UNIVERSITY OF BOLOGNA
SCHOOLOF ENGINEERING ANDARCHITECTURE
Master of Siene Degree inAerospae Engineering
Testing of subgrid sale (SGS)
models for large-eddy simulation
(LES) of turbulent hannel ow
Master thesis in Applied Aerodynamis
SUPERVISORS:
Prof. Alessandro Talamelli
Prof. Arne V. Johansson
ASSISTANT SUPERVISOR:
Dr. Geert Brethouwer
STUDENT:
Matteo Montehia
Aademi Year 2013-2014
andthe onlywaytobetrulysatised
istodo what youbelieveisgreatwork.
And the onlywaytodo greatwork
istolovewhat youdo.
If youhaven'tfoundityet,keep looking.
Don'tsettle.
As withall mattersof the heart, you'llknow whenyound it.
SteveJobs
Vedi aroamio osatisrivo etidio
e omesonoontento
di essere quiin questomomento,
vedi aroamioosa sideve inventare
per poteri ridere sopra,
Sub-grid sale(SGS) modelsare requiredin order to model theinuene of theunresolved small
sales on the resolved sales in large-eddy simulations (LES), the ow at the smallest sales of
turbulene.
In the following work two SGS models are presented and deeply analyzed in terms of auray
throughseveralLESswithdierentspatialresolutions,i.e. gridspaings.
The rstpart ofthis thesisfouses onthe basitheory ofturbulene, the governingequations of
uiddynamisandtheiradaptationtoLES.Furthermore,twoimportantSGSmodelsarepresented:
oneistheDynamieddy-visositymodel(DEVM),developedby[Germanoetal.,1991℄,whilethe otheristheExpliitAlgebraiSGS model(EASSM),by[Marstorpetal.,2009℄.
In addition, somedetails about the implementation of the EASSM in a Pseudo-Spetral
Navier-Stokesode[Chevalieret al.,2007℄arepresented.
Theperformaneof thetwoaforementionedmodelswillbeinvestigatedin thefollowinghapters,
bymeansofLESofahannel ow,withfritionReynoldsnumbers
Re
τ
= 590
uptoRe
τ
= 5200
, with relatively oarse resolutions. Data from eah simulation will be ompared to baseline DNSdata.
Resultshaveshownthat,in ontrasttotheDEVM, theEASSM haspromising potentialsforow
preditionsathighfritionReynoldsnumbers:thehigherthefritionReynoldsnumberisthebetter
theEASSM willbehaveandtheworsetheperformanesoftheDEVMwillbe.
The betterperformaneof theEASSM is ontributedto theabilityto aptureowanisotropyat
thesmallsalesthroughaorretformulationfortheSGSstresses.
Moreover,aonsiderableredutionintherequiredomputationalresouresanbeahievedusing
the EASSM ompared to DEVM. Therefore, the EASSM ombines auray and omputational
VersioneItaliana
Nelle Simulazioni agrandi vortii (Large-eddy simulations, LES), dei modelli per la sala di
sottogriglia(Sub-grid sale, SGS) sono neessariper riprodurrel'inuenza delle piúpiole sale
dellaturbolenza,quindinonrisolte,suquelleheinveevengonorisoltedirettamente.
Nel seguente lavorosono presentati due modelli SGS, laui auratezzaverrá poi analizzata
at-traversovarieLESadiverserisoluzionispaziali,equindidiversiintervallididierenziazione.
Laprima partedellatesisionentrasullateoriadellaturbolenza,partendodalleequazioni
osti-tutivedellauidodinamia,noallaloroversioneperLES.
Due importanti modelli SGS sono stati presentati: il primo é il modello Dynami eddy-visosity
model(DEVM),di[Germanoetal.,1991℄,ilseondoéilmodelloSGSEspliitoAlgebrio,Expliit AlgebraiSGS model (EASSM),di[Marstorpetal.,2009℄.
Sarannoinoltrefornitidettagliaggiuntivisull'implementazionedelmodelloEASSMsuunodiedi
FluidodinamiaComputazionale(ComputationalFluidDynamis,CFD),Pseudo-Spettrale
Navier-Stokessviluppato nel Linné Flow Centre del Dipartimento di Ingegneria Meania del KTH di
Stoolma,da[Chevalieret al.,2007℄.
I seguentiapitoli verterannosull'analisi dellastimadi unussoin unanale,hannel ow, fatta
daimodellidesrittiinpreedenza,perunbassonumerodiReynoldsbasatosull'attrito,
Re
τ
= 590
, no aRe
τ
= 5200
. I datiottenutida iasunasimulazioneverrannoonfrontationdatidi Simu-lazioni NumeriheDirette (DiretNumerial Simulations, DNS).Dai risultati ottenuti si puó onludere he, dierentemente dal modello DEVM, l'EASSM ha
promettenti potenzialitá nella stima del usso ad alti numeri di Reynolds
Re
τ
: piú alto é tale numero,piúilmodelloEASSMdarárisultatiaurati,mentreleperformanesdelDEVMpeggior-eranno.
LemiglioriperformanedelmodelloEspliitoAlgebriopossonosenz'altroessereattribuiteallasua
abilitádialolareinmanieraorrettal'anisotropiaallepiolesaletramiteunaformulazione
or-rettadeglistressdisottogriglia,SGS.
Inonlusione,datalaridottaquantitádirisorseomputazionalirihiestapereettuaresimulazioni
rispetto alDEVM, tale modelloombina auratezza eeienza omputazionale,tanto he puo'
1 Turbulene 5
1.1 IntrodutiontoTurbulene . . . 5
1.2 NumerialapproahestoN-S solution . . . 12
1.3 Apartiularase: TurbulentChannelFlow . . . 17
2 Large Eddy Simulation 25 2.1 Thelteringoperation . . . 26
2.2 GoverningequationsofLESandthelosure problem . . . 27
2.3 An exampleofalosure: theEddyVisosityModel(EVM) . . . 27
2.4 Subgrid-saledissipation . . . 28
3 Subgrid-salestress modelsfor LES 31 3.1 TensorialpolynomialformulationoftheSGSstresstensor . . . 31
3.2 ThedynamiproedureforEVM(DEVM) . . . 33
3.3 ExpliitAlgebraiSGS stressmodel . . . 34
4 Implementationin a CFD ode 39 4.1 Theneedforauray: spetralmethods. . . 39
5.1 MeanValues . . . 49
5.2 Root-meansquaredValues,rms . . . 50
5.3 Vortialstrutures . . . 51
6 Results 53
6.1 LESat
Re
τ
= 590
. . . 546.2 LESat
Re
τ
= 2000
. . . 646.3 LESat
Re
τ
= 5200
. . . 72Provaad immaginare unpennahio di fumo fuoriusente dauna pipa di un uomopensante,
seduto su diuna sediaadondolo. Anhe in questopioloaspettodellavita quotidiana la
turbu-lenza gioaunruolofondamentale.
Il noto sioRihardFeynman denii laturbulenza ome il piúimportante problema dellasia
lassiaheanoranonéstatorisolto.
Nelorsodegliannidiversisienziatihannoprovatodi omprendereil veroomportamentoditale
fenomeno, la ui omplessitá giae nelle equazioni he lodesrivono. Per unuido Newtoniano,
la turbolenzaé denita dalle equazioni di Navier-Stokes, unsistema di equazioni non-lineari alle
derivateparziali,lauisoluzioneanalitiaanoranonéstataottenuta.
Tuttavia, degli approi alternativi per ottenere delle soluzioni sono stati sviluppati nora: un
metodoonsistenell'eettuaredegliesperimenti: tramitegalleriedelventosiamoingradodi
ripro-durreussiinsvariateondizioni,dallesituazionilassihedistratolimite(motosudiunaparete)
eussoinunanale,noaasipiúompliati, omeilussoattornoadunorpotozzo(omeuna
automobile)eunorpoaerodinamio,omeunaeroplano.
Gliesperimentihannoomeneprinipalel'analisidellequalitádelusso(omepressionee
velo-itá) attraversodelle sondeedelle tenihe di visualizzazione, omela Partile ImageVeloimetry
(PIV).
[Talamellietal.,2009℄nell'Universitádi Bologna.
D'altroanto,unapproiodiversoéquellodisfruttarel'analisinumeriaepassarealla
implemen-tazionedelsistemaN-S suodii diFluidodinamiaComputazionale.
Tuttavia, la soluzione numeria e ompleta fornita dalle Simulazioni Numerihe Dirette (DNS),
in grado di desriverela turbolenza aqualsiasi sala, non ésemprepossibileaausa dellegrandi
risorse omputazionalirihieste,in partiolare adelevati numeridi Reynolds. Perquesto motivo,
l'obiettivoprinipaledelprogettoCICLoPEéquellodiomprendereapienoilfenomenodella
tur-bolenzaqualoranonsiapossibileotteneredatitramiteDNS.
Comunque,latenia DNSnon élasola in gradodi riprodurre numeriamente unpreiso usso.
Una seondapossibilitási hiamala simulazione agrandi vortii (Large-eddysimulation -LES);
questateniaéingradodiraggiungere,onopportunimodelliperlepiolesaledellaturbolenza,
unasoddisfaenteauratezzaomputazionaleonunaragionevolequantitádirisorse
omputazion-ali.
A tale avviso, la tesi verterá sull'impiego e il test di due modelli innovativi per le piole sale
dellaturbolenza;inpartiolareverrámessaaonfrontol'auratezzadiiasunmodelloonl'altro,
rispettoanhe i risultatiforniti da DNS. Lesimulazioni sonostate ompiuteatre diversi numeri
di Reynolds;in partiolarel'ultimoaso,dimaggioreimportanza, érelativoalmassimonumerodi
ReynoldsraggiuntonoradaDNS.
Laprimapartedellatesiomprenderáladesrizionedellaturbolenza,siadaunpuntodivistasio
hematematio. Saranno onsiderateleequazionidiN-S perunuidoNewtoniano,onusso
in-omprimibileeturbolento, inunanale.
NelseondoapitololateniaLESsaráespressaneidettagli,mentrenelterzoverrannopresentati
duemodellidisalasotto-griglia(sub-gridsale-SGS).Ilquartoapitolospiegherádeidettagli
te-niisull'implementazionedellesimulazioni,deiduemodelli,edellaparallelizazioneMPIutilizzata.
Nel quinto apitoloenni di metodi statistii per l'analisi della turbolenza saranno arontati. I
risultativerrannopresentatinelsestoapitolo,einnesarannofornitedelleonlusioninelsettimo
Imagineasmokingplumeomingoutfrom apipeofathinking man,thisisatypialeveryday
life senariowhereturbuleneplaysamain role. Thephysiist RihardFeynmandened itasthe
most important unsolvedproblemof lassial physis.
Sientistsalongtheyearshavetriedto understand therealbehaviourof thisphenomenon, whih
gains itsomplexitybeauseoftheequationsthat desribesit. ForaNewtonian uid,turbulene
istraedbyNavier-Stokesequations,asystemofnon-lineardierentialequationswhoseanalytial
solutionhasnotbeenprovidedyet.
Anyway,alternativeapproaheshasbeendevelopedarosstheyearsuptonow: themostintuitive
andoldwaytounderstanduidmotionistomakeexperiments: throughtheemploymentofwind
tunnelsweanreproduetheowinseveralonditions,fromthebasialhannelowandboundary
layerto themoreomplexones,liketheowarossablu body (likeaar)oranairplane. This
tehniqueinvolves theanalysis of the owproperties(likepressure and veloity) throughprobes
andadvaned visualizationtehniquessuhasPartileImageVeloimetry(PIV).
A reentexamplewhoseaimistodesribehighReynoldsnumberturbulenearossapipeis
nev-erthelesstheCICLoPEexperiment,abigprojetdevelopedby
[Talamellietal.,2009℄inUniversityofBologna.
Theseondapproahisthustousenumerialanalysisinordertoimplementodesableto
mainCICLoPEprojet'saimistounderstandturbulenephenomenafurthertheReynoldsnumber
limitditatedbyDiretNumerialSimulations.
However, DNS is not the only tehnique able to reprodue numeriallya given ow. A possible
seondhoieisalledLarge-eddysimulation;thistehniqueisabletoreahasatisfying
omputa-tional auraywithareasonableamountof omputationalresoures.
InthisthesisNavier-StokesequationsforaNewtonianuidandaninompressibleowina
turbu-lenthannelowsenariowillbedesribedinthersthapter,togetherwithaphysialdesription
ofturbulene. IntheseondhaptertheLarge-eddysimulation(LES)tehniquewill beexplained
in detail,while in the third onetwoLES sub-grid salemodelsare shown. Inhapter foursome
tehnialdetails oftheimplementationof theLES,ofthemodels,with apartiularfousonMPI
parallelizationare pointedout. Sometheory aboutturbulene statistiswill befaedin thefth
1
TURBULENCE
Everybodyisagenius. Butif youjudgeash byitsability tolimb atree,
itwilllive itswholelifebelieving thatitisstupid. (A.Einstein)
1.1 Introdution to Turbulene
Thephenomenonofturbuleneisfoundinseveralappliations, forexample,ombustiontumbling
in Internal CombustionEngines(ICEs),thewakeofaFormula1ar,thejetspreadbythenozzle
ofasupersoniairraftengine.
Inautomotiveengineering,forexample,thestudyofaerodynamisaroundaarinvolvesthe
har-aterizationofaturbulentwall-boundedow,alledaBoundary Layer. Itwasintensivelystudied
byL.Prandtlin1904;hereturbuleneisthemain responsibleforfritionandwakedrag.
In asolid roketmotor nozzle,there's ageneration ofaplume, whereturbulentmotionsofmany
salesanbeobserved;fromeddiesandbulgesomparableinsizetothewidthoftheplumetothe
andhaotinature,sinethemotionofeveryeddy isunpreditable.
Figure1.1Large-eddysimulationofjetfromaretangularnozzle. Theretangularnozzleis
showningraywithanisosurfaeoftemperature(gold)utalongtheenterplaneofthenozzle
showingtemperature ontours (red/yellow). Theaousti eldisvisualizedby(blue/yan)
ontoursofthepressureeldtakenalongthesameplane,fromP.Moin.
Whilelaminarowisa smoothand steadyowmotion,where any induedperturbationsare
dampedoutduetotherelativelystrongvisousfores,inturbulentowsotherforesmaybeating
that ounterattheationofvisosity. Ifsuh foresarelargeenough,theequilibrium oftheow
is upset and theuid annot adapt suddenlyto visosity. Thefores that upset this equilibrium
aninlude buoyany,inertia,oreven rotation. Inahannel, visousandinertialforesatingon
theuidareproportionalto
F
v
∝
νL
(1.1)F
i
∝
V L
2
(1.2)where
ν
istheuidvisosity,andL
andV
aretheharateristiveloityand lengthsales. Ifthevisous foresontheuidarelargeomparedwithothers, anydisturbanesintroduedin the
owwilltend to bedamped out. Ontheother hand,iftheinertialfores beome large,theuid
will tend to break upinto eddies. For greater inertialfores, the eddies will break upinto even
smallereddies. This willontinueuntilwereahasmallenoughlengthsale(eddy size)onwhih
The largest of these eddies will be onstrained by the physial size onstraints on the ow (like
hanneldiameter);thesmallereddieswillbeonstrainedbythevisousforeswhihatstrongest
at the smallest length sales. Therefore, one of the diulties assoiated with the predition of
turbulentowis thattherangeoflengthsalesanbeverylarge.
Thedesriptionofturbuleneinvolvesdierentoneptsliketurbuleneenergyprodution,transfer
anddissipation. Rihardson'sfamouspoemgivesagoodideaaboutturbulene:
Big whorlshave littlewhorls
That feed ontheir veloity
Andlittle whorls havelesserwhorls
Andso ontovisosity (in the moleularsense)
Thisisthedesriptionoftheenergyasade onept. Itstatesthatturbulentowsanbe
on-sideredasanagglomerateofeddies ofdierentsizes. Largeenergyontainingeddiesareunstable
and break down andtransferenergy to smallereddies. The proessgoesontill thesmallestone,
theKolmogorovsale,where energyis dissipatedinto heatbyvisouseets.
This isthegreatonlusion thatKolmogorovmadein 1941,andhis rsthypothesis isompletely
basedonthat: atsuientlyhighReynoldsnumber,thestatistisofthesmallsalesareuniversal
andaredeterminedsolely byvisosity,
ν
,andtheenergydissipationrate,ε
.Usingdimensionalanalysis,itispossibleto derivetheKolmogorovlengthsale
η
,timesalet
η
andveloitysalev
η
:η
=
ν
3
ε
!1
/
4
,
t
η
=
ν
ε
!1
/
2
,
v
η
= (
νε
)
1
/
4
(1.3)Then, in aordaneto Rihardson'spoem, Kolmogorovmade aseond hypothesis, basedon
thefat thatatsuientlyhighReynoldsnumberthestatistisofthesaleswhiharesuiently
larger than
η
and muh smallerthan the largestenergeti sales are solely desribed byε
. Thishypothesisreferstotheinertialrangeofsales. Thekinetienergyspetrumofthesesalesanbe
desribedby
Figure1.2Turbuleneenergyvswavenumberspetrum,fromJ.M.MDonough.
where
k
isthewavenumberandC
k
≈
1
.
5
[Sreenivasan,1995℄istheKolmogorovonstant. Fromthelastdiagram,it'slearthatturbulenehasdierentbehavioursaordingtothewavenum-ber. Basingontheseonepts,weandistinguishfourdierentregions:
1. thelargesale,determinedbytheproblem domaingeometry;
2. theintegralsale
(Λ)
,whih isanO
(1)
fration(oftentakentobe∼
0
.
2
)ofthelargesale;3. the Taylor mirosale whih is an intermediate sale, found in the Kolmogorov's inertial
subrange
(
η
≪
2
π
κ
≪
Λ)
4. theKolmogorov(ordissipation)sale
(
η
)
whihisthesmallestofturbulenesales,theinnersale
1.1.1 3D Nature of Turbulene
Turbulene is rotational and a three-dimensional phenomenon. It is haraterized by large
utuationsinvortiity,whihareresponsibleforvortexstrethingandlengthsaleredution. These
harateristisareidentially zeroin twodimensionsand theseare thereasonswhyturbulene is
hardtodesribebothanalytiallyandnumerially.
Thesethree-dimensionaldynamialmehanismsarehighlyomplexandnonlinear,howevertheow
anbeassumedas bi-dimensionalforlargesale2Dstrutures. These struturesplayadominant
inthesmallersale,wheretheyarefundamentalformixing,mostofallatmoleularsales(e.g. in
ombustionproblems).
1.1.2 Order & Randomness
Despiteturbulene is haoti, itonsists of ompletely randommotionsthat anaggregatein
oherentstrutures. Typialexamplesareturbulentboundarylayersand homogeneousturbulent
shearows,whihexhibithorseshoe,orhairpinvorties(seegure1.3)thatappeartobeinherent harateristis. Freeshearowslikethemixinglayerrevealoherentvortexstruturesverylearly,
againevenforveryhighturbuleneintensities.
The onepts of order and randomness have also led to some new analyti approahes and new
interpretations in the study of turbulene. The names of these disiplines are known as Chaos,
Bifuration Theory, and DynamialSystems [MDonough,2007℄. These theorieshavebeenfaed forthestudyofturbulene,in partiularintheareaofhydrodynamistabilityandtransitionfrom
laminarowtoturbulentone. Cometoattentionofmathematiians,physiistsandengineers,these
phenomenaisofaremarkablenon-linearity,whihmakesturbuleneunpreditableandomplexto
desribe.
Asanonlinearproblem,itanbeseenthatthesolutionstotheseproblemswiththesamenonlinear
equationswithonlyslightdierenesininitialonditions,willrapidlydiverge. Therefore,asuitable
denition ofturbulene must neessarilyinvolveaomplexdynamial systemwith many degrees
offreedom.
1.1.3 The Reynolds number
Turbulene an be seen also a play between inertial fores and visous. Therefore the ratio
betweenthemisruialinorderto haraterizeaow. TheReynoldsnumberplaysthat role,and
in turbulentowsholds
Re
def=
U L
ν
≫
1
.
(1.5)Formany ows of pratialimportane (e.g. a ow on airplanewings) the Reynoldsnumber
anbeontheorderof
Re
∼
10
6
. Thismeansthat thevisousfores,thataremoleularfores,at
in smallersalesthanin thelargeones. However,in anyturbulentowthemoleularvisosityis
alwaysimportantatsomesale. AstheowReynoldsnumberinreases,theregionwhere visous
eetsareremarkable,dereasesinthiknessandtheveloityoftheowhangesveryrapidlyfrom
zeroatthesurfaeto thefree-streamveloityat theouteredgesoftheboundarylayer. Again,we
see thetendenyofthenonlinearinertialtermstogeneratedisontinuitiesat highwavenumbers.
One more,theReynoldsnumberanbeinterpretedalso intermsoflengthandtimesaleratios.
Let'sonsideradutofwidth
L
,withaowveloityU
. Thetimeauidpartile,withtransverseveloity
u
′
takesto rossthedut isalledtheinertial time,
T
i
∼
L/u
′
. Atthesameway,visous
foreshaveatimesale,
T
v
∼
L
2
/ν
.
Inaturbulentow,theinertialtime-salewillbemih lessthanthediusivetime-sale,
T
v
T
in
=
u
′
L
ν
>
1
.
(1.6) 1.1.4 Navier-Stokes equationsTurbulene behaviouris ompletely desribed by the governing equations of uid mehanis,
i.e. the ontinuity and Navier-Stokes(N-S) equations. In ase of inompressible ows, they are
expressedin thefollowingwayonnon-dimensionalform:
∂u
i
∂x
i
= 0;
∂u
i
∂t
+
∂u
i
u
j
∂x
j
=
−
∂x
∂p
i
+
1
Re
∂
2
u
i
∂x
j
∂x
j
(1.7)where
Re
istheharateristiReynoldsnumberoftheow,u
i
, i
= 1
,
2
,
3
aretheveloity om-ponents,p
isthepressure. NotethatEinstein'ssummationonventionisusedhere,wherei
= 1
,
2
,
3
.Together withthemainoweld,sometimes weneedalsoto understandphenomenasthat
Bydenition,thewordpassiverefersto theonditionthattheresultingdensitydierenes areso
small that theeet from the salaron theowis negligible. Soapassivesalaranbe heat or
temperatureinaoworaonentrationofasubstane.
ThereforeisalsopossibletouseNavier-Stokesequations1.7to desribethedevelopmentofa pas-sivesalar,
θ
:∂θ
∂t
+
∂u
i
θ
∂x
j
=
1
ReP r
∂
2
θ
∂x
j
∂x
j
(1.8)where
P r
is the Prandtl number, dened as the ratio of momentum diusivity to thermaldiusivity:
P r
def=
µc
p
1.2 Numerial approahes to N-S solution
Thereareseveralnumerialapproahestosolvethesystem1.7. Themostintuitiveoneisalled DiretNumerial Simulation, DNS,wherebythe governingeuquationsare solvedwithoutmaking
anyassumption, resolvingallthesalesfrom thesmallesttothelargestone. Thereforeitprovides
alltheinformationofaturbulentow,withoutanyapproximations. Sinetheomputationalost
ofDNSsaleswiththeReynoldsnumberis
∼
Re
37
/
14
[ChoiandMoin,2012℄,thisisnotaordable forpratialengineeringanalysesathigh Reynoldsnumber.
Soamorepratialapproahhasbeendeveloped,basedontheReynolds'deomposition[Reynolds,1894℄:
u
i
=
u
i
+
u
′
i
,
θ
=
θ
+
θ
′
(1.9)
wheretheoverlinerepresentstheensembleaveragedquantity,and
u
′
i
andθ
′
aretheveloityand
salaromponentsutuations,respetively.
UsingtheReynoldsdeompositionin equations1.7and1.8,andtakinganensembleaverageofall terms,theReynolds-averagedN-S (RANS)equationsarederived:
∂u
i
∂t
+
∂u
i
u
j
∂x
j
=
−
∂p
∂x
i
+
1
Re
∂
2
u
i
∂x
j
∂x
j
−
∂u
′
i
u
′
j
∂x
j
,
∂u
i
∂x
i
= 0
(1.10)∂θ
∂t
+
∂u
i
θ
∂x
j
=
1
ReP r
∂
2
θ
∂x
j
∂x
j
−
∂u
′
j
θ
′
∂x
j
(1.11)Note that here turbulene is solely desribed by the Reynolds stress tensor
u
′
i
u
′
j
and salar ux vetoru
′
j
θ
. These terms have to beproperlymodelled in order to lose the problem. Eah model involvesapproximationswhih limitsits auray. Therefore, this approah providesonlyanapproximatesimulationofthemeanow.
The rst simple model was developed by Boussinesq in 1877. It is based on an eddy visosity
formulation
u
′
i
u
′
j
−
2
3
Kδ
ij
=
−
2
ν
T
S
ij
,
S
ij
=
1
2
∂u
i
∂x
j
+
∂u
j
∂x
i
!
(1.12) whereK
=
u
′
i
u
′
i
/
2
isthemeanturbulentkinetienergy,S
ij
isthemeanstrain-ratetensor,the symmetripartofthemeanveloitygradienttensor,ν
T
istheeddyvisosityweusetomodelasthe produtofaertainsalewith lengthΛ
of theeddiesandveloityV
. Usingdimensional analysisν
T
∼
Λ
V.
(1.13) Eddyvisosityisthenmodeledaordinglyto someharateristisoftheow.Forinstane, algebrai models(orzero equationmodels) relatelength
Λ
andveloityV
to themeanveloityeldandtheowgeometryharateristislikeveloitygradient,distanetothewall,
thiknessoftheshearlayeret. Thesekindofmodelsworkquitewellforthespeiasethatthey
are designedfor,e.g. attahedboundarylayersanddierenttypesof thin shearlayers. However,
theydon'tgivesatisfatoryresultsforgeneralases.
Better results an be ahieved using one-equation models, they typially solve an additional
transport equation for the turbulent kineti energy,
K
, or the eddy visosity,ν
T
. One-equation models give good results for attahed boundary layers and other thin shear layer ows, but foromplexows. A goodexampleistheSpalart-Allmaras[SpalartandAllmaras,1992℄model (SA), that solvesfor theeddy visosity. This isverysuitablefor aeronautialappliationsand atually
isthestandardmodelforexternalaerodynamisCFDanalysesatBoeing.
Two-equations models solve two transport equations for two quantities that an be used for
determiningthelengthandveloitysaleneededtoomputetheeddyvisosity. Themostommon
are
K
−
ε
andK
−
ω
models,wheretransportequationsfortheturbulentkinetienergy,K
andforthedissipationrate,
ε
orthe turbulenefrequenyω
aresolved. Nowadaystheimplementationofthese modelsin ommerialCFDodes(e.g., ANSYSFluent)presentsadditionalorretionsthat
might be dependent on non-loal quantities suh as thewall distane. Oneimportant and most
reent example is the Menter SST
K
−
ω
model [Menter,1994℄, whih is suitable for separated ows;itis thestandardturbulenemodelusedatAirbus.Despiteeddyvisositytwo-equationsmodelsarestilldominatinginindustrialCFD,there'sabig
de-mandformoreauratepreditionofomplexowsituations,inludingonsetofseparation,highly
urvedows,rapidly rotatingowset. Inthesesituations, eddyvisosityBoussinesq'hypothesis
(1.12)doesnotdesribetherealphysiswell. Aneet ofdeorrelation aused byrotationours at high rotation rates, and generally the alteration of prodution to dissipation ratio is a diret
onsequeneofthat.
This phenomena, for instane, is animportant aspetwhih eddy visositymodels doesnot take
into aount,beausethemodelisinsensitivetosystemrotation.
A better alternative to these models are the Reynolds Stress Models. They solve transport
equationsforeahReynoldsstressomponentsderivedfrom themodelledN-S equations,in order
ompu-tationally moreexpensiveandompliated than theothers. However,those dierentialReynolds
stressmodelsanbesimpliedusingtheweakequilibriumassumptionbyRodi[Rodi,1992℄. Details aboutthiswillbeshownin 2. Thealgebrairelationisimpliitin Reynoldsstresses,butthereare some expliit solutions ([Pope,1975℄; [GatskiandSpeziale,1993℄; [WallinandJohansson,2000℄). These modelsarealledExpliitAlgebrai ReynoldsStress Models(EARSM).
Inpartiular,in theEA model[Wallin andJohansson,2000℄ theowanisotropy(
a
ij
)isdesribed asanexpliitexpressionintermsofthe(normalized)meanstrainandrotationtensorswithaddi-tional salarparameters. This leadsto aomparable omputational eorts, as ompared to eddy
visositytwo-equationmodels.
There'salsoaninterestinganalogyforthesalar
θ
modeling. Taylordevelopedananaloguewaytoformulate theeddydiusivitymodel(EDM)forthemeanturbulentsalarux
u
′
i
θ
′
[Taylor,1915℄u
′
i
θ
′
=
−
D
T
∂θ
∂x
i
,
D
T
=
ν
T
P r
T
(1.14)where
D
T
istheeddy diusivityoeientandP r
T
theturbulentPrandtlnumber.Inananalogouswayto
K
−
ε
model,Nagano&
Kim[Kim,1988℄developedtheK
θ
−
ε
θ
model. Thetimesaleτ
θ
=
K
θ
/ε
θ
,isusedtoomputeD
T
∼
Kτ
θ
.Still, model 1.14 is not ompletely orret. Aording to Bathelor, eddy diusivity assumes an alignmentbetweenthe
u
′
i
θ
′
vetorandthemeansalargradient,soithasto beonsidereditselfa
tensor. Thefollowingexpressionwillholdthen:
u
′
i
θ
′
=
−
D
ij
∂θ
∂x
j
(1.15)
Wheretheeddydiusivitytensoranberewrittenas[DalyandHarlow,1970℄
D
ij
=
−
C
θ
τ
θ
u
′
i
u
′
j
∂θ
∂x
j
(1.16)and
C
θ
isamodeloeient.Inthesameway, salaranbemodelled withanExpliit Algebraimodel,in thisasealled the
Expliit AlgebraiSalarFluxModels(EASFM).
Atrade-obetweenaurayandomputationaleortisLarge-eddysimulations(LES)ofturbulent
ows. InLES there'saseparationofsales, inthesense that onlylarge-saleeddiesareresolved,
while the remaining small sales (whih are alled sub-grid sales, SGS) are modelled, one the
resolvedsales havebeenomputed. Theseparationof salesis generallydoneusing agrid. T
of the Navier-Stokesequations. This is thereforeaphysially-oherentapproah sineturbulene
isunstationary.
DespitemoreomputationallyexpensivethanRANS,LESgivesabetterdesriptionofturbulene,
and unlikeDNS,is abletoprovideagoodresolution oftheowin anaeptableamountof
om-putational time. Reently, that timehas been estimatedby Choi&Moin[ChoiandMoin,2012℄, to saleas
∼
Re
26
/
14
.
AfairandsimplieddistintionbetweenDNSandLESanbenotiedhavingalooktothevortial
strutures, for bothof the ases,in gure 1.4. Ata rstglane, we ansee that in LES vortial strutures are underestimatedand fewer,ompared to the DNS,whih is able,instead, to givea
ompleteanddetailed desriptionofthem.
Therst LESmodelwasdevelopedbySmagorinsky[Smagorinsky,1963℄formeteorologial appli-ationsusinganeddyvisosityassumptionintheSGSmodel. Thismodelhasbeenimprovedlater
on by Germano [Germanoetal.,1991℄ introduing the dynami proedure, whih givesa orret asymptotinear-wallbehaviouroftheeddyvisosity,andimprovedtransitionalowspreditions.
However, eddy-visosity models are not anisotropi, that is, they are not able to apture ow
anisotropywell,afeatureoftheow,whih isnotnegletablenearthewalls.
Forthat reasonseveralnon-linear models (whih are ofourse anisotropi),have been developed
reently,and someof themwill bedesribed in detailin hapter 3. In partiular,the aimof this thesis is to test the auray in ow predition of the Expliit Algebrai SGS model (EASSM),
a)
0
0.2
0.4
0.6
0.8
vel_u
-0.0112
0.864
b)0
0.2
0.4
0.6
0.8
vel_u
-0.0112
0.864
Figure 1.4 Vortial strutures inturbulent hannel ow at
Re
τ
= 590
, visualized by isosurfaes ofλ2
, olored by the veloity magnitude, froma) DNS by P.Shlatter b) LES simulationwiththeExpliitAlgebraimodel.1.3 A partiular ase: Turbulent Channel Flow
Mathematialmodelsforuiddynamishavebeenalreadydened. Inordertosolvethemodels,
amathematialproblemneedsto bedened inapropertimeandspaedomains.
Aordingtothespaedomain,inadierentgeometry,physialquantitiesandtheowwillbehave
inadierentway. Forthisreasonweneedtosetapartiularase,sothattheowanbeunivoally
lassied.
InthisthesisweareonsideringtheaseofChannelFlow,withthefollowingproperties:
turbulent,in thesense that theReynoldsnumberis suientlyhigh suhthat the regime
anbeassumedas turbulent;
fully developed, sothat veloitystatistis areonstant along
x
-axis. Inother words,theow is statistially stationary and statistially one-dimensional [Pope,2000℄, with veloity statistisonlyvariablealongthe
y
-axis. Inotherwords:h
u
i
=
U
=
U
(
y
)
,
h
v
i
=
V
= 0
,
u
′
v
′
=
u
′
v
′
(
y
)
(1.17)
Channelowbelongsto the wall-boundedshear ows lass: owmotionis ontainedbetween
twosolidsurfaes. Therefore,no-slip onditions areimposedonthewalls,wheretheuidveloity
isassumedto bezero. Thefollowingpitureshowsthequalitativebehaviouroftheveloity.
Figure1.5ChannelFlowmeanveloityprole,with
u
′
utuationsontour plotinthe
Furthermore,astatistiallysymmetriowgeometryw.r.t. themid-plane
y
=
δ
isonrmedbyexperiments;thereforethestatistisof
(
u, v, w
)
aty
arethesameasthoseof(
u,
−
v, w
)
at2
δ
−
y
.Reynolds number is always used to haraterize the ow, in this ase we will refer to two
partiularReynoldsnumbers,
Re
τ
def=
u
τ
δ
ν
(1.18)Re
b
def=
uδ
ν
(1.19)where1.18is basedonthefritionveloity
u
τ
,denedasfollows:u
τ
def
=
p
τ
W
/ρ
(1.20)τ
W
isthemeanwallshearstressandδ
isthehannelhalf-width.Thebulkveloity
u
in 1.19,isdened asu
=
1
2
δ
2
δ
Z
0
h
u
i
dy
(1.21)Foraturbulenthannelow,thefollowingresultholds:
Re
τ
≈
0
.
166
Re
0
b
.
88
(1.22)Notethat thespeiedformula 1.22willbeused forthederivation of
Re
b
, whih will be on-sideredasaninputquantityintheomputations.As previouslystated, hannel ow is a wall-bounded shear ow. In a boundary layer ora
wall-bounded shear ow theharateristilengthfor streamwisedevelopmentis muhlarger thanthe
ross-streamextentoftheregionwithsigniantveloityvariation.
So its behaviour an be observed studying a two-dimensional steady owsenario with the thin
shearlayerapproximation [JohanssonandWallin, 2012℄. Thisapproximationstatesthatthe har-aterististreamwisedevelopmentlength,
L
ismuhlargerthanδ
,theshearlayerthikness. Aswewill onrm later,in hannelowthere willbealayer,whose thikness issmall omparedto the
harateristilength,wherebyvisouseets fritiondependson.
Thankstothisapproximation,theNavier-Stokesequationsanbesimpliedtothethinshearlayer
U
∂U
∂x
+
V
∂U
∂y
=
−
1
ρ
dP
0
dx
+
∂
∂y
ν
∂U
∂y
−
u
′
v
′
!
|
{z
}
total shear stress
(1.23)
where
P
0
if thepressureat thewall. Notethat allthe veloities in 1.23aremean values,and theonlyomponentresponsibleforturbuleneisthelastoneontheright-hand-side,whihisalledtheReynoldsstress. Togetherwiththevisousstress,theReynoldsstressgeneratesthetotalshear
stress.
In partiular, usingthe hannel owassumptions given in 1.17, the thin-shearlayerequation be-omes
0 =
−
1
ρ
dP
dx
0
+
d
dy
ν
dU
dy
−
u
′
v
′
!
(1.24)Integratedinthewall-normaldiretionitreads
0 =
−
1
ρ
dP
dx
0
y
+
ν
dU
dy
−
ν
dU
dy
y
=0
| {z }
u
2
τ
−
u
′
v
′
+ 0
(1.25)Attheenterline
(
y
=
δ
)
thetotalshearstressiszero,thereforewehavethefollowingondition:1
ρ
dP
0
dx
=
−
u
2
τ
h
(1.26)meaningthatthepressuregradientisrelatedtothefritionveloityandthewidthofthehannel.
Plugging this relation in 1.25, we ansee that thetotal shear stress developslinearly arossthe hannel:
ν
dU
dy
−
u
′
v
′
=
u
2
τ
1
−
y
δ
!
(1.27)whih inwallunitsbeomes
dU
+
dy
+
−
u
′
v
′
+
=
1
−
y
+
δ
+
!
(1.28)wherethequantities
y
+
and
δ
+
aresaled bytheinner(visous)lengthsale
y
+
def=
y
l
∗
=
yu
τ
ν
Considerations. Dependingontheregiononsidered,equation1.28assumesdierentforms. Intheouterlayer,wherevisous eetsarenegligible,theleft-hand-sideof1.28beomes
−
u
′
v
′
+
= 1
−
y
+
δ
+
(1.29)Ontheotherhand,loseto thewall,
y/δ <<
1
,visous eetswill benotnegligibleanymore,thereforethefollowingrelationholds
dU
+
dy
+
−
u
′
v
′
+
≈
1
(1.30)thatisompatiblewiththelawofthewall,sinethere'snotinuene oftheReynoldsnumber.
It'saonstantstress region:
U
+
def=
U
u
τ
= Φ
1
(
y
+
)
(1.31)u
′
v
′
+
= Φ
2
(
y
+
)
(1.32)Moreover,for largeReynoldsnumbers,weanalsoassume that thereis an overlap region for
walldistanes
y
,ℓ
∗
<< y << δ
where
δ
isboundarylayerthiknessandℓ
∗
thevisous lengthsale. This isapartiularregion where outerandinnerlayerdesriptionshold simultaneously.Derivatingthefollowing1.31w.r.t.
y
+
we'llhaveanexpressionwhihisindependentoflengthsale.
Thereforeholds
y
+
d
Φ
1
(
y
+
)
dy
+
≡
const
(1.33)sothat,oneintegrated,itgivesalogarithmi law:
Φ(
y
+
) =
1
κ
lny
+
+
B
(1.34)
where
κ
= 0
.
38
istheKármánonstantandB=4.1,aordingtoobservations. Inthefollowinggure6.22ameanveloitydiagramisshown,togetherwiththelawof the wall andtheloglaw, at
Re
τ
= 5200
.PSfragreplaements
y
+
h
u
i
+
10
0−
2
10
−
1
10
0
10
1
10
2
10
3
10
4
5 10 15 20 25 30 Figure1.6:< u >
+
vsy
+
,at
Re
τ
= 5200
,from[LeeandMoser,2014℄,.:lawofthe wall,:loglawAsweanseefromthepreviouspiture,therstveloityregionalledvisoussublayer,follows
thelawofthewall,anditextendsoutto approximately
y
+
= 5
. Despite theabsolutemagnitude
of the turbulent utuations are small in this region, the relative (wall-parallel) intensities are
large. Asweinrease
y
+
,wewillhaveabuer region, wherethemaximumturbuleneprodution
is at
y
+
= 12
and the maximum turbulene intensity at
y
+
= 15
. Log-layer starts between
50
< y
+
<
200
, andit extendsto
y/δ
≈
0
.
15
, whereδ
isthehannel half-width. Beyondthe loglayer,there'snallytheouterregion.
The maximumturbulene prodution and Reynolds stress.
Theturbuleneprodutioninthenear-wallregionofwall-bounded owsanbeformulatedas
P
+
=
−
u
′
v
′
u
2
τ
dU
+
dy
+
≃
1
−
dU
+
dy
+
!
dU
+
dy
+
(1.35)Thelatterapproximationis validifvisouseetsare small. Fromthis weanderivethatthe
maximumprodutionisfoundwhere
dU
+
dy
+
=
1
2
(1.36) whih leadstoP
max
+
=
1
4
(1.37)ThereforetheturbulentprodutionismaximumwhereasbothvisousandReynoldsstressesare
exatlythesame,thatisinthenear-wallregion;itgenerallyhappenswhen
y
+
≈
12
alsopossibletoestimatewherethemaximumReynoldsshearstressours. Derivingequation1.28
w.r.t.
y
+
, and denotingthe Reynoldsshear stress as
τ
, normalized by inner units, the followingrelationholds:
d
2
U
+
dy
+
2
+
dτ
+
dy
+
=
−
1
δ
+
(1.38)ifwealsoassumethattheReynoldsstressismaximumin thelog-region,taking thederivative
1.34w.r.t
y
+
wehaved
2
U
+
dy
+
2
=
−
1
κy
+
2
(1.39)meaningthat themaximumoftheReynoldsstressisat
y
max
+
=
r
h
+
κ
=
r
Re
τ
κ
(1.40)Then,intermsofoutersaleweansaythatthepositionwheretheReynoldsstressreahesits
maximumisproportionalto
Re
−
1
/
2
τ
:y
max
h
=
κ
−
1
/
2
Re
−
1
/
2
τ
∝
Re
−
1
/
2
τ
(1.41)ThisalsoanbeprovedusingDNSresultsatdierentfritionReynoldsnumbers. Anexample
ofReynoldsstressprolesis showningure1.7. PSfragreplaements
y/δ
−
u
′
v
′
u
2
τ
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 Figure1.7−
u
′
v
′
normalizedby
u
τ
vsy/δ
,from[LeeandMoser,2014℄,· · ·
:Re
τ
= 550
,Themaximum Reynoldsshear stress anbe derived aswell, using the relation1.38and 1.40, togetherwiththelog-law:
τ
max
+
= 1
−
y
+
max
h
+
−
1
κy
max
+
= 1
−
√
2
κh
+
= 1
−
2
√
κRe
τ
.
(1.42)Wealreadyhavetalkedaboutturbuleneanditsprodutioninaqualitativesense,but,inorder
todesribeitproperlyandtounderstandthephenomenoninaompleteanddetailedsense,afous
on the three omponents of theveloity, theyare
u
′
,
v
′
and
w
′
, is neessary. As we'll see in the
nexthapters,theomputationsoftheseomponentsplayanimportantrolein thevalidation ofa
LESmodel.
Inpartiular,it'spossibletoderivetheReynoldsstresstensormakingthesquareofthe
root-mean-squarevalueofthesethree omponents:
u
′
i
u
′
j
=
u
′
u
′
u
′
v
′
u
′
w
′
u
′
v
′
v
′
v
′
v
′
w
′
u
′
w
′
v
′
w
′
w
′
w
′
.
(1.43)InthefollowingpiturealltheomponentsoftheReynoldsstressareshown. Importantto
un-derlineistherelationbetweenthemaximumturbuleneprodution,andthepeaksintheReynolds
stressproles. PSfragreplaements
y
+
0 5 10 15 20 25 30 35 40 45 50 0 1 2 3 4 5 6 7 8 9Figure1.8Varianesofthe veloityutuations, normalizedinwall units,fromDNS of
[LeeandMoser,2014℄at
Re
τ
= 1000
,:u
′
u
′
+
,:v
′
v
′
+
,:w
′
w
′
+
'Anisotropy of the ow. A fundamental aspet, whih is stronglyrelevantfor theomplete
understandingoftheaforementionedwork,is theoneptofanisotropyin turbulene.
Fromaphysialpointof view,anisotropyis aproperty of theowwhih is notalignedwith the
veloitydiretion andtheveloitygradient.
Forafully-developed hannelowthequantities
u
′
w
′
and
v
′
w
′
= 0
arezero. TheReynoldsstress
tensorbeomes:
τ
ij
=
−
ρ
τ
11
u
′
v
′
0
u
′
v
′
τ
22
0
0
0
τ
33
.
(1.44)in ase ofisotropi turbulene, the diagonal termsare equal, i.e.
τ
11
=
τ
22
=
τ
33
andall the deviatoritermsarezero. Therefore,foraspeiLESmodel,theabilitytoapturetheanisotropyoftheowonsistsin reproduingdierentdiagonalterms,i.e.
u
′
u
′
6
=
v
′
v
′
6
=
w
′
w
′
2
LARGE EDDY SIMULATION
Anxietyisthe handmaiden of reativity. (T.S.Eliot)
Reynolds-averaged Navier Stokes (RANS) equations-based simulations are able to solve only
themeanveloityeld oftheow. RANSsimulationsrelyheavilyonmodellingsineallturbulent
motionsaremodelledandthereforetheyarenotalwaysaurate.
InDiretNumerialSimulations(DNS), theunsteadyNavier-Stokesequationsaresolvedwithout
using models. Therefore, DNS is very aurate. However, the omputational powerdemand for
DNSistoolargeforindustrialappliations. Therefore,anewmethodwhihombinesareasonable
owpreditionauray withalimitedamount ofomputationalost hasbeenstronglyrequired
in reentyears.
Large-eddysimulation(LES)representsanalternativethattsthoserequirements. Thispartiular
methodemploysaseparationofsales[Rasam,2014℄: alteringoperationdeomposestheveloity eld (generallytogether withasalarone) intoaresolvedpart,representedbytheomputational
grid,andanunresolvedpart,whihismodelledthroughphysiallyrealistimodels. Inotherwords,
1. thesolutionofthelargesalesofturbuleneonarelativelyoarsegrid;
2. themodelling ofthe smallerunresolvedsales, the so-alledsubgrid-sales (SGS), basedon
theresolvedveloityeld.
2.1 The ltering operation
Step1. Let's onsidernowageneraltimeand spae-dependentfuntion
φ
(
x, t
)
. Thelteringoperationonsistsofaonvolutionofakernel
G
∆
onthatfuntion(Leonard,1975),overthedomainD
ofthegrid:e
φ
(
x, t
) =
Z
D
φ
(
x, t
)
G
∆
(
x
−
ξ
)
dξ
(2.1)There are several optionsfor the lter: the ommonly used ones are spetral uto, box and
Gaussianlters. Inspetralodes,spetralltersarethemostsuitableonesforltering,sinethey
at onaspetralspae(where theodeworkson)and aremorepreise. Figure below showsthe
dierenebetweenthedierentkindsoflteringmethods.
Figure2.1.: Sharp-spetrallter,-: Gaussianlter, : Boxlter,
r
=
ξ
,from [Pope,2000℄eld
e
u
(
x, t
)
is notadeterministivariable, implyingthatee
u
(
x, t
)
6
=
u
e
(
x, t
)
,
u
e
′
(
x, t
)
6
= 0
.
2.2 Governing equations of LES and the losure problem
Thelteringoperationleadsto thelteredNavier-Stokesequations:
∂
e
u
i
∂x
i
= 0;
∂
e
u
i
∂t
+
∂
e
u
i
u
e
j
∂x
j
=
−
∂x
∂
p
e
i
+
1
Re
∂
2
e
u
i
∂x
j
∂x
j
−
∂τ
SGS
ij
∂x
j
(2.2)here
τ
ij
isthe SGS stress tensor. Leonard[Leonard,1974℄ proposed a possibledeomposition ofnon-linearterms(i.e. theSGSstresstensor),inthefollowingway:τ
SGS
ij
=
u
g
i
u
j
−
e
u
i
u
e
j
=
L
ij
+
C
ij
+
R
ij
(2.3) whereL
ij
=
u
e
g
i
u
e
j
−
u
e
i
e
u
j
,
(2.4)C
ij
=
e
u
g
i
u
j
′
−
u
e
g
j
u
′
i
(2.5)R
ij
=
u
g
′
i
u
′
j
,
(2.6) andu
′
i
=
u
i
−
u
e
i
.In the Leonard deomposition
R
ij
is the Reynoldssubgrid tensor and represents the interation betweensubgrid-sales;C
ij
is theross-stresstensor andaountsforlargevssmall sale intera-tions;nallyL
ij
,theLeonardtensor,givestheinterationsbetweenthelargesales.τ
SGS
ij
is theunknown additivepartin theltered Navier-Stokesequation, and thereforeneeds to bemodelled.2.3 An exampleofa losure: theEddy Visosity Model (EVM)
Step2. Thereareseveralwaystomodelthesubgridsalestresstensor. Aspreviouslywritten,
the rstandsimplemodel wasdevelopedby Smagorinsky[Smagorinsky,1963℄, formeteorologial appliations. ItoriginatesfromtheRANSmodel,takingintoaountBoussinesq'shypothesis. The
EVMonsistsofalinearformulationofthedeviatoripartof
τ
SGS
ij
,τ
ij
SGS
=
τ
kk
3
δ
ij
| {z }
isotropic
−
2
ν
SGS
| {z }
EV M
contribution
e
S
ij
(2.7) whereν
SGS
= (
C
s
∆)
e
2
|
S
e
|
(2.8)e
S
ij
and|
S
e
|
are theresolved strain-ratetensorand its magnitude, respetively.∆
e
is thelter sale,ν
SGS
istheSGSeddyvisosity,andC
s
isthemodeloeient,theSmagorinskyoeient.Notethat
ν
SGS
isaonstantratherthandependentofdiretion. Aordingtothereentpaper of Spalart [Spalart,2015℄, in a simple shear ow with two of the axes aligned with the veloity diretion and thegradientdiretion,suhthat thestraintensorS
ij
havezeroandequaldiagonal terms, this model predits onstant and equal diagonal terms of the Reynolds stress tensor, i.e.τ
11
=
τ
22
=
τ
33
.ThereforeEVManbeonsideredanisotropi model,sine
ν
SGS
doesn'ttakeintoaountany eet ofanisotropy.As we'll disuss further, this model an be improved using a dynami proedure, where the
C
s
oeientisomputedduringthesimulation.2.4 Subgrid-sale dissipation
Animportantaspet ofLESistheimpatofSGS ontheresolvedsales. Inotherwords,LES
preditionsarestronglydependentontheontributionof
τ
SGS
ij
tensorontheresolvedkinetienergyK
=
u
e
i
u
e
j
/
2
. Considerthekinetienergyequation∂K
∂t
+
∂
∂x
j
(
e
u
j
K
)
|
{z
}
advection
=
−
ν
∂
e
u
i
∂x
j
∂
u
e
i
∂x
j
|
{z
}
viscous
dissipation
−
∂x
∂
i
u
e
i
p
e
+
ν
∂K
∂x
i
−
u
e
i
τ
ij
!
|
{z
}
dif f usion
+
τ
ij
SGS
S
e
ij
| {z }
−
SGS
dissipation
(2.9)SineLESbydenition,arearriedoutforlargesales,andthegridsaleismuhlargerthan
Kolmogorovone,the visous dissipationterm is sosmall so that it is negligibleompared to the
others. Fromaphysialpointofview,diusiontermtransfersenergyinspae,butnotina
volume-averagedsense [Rasam,2014℄. Forthatreason,anadditionaltermisthereforerequiredinorderto reproduetheorretenergy transferfrom thelargeto thesmallersales. SGS dissipationovers
Π =
−
τ
SGS
ij
S
e
ij
(2.10)ThemeanSGS dissipationbehavesasasink term, whiletheinstantaneousonegivesnegative
(baksatter)andpositive(forwardsatter)ontributionsinthetransfer.
The following piture showsthe SGS dissipation (red line), together with the dissipation of the
resolvedsales(blue line);theirtotalontributionisshownbytheredline.
PSfragreplaements
y
+
0 -100 -200 -300 -400 -500 -600 -0.25 -0.2 -0.15 -0.1 -0.05 0 0Figure2.2Dissipationoftheresolvedsales,
ε
,togetherwiththeSGSdissipation,Π
,and theirsumatinreasingresolutions,fromLESatRe
τ
= 590
showninhapter6.ThepreviousonsiderationleadstothefatthataninvestigationonSGSdissipationbehaviour
isveryimportantinanaurayassessmentofLES.
Basedonthispriniple,severalinvestigationshavebeenarriedoutbyChow&Moin[ChowandMoin,2003℄, Ghosal[Ghosal, 1996℄andKravhenko&Moin[Kravhenkoetal.,1996℄.
Geurts &Fröhlih [Geurtsand Fröhlih, 2002℄introdued theSGS ativityparameter, dened as follows
s
=
<ε
SGS
>
<ε
SGS
>
+
<ε
µ
>
(2.11) where<ε
µ
>
= 2
µ
S
e
ij
S
e
ij
isthevisousdissipation.With inreasingresolution, SGS dissipation dereaseswhile visous oneinreases, therefore
s
be-omes smaller. A remarkable aspet isthat, theoarser is theresolution , the biggeris
s
. Thisan be proved having alook at 2.11: if we use aoarse grid, the resolution will be smallersuh that visous dissipationwill benegligible, and
s
will reah itsmaximumunity value. This ourssmall.
Starting from the previous denition, Geurts & Fröhlih [GeurtsandFröhlih,2002℄ dened an errornormas
δ
E
=
E
LES
−
E
e
DN S
e
E
DN S
(2.12)where
E
LES
isthemeanresolvedkinetienergyintegratedovertheowdomain,whileE
e
DN S
is the samequantity, omputed from ltered DNS datainstead, with the samelter width asintheLES.
The auray assessment onsists of omputing the relation between
δ
E
ands
: in the piture below[Rasametal.,2011℄ it is shown that therelativeerrorδ
E
drops almost exponentiallywith dereasings
(i.e. high resolution); results referto a LES with the expliit algebraiSGS stressmodel (EASSM) [Marstorpetal.,2009℄ at
Re
τ
= 934
, for six dierent resolutions. This result showsthat with inreasing resolution theSGS ontribution beomessmaller and theauray ofthe LES higher. Inother words,the resolution of the LES mustbe suiently high to obtain an
aeptable solution.
Figure2.3Normalizederror
δ
E
oftheresolvedkinetienergy,integratedoverthehannel width[Rasametal.,2011℄w.r.t. the ltered DNSvalue vss
. Arrow points toinreasing resolutionases.3
SUBGRID-SCALE STRESS MODELS
FOR LES
Logi willgetyoufrom AtoB.
Imagination willtake youeverywhere. (A.Einstein)
InthishaptertwodierentLESmodelsaregoingtobeanalyzed;therstoneisanimprovement
ofthelassiSmagorinskymodel, theDynami EVM.
Theseondoneisapartiularnon-linearmodelusinganexpliitalgebraiformulationfortheSGS
stresstensor.
Inthefollowingsetionadeomposition oftheSGS stresstensorisgiven.
3.1 Tensorialpolynomial formulation of the SGS stresstensor
A usefulapproah that leadsto several non-linear models is basedon adeomposition of the
SimilartoRANSapproahforReynoldsstresslosures,SGSstresstensoranbeexpressedinterms
ofstrainandrotationratetensors.
Lund &Novikov[LundandNovikov,1993℄ expressedthedeviatoripartof theSGS stresstensor
τ
ij
SGS,d
asageneraltensorialfuntion of theltered strain-rateS
e
ij
androtation-rateΩ
e
ij
tensors, theKronekerdeltaδ
ij
andtheltersize∆
e
asτ
ij
SGS,d
=
τ
ij
SGS
−
1
3
τ
kk
δ
ij
=
f
(
S
e
ij
,
Ω
e
ij
, δ
ij
,
∆)
e
(3.1) wheree
S
ij
=
1
2
∂
e
u
i
∂x
j
+
∂
e
u
j
∂x
i
!
,
Ω
e
ij
=
1
2
∂
e
u
i
∂x
j
−
∂
e
u
j
∂x
i
!
.
(3.2)AsPope[Pope,1975℄showedtheReynoldsstressesinRANS,
τ
ij
anbeformulatedasatensor polynomialwithtenelementsofdierentspowersofS
e
ij
andΩ
e
ij
andtheirombination.Coeientsarefuntions of
S
e
ij
andΩ
e
ij
invariantsorboth,andtheyarederivedusingthe Cayley-Hamiltontheorem:τ
d
=
10
X
k
=1
β
k
T
(
k
)
,
(3.3)sothatthetenpolynomialtensorsare:
T
(1)
=
S
e
T
(2)
=
S
e
2
−
1
3
II
S
I
T
(3)
=
Ω
e
2
−
1
3
II
Ω
I
,
T
(4)
=
S
e
Ω
e
−
Ω
e
S
e
T
(5)
=
S
e
2
Ω
e
−
Ω
e
S
e
2
T
(6)
=
S
e
Ω
e
2
+
Ω
e
2
S
e
−
2
3
IV
I
T
(7)
=
S
e
2
Ω
e
2
+
Ω
e
2
S
e
2
−
2
3
V
I
T
(8)
=
S
e
Ω
e
S
e
2
−
e
S
2
Ω
e
S
e
T
(9)
=
Ω
e
e
S
Ω
e
2
−
Ω
e
2
e
S
Ω
e
T
(10)
=
Ω
e
e
S
2
Ω
e
2
−
Ω
e
2
S
e
2
Ω
e
and
β
k
aresalaroeientsthat arefuntionsofvetensorialinvariantsII
S
=
tr
(
S
e
2
)
,
II
Ω
=
tr
(
Ω
e
2
)
,
V
=
tr
(
e
S
2
Ω
e
2
)
.
Thus, equation3.3isthemostgenerlaformulationfor
τ
d
in termsof
S
˜
ij
andΩ
˜
ij
.3.2 The dynami proedure for EVM (DEVM)
Regardingto themodeloeient
C
s
forthe eddy-visosity model, it hasbeendemonstrated that adynami omputation ofC
s
(that is, during the simulation),an signiantly improvethe owpreditionauray.The omputation of
C
s
is done by taking into aount the resolved sales, aordingto a sale invarianeassumption.Theruialpointofthisdynamiproedure istheso-alledGermanoidentity. Let'snowdenotea
testlter,
∆ = 2
b
∆
e
,∆ =
e
3
√
Ω
,whereΩ
isthevolumeofaomputationalell. TheGermanoidentity isdened asfollows:L
ij
=
T
ij
−
b
τ
ij
,
(3.4)where
T
ij
isthe SGSstress tensorlteredat thetest lterlevel,andL
ij
=
e
u
d
i
e
u
j
−
be
u
i
be
u
j
ThenL
ij
isappliedin thisway:L
ij
−
1
3
L
kk
δ
ij
=
−
2
C
s
M
ij
(3.5)where
M
ij
=
∆
b
2
|
S
be
|
S
be
ij
−
∆
e
2
|
\
S
e
|
S
e
ij
.
(3.6) Thesystemofequations3.5isover-determined. InordertohaveauniquevalueofC
s
Germano ontrateditin:C
s
=
−
1
2
h
L
ij
S
e
ij
i
h
S
e
ij
M
ij
i
(3.7)Moreover,tomake
C
s
variationssmoother,aspatial averagingh
.
i
hasbeenapplied.Togetherwithabetterperformaneof theEVM,theGermanoidentitygivesaorretasymptoti
near-wallbehaviourfor
ν
SGS
.Ithasalsobeenshownthatispossibleto applyGermanoidentityforthedynamiomputationof
3.3 Expliit Algebrai SGS stress model
Anisotropi eets of turbulene are important in several onditions: examplesare near-wall
owbehaviourandboundarylayesseparationaseswithurvature,swirl,rotation.
Asitwasdisussedbefore,theDynamiSmagorinskymodelisanisotropimodel,inthesensethat
theSGSvisosity
ν
SGS
isdiretion-independent.Therefore,in ordertoaptureowanisotropy,DEVM isnotsuitable.
InthesamespiritasReynoldsstress-basedmodelsareneessaryforRANS,non-linearSGSmodels
areneededforLES.
A reent example of those is the nonlinear dynami SGS stress model by Wang & Bergstrom
[WangandBergstrom,2005℄, whihonsists of three base tensorsand three dynami oeients.
OneofthetermsinthemodelissimilartotheDEVMmodel. Wang&Bergstrom[Wangand Bergstrom,2005℄ showedthatthedynaminon-linearmodelpreditsamorerealistitensorialalignmentoftheSGS
stressthaneddy-visositymodelsandanprovideforbaksatterwithoutlipping oraveragingof
thedynamimodelparameters.
The model here disussed is alled Expliit Algebrai SGS stress model (EASSM), was
devel-oped by Marstorp [Marstorpetal.,2009℄ and is similar to the EARSM by Wallin & Johansson [WallinandJohansson,2000℄, whih is based on a modelled transport equation of the Reynolds stressesandontheassumption thattheadvetionanddiusionof theReynoldsstressanisotropy
arenegligible.
AnalogoustotheReynoldsstressanisotropytensorwedenetheSGSstressanisotropytensoras
a
ij
=
τ
ij
K
SGS
−
2
3
δ
ij
(3.8)Forsimpliity,we'llonsider
τ
SGS
ij
=
τ
ij
. Moreover,K
SGS
= (
u
g
i
u
i
−
u
e
i
e
u
i
)
/
2
istheSGSkineti energy.Inanon-rotatingframethetransportequationfor
a
ij
readsK
SGS
Da
ij
Dt
|
{z
}
advection
−
∂D
τ
ij
ijk
∂x
k
−
τ
ij
K
SGS
∂D
K
SGS
k
∂x
k
!
|
{z
}
dif f usion
=
−
K
τ
ij
SGS
(
P −
ε
) +
P
ij
−
ε
ij
+ Π
ij
,
(3.9) where−
D
K
SGS
k
=
−
D
τ
ij
ijk
/
2
is thesumoftheturbulent andmoleularuxes oftheSGS stress andSGS kinetienergy,respetively.Although the prodution of the SGS stress
P
ij
and SGS kineti energyP
=
P
ii
/
2
are given in termsofτ
ij
andlteredgradients,theSGSpressurestrainΠ
ij
andSGSdissipationratetensorε
ij
needto bemodelled. Thosetermsare
P
ij
=
−
τ
ik
∂
u
e
j
∂x
k
−
τ
jk
∂
u
e
i
∂x
k
=
K
SGS
"
−
4
3
S
e
ij
−
(
a
ik
S
e
kj
+
S
e
ik
a
kj
) + (
a
ik
Ω
e
kj
−
Ω
e
ik
a
kj
)
#
,
(3.10)Π
ij
=
2
ρ
(
S
g
ij
p
−
S
e
ij
p
e
)
,
(3.11)ε
ij
= 2
ν
^
∂u
i
∂x
k
∂u
j
∂x
k
−
∂
u
e
i
∂x
k
∂
e
u
j
∂x
k
!
(3.12)andtheirmodellingleadsto [Launderetal.,1975℄
Π
ij
=
−
εc
1
a
ij
+
K
SGS
"
3
5
S
e
ij
+
9
c
2
+ 6
11
a
ik
S
e
kj
+
S
e
ik
a
kj
−
2
3
a
km
S
e
mk
δ
ij
!
+
7
c
2
−
10
11
(
a
ik
Ω
e
kj
−
Ω
e
ik
a
kj
)
#
,
(3.13)ε
ij
=
ε
2
3
δ
ij
,
(3.14)where
c
1
isarelaxationoeientandc
2
= 5
/
9
isaparameterofthemodelfortherapid part ofΠ
ij
,whihdependsdiretlyonhangesin theresolvedveloitygradients,andε
=
ε
ii
/
2
. Like in Wallin & Johansson's [Wallinand Johansson,2000℄ RANS model, the derivation of the EASSM model involves the weak equilibrium assumption, whih implies that the advetion anddiusion terms of the Reynolds stress anisotropy are negleted. In order to simplify the model,
togetherwiththeweakequilibriumassumption,weassumealso that
P
=
ε
. Thisleadsto0 =
P
ij
−
ε
ij
+ Π
ij
.
(3.15)Usingthemodellinggivenby3.13and3.14in3.15wehave
c
1
a
ij
=
τ
∗
"
−
11
15
S
e
ij
+
4
9
(
a
ik
Ω
kj
−
Ω
ik
a
kj
)
#
(3.16) notethatτ
∗
=
K
SGS
/ε
istheSGStime sale.Finally,equation3.16hasbeensolvedusinganansatz
a
ij
=
β
1
τ
∗
S
e
ij
+
β
4
τ
∗
2
(
S
e
ik
Ω
e
kj
−
Ω
e
ik
S
e
kj
)
,
(3.17) whereβ
1
andβ
4
aremodelparametersandfuntionsofthelteredstressandstrain-rate. Using that ansatz,equation3.16beomesτ
ij
=
K
SGS
"
2
3
δ
ij
+
β
1
τ
∗
e
S
ij
+
β
4
τ
∗
2
(
S
e
ik
Ω
e
kj
−
Ω
e
ik
S
e
kj
)
#
,
(3.18)whih isthemain EASSM modelformulation. Usingnormalizedstrainandstress-ratetensors
itanberewrittenas
τ
ij
=
2
3
δ
ij
K
SGS
+
β
1
K
SGS
S
e
∗
ij
|
{z
}
eddy
−
viscosity
+
β
4
K
SGS
(
S
e
∗
ik
Ω
e
∗
kj
−
Ω
e
∗
ik
S
e
∗
kj
)
|
{z
}
anisotropy of
SGS stresses
.
(3.19)Theseondtermontheright-hand-sideisaneddyvisositytermresponsibleforSGSdissipation,
whereasthe third termreproduesanisotropieets of SGS stressesandgivesadisalignmentof
theSGSstressandresolvedstrain-ratetensors.
β
1
andβ
4
oeientshavetheform:β
1
=
−
33
20
9
c
1
/
4
[(9
c
1
/
4)
2
+
|
Ω
e
∗
|
2
]
,
β
4
=
−
33
20
1
[(9
c
1
/
4)
2
+
|
Ω
e
∗
|
2
]
(3.20) where|
Ω
e
∗
|
=
p
2
II
∗
Ω
=
q
2
τ
∗
2
e
Ω
ik
Ω
e
ki
≤
0
istheSGStimesale-normalizedseond invariant. TheunknownquantitiesK
SGS
andτ
∗
anbedynamiallyornon-dynamiallyomputed.
Theequation 3.19analso berelatedto thetensorialformulationoftheSGS stresstensor, given in theprevioussetion: therst termontheright-handside reproduestheisotropipartofthe
SGS stress,whiletheseond andthirdtermsanbeonsidered astwopolynomialtensors,forthe
ase
k
= 1
andk
= 4
.ThedynamiversionoftheEASSM involvesGermano's dynamiproedure;heretheSGSkineti
energyismodelled intermsofthesquaredSmagorinskyveloitysale
∆
|
S
e
ij
|
[Yoshizawa,1986℄:K
SGS
=
c
∆
2
|
S
e
ij
|
,
(3.21)where
∆
e
istheltersale;|
S
e
ij
|
= (2
S
e
ij
S
e
ij
)
1
/
2
<