Performance of a crossflow membrane bioreactor (CF MBR) when treating refinery wastewater

Transcription

1 Desalination 191 (2006) Performance of a crossflow membrane bioreactor (CF MBR) when treating refinery wastewater Muhammad Muhitur Rahman, Muhammad H. Al-Malack* Box 1150, King Fahd University of Petroleum & Minerals, Dhahran 31261, Saudi Arabia Tel (3) ; Fax: +966 (3) ; mhmalack@kfupm.edu.sa Received 21 March 2005; accepted 3 May 2005 Abstract The use of a crossflow membrane bioreactor (CF MBR) in treating wastewater discharged by a petroleum refinery was investigated. The performance of the CF MBR process was evaluated at MLSS concentrations of 5000 and 3000 mg/l. The process performance was measured in terms of the hydraulic efficiency as well as the COD removal efficiency. A laboratory-scale experimental set-up comprised mainly of tubular ceramic membranes, aeration tank, and circulation pump was used throughout the investigation. The results of the investigation showed that a COD removal efficiency of more than 93% was obtained at both MLSS values. The study also showed that hydraulic retention time did not have a significant effect on the system s performance. The relationship between permeate flux and crossflow velocity was found to be best described by a power relationship (J = kv n ) where constants k and n were affected by MLSS concentration. The cleaning mechanism investigation showed that cleaning the membrane with an acidic detergent, Superclean, with a ph value of about 1.5, produced the best results. Keywords: Crossflow filtration; Membrane bioreactor; Ceramic membrane; Refinery wastewater 1. Introduction The occurrence of oil-containing wastewater and the corresponding contamination of water sources by oil began with the production and utilization of petroleum and its products. The *Corresponding author. Presented at the International Congress on Membranes and Membrane Processes (ICOM), Seoul, Korea, August major problem of oily wastewater is associated with its proper disposal. The reclamation and reuse of wastewater is needed especially in the oil-producing arid regions due to water scarcity. With the development of membrane technology, membrane bioreactors have recently attracted great attention in the field of industrial wastewater treatment [1 5]. The performance of cross /06/$ See front matter 2006 Published by Elsevier B.V. doi: /j.desal

2 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) flow membrane bioreactor (CF MBR) processes is affected by environmental and operating conditions as measured by the quality and quantity of permeate [6,7]. Hydraulic retention time (HRT) plays an important role in the removal of pollutants in activated sludge processes coupled with membranes. In case of activated sludge systems, it is a common conviction that with the increase of HRT, the pollutant removal efficiency of the system increases. The conflicting scenario is also available where insignificant improvement was observed with the variation of HRT [8,9]. The MBR process was found to be useful when a long solid retention time is required and physical retention and subsequent hydrolysis are critical [4]. This process provides benefits over conventional activated sludge systems in terms of high effluent quality, reduced sludge wasting and production, reduced vulnerability to upsets, and improved biological degradation. It was observed that the MBR process improved the quality of degreasing solutions from surface refining processes in the metal-working industry. Permeate was found free of solid matter and hydrocarbon concentration was reduced by 85 90%. Compared to conventional biological regeneration, a five-fold increase in the volumetric biodegradation rate was achieved due to higher biomass concentration [10]. In a feasibility study of applying CF MBR to treat surfactants containing oil water emulsion, high efficiency (94 98%) in removing COD and TOC was observed at a MLSS concentration of 48 g/l [11]. In another study, hydrocarbon aggregation on bacterial flocs was observed leading to larger particles, which significantly increased the flux [12]. Fouling of membranes occurs with the application of MBR. Fouling depends on the characteristics of foulant and membrane materials. The major contribution to fouling is the different solute fractions resulting from activated sludge. Suspended solid constituents that contribute to membrane fouling consist mainly of bacterial flocs with a concentration depending upon the sludge age, colloids (polymers, fragments of lysed cells) and dissolved molecules [13]. Many inorganic elements dissolved in oily wastes can also play a significant role in fouling membranes. As a rule, mineral deposits are removed by acidic solutions, while organic compounds are removed by alkaline solutions [14]. Chemicals used as cleaning agent for fouling control of ceramic membrane include NaOH, HNO 3, H 2 O 2, Ultrasil 11, oxalic acid solution, citric acid solution, HCl solutions and saturated KHCO 3 solutions [15 18]. Several investigators studied the performance of CF MBR in terms of removal efficiencies and stability of flux with variation of different operating parameters for treating industrial wastewater [19,20]. In the literature, there is a lack of understanding of the interaction between the biological and filtration unit for treating refinery wastewater. Based on the above, the main objective of the study was to investigate the performance of the CF MBR process when treating refinery wastewater at different MLSS concentrations. The performance of the process was measured in terms of the hydraulic performance as well as the COD removal efficiency. Additionally, the effect of HRT on process performance and the membrane cleaning mechanism was investigated. 2. Materials and methods 2.1. Experimental system Fig. 1 shows a schematic diagram of the experimental set-up used throughout the investigation. It is comprised of two main parts: the crossflow membrane separation unit and the activated sludge bioreactor. The effective volume of the aeration tank was 20 l. The general characteristics of the membrane are shown in Table 1.

3 18 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) Fig. 1. Schematic diagram of the CF MBR system. Table 1 Characteristics of the membrane Configuration Hollow tubular Material Alumina (ceramic) Pore size, µm 0.2 Outer diameter, mm 10 Inner diameter, mm 7 Length, cm 5 20 Cross-sectional area, mm Total surface area, m Effective surface area, m Max. thermal stability, EC 120 Max. filtration pressure, bar 15 ph range Bioreactor feed The oily wastewater used in the investigation was collected from a petroleum refinery. The oil content and COD of oily wastewater were found to be mg/l and to mg/l, respectively. The COD was determined by a modified approach of the closed reflux titrimetric method. Essential nutrients such as glucose, peptone and yeast extract were added to the bioreactor Experimental Procedure and analytical methods Nutrients and oily wastewater were added to the bioreactor and mixed completely. It is worth mentioning that nutrients were supplied continuously while the oily wastewater was pumped intermittently for 2 min every 2 h. Consequently, the influent COD calculations were based on the mass loading per day rather than concentration. Initially the reactor was seeded using returned activated sludge that was collected from a municipal wastewater treatment plant. The circulation pump was used to pump the MLSS through the membrane separation unit. Permeate was collected in a permeate tank, while the concentrate was returned to the aeration tank. Permeate flux and inlet and outlet pressures were recorded. The fouled membrane was chemically cleaned to recover its permeability. The cleaning was achieved by using three different chemicals, namely Clorox (5.25% sodium hypochlorite), Persil (washing detergent) and Superclean (acidic detergent with a ph range of ). The chemicals were used individually and in combination with each other, followed by water rinsing. Flux recovery (J cleaned /J uncleaned ) was calculated

4 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) Table 2 Analytical methods of various parameters Parameter Technique Methods Turbidity Nephelometric SM-2130B ph Potentiometric SM-4500-H + MLSS Filtration 4.5 µm SM-2540D DO Oxygen Probe SM-4500-O G COD Closed reflux SM-5220C BOD 5 days SM-5210B TOC Combustion infrared SM-5310B Phenol Mass spectrometric SM-6420C Oil and grease Gravimetric EPA 1664 Ammonia Ion selective SM-4500-NH 3 D electrode Microbial Heterotrophic plate count SM-9215B based on the permeate flux of the cleaned and uncleaned membranes [14]. During the experimental period, samples from the bioreactor and permeate were collected periodically and analyzed for different physical and chemical parameters by standard analytical methods [21]. Table 2 shows the analytical methods used in this study. 3. Results and discussion 3.1. Hydraulic performance of the process The hydraulic performance was assessed by investigating the effect of operating conditions on flux rate of the membrane unit. In the study of a crossflow filtration system, selection of the circulation pump plays a significant role. The pump is responsible for maintaining sufficient transmembrane pressure (TMP) as well as flow, which are directly related to the variation of flux. At the beginning of this study (1st to 12th day), a pump with a cast iron impeller was used for circulation purposes. When not in operation, the impeller of the pump becomes corroded. In this case and when the pump was used, an enormous quantity of fine particulates was noticed to be produced and came in contact with the membrane, which resulted in rapid membrane fouling. This phenomenon can be observed in Fig. 2 where the variation of flux, TMP and the MLSS during the whole study period are shown. The data presented in this figure were recorded from the beginning of the operation. The above-mentioned reason might be the cause of lower initial maximum flux (65 l/m 2 /h on the first day) than the latter part of the study period (123, 123, 140 and 114 l/m 2 /h on the 13th, 26th, 29th and 83rd days, respectively) when the pump with a plastic impeller was used. The variation of flux showed a decreasing pattern that can be attributed to membrane fouling. When the flux dropped below the critical level (35 l/m 2 /h), the membrane unit had to undergo chemical cleaning. For the determination of MBR performance, HRT is a key issue. The system was operated at an average HRT of 21 h. During most of the experimental period, it was observed that the flux increased with the increase of pressure and vice versa. This phenomenon can be observed in Fig. 2 where, before the 47th day, the flux had a decreasing pattern that increased suddenly when the pressure increased from 12.1 to 18.6 psi. Similar phenomena were observed on the 40th and 74th days. Thus, in this case the flux can be described as pressure dependent, but some exceptions were also found. From the 83rd to the 92nd day the flux declined exponentially even though the pressure remained constant. This might happen solely because of fouling of the membrane, and the flux can be described as pressure independent. Throughout the study period, the TMP varied within a range of 11.5 to 24.0 psi.

5 20 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) Fig. 2. Variation of cumulative flux, TMP and MLSS with time. The biomass content of the reactor was measured twice a day. One measurement was to monitor the MLSS concentration present in the reactor and then calculate the volume of MLSS to be wasted in order to keep the suspended solids around 5000 or 3000 mg/l. The other measurement was taken after the wasted volume was replaced by tap water in order to check the adjusted MLSS concentration. For that reason, the variation of MLSS in Fig. 2 has a crisscross shape. In this regard it should be mentioned that the error associated for replacing the MLSS by tap water was not calculated in this study. In Fig. 2 a decreasing shape of MLSS variation is observed during the period of the 29th to the 34th day. The reason behind this is the excessive foam that occurred in the reactor. The foam was full of attached biomass and carried a considerable amount of MLSS out of the reactor. Sometimes, at the beginning of the run and after cleaning, the flux was found to decrease sharply with time, which is a classic phenomenon in membrane filtration, but later the flux started to recover slightly. This increase in flux could be attributed to the increase in temperature. This increase in temperature resulted in a reduction in viscosity of the fluid, thus allowing more fluid to pass through. To keep the aerobic condition in the bioreactor, air was supplied continuously and the dissolved oxygen (DO) was measured frequently using a DO probe. The DO level was never less than 4.0 mg/l, which shows that aeration provided in the reactor was in excess of the DO requirement. The ph of the aeration tank content was also within the limit of 6 to 8, which ascertains the suitable condition for biomass growth COD removal efficiency During the same set of experiments, the performance of the CF MBR was studied to assess the ability and stability of the system to provide the required COD removal efficiency. At the beginning of the study, the MLSS concentration was maintained at 5000 mg/l. The liquor

6 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) Fig. 3. Variation of influent and effluent organic mass loading at MLSS 5000 and 3000 mg/l, respectively, with time. was light brown in color and made up of dispersed non-flocculent particles. Fig. 3 represents the variation of influent and effluent mass loading at MLSS value of 5000 and 3000 mg/l, respectively. The influent mass loading was calculated as follows: Mass loading = [Concentration of oily waste (mg/l) flowrate of oily waste (l/d)] + [Concentration of nutrient (mg/l) flowrate of nutrient (l/d)] The influent mass loading presented in Fig. 3 is the average of the mass loading applied during a certain period of time needed for getting the steady-state condition at each adopted loading. When the steady-state period was obtained, the flowrate of oily waste was increased to get a higher mass loading. However, oil flowrate was insignificant to the total feed flow and did not affect the permeate flux and HRT. The variation of influent substrate can be clearly noticed by a steady horizontal line. To keep a resemblance, the effluent loading is also presented in average values. The sharp peaks in the permeate COD are due to the sudden increase in influent COD concentration. Occasionally it was found that the increase of effluent COD due to the change of influent loading was not rapidly responding and was apparent after one or two days. On the 42nd day, the effluent loading was noticed to be lower than the previous days, although the influent loading increased, and the effluent COD concentration remained the same. This was due to the calculation of the effluent mass loading with a lower volume of effluent (due to low flux) at this higher adopted loading stage. The influent mass loading at MLSS of 5000 mg/l varied from 24 to 67 g/day. The COD removal efficiency ranged from 82 to 97% with an average of 93%. Performance of the unit when operated with MLSS of 3000 mg/l was impressive with changing the influent mass loading from 30 to 65 g/day (Fig. 3). The average COD removal efficiency at this MLSS (3000 mg/l) was 94%. The food to microorganism (F/M) ratio during the whole study period was found to range between 0.20 and 1.15 d!1.

7 22 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) Effect of HRT on the performance of the process Experiments were carried out to investigate the effect of variation in HRT on the system performance in terms of flux stability and COD removal efficiency at different MLSS concentrations. To calculate the HRT, the following formula was used: HRT = [Reactor Volume] / [Permeate Flux Membrane Surface Area] The experiments were conducted at three HRT values. After finishing each experiment, the membranes were cleaned to restore the flux. The experiment began with the MLSS concentration of 5000 mg/l. The corresponding average HRTs were 17, 22, and 34 h, and CFVs were 3.24, 2.69, and 2.21 m/s. To find the COD removal performance of the system, various organic mass loadings were applied under different HRT conditions. The system was put in with an average influent mass loading of 42.75, and g/day that resulted in the effluent mass loading of 2.66, 2.45 and 1.71 g/day at the HRTs of 17, 22 and 34 h, respectively. Fig. 4 presents the COD removal efficiency for different HRTs at the MLSS of 5000 mg/l. Although the highest removal efficiency (95%) was observed at the HRT of 34 h, the removal efficiencies at HRTs of 22 and 17 h were very close (94%) to this value. Therefore, it can be postulated that HRT did not affect the system in COD removal efficiency, which varied in a narrow range of 93 95%. After finishing the experiment at the MLSS value of 5000 mg/l, the biomass was treated to maintain an MLSS concentration of 3000 mg/l. At this MLSS concentration, the HRTs were 16, 20, and 33 h and CFVs were 3.39, 2.76, and 2.25 m/s. The system was fed with an average influent mass loading of 42.75, and g/d resulting in the effluent mass loadings of 2.61, 2.26 and 1.84 g/d, respectively. Fig. 5 shows the COD removal efficiency for different HRTs at a MLSS of 3000 mg/l. The highest removal efficiency (95%) was observed at the highest HRT of 33 h. The removal efficiency at rest of the HRT values remained same as previously (94%). It was observed that COD removal efficiency is independent of HRT at different MLSS concentrations in this study. This might happen because the adopted HRTs were too close to each other to demonstrate the variation in COD removal efficiency, and the experiments ran for a short time to allow degradation of the high-molecular-weight compound derived from the oily waste. Fig. 4. COD removal efficiency with time at different HRTs at MLSS 5000 mg/l.

8 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) Fig. 5. COD removal efficiency with time at different HRTs at MLSS 3000 mg/l. To investigate the effect of CFV on the hydraulic performance of the process, flux was correlated with CFV described by the following power relation [22]: J = kv n where J is the permeate flux (l/m 2 /h), V is the crossflow velocity (m/s), and k and n are constants. To find k and n, the steady-state flux values were used at the corresponding MLSS concentration and CFV value. The values of (log J) were plotted against (log V), and by linear regression k and n were determined. The values of k were and 11.03, while the values of n were 1.75 and 1.6 at the MLSS concentrations of 3000 and 5000 mg/l, respectively. It can be seen that as the MLSS concentration increased, the values of k and n decreased. Since two MLSS concentrations were investigated, the relationship between the MLSS concentration and the constants k and n was not developed. It is worth mentioning that the values of constants k and n may only be used with this type of membrane and under similar conditions Permeate quality Throughout the study period, different parameters indicating the quality of permeate were Table 3 Summary of permeate quality parameters Parameter Range Mean value BOD, mg/l COD, mg/l TOC, mg/l Phenol, mg/l Oil and grease, mg/l Ammonia, mg/l Turbidity, NTU Heterotrophic plate count, CFU/ml examined. The summary of organic constituent analyses is shown in Table 3. To find the viable bacterial count in the reactor as well as permeate, a heterotrophic plate count method was adopted. Following the incubation, two types of surviving colonies were found, i.e., large spongy-white colonies and small whitish-yellow colonies. On average, one log reduction in permeate colony forming units was observed during the experimental period Membrane fouling control At the beginning of the study, the cleaning procedure was attempted by the use of Clorox

9 24 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) Fig. 6. Variation of flux before and after cleaning with Persil Clorox and Superclean. Fig. 7. Summary of the cleaning procedure with Persil Clorox and Superclean. only. The cleaning process continued for about 10 h to restore the flux of the new membrane. Although cleaning the membrane with Clorox restored the flux significantly, the cleaning time was not satisfactory. To obtain a reasonable cleaning time it was decided to use Persil and Clorox in different sequences followed by water rinsing, but this did not improve the cleaning time. Another cleaning agent, Superclean, followed by a backwash, was tested to improve the cleaning time and the flux restoration. It can be observed in Fig. 6 that washing the membrane with Superclean caused the peak flux to be constant for around 27 h, thus establishing a wider peak than the previous cleaning agents (Persil and Clorox). In relation to cleaning time,

10 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) the Superclean needed 85% less time than that of the other cleaning agents (Fig. 7). Therefore, chemical washing with Superclean was regarded as the best solution for fouling control and was adopted as the cleaning technique throughout the study period. 4. Conclusions The performance of a laboratory-scale crossflow MBR treating wastewater discharged by a petroleum refinery was investigated. The study showed that a COD removal efficiency of more than 93% was obtained, and removal efficiency was not significantly affected by the increase in MLSS concentration. The flux CFV relationship was best described by a power relationship whose constants were found to be affected by MLSS concentration. The study also showed that COD removal efficiency was independent of HRT. Chemical cleaning of the membrane with the acidic detergent Superclean, followed by a backwash, was found to produce the best results in terms of cleaning time and recovered flux. Acknowledgements The authors would like to thank King Fahd University of Petroleum & Minerals (Dhahran, Saudi Arabia) for financial and technical support. References [1] F. Malpei, L. Bonomo and A. Rozzi, Feasibility study to upgrade a textile wastewater treatment plant by a hollow fiber membrane bioreactor for effluent reuse, Water Sci. Technol., 47(10) (2003) [2] D.M.Stamper, M. Walch and N.R. Jacobs, Bacterial population changes in a membrane bioreactor for gray water treatment monitored by denaturing gradient gel electrophoretic analysis of 16S RNA gene fragments, Appl. Environ. Microbiol., 69(2) (2003) [3] D.S. Kim, J.S. Kang and Y.M Lee, Microfiltration of activated sludge using modified PVC membranes: Effect of pulsing on flux recovery, Sep. Sci. Technol., 38(3) (2003) [4] M.D. Knoblock, P.M. Sutton, P.N. Mishra, K. Gupta and A. Janson, Membrane biological reactor system for treatment of oily wastewaters, Water Environ. Res., 66(2) (1994) [5] K.D. Zoh and M.K. Stenstrom, Application of a membrane bioreactor for treating high explosives process wastewater, Proc. 71st Water Environment Federation Annual Conference and Exposition, Orlando, FL, 1998, pp [6] Y.B. Fan, J.S. Wang and Z.C. Jiang, Test of membrane bioreactor for waste water treatment of a petrochemical complex, J. Environ. Sci. (China), 10(3) (1998) [7] A.H. Kuljian, J.R. Porter and T. Chen, Biodegrading machining wastewater: a compressor manufacturer uses a membrane biological reactor to treat mediumstrength wastewater, Ind. Wastewater, 6(6) (1998) [8] J.C. Campos, R.M.H. Borges, A.M. Filho, N.R. Oliveira and G.L. Sant Anna, Oilfield wastewater treatment by combined microfiltration and biological processes, Water Res., 36 (2002) [9] J.H. Tay, J.L. Zeng and D.D. Sun, Effects of hydraulic retention time on system performance of a submerged membrane bioreactor; Sep. Sci. Technol., 38(4) (2003) [10] C. Blöcher, U. Bunse, B. Sessler, H. Chmiel and H.D. Janke, Continuous regeneration of degreasing solutions from electroplating operations using a membrane bioreactor, Desalination, 162 (2004) [11] W. Scholz and W. Fuchs, Treatment of oil-contaminated wastewater in a membrane bioreactor, Water Res., 34(14) (2000) [12] S. Elmaleh and N. Ghaffor, Upgrading oil refinery effluents by cross-flow ultrafiltration, Water Sci. Technol., 34(5) (1996) [13] L.J. Defrance, Y. Michael, B. Gupta, P. Paullier and V. Geaugey, Contribution of various constituents of activated sludge to membrane bioreactor fouling, Bioresource Technol., 73 (2000) [14] J. Lindau and A.S. Jonsson, Cleaning of ultrafiltration membrane after treatment of waste water, J. Membr. Sci., 87 (1994)

11 26 M.M. Rahman, M.H. Al-Malack / Desalination 191 (2006) [15] W.P. Bedwell, S.F. Yates and I.M. Brubaker, Crossflow microfiltration: Fouling mechanism studies, Sep. Sci. Technol., 23(6 7) (1988) [16] Z. Yijiang, Z. Jing, L. Hong, X. Nanping and S. Jun, Fouling and regeneration of ceramic microfiltration membranes in processing acid wastewater containing fine TiO 2 particles, J. Membr. Sci., 208 (2002) [17] Q. Gan, J.A. Howell, R.W. Field, R. England, M.R. Bird and M.T. McKechinie, Synergetic cleaning procedure for a ceramic membrane fouled by beer microfiltration, J. Membr. Sci., 155(2) (1999) [18] P. Heineman, J.A. Howell and R.A. Bryan, Microfiltration of protein solutions: Effect of fouling on rejection, Desalination, 68 (1988) [19] I. Daubert, M. Mercier, C. Maranges, G. Goma, C. Fonade and C. Lafforgue, Why and how membrane bioreactors with unsteady filtration conditions can improve the efficiency of biological processes, Adv. Membr. Technol., 984 (2003) [20] P.M. Sutton and P.N. Mishra, The membrane biological reactor for industrial wastewater treatment and bioremediation, Proc. International Symposium on the Implementation of Biotechnology in Industrial Waste Treatment and Bioremediation, Grand Rapids, MI, 1992, pp [21] APHA, Standard Methods for the Examination of Water and Wastewater; 19th ed., American Public Health Association, Washington, DC, [22] J. Murkes and C.G. Carlsson, Crossflow Filtration: Theory and Practice, Wiley, New York, 1988.