Lei Zhang ,Mengyu Wu ,Yanhe Han ,*,Meili Liu ,Junfeng Niu

1 Department of Environmental Engineering,Beijing Institute of Petrochemical Technology,Beijing 102617,China

2 Research Center for Eco-Environmental Engineering,Dongguan University of Technology,Dongguan 523808,China

Keywords:Iron-carbon micro-electrolysis Internal cycling Computational fl uid dynamics Structure design

ABSTRACT It isgenerally recognized that internal-loop reactorsare well-developed massand heat-transfer multiphase fl ow reactors.How ever,the internal fl ow fi eld in the internal-loop reactor is in fl uenced by the structure parameter of the reactor,w hich has a great effect on the reaction ef fi ciency.In this study,the computational fl uid dynamics simulation method was used to determine the in fl uence of reactor structure on fl ow fi eld,and a volume-offl uid model w as employed to simulate the gas-liquid,tw o-phase fl ow of the internal-loop micro-electrolysis reactor.Hydrodynamic factors were optimized w hen the height-to-diameter ratio w as 4:1,diameter ratio w as 9:1,draft-tube axial height was 90 mm.Three-dimensional simulations for the w ater distributor w ere carried out,and the results suggested that the optimal conditions are as follow s:the number of water distribution pipes w as four,and an inhomogeneous w ater distribution w as used.According to the results of the simulation,the suitable structure can be used to achieve good fl uid mechanical properties,such asthe good liquid circulation velocity and gas holdup,which provides a good theoretical foundation for the application of the reactor.

1.Introduction

Iron-carbon micro-electrolysisconsisting of scrap iron and activated carbon is considered a low cost,high-performance pre-treatment option for the puri fi cation of wastew ater[1-3].It can degrade the pollutants by the action of redox,fl occulation,and adsorption of ferric ions[4].

Because of its advantages,iron-carbon micro-electrolysis has become an ef fi cient pre-treatment tool and the technology is w idely applied for w astew ater from many industries,including the dye[5],petroleum[6],and coke[7]industries.How ever,there are practical problems with such reactor over long periods of operation.Tw o such problems are(1)the bed easily succumbs to passivation due to iron rust,resultingin short circuitsand dead zones,and(2)scrap iron isconsumed too quickly,causingthe fi llingprocessto betime-consumingand labour-intensive.

To overcome the problems identi fi ed above,we have developed a novel reactor w hich combines Internal-Loop Airlift Reactors(ILARs)and iron-carbon micro-electrolysis(Fig.1).In the developed reactor,the driving force is produced by the difference in the average gas densities betw een the riser and dow n-comer,and the multiphase fl ow inside and outside of the draft-tubes generates circulation.In contrast to mechanical stirring devices[8],ILARs have no rotating parts and reducecost.More importantly,there isan increase in the rate of particle collision,mixing,and masstransfer[9].Our research hasshown that our novel reactor possesses good hydraulic performance and ability of sew age treatment[5].

Relatively few studies have been devoted to the complexities of the fl ow fi eld of internal-loop micro-electrolysis reactors.Even though empirical and semi-empirical correlationshavebeen found predicting their hydrodynamic parameters[10-12],their design,operation,and control still present dif fi culties.

In recent years,a number of studies have focused on the structure optimization of the ILARs,w hich provides a good foundation for the design and ampli fi cation of the reactor.Generally,gasholdup and liquid velocity are usually used to evaluate the hydraulic performance of reactor with the experiment and simulation methods.The in fl uence of operating parameters and physical properties on the hydrodynamic property w as investigated by experiments(Tao Yang et al.[13],Cao et al.[14]).How ever,if the in fl uences of structure parameters on the hydrodynamic property w ere only carried out by experiment,a large amount of material and labour force will be w asted.Therefore,the computational fl uid dynamics(CFD)simulation methods have become the main w ay of studying the hydrodynamic property of the ILARs.

Fig.1.Schematic of the structure of an internal-loop,iron-carbon,micro-electrolysis reactor.1)gas sparger;2)horn-mouth;3)w ater inlet pipe;4)draft-tube;5)reactor w all;6)reactor cap;H:height of reactor;H d:axial position height of the draft-tube;D r:diameter of reactor;D d:diameter of draft-tube;A:local enlarged draw ing;A1:2 w ater distribution pipes;A2:4 w ater distribution pipes.

Davarnejad et al.[15]investigated theoverall gashold-up in theriser and dow n-comer of three different volume reactors,CFD simulation results and experimental data show ed that the gas holdup increases with enlarging of reactor scale,and the simulation results are in good agreement w ith the experimental data,and numerical simulation can further study the microscopic fl ow of gas-liquid two-phase fl ow.Chen and Bai[16]simulated the hydrodynamics in a cyclohexane oxidation airlift loop reactor using a tw o-fl uid model.The results indicated the key hydrodynamic parameters are liquid circulation velocity and liquid circulation fl ux,and liquid circulation fl ux can enhance gas holdup in the dow n-comer.Wadaugsorn et al.[9]systematically study hydrodynamic behaviors and mixing characteristics w ith a computational approach.Simulation results show ed structural parameters(heightto-diameter ratio,dow n-comer diameter and reactor scale)have signi fi cant effects on hydrodynamic parameters,and the detailed distribution of fl ow fi eld in the reactor is obtained[17].How ever,in speci fi c fi elds,the range of the structural parameters is certain scope,but not universally applicable to any ILARs.

In this study,a two-dimensional(2D)numerical simulation w as used to study the in fl uence of the height-to-diameter ratio(H/D),diameter ratio of riser to down-comer(Dr/Dd),and the axial position height of the draft-tube(Hd)w ater distribution w ays to the hydrodynamic parameters of a novel internal-loop iron-carbon micro-electrolysis reactor.Due to the in fl uence of the number of distribution pipes,a 2D model cannot be carried out in 2D model,;thus,a three-dimensional(3D)model w as employed to discuss this structure parameter and veri fi ed the reliability of 2D simulation results through the height-todiameter ratio.

2.Methodology

2.1.Geometry and meshing

The geometry w ascreated w ith the software GAMBITfor Fluent Preprocessing.The initial structural dimensions w ere a height(H)of 600 mm,a reactor diameter(Dr)of 300 mm,and the H/D of 2:1,3:1,and 4:1.The draft-tube had a diameter(Dd)of 30 mm and a height(H1)of 400 mm.The w ater distributor comprised tw o pipes,w ith uniform w ater distribution,located 90 mm from the draft-tube's bottom.The diameter of the distributor holes was 3 mm.

The quality of the CFD simulation grid has a critical impact on the simulation results,as does the ef fi ciency of computing[18].During the meshing process,an ideal division method not only produces highquality grids,but also conserves computational resources.Close attention must therefore be paid to the division of grids.

In order to maintain a high-quality grid for the reactor analysis,polygonal mesh generation w as used.The “Map”meshing method w as used to mesh the geometry w ith the exception of the cone part.As shown in Fig.2,the mesh interval size w as 3 mm(cell number 16502,13925,and 14942).

2.2.Boundary condition

In this simulation,the boundary conditions at the gas-inlet of the model w ere set as the velocity inlet.The turbulence intensity was 5%,the hydraulic diameter w as 20 mm,and the gas volume fraction w as one.The outlet was set to a pressure outlet with a turbulence intensity of 5%,and a hydraulic diameter of 20 mm.The w ater distributor w as used as the velocity inlet boundary.Its turbulence intensity w as 5%,and the hydraulic diameter w as 3 mm.The moving wall was applied at the top tw o sides of the reactor to better simulate the impact of w ater fl ow on the top cap.The other w alls w ere de fi ned as the no-slip boundary.Based on subsequent calculation,the liquid velocity w as set to 0.04 m·s-1,and the gas velocity w as 0.7 m·s-1.

2.3.Mathematical model

The Reynolds number w as used to select the turbulence model for the fl uid fl ow.When the Reynolds number is less than 2100,the fl ow in the pipe isa laminar fl ow[19].When the Reynolds number isgreater than 4000,the fl ow in the pipe is turbulent.According to Eq.(1),the Reynolds number at the entrance of the reactor was11045.

Fig.2.Reactor geometry for H/D:a)2:1;b)3:1;c)4:1.The geometry's meshing H/D ratios w ere d)2:1;e)3:1;and f)4:1.

w hereρis the liquid phase density,v is the liquid velocity,d is the diameter of inlet,μthe liquid phase viscosity.

The factors affecting simulation results come mainly from the shear stress,eddy,and separation produced by the tw o-phase circulation fl ow.How ever,the effect of the Near-Wall Grid Size of ILARs on the fl uid fl ow can be ignored.Hence,the standard k-εmodel of turbulence wasselected.This model was proposed by Launder and Spalding[20]in 1972,ow ing to its advantageous simulation of a slightly diffused fl ow fl uid as w ell as its economic ef fi ciency and relative stability.It has already become the primary tool for the calculation of engineering fl ow fi elds.

The turbulent kinetic energy k and the dissipation rateεfor the standard equation k-εmodel are as follow s[21]:

w here coef fi cient C1ε=1.44,C2ε=1.92,C3ε=0.09,σk=1.0,and σε=1.3.

In numerical simulations,the multiphase fl ow model should be introduced to describe the force and motion betw een tw o or more phases.The hydrodynamic modelling of tw o-phase fl ow uses a dual approach:a)the Euler-Lagrange formula and b)the Euler-Euler formula.If the tw o phases are treated as a continuous phase,the fi rst approach is chosen;otherw ise,the latter is selected.The tw o phases are composed of a primary phase(w ater)and secondary phase(air)w ithin the ILAR.Both are treated as continuous(similar to Xu et al.[17]).This volume-of-fl uid(VOF)simulation under the Euler-Euler approach,w hich can model tw o or more immiscible fl uids,relies on solving a single set of momentum equations and tracking the volume fraction of each of the fl uids throughout the domain[22].

3.Results and Discussion

3.1.Grid independence

This simulation used differently sized grids,which helps exclude the in fl uence of grid size on the simulation results.In other w ords,this reinforces the idea that hydrodynamic parameter differences w ere caused by structural or operating parameters,rather than the grid quality[23].The simulated liquid velocities w ith different cell sizes are show n in Fig.3.As the sizes of the cells increase from 1 mm to 5 mm,the velocity changes approximately 0.02 m·s-1for every 1 mm.For a cell size less than 3 mm,the hydrodynamic parameters w ere predominantly stable.Therefore,this simulation w as conducted w ith a cell size of 3 mm.

Fig.3.Veri fi cation of grid independence.

3.2.CFDresults

3.2.1.Effects of height-to-diameter(H/D)ratio

Fig.4 show s the radial distribution of liquid velocities at different H/D ratios at an axial height of 300 mm.We can see that with an increasing H/D ratio,the liquid circulation velocity increases in the riser zone,and the liquid velocity decreases in the down-comer zone.This makes the liquid velocity uniform at radial and axial distributions,increases the bubble residence time and total gas holdup[24],and improves the degree of liquid circulation.Fig.5 shows the velocity distribution of the liquid phase at different H/D ratios.With an increasing H/D ratio,the velocity distribution of the gas-liquid phases is more uniform in the dow n-comer zone,w hich is useful to the iron-carbon micro-electrolysis.

When the H/D ratio is2:1,themaximum liquid velocity in theriser is about 0.7 m·s-1,w hich ishalf at the H/D ratio of 4:1.Theliquid velocity outside the draft-tube was higher,up to 0.5 m·s-1,and the average liquid velocity on the lift side w as higher than the right side.This suggested that the liquid velocity distribution is not uniform.Fig.5 illustrates the liquid velocity contour of different H/D ratios.It can be seen that there is obvious bias fl ow at the bottom of the reactor,a portion of the air spilling directly through the tw o sides of the drafttube(see from Fig.5(a))under the wall effect.As a result,the liquid velocity in the riser is reduced,w hich ultimately affects the gas-liquid circulation performance.

Fig.4.Radial distribution of liquid velocity of different H/D ratios.

With an increase in the H/D ratio to 3:1,the reactor fl uid fl ow form w as better,and the speeds w ithin the draft-tube considerably improved.The liquid velocity distribution is more uniform here,but a portion of the gas is still entering the dow n-comer(as show n in Fig.5(b)).When the H/D ratio is approximately 4:1,the maximum liquid velocity in the draft-tube is about 1.4 m·s-1,which is about 75%higher than that at 3:1.The fl ow state is similar to the plug fl ow type,w ith the maximum velocity of the gas-liquid mixture concentrated in the draft-tube.The gas holdup is more uniform in the reactor,and the refracted fl ow is virtually eliminated.This is the ideal state of mass transfer.Therefore,the best H/D ratio is 4:1.

3.2.2.Effectsof the diameter ratio of riser to down-comer(Dr/Dd)

In this section,w e examine the effect of the Dr/Ddratio using an H/D ratio of 4:1.Three draft-tube diameters were used:28 mm(Dr/Dd=7:1),25 mm(Dr/Dd=8:1),and 22 mm(Dr/Dd=9:1).These diameters w ere examined w ith the value of the other structural parameters unchanged.From Fig.6(a),we can see that when a 7:1 diameter ratio is used,much gas enters the down-comer zone,inducing that it is dif fi cult for the w astew ater to form a cycle inside the reactor.

It can be seen from Fig.7 that as the ratio increases,the average liquid velocity rises in the riser.When the ratio is 9:1,the maximum velocity is about 2.13 m·s-1,w hich w as almost one time higher than that of the Dr/Ddratio of 7:1.A good fl ow regime can generate in the dow n-comer zone w hen the diameter ratios are 8:1 and 9:1.The distribution of the gas-liquid tw o-phase fl ow is especially symmetric w hen the diameter ratio is 9:1.

3.2.3.Effects of the axial position height of the draft-tube(Hd)

The axial position height of the draft-tube(Hd)refersto the vertical distance from the reactor's bottom to the draft-tube and mainly affects the gas and liquid circulation in the reactor.Models w ith Hdat 50 mm,90 mm,and 150 mm w ere studied in our simulation.From Fig.8,more gas enters the dow n-comer zone w hen the Hdw as increased.Especially at 150 mm,only a small portion of the gas can enter the riser zone.Hence,a large amount of gas that enters the dow n-comer zone increases the resistance of fl uid fl ow.

Fig.5.Liquid velocity contour of different H/D ratios.

Fig.6.Liquid velocity contour of different D r/D d ratios.

Hencethe height of the draft-tubecan affect the fl ow direction of the gas.In order to quantitatively investigate the effect of Hdon the liquid velocity,three cross sections of difference positions(H1:the connection area of draft-tube and horn-mouth;H2:the end edge of the hornmouth;H3:the middle position betw een H1and the bottom of the reactor)w ere selected.Although Hdis different,three pictures(Fig.9(a)-(c))have the same trend that the average liquid velocity possesses high level(about 0.7 m·s-1)at the location of H3.How ever,the liquid velocity of Fig.9(a)is obviously low er than(b)and(c)at H1.As mentioned earlier,with the decrease of Hdmore gas will enter the draft-tube and then improve liquid velocity.How ever,the maximum liquid velocity is only 1.8 m·s-1.It may be that the Hdis too low to raise the circulating resistance of the dow n-comer.

When the Hdis 150 mm,the maximum liquid velocity at the H1position is about 2.6 m·s-1,w hich is better than that(about 1.8 m·s-1)w hen the Hdis 50 mm.How ever,it is low er than that w hen Hdis 90 mm.This phenomenon can also be seen from the liquid velocity contour of different Hd,asshown in Fig.8.It may bedue to thegasenters into the dow n-comer zone w hen the Hdis higher than a certain value,w hich w ill increase the resistance of circulating fl ow.Therefore,the Hdhasasigni fi cant impact on thecirculation drivingforce,astheresultsreported by Chen and Bai[16].The Hdw as set at 90 mm in the follow-up study.

Fig.7.Radial distribution of liquid velocity at different D r/D d ratios.

3.2.4.Effects of water distribution

Thefunction of thew ater distributor isto distributethew ater evenly w ithin and make full use of the packing layer.The fl ow rate of the fl uid w ithin the w ater distributor should be uniform,neither too fast nor too slow.In this section,only the radial distribution of liquid w as investigated,therefore,tw o-dimensional simulation w ith 2 pipesw asselected.We found that uneven distribution often appears in the simulation as a result of different structural changes,short circulation currents,and bias w ithin the current.In Fig.10(w here the liquid velocity of the distribution pipe constitutesthevertical axisand theradial distanceisshown as the horizontal),w e can see that the average liquid velocity of the uniform w ater distribution(the inlet holes in the pipes have an equal distance)is low,at approximately 0.3 m·s-1.This makes it dif fi cult to form a good cycle.By changing the uniform distribution of w ater to a calculated non-uniform distribution(non-uniform w ater distribution has an equal distribution area),the ori fi ce spacing is the largest at the centre of the draft-tube and gradually decreases tow ards the reactor's side w all.The liquid velocity is above 0.4 m·s-1,hence uniform and full circulation of w ater can be achieved.

Fig.8.Liquid velocity contour of different H d.

Fig.9.Radial distribution of liquid velocity at three different axial heights in reactor.((a)H d=50 mm;(b)H d=90 mm;(c)H d=150 mm.)

Fig.10.Velocity distribution of uniform and non-uniform distribution of w ater.

3.2.5.Effects for the number of water distribution pipes

The w ater distributor,located at the bottom of the reactor,is one of the most important parts of the internal circulation of iron-carbon micro-electrolysis reactors.Liquid out fl ow s from the w ater distribution pipes into the reactor and mixes w ith air bubbles;then,gas-liquid cycling occurs in the reactor.Therefore,the fl ow number and distribution holes of the water distribution pipe reactor also have an important impact on the w ater distributor[25].

The uniform distribution of w astew ater in the dow n-comer zone is desired,w hich involves three-dimensional space.Therefore,the GAMBITsoftw are for three-dimensional modelling and mesh generation w as employed to investigate the in fl uence of number of w ater distribution pipes on the w astew ater distribution.A high-accuracy hexahedral mesh division w as used to allow for better simulation regarding how the number of water pipes in fl uences the reactor.The overall liquid phase circulation speed represents the hydraulic performance of the reactor to a certain extent and is related to the apparent gasvelocity,the geometric structure,and thephysical properties(viscosity,density,etc.)of the liquid phase.As can be seen from Fig.11,the overall liquid speed of four tubes is clearly greater than that of two(because of the greater number of w ater holes).An increase in the number of w ater holes corresponds to a more uniform liquid in fl ow.This means that less gas is carried into the dow n-comer by the liquid phase,low ering the gas content of the dow n-comer,increasing the gas content w ithin the riser,and making the gas-liquid circulation more uniform.

Fig.11.Effect of the number of water distributing pipes on the overall liquid velocity.

Fig.12.Liquid velocity contour of different H/D ratios.

3.3.Three-dimensional veri fi cation

3D simulation of the height-to-diameter ratio w as carried out to verify the reliability of the 2D model.The same model and boundary conditions as the 2D simulation w ere used.Fig.12 show s the effect on the liquid velocity at H/D ratios of 2:1,3:1,and 4:1.At the same superfi cial gas velocity,the velocity in the draft-tube increases w ith increasing H/D ratio.When the H/D ratio is 4:1,the average liquid velocity w ithin the draft-tube is approximately 1.35 m·s-1for an increase of approximately 35%over an H/D ratio of 2:1.It can also be seen from Fig.13 that as H/D increases,the liquid velocity in the dow n-comer zone gradually increases.

Fig.13.Effect of H/D ratio on the overall liquid velocity.

In thethree groupsof simulations,the resultsindicatethat the liquid velocity value of the raiser is directly proportional to the liquid velocity of the dow n-comer zone,w ith the velocity of the dow n-comer zone being highest at 4:1.The circulation effect is most obvious at these levels.This phenomenon show s that the H/D ratio has an obvious in fl uence on the fl ow fi eld and that the in fl uence law is the same as in the 2D simulations.Hence,the 2D simulation can be replaced by the 3D simulation under certain conditions,reducing time and ensuring accurate results.

4.Conclusions

For internal-loop iron-carbon micro-electrolysis reactors,the VOF tw o-fl uid model is an ef fi cient CFD mathematical model for the complexities of a fl ow fi eld inside the reactor.The model accurately re fl ects the numerical values of the gas-liquid,tw o-phase fl ow characteristics within the reactor.It can,therefore,be used as the basis of understanding for the fl uid mechanics performance of the reactor,and as a basic tool for industrial ampli fi cation design.

The numerical calculations show that the H/D ratio makes the fl ow of the gas and liquid more uniform.The ratio of the reactor to the draft-tube diameter(Dr/Dd)predominantly affects the axial velocity w ithin the internal circulation tube and the circulation effect of the reactor.The diameter ratio is optimized at 9:1.The axial height of the draft-tube predominantly affects the degree of gas-liquid mixing in the mixing zone.If it is too high,the gas w ill enter the dow n-comer zoneand will affect circulation.If thediameter ratio istoo low,it w ill increase the resistance of thecirculation.Thebiascurrent may be reduced to a certain extent by uneven water distribution.

The number of w ater distribution pipes w as analysed via 3Dsimulation,w hich determined that a four-pipe w ater distribution system is better than a two-pipe system.The simulation results con fi rm that an inhomogeneous w ater distribution is obtained w hen the H/D ratio is 4:1,Dr/Ddis 9:1,Hdis 90 mm,and the number of water pipes is four.

Through simulation,w e ascertained that the relationship betw een the fl ow fi eld distribution,velocity variation,and mass transfer offers a good foundation for future research on the internal reaction mechanism.CFDnumerical simulation is useful for the continued exploration of different structural designs and operational parameters regarding increasing fl ow models for industrial production.