• Aucun résultat trouvé

The effect of column tilt on flow homogeneity and particle agitation in a liquid fluidized bed

N/A
N/A
Protected

Academic year: 2021

Partager "The effect of column tilt on flow homogeneity and particle agitation in a liquid fluidized bed"

Copied!
12
0
0

Texte intégral

(1)



$Q\FRUUHVSRQGHQFHFRQFHUQLQJWKLVVHUYLFHVKRXOGEHVHQWWRWKHUHSRVLWRU\DGPLQLVWUDWRU

WHFKRDWDR#OLVWHVGLIILQSWRXORXVHIU

2

SHQ

$

UFKLYH

7

RXORXVH

$

UFKLYH

2

XYHUWH

2$7$2



2$7$2 LV DQ RSHQ DFFHVV UHSRVLWRU\ WKDW FROOHFWV WKH ZRUN RI VRPH 7RXORXVH

UHVHDUFKHUVDQGPDNHVLWIUHHO\DYDLODEOHRYHUWKHZHEZKHUHSRVVLEOH

7RFLWHWKLVYHUVLRQ











2IILFLDO85/

7KLVLV

an



author's

 YHUVLRQSXEOLVKHGLQ

http://oatao.univ-toulouse.fr/20380

http://doi.org/10.1016/j.ijmultiphaseflow.2017.02.008

Aguilar-Corona, Alicia and Masbernat, Olivier and Figueroa, Bernardo and Zenit, Roberto The effect of column tilt on

flow homogeneity and particle agitation in a liquid fluidized bed. (2017) International Journal of Multiphase Flow,

92. 50-60. ISSN 0301-9322

(2)

The

effect

of

column

tilt

on

flow

homogeneity

and

particle

agitation

in

a

liquid

fluidized

bed

A. Aguilar-Corona

a

,

O.

Masbernat

b

,

B.

Figueroa

c

,

R.

Zenit

d,∗

a Facultad de Ingeniería Mecánica, Universidad Michoacana de San Nicolás de Hidalgo, Francisco J. Mujica s/n C.P. 58030, Morelia-Michoacán, México b Laboratoire de Génie Chimique, Université de Toulouse, CNRS/INPT-UPS, 4, allée Emile Monso BP 44362, 31030 Toulouse Cedex 4, France

c Laboratorio de Ingeniería y Procesos Costeros, Instituto de Ingeniería, Universidad Nacional Autónoma de México, Puerto de Abrigo S/N, Sisal, Yucatán

97355, México

d Instituto de Investigaciones en Materiales, Universidad Nacional Autónoma de México, Apdo, Postal 70-360, México D.F., 04510, México

Keywords:

Agitation Liquid fluidized bed Homogenization Solid fraction Column tilt

a

b

s

t

r

a

c

t

Themotionofparticles inasolid-liquidfluidizedbedwasexperimentallystudiedbyvideotrackingof markedparticlesinamatchedrefractiveindexmedium.Inthisstudy,twofluidizedstatesarecompared, onecarefullyalignedintheverticaldirectionensuringahomogeneousfluidisationandanotheronewith anon-homogeneousfluidisationregimethatresultsfromaslighttiltofthefluidisationcolumnof0.3° withrespecttothevertical.Asaresultofthemisalignment,largerecirculationloopsdevelopwithinthe bedinawell-definedspatialregion.Itisfoundthatinthatrangeofsolidfraction(between0.3and0.4), theinhomogeneousmotionoftheparticlesleadstosignificantdifferencesinvelocityfluctuationsaswell as inself-diffusioncoefficientoftheparticles intheverticaldirection, whereasthe fluidisationheight remainsunaffected.Atlower(lessthan0.2)orhigher(higherthan0.5)concentration,particleagitation characteristicsarealmostunchangedintheverticaldirection.

1. Introduction

Thestudyofliquidfluidisationissignificantfromboth theoret-ical andpractical viewpoints. It allows forthe experimental ver-ification of two-fluid modelling concepts in gravity driven solid-liquidflows, wheretheslipvelocity isofthesameorderof mag-nitudeasthatofthecontinuousphaseandwheresolidphase agi-tationisinducedbycollisionalandhydrodynamicinteractions(the contribution dueto thecontinuous phaseturbulencebeing negli-giblecompared totheaforementionedinteractions).Evenifa liq-uidfluidizedbeddoesnotexhibit chaoticmixingorsharpregime transitions(asinthecaseofbubblyflows),astateofhomogeneous fluidisationisseldomobserved.

The concept ofa homogeneous bed is usually referred to the appearance of solid fraction fluctuationswhich are observable at macroscopic scale, characterized by particle-free regions or voids ofdifferentshapes:dependingonthefluidisationconditions,a flu-idizedbedcan destabilizeinthesense oftheappearanceof one-dimensionaltravelingwaves(ODTW)(AndersonandJackson,1969) whichmaylaterdevelopintotwo-dimensionalstructuresand tran-sitionstootherflowregimessuchastheformationofbubbles,and

Corresponding author.

E-mail address: zenit@unam.mx (R. Zenit).

structuresthatremindobliquewaves(DidwaniaandHomsy,1981;

Duruand Guazzelli,2002; El-Kaissy andHomsy, 1976; Homsyet al.,1980).Therearemanyinvestigationsthattrytomapsuch tran-sitionsthroughexperimental,theoreticalandnumericalsimulation intheliterature,thereadermayreferto(DiFelice, 1995; Homsy, 1998;Sundaresan,2003)foramorecomprehensivereview.

Inthisworkwestudyhomogeneityfromadifferentviewpoint; the column is (or is not) homogeneous in terms ofthe absence oflargescalerecirculationandthebreakofsymmetryoftheflow velocityfield interms ofsome statisticalparametersthat charac-terizesitinspaceorintime(Gordon,1963;Handleyetal., 1966). Thehomogenizationconceptisnoteasilydefinedbecauselow fre-quency motion is always present due to confinement (Buyevich, 1994). Thiseffect is evident when the particlemotion is studied forrelativelylongtime,so theparticleshaveenough timeto tra-verse many times the column length and diameter. Two dimen-sionalanalysesoftenturnouttobeinsufficientbecausethesystem isnot guaranteedto be symmetricandtheformation oflow fre-quencystructures maynot be observedfroma givenobservation direction. This iswhy a full three dimensional analysisis funda-mentallyimportantto characterizea liquidfluidisation system. It alsoallowsfordirectcomparisonswithnumericalresults.

The effect of inclination on a fluidized bed is of particular relevance in the context of this study. It has been investigated from several viewpoints in the literature: for gas-solid

(3)

fluidisa-tion one canfindmany previous relevantworks:in particular,in

Yamazaki et al.(1989) the minimum fluidisation velocity for in-clined columns was studied. The authors observed three distinct flow regimes (fixed bed, partially fluidized and completely flu-idized), andnoted the effectof columninclination on the corre-spondingregimetransitions.Yakubovetal.(2005) studiedthe ef-fectofinclinationofaliquid-solidfluidizedbedonseveralworking parameterssuchascriticalflowrate,bedheightanddynamic pres-suredrop.Theyobservedapatternofconcentrationwaves(forthe effectofinclinationinthecaseofcohesivepowders,see(Valverde etal.,2008)).Numericalinvestigations havealsobeencarriedout onthesubject;inChaikittisilpetal.(2006)DiscreteElement Sim-ulations (DEM) were used to study gas-solid two-phase flow, in orderto investigatethe mixingbehaviorofthesolid phasein in-clined fluidized beds. A large scale recirculationpattern was ob-served. Low concentration bubbles tend to move upwards along the uppermost wall, contrary to the particles that moved down-ward,closertothewallbelowit,enhancingbackmixing.This be-havior has also been observed experimentally in gas-liquid bub-blyflows, dueto buoyancy,forvery smalltiltangles(Zenit etal., 2004).

Forliquid-solidfluidizedbeds,littleattentionhasbeenpaidto the effect of inclination on the columnhomogeneity. Hudsonet al.(1996)usedsalttracermeasurementstoconcludethatfluidized bedinclinationstronglyaffectsthecolumnhydrodynamics. More-over, inDel Pozo etal.(1992) itwas shownthat a smalltilt an-gleof1.5°onathree-phasefluidizedbedaffectstheparticle-liquid mass and heat transfer coefficientssignificantly. Other important aspectinthecomplexinteractionbetweensolidandliquidphases istheeffectoftheinclinationangleonmixinganddiffusion,asa functionoftherelevantparameterssuchastheReynoldsnumber, Stokes number, andparticle-column width (orheight) ratio. Sev-eralstudieshavebeendevotedtothediffusioninaliquid-solid flu-idizedbed(Al-DibouniandGarside,1979;CarlosC.andRichardson J.,1968;Dorgeloetal.,1985;JumaandRichardson,1983;Kennedy S. andBretton R.,1966;Van DerMeeretal.,1984; Willus, 1970). Twotrendscanbeidentifiedintheliteratureconcerningdiffusion: firstly, withrespecttosolid concentration,andsecondly,with re-spectto theparticle-to-columndiameterratio.Insome investiga-tions (CarlosC. andRichardson J., 1968; Willus, 1970; Dorgeloet al.,1985)itisfoundthatthediffusioncoefficientdecreasesassolid fractionincreases.Ontheotherhand,otherstudiesfromthe liter-ature(Kangetal.,1990;Yutanietal.,1982)foundasmallpeakon theauto-diffusioncoefficientasthesolidfractionconcentration in-creases.Concerningtheparticlesizeratio,theexperimentsshowed that diffusiondecreases astheparticlesize ratioincreases.Those experimentswere carriedout fordifferentflowregimes compris-ing superficial Reynolds numbers of O(10–1000) and two-phase StokesnumbersofO(1–10).Althoughtherearemanyinvestigations devoted tothe effectofinclination onliquid-solid fluidizedbeds, none ofthe aforementioned worksremarked the highsensitivity ofthefluidized bedcharacteristics toasmallinclination; mostof those studies comprisedranges ofinclination betweenhorizontal tovertical,butincrementedthetiltinlargesteps,ignoringthe ef-fectsofverysmallinclinationangles.

Thisworkisdevotedtostudytheeffectofasmalltiltofthe flu-idisationcolumn(0.3°withthevertical),comparedto avertically aligned column. Low frequencystructures are detected andtheir effectonthedispersedphasevelocityisassessedthroughthe anal-ysisof:a)Theparticletrajectories,b)Thespatialdistributionofthe verticalspeed,c) Theparticlevelocityvariancesandd)The diffu-sioncoefficient.Thetechniqueusedtocalculatethemeanvelocity andagitation(velocityvariances)alongthethreedirectionsis sim-ilartothatusedinHandleyetal.(1966)andCarlosandRichardson (1968) and later revisited in Buyevich (1994), Willus (1970) and

Latif andRichardson(1972), who useda Lagrangiantracking ofa

colored particleinthebulkofa transparentbed.Morerecentlya similarparticletrackingtechniquewasusedbutinacarefully con-trolledopticallymatchedsystem(Aguilar,2008;AguilarCoronaet al., 2011;HassanandDominguez-Ontiveros,2008).Acamerawith highresolutionwas used(bothintimeandspace),whichallowed for the determination of detailed information about the particle phasemotionwithinthefluidizedbed.

2. Experimentalset-up

The experimental device isshownschematically inFig. 1.The fluidisation section iscomposed ofa 60cmhighcylindricalglass columnof8cminnerdiameter.Aflowhomogenizer,consistingof a fixed bedof packedbeads covered by syntheticfoam layers,is mounted atthe bottom ofthe columnto ensurea homogeneous flowentry.Theflowtemperatureismaintainedat20°Cbya con-trolled heat exchanger. Twoparticular cases were studied during thiswork: 1) Averticallyalignedcolumnand2)Atilted column, forminganangleinthe(y,z)planeof0.3°withrespecttothe ver-ticalaxisz.ThereferenceframeisshowninFig.2.

2.1. Particlesandfluid

Calibrated 3mmpyrexbeadswere fluidizedby aconcentrated aqueous solution of Potassium Thiocyanate (KSCN, 64%w/w). At 20°C,thefluidandtheparticlesandfluidhavethesamerefractive index(∼1.474),sothatatagged(colored)particlecouldbetracked individually inanearly transparentsuspension (Aguilar-Corona et al.,2011).ParticleandfluidpropertiesarereportedinTable1.The particleStokesandReynoldsnumbers,basedontheterminal (sed-imentation)velocity,areSt=4.8andRe=160,respectively.

2.2. Particletrackingtechnique

The analysisof particle motion in the fluidized bedwas per-formed by means of highspeed 3-D trajectography. The fluidisa-tioncolumnisequippedwithanexternalglassboxfilledwiththe aqueousphaseinordertoreduce opticaldistortion(seeFig.2).A mirror orientedat45°totheside ofthebox allowed forthe ob-servationoftheparticlepathinthreedimensions,providingan ad-ditionalside view.APhotronAPXcameraequippedwithaCMOS sensorwasusedtorecordthefront((x,z)plane)andthesideview fromthemirror((y,z)plane)inasingleframe(512pix×1024pix). Images were recorded over periods of 204 seconds, starting af-terthestationaryregimehadbeenreached.Takingacharacteristic particlevelocityof3cm/s(fromthestandarddeviationofthe par-ticlevelocityofatypicalexperiment),thistotaltimewould repre-sent morethan 70timesthetimea particlewouldtake totravel one columndiameter.The averageresidencetime(the averageof the time it takes to a particle to travel one column height) for thealignedcasewas 6seconds,whilethecorrespondingvaluefor the tilted case was 4.5s, so one canexpect the average absolute speedtoincreasewithinclination.Ablackcoloredparticlewas in-troducedinthe bedandits trajectory wasrecorded at60frames persecond(fps).

Fig. 3showsboth front (x-z plane)andside (y-zplane) views as captured by the camera,for solid fractions of

h

α

p

i

=0.50 and

h

α

p

i

=0.14 (sub-figures (a) and (b), respectively). All the experi-mentswerecarriedoutwithaparticlesizeofdp=3mm.Theblack linebetweentheimagesis justthe spacebetweenthe frontwall and themirror, whichwas maskedin orderto avoida confusing view oftheadjacentwallofthecolumn.Theimage fromthe mir-ror hadaslightlydifferentscaleduetotheoptical pathsbetween the (direct)front view andthat comingfromthe mirror,so each planehaditscorresponding(horizontalandvertical)scalefactorin

(4)

Fig. 1. Scheme of the experimental set-up and the entry section.

Table 1.

Fluid and particle properties at 20 °C.

Pyrex beads dp = 3mm ρp = 2230 kgm −3 nD = 1.474 KSCN solution 64% w/w µf = 3.8 × 10 −3 Pa s ρf = 1400kgm −3 nD = 1.474

Fig. 2. Top view scheme of the trajectography 3D system.

ordertoobtaintheactualpositionincentimeters afterimage dig-italprocessing.Notethatthecoloredparticleisclearly discernible evenwhen itislocateddeepinsidethebulk ofthecolumn(even forlargesolidfraction).ItcanbeseenfromtheFigure(d.1andd.2) thattheparticlediameteroccupiesapproximately13to15pixels.

3. Results

3.1. Globalsolidfraction

The initialvolume ofparticlesinthefixed bedcorrespondsto an initial height h0 of9.5cm,slightlylarger thanthe column

di-ameter. The maximumcompactness concentration

h

α

c

i

was esti-matedtobe0.56,correspondingtoarandompacking.Eventhough thesystemisopticallyhomogeneous,beadsinterfacesarestill de-tectable;acarefulobservationoftheimagesallowedforthe deter-mination ofthemaximum height reachedby theparticles hb for eachsolidfraction.Themeansolidfractionwasobtainedas:

h

α

p

i

=

h

α

c

i

h0

hb

. (1)

The bracket symbol represents the time and space averaging over the bed volume of the local instantaneous solid fraction. It wasobservedthat thehbfluctuationsdecreasedasthesolid frac-tion increased, with a relative error of lessthan 5% in all cases. Inthiswork, fivefluidisation velocitiesweretested (0.095,0.078, 0.053,0.038 and0.02m/s) correspondingto globalsolid fractions

h

α

p

i

of0.14,0.2,0.3,0.4and0.5respectively.Forthehomogeneous case,the fluidisation velocityUF (fluidvelocity inempty column) wasfoundtobeadecreasingpowerlawofvoidfraction:

UF=0.145

(

1−

h

α

p

i

)

2.78 (2) where the prefactor is close to the particle terminal velocity (0.135m/s) andtheexponentvalue, n=2.78, iscloseto that pre-dictedbyRichardson-Zaki’scorrelation(n=4.4Ret−0.1=2.67).

A first visual observationindicates that the slight tilt didnot haveanymeasurableeffectonthebedexpansion,soforeach flu-idisation velocity studied, the global solid fractionremained un-changedinbothhomogeneous (verticallyaligned) and inhomoge-neous(tiltedcolumn)cases.Thisobservationisconsistentwiththe averaged momentum balance in the bed volume. At first order, theeffectoffluidandparticlefrictionatthewallbeingneglected, thisbalancereducestoequilibriumbetweenbuoyancyforce term anddragforce termbaseduponmeanslipvelocity, i.e. themean liquidvelocity. Theaveragesolid fraction, orequivalently thebed height,resultsfromthisbalance.However,whenspatially

(5)

averag-Fig. 3. Raw images as captured by the camera: a) Large solid fraction: < αp > = 0.5; to the left of the black division: front view ( x-z plane). To the right of the division is the lateral view ( y-z plane). b) Moderate solid fraction: < αp > = 0.14 (same views as in (a)); c) Particle close-up; c.1) front view and c.2) side view; d) Binarized particle image. (d.1) front view and (d.2) side view; e) Centroid detection: (e.1) and (e.2) correspond to front view and mirror image, respectively.

ing thelocal two-phase momentumtransport equation, two con-tributions arisingfromthefluctuatingmotionoftheparticlesand the fluid need alsoto be considered:one is the non-linear drag forcetermthroughthevelocityfluctuationsandthesecondisthe cross-correlationbetweenthespatial fluctuationsofsolidfraction andpressuregradientinthebed.Itcanbeshownthatinallrange offluidisationvelocities,thefirstcontributionisalwayslargerthan thesecond one,whichroughlyscales asfew percentofthemean drag term. As a consequence, in a homogeneous liquid fluidized bed,thebedheightisweaklydependentuponphaseagitation.In areference framewheretheaxialdirection istheaxisofthe col-umn,thebuoyancyforcecomponentis

1

ρ

gcos

θ

whereisthe an-gle with the vertical (0.3°), so therelative variation of thisterm is of order of 10−5 and can be neglected. Therefore, tilting the

columna smallangle(0.3°)will notmodify thebedheight,even though thisperturbationinduces importantflow inhomogeneities and significant variations of particle fluctuations, as discussed in thenextsections.

3.2. Particletrajectories

Fig. 4 shows particle trajectories recorded at three different concentrations. The left panel of the figure shows the trajecto-ries projection in thehorizontal(x,y) plane ofthe column, while therightpanel displaysthe projectionsinthe vertical(x,z) plane. For moderatesolid fractions (<

α

p>=0.14and<

α

p>=0.20)the particle path was observedto span the wholebed volume with-outexhibitingclearcoherentstructures.Atlargersolidfractions,a toroidalstructurewasobservedinthelowerpartofthebed,along with a corresponding increase ofthe low frequency fluctuations. The originofthissteadystructurehasnotbeenclarifiedyet.One possibleexplanationcould bethat duetoawall effect:a slip ve-locitydifferencebetweenthemiddleandthenear-wallregion de-velops intheentrysection,resultinginasolid fractionhorizontal gradient.Thissolidfractiongradientwouldtheninducea horizon-talpressuregradientthatwouldgeneratethisrecirculationpattern. But such a mechanismneeds a more indepth analysis, which is beyondthescopeofthispaper.

Fig.5 showsa comparisonbetweentheparticletrajectoriesof theverticalcolumnandthetiltedone, correspondingtothe(x,z), (y,z) and(x,y) planes fora solid fractionof <

α

p>=0.30. In the tilted columncase(forthatconcentration)awell-defined recircu-lationloop inthe(y,z) planewas observed,wherethetracer par-ticle trajectory forms an annulus. For thesame case, inthe (x,z) plane the particle path spans across the whole column volume

without any preferential motion of the dispersed phase. Inclina-tioninducesabuoyancyforcecomponentnormaltothewall; how-ever,thecounterbalanceofthisforcecannotbereadilyidentifiedif therearenosignificantchangesinconcentrationorvelocity. There-fore, thissmall imbalancemay generatea radial drift velocity at thescaleofeachparticle.Nowasthisdriftvelocityislikelyto in-duce a radial concentration gradient, collectiveeffects (such asa radial apparentdensitygradient) areprobablydrivingthe recircu-latingmotionatthebedscale thatis observedontrajectory pat-terns(similarinthatsensetotheso-calledBoycotteffect).

3.3. Testofhomogeneity

In order to characterize fluidisation homogeneity, for each mean fluidisation velocity studied,the spatial distribution of up-ward and downward particle motion was analyzed in four dis-tinct cross sections Si, (i=1 to 4) regularly distributed along the bed height (0≤zS1≤0.25hb; 0.25hb<zS2≤0.50hb; 0.50hb<

zS3≤0.75hb;0.75hb<zS4hb),asschematizedinFig.6.

Fig. 7shows thevelocity signdistributions following particles trajectoriesineachtestsectionSi,foraglobalsolidfractionof0.3. Bothaligned(leftcolumn)andtilted(rightcolumn)casesare dis-playedinthisfigure.Thedirectionofthemotionisindicatedwith a crosssymbol if the particlemoved downwards ora circle ifit movedupwardsasitcrossedtheplaneSi.Fortheverticallyaligned casethesignatureofan axisymmetrictoroidalmotionatthe low-ermostpartofthebedcanbe identified,withapreferential con-centration ofascending velocities atthe centerof thebed cross-section,anddescendingvelocityinthenear-wallregion.Inthe up-permost section, the distribution appears homogeneous over the cross-section. Forthetilted casethere isapreferential motionin all test sections,whichconsistsofa large-scalerecirculationover thewholebedvolume,wheretheparticletendstoriseinone half-sectioninFig.7-(ii),andtodescendintheotherone.Notethatthis motionisquiteparalleltothe(y,z)plane asexpected. Theseplots clearly demonstratetheeffectofthetiltontheparticlemotionin thefluidizedbed.

In order to quantify homogeneityin the cylindricalgeometry, the (circular) cross-section Si was divided into 12 sectors of 30° eachintheangulardirection.Theprobabilityofparticlecrossingin aparticularsectorjwithascendingverticalmotionwascalculated as:

ϕ

up,j=

nup,j

nj

(6)

Fig. 4. Particle trajectories projection for the vertically aligned case at a) < αp > = 0.14, b) < αp > = 0.20 and c) <αp > = 0.40: left and right columns correspond to the horizontal ( x,y ) and vertical ( x,z ) planes, respectively.

withnjthetotalnumberofparticlecrossinginsectorj.This prob-abilityismutuallyexclusivewithrespecttoitscounterpart (down-wards crossing)

ϕ

down,j. A perfectly homogeneous columnwould attainavalueof

ϕ

up, j=0.5forallsectors(j=1to12inthiscase), which ina polarrepresentations wouldgive a circleof radius ½.

Figs. 8and9showthistypeofrepresentation,forthecaseofthe

alignedandtiltedcolumns,respectively. Thedotted circlehas ra-diusr=0.5,forcomparison.

Fig.8showsanangulardistributionclosetohomogeneous for

S4,whilethereseemstobemoredownwardsmovingparticlesfor

S1. Thereare smalldeviationsfrom½forS2 andS3.Fig.9shows thedistribution

ϕ

up,jforthetiltedcolumn.Itcanbeobservedthat

(7)

Fig. 5. Particle trajectories projection at < αp > = 0.30 in a) ( x,z ) plane; b) ( y,z ) plane; c) ( x,y ) plane. Left and right columns correspond to the aligned and tilted cases, respectively.

thedistributionisclosetothecenterforthefirstthreesectors,and morehomogeneous inS4. Howeverthereisstill atrend,showing

ϕ

up,j> 0.5inthesecond quadrant,withvaluesbelow0.5forthe third andfourthquadrants(values closertoone meanthat more particlesmove upwards,consistentwithFig.7(ii)).Iftheparticles were less dense than the liquid,the particles would descend (in average)onthefirstandthirdquadrants.

Another way to represent Figs. 8 and 9 is to plot the angu-larstandard deviationof

ϕ

up,j ineach sector forthe verticaland tiltedcases.ThisquantityisplottedalongbedheightinFig.10for

the verticalandtiltedcases.Anincrease bya factorcloseto 3or 4 of thereference values inthe vertical casecan be observed in the tilted case. Note there isa correspondence betweenthe four points inFigs. 10and9forS1,S2,S3 andS4.Themore heteroge-neousdistributionoftheverticalvelocitycomponentwasobserved at S2 and S3, where the value of

ϕ

up,j is very small in the 3rd

and4thquadrants,indicatingdownwardsverticalmotioninthese quadrants.ThemosthomogeneoussectionwasS4,consistentwith

Fig. 9, where

ϕ

up,j is close to 0.5 in the 3rd and 4th quadrants.

(8)

s

1

s

2

s

3

1h

b

s

4

0.75

h

b

0.50

h

b

0.25

h

b

Fig. 6. Bed test cross sections S 1 , S 2 , S 3 and S 4. .

Consequences of column tilt-induced flow inhomogeneities upon particleagitationareexamined inwhatfollows,by comput-ingtheparticlesvelocityvarianceandself-diffusioncoefficientand comparingtheirintensitieswiththoseofthealignedcase.

3.4. Effectofcolumntiltingonparticlevelocityvariance

The varianceoftheithparticle velocitycomponentinthebed iscomputedas:

D

u′2 pi

E

=

D

¡

upi

(

t,x

(

t

)

)

­

upi

®¢

2

E

(6)

whereupi(t,x(t))istheinstantaneousvelocityi-component follow-ing particle trajectory x(t), and the bracketsymbol denotes here theaverageofparticlevelocityith-componentoveralltrajectories (equivalent to an ensemble average operator). Note that

h

upi

i

is closetozero,theaverageparticlevelocityinthebedbeingzerofor a steadyfluidizedbed(sou

piisvery closetoupi). InFig.11,the variance ofeach velocity componentis reportedas a function of globalsolidfraction,inbothhomogeneous(verticallyaligned)and inhomogeneous (tilted) cases. In both cases, particle agitation is stronglyanisotropicasexpectedingravity-driventwo-phaseflows. In thehomogeneous case, theaxial componentofparticle ve-locityvariance(z-component)beingabout2timeslargerthanthe components in the horizontal plane (x,y) for all concentrations. Particle velocity variance isa continuouslydecreasing functionin therangeofconcentrationinvestigated[0.14–0.5],withaprobable maximumlyingintherange[0–0.14].

In the non-homogeneous case, the evolution of the axial ve-locity variance is quite different compared to the homogeneous case, mainly in the range ofsolid fraction[0.3–0.4]. At the low-est concentration (<

α

p>=0.14), the axial velocity variance is smallerthaninthehomogeneouscase,thenabruptlyincreasesup to a maximum at <

α

p>=0.3, then strongly decreases between <

α

p>=0.3 and 0.5. This evolution results fromthe progressive development of the large-scale loop induced by the column tilt as the particle concentration increases. In the horizontal plane, particle velocity variance is a continuous decreasing function of solid fraction, slightlybelowthehomogeneous casevaluesinthe range [0.14–0.3] with a similar behavior at higher concentration (<

α

p>=0.5).

Table2reportstherelativedifference betweentheparticle ve-locity component variances forthe homogeneous (noted

h

u

2pi

i

H) andnon-homogenous (noted

h

u

2pi

i

H) fluidisation cases,measured

Table 2.

Relative difference δupi of h u ′2piiH between homogeneous and inho- mogeneous cases. <αp > δux δuy δuz 0 .14 0 .12 0 .22 0 .15 0 .2 0 .10 0 .16 −0 .015 0 .3 0 .20 0 .08 −0 .59 0 .4 0 .18 −0 .04 −0 .6 0 .5 −0 .08 −0 .1 0 .09

atfivedifferentglobalsolidfractions

h

α

p

i

,anddefinedas:

δ

upi=1−

D

u′2 pi

E

nH

­

u′2 pi

®

H (7)

Positivevaluesof

δ

upiindicatethattheith-componentvelocity varianceinthenon-homogeneous caseissmallerthanthatofthe homogeneouscasewhereasnegative valuesof

δ

upireveal the op-positetrend.Notethatinthehomogeneouscase,thevelocity vari-ance inxandy-direction shouldbe equal.The relativedifference betweenthesevaluesisinaverageoftheorderof0.05forall con-centrations,so the relative difference betweennon-homogeneous andhomogeneouscaseisconsidered assignificantwhenitsvalue exceeds0.1.Largenegativevaluesareobservedfortheaxial com-ponents of the variance, reaching

δ

upz=0.6 at<

α

p>=0.3 and 0.4, which confirms the predominance of large-scale motions in thatrange ofconcentration.In thehorizontal(x,y) plane, the rel-ative difference is smaller than in the homogeneous case when <

α

p>≤0.3. At the highestconcentration (<

α

p>=0.5), the dif-ferencebecomesslightlynegative.

If, in the non-homogeneous case, particles are globally accel-erated in the vertical direction by a large scale motion induced bycollective effects,thenit canbe understoodthat velocity fluc-tuations in the transverse directions will diminish in the ascen-dantand descendantparts ofthe loop, andincrease in the hori-zontalpart.Inaverage,thetransversecomponentvariancewill de-crease,probablybecausetheweightoftheascendingand descend-ingpartsisstrongerthanthatinthehorizontalplane.Athigh con-centration(0.5), thesignofthecriterionisreversed,likelydueto anaspectratioeffect(inthiscase,theheightofthebedisindeed closetothecolumndiameter).Intherange[0.14,0.2],the concen-tration seem tobe too smallto induce a large recirculationloop inthebed,butanon-zerotransversecomponentofbuoyancystill existsandisabletodamp inthehorizontalplane thefluctuating motionofparticlesproducedbythemeandragforce.

3.4.Effectofcolumntiltingonparticlediffusioncoefficient

Particle diffusioncoefficient is determined fromthe computa-tionof Lagrangianvelocity autocorrelationcoefficient, definedfor eachvelocitycomponentupias:

Rii

(

t

)

=

­

upi

(

τ

)

upi

(

τ

+t

)

®

­

u2pi

(

τ

)

®

(8)

In the range of globalsolid fraction investigated, particle La-grangianvelocity decorrelateswithin atime intervalsmallerthan 4seconds,asillustrated in Fig.12. Thecurves insuch figure can be fitted to a decaying exponential of the form Rii(t)=exp(-bt); Thefitted curve hasan exponentb=5.541s−1 (withR square of 0.990)forthez component, whileforthe xandycomponents it givesb=11.15 s−1 (withR square of 0.987).The time integration oftheautocorrelationcoefficient overthistime intervalgives the Lagrangianintegraltimescaleforeachcomponent:

TL ii =

Tmax

(9)

Fig. 7. Projection of trajectories in test sections S i for < αp > = 0.30 a) S 1 ; b) S 2 ; c) S 3 ; d) S 4 ; in the case of i) aligned case and ii) tilted case. Symbols (o) and ( + ) indicate

the directions (ascending and descending, respectively). < αp > = 0.3.

(10)

Fig. 9. Homogeneity analysis in terms of particle crossing moving upwards ϕup, j , for different cross sections S i , for the tilted column.

Fig. 10. Evolution of the standard deviation of ϕup, j as a function of the normalized bed height, z/hb .

Thediffusioncoefficientineachdirectionisthengivenby:

Dii=

­

u2pi

(

t

)

®

TL

ii (10)

It isthenclearfromplotsofFigs. 11and12thatthe diffusion in the vertical direction z is stronger than that ofthe transverse plane (x,y), resultingfromboth alarger decorrelationtime anda largervelocityvarianceinz-directionthaninxandy-directions.

Diffusion coefficients in transverse (Dxx and Dyy) and vertical (Dzz) directions asa function of global solid fraction are plotted inFig. 13,forboth homogeneousandnon-homogeneouscases.In

Fig. 11. Variance of particle velocity component as a function of <αp > . Comparison between homogeneous (vertically aligned) and non-homogeneous (tilted column).

both cases as expected, particle diffusion is strongly anisotropic, the diffusion in z-direction being an order of magnitude larger thanthatinxandy-directionsatallsolidfractions.Inthe homo-geneous case, thediffusion coefficientis a decreasingfunction of solid fraction butexhibitsa slightmaximum around <

α

p>=0.2 (open symbols in Fig. 13) for the three components. As for the evolutionofparticleaxialvelocityvariancewithsolidfraction(Fig. 11),thismaximumisaround<

α

p>=0.3inthenon-homogeneous case, and maximum differenceswith the homogeneous case are

(11)

Fig. 12. Particle Lagrangian velocity autocorrelation coefficient versus time for < αp > = 0.3. Homogeneous case. ( ο) z, ( 1) y and (—) x components.

Fig. 13. Particle diffusion coefficient D ii in 3 directions ( D zz , D yy and D xx ). Compari- son between homogeneous (open symbols) and non-homogeneous (filled symbols) cases.

Table 3.

Relative difference δDii between homo- geneous and inhomogeneous cases.

p > δDxx δDyy δDzz 0 .14 0 .17 0 .10 −0 .05 0 .2 0 .02 0 .06 −0 .07 0 .3 −0 .07 −0 .23 −2 .62 0 .4 −0 .07 −0 .75 −2 .14 0 .5 −0 .02 0 .14 0 .07

observed in thesame rangeof

h

α

p

i

, between0.3and0.4. Maxi-mum relativedifferencesare reachedforthez-componentinthat rangeofconcentration,duetothedevelopmentofalarge recircu-lationpatternevidencedby thetrajectoriesenvelopedisplayedin

Fig.5.

Relativedifferences

δ

Dii=1-DiinH/DiiHarereportedinTable3for allsolidfractioninvestigated.Atlowerconcentration(<

α

p>=0.1 and0.2),differencesbetweenbothcasesarenotsignificantinthe vertical direction, suggesting that the large-scale coherent struc-tureis notfullydeveloped,probablydueto atoo smallapparent density-induced collective effect.However, at the lowest concen-tration,theeffectofthetiltistodecreasethediffusivityof parti-cles inthe horizontalplane. Athighconcentration (<

α

p>=0.5),

this coherentmotion ofparticles is damped probablydueto the bedaspect ratio(heightofthebedcompares withcolumn diam-eter inthat case) andthe differencesbetweenthe two casesare also negligible. The maximum difference is reached in therange ofconcentration0.3–0.4,thediffusioncoefficientinz-direction be-ing morethan2timeslargerinthenon-homogeneous (tilted col-umn) casethan in the homogeneous (vertically aligned column) case. Note alsothat in that range ofconcentration, the diffusion inthe(x,y) planeisnotisotropicduetothefact thattheplaneof inclination isthe (y,z) plane, andthe relative difference of diffu-sion coefficient in the y-direction is larger than that observedin the x-direction.Inthe tiltedcase, thetransverse component vari-anceisveryclosetothevalueobtainedintheverticalcase.Itwas alsoshownpreviouslythatthevelocityfluctuationinthe horizon-tal plane was the correct scaling velocity for collisions ( Aguilar-Corona et al., 2011); or in other words, the horizontal fluctua-tionsdeterminedtheuncorrelatedmotionoftheparticles(notonly Gaussian but also Maxwellian, hence isotropic). As a result, tilt-ing the columndoes not significantly affectthe velocity variance (hencethepdf)oftheuncorrelatedpartofparticlesmotion.In re-turn,asshownbyourmeasurements, thediffusivemotioninthe

y-direction is slightly affected by the column tilt. Therefore, the decorrelation time isincreasedby the tiltdue tothe small grav-itycomponentnormaltothewall.

Itcanbe concludedthatwhenarecirculationloopdevelopsin thewholebedvolume,itmainlycontributestotheincreaseof par-ticlevelocity fluctuationsinthe verticaldirection andalsointhe decorrelationtime,leadingtoasignificantincreaseofparticle dif-fusivity in that direction. In this range of highparticle Reynolds and finiteStokes numbers, thevertical alignment ofthe fluidisa-tion column isan important criterion regardingthe validationof numerical methodsinconcentrated two-phase flows.The present experimentshavebeencarriedoutinaliquidfluidizedbedwhere theagitationoffluid,andconsequentlythatofparticles,ismainly induced by wake effects (also referred to aspseudo-turbulence). Note that thissituation is quite differentfrom gas-solid fluidisa-tion whereasgeneralcase, particleagitationisdrivenby the tur-bulenceofthecontinuousphase,modulatedbyparticleinertiaand finitesizeeffects.Inthelattercase,theeffectofasmalltiltofthe columnwouldprobablynotbethesame,becauseoftheturbulent large-scale induced intensemixingthat wouldprevent the devel-opmentofcoherentstructures atthebedscale.Hence,this situa-tion isparticulartogravitydrivendispersed flowathigh concen-tration forwhichproperturbulenceofthecarryingphaseremains smallcomparedtothatinducedbywakeeffects.

4. Conclusions

In this work, we carried out an experimental investigation of the 3-D particle fluctuating motion in a liquid fluidized bed andcomparetwodifferentsituations:ahomogeneous fluidisation regime (homogeneous feeding in the entry section and carefully verticallyalignedbed)andanon-homogeneousfluidisationregime resultingfromasmalltilt(0.3°)ofthefluidisationcolumnwiththe vertical.

The bedexpansion is not modifiedsignificantly by thetilt, as a resultof momentumconservation averagedin thebedvolume, whichatfirstorderbalancesthebuoyancyforceandthedragforce basedonaveragedslipvelocity.Inturn,weshowthattheparticle trajectoriesinthebedarestronglymodified,shiftingfroman over-all uniformlydistributedrandommotioninthebedwith axisym-metrictoroidalstructureinthebottompart,tolarge-scale recircu-lation patternsina givenrangeofbedexpansion(solidfraction). As aconsequence,theparticlevelocityvariance andself-diffusion coefficientaresignificantlyaffectedbythetiltinverticaldirection. Inparticular,itisshownthatthechangeobservedonthese

(12)

quan-titiesdependsonthe fluidisationvelocity andthatthemaximum variationoccursforsolidfractionsrangingbetween0.3and0.4.It is also possible, although not investigated in thisstudy, that the particle velocity fluctuations depend also on the particle inertia (Stokesnumber).Theseresultsarerelevantwhencomparisons be-tweenexperimentsandnumericalsimulationsareconductedatan industrial scale. Ifthe experiment is not accurately aligned with thevertical,mixingofpassivescalarand/or thetransport ofmass or heat could be significantly affected by the non-homogeneous state offluidisation. It isalso clearfromthe resultspresentedin this work that thethree dimensionalcharacter of thefluctuating motionhastobetakenintoaccountwhencomparingwiththe nu-mericalsimulationsandmodels.

This investigationshowedevidence oflargesensitivity tovery smallmisalignmentswithrespecttotheverticalinfluidizedbeds. The implications of such an effect are very important when de-signingamodelexperimentsforthevalidationofnumerical simu-lations.

Acknowledgements

TheauthorswishtoexpresstheirgratitudetotheNational Sci-ence and Technology Council of Mexico (CONACYT) and the re-searchfederationFERMaT(FRCNRS3089)forfundingthisproject.

References

Aguilar, A. , 2008. Agitation Des Particules Dans Un Lit Fluidisé liquide. Étude expéri- mentale Ph.D. thesis. Institut National Polytechnique de Toulouse, France . Aguilar-Corona, A. , Zenit, R. , Masbernat, O. , 2011. Collisions in a liquid fluidized bed.

Int. J. Multiphase Flow 37 (7), 695–705 .

Al-Dibouni, M.R. , Garside, J. , 1979. Particle mixing and classification in liquid flu- idized beds. Trans. Inst. Chem. Eng. 57 (2), 94–103 .

Anderson, T.B , Jackson, R. , 1969. Fluid mechanical description of fluidized beds. Comparison of theory and experiments. Ind. Eng. Chem. Fundam. 8, 137–144 . Buyevich, Y.A. , 1994. Fluid dynamics of coarse dispersions. Chem. Eng. Sci. 49 (8),

1217–1228 .

Carlos, C.R. , Richardson, J.F. , 1968. Solids movements in liquid fluidized beds-I Par- ticle velocity distribution. Chem. Eng. Sci. 23 (8), 813–824 .

Chaikittisilp, W. , Taenumtrakul, T. , Boonsuwan, P. , Tanthapanichakoon, W. , Charin- panitkul, T. , 2006. Analysis of solid particle mixing in inclined fluidized beds using DEM simulation. Chem. Eng. J. 122 (1), 21–29 .

Del Pozo, M. , Briens, C.L. , Wild, G. , 1992. Effect of column inclination on the perfor- mance of three-phase fluidized beds. AIChE J. 38 (8), 1206–1212 .

Didwania, AK , Homsy, GM , 1981. Flow regime and flow transitions in liquid-flu- idized beds. Int. J. Multiphase Flow 7, 563–580 .

Di Felice, R. , 1995. Hydrodynamics of liquid fluidisation. Chem. Eng. Sci. 50 (8), 1213–1245 .

Dorgelo, E.A.H. , Van Der Meer, A.P. , Wesselingh, J.A. , 1985. Measurement of the axial dispersion of particles in a liquid fluidized bed applying a random walk method. Chem. Eng. Sci. 40 (11), 2105–2111 .

Duru, P. , Guazzelli, É. , 2002. Experimental investigation on the secondary instabil- ity of liquid-fluidized beds and the formation of bubbles. J. Fluid Mech. 470, 359–382 .

El-Kaissy, M.M. , Homsy, G.M. , 1976. Instability waves and the origin of bubbles in fluidized beds. Int. J. Multiphase Flow 2-4, 379–395 .

Gordon, L.J. , 1963. Solids Motion in a Liquid Fluidized Bed PhD thesis. University of Washington .

Handley, D. , Doraisamy, A. , Butcher, K.L. , Franklin, N.L. , 1966. A study of the fluid and particle mechanics in liquid-fluidized beds. Trans. Inst. Chem. Eng. 44, T260–T273 .

Hassan, A.Y. , Dominguez-Ontiveros, E.E. , 2008. Flow visualization in a pebble bed reactor experiment using PIV and refractive index matching techniques. Nucl. Eng. Des. 238 (11), 3080–3085 .

Homsy, G.M. , 1998. Nonlinear waves and the origin of bubbles in fluidized beds. App. Sci. Res. 58 (1), 251–274 .

Homsy, G.M. , El-Kaissy, M.M. , Didwania, A. , 1980. Instability waves and the origin of bubbles in fluidized beds - II. Int. J. Multiphase Flow 6, 305–318 .

Hudson, C. , Briens, C.L. , Prakash, A. , 1996. Effect of inclination on liquid-solid flu- idized beds. Powder Tech. 89 (2), 101–113 .

Juma, A .K.A . , Richardson, J.F. , 1983. Segregation and mixing in liquid fluidized beds. Chem. Eng. Sci. 38 (6), 955–967 .

Kang, Y. , Nah, J.B. , Min, B.T. , Kim, S.D. , 1990. Dispersion and fluctuation of fluidized particles in a liquid-solid fluidized bed. Chem. Eng. Comm. 97 (1), 197–208 . Kennedy S., C. , Bretton R., H. , 1966. Axial dispersion of spheres fluidized with liq-

uids. AIChE J. 12 (1), 24–30 .

Latif, B.A.J. , Richardson, J.F. , 1972. Circulation patterns and velocity distribution for particles in a liquid fluidized bed. Chem. Eng. Sci. 27, 1933–1949 .

Sundaresan, S. , 2003. Instabilities in fluidized beds. Annu. Rev. Fluid Mech 35 (1), 63–88 .

Valverde, J.M. , Castellanos, A . , Quintanilla, M.A .S. , Gilabert, F.A . , 2008. Effect of in- clination on gas-fluidized beds of fine cohesive powders. Powder Tech. 182 (3), 398–405 .

Van Der Meer, A.P. , Blanchard, C.M.R.J.P. , Wesselingh, J.A. , 1984. Mixing of particles in liquid fluidized beds. Chem. Eng. Res. Des. 62, 214–222 .

Willus, C.A. , 1970. An Experimental Investigation of Particle Motion in Liquid Flu- idized Bed. Thesis of California Institute of Technology, United States of Amer- ica .

Yakubov, B. , Tanny, J. , Maron, D.M. , Brauner, N. ,2005. The effect of pipe inclination on a liquid-solid fluidized bed. HAIT J. Sci. Eng. B 3, 1–19 .

Yamazaki, R. , Sugioka, R. , Ando, O. , Jimbo, G. , 1989. Minimum velocity for fluidisa- tion of an inclined fluidized bed. Kagaku Kogaku Ronbunshu 15 (2), 219–225 . Yutani, N. , Ototake, N. , Too, J.R. , Fan, L.T. , 1982. Estimation of the particle diffusivity

in a liquid-solids fluidized bed based on a stochastic model. Chem. Eng. Sci. 37 (7), 1079–1085 .

Zenit, R. , Tsang, Y.H. , Koch, D.L. , Sangani, A.S. , 2004. Shear flow of a suspension of bubbles rising in an inclined channel. J. Fluid Mech. 515, 261–292 .

Figure

Fig. 1. Scheme of the experimental set-up and the entry section.  Table 1.
Fig. 3. Raw images as captured by the camera: a) Large solid fraction: &lt;  α p  &gt; = 0.5; to the left of the black division: front view ( x-z plane)
Fig. 4. Particle trajectories projection for the vertically aligned case at a) &lt;  α p  &gt; = 0.14, b) &lt;  α p  &gt; = 0.20 and c) &lt; α p  &gt; = 0.40: left and right columns correspond to the horizontal ( x,y ) and vertical ( x,z ) planes, respecti
Fig. 5. Particle trajectories projection at &lt;  α p  &gt; = 0.30 in a) ( x,z ) plane; b) ( y,z ) plane; c) ( x,y ) plane
+5

Références

Documents relatifs

In the absence of additional information at this time on lipid composition of seston and taxonomic composition of phytoplankton from both treatments, no clear conclusion

In this thesis, we use shear rheology, tack tests, and spherical probe indentation tests to explore correlations between linear viscoelastic properties (i.e., plateau modulus,

Interroger les raisons du maintien dans les métiers politiques ou, au contraire, du choix de l’exit (Hirschman, 2011) nécessite de prendre en compte deux éléments dans l’analyse

Figure 8 presents profiles of the vertical component of solid velocity obtained by different averaging techniques from the bulk flow, as well as velocity profiles from the analysis

We distinguish two broad classes of empirical approaches: approaches based on individuals’ revealed demand for moving resources across health states, and approaches based on

The main findings of this study are as follows: (i) PBS significantly decreases after BMS implantation, whereas it increases after DES implantation irrespective of underlying lesion

Employing Information Extraction and Information Retrieval techniques, along with mining file properties, text keywords, and metadata, I-Document is able to

In order to evaluate the purely mechan- ical contribution of the particles to the paste behavior, independently of the physicochemical properties of the materials, we have