of 18

Application of coagulation–ultrafiltration hybrid process for drinking water treatment: Optimization of operating conditions using experimental design

All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Application of coagulation–ultrafiltration hybrid process for drinking water treatment: Optimization of operating conditions using experimental design
  Separation and Purification Technology 65 (2009) 193–210 Contents lists available at ScienceDirect Separation and Purification Technology  journal homepage: www.elsevier.com/locate/seppur Application of coagulation–ultrafiltration hybrid process for drinking watertreatment: Optimization of operating conditions using experimental design A.W. Zularisam a , b , c , A.F. Ismail a , ∗ , M.R. Salim c , Mimi Sakinah a , T. Matsuura d a  Advanced Membrane Technology Research Center, Universiti Teknologi Malaysia, 81300 Skudai, Johor, Malaysia b Faculty of Civil & Earth Resources, Universiti Malaysia Pahang, UMP, 25000 Gambang, Pahang, Malaysia c Faculty of Civil Engineering, Universiti Teknologi Malaysia, 81300 Skudai, Johor, Malaysia d Industrial Membrane Research Institute, Department of Chemical Engineering, University of Ottawa, 161 Louis Pasteur Street, P.O. Box 450,Station A, Ottawa K1N 6N5, Canada a r t i c l e i n f o  Article history: Received 21 July 2007Received in revised form15 September 2008Accepted 4 October 2008 Keywords: NOMFlocculationCoagulationMembraneDrinking water a b s t r a c t Coagulation application prior to ultrafiltration process was carried out to increase NOM removal andmembrane permeability. A systematic experimental design based on first order model of 2 4 full factorialdesign was used as an initial screening process to determine the significant variable factors and theirinter-relationshiptowardsnaturalorganicmatter(NOM)removalefficiency.ThefactorsconsideredwereSUVA (specific UV 254nm  absorbance), FRT (flocculation retention time), alum dosage and pH. Statisticalanalysisofresulthasshownthemaineffectofalum,pH,SUVAandthetwolevelinteractionsofalumandFRT, alum and SUVA, alum and pH, FRT and SUVA, and SUVA and pH were the significant model terms.Thereafterasecondordermodelwhichwasthe2 3 centralcompositedesigns(CCD)wasfurtheremployedto develop a mathematical correlation model between the significant factors for the optimum modes of operatingcondition,withrespecttoNOMremovalandmembranepermeability.Thealumdosageisfoundto be the most significant factor that influences the NOM removal and this is followed by the two levelinteractions of pH and SUVA, the main effect of pH, the main effect of SUVA, the two level interactionof SUVA and alum, the second order effect of SUVA and the second order effect of pH. In the case of membrane permeability, the main effect of alum dosage and the second order effect of pH provided theprincipal effect, whereas the second order effect of alum, the main effect of pH, the two level interactionof pH and SUVA, the two level interaction of SUVA and alum dose, and the main effect of SUVA providedthe secondary effect. The optimized values of 5.48, 3.24L/(mmg) and 3.0mgAl/L of pH, SUVA and alumdosewereobtainedrespectively,andwiththeseproposedoptimizedconditions,aNOMremovalof81.28%and permeability of 30.61 LMHBar was predicted. Thereafter via experimental validation process, 79.50%and 31.29 LMHBar of NOM removal and permeability was attained, respectively. In conjuction treatmentefficiency has shown excellent water quality that is well beyond the Malaysian and WHO drinking waterqualityregulationswith>96%colourremoval,about87%UV 254  removal,<0.2NTU,>99%suspendedsolidsremoval. and>99% of turbidity removal.© 2008 Elsevier B.V. All rights reserved. 1. Introduction Membrane filtration process involving microfiltration (MF),ultrafiltration(UF),nanofiltration(NF)andreverseosmosis(RO)inpotablewaterproductionhaveincreasedrapidlyforthepastdecadeand would potentially replace the conventional treatment whichconsist a complex series of unit operations, such as ozonation–precipitation–coagulation–flocculation–chlorination–gravel filtra-tion [1,2]. As part of drinking water treatment processes, ∗ Corresponding author. Tel.: +60 7 5535592; fax: +60 7 5581463. E-mail address:  afauzi@utm.my (A.F. Ismail). Ultrafiltration (UF) has been known to be effective for the removalof suspended solids, colloidal material (>0.1  m), inorganic par-ticulates and fatal microorganisms such as colliform, protozoa,giardia and cryptosporidium [3,4]. In spite of offering high per- meate flux and low pressure requirement, this filtration type isrelatively found to be less successful for wide application, par-ticularly in removing the dissolved organic matters (DOC) suchas natural organic matter (NOM) which is regarded as a majorfactor for membrane fouling and poor permeate quality [5,6]. In fact removal of NOM which is inherently present in the formof DOC or colloidal material is required, since NOM is a pre-cursor to the formation of carcinogenic disinfection by-products(DBP), aesthetically unattractive, carbon source for biofilm growth 1383-5866/$ – see front matter © 2008 Elsevier B.V. All rights reserved.doi:10.1016/j.seppur.2008.10.018  194  A.W. Zularisam et al. / Separation and Purification Technology 65 (2009) 193–210 in the distribution network and often reclaimed as an impor-tant factor for both reversible and irreversible membrane fouling[7,8]. In the case of drinking water filtration the colloidal mate-rial may reduce the membrane permeability (reversible fouling)by accumulating solutes on the adjacent membrane surface (con-centration polarization), precipitating and forming a cake or gellayer as they become supersaturated [9], while the DOC may cause irreversible fouling by altering the effective membranepores through precipitating and absorbing onto the membranematrix and within the pores space. Hence coagulation pretreat-ment prior to membrane filtration has been widely researched asa mean of removing non-settleable NOM content and reducingmembrane fouling through agglomeration of alum salts with thedestabilized particles before gravity precipitation [10–13]. Previ- ous studies using integrated conventional coagulation followed bydirect membrane filtration or an inline coagulation (without set-tling) combined with direct membrane filtration [10,11,14], have demonstrated effective control of fouling, improved membranepermeabilityandsuperiorpermeatequalitydespitehavingappliedon low quality water sources. In details, coagulation pretreatmentof NOM with metal salt ions of Al 3+ has been used to alter theNOM characteristics such as the relative content of NOM com-position (hydrophilic versus hydrophobic), NOM flocculant sizeand charge density possession, prior to UF membrane filtration.Changes in NOM characteristics due to the complexation of NOM-metal ions are greatly influenced by the type and alum dosage,pH, ionic strength, NOM solubility and hydrodynamic condition,which contributed to the overall efficiency of a coagulation–UFmembrane process [15–17]. In this study the coagulation process before membrane filtration was primarily employed to create sus-pended coagulated micro-flocs which were relatively bigger thanthe membrane molecular weight cut-off. Such approach was pre-sumably effective in promoting higher NOM removal efficiencythrough micro-floc cake layer rejection, thus would imply incoagulant dosage lower than the optimized level for aggregatesprecipitation, reduce flocculation retention time, increase mem-brane flowrate capacity, eliminate non-settleable flocs concernsand could effectively reduce pore clogging by particles. Particu-larly it was hypothesized that the hydrophobic fraction havingmore anionic binding sites than the hydrophilic fraction towardscationic aluminium hydrolysis product, was mainly contributingin forming the micro-flocs via charge neutralization. In the caseof hydrophilic fraction which has less electron-rich sites, wassupposed not participating in the charge neutralization with thealuminium hydrolysis product, thus apparently it was consideredplaying only a minor role in forming the micro-flocs and in beingassociated with pore adsorption and poor removal by membranefiltration. However with optimal setting of experimental factorsthe hydrophilic fraction could be effectively removed through sur-face adsorption on the neutralized flocculant or entrapped withinthe coagulated suspension, before being retained or absorbed onthe micro-floc cake layer that is formed on the membrane surface.In particular a systematic experimental design based on full fac-torial design was employed to determine the significant effect of pH, SUVA, FRT and alum towards NOM removal. A second leveldesignstudywasfurtherconductedwithcentralcompositedesignof response surface methodology to develop an empirical predic-tive model of NOM removal and membrane permeability, basedon the significant factors obtained from previous factorial designstudy. In view of the fact that only a considerable amount of research had been done, emphasis is placed towards quantitativeidentification of significant and interaction factors, and determi-nation of the best setting of variable factors that produce theoptimal outcome with respect to NOM removal and membranepermeability.  Table 1 Characteristics of Yong Peng NOM source water.UV 254nm  (cm − 1 ) 0.174 ( ± 0.015)DOC (mg/L) 5.36 ( ± 1.27)SUVA (L/(mgm)) 3.24Turbidity (NTU) 14.1 ( ± 5.1)Mn 2+ (mg/L) 0.28Ca 2+ (mg/L) 2.16Al 3+ (mg/L) 5.95The numbers in parenthesis are standards deviation. 2. Materials and methods  2.1. NOM source water, isolation and concentration apparatus Selected Malaysian surface water source of Yong Peng waterwas used as a model solution for the membrane permeability andNOMrejectionefficiencyofcoagulation-directUFexperiments.ThesurfacewaterqualityanditsNOMfractionscharacteristicsaresum-marized in Table 1, Table 2 and Fig. 1, respectively. NOM of Yong Peng water was pretreated with 0.45  m filter paper (WhatmanGC) prior to coagulation process while preparation of surrogatedSUVA was carried out by concentrating the NOM source using cel-lulose acetate reverse osmosis (RO) membrane. The concentrationmodule unit was designed and modified based on the principledeveloped by Serkiz and Perdue [18]. The detailed characteristics of the RO membrane and module used for concentrating the NOMsourcesaregivenelsewherebyIdrisetal.[19]andthewholeexper- imentset-upisillustratedinFig.2,respectively.Moreoverchemical properties of the feed stream solution was adjusted to the approx-imate value as the experimental designed requirement, by using0.1NHCl,0.1NNaOHanddeionizedwater.Therelationshipbetweenconcentrations of Yong Peng NOM and its UV absorbance recordedat 254nm were plotted in standard curves as tabulated in Fig. 3.The slope of line (Fig. 3.) represents SUVA [3.38L/(mmg)] which denotes an index of NOM relative aromaticity. In particular, SUVAis a function of humic fraction content within the NOM source andcan be employed as the surrogate for the relative humic concen-tration in the surface water. As can be inferred from Fig. 3 thereseems to appear a significant correlation between humic fractionand SUVA ( R 2 =98.15%) which can be substituted for the relativehumicfractionestimationbyusingUVandDOCmeasurements.TheSUVA value [3.38L/(mmg)] was found consistent with the in situSUVA of Yong Peng which was about 3.24 ± 0.17 (Table 1). In gen- eral high SUVA possession could be correlated with greater NOMaromaticity and negative charge density (Fig. 4) caused by the car- boxylic and phenolic moieties of hydrophobic acids [5], for NOM Fig.1.  Apparentmolecularweightdistributions(AMWD)ofYongPengNOMsourcewaterbyUFfractionationusingaseriesofcelluloseacetatemembraneswithMWCOsof 1kDa, 5kDa, 10kDa and 30kDa (Ultracel Millipore membrane).   A.W. Zularisam et al. / Separation and Purification Technology 65 (2009) 193–210  195  Table 2 DOC concentrations of NOM fractions for Yong Peng water (based on DOC and mass balance technique).Source water Hydrophobic DOC (%) Transphilic DOC (%) Hydrophilic DOC (%) Unfractionated DOC (%) SUVA [L/(mgm)]Yong Peng Intake 57 (3.05) 21 (1.12) 22 (1.17) 100 (5.34) 3.24The numbers in parenthesis are DOC concentration in mg/L. Fig.2.  SchematicdiagramofconcentrationunitofROmembrane(1)feed;(2)valve;(3) pump; (4) hollow fiber pressure housing; (5) flow meter; (6) permeate; (7)pressure gauge. Fig. 3.  Correlation curve of Yong Peng NOM concentration and its UV 254nm absorbance. Fig.4.  ThechargedensityperunitNOMmassofYongPengwaterbypotentiometrictitration [milliequivalent per gram of Carbon from DOC measurement; mequiv./gC,for carboxylic and phenolic groups of hydrophobic (HPO) fraction of NOM sourcewater]. source which possessed higher SUVA was supposed to imply ingreater NOM rejection due to higher coagulant demand towardscationic aluminium (Al) hydrolysis and subsequently would pro-duce micro-flocs of greater size. Thus SUVA factor was taken intoaccount in this study as one of the important factors since it wasrelatively influential in governing the NOM removal efficiency inthe coagulation-submerged UF system. The spectrophotometricbehaviour of Yong Peng NOM was further studied as a function of the pH solution (Fig. 5). It was shown that increase in solution pH implied in increase in UV absorbance which directly increased theSUVA,particularlyathighpHvalues.ThischangeinUVabsorbancewith pH incremental is attributed to ionization (dissociation) of hydrophobic NOM’s carboxylic, hydroxyl and phenolic functionalgroupswhichenhancetheNOMintra-molecularrepulsion[20,21],thussubsequentlyincreasingtheNOMmoleculessolubilityanditsdegree of hydrophobicity. However, the influence of pH on NOMsolubility may differ at low pH values. The fact that UV absorbancedecreasesasthepHdecreasesisthesuppositionthatatlowpHval-uestheionizablefunctionalgroupsofNOMareprotonated,reducedinelectrostaticrepulsioninbetweentheNOMintra-chains,therebydecreasingtheirsolubilityandfollowedbyformationofcoiledcom-plexes and more compacted NOM molecules configuration. Theseunchargedcomplexesarelesssolubleandtendtoagglomerate(dueto hydrophobicity interactions) before precipitated from solution.Subsequently this condition resulted in lesser NOM residual, espe-cially the hydrophobic NOM fraction, which is accompanied byreductionintheUVabsorption(shrinkageofthearomaticrings).Itis worth to note that results given in Figs. 3–5 are essentially use- ful in correlating the membrane performance (NOM rejection andfluxrate)withtheNOMfeedcharacteristics,asthecoagulation–UFsystem was operating at variety of pH and SUVA concentrations.The relationship between higher UV absorbance at greater pH andhigherNOMchargedensity(mequiv./gC)werecarefullyconsideredand would be used to elucidate the critical mechanisms involvedduringcoagulation–flocculation–ultrafiltrationprocessesaswellasproviding reasons for the coagulation–UF system performances. Fig. 5.  UV absorbance spectrophotometric characteristics of Yong Peng NOM as afunction of pH values.  196  A.W. Zularisam et al. / Separation and Purification Technology 65 (2009) 193–210 Fig. 6.  Schematic diagram of the coagulation–direct UF system.  2.2. Submerged coagulation–UF membrane Theexperimentalset-upisschematicallyillustratedinFig.6.The system consists of rapid mix coagulation, flocculation and directultrafiltration without the need of sedimentation process. The UFmembrane characteristics, SEM images of clean membrane andits zeta potential property are shown in Table 3, Figs. 7 and 8, respectively. Details of the UF membrane fabrication process andits properties determination procedure are given by Zularisam et  Table 3 UF membrane specification.Parameter PSF UF membraneMembrane type Hollow fiberMembrane material PolysulfoneExternal diameter (  m) 600(Based on spinneret opening)Internal diameter (  m) 300(Based on bore fluid needle)Contact angle ( ◦ ) 56Zeta potential (mV @ pH 7)  − 27MWCO 68kDaPure water flux;  J  pwf   at TMP of 250mmHg (Lm − 2 h − 1 ) 13.7Pure water permeability (Lm − 2 h − 1 bar − 1 ) 43 ± 5 al. [22]. Continuous feed supply controlled by a buoyant water level controller was used to ensure sufficient loading of coagu-latedNOMsourceintotheprocesstank.Bubblingsystemcontrolledby adjustable air flow regulator continuously supplied air bubbleswithin the fibers network at the bottom of membrane module toprovide a continuous up-flow circulation of micro-flocs suspen-sion for hindering any micro-particles settlement. In particulara constant air scouring bubble of 200L/(m 2 min) was applied toexert shear stress to suppress potential particles deposition onthe membrane surface. A commercial alum of Al 2 (SO 4 ) 3 · 18H 2 Owas employed throughout this experiment as it is widely usedin Malaysian water treatment plants. Final NOM source pH wasadjustedfrom5to10byadditionof0.1NNaOHand0.1NHClbefore Fig. 7.  SEM images of the cross section of clean asymmetric PSF membrane.   A.W. Zularisam et al. / Separation and Purification Technology 65 (2009) 193–210  197 Fig. 8.  Zeta potential curves of PSF membrane by streaming potential technique. coagulant mixing. The NOM source water spiked with the desiredcoagulant dose was hydraulically pumped from the influent tankand over-flowed into the process tank (Fig. 6). The flowrate of the coagulated NOM into the process tank was designed to range from2.67L/min to 8L/min, which was technically equivalent to 15minand5minofflocculationretentiontime(FRT),respectively.Floccu-lation of feed water was done in the process tank according to thedesigned FRT in order to maintain a rapid development of micro-flocs (pin-sized flocs). A PSF hollow fiber module with an effectivearea of 0.11m 2 was immersed in the process tank and a constantTMP (250mmHg) was maintained to extract the flocculated waterfromtheoutsidetoinsideofthemembranefibers.Thefluxdeclinewould be expected to increase in the course of time due to mem-brane fouling, thus operational permeate flux was monitored overthe time to determine the degree of membrane fouling to mem-brane permeability. Parameters used to quantify the efficiency of membrane processes are flux (  J  ), permeability and solute rejection( R ), where the flux is defined as  J   = Q  A  (1)where  Q   is the permeate flowrate (Lh − 1 ) and  A  is the membranearea (m 2 ) and permeability asPermeability = Q  AP   = Q NPdl  (2)where  Q   is the permeate flowrate (Lh − 1 ),  A  is the effective mem-brane area (m 2 ),   P   is the transmembrane pressure (Pa),  N   is thefiber quantity,  d  is the membrane OD and  l  is the membrane effec-tive length (m), the rejection ( R %) as R (%) =  1 −  C  p C  f   × 100 (3)where  C  p  is the permeate concentration in mg/L and  C  f   is the feedDOC concentration (mg/L) measured by DOC analyzer (ShidmadzuTOC-VE).With optimal coagulant dosing, pH adjustment, appropriateSUVA and proper FRT, the evolved suspended pin-sized flocs(instead of settleable pin-sized flocs) could be effectively rejectedby the 68kDa MWCO PSF membrane, resulting in higher NOMremoval with reasonable flux rate. This may also lead to loweroperating costs associated with lesser coagulant consumption andsludgeproduction,andshortenedoperationtimeassedimentationprocess was eliminated.  Table 4 Independent variables for first order model (screening design).Factors Unit Levels − 1 +1  A  (  x 1 ) Alum mgAl/L 1.8 3.0 B  (  x 2 ) FRT min 5 15 C   (  x 3 ) SUVA L/(mmg) 1.27 3.3 D  (>  x 4 ) pH pH 5 10  2.3. Analysis of data Full factorial design and central composite design were usedto quantitatively measure the significant effect of variable factorsand their interactions with each other on the membrane perfor-mance with respect to the NOM rejection efficiency and relativemembrane permeability (responses). In addition, analysis of vari-ance(ANOVA)wasusedtoanalyzedatacollectedintheperspectiveof the model adequacy corresponding to the significance of theregressionmodel,significanceofindividualmodelcoefficientsandmodel’s lack of fit possession. In fact experimental plan (at secondlevel design study) with the central composite design was used todevelopapredictionmathematicalcorrelationmodelbetweenthesignificantfactorswithrespecttotheoptimumNOMrejectionandmembrane permeability.  2.3.1. Full factorial design (first order model) DesignExpertSoftware(Stat-EaseInc.,Statisticmadeeasy,Min-neapolis, MN, USA, version 6.04) was used for the experimentaldesign throughout this screening process study. A total of sixteensets of experiments (2 4 full factorial design) and three replicatesat the center point were used to demonstrate the statistical signif-icance of the alum dosage (  A ; pretreatment), FRT ( B ; operationalcondition), SUVA ( C  ; NOM characteristic), and pH ( D ; solutionchemistry)onaffectingthepermeatequality.InthiscaseDOCrejec-tion was chosen to represent the quality characteristic performedbythemembraneasoptimalconfigurationofoperatingconditionswascapableofachievingasclosetoa100%rejection.Afluxratewasdecided to be inappropriate as the quality characteristic since highrejectionratesareusuallyaccompaniedbylowfluxrates.Therangeand levels of the variables investigated in this study are shown inTable4whileTable5showstheexperimentaldesignandpredicted values of the screening process. Range settings for variable factorswere adjusted based on previous findings and literature.  2.3.2. Response surface methodology (second order model) The factorial design was further continued with the responsesurface methodology (RSM) developed based on the centralcomposite design (CCD) with NOM removal and membrane per-meability as the dependent variables (responses) while significantterm from preliminary screening process were chosen as the inde-pendent factors. The CCD was conducted with a 2 3 full factorialcentral composite design of combinations factors at two levels(high, +1 and low,  − 1 levels), included with six star points (axial)corresponding to an  ˛  value of 2 and six replicates at the centerpoints (coded level 0, midpoint of high and low levels). In thisdesign, due to their relative insignificancies for NOM removal, theFRTdurationandaerationintensityweresetatitscenterpointset-ting(10min)and200L/(m 2 min),respectively.TherangeandlevelsoftheprocessingparametersinvolvedaretabulatedinTable6while the central composite design matrices and experimental responseof each individual experiment are shown in Table 7. In this study the independent variable for pH is represented by variable  A  whileSUVAandalumby B and C  ,respectively.Inthisdesign,theFRTfac-torwasconstantlysetatthecenterpointsettings(10min)asitwasfound to be insignificances in the organic removal (Table 8).
Related Search
Related Docs
View more...
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks