Membrane fouling mechanisms during ultrafiltration of skimmed coconut milk

Membrane fouling mechanisms during ultrafiltration of skimmed coconut milk

Accepted Manuscript Membrane Fouling Mechanisms during Ultrafiltration of Skimmed Coconut Milk Ching Yin Ng, Abdul Wahab Mohammad, Law Yong Ng, Jamali...

2MB Sizes 8 Downloads 135 Views

Accepted Manuscript Membrane Fouling Mechanisms during Ultrafiltration of Skimmed Coconut Milk Ching Yin Ng, Abdul Wahab Mohammad, Law Yong Ng, Jamaliah Md. Jahim PII: DOI: Reference:

S0260-8774(14)00246-5 http://dx.doi.org/10.1016/j.jfoodeng.2014.06.005 JFOE 7821

To appear in:

Journal of Food Engineering

Received Date: Revised Date: Accepted Date:

27 January 2014 18 April 2014 8 June 2014

Please cite this article as: Ng, C.Y., Mohammad, A.W., Ng, L.Y., Jahim, J.M., Membrane Fouling Mechanisms during Ultrafiltration of Skimmed Coconut Milk, Journal of Food Engineering (2014), doi: http://dx.doi.org/ 10.1016/j.jfoodeng.2014.06.005

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Nomenclature UF MWCO PSF VCO SCM R2 Jvt Jvo c s i cf Kc Ks Kck Ro Rc Rr Vo t

Ultrafiltration Molecular weight cut-off (kDa) Polysulfone Virgin coconut oil Skimmed coconut milk Coefficient of determination Volume flow at a particular time (m3/m2.h) Initial volume flow rate (m3/m2.h) Constant of the complete blocking model (s-1) Constant of the standard blocking model (s-1/2m-1/2) Constant of the intermediate blocking model (m-1) Constant of the cake formation model (sm-2) Membrane surface area blocked per unit of permeate volume (m-1) Reduction in the cross-sectional area of the pores per unit of permeate volume (m-1) Area of cake formed per unit of permeate volume (m-1) Resistance of the fresh or clean membrane (m-1) Resistance of the cake layer formed (m-1) Rr=Rc/Ro, Ratio of resistance of the cake over the resistance of the clean membrane (dimensionless) Initial mean velocity of the filtrate through the membrane (m/s) Time (s)

Membrane Fouling Mechanisms during Ultrafiltration of Skimmed Coconut Milk Ching Yin Nga, Abdul Wahab Mohammada,b,*, Law Yong Nga, Jamaliah Md. Jahima, a

Department of Chemical and Process Engineering, Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Selangor, Malaysia b Research Centre for Sustainable Process Technology (CESPRO), Faculty of Engineering and Built Environment, Universiti Kebangsaan Malaysia, Selangor, Malaysia *Corresponding author. Tel: +603-89216102; Fax: +603-89252546 E-mail: [email protected] ABSTRACT Ultrafiltration is a promising technique to produce value-added products from skimmed coconut milk. Unavoidably, membrane fouling always hinders the membrane performance. In this work, Hermia’s models were used to investigate the fouling mechanisms. Effects of the molecular weight cut-off of the polysulfone membrane (10, 20 kDa), feed solution temperature (50, 55, 60°C) and operating pressure (1.8, 2.0, 2.2, 2.4 bar) towards the membrane fouling were analysed. The results showed that the best fit (R2≥0.98) of the experimental data to all fitted fouling mechanisms (complete blocking, standard blocking, intermediate blocking and cake formation) occurred for experiments using a 20 kDa polysulfone and 60°C feed temperature. All fouling mechanisms were present during the ultrafiltration but dominated by complete blocking, followed by standard, intermediate blocking and cake layer formation. The characteristics of the membrane and feed solution were found to be highly influential on the membrane fouling mechanisms in this study. Keywords: Fouling mechanisms; membrane; ultrafiltration; skimmed coconut milk. 1. Introduction Coconut (Cocos nucifera L.) is an important palm species due to its high oil content (around 37–50% oil by weight of coconut kernel) (Santoso et al., 1996). Thus, virgin coconut oil (VCO) has become the major commercial coconut product and has gained in popularity in recent times (Marina et al., 2009). During the production of VCO, numerous by-products are discarded such as skimmed coconut milk (SCM), soluble proteins, etc. The abundance of these discarded by-products has contributed to environmental issues. These discarded coconut residuals still contain numerous nutritional compounds such as proteins, carbohydrates, sugars, vitamins and so forth. SCM contains high quality proteins (70% of the total coconut protein) whereas it possesses relatively well-balanced amino acid profiles (Naik et al., 2012). The dominant types of protein in SCM are globulin and albumins. The proteins in the SCM need to be processed using certain reliable and mild processes in order to make them suitable for use as functional foods, dietary supplements and formulated milk. However, the utilization of these by-products to produce value-added products has not yet been given much attention. This proteinaceous solution (SCM) has been chosen as the solution medium in this study. So far, a few attempts have been undertaken to extract the coconut proteins. (Chen and Diosady, 2003; Samson S.J et al., 1971). However, such processes (enzymatic extraction and chemical extraction of coconut proteins) have been reported to be less productive in recovering proteins.

Ultrafiltration (UF) has been employed to concentrate and recover the proteins from skim milk (Makardij et al., 1999), whey proteins (Nigam et al., 2008), soy flour (Krishna Kumar et al., 2004) and even wastewaters (Wu et al., 2013; Wu et al., 2006; Wu et al., 2009). However, too little attention has so far been paid on coconut proteins. A UF process has been used in this study in order to obtain high quality protein concentrates from SCM. Membrane can be used to produce coconut protein concentrates that possess high functional and nutritional characteristics. The UF membrane performs based on a size-sieving mechanism. It has been postulated that the UF polymeric membrane manages to concentrate the coconut proteins while allowing the permeation of water and other smaller sized compounds. However, a complex protein mixture (such as SCM) can easily reduce the efficiency of the UF process due to the tendency of the membrane to foul. The permeate flux and selectivity of membrane are hindered due to the membrane fouling issue, especially when the involved subject is a complicated proteinaceous solution (Chan et al., 2002; Wu et al., 2007). The characterization of membrane fouling mechanisms is highly important especially during the UF process. In order to produce higher permeate flux and greater solute rejections, a clear understanding on the membrane fouling mechanisms is essential. Besides, a loss in selectivity of the main products can occur during the occurrence of membrane fouling in UF process (Mohammad et al., 2012). Membrane fouling can either occur by the deposition of particles inside the membrane pores, on the membrane surface or a combination of both (Ho and Zydney, 2000; Kelly and Zydney, 1997; Maruyama et al., 2001). When the particles deposit on the membrane surface and cover up the membrane surface pores, the permeability of the membrane can be reduced as these particles will restrict the passage of the water molecules through the membrane surface pores. The extent of fouling strongly depends on the feed solution properties, membrane materials and operating conditions (Ramesh Babu and Gaikar, 2001). Many studies had attempted to improve the filtration processes of complex solutions by various methods such as varied experimental conditions, adjusted feed solution behaviour, modified operational design and the used of modified membrane (Boyd and Zydney, 1998; De Bruijn et al., 2005; Palacio et al., 2002; Rahimpour, 2011). However, the characterization and prediction of membrane fouling in typical condition remain as a challenge in the membrane technology. In order to minimize membrane fouling during the ultrafiltration of SCM, various membrane fouling mechanisms need to be pre-evaluated. All the fouling mechanisms of the membrane can be described by blocking filtration laws. These blocking filtration laws consist of complete blocking, standard blocking, intermediate blocking and cake filtration mechanisms. This fouling model was firstly introduced by Hermia and named Hermia’s model (Hermia, 1982). There have already been a few studies that analysed the fouling of a membrane using Hermia’s model. Recently, researchers (Nourbakhsh et al., 2013) employed Hermia’s model to describe fouling mechanisms during the clarification of red plum juice. They concluded that cake formation was the predominant fouling mechanism during juice clarification. However, other fouling mechanisms (complete, standard and intermediate pore blocking) were also involved when the filtration time was prolonged . In addition, Hermia’s model has also been used in the UF of polysaccharide macromolecules (Sarkar, 2013). In this case, they found that a proposed flux model in which complete pore blocking predominated in the early stage of the UF process could satisfactorily describe the flux decline. This phenomenon has been frequently studied during past decade for the filtration of protein and polysaccharide solutions (Feng et al., 2009; Palacio et al., 2002). In addition, Hermia’s model has also been used to determine nature of membrane fouling during the application of wastewater (Salahi et al., 2010), glycerine solution (Amin et al., 2010), colloidal suspensions

(Wang and Tarabara, 2008), polyethylene glycol (Vela et al., 2008) and oil in water emulsions (Mohammadi et al., 2003). To minimize fouling while using polymeric membranes, experimental parameters that influence reductions in flux need to be studied. Therefore, the fouling mechanisms involved in this study needed to be identified during ultrafiltration of SCM under different operating conditions including pressure (1.8, 2.0, 2.2, 2.4 bar), the membrane’s MWCO (10 kDa, 20 kDa) and temperature (50°C, 55°C, 60°C). Effects on membrane fouling due to the changes in operating conditions were also studied. The fouling mechanisms were analysed using Hermia’s model and the predictions compared with the experimental data obtained. 1.1 Prevailing fouling mechanisms Hermia (Hermia, 1982) derived a mathematical model (Eq.(1)) to describe permeate flux decline phenomena. The derivation of this theoretical model is typically based on the classic constant-pressure filtration process. The fouling mechanism can be identified using the socalled blocking filtration law or Hermia’s model. (d2t/dV2)=k(dt/dV)n

(1)

The exponent n in Eq.(1) characterizes the type of filtration mechanism. In the following sessions, further description of each fouling mechanism will be given. 1.1.1 Complete pore blocking (n=2) Complete blocking normally occurs when the sizes of filtration solutes are greater than the pore openings in the membrane. The solutes will completely obstruct or seal the openings of the membrane pores without superposition of the solutes. The filtration resistance will increase as the number of unblocked membrane pores decreases (Hwang and Lin, 2002). As a result, the permeate flow rate will decrease exponentially with time. The filtration volume flow can be related to the time using Eq.(2): Jvt = Jvo[exp(- ct)] (2) where εc = KcVo Kc is the area of membrane surface blocked per unit of total volume permeated through the membrane and Vo is the mean initial velocity of the filtrate or initial volume flow per unit of porous membrane surface area. Thus, the estimated evolution over time of the permeate volume flow is given by Eq.(3): ln Jvt = ln Jvo -

ct

(3)

1.1.2 Standard pore blocking (n=1.5) Standard blocking, also called internal pore blocking. This fouling mechanism occurs when small size solutes deposit or adsorb onto the pore walls in the membrane (Bowen et al., 1995). It only takes place when the sizes of solutes are less than the size of the pore entrances in the membrane. The deposition of solutes onto the pore walls or within the

membrane supports will greatly increase the filtration resistance and attenuate the filtration rate as the membrane pore volumes are reduced . The correlation between the volume flow and time is given by Eq.(4): Jvt = Jvo/(1+ st)2

(4)

1/2

where s = KsVo/(Jvo) Ks is the reduction in cross-sectional area of the pores per unit of total permeate flow. The linearized equation that corresponds to the volume flow versus time is given by Eq.(5): 1/(Jvt)1/2 = 1/(Jvo)1/2 +

st

(5)

1.1.3 Intermediate pore blocking (n=1) In the case of intermediate pore blocking, the diameters of the solutes are very similar to the pore size of the membrane. It presumes that the solutes can root steadily on previously deposited solutes. Some particles may directly obstruct the pores and cover some active areas of the membrane (Hwang and Lin, 2002). Based on this presumption, each point on the membrane surface has an equal chance to be covered by the solutes . Therefore, it is the socalled “intermediate pore blocking” phenomenon. The volume flow versus time law is given by Eq.(6): Jvt = Jvo/(1+ it) Where i = KcVo/Jvo Here, the predicted evolution of the permeate flow over time is given by Eq.(7): 1/Jvt = 1/Jvo +

it

(6)

(7)

1.1.4 Cake formation (n=0) The cake formation model can be employed when the accumulated solutes on the membrane surface display a layer form. Normally this phenomenon occurs when the solutes involved are larger in size compared to the membrane pore size. It is postulated that the particles will settle on other pre-deposited solutes that arrived earlier and already cover the membrane surface. In this case, there is no more space on the membrane surface as the filtration time prolong (Palacio et al., 2002). Thus, it is expected that there is a high concentration of solute in this fouling pattern. For this reason it is known as the “cake formation or cake filtration” phenomenon. The relation between time and volume flow is given by Eq.(8): Jvt = Jvo/(1+ cft)1/2 where

cf = C/(Jvo)2

(8) (9)

and C = (2Rr)KckVo By introducing Eq.(10) into Eq.(9), Eq.(11) can thus be obtained as: cf = (2Rr)KckVo/[(Jvo)2]

(10)

(11)

where 1/Kck is determined as the volume of permeate collected per unit area of membrane and Rr is the ratio of the resistance of the cake layer formed over the resistance of the fresh or clean membrane. In this case, the correlation between the total permeate volume and time is given by Eq.(12): 1/(Jvt)2 = 1/(Jvo)2 +

cft

(12)

Fig.1. Schematic diagrams showing the types of fouling mechanisms: (a) complete pore blocking, (b) intermediate pore blocking, (c) standard pore constriction and (d) cake layer formation Fig.1 shows a schematic representation of each fouling mechanism or blocking pattern during the filtration process, whereas Table 1 shows the fitting equation and the value of n for each model. Table 1. Summary of the fouling mechanisms or blocking models during membrane fouling 2. Material and Methods 2.1 Preparation of skimmed coconut milk (SCM) The coconut milk used in the experiments was the Malayan Tall variety. The coconut milk was produced from grated white coconut endosperm using a coconut extraction machine. Based on the perishability properties of coconut milk, a pasteurization process was conducted to reduce the microbial load. The method used was according to previously reported work (R.Hagenmaier, 1980). The solution employed in this study was SCM or so-called defatted coconut milk. Table 2 shows the basic nutrients in SCM. SCM is enriched with various nutrients such as protein, carbohydrate, fat, ash, moisture and others. The fat content of the coconut milk was removed using a cream separator (Elecrem, France) prior to the filtration process. The extracted SCM was stored at a temperature of 4°C. SCM can only be kept up to 5 days after the extraction process as it is a highly perishable liquid. Table 2. Content of skimmed coconut milk 2.2 Ultrafiltration process Polysulfone membrane was employed throughout the present study. This is due to its several superior properties such as high material toughness, good stability at high temperature, high resistant to various solvent, good resistant to wide pH range (pH 2-13) and low protein binding tendency. Two types of PSF membranes with different MWCO values (10 kDa and 20 kDa) are used during the filtration experiments. The 10 kDa PSF membrane was purchased from Koch while the 20 kDa PSF membrane was purchased from Alfa Laval. The active membrane surface area used was 15.2 cm2. Fresh membranes were soaked in pure water overnight prior to each experimental run. This step was aimed at removing preservatives on the membrane surfaces and also acted as a wetting process for the membranes.

A stirred filtration cell (Sterlitech HP4750) was used to perform the ultrafiltration process. The dead-end cell was equipped with a magnetic stirrer. A stirring rate of 400 rpm was set in each run to reduce the formation of high solution concentration in the adjacent areas at the membrane surface. In addition, homogeneity was achieved in the solution by the stirring effect. The desired pressure was supplied by a nitrogen cylinder connected to the filtration system. The experiments were carried out in a water bath in order to maintain a constant temperature throughout the filtration process. The pH of the solution was kept at its origin form (pH 6). Membrane compaction was carried out before any filtration process. This was done by filling the cell with pure water and pressurizing the cell at a constant pressure of 3 bar. All the experimental runs in present work were replicated at least twice to ensure the reproducibility of the results. 2.3 Experimental design After the compaction procedure, a membrane permeability test was conducted at different pressures (1.8 to 2.4 bar) using pure water. The effects of the operating parameters on the membrane filtration performance were evaluated based on the pressure and temperature changes listed in Table 3. In engineering application, thermal processing is important during the handling of food fluids. Therefore, temperature ranges from 50oC to 60oC were chosen in order to evaluate the fouling extent of membrane during UF of skimmed coconut milk. If the operating temperature was fixed at lower value, severe fouling may take place. It is contributed mostly by the combination of chemical reaction fouling from proteins and precipitation fouling of fat (Narataruksa et al., 2010). In addition, at these temperatures (5060oC), the deterioration of skimmed coconut milk by micro-organisms can be prevented. The operating conditions of the flux decline experiments were: temperature (50°C, 55°C and 60°C), pressure (1.8, 2.0, 2.2 and 2.4 bar) and constant stirring rate (400 rpm). The pH of the SCM milk was maintained where it started (pH 6). The feed solution used during the filtration process was SCM. The solutions were heated to a desired temperature before the ultrafiltration process using PSF membranes. The duration of each ultrafiltration process was set at 60 min. The volume of permeate per minute was continuously monitored. The curve of flux versus time was then plotted using the data obtained. Table 3. Experimental parameters for ultrafiltration of skimmed coconut milk 3. Results and Discussion 3.1 Effect of membrane molecular weight cut-off (MWCO) SCM milk is a type of proteinaceous solution that contains 70% of total coconut proteins. Thus, the main foulant involved in this UF process is protein molecules. In this study, blocking filtration laws or Hermia’s models were employed to interpret the fouling mechanisms that occurred during the ultrafiltration process using SCM. The fit of the fluxtime experimental data to each model (Eq.(3),(5),(7),(12)) allows one to distinguish whether the drop in the permeate flux is governed by cake layer formation or pore blocking. To investigate the effect of the membrane molecular weight cut-off upon the membrane fouling, two PSF membranes with different MWCO values (10 kDa, 20 kDa) were employed. Fig.2 shows the declines in permeate flux versus time using the 10 kDa and 20 kDa PSF

membranes. The flux decline was more significant using the low MWCO (10kDa) membrane compared to the higher MWCO (20 kDa) membrane. Reductions in the permeate flux can therefore be explained by total blocking of the membrane pores, resulting in less fluid being passed through the membrane (Grenier et al., 2008). Fig.2. Normalized flux declines of skimmed coconut milk using 10 kDa and 20 kDa PSF membranes Fig.3(a-d) illustrates the fitting of the obtained experimental data using the 10 kDa and 20 kDa PSF membranes to the different predicted fouling mechanisms (complete pore blocking, standard pore blocking, intermediate pore blocking and cake formation). All the experimental data were well fitted to each fouling mechanism. In Fig.3, it is apparent that the experimental data obtained with the 20 kDa membrane shows the best fit to each fouling model. It can therefore be inferred that the PSF membrane with a higher MWCO is more susceptible to fouling. This observation is supported by other work (Marshall et al., 1993). The researchers claimed that membrane fouling is more severe when the membrane pore size increases. Thus, the accumulation of proteins and other minor coconut components were more significant when using a membrane with a higher MWCO (20 kDa). This condition allowed inner pore blockage to occur easily and thus caused more severe fouling compared to the 10 kDa PSF membrane. The total effective internal pore surface area of the 20 kDa PSF membrane is larger than the internal surface area of the pores in the 10 kDa PSF membrane (Marshall et al, 1993). This condition encouraged a larger number of smaller sized proteins to adsorb onto the pore walls of the 20 kDa PSF membrane. Thus, fouling is more severe for the 20 kDa PSF membrane. From Fig.2 it can be seen that the normalized permeate flux produced by the 20kDa PSF membrane was less than that for the 10 kDa PSF membrane. According to previous studies, researchers have claimed that coconut proteins exist with various molecular sizes which ranging between 10kDa to >100kDa (DeMason and Sekhar, 1990; Kwon et al., 1996; Samson S.J et al., 1971) whereas molecular weights of dominant proteins ranging from 14kDa to 52kDa (Seow and Gwee, 1997). The sizes of the coconut proteins therefore may be smaller than, the same or even larger than that of the UF membrane pores (10kDa and 20kDa). Moreover, there are some micro-nutrients in the SCM such as vitamins, nitrogenous compounds, organic acids and so forth. For instance, vitamin B12 (molecular size, 2.4nm), phytohormones (molecular size, 0.05-0.4μm) and other low molecular weight biologically active compounds (e.g. pyridoxine 1 nm) (Yin et al., 2013; Yong et al., 2009). Thus, it is reasonable that both pore blocking and cake formation mechanisms contributed to the membrane fouling observed. Fig.3. Effect of the membrane molecular weight cut-off upon the permeate flux according to the predicted fouling mechanisms: (a) complete pore blocking, (b) standard pore blocking, (c) intermediate pore blocking and (d) cake formation Value of R2 and fitted parameters for each Hermia’s model by studying the effect of membrane molecular weight cut-off (10 kDa, 20 kDa) upon membrane fouling were shown in Table 4 (No. 1-2) and Table 5 (No.1-2). A higher value of R2 indicates that a very good fit of the experimental data to the Hermia’s models was obtained. Meanwhile, the values of the fitted parameters should be higher for the membrane that exhibits a more severe flux decline (Vela et al., 2008). The values of R2 obtained using the 20 kDa PSF membrane with each of the predicted fouling mechanism are higher (>0.98) compared to those obtained using the 10 kDa PSF membrane. In addition both membranes, with different MWCO values, showed the smallest values of the fitted parameters for the cake formation mechanism when compared to

the other fouling mechanisms (Table 5 (No.1-2)). Thus, it can be postulated that the blocking mechanisms (complete, standard and intermediate blocking) dominate the UF processes for both membranes using SCM. The fitted parameters ( s, i, cf) obtained with the 10 kDa PSF membrane were slightly higher compared to those for the 20kDa PSF membrane except for the complete blocking parameter. This proves that internal blocking within the 10 kDa PSF membrane was more severe than with the 20 kDa PSF membrane. The occurrence of internal blocking is probably due to the adsorption of smaller sized proteins and other components (such as vitamins, anti-oxidants and minerals) on the membrane pore walls. Fouling is a complex phenomenon that is likely to occur both on the membrane surface and within the membrane pores. The predominant fouling mechanism is a function of the experimental conditions (mostly influenced by the operating conditions and membrane properties). Therefore, in the following sections, the effects of the operating conditions upon membrane fouling were evaluated. 3.2 Effect of feed temperature The feed solution (SCM) was filtered using a PSF membrane with a 10 kDa MWCO at different temperatures (50, 55, 60°C) but constant pressure (1.8 bar) to evaluate the effect of feed temperature upon membrane fouling. In Fig.4, the normalized flux declines for the SCM milk at varying temperature values are displayed. From Fig.4 it can be seen that the normalized flux decline curve obtained at 60°C is significantly higher than the other two curves (at 50 and 55°C). This result showed that a larger amount of the proteins within the SCM managed to pass through the membrane at a higher feed temperature. This can be explained by the fact that an increase in temperature increases the diffusion coefficient of some molecules (for example polysaccharides and proteins) to penetrate into the pores and/or deposit along the pore walls (Wang and Tang, 2011). Besides that, the physicochemical properties of SCM can easily be affected when the temperature is increased. The viscosity of the feed solution decreases as the temperature increases. Thus, the flow of solutes within the SCM becomes smoother so it is easier for them to pass through the semi-permeable membrane. This observation has previously been noted and explained by several researchers (Makardij et al., 2002).

Fig.4. Normalized flux declines of skimmed coconut milk at various operating temperatures Fig.5 illustrates the fit of the permeate fluxes to the predicted fouling mechanisms in order to study the effect of various feed temperatures upon membrane fouling. It can be seen that deviations of the experimental data from the predicted fouling mechanisms occurred at feed temperatures of 50°C and 55°C. However, the permeate fluxes obtained at a temperature of 60oC best fit all the predicted fouling mechanisms. These results indicate that at a temperature of 60°C all the fouling mechanisms were taking place on the membrane surface and inside the membrane pores. At a feed temperature of 50°C, standard and intermediate blocking were the dominant fouling patterns because the fits of the experimental data deviated less from those Hermia’s models (Eq.(5) and (7)). At lower feed temperatures (50, 55°C) poor fitting for the cake formation mechanism was present (Fig.5(d)). It can therefore be seen that fouling during UF of SCM milk was dominated throughout the filtration process by the standard blocking and intermediate blocking mechanisms instead of cake layer formation. This can be seen from the small deviation achieved in Fig.5(a-c). The adsorption and deposition of proteins and other smaller sized solutes such as biomolecules in pore walls can contribute to this irreversible internal fouling (James et al., 2003). They (James et al.,

2003) claimed that the deposited protein particles would interact with the pore walls of the membranes (protein-polymer reaction). In addition, protein particles would form agglomerates through protein-protein reactions. These reactions may lead to the narrowing and finally blocking of the pores. The pore blocking mechanisms (complete blocking, standard blocking and intermediate blocking) corresponded to the fitted Hermia’s models in Fig.5(a-c) These pore blocking phenomena contributed to low permeates fluxes at feed temperatures of 50°C and 55°C (as shown in Fig.4). Fig.5. Effect of various feed temperatures on permeates fluxes according to the predicted fouling mechanisms: (a) complete pore blocking, (b) standard pore blocking, (c) intermediate pore blocking and (d) cake formation. R2 for the fit of the experimental data to the predicted fouling mechanisms from Hermia’s models by studying the effect of various feed temperatures upon membrane fouling were shown in the Table 4 (No.3-5). Generally, all the experimental data of permeate flux fit well in Hermia’s models, with all the R2 values above 0.90 except for the experimental run with a feed temperature of 55°C using the predicted cake formation mechanism (Table 4 (No.3-5)). The best fittings of the experimental data to each fouling mechanism (Eq.(3), (5), (7) and (12)) were obtained for experiment sets conducted at a feed temperature of 60°C. Values of R2 near or above 0.99 were achieved with permeate flux data obtained when the analyses were conducted at 60°C. The functions, which were linearly dependent on time (lnJvt, 1/(Jvt)-1/2, 1/Jvt and 1/Jvt2), verify that the fouling mechanisms induced the initial declines in the permeate fluxes. The values of fouling constants or Hermia’s model parameters, , for the fitted fouling mechanisms (Hermia’s models) that were employed in this work were shown in Table 5 (No.3-5). According to the physical meanings and the constant definitions of Hermia’s models (Eq.(3), (5), (7)) and (12)), the values of these parameters are expected to be higher for experimental conditions that correspond to more severe membrane fouling. As can be seen from Table 5, the fouling is less severe when the feed solution temperature is higher. Thus, the normalized flux declines obtained with feed temperatures of 50°C and 55°C were more significant compared to the case with a higher feed temperature (60°C) (Fig.4). By comparing all the fouling constants obtained, it can be seen that the fouling constant for the complete blocking model exhibited the highest values (Table 5 (No.3-5)). It can be noted that the predominant fouling mechanisms due to the effect of various feed temperatures were pore blocking phenomena. The pore blocking phenomena initiated with complete blocking, followed by standard blocking and intermediate blocking. The low value of the cake formation parameter indicates that the extent of fouling contributed by the cake layer effect is insignificant compared to the pore blocking models when studying the effect of various feed temperatures. The present findings seem to be consistent with other research works, which also concluded that pore blocking mechanisms dominated membrane fouling during the filtration of proteinaceous solutions (James et al., 2003; Krishna Kumar et al., 2004). It is notable that the greatest drop in the cake layer constant occurred at the 60°C temperature. This measured parameter revealed that the lesser intensity of cake layer fouling was possibly due to the reduced solution viscosity when the feed temperature increased (Vetier et al, 1988). This condition encouraged more proteins and other soluble molecules to pass through the membrane and thus increased the flow of the permeate stream. This explanation is in good agreement with the obtained result (Fig.4). On the other hand, the reduced intensity of cake layer fouling may signify a greater internal fouling.

3.3 Effect of operating pressure To investigate the effect of various operating pressures upon membrane fouling, the UF processes using SCM were carried out using 10 kDa molecular weight cut-off PSF membranes at constant temperature (60°C) and varying pressures (1.8, 2.0, 2.2 and 2.4 bar). In Fig.6, the normalized flux declines for SCM at varying operating pressures are illustrated. It can be seen that at a pressure of 2.4 bar, the normalized flux decline obtained is the most significant compared to the others. At the other operating pressures of 1.8, 2.0 and 2.2 bar, the declines in normalized fluxes were less significant. Similar patterns in the normalized flux declines were exhibited when employing 10 kDa PSF membranes at various operating pressures. The most drastic declines in normalized fluxes occurred in the first 10 minutes of the UF processes and were followed by more gradual declines (Grenier et al., 2008). Membrane fouling is believed to induce the initial decline in normalized fluxes. The declining normalized fluxes could be due to concentration polarization, accumulation of protein molecules or the deposition of other coconut components on the membrane surface. In addition, internal pore blocking also contributes to the initial decline in the normalized flux of SCM milk at various operating pressures (Tong et al., 1988). Fig.6. Normalized flux declines of skimmed coconut milk using UF membranes at various operating pressures. Fig.7(a-d) shows the fitting of the experimental results to the Hermia’s models according to Eq.((3), (5), (7) and (12)). Small deviations were observed between the experimental results and the predicted flux declines for various fouling mechanisms. The best fit of the permeate flux data to a fouling mechanism was obtained in the experiment carried out at a pressure of 2.0 bar. Complete blocking can occur when the sizes of solute molecules in the feed solution are greater than those of the membrane pore openings. Under this condition, the solute molecules cannot either pass through the membrane. This fouling condition is reasonable as SCM milk consists of a broad size distribution of albumin, globulins and other types of protein molecules (Naik et al., 2012). This contributes to the small deviations observed between the predicted data and the experimental results. Internal fouling of the membrane occurs when the standard blocking mechanism is involved. Standard blocking phenomena occur when the sizes of the solute molecules are smaller than those of the membrane pores (Cheng et al., 2011). This encourages the adsorption of smaller sized solutes onto the membrane pore walls. Fig.7(b) displays the experimental results that fit well to the predicted standard blocking mechanism according to Eq.(5). This can be explained by the fact that some solute molecules in the SCM milk were smaller in size compared to the membrane pores. Some of these possible solutes are vitamins, minerals, plant hormones, anti-oxidants and nitrogenous compounds (Yin et al., 2013). It is thus reasonable for standard blocking to participate in membrane fouling during the filtration of SCM. Fig.7(c) shows the good fit of the experimental permeate flux to the intermediate blocking model for various operating pressures. When solute molecules possess sizes similar to the membrane pore size, intermediate blocking can easily occur. In this phenomenon, some of the entrances to the membrane pores will be blocked by deposited solutes. This fouling mechanism is more suitable for use in describing real-life situations of UF processes because the solute molecules can easily deposit on other previously deposited solutes (Hermia, 1982).

Therefore, the intermediate blocking mechanism provides better agreement with the experimental data than complete blocking in this study. Cake layer formation occurs when the solute molecules are unable to enter the membrane pores. Consequently, the gel layers will cover the membrane surface and build up the thickness of the cake formed. Molecular deformation may occur under this fouling condition. As a result, a lower permeate flux and higher flux decline rate will be obtained (Bowen et al., 1995). Cake layer formation usually happens after a prolonged filtration period (Saha et al., 2007). The overlapping cake layers prohibit proteins from entering into the membrane pores and this contributes to the low permeate flux produced. The protein molecules with sizes larger (e.g. globulins with sizes 34-150kDa (Garcia et al., 2005; N.Angelia et al., 2010)) than membrane pore size also were retained by PSF membranes. The thickness of cake layer increased as the filtration time prolonged. This makes a cake layer more likely to form. This circumstance is supported by Fig.7 (d), which shows the experimental permeate flux fit well to the predicted cake layer mechanism. Moreover, the fouling constant of the cake formation mechanism in Table 5 also indicates the participation of cake layer fouling in this study of the effect of various operating pressures. Fig.7. Study of the effect of various operating pressures upon the permeate flux during filtration according to the predicted fouling mechanisms: (a) complete pore blocking, (b) standard pore blocking, (c) intermediate pore blocking and (d) cake formation Tables 4 (No.6-9) and Table 5 (No.6-9) present the values of R2 and fitted parameters, respectively, for the fouling mechanisms in the study of the effect of various operating pressures upon membrane fouling. The results show that the best fit of the experimental results to a predicted fouling mechanism was achieved using intermediate blocking, with a value of R2 greater than 0.97 (Table 4 (No.6-9)). Besides intermediate blocking, other fouling mechanisms (complete blocking, standard blocking and cake formation) also fit well with the experimental permeate flux data. These findings are supported by the data in Fig.7(a-d). Table 5 shows the values of fouling constants ( ) for each fitted fouling mechanism obtained during UF of SCM milk using a PSF membrane. As discussed previously, the values of the fouling constants represent the intensity of each membrane fouling mechanism. The highest value of the fitted parameter was seen with the complete blocking mechanism among all the experimental runs in the study using various operating pressures. This was followed by, in descending order, standard blocking, intermediate blocking and cake formation. The pore blocking mechanisms dominated the UF of SCM using 10 kDa PSF membranes at various operating pressures. This can be explained by the fact that finer solutes (vitamins, minerals, anti-oxidants and etc.) can enter the membrane pores much faster than other larger sized molecules (proteins). As a result, the fouling caused by pore blocking mechanisms (complete blocking, standard blocking and intermediate blocking) was more severe when compared to cake layer formation. This finding was consistent with the results shown in Table 5 (No.6-9). Similar findings have also been reported by others (Li and Chen, 2004). Table 4. The values of R2 obtained from the experimental data in the study of the effect of membrane MWCO and operating parameters upon membrane fouling Table 5. The parameters of fitted Hermia’s model from the study of the effect of different membrane MWCO and various operating parameters upon membrane fouling 4. Conclusion

In this study, fouling behaviours of polysulfone (PSF) membranes during ultrafiltration (UF) of SCM were studied. The effects of various membrane MWCO values, feed temperatures and operating pressures have been evaluated. Hermia’s models (complete blocking, standard blocking, intermediate blocking and cake formation) were used to analyse and predict the fouling mechanisms occurring during UF of SCM. The permeate flux decline profiles of all the experimental runs were compared to the Hermia’s models. In brief, good fits of the experimental results to the fouling models were obtained. Based on the results from all the experimental conditions tested, it can be concluded that fouling was predominantly contributed to by the pore blocking models, followed by cake layer formation. On the other hand, a combination of two or more fouling mechanisms may take place during filtration of a proteinaceous solution (skimmed coconut milk). This could be the reason for the deviation obtained between the experimental and predicted fouling models. The best fits were obtained when the experimental results using a 20 kDa membrane and a feed temperature of 60°C were fitted to the predicted fouling mechanisms (Fig.3 and 5). During two different experimental runs using a 20 kDa PSF membrane and a feed solution temperature of 60 °C, the values of R2 obtained in both cases were around 0.99 (Tables 4 (No.2 and 5)). Under all the experimental conditions tested, it was observed that the fouling constants of the complete blocking mechanism were always greater than those obtained from other fouling mechanisms (Table 5). This indicates that the complete blocking mechanism dominated fouling studies that were conducted under various conditions (by manipulation of membrane MWCO, feed temperature and operating pressure). Complete blocking was followed by standard blocking, intermediate blocking and cake layer formation, in descending order. Cake layer formation showed the least severe fouling (with the smallest value of the model parameter under each experimental condition) compared to pore blocking (Tables 5). In brief, the experimental results suggest different fouling mechanisms are operative during UF of SCM. In addition, the extent of membrane fouling was successfully determined through employing predictive fouling models. Acknowledgements The authors of this work wish to gratefully acknowledge the financial support of the Ministry of Science and Technology (ScienceFund) through project no. 02-01-02-SF1021. References Amin, I.N.H.M., Mohammad, A.W., Markom, M., Peng, L.C., Hilal, N., (2010). Analysis of deposition mechanism during ultrafiltration of glycerin-rich solutions. Desalination 261(3), 313-320. Bowen, W.R., Calvo, J.I., Hernández, A., (1995). Steps of membrane blocking in flux decline during protein microfiltration. Journal of Membrane Science 101(1–2), 153-165. Boyd, R.F., Zydney, A.L., (1998). Analysis of protein fouling during ultrafiltration using a two-layer membrane model. Biotechnology and Bioengineering 59(4), 451-460. Chan, R., Chen, V., Bucknall, M.P., (2002). Ultrafiltration of protein mixtures: measurement of apparent critical flux, rejection performance, and identification of protein deposition. Desalination 146(1-3), 83-90.

Chen, B.-K., Diosady, L.L., (2003). Enzymatic Aqueous Processing of Coconuts. International Journal of Applied Science and Engineering 1(1), 55-61. Cheng, Y.-L., Lee, D.-J., Lai, J.-Y., (2011). Filtration blocking laws: Revisited. Journal of the Taiwan Institute of Chemical Engineers 42(3), 506-508. De Bruijn, J.P.F., Salazar, F.N., Borquez, R., (2005). Membrane Blocking In Ultrafiltration: A New Approach to Fouling. Food and Bioproducts Processing 83(3), 211-219. DeMason, D.A., Sekhar, K.C., (1990). Electrophoretic characterization and immunological localization of coconut (Cocos nucifera L.) endosperm storage proteins. Botanical gazette, 302-313. Feng, L., Li, X., Du, G., Chen, J., (2009). Characterization and fouling properties of exopolysaccharide produced by Klebsiella oxytoca. Bioresource Technology 100(13), 33873394. Garcia, R.N., Arocena, R.V., Laurena, A.C., Tecson-Mendoza, E.M., (2005). 11S and 7S Globulins of Coconut (Cocos nucifera L.): Purification and Characterization†. Journal of Agricultural and Food Chemistry 53(5), 1734-1739. Grenier, A., Meireles, M., Aimar, P., Carvin, P., (2008). Analysing flux decline in dead-end filtration. Chemical Engineering Research and Design 86(11), 1281-1293. Hermia, J., (1982). Constant pressure blocking filtration laws-application to power-law nonnewtonian fluids. Icheme 60, 183-187. Ho, C.-C., Zydney, A.L., (2000). A combined pore blockage and cake filtration model for protein fouling during microfiltration. Journal of Colloid and Interface Science 232(2), 389399. Hwang, K.-J., Lin, T.-T., (2002). Effect of morphology of polymeric membrane on the performance of cross-flow microfiltration. Journal of Membrane Science 199(1–2), 41-52. James, B.J., Jing, Y., Dong Chen, X., (2003). Membrane fouling during filtration of milk--a microstructural study. Journal of Food Engineering 60(4), 431-437. Kelly, S.T., Zydney, A.L., (1997). Protein fouling during microfiltration: Comparative behavior of different model proteins. Biotechnology and Bioengineering 55(1), 91-100. Krishna Kumar, N.S., Yea, M.K., Cheryan, M., (2004). Ultrafiltration of soy protein concentrate: performance and modelling of spiral and tubular polymeric modules. Journal of Membrane Science 244(1–2), 235-242. Kwon, K.S., Bae, D., Park, K.H., Rhee, K.C., (1996). Aqueous Extraction and Membrane Techniques Improve Coconut Protein Concentrate Functionality. Journal of Food Science 61(4), 753-756.

Li, C.-W., Chen, Y.-S., (2004). Fouling of UF membrane by humic substance: Effects of molecular weight and powder-activated carbon (PAC) pre-treatment. Desalination 170(1), 59-67. Makardij, A., Chen, X.D., Farid, M.M., (1999). Microfiltration and Ultrafiltration of Milk: Some Aspects of Fouling and Cleaning. Food and Bioproducts Processing 77(2), 107-113. Makardij, A.A., Farid, M.M., Chen, X.D., (2002). A simple and effective model for crossflow microfiltration and ultrafiltration. The Canadian Journal of Chemical Engineering 80(1), 28-36. Marina, A.M., Che Man, Y.B., Amin, I., (2009). Virgin coconut oil: emerging functional food oil. Trends in Food Science & Technology 20(10), 481-487. Marshall, A.D., Munro, P.A., Trägårdh, G., (1993). The effect of protein fouling in microfiltration and ultrafiltration on permeate flux, protein retention and selectivity: A literature review. Desalination 91(1), 65-108. Maruyama, T., Katoh, S., Nakajima, M., Nabetani, H., (2001). Mechanism of bovine serum albumin aggregation during ultrafiltration. Biotechnology and Bioengineering 75(2), 233-238. Mohammad, A., Ng, C., Lim, Y., Ng, G., (2012). Ultrafiltration in Food Processing Industry: Review on Application, Membrane Fouling, and Fouling Control. Food and Bioprocess Technology 5(4), 1143-1156. Mohammadi, T., Kazemimoghadam, M., Saadabadi, M., (2003). Modeling of membrane fouling and flux decline in reverse osmosis during separation of oil in water emulsions. Desalination 157(1–3), 369-375. N.Angelia, M.R., N.Garcia, R., P.Caldo, K.M., Prak, K., Utsumi, S., Tecson-Mendoza, E.M., (2010). Physicochemical and Functional Characterization of Cocosin, the Coconut 11S Globulin. Food Science and Technology Research 16(3), 225-232. Naik, A., Raghavendra, S.N., Raghavarao, K.S.M.S., (2012). Production of Coconut Protein Powder from Coconut Wet Processing Waste and its Characterization. Applied Biochemistry and Biotechnology 167(5), 1290-1302. Narataruksa, P., Pichitvittayakarn, W., Heggs, P.J., Tia, S., (2010). Fouling behavior of coconut milk at pasteurization temperatures. Applied Thermal Engineering 30(11-12), 13871395. Nigam, M.O., Bansal, B., Chen, X.D., (2008). Fouling and cleaning of whey protein concentrate fouled ultrafiltration membranes. Desalination 218(1–3), 313-322. Nourbakhsh, H., Emam-Djomeh, Z., Mirsaeedghazi, H., Omid, M., Moieni, S., (2013). Study of different fouling mechanisms during membrane clarification of red plum juice. International Journal of Food Science & Technology, n/a-n/a.

Palacio, L., Ho, C.-C., Zydney, A.L., (2002). Application of a pore-blockage—Cake-filtration model to protein fouling during microfiltration. Biotechnology and Bioengineering 79(3), 260-270. R.Hagenmaier, (1980). Coconut Aqueous Processing, 2nd ed. Publications, University of San Carlos, Cebu City,Philippines. Rahimpour, A., (2011). Preparation and modification of nano-porous polyimide (PI) membranes by UV photo-grafting process: Ultrafiltration and nanofiltration performance. Korean Journal of Chemical Engineering 28(1), 261-266. Ramesh Babu, P., Gaikar, V.G., (2001). Membrane characteristics as determinant in fouling of UF membranes. Separation and Purification Technology 24(1-2), 23-34. Saha, N.K., Balakrishnan, M., Ulbricht, M., (2007). Sugarcane juice ultrafiltration: FTIR and SEM analysis of polysaccharide fouling. Journal of Membrane Science 306(1-2), 287-297. Salahi, A., Abbasi, M., Mohammadi, T., (2010). Permeate flux decline during UF of oily wastewater: Experimental and modeling. Desalination 251(1–3), 153-160. Samson S.J, A.S., Khaund, R.N., Cater, C.M., Mattil, K.F., (1971). Extractability of coconut proteins. Journal of Food Science 36(5), 725-728. Santoso, U., Kubo, K., Ota, T., Tadokoro, T., Maekawa, A., (1996). Nutrient composition of kopyor coconuts (Cocos nucifera L.). Food Chemistry 57(2), 299-304. Sarkar, B., (2013). A combined complete pore blocking and cake filtration model during ultrafiltration of polysaccharide in a batch cell. Journal of Food Engineering 116(2), 333-343. Seow, C.C., Gwee, C.N., (1997). Coconut milk: chemistry and technology. International Journal of Food Science & Technology 32(3), 189-201. Tong, P.S., Barbano, D.M., Rudan, M.A., (1988). Characterization of Proteinaceous Membrane Foulants and Flux Decline During the Early Stages of Whole Milk Ultrafiltration. Journal of Dairy Science 71(3), 604-612. Vela, M.C.V., Blanco, S.Á., García, J.L., Rodríguez, E.B., (2008). Analysis of membrane pore blocking models applied to the ultrafiltration of PEG. Separation and Purification Technology 62(3), 489-498. Wang, F., Tarabara, V.V., (2008). Pore blocking mechanisms during early stages of membrane fouling by colloids. Journal of Colloid and Interface Science 328(2), 464-469. Wang, Y.-N., Tang, C.Y., (2011). Protein fouling of nanofiltration, reverse osmosis, and ultrafiltration membranes—The role of hydrodynamic conditions, solution chemistry, and membrane properties. Journal of Membrane Science 376(1–2), 275-282. Wu, T., Mohammad, A., Lim, S., Lim, P., Hay, J., (2013). Recent Advances in the Reuse of Wastewaters for Promoting Sustainable Development, in: Sharma, S.K., Sanghi, R. (Eds.), Wastewater Reuse and Management. Springer Netherlands, pp. 47-103.

Wu, T.Y., Mohammad, A.W., Jahim, J.M., Anuar, N., (2006). Investigations on protease production by a wild-type Aspergillus terreus strain using diluted retentate of pre-filtered palm oil mill effluent (POME) as substrate. Enzyme and Microbial Technology 39(6), 12231229. Wu, T.Y., Mohammad, A.W., Md. Jahim, J., Anuar, N., (2007). Palm oil mill effluent (POME) treatment and bioresources recovery using ultrafiltration membrane: Effect of pressure on membrane fouling. Biochemical Engineering Journal 35(3), 309-317. Wu, T.Y., Mohammad, A.W., Md. Jahim, J., Anuar, N., (2009). Optimized reuse and bioconversion from retentate of pre-filtered palm oil mill effluent (POME) into microbial protease by Aspergillus terreus using response surface methodology. Journal of Chemical Technology & Biotechnology 84(9), 1390-1396. Yin, N.C., Mohammad, A.W., Yong, N.L., (2013). Membrane Performance and Potential Separation of Cytokinins During Ultrafiltration of Skimmed Coconut Milk. Advanced Science Letters 19(12), 3620-3624. Yong, J.W.H., Ge, L., Ng, Y.F., Tan, S.N., (2009). The chemical composition and biological properties of coconut (Cocos nucifera L.) water. Journal of Molecules 14, 5144-5164.

List of Figure Figure 1. Schematic diagrams showing the types of fouling mechanisms: (a) complete pore blocking, (b) intermediate pore blocking, (c) standard pore constriction and (d) cake layer formation Figure 2. Normalized flux declines of skimmed coconut milk using 10 kDa and 20 kDa PSF membranes Figure 3. Effect of the membrane molecular weight cut-off upon the permeate flux according to the predicted fouling mechanisms: (a) complete pore blocking, (b) standard pore blocking, (c) intermediate pore blocking and (d) cake formation Figure 4. Normalized flux declines of skimmed coconut milk at various operating temperatures Figure 5. Effect of various feed temperatures on permeates fluxes according to the predicted fouling mechanisms: (a) complete pore blocking, (b) standard pore blocking, (c) intermediate pore blocking and (d) cake formation. Figure 6. Normalized flux declines of skimmed coconut milk using UF membranes at various operating pressures. Figure 7. Study of the effect of various operating pressures upon the permeate flux during filtration according to the predicted fouling mechanisms: (a) complete pore blocking, (b) standard pore blocking, (c) intermediate pore blocking and (d) cake formation

List of Table Table 1. Summary of the fouling mechanisms or blocking models during membrane fouling Table 2. Content of skimmed coconut milk Table 3. Experimental parameters for ultrafiltration of skimmed coconut milk Table 4. The values of R2 obtained from the experimental data in the study of the effect of membrane MWCO and operating parameters upon membrane fouling. Table 5. The parameters of fitted Hermia’s model from the study of the effect of different membrane MWCO and various operating parameters upon membrane fouling.

Table 1. Summary of the fouling mechanisms or blocking models during membrane fouling Type of Fouling Characteristic Exponent Equation blockage Concept equation n Complete pore Pore sealing ln Jvt = ln Jvo - ct 2 (3) blocking Standard blocking

Intermediate blocking

Cake formation

1

1

pore walls restricted

1/(Jvt)1/2 = 1/(Jvo)1/2 + st

Pore sealing and membrane surface deposition

1/Jvt = 1/Jvo +

Formation of cake layers on surface

it

1/(Jvt)2 = 1/(Jvo)2 + cft

3/2

(5)

1

(7)

0

(12)

Table 2. Content of skimmed coconut milk Content Result Protein % (w/w) 3.7 Fat % (w/w) 0.1 Carbohydrate % (w/w) 2.8 Moisture % (w/w) 92.1 Ash % (w/w) 1.3 2

2

Table 3. Experimental parameters for ultrafiltration of skimmed coconut milk Membrane Experimental Pressure Molecular Weight Run (bar) Cut Off (kDa)

Temperature (°C)

1 2

10 20

1.8 1.8

60 60

3 4 5

10 10 10

1.8 1.8 1.8

50 55 60

6 7 8 9

10 10 10 10

1.8 2.0 2.2 2.4

60 60 60 60

3

3

4

No.

MWCO (kDa)

Tempe r-ature (oC)

Pressure (bar)

1 2

10 20

60 60

3 4 5

10 10 10

6 7 8 9

10 10 10 10

4

Standard pore blocking (n=3/2) 0.9865 0.9938

Intermediate pore blocking (n=1) 0.9940 0.9932

Cake formation (n=0)

1.8 1.8

Complete pore blocking (n=2) 0.9723 0.9932

50 55 60

1.8 1.8 1.8

0.9820 0.9664 0.9950

0.9872 0.9438 0.9957

0.9758 0.9054 0.9939

0.9130 0.8072 0.9829

60 60 60 60

1.8 2.0 2.2 2.4

0.9406 0.9730 0.9207 0.9405

0.9734 0.9912 0.9536 0.9680

0.9885 0.9934 0.9737 0.9843

0.9871 0.9605 0.9893 0.9842

0.9929 0.9880

5

(s-1)

(s-1/2m-1/2)

(m-1)

(sm-2)

No.

MWCO (kDa)

Tempe r-ature (oC)

Pressure (bar)

1 2

10 20

60 60

1.8 1.8

0.0017 0.0083

0.0025 0.0010

0.0015 0.0005

0.0003 6.7901x10-5

3 4 5

10 10 10

50 55 60

1.8 1.8 1.8

0.0274 0.0215 0.0083

0.0044 0.0036 0.0010

0.0029 0.0024 0.0005

0.0007 0.0006 6.6415x10-5

6 7 8 9

10 10 10 10

60 60 60 60

1.8 2.0 2.2 2.4

0.0183 0.0223 0.0179 0.0203

0.0026 0.0031 0.0023 0.0027

0.0016 0.0018 0.0012 0.0014

0.0003 0.0003 0.0002 0.0002

5

c

s

i

cf

Highlights • To investigate the membrane fouling during UF of skimmed coconut milk. • Fouling mechanism was evaluated using Hermia’s model ((d2t/dV2) =k (dt/dV) n). • 20kDa PSF membrane at 60oC and 1.8 bar encouraged the most severe membrane fouling. • All fouling were dominated by standard, intermediate blocking and cake formation.