• Aucun résultat trouvé

A consistent dimensional analysis of gas-liquid mass transfer in an aerated stirred tank containing purely viscous fluids with shear-thinning properties

N/A
N/A
Protected

Academic year: 2021

Partager "A consistent dimensional analysis of gas-liquid mass transfer in an aerated stirred tank containing purely viscous fluids with shear-thinning properties"

Copied!
17
0
0

Texte intégral

(1)

HAL Id: hal-00744336

https://hal.archives-ouvertes.fr/hal-00744336

Submitted on 22 Oct 2012

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

A consistent dimensional analysis of gas-liquid mass

transfer in an aerated stirred tank containing purely

viscous fluids with shear-thinning properties

Raouf Hassan, Karine Loubiere, Jack Legrand, Guillaume Delaplace

To cite this version:

Raouf Hassan, Karine Loubiere, Jack Legrand, Guillaume Delaplace.

A consistent dimensional

analysis of gas-liquid mass transfer in an aerated stirred tank containing purely viscous fluids

with shear-thinning properties. Chemical Engineering Journal, Elsevier, 2012, vol. 184, pp.42-56.

�10.1016/j.cej.2011.12.066�. �hal-00744336�

(2)

O

pen

A

rchive

T

OULOUSE

A

rchive

O

uverte (

OATAO

)

OATAO is an open access repository that collects the work of Toulouse researchers and

makes it freely available over the web where possible.

This is an author-deposited version published in :

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

Eprints ID : 6336

To link to this article

: DOI:10.1016/j.cej.2011.12.066

URL :

http://dx.doi.org/10.1016/j.cej.2011.12.066

To cite this version

:

Hassan, Raouf and Loubiere, Karine and Legrand, Jack and

Delaplace, Guillaume A consistent dimensional analysis of gas–

liquid mass transfer in an aerated stirred tank containing purely

viscous fluids with shear-thinning properties. (2012) Chemical

Engineering Journal, vol. 184 (n°1). pp. 42-56. ISSN 1385-8947

Any correspondance concerning this service should be sent to the repository

administrator: staff-oatao@inp-toulouse.fr



(3)

A

consistent

dimensional

analysis

of

gas–liquid

mass

transfer

in

an

aerated

stirred

tank

containing

purely

viscous

fluids

with

shear-thinning

properties

Raouf

Hassan

a,b

,

Karine

Loubiere

c,d,∗

,

Jack

Legrand

a,b

,

Guillaume

Delaplace

e

aUniversitédeNantes,LaboratoireGEPEA,CRTT,37boulevarddel’université,BP406,F-44602Saint-Nazaire,France bCNRS,LaboratoireGEPEA,F-44602Saint-Nazaire,France

cUniversitédeToulouse,INPT,UPS,LaboratoiredeGénieChimique(LGC),4alléeEmileMonso,BP84234,F-31432Toulouse,France dCNRS,LaboratoiredeGénieChimique(LGC),F-31432Toulouse,France

eINRA,LaboratoirePIHMUR638,369RueJulesGuesde,BP39,F-59651Villeneuved’AscqCedex,France

Keywords:

Dimensionalanalysis Variablematerialproperty Gas–liquidmasstransfer Shear-thinningfluids Aeratedstirredtank ModelofWilliamson–Cross

a

b

s

t

r

a

c

t

Thispaperdealswithgas–liquidmasstransferinanaeratedstirredtankcontainingNewtonianor

shear-thinningfluids.Theaimistodemonstratethat,foragivenmixingsystem,anuniquedimensionless

correlationgatheringallthemasstransferrates(150klameasurements)canbeobtainedifandonlyifthe

variabilityoftherheologicalmaterialparametersiscorrectlyconsideredwhenimplementingthetheory

ofsimilarity.Moreparticularly,itisclearlyillustratedthatatoogrosssimplificationintherelevantlist

oftheparameterscharacterizingthedependenceofapparentviscositywithshearratesleadstopitfalls

whenbuildingthe-spaceset.Thisisthenastrikingexampleshowingthatarobustpredictivecorrelation

canbeestablishedwhenthenon-constancyoffluidphysicalpropertiesceasestobeneglected.

1. Introduction

Thedispersionofagaseousphase inaliquidphaseformass transferpurposes isinvolved in manyprocesses, inthefield of chemicalreactionengineering(e.g.forchlorinations, hydrogena-tions,oxidations,alkylations,ammonolysisandsoforth)butalso inbiochemicalengineering(includingfermentation,wastewater treatment).Theuseofagitatedtanksisawidespreadpracticefor operatingsuchabsorptionprocessesasofferingtheadvantagesto generatehighinterfacialareasandintensemixingofliquidphase. Understandingandmodellingmasstransferbetweenphasesisof importance,becauseitmayoftenbecomethecriticalstep deter-miningtheachievementoftheapplication,andthusmaygivethe mainguidelinesonwhichthedesignandthescale-upofthe pro-cesswillbebased.Thetransferredmassquantitydependsonthe solutesolubilityintheliquidphase,butaboveallontheinterfacial area,a,andontheoverallliquid-phasemasstransfercoefficient,Kl. Theproductoftheselatterparametersiscommonlycalled absorp-tionratecoefficientoroverallvolumetricgas–liquidmasstransfer coefficient (Kla or kla for low soluble gases). The factors influ-encingof such unit operation (inparticularKla) arevery large,

∗ Correspondingauthor.Tel.:+330534323619;fax:+330534323697. E-mailaddress:Karine.Loubiere@ensiacet.fr(K.Loubiere).

includingtankgeometryanddimensions,impellertype, dimen-sionsandrotationalimpellerspeed,aerationsystemandgasflow rate,physicalandrheologicalpropertiesofgasandliquidphases, temperature,andpressure.Inanattempttoelucidatetheeffects ofthelatterparametersontheabsorptionratecoefficient,agreat amountofpublishedinvestigationsareencounteredinthe litera-ture;someinterestingoverviewsaregivenin[1–5].Theyprovide improvedknowledgeongas–liquidmasstransferthrough experi-mentaldata,empiricalcorrelations,mechanisticanalysisormore recentlynumericalsimulations.

Mostofthemdealwiththecaseswhentheliquidphaseisa Newtonianfluidwithlowviscosities.Fromvarioussetsof experi-mentscarriedoutatlab-scaleand/oratlargerscale,someempirical correlationsforklaareproposed,involvingeitherdimensionalor dimensionlessparameters.Atpresent, themostfrequentlyused dimensionalcorrelationremainstheoneofVan’tRiet[6]orsome variantsinwhichtheconstantandexponentshavebeenmodified. Theyareexpressedsuchas:

kla=C·(Ug)C1·



P

Vl



C2

·()C3 (1)

wheretheconstantCisstronglyaffectedbythegeometrical param-etersoftheagitationsystem,and is theNewtonian viscosity. Garcia-Ochoa andGomez[3]recentlysummarized thedifferent exponentsassociatedwithEq.(1)thatareavailableintheliterature.

(4)

Nomenclature

a specificinterfacialarea(m−1)

CO2 concentrationindissolvedoxygen(kg/m3) C∗

O2 concentration in dissolved oxygen at saturation

(kg/m3)

d airspargerdiameter(m) D impellerdiameter(m)

D oxygen diffusion coefficient in the liquid phase (m2/s)

G/Vl massthroughoutperunitofliquidvolume(kg/s/m3) g gravityacceleration(m/s2)

H Henry’sconstant(Pa) Ht tankheight(m)

kl liquid-sidemasstransfercoefficient(m/s)

K consistencyindexfromthemodelof Ostwald–de-Waele(Eq.(14))(Pasnost)

Kl overallmasstransfercoefficient(m/s)

kla overallvolumetricgas–liquidmasstransfer coeffi-cient(s−1)

KMO Metzner–Ottoconstant(Eq.(5)) m Henry’sconstant

nost flow indexfromthemodelof Ostwald–de-Waele (Eq.(14))

nw flowindexfromthemodelofWilliamson–Cross(Eq.

(15))

N rotationalimpellerspeed(s−1)

P/Vl powerdissipatedperunitofvolume(W/m3) Ps pressureinthesystem(Pa)

Qg gasflowrate(m3/s) t time(s)

tw time parameter from the model of Williamson–Cross(Eq.(15))(s)

Tl temperatureoftheliquidphase(K) Tt vesseldiameter(m)

Ug superficialgasvelocity(m/s) V tankvolume(m3)

Vl liquidtankvolume(m3) Greekletters

ε meanstandarddeviation(Eq.(50)) ˙ shearrate(s−1)

˙ o referenceshearrate(s−1)

˙ av averageshearratedefinedfromtheMetzner–Otto

concept(Eq.(4))(s−1)  dynamicviscosity(Pas) a apparentviscosity(Pas)

o referenceapparentviscosityparameterdefinedat referenceshearrate(Pas)

w viscosity parameter from the model of Williamson–Cross(Eq.(15))(Pas)



cinematicviscosity(Pas)  density(kg/m3)

 surfacetension(N/m) Dimensionlessnumbers

Fr Froudenumber(Eq.(27))

kla* dimensionlessvolumetricmasstransfercoefficient (Eq.(25))

* dimensionlessviscosity(Eq.(29))

i dimensionlessnumberdeducedfromthetheoryof similarity

* dimensionlessdensity(Eq.(28)) * dimensionlesssurfacetension(Eq.(30))

Sc Schmidt number defined according to gas phase properties(Eq.(31))

t∗

w dimensionless time number issued from the Williamson–Cross’smodel(Eq.(45))

U∗

g dimensionlesssuperficialgasvelocity(Eq.(26)) Subscripts

g gasphase l liquidphase

Another approach is to use correlations with dimensionless groups;contrarytodimensionalcorrelations,theyguaranteefirm basisforprocessscale-up,providedthattheymustbeestablished withrespecttothetheoreticalcontextofthetheoryofsimilarity. Thepioneerwork’sofZlokarnik[7]hasestablishedtherelevant listofinfluencingintensiveparametersandproposedthefollowing dependencebetweendimensionlessnumbers:

(kla)∗=f1

n

P Qg



∗ ,



Q g Vl



∗ ,∗,Sc,Si

o

where

(kla)∗=kla·



 l g2



1/3 ,



P Qg



∗ =



P Qg



·[l·(l·g)2/3] −1 , ∗=·[ l·(4l·g) 1/3 ]−1,



Q g Vl



∗ =



Q g Vl



·



 l g2



1/3 (2)

Notethat,inEq.(2),ScistheSchmidtnumberandSi*isamaterial dimensionlessparameterwhichdescribescoalescencebehaviour ofsolutions(i.e.ionicstrength,electricalchargeofions,...).Thanks tothis approach, Zlokarnik[7] couldrigorously distinguish dif-ferent process relationships depending whether the system is coalescentornon-coalescent.

Fewyearslater,Judat[8]hascriticallyexaminedtheexisting publicationsongas–liquidmasstransfer(coalescingsystems)in stirredvessels.Descriptionofexperimentaldatawiththeaidof intensiveparametershasleadedthisauthorto(±30%deviation): (kla)∗=9.8×10−5· (P/Vl)

∗0.40

B−0.6+0.81×10−0.65/B (3) whereB=(Qg/T2

t)·(l·g)−1/3,theothersnumbersbeingdefined asin[7].Judat[8]hasthenshownthatamonoparametric represen-tationof(kla)*versus(P/Vl)*isinadequate,andthatonlya-space representationcontainingbothdimensionlesspowerperunit vol-umeandsuperficialgasvelocitycansatisfactorilycorrelate(kla)*. Thisauthorhasalsoputforwardthatanother-space, contain-ingnonintensiveparameters (rotationalimpeller speedinstead ofpowerperunitliquid volume)couldbeusedtodescribethe measuresof(kla)*,butthisi-spaceislargerthanthepreviousone. Fewauthors(forexample[9,10])haveconservedthe dimen-sionlessgroup(kla)∗

=kla·(l/g2)1/3definedbythesetwopioneer works for modelling absorption processes. Most of them have adopted,withorwithouttheoreticalbackgrounds,others defini-tionsformakingdimensionlesskla[11–17].Theyincludenotably a modified Sherwood number (kla·T2

t/D) or Stanton number (kla·Vl/Qg)[3].

Whenshear-thinningfluidsareinvolved,twoadditional ques-tions arise unfortunately in a point of view of the theory of similarity:

-Howshouldweproceedtoguaranteethattheresultsobtained withNewtonianfluidscanbeextendedtothesenon-Newtonian liquids?Thespatialdistributionofliquidviscosityinthetank,due toitsdependencywithshearrates,constitutesamajordifficulty withregardtothechoiceofarepresentativeviscosity.

(5)

Table1

Dimensionsoftheexperimentalsetup.

Tanksize Tt=0.212m Ht=0.316m V=10L Hl=0.212m Vl=7.4L Baffles wb=Tt/10(width,inm)

bb=Tt/50(distancefromwalls,inm)

Impeller(six-concave-blade turbine)

D=0.4·Tt(impellerdiameter,inm)

Ds=3·Tt/4(diskdiameter,inm)

C=Tt/4(clearancefromthebottom,inm)

b=D/5(bladeheight,inm) w=0.5D/5(bladewidth,inm) l=D/4(bladelength,inm) Sparger d=D(spargerdiameter,inm)

-Howshouldweproceedtoguaranteethattheresultsobtainedat lab-scalewillbealsoscalableatindustrialscale?

Untilnow,mostofworksencounteredintheliteraturedidnot takecareaboutthesequestions:theysimplyconsistinreplacing theNewtonianviscositybyanapparentviscositydefinedfromthe Ostwald–de-Waele’smodelinwhichanaverageshearrateis con-sideredaccordingtothewell-knownconceptofMetzner–Otto:

˙ av=KMO·N (4)

where theconstant KMO dependsontheagitationsystem.Such choiceofapparentviscosityisinmanycasesquestionable,in par-ticularwhen(i)theflowregimeisnotlaminar(theuseofEq.(4)

becomesthenquitehaphazard)and(ii)therheologicalbehaviour offluidscannotbedescribedinthewholerangeofshearratesby theOstwald–de-Waele’smodel.

Thepresentpaperaimsatrigorouslyansweringthesetwo lat-terquestions,startingfromthetheoreticalbackgroundunderlying thedimensionalanalysisandextendingittothecasesofvariable materialproperties.Moreaccurately,theobjectiveistoshowhow toproceed:(i)toconstructacompletelistofrelevantparameters abletoconsidervariablerheologicalparameters,andconsequently (ii)toelaborate,withoutpitfalls,asetofdimensionlessnumbers characterizingallthefactorsgoverningabsorptionratecoefficients (kla)inanaeratedstirredtankwherepurelyviscousfluidswith orwithoutshear-thinningpropertiesareinvolved.Tosupportthis theoreticalapproach,asetofexperimentswascarriedoutto mea-sureklainastirredtankaeratedinvolume.Differentoperating conditions(rotationalimpellerspeed,gasflowrate)werecovered aswellasvariousfluids(sevenpurelyviscousfluidsofwhichfour haveshear-thinningproperties).

2. Materialsandmethods 2.1. Experimentalset-up

AsshowninFig.1,theexperimentalset-upconsistedofa cylin-dricalPMMAvesselof10Lwithacurvedbottom.Itwasequipped withasquaredoublejacket(27.2cm×27.2cm)andfourbafflesin stainlesssteelmountedperpendiculartothevesselwall.Table1

presentsthegeometricaldetailsofthetank.Theagitationsystem wascomposedofahome-madesix-concave-blades diskturbine whichdimensions wererespectfulfor theonesimplementedin commercial CD6Chemineer®

impellers. Therotational impeller speed(N)wasregulatedbyusinganelectricalmotor(Ikavisc MR-D1 Messrührer, Janke &Kunkel, Ika®), and varied from200to 1000rpm.

Gas(airornitrogen)wasfedintothetankusingaringsparger withadiameterequaltotheimpellerdiameter,asrecommendedby

[18].Thelatterwascomposedby20holesof0.5mmindiameter.

Thespargerwaslocated30mmfromthebottomofthetank,in theaxisoftheimpeller.Thegasflowrate(Qg)wasregulatedby using amanometer (Samson® 4708-1155) and measuredwith avolumetric flowmeter(Brooks® R2-25-C)withanaccuracyof 0.05L/min.Rangedfrom0.33to3.33L/min,thesevaluesremained quitenarrowwhencomparedtotheavailableliterature,theywere initiallyimposedbytheapplicationunderlyingthiswork(namely the AutothermalThermophilic AerobicDigestion of sludge, see

[19]).Intermsofgas–liquidregime,itcanbenoticed(visual obser-vations) thattheoperating conditions undertest (N,Qg, fluids) leadedtoacompletedispersionregime,meaningthusthat bub-bleswerealmostuniformlydistributedthroughoutthetank.One exceptionwasforN=200rpmwheretheloadingregimetookplace (presenceofbubblesonlyintheupperpartofthetank).

2.2. Methodsofoverallvolumetricgas–liquidmasstransfer coefficientmeasurement

2.2.1. Standarddynamicmethod

For implementing the standard dynamic method,the liquid phase wasdeoxygenated by flushingwithnitrogen.Then, after replacingnitrogenbyair,thevariationindissolvedoxygen con-centrationswithtimewasmeasureduntilreachingthesaturation. Forthat,twoprobes(InPro-6050,Mettler-Toledo®)andan acqui-sitioncardwereimplemented,aswellastheLabView®

software fordataacquisition.Thepositionsoftheprobesarerepresentedin

Fig.1:theyarelocatedverticallyat4cmand18cmabovethe bot-tomofthevessel,andhorizontallyat1.5cmfromwalls.Assuming awellmixedliquidphase,themassbalanceindissolvedoxygen concentrationisgivenby

dCO2

dt =Kl·a·(C ∗

O2−CO2) (5)

whereKlistheoverallmasstransfercoefficientintheliquidside andaistheinterfacialareabetweengasandliquidphases.The two-filmtheoryofLewisandWhitman[20]assumesthatKlistheresult oftwolocalmasstransfercoefficients(klandkg):

1 Kl= 1 kl+ 1 m·kg (6)

where m is the Henry’s constant (m=H/Ps). The solubility of oxygen is low: H is equal to 4.05×109Pa (corresponding to C*=9.09mgL−1)indeionisedwaterat293Kinequilibriumwith airunderatmospheric pressure.So,alltheresistancetooxygen masstransferislocatedintheliquidfilm,leadingtoKl≈kl.

Asaconsequence,thevolumetricgas–liquidmasstransfer coef-ficient,kla,canbedirectlydeducedfromtheslope ofthecurve relatingln(C∗

O2−CO2)totime,obtainedwhenintegratingEq.(5).

Thedynamicsoftheoxygenprobecanbedescribedusinga first-orderdifferentialequation[21]as:

dCp dt =

1

tp(CO2−Cp) (7)

whereCpisthedissolvedoxygenconcentrationinsidetheprobe. Thetimeconstantoftheoxygenprobe,tp,wasmeasuredusinga methodbasedonprobe responsetonegativeoxygensteps[22]

andfoundequal to16s.Thislattervalueremainedsmallwhen comparedtomasstransfercharacteristictimes,1/kla.Asthe tem-peratureTlslightlyvaried(20±3◦C)withpowerdissipation,the usualtemperaturecorrectionwasapplied[23]:

(6)

Fig.1. Experimentalset-up.

Foragivenoperatingcondition(NandQg),themeanklainthetank wascalculatedbyaveragingthevaluesmeasuredbythetwoprobes andforthethreeruns(N′=3),suchas:

kla=<kla>= 1 N′ ·

X

N′ |klatop+ klabottom| 2 (9)

Theaxialhomogeneity(h)ofvolumetricgas–liquidmasstransfer coefficientswasalsoevaluatedbyusingthecriterionhdefinedas follows[24]: h= 1 N′′ ·

X

N′′ |klatop−klabottom| <kla> (10)

whereN′′wasthenumberofexperiments(N′′=60foreachliquid phase).

2.2.2. Chemicalmethod

Whenapplyingthelatterdynamicmethodinviscousfluids,the impactoftheprobedynamicsandoftheliquidfilminfrontofthe membraneontheprobecannomorebeignored,aspossibly bias-ingthemeasurementsofkla[25].Inaddition,insuchfluids,the gashold-upstructureisknowntobeverydifferentfromtheone observedinwaterandotherlow-viscosityliquids:manytiny bub-blesappeartoaccumulateduringaerationandcirculatewiththe liquidwhilelargebubblesarealsoobserved(bimodalbubble popu-lation).Thesetinybubblescanactivelycontributetomasstransfer, dependingwhethertheyareinequilibriumwiththelevelof dis-solvedsoluteintheliquidphase(highresidencetimes)ornot[26]. Forthesereasons,ithasbeenchosentoimplementasecond methodforklameasurement.Itwillthenenabletotestthe valid-ity and accuracy of the dynamic method in the viscous fluids involved.Thisalternativemethodwasthechemicalmethod devel-opedby[27],basedonamassbalanceonsodiumsulphite(Na2SO3) concentrationsduringagivenaerationtime.Nitrogenwasfirstly injectedintotheliquidphaseinordertoeliminatethedissolved oxygenpresentinthetank.Whentheconcentrationofdissolved oxygenreachednearlyzero,anadequateamountofNa2SO3was introduced;air wasthenintroducedin thetankandwillreact,

duringanaerationtimetaeration,withthesmallquantityofNa2SO3 introduced:

Na2SO3+1

2O2→Na2SO4 (11)

Themass ofNa2SO3 toinitially introduce (mt)must bechosen carefully,asitshouldenabletokeepazerooxygenconcentration duringtheaerationtime(taeration)whileavoidinganexcessiveuse ofNa2SO3.Indeed,itisimportanttoguarantythatthecoalescing propertiesoftheliquidphasewerenotaffectedbythepresence ofNa2SO3.Agoodcompromisewastomaintainaninitial concen-trationbelow0.5g/L(i.e.mt<3.5gwithVl=7.4L)[25].Notethat tominimizemt,itwasalsopossibletoplayontheaerationtime (taeration),whichwasheretypicallyrangedfrom1.5minto9min. Theoptimizationofbothmt andtaerationwasmadeeasierbythe factthattheordersofmagnitudeofkLawereknownthankstothe measurementsissuedfromthedynamicmethod.

Whensuch conditionsarerespected,Painmanakuletal.[27]

haveshownthattheoverallvolumetricgas–liquidmasstransfer coefficientcanbededucedfrom:

kla=(1/2)(MO2/MNa2SO3)·(mt−mr) taeration·Vl·C∗

O2

(12) wheremt isthetotalmassofNa2SO3 initiallyintroduced,mr is themassofNa2SO3remaininginthetankafteranaerationperiod taeration,andCO∗2istheconcentrationindissolvedoxygenat

satura-tion.Forglycerinesolutions,C∗

O2wasbydefaultconsideredequalto

8.8mgL−1asindeionisedwaterat20C.Anidenticalassumption wasmadeforCMCandxanthangumsolutionsinagreementwith thedatareportedby[28]whoshowedthat,intherangeof con-centrationshereinvolved,nomajorvariationofC∗

O2 occurswhen comparedtowater.

Atlast,foreachcondition,threesamples(10mL)weretakenin thetank,immediatelymixedwith10mLofstandardiodinereagent at0.12equiv./L.Thetitrationofthesesampleswithasodium thio-sulphatesolution(0.05equiv./L)andastarchindicator(iodometry titration)gaveaccesstotheconcentrationofNa2SO3remainingin thetankafteranaerationperiodtaeration,andthustomr.

(7)

2.2.3. Surface-andvolume-aerations

Theoverallvolumetricgas–liquidmasstransfercoefficient mea-suredbythelattermethodsisinrealitytheglobalresultofboth contributions:

- thesurface-aeration:itcorrespondstothemasstransferoccurring atthefreesurfacewhichimportancedependsstronglyonthe surfacemotion.Italsoincludestheaerationassociatedwiththe bubblesentrainedfromthesurfaceintotheliquidbulk; -thevolume-aeration:itisinducedbythebubblesgeneratedatthe

spargerdirectlyinsidetheliquidbulk. Thiscanbeexpressedsuchas:

kla|t=kla|surf+kla|vol (13)

Somemeasurementsaremadewithmechanicalagitationand with-outbubblingatthesparger.Theyenabletogetanideaoftherelative importanceofeachcontribution.

2.3. Fluids

Theapplicationunderlyingthisstudydealtwithinvestigations onaerationperformancesinanAutothermalThermophilic Aero-bicDigestion(ATAD)processfortreatingsludgeissuedfromwaste watertreatmentplant[19].Assludgewasaverycomplex mate-rial,itwasdecidedinafirststeptoworkwithmodelfluidsinstead ofsludge.Theirformulationwaschosensoastoobtain rheologi-calbehavioursasclosetosludgeaspossible,inparticularinterms ofshear-thinningproperties.Forthesereasons,andwithregard toliterature,aqueoussolutionsofcarboxymethylcellulose(CMC) and xanthangumwereselected.Theuseoftwo typesoffluids offeredtheadvantagetocoverawiderrangeofcombinationsof flowandconsistencyindexes.Toensuretheirstabilityintime (dur-ingseveraldays),NaClwasaddedat0.1%(w/v)tothesolutions ofxanthangumsolutions[29]andNaHCO3(0.1mol/L)+Na2CO3, 10H2O(0.1mol/L)tothesolutionsofCMC[30].Various concentra-tionsofCMCandxanthangumweretested,andfinallyconverged towardsthefollowingones:4and6g/LforCMC,and1and2g/Lfor xanthangum.Inadditiontothesenon-Newtonianfluids,deionised waterandtwoaqueoussolutionsofglycerine(50%and70%,v/v) werealsochosenasNewtonianfluids.

The rheology of these fluids was measured, at 20◦C, by a rotationalstress-controlledrheometrer(MCR500,PAARPhysica®) equippedacone-platedevice(50mmindiameter,3degreeincone angle).Theirdensityandsurfacetensionweredeterminedusing adensimeterERTCO®andatensiometerinvolvingtheWilhelmy plate method (3S GBX®).The physical and rheological proper-tiesofbothNewtonianandnon-Newtonianfluidsarecollectedin

Table2.

Therheologicalbehavioursofthenon-Newtonianfluidswere firstlycharacterisedbymeasuringthevariationofshearstress() orapparentviscosity(a)asafunctionofshearrates(˙ )which range variedfrom0.1 to3000s−1.Whencomparing thecurves obtainedforincreasinganddecreasingshearrates,nodifference wasobservedwhateverthenon-Newtonianfluids:nohysteresis phenomenonthenexisted.ShowninFig.2,therheogramsobtained (issued fromseveral trials)clearly illustrate theshear-thinning propertiesofthesefluids.Foreachfluid,theyieldstresswasalso determined,byapplyingthemethodproposedby[31]which con-sisted in oscillating stress sweep testsat a constant frequency (1Hz);thedynamicyieldstresswasthendefinedattheendofthe linearviscoelasticregion,namelyfromtheabscissa correspond-ingtotheintersectionpointbetweenthetangenttotheplateau andthetangenttotheinflexionpointofthecurvelinkingcomplex modulusandshearstress.Dependingonthefluid,theyieldstress

0.001 0.01 0.1 1 10 100 0.1 1 10 100 1000 10000 shear rate (s-1) v is co si ty ( P a.s ) Exp. Ostwald Williamson

(a)

0.001 0.01 0.1 1 10 100 0.1 1 10 100 1000 10000 shear rate(s-1) v is co si ty ( P a. s) Exp. Ostwald Williamson

(b)

0.001 0.01 0.1 1 10 100 0.1 1 10 100 1000 10000 shear rate (s-1) v is co si ty ( P a. s) Exp. Ostwald Williamson

(c)

0.001 0.01 0.1 1 10 100 0.1 1 10 100 1000 10000 shear rate (s-1) v is co si ty ( P a. s) Exp. Ostwald Williamson

(d)

Fig.2.Rheograms:(a)xanthangumat1g/L,(b)xanthangumat2g/L,(c)CMCat 4g/L,and(d)CMCat6g/L.

wasfoundtovarybetween0.1and1Pa,andremainedthus neg-ligible.Tocompletetherheologicalcharacterisationofthefluids, theviscoelasticpropertieswereinvestigated,bymeansofcreeping tests,relaxationtestsand/ordynamicoscillatingmeasurements.In

(8)

Table2

Physicalandrheologicalpropertiesofthefluids.

l(kg/m3) l(Pas) (N/m) Ostwald–de-Waele’smodel: Williamson–Cross’smodel:

K(Pasn) n ost w(Pas) tw(s) nw Air 1.18 1.85×10−5 Newtonianfluids Deionisedwater 998 0.001 0.0728 – – – – – Glycerine50%[Gly50] 1145 0.0109 0.0456 – – – – – Glycerine70%[Gly70] 1195 0.0349 0.0503 – – – – – Non-Newtonianfluids CMC4g/L[CMC4] 997 – 0.0717 0.1914 0.642 0.091 0.029 0.546 CMC6g/L[CMC6] 1006 – 0.0771 0.9470 0.527 0.948 0.844 0.514 Xanthangum1g/L[XG1] 1013 – 0.0753 0.0890 0.543 1.885 57.85 0.411 Xanthangum2g/L[XG2] 1032 – 0.0767 0.5084 0.373 29.505 150.36 0.281

therangeofshearrateinvestigated,nomajorelasticpropertywas highlighted.

Basedonthesefindings,somerheologicalmodelswerechosen tomathematicallydescribethevariationofapparentviscosity(a) withshear ratesrangingfrom0.1 to3000s−1.Amongthelarge variabilityavailableintheliterature,twomodelswereselected: - theOstwald–de-Waele’smodel

a=K· ˙ nost−1 (14)

whereKandnostarerespectivelytheconsistencyandflowindexes respectively.

-theWilliamson–Cross’smodel

a= w

1+ (tw· ˙ )1−nw (15)

where w is a parameter describing a pseudo-Newtonian behaviourforthesmallestshearrates,twisa timeparameter characterizingthetransitionbetweenthe“pseudo-Newtonian” andpurelyshear-thinningbehaviours,andnwistheconsistency index.

Thevalues of K,nost,w,tw,nw are reportedin Table2 for eachfluid;theyhavebeenobtainedbymulti-parameter optimiza-tionsusingthesoftwareAuto2fit®.Intermsofconsistencyindex, themostshear-thinningfluidappearedtobethesolutionof xan-thangumat2g/L.Basedonthelatterparameters,theapparent viscositiespredictedbyEqs.(14)and(15)werecomparedtothe experimental ones in Fig. 2. For all the fluids, a better agree-mentwithexperimentswasobtainedwiththeWilliamson–Cross’s model,insofarasitenablestodescribethemostfaithfullypossible theshapeoftherheogramsoverthewholerangeofshearrates.This demonstratesthatthefluidsundertestwerenotshear-thinningon thewholerangeofshearratesinvestigated,threeparametersbeing requiredtowelldescribetheirbehaviour.

3. DimensionalanalysisforNewtonianandnon-Newtonian Fluids

3.1. Generationofi-setsgoverningaerationprocessfor Newtonianfluids

Inthepresentstirredtank, bubblesweredirectlygenerating insidetheliquidbulkbymeansofagassparger.Surface-aeration wasofcoursepresent,butitscontributionremainedminorwhen comparedtovolume-aeration(seeSection4).Consequently,the overallvolumetricgas–liquidmasstransfercoefficient,kla,canbe consideredasthetractablequantitywhich issignificantly influ-encedbyaerationconditions:itwillbethustakenastargetvariable. Remindthat such choiceisbased onthefollowing relationship

describingthephysicalabsorptionprocessaccordingtothe two-filmtheory:

G Vl

= kl· a· 1C or kl· a= G

Vl·1C (16)

whereG/Vlisthemassthroughputperunitvolumeofliquidand 1Cisacharacteristicconcentrationdifference.Eq.(16)implicitly assumesthat(i)theintensityofgas–liquidcontactingissohigh thataquasi-uniformsystemisproduced,(ii)thegas-phasemass coefficientkgisnegligiblewhencomparedtokl(lowsolublegases), (iii)theabsorptionrateattheinterfaceisextremelyfast,resultingin anequilibriumconcentrationofthedissolvedgasattheinterfaceC* (e.g.1C=C*− C).Hence,theestablishmentofthelistofparameters influencingthemainparameterklashouldrespectthefollowing rules[7]:(i)klamustbeindependentofallgeometricalparameters (i.e.diametersofstirrerandtank,etc.),(ii)klamustbeindependent ofthematerialparametersofgasphase,and(iii)klaisanintensive quantitybecauseofitsvolume-relatedformulation.

Aspreviouslymentionedthenumberoftheparameters influ-encingkla,evenperformedinNewtonianliquids,islargeandcan bedecomposedaccordingto:

• Thegeometricparameters(seelegendinTable1),characterizing -thetank:Tt,Hl,curvatureradiusandangle(fortank’sbottom),

...

-theimpeller:D,Ds,C,w,b,l,...

-thesparger:d,number,diameterandshapeofholes,... • Thematerialparameters:

l,l,g,g,s,D,C* • Theprocessparameters:

g,N,Ug= Qg ·T2

t/4

,Tl,Ps,...

Notethat,inthepresentstudy,alltheexperimentswere con-ductedatroomtemperatureandatmosphericpressure,inducing thusthattemperatureandabsolutepressurewillnotbelisted.The geometryofthetankwasalsounchanged,aswellasthetypeand positionofbothspargerandimpeller.Asaconsequence,thelist ofindividualphysicalquantitiescouldbereduced:Table3shows thedimensionalmatrixobtainedwiththereducedlistofrelevant parameters.

Itisvoluntarilychosentolistthenon-intensiveparametersN andDinsteadofpowerperunitofliquidvolume(P/Vl).Themain motivationisthat P/Vlisan intermediaryvariablewhichis not alwaysavailable(inparticularatindustrialscale),andthususing suchintensivevariablewouldrestrictthefieldofapplicationsof thefinaldimensionalcorrelationwhichwillbeestablished.

InTable3,thecolumnsareassigned totheindividual phys-icalquantities and therows to theexponents appearing when

(9)

Table3

Dimensionalmatrixoftheinfluencingparameters(Newtonianfluids).

Corematrix Remnantmatrix

g g g kla Ug N D l l  D

Mass,M(kg) 1 1 0 0 0 0 0 1 1 1 0

Length,L(m) −3 −1 1 0 1 0 1 −3 −1 0 2

Time,T(s) 0 −1 −2 −1 −1 −1 0 0 −1 −2 −1

each quantityis expressedasan appropriatepower productof thebasedimensions(mass,length,time).Thistableisstructured inacorematrixandaresidualmatrix.Thecorematrixregroups theindividualphysicalquantitiesputforwardbytheusertoform thedimensionlessratiosfromotherindividualphysicalquantities (namely kla,Ug, N, D,l,l,, D).Depending onthe individ-ualphysicalquantitiesassignedinthecorematrix,severalsetof dimensionlessnumbersicanbeobtained.Ithasbeenshown[32]

thatallthei-setsobtainedfromasingleandidenticalrelevance listareequivalenttoeachotherfromapointofviewof dimen-sionalanalysis,andcanbemutuallytransformedatleisure.The finalformforthei-setshouldbelaiddownbytheusersoastobe the“best”suitableforevaluatingandpresentingthe experimen-taldata.Contrarytowhatcommonlyfoundintheliterature,the gaspropertiesgandg(andnottheliquidproperties)werehere chosenasindividualphysicalquantities;theassociated motiva-tionwastogeneratedimensionlessnumbersdependentofasingle influencingparameter,thegasphasebeingkeptunchangedinthis study(air).

Generatingthesetofdimensionlessnumbers(andpossiblytheir futuretransformation)representsanextremelyeasyundertaking whencomparedtothedrawingupofareliableandasaccurateas possiblerelevancelist;thiscanbemadebymatrixtransformation. Thestartingpointconsistsincarryingouttheso-calledGaussian algorithminordertoobtainaunitcorematrixbylinear transfor-mations(zero-freemaindiagonal,beneathitzeros).Table4reports theunit corematrixassociated withthedimensionalmatrix of

Table3.Theanalysisoftheunitcorematrixleadstothe dimen-sionlessratios.Indeed,therowsofresidualmatrixareassignedto theexponentswithwhomtheelementsofthecorematrixappear whentheindividualphysicalquantitiesoftheresidualmatrixare expressedasanappropriatepowerproductofthephysical quanti-tiesofthecorematrix.Literature[33]offersdetailedexamplesof howtohandlematrixtransformationandrecombinationinorder toquicklyobtainthecompletesetofdimensionlessnumbers.This aspectwillbethenonlybrieflydescribedinthispaper.For New-tonianfluids,thematrixanalysisleadstotheeightdimensionless numbers,1to8: 1= kla g1/3·−1/3g ·g2/3 (17) 2= Ug g−1/3·1/3g ·g1/3 (18) 3= N g1/3·−1/3g ·g2/3 (19) 4= D g−2/3·2/3g ·g−1/3 (20) 5= l g (21) 6= l g (22) 7=  g−1/3· 4/3g · g1/3 (23) 8= D g−1·g (24) Notethat1isnothingotherthatthedimensionlessvolumetric masstransfercoefficientdefinedby[7],thecombinationof4with (3)2leadstotheFroudenumber,and8istheinverseofaSchmidt numberdefinedaccordingtogasphaseproperties.Byintroducing thegaskinematicviscosity(g=g/g)andgivingexplicit nota-tions,thelatternumbersbecome:

1=kla∗=kla·



g g2



1/3 (25) 2=U∗ g= Ug (g·g)1/3 (26) 3= Fr=N 2·D g (27) 5=∗= l g (28) 6=∗= l g (29) 7=∗=  (3 g·4g·g) 1/3 (30) 8=Sc= g D (31)

Thus,atagiventemperature,underagivenpressure,forthe geom-etryoftheaeratedstirredtankundertest(inparticularforHl/Tt=1, D/Tt=0.4, Ds/Tt=0.75and fortheothergeometricalratios char-acteristics of thesystem), the dimensionalanalysis states that, when Newtonian fluids are involved, dimensionless volumetric masstransfercoefficient(kla*)ispotentiallyaffectedbysix dimen-sionlessnumbers,respectivelydescribingtheeffectsofsuperficial gasvelocity,rotationalimpellerspeed,liquiddensity,Newtonian viscosity,liquidsurfacetensionandoxygendiffusivity:

kla∗=f



U∗ g,Fr,∗,∗,∗,Sc

(32) Table4

Unitcorematrixobtainedbylineartransformationsofdimensionalmatrix(Newtonianfluids). Corematrix Residualmatrix

g g g kla Ug N D l l  D M+T+2A 1 0 0 1 3 − 1 3 1 3 − 2 3 1 0 − 1 3 −1 3M+L+T+A 0 1 0 −13 13 −13 23 0 1 4 3 1 A=−13×(3M+L+2T) 0 0 1 2 3 13 23 −13 0 0 13 0

(10)

Suchformulationofdimensionlessnumbersofferstheadvantageto enabletheimpactofallinfluencingparameterstobestudied sepa-rately,orinotherswordseachdimensionlessnumberisdefinedfor asingleinfluencingvariable.Thisisnotthecaseintheliterature where,forexample,mostoftheauthorsusedtheaeration num-ber,Na=Flg=Qg/(N·D3),inwhichtheeffectofgasflowrateisnot decoupledfromtheoneofrotationalimpellerspeed.

Atthisstate,thedependenceofEq.(32)isallthatcanbe con-tributedbythetheoryofsimilarity.Themathematicalexpression forthefunctionf,namelyfortheprocessrelationship,hastobe determinedexperimentally.

3.2. Extensionofthetheoryofsimilaritytothecasesofvariable materialproperties

Whenusingthedimensionalanalysistomodelsystemanswers, itisgenerallyassumedthatthematerialpropertiesremain unal-teredinthecourseoftheprocess.However,theinvariabilityof materialpropertiescannotbeassumedwhennon-Newtonian flu-idsareinvolved.Indeed,attheleastoneofthematerialproperties, theapparentviscosity,cannolongerbeconsideredasaconstant insidethewholevolumeoftheaeratedstirredtank,insofarasthe dependencyofthislatterwiththeshearrates(˙ )generatesa spa-tialdistributionoftheliquidviscosity.Theunderlyingquestion addressedtothedimensionalanalysisisnow:howmustthespace ofdimensionlessnumbers,i,bebuilt inpresenceofsuchvariable materialproperty?

Inthecaseofmaterialswithconstantproperties,nospecial pre-cautionshouldbemadetoguaranteethataprocessrelationship correlatingasetofdimensionlessratiosisalsoapplicabletoanother material.Thisisnottrueformaterialswithvariablesproperties,as demonstratedby[34].Inthiscase,weshouldfirstensureas prior-itythatacertainsimilarityexistsformaterialsinordertoextend therangeofvalidityoftheprocessrelationshiptoother materi-als.Despitethat,thetheoryofsimilarityhaslittlechangedsince itsbeginning,andthedimensionalmodellinginvolvingmaterials withvariablephysicalpropertiesremainstreated,inmostofthe papers,asthecasewithconstantmaterialproperties.Theauthors ignorethenthefactthatthespatio-temporalvariabilityof mate-rialpropertiesinfluencesthecourseoftheprocess!Oneexception isthemodellingofthetransformationprocesseswhere a mate-rialhavingatemperature-dependenceinviscosityissubmittedto heattransfercondition.Mostofattemptsmadetotakeintoaccount thevariabilityofproductpropertiesinthereactorhaveconsisted inaddinganewratioraisedtoa certainexponentto character-izethesystemresponse.Thisratioisdefinedbytheratiobetween theviscositiesatbulktemperatureandatwalltemperature. How-ever,thiskindofenlargementofthesetofdimensionlessratios istheoreticallyvalidonlyforfewlimitedcases,thatistosayonly ifthematerialfunction(hereviscosityversustemperature) sat-isfiesspecificcriteria[34].Inotherwords,suchmethodleadsto biasedpredictionsandthus,cannotbegeneralisedwhenhandling otherfluids. Despite this fact, this ratio remains systematically used,whatevertheproductsinvestigated,inmostofthestudies since[35].

The theoretically consistent way of proceeding has been introducedbyPawlowskiin1971[34]andrememberedby[32,36]. Themethodconsistsinintroducingsomeadditionalparametersin therelevancelistsoastotakeintoaccountthevariationinthe flowdomainofthephysicalproperty,noteds(forexample viscos-ity),asafunctionofaparameternotedp(forexampletemperature orshearrate);s(p)iscalledthematerialfunction(forexample(T) or(˙ )).Thisimpliesthatthei-spacegoverningtheprocesswill beextendedincomparison tofluidshavingconstantproperties. Theguidelinestointroducetherightnumberofadditional dimen-sionalparametersaredetailedin[34].Hence,whendealingwitha

material function s(p) which is apparent viscosity a(˙ ), the Pawlowski’swork[34]canbesummedup,asfollows:

• Firstly,allthedimensionalparametersdefinedintherelevantlist withfluidshavingconstantproperties(Newtoniancase)should beconserved,exceptforthevariablephysicalpropertiesin ques-tion(hereapparentviscositya(˙ )).

• Secondly,theNewtonianviscosityshouldbereplacedbya refer-enceapparentviscosity,o,calculatedatareferenceshearrate, ˙ o.Itisimportanttopointoutthatanyvalueforthereference shearratecanbechosen.

• Thirdly,thereferenceshearrateshouldbeaddedinthelistof relevantparameters.Atthisstage,wecanpointoutthatsome exceptionstothisruleexist.Indeed,ithasbeendemonstrated thataddingthereferencepointisnotnecessarywhenthe mate-rialfunctions(p),herea(˙ ),canbedescribedbythefollowing familyofcurves:

s(p)=(A+B·p)c or s(p)=exp(A+B·p) (33) whereA,BandCarethreeindependentconstants.

• Finally, a set of additional dimensionless numbers shouldbe addedintherelevant(dimensional)listtotakeintoaccountthe dependencyofmaterialfunction.Thesedimensionless parame-ters,calledrheol,correspondtoallthedimensionlessratiosi whichappearintheexpressionofthefunctionu,exceptforthe ratio ˙ /˙ 0: {rheol}=



{i}suchasu=(˙ − ˙ 0) 1 a(˙ 0)·

h

da

i

˙ =˙ o =g



˙ ˙ 0;{i}





(34) Whenintegrating the previousguidelines,the listof the influ-encingparametersestablishedforNewtonianfluidsbecomesfor non-Newtonianfluids:

{kla,g,g,g,Ug,N,D,l,,D,a(˙ o), ˙ o,rheol} (35) Thenextstepistofindthetypeofmaterialfunctiondescribingthe rheologicalbehaviourofviscousfluidshavingshear-thinning prop-erties.AspresentedinSection2,twomodelsarewelladaptedfor theinvestigatedfluids:theOstwald–de-Waele’smodel(Eq.(14)) andtheWilliamson–Cross’smodel(Eq.(15)).Asthesemodelsare abletodescribethevariationofviscositywithshearrateforallthe aqueoussolutionofCMCandxanthangum,eachofthemconstitute onepossiblematerialfunction.

3.2.1. CaseNo.1:Ostwald–de-Waele’smodel.

Whenthematerialfunction correspondstothe Ostwald–de-Waele’smodel,thefunctionucanbeexpressedas:

u=



˙ ˙ 0−1



(nost−1) (36) Consequently, {rheol}={nost} (37)

TheOstwald–de-Waele’smodel, definingasa=K· ˙ nost−1 (Eq.

(14)),verifiesEq.(33),implyingthusthatthereferenceshearrate, ˙ o,canberemovedfromtherelevancelist.Asaconsequence,for purelyviscousfluidshavingshear-thinningproperties,the addi-tionalparametersisrestrictedtonostandtheNewtonianviscosity isreplaced bya(˙ o).Eq.(32)establishedfor Newtonianfluids becomesthen: kla∗=g



U∗g,Fr,,,Sc,=a( o) g ,nost



(38)

(11)

Table5

Dimensionalresultsforkla(expressedins−1):Newtonianandnon-Newtonianfluids.

N(rpm) Qg(L/min) Newtonianfluids Non-Newtonianfluids

Water Gly50 Gly70 CMC4 CMC6 XG1 XG2

200 0.33 9.05×10−4 0.9 1.91×10−3 –– 1.6 3.35×10−3 7.75×10−4 3.13×10−4 1.64×10−3 1.26×10−3 1.55×10−3 1.31×10−3 2.33 4.58×10−3 1.05×10−3 4.22×10−4 2.08×10−3 1.85×10−3 2.06×10−3 1.47×10−3 3 5.33×10−3 1.45×10−3 5.15×10−4 2.58×10−3 2.01×10−3 2.68×10−3 2.16×10−3 3.33 5.88×10−3 1.77×10−3 5.75×10−4 2.82×10−3 2.3×10−3 2.85×10−3 2.56×10−3 400 0.33 2.31×10−3 0.9 4.91×10−3 1.6 8.69×10−3 2.90×10−3 6.82×10−4 3.68×10−3 2.54×10−3 4.83×10−3 3.61×10−3 2.33 1.12×10−2 3.50×10−3 8.00×10−7 4.43×10−3 4.40×10−3 5.93×10−3 4.81×10−3 3 1.29×10−2 4.28×10−3 9.99×10−4 4.98×10−3 4.98×10−3 6.59×10−3 6.51×10−3 3.33 1.31×10−2 4.63×10−3 1.19×10−3 5.19×10−3 5.53×10−3 7×10×10−3 8.01×10−3 600 0.33 4.30×10−3 0.9 9.19×10−3 1.6 1.62×10−2 4.72×10−3 1.38×10−3 7.51×10−3 3.99×10−3 9.52×10−3 7.68×10−3 2.33 1.82×10−2 5.73×10−3 1.77×10−3 8.60×10−3 6.90×10−3 1.13×10−2 9.57×10−3 3 2.06×10−2 6.39×10−3 2.25×10−3 9.23×10−3 8.20×10−3 1.30×10−2 1.11×10−2 3.33 2.26×10−2 6.78×10−3 2.55×10−3 9.44×10−3 9.05×10−3 1.39×10−2 1.20×10−2 800 0.33 4.94×10−3 0.9 1.39×10−2 1.6 2.26×10−2 6.87×10−3 2.35×10−3 1.14×10−2 6.40×10−3 1.29×10−2 1.08×10−2 2.33 2.68×10−2 7.61×10−3 3.31×10−3 1.32×10−2 8.30×10−3 1.51×10−2 1.28×10−2 3 2.81×10−2 8.14×10−3 3.97×10−3 1.45×10−2 9.17×10−3 1.79×10−2 1.43×10−2 3.33 3.21×10−2 8.45×10−3 4.46×10−3 1.48×10−2 9.75×10−3 1.87×10−2 1.57×10−2 1000 0.33 1.05×10−2 0.9 1.43×10−2 1.6 2.25×10−2 8.36×10−3 4.41×10−3 1.54×10−2 8.90×10−3 1.70×10−2 1.38×10−2 2.33 2.70×10−2 8.87×10−3 5.21×10−3 1.75×10−2 1.05×10−2 2.06×10−2 1.68×10−2 3 2.84×10−2 9.18×10−3 5.74×10−3 1.88×10−2 1.17×10−2 2.36×1010−2 1.89×10−2 3.33 3.34×10−2 9.48×10−3 6.21×10−3 1.90×10−2 1.29×10−2 2.61×10−2 2.06×10−2

WhendegeneratedtotheNewtoniancase,thematerialfunction associatedwiththeOstwald–de-Waele’smodelleadstonost=1and *=a/gwherea=listheNewtonianviscosity.

3.2.2. CaseNo.2:Williamson–Cross’smodel

When the material function corresponds to the Williamson–Cross’s model, the function u can be expressed as: =



˙ ˙ 0 ˙ 0



·(nw−1)·(tw· ˙ 0) (−nw) 1+(tw· ˙ 0)1−nw (39) Consequently, {rheol}={nw,tw· ˙ 0} (40)

The reference shear rate, ˙ o, should be listed as the Williamson–Cross’smodeldoesnotverifyEq.(33).Finally,thenew variablestoaddintherelevantdimensionallistare:

{˙ o,nw,tw· ˙ 0} (41)

Nevertheless,as ˙ ocantheoreticallytakeanyvalue,itispossible tochoose:

˙ o= 1

tw (42)

Bythisway,thesupplementaryvariablesarerestrictedto:

{tw,nw} (43)

As a consequence, for the fluids described by the Williamson–Cross’s model, the additional parameters are tw andnw,andtheNewtonianviscosityshouldbereplacedby a



1

tw



=0= w

2 (44)

After introducing these variables in the core matrix (Table 3) andapplyingthelineartransformationsfromTable4,thesetof

dimensionlessnumbersdefiningaeration processisenlargedby twodimensionlessnumbers:

9= nw and 10= t∗ w= 1 tw·



g g2



1/3 (45) Eq.(32)establishedforNewtonianfluidsbecomesthen:

kla∗=h



U∗ g,Fr,∗,∗,Sc,∗= a(1/tw) g ,nw,t ∗ w



(46) Note that theWilliamson–Cross’s model leads toa Newtonian behaviourforparticularvaluesofnwandtw:namelynw=1and tw=1.Inthiscase,l=w/2.

4. Results

4.1. Validationofklameasurements

InTable5arecollectedthesetofexperimentswhichwillserve asdatabaseforthedimensionalanalysis.Itisconstitutedby150 measuresofkla(including70valuesfortheNewtoniancase), car-riedout at five rotational impellerspeeds (200≤N≤1000rpm) and fourflow rates (0.33≤Qg≤3.33L/min),and for several flu-ids(threeNewtonianfluids,fournon-Newtonianfluids).Notethat theseoverallvolumetricmasstransfercoefficientscorrespondto thevaluesobtainedwiththedynamicmethod,andaveragedfrom themeasuresforbothprobesandthreeruns(Eq.(9)).

Whateverthefluids,thecriterionforaxialhomogeneity(defined inEq.(10))hasbeenfoundvaryingfrom6to8%forN=200rpm, andfrom1to3%forN=1000rpm[19,37],thesmallestvaluebeing obtainedfortheprobelocatedatthetopofthetank.Then,no signif-icantspatialheterogeneitytakesplaceinthetank,andthesevalues ofklacanbeconsideredrepresentativeoftheaerationstateinthe wholetank.

(12)

kl a (s

-1

)

with physical method

kl a (s

-1

)

with chemical method Water 400 (1.6) Water 400 (3.33) Water 800 (1.6) Water 800 (3.33) Gly50 400 (3.33) Gly50 800 (3.33) XG2 400 (3.33) CMC4 400 (3.33) 0 10-2 2.10-2 3.10-2 0 10-2 2.10-2 3.10-2

Fig.3.Comparisonbetweenchemicalandphysicalmethodsformeasuringkla(the

dottedlinescorrespondtoadeviationof±15%;inthelegend,thefirstnumberisN inrpm,thenumberintobracketsQginL/min).

Fig.3 presents,for somerepresentativecases, a comparison betweenthephysicalandchemicalmethodsformeasuringkla.A goodagreementbetweenbothmethodsisobserved:thedeviation never exceeds 15% which corresponds to the order of magni-tudeassociatedwiththeexperimentaluncertaintyinthechemical method[27].Asaconsequence,thevaluesofklareportedinTable5 canbeassumedrelevantasvalidatedbytwomethods.

Thecontributionofsurfaceaerationtotheoverallvolumetric gas–liquidmasstransfercoefficienthasbeenestimatedby mea-suringklawithoutbubblingattheringsparger(seeSection2).For 200≤N≤1000rpmand1.6≤Qg≤3.33L/min,thefollowingtrends havebeenobtained[37]:

- forwater,kla

surf<0.14×kla

t,

-forglycerineat50%, kla

surf<0.15×kla

t,andforglycerineat

70%,kla

surf <0.23×kla

t,

-for CMCat4g/L, kla

surf <0.08×kla

t,and for CMCat 6g/L,

kla

surf≪kla

t,

-forxanthangumat1g/L, kla

surf<0.17kla

t,andforxanthan

gumat2g/Lkla

surf<0.14×kla

t.

Thisdemonstratesthatthesurfaceaerationremainssmallwhen comparedtovolumeaeration,confirmingthusthatkla|tisthe ade-quatetargetparametertotractinthedimensionlessanalysisfor qualifyingtheaerationstateinthetank.

Whenanalysedindetail[19],Table5pointsthat,foragiven fluid, theoverall volumetric mass transfer coefficients logically increase for increasingrotational impeller speeds and gas flow rates.Therelativecontributionsofgassparging(Qg)and mechan-icalagitation(N)onthevariationsofklaarecomparable,evenif therotationalimpellerspeedplaysamorepronouncedrole.The presentinvestigationsarethenperformedundertheintermediary conditiondefinedby[13,14],namelytheconditionrangedbetween the bubbling-controlling condition (at relatively high gas flow rates)andtheagitation-controllingcondition(atrelativelyhigh rotationalimpellerspeeds).Moreimportantisthemajor reduc-tioninklaobservedinpresenceofviscousfluids(Newtonianand non-Newtonian)whencompared towater.Tobetterappreciate thisphenomenon,thefollowingratiocanbedefined:

R=kla|viscousfluid

kla|water (47)

Forinstance,atQg=3L/minandfor200≤N≤1000rpm,thelatter ratioRvaries:

-from27to32%inglycerineat50%,andfrom9.7to20%inglycerine at70%,

-from43to66%inCMCat4g/L,andfrom38to43%inCMCat6g/L, -from50to83%inxanthangumat1g/Landfrom40to66%in

xanthangumat2g/L.

Thecomparisonof suchvalues ofkla,or anyotherattempts for their modelling, is usually made (see Section 1) by calcu-latingtheapparentviscositybasingonthewell-knownconcept ofMetzner–Ottoand theOstwald–de-Waele’s model (Eq.(14)). Whenapplyingthismethod[19],itispossibleneithertoexplain nortounderstandsatisfactorilytheseresults,confirmingthusthe requirement to perform more consistent investigations on the influencingparameters.

4.2. DimensionlessresultsforNewtonianfluids

The oxygen diffusivity, D, was unknown for the viscous Newtonian and non-Newtonian fluids under test. Indeed, such informationremainsunavailableintheliterature,andtheusual correlations(forexampletheWilke–Changone)cannotbeapplied bylackofsomerequireddata(forexampletheassociation fac-tor of solvent). So, the contributionof Schmidt number in the process relationship (Eq. (32)) cannot be rigorously sought. By default,whatevertheliquidphases,theoxygendiffusivity,D,will beassumedequaltotheoneinwaterat20◦C(i.e.to2×10−9ms−2). Inthiscase,theSchmidtnumberSc(definedwithrespecttogas cin-ematicviscosity,Eq.(31))isthenequalto7850.Asaconsequence, itseemsreasonabletoassertthattheprocessrelationship estab-lishedwillbeinsuredforvaluesofSchmidtnumbersclosetothis latter.Inaddition,thevariationofdensityforthefluidsinvestigated isnotimportant(Table2),implyingthusthatthechangein*is weak(847<*<1015).Asaconsequence,forsecuringtheprocess relationship,itischosentoignorethepossiblealterationsof*and ScinEq.(32),leadingto:

kla∗= f{Ug,Fr,,} (48) Havingnomechanisticindicationontheformofthef′-relation,the simplestmonomialformislookedfor:

kla∗=˛·(Fr)a ·(U∗

g)b·(∗)c·(∗)d (49) where ˛, a, b, c and d are respectively the constant and the exponentstowhichthedimensionlessFroude,gasvelocity, vis-cosityandsurfacetensionnumbersareraised.Thedimensionless viscosity, *, is calculated using the Newtonian viscosity (l) reportedin Table2. ThesoftwareAuto2fit® is usedtoperform themulti-parameteroptimizationrequiredtodetermine˛,a,b,c andd.Differentmathematicalalgorithmsaresystematicallytested (globalLevenberg–Marquardt,globalQuasi-Newton,standard dif-ferentialevaluation,geneticalgorithm)toverifythestabilityofthe resultsandtheirindependencywithinitialconditions.Themean standarddeviationiscalculatedfrom:

ε= 1 N

X

i=1,N

(kla)∗exp,i−(kla)∗mod,i (kla)∗ exp,i

(50)

Ina firsttime,it isinterestingtovisualizetheeffectof Newto-nianviscosity(*)on(kla)*separatelyfromtheothervariables. TheproblemisthatthegraphicrepresentationassociatedwithEq.

(49)requiresfivedimensionsas(kla)*dependsonFr,Ug∗,*and *.Asimplewaytosidestepthisproblemistocomedowntoa2D representation,inwhichtheimpactof*on(kla)*isneglectedand anidenticalexponentisimposedfortheFroudeanddimensionless superficialgasvelocitynumbers.ThisisillustratedinFig.4,where (kla)*isplottedasafunctionof(Ug∗·Fr)2/3forNewtonianfluids. Thevalueof2/3ischosenasafirstapproximationfortheexponent

(13)

Fig.4. EffectofdimensionlessNewtonianviscosity,kla∗=kla·(g/g2)1/3,versus

(U∗ g·Fr)

2/3

(thedotted/continuouslinescorrespondtothevaluespredictedbyEq.

(51)).

onFrandU∗

g,asalreadyencounteredintheliterature(seereviewof

[3]).Thus,Fig.4offerstheadvantagetoeasilyappreciatethe neg-ativeeffectof*onaerationperformances:whatever(U∗

g· Fr)2/3, anincreaseof*from54(water)to1886(glycerine70%)leadsto adrasticreductionof(kla)*(morethan85%).

When imposing the exponent 2/3 to (U∗

g· Fr), the best fit-tingbetweenexperimentaldataandEq.(49)withouttakinginto account*leadsto:

kla∗=0.1420·(Fr·U

g)2/3·(∗)−0.591 (51) Theassociatedmeanstandarddeviationisequalto17.1%.Totest therobustnessofsuchcorrelation(inparticularoftheexponents found),severalcasesarenowtestedwhenimplementingthe multi-parameteroptimization:

-CaseNo.1:alltheexponentsinEq.(49)arekeptfreeand*is neglected;

-CaseNo.2:theexponentsofFrandofU∗

gareimposedidentical withoutanyspecifiedvalue,and*isneglected;

-CaseNo.3:theexponentsofFrandU∗

gareimposedequalto2/3 (asinEq.(51)),butnowtheeffectof*istakenintoaccount.

InTable6arecollected,foreachcase,theconstantand expo-nentsofEq.(49)deducedfromthemulti-parameteroptimization. Inageneralpointofview,nomajordifferenceappearsbetweenthe differentcases:theexponentof*remainscloseto−0.59 (devia-tion<1.3%),andtheexponentsofFrandU∗

gdonotvarysignificantly whethertheyareimposedidenticalornot(deviation<6.5%).These findingsconfirmthattheordersofmagnitudeoftheexponents foundinEq. (51)are relevant,and thus,thatthis correlationis mathematicallyrobust.Notethat,evenifFrandU∗

ghavethesame exponent,thecontributionofNistwotimeshigherthantheone ofsuperficialgasvelocity,astheFroudenumberisexpressedin squaredrotationalimpellerspeed(Eq.(27));thisiscoherentwith theseobservationsmadeonthedimensionalresults(Section4.1). ThecasesNo.1andNo.2leadstoaslightimprovement(smaller than1%)ofthemeanstandarddeviation(ε);however,the expo-nentsofEq.(51)willbethereafterconserved,insofarastheyleadto themostsimpleformulation,andalsoasthevalueof2/3isalready

Table6

DimensionlessmodellingforNewtonianfluidsusingEq.(49)(inboldthecase retained). Case ˛ a b c d ε(%) No.1 0.02125 0.747 0.66 −0.605 – 16.1 No.2 0.01535 0.683 0.683 −0.592 – 16.8 No.3 0.2097 2/3 2/3 −0.591 −0.245 16.8 (kla )*exp (kla )*mod Water Gly50 Gly70 [µ * = 54, σ* = 73385] [µ * = 589, σ* = 45066] [µ * = 1886, σ* = 50704] 10-6 10-5 10-4 10-3 10-6 10-5 10-4 10-3

Fig.5.Experimentaldimensionlessoverallgas–liquidmass transfercoefficient versuskla*predictedfromEq.(52)(Newtonianfluids).Thedottedlinescorrespond

toadeviationof±20%.

encounteredintheliterature.Basedonthat,theeffectofsurface tension, ∗=/(3

g·4g·g) 1/3

,hasbeenaddedin Eq.(51)(Case No.3):theconstantisinreturnincreasedandanegativeexponent appearsfor*(−0.245).Thislatterlogicallyconfirmsthepositive impactofdecreasingsurfacetensiononaeration:whensurface ten-sionisreduced(forexamplewhenaddingsurfactantoralcoholin theliquidphase),smallerbubblesizesaregenerated,leadingtoa riseininterfacialareaandthusinoverallvolumetricmass trans-fercoefficient.Intheliterature[7],thisusuallyresultsinpositive exponentsfortheWebernumber.

Atlast,atagiventemperature,underagivenpressure,forthe geometryoftheaeratedstirredtankundertest(inparticularfor Hl/Tt=1,D/Tt=0.4,Ds/Tt=0.75andfortheothergeometricalratios characteristicsofthesystem),thedimensionalanalysisstatesthat theprocessrelationshipforNewtonianfluidsisexpressedby:

kla∗=0.2097·(Fr·Ug∗) 2/3 ·(∗)−0.591 ·(∗)−0.245 validfor



0.096<Fr<2.4, 0.0029<U∗ g<0.029, 54<∗<1886 847<∗<1015, 50704<<73385, Sc=7850 (52)

Fig.5comparestheexperimentaldatawiththedimensionless overallvolumetricgas–liquidmasstransfercoefficientspredicted byEq.(52).Agoodagreementisobserved,asthemeanstandard deviationremainssmallerthan17%.Itcanbeenobservedthatsome datacorrespondingtothesmallestkla*(obtainedforlowspeedsin glycerine)tendstomoveawayfromthestraightlinesrepresenting (kla∗)

exp= ±20%· (kla∗)pred.Thisshouldbelinkedtothefactthat, insuchconditions,thegas–liquidregimecorrespondstoaloading regime,andnottoacompletedispersionregimecharacterizingall theotheroperatingconditions.Themechanicalagitationisnotthen sufficienttodisperseuniformlythebubblesinthewholevolumeof tank,inparticularinthelowerpartofthetank(belowtheimpeller). Hence,therelativecontributionsofthemechanismscontrollingkla (mechanicalagitationagainstspargeraeration)deviatesfromthe onesactingwhenacompletedispersionregimetakesplace.When suchchangesinregimeoccurs,itbecomesthendifficulttodefine anuniqueandaccurateprocessrelationshipabletodescribethe entirerangeofoperatingconditions.Somedeviationscanbealso observedforthehighestvalesofkla;theycanbehereassociated withanincreasingcontributionofthesurfaceaeration(mass trans-feroccurringatthefreeliquidsurface)whenrotationalimpeller speedrises.Indeed,forthehighestN,asignificantvortexappears inthecentreofthetankandentrainsmanybubbles.Insuch condi-tions,themechanism(orregime)ofaerationisdeviatedtoapure volume-aeration,andthus,thechoiceofkla

t(andnotkla

surf)for

tractingvolume-aerationstateislessrepresentative.

Toconclude,itshouldbekeptinmindthatthevalidityofEq.

(14)

Table7

Dimensionlessnumberscharacteristicsforliquidproperties.

* * nost ∗ost nw t∗w ∗w Water 848 73385 1 54 1 1 108 Gly50 973 45966 1 589 1 1 1178 Gly70 1015 50704 1 1886 1 1 3773 CMC4 861 75905 0.642 1866 0.546 0.201 4944 CMC6 877 77316 0.527 5309 0.513 6.84×10−3 5.12×104 XG1 847 72276 0.543 539 0.411 9.98×10−5 1.02×105 XG2 855 77719 0.373 1366 0.281 3.84×10−5 1.59×106

numbersabovementioned,atgiventemperatureandpressure,and inthegeometrydefinedinSection2.

4.3. Dimensionlessresultsfornon-Newtonianfluids 4.3.1. WhenconsideringthemodelofOstwald–de-Waele

Asany valuecan beconsidered(see Section 3.2), the refer-enceshearrate, ˙ o,hasbeenarbitrarychosenequalto120s−1.The dimensionlessnumbersdescribingtherheologicalproperties asso-ciatedwiththeOstwald–de-Waele’smodel(Table2)andsuch ˙ o arecollectedinTable7(fourthandfifthcolumns).Itcanbe logi-callyobservedthattheflowindex(nost)decreaseswhenincreasing theconcentration,thesmallestvaluebeingobtainedforthemost concentrated solutionof xanthangum.Dimensionless apparent viscosity(∗

ost)arehigherforaqueoussolutionsofCMCthanthe onesofxanthangum.

BasingonthereasonsmentionedinthecaseofNewtonianfluids, thepossiblechangesin*andinScwillbeneglectedinEq.(38)

previouslyestablishedwhenthemodelofOstwald–de-Waeleis considered.Thus,Eq.(38)becomes:

kla∗=g



U∗g,Fr,,= a( o) g ,nost



(53) Thesimplestmonomialformisherealsolookedfortheg′-relation (nomechanisticinformationavailable):

kla∗=˛·(Fr)a′ ·(U∗ g)b ′ ·(∗)c′ ·(∗)d′ ·(nost)e (54) where ˛′, a, b, c, dand e are respectively the constant and theexponentstowhichthedimensionlessFroude,superficialgas velocity, viscosity,surface tensionand flow indexnumbers are raised(Auto2fit®).However,itisimportanttokeepinmindthat themodellingforNewtonianfluidsisadegradedcaseoftheone fornon-Newtonianfluids.Asaconsequence,theexponentsofFr, U∗

g,*and*(namelya′,b′,c′,d′)arealreadydefined,andwillbe thustakenequalrespectivelyto2/3,2/3,−0.591and−0.245asin theNewtoniancase(Table6,CaseNo.3).

Fig. 6.Effect of dimensionless apparent viscosityand flow index when the Ostwald–de-Waele’smodelisconsidered:(kla)*versus(Ug∗·Fr)2/3.

InFig.6,(kla)*isplottedasafunctionof(Ug∗·Fr)2/3forallthe fluids.Suchfigureisparticularlyimportantastheeffectsof*and nostcanbevisualizedseparately.Indeed,atagiven*,andecrease in nost clearly induces an increase in (kla)*, or in otherwords, theshear-thinningcharacteroffluidsfavourstheaeration perfor-mances,probablyduetothespatialheterogeneityofviscosity.This isillustratedwhencomparingeither(i)theglycerineat70%and theaqueoussolutionsofCMCat4g/Lwhichhavealmostthesame dimensionlessapparentviscosity(1886against1866),but differ-entflowindexes(1against0.642),or(ii)theglycerineat50%and theaqueoussolutionsofxanthangumat1g/Lwhichhavealmost thesamedimensionlessapparentviscosity(589against538),but differentflowindexes(1against0.543).Theimpactofnostseems howeverlesspronouncedthantheoneof*.

Fig. 7 compares the dimensionless overall volumetric mass transfercoefficientsmeasuredinpresenceofnon-Newtonianfluids with the ones predicted using the process relationship estab-lishedforNewtonianfluids(Eq.(52)).Thepointsrelatedtoeach non-Newtonianfluidareregroupedalongindividualstraightlines which are parallelto each others and tothe curves previously obtainedforNewtonianfluids.Thisillustratesthatthe introduc-tionofnost intothedimensionlessmodelling(Eq.(54))couldbe a meanfor differentiatingthe non-Newtonian fluidsfromeach othersandfromtheNewtonianones.Nevertheless,thisgroupof fourstraightlines(oneforeachfluid)isnotclassifiedaccording tothevalues offlowindex,in particulartheassociatednost are notdecreasingwhendeviatingfromtheNewtoniancurve(nost=1). Indeed, thestraight linethe closestfrom theNewtonian curve correspondstonost=0.543(XG1g/L),followedbynost=0.373(XG 2g/L),thennost=0.642(CMC4g/L),untilreachingthemostdistant line,nost=0.527(CMC6g/L).Thisfindingtendstoshowthat gather-ingallthedatarelatedtonon-NewtonianfluidsovertheNewtonian oneswillbedifficult byadding onlythe flow indexnost inthe relevantlist.Inotherswords,thiswouldmeanthatthematerial functionassociatedwiththemodelofOstwald–de-Waelecouldbe insufficientand/orunsuitablefordescribingcompletelytheeffect oftheshear-thinningpropertieson(kla)*.

1.0E-04 (kla )*exp (kla )*mod Water Gly50 Gly70 CMC4 CMC6 XG1 XG2 nost=0.543 nost=0.373 nost=0.642 nost=0.527 10-6 10-5 10-4 10-3 10-6 10-5 10-4 10-3

Fig.7.ModellingusingtheOstwald–de-Waele’smodel:comparisonbetweenthe experimental(kla)*andthevaluespredictedbyEq.(52)forNewtonianfluids.The

Figure

Fig. 1. Experimental set-up.
Fig. 2. Rheograms: (a) xanthan gum at 1 g/L, (b) xanthan gum at 2 g/L, (c) CMC at 4 g/L, and (d) CMC at 6 g/L.
Fig. 3. Comparison between chemical and physical methods for measuring k l a (the dotted lines correspond to a deviation of ±15%; in the legend, the first number is N in rpm, the number into brackets Q g in L/min).
Fig. 4. Effect of dimensionless Newtonian viscosity, k l a ∗ = k l a · ( g /g 2 ) 1/3 , versus (U ∗ g · Fr) 2/3 (the dotted/continuous lines correspond to the values predicted by Eq.
+4

Références

Documents relatifs

Kinetic parameter estimation of compounds in low-intermediate moisture food processed at high temperature (T &gt; 100 °C) represents a challenge since two aspects must be

 Remark 3.7 The relation between the two flavours of designs (dessins and desseins, or typing proofs and typable pseudo-designs) is of the same nature as the relation between proofs

Nous montrons que les équations du système discret endommageable sont équivalentes à une formulation en différences finies centrée d’un problème d’endommagement continu

The intensity gradient at relative level l ∗ corresponding to the interface (i.e. interface identification criterion) is plotted in figure 7 as a function of the local contrast for

From these equations, it could be deduced that to improve aeration performances in a microporous hollow fi ber membrane aerated bioreactor, the set of parameters should be chosen

Besides, bubbles tended to coalescence close to the diffusion mesh, leading to bigger bubbles in the bulk of the column and resulting in slug flow or annular-slug flow for some of

Jacques Coupry. Le Maître du ciel et la Dame de vie ?.. Des centurions primipiles peuvent être chargés du commandement de détachements ou, comme ici, de groupes de

The future efficiency of satellite networks will not only depend on the architecture or on the bandwidth available: transport protocols have to be aware of the network below or to