Computational fluid dynamics simulation of a novel bioreactor for sophorolipid production☆

Xiaoqiang Jia ,Lin Qi,Yaguang Zhang ,Xue Yang ,Hongna Wang ,Fanglong Zhao ,Wenyu Lu ,*

1 Department of Biochemical Engineering,School of Chemical Engineering and Technology,Tianjin University,Tianjin 300072,China

2 Key Laboratory of Systems Bioengineering(Tianjin University),Ministry of Education,Tianjin 300072,China

3 Synthetic Biology Platform,Collaborative Innovation Center of Chemical Science and Engineering(Tianjin),Tianjin 300072,China

1.Introduction

Surfactants are a group of amphiphilic chemicals consisting of both hydrophilic and hydrophobic regions that partition preferentially at the interface between fluid phases[1,2].Annual consumption of surfactants in the world is more than 13 million tons,and most of them are synthesized by chemical methods.The large majority of the currently used surfactants are petroleum-based and are produced by chemical means.

Biosurfactants,mainly produced by microorganisms,have advantages over chemical surfactants for their lower toxicity,higher biodegradability,better environmental compatibility and higher selectivity[3,4].Biosurfactants are grouped as glycolipids,lipopeptides,phospholipids,fatty acids,neutral lipids,polymeric and particulate compounds[5].Sophorolipid is a kind of extracellular glycolipid biosurfactants which comprise a hydrophilic carbohydrate section and a hydrophobic fatty acid chain[6].Sophorolipid is widely used in the environmental remediation and cleaning industries for the advantages of biodegradability,high surfactivity,low ecotoxicity and the production on renewableresource substrates[4,7,8].In the environmental remediation area,SLs are among one of the most promising biosurfactants for heavy metal removal from soil sediments.Experiments have indicated that sophorolipids could enhance biodegradation of the insoluble aromatic compounds like phenanthrene through enhanced solubilization[9-11].In addition,SLs have been successfully applied in the petroleum industry such as secondary oilrecovery,remove hydrocarbons from drillmaterial,and the regeneration of hydrocarbons from dregs and muds[12].

For the required characteristics and great potential to replace the existing chemical surfactants,many studies have focused on reducing production costs or improving the yield of SLs to make it competitive with existing chemical surfactants[3,4].As the SL production is a complex,multiphase chemical,biological and physical process,a lot of research has been performed on the optimization of the fermentation process,including fermentation type,culture conditions,carbon sources,medium components,and so on.Ribeiroet al.[13]observed that the production of SLs was favored when culture media was supplied with avocado,argan,sweet almond and jojoba oil or when NaNO3was supplied instead of urea,which indicated the potential of the selective production of SLs based on the selection of carbon and nitrogen sources to culture media.Maet al.[14]used cell lysate ofC.curvatus,oleic acid,and delignined corncob residue hydrolysate(DCCRH)/detoxified DCCRH as nitrogen and carbon sources and results demonstrated that renewable DCCRH can be utilized for the production of high-value SLs.Davereyet al.[15]studied low cost media based on sugarcane molasses and three different oils for the production of sophorolipids(SLs)from the yeastCandida bombicolain batch shake flasks.However,hardly any work has been published on the physical characteristics affecting the efficiency of SL production.

During recent decades,many studies have been made including structural improvement,hydrodynamic condition optimization,in order to further improve various types of bioreactors to meet the demands of microorganism cultures and industrialized application.The hydrodynamic behaviors such as gas volume fraction,velocity fields,distributions of shear stresses,turbulentintensity which significantly affect the growth of microorganisms and the production of fermentation product,have been investigated.Kleinet al.[16]assess the effect of operational conditions(air- flow rate,biomass concentration)on hydrodynamic behavior ofan airliftbioreactor and the results on hydrodynamics can be useda priorior during the bioprocess to optimize operational parameters to avoid the occurrence of undesirable bioreactor stalling and to maximize the process productivity.The rheological properties of a fermentation with the fungusBeauveria bassianaunder different hydrodynamic conditions were studied and the simulated results will be helpful in the optimization of scale-up production of these fungi[17].Hence,it is necessary to investigate the hydrodynamic phenomena involved in SL production for the industrialscale application.Research of SL production from the perspective of bioreactor design and separation process optimization will be interesting and promising.

As the laboratory-scale yeast-bioreactor system for the production of SLs[18],traditional semi-empirical approach to bioreactor design is time-consuming and the application of experimental techniques is limited to large amount spend on investigating flow fields,mass concentration fields,etc.Numerical simulation based on computational fluid dynamics(CFD)calculations can provide a feasible way to explain the hydrodynamic behavior of bioreactors under different conditions and has been employed to optimize the design of bioreactor by researchers[19-21].Dinget al.[22]applied CFDsimulations to evaluate the role of hydrodynamics in reactor design and optimize the reactor con figuration in a laboratory-scale continuous stirred-tank reactor used for biohydrogen production.Liuet al.[23]developed a two dimensional CFD model for optimizing the structure design of an airlift sonobioreactor for hairy root culture.

The aims of the presentwork are to investigate the flow characteristics of a new bioreactor with dual ventilation-pipe and double sieveplate(DVDSB),in which the semi-continuous fermentation of SLs was achieved and established using CFD simulation and experimental validation.The hydrodynamic behavior including the velocity field,volume-averaged overall and time-averaged local gas volume fraction and liquid phase turbulent kinetic energy was predicted.Comparisons were then made between conventional fed-batch fermenter and DVDSB of hydrodynamic behavior on SL production was predicted and validated.

2.Materials and Methods

2.1.Reactor configuration and culture conditions

The dual ventilation-pipe and double sieve-plate bioreactor(DVDSB)with a total volume of 5.7 L was operated in a semicontinuous flow mode for SL fermentation[Fig.1].Compared with the conventional fermenter,there is a sieve-plate in the middle of new designed bioreactor.The reactor is divided into cylinder part and cone part by the sieve-plate.The upper cylinder part and the tank of traditional reactor is the same,while the lower conical part is mainly for the sedimentation and collection of SLs.Furthermore,there are two ventilation pipelines in the novel reactor.The main oxygen supply pipeline and assistant oxygen pipeline are in the cone area and cylinder area respectively.The main oxygen supply pipeline can provide oxygen for the whole reactor while the assistant oxygen pipeline only supplies oxygen for the cylinder area.

Candida albicansO-13-1 obtained from the Ocean University of China was used in this study.The temperature was maintained at 30°C throughout the fermentation process.Seed medium contains:100 g·L-1of glucose,10 g·L-1of yeast extract,1 g·L-1of(NH2)2CO,pH is 6.0.The initial fermentation medium contains:120 g·L-1of glucose,120 g·L-1of oleic acid,3.5 g·L-1of yeast extract,0.5 g·L-1of peptone,5 g·L-1of sodium citrate,4 g·L-1of MgSO4·7H2O,2 g·L-1of(NH4)2SO4,1 g·L-1of KH2PO4,0.1 g·L-1of NaCl,0.1 g·L-1of CaCl2·2H2O[24].Supplemented medium contains:200 g·L-1of glucose,5 g·L-1of yeast extract,2 g·L-1of(NH2)2CO.In the first twenty hours,pH was maintained at 5.8-6.2,then maintained at 3.5-4.0.The air flow rate was 8 L·min-1,and the stirring speed was 450 r·min-1.The same operation and fermentation conditions as mentioned above were operated in new bioreactor and conventional fermenter.

2.2.Analytical methods

The key parameters including temperature,pH,dissolved oxygen concentration,cell population,SL concentration,were observed during SL fermentation.The content of yeast cells was determined by the dry weight method and the microscopic counting method[25,26].An oxygen electrode(Hamilton FDA 120,Bonaduz,Switzerland)was equipped in the bioreactorto monitor the concentrations of dissolved oxygen.The concentration of SLs in the culture was measured according to previous studies[8],which was repeated three times for every experiment.

3.Computational Fluid Dynamic Model

3.1.Model assumption

In this study,we assume that yeast cells distribute uniformly in liquid,and yeast cells and the mixture of the medium are regarded as one liquid phase,while the fluid phase characteristic such as density and viscosity was specified according to measurement results.From the above assumptions,there are two phases in the reactor(one gas phase and one liquid phase),and an Eurlerian-Eurlerian multiphase model in ANSYS CFX is appropriate to describe the flow behaviors of two different phases.

3.2.Governing equations

The Eulerian approach was adopted to describe flow behaviors of the gas and liquid phases,which were considered to be the dispersed and continuous phases,respectively.

The two phase holdups satisfied the compatibility condition:

Where αgand αlare volume fraction of gas phase and liquid phase,respectively.

The continuity equations are:

Heret,ρ,and uiare the time,density and velocity of each phase,respectively.The subscriptsi=g andlrepresentgas and liquid.

The gas-liquid two phase momentum conservation equations are:

Note thatp,μeff,i,g and MI,liare the pressure,effective viscosity for phasei,gravity acceleration vector,and inter-phase momentum transfer force.

Fig.1.Schematic diagram of the new Bioreactor with dual ventilation pipe and double sieve-plate(DVDSB)(unit:mm)(a:side view;b:top view;c:the detail of sieve).

3.3.Inter-phase momentum transfer

In this study,drag force and lift force between the continuous phase and dispersed phase were considered among the inter-phase momentum transfer forces,while virtual mass force and turbulent dispersion force was neglected,as adding of them did not bring any obvious refinement to the current simulation results,but only convergence difficulties[27].

Drag force exerted by the dispersed phase on the continuous phase was calculated by:

Here,MD,CD,anddgare the interphase drag force,drag coefficients and bubble diameter,respectively.

Drag coefficient exerted by the gas phase on the liquid phase was obtained by the Ishii-Zuber drag model:

Note thatRem,EoandE(αg)are the mixture Reynolds number,Eotvos number,and correction term.

Here,um,σ,andE(αg)are the gas-liquid mixture velocity and surface tension between liquid phase and gas phase and correction term,respectively[27].

Lift force acting perpendicular to the direction of relative motion of two phases was given by:

AndCLis the dimensionless lift coefficient with a value of 0.5.

3.4.Initial and boundary conditions

Transient calculation started from assuming that gas holdup was zero in the reactor.It was also assumed that the gas bubbles supplied by the oxygen pipeline were distributed in the mixture with a fixed mean diameter of 5 mm.The multiple reference frame(MRF)boundary condition was adopted at the motion region around the impeller and the reactor was divided into a rotating domain and a stationary domain.The boundary condition for reactor in walls was defined as no-slip for the liquid phase and free-slip for the gas phase.At the top of the computational domain a degassing condition was defined for the outlet boundary so that only gas phase can leave the domain.

3.5.Numerical solution

A commercial computational fluid dynamics code ANSYS CFX 13.0 was used to establish the model to investigate hydrodynamics in two bioreactors.The geometry and the unstructured grid of the bioreactors were generated by ANSYS ICEM.The number of nodes and tetrahedral cells generated for the conventional fed-batch fermenter was 41370 and 204566,while for the new bioreactor was 51055 and 258953,respectively.Preliminary simulations were performed to ensure that the simulation results were independent of mesh size with this number of cells.The mesh layout for reactors geometry is shown in Fig.2.The total simulation time for each three-dimensional transient CFD case was 120 s.

4.Results and Discussion

4.1.Model validation

Considering the limitation of available experimental techniques,overall gas holdups(ε)were measured by the volume expansion method during the steady state condition in the bioreactor(Eq.(13)).

As far as the CFD model validation is concerned,Fig.3 presents a comparison between the experimentally measured and the simulated values of the total gas holdups at fixed agitation speed of 450 rpm and varied inlet air flow rate of 2,4,6,and 8 L·min-1.It was noticed that model simulations were in very good agreement with experimental measurements,which indicated the reliability of the CFD model.

4.2.Hydrodynamics simulation

Fig.3.Modelsimulated and experimental measured volume-averaged overallgas holdups in the conventional fermenter and new bioreactor with DVDSB.

Gas holdup is defined as the fraction ε of gas volume in gas-liquid dispersion and is commonly used to characterize oxygen mass transfer and mixing of aerated vessels[28].Fig.4 shows the model simulated volume-averaged overall gas phase volume fraction in two bioreactors under fixed inlet air flow rate of 8 L·min-1and agitation speed of 450 r·min-1.After 20 s of agitation and aeration,the gas phase volume fraction distribution in the reactors reaches steady state.The gas phase volume fraction in the new bioreactor presents higher than conventional fermenter.As the two reactors operate under same conditions,this could be attributed to the design of the cone area and sieve plate.It contributes to increasing the moving distance of the air from the main oxygen supply pipeline to the outlet(as shown in Fig.1),and the mean residence time of gas in the reactor is increased with the increase of reactor height.This provides sufficient resident time of the water in reactor to permit the injected oxygen gas to transition into the dissolved state prior to reaching the top of the reactor,which is meaningful for the enhancement of the utilization ratio of oxygen for the microorganisms' cultivation.

Fig.2.Unstructured mesh layout for reactors geometry.

Fig.4.Model simulated reactor volume-averaged overall gas holdups along time in the conventional fermenter and new bioreactor with DVDSB.

Fig.5 shows the transversal distributions of time-averaged local gas holdups at five different vertical positions.Considering the geometry symmetry of the reactors(Fig.2),half of the vertical section(Z=0 mm,X=20-80 mm)was presented in this figure and five typical vertical positions were selected:5 mm below the lower impeller(Y=30 mm),5 mm above the lower impeller(Y=60 mm),5 mm below the upper impeller(Y=170 mm),5 mm above the upper impeller(Y=200 mm),middle of the lower and upper impellers(Y=115 mm).It was evident from the figure that the local gas holdups in two reactors concentrate in the zone close to the center axis and disperse in horizontal direction,and the local gas phase volume fraction was also influenced by vertical height and the largest local gas phase volume fraction was at the position below the lower impeller(Y=30 mm).It might be that air from the inlet is hard to be brought to the tank under current agitation speed of 450 r·min-1which results in the bubble accumulation near the impeller,and the increase in agitation speed would have a significant effect on increasing local gas holdup which was proven in our previously published work[27].Also,the local gas holdups in the new reactor are higher than that in the conventional reactor at similar position,which is consistent with the previous conclusion about total gas holdup.A more uniform distribution of gas phase in the new reactor is important under such circumstances which is efficient and beneficial to the utilization of microorganisms.

The three-dimensional CFD model was also used to predict transient gas holdup distributions at vertical sections(X=0 mm)as shown in Fig.6,with three time points selected:t=10,60,110 s(from left to right).It can be seen that the values of gas phase volume fraction around the central axis is higher than those in the bulk flow region,and the feed air and the corresponding dissolved oxygen get a wider distribution with the agitation of two impellers in the horizontal direction.It is the same for traditional reactor and the cylinder area of the novel reactor for the same air flow rate and agitation speed.The difference is that due to the isolation of sieve plate,the ventilation from the bottom plays a great role in increasing gas phase concentration in cone area.Also,it can be seen that the vertical distance for the air injected from the main oxygen supply pipeline is vertically oriented and longitudinally extended,which conduces to improve the reactor total gas holdup.

The transient gas holdup distributions at five typical horizontal sections:Y=30,60,115,170,200 mm(from left to right)were carried out as shown in Fig.7,with a specified time point:t=60 s.From the comparison of transient local gas holdup distributions of five typical horizontal sections,it can be found that regions near the impellers had a better gas dispersion compared with regions in the middle of the lower and upper impellers,which indicates that the design and location of impeller have an obvious effect on the gas holdup distributions.

The transient liquid phase velocity distributions at vertical sections(X=0 mm)were carried out as shown in Fig.8,with several time points selected:t=10,60,110 s(fromleftto right).The flow characteristics,e.g.the presence of vortices,recirculation zones,and dead zones,which were important for the mixing of substrate,microbialcell and oxygen allover the tank in fermentation processes,should be concerned.It can be seen that the two reactors generate powerful vortex area and liquid recirculation was observed in regions of the bottom of the tank,which was often considered to be “dead”regions,and it could ensure better mixing of fermentation substrates and gas and promise higher mass-transfer efficiency in the reactor.But the velocity in cone area in the new reactor is dramatically lower than cylinder area which could be interpreted as an obvious“dead”zones.The dead zones meant that the mixing was poor in this area whose volume should be reduced,but it provides good conditions for the separation of sophorolipids during the fermentation process.

Sophorolipids are heavier than water and the solubility of the sophorolipids is very low in the acidic fermentation environment.The sophorolipids would sedimentate onto the bottom of the tank naturally once the concentration reaches the saturation in the stationary state.Thus a semicontinuous fermentation process of sophorolipids was established in the new designed bioreactor,during which the production and the product separation were combined successfully.

Fig.5.Modelsimulated time-averaged localgas holdups along transversalcourse(Z=0 mm,X=20-80 mm)at five verticalpositions(Y=30,60,115,170,200 mm)in the conventional fermenter and new bioreactor with DVDSB.

Fig.6.Model predictions of transientgas phase volume fraction distributions at the vertical sections(Z=0 mm)in the conventional fermenter and new bioreactor with DVDSB with t=10,60,110 s.

Fig.7.Model predictions of transient gas holdup distributions at the horizontal sections(Y=30,60,115,170,200 mm)in the conventional fermenter and new bioreactor with DVDSB with t=60 s.

Fig.8.Model predictions of transient liquid velocity distributions at the vertical sections(Z=0 mm)in the conventional fermenter and new bioreactor with DVDSB with t=10,60,110 s.

In addition,the simulated distribution contour of turbulent kinetic energy(k)shows that the values of turbulent kinetic energy in the lower impeller region were higher than those in the bulk flow region(Fig.9),which is similar in the two reactors for the same stirring power.

4.3.Effect on fermentation results of SLs

Fig.10 summarizes the comparison of the sophorolipid fermentation results between new bioreactor and conventional fermenter.Compared with 96 h in the traditional reactor,the fermentation period in the new bioreactor was dramatically prolonged to 320 h.The fermentation yield of SLs and the product formation rate in the DVDSB were increased significantly to 484 g·L-1and 1.51 g·L-1·h-1compared with the traditional reactor of 120 g·L-1and 1.25 g·L-1·h-1,respectively.In addition,the conversion efficiency of carbon source of the new reactor was 60.0%,while the traditional reactor is 51.0%.

By integrating with the previous results of simulation,a qualitative relationship between hydrodynamics and sophorolipid production can be obtained.As the overall gas holdup in the new reactor is improved,along with the increase of local gas holdup,the yield of sophorolipids increases in the DVDSB.But the velocity distribution in the new reactor is inhomogeneous,especially the formation of“dead zones”in the cone area.It would negatively affect the mass-transfer efficiency,but it is obviously convenient for the natural sedimentation for SLs.What more,it is noteworthy that the DVDSB can be operated under different conditions on hydrodynamic behaviors,e.g.:with higheragitation speed orrelatively larger amount of air flow to a more dispersed gas distribution all over the whole reactor;however,this would have a negative effect on cultivation of microorganism.Thus,further investigation on the relationship between hydrodynamics and sophorolipid production is necessary.

Fig.9.Turbulence kinetic energy distribution contour of model in the conventional fermenter and new bioreactor with DVDSB.

5.Conclusions

Fig.10.Comparison of the sophorolipid fermentation results in new bioreactor with conventional fermenter.

The flow characteristics of the novel bioreactor with DVDSB for the SL production were investigated and compared with conventional fedbatch fermenter rectors by using CFD simulation and experimental validation.The comparison results on the hydrodynamic behavior show that due to the design of the sieve plate,the new bioreactor generates a better gas-liquid dispersion characteristics than the conventional fed-batch fermenter,and the fermentation yield of SLs and the product formation rate in the DVDSB was increased significantly to 484 g·L-1and 1.51 g·L-1·h-1compared with the traditional reactor of 120 g·L-1and 1.25 g·L-1·h-1,respectively.

The results on hydrodynamic behaviors in the DVDSB can be used to optimize operational parameters and the design of bioreactor to maximize the product productivity and indicates the DVDSB has the potential for application on a bigger scale fermentation of SLs.

Acknowledgements

The authors specially thank Dr.Shengkang Liang from Ocean University of China who provided theCandida albicansO-13-1.

[1]A.E.Elsha fie,S.J.Joshi,Y.M.Al-Wahaibi,A.S.Al-Bemani,S.N.Al-Bahry,D.A.Al-Maqbali,I.M.Banat,Sophorolipids production byCandida bombicolaATCC 22214 and its potential application in microbial enhanced oil recovery,Front.Microbiol.6(2015)1324.

[2]C.N.Mulligan,Environmental applications for biosurfactants,Environ.Pollut.133(2)(2005)183-198.

[3]R.Marchant,I.M.Banat,Biosurfactants:A sustainable replacement for chemical surfactants?Biotechnol.Lett.34(9)(2012)1597-1605.

[4]I.N.A.V.Bogaert,K.Saerens,C.D.Muynck,D.Develter,W.Soetaert,E.J.Vandamme,Microbial production and application of sophorolipids,Appl.Microbiol.Biotechnol.76(1)(2007)23-34.

[5]A.M.Davila,R.Marchal,J.P.Vandecasteele,Sophorose lipid fermentation with differentiated substrate supply for growth and production phases,Appl.Microbiol.Biotechnol.47(5)(1997)496-501.

[6]A.Daverey,K.Pakshirajan,P.Sangeetha,Sophorolipids production byCandida bombicolausing synthetic dairy wastewater,Int.J.Environ.Sci.Eng.13(3)(2009)481-488.

[7]D.A.P.Bramwell,S.Laha,Effects of surfactant addition on the biomineralization and microbial toxicity of phenanthrene,Biodegradation11(4)(2000)263-277.

[8]I.N.A.V.Bogaert,J.Zhang,W.Soetaert,Microbial synthesis of sophorolipids,Process Biochem.46(4)(2011)821-833.

[9]C.Schippers,K.Gessner,T.Müller,T.Scheper,Microbial degradation of phenanthrene by addition of a sophorolipid mixture,J.Biotechnol.83(3)(2000)189-198.

[10]C.N.Mulligan,R.N.Yong,B.F.Gibbs,Heavy metal removal from sediments by biosurfactants,J.Hazard.Mater.85(1-2)(2001)111-125.

[11]I.M.Banat,Biosurfactants production and possible uses in microbial enhanced oil recovery and oil pollution remediation:A review,Bioresour.Technol.51(1)(1995)1-12.

[12]L.Pesce,A Biotechnological Method for the Regeneration of Hydrocarbons From Dregs and Muds,on the Base of Biosurfactants,European Patent.,EP 1427547 A1[P],2004.

[13]I.A.M.Ribeiro,B.Rosário,M.F.Castro,M.H.L.Ribeiro,Sophorolipids:Improvement of the selective production by Starmerella bombicola through the design of nutritional requirements,Appl.Microbiol.Biotechnol.97(5)(2013)1875-1887.

[14]X.J.Ma,H.Li,D.X.Wang,X.Song,Sophorolipid production from delignined corncob residue byWickerhamiella domercqiaevar.sophorolipid CGMCC 1576 andCryptococcus curvatusATCC 96219,Appl.Microbiol.Biotechnol.98(1)(2014)475-483.

[15]A.Daverey,K.Pakshirajan,Production of sophorolipids by the yeastCandida bombicolausing simple and low cost fermentative media,Food Res.Int.42(4)(2009)499-504.

[16]J.Klein,J.Maia,A.A.Vicente,L.Domingues,J.Teixeira,M.Jurascik,Relationships between hydrodynamics and rheology of flocculating yeast suspensions in a high-celldensity airlift bioreactor,Biotechnol.Bioeng.89(4)(2005)393-399.

[17]D.M.Nú?ezramírez,L.Medinatorres,J.J.Valencialópez,F.Calderas,M.J.López,H.Medranoroldán,A.Solíssoto,Study of the rheological properties of a fermentation broth of the fungusBeauveria bassianain a bioreactor under different hydrodynamic conditions,J.Microbiol.Biotechnol.22(11)(2012)1494-1500.

[18]A.Daverey,K.Pakshirajan,Treatment of dairy wastewater containing high amount of fats and oils using a yeast-bioreactor system under batch,fed-batch and continuous operation,Desalin.Water Treat.(2015)1-7.

[19]V.Nagarajan,V.Ponyavin,Y.Chen,M.E.Vernon,P.Pickard,A.E.Hechanova,CFD modeling and experimental validation of sulfur trioxide decomposition in bayonet type heat exchanger and chemical decomposer for different packed bed designs,Int.J.Hydrog.Energy34(6)(2009)2543-2557.

[20]B.N.Murthy,R.S.Ghadge,J.B.Joshi,CFD simulations of gas-liquid-solid stirred reactor:Prediction of critical impeller speed for solid suspension,Chem.Eng.Sci.62(24)(2007)7184-7195.

[21]F.Zhu,J.Huang,J.Chen,L.I.Yuanguang,CFD simulation and optimization of an open raceway photo-bioreactor,Chem.Ind.Eng.Prog.32(4)(2012)922-941.

[22]D.Jie,W.Xu,Z.Xue-Fei,R.Nan-Qi,G.Wan-Qian,CFD optimization of continuous stirred-tank(CSTR)reactor for biohydrogen production,Bioresour.Technol.101(18)(2010)7016-7024.

[23]R.Liu,Y.Liu,C.Z.Liu,Development of an efficient CFD-simulation method to optimize the structure parameters of an airlift sonobioreactor,Chem.Eng.Res.Des.91(2)(2013)211-220.

[24]Y.Xue,L.Zhu,C.Xue,C.Yu,Q.Liang,W.Lu,Recovery of purified lactonic sophorolipids by spontaneous crystallization during the fermentation of sugarcane molasses withCandida albicansO-13-1,Enzym.Microb.Technol.51(6-7)(2012)348-353.

[25]W.Pan,J.A.Perrotta,A.J.Stipanovic,C.T.Nomura,J.P.Nakas,Production of polyhydroxyalkanoates byBurkholderia cepaciaATCC 17759 using a detoxified sugar maple hemicellulosic hydrolysate,J.Ind.Microbiol.39(3)(2012)459-469.

[26]N.Szeszenia-D?browska,W.Sobala,B.?wi?tkowska,G.Stroszejn-Mrowca,U.Wilczyńska,Environmental asbestos pollution — Situation in Poland,Int.J.Occup.Med.Environ.Health25(1)(2012)3-13.

[27]H.Wang,X.Jia,X.Wang,Z.Zhou,J.Wen,J.Zhang,CFD modeling of hydrodynamic characteristics of a gas-liquid two-phase stirred tank,Appl.Math.Model.38(1)(2014)63-92.

[28]S.García-Salas,C.Orozco-Alvarez,R.M.Porter,F.Thalasso,Measurement of local gas holdup in bubble columns via a non-isokinetic withdrawal method,Chem.Eng.Sci.60(24)(2005)6929-6938.