335

340

345

350

355

T(K) Fig. 7. Effect of feed temperatures on theoretical and experimental Rj (XNMP ¼10 wt%, Pvacuum ¼ 0.09 MPa).

F. Shao et al. / Journal of Membrane Science 452 (2014) 157–164

are the theoretical linear fit and the points are the experimental data. H2O flux and NMP flux increased with the temperature increasing both theoretically and experimentally. Higher temperature can lead to higher vapor pressure which results in the increasing of H2O and NMP flux, but no Changes in NMP rejection. Therefore, a high feed temperature was favor for the enhancement of H2O removal, and also dramatically yielded a high permeation flux. Comparing the three different membranes, PP-2 has the largest membrane flux, while PP-3 has the minimum one, because of the largest porosity of PP-2 and minimum porosity of PP-3, although pore sizes of PP-3 are slightly larger than the other two membranes. The flux was much larger than the max flux of 1.5 L/m2 h at 70 1C in the literature [13], and smaller than the flux of 18 L/m2 h at 90 1C in the literature [14] in concentrating NMP aqueous solution.

NNMP( L/m2h)

162

0.030 0.028 0.026 0.024 0.022 0.020 0.018 0.016 0.014 0.012 0.010 0.008 0.006 0.004 0.002 0.000 0.00

T=55 °C T=65 °C T=75 °C

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

feed concentration

4.5. Effect of the feed concentration on VMD performance for NMP aqueous solution

Fig. 9. Effect of feed concentration on theoretical and experimental flux of NMP (T ¼ 65 1C, Pvacuum ¼ 0.09 MPa).

1.00 0.99 0.98 0.97 0.96

Rj

Analogously, the value of γ and the saturation vapor pressure in different mass fractions solution are listed in Table 7. Likewise, the saturation vapor pressure at other temperatures can be estimated. Analogous experiments were carried out using aqueous solution of NMP (5–40 wt%) at three different temperatures 75 1C, 65 1C and 55 1C at Pvacuum ¼ 0.09 MPa. The membrane used is PP-1. Figs. 8, 9 and 10 respectively show the effect of the feed solution concentration on VMD performance theoretically and experimentally at three different temperatures. With the feed concentration increasing, water flux decreased and NMP flux increased. The NMP rejection decreased and it clearly can be seen that NMP molar fraction increased in the permeate side. That because, the saturated

0.95

T=55 °C T=65 °C T=75 °C

0.94 0.93

Table 7 Values of γ in different mass fractions solution (Pvacuum ¼ 0.09 MPa, T ¼ 65 1C). X1 (%)

X2 (%)

λ1

λ2

0 5 10 15 20 25 30 35 40

100 99.1 98.0 96.9 95.7 94.3 92.4 91.1 89.2

0 0.9 2.0 3.1 4.3 5.7 7.2 8.9 10.8

1 1 1.002 1.006 1.009 1.014 1.02 1.08 1.17

0 2.779 2.445 2.192 1.982 1.795 1.636 1.471 1.312

NH2O ( L/m2h)

Mass fraction of NMP (%)

7.0 6.5 6.0 5.5 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0

P1 (kPa)

P2 (kPa)

24.797 24.571 24.392 24.162 23.926 23.583 23.043 22.321

0 0.015 0.029 0.040 0.051 0.060 0.069 0.073 0.087

0.92 0.91 0.00 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45

feed concentration Fig. 10. Effect of feed concentration on theoretical and experimental Rj (T ¼65 1C, Pvacuum ¼ 0.09 MPa).

vapor pressure of NMP increased and the saturated vapor pressure of H2O decreased with the increasing of the mass concentration which is shown in Table 7, and led to the increasing of the NMP molar fraction in the permeate side. It shows that the theoretical mass flux fits well with experimental results, and the theoretical NMP rejection was slightly higher than the experimental one. 4.6. Effect of vacuum degree on the VMD performance for NMP aqueous solution Figs. 11, 12 and 13 respectively show the theoretical and experimental changes of VMD performances with the increasing of vacuum degree. The condition of feed solution was 75 1C and XNMP ¼ 10 wt%. The vacuum degree increasing, the flux of H2O and NMP both increasing, can be explained by the growing transmembrane driving force along with the vacuum degree increasing. NMP rejection remained unchanged, and the theoretical data fits well with the experimental one.

T=55 °C T=65 °C T=75 °C

5. Conclusions

0.0

0.1

0.2

0.3

0.4

feed concentration Fig. 8. Effect of feed concentration on theoretical and experimental flux of H2O (T¼ 65 1C, Pvacuum ¼0.09 MPa).

The work presented a methodology in N-methyl-2-pyrrolidone concentration by vacuum membrane distillation process using PP hollow fiber membrane. Wilson equation method was used for estimating γi of NMP aqueous solution, and the “Dusty-Gas” model

F. Shao et al. / Journal of Membrane Science 452 (2014) 157–164

NMP rejection. H2O flux and NMP rejection decreased with the feed concentration increasing, while NMP flux increased. Finally, the theoretical data simulated by the mathematical “Dusty-Gas” model fitted well with the experimental data and it indirectly showed that the value of γi estimated by Wilson equation method was accurate.

7 PP-1 PP-2 PP-3

NH2O( L/m2h)

6

163

5 4

Acknowledgments

3

This work was financially supported by the China High-Tech R&D Program (863 Program, 2012AA03A602), the National Basic Research Development Program of China (973 Program, 2012CB722706) and the National Natural Science Foundation of China (51073120).

2 1 0.070

0.075

0.080

0.085

0.090

vacuum degree(MPa) Fig. 11. Effect of vacuum degree on theoretical and experimental flux of H2O (T ¼75 1C, XNMP ¼0.02).

PP-1 PP-2 PP-3

0.014

NNMP( L/m2h)

viscous flux coefficient (m2) weight concentration of the solution in the permeate side (kg/L) Cf weight concentration of the solution in the feed side (kg/L) (g12  g11), (g21  g22) interaction energy parameters of H2O and NMP Km parameter defined in Eq. (1) (s mol1/2/m kg1/2) K0 Knudsen flux coefficient (m) m mass (kg) M mole mass (kg/mol) Mi mole mass of component i (kg/mol) N mass flux (kg/m2 s) Ni mass flux of component i (kg/m2 s) n Pi saturation vapor pressure of pure substance (Pa) ΔP pressure difference (Pa) Pavg average pressure (Pa) P1 total pressure in heat side (Pa) P2 total pressure (Pa) in vacuum side (Pa) PV saturation vapor pressure of pure NMP (Pa) Pvacuum vacuum degree (Pa) Q gas flow capacity (L/min) R gas constant (J/mol K) Rj NMP rejection T thermodynamic temperature (K) t Celsius temperature (1C) V volume (m3) V gas's mean molecular speed (m s  1) V1 liquid molar volume of H2O (L/mol) V2 liquid molar volume of NMP (L/mol) Λ12 , Λ21 parameters of Wilson equation X defined in Eq. (4) (m  1) X1 mole fraction of 1 in the liquid phase X2 mole fraction of 2 in the liquid phase Xi mole fraction of i in liquid phase Yi1 mole fraction of i in the vapor phase of the heat side Yi2 mole fraction of i in the vapor phase of the vacuum side Z defined in Eq. (4) (m) γ the activity coefficient δ membrane thickness (m) μ gas viscosity (Pa s) ε porosity ρm density of PP membrane (kg/m3) ρs density of PP entity (kg/m3) B0 Cp

0.016

0.012 0.010 0.008 0.006 0.004 0.002 0.070

0.075

0.080

0.085

0.090

vacuum degree(MPa) Fig. 12. Effect of vacuum degree on theoretical and experimental flux of NMP (T ¼75 1C, XNMP ¼0.02).

1.00 0.99 0.98 0.97

Rj

Nomenclature

0.96

PP-1 PP-2 PP-3

0.95 0.94 0.93 0.92 0.91 0.070

0.075

0.080

0.085

0.090

vacuum degree(MPa) Fig. 13. Effect of vacuum degree on theoretical and experimental Rj (T ¼ 75 1C, XNMP ¼0.02).

of Knudsen-viscous Flow Diffusion was used to predict the process properties. The effects of operating variables including feed concentration, feed temperature and vacuum degree on VMD performance were incorporated in the simulation. Under the experimental conditions, the maximum rejection rate was approximately 98%, and the total flux was 9.5 L/m2 h. The flux of H2O and NMP both increased with the increasing of the feed temperature and vacuum degree, but no obvious variation in

164

F. Shao et al. / Journal of Membrane Science 452 (2014) 157–164

References [1] R.B. Jorgensen, A.S. Meyer, C. Varming, Recovery of volatile aroma compounds from black currant juice by vacuum membrane distillation, J. Food Eng. 64 (2004) 23–31. [2] M. Khayet, T. Matsuura, Membrane Distillation Principles and Applications, Elsevier, Amsterdam, 2011. [3] A.M. Urtiaga, G. Ruiz, I. Ortiz, Kinetic analysis of the vacuum membrane distillation of chloroform from aqueous solutions, J. Membr. Sci. 165 (2000) 99–109. [4] B. Wu, X.Y. Tan, K. Li, W.K. Teo, Removal of 1,1,1-trichloroethane from water using a polyvinylidene fluoride hollow fiber membrane module: vacuum membrane distillation operation, Sep. Purif. Technol. 52 (2006) 301–309. [5] V. Soni, J. Cabildo, G. Jonsson, R. Gani, Modeling and analysis of vacuum membrane distillation for the recovery of volatile aroma compounds from black currant juice, J. Membr. Sci. 320 (2008) 442–455. [6] B. Fabiano, R. Pastorino, M. Ferrando, Distillation of radioactive liquid organic wastes for subsequent wet oxidation, J. Hazard. Mater. 57 (1998) 105–125. [7] I.D. Gil, J.M. Gómez, G. Rodríguez, Control of an extractive distillation process to dehydrate ethanol using glycerol as entrainer, Comput. Chem. Eng. 39 (2012) 129–142. [8] S. Bandini, G.C. Sarti, Concentration of must through vacuum membrane distillation, Desalination 149 (2002) 253–259. [9] M. Tomaszewska, M. Gryta, A.W. Morawski, Study on the concentration of acids by membrane distillation, J. Membr. Sci. 102 (1995) 113–122. [10] J. Ma, M.H. Zhang, H. Wu, X. Yin, J. Chen, Z.Y. Jiang, Mussel-inspired fabrication of structurally stable chitosan/polyacrylonitrile composite membrane for pervaporation dehydration, J. Membr. Sci. 348 (2010) 150–159. [11] Z.W. Ding, L.Y. Liu, Z. Liu, R.Y. Ma, The use of intermittent gas bubbling to control membrane fouling in concentrating TCM extract by membrane distillation, J. Membr. Sci. 372 (2011) 172–181. [12] L. Xu, M. Rungta, M.K. Brayden, M.V. Martinez, B.A Stears, G.A. Barbay, W.J. Koros, Olefins-selective asymmetric carbon molecular sieve hollow fiber

[13]

[14]

[15]

[16] [17]

[18]

[19] [20]

[21]

[22]

[23]

membranes for hybrid membrane-distillation processes for olefin/paraffin separations, J. Membr. Sci. 423–424 (2012) 314–323. H.A. Tsai, Y.L. Chen, K.R. Lee, J.Y. Lai, Preparation of heat-treated PAN hollow fiber membranes for pervaporation of NMP/H2O mixtures, Sep. Purif. Technol. 100 (2012) 97–105. K. Sato, K. Sugimoto, N. Shimotsuma, T. Kikuchi, T. Kyotani, T. Kurata, Development of practically available up-scaled high-silica CHA-type zeolite membranes for industrial purpose in dehydration of N-methyl pyrrolidone solution, J. Membr. Sci. 409–410 (2012) 82–95. ́ C. Fernández-Pineda, M.A. Izquierdo-Gil, M.C. Garcıa-Payo, Gas permeation and direct contact membrane distillation experiments and their analysis using different models, J. Membr. Sci. 198 (2002) 33–49. M. Khayet, Membranes and theoretical modeling of membrane distillation: a reviews, Adv. Colloid Interface 164 (2011) 56–88. M.A. Izquierdo-Gil, G. Jonsson, Factors affecting flux and ethanol separation performance in vacuum membrane distillation (VMD), J. Membr. Sci. 214 (2003) 113–130. F.X Gao, X.C. Chen, G.G. Yu, C. Asumana, Compressible gases transport through porous membrane: a modified dusty gas model, J. Membr. Sci. 379 (2011) 200–206. F. Wu, The Purification of the N-methyl Pyrrolidone. Doctor's Degree, East China University of Science and Technology Press, Shanghai (2012) 2012; 1–5. F.F. Shao, Y.F. Zhang, Study on lamellar structures and parameters of polyethylene hollow fiber membranes made by melt-spinning and stretching method, Appl. Mech. Mater. 178–181 (2012) 491–494. K.W. Lawson, M.S. Hall, D.R. Lloyd, Compaction of microporous membranes used in membrane distillation. I. Effect on gas permeability, J. Membr. Sci. 101 (1995) 99–108. M.A. Izquierdo-Gil, G. Jonsson, Factors affecting flux and ethanol separation performance in vacuum membrane distillation (VMD), J. Membr. Sci. 214 (2003) 113–130. E.A. Mason, A.P. Malinauskas, R.B. Evans, Flow and diffusion of gases in porous media, J. Chem. Phys. 46 (1967) 3199–3217.