- Email: [email protected]
Interpretation of fouling characteristics of ultrafiltration membranes during the filtration of membrane bioreactor mixed liquor
Journal of Membrane Science 264 (2005) 151–160
Interpretation of fouling characteristics of ultrafiltration membranes during the filtration of membrane bioreactor mixed liquor Tae-Hyun Bae, Tae-Moon Tak ∗ School of Biological Resources and Materials Engineering, Seoul National University, San 56-1, Shinlim-dong, Kwanak-gu, Seoul 151-921, South Korea Received 11 February 2005; received in revised form 14 April 2005; accepted 19 April 2005 Available online 9 June 2005
Abstract In this study, ultrafiltration (UF) membranes with various pore sizes were prepared by the phase inversion method, and their fouling characteristics were investigated with membrane bioreactor (MBR) mixed liquor. MBR sludge was fractionated into three parts, suspended solids, colloids and solutes, and their individual contributions to membrane fouling were quantified in order to asses the mechanisms by which fouling occurs. Cake layer formation on the membrane surface constituted the main fouling mechanism and the incidence of irreversible fouling was so small as to be negligible. The degree of fouling correlated very strongly with membrane permeability. This implies that hydrodynamic conditions are important factors affecting membrane fouling. The fouling contribution of each sludge fraction appeared to depend on particle size, as both permeation drag and back transport velocity are particle size-related functions. Solutes played a significant role in the initiation of cake layer formation, because they were deposited onto the membrane surface and pore wall immediately upon initial filtration, but were dislodged only in small amounts by cross flow. Suspended solids were consistently deposited onto the membrane surface, until flux reached a steady state and colloids exhibited characteristics commensurate with an intermediated state between solutes and suspended solids. Suspended solids were, in fact, found to be the main contributor to the fouling process. However, the relative contribution of each of the sludge fractions to membrane fouling varied with the permeability of the membrane, and also with the hydrodynamic condition. © 2005 Elsevier B.V. All rights reserved. Keywords: Membrane bioreactor (MBR); Membrane fouling; Ultrafiltration; Suspended solids; Colloids; Solutes
1. Introduction Membrane bioreactors (MBRs) have been widely used in advanced wastewater treatment schemes. The substitution of a secondary clarifier with membranes enables the total rejection of suspended solids without a settling process, and offers numerous advantages over the conventional biological process [1]. In particular, the absolute rejection of sludge by the membrane makes it possible to overcome a problem of dependence on settleability [2]. Furthermore, membrane separation enables a significant increase in the concentration of mixed liquor suspended solids (MLSS) in the bioreactor, thus reducing its size for a given sludge [1]. ∗
Corresponding author. Tel.: +82 2 880 4621; fax: +82 2 873 2285. E-mail address: [email protected] (T.-M. Tak).
0376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2005.04.037
However, wide acceptance of the MBR system in the wastewater treatment process has been hindered by problems of membrane fouling, which decreases permeate flux during the filtration of activated sludge. Membrane fouling increases the operation costs of MBR processes, due to the necessity of cleaning and replacing fouled membranes. Because of this, various techniques have been utilized in order to reduce membrane fouling in MBR processes. Since MBR mixed liquor includes living microorganisms and their metabolites, the fouling mechanism is even more complex than that of conventional membrane separation processes. Many researchers have focused their attentions on the fouling mechanism in MBR systems. It has been reported that MLSS concentration [3–5], sludge characteristics [6–9], and the amount and composition of microbial products [6,9–11] are the key factors in membrane fouling. In addition,
152
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
membrane fouling may be directly influenced by properties of the membrane [7,12] and the hydrodynamic conditions of the membrane process, most notably, permeation drag and back transport [13–16]. Several studies have attempted to quantify the degree and character of fouling caused by each fraction of the activated sludge, namely, suspended solids, colloids and solutes. Wisniewski et al. reported that solutes play a major role in membrane fouling in the MBR process [17]. Another study presented evidence that suspended solids cause the greatest amount of fouling among the sludge fractions [18]. However, another study reported that colloids are the main contributors to membrane fouling [19]. Clearly, such a panoply of conflicting results can only deepen the controversy. However, by reviewing most of the previous relevant literature, we can extend the hypothesis that fouling contributions from various fractions of the activated sludge may be variable, and change according to membrane properties, hydrodynamic conditions and the physiological properties of the biomass. This present work is focused on the effects of membrane pore size, permeability and filtration time on membrane fouling, and should be considered the first in a series of inquiries into this issue. In this study, ultrafiltration (UF) membranes of various pore sizes were prepared using the phase inversion method, and their fouling characteristics were assessed by testing them with activated sludge mixed liquor from an MBR plant. The objectives of this study were to investigate the roles of various sludge fractions in membrane fouling, and also to evaluate the effects of membrane pore size, permeability and filtration time on the behaviors of the aforementioned sludge constituents. The results of this study would help us elucidate the fouling mechanism of UF membranes during the filtration of MBR mixed liquor.
2. Materials and methods 2.1. Membrane preparation and characterization Membranes used in this study were prepared using phase inversion method [20]. The membrane material was cellulose acetate (CA; CA398-3, Eastman-Kodak), and N-methyl-2pyrrolidone (NMP, Aldrich) and acetone (Aldrich) were used as the solvent and co-solvent, respectively. In order to change the membrane pore size, the solvent/co-solvent ratio of the dope solution was controlled. The dope solution was kept at room temperature for 24 h, and then cast on a polyester non-woven fabric with a casting knife having 200 m thickness. The nascent membrane was evaporated at 25 ± 1 ◦ C and 65 ± 5% relative humidity for 30 s, and then immersed in an 18 ± 1 ◦ C deionized water coagulation bath. Pure water flux was measured under a transmembrane pressure (TMP) of 100 kPa at 20 ± l◦ C after the flux reached steady state. The molecular weight cut off (MWCO) was characterized by the rejection performance of
Table 1 The composition of standard wastewater Components
Concentration (mg/l)
Glucose (NH4 )2 SO4 KH2 PO4 MgSO4 ·7H2 O CaCl2 ·2H2 O FeCl3 ·6H2 O NaHCO3
400 220 23 50 10 4 50
1000 mg/l polyethylene glycol (PEG) or polyethylene oxide (PEO) aqueous solution [21]. Rejection was calculated by the following equation: Cper R (%) = 1 − × 100 (1) Cfeed where Cper is the concentration of the permeate and Cfeed the feed concentration. Since new polymeric membranes can be compacted during filtration by applied pressure, the flux can be declined without membrane fouling. In order to remove the compaction effect on flux decline, all membranes had been previously filtered by pure water until flux reached steady state, prior to the fouling tests. 2.2. Activated sludge The activated sludge used in this study had been cultivated in a submerged MBR plant, with synthetic substrate. Standard wastewater composition is presented in Table 1. Glucose, (NH4 )2 SO4 and KH2 PO4 were used as carbon, nitrogen and phosphorus sources, respectively. Additional nutrients and alkalinity (NaHCO3 ) were also supplied to the reactor. The MBR plant, which was operated by the authors, was a membrane coupled sequencing batch reactor for biological nutrient removal study. MLSS concentration was maintained at levels between 3100 and 3400 mg/l. Sludge retention time was about 20 days, and the organic loading rate was 0.16–0.17 gCOD/gMLSSday. 2.3. Experimental system and analysis Flux decline behavior was measured using membrane cells, the membrane surface area of which was 18.1 cm2 . A schematic diagram of the filtration system is shown in Fig. 1. Filtration tests were performed at 100 kPa and 20 ± 2 ◦ C. Crossflow velocity was controlled at 1.2 m/s, and the flow rate was 2.5 l/min. Four membranes were tested simultaneously with the same feed, and air was supplied to the biomass in the feed tank during the filtration test. Flux was determined by weighing the permeate on a top-loading balance at timed intervals. Stirred batch cell (Amicon) was used for dead end filtration experiments under different conditions. MLSS concentration was measured in accordance with Standard Methods [22]. The particle size of the activated
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
153
Fig. 1. The schematic diagram of MBR mixed liquor filtration system.
sludge was measured with a Multisizer IIe (Coulter). The COD of the MBR mixed liquor and membrane permeate was measured by the absorbance method, using a Fotometer AL282 (AQUALYTIC). 2.4. Membrane fouling analysis The degree of membrane fouling was quantitatively calculated, using the resistance in series model [6,20–21]: Rt =
PT ηJ
Rt = Rm + Rc + Rf
Rf =
PT ηJw PT − Rm ηJw
PT − Rm − Rf Rc = ηJAS
In this study, we assumed that the activated sludge consisted of solutes, colloids and suspended solids, and that these components independently contribute to membrane fouling, that is to say, we neglected any coupling or synergistic effects which might occur among the components. Therefore, the resistance of the activated sludge can be considered to be equal to the sum of the resistances of the suspended solids, colloids and solutes:
(2)
RAS = Rss + Rcol + Rsol
(3)
where RAS is the resistance of the activated sludge, Rss the resistance of the suspended solids, Rcol the resistance of the colloids and Rsol the resistance of the solutes. The resistance of the activated sludge could be measured from the activated sludge filtration, and calculated by the following equation:
where J is the permeation flux (m3 /m2 s), PT the TMP (Pa), η the viscosity of the permeate (Pa s), Rt the total filtration resistance (m−1 ), Rm the membrane resistance (m−1 ), Rc the cake layer resistance (m−1 ) and Rf the fouling resistance (m−1 ). The intrinsic membrane resistance (Rm ), the cake resistance by cake layer formed on the membrane surface (Rc ) and the fouling resistance caused by pore plugging and irreversible adsorption of foulants onto the membrane pore wall or surface (Rf ) can be calculated using the following equations: Rm =
2.5. Fouling contribution of sludge fractions
(4) (5)
RAS = Rt − Rm =
where JAS is the flux of activated sludge at steady state, Jw the initial water flux and Jw the final water flux after removing the cake layer by flushing with tap water.
(8)
Colloids and solutes could be obtained by extracting the supernatants generated after 4 h of gravitational sedimentation of the activated sludge mixed liquor. In fact, the colloids in the present study included large non-settling particles, which were derived mainly from the breakage of sludge flocs. The sum of the resistances of the colloids and solutes could be determined by a filtration test of the supernatant: Rcol + Rsol =
(6)
PT − Rm ηJAS
(7)
PT ηJsup
(9)
where Jsup is the permeate flux of the supernatant. Finally, the soluble fraction was acquired via the filtration of the supernatant with a 0.45 m microfiltration membrane (Millipore). The resistance of the solutes could be measured
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
154
Table 2 The compositions of the dope solutions and the characteristics of the UF membranes Membranes
CA-1 CA-2 CA-3 CA-4 a
Composition (wt.%) (CA/NMP/acetone)
(15/35/50) (15/55/30) (15/75/10) (15/85/0)
PWFa (l/m2 h)
34.1 193.0 224.3 282.2
Rejection (%) 10000
20000
35000
100000
200000
300000
52 7 8 6
75 11 9 11
89 19 16 17
95 86 76 50
98 96 91 71
99 99 96 93
PWF: pure water flux.
by the filtration of the soluble fraction: Rsol =
PT ηJsol
(10)
where Jsol is the permeate flux of the soluble fraction. RAS , Rss , Rcol and Rsol were calculated by Eqs. (8)–(10).
3. Results and discussion 3.1. Characterization of the membranes It has been established that the solvent–water exchange rate in the coagulation bath can alter membrane pore size and permeability. In general, if the exchange rate is delayed, pore size and permeability decrease, and the top layer, which forms on the support layer, becomes denser and thicker [20]. Since the addition of acetone can serve to slow the rate of solvent–water exchange, we controlled the NMP/acetone ratios in order to vary the pore sizes of the UF membranes [7]. Table 2 shows the pure water flux and solute rejection rate of the membranes. Permeability of the membranes increased with NMP content in the dope solutions from membranes CA1 to CA-4. Membrane pore size also increased with NMP concentration, thus the MWCO of CA-1, CA-2, CA-3 and CA-4 were 35,000, 100,000, 200,000 and 300,000 Da, respectively. 3.2. Particle size distribution and COD measurement Generally, the aeration intensity of submerged MBR systems is much higher than that of conventional activated sludge processes due to the necessity of control of membrane fouling. Thus, sludge particles can be broken up by shear stress resulting from intensive aeration. Fig. 2(a) shows the particle size distribution of the MBR mixed liquor. Most particles existed in a size range of between 10 and 40 m, and the mean particle size was 25 m, which was much lower than that of conventional activated sludge. The particle size range of the supernatant is shown in Fig. 2(b), and the mean diameter there was 9 m. Relatively large particles, generated by shear stress, increased both the concentration and mean particle diameter of supernatant. As mentioned above, those non-settling large particles were classified, in this study, as colloids. Soluble COD (SCOD) of feed and membrane permeates are shown in Fig. 3. More than 50% of SCOD was removed
Fig. 2. The particle size distributions of activated sludge and supernatant. (a) Activated sludge; (b) supernatant.
by the UF membranes. The removal rate increased slightly, with decreasing membrane pore size. In this study, the carbon source used in the substrate was glucose, which could easily pass through the membranes. This result indicated that the mixed liquor contained macromolecules, derived from microbial metabolism, and that they were rejected by the UF membrane. 3.3. Flux decline during the filtration of MBR mixed liquor Flux decline behaviors of the membrane are illustrated in Fig. 4. All filtration tests were performed until flux reached
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
155
case of the other membranes, however, membrane fouling occurred and flux declined to steady state at a lower level than the initial. Interestingly, all membranes, except CA-1, exhibited the similar steady state flux values for the filtration of the supernatant and the soluble fraction. This indicated that the sludge components were consistently deposited onto the membrane surfaces during filtration, until total filtration resistances reached the same level. 3.4. Filtration resistances of the membranes
Fig. 3. The soluble COD of feed and membrane permeate.
stable state. The rate of flux decline was very high during initial filtration, but decreased with time, finally reaching steady state. The amount of flux decline increased with both membrane pore size and permeability from CA-1 to CA-4. Although the initial flux characteristics of the membranes were quite different, steady state flux values were practically uniform during the filtration of the activated sludge. Membrane CA-1 exhibited no significant decline in flux during filtration of both the supernatant and the soluble fraction. In the
Filtration resistances, calculated from activated sludge filtration, are presented in Table 3. As mentioned above, total resistances were equal for all membranes. Since intrinsic membrane resistances vary inversely with pure water flux, they showed a tendency to decrease, from CA-1 to CA-4. However, the resistance of the cake layer increased with membrane permeability, due to the large amount of sludge deposition on the membrane surface. Cake resistance was observed to make the highest contribution to flux decline in all membranes. In particular, from CA-2 to CA-4, the contribution of cake resistance to fouling reached more than 90%. Thus, it can be concluded that the main fouling mechanism, in this study, was the formation of the cake layer, which results from the deposition of sludge components on the membrane surface.
Fig. 4. Behaviors of the flux declines of UF membranes during the filtration of various sludge fractions (MLSS: 3100–3400 mg/l, SS of supernatant: 90 mg/l and COD of soluble fraction: 23 mg/l). (a) CA-1, (b) CA-2, (c) CA-3 and (d) CA-4.
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
156
Table 3 The filtration resistances of the UF membranes
CA-1 CA-2 CA-3 CA-4
Rm (×1012 m−1 )
Rf (×1012 m−1 )
Rc (×1012 m−1 )
Rm /Rt (%)
Rf /Rt (%)
Rc /Rt
10.5 1.9 1.6 1.3
7.8 1.1 0.8 0.6
15.5 32.2 35.0 38.0
31 5 4 3
23 3 2 2
46 92 94 95
Irreversible fouling resistance increased as the permeability of the membrane decreased. Pore clogging and permanent adsorption of macromolecules are considered to constitute a main cause of irreversible fouling of UF membranes. It was reported that cake layers forming on the membrane surface act as barriers which protect membrane surfaces and pores [5]. Thus, irreversible fouling of the membrane increased as its cake layer resistance decreased. As a result, Rf shows a gradual decrease from CA-1 to CA-4. 3.5. Contributions of sludge constituents to membrane fouling Filtration resistances of suspended solids, colloids and solutes were calculated for all membranes. Filtration resistances of sludge constituents versus permeability of membrane measurements are presented in Fig. 5. From membranes CA-1 to CA-4, RAS exhibited a linear relationship to membrane permeability, and our results showed there to be a very high correlation coefficient (R = 0.99) between the two parameters. This higher correlation was also observed for individual sludge fractions: 0.99, 0.98 and 0.99 for Rss , Rcol and Rsol . As all membranes were filtered under identical conditions, membrane pore size and permeation drag must necessarily be the main factors operant in membrane fouling. The correlation between filtration resistance and membrane permeability prove that increases in filtration resistance are mainly
Fig. 5. The relationship between the resistances of sludge constituents and membrane permeability.
affected by hydrodynamic conditions, such as permeation drag and back transport, and also that effect of pore size can be considered to be negligible in the UF range from MWCO 35,000 to 300,000. Generally, as the pore size of a membrane increase, fine particles and macromolecules can more readily plug the pore and be adsorbed onto the pore walls [13]. In this study, however, the formation of the cake layer proved to be the main fouling mechanism, and it grew rapidly from the initial period of filtration. As mentioned above, the cake layer can function as a protective barrier. Due to this feature of the cake layer, the influence of pore size on membrane fouling was found to be so small to be negligible in this study. All sludge constituents are attached to the membrane surface by permeation drag, but are also detached by back transport. Permeation drag, which is generated by flux, increased with operation flux and particle size. Back transport in a cross flow system consists of Brownian diffusion, inertial lift and shear-induced diffusion [23]. Although Brownian diffusion, the driving force of which is the concentration gradient between the membrane surface and the bulk side of feed, tends to decrease as particle size increases, the other forces, which are generated by cross flow, tend to increase with cross flow velocity and particle size [23]. MBR mixed liquor contains large sludge flocs as a major fraction, although it contains a small amount of solutes. Furthermore, as these solutes consist mainly of high molecular weight macromolecules, the influence of Brownian diffusion is also extremely small, or negligible. Therefore, it can be concluded that the back transport velocity of the particles increased with their size. As shown in Fig. 5, regardless of membrane permeability, the contribution of Rss to total resistance was higher than that of any other sludge constituent, because suspended solids comprised the main fraction of the activated sludge, and played a key role in the deposition and formation of the cake layer. Although the resistances of all sludge constituents were proportional to the water flux of the membrane, the degree of this proportion varied among the sludge constituents. The slopes of the lines in Fig. 5 are 0.035, 0.019 and 0.009 for Rss , Rcol and Rsol , respectively. This indicates that the influence of permeability on membrane fouling increased with the particle size of foulants at the same cross flow velocity. On the other hand, fouling control appears to be a simpler matter for suspended particles than for other components, because back transport velocity, which is generated by cross flow, also increases with particle size.
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
157
Fig. 6. The contributions of each sludge fraction to membrane fouling according to relative filtrate volume. (a) CA-1, (b) CA-2, (c) CA-3, and (d) CA-4.
3.6. Deposition of sludge constituents on membrane surface according to filtration time
lutes. They take a shorter time than suspended solids but a longer time than solutes to reach steady state.
The filtration resistances of each sludge constituent, according to the relative filtrate volume, are shown in Fig. 6. All membranes exhibit a similar pattern in terms of resistance increase behavior, although filtration resistance values were different. Solutes are rapidly deposited onto the membrane surface, and then filtration resistances reach nearly steady state at an early stage of filtration. This result is consistent with the previous studies of other researchers [24]. Rapid solute deposition is induced by the inherent characteristics of soluble microbial products in the sludge mixed liquor. Since soluble macromolecules have a low back transport velocity and strongly adsorb onto the membrane surfaces, solutes are readily deposited on the membrane surfaces by permeation drag, and not readily detached by shear force. On the other hand, Rss gradually increased until the flux reached steady state. Although a huge amount of suspended solids were deposited onto the membrane surfaces by permeation drag, those particles are readily removed by cross flow due to their larger size. Thus, it takes the longest time for suspended solids to reach an equilibrium between back transport and permeation drag. Colloids and fine particles exhibited intermediate characteristics between suspended solids and so-
3.7. Hypothetical mechanism for initial cake layer formation According to the results described in the previous section, a mechanism underlying the formation of the cake layer can be proposed. A schematic illustration of cake formation according to flux decline is presented in Fig. 7. 3.7.1. Phase-1 Rapid flux decline occurs during initial filtration. All sludge particles are deposited onto the membrane surface. Large sludge flocs attached to the membrane surface can be removed by cross flow due to their size, while solutes are deposited onto the membrane surface and pore wall, and cause irreversible fouling. The influence of colloids is insignificant compared to the influences of large sludge flocs and solutes, since the amount of colloids are much smaller than the amount of suspended solids, and also because some of the attached colloidal particles are removed by cross flow. Consequently, all sludge particles take part in the formation of the cake layer during the initial stage, thus inducting rapid flux decline. The solutes are the main
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
158
Fig. 7. The schematic illustration of initial cake layer formation.
factor in the flux decline and induce irreversible membrane fouling. 3.7.2. Phase-2 Rate of flux decline is alleviated compared to that of Phase1. In this stage, sludge constituents are accumulated, not only on the membrane surface, but also on previously deposited layers. Permeation drag of this stage has significantly declined, because of the initial flux decline in Phase-1, thus the amount of particle deposition decreases from the rates associated with Phase-1. Permeation drag and back transport of solutes reach equilibrium state in this stage, and colloids and suspended solids contribute to an increase in filtration resistance. While the influence of colloids on fouling gradually decreases, suspended solids are continuously deposited onto the surface. Finally, the resistance of the colloids increases nearly to steady state at the end of this stage, and the rate of flux decline is depressed even further. Most fouling occurring in this stage is reversible fouling, which can be ameliorated by cleaning. 3.7.3. Phase-3 The fouling rate decreases further and reaches steady state. Most deposited particles in this stage are large sludge flocs. Particle deposition is significantly reduced from the levels associated with Phase-1 or Phase-2, due to dramatically reduced
permeation drag. Another important factor influencing flux decline is the compaction of the cake layer. The deposited cake layer can be compacted by TMP. Compaction of the cake layer also increases filtration resistance and decreases permeability. Compaction takes place throughout the whole process. However, its effects become far more relevant during Phase-3 than they were during Phases-1 and 2, because the largest amounts of particles are now being pressurized for a longer time. When the permeation drag of the suspended solids decreases until it becomes equal to the back transport velocity, flux reaches steady state. 3.8. Discussion — relative contribution of each sludge fractions to membrane fouling Results from pervious studies and the present work are summarized in Table 4 [9,17–19]. Results were quite different from study to study. Bouhabila et al. speculated that those differences could be derived from the nature of the substrate, physiological characteristics of the biomass, fractionation method and membrane materials [19]. In this study, the relative fouling contributions of each sludge fraction varied according to membrane permeability. The main fouling contributor, in this present work, proved to be suspended solids, which is consistent with the reports of Lee et al. [9] and Defrance et al. [18], but the relative
Table 4 The relative contributions of various sludge fractions to membrane fouling in MBR systems (%) Fraction
Wisniewski
Defrance
Bouhabila
Lee
Our results CA-1
Suspended solids Colloids Solutes a
24 24 52
The contribution of supernatant to fouling.
65 30 5
24 50 26
63–71 29–37a 29–37a
83 4 13
CA-2
CA-3
CA-4
76 10 14
74 13 13
72 14 14
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
Fig. 8. Flux decline behaviors during the dead end filtration of membrane CA-2.
contribution of suspended solids to fouling decreased with membrane permeability. On the other hand, the relative contribution of the colloids increased with membrane permeability, because their back transport velocity was relatively low compared to that of the suspended solids. Considering the results of this study, it does not seem overly speculative to surmise that the relative fouling contributions of each sludge fraction can also be influenced by hydrodynamic conditions. In this study, since the membrane was operated at a high flux and a low cross flow velocity, the relative contribution of suspended solids was very high, and the contributions of the colloids and solutes were relatively low. Because the deposition of suspended solids is easily controlled by the cross flow, their relative contribution to fouling can be decreased by ensuring conditions of low flux and high shear rate. This was proved by performing a simple dead end filtration test. The dead end filtration tests were carried out at a TMP of 20 kPa and a stirring rate of 600 rpm, using stirred batch cell. Results from the filtration of membrane CA-2 are presented in Fig. 8 and Table 5. Fig. 8 shows that the flux can be gradually decreased, nearly to the steady state flux values of activated sludge by soluble fraction. This implies that the deposition of large particles can be controlled by the creation of conditions of low permeation drag and high shear rate, but more solutes will then be directly deposited onto the membrane surfaces and pore walls without the protective effects of the cake layer. As a result, as shown in Table 5, the main fouling contributor shifted from the suspended solids to the solutes. It should be noted that hydrodynamic control is powerful method for the mitigation of macroscopic fouling, but it cannot prevent the deposition of soluble macromolecules. Table 5 The filtration resistances and relative fouling contributions of each sludge fraction during dead end filtration of membrane CA-2 Resistance (×1012 m−1 ) R/RAS (%)
Suspended solids
Colloids
Solutes
1.1 24
0.4 9
3.1 67
159
As discussed in the previous section, the contributions of various sludge fractions to fouling can be affected by filtration time, as well as membrane permeability and hydrodynamic conditions. Many other factors also may affect the fouling contributions of the various sludge fractions. Thus, it is futile to compare relative fouling contributions from study to study, because varying experimental conditions clearly produce discrepant results. However, a series of experiments with controlled conditions could result in a uniform and relevant set of results. With regard to a concerted investigation into membrane fouling, this seems the most useful approach. Therefore, we will conduct further studies, focusing on other factors affecting membrane fouling, such as membrane material, system configuration and hydrodynamic conditions, and the physiological properties of the biomass.
4. Conclusions In this study, ultrafiltration membranes with various pore sizes were prepared by the phase inversion method, and their fouling characteristics were investigated with fractionated MBR mixed liquor. Cake layer formation on the membrane surface constituted the main fouling mechanism and the incidence of irreversible fouling was so small as to be negligible. The degree of fouling correlated very strongly with membrane permeability. The fouling contribution of each sludge fraction appeared to depend on particle size, as both permeation drag and back transport velocity are particle size-related functions. Solutes played a significant role in the initiation of cake layer formation. Suspended solids were consistently deposited onto the membrane surface, until flux reached a steady state and colloids exhibited characteristics commensurate with an intermediated state between solutes and suspended solids. Suspended solids were, in fact, found to be the main contributor to the fouling process. However, the relative contribution of each sludge fraction to membrane fouling varied with the permeability of the membrane, and also with the hydrodynamic condition.
5. Acknowledgement This work was supported by the Seoul National University Foundation (Project No. 02-02-31-4).
References [1] J. Manem, R. Sanderson, Water Treatment Membrane Process, AWWA Research Foundation, McGraw-Hill, 1996, Chapter 17. [2] K. Brindle, T. Stephenson, The application of membrane biological reactors for the treatment of waste waters, Biotechnol. Bioeng. 49 (1996) 601–610. [3] Y. Magara, M. Itoh, The effect of operational factors on solid/liquid separation by ultramembrane filtration in a biological denitrification
160
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160 system for collected human excreta treatment plants, Water Sci. Technol. 23 (1991) 1583–1590. P. Le Clech, B. Jefferson, S.J. Judd, Impact of aeration, solids concentration and membrane characteristics on the hydraulic performance of a membrane bioreactor, J. Membr. Sci. 218 (1–2) (2003) 117–129. J.M. Lee, W.Y. Ahn, C.H. Lee, Comparison of the filtration characteristics between attached and suspended growth microorganisms in submerged membrane bioreactor, Water Res. 35 (2001) 2435– 2445. I.S. Chang, C.H. Lee, Membrane filtration characteristics in membrane-coupled activated sludge system — the effect of physiological states of activated sludge on membrane fouling, Desalination 120 (1998) 221–233. J.G. Choi, T.H. Bae, J.H. Kim, T.M. Tak, A.A. Randall, The behavior of membrane fouling initiation on the crossflow membrane bioreactor system, J. Membr. Sci. 203 (2002) 103–113. H.S. Shin, S.T. Kang, Characteristics and fates of soluble microbial products in ceramic membrane bioreactor at various sludge retention times, Water Res. 37 (2003) 121–127. W.T. Lee, S.T. Kang, H.S. Shin, Sludge characteristics and their contribution to microfiltration in submerged membrane bioreactors, J. Membr. Sci. 216 (2003) 217–227. H. Nagaoka, S. Ueda, A. Miya, Influence of bacterial extracellular polymers on the membrane separation activated sludge process, Water Sci. Technol. 34 (1996) 165–172. X. Huang, R. Liu, Y. Qian, Behaviour of soluble microbial products in a membrane bioreactor, Process. Biochem. 36 (2000) 401– 406. I.J. Kang, S.H. Yoon, C.H. Lee, Comparison of the filtration characteristics of organic and inorganic membranes in a membranecoupled anaerobic bioreactor, Water Res. 36. (2002) 1803– 1813.
[13] S.P. Hong, T.H. Bae, T.M. Tak, S. Hong, A. Randall, Fouling control in activated sludge submerged hollow fiber membrane bioreactor, Desalination 143 (2002) 219–228. [14] E. Tardieu, A. Grasmick, V. Jaugey, J. Manem, Hydrodynamic control of bioparticle deposition in a MBR applied to wastewater treatment, J. Membr. Sci. 147 (1998) 1–12. [15] E. Tardieu, A. Grasmick, V. Jaugey, J. Manem, Influence of hydrodynamics on fouling velocity in a recirculated MBR for wastewater treatment, J. Membr. Sci. 156 (1999) 131–140. [16] J.S. Kim, C.H. Lee, I.S. Chang, Effect of pump shear on the performance of a crossflow membrane bioreactor, Water Res. 35 (2001) 2137–2144. [17] C. Wisniewski, A. Grasmick, Floc size distribution in a membrane bioreactor and consequences for membrane fouling, Colloids Surf. A: Physicochem. Eng. Aspects 138 (1998) 403–411. [18] L. Defrance, M.Y. Jaffrin, B. Gupta, P. Paullier, V. Geaugey, Contribution of various constituents of activated sludge to membrane bioreactor fouling, Bioresour. Technol. 73 (2000) 105–112. [19] E.H. Bouhabila, R.B. Aim, H. Buisson, Fouling characterization in membrane bioreactors, Sep. Purif. Technol. 22–23 (2001) 123–132. [20] M. Mulder, Basic Principles of Membrane Technology, Kluwer Academic Publisher, 1996. [21] M. Cheryan, Ultrafiltration and Microfiltration Handbook, Technimic Publishing, 1998. [22] Standard Methods for the Examination of Water and Wastewater, 20th ed., American Public Health Association, Washington, DC, 1998. [23] G. Belfort, R.H. Davis, A.L. Zydney, The behavior of suspensions and macromolecular solutions in cross-flow microfiltration, J. Membr. Sci. 96 (1994) 1–58. [24] S. Ognier, C. Wisniewski, A. Grasmick, Influence of macromolecule adsorption during filtration of a membrane bioreactor mixed liquor suspension, J. Membr. Sci. 209 (2002) 27–37.
Interpretation of fouling characteristics of ultrafiltration membranes during the filtration of membrane bioreactor mixed liquor Tae-Hyun Bae, Tae-Moon Tak ∗ School of Biological Resources and Materials Engineering, Seoul National University, San 56-1, Shinlim-dong, Kwanak-gu, Seoul 151-921, South Korea Received 11 February 2005; received in revised form 14 April 2005; accepted 19 April 2005 Available online 9 June 2005
Abstract In this study, ultrafiltration (UF) membranes with various pore sizes were prepared by the phase inversion method, and their fouling characteristics were investigated with membrane bioreactor (MBR) mixed liquor. MBR sludge was fractionated into three parts, suspended solids, colloids and solutes, and their individual contributions to membrane fouling were quantified in order to asses the mechanisms by which fouling occurs. Cake layer formation on the membrane surface constituted the main fouling mechanism and the incidence of irreversible fouling was so small as to be negligible. The degree of fouling correlated very strongly with membrane permeability. This implies that hydrodynamic conditions are important factors affecting membrane fouling. The fouling contribution of each sludge fraction appeared to depend on particle size, as both permeation drag and back transport velocity are particle size-related functions. Solutes played a significant role in the initiation of cake layer formation, because they were deposited onto the membrane surface and pore wall immediately upon initial filtration, but were dislodged only in small amounts by cross flow. Suspended solids were consistently deposited onto the membrane surface, until flux reached a steady state and colloids exhibited characteristics commensurate with an intermediated state between solutes and suspended solids. Suspended solids were, in fact, found to be the main contributor to the fouling process. However, the relative contribution of each of the sludge fractions to membrane fouling varied with the permeability of the membrane, and also with the hydrodynamic condition. © 2005 Elsevier B.V. All rights reserved. Keywords: Membrane bioreactor (MBR); Membrane fouling; Ultrafiltration; Suspended solids; Colloids; Solutes
1. Introduction Membrane bioreactors (MBRs) have been widely used in advanced wastewater treatment schemes. The substitution of a secondary clarifier with membranes enables the total rejection of suspended solids without a settling process, and offers numerous advantages over the conventional biological process [1]. In particular, the absolute rejection of sludge by the membrane makes it possible to overcome a problem of dependence on settleability [2]. Furthermore, membrane separation enables a significant increase in the concentration of mixed liquor suspended solids (MLSS) in the bioreactor, thus reducing its size for a given sludge [1]. ∗
Corresponding author. Tel.: +82 2 880 4621; fax: +82 2 873 2285. E-mail address: [email protected] (T.-M. Tak).
0376-7388/$ – see front matter © 2005 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2005.04.037
However, wide acceptance of the MBR system in the wastewater treatment process has been hindered by problems of membrane fouling, which decreases permeate flux during the filtration of activated sludge. Membrane fouling increases the operation costs of MBR processes, due to the necessity of cleaning and replacing fouled membranes. Because of this, various techniques have been utilized in order to reduce membrane fouling in MBR processes. Since MBR mixed liquor includes living microorganisms and their metabolites, the fouling mechanism is even more complex than that of conventional membrane separation processes. Many researchers have focused their attentions on the fouling mechanism in MBR systems. It has been reported that MLSS concentration [3–5], sludge characteristics [6–9], and the amount and composition of microbial products [6,9–11] are the key factors in membrane fouling. In addition,
152
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
membrane fouling may be directly influenced by properties of the membrane [7,12] and the hydrodynamic conditions of the membrane process, most notably, permeation drag and back transport [13–16]. Several studies have attempted to quantify the degree and character of fouling caused by each fraction of the activated sludge, namely, suspended solids, colloids and solutes. Wisniewski et al. reported that solutes play a major role in membrane fouling in the MBR process [17]. Another study presented evidence that suspended solids cause the greatest amount of fouling among the sludge fractions [18]. However, another study reported that colloids are the main contributors to membrane fouling [19]. Clearly, such a panoply of conflicting results can only deepen the controversy. However, by reviewing most of the previous relevant literature, we can extend the hypothesis that fouling contributions from various fractions of the activated sludge may be variable, and change according to membrane properties, hydrodynamic conditions and the physiological properties of the biomass. This present work is focused on the effects of membrane pore size, permeability and filtration time on membrane fouling, and should be considered the first in a series of inquiries into this issue. In this study, ultrafiltration (UF) membranes of various pore sizes were prepared using the phase inversion method, and their fouling characteristics were assessed by testing them with activated sludge mixed liquor from an MBR plant. The objectives of this study were to investigate the roles of various sludge fractions in membrane fouling, and also to evaluate the effects of membrane pore size, permeability and filtration time on the behaviors of the aforementioned sludge constituents. The results of this study would help us elucidate the fouling mechanism of UF membranes during the filtration of MBR mixed liquor.
2. Materials and methods 2.1. Membrane preparation and characterization Membranes used in this study were prepared using phase inversion method [20]. The membrane material was cellulose acetate (CA; CA398-3, Eastman-Kodak), and N-methyl-2pyrrolidone (NMP, Aldrich) and acetone (Aldrich) were used as the solvent and co-solvent, respectively. In order to change the membrane pore size, the solvent/co-solvent ratio of the dope solution was controlled. The dope solution was kept at room temperature for 24 h, and then cast on a polyester non-woven fabric with a casting knife having 200 m thickness. The nascent membrane was evaporated at 25 ± 1 ◦ C and 65 ± 5% relative humidity for 30 s, and then immersed in an 18 ± 1 ◦ C deionized water coagulation bath. Pure water flux was measured under a transmembrane pressure (TMP) of 100 kPa at 20 ± l◦ C after the flux reached steady state. The molecular weight cut off (MWCO) was characterized by the rejection performance of
Table 1 The composition of standard wastewater Components
Concentration (mg/l)
Glucose (NH4 )2 SO4 KH2 PO4 MgSO4 ·7H2 O CaCl2 ·2H2 O FeCl3 ·6H2 O NaHCO3
400 220 23 50 10 4 50
1000 mg/l polyethylene glycol (PEG) or polyethylene oxide (PEO) aqueous solution [21]. Rejection was calculated by the following equation: Cper R (%) = 1 − × 100 (1) Cfeed where Cper is the concentration of the permeate and Cfeed the feed concentration. Since new polymeric membranes can be compacted during filtration by applied pressure, the flux can be declined without membrane fouling. In order to remove the compaction effect on flux decline, all membranes had been previously filtered by pure water until flux reached steady state, prior to the fouling tests. 2.2. Activated sludge The activated sludge used in this study had been cultivated in a submerged MBR plant, with synthetic substrate. Standard wastewater composition is presented in Table 1. Glucose, (NH4 )2 SO4 and KH2 PO4 were used as carbon, nitrogen and phosphorus sources, respectively. Additional nutrients and alkalinity (NaHCO3 ) were also supplied to the reactor. The MBR plant, which was operated by the authors, was a membrane coupled sequencing batch reactor for biological nutrient removal study. MLSS concentration was maintained at levels between 3100 and 3400 mg/l. Sludge retention time was about 20 days, and the organic loading rate was 0.16–0.17 gCOD/gMLSSday. 2.3. Experimental system and analysis Flux decline behavior was measured using membrane cells, the membrane surface area of which was 18.1 cm2 . A schematic diagram of the filtration system is shown in Fig. 1. Filtration tests were performed at 100 kPa and 20 ± 2 ◦ C. Crossflow velocity was controlled at 1.2 m/s, and the flow rate was 2.5 l/min. Four membranes were tested simultaneously with the same feed, and air was supplied to the biomass in the feed tank during the filtration test. Flux was determined by weighing the permeate on a top-loading balance at timed intervals. Stirred batch cell (Amicon) was used for dead end filtration experiments under different conditions. MLSS concentration was measured in accordance with Standard Methods [22]. The particle size of the activated
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
153
Fig. 1. The schematic diagram of MBR mixed liquor filtration system.
sludge was measured with a Multisizer IIe (Coulter). The COD of the MBR mixed liquor and membrane permeate was measured by the absorbance method, using a Fotometer AL282 (AQUALYTIC). 2.4. Membrane fouling analysis The degree of membrane fouling was quantitatively calculated, using the resistance in series model [6,20–21]: Rt =
PT ηJ
Rt = Rm + Rc + Rf
Rf =
PT ηJw PT − Rm ηJw
PT − Rm − Rf Rc = ηJAS
In this study, we assumed that the activated sludge consisted of solutes, colloids and suspended solids, and that these components independently contribute to membrane fouling, that is to say, we neglected any coupling or synergistic effects which might occur among the components. Therefore, the resistance of the activated sludge can be considered to be equal to the sum of the resistances of the suspended solids, colloids and solutes:
(2)
RAS = Rss + Rcol + Rsol
(3)
where RAS is the resistance of the activated sludge, Rss the resistance of the suspended solids, Rcol the resistance of the colloids and Rsol the resistance of the solutes. The resistance of the activated sludge could be measured from the activated sludge filtration, and calculated by the following equation:
where J is the permeation flux (m3 /m2 s), PT the TMP (Pa), η the viscosity of the permeate (Pa s), Rt the total filtration resistance (m−1 ), Rm the membrane resistance (m−1 ), Rc the cake layer resistance (m−1 ) and Rf the fouling resistance (m−1 ). The intrinsic membrane resistance (Rm ), the cake resistance by cake layer formed on the membrane surface (Rc ) and the fouling resistance caused by pore plugging and irreversible adsorption of foulants onto the membrane pore wall or surface (Rf ) can be calculated using the following equations: Rm =
2.5. Fouling contribution of sludge fractions
(4) (5)
RAS = Rt − Rm =
where JAS is the flux of activated sludge at steady state, Jw the initial water flux and Jw the final water flux after removing the cake layer by flushing with tap water.
(8)
Colloids and solutes could be obtained by extracting the supernatants generated after 4 h of gravitational sedimentation of the activated sludge mixed liquor. In fact, the colloids in the present study included large non-settling particles, which were derived mainly from the breakage of sludge flocs. The sum of the resistances of the colloids and solutes could be determined by a filtration test of the supernatant: Rcol + Rsol =
(6)
PT − Rm ηJAS
(7)
PT ηJsup
(9)
where Jsup is the permeate flux of the supernatant. Finally, the soluble fraction was acquired via the filtration of the supernatant with a 0.45 m microfiltration membrane (Millipore). The resistance of the solutes could be measured
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
154
Table 2 The compositions of the dope solutions and the characteristics of the UF membranes Membranes
CA-1 CA-2 CA-3 CA-4 a
Composition (wt.%) (CA/NMP/acetone)
(15/35/50) (15/55/30) (15/75/10) (15/85/0)
PWFa (l/m2 h)
34.1 193.0 224.3 282.2
Rejection (%) 10000
20000
35000
100000
200000
300000
52 7 8 6
75 11 9 11
89 19 16 17
95 86 76 50
98 96 91 71
99 99 96 93
PWF: pure water flux.
by the filtration of the soluble fraction: Rsol =
PT ηJsol
(10)
where Jsol is the permeate flux of the soluble fraction. RAS , Rss , Rcol and Rsol were calculated by Eqs. (8)–(10).
3. Results and discussion 3.1. Characterization of the membranes It has been established that the solvent–water exchange rate in the coagulation bath can alter membrane pore size and permeability. In general, if the exchange rate is delayed, pore size and permeability decrease, and the top layer, which forms on the support layer, becomes denser and thicker [20]. Since the addition of acetone can serve to slow the rate of solvent–water exchange, we controlled the NMP/acetone ratios in order to vary the pore sizes of the UF membranes [7]. Table 2 shows the pure water flux and solute rejection rate of the membranes. Permeability of the membranes increased with NMP content in the dope solutions from membranes CA1 to CA-4. Membrane pore size also increased with NMP concentration, thus the MWCO of CA-1, CA-2, CA-3 and CA-4 were 35,000, 100,000, 200,000 and 300,000 Da, respectively. 3.2. Particle size distribution and COD measurement Generally, the aeration intensity of submerged MBR systems is much higher than that of conventional activated sludge processes due to the necessity of control of membrane fouling. Thus, sludge particles can be broken up by shear stress resulting from intensive aeration. Fig. 2(a) shows the particle size distribution of the MBR mixed liquor. Most particles existed in a size range of between 10 and 40 m, and the mean particle size was 25 m, which was much lower than that of conventional activated sludge. The particle size range of the supernatant is shown in Fig. 2(b), and the mean diameter there was 9 m. Relatively large particles, generated by shear stress, increased both the concentration and mean particle diameter of supernatant. As mentioned above, those non-settling large particles were classified, in this study, as colloids. Soluble COD (SCOD) of feed and membrane permeates are shown in Fig. 3. More than 50% of SCOD was removed
Fig. 2. The particle size distributions of activated sludge and supernatant. (a) Activated sludge; (b) supernatant.
by the UF membranes. The removal rate increased slightly, with decreasing membrane pore size. In this study, the carbon source used in the substrate was glucose, which could easily pass through the membranes. This result indicated that the mixed liquor contained macromolecules, derived from microbial metabolism, and that they were rejected by the UF membrane. 3.3. Flux decline during the filtration of MBR mixed liquor Flux decline behaviors of the membrane are illustrated in Fig. 4. All filtration tests were performed until flux reached
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
155
case of the other membranes, however, membrane fouling occurred and flux declined to steady state at a lower level than the initial. Interestingly, all membranes, except CA-1, exhibited the similar steady state flux values for the filtration of the supernatant and the soluble fraction. This indicated that the sludge components were consistently deposited onto the membrane surfaces during filtration, until total filtration resistances reached the same level. 3.4. Filtration resistances of the membranes
Fig. 3. The soluble COD of feed and membrane permeate.
stable state. The rate of flux decline was very high during initial filtration, but decreased with time, finally reaching steady state. The amount of flux decline increased with both membrane pore size and permeability from CA-1 to CA-4. Although the initial flux characteristics of the membranes were quite different, steady state flux values were practically uniform during the filtration of the activated sludge. Membrane CA-1 exhibited no significant decline in flux during filtration of both the supernatant and the soluble fraction. In the
Filtration resistances, calculated from activated sludge filtration, are presented in Table 3. As mentioned above, total resistances were equal for all membranes. Since intrinsic membrane resistances vary inversely with pure water flux, they showed a tendency to decrease, from CA-1 to CA-4. However, the resistance of the cake layer increased with membrane permeability, due to the large amount of sludge deposition on the membrane surface. Cake resistance was observed to make the highest contribution to flux decline in all membranes. In particular, from CA-2 to CA-4, the contribution of cake resistance to fouling reached more than 90%. Thus, it can be concluded that the main fouling mechanism, in this study, was the formation of the cake layer, which results from the deposition of sludge components on the membrane surface.
Fig. 4. Behaviors of the flux declines of UF membranes during the filtration of various sludge fractions (MLSS: 3100–3400 mg/l, SS of supernatant: 90 mg/l and COD of soluble fraction: 23 mg/l). (a) CA-1, (b) CA-2, (c) CA-3 and (d) CA-4.
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
156
Table 3 The filtration resistances of the UF membranes
CA-1 CA-2 CA-3 CA-4
Rm (×1012 m−1 )
Rf (×1012 m−1 )
Rc (×1012 m−1 )
Rm /Rt (%)
Rf /Rt (%)
Rc /Rt
10.5 1.9 1.6 1.3
7.8 1.1 0.8 0.6
15.5 32.2 35.0 38.0
31 5 4 3
23 3 2 2
46 92 94 95
Irreversible fouling resistance increased as the permeability of the membrane decreased. Pore clogging and permanent adsorption of macromolecules are considered to constitute a main cause of irreversible fouling of UF membranes. It was reported that cake layers forming on the membrane surface act as barriers which protect membrane surfaces and pores [5]. Thus, irreversible fouling of the membrane increased as its cake layer resistance decreased. As a result, Rf shows a gradual decrease from CA-1 to CA-4. 3.5. Contributions of sludge constituents to membrane fouling Filtration resistances of suspended solids, colloids and solutes were calculated for all membranes. Filtration resistances of sludge constituents versus permeability of membrane measurements are presented in Fig. 5. From membranes CA-1 to CA-4, RAS exhibited a linear relationship to membrane permeability, and our results showed there to be a very high correlation coefficient (R = 0.99) between the two parameters. This higher correlation was also observed for individual sludge fractions: 0.99, 0.98 and 0.99 for Rss , Rcol and Rsol . As all membranes were filtered under identical conditions, membrane pore size and permeation drag must necessarily be the main factors operant in membrane fouling. The correlation between filtration resistance and membrane permeability prove that increases in filtration resistance are mainly
Fig. 5. The relationship between the resistances of sludge constituents and membrane permeability.
affected by hydrodynamic conditions, such as permeation drag and back transport, and also that effect of pore size can be considered to be negligible in the UF range from MWCO 35,000 to 300,000. Generally, as the pore size of a membrane increase, fine particles and macromolecules can more readily plug the pore and be adsorbed onto the pore walls [13]. In this study, however, the formation of the cake layer proved to be the main fouling mechanism, and it grew rapidly from the initial period of filtration. As mentioned above, the cake layer can function as a protective barrier. Due to this feature of the cake layer, the influence of pore size on membrane fouling was found to be so small to be negligible in this study. All sludge constituents are attached to the membrane surface by permeation drag, but are also detached by back transport. Permeation drag, which is generated by flux, increased with operation flux and particle size. Back transport in a cross flow system consists of Brownian diffusion, inertial lift and shear-induced diffusion [23]. Although Brownian diffusion, the driving force of which is the concentration gradient between the membrane surface and the bulk side of feed, tends to decrease as particle size increases, the other forces, which are generated by cross flow, tend to increase with cross flow velocity and particle size [23]. MBR mixed liquor contains large sludge flocs as a major fraction, although it contains a small amount of solutes. Furthermore, as these solutes consist mainly of high molecular weight macromolecules, the influence of Brownian diffusion is also extremely small, or negligible. Therefore, it can be concluded that the back transport velocity of the particles increased with their size. As shown in Fig. 5, regardless of membrane permeability, the contribution of Rss to total resistance was higher than that of any other sludge constituent, because suspended solids comprised the main fraction of the activated sludge, and played a key role in the deposition and formation of the cake layer. Although the resistances of all sludge constituents were proportional to the water flux of the membrane, the degree of this proportion varied among the sludge constituents. The slopes of the lines in Fig. 5 are 0.035, 0.019 and 0.009 for Rss , Rcol and Rsol , respectively. This indicates that the influence of permeability on membrane fouling increased with the particle size of foulants at the same cross flow velocity. On the other hand, fouling control appears to be a simpler matter for suspended particles than for other components, because back transport velocity, which is generated by cross flow, also increases with particle size.
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
157
Fig. 6. The contributions of each sludge fraction to membrane fouling according to relative filtrate volume. (a) CA-1, (b) CA-2, (c) CA-3, and (d) CA-4.
3.6. Deposition of sludge constituents on membrane surface according to filtration time
lutes. They take a shorter time than suspended solids but a longer time than solutes to reach steady state.
The filtration resistances of each sludge constituent, according to the relative filtrate volume, are shown in Fig. 6. All membranes exhibit a similar pattern in terms of resistance increase behavior, although filtration resistance values were different. Solutes are rapidly deposited onto the membrane surface, and then filtration resistances reach nearly steady state at an early stage of filtration. This result is consistent with the previous studies of other researchers [24]. Rapid solute deposition is induced by the inherent characteristics of soluble microbial products in the sludge mixed liquor. Since soluble macromolecules have a low back transport velocity and strongly adsorb onto the membrane surfaces, solutes are readily deposited on the membrane surfaces by permeation drag, and not readily detached by shear force. On the other hand, Rss gradually increased until the flux reached steady state. Although a huge amount of suspended solids were deposited onto the membrane surfaces by permeation drag, those particles are readily removed by cross flow due to their larger size. Thus, it takes the longest time for suspended solids to reach an equilibrium between back transport and permeation drag. Colloids and fine particles exhibited intermediate characteristics between suspended solids and so-
3.7. Hypothetical mechanism for initial cake layer formation According to the results described in the previous section, a mechanism underlying the formation of the cake layer can be proposed. A schematic illustration of cake formation according to flux decline is presented in Fig. 7. 3.7.1. Phase-1 Rapid flux decline occurs during initial filtration. All sludge particles are deposited onto the membrane surface. Large sludge flocs attached to the membrane surface can be removed by cross flow due to their size, while solutes are deposited onto the membrane surface and pore wall, and cause irreversible fouling. The influence of colloids is insignificant compared to the influences of large sludge flocs and solutes, since the amount of colloids are much smaller than the amount of suspended solids, and also because some of the attached colloidal particles are removed by cross flow. Consequently, all sludge particles take part in the formation of the cake layer during the initial stage, thus inducting rapid flux decline. The solutes are the main
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
158
Fig. 7. The schematic illustration of initial cake layer formation.
factor in the flux decline and induce irreversible membrane fouling. 3.7.2. Phase-2 Rate of flux decline is alleviated compared to that of Phase1. In this stage, sludge constituents are accumulated, not only on the membrane surface, but also on previously deposited layers. Permeation drag of this stage has significantly declined, because of the initial flux decline in Phase-1, thus the amount of particle deposition decreases from the rates associated with Phase-1. Permeation drag and back transport of solutes reach equilibrium state in this stage, and colloids and suspended solids contribute to an increase in filtration resistance. While the influence of colloids on fouling gradually decreases, suspended solids are continuously deposited onto the surface. Finally, the resistance of the colloids increases nearly to steady state at the end of this stage, and the rate of flux decline is depressed even further. Most fouling occurring in this stage is reversible fouling, which can be ameliorated by cleaning. 3.7.3. Phase-3 The fouling rate decreases further and reaches steady state. Most deposited particles in this stage are large sludge flocs. Particle deposition is significantly reduced from the levels associated with Phase-1 or Phase-2, due to dramatically reduced
permeation drag. Another important factor influencing flux decline is the compaction of the cake layer. The deposited cake layer can be compacted by TMP. Compaction of the cake layer also increases filtration resistance and decreases permeability. Compaction takes place throughout the whole process. However, its effects become far more relevant during Phase-3 than they were during Phases-1 and 2, because the largest amounts of particles are now being pressurized for a longer time. When the permeation drag of the suspended solids decreases until it becomes equal to the back transport velocity, flux reaches steady state. 3.8. Discussion — relative contribution of each sludge fractions to membrane fouling Results from pervious studies and the present work are summarized in Table 4 [9,17–19]. Results were quite different from study to study. Bouhabila et al. speculated that those differences could be derived from the nature of the substrate, physiological characteristics of the biomass, fractionation method and membrane materials [19]. In this study, the relative fouling contributions of each sludge fraction varied according to membrane permeability. The main fouling contributor, in this present work, proved to be suspended solids, which is consistent with the reports of Lee et al. [9] and Defrance et al. [18], but the relative
Table 4 The relative contributions of various sludge fractions to membrane fouling in MBR systems (%) Fraction
Wisniewski
Defrance
Bouhabila
Lee
Our results CA-1
Suspended solids Colloids Solutes a
24 24 52
The contribution of supernatant to fouling.
65 30 5
24 50 26
63–71 29–37a 29–37a
83 4 13
CA-2
CA-3
CA-4
76 10 14
74 13 13
72 14 14
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160
Fig. 8. Flux decline behaviors during the dead end filtration of membrane CA-2.
contribution of suspended solids to fouling decreased with membrane permeability. On the other hand, the relative contribution of the colloids increased with membrane permeability, because their back transport velocity was relatively low compared to that of the suspended solids. Considering the results of this study, it does not seem overly speculative to surmise that the relative fouling contributions of each sludge fraction can also be influenced by hydrodynamic conditions. In this study, since the membrane was operated at a high flux and a low cross flow velocity, the relative contribution of suspended solids was very high, and the contributions of the colloids and solutes were relatively low. Because the deposition of suspended solids is easily controlled by the cross flow, their relative contribution to fouling can be decreased by ensuring conditions of low flux and high shear rate. This was proved by performing a simple dead end filtration test. The dead end filtration tests were carried out at a TMP of 20 kPa and a stirring rate of 600 rpm, using stirred batch cell. Results from the filtration of membrane CA-2 are presented in Fig. 8 and Table 5. Fig. 8 shows that the flux can be gradually decreased, nearly to the steady state flux values of activated sludge by soluble fraction. This implies that the deposition of large particles can be controlled by the creation of conditions of low permeation drag and high shear rate, but more solutes will then be directly deposited onto the membrane surfaces and pore walls without the protective effects of the cake layer. As a result, as shown in Table 5, the main fouling contributor shifted from the suspended solids to the solutes. It should be noted that hydrodynamic control is powerful method for the mitigation of macroscopic fouling, but it cannot prevent the deposition of soluble macromolecules. Table 5 The filtration resistances and relative fouling contributions of each sludge fraction during dead end filtration of membrane CA-2 Resistance (×1012 m−1 ) R/RAS (%)
Suspended solids
Colloids
Solutes
1.1 24
0.4 9
3.1 67
159
As discussed in the previous section, the contributions of various sludge fractions to fouling can be affected by filtration time, as well as membrane permeability and hydrodynamic conditions. Many other factors also may affect the fouling contributions of the various sludge fractions. Thus, it is futile to compare relative fouling contributions from study to study, because varying experimental conditions clearly produce discrepant results. However, a series of experiments with controlled conditions could result in a uniform and relevant set of results. With regard to a concerted investigation into membrane fouling, this seems the most useful approach. Therefore, we will conduct further studies, focusing on other factors affecting membrane fouling, such as membrane material, system configuration and hydrodynamic conditions, and the physiological properties of the biomass.
4. Conclusions In this study, ultrafiltration membranes with various pore sizes were prepared by the phase inversion method, and their fouling characteristics were investigated with fractionated MBR mixed liquor. Cake layer formation on the membrane surface constituted the main fouling mechanism and the incidence of irreversible fouling was so small as to be negligible. The degree of fouling correlated very strongly with membrane permeability. The fouling contribution of each sludge fraction appeared to depend on particle size, as both permeation drag and back transport velocity are particle size-related functions. Solutes played a significant role in the initiation of cake layer formation. Suspended solids were consistently deposited onto the membrane surface, until flux reached a steady state and colloids exhibited characteristics commensurate with an intermediated state between solutes and suspended solids. Suspended solids were, in fact, found to be the main contributor to the fouling process. However, the relative contribution of each sludge fraction to membrane fouling varied with the permeability of the membrane, and also with the hydrodynamic condition.
5. Acknowledgement This work was supported by the Seoul National University Foundation (Project No. 02-02-31-4).
References [1] J. Manem, R. Sanderson, Water Treatment Membrane Process, AWWA Research Foundation, McGraw-Hill, 1996, Chapter 17. [2] K. Brindle, T. Stephenson, The application of membrane biological reactors for the treatment of waste waters, Biotechnol. Bioeng. 49 (1996) 601–610. [3] Y. Magara, M. Itoh, The effect of operational factors on solid/liquid separation by ultramembrane filtration in a biological denitrification
160
[4]
[5]
[6]
[7]
[8]
[9]
[10]
[11]
[12]
T.-H. Bae, T.-M. Tak / Journal of Membrane Science 264 (2005) 151–160 system for collected human excreta treatment plants, Water Sci. Technol. 23 (1991) 1583–1590. P. Le Clech, B. Jefferson, S.J. Judd, Impact of aeration, solids concentration and membrane characteristics on the hydraulic performance of a membrane bioreactor, J. Membr. Sci. 218 (1–2) (2003) 117–129. J.M. Lee, W.Y. Ahn, C.H. Lee, Comparison of the filtration characteristics between attached and suspended growth microorganisms in submerged membrane bioreactor, Water Res. 35 (2001) 2435– 2445. I.S. Chang, C.H. Lee, Membrane filtration characteristics in membrane-coupled activated sludge system — the effect of physiological states of activated sludge on membrane fouling, Desalination 120 (1998) 221–233. J.G. Choi, T.H. Bae, J.H. Kim, T.M. Tak, A.A. Randall, The behavior of membrane fouling initiation on the crossflow membrane bioreactor system, J. Membr. Sci. 203 (2002) 103–113. H.S. Shin, S.T. Kang, Characteristics and fates of soluble microbial products in ceramic membrane bioreactor at various sludge retention times, Water Res. 37 (2003) 121–127. W.T. Lee, S.T. Kang, H.S. Shin, Sludge characteristics and their contribution to microfiltration in submerged membrane bioreactors, J. Membr. Sci. 216 (2003) 217–227. H. Nagaoka, S. Ueda, A. Miya, Influence of bacterial extracellular polymers on the membrane separation activated sludge process, Water Sci. Technol. 34 (1996) 165–172. X. Huang, R. Liu, Y. Qian, Behaviour of soluble microbial products in a membrane bioreactor, Process. Biochem. 36 (2000) 401– 406. I.J. Kang, S.H. Yoon, C.H. Lee, Comparison of the filtration characteristics of organic and inorganic membranes in a membranecoupled anaerobic bioreactor, Water Res. 36. (2002) 1803– 1813.
[13] S.P. Hong, T.H. Bae, T.M. Tak, S. Hong, A. Randall, Fouling control in activated sludge submerged hollow fiber membrane bioreactor, Desalination 143 (2002) 219–228. [14] E. Tardieu, A. Grasmick, V. Jaugey, J. Manem, Hydrodynamic control of bioparticle deposition in a MBR applied to wastewater treatment, J. Membr. Sci. 147 (1998) 1–12. [15] E. Tardieu, A. Grasmick, V. Jaugey, J. Manem, Influence of hydrodynamics on fouling velocity in a recirculated MBR for wastewater treatment, J. Membr. Sci. 156 (1999) 131–140. [16] J.S. Kim, C.H. Lee, I.S. Chang, Effect of pump shear on the performance of a crossflow membrane bioreactor, Water Res. 35 (2001) 2137–2144. [17] C. Wisniewski, A. Grasmick, Floc size distribution in a membrane bioreactor and consequences for membrane fouling, Colloids Surf. A: Physicochem. Eng. Aspects 138 (1998) 403–411. [18] L. Defrance, M.Y. Jaffrin, B. Gupta, P. Paullier, V. Geaugey, Contribution of various constituents of activated sludge to membrane bioreactor fouling, Bioresour. Technol. 73 (2000) 105–112. [19] E.H. Bouhabila, R.B. Aim, H. Buisson, Fouling characterization in membrane bioreactors, Sep. Purif. Technol. 22–23 (2001) 123–132. [20] M. Mulder, Basic Principles of Membrane Technology, Kluwer Academic Publisher, 1996. [21] M. Cheryan, Ultrafiltration and Microfiltration Handbook, Technimic Publishing, 1998. [22] Standard Methods for the Examination of Water and Wastewater, 20th ed., American Public Health Association, Washington, DC, 1998. [23] G. Belfort, R.H. Davis, A.L. Zydney, The behavior of suspensions and macromolecular solutions in cross-flow microfiltration, J. Membr. Sci. 96 (1994) 1–58. [24] S. Ognier, C. Wisniewski, A. Grasmick, Influence of macromolecule adsorption during filtration of a membrane bioreactor mixed liquor suspension, J. Membr. Sci. 209 (2002) 27–37.