Particle Size Distribution in CFD Simulation of Gas ...

1 Timo NiemiParticle size Distribution in CFDS imulation of Gas- Particle FlowsSchool of ScienceThesis submitted for examination for the degree of Master ofScience in supervisor:Prof. Rolf StenbergThesis instructor:Lic. (Tech.) Sirpa KallioAALTO UNIVERSITYSCHOOL OF SCIENCEABSTRACT OF THEMASTER S THESISA uthor: Timo NiemiTitle: Particle size Distribution in CFD Simulation of Gas- Particle FlowsDate: : EnglishNumber of pages:8+84 Department of Mathematics and Systems AnalysisProfessorship: MechanicsCode: Mat-5 Supervisor: Prof. Rolf StenbergInstructor: Lic. (Tech.) Sirpa KallioFluidized bed combustion (FBC) boilers have become one of the leading technolo-gies in environment-friendly biomass combustion. In fluidized bed combustorsthe solid fuel particles are suspended on upward-blowing air resulting in a tur-bulent mixing of gas and solids. This mixing process allows efficient chemicalreactions and heat transfer which in turn help to reduce emissions and allow toutilize many different types of order to develop even better FBC designs, numerical simulations could be usedto model the flow behaviour inside the boilers.

2 However, the complex nature ofthe multiphase flow of particles and combustion air makes the modelling verychallenging. One of the issues that requires consideration is the size distributionof the particles . Traditionally only the average size of the particles has beenused in the simulations in order to keep them simpler but in reality the sizedistribution also affects the flow behaviour and should be taken into this thesis the relatively recently introduced approaches for Particle size dis-tribution (PSD) modelling in CFD setting are studied. The applicability of theapproaches to fluidized bed Simulation are examined and the computational re-quirements of the approaches are compared. Based on the these factors, currentlythe moment methods and of these especially the DQMOM-method appear to bemost suitable for fluidized bed many of the studied approaches can increase the computational requirementsof the simulations substantially, a new, simplified method to model the PSD is alsointroduced in this thesis.

3 The new method is tested by simulating a small scalefluidized bed and the results are compared to measurements and to the resultsobtained from simulations with a single Particle : Particle size Distribution , fluidized bed, CFB, multiphase, DQMOM,Euler-Euler, CFDAALTO-YLIOPISTOPERUSTIETEIDEN KORKEAKOULUDIPLOMITY NTIIVISTELM Tekij : Timo NiemiTy n nimi: Partikkelikokojakauman huomioiminenkaasu-partikkelivirtausten iv m r : : EnglantiSivum r :8+84 Matematiikan ja systeemianalyysin laitosProfessuuri: MekaniikkaKoodi: Mat-5 Valvoja: Prof. Rolf StenbergOhjaaja: TkL Sirpa KallioLeijupetikattilat ovat yksi t rkeimmist biopolttoaineille soveltuvista katti-latyypeist . Leijupetikattiloissa polttoainepartikkelit leijuvat alhaalta p in tule-van ilmavirtauksen varassa, mik johtaa polttoaineen ja ilman tehokkaaseensekoittumiseen. Hyv sekoittuminen tasaa l mp tiloja kattilassa ja mahdollis-taa p st jen v henemisen ja monipuolisen polttoainevalikoiman k yt kehityksen nopeuttamiseksi olisi hy dyllist , jos kattiloissatapahtuvaa polttoprosessia voitaisiin mallintaa numeerisesti.

4 Partikkeleiden japolttoilman muodostama monimutkainen monifaasivirtaus on kuitenkin hyvinhaasteellinen mallinnettava ja laskentamenetelmi on edelleen kehitett v .Er s ongelma, joka mallinnukseen liittyy on partikkeleiden kokojakaumanhuomioon ottaminen. Perinteisesti partikkelit on mallinnettu pelk st n niidenkeskikoon avulla mallinnuksen yksinkertaistamiseksi, mutta my s kokojakaumatulisi ottaa huomioon, koska se vaikuttaa virtauksen k ytt ss diplomity ss tutkitaan viime aikoina kehitettyj menetelmi partikke-likokojakauman mallintamiseen monifaasivirtauslaskennassa. P asiallisenavertailukohtana eri menetelmien v lill k ytet n sek niiden soveltuvuutta lei-jupetien mallintamiseen ett menetelmien vaatimaa laskentaty n m r . Ver-tailun perusteella momenttimenetelm t ja n ist erityisesti DQMOM-menetelm vaikuttaa t ll hetkell soveltuvimmalta l monet tutkituista menetelmist lis v t tarvittavan laskentaty n m r merkitt v sti, t ss ty ss kehitet n my s vaihtoehtoinen, kevyt menetelm kokojakauman menetelm testataan simuloimallapienen kokoluokan leijupeti ja saatuja tuloksia verrataan mittauksiin ja yhdell partikkelikoolla laskettuihin : partikkelikokojakauma, leijupeti, CFB, monifaasi, DQMOM,Euler-Euler, CFDivPrefaceThis Master s thesis has been written at VTT Technical Research Centre of Finlandas a part of Tekes project: CFD based on-line process analysis - applied to circulat-ing and bubbling fluidized bed processes (OnlineFB-CFD).

5 I gratefully acknowledgethe financial support of Tekes and all the industrial partners which are involved inthe would like to thank Sirpa Kallio for her excellent guidance and feedback through-out this work and for introducing me to the world of multiphase flow modelling. Iam also grateful to Juho Peltola and other colleagues here at VTT for sharing theirknowledge in many aspects related to this work. The experimental measurementsused in this thesis were conducted by Markus Honkanen. Alf Hermanson helpedto operate the experimental devices and has constructed much of the would like to thank my supervisor Rolf Stenberg for proof-reading this thesis andfor the general guidance during my , I would also like to thank my parents for their encouragement, my friendsfor bringing balance to work and especially I would like to thank Ella, for her loveand , NiemivContentsAbstractiiAbstract (in Finnish)iiiPrefaceivContentsvSymbols and abbreviationsvii1 Background.

6 Goals and Objectives .. Structure of the Thesis ..22 The fluidization phenomena .. Applications of fluidization .. Particle classification .. Minimum fluidization velocity and terminal velocity .. Effect of the Particle diameter and the size Distribution ..123 Multiphase Flow Introduction .. Eulerian-Eulerian approach .. Gas-solid drag models .. Eulerian-Lagrangian approach .. The dense DPM approach .. Turbulence modelling .. Time averaged modelling ..254 Modelling The population balance equation .. The class methods .. The quadrature method of moments .. The direct quadrature method of moments .. The PD algorithm ..395 The general idea of the approach .. The mixture formulation .. Volume fraction corrections .. Implementation details ..506 Introduction .. Geometry description.

7 Boundary conditions .. Used models and solution strategies .. Experimental measurements .. Validity study of the local equilibrium approximation .. Case 1: Binary mixture .. Case 2: Wide size Distribution .. The effect of the PSD ..757 Conclusions77 References79viiSymbols and abbreviationsLatin SymbolsAAreaCDDrag coefficientdsParticle diameteressRestitution coefficientfNumber density functiongGravitational accelerationg0,ssRadial Distribution functionIIdentity matrixI2 DSecond invariant of the deviatoric stress tensorKgsInterphase drag coefficientLParticle length, ie. Particle diameterpPressureReReynold s numberSSource termUmfMinimum fluidization velocityVVolumevVelocity vectorwQuadrature weightGreek Symbols Volume fraction The Dirac delta function qsKronecker delta Voidage Bulk viscosity Viscosity Sphericity fAngle of internal friction Density Stress tensor Granular temperature Internal coordinate, quadrature abscissaviiiSubscriptscolCollisionalfFlu idfrFrictionalgGaskinKineticktgfKinetic Theory of Granular FlowmfMinimum fluidization statepParticleqPhase indicatorsSolid.

8 Solid mixturetTerminalAbbreviationsBFBB ubbling Fluidized BedCFBC irculating Fluidized BedCFDC omputational Fluid DynamicsCMClass methodsCPFDC omputational Particle Fluid DynamicDEMD iscrete Element MethodDPMD iscrete Particle ModelDQMOM Direct Quadrature Method Of MomentsFBCF luidized Bed CombustionKTGFK inetic Theory of Granular FlowMOMM ethod Of MomentsMP-PICM ultiphase Particle -in-Cell methodMUSIGM ultiple size Group modelN-SNavier-Stokes (equations)PBEP opulation Balance EquationPDProduct-Difference (algorithm)PSDP article size DistributionQMOMQ uadrature Method Of MomentsSMDS auter Mean DiameterTFMTwo Fluid ModelUDFUser Defined FunctionVOFV olume Fraction1 BackgroundSince their introduction in the 1920s, fluidized bed processes have become an impor-tant and widely used technology in chemical and metallurgical industries as well asin power generation. These industrial processes are typically large and complex andthe prototyping and development of new systems is expensive and time order to make the development process more efficient, computational simulationof fluidization has gained interest.

9 The computational fluid dynamics (CFD) ap-proach has emerged as a popular tool in assisting the development of fluidized flows in fluidized beds are inherently multiphase consisting of at least one fluidphase and one particulate solid phase. The CFD modelling of single phase flows isalready quite a challenging task and the multiphase aspect only adds in modelling of dense particulate flows is especially complicated since the flowbehaviour in these cases is quite different from the ordinary fluid flows. Thesechallenges make CFD Simulation less straightforward to apply, but with propersimulation models good results can still be achieved. However, careful validation ofthe models and comparisons with experimental data are still very of the key problems in particulate flow modelling is the Particle size distri-bution (PSD). In virtually all real life situations the particles in fluidized beds arepolydisperse.

10 Typically also the chemical reactions and physical processes insidethe reactors constantly affect the sizes of the particles . The shape and the prop-erties of the Particle size Distribution can have dramatic impact on the behaviourof the particulate flow. Thus, in order to accurately model a fluidization process,the Particle size Distribution and the size change mechanisms have to be taken intoaccount in the computational the Particle phase has been modelled by using only one, mean particlesize disregarding other important characteristics of the Particle size has been done for efficiency reasons as the modelling of the size distributioncan be computationally intensive. At present the increased computational powerand improved modelling methods have made more accurate PSD modelling subject is, however, still under active research and there are various alternativeapproaches available. In this master s thesis the issue of modelling Particle sizedistributions in gas- Particle flows is Goals and ObjectivesThe goal of this thesis is to implement a method for fluid dynamic modelling ofparticle size Distribution in a fluidized bed reactor.