a r t i c l e
i n f o
Article history:
Received 7 October 2009
Received in revised form
10 December 2009
Accepted 12 December 2009
Available online 22 December 2009
Keywords:
Mixing
Hydrodynamics
Residence time distribution
Membrane bioreactor
Energy
a b s t r a c t
The design and optimisation of MBR units require knowledge of the biokinetics, fouling potential and
mixing. Although the mixing within an MBR system is of critical importance to the performance, MBRs
are mainly designed on the basis of biokinetics and fouling potential of the treatment system while
assuming the hydrodynamic characteristics. One method to characterise the mixing is the residence
time distribution (RTD). In this work, tracer studies using lithium chloride were performed to acquire RTD
proles of two full-scale MBR systems with different membrane congurations (at sheet and hollow
bre). Analysis of the RTD proles indicated that that both MBRs, including their respective ltration
tanks, are very close to completely mixed. The mixing energy per volume of permeate used by the hollow
bre membrane vessel was lower than that of at sheet module MBR; both in terms of whole MBR energy
usage and membrane blower only energy usage. Hence, it is possible to conclude that the at sheet MBR,
per square metre of membrane, in this case, requires more energy to achieve a similar degree of mixing.
2010 Elsevier B.V. All rights reserved.
1. Introduction
Optimisation of membrane bioreactors (MBR) requires detailed
understanding of the kinetics of biological nutrient removal (BNR),
ltration separation performance of the microporous membranes
and the hydraulic conditions in the bioreactor. Although MBR performance has been the subject of many investigations, the literature
is silent on the hydrodynamics of MBR design, particularly the
impact of membrane conguration on mixing conditions [1,2].
Consequently, the mixing within the MBR process has been an
insufciently understood aspect of MBR design [3,4].
The performance of a reactor is largely inuenced by the retention time of a reactant in the reactor vessel. The retention time
of the reactant is determined by the mixing in the reactor [5].
Many researchers have reiterated the importance of mixing in
achieving efcient conversion of reactants in wastewater treatment processes [57] and to provide sufcient shear to prevent
fouling of membranes [8,9]. One method to characterise the mixing is based on the concept of residence time distributions (RTDs).
The degree of mixing energy input and the bioreactor and membrane conguration affects the output response (or RTD) which
describes the mixing regime occurring in the system. A number of
authors have employed residence time distribution analysis in the
Corresponding author. Tel.: +61 2 9385 6092; fax: +61 2 9385 5966.
E-mail address: g.leslie@unsw.edu.au (G. Leslie).
0376-7388/$ see front matter 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.memsci.2009.12.016
102
Fig. 1. Overview of the process setup with sample points marked with letters: one of the two parallel process streams of Site 1 (FS) MBR (left)(1) bioselector, (2) swing aerobic
& anoxic zones, (3) aerobic zone, (4) membrane ltration vessel; Site 2 (HF) MBR (right)(1) anoxic zone, (2) aerobic zones, (3) membrane ltration zone, (4) de-aeration
zone.
V
Q
(3)
The second moment of the function E(t) is referred to as the variance, 2 , which describes the spread of the RTD curve where a
higher value indicates more spread:
2 =
s3 =
2.1. Residence time distribution (RTD)
0
C(t) dt
(1)
tm =
tE(t) dt
0
(5)
(2)
2 =
(6)
2
( 1) E () d =
s 3 =
( 1) E () d =
0
(t tm )3 E(t) dt
RTD can be used to assess the degree of mixing and ow patterns within any reactors. Analysis of the RTD prole can provide
information on the degree of nutrient conversion at different locations in the reactor and the effect of the use of hollow bre or at
sheet membrane congurations on the degree of mixing.
RTD proles may be measured by monitoring the evolution of
an inert tracer through the reactor. The residence time distribution
function, E(t), can be evaluated by dividing temporal variation of
tracer concentration in the membrane ltrate, i.e. C(t), by the total
mass of tracer injected in the feed (i.e. the area under the C(t) curve)
as shown by Eq. (1).
E(t) =
(4)
2. Theory
C(t)
(t tm )2 E(t) dt
2
tm 2
s3
tm 3
(7)
(8)
where
=
t
tm
(9)
Burrows et al. [19] and Thirumurthi [20] have used various relationships between the hydraulic residence time (), peak time (tp )
and the mean residence time (tm ) to quantitatively assess the RTD
curves and the degree of plug ow, amount of dead zones and
Per
2
Per 2
(1 ePer )
(10)
Pi
PS,V =
Vi
PS,M =
AM
PS,P =
Pi
Pi
QP
(11)
(12)
(13)
103
P 0.5
(14)
V
0.41
dw 0.23X 0.37
dy
(15)
0.41
G0.23X
0.37
(16)
104
Table 1
Operating process parameters of the two MBRs during each trial.
Parameters
Units
ML/day
m3
m3
g/L
m2
L/m2 /h
m3 /h
days
Nm3 /h
Nm3 /h
1.09
852
392
11.3
Flat sheet
3835
11.8
461
16.6
109
992
1.10
435
36.0
5.0
Hollow bre
3200
14.3
433
9.9
419
918
produces recycled water for the site and local area. The plant is sized
to treat 2 ML/day of inuent and is designed for nutrient removal
so it possesses an anoxic zone, aerobic zone and an internal recycle.
The biological parameters of feed and efuent of both plants during
the tracer studies are shown in Tables 2 and 3.
The tracer studies were carried out using a pulse input of lithium
chloride delivered at the MBR inlet (post-screening) with the tracer
response being measured in the permeate and other relevant sample points. Lithium chloride is commonly used for tracer studies of
wastewater processes due to its inert nature [5,7,27]. The amount
of tracer used corresponded to a bulk concentration of 1.5 mg Li+ /L
(i.e. mass of lithium divided by volume of the MBR). The bulk
concentration was comparable to that used by other researchers
who have investigated mixing in wastewater treatment processes
[5,10,15,2729]. This ensured that the tracer response is much
greater than the detection limit of the analysis technique and the
lithium ion background concentration. The Li+ concentration was
measured using ICP-AES (inductively couple plasma-atomic emission spectrophotometry) and had a detection limit of 0.008 mg Li+ /L
Dosing solutions were prepared with concentrations of 4060 g
Li+ /L and a maximum dosage volume of 25 L; this ensured a small
dosage volume, low dosage time yet at a small density difference
between the dosage solution and mixed liquor. The dosage solution was pumped at approximately 75 L/min over 20 s into the inlet
stream. For Site 2 (HF) MBR this is 0.02% of the HRT and for Site 1
(FS) MBR is 0.004% of the HRT. This enabled effectively instantaneous delivery of the tracer. To obtain reproducible results at both
sites, the tracer studies were undertaken with as many constant
process parameters possible. The intermittent inuent ow and the
switching on/off of the aeration were still experienced. The tracer
studies commenced at exactly the same time of the day. Sampling
was undertaken for four hydraulic residence times ensuring close
to 100% tracer recovery. The sample points are marked with letter
in Fig. 1 with Sample Point A being the combined permeate which
is used to derive the overall MBR RTD and Sample Point B being
Table 2
Average feed characteristics during each trial.
Site
COD (mg/L)
BOD5 (mg/L)
NH3 -N (mg/L)
SS (mg/L)
608
482
260
200
55.0
33.0
284
325
TDS (mg/L)
1175
749
Temp. ( C)
pH
384
247
24.4
22.1
7.3
6.9
Table 3
Efuent characteristics during each trial.
Site
COD (mg/L)
BOD5 (mg/L)
NH3 -N (mg/L)
NOX -N (mg/L)
SS (mg/L)
TDS (mg/L)
Temp. ( C)
pH
48.8
29.0
2.0
4.5
0.7
0.1
1.5
16.1
1.5
1.0
960
786
112
78
25.6
23.3
7.5
7.2
105
Fig. 2. Normalised residence time distribution (RTD) from the combined ltrate for the at sheet MBR at Site 1 (left) and the hollow bre MBR at Site 2 (right). The RTDs were
measured on two separate occasions.
Table 4
Quantitative RTD properties.
Site
Trial
Rec. (%)
(h)
tm (h)
2
s 3
tp (h)
tp /
tm /
(1 frame=t opbotrowsep=0
tp /tm )
colsep=0 >
Per
1
2
99.7%
99.5%
27.5
27.5
31.3
31.4
0.882
0.897
1.75
1.79
3.33
3.92
0.121
0.142
1.14
1.14
0.894
0.875
0.388
0.334
1.13
1.11
1
2
82.7%
96.0%
10.3
10.3
10.4
10.2
0.806
0.803
1.21
1.24
0.67
0.83
0.065
0.081
1.01
0.99
0.936
0.918
0.662
0.697
1.24
1.25
2 (HF) MBR (Table 4). This is likely due to the Site 1 RTD having a longer tail than the Site 2 RTD curve which decreases more
abruptly.
The degree of mixing within the membrane ltration vessels
was also assessed through measurement of tracer response at an
alternative location in the tank. Any large difference between tracer
response in permeate and other point would indicate a deviation
from complete mixing. The position measured for Site 1 (FS) was
the Top Left Deck, that is the permeate extracted from the top
deck closest to the membrane ltration inlet from the left membrane ltration vessel. The position measure for Site 2 (HF) was the
RAS, that is the return activated sludge owing from the membrane ltration vessel over a weir into the de-aeration vessel. These
points are indicated in Fig. 1 as Sample Point B.
The largest difference between the permeate response and Sample Point B response has been found to be 0.4% of the mean
residence time (Fig. 3). This translates to at most a lag time between
the two sample points, as a fraction of the membrane ltration vessel theoretical residence time, of 1.5% for Site 1 and 5% for Site 2.
If the system vessel was completely mixed this lag time would be
equal to 0% between any points in the vessel and if it was a plug
ow system it would be up to 100%. This indicates that the mixing energy from aeration and recirculated sludge is very close to
complete mixing for both MBR membrane ltration tanks.
4.2. Mixing energy
The mixing energy was determined for each MBR (Table 5). The
drivers of mixing are considered to be mechanical mixing, bioreTable 5
MBR power usage.
Parameters
Units
Site 1 (FS)
MBR
Site 2 (HF)
MBR
Powermixer
Powerbioreactor blower
Powermembrane vessel blower
Powerrecirculation pump
Powerpermeate/backwash pump
Powertotal
kW
kW
kW
kW
kW
kW
7.1
3.3
29.5
16.0
55.8
2.2
8.5
13.8
18.5
2.0
44.9
106
Fig. 3. Comparison of residence time distributions obtained from specic sampling points within the at sheet MBR at Site 1 (left) and the hollow bre MBR at Site 2 (right).
Table 6
MBR specic power usage.
Parameters
PS,V (
Pi = membrane blower; Vi = Membrane VESSEL)
PS,V (
Pi = membrane blower, recirculation
pump and membrane ltration pump; Vi = membrane vessel)
PS,V (
Pi = total MBR energy usage; Vi = total MBR volume)
PS,P (
Pi = membrane blower)
PS,P (
Pi = membrane blower, recirculation pump and membrane ltration/backwash pump)
PS,P (
Pi = total MBR energy usage)
PS,M (
Pi = membrane blower)
PS,M (
Units
75.2
W/m
W/m3
116
W/m3
44.8
kWh/m3
0.651
0.301
kWh/m3
1.00
0.748
kWh/m3
1.23
0.982
W/m2
7.69
5.78
W/m2
11.9
10.7
Table 7
MBR average velocity gradient.
Parameter
Membrane vessel average velocity gradient (membrane blower only)
Membrane vessel average velocity gradient (membrane blower and
return activated sludge pumping)
Overall MBR average velocity gradient (G)
signicantly higher (Table 6). This again is due to the larger bioreactor and ltration vessel.
This pattern is repeated for the specic energy usage of the ltration vessel itself, although the extra vessel volume is solely due to
the lower packing density of the membrane vessels. It appears that
Site 2 requires less energy (per unit volume of permeate produced)
for membrane ltration mainly due to the packing density of membranes. Apart from a higher membrane packing density the other
advantage of Site 2 is that it is able to operate at a higher ux. This
leads to efciency gains on a permeate production basis is lower
for the Site 2 (HF) MBR. The ability to operate at a higher ux is
heavily inuenced by the lower MLSS and the resultant higher viscosity. Site 2 does however use more energy per unit vessel volume
perhaps indicating that efciency gains can be made on a energy
usage per unit volume of vessel. However, in terms of mixing energy
required per unit area of membrane (PS,M ) Site 1 (FS) MBR uses a
third extra power using the membrane blower motor as a basis
and only 10% extra if all membrane ltration vessel motors are
accounted for.
Units
s
s1
47.2
63.9
243.5
417.1
s1
33.0
101.2
The overall MBR average velocity gradient for the Site 2 (HF)
MBR is 5.2 times greater than for the Site 1 (FS) MBR (Table 7) which
is due to the lower MLSS at Site 2 (less than half of Site 1) and
therefore viscosity. This higher G factor translates into an earlier
peak in the RTD curve for Site 2, however the Site 1 RTD is, according
to the number of tanks in series and the Peclet Number, closer to
complete mixing than for Site 2. The average velocity gradient for
the ltration tank at Site 2 is also substantially greater than for
Site 1. While in the early research on pilot plant MBRs [10], the
G factor was evaluated using a constant viscosity for both MBRs,
hence the hollow bre MBR which requires less power input has a
lower overall MBR average velocity gradient.
5. Conclusions
The experimental methodology employed provided reproducible results with high recovery of tracer. The results show that
different MBR designs, both with respect to the bioreactor and
membrane ltration vessel, have differing effects on the RTD. Analysis of the RTD proles indicated that both MBRs are close to
complete mixing conditions, while Site 1 (FS) MBR is marginally
closer according to Peclet and the Tanks-in-Series measures. In
terms of energy usage per unit volume of permeate produced, the
Site 1 (FS) MBR has higher requirements. This is due to a number of factors. The at sheet membrane requires a larger volume
of ltration vessels due to the larger size of membrane modules
(i.e. lower packing density), while the bioreactor of Site 1 (FS)
MBR also requires a larger volume for higher inuent concen-
Nomenclature
List of symbols
AM
membrane area (m2 )
C
tracer concentration (mg/L)
E
dimensionless function of residence time distribution
dw/dy
applied velocity gradient (s1 )
G
average velocity gradient (s1 )
N
number of equal-size mixed ow reactors in series
P
Power input (W/kW)
Peclet number
Per
Pi
power draw of different motors (i.e. mixers, blower
or pumps) being considered (W/kW)
power per unit area of membrane (W/m2 )
PS,M
power per unit volume of permeate produced
PS,P
(kWh/m3 )
PS,V
power per unit volume of vessel (W/m3 )
Q
volumetric owrate (m3 /s)
permeate owrate (m3 /s)
Qp
s3
skewness of a tracer curve or distribution function
t
current time (s)
t50
the time at which 50% of the tracer has exited from
the reactor (s)
mean residence time (s)
tm
tp
peak time (s)
V
reactor volume (m3 )
volumes of the vessels being considered (m3 )
Vi
X
mixed liquor suspended solids (g/L)
Greek symbols
dynamic viscosity of uid (Pa s)
dimensionless time
2
variance of a tracer curve or distribution function
theoretical mean residence time/hydraulic retention time (HRT) (s)
Subscripts
i
the items being considered
M
membrane
P
permeate
V
volume of vessels
using dimensionless time units
107
presented here will also form part of the validation of a computational uid dynamics model of an MBR.
Acknowledgements
This project is proudly supported by the International Science
Linkages programme established under the Australian Governments innovation statement, Backing Australias Ability. This work
was also supported by the European Commission under the 6th
Framework Programme (AMEDEUS project, contract 018328). The
authors are grateful for the assistance provided by Sydney Water
and the South Australian Water Corporation during this project.
References
[1] A.N.L. Ng, A.S. Kim, A mini-review of modeling studies on membrane bioreactor (MBR) treatment for municipal wastewaters, Desalination 212 (2007)
261.
[2] H. Fletcher, T. Mackley, S. Judd, The cost of a package plant membrane bioreactor, Water Research 41 (2007) 2627.
[3] T.J. Rector, J.L. Garland, S.O. Starr, Dispersion characteristics of a rotating hollow
ber membrane bioreactor: effects of module packing density and rotational
frequency, Journal of Membrane Science 278 (2006) 144.
[4] S. Judd, The MBR Book: Principles and Applications of Membrane Bioreactors
in Water and Wastewater Treatment, Elsevier, Amsterdam/Boston/London,
2006.
[5] M.R. Pena, D.D. Mara, G.P. Avella, Dispersion and treatment performance analysis of an UASB reactor under different hydraulic loading rates, Water Research
40 (2006) 445.
[6] O. Levenspiel, Chemical Reaction Engineering, Wiley, New York, 1999.
[7] O. Potier, J.P. Leclerc, M.N. Pons, Inuence of geometrical and operational
parameters on the axial dispersion in an aerated channel reactor, Water
Research 39 (2005) 4454.
[8] N. Ndinisa, A. Fane, D. Wiley, Fouling control in a submerged at sheet membrane system: Part Ibubbling and hydrodynamic effects, Separation Science
and Technology 41 (2006) 1383.
[9] T. Taha, W.L. Cheong, R.W. Field, Z.F. Cui, Gas-sparged ultraltration using horizontal and inclined tubular membranesa CFD study, Journal of Membrane
Science 279 (2006) 487.
[10] Y. Wang, K.W. Ong, M.W.D. Brannock, G.L. Leslie, Evaluation of membrane
bioreactor performance via residence time distribution: effects of membrane conguration and mixing, Water Science and Technology 57 (2008)
353.
[11] Q. Yang, A. Drak, D. Hasson, R. Semiat, RO module RTD analyses based on directly
processing conductivity signals, Journal of Membrane Science 306 (2007)
355.
[12] M. Curlin, M. Matic, M. Matosic, I. Mijatovic, Z. Kurtanjek, Effects of Hydraulic
Residence Time and Mixing on Wastewater Treatment in a Membrane Bioreactor, Chemical and Biochemical Engineering Quarterly, Croatian Society of
Chemical Engineers, 2004, pp. 97104.
[13] J. Alex, R. Tschepetzki, U. Jumar, F. Obenaus, K.H. Rosenwinkel, Analysis and
design of suitable model structures for activated sludge tanks with circulating
ow, Water Science and Technology 39 (1999) 55.
[14] J. Alex, G. Kolisch, K. Krause, Model structure identication for wastewater
treatment simulation based on computational uid dynamics, Water Science
and Technology 45 (2002) 325.
[15] D. Olivet, J. Valls, M.A. Gordillo, A. Freixo, A. Sanchez, Application of residence
time distribution technique to the study of the hydrodynamic behaviour of a
full-scale wastewater treatment plant plug-ow bioreactor, Journal of Chemical Technology and Biotechnology 80 (2005) 425.
[16] G.C. Glover, C. Printemps, K. Essemiani, J. Meinhold, Modelling of wastewater treatment plantshow far shall we go with sophisticated modelling tools?
Water Science and Technology 53 (2006) 79.
[17] S. Kawamura, Integrated Design of Water Treatment Facilities, Wiley, New
York, 1991.
[18] Metcalf & Eddy, Wastewater Engineering: Treatment and Reuse, McGraw-Hill,
Boston, 2003.
[19] L.J. Burrows, A.J. Stokes, J.R. West, C.F. Forster, A.D. Martin, Evaluation of different analytical methods for tracer studies in aeration lanes of activated sludge
plants, Water Research 33 (1999) 367.
[20] D. Thirumurthi, A break-through in the tracer studies of sedimentation tanks,
Journal of Water Pollution Control Federation 41 (1969) 405.
[21] L.C. Smith, D.J. Elliot, A. James, Characterization of mixing patterns in an anaerobic digester by means of tracer curve analysis, Ecological Modelling 69 (1993)
267.
[22] H.S. Fogler, Elements of Chemical Reaction Engineering, Prentice Educational
International, Upper Saddle River, NJ, 2006.
[23] P.V. Danckwerts, Continuous ow systems: distribution of residence times,
Chemical Engineering Science 2 (1953) 1.
[24] S. Judd, The status of membrane bioreactor technology, Trends in Biotechnology
26 (2008) 109.
108
[29] M.R. Pena, D.D. Mara, A. Sanchez, Dispersion studies in anaerobic ponds: implications for design and operation, Water Science and Technology 42 (2000)
273.
[30] P. Cote, M. Masini, D. Mourato, Comparison of membrane options for water
reuse and reclamation, Desalination 167 (2004) 1.
[31] A. Garcs, W. De Wilde, C. Thoeye, G. De Gueldre, Operational Cost Optimisation
of MBR Schilde IWA 4th International Membrane Conference, IWA, Harrogate,
UK, 2007.