Simulating confined swirling gas-solid two phase jet

SSN 1009-3095 Joumal of Zhejiang University SCIENCE V 3 No. 2 P 157-161 Apr. -June, 2002http:/www.periodicals.comcn;http:/www.zju.educ/englishhttp:/www.zjupress.com;httpV/lib.Zju.educneindex.htm:jzu_s@mail.hz.zj.cn157Simulating confined swirling gas-solid two phase jetJIN Han-hu(金晗辉), XIA Jur(夏钧), FAN Jian-ren(樊建人) cen Ke-f岑可法)Department of Energy Engineering Zhejiang Universily, Hangzhou, 310027, ChinaReceived May 21, 2001 i revision accepted July 28, 200Abstract: a h-E-k, multi-fluid model was used to simulate confined swirling gas-solid two phcomprised of particle-laden flow from a center tube and a swirling air stream entering the testfrom the coaxial annular. After considering the drag force bet ween the two phases and gravity a seriesof numerical simulations of the two-phase flow of 30um, 45]m, 60jm diameter particleformed on a xx r=50 x 50 mesh grid respectively. The results showed that the k-E-kp multi-fluid model can be applied to predict moderate swirling multi-phase flow. When the particle diameter is large, theles with the wall will influence the predictithe particles the stronger the collision with the wall and the more obvious the difference bet ween meaured and calculated resultsKey words: A k-E-kp multi-fluid model Two phase flow, Confined s wiring jetDocument code: ACLC number: TK16INTRODUCTIONcal simulation by using the h-E-hp multi-fluidmodel with the particle phase considered asConfined swirling two phase flows are pseudo-fluid phasd Spalding, 1981)widely utilized in engineering applicationssuch as combustion systems cyclone separa- MATHEMATICAL MODELtors etc. They are used to enhance the flamestability and to mix properly the fudizer in combustion systems. In cyclone sepa-The schematic diagram of the flow confirators they are used to separate particles by uration is shown in Fig. 1. The central primaentrifugal force. In all of those engineeringry jet is loaded with particles and the annularSystems, the behavior of both the particles jet provides s wiring air stream. The 2-d gas-and the air is of great importancesolid two phase flo w model can be obtained onthe basis of the standard k-E model ofStudies were carried out to find the flowphase. The conservation equations ofcharacteristics of the particles and the air model in cylindrical coordinates can be writtenthrough experiment and numerical simulationSommerfeld et al., 19932000;Heas follo wing( Huang et al., 1991)Gas pha2000). In their numericales, the flowcharacteristics of the particles were investigat- ax/e)+/ed by using the Lagrangian approach which a(a1showed that interaction bet ween the particleF)+Sn+S(1)phase and the air phase mainly depended onhe local eddy lifetime and the stokesian reParticle phasesponse time of the particles. More than a中国煤化工100 000 particles were needed to describe the dxCNMHGflow characteristics of the particle phase_aCrowe, 1985). This paper presents nume)+12(n2y)+sx(2)Project supported by Zhejiang Provincial Natural Science Foundation China( No. 598017)A数据epFax:+86-0571-87951358,E-mail:fanir@mailhz.zj.cn15JIN Hanhui xIA Jun et alφ, s. of tIabetheirspesource terms ( STable 1 where s stands for the source termof the fluid phase acting on the particle phaseDf the particleD=70mDenergy of the gas phase, dimension analysis2. Annular swirling flowD,=194mmcan be used to describe the particle phase turFig 1 Flow configuration of swirling gas-solid bulent viscosity y as Y= CHok k2/E. Thetwo phase flowturbulent diffusion of the particle phase is ex-Fick lThe ex pressions of the different variables particle turbulent viscosity and the gradient ofthe particle mean densitTable 1 Expressions of p, r, s in eqs. (1)and(2)[/,][F/FSource term[S +S)S]3()-)r(G1+Gn)+C2np(…2mx)+1npgGWhereu+UT, AT=Cupk/h2/a wtdu dar dGralwd/a中国煤化工刁GnpPIxh-Ch√h)CNMHGr」1万数据((1+R6)1,Rp=n-/p 18Simulating conf ined swirling gas-soild two phase jet159Where up ,u= particle and gas axial mean ve- is very high( Elghobashi et al., 1992, Shilocity i u,, v= particle and gas radial mean ve- et al., 1989), so it was neglected toolocity wparticle andgentialAs conclued above, the reactions bet weenmean velocity ;hp ,k= particle and gas turbu- the two phases considered in the equations arelent energy iyp wue= particle and gas turbulent the drag force and the gravity. The drag forcevIscosIty ,Irparticle stokesian response between the two phases and the gravity aretime ' x,r= coordinates i n.= particle mean added to the source term of the both phasesnumber density subscript p= particlemomentum equations( Mclaughlin, 1994C.=009,C1=1.44,C2=1.92,ak= Stock1995). The dragorce cane ex-l,σε=1.33,σn=0.7p=0.75, Cm= pressed as F0.0064,=18.08×10(Pas)The flow parameters of different cases areshown in Table 2. The boundary conditionsSince the density ratio Pp/p>1000, theBasset force the added force, can be neglectare specified as follows. ThAnd the collisions bet ween the particlescified according to the experimentalcan be ignored because of low particle load- conditions( Sommerfeld et al., 1993). Sym-ng. Since the gradient of the velocity is notmetrical condition at the axis, fully develebig, the Magnus force is neglectedconditions at the outlet and no slip conditionswas concluded that the Saffman force should at the wall are taken for both phases. And thebe considered only when the particle rotation wall function approximations are adopted fornear-wall grid nodesTable 2 Flow parameters of both phasesPhaseFlow conditionFlow paranMass flow rate of the primary jet(g/s)Mass flow rate of the secondary flow(g/s)44,6Swirl numberParticle phaseParticle loading in the primary jet0,17Particle mean diameter (umParticle material der2500could be due to the fact that the turbulenceRESULTS AND DISCUSSIONSanisotropic characteristic becomes most inten-sive at those cross sections. And the standardThe computational results were obtainedk-E model based on the isotropic hypothesisonax×r=50×50 mesh grid. The span ofcould not predict it exactly. Simulations usinmodels based on anisotropic hypothesis or usthe computational domain in the stream wisedirection was 1.0 m so that the outflow con-( direct numerical simulation )may yield moreThe results were compared with published precise resultFigs.6 to 9 show the measurements andexperimental data Sommerfeld et. al calculations when the particle mean diameter1993)to 5 show the measuredcalculated profiles of the gas phase when the was 30um. The agreement was fairly goodes of the exnerimental data col-particle mean diameter was 30um. The a- lapsed中国煤化工 the last cross sgreement for velocity in the three directions tionCN MH Gwere too few partiwas very good except for the axial velocity cles to be detected by LD V( laser Doppler vefluctuation between x= 52 mm and x =85 locimeter ) a gas bubble was observed in themm. The agreement for velocity fluctuationexperiment in this region. For the same reawas reasonably good. The disagreement be- son as that for the gas phase the predictionstween tos measurements and calculations of the axial velocity fluctuation at the core re-160JIN Hanhui xIA Jun et al007500750075mmFig 2 Measured and calculated axial gas velocityFig 3 Measured and calculated axial gas velocityfluctuation&m/sASm/s007500250.075Fig 4 Measured and calculated radial gas velocityFig 5 Measured and calculated tangential gasvelocit00750.025-0075Fig 6 Measured and calculated axial particle velociFig 7 Measured and calculated axial particle velocity 30umty fluctuation( 30um网乃2-0.02500250075Ⅺmm)Fig 8 Measured and calculated radial particle veig.9 Measured and calculated tangential particlelocity( 30umvelocity( 30umgion near the cross section x=85 mm were a very strong( which could be concluded fromlittle lower than the experiments. In some ar- the ldeas, the radial velocity in the experinegative but positive in the predictions. It canH中国煤化工y), and wohCNMHGbe explained that the particles near the wall could also be seen when the mean diameterregion show particle characteristics more in- was 45um( Fig. 10) and 60 um( Fig. 11tensively w hen colliding with the wall. After The bigger the mean diameter, the strongercollision with the wall the particles radial ve- the collisions of the particles with the walllocity becateffegative. The collision was not were. As a result, the difference betweenSimulating conf ined swirling gas-soild two phase jet161measured and calculated radial velocity in to using the model due to the isotropic hthose regions was more obviouspothesis. Further studies should concentrateon establishing a model in5m4anisotropic characteristic of turbulence is corsidered. LES or DNS can be the appropriateway to solve this problemThe differences bet ween measurements003and calculations of the particle radial velocity52852155195315sho w the particle phase has its own particlecharacteristic besides pseudo-fluid characteris-Fig10 Measured and calculated radial particle velocity tic. The pseudo-fluid model of the particlephase can also be improved by considering thearticle characteristics during collision with a00257ReferencesCrowe f. T, Troutt ,TR., Chung ,J N.,1996. N-0025erical models for two-phase flow. Annu. R0.075luid.Mech.,28:149-15812155195315Elghobashi S E., Truesdell G. C., 1992. Direct simulaX(mm)tion of particle dispersion in decaying isotropic turbulence,J. Fluid Mech, 242: 655-700Fig11 Measured and calculated radial particle velocity He, H, Fan J. R. Cen, K F., 2000. A numericalAs shown in Figs. 10 and 1 1 when particlesystem. Journal of Zhejiang University SCIENCE1(2mean diameters were 45 um and 60 um, al- Huang .X.Q., 1989. Study of turbulent gas-particle jetsthough the particles collisions with the waland 3-D turbulent recirculating gas-particle flowswere not considered, the calculated resultstill agree well with the experimental resultsHuang ,X Q., Zhou ,L.x., 1991. Simulation of three-dimensional turbulent recirculating gas-particle flowsIt sho ws that the pseudo-fluid characteristic ofby an energy equation model of particle turbulencethe particle phase is the leading characteristicGas-Solid Flows 121: 261-265in the particles' movement. If the mean di- Jin, HH., 1999. Research of charged gas-liquid twoameter of the particles is not very big and thehase turbulence jet and its application in pest concollisions of the particles with the wall are nottrolling. Master discourse of Jiangsu University ofvery strong, the multi-fluid model can work Mclaughlin F.B., 1994. Numerical computation of par-fairly wellticle-turbulence interaction. Int. J. MultiphaseFlow 20 Suppl): 211-232CONCLUSIONSShi,XG., Xu, X.C., Feng ,JK., 1989. Analysoving in turbulent flow. E7gineering Thermalphysics 103 ): 320-325A multi-fluid model was derived based on Sommerfeld, M., Qiu,H.H., 1993. Characterizationlel. the belof particle-laden, confined swirling flows by phaof both phases in a confined swirling tdoppler anemometry and numerical calculationphase iet as simulated detailedly. The folInt.J. Multiphase Flow 19: 1093-1127Spalding, D.B., 1981. A general purpose computer ro-conclusionsThe numerical simulations for different中国煤化工mml.23:267-276particle diameter showed good agreementStockCNMHGr lecture: Particle diswith the experimental results for both phases. Zha xD,Fan AR,Sun,P,2000.NumericalIt shows the k-E-An model can be applied tolation on dense gas-particle riser flow. Journal ofpredict moderate swirling multi-phang Universily SCIENCE 1(1): 29-38differences bet ween calculation Note: Explanation an apology to our readers: we ll use this articleand expe流落据 show there are some limitsplace that of 2002, first issue p 82-85,JZU(S)