UNIVERSITATPOLITCNICADEVALNCIA

INSTITUTODEINGENIERAENERGTICA
DEPARTAMENTODEINGENIERAMECNICAYMATERIALES
DoctoralprogramIngenierayProduccinIndustrial

NumericalModelforMicrochannel
CondensersandGasCoolerswith
anImprovedAirSideApproach

PhDcandidate: SantiagoMartnezBallester
PhDdirectors: Dr.JosMiguelCorbernSalvador
Dr.JosGonzlvezMaci

Valencia,October2012


ACKNOWLEDGEMENTS

Iwouldliketostarttheacknowledgementsingeneralterms,givingspecialthanksto
allthepeopleorinstitutionswhotrytodotheirbestforthisworld,tryingtoimprove
or just maintaining our society. And especially to those people who appreciate the
knowledge and try to get answers and solutions to our societys questions and
problems. These people, organizations and ideas have been my main motivation to
carryoutthisresearchwork.
Of course, I needed something else than ideas or motivations, e.g. friendship, fun,
love,advice,timethereforeIwantsincerelytoacknowledgetoallpeoplewhohave
contributedtothisthesis,indifferentways,withtheirsupport,attentionandconcern:
Iamdeeplyindebtedtomysupervisor,Prof.JosGonzlvezMaci,forgivingme
the opportunity and trust to conduct research works at Institute for Energy
Engineering(IIE).Togetherwithhim,alsospecialthankstoProf.JosM.Corbern
Salvadorbecauseofhisguidanceandadvisetodoagoodjob.Bothofthemwere
directorsofmythesisandIamreallygratefulfortheirvaluableteachingandadvice
onnumericalmodeling.
MostofmyresearchwasdoneatIIE,whereweenjoyedamarvelousatmosphere
thankstothefriendlyrelationshipwehavethere.Thus,itisfairtocongratulateto
the people in charge of this atmosphere: Rafa, Israel, Jess, Carla, Jorge, Emilio,
Pepe,Gerardo,Alicia,andmanyothersSpecialthankstoRafaandDavidbecause
they did a great job in the first stage of my thesis helping me with the C++
programming.
IhadthegreatopportunitytobehostedbyPiotrA.DomaskiatNationalInstituteof
Standards and Technology in USA for working in the thesis for 6 months. I am
deeplygratefultohimforhostingmeatNISTaswellasforgivinghisvaluabletime
andadvice.IwillrememberalwaysfondlymystayatGaithersburgthankstopeople
like:Joanne,Susanne,Nathalie,P.Domanski,D.Yashar,V.Payne,Sunil
Themostimportantsupportandspecialaffectivegratitudeareforthosepeoplewho
hadtosupportthebestandtheworstofme,before,duringandafterthisthesis,even
betterthanIcouldhaveaskedforordeserved:

Tomyparents,mybrotherandmywifePuri




Thisthesisworkhasbeensupportedbythefollowingresearchgrants:

S. Martnez Ballester received a research fellowship from the Program for Training
UniversityTeachingStaff(FPU)oftheMinistryforEducationofSpain.
PartofthecontentsofthethesiswasperformedduringSantiagosstayfor6monthsat
National Institute of Standards and Technology (NIST), Gaithersburg (USA). The stay
(January2010July2010)wassupportedbytheSantiagosFPUprogramandtheNIST.
AndithasbeendevelopedundertheframeofthefollowingR&Dprojects:
Study and Optimization of Refrigerating and Air Conditioning Systems for Working
withNaturalHydrocarbonRefrigerants,fundedbytheMinistryforEducationofSpain
(ContractRef.DPI200806707C0201).
Study about Evaporators and Condensers based on Minichannel Technology for
applicationinRefrigerating,AirConditioningandHeatPumpSteadySystems,funded
bytheMinistryforScienceandInnovationofSpain(ContractRef.DPI201126771C02
01).



SUMMARY

ThefollowingPhDthesishasbeencarriedoutintheInstituteforEnergyEngineeringof
Universitat Politcnica de Valncia and during a stay at at the National Institute of
StandardsandTechnology(NIST).Themainobjectiveofthisthesisistodevelopahigh
accuracymodelformicrochannelheatexchangers(MCHX),whichhastobeusefulfor
designingpurposesintermsofcomputationalcost.
Intheauthorsopinion,therearesomedrawbackswhenexistingmodelsareappliedto
somerecentdesignsofheatexchangersuchasserpentineandparalleltubesMCHXs.
Thus,thefirststageofthesisidentifiesthephenomenathathavethelargesteffecton
theaccuracyofaMCHXmodel.Itwasalsoevaluatedthedegreeofaccomplishmentof
some classical assumptions and approaches. To this end, the high accuracy model
Fin2Dwasdevelopedasatooltocarryoutthementionedresearch.
Fin2D model is a useful tool to analyze phenomena that takes place but requires a
largecomputationalcost;notbeingfeasiblefordesignpurposes.Therefore,basedon
theknowledgeacquiredwithFin2D,anewmodelwasdeveloped:Fin1Dx3model.This
modelonlyaccountsforthemostimportantphenomenapreservingnearlythesame
accuracy as Fin2D but with a reduction of one order of magnitude in the required
simulationtime.Itintroducesanoveldiscretizationandauniquenumericalschemefor
modelingtheairsideheattransfer.Thisnovelapproachallowsmodelingconsistently
existing phenomena with great accuracy and with much less simplifying assumptions
than current models of literature. Furthermore, it achieves a reasonable
computational cost for the objective set. The thesis includes the experimental
validationofthismodelforbothacondenserandagascooler.
WiththeaimtopresentFin1Dx3modelasasuitabledesigntoolforMCHX,ithasbeen
compared in terms of accuracy and computational cost against Fin2D model,
simplificationsoftheFin1Dx3modelandotherrepresentativemodelsfromliterature.
Finally, as application of the proposed model for design purposes a set of numerical
studies were carried out. The studies are about the influence of some design
parametersontheMCHXsperformance.Theimportanceofthesestudiesisthatthey
cannot be carried out by a model that do not take into account the phenomena
modeledbyFin1Dx3.

RESUMEN

La presente tesis se ha llevado a cabo en el Instituto de Ingeniera Energtica de la

Universitat Politcnica de Valncia y durante una estancia en el National Institute of
Standards and Technology (NIST). El objetivo principal de la tesis es desarrollar un
modelodealtaprecisinparaintercambiadoresdecalordemicrocanales(MCHX),que
tienequesertil,entrminosdecostecomputacional,paratareasdediseo.
Enlaopinindelautor,existenalgunosinconvenientescuandolosmodelosexistentes
seaplicanaalgunosdiseosrecientesdeintercambiadordecalor,talescomoMCHXs,
bien de tubos en serpentn o en paralelo. Por lo tanto, la primera etapa de la tesis
identifica los fenmenos que tienen el mayor impacto en la precisin de un modelo
para MCHX. Adicionalmente, se evalu el grado de cumplimiento de varias
simplificaciones y enfoques clsicos. Con este fin, se desarroll el modelo de alta
precisinFin2Dcomounaherramientaparallevaracabolainvestigacinmencionada.
ElmodeloFin2Desunaherramientatilparaanalizarlosfenmenosquetienenlugar,
sinembargorequiereungrancostecomputacional,yportantonoestilparatrabajos
dediseo.EsporelloqueenbasealosconocimientosadquiridosconelmodeloFin2D,
sehadesarrolladounnuevomodelo,elFin1Dx3.Estemodelotanslotieneencuenta
los fenmenos ms importantes, reteniendo casi lamisma precisin queFin2D,pero
conunareduccineneltiempodeclculodeunordendemagnitud.Seintroduceuna
novedosa discretizacin y un esquema numrico nico para el modelado de la
transferencia de calor del lado del aire. Este nuevo enfoque permite modelar los
fenmenosexistentesdeformaconsistenteconmayorprecisinyconmuchomenos
simplificacionesquelosmodelosactualesdelaliteratura.Porotraparte,selograun
coste razonable de clculo para el objetivo fijado. La tesis incluye la validacin
experimentaldeestemodelotantoparauncondensadoryunenfriadordegas.
ConelobjetivodepresentarelmodeloFin1Dx3 comounaadecuadaherramientade
diseo para MCHX, ste ha sido comparado en trminos de precisin y coste
computacionalconelmodeloFin2D,otrassimplificacionesdelmodeloFin1Dx3yotros
modelosrepresentativosdelaliteratura.
Por ltimo, como aplicacin del modelo propuesto a tareas de diseo, se llevaron a
cabo una serie de estudios numricos. Los estudios analizan la influencia de algunos
parmetrosdediseoenlaeficienciadelMCHX.Laimportanciadeestosestudioses
quenopuedenllevarseacaboporunmodeloquenotengaencuentalosfenmenos
modeladosporFin1Dx3.


RESUM

Aquesta tesi s'ha dut a terme a l'Institut d'Enginyeria Energtica de la Universitat

Politcnica de Valncia i durant una estada al National Institute of Standards and
Technology (NIST). L'objectiu principal de la tesi s desenvolupar un model d'alta
precisi per bescanviadors de calor de microcanals (MCHX), que ha de ser til, en
termesdecostcomputacional,peratasquesdedisseny.
Enl'opinidel'autor,hihaalgunsinconvenientsquanelsmodelsexistentss'apliquena
alguns dissenys recents d'bescanviador de calor, com ara MCHXs, b de tubs en
serpentoenparallel.Pertant,laprimeraetapadelatesiidentificaelsfenmensque
tenen el major impacte en la precisi d'un model per MCHX. Addicionalment, es va
avaluarelgraudecomplimentdediversessimplificacionsienfocamentsclssics.Amb
aquestafinalitat,esvadesenvoluparelmodeld'altaprecisiFin2Dcomunaeinaper
duratermelainvestigaciesmentada.
El model Fin2D s una eina til per analitzar els fenmens que tenen lloc, per
requereixungrancostcomputacional,ipertantnostilperatreballsdedisseny.s
per aix que en base als coneixements adquirits amb el model Fin2D, s'ha
desenvolupat un nou model, el Fin1Dx3. Aquest model noms t en compte els
fenmensms importants, retenint gaireb la mateixa precisi que Fin2D, per amb
una reducci en el temps de clcul d'un ordre de magnitud. S'introdueix una nova
discretitzaciiunnicesquemanumricperalamodelitzacidelatransfernciade
calor del costat de l'aire. Aquest nou enfocament permet modelar els fenmens
existentsdeformaconsistentambmsprecisiiambmoltmenyssimplificacionsque
els models actuals de la literatura. D'altra banda, s'aconsegueix un cost raonable de
clculperal'objectiufixat.Latesiincloulavalidaciexperimentald'aquestmodeltant
perauncondensadoriunrefredadordegas.
Ambl'objectiudepresentarelmodelFin1Dx3comunaadequadaeinadedissenyper
MCHX, aquest ha estat comparat en termes de precisi i cost computacional amb el
modelFin2D,altressimplificacionsdelmodelFin1Dx3ialtresmodelsrepresentatiusde
laliteratura.
Finalment, com a aplicaci del model proposat a tasques de disseny, es van dur a
terme una srie d'estudis numrics. Els estudis analitzen la influncia d'alguns
parmetres de disseny en l'eficincia del MCHX. La importncia d'aquests estudis s
que no es poden dur a terme per un model que no tingui en compte els fenmens
modelatsperFin1Dx3.

11


 TABLEOFCONTENTS

TABLE OF CONTENTS
LISTOFFIGURES...........................................................................................17
LISTOFTABLES.............................................................................................19
NOMENCLATURE..........................................................................................21

1 INTRODUCTION.........................................................................................24
1.1 MicrochannelHeatExchangers.........................................................25
1.2 LiteratureReviewandBackground...................................................27
1.3 MotivationsandResearchObjectives...............................................31
1.4 ThesisOrganization...........................................................................33
2 FIN2DMODEL............................................................................................36
2.1 Fin2DHeatExchangerModel............................................................36
2.1.1 Heatexchangerdiscretization...................................................36
2.1.2 Governingequations.................................................................38
2.1.3 NumericalSolution....................................................................39
2.2 CaseStudyDefinition........................................................................40
2.3 NumericalVerificationoftheFin2dModel.......................................42
2.4 Fin2DSolutionfortheCaseStudy.....................................................46
2.5 AnalysisoftheSegmentbySegmentNTUModelingandEffectof
ClassicalAssumptions....................................................................................51
2.5.1 ComparisonofFin2DmodelagainstNTUapproaches...........53
2.5.2 AnalysisofclassicalassumptionswithFin2Dmodel.................55
2.6 Conclusions........................................................................................59
3 FIN1Dx3MODEL........................................................................................62
3.1 Fin1Dx3HeatExchangerModel........................................................63
3.1.1 HeatExchangerDiscretization...................................................65
3.1.2 GoverningEquations.................................................................67

13

TABLEOFCONTENTS

3.1.3 NumericalScheme.....................................................................74
3.1.4 SolutionMethodology...............................................................80
3.2 ModelValidation...............................................................................83
3.2.1 Microchannelcondenservalidation..........................................84
3.2.2 Microchannelgascoolervalidation...........................................86
3.2.3 Impactofparameteronpredictedcapacity..........................88
3.3 SimplificationofFin1Dx3model:Fin1D............................................89
3.4 Conclusions........................................................................................90
4 NUMERICALCOMPARISONOFMODELS...................................................94
4.1 Comparisonamongdifferentdevelopedmodels:Fin2D,Fin1Dx3and
Fin1D ...........................................................................................................95
4.2 Comparisonwithotherauthorsapproaches....................................98
4.3 Conclusions......................................................................................102
5 SIMULATIONSTUDIES..............................................................................106
5.1 Simulationmethodologyandcasestudydescription.....................107
5.2 Numberofrefrigerantpasses..........................................................109
5.3 Influenceofthefincuts...................................................................111
5.4 Influenceofaspectratioforaserpentinegascooler......................114
5.5 Conclusions......................................................................................115
6 CONCLUSIONS.........................................................................................118
6.1 GlobalConclusions..........................................................................118
6.2 Contributionsandpublications.......................................................122
6.3 Futurework.....................................................................................124
7 APPENDICES.............................................................................................128
7.1 AppendixA:Matrixcoefficientsforbothcontinuousfinandcutfin....
.........................................................................................................128
7.1.1 [B]forcontinuousfin...............................................................129
7.1.2 [B]forcutfin............................................................................135
14

 TABLEOFCONTENTS

7.1.3 [C]forcontinuousfin...............................................................137
7.1.4 [C]forcutfin............................................................................141
7.2 AppendixB:Experimentaldatausedforvalidationofmodel........142
7.2.1 Gascooler................................................................................142
7.2.2 Condenser...............................................................................145
REFERENCES....................................................................................................148

15

TABLEOFCONTENTS

16

 LISTOFFIGURES

LIST OF FIGURES

Fig.1.1TypesofairtorefrigerantheatexchangerscommonlyusedinHVAC&R.....................24
Fig.1.2DifferentviewsofmicrochannelHXs,microchanneltubesandfinsurfacebetween
tubes.............................................................................................................................................26
Fig.2.1(a)Pieceoftheheatexchangerstudiedinthepaper.(b)Schematicofthediscretization
appliedinasegmentoftheheatexchanger.................................................................................37
Fig.2.2(a)Cellsschematicanddefinitionofthecellnodes.(b)Directionreferencesforfinand
tubewallcells................................................................................................................................37
Fig.2.3(a)Schematicoftheequivalentheatexchangerstudied(b)Schematicoftheequivalent
heatexchangerusedinthestudyoftheadiabaticfintipassumption.........................................41
Fig.2.4Validationresultsfortwoscenarios:airsidewhenthenumberofcellsintheZdirection
isvaried(V1),andrefrigerantsidewhenthenumberofsegmentsintheXdirectionischanged
(V2)................................................................................................................................................43
Fig.2.5Fintemperatureprofilevalidation(V3):Erroroftheheattransferredfromthefintothe
air,fortwocases:tubeswiththesametemperatureandwithatemperaturedifferenceof15K,
withthegrid:{1,1,1,N,1}..............................................................................................................44
Fig.2.6Fintemperatureprofilevalidation(V3):(a)thecasewiththesameinlettube
temperaturesand(b)thecasewithatemperaturedifferencebetweentubesof15K(fivegrids
considered)...................................................................................................................................45
Fig.2.7Twodimensionalconduction(V4):(a)Tubetemperatureprofilepredictedwiththe
Fin2Dmodel.(b)Error,evaluatedastemperaturedifference,oftheFin2Dmodelwithrespect
theanalyticalsolution...................................................................................................................46
Fig.2.8(a)Refrigeranttemperatureevolutionalongthetubelengthfortwovaluesoftheair
sideheattransfercoefficient.(b)RefrigeranttemperatureprofilealongtheZdirectionfortwo
valuesoftheairsideheattransfercoefficient.Eachscenariowasstudiedusingtheactual
numberofchannels(b=10)andoneequivalentchannel(b=1).....................................................48
Fig.2.9AirtemperatureprofilesalongtheYdirectionattherefrigerantinlet(X=0)forthree
locationsalongtheZdirection......................................................................................................49
Fig.2.10Finwalltemperatureprofileattherefrigerantinletsection(X=0)forthecasestudy
with:(a)=air(b)=3air.........................................................................................................50
Fig.2.11Tubewalltemperatureprofileforthecasestudywith:(a)=air(b)=3air..........51
Fig.2.12ComparisonofFin2DmodelandNTUmodelfordifferentnumberofrefrigerant
segments(Ns)intheXdirectionusedbytheNTUmodel:(a)usingRMAUrelationships,(b)
usingBUrelationships...................................................................................................................54
Fig.2.13Walltemperatureprofile(finandtubes)alongtheYdirectionattherefrigerantinlet
section(X=0)inthemiddleofthetubedepth(Z/W=0.5)forbothscenariossolvedwithFin2D
model:fincutandwithoutcut,=air.........................................................................................58
Fig.2.14AirtemperatureprofilesalongtheYdirectionattherefrigerantinlet(X=0)forthree
differentlocationsalongtheZdirectionwhenadifferencetemperatureof40Kexistsbetween
refrigerantinlets...........................................................................................................................59

17

LISTOFFIGURES

Fig.3.1Detailofalouveredfinsurfaceinamicrochannelheatexchanger,wherethenon
louveredheightandthetotalfinheightaredepicted..................................................................65
Fig.3.2ExampleofamicrochannelheatexchangerthatcanbesimulatedbyFin1Dx3.............65
Fig.3.3DiscretizationinsegmentsoftheheatexchangershowedinFig.3.2,whichincludesthe
thermalconnectionsbetweendifferentsegmentsandflowarrangement...................................66
Fig.3.4Schematicofasegmentdiscretizationintocells............................................................67
Fig.3.5Differentviewsofadiscretizedportionofheatexchanger:(a)globalviewillustrating
fluidnodesandtubedirections;(b)ZYplane,whichshowsmaingeometricdataofthefinand
regionswhereisdefinedthecorrespondingTf(Y)and T a ;(c)XYplane,whichshowsthe
locationoftheTfTandTfBtemperatures.......................................................................................68
Fig.3.6Solutionmethodologyforsolvingtheproblem..............................................................81
Fig.3.7Modelvalidationfortwocondensersbymeansofcomparisonbetweenexperimental
andpredictedcapacity..................................................................................................................85
Fig.3.8Modelvalidationforagascoolerbymeansofcomparisonbetweenexperimentaldata
against(a)predictedcapacity;(b)predictedrefrigerantoutlettemperature...............................87
Fig.3.9Influenceoffinheightratioonheatdutyofgascoolerfordifferentfinsdensities....88
Fig.3.10DifferentviewsofportionofheatexchangerdiscretizedbyFin1Dmodelwhichuses
onlyonecellalongYdirectionforfinandair................................................................................90
Fig.4.1Heattransferdeviation,fordifferenttestconditions,ofFin1Dx3andFin1Dmodelswith
regardtoFin2Dmodel...................................................................................................................96
Fig.4.2Comparisonofthesimulationtimeemployedbyeachmodel.......................................97
Fig.4.3AnalogybetweenafinnedtubeandaMCHXfortheheatconductionresistance
evaluationbetweentwoneighborstubesalongthefin................................................................99
Fig.4.4DeviationofpredictedheattransferofmodelsFin1D_CutandCorrectedFinwith
regardtoFin1Dfordifferenttestconditions..............................................................................101
Fig.4.5DeviationofpredictedheattransferofmodelsFin1D,Fin1D_CutandCorrectedFin
withregardtoFin1Dx3fordifferenttestconditions...................................................................102
Fig.5.1Schematicsoftwogascoolerarrangementsstudied:3and12refrigerantpasses......109
Fig.5.2Heattransferwhennumberofrefrigerantpassesischangedintwoscenarios:air
velocityof3m/sand1m/s.........................................................................................................110
Fig.5.3Refrigerantpressuredropalongheatexchangerwhennumberofrefrigerantpassesis
changed.......................................................................................................................................110
Fig.5.4Schematicofthefincutarrangementstudied..............................................................111
Fig.5.5Improvementofheattransferbycuttingfinswithrespecttothesameconditionsbut
withcontinuousfinfordifferentnumberofrefrigerantpassesandfortwoscenarios:airvelocity
of1m/sand3m/s......................................................................................................................112
Fig.5.6Heattransferofthegascoolerwhentheaspectratioisvariedfortwoscenarios:
continuousfinandfinwithcuts..................................................................................................115

18

 LISTOFTABLES

LIST OF TABLES

Table1.1FinandtubeHXsclassification(KandlikarandGrande,2002)....................................25
Table2.1Geometryofthemicrochannelheatexchanger..........................................................41
Table2.2Operatingconditions;Testforgascoolingn3b,HX1(Zhaoetal.,2001)...................41
Table2.3Effectof2DLHConcapacity........................................................................................55
Table2.4Effectofassumingadiabaticfintipefficiencyoncapacity..........................................57
Table3.1Correlationsusedinthemodelforcoefficientsevaluation.........................................83
Table3.2.Geometriccharacteristicsofthecondensersusedforthemodelvalidation(Garca
Cascalesetal.,2010).....................................................................................................................84
Table3.3.Geometriccharacteristicsofgascooler(Yinetal.,2001).............................................86
Table5.1.Geometriccharacteristicsofgascoolerforsimulationstudies..................................108
Table5.2.Operatingconditionsforsimulationstudies:basedontestn2(Yinetal.,2001).....108
Table5.3.Geometricvariablesintheaspectratiostudyforaserpentinegascooler................114

19

LISTOFTABLES

20

 NOMENCLATURE

NOMENCLATURE

A heattransferarea(m2)
Ac crosssectionalarea(m2)
a,b,c,d,e griddimensions
Cp specificheat(Jkg1K1)
D hydraulicdiameter(m)
f DarcyWeisbachfrictionfactor()
G massvelocity (kgm2 s1)
g accelerationofgravity(9.81ms2)
H height(m)
h specificenthalpy(Jkg1)
Hp tubepitch(m)
k thermalconductivity(Wm1 K1)
L tubelength(m)
l distancebetweentwowallcells(m)
m massflowrate(kgs1)
N numberofcells
Np numberofrefrigerantpasses
Ns numberofsegments
P pressure(Pa)
pw wettedperimeter(m)
q heatflux(Wm2)
Q heatcapacity(W)
R thermalresistance(KW1)
s lengthintheforwarddirectionofafluid(m)
T temperature(K)
t thickness(m)
U overallheattransfercoefficient(Wm2K1)
V volume(m3)
v airvelocity(ms1)
W tubedepth(m)
X,Y,Z spatialcoordinates(m)
Greeksymbols
convectiveheattransfercoefficient(Wm2K1)
tubeinclination (deg)
heatexchangereffectiveness()
finheightratio()
finefficiency()
temperaturedifference(K)
density(kgm3)
Subscript
a air,aircellindex
acc acceleration
cont contraction
21

NOMENCLATURE

exp expansion
f fin,fincellindex
fB fincellatbottom
fr friction
fT fincellattop
g gravitational
i fluidcellindex
in inlet
j matrixcolumnindex
k directionindex
N,S,W,E,JB,JT directionsofneighbourwallcell
out outlet
r refrigerant,refrigerantcellindex
t tube,tubecellindex
w wallcellindex
X,Y,Z spatialcoordinatesdirections
Abbreviations
BU bothunmixed:airandrefrigerant
HX heatexchanger
LHC longitudinalheatconduction
MCHX microchannelheatexchanger
NTU numberoftransferunits
RMAU refrigerantmixedandairunmixed
RTPF roundtubeandplatefinheatexchanger

22

CHAPTER1
INTRODUCTION

23

1.INTRODUCTION

1 INTRODUCTION
Heat exchangers (HXs) are components which are present in lots of industrial
processes, ranging from low to high thermal capacities. Depending on the working
fluids,applicationandoperatingconditionsthereareseveraltypesofheatexchangers.
The applications in which the present work focuses are heating, ventilation, air
conditioningandrefrigeration(HVAC&R)usingairassecondaryfluid.(Fig.1.1)shows
someoftheheatexchangersusedintheseapplications.

Fig.1.1TypesofairtorefrigerantheatexchangerscommonlyusedinHVAC&R.

As energy costs become very important in todays industrial, residential, and

commercial settings, the rational use of energy is now a primary design and
management objective. The airconditioning, refrigeration, and heating equipment
consumealargepartofelectricalenergyonagloballevel.Forexample,11.1%ofthe
total Spain energy consumption is in HVAC&R applications, and 98 % of this
consumption corresponds to the commercial sector (IDAE, 2005). The research and
progress have been going a long way toward improving the energy efficiency of the
systems,bymeansofinnovativesystemandcomponentdesign.

Aswithotherproducts,reliablesimulationmodelscanprovidesubstantialcostsavings
during the design and optimization process of heat exchangers, mainly due to the

24

 1.INTRODUCTION

reduction of experimental tests. The experimental tests are carried out in expensive
laboratories with large human resources requirements, which would have a large
impact on final cost. A suitable simulation tools can achieve designs that perform as
expectedonfewtries,whichwouldrequirelessnumberofexperimentaltests.

1.1 Microchannel Heat Exchangers

FinandtubeHXisthemostimportantkindofairtorefrigerantHXs.Manyparameters
canbeusedtoclassifyfinandtubeHXs,buthydraulicdiameteristhemostaccepted
criteria. There are also many classifications as function of the hydraulic diameter:
Kandlikar and Grande (2002) propose the classification shown in Table 1.1. It is
important notice that this is not the only classification, and there is not a complete
agreementtoclassifyfinandtubeHXs.

Table 1.1 Finandtube HXs classification (Kandlikar and Grande,

2002)
Conventionalchannels Dh>3mm
Minichannels 200m<Dh<3mm
Microchannels 10m<Dh<200m
TransitionalChannels: 0.1m<Dh <10m
TransitionalMicrochannels 1m<Dh<10m
TransitionalNanochannels 0.1m<Dh <1m
MolecularNanochannels Dh <0.1m

The two most common types of finandtube HXs are roundtube and platefin heat
exchangers (RTPFs) and microchannel heat exchangers (MCHXs). According to Table
1.1,RTPFscorrespondtotheconventionalchannelcaseshowninthetable.Thiskind
ofHXsarepreferredbecauseoftheirperformance,easeofmanufacturingandproven
reliability.

Regarding MCHXs, the first discussion is the name used to reference them. MCHX
commonlyusedinHVAC&Rapplicationsconsistofflatmultiporttubes,withindividual
tube hydraulic diameters between 0.5 mm and 1 mm. According to Table 1.1, these
HXswouldcorrespondtominichannelsHX.However,sincethereisnofullagreement
intheclassification,microchannelHXisthemostcommonwaytoreferencethesekind
of HX in the technical literature. Fig. 1.2 shows many MCHX samples and the cross
sectionofsomeoftubesemployedintheseheatexchangers.

25

1.INTRODUCTION

Fig.1.2DifferentviewsofmicrochannelHXs,microchanneltubesandfinsurfacebetween
tubes.

Currently,anincreasinginterestonmicrochannelheatexchangers(MCHXs)isarisingin
refrigeration and air conditioning applications due to their high compactness (high
ratiobetweenUAandvolume).Itisconsequenceoflargeheattransfercoefficientsas
result of using small hydraulic diameters. Given an air side heat transfer area, high
compactnessmeansreducedvolumeswhatwillresultinlightheatexchangersandlow
refrigerantcharges.

Getting low refrigerant charges plays an important role on the use of natural
refrigerants which are flammable like propane (Palm, 2007; Hrnjak, 2010). Natural
refrigerantsareconsideredasmoreenvironmentallyfriendlythanothersrefrigerants
commonly used, with a similar or even better performance. The main drawback of
workingwithsomenaturalrefrigerantsisthattheyaredangerousinlargequantities:
ammonia can be toxic and propane is highly flammable, in fact IEC 603351 restricts
the amount of a hydrocarbon that can be used in a system to 150 g. To this end, a
suitable heat exchanger design is a serpentine MCHX (Fernando et al., 2004; Hrnjak
andLitch,2008;Palm,2007;Hrnjak,2010).Thiskindofheatexchangersminimizesthe
refrigerant charge because it has no headers, thus saving all this volume and the
correspondingrefrigerantcharge.

26

 1.INTRODUCTION

In the case of transcritical CO2 systems, microchannels have an additional merit

relatedtotheirhighmechanicalstrength.

However, microchannels also have disadvantages, such as: high cost of manufacture
andincaseofevaporators:problemstodrainthecondensatewater(Pettersenetal.,
1998; Qi et al., 2009; Zhang and Hrnjak, 2010; Moallem et al., 2012a, 2012b) and
refrigerantflowmaldistribution(Kulkarnietal.,2004;Brixetal.,2009;Yeetal.,2009;
Brixetal.,2010;Nielsenetal.,2012).Additionally,intwophaseflowsomeoftheheat
transfer and pressure drop models and/or correlations, traditionally worked out for
RTPF heat exchangers, do not work accurately for MCHX because the fluid flow and
heattransferinmicrochannelsissubstantiallydifferentfromthoseencounteredinthe
conventionaltubesandchannels.Manyauthorshaveworkedonthistopic: Kandlikar
(2002); Thome (2004); Garimella et al. (2005); Cavallini et al. (2005); Revellin and
Thome (2007); Bertsch et al., (2008); Cavallini et al. (2009); Revellin et al. (2009);
Agarwaletal.(2010).Theseproblemsarethegoalofmuchoftheresearchinvolving
microchannelsrightnow.

Microchannels have been widely used in the automotive industry, where RTPF has
beenalmostcompletelyreplacedwithMCHXs.Thisapplicationhaswidelystudiedby:
JoardarandJacobi,(2005);Zhongetal.(2005);Xiaetal.(2006);Liaetal.(2011);Qu
et al. (2011); Zilio et al. (2011); Ayad et al. (2012). It is a suitable option for this
application because it is light and saves a lot of space, so needed in automobiles.
Lately, HVAC&R industry is starting to use MCHX in some systems (Fernando et al.
2008;Shaoetal.,2010;KewandReay,2011).Infact,MCHXsareveryusedinsystems
thatworkwithCO2(Pettersenetal.,1998;KimandBullard,2001;Kimetal.2002;Kim
et al. 2004; Veje and Sss, 2004). Other interesting applications are found in highly
specialized fields (Singh, 2009): bioengineering, microfabricated fluidic systems,
electronics (Zhao and Lu, 2002; Chen and Garimella, 2011; Ramos Alvarado et al.,
2011),andhightemperaturesolarreceivers(Lietal.,2011).

1.2 Literature Review and Background

Nowadays, simulation software is a very suitable tool for the design of products in
whichcomplexphysicalprocessesoccur.Thesetoolsallowthesavingoflotsofcosts
andtimeinthelaboratoryworkingwithexpensivetestbenches.

Currently, several models or simulation tools for heat exchanger are available in the
literature:forfinnedtubes(LeeandDomanski,1997;Corbernetal.,2002;Jiangetal.,
2006; Oliet et al., 2007b; Singh et al., 2008; CoilDesigner, 2010; EVAPCOND, 2010;

27

1.INTRODUCTION

IMSTART, 2010; Oliet et al., 2010) and microchannel heat exchangers (Kim and
Bullard,2001;Yinetal.,2001;Asinari,2004;Jiangetal.,2006;Olietetal.,2007a;Shao
etal.,2009;GarcaCascalesetal.,2010;MPower,2010;FronkandGarimella,2011).
Someofthem(KimandBullard,2001;Yinetal.,2001;Corbernetal.,2002;Asinari,
2004;Olietetal.,2007a;Singhetal.,2008;Shaoetal.,2009;CoilDesigner,2010;IMST
ART, 2010; Oliet et al., 2010) apply energy conservation equations to each control
volume,whileothers(LeeandDomanski,1997;EVAPCOND,2010;Jiangetal.,2006;
Oliet et al., 2007b; GarcaCascales et al., 2010; Fronk and Garimella, 2011) apply
directly the solution given by the NTU methodology. The main difference between
both methodologies is that the NTU model uses several implicit assumptions
resultinginlessfreedomtodescribetheactualprocesses.However,themodelsbased
onenergyconservationequationsusuallyapplythesameassumptionsmadeinNTU
approaches.Theseclassicalassumptionsarethefollowing:

Steadystate.
Uniformfluidproperties.
Use of fin efficiency: application of the fin theory, which assumes uniform
temperaturethroughouttheairincontactwiththefin.
Adiabaticfintip assumption for the fin efficiency evaluation: no heat
conductionbetweentubestroughthefin.
Onedimensional heat conduction: negligible effect of 2D longitudinal heat
conduction(2DLHC).

Steadystateisarealassumption,whichissatisfiedforconditionsusuallystudied.

The fluid properties issue is easily addressed by splitting the heat exchanger into
segments.Ontheotherhand,notusingtheNTUmethodologyhasthedisadvantage
of losing an accurate fluid temperature function, which requires assuming some
temperature profile for the fluids. This problem can be solved by dividing the heat
exchangerintosmallersegments,whichimprovestherepresentationofnonuniform
air and refrigerant properties. Inmost published models, this methodology improves
only the representation of the refrigerant properties because no discretization is
provided in the air flow direction. This leads to approximated air properties for the
heat exchanger depth (air flow path) based on the average of the inlet and outlet
temperatures.

The fin efficiency is calculated following the fin theory, which is developed assuming
uniformtemperatureforthesurroundingairincontactwiththefin,anduniformheat
transfercoefficients.Theassumptionofuniformairtemperaturealongthefinheightis
violatedsincethereisatemperaturevariationalongthefinheightandfindepth.The

28

 1.INTRODUCTION

NTUmethodologyneedstheuseoffinefficiency.Totheknowledgeoftheauthor,all
modelsavailableintheliterature,whichuseafinitevolumemethod(FVM)(Patankar,
1980),applythefinefficiency.

One of the most important effects to capture to the authors opinion is the heat
conductionbetweentubes.Allthemodelsthatapplyfintheoryassumeadiabaticfin
tip assumption in order to use the corresponding relationship for the fin efficiency
(Incropera and DeWitt, 1996). It is a quite simple expression but this efficiency,
fundamentally, does not lend itself to accounting for heat transfer via fins between
tubesofdifferenttemperatures.Someauthors(LeeandDomanski,1997andSinghet
al,2008)haveworkedinordertoimplementthisheatconductionintheirmodelsfor
finned tubes. In order to implement it, they apply similar approaches that consist in
addingaheatconductiontermtotheenergyconservationequationinthewall.Singh
et al. (2008) presented a model, referred to as a resistance model, to account for
heattransferbetweentubesthroughthefinsinfinnedtubeheatexchangersusinga
segmentbysegmentapproach.Theenergyconservationequationinthewallincludes
the heat transfer from fin to air that is modeled by using fin efficiency. This fin
efficiencyisbasedonthefintheory,whoseapplicationwasdiscussedabove,assuming
adiabaticfintipinordertoevaluatethefinefficiency.Then,theyintroducetheheat
conduction term as function of temperature gradient between neighbouring tubes
despiteusingtheconceptofadiabaticfintipefficiency.Singhetal(2008)applytothis
termamultiplierwhichhastobeadjustedeithernumericallyorexperimentallywhich
dependsontheheatexchangersimulated.Basically,thisapproachcorrectsascenario
whereadiabaticfintipassumptionisnotcorrectbyapplyingacorrectiontermtotake
intoaccountheatconductionbetweentubes.Theauthorsexplainedthattheuseofa
set of energy conservation equations is better than the use of NTU methodology
with the included heat conduction term because the NTU relationship assumes all
heatistransferredfromonefluidtoanotherwithoutinternalheattransferwithinthe
heat exchanger wall structure itself. The latter methodology was used by Lee and
Domanski, 1997. Validation effort of Singh et al (2008) showed improved model
predictions when heat conduction effects were included: predicted heat load
agreementwithin3%oftheexperimentaldatainsteadof5%correspondingtothe
model without heat conduction between tubes; the temperature distribution
prediction showed an agreement within 3.3 C of the experimental data instead of
8.5Ccorrespondingtothemodelwithoutheatconductionbetweentubes.

To the knowledge of the author there is no model for MCHXs that can evaluate the
effect of cutting fins by using a finite volume method (FVM) (Patankar, 1980) for
discretizing the equations. This capability means a model able to simulate both
scenarioswithandwithoutfincut.Asinarietal.(2004)proposedathreedimensional
29

1.INTRODUCTION

modelformicrochannelgascoolersusingCO2asrefrigerantwhichcanevaluatesuch
effects but the model employs a finitevolume and finiteelement hybrid technique.
This model discretizes the equations by means of a finitevolume and finiteelement
hybrid technique taking into account longitudinal heat conduction (LHC) along all
directionsforallelements(finsandtubes),thusitdoesnotemploytheadiabaticfin
tip assumption. They investigated the impact of longitudinal heat conduction effects
on capacity, and also studied the prediction error due to the adiabaticfintip
assumption. The authors concluded that when tube temperatures are different, the
use of the adiabaticfintip efficiency gives accurate predictions of the total heat
capacityalthoughitdoesnotaccuratelyrepresenttheactualdistributionofheatflow
betweenthefinroots.Buttheydidnotreporttheeffectsinthepredictedcapacityfor
theindividualtubesoftheheatexchanger.Ithastobenotedthataconsequenceofa
wrongpredictionoftheindividualtubecapacityintroducesawrongevaluationofthe
fluidtemperatureandpressureatthetubeoutletsection.

Cutting the fins between tubes for airtorefrigerant heat exchangers is an

improvement studied in literature. Cutting the fins avoids the heat conduction
between tubes trough the fins, which degrades the heat exchanger effectiveness. In
factseveralexperimentalstudiesindicatedthattheheatexchangerperformancecan
be significantly degraded by the tubetotube heat transfer via connecting fins.
Domanskietal.(2007)measuredasmuchas23%reductioninfinnedtubeevaporator
capacity when different exit superheats were imposed on individual refrigerant
circuits.ForafinnedtubegascoolerSinghetal.(2010)reportedheatloadgainofup
to12%andfinmaterialsavingsofupto40%.However,improvementsnosolargehave
beenachievedforMCHXs,namely:Asinarietal.(2004)whoconcludedthattheimpact
ofthisheatconductioncanbeassumedasnegligibleinawiderangeofapplications;
Park and Hrnjak (2007) reported measurements of capacity improvements of up to
3.9%bycuttingthefinsinaCO2serpentinemicrochannelgascooler.AlsoZilioetal.
(2007) concluded that heat conduction through fins in a CO2 gas cooler had a
significantimpactonthecapacity.Infact,cutfinsurfacesareincreasinglybeingused
in heat exchangers to reduce the heat conduction between tubes and improve the
heatexchangerperformance.

The onedimensional heat conduction assumption only accounts for the transverse
heatfluxtroughthewallbetweentwofluids.Itdoesnotaccountfor2Dlongitudinal
heatconductioninthetubeanditneglectsthelongitudinalheatconductioninthefin,
alongtheairflowdirection.Asinarietal.(2004)concludedthattheseeffectsproducea
negligibleeffectontheperformanceoftheclassofCO2gascoolertheystudied.

30

 1.INTRODUCTION

AgascoolerworkingwithCO2insupercriticalpressuresisanapplicationwherealarge
impactontheperformancecouldbeexpecteddueto2DLHCinthetubewallandheat
conduction between tubes through fins. The reasons are based on the temperature
glideofCO2duringasupercriticalgascoolingincontrastwithacondenserwherethe
temperature during condensation is approximately constant. Representative values
canbeextractedfromexperimentalresultsofZhaoetal.(2001)whereCO2undergoes
temperature variations along a single tube from 25 K up to 85 K while maximum
temperaturedifferencebetweentwoneighbortubesrangefrom30Kto100K.Heat
conduction between tubes trough fins appears when a temperature difference
between tubes exists. This temperature difference will vary depending on the
refrigerant circuitry and its temperature difference along the heat exchanger, which
willbelargerinagascoolerthaninanevaporatororcondenser.Thus,gascoolerisan
interestingapplicationtobestudiedinthepresentthesis,becauseitiswherelargest
impactoftheseeffectsisexpected.

Fewmodelstakeintoaccountallthephenomenaexplainedabove.Asinarietal.(2004)
presented a model for MCHX gas coolers that included these phenomena but an
importantdifferenceexistsbetweenthesesimilarmodelsandtheproposedoneinthis
workrelatedtothecomputationalcost.Themethodusuallyappliedbythesemodels
todiscretizegoverningequationsisafiniteelementmethod(FEM)(Patankar,1980)or
evenahybridoneasAsinarietal.(2004)employed,whichuseboththeFEMandthe
finitevolumemethod(FVM)(Patankar,1980).

1.3 Motivations and Research Objectives

Due to the relatively recent application of MCHXs in HVAC&R industry, a need of
developing a simulation tool for this kind of heat exchanger arises in order to aid its
design. Thus, the main objective of present thesis is the development of a high
accuracy simulation tool for design purposes of MCHX. To this end, this tool should
havefollowingfeatures:itshouldrequireareasonablecomputationalcostfordesign
purposes; it has to be robust; it must be able to be used in standard personal
computers; and it has to be able to assess the impact on performance of those
parametersthatthedesignerhastodefine.

Other motivation is based on the drawbacks that, in the authors opinion, existing
modelshavewhentheyareappliedtosomerecentdesignsofheatexchangersuchas
paralleltubesandserpentineMCHXs.Themaindrawbackofapplyingexistingmodels
to new heat exchangers designs is the implicit application of classical assumptions,
whichwerestudiedanddevelopedforaspecificheatexchangerdesignandmaynot
31

1.INTRODUCTION

be correct for these new MCHX designs. Some of these assumptions, which were
discussed in previous subsection, have been studied in the literature for many heat
exchangers topologies such as RTPF heat exchangers, whilst the effects of these
assumptionsarenotstudiedsoextensivelyforparalleltubesandserpentineMCHXs.
These heat exchangers have a different thermal behavior since the thermal and
geometric conditions are different. Thus, the author thinks that it is interesting to
evaluatetheimpactoftheclassicalassumptions,whichwereexposedpreviously,for
MCHXsworkingintypicaloperatingconditions.

Giventhesemotivations,themainobjectivesforthethesisarethefollowing:

TodevelopahighaccuracymodelabletopredicttheperformanceofaMCHX
with a low computational cost. This model has to be useful for design
purposes.
Toidentify,understandandquantifyallthephenomenathathavethelargest
effectontheheattransferofaMCHX.

Inordertoaccomplishthepreviousobjectives,additionalpartialobjectivesare:

Todevelopa model(Fin2D)which cantakeintoaccountallthe phenomena

takingplaceinMCHX,withregardtoheattransfer,e.g.2Dlongitudinalheat
conduction, heat conduction between tubes and the unmixed air along the
finheight.Ithastoallowevaluatingtheisolateimpactofeachphenomenon
withhighaccuracy.
Tochecktheconsistencyandaccuracyofresultsforthismodel.
To analyze and quantify, by using the Fin2D model, the effect of each
phenomenon on the MCHX performance in order to identify the most
importanteffectstoincludeinamodelforMCHXs.
To evaluate the error obtained by using classical approaches for heat
exchangersmodeling.
To develop a model (Fin1Dx3) that retains only the most important effects,
whichwereidentifiedpreviouslywiththeFin2Dmodel,inordertogetalow
computationalcostwithhighaccuracy.
To carry out a validation of the model with experimental data for different
scenarios.
Tocompareintermsofaccuracyandcomputationalcosttheproposedmodel
withmorecomplicatedmodels,simplifiedmodelsandrepresentativemodels
ofliterature.
To use the developed model for analyzing the influence of some design
parametersontheMCHXperformance.
32

 1.INTRODUCTION

1.4 Thesis Organization

The first chapter of the thesis reviews the state of the art related with the problem
thattheauthorwanttostudyandpresentstheobjectivesoftheresearch.

Inordertoevaluatetheeffectsexposedinprevioussubsectionamodelisgoingtobe
developed.FirstmodeldevelopedisreferredtoasFin2Dandispresentedinchapter2.
After many consistency tests in order to validate the model, Fin2D was used to
evaluateina morefundamentalwayalltheproposedeffectsandassumptions,their
impact on predicted results. It also allowed studying differences between classical
modeling approaches from literature. Results and conclusions by using Fin2D model
willbediscussedinchapter2.

Fin2Dmodelwascreatedtoobtainaccurateresultsbymodelingphenomenainavery
detailed way in order to detect inaccuracy sources and to evaluate the degree of
accomplishment of some classical assumptions. On the other hand, Fin2D model
required a large computational cost. Therefore, according to the objective of
developing a suitable model for design purposes, a new model was developed:
Fin1Dx3 model. This model only accounts for the most important phenomena
preservingnearlythesameaccuracy.Itintroducesanoveldiscretizationmethodology,
which allows be consistent with the assumptions done and it achieves a large
reduction of the simulation time required for simulations. Chapter 3 will present
Fin1Dx3model,validationstudieswithexperimentaldataforbothcondenserandgas
cooler.

Chapter 4 introduces a comparison between different model approaches, namely:

some approaches available in literature and some otherfrom Fin1Dx3 modifications.
Thiscomparisonwillshowdifferencesconcerningcomputationalcostandaccuracy.

Once the advantages of using Fin1Dx3 were shown in the comparison of models,
chapter 5 presents numerical studies which analyze influence of some design
parametersontheperformanceofMCHXsbyusingFin1Dx3model.

Finally, chapter 6 will summarize the main conclusions and contributions of the
research presented in this thesis. Chapter 6 also lists the publications resulting from
thisthesis.

33

1.INTRODUCTION

34

CHAPTER2
FIN2DMODEL

35

2.FIN2DMODEL

2 FIN2D MODEL
Thischapterpresentsadetailedmodelformicrochannelheatexchangersusedasgas
coolers which does not use the fin efficiency, accounts for 2D LHC in fins and tubes,
accountsfortheheatconductionbetweentubes,andappliesadetaileddiscretization
fortheair,whichisindependentoftherefrigerantdiscretization.Themodel,referred
toasFin2D,subdividestheheatexchangerintosegmentsandcells(air,refrigerant,fin,
tube wall), to which a system of energy conservation equations is applied without
traditional heat exchanger modeling assumptions. Fin2D differs from other models
referred to previously in the number of classical assumptions made. The model of
Asinarietal.(2004)isthemostsimilartothepresentmodelregardingthenumberof
assumptions, although the discretization method applied by them is a hybrid one,
which uses both the finite element method (FEM) and the finite volume method
(FVM),whereasthemethodologyusedintheFin2DmodelistheFVM.

Afteranumericalverification,thesolutionobtainedwiththeFin2Dmodelisemployed
to assess the impact of the classical heat exchanger modeling assumptions on the
accuracyoftheperformancepredictionsforsuchconditions.Thegoalofthepresent
chapter is to study the heat transfer processes in a microchannel gas cooler by
evaluation of each of individual heat transfer effects described above, rather than
propose a model able for heat exchangers design. This chapter will also provide a
deeperunderstandingofthemicrochannelCO2gascoolers.

2.1 Fin2D Heat Exchanger Model

2.1.1 Heat exchanger discretization

Fig. 2.1(a) presents a piece of the studied microchannel heat exchanger. It is
discretized along the X direction (refrigerant flow) in a number of segments a. Each
segment (Fig. 2.1(b)) consists of: two streams of refrigerant (top and bottom flows)
that are split into b channels in the Z direction (air flow); two flat tubes (top and
bottom)thatarediscretizedintoccellsintheZdirection;andbothairflowandfins,
whicharediscretizedintwodimensions:dcellsintheYdirectionandecellsintheZ
direction. This discretization is summarized in the text as; grid: {a,b,c,d,e}. For
illustration of the nomenclature, the numerical example shown in Fig. 2.1(a) and (b)
correspondstoagrid:{3,5,3,7,4}.

36

 2.FIN2DMODEL

Fig.2.1(a)Pieceoftheheatexchangerstudiedinthepaper.(b)Schematicofthediscretization
appliedinasegmentoftheheatexchanger.

The refrigerant flows inside the channels (b=5 in Fig. 2.1(a)) along the X direction
withoutanymixingbetweenthechannels,anditexchangesheatwiththetubecellsin
contact;thesetubecellstransferthisheattotheaircellsincontactbyconvection,and
to its neighbouring tube cells on the plane XZ and to the fin roots in contact by
conduction.Theairexchangesheatbyconvectionwiththefincells,andtheaircellsat
the bottom and top also exchange heat with the tube cells in contact. The fin cells
conduct the heat along the plane YZ, and the bottom and top fin cells also conduct
heattothetubewall.

Fig.2.2(a)Cellsschematicanddefinitionofthecellnodes.(b)Directionreferencesforfinand
tubewallcells.

37

2.FIN2DMODEL

2.1.2 Governing equations

Everyfluidcell(refrigerantorair)hastwonodes,whichcorrespondtotheinletandthe
outletinthefluidflowdirection.Thewallcells(tubeorfin)haveonlyonenodelocated
inthecentroidofthecell,asisshowninFig.2.2(a).Allcellslocalvariablesarereferred
tothevalueinthesenodes,e.g.Ti,inandTi,oarethetemperatureattheinletandatthe
outlet, respectively, of a fluid cell i, either refrigerant or air. Tw is the temperature
definedforthewallcellw,whichcouldbeeitherfinortube.

Inthissituationthegoverningequationsateachfluidcell(refrigerantandair)andat
eachwallcell(tubeandfin)canbewrittenasfollows:

ni
m i dh i q w,i pww,i dsi (2.1)
w 1

qw,i Uw,i (Tw Ti ) (2.2)

1 / Aw,i
U w, i
tw / 2 1

Aw,i k w,k Aw,i w,i

nw
k w, k t w Tw q w,i 0 (2.3)
i 1

whereanywallcellwisincontactwithnwfluidcellsi=1,nw;anyfluidcelliisincontact
with ni wall cells w=1, ni; kw,k is the thermal conductivity of the wall cell w in the k
direction,thusitispossibletostudytheinfluenceof2DLHCatbothfinandtubewalls.
Eq. (2.1) states the energy conservation for a fluid cell, whereas Eq. (2.3) states the
energyconservationforawallcell.Eq.(2.2)representstheheatflowbetweenawall
cell and a fluid cell. Neither pressure losses nor dehumidification has been modeled
since it is a gas cooler and the paper only focuses on the understanding of possible
differencesinheattransfer.

For solving the system of equations a set of boundary conditions is needed. Inlet
conditions and velocity distributions are known for both fluids, and velocity
distribution is assumed as uniform. Since the heat exchangers are normally well
insulated, the heat transferred by the wall edges to the surrounding is considered
negligible,andthewallcellsareconsideredtobeadiabaticwiththesurrounding.Only
two tubes of the whole gas cooler are going to be modeled in this work, so an
additionalboundaryconditionisnecessary:bothtubeshavesymmetrycondition.This
symmetry condition implies that the heat transferred from a tube to each of the

38

 2.FIN2DMODEL

neighboring tubes is the same. This assumption approximates simulations at central

tubesofamicrochannelslab.Ithastobenotedthatthissymmetryconditiondoesnot
mean that the heat transferred by each tube have to be the same, in fact it will be
studiedinsection2.5.

2.1.3 Numerical Solution

For the discretization of equations the finite volume method (FVM) (Patankar, 1980)
has been applied along with the semiexplicit method for wall temperature linked
equations (SEWTLE) proposed by Corbern et al. (2001). The discretization of
governingequationsdoesnotpresentanyspecialdifficulty,exceptfortheestimation
oftheintegraloftheheattransferredtothefluidsincontactwithaconsideredpiece
ofwall(Eq.(2.2)and(2.3)).Thisintegrationmustbeconsistentwiththeintegrationof
thecoincidenttermsoffluidenergyEq.(2.1).Thenumericalschemecorrespondingto
a linear fluid temperature variation (LFTV), as explained in Corbern et al. (2001), is
employedforthediscretizationofEq.(2.2).Thisnumericalschemeisbasicallybased
on assuming a piecewise distribution of the fluid temperature along the fluid cell,
leadingtothefollowingexpression:

Ti,in Ti,out
Aw,i q w,i U w,i pww,i Tw si (2.4)
2

ThediscretizationoftheLaplacianoperatorinEq.(2.3)hasbeenmadebytheclassical
finitedifferenceapproachaccordingtotheadoptedFVM.TheEq.(2.3)discretization
usedinthismodelisshowninEq.(2.5).

nj
a w Tw a w,k Tw,k a w,J Tw,J pww,i U w,i (Tw Ti )ds w,i
k W, E,S, N i0
(2.5)

k w, W Aw, W k w,E Aw,E k w,S Aw,S

a w, W a w,E a w,S
l w, W l w,E l w,S

k w, N Aw, N k w, J Aw, J
a w, N
l w, N
a w, J
l w, J
aw a w,k
k W,E,S,N,J

Allaw,ktermsrefertotheconductancebetweenawallcellwandtheneighbouringwall
cell,adjoinedtothisone,inthedirectionk.Thedirectionreferenceisdifferentinthe
tubeandfincells;theschematicusedinthemodelisshowninFig.2.2(b).Thereisone

39

2.FIN2DMODEL

exception: aw,J , which means the conductance of the joint between a tube wall cell
andafinwallcell.

TheglobalsolutionmethodisoutlinedinCorbernetal.(2001).Basically,thismethod
is based on an iterative solution procedure. First, a guess is made about the wall
temperature distribution, and then the governing equations for the fluid flows are
solved in an explicit manner, getting the outlet conditions at any fluid cell from the
values at the inlet of the heat exchanger and the assumed values of the wall
temperature field. Once the solution of the fluid properties is obtained for any fluid
cell,thenthewalltemperatureateverywallcellisestimatedfromthebalanceofthe
heat transferred across it (Eq.(2.3)). This procedure is repeated until convergence is
reached. The numerical method employed for calculating the temperature at every
wallcellisbasedonthelinebylinestrategy(Patankar,1980)followingtheYdirection
forfincellsandtheXdirectionfortubecells,sothattheglobalstrategyconsistsofan
iterative series of explicit calculation steps. This method can be applied to any flow
arrangementandgeometricalconfiguration,andoffersexcellentcomputationalspeed.
Additionally,itcaneasilybeextendedtoothercases,suchastwophasefloworhumid
air.

2.2 Case Study Definition

Inthiscasestudywemodelledamicrochannelgascoolerforwhichdimensionswere
extractedfromZhaoetal.(2001).Sincetheobjectiveofthisworkwastoevaluatethe
effectsofdifferentclassicalassumptionsinthepredictedresults,operatingconditions
that produce large temperature variations and high heat fluxes were of interest.
Consequently,thechosenoperatingconditionscorrespondtotheexperimentaldataof
the test for gas cooling n 3b, HX1, from the same work. Table 2.1 shows the most
importantgeometricdatawhileTable2.2showstheconsideredoperatingconditions.
Some data were estimated from the reported experimental values; namely, the CO2
sideheattransfercoefficientwasestimatedtobe537Wm2 K1.Thiscoefficientwas
estimatedbyusingtheNTUrelationshipforcrossflow(IncroperaandDeWitt,1996)
working in the mentioned test conditions. For these calculation the air side heat
transfercoefficientwasrequiredanditwasevaluatedwithconvectioncorrelationfor
fullylaminarflowinnoncirculartubes(IncroperaandDeWitt,1996)resultingtobe66
Wm2K1(thisvaluewillbeusedonlyintheverificationstudies).

The heat transfer mechanisms that take place along a tube in a gas cooler depend
neither on the tube length nor the number of tubes. For this reason, only an

40

 2.FIN2DMODEL

equivalentpieceoftheheatexchangerhasbeenconsideredinthedetailedanalysisof
theheattransfer.

Table2.1Geometryofthemicrochannelheatexchanger
Tubelength
8 Finpitch(mm) 1.56 Channeldiameter(mm) 1
(cm)
Tubedepth
16 Finthickness(mm) 0.152 Channelsnumber 10
(mm)
Tubethickness
1 Finheight(mm) 8
(mm)

Table2.2Operatingconditions;Testforgascoolingn3b,HX1(Zhaoetal.,2001).
Inletpressure(kPa) Inlettemperature(C) Outlettemperature G
2
(C) (kg/m s)
CO2 8937 79.9 42.4* 132.56
Air 100 23.74* 32.4 3.05
*estimatedvalue

ThereferencecasestudyisshowninFig.2.3(a).Itconsistsoftwocentraltubeswith
their fins attached. The total length of the tubes is five times the tube depth. The
refrigeranthasonlyonepassalongtheheatexchangerwiththesamemassflowrate
inbothtubes.

Fortheevaluationofthethermodynamicandtransportpropertiesoffluids,REFPROP
(Lemmon et al., 2002) was used. The air, properties were locally evaluated whereas
therefrigerantpropertieswereassumedasuniformandevaluatedasaveragedvalues
between the corresponding values at the inlet and the outlet of the heat exchanger
studied in the present work. The thermal conductivity of the fin and tube walls was
estimatedtobe173Wm1K1.

Fig.2.3(a)Schematicoftheequivalentheatexchangerstudied(b)Schematicoftheequivalent
heatexchangerusedinthestudyoftheadiabaticfintipassumption.

41

2.FIN2DMODEL

2.3 Numerical Verification of the Fin2d Model

Before employing the newly developed model to produce detailed solutions of heat
transferintheequivalentpieceofthemicrochannelgascooler,showninFig.2.3(a),it
isnecessarytovalidatethemodel.Withthispurposeinmindweperformedaseriesof
systematic checks against operational cases for which an analytical solution can be
obtained.

ThedetaileddiscretizationoftheairflowintheYdirectionadoptedinFin2Dmakesit
difficult to compare Fin2D predictions with those of analytical solutions. In order to
validate the model, many scenarios, listed below, were simulated. These scenarios
haveananalyticalsolution,andthissolutionwasadoptedasareferencetoevaluate
theerroroftheFin2Dmodel.Thefollowing,aredifferentstudiedscenarios:

Airsideverification(V1):Fortherefrigerant,theinfiniteheatcapacityrate(
m Cp ) was imposed, which means no temperature change for the
refrigerant. Also, we disabled the 2D LHC in the tube walls (LHCX and LHCZ)
and the LHCZ in the fin, since these effects are not accounted for by the
availableanalyticalsolutions.
Thedetaileddiscretizationoftheairvolume,intheFin2Dmodel,accountsfor
a nonuniform air temperature along the Y direction. Since this effect is not
taken into account by any analytical solution, the fin and air were only
discretizedintheXandZdirections,usingonlyonecellalongtheYdirection,
to make a valid comparison of the Fin2D model with analytical solution. On
theotherhand,itisnotpossibletocapturethefintemperaturevariationwith
only one fin cell along the Y direction. Thus, the value of the thermal
conductivityforthefininthisdirectionwassetasinfinite.Inthissituationthe
finefficiencywasequalto1,andthefinwalltemperaturewasuniformalong
the Y direction. Finally, constant properties and heat transfer coefficients
wereused,whichcorrespondtothoseexposedinsection3.Forthisscenario
the analytical solution for the heat exchanger effectiveness is
1 exp ( NTU) .
Refrigerantsideverification(V2):Themethodologyappliedwasthesameas
for V1, but the fluid with infinite heat capacity rate was the air. Now, the
resultswerenotassensitivetotheairdiscretizationasitwasintheV1case
becausetheairhasinfiniteheatcapacityrateanditstemperaturechangeis
negligible.
Fin temperature profile verification (V3): The fin conductivity in the Y
direction had a value corresponding to the case study. Two cases were
studied: with the same and different fin root temperatures. The analytical
42

 2.FIN2DMODEL

solutionsforbothcasesweretakenfromIncroperaandDeWitt(1996).These
relationships assume a uniform air temperature along the Y direction, and
uniform air properties and heat transfer coefficient. Thus, to avoid the air
temperature change along the Y direction, the infinite air flowstream heat
capacityrate( m Cp )wasimposed.Therefrigerantflowstreamcapacityrate
was also assumed to be infinite to obtain a uniform tube wall temperature
alongallfinroots.
ThefinwasdiscretizedonlyinonecellalongtheZdirectionsotherewasno
LHCZinthefin.
Twodimensional heat conduction in the wall (V4): This case validates the
discretization of the Laplacian term of Eq. (2.3). This case studies 2D LHC in
the tube assuming no convection and no thermal joint between the fin and
the tube. A set of temperatures for each walls edge was imposed. The
analytical solution for this situation can be obtained solving the Laplacian
equationforaflatplategiventemperaturesattheedges.

0.7
HeatExchanger Error[%]

V1,Grid:{1,10,N,1,N}
0.6
V2,Grid:{N,10,1,1,1}
0.5
0.4
0.3
0.2
0.1
0
0 5 10 15 20 25
N
Fig.2.4Validationresultsfortwoscenarios:airsidewhenthenumberofcellsintheZdirection
isvaried(V1),andrefrigerantsidewhenthenumberofsegmentsintheXdirectionischanged
(V2).

Fig. 2.4 shows the error of the numerical solution with reference to the analytical
solutionforV1andV2cases.Theerrortendstodiminishveryquicklywiththenumber
of cells used (N). In the case of V1, the abscissa shows the number of cells in the Z
direction.Asitcanbeobserved,theerrorisverysmallalreadyforN=5.Inthecaseof
V2,wheretheairhastheinfiniteflowstreamcapacityrate,theabscissawastakenas
the number of cells along the X direction. Again the analytical solution is almost
reachedwithonlyfivecells.

43

2.FIN2DMODEL

0
0.2
0.4 (Y=0)=(Y=H)
Error[%] 0.6 (Y=0)=(Y=H)+15K
0.8
1
1.2
1.4
0 5 10 15 20 25 30 35
N
Fig.2.5Fintemperatureprofilevalidation(V3):Erroroftheheattransferredfromthefintothe
air,fortwocases:tubeswiththesametemperatureandwithatemperaturedifferenceof15K,
withthegrid:{1,1,1,N,1}.

RegardingthecaseV3verification,Fig.2.5showstheerrorofthenumericalsolution
fortheheattransferredfromtheairfilmtothefinwallasafunctionofthenumberof
cellsintheYdirectionfortwosituations:equaltemperaturesofthebottomtubeand
the top tube, and atemperature difference between tubes of 15 K. is a difference
betweenthefintemperatureandtheairtemperature.Ascanbeobserved,theerroris
small,0.2%,withonlyfivecellsintheYdirection,andquicklyapproacheszero.The
calculatedfintemperatureprofileisshowninFig.2.6(a)and(b)fortheV3studyinthe
sametwopreviouscases.ThegridsdifferbythenumberoffincellsintheYdirection.
Inthismannertheaccuracyofthenumericalmodelisproved.

44

 2.FIN2DMODEL

1
0.9
0=H
0.8
0.7
0.6

Y/Hf
0.5
0.4
0.3
0.2
0.1
0
30.5 31 31.5 32 32.5
a) (Y)[K]
Theoretical 1,1,1,20,1 1,1,1,10,1

1,1,1,5,1 1,1,1,3,1 1,1,1,2,1

1
0=H+15K
0.9
0.8
0.7
0.6
Y/Hf

0.5
0.4
0.3
0.2
0.1
b) 0
26 27 28 29
(Y)[K]
Theoretical 1,1,1,20,1 1,1,1,10,1
1,1,1,5,1 1,1,1,3,1 1,1,1,2,1

Fig.2.6Fintemperatureprofilevalidation(V3):(a)thecasewiththesameinlettube
temperaturesand(b)thecasewithatemperaturedifferencebetweentubesof15K(fivegrids
considered).

Finally,inordertovalidatethe2DLHCintheFin2Dmodel,Fig.2.7(a)presentsresults
forcaseV4,wheretheboundaryconditionswere:T(X/L=0)=70C,T(X/L=1)=50C,
T(Z/W=0)=25C,T(X=0)=35C.Fig.2.7(b)showstheerrorinthewalltemperature
field evaluated as a deviation of the Fin2D results from the theoretical solution. It is
noticeablethattheerrorrangesfrom0.1Kand+0.1Koveralmosttheentireplate,
whichisconsideredasacceptable.Theerrorincreasesupto1.3Konlylocallynearthe

45

2.FIN2DMODEL

corners.Thereasonisthattheactualtemperaturefieldimposedalongthetubeedge
is discontinuous just at the corners. The Fin2D model can obtain only continuous
solutions,sonearthecornertheerrorisincreased.Thisstudywasalsocarriedoutfor
thefin,resultinginsimilarresults.

Fig.2.7Twodimensionalconduction(V4):(a)Tubetemperatureprofilepredictedwiththe
Fin2Dmodel.(b)Error,evaluatedastemperaturedifference,oftheFin2Dmodelwithrespect
theanalyticalsolution.

2.4 Fin2D Solution for the Case Study

InthissectiontheFin2Dmodelisusedtosolvethecasestudyproblem.Manythermal
variablesareanalyzedinordertounderstandtheactualheattransfermechanismsin
the case study using a very detailed model able to capture fin and tube two

46

 2.FIN2DMODEL

dimensionaltemperatureprofiles,therefrigeranttemperatureprofileineachchannel,
andtheairtemperatureprofilealongthedirectionbetweentubes.

This study used test conditions presented in Table 2.2. The airside heat transfer
coefficient was estimated by correlations for a plain fin following the
recommendations of Webb (1994). The heat transfer coefficient obtained with this
correlationisreferredtoasair.Twoscenarioswereconsidered:withtheairsideheat
transfercoefficientequaltoair,andwithavaluethreetimeslarger,consequentlythe
airsideheattransfercoefficientrangedfrom60Wm2K1to180Wm2K1.Thischoice
was made to cover large variations of possible fin surfaces including enhanced fin
surfaceswithahighheattransfercoefficient.

Regardingtherefrigerantside,constantpropertiesandheattransfercoefficientswere
used,aslistedinSection2.2.Sincethetubelengthwasshort,therefrigerantproperty
variations are expected to be negligible; the refrigerant is a gas far from the critical
pointatwhichpropertieschangedrastically.

In order to set a grid size to obtain the solution in each scenario, with the required
accuracy for the comparisons done in this work, the author studied the results
accuracy when the grid dimensions were changed. From a very detailed grid, the
differentgriddimensionswerereduceduntilafurtherrefinementofthegriddidnot
lead to a significant increase in accuracy. The adopted grid dimensions were:
{3,10,10,30,10},followingthenomenclatureexplainedinsection2.1.1.Thecapacities
obtainedforthiscasestudywere24.21Wand33.6Wforthescenarioswithair=air
andair=3air,respectively.

First,resultsfortherefrigerantareshowninFig.2.8(a)and(b).Fig.2.8(a)presentsthe
refrigerant temperature evolution along the X direction. Fig. 2.8 (b) presents the
temperature profile as a function of the dimensionless tube depth (Z direction) at
X=0.833. In this figure, a case with one equivalent channel (same hydraulic diameter
and crosssectional area) is also plotted in order to study the differences between
modeling the actual number of channels and modeling all the fluid as an equivalent
fluidcellwiththeassumptionofmixedrefrigerantalongthetube.

47

2.FIN2DMODEL

a) 85
Z=0;=air
Z=0;=3air
80

Temperature[C]
75

70

65

60
0 0.2 0.4 0.6 0.8 1
X/L
b) 70

69 b=10;=air
Temperature[C]

b=10;=3air
68 b=1;=air
b=1;=3air
67

66

65
0 0.2 0.4 0.6 0.8 1
Z/W
Fig.2.8(a)Refrigeranttemperatureevolutionalongthetubelengthfortwovaluesoftheair
sideheattransfercoefficient.(b)RefrigeranttemperatureprofilealongtheZdirectionfortwo
valuesoftheairsideheattransfercoefficient.Eachscenariowasstudiedusingtheactual
numberofchannels(b=10)andoneequivalentchannel(b=1).

InFig.2.8(b)thetemperatureprofilefor10channelsdescribesatypicaltrendwhen
2D LHC in the tube is present. It is noticeable how small the temperature variation
between different refrigerant channels is in both scenarios (at most 0.5 K). The
difference in the total capacity calculated resulted to be less than 0.005 % for both
scenarios.Thisisduetotworeasons:theuniformrefrigeranttemperaturefortheone
channelcasealmostcoincideswiththeaveragedvalueoftherefrigeranttemperature
in the multichannel case, and the uniformity of the tube temperature along the Z
direction.Thecombinationofthesetwofactsproducesanequalaverageddifference
of temperatures between the tube and the refrigerant, which produces the same

48

 2.FIN2DMODEL

capacity transferred by the fluid. Thus, for the scenario studied, the modeling of a
minichanneltubeasoneequivalentchannelintroducesanegligibleerror.

AirTProfile,X=0
1
0.9
0.8
Z/W=0.1;=air
0.7
0.6 Z/W=0.5;=air
Y/Hf

0.5 Z/W=1;=air
0.4 Z/W=0.1;=3air
0.3 Z/W=0.5;=3air
0.2
Z/W=1;=3air
0.1
0
20 25 30 35 40 45 50 55
AirTemperature[C]

Fig.2.9AirtemperatureprofilesalongtheYdirectionattherefrigerantinlet(X=0)forthree
locationsalongtheZdirection.

Toanalyzethethermalevolutionoftheair,Fig.2.9presentsairtemperatureprofiles
alongtheYdirectionattherefrigerantinlet(X=0)atthreedifferentlocationsalongthe
Z direction.The detailed air discretization makes it possible to study the variation of
the air temperature not only along its flow rate direction but also in the direction
betweentubes.InFig.2.9,wecanobservethatthetemperatureofmostoftheairis
uniform,excepttheairclosetothetubewall.Onlyforahighvalueoftheairsideheat
transfer coefficient (about 180 W m2 K1) the air undergoes a small temperature
variationalongtheYdirection.Thisobservationagreesquitewellwiththeassumption
usedinthefintheorydevelopment.But,thetemperatureofairclosetothetubewall
ishigherbyupto15Kwithrespecttotherestoftheair,andthisfactisnottakeninto
accountinthefintheorydevelopment.

Finally,similarlytotheaimofFig.2.9tostudytheairflowevolution,Fig.2.10(a)and
(b) were plotted to study the fin temperature field at the refrigerant inlet section
(X=0).When=airthetemperaturefieldisquitesimilartothatforaonedimensional
field since the temperature gradient is almost negligible along the Z direction.
However,whentheairsideheattransfercoefficientincreases,duetothefactthatthe
air temperature variation along the Z direction significantly increases, a strong

49

2.FIN2DMODEL

temperaturegradientintheZdirectionappearsalongthefin,leadingtoaconsiderable
effect of the LHCZ. This fact points out the big impact of the airside heat transfer
coefficientontheseprofiles.

Fig.2.10Finwalltemperatureprofileattherefrigerantinletsection(X=0)forthecasestudy
with:(a)=air(b)=3air.

Fig.2.11(a)and(b)presentthetemperaturefieldsforthetubewall.Again,theresults
depend strongly on the airside heat transfer coefficient. When the airside heat
transferislow,basicallyonlytheLHCXinthetubeispresent,butwhenthiscoefficient
increases the effects of LHCZ in the tube also become visible since there is a
temperaturegradientonthewalltubealongtheZdirection.

Theimpactof2DLHCinthetubeandLHCZinthefinonthesolutionarediscussedin
thefollowingsection.
50

 2.FIN2DMODEL

Fig.2.11Tubewalltemperatureprofileforthecasestudywith:(a)=air(b)=3air.

2.5 Analysis of the SegmentbySegment NTU Modeling and Effect

of Classical Assumptions
OncetheFin2Dmodelhasbeenvalidateditcanbeusedasareferencetocheckthe
relative error made by the segmentbysegment NTUmodeling of a gas cooler and
for studying the impact of the classical assumptions, which are implicit in this
methodology. In order to evaluate the relative error, the reference used in this
evaluation was the solution for the case studyat the same operating conditions and
applyingthesamegridsize.

The classical NTU modeling approach divides each heat exchanger tube into
segmentsalongtherefrigerantflowwithitscorrespondingfins.Someauthorsuseonly
51

2.FIN2DMODEL

onesegmentpertube,whichiscommonlyreferredtoasthetubebytubeapproach.
Whenthetubeisdiscretizedinmorethanonesegment(Ns)theapproachisdefinedas
thesegmentbysegmentapproach.Oncetheheatexchangerisdividedintosegments,
the NTU relationships for heat exchangers (Incropera and DeWitt, 1996) are
employed for each segment. For multichannel crossflow heat exchangers the air is
alwaysconsideredtobeunmixedbecausethefinspreventthemixing,butthereare
two options for the refrigerant: to assume the refrigerant as mixed (RMAU) or as
unmixed (BU). In a multichannel tube, the refrigerant is actually unmixed, but some
authors assume the refrigerant flow as mixed, applying the RMAU relationship in a
segmentbysegment approach, e.g. Jiang (2003). However some other authors, e.g.
Fronk and Garimella (2011), apply the BU relationship using also a segmentby
segment approach. Thus, there is no full agreement in the literature regarding using
theRMAUandBUoptions.

The NTU models used in this analysis were developed within Engineering Equation
Solver (Klein, 1995). Both options available within the NTU modeling methodology
were included in this study: BU and RMAU. The NTU models used the same
propertiesandheattransfercorrelationsasthoseusedintheFin2Dmodel.

TheclassicalNTUmodelingpresentsthefollowingdrawbacks:

2DLHC:Asitwasexplainedintheintroduction,theNTUmethoddoesnot
accountfor2DLHCinthetube(LHCXandLHCZ)andLHCZinthefin.
Adiabaticfintip efficiency: This assumption is widely used even when a
temperaturedifferencebetweentubesexists.
Discretization inconsistency of the BU option: discretizing along the X
direction, i.e. introducing number of segments (Ns), involves an implicit
mixingoftherefrigerantstreamsincetheinlettemperatureatonesegmentis
evaluated as the averaged value at the outlet section of the preceding
segment. Consequently, for the BU NTU case, increasing the number of
segments is inconsistentwith the hypothesis of unmixed refrigerantstream.
Therefore,iftheunmixedconditionfortherefrigerantistheonewhichbetter
representstheactualprocess,thebestoptionforthediscretizationalongthe
Xdirectionwould betoemployatubebytubeapproach.Thiswillleadtoa
fullconsistentBUsolutionateachtubewithmixingattheoutlet.Thismixing
wouldbeperfectlyconsistentwiththerealoperationinthosemicrochannel
heat exchangers where the tubes end in the collector/distributor head. For
serpentine heat exchangers, the BU solution is not consistent because the
refrigerantismixed.

52

 2.FIN2DMODEL

Air temperature variation along the Y and Z directions: the NTU approach
assumesthattheairtemperatureisconstantalongtheYdirectionsincethe
NTUapproachusesthefintheory,whichisdevelopedunderthisassumption.
Thisassumptiondeviatesfromtherealitybecausethetemperatureoftheair
flowingclosetothetubeandthefinrootsbecomesmuchclosertothewall
temperature, as it was shown earlier. Additionally, fin theory assumes
uniform air temperature along Z direction. The impact of temperature
variationalongtheZdirectiononthefinefficiencyevaluationcanbereduced
by discretizing the fin along this direction. Given this discretization along Z,
theassumptionofuniformairtemperaturewouldbeonlyappliedtotheair
alongtheYdirection.

Itisimportanttonoticethatmostofthemodelsforheatexchangersarebasedonthe
classicalassumptionsanalyzedabove.Therefore,althoughtheydonotemployNTU
approach, they suffer of some of the drawbacks commented above, except the BU
discretizationinconsistency,thatisexclusiveforNTUmodels.

2.5.1 Comparison of Fin2D model against NTU approaches

ThescenariosusedtoanalyzedifferencesbetweensimulationpredictionsbytheFin2D
model and the classical NTU approaches are the same as those used for the case
study solution. In gas coolers such as serpentine or multitube heat exchangers with
largenumberofrefrigerantpasses,largetemperaturedifferencescanappearbetween
the refrigerant in neighboring tubes. In order tostudy the heat transfer in these gas
coolers, a new scenario has been added to the simulation studies. This scenario
modifies the cases presented previously by introducing a temperature difference
betweenrefrigerantinletsof40K,asshowninFig.2.3(b).

Fig. 2.12 (a) and (b) quantify the relative errors obtained when using the classical
NTU approaches. In these figures Ns represents the number of segments used to
discretize the tube length. For the RMAU case, the trend of the NTU model is
asymptotic to the Fin2D solution with a final error of 2.5 % for the air case, which
increases to 3.5 % for the airside heat transfer coefficient value increased threefold
(to about 180 W m2 K1). The simulations carried out for the scenario with different
refrigerantinlettemperaturesresultedinidenticalresults,whichmeansthattheerror
doesnotdependonatemperaturedifferencebetweenthetubes.

53

2.FIN2DMODEL

NTURefrigerantMixedAirUnmixed
7

6
=air
5 =3air

Qerror[%]
4

0
a) 0 2 4 6 8 10 12
Ns
NTUBothUnmixed
7

6 =air
=3air
5
Qerror[%]

0
0 2 4 6 8 10 12
b) Ns
Fig.2.12ComparisonofFin2DmodelandNTUmodelfordifferentnumberofrefrigerant
segments(Ns)intheXdirectionusedbytheNTUmodel:(a)usingRMAUrelationships,(b)
usingBUrelationships.

FortheBUcase,theerrorsaresmaller,below1.5%,indicatingthatthisapproachis
muchclosertotheFin2Dsolution.However,asitcanbeobservedinFig.2.12(b),the
error increases with the increasing number of cells. This problem is a result of the
modelling inconsistency that was pointed out and explained above. Following that
explanation,itwouldbeconsistentwiththeassumptionofunmixedrefrigerantmade
fortheBUcasetouseonlyonecellpertube;however,Fig.2.12(b)showsthemost
accurate solution when Ns=2. The reason for this result and trend is not clear since
manyeffectsoccursimultaneously.Themainreasoncouldbethatonesegment(Ns=1)
produces a linear temperature distribution for the fluids and a poor discretization of
the problem, thus when the number of segments is increased, the accuracy is
improved despite the modeling inconsistency. But with values Ns larger than two
54

 2.FIN2DMODEL

segments, the modeling inconsistency takes an overriding influence on the accuracy,

andtheerrorbeginstoincrease.

After observing these results it could be thought that the effect of the temperature
differenceoftherefrigerantbetweenneighboringchannelsisimportant.However,Fig.
2.8(b)andthecommentsmaderegardingFig.2.8(b)intheprevioussectionsdonot
supportthisidea.

2.5.2 Analysis of classical assumptions with Fin2D model

This section analyzes the impact of individual classical assumptions used in a heat
exchangermodelonthesimulationresults.Theimpactwasevaluatedwithrespectto
the complete Fin2D model prediction by imposing selected assumptions within the
Fin2Dmodel,andperformingadditionalsimulations.Theconsideredassumptionsare:
no 2D LHC in the tube, no LHCZ in the fin, adiabaticfintip, and uniform air
temperaturealongthefinheight(Ydirection).ThecasestudyCO2gascoolerwasused
inthesesimulations.

2.5.2.1 LHC effects

ToevaluatetheimpactoftheLHCeffectsandtoidentifythemostdominantareaswith
heat flow and its direction, four cases of simulations were performed with the
followingmodelingconstraints:(1)noLHCZinthefin,(2)noLHCXinthetubes,(3)no
LHCZ in the tubes, and (4) all LHC effects disabled for all wall elements, except
conductionalongtheYdirectioninthefin.Table2.3showstheerrorinthecapacity
predictions associated with eliminating from consideration selected LHC phenomena
with respect to the complete solution (case study solution which includes LHC in all
elementsanddirectionsenabled)andsamerefrigerantinlettemperature.

Table2.3Effectof2DLHConcapacity
Qerrorcase1 Qerrorcase2 Qerrorcase3 Qerrorcase4
[%] [%] [%] [%]
=air 0.03 0.12 0.09 0.66
=3air 0.24 0.10 0.55 2.54

TheeffectofLHCdependsstronglyontheairsideheattransfercoefficient.Whenthe
airsideheattransfercoefficientisequaltothereferencevalue,air,theinfluenceof
LHCisnegligible.Butwhentheairsideheattransfercoefficienthasathreetimesas
highvalue(about180Wm2K1),theeffectisnoticeable,2.54%.Thisimpactincrease,

55

2.FIN2DMODEL

whentheairsideheattransferisincreased,canbeexplainedbyobservingFig.2.10(b)
and 12(b). In these figures the temperature gradient along the Z direction in the fin
and along the X and Z directions in the tube rises when the airside heat transfer
coefficient increases, whereas the temperature gradient along the Z direction in the
tubeandfinisalmostnegligibleforthelowestairsideheattransfercoefficientvalue.

This increase in the prediction error due to neglecting the LHC effects when the air
side heat transfer coefficient is increased is consistent with the increase in the
prediction error shown in Fig. 2.12(a) for the NTU models. When the LHC has the
largestinfluence,thedominantcomponentistheLHCZinthetube.Itisimportantto
noticethattheLHCeffectsarestronglynonlinear.

Thecasewithatemperaturedifferencebetweentubeswasalsostudiedinthesame
wayasdescribedabove.Theresultsandconclusionsarethesame.Thisfactindicates
thattheLHCeffectsinanelementdonotdependontheconditionsofitsneighboring
elements.Theseconclusionsarenotvalidfortheheatconductioninthefinbetween
tubes,whichisstudiedbelow.

2.5.2.2 Adiabatic fin tip

Tostudytheeffectofassumingtheadiabatictipatthehalflengthofthefin,asitis
usuallyaccepted,thecasewiththe40Ktemperaturedifferencebetweenrefrigerant
inletswaschosen,sincetheadiabaticfintipassumptionisexactforthecasewiththe
samerefrigerantinlettemperature.Toquantifythiserrorandisolateonlytheeffectof
theadiabaticfintipassumption,thisscenariowasmodeledbyintroducingacutalong
thefinsurfaceandleavingallremainingLHCeffectsenabled.Thecutwasmodeledasa
cutalongtheairdirectioninthemiddlesectionofthefinsurfacebetweentubes(Fig.
2.3(b)).Undertheseconditionstheadiabaticfintipassumptionisstrictlycorrect,even
though there is a temperature difference between neighbouring tubes. Thus, the
difference between results for a scenario solved with and without modelling a cut
along the fin corresponds to assuming an adiabaticfintip when a temperature
differencebetweentubesexists.

Table2.4containsasummaryoftheobtainedresults.Tubes1and2referredtointhis
tablearedepictedinFig.2.3(b).Thedifferencebetweenthecapacitywiththefincut
and without it is negligible, which means that the improvement in the capacity is
almostzero.Nevertheless,Table2.4showslargeerrors,morethan300%,intheheat
capacitypertubecalculatedassumingadiabaticfintipefficiencywithrespecttousing
theactualfinefficiency.Thisfactmeansthatrefrigeranttemperatureattheoutletof
one tube would be wrongly predicted in case of using adiabaticfintip efficiency;
56

 2.FIN2DMODEL

nevertheless the impact on total capacity would result negligible for the analyzed
scenario.Thesedeviationsproduceadifferentheatfluxdistributioninthegascooler.
In fact, these results agree with the findings of Asinari et al. (2004). They also
concludedthattheheatfluxdistributionpertuberesultsdifferentdependingonthe
finefficiencyused,butthetotalheattransferredbythefinbetweentwotubes(sumof
theheatfluxforbothfinroots)isexactlythesameassumingeithertheadiabaticfin
tip or the actual one. This conclusion is independent of the temperature difference
between neighboring tubes. They studied a three passes gas cooler, and concluded
that the impact of adiabaticfintip assumption involves a modest effect on the total
capacity prediction, about 1%. However, Park and Hrnjak (2007) reported
improvementsincapacityupto3.9%bycuttingfinsforamicrochannelserpentinegas
cooler.Apossibleexplanationofthiscontradictionisthattheeffect,oftheheatflux
distribution in the gas cooler on the total heat capacity, depends of the number of
passes.Itresultsinanoticeabletotalcapacitydifferencewhenthenumberofpassesis
large, as is the case of the serpentine gas cooler studied by Park and Hrnjak (2007).
Currently, Fin2D does not have the capability to simulate complex circuitry
arrangementtovalidatetheabovehypothesis,buttheworktoenhanceFin2Dinthis
directionisunderway.

Table2.4Effectofassumingadiabaticfintipefficiencyoncapacity
Qwithout Qwith Qwithoutfin Qerror Qwithoutfin Qerror
fincut fincut cuttube1 tube1 cuttube2 tube2
[W] [W] [W] [%] [W] [%]
=air 15.35 15.37 1.87 274.87 17.22 29.73
=3air 21.32 21.35 1.1 313.64 20.23 16.96

Thewalltemperatureprofilesforeachsolution(when=air)areplottedinFig.2.13.
TheprofilesareshownalongtheYdirectionattherefrigerantinletsection(X=0)inthe
middleofthetubedepth.Theerrorinthecapacityofthefinroots,explainedabove,
can be interpreted from Fig. 2.13. It can be observed how different the actual
temperature profile is from the temperature profile when the adiabaticfintip is
assumed.TheslopeofthesecurvesintheYdirectiongivesthelocalheatfluxalongthe
finandfromthefintothetubes.Consequently,iftheslopeofthecurvesisanalyzed,it
iseasytonoticethedeviationoftheadiabaticfintipassumptionfromthereality;the
Fin2Dsolutionpresentsasignificantslopeinthemiddlesectionofthefinwhereasthe
adiabaticfintipassumptionimposesanullslopeinthemiddlesection.Fig.2.13shows
thatthesolutiontemperatureslopedoesnotchangeitssigninanysectionalongthe
fin height, which means that the fin receives heat from tube 2 and transfersheat to
thetube1(andtotheair).Theslopeforthecutfinchangesitssigndependingthefin
rootanalyzed,hencetheadiabaticfintipassumptionresultsinawrongheatfluxsign
57

2.FIN2DMODEL

calculation(notonlytheabsolutevalue)forthefinrootoftube1.Theconsequenceof
these differences is a large error in the heat capacity predicted for each tube and,
therefore,intherefrigerantoutletproperties.

1
0.9
0.8
0.7
0.6
Y/Hp

0.5
0.4
0.3 WithFinCut
0.2 WithoutFinCut
0.1
0
30 35 40 45 50 55 60
WallTemperature[C]

Fig.2.13Walltemperatureprofile(finandtubes)alongtheYdirectionattherefrigerantinlet
section(X=0)inthemiddleofthetubedepth(Z/W=0.5)forbothscenariossolvedwithFin2D
model:fincutandwithoutcut,=air.

2.5.2.3 Uniform air temperature along fin height

To study the assumption of constant air temperature along the Y direction, Fig. 2.14
presents the corresponding air temperature profile in the same locations as those
studied in Fig. 2.9, but now the studied scenario includes a 40 K temperature
differencebetweenrefrigerantinlets.

The results shown in Fig. 2.14 are similar to those shown for Fig. 2.9 except two
differences:thetemperaturedifferencebetweentheairclosetothetubeandtherest
oftheairisnowwithin10K,andtheairtemperatureprofileislessflat,particularlyat
the air outlet with the highest airside heat transfer rate, due to the temperature
difference between refrigerant inlets. This aspect is not accounted for by the fin
theory, since it assumes a uniform air temperature. Although not studied here, an
additional impact on the prediction results can be expected in an evaporator
simulation due to the large temperature difference between the bulk air and the air
closetothetubewall.Inanevaporatormodelinthepresenceofdehumidification,the
heat and mass transfer processes are strongly a function of local properties, which
dependonthelocaltemperatures.

58

 2.FIN2DMODEL

AirTProfile,X=0
1
0.9
0.8
Z/W=0.1;=air
0.7
Z/W=0.5;=air
0.6
Y/Hf

0.5 Z/W=1;=air

0.4 Z/W=0.1;=3air

0.3 Z/W=0.5;=3air
0.2 Z/W=1;=3air
0.1
0
20 25 30 35 40 45
AirTemperature[C]
Fig.2.14AirtemperatureprofilesalongtheYdirectionattherefrigerantinlet(X=0)forthree
differentlocationsalongtheZdirectionwhenadifferencetemperatureof40Kexistsbetween
refrigerantinlets.

2.6 Conclusions
Amodelformicrochannelheatexchangers,Fin2D,accountingforheatconductionin
alldirectionsandinallheatexchangerelementswaspresented.Themodelallowsfor
independent discretization for the refrigerant, tube and fins. The air has the same
discretization as the fins. After verification against known analytical solutions, the
modelwasemployedtoquantifypredictionerrorsassociatedwiththeclassicalNTU
modelling approach. Also, the classical assumptions were studied to evaluate their
impactontheaccuracyofsimulationresults.Thefollowingarethemainconclusionsof
thestudy:

The error obtained using the NTU method depends on the NTU
relationshipemployedtocalculatetheeffectivenessofeachsegment.Forthe
studied case it is smaller than 3.5% for RMAU, smaller than 1% for BU and
becomeslargerastheairsideheattransfercoefficientincreases.Ingeneral,
thebestoptionforthestudiedcaseistousethetubebytubeapproachand
to consider both fluids as unmixed although the effect of the mixed
refrigerant assumption turned out negligible in the scenarios studied.
However, this option can lead to larger errors when long length tubes are
59

2.FIN2DMODEL

simulated because refrigerant properties and heat transfer coefficients can

have significant variations, particularly when the refrigerant undergoes a
phase change. It is not consistent to apply a segmentbysegment approach
whentheRMAUrelationshipisadopted.
For the operating conditions studied, the impact of LHC effects along each
directioninfinsandtubewalls,ifconsideredseparately,isnotsignificant.The
combined effect is more noticeable and may result in a capacity prediction
errorofasmuchas2.5%,withtheLHCZinthetubebeingthedominanteffect.
TheimpactofLHCdependsontheairheattransfercoefficient.
Using the adiabaticfintip efficiency, which is commonly applied, leads to
large errors in heat distribution per tube, and therefore in the prediction of
temperature at the tube outlet, when a temperature difference between
tubes exists. However impact of this assumption in the total gas cooler
capacityismuchless.Thisassumptionaffectstheglobalcapacitypredictionof
gas coolers with large number of refrigerant passes. Thus, the fin cuts are
justifiedintheseheatexchangertopologies.
Thetemperatureofairclosetothetubewallisverydifferentthanthebulkair
temperature. This fact could have an important impact on local effects
controlling the heat and mass transfer, e.g. dehumidification. It would have
been interesting to evaluate the isolated effect of the nonuniform
temperatureprofileoftheairalongthefinheight.
Thedevelopedmodelisabletocapturemostofthesecondaryheattransfer
effectsnottakenintoaccountbytheclassicalNTUapproachoranymodel
whichappliesthedescribedclassicalassumptions;however,simulationofthe
2D LHC problem in the wall requires a considerable computation time. The
authorswillcontinueworkingonasimplifiedmodelthatwillretainthemost
important effects. This will lead to much lower computation times while
providing high accuracy of prediction of the complex heat transfer
phenomenatakingplaceinairtorefrigerantmicrochannelheatexchangers.

60

CHAPTER3
FIN1Dx3MODEL

61

3.FIN1Dx3MODEL

3 FIN1Dx3 MODEL
PreviouschapterproposedamodelforamicrochannelgascoolerreferredtoasFin2D
model. The aim of developing Fin2D model was to evaluate the prediction errors of
classical modelling assumptions and techniques described above, in an equivalent
pieceofamicrochannelgascooler,andidentifyerrorsources.

Fin2D model was sufficient to identify the deficiency sources of the classical
methodologiesinsuchkindofheatexchangers,butwasnogoodtoevaluatetheglobal
performance prediction errors when an actual MCHX is simulated regarding
dimensions, number of tubes and number of refrigerant passes. The main reason
whichdidnotallowstudyinganactualmicrochannelgascoolerwasthecomputational
costoftheFin2Dmodel.Thismodelhasalargecomputationalcost,mainlyduetothe
fin surface discretization: the model needs to employ a large number of fin cells
because any fin efficiency is used to solve the heat transfer equation along the fin,
which involves heat convection and heat conduction. Furthermore the Fin2D model
applies the same discretization for both air and fin, so the air also introduced an
importantcomputationaleffort.

ThegoalofthepresentworkistodevelopamodelbasedontheFin2Dmodel,without
modellingthenegligibleeffectsandchangingthemodelstructureand/ordiscretization
inordertoreducethecomputationalcost,providingasimulationtoolforMCHXswith
reasonable computational cost for design purposes. To this end, the authors
developed the Fin1Dx3 model, with a novel formulation to take into account heat
conduction between tubes without applying adiabaticfintip assumption, thus any
correctiontermwasneeded.Theresultisamuchfastermodelwithalmostthesame
accuracyasFin2Dmodel.

Given this computational time reduction, the proposed model is able to simulate
either a microchannel gas cooler or a condenser with any refrigerant circuitry. Thus,
experimental data from measurements of an actual condenser and a gas cooler was
usedtovalidatethemodel.Thismodelwillalsoallowevaluatetheisolatedeffectof
the nonuniform air temperature profile along the fin height; fact that could not be
studiedwithFin2Dmodel.

62

 3.FIN1Dx3MODEL

3.1 Fin1Dx3 Heat Exchanger Model

Themodelproposedinthischapter,whichwillbereferredtoasFin1Dx3,isbasedon
theFin2Dmodel,performingsomechangesinordertoreducethecomputationalcost
butpreservingaccuracy.Thechangesarebasedonthefollowingconsiderations:

Chapter 2 revealed that the longitudinal conduction in the fin along the air
flow direction resulted in a negligible effect on the predicted performance
results.Inaddition,severalcurrentfinsurfaceshavecutsalongthisdirection
(louvered,slit,lanced)cancellingLHCinthatdirection.Thus,inthepresent
modelthiseffectiscancelled,whatmeansinpracticenothermalconnections
between neighbouring fin cells along the air flow direction, even though a
discretizationofboththefinandairexistsalongthisdirection.
Chapter 2 revealed that the air temperature profile is quite flat along the
direction between tubes, excepting the air close to the tube wall. In those
studies the height of the fin occupied by the air close to the tube with a
temperature differentfrom the rest of air was about 1/30 of the fin height.
Thediscretizationofairalongthatdirectionincreasesthecomputationalcost.
On the other hand it would be quite interesting to capture the effect of
accountingforthetemperaturedifferencebetweentheairclosetothetube
wallandtherestoftheair.Apossiblesolutionforthisconflictofinterestsis
todiscretizetheairwiththreeaircellsalongtheYdirection,asshowninFig.
3.1.Forthisdiscretization,theheightoftheaircellsclosetothetubewallis
unknown: this dimension (Hf) should be that one which provides the best
results, so this dimension is a parameter to adjust either experimentally or
numerically, even by observation. The only restriction is that both fin cells
(cells close to the tubes), for the same Z, will measure the same. What
actually these three air cells represent is the consideration of nonmixed air
along the Y direction between them. This idea makes sense for air flowing
throughlouveredfinsandforlaminarflows,i.e.:theaircouldbeassumedas
mixedalongallthelouverheightbutasnonmixedwiththerestofaircloseto
thetubewalls,sincethefinheightisgreaterthanlouverheight,asitisshown
in Fig. 3.1; laminar flow represents a nonuniform temperature distribution
alongtheYdirection.
TheFin2Dmodelsolvedtheheattransferredfromtheairtothefinwithout
applyingthefintheory.ConsequentlyFin2Dmodelneedsalargenumberof
fin cells along the Y direction to solve accurately the 2D heat conduction in
thefin.Thiscalculationistheprocedurewhichrequiredmorecomputational
costwithintheFin2Dmodel.

63

3.FIN1Dx3MODEL

Thefintemperature,forauniformfin,isgovernedbyEq.(3.1).Onlywhenthe
airtemperatureisconstantthroughoutthefin,Eq.(3.1)canbeexpressedas
Eq. (3.2), (Incropera and DeWitt, 1996). If heat transfer properties are
constant,ortheyareevaluatedwithmeanvalues,thegeneralsolutionforthe
Eq.(3.2)isEq.(3.3).

d 2T f
m 2 T f Ta 0 (3.1)
dY 2

d 2 f ,a
m 2 f ,a 0 (3.2)
dY 2

pw
m2
kA

f,a(Y) C1emY C2emY (3.3)

First, fin theory assumes uniform air temperature along Z direction. The
impact of temperature variation along the Z direction on the fin efficiency
evaluationisreducedbydiscretizingthefinalongthisdirection,asthemodel
proposed in this chapter does. Given this discretization along Z, the
assumptionofuniformairtemperaturewouldbeonlyappliedtotheairalong
theYdirection.Thus,whenequations(3.1),(3.2)and(3.3)areused,themain
assumptionofthefintheorywhichisnotsatisfiedinan actualfinsurfaceis
that the air temperature is not uniform along the Y direction. In the model
proposedinthischapter,thediscretizationforthefinandtheairisthesame
andthediscretizationintheairhasbeenchoseninordertorepresentaircells
with uniform temperature, so it could be possible to apply the fin theory
solution (Eq. (3.3)) for each airfin cell connection without failing the
assumption of uniform air temperature. The result of this methodology is a
greatreduction,incomparisonwithFin2D,ofthegridsizeandconsequently
ofthecomputationalcost.

ItisshouldbenotedthatEq.(3.3)doesnotimplytheclassicaladiabaticfintip
assumption(inthecrosssectionathalffinheight),sinceboundaryconditions
havenotbeenappliedyet.TheevaluationoftheconstantsC1andC2 willbe
exposedinthesubsection3.1.2.

64

 3.FIN1Dx3MODEL

Fig.3.1Detailofalouveredfinsurfaceinamicrochannelheatexchanger,wherethenon
louveredheightandthetotalfinheightaredepicted.

3.1.1 Heat Exchanger Discretization

Fig.3.2presentsasampleofaMCHXthatcanbesimulatedwiththeproposedmodel.
The model can simulate any refrigerant circuitry arrangement: any number of
refrigerant inlets and outlets; and any connection between different tube
outlets/inletsatanylocation.

Fig.3.2ExampleofamicrochannelheatexchangerthatcanbesimulatedbyFin1Dx3.

Fig.3.3showsthediscretizationinsegmentsoftheheatexchangershowninFig.3.2,
where the thinner lines correspond to thermal connections between wall cells,
whereas the thicker lines correspond to the refrigerant flow path. First, the heat
exchangerisdiscretizedalongtheXdirection(refrigerantflow)resultingNssegments
pertube.ThediscretizationforeachsegmentisthesameanditisshowninFig.3.4.
Eachsegmentconsistsof:arefrigerantstreamthatissplitintoNr,ZchannelsintheZ
direction;aflattubewhichisdiscretizedintoNt,ZcellsintheZdirection;andbothair
flow and fins, which are discretized in two dimensions:Na,Y=3 cells in the Y direction
andNa,ZcellsintheZdirection.Sincethediscretizationfortheairandfinwallisthe
same,Nf,Y=Na,YandNf,Z=Na,Z.

65

3.FIN1Dx3MODEL

The discretization fora heat exchanger is summarizedin the following as agrid: {Ns,
Nr,Z, Nt,Z, Na,Z}. For illustration of the nomenclature, the numerical example shown in
Fig.3.3andFig.3.4correspondstoagrid:{3,4,3,2}.

The refrigerant flows inside the channels along the X direction without any mixing
betweenthechannels,anditexchangesheatwiththetubecellsincontact;thesetube
cellstransferheattotheaircellsincontactbyconvection,toitsneighboringtubecells
ontheplaneXZbyconduction,andtothefinrootsincontactbyconduction.Theair
exchangesheatbyconvectionwiththefincells,andtheaircellsatthebottomandtop
alsoexchangeheatwiththetubecellsincontact.Thefincellsconducttheheatalong
the Y direction, and the bottom and top fin cells also conduct heat to the neighbor
tubewall.

Fig.3.3DiscretizationinsegmentsoftheheatexchangershowedinFig.3.2,whichincludesthe
thermalconnectionsbetweendifferentsegmentsandflowarrangement.

Regardingthefluidcells,eitherairorrefrigerant,therearetwotypologies:elemental
cell and mixture cell. The elemental cell corresponds to the one described above,
exchanging heat with the surrounding wall cells. The mixture cell is adiabatic and its
function is collecting the fluid from a number of tubes and distributing into the next
tubesaccordingtheheatexchangercircuitry.Theinletandoutletportsofeachtube
areconnectedtothecorrespondingmixturecells.Thedistributionofthesefluidcells
and the definition of the tubes connected to these cells determine the flow path of
each fluid. In the proposed model, any configuration can be fixed, thus heat
exchangers such as serpentine or parallel tubes MCHXs can be simulated with any
refrigerantcircuitry.

66

 3.FIN1Dx3MODEL

Fig.3.4Schematicofasegmentdiscretizationintocells.

The model is designed to allow this methodology also with the air, but in this work,
thesemixturecellswereonlyusedfortherefrigerantfluid.InFig.3.3,theserefrigerant
cellsarerepresentedwiththeroundshapeboxes.

3.1.2 Governing Equations

Everyfluidcell(eitherrefrigerantorair)hastwonodes,whichcorrespondtotheinlet
andoutletsectionsinthefluidflowdirection.Thetubewallcellshaveonlyonenode
located in the centroid of the cell, as shown in Fig. 3.5(a). All cell local variables are
referredtothevalueinthesenodes,e.g.Tr,inandTr,outarethetemperatureattheinlet
andoutlet,respectively,ofarefrigerantcellr;fortheairflowwouldbethesamebut
with subscript a; Tt is the temperature of the tube cell t. The fin does not have any
nodebecauseacontinuousfunctiongovernsinthefin.

67

3.FIN1Dx3MODEL

Fig.3.5Differentviewsofadiscretizedportionofheatexchanger:(a)globalviewillustrating
fluidnodesandtubedirections;(b)ZYplane,whichshowsmaingeometricdataofthefinand
regionswhereisdefinedthecorrespondingTf(Y)and T a ;(c)XYplane,whichshowsthe
locationoftheTfTandTfBtemperatures.

According to the assumptions and methodology explained above, the governing

equationsforthetubewallwillbedifferentfromthoseappliedtothefinwall.So,the
description of the governing equations is going to be structured in four blocks:
refrigerant flow, air flow, tube wall and fin wall with their corresponding thermal
connectionsbetweenneighborfluidsorwallcells.

3.1.2.1 Tube wall and its thermal connections

Theenergyconservationequationforatubecellcanbewrittenas:

nr na
kt,k Tt dV qt,r pwt,r dst,r qt,a pwt,a dst,a 0 (3.4)
r 1 a 1

68

 3.FIN1Dx3MODEL

whereanytubewallcelltisincontactwithnrrefrigerantcellsr=1,nrandwithnaair
cells a=1, na; kt,k is the thermal conductivity of the tube cell t in the k direction. The
Laplaciantermisthetermthatallowsstudyingtheinfluenceof2DLHCalongthetube
walls, and it also takes into account the heat transferred by conduction between a
tubecellandtheadjoinedfincells.Theheatfluxes qt,r and qt,a forathermalconnection
betweenatubewallcelltandafluidcelli,eitherairorrefrigerant,areevaluatedas
follows,

qt,i U t,i Tt Ti (3.5)

wheretheoverallheattransfercoefficient U t, i forthisconnectioncorrespondsto:

1/At,i
Ut, i
tt / 2 1

At,i kt,k At,i t,i

ByusingEq.(3.5),Eq.(3.4)canberewrittenasfollows,

ni
kt,k Tt dV pwt,iU t,i Tt Ti dst,i 0 (3.6)
i 1

whereni=nr+na.

3.1.2.2 Refrigerant flow

Eq.(3.7)statestheenergyconservationforarefrigerantcellrincontactwithnttube
wallcellst=1,nt.

nt
m r dhr q t,r pwt,r dX (3.7)
t 1

Thefluidpressuredropalonglengthoftherefrigerantcellr,isobtainedbyintegration
ofthemomentumequation,

Pr Pr , fr Pr ,acc Pr , g (3.8)

69

3.FIN1Dx3MODEL

wherethefrictiontermisevaluatedintheform:

X r
Pr , fr f r m r2 (3.9)
Dr Ac r ,in r ,out
2
r

Thefrictionfactorfrcanbeevaluatedbyseveralcorrelationsavailableintheliterature
(KaysandLondon,1984).Theaccelerationterm,canbeexpressedby:

1 1
2
m
Pr ,acc r (3.10)

Acr r ,out r ,in

Finally,thegravitationalterm,canbeevaluatedasfollows:

g r ,in r ,out X r sin

1
Pr , g (3.11)
2

The mixture cells, which in this work are only used for the refrigerant flow, have
another formulation since they are adiabatic and only collect and/or distribute the
refrigerant flow. The governing equations implemented in these cells determine the
fluid flow distribution: uniform or nonuniform. In the present work, uniform flow
distributionwasassumed.Anobjectorientedprogrammingallowsaneasierchangeof
amodelwhichassumesuniformflowdistributiontoonewhichdoesnot,becausethe
equationstodescribethisphenomenaareonlylocatedinthemixturecellduetothe
modular capability of this programming technique. According to the assumption of
uniform flow distribution, any outlet of the mixture cell r has the same conditions
whicharecalculatedwith:

nr,in

m r,in
m r,out 1 (3.12)
nr,out

hr,out hr,in (3.13)

whereamixturecellrisconnectedtonr,ininlettubesandnr,outoutlettubes.

A mixture cell in this kind of heat exchanger would be a portion of header tube, so
thereisapressuredropduetotheinsertionofthetubesintotheheader,whichcanbe
treatedasasuddenexpansionorcontraction.Additionally,frictionalandgravitational
pressuredropalongtheheaderscouldbeevaluatedapplyingEq.(3.9)andEq.(3.11)to
thecorrespondingmixturecell.

70

 3.FIN1Dx3MODEL

3.1.2.3 Fin wall and its thermal connections

The present model discretizes the fin height, together with the air in contact, into
threecells(Fig.3.5(b));twoshortcellsofequalheight,whichareincontactwiththe
correspondingtubes,andacentralcell.Thereasontoapplythisdiscretizationisthatin
this way, the assumption of uniform temperature along Y direction of the air cell in
contactwitheachfincellismorecorrect.Inthismanner,Eq.(3.3)canbeappliedmore
fundamentallytoeachfinairconnectionthaninasituationwithjustoneaircell.Thus,
intheproposedmodelthefintheoryisappliedtoeachfinairconnection,whatmeans
applyingEq.(3.3)foreachfincell,resultingforacolumnoffincellsthesystemofEq.
(3.14). That is the reason for the models name: 1D because it applies a one
dimensional equation for each finair connection and x3 because it is applied for
threeconnectionsperfin.

f , a1 (Y ) C1 e m f , a1Y C 2 e m f , a1Y , 0 Y H f

m Y H f m Y H f (3.14)
f ,a (Y ) f , a 2 (Y ) C 3 e f , a 2 C 4 e f ,a 2 , H f Y (1 ) H f
m f , a 3 Y 1 H f m Y 1 H f
f , a 3 (Y ) C 5 e C 6 e f ,a 3 , (1 ) H f Y H f

InEq.(3.14)Hfisthefinheightandisthenondimensionalheightoffinandaircells
at the bottom and top of the fin. Eq. (3.14) assumes uniform air temperature inside
eachregion,ontheYZdirections,thustoobtainthefinwalltemperature(Eq.(3.15)),
an integrated mean valueisused for the air temperature, which corresponds to T a .
ActuallythisvaluecorrespondstoanintegratedvaluealongtheZdirectionsincethe
airtemperatureforeachcorrespondingregionisuniformalongtheYdirection.

T f 1 (Y ) f ,a1 (Y ) Ta1 , 0 Y H f

T f (Y ) T f 2 (Y ) f ,a 2 (Y ) Ta 2 , H f Y (1 ) H f (3.15)

T f 3 (Y ) f ,a 3 (Y ) Ta 3 , (1 ) H f Y H f

71

3.FIN1Dx3MODEL

The unknown constants: C1, C2, C3, C4, C5, C6 must be evaluated from the boundary
conditionsoftheheattransferproblemalongthefinheight,i.e.thetemperaturefield
must be continuous and derivable. Therefore, for a continuous fin, the conditions to
evaluatetheconstantsare:

T f 1 (Y 0) T fB

T f 1 (Y H f ) T f 2 (Y H f )
T (Y H ) T
f3 f fT

T f 2 (Y (1 ) H f ) T f 3 (Y (1 ) H f ) (3.16)

dT dT f 2
f1
dY Y H dY Y H
f f

dT f 2 dT f 3

dY Y (1 ) H f dY Y (1 ) H
f

where TfB and TfT, illustrated in Fig. 3.5(c), correspond to the temperature at the
bottomandtopofthefininthebaseofcontactwiththebottomandtoptubecells,
respectively.Inthiswayitispossibletodefine T f asfollows,

Ta1

T f 1 (Y ) Ta 2
(3.17)
T f (Y ) T f 2 (Y ) [ A(Y )]Ta 3
T
T f 3 (Y ) fB
T fT

[A(Y)] is a 3x5 matrix that depends on Y, geometry, airside heat transfer coefficient
and fin conductivity. Note that Tf has the interesting feature that is a pseudolinear
functionwithrespectto Ta ,TfBandTfT.

Governingequationsarepresentedforacontinuousfin,asithasbeenexposedabove,
butthismethodologyallowsmodelingeitheracontinuousfinorafinwithacutalong
the air flow direction and located at half the fin height. Motivations for this fin cut
were mentioned in Introduction and they will be widely discussed in subsection 3.3.
Thewaytoincludethiscutonlyaffectstothedefinitionoftheboundaryconditionsin
order to get the unknown constants of Eq. (3.14): C1, C2, C3, C4, C5, C6, and
consequently[A(Y)].AppendixAincludesthedefinitionoftheseboundaryconditions
foraparticularcase.Restofmodelwouldbejustthesame.

72

 3.FIN1Dx3MODEL

Once we have an expression for the fin wall temperature function, the governing
equationsforthethermalconnectionsbetweenfinandtube(Eq.(3.18)and(3.19))can
beobtainedbyimposingthattheheatconductionatthebottom(Eq.(3.18))andtop
(Eq.(3.19))ofthefinareequaltotheheatconduction,intheYdirection,throughthe
ntwalltubecellsincontact.Theconductionareasbetweenatubecelltandthetop
andbottomofthefinare A J and A J ,respectively;thecorrespondingconduction
T B

areasforthebottomandtopofthefinare A fB and A fT ,respectively.

k f 1T f 1
nt
d
A
t
t ,J B
Y
k t , J B Tt A fB
dY
(3.18)
Y 0 Y 0

nt
A t , JT kt , JT Tt A fT
d
k f 3T f 3 (3.19)
t Y Y H f dY Y H f

3.1.2.4 Air flow

Eq.(3.20)statestheenergyconservationinanaircellaincontactwithafincellfand
nttubecellst=1,nt.

nt
m a dha dQ f,a q t,a pwt,a dZ (3.20)
t 1

The heat flux qt,a exchanged with each tube cell t is calculated by applying Eq. (3.5),
whiletheheattransferredtotheneighbourfincellf,canbeevaluatedby:

dQ f ,a f ,a pwf ,a f ,a dY (3.21)

The air pressure drop along the fluid cell i length is obtained by applying the
momentumequation,

Pa Pa , fr Pa ,acc Pa ,cont Pa ,exp (3.22)

where Pa , fr and Pa,acc areevaluatedwiththeEq.(3.9)and(3.10),respectively.The

pressure drop terms due to the sudden contraction and expansion in the heat
exchangerareobtainedfollowingKaysandLondon(1984).

For solving the system of equations a set of boundary conditions is needed. Inlet
conditionsandvelocitydistributionsareknownforbothfluids.Additionally,theheat

73

3.FIN1Dx3MODEL

transferredbythewalledgestothesurroundingsisusuallyconsiderednegligible,so
the edges of the extreme wall cells are adiabatic with the surrounding. The two
extremetubes,intheYdirection,oftheheatexchangerareusuallymodelledusinga
finpackage,whichisplacedinthesurroundingside,withaheightofhalfafinheightof
therestoffins.However,inrealheatexchangers,theseextremetubesarefinnedwith
wholefins,inbothsidesofthetube,andtheheatexchangerisclosed,atthebottom
andtop,withtwometalplates.Intheproposedmodel,thesetwometalplateshave
been modelled as empty tubes, i.e. tubes with the same geometry as the rest but
withoutrefrigerantflowinginside,whichareadiabaticwiththesurrounding.

3.1.3 Numerical Scheme

Inordertodiscretizethegoverningequationspresentedintheprevioussubsection,a
finite volume method (FVM) was applied. In the governing equations the wall
temperature has been considered as the iterative variable of the problem, and the
semiexplicit method for wall temperature linked equations (SEWTLE) proposed by
Corbern et al. (2001) has been employed to solve the problem. The use of the wall
temperature as independent variable gives more freedom to formulate the heat
transfer phenomena, allowing the formulation of equations for energy conservation
withfewerassumptionsthanclassicalNTUapproaches.Additionally,usingthewall
temperature as independent variable of the thermal problem, converts an implicit
problem into a semiexplicit problem, by solving at each iteration a series of explicit
steps.

In order to integrate the Laplacian operator in Eq. (3.6), it has been discretized by a
classicalfinitedifference(finitevolume)approach.

ni
at Tt a T at,J B Tt , fB at,JT Tt , fT pwt,iU t,i Tt Ti dst,i
t,k t,k
(3.23)
k W,E,S,N i 1

k t, W At,W k t,E At,E k t,S At,S k t, N At, N k t,J B At,J B

a t,W a t,E a t,S a t, N at,J B
l t,W lt,E l t,S l t, N lt,J B

k t,JT At,JT
at,JT
l t,JT
at a t, k
k W,E,S,N,J B ,J T

Notice that the term in Eq. (3.23) of the heat transferred to the fluids in contact is
applied to both refrigerant and air cells in contact with a tube cell t. The direction
reference used in the model for k is shown in Fig. 3.5(a). All at,k terms refer to the

74

 3.FIN1Dx3MODEL

conductancebetweenatubecelltandtheneighbourtubecellinthekdirection. a t, J
B

and a t, J are the conductance of the connection between a tube cell t and the
T

correspondent fin base, at either the bottom or the topof each fin respectively. For
thisformulationithasbeenassumedthatthetubewalltemperatureTtisuniformin
theZXdirection.

Tocontinuediscretizingthesetofgoverningequations,firstitisnecessarytoassumea
temperature profile for the tube walls, or for the fluids, in order to obtain the
estimation of the integral of the heat transferred to the fluids in contact with a
consideredpieceofwall(Eq.(3.5)and(3.6))inthefluidflowdirection.Thisintegration
must be consistent with the integration of the coincident terms of fluid energy Eq.
(3.7) and Eq. (3.20). The linear fluid temperature variation scheme (LFTV) has been
assumed for both fluids, as Corbern et al. (2001) suggested for this application,
leadingtothefollowingexpression:

T T
At,i qt,i U t,i pwt,i Tt i,in i,out st,i (3.24)
2

Eq. (3.24) is valid for both refrigerant and air cells in contact with a tube cell t.
Substituting Eq. (3.24) in Eq. (3.18), the tube wall temperature can be evaluated as
follows,

ni

at,kTt,k at,JB Tt,fB at,JH Tt,fH pwt,iU t,i 0.5 (Ti,out Ti,in )st,i
Tt k W,E,S,N i 1 (3.25)
ni
at pwt,iU t,i st,i
i 1

BycombinationofEq.(3.7)andEq.(3.24),theoutgoingtemperatureofarefrigerant
cellr,canbeexpressedby,

nt
nt
Tr,in 1 0.5 NTU t,r NTU t,r Tt

Tr,out t 1 t 1 (3.26)
nt

1 0.5 NTU t,r
t 1

pwt,iU t,i st,i

NTU t ,i
m i Cpi

75

3.FIN1Dx3MODEL

Eq. (3.26) is used for a onephase flow whereas for twophase flow the outlet
temperaturedependsontheoutletpressure.

To obtain the outgoing temperature of the air, Eq. (3.20) has to be solved, so the
integrationofEq.(3.21)mustbedonepreviously.Thetotalheattransferalongthefin
cellcanbeexpressedas:

dQ f ,a f ,a pwf ,a f ,a dY f ,a Af ,a f ,a (3.27)

where f , a istheintegratedmeanvalueof f ,a (Y ) .Anovelaspectofthismodelisthat

inordertoincludeheattransferfromfintoair,integrationoftemperaturedifference
f , a isimplementedinthemodel,whilerestofmodelsusedirectlyafinefficiencyby
applyingtheanalyticalrelationshipforadiabaticfintipassumption, tanh( mL ) / mL .

The advantage of using the integration of f , a is that allows taking into account the
heat conduction between tubes more easily and fundamentally than other fin
efficiencybasedapproaches.Furthermoreafinefficiencycannotbealwaysbedefined,
e.g.whentemperatureatfinrootsarenotidentical.Thisfactleadstosomemodels,
whichusetheadiabaticfintipefficiency,toapplymoreorlessartificialapproachesin
ordertoincludeheatconductionbetweentubes.Itisimportantnoticethatthisideais
independent on the discretization applied in air and fin, i.e. with just one air cell,
instead of three as this paper proposes, applying this idea is possible. Thus, there is
neitheraccuracynorcomputationalcostreasontoapplyanapproachbasedontheuse
of an adiabaticfintip efficiency instead of the previous methodology, which is
fundamentally more appropriated. Following section 3.2 will present an accuracy
comparisononresultsbetweenbothmethodologies,whenoneairandfincellisused
insteadthree.

ByusingEq.(3.27),Eq.(3.20)canbewrittenforeachregionofthefinonthefollowing
way,

nt

m a1h a1 f,a1 A f,a1 f,a1 t 1

q t,a1 pwt,a 1 Z t,a1
(3.28)
m h A
a 2 a 2 f,a 2 f,a 2 f,a 2 0
m h nt
a 3 a 3 f,a3 A f,a 3 f,a 3 q pw Z

t 1
t,a 3 t,a 3 t,a 3

76

 3.FIN1Dx3MODEL

The linear fluid temperature variation (LFTV) approach was also assumed for the air,
alongtheZdirection,thusthesetofEq.(3.28)isrewrittenas,

nt
2 T a1 T a1,in NTU f,a1 f,a1 t1
NTU t,a1 Tt T a1

2 T a2 T a 2,in NTU f,a2 f,a2 0

(3.29)

2 T a3 T a3,in NTU f,a3 f,a3 NTUt,a3 Tt T a3
nt
t 1

f,a Af,a
NTU f ,a
m aCpa

where,NTUt,aisdefinedforthethermalconnectionsbetweenatubecelltandanair
cella,whereasNTUf,aisdefinedforthethermalconnectionbetweentheaircellaand
theattachedfincellf.

In Eq. (3.29) the term f , a corresponds to T f T a , whilst T f can be obtained by

integrationofEq.(3.17)resultingthefollowingequations,

H f [ A(Y )]
0 1, j

Hf Ta1

T f 1 (1 ) H f Ta 2
H f [ A(Y )]2, j (3.30)
T f T f 2 Ta3
( 1 2 ) H
T f 3
f
T fB
H

f

(1 ) H f [ A( Y )] 3, j T fT

Hf

Now,if T a issubtractedfrom T f andrearrangingtheresult, f ,a canbeexpressedas,

Ta1

f ,a1 T f 1 Ta1 Ta 2
(3.31)
f ,a 2 T f 2 Ta 2 B Ta 3
T
f ,a 3 T f 3 Ta 3 fB
T fT

77

3.FIN1Dx3MODEL

[B]isa3x5matrixthatdependsonthesameparametersas[A(Y)]exceptingY. f ,a
depends on the outlettemperatures of all the air cellslocated at thesame Z (of the
same segment), however note that f ,a has the interesting characteristic, same as

T f (Y ) ,thatisapseudolinearfunctionwithrespectto T a , TfB and T fT .Thisfactwill

reportinterestingcomputationcapabilities,whichwillbedescribedinnextsection.In
ordertogetthematrix[B]manyalgebraicoperationshavetobedone,thusAppendix
Acontainsthedefinitionoftheseterms,forbothcontinuousfinandcutfin.

If Eq. (3.31) is substituted in Eq. (3.29), and rearranging, the system of equations to
solveisthesetofEq.(3.32).Theaverageairtemperatureandconsequentlytheoutlet
airtemperatureforeachsegmentareobtainedsimultaneouslybysolvingthesystem
of Eq. (3.32). The solution for a system of 3 linear equations is known and easy to
compute.

nt
2 NTU t ,a1
t 1
B1,1 B1, 2 B1,3
NTU f ,a1
T a1
2 T a 2
B2,1 B2, 2 B2,3
NTU f ,a 2
nt T a 3
2 NTU t ,a 3

B3,1 B3, 2 t 1
B3,3
NTU f ,a 3
nt
2Ta1,in Tt NTU t ,a1
t 1
T fB B1, 4 T fT B1,5
NTU f ,a1
2Ta 2,in
T fB B2, 4 T fT B2,5
NTU f ,a 2
nt
2Ta 3,in Tt NTU t ,a 3


t 1
T fB B3, 4 T fT B3,5 (3.32)
NTU f ,a 3

78

 3.FIN1Dx3MODEL

Now,wehavetodiscretizeEq.(3.18)and(3.19)inordertorelatetemperatureatthe
bottom and top of a fin with the temperature at the corresponding connected tube
cells.Thediscretizationoftheseequations,consideringtheproposedheatexchanger
discretizationandusingsometermsofEq.(3.23),is

nt dT f 1
at , J B Tt T fB A fB k fB
t 1 dY Y 0 (3.33)
nt
a T T A k dT f 3
t , JT t fT fT fT
dY Y H
t 1 f

where the derivative of the temperature at the bottom and top of the fin can be
evaluatedfromEq.(3.17),resulting

T a1
dT f 1
T a 2
dY Y 0
C T a 3
(3.34)

dT f 3 T
dY fB
Y H f
T fT

where [C] is obtained by deriving the corresponding rows of [A(Y)] (Eq. 3.17) and
evaluating them at Y=0 or Y=Hf, respectively. Because of same reasons as for matrix
[B],thecoefficientsofmatrix[C]areexposedintheAppendixA,forbothcontinuous
finandcutfin.

BysubstitutionofEq.(3.34)inEq.(3.33)resultsthefollowingsystemofequations,

nt

1, 4 t ,B
C C1,5 T
t 1 fB
nt
T
C2, 4 t ,T C2,5 fT
(3.35)
t 1
nt

C T
1,1 a1 C T
1, 2 a 2 C T
1, 3 a 3 t ,BTt
t 1
n

C T a1 C T a 2 C T a3 t T
2,1 2, 2 2,3 t 1
t ,T t

79

3.FIN1Dx3MODEL

Wherethefollowingexpressionshavebeenused,forthethermalconductanceatthe
bottomandtopofthefin:

at , J B at , J T
t ,B , t ,T
A fB k fB A fT k fT

Thissystemisagainalinearsystemofequations,anditisemployedtorelate TfB and

T fT with T a and Tt .

At this point all the equations which describe the thermal problem have been
discretized.Thepressuredropequations,presentedintheprevioussubsection3.1.2,
werealreadyindiscretizedform.

To summarize, the proposed model applies to each segment the equations: (3.25),
(3.26),(3.32),(3.35),with TfB , TfT , Tt , Tr ,out and Ta,out beingtheunknownvariablesof
the problem for each of the cells of the heat exchanger. Once these variables are
known,anyperformanceparameteroftheheatexchangercanbecalculated,including
thefintemperatureatanypositionYbymeansofEq.(3.17).

3.1.4 Solution Methodology

TheglobalsolutionmethodisbasedontheSEWTLEmethod,outlinedbyCorbernet
al.(2001),withsomedifferencesduetotheparticularitiesofthepresentmodel.

Theproposedsystemofequationsconsistsofasystemofnonlinearequationssince
coefficients and properties depend on the temperature and pressure field. The
functionsofthepropertiesandcoefficientsarestronglynonlinearandtoocomplexto
introduce directly in the system of equations. Thus, the solution needs an iterative
process. A first option could be to start solving the problem assuming constant
coefficients and properties. Corbern et al. (2001) concluded that it is not worth
finding the exact solution for such a system, since even with the exact solution the
properties/coefficients have to be recalculated and the system must be solved again
and again. They proposed that a better strategy would be combine the iterative
calculationofthe solution with the continuous updatingof the coefficients, insuch a
way that both calculations progress together toward the solution of the nonlinear
problem. The proposed numerical scheme fits quite well with this strategy, since it
consistsofasetofexplicitequations.

80

 3.FIN1Dx3MODEL

Fig.3.6Solutionmethodologyforsolvingtheproblem.

ThesolutionmethodologyappliedinthepresentworkissummarizedinFig.3.6.Inthe
firststep,thefluidoutlettemperaturesateachcellareinitializedforbothfluids,with
the corresponding inlet temperature. Any wall cell is initialized using the average
temperatureofthefluidcellsincontactwithit.Sincethetemperaturesatthebottom
andtopofthefinarerequired,thesetemperaturesareinitializedwiththesamevalue
asthoseadoptedattheattachedtubecells.

Thefirststepintheiterativeprocedureisthefluidoutlettemperaturecalculationfor
bothstreams:refrigerantandair.First,Eq.(3.26)isusedtoobtaintherefrigerantfield
temperature. Secondly, for the air flow, the solution of Eq. (3.32) gives the outlet
temperatureofeachaircellalongthefinheightofafin.AlthoughEq.(3.32)represents
a system of three equations, note that the solution for such a system is known, and
can be easily expressed as an explicit equation for each air cell, so that it is not
necessary solving numerically the inverse matrix of Eq. (3.32). Thus, the outlet
temperatures of the three air cells of each fin column (in a segment) are obtained
explicitly.

Oncethefluidtemperaturehasbeenevaluatedateachcell,Eq.(3.25)mustbesolved
forthetubewalltemperature.Notethatatthisstepthetemperaturefieldoffluids,fin
81

3.FIN1Dx3MODEL

bottomandfintoptemperaturesareknown.InpresenceofLHC,Eq.(3.25)showsthat
the wall temperature of a wall cell t depends on the wall temperature of the
neighbouringwallcells.WhenthereisnoLHCtheexactsolution,atthisstep,consists
ofexplicitcalculations.

In case of LHC being present, it is necessary to solve a system of linear equations,

involvingallthetubecelltemperatures.Ifwetakeintoaccountthatitisnotworthto
obtain the exact solution in each step, as it was explained above, the best
methodologiestodealwiththiscalculationare:

Usingforeachtubecellequationvaluesofthetemperatureofneighbouring
tube cells corresponding to the previous iteration. This way converts Eq.
(3.25)intoanexplicitequation,butthismethodwillincreaseconsiderablythe
timetoreachtheconvergence.
LinebyLineiterationmethod(Patankar,1980).Duetothecharacteristicsof
the system, it converts the system of equations into a tridiagonal system of
equations, which is easily solved. When the LHC is only one direction, this
methodologygivestheexactsolution.
BlockbyBlock(Patankar,1980).Itisbasedonthelinebylinemethod,butit
adds a correction to the solution after finishing each iteration. It has the
advantageofafasterconvergencethanthelinebylinemethodbutitneeds
twicethenumberofoperations.

Regardless of the selected methodology, this step calculates the wall tube
temperature field along a set of explicit calculations. Particularly, these calculations
willbetheexactsolutionincasethatnoLHCispresentorwhenLHCispresentonlyin
thetubealongonedirectionandthelinebylinemethodisused.

The last step consists of solving the temperature field for the fin wall cells at the
bottom and top of the fins. For each fin, both temperatures are obtained from Eq.
(3.35).Thissystemhasonlytwoequationsandthesolutioncanbeeasilyexpressedas
twoexplicitequations.Notethatduetonoexistenceoflongitudinalheatconduction
in the fin, along the Z direction, the fin wall temperature field does not need an
iterativeresolutionasitoccurredintheFin2Dmodel,sothattheprocesstoobtainthe
finwalltemperaturefieldresultsexplicit.Aninterestingpointofthefindiscretization
isthatthoughfinwallisdiscretizedintothreecells,computationallyitbehavesasjust
onefincell.Infact,thetotalnumberofunknownvariablesforeachfin,regardlessthe
numberoffincells,aretwo:temperaturesatfinandbottomoffin.

At the end of the iterative process, the residual is compared with the accepted
tolerance. Several variables can be used to evaluate the residual such as wall
82

 3.FIN1Dx3MODEL

temperature, heat exchanged by a fluid flow In this model, the authors used as
residualforeachcellthescaledheatbalance(absolutevalue)inthatcell,applyingthe
maximumoneforallthecellsasthedecisionparameter.

3.2 Model Validation

In order to validate the proposed model, a set of existing experimental results are
going to be compared with the thermal capacity predicted by the model, when inlet
conditions and mass flow rates are provided for both fluids. The model is able to
simulatebothgascoolerandcondenser,thusbothscenariosarevalidated.

Table3.1Correlationsusedinthemodelforcoefficientsevaluation.
Heattransfer Pressuredrop Expansion/Contraction
coefficient pressurelosses
Refrigerant
Onephase Gnielinski(1976) Churchill KaysandLondon(1984)
(1977)
Twophase Cavallinietal. Friedel(1980) KaysandLondon(1984)
(2002)
Air
KimandBullard Kimand KaysandLondon(1984)
(2002) Bullard(2002)

Thegridsizewaschosenastheonethatgaveagoodbalancebetweenaccuracyand
computational cost. According to the definition exposed in subsection 3.1.1, the grid
employedforallthepredictedresultswas:{5,1,3,3}.Theauthorsstudiedinsection2
the effect of simulating the actual number of channels or just one channel with
identicalhydraulicdiameter,andtheyconcludedthatthedifferenceswerenegligible.
Therefore, regardless the actual number of channels per tube, only a hydraulic
equivalentchannelwasmodeled.

Thefinheightratio,hasstillnotbeenevaluated.Thisparametercouldbeadjusted
experimentally, numerically or even by observation. According to the corresponding
explanations in section 3.1, it is possible to get a first approach from: typical
dimensionsoflouveredfinsusedinthistypeofheatexchangers;thevaluereportedin
section2wasabout3%.Thus,avalueofequalto4%wasassumedforthevalidation
and for the different scenarios studied in next section. At end of this section, a
simulationstudywascarriedoutinordertoanalyzetheinfluenceofthisparameteron
thesolution.

83

3.FIN1Dx3MODEL

The different correlations employed to evaluate the heat transfer and pressure drop
coefficientsarelistedinTable3.1.

Table3.2.Geometriccharacteristicsofthecondensersusedforthemodelvalidation(Garca
Cascalesetal.,2010)
Condenser#1
2 2
Facearea(cm ) 1604 Refrigerantsidearea(m ) 1.16
2
Airsidearea(m ) 3.9 Tubesnumber 33
Tubelength(mm) 483 Refrigerantpasses 4
Fintype Louvered Tubedepth(mm) 19
Numberofports 19 Findepth(mm) 21.5
Wallthickness(mm) 0.32 Findensity(fins/in) 14
Hydraulicdiameter(mm) 1.276 Finheight(mm) 8.1
Finthickness(mm) 0.11
Condenser#2
2 2
Facearea(cm ) 5939 Refrigerantsidearea(m ) 3.76
2
Airsidearea(m ) 16.51 Tubesnumber 66
Tubelength(mm) 889 Refrigerantpasses 2
Fintype Louvered Tubedepth(mm) 19
Numberofports 19 Findepth(mm) 21.5
Wallthickness(mm) 0.32 Findensity(fins/in) 14
Hydraulicdiameter(mm) 1.276 Finheight(mm) 8.1
Finthickness(mm) 0.11

3.2.1 Microchannel condenser validation

The experimental data used was measured by GarcaCascales et al. (2010). They
measured two condenser arrangements (onerow and tworow) for two refrigerants
(R410A and R134a), but for the present work only the experimental data
correspondingtotheonerowcondenserisgoingtobeusedforthemodelsvalidation.
Forthisarrangement,twocondensersworkingwithR410Aweretestedandhavebeen
simulated; their main geometry is described in Table 3.2. Differences between both
condensersaremainlythenumberoftubes,finnedlengthandthereforethecapacity.

Usually,predictedcapacityisreportedasthemeanvaluebetweenrefrigerantandair
sidecapacities.However,measuredcapacityforrefrigerantsidehaddifferencesabout
10%withrespecttomeasuredcapacityintheairside.Sinceairishardertomeasure,
theauthorsconsiderthattheactualcapacityshouldbeclosertotherefrigerantside.
Thus, in order to present the models validation, predicted capacity is compared
against measured refrigerant side capacity and measured mean capacity (arithmetic
averageofbothcapacities).

84

 3.FIN1Dx3MODEL

20
+5%

PredictedCapacity[kW]
15 5%

10

RefrigerantCapacity
MeanCapacity
0
0 5 10 15 20
MeasuredCapacity[kW]
Fig.3.7Modelvalidationfortwocondensersbymeansofcomparisonbetweenexperimental
andpredictedcapacity.

Fig.3.7presentsthepredictedcapacityagainstthewholesetofexperimentalvalues,
forbothcondensers.Thefigureshowstwoplots,whichcorrespondtothemeasured
heat capacities mentioned above. The model agrees well with the mean measured
capacitywhileitunderestimatestherefrigerantcapacity,whenanerrorbandof5%is
considered.

Inordertoobtaintheseresultsanadjustmentfactorof1.15wasappliedtotheairside
heattransfercoefficientobtainedwiththecorrespondingcorrelationofTable3.1.The
mainreasontousethisfactorconcernstoanadjustmentofthefactortotherealfin
surface measured whose meaning was introduced in subsection 3.1, but basically it
represents the effects of nonmixed air along Y direction and it depends on the fin
surface and operating conditions. Next subsection will assess the influence of this
parameter on the predicted results. In addition to the influence of factor on the
validationstudy,thecorrelationusedfortheairsideheattransfercoefficient(Kimand
Bullard,2002)reportedrmserrorsof14.5%fortheColburnfactor.Theauthorsfound
asasuitablewaytouseanadjustmentfactorfortheairsideheattransferof15%to
takeintoaccountallthedeviationsexplainedabove.

The predicted pressure losses of refrigerant were quite underestimated with regard
the experimental data, with mean errors of 50% and 75% for condenser #1 and
condenser#2respectively.Microchanneltubegeometryishardtomeasureaccurately
since small manufacturing defects produce large geometry variations, thus pressure
lossesvariations.Adifferentmeanerrorforbothcondensersalsosupportsthisidea.

85

3.FIN1Dx3MODEL

3.2.2 Microchannel gas cooler validation

TheexperimentaldatausedcorrespondstoaCO2gascoolertestedbyYinetal.(2001),
who measured the performance of the gas cooler over a wide range of operating
conditions. The uncertainty for the capacity measurement in those experiments was
5%. The gas cooler modeled is a parallel tube MCHX of three refrigerant passes,
whosemaingeometricdataissummarizedinTable3.3.

Table3.3.Geometriccharacteristicsofgascooler(Yinetal.,2001)
Facearea(cm2) 1950 Refrigerantsidearea(m2) 0.49
Airsidearea(m2) 5.2 Tubesnumber 34
Tubelength(mm) 545 Refrigerantpasses 3
Fintype Louvered Coredepth(mm) 16.5
Numberofports 11 Findensity(fins/in) 22
Wallthickness(mm) 0.43 Portdiameter(mm) 0.79
Finthickness(mm) 0.1 Finheight(mm) 8.89

Fig. 3.8(a) presents the predicted gas cooler capacity against the whole set of
experimental values. As it can be observed, all the predicted values are within the
error bound of 5 %. The accuracy is quite high since a linear function fitted to the
predictedcapacityhadaslopeof0.9972,whatrepresentsanerrorof0.28%,forthe
whole set of experimental data. Overall, the model slightly underpredicts the gas
coolercapacity.Forthisscenario,itwasnotappliedanyadjustmentfactortotheheat
transfercoefficients.

The outlet refrigerant temperature was also compared against experimental data in
Fig.3.8(b).Thefigureincludestheboundsof2Karoundthemeasuredtemperatures.
Asitcanbeobserved,allthepointsdeviatefromtheexperimentaldatalessthan2K.

86

 3.FIN1Dx3MODEL

8
+5%
7

PredictedCapacity[kW]
5%
6

3
Linearregression(0.28%)
a) 2
2 3 4 5 6 7 8
MeasuredCapacity[kW]
60
+ 2K
55
Predictedref.outletT[C]

2 K
50

45

40

35

b) 30
30 35 40 45 50 55 60
Measuredref.outletT[C]
Fig.3.8Modelvalidationforagascoolerbymeansofcomparisonbetweenexperimentaldata
against(a)predictedcapacity;(b)predictedrefrigerantoutlettemperature.

Thepredictedpressurelossesofrefrigerantwerefarfromtheexperimentaldata,with
ameanerrorof80%.TheseerrorsaresimilartothoseerrorsreportedbyAsinarietal.
(2004)andYinetal.(2001)whentheyevaluatedthiserrorwiththeirownmodelsfor
thesamecases.Yinetal.(2001)solvethisdisagreementintroducingsomedimensional
changes in ports, produced by manufacturing defects. Asinari et al. (2004)
demonstrate that introducing arbitrary multiplying factors pressure losses agree well
with a negligible effect on the capacity results. They argue that the reason for the
disagreement between predicted and experimental pressure drop is based on
underestimationofthepressurelosseswhentraditionalcorrelationsareusedforthis
particulargeometryandconditions.

87

3.FIN1Dx3MODEL

Animportantfactortotakeintoaccountinamodelvalidationistheeffectivenessof
the gas cooler employed for obtaining the experimental data. If we define the
temperature approach as the temperature difference between the refrigerant outlet
andtheairinlet,thisfactorwillbequiterepresentativeofthegascoolereffectiveness.
Arobustvalidationofamodelwillimplylargeapproachvalues.Fortheexperimental
dataused,theapproachrangedfrom1Kto7K,withanaveragevalueof4.1K.This
value indicates a high gas cooler effectiveness, in fact, for the data used it had an
averagevalueof83%.

3.2.3 Impact of parameter on predicted capacity

Themeaningofthefinheightratiointhemodelhasbeendiscussedinsection3.1
whilst its estimation has been discussed in subsection 3.2. A value of 4% was finally
proposedforthevalidationcases,eventhoughthisvaluedependsonthefinsurface
andoperatingconditions.Tothisend,authorscarriedoutanumericalstudyaboutthe
impact of on the model results in order to analyze the model sensitivity to this
parameter,whatimplicitlymeanstheimpactofunmixedairflowalongYdirection.

Thescenariochosenforthisstudycorrespondstothegascoolerintroducedinthegas
coolervalidationworkingintheoperatingconditionsoftest#2(Yinetal.,2001),which
hasthelargestcapacityofallthetests.

0
Heatcapacitydeviation[%]

2
22fpi
3 14fpi

5
0 10 20 30 40 50
[%]
Fig.3.9Influenceoffinheightratioonheatdutyofgascoolerfordifferentfinsdensities.

Fig.3.9showsthedeviationbetweentheheatexchangedbythegascoolerwhenthe
parameterismodifiedwithrespecttotheheatexchangedwhentendstobe50%.
This study is carried out for two fins densities: 22 fpi and 14 fpi. The variation of
rangesbetweentwolimitsituations:0%and50%.Avaluecloseto50%meansthatair

88

 3.FIN1Dx3MODEL

flowhasatemperatureprofilecorrespondingtoamixedairflowalongthefinheight.
Infact,forthisvalue,iftemperaturesofattachedtubestothefinareidentical,air
flow has uniform temperature along the whole fin height. The opposite behaviour is
whenthetendstobe0%whichmeansthatthenonmixedairalongYdirectionhas
themaximumeffect.Inthissituation,athinairlayerisincontactwitheachwalltube.
These air layers have a quite different temperature from the rest of air because air
flowrateinthesecellsisverysmallanditstemperaturewillbeveryclosetothetube
walltemperature.Therefore,theseairlayerswillhaveaninsulatingeffectonthetube
wallincontact.Bothsituationsarenotrealandwhatactuallyhappensisbetweenboth
ofthem.

The effects described above for extreme values of valueagree quite wellwith the
trendshowninFig.3.9;thelowerthevalue,thelesscapacityisexchangedbythe
gas cooler. It is interesting notice that the trend is asymptotic for large values
whereas the effect of changes sharplyfor low values. Louvered fins would have
low values while it would be much greater for plain fins. Fig. 3.9 also shows that
impactofonthesolutiondependsonthefindensity,resultingvaluesfrom2%to5%
forfindensitiesof22fpiand14fpirespectively.

Thisparameterisactuallyunknown,butfortunatelythedeviationinpredictedcapacity
wouldbelessthan5%,forthefinsdensitiesevaluated.

3.3 Simplification of Fin1Dx3 model: Fin1D

TheobjectiveofthethesisworkistodevelopadetailedMCHXmodelconsideringnot
onlyaccuracyreasonsbutalsotocomputationalcost.Tothisend,reductionofnumber
ofcellsplaysanimportantrole.Thus,authorsdecidedtosimplifytheFin1Dx3modelin
ordertoassessinfurtherstudiesthepotentialofthissimplificatedmodel.

ThesimplificationconsistsindiscretizingairandfinonlyinonecellalongYdirection.
Fig.3.10showsanexampleofthisdiscretization.

GoverningequationsandnumericalschemeforFin1Dmodelarenotpresentedinthe
sectionbecausetheyareobtainedinsamewayandwithsameassumptionsasmadein
section 3.1. Instead of the piecewise function proposed for Fin1Dx3, now only one
continuousfunctionforthefinwalltemperatureisused.Thisfunctionisobtainedby
imposingsameboundaryconditionsinthefin:temperaturesatfinrootsaregiven.In
the air flow, now there is only a temperature variable for one airfin column.

89

3.FIN1Dx3MODEL

Integrationoftheairtofin heattransferisobtained inthesamewayasforFin1Dx3

butnowismuchsimpler.

Tube

TfT
Tf(Y),
Hf

Tube TfB

Y Y

Z X

Fig.3.10DifferentviewsofportionofheatexchangerdiscretizedbyFin1Dmodelwhichuses
onlyonecellalongYdirectionforfinandair.

PhenomenathatFin1DmodeltakesintoaccountarethesameasforFin1Dx3,withjust
oneexception:Fin1Dcannottakesintoaccounttheeffectontheheattransferofthe
unmixedairYdirection.

ThevalidationforthisFin1Dmodelisnotcarriedoutsinceanaccuracystudyagainst
Fin1Dx3modelisworkedoutinthefollowingchapter.

3.4 Conclusions
This chapter presents the Fin1Dx3 model which reduces significantly the
computational cost of the Fin2D model retaining its accuracy. In this way, it allows
using that model to analyze microchannel condensers and gas coolers with any
refrigerantcircuitry,includingserpentineheatexchangers.

The main conclusion is that it is possible to take into account the heat conduction
betweentubesinamorefundamentalwaythanotherfinefficiencybasedapproaches,
whichhavetoapplyheatconductiontermstoanapproachthatusestheadiabaticfin
tip assumption, which is not satisfied in such cases. The alternative methodology,
proposed in this work, consists in evaluating the heat transfer by integration of the
corresponding fin temperature profile instead of using a fin efficiency which cannot
alwaysbedefined,e.g.whentemperatureatfinrootsaredifferent.Ithasbeenshown
that this integration does not represent an obstacle since it can be easily discretized
consistentlywiththerestofgoverningequations;thereforethereisneitheraccuracy
nor computational cost reason to apply adiabaticfintip assumption when it is not
90

 3.FIN1Dx3MODEL

satisfied. The approach proposed in this paper is developed for a three airfin cells
discretizationbutthisconclusionisequallyapplicabletoasingleairfindiscretization.

The rest of conclusions regarding the Fin1Dx3 model and its validation that can be
drawnfromthissectionarethefollowing:

TheFin1Dx3modelaccountsforallthesameeffectsthanFin2Dmodelexcept
theLHCinthefinalongZdirectionwhich,inanycase,hasbeenfoundnotto
beimportant.Fin1Dx3isbasedonanoveldiscretizationmethodologyforthe
air and fin wall that only needs three air cells along the Y direction. This
methodologyallowsreducingdrasticallythenumberofcellstocompute,with
regardtotheFin2Dmodel,andconsequentlythesimulationtimebutkeeping
ahighresolutionalongYdirection.
ThelargenumberoffincellsneededbyFin2Dmodeltosolveaccuratelythe
airsideheattransfer,iscompensatedinFin1Dx3withanovelmethodologyto
describe the airside heat transfer, using a piecewise function for the fin
temperature. This piecewise function, together with the employed air
discretization, allows applying in a more fundamental way the analytical
solutiongivenbythefintheory.
ThemaincapabilitiesofFin1Dx3are:2DLHCinthetubewall;nonmixedair
effects due to temperature difference between bulk air and the air close to
thetubes;anditaccountsfundamentallyforheatconductionbetweentubes
sinceitdoesnotapplyadiabaticfintipassumption.
The equations have been discretized, with the interesting characteristic of
resultinginasystemofpseudolinearequationswithrespecttothevariables
oftheproblem.Anumericalschemehasbeenproposedtosolvetheproblem
asaseriesofexplicitsteps.Thenumericalschemeproposedallowscomputing
thethreefincellswiththecomputationaleffortofjustonefincell.
Fin1Dx3modelwasvalidatedwithexperimentaldata,forbothcondenserand
gas cooler. The predicted capacity is within 5% error, being much more
accurate for the gas cooler scenario. Although pressure drop was
underpredicted,itdidnotaffecttheheattransferresults.
Thestudyabouttheinfluenceofthefactor,whichaccountsfortheeffects
of unmixed air flow along Y direction, showed deviations less than 5% for
extremevaluesofandforthesimulatedconditions.

Finally,asimplificationofFinDx3hasbeenpresented:Fin1D.Itreducesthenumberof
cells employed in the discretization at the expense of accuracy. In next chapter it is
comparedagainstFin1Dx3modelregardingtoaccuracyandcomputationalcost.

91

3.FIN1Dx3MODEL

92

CHAPTER4
NUMERICALCOMPARISONOF
MODELS

93

4.NUMERICALCOMPARISONOFMODELS

4 NUMERICAL COMPARISON OF MODELS

This section will discuss and compare accuracy skills and computation time for two
groupsofmodelsforMCHX.

FirstgroupofmodelstocomparearethosedevelopedbypresentauthorsforMCHX
modeling:Fin2D,Fin1Dx3andFin1D.ThegoalofFin1Dx3modelistobeasuitabletool
for simulation and design of MCHXs. It has been explained that the Fin1Dx3s
discretization is based on the Fin2D model, which had large computational
requirements, but the new discretization of Fin1Dx3 allows, for same accuracy,
reducing considerably the number of both air and fin cells, so that a large
computationalcostreductionisexpected.Therefore,thefirstresultspresentedinthis
sectionareorientedtoassessthedegreeofaccomplishmentgottothisend.Another
waytoreducethecomputationalcostistoreducethenumberofcellsemployedinthe
discretization, which a priori means accuracy degradation. To this end authors
developed a model referred to as Fin1Dwhich applies same assumptions as Fin1Dx3
modelbutitdiscretizesthewholefinandaircolumnofeachsegmentintojustonecell
alongthefinheightdirection.

In Introduction, it was explained that most of models available in literature do not

account for heat conduction between tubes. A model that uses the adiabaticfintip
withoutanycorrectionterm,totakeintoaccounttheheatconductionbetweentube,
is always predicting results as if the heat exchanger had all fins cut, hence these
models are always overpredicting the heat transfer. Only few authors (Lee and
Domanski,1997;andSinghetal.,2008)model,forfinnedtubeheatexchangers,this
phenomenon by using a correction term which takes into account, in a more or less
artificial way, the heat conduction between tubes, despite using the adiabaticfintip
assumptioninthegoverningequations.Thus,thesecondgroupofmodelscomparedin
thissectionconsistsofthemodelsproposedinthispaper(Fin1Dx3andFin1D)anda
modelwhichrepresentswhatotherauthors(LeeandDomanski,1997;andSinghetal.,
2008)dointheirmodelsinordertotakeintoaccounttheheatconductionbetween
tubesinaMCHX.Althoughtheseapproacheswereoriginallyproposedforfinnedtubes
heatexchangers,inthischaptertheyhavebeenadaptedtoMCHXs.

The accuracy comparison between the different models has been worked out by
comparingthemodelswithregardtothemostdetailedmodel,whichdependsonthe
scenariostudied.Experimentalresultshavenotbeenusedasreferencetoperformthe
accuracycomparisonsincethedeviationbetweenresultsisaffectedbyseveralfactors
hard to identify such as experimental uncertainty, moreover this deviation could be

94

 4.NUMERICALCOMPARISONOFMODELS

nonlinear adding a complex factor in order to draw conclusions about the models
comparison.

4.1 Comparison among different developed models: Fin2D, Fin1Dx3

and Fin1D
Below are listed and briefly summarized each of the models compared in this
subsection:

Fin2D:Itcorrespondstothemodelpresentedinchapter2.Itisaverydetailed
model which discretizes fin and air into a twodimensional grid. Its main
capabilitiesare:ittakesintoaccount2Dlongitudinalheatconduction(LHC)in
bothfinandtubewall,itdoesnotappliesfintheoryanditaccountsforheat
conduction between tubes. Its main drawback is the simulation time
employedtosolveacase,duetothedetailedgridadoptedinthefinandair
elements.
Fin1Dx3: It corresponds to the model presented in chapter 3. For each
segmentitdiscretizesairandfinintothreecellsalongthedirectionbetween
tubes, while for the tube wall it applies the same discretization as Fin2D
model.ThephenomenamodeledarethesameasFin2Dmodelexceptingthe
LHCinthefinalongairflowdirectionwithalargereductioninthenumberof
cellsemployed.
Fin1D: It is basically the same model as Fin1Dx3, and same phenomena
modeled,butnowneitherairnorfinsarediscretizedalongthefinheight.The
analyticalsolutiongivenbyfintheory,forthecaseofgiventemperaturesat
fin roots, is used to get the fin temperature profile. Thus, it also takes into
accountheatconductionbetweentubes.

Thegeometryofthetubesandfins,ofthecasestudyforthissection,arethesameas
usedbyYinetal.(2001)(Table3.3).Theoperatingconditionsforthesimulationsare
those used for the tests n: 9, 17, 25, 33 and 41 (Yin et al., 2001). All the models
appliedthesamegrid,withtheexceptionofthefinandaircellsalongYdirection.Due
tothemodeldifferences,Fin2Dmodelneedsalargenumberofthesecells;inchapter
2itwasproposedusing30cellsintheYdirection.Thegridsappliedforthesescenarios
are: {5,1,3,30,3} for Fin2D model and {5,1,3,3} for Fin1Dx3 and Fin1D models.
Regardingthecorrelationsusedbythemodel,theyarelistedintheTable3.1.

TheresultsfortheaccuracycomparisonarepresentedinFig.4.1.Thefigureshowsthe
deviation on predicted capacity for models Fin1D and Fin1Dx3 with respect the

95

4.NUMERICALCOMPARISONOFMODELS

predictedresultsofFin2Dmodel.Therefore,thezeroforordinatesaxiscorrespondsto
the predicted results with Fin2D model. Fin2D model has been chosen as reference
becauseitisthemostaccurateofthemsinceitappliesthefinestdiscretizationtothe
heatexchanger.

2
1.8 Fin1Dx3 Fin1D
1.6
1.4
Deviation[%]

1.2
1
0.8
0.6
0.4
0.2
0
1 9 17 25 33 41
Test#
Fig.4.1Heattransferdeviation,fordifferenttestconditions,ofFin1Dx3andFin1Dmodelswith
regardtoFin2Dmodel.

Fig.4.1showsthatthedeviationbetweenFin2DandFin1Dx3modelisatmost0.2%,
whichmeansthatpredictedresultscouldbeconsideredasthesameones.However,
this deviation turns to be as much 2% in the case of Fin1D model. The negligible
difference between Fin2D and Fin1Dx3 models means that longitudinal heat
conductioninfinsurfacealongairdirection,whichisnotmodeledinFin1Dx3model,
can be neglected for this scenario. These results also confirms that the approach of
usingthreefin/aircellswithapiecewisefunctionforthefintemperatureprofilegives
agoodsolutionwithmuchlowercomputationalcost.

Fin1DandFin1Dx3takeintoaccountsamephenomenaanddifferencesbetweenthem
are only due to the fin/air discretization. According to this, the deviation between
predictedresultsofbothmodelsisconsequenceonlyof amoreaccurateapplication
by Fin1Dx3 model of fin theory for the airtofin heat transfer evaluation. In other
words, this difference could be interpreted as the effect of nonmixed air along Y
direction. Nevertheless, this deviation can be interpreted as small, though the effect
would depend on the operating conditions, heat exchanger and application. The
present work, analyses the case of a gas cooler that corresponds to a case with an
expectedimpactofthesephenomenalargerthanforthecaseofacondenser.Foran
evaporator,dehumidificationappearsandplaysanimportantroleandwhathappens
dependsstronglyonlocalproperties,thusauthorsforeseetoincludedehumidification
infutureworks.
96

 4.NUMERICALCOMPARISONOFMODELS

Withregardtothecomputationalcost,Fig.4.2presentsthesimulationtimeemployed
byeachmodeltosolveseveralcaseswhichweredescribedabove.Inthefigure,alarge
computing time reduction, from Fin2D model to Fin1Dx3 model, is noticeable. This
reductionrepresentsoneorderofmagnitude.Themainreasonisthelargedifference
of air and fin cells used by both models. In the case of Fin1Dx3 model, a piecewise
function which consists of three onedimensional functions is enough to capture
accuratelytheactualfintemperatureprofileandconsequentlytheheattransferfrom
fin to air. However, as explained in the introduction, Fin2D needs to apply a large
discretization to the fin height, in practice 30 fin and air cells are required to get
accurateresults.

1000

100
Time[s]

10

1
Fig.4.2Comparisonofthesimulationtimeemployedbyeachmodel.

The simulation time reduction from Fin1D to Fin1Dx3 model is not as drastic as for
Fin2D case, Fin1D needs half the timespent by Fin1Dx3. A priori, a largersimulation
timereductioncouldbeexpectedsinceFin1Dusesjustoneairandfincellalongthefin
heightdirectioninsteadofthreeairandfincellsasFin1Dx3modeldoes.Howeveran
interesting fact of the piecewise function applied in the Fin1Dx3 model is the
following;thepiecewisefunctionusesasunknownvariablesthetemperaturesofthe
threeaircellsandthefinroots.TheFin1Dmodelalsoincludesasunknownvariables
thefinrootstemperatures,sinceittakesintoaccountheatconductionbetweentubes,
but only one air temperature value. Thus only two variables are added to the Fin1D
model with regard to Fin1Dx3 model, which corresponds to the air temperature
values.ThesetemperaturevaluesareobtainedinthesamewayasFin1Dmodeldoes,
i.e. with an explicit calculation given the wall temperature field, so that in Fin1Dx3
modelthereareonlytwomoreexplicitcalculations.Inotherwords,theonlycellsthat
addcomputationalcosttotheFin1Dx3modelsaretheaircellswhilstthethreefincells
behavenumericallyasjustone.

97

4.NUMERICALCOMPARISONOFMODELS

Ifbothfactorsaccuracyandcomputationalcostaretakenintoaccount,Fin2Dmodelis
notacosteffectivesolutionsinceFin1Dx3providesthesameresultswithasimulation
time reduction of one order of magnitude. In contrast, Fin1Dx3 offers better results
thanFin1Dwithonlydoublesimulationtime,thusFin1Dx3modelisconsideredbythe
authorsasthebestoptionformodelingthiskindofheatexchangers.Despitethefact
thatforacondenseroragascoolertheFin1Dcouldbe aninterestingoption,foran
evaporatortheaccuracydifferencebetweenbothapproachesisexpectedtobelarger
thanforthepresentedresultsduetolocaleffectsondehumidification.

4.2 Comparison with other authors approaches

Thissubsectioncompares,inasimilarwayastheprevioussubsectiondid,themodels
proposedinthepaper(Fin1DandFin1Dx3)againstotherapproachesusedinliterature
forheatexchangermodelling.Tothisend,ithasbeennecessarytodeveloptwonew
models:

Fin1D_Cut:Itreproducestheresultsofthemostcommonmodelsavailablein
literature (Yin et al., 2001; Corbern et al., 2002; Jiang et al., 2006; Garca
Cascales et al., 2010; Fronk and Garimella, 2011). It applies a segmentby
segmentdiscretization,usestheadiabaticfintipassumptionanditdoesnot
takeintoaccountheatconductionbetweentubes.Themodelisjustthesame
asFin1Dbutitincludesacutalongthefintoreproducealwaysadiabaticfin
tipassumption.The neededchangesinthemodeltoincludethisfincutare
the same as those explained in previous chapter, when the Fin1Dx3 model
wasmodifiedtosimulateaMCHXwithfincuts.
CorrectedFin:ItisbasedontheapproachesproposedbySinghetal.(2008),
and Lee and Domanski (1997). They have been chosen as references since
they account for heat conduction between tubes in a different way to the
proposed in the present paper, though these approaches model that
phenomenoninamoreartificialway.Thismodeltriestoberepresentativeof
whatreferredauthorsmodelsdo.ItisbasedonFin1Dmodelanditapplies
the same discretization but now it uses the analytical solution given by fin
theory when adiabaticfintip is assumed. In order to account for the heat
conduction in the same way as referred authors do, correction terms are
included in the corresponding energy conservation equations, which will be
describedindetailbelow.

The approaches of Singh et al. (2008), and Lee and Domanski (1997) were originally
developedforfinandtubeheatexchangersbuttheyhavebeenadaptedinthispaper
98

 4.NUMERICALCOMPARISONOFMODELS

foraMCHX.Fig.4.3showsthegeometricparametersofbotharrangementsregarding
theheatconductionphenomenon.

Tt

Tt Tt*
Lt,t*

Lt,t* W

Tt*

Fig.4.3AnalogybetweenafinnedtubeandaMCHXfortheheatconductionresistance
evaluationbetweentwoneighborstubesalongthefin.

Theseapproaches(Singhetal.,2008;LeeandDomanski,1997)applythefintheoryto
eachvolumecontrolanduseafinefficiencytoincludethefintoairheattransferthat
isevaluatedwithEq.(4.1),where fB,a isthetemperaturedifferencebetweenthebulk
air temperature and the corresponding fin root temperature, and f is the fin
efficiency.Therelationshipusedforevaluationofthefinefficiencycorrespondstothe
caseofadiabaticfintipassumption,Eq.(4.2).

Q f ,a f f ,a Af ,a fB,a (4.1)

tanh(m f ,a H f / 2)
f (4.2)
m f ,a H f / 2

f ,a pw f ,a
m 2f ,a
k f Af

Eq.(4.3)establishestheenergyconservationinasegment.Thesegmentconsistsofthe
walltubecellt,thecorrespondingfinwallcellfattachedtothetube,andthefluidsin
contactwithit:refrigerantcellrandaircella.

Q f ,a Q t ,a Q t ,r Q t ,t* 0 (4.3)

99

4.NUMERICALCOMPARISONOFMODELS

Sincetubeshavedifferenttemperatures,thecorrectionterm Q t ,t* isintroducedinEq.

(4.3) in order to take into account the heat conduction between tubes, which
correspondstothetotalheattransferbyconductionbetweenneighboringtubes.Fig.
4.3shows4tubest*connectedtoacentraltubetbythefinsurface.Forthisexample
thetotalheatconductionbetweencentralandneighboringtubescanbemodeledas
Eq.(4.4).

T T
Q t ,t* t t* (4.4)
t* Rt ,t*

Different approaches could be applied to get the value of thermal resistance Rt,t*
together the use of ,which is a multiplier that can be used to adjust the heat
conduction term. Singh et al. (2008) explain that this multiplier has to be adjusted
eithernumericallyorexperimentallywhichontheheatexchangersimulated.Theneed
to use this correction factor which a priory is unknown and its dependency on the
modeledcasearethemaindrawbacksofthismethodology.

CorrectedFinmodelevaluatesRt,t*withEq.(4.5)andapplies=1.

Lt ,t* (4.5)
Rt ,t*
tf W kf

Thesimulationswerecarriedoutforthegascooler(Yinetal.,2001)thatwasvalidated
in previous chapter. The operating conditions for the simulations are those used for
thetestsn:9,17,25,33and41byYinetal.(2001).Thecorrelationsforheattransfer
and pressure losses coefficients were also the same as described in previous
subsection.

Allthecasesanalyzedhavetubeswithdifferenttemperaturesandheatconductionis
present,thereforeFin1DwillbemoreaccuratethanFin1D_Cutsinceadiabaticfintip
assumptionisnotvalid.Fin1DmodelshouldbealsomoreaccuratethanCorrectedFin
because the latter applies a correction term to take into account heat conduction
betweentubes,whileFin1Dtakesintoaccountimplicitlytheheatconductionwithout
simplifyingassumptions.

First study compares the models that apply same discretization level, i.e. fin is
discretized just in one cell. Fig. 4.4 shows the deviation on predicted capacity for
models Fin1D_Cut and CorrectedFin with regard to the predicted results of Fin1D
model,whichisexpectedtobethemostaccurateofthem.First,itisnoticeablethat
deviations between these models for these conditions are quite small, what means
100

 4.NUMERICALCOMPARISONOFMODELS

thatadiabaticfintipassumption,despitenotbeingvalid,doesnothavealargeimpact
onthesolutionwhichislessthan0.8%.ThedeviationbetweenFin1D_CutandFin1Dis
alwayspositivewhatimpliesthatbycuttingthefins,heattransferisalwaysincreased.
As can be observed in Fig. 4.4, CorrectedFin can take into account heat conduction
betweentubeswithnegligibledeviations,whatmeansthattheapproachesofSinghet
al. (2008) and Lee and Domanski (1997) are a good alternative for modeling finned
tubesheatexchangersinpresenceofheatconductionbetweenneighboringtubes.

0.8 CorrectedFin
0.7 Fin1D_Cut
0.6
Deviation[%]

0.5
0.4
0.3
0.2
0.1
0
0.1
0.2
1 9 17 25 33 41
Test#
Fig.4.4DeviationofpredictedheattransferofmodelsFin1D_CutandCorrectedFinwith
regardtoFin1Dfordifferenttestconditions.

Thefollowingstudycomparesthemodelswhichapplysamediscretizationlevel(Fin1D,
Fin1D_Cut and CorrectedFin) with Fin1Dx3 model that applies a more detailed
discretization,resultinginthemostaccurateofthem.Fig.4.5presentsthedeviationin
capacityformodelsFin1D,Fin1D_CutandCorrectedFinwithregardtothepredicted
results of Fin1Dx3 model. Therefore, the zero for ordinates axis corresponds to the
predictedresultswithFin1Dx3model.

AscanbeobservedinFig.4.5theaccuracyofallthemodelsisgood,resultingerrorsas
muchof2%withrespecttoFin1Dx3model.Thelargestdeviationisproducedbythe
Fin1D_Cut model which uses the adiabaticfintip. Fin1D and CorrectedFin models
haveasimilardeviationrangingfrom1%to2%.Thisdeviationindicatesthatthemajor
impactinthepredictionerrorisconsideringtheairasmixedinthedirectionbetween
tubes;infact,thisistheonlydifferencebetweenFin1DandFin1Dx3models;Fin1Dx3
discretizesthefinheightinto3cellsbeingabletoaccountfornonmixedairalongfin
height.

101

4.NUMERICALCOMPARISONOFMODELS

2.5
Fin1D CorrectedFin Fin1D_Cut
2

Deviation[%] 1.5

0.5

0
1 9 17 25 33 41
Test#
Fig.4.5DeviationofpredictedheattransferofmodelsFin1D,Fin1D_CutandCorrectedFin
withregardtoFin1Dx3fordifferenttestconditions.

Regarding computational cost, it is not necessary to perform an evaluation of the

simulationtimerequiredbyFin1D_CutandCorrectedFinmodelsbecausetheyapply
same discretization as Fin1D and therefore they have the same computational cost,
whichwaspresentedabove(Fig.4.2).

4.3 Conclusions
Numericalstudiesaboutaccuracyandcomputationalcostwerepresentedinorderto
compare the different developed models: Fin2D, Fin1Dx3 and Fin1D, and other
representative models from literature. The main conclusions of these comparisons
werethefollowing:

ThesolvingtimeofFin1Dx3hasbeenreducedoneorderofmagnitudewith
regardtothetimeemployedbyFin2D,whereasthedifferencesontheresults
are less than 0.3%, which are considered as negligible for practical
applications.ThesolvingtimedifferencebetweenFin1Dx3andFin1Dturned
outtobetwice.
CorrectedFincanleadtoaccurateresultswhencomparedwithanequivalent
approach model that models heat conduction between tubes in a more
fundamentalwaylikeFin1Dx3does.Differencebetweenpredictedresultsof
both models was between 1% and 2%. Computational cost of Fin1D and
CorrectedFin models are the same. Nevertheless, author would like to
emphasizethefactthatthepresentworkshowsnocomputationalsavingor
accuracy increase by adding correction terms to an approach that uses the

102

 4.NUMERICALCOMPARISONOFMODELS

adiabaticfintip efficiency with regard to a more fundamental approach,

regardingtheheatconductionbetweentubesphenomena,likeFin1D.
By comparison of deviations between Fin1D, CorrectedFin and Fin1Dx3
models,itwasconcludedthatthemainresponsiblefordifferencesbetween
themweretheeffectsofnonmixedairalongfinheight.

Accordingtopreviousconclusions,thefinaldecisionofauthorsaboutwhichmodelto
use in the following for modeling MCHXs, it corresponds to the Fin1Dx3 model with
continuousfin.

103

4.NUMERICALCOMPARISONOFMODELS

104

CHAPTER5
SIMULATIONSTUDIES

105

5.SIMULATIONSTUDIES

5 SIMULATION STUDIES
Inpreviouschapter,Fin1Dx3modelhasbeenchosenasthebestmodelwithregardto
computational cost and accuracy reasons. This chapter uses Fin1Dx3 model as
simulation tool to study the impact of some design parameters of a MCHX on its
performance. The following proposed numerical studies are hardly available in the
literatureforaMCHX.Thegoaloftheselectedcasestudiesiscontributingtoabetter
understandingoftheinfluenceofsomedesignparametersonMCHXperformance.

AlongthedesignprocessofaMCHX,thefirststepischoosingthegeometricdataof
tubesandfinssuchasminortubedimension,majortubedimension,finheightandfin
depth. This choice is based on manufacturing requirements, e.g. costs, tooling and
volumeproduction.Giventheinletconditionsandmassflowratesforbothrefrigerant
andair,heattransferareacanbegenerallyfixedbyimposingatargetheatloadwhile
faceareaoftheMCHXisobtainedfromapressuredropcriteria.Finpitchwouldhave
tosatisfyacompromisebetweenheattransferandpressuredrop.

Once these areas have been chosen, there are multiple circuitry designs that satisfy
the target heat load so that the refrigerant circuitry can be designed in order to
optimize the heat exchanger effectiveness by maximizing heat transfer, with some
restrictionsregardingpressuredrop.Inthesameway,parametersliketheaspectratio
(L/H)playthesameroleasthecircuitry:severalaspectratiossatisfytheperformance
requirements but just one optimizes the effectiveness. In fact, some simulation
softwarelikeEVAPCOND(2010) hasthecapabilitytooptimizetheheatloadvarying
thecircuitryofafinnedtubeheatexchanger.Shaoetal.(2009)studiedtheeffectof
the number of refrigerant passes for a serpentine MCHXworking as condenser, with
thesamefaceareaandheattransferarea.Theauthorsobtainedupto30%differences
on heat load only by changing the number of refrigerant passes. Given that the
circuitry has an important influence on the heat exchanger performance, the
usefulness of simulation software for this purpose is clearly justified, since the
optimizationviaexperimentationwouldtaketoolong,itisdifficultandexpensive.

FincutsareanotherpossibleimprovementtointroduceinaMCHXdesign.Obviously,
the improvement on the heat transfer will be null for a singlepass heat exchanger
sinceallthetubeshavethesametemperatureandtheheatconductionbetweenthem
wouldbezero.Thus,itisworthassessingtheimprovementduetofincutsinaMCHX
fordifferentnumberofrefrigerantpasses.

On the other hand, depending on the models assumptions some parameter can be
studiedornot,e.g.theimpactoftheaspectratio(L/H)ontheheattransferofaheat
106

 5.SIMULATIONSTUDIES

exchangerwouldbenullifitisevaluatedwithamodelwhichappliestheadiabaticfin
efficiency. This design parameter can only be assessed ifmodel adequately accounts
fortheheatconductionbetweentubes.

According to the ideas previously exposed, the authors considered to study some
designparametersofaMCHXsuchas:aspectratioandnumberofrefrigerantpasses.
Theinfluenceoffincutswasalsostudiedfordifferentrefrigerantcircuitry.Theimpact
ofalltheseparametersdependsstronglyontheheatconductionbetweentubes,LHC
andairsideheattransfer.Hencetheneedofusingamodelwhichtakesintoaccount
accuratelyallpreviousphenomena,otherwiseitwouldnotbepossibleevaluateeffects
ofsomeofmentionedparametersonMCHXperformance.Tothisend,thesimulation
studies were carried out with the previously proposed model Fin1Dx3. The Fin1Dx3
modeltakesintoaccountalltheseeffectsanditcansimulateanyrefrigerantcircuitry
regarding the number of refrigerant passes, tubes number and tube connections. In
addition,themodelhastheoptiontoworkintwodifferentmodes:continuousfinor
cutfin.Thereasonforthesetwomodesistobeabletoevaluatetheimprovementsby
cuttingthefinsontheheattransfer.

ThemoresensitivethecasestudyistoLHCandtoheatconductionbetweentubes,the
larger the impact will be on the performance due to variations of the defined
parameters. Impact of LHC and heat conduction between tubes will increase as the
temperature gradient on a tube and temperature difference between tubes become
larger. That is the reason why a microchannel gas cooler working with CO2 in
transcriticalpressureshasbeenchosenascasestudy.Reasonswereexplainedinthe
Introductionchapter.

5.1 Simulation methodology and case study description

The MCHX chosen for these studies corresponds to a gas cooler, according to the
reasons exposed in the introduction. The gas cooler geometry is based on the gas
coolertestedbyYinetal.(2001),whichcorrespondstoamicrochannelgascoolerused
inautomotiveapplicationswithCO2asworkingfluidintranscriticalconditions.Thisgas
coolerconsistsof34tubeswith3refrigerantpasses.Thenumberofrefrigerantpasses
is a parameter to be studied, from one pass up to the limit that corresponds to a
serpentine gas cooler, i.e. refrigerant passes equals the tubes number, without
changingtherestofgascoolerdimensionsandinletconditions.Increasingnumberof
refrigerant passes leads to larger velocities of the refrigerant flow. This fact, besides
theincreaseofrefrigerantpathlengthproducesmuchlargerpressuredrop.Thelimit
case(serpentineMCHX)wouldbe,forthisreason,ofnopracticaluse.
107

5.SIMULATIONSTUDIES

Table5.1.Geometriccharacteristicsofgascoolerforsimulationstudies.
Facearea(cm2) 242.5 Refrigerantsidearea(cm2) 609
Airsidearea(cm2) 6465 Tubesnumberoftubes 12
Tubelength(mm) 192 Coredepth(mm) 16.5
Fintype Louvered Findensity(fins/in) 22
Numberofports 11 Portdiameter(mm) 0.79
Wallthickness(mm) 0.43 Finheight(mm) 8.89
Finthickness(mm) 0.1

Thetotalnumberoftubesandsomegeometricdimensionsofthegascoolertestedby
Yinetal.(2001)(Table3.3)havebeenmodifiedsothatthechangeinthenumberof
refrigerantpasseswillnotproduceexcessivepressurelossesfortheserpentinecase.
Thetotalnumberoftubeswasreducedto12andtherestofdimensionssuchasgas
coolerwidthandheightwereobtainedrescalingtheoriginalonesproportionallytothe
tubes number. The resulting geometric data is shown in Table 5.1. The rest of
geometricdataforfinsandtubeswerethesameasYinetal.(2001)tested(Table3.3).

Table 5.2. Operating conditions for simulation studies: based on test n 2 (Yin et al.,
2001).
InletPressure Pressuredrop Inlet Outlet Massflow
(kPa) (kPa) temperature temperature rate(g/s)
(C) (C)
CO2 10792 421.6 138.6 48.2 5.64
Air 100 61103 43.5 87.3

Forallscenariostherefrigerantandairsideareas,faceareaandrestofthegeometry
arethesame.Inletconditionsforbothfluidsinthegascooleraregoingtobeidentical
forallsimulationstudies.Regardingtheoperatingconditions,thosecorrespondingto
testn2fromYinetal.(2001)havebeenchosen.Boththemassflowrateandairflow
ratehavebeenmodifiedinordertogetthesamemassvelocitiesastheoriginalvalues
according to the new geometry. The operating conditions are listed in Table 5.2.
Regardingtheair,therearetwoscenarios:withthemassflowrategiveninTable5.2,
andwithamassflowratethreetimestheindicatedinTable5.2.

Regardingthecorrelationsusedbythemodel,theyarelistedintheTable3.1.

108

 5.SIMULATIONSTUDIES

5.2 Number of refrigerant passes

Thenumberofrefrigerantpassesisvariedfromonepassuptothemaximumpossible
number, i.e. 12 passes which corresponds to a serpentine configuration. Fig. 5.1
depicts two samples of the cases studied. The performance differences will be only
duetothenumberofpassessincerefrigerantarea,airsidearea,faceareaandrestof
thegeometrydonotchange.

IN
IN
a
c

H
b

OUT OUT

3 passes 12passes

Fig.5.1Schematicsoftwogascoolerarrangementsstudied:3and12refrigerantpasses.

Fig.5.2showstheresultsofthisstudyfortwodifferentvaluesoftheairvelocity.As
theairvelocityisincreased,theheattransferisalsoincreasedforallcasesdueto:the
mass flow rate rises since the air velocity is increased with the same face area; the
overall heat transfer coefficient increases because the greater the air velocity the
larger the air side heattransfer coefficient. When the number of passes is increased
the total refrigerant crosssectional area is reduced so that the refrigerant velocity
risestokeepconstantthemassflowrate,anditimprovestheheattransfercoefficient.
Thus, for this case study, the figure shows clearly that the heat transfer is always
raised,withanasymptotictrend,byincreasingthenumberofpasses.

Regardingrefrigerantpressurelosses,Fig.5.3showsthetotalpressuredropalongheat
exchangerwhenthenumberofrefrigerantpassesismodified.Ithasbeenplottedonly
the scenario corresponding to the air velocity of 3 m/s because these results do not
dependonairvelocity.Itshouldbenoticedthatthecasestudycorrespondstoagas
cooler,whichdoesnotundergoaphasechange.

109

5.SIMULATIONSTUDIES

v=1m/s v=3m/s
1100

1000

Heattransfer[W] 900

800

700

600

500
0 2 4 6 8 10 12 14
NumberofPasses
Fig.5.2Heattransferwhennumberofrefrigerantpassesischangedintwoscenarios:air
velocityof3m/sand1m/s.

In a condenser, the pressure drop leads to a temperature drop during the phase
change, therefore the temperature difference between air and refrigerant would go
downandtheheattransferwouldbereduced.Inthisway,forcondensers/evaporators
the pressure drop plays an important role in the heat transfer, in fact there is an
optimum on the heat transfer when the number of refrigerant passes is studied,
becauseoftheoppositeinfluenceoftheheattransfercoefficientsandpressuredrop.
ThisconclusionwasalsoexposedbyShaoetal.(2009)intheirstudiesforaserpentine
microchannel condenser, where they studied the influence on heat transfer of the
numberofpasses.

160
140
Pressuredrop[kPa]

120
100
80
60
40
20
0
0 2 4 6 8 10 12 14
NumberofPasses
Fig.5.3Refrigerantpressuredropalongheatexchangerwhennumberofrefrigerantpassesis
changed.

110

 5.SIMULATIONSTUDIES

5.3 Influence of the fin cuts

A technique to improve the effectiveness in airtorefrigerant heat exchangers is by
cuttingthefins.Theheatconductionbetweentubes,duetotemperaturedifferences
from bottom to top fin roots, degrades the heat exchanger effectiveness. By cutting
thefins,thisheatconductionisavoided.

This technique is indicated for heat exchangers which have large temperature
differences between tubes. For example, in a condenser there are tubes with
superheatedvaporflowinginsidewhichareconnectedthroughfinstoothertubewith
saturated vapor inside. Under these conditions large temperature differences can be
expected. An extreme case corresponds to a gas cooler arrangement, in which the
refrigerantundergoesatemperaturevariationalongthewholegascoolerlength,since
thereisnophasechange.Thusthetemperaturedifferencebetweentwoneighboring
tubescanbeaslargeas50K.

As it was mentioned in the introduction there only exist few models that take into
account heat conduction between tubes. The rest of models always overpredict the
heattransferforthesameconditionssincetheydonotaccountfortheeffectiveness
degradationcausedbytheheatconduction.Theimpactexpectedontheeffectiveness
bycuttingfinsisnotthesameforafinnedtubeasforaMCHX.Inafinnedtubeheat
exchangerthefincutscanbemadeperpendicularlytotheairflowdirection,thusthe
longitudinalheatconductionbetweenrowsoftubesisavoided,whichalwaysdegrades
the effectiveness. In a MCHXthe fins arecut along the air flow direction so that the
effectintroducedbythemisnotfundamentallythesameasforthefinnedtubecase,
in fact the improvements on the capacity are lower: Singh et al. (2010) reported
capacityimprovementsofupto12%fora finnedtubeheatexchangerwhereasPark
and Hrnjak (2007) obtained 3.9% for a serpentine microchannel gas cooler. Notice
that, fin surfaces commonly used for MCHXs are louvered which have louvers that
alreadypreventthelongitudinalheatconductioninfinalongtheairdirection.

Fig.5.4Schematicofthefincutarrangementstudied.

111

5.SIMULATIONSTUDIES

The fin cuts can be customized according to the working conditions and heat
exchangercircuitry.Singhetal.(2010)analyzeddifferentfincutsarrangementsfora
finnedtubegascooler.Inthepresentstudythefincutsstudiedaredisposedalongthe
middlesectionbetweentwoneighbortubesforallthefinsoftheheatexchanger.Fig.
5.4showsanexampleofthisfincutarrangement.TheFin1Dx3modelisdevelopedfor
acontinuousfin,butcanbeslightlymodifiedtoincorporateacutinasectionathalf
thefinheight.Thischangeimplieschangingtwoboundaryconditionofthepiecewise
functionforthefintemperature,whichwaspresentedinsection3.1.Asconsequence
of changingthe boundary conditions it is also needed to obtain the newmatrixes of
themodel:[A],[B],[C](section3.1).

To the authors knowledge there are no numerical studies for MCHXs about the
influenceoftherefrigerantcircuitryontheimpactoffincuts.Tothisend,theimpact
of cutting the fins has been evaluated for the same refrigerant passes studied in
previoussubsection.

The results are shown in Fig. 5.5, where it has been plotted the heat transfer
improvement by cutting fins with respect to the solution given by same model and
same conditions but without fin cuts, i.e. continuous fins. The heat improvementfor
one pass is zero because for this arrangement all the tubes have same temperature
evolution, resulting null the temperature difference between tubes at the same X
coordinate.Insuchacasetheadiabaticfintipassumptionisfundamentallycorrect.

v=1m/s v=3m/s
3.5
3
Qimprovement[%]

2.5
2
1.5
1
0.5
0
0 2 4 6 8 10 12 14
NumberofPasses
Fig.5.5Improvementofheattransferbycuttingfinswithrespecttothesameconditionsbut
withcontinuousfinfordifferentnumberofrefrigerantpassesandfortwoscenarios:airvelocity
of1m/sand3m/s.

The first interesting fact is that the influence of the air velocity on the parameter
studieddoesnotchangethetrendofthecurves,itonlymovesthemvertically.Thus,if

112

 5.SIMULATIONSTUDIES

westudytheplotforv=1m/s,whenthenumberofpassesisdifferentfromonethere
isalwaysanimprovementontheheattransferbycuttingthefinsand,forthestudied
conditions, there is a maximum value for 3 passes, regardless the air velocity. A
possible explanation for the presence of a maximum in the heat improvement is
describedbelow.

Whenthenumberofpassesistwo,thefinrootswhichconnecttwotubesofdifferent
passes(centraltubesoftheheatexchanger)havealargetemperaturedifferencethat
produces a heat conduction flux. As the number of passes is increased the
temperaturedifferencebetweentubesdecreasesbutthenumberoffinswithsucha
temperature difference rises. Fig. 5.1 illustrates this explanation, where the heat
exchangerwith3passeshastwozoneswithlargetemperaturedifference,regionsa
andb.Theserpentineheatexchangerhasasimilarandsmalltemperaturedifference
between all the tubes, which can be represented by the temperature difference at
zone c. Heat exchanger with 3 passes will have only two zones with temperature
difference, but the temperature difference between bottom and top of zones a or
b is much higher than corresponding value for region c of the serpentine case,
thoughserpentineMCHXhas11regionswithasimilartemperaturedifferencetothe
czone.Theseoppositeeffectscouldbeoneofthereasonstoexplainthepresenceof
themaximumdepictedinFig.5.5.

Regarding the influence of the air velocity on these results, Fig. 5.5 shows that the
lower the velocity the larger the improvement. This fact was already pointed out by
Singhetal.(2010)intheirsimulationstudiesforafinnedtubegascooler.

Themaximumimprovementthatcanbeobtaineddependsontheairvelocity,butfor
thescenariosstudiedthisimprovementisasmuch3%.Similarvalueswerereportedby
Park and Hrnjak (2007) who measured capacity improvements of up to 3.9% for a
serpentinegascooler.

113

5.SIMULATIONSTUDIES

5.4 Influence of aspect ratio for a serpentine gas cooler

AserpentineMCHXcorrespondstoaMCHXwithasingletubewhichisbendedinorder
to get a specific number of refrigerant passes. It has the particularity of not having
headersthereforeitishighlyrecommendedforsavingrefrigerantchargethankstoits
reducedinternalvolume.

Arestrictionforthesestudiesisthatairsideandfaceareaareconstantswhileaspect
ratio(L/H)changes.ByobservationoftheserpentineMCHXdesign,itisdeduciblethat
theairsideheattransferareaisproportionaltotheproduct:NpL.Therefore,tostudy
the isolated effect of the aspect ratio on the performance, Np L will have to remain
unchanged for all the cases studied. The baseline gas cooler corresponds to the
twelvepass gas cooler studied in subsection 5.2. When aspect ratio changes, gas
coolerlengthbecomeslargerorshortersothatthenumberofrefrigerantpasseswill
havetochangetokeepNpLconstant.Table5.3liststhecorrespondinglengthL,gas
coolerheightH,andaspectratio,whenthenumberofrefrigerantpassesNpisvaried
accordingtopreviousrestrictions.

Table5.3.Geometricvariablesintheaspectratiostudyforaserpentinegascooler.
Numberofpasses Length(m) Height(m) Aspectratio
2 1.15E+00 2.02E02 5.70E+01
4 5.76E01 4.04E02 1.43E+01
6 3.84E01 6.06E02 6.34E+00
8 2.88E01 8.08E02 3.56E+00
10 2.30E01 1.01E01 2.28E+00
12 1.92E01 1.21E01 1.58E+00
14 1.65E01 1.41E01 1.16E+00
16 1.44E01 1.62E01 8.91E01

Since the tube length changes, the number of segments used by the model to
discretize the gas cooler were also changed in order to keep same accuracy for all
cases. Fig. 5.6 shows the results for the predicted heat transfer as function of the
aspectratio.Thefigureshowstheresultsforthetwoanalyzedcases:thefiniscutand
a continuous fin. The figure shows that aspect ratio has no effect on heat transfer
whenthefiniscut,thusmodelsthatapplyadiabaticfintipassumptionwillnotbeable
tostudythisinfluencesinceresultsarealwaysthesame.

114

 5.SIMULATIONSTUDIES

1100

1050

Heat transfer [W]

1000

950

900
Without fin cuts
With fin cuts
850
1 10 100
Aspect ratio: L/H

Fig.5.6Heattransferofthegascoolerwhentheaspectratioisvariedfortwoscenarios:
continuousfinandfinwithcuts.

For the case of continuous fin, Fig. 5.6 shows that heat transfer has a strong
dependenceonthegascooleraspectratio.AccordingtoTable5.3,thehighestvalueof
aspectratiocorrespondstoNp=2whilelowestvaluecorrespondtoNp=16,therefore
Fig. 5.6 shows that is preferably using lots of refrigerant passes with a short heat
exchangerlengthinsteadoffewpasseswithalargelength,resultinginanasymptotic
trend.Aninterestingobservationisthattheasymptotelookstobethecapacityforthe
cutfincase.Thisfactmeansthattheaspectratiowhichmaximizestheheattransfer
correspondstothevaluethatminimizestheheatconductionbetweentubes.

Noticethattheseconclusionsarenotaffectedbypressurelossesphenomenabecause
therefrigerantcrosssectionalareaandtotallengthofrefrigerantpath arethesame
forallcases,thereforethepressurelosseswillberatherthesame.

5.5 Conclusions
For application where natural refrigerants are used, MCHX minimizes the impact of
their disadvantages, e.g. flammability and high pressures, thanks to its high
compactness,reducedvolumesandhighmechanicalstrength.

From the designers point of view, some parameters of MCHXs such as number of
refrigerant passes, aspect ratio and effect of fin cuts, are hard and expensive to
determinebyexperimentation.Ontheotherhand,successofsimulationtoolstothis
enddependsonthemodelsassumptions,i.e.someparametersproduceeffectsthat
areduetophenomenathatarenottakenintoaccountbythemodel.Forinstance,the

115

5.SIMULATIONSTUDIES

effectoftheaspectratiocannotbestudiedbyamodelthatdoesnotaccountforheat
conduction between tubes; otherwise the results would be always the same. The
capabilities of Fin1Dx3 model allowed assessing numerically the impact on the heat
exchangerperformanceof:numberofrefrigerantpasses,aspectratioandeffectoffin
cuts. Thus, the presented numerical studies are hardly available in literature for
MCHXs.

ForagascoolerworkingwithCO2undertranscriticalpressures,themainconclusions
ofthesimulationsstudieswere:

For a gas cooler where no phase change occurs, heat transfer is always
increased by increasing the number of refrigerant passes regardless the
increaseofpressuredrop.
Thefincutsalwaysincreasetheheattransfer.Inthegascooleranalyzed,the
improvement with regard to the continuous fin depends on the air velocity
and number of refrigerant passes. The lower the velocity, the greater the
capacity improvement. When the number of refrigerant passes is varied,
thereisanoptimumvalueregardlesstheairvelocity,whichis3passesforthe
analyzedcase.Theimprovementinheattransferwasasmuchas3%.
Regarding the aspect ratio of a serpentine heat exchanger, given a heat
transfer area and a face area, the best aspect ratio corresponds to a gas
coolerwithreducedlength(L)andlargeheight(H).Thereasonisbasedonthe
factthatthisconfigurationreducestheheatconductionbetweentubes.

116

CHAPTER6
CONCLUSIONS

117

6.CONCLUSIONS

6 CONCLUSIONS
Belowaresummarizedthemostimportantconclusionsofeachchapter.Finally,a
summaryofthecontributionsofthethesistothestateoftheartinMCHXmodelingis
presented,andpublicationsresultingofthisPhDworkarelisted.

6.1 Global Conclusions

AlongtheintroductiontheusefulnessofMCHXanditsrelativelyrecentandincreasing
application in HVAC&R industry was discussed. A simulation model is an interesting
toolforthedesignerandnowadaysthereonlyexistfewmodelsforthiskindofHXs.In
the introduction, these models were analyzed and some problems were identified
related to the phenomena modeled. The problems and effects identified in the
introductionwere:

Differentvariationoffluidpropertiesforairandrefrigerantalongasegment.
Effectof2Dlongitudinalheatconduction(2DLHC).
Heatconductionbetweentubesalongthefin:applicationofadiabaticfintip
assumption.
Applicationofthefintheory,whichassumesuniformtemperaturethroughout
theairincontactwiththefin.

The main objective of the thesis is to develop an accurate and suitable model for
design purposes of MCHX. To this end the authors decided to study the previous
mentioned effects in some applications of MCHX and once the impact of them was
assessed,todevelopthedefinitivemodel.

Inordertostudyandevaluateimpactofeachphenomenon,Fin2Dmodelwascreated
to obtain very accurate results. It is a very detailed model which allows detecting
inaccuracy sources by evaluating the isolate impact of each effect previously
mentioned. Fin2D does not use the fin efficiency, accounts for 2D LHC in fins and
tubes, accounts for the heat conduction between tubes, and applies a detailed
discretizationfortheair,whichisindependentoftherefrigerantdiscretization.Fin2D
subdividestheheatexchangerintosegmentsandcells(air,refrigerant,fin,tubewall),
towhichasystemofenergyconservationequationsisappliedwithouttraditionalheat
exchanger modeling assumptions. Few models in literature takes into account these
phenomenaandallofthemapplyFEMtechniquesinsteadofFVMthatisemployedby
Fin2D,whichapriorirequireslesscomputationaleffort.

118

 6.CONCLUSIONS

The scenario analyzed with Fin2D model corresponded to a gas cooler working with
CO2 under transcritical pressure. Main conclusions of the studies carried out with
Fin2Dmodelare:

Whenclassicalapproacheswereanalyzed,theerrorobtainedusingtheNTU
method depends on the NTU relationship employed to calculate the
effectivenessofeachsegment.Forthestudiedcase,theerrorissmallerthan
3.5%whentherelationshipforrefrigerantmixedandairunmixed(RMAU)is
used, whereas the error is smaller than 1% when the relationship for both
fluids considered as unmixed (BU) is used. The error becomes larger as the
airsideheattransfercoefficientincreases.Ingeneral,thebestoptionforthe
studiedcaseistousethetubebytubeapproachandtoconsiderbothfluids
asunmixed,thoughtheeffectofthemixedrefrigerantassumptionturnedout
negligible in the scenarios studied. On the other hand, a tubebytube
approach can lead to larger errors when long length tubes are simulated
because refrigerant properties and heat transfer coefficients can have
significant variations, particularly when the refrigerant undergoes a phase
change.It wasreportedthatapplicationofasegmentbysegmentapproach
whentheRMAUrelationshipisadoptedproducesaninconsistencyinresults.
For the operating conditions studied, the impact of LHC effects along each
directioninfinsandtubewalls,ifconsideredseparately,isnotsignificant.The
combined effect is more noticeable and may result in a capacity prediction
errorofasmuchas2.5%,withtheLHCZinthetubebeingthedominanteffect.
Using the adiabaticfintip efficiency, which is commonly applied, leads to
large errors in heat distribution per tube, and therefore in the prediction of
temperature at the tube outlet, when a temperature difference between
tubesexists.
Thetemperatureofairclosetothetubewallisverydifferentthanthebulkair
temperature. This fact could have an important impact on local effects
controllingtheheatandmasstransfer,e.g.dehumidification.

Fin2Dmodelwasgoodtoidentifythedeficiencysourcesoftheclassicalmethodologies
insuchkindofheatexchangers,butwasnogoodtoevaluatetheglobalperformance
when an actual MCHX is simulated regarding number of tubes and number of
refrigerant passes. The main reason which did not allow studying the whole
microchannel gas cooler was the large computation time required by Fin2D model.
Thiscomputationalcostismainlyduetothefinsurfacediscretization:themodelneeds
toemployalargenumberoffincellsbecausenofinefficiencyisusedtosolvetheheat
transferequationalongthefin.

119

6.CONCLUSIONS

Given the main conclusions about the impact of different phenomena and modeling
assumptions in a MCHX, a new model was proposed in order to reduce the
computational cost but preserving same accuracy. To this end, Fin1Dx3 model was
proposed which is based on the Fin2D model but without modeling the negligible
effects and changing the model structure and/or discretization. The result is a much
fastermodelwithalmostthesameaccuracyasFin2Dmodel.Duetothiscomputation
time reduction, Fin1Dx3 allows analyzing microchannel condensers and gas coolers
withanyrefrigerantcircuitry,includingserpentineheatexchangers.

MainconclusionsaboutFin1Dx3modelanditsvalidationarethefollowing:

TheFin1Dx3modelaccountsforallthesameeffectsthanFin2Dmodelexcept
theLHCinthefinalongZdirectionwhich,inanycase,hasbeenfoundnotto
beimportant.Fin1Dx3isbasedonanoveldiscretizationmethodologyforthe
air and fin wall that only needs three air cells along the Y direction. This
methodologyallowsreducingdrasticallythenumberofcellstocompute,with
regardtotheFin2Dmodel,andconsequentlythesimulationtimebutkeeping
ahighresolutionalongYdirection.
ThelargenumberoffincellsneededbyFin2Dmodeltosolveaccuratelythe
airsideheattransfer,iscompensatedinFin1Dx3withanovelmethodologyto
describe the airside heat transfer, using a piecewise function for the fin
temperature. This piecewise function, together with the employed air
discretization, allows applying in a more fundamental way the analytical
solutiongivenbythefintheory.
ThemaincapabilitiesofFin1Dx3are:2DLHCinthetubewall;nonmixedair
effects due to temperature difference between bulk air and the air close to
thetubes;anditaccountsfundamentallyforheatconductionbetweentubes
sinceitdoesnotapplyadiabaticfintipassumption.
The equations have been discretized, with the interesting characteristic of
resultinginasystemofpseudolinearequationswithrespecttothevariables
oftheproblem.Anumericalschemehasbeenproposedtosolvetheproblem
asaseriesofexplicitsteps.Thenumericalschemeproposedallowscomputing
thethreefincellswiththecomputationaleffortofjustonefincell.
Fin1Dx3modelwasvalidatedwithexperimentaldata,forbothcondenserand
gas cooler. The predicted capacity is within 5% error, being much more
accurateforthegascoolerscenario.
Thestudyabouttheinfluenceofthefactor,whichaccountsfortheeffects
of unmixed air flow along Y direction, showed deviations less than 5% for
extremevaluesofandforthesimulatedconditions.

120

 6.CONCLUSIONS

AnevaluationofcapabilitiesofFin1Dx3modelassimulationtoolwascarriedout.To
this end, Fin1Dx3 model was compared in terms of computational cost and accuracy
against others modeling approaches: Fin2D, a simplified approach based on Fin1Dx3
(Fin1Dmodel),andanapproach(CorrectedFinmodel)representativeofothermodels
fromliterature(LeeandDomanski,1997;andSinghetal.,2008).Mainconclusionsof
thesecomparisonswerethefollowing:

ThesolvingtimeofFin1Dx3hasbeenreducedoneorderofmagnitudewith
regardtothetimeemployedbyFin2D,whereasthedifferencesontheresults
are less than 0.3%, which are considered as negligible for practical
applications.ThesolvingtimedifferencebetweenFin1Dx3andFin1Dturned
outtobetwice.
CorrectedFincanleadtoaccurateresultswhencomparedwithanequivalent
approach model that models heat conduction between tubes in a less
artificialwaylikeFin1Dx3does.Differencebetweenpredictedresultsofboth
modelswasbetween1%and2%.ComputationalcostofFin1DandCorrected
Fin models are the same. Nevertheless, author would like to emphasize the
fact that the present work shows no computational saving or accuracy
increase by adding correction terms to anapproach that usesthe adiabatic
fintipefficiencywithregardtoamorefundamentalapproach,regardingthe
heatconductionbetweentubesphenomena,likeFin1D.
By comparison of deviations between Fin1D, CorrectedFin and Fin1Dx3
models,itwasconcludedthatthemainresponsiblefordifferencesbetween
themweretheeffectsofnonmixedairalongfinheight.
Basedontheconclusions,authoranddirectorsofthethesisdecidedtousein
thefollowingtheFin1Dx3modelwithcontinuousfinforsimulationofMCHXs.

Finally, in order to show the application of the Fin1Dx3 model as simulation tool for
design purposes, it was used to carry out numerical studies about the influence of
somedesignparametersofaMCHXonitsperformance.Fromthedesignerspointof
view, some parameters of MCHXs are hard and expensive to determine by
experimentation.Ontheotherhand,successofsimulationtoolstothisenddepends
onthemodelsassumptions,e.g.theeffectoftheaspectratiocannotbestudiedbya
modelthatdoesnotaccountforheatconductionbetweentubes;otherwisetheresults
would be always the same. The capabilities of Fin1Dx3 model allowed assessing
numerically the impact on the heat exchanger performance of some design
parameters for a microchannel gas cooler working with CO2 under transcritical
pressures.Thesestudiesconsistedof:

Influenceofthenumberofrefrigerantpassesonthecapacity.
121

6.CONCLUSIONS

Influenceofthefincutsonthecapacity.
Influenceoftheaspectratio(L/H)foraserpentinemicrochannelgascooler.

Themainconclusionsofthesestudieswere:

For a gas cooler where no phase change occurs, heat transfer is always
increased by increasing the number of refrigerant passes regardless the
increaseofpressuredrop.
Thefincutsalwaysincreasetheheattransfer.Inthegascooleranalyzed,the
improvement with regard to the continuous fin depends on the air velocity
and number of refrigerant passes. The lower the velocity, the greater the
capacity improvement. When the number of refrigerant passes is varied,
thereisanoptimumvalueregardlesstheairvelocity,whichis3passesforthe
analyzedcase.Theimprovementinheattransferwasasmuchas3%.
Regarding the aspect ratio of a serpentine heat exchanger, given a heat
transfer area and a face area, the best aspect ratio corresponds to a gas
coolerwithreducedlength(L)andlargeheight(H).Thereasonisbasedonthe
factthatthisconfigurationreducestheheatconductionbetweentubes.

6.2 Contributions and publications

Inconclusion,fromtheauthorspointofview,themostimportantcontributionsofthis
thesistothestateoftheartinheatexchangersmodelingare:

TheeffectofeachphenomenoninvolvedintheheattransferforaMCHXhas
been evaluated. This analysis allowed identifying the phenomena with the
largesteffectintheperformanceofMCHXs.

If the classical approach NTU is employed to model heat exchangers, the

relationship used to calculate the effectiveness of each segment has to be
chosen carefully. An important finding in the thesis is that using the
relationship for the arrangement of mixed refrigerant and unmixed air
(RMAU) produces an inconsistency in results if a segmentbysegment
approachisadoptedorviceversa.

AuniquemethodologyhasbeendevelopedtomodelMCHXsthatreducesthe
computational cost retaining the accuracy. The high accuracy is achieved by
usinganovelapproachtomodeltheairsideheattransfer.Thisapproachalso
allows getting low computational cost by applying the proposed solving
methodology.

122

 6.CONCLUSIONS

AhighaccuracymodelforsimulationofMCHXs:condensersandgascoolers,
has been developed. It has a low computational cost, which is useful for
designpurposes.Thismodelhasbeenvalidatedagainstexperimentaldata.

Theproposedmodelallowsassessingtheinfluenceofsomeimportantdesign
parametersthatarenotevaluablebyanyothermodelfromliterature,witha
similarcomputationalcostandwithsuchahighaccuracy.Themodelhasbeen
applied to perform numerical studies about the influence of these design
parametersontheperformanceofaMCHX.Thesekindofnumericalstudies
arenotpresentinliterature.

The thesis has shown that it is possible to take into account the heat
conduction between tubes in a less artificial way than other fin efficiency
basedapproaches,whichhavetoaddheatconductiontermsafterassuming
adiabaticfintip, what is inconsistent to a certain extent. The alternative
methodology,proposedinthiswork,consistsinevaluatingtheheattransfer
byintegrationofthecorrespondingfintemperatureprofileinsteadofusinga
fin efficiency which cannot always be defined, e.g. when temperature at fin
rootsaredifferent.Ithasbeenshownthatthisintegrationdoesnotrepresent
a serious drawback for the computation, since it can be easily discretized
consistently with the rest of governing equations, regardless the number of
airfin cells used. Thus, there is neither accuracy nor computational cost
reasons to apply adiabaticfintip assumption when in general it is not
satisfied.

Thefollowingjournalpaperswerepublishedorarealreadysubmittedbutpendingof
acceptationforpublicationasanoutcomeoftheresearchconductedinthisthesis:

MartnezBallesterS.,Corbern,JosM.,GonzlvezMaci,J.,Domanski,
P.A.,2011.ImpactofClassicalAssumptionsinModellingaMicrochannel
GasCooler,InternationalJournalofRefrigeration,34,18981910.
MartnezBallesterS.,Corbern,JosM.,GonzlvezMaci,J.Numerical
ModelforMicrochannelCondensersandGasCoolers:PartIModel
DescriptionandValidation,InternationalJournalofRefrigeration,In
press,DOI:10.1016/j.ijrefrig.2012.08.023.
MartnezBallesterS.,Corbern,JosM.,GonzlvezMaci,J.Numerical
ModelforMicrochannelCondensersandGasCoolers:PartIISimulation
StudiesandModelsComparison,InternationalJournalofRefrigeration,In
press,DOI:10.1016/j.ijrefrig.2012.08.024.

123

6.CONCLUSIONS

The following conference papers were published as an outcome of the research

conductedinthisthesis:

MartnezBallester,S.,JosM.Corbern,J.,GonzlvezMaci,J.,2012.A
NovelApproachtoModeltheAirSideHeatTransferinMicrochannel
Condensers,6thEuropeanThermalSciencesConference,Poitiers,France.
MartnezBallesterS.,GonzlvezMaci,J.,Corbern,JosM.,2012.Impactof
theRefrigerantLayoutandFinCutsonthePerformanceofaMicrochannel
CondenserandaGasCooler,14thInternationalRefrigerationandAir
ConditioningconferenceatPurdue,IN,USA.
MartnezBallesterS.,Corbern,JosM.,GonzlvezMaci,J.,2012.Impactof
FinCutsandRefrigerantLayoutonthePerformanceofaMicrochannelGas
CoolerWorkingwithTranscriticalCO2,10thIIRGustavLorentzenConference
onNaturalRefrigerants,Delft,TheNetherlands.
MartnezBallesterS.,Corbern,JosM.,GonzlvezMaci,J.,2011.
NumericalModelforaMicrochannelGasCoolerwithanyRefrigerant
Circuitry,23rdIIRInternationalCongressofRefrigeration,Prague,Czech
Republic.
MartnezBallesterS.,GonzlvezMaci,J.,Corbern,JosM.,2011.Influencia
delaCircuiterayCortesenlasAletasdeunEnfriadordeGasde
Microcanales,VIICongresoNacionaldeIngenieraTermodinmica,Bilbao,
Spain.
MartnezBallesterS.,Corbern,JosM.,GonzlvezMaci,J.,Domanski,P.A.,
2010.AnalysisofSegmentbySegmentNTUModellingofaMinichannel
CO2GasCooler,13thInternationalRefrigerationandAirConditioning
conferenceatPurdue,IN,USA.

6.3 Future work

In order to use this model as a design tool by manufacturers, researchers and
academicpersonnelithastobeslightlydevelopedinfollowingpoints:

Users graphical interface (GUI): Currently, definition of heat exchanger

geometry,circuitry,operatingcorrelationsandcorrelationswasperformedby
using external files (excel files) or just directly in the source code. A GUI is
plannedtobedevelopedfordefinitionsofalltheseparameters.
Simplification of source code: Thanks to the object oriented programming
language used in the thesis, the source code is quite robust and compact.
Nevertheless, the model was developed for research on modeling

124

 6.CONCLUSIONS

approaches,sotheexistingsourcecodecanberearrangedandmoreclosed
to changes, i.e. less flexible. This task will allow having asimpler, faster and
morecompactsourcecode.
Evaporator Model: Currently, Fin1Dx3 only is valid for condensers since it
doesnotincludedehumidification.Fin1Dx3modelwillbeextendedtoinclude
dehumidification and then serve also as simulation tool for MCHX
evaporators.Foranevaporatorauthoranddirectorsforeseethatdifferences
in accuracy with regard to other existing models will be larger due to
dehumidification. Dehumidification depends strongly on what happens in
tube and fin wall, and Fin1Dx3 models much better these phenomena at
these locations. Thus the impact of dehumification on performance of a
MCHXevaporatorisaphenomenontobestudiedwithFin1Dx3modelinnext
future.
Flow maldistribution: In the governing equations of the Fin1Dx3 model,
uniform flow distribution for both fluids was assumed. This assumption is
nearlytruein acondenserbutnotforan evaporator.Intheairside,theair
flowmaldistributionisreallyaproblemindefiningtheairdistributionatinlet
ratherthanaproblemingoverningequations.DespitethefactthatFin1Dx3
model assumed uniform flow distribution for both fluids, the governing
equationsformodelingtheflowdistributionwasplacedatthemixturecells.
Thus,theirgoverningequationscanbeeasilymodifiedwithoutaffectingthe
restofsourcecode.

125

6.CONCLUSIONS

126

CHAPTER7
APPENDICES

127

7.APPENDICES

7 APPENDICES

7.1 Appendix A: Matrix coefficients for both continuous fin and cut
fin.
Thisappendixcontainsthematrixcoefficientsof[B]and[C].Followingnomenclature
hasbeenused:

f ,a ( i ) pw f ,a ( i )
ma ( i ) m f ,a ( i )
k f Af , a ( i )

128

 7.APPENDICES

7.1.1 [B] for continuous fin

B1,j
Hf ma1 Hf ma1 4 ma2 2 ma3 2 Hf ma2
B1,1=-((-1+ ) ( (ma1-2 ma2) (ma2-ma3)+ (-ma1+2
Hf ma1 2 Hf ma2 2 Hf 2 ma2 ma3
ma2) (ma2-ma3)- (ma1+2 ma2) (ma2-ma3)+ (ma1+2 ma2)
Hf ma1 4 ma2 2 Hf ma2 ma3
(ma2-ma3)- (ma1-2 ma2) (ma2+ma3)- (-ma1+2 ma2) (ma2+ma3)-

4 Hf ma2 (m +2 m ) (m +m )+Hf 2 ma2 ma1 2 ma3 (m +2 m ) (m +m )))/( H

a1 a2 a2 a3 a1 a2 a2 a3 f

2 Hf ma1 2 ma2 ma3 2 Hf ma2

ma1 ( (ma1-ma2) (ma2-ma3)+ (-ma1+ma2) (ma2-ma3)-

2 Hf ma1 ma2 (m +m ) (m -m )+2 Hf 2 ma2 ma3 (m +m ) (m -m )-

2 Hf ma1 2 ma2 (m -m ) (m +m )-2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3

a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3)))

Hf ma1 Hf ma2 2 Hf ma2 2 Hf ma2 ma3 Hf ma2

B1,2=(((-1+ ) ( - ) ma2 ( (ma2-ma3)+
2
(-
2 Hf ma2 Hf ma2 2 ma3
ma2+ma3)- (ma2+ma3)+ (ma2+ma3)))/( Hf ma1 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

a1 a2 a2 a3

12 Hf ma2 Hf ma1 Hf ma3

B1,3=((2 (-1+ ) (-1+
2 2
) ma2 ma3)/( Hf ma1 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

a1 a2 a2 a3

Hf ma1 Hf ma1 4 ma2 2 ma3 2 Hf ma2

B1,4=(((-1+ ) ( (ma1-ma2) (ma2-ma3)+ (-ma1+ma2)
Hf ma1 2 Hf ma2 2 Hf 2 ma2 ma3
(ma2-ma3)- (ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)-

129

7.APPENDICES

Hf ma1 4 ma2 (m -m ) (m +m )-2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2

Hf 2 ma2 ma1 2 ma3
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3)) )/( Hf ma1 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

a1 a2 a2 a3

Hf ma2 2 ma2 ma3 Hf ma1

B1,5=((4 (-1+
2
) ma2 ma3)/( Hf ma1 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

a1 a2 a2 a3

130

 7.APPENDICES

B2,j
Hf ma1 Hf ma2 2 Hf ma2 2 Hf ma2 ma3 Hf ma2
B2,1=-((-1+ ) ( - ) ma1 ( (ma2-ma3)+
2
(-
2 Hf ma2 Hf ma2 2 ma3
ma2+ma3)- (ma2+ma3)+ (ma2+ma3)))/((-1+2 ) Hf ma2 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m )))

a1 a2 a2 a3

Hf ma2 2 Hf ma2 2 Hf ma2 ma3 Hf ma2

B2,2=((( - ) ( (ma1 (ma2-2 ma3)-ma2 ma3)- (ma1
2 Hf ma1 ma2 ma3 Hf 2 ma1 ma2
(ma2-2 ma3)+ma2 ma3)+ (ma1 (ma2-2 ma3)+ma2 ma3)+ (-ma1
2 Hf ma1 ma2 Hf ma2 2 ma3
ma2+2 ma1 ma3+ma2 ma3)+ (ma2 ma3-ma1 (ma2+2 ma3))+ (-ma2
2 Hf ma2 Hf ma2 2 ma1 ma3
ma3+ma1 (ma2+2 ma3))- (ma2 ma3+ma1 (ma2+2 ma3))+ (ma2
2 Hf ma1 2 ma2 ma3
ma3+ma1 (ma2+2 ma3))))/((-1+2 ) Hf ma2 ( (ma1-ma2) (ma2-ma3)+

2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2 (m +m ) (m -m )+2 Hf 2 ma2 ma3

2 Hf ma1 2 ma2 2 Hf ma2 ma3
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)- (ma1-ma2) (ma2+ma3)- (-ma1+ma2)

4 Hf ma2 2 Hf ma2 ma1 ma3
(ma2+ma3)- (ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3))))

Hf ma2 2 Hf ma2 Hf ma3 Hf ma2 2 Hf ma1 ma2

B2,3=(-(( - ) (-1+ ) ( (ma1-ma2)+
2
(-
2 Hf ma2 Hf 2 ma1 ma2
ma1+ma2)- (ma1+ma2)+ (ma1+ma2)) ma3)/((-1+2 ) Hf ma2 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

2 Hf ma2 ma3
a1 a2 a2 a3

Hf ma1 Hf ma2 2 Hf ma2 Hf ma2

B2,4=(-(2 ( - ) ma1 ( (ma2-ma3)+ (-
2 Hf ma2 Hf ma2 2 ma3
ma2+ma3)- (ma2+ma3)+ (ma2+ma3)))/((-1+2 ) Hf ma2 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

a1 a2 a2 a3 a1 a2 a2 a3

131

7.APPENDICES

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)
2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

a1 a2 a2 a3

Hf ma3 Hf ma2 2 Hf ma2 Hf ma2 2 Hf ma1 ma2

B2,5=(-(2 ( - ) ( (ma1-ma2)+ (-
2 Hf ma2 Hf 2 ma1 ma2
ma1+ma2)- (ma1+ma2)+ (ma1+ma2)) ma3)/((-1+2 ) Hf ma2 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

a1 a2 a2 a3

132

 7.APPENDICES

B3,j
12 Hf ma2 Hf ma1 Hf ma3
B3,1=(2 (-1+ )2 (-1+ )2 ma1 ma2)/( Hf ma3 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m )))

a1 a2 a2 a3

Hf ma2 2 Hf ma2 Hf ma3 Hf ma2 2 Hf ma1 ma2

B3,2=((( - ) (-1+ ) ma2 ( (ma1-ma2)+
2
(-
2 Hf ma2 Hf 2 ma1 ma2
ma1+ma2)- (ma1+ma2)+ (ma1+ma2)))/( Hf ma3 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

a1 a2 a2 a3

Hf ma3 Hf 2 ma1 4 ma2 ma3 2 Hf ma2

B3,3=(-((-1+ ) ( (ma1-ma2) (2 ma2-ma3)+ (-ma1+ma2)
2 Hf ma1 ma2 Hf 4 ma2 ma3
(2 ma2-ma3)- (ma1+ma2) (2 ma2-ma3)+ (ma1+ma2) (2 ma2-
2 Hf ma1 2 ma2 2 Hf ma2 Hf ma3
ma3)- (ma1-ma2) (2 ma2+ma3)- (-ma1+ma2) (2 ma2+ma3)-

4 Hf ma2 (m +m ) (2 m +m )+Hf 2 ma2 2 ma1 ma3 (m +m ) (2 m +m )))/( H

a1 a2 a2 a3 a1 a2 a2 a3 f

2 Hf ma1 2 ma2 ma3 2 Hf ma2

ma3 ( (ma1-ma2) (ma2-ma3)+ (-ma1+ma2) (ma2-ma3)-

2 Hf ma1 ma2 (m +m ) (m -m )+2 Hf 2 ma2 ma3 (m +m ) (m -m )-

2 Hf ma1 2 ma2 (m -m ) (m +m )-2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3

a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3))))

Hf ma2 ma1 2 ma2 Hf ma3

B3,4=(-(4 (-1+
2
) ma1 ma2)/( Hf ma3 (-

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )-2 Hf ma2 (-m +m ) (m -m )+2 Hf ma1 ma2

a1 a2 a2 a3 a1 a2 a2 a3

133

7.APPENDICES

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2

(ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+ (ma1-ma2)
2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)+ (-ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3)-

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

a1 a2 a2 a3

Hf ma3 Hf 2 ma1 4 ma2 ma3 2 Hf ma2

B3,5=(((-1+ ) ( (ma1-ma2) (ma2-ma3)+ (-ma1+ma2)
2 Hf ma1 ma2 Hf 4 ma2 ma3
(ma2-ma3)- (ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)-

2 Hf ma1 2 ma2 (m -m ) (m +m )-2 Hf ma2 Hf ma3 (-m +m ) (m +m )-4 Hf ma2

Hf 2 ma2 2 ma1 ma3
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3)))/( Hf ma3 (

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))))

a1 a2 a2 a3

134

 7.APPENDICES

7.1.2 [B] for cut fin

When cut fin is modeled a simplified approach is used in Fin1Dx3_Cut. The required
evaluationofboundaryconditionsturnsoutverysimpleifthefincuthasatotalheight
givenby(12)Hf,i.e.thefincutisnotjustalinebutithasfinitedimensions,whatitis
totallytrue.Averythincutwillcorrespondtothecasewhen(12)Hftendstobezero,
whatimpliesthattendtobe0.5.

ForthismodelthefincutisadiscontinuitysothatcalculationofTf2(Y)isnotdoneand
Ta2doesnotchangealongYanditsimpactonHXperformanceisnullsinceairflowrate
of this cell tends to be zero. Important consequences of this discontinuity is that
equations for fin f1 and fin f3 are decoupled and the number of the unknown
constantsofEq.(3.14)(C1,C2,C3,C4,C5,C6)isreducedtofourconstants:(C1,C2,C5,C6),
whicharecalculatedwithfollowingboundaryconditions:

T f 1 (Y 0) T fB

T f 3 (Y H f ) T fT
dT
f1
dY 0
Y H f

dT
f3 0
dY Y (1 ) H
f

135

7.APPENDICES

B1,j
2 Hf ma1 2 Hf ma1
B1,1=(1- )/( Hf ma1 ( +1))

B1,2= 0

B1,3= 0

B1,4= - B1,1
B1,5=0

B2,j
Tf2willnotbeevaluatedbecausef2correspondstothefincutanditisadiscontinuity.
In order to maintain same equations structure and solving methodology the way to
solvethisdiscontinuityistomakenullalltheelementsrelatedtoTf2:

B2,1= 0

B2,2= 0

B2,3= 0

B2,4= 0

B2,5= 0

B3,j
B3,1= 0

B3,2= 0
2 Hf ma3 2 Hf ma3
B3,3=(1- )/( Hf ma3 ( +1))

B3,4= 0
B3,5= -B3,3

136

 7.APPENDICES

7.1.3 [C] for continuous fin

C1,j
Hf ma1 Hf ma1 4 ma2 2 ma3 2 Hf ma2
C1,1=((-1+ ) ma1 ( (ma1-ma2) (ma2-ma3)+ (-ma1+ma2)
Hf ma1 2 Hf ma2 2 Hf 2 ma2 ma3
(ma2-ma3)- (ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)-

Hf ma1 4 ma2 (m -m ) (m +m )-2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2

Hf 2 ma2 ma1 2 ma3 2 Hf ma1 2 ma2 ma3
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3)))/(

2 Hf ma2 2 Hf ma1 ma2
(ma1-ma2) (ma2-ma3)+ (-ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+

2 Hf 2 ma2 ma3 (m +m ) (m -m )-2 Hf ma1 2 ma2 (m -m ) (m +m )-

2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2 (m +m ) (m +m )+

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))

a1 a2 a2 a3 a1 a2 a2 a3

a1 a2 a2 a3

C1,2=((2 Hf ma1 (Hf ma2 -2 Hf ma2 ) m m (2 Hf ma2 ma3 (m -m )+Hf ma2 (-

Hf ma2 2 ma3 2 Hf ma1 2 ma2 ma3
a1 a2 a2 a3

2 Hf ma2
ma2+ma3)- (ma2+ma3)+ (ma2+ma3)))/( (ma1-
2 Hf ma2 2 Hf ma1 ma2
ma2) (ma2-ma3)+ (-ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+

2 Hf 2 ma2 ma3 (m +m ) (m -m )-2 Hf ma1 2 ma2 (m -m ) (m +m )-

2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2 (m +m ) (m +m )+

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3 (m +m ) (m +m )))

a1 a2 a2 a3 a1 a2 a2 a3

a1 a2 a2 a3

C1,3=((4 Hf ma2 ma1 2 ma2 (-1+ Hf ma3 )2 m m m )/(2 Hf ma1 2 ma2 ma3
2 Hf ma1 ma2
a1 a2 a3

2 Hf ma2
(ma1-ma2) (ma2-ma3)+ (-ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+

2 Hf 2 ma2 ma3 (m +m ) (m -m )-2 Hf ma1 2 ma2 (m -m ) (m +m )-

2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2 (m +m ) (m +m )+

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3 (m +m ) (m +m )))

a1 a2 a2 a3 a1 a2 a2 a3

a1 a2 a2 a3

2 Hf ma1 2 ma2 ma3 2 Hf ma2

C1,4=(-(ma1 ( (ma1-ma2) (ma2-ma3)- (-ma1+ma2) (ma2-ma3)-

2 Hf ma1 ma2 (m +m ) (m -m )-2 Hf 2 ma2 ma3 (m +m ) (m -m )-

a1 a2 a2 a3 a1 a2 a2 a3

137

7.APPENDICES

2 Hf ma1 2 ma2 (m -m ) (m +m )+2 Hf ma2 ma3 (-m +m ) (m +m )+4 Hf ma2

2 Hf ma2 ma1 ma3 2 Hf ma1 2 ma2 ma3
a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3)))/(

2 Hf ma2 2 Hf ma1 ma2
(ma1-ma2) (ma2-ma3)+ (-ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+

2 Hf 2 ma2 ma3 (m +m ) (m -m )-2 Hf ma1 2 ma2 (m -m ) (m +m )-

2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2 (m +m ) (m +m )+

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3 (m +m ) (m +m )))

a1 a2 a2 a3 a1 a2 a2 a3

a1 a2 a2 a3

C1,5=((8 Hf ma2 ma1 2 ma2 ma3 m m m )/(4 Hf ma2 (m +m ) (-m +2 Hf ma3

a1 a2 a3 a1 a2 a2

2 Hf ma2 2 Hf ma3 2 Hf ma1 4 Hf ma2

(ma2-ma3)-ma3)+ (ma1-ma2) (-ma2+ma3+ (ma2+ma3))+ (
2 Hf ma3 2 Hf ma2 2 Hf ma3
(ma1-ma2) (-ma2+ (ma2-ma3)-ma3)+ (ma1+ma2) (-ma2+ma3+

(ma2+ma3)))))

138

 7.APPENDICES

C2,j
Hf ma2 2 ma2 ma3 Hf ma1 2 Hf ma1 2 ma2 ma3
C2,1=-(4 (-1+ ) ma1 ma2 ma3)/(
2
(ma1-
2 Hf ma2 2 Hf ma1 ma2
ma2) (ma2-ma3)+ (-ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+

2 Hf 2 ma2 ma3 (m +m ) (m -m )-2 Hf ma1 2 ma2 (m -m ) (m +m )-

2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2 (m +m ) (m +m )+

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3 (m +m ) (m +m ))

a1 a2 a2 a3 a1 a2 a2 a3

a1 a2 a2 a3

Hf ma3 Hf ma2 2 Hf ma2 Hf ma2 2 Hf ma1 ma2

C2,2=(-(2 ( - ) ma2 ( (ma1-ma2)+ (-
2 Hf ma2 Hf 2 ma1 ma2 2 Hf ma1 2 ma2 ma3
ma1+ma2)- (ma1+ma2)+ (ma1+ma2)) ma3)/(
2 Hf ma2 2 Hf ma1 ma2
(ma1-ma2) (ma2-ma3)+ (-ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+

2 Hf 2 ma2 ma3 (m +m ) (m -m )-2 Hf ma1 2 ma2 (m -m ) (m +m )-

2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2 (m +m ) (m +m )+

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3 (m +m ) (m +m )))

a1 a2 a2 a3 a1 a2 a2 a3

a1 a2 a2 a3

Hf ma3 Hf 2 ma1 4 ma2 ma3 2 Hf ma2

C2,3=(-((-1+ ) ma3 ( (ma1-ma2) (ma2-ma3)+ (-
2 Hf ma1 ma2 Hf 4 ma2 ma3
ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-
2 Hf ma1 2 ma2 2 Hf ma2 Hf ma3
ma3)- (ma1-ma2) (ma2+ma3)- (-ma1+ma2) (ma2+ma3)-

4 Hf ma2 (m +m ) (m +m )+Hf 2 ma2 2 ma1 ma3 (m +m ) (m +m )))/(

2 Hf ma1 2 ma2 ma3 (m -m ) (m -m )+2 Hf ma2 (-m +m ) (m -m )-2 Hf ma1 ma2

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf 2 ma2 ma3 2 Hf ma1 2 ma2

a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2-ma3)+ (ma1+ma2) (ma2-ma3)- (ma1-ma2)

2 Hf ma2 ma3 4 Hf ma2
(ma2+ma3)- (-ma1+ma2) (ma2+ma3)- (ma1+ma2) (ma2+ma3)+

2 Hf ma2 ma1 ma3 (m +m ) (m +m )))

a1 a2 a2 a3

Hf ma2 ma1 2 ma2 ma3 2 Hf ma1 2 ma2 ma3

C2,4=(-(8 ma1 ma2 ma3)/( (ma1-ma2) (ma2-
2 Hf ma2 2 Hf ma1 ma2 2 Hf 2 ma2 ma3
ma3)+ (-ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+

139

7.APPENDICES

2 Hf ma1 2 ma2 2 Hf ma2 ma3

(ma1+ma2) (ma2-ma3)- (ma1-ma2) (ma2+ma3)- (-ma1+ma2)
4 Hf ma2 2 Hf ma2 ma1 ma3
(ma2+ma3)- (ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3)))

2 Hf ma1 2 ma2 ma3 2 Hf ma2

C2,5=((ma3 ( (ma1-ma2) (ma2-ma3)- (-ma1+ma2) (ma2-ma3)+

2 Hf ma1 ma2 (m +m ) (m -m )+2 Hf 2 ma2 ma3 (m +m ) (m -m )+

2 Hf ma1 2 ma2 (m -m ) (m +m )-2 Hf ma2 ma3 (-m +m ) (m +m )+4 Hf ma2

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3 2 Hf ma1 2 ma2 ma3

a1 a2 a2 a3 a1 a2 a2 a3

(ma1+ma2) (ma2+ma3)+ (ma1+ma2) (ma2+ma3)))/(

2 Hf ma2 2 Hf ma1 ma2
(ma1-ma2) (ma2-ma3)+ (-ma1+ma2) (ma2-ma3)- (ma1+ma2) (ma2-ma3)+

2 Hf 2 ma2 ma3 (m +m ) (m -m )-2 Hf ma1 2 ma2 (m -m ) (m +m )-

2 Hf ma2 ma3 (-m +m ) (m +m )-4 Hf ma2 (m +m ) (m +m )+

a1 a2 a2 a3 a1 a2 a2 a3

2 Hf ma2 ma1 ma3 (m +m ) (m +m )))

a1 a2 a2 a3 a1 a2 a2 a3

a1 a2 a2 a3

140

 7.APPENDICES

7.1.4 [C] for cut fin

Thesameapproachappliedforgettingcoefficientsofmatrix[B]incaseofcutfinhas
beenappliedtogetthecoefficientsofmatrix[C].

C1,j
2 Hf ma1 2 Hf ma1
C1,1= ma1 ( -1)/( +1)
C1,2= 0

C1,3= 0
C1,4= -C1,1

C1,5= 0

C2,j
C2,1= 0

C2,2= 0
2 Hf ma3 2 Hf ma3
C2,3= ma3 (1- )/( +1)

C2,4= 0

C2,5= -C2,3

141

7.APPENDICES

7.2 Appendix B: Experimental data used for validation of model

7.2.1 Gas cooler

Alldatarelatedtothegascoolerusedinthisthesisformodelsvalidationandother
numericalstudiesisextractedfromYinetal.(2001),whichissummarizedbelow.

Gascoolergeometry

GasCoolerCharacteristics
Mass(kg) 2.3
2
FaceArea(cm ) 1950
CoreDepth(cm) 1.65
3
CoreVolume(cm ) 3320
2
AirsideArea(m ) 5.2
RefrigerantSideArea(m2) 0.49
FinDensity(fins/in) 22
LouverAngle(deg) 23
TubeLength(mm) 545
NumberofPorts 11
PortDiameter(mm) 0.79
WebThickness(mm) 0.70
WallThickness(mm) 0.43
FinHeight(mm) 8.89
FinThickness(mm) 0.10
LouverHeight(mm) 7.16
LouverPitch(mm) 0.99
NumberofLouvers 2x6
LouverredirectionLength(mm) 1.7
LouverentryLength(mm) 1.7
Headertubediameter(mm) 2@7
Inlet/exitelbowtubediameter(mm) 7

142

 7.APPENDICES

Operatingconditions

Test mr Pr,in Pr Tr,in Tr,out ma Ta,in Pa

[#] [g/s] [kPa] [kPa] [C] [C] [g/s] [C] [Pa]
1 34.74 11007 137.7 108.9 48.2 542 43 43
2 56.36 10792 421.6 138.6 50.2 701 43.5 61
3 39.98 12014 155.4 125.4 49 699 44.3 62
4 26.39 12464 62.7 115.8 58 457 55.1 45
5 37.84 10937 116.3 124.7 49 537 42.7 46
6 22.37 9556 60.9 97.9 45.8 453 43.1 39
7 27.12 10982 97.5 101.2 56 459 53.9 46
8 27.48 9333 99 91 45.5 452 43 39
9 31.49 8858 103.6 66.5 38.4 452 31.8 34
10 24.07 14390 25.4 113.3 45.4 452 43 40
11 22.08 10555 28.2 103 35.9 453 32.4 34
12 22.06 11392 37.2 114.8 46.8 437 43.4 38
13 25.17 8386 91.1 87.4 36.8 451 31.5 32
14 38.49 10278 197.9 118.9 48.4 537 43.3 47
15 19.6 12460 27.6 126.6 46 454 43.5 40
16 44.31 9061 337.3 104 46.4 541 43.4 47
17 21.67 9514 24 102 36.9 451 32.4 34
18 23.56 9841 52.7 107.1 46 452 42.8 39
19 45.58 8587 344.3 101.7 45 536 42.7 46
20 24.68 8460 67.3 89 37.2 451 31.8 33
21 47.53 8677 389.9 97.8 45.1 535 42.7 46
22 25.92 9387 92.2 97 45.7 455 43.1 40
23 26.6 8435 78 81.3 37.2 453 32.2 34
24 25.98 10014 84.4 96.6 46.5 461 42.9 40
25 42.73 9713 239.4 108 47.5 539 43.2 48
26 25.25 8328 66.2 84.4 36.8 453 32.4 34
143

7.APPENDICES

Test mr Pr,in Pr Tr,in Tr,out ma Ta,in Pa

[#] [g/s] [kPa] [kPa] [C] [C] [g/s] [C] [Pa]
27 25.11 9593 78.5 96.9 45.7 453 43 41
28 43.12 9200 286.9 105.4 46.3 539 42.9 47
29 26 8331 59.1 81.2 36.9 453 32.1 35
30 37.69 10259 166 120.9 48.6 540 43.7 49
31 22.83 9537 18.9 93.8 37.4 450 32.1 34
32 26.04 7659 73.8 78.2 32.5 454 26.8 32
33 25.45 8359 66.1 87.3 36.7 456 31.7 35
34 32.63 9242 145 83.3 45.5 455 43.3 38
35 26.06 9349 93.2 94.1 45.2 536 43.8 47
36 25.55 9385 85.3 95.4 44.9 710 43.6 66
37 24.72 7826 57.1 79.9 33.4 448 26.8 31
38 26.23 8204 81.9 82.3 36.4 449 32.4 34
39 20.78 9833 46.2 106.8 46 451 43.6 40
40 34.86 10735 168.9 129.8 48.7 540 43.7 48
41 32.91 9879 150.5 110.5 47.3 502 43.7 44
42 31.47 10315 127.9 116 48.3 501 44.1 45
43 29.95 10772 109.4 121.1 48.2 502 43.7 45
44 28.82 11251 92.8 125.8 48.1 501 43.6 45
45 27.94 11745 84 130.3 48 501 43.7 45
46 23.02 8303 29.6 83.3 33.1 455 26.7 29
47 22.9 8413 31 85.5 33.7 447 27 31

144

 7.APPENDICES

7.2.2 Condenser
Alldatarelatedtothegascoolerusedinthisthesisformodelsvalidationandother
numericalstudiesisextractedfromGarcaCascalesetal.(2001),whichissummarized
below.

Condensergeometry

145

7.APPENDICES

HeatExchanger1 HeatExchanger2
Tubes 33 66
Fintype Louveredfins Louveredfins
Finsperinch 12 10
Corewidth(mm) 534.56 889
Finnedlength(mm) 482.6 641

146

 APPENDICES

Operatingconditions

Test t1.1 t1.2 t1.3 t1.4 t2.1 t2.2 t2.3 t2.4

Secondaryfluid Air
InletairdrybulbT(C) 35 34.98 34.92 34.99 35.02 35.01 35.01 35
InletairwetbulbT(C) 22.74 22.69 22.7 22.79 21.62 21.15 21.61 21.71
OutletairdrybulbT(C) 42.18 42.66 43.21 43.9 42.24 42.79 43.41 44.19
Atmosphericpressure(kPa) 98.6 98.6 98.64 98.62 99.22 99.22 99.21 99.21
3
Volumetricflowrate(m /h) 1877.19 1625.38 1375.56 1126.41 7334 6342 5369 4388
Refrigerant R410A
Inlettemperature(C) 79.57 79.47 79.5 79.85 79.24 79.44 79.73 79.65
Inletpressure(kPa) 2977.29 2977.88 2975.76 2977.6 2978.78 2980.38 2974.51 2975.85
Outlettemperature(C) 42.77 42.7 42.93 42.68 42.71 42.39 42.5 42.58
Outletpressure(kPa) 2955.81 2959.47 2960.29 2965.82 2943.88 2950.35 2949.01 2954.95
Outletsaturation(C) 48.35 48.4 48.41 48.49 48.17 48.27 48.25 48.33
Outletsubcooling(C) 5.577 5.705E+00 5.487E+00 5.812E+00 5.460E+00 5.880E+00 5.740E+00 5.750E+00
Massflowrate(kg/s) 0.023765 0.019645 0.019645 0.016881 0.09236 0.08508 0.07681 0.0681
PressureDrop(kPa) 2.066E+01 1.763E+01 1.464E+01 1.103E+01 3.419E+01 2.944E+01 2.490E+01 2.040E+01
HeatTransferred(kW) 4.754 4.362 3.923 3.386 18.446 17.064 15.424 13.657

147

REFERENCES

REFERENCES
Agarwal,A.,Bandhauer,T.M.,Garimella,S.,2010.Measurementandmodelingof
condensationheattransferinnoncircularmicrochannels,Int.J.Refrigeration,33(6),
11691179.

Asinari,P.,2004.FinitevolumeandFiniteelementHybridTechniqueforthe
CalculationofComplexHeatExchangersbySemiexplicitMethodforWallTemperature
LinkedEquations(SEWTLE).NumericalHeatTransferPartBFundamentals,45,221
247.

Asinari,P.,Cecchinato,L.,Fornasieri,E.,2004.EffectsofThermalConductionin
MicrochannelGasCoolersforCarbonDioxide.Int.J.Refrigeration27(6),577586.

Ayad,F.,Benelmir,R.,Souayed,A.,2012.CO2evaporatorsdesignforvehicleHVAC
operation,AppliedThermalEngineering,36,330344.

Bertsch,S.S.,A.Groll,E.A.,Garimella,S.V.,2008.Refrigerantflowboilingheattransfer
inparallelmicrochannelsasafunctionoflocalvaporquality,InternationalJournalof
HeatandMassTransfer,51(1920),47754787.

BrixW.,Krn,M.R.,Elmegaard,B.,2009.Modellingrefrigerantdistributionin
microchannelevaporators,Int.J.Refrigeration,32(7),17361743.

BrixW.,Krn,M.R.,Elmegaard,B.,2010.ModellingdistributionofevaporatingCO2in
parallelminichannels,Int.J.Refrigeration,33(6),10861094.

CavalliniA.,CensiG.,DelColD.,DorettiL.,LongoG.A.,RossettoL.,2002.Intube
condensationofhalogenatedrefrigerants.ASHRAETrans.,108(1),pp.146161.

Cavallini,A.,DelCol,D.,Doretti,L.,Matkovic,M.,Rossetto,L.,Zilio,C.,2005.
CondensationHeatTransferandPressureGradientInsideMultiportMinichannels,
HeatTransferEngineering,26(3),4555.

Cavallini,A.,DelCol,D.,Matkovic,M.,Rossetto,L.,2009.Frictionalpressuredrop
duringvapourliquidflowinminichannels:Modellingandexperimentalevaluation,
InternationalJournalofHeatandFluidFlow,30(1),131139.

Chen,T.,Garimella,S.V.,2011.Localheattransferdistributionandeffectof
instabilitiesduringflowboilinginasiliconmicrochannelheatsink,International
JournalofHeatandMassTransfer,54(1516),31793190.

148

 REFERENCES

ChurchillS.W.,1977.FrictionfactorEquationSpansAllFluidFlowRegimes.Chemical
Engineering,7:9192.

CoilDesigner,2010.TooltoAidintheDesign,SimulationandOptimizationofAir
CooledHeatExchangers,http://ceee.umd.edu/isoc/software/index_coildesigner.htm,
UniversityofMaryland,CenterforEnvironmentalEnergyEngineering,MD,USA.

CorbernJ.M.,GonzlvezJ.,MontesP.,BlascoR.,2002.ARTaComputerCodeto
AssisttheDesignofRefrigerationandA/CEquipment.InternationalRefrigerationand
AirConditioningConferenceatPurdue,IN,USA.

Corbern,J.M.,DeCordoba,P.F.,Gonzalvez,J.,Alias,F.,2001.SemiexplicitMethodfor
WallTemperatureLinkedEquations(SEWTLE):AGeneralFiniteVolumeTechniquefor
theCalculationofComplexHeatExchangers.Numer.HeatTransfer,PartB40,3759.

Domanski,P.A.,Choi,J.M.,Payne,W.V.,2007.LongitudinalHeatConductioninFinned
TubeEvaporator.22ndIIRInternationalCongressofRefrigeration,Beijing,China.

EVAPCOND,2010.SimulationModelsforFinnedTubeHeatExchangerswithCircuitry
Optimization,http://www.nist.gov/el/building_environment/evapcond_software.cfm,
NationalInstituteofStandardsandTechnology,BuildingandFireResearchLaboratory,
Gaithersburg,MD,USA.

Fernando,P.,Palm,B.,Ameel,T.,Lundqvist,P.,Granryd,E.,2008.Aminichannel
aluminiumtubeheatexchangerPartII:Evaporatorperformancewithpropane.Int.J.
Refrigeration,31(4),681695.

Fernando,P.,Palm,B.,Lundqvist,P.,Granryd,E.,2004.Propaneheatpumpwithlow
refrigerantcharge:designandlaboratorytests,Int.J.Refrigeration,27(7),761773.

FriedelL.,1980.Pressuredropduringgas/vaporliquidflowinpipes.Int.Chem.Eng.,
July,20,pp.352367.

Fronk,B.M.,GarimellaS.,2011.WaterCoupledCarbonDioxideMicrochannelGas
CoolerforHeatPumpWaterHeaters:PartIIModelDevelopmentandValidation,Int.
J.Refrigeration,34,1728.

GarcaCascales,J.R.,VeraGarca,F.,GonzlvezMaci,J.,CorbernSalvador,J.M.,
Johnson,M.W.,Kohler,G.T.,2010.CompactHeatExchangersModeling:Condensation.
Int.J.Refrigeration33,135147.

149

REFERENCES

Garimella,S.,Agarwal,A.,Killion,J.D.,2005.CondensationPressureDropinCircular
Microchannels,HeatTransferEngineering,26(3),2835.

GnielinskiV.,1976.NewEquationsforHeatandMassTransferinTurbulentPipeand
ChannelFlow.Int.Chem.Eng.16(2),359368.

Hrnjak,P.,2010.DevelopmentsinChargeReductionandMicrochannelTechnology,
SustainableRefrigerationandHeatPumpTechnologyConference,Stockholm,Sweden,
2010.

Hrnjak,P.,Litch,A.D.,2008.Microchannelheatexchangersforchargeminimizationin
aircooledammoniacondensersandchillers,Int.J.Refrigeration,31(4),658668.

IDAE,2005.Aireacondicionadoanuestrasnecesidadesreales.http://www.idae.es/,
InstitutoparalaDiversificacinyAhorrodelaEnerga(IDAE).

IEC603351,2010.InternationalStandard:Householdandsimilarelectricalappliances
SafetyPart1:Generalrequirements.InternationalElectrotechnicalCommission.

IMSTART,2010.Simulationtooltoassisttheselection,designandoptimizationof
refrigerationequipmentandcomponents,http://www.imstart.com,
UniversitatPolitcnicadeValncia,InsitutodeIngenieraEnergtica,Spain.

Incropera,F.P.,DeWitt,D.P.,1996.FundamentalsofHeatandMassTransfer,fourth
ed.JohnWileyandSons,NewYork.

Jiang,H.B.,2003.Ph.D.Thesis,DevelopmentofaSimulationandOptimizationToolfor
HeatExchangerDesign.UniversityofMaryland,USA.

Jiang,H.B.,Aute,V.,Radermacher,R.,2006.Coildesigner:aGeneralPurpose
SimulationandDesignToolforAirToRefrigerantHeatExchangers.Int.J.
Refrigeration29(4),601610.

Joardar,A.,Jacobi,A.M.,2005.Impactofleadingedgedeltawingvortexgeneratorson
thethermalperformanceofaflattube,louveredfincompactheatexchanger,
InternationalJournalofHeatandMassTransfer,48(8),14801493.

Kandlikar,S.G.,2002.Fundamentalissuesrelatedtoflowboilinginminichannelsand
microchannels,ExperimentalThermalandFluidScience,26(24),389407.

Kandlikar,S.G.,andGrande,W.J.,2002.EvolutionofMicrochannelFlowPassages
ThermohydraulicPerformanceandFabricationTechnology.HeatTransferEng.,25(1),
pp.317.
150

 REFERENCES

KaysW.M.,LondonA.L.,1984,CompactHeatExchangers,3rded.McGrawHill,New
York.

Kew,P.A.,Reay,D.A.,2011.Compact/microheatexchangersTheirroleinheat
pumpingequipment,AppliedThermalEngineering,31(5),594601.

Kim,M.H.,Bullard,C.W.,2001.Developmentofamicrochannelevaporatormodelfora
CO2airconditioningsystem,Energy,26(10),931948.

KimM.H.,BullardC.,2002,AirSideThermalHydraulicPerformanceofMultiLouvered
FinAluminumHeatExchangers,Int.J.Refrigeration,25:390400.

Kim,M.H.,Pettersen,J.,Bullard,C.W.,2004.Fundamentalprocessandsystemdesign
issuesinCO2vaporcompressionsystems,ProgressinEnergyandCombustionScience,
30(2),119174.

Kim,M.H.,Sumin,S.,Bullard,C.W.,2002.Effectofinlethumidityconditionontheair
sideperformanceofaninclinedbrazedaluminumevaporator,Int.J.Refrigeration,
25(5),611620.

Klein,S.A,2004.EngineeringEquationSolver,FChartSoftware,Madison,WI(USA).

Kulkarni,T.,Bullard,C.W.,Cho,K.,2004.Headerdesigntradeoffsinmicrochannel
evaporators,AppliedThermalEngineering,24(56),759776.

Lee,J.,Domanski,P.A.,July1997.ImpactofAirandRefrigerantMaldistributionsonthe
PerformanceofFinnedTubeEvaporatorswithR22andR407C.ReportNo.:
DOE/CE/2381081.

Lemmon,E.W.,McLinden,M.O.,M.L.Huber.2002.REFPROP,Version7.0.U.S.
DepartmentofCommerce,Maryland.

Li,Q.,Flamant,G.,Yuan,X.,Neveub,P.,Luo,L.2011.Compactheatexchangers:A
reviewandfutureapplicationsforanewgenerationofhightemperaturesolar
receivers,RenewableandSustainableEnergyReviews,15(9),48554875.

Lia,B.,Peuker,S.,Hrnjak,P.S.,Alleyne,A.G.,2011.Refrigerantmassmigration
modelingandsimulationforairconditioningsystems,AppliedThermalEngineering,
31(10),17701779.

Moallem,E.,Cremaschi,L.,Fisher,D.E.,Padhmanabhan,S.,2012.Experimental
measurementsofthesurfacecoatingandwaterretentioneffectsonfrosting

151

REFERENCES

performanceofmicrochannelheatexchangersforheatpumpsystems,Experimental
ThermalandFluidScience,39,,176188.

Moallem,E.,Padhmanabhan,S.Cremaschi,L.,Fisher,D.E.,2012.Experimental
investigationofthesurfacetemperatureandwaterretentioneffectsonthefrosting
performanceofacompactmicrochannelheatexchangerforheatpumpsystems,Int.J.
Refrigeration,35(1),Pages171186.

MPower,2010.Modine'sCustomVaporCompressionSystemDesign,
http://www.modine.com/v2portal/page/portal/hvac/hvacCoolingCoilsDefault/hvac_co
m/cooling_coils/level_3_content2_040.htm,ModineManufacturingCompany,Racine,
WI,USA,andUniversitatPolitcnicadeValncia,Spain.

Nielsen,K.K.,Engelbrecht,K.,Christensen,D.V.,Jensen,J.B.,SmithA.,Bahl,C.R.H.,
2012.Degradationoftheperformanceofmicrochannelheatexchangersduetoflow
maldistribution,AppliedThermalEngineering,40,236247.

Oliet,C.,Oliva,A.,Castro,J.,PrezSegarra,C.D.,2007a.Parametricstudieson
automotiveradiators,AppliedThermalEngineering,27(1112),20332043

Oliet,C.,PrezSegarra,C.D.,Castro,J.,Oliva,A.,2010.Modellingoffinandtube
evaporatorsconsideringnonuniformintubeheattransfer,InternationalJournalof
ThermalSciences,49(4),692701.

Oliet,C.,PrezSegarra,C.D.,Danov,S.,Oliva,A.,2007b.Numericalsimulationof
dehumidifyingfinandtubeheatexchangers:Semianalyticalmodellingand
experimentalcomparison,InternationalJournalofRefrigeration,30(7),12661277.

Palm,B.,2007.Refrigerationsystemswithminimumchargeofrefrigerant,Applied
ThermalEngineering,27(10),16931701.

Park,C.Y.,Hrnjak,P.,2007.EffectofHeatConductionthroughtheFinsofa
MicrochannelSerpentineGasCoolerofTranscriticalCO2System.Int.J.Refrigeration
30(3),389397.

Patankar,S.V.,1980.NumericalHeatTransferandFluidFlow.Hemisphere,NewYork.

Pettersen,J.,Hafner,A.,Skaugen,G.,1998.Developmentofcompactheat
exchangersforCO2airconditioningsystem,Int.J.Refrigeration,21(3),180
193.

152

 REFERENCES

Qi,Z.,Chen,J.,Radermacher,R.,2009.Investigatingperformanceofnewminichannel
evaporators,AppliedThermalEngineering,29(1718),35613567.

Qu,X.,Shi,J.,Qi,Z.,Chen,J.,2011.Experimentalstudyonfrostingcontrolofmobile
airconditioningsystemwithmicrochannelevaporator.AppliedThermalEngineering,
31(1415),27782786.

RamosAlvarado,B.,Li,P.,Liu,H.,HernandezGuerrero,A.,2011.CFDstudyofliquid
cooledheatsinkswithmicrochannelflowfieldconfigurationsforelectronics,fuelcells,
andconcentratedsolarcells,AppliedThermalEngineering,31(1415),24942507.

Revellin,R.,Mishima,K.,Thome,J.R.,2009.Statusofpredictionmethodsforcritical
heatfluxesinminiandmicrochannels,InternationalJournalofHeatandFluidFlow,
30(5),983992.

Revellin,R.,Thome,J.R.,2007.Adiabatictwophasefrictionalpressuredropsin
microchannels,ExperimentalThermalandFluidScience,31(7),673685.

Shao,L.L.,Yang,L.,Zhang,C.L.,Gu,B.,2009.NumericalModelingofSerpentine
MicrochannelCondensers.Int.J.Refrigeration32(6),11621172.

Shao,L.L.,Yang,L.,Zhang,C.L.,2010.Comparisonofheatpumpperformanceusing
finandtubeandmicrochannelheatexchangersunderfrostconditions,Applied
Energy,87(4),11871197.

Singh,V.,2009.DevelopmentofanAdvancedHeatExchangerModelforSteadyState
andFrostingConditions,DoctoralThesis,UniversityofMaryland,CollegePark,USA.

Singh,V.,AuteV.,Radermacher,R.,2008.NumericalApproachforModelingAirTo
RefrigerantFinAndTubeHeatExchangerwithTubeToTubeHeatTransfer.Int.J.
Refrigeration31(8),14141425.

Singh,V.,AuteV.,Radermacher,2010.InvestigationofEffectofCutFinsonCarbon
DioxideGasCoolerPerformance,HVAC&RReasearch,16(4),513527.

Thome,J.R.,2004.Boilinginmicrochannels:areviewofexperimentandtheory,
InternationalJournalofHeatandFluidFlow,25,128139.

Veje,C.,Sss,J.,2004.TheTranscriticalCO2CycleinLightCommercialRefrigeration
Applications.6thG.LorentzenConf.onNaturalWorkingFluids,Glasgow,Scotland.

Webb,R.L.,1994.PrinciplesofEnhancedHeatTransfer.JohnWileyandSons,New
York.
153

REFERENCES

Xia,Y.,Zhong,Y.,Hrnjak,P.S.,Jacobi,A.M.,2006.Frost,defrost,andrefrostandits
impactontheairsidethermalhydraulicperformanceoflouveredfin,flattubeheat
exchangers,Int.J.Refrigeration,29(7),10661079.

Ye,L.,Tonga,M.W.,Zeng,X.,2009.Designandanalysisofmultipleparallelpass
condensers,Int.J.Refrigeration,32,11531161.

YinJ.M.,BullardC.W.,HrnjakP.S.,2001.R744GasCoolerModelDevelopmentand
Validation.Int.J.Refrigeration24(7),692701.

Zhang,P.,Hrnjak,P.S.,2010.Airsideperformanceofaparallelflowparallelfin(PF2)
heatexchangerinsequentialfrosting,Int.J.Refrigeration,33(6),11181128.

Zhao,C.Y.,LuT.J.,2002.Analysisofmicrochannelheatsinksforelectronicscooling,
InternationalJournalofHeatandMassTransfer,45(24),48574869.

Zhao,Y.,Ohadi,M.M.,Radermacher,R.,2001.MicrochannelHeatExchangerswith
CarbonDioxideReportNo.:ARTI21CR/1002001.

Zhong,Y.,Joardar,A.,Gu,Z.,Park,Y.G.,Jacobi,A.M.,2005.Dynamicdiptestingasa
methodtoassessthecondensatedrainagebehaviorfromtheairsidesurfaceof
compactheatexchangers,ExperimentalThermalandFluidScience,29(8),957970.

Zilio,C.,Brown,J.S.,Schiocheta,G.,Cavallini,A.,2011.TherefrigerantR1234yfinair
conditioningsystems,Energy,36(10),61106120.

Zilio,C.,Cecchinato,L.,Corradi,M.,Schiochet,G.,2007.AnAssessmentofHeat
TransferThroughFinsinaFinandTubeGasCoolerforTranscriticalCarbonDioxide
Cycles.HVAC&RRes.J.13(3),457469.

154