Contents lists available at SciVerse ScienceDirect
Desalination
j ourna l homepage: www.e lsev ie r .com/ locate /desa l
Preparation and characterization of low cost ceramic membranes formosambi juice clarification
Sriharsha Emani, Ramgopal Uppaluri ⁎, Mihir Kumar Purkait ⁎Department of Chemical Engineering, Indian Institute of Technology Guwahati, Guwahati, 781039, India
H I G H L I G H T S
• Kaolin based low cost ceramic membranes were prepared for juice clarification.• Membrane morphology varies with the particle size of inorganic precursors and fabrication pressure.• Enzyme treatment followed by centrifugation is recommended prior to microfiltration of mosambi juice.• Membrane fabricated at a pressure of 29MPa provided maximum flux of 20×10−6m3/m2·s using enzyme treated juice.
Article history:Received 18 November 2012Received in revised form 21 February 2013Accepted 22 February 2013Available online 22 March 2013
Keywords:MicrofiltrationJuiceFlux declineCeramic membraneKaolinMosambi juice
This work addresses the preparation and characterization of microfiltration membranes using low cost inor-ganic precursors for juice clarification. Low cost ceramic membranes have been fabricated using the uniaxialdry compaction method at various fabrication pressures (29, 39 and 49 MPa). The fabrication research hasbeen dovetailed to identify the most critical parameters (such as fabrication pressure and particle size ofthe precursors) to achieve the desired membrane morphologies. The prepared membranes possessed aver-age pore size (based on FESEM) varying from 1.85 to 0.89 μm, hydraulic pore size varying from 1.69 to0.72 μm, porosity ranging from 35.4 to 39.4%, and flexural strength ranging from 7.81 to 11 MPa. Amongstseveral membranes, the MF performance of M3membrane (fabricated at 49 MPa) for mosambi juice is highlysatisfactory. The M3 membrane provided a membrane flux of 90 to 44 × 10−6 m3/m2·s at 206.7 kPa with apermeate juice quality of negligible AIS content for the enzyme treated centrifuged juice (ETCJ). Flux declinemodelling analysis indicated that enzyme treatment followed by centrifugation minimizes irreversible foul-ing and is highly recommended before the MF.
Compared to polymeric membranes, ceramic membranes offer sev-eral beneficial options. These include good combinations of thermal,mechanical and chemical stability that enable them to possess longlife. Membrane technology research primarily aims at the developmentof functional materials to achieve low cost processing of industrial sep-aration schemes. Microfiltration of oil–water emulsions and fruit juicesis regarded to be an important avenue for the application of ceramicmembrane technology in industrial processing schemes.
Amongst various ongoing research efforts, the development andapplication of α-alumina/zirconia/titania membranes received signif-icant attention towards juice clarification. Jegatheesan et al. [1]reported the promising and optimal performance of zirconia ceramicmembranes (average pore size of 0.10 μm and 35% porosity) for the
ultrafiltration of sugarcane juice. Vaillant et al. [2] studied Membralox(alumina supported membrane of average pore size of 0.2 μm) for thecrossflow microfiltration of banana, pineapple and blackberry juices.Wang et al. [3] reported commercial CARBOSEP (zirconium oxideand titania with an average pore size of 0.14 μm) tubular ceramicmembrane for the crossflow microfiltration of West Indian Cherryjuice. On the other hand, the development and application of lowcost ceramic membranes for juice clarification is gradually gainingattention in the research community. Nandi et al. [4] reported aceramic composition (kaolin 40%, quartz 15%, calcium carbonate25%, sodium carbonate 10%, boric acid 5% and sodium metasilicate5%) that enabled them to prepare a membrane with an averagepore size of 1.54 μm and porosity of 42% using the paste method.The membrane provided optimal performance for the clarification ofmosambi juice. Primarily, the membranes have been specified aslow cost materials due to the usage of inexpensive inorganic precur-sors and low sintering temperature (850–1000 °C). Vasanth et al.[5] reported a ceramic precursor composition (kaolin 40%, quartz15%, calcium carbonate 25%, sodium carbonate 10% boric acid, sodium
metasilicate 5% and 2 wt.% polyvinyl alcohol) to achieve low costceramic membranes using uniaxial dry compaction method. Theauthors reported that the membrane with an average pore size of3.45 μm, porosity of 30% and mechanical strength of 28 MPa is suit-able for various separation applications.
Various other microfiltration range ceramic membranes havebeen studied towards the clarification of watermelon [6], hamimelon [7], orange [8], mosambi [9], apple [10] and pineapple [11].Gomes et al. [6] reported the MF data using a ceramic membranepossessing an average pore size of 0.1 μm. Zhang et al. [7] concludedthat the ceramic membrane with an average pore size of 0.2 μm andan operating pressure of 2 bar is optimal for the clarification ofhami melon juice. Nandi et al. [8,9] reported kaolin based low costceramic membrane with an average pore size of 0.285 μm for theMF of orange and mosambi juice. Fukumoto et al. [10] reported0.2 μm range MF ceramic membranes for apple juice clarification. deCarvalho et al. [11] carried out pineapple juice clarification using0.22 μm ceramic membranes.
Based on the ongoing research trends, it can be observed thatthe development of low cost ceramic membranes for juice clarifica-tion is an attractive option towards the affirmed commercializationof ceramic membrane technology. In the contemporary scenario,other than Nandi et al. [4], there has not been much effort withrespect to the development of low cost ceramic membrane formula-tions that can be experimented for juice clarification. Further, it canbe also noted that while Nandi et al. [4] adopted paste method, theuniaxial dry compaction method adopted by Vasanth et al. [5] ismore promising due to its simplicity and ability to achieve defectfree membranes. So far, there have not been efforts that addressedthe fabrication of low cost ceramic membranes using uniaxial methodfor juice clarification. In this regard, it can be also analyzed that themembrane morphological parameters (average pore size of 3.45 μm,porosity of 30%) reported by Vasanth et al. [5] are significantly highand hence may not be useful for juice clarification.
Considering the above research limitations and avenues, this workattempts to extend the research towards the identification of optimalfabrication parameters using uni-axial dry compaction method. Thiswork adopted the compositions reported by Vasanth et al. [5] alongwith few procedural modifications to screen size and fabrication pres-sures to achieve membrane morphologies desired for mosambi juiceclarification. The aim of this work is to develop low cost ceramicmembranes with appropriate pore size and porosity to obtain clari-fied mosambi juice with good combinations of flux and juice quality.In other words, the clarified juice shall possess minimal pectin con-tent (expressed in alcohol insoluble solids), higher degree of clarityand citric acid and minimal reduction in °Brix. Therefore, themicrofiltration operation shall retain sugars and flavoured compo-nents and shall eliminate haze and colloidal substances that contrib-ute to the deterioration of the juice during long term storage. Thenext section elaborates upon the experimental investigations carriedout to realize the membrane morphologies and juice microfiltration.
2. Experimental
2.1. Ceramic membrane fabrication
In addition to polyvinyl alcohol (PVA, M.wt. 1,15,000, LobaChemicals Ltd., India), six different ceramic precursors namely kaolin(CDH, India), quartz (Research Lab Fine Chem Industry, India), calciumcarbonate (Merck, India), sodium carbonate (Merck, India), boric acid(Merck, India) and sodium metasilicate (CDH, India), without anyfurther purification have been used to fabricate the low cost porousceramic disk shaped membranes. The membrane fabrication procedurefollowed the methodology outlined by Vasanth et al. [5] and consistsof steps similar to their work that involved a sequential procedure ofraw-material preparation, preparation of membrane mould at various
fabrication pressures, drying, sintering, and polishing with siliconcarbide abrasive paper of the prepared membranes followed with son-ication using Millipore water to remove loose particles to obtain thefinal membranes. Two important variations of this work from the pro-cedures outlined by the authors [5] are presented as follows. Firstly,the raw-materials have been screened through a 36 mesh standardscreen instead of using 30 mesh standard screen [5]. Secondly, thehydraulic press (Make: Velan Engineering, Tamilnadu, India), fabrica-tion pressures for various membranes are 29 (M1), 39 (M2) and 49(M3) MPa instead of using 50 MPa. Hence, the fabricated membranesM1, M2 and M3 correspond to the membranes fabricated at 29, 39and 49 MPa fabrication pressures respectively.
2.2. Membrane characterization
Membrane characterization was carried out using thermogravimetricanalyzer (TGA, Make: NETZ5CH,Model No. STA 449F3, Jupiter, Switzer-land), X-ray diffraction (XRD, Make: BRUKER, Model No. D8 Advance,Germany), field emission scanning electron microscopy (FESEM,Make: ZEISS,Model No.∑ IGMA, USA), porositymeasurement, hydrau-lic permeability measurement and corrosion resistance to acidic andbasic media. Procedures adopted for membrane characterization havebeen presented elsewhere [12].
2.3. Microfiltration studies using mosambi juice
Fresh mosambi juice (FJ) was obtained using depulped mosambifruits (sweet orange, Citrus sinensis (L) Osbeck) in a manually operatedscrew type juice extractor. Centrifuged juice (CJ) was obtained usinga centrifuge (Make: SORVALL, RC 5C PLUS) and enzyme treatedcentrifuged juice (ETCJ) was prepared by first carrying out enzymaticpretreatment of the juice followed with centrifugation at 4000 rpmfor 20 min. The enzymatic pretreatment involved heating the juiceat 42 °C for 100 min with a pectinase enzyme concentration of0.0004 w/v% [13] followed by heating at 90 °C for 5 min to inactivatethe remaining enzyme. Following this step, the juice was cooled toambient temperature (25 °C) to undergo centrifugation. For all mem-branes (M1, M2 and M3), dead end MF was carried out using CJ andETCJ at pressure differentials (ΔP) of 68.9, 137.8 and 206.70 kPa. Theeffective membrane area for permeation is 1.45 × 10−3 m2. Furtherdetails with respect to the microfiltration experimental methodologyare presented elsewhere [9].
2.4. Analytical methods
Various analytical techniques are used tomeasure relevant param-eters in FJ, CJ, ETCJ and permeate juice samples. The analytical proce-dures involved the measurement of particle size distribution usinglaser particle size analyzer (Make: Malvern; Mastersizer 2000, UK);absorbance (at 420 nm) and transmittance (at 660 nm) using UV–vis spectrophotometer (Make: Perkin Elmer Precisel; Model: Lambda35, USA); total soluble solids (TSS) using digital refractrometer(Make: Atago; Model: DR-A1, India); pH using water and soil analysiskit (Make: VSI Electronics, Model: VSI-06D1, India); viscosity using arheometer (Make: Thermo Electron Corporation; Model: HAAKERheostress 1) and density using pycnometer. Other analytical mea-surements such as acidity (in terms of equivalent citric acid wt.%)and alcohol insoluble solids (AIS) were measured using proceduressummarized in the literature [9].
2.5. Fouling studies
After each MF experiment, membrane cleaning was conducted toregain the membrane permeability. It was observed that after eachMF run, the membrane possessed a thick gel layer on its surface,which was removed by treating the membrane with commercial
060
70
80
90
100
110
120
Carbonate decomposition
Moisture removal
Wei
gh
t lo
ss (
%)
Temperature (oC)
710oC
830oC
120oC
(a)
10
I
Inte
nsi
ty (
a.u
.)
2θ angle (degrees)
Unsintered powder
K
Q
Q
Q
I
Q- quartzI - InyoiteK - KaolinN - Nephiline
QK
K I I Q K K Q
Sintered powder at 900OC
N
Q
N QN
I
Q N
(b)
12001000800600400200
80706050403020
Fig. 1. (a) Thermogravimetric analysis of the raw material mixture. (b) X-ray diffrac-tion patterns of un-sintered and sintered powder at 900 °C.
34 S. Emani et al. / Desalination 317 (2013) 32–40
detergent solution (surf excel with a concentration of 2 g/L) for10 min in a sonication bath. Following this, the membrane was thor-oughly cleaned with Millipore water and hydraulic permeability wasonce again measured using the MF setup. The subsequent reductionin pure water flux (or permeability) was eventually expressed interms of the fouling index defined as:
FI ¼ PWi−PWf
PWi� 100 ð1Þ
where PWi and PWf correspond to the pure water hydraulicpermeability values for the fresh and the cleanedmembrane, respectively.
3. Theory of membrane fouling mechanism
To analyze the existent fouling mechanism during dead end MF atconstant pressure, Hermia [14] proposed four different models,namely; cake filtration, intermediate, standard and complete poreblocking models. During cake filtration, particles larger than thepore size of the membrane retain and enable the formation of acake over the membrane surface. Hence the cake layer offers addi-tional resistance to the liquid permeation. Also, as time proceedsthe cake layer grows with time and is intrinsically related to flux de-cline. Complete, standard and intermediate pore blocking occur,when the particles enter the membrane pores, which is possible ifthe particle size is less than the membrane pore size. During completepore blocking, the particles enter and block the pores, thus enablingthe drastic reduction in the size of the open pores through whichthe liquid passes. During standard pore blocking, the permeating par-ticles adhere on the membrane pore walls. Intermediate poreblocking is a case that corresponds to be between standard poreblocking and cake filtration where the particles may adhere to thesurface or to the pores or block the pores. Further details with respectto the derivations of flux decline models for all four cases have beenpresented by Hermia [14] and are not discussed here. The linearizedflux decline models for cake filtration, complete, standard and inter-mediate pore blocking phenomena are expressed as
(a) Complete pore blocking:
ln J−1� �
¼ ln Jo−1
� �þ kbt ð2Þ
(b) Standard pore blocking:
J−0:5 ¼ Jo−0:5 þ kst ð3Þ
(c) Intermediate pore blocking:
J−1 ¼ Jo−1 þ kit ð4Þ
(d) Cake filtration:
J−2 ¼ Jo−2 þ kct: ð5Þ
Thus, the fitness of the flux decline data with any one of the above4 models can be obtained by analyzing the coefficient of correlation(R2) values corresponding to the experimental data. The identifica-tion of most competent flux decline model will enable to examinethe extent of reversible and irreversible fouling. For instance, com-plete pore blocking may correspond to maximum irreversible foulingand cake filtration may correspond to maximum reversible foulingand hence the onset of cake filtration during juice MF is highly appre-ciable to minimize fouling effect and easy restoration of flux aftercleaning operation.
4. Results and discussion
4.1. Membrane characterization
Fig. 1(a) presents the variation of weight loss (%) as a function oftemperature for the raw material mixture. It can be observed thatwhile sufficient weight losses existed below 800 °C, very significantweight loss existed from 900 to 1000 °C. Hence, the maximumsintering temperature of 900 °C is acceptable for the fabrication ofthe membranes. Also, it can be observed that weight decline up to120 °C was due to the loss of moisture. Further, weight reduction oc-curred at 710 °C, where decarbonization of carbonate salts occurs.Fig. 1(b) presents the XRD patterns for unsintered raw material mix-ture and M3 membrane sintered at 900 °C. It can be observed fromthe figure that kaolin related peaks disappeared in the membrane
35S. Emani et al. / Desalination 317 (2013) 32–40
sample sintered at 900 °C, thus indicating the conversion of kaoliniteand metakaolinite at the sintering temperature. The XRD patternsalso indicate the existence of several nepheline and quartz peaksalong with the inyoite phase for the sintered membrane. Fig. 2 pre-sents the FESEM images for all membranes and indicates highly po-rous and rough surface morphologies. The images also indicate the
(a)
(b)
(C)
Fig. 2. FESEM images of (a) M1, (b) M2 and (c) M3 membranes.
absence of pin holes and cracks with maximum observable pore sizeof 2–4 μm. Fig. 3(a) illustrates the variation of pure water flux withapplied pressure for all the membranes. The pure water flux variedfrom 4.63 to 18.64 × 10−4, 5.33 to 21.47 × 10−4 and 1.31 to5.19 × 10−4 m3/m2·s with a variation in applied pressures from68.9, 137.8, 206.7 to 276 kPa for M1, M2 and M3 membranes respec-tively. The variation in the average porosity of the membranes (deter-mined using Archimedes principle) with fabrication pressure isshown in Fig. 3(b). With an increase in fabrication pressure from 29to 49 MPa, the membrane porosity reduced from 39.4 to 35.4%. Thereduction in the porosity with fabrication pressure is due to thefiner arrangement of the raw materials in the porous texture atincreased fabrication pressure. Using Hagen–Poiseuille equation,average porosity and hydraulic permeability, the average pore sizeof the membrane was evaluated to vary from 1.69 to 0.72 μm forthe membranes. Vasanth et al. [5] reported the average pore size of2.60 μm and porosity of 40% at a fabrication pressure of 50 MPa.The variation in membrane morphology in this work is possibly dueto the reduction in particle size of the inorganic precursors.
Subsequent analysis for the pore size distribution by using Image Jsoftware (Version 1.40) was performed and the evaluated pore sizedistributions indicated that less than 10% of the pores had diameters
00
5
10
15
20
25Membrane type:
M1M2M3
Pu
re w
ater
flu
x x
104
(m3 /
m2
S)
Applied pressure (kPa)
(a)
2028
30
32
34
36
38
40
42
PorosityAverage pore size
Fabrication membrane pressure (MPa)
Po
rosi
ty (
%)
0.6
0.8
1.0
1.2
1.4
1.6
1.8
Averag
e po
re size (μm)
(b)
30025020015010050
60504030
Fig. 3. (a) Pure water flux vs applied pressure for membranes M1–M3. (b) Variation ofaverage porosity and pore size with membrane fabrication pressure.
0.01
0
1
2
3
4
5
6
7
8
9
10
Vo
lum
e (%
)
Particle size (µm)
Juice type :FJCJETCJ
1000010001001010.1
Fig. 4. Particle size distributions in various FJ, CJ and ETCJ.
36 S. Emani et al. / Desalination 317 (2013) 32–40
within the range of 3–15 μm for all the membranes (M1, M2 and M3)and distinct pore size distributions existed for the membranes withinthe range of 0–3 μm. For M1, M2 and M3 membranes, up to 45%, 35%and 25% pores have been analyzed in the range of 0–2, 0–1.75 and 0–1 μm respectively. Based on the FESEM image analysis using Image Jsoftware, the average pore size of the membranes is 1.85, 1.78 and0.89 μm for membranes M1, M2 and M3 respectively, which is inagreement with the pore size determined from hydraulic permeabil-ity and porosity studies.
The flexural strength values of the membranes M1–M3 variedfrom 7.81 to 11 MPa, which is lower than the values reported byVasanth et al. [5] (22 to 31 MPa) and higher than the values reportedby Nandi et al. [12] (3 to 8 MPa). This indicates that the fabricatedmembranes have acceptable values of flexural strength and furtherresearch is expected to enhance the flexural strength values.
4.2. Microfiltration of mosambi juice
The particle size distributions of FJ, CJ and ETCJ have beenpresented in Fig. 4. It can be observed that FJ had unimodal particlesize distribution profile ranging from 0.1 to 2000 μm. The averageparticle size of FJ corresponded to a value of 526 μm, which is higherthan that reported by Nandi et al. [4]. The particle size distributionprofiles for CJ and ETCJ are also shown in the figure. As shown, forboth CJ and ETCJ, the particle size distributions followed a bimodaldistribution ranging from 0.1 to 600 μm. It can be also observed thatlarger colloidal particles are successfully eliminated by pretreatment.The average particle size of CJ and ETCJ have been determined to beabout 34 and 22 μm, respectively, which is significantly higher thanthose reported by Nandi et al. [4]. Therefore, the chosen feed juicein this work is regarded to be characterized with higher concentra-tion of larger particles, and thus pre-treatment appears to be neces-sary prior to the microfiltration tests. Further, it has also beenobserved that the maximum particle size of the CJ and ETCJ has
Table 1Physico-chemical properties of FJ, CJ and ETCJ samples.
been analyzed to be 677 and 590 μm as opposed to the original max-imum particle size range of about 2000 μm. Thus, pretreatment oper-ation can be concluded to be very important to alter the particle sizedistribution characteristics of the juice samples.
Table 1 presents the physicochemical properties of FJ, CJ and ETCJ.It can be observed that pretreatment enabled significance reductionin colour, viscosity, AIS and enhancement in the clarity of the juice.The TSS, pH and citric acid content of the juice remained fairly con-stant and the juice density reduced slightly after pretreatment. There-fore, it is observed that pretreatment enabled the removal of pecticand colloidal material from the juice.
The CJ flux decline profiles for M1, M2 and M3 membranes havebeen presented in Fig. 5(a). It can be seen that the flux decline profilefor M1 is located above M2 and M2 above M3. This is due to theincrease inmembrane pore sizewith fabrication pressure. ForM1mem-brane, the flux reduced from 80 to 20 × 10−6 m3/m2·s with in a timeperiod of 30 min at 206.7 kPa. For M3 membrane, the flux reducedfrom 55 to 9 × 10−6 m3/m2·s with in a time period of 30 min at206.7 kPa. Corresponding flux decline profile reported by Nandi et al.[4] varied from 68 to 15 × 10−6 m3/m2·s at 165.5 kPa. Therefore, it isconcluded that the bestmembrane performance in terms offlux declineis comparable to the literature values reported with paste method.
The ETCJ flux decline profiles for M1, M2 and M3 at 206.7 kPa arepresented in Fig. 5(b). Again, the membrane pore size contributedsignificantly to the flux decline performance. For M1 membrane, theflux varied from 120 to 70 × 10−6 m3/m2·s for a time period of30 min at 206.7 kPa. For M3 membrane, the flux varied from 90 to44 × 10−6 m3/m2·s for a time period of 30 min at 206.7 kPa. Fluxvalues in the literature (Nandi et al. [4]) correspond to 137.5 to50 × 10−6 m3/m2·s at 165.5 kPa. Therefore, the performance of themembranes is comparable with those presented in the literature.
Fig. 5(c) and (d) presents the variation of trans membrane fluxwith time for the M3 membrane at different pressures. It can beobserved that for CJ the membrane flux varied from 41 to3 × 10−6 m3/m2·s at 68.9 kPa which increased to 55 to 9 × 10−6
m3/m2·s at 206.7 kPa. This is due to increase in the driving forceacross the membrane. On the other hand, for ETCJ the membraneflux varied from 66 to 17 × 10−6 m3/m2·s at 68.9 kPa which in-creased to 90 to 44 × 10−6 m3/m2·s at 206.7 kPa. It can be ob-served that the flux decline profiles are more distinctly placed forthe ETCJ when compared to CJ and hence enzyme treatment is ahighly promising option to increase the shelf life of the membrane.
Table 2 summarizes the physicochemical properties of clarifiedmosambi juice obtained from the MF using M1, M2 and M3 mem-brane. It can be observed that several juice properties remained sim-ilar for all membranes. The physicochemical properties of clarifiedmosambi juice after MF at 206.70 kPa for all membranes have beenevaluated as follows: colour index of 0.12, clarity index from 0.01 to0.03, °Brix of 9.1, citric acid wt.% from 0.78 to 0.80, pH of 3.93–3.95,density of 1.24 to 1.25 g/cm3, viscosity of 1.255 to 1.265 MPa·s andAIS of 0.04–0.05 wt.% for CJ samples. For the ETCJ case, distinct proper-ties were observed for colour index (0.01 to 0.03), density (1.17 g/cm3)and negligible AIS content. Therefore, enzyme treatment followed withcentrifugation enabled to achieve good quality feed for the micro-filtration operation. The observed trends for all parameters are ingood agreement with those reported by Nandi et al. [4].
Fig. 5. Variation of permeate flux with time for (a) membranes M1, M2 andM3 at ΔP = 206.70 for CJ, (b) membranes M1, M2 andM3 at ΔP = 206.70 for ETCJ, (c) membrane M3 atpressures ΔP = 68.9, 137.8 and 206.70 kPa for CJ and (d) membrane M3 at pressures ΔP = 68.9, 137.8 and 206.70 kPa for ETCJ.
37S. Emani et al. / Desalination 317 (2013) 32–40
The effect of increasing transmembrane pressure on the juice qualityhas also been studied for the M3 membrane. The physical–chemicalproperties of clarified mosambi juice after MF for M3 membrane atdifferent pressures (68.9–206.7 kPa) have been evaluated as follows:colour index of 0.11 to 0.12, clarity index ranging from 97.44 to 98.01,°Brix of 9.1, citric acid wt.% from 0.77 to 0.78, pH of 3.93–3.95, densityof 1.24 g/cm3, viscosity of 1.255–1.265 MPa·s and AIS of 0.03–0.04 wt.% for CJ samples. For the ETCJ case, distinct properties wereobserved for colour index (0.02 to 0.03), density (1.17 g/cm3) and AIS(negligible). Thus, it can be observed that important parameters suchas density, AIS, TSS, and acidity did not vary much with variation in thepressure and confirmed that the membrane performance is satisfactoryfor the desired juice quality and is fairly independent of the appliedpressure.
4.3. Fitness of fouling models
The fitness of experimental data for all the membranes to repre-sent various pore blocking models is presented in Fig. 6(a) and (b)for CJ and ETCJ, respectively. It can be observed that for both CJ andETCJ, none of the models could very closely fit to the experimentaldata except for the M1 and CJ combination. A paradigm shift
involving the onset of cake filtration after a preliminary pore blockingmechanism can be typically observed from the data presented inFig. 6(b) for ETCJ. Corresponding coefficient of correlation (R2) valuesalong with slope and intercept values presented in Table 3 for CJ andETCJ, respectively. From Table 3, it is apparent that the R2 value is notvery close to 1 for any case. Consequently, it is concluded that themembrane morphology and pore size distribution favours initiallypore blocking followed by cake filtration for CJ and ETCJ.
The fitness ofM3membraneflux data (obtained at various pressures)towards cake filtration model is presented in Fig. 6(c) and (d) for CJ andETCJ respectively. It can be observed that for ETCJ, a very good fitness ofthe cake filtration model exists. However, for the CJ case, the flux dataobtained at 68.9 kPa did not fit well with the experimental data. On theother hand, cake filtration model fit well for the flux data obtained at137.8 kPa and 206.7 kPa. Incidentally, the flux data at 68.9 kPa fit wellwith the intermediate pore blocking model. Therefore, it is apparentthat operating pressure plays an important role in defining the pertinentflux decline mechanism and lower pressures tend to maximize poreblocking and hence fouling. The same is also confirmed with the foulingindex values that are also presented in this work.
The cake filtration constant (kc in Eq. (5)) is analyzed to verylinearly with fabrication pressure. The cake filtration constant varied
Table 2Physicochemical properties of clarified mosambi juice obtained after MF at 206.70 kPa using various membranes.
38 S. Emani et al. / Desalination 317 (2013) 32–40
from 10–50 × 10−7 for CJ and 0.6–1.0 × 10−7 for ETCJ, respectively.Thus it can be observed that a drastic reduction in cake filtrationparameter values can be achieved by enzyme treatment followedwith centrifugation. The cake filtration constant is analyzed to varylinearly with fabrication pressure and is suitably modelled to thefollowing expression.
kCJC ¼ 2:00� 107ΔPfab−5� 108; R2 ¼ 0:999 ð8Þ
kETCJC ¼ 20000ΔPfab þ 20000; R2 ¼ 1 ð9Þ
The cake filtration constant varied from 10 to 50 × 10−7 for CJ and0.6–1.0 × 10−7 for ETCJ, respectively. Thus it can be concluded that a
00
20
40
60
80
100
120
140Juice type : CJPressure : 206.70kPaMembrane type :
M1M2M3
J-1 x
10-3
, m2 .
s/m
3J-2
x 1
0-8, m
4 .s2
/m6
Time (min)
(a)
0
250
500
750
1000
1250
1500
1750Juice type : CJPressure : kPaMembrane type : M3
68.90137.80206.70
Time (min)
(c)
0 3530252015105
30252015105
Fig. 6. Linearized fitness plots of various membranes for the MF flux decline data obtained206.70 kPa, (b) intermediate pore blocking model for M1–M3 and ETCJ at ΔP = 206.70(d) cake filtration model for M3 for ETCJ at ΔP = 68.7, 137.8 and 206.7 kPa.
drastic reduction in cake filtration constant can be achieved byenzyme treatment followed with centrifugation.
Since the MF flux decline profiles for both CJ and ETCJ could not berepresented using any one of the chosen models, combination modelshave been analyzed to represent the flux decline data. To do so, twodistinct sets of flux decline data have been subjected to the modellinganalysis. These correspond to the initial flux decline profiles up to atime period of 10 min and the flux decline profiles from 11 to30 min of the microfiltration runs. Table 4 summarizes the coefficientof correlation (R2) for the combination models for both CJ and ETCJ. Itcan be noted that cake filtration with different combinations ofparameters (initial flux (J0) and cake filtration constant (kc)) in bothregimes suited well with the experimental data obtained using ETCJfor all the membranes. In other words, the flux data indicated thedynamic variations of cake layer morphology with time. For CJ using
J-2 x
10-8
, m4 .
s2 /m
6
00
1
2
3
4
5
6
7
8Juice type : ETCJPressure : 206.70 kPaMembrne type :
M1 M2M3
Time (min)
(b)
(d)
3530252015105
for various cases: (a) intermediate pore blocking model for M1–M3 and CJ at ΔP =kPa, (c) cake filtration model for M3 for CJ at ΔP = 68.7, 137.8 and 206.7 kPa, and
Table 3Summary of parameters associated to various pore blocking models for the M3 membrane.
Juice type Pressure (kPa) Complete pore blocking Standard pore blocking Intermediate pore blocking Cake filtration
M1 membrane, cake filtration followed with intermediate poreblocking fit well with the data. For M2 and M3 membranes, interme-diate and complete pore blocking represent the experimental datawith good fitness. Thus, it can be observed that pore blocking modelshave been representing the CJ flux decline during the time period of11–30 min of MF runs. This indicates the significance of irreversiblefouling in the early stages of MF which is also confirmed by the foul-ing studies. Hence, MF of CJ can be inferred to be not suitable for theprepared membranes.
The average error of MF flux data obtained for all membranes forthe combination models has been analyzed to be no more than 5%with respect to the experimental data. Further, it has been alsoobserved that several predicted data sets deviated significantly dur-ing the initial stages of microfiltration runs for all membranes, thusindicating that the initial pore blocking mechanism is highly complexand could not be fit with simplistic models. Nonetheless, the overallfitness of the two combination model with the experimental fluxdata is satisfactory.
4.4. Fouling index
The fouling index for the M3 membrane has been evaluated tovary from 41.6 to 33.5% for a variation in pressure from 68.9 to206.7 kPa. On the other hand, for the ETCJ case, the correspondingvalues varied from 1.51 to 1.04%. This indicates that enzyme treat-ment prior to centrifugation is highly effective towards long termusage of the ceramic membranes for juice MF. Also, for the mem-branes whose MF runs carried out at 206.7 kPa, the FI values variedfrom 43.9 (M1) to 33.5% (M3) for the CJ case and 1.73 (M1) to1.04% (M3) for the ETCJ case. This indicates that the M3 membraneis the most suitable for juice MF. This is possibly due to the earlyonset of pore blocking which is expected to contribute significantlytowards fouling for membranes with higher pore size and porosities(M1 and M2).
4.5. Cost estimation
Based on retail cost of various inorganic precursors, the averagecost of the membranes has been estimated. Considering the unit
Table 4A Summary of correlation coefficient parameters (R2) associated with combinationpore blocking models for all membranes at ΔP = 206.70 kPa.
price of kaolin, quartz, calcium carbonate, sodium carbonate, boricacid, sodium meta silicate and polyvinyl alcohol as 520, 358, 408660, 760, 578 and 13,500 Rs./kg respectively, the total cost of themembrane has been estimated as Rs. 10.15/- per one circular discshaped membrane. This corresponds to the membrane cost of 78($/m2). Comparatively, the membrane cost reported by Vasanthet al. [5] is 67 ($/m2) and that reported by Nandi et al. [4] is 130($/m2). In other words, the cost of the membrane is comparablewith the low cost membranes reported in the literature.
5. Conclusions
For the first time, this work reported the fabrication of low costceramic membranes for juice clarification using uniaxial dry compac-tion method. It was observed that both mesh screen size (to obtainprecursors with finer pore size) and fabrication pressure significantlyinfluenced the membrane morphological parameters and enabled theachievement of membranes that can be deployed for mosambi juiceclarifications. Several membrane characterization parameters suchas hydraulic permeability and average pore diameter, porosity, chem-ical stability, FESEM, XRD, TGA and flexural strength are in agreementwith those reported by Nandi et al. [4].Amongst all membranes, M1membrane provided maximum flux where as membrane M3 provid-ed minimum fouling index for CJ and ETCJ. However, membraneM3 isrecommended for further process engineering studies due to itshigher mechanical strength, lower fouling and acceptable combina-tions of membrane flux and juice quality. The effect of pressure onthe microfiltration studies for M3 membrane indicated that whileflux was highly dependent on pressure, the juice quality was fairly in-dependent of the applied pressure. Amongst CJ and ETCJ, ETCJ provid-ed minimal fouling and clarified juice with negligible AIS content forall membranes. Cake filtration has been identified to be the mostcompetent fouling mechanism for membrane M3 operated at206.7 kPa. Future research is expected to further reduce the averagemembrane pore size up to 0.1–0.2 μm. Further, dead end and crossflowmicrofiltration for other types of juices (apple, pineapple, water-melon and sugarcane juices) will be also taken up to identify themicrofiltration performance of the developed low cost membranesusing uniaxial dry compaction method.
Nomenclatureε porosity of the membrane (dimension less)ΔPfab fabrication membrane pressure (MPa)W1 dry weight of the membrane (g)W2 wet weight of the membrane (g)J permeate flux (m3/m2 s)Q volume of permeate (m3)S permeable area of membrane (m2)FI fouling index (dimension less)PWi pure water hydraulic permeability values for the cleaned
membranePWf pure water hydraulic permeability values for the fresh
membraneJo initial permeate flux (m3/m2 s)
40 S. Emani et al. / Desalination 317 (2013) 32–40
kb complete pore blocking model constant (s−1)kc cake filtration model constant (s·m−2)kcCJ cake filtrationmodel constant for centrifuged juice (s·m−2)
kcETCJ cake filtration model constant for enzyme treated centri-
fuged juice (s·m−2)ki intermediate pore blocking model constant (m−1)ks standard pore blocking model constant (m−0.5 s−0.5)R2 square of correlation coefficient (dimensionless)
Acknowledgement
The authors sincerely acknowledge the Department of Biotechnol-ogy (DBT), New Delhi for providing the financial assistance towardscarrying out the research reported in this article. Any opinions, find-ings or conclusions expressed in this article correspond to the authorsviews and do not necessarily reflect the views of the DBT.
References
[1] V. Jegatheesan, D.D. Phong, L. Shu, R. Ben, Performance of ceramic micro- andultrafiltration membranes treating limed and partially clarified sugar cane juice,J. Membr. Sci. 327 (2009) 69–77.
[2] F. Vaillant, A.M. Perez, O. Acosta, M. Dornier, Turbidity of pulpy fruit juice: a keyfactor for predicting cross-flow microfiltration performance, J. Membr. Sci. 325(2008) 404–412.
[3] B.J. Wang, T.C. Wei, Z.R. Yu, Effect of operating temperature on component distri-bution of West Indian cherry juice in a microfiltration system, LWT- Food Sci.Technol. 38 (2005) 683–689.
[4] B.K. Nandi, R. Uppaluri, M.K. Purkait, Identification of optimal membrane mor-phological parameters during microfiltration of mosambi juice using low costceramic membranes, LWT- Food Sci. Technol. 44 (2011) 214–223.
[5] D. Vasanth, R. Uppaluri, G. Pugazhenthi, Influence of sintering temperature on theproperties of porous ceramic support prepared by uniaxial dry compaction methodusing low cost rawmaterials formembrane applications, Sep. Sci. Technol. 46 (2011)1241–1249.
[6] F.S. Gomes, P.A. Costa, M.B.D. Campos, R.V. Tonon, S. Couri, L.M.C. Cabral, Water-melon juice pretreatment with microfiltration process for obtaining lycopene, Int.J. Food Sci. Technol. 48 (2013) 601–608.
[7] J. Zhang, A. Rahman, M. Atergul, Application of microfiltration sterilizing ofceramic membrane in hami melon juice, Shipin Gongye Keji 33 (2012) 243–245.
[8] B.K. Nandi, B. Das, R. Uppaluri, Clarification of orange juice using ceramicmembraneand evaluation of fouling mechanism, J. Food Process Eng. 35 (2012) 403–423.
[9] B.K. Nandi, B. Das, R. Uppaluri, M.K. Purkait, Microfiltration of mosambi juiceusing low cost ceramic membrane, J. Food Eng. 95 (2009) 597–605.
[10] L.R. Fukumoto, P. Delaquis, B. Girard, Microfiltration and ultrafiltration ceramicmembranes for apple juice clarification, J. Food Sci. 63 (1998) 845–850.
[11] L.M.J. de carvalho, C.A. Bento da silva, A.P.T.R. Pierucci, Clarification of pineapplejuice (Ananas comosus, L. Merryl) by ultrafiltration and microfiltration: physico-chemical evaluation of clarified juices, soft drink formulation and sensorial eval-uation, J. Agric. Food Chem. 46 (1998) 2185–2189.
[12] B.K. Nandi, R. Uppaluri, M.K. Purkait, Preparation and characterization of low costceramic membranes for micro-filtration applications, Appl. Clay Sci. 42 (2008)102–110.
[13] P. Rai, G.C. Majumder, G. Sharma, S. DasGupta, S. De, Effect of various cut-offmembranes on permeate flux and quality during filtration of mosambi (Citrussinensis (l.) Osbeck) juice, Food Bioprod. Process. 84 (2006) 213–219.
[14] J. Hermia, Constant pressure blocking filtration laws — application to power-lawnon-Newtonian fluids, Trans. Inst. Chem. Eng. 60 (1982) 183–187.
