Process Biochemistry 38 (2002) 279
/285

www.elsevier.com/locate/procbio

Design and operation considerations for wastewater treatment using

a flat submerged membrane bioreactor
Jin Kie Shim a, Ik-Keun Yoo b,*, Young Moo Lee a
a
School of Chemical Engineering, College of Engineering, Hanyang University, Seoul 133-791, South Korea
b
School of Chemical Engineering and Bioengineering, University of Ulsan, Ulsan 680-749, South Korea

Received 13 November 2001; received in revised form 14 March 2002; accepted 21 March 2002

Abstract

Combining a flat submerged membrane with a biological reactor was investigated for synthetic wastewater treatment of high
concentration. The increase of suction pressure as a result of fouling of the submerged membrane, nitrification and chemical oxygen
demand (COD) removal efficiencies were monitored during 4
/9 months operation. In order to find a proper geometry of the reactor
with respect to the long-term stability of membrane filtration, three types of membrane bioreactor with a different ratio of cross-
sectional area of the downcomer to the riser were operated and the time profiles of suction pressure increase were compared. While
maintaining a filtration flux at 12
/16 l m 2 h1, a higher ratio of downcomer to riser area gave a better filtration stability of
submerged membrane over several months of operation. A sharp increase of suction pressure was also observed when sludge bulking
occurred during normal operation. Recovery of initial pure water flux for the membrane, which has been used over 6 months, was
possible after 96 h of chemical cleaning. Under operating conditions with 12
/16 h hydraulic retention time and 8000
/16 000 mg l 1
sludge concentration in the bioreactor, COD removal efficiency was above 98% for 900
/1600 mg l 1 influent COD and nitrification
above 95% with 50
/600 mg l 1 total nitrogen fed to the membrane bioreactor. # 2002 Elsevier Science Ltd. All rights reserved.

Keywords: Membrane bioreactor; Sludge bulking; Membrane fouling; Filtration stability; Flat membrane

1. Introduction istics and other design parameters. This higher sludge

biomass enables stable wastewater treatment regardless
The activated sludge process, which combines a of organic loading fluctuation and a reduction of plant
biological reactor with a secondary clarifier, is the size is possible. The advantage of the membrane
most common method for wastewater treatment. bioreactor (MBR) also includes the development of
Although the successful operation of the conventional slow growth rate bacteria such as nitrogen oxidizing
activated sludge process mainly depends on the perfor- bacteria (nitrifiers).
mance of gravitational secondary clarifiers, the sludge Among various types of membranes, the first genera-
settleability cannot be easily controlled in real waste- tion of MBR operated with tubular membranes placed
water treatment plants. The membrane separation in external recirculation loops. The use of recirculation
technique in the field of wastewater treatment, which loops leads to increased energy costs, 2
/10 kW h m 3
was first reported with ultrafiltration membranes [1], of water produced, depending on the internal diameter
avoids these problems of non-settling sludge and of the tubes used. In addition, the high shear stresses in
replaces the secondary clarifier. Due to the complete the tubes and recirculation pumps can contribute to the
separation of solid
/liquid, the mixed liquor suspended destruction of bioflocs and this has been linked to a loss
solids (MLSS) can be increased up to whatever con- of biological activity [2]. The submerged MBR, which
centration is appropriate for the wastewater character- was first developed and popularly installed in Japan, is
an alternative system of overcoming these limits [3]. In
* Corresponding author. Tel.: /82-52-259-2251; fax: /82-52-259-
this configuration, the membranes are placed directly
1689. into a bioreactor and the treated permeate is extracted.
E-mail address: ikyoo@mail.ulsan.ac.kr (I.-K. Yoo). This operating mode limits the energy consumption
0032-9592/02/$ - see front matter # 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 3 2 - 9 5 9 2 ( 0 2 ) 0 0 0 7 7 - 8
280 J.K. Shim et al. / Process Biochemistry 38 (2002) 279
/285

associated with the filtration to 0.2
/0.4 kW h m 3

treated.
The widely used membranes to date for a submerged
type are either hollow fibres or flat membranes [4
/6].
The key to the success of internal membrane filtration is
the direct application of air for membrane cleaning at
the base of the membrane modules. Although many
applications using submerged type membranes have
been reported [3
/9], the considerations in terms of
bioreactor design were not enough for MBR. For flat
membrane, the air bubbles and air-lifted liquid move
upward along the membrane surface. Because the
membrane elements are very close together in the
membrane module, this uprising motion causes a wiping
or rubbing action between the flat membranes that Fig. 1. Schematic diagram of the flat submerged membrane bioreac-
provides surface cleaning. In addition to the riser section tor.
sparged with air, this kind of airlift reactor also has a
downcomer section with liquid moving downward with each vertical plate. The pressure gauge was
between the membrane module and the wall of the installed in order to monitor the variation of the
bioreactor. For the successful control of the membrane transmembrane pressure between the membrane and a
fouling problem, this circulation velocity of air-induced suction pump. To maintain a constant level in the
two-phase flow of bubbling air and sludge should reactor, a peristaltic pump providing influent waste-
generate an appropriate turbulence for the cleaning of water and a level sensor were used. Air was supplied
the membrane surface. Various design variables on through the diffuser below the membrane module. The
airlift-type reactor were studied in terms of liquid coarse bubble diffuser was a 1.5-cm inner diameter tube
circulation velocity, which is determined by an energy with 1 mm openings so that coarse bubbles would be
balance [10]. produced. The riser section is a region of upflow air
The objective of this study was to investigate the effect bubbles from the diffuser and air-induced flow of mixed
of some design and operational variables on the proper liquor, while the downcomer section is a downflow
control of membrane clogging in MBR wastewater region of mixed liquor between the reactor wall and
treatment process. The time profiles of suction pressure both vertical plates.
increase, which indicates the progress of membrane As shown in Table 1, three rectangular types of
clogging, were compared between three types of bior- bioreactor with different cross-sectional area ratio of
eactor with different riser-to-downcomer cross-sectional downcomer (Ad) to riser (Ar): reactor 1 (Ad/Ar ratio of
area ratio and size of bubbling air supplied to the 4.5 and fine bubble diffuser); reactor 2 (Ad/Ar ratio of
reactor. During operation, the sludge settlability repre- 1.6 and fine bubble diffuser); and reactor 3 (Ad/Ar ratio
sented by sludge volume index (SVI) value was mon- of 3.6 and coarse bubble diffuser) were constructed and
itored to determine whether sludge characteristics might operated for several months. One element of flat
affect membrane clogging. The organic and nitrogen microfiltration membrane (chlorinated polyethylene,
loads were varied during several months of operation to pore size of 0.4 mm, Yuasa Co., Japan) in each reactor
verify the high volumetric efficiencies of MBR for COD was used in this study. The filtration area of each side of
removal and nitrification. flat membrane was 0.05 m2, creating a total filtration
area of 0.1 m2.

2. Materials and methods 2.2. Operating conditions and analysis

2.1. Construction of the membrane bioreactor system The operational periods of each MBR were about
250, 40, 70 days for reactor 1, 2 and 3, respectively.
Fig. 1 shows the experimental setup for a flat-type Since intermittent suction gave a stable filtration
submerged membrane bioreactor system. For effective performance for a long period of time [3], the permeate
circulation of air-induced liquid in the bioreactor, two was extracted by a suction pump under intermittent
vertical flat plates separating the cross-section of the operation in a 10-min cycle; 8-min on and 2-min off.
bioreactor into the riser section and the downcomer MLSS was measured in daily samples and excess sludge
section were installed between each side of the flat was removed once a week from the reactor to maintain
membrane and the reactor wall. The membrane was the MLSS concentration at around 10 000
/13 000 mg
placed in the centre of the bioreactor at a gap of 7 mm l 1. The sludge used in this study was collected from a
 J.K. Shim et al. / Process Biochemistry 38 (2002) 279
/285 281

Table 1
Specification of bioreactors with different geometry and air diffuser

Reactor Length (cm) Width (cm) Heighta (cm) Working volume (l) Ad/Arb Air bubble

R1 40 13.5 39 21 4.5 Fine bubble

R2 22 6 160 21 1.6 Fine bubble
R3 22 10 115 26 3.6 Coarse bubble
a
The height means the working level of sludge mixture in the bioreactor.
b
Ad: cross-sectional area of the downcomer (downflow section of degassed sludge between the guard plate and the reactor wall); Ar: cross-
sectional area of the riser (upflow section of bubbling air and sludge between the membrane and both guard plates).

municipal wastewater treatment plant and acclimated capacity per membrane area, and so on. Among these,
with synthetic wastewater for about 1 month. Synthetic this study mainly focused on bioreactor geometry that
wastewater was made with tap water and the compo- may affect on the circulation velocity of air-induced
nents as follows (mg l1): glucose 670
/1080, glutamic liquid and also on the change of sludge characteristics.
acid 285
/460, CH3COONH4 220
/350, NaHCO3 750
/ In submerged MBRs, the turbulence induced by air and
3000, NH4Cl 33
/53, KH2PO4 60, K2HPO4 80, MgSO4 / liquid flow has been known to be the factor preventing
7H2O 33, MnSO4 /H2O 10, FeCl3 /6H2O 2, CaCl2 /2H2O membrane clogging during wastewater treatment.
20, NaCl 25. The COD and nitrogen concentrations of Therefore, it is important to determine a proper design
influent wastewater were adjusted by changing the parameter for long-term operation of a MBR as well as
concentrations of glucose, glutamic acid, CH3COONH4, the development of the membrane itself with good anti-
and NH4Cl. The flow rate of air supplied to the reactor fouling properties for wastewater treatment.
was maintained in the range of 12
/15 l min 1, so the Three different types of rectangular bioreactors with
dissolved oxygen (DO) was not exactly controlled at a one flat membrane element in each reactor have been
fixed value. The MLSS concentration and sludge tested for several months of operation as shown in Table
volume index were determined by Standard Methods 1. The bioreactor was classified as R1, R2 and R3
[11]. Ammonium was measured by a Nesslerization depending on the riser-to-downcomer cross-sectional
method by reading absorbance at 425 nm, and nitrate area ratio and the type of diffuser for air supply. The
was measured by an ultraviolet spectrophotometric effect of air flow rate, which was maintained at 12
/15 l
method at 220 nm as described in Standard Methods min 1 in all three reactors, was not quantitatively
[11]. Nitrate concentration was measured by ion chro- investigated although the air flow rate is one of the
matography (DX500, Dionex). COD was measured in most important factors determining the circulation
accordance with a US EPA approved method utilizing velocity in air lift reactor. The slow increase of suction
Hach Laboratory Method 8000 (DR/2010, Hach). The pressure applied to the submerged membrane may have
temperature, pH, permeate flow rate from the mem- resulted from an irreversible clogging of the membrane,
brane, and MLSS concentrations were monitored in which requires chemical cleaning for the recovery of
daily samples. initial flux. In this study, an abrupt change of suction
pressure rather than a slow increase of suction pressure
was considered as a failure of membrane cleaning by air-
3. Results and discussion induced flow. The circulation velocity of two-phase flow
of bubbling air and sludge was not measured directly in
3.1. Suction pressure and filtration stability of membrane this study. Fig. 2 shows the filtration stability of R1 in
terms of suction pressure for about 8 months of
Since air was fed to the reactor only in the riser operation. The flux was maintained at 12
/16 l m 2
section as shown in Fig. 1, the sludge mixture was h 1. No increase of suction pressure during an initial 5
circulated between the air-sparged riser and the un- months clearly showed the effective cleaning of mem-
sparged downcomer section in the reactor, so this MBR brane surface by an uplifting two-phase flow of air and
can be regarded as an airlift-type reactor. In general, the sludge. Near day 150, the increase in suction pressure
magnitude of liquid circulation is one of the most became serious and operation was temporarily inter-
important design parameter for airlift reactor because rupted for membrane washing.
it controls various conditions such as the extent of gas
/ During operation, the SVI was monitored to check
liquid mass transfer, heat transfer, mixing and turbu- the sludge characteristics. SVI was maintained around
lence [10]. Depending on the characteristics of the 80
/150 for an initial 140 days, but a sharp increase of
membranes used, the submerged MBRs have different SVI up to 500 was observed between day 140 and day
design and operating parameters such as the flow rate of 155 of operation. At that time, the bulking sludge state
supplied air, the bioreactor geometry, the treatment caused by filamentous organisms could be confirmed by
282 J.K. Shim et al. / Process Biochemistry 38 (2002) 279
/285

for proper cleaning of the membrane surface compared

with the ratio of 4.5 in R1. After operating for 39 days,
the sludge in R2 was transferred to R3 and the operation
of R3 was started with no delay. Compared with the R1
and R2, respectively, the coarse bubble diffuser was
applied in R3 and the Ad/Ar ratio was two times larger
than that of R2. During 80 days of operation with R3,
the suction pressure did not increase from its initial zero
value and the Ad/Ar ratio of 3.6 in R3 was sufficient for
the prevention of membrane clogging.

3.2. Chemical cleaning and flux recovery

As shown in Fig. 2, the upward tendency of suction

pressure became serious over a short period near day
Fig. 2. Variation of suction pressure and flux in reactor 1 (R1).
150, which implied rapid acceleration of membrane
fouling. Since an increase in air flow rate up to 30 l
microscopic observation. If this bulking did not occur,
min 1 for some hours was not effective for lowering the
the suction pressure might be maintained at almost zero
suction pressure, the operation was stopped on day 155
value for a longer operation time. In fact, the sludge
and the membrane element was taken from the reactor.
characteristics in the MBR can be in many physiological
The formation of a thick gel layer was observed on the
states during normal operation of the MBR process
membrane surface. At first, sponge washing with tap
because the sludge is a living microorganism. Although
water was carried out and then the membrane surface
MBRs have been said to be a stable system regardless of
became visually clean. After the washed membrane was
sludge settlability, this study suggests that the abrupt
re-installed in the reactor, the operation re-started. The
change of sludge characteristics such as a bulking state suction pressure was restored almost to its initial zero
should be prevented in MBRs for long-term stability of value and maintained for about 3 days, but then started
membrane filtration. to increase again. Until then, the SVI was maintained
Fig. 3 shows the operation result for R2 and R3. The around 400
/500, which indicates the bulking state of
operation of R2 was started at the same time as R1. sludge. It is not clear if this second increase of suction
While R1 was stably operated for an initial 5 months, pressure was due to the hard deposition and internal
the filtration stability of R2 was maintained for less than pore plugging by particles that were not easily cleaned
1 month. As shown in Table 1, the difference of design by physical washing.
and operation variables between R1 and R2 was only in When the suction pressure reached about 190 mmHg
the ratio of cross-sectional area of the downcomer to the on day 166, the operation was stopped for the second
riser. The SVI value was also monitored during the cleaning of membrane. Chemical cleaning was done
operation of R2 and it was stably maintained at around after washing as before. The membrane was immersed in
80
/120. From this result, it can be assumed that the Ad/ a 0.5 wt.% sodium hypochlorite solution for 3 h
Ar ratio of 1.6 in R2 did not generate a sufficient flow followed by 0.5 wt.% oxalic acid solution for another
3 h. Before the start of operation with this chemically
washed membrane, the whole sludge in the reactor was
removed and the reactor was filled with new sludge of
low SVI. Although the restored zero value of suction
pressure was maintained for a longer time compared
with the physical sponge cleaning, the increase of
suction pressure was also observed from day 190. The
operation was stopped on day 200. At the third cleaning
step, the concentration of cleaning chemicals was
increased to 1.0 wt.% and the immersion time was
varied while the immersion time was fixed at overall 6
hours in the previous second cleaning. The membrane
was immersed in a 1.0 wt.% sodium hypochlorite
solution followed by a 1.0 wt.% oxalic acid solution.
As shown in Fig. 4, the recovery of the original pure
Fig. 3. Variation of suction pressure and flux in reactors 2 and 3 (R2, water flux was measured with the variation of immer-
R3). sion time, which means overall immersion time into a
 J.K. Shim et al. / Process Biochemistry 38 (2002) 279
/285 283

Fig. 4. Effect of chemical cleaning time on the recovery of the initial

pure water flux. Chemical cleaning was performed with 1.0 wt.%
NaOCl solution followed by 1.0 wt.% oxalic acid solution.

sodium hypochlorite solution followed by an oxalic acid Fig. 5. Variation of MLSS and F/M ratio in reactor 1 (R1).
solution. The recovery of the pure water flux is defined
as, was the domestic wastewater of low concentration [5
/7].
J The above F/M ratio in this study corresponds to a
Recovery of pure water flux (%)  t 100 similar operating condition to the conventional acti-
Ji vated sludge process. Consequently, this study clearly
where Ji and Jt are the pure water flux of before used demonstrated that the wastewater treatment capacity of
and after cleaned. the MBR process could be increased as much as the
The recovery rate of pure water flux sharply increased increase of sludge concentration if the MBR was
with the overall immersion time and led to almost the operated with the same F/M ratio as in the conventional
original value after 96 h. In this study, the required time process.
for successful membrane cleaning was too long since the The SRT (sludge retention time), which was at the
membrane was just immersed in a chemical solution. So, range of 20
/30 days, was determined from the inter-
the other appropriate methodology will be necessary for mittent removal ratio of sludge. This higher SRT
the effective short term cleaning of membrane. After the compared with that of the conventional process resulted
third cleaning for 96 h, the operation was started and the in a low sludge yield of 0.2
/0.25 kg VSS kg BOD 1 in
suction pressure did not increase until day 246 as shown this study. In the case of R2 and R3, the MLSS
in Fig. 2. concentrations and F/M ratio were maintained at
similar ranges with R1 as shown in Fig. 6.

3.3. Sludge concentration and F/M ratio

3.4. COD removal and nitrification efficiency
Due to the complete retention of all suspended solids
by the membrane, the MBR has the advantage of a The concentrations of organic compounds during the
higher biomass concentration compared with the con- operation of R1 are shown in Fig. 7. Influent COD
ventional process. Fig. 5 shows the variation of MLSS concentration of the synthetic wastewater was varied
concentration and the resulting F/M ratio (kg COD kg between 900 and 1600 mg l 1 and effluent COD was
SS1 day1) during the operation of R1. The F/M ratio mostly maintained at less than 30 mg l 1. Regardless of
is defined as S/uX , where S is influent COD concentra- the changes in the operating conditions such as the feed
tion (kg m 3), u is hydraulic retention time of the concentrations, the membrane flux determining a hy-
reactor (day), and X is the MLSS concentration (kg draulic retention time and MLSS concentration, an
m 3). Sludge was intermittently withdrawn from the excellent performance for COD removal was observed.
reactor in order to maintain its concentration around Since the pore size of the flat membranes in this study
9000
/12 000 mg l 1, so the F/M ratio was maintained was less than 0.4 mm, suspended solids (SS) was not
around 0.15
/0.3 depending on the membrane flux and detected in the effluent and effluent BOD was lower
the feed wastewater concentration. Many previous than 3 mg l 1 (data not shown).
studies on the submerged MBR have been performed Fig. 8 shows the concentrations of nitrogen com-
at a F/M ratio less than 0.1 since the target wastewater pounds for the operation of R1. Over the whole period,
284 J.K. Shim et al. / Process Biochemistry 38 (2002) 279
/285

the nitrification efficiency was more than 95% and the

nitrogen content in the permeate was almost all nitrate
(NO3). During the initial 80 days of R1 operation, the
influent wastewater was fed with a COD: N ratio of 16:1
for the first 10 days, 8:1 between day 11 and day 40, 4:1
between day 41 and day 60, and 2:1 between day 61 and
day 80, respectively. From day 61 to day 80, the effluent
ammonia concentration was less than 30 mg l 1 even at
an influent nitrogen concentration of 600 mg l1. This
high efficiency of nitrification resulted from a longer
SRT and a higher sludge concentration compared with
the conventional process.

4. Conclusions

A flat-type submerged membrane bioreactor for

wastewater treatment was studied in terms of design
Fig. 6. Variation of MLSS and F/M ratio in reactor 2 (R2). and operational considerations. Among the variables
determining the long-term stable membrane filtration,
this study mainly focused on the geometry of a
bioreactor in which the membrane is immersed. Filtra-
tion stability was maintained for a longer period when
the cross-sectional area of the downcomer (Ad) was
larger than that of the riser section (Ad). While an Ad/Ar
ratio of 1.6 in an aerated bioreactor did not generate a
sufficient flow for proper cleaning of membrane surface,
the ratio between 3.6 and 4.5 seemed appropriate for the
prevention of membrane clogging. It could be assumed
that sludge bulking, which was confirmed by SVI value
and microscopic observation, might affect the abrupt
deterioration of filtration performance of membrane
even though a systematic study was not carried out.
Chemical immersion was an effective method for the
restoration of a fouled membrane, but a more extensive
study on the strategy for membrane washing would be
Fig. 7. Variation of COD concentration of influent and effluent in necessary to shorten the cleaning time. Due to the
reactor 1 (R1). distinctive advantages of the membrane bioreactor, the
COD removal efficiency and the nitrification perfor-
mance was satisfactory for a high concentration of
organics and nitrogen compounds.

Acknowledgements

This work was supported by University of Ulsan

Research Fund of 2002.

References

[1] Hardt FW, Clesceri LS, Nemerow NL, Washington DR. Solids
separation by ultrafiltration for concentrated activated sludge.
JWPCF 1970;42(12):2135
/48.
Fig. 8. Variation of nitrogen concentration of influent and effluent in [2] Brockmann M, Seyfried CF. Sludge activity under the conditions
reactor 1 (R1). of crossflow microfiltration. Wat Sci Technol 1997;35(10):173
/81.
 J.K. Shim et al. / Process Biochemistry 38 (2002) 279
/285 285

[3] Yamamoto K, Hiasa M, Mahmood T, Matsuo T. Direct solid
/ [7] Chiemchaisri C, Yamamoto K, Vigneswawan S. Household
liquid separation using hollow fiber membrane in an activated membrane bioreactor in domestic wastewater treatment. Wat
sludge aeration tank. Wat Sci Technol 1989;21:43
/54. Sci Technol 1993;27:171
/8.
[4] Chiemchaisri C, Yamamoto K. Performance of membrane [8] Cote P, Buisson H, Pound C, Arakaki G. Immersed membrane
separation bioreactor at various temperatures for domestic waste- activated sludge for the reuse of municipal wastewater. Desalina-
water treatment. J Membr Sci 1994;87:119
/29. tion 1997;113:189
/96.
[5] Ueda T, Hata K. Domestic wastewater treatment by a submerged [9] Nah YM, Ahn KH, Yeom IT. Nitrogen removal in household
membrane bioreactor with gravitational filtration. Wat Res wastewater treatment using an intermittently aerated membrane
1999;33(12):2888
/92. bioreactor. Environ Technol 2000;21:107
/14.
[6] Gunder B, Krauth K. Replacement of secondary clarification by [10] Chisti Y, Moo-Young M. Improve the performance of airlift
membrane separation */results with tubular, plate and hollow reactors. Chem Eng Prog 1993;89(6):38
/45.
fiber modules. Wat Sci Technol 1999;40(4-5):311
/20. [11] APHA/AWWA/WEF. Standard methods for the examination of
water and wastewater, published by APHA, 18th ed., 1992.