journal of MEMBRANE SCIENCE

ELSEVIER
Journal of Membrane Science 90 ( 1994 ) 131- 139

Application of ceramic membrane as a pretreatment in anaerobic digestion of alcohol-distillery wastes*

(Received November 27, 1992; accepted in revised form December 23, 1993)

Abstract
Pretreatment of alcohol-distillery wastes with ceramic membranes was performed prior to anaerobic digestion. Ceramic membranes with 0.05 pm pore size were chosen based on the particle size distribution in raw wastes. In this pretreatment, chemical oxygen demand (COD) was reduced from 36,000 to 18,000 mg/l and suspended solids were almost completely removed. The permeates from the ceramic membrane were further separated by ultrafiltration, but further COD reduction by using the PM30 and PM10 membranes was not achieved. Mixed stillages exhibited higher fouling tendency than pure naked barley stillage. Several cleaning methods were attempted to recover water flux. Although lumen flushing was effective, hydrogen peroxide proved to be the most effective cleaning agent. The negative flux recovery after nitric acid cleaning could be explained by the ligand exchange theory. The performance of digester was greatly improved with membrane pretreatment, specially in the case of naked barley based stillage.
Key words: Pretreatment; Ceramic membrane cleaning; Fouling tendency; L&and exchange model; Anaerobic digestion

1. Introduction Anaerobic digestion is a microbiological process that converts complex organics into methane gas. It is a proven technology for the stabilization of domestically derived organic sludges and offers tremendous potential as an industrial wastewater treatment process [ l-41. So far it has been used successfully in treating alcohol-distillery wastes (stillages) in Korea. However, re*Paper presented at the Int. Membrane Science and Technol-

ogy Conference, November lo- 12,1992, Sydney, Australia.

placement of naked barley instead of tapioca or sweet potato as the raw material for alcohol fermentation reduced both chemical oxygen demand (COD) removal efficiency and methane gas production of anaerobic digestion [ 5,6]. These phenomena seem to be attributed to the presence of toxins and inhibitors in stillages based on naked barley. To overcome these partitular problems dilution or chemical precipitation was attempted prior to anaerobic digestion, but negative effects such as volume increase of wastes or consumption of chemicals were too setious to justify these attempts [ 7,8]. Therefore, in this study, pretreatment of wastes

0 1994 Elsevier Science B.V. All rights reserved

SSDI 0376-7388 (94)00013-O

132

I.-S. Changet al. /Journal ofMembrane Science 90 (1994) 131-139

by ceramic membranes was performed prior to anaerobic digestion to remove toxins and inhibitors from the wastes based on naked barley as shown in Fig. 1. The concentrates obtained from the membrane separation could be recycled back to the screw decantor, while the cake formed from the screw decantor could be used as feedstock. Ceramic membranes being more endurable than polymeric ones at high temperature are employed because the temperatures of stillages discharged from alcohol fermentation industry is over 70 C. Membrane separation characteristics were studied with different types of stillages and inorganic membranes. The efficiency of physicochemical cleaning methods was studied, together with the cleaning mechanism. The performance of digesters with and without membrane pretreatment was also compared. 2. Experimental 2.1. Equipment and procedure Fig. 2 shows the simplified schematic of the membrane system (TIA, France) used in pretreating stillages. The transmembrane pressure was adjusted by the V1 or VZ valve playing an equal role in controlling the pressure. From the inlet and outlet gauge pressures an average transmembrane pressure was calculated, while the feed

Fig. 2. Schematic equipment.

diagram of laboratory-scale

membrane

Fig. 1. Proposed schematic diagram of waste treatment with a novel pretreatment involving ceramic membrane separation.

flow velocity was regulated by the F2 valve and ranged from 1.52 to 6.08 m/s. The membranes used were TIA 004 CF (0.4 pm pore) and TIA 0005 CF (0.05 pm pore) tubes, with 19 channels of 3.5 mm inner diameter in a single tube, which are composed of an alumina surface layer on a silicon carbide support. The total surface area of the membrane was 0.2 m2. To estimate the effect of pore sizes on permeation flux, flux decline during the concentration of the same feed ( 1: 1 mixture of naked barley and sweet potato) was checked using the 0.05 and 0.4 pm membranes. The operating conditions were as follows; transmembrane pressure, 5 bar; fluid velocity, 6.08 m/s; and temperature, 70C. Pretreatment of different types of stillages with the 0.05 pm membranes was performed using the following feed streams: naked barley ( 100%); a mixture of naked barley and sweet potato ( 1: 1); and a mixture of naked barley and tapioca ( 1: 1) . The operating conditions were the same as above. After emptying and flushing all the equipment

I.-S. Changet al. / JournalofMembraneScience

90 (1994) 131-139

133

with tap water ( 50 C) at the end of each experiment, water flux was always measured under the fued condition of 2 bar and 30C. Cleaning of the membrane was performed by recirculating solutions ( 1%) of neutral detergents (commercial detergents for kitchen-use), enzyme detergents (Novadan Kemi AIS, Ro-dazym, Denmark), hydrogen peroxide, and nitric acid. The common cleaning conditions were as follows: fluid velocity, 6.08 m/s; transmembrane pressure, 2 bar; temperature, N 70C; and cleaning time, 1 h. Other specific cleaning conditions are given in Table 3. In order to compare the cleaning efficiency, other ceramic membranes (Techsep, M6, France) and equipment (Techsep, CSV3R, France) were used. The M6 membrane having 0.08 p pores consists of a zirconia surface layer on a carbon support. The lumen flushing effect was also checked by closing the permeate side valves for 25 min during the concentration of naked barley wastes. The permeates from the ceramic membranes were further separated by ultrafiltration (Stirred cell, Amicon model 8200). The UF membranes used were PM 10 and PM30 with an effective area of 30.2 cm*. The applied pressure was controlled by nitrogen gas. The pretreated stillages were anaerobically digested at a thermophilic temperature of 55 oC. The digester was made by acrylic cylinders with a liquid volume of 8 1 and a gas space of N 2 1. The seed sludge was obtained from a thermophilic anaerobic digester in an alcohol-distillery wastewater treatment plant in Korea. The contents of the digester were continuously mixed by mechanical stirring at 50-60 rpm. The digester substrate was the permeate from the stillages based on naked barley. The digester was operated by gradually decreasing the hydraulic retention time (HRT) calculated as the ratio of reactor volume to feed flow rate. The gases produced in the digester were collected in an inverted graduated cylinder that was submerged into an acid brine confining solution ( 10% NaCl, 2%
H2S04)

2.2. Analytical methods COD (dichromate method), total solids, volatile solids, suspended solids, volatile suspended solids, and pH were measured according to the procedures as described by the Standard Methods [ lo]. Volatile fatty acids were determined using a commercial spectrophotometric method according to the manufacturers instructions (Hach, DR2000) and expressed as mg CH,COOH/l. Before the analysis of volatile fatty acids, samples were centrifuged at 3000 rpm ( N 1,200 g) for 15 min. The particle size distribution of stillages was measured using a microscope (Karlzeiss, Jeneval, Germany) and an image analyzer (Pias, Kit-500, Japan). 3. Results and discussion 3. I. Membrane pretreatment Fig. 3 shows the effect of membrane pore size on flux. In each pore size (0.4,0.05 p), the flux, which was initially kept at 250 and 280 l/m* h, respectively, gradually declined as the feed solution was concentrated. Although the initial water flux of the 0.4 pm membrane was much higher than that of the 0.05 pm membrane, the flux pro-

2 3 Volume Concentration Factor(log scale)

PI.
Fig. 3. Effect of membrane pore size on flux. Feed, naked barley and sweet potato based stillages.

134

I.-S. Chang et al. /Journal ofMembrane Science 90 (1994) 131-139

duced by the membrane with a 0.05 pm pore size was always higher than that of the 0.4 pm pore size, all other factors being equal. This result appeared to be associated with the relationship between particle size and pore size. As shown in Fig. 4, particles of 0.4-0.8 pm in stillages were dominating in the particle size distribution of the raw feed. Particles whose sizes are of the same order of magnitude as the membrane pore sizes are more likely to plug up the pores and cause pore blockage and fouling. In contrast, if the membrane pore size is much smaller than that of the particles to be separated, the particles will not be able to plug the pores, but to simply roll off under the shear forces generated by the flow. Thus, the 0.4 ,um membrane could be clogged more easily than the 0.05 pm membrane because the pore size of the former is similar to that of the dominating particles. Similar results were reported by others [ 111. This could be the reason for the unexpected flux behavior. On the basis of the above results, the 0.05 pm membrane was selected in this study. Solute rejections for different types of stillages are given in Table 1. Rejections of COD, total solids, and volatile solids were -40-609/o, but those of suspended solids and volatile suspended solids were almost complete. The composition of the permeates was independent of the feed stillages of different quality. This is an important fact to note because it implies the membrane
1 2 10/2.44

pretreatment serves as an equalization tank as well as a filtration process. In an alcohol-distillery wastewater treatment plant, operating conditions of anaerobic digesters, such as HRT, must be altered according to the change in the composition of the feed. However, if a membrane plant is introduced as a part of the pretreatment system, such a cumbersome change in digester conditions could be avoided. The variation of flux according to the concentration factor for the different types of stillages is shown in Fig. 5. Although the initial flux of stillages based on naked barley was the lowest, the final flux was the highest. Its production rate decreased slower compared to the curves based on mixed naked barley and sweet potato ( 1: 1) and mixed naked barley and tapioca ( 1: 1). According to the resistance-in-series model, flux (J) is expressed as: J=dP/(/&) (1) where AP is the transmembrane pressure, p is the viscosity of permeate, and RT is the total resistance. The total resistance is the sum of intrinsic membrane resistance, polarization resistance, and fouling resistance. However, as the polarization resistance is a function of the applied pressure which was kept constant during this ex-

177I2.10

. n .

naked barley (100%) naked barley 8 tapioca (1 :l) naked barley 8 sweet potato

(11)
4

o.oom tll 0 5 IO 15 20 25 U) Pamcle Size Dstribution(To) 35

2 ; Volume Concenwation scale) Pnctor(log

Fig. 4. Particle size distribution for alcohol-distillery wastes.

Fig. 5. Flux versus concentration factor with different types of stillages. Membrane, TIA 0005 CF (0.05 pm).

I.-S. Chang et al. /Journal ofMembrane Science 90 (1994) 131-139 Table 1 Composition Analysis

135

of alcohol-distillery

wastes before and after membrane treatment

Raw materials for fermentation Naked barley and sweet potato (1:l) Feedb Permeateb 20,600-26,500 (50.8-61.8) 19,900-24,600 (43.3-54.1) 16,000- 18,600 (51.6-58.3) negligible (-100) negligible (-100) NM 3.9 Naked barley and tapioca ( 1: 1) Naked barley ( 100%)

Feed 47,000 33,600 27,520 8,227 7,753 NM 3.3

Permeate 23,600-28,300 (39.8-49.8) 22,400-23,700 (29.5-33.3) 19,100-19,300 (29.9-30.6) negligible (-100) negligible (-100) NM 3.3

Feed 35,600 27,700 23,000 9,750 9,270 3,360 3.3

Permeate 17,300-19,300 (45.8-51.4) 15,900-18,300 (33.9-42.6) 13,400-14,200 (38.3-41.7) negligible (-100) negligible (-100) 2,850 3.3

Chemical oxygen demand Total solids Volatile solids Suspended solids Volatile suspended solids Volatile fatty acids (as mg CH&OOH/l) PH

53,900 43,400 38,400 18,600 17,800 NM* 3.8

Membrane, TIA 0005 CF (0.05 ym). bUnit, mg/l. Rejection = ( 1 -concentration of permeate/concentration *Not measured. Table 2 Fouling tendency for different types of stillages Raw material for fermentation Naked barley ( 100%) Naked barley and tapioca ( 1: 1) Naked barley and sweet potato ( 1: 1)

of feed) x 100, %.

Pore size of ceramic membranes (,um) 0.05 0.05 0.05

JeS (l/m2 h) 1,098 1,006 2,210

JlVf
(l/m2 h) 892b 45Sb 463

1-J:

0.188 0.545 0.890

*Initial water flux. bFinal water flux after water washing. Final water flux after cleaning with enzyme detergent.

periment, the different profiles of flux decline may be in close association with the variations of fouling resistance during concentration. The fouling tendency or fouling level, defined as 1 - Jti/Jti, is given for each type of stillage in Table 2. The fouling tendency of 100% naked barley stillage was the lowest, resulting in the slowest flux reduction as concentration proceeded. However, the mixed naked barley and sweet potato stillage exhibited the highest fouling tendency even after chemical cleaning, resulting in severe flux reduction.

Fig. 6 shows the extent of flux recovery (water flux after cleaning/initial water fluxx 100, %) after flushing with three types of cleaning agents. Hydrogen peroxide was the most effective and provided almost complete flux recovery for the inorganic membrane based on alumina. The cleaning effkiency of hydrogen peroxide was also tested for other inorganic membranes. When fouling levels for alumina and zirconia membranes were equal to 0.895 and 0.897, respectively, the degree of flux recovery of a zirconia based membrane (83%) was less than that of the

136

I.-S. Chang et al. /Journal ofMembraneScience 90 (1994) 131-139

Table 3 Cleaning efficiency with cleaning sequence of hydrogen peroxide followed by nitric acid Membrane materials Cleaning agents A120: 1% H202
Flux recovery %)

ZrOc, 0.5% HN09 -18 15 30 2 1% H202 83 60 70 1.6 1% HN09 -22 30 30 1.5

76 20 30 2

Cleaning conditions time (min) temp. (C) AP (bar)

Feed, 100% naked barley based stillages. bMembrane, TIA 0005 CF (0.05 pm). Membrane, M6 (0.08 pm).

predicted lme without lumen flushmg

10 n LI

Neutral detergent

Enzyme detergent Cleanmg Reagents Used

Hydrogen peroxide

6,

Volume Concen~atron Factor(log scale)

Fig. 6. Efficiencies of different cleaning reagents. Membrane, TIA 0005 CF (0.05 pm); feed, 100% naked barley based stillages.

Fig. 7. Effect of lumen flushing on flux enhancement. Membrane, TIA 0005 CF (0.05 pm); feed, 100% naked barley based stillages.

alumina based membrane ( N 100%) at the same cleaning conditions: temperature, 70 C; cleaning time, 1 h. This discrepancy between the two membranes could be attributed to the different adsorption and desorption mechanisms of the foulants. But the flux recovery was only 76% even for the alumina based membrane at milder cleaning conditions: temperature, 30 C; cleaning time, 20 min (Table 3 ) . Nitric acid is a commonly used cleaning agent which effectively removes fouling due to mineral scaling. But when flux recovery to initial water flux was not completely achieved by hydrogen

peroxide for both types of membranes, the further cleaning with nitric acid surprisingly gave a negative flux recovery to water flux after hydrogen peroxide cleaning, that is, a decrease of water flux after cleaning (Table 3). This unexpected phenomenon was observed in cleaning both the alumina based membrane ( - 18%) and the zirconia based one ( -22%) under the specific cleaning conditions. Other researchers also reported the poor cleaning efftciency of nitric acid, but the causes were not clearly illuminated [ 12,13 1. Here, it was hypothesized by a ligand exchange model [ 14,15 1, where anionic ligands

I.-S. Chang et al. /Journal ofMembrane Science 90 (1994) 131-139 Table 4 The permeate composition

137

(mg/l)

according to membrane pore size Alcohol-distillery wastes* Permeate 1 Permeate 2b Permeate 3b

Membrane used (pore size) Chemical oxygen demand Total solids Volatile solids Suspended solids Volatile suspended solids

53,900 43,400 38,300 18,600 17,800

TIA 00005 (0.05 pm) 20,600 19,850 16,030 negligible negligible of permeate 1.

PM 30 (MWCO 30,000) 20,400 17,926 12,000 negligible negligible

PM 10 (MWCO 10,000) 15,900 NM NM negligible negligible

Pretreated by a screw decantor. bPermeates obtained from further ultrafiltration Not measured. Table 5 Performance

of the anaerobic digester with an without membrane pretreatment Without membrane pretreatment Stillage Efnuent HRT 15 days HRT 23 days

of naked barley based stillages With membrane pretreatment Permeate of stillage HRT 30 days Effluent HRT 6.7 days

[5]

Volatile fatty acids (asmg (CH,COOH/l) COD (mg/l) COD removal (% ) Organic loading rate (kg COD/m3 day)

680 45,000

<7,050 18,500 58.8 3.1

2,764 14,100 68.7 2.0

1,580 11,000 75.6 1.5

2,170 20,280

300-500 2,000-2,300 89-91 3.03

*Membrane, TIA 0005 CF (0.05 ,um).

replace the surface hydroxyl groups of alumina or zirconia, surface as the following reaction: S-OH+L-=S-L+OH(2)

where S is the surface of the inorganic membrane and L- is the foulant or nitrate as ligand. The lower pH due to nitric acid would make this reaction go to the right and this could promote the adsorption of the foulants remaining in the cleaning solution. This might be a possible reason for the flux decrease. Nitrate groups can also substitute for hydroxyl groups at the membrane surface under acidic conditions [ 161, which could lead to the flux decrease. In order to examine whether or not the permeation flux can be improved during concentration, lumen flushing was attempted in this tubu-

lar module, although it is generally carried out in a hollow fiber module. As shown in Fig. 7, flux was increased by N 20-25 l/m* h for each lumen flushing as compared to predicted values. Further ultrafiltration of the permeates from the ceramic membrane was performed using polymeric membranes. The results are given in Table 4. Using the PM10 membrane, COD was reduced from 20,600 to 15,900 mg/l, which gave only a 20% removal efficiency. This result indicates that the permeates from the ceramic membrane contained mostly low molecular weight soluble organics of less than 10,000 daltons. The permeates from the ceramic membrane must be further treated either by biological wastewater treatment or nanofiltration method to meet discharge limits for industrial wastewaters.

138

I.-S. Changet al. / JournalofMembraneScience

90 (1994) 131-139

3.2. Anaerobic digestion The pretreated stillages based on naked barley were anaerobically digested at a thermophilic temperature of 5 5 C. The digester was operated for 300 days by gradually decreasing HRT. Table 5 illuminates the performance of the digester with and without membrane pretreatment. With membrane pretreatment, the volatile fatty acids concentration which indicates the digesters stability was in the range of 300-500 mg/l and COD removal efficiency was around 90% even for a HRT of 6.7 days. The reason for the improved performance of the digester including membrane pretreatment seems to be that inhibitorproducing materials (e.g., suspended solids) present in the stillages have been removed by membrane separation. As an intermediate metabolite from the anaerobic decomposition of suspended solids fraction from naked barley based stillages, propionic acids, thermodynamically hard to be degraded, were more produced than other types of stillages [ 17 1. Furthermore, the removal of the suspended particles may have shortened the overall biochemical reaction rate. This result agreed with the other findings which showed that the feed solution free of particulates enhanced the performance of the anaerobic digestion [ 18,19 1.

lages composition. Membrane separation could effectively be used instead of an equalization tank in the alcohol-distillery waste treatment. The reason for the improved performance of the digester including membrane pretreatment seems to be that inhibitor-producing materials present in the stillages have been removed by membrane separation, and/or removal of the suspended particles may have shortened the overall biochemical reaction rate.

5. List of symbols J J wf Jwi LAP RT S p flux (l/m* h) final water flux (l/m* h) initial water flux (l/m* h) foulant or nitrate as ligand transmembrane pressure (bar) total resistance surface of inorganic membrane viscosity of permeate

7. Acknowledgements We are grateful to Mr. J.S. Yum and Miss H.S. Lee for their assistance during the experiments at the Institute of Environmental Science & Engineering, Seoul National University.

4. Conclusions The higher flux profile obtained with the membrane of smaller pore (0.05 pm) during concentration of stillages could be in close association with the particle size distribution of the stillages. Permeation flux during concentration of mixed stillages was prone to decrease more rapidly due to higher fouling tendency than that of naked barley based stillage. The cleaning sequence of hydrogen peroxide followed by nitric acid gave a negative flux recovery, which could be explained by a ligand exchange model. The quality of permeates produced by the ceramic membrane was independent of the stil8. References
[ 1 ]C.P. Leslie Grady, Jr. and H.C. Lim, Biological Wastewater Treatment, Marcel Dekker, New York, 1980, pp. 833-886. [ 2 ] G.F. Parkin and W.F. Owen, Fundamentals of anaerobic digestion, J. Environ. Eng., 112 (1986) 867. [ 3lD.A. Stafford, Methane Production from Waste Organic Matter, CRC Press, Boca Raton, FL, 198 1. [ 41 D.P. Chynoweth and R. Isaason (Eds.), Anaerobic Digestion of Biomass, Elsevier, Amsterdam, 1987, pp. 91-107. [ 5 ]Y. Chung, M. Park, J. Cho and C. Kim, The effect of different substrate of anaerobic digestion of distillery wastewater, J. Korea Sot. Environ. Eng., 12 ( 1990) 63. [ 61 K. Baek, Y. Cho, J. Lee, D. Chang and Y. Chung, Inhibition of anaerobic digestion for alcohol distillery waste, J. Korea Sot. Water Pollut. Res.Control, 6 ( 1990) 28.

I.-S. Chang et al. / JournalofMembraneScience [ 7]Y. Cho, K. Baek and D. Chang, Effect of influent dilution of the inhibition of methane fermentation for naked barley distillery wastes, J. Korea Sot. Environ. Eng., 12 (1990) 21. [ 8]Y. Kim, C. Won and P. Chung, On the treatment conditions and analysis of distillate wastewater, J. Korea Sot. Water Pollut. Res. Control, 7 ( 199 1) 1. [ 91 J.A. Eastman and J.F. Ferguson, Solubilization of particulate organic carbon during the acid phase of anaerobic digestion, J. Water Pollut. Control Fed., 53 ( 198 1) 352. [ 10] L.S. Clescett, A.E. Greenberg and R.R. Trussell ( Eds. ), Standards Methods for the Examination of Water and Wastewater, 17th ed., American Public Health Association, 1989. [ 11 IT. Imasaka, N. Kanekuni, H. So and S. Yoshino, Crossflow filtration of methane fermentation broth by ceramic membrane, J. Ferment. Bioeng., 68 ( 1989) 200. [12]G. Dautin, U. Merin, F.L. Kerherve, J.P. Labbt, A. Quemerais and C. Bousser, Efficiency of cleaning agents for an inorganic membrane after milk ultrafiltration, J. Dairy Res., 59 (1992) 29. [ 13 ] G. Dautin, U. Met-in, J.P. LabbC, A. Qutmerais and F.L. Kerhervt, Cleaning of inorganic membranes after whey

90 (1994) 131-139

139 Biotechnol. Bioeng., 38 ( 199 1)

and milk ultrafiltration,

Yates and T.W. Healy, Mechanism of anion adsorption of the ferric and chromic dioxide/water interface, J. Colloid Interface Sci., 52 (1975) 222. [ 15 ] R. Kummert and W. Stumm, The surface complexation of organic acids on hydrous y-AlzOs, J. Colloid Interface Sci., 75 (1980) 373. [ 161 S. Dumon and H. Bamier, Ultrafiltration of proteins on ZrG2 membranes. The influence of surface chemistry and solution chemistry on adsorption, J. Membrane Sci., 74 ( 1992) 289. [ 17lI.S. Chang, J.S. Kim, C.H. Lee and K.S. Park, A Study on the inhibition of anaerobic digestion for alcohol distillery wastes, J. Korean Ind. Eng. Chem., 4 ( 1993) 576. [ 181 K.O. Baugh, J.A. Levy and P.L. McCarty, Thermochemical pretreatment of lignocellulose to enhance methane fermentation. II. Evaluation and application of pretreatment model, Biotechnol. Bioeng., 31 ( 1988) 62. [ 191 H.H. Hahn and R. Klute (Eds. ), Pretreatment in Chemical Water and Wastewater Treatment, SpringerVerlag, Berlin, 1988, pp. 249-260.

[ 141;;.