Introduction

Hybrid membrane bioreactors (HMBRs) are specific membrane bioreactors that combine certain materials, agents, or equipment with a bioreactor (Borea et al. 2017; Mei et al. 2014; Palmarin and Young 2019). They have many advantages, including biostability, a small footprint, and a stable effluent quality despite high organic load shock (Aslam et al. 2017; Ensano et al. 2019; Park et al. 2017). However, with prolongation of the HMBR operation time, some inorganic substances, organic substances, and microorganisms including fungi, bacteria, and viruses gradually adhere to or are deposited on the inner and outer surfaces of the micromembrane, resulting in elevated filtration drag force and ultimately membrane fouling. Membrane fouling has several negative consequences, which include not only a reduction in the water production rate but also an increase in the complexity of the operation procedure. Further, the maintenance and operation cost increases, as fouling requires periodic stoppage of the membrane module from running.

HMBRs have drawn increasing research attention in the last few years to improve the design and operation conditions to avoid membrane fouling (Asif et al. 2018; Bilal et al. 2013; Jin et al. 2013; Liu et al. 2012; Loulergue et al. 2014). Despite extensive research on the membrane fouling process, including controlled experiments and simulations, it remains an inevitable outcome during HMBR operation. However, the process of membrane fouling can be slowed down by varying degrees by properly controlling the operating conditions of the reactor. Among the control parameters, sludge retention time (SRT) has been verified to have a negligible effect on the biodegradability of the bioreactors. SRT can dominantly impact substrate concentrations, including microsecretions, biomass community, and properties of activated sludge (Gong et al. 2019; Huang et al. 2019; Huang et al. 2000; Zhang et al. 2017). Unfortunately, the association of SRT with extracellular polymeric substance (EPS) content, hydraulic resistance, and membrane fouling in HMBRs remains to be determined.

Moreover, some scholars (Cai et al. 2019; Qu et al. 2013) reported that in the process of treating domestic sewage, the treatment effects of a bioreactor are largely affected by the change in hydraulic retention time (HRT), which has a noticeable impact on mixed liquor, especially the biomass characteristics in the reaction tank, which in turn change the procedure of membrane fouling (Cai et al. 2019; Gkotsis et al. 2018; Teng et al. 2019; Yu et al. 2017). Some researchers deem that fluctuations in the HRT alter the organic loading rate (OLR); thus, the OLR varies inversely with HRT (Rahman et al. 2014; Zhang et al. 2018). However, there has been minimal research concerning the interaction between SRT and HRT, and their relative contributions to the membrane fouling mechanism. Therefore, in this study, we investigated the interaction between SRT and HRT and their influence on the biomass characteristics and membrane fouling of HMBR, including variations in the EPS content and membrane resistance.

EPS is a general term used for a variety of high molecular weight substances, and the relative molecular mass of EPSs exceeds 10,000 Da. EPSs are mainly produced by biofilm-forming bacteria through metabolism and polymerization. EPSs are generally divided into soluble EPS (S-EPS) and bound EPS (B-EPS), with the latter being subdivided into loosely bound EPS (LB-EPS) and tightly bound EPS (TB-EPS). To date, EPS and soluble microbial products (SMPs) have been considered to be the dominant substances responsible for membrane fouling (Chang and Lee 1998; Liu and Wang 2014; Ni et al. 2017). The deposition of EPS and activated sludge flocs onto the membrane surface lead to the production of a cake layer and reduction of porosity. Some scholars (Kimura et al. 2005) found that excitation/emission (Ex/Em) matrix (EEM) was effective to investigate and analyze the fouling process in membrane bioreactors. They identified a protein-like substance and humic acid-like substances at different Ex/Em by analyzing the SMP and foulants of an MBR running at various HRTs. They also indicated that SRT plays a dominant role in the characteristics of S-EPS in reactors. In the present study, both forms of EPS were detected by measuring the EPS contents in mixed liquor and biofilms of HMBRs.

Previous single-factor experiments have demonstrated that both HRT and SRT can affect EPS constitution as well as the membrane fouling process (El-Fadel et al. 2018; Liu et al. 2010). To establish optimized operating conditions and identify factors that are the main contributors of HMBR fouling, a response surface methodology (RSM) approach, which is a comprehensive experimental technique, was used. Combining systematics with statistics, an RSM approach establishes a surface model of continuous variables and also analyzes factors of the experimental design and their interactions (Antonopoulou et al. 2012; Chen et al. 2011). Thus, this approach includes experimental design, modeling, and model checking and seeks the best combination conditions along with other experimental and statistical techniques. Response surface optimization considers the random errors in experiments and fits the complex undiscovered function relations in a small range with a simple first- or second-order polynomial model. Thus, RSM serves as an efficient method to cut the expense of experiments, improve the quality of experiments, and solve practical problems in production processes, especially when the system characteristics are affected by a large number of variables. RSM approaches have been widely applied in the chemical industry, food engineering, and pharmaceutical engineering (Bhateria and Dhaka 2019; Manwar et al. 2018; Polat and Sayan 2019; Sun et al. 2019; Thakur et al. 2019). Increasing attention has been focused on industrial wastewater treatment (Rout et al. 2017; Zheng et al. 2017) instead of domestic wastewater treatment by HMBR technology. Both Box-Behnken designs (BBDs) and central composite designs (CCDs) are commonly used in RSM. BBD was selected in this study as it can be used with a lower number of experiments than CCD with the same number of factors. Therefore, we sampled five HMBRs varying in SRTs and HRTs, and determined the EPS contents in the mixed liquor and biofilm samples, along with their effects on membrane resistance. The findings of this study and the established model determined through BBD for predicting EPS may serve as a useful guide for the determination of the optimal HMBR conditions to help prevent fouling and maintain long-term membrane function.

Materials and methods

Raw wastewater

Effluent from the primary settling tank at the urban sewage treatment plant in Xuzhou, Jiangsu Province, China (longitude 117.18°, latitude 34.27°) served as the raw water in this investigation. Domestic wastewater after precipitation was used as the raw feed wastewater over a 6-month operating period, for a better analysis of the effluent organic matter characteristics under a stable condition. Table 1 lists the composition of the raw feed wastewater.

Table 1 Characteristics of the raw feed wastewater
Full size table

HMBR operational conditions

Five identical HMBRs worked under different SRTs and HRTs over a period of 6 months. All the HMBRs had an effective volume of 100 L; their design is schematically presented in Fig. 1, and the operating conditions are shown in Table 2. A polyvinylidene (PVDF) hollow fiber membrane with an average pore size of 0.10 mm was immersed in the mixture. Polyethylene K3 biological fillers, with a porous interior cylinder (diameter 25 mm and height 12 mm), were placed in the tank, with a 50% (volume fraction) dose. Air was transferred into the liquid phase through the diffusers at the bottom of the vessel with a velocity of 2 m3 per membrane area (m2) per hour. This method of aeration not only provides oxygen for microbial metabolism but also able to scour the membranes and alleviate the attachment of deposits. The concentration of dissolved oxygen (DO) in the mixture was controlled at approximately 4 mg L−1 by aeration from the diffuser on the bottom through an air flow meter. The temperature of the mixture was controlled at approximately 20 °C using a heating rod. The membrane module was operated by a self-priming pump, under constant water flux of 10.0, 12.5, and 16.7 L m−2 h−1, respectively, according to the different HRTs with a 10-min operation cycle (8 min on, 2 min off). When TMP achieved 20 kPa, the membrane module was considered to be contaminated and then was brought out for chemical cleaning by using Deng’s method (Deng et al. 2016).

Fig. 1
figure 1

Schematic of the HMBR system

Full size image
Table 2 HMBR operating conditions
Full size table

Analysis methods

Water quality measurement

To determine the quality of the raw feed wastewater, the temperature and pH values were determined by a pH analyzer (pH Meter, ARCO Electronics LTD, Guangdong). CODCr, BOD5, SS, NH4+-N, TP, and TN were analyzed according to standard methods (APHA, 2005). Our previous studies showed that HMBR performed well in treating domestic wastewater (Li et al. 2019), and thus, the optimum operating condition based on effluent quality analysis were not a concern in this study. Instead, HRT and SRT were applied to analyze the mechanism of membrane fouling influence.

EPS extraction and detection

Five milliliters of the mixed liquor was filtered through a 0.22-mm pore-size filter paper, and the DOC of the filtrate was considered to be S-EPS. LB-EPS quantity, calculated by subtracting the S-EPS amount from the EPS amount of the filtrate, was obtained by determining the supernatant DOC value after high-speed centrifugation. The specific operating conditions were described previously (Li et al. 2019). To supplement the loss of water after the previous filtration, 5 mL of pure water was injected into the centrifuge tube. A water bath under 80 °C was used to heat the well-mixed solution for 30 min. The sample was then centrifuged at 8311.68g for 20 min, and the supernatant was filtered through a 0.22-μm pore-size filter paper. This filtrate was used to measure the EPS components of mixed liquor.

The biofilm slicked to the suspended filter was scraped and enriched smoothly by a plastic sheet. A mixture of a certain concentration composed of the enriched biofilm and deionized water was placed into a test tube to detect the quantity of the EPS. Because S-EPS was absent in the biofilm, only LB-EPS and TB-EPS were measured, which were extracted from the biofilms in the same manner as described above for the mixed liquid. The amount of EPS was typically quantified by milligrams per unit mass volatile suspended solid (VSS).

The EEMs of the obtained EPS from the mixed liquid and biofilm were assessed using a fluorescence spectrophotometer (Shanghai Sanke Electrical Appliances Co., Ltd.). The excitation wavelength ranged from 200 to 500 nm, and the emission wavelength ranged from 280 to 500 nm; 5-nm increments for both Ex and Em models were applied to obtain the spectra.

Test of membrane resistance

After each operation cycle, the contaminated membrane module was brought out from the tank and washed by pure water to remove the biofilm and sludge attached to the membranes. The membrane resistance was calculated by laying the membrane module in deionized water and operating it at various fluxes. Total resistance (Rt) during membrane filtration mainly comprises inherent resistance (Rm), pore resistance (Rp), and cake layer resistance (Rc), which were respectively calculated by using Lee’s method (Lee et al. 2003).

RSM and statistical analysis

The RSM was developed by Box and collaborators in the 1950s (Polat and Sayan 2019). Mathematical and statistical techniques are used in RSM to match empirical models to experimental data. To achieve this goal, linear or square functions were used to represent the system studied and, consequently, to search the optimum solution. Box and Behnken (1960) proposed methods to opt points from three-level factorial arrays, which are called BBD and can make models more efficient and profitable compared to their corresponding 3k designs, mainly due to consideration of a large number of factors. More information on the theoretical principles of RSM and procedures for its function can be found in the report by Dun et al.’s (2014). According to the BBD principle, the response surface experiments of three factors and three levels were designed using Design Expert software (Version 12) with EPS content and the membrane resistance coefficient as the response values Y1 and Y2, respectively. HRT (A) and SRT (B) were selected as the two influencing factors. Table 3 shows the factors and levels used in the response surface analysis.

Table 3 Factors and levels of response surface analysis
Full size table

The experimental data of EPS content shown in Table 3 were fitted to the model and tested using analysis of variance (ANOVA), and the quadratic regression models of A and B for EPS content were obtained.

Results and analysis

Response surface design

EPS, as an important component of membrane pollutants in an HMBR, together with the membrane resistance coefficient, is an important comprehensive index to characterize the degree of membrane fouling. HRT and SRT are important indicators of membrane fouling.

According to the experimental factors and levels established using Design Expert software, the experimental values of surface modification conditions optimized by RSM are shown in Table 4, including a total of 15 experimental points.

Table 4 RSM experimental design and results of membrane fouling
Full size table

Regression model and ANOVA

The final equations of the regression models with the coded factors (Table 3) were as follows:

$$ \mathrm{EPS}\ \mathrm{content}=87.67-14.53A-23.44B-18.577 AB $$
$$ \mathrm{Membrane}\ \mathrm{Resistanceresistance}=5.24+0.1587A-1.38B-1.36 AB $$

The ANOVA results for model fitting are shown in Tables 5, 6, 7, and 8. For EPS, the F value of the model was 11.67, demonstrating that the model is statistically significant (P = 0.0010). The lack of fit F value of 1.01 signified that the lack of fit is not significant relative to the pure error probability. Specifically, there was a 48.50% chance that a lack of fit F value this large could occur due to noise. Nonsignificant lack of fit is good.

Table 5 ANOVA for response surface linear model: response 1 EPS content
Full size table
Table 6 ANOVA for response surface linear model: response 2 membrane resistance
Full size table
Table 7 Test of significance for coefficients of the quadratic model: response 1 EPS content
Full size table
Table 8 Test of significance for coefficients of the quadratic model: response 2 membrane resistance
Full size table

Similarly, for membrane resistance, B was found to be significant (P = 0.0119), with a lack of fit F value of 2.80 indicating no significant lack of fit relative to the pure error, and a 12.14% chance that this lack of fit could be due to noise.

ANOVA of the regression model showed that the linear terms A and B of the model could significantly influence response factor 1, and linear term B could significantly influence response factor 1 but not factor 2. In the quadratic term, the effect of AB could significantly correlate with response factors 1 and 2.

3D response surface graph and contour graph analysis

Effect of HRT and SRT on EPS

The effect of HRT and SRT on the EPS contents is presented in Figs. 2, 3, 4, and 5 in the form of response surface graphs. These results demonstrated that when the HRT is less than 10 h and is kept constant, the change in EPS contents decreases slowly with the extension of SRT. However, when the HRT is greater than 10 h, the EPS content exhibits an initial decrease, followed by an increase with the extension of SRT. Higher fouling rates in terms of EPS content can be ascribed to higher pollutant loading rates (increased F/M) and inadequate retention time with shorter HRT (Deng et al. 2016). As the supply of organic matter exceeds the demand of consumption, immobilized microorganisms grow at the maximum speed on the carrier surface, and the yield of biofilm polysaccharide and protein increases (Cai et al. 2019; Gkotsis et al. 2018; Huang et al. 2019). The B-EPS is closely related to growth and is generated in a direct proportion to substrate utilization (Qu et al. 2013). Therefore, the decreased HRT facilitates the generation of bound EPS and eventually increases the total content of EPS (Hao et al. 2016).

Fig. 2
figure 2

Contour graph of HRT and SRT on EPS content

Full size image
Fig. 3
figure 3

Contour graph of HRT and SRT on membrane resistance

Full size image
Fig. 4
figure 4

3D response surface graph of HRT and SRT on EPS content

Full size image
Fig. 5
figure 5

3D response surface graph of HRT and SRT on membrane resistance

Full size image

Excessive and low F/M affects the microbial count, which in turn may affect the production of EPS. The change in SRT may lead to a change in MLSS, MLVSS, and MLVSS/MLSS. A high MLVSS/MLSS ratio after a long SRT (30 days) indicates high viability of biomass from the perspective of assimilation and decomposition (Li et al. 2006; Liu et al. 2012).When SRT is too short or F/M is too high, organic matter, and nutrients are converted into EPS before they can be completely consumed by microorganisms (Burman and Sinha 2018; Choi et al. 2013; Li et al. 2019); on the contrary, endogenous respiration of microorganisms can intensify and produce a substantial amount of EPS. In this study EPS increased with decreasing SRT, indicating that there may be more utilization-associated SMPs at short SRTs (high F/M ratio). The microbial activities of biomass were positively correlated with bound EPS and SMP, which is consistent with the study of Choi’s studies (Choi et al. 2013).

Effect of HRT and SRT on membrane resistance

The total hydraulic resistance for each membrane module was calculated based on Eq. (1).

$$ {R}_T={R}_M+{R}_C+{R}_P $$
(1)

where RT (m−1) is the total membrane resistance, RM (m−1) is the intrinsic membrane resistance, RC (m−1) is the cake resistance, and RP (m−1) is the pore blocking resistance.

As shown in Figs. 2, 3, 4, and 5, RT showed the same pattern of change as that of EPS content to a great extent. This indicated that EPS is closely related to membrane resistance, which is consistent with previous reports (Dong et al. 2018; Teng et al. 2019). The main reason for the increase in membrane resistance is that EPS is adsorbed on the surface of the membrane or inside the channel (Nouha et al. 2018; Tansel et al. 2006). With the increase in adsorption capacity, the pore size of the membrane becomes increasingly and more severely blocked, resulting in a continuous increase in membrane resistance (Chabaliná et al. 2013). Among the components of EPS, S-EPS, and LB-EPS play an important role in membrane fouling. Reduction of S-EPS may lead to the reduction of RP and RC (Borea et al. 2017). Reduction of LB-EPS can ameliorate the flocculation and sedimentation performance of activated sludge (Su et al. 2014). Flocculation performance is improved when the number of colloids in water decreases and RP decreases further. Therefore, RC, RP, and RT gradually decrease with the reduction of S-EPS and LB-EPS.

Lower operating flux (longer HRT) generated small drag force for the particles, which could help form a loose gel layer rather than a compact cake layer (Gkotsis et al. 2018). Appropriate extension of SRT was considered to be conducive to the accumulation of total biomass, but continuous prolongation of SRT (lower F/M) aggravates endogenous microbial respiration and causes elevation of EPS levels. The increased EPS would strengthen the polymer bridging of sludge flocs and embed small flocs in the EPS matrix, thus hindering deflocculation and small floc production (Huang et al. 2019; Liu et al. 2012). Thus, our results suggested that appropriate extension of HRT and SRT is conducive to controlling membrane fouling; however, when the HRT exceeds 8 h, an SRT of over 30 days may have the opposite effect.

Determination and verification of optimum operating conditions

Based on the above results, the optimum conditions of HMBR were determined to be an HRT of 9.53 h and SRT of 29.77 days, corresponding to an EPS content and membrane resistance of 39.80 mg g−1 VSS and 3.21 × 1012 m−1, respectively. Considering the feasibility and convenience of practical operation, the simulated operating conditions were revised as follows: HRT of 8 h and SRT of 30 days. To test and verify the reliability of the simulation based on RSM, membrane fouling control experiments were carried out under the optimum operating conditions. The EPS content of the three repeated experiments was found to be 43.17 mg g−1 VSS and the membrane resistance was 3.46 × 1012 m−1, which are consistent with the predicted values of the model. Therefore, the optimal operating conditions obtained by the RSM were confirmed to be accurate and reliable, with potential practical value.

Characteristics of membrane fouling components under the optimal operating conditions

The EEM analysis was further applied to investigate EPS in the components and structures of the suspended carrier, membrane, and sewage sludge under the optimum operating conditions (Fig. 6). Three major peaks could be read from the fluorescence spectra. The first major peak was located at the Ex/Em wavelengths of 340–360/420–445 nm (peak A), which is considered a visible humic acid-like fluorescence. The second major peak was located at the Ex/Em wavelengths of 250–260/430–460 nm (peak B), which is regarded as a fulvic acid-like or unknown fluorescent component of humus. Moreover, a third major peak was observed at Ex/Em of 240–260/570–600 nm (Peak C), which was used to characterize the unknown fluorescent components. The peaks located at Ex/Em of 225/340 nm and 230/430 nm can be used for online analysis and real-time monitoring of NH4+-N and TN in municipal domestic sewage. Since no such peaks were detected in the three samples in this research, the nitrogen removal effect of the HMBRs could be considered to be good. The characteristic humic acid peak (peak A) showed a relatively high content in suspended carriers, followed by the membrane and sludge. The main components of humic acid are high polymers of aliphatic and aromatic groups. Functional groups, such as carboxyl and phenol groups, which originate from ion dissociation and exchange with hydrogen ions in solution, result in high negative charges. This consequently weakens the flocculation ability of the sludge, and is not conducive to sludge granulation. Optimum SRT should be confirmed between 20 days and 30 days to improve settling performance.

Fig. 6
figure 6

Excitation emission matrix (EEM) spectra of membrane foulant of suspended carrier (a), membrane (b), and sludge (c) in wastewater treatment

Full size image

Conclusion

Based on a previous single-factor experiment, a quadratic multiple regression model of the influence of HRT and SRT changes on the EPS content and membrane resistance of HMBRs was established using an RSM method. The influences of the two factors were pronounced, as was their interaction. The model was confirmed to be reasonable and reliable. According to the model optimization and practical feasibility, the optimal HRT is 10 h and SRT is 30 days. Under the optimal working condition, the EPS content was 43.17 mg g−1 VSS and the membrane resistance was 3.46 × 1012 m−1, which indicated that the empirical model established by the RSM has a better regression fit. The EEM results further indicated that humic acid-like and fulvic acid-like were the dominant composition of EPS in the HMBR, which likely conducive to format and keep structural stability of the biofilm. Later studies for membrane fouling control can be carried out from the perspective of controlling the secretion of humic acids and humic acids by microorganisms.