w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 9 9 e4 5 0 7

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Efficacy of relaxation, backflushing, chemical cleaning and

clogging removal for an immersed hollow fibre membrane
bioreactor

T. Zsirai a, P. Buzatu b, P. Aerts c, S. Judd a,*

a
Cranfield Water Science Institute, Cranfield University, Bedfordshire MK43 0AL, United Kingdom
b
University Politehnica of Bucharest, Romania
c
The Dow Chemical Company, USA

article info abstract

Article history: A pilot-scale hollow fibre immersed MBR, challenged with real municipal wastewater, was
Received 23 February 2012 studied and operated under conditions identical to those prevailing at full-scale to assess
Received in revised form the relative influence of backflushing, relaxation, chemical enhanced backflushing (CEB)
19 April 2012 and declogging on permeability decline and recovery. The influence of relaxation and
Accepted 3 May 2012 backflushing was initially assessed using the conventional flux step method; results indi-
Available online 16 May 2012 cated reversible fouling to be similar for each method, whilst the irreversible fouling rate
was significantly reduced by backflushing. For a given total backflush volume, fouling
Keywords: mitigation was found to be marginally better through employing higher backflush fluxes
Membrane bioreactor than longer backflush durations.
Hollow fibre The impact of the CEB on permeability recovery assessed at low and high fluxes indi-
Backflush cated operation at more conservative fluxes to yield more sustained permeability. Under
2 1
Chemical enhanced backflush more aggressive operating conditions - fluxes of up to 35 L m h at specific aeration
3 2
Clogging demand values of 0.25 Nm /(m h) e long-term permeability decline took place which was
Permeability recovery not significantly ameliorated by chemical cleaning. On declogging the membrane through
gentle agitation permeability recovery was significant, but was followed by a rapid
permeability decline over the course of a few hours. Results suggested control of clogging
to be of greater importance than that of fouling in sustaining permeability.
ª 2012 Elsevier Ltd. All rights reserved.

1. Introduction (Brepols, 2010; Zheng et al., 2011). Sustainable operation

Economical operation of microfiltration/ultrafiltration (MF/
UF) systems relies on optimisation of operating conditions,
normally through fouling minimisation. This entails identi-
fying the maximum flux at which the transmembrane
pressure (TMP) increases at an acceptable rate. The selection
of operation and maintenance (O&M) parameters for
sustaining membrane permeability is then generally based
on rules of thumb, pilot plant and full-scale plant studies
relies on physical cleaning through relaxation or back-
flushing, or a combination of both, supplemented with
periodic chemical cleaning in place (CIP). Relaxation, the
intermittent cessation of permeation, permits some flux
recovery if the membrane is immersed and scoured with air,
such as for immersed membrane bioreactors (iMBRs). Back-
flushing (BF) is the reversing of permeate back through the
membrane, and is routinely applied to hollow fire (HF) iMBRs
(Judd, 2010).

* Corresponding author. Tel.: þ44 1234 754842.

E-mail address: s.j.judd@cranfield.ac.uk (S. Judd).
0043-1354/$ e see front matter ª 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.watres.2012.05.004
4500 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 9 9 e4 5 0 7

During BF, the loosely attached cake detaches from the
membrane together with some colloidal and soluble material,
such as bound EPS (extracellular polymeric substances),
entrapped in it (Jiang et al., 2003; Smith et al., 2006; Kim and
DiGiano, 2006; Geng et al., 2007; Wu et al., 2008; Hwang
et al., 2009). Cake materials removed by physical cleaning
comprise “reversible fouling” (Judd, 2010). So-called “irre-
versible fouling”, thought to relate to membrane pore blocking
in the early period of the filtration cycle prior to cake forma-
tion (Jiang et al., 2003; Hwang et al., 2009), is removed only be
chemical cleaning (Judd, 2010). Reversible fouling can become
irreversible at overly challenging operating fluxes and
hydrodynamic conditions (Lin and Bérubé, 2007). For MBRs,
the CIP normally employs a combination of hypochlorite
(primarily for removing organic polymers through oxidation),
and mineral and/or organic acids (for dislodging scales and
metal dioxides through solubilisation) (Judd, 2010). For HF
iMBRs the CIP “maintenance” clean takes the form of
a chemically enhanced backflush (CEB), where chemicals are
added to the BF water to enhance foulant removal.
MBR operation thus relies upon appropriate selection of
physical and chemical cleaning protocols. Cleaning parameter
values taken from existing full-scale operating plants (Table 1,
predominantly from Judd, 2010) indicate filtration, relaxation

Table 1 e Summary of O&M protocols and fouling trends in full-scale installations.

Installation Feeda SADm, Filtration/backflush Filtration/backflush TMP Maintenance clean
Nm3/(hr m2) periodb min/sec flux, L/(m2h) increasec, bar (CEB) cycle time,
day (protocol)

Wenyu river R 0.2e0.28 9 min/60sec 20 (summer), 0.1e0.3 7 (0.5 g/l NaOCl)

5 (winter)/24
Avranches M 0.15 6 min/30 sec 18.5/30 e 7 (1 g/L NaOCl,
citric acid at pH 3)
Bei Xiao He M 0.33 12 min/60 sec 16.5/x e 14 (0.5 g/L NaOCl)
relaxation
Beijing Miyun M 0.22e0.35 7 min/60 sec 28 (summer), 0.1e0.6 7 (0.5e3 g/L)
relaxation 21 (winter)/x
Brescia M 0.127 11.9 min/85 sec 20e24/x e 7 (0.5 g/L NaOCl,
relaxation 0.2 wt% citric acid)
Dalsung M 0.12e0.16 12 min/3 min 12.5/x 0.065e0.56 14 (2 g/L NaOCl)
relaxation
Hammatsu M 0.2 9 min/60 sec 19/24 0.15 mean 14 (3 g/L NaOCl)
City
Jinqiao M e 9 min/60 sec 20e25/- e 7
Karcag M 0.64 12 min/60 sec 12/33 e e
Monheim M 0.3 6 min/30 sec 10/33 0.6 max 14 (2.5 g/L H2O2,
2 g/L NaOCl)
Nordkanal M 0.43 6.6 min/50 sec 23.7/38 e 7 (0.5 g/L NaOCl,
2 g/L citric acid)
Okchun M 0.2e0.232 12 min/3 min 10.3/x 0.065e0.53 14 (2 g/L NaOCl)
relaxation
Oroszlány M 0.67 12 min/60 sec 13e15/33 0.2e0.5 7 (0.15 g/L, citric
acid at pH 2)
Peoria M 0.2 12 min/45 sec 17.5/x 0.03e0.1 7 (0.20 g/L NaOCl)
relaxation
Perthes en M 1 9 min15 sec/ e 0.2e0.5 7 (NaOCl)
Gatines 45 sec
ShenDingHe M 0.26e0.31 e 25 0.1e0.25 7 (0.5 g/L NaOCl)
Sinopec Jinling M e 8 min/60 sec e/x 0.6 max 7 (0.5 g/L NaOCl)
relaxation
Traverse M 0.18 11.33 min/40 sec 19.2/x e 10 (0.2 g/L NaOCl);
relaxation 20 (2 g/L citric acid)
Ulu Pandan M 0.2 12 min/30 sec 25/x 0.015/year 14 (0.2 g/L NaOCl)
Basic American I 0.21 12 min/35 sec 17.5/x e 7 (0.1 g/L NaOCl)
foods
Machida Food I 0.2 9 min/60 sec 16/16 0.15 mean 14 (1 g/L NaOCl)
Factory
Shenghong I 0.3 9 min/60 sec 21/20 0.2e0.6 7 (0.2 g/L NaOCl)
Sinopec I e 9 min/60 sec 12/x 0.03e0.3 10 (0.3 g/L NaOCl)
Guangzhou relaxation
x
Backflushing is not applicable for relaxation mode or data has not been found.
a Feedwater type: R river, M municipal, I industrial.
b Backflushing applied unless relaxation stipulated.
c Increase in pressure over the cleaning cycle or time period.
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 9 9 e4 5 0 7
and BF durations of 5.8e15, 0e3, and 0e1 min respectively. In
general, the change in TMP over the course of a cycle should
not exceed 0.6 bar and, according to this data set, chemical
cleaning is applied every 7e14 days. It follows that the
maximum allowable rate of pressure change is 0.6 bar/week,
or 0.42 mbar/h, and that the CIP must be expected to return
the membrane to a condition whereby this underlying irre-
versible fouling rate is not exceeded.
The precise physical and chemical cleaning protocol
demanded for sustaining operation is dependent to a large
extent on the membrane flux and air scouring rates selected,
and on the fouling propensity of the mixed liquor. Whilst the
nature and characterisation of membrane foulant compo-
nents has formed the focus of many studies, with inconclu-
sive trends reported for full-scale plants (Reid et al., 2008;
Delrue et al., 2011), fewer studies have focused on cleaning
efficacy.
Studies of physical cleaning by BF have generally revealed
net permeability to increase more with BF flux and frequency
than with duration (Decarolis et al., 2001; Schoeberl et al.,
2005; Smith et al., 2006; Kim and DiGiano, 2006). In a study
of an HF iMBR the BF frequency (8e16 min) was found to have
a greater impact on foulant removal than either aeration
intensity (0.3e0.9 m3/h per m2) or BF duration (25e45 s)
(Schoeberl et al., 2005). Against this, it has been claimed that
BF may exacerbate pore plugging (Wu et al., 2008), and that an
extended filtration cycle time can increase BF efficiency if the
cake formed suppresses membrane pore plugging and so
enhance permeability recovery (Hwang et al., 2009). Optimi-
sation of the BF protocol can significantly improve perfor-
mance; initiation of a BF at an optimal threshold pressure has
been shown to reduce the BF volume by up to 40% and
maintain consistent 80e85% removal of organic matter (Smith
et al., 2006).
There have been relatively few studies of chemical clean-
ing impacts on MBR membrane permeability recovery. Recent
examples include practical, heuristic data from actual MBR
installations (Brepols et al., 2008; Brepols, 2010; Stüber et al.,
2009), optimisation methods such as statistical quality
control (Kim et al., 2011), and small-scale/ex-situ practical
determination (Raffin et al., 2011; Mahendran et al., 2011).
Practical studies have included an assessment of the impact if
chemical cleaning sequence (Brepols et al., 2008; Brepols,
2010) and, for a bench-scale anaerobic MBR, a study of phys-
ical and maintenance/recovery chemical cleaning efficacy
(Mahendran et al., 2011). Whilst Table 1 suggests a degree of
variability in operating and maintenance (O&M) conditions, in
practice these mean O&M parameter values do not appear to
correlate with sludge quality. Notwithstanding this, there
remains little understanding of the basic differences in effi-
cacy of backflushing, relaxation and chemical cleaning by
CEB, particularly for full-scale modules operating under
hydrodynamic conditions representative of those encoun-
tered at full scale.
This work presents the results of such a study, focussing on
(a) permeability decline over the course of the cleaning cycle
(i.e. the period between chemical cleans), and (b) permeability
recovery from chemical cleaning using chemically enhanced
backflushing. The work includes analysis of both fouling and
clogging, the latter being assessed qualitatively.
4501
2. Material and methods
2.1. Pilot plant
The pilot plant comprised a 2.4 m3 bioreactor with a vertically
mounted module containing seven membrane elements
(Table 2), each fitted with a single header at the base. The three
central elements (E1eE3) of differing permeabilities, E1 having
a substantially lower surface porosity, but otherwise similar
specifications (Table 3) were used for testing. The elements
were individually fitted with peristaltic pumps (Watson Mar-
low, 520 Du) and pressure sensors for permeate extraction and
transmembrane pressure (TMP) monitoring respectively. The
two pairs of outer elements were each connected to further
peristaltic pumps (Watson Marlow, 620 Du). The module
frame was fitted with a skirt to ensure air, provided by eight
coarse-bubble aerators having 6 mm ports and placed 10 cm
below the membrane elements, was directed through the
membrane element channels. Air flow was applied intermit-
tently (Table 4) through a Programmable Logic Control
(Schneider) actuated solenoid valve and regulated with flow-
meters. Scoping trials conducted on a module fitted with three
identical test elements revealed the recorded fouling rate
standard deviation to be generally within 10% across the three
modules.
The biotank base was fitted with both a fine-bubble diffuser
(FBD) for providing dissolved oxygen (DO) and a ring aerator
for mixing. The FBD and ring aerators collectively delivered air
at a rate of 110 L/min; supplementary tests revealed this air
flow not to significantly contribute to membrane scouring.
The DO concentration was continuously monitored, and
aeration actuated via a solenoid valve with the PLC.
The TMP and permeate flow were respectively monitored
every 10 s by immersed pressure transducers and ultrasonic
flow metres, connected to a data logger (Pico data logging
system, Pico Technology) and monitored by a PC. The pressure
transducers were installed adjacent to the permeate header to
ensure accuracy of the readings. Permeate was collected in de-
aeration/backflush tanks equipped with ultrasonic flow
metres at the discharge, preventing suspended air bubbles
from interfering with measurements. An additional tank was
Table 2 e Hollow fibre membrane element specifications.
Parameter Value
Membrane material PVDF
Pore size, mm 0.2
Internal fibre, mm 1.2
Extenal fibre, mm 2.2
Fibre length, mm 1500
Elements dimensions 169  40  1620
(L  W  H), mm
Number of fibres per element 315
Fibre separation, mm 1
Element separation, mm 20 mm
Element membrane area, m2 3.33
Module (seven elements) total 23.4
membrane area, m2
4502 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 9 9 e4 5 0 7

2.3. Plant operation

Table 3 e Membrane material specification.
Sample K, BP pore dia, Mean flow All filtration cycles were 10 min long (Table 4). Initial studies
LMH/bar mm pore dia, mm
compared the irreversible fouling rate dPirr/dt (based on the
E1 300 0.670 0.128 physically-recovered TMP at the start of the 10 min filtration
E2 2160 0.981 0.216 cycle) sustained by a 1 min relaxation compared with 30 s of
E3 1560 0.845 0.185 backflushing at a flux of 25 LMH. A standard flux step method
was used, with step length of 1 h. Subsequent trials employed
backflushing exclusively, step durations of 16e23 h. The
membrane elements were subjected to daily CEBs after each
connected to the three test permeate streams via three way test using 500 mg/l sodium hypochlorite applied at 25 LMH
valves for daily chemically enhanced backflushing (CEB). based on 10 pulses of 30 s with 2 min relaxation between
The test and the control pumps were connected to and pulses e similar to protocols applied at full scale (Table 1).
controlled by programmed software (Debaru Kft) through the The impact of BF time and flux on fouling was studied over
PLC system. An additional peristaltic pump was used for a 21 h period at a constant operating flux J of 25 LMH and total
hourly sludge wasting to control the solid retention time (SRT) filtration cycle time (filtration þ backflush) of 10 min. BF
at around 30 days to maintain the mixed liquor suspended duration sp was altered from 10 to 30 s at fixed backflush flux
solids (MLSS) concentration at 10 g/l. The bioreactor was fed of 25 LMH, and subsequently BF flux Jb altered from 10 to
with settled sewage (427  20% mg/L TSS, 164  23% mg/L COD 25 LMH at a BF duration of 30 s. The net flux Jnet is then given
and 35  27% mg/L NeNH3) from the primary settlement tank by (Judd, 2010):
at the Cranfield University Wastewater Treatment Works,

with a float switch used to control the maximum bioreactor n Jtp Jb sp
Jnet ¼ (1)
sludge level. At the 8 h hydraulic residence time (HRT) and tc þ s p
10 g/L MLSS (w7 g/L MLVSS), the calculated F:M ratio was
where tp and tc are the filtration cycle and CEB cycle times, sp
w0.07/d e comparable with reported full-scale plants (Delrue
and sc and are the BF and CEB durations and n is the number of
et al., 2011).
filtration cycles per CEB (i.e. tc/(tp þ tc)).
The permeability recovery from backflushing for constant
2.2. Analytical methods flux operation is given by:

ð Þ
Feed, effluent and sludge were monitored twice weekly from dP
 
tp
samples collected from the reactor vessel top and base to dt
Rback flush ¼ 1  100% (2)
assess mixing efficiency. Measured feedwater quality dpp
parameters included total suspended solids (TSS), total and
where dP/dt is the irreversible fouling rate in mbar/min (Fig. 1),
soluble chemical oxygen demand (COD), and ammonia
tp the BF cycle time of 10 min and dPp the change in pressure
(NHþ4 N), all according to standard methods (APHA, 1999).
across the filtration cycle.
Permeate was monitored for ammonia and nitrate (NO3 N)
Permeability recovery from the chemical clean (CEB) or
to ensure nitrification, as well as turbidity for membrane
declogging is defined as:
integrity. Sludge quality was monitored with reference to non-
volatile and voliaite mixed liquor suspended solids (MLSS and Lc Lend
Rclean ¼  100% (3)
MLVSS respectively), sludge volume index (SVI), capillary Lstart Lend
suction time (CST), and soluble chemical oxygen demand
where Lstart, Lend and Lc refer to the respective permeabilities
(sCOD), all according to standard methods (APHA, 1999). COD
(J/P) in L/(m2h bar) at the start and end of the filtration cycle,
was determined using a Merck Spectroquant (Nova 60 Spec-
and after cleaning. The effect of successive CEB applications
trophotometer, Spectroquant).
using sodium hypochlorite at 500 mg/L concentration was
assessed, along with declogging by ex-situ washing with low-
pressure water combined with gentle agitation. For the CEB
trials the plant was operated at gross fluxes of 22.5e27.4 and
32e38.8 LMH for 21 h under constant BF conditions. Four CEBs
Table 4 e Operation conditions.
in succession were applied, interspersed with 15 min of
Parameter Value(s) filtration at 25 LMH to allow measurement of Rclean following
Hydraulic retention time, 8 each clean.
HRT, h
MLSS concentration, g/L 10
Flux range, J, L/(m2h) or LMH 5e40
Specific aeration demand, 0.25 3. Results and discussion
SADm, Nm3/(hr m2)
Aeration mode Intermittent (10ʺ on/ 10ʺ off) 3.1. Permeate and sludge quality
Filtration mode, relaxation 10 min suction/60 s relaxation
Filtration mode, backflush 9.5 min suction/30 s backflush
The permeate turbidity remained below 0.5 NTU throughout,
Backflush flux 25
and 0.3 NTU 30% on average: there was no evidence of
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 9 9 e4 5 0 7 4503

a b

Fig. 1 e Examples of fouling rate responses: (a) 1 h flux step (E1 membrane), and (b) 21 h run (E1eE3 membranes) indicating
the irreversible fouling rate.

membrane integrity impairment throughout this study.
Ammonia and COD removal were 96 and 98% on average.
Sludge quality was reasonably consistent at 10.8  7% g/L
MLSS, 119  29% mL/g SVI, and 10.9  29% s CST.
3.2. Relaxation vs. backflush (BF)
Typical transients are shown in Fig. 1a and b for flux steps of 1
and 21 h respectively. Irreversible fouling rates (dPirr/dt) were
determined by linear regression of the maximum pressure
reading (Pmax) data (i.e. those recorded prior to the BF).
Reversible fouling rates (dPrev/dt), i.e. the rate of pressure
increase during a 10-min filtration cycle, followed a similar
exponential relationship with flux to that recorded for
relaxation-based operation on the same membrane material
(Monclús et al., 2010). Small differences (Fig. 2a) are attribut-
able to differences in sludge concentration and quality. For
irreversible fouling, however, relaxation was far less effective
in maintaining permeability than backflushing (Fig. 2b), which
suppressed fouling to below 0.08 mbar/min (w0.8 bar/week).
This was significantly less than that reported from previous
similar pilot-scale studies employing the same membrane
scouring conditions (Yigit et al., 2009, tests 1e2, Fig. 2b) and
slightly lower than that reported for a bench-scale system
operating at much higher specific aeration demand, SADm
(Wu et al., 2008), though comparable to data reported by Geng
et al. (2007) for their pilot-scale study conducted at lower MLSS
levels (3.3e4.1 g/L).
Increasing the step length to 21 h produced measurable
increases in irreversible fouling (Fig. 1b), the rate of which (i.e.
dPirr/dt) was found to increase roughly linear with applied flux
(Fig. 3a) e albeit with significant data scatter e with a corre-
sponding decrease in permeability (Fig. 3b). This is at odds
with the quasi-exponential behaviour of reversible fouling
shown in Fig. 2 and reported many times previously (Judd,
2010). However, even at fluxes exceeding 35 LMH the fouling
rate was below 0.08 mbar/min. Moreover, this fouling rate is
somewhat conservative, since dPirr/dt tended to decrease with
time (Fig. 1b).
Whilst the two higher-permeability membranes E2 and E3
provided a correspondingly higher mean operating perme-
ability than the E1 membrane (Fig. 3b), the difference
decreased with increasing flux. This derives from the more
rapid fouling exerted at the higher membrane permeabilities
arising at the start of each cleaning cycle, with permeabilities
decreasing more significantly over the course of the 21 h cycle
at the higher imposed fluxes. The difference in permeability
between the fibres was thus greatest at the most conservative
operating flux of 25 LMH, where the E1 permeability was
respectively 31e41% and 21%e25% lower at start and end of

Fig. 2 e Reversible (a) and irreversible (b) fouling rate vs. operating flux, E1.
4504 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 9 9 e4 5 0 7

Fig. 3 e (a) Irreversible fouling rate and (b) permeability vs. operating flux for membrane E1, E2 and E3, backflushing for 30 s
at 25 LMH every 9.5 min.

the 36 LMH flux step than that of the two higher-permeability
membranes. The impact of membrane permeability thus
decreased with increasing flux.
3.3. Backflush trials
The impact of BF duration (30e10 s) on baseline fouling rate for
a BF flux of 25 LMH and a total filtration time fixed at 10 min is
shown in Fig. 4a; the impact of BF flux (25e10 LMH) at a fixed
duration of 30 s is shown Fig. 4b. Data refer to the mean linear
fouling rate over the 21 h cycle, and the change in backflush
condition is expressed as delivered backflush water volume.
Duplicate data are provided for both increasing and
decreasing backflush condition, with CEBs conducted at the
end of each run to recover permeability.
Plots indicate an increased BF volume to generally decrease
irreversible fouling rate. Marginally greater efficiency was
demonstrated from increasing the BF flux (Jb) (Fig. 4b) than
from the duration (sp) (Fig. 4a) as reflected in the steepness of
the dPirr/dt:BF volume gradient, corroborating other reported
findings (Kim and DiGiano, 2006; Huang et al., 2007; Wu et al.,
2008; Zheng et al., 2011). Increasing the total BF volume from
w0.25 to 0.7 L, correspondingly decreasing the net flux from
w23.6 to 22.4 LMH, (i.e. by 5%) approximately halved the
fouling rate. To obtain the same reduction in fouling rate by
adjusting the flux would, according to Fig. 3, require the
operating flux to be reduced to around 18 LMH (i.e. by 40e50%).
Clearly, it is more cost effective to apply more backflush water
than to reduce the flux to attain the same sustainable state in
this instance.
Significant differences in the fouling rate trend with
applied BF water volume were observed across the three
membranes, with 35e43% increased recovery for the higher-
permeability membranes over that of the E1 membrane.
Also, no significant change in the absolute permeability
recovery (at 25 LMH for 30 s) was evident with time: conse-
quently, the % permeability recovery tended to decrease with
increasing fouling, in accordance with previously reported
studies (Hwang et al., 2009; Yigit et al., 2009; Yang et al., 2011).
3.4. CEB trials
The effect of successive CEB applications with sodium hypo-
chlorite (500 mg/l) was assessed at low (22.5e27.4 LMH, Fig. 5a)
and high (32e38.8 LMH, Fig. 5b) operating fluxes. The initial
CEB yielded a similar Rclean (Eq. (3)) of around 65e75% for all
membranes at both fluxes. However for subsequent CEBs the
% recovery attained for low-flux operation greatly exceeded
that from high-flux operation e by around a factor of two e
again indicating the benefit of operation at more conservative
fluxes. The increased Rclean attained by the low-permeability
membrane (E1) reflects its more highly fouled condition,

Fig. 4 e Irreversible fouling rate against (a) backflush duration (30e10 s), at a backflush flux of 25 LMH, and (b) backflush flux
(25e10 LMH), at a backflush duration of 30 s, expressed in applied backwash volume at a 25 LMH operating flux.
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 9 9 e4 5 0 7
Fig. 5 e Average Rclean (%), (a) low-flux (22.5e27.4 LMH), and (b) high-flux (32e38.8 LMH) operation vs. CEB for each
membrane (E-1, black; E-2, grey; E-3, white; Average, hatched).
providing greater relative Rclean. The increased recovered
permeability with successive CEBs is intuitive and in keeping
with reported lab-scale data indicating increased permeability
recovery with treatment duration (Yan et al., 2010).
The standard CEB yielded an increased Rclean throughout
the campaign, from between 70e110% at the start to
130e220% by the end (Fig. 6). This provides an indication
accumulation of materials suppressing permeation over the
course of time, notwithstanding the apparent efficacy of
the CEB.
3.5. Clogging
It is widely recognised by practitioners that clogging forms
a key limitation to iMBR operation (Santos et al., 2011; Lesjean
et al., 2011), with its most obvious manifestation being the
failure of chemical cleaning to recover permeability. Since
clogging is otherwise indistinguishable from irrecoverable
fouling as far as performance impacts are concerned, both
leading to an inexorable permeability decline, clogging can
currently only be monitored through periodic inspection of
the membrane module.
On withdrawing and inspecting the test elements, clogging
was clearly visible as dark brown particulate matter partially
filling the voids space between the fibres along the length of all
the membrane elements. Manual separation of the individual
Fig. 6 e % Rclean with standard CEB vs. average baseline permeability (mbar/min) over the study.
4505
fibres revealed relatively hard material (compared with the
biotank flocs), characteristic of widely encountered “sludging”
or “localised dewatering” (Stone and Livingston, 2008; Mason
et al., 2010) rather than “ragging” by long rags or braids from
agglomerated cellulosic matter (Mason et al., 2010; Stefanski
et al., 2011).
The requirement of balancing the flux with membrane
aeration to suppress localised dewatering is well understood,
and yet the impact of these two key operating parameters on
sludging specifically has rarely been reported. A bench-scale
study of the impact of membrane aeration rate on clogging
within fibre bundles showed that a specific aeration demand
(SADm) of 1.36 Nm3/(h m2) allowed non-sludging operation,
whereas operation under high-flux (30 LMH) conditions at
0.45 Nm3/(h.m2) did not (Lebegue et al., 2009). At the lower
aeration rate “dead zones” were reported, corresponding to
agglomerated solids accumulation. Clogging must clearly also
be influenced by design parameters, and specifically packing
density, aerator type/positioning and membrane module
dimensions.
3.6. Declogging
Whilst manifestation of clogging is widely recognised, its
relative impact compared with fouling has not been previ-
ously assessed. Declogging of the membrane was attempted
4506 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 9 9 e4 5 0 7
Fig. 7 e 21 h trial, 25 LMH: (a) prior to, and (b) after declogging, with CEB using 500 mg/l NaOCl at the end of each trial.
by manual agitation accompanied by low pressure washing to
remove the less tenacious of the clogged material. Membranes
were then re-inserted into the tank and a further test con-
ducted at 25 LMH for 21 h following the standard CEB.
Visual inspection revealed more than 90% of the solids to
be removed by the manual declogging. The TMP transients
produced prior to and following declogging are shown in Fig. 7.
The clogged membranes (Fig. 7a) had a similar dPirr/dt profile
to those recorded originally (Fig. 1b), but with a lower
permeability. Across the two months duration of the
campaign, the permeability decreased from 141 to 96, from
241 to 146 and from 206 to 106 LMH/bar for membranes E1, E2
and E3 respectively. Given that E2 and E3 had been operating
under nominally the same conditions, and were also initially
demonstrated as having similar permeability, the subsequent
divergence on operating permeability can only be attributed to
either permanent fouling or clogging.
Following declogging, the permeability of the E1, E2 and E3
membranes increased to 150, 310 and 290 LMH/bar, respec-
tively - the highest permeabilities attained across the whole
study other than for the virgin membrane (Fig. 8) - with cor-
responding permeability recoveries of w50%, albeit with
subsequent rapid fouling (Fig. 7b). As shown in Fig. 8, the
absolute permeability recovery (normalised against that of the
virgin membrane) from declogging is 61e77%, with the less
Fig. 8 e Absolute membrane permeability recovery from
CEBs over the course of the campaign.
innately permeable membrane providing the lowest absolute
permeability recovery. Suppression of clogging alone can thus
potentially maintain a higher permeability than that from
conventional fouling control.
4. Conclusions
Results from a two-month study of full-scale hollow fibre
membrane modules in an immersed membrane bioreactor
have demonstrated the following overall trends in perme-
ability and its recovery by physical and chemical cleaning:
1. The high-permeability membranes initially provided
higher net permeability operation than the low-
permeability material, but this difference diminished at
higher fluxes and over the course of time due to the onset of
clogging.
2. An increased backflush flux was found to be slightly more
effective than increased backflush duration for the same
overall backflush volume.
3. Permeability could be largely recovered through sequential
CEBs, with little or no difference in behaviour between
membranes of differing innate permeability, though an
insidious decline on permeability was evident.
4. All membranes were found to become significantly clogged
after two months of operation, with the least innately
permeable membrane providing the lowest recovery of
permeability to its nominally clean condition.
5. A cursory examination of simple mechanical declogging
revealed it to be generally more effective than the CEB in
recovering permeability, albeit with a more rapid subse-
quent fouling rate.
Bulk membrane permeability appeared to have minimal
impact on sustainable performance of the MBR, since the
diminution of available membrane area due to clogging
becomes more significant at high fluxes regardless of the
membrane material properties. However, the CEB is effective
in sustaining permeability under more conservative operating
conditions (25 LMH flux, 0.25 SADm). Control of clogging over
the long term becomes critically importance if more aggres-
sive operating conditions of flux are to be sustained.
 w a t e r r e s e a r c h 4 6 ( 2 0 1 2 ) 4 4 9 9 e4 5 0 7
Acknowledgements
The authors are grateful to Dow Chemical, who funded the
work.
references
APHA, 1999. American Public Health Association/American Water
Works Association/Water Environment Federation. Standard
Methods for the Examination of Water and Wastewater,
twenty first ed. APHA/AWWA/WEF, Washington, District of
Columbia, US.
Brepols, C., Drensla, K., Janot, A., Trimborn, M., Engelhardt, N.,
2008. Strategies for chemical cleaning in large scale
membrane bioreactors. Water Science and Technology 57 (3),
457e463.
Brepols, C., 2010. MBR Practice Report: Operating Large Scale
Membrane Bioreactors for Municipal Wastewater Treatment.
IWA Publishing.
Decarolis, J., Hong, S., Taylor, J., 2001. Fouling behavior of a pilot
scale inside-out hollow fiber UF membrane during dead-end
filtration of tertiary wastewater. Journal of Membrane Science
191 (1e2), 165e178.
Delrue, F., Stricker, A.E., Mietton-Peuchot, M., Racault, Y., 2011.
Relationships between mixed liquor properties, operating
conditions and fouling on two full-scale MBR plants.
Desalination 272 (1e3), 9e19.
Geng, Z., Hall, E.R., Bérubé, P.R., 2007. Modeling the impact of
permeate flux and hydrodynamic conditions on fouling in
submerged hollow fiber membranes. Water Science and
Technology: Water Supply 7 (4), 111e118.
Huang, H., Lee, N., Young, T., Gary, A., Lozier, J.C., Jacangelo, J.G.,
2007. Natural organic matter fouling of low-pressure, hollow-
fiber membranes: effects of NOM source and hydrodynamic
conditions. Water Research 41 (17), 3823e3832.
Hwang, K.-J., Chan, C.-S., Tung, K.-L., 2009. Effect of backwash on
the performance of submerged membrane filtration. Journal
of Membrane Science 330 (1e2), 349e356.
Jiang, T., Kennedy, M.D., Van der Meer, W.G.J., Vanrolleghem, P.A.,
Schippers, J.C., 2003. The role of blocking and cake filtration in
MBR fouling. Desalination 157 (1e3), 335e343.
Judd, S.J., 2010. The MBR Book, Principles and Applications of
Membrane Bioreactors in Water and Wastewater Treatment,
second ed. Elsevier.
Kim, J., DiGiano, F.A., 2006. A two-fiber, bench-scale test of
ultrafiltration (UF) for investigation of fouling rate and
characteristics. Journal of Membrane Science 271 (1e2), 196e204.
Kim, M.J., Sankararao, B., Yoo, C.K., 2011. Determination of MBR
fouling and chemical cleaning interval using statistical
methods applied on dynamic index data. Journal of Membrane
Science 375 (1e2), 345e353.
Lebegue, J., Heran, M., Grasmick, A., 2009. Membrane air flow
rates and HF sludging phenomenon in SMBR. Desalination 236
(1e3), 135e142.
Lesjean, B., Tazi-Pain, A., Thaure, D., Moeslang, H., Buisson, H.,
2011. Ten persistent myths and the realities of membrane
bioreactor technology for municipal applications. Water
Science and Technology 63 (1), 32e39.
Lin, H., Bérubé, P.R., 2007. Modeling the impact of permeate flux
and hydrodynamic conditions on fouling in submerged hollow
4507
fiber membranes. Water Science and Technology: Water
Supply 7 (4), 111e118.
Mahendran, B., Lin, H., Liao, B., Liss, S.N., 2011. Surface properties
of biofouled membranes from a submerged anaerobic
membrane bioreactor after cleaning. Journal of Environmental
Engineering 137 (6), 504e513.
Mason, S., Ewert, J., Ratsey, H., Sears, K., Beale, J., 2010. Flat sheet
membrane bioreactors operational experiences e a New
Zealand perspective. In: Proceedings of NZWWA Conference,
Christchurch, Canterbury, New Zealand, September 2010.
Monclus, H., Zacharias, S., Santos, A., Pidou, M., Judd, S., 2010.
Criticality of flux and aeration for a hollow fiber membrane
bioreactor. Separation Science and Technology 45 (7),
956e961.
Raffin, M., Germain, E., Judd, S.J., 2011. Optimisation of
microfiltration cleaning for indirect potable reuse. Separation
and Purification Technology 80, 452e458.
Reid, E., Liu, X., Judd, S.J., 2008. Sludge characteristics and
membrane fouling in full-scale submerged membrane
bioreactors. Desalination 219 (1e3), 240e249.
Santos, A., Ma, W., Judd, S.J., 2011. Membrane bioreactor: two
decades of research and implementation. Desalination 273 (1),
148e154.
Schoeberl, P., Brik, M., Bertoni, M., Braun, R., Fuchs, W., 2005.
Optimization of operational parameters for a submerged
membrane bioreactor treating dyehouse wastewater.
Separation and Purification Technology 44 (1), 61e68.
Smith, P.J., Vigneswaran, S., Ngo, H.H., Ben-Aim, R., Nguyen, H.,
2006. A new approach to backwash initiation in membrane
systems. Journal of Membrane Science 278 (1e2), 381e389.
Stefanski, M., Kennedy, S., Judd, S., 2011. The determination and
origin of fibre clogging in membrane bioreactors. Journal of
Membrane Science 375 (1e2), 198e203.
Stone, M, Livingston, D, 2008. Flat plate MBR energy consumption
e village of Dundee, MI. In: Membrane Technology 2008
Conference Proceedings of the Water Environment
Federation, Alexandria (VA) Jan 27e30.
Stüber, J., Lüdicke, C., Lesjean, B., Gnirss, R., Kraume, M., 2009.
Operation of MBR membrane modules used in a decentralised
wastewater treatment plant: field study and comparison of
different cleaning strategies. Desalination and Water
Treatment 9 (1e3), 181e188.
Wu, J., Le-Clech, P., Stuetz, R.M., Fane, A.G., Chen, V., 2008. Effects
of relaxation and backwashing conditions on fouling in
membrane bioreactor. Journal of Membrane Science 324 (1e2),
26e32.
Yan, X., Gerards, R., Vriens, L., Vankelecom, I., 2010. Hollow fiber
membrane fouling and cleaning in a membrane bioreactor for
molasses wastewater treatment. Desalination and Water
Treatment 18 (1e3), 192e197.
Yang, L., Wang, Z., Sun, Y., Hu, Z., Zhao, S., Wang, X., Li, W., Xi, X.,
Zhang, J., Pei, Z., 2011. Influence of various operating
conditions on cleaning efficiency in sequencing batch reactor
(SBR) activated sludge process. Part II: backwash and water
rinsing introduced membrane filtration process. Desalination
272 (1e3), 76e84.
Yigit, N.O., Civelekoglu, G., Harman, I., Koseoglu, H., Kitis, M.,
2009. Effects of various backwash scenarios on membrane
fouling in a membrane bioreactor. Desalination 237 (1e3),
346e356.
Zheng, X., Ernst, M., Jekel, M., 2011. Stabilizing the performance of
ultrafiltration in filtering tertiary effluent-Technical choices
and economic comparisons. Journal of Membrane Science 366
(1e2), 82e91.