3. Experimental methods
3.1 Material
3.2.5 Fractionation of supernatant solutes
Supelite DAX-8 and Amberlite XAD-4 resins were used to fractionate supernatant solutes into hydrophobic acids (HPO-A) and hydrophobic neutrals (HPO-N) adsorbing onto the DAX-8 resin, transphilic acids (TPI-A) and transphilic neutrals (TPI-N) absorbing onto the XAD-4 resin, and the hydrophilic faction which does not adsorb on either theDAX-8 or XAD-4 resin. Only the hydrophobic acids and the transphilic acids fractions were used in this study. However, the separation of the different organic matter fractions with XAD resins is not sharp, instead the fractions overlap to a certain degree (Aiken & Leenheer, 1993)
. It should be mentioned that the so-called hydrophobic fraction (adsorbing onto the DAX-8 resin) do not exhibit a truly hydrophobic character in chemistry. The organic matter found in the hydrophobic fraction merely exhibits a more hydrophobic character in comparison to
there was no interaction among these three parts and the total resistance was the sum of the three resistances. Sludge was centrifuged (4,000 rpm) at 4℃ and the supernatant was considered to contain colloids and solutes. The supernatant was then filtered through a 0.45 µm membrane (Mixed cellulose ester, Advantech) to obtain the solutes. The resistance of individual sludge component was determined by the is resistance caused by membrane itself (1/m), Ras is resistance by sludge (1/m), Rss is resistance by suspended solids (1/m), Rcol is resistance by colloids (1/m), Rsol is resistance by solutes (1/m), Jiw is stable flux by filtering Milli-Q water (clean water flux), Jas is flux by filtering sludge, Jsup is flux by filtering supernatant and Jsol is flux by filtering solutes.
First, the Milli-Q water was filtered through the membrane to determine Rm by using equation 2. Sludge was then filtered to determine Ras by using equation 4 after a stable flux was reached. Supernatant and solutes were filtered through and the Rcol+sol and Rsol were determined by using equation 5 and 6, respectively. The difference between Rcol+sol and Rsol was Rcol (Subtract equation 6 from equation 5). Once Rsol and Rcol are known, Rss can be calculated by equation 3.
3.2.7 Specific cake resistance
Constant pressure filtration using the dead-end cell system under unstirred condition was used to calculate specific cake resistance (α). Plotting t/V vs. V, knowing other parameters, α can be calculated as follows:
PV membrane (m2), C is concentration of MLSS and α is specific cake resistance (m/kg).
t/V versus V plot is depicted linearly and the slope can be obtained by linear regression analysis. Then the specific resistance is calculated from the slop value.
3.2.8 Fourier-transform infrared spectrometer
Attenuated total reflectance-FTIR (ATR-FTIR) (Bomem DA8.3, Canada) was used to characterize foulant on the membrane surface. Samples were prepared in 2 cm
× 2 cm rectangles and dried at a vacuum box overnight. Samples were examined to a resolution of 4 1/cm.
3.2.9 Characterization of nanosized TiO2 particles
The crystal structure of TiO2 particles was characterized by X-ray diffraction (XRD) with a Mac Science MXP-18 X-ray diffractometer using Cu Kα (voltage: 30 kV; current: 20mA; λ = 0.154056 nm) radiation. The particle size of TiO2 was determined by a Philip transmission electron microscope (TEM, Philip CM-200 TWIN) at 200 kV. For TEM observation, TiO2 suspension was dropped on a carbon-coated grid and then dried at room temperature. The particle size distribution of TiO2 particles was also measured by a dynamic light scattering particle size distribution analyzer (Zetasizer Nano ZS, Malvern, UK).
3.2.10 Characterization of morphology and chemical composition of membrane surface
The surface topography of the TiO2 composite membrane was observed by JEOL JSM-6700F field emission scanning electron microscopy (FE-SEM). For SEM observation, the membrane samples were cut into appropriate size and the surfaces were coated with gold by a sputter coating machine.
X-ray photoelectron spectroscopy (XPS) was conducted to determine the
The contact angle goniometer (MagicDroplet model 100, Future digital scientific, USA) was used to characterize the hydrophilicity of the composite membranes by sessile drop method. The contact angles were determined by taking the average of three measurements.
3.2.11 Fouling test of composite membranes
The membrane fouling test of the composites membranes were conducted using a stirred cell system, as shown in Figure 3.4. The activated sludge used as the feed of the fouling test was taken from a 30-L submerged MBR system with synthetic influent.
The samples in the filtration cell were stirred at a constant stirring rate over the entire experiment and all the data were automatically logged in a computer. All the experiments were carried out at 0.5 bar constant pressure by using a nitrogen cylinder.
Resistance-in-series model was used to assess the degree of membrane fouling:
Rt J PT
µ
= ∆ (8)
where J is the permeate flux (m3/m2.s), △PT the trans-membrane pressure (Pa), µ the viscosity of the permeate (Pa.s), and Rt the total filtration resistance (1/m).
3.2.12 Ultrasonic wash of the TiO2 composite membrane
To evaluate the fixation of TiO2 coating on membrane, ultrasonic washing (40 KHz) was applied. The relative atomic concentrations of elements on the membrane surface were quantified by XPS. The relative atomic concentrations of the individual elements can be calculated:
where Ai is the photoelectron peak area of the element i, Si is the sensitivity factor for the element i, and m is the number of the elements in the sample.
3.2.13 Other analytical methods
MLSS was measured following the standard method (APHA, 1998)
. TOC was measured using a TOC analyzer (TOC-5000A, Shimadzu). Each TOC sample was measured at least two times with a standard deviation of less than 5%. Ammonia nitrogen was measured using a spectrophotometer (DR/4000U, HACH) according to salicylate method (method 10031). All the samples for TOC and ammonia nitrogen measurements were filtered through a 0.45 µm membrane filter first (Mixed cellulose ester, Adventec). Capillary suction time (CST) was determined to evaluate the
filterability of sludge. Five milliliters of sludge was sampled from the bioreactor and the CST (304B CST, Triton) was measured immediately. Each CST measurement was performed at least three times with a standard deviation of less than 5%.
Chapter 4
Effect of sludge characteristics on membrane fouling in MBRs
4.1 Performance and fouling characteristics of membrane bioreactor under different sludge characteristics
MBRs with different sludge characteristics, bulking sludge and normal sludge, were investigated in this study to evaluate their effects on membrane performance.
4.1.1 Performance of membrane bioreactor treatment under different sludge conditions
In the beginning of the MBR operation, sludge bulking due to overgrowth of filamentous bacteria was observed. The excessive growth of filamentous bacteria when sludge bulking became serious is clearly shown in Figure 4.1 (a) and (b). Ideally, the sludge contains both filamentous bacteria and floc-forming bacteria. When the two are in balance, the filamentous bacteria act as the backbone of activated sludge flocs without causing sludge bulking (Jenkins et al., 1993)
. The theories of the overgrowth of filamentous bacteria are: (1) The surface/volume theory: filamentous bacteria have easier access to substrate, oxygen and nutrients than floc-forming bacteria owing to the long filaments, (2) The kinetic theory: filamentous and floc-forming bacteria have different maximum growth rates, (3) The accumulation/regeneration theory:
floc-forming bacteria have greater capacity of energy storage, (4) The starvation theory:
organisms with higher storage capacity are more resilient under limited substrate conditions (Dalentoft & Thulin, 1997)
. Since the majority of the nutrient compounds in the simulated feed are readily biodegradable, which are much more readily accessible to the filamentous bacteria. As a result, the filamentous bacteria became the dominant species.
To correct this problem an aerobic selector was installed. A selector is a separate mixing zone upstream of the aerobic basin in which the recycled activated sludge and influent wastewater are mixed. Three types of selectors are used in dealing with filamentous bulking: aerobic, anoxic and anaerobic. The key in preventing filamentous bulking by selector is the substrate utilization characteristics of the bacteria (Tsai & Lee, 1998)
. Filamentous bacteria have lower half-saturation constant (Ks) and maximum growth rate (µmax) than floc-forming bacteria, which therefore is the main theory of aerobic selector. In this way the sludge was successfully shifted from filamentous bacteria to flco-forming bacteria as seen in Figure 4.1 (c) and (d).
In conventional activated sludge process, sludge settleability is the key factor in maintaining effluent quality. Sludge bulking which is usually due to overgrowth of filamentous bacteria often deteriorates the performance of activated sludge. Figure 4.2
shows the removal of TOC and ammonia nitrogen. The selector was installed after the MBR was operated for 20 days. Despite the serious sludge bulking caused by overgrowth of filamentous bacteria, the effluent quality remained the same, as shown in Figure 4.2 (a) and (b). The average TOC and NH3-N in the MBR influent was 158±20.0 mg/L and 32.0±0.67 mg/L, respectively, over the entire period of operation. Nearly 98% of the organics were removed by the MBR treatment regardless of the sludge characteristics (Figure 4.1). Biological nitrification was also excellent. Almost 99% of ammonia nitrogen was nitrified during the experiment. The NH3-N of the effluent was reduced to 0.24±0.37 mg/L even when sludge bulking occurred (Figure 4.2 (b)). The result indicates that membrane bioreactor is a reliable wastewater treatment process.
The excellent pollutant removal renders MBR a promising process for wastewater reuse.
Figure 4.1. Microscopic images of sludge flocs: (a) and (b) overgrowth of filamentous bacteria without installation of the selector; (c), and (d) floc-forming bacteria after installation of the selector.
100 µm
100 µm 100 µm
100 µm
(a) (b)
(c) (d)
Time (day) bulking sludge and normal sludge.
(a)
(b)
Installing selector
4.1.2 Impact of bulking sludge on membrane fouling
Bulking sludge, on the other hand, had significant impact on membrane fouling, as illustrated in Figure 4.3 (a). In the initial period of operation, the TMP profile exhibited a typical two-stage pattern under subcritical flux operation when filamentous bacteria started to become dominant. A slow and progressive membrane fouling was observed in the initial 100 h followed by a sudden TMP increase. After the TMP reached -60 kPa, the membrane was chemically cleaned by 0.5% sodium hypochlorite for 2 hours. However, the membrane fouled right after the membrane cleaning with a fouling rate of up to 28.7 kPa/h. Therefore, frequent membrane cleaning was performed afterwards. The TMP profile of the MBR after the aerobic selector was installed was shown in Figure 4.3 (b). Membrane fouling decreased gradually and the TMP profile changed completely when floc-forming bacteria were dominant in the bioreactor. The TMP profile changed to a typical two-stage pattern in subcritical flux operation. The first stage of slow fouling rate lasted for about 200 h before the second stage of TMP jump appeared. The fouling rate was greatly reduced to 0.03 kPa/h in the progressive and slow fouling stage. After the floc-forming bacteria stabilized and became steadily dominant in the bioreactor, the fouling rate became steady and relatively slow. Meng et al (2006b) reported that the excess growth of filamentous bacteria formed a non-porous cake layer on the membrane surface which interfered with the membrane filtration. Meng et al (2006 a) and Meng & Yang (2007) further suggested that bulking sludge caused the formation of a dense cake layer on the membrane surface due to the fixation of filamentous bacteria. Chang et al (1999) also reported that bulking sludge have higher fouling tendency than normal sludge and pinpoint sludge. However, Li et al (2008) had found an opposite result that filamentous bacteria had negligible effect on membrane fouling. The contradict results might be due to the different influent wastewater and processes discussed in 2.1.2.5.
Particle size distributions of normal sludge and bulking sludge are shown in that bulking sludge caused by overgrowth of filamentous bacteria had larger particle size distribution. It contradicts the common knowledge that smaller particles are generally more easily to deteriorate membrane filtration (Chang et al., 2002; Rosenberger & Kraume, 2002)
. According to Carmen-Kozeny equation, specific cake resistance is a function of particle diameter, porosity of cake layer, and particle density. The specific cake resistance is inversely proportional to the square of the particle diameter. Thus smaller particles size will result in greater cake resistance. However, the severe fouling in
bulking sludge cannot be explained by particle size alone. There are some other important factors resulting in the severe fouling.
The distinct TMP profiles of normal sludge (floc-forming bacteria) and bulking sludge (filamentous bacteria) must be answered by the difference in sludge characteristics, which is summarized in Table 4.1. The supernatant TOC, representing SMP in mixed liquor (Liang et al., 2007)
, was about 12 times higher than that of normal sludge. The soluble EPS of bulking sludge was about 6 times higher. It strongly implies that SMP or other organic compounds in bulking sludge might be responsible for the higher fouling rate. On the contrary, the concentrations of bound EPS in normal sludge and bulking sludge were about the same. The detail will be discussed in 4.1.3. CST is commonly used to represent dewaterability of sludge. The CST of the bulking sludge was significantly larger than that of normal sludge, which echoes the findings by Wang et al (2006) and Wu et al (2007) that CST values were positively correlated to membrane fouling. Rosenberger and Kraume (2002) have also reported that soluble EPS affected the filterability of activated sludge most significantly, in agreement with our result. As a result, CST seems to be a good indicator of sludge filterability.
Operating time (h)
0 100 200 300
TMP (kPa)
-60
-50
-40
-30
-20
-10
0
Operating time (h)
0 100 200 300 400 500
TMP (kPa)
-60
-50
-40
-30
-20
-10
0
Figure 4.3. TMP profiles of different sludge properties: (a) filamentous bacteria and (b) floc-forming bacteria.
(a)
(b)
Particle size (µm)
0.1 1 10 100 1000
Volume (%)
0 2 4 6 8
Bulking sludge Normal sludge
Figure 4.4. Particle size distributions of normal sludge and bulking sludge.
Table 4.1. Comparison of sludge characteristics between normal sludge and bulking sludge
Supernatant TOC (mg/L)
CST (s)
Soluble EPSa (mg/L)
Bound EPS a (mg/g MLSS) Normal
sludge
5±2 16±2 25±16 130±13
Bulking sludge
66±9 30±28 145±37 133±20
a The concentration of EPS is expressed as the sum of proteins and polysaccharides as BSA and glucose, respectively.
4.1.3 Effect of sludge properties on EPS
Since EPS has been widely accepted as the major foulant in MBR (Nagaoka et al., 1996a;
Cho & Fane, 2002; Kimura et al., 2005; Zhang et al., 2006a)
, the EPS components in the mixed liquor were monitored and compared with the performance of the MBR operation for fouling study. Four components were monitored: soluble polysaccharides, soluble proteins, cell-bound polysaccharides and cell-bound proteins. In this study, total soluble EPS or SMP is the sum of soluble polysaccharides and soluble proteins. And the sum of cell-bound polysaccharides and cell-bound proteins represents the total bound EPS.
Figure 4.5 compares the concentration of soluble and cell-bound EPS in various sludge conditions. There was no significant difference in the production of bound EPS between bulking and normal sludge. On the other hand, much more soluble polysaccharides and soluble proteins were produced in bulking sludge, especially the soluble polysaccharides. Higher membrane fouling caused by overgrowth of filamentous bacteria seems to relate to the increased amount of SMP in bulking sludge, which echoes the observations by other researches that soluble polysaccharides or soluble proteins in SMP influence the membrane performance in MBR (Mukai et al., 2000;
Hernandez Rojas et al., 2005; Kimura et al., 2005; Rosenberger et al., 2005; Fan et al., 2006; Zhang et al., 2006b)
. However, Meng et al (2006a; 2006b; 2007) later made an observation that contradicts our results.
They concluded that severe membrane fouling caused by excessive growth of filamentous bacteria might be caused by the production of more bound EPS. The result differed from our observation, possibly because that they obtained the sludges from different MBR processes. Lately, Li et al (2008) also reported that bound EPS was the major contributor to membrane fouling. They pointed out that filamentous bacteria have no significant influence on membrane fouling. The contradictory findings could come from the difference in operation conditions or the difference in filamentous species. To verify the cause for severe fouling associated with bulking sludge, the foulants on membrane surface were identified by FTIR. FTIR spectra of fresh and fouled membranes are shown in Figure 4.6 to show the functional groups of the foulants on the membrane surface. The peaks at wave number 1647 and 1533 cm-1 are assigned to the amide-Ⅰ and amide-Ⅱ bands (Kimura et al., 2005; Jarusutthirak & Amy, 2006)
, respectively. The absorption band at 3286 cm-1 is N-H stretching. The peak at wave number 1041 cm-1 is assigned to bond vibrations of polysaccharides (Omoike & Chorover, 2004)
and the peak at wave number 2925 cm-1 is also a character of polysaccharides. The result suggests that
Protiens and polysaccharides concentration (mg/L)
0 20 40 60 80 100 120
Proteins and polysaccharides concentration (mg/g MLSS)
0 20 40 60 80 100 120 140 Soluble proteins
Soluble poolysaccharides Cell-bound proteins Cell-bound polysaccharides
Bulking sludge
Normal sludge
Bulking sludge
Normal sludge
Figure 4.5. Comparison of EPS components in filamentous bacteria and floc-forming bacteria.
Wave number (cm-1)
1000 2000
3000 4000
Absorbance
0.00 0.02 0.04 0.06 0.08 0.10 0.12
Fouled membrane Fresh membrane
1041
3286 2925
1647 1533
Figure 4.6. FTIR spectra of fresh and fouled membranes.
4.1.4 Effect of sludge fractions on membrane fouling
To provide more information concerning the contradictory results among other studies (Meng et al., 2006a; Meng et al., 2006b; Meng & Yang, 2007; Li et al., 2008)
and ours, fouling by sludge components was also investigated. Sludge was separated into three fractions by particle size: suspended solids, colloids and solutes. The experiment was first operated at two stirring rates to create different shear forces for the evaluation of the contributions of different sludge fractions on membrane fouling. Figure 4.7 illustrates the resistances of sludge fractions at different stirring rates. The difference between activated sludge and colloids + solutes represents the resistance of suspended solids.
The difference between colloids + solutes and solutes represents the resistance of colloids. In order to compare the relative contribution of sludge fraction on membrane fouling in detail, the results obtained in Figure 4.7 was summarized in Table 4.2. At low shear stress (stirring rate of 400 rpm), the major fouling contributors are colloids and solutes. The resistance contributed by colloids and solutes were 36.52% and 36.15%, respectively. When the stirring rate was increased to 1,000 rpm, the resistance contributed by suspended solids disappeared completely while the majority of resistance came from the solutes. As shown in Table 4.2, increasing the stirring rate increased the contribution of solutes to fouling. The finding of this study also proved that different operation conditions might lead to different results, which might explain the disagreement between studies (Wisniewski & Grasmick, 1998; Defrance et al., 2000; Bouhabila et al., 2001;
Lee et al., 2003; Bae and Tak, 2005a)
. Defrance et al (2000) and Bae and Tak (2005) concludes that suspended solids are the main contributor to membrane fouling because their systems were operated under relatively high flux and low cross flow. On the other hand, Wisniewski & Grasmich (1998) reported differently, that solutes were the main contributor to membrane fouling since their system was operated under high shear stress condition.
At high shear force, smaller components, namely, colloids and solutes, dictated the resistance. Back transport caused by Brownian diffusion is dominant for small particles and at low shear stress, while back transport caused by shear-induced diffusion and inertial lift increase with shear stress rate and is proportional to particles size (Belfort et al., 1994)
. As a result, the shear-induced diffusion and inertial lift of larger particles such as suspended solids and colloids keeps them away from the membrane, resulting in reduced resistance. In contrast, shear-induced diffusion and initial lift is negligible for small molecules. Back transport of small molecules is caused by Brownian diffusion. In membrane filtration when the drag force due to filtration balances the back transport, membranes are free of deposit. In subcritical flux, the back transport was equal or greater than the permeation drag, therefore, no sharp TMP increase would be observed.
Table 4.2 implies that when the membrane was operated at subcritical flux, larger
particles such as suspended solids would not deposit on the membrane to form a sludge cake. On the other hand, smaller particles such as soluble EPS and other macromolecules would be continuously attracted onto the membrane regardless of the strength of the shear force. TMP jump will be observed when local flux exceeds critical flux. The result is in agreement with the results in 4.1.3 that SMP dominated the membrane fouling in MBR, and, therefore, membrane fouling will occur eventually even though the MBR is operated under subcritical condition.