Caihong Wang ,Aishu Wei,Hao Wu ,Fangshu Qu ,Weixiong Chen ,Heng Liang ,*,Guibai Li

1 State Key Laboratory of Urban Water Resource and Environment(SKLUWRE),Harbin Institute of Technology,73 Huanghe Road,Nangang District,Harbin 150090,China

2 The Architecture Design and Research Institute of Guangdong Province,Guangzhou 510100,China

3 China Urban Construction Design&Research Institute,Beijing 100120,China

4 School of Environment and Energy,South China University of Technology,Guangzhou 510006,China

1.Introduction

Low-pressure hollow membrane filtration is considered to be one of the most promising technologies in drinking water production[1].Although MF and UF technologies have been widely studied and implemented,membrane fouling is still the technical bottleneck of these technologies[2,3].Recently in China,due to the gradual degradation of source water quality,the upgrades of water treatment processes have become the top priority for water supply companies.UF technology,which is recognised as an efficient process with great effluent qualities,has become one of the preferred drinking water treatment technologies for upgrading and expanding existing facilities.Since 2009,a large number of water treatment plants(WTPs)using membrane filtration technology have been constructed,such as the Dongying Nanjiao WTP(50000 t·d-1),the Beijing No.9 WTP(90000 t·d-1),the Shanghai Qingpu No.3 WTP(100000 t·d-1)and the Wuxi Zhongqiao WTP(150000 t·d-1).With the application of the membrane filtration process in a large numberofWTP upgrade projects,the development of effective chemical cleaning processes with fewer experiments has become a challenging day-to-day topic.However,the work on the membrane cleaning was much less than the work on understanding the membrane fouling mechanism[4].Moreover,a majority of the research on membrane cleaning focused on physical cleaning[5].

Currently,a model developed by Feng et al.for cleaning of fouled membranes has been widely recognised.The model suggested that membrane fouling was governed mainly by the electrostatic interaction and the hydrophobic/hydrophilic interaction between the membranes and foulants.Electrostatic repulsion majorly enhances the cleaning efficiency[6].The membrane cleaning options can be categorised as chemically enhanced backwashing(CEB)and cleaning in place(CIP),or as regeneration cleaning and maintenance cleaning,according to its purpose.Caustics(NaOH),oxidants(H2O2and NaClO),acids(weak HCl,citric acid,etc.)and surfactants were widely used chemical cleaning agents.Recently,studies suggested that combination cleaning(the sequential use of caustics,oxidants and acids)could obtain a relatively high cleaning efficiency,due to the electrostatic interactions between foulants and membrane surfaces[7-11].

Membrane chemical cleaning can be illustrated in 6 steps[5,12,13]:1)the introduction of the cleaning agent to the filter feed;2)the cleaning agent makes contact with the foulant layers;3)the cleaning agent travels through the foulant layers to the membrane surface;4)the cleaning agent dissolves and detaches the foulant;5)the reacted cleaning agent and suspended foulants are transported to the interface;and 6)the waste matter(used cleaning agent and detached foulants)is transported to the feed.

It is difficult to develop models from these complex factors to precisely evaluate the effects of chemical cleaning to the efficiency[5].Therefore,current studies on chemical cleaning have been limited to a small number of selected schemes[14,15].Bartlett et al.attempted to use the recovery rate of membrane flux as the indicator to quantitatively evaluate the impact of various cleaning factors on cleaning efficiency and suggested that it was possible to optimise the chemical cleaning process by using appropriate modelling methodology[16].Chen et al.introduced factorial study in the testing of chemical cleaning processes and used the analysis of variance(ANOVA)to assess the impact of each factor[17].Tian et al.considered that the chemical cleaning efficiency of the UF membranes could be indicated by the removal rate of the irreversible resistance of membrane[18].

In this study,software Design Expert Version 8.0 was used to develop the experimental plan and to response surface model within the limits of the experiment.The significance of the model was analysed by performing an analysis of variance(ANOVA).The interactions of the four factors,i.e.,sodium hydroxide concentration,sodium hypochlorite concentration,citric acid concentration and cleaning duration,affecting the chemical cleaning efficiency were assessed.The objective was to demonstrate that fitted response surface model could serve as a tool to perform and optimise control factors in the chemical cleaning process of a UF membrane.

2.Materials and Methods

2.1.Materials

The PVDF UF membrane,produced by the Suzhou Litree Ultrafiltration Membrane Technology Co.Ltd.,was used in the experiment.According to the membrane producer,the average pore size of the given UF membrane was approximately 0.01 μm.The effective surface area of the membrane module was 0.003 m2.Detailed information on the PVDF membrane module used in the experiment can be found in a previous paper[18].The humic acids,bought from Sigma-Aldrich Co.,were used as the foulant in the modelling.The sodium hydroxide(NaOH),NaClO and citric acid were of analytical purity.Milli-Q water was used to prepare the feed water and the cleaning solution.

2.2.Contamination and cleaning experiments

The resistance-in-series model[2]is widely used to describe the fouling properties of membranes and can be expressed as the formula below[19,20]:

where Rmis the intrinsic membrane resistance(m-1);Rfis the total foulant resistance,including reversible(Rrev)and irreversible(Rirr)fouling resistance;ΔP is the transmembrane pressure(TMP,Pa);η is the dynamic coefficient of viscosity(Pa?s);and J is the filtrate flux(m3·m-2?s-1).

Prior to the experiments,the virgin membranes were placed into the filtering vessel after soaking in demineralised water for 30 min.To accelerate the fouling process,a relatively high membrane flux of 40 L·m-2was applied.First,the membrane was used to filter the demineralised water for 1 h to reach a stable transmembrane pressure,which was used to calculate the intrinsic membrane resistance[21].Then,a 20 mg·L-1humic acid solution at pH 7 was fed to the membrane for a 12-hour fouling test.After the fouling test,the residual foulants on the surface of the membrane were wiped off by a sponge to physically eliminate the reversible fouling resistance before the chemical cleaning test.It was followed by another 1 h of filtration of demineralised water to measure the irreversible fouling resistance prior to the chemical cleaning.The chemical cleaning process was then conducted according to the central composite experimental design.Finally,the irreversible fouling resistance was measured again by filtering demineralised water for 1 h.Each experiment set was applied to 4 sets of new membrane modules in parallel.

2.3.Experiment design

Central composite design(CCD)is a method that can be efficiently applied to develop second-order response models with limited numbers of factors n(2＜n＜6).Based on the CCD,the experimental design was used to develop a response surface model by quadratic approximation.In this experiment,the central composite design(CCD)together with the response surface methodology was used to simulate the PVDF membrane cleaning process.Four independent variables were considered(NaOH concentration,NaClO concentration,citric acid concentration and the cleaning duration).According to existing studies on chemical cleaning of membranes,the operating ranges and the levels of the considered variables were chosen and are given in Table 1[17,18].The experimental design is shown in Table 2.The design involved 30 runs.The removal rate of the irreversible membrane resistance[as described by Eq.(2)]was the response variable.The chemical cleaning efficiency of the UF membranes could be indicated by the removalrate of the irreversible resistance of membrane[18].

where RirrR is the removal rate of the irreversible resistance of membrane,is the irreversible fouling resistance before chemical cleaning(m-1),and Rairris the irreversible fouling resistance after chemical cleaning(m-1).

3.Results and Discussion

3.1.Model establishment and sensitivity analysis

As shown in Table 2,the influence of the four variables[NaOH concentration(X1),NaClO concentration(X2),citric acid concentration(X3)and cleaning duration(X4)]on chemical cleaning process was evaluated in terms of cleaning efficiency(RirrR)of the UF membrane.When a statistical analysis using the Design Expert Version 8.0 was performed on the four factors of study,the interactions among the factors could be determined.The regression equation for the response variables(in coded terms)obtained from the experimental data based on the interaction effects between the factors was as follows:

An ANOVA was applied to the regression model.The ANOVA result,with F=2.44 and the p-value＜0.05,showed that the fitting equation was significant,which indicated that the four factors had an effect on the cleaning efficiency.The NaClO concentration and the cleaning duration were the significant factors.As shown in Fig.1,the residuals generally falling on a straight line were distributed normally.The predicted value was close to the actual one.This observation implies

that the models were well fitted to the experimental samples and they are able to be applied in the analysis and prediction of the efficiency of the chemical cleaning of the PVDF UF membranes.

Table 1 Actual and coded levels of the variables used for the experiment design

Table 2 Actual and coded values of the variables used for the experiment design

Fig.1.R irr R of residuals plots.

3.2.Interaction of factors

The Design Expert Version 8.0 was used to depict three dimensional(3-D)response surfaces and contour plots of the dataset in Table 2.The graphs of the 3-D response surfaces and the contour plots of the different factors were shown in Fig.2 and Fig.3,respectively.A combination of 3-D response surfaces and contour plots was employed to assess the mutual interaction between of the following six sets of factors affecting the PVDF chemical cleaning efficiency of the PVDF UF membrane:the concentrations of NaOH and NaClO(Fig.2a),the concentrations of NaOH and citric acid(Fig.2b),the concentration of NaOH and the cleaning duration(Fig.2c),the concentrations of NaClO and citric acid(Fig.3a),the concentration of NaClO and the cleaning duration(Fig.3b),and the concentration of citric acid and the cleaning duration(Fig.3c).The plots were also used to identify the optimum range of each factor.

Fig.2.Response surfaces and contour plots showing the influence factors on R irr R(a.NaOH and NaClO concentrations,b.NaOH and Citric acid concentrations,c.NaOH concentration and cleaning duration).

Fig.3.Response surfaces and contour plots showing the influence factors on R irr R(a.NaClO and Citric acid concentrations,b.NaClO concentration and cleaning duration,c.Citric acid concentration and cleaning duration).

Utilising the shape of the contour plots to assess the interactions between the factors,a greater eccentricity ratio relates to a stronger interaction between two variables,whereas,a nearly round shape of the contour plot(on the right)indicates a weak interaction between two factors.The lines with labels on the contour plot,as well as the different colours of the response surfaces and the contour plots,showed a variety of degrees of interaction based on the RirrR[see Eq.(3)].

The mutual effect of the concentrations of NaOH and NaClO on the chemical cleaning efficiency was more significant than that of the concentrations of NaOH and citric acid showed in Fig.2.The mutual effect of the concentration of NaOH and the cleaning duration was similar to that of the concentrations of NaOH and NaClO.However,the slope of the curved surface on the 3-D response surface(see Fig.2a and c)indicated that the concentration level of NaClO and the cleaning duration demonstrated a greater impact on the chemical cleaning efficiency of the PVDF UF membrane compared with the concentration level of NaOH.In the result,these two factors played a major role in affecting the chemical cleaning efficiency,which turned out to be consistent with the results obtained from the ANOVA of the regression model.From the existing research works,the main effect of NaOH chemical cleaning is the increase of membrane hydrophoicity,while,the main effect of NaClO chemical cleaning is the oxidation ofmembrane organic pollution[18,19].Under the condition of this experiment,the main reason of membrane pollution came from the humic acids,the foulant,which result serious organic pollution,therefore,the concentration level of NaClO and the cleaning duration demonstrated a greater impact on the chemical cleaning efficiency of the PVDF UF membrane compared with the concentration level of NaOH.

Table 3 Optimised cleaning groups by RSM

The cleaning efficiency was improved with the increase of NaClO concentration,although the rate of improvement was gradually slowed.On the contrary,the cleaning efficiency was improved to a point and then reduced with the increase in NaOH concentration.A similar pattern was identified when evaluating the interaction between the concentration of NaOH and the cleaning duration.However,when assessing the interaction between the concentrations of NaOH and citric acid,the results suggest that concentrations that are either too high or too low(for both factors)could impede the increase of the cleaning efficiency.In addition,according to the response surface diagram,the cleaning efficiency declined significantly(blue area in Fig.2c)when the concentration of NaOH decreased to a certain level,which suggests that NaOH remained the essential factor in controlling the chemical cleaning efficiency,although increasing it alone could not significantly improve the cleaning efficiency.

The usage of sodium hydroxide during chemical cleaning could accelerate the dissolution of slight acidic organic matter,leading to the decomposition of the polysaccharide and protein foulants[5].Moreover,the sodium hydroxide could cause the expansion of natural organic matter attached to the membrane surface and thus promote the mass transformation of other cleaning agents across the cake layer[22].However,other studies have also revealed that the sodium hydroxide reacts with the hydrophilic groups of the membrane during chemical cleaning,which in turn decreased the permeability of the membrane[17,23].Overall,the sodium hydroxide was essential in PVDF membrane chemical cleaning,and its concentration needed to be controlled ate relatively low level.

Among the 3-D response surface and contour plots shown in Fig.3,Fig.3b shows that there was a significant relationship between the concentration of NaClO and the cleaning duration.The efficiency of chemical cleaning was improved with increases in the concentration of NaClO and the cleaning duration,and the efficiency reached the maximum value(approximately 150%)when both controlling variables were at their maxima within the boundaries of the experiment.The efficiency of the chemical cleaning could exceed 100%because the free chlorine provided by sodium hypochlorite could effectively oxidise organic foulant.Furthermore,the cleaning duration contributed to the diffusion of sodium hypochlorite within the fouling layer[1].However,although it could effectively remove the organic foulant,sodium hypochlorite inevitably harms the membrane material.Sodium hypochlorite increased the hydrophilicity of the membrane material and reduced the degree of cross linking of the membrane during the chemical cleaning process,which in turn improved the permeability of the membrane and decreased the pollutant removal capacity of the UF membrane[24].This might be the main reason why the efficiency of chemical cleaning exceeded 100%.

The significance of interaction between NaClO and citric acid(see Fig.3a)was the same as that between NaClO and NaOH(see Fig.2a).There was no significant interaction between the concentration of citric acid and the cleaning duration(see Fig.3c).The efficiency of the chemical cleaning initially increased and then decreased with continual increases in the concentration of citric acid.Although citric acid had a secondary impact on the efficiency of the chemical cleaning,it was necessary to maintain a moderate level of the concentration of citric acid in the experiment to ensure a satisfactory cleaning.Typically,an acid cleaning agent is used during chemical cleaning to remove salt and metallic hydrate foulants[9].In this experiment,however,humic acid was used as the tracer,which resulted in only a small amount of such inorganic chemicals in the membrane fouling.This might account for the poor efficiency of using citric acid in the chemical cleaning.

3.3.Chemical cleaning optimisation

As shown in Table 2,several cleaning schemes achieved excellent cleaning efficiencies as high as 100%.However,a high concentration of cleaning agents and a long cleaning duration would increase the operation cost and reduce the life span of the membrane[5].Therefore,it was illogical to seek only the maximisation of cleaning efficiency.Taking into account the widely accepted level of effective cleaning efficiency(over 80%)and the life-time of a membrane,a cleaning efficiency between 80%and 100%was set as the criteria for determining the optimum cleaning schemes for the PVDF UF membranes.Table 3 presents the results using the response surface model to predict different cleaning schemes,i.e.,54 satisfactory sets of chemical cleaning experiments.

In general,the experimental study showed that using the RSM to predict chemical cleaning process was technically feasible.However,in the experimental study,the response surface model accounts for only one type of pollutant,namely humic acid,and the predicted results have not yet been verified by testing.Therefore,in practise,the model has only limited implications,combined with a high level of uncertainty.In future studies,the predicted results will be verified by actual tests.The results would then be used to further modify the model.In addition,future studies could also use real source water as the feed water to obtain models for chemical cleaning prediction and optimisation with a better reflection of the reality.

4.Conclusions

This study focused on the optimisation of the chemical cleaning process for fouled UF membranes using the response surface model together with a central composite experimental design.Four factors,namely the NaOH concentration,NaClO concentration,citric acid concentration and cleaning duration,were considered in the experiment.The following conclusions can be drawn.

(1)The chemical cleaning process of the PVDF UF membrane scan be predicted and optimised by the RSM.

(2)The effect of the interaction between the NaOH concentration and the cleaning duration was the most significant on the cleaning efficiency of the PVDF UF membranes.

(3)The increase in sodium hydroxide showed little effect on improving the cleaning efficiency.However,lowering the NaOH concentration led to a significant drop in the cleaning efficiency,which suggests the necessity of controlling NaOH concentration below a certain but relatively low level during PVDF membrane cleaning.

[1]H.Huang,N.Lee,T.Young,A.Gary,J.Lozier,J.Jacangelo,Natural organic matter fouling of low-pressure,hollow- fiber membranes:Effects of NOM source and hydrodynamic conditions,Water Res.41(17)(2007)3823-3832.

[2]R.Fabris,E.K.Lee,C.W.K.Chow,V.Chen,M.Drikas,Pre-treatments to reduce fouling of low pressure micro- filtration(MF)membranes,J.Membr.Sci.289(1-2)(2007)231-240.

[3]X.Huang,M.Leal,Q.Li,Degradation of natural organic matter by TiO2photocatalytic oxidation and its effect on fouling of low-pressure membranes,Water Res.42(4-5)(2008)1142-1150.

[4]N.Porcelli,S.Judd,Chemical cleaning of potable water membranes:The cost benefit of optimisation,Water Res.44(5)(2010)1389-1398.

[5]N.Porcelli,S.Judd,Chemical cleaning of potable water membranes:A review,Sep.Purif.Technol.71(2)(2010)137-143.

[6]K.Feng,L.Hou,B.Tang,P.Wu,A self-protected self-cleaning ultra filtration membrane by using polydopamine as a free-radical scavenger,J.Membr.Sci.490(2015)120-128.

[7]Z.Zhang,M.W.Bligh,T.D.Waite,Ascorbic acid-mediated reductive cleaning of iron fouled membranes from submerged membrane bioreactors,J.Membr.Sci.477(2015)194-202.

[8]S.Platt,M.Nystr?m,Cleaning of membranes fouled by proteins to evaluate the importance of fully developed flow,Desalination 208(1-3)(2007)19-33.

[9]S.Strugholtz,K.Sundaramoorthy,S.Panglisch,A.Lerch,A.Brügger,R.Gimbel,Evaluation of the performance of different chemicals for cleaning capillary membranes,Desalination 179(1-3)(2005)191-202.

[10]A.Al-Amoudi,R.W.Lovitt,Fouling strategies and the cleaning system of NF membranes and factors affecting cleaning efficiency,J.Membr.Sci.303(1-2)(2007)4-28.

[11]F.Qu,H.Liang,Z.Wang,H.Wang,H.Yu,G.Li,Ultra filtration membrane fouling by extracellular organic matters(EOM)of Microcystis aeruginosa in stationary phase:Influences of interfacial characteristics of foulants and fouling mechanisms,Water Res.46(5)(2012)1490-1500.

[12]C.Ramos,F.Zecchino,D.Ezquerra,V.Diez,Chemical cleaning of membranes from an anaerobic membrane bioreactor treating food industry wastewater,J.Membr.Sci.458(2014)179-188.

[13]I.Levitsky,A.Duek,R.Naim,E.Arkhangelsky,V.Gitis,Cleaning UF membranes with simple and formulated solutions,Chem Eng.Sci.69(1)(2012)679-683.

[14]R.Field,D.Hughes,Z.Cui,U.Tirlapur,Some observations on the chemical cleaning of fouled membranes,Desalination 227(1-3)(2008)132-138.

[15]S.R.Gonzalez-Avila,F.Prabowo,A.Kumar,C.Ohl,Improved ultrasonic cleaning of membranes with tandem frequency excitation,J.Membr.Sci.415-416(2012)776-783.

[16]M.Bartlett,M.R.Bird,J.A.Howell,An experimental study for the development of a qualitative membrane cleaning model,J.Membr.Sci.105(1-2)(1995)147-157.

[17]J.P.Chen,S.L.Kim,Y.P.Ting,Optimization of membrane physical and chemical cleaning by a statistically designed approach,J.Membr.Sci.219(1-2)(2003)27-45.

[18]J.Tian,Z.Chen,Y.Yang,H.Liang,J.Nan,G.Li,Consecutive chemical cleaning of fouled PVC membrane using NaOH and ethanol during ultra filtration of river water,Water Res.44(1)(2010)59-68.

[19]X.Zheng,M.Ernst,M.Jekel,Identification and quantification of major organic foulants in treated domestic wastewater affecting filterability in dead-end ultra filtration,Water Res.43(1)(2009)238-244.

[20]H.K.Shon,P.J.Smith,S.Vigneswaran,H.H.Ngo,Effect of a hydrodynamic cleaning of a cross- flow membrane system with a novel automated approach,Desalination 202(1-3)(2007)351-360.

[21]L.Bai,F.Qu,H.Liang,J.Ma,H.Chang,M.Wang,G.Li,Membrane fouling during ultra filtration(UF)of surface water:Effects of sludge discharge interval(SDI),Desalination 319(2013)18-24.

[22]S.Hong,M.Elimelech,Chemical and physical aspects of natural organic matter(NOM)fouling of nanofiltration membranes,J.Membr.Sci.132(2)(1997)159-181.

[23]A.Idris,F.Kormin,M.Noordin,Application of response surface methodology in describing the performance of thin film composite membrane,Sep.Purif.Technol.49(3)(2006)271-280.

[24]M.Corbatón-Báguena,S.álvarez-Blanco,M.Vincent-Vela,J.Lora-García,Utilization of NaCl solutions to clean ultra filtration membranes fouled by whey protein concentrates,Sep.Purif.Technol.150(2015)95-101.