Performance of Anaerobic Membrane Bioreactor (AnMBR) with Sugarcane Bagasse Ash-based Ceramic Membrane treating Simulated Low-strength Municipal Wastewater: Effect of Operation Conditions

Abstract

This study assesses the performance of waste sugarcane bagasse ash (SBA)-based ceramic membrane in anaerobic membrane bioreactor (AnMBR) treating low-strength wastewater. The AnMBR was operated in sequential batch reactor (SBR) mode at hydraulic retention time (HRT) of 24 h, 18 h, and 10 h to understand the effect on organics removal and membrane performance. Feast-famine conditions were also examined to evaluate system performance under variable influent loadings. An average removal of >90% chemical oxygen demand (COD) was obtained at each HRT and starvation periods up to 96 days did not significantly affect removal efficiency. However, feast-famine conditions affected extracellular polymeric substances (EPS) production and consequently the membrane fouling. EPS production was high (135 mg/g MLVSS) when the system was restarted at 18 h HRT after shutdown (96 days) with corresponding high transmembrane pressure (TMP) build-up; however, the EPS content stabilized at ~60–80 mg/g MLVSS after a week of operation. Similar phenomenon of high EPS and high TMP was experienced after other shutdowns (94 and 48 days) as well. Permeate flux was 8.8±0.3, 11.2±0.1 and 18.4±3.4 L/m² h at 24 h, 18 h and 10 h HRT, respectively. Filtration-relaxation (4 min - 1 min) and backflush (up to 4 times operating flux) helped control fouling rate. Surface deposits (that significantly attributed to fouling) could be effectively removed by physical cleaning, resulting in nearly complete flux recovery. Overall, SBR-AnMBR system equipped with waste-based ceramic membrane appears promising for treatment of low-strength wastewater with disruptions in feeding.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11270-023-06173-3.

Introduction

Many countries are facing the challenges of water stress and inadequate wastewater treatment. In a paradigm change, wastewater is being perceived as a resource with the focus shifting from treatment towards recovery of resources (e.g., water, energy, nutrients) (Coats and Wilson, 2017). Wastewater flowing in urban drains is often characterized by low organic load (in terms of chemical oxygen demand, COD) (Dalvi et al., 2021; Piaggio et al., 2022). Considering the low energy consumption in anaerobic treatment and the advantage of retaining the slow-growing anaerobic microorganisms in the reactor, anaerobic membrane bioreactors (AnMBRs) have been tested for treating as-is and diluted municipal wastewater (Baek et al., 2010; Ji et al., 2020; Lin et al., 2011). Decentralized AnMBRs are being investigated in developing countries for cost-effective treatment and possible on-site reuse of the treated wastewater (Bair, 2016). However, decentralized treatment systems can experience long starvation during holiday seasons and short periods of starvation during maintenance closure (Palmarin et al., 2020). The consequent famine conditions lead to increased production of biomass-associated products through endogenous respiration and cell debris accumulation that enhances membrane fouling as experienced in aerobic MBRs (Palmarin et al., 2020; Wu and Lee, 2011).

Membrane filtration is an integral component of MBRs for retaining the biomass (sludge) in the reactor and producing high quality permeate. The membrane material can either be polymeric or ceramic. Though polymeric membranes are relatively inexpensive and more widely used in wastewater treatment applications, they have a shorter life and are more susceptible to thermal and chemical damage during operation (Dvořák et al., 2016). Ceramic membranes are characterized by narrow, well-defined pore-size distribution and higher hydrophilicity (He et al., 2019); they can also withstand variable and harsh thermal and chemical cleaning conditions more effectively (Zou et al., 2019). Commonly used ceramic membranes in AnMBRs are composed of alumina (Yue et al., 2015; Jeong et al., 2017; Chen et al., 2021) and silicon carbide (Cho et al., 2019; Ying et al., 2021); other membrane materials such as yttria and zirconia (Nilusha et al., 2020), pyrophyllite (Jeong et al., 2017) have also been reported.

Fouling is a major drawback in membrane separation systems (Liu et al., 2012); AnMBR systems particularly foul highly (Dvořák et al., 2016). Furthermore, membranes can account for up to 56% of AnMBR life cycle cost (Lin et al., 2011) motivating the need to develop improved and affordable membranes for AnMBR applications. One approach to lower ceramic membrane cost is to employ suitable secondary raw materials such as industrial solid wastes - together with sustainable management of such wastes, this would assist in the transition towards a circular economy. An example of a large-scale agro-industrial waste is sugarcane bagasse ash (SBA) generated through combustion of bagasse (the solid residue from crushing sugarcane) used as a boiler fuel in sugar mills. With annual global production of sugarcane around 1.9 billion metric tonnes, the yearly generation of SBA is about 12.3 million metric tonnes (Neto et al., 2021). SBA is primarily silica (65% by weight), with small fraction of other oxides such as alumina (0.49%) and titania (0.08%) (Umamaheswaran and Batra, 2008); the high silica content contributes to membrane strength (Jamalludin et al., 2020). SBA based membranes have been developed and tested in MBRs for treatment of industrial effluents such as distillery wastewater (Gupta et al., 2008) and oily wastewater (Jamalludin et al., 2018, 2020) as well as synthetic wastewater (Basu et al., 2014) and sewage (Tewari et al., 2012). To eventually mainstream the use of SBA waste-based ceramic membranes in MBRs, more research is required for optimizing the operation parameters and understanding fouling behaviour.

This study examines the treatment of low strength sewage (mimicking wastewater in urban drains) in an AnMBR equipped with ceramic membranes prepared from SBA. The following aspects were examined: (a) AnMBR operation in sequential batch reactor (SBR) configuration (b) variable hydraulic retention time (HRT) and startup/shutdown in SBR-AnMBR considering situations encountered on-ground in non-automated decentralized treatment systems and (c) fouling control strategies viz. filtration-relaxation, permeate backflushing and physical cleaning. The scope of biogas generation was not explored as no measurable biogas production has been reported in AnMBRs treating real low strength wastewater with COD < 225 mg/L (Smith et al., 2013). Considering the context of decentralized wastewater treatment systems operational in residences/small communities as well as in residential schools/higher education institutions, highly variable shutdown durations ranging from a few days for maintenance up to 3 months for summer breaks (Peralta et al., 2022) can be encountered. With this background, the experimental conditions were set to answer the following questions: (a) Is the SBR configuration suitable for AnMBR operation? (b) What is the effect of feast-famine conditions on biological treatment (i.e. COD removal) and membrane fouling (i.e. TMP build-up)? (c) Which of the fouling control strategies tested work best under the conditions of this study? The findings will assist in identifying appropriate operation parameters leading to satisfactory long term performance of AnMBR equipped with waste SBA based ceramic membranes.

Materials and Methodology

AnMBR Set-up and Operation

The AnMBR was tested in SBR mode (Fig. 1). The system consisted of 3 units viz. a continuous stirred tank reactor (CSTR) (diameter 17 cm × height 46 cm), a holding tank (length 20 cm × width 12 cm × height 50 cm) and a membrane filtration tank (length 15.5 cm × width 9 cm × height 30 cm). These 3 units were integrated using appropriate piping and valves. The CSTR (8 L working volume) was operated in sequential batch mode and the effluent was decanted into the holding tank (5 L working volume); the balance 3 L of the CSTR effluent was discarded. The effluent in the holding tank was then pumped into the membrane filtration unit (3.5 L working volume). All tanks were constructed of polymethyl methacrylate (acrylic) and fabricated locally.

Fig. 1.

Illustration of lab scale AnMBR set-up operated in SBR mode

The CSTR was fed using a reverse osmosis (RO) booster pump (Kemflo, India) at a flow rate of 0.8 L/min. The influent and permeate flow in the membrane filtration tank were maintained by peristaltic pumps (Electrolab India Pvt. Ltd., India). A mercury manometer in the permeate line was used to measure the transmembrane pressure (TMP). Temperature of the CSTR contents was monitored with a digital thermometer (Aeoss, India). SBA based flat-sheet ceramic membrane modules (pore size 7.14±1.19 μm and 0.03125 m² effective filtration area) were manufactured in-house at The Energy and Resources Institute (TERI), Delhi. The ceramic membrane module was immersed in RO water for 2 h to remove any loose deposits before submerging in the filtration tank.

The AnMBR system was operated at 24 h, 18 h, and 10 h HRT at ambient temperature (22–25°C). The reactor was studied for 377 days (including shutdowns). Table 1 summarizes the operation conditions of the AnMBR at various HRTs. Unforeseen shutdowns as well as stoppages during regular maintenance can lead to food deprivation and low food-to-microorganisms (F/M) ratio in biological systems. To mimic such starvation experiences, the system was operated for 18 h and 10 h in a day during 18 h and 10 h HRT respectively and the following cycle was started the next day. The AnMBR system was thus idle during the balance time of 6 h and 14 h at 18 h and 10 h HRT respectively. Nitrogen (N₂) purging of the CSTR was done for 0.5 h before initiating each HRT to maintain anaerobic conditions. Filtration was conducted in the filtration (4 min)-relaxation (1 min) mode. Backflush with permeate was performed by manually reversing the peristaltic pump flow after every HRT cycle for 4 min. Magnitude of the backflush flux (J_b) was adjusted with respect to operating flux (J) and J_b/J ratio up to 4 were tested.

Table 1.

Operation conditions of AnMBR at various HRTs

HRT (h)	Days (d)	SBR cycle details	Desired flux (L/m2 h)	Filtration-relaxation (min-min)	Backflush flux/operating flux ratio (Jb/J) (-)
24	1–3	Feeding: 10 min	8.9	4–1	2 to 3
		Reaction: 22.5 h
		Settling: 1 h
		Decantation: 20 min
-	4–97 (94 d SD) + 98–108 (11 d SR)	-	-	-	-
24	109–145	Feeding: 10 min	8.9	4–1	2 to 3
		Reaction: 22.5 h
		Settling: 1 h
		Decantation: 20 min
-	146–241 (96 d SD)	-	-	-	-
18	242-286*	Feeding: 10 min	11.1	4–1	2
		Reaction: 16.5 h
		Settling: 1 h
		Decantation: 20 min
-	287–334 (48 d SD)	-	-	-	-
10	335–377	Feeding: 10 min	21.1	4–1	2 to 4
		Reaction: 8.5 h
		Settling: 1 h
		Decantation: 20 min

*System was shutdown between days 273–282 because of COVID-19 restrictions; SD shutdown; SR sludge revival

During the shutdown period, anaerobic sludge was stored (within the CSTR) at ambient temperature and was subsequently used to restart the reactor after shutdown. The CSTR was wrapped with aluminium foil to prevent photo assisted reactions during shutdown. For the first shutdown period (day 4–97), the membrane was left submerged in the filtration tank and was physically cleaned using a sponge before reactor restart. AnMBR was operated only during weekdays (Monday to Friday) due to site unavailability on weekends following COVID-19 restrictions. As sludge from CSTR was removed only for sample analysis, the loss was negligible and sludge retention time (SRT) was assumed to be infinite.

Wastewater Feed and Seed Sludge

Synthetic wastewater composition as per Shim et al. (2002) was modified to make it low strength (~170–200 mg COD/L) (Table 2), simulating real wastewater properties from an urban drain in Delhi, India. For each experiment, feed of the required strength was prepared fresh daily with tap water. Fresh cow dung slurry (sieved sequentially through 2.5 mm followed by 1 mm mesh) cultured at 37°C was used as seed sludge. Volatile solids (VS) content of seed sludge was 7.43 g/L and inoculum volume to be added in CSTR was calculated assuming food to microorganisms (F/M) ratio of 0.6.

Table 2.

Composition of low strength synthetic wastewater feed

Chemical	Concentration (mg/L)
D-glucose (C6H12O6)	100
L-glutamic acid (C5H9NO4)	50
Ammonium acetate (CH3COONH4)	40
Sodium hydrogen carbonate (NaHCO3)	65
Ammonium chloride (NH4Cl)	35
Potassium dihydrogen phosphate (KH2PO4)	28
Sodium chloride (NaCl)	12
Magnesium chloride hexahydrate (MgCl2.6H2O)	8
Calcium chloride dihydrate (CaCl2.2H2O)	6
Ferric chloride hexahydrate (FeCl3.6H2O)	2

Analytical Methods

Samples (feed, CSTR effluent and permeate) were collected twice a week for analysis of COD (closed reflux method), ammonia, phosphate, pH and turbidity. The CSTR effluent was filtered through Whatman grade 1 filter to remove any suspended solids before analysis. All the analyses were done in duplicate as per APHA, 2005. Samples were collected from CSTR for mixed liquor suspended solids (MLSS), mixed liquor volatile suspended solids (MLVSS) and EPS (extracellular polymeric substances) measurement once every 7–10 days. Loosely bound (LB) and tightly bound (TB) EPS fractions were extracted as per method used by Li and Yang (2007) and Liu et al. (2012), respectively and analysed for both proteins and carbohydrates. Protein was quantified using the method as per Hartree (1972) and carbohydrate as per Dubois et al. (1956) and Albalasmeh et al. (2013). Surface deposits on used membrane were collected by washing the membrane surface with 100 mL RO water and analysed for MLVSS and EPS. Porosity of virgin and used ceramic membranes was measured using mercury porosimeter (Micromeritics, AutoPore IV, USA).

Total hydraulic resistance (R_t) was calculated using Darcy’s law using Eq. (1) as given below (Martinez-Sosa et al., 2011).

$$\mathrm{J}=\mathrm{TMP}/\eta \mathrm{Rt}$$ 1

where J is permeate flux (m/s), TMP is transmembrane pressure (Pa), η is the dynamic viscosity of the permeate (Pa s) and R_t is the total filtration resistance (1/m). Dynamic viscosity of the permeate was assumed to be that of water at ambient temperature.

Results and Discussion

The AnMBR was initially operated in continuous mode at 48 h HRT (data not shown). After 11 days operation, MLSS and MLVSS in the bioreactor were 202 mg/L and 74 mg/L, respectively. At this stage, direct exposure of membrane to sludge was experienced leading to sharp TMP rise (Fig. S1 in supplementary sheet). To avoid frequent transfer of biomass back to the CSTR from the filtration unit, the system configuration was modified by introducing a holding tank between CSTR and the filtration unit; the CSTR was subsequently operated in the SBR mode (as detailed in Section 2.1) for the rest of this study. HRT was reduced to 24 h as supported by substrate utilization kinetics results which followed zero-order kinetics (R² 0.9561) with decay coefficient 8.1883 h⁻¹ (Fig. S2 in supplementary sheet). The following sections describe the biological and filtration performance of the SBR-AnMBR system.

Organics Removal

Figure 2 shows the organics removal obtained under the operation conditions studied. The corresponding nutrients removal is summarized in Fig. S3 (supplementary sheet).

Fig. 2.

Organics concentration and removal profile at 24 h, 18 h and 10 h HRT in SBR-AnMBR

Before commencing AnMBR operation at 24 h HRT after switching from continuous operation, the biomass in the CSTR was intermittently fed with low strength wastewater for 74 days. Once fed, stirring was kept ON for few days without any discharge followed by settling (1 h) and decanting before next feeding cycle; total of 17 cycles were performed in 74 days. As a result, MLVSS increased to 430 mg/L on day 74 wherein filtration was initiated. The COD removal efficiency was 94.9% on day 1 of AnMBR operation that increased to 99% the next day. Organic loading rate (OLR) was 0.17±0.0 gCOD/L d and the F/M ratio varied between 0.30-0.39 gCOD/g MLVSS. The system was shutdown for 94 days after 3 days operation due to COVID restrictions. At the shutdown, MLSS and MLVSS were 767 mg/L and 550 mg/L, respectively.

After 94 days of shutdown, the MLSS and MLVSS dropped to 353 mg/L and 169 mg/L, respectively with ~70% loss in MLVSS content. To revive the sludge, the CSTR was fed and decanted twice a week; within a week, the values increased to 433 mg MLSS/L and 305 mg MLVSS/L. After 11 days, AnMBR operation was commenced and showed COD removal of 89% indicating quick revival of the sludge (Fig. 2). This was unlike longer revival periods (1-1.5 months) experienced in other anaerobic reactors treating high strength wastewater for somewhat similar shutdown periods (Dong et al., 2010). Our results are similar to that of Renuka et al. (2016) who also observed 85% COD removal efficiency within 2 weeks after a 3-month shutdown in a panelled anaerobic baffle-cum-filter treating municipal wastewater. Over the next 37 days (day 109–145) of AnMBR operation at 24 h HRT, COD removal varied between 89 and 95% which is marginally lower compared to before shutdown. Hwang et al. (2010) also observed a decrease in COD removal efficiency from 70 to 63% after prolonged starvation of anaerobic biomass for 4 months. OLR was similar to pre-shutdown period and F/M ratio varied between 0.25–0.56 gCOD/g MLVSS. At day 138, MLSS and MLVSS were 1082 mg/L and 680 mg/L, respectively.

After completion of 24 h HRT experiments, the AnMBR was again shutdown for 96 days. The sludge storage conditions were similar to that during the first shutdown period. Upon restart, HRT was reduced to 18 h. Post shutdown, MLVSS was 584 mg/L with ~14% content loss. This loss is far less compared to earlier shutdown period of 94 days and sludge revival strategy was therefore not applied. The first shutdown of 94 days occurred during summer, while the later ones were during the monsoons and winter where the lower ambient temperature is likely to have suppressed the microbial activities (Shin et al., 2014) and lysis. Also, anaerobes are known to adapt to repetitive short starvations (3–4 days) (Ye et al., 2018). When operation commenced at 18 h HRT, COD removal was 91% initially that rapidly reached up to 98% in the next two days (Fig. 2). The COD removal is higher compared to that obtained at 24 h HRT. This could be due to improved OLR of 0.24 gCOD/L d compared to 0.17 gCOD/L d at 24 h HRT (Baek et al., 2010; Nilusha et al., 2020). 18 h HRT was concluded on day 286 with average COD removal 95.4±3.4%. MLVSS was 1086 mg/L with MLVSS/MLSS ratio of 0.71, that is marginally higher than at previous HRTs.

After 18 h HRT experiments, the AnMBR was subjected to the next shutdown of 48 days before initiating the 10 h HRT run. There was ~8% loss in MLVSS content over this shutdown which is the lowest compared to that experienced in the earlier shutdown periods, possibly due to the shorter starvation period combined with lower ambient temperature (winter season). As the MLVSS loss was minimal, no separate time for biomass revival was provided and the experiments were initiated maintaining an OLR of 0.47±0.014 gCOD/L d. After 1 week into the operation, COD removal was 93%, subsequently reaching up to 95%. Though HRT was lower, the higher OLR probably helped achieve high organics removal. Average COD removal was 91.7±4.39%. F/M was however very low and varied between 0.09 and 0.14 gCOD/g MLVSS; this showed that AnMBR can achieve high efficiency even at very low F/M (Martinez-Sosa et al., 2011). On day 377, MLVSS was 2088 mg/L with improved MLVSS/MLSS ratio of around 0.8 compared to that observed at 18 h HRT.

In this system, bulk of the COD removal occurs biologically in the CSTR with 87.4±2.8%, 92.4±3.9% and 89.3±6.5% removal at 24 h (post-shutdown), 18 h and 10 h HRT, respectively. Marginal COD removal (< 5%) takes place in the rest of the system (holding tank + membrane filtration). Even though sludge deposition occurs progressively on the membrane surface, it does not appear to contribute noticeably towards COD removal.

Filtration Performance

Figure 3 shows the TMP, flux and total filtration resistance profiles at varying HRTs and backflushing conditions in the SBR-AnMBR.

Fig. 3.

Filtration performance at 24 h, 18 h, and 10 h HRT in the SBR-AnMBR: (a) variation of TMP and flux at different HRTs, (b) Location and magnitude of backflush flux (J_b), and physical cleaning effect on total filtration resistance

The CSTR was intermittently fed for 74 days before switching to SBR mode of operation as explained earlier. Thereafter, the AnMBR system was restarted at 24 h HRT and the flux maintained at 8.8±0.1 L/m² h. The membrane module, previously used in continuous operation, was reused after wiping off surface deposits with a wet sponge and rinsing with water. Once filtration was initiated at 24 h HRT, the initial TMP was 29.3 mbar which was nearly one-third of the TMP value at the conclusion of the earlier continuous run (Fig. S1 in supplementary sheet). This suggests that membrane surface deposits had a major contribution towards TMP build-up. As the experiment continued, TMP increased at a rate of 11.8 mbar/d reaching 63.2 mbar on day 3 (Fig. 3a). The system was shutdown for the next 94 days plus 11 days for sludge revival; before restarting, the membrane module was cleaned with water to remove the surface deposits. Filtration was conducted in filtration-relaxation mode, with backflushing flux (J_b) twice the operating flux, i.e., J_b/J ratio of 2. TMP increased at a rapid rate (15.0 mbar/d) reaching 114.67 mbar within a week, with a corresponding total filtration resistance of 5.14 × 10¹² m^-1. To lower the TMP build-up rate, at day 118, backflush flux was increased with J_b/J ratio of 3. This approach was evidently successful as both TMP build-up rate and total filtration resistance dropped by ~47% and ~30%, respectively. On day 145 of the experiment, the TMP reached 173 mbar. Overall, SBR-AnMBR operation with filtration-relaxation and backflush (3 times operating flux) helped in controlling membrane fouling. The quality of the permeate obtained was excellent with average turbidity 1.2±0.9 NTU and pH 7.9±0.19.

For the next HRT of 18 h after the 96 days shutdown period, a virgin membrane module was used applying both filtration-relaxation and backwashing. Flux was maintained at 11.2±0.11 L/m² h. High TMP build-up rate of 19.56 mbar/d was observed during the first three days despite using a virgin membrane and application of fouling control strategies (Fig. 3a). The possible reasons could be (a) the higher generation of EPS in the CSTR due to stress experienced by the microbes during starvation (discussed in Section 3.3) and (b) development of foulant layer due to surface deposits (Ugarte et al., 2022). Over the next few days, TMP build-up rate reduced and was 4.57 mbar/d until day 272 of operation. The system had to be shutdown for the next 10 days due to COVID-19 issues. Upon restarting on day 283, a low TMP of 36.4 mbar was registered. This was attributed to dislodging of the loose surface deposits when no filtration occurred over the 10 days shutdown. Over the next 4 days, TMP build-up reached 87.56 mbar (day 286). Overall TMP build-up rate at 18 h HRT was 5.2 mbar/d. This is much lower compared to build-up rate observed at 24 h HRT. Total filtration resistance was also correspondingly lower (3.14 × 10¹² m⁻¹). Average permeate turbidity remained low (1.3±0.5 NTU) showing good particulate rejection by the membrane.

The final shutdown period of 48 days was experienced at this stage. Upon restarting the AnMBR operation (day 335), the HRT was reduced to 10 h. A virgin membrane module was employed and the same fouling control strategies used at 18 h HRT were applied here as well. By day 343, the TMP reached 146.67 mbar with a high TMP build-up rate of 42.2 mbar/d and filtration resistance of 3.1 × 10¹² m^-1 (Fig. 3b). Average flux in this period was 20.8±0.8 L/m² h that was close to the desired value (21.05 L/m² h). To lower the TMP build-up rate and to maintain the flux, J_b/J ratio was increased up to 3. Though there was an initial improvement at J_b/J ratio of 3, a sharp flux decline was observed after day 364 with corresponding high filtration resistance of 4.76 × 10¹² m^-1. The membrane module was taken out of the filtration tank and the surface wiped clean using sponge and water (Fig. S4 in supplementary sheet). The membrane surface deposits were collected for subsequent MLVSS and EPS analysis. After this physical cleaning, filtration was restarted at J_b/J ratio of 3; nearly complete flux recovery (99.8%) was observed and the TMP dropped to 75.3 mbar indicating that surface deposition was a major cause for fouling. The trend of TMP build-up with corresponding flux drop continued and further increasing the J_b/J ratio (to 3.5 and 4) was not effective in maintaining the flux. The average permeate flux for days 372-377 at J_b/J ratio 4 declined to 12.0±1.8 L/m² h i.e., nearly half of the desired flux, with corresponding total filtration resistance 5.92 x 10¹² m^-1 at day 377. Thus, the flux at 10 h HRT could not be maintained under the conditions tested in this study. 10 h HRT experiments were terminated on day 377 with the overall average flux for this HRT being 18.4±3.36 L/m² h (~13% less than the desired value). The membrane showed high rejection (~94%) of particulate matter and turbidity of permeate remained low around 1.3±0.4 NTU.

Figure 4 summarizes the filtration results (average flux and TMP build-up rate) at different HRTs with varied J_b/J ratios. For flux up to 11 L/m² h (at 24 h and 18 h HRT), J_b/J ratio of 2 or 3 was adequate to maintain the flux with low TMP build-up rates. The same J_b/J ratios displayed a marginal flux decline in the initial period of operation (day 335–370) at higher flux (21 L/m² h at 10 h HRT). The flux decline was steeper after day 371 and increasing J_b/J ratio (up to 4) was not effective in maintaining the required flux, explained by the high TB-EPS fraction in membrane surface deposits (explained in Section 3.3). Consequently, the experiment was terminated at this stage.

Fig. 4.

Filtration performance summary at 24 h, 18 h, and 10 h HRT, in the SBR-AnMBR: variation of TMP build-up rate w.r.t J_b/J ratio and average flux (*physical cleaning employed before starting at day 365)

EPS and Membrane Fouling

Figure 5 shows the specific EPS profile of bulk sludge suspension in CSTR at variable operating conditions.

Fig. 5.

Specific EPS (proteins and carbohydrates) and TB-EPS/LB-EPS ratio profile in CSTR bulk sludge suspension in the SBR-AnMBR at 24 h, 18 h, and 10 h HRT

EPS, a major fouling component in AnMBR systems, is composed of LB-EPS and TB-EPS (Ding et al., 2015). Specific EPS (i.e., EPS generation per gram MLVSS) was quantified for CSTR bulk sludge suspension and membrane surface deposits; for both the LB-EPS and TB-EPS fractions, both protein and carbohydrate content were determined. Overall average specific EPS in CSTR bulk sludge suspension for different HRTs was between 70 and 80 mg EPS/g MLVSS with no clear relation between average specific EPS concentration and extent of fouling. Similarly, in an AnMBR treating low strength wastewater of 500 mg COD/L, Liu et al. (2012) found no correlation between specific EPS content and fouling in operation at both low and high F/M ratios. In our work, when the AnMBR system was reactivated after each shutdown period, specific EPS concentration in CSTR bulk sludge suspension was initially high but became steady within 2 weeks of operation regardless of the operating HRT (Fig. 5). The effect was reflected in the higher TMP build-up rate in the initial week of operation at each HRT after the shutdown period. Palmarin et al. (2020) investigated fouling propensity of microbial byproducts after starvation and reported rapid fouling rate for the initial 4 days after reactivation. For longer starvation periods, beyond a stage, biomass lysis can occur (as indicated by MLVSS content loss during each starvation period), producing high molecular weight EPS possessing high fouling potential (Wu and Lee, 2011).

Protein to carbohydrate (P/C) ratio for HRT of 24 h (post-shutdown), 18 h and 10 h was 4.3±1.6, 6.5±1.6, and 2.0±0.4, respectively. Data for 24 h HRT (pre-shutdown) has been excluded as the system was operated for limited period. Lower P/C ratio corresponds to higher total filtration resistance in the following order of HRT: 10 h > 24 h > 18 h. There is no agreement in literature with both carbohydrates (Ni et al., 2011) and proteins (Huang et al., 2011) identified to induce relatively more fouling. EPS concentration affects membrane fouling differently depending upon feed composition, reactor configuration, operating conditions and membrane properties (Ni et al., 2011). High concentration of carbohydrates and increased fouling rate was observed by Wu and Lee (2011) at very low F/M ratio in aerobic SBR. In this work, at all HRTs, F/M ratio was anyway low; further, over a 24 h cycle, the 10 h HRT creates 14 h of starvation conditions. The alternate low F/M ratio and starvation conditions therefore altered the EPS composition with higher carbohydrate content at 10 h HRT.

EPS has a multi-layered structure. TB-EPS is strongly associated with sludge flocs and helps in maintaining flocs structural integrity (Ramesh et al., 2007). Average TB-EPS/LB-EPS ratio for HRT of 24 h (post-shutdown) and 18 h were highly variable, 44.7±78 and 24.45±23.99, respectively. But during 10 h HRT, average TB-EPS/LB-EPS ratio was consistently low (2.42±0.75) (Fig. 5). On day 364, when there was a considerable decline in permeate flux despite increasing backflush flux, the membrane was taken out of the tank, surface deposits were collected and analyzed for specific EPS. MLVSS of membrane surface deposits was 1.14 g/L and specific EPS was 58.4 mg/g MLVSS. Proteins accounted for 70% of the specific EPS fraction which is equivalent to specific proteins concentration in CSTR bulk sludge suspension. But the TB-EPS/LB-EPS ratio in surface deposits (5.3) was twice the value for the CSTR bulk sludge suspension (2.42). LB-EPS contains bound water in abundance and can alter the sludge characteristics (Li and Yang, 2007). High TB-EPS content in surface deposits could be due to the loss of bound water in LB-EPS under compression in the deposited layer, converting LB-EPS into TB-EPS and resulting in tighter link to the membrane surface (Liu et al., 2012) (Fig. 6). This could be one reason that permeate flux could not be recovered even at higher backflush at 10 h HRT. Additional studies, in terms of the scanning electron microscopy (SEM) and Fourier-transform infrared spectroscopy (FTIR) of the used (fouled) membranes would be required to confirm this hypothesis. This aspect is being investigated. Yet another reason could be that backflushing was limited to 4 min once every HRT cycle in this work. Ugarte et al. (2022) found frequent backflushing (1 min backflushing after every 9.5 min) to be effective in lowering fouling rate for waste-based ceramic membrane (made from clay, almond shell and chamotte), operated at 15 L/m² h in aerobic MBR. Regular backflush has the tendency to reduce surface deposition rate (Smith et al., 2006); the effect can be enhanced by chemicals addition during backflushing to weaken EPS bonding to membrane surface (Yu et al., 2019). These approaches need to be examined.

Fig. 6.

Illustration of conversion of LB-EPS to TB-EPS under compression

Membrane Characterization

Characterization (porosity, bulk density, and median pore diameter) of used (fouled) membranes was performed and compared with the virgin membrane. Table 3 summarizes the results and Fig. 7 shows the pore size distribution.

Table 3.

Characterization of virgin and used membranes

Property	Virgin membrane*	Used membrane from SBR operation at various HRTs (h)**
		48 h & 24 h	18 h	10 h
Porosity (%)	43.37±7.17	38.23±2.11	44.46±0.83	40.86±3.09
Bulk density (g/mL)	1.38±0.18	1.39±0.08	1.35±0.01	1.43±0.08
Median pore diameter (μm)	7.14±1.19	5.78±1.03	6.51±1.07	7.17±0.95
Total intrusion volume (mL/g)	0.325±0.097	0.275±0.031	0.329±.009	0.287±0.038

* average of 6 samples; ** average of three samples

Fig. 7.

Plot of log differential intrusion vs pore diameter of virgin and used membranes

As mentioned earlier, the same membrane module was used for experiments at HRTs of 48 h (continuous mode) and 24 h (SBR-AnMBR); also, no fouling control strategies were applied in operation at 48 h HRT. Compared to the virgin membrane porosity (43.37±7.17%), the used membrane displayed lower porosity (38.23±2.11%), marginally higher bulk density, reduced median pore diameter and lower pore volume (as indicated by reduction in total mercury intrusion volume). This module showed the maximum peak shift in the log differential intrusion vs pore size plot (Fig. 7). For the experiments at 18 h and 10 h HRT, virgin membrane module with fouling control strategies were employed; there was no noticeable deviation between the properties of the virgin and used membranes. Thus, higher total filtration resistance experienced at 10 h HRT was attributed to surface deposits with high TB-EPS concentration.

Conclusions

This work evaluated the performance of AnMBR system equipped with SBA waste-based ceramic membrane for treating low strength simulated municipal wastewater at variable HRTs and intermittent/complete starvation conditions. It was concluded that SBR mode of operation is suitable for this AnMBR system. Variations in the HRT and intermittent starvation did not affect the average COD removal indicating the AnMBR could tolerate unregulated or seasonal influent variations. However, starvation led to high EPS generation and correspondingly higher membrane fouling. Fouling control strategies (filtration-relaxation and backflushing) were able to minimize the surface deposition up to an extent and maintain the flux, especially at lower fluxes (below 11 L/m² h). Nevertheless, more effective approaches to control surface deposition are required for maintaining higher flux (>11 L/m² h). Overall, these SBA waste-based ceramic membranes appear promising in AnMBRs treating low-strength wastewater under varying operation conditions. Further research is required to understand sludge properties and its correlation with the membrane fouling behaviour especially upon system reactivation after shutdown. This would require monitoring changes in CSTR parameters such as fatty acids content, alkalinity and EPS during the starvation period. Additional characterization of fouled membranes using techniques such as SEM and FTIR is also recommended to obtain insights on the membrane foulants and thus selection of corresponding fouling mitigation techniques.

Author Contribution

S. Dhiman: Data curation, Formal analysis, Methodology, Writing—original draft. M. Balakrishnan: Conceptualization, Funding acquisition, Project administration, Supervision, Writing—review & editing. V. Naddeo: Funding acquisition, Project administration. N. Ahsan: Supervision.

Funding Information

Funding from the project n. INT/Italy/P-17/2016 (SP) and IN17GR09 from Department of Science and Technology, Ministry of Science and Technology, Government of India and Italian Ministry of Foreign Affairs and International Cooperation is gratefully acknowledged. S. Dhiman is grateful to University Grants Commission for Junior and Senior Research Fellowship (UGC ref. no. 3781/NET-DEC2018).

Data Availability

Not applicable.

Declarations

Ethical Approval

No applicable.

Consent to Participate

Not applicable.

Consent for Publication

All co-authors agree to the publishing of this work in Water, Air, & Soil Pollution.

Competing Interests

The authors declare no competing interests.