Sie sind auf Seite 1von 11

Desalination 206 (2007) 215225

Modelling the effects of nanofiltration membrane properties on system cost assessment for desalination applications
A.W. Mohammada, N. Hilalb*, H. Al-Zoubib, N.A. Darwish3, N. Alia
Department of Chemical and Process Engineering, Universiti Kebangsaan Malaysia, 43600 UKM Bangi, Selangor Darul Ehsan, Malaysia b Centre for Clean Water Technologies, School of Chemical, Environmental and Mining Engineering, The University of Nottingham, NG7 2RD, United Kingdom Tel. +44 (115) 9514168, Fax +44 (115) 9514115; email: nidal.hilal@nottingham.ac.uk c Chemical Engineering Program, The Petroleum Institute, P.O. Box 2533, Abu Dhabi, United Arab Emirates
a

Received 2 February 2006; accepted 16 February 2006

Abstract Nanofiltration (NF) membranes have been successfully used in various applications in water and wastewater treatment as well as other industries. In desalination plants, NF membranes were found to be suitable for pretreatment application prior to the desalination process. In general NF membranes can be characterized in terms of the effective pore radius (rp), effective charge density (Xd) and permeability (Pm). NF membranes available in the market show a wide range of these properties and thus this variation affects the membrane performance significantly. Through a predictive model such as the Donnan-Steric-Pore-Dielectric-Exclusion (DSPM-DE) model, the performance of NF membranes can be predicted in terms of the solute rejections, fluxes and applied pressure. In addition, for desalination application, the osmotic effect due to the high concentration of salt can also be included in such model. The osmotic effect is taken into account by using the Vant Hoff equation and this is incorporated in the modified DSPM-DE model to allow for approximation of the reduced permeate flux. In this study, the modified DSPM-DE model was used in conjunction with an economic assessment model to analyse the effect of the NF membrane characteristics (rp, Xd and Pm) on the cost of the NF membrane system with a particular emphasis on desalination application. Detailed analyses of the economic assessment shows that rp can play a major role in reducing the cost of an NF system provided that the minimum rejection requirement is met. Keywords: Nanofiltration; Economic assessment; Modeling; Rejection; Osmotic effect *Corresponding author. Presented at EuroMed 2006 conference on Desalination Strategies in South Mediterranean Countries: Cooperation between Mediterranean Countries of Europe and the Southern Rim of the Mediterranean. Sponsored by the European Desalination Society and the University of Montpellier II, Montpellier, France, 2125 May 2006.
0011-9164/07/$ See front matter 2007 Published by Elsevier B.V.
doi:10.1016/j.desal.2006.02.068

216

A.W. Mohammad et al. / Desalination 206 (2007) 215225

1. Introduction Nanofiltration (NF) membranes have found application in several areas such as water softening and water treatment, pharmaceutical and biotechnology [1,2]. One of the main applications which is relevant to desalination plants is in using NF as a pretreatment of either the seawater or brackish water feed to the plant [3,4]. The ability to model the NF membrane processes will be very useful, especially for process prediction, optimization and economic assessment. Various models have been proposed in modelling the rejection performance of salts and charged organics in NF membranes [58]. Most of these models were based on the extended Nernst Planck equation to describe the transport inside the membrane. One of the most commonly cited models is the so-called Donnan-Steric-Pore model (DSPM) [9]. The DSPM model has recently been modified to include the dielectric exclusions effect for charged ions [10,11]. Such a model has been termed DSPM-DE model and has been shown to be able to predict rejections of divalent ions including Mg2+ and Ca2+ quite accurately compared to DSPM model [10]. One of the shortcomings of the previous works however, is that the prediction of the ion rejections was based on specified flux without taking into account the osmotic effect. This may be correct for dilute solution. However, in more concentrated solutions such as to be found in desalination plants, the flux will depend on the osmotic effect caused by the salt concentrations at the membrane wall. Therefore, the ability to predict rejection and flux simultaneously will further improve the prediction capability [12,13]. Subsequently, the models can be used to analyse the cost impact of an NF system [14] with greater reliability. In our recent work to be published separately [15], we have modified the DSPM-DE model to include the osmotic effect due to the high concentration of the salt solutions. The modified model has been found to be able to predict the rejection and flux of single salts solutions such as NaCl,

Na2SO4, and MgCl2 at higher concentration ranging from 5000 to 2000 ppm, which are the typical concentrations in seawaters. In this work, the modified DSPM-DE model will be used to look at the cost impact (capital and operating cost) of an NF system. Firstly, the modified DSPM-DE model will be used to show the effect of varying effective pore radius (rp) and effective charge density (Xd) on the rejection and permeate fluxes of MgCl2 solutions. Secondly, by combining the modified DSPM-DE model with a cost model such as that proposed by Verberne et al. [16], the effect of the NF membrane characteristics on the cost of an NF system can be analysed. Similar study has been carried out previously [14] using the DSPM model. However in that study, the osmotic effect of the solutions was not taken into account. It is hoped that through this study better understanding can be achieved on the impact of the NF membrane characteristics on the performance and economics of the NF system. This will lead further towards improved membranes and system design in the future.

2. Theory 2.1. Modified DSPM-DE model The DSPM-DE model [10,11] is based on the following extended NernstPlanck equation whereby the concentration gradient inside the membrane pore is related to the potential gradient:
dci J z c F d = v i i K ic ci C pi dx Di , p Rg T dx

(1)

The potential gradient is

d = dx

z
i =1

Jv [ K i ,c ci C pi ] Di , p
n i =1

( F / Rg T ) zi2 ci

(2)

A.W. Mohammad et al. / Desalination 206 (2007) 215225

217

The main difference between the DSPM model and the DSPM-DE model is due to the inclusion of the dielectric exclusion effect in the partitioning coefficients of each ion as follows
Ci , w ci ,0 = i exp ( zi D ,0 ) exp ( zi2 Wi ,0 )

2.2. Cost assessment model Cost of each membrane separation systems is definitely varied and depends on production capacity, type of treatment, design criteria, climate condition, characteristics of land and building, etc. Most of the cost models were established on the basis of actual project data and tender provided by supplier and vendor [17,18]. In this study, the model proposed by Verberne et al. [16], and the latter will be known as Verberne Cost Model, will be used. Verberne Cost Model has been used previously to assess the economic aspect of a treatment system for elimination of pesticides, nitrate and hardness from soil water [19]. All equations in the cost model are created based on project practical data and actual tender from suppliers. Therefore Verbene Cost Model is deemed to be a suitable model for the purpose of this study. Details regarding the model are presented in Table 1. In this study, a simulation based on the modified DSPM-DE model will be carried out to remove at least 80% (R 80%) of 100 molm3 MgCl2 solution. The desired production rate is 2000 m3/h (= 48,000 m3/d = 17.5 106 m3/y). The system is operated at recovery of 0.8. Other parameters such as temperature and pH are assumed constant during the process and fouling factor is not taken into account. Flux through the membrane is a function of the operating pressure and the osmotic effect due to the salt concentration at the membrane wall. In the case whereby the concentration polarisation is significant, much higher pressure is required to obtain similar flux. The higher the flux, the lesser the number of membrane module needed to achieve a certain production rate. In this case, each membrane module is assumed to have the same size and geometry with surface area of 30 m2. 3. Results and discussion 3.1. DSPM-DE model prediction In order to show the capability of the modified DSPM-DE model to predict rejection and flux of

(3)

Thus in Eq. (3) above, the partitioning of ions at the membrane interface is due to the steric effect (), Donnan exclusion effect () and the dielectric exclusion effect (W). The dielecteric exclusion effect which is represented by the Born solvation energy barrier, W, has been calculated based on the method proposed by Bowen and Welfoot [10]. Our approach has made use of the Hagen Poiseulle equation to correlate the permeate flux with the applied pressure.

Jv =

rp2 8(x / Ak )

(P )

(4)

For pure water, is zero and Eq. (4) in conjunction with pure water permeability data can be used to determine x/Ak provided that rp is determined independently by other methods such as neutral solute rejection data or atomic force microscope (AFM) imaging [9]. The osmotic pressure for the solution, , can be estimated using equation such as the Vant Hoff equation or Pitzer equation [12]. Our calculation has shown that [15] for estimating the osmotic pressure difference, , both equations yielded similar results. However, for calculation of osmotic effect only, the Pitzer equation is preferred since the Vant Hoff equation tends to overestimate the value of . In this study the Vant Hoff equation will be used for its simplicity. The Vant Hoff equation is given as follows
= R g T ( C i , w C i , p )

(5)

Details of the model simulation will be presented elsewhere [15].

218

A.W. Mohammad et al. / Desalination 206 (2007) 215225

Table 1 List of equation for the cost model [19]


Capital cost items Operating cost items (i) Civil investment: For instance, a building where (i) Depreciation costs: depreciation rate upon investment cost. The investments are linearly depreciated and installation of membrane system is be positioned. interest is neglected. Depreciation period is 30 years. Ccivil = 862 QF + 1239 n (ii) Mechanical engineering: costs for pumps, filters, piping, etc. Depreciation period is 15 years.
0.85 + 908 n Cmech = 3608 QF

Cdeprec =
(ii)

Ccivil 30

1 15

(Cmech + Celectro ) +

Cmembrane 5

(iii) Electrotechnical costs: costs for energy supply, control engineering and all electronic components. Celectro = 14106 + 54 PQF

Energy cost: energy required to pump the feed stream into membrane system. It is assumed that membrane system uses 40 Wh/m3 for each m3 feeding and feeding pressure (bar). An electric cost is estimated as 0.05 /kWh. Cenergy = [((QF P) / ) + ((QR P)/)] (0.05/1000)

(iv) Membrane investments: cost for membrane installation. The membrane lifetime is taken as 5 years. It was assumed that one membrane module (iv) Maintenance cost: 2% of the total investment costs. costs about 1000 . Cmaint = 0.02 Ccapital Cmembrane = 1000 n (v)

(iii) Chemical cost: Cost needed by a total of chemical materials which be filtered. This cost is used as 0.023 /m3 filtration. Cchemical = QF 0.023

Quality control cost: 2% of the total investment costs. Cquality control = 0.02 Ccapital costs. Cinstallation = 0.02 Ccapital

(vi) Operation of installation: 2% of the total investment

ions at high concentration, one of the results from our study to be published separately [15] is shown in Fig. 1. Fig. 1 shows the comparison between experimental data and predicted rejection (Fig. 1a) and flux (Fig. 1b) of MgCl2 for membrane NF-90. The ability of the model is quite good in estimating the flux reduction of the MgCl2 solutions as a function of concentration. Once the accuracy of the modified DSPM-DE model has been established, the model can be used to simulate the rejection at various membrane characteristics. Figs. 25 show the prediction of MgCl2 rejection and fluxes for different rp ranging from 0.6 to 1.0 nm and Xd (represented by which

is the ratio of Xd/Cb). Tighter membrane (low rp) provided very high rejection of MgCl2 even at lower permeate fluxes (corresponding to lower P). At higher rp, the rejection of MgCl2 decreases significantly with the rejection below 70% at rp = 1.0 nm even for = 10. For desalination application, the low rejection of MgCl2 is not preferable since the purpose of NF is to remove hard ions such as Mg2+ and Ca2+. Alternatively for loose NF membrane (rp > 1.0 nm), high membrane charge density (exceeding > 10) is required in order to increase the rejection of MgCl2. The prediction of applied pressure (at the corresponding permeate fluxes) as shown in Figs. 2b

A.W. Mohammad et al. / Desalination 206 (2007) 215225


120.00 100.00 Rreal (%) 80.00 60.00 40.00 20.00 0.00 0 2 4 P (bar) Cb=86.82 mM Cb=344.83 mM Cb=172.41 mM DSPM-DE Cb=258.62 mM 6 8 10

219

(a)

30.00 Jv x 10 (m m s )
-1

25.00 20.00 15.00 10.00 5.00 0.00 0

(b) (b)

-6

-2

4 P (bar)

10

Cb=86.82 mM Cb=344.83 mM

Cb=172.41 mM DSPM-DE

Cb=258.62 mM

Fig. 1. Experimental verification of the modified DSPM-DE model using NF-90 membrane; prediction of (a) rejection and (b) permeate flux [15].

5b shows that significantly higher applied pressure is required in order to overcome the osmotic effect of the MgCl2 solutions. The tighter the membrane, the higher P required. In this case, a looser membrane is better since lower P is required to provide a very good permeate flux. But as described above, the rejection of MgCl2 may not fulfill the requirement. Which NF membrane should be chosen for our application? The answer may be obtained after analysing the capital and operating cost of the NF system as will be shown in the following section.

3.2. Estimation of capital and operating cost The question on which NF membrane should be selected for desalination application can be answered by analysing the effect of membrane characteristics on the capital and operating cost of the NF system. It will be assumed that the major requirement for the selection of the NF membrane is that the MgCl2 rejection must be greater than 80%. If the membrane fails to provide the level of rejection, then it will not be considered at all. Based on Figs. 25, only the membranes with the

220
100 90 80 70 R (%) 60 50 40 30 20 10 0 0 20 40 60

A.W. Mohammad et al. / Desalination 206 (2007) 215225


100 90 80 70 R (%)
80 100 120 140 160

=10 =5 =1

60 50 40 30 20 10 0 0 20 40 60 80 100 120 140 160

Jv (L.m -2h-1)

Jv (L.m-2h-1)

(a)
50
60 50 40 P (bar) 30 20 10 0 0 20 40 60 80 100 120 140 160 Jv (L.m -2h-1)

(a)
45
=10 =5 =1

40 35 P (bar) 30 25 20 15 10 5 0 0 20 40 60 80
-2 -1

=10 =5 =1

pure water

pure water 100 120 140 160

J v (L.m h )

(b) Fig. 2. Simulation results from DSPM-DE model for MgCl2 for rp = 0.6 nm and varying , (a) predicted rejection, and (b) predicted applied pressure.

(b) Fig. 3. Simulation results from DSPM-DE model for MgCl2 for rp = 0.7 nm and varying , (a) predicted rejection, and (b) predicted applied pressure.

characteristics as shown in Table 2 can be considered for selection. Based on the results shown in Table 2, the membrane with rp = 0.6 will meet the rejection requirement at much lower P. However at this low P, the permeate flux will be quite small resulting in the large number of modules required for the applications. Higher rp (0.7 and 0.8) will meet the rejection requirement at higher applied pressure and higher permeate fluxes and thus the number of modules will also be significantly reduced as compared to the membrane of rp = 0.6 nm.

It is also possible to consider the tighter membrane, rp = 0.6 and rp = 0.7, but to be operated at higher pressure in order to provide much higher permeate flux. As shown in the last five rows of Table 2, the rejection obtained at this higher P will be higher than the minimum requirement of 80%. The applied pressure is significantly higher but the number of modules will be minimal. A detailed analysis of the capital and operating cost of the system will provide a better justification for the selection of the best membranes. Table 3 shows the detailed breakup of the capital and

A.W. Mohammad et al. / Desalination 206 (2007) 215225


100 90 80 70 R (%) 60 50 40 30 20 10 0 0 20 40 60 80 100
-2 -1

221

100 90
=10 =5 =1

80 70 R (%) 60 50 40 30 20 10 0
=10 =5 =1

120

140

160

20

40

60

80

100
-2 -1

120

140

160

Jv (L.m h )

J v (L.m h )

(a)
40 35 30 P (bar) 25 20 15 10 5 0 0 20 40 60 80 100 120 140 160 Jv (L.m-2h-1) pure water

(a)
40
=10 =5 =1

35 30 P (bar) 25 20 15 10 5 0 0 20 40 60 80 100
-2 -1

=10 =5 =1

pure water 120 140 160

Jv (L.m h )

(b) Fig. 4. Simulation results from DSPM-DE model for MgCl2 for rp = 0.8 nm and varying , (a) predicted rejection, and (b) predicted applied pressure.

(b) Fig. 5. Simulation results from DSPM-DE model for MgCl2 for rp = 1.0 nm and varying , (a) predicted rejection, and (b) predicted applied pressure.

operating cost based on the Verbenne cost model. The major cost contributor will be due to the applied pressure of the NF membrane system. As shown in Table 3, the lowest overall cost is obtained for case 8, which is for NF membrane of rp = 0.8 and = 10. This is similar to our previous finding [14] which concluded that a looser membrane would result in the lowest system cost provided that the membrane can meet the rejection requirement of the ions. In the same study it was also concluded that the membrane charge density will only affect the degree of rejection but will

not contribute to the effect on the overall cost of the system. However this may not be true for the case considered in this current study. As shown for cases 46 in Table 3, the variation of Xd or can have an impact on the overall cost. In this case, the membrane with = 5 provides the most optimum cost as compared to those with = 1 and = 10. This is an interesting finding which requires a more detailed analysis in the future work. Another observation based on Table 3 is that the increase in applied pressure with a corresponding increase in permeate flux and rejection

222

A.W. Mohammad et al. / Desalination 206 (2007) 215225

Table 2 Membranes that fulfilled the rejection requirement


Case 1 2 3 4 5 6 7 8 9 10 11 12 13 rp (nm) 0.6 0.6 0.6 0.7 0.7 0.7 0.8 0.8 0.6 0.6 0.6 0.7 0.7 = Xd/Cb 1 5 10 1 5 10 5 10 1 5 10 5 10 Rejection (%) 80 80 80 80 80 80 80 80 88 94 96 88 91 P (bar) 21 12 10 41 20 15 34 24 55 56 56 43 43 Jv (Lm2h1) 44 17 11 144 54 35 144 87 144 144 144 144 144 Number of modules required 1525 3959 5814 463 1224 1918 463 769 463 463 463 463 463

(cases 913) will not result in a lower overall cost for the system. Thus unless the higher rejection is favourable (compared to the minimum 80% rejection), cases 913 should not be considered for selection at all. 4. Conclusions Based on this preliminary study on the impact of the NF membrane characteristics on the cost of the NF membrane system, the following conclusions can be made: The modified DSPM-DE model is a reasonably good model to predict the rejection and permeate fluxes of single salt solution at higher concentration when the osmotic effect is significant. Based on the simulation study for MgCl2 solution, a membrane with tighter pores (0.60.7 nm) could provide the required high rejection of MgCl2 but the rejection will be much lower for a loose membrane (0.81.0 nm). The overall cost study shows that the lowest system cost will be obtained for the membrane with the largest pore size provided that the required minimum rejection can be achieved.

The effective charge density may also affect the overall cost and it should be considered as well for the membrane selection. Acknowledgement This work has been accomplished as part of a research project sponsored and funded by the Middle East Desalination Research Center (MEDRC). Symbols ci Ci,b Ci,w Ci,p Ccapital Cchemical ccivil Cdeprec celectro Concentration in membrane, mol m3 Concentration in the bulk solution, mol m3 Concentration at the membrane wall solution, mol m3 Concentration in permeate, mol m3 Capital cost, Cost required by a total of chemical filtered, Civil investement cost, Depreciation rate upon investment cost, Electrochemical cost,

Table 3 Detailed analysis of the capital and operating cost of the NF membrane system for different cases
Maintenance (106 ) 0.281 0.409 0.521 0.268 0.259 0.289 0.247 0.238 0.304 0.306 0.307 0.271 0.273 0.561 0.819 1.042 0.535 0.518 0.579 0.494 0.477 0.608 0.613 0.614 0.543 0.546 (106 ) Operation Total oprt (106 ) 3.1712 3.8147 4.6102 3.8653 2.9597 3.0064 3.4016 2.9245 4.6849 4.7414 4.7602 3.9533 3.9873 Total cost 17.20 24.28 30.66 17.24 15.90 17.48 15.75 14.85 19.88 20.06 20.12 17.53 17.64

Case

Civil

Mechn.

(106 ) 1.525 3.959 5.814 0.463 1.224 1.918 0.463 0.769 0.463 0.463 0.463 0.463 0.463 14.026 20.465 26.049 13.378 12.938 14.470 12.350 11.923 15.195 15.320 15.362 13.573 13.648 1.004 1.657 2.200 0.863 0.903 1.069 0.794 0.794 0.984 0.992 0.995 0.876 0.881 1.326 0.930 0.847 2.200 1.280 1.069 1.866 1.415 2.789 2.830 2.844 2.263 2.288

Electrotech. (106 ) (106 )

Membrane Total Depreciation Consumption installation capital (106 ) (106 ) (106 ) (106 )

A.W. Mohammad et al. / Desalination 206 (2007) 215225

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

4.044 7.060 9.359 2.729 3.672 4.531 2.729 3.108 2.729 2.729 2.729 2.729 2.729

4.174 6.384 8.069 3.210 3.901 4.531 3.210 3.488 3.210 3.210 3.210 3.210 3.210

4.283 3.062 2.808 6.976 4.141 3.490 5.949 4.558 8.793 8.919 8.960 7.172 7.247

223

224

A.W. Mohammad et al. / Desalination 206 (2007) 215225 123129. [4] M. Al-Sofi. Seawater desalination SWCC experience and vision. Desalination, 135 (2001) 121 139. [5] R. Levenstein, D. Hasson and R. Semiat, Utilisation of the Donnan effect for improving electrolyte separation with nanofiltration membranes. J. Membr. Sci., 116 (1996) 7792. [6] M. Perry and C. Linder, Intermediate RO UF membranes for concentrating and desalting of low molecular weight organic solutes. Desalination, 71 (1989) 233245. [7] T. Tsuru, S. Nakao and S. Kimura, Calculation of ion rejection by extended NernstPlanck equation with charged reverse osmosis membranes for single and mixed electrolyte solutions. J. Chem. Eng. Japan, 24 (1991) 511517. [8] J. Schaep, C. Vandecasteele, A.W. Mohammad and R. Bowen, Modelling the retention of ionic components for different nanofiltration membranes. Separ. Purif. Technol., 2223 (2001) 169179. [9] W.R. Bowen, A.W. Mohammad and N. Hilal, Characterization of nanofiltration membranes for predictive purposes use of salts, uncharged solutes and atomic force microscopy. J. Membr. Sci., 126 (1997) 91105. [10] W.R. Bowen and J.S. Welfoot, Modelling the performance of membrane nanofiltration critical assessment and model development. Chem. Eng. Sci., 57 (2002) 11211137. [11] S. Bandini and D. Vezzani, Nanofiltration modeling: the role of dielectric exclusion in membrane characterization. Chem. Eng. Sci., 58 (2003) 3303 3326. [12] A.W. Mohammad, A modified Donnan-Steric-Pore model for predicting flux and rejection of dye/NaCl mixture in nanofiltration membranes. Separ. Sci. Technol., 37(5) (2002) 10091030. [13] A.W. Mohammad and M.S. Takriff, Predicting flux and rejection of multicomponent salts mixture in nanofiltration membranes. Desalination, 157 (2003) 105112. [14] N. Ali, A.W. Mohammad and A.L. Ahmad, Use of nanofiltration predictive model in developing criteria for membrane selection and analyzing cost impact. Separ. Purif. Technol., 41 (2005) 2937. [15] A.W. Mohammad, N. Hilal, H. Al-Zoubi and N. Darwish, Prediction of rejection and permeate fluxes for highly concentrated salt solutions. J. Membr. Sci.,

Cost associated with energy for pumping, Cinstall Cost associated with installation of equipment, Cmaint Maintenance cost, Mechanical equipment cost, cmech Cmembrane Cost for membrane installation, Cost associated with quality control, Cqc Hindered diffusivity, m2 s1 Di,p Di, Bulk diffusivity, m2 s1 F Faraday constant, C mol1 Jv Permeate flux (based on membrane area), m3 m2 s1 Ki,c Hindrance factor for convection Hindrance factor for diffusion Ki,d n Number of membrane module P Applied Pressure difference, bar Osmotic pressure difference, bar Flowrate, m3 s1 Qf rp Effective pore radius, m Gas constant, J mol1 K1 Rg R Real rejection, % T Absolute temperature, K x Distance normal to membrane, m x/Ak Ratio of effective membrane thickness over porosity, m Effective membrane charge, mol m3 Xd W Born solvation energy barrier, J Valence of ion zi Donnan potential difference, V Energy efficiency References
[1] L.P. Raman, M. Cheryan and N. Rajagopalan, Consider nanofiltration for membrane separation. Chem. Eng. Prog., 90 (1994) 6874. [2] P. Eriksson, Nanofiltration extends the range of membrane filtration. Environ. Prog., 7 (1988) 58 61. [3] M. Al-Sofi, A. Hassan, G. Mustafa, A. Dalvi and M. Kither, Nanofiltration as a means of achieving higher TBT of 120C in MSF. Desalination, 118 (1998)

Cenergy

A.W. Mohammad et al. / Desalination 206 (2007) 215225 submitted for publication. [16] A.J.P. Verberne and J.W. Wouters, Membranfiltratie voor de drinkwaterbereiding: economische optimalisatie van ontwerpparameters. H2O, 26(14) (1993) 383387. [17] J.S. Taylor, L.A. Mulford, S.J. Duranceau and W.M. Barrett, Cost and performance of a membrane pilot plant. J. AWWA, 81(11) (1989) 5260. [18] G.W. Dunkelberger and J.S. Taylor, Cost optimiza-

225

tion model for low pressure membrane applications. Proc. American Water Works Association, Membrane Process Conf., 1991, pp. 6378. [19] V.D. Bruggen, D. Everaert, D. Wilms and C. Vandecasteele, Application of nanofiltration for removal of pesticides, nitrate and hardness from ground water: rejection properties and economic evaluation. J. Membr. Sci., 193 (2001) 239248.

Das könnte Ihnen auch gefallen