CFD simulations of membrane filtration zone in a submerged hollow fibre membrane bioreactor using a...

KMHCPH

cbciutw(thttc

Journal of Membrane Science 363 (2010) 57–66

Contents lists available at ScienceDirect

Journal of Membrane Science

journa l homepage: www.e lsev ier .com/ locate /memsci

FD simulations of membrane filtration zone in a submerged hollow fibreembrane bioreactor using a porous media approach

uan Wang, Matthew Brannock, Shane Cox, Greg Leslie ∗

NESCO Centre for Membrane Science & Technology, School of Chemical Engineering, University of New South Wales, Sydney, NSW 2052, Australia

r t i c l e i n f o

rticle history:eceived 30 March 2010eceived in revised form 1 July 2010ccepted 3 July 2010vailable online 13 July 2010

eywords:embrane bioreactorsollow fibreFDorous mediaydrodynamics

a b s t r a c t

The current membrane bioreactor (MBR) design methods and the popular bio-kinetic models rely on theassumption that membrane bioreactor is completely mixed, neglecting the real hydrodynamic conditionwithin the reactor. MBRs differ from conventional reactors in so far as the spatial distribution of reactordischarge points is very broad for an MBR compared with a conventional bioreactor. ComputationalFluid Dynamics (CFD) provides a possibility to investigate the hydrodynamic behaviour of large scaleMBRs. The CFD modelling of whole MBR plant requires a macro-scale approximation which can keepthe mesh size and computation effort within the reasonable limit. However, the simulation of the flowbehaviour surrounding the membranes requires high mesh resolutions and hence large element numbers.Therefore, it is impossible to model each individual hollow fibre using the Navier–Stokes equations asit is too computationally costly. In this paper, the effects of Siemens Memcor Memjet® hollow fibremembrane bundle on flow field were transferred to a porous media model. This porous media model

. Introduction

For a given wastewater, the parameters affecting the MBR designan be categorized into three groups, (i) biological factors, (ii) mem-rane factors, and (iii) hydrodynamic factors [1]. The hydrodynamiconditions of a reactor can be characterised by the degree of mix-ng [2]. Current MBR design tools such as BioWin® or WEST® [3]tilise variants of the Activated Sludge Model (ASM) [4–7] andherefore usually assume that the mixing characteristics confirmith either complete mixing (CSTR) for aeration tanks, or plug flow

PFR) for some anoxic channels. Unlike conventional wastewaterreatment process that have a single discharge point, MBR mayave multiple discharge points resulting from its variable posi-ions along the reactor length [8]. Therefore, non-ideal mixing (e.g.

Research has been carried out on the characterisation of mixingegimes of the whole full-scale MBR plants by acquiring residence

∗ Corresponding author. Tel.: +61 2 9385 6092; fax: +61 2 9385 5966.E-mail address: g.leslie@unsw.edu.au (Greg Leslie).

time distribution profiles through tracer studies [9] and compu-tational fluid dynamics (CFD) modelling [10]. Results from thesestudies suggested that the full-scale MBRs being examined wereclose to CSTR conditions but not 100% completely mixed and there-fore the ideal flow models cannot be used to predict the flowregimes within the reactor. Another study [11] tried to couple theASM No. 1 with the CFD model to predict the biological treatmentperformance and compared various biological process variablesagainst COST benchmark [3]. Minor difference on nutrient conver-sion was observed due to the deviation from CSTR condition. Thesewere the first efforts ever to investigate the mixing behaviour andits impact on nutrient removal of the entire full-scale MBR.

The mixing of entire MBR plant is affected by the power inputsrequired by the aeration for biological reactions, pump required forreturned activated sludge and aeration for membrane scouring [9].By contrast, the mixing behaviour of membrane filtration zone itselfis only affected by the aeration energy for membrane scouring. To

5 mbrane Science 363 (2010) 57–66

tfiTaeishmttomEtbis[flflepeh

trmmpBd

8 Y. Wang et al. / Journal of Me

he MBRs [9], it is hard to measure residence time distribution pro-les from the lead and tail element separately in large-scale MBRs.he CFD modelling of whole MBR plants requires a macro-scalepproximation which can keep the mesh size and computationffort within the reasonable limit. However, the simulation of mix-ng behaviour of membrane filtration vessels or the flow behavioururrounding the membranes requires high mesh resolutions andence large element numbers [12]. Therefore, it is impossible toodel each individual hollow fibre using the Navier–Stokes equa-

ions as it is too computationally costly. Since we are interested inhe flow resistance induced by the hollow fibre array and its impactn mixing profile, the whole membrane module can therefore beodelled as a porous media with macroscopical characteristics.

ffects on the flow field caused by the membrane module can beransferred to the porous zone [12,13]. The porous media model haseen used for modelling the pressure drop across tube banks and

ts impact on heat transfer [14–16]. The tube banks have geometryimilarity as the hollow fibre array. Therefore, there was attempt13] to apply the empirical correlations of tube banks to model theow resistance induced by hollow fibre. The challenge is that theow resistance induced by different membrane modules is differ-nt, leading to different inertial loss coefficients required in theorous media model. There is no information available in the lit-rature on how to characterise the flow resistance caused by theollow fibre bundle.

This paper demonstrates an experimental approach to measurehe flow resistance of the hollow fibre membrane bundles. The cor-elations obtained from measurements were applied to the porousedia model which was used to simulate the flow behaviour of theembrane filtration zone of a 2.2 MLD MBR located in Sydney. The

orous media model was coupled with the CFD model developed byrannock et al. [10] to show the improvement on the hydrodynamicescription by using porous media model.

The packing density of the membrane bundle used in the exper-ments was as same as the hollow fibre membrane module usedor Sydney Water’s North Head MBR plant [9]. The porous media

odel was then applied to the hollow fibre membrane module ofhis full scale submerged MBR using the lab calibration data.

. Theory

.1. Porous media model

The porous media model is widely used for determining theressure loss in the flows through packed beds, filter papers, perfo-ated plates, flow distributors, and tube banks [16]. A momentumource term is added to the governing momentum equations whichreates a pressure drop that is proportional to the fluid velocity:

i = −

⎛⎜⎜⎜⎜⎝

3∑j=1

Dij�vj

︸︷︷︸viscous loss term

�vmagvj

︸︷︷︸inertial loss term

⎞⎟⎟⎟⎟⎠ (1)

here Si is the source term for the ith (x, y, or z) momentum equa-ion, and D and C are prescribed matrices.

In laminar flows through porous media, the pressure drop isypically proportional to velocity. Ignoring convective accelerationnd diffusion, the porous media model then reduces to Darcy’s law:

p = −�

˛−→v (2)

here ˛ is the permeability.At high flow velocities, the inertial resistance factor, C2ij, can be

iewed as a loss per unit length along the flow direction, thereby

Fig. 1. Tube layout and dimensions; (A) staggered arrangement; (B) in-line arrange-ment.

allowing the pressure drop to be specified as a function of dynamichead.

The aim of current work is to model the effects of pressure dropcaused by membrane bundles, which is similar to the situation ofmodelling a perforated plate or tube bank, in which cases the vis-cous term can be eliminated, yielding the simplified form of theporous media equation [16]:

∇p = −3∑

�vjvmag

where C2 is the inertial resistance factor.Eq. (3) can be written in terms of the x, y, z directions:

�px =3∑

C2xj�nx12

�vjvmag

�py =3∑

C2yj�ny12

�vjvmag

�pz =3∑

C2zj�nz12

�vjvmag

where �nx, �ny, and �nz is the actual thickness of the porousregion.

2.2. Empirical correlations for tube banks

The challenge of using porous media model is that the flowresistance/pressure drop induced by different membrane moduleis different and therefore the inertial resistance factor C2 is hardto determine. One way to determine the pressure drop across thehollow fibre membrane bundle is using the empirical correlationswhich are used to estimate pressure drop across tube banks as theconfiguration of tube banks is similar to that of hollow fibres.

Pressure drop for tube banks can be presented in the form of afriction factor vs. Reynolds number and affected by the arbitrarycharacteristic of the tube bank such as tube spacing and configu-ration (i.e. staggered arrangement or in-line arrangement, Fig. 1)[17,18].

For the turbulent flow with the flow direction perpendicular totube banks:

�umax2

�p = 4fNt 2(5)

where Nt is the number of tube rows in the direction of fluid travel,f is friction factor. The configuration of the hollow fibre membranebundle has the similarity with the configuration of tube banks with

Y. Wang et al. / Journal of Membrane Science 363 (2010) 57–66 59

cale s

density as that used in the full-scale MBR being investigated.The two ends of fibres are fixed on two flyscreens (Fig. 3) and

therefore the packing density of the fibres can be controlled. There-fore, although the correlations obtained from this experiment are

Fig. 2. Schematic diagram of the bench s

taggered arrangement. The friction factor for staggered arrange-ent can be calculated by:

= 0.75(

Dcumax�

)−0.2(6)

here Dc is the clearance between tube banks and umax is the veloc-ty through Dc. Eq. (6) is recommended for the common commercialpacings of tube banks, which is in the range of 1.25–1.50. Theccuracy in this range is ±25%.

For the hollow fibre membrane bundle, the lumen spacing (a inig. 1) can be determined by Eq. (7) [19]:

√1 − 0.046

here DHF is the diameter of a hollow fibre membrane module, n ishe number of fibres per module. For the Siemens Memcor Memjet®

10R HF module, which is studied in this work, DHF = 0.1114 m and= 1692 [19], the lumen spacing a = 0.0026 m and therefore thelearance between fibres Dc = 0.0013 m.

For the flow parallel to the tube banks, the pressure drop weresually estimated using the pressure drop for pipe flow [20,21]:

p = fL

here the friction factor f for the turbulent flow is a function ofeynolds number:

f = 0.079 Re−0.25 (9)

The hydraulic diameter DH for the hollow fibre bundle can bestimated by Eq. (10):

H = 4Across-section

Lwetted(10)

here Across-section is the cross sectional area of the hollow fibreembrane module, which can be estimated by:

cross-section = �DHF2 − n �d2 (11)

nd Lwetted is the wetted perimeter, which can be determined by:

wetted = �DHF + n �d (12)

In this work the inertial resistance factor of hollow fibreembrane bundle was determined experimentally and compared

gainst the values estimated by these empirical correlations.

etup for the flow resistance experiment.

3. Methodology

3.1. Experimental calibration of inertial loss factor

A bench scale rig (Fig. 2) was built for the calibration of iner-tial resistance factors. It contains a transfer pump (DAVEY, XF221)which flowrate is up to 225 L/min, a by-pass loop, control valves,a manometer, a surge tank, a cooling water system, a thermome-ter and a hollow fibre membrane bundle with the same packing

Fig. 3. A photo of hollow fibre membrane bundle used for the calibration of inertialloss factor.

60 Y. Wang et al. / Journal of Membrane Science 363 (2010) 57–66

olblnudtwfotmutc

Table 1Summary of flow resistance test conditions.

Flow direction Test solution Velocity range(m/s)

Flow parallel tomembranes

Water 0–0.340

Xanthan gum solution 0.5 g/L 0–0.319Xanthan gum solution 1 g/L 0–0.278

Flow perpendicularto membranes

Water 0–0.24

Xanthan gum solution 0.5 g/L 0–0.264Xanthan gum solution 1 g/L 0–0.317

Table 2Summary of FLUENT® modelling settings.

Dimension of the fluid domain Three-dimensionalPhase Two-phase: water and airTemperature 25 ◦CTurbulent model Standard k − εMultiphase model Eulerian–EulerianAir bubble size 5 mmOperating conditions Pressure: 101,352 Pa (default)

Gravity: −9.81 m/s2 (in z direction)Discretisation scheme First order upwind differencing

ig. 4. Process diagram of the MBR: (1) anoxic zone; (2) aerobic zones; (3) mem-rane filtration zones; (4) de-aeration zone.

nly valid for the MBR plants using Memcor Memjet® B10R hol-ow fibre membrane module, the experimental methodology cane easily applied for the measurement of friction loss of any hol-

ow fibre modules by changing the membranes and adjusting theumber of membranes according to the packing density of the mod-le. The flyscreens were fixed on a metal frame which has the sameimensions of the acrylic tank. The internal dimension of the acrylicank is 1560 mm × 56 mm × 200 mm (L × W × H). The metal frameas positioned at as far as possible from the inlet in order to achieve

ully developed turbulent flow. The pressure drop due to the wallf the acrylic tank was estimated to be less than 0.1% of that due to

etal frame through connectors. The flowrates were determinedsing a measuring cylinder and a stopwatch. The temperature ofhe fluid in the surge tank was maintained at 27 ± 2 ◦C through theooling water system.

ig. 5. Computational geometry of HF MBR; (A) 3D: (1) anoxic zone; (2) aeration zone; (c) overflow weir; (d) aerator; (e) underflow weir; (f) mixed liquor recycle outlet; (g) recy

Convergence 1 × 10−3

Pressure-correction method Phased-coupled SIMPLE scheme

The resistance of the flow perpendicular and parallel to the hol-low fibres were measured separately to obtain the data for thethree spatial directions. The pressure drop at various flowratesranging from 12–200 L/min was measured. This range was deter-mined based on the CFD simulation results from previous work[10] which showed that 90% of the cell has liquid velocity in thisrange. Moreover, it was difficult to achieve higher flowrates dueto the limitation of the pump and high volume of the water. Threedifferent testing fluids were used, which were tap water, 0.5 g/Lxanthan gum solution and 1 g/L xanthan gum solution, to repre-sent different mixed liquor (MLSS) concentration. Xanthan gum

3) membrane filtration vessels; (4) de-aeration zone and (a) main inlet; (b) mixer;cle inlets for RAS; (B) plan-view of the computational geometry.

CMpcet

tEbulhm

the porous media model input would lead to inaccurate modellingresults.

For the perpendicular flow, the difference was not as significantas that of the parallel flow (Table 4). Only minor difference was

ig. 6. Computational geometry of HF MBR with membranes modelled as clovers.

.2. CFD modelling methods

A three-dimensional model was developed using commercialFD code FLUENT® to study the hydrodynamic behaviour of a 2.2LD submerged hollow fibre (HF) MBR located in Sydney. The

lant possesses an anoxic zone, aerobic zone and an internal recy-le (Fig. 4). The mixed liquor from the membrane filtration vesselsnters into the de-aeration zone ((4) in Fig. 4) before returning backo the anoxic zone.

The membrane zone of this MBR ((3) in Fig. 4) is split into tworains providing redundancy and allowing maintenance activities.ach zone consists of 4 racks of 40 Memcor Memjet® B10R mem-rane modules. The membrane module block is positioned at thepper and right part of the membrane filtration vessels rather than

ocated in the centre. The modules are grouped in clovers of four,ence 10 clovers per rack. Therefore, there are 160 HF membraneodules per train and hence 320 modules in total.The geometry was built using Gambit® v2.4.6 using the true

imensions of the plant (Fig. 5). A Eulerian–Eulerian approachas implemented to describe the liquid and gas components of

he multiphase flow. Detailed turbulent and multiphase modellingethods can be found in a previous published work [10]. A sum-ary of modelling settings is shown in Table 2.The membrane module was modelled as a whole block or 40

lovers in each train (Fig. 6). The porous zone was enabled on eitherhe whole block or each clover and therefore the thickness (�nn Eq. (4)) was varied. The experimentally calibrated correlationsbtained at different conditions (e.g. different testing solutions)ere applied through a user-defined-function (UDF).

. Results and discussion

.1. Calibration of inertial loss factor

Six correlations (3 pairs) representing the pressure drop acrossollow fibre bundle at flow direction either parallel or perpendic-lar to the fibre bundle and various fluid viscosity were obtainedrom flow resistance experiments (Table 3 and Fig. 7) with goodeproducibility (Fig. 7).

It can be seen that the flow resistance increases with (1) the

ore pronounced with the increasing of fluid velocity.

Fig. 7. Pressure drop vs. velocity at different flow direction and viscosity.

The pressure drop values estimated from the experimentalcalibration were compared against those from the empirical cor-relations used for tube banks (Eqs. (5)–(12)). It can be seen that thecorrelations for tube banks more or less underestimate the inertialloss caused by the fibres. Dramatic difference was observed for theflow direction parallel with fibres (Table 4 and Fig. 8). The pres-sure drop per unit of length calculated from experimental data wasaround 260 times of that estimated from the correlations for tubebanks, with a maximum difference up to 350 times. This large dif-ference is due to the huge amount of fibres compared to the tubesand fibre movement which was not taken into account by the corre-lations used to estimate the flow resistance of tube banks (smoothpipes). The friction loss caused by the fibres in the direction of flowparallel with fibres cannot be neglected for the current study as themixed liquor is mixed with air in the two-phase jet (Memjet®) andentering into the membrane filtration zone from bottom. There-fore, the main flow direction is from bottom to up (in the positive zdirection in the computational domain) and is parallel with hollowfibres. Therefore, using the empirical correlations for tube banks as

Fig. 8. Comparison of the logarithmic scale of pressure drop obtained from experi-mental calibration and empirical correlations for tube banks.

Table 3Pressure drop vs. velocity at different conditions.

Flow direction Testing fluid Correlation (variables in SI units) R2

Parallel to fibreWater �p

L = (121.16 + 6.74u−1) 12 �u2 0.9853

Xantan gum 0.5 g/L �pL = (200.54 + 16.71u−1) 1

2 �u2 0.9967

Xanthan gum 1 g/L �pL = (150.77 + 49.94u−1) 1

2 �u2 0.9885

Perpendicular to fibreWater �p

L = (374.95 + 40.75u−1) 12 �u2 0.9981

Xantan gum 0.5 g/L �pL = (410.4 + 17.3u−1) 1

2 �u2 0.9901

Xanthan gum 1 g/L �pL = (220.21 + 116.31u−1) 1

2 �u2 0.9853

Table 4Comparison of pressure drop estimated by the correlations for tube banks and friction loss experimental calibration at various liquid velocities.

Liquid velocity (m/s) Parallel flow, �P/L estimated by Perpendicular flow, �P/L estimated by

Correlations for tube banks Experimental calibration Correlations for tube banks Experimental calibration

0.03 0.44 155.34 348.56 778.570.06 1.48 419.53 1213.76 1893.990.09 3.01 792.57 2518.23 3346.260.12 4.97 1274.45 4226.54 5135.380.15 7.35 1865.19 6315.72 7261.340.18 10.11 2564.77 8768.99 9724.15

12584.03 13531.8915839.85 16780.6820690.71 21636.3824701.82 25671.15

ogusrodnr

tattpwtwflwfi

0.22 14.36 3666.860.25 17.96 4620.420.29 23.29 6061.150.32 27.67 7268.68

bserved at high liquid velocities (Fig. 8). This is due to the lack ineometry similarity between the fibres and tube bank array. These of Eqs. (5)–(7) is recommended for the common commercialpacing of tube banks, which is in the range of 1.25–1.50. The accu-acy in this range is ±25%. The spacing between hollow fibres isbviously much smaller than this range and hence leading to theifference. Moreover, the large amount of fibres compared to theumber of tubes and fibre movement would also lead to differentesistance values.

It is important to note that the calibration of the inertial fac-or in the porous zone was conducted in single phase flow in thebsence of aeration and at zero membrane flux. Under these condi-ions the viscosity and density of the fluid will at a maximum so aso not underestimate the inertial loss. The effects of aeration in theorous zone have been accounted for in the full CFD model whichas developed for two-phase flow. These experimental correla-

ions were applied to the liquid phase and the effects of bubbles

Fig. 10. Comparison of liquid velocity magnitude of membrane filtration zone at the plane 1 m below surface; (A) porous zone disabled; (B) porous zone enabled.

ne 1 m

dwevibebtw

Fig. 11. Comparison of gas hold-up of membrane filtration zone at the pla

le of hollow fibres during the filtration process (i.e. flux > 0 L/m2/h)ill differ from conditions without filtration (flux = 0 L/m2/h). How-

ver, we anticipate that the differences would be negligible as theelocity of filtrate flow is approximately 2–5% of the flow veloc-ty surrounding the membranes. Notwithstanding this, it would

able 5as hold-up of the membrane filtration zone at various cross-sections.

z = 3.392 (0.392 m below surface)

z = 2.782 (1 m below surface)

z = 1.892

z = 1.65 (0.179 m above the bottom of membrane zone)

below surface; (A) porous zone disabled; (B) with porous zone enabled.

4.2. CFD modelling results

4.2.1. Effects of porous zoneA liquid velocity contour at the plane of 1 m below the sur-

face of the simulated MBR is shown in Fig. 9. The employment

Table 6Comparison of liquid velocity within the membrane module at different MLSS concentrations.

Testing fluid Volume-average of liquid velocitywithin the porous zone

Liquid velocity magnitude at the plane of 1 m below surface

Water 0.0731

Xanthan gum solution, 0.5 g/L 0.0485

Xanthan gum solution, 1 g/L 0.0264

ulbfieitwa(

sing the porous media conditions. It can be seen that the maximumiquid velocities occurred in the upper and right corner of the mem-rane filtration vessels, corresponding to the position of hollowbre membrane modules and the two-phase jets (Fig. 10(A)). How-ver, by using the porous media conditions, the liquid velocitiesn this region were significantly reduced due to the flow resis-

rea of membrane aeration were found to be significantly increasedFig. 11).

able 7omparison of liquid velocity magnitude obtained from various geometries of the memb

Geometry of the membranes Volume-average of liquid velocitywithin the porous zone

Whole block 0.0485

4 × 10 clovers 0.0534

The inclusion of a porous zone appears to reconcile the mod-elled hydrodynamic descriptions of the membrane filtration vesselwith expectations for fluid flow based on visual observations of fullscale MBRs. For example, calculated liquid velocities were higher inthe region occupied by the fibre bundles for simulations performedwithout a porous zone (Fig. 10(A)) compared with simulations

rane bundles using porous media model.

Liquid velocity magnitude at the plane of 1 m below surface

Table 8Comparison of gas hold-up obtained from various geometries of the membrane bundles using porous media model.

Geometry of the membranes Liquid velocity magnitude at the plane of 1 m below surface

Whole block

4 × 10 clovers

ettottHmiaegTu

ewottt

cbctetwtt

ffectiveness of the two-phase jet flow that distributes more air inhe region occupied by the fibre module (Fig. 11). For example, ashe flow direction is towards de-aeration zone and the resistancef the flow is not presented, the air bubbles were concentrating onhe right corner of the reactor (Fig. 11(A)) rather than evenly dis-ributed in the zone where membrane module located (Fig. 11(B)).owever, the importance of including the porous zone in the CFDodel can only be quantified by direct measurement of fluid veloc-

ty surrounding the membrane module and between the racks onfull scale plant in two-phase flow. This could be done using an

lectromagnetic flow meter and invoking Faraday’s law to distin-uish the liquid velocities from the bubble rising velocities [23].he experimental procedure and theoretical treatment of the datasing this method is not trivial and is the subject of ongoing work.

The effects of membrane aeration can be more clearly demon-trated by comparing the gas hold-up at different cross-sectionsTable 5). As membrane aerators are located at the bottom of theessel, the air was becoming more dispersed in the positive z direc-ion.

.2.2. Effects of MLSS concentrationThe friction loss correlations obtained from flow resistance

xperiments using water, 0.5 g/L and 1 g/L xanthan gum solution,ere applied to the porous zone respectively to assess the effects

f MLSS concentration on the hydrodynamic results. As expected,he liquid velocities within the membrane module decreased withhe increasing of mixed liquor concentration (i.e. the viscosity ofhe fluid) (Table 6).

.2.3. Effects of geometryThe HF membrane zone was modelled as a whole block and 40

lovers, respectively. The results suggested by modelling the mem-rane zone as clovers more accurate hydrodynamic descriptionsan be obtained (Tables 7 and 8). Different from modelling the fil-ration zone as a whole block, the use of inertial loss correlation on

ributing in the whole filtration area which should be more closeo reality (the jet effect of Memjet®).

5. Conclusions

The hydrodynamic conditions of membrane filtration zone ofa full-scale submerged hollow fibre MBR was studied using aporous media model. The inertial resistance factor required bythe porous media model representing the correlations betweenthe pressure drop across hollow fibre membrane bundle at vari-ous fluid velocities was calibrated in a bench scale set-up usinga self-made hollow fibre bundle having the same packing den-sity as the Memcor Memjet® B10R hollow fibre membrane moduleused in the full-scale MBR. The measurement was conducted at theflow direction either perpendicular or parallel to the membranebundle and at different MLSS concentrations. Comparison betweenthe pressure drop values calculated from the experimental calibra-tion and the empirical correlations used for tube banks indicatedthat the latter would underestimate the flow resistance inducedby the fibres, particularly for the flow direction parallel with fibrebundle.

The porous media model was incorporated with the three-dimensional and two-phase CFD model developed in previousresearch [10] to account for the flow resistance induced by thehollow fibre membrane bundle. It was found that the hydrody-namic descriptions obtained from the model coupled with theporous zone were more consistent with the observations made onfull-scale plants that the fibres present a resistance to flow in thefiltration zone. The inertial loss caused by the hollow fibre mod-ule and the effects of sludge viscosity on the flow regimes can beclearly presented. It also enables the more effective assessment ofthe two-phase jet effects. Therefore, more accurate results wouldbe achieved on the subsequent investigations on the mixing ofthe membrane filtration zone and the aeration energy required formembrane scouring. In addition, the CFD models could be furtherimproved by applying the porous media model on the geometrycloser to the real setup, i.e. by applying the experimental correla-tions of inertial loss on each clover rather than on the whole block.The validation of the porous media model approach requires direct

6 mbran

aIAAu0

[Vossenkaul, Permeate flux decline in cross-flow microfiltration at constantpressure, in: 12th Aachener Membran Kolloquium (AMK), Aachen, Germany,2008.

[23] P.R. Marshall, Computational fluid dynamics modelling of an aerated wastew-

6 Y. Wang et al. / Journal of Me

etical treatment of the data using this method is not trivial and wille difficult to implement on a full scale plant, which is the subjectf ongoing work.

cknowledgements

The authors acknowledge the support of Memcor Australiand Sydney Water. This project is proudly supported by thenternational Science Linkages programme established under theustralian Government’s innovation statement, Backing Australia’sbility. This work was also supported by the European Commissionnder the 6th Framework Programme (AMEDEUS project, contract18328).

Nomenclature

Across-section cross-sectional area of the hollow fibre mem-brane module

a fibre (lumen) spacing or tube spacingC2ij inertial resistance factorDij viscous resistance factorDc clearance between tube banks or fibresDH hydraulic diameterDHF diameter of hollow fibre membrane moduled outside diameter of fibre or tubef friction factorL lengthLwetted wetted perimeter of the cross sectionNt number of tube rows in the direction of fluid traveln number of fibres�n actual thickness of the porous regionp pressure�p pressure dropRe Reynolds numberSi source term for the ith momentum equationu liquid velocityumax velocity through clearance between the tube banksvj liquid velocity in the jth direction

Greek symbols˛ permeability� dynamic viscosity� density of fluid

eferences

[1] J. Manem, R. Sanderson, Membrane bioreactors, in: J. Malievialle, P.E. Odendaal,M.R. Wiesner (Eds.), Water Treatment Membrane Processes, McGraw-Hill, NewYork, 1996 (Chapter 17).

e Science 363 (2010) 57–66

[2] Metcalf, Eddy, Wastewater Engineering, Treatment, Disposal, and Reuse,McGraw-Hill, New York, 1991.

[3] COST, The COST ‘Simulation Benchmark’: Description and Simulator Manual,European Co-operation in the Field of Scientific and Technical Research, 2002.

[4] W. Gujer, M. Henze, T. Mino, M. van Loosdrecht, Activated sludge model No. 3,Water Science and Technology 39 (1999) 183.

[5] M. Henze, C.P.L. Grady, W. Gujer Jr., G.v.R. Marais, T. Matsuo, Activated sludgemodel No. 1, IAWPRC scientific and technical report No. 1, IAWPRC, London,1987.

[6] M. Henze, W. Gujer, T. Mino, T. Matsuo, M.C. Wentzel, G. Marais, Activatedsludge model No. 2, IAWQ scientific and technical report No. 3, IAWQ, London,1995.

[7] M. Henze, W. Gujer, T. Mino, T. Matsuo, M.C. Wentzel, G.v.R. Marais, M. van Loos-drecht, Activated sludge model No. 2d, ASM2d, Water Science and Technology39 (1999) 165.

[8] Y. Wang, Sanly, M. Brannock, G.L. Leslie, Diagnosis of membrane bioreactor per-formance through residence time distribution measurements—a preliminarystudy, Desalination 236 (2009) 120.

[9] M. Brannock, Y. Wang, G. Leslie, Evaluation of full-scale membrane bioreac-tor mixing performance & the effect of membrane configuration, Journal ofMembrane Science 350 (2010) 101.

10] M. Brannock, Y. Wang, G. Leslie, Mixing characterisation of full-scale membranebioreactors: CFD modelling with experimental validation, Water Research 44(2010) 3181.

11] Y. Wang, M. Brannock, G.L. Leslie, CFD simulations of mixing andnutrient removal in full-scale membrane bioreactors with experimentalvalidation, Conference proceedings WEFTEC 2009, Orlando, Florida, USA,2009.

12] J. Saalbach, M. Hunze, Flow structures in MBR-tanks, in: 2nd IWANational Young Water Professionals Conference, IWA Berlin, Germany,2007.

13] C. Kang, J. Hua, J. Lou, W. Liu, E. Jordan, Bridging the gap between membranebio-reactor (MBR) pilot and plant studies, Journal of Membrane Science 325(2008) 861.

14] M. Jamialahmadi, H. Muller-Steinhagen, M.R. Izadpanah, Pressure drop,gas hold-up and heat transfer during single and two-phase flow throughporous media, International Journal of Heat and Fluid Flow 26 (2005)156.

15] A. Narasimhan, K. Sumithra Raju, Effect of variable permeability porousmedium inter-connectors on the thermo-hydraulics of heat exchanger mod-elled as porous media, International Journal of Heat and Fluid Flow 50 (2007)4052.

16] FLUENT, Fluent 6 User’s Guide, FLUENT, Lebanon, USA, 2005.17] D.F. Boucher, Pressure drop across tube banks, flow characteristics of tube

banks, both staggered and in-line, can be calculated by modifying pipe flowapproach, Chemical Engineering (May) (1949) 118.

18] D.F. Boucher, C.E. Lapple, Pressure drop across tube banks, critical comparisonof available data and of proposed methods of correlation, Chemical EngineeringProcess 44 (1948) 117.

19] S. Judd, The MBR Book: Principles and Applications of Membrane Bioreactors inWater and Wastewater Treatment, Elsevier, Amsterdam/Boston/London, 2006.

20] R.H. Perry, D.W. Green, J.O. Maloney, Perry’s Chemical Engineers’ Handbook,McGraw-Hill, New York, 2008.

21] H. Hausen, The pressure drop accompanying fluid flow through tubes and ducts,in: Heat Transfer in Counterflow, Parallel-flow, and Cross-flow, McGraw-Hill,New York, 1983.

22] S. Buetehorn, F. Carstensen, T. Wintgens, T. Melin, D. Volmering, K.

ater pond—development and validation, PhD Thesis, Dept. of ChemicalEngineering, University of Queensland, 1999.

CFD simulations of membrane filtration zone in a submerged hollow fibre membrane bioreactor using a...

Documents

[2012] Development of hybrid membrane bioreactor for tannery effluent treatment

Troubleshooting Foaming in Membrane Bioreactor

Membrane bioreactor technology for wastewater treatment and reuse

Submerged Arc Welding Consumables - PDF4PRO

Development of a Submerged Anaerobic Membrane Bioreactor for Concurrent Extraction of Volatile Fatty Acids and Biohydrogen Production

Performances of submerged membrane photocatalysis reactor during treatment of humic substances

Contribution of microfiltration on phosphorus removal in the sequencing anoxic/anaerobic membrane bioreactor

The forward osmosis membrane bioreactor A low fouling alternative to MBR processes

Correlation between dissolved oxygen concentration, microbial community and membrane permeability in a membrane bioreactor

Submerged Membrane Bioreactor at Substrate-Limited Conditions: Activity and Biomass Characteristics

Comment on \"Comparison of the filtration characteristics between attached and suspended microorganisms in submerged membrane bioreactor

Effects of coagulation on submerged ultrafiltration membrane

Fate of organic pollutants in a pilot-scale membrane bioreactor-nanofiltration membrane system at high water yield in antibiotic wastewater treatment

Semi-permeable membrane retention of synovial fluid lubricants hyaluronan and proteoglycan 4 for a biomimetic bioreactor

A PERFLUOROCARBON-BASED OXYGEN DELIVERY SYSTEM TO A MEMBRANE BIOREACTOR BY SETENO KARABO OBED NTWAMPE FACULTY OF ENGINEERING

Fetuin-A gene expression, synthesis and release in primary human hepatocytes cultured in a galactosylated membrane bioreactor

Immediately loaded non-submerged versus delayed loaded submerged dental implants: A meta-analysis

Zero net growth in a membrane bioreactor with complete sludge retention

Human hepatocyte functions in a crossed hollow fiber membrane bioreactor

Analysis of the microbial community structure in a membrane bioreactor during initial stages of filtration