• Aucun résultat trouvé

Implementation of ‘chaotic’ advection for viscous fluids in heat exchanger/reactors

N/A
N/A
Protected

Academic year: 2021

Partager "Implementation of ‘chaotic’ advection for viscous fluids in heat exchanger/reactors"

Copied!
11
0
0

Texte intégral

(1)

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 : 17719

To link to this article :

DOI: 10.1016/j.cep.2016.07.010

URL :

http://doi.org/10.1016/j.cep.2016.07.010

To cite this version :

Anxionnaz-Minvielle, Zoé and Tochon, Patrice

and Couturier, Raphael and Magallon, Clément and Théron, Felicie

and Cabassud, Michel and Gourdon, Christophe Implementation of

‘chaotic’ advection for viscous fluids in heat exchanger/reactors.

(2016) Chemical Engineering and Processing: Process

Intensification, 113. pp. 118-127. ISSN 0255-2701

Any correspondence concerning this service should be sent to the repository

administrator:

staff-oatao@listes-diff.inp-toulouse.fr

(2)

Implementation

of

‘chaotic’

advection

for

viscous

fluids

in

heat

exchanger/reactors

Z.

Anxionnaz-Minvielle

a,

*

,

P.

Tochon

a

,

R.

Couturier

a

,

C.

Magallon

a

,

F.

Théron

b

,

M.

Cabassud

b

,

C.

Gourdon

b

aCEA,LITEN,DTBH,17ruedesMartyrs38054Grenoble,France

bUniversityofToulouse,LaboratoiredeGénieChimiqueUMR5503CNRS/INPT/UPS,31432Toulouse,France

Keywords: Heatexchanger/reactor Chaoticadvection Viscousfluids Continuousmode Processintensification Split-And-Recombinepattern ABSTRACT

Whenviscousfluidsareinvolved,laminarhydraulicconditionsandheatandmasstransferintensification are conflicting phenomena. A channel geometry based on Split-And-Recombine (SAR) patterns is experimentally investigated. The principle implements the Baker’s transformation and ‘chaotic’ structuresaregeneratedtopromoteheatandmasstransfer.Thisworkassessestheenergyefficiency ofdifferentheatexchanger/reactorsintegratingtheseSARpatterns.

Theheattransfercapacityisassessedandcomparedwiththeenergyconsumptionofeachmock-up.It issensitivetothecoolingmodeandtothenumberofSARpatternsperlengthunitaswell.

Thecontinuousoxidationof sodium thiosulfatewithhydrogenperoxide hasbeenimplemented. Conversionsupto99%arereachedaccordingtotheutilityfluidtemperatureandtheresidencetime.

Finally,thewholeperformancesoftheSARgeometriesarecomparedtoaplate-typeheatexchanger/ reactorwithacorrugatedpattern.Themoreviscousthefluid,themoretheenergyefficiencyoftheSAR designincreasescomparedtothecorrugateddesignbecauseofthebalancebetweenadvectionand diffusionmechanisms.TheinterestintermsofenergyefficiencyinworkingwithSARheatexchanger/ reactorappearsfromReynoldsnumbersbelow50.

1.Introduction

Among the technologies promoting process intensification

[1,2],heat-exchanger/reactor(HEXreactor)isapromisingone[3].

This device combines the benefits of a large heat transfer

performanceandaplug-flowregimeallowinganintensiveradial

mixing,asaresultofthespecificdesignofprocesschannels[4–9].

These geometries generate instabilities even in laminar flow

regime(50<Re<2000).Howevermostofthestudieshavebeen

carriedoutwithinviscidfluidsandtheintensificationof

perform-anceswhenthefluidviscosityincreasescomesattheexpenseof

pumpingcosts.Processesoffoodindustry,intermediatechemistry

likesiliconesorpolymers,... involveviscousfluidsandastudy,

carriedoutintheframeoftheindustrialnetworkEUROPIC[10],

pointedoutthatheatandmasstransfersinviscousmediaareone

ofthemainindustrialconcern.Intheverylaminarflowregime,

precludingturbulenceasamixingmechanism,mixingbydiffusion

canbeefficientprovidedthatthecontactareaissufficient.Thiscan

beachievedwithsystemsbasedonmulti-laminationmechanism

and baker’s transformation [11,12]. These have been proposed

especiallyinthecontextofmicrofluidics[13–16].Inmicrofluidic

devices,thetypicalchanneldimensionsandflowratesaresolow

that all flow is laminar (Re!0.1) and turbulence cannot be

achieved.Diffusionisalsotooslowtobeeffectiverequiringtoo

longchannellengths.Theimplementationofviscousfluidstreams

in HEX reactor and the typical millimetre dimensions of the

channel cross-section lead to similarconclusions. However,by

applyingaseriesof Baker’stransformationsindedicated

three-dimensionalmixingelements,namedSplit-And-Recombine(SAR)

patterns,therequiredchannellengthcanbereducedexponentially

with thenumber of mixing patterns.The involved separation/

stackingmechanismslieontheBaker’stransformationprinciple.

Two fluid streams are combined, split out-of-plane, rotated in

oppositedirectionsandrecombinedasillustratedwithFig.1.Then,

a 2-strips domainbecomes 2nalternatingstrips afternmixing

patterns.

*Correspondingauthor.

E-mailaddress:zoe.minvielle@cea.fr(Z. Anxionnaz-Minvielle).

(3)

AveryefficientmixingisachievedbydiffusionandSchönfeld

etal.[13]evenproposedasanoutlooktousesimilarSARstructures

toimproveheattransferinminiheatexchangers.Theinfluenceof

chaotic advection on heat transfer has been numerically and

experimentallyinvestigated[17–21].Resultsina singlechannel

arepromising.

TheinterestofHEXreactorsistheclosecontrolofthereaction

temperature.Theveryshortcharacteristiclengthsbetweenboth

theprocessandtheutilitystreamsareofparamountimportance.

Withthree-dimensionalSARstructures,characteristicdimensions

are quite different from ‘classical’ two-dimensional patterns

(plate-typeHEXreactor)andheattransfermightbeaffected.

ThustoinvestigatetherelevanceofimplementingSARpatterns

ina HEXreactordevice,performanceindicatorsbesides mixing

havetobeassessed.Theyaccountfor heattransferand energy

consumption.Theobjectiveistopromoteheatandmasstransfer

eveninviscousconditions.

For that purpose, we experimentally characterized three

stainlesssteelmock-ups"assembledbydiffusionbonding. Two

differentflowpatternsaccordingtothenumberofSARstructures

areconsidered.Theirdesignscomefrom[14,15].

For oneofthesegeometries,two coolingsystemshavebeen

investigated, external cooling Plates "isothermal wall- and

integrated cooling tubes. For each mock-up, the 3mm square

cross-sectionchannelsarearound3.5mlong.Then,acasestudyin

reactiveconditionsisimplementedwiththecontinuousoxidation

ofsodiumthiosulfatewithhydrogenperoxide.Finally,thewhole

performancesoftheSARgeometrieshavebeencomparedwiththe

onesofacorrugatedpatternplate-typeheatexchanger/reactor[6].

2.Heat-exchangerreactordesignandintegration

2.1.SARpatterns

ThreeSARpatterns,basedonthedesignsof[14,15]havebeen

considered. Fig. 2 depicts four patterns of each considered

geometry.

The square cross-sections of the channels are 3#3mm2

correspondingtoanequivalenthydraulicdiameterof3mm.The

Reynoldsnumberassessmentisbasedonthishydraulicdiameter

andonthefluidvelocityattheinletoftheprocesschannel,i.e.

upstreamthefirstfluidstreamseparation.Thisreferenceistaken

since the inlet flowrate is a major process parameter for the

industrialapplication.

TwochannellengthscanbedefinedintheSARgeometries.We

considerthedeveloped length,Ldevand thetotallength,L. The

former corresponds tothe distance travelled by a single fluid

particle between the inlet and the outlet of the HEX reactor.

Whereas the latter corresponds to the sum of every stream

branches,i.e.thetotalfluidvolumedividedbythecrosssection.

PatternsSAR-1andSAR-2areverysimilar.Thedifferencestems

fromthepatterndimensiononthez-axiswhen fluidstreamis

split.IncomparisonwithSAR-2,theSAR-1patternisexpandedto

allowtheinsertionofcross-flowcoolingtubes(seeSection2.2).

SAR-3patternislesscomplexthanSAR-1andSAR-2patterns.It

makeseasieritsmanufacturing.It includes3bendsperpattern

versus6bendsperpatterninbothSAR-1andSAR-2geometries.

2.2.HEXreactorintegrationandmanufacturing

ThethreeSARgeometriesareintegratedinrespectivelythree

wholeHEXreactorprototypes.TomanufacturetheHEXreactors,

theSARchannelisfirstdividedinseverallayerswhicharethen

transposedincorrespondingplatesasdepictedinFig.3.

The HEX reactors are made of one process plate including

severalrowsofSARpatternsinseriesandsensorconnectionsat

eachendoftherows(seeFig.4b).

ThestratificationstepdepictedinFig.3 isduplicatedonthe

platetoobtaintherequirednumberofrows(seeFig.4a).Laser

machiningisusedandplatesarethenstackedintoacontainer.The

assembling process is based on diffusion bonding (High

Nomenclature

A [m2]heatexchangearea

Cp [Jkg"1K"1]thermalcapacity

dh [m]hydraulicdiameter

Fp [kgh"1]processflowrate

Fu [kgh"1]utilityflowrate

L [m]totallengthoftheprocesschannel

Ldev [m]developedlengthoftheprocesschannel

_ni [mols"1]initialmolar

flowrate

Ploss [W]thermalloss

Pth [W]thermalpower

Preaction [W]heatofreaction

Pe [-]pecletnumber(=RePr)

Pr [-]prandtlnumber

Re [-]reynoldsnumber

T [K]temperature

u [ms"1]

fluidvelocity

U [Wm"2K"1]globalheattransfer

coefficient

V [m3]volumeof

fluid

Greekletters

D

Hr [kJmol"1]enthalpyofreaction

D

P [Pa]pressuredrop

e

[Wm"3]energydissipationrate

L

[-]darcycoefficient

l

[Wm"1K"1]thermalconductivity

m

[Pas]viscosity

r

[kgm"3]density

x

[%]conversionrate

Fig.1.Illustrationofa3Dmixingpattern.Top:Cross-sectionillustratingtheBaker’s transformationaftertwoiterations.Bottom:sketchofmixingactionsaftertwo elements[11].

(4)

temperatureIsostaticPressing–HIP–assembling,[22]).Thefinal

prototypeismadeofstainlesssteel.

Twocoolingconfigurationshavebeenconsidered.Thefirstone

assumesisothermalwalltemperature. Theprocessplate(SAR-2

andSAR-3)issandwichedbetweentwocoolingplates.Thesecond

cooling configuration involvescross-flow coolingtubes directly

integratedthroughtheSARpatterns(SAR-1).Fig.5illustrates

SAR-1andSAR-3integration.

Thefinalprototypesarearound185mmlongand90mmwidth.

ThegeometricalcharacteristicsarelistedinTable1.

2.3.Corrugatedpattern

TheperformancesofthepreviousSARpatternsarecompared

withaclassical2Dcorrugatedonewhichdesignisbasedontheone

studiedby[5,6].Themanufacturingprocessofthe2DHEXreactor

isthesameandthecoolingsystemisbasedonisothermalwalls

(withcoolingplates).Thesquarecross-sectionofthecorrugated

channelis2#4mm2andthehydraulicdiameteris2.67mm.The

channelis2.2mlongincluding103bends(90$)perlengthunit.The

corrugatedchannelisshowninFig.6.

3.Thermalandhydraulicinvestigations

ToassesstheHEXreactorsperformances,thestudiedReynolds

number rangesfrom0.1 to10,000.Waterand twosolutions of

glycerolintowater(70%w.and90%w.ofglycerol)areused.

3.1.Pressuredrop

3.1.1.Experimentalset-upandprocedure

Pressuredropsaremeasuredwithadifferentialpressuresensor

betweentheinletandtheoutletoftheHEXreactors.Tocovera

widerangeof Reynoldsnumber,distilled water,ethylene glycol

andtwosolutionsofglycerolareused(70%w.and90%w.).The

tests are implemented in isothermalconditions and the mean

physico-chemicalpropertiesofthefluidsarelistedinTable2.Since

thetemperatureismeasuredduringthepressuredrop

measure-ments,boththeviscosityandthedensityareassessedattheexact

testtemperature.

FrompressuredropmeasurementstheDarcycoefficient,

L

,is

assessedaccordingtothefollowingexpression:

L

¼ 2&

D

P&dh

r

&Ldev&u2

ð1Þ

AndtheReynoldsnumberisdefinedas:

Re ¼

r

&u

m

&dh ð2Þ

whereLdevisthedevelopedlengthoftheprocesschannel,anduthe

processfluidvelocity.TheSARgeometriesaredesignedsuchasthe

pressuredropsineachfluidbranchareequal.Moreover,sincea

fluid particle flows successively through the inlet zone (flow

velocity, u), a split branch (flow velocity, u/2), a stream

Fig.2.SARpatterns.ThereddottedlinesdelimitoneSARpattern.(Forinterpretationofthereferencestocolorinthisfigurelegend,thereaderisreferredtothewebversionof thisarticle.)

Fig.3. StratificationoftheSAR-2patternforthemanufacturingprocess.

(5)

recombinationzone(flowvelocity,u),asplitbranch(flowvelocity,

u/2),... Wedecidedtoconsiderthedevelopedlengthtoassess

theDarcycoefficientintheSARgeometries.

3.1.2.Resultsanddiscussion

Fig. 7 showstheevolution of theDarcycoefficientwiththe

ReynoldsnumberinboththeSARandthecorrugatedchannels.

Nomajordiscontinuityisobservedbetweentheglycerol0.7w.

curve(Reynoldsnumberfrom5to90)andtheglycerol0.9w.one

(Reynoldsnumberfrom0.1to3).Whateverthegeometry,alinear

sharpdecreaseisobservedasforstraighttubes(

L

)64/Re)inthe

laminarflowregime.Aweakdecreaseoftheslopecanbenoticed

onFig.7whereasReynoldsnumberisfarbelowtheclassicalflow

regime transition around Re=2300 for straight channel. To

illustrate this low Reynolds number flow regime transition,

Fig.8showstheevolutionoftheDarcycoefficientintheSAR-1

channelonawiderReynoldsnumberrange.

AsobservedinFig.7,thelinearsharpdecreaseismeasuredfor

Reynoldsnumberrangingfrom10toaround100.Above100,the

transitionalzoneseemstoappearandfinallyamoderateslopeis

observedforhigherReynoldsnumbers.Thesetrendsaresimilarto

straightchannelsexceptthatthetransitionalzoneinourgeometry

isaroundRe=100ratherthanRe=2300.Flowinstabilitiesabove

these Reynolds number may certainly promote turbulence-like

vortices. This is interestingto promote heat and mass transfer

intensificationwhileworkingwithlowReynoldsnumbers.Similar

trendshavebeenobservedbyTheronetal.[6]inthecorrugated

channel.

InthelowReynoldsnumberzone,Fig.7showsdiscrepancies

betweentheSARpatterns.LossesintheSAR-3geometryarehigher

thaninSAR-1andSAR-2geometries(uptox2).Thismaybedueto

thenumberofpatternsperdevelopedlengthunitwhichisaround

50%higherinSAR-3thaninSAR-1and2(39.5vs25patterns/Ldev).

Thefluidundergoesahighernumberofsplittingand

recombina-tionwhichproduceslosses.Moreover,theratiobetweenthelength

ofthebrancheswherethefluidflowswithavelocityu/2andthe

Table1

GeometricalcharacteristicsofSARHEXreactors.

SAR-1 SAR-2 SAR-3

Coolingconfiguration cross-flowcoolingtubes Coolingplates(isothermalwall) Coolingplates(isothermalwall)

Cross-section(mm2) 3#3

Hydraulicdiameter(mm) 3

Developedlength,Ldev(m) 1.9 1.7 2.1

Totallength,L(m) 3.4 2.8 3.4

Totalvolume(mL) 28.3 24.7 28.5

Numberofrowsinseries 4 4 7

Numberofpatternsperrow 12 12 12

Numberofpatternsperunitofdevelopedlength 25 29 39.5

Fig.6.PhotographofthecorrugatedchannelbeforeHIPassembling.

Table2

Physico-chemicalpropertiesofthefluidsusedforpressuredropmeasurements.

Fluid Densityr(kgm"3) Viscositym(Pas) FlowrateFp(kgh"1) Reynolds,Re

Water"Glycerol(70%weightglyc.) 1030 0.018 1.4–18.0 5.3–110.9

Water"Glycerol(90%weightglyc.) 1015 0.210 0.3–7.1 0.1–3.3

Water"Ethyleneglycol(98%weighteth.Glyc.) 1110 0.010 1.0–27.5 9.0–316.0 Fig.5. CADviewsofSAR-1(aandb)andSAR-3(c)HEXreactors.(a)ZoomofthecrossingbetweenthecoolingtubesandtheSAR-1channel.(b)Thecross-flowcoolingtubesin SAR-1HEXreactoraredepictedinred.

(6)

developed length is lower in SAR-3 than in SAR-1 and 2. The

pressuredropisdirectlyproportionaltothesquareofthevelocity

andthisalsogeneratesadditionallossesinSAR-3pattern.

TheslightdiscrepancybetweenSAR-1andSAR-2curvesmight

beexplained aswellsince SAR-1 geometryhasbeenexpanded

compared withSAR-2.ThismeansthatthelengthofoneSAR-1

patternisslightlyhigherthantheoneoftheSAR-2pattern,i.e.the

fluidundergoesalittlebitlesssplittingandrecombinationperunit

ofdevelopedlength(29vs25patterns/Ldev).

Finally,the corrugated channel generateshigher losses than

SARgeometriesinthelowReynoldsnumberrange(Re<100).The

chaoticstructuresseemtogeneratemoderatelossescomparedto

theturbulent-likestructuresofthecorrugatedgeometry.

3.2.Thermalcharacterization

3.2.1.Experimentalset-upandprocedure

The HEXreactorsareequippedwithtemperaturesensorsto

characterizethetemperatureprofilesversustheflowregime.The

temperatureismeasuredbetweentheinletandoutletofeachrow

(seeFigs.4and5).Eitheracounter-currentflow(SAR-2,SAR-3)ora

cross-flow(SAR-1,corrugatedchannel)patternisimplementedfor

the cooling stream. The intermediate temperature probes are

required to locate the temperature pinch and to avoid

under-estimatingtheheattransfercapacity.

Tocharacterizethethermal behaviouroftheHEXreactors,a

heat transfer capacity, or thermal intensification factor, UA/V

(kWK"1m"3),isassessed.Uistheglobalheattransfer

coefficient

(Wm"2K"1),Aistheheatexchangerarea(m2)andVisthevolume

offluid(m3).

InboththeSARpatternsandthe2Dcorrugatedgeometry,the

heatexchangeareacanbedefinedaccordingtovariousreferences.

Itcanrepresentthedevelopedheatexchangeareaofthe4facesof

thechannel,or aprojected area(i.e.assumingnoheattransfer

limitations in the process plate material). The 3-dimensions

characteristic of the SAR structures add an extra level of

complexityand theresultsinterms of heattransfer coefficient

(and the resultant Nusselt number) depends on the chosen

reference.It mayleadtointerpretationmistakeswhentheHEX

reactorsarecompared.

Fig.7.DarcycoefficientvsReynoldsnumber(glycerol70%andglycerol90%at25$C).

(7)

Thethermalintensificationfactoristhusdirectlyassessedfrom

temperaturemeasurementsandnoheatexchangeareareferenceis

required: U&A ¼ Pth

D

Tml

ð3Þ

wherePth(W)istheexchangedpowerbetweentheprocessfluid

andthecoolingfluid:

Pth¼Fp:Cp:!Tp;in "Tp;1" ð4Þ

and

D

Tmlisthelogarithmicmeantemperature:

D

Tml ¼ Tp;in "Tu;out ! " "!Tp;1 "Tu;in" ln Tp;in"Tu;out Tp;1"Tu;in # $ ð5Þ

Tp,inaswellasTu,inaretheinlettemperaturesofprocessand

cooling (also named utility) fluids respectively. Tp,1 is the

temperaturemeasuredatthefirsttemperaturesensorconnection.

Tu,outistheoutlettemperatureoftheutilitystream.

3.2.2.Resultsanddiscussion

Fig.9displays theheattransfercapacityasa functionofthe

ReynoldsnumberforthethreeSARheatexchanger/reactors.

ThebehaviourofthethreeSARgeometriesisclearlydistinct.

Thehighratioofsplittingandrecombinationperunitoflengthof

SAR-3geometry(+50%comparedwithSAR-1andSAR-2

geome-tries) seems clearly to favour heat transfer. Despite a small

differencebetweenthenumberofpatternperunit oflengthin

SAR-1andSAR-2geometries,respectively25pattern/Ldevand29

pattern/Ldev,anincreaseoftheheattransfercapacityupto30%is

measuredinSAR-2 heat exchangerreactor.To integratethe 22

cooling tubes through the SAR-1 channel in a cross-flow

configuration,theirdiameterislimited(2mm).Becauseofhigh

pressuredrops,thislimitstheutilityflowrate.Thecontributionof

theutilitystreamthermalresistancemaybenolongernegligiblein

the global heat transfer assessment as it is for the plate-type

cooling(SAR-2configuration).Asaconsequence,theincreaseof

theprocessflowrateinSAR-1channeldoesnotcontributedirectly

totheincreaseoftheheattransfercapacity.Thisemphasizesthe

differencebetweenSAR-1andSAR-2geometries,especiallywhen

theReynoldsnumberincreasesaboveRe=200.

AboveRe=150,theslope oftheSAR-3curvediminishesand

heattransfercapacityofSAR-2configurationbecomeshigher.For

suchReynoldsnumber,advectionmaybecomethemostdominant

phenomena(overdiffusion)andtheflowismainlygovernedby

secondaryandDeanvortices.Heattransferthendependsonthe

numberof90$bendsperunitofdevelopedlengthwhichishigher

intheSAR-2patternthanintheSAR-3one(respectively173bends/

mvs118bends/m).

Fig.10illustratestheeffectoffluidviscosity.Theheattransfer

capacityisplottedasafunctionoftheReynoldsnumberforamore

viscousfluid(Pr)200).

ItisconsistentwithresultsdisplayedinFig.9belowRe=150.

ForlowReynoldsnumber,advectionisnomorethedominantflow

mechanism and a balance is made between advection and

diffusion. As a consequence, the number of SAR patterns/Ldev

influencestheheattransfercapacity.

3.2.3.Comparisonwiththe2Dcorrugatedgeometry

Fig.11comparestheheattransfercapacityofeachHEXreactor

(bothSARgeometriesand2Dcorrugatedone)asafunctionofthe

energydissipationrate

e

(Wm"3).Thisparameterrepresentsthe

requiredpumpingpowerandisdefinedas:

e

¼Fp:

D

P

r

:V ð6Þ

Above 1000Wm"3, the heat transfer capacity of the 2D

corrugated channel is clearly higher than the one of the SAR

geometries.It corresponds toReynoldsnumber around100. As

observedin Fig.9,theflowis governedbyadvectionandDean

vorticesallowtheintensificationofheattransferwhilegenerating

moderatefrictionlosses.

Below 1000Wm"3, heat transfer capacities are similar in

functionoftheenergydissipationratewhatevertheflowpattern.

This is an interesting result considering both mixing and heat

transfercapacity.Indeed,Ghanemelal.[23]showedthattheSAR

geometries promote mixing when compared to 2D corrugated

geometryandFig.11showsthatheattransferinSARpatternscan

competewithcorrugatedchannelperformances.Wecanconclude

thattheyaregoodcandidateforchemicalsynthesis

implementa-tionwithviscousfluids.

The Fig.12 illustrates the effect of Prandtl number on the

thermalandhydraulicperformancesofthestructuredflows.The

exchangedthermalpowertothepumpingpowerratioisplotted.It

isfunctionofthegeometry,thePrandtlnumberandtheReynolds

number.

(8)

Fig.12 showsthatwhateverthegeometryandthefluid,the

Pth/

e

ratiodecreaseswhen theReynoldsnumberincreases.This

means that the heat transfer intensification regarding energy

saving ismoresignificantatlowerflow regime.WhenSAR and

corrugatedgeometriesarecompared,twobehavioursareobserved

dependingonthePrandtlnumber.ForlowPrandtlnumber(water),

thepowerratioishigherinthecorrugatedchannelthanintheSAR

patterns(x2)whereasforhighPrandtlnumbers(glycerol70%w.)

powerratiosaresimilar.Inourcase,theincreaseinPrandtlnumber

favours the diffusion effect vs advection. As a consequence,

advectionwhich appearsas thepredominantphenomenon and

seemstogovernthewaterflow(lowPrandtlnumber)favoursthe

powerratiointhe2DcorrugatedgeometrycomparedwiththeSAR

ones.Whentheviscosityincreases,thebalancebetweenadvection

anddiffusioneffectsisshiftedandtheSARmechanismseemsto

becomeinterestingoveradvection.SAR-3patternshowsalower

powerratiothanothergeometries.Thisisconsistentwithresults

displayedinFigs.7and9,thehigherheattransfercapacity(see

Fig.9)doesnotcompensatethehigherpressuredrop(seeFig.7).

4.Implementationofacasestudy"exothermalreaction

4.1.Theoxidationreaction

Basedonthepromisingresultsofthethermalcharacterization,

anexothermicreactionisimplementedintheSAR-1HEXreactor.

ThisistheoxidationofsodiumthiosulfateNa2S2O3byhydrogen

peroxideH2O2:

2Na2S2O3+4H2O2!Na2S3O6+Na2SO4+4H2O

Thisreactionoccursinhomogeneousliquidphase,is

irrevers-ible,fastanditsreactionheatis

D

Hr="586.2kJmol"1.Sinceitis

temperature sensitive, i.e. the conversionrate depends on the

Fig.10. HeattransfercapacityvsReynoldsnumber(glycerol90%weight"Pr)200).

(9)

operatingtemperature,thisreactionisanidealcandidateforcase

study[4,24–28].

4.2.Experimentalset-up

The implemented experimental set-up and procedure are

similartotheonesdescribedbyTheronetal.[6].Theoxidation

reactioninvolvestwoaqueoussolutionsofsodiumthiosulfateand

hydrogenperoxideasreactants.Toavoidtoohightemperatures

insidethereactorincaseofanadiabaticrisethetwosolutionsare

bothpreparedinorder toreach9%inmass ofeach reactantin

water. Two operating temperatures, controlled by the utility

streamtemperature,aretested: 40 and50$C.For these

experi-mentsthereactoristhermallyinsulatedthankstoapolystyrene

jacket.

TheoperatingconditionsarelistedinTable3.

Theconversionrate

x

inthereactorisdeducedfromthereactor

thermalbalancebetweentheprocessandutilityinletsandoutlets

atsteadystate.Theconversionisthuscalculatedasfollows:

x

¼Ploss þFp:Cpp: Tp;out "Tp;in

!

" þFu:Cpu: Tu;out "Tu;in

! "

Preaction ð11Þ

where

FpandFuarethemassflowrateinrespectivelytheprocessand

theutilitychannels

Preactionisthetotalheatofreaction:

Preaction ¼ _ni:

D

Hr ð12Þ

where _ni is the initial molar flowrate of the limiting reactant

(Na2S2O3).

Plossistheheatloss.Itismeasuredfrompreliminaryunreactive

tests.Boththeprocessandtheutilitylinesarefedwithwateratthe

temperatures and flowrates targetedfor thereactiontests. The

heatlossesareassessedwhensteadystateisreached.

4.3.Resultsanddiscussion

ThemeasuredconversionratesaregiveninTable3 foreach

experiment.

As expected the conversion rate increases when increasing

eithertheresidencetimeinthereactororthetemperatureofthe

utilitystream.Itisinterestingtonotethatautilitytemperature

below40$CalmostinhibitsthereactionintheHEXreactor.Asa

consequencetheminimalutilitytemperatureof40$Callowsthe

initiationofthereactionandthenthecontrolofthetemperature

riseduetothereactionexothermicity.

TheFig.13showstheconversionprofileversustimeobtained

fromtheGrauetal.[29]kineticlawandtheexperimentaldatafor

boththe2DcorrugatedHEXreactor(see[6],T=42.9$C)andthe

SAR-1one(T=47.5$C).

At fixed utility inlet temperature the different flowrates

experimentallytested enabletocompare theconversiontothe

kineticdata reportedintheliterature at differentstagesof the

reaction.

TheexperimentaldatadisplayedinFig.13areobtainedwitha

utilitytemperatureequalto40$C.Theequilibriumtemperatureat

theprocessoutletishigherintheSAR-1HEXreactor(47.5$C)than

inthecorrugatedone(42.9$C).Thisis consistentwiththeheat

transfer characterisation (see Fig. 10) which showed that for

inviscidfluidheattransfercapacityofthecorrugatedgeometryis

higher.

TheFig.13showsthatforbothconversionratesof88and94%

thekineticmodelresultsfitwellwiththeexperimentaldatasince

thediscrepancyislowerthan6%.Ahigherdiscrepancy (15%)is

Fig.12.ThermalexchangedpowertopumpingpowerratiovsReynoldsnumber(glycerol70%weight"Pr)40andwater"Pr)7).

Table3

Operatingconditionsoftheoxidationtests.

No.exp. Processside Utilityside Conversionrate(%)

FH2O2(kgh"1) FNa2S2O3(kgh"1) Reynoldsnumber Residencetime(s) Tutility($C)

1 1.7 3.4 472 20 40 94

2 2.4 4.8 667 14 40 88

3 3.4 6.8 944 10 40 59

(10)

obtainedatthelowresidencetime.Theprofileofconversionvs

time obtained from Grau et al. data [29] assumes a constant

temperature.HoweverintheSARHEXreactor,whentheresidence

timedecreases,i.e.whentheflowrateincreases,thetemperature

profile along the SAR channel is less uniform than for longer

residencetimes.Almost0.5m,i.e.aquarteroftheprocesschannel

developedlength,isrequiredtoheatthereactantuptothedesired

temperature(40$C).Inthispartofthechannel,thetemperatureis

thustoolowtoinitiatethereaction.Asaconsequencetheeffective

residencetime during which reactantsareconverted is shorter

than the theoretical one reported in Fig.13 (Fp/V=10s). This

explainswhythediscrepancybetweentheconversionpredicted

fromGrauetal.kinetics[29]andthemeasuredonemayincrease

whentheflowrateincreases.

However,theresultsobtainedinthisstudyshowreasonable

agreementwithGrauet al.[29] kineticparameterswhich have

been obtained from batch experiments, especially for long

residence time, ie low flowrates (Re=573 and 802). As a

consequence itconfirms that themixing time is not a limiting

parameterforsuchafastreaction.Moreover,sinceheattransfer

capacityintheSARHEXreactorishighenough,thereactioncanbe

implemented at higher temperature to accelerate the rate of

reactionwithoutsafetyissues.

5.Discussion/Conclusion

Thermalandhydraulicbehaviourofastructuredchaoticflow

hasbeencharacterizedinheatexchanger/reactorswithchannel

designsbasedonSplit-And-Recombinepatterns.Theprincipleis

based ontheBaker’s transformation and chaotic structuresare

generated to promote heat and mass transfer intensification.

Intensive mixing performances of such geometry have been

demonstratedinpreviousstudies[13,23]and thiswork

investi-gatedtheenergyefficiencyofaheatexchanger/reactorintegrating

theseSARpatterns.

Three stainless steel mock-ups "assembled by diffusion

bonding-havebeenexperimentallycharacterized.Twodifferent

flow patterns according to the number of SAR structures are

available.Foroneofthesegeometries,twocoolingsystemshave

beeninvestigated,externalcoolingPlates"isothermalwall-and

integrated cooling tubes. For each mock-up, the 3mm square

cross-sectionchannelsarearound3.5mlong.

The SAR based reactors have been compared with a 2D

corrugatedgeometry.Thenumberoffluidsplittingand

recombi-nation per unit of developed length is a design parameter

promotingheattransferbutalsogeneratinghigherfrictionlosses.

Itresultsinsimilarperformancesconsideringtheenergy

dissipa-tionrate.Likewise,similarheattransfercapacitiesaremeasured

betweenSAR-1 and SAR-2geometriesregardless of thecooling

system. The integrated cooling tubes system which is more

complex to implement than the sandwiched plates seems not

necessary.

Accordingtotheflowregimeabalancebetweenadvectionand

diffusion exists. The former becomes predominant for high

ReynoldsnumberorlowPrandtlnumber.Turbulent-likestructures

(likeDeanvortices)appearandgovernthetransfermechanisms.In

these flow conditions, the design parameter seems to be the

number of 90$ bends/m and thecorrugated geometry is much

moreperformant.Butwhenthefluidviscosityincreases,the

Split-And-Recombinemechanismseemstopromotethechaoticnature

oftheflow andcontributes withadvection totheheat transfer

intensification.

Considering theintensification of mixing in SAR geometries

comparedwithcorrugatedone,maintaininghighheattransferata

similarlevelthanthecorrugatedgeometryisaninterestingresult

regardingapplicationsinchemicalindustryhandlingviscousfluids

(food industry, bulk chemistry, polymers,...). Implementing

moreviscousfluidscouldpossiblyhelptocompletelydissociate

theinfluenceofadvectionversusdiffusion.

A case study with the continuous oxidation of sodium

thiosulfatewithhydrogenperoxidehasbeenstudied.Conversions

havebeenmeasuredaccordingtothecoolingfluidtemperature

andtheresidencetime.Conversionsupto99%havebeenreached

withresidencetimearound13sandconfirmedthatmixingtimeis

nota limiting parameterforsucha fastreaction.Howeverthis

reaction involves inviscid fluids (viscosity close to water) and

advectiongoverns probablythe flow.It could beinteresting to

completetheseresultswithanexothermicreactionwithviscous

media.

Finally,SARpatternscouldbecandidateforscale-upprocedure.

Whenincreasingthehydraulicdiameterabalancebetweenthe

increaseoftheDeannumber(coupledtoanincreaseofthelosses)

andtheuniformityofthemulti-laminatedfluidstreamshouldbe

thekeyissue.

References

[1]A.I.Stankiewicz,J.A.Moulijn,Processintensification:transformingchemical engineering,Chem.Eng.Prog.96(2000)22–34.

[2]A.Cybulski,J.A.Moulijn,A.I.Stankiewicz,NovelConceptsinCatalysisand ChemicalReactors:ImprovingtheEfficiencyfortheFuture,Wiley-VCH, Weinheim,2010.

[3]Z.Anxionnaz,M.Cabassud,C.Gourdon,P.Tochon,Heatexchanger/reactors (HEXreactors):conceptstechnologies:state-of-the-art,Chem.Eng.Proc.47 (2008)2029–2050.

[4]Z. Anxionnaz, M. Cabassud, C. Gourdon, P. Tochon, Transposition of an exothermicreactionfromabatchreactortoanintensifiedone,HeatTransfer Eng.31(2010)788–797.

[5]Z.Anxionnaz-Minvielle,M.Cabassud,C.Gourdon,P.Tochon,Influenceofthe meanderingchannelgeometryonthethermo-hydraulicperformancesofan intensifiedheatexchanger/reactor,Chem.Eng.Process.ProcessIntensif.73 (2013)67–80.

[6]F. Theron,Z. Anxionnaz-Minvielle, M.Cabassud, C. Gourdon, P. Tochon, Characterizationoftheperformancesofaninnovativeheat-exchanger/reactor, Chem.Eng.Process.ProcessIntensif.82(2014)30–41.

[7]M.Moreau,N.DiMiceliRaimondi,N.LeSauze,M.Cabassud,C.Gourdon, Pressuredropandaxialdispersioninindustrialmillistructuredheatexchange reactors,Chem.Eng.Process.ProcessIntensif.95(2015)54–62.

[8]A.Ghanem,C.Habchi,T.Lemenand,D.DellaValle,H.Peerhossaini,Mixing performancesofswirlflowandcorrugatedchannelreactors,Chem.Eng.Res. Des.92(2014)2213–2222.

[9]C.Habchi,S.Russeil,D.Bougeard,J.L.Harion,T.Lemenand,D.DellaValle,H. Peerhossaini,Enhancingheattransferinvortexgenerator-type

multifunctionalheatexchangers,App.ThermalEng.38(2012)14–25. [10]EuropeanProcessIntensificationCenter,EUROPIC,http://europic-centre.eu/,

2016(accessed02.29.2016).

[11]P. Carrière, On a three-dimensional implementation of the Baker’s transformation,Phys.Fluids19(2007)118110.

[12] F.Raynal,P.Carrière,Thedistributionoftimeofflightinthreedimensional stationarychaoticadvection,Phys.Fluids27(2015)1.4918750.

[13]F.Schönfeld,V.Hessel,C.Hofmann,Anoptimisedsplit-and-recombine micro-mixerwithuniform‘chaotic’mixing,LabChip—Miniat.Chem.Biol.4(2004) 65–69.

Fig. 13.Conversionratevs.residencetime(Tu=40$C).Comparisonofthetheoretical conversionwiththeexperimentalones.

(11)

[14]B.L.Gray,D.Jaeggi,N.J.Mourlas,B.P.vanDrieënhuizen,K.R.Williams,N.I. Maluf,G.T.A.Kovacs,Novelinterconnectiontechnologiesforintegrated microfluidicsystems,Sens.ActuatorsAPhys.77(1999)57–65.

[15]H.Chen,J.-S.Meiners,Topologicmixingonamicrofluidicchip,Appl.Phys.Lett. 84(2004)2193–2195.

[16]J.Branebjerg,P.Gravesen,J.P.Krog,C.R.Nielsen,Fastmixingbylamination, Proc.IEEEMicroElectroMech.Syst.(MEMS)(1996)441–446.

[17]M.Creyssels,S.Prigent,Y.Zhou,X.Jianjin,C.Nicot,P.Carrière,Laminarheat transferintheMLLMstaticmixer,Int.J.HeatMassTransfer81(2015)774–783. [18]A.Ghanem,T.Lemenand,D.DellaValle,H.Peerhossaini,Optimizedchaotic heatexchangerconfigurationsforprocessindustry:anumericalstudy,Proc. ASMEFluidsEng.Div.SummerMeet.(2013),InclineVillage,USA. [19]A.Mokrani,C.Castelain,H.Peerhossaini,Theeffectsofchaoticadvectionon

heattransfer,Int.J.HeatMassTransfer40(1997)3089–3104.

[20]C. Chagny, C. Castelain, H. Peerhossaini, Chaotic heat transfer for heat exchangerdesignandcomparisonwitharegularregimeforalargerangeof Reynoldsnumbers,Appl.Therm.Eng.20(2000)1615–1648.

[21]V.Kumar,S.Saini,M.Sharma,K.D.P.Nigam,Pressuredropandheattransferstudy intube-in-tubehelicalheatexchanger,Chem.Eng.Sci.61(2006)4403–4416. [22]R.Couturier,C.Bernard,J.M.Leibold,F.Vidotto,P.Tochon,Deviceforminga

chemicalreactorwithimprovedefficiency,comprisingaheatexchanging circuit,PatentnoWO2011/083163Al,2011.

[23]A.Ghanem,T.Lemenand,D.DellaValle,H.Peerhossaini,Transportphenomena inpassivelymanipulatedchaoticflows:Split-and-recombinereactors,Proc. ASMEFluidsEng.Div.SummerMeet.(2013),InclineVillage,USA.

[24]L.Prat,A.Devatine,P.Cognet,M.Cabassud,C.Gourdon,S.Elgue,F.Chopard, Performanceevaluationofanovelconcept“openplatereactor”appliedto highlyexothermicreactions,Chem.Eng.Technol.28(9)(2005)1028–1034. [25]M.D. Grau, J.M. Nougues, L. Puigjaner, Batch and semibatch reactor

performanceforanexothermicreaction,Chem.Eng.Proc.39(2000)141–148. [26]A.M.Benkouider,J.C.Buvat,J.M.Cosmao,A.Saboni,Faultdetectionin semi-batchreactorusingtheEKFandstatisticalmethod,J.LossPrev.ProcessInd.22 (2009)153–161.

[27]L.Vernieres-Hassimi,M.A.Abdelghani-Idrissi,D.Seguin,Experimentaland theoreticalsteadystatemaximumtemperaturelocalizationalongan exothermictubularreactor,OpenChem.Eng.J.2(2008)57–65.

[28]W.Benaissa,S.Elgue,N.Gabas,M.Cabassud,D.Carson,M.Demissy,Dynamic behaviourofacontinuousheatexchanger/reactorafterflowfailure,Int.J., Chem.React.Eng.6(2008).

[29]M.D. Grau, J.M. Nougues, L. Puigjaner, Batch and semibatch reactor performanceforanexothermicreaction,Chem.Eng.Proc.39(2000)141–148.

Figure

Fig. 1. Illustration of a 3D mixing pattern. Top: Cross-section illustrating the Baker’s transformation after two iterations
Fig. 3. Stratification of the SAR-2 pattern for the manufacturing process.
Fig. 7 shows the evolution of the Darcy coefficient with the Reynolds number in both the SAR and the corrugated channels.
Fig. 8. Evolution of the Darcy coefficient vs Reynolds number in the SAR-1 channel (Water and ethylene glycol &#34; 25 $ C).
+4

Références

Documents relatifs

Among the numerous options to intensify a process, the transposition from a batch reactor to a continuous plug flow reactor is a good alternative when the selectivity and the

Despite a small difference between the number of pattern per unit of length in SAR-1 and SAR-2 geometries, respectively 25 pattern/L dev and 29 pattern/L dev , an increase of the

In the case when the Mach number is of higher order than the Rossby number, we prove that the limit dynamics is described by an incompressible Oberbeck- Boussinesq system, where

To test these hypotheses I examined the variation in tree root functional traits (across- and within-species), and its consequences for fluxes of C, N and P at the

There are thirteen major steps included in this process (Fig. Because the Pyrex wafer is about 100m thick, the bonding voltage is reduced to 500V. The entire stack is

It is shown that by adding a suitable perturbation to the ideal flow, the induced chaotic advection exhibits two remarkable properties compared with a generic perturbation :

The efficiency of this DCE was compared to a smooth duct heat exchanger of equal length and hydraulic diameter, at the same flow rate, and under the most favorable

Le calcul des poteaux mixtes acier béton soumis à la flexion composée uni axiale et bi axiale avec un effort normal de compression tel que présenté par l’Eurocode 4 se base