Contents lists available atScienceDirect
Chemical Engineering Research and Design
j o u r n a l h o m e p a g e :w w w . e l s e v i e r . c o m / l o c a t e / c h e r d
Mixing performance in Split-And-Recombine Milli-Static Mixers—A numerical analysis
Charbel Habchi
a, Akram Ghanem
b, Thierry Lemenand
c, Dominique Della Valle
d, Hassan Peerhossaini
e,f,∗aNotreDameUniversity—Louaize,ThermofluidsResearchGroup,P.O.Box:72ZoukMikael,ZoukMosbeh,Lebanon
bELISAAerospace,48RueRaspail,02100Saint-Quentin,France
cAngersUniversity,ISTIA,LARISEA7315,49000Angers,France
dONIRIS,44300Nantes,France
eUniversitéParisDiderot,SorbonneParisCitéUniversity—EnergyPhysicsGroup,AstroParticlesand Cosmology-CNRSUMR7164,75013Paris,France
fUniversityofWesternOntario,MechanicsofActiveFluidsLaboratory,Civil&EnvironmentalEngineeringand Mechanical&MaterialsEngineering,London,ONN6A3K7,Canada
a r t i c l e i n f o
Articlehistory:
Received24May2018 Receivedinrevisedform2 December2018
Accepted12December2018 Availableonline21December2018
Keywords:
Mixingenhancement CFDstudy
Staticmixer Creepingflow Baker’stransform
a bs t r a c t
VerylowReynoldsnumber laminar flowin Split-And-Recombine(SAR)static mixersis numericallyinvestigatedbyusingafinitevolumemethod.Intheseconfigurations,advective chaoticflowstructuresarecreatedbythepassivecontroloftheflow,mimickingthebaker’s transformation,throughaseriesofflowsplitting,rotations,andre-combinationsthatgen- erateintertwinedlamellarstructures.Thisprocessleadstoextrasurfacecreation,which ultimatelyintensifiesmasstransfer.Themainaimofthisstudyisfirstlytodemonstratethe interestofthistechniqueforenhancingmixing,whilekeepingpressuredropsmoderate, andsecondlytooptimizethistypeofgeometry.Forthelatterstake,twodifferentSARs, namelySAR1(Gray’sconfiguration)andSAR2(Chen’sconfiguration),arecomparedinterms ofdistributiveanddispersivemixingandenergyexpendituresevaluatedbythepressure drop.Theenhancementeffectofsplitting/recombinationisisolatedthroughthecompari- sonwithachannelcomposedofsmallstraightsectionsatrightanglebendswithalternate chiralities(referredas3D-Flow),withoutsplitting.Aplainsquare-sectionchannelisusedas base-linereferencegeometrytoassesstherelativemixingefficiencyofeachconfiguration.
ResultsshowthattheSARtechniqueiscapableofenhancingmasstransferincreepingflows comparedtonon-splitting-flowmixers/exchangers,thusallowingagaininresidencetime andmixersizeforthesamefinalresults.BetweenthetwoSARgeometries,thesuperiority ofChen’sconfigurationregardingtherelativemixingefficiencyisdemonstrated,with83%
mixingintensification,accompaniedbyacostincreaseof68%inthefrictioncoefficient.
©2018InstitutionofChemicalEngineers.PublishedbyElsevierB.V.Allrightsreserved.
1. Introduction
Laminarmixingisimportantformanyengineeringapplicationswhere turbulencecannotbeusedorgenerated.Thisisacommonissuein pharmaceutical,cosmetic,andbiologicalapplicationswherehighly viscousandfragilefluidsare frequentlyused. Inthiscase,mixing
∗ Correspondingauthorat:UniversitéParisDiderot,SorbonneParisCitéUniversity—EnergyPhysicsGroup,AstroParticlesandCosmology- CNRSUMR7164,75013Paris,France.
E-mailaddress:[email protected](H.Peerhossaini).
shouldbeenhancedatverylowReynoldsnumberswithoututilizing vortexgeneratorsorturbulencepromoters. Compactnessofmixers inlaminarregimeisalsoofgreatimportance(Neerincxetal.,2011) sincethemixingtime,i.e.thetimerequiredforachievingacertain degreeofhomogeneity,isusuallylargerthanthatinturbulentflow regimes.
https://doi.org/10.1016/j.cherd.2018.12.010
0263-8762/©2018InstitutionofChemicalEngineers.PublishedbyElsevierB.V.Allrightsreserved.
Table1–Characteristicsoftheflowconfigurations.
Plainchannel 3D-Flow SAR1(Grayetal.,1999) SAR2(ChenandMeiners,2004)
Crosssection(mm2) 3×3 3×3 3×3 3×3
Numberofelements – 18 13 10
Elementdevelopedlength(mm) – 39 54 69
Totaldevelopedlength(mm) 700 702 702 690
Pre-conditionerlength(mm) – 30 30 30
Post-conditionerlength(mm) – 30 30 30
Thischallengehasledtothedevelopmentandoptimizationofthe so-called“SplitAndRecombine”(SAR)staticmixers(Grayetal.,1999;
ChenandMeiners,2004;Jarrahietal.,2016;HirataandOhkawa,2016).
SARmixerconsistsofanetworkofdividedandthenrecombinedchan- nelsinwhichseveralfluidsareintroducedseparatelyandmixedby amulti-laminationprocess.ThisSARtopologyperformsaseriesof baker’stransformsontheconcentrationprofile(Carriere,2007).The flowstream-tubesare split out,rotatedin oppositedirections and thenrecombined,foldingovertheconcentrationprofileanddoubling thetransversgradient.Successiveverticalseparationandhorizontal reunitingoffluidstreamsincreasethenumberoflaminateswitheach stage;hencethecontact-surfaceareabetweenthetwofluidsisexpo- nentiallyincreased,resultinginfastermixing(Ghanemetal.,2014;
HossainandKim,2015).Recentinvestigations(Carriere,2007;Ghanem etal.,2013a;Creysselsetal.,2015)revealedthechaoticnatureofthe SARflowandunderfavorableconditions,themaximumvalueofLya- punovexponentofln (2) wasrecordedintheSARstaticmixerdesigned byGrayetal.(1999).
SARmixerscanbeclassifiedintwotypes:theplanaror2Dmixers, andthe3Dconfigurations.AnsariandKim(2010)investigatednumer- icallythe mixingprocessin unbalancedsplitsandcross-collisions planarcircularandrhombicmicro-channels forReynoldsnumbers rangingfrom1to80.Theyshowedthatthelowestmixingperformance isobservedinthebalancedcollisionconfigurationswhereasthehigh- estmixingindexwasobtainedinthecircularmicro-channelsinwhich Deanroll-cellsaregeneratedatReynoldsnumbershigherthan10.Chen etal.(2011)studiedthemixingprocessinstaggeredplanarDeanvor- texmicromixersforReynoldsnumbersrangingfrom0.5to50.Itwas foundthatforReynoldsnumberslowerthan5,diffusionisdominant overadvectivetransport,whileforReynoldsnumbershigherthan10 mixingisgovernedbyDeanroll-cells.
Anxionnaz-Minvielle et al. (2017) studied experimentally heat transferinthreedifferentconfigurationsof3DSARstaticmixersand comparedtheirefficiencytoplanarcorrugatedchannels.Itwasfound thattheenergyefficiencyofSARmixersincreasescomparedtothat of2Dcorrugatedchannelswithincreasingviscosities,especiallyfor Reynoldsnumberslowerthan50.Ghanemetal.(2013a,2013b)used bothnumericalandexperimentaltechniquestostudythemassand heattransferindifferentmilli-3DSARflowconfigurationsfordiffer- entflowregimes.Comparisonoftherelativemixingefficiencyandthe energyconsumptionamongthedifferentflowconfigurationsshowed thatthemixerproposedbyJarrahietal.(2016)ismoreefficientinthe laminarandmoderatelyturbulentflowregimes,forReynoldsnumbers rangingfrom40to5000Grayetal.(1999).Fromtheheattransferpoint ofview,theSARconfigurationproposedbyChenandMeiners(2004) showsanadvantageovertheotherconfigurationsforcreepingand deeplylaminarflows,withReynoldsnumbersbetween10−4 and10 (Ghanemetal.,2013a,2013b).
Inthepresentpaper,numericalsimulationsareconductedtoana- lyzepassivescalarmixingindifferent3DSARflowconfigurationsfor averylowReynoldsnumbervalueof10−3,basedontheinletmean flowvelocity.ThislowReynoldsnumberallowstohighlightthebaker’s transformsoccurringinSARconfigurationsandtoexhibititseffectson mixingversustheincreaseinthepressuredrop.Thisanalysisallows establishingahierarchyofmixersbasedandtheirmixingefficiency andopensintoanoptimizationprocess.
Thepaperisorganizedasfollows:Section2isdevotedtoproblem descriptionandnumericalmodelused,theresultsarediscussedin Section3,andfinallytheconcludingremarksaregiveninSection4.
2. Problem description
2.1. GoverningequationsFor an incompressible,steady, Newtonianfluid, the flowis governedbythecontinuity(Eq.(1))andNavier-Stokes(Eq.(2)) equations:
∇·u=0 (1)
u·∇u=−∇p
+∇2u (2)
whereuisthe velocityvector,and arerespectively the densityandkinematicviscosityoftheworkingfluid,andpis thepressure.
Themixingprocessisanalyzedbytrackingapassivescalar Cthatwillallowcomputingthevariancedestructiondown- streamandtocharacterizemixing.Themassbalanceshould betreatedinapureconvectivetransporttoquantifythekine- maticmixing.Actually,CFDsimulationsgenerateanumerical diffusionthatdependsonthenumericalscheme,thesolver, and mostlythemeshing.Inordertoobviatenumericaldif- fusion,wehaveintroducedaphysicaldiffusioncoefficient˛, whichindeeddampsthepureSARmixingeffect(butremains identicalinallsystemsfortheglobalmixingcomparison),and couldbeclosertotheeventualexperimentalresults.Themass transferequationishencegiveninEq.(3):
∇·(uC)=␣∇2C (3)
2.2. Numericalmethod
The computational fluid dynamics (CFD) code ANSYS Flu- ent15 (ANSYS,2018)isusedforthepresent simulations.It isbasedonanEulerianapproachtocomputeNavier–Stokes equationsthroughcell-centeredfinitevolumediscretization.
Theflowandscalarequations aresolvedsequentiallywith doubleprecision.ThirdorderMUSCLschemeisusedforspa- tialdiscretizationofthelinearmomentumandmasstransfer equations.Diffusiontermsarecentral-differenceandsecond- orderaccurate.Pressure-velocitycouplingisachievedbythe SIMPLEalgorithmproposedbyPatankar(1980).
2.3. Flowconfigurationsandoperatingconditions
Numericalsimulationsareperformedforfourdifferentflow configurationssummarizedinTable1:Plainstraightchannel, 3D-Flow,SAR1andSAR2geometries.Allconfigurationshave thesameflowcross-sectionalarea.Thenumberofunitsper mixerischosensothatallconfigurations leadtothesame residencetime(definedbythereactivevolumeovertheflow rateratio).Foreachconfiguration,pre-andpost-conditioners areaddedtoattenuatetheentranceandexiteffectsonthe
Table2–Workingfluidpropertiesandoperating conditions.
(Pas) 100
(kg/m3) 1000
Pe 106
Re 10−3
˛(m2/s) 10−10
resultsandtoobtainafullydevelopedflowattheinletofthe mixerelements.
The first flow configuration, plain channel, is a plain straightduct ofsquarecross section.The3D-Flowconfigu- rationconsistsofacontinuousductflowwithalternatebends atrightangles.ThetwoSARconfigurations are referredas SAR1 and SAR2. Basic elements ofthe 3D-Flow, SAR1 and SAR2configurationsareshowninFig.1.SAR1isthegeometry describedbyGrayetal.(1999),wedenotedthisgeometrySAR1 since there is one separation/recombination per element.
SAR2 is the configuration described by Chen and Meiners (2004),wedenoted thisgeometrySAR2sincethereare two separations/recombinationsperelement.Resultsarefurther normalizedbythoseobtainedintheplainsquareductusedas thereferencegeometry.
No-slip boundary conditions are prescribed on all wall boundaries.Auniformvelocityprofilewithaveragevelocity equal to 3.33cm/s is imposed at the inlet of the pre- conditioner. The scalar field at the inlet cross-section is dividedintotwoequalparts,leftandright,withCleft=0and Cright=1.Attheoutlet,streamwisegradientsofallthevariables are settozero. Theworking fluidpropertiesandoperating conditionsareshowninTable2.TheReynoldsnumberRe,is calculatedbasedonthehydraulicdiameteroftheductcross- sectionandtheinletmeanflowvelocity.Theviscosityofthe workingfluidischosentomatchthatofviscousfluidproducts encounteredinindustrialapplicationswhereSARdevicesare usuallyusedforcreepingflows(Thakuretal.,2003).Numeri- calsimulationsarecarriedoutinisothermalconditions,with alowmassdiffusivitycoefficient(highPécletnumber)sothat diffusioncannotdominateoveradvectivemixing.InTable2,Pe isthePécletnumberrepresentingtheratiobetweenadvective anddiffusivetransportrates.Asnoticed,theadvectivetrans- portisdominantoverdiffusionsincethediffusivitycoefficient
˛isverysmall.
Scaled residualvalues of10−8 are set as aconvergence criterionforthe solutionofmomentumand passive scalar equations.
2.4. Mesh
SinceintheSARconfigurationsmixingelementsarerepeated, aseriesofnumericalsimulationswereconductedforasingle elementinSAR1configurationwithdifferentmeshdensities inordertodeterminetheappropriatefinalmeshtobeused inthestudy.Themeshsensitivityanalysisisdoneaccording totheprocedureproposedbyCeliketal.(2008);fivedifferent meshdensitieswereconsideredasshowninTable3.Thegrid sizeratioisthegridsizeofmeshi+1dividedbythatofmesh i;itisrecommendedtobeatleastequalto1.3(Celiketal., 2008).Forallcasesthemeshisstructuredincubicalcellsto reducenumericaldiffusion.Thecriteriaforthemeshsensitiv- ityarethepressuredrop,i.e.thepressuredifferencebetween theinletandoutletoftheSARgeometry,andthenormalized
standarddeviationofthecomputedscalarvaluewhichisused toquantifymixinginthedifferentgeometriesinthispaper:
0=
C2−C¯2
C20−C0 2
(4)
where ¯C0correspondstotheaveragescalarconcentrationata givencrosssectioncomputedinplainductflow.
Table3representsthemeshsensitivityanalysisperformed ononeelementintheSAR1configuration.Asshowninthis table, thevalue ofGCI on/0 is0.33% and it is0.48%for pforthe finermesh(meshnumber 5).TheGCIissimilar todiscretizationerrorandcouldbeusedaserrorbarwhen presentingtheresults.Finally,thefinermesh5isadoptedfor allgeometries.
3. Results and discussion
3.1. FlowstructureTheflowstreamlinesthroughthelastelementinthethree different geometries are shown in Fig. 2. In the 3D-Flow configuration (Fig. 2(a)), the streamlinesare simplytwisted periodicallywhilecirculatinginthesuccessivebends.Since the flow inertial forces are small relative to the viscous forces,thereisnogenerationofhydrodynamicinstabilityor vortices and thus the only irreversible mixing mechanism presentisthemoleculardiffusion.IntheSAR2configuration (Fig.2(c)),thestreamlinesareevenlysplitintotwoperpendic- ulardirections,thenrotatedby90◦twotimesbeforetheyare recombined.Thisprocessisrepeatedtwotimesineachele- ment.Herealsotherearenovorticesorinstabilities dueto highfluidviscosity.However,itwillbeshownlaterthatthe mixingoccursaccordingtobaker’stransformationinthesuc- cessiveelements.Theflowstructureinthethirdconfiguration SAR1(Fig.2(b))issomehowmorecomplexthanthefirsttwo;
streamlinesareinitiallysplitintwooppositedirectionsthen rotatedby90◦fourtimesinoppositedirectionsbeforetheyare recombinedtogether.Vorticesarealsoabsentinthisconfig- urationduetothelowReynoldsnumber.Inthenextsection, weexamineindetailshowthis successionofthesplitand recombineprocessesaffectspassivescalarmixing.
3.2. EquivalentPoincarésectionsanddistributive mixing
Chaoticpropertiesoftheflowcanbecharacterizedbymeans of Poincaré sections. For this purpose, the outlet section- plane isconsidered for each ofthe four configurations.At thecenteroftheinlet-plane,around1000inertparticlesare releasedstartingfromaninitialpositionintheformofadisk of0.15mmdiameter.Thetrajectoriesoftheindividualparti- clesaretrackedandtheircorrespondingpositionsattheoutlet sections are exported.Theset ofpositions ofthe particles inthecomputationaldomainoutletconstitutethePoincaré sections, which are shown in Fig. 3. For the straight plain duct, the particles leavethe outletatthesame positionin whichtheywerereleasedattheinlet-sectionsincetheflow islaminarandthestreamlinesarestraight.Therefore,dueto highfluidviscosityanddeeplaminarflow,fluidparticlesare organizedinuniformparallelstratathatmovealongthelon- gitudinaldirectionwithnoradialdeviationinthetrajectories.
Undertheseconditions,the3D-Flowconfigurationalsoshows
Fig.1–Isometricviewofoneelementfromeachflowconfiguration:(a)3D-Flowconfiguration,(b)SAR1(Grayconfiguration (Grayetal.,1999)),and(c)SAR2(Chenconfiguration(ChenandMeiners,2004)).
Table3–MeshsensitivityanalysisperformedononeelementintheSAR1geometry.
Meshnumber 1 2 3 4 5
Cellnumber 1.07×104 7.09×104 5.67×105 1.15×106 2.25×106
Gridsize(mm) 0.363 0.193 0.097 0.076 0.061
Gridsizeratio 1.9 2.0 1.3 1.3
/0 0.7683 0.8373 0.9227 0.9344 0.9351
GCI(%) 13.17 1.57 0.33
p (bar) 1.632 1.749 1.806 1.969 1.971
GCIp(%) 4.49 10.35 0.48
Fig.2–Streamlinesthroughthelastelementof(a)3D-Flowconfiguration,(b)SAR1and(c)SAR2.
apoordispersingperformanceclosetothestraightduct.The SARconfigurationshoweverexhibit,asexpected,aneffective signofchaoticadvection.Theparticlesarelaterallydispersed andarecapableofvisitingvariousradialpositionsintheflow cross-section,revealingthemixingenhancementbytheSAR mechanism.Atafirstglance,theSAR2configurationshows thebestperformance:particlescoverthelargestareaofthe flowcross-sectionandextendtowardsthewalls,comparedto SAR1geometryinwhichwidergapscanbeseen.
For a better quantitativecomparison ofthe distributive mixingeffects(SomanandMadhuranthakam,2017;Parketal., 2018),thecoefficientofvariationoftheparticle locationsis computedusingtheradiusasthecharacteristiclength:
CoVr=r
¯
r (5)
where ¯ristheaverageradiusandrthestandarddeviation obtainedasfollows:
r=
N
i=1
(ri−r)¯2
N−1 (6)
Fig.4comparesthecoefficientofvariationofthedistribu- tivemixingamongallthe geometriesstudiedinthiswork.
Theplainpiperepresentsthelowestdistributivemixingper- formance with a CoVr almost equal to zero. The 3D-Flow configuration comessecondwithaverylowCoVr,whichis consistentwiththePoincarésectionshowninFig.3(b).The SAR2 geometry shows the highest performancewith CoVr around 0.347comparedto0.299 intheSAR1 configuration, showingaround17%relativeincreasecomparedtothatofthe
Fig.3–Poincarésectionsforthestudiedflowconfigurations.Theinitialpositionisadisk(inblue)ofradius0.15mmatthe inletcenter:(a)Plainchannel,(b)3D-Flow,(c)SAR1,and(d)SAR2configurations.(Forinterpretationofthereferencestocolor inthisfigurelegend,thereaderisreferredtothewebversionofthisarticle.)
Fig.4–Distributivemixinganalysisusingthecoefficientof variationoftheparticlelocationascharacteristicdistance.
SAR1.Inthenextsectionthemixingperformanceisstudied bytheanalysisofpassivescalarmixinginthestudiedgeome- tries.
3.3. Scalarmixing
Pure kinematic mixing produced by the baker’s transform leads tothe formationof parallelfluid layerswithvarying propertiesfollowingtheinitial(inlet-section)stateofsegre- gation(Carriere,2007).However,usingfinitevolumemethod resultsinnumerical diffusionduetotheinterpolation and discretizationschemesadopted(HossainandKim,2015).To further reduce the numerical diffusion, in addition to the meshsensitivityanalysis,weusehigherorderdiscretization schemes.Thereforea3rdorderMUSCLschemewasadopted fortheconvectivetermsintheNavier–Stokesequations.
Scalar mixingis studiedbydividingthe inlet-sectionof each configurationintotwopartsasshowninFig.5where, theleftpartcorrespondstoanarbitraryscalarvalueequalto 0andtherightpartcorrespondstoascalarvalueattheexit- planeofthe firstfiveelementsofthe3D-Flow andthetwo
Fig.5–Scalarmixingattheoutlet-sectionofeachofthefirstfiveelementsfor(a)the3D-Flow,(b)theGray(SAR1),and(c) theChen(SAR2)configurations.
SARconfigurations.Inspiteofthelowfluiddiffusivity,with thelowvelocityand relativelylarge contacttime intervals, smalldiffusionoccursattheinterfacebetweendifferentscalar layersproducingratherintertwinedlamellarstructures.Yet thesignatureoftheSARmechanismcanbeclearlydetected andthearrangementofscalarstratafollowingthetheoret-
ical predictions of the baker’s transform is visible, though accompaniedbythedistortioncausedbymoleculardiffusion asshowninFig.5(b)and(c).FromthecontoursinFig.5(a),it canbenoticedthatforthisflowregimethebendsanddirec- tionchangesinthe3D-Flowconfigurationhavenoimportant effectonthemixingofthefluidstreams.Theonlyvisibleinflu-
Fig.6–Normalizedscalarstandarddeviation(a)atthe outletofeachelementand(b)versustheresidencelength forthethreeconfigurations:3D-Flow,SAR1,andSAR2.
enceisthatoftheinterfacialdiffusionspreadingsidewaysas thecontacttimebetweenthedifferentscalarlayersincreases downstreamintheflows.Bycomparingqualitativelythetwo SARconfigurations,itcanbeobservedthattheSAR2shows abettermixingperformance,wherethescalargradientsare rapidlyhomogenizedcomparedtoSAR1andaquasi-uniform distributionofscalarisobtainedattheoutletcross-sectionof thefifthelement.
Variationofthenormalizedstandarddeviation/0ofthe scalarattheexitofeachelementinthethreedifferentconfigu- rationsisshowninFig.6,plottedagainsttheelementnumber inFig.6(a)andtheresidencelengthinFig.6(b),sincetheele- mentsofthedifferentconfigurationsdonotsharethesame developedlength. The 3D-Flow configuration shows quasi- constantlevelsofsegregationovertheentirelength,witha relativestandarddeviationequalto1,meaningafully seg- regated state. Conversely, the segregation index decreases continuously in the SAR configurations, as the fluid flows through the different elements. As observed in the scalar distribution,SAR2geometryclearlyexhibitsthebestmixing quality,withafasterdestructionratethanthatoftheSAR1 configuration.Startingfromthe3rdelement,themagnitude ofthecurve’sslopeincreases;thegradientsarerapidlydissi-
Fig.7–(a)Normalizedpressuredropand(b)normalized Fanningfrictioncoefficientforthethreeconfigurations:
3D-Flow,SAR1,andSAR2.
patedbyvirtueoftheexponentialincreaseintheinterfacial area.Theexponentialfittingafterthe3rdelementshowsthat thedecayin/0fortheSAR2isalmost2.4timesgreaterthan thatinSAR1.Attheinlet-sectionofthe5thelementinSAR2 geometry,thedeviationfromthemeanbecomesinferiorto 1%,avaluelargelyacceptableforawiderangeofprocessengi- neeringapplications,whilethedeviationisequalto27%inthe SAR1configurationatthesamelocation.
3.4. Pressurelossesandfrictionfactors
Fig.7(a)representsthenormalizedpressuredropbetweenthe inletandtheoutletofeachmixingelementinthethreecon- figurations,wherep0correspondstothepressuredropinthe plainductflow.Theseeminglysurprisingresultfromthisfig- ureisthatthepressuredropinthe3D-Flowconfigurationis largerthanthatintheSARconfigurations andthattherel- ative pressure dropin thelattercaseislower than 1.This implies thatthepressure lossesacrosstheSARmixingele-
mentsaresmallerthanthoseinthestraightplainductwith thesameinletmeanflowvelocity.However,thereasonbehind this behavioristhat inthe SARconfigurations,the flow is dividedintotwochannelsofthesamecrosssection,sothat theflowrateisdividedby2ineachchannel,inducinglower pressuredropasitcanbeobservedfromthestreamlinescol- oredbythemeanflowvelocityinFig.2.Therefore,toaccount forthisissue,itispreferabletocomparethefrictionfactors thatcanscaleouttheflowrateissue.
TheFanningfrictioncoefficientisobtainedfromthepres- suredropacrossthemixerinletandoutletviathefollowing relation:
f= pDh
2Lu2m (7)
whereumisthemeanflowvelocity.
Inthestraightplainpipeand3D-Flowconfigurations,umis simplythemeanflowvelocity,whichisuniformthroughout theentiremixerlength.However,intheSARconfigurations, thecalculationofthefrictioncoefficientshouldbeweighted bythecorrespondingflowvelocity.InSAR1configuration,10%
oftheflowvolumecorrespondstoameanflowvelocityum
while90%correspondstothehalfofthemeanflowvelocity um/2.InSAR2,thevolumecorrespondingtoumis13%while thatcorrespondingtoum/2is87%.Letusdenoteˇthefraction ofthevolumecorrespondingtoum/2;thefrictioncoefficient wouldbecalculatedusingthefollowingexpression:
fSAR= pDh
2Lu2m(1−ˇ)+ pDh
2L
u2m
2ˇ (8)
whichleadstothefollowingequation:
fSAR= pDh
2Lu2m(1+3ˇ) (9)
ThenormalizedfrictioncoefficientsareshowninFig.7(b) wheref0correspondstothefictioncoefficientoftheplainduct flow.Thefrictionratios aregreater than1,sincethelosses instraightductarealwaysthelowest.Moreover,thefriction inthe3D-FlowconfigurationissmallerthanintheSARcon- figurationsduetosmallernumberofbends.Theincreasein thefrictionlossesinthe3D-Flowrelativetothestraightduct isabout11%,whileit islargerinthe SARcases:SAR1 has anincreaseofabout65%andSAR2hasanincreaseofabout 68%,whichisquitesimilaralthoughslightlyhigherthanSAR1 configuration.
3.5. Relativemixingefficiency
Analternativewayofassessingthemixingperformanceisby comparingtherelativemixingefficiencyofthedifferentcon- figurationsatconstantpumpingpower.Therelativemixing efficiencyisdefinedastherelativedifferenceinthestandard deviationsbetweenthemixerandthatofastraightductat constantpumpingpower:
= 0− 0 |
pp
(10)
Forequalpumpingpower:
V˙0p0=V˙ (11)
Fig.8–Relativemixingefficiencyversustheresidence lengthinthethreeconfigurations:3D-Flow,SAR1,and SAR2.
where ˙Visthevolumetricflowrate.
Substitutingpbyanexpressionintermsoffandsubsti- tuting ˙VbyanexpressionintermsofReynoldsnumberRe,it followsthat:
Re Re0 =
ff0
−13(12)
Thusatconstantpumpingpower,therelativemixingeffi- ciencybecomes:
=
0− 0
f
f0
−13(13)
Fig.8representsthevariationoftherelativemixingeffi- ciency versus the residence length for the three different configurations.TheSAR2geometryhashigherrelativemix- ingefficiencythanSAR1and3D-Flowconfigurations.Inthe 3D-Flow,theefficiencyincreasesveryslowlytoreachamax- imumvalueofabout22%atthemixerexitsincethemixing inthisconfigurationisalmostentirelyduetomoleculardiffu- sion.IntheSARconfigurations,theefficiencyincreasesrapidly toreachaplateauafterabout300mmfromtheinlet.Themax- imumefficiencyinSAR1andSAR2geometriesisalmostthe same;itisaround83%.However,mixingincreasesfasterin theSAR2comparedtoSAR1.
4. Conclusions
Mixing quality in chaotic flux recombination reactors is numerically investigated. The split/recombination mecha- nism isapractical realizationofthe mathematical baker’s transformthatgenerateschaoticstructuresandexponentially increasestheinterfacialareabetweenfluidlayers.Thispro- cess significantly promotes mixing by molecular diffusion.
Between two subsequent split and recombination phases, several rotations and flow direction changes are passively imposedontheflowcontributingtothecontinuousstretching andfoldingoftheviscousfluid.
ThestudiedcompactSARconfigurations(SAR1andSAR2), first proposed byGray et al. (1999) and Chen and Meiners (2004)respectively,arereproducedonamilli-scaleexploitable
in the industrial applications of handling viscous fluids and deeply laminar flows. A three-dimensional flow con- figuration with no split/recombination (3D-Flow) is also studied in terms of relative mixing efficiency and pres- sure drop, the baseline geometry being a plain square channel.
Thegoverningmomentum andpassive scalarconserva- tion equationsare solved usingaEulerian approachand a finitevolumemethodwiththeSIMPLEalgorithmforpressure- velocitycoupling.
Astheflowisdeeplylaminarandthefluidishighlyviscous, the3D-Flowconfigurationshowslittlemixingenhancement withrespecttoplainpipeflowandgeneratesahigherpres- suredrop.Thefluxrecombinationmixershowever,showan interestingrapiddecreaseinscalarconcentrationgradients.In SAR2configuration,forexample,thenormalizedscalarcon- centrationstandarddeviationfromthemeanattainsvalues aslowas1%asearlyasattheinletofthefifthelement.At thesamelocation,thoseforSAR1geometryarecalculatedat 27%.Indeed,SAR2mixershowssuperiormixingperformance comparedtotheotherconfigurations.ThePoincarésections, producedbyreleasinginertparticlesattheinletandfollowing theirpathsthroughthemixeroutlet,showthecapacityofSAR configurationstogeneratechaotic-typebehaviorinlaminar flow.Particlesremaingroupedinplainpipeand3D-flowwhile theyget dispersedand visit severallocations in thecross- sectionoftheSARconfigurations,especiallyinSAR2mixer, confirmingtheconclusionsbasedonscalarstandarddeviation values.
Toproperlyevaluatetheinterestofthesedevicesinprocess industry,energyexpenditures behind the mixingenhance- mentareevaluatedintermsofpressuredrop.Forsimilarinlet Reynoldsnumber,SARmixersgeneratelowerpressuredrops than3D-Flowandtheplainchannelduetothefactthatthe fluidflowsinthe majorityoftheir volumewithhalfofthe inletvelocitydueto splitting,thus decreasinghead losses.
Toaccountforthis point,the normalizednon-dimensional longitudinal Fanning friction coefficient is calculated and compared:3D-Flowshowsvalues11%increasecomparedto theplainchannel;theincreaseinSAR1andSAR2mixersbeing 65%and68%respectively.
Mixingperformanceandenergyconsumptionaresimulta- neouslyevaluatedintherelativemixingefficiencyparameter which gives the relative increase in mixing quality with respecttoaplainsquarechannelatconstantpumpingpower.
Whilethemaximumefficiencyin3D-Flowremainslimitedto 22%,SAR1and SAR2configurationsshowvaluesashigh as 83%withtheSAR2mixerfinallymanifestingthefastestmix- ingrates,dissipating concentrationgradientswithminimal residencetimeandmoderatepressuredrops.
Inthiscontext,thenumericalsimulationsofferapower- fuldesignandoptimizationtoolgivinginsighttotheoptimal numberofelementsasacompromisebetweenproductquality andenergyexpenditures.
Acknowledgments
C. Habchi and T. Lemenand would like to thank the PHC CEDREprogramandtheResearchCommissionofAngersUni- versityforthegrantsallowingresearcherexchangebetween FranceandLebanon.Theauthorsdedicatethisworktotheir dearcolleaguePhilippeCarrière,seniorscientistintheCNRS,
deceasedinFebruary2015.Amongmanyothercontributions, hehasintroducedustotheartofthebaker’stransform,inthe frameworkoftheEnergyProgram2006ACPR1-2-22.
References
Ansari,M.A.,Kim,K.-Y.,2010.Mixingperformanceofunbalanced splitandrecombinemicomixerswithcircularandrhombic sub-channels.Chem.Eng.J.162,760–767.
ANSYS,Fluent,AcademicResearch,version2018.
Anxionnaz-Minvielle,Z.,Tochon,P.,Couturier,R.,Magallon,C., Théron,F.,Cabassud,M.,Gourdon,C.,2017.Implementation of‘chaotic’advectionforviscousfluidsinheat
exchanger/reactors.Chem.Eng.Process.:ProcessIntensif.
113,118–127.
Carriere,P.,2007.Onathree-dimensionalimplementationofthe baker’stransformation.Phys.Fluids19,118110–118114.
Celik,I.B.,Ghia,U.,Roache,P.J.,Freitas,C.J.,Coleman,H.,Raad, P.E.,2008.Procedureforestimationandreportingof
uncertaintyduetodiscretizationinCFDapplications.J.Fluids Eng.130,078001.
Chen,H.,Meiners,J.-C.,2004.Topologicmixingonamicrofluidic chip.Appl.Phys.Lett.84,2193–2195.
Chen,J.J.,Chen,C.H.,Shie,S.R.,2011.Optimaldesignsof staggereddeanvortexmicromixers.Int.J.Mol.Sci.12, 3500–3524.
Creyssels,M.,Prigent,S.,Zhou,Y.,Jianjin,X.,Nicot,C.,Carrière, P.,2015.LaminarheattransferintheMLLMstaticmixer.Int.J.
HeatMassTransf.81,774–783.
Ghanem,A.,Lemenand,T.,DellaValle,D.,Peerhossaini,H., 2013a.Optimizedchaoticheatexchangerconfigurationsfor processindustry:anumericalstudy.In:ASME,FEDSM,Incline Village,Nevada,USA.
Ghanem,A.,Lemenand,T.,DellaValle,D.,Peerhossaini,H., 2013b.Transportphenomenainpassivelymanipulated chaoticflows:split-and-recombinereactors.In:ASME,FEDSM, InclineVillage,Nevada,USA.
Ghanem,A.,Lemenand,T.,DellaValle,D.,Peerhossaini,H.,2014.
Staticmixers:mechanisms,applications,andcharacterization methods—areview.Chem.Eng.Res.Des.92,205–228.
Gray,B.L.,Jaeggi,D.,Mourlas,N.J.,Drieenhuizen,B.P.v.,Williams, K.R.,Maluf,N.I.,Kovacs,G.T.A.,1999.Novelinterconnection technologiesforintegratedmicrofluidicsystems.Sens.
Actuators77,57–65.
Hirata,Y.,Ohkawa,K.,2016.Developmentofchannelmixers utilising180◦fluidrotationcombinedwithsplitand recombination.Chem.Eng.Res.Des.108,118–125.
Hossain,S.,Kim,K.-Y.,2015.Mixinganalysisina three-dimensionalserpentinesplit-and-recombine micromixer.Chem.Eng.Res.Des.100,95–103.
Jarrahi,M.,Thermeau,J.P.,Peerhossaini,H.,2016.Heattransfer enhancementinsplitandrecombineflowconfigurations:a numericalandexperimentalstudy.In:ASME,FEDSM, Washington,DC,USA.
Neerincx,P.E.,Denteneer,R.P.J.,Peelen,S.,Meijer,H.E.H.,2011.
Compactmixingusingmultiplesplitting,stretching,and recombiningflows.Macromol.Mater.Eng.296,349–361.
Park,C.,Lee,J.,Cho,H.,Kim,Y.,Cho,S.,Moon,I.,2018.Strategies forevaluatingdistributivemixingofmultimodalLagrangian particleswithnovelbimodalbincountvariance.Powder Technol.325,687–697.
Patankar,S.,1980.NumericalHeatTransferandFluidFlow.
HemispherePublishingCo.,NewYork.
Soman,S.S.,Madhuranthakam,C.M.R.,2017.Effectsofinternal geometrymodificationsonthedispersiveanddistributive mixinginstaticmixers.Chem.Eng.Process.:ProcessIntensif.
122,31–43.
Thakur,R.K.,Vial,C.,Nigam,K.D.P.,Nauman,E.B.,Djelveh,G., 2003.Staticmixersintheprocessindustries—areview.Chem.
Eng.Res.Des.81,787–826.