This study focuses on the use of fluidized glass beads as turbulence promoters in a laboratory-scale anaerobic membrane bioreactor treating municipal wastewater at 20 °C. The addition of fluidized glass beads into an external tubular ceramic membrane enabled the operation at low crossflow velocities of 0.053–0.073 m/s (mean fluxes between 5.5 and 9.7 L/(m2·h)) with runtimes >300 h. Glass beads with a diameter of 1.5 mm were more effective than smaller ones with a diameter of 0.8–1.2 mm. Increasing the bed voidage from 74 to 80% did not show any beneficial effect. As scanning electron microscope examination showed, the fluidized glass beads damaged the used membrane by abrasion. The overall total chemical oxygen demand (COD) removal was between 77 and 83%, although mean hydraulic retention times were only between 1.3 and 2.3 h. The production of total methane was increased about 30% in comparison to the bioreactor without membrane. The increased methane production is presumably attributed to biological conversion of rejected, dissolved and particulate organic matter. The total required electrical energy was predicted to be about 0.3 kWh/m3.

INTRODUCTION

The anaerobic treatment of municipal wastewater still represents a challenge for wastewater engineers. At low temperatures and substrate concentrations the anaerobic membrane bioreactor (AMBR) might be a promising technology to achieve high quality effluents (Smith et al. 2012; Stuckey 2012; Lin et al. 2013). However, control of concentration polarization and fouling, respectively, is a crucial challenge in launching the AMBR for municipal wastewater treatment (Smith et al. 2012; Stuckey 2012; Lin et al. 2013). A promising approach might be the addition of fluidized particles, such as resin beads (Imasaka et al. 1989; Düppenbecker et al. 2016), glass beads (Stuckey 2012), or granular activated carbon (GAC) (Kim et al. 2011; Seib et al. 2016) as turbulence promoter. Furthermore, previous studies showed that fluidized glass and steel beads effectively promoted the turbulence in tubular ultrafiltration (UF) membranes (Rios et al. 1987; Noordman et al. 2002). In the present study, a novel AMBR configuration is introduced using fluidized glass beads as turbulence promoters in a tubular ceramic membrane module coupled to an anaerobic fluidized bed reactor.

The impact of fluidized bed parameters (bed voidage, superficial velocity, particle diameter and density, particle shape, etc.) on fouling mitigation is not yet understood in detail (Aslam et al. 2017). Bixler & Rappe (1970) stated that fluidized particles might mitigate fouling (1) by decreasing concentration polarization via increased turbulence, caused by the mixing action of the particles and (2) by scouring via the mechanical action of fluidized particles. Previous studies showed that the bed voidage (the ratio of the void volume to the total volume of the fluidized bed) represent a crucial parameter to optimize fouling mitigation by means of fluidized particles. Rios et al. (1987) used fluidized steel beads during UF of macromolecular solutions and observed an optimum bed voidage (highest flux at constant pressure filtration) of 68%. This might be explained by the fact that the wall-to-liquid mass transfer in fluidized beds depends on bed voidage and reaches a maximum between 64 and 72% according to the correlation of Schmidt et al. (1999). However, it should be noted that the correlation of Schmidt et al. (1999) is only valid if diffusion is the driving force for mass transfer and does not account for the scouring effect of the fluidized particles. Moreover, de Boer et al. (1980) observed an optimum bed voidage between 65 and 80% during filtration of cheese whey using fluidized glass beads with diameters between 0.7 and 3 mm. Mikulášek & Filandrová (1995) observed an optimum at a bed voidage of 80% during filtration of alumina suspensions. Furthermore, de Boer et al. (1980) showed that glass beads with a diameter of 3 mm were more effective than glass beads with a diameter of 2 and 0.7 mm. In summary, there exists little information regarding the impact of the bead diameter and bed voidage on mass transfer during the filtration of complex media, such as anaerobically treated wastewater. Therefore, a further focus of the present study is to determine the most promising fluidized bed parameters during UF of anaerobically treated municipal wastewater using fluidized glass beads as mechanical cleaning agent and turbulence promoters, respectively.

MATERIAL AND METHODS

Experimental setup

Figure 1 shows a schematic diagram of the laboratory setup. An anaerobic bioreactor was coupled to a tubular ceramic UF membrane (ID = 16 mm; L = 500 mm; A (membrane surface area) = 0.025 m2; 150,000 Da; ZrO2/Al2O3; atech innovations, Germany). The membrane was mounted vertically in a PVC housing and operated in inside-out mode. The transmembrane pressure (TMP) was generated by suction on the permeate side. Glass beads (soda-lime glass; density = 2,500 kg/m3; Worf Glaskugeln, Germany) with a diameter of 0.8–1.2 or 1.5 mm were added as turbulence promoters and were fluidized by the upflow to the top of the membrane module. The height of the fluidized bed was expanded to 610 mm in each mode. The bed voidage was varied by changing the mass of glass beads and the crossflow velocity, respectively. Since the fluidized bed showed a sharp boundary, further efforts to avoid discharge of the glass beads from the membrane module were not required. To enable fluidization, a wire cloth was fixed at the bottom of the membrane module. Furthermore, glass beads (2 mm and 4 mm) were filled in as support layer. The anaerobic bioreactor was a fluidized bed bioreactor described in detail by Düppenbecker & Cornel (2016), with a volume of about 317 mL. This volume was defined as reaction zone for calculating the organic loading rate (OLR) and hydraulic retention time (HRT). Since no biofilm formation on the fluidized glass beads occurred, biological degradation was attributed to the fluidized bed bioreactor. A cylindrical settler (H = 150 mm, D = 72 mm) was attached on the top of the reactor to catch carry-over GAC. The main function of the second settler (H = 180 mm, D = 72 mm), placed between bioreactor and membrane module was to distribute the volume flows. Furthermore, in case of GAC and glass beads discharge the settler prevented clogging of the peristaltic pumps. The total volume of the AMBR (excluding membrane module and recirculation lines) was about 1,600 mL. GAC (Epibon A, 8 × 30 mesh, Donau Carbon, Germany) was chosen as carrier material. GAC was sieved to sort out particles larger than 1 mm.
Figure 1

Schematic diagram of the laboratory setup (dimensions in mm). The height of the fluidized bed (610 mm) was constant in each phase.

Figure 1

Schematic diagram of the laboratory setup (dimensions in mm). The height of the fluidized bed (610 mm) was constant in each phase.

Peristaltic pumps were used for recirculating (ProMinent, Germany) and introducing feed or drawing permeate, respectively (IDEX, USA). The recirculation rates were adjusted using flow meters (GEMÜ, Germany). Pressure was monitored at the top and the bottom of the membrane module and at the permeate line using pressure transmitters (PAA-33X, Keller). The permeate flow rate was monitored gravimetrically using a balance (Entris, Sartorius, Germany). The biogas flow rate was measured volumetrically (MilliGascounter, Ritter, Germany). ORP and pH (Sensolyt, WTW, Germany) electrodes were installed in the settler and connected to a pH meter (pH 191, WTW, Germany).

Operating conditions

The AMBR was fed with raw municipal wastewater (160 μm pre-screened) from the wastewater treatment plant Darmstadt Süd (Darmstadt, Germany). Wastewater was stored in a refrigerated, stirred feed tank at about 5–10 °C and replaced weekly. The mean total chemical oxygen demand (COD) was 446 mg/L, and a mean fraction of 65% was particulate COD. The feed concentrations were subject to large fluctuations between 155 and 892 mg/L. All experiments were carried out at 20 °C in a climate chamber. Before the test phase, the biofilm on the GAC was cultivated for more than 6 months under comparable operating conditions as described in the present study. The volume of the settled GAC (covered with biofilm) at start of operation and start of mode IIIw/B was 129 mL (corresponding to a fixed bed height of about 0.3 m in the bioreactor). The volume was determined in a 100 mL graduated cylinder (diameter 27 mm) filled with tap water, after sedimentation to constant volume. The upflow velocity of the fluidized bed reactor was adjusted between 28.5 and 36.5 m/h (corresponding to a volume flow between 13.0 and 15.5 L/h).

The study was divided into three phases. Each phase started with a filtration test using fluidized glass beads (marked by w/). During phase I, glass beads with a diameter of 0.8–1.2 mm were tested (bed voidage = 74%). Afterwards, during phase II, the bed voidage was increased to 80% by decreasing the mass of glass beads and increasing crossflow velocity. Finally, glass beads with a diameter of 1.5 mm were tested at a bed voidage of 74% (phase III). At the end of each phase, a run without glass beads was carried out (marked by w/o). Due to the fast increase in TMP, the runs without fluidized glass beads only took about 24 h. The operating conditions are compiled in Table 1. The OLR varied in each mode as a result of fluctuating COD influent concentrations and fluxes, respectively. When required, the membrane was cleaned with a NaOCl solution (195 ppm, 1 h, 20 °C) and citric acid (1% monohydrate by mass, 1 h, 20 °C), however, the membrane was not replaced.

Table 1

Operating conditions of the AMBR

Phase I
 
II
 
III
 
Mode Iw/ Iw/o IIw/ IIw/o IIIw/A IIIw/B IIIw/o 
Operating time (d) 0–56 56–57 57–77 77–78 78–111 111–153 153–154 
Bead diameter (mm) 0.8–1.2 – 0.8–1.2 – 1.5 1.5 – 
Mass of beads (g) 80 60 80 80 
Bed voidage (%) 74 100 80 100 74 74 100 
Crossflow velocity (m/s) 0.053 0.054 0.058 0.058 0.071 0.073 0.073 
Flux (mean) (l/(m2 h) 6.5/7.7/6.7/5.5 6.3 6.3 5.7 7.2/9.7 7.9 6.8 
HRTa (mean) (h) 2.0/1.7/1.9/2.3 2.0 2.0 2.2 1.8/1.3 1.6 1.9 
OLRa (mean) (kg COD/(m3·d)) 4.8/6.4/7.2/5.7 9.3 5.0 4.0 5.3/6.5 6.8 5.4 
Phase I
 
II
 
III
 
Mode Iw/ Iw/o IIw/ IIw/o IIIw/A IIIw/B IIIw/o 
Operating time (d) 0–56 56–57 57–77 77–78 78–111 111–153 153–154 
Bead diameter (mm) 0.8–1.2 – 0.8–1.2 – 1.5 1.5 – 
Mass of beads (g) 80 60 80 80 
Bed voidage (%) 74 100 80 100 74 74 100 
Crossflow velocity (m/s) 0.053 0.054 0.058 0.058 0.071 0.073 0.073 
Flux (mean) (l/(m2 h) 6.5/7.7/6.7/5.5 6.3 6.3 5.7 7.2/9.7 7.9 6.8 
HRTa (mean) (h) 2.0/1.7/1.9/2.3 2.0 2.0 2.2 1.8/1.3 1.6 1.9 
OLRa (mean) (kg COD/(m3·d)) 4.8/6.4/7.2/5.7 9.3 5.0 4.0 5.3/6.5 6.8 5.4 

aRelated to volume of the fluidized bed bioreactor (317 mL).

Sampling and assays

COD concentrations were determined by cuvette tests (Hach, USA). Dissolved COD (CODdiss) was analyzed after 0.45 μm filtration with a syringe filter (VWR, USA). Samples were homogenized with a dispersion tool (IKA, Germany) before determining total COD (CODtot). Before analyzing the effluent, COD 10 μL of sulfuric acid (4 mol/L) per mL sample were added. The samples were then shaken in open vials for 10 min on a platform shaker (Heidolph, Germany) to strip hydrogen sulfide and dissolved methane. The particulate COD (CODpart) results from the difference of CODtot and CODdiss. Dissolved organic carbon (DOC) and sulfate were determined by cuvette tests (Hach, USA). DOC and sulfate samples were filtrated using a 0.45 μm syringe filter (VWR, USA). No efforts were made to control biofilm formation and withdraw continuously suspended solids from the reactor. However, due to sampling between 30 and 60 mL liquid were removed from the reactor per day. This corresponded to 0.7 to 1.5% of the influent volume flow. Therefore, a distinct influence of sampling on the COD concentration within the reactor can be excluded. At the end of mode Iw/, IIw/, IIIw/A and IIIw/B, the liquid from the reactor (including accumulated settled and suspended particulate matter) was completely removed. To determine the COD of the accumulated particulate matter, the removed suspension was 0.45 μm filtered (Whatman ME 25, GE, UK) using pressure filtration equipment. Finally, the remaining filter cake was suspended in deionized water, and the COD of the suspension was determined.

The gas composition (CH4, CO2, N2, O2) was determined using a gas chromatograph equipped with a flame ionization detector (FID) and thermal conductivity detector (TCD) (Agilent Technologies, USA). For analysis, a 5 mL gas sample was taken with a syringe from the gas collection system. Determination of dissolved methane was carried out according to Düppenbecker & Cornel (2016). Samples were taken from the bioreactor top and recirculation line. Henry's law constants for methane were calculated based on data from NIST (2015). To predict the potential electrical energy of the produced methane a conversion factor of 0.33 (heat from combustion to electrical energy) and lower heating value of 800 kJ/mol were assumed.

The used membrane was examined by scanning electron microscopy (SEM) after the end of operation. Before examination, the membrane was cleaned as mentioned above and ignited at 650 °C. SEM examinations were carried out by the membrane manufacturer (atech innovations, Germany).

Prediction of pumping power required for fluidization

The pumping power required for the fluidization P (W) of glass beads was determined according to Equation (1) (Martin et al. 2011). Here, Q is the volume flow (m3/s) of the recirculation required for fluidization and Δp (Pa) the pressure loss across the fluidized bed. 
formula
1
The pressure loss of the fluidized bed can be calculated according to Equation (2) (Epstein 2003). Here, H (m) is the height of the fluidized bed and ɛ (−) the bed voidage (the ratio of the void volume to the total volume of the bed). Due to the fact that the product of bed height and particle concentration is constant for a fluidized bed (Epstein 2003), the pressure drop is determined via the initial depth of the packed bed. 
formula
2
The volume flow Q in Equation (1) can be calculated predicting the superficial liquid velocity u (m/s), which is required to achieve the desired expansion of the fluidized bed, according to Equation (3) (Epstein 2003). By assuming that the glass beads are smooth spheres the free-settling velocity u0 was calculated according to Turton & Clark (1987). Furthermore, a mean particle diameter of 1 mm (for particles with diameters between 0.8 and 1.2 mm), and a density of 1,000 kg/m3 for water (ρL) were assumed. The parameters n (−) and k (−) in Equation (3) were calculated according to Khan & Richardson (1989). 
formula
3

To estimate the pumping power required for fluidization of GAC in the bioreactor the pressure lost at onset of fluidization (at minimum fluidization velocity) was predicted. For this purpose, the fixed GAC bed height (about 0.3 m) was equated with the bed height at minimum fluidization velocity. Furthermore, a bed voidage of 0.453 at onset of fluidization (according to Wang et al. 2017) and an apparent density of 1,200 kg/m3 (according to Zamani et al. 2015) were assumed.

RESULTS AND DISCUSSION

Effect of fluidized glass beads on filtration performance

Figure 2 (top) shows TMP and flux over operating time for all phases. Firstly, glass beads with a diameter of 0.8–1.2 mm and a bed voidage of 73% were tested (mode Iw/). At the beginning, the flux was set to 6.5 L/(m2·h) without any TMP increase for approximately 21 days. From operating days 21 to 35 and a mean flux of 7.7 L/(m2·h) was adjusted and the TMP started to increase rapidly after 3 days. After cleaning (marked as #1 in Figure 3, top), the flux was set at 6.7 L/(m2·h). In contrast to the beginning of the test phase, an increase in TMP was only prevented for approximately 12 days after the first cleaning. It should be considered that OLR and CODpart concentrations varied; for instance, the CODpart concentration increased continuously during mode I up to about 1000 mg/L (Figure 2, bottom). Increasing OLR (Huang et al. 2011) and particulate matter concentrations (Jeison & van Lier 2006) can accelerate membrane fouling. This might be one reason for the accelerated TMP increase observed at the end of mode Iw/. By decreasing the flux to about 5.5 L/(m2·h), a further increase in TMP could be prevented. Without fluidized glass beads, TMP increased rapidly within one day (mode Iw/o). Increasing the bed voidage from 74 to 80% and the crossflow velocity from 0.053 to 0.058 m/s, respectively, did not show any beneficial effect (mode IIw/): the TMP started to increase after approximately 10 days of operation. At a crossflow velocity of 0.058 m/s in absence of fluidized glass beads, TMP reached within 16 h to about 30 kPa (mode IIw/o). Glass beads with a diameter of 1.5 mm were distinctly more effective as mode IIIw/A showed. At a bed voidage of 74% (crossflow velocity of 0.073 m/s), mean fluxes of 7.2 and 9.7 L/(m2·h) could be adjusted for approximately 20 and 12 days, respectively, without any distinct increase in TMP (mode IIIw/A). Therefore, glass beads of 1.5 mm diameter and a bed voidage of 74% were chosen for the final mode of about 41.5 days (IIIw/B). At a mean flux of 7.9 L/(m2·h), chemical cleaning (#7) was necessary after 17 days when TMP reached approximately 30 kPa. The first cleaning (#7) was carried out in the presence of fluidized glass beads and failed. Presumably, silica dissolved by caustic NaOCl solution fouled the membrane. In the absence of fluidized glass beads (#8), membrane cleaning was successful. During the last 21 operation days, no further cleaning was necessary. However, the TMP reached about 14 kPa at end of mode IIIw/B. In absence of fluidized glass beads at a crossflow velocity of 0.073 m/s, a rapid increase of TMP was observed (mode IIIw/o). It should be noted that chemical cleaning #9 after mode IIIw/B was not successful (see Figure 3, top). The reason for that is not clear. To verify whether the fouling is removable by the applied chemical cleaning procedure, after mode IIIw/o and cleaning #10, a clean water filtration test was carried out. As the lower TMP after cleaning #10 shows, the fouling could be distinctly reduced. Hence, the fouling seemed to be removable by the applied chemical cleaning.
Figure 2

TMP and flux over operating time (top). OLR, CODpart and CODdiss (at membrane inlet) and permeate COD over operating time (bottom). Flux and TMP after mode IIIw/o were measured during clean water (CW) filtration without fluidized bed to verify success of cleaning #10.

Figure 2

TMP and flux over operating time (top). OLR, CODpart and CODdiss (at membrane inlet) and permeate COD over operating time (bottom). Flux and TMP after mode IIIw/o were measured during clean water (CW) filtration without fluidized bed to verify success of cleaning #10.

Figure 3

SEM images of breaking edge (left) and membrane surface (right). The bright spots are the residual active layer (150,000 Da). The support membrane (0.1 μm) is damaged as well and the support layer appears (coarse structure with mean pore size of 3.9 μm).

Figure 3

SEM images of breaking edge (left) and membrane surface (right). The bright spots are the residual active layer (150,000 Da). The support membrane (0.1 μm) is damaged as well and the support layer appears (coarse structure with mean pore size of 3.9 μm).

The results show that by using glass beads as turbulence promoters, crossflow velocities are reduced distinctly to 0.053–0.073 m/s, compared to reported crossflow velocities of 2–5 m/s for conventional AMBRs with external crossflow membrane module (Liao et al. 2006). Several studies operated at lower crossflow velocities treating municipal wastewater. For instance, An et al. (2009) operated an AMBR at crossflow velocities between 0.22 m/s and 1.45 m/s at a flux of 10.5 L/(m2·h). However, at crossflow velocities of 0.22 m/s, TMP rapidly reached 50 kPa within 20 h. More recently, Seib et al. (2016) achieved runtimes (TMP ≤ 50 kPa) of about 50 h, at crossflow velocities between 0.018 and 0.024 m/s and fluxes between 5.9 and 7.4 L/(m2·h). By adding fluidized GAC (0.6–1.7 mm), run times were extended to 73–84 h (Seib et al. 2016). In the present study, the achieved run times were >300 h. Hence, fluidized glass beads seem to be more effective than fluidized GAC. This might be attributed to the higher density of the glass beads (2,500 kg/m3) in comparison to the density of GAC (about 1,480 kg/m3 according to Wang et al. (2017)). Due to the higher density, a higher superficial velocity is required to fluidize the glass beads. Since momentum transfer and shear rate increase with increasing particle density and superficial velocity (Charfi et al. 2017) glass beads might be more effective than GAC.

Membrane damage and rejection

Figure 3 shows SEM images of the breaking edge (left) and surface (right) of the used membrane after end of operation. It can be seen that the active layer (150,000 Da) was almost completely abraded and the support membrane (0.1 μm) was damaged seriously, as well. A result of the severe membrane damage is presumably the observed decrease in DOC rejection from 62% and 57% (mode Iw/ and IIw/, respectively) to 46% (mode IIIw/A and IIIw/B). Due to the fact that the DOC rejection decreased distinctly in phase III, the use of the 1.5 mm glass beads might be the reason for severe membrane damage. During phases I and II using 0.8–1.2 mm glass beads, there was no distinct decrease in DOC rejection. The lowest DOC rejection was observed during mode IIIw/B after cleaning #8. Interestingly, as Figure 4 shows, increased the DOC rejection with increasing TMP after cleaning #8. Smith et al. (2015) observed a similar phenomenon and attributed it to the biogical activity of a biofilm growing onto the membrane surface. However, in the present case, the increase in DOC is unlikely to result from biological activity of a biofilm: after cleaning #8, the increase in TMP occurred without any increase in DOC rejection. It is assumed that before cleaning #8 the destruction of the support membrane (pore size 0.1 μm) began and the pores of the support layer (mean pore size of 3.9 μm) were locally exposed. Intermediately, deposits on the surface of the support layer pores' walls started to diminish the pore sizes and maybe blocked them over time. As a result, the TMP increased continuously. Assuming that the diminished or blocked pores still rejected organic matter <0.45 μm, DOC rejection did not vary significantly. During chemical cleaning #8, all deposits were removed at once and consequently hydraulic resistance and DOC rejection declined abruptly. Thereafter, the support layer pores were continuously diminished or blocked and DOC rejection and TMP increased simultaneously.
Figure 4

DOC rejection as a function of TMP during mode IIIw/B before and after cleaning #8.

Figure 4

DOC rejection as a function of TMP during mode IIIw/B before and after cleaning #8.

Organic removal and COD balance

Overall CODtot removal of the AMBR was between 77 and 83% resulting in mean COD permeate concentration of 69–83 mg/L (Table 2). Calculating COD removal of the fluidized bed bioreactor alone – based on the COD influent concentration – is not meaningful, since the membrane outlet is recirculated to the bioreactor inlet. Note, that the present system is not a staged system. The COD removal of the bioreactor alone was evaluated in a previous study (Düppenbecker & Cornel 2016) and is compared with the results of the present study in Table 3. The observed COD removal agrees well with previous studies of AMBRs treating municipal wastewater. Yoo et al. (2012) reported a CODtot removal of about 84% and mean COD permeate concentrations of about 25 mg/L at slightly higher HRT (2.3–3.4 h) and 25 °C. Smith et al. (2013) observed a CODtot removal of 69% and permeate COD of 76 mg/L, respectively, at lower temperature (15 °C) and higher HRT between 16 and 24 h. A CODtot removal of 87% (permeate COD about 100–120 mg/L) at 18 °C (HRT = 7 h) was reported by Gouveia et al. (2015).

Table 2

Summary of COD concentrations of influent, membrane inlet and permeate. Overall removal is calculated based on influent and permeate concentrations. Membrane rejection is calculated based on membrane inlet and permeate concentration

Mode Influent
 
Membrane inlet
 
Permeate Overall removal Membrane rejection
 
CODtot CODdiss CODtot CODdiss COD CODtot CODtot CODdiss 
( − ) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (%) (%) (%) 
Iw/ 470 146 918 206 75 83 91 62 
IIw/ 422 131 748 164 69 80 91 58 
IIIw/A 376 134 629 156 82 77 86 47 
IIIw/B 461 174 696 150 83 81 87 42 
Mode Influent
 
Membrane inlet
 
Permeate Overall removal Membrane rejection
 
CODtot CODdiss CODtot CODdiss COD CODtot CODtot CODdiss 
( − ) (mg/L) (mg/L) (mg/L) (mg/L) (mg/L) (%) (%) (%) 
Iw/ 470 146 918 206 75 83 91 62 
IIw/ 422 131 748 164 69 80 91 58 
IIIw/A 376 134 629 156 82 77 86 47 
IIIw/B 461 174 696 150 83 81 87 42 
Table 3

CODtot balance of AMBR (mode IIIw/B of present study) and anaerobic fluidized bed reactor without membrane. OLR (kg CODtot/(m3·d))/HRT(h) were 6.8/1.6 (AMBR) and 6.1/1.9 (fluidized bed), respectively

  AMBR (present study)
 
Fluidized beda (without membrane)
 
(mg/l) (%) (mg/l) (%) 
Influent CODtotal 460 100 480 100 
Effluent CODtotal 82 18 193 40 
Difference 379 82 286 60 
Methane total 134 29 104 22 
Methane gaseous 77 17 53 11 
Methane dissolvedb 57 12 51 11 
Accumulated CODpartc 60 13 92 19 
Sulfate reductiond 70 15 66 14 
Unknown 114 25 25 
  AMBR (present study)
 
Fluidized beda (without membrane)
 
(mg/l) (%) (mg/l) (%) 
Influent CODtotal 460 100 480 100 
Effluent CODtotal 82 18 193 40 
Difference 379 82 286 60 
Methane total 134 29 104 22 
Methane gaseous 77 17 53 11 
Methane dissolvedb 57 12 51 11 
Accumulated CODpartc 60 13 92 19 
Sulfate reductiond 70 15 66 14 
Unknown 114 25 25 

bCorresponding to the solubility of methane in pure water at 20 °C calculated by Henry's law.

cCorresponding to suspended and settled CODpart (>0.45 μm) inside the reactor; determined at end of mode.

dAssuming total reduction of influent sulfate.

As Table 2 shows, membrane total COD rejection varied between 87 and 91%. In contrast, dissolved COD rejection was subject to large fluctuation and decreased over operating time from 62% (mode Iw/) to 42% (mode IIIw/B). By comparing influent and membrane inlet CODtot concentrations (Table 2), it is conspicuous that the low-strength wastewater is concentrated by membrane rejection. CODtot concentrations in the reactor are higher by 44 to 95% than in the influent. Since the CODdiss remained on a constant level, a concentration of the CODpart took place. As Figure 2 shows, during mode Iw/ there was a continuous increase in CODpart within the reactor. The CODpart seemed not to have been degraded at the same rate as it was accumulated due to the rejection by the membrane. An accumulation of CODpart was observed during the other modes, as well. However, in modes IIw/ and IIIw/B, it was observed as well that at low OLRs the concentration of the CODpart decreased. In contrast, the CODdiss increased only at the beginning of each mode and remained on a constant level, presumably due to the fact that the rejected dissolved COD (>150,000 Da) is biologically degraded. During modes Iw/, IIw/ and IIIw/A, CODdiss concentrations were higher by 16 to 41% than the influent concentrations. During mode IIIw/B, the CODdiss concentration at the membrane inlet is lower than in the influent. This might be attributed to the decrease of CODdiss rejection over operating time, presumably due to membrane damage. As a result, in mode IIIw/B, an accumulation of the CODdiss within the reactor did not occur. Frequent sampling did not influence COD concentration within the reactor: for instance, only 2.4 and 1.5% of the rejected CODdiss and CODpart mass flow were removed by sampling during mode IIIw/B. The results indicate that, due to membrane rejection, the effluent quality can be improved distinctly. Moreover, CODdiss rejection by the membrane might be a key to increase biological conversion at low HRTs.

This fact is confirmed by comparing the COD balance of the AMBR (mode IIIw/B) and the fluidized bed reactor without membrane (Table 3). Both reactors were operated under similar conditions and fed with the same wastewater. As Table 2 shows, effluent concentrations were reduced from 40 to 18% related to the influent concentrations. Furthermore, the production of total methane was increased by nearly 30%. It is assumed that the increased methane production is attributed to biological conversion of rejected CODpart and CODdiss. However, about 43% of the produced methane remained dissolved in the effluent. Methane supersaturation within the reactor was not observed during mode IIIw/B: saturation values of 91, 100, and 108% were measured. Therefore, the dissolved methane concentration for the COD balance was predicted by Henry's law. The noticeable gap of 25% in the case of the AMBR might be mainly attributed to inaccurate sludge sampling: COD of accumulated particulate COD is distinctly lower in comparison with fluidized bed reactor without membrane. The fraction of total influent COD removed by sampling is less than 2% and accumulation of dissolved COD can be neglected (less than 0.5% of total influent COD).

Energy considerations

Table 4 shows the required energy for fouling control, permeate drawing and bioreactor operation. A total required pumping energy of 0.190 kWh/m3 and electrical energy of 0.318 kWh/m3, respectively, was predicted. By using the produced gaseous methane, 0.09 kWh/m3 electrical energy could be generated; sufficient to cover about 30% of the total electrical energy demand. In absence of sulfate and dissolved methane recovery, about 0.15 kWh/m3 electrical energy could be generated additionally.Seib et al. (2016), using fluidized GAC, achieved values of 0.05–0.13 kWh/m3 in laboratory scale. The distinctly lower required electrical energy can be attributed to the lower density of GAC which reduce the required volume flow for recirculation and pressure loss across the fluidized bed. However, the crossflow velocity of 0.073 m/s and the energy demand of about 0.3 kWh/m3 are distinctly lower than reported values of 2–5 m/s and 3–7.3 kWh/m3 for conventional AMBRs with external crossflow (Liao et al. 2006). Note that predicted energy demand of the present study based on laboratory-scale experiments. Therefore, comparison of these data with data from pilot- and large-scale plants is only possible to a limited extent.

Table 4

Required energy for the AMBR during mode IIIw/B

Fluidization GAC (bioreactor) 
 Volume flow recirculation (l/h) 14.4 
 Pressure loss (kPa)a 0.3 
 Required pumping power (W) 0.001 
 Required pumping energy (kWh/m30.007 
Fluidization glass beads 
 Volume flow recirculationb (l/h) 52.6 
 Pressure lossa (kPa) 2.3 
 Required pumping power (W) 0.034 
 Required pumping energyc (kWh/m30.180 
Permeate drawing 
 Permeate volume flow (l/h) 0.190 
 Pressure loss (kPa) 10.2 
 Required pumping power (W) 0.0005 
 Required pumping energyc (kWh/m30.0028 
Total 
 Required pumping energyc (kWh/m30.190 
 Required electrical energyc,d (kWh/m30.318 
Fluidization GAC (bioreactor) 
 Volume flow recirculation (l/h) 14.4 
 Pressure loss (kPa)a 0.3 
 Required pumping power (W) 0.001 
 Required pumping energy (kWh/m30.007 
Fluidization glass beads 
 Volume flow recirculationb (l/h) 52.6 
 Pressure lossa (kPa) 2.3 
 Required pumping power (W) 0.034 
 Required pumping energyc (kWh/m30.180 
Permeate drawing 
 Permeate volume flow (l/h) 0.190 
 Pressure loss (kPa) 10.2 
 Required pumping power (W) 0.0005 
 Required pumping energyc (kWh/m30.0028 
Total 
 Required pumping energyc (kWh/m30.190 
 Required electrical energyc,d (kWh/m30.318 

aPredicted according to Equation (2).

bBased on measured data of mode IIIw/B.

cRelated to permeate volume flow at flux of 7.8 l/(m2·h).

dAssuming pump efficiency of 60% (Martin et al. 2011).

Predicted (theoretical) volume flow, pressure loss (of the fluidized bed), and required pumping power for the membrane module are shown in Figure 5. The pressure loss across the wire cloth and the support layer, respectively, and wall friction were neglected. The depth of the fluidized bed was kept constant at 0.61 m to cover the membrane with the fluidized bed. Consequently, the initial packed bed depth and the pressure drop decreased with increasing bed voidage. In contrast, the volume flow required for fluidization increased with increasing bed voidage. The required pumping power (the product of both) increased initially with increasing bed voidage and reached a maximum at about 72.5%. A further increase in bed voidage results in a decreasing pumping power demand. With increasing bead diameter, the volume flow required for fluidization increased. In contrast, the pressure loss does not depend on the particle diameter. At a bed voidage of 74%, the required pumping power for fluidization of 1 mm beads, (phase I) is about 25% lower as for 1.5 mm beads (phase III). Furthermore, for 1 mm beads, an increase of the bed voidage from 74 (phase I) to 80% (phase II) decreased the required pumping power by 6%. It should be considered that in the present study lower fouling rates were observed at a bed voidage of 74% (for 1 mm beads) and using larger beads (at a bed voidage of 74%), respectively. Hence, the fouling rate seems to decrease with increasing power input. In terms of saving energy, the operation at a bed voidage lower than 72.5% might be more beneficial. Due to the lower crossflow velocity, the pressure loss across the support layer and recirculation line could be reduced as well. However, previous studies showed an optimum bed voidage (highest flux at constant pressure filtration) between 65 and 80% during filtration of cheese whey (de Boer et al. 1980), macromolecular solutions (Rios et al. 1987) and alumina suspensions (Mikulášek & Filandrová 1995). Consequently, at a bed voidage distinctly below 72.5% an increase of the fouling rate has to be expected. However, the impact of bed voidage on fouling mitigation during filtration of complex media is not clear and should be investigated in future. In particular, the impact of fluidized particles on the deposition of particulate matter is not yet well understood.
Figure 5

Predicted (theoretical) volume flow, pressure loss, and required pumping power. A mean particle diameter of 1.0 and 1.5 mm and a constant fluidized bed height of 0.61 mm was assumed.

Figure 5

Predicted (theoretical) volume flow, pressure loss, and required pumping power. A mean particle diameter of 1.0 and 1.5 mm and a constant fluidized bed height of 0.61 mm was assumed.

CONCLUSIONS

This study shows that fluidized glass beads can distinctly reduce fouling at low crossflow velocities of 0.053–0.073 m/s in an AMBR with an external tubular module, treating municipal wastewater. The crossflow velocities, applied in the present study, are distinctly lower in comparison to data for conventional AMBR with external crossflow membrane modules. A glass bead diameter of 1.5 mm and a bed voidage of about 74% seemed to be appropriate. However, further research is necessary to understand the impact of bed voidage and particle diameter on the fouling rate in more detail. As SEM examinations showed, the fluidized glass beads damaged the used ceramic membrane distinctly by abrasion. Obviously, the used membrane is not suitable for the intended purpose; ceramic membranes with other properties will be tested in future. The overall CODtot removal of the AMBR was between 77 and 83%, although HRTs were only between 1.3 and 2.3 h. The resulting COD permeate concentrations were between 69 and 83 mg/L. Furthermore, the production of total methane was increased by about 30% in comparison to the bioreactor without membrane. It is assumed that the increased methane production is attributed to the biological conversion of rejected dissolved and particular COD. The total required electrical energy was predicted to be about 0.3 kWh/m3. By using the produced gaseous methane 0.09 kWh/m3 electrical energy could be generated, sufficient to cover about 30% of the total electrical energy demand.

ACKNOWLEDGEMENTS

This research project is funded by the Willy Hager Foundation, Stuttgart, Germany.

REFERENCES

REFERENCES
An
Y. Y.
Yang
F. L.
Bucciali
B.
Wong
F. S.
2009
Municipal wastewater treatment using a UASB coupled with cross-flow membrane filtration
.
Journal of Environmental Engineering
135
,
86
91
.
Bixler
H. J.
Rappe
G. C.
1970
Ultrafiltration Process
.
United States Patent 3,541,006
,
Amicon Corporation
,
Lexington, MA
,
USA
.
Charfi
A.
Aslam
M.
Lesage
G.
Heran
M.
Kim
J.
2017
Macroscopic approach to develop fouling model under GAC fluidization in anaerobic fluidized bed membrane bioreactor
.
Journal of Industrial and Engineering Chemistry
49
,
219
229
.
De Boer
R.
Zomerman
J. J.
Hiddink
J.
Aufderheyde
J.
van Swaay
W. P. M.
Smolders
C. A.
1980
Fluidized beds as turbulence promoters in the concentration of food liquids by reverse osmosis
.
Journal of Food Science
45
,
1522
1528
.
Düppenbecker
B.
Behnisch
J.
Cornel
P.
2016
Fluidized resin beads as turbulence promoters in anaerobic membrane bioreactors
. In:
IWA Specialist Conference on Advances in Particle Science and Separation: Meeting Tomorrow's Challenges
,
Oslo, Norway
,
22–24 June 2016
.
Epstein
N.
2003
Liquid-solids fluidization
. In:
Handbook of Fluidization and Fluid-Particle Systems
(
Yang
W.-C.
, ed.).
Marcel Dekker
,
New York, NY
,
USA
.
Imasaka
T.
Kanekuni
N.
So
H.
Yoshino
S.
1989
Cross-flow filtration of methane fermentation broth by ceramic membranes
.
Journal of Fermentation and Bioengineering
68
(
3
),
200
206
.
Kim
J.
Kim
K.
Ye
H.
Lee
E.
Shin
C.
McCarty
P. L.
Bae
J.
2011
Anaerobic fluidized bed membrane bioreactor for wastewater treatment
.
Environmental Science and Technology
45
,
576
581
.
Liao
B.-Q.
Kraemer
J. T.
Bagley
D. M.
2006
Anaerobic membrane bioreactors: applications and research directions
.
Critical Reviews in Environmental Science and Technology
36
(
6
),
489
530
.
Martin
I.
Pidou
M.
Soares
A.
Judd
S.
Jefferson
B.
2011
Modelling the energy demands of aerobic and anaerobic membrane bioreactors for wastewater treatment
.
Environmental Technology
32
(
9–10
),
921
932
.
Mikulášek
P.
Filandrová
L.
1995
Use of fluidized bed for intensification of microfiltration on ceramic membranes
.
Developments in Chemical Engineering and Mineral Processing
3
(
3–4
),
139
149
.
NIST
2015
National Institute of Standards and Technology Chemistry WebBook
.
http://webbook.nist.gov/chemistry/ (accessed 16 December 2016)
.
Noordman
T. R.
de Jonge
A.
Wesselingh
J. A.
Bel
W.
Dekker
M.
Ter Voorde
E.
Grijpma
S. D.
2002
Application of fluidised particles as turbulence promoters in ultrafiltration: improvement of flux and rejection
.
Journal of Membrane Science
208
(
1–2
),
157
169
.
Rios
G. M.
Rakotoarisoa
H.
Tarado de la Fuente
B.
1987
Basic transport mechanisms of ultrafiltration in the presence of fluidized particles
.
Journal of Membrane Science
34
(
2
),
331
343
.
Schmidt
S.
Büchs
J.
Born
C.
Biselli
M.
1999
A new correlation for the wall-to-fluid mass transfer in liquid-solid fluidized beds
.
Chemical Engineering Science
54
,
829
839
.
Seib
M. D.
Berg
K. J.
Zitomer
D. H.
2016
Low energy anaerobic membrane bioreactor for municipal wastewater treatment
.
Journal of Membrane Science
514
,
450
457
.
Smith
A. L.
Stadler
L. B.
Love
N. G.
Skerlos
S. J.
Raskin
L.
2012
Perspective on anaerobic membrane bioreactor treatment of domestic wastewater: a critical review
.
Bioresource Technology
122
,
149
159
.
Smith
A. L.
Skerlos
S. J.
Raskin
L.
2013
Psychrophilic anaerobic membrane bioreactor treatment of domestic wastewater
.
Water Research
47
,
1655
1665
.
Stuckey
D. C.
2012
Recent developments in anaerobic membrane reactors
.
Bioresource Technology
122
,
137
148
.