Majid Peyravi*,Mohsen Jahanshahi,Soodabeh Khalili

Nanoenvironment Research Group,Nanotechnology Institute,Faculty of Chemical Engineering,Babol Noshirvani University of Technology,Babol,Iran

1.Introduction

In membrane separation processes,membranes are always subjected to fouling to a greater or lesser extent depending on fluid composition,membrane properties or operating conditions such as trans membrane pressure(TMP),temperature and flow velocity.This fouling issue is assumed to be a major drawback of membrane technology,especially in the case of using membranes for wastewater treatment applications.Therefore,it is of great importance to identify and understand the fouling mechanisms in order to predict and control membrane behavior over time.

On the one hand,empirical fouling models fail to explain the underlying fouling mechanisms in membrane filtration in spite of their precision.On the other hand,completely theoretical models can rather provide an explanation about underlying phenomena,though they are inaccurate in quantitative prediction of flux decline.Thus,in order to achieve both the accurate prediction and better understanding of the fouling phenomena,semi-empirical models whose parameters are physically meaningful may be exploited[1].Hermia[2]once proposed a mathematical expression from which different mechanisms could be derived by simply assigning its generalindex(n)in the following equations:

wheretis filtration time,Vis accumulated permeate volume andKis a phenomenological coefficient,Jis the permeate flux andAmis the membrane effective area.Depending on the mechanism of the fouling,ncan adopt different values of 2,3/2,1 and 0 for complete blocking model,standard blocking model,intermediate blocking model and cake layer formation model,respectively.The equations corresponding to these mechanisms are summarized in Table 1.

It is noteworthy that although Hermia model was originally developed for the dead-end filtration,many researchers have successfullyexploited this model without any modifications in applying to cross-flow micro filtration[3],cross- flow ultra filtration[1,4,5]and even reverse osmosis systems[6].Others used the modified form of Hermia's model in cross- flow micro filtration[7,8],ultra filtration[9]and nano filtration[10,11].

Table 1Values of n in Eqs.(1)and(2)and their corresponding fouling mechanisms and parameters

To date,many researchers have employed Hermia fouling models in filtration of a variety of feeds.For instance,Aminet al.[12]studied the flux decline in the UF of glycerin-rich solutions.They used high purity fatty acids as foulant model in the feed solution.Velaet al.[4,9]examined Hermia models in UF of polyethylene glycol with known molecular weights which were dissolved in deionized water.As realized by these cases,fouling mechanisms are usually studied with analytical-grade artificial solutions of known composition in the literature.According to Velaet al.[4],complete blocking and cake layer formation mechanisms occur when the sizes of solute molecules are greater than the sizes of membrane pores,intermediate blocking happens when the solute molecule size is similar to the membrane pore size and standard blocking is caused by molecules smaller than the membrane pore size.However,in cases where there are a variety of particle sizes,some fouling mechanisms may take place simultaneously.In other words,fouling mechanisms may differ from a membrane to another membrane due to the change in membrane pore size distributions for an identical feed.In this research,nano-WO3loaded Polysulfone(PSf)membranes with different concentration of nanoparticle loading were synthesized as a model membrane to evaluate the role of fouling mechanisms.

One ofthe methods to prevent the membrane fouling is the incorporation of metal oxide nanoparticles,e.g.WO3[13]and TiO2[14].Acting as a photocatalyst,the UV-irradiated nanoparticles lead to the degradation of organic pollutants,which is also known as self-cleaning property[15].Having recently received growing attention,WO3nanoparticles show both hydrophilic and photocatalytic properties,enabling the membrane to decompose organic chemicals present in the environmental liquid waste[16,17].What is important to note is that the UV irradiation causes the semiconductor material,i.e.WO3,to involve in lightinduced redox processes.Upon irradiation,valence band electrons of the metal promote to the conduction band and leave a hole behind.These electron-hole pairs can either recombine or can interact separately with other molecules.The holes may react either with electron donors in the solution,or with hydroxide ions to produce powerful oxidizing species[18].

An excellent review of coating materials which show self-cleaning properties and the underlying phenomena that bring about such property may be found elsewhere[19].

As mentioned earlier,applying one of the Hermia fouling models to a filtration experiment where there is a feed containing a variety of particle sizes or a membrane having a wide pore size distribution may not provide a good fit and consequently,would not describe the underlying fouling mechanisms.Although Hermia models offer four basic easy-touse equations,they are usually precise only when there is a moderate variation in flux over time;otherwise,they suffer from lack of fit.Therefore,a multi-mechanism fouling model which takes into account the co occurrence of different fouling mechanisms may be favorable for accurate prediction of the permeate flux.

To the best of our knowledge,only a few studies have been dedicated to the development of a combined fouling model[20-22].This paper therefore aims to develop a robust multi-mechanism model to describe the time dependence of permeate flux in UF of two different feeds as prototypes of fluids which have relatively wide particle(or solute)size distribution.Towards this end,a resistance in series-parallel approach was taken in order to account for the four basic types of fouling mechanisms commonly assumed to be dominant during a course of filtration.

2.Materials and Methods

2.1.Model development

Fouling can be the result of many phenomena.On the one hand,complete blocking and cake layer formation mechanisms occur when solute sizes are larger than that of membrane pores.They simultaneously contribute to the surface deposition of solute molecules and hence increase the resistance to permeate flow,which constitute external fouling.On the other hand,standard blocking causes pore constriction and intermediate blocking leads to the partial bridging of membrane pores.The latter two may be categorized as fouling mechanisms which occur within the membrane structures,i.e.,internal fouling.There is an electrical analogy with mass transfer through the membrane that can be exploited in simulating fouling resistance concept.Given these hypotheses,a resistance-in-series model can be written as follows:

whereRtis the total resistance(m-1),Rs,Ri,RcandRglare resistances caused by standard blocking,intermediate blocking,complete blocking and cake layer,respectively.Darcy's law which relates the hydraulic resistance to the permeate flux during filtration is given by:

where ΔPis the trans-membrane pressure(Pa),μ is the viscosity of the filtered media(Pa·s)andJis the permeate flux(m·s-1).By writing resistances of each mechanism in the form ofR=ΔP/μJ,Eq.(3)can be rewritten as:

whereJs,Ji,JcandJglare the permeate fluxes governed by standard blocking,intermediate blocking,complete blocking and cake layer mechanisms,respectively.By simplifying and rearranging the latter equation:

Again,by applying Darcy's law,we have:

whereJpis the overall permeate flux simultaneously governed by all the above-mentioned mechanisms.When the pressure is kept constant,Eq.(8)can be substituted into Eq.(7)to reduce to the following equation:

Hermia equations(Table 1)can be inserted into Eq.(9)to determine the ultimate permeate flux as a function of time:

The multi-mechanism model explicitly takes into account the co occurrence of different fouling mechanisms throughout a run.As listed in Table 1,Ks(s-0.5·m-0.5),Ki(m-1),Kc(s-1)andKgl(s·m-2)represent the parameters for standard blocking,intermediate blocking,complete blocking and cake layer formation,respectively andJ0(m·s-1)is the initial permeate flux.In this model,it was assumed that each mechanism occurred independently and simultaneously,depending on the nature of feed and membrane types.

2.2.Materials and instrumentation

Whey solutions were supplied from Kalleh factory of dairy products(Amol,Iran).Land fill leachate was obtained from a land fill waste of municipal site located in Kiasar region of Sari,Mazandaran province,Iran.Characteristics of the whey and land fill leachate waste waters used in this study are listed in Table 2.Polysulfone(PSf)pellets were supplied by Udel P-1700 NT LCD and were used for UF membrane formingmaterial.N,N-dimethyl acetamide(DMAc)as a solvent was provided by BASF Co.(Germany).Polyethylene glycols(PEG)with molecular weight of 600 Da as the viscosity adjustment agent were purchased from Merck.WO3and Ammonium peroxydisulfate(APS)were used with the analytical grade.

A UV-Vis spectrophotometer(U-2800,Hitachi,Japan)was employed for protein absorption measurements of whey solutions.Among the proteins in whey(beta-lactoglobulin,alpha-lactalbumin,bovine serum albumin and immunoglobulins),beta-lactoglobulin contains 65%of total proteins.Regarding to the high concentration of beta-lactoglobulin in whey,this protein is chosen as a characteristic protein in our experiment and the protein concentration was measured by UV-Vis spectrometer at ambient temperature(22°C)and wavelength of 287 nm.Chemical Oxygen Demand(COD)levels of leachate were measured using a spectrophotometer of AL250 AQUALYTIC(Germany)at600 nm.Scanning Electron Morphology(SEM)analyses were utilized to characterize the morphology and surface of membranes by Philips-X130.In order to characterize the property of hydrophilic polymeric surfaces,contact angle measurement was used[23].The contact angle between water and the membrane surface was measured with Dataphysics-OCA 15 plus.De-ionized water was used as the probe liquid in all measurements.To investigate the effect of UV light on hydrophilicity,the modified membrane was illuminated by a UV lamp for 10 min at the distance of 5 cmand the contactangle was measured subsequently.The membranes' pore size distributions were studied by Atomic Force Microscopy(AFM)device:Nanosurf scanning probeoptical microscope(Easy Scan II,Swiss).

In order to measure the COD of feed and permeate solutions,the samples were micropipetted into the digestion solution vials.The vials were subsequently shaken vigorously by hand for 30 s in order for the contents to be mixed.Then,they were placed in the wells of a COD Reactor(Model Aqua Lytic,ET 108)and were heated at 150°C for 2 h.Every 10 min,the vials were shaken by hand to mix the condensed water and clear insoluble matter from the walls of the vials and put back in the reactor.After 2 h,the vials were removed from the reactor and cooled to room temperature.Afterward,the absorbance of the vial contents was measured at 600 nm using a spectrophotometer.The COD of land fill leachate used as a feed was obtained as 12420 mg·L-1.

2.3.Membrane preparation

PSf membranes were prepared by phase inversionviaimmersion precipitation method.Table 3 presents the composition of four nano composite membranes prepared in this study using PSf-based dope solutions.Casting solutions were prepared by dispersing WO3in DMAc and sonicated for about 30 min using the ultrasonic device.Then,PSf was dissolved in the resulting solution while being stirred at 200 r·min-1to prepare PSf/WO3solutions with different mass ratio.Homogeneous polymer solutions were kept overnight without stirring to allow complete release of the bubbles.The solutions were cast by film applicator on polyethylene/polypropylene non-woven fabric.The thickness of the cast film was adjusted to 75 μm thickness by adjusting the position of the casting knife.The membranes then immersed in a coagulation bath of pure water.Prepared membranes were keptovernight in the water bath to let the residual solvents and additives completely leach out.Finally,all of the membranes were washed with distilledwater and the modified ones were illuminated by a UV lamp for 10 min prior to use.

2.4.Bench scale membrane testing

The UF of whey wastewater was conducted using locally house made and assembled system.In the experimental trials,the cross flow batch concentration process was selected.The permeate flow was taken out of the loop and retentive flow was returned by a high pressure pump to the tank completely.The system consisted of a valve to control the applied pressure by the pump and a by-pass valve.These valves were used to control the flow and the pressure.The cell consisted of two cubic parts and was made of a specific alloy.A membrane with the area of 40 cm2was sandwiched between two parts.In the flow line,there were oil pressure gauges(0-2.5 MPa)to show the pressure of concentrated phase before the cell.There was a by-pass before feed inlet to recycle extra feed to the tank.There were two valves in the by-pass flow and retentate flow to adjusts the main flow rate and desired operating pressure.The flux of membranes during wastewater treatment was determined by weighing of permeate during 15 min under the TMP of3 and 5 bars.In addition,the experiments were carried out at temperature of 25 °C with flow rate of 10 L·min-1and the cross flow velocity of 0.42 m·s-1over the membrane.

2.5.Performance and fouling assay

The flux(J),through the membrane,could be described by the following equation:

whereVis the volume of permeate,Ais the membrane area and Δtis the permeation time.The retention ratios were calculated following the equation below:

whereCpandCf(g·L-1)are the concentrations of either COD or protein in permeate and feed solutions,respectively.Allprotein retentions were determined by spectrophotometry method at the wavelength of 287 nm to account for the major protein component available in the whey solutions,β-Lactoglobulin[24].

In order to evaluate the fouling resistant ability of membranes, flux recovery(FR)was introduced and calculated as follows:

whereJ0andJ1are the pure water of virgin and fouled membrane,respectively.The total fouling ratio(Rt)was defined and calculated as follows:

whereJpis the permeate flux of either whey or leachate andRtis the degree of total flux loss caused by total fouling.Reversible fouling ratio(Rr)and irreversible fouling ratio(Rir)were also calculated by following equations,respectively.

Obviously,Rtis the sum ofRrandRir[25].

2.6.Overall porosity and mean pore size measurements

Mean pore size measurements of membranes were performed following a gravimetric method reported elsewhere[26].Typically,3 membrane sheet replicates were stored 24 h in deionized water,then they were weighed after wiping excessive water.Then the membranes were dried in a vacuum oven at 80°C for 24 h and weighed again.The overall porosity(ε)was calculated by the following equation:

wherew1is the mass of the wet membrane,w2is the mass of the dry membrane(kg),Ais the membrane effective area(m2),dwis the water density(998 kg·m-3)andlis the membrane thickness(m).By having pure water flux and porosity data,the membrane mean pore radius(rm)can be determined using the following equation:

where η is the water viscosity(8.9 × 10-4Pa·s),Qis the volume of the permeate pure water per unit time(m3·s-1),and ΔPis the transmembrane pressure(Pa).

3.Results and Discussion

3.1.Characterization of the WO3/PSf membrane

For the morphological study of membranes,the surfaces of PSf and WO3/PSf ultra filtration membranes were characterized by SEM.The SEM micrographs of samples are displayed in Fig.1.It can be seen that surface of the neat membrane(Fig.1A)is totally uniform without any defects and cracks,which is a common morphology for the commercial dense asymmetric membrane.Compared to the neat membrane,WO3incorporated membrane clearly shows that nanoparticles are embedded on the membrane surface with the size range of 70-150 nm(Fig.1B).Atlower magnification(Fig.1C),for the membrane containing 1 wt%WO3,nanoparticles were also uniformly distributed on the membrane surface without any significant agglomeration which is believed to be due to the proper sonication during the preparation of dope solution.The AFM pore size distribution plots of neat and WO3-1%membranes at a scan size of 5 μm × 5 μm are shown in Fig.2.In these images,it seems that the bimodal distribution of pore sizes shifted to unimodal by addition of WO3nanoparticles in the casting solution.

The hydrophilic property of PSf/WO3membrane with WO3(WO3-1%)was characterized and compared with the neat PSf membrane through the measurement of water contact angle.Contact angles of the neat and modified membranes under UV radiation were 74.3°and 73.7°for the neat membrane before and after UV radiation,respectively.However,this value was 71.6°and 39.7°for the WO3-1%membrane before and after UV radiation,respectively.As observed,the contact angle had a sharp decrease in WO3-containing membranes after UV radiation,but had no significant decrease in the case where neat membranes were exposed to UV radiation.The reason why this decrease in contact angle(or increase in hydrophilicity and wettability)does not happen in neat membranes is the absence of WO3particles in neat membranes.That is,if there are no WO3particles on membranes,then they are not revived by UV radiation.As a result ultrahydrophilicity properties are only created on the membrane surface of UV-radiated WO3-containing membranes[27].As a brief explanation,photo-generated holes firstproduce OH radicals in the modified membrane which can resultin an increase of OH groups.This increases the surface energy which leads to the decrease of contactangle,and a super-hydrophilic surface is formed as a result[28].

Fig.1.SEM images of membrane surface obtained from(A)Neat membrane,(B)modified membrane(WO3-1%),(C)lower magnification of modified membrane(WO3-1%).

Fig.2.Pore size distribution plots of neat(left)and WO3-1%(right)membranes.

3.2.UF of land fill leachate

3.2.1.Model validation and fouling assessment

Model validation is possibly the most important step in the model building sequence.As a first step to checking the conformity of the proposed model with experimental data,the flux of leachate as a function of time was plotted for membranes with different WO3contents.Fig.3 shows the experimental data obtained during constant pressure UF of leachate with four membranes incorporating WO3nano particles whose mass fractions ranged from 0 to 2 wt%.The membrane synthesized with 0 wt%WO3is named ‘neat’membrane throughout this work.Nonlinear regression optimization was performed using MATLAB R2013a(Math Works)software.As illustrated,there is an excellent agreement between the model predictions and experimental data.Further,these results show that the combined model is valid regardless of membranes with different average pore size for an identical feed.

It must be mentioned that the flux value att=0 in every experimental run was obtained by measuring the flux in the first 30 s period of the run.Fig.3(b-d)imply that the fluxes of leachate through WO3-incorporating membranes which underwent the UV irradiation were all higher than non-UV-radiated membranes.The presence of WO3particles on the membrane surface plays two majorroles when activated by UV irradiation:photocatalysis and ultrahydrophilicity[29].The photocatalysis of WO3particles on the membrane surface produces oxidant agents such as hydroxyl radicals and superoxide radical anions.These strong oxidant groups repel the natural organic matter(NOM)and prevent their deposit on the membrane surface.It is noteworthy that NOM is the most fouling-causing substance available in the leachate[30]which deposits on the membrane surface and form a cake layer and consequently increases the hydraulic resistance over time.In addition to photocatalysis,the photoinduced hydrophilicity of WO3results in the spread of water layer all over the membrane surface,which consequently lowers the membrane affinity to organic matter due to van der Waals forces and hydrogen bonds.This makes the membrane less likely to be deposited by(often organic)particulates and slows down the formation of the cake layer,which is known as self-cleaning effects[31].

In Fig.3(a),the flux data of leachate through neat membranes with and without UV fall exactly on each other.This was expected because no WO3was present on neat membranes to show photocatalytic activity and as a result flux values showed no variation before and after UV irradiation.It indicates that mere UV radiation on the membrane surface is not enough for creating the self-cleaning property.

Looking at Fig.3 carefully,it can be observed that by moving from lower WO3concentrations to higher ones, flux of leachate is more enhanced.The highest flux,i.e.35.92 L·m-2·h-1,corresponded to the membrane containing 2 wt%WO3.This may be attributed to the surface chemistry of the nano composite membranes.That is,at the defined area of membrane surface,by reduction of the pore size,the number of pores is increased and this results in a higher porosity[32].The overall porosities of the neat membrane,WO3-0.5%,WO3-1%and WO3-2%were obtained as 28.4%,13.0%,49.2%and 84.9%and mean pore radii were 10.4,8.7,8.1 and 7.7 nm,respectively.Although the average pore size is the least in the case of WO3-2%membrane(7.7 nm),the porosity corresponding to this membrane is the highest of all(84.9%).This higher porosity is believed to be the reason why the flux was higher under this condition.The same behavior is observed in a similar study by Yanget al.[33],who studied Polysulfone/TiO2nanocomposite UF membranes.In fact,the porosity of membrane is increased after modification by metaloxide nanoparticles[34].Table 4 compares the values ofR2,relative errors of prediction(REP),initial permeate fluxes(J0)and total resistances(Rt)obtained in UF of leachate by membranes with different WO3contents.As observed byR2values,the proposed model represents a perfect match with the experimental data.The precision of the model studied in this work was evaluated as the relative error of prediction(REP)calculated following the equation below[35]:

Fig.3.Permeate flux predictions by the combined model for the UF of land fill leachate through different membranes at 300 kPa.

Table 4Comparison of the measures of fit calculated for UF of leachate by different membranes

wheren,Jpred,andJactare the total number of flux data samples,predicted flux and the actual flux of the membrane,respectively.All relative errors of flux predictions committed by the model were less than 2.5%except for the WO3-2%UV membrane which was 7.65%.

Initial permeate fluxes predicted by the model were too close to the experimental data obtained in almost every cases.Total resistances of the membrane at the end of filtration time(120 min)were calculated withRtdefined byRt=ΔP/μJ.Also as seen in Table 4,there is a close agreement between total hydraulic resistances predicted by the model and the actual values.

Fig.4 illustrates the changes in COD removal of membranes as a function of WO3concentration in the casting solutions.COD removal enhanced with increasing WO3concentration.This is because the addition of WO3nanoparticles in the PSf casting solution leads to the formation of membranes with smaller surface pore sizes.Needless to say,the membranes with smaller pore sizes exhibit higher COD removal values.As shown in Fig.4,UV radiation clearly had an important impact on improving COD removal in all modified membranes because of photocatalytic property of WO3[18].Again in the case of neat membranes,no significant changes in COD removal is seen before and after UV irradiation due to the reasons mentioned before.

Fig.4.Leachate COD removal ratios of different membranes with and without UV radiation.

3.2.2.Effect of WO3on the antifouling properties of membranes

For further investigation of flux behavior and the anti fouling property of the membranes,several parameters such as reversible fouling ratio(Rr),irreversible fouling ratio(Rir), flux recovery ratio(FR)and total fouling ratio(Rt),were calculated and listed in Table 5.Irreversible fouling is usually attributed to the pore blocking mechanisms and the reversible fouling is attributed to the cake filtration[36]andRtis the sum of these parameters and shows the total resistance of fouling against permeation flux.Table 5 shows that the amounts of FR increased from 51.5%for the neat membrane,to 69.9%for WO3-2%UV-irradiated membrane.Also,irreversible fouling was decreased from 48.5%to 30.1%for the said membranes.

Table 5Flux recovery,reversible fouling,irreversible fouling and total fouling ratios for different membranes in filtration of leachate wastewater

As observed in Table 5,mere addition of WO3to the membrane without UV irradiation has a negative effect on flux recovery in all cases.For instance,FRof WO3-1%membrane drops sharply from 58.5%to 15.6%for with and without UV radiated membranes,respectively,orRtincreases from 53.6%to 92.4%.This is because the WO3nanoparticles on the membrane surface without UV radiation blocks the fluid passage and decreases the average pore size which will result in higher rejections than neat membrane,as confirmed by Fig.4.

By looking more carefully at the fouling parameters represented in Table 5,it is seen that a reversible fouling of 23.5%corresponds to the WO3-2%UV-irradiated membrane.It can be deduced that the overall fouling resistance can even be further reduced by increasing the cross- flow velocity of feed over the membrane surface.However,fouling due to the adsorption of foulants on the membrane surface or within the membrane matrix is essentially irreversible.Therefore,Back flushing,chemical or biological cleaning may then be the only effective measures to mitigate the fouling in cases with high irreversible fouling[37].

It can be concluded that the addition of WO3in the PSf casting solution has a very important effect on the membrane properties:an increase in the hydrophilicity of PSf membrane.As hydrophilic surfaces tend to have a strong affinity with polar components,they lead to the higher permeability of pure water and repel the hydrophobic substances such as NOM.Hydrophilic property therefore as a key factorprevents form cake layer formation on the membrane surface[38].

3.3.UF of dairy wastewater

In the second part of this work,the filtration data obtained from cross- flow UF of dairy wastewater by means of neat membranes was analyzed in order to both scrutinize the underlying fouling mechanisms and to further examine the model validation.Waste whey solutions were obtained from cheese processing unit of a local dairy factory.The flux of whey solutions as a function of time was plotted for different TMPs and concentrations in Fig.5.In addition,a comparison between the fitting of the experimental data to the proposed model and Hermia models has also represented this figure.An excellent agreement is seen between the flux data with the combined model in every operating condition tested while Hermia models showed a significant lack of fit in all conditions except in the case of Feed concentration and TMP=300 kPa.

Fig.5.Permeate flux predictions by the combined model vs.Hermia models for the UF of whey solutions at different operating conditions.

Table 6 compares the calculatedR2values of the models studied in this work and Table 7 lists the predicted and actual initial permeate fluxes.For all Hermia models applied to our experimental conditions,high precisions were obtained only when permeate flux varied slightly with time which was the case with Feed concentration with TMP=300 kPa.When evaluating the performance of different models,it should be noted that comparison ofR2values is not to be made across samples.In other words,R2values must be compared to the same test and between different models,and not at the same model and between different tests[1].Therefore,it can be concluded that the cake layer formation model,among all Hermia models,provided the best fits to data for all the experimental conditions tested except for the test with Feed concentration and TMP=300 kPa.The presented model however strongly enhanced the fitting to experimental data in comparison to Hermia models.Whey waste waters usually contain a variety of proteins including β-Lactoglobulin(β-Lg),α-Lactalbumin and Bovine Serum Albumin(BSA),in the order of abundance.Moreover,whey contains lactose,minerals and small amounts of fat.It is believed that these components simultaneously contribute to the occurrence of fouling of different mechanisms.As discussed earlier in this paper,particles smaller than membrane pores such as BSA with stokes radius of3.48 nm[39],lactose or minerals may deposit on the pore walls to cause standardblocking;particles larger than membrane pores such as protein macromolecules and fat globules may completely seal the entrance of the membrane,i.e.,complete blocking,or form a cake layer;the particles which have the same size as membrane pores may bridge the pores but not completely block them,i.e.,intermediate pore blocking.Altogether,it seems that the combination of all pore blockage and cake formation mechanisms lead to the fouling of membrane,which is explicitly accounted for in the proposed model.This is probably the reason why the combined model was more successful in fitting to the experimental data.Fig.6 illustrate the relative errors committed by different Hermia modelsvs.combined model in the prediction of permeate flux at TMP of 300 kPa.As illustrated in Fig.6,an increase in concentration at a low pressure significantly decreases the error committed by all Hermia models while variation of the error of combined model is mild.However,increased concentration slightly changes the error at the upper pressure.Clearly,the best results corresponded to the combined model with REP＜2.3%for all the conditions tested.

Table 6Comparison of R2 values calculated for Hermia models at different operational conditions for UF of dairy wastewater through neat membranes

Table 7Initial permeate flux values predicted by Hermia models and the proposed model vs.experimental values

Fig.6.Comparison between relative errors of predictions for different models studied in the present work at TMP=300 kPa.

The protein retention values at different operating conditions obtained by neat membrane were between 69%-83%.Comparing this result with Fig.4,it can be deduced that neat PSf membranes show an improved retention behavior for filtration of dairy wastewater than land fill leachate.In all cases,by having the concentration constant,it was observed that retention slightly decreased by increasing the TMP.This may be due to the fact that higher TMPs favor the accumulated solutes which are trapped in the pores to pass through the membrane and facilitates the permeation of solutes across the membrane.It should be noted that calculated retention values only assign to β-Lg protein which is the most abundant whey protein in bovine milk and many other mammals[24].

Table 8 lists several fouling parameters obtained during UF of dairy wastewater at different operational conditions.As expected,when moving from high concentrations to lower ones,Rtattains lower values.This is because of low convective transport of proteins to the membrane surface at low feed concentrations and thus forming a thinner layer of cake compared to high feed concentrations.Moreover,by considering the low values of FR,it appears that neat membranes are susceptibleto protein fouling and they suffer from severe fouling.The authors believe that if the PSf membranes are not to be modified,then whey wastewater should be treated by pretreatment methods[40-42]prior to UF by the synthesized membrane to protect the membrane and make the filtration a sustainable and economical process.

Table 8Flux recovery,reversible fouling,irreversible fouling and total fouling ratios for different operational conditions in filtration of whey wastewater

Fig.7.Rate of flux decline for the UF of whey solutions at different TMPs and 15-times diluted feed solution.Solid curves are model calculations.

Fig.7 shows the rate of flux decline(RFD)evaluated from the filtrate flux as:solid curves in Fig.7 represent model calculations.As observed,at the beginning of the filtration,RFD decreased with increasing TMP.Both curves are concave down during the early stages of filtration and concave up at longer times.This was in agreement with the similar results reported by Kelly and Zydney[43].These results indicate that pore blockage mechanism is the dominant phenomenon in the early filtration times,whereas cake layer formation mechanism plays a more important role in the long term.The inflection point between these two regions reveals the transition between the two mechanisms.The observed initial increase in RFD is justifiable by the two-step fouling model concept previously proposed by Kellyet al.[44]which suggests the existence of two separate fouling mechanisms.The proposed model provides a smooth transition from the pore blockage mechanism gradually giving way to the cake layer formation during the course of the filtration.The model calculations are in good agreement with the RFD data,although it tends to underestimate the data obtained from the case with TMP=300 kPa.

In order to further analyze the underlying fouling mechanisms, filtrate flux data was replotted as d2t/dV2versusdt/dVto obtain the value ofnin Eq.(1).The required derivatives were evaluated by following equations[22]:

As observed in Fig.8,both the experimental data yielded a negative slope followed by zero slope on the log-log plot.The same results are seen in the work conducted by Duclos-Orselloet al.[20].According to the authors,the initial decrease in d2t/dV2indicates a reduction of flux decline(dJ/dt).During the initial period of filtration,dJ/dtdeclines much more rapidly thanJ.As all the classical fouling mechanisms offer non-negative slopes,these results cannot be explained by any of them.However,this behavior is accurately predicted by the proposed fouling model.The data at longer filtration times yields a zero slope which suggests the cake layer formation mechanism.It is noteworthy that the data obtained at higher whey concentration started at higher value of dt/dV.This is because of the more rapid flux decline for higher feed concentration which gives a lower initial flux,as observed in Fig.5(b)and(d).

Fig.8.Flux decline analysis for UF of whey solutions of different concentrations at TMP=500 kPa.

4.Conclusions

In this study,Polysulfone(PSf)membranes were modified by incorporating different contents of WO3nanoparticles.This led to the formation of membranes with smaller surface pore sizes,i.e.,from 10.4 nm to 7.7 nm.Furthermore,the results obtained from water contact angle measurements indicated that high hydrophilicity and wettability have been provided by UV radiation on the surface of WO3/PSf membranes.UV-irradiated WO3/PSf membranes were shown to cause a drastic increase in both COD removal and flux of leachate because of selfcleaning and anti-fouling properties.These UV-modified membranes also exhibited higher flux recovery and less irreversible fouling than neat membranes.

In order to fur ther analyze the fouling mechanisms,a semi-empirical model was developed based on a resistance in series approach.The model explicitly accounts for both internal and external fouling mechanisms by combining classical fouling models.Applying the model to the cross- flow UF data of land fill leachate and dairy wastewater has shown good precisions,with best and worst relative error of prediction of0.29%and 7.6%,whereas those values were 1.6%and 25%for Hermia classical fouling models.Moreover,the model exhibited a smooth transition between the common two-step successive pore blockage—cake filtration fouling phenomena.

[1]A.Salahi,M.Abbasi,T.Mohammadi,Permeate flux decline during UF of oily wastewater:Experimental and modeling,Desalination251(1)(2010)153-160.

[2]J.Hermia,Constant pressure blocking filtration law application to powder-law non-Newtonian fluid,Trans.Inst.Chem.Eng.60(1982)183-187.

[3]K.-J.Hwang,T.-T.Lin,Effect of morphology of polymeric membrane on the performance of cross- flow micro filtration,J.Membr.Sci.199(1)(2002)41-52.

[4]M.C.V.Vela,S.á.Blanco,J.L.García,E.B.Rodríguez,Analysis of membrane pore blocking models applied to the ultra filtration of PEG,Sep.Purif.Technol.62(3)(2008)489-498.

[5]A.Salahi,A.Gheshlaghi,T.Mohammadi,S.S.Madaeni,Experimental performance evaluation of polymeric membranes for treatment of an industrial oily wastewater,Desalination262(1)(2010)235-242.

[6]T.Mohammadi,M.Kazemimoghadam,M.Saadabadi,Modeling of membrane fouling and flux decline in reverse osmosis during separation of oil in water emulsions,Desalination157(1)(2003)369-375.

[7]Y.Pan,W.Wang,T.Wang,P.Yao,Fabrication of carbon membrane and micro filtration of oil-in-water emulsion:An investigation on fouling mechanisms,Sep.Purif.Technol.57(2)(2007)388-393.

[8]T.Arnot,R.Field,A.Koltuniewicz,Cross- flow and dead-end micro filtration of oily water emulsions:Part II.Mechanisms and modelling of flux decline,J.Membr.Sci.169(1)(2000)1-15.

[9]M.C.V.Vela,S.á.Blanco,J.L.García,E.B.Rodríguez,Analysis of membrane pore blocking models adapted to cross flow ultra filtration in the ultra filtration of PEG,Chem.Eng.J.149(1)(2009)232-241.

[10]E.-E.Chang,S.-Y.Yang,C.-P.Huang,C.-H.Liang,P.-C.Chiang,Assessing the fouling mechanisms of high-pressure nano filtration membrane using the modified Hermia model and the resistance-in-series model,Sep.Purif.Technol.79(3)(2011)329-336.

[11]E.-E.Chang,Y.-C.Chang,C.-H.Liang,C.-P.Huang,P.-C.Chiang,Identifying the rejection mechanism for nano filtration membranes fouled by humic acid and calcium ions exemplified by acetaminophen,sulfamethoxazole,and triclosan,J.Hazard.Mater.221(2012)19-27.

[12]I.N.H.M.Amin,A.W.Mohammad,M.Markom,L.C.Peng,N.Hilal,Analysis of deposition mechanism during ultra filtration of glycerin-rich solutions,Desalination261(3)(2010)313-320.

[13]N.Shafaei,M.Peyravi,M.Jahanshahi,G.Najafpour,Self-cleaning behavior of nanocomposite membrane induced by photocatalytic WO3nanoparticles for land fill leachate treatment,Korean J.Chem.Eng.33(10)(2016)2968-2981.

[14]G.Romanos,C.Athanasekou,V.Likodimos,P.Aloupogiannis,P.Falaras,Hybrid ultrafiltration/photocatalytic membranes for efficient water treatment,Ind.Eng.Chem.Res.52(39)(2013)13938-13947.

[15]A.Rahimpour,S.Madaeni,A.Taheri,Y.Mansourpanah,Coupling TiO2nanoparticles with UV irradiation for modification of polyethersulfone ultra filtration membranes,J.Membr.Sci.313(1)(2008)158-169.

[16]O.Carp,C.L.Huisman,A.Reller,Photoinduced reactivity of titanium dioxide,Prog.Solid State Chem.32(1)(2004)33-177.

[17]P.Ragesh,V.A.Ganesh,S.V.Nair,A.S.Nair,A review on‘self-cleaning and multifunctional materials’,J.Mater.Chem.A2(36)(2014)14773-14797.

[18]S.Chakrabarti,B.K.Dutta,Photocatalytic degradation of model textile dyes in wastewater using ZnO as semiconductor catalyst,J.Hazard.Mater.112(3)(2004)269-278.

[19]I.P.Parkin,R.G.Palgrave,Self-cleaning coatings,J.Mater.Chem.15(17)(2005)1689-1695.

[20]C.Duclos-Orsello,W.Li,C.-C.Ho,A three mechanism model to describe fouling of micro filtration membranes,J.Membr.Sci.280(1)(2006)856-866.

[21]G.Bolton,D.LaCasse,R.Kuriyel,Combined models of membrane fouling:Development and application to micro filtration and ultra filtration of biological fluids,J.Membr.Sci.277(1)(2006)75-84.

[22]C.-C.Ho,A.L.Zydney,A combined pore blockage and cake filtration model for protein fouling during micro filtration,J.Colloid Interface Sci.232(2)(2000)389-399.

[23]A.Asatekin,S.Kang,M.Elimelech,A.M.Mayes,Anti-fouling ultra filtration membranes containing polyacrylonitrile-graft-poly(ethylene oxide)comb copolymer additives,J.Membr.Sci.298(1)(2007)136-146.

[24]S.Brownlow,J.H.M.Cabral,R.Cooper,D.R.Flower,S.J.Yewdall,I.Polikarpov,A.C.T.North,L.Sawyer,Bovine β-lactoglobulin at 1.8 ? resolution — Still an enigmatic lipocalin,Structure5(4)(1997)481-495.

[25]M.Peyravi,A.Rahimpour,M.Jahanshahi,Thin film composite membranes with modified polysulfone supports for organic solvent nano filtration,J.Membr.Sci.423-424(2012)225-237.

[26]H.Yu,X.Zhang,Y.Zhang,J.Liu,H.Zhang,Development of a hydrophilic PES ultrafiltration membrane containing SiO2@N-Halamine nanoparticles with both organic antifouling and antibacterial properties,Desalination326(2013)69-76.

[27]S.Madaeni,N.Ghaemi,Characterization of self-cleaning RO membranes coated with TiO2particles under UV irradiation,J.Membr.Sci.303(1)(2007)221-233.

[28]H.Irie,K.Hashimoto,Photocatalytic active surfaces and photo-induced high hydrophilicity/high hydrophobicity,Environmental Photochemistry Part II,Springer 2005,pp.425-450.

[29]S.You,C.Wu,Fouling removal of UF membrane with coated TiO2nanoparticles under UV irradiation for effluent recovery during TFT-LCD manufacturing,Int.J.Photoenergy2013(2013).

[30]C.Liu,S.Caothien,J.Hayes,T.Caothuy,T.Otoyo,T.Ogawa,Membrane Chemical Cleaning:From Art to Science,Pall Corporation,Port Washington,NY 11050,2001.

[31]K.Guan,Relationship between photocatalytic activity,hydrophilicity and selfcleaning effect of TiO2/SiO2films,Surf.Coat.Technol.191(2)(2005)155-160.

[32]M.Peyravi,A.Rahimpour,M.Jahanshahi,Developing nanocomposite PI membranes:Morphology and performance to glycerol removal at the downstream processing of biodiesel production,J.Membr.Sci.473(2015)72-84.

[33]Y.Yang,H.Zhang,P.Wang,Q.Zheng,J.Li,The influence of nano-sized TiO2fillers on the morphologies and properties of PSF UF membrane,J.Membr.Sci.288(1)(2007)231-238.

[34]N.Hilal,M.Khayet,C.J.Wright,Membrane Modification:Technology and Applications,Taylor&Francis,New York,2012.

[35]Y.S.Al-Degs,A.H.El-Sheikh,M.A.Al-Ghouti,B.Hemmateenejad,G.M.Walker,Solidphase extraction and simultaneous determination of trace amounts of sulphonated and azo sulphonated dyes using microemulsion-modified-zeolite and multivariate calibration,Talanta75(4)(2008)904-915.

[36]T.Jiang,M.D.Kennedy,W.G.van der Meer,P.A.Vanrolleghem,J.C.Schippers,The role of blocking and cake filtration in MBR fouling,Desalination157(1)(2003)335-343.

[37]Z.Wang,J.Ma,C.Y.Tang,K.Kimura,Q.Wang,X.Han,Membrane cleaning in membrane bioreactors:A review,J.Membr.Sci.468(2014)276-307.

[38]M.Peyravi,A.Rahimpour,M.Jahanshahi,A.Javadi,A.Shockravi,Tailoring the surface properties of PES ultra filtration membranes to reduce the fouling resistance using synthesized hydrophilic copolymer,Microporous Mesoporous Mater.160(2012)114-125.

[39]I.Axelsson,Characterization of proteins and other macromolecules by agarose gel chromatography,J.Chromatogr.A152(1)(1978)21-32.

[40]S.A.Patil,V.V.Ahire,M.H.Hussain,Dairy wastewater—A case study,Int.J.Res.Eng.Technol.03(09)(2014)30-34.

[41]T.V.Adulkar,V.K.Rathod,Ultrasound assisted enzymatic pre-treatment of high fat content dairy wastewater,Ultrason.Sonochem.21(3)(2014)1083-1089.

[42]B.Demirel,O.Yenigun,T.T.Onay,Anaerobic treatment of dairy wastewaters:A review,Process Biochem.40(8)(2005)2583-2595.

[43]S.T.Kelly,A.L.Zydney,Mechanisms for BSA fouling during micro filtration,J.Membr.Sci.107(1)(1995)115-127.

[44]S.T.Kelly,W.S.Opong,A.L.Zydney,The influence of protein aggregates on the fouling of micro filtration membranes during stirred cell filtration,J.Membr.Sci.80(1)(1993)175-187.