Bioresource Technology 97 (2006) 114

Combined biological and membrane treatment of food-processing wastewater to achieve dry-ditch criteria: Pilot and full-scale performance
George Nakhla
a

a,*

, Andrew Lugowski b, Javnika Patel b, Victor Rivest

Department of Chemical and Biochemical Engineering, University of Western Ontario, London, ON, Canada N6A 5B9 b Conestoga-Rovers & Associates, 651 Colby Street, Waterloo, ON, Canada N2V 1C2 c Sun-Brite Canning, P.O. Box 70, Ruthven, ON, Canada N0P 2G0 Received 6 December 2004; received in revised form 10 March 2005; accepted 10 March 2005 Available online 31 May 2005

Abstract This study tested the applicability of a submerged vacuum ultraltration membrane technology in combination with the biological treatment system to achieve dry-ditch criteria stipulated as follows: BOD5, TSS, NH3-N, and total phosphorous (TP) concentration not exceeding 10, 10, 1, and 0.5 mg/L respectively for the treatment of high strength food-processing wastewater. During the study, the biological system, operated at average hydraulic retention time of 56 days, achieved 9596.5% BOD removal and 9699% COD removal. The external membrane system ensured the achievability of the BOD and TSS criteria, with BOD and TSS concentrations in the permeate of 12 and 18 mg/L respectively. Nitrate, and nitrite concentrations increased during membrane ltration, while ammonia concentrations decreased. The most salient nding of this study is that, contrary to common belief, for industrial wastewaters, the lterability of the mixed liquor is inuenced by the soluble organics, and may be low, thus necessitating operation of bioreactors at low mixed liquor solids. This study demonstrated that bioreactors operated at low SRTs and in combination with ultraltration can still achieve superior euent quality that may meet reuse criteria at reasonable cost. 2005 Elsevier Ltd. All rights reserved.
Keywords: Suspended solids; Biological treatment; Ultraltration membrane; Food-processing

1. Introduction Membrane technology has been used for the treatment of municipal and industrial wastewaters from tanneries (Krauth, 1996), textiles (Rozzi et al., 2000), chemical (Livingston et al., 1998; Greene et al., 2000), lie ` res, 2000; Cupeand food-processing (Mavrov and Be rus, 1998; Mavrov et al., 1997) facilities. Membrane technology has been applied successfully to the foodprocessing industry as membrane bioreactors (Cantor et al., 2000) and physical separators (Kuemmel et al.,
Corresponding author. Tel.: +1 519 661 2111/85470; fax: +1 519 850 2921. E-mail address: gnakhla@eng.uwo.ca (G. Nakhla). 0960-8524/$ - see front matter 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.biortech.2005.03.034
*

2000; Fakhrul-Razi and Noor, 1999). By far, most of the MBRs applied for municipal and industrial wastewater treatment utilize external membranes in contrast to the submerged vacuum membranes, which are employed in this study. The vacuum submerged membranes, used in this study were provided by Zenon Environmental (Oakville, Ontario, Canada), have a nominal pore opening of 0.036 lm, molecular cuto point of 300 kilodaltons, and operate at low vacuum pressures (less than 0.3 atm). The main advantages of these membranes include lower energy consumption, better euent quality, better retention of microbes and viruses, and less fouling due to continuous cleaning of the membranes by air. Due to operation at high solids retention times (SRTs) and consequently high mixed

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

liquor suspended solids (MLSS) concentrations, membrane bioreactors have been reported to enhance biodegradation of non-readily biodegradable compounds including high molecular weight compounds (Reemtsma et al., 2002; Schro der, 2002). However, even when used for solids separation only, biological activity on membranes cannot be ruled out due to the formation of a biolm on the membrane surface. Despite the very brief contact time in the membrane unit, the very high biosolids concentrations will inevitably inuence euent quality. While MBRs have recently elicited signicant interest, biological activity during membrane separation and its impact on euent quality has not received attention. While the operation of membrane bioreactors at high SRTs and consequently high MLSS concentrations has been thoroughly reported in the literature, the performance of membranes at low SRTs has not been thoroughly investigated. Chaize and Huyard (1990) ran a pilot plant MBR with external ultraltration membranes on municipal wastewaters at hydraulic retention times (HRT) of 28 h, long SRT of 50100 days, and high biomass concentrations of 810 g/L, and achieved virtually complete removal of organics and nitrication. Similarly Muller et al. (1995) have operated an MBR system with external ultraltration on domestic wastewater at biomass concentrations of 4050 g/L and reported greater than 90% removal of carbon and complete nitrication. On the other hand, Ng and Hermanowicz (2003) have operated an MBR system employing submerged pressure membranes on synthetic

wastewater with average COD of 400 mg/L at HRTs of 36 h and SRT of 0.255 days, and average mixed liquor suspended solids concentration of 3462300 mg/L and reported complete removal of carbon and nitrication at an SRT of 5 days. Cicek et al. (2001) have shown that the lterability of wastewater by membranes is inuenced not only by suspended solids but also by soluble products. Thus, the performance of membranes in industrial wastes applications may dier markedly from municipal wastewater treatment. Furthermore, the achievability of strict surface discharge criteria with ultraltration membrane bioreactors has not solicited much attention since this feat is readily accomplished by nanoltration and reverse osmosis systems, despite the high energy requirements. The primary objective of this study was to test the applicability of a submerged vacuum ultraltration membrane in combination with biological treatment in the food-processing industry to achieve stringent dry-ditch criteria. A detailed pilot study was undertaken at a food-processing facility to investigate the achievability of euent discharge criteria of 5-day biochemical oxygen demand (BOD5), total suspended solids (TSS), ammonia nitrogen (NH3-N), and total phosphorous (TP) concentration not exceeding 10, 10, 1, and 0.5 mg/L respectively from the treatment of wastewaters generated primarily during the o-season from canning of various products, predominantly beans. Preliminary data from full-scale operations is also included. This study demonstrated that biological activity occurred in membrane ltration systems even at relatively low inuent biomass concentrations.

UF UNIT BLOWER/PUMP HOUSE Concentrate Waste Activated Sludge ANOXIC TANK SCREEN SPRAY IRRIGATION FIELD AERATION TANK FINAL CLARIFIER Permeate

Final Effluent (Direct Discharge)

PUMP HOUSE

Internal Recirculation Influent

PRIMARY CLARIFIER

FACULTATIVE LAGOON SYSTEM NOTE: Future Additions

Fig. 1. Process ow diagram of treatment system including pilot-scale membrane.

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

2. Methods
VSS (mg/L)

2.1. Description of existing treatment system The ultimate wastewater treatment system to be implemented at the site includes primary clarication to remove suspended solids, anoxic/oxic biological treatment for removal of organics followed by secondary clarication for solid/liquid separation prior to discharge to a dry-ditch. The system has been built in phases and the existing facilities include primary clarication, a 480,365 US gal aeration tank and a secondary clarier. Agitation and oxygen in the aeration tank are provided by a diused aeration system consisting of 284 medium-bubble diusers and two 100-hp blowers. Dissolved oxygen (DO) concentrations in the aeration tank varied from 2 to 4 mg/L. Due to the lack of sludge recirculation systems, daily disposal of the entire inuent wastewater quality was exercised. Settled sludge was hauled directly from the aeration tank to the local wastewater treatment plant. The total volume of settled sludge hauled was equal to the daily wastewater ow. Mixing in the aeration tank was stopped for 8 h, during which time the supernatant was pumped out of the aeration tank and trucked to the local wastewater treatment plant. Due to the lack of full-scale solid/liquid separation process, the biological system was operated as a fed sequencing batch reactor as opposed to an activated sludge, i.e. the SRT was not equal to the hydraulic retention time (HRT) of approximately 56 days, based on an average wastewater ow rate of 60,00075,000 US gpd, due to allocation of 120,091 US gal of the aeration tank capacity for storage of wastewater in case of emergency. The pilot-scale membrane system used in this study was provided by Zenon Environmental Inc. of Oakville, Ontario and utilized a 180 US gal aerated tank to accommodate the 500 ft2 ZeeWeed hollow ber, vacuum ltration membrane. The membrane tank was continuously aerated to reduce fouling and cleaning requirements, and as a result the DO concentration in the permeate typically ranged from 2 to 3 mg/L. Due to visible quantities of ne colloidal material in the raw wastewater, the selected membrane opening was 0.04 lm to remove colloidal material, bacteria, and viruses, thus rendering the euent amenable for recycle back to the production facility to be used for cleaning and other operations. The membrane ltration system was operated at four distinct permeate ows: 2.0, 2.5, 3.5, and 3.8 US gpm corresponding to uxes of 5.8, 7.2, 10, and 11 gpd/ft2. However, due to the close ows of 3.5 and 3.8 US gpm, the data has been combined. Thus the system operation is divided into three operating periods: OP-1 at 2.0 US gpm which lasted for 25 days, OP-2 at 2.5 US gpm which lasted for 9 days, and OP-3 at permeate ows was 3.53.8 US gpm, which

1262242 1882374 1.4170.0 0.555.0 0.0020.212 0.16.5 0.29.0 10481460 2703540 5917450 2881400

TSS (mg/L)

PO4 (mg/L)

PO4-P (mg/L)

NO2 (mg/L)

NO3 (mg/L)

NH3 (mg/L)

TKN (mg/L)

COD (mg/L)

Soluble

Total

Soluble

920 326 (22) Mean SD (n) 1448 475 (22)

3120 1619 (38) (n): Represents the number of data points.

BOD (mg/L)

Table 1 Wastewater characterization

Range (minmax)

2342256

Total

1965 996 (23)

1254 291 (2)

3.0 2.0 (39)

1.5 2.0 (38)

0.026 0.044 (38)

20.6 13.0 (36)

68.5 37.0 (33)

488 429 (24)

447 409 (24)

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

lasted for 42 days, following complete start-up and commissioning of the full-scale biological system and pilotscale membrane unit. The maximum capacity of the pilot unit was 5 US gpm corresponding to 7205 US gpd or 7.5% of the total wastewater ow. A process ow diagram of the system including the pilot-scale membrane system is shown in Fig. 1. The membranes were cleaned in accordance with the procedures specied by Zenon Environmental (Burlington, Ontario). Two types of cleanings were performed: maintenance clean and a recovery clean. A maintenance clean is a preventative cleaning, that is performed on a regular basis, and basically involves 10 cycles each comprised of stopping the permeate while continuing to aerate for 510 min, backpulsing (BP) with a 1000 mg/L sodium hypochlorite solution at 12 gallons/ft2/days for 30 s, and relaxing for 4.5 min. A recovery clean is performed when the suction required to permeate through the membrane reaches 79 psi, and involves alternate soaking in concentrated solutions of sodium hypochlorite (1000 mg/L) to remove organic debris, and 2 g/L of citric acid to remove inorganics. Two recovery cleans were performed on day 13 and day 53, while weekly maintenance cleans were initiated on day 42. The full-scale membrane system designed for a permeate ow of 82.3 US gpm (120,000 US gpd) consisted of forty-eight (48) 500-ft2 membrane cassettes for a total

area of 24,000 ft2, submerged in a 12 0 (L) 7.8 0 (W) 8.67 0 (side water depth) aerated epoxy coated carbon steel tank. At full operating capacity, with a permeate to concentrate ow ratio of 1:1, the HRT in the membrane tank was about 36 min only. The full-scale system was commissioned and fully operational by day 687 from the start of the pilot plant. 2.2. Wastewater characteristics Wastewater generation and quality from the foodprocessing facilities varies considerably throughout the year. Peak wastewater ows at approximately 480,365 US gal are generated during the tomato canning season extending from August to October while much lower wastewater ows at approximately 60,00075,000 US gal result from the processing of other vegetable products during the rest of the year. The organic matter concentrations during tomato canning are higher than during the rest of the year. The results presented here pertain to the treatment of wastewater from processing of vegetable products exclusive of tomato canning wastewater. Table 1 presents the variations in primary euent (inuent to biological system) quality. BOD total mostly ranged from 234 to 2256 mg/L while COD varied more widely between 591 and 7450 mg/L. Phosphorus ranged

Table 2 Pilotplant operational conditions Permeate ow (US gpm) Operation (days) Aeration tank HRT (h) Average F/M ratio in aeration tank BOD g BOD/g MLVSS/day 2.0 2.5 3.53.8 25 9 42 215 185 187 0.10 0.10 0.31 COD g COD/g MLVSS/day 0.29 0.25 0.35 0.74 0.59 0.42 MBR HRT (h) Average F/M ratio in MBR BODs g BODs/g VSS/day 1.26 0.55 0.65 CODs g CODs/g VSS/day 1.28 1.06 NA BOD g BOD/g VSS/day 6.31 7.20 7.15 COD g COD/g VSS/day 13.7 14.5 NA

NA: Not available.

Table 3 Overall pilot plant system performance Permeate ow Aeration tank BODT % removal Operating period 1 Range (minmax) Mean SD (n) Operating period 2 Range (minmax) Mean SD (n) Operating period 3 Range (minmax) Mean SD (n) 4172 59 11 (7) 4168 54 19 (2) 7490 85 7 (4) BODS % removal 4995 84 18 (6) 9397 95 3 (2) 8396 91 6 (4) CODT % removal 1389 68 24 (10) 6465 64 1 (2) 1394 55 29 (13) CODS % removal 9498 96 2 (5) 9697 97 0 (2) 8698 93 5 (11) Membrane BOD % removal 4899 75 27 (4) 9697 96 1 (2) 2596 62 31 (4) COD % removal 4469 56 13 (3) 4978 63 21 (2) 1375 51 22 (8)

(n): Represents the number of data points.

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

from 0.5 to 55 mg/L depending on the production process with an average of 20 mg/L, which was satisfactory for maintaining biological activity. As presented in Table 1, highly uctuating chemical characteristics of wastewater required a combined biological and membrane treatment system to operate over a wide range of suspended solids concentrations, as well as to achieve stringent euent quality objectives. 2.3. Sampling and monitoring program The pilot system was commissioned on April 26, 2002 and the study lasted for 3 months. Daily grab samples of the primary clarier euent (inuent to biological treatment), aeration tank mixed liquor (inuent to membrane system), and ZeeWeed permeate were collected

and analyzed for COD (total and soluble), NH4-N, ni trate nitrogen NO 3 -N, nitrite nitrogen NO2 -N and 3 phosphate phosphorus PO4 -P concentrations. Grab samples of the aforementioned three waste streams were collected twice a week, and sent o-site for analysis of the 5-day BOD (total and soluble), total suspended solids (TSS), and volatile suspended solids (VSS) concentrations. Weekly grab samples from the membrane tank were collected and analyzed o-site for TSS and VSS. For the full-scale system, the same streams were analyzed and for the same parameters as the pilot unit, albeit at a lower frequency. O-site analysis was done once a week and parameters determined on-site, i.e. 3 COD, NH4-N, NO 3 -N, NO2 -N, and PO4 -P were analyzed twice a week. For the membrane pilot plant, the permeate ow, the reject ow, vacuum before and after

10000

Above MAC Aeration Tank: 15/15 Permeate: 5/17

Influent Aeration Tank Aeration Tank - Soluble Permeate

1000

Log BOD (mg/L)

100

MAC = 10 mg/L 10

0.1 0 20 40 60 80 100 120

(a)
10000 Above MAC Aeration Tank Soluble: 2/4 Permeate: 0/4

Time (days)
Influent Aeration Tank Aeration Tank - Soluble Permeate

1000

Log BOD (mg/L)

100

MAC = 10 mg/L 10

1 0 2 4 6 8 10 12 14

(b)

Time (days)

Fig. 2. Temporal variation of inuent, aeration tank and permeate BOD: (a) pilot data; (b) full scale data.

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

backwash, ZeeWeed tank temperature, and pH were recorded several times a day. Due to the high concentrations of colloidal matter in the wastewater, the system was based on a permeate to concentrate ratio of 1:1. 2.4. Testing methodology
COD (total and soluble), NH4-N, NO 3 -N, NO2 -N 3 and PO4 -P concentrations were measured on-site using a Hach DR 2500 system. Standards with known concentrations of the various analytes were run with all analysis, and tests were repeated when the accuracy of standards determinations was below 95%. The chemical analyses for BOD (total and soluble), TSS, and VSS concentrations were carried out o-site at certied laboratories in accordance with the Standard Methods SM5210 B, SM2540 B, and SM2540 E respectively (APHA, 1985). It must be asserted that all BOD results reported here are 5-day BOD. Samples were transported to the o-site laboratory in ice-packed coolers, and stored at the lab in a cold room for a maximum period of 1 day prior to analysis. Standard deviations listed in the various tables were calculated using all the pertinent analytical data using the statistical formulae in Microsoft Excel, and reect the variability of the water quality, not the accuracy of the analytical tests.

3. Results and discussion 3.1. Organics removal A summary of the operational conditions for the three operational period OP-1 to OP-3 during the pilot

plant testing is presented in Table 2. It is apparent that, despite the long HRT in the aeration tank, it operated at loadings similar to conventional activated sludge systems with average food-to-microorganisms (F/M) ratios ranging from 0.1 to 0.3 g BOD/g VSS-d. For the membrane tank, the HRT was based on the total inuent ow rate from the aeration system and not based on the permeate ow. Due to the low HRT in the membrane tank, and despite the excellent removal of BOD in the aeration tank, the soluble organic loadings were still high in the 0.551.26 g BOD/g VSS-d range, typical of high rate systems. It should be noted that ammonia nitrogen in the raw wastewater (Table 1) was extremely low relative to organic matter with the BOD:NH3-N ratio, based on average observed concentrations of 483:1, thus indicating a severe nitrogen deciency. This deciency was corrected by addition of 751 kg of dry urea (40% nitrogen) over the course of the study to both the inuent as well as the aeration tank. This translates to an average nitrogen concentration based on a ow of 75,000 US gpd and 76 days of net operation of 10 mg/L, thus reducing the BOD:NH3-N ratio to a more typical 144:1. A summary of the removals of total and soluble BOD and COD in the full-scale aeration tank and across the pilot-scale membrane during the pilot plant study is depicted in Table 3. BOD and COD removals by the membrane system were based on inuent soluble BOD and COD. Total BOD removal eciency in the biological system ranged from 41% to 90% averaging at 65% while total COD removal eciencies were generally similar varying from 13% to 94% with an average of 62%. Removal of soluble organics in the biological treatment system on the other hand was much higher, averaging

10000

1000

Log COD (mg/L)

100

10 Influent Aeration Tank Aeration Tank - Soluble Permeate 1 0 10 20 30 40 50 60 70 80 90 100

Time (days)

Fig. 3. Temporal variation of inuent, aeration tank and permeate COD.

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

7 0.520 3.7 5.4 (12) 2342178 1523 552 (11) Operating period 3 Range (minmax) Mean SD (n) 3.54.0 3.8 0.1 (30) 5914370 2199 1153 (16) (n): Represents the number of data points. 2881400 990 331 (11) 2703470 1568 1106 (12) 156698 404 195 (8) 2112151 974 574 (15) 20170 86 47 (11) 32196 88 47 (12) 0.12.2 1.3 0.7 (7) 1377 37 21 (8) 94100 98 2.1 (7) 4675 64 11 (5)

at 88% and 94% based on BOD and COD respectively. Approximately 2/3 of the organics were in soluble form and the organics were moderately biodegradable as reected by a BOD5-to-COD ratio of around 0.4:1. The dierence between total and soluble organics removal in the aeration tank is attributed to the slower biodegradation kinetics of particulate organics relative to the soluble organics, and the absence of full-scale solidliquid separation, i.e. secondary clarication, which would have increased particulates removal. It must be emphasized that the organic removal eciencies are remarkable considering that the biological system was run at low SRT and consequently low MLSS of mostly approximately 10001500 mg/L despite the long HRT. The diurnal variation of BOD, COD respectively, in the inuent, aeration tank, and permeate together with the anticipated discharge criteria is illustrated in Figs. 2 and 3. BOD results from the full-scale operation are also included in Fig. 2. As shown in Fig. 2, 100% and 50% of the aeration tank euent samples taken for soluble BOD analysis during the pilot study and full-scale study, respectively, exceeded the euent BOD criteria of 10 mg/L. During the pilot study, the nal euent BOD criteria was exceeded in ve out of 17 samples, four of them during the early commissioning and start-up phase, i.e. the system readily met the criterion for 94% of the time producing BOD of 12 mg/L, as shown in Table 4. Thus, in any of the three operating conditions listed in Table 4, the membrane would easily meet the 10 mg/L criterion, as conrmed by the permeate BOD data of Table 4 as well as the results of the fullscale operation. As depicted in Fig. 3, despite the highly uctuating inuent COD, and the aeration tank soluble COD varying from 30 to 200 mg/L, the permeate COD was stable at 3050 mg/L. It is noteworthy that the typical euent soluble COD from municipal wastewater treatment plants, is 3050 mg/L, which constitutes the inert soluble COD component of municipal waste (IAWQ, 1995). Thus, the achievability of a similar euent COD concentration in the context of high strength wastes is remarkable. Details of the performance of the membrane system during the pilot study are presented in Table 4. The membrane was operated at three dierent ow rates of 2.0, 2.5, and 3.53.8 US gpm. It should be noted that the BOD and COD removals calculated for the membrane were based on the soluble fractions of the inuent BOD and COD to the system. At a ow rate of 2.5 US gpm corresponding to a ux of 7.2 gpd/ft2, the average BOD and COD removals achieved were 96% and 63% respectively. Euent BOD and COD concentrations in the permeate at 22.5 US gpm were mostly in the 1.3 1.8 mg/L while euent CODs were in the range of 23 35 mg/L. At 3.53.8 US gpm or 10 gpd/ft2, average permeate BOD and COD concentrations increased

11 1 0.1 (2) 4469 56 18 (2) 9899 99 0.6 (2) 2335 28 6 (3) 1.31.8 1.5 0.3 (3) 6290 80 15 (3) 72196 134 88 (2) 720800 760 40 (3) 24103210 2900 429 (3) 10831354 1210 136 (3) 31906540 4903 1676 (3) 14282256 1770 432 (3) 2.02.0 2.0 0.0 (9) Operating period 1 Range (minmax) Mean SD (n) 542699 614 79 (3)

TSS (mg/L)

COD % removal

BOD % removal

COD (mg/L)

Permeate

Aeration tank

Table 4 Summary of membrane operational and performance dataBOD and COD (total and soluble)

Permeate (US gpm)

Operating period 2 Range (minmax) Mean SD (n)

2.52.5 2.5 0.0 (8)

11911347 1269 110 (2)

Inuent

BODT (mg/L)

29203610 3287 347 (3)

CODT (mg/L)

860939 900 56 (2)

BODS (mg/L)

21102750 2350 349 (3)

CODS (mg/L)

437705 571 190 (2)

BODT (mg/L)

10301300 1165 191 (2)

CODT (mg/L)

2465 45 29 (2)

BODS (mg/L)

7099 85 21 (2)

CODS (mg/L)

1.02.1 1.6 0.8 (2)

BOD (mg/L)

2236 30 7 (3)

9697 96 0.7 (2)

4978 63 21 (2)

314 98 (2)

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

sharply to 5.5 and 68.5 mg/L. At the low ows, the membrane ltration aected removal of approximately 43 mg BOD/L and approximately 50 mg COD/L. Since the soluble BOD and COD concentrations in the inuent were based on ltration through 0.45 lm while the membrane opening is 0.04 lm, it is possible that physical separation of particles in the 0.040.45 lm range was the primary removal mechanism. Alternatively, enhanced biodegradation of slowly biodegradable waste constituents that escaped biological treatment in the aeration tank may have contributed to the organics removal. Oxygen uptake rates in the membrane tank averaged 0.64 0.07 mg/L min, and it is thus estimated that the average COD removed by biodegradation in the membrane tank at an average HRT of 5090 min is 3257 mg/L. However, in the absence of the fraction of BOD

and COD in the aeration tank euent that is lterable through 0.04 lm lter, the predominance of physical ` -vis enhanced biodegradation cannot separation vis-a be precisely delineated. While the short hydraulic contact time in the membrane tank of 5090 min may not be conducive to the biodegradation of slowly biodegradable organics escaping a 56-day HRT biological treatment system, the relatively much higher biomass concentrations and dierent microbial groups (Cicek et al., 2001) may have induced biodegradation. 3.2. Suspended solids removal The temporal variation of the inuent, aeration tank euent, and permeate total suspended solids concentrations both during the pilot study and full-scale operation

10000

1000

Log TSS (mg/L)

100

MAC = 10 mg/L 10

1 Above MAC Aeration Tank: 26/26 Permeate: 3/19 0.1 0 20 40 60 80 100 Influent Aeration Tank Permeate 120

(a)
10000 Influent Aeration Tank Permeate

Time (days)

1000

Log TSS (mg/L)

100

10

MAC = 10 mg/L

0.1 Above MAC Aeration Tank: 4/4 Permeate: 0/4 0.01 0 2 4 6 8 10 12 14

(b)

Time (days)

Fig. 4. Temporal variation of inuent, aeration tank and permeate TSS: (a) pilot data; (b) full scale data.

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

9 7.451.9 33.9 19.3 (4) 4.219.6 10.2 8.3 (3) 2.347.1 23.3 22.5 (3) 5.821.6 11.4 8.9 (3) 2.741.6 21.0 19.5 (3) 6.621.0 15.7 8.0 (3) 0.16.5 2.3 3.6 (3) 2.84.7 3.5 1.0 (3) 1.013.0 6.4 6.1 (3) 1.322.6 10.5 11.0 (3) 0.221.8 7.4 6.0 (16) 0.29.0 2.9 2.6 (17) Operating period 3 Range (minmax) 3.54.0 Mean SD (n) 3.8 0.1 (30) 0.24.4 0.9 1.4 (16) (n): Represents the number of data points. 3.541.3 22.5 11.1 (13) 0.15.4 1.3 1.6 (11) 0.373.7 22.5 18.6 (16) 0.226.9 11.7 8.2 (16) 0.13.4 1.0 1.0 (13) 0.397.3 27.6 23.3 (16)

Table 5 Summary of membrane operational and performance datanitrogen and phosphorus

NO3-N + NO2-N (mg/L) Inuent NH3-N (mg/L)

is depicted in Fig. 4. The euent TSS criterion of 10 mg/ L was met in 85% of the samples during the pilot and 100% of the samples during full-scale operation (Fig. 4). Only three exceedances occurred, two being slightly higher than 10, i.e. 11 and 14 mg/L. Additionally, two out of the three exceedances of the TSS criteria occurred during the early commissioning and start-up phase of the pilot study. This data clearly demonstrates the achievability of the euent TSS criterion of 10 mg/L. While the mixed liquor suspended solids concentration in the aeration tank varied mildly, from 1000 to 2000 mg/L, the nal euent TSS concentrations were mostly in the 12 mg/L during stable operation of the membrane system. It is note-worthy however that the performance of the membrane with respect to TSS removal at much higher inuent TSS, i.e. 40006000 mg/L has not been tested in this study. Sludge yields in the aeration tank during the entire operating period of the pilot plant study were derived from charts of cumulative sludge produced (as VSS) versus cumulative COD/BOD removals. The calculated yields were 0.09 g VSS/g COD (R2 = 0.89), 0.15 g VSS/ g SCOD (R2 = 0.89), 0.215 g VSS/g BOD (R2 = 0.84), and 0.363 g VSS/g SBOD (R2 = 0.88). Thus, it is apparent that the system biological sludge yield is relatively low compared with the 0.5 g VSS/g BOD observed in municipal wastewater treatment. This low yield is attributed to the relatively low F/M ratio that the system operated at. It must be asserted that the volatile fraction of the aeration tank MLSS during the pilot study was 0.85, i.e. MLVSS was 85% of the MLSS. This atypically high ratio is attributed to two factors: the low concentration of inuent inorganic suspended solids, as depicted in Table 1, and the low operational sludge age. The high volatile fraction of mixed liquor suspended solids, combined with the low sludge yield indicates that particulate organics were biodegraded in the system. 3.3. Nutrient removal The performance of the combined biological and pilot-scale membrane system with respect to nitrogen and phosphorous removal is presented in Table 5 while the diurnal variations of the two aforementioned parameters are depicted graphically in Figs. 5 and 6. The adequacy of soluble phosphorous and deciency of soluble nitrogen is noteworthy. No chemicals to remove phosphorous were added during the pilot study. Nitrogen deciency was corrected by addition of urea as elaborated upon earlier. As shown in Fig. 5, during the pilot study the system exceeded the 1 mg/L NH4-N criteria in 13 of 28 samples. Two out of 13 samples that exceeded the NH4-N criterion occurred during the early commissioning and start-up phase. In the aeration tank euent, 50% of the samples exceeded the euent NH4-N criterion. Considering that the biological system

NO3-N + NO2-N (mg/L) Permeate NH3-N (mg/L) NO3-N + NO2-N (mg/L) Aeration tank NH3-N (mg/L) PO4-P (mg/L) PO4-P (mg/L)

PO4-P (mg/L)

Operating period 1 Range (minmax) 2.02.0 Mean SD (n) 2.0 0.0 (9)

2.28.1 4.5 2.6 (4)

0.30.5 0.4 0.1 (4)

22.155.0 41.1 15.2 (4)

0.43.7 1.9 1.7 (4)

2.919.4 14.4 7.7 (4)

6.453.9 32.5 20.6 (4)

0.13.4 1.1 1.6 (4)

2.332.5 17.9 13.5 (4)

Operating period 2 Range (minmax) 2.52.5 Mean SD (n) 2.5 0.0 (8)

Permeate (US gpm)

10

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

operated at very low SRTs and low biological solids concentrations that are not conducive to nitrication, the inadequacy of the nitrication is not surprising. Furthermore, since the waste is nitrogen decient, more controlled addition of urea may alone be sucient to meet the criteria. Thus, based on the pilot study results, for the three owrates 2.5, 2.0, 3.53.8 US gpm based on 3-, 4-, 14-sample averages, respectively, the membrane will not meet the 1 mg/L NH4-N criterion, as shown in Table 5. However, the results of the full-scale system indicate that the combined biological and membrane system met the criterion in all ve samples (Fig. 5). As anticipated, phosphorus removal without chemical addition was not adequate for meeting TP criterion of less than 0.5 mg/L. The euent TP criterion was exceeded in 86% of the samples. A chemical addition system is needed for phosphorus removal during the
100

full-scale operation. For all three operating permeate ow rates during the pilot study, shown in Table 5, the membrane did not meet the 0.5 mg/L TP criterion. 3.4. Membrane system performance The performance of the pilot-scale membrane ltration system with respect to nitrogen and phosphorus removals is presented in Table 5. Orthophosphate concentrations decreased during membrane ltration at the three operational conditions. At uxes of 7.2 and 1011 gpd/ft2, based on average conditions, 1.2 and 3.4 mg/L of PO3 4 -P were removed. Interestingly, nitrates and nitrites concentrations in the permeate increased, while ammonia concentrations decreased. The relationship between ammonia and nitrates concentrations in the permeate, and the inuent to the membrane

10

Log NH3-N (mg/L)

MAC = 1 mg/L

0.1

Above MAC Aeration Tank: 15/32 Permeate: 13/28 0.01 0 20 40 60 80 100

Influent Aeration Tank Permeate 120

(a)
10

Time (days)

Log NH3-N (mg/L)

MAC = 1 mg/L

0.1

Above MAC Aeration Tank: 4/5 Permeate: 0/5 0.01 0 2 4 6 8 10 12

Influent Aeration Tank Permeate 14 16

(b)

Time (days)

Fig. 5. Temporal variation of inuent, aeration tank and permeate NH3-N: (a) pilot data; (b) full scale data.

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

100

11

10

PO 4 -P (mg/L)

1 MAC = 0.5 mg/L

3-

0.1 Above MAC Influent: 35/36 Aeration Tank: 36/38 Permeate: 24/28 0.01 0 10 20 30 40 50 60 70 80 90 100 Influent Aeration Tank Permeate

Time (days)
Fig. 6. Temporal variation of inuent, aeration tank and permeate PO3 4 -P.

system both during the pilot study as well as during the full-scale operation is depicted in Figs. 7 and 8. It is evident that strong statistical correlations, as reected by R2 values of 0.81 and 0.94 respectively were observed. Changes to these two soluble species are not anticipated to occur during physical separation since ultralters retain only particles with molecular weights of 1000 1,000,000 (Scott, 1997) and the Zenon lter had a molecular cut-o point of 300 kilodaltons. The data of Fig. 7 indicates that in the range of 0.110 mg/L inuent ammonia concentration, ammonia decreased by about 24% during membrane ltration. The results of the full-scale operation were also similar with permeate ammonia concentrations of less than 0.1 mg/L at inuent concentrations of up to 2 mg/L. The concentration of nitrates and nitrites in the permeate, during the pilot
10 9 8

study, increased by about 15% (Fig. 8). The relatively good agreement between the full-scale data and the pilot study correlation conrms the trend of substantial increase in nitrates and nitrites during membrane ltration. Thus, nitrication was achieved in the membrane tank, despite the very short HRT of 0.40.75 h. This is consistent with the ndings of Muller et al. (1995) who observed up to 86% conversion of inuent nitrogen in domestic wastewater to nitrates in an MBR operating at MLSS concentrations of 1040 g/L. It is interesting to note that nitrication in the membrane tank occurred rapidly, after 20 days following start-up of the pilot system. While the change in ammonia concentration of only 24% may not be signicant considering the observed concentrations of about 1 mg/L, the change in nitrates (Fig. 8) and nitrites (Table 5) is signicant.

Membrane NH3-N (mg/L)

7 6 5 4 3 2 1 0 0 2 4 6 8 10 12 Pilot Data Full Scale Data y = 0.7646x R2 = 0.8107

Aeration Tank NH3 (mg/L)

Fig. 7. Relationship between permeate NH3-N and aeration tank NH3-N.

12
120

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

100

Membrane NO3 (mg/L)

80 y = 1.1983x R2 = 0.9377 60

40 Pilot Data Full Scale Data 20

0 0 10 20 30

Aeration Tank NO3 (mg/L)

40

50

60

70

80

Fig. 8. Relationship between permeate NO 3 and aeration tank NO3 .

The increase in nitrates plus nitrites across the membrane varied from 20% to 40%. The implications of this for denitrifying membrane bioreactors may be serious as it clearly manifests the potential for secondary nitrication. 3.5. Membrane cleaning The temporal variation of vacuum prior to BP and aeration tank MLSS concentrations is illustrated in Fig. 9. It should be noted that due to employment of a 1:1 permeate to retentate ratio, the ambient solids concentrations in the membrane tank were essentially twice the aeration tank or inuent MLSS. It is conspicuous

from Fig. 9 that generally BP started to increase, as the concentration of MLSS increased. However attempts to directly correlate BP vacuum with MLSS concentrations, as depicted in Fig. 10 clearly indicate that the correlation is not very strong. This implies that the lterability of the mixed liquor is not solely related to MLSS, i.e. soluble products aect lterability. This is consistent with the observations of Cicek et al. (2003), who provided evidence that the soluble fraction of mixed liquor, i.e. smaller size solutes, proteins, and sugars have a great impact on ltration performance in MBRs. For this particular waste, the BP vacuum of 8 to 9 psi, prescribed by the supplier as a trigger for recovery cleans, would be achieved at MLSS concentrations

14.0

3500

12.0

3000

Before BP Vacuum (" Hg)

10.0

2500

6.0

1500

4.0 Group 1 - 3.5 US gpm Group 2 - 2.5 US gpm Group 3 - 2.0 US gpm Group 4 - 3.8 US gpm MLSS 0 20 40 60 80 100

1000

2.0

500

0.0

0 120

Time (days)

Fig. 9. Temporal variance of before BP vacuum and MLSS.

MLSS (mg/L)

8.0

2000

G. Nakhla et al. / Bioresource Technology 97 (2006) 114

14.0

13

12.0

Before BP Vacuum (" Hg)

10.0

8.0

6.0

4.0 Group 1 - 3.5 US gpm Group 2 - 2.5 US gpm Group 3 - 2.0 US gpm Group 4 - 3.8 US gpm 0 500 1000 1500 2000 2500 3000 3500

2.0

0.0

MLSS (mg/L)

Fig. 10. Relationship between before BP vacuum and MLSS.

of about 3000 mg/L, which translates to 6000 mg/L in the membrane tank. This is much lower than the typical concentration of 12,000 mg/L. Since one of the fundamental premises for using membranes, is the ability of the MBR process to operate at high MLSS concentrations, the lterability of this particular waste may constrain this kinetic advantage of MBRs over conventional activated sludge systems. Under such circumstances, contrary to common practice, it is only possible to operate the bioreactor at relatively low MLSS concentrations in the 20003000 mg/L range. Thus the achievability of stringent soluble organics and nutrients criteria at low biomass concentrations is critical, and therefore the evidence provided in this study to the aforementioned eect is highly pertinent.

4. Summary and conclusions Based on the results of this pilot study, the following conclusions can be drawn:

Ammonia decreased, while nitrates, and nitrites increased across the membrane, indicating that biologically mediated nitrication occurred in the membrane tank despite the short hydraulic retention time of 5090 min. For this particular wastewater, the lterability of the biological sludges was fairly low and did not correlate well with biomass concentrations. Accordingly, operation of the bioreactor at low MLSS concentrations in the range of 20004000 mg/L, may be required to sustain satisfactory membrane performance without signicantly increasing cleaning frequency. Contrary to common applications of membranes wherein high biomass concentrations are employed, in this case the combination of biological treatment at low biomass concentrations in conjunction with membrane separation yielded excellent performance meeting stringent criteria.

Acknowledgements The membrane ultraltration system has achieved complete removal of suspended solids and colloidal matter, thus facilitating compliance with the dryditch discharge criteria. Approximately 9099% of the soluble BOD as well as 4050% of the soluble COD were removed by the membrane. Membrane ux was relatively low with satisfactory performance achieved at 5.87.2 US gpd/ft2. Orthophosphates removal across the membrane was minimal varying between 0.5 and 2.6 mg/L, thus necessitating chemical addition to meet the phosphorus criteria. The analytical assistance provided by the Leamington Water Pollution Control Plant and Ontario Clean Water Agency (OCWA) is greatly appreciated.

References
APHA, 1985. Standard Methods for the Examination of Water and Wastewater, sixteenth ed. American Public Health Association, Washington, DC. Cantor, J., Sutton, P.M., Steinheber, R., Novachis, L., 2000. Industrial Biotreatment: Plant capacity expansion and upgrading through

14

G. Nakhla et al. / Bioresource Technology 97 (2006) 114 Livingston, A.G., Arcangeli, J., Boam, A.T., Zhang, S., Marangon, M., Freitas dos Santos, L.M., 1998. Extractive membrane bioreactors for detoxication of chemical industry wastes: process development. J. Membrane Sci. 151 (1), 2944. lie ` res, E., 2000. Reduction of water consumption and Mavrov, V., Be wastewater quantities in the food industry by water recycling using membrane processes. Desalination 131, 7586. Mavrov, V., Fa hnrich, A., Chmiel, H., 1997. Treatment of lowcontaminated waste water from the food industry to produce water of drinking quality for reuse. Desalination 113, 197203. Muller, E.B., Stouthamer, A.H., van Verseveld, H.W., Eikelboom, D.H., 1995. Aerobic domestic waste water treatment in a pilot plant with complete sludge retention by cross-ow ltration. Water Res. 29 (4), 11791189. Ng, H.Y., Hermanowicz, S.W., 2003. Membrane bioreactor performance at short mean cell residence times. Proc of WEFTEC 2003 Conf., Los Angeles, CA. Reemtsma, T., Zywicki, B., Stueber, M., Kloepfer, A., Jekel, M., 2002. Removal of sulfur-organic polar micropollutants in a membrane bioreactor treating industrial wastewater. Environ. Sci. Technol. 36, 11021106. Rozzi, A., Malpei, F., Bianchi, R., Mattioli, D., 2000. Pilot-scale membrane bioreactor and reverse osmosis studies for direct reuse of secondary textile euents. Water Sci. Technol. 41 (1011), 189195. Schro der, H., 2002. Mass spectrometric monitoring of the degradation and elimination eciency for hardly eliminable and hardly biodegradable polar compounds by membrane bioreactors. Water Sci. Technol. 46 (3), 5764. Scott, K., 1997. Handbook of Industrial Membranes, rst ed. Elsevier Advanced Technology.

application of membrane biomasseuent separation. Paper presented at WEFTEC Conference, Anaheim, California. Chaize, S., Huyard, A., 1990. Membrane bioreactor on domestic wastewater treatment sludge production and modeling approach. Water Sci. Technol. 23, 15911600. Cicek, N., Macomber, J., Davel, J., Suidan, M.T., Audic, J.M., 2001. Eect of solids retention time on the performance and biological characteristics of a membrane bioreactor. Water Sci. Technol. 43 (11), 4350. Cicek, N., Suidan, M.T., Ginestet, P., Audic, J.M., 2003. Impact of soluble organic compounds on permeate ux in aerobic membrane bioreactor. Environ. Technol. 24 (2), 249256. Cuperus, F., 1998. Membrane processes in agro-food: state-of-the-art and new opportunities. Sep. Purif. Technol. 14, 233239. Fakhrul-Razi, A., Noor, M.J.M.M., 1999. Treatment of palm oil mill euent (POME) with the membrane anaerobic system (MAS). Water Sci. Technol. 39 (1011), 159163. Greene, M.R., Guarini, W.J., Manning, J.W., III, Togna, P.A., 2000. Wastewater treatment of a batch chemical plant euent with a membrane biological reactor. Paper presented at WEFTEC Conference, Anaheim, California. IAWQ, 1995. Activated Sludge Model No. 2, IAWQ Scientic and Technical Report No. 3. Krauth, K., 1996. Sustainable sewage treatment plantsapplication of nanoltration and ultraltration to a pressurized bioreactor. Water Sci. Technol. 34 (34), 389394. Kuemmel, R., Robert, J., Loewen, A., 2000. Improvement of waste water treatment, water recycling, and eciency in the dairy industry by combined membrane processes. Paper presented at WEFTEC Conference, Anaheim, California.