CFD-PBM Coupled Simulation of an Airlift Reactor with Non

IFP Energies nouvelles International ConferenceRencontres Scientifiques d'IFP Energies nouvelles

Dynamics of Evolving Fluid Interfaces – DEFI Gathering Physico-Chemical and Flow PropertiesDynamiques des écoulements à interfaces fluides – au croisement de la physico-chimie et de la mécanique de fluides

CFD-PBM Coupled Simulation of an Airlift Reactor

with Non-Newtonian Fluid

Mei Han1,2

*, Zuoliang Sha2, Arto Laari

1and Tuomas Koiranen

1

1 School of Engineering Science, Lappeenranta University of Technology, P.O. Box 20, Lappeenranta 53851 - Finland2 Tianjin Key Laboratory of Marine Resources and Chemistry, Tianjin University of Science and Technology, Tianjin 300457 - PR China

e-mail: [email protected]

* Corresponding author

Abstract — Hydrodynamics of an AirLift Reactor (ALR) with tap water and non-Newtonian fluid wasstudied experimentally and by numerical simulations. The Population Balance Model (PBM) withmultiple breakup and coalescence mechanisms was used to describe bubble size characteristics inthe ALR. The interphase forces for closing the two-fluid model were formulated by considering theeffect of Bubble Size Distribution (BSD). The BSD in the ALR obtained from the coupledComputational Fluid Dynamics (CFD)-PBM model was validated against results from digitalimaging measurements. The simulated velocity fields of both the gas and liquid phases werecompared to measured fields obtained with Particle Image Velocimetry (PIV). The simulated resultsshow different velocity field profile features at the top of the ALR between tap water and non-Newtonian fluid, which are in agreement with experiments. In addition, good agreement betweensimulations and experiments was obtained in terms of overall gas holdup and bubble Sauter meandiameter.

NOMENCLATURESymbols

ag/l Volume fraction of gas/liquid phase, 1Cl, Cɛ1, Cɛ2 k-ɛ Model parametersClb Sato model parameterCVM Added mass coefficientCD Drag coefficientCL Lift force coefficient

CTD Turbulent dispersion force coefficientCW Wall lubrication force coefficientds Bubble Sauter mean diameter, mdb Bubble equivalent diameter, mfi Volume fraction of the bubble phase with size

class i based on the gas phasefv Volume fraction of bubbles with volume v

FD Drag force, Nm�3

FL Lift force, Nm�3

FTD Turbulent dispersion force, Nm�3

FW Wall lubrication force, Nm�3

kl Liquid shear-induced turbulence kinetic energy,m2 s�2

kl,g Turbulence kinetic energy due to bubble-induced turbulence, m2 s�2

j Consistency index, Pa sn

n Bubble numbern Flow indexUg Gas superficial velocity, m s�1

Eo Eotvos numberRe Reynolds numberb(d) Bubble breakup rate due to turbulent eddies, s�1

b(fv|d) Daughter bubble size distribution

b2(d) Bubble breakup rate due to instability, s�1

-T (di, dj) Collision rate due to turbulent eddies, s�1

PT (di, dj) Coalescence efficiency due to turbulent eddies

Oil & Gas Science and Technology – Rev. IFP Energies nouvelles (2017) 72, 26� M. Han et al., published by IFP Energies nouvelles, 2017DOI: 10.2516/ogst/2017017

This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0),which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

http://ogst.ifpenergiesnouvelles.fr/

http://ifpenergiesnouvelles.fr/

https://doi.org/10.2516/ogst/2017017

-W (di, dj) Collision rate due to bubble wake, s�1

PW (di, dj) Coalescence efficiency due to bubble wake-U (di, dj) Collision rate due to different rise velocities, s�1

PU (di, dj) Coalescence efficiency due to different risevelocities

Greek Symbols

lapp Liquid apparent viscosity, kg m�1 s�1

llam Laminar viscosity, kg m�1 s�1

ltl Dynamic viscosity of liquid due to shear-induced turbulence, kg m�1 s�1

ltb Dynamic viscosity of liquid due to bubble-induced turbulence, kg m�1 s�1

ltg Turbulence viscosity of the gas phase,kg m�1 s�1

leff Effective viscosity, kg m�1 s�1

c Shear rate, s�1

q Liquid/gas density, kg m�3

ɛg,o Overall gas holdup, %ɛl Turbulent dissipation rate due to shear-induced

turbulence, m2 s�3

ɛl,g Turbulent dissipation rate due to bubble-induced turbulence, m2 s�3

Subscripts

m Gas-liquid mixturel,L Liquid phase

g Gas phasei,j Classes of bubble size

INTRODUCTION

Airlift Reactors (ALR) attract more and more interests inmany process applications, such as syngas fermentation [1],direct coal liquefaction [2] and waste water/gas treatment[3]. The advantages of ALR over bubble columns includeenhanced mixing efficiency and good suspension ofparticles due to improved liquid circulation. In comparisonto mechanically stirred tanks, simple structure and lowenergy consumption make ALR even more attractive,particularly considering the current stringent energy andsafety demands.

Airlift loop reactors are an important type of pneumati-cally agitated multiphase reactors where the bubble phasefunctions as the agitation source and reactant. Bubble sizecharacteristics, determined by the initial bubble size, bubblebreakup and coalescence, is an important hydrodynamicparameter and closely related to mass transfer performance

and even production yield in the reactor. Bubble sizeinfluences gas holdup, gas residence time, specific interfacialarea, and mass and momentum transfer between the gas andliquid phase. Meanwhile, bubble size and its behavior areaffected by several factors, such as gas holdup, liquidvelocity field and liquid physical property. Extensive studies[4-8] have shown that bubble size characteristics havesignificant effects on the hydrodynamics and mass transferperformance of ALR, especially for liquids with highviscosity. For example, Hwang and Cheng [5] found thatgas holdup decreased with increasing Carboxyl MethylCellulose (CMC) concentration due to formation of largebubbles in highly viscous liquid. Deng et al. [6] showed thatBubble Size Distribution (BSD) for highly viscous liquidsbecame broader in heterogeneous flow regime and increaseof gas holdup slowed down with aeration rate due toformation of large bubbles. Gumery et al. [8] reported thatliquid circulation velocity decreased in the ALR by increas-ing liquid apparent viscosity due to enhanced bubblecoalescence.

Computational Fluid Dynamics (CFD) simulation is apowerful tool to investigate and predict hydrodynamics inmultiphase reactors. However, it is required that the CFDmodels describe the phenomena behind the physics accu-rately in order to predict results with good confidence. Thisis still difficult and challenging since fluid mechanics inmultiphase flow is quite complex and far from fully under-stood. Therefore, it is necessary to validate the CFD simula-tions of ALR experimentally, particularly with differentliquid physical properties. The two-fluid model with theassumption of a single bubble size can give reasonablepredictions for multiphase reactor only in cases where thebubble size is narrowly distributed. Coupling of PopulationBalance Model (PBM) into CFD models could make it pos-sible to extend the simulation and prediction of flow regimesfrom homogeneous to heterogeneous where the bubble sizeis widely distributed. PBM is an effective method to calcu-late BSD for gas-liquid flows on the condition that bubblecoalescence and breakup models are provided as model clo-sures. Many researchers have used the coupled CFD-PBMmodel to predict the hydrodynamics and mass transfer inmultiphase reactors [9-14]. Their studies showed that abetter agreement between simulations and experiments wasobtained when BSD was considered compared to using aconstant mean bubble diameter. Chen et al. [10] found thatthe coupled CFD-PBM simulation was able to predict theeffects of aeration rate and liquid surface tension on thegas holdup in bubble columns. They pointed out that simu-lations made by implementing PBM produced much bettercomparison to experiments for gas holdup and liquidvelocity than by using mean bubble diameter. Wang andWang [12] investigated mass transfer in bubble columnwith the coupled CFD-PBM model and showed that the

Page 2 of 12 Oil & Gas Science and Technology – Rev. IFP Energies nouvelles (2017) 72, 26

CFD-PBM coupled model could predict effectively thehydrodynamics and mass transfer. Liu et al. [13] performedCFD-PBM coupled simulations of airlift bioreactor for hairyroot culture and presented good agreement betweensimulations and experiments in terms of bubble numberdensity. Xing et al. [14] investigated the ability of theCFD-PBM coupled model to account for the effect ofviscosity on the gas holdup in a bubble column. The featurethat the increased viscosity resulted to a decrease in thevolume fraction of small bubbles and an increase in thevolume fraction of large bubbles was captured well usingthe developed CFD-PBM model [12, 15]. The coupledCFD-PBM simulations of ALR are relatively rare, espe-cially with the viscous or non-Newtonian fluid. In industrialapplications, many processes [16-18] are operated in theviscous mediums or with catalysts showing non-Newtonianbehavior.

The aim of this study was to develop a CFD model forALR with non-Newtonian fluid in order to contribute tothe design and optimization of ALR in practical applica-tions. A PBM with the coalescence mechanisms by turbu-lent eddies, wake entrainment and different rise velocitiesof bubbles is coupled into the two-fluid model. Bubblebreakups due to eddy collision and eddy instability are takeninto account. Gas holdup, bubble size and liquid and gasvelocity fields obtained by Particle Image Velocimetry(PIV) were used to validate the CFD model.

1 EXPERIMENT

1.1 Experimental Setup and Material

The schematic diagram of the experimental setup used in thiswork is shown in Figure 1.

The AR_ALR is composed of a riser (outer column),downcomer (draft tube) and gas distributor. The riser is0.15 m in inner diameter and 1.2 m in height. The down-comer is 0.60 m high with the inner diameter of 0.10 m.The downcomer was co-axially mounted with the riserlocating at 0.05 m above the column bottom. A plasticO-ring gas distributor with a diameter of 0.12 m wasinstalled in the annulus between the riser and the down-comer and it was at 0.07 m above the bottom. Gas is fedto the column through 36 pores, with hole size of0.5 mm distributed equidistantly at the top of the O-ring.The AR_ALR is located inside a transparent rectangularbox in order to diminish optical distortions during visual-ization measurements.

Tap water and 0.14 wt% CMC (Finnfix 50000) aqueoussolution were employed to study the Newtonian and non-Newtonian fluids. The dynamic viscosity of the CMC solu-tion was measured with the Modular Compact Rheometer

302 (Anton Paar) and the curve is shown in Figure 2.As the reference, the viscosity of tap water was also testedand is shown in Figure 2. The CMC solution exhibited shear

1 Riser2 Downcomer (draft tube)3 O-ring gas distributor4 Gas mass flow controller5 Square tank6 Davis 8.3 software7 Laser8 Camera for gas phase9 Camera for liquid phase

1 Riser2 Downcomer (draft tube)3 O-ring gas distributor4 Gas mass flow controller5 Square tank6 Davis 8.3 software7 Laser8 Camera for gas phase9 Camera for liquid phase

1

2

4

5

3

7

68

9

Figure 1

Schematic diagram of the experimental setup.

Figure 2

Dynamic viscosity of water and CMC solution measured byrheometer.

Oil & Gas Science and Technology – Rev. IFP Energies nouvelles (2017) 72, 26 Page 3 of 12

thinning behavior and the apparent viscosity, lapp, could bedescribed by the following power law model:

lapp ¼ jcn�1 ð1Þ

where j and n are the consistency index and the flow indexof the fluid, respectively. j is 0.118 Pa sn and n is 0.68 for theCMC solution obtained by fitting Equation (1) with the least-squares method.

Tap water and CMC solutions were operated in experi-ments in batch mode. The static level for both liquid phaseswas 0.8 m. Oil-free compressed air as the gas phase was runin once-through mode. Gas flow rate was controlled by a

mass flow controller (Bronkhorst E-7000). Gas superficialvelocity,Ug, was calculated based on the cross-sectional areaof the riser (the annulus). All experiments were carried out atatmospheric pressure and room temperature.

1.2 Measurement Techniques

The overall gas holdup in the reactor, ɛg,o, was determinedby the height expansion method:

eg;o ¼ Hm � H lð Þ=Hm ð2Þwhere Hl is the static liquid level and Hm is the gas-liquidmixture height after aeration.

TABLE 1

Governing equations and discrete PBM for the ALR.

Models Equations

Mass conservation r qkakukð Þ ¼ 0; k ¼ l; g

Momentum conservationr qkakukukð Þ ¼ �akrP0 þ r akleff ruk þruTk

� �� þ qkakg

þFD;k þ FL;k þ FW;k þ FT;k

Modified k-ɛ turbulenceequation [20, 21]

rðqlalklulÞ ¼ r alllam;l ltl þ ltbð Þ=rk� �rkl� �þ alðGk;l � qlelÞ

r qlalelulð Þ ¼ r al llam;l þ ðltl þ ltbÞ=re� �rel

� �þ alel=klðCe1Gk;l � Ce2qlelÞ

lt;l ¼ Clðql=k2l el Þ; leff ¼ lapp;l þ lt;l þ ltb; ltb ¼ Clbqlagdbs ug � ul��

kl;t ¼ kl þ kl;g; el;t ¼ el þ el;g; kl;g ¼ 1=2agCVMu2slip; el;g ¼ agguslip

ltg ¼ lt;lqg=ql

Drag force [22]FD;k ¼

PMi¼1fiagql

3CDi4dbi

ug � ul� �

ug � ul��

CDi ¼ max 24Re�1i 1þ 0:15Re0:687i

� �; 8=3Eo=ðEoþ 4Þ� �

Lift force [23]

FL;k ¼ �PMi¼1CLifiagql ug � ul

� � oulor

CLi

min 0:288 tanhð0:121Rei; f Eo0i� ��

; Eo0i < 3:4f ðEo0iÞ; 3:4 < Eo0i < 5:3�0:29; Eo0i > 5:3

8<:

f ðEo0iÞ ¼ 0:00925Eo03i � 0:099Eo02i þ 1:088

Wall lubrication force [23] FW;k ¼ �PMi¼11=2CWifiagdbiql ug � ul

� �2 ðR� rÞ�2 � ðRþ rÞ�2h i

Turbulent dispersion force [24] FTD ¼ �CTDagqlkloaor

Population balance model forbubbles in the ALR [11]

oot ðagfiÞ þ r agubfi

� � ¼ Pj�k

j; kxi�1 � ðxj þ xkÞ � xiþ1

1� 12 dj;k

� �gi;jkcj;kagfiagfkvi=vj=vk

� agfiXMk¼1

ci;jagfk=vk þXMk¼i

ni;kagfkbkvi=vk � biagfi


PIV method was adopted to measure flow velocity fieldsfor both gas and liquid phases simultaneously. The PIV sys-tem (LaVision GmbH) mainly included a double pulsed NdYag laser, two high resolution CCD cameras (1600 pixel 91200 pixel) and DaVis 8.3 software. The liquid velocity fieldwas traced by PMMA-RhB fluorescent particles with thediameter range of 20-50 lm. The measuring range of theflow field was from the wall to the centerline of the ALRin the radial direction. Each measurement window is75 9 50 mm in physical size. Eight hundred (800) pairs ofimages were acquired by each camera within 10 min for eachtest when the flow field was in steady state. DaVis 8.3 soft-ware was utilized to analyze the acquired images for eachphase. The spatial resolution of the measured velocity fieldwas 3 9 3 mm (grid size). Ensemble- and time-averagedvelocity fields were calculated and used to compare to thesimulations.

Digital imaging method was employed to evaluate thebubble size and size distribution in the ALR. One thousand(1000) digital images were acquired and used to analyze thebubble size for each condition. An overlapping object recog-nition algorithm (Pixact Object Recognition and Analysissoftware, Pixact Ltd) developed by Eloranta et al. [19] wasused to recognize and analyze the bubbles from the acquireddigital images. More than 104 bubbles were detected for

each condition and used to calculate the bubble Sauter meandiameter, d32:

d32 ¼XNi¼1

nid3bi

.XNi¼1

nid2bi ð3Þ

where dbi and ni are the bubble equivalent diameter andthe number of bubbles corresponding to the equivalentdiameter, and i indicates the size class that refers to thebubble equivalent diameter.

2 MODELING AND SIMULATION

2.1 Coupled CFD-PBM Model

The governing equations of the time-averaged two-fluidmodel and the discrete PBM model from Wang et al. [11]are listed in Table 1. BSD was resolved from the PBM usingthe gas holdup and the kinetic energy dissipation ratecalculated in CFD. Then BSD obtained from the PBM wasused to formulate the interphase forces and the turbulencemodification closure terms for the two-fluid model. A mod-ified k-ɛ model based on the two-time constant model fromLahey et al. [20] was used to describe the liquid phase

TABLE 2

Models of bubbles breakup and coalescence rate.

Mechanisms Equations

Bubble breakup dueto turbulent eddies

bðdÞ ¼ R 0:50 b fv djð Þdfv; bðfv; dÞ ¼ 2bðfv dÞj =

R 10 bðfv dÞdf vj

bðfv dj Þ ¼ 0:923ð1� agÞne1=3R dbkmin

Pb fv d; kjð Þðkþ dÞ2k�11=3dk

Pb fv d; kjð Þ ¼ R10 Pb fv d; e kð Þ; kjð ÞPe e kð Þð ÞdeðkÞ; Pe e kð Þð Þ ¼ 1=�e kð Þð Þ exp �e kð Þ=�e kð Þð Þ

Pb fv d; eðkÞ; kjð Þ ¼ 1 fv;max � fv;min

� �fv;max � fv;min � d fv;min < fv < fv;max

0 else

�

fv;min ¼ pk3r=ð6eðkÞdÞ� �3; ðf 2=3v;max þ ð1� fv;maxÞ2=3 � 1Þ ¼ min ð21=3 � 1Þ; eðkÞ=ðpd2rÞ� �

; �eðkÞ ¼ p6 k

3ql�u2k2

Bubble breakup rate dueto instability b2ðdÞ ¼ b� d�dc2ð Þm

ðd�dc2Þmþ dmc2; b fv; dð Þ ¼ 2dð0:5Þ

Coalescence rate dueto turbulent eddies

-Tðdi; djÞ ¼ ðp=4Þamaxðamax � aÞ�1Cij

ffiffiffi2

pe1=3ðdi þ djÞ2ðdi2=3 þ dj

2=3Þ1=2

PTðdi; djÞ ¼ exp� 0:75ð1þ n2ijÞð1þ n3ijÞ 1=2

ðqg=ql þ cÞ�1ð1þ nijÞ�3We1=2ij

�Cij ¼ lmbt;ij=ðlmbt;ij þ hmb;ijÞ; lbt;ij ¼ l2bt;i þ l2bt;j

1=2; lbt ¼ 0:89db; hb;ij ¼ Ni þ Nj

� �1=3

Coalescence rate dueto bubble wake

-W di; dj� � ¼ 12:0Hd2i �uslip;i; �uslip;i ¼ 0:71

ffiffiffiffiffiffiffigdi

p

PWðdi; djÞ ¼ expð�0:46q1=2l e1=3r�1=2ðdidj=ðdi þ djÞÞ5=6Þ

H ¼ dj � dc=2� �6

= dj � dc=2� �6 þ dc=2ð Þ6

dj � dc=2

0 else

(

-Uðdi; djÞ ¼ p=4ðdi þ djÞ2 uri � urj�� ; PUðdi; djÞ ¼ 0:5


turbulence, where both the shear-induced turbulence andthe bubble-induced turbulence were considered. The turbu-lent viscosity of the gas phase was formulated using thecorrelation proposed by Jakobsen [21]. The drag, lift, walllubrication and turbulent dispersion forces were taken intoaccount for gas and liquid momentum transfer. The dragcoefficient model from Tomiyama et al. [22] was used tocalculate drag force for each bubble size group. TheTomiyama drag coefficient model, being developed basedon the balance of forces acting on a bubble and availabletheoretical and empirical correlations of terminal risingvelocity, is well suited for gas-liquid flows where thebubbles can be of different size to a certain extent. The liftforce was calculated with the lift force coefficient model ofTomiyama et al. [23], which is applicable to larger-scaledeformable bubbles in the ellipsoidal and spherical capregimes. The Tomiyama wall lubrication force model [24],which modified the formulation of Antal et al. [25], wasemployed for wall lubrication force calculation, as listedin Table 1.

The coupling of PBM into CFD model was implementedin ANSYS Fluent 14.5 using user defined scalars that weredefined for volume fractions of bubble group i in the gasholdup. Bubble breakup and coalescence rate models wereloaded by the user defined function. All bubble groups wereassumed to move at identical velocity that equals to the localensemble averaged gas velocity.

To describe BSD characterization in viscous fluid, themultiple bubble breakup and coalescence models from Xinget al.’s [14] were used to close the PBM. The total bubblebreakup rate was calculated as the sum of the rates from eddycollision and from the instability of large bubbles. For thetotal coalescence rate three mechanisms were included,namely coalescence due to turbulent eddies, bubble wakeentrainment, and different bubble rise velocities. The detailsof the breakup and coalescence models from Xing et al. [14]used are listed in Table 2.

2.2 Numerical Details

Steady state simulations were performed for the ALR.Two liquid phases, which are the CMC solution as thenon-Newtonian fluid and tap water as the reference, weresimulated. The apparent viscosity of the CMC solutionwas described with the power law equation, as measuredin the experiments. The numerical simulations were per-formed at two gas superficial velocities of 1.18 cm s�1 and2.04 cm s�1, which corresponded to the homogeneous andheterogeneous flow regimes, respectively. The bubble sizewas divided into 30 classes with the geometric method(vi+1 = viq), where the smallest bubble volume v1 was1.0 9 10�10 m3 with the factor q of 1.7.

The two-dimensional axis-symmetry assumptionwas usedin this work due to the geometric symmetry of the ALR andlow computational costs. The computational domain isshown in Figure 3a. The computational grid was created withANSYS ICEM and the structural hexahedra gridding atthe bottom and top is shown in Figure 3b and 3c, respectively.The grid density and the corresponding cell size in the

a)

b)

c)

Pressure outlet boundarycondition at top surface

Center axis

Periodic plane

Velocity inlet boundary conditionon surface of gas distributor

Figure 3

Computational domain and grid profile: a) computationaldomain, b) grid profile at the top, c) grid profile at the bottom.

Figure 4

Overall gas holdup from experiments and simulations.


three modeled zones along the ALR height were 23 9 3 mm(below the gas distributor), 21 9 1 mm (around the gasdistributor), and 284 9 2.5 mm (above the gas distributor).In the radial direction, the grid was fined near the wall withthe minimum edge size of 0.5 mm and the maximum edgesize of 2 mm. Grid independent solution was verified bycomparing to the experiments in terms of gas holdup andliquid velocity.

The wall function was used for the liquid phase and thenon-slip condition was used for the gas phase. The veloc-ity inlet boundary was applied to the upper face of the gasdistributor, where the inlet velocity, gas holdup andvolume fractions of bubble groups were specified accord-ing to the experiments. At the top surface of disengage-ment zone, pressure outlet boundary condition was used.The second order upwind scheme was used for themomentum, turbulence kinetic energy and dissipationrate equations and the first-order scheme for the volumefraction equations.

a)

b)

c)

d)

Figure 5

Bubble volume fraction of experiment and simulation.

Figure 6

Bubble Sauter mean diameter from experiments andsimulations.


3 RESULT AND DISCUSSION

3.1 Overall Gas Holdup

The overall gas holdups for tap water and CMC solution attwo different gas superficial velocities Ug, are shown inFigure 4. The overall gas holdups increase with increasing

gas superficial velocity Ug for both in tap water andin CMC solution. The value of the overall gas holdupis lower in CMC solution than in tap water at the sameUg. The simulated results for the overall gas holdupagrees well with the experimental data for both thetap water and CMC solution with the relative error beingless than ±10%.

a) b) c) d)

Figure 7

Velocity fields of AR_ALR with water at Ug = 2.04 cm s�1: a) liquid velocity – PIV, b) gas velocity – PIV, c) liquid velocity – simulation, d) gasvelocity – simulation.


3.2 Bubble Size

The simulation results of the BSD in the ALR with differentliquid phases were obtained from the discrete PBM. Thecomparison between the simulation results and experimen-tally determined bubble number densities in the ALR with

different liquid phases is shown in Figures 5a-5d. It can beseen that the bubble sizes are narrowly distributed both inwater and in CMC solution at low Ug. With increasing Ug,BSD becomes broad in water and even bimodal in CMCsolution. All these features are captured by the PBM withthe bubble breakup and coalescence models including

a) b) c) d)

Figure 8

Velocity fields of AR_ALRwith CMC solution atUg = 2.04 cm s�1: a) liquid velocity – PIV, b) gas velocity – PIV, c) liquid velocity – simulation,d) gas velocity – simulation.


several mechanisms. The relative errors between the exper-iments and the simulations is less than ±10% for most ofthe bubble groups, as shown in Figure 5.

Figure 6 shows a good agreement between the simula-tions and experiments for bubble Sauter mean diameter in

the ALR with different fluids at two investigated Ug. It canbe found that the bubble Sauter mean diameter is slightlyunder-predicted in water and over-predicted in CMC solu-tion. These results are obtained most probably because thebreakup models in the PBM are much more sensitive to

c)

b)

a)

Figure 9

Radial profile of liquid axial velocity of experiments andsimulations at Ug = 1.18 cm s�1.

c)

b)

a)

Figure 10

Radial profile of liquid axial velocity of experiments andsimulations at Ug = 2.04 cm s�1.


the viscosity compared to the coalescence models, as shownby Xing et al. [14]. In addition, the breakup model due toturbulent eddies by Wang et al. [15] is able to give muchbetter prediction for unequal daughter BSD in breakup.

3.3 Velocity Profiles

The velocity fields of both liquid and gas phase in the ALRwere measured with PIV method to compare the velocitiesagainst simulations. Figures 7 and 8 show the comparisonof the velocity fields for water and CMC solution at Ug

2.04 cm s�1. It can be seen that the simulated velocityprofile generally matched well to that measured with PIV.In addition, the magnitudes of the velocities are in agreementbetween the simulations and experiments.

By comparing Figures 7a and 8a, it can be found that themeasured liquid flow structures at the top of the ALR forwater and CMC solution are different from each other. Thesestructures are also captured by the coupled CFD-PBMmodel, as shown in Figures 7c and 8c. This illustrates thatthe CFD-PBM model is able to describe the hydrodynamicsin the ALR with reasonable accuracy for non-Newtonianfluid at the investigated Ug.

Figures 9 and 10 show the radial profile of the axial liquidvelocity at three axial heights in the ALR located at 0.15,0.45, and 0.7 m from the bottom of the ALR. It can be seenthat the radial profile of liquid velocity is more uniform inwater compared to CMC solution, particularly in thedowncomer. At the top of the ALR, the liquid velocityprofile in CMC solution is different from that of water athigh Ug, most probably due to the effect caused by move-ment of large bubbles and bubble wakes.

CONCLUSION

The hydrodynamic characteristics of fluid flow was studiedby experiments and numerical simulations in an ALR fortap water and non-Newtonian fluid. Coupled CFD-PBMsimulations were carried out taking into account multiplebreakup and coalescence models. Simulated overall gasholdup, BSD, Sauter mean bubble diameter, and velocityfields for both phases were compared and validated againstexperimental data. PIV and digital imaging techniqueswere employed in experiments in order to validate thesimulations.

It can be concluded that the coupled CFD-PBM modelcan reasonably well predict the effect of apparent viscosityon the overall gas holdup, bubble Sauter mean diameterand bubble number density in the range of the investigatedUg. The overall gas holdup and bubble Sauter mean diameterare smaller in CMC solution than in water. The BSDbecomes broad in water and even bimodal in CMC solution

at Ug of 2.04 cm s�1. All these features were captured wellby the coupled CFD-PBM model. A difference in the liquidvelocity profile was observed between the CMC solutionand tap water at the top of ALR with the PIV method. Thesame difference was also predicted well by the coupledCFD-PBM simulations.

ACKNOWLEDGMENTS

This work was financially supported by Outotec Oyjand Finnish Funding Agency for Innovation (TEKES)[908/31/2016] and [958/31/2016]. The authors would liketo acknowledge CSC-IT Center for Science, Finland, forcomputational resources.

REFERENCES

1 Munasinghe P.C., Khanal S.K. (2010) Syngas fermentationto biofuel: evaluation of carbon monoxide mass transfercoefficient (kLa) in different reactor configurations, Biotechnol.Prog. 26, 1616-1621.

2 Huang Q.S., Zhang W.P., Yang C. (2015) Modelingtransport phenomena and reactions in a pilot slurry airlift loopreactor for direct coal liquefaction, Chem. Eng. Sci. 135,441-451.

3 Heijnen J.J., Hols J., van der Lans R.G.J.M., van Leeuwen H.L.J.M., Mulder A., Weltevrede R. (1997) A simple hydrodynamicmodel for the liquid circulation velocity in a full-scale two- andthree-phase internal airlift reactor operating in the gasrecirculation regime, Chem. Eng. Sci. 52, 2527-2540.

4 Li G.Q., Yang S.Z., Cai Z.L., Chen J.Y. (1995) Mass transferand gas-liquid circulation in an airlift reactor with viscousnon-Newtonian fluids, Chem. Eng. J. 56, B101-B107.

5 Hwang S.J., Cheng Y.L. (1997) Gas holdup and liquid velocityin three-phase internal-loop airlift reactors, Chem. Eng. Sci. 52,3949-3960.

6 Deng Z.H., Wang T.F., Zhang N., Wang Z.W. (2010) Gasholdup, bubble behavior and mass transfer in a 5 m highinternal-loop airlift reactor with non-Newtonian fluid, Chem.Eng. J. 160, 729-737.

7 Han M., González G., Vauhkonen M., Laari A., Koiranen T.(2017) Local gas distribution and mass transfer characteristicsin an annulus-rising airlift reactor with non-Newtonian fluid,Chem. Eng. J. 308, 929-939.

8 Gumery F., Ein-Mozaffari F., Dahman Y. (2011) Macromixinghydrodynamic study in draft-tube airlift reactors using electricalresistance tomography, Bioprocess Biosyst. Eng. 34, 135-144.

9 Chen P., Sanyal J., Dudukovic M.P. (2005) Numerical simula-tion of bubble columns flows: effect of different breakup andcoalescence closures, Chem. Eng. Sci. 60, 1085-1101.

10 Chen P., Sanyal J., Dudukovic M.P. (2004) CFD modeling ofbubble columns flows: implementation of population balance,Chem. Eng. Sci. 59, 5201-5207.

11 Wang T.F., Wang J.F., Jin Y. (2005) Population balance modelfor gas-liquid flows: influence of bubble coalescence andbreakup models, Ind. Eng. Chem. Res. 44, 7540-7549.


12 Wang T.F., Wang J.F. (2007) Numerical simulations ofgas-liquid mass transfer in bubble columns with a CFD-PBMcoupled model, Chem. Eng. J. 62, 7107-7118.

13 Liu R., Sun W., Liu C.Z. (2011) Computational fluid dynamicsmodeling of mass transfer behavior in a bioreactor for hairy rootculture. I. Model development and experimental validation,Biotechnol. Prog. 27, 1661-1671.

14 Xing C.T., Wang T.F., Wang J.F. (2013) Experimental studyand numerical simulation with a coupled CFD-PBM model ofthe effect of liquid viscosity in a bubble column, Chem. Eng.Sci. 95, 313-322.

15 Wang T.F., Wang J.F., Jin Y. (2003) A novel theoretical breakupkernel function for bubbles/drops in a turbulent flow, Chem.Eng. Sci. 58, 4629-4637.

16 Lennartsson P.R., Niklasson C., Taherzadeh M.J. (2011) A pilotstudy on lignocellulose to ethanol and fish feed using NMMOpretreatment and cultivation with zygomycetes in an air-liftreactor, Bioresour. Tech. 102, 4425-4432.

17 Chavez-Parga M.C., Gonzalez-Ortega O., Negrete-RodriguezM.L.X., Medina-Torres L., Escamilla-Silva E.M. (2007)Hydrodynamics, mass transfer and rheological studies ofgibberellic acid production in an airlift bioreactor, World J.Microb. Biot. 23, 615-623.

18 Chavez-Parga M.C., Gonzalez-Ortega O., Negrete-RodríguezM.L.X., Vallarino I.G., González Alatorre G., Escamilla-SilvaE.M. (2008) Kinetic of the gibberellic acid and bikaverinproduction in an airlift bioreactor, Process Biochem. 43,855-860.

19 Eloranta H., Honkanen M., Marjanen K. (2008) PORA-Objectrecognition and analysis software 1.0 users manual, version1.0, �Pixact Ltd., www.pixact.fi.

20 Lahey R.T., Lopez de Bertodano M., Jones O.C. (1993) Phasedistribution in complex geometry conduits, Nuclear Eng. Des.141, 177-201.

21 Jakobsen H. (1993) On the modelling and simulation of bubblecolumn reactors using a two-fluid model, PhD Thesis,Norweigian Institute of Technology, Norway, 110.

22 Tomiyama A., Kataoka I., Zun I., Sakaguchi T. (1998) Dragcoefficients of single bubbles under normal and micro gravityconditions, JSME Int. J. 41, 472-479.

23 Tomiyama A., Tarnai H., Zun I., Hosokama S. (2002)Transverse migration of single bubbles in simple shear flow,Chem. Eng. Sci. 57, 1849-1858.

24 Tomiyama A. (1998) Struggle with computational bubbledynamics, Multiphase Sci. Tech. 10, 369-405.

25 Antal S.P., Lahey R.T., Flaherty J.E. (1991) Analysis of phasedistribution in fully developed laminar bubble two-phase flow,Int. J. Multiphase Flow 17, 635-652.

Manuscript submitted in November 2016

Manuscript accepted in May 2017

Published online in September 2017

Cite this article as: M. Han, Z.L. Sha, A. Laari and T. Koiranen (2017). CFD-PBM Coupled Simulation of an Airlift Reactorwith Non-Newtonian Fluid, Oil Gas Sci. Technol


http://www.pixact.fi

Documents

CFD-PBM Coupled Simulation of an Airlift Reactor with Non