Performance comparison of ultrafiltration, nanofiltration and reverse osmosis on whey treatment

Transcription

1 Desalination 9 (8) 1 Performance comparison of ultrafiltration, nanofiltration and reverse osmosis on whey treatment M.S. Yorgun a, I. Akmehmet Balcioglu b *, O.Saygin b a Michigan State University, Department of Civil and Environmental Engineering, East Lansing, MI 88, USA b Bogazici University, Institute of Environmental Sciences, Bebek, Istanbul 33, Turkey Tel. +9 (1) ; Fax +9 (1) 3--77; [email protected] Received 9 July 7; accepted revised 3 September 7 Abstract Treatment of whey, which is a by-product of cheese manufacturing process, has been a significant problem due to its high organic load with 1, mg O /L COD. In this study, treatment of two different types of whey by using different membrane processes namely ultrafiltration (UF), nanofiltration (NF), and reverse osmosis (RO) was investigated to produce cleaner discharge and to recover the proteins in whey for re-use. Membrane modules were tested as one stage operations and cascade operations by employing a combination of membrane modules in series. Nanofiltration, when operated in one stage, produced the best results from the point of treatment capacity, COD removal, and protein recovery. 3.8 L/m h of permeate flux value at transmembrane pressure (TMP) of 8 bar was reached with nanofiltration which produced permeate with COD load of,787 mg O /L, and the protein rejection was 88%. Additionally, the influent whey was concentrated.8 times its original volume. Among the applied cascade operations, the NF + RO combination produced the best results. Another achievement of this combination is its capability of recovering both protein and lactose separately, protein recovery in the first stage and lactose recovery in the second stage. Keywords: Whey; Nanofiltration; Ultrafiltration; Reverse osmosis; COD reduction; Volume reduction 1. Introduction Whey is a by-product of cheese production which is used mainly as animal feed or released into the wastewater treatment process, although it is rich in valuable components. It contains lactose, minerals (e.g., calcium, magnesium, *Corresponding author. phosphorus), vitamins, non-casein protein (except glycomacropeptide), and traces of milkfat [1,]. Because of its content of organic compounds, whey can not be discharged to receiving environments. It is therefore necessary to process the whey even it may be uneconomic. Also, when its considered that on cheese making about half of the total milk finds its way into the whey, it is 11-91/8/$ See front matter 8 Published by Elsevier B.V. doi:1.11/j.desal.7.9.8

2 M.S. Yorgun et al. / Desalination 9 (8) 1 more understandable that the processing of whey and in particular its organic constituents is regarded as very important [3]. Therefore recovery of valuable compounds in whey such as protein and lactose has received intense attention recently. In Turkey, major dairy companies that process 19% of Turkish raw milk are present, whereas there are hundreds of so called mandras possessing small traditional dairy processors which generate huge amounts of whey. Whey treatment processes include traditional techniques such as evaporating and drying which are widely employed in Turkish companies. These processes do not contribute to recovery of valuable products in whey. These methods are used to remove some part of the water in whey to diminish the volume and to enhance the keeping quality. Anaerobic treatment is another process employed for organics removal from whey. This process is preferred instead of conventional aerobic wastewater treatment since cheese whey has a very high organics content ( 8 g COD/L) and may impair biomass granulation during biological treatment. This would in turn result in biomass wash-out []. Further purification of whey can be achieved via ion exchange, affinity chromatography and selective precipitation []. Recent developments [,] in membrane filtration have provided exciting new opportunities for large-scale whey treatment to produce cleaner discharge as well as protein and lactose fractionation. Recently, a few studies of material recovery from whey by nanofiltration, comparison of ultrafiltration and nanofiltration for the utilization of whey protein and lactose, and development of complex membrane systems for whey processing have been conducted [7 11]. One-stage and twostage operations of NF and RO were also investigated for dairy wastewaters [1]. However, no published report has been available on white cheese whey recovery, which can be accepted as the utmost consumed cheese product in Turkey. The present study reports on the application of 7 different membrane modules (1 UF, NF, RO) to produce cleaner discharge as permeate to be treated with conventional methods and to recover whey proteins from the white cheese whey in its retentate. Treatment of curd cheese whey was also investigated. Volume reduction and consequent economical benefit due to less volumetric feed for evaporating and drying processes will also be achieved. Each membrane module was tested as one stage operations and different combinations of these modules were tested as cascade operations. The main focus of the study was to investigate and obtain success in one stage NF performance of whey treatment and protein recovery since employing one stage would be economically beneficial for full-scale applications.. Experimental.1. Membrane modules Four different nanofiltration modules, one ultrafiltration module, and two reverse osmosis modules (all commercially available) were used during the experiments. The specifications of each module are given in Table 1. All modules are in spiral-wound configuration... Cleaning agents Membranes were reused after each experiment with ensuring no fouling remained in the module and the module would act as a virgin membrane in each experiment. The cleaning efficiency is checked by measuring the pure water flux of the modules which are given in Table. The pure water flux of modules measured after cleaning indicated very small differences than the initial values ensuring effective cleaning. Three commercially available cleaning agents were used for the clean in place (CIP) procedure. The procedure consisted of the initial acidic cleaning (Ultrasil 7) followed by alkaline cleaning (Ultrasil 11) then the final step of disinfection (Ultrasil ). The cleaning agents were provided by Henkel Corp.

3 M.S. Yorgun et al. / Desalination 9 (8) 1 Table 1 The specifications of membrane modules Module/type Supplier Material/P max (bar)/t max (C ) Nominal solute rejection (%) Module area (m ) Module length (mm) Cartridge diameter (mm) Feed spacer (mm) A/NF Nadir GmbH. Polyethersulfone// Na SO B/NF Nadir GmbH Polyethersulfone// 8 9 Na SO C/NF Trisep Corp. Polyamide-urea/1/ 9 MgSO D/NF Filmtec Polysulfone/.8/ E/UF Nadir GmbH Polyethersulfone// F/RO Trisep Corp. Polyamide-urea/1/ 99. NaCl G/RO Trisep Corp. Polyamide-urea/1/ 99. NaCl MWCO of module E = kda Module D rejects organics with a molecular weight above while passing monovalent salts

4 M.S. Yorgun et al. / Desalination 9 (8) 1 7 Table Operational parameters of modules used in the experiments Membrane module TMP (bar) RFR (L/h) Pure water flux(l/m /h) Nanofiltration A, 3 B 8, C 8, D 8, 1 Ultrafiltration 3, 1 Reverse osmosis 1 3, Apparatus and procedure The system by which the experiments were conducted, consisted of a L feed tank, stainless steel membrane housing, feed and recycling pumps, pressure indicators, and flow meters (Fig. 1) [13]. The feed was pumped to the membrane module by means of a feed pump (3), which was the main pump that provided high pressure to the system (Wilo). V was an emergency valve to discharge the water in the tank. V3 and V were used to adjust the pressure of the feed pump (3). V3 was used for rough and V was used for fine ad- Fig. 1. Schematic diagram of the membrane system.

5 8 M.S. Yorgun et al. / Desalination 9 (8) 1 justment of the inlet pressure. V adjusted the inlet pressure of the recirculation pump () which was employed to create sufficient cross-flow velocity through the membrane. The transmembrane pressure was adjusted by throttling valve V7... Feed solutions Experiments were conducted by using white cheese whey and curd cheese whey as feed for membrane modules. The original whey samples used in this study were obtained from a company which uses 1, L raw milk for one cycle production of white cheese and generate approximately 78, L of whey. This company has 3 cycles/week for each of the four cheese types manufactured. Characteristics of both solutions are given in Table 3... Filtration experiments One stage and cascade operations of membrane modules were investigated under different operational conditions, namely different transmembrane pressures and different recirculation flow rates for comparison purposes. The pure water fluxes of each membrane module were determined before the experiments in order to have a reference value for flux. The deviation of pure water flux after the experiment gives a clear indication of membrane fouling and need for chemical cleaning. Simple sieving was applied for feed solutions Table 3 Characteristics of curd and white cheese whey Curd cheese whey White cheese whey Fat content, %. COD, g/l Approx. 1 Approx. 11 Protein, g/l 1..3 Lactose, g/l Minerals, g/l.1. before membrane treatment in order to remove coarse protein particles that would cause severe membrane fouling. The permeate stream of each experiment was steadily removed and collected separately but the concentrate stream was recycled back to the feed tank in order to ensure maximum solute concentration. Different arrangements were tested as cascade operations. Two stage and three stage operations were analyzed, the permeate stream of one module was filtered by the consequent module. In the cascade operation with heating, the curd cheese whey was first heated up to 1 o C in order to precipitate the protein portion. The curd cheese whey was cooled and the de-solubilized protein molecules were left to be settled. The supernatant was then used as the feed stream of reverse osmosis module G. The effectiveness of all membrane processes was determined by measuring the permeate flux during the experiments, calculating the protein rejection and the COD removal of each membrane module. The protein rejection percentage (R %) of each module was calculated by: 1 C p R% = 1 (1) C where C and C P are protein concentrations in the feed and permeate respectively. Volume reduction ratios () obtained by the membrane modules were calculated by: V V = = V t V V t R () () P () where V is the initial feed volume, V R (t) and V P (t) are the retentate and permeate volumes at t time, respectively. COD and total Kjeldahl nitrogen (TKN) were determined in accordance with the Standard Methods [1]. In order to calculate the protein fraction of samples, the TKN value was than multiplied by the coefficient. [1].

6 M.S. Yorgun et al. / Desalination 9 (8) Results and discussion 3.1. One stage concentration operations of membrane modules Permeate flux of ultrafiltration, nanofiltration and reverse osmosis modules In the first part of the study NF, UF, and RO membrane modules were operated as a single stage. Operational parameters of the modules are presented in Table. The operational TMP and recirculation flow rate (RFR) for nanofiltration modules B, C, and D were lower than those for module A due to higher solute rejection values of these modules (Table 1). The recycle flow rate created the cross flow of the whey through the membrane surface. In order to ensure good scouring of the protein molecules deposited on the membrane surface, high RFR was used for all membrane modules. The variation of the permeate flux and with time for four different nanofiltration modules is given in Fig. for curd and white cheese whey. The initial permeate flux values for white cheese whey were slightly lower than those for curd cheese whey due to the fat content of white cheese whey. The initial permeate flux values of modules A, B, C, and D were 3,, 3.8, and Permeate Flux (L/m h) Permeate Flux (L/m h) pflux(curd) pflux(white) (white) (curd) Module B Operation Time (min) Operation Time (min) Permeate Flux (L/m h) Operation Time (min) Permeate Flux (L/m h) Module D Operation Time (min) Fig.. Permeate flux and vs. time graphs for nanofiltration modules A, B, C, and D.

7 1 M.S. Yorgun et al. / Desalination 9 (8) 1 3. L/m.h, respectively for curd cheese whey. As can be seen from Fig., the permeate flux values declined mildly over time for modules A, B, and C. This decline in the permeate flux was well expected due to the gel layer formation on the membrane surface as the feed concentration increased and deposition occured. For pressuredriven processes, the greater the flux, the greater the build-up of the solute at the interface; the greater the solute build-up, the higher the concentration gradient; the steeper the concentration gradient, the faster the diffusion. Under normal steady-state operating conditions there is a balance between those forces transporting the water and constituents within it towards, through and away from the membrane. This balance is determined by concentration polarization (CP). CP also raises the effect of osmotic pressure at the membrane solution interface, increasing the required transmembrane pressure for operation. It is thus always desirable to suppress CP by promoting turbulence and/or operating at a flux below that at which CP starts to be significant [1]. However, these mild declines show that no extreme deposition occurred on the surface of these modules to create concentration polarization, thus severe fouling. But this is not the case for module D treating white cheese whey. As can be seen from Fig., a steeper decline occurred in the first min of the experiment. is an important parameter for concentration of whey and it is chosen depending upon the desired protein and lactose content in the final retentate. The values obtained over time for modules A, B, C, and D are also presented in Fig.. The highest (.8) was reached by module C in the shortest period of time (3 min). The permeate flux values show a steep decline while concentrating whey times of its original concentration. Further concentration of whey did not affect the permeate flux as much. In this study, modules A, B, C, and D all produced better permeate flux values in different values when compared to previous studies where NF, together with MF as pretreatment, produced 1.3 L/m h of permeate flux at = [7,9]. Although the operating TMP values for ultrafiltration for various applications are given as 1.. bar [1], in this study operating TMP of ultrafiltration module E was chosen as 3 bar (Table ). This low value was set to achieve comparison with the test conditions implied by the manufacturer. The experiment with module E produced a very high overall permeate flux value (3 L/h) due to its larger membrane area. However, the permeate flux calculated per unit area of the membrane, was smaller compared to the nanofiltration modules. There was a mild decline for the permeate flux values by the operation time as seen in Fig. 3. No critical drop of the permeate flux between values of 1 and was observed. A high value of 7.8 was reached in a very short period of time, namely 13 min. During the treatment of highly concentrated wastewaters such as whey, the gel layer that is responsible for concentration polarization is highly concentrated, and when the inlet pressure is high, this layer gets even denser thus quickly fouling the membrane [1]. Therefore in this study for reverse osmosis module F the chosen pressure was low (Table ) although the maximum allowable pressure was 1 bar. In Fig. variation of the permeate flux and are presented as a function of the operation time. Fig. indicates that the initial permeate flux has a very low value (1. L/m h) compared to the ultrafiltration and nanofiltration modules. Moreover the permeate flux of module F dropped to. L/m h with a very sharp decline in the first min while concentrating whey 1. times of its original concentration. Further concentration of whey did not affect the permeate flux as much. This phenomenon shows that the membrane is quickly fouled in the first min, and no significant effect of on the permeate flux was observed after min of operation. This result reveals that the limiting flux of the membrane was reached where no effect of TMP on the permeate

8 M.S. Yorgun et al. / Desalination 9 (8) 1 11 Permeate Flux (L/m h) 1 1 Module E pflux(white) (white) Operation Time (min) Fig. 3. Permeate flux and vs. time graph for ultrafiltration module E. Permeate Flux (L/m h) Module F pflux(curd) (curd) Operation Time (min) Fig.. Permeate flux and vs. time graph for reverse osmosis module F. flux existed due to the gel layer formed on the membrane surface as a result of fouling. The occurrence of fouling affects the performance of the membrane either by deposition of a layer onto the membrane surface or by blockage or partial block- age of the pores. This changes the effective pore size distribution [17]. In our experiment, the unaltered permeate flux can be attributed both to layer formation and blockage due to fouling.

9 1 M.S. Yorgun et al. / Desalination 9 (8) Cascade concentration operations of membrane modules To achieve both satisfactory operational performances and clear permeate streams, combinations of ultrafiltration, nanofiltration and reverse osmosis modules were tested. Among the various combinations tested using available modules, NF + RO treatment of curd cheese whey turned out to be the best combination. In this two stage operation, nanofiltration membrane module A was chosen as the first stage and the permeate stream of module A was subjected to reverse osmosis module F. The permeate flux values were given the most importance while making the choice of which module was going to be the initial stage, and as can be seen from Fig., module A, due to its higher pore size, had the most ease of operation from the point of view of TMP and the consequent permeate flux. As can be seen in the one stage operation, this module behaved as an ultrafiltration module rather than nanofiltration one due to the high lactose content in the permeate. The operating transmembrane pressure (8 bar) and the recycle flow rate (, L/h) were kept low compared to the one stage operation with the same module. Since the permeate stream of module A had a low protein content this reduced the organics load in the influent water, so low pump energy could create a more economically feasible situation. As can be seen in Fig., reverse osmosis module F produced nearly the same results as in its one stage operation experiment. The initial permeate flux was 1. L/m h, and it showed a steep decline in the first min. However, the values did not change at all until the end of the experiment. The last permeate flux value recorded, which could be accepted as the limiting flux, was.9 L/m /h for this experiment since no significant flux change occurred. High lactose concentration in the influent stream may be the reason of the results obtained. Adjusting the operating parameters, such as TMP and recycle flow rate, would not be the remedy of this situation since higher TMP would cause the gel layer formed by lactose to be more concentrated on the membrane surface [17]. The recycle flow rate can be adjusted to a very high level to overcome this situation, however this would cause increased energy consumption that would make this operation economically unfeasible. The influent stream of re Module A+E. Permeate Flux (L/m /h) pflux(nfpermeate) (NFpermeate) Operation Time (min.) Fig.. Permeate flux and vs. time graph for reverse osmosis module F (NF + RO operation).

10 M.S. Yorgun et al. / Desalination 9 (8) 1 13 verse osmosis module F was low in both protein and lactose content and could be employed only for further COD removal, not the concentration purposes. Heating is an old technique which is being widely used for precipitating the curd fines in the whey. In this part this technique is combined with RO. The production of curd cheese whey involves heating the product up to 9 C and collecting the precipitate as the curd cheese. The remaining part is called the curd cheese whey which is also rich in protein content. In this two stage operation, heating the curd cheese whey up to 1 C in order to precipitate the protein portion was applied as the first stage. The supernatant was than used as the feed stream of reverse osmosis module G. The curd cheese whey, after heating up to 1 C, was cooled, and the de-solubilized protein molecules were left to be settled. Since the temperature applied in the first stage operation exceeded the denaturating temperature of proteins ( C), the proteins in the curd cheese whey were de-solubilized losing their structure and leaving a clear supernatant. However, the nutritional value of the precipitated product remains the same, so it can be used further for animal feed or as raw material [18]. L of curd cheese whey, after heating, produced 3 L of precipitate with a protein concentration of 31 mg/l. The total solid content of the precipitate was 11 g/l. By precipitating the proteins, the organics load of the curd cheese whey was decreased. The COD removal efficiency of the first stage heating operation was 3% producing a supernatant with the COD value of 8,7 mg O /L. This supernatant was then treated with reverse osmosis module F. The operating transmembrane pressure was 1 bar and the recycle flow rate was, L/h. These values were set higher than those for module F in the NF + RO operation, since module G possessed a combination of properties inherent to its structure that were uniquely advantageous in significantly resisting colloidal particulate fouling and biofouling, therefore applying higher TMP would not create a much thicker gel layer on the surface. Module G showed somewhat better permeate flux values when compared to other reverse osmosis module F treating NF permeate. As can be seen in Fig., the initial permeate flux value was 1. L/m h. This can be attributed 18. Permeate Flux (L/m /h) Module F pflux(supernatant) (Supernatant) Operation Time (min.). Fig.. Permeate flux and vs. time graph for reverse osmosis module G (heating + RO operation).

11 1 M.S. Yorgun et al. / Desalination 9 (8) 1 to the TMP difference between the two experiments (the operating TMP for reverse osmosis module E was 8 bar). However it should be noted that the influent stream of module F was organically more loaded, including some protein portion that had not precipitated in the heating stage. L of feed water were finished in a relatively short time ( min) and the permeate flux value at the end of the experiment was 1.3 L/m h. The limiting flux was reached at the th min COD removal and protein rejection efficiencies The permeate COD values and % COD rejection for modules A, B, C, D, E, F, and cascade operation A + F are given in Fig. 7. As can be seen in Fig. 7, the best COD removal was achieved by module D. The permeate stream of module D treating influent white cheese had the COD load of.787 mg O /L. Very poor COD removal were achieved by modules A, B, and C. These modules behaved as ultrafiltration modules, leaving the lactose content in the permeate stream (the yellowish color of the permeate streams was the evidence while the permeate stream of module D was clear). This also explains the steep permeate flux decline during the first min of the experiment conducted by module D. The additional lactose concentration on the membrane surface increased the CP thus reducing the flux. The protein rejection percentages for module A for curd cheese whey was 87% and for white cheese whey it was 83% at =.. In the experiment conducted by module B with curd cheese whey as the feed, the protein rejection was 83% at =.8, the rejection for white cheese whey was 81% for the same. Module C produced slightly higher results (88%) for both whey types at =.8. The best rejection among the nanofiltration modules was obtained by module D (9%) for white cheese whey for = 3.7. Rejection percentages for modules E, F and cascade operations with A+F were 78%, 9% and 9% respectively. As can be seen from Fig. 7, high protein rejec- Modules COD (mg/l), 1, 1,, Heat.+G 7, 89.7 % A+F, 9.38 % F 3, % E,331.8 % D, % C 3, 77.1 % B A 37,8 3,8. 8. % % removal COD 8 1 COD removal (%) Fig. 7. The permeate COD values and % COD removal for each module including cascade operations.

12 M.S. Yorgun et al. / Desalination 9 (8) 1 1 tion did not contribute to high COD removal since lactose was the main component of whey that was responsible for high COD. Theoretically, protein removal reduced by only 1% of the whey COD [19]. The COD removal efficiency of the ultrafiltration module E was low. The module produced the permeate stream with the COD value of,331 mg O /L. The protein rejection of this module appeared to be 78%. This value is slightly smaller when compared to the rejections of the nanofiltration modules since the pore size of module E was higher than that of the nanofiltration modules. The COD removal efficiency of module F was again quite high. The module produced the permeate stream with the COD value of 3,87 mg O /L. This COD value does not comply with usual RO effluent values. Typical COD values of reverse osmosis permeate are generally 1, mg O /L []. Protein rejection of this membrane module was very good, 9% at = 1.7. The permeate stream was clear without any indication of lactose content. As for the cascade operations, the COD removal efficiency of module F in NF+RO operation was high, 9%, as expected. The COD value of the permeate of this second stage was, mg O /L. In the heating + RO operation, the COD removal efficiency was lower when compared to that of reverse osmosis module F, namely 89%, while producing a permeate stream with the COD value of 7, mg O /L. As implied in the experimental section, the protein retentions of modules were evaluated via TKN measurements. Due to the nature of this method, nitrogen compounds other than proteins (such as amino acids) can also be detected. Consequently, the incomplete protein rejection of the RO modules in this study can be attributed to nonprotein compounds in the permeate streams of the modules. Also, higher protein rejection values for NF and UF modules can be expected due to this phenomenon.. Conclusion Although the protein rejection and the permeate flux values of nanofiltration modules A, B, C and ultrafiltration module E were very promising, the COD removal efficiencies of these modules were very low due to poor lactose rejection. Nanofiltration module D produced very good results in terms of permeate COD, but a relatively poor operational performance. Considering the economical benefit of using one stage system, employing nanofiltration module D for recovery of protein and reducing the effluent organics load is an appropriate choice when considered that the effluent stream is to be further concentrated or treated by means of conventional treatment. As well as being economically beneficial due to only one module as one stage is employed, this solution also provides high volume reduction and consequent less load for drying and evaporating that are widely used among Turkish companies. The experimental data also indicated that if further COD removal is aimed by employing membranes only, one additional RO stage is necessary as a polishing step. However, the problem with cascade operations with RO as the second stage is the performance of RO in terms of the permeate flux although the decreased influent load was caused by the first NF stage. The permeate flux achieved by second stage RO module F with this influent was only 1. l/m /h. This value dropped to.9 l/m /h after 13 min of operation which is a very low value. The reason for this is that the maximum transmembrane pressure that the experimental set up could withstand was just over 1 bar, and the transmembrane pressure for the experiment was 8 bar. 8 bar of driving pressure for the RO module while treating high loaded influents would not be enough to exceed the osmotic pressure that the influent solution creates across the membrane. High TMP such as 3 bar should be applied in order to achieve better results in terms of the permeate flux.

13 1 M.S. Yorgun et al. / Desalination 9 (8) 1 Another benefit of using NF + RO cascade treatment with available modules appeared as the recovery of protein and lactose separately. As discussed above, the first stage NF behaved more like an UF module, concentrating only the protein portion and leaving the high lactose content in the permeate. With this permeate stream treated by the second RO stage, lactose was recovered separately producing a clean effluent. This phenomenon is to be further investigated. Atra et al. [8] investigated NF performance as the following step after UF. Vourch et al. [1] investigated NF as the first stage but for treatment of model process waters, whose organics load was very low when compared to whey that was used in this study. Rektor and Vatai [9] investigated the performance of NF for whey treatment as the first stage, but this study obtained better results in comparison. One stage operations of UF and RO were also conducted for information to be used while designing suitable cascade operations. An additional heating step was also investigated. Acknowledgement The authors acknowledge Bogazici University BAP for funding this research ( Y1) and SUTAS A.S. for supplying wastewater. References [1] R.L. Walzem, C.J. Dillard and J.B. German, Whey components: millennia of evolution create functionalities for mammalian nutrition: What we know and what we may be overlooking, Crit. Rev. Food Sci. Nutrition, () () [] W.J. Harper, Biological Properties of Whey Components: A Review, The American Dairy Products Institute. Monograph. American Dairy Product Institute, Chicago, IL,, pp [3] H.G. Kessler, Food Engineering and Dairy Technology, Veglas A. Kessler, Germany, [] G. Mockaitis, S.M. Ratusznei, J.A.D. Rodrigues, M. Zaiat and E. Forest, Anaerobic whey treatment by a stirred sequencing batch reactor (ASBR): effects of organic loading and supplemented alkalinity, J. Environ. Manage., 79() () 198. [] A. Zydney, Protein separations using membrane filtration: New opportunities for whey fractionation, Int. Dairy J., 8(3) (1998) 3. [] G. Brans, C.G.P.H. Schroen, R.G.M. van der Sman and R.M. Boom, Membrane fractionation of milk: state of the art and challenges, J. Membr. Sci., 3 () 3 7. [7] M. Nguyen, N. Reynolds and S. Vigneswaran, By-product recovery from cottage cheese production by nanofiltration, J. Cleaner Production, 11 (3) [8] R. Atra, G. Vatai, E. Bekassy-Molnar and A. Balint, Investigation of ultra- and nanofiltration for utilization of whey protein and lactose, J. Food Eng., 7 () [9] A. Rektor and G. Vatai,. Membrane filtration of mozzarella whey, Desalination, 1 () [1] O. Akoum, M.Y. Jaffrin, L.H. Ding and M. Frappart, Treatment of dairy process waters using a vibrating filtration system and NF and RO membranes, J. Membr. Sci., 3 () [11] B. Balannec, M. Vourch, M. Rabiller-Baudry and B. Chaufer, Comparative study of different nanofiltration and reverse osmosis membranes for dairy effluent treatment by dead-end filtration, Separ. Purif. Technol., () 19. [1] M. Vourch, B. Balannec, B. Chaufer and G. Dorange, Nanofiltration and reverse osmosis of model process waters from the dairy industry to produce water for reuse, Desalination, 17 (). [13] E. Çaloglu, Treatment of wastewater containing emulsified oil, M.Sc. thesis, Bogaziçi University, 3. [1] APHA/AWWA/WPCF, Standard Methods for the Examination of Water and Wastewater. 19th ed., American Public Health Association, Washington, DC, 199. [1] S. Kirdar, Analysis Methods for Milk and Products, Süleyman Demirel University, Isparta, 1. [1] S. Judd and B. Jefferson, Membranes for Industrial Wastewater Recovery and Re-Use, Elsevier Advanced Technology, UK, 3. [17] R.W. Field, D. Wu, J.A. Howell and B.B. Gupta, Critical flux concept for microfiltration fouling, J. Membr. Sci., 1 (199) 9 7. [18] N. Qureshi and G.J. Manderson, Bioconversion of renewable resources into ethanol: an economic evaluation of selected hydrolysis, fermentation and membrane technologies, Energy Sources, 17() (199) 1. [19] R.K. Scopes, Protein Purification: Principles and Practice, 3rd ed., Springer-Verlag, New York, 199, pp [] G.M.I. Siso, The biotechnological utilization of cheese whey: a review, Biores. Technol., 7 (199) 1 11.