Professional Documents
Culture Documents
F DF302 Slides
F DF302 Slides
Poiseuille
The friction coefficient
16 ≠1 2·p Um h
Cf = Re m , with Cf = and Re m = .
3 flU02 ‹
petition
asso
Turbulence is
instationnaire
random and unsteady pqcoupure
spreading on a continuous and finite range of scales
three-dimensional
turbulence_ t ix
gg
termedadvection
ˆu ˆu
+u =0 one-dimensional Burger’s equation
ˆt ˆx
u(x , t0 ) = A cos(kx ) with a first order harmonic initial condition
Transition to turbulence
U2 ‹U
≥ ≥
L L2
Reynolds number Re = U¸/‹ drives the transition from laminar to turbulent regime via the
development of instabilities
ante
chemin
Linear growth : exponential amplification and symmetry breaking
Nonlinear saturation : finite amplitude and bifurcation
gun Secondary instabilities : growth of secondary mode(s), successive symmetry
breaking and bifurcations
Breakdown : rapid excitation of a broad band of length scales and frequencies
Lemergencetrèsrapide
Max Planck Institute Fontane, Joly & Reinaud, Phys. Fluids, 2008
A
Turbulent flows are
strongly non-linear
Extra diffusion
(Wikimedia commons)
”L (x ) ”T (x )
= 4.92 Re ≠1/2
x = 0.37 Re ≠1/5
x
x x
Effects of turbulence 16 / 28
Statistical approach
Fluctuation
Ensemble
averaging
1 2 1 2
ˆF ˆF ˆF ˆF
= and = .
ˆt ˆt ˆxi ˆxi
Statistical approach 21 / 28
ˆU j
= 0,
ˆxj
ajoutépourRANS
ˆU i ˆU i 1 ˆP ˆ # $
+ Uj =≠ + 2‹S ij ≠ ui uj with i = 1, 2, 3.
ˆt ˆxj fl ˆxi ˆxj
Reynoldssten tenor
Statistical approach 22 / 28
Cas d’un écoulement
12
Exemple « artificiel »
Equation de Burgers V v v
v
Entr observable
mon
détend NEE perturbation
Ecoulement de base
opposée petite
Relation de dispersion
15
Bilan pour l’exemple artificiel
Moninjecte
Perturbation
Problème linéarisé
16
Théorie temporelle
Théorie temporelle
Ni _intaboloté
un amplification terme d’onde
17
Théorie spatiale
Théorie spatiale
18
20
Définitions
21
Formalisme
pasexamen 1
0 MG a
II de IER CA SI
• Formalisme (opérateurs)
22
Ecoulement constant en non visqueux (1)
à
I tu o
Écoulement de base : constants
24
Solution intuitive
Notation
Mode entropique
avec
Afaf
Mode de vorticité
Modes acoustiques
25
Ecoulement constant en non visqueux (3)
Pourannuler x
26
Si en plus
27
Instabilité de Kelvin-Helmholtz (1)
instabilités aux
y
cisaillements
denitrure
r interface
Gravité : g
phénomène capillaité
eau remonte
qui
sur lesbords
Changement de référentiel :
29
30
Instabilité de Kelvin-Helmholtz (8)
Modèle physique
Position de l’interface :
36
Traitement de l’interface
Position de l’interface :
reçues
Loi de Laplace :
37
Instabilité de Kelvin-Helmholtz (9)
Diagramme de stabilité
W
flh Dieu et g
instable
stable
38
Diagramme de stabilité
stable
instable
39
Instabilité de Kelvin-Helmholtz (10)
Diagramme de stabilité
il exoste I c Juit
ET àtoutet stable
40
44
Cas visqueux : Modèle physique (1)
Grandeurs de référence
45
Nombre de Reynolds
instabilitéOu Sommerfeld
46
Écoulement de base
constant
47
48
Équation d’Orr-Sommerfeld
Équation d’Orr-Sommerfeld
49
Calcul du spectre
Théorie temporelle
Discrétisation
50
Exemple de spectre
instable
stable
51
Formalisme
opérateur linéaire
52
Diagramme de stabilité
55
Y Ue
X
D
U∞ U
if
f
Convection u
si
o
n
L
U
Epaisseur de déplacement: 1 (1 ) dy
0 e Ue
• On montre que physiquement l’épaisseur de déplacement représente la hauteur dont il
faudrait déplacer la paroi pour obtenir le débit inchangé en fluide parfait:
U U
Epaisseur de quantité de mouvement: (1 ) dy
0 e Ue Ue
1
Facteur de Forme: H
Mdr
Gaf
H≈2.59 H≈1.4
ue 2 ue
• Exact: solution de Blasius • Exact: Non
H 2.591 H 1.4
x 0,0309 x
1 1.7208 1
Rx1/ 6
Rx
0.0221 x
0.664 x Rx1/ 6
Rx
0.16 x
x Rx1/ 7 23,6 mm
5 6 mm
Rx
Application: Calculer les épaisseurs d’une couche limite laminaire et
turbulente à l’extrémité d’une plaque plane de 1m pour une vitesse de 10 m/s
olivier.vermeersch@onera.fr Dyn. des Fluides: Instabilités et transition 3A 7
Epaisseur de déplacement 1
Epaisseur de quantité de mouvement
Coefficient de frottement Cf
ou flux de chaleur pariétal p
Phénomène de transition
Déclenchement de la transition
région de transition
Y U∞
X
Instabilité Turbulent
Laminaire Transition
Principe de la théorie :
Introduire des petites perturbations sinusoïdales dans les équations de Navier-
Stokes linéarisées et rechercher les fréquences instables
olivier.vermeersch@onera.fr Dyn. des Fluides: Instabilités et transition 3A 16
U
et sont complexes, réel
2 2 d 2U d 2W
vIV 2 2 2
v'' 2
v iR U W v'' ( 2 2
)v v
dy2 dy2
y 0 v v' 0
y v 0, v' 0
IV 2 4 2 d 2U
v 2 v' ' v iR U v' ' v v Orr-Sommerfeld
dy2
Se simplifie en:
2 2 d 2U
U v' ' v v 0 Rayleigh
dy2
-102 i =
H = 2.59
R H = 3.05
1
R 1
- i >0 - i >> 0
Instabilité Turbulent
Ondes de TS:
Amplitude initiale A0
u '2
Tu
Ue
taux deturbulence
extérieure
Prévision de la transition
Méthode du eN
On est capable de calculer l’amplification d’une
perturbation: x
A
N ln( ) i dx
A0 x0
Experiments
NT 8.43 2.4 ln(Tu ) Mack1977
(flat plate)
Mack Essais en vol: Faible Tu, NT≈10
Etauxturb
Soufflerie: Dépendant fortementext
de la qualité,
3<NT<10 !
Prévision de la transition
« Philosophie » des critères de transition
1.5
0.5
0
0 0.01 0.02 0.03 0.04 0.05
Instabilité Turbulent -0.5
-1
-1.5
t(s)
Instabilité Turbulent
Soufflerie F2 (1998)
olivier.vermeersch@onera.fr Dyn. des Fluides: Instabilités et transition 3A 56
1
T Tw,turb Tw,lam rturb rlam M e2 Te
2
tant É
Applications: Calculer la différence de
Peinture thermosensible (plaque plane à M = 7)
température pariétale pour:
Une soufflerie hypersonique (M=7,Ti=750K) T=34K
Une soufflerie subsonique (M=0.2, Ti=290K) T=0.1K
Soufflerie F2 (2009)
olivier.vermeersch@onera.fr Dyn. des Fluides: Instabilités et transition 3A 58
Défaut de surface
Les ailes d’avions (nacelles, dérives horizontale & verticale)
présentent des défauts de surfaces inévitables:
Liés à la « manufacture » : rivets, jonction entre deux panneaux
(rainures, marches), …
Liés au vol: impacts d’insectes, de glace, …
Liés à la dégradation: vieillissement du matériel, altérations dues
aux intempéries, aux mauvaises réparations …
Dryden (1959)
Ue h
270 m Rh
e
1800
2.3mm
VIE État
Application: Hauteur maximale d’une marche (descendante et montante) sur l’aile
FI
d’un A380 (h=35000ft, M=0.85)
Model d’atmosphère standard: =0.38[kg/m3], T=218[K], =1.445×10-5 [Pl]
Xk (position de la rugosité) k/ 1
U V W Q∞
0
X Y Z
U U U 1 P U
U V W u ' v'
X Y Z X Y Y
W W W 1 P W
U V W v' w'
X Y Z Z Y Y
Longitudinales
u u u
1 1 dy 11 1 dy
0 e ue u
0 e 0 e ue
Transversales
2
w w
2 dy 22 dy
0 e ue 0 e ue
" Mixtes"
w u w u
12 1 dy 21 dy
0 e ue ue 0 e u e ue
Dans la couche limite tg = w/u
A la paroi, valeur limite :
w pz
tg 0 lim
y 0 u px
Ecoulement tridimensionnel
Instabilités longitudinale (TS) et transversale (CF)
On a vu qu’en 2D la transition était déclenchée par des instabilités
longitudinales les ondes de Tollmien-Schlichting
Sur une aile en flèche s’ajoutent les instabilités transversales les
tourbillons Cross-Flow
Instabilité longitudinale :
ondes TS
Instabilité transversale :
ondes CF (crossflow)
linear
V0
non
linear
Instabilités transversales
Equations aux petites perturbations
Les mêmes qu’en 2D mais avec W non nul (et non nul aussi)
i( u w) v' 0 En y 0 : u(0) v(0) w(0) 0
dU 1 2 2
i( U W )u v i p u'' u
dy R y : u( y), v( y), w( y), p( y) 0
1 2 2
i( U W ) p' v'' v
R
dW 1 2 2
i( U W )w v i p w'' w
dy R
Instabilité longitudinale :
ondes TS
Instabilité transversale :
ondes CF (crossflow)
Instabilités transversales
Localisation des instabilités (1)
dp/dx<0 dp/dx>0
-Kp
gradient
favorable
CF destabilizing TS destabilizing
Attachment line
2U
Ue kX avec k ,We Q sin
R
1/ 2
longueur caractéristique
k
We
R nombre de Reynolds caractéristique
We e
Contamination du bord d’attaque si: R 250 with
e dU e
dx
A basse vitesse (subsonique): x 0
2U
Ecoulement au voisinage d' un point d' arret d' un cylindre : U e kX avec k
R
On a : U Q sin , et We Q sin
longueur caractéristique
dU e 2U 2Q cos
dx R R
1/ 2
R
Q R sin tan Risque de contamination augmente avec la flèche,
2 le rayon du BA, et la vitesse
olivier.vermeersch@onera.fr Dyn. des Fluides: Instabilités et transition 3A
laminaire
Point d’arrêt
O Gate
Ligne de partage
olivier.vermeersch@onera.fr Dyn. des Fluides: Instabilités et transition 3A 95
Turbulent
Laminaire
Inflow
l
Cross-Flow instabilities
Görtler instabilities
Attachment-line contamination
R R
La diminution de H a pour effet: sans avec
Diminution des i
Application pratique :
La transition est ainsi retardée Réduction de traînée visqueuse
Fentes mais surtout tôles perforées
Recul de la transition
Transition mesurée
Déstabilisation locale
Le chauffage doit être appliqué près de l’abscisse critique
olivier.vermeersch@onera.fr Dyn. des Fluides: Instabilités et transition 3A 117
Soufflage gaz
chaud
Tw=493 K
Flow Plasma
HT
Dielectric materialDielectric material
Electrodes
Recul de la transition
Facile à implémenter (collage des électrodes)
mais nécessite une alimentation en haute tension
A l’heure actuelle limité à des faibles vitesses
120 olivier.vermeersch@onera.fr Dyn. des Fluides: Instabilités et transition 3A 120
Nouvelles méthodes de contrôle de la transition
Contrôle des instabilités Cross-Flow par Micro Rugosités
Bilan
Phénomènes de transition aujourd’hui assez bien compris dans des cas
« standard » (exemple : conditions de vol)
Moyen de mesure expérimentale efficace en vol et en soufflerie
Prévision possible grâce à la théorie de stabilité linéaire, critères de
transition
Capacité à prévoir les défauts admissibles
Application:
Laminarité (avion de transport)
Turbomachines
Véhicule de rentrée atmosphérique (en sommeil)
Efforts actuels:
Introduction des critères dans les grands codes
Modélisation de la réceptivité et de la phase non-linéaire
Modélisation plus réaliste des défauts de surface
ˆU j
= 0,
ˆxj
ˆU i ˆU i 1 ˆP ˆ # $
+ Uj =≠ + 2‹S ij ≠ ui uj with i = 1, 2, 3.
ˆt ˆxj fl ˆxi ˆxj
added for
The Reynolds stress tensor R = [Rij ] = [ui uj ] RANS équation
has an advective (non linear) origin
accounts for 6 extra unknowns which must be modelled in order to close the
system
4 / 58
MACRO micro
Change of point of view :
macro/micro vs. mean/fluctuation Advection by + Advection by molecular
the mean flow the fluctuation diffusion
5 / 58
Turbulence intensity and turbulent Reynolds number
6 / 58
Two solutions...
ˆ ui uj ˆ ui uj
+ Uk = Pij + Dij + Dijp + dij‹ + „ij ≠ Áij
ˆt ˆxk p
3 4
ˆU j ˆU i
Pij = ≠ ui uk + uj uk . . . . . . . . . . . . . . . . . . . . . . . is the production term ;
ˆxk ˆxk
ˆ ui uj uk
Dij = ≠ . . . . . . . . .is the diffusion by velocity fluctuation ;
ˆxk
3 4
1 ˆ pui ˆ puj
Dijp =≠ + . . . . . . . . is the diffusion by pressure fluctuation ;
fl ˆxj ˆxi
ˆ ! "
dij‹ = 2‹ ui sjk + uj sik . . . . . . . . . . . is the molecular (viscous) diffusion ;
ˆxk
3 4
1 ˆui ˆuj
„ij = p + . . . . . . . . . is the pressure-strain correlation term ;
fl ˆxj ˆxi
3 4
ˆui ˆuj
Áij = 2‹ sjk + sik . . . . . . . . . . . . . . . . . . . . . . is the dissipation tensor.
ˆxk ˆxk
ˆ ui uj ˆ ui uj
+ Uk = Pij + Dij + Dijp + Dij‹ + „ij ≠ ‘ij
ˆt ˆxk
p
3 4
ˆU j ˆU i
Pij = ≠ ui uk + uj uk . . . . . . . . . . . . . . . . . . . . . . . is the production term ;
ˆxk ˆxk
ˆ ui uj uk
Dij = ≠ . . . . . . . . .is the diffusion by velocity fluctuation ;
ˆxk
3 4
1 ˆ pui ˆ puj
Dijp =≠ + . . . . . . . . is the diffusion by pressure fluctuation ;
fl ˆxj ˆxi
ˆ 2 ui uj
Dij‹ = ‹ . . . . is the molecular (viscous) pseudo diffusion ;
ˆxk ˆxk
3 4
1 ˆui ˆuj
„ij = p + . . . . . . . . . is the pressure-strain correlation term ;
fl ˆxj ˆxi
1 2 1 ˆu 2
ˆui j
‘ij = 2‹ . . . . . . . . . . . . . . . is the pseudo dissipation tensor.
ˆxk ˆxk
f
D ‘
k k
ke
Disappearance of the
pressure/deformation
correlation terms
∆ they correspond to energy
transfers between the
Reynolds stress tensor
components without
production/destruction of
energy
Production of
turbulent kinetic energy
Mean motion ≠ui uj S ij Turbulent motion
-
Ec k
Energy transfers
Molecular motion
2‹S ij S ij 2‹sij sij
-
E
Mean Turbulent
dissipation dissipation
A turbulent flow is said homogeneous when all its statistical properties are
invariant under translation
Isotropic turbulence
A turbulent flow is said isotropic when all its statistical properties are
invariant under reflection symmetry and rotation
æ Isotropic turbulence is necessarily homogeneous (Taylor)
Homogeneous Isotropic Turbulence (HIT)
ˆk
= ≠‘
ˆt
with
ˆ ˆ 2 ui uj
dij‹ = 2‹ (ui sjk + uj sik ) Dij‹ = ‹
ˆxk ˆxk ˆxk
3 4 1 21 2
ˆui ˆuj ˆui ˆuj
Áij = 2‹ sjk + sik ‘ij = 2‹
ˆxk ˆxk ˆxk ˆxk
ˆ ˆ 2 ui uj
Áij ≠ ‘ij = 2‹ (ui sjk + uj sik ) ≠ ‹
ˆxk ˆxk ˆxk
Both dissipations are strictly equivalent only in homogeneous flows. In the
TBL, differences are less than 2% (Bradshaw & Perot, Phys. Fluids, 1993)
Turbulent Boundary
layer
The turbulent motion Homogeneous and isotropic turbulence 14 / 58
U = U0
-
-
U0 -
- -
- -
-
- -
- Sr Sz -
-
-
-
U0
-
- U > U0
- -
- -
- -
- 52 -
- -
-
-
Si
UniformlySheared Turbulence
¸=0
r ˆf
g =f +
2 ˆr
k = ≠2h
r ˆh
q = ≠h ≠
2 ˆr
L,i = 0
1 2
1 ˆf r ˆf
Ri,j = ≠ ri rj + f + ”ij
2r ˆr 2 ˆr
1 2
1 1 ˆr 2 k 1 ˆr 2 k k
Tij,k = 3k ≠ ri rj rk + (ri ”jk + rj ”ik ) ≠ rk ”ij
2r 3 r ˆr 4r 2 ˆr 2r
Taylor micro-scales
Defined as the intersection of the osculating parabola with the horizontal axis
-
1 1 ˆ 2 f (r , t) --
= ≠
⁄2f (t) 2 ˆr 2 -
r =0 fgentoconelation
functions
Similarly
-
1 1 ˆ 2 g(r , t) --
= ≠
⁄2g (t) 2 ˆr 2 - 1
r =0
Kármán–Howarth formula
f (r, t)
Ô
⁄f (t) = 2 ⁄g (t).
Intermediate characteristic 0
f
(« dissipation ») scale
r
Similarly
1
⁄ Œ
g (t) = g(r , t) dr
0
f (r, t)
Kármán–Howarth formula
f (t) = 2 g (t).
0
⇤f
Derivation of the transport equation for the two-point double velocity correlation
known as the Kármán–Howarth equation :
1 [Eq : ui ]P ◊ uj P Õ + [Eq : uj ]P Õ ◊ ui P
2 Averaging
3 Recasting the derivatives using rk = xk |P Õ ≠ xk |P
4 Simplifying according to properties of divergence free HIT
The equation is open and the extra unknowns (triple correlations) come from the
non-linear inertial terms again !
u Õ2
∆ ‘ = 30 ‹
⁄2f
Asymptotic solutions
Linear solution
Linear Kármán–Howarth equation corresponds to a small Reynolds number of turbulence
1 2
ˆu Õ2 f u Õ2 ˆ 4 ˆf
= 2‹ 4 r
ˆt r ˆr ˆr
2 ≠5/2
Ô
u (t) = – t and ¸(t) = 8‹t
small Rez
Õ2
Inviscid solution
Inviscid Kármán–Howarth equation corresponds to a large Reynolds number of turbulence
ˆ ! " u Õ3 ˆ ! 4
"
u f r k =0
Õ2
≠
ˆt r 4 ˆr
2/7
u (t) = – (t ≠ t0 ) and ¸(t) = — (t ≠ t0 )
Õ2 ≠10/7
lage Re
Both solutions are in agreement with Loitsyanski invariance
Several issues
Is the energetic decrease is the same for all vortices whatever their size ?
Energy spectrum
What is the spectral content of the kinetic turbulent energy k ?
⁄ ⁄+Œ⁄
1 1 1
k(t) = ui ui = lim Qi,i (˛r , t) = lim „ii (˛Ÿ, t) exp (i˛Ÿ · ˛r ) d 3˛Ÿ
2 2 r æ0 2 r æ0
≠Œ
E (Ÿ, t) is not the Fourier transform of k but its integration over all wavenumbers
leads to the turbulent kinetic energy
ˆQi,j ˆ ˆ 2 Qi,j
≠ (Sik,j + Sjk,i ) = 2‹
ˆt ˆrk ˆrk ˆrk
Fourier transform :
ˆ
„ij (Ÿ, t) ≠ Ÿk [Âikj (Ÿ, t) + Âjki (Ÿ, t)] = ≠2‹Ÿ2 „ij (Ÿ, t)
ˆt
ˆ
E (Ÿ, t) = T (Ÿ, t) ≠ 2‹Ÿ2 E (Ÿ, t)
ˆt
Dissipation spectrum :
1
10
E D
2
10
E(, t)/k(t)
3
E(, t)/✏(t)
10
ŸE : most energetic
wavenumber 10 4
5
10
ŸD : most dissipative
wavenumber 10 6
7
10
8
10
0.1 1 10 100 1000
⇥ k 3/2 /✏
0.4
⁄ Ÿ
0.2
W (Ÿ, t) = ≠ T (Ÿ, t) dŸ
0
0.1
Spectral energy flux is always positive ∆ energy is transferred down the scales
Small eddies
Heat (dissipation)
0
The energy is injected at large scales from the mean flow and transferred towards
small scales
The amount of energy injected at large scale (and thus cascading and dissipated)
is driven by the mean flow
Z
u Õ2 _
injection rate : ‘≥ ^
· u Õ3
‘≥
f
_
\ f
lifetime of largest eddies : · ≥
uÕ
Z
ŸI π ŸD ‘I (t) ¥ ‘(t) ^
∆ ‘(t) ¥ W (ŸI , t)
ˆEI (t) ˆk \
ŸE π ŸI ≠ π≠ = ‘(t)
ˆt ˆt
3 4≠5/3
E (Ÿ, t) Ÿ‹ 3/4
=A◊
‘(t)1/4 ‹ 5/4 ‘(t)1/4
Kolmogorov normalisation
‚K = [‹‘]1/4
·K = [‹/‘]1/2
Taylor macro-scale (energetic vortices)
u Õ3
f ≥
‘
Taylor micro-scale (intermediate vortices)
u Õ2
⁄2f = 30‹ Scales in Turbulent Motions, Jacquin, 2009
‘
f 3/4 f 1/2
≥ Re ≥ Re
÷K ⁄f
Kolmogorov spectrum
107
‚ K ÷K 105
Re K = =1
‹
103
Kolmogorov constant cut-o↵
10 Re
A¥1
10 1
F æ AŸ̃≠5/3 when Ÿ̃ æ 0 10 3
10 6 10 4 10 2 1
̃
The width of the intertial range increases with turbulence Reynolds number
1
107
2 3
105 1 : Statistical quasi-equilibrium
4
103
E(, t)
2 : Inertial range
10 3 : Dissipative range
10 1
4 : Energetic range
10 3
10 6 10 4 10 2 1
̃
2 160
10
10 3
t = x/U0 140
x/M : 30 ! 50
4
120
10
100
x/U0
5
10 x/M : 50 ! 70
E
E(, t)
80
⇥
6
10
E
1
60
7
10
x/M = 30 40
x/M = 50
8 x/M = 70
10 20
9 0
10
1 10 100 1000 10000 0 500 1000 1500 2000 2500 3000 3500 4000 4500
3 / 46
Framework
-
-
.............
evolution
-
” x
Ïa
π1
L
Turbulent boundary layer : the mean velocity profile The turbulent Prandtl equations 5 / 46
The turbulent Prandtl equations
ˆU ˆV
+ = 0 ai
U
ˆU
+V
ˆx
ˆU
ˆy
= ≠
1 dPe
+
ˆ
3
‹
ˆU
≠ uv
4 j
ˆx ˆy fl dx ˆy ˆy
Pe (x ) = P + flv 2
L peutêtreévaluer à la loute
de la couche bite
The TBL equations are parabolic : integration using space marching in the
streamwise direction
3 4 ⁄ y
ˆU ˆU 1 dPe ˆ ˆU ˆU
U = ≠V ≠ + ‹ ≠ uv with V = dy
ˆx ˆy fl dx ˆy ˆy 0
ˆx
Turbulent boundary layer : the mean velocity profile The turbulent Prandtl equations 6 / 46
Self-similar solutions ?
0.006
- 0.004
ˆU -- Ue
·w = µ - = µ f Õ (0) Cf
ˆy y =0 ”
0.002
2·w 2 Õ A
∆ Cf = = f (0) =
flUe2
Re ” Re ◊ 0
0 4000 8000 12000 16000
Re
Turbulent boundary layer : the mean velocity profile The turbulent Prandtl equations 7 / 46
A multiple scale problem
Turbulent boundary layer : the mean velocity profile A multiple scale problem 8 / 46
Te ” ·w Ue ” ·w Cf
= ◊ = ◊ = Re ”
Ti Ue µ ‹ flUe 2
2
Turbulent boundary layer : the mean velocity profile A multiple scale problem 9 / 46
The ratio of time scales
Laminar regime
K
Te
= cst ¥ 0.6
Ti
CST
Turbulent regime
Te Â
”= cst = f (Re ” )
Ti Ref
Turbulent boundary layer : the mean velocity profile A multiple scale problem 10 / 46
Turbulent boundary layer : the mean velocity profile A multiple scale problem 11 / 46
An approximate solution
Despite the absence of self-similar velocity profile, an approximate solution
supported by experimental data exists
1 2 17
U y
=
UE ”
9
with a shape factor of H12 = 7
= 1.29
Turbulent boundary layer : the mean velocity profile A multiple scale problem 12 / 46
A piecewise approach
‹/u· ”
Turbulent boundary layer : the mean velocity profile A multiple scale problem 13 / 46
The inner layer : the surface layer
y u·
Wall-normal distance : y + =
‹
Turbulent boundary layer : the mean velocity profile The surface layer 14 / 46
An even more severe approximate in the viscous sub-layer ∆ closer to the wall the
mean viscous friction dominates : if y + æ 0, · ‹ ∫ ·t
ˆU
‹ = u·2 ∆ U+ = y +
ˆy
Turbulent boundary layer : the mean velocity profile The surface layer 15 / 46
Local self-similarity in the inner layer
Self-similarity beyond the viscous sub-layer ?
ˆU
‹ = u·2 + uv
ˆy
+
ˆU
= 1 + uv +
ˆy +
⁄ y+
+ +
U = y + uv + dy +
0
uv + = uv + (y + )
experimentally verified
Turbulent boundary layer : the mean velocity profile The surface layer 16 / 46
y
Par itition
6
U (y)
- uv is negative
uv is independent of ” if
y π”
Turbulent boundary layer : the mean velocity profile The surface layer 17 / 46
The outer layer
Turbulent boundary layer : the mean velocity profile The outer layer 18 / 46
U ≠ Ue y
= F (÷) with ÷=
u· ”
Turbulent boundary layer : the mean velocity profile The outer layer 19 / 46
Existence of an overlap region ?
L zonederecouvrement
Surface layer f inner Outer layer
layer
Small ÷ = y /” large y + = yu· /‹
÷æ0 y+ æ Œ
independent of ” Independent of ‹
1. Overlap layer exists when both conditions hold, small ÷ and large y + :
y+ yu· ” Ue ” u·
∫1∆ ◊ = ◊ ≥ Re 7/8 ∫ 1
÷ ‹ y ‹ Ue
Turbulent boundary layer : the mean velocity profile The overlap region 20 / 46
1
y + f Õ (y + ) = ÷F Õ (÷) =
Ÿ
Y 1
_
] F (÷) = ln (÷) + B
Ÿ
_
[ 1
f (y + ) = ln(y + ) + C
Ÿ
Turbulent boundary layer : the mean velocity profile The overlap region 21 / 46
Resulting velocity profile
Turbulent boundary layer : the mean velocity profile The overlap region 22 / 46
Inner layer
20
22
+ + +
U
+
= ln y + /0.41 + 5.2 pp 19
U 15 U =y
Outer layer
10
0
0.1 1 10 100 1000 10000
y+
1
velocity profile
0.9
0.8
0.7
Logarithmic profile
0.6
y/ 0.5
0.4
Wake
0.3 component
0.2
0.1
0
-0.2 0 0.2 0.4 0.6 0.8 1 1.2 1.4
U /Ue
Turbulent boundary layer : the mean velocity profile The overlap region 24 / 46
30
Re
25
20
+
15
U
10
Re✓ = 3250
5 Re✓ = 7500
Re✓ = 15000
0
0.1 1 10 100 1000 10000
y+
Turbulent boundary layer : the mean velocity profile The overlap region 25 / 46
Outer self-similar profiles
-5
-10
Ue )/u⌧
-15
(U
-20 Re
Re✓ = 3250
-25 Re✓ = 7500
Re✓ = 15000
-30
0.0001 0.001 0.01 0.1 1 10
y/
Turbulent boundary layer : the mean velocity profile The overlap region 26 / 46
The inner layer remains insensitive to a large range of the external pressure
gradients while the outer layer is strongly affected.
The wake parameter increases with the intensity of the external pressure gradient
(Ligrani, 1989)
= 0.8 (— + 0.5)0.75 1989 Lignan
wakeparameter
1.2 8
7
1
H = 1,33 6
0.8 H = 1,53
H = 1,65 5
H = 1,81
0.6 ⇧ 4
y/
0 0
0 0.2 0.4 0.6 0.8 1 1.2 0 5 10 15 20 25
U /UE
Turbulent boundary layer : the mean velocity profile The overlap region 27 / 46
The friction law
1 2
+ 1 y Ue 1 ”u· 2
U = ln y + + C + w ≠æ = ln +C +
Ÿ Ÿ ” y æ” u· Ÿ ‹ Ÿ
1 2≠1/2 1 2
Cf 1 1 Cf 2
Coles1956 = ln Re ” + ln +C +
2 Ÿ 2Ÿ 2 Ÿ
0.006
0.005
0.004
Cf
0.003
0.002
0.001
0 5000 10000 15000 20000
Re✓
Turbulent boundary layer : the mean velocity profile The friction law 28 / 46
Turbulent boundary layer : the mean velocity profile The friction law 30 / 46
Back to the turbulent kinetic energy transport equation
ˆU i
Pk = ≠ui uj . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . is the production term
ˆxj
ˆ kuj
Dk = ≠ . . . . . . . . . . . . . . . . is the diffusion by velocity fluctuation
ˆxj
p 1 ˆ puj
Dk = ≠ . . . . . . . . . . . . . . . . is the diffusion by pressure fluctuation
fl ˆxj
ˆ2k
Dk = ‹ . . . . . . . . . . . .is the molecular (viscous) pseudo diffusion
‹
ˆxj ˆxj
1 21 2
ˆui ˆui
‘=‹ . . . . . . . . . . . . . . . . . . . . . . . . . . . . is the pseudo dissipation
ˆxj ˆxj
Turbulent motion in the vicinity of the wall Energy and dissipation transport equations 32 / 46
Transport equations for normal Reynolds stress under Prandtl thin layer assumption
1 2
Du 2 ˆU ˆ ! 2" ˆ 2 u2 2 ˆu ˆu ˆu
= ≠2uv ≠ vu + ‹ + p ≠ 2‹
Dt ˆy ˆy ˆy 2 fl ˆx ˆxk ˆxk
3 4 1 2
Dv 2 ˆ pv ˆ2v 2 2 ˆv ˆv ˆv
= ≠ v3 + 2 +‹ + p ≠ 2‹
Dt ˆy fl ˆy 2 fl ˆy ˆxk ˆxk
1 2
Dw 2 ˆ ! 2" ˆ2w 2 2 ˆw ˆw ˆw
= ≠ vw + ‹ + p ≠ 2‹
Dt ˆy ˆy 2 fl ˆz ˆxk ˆxk
Production Diffusion
2
1. Production of u only 1. In the crosswise direction only
2. Feeding of v 2 and w 2 ? 2. D p ”= 0 (v 2 transport equation)
Dissipation
Always negative
Turbulent motion in the vicinity of the wall The Reynolds stress equations 33 / 46
Shear Reynolds stress and turbulent kinetic energy
Transport equation for shear Reynolds stress under Prandtl thin layer
assumption
∆ statistical symmetry : uw = vw = 0 but uv ”= 0
1 2 1 2 1 2
Duv ˆU ˆ pu ˆ 2 uv p ˆu ˆv ˆu ˆv
= ≠v 2 ≠ v 2u + +‹ + + ≠ 2‹
Dt ˆy ˆy fl ˆy 2 fl ˆy ˆx ˆxk ˆxk
Transport equation for turbulent kinetic energy under Prandtl thin layer
assumption
3 4
Dk ˆU ˆ ui ui pv ˆ2k
= ≠uv ≠ v + +‹ ≠‘
Dt ˆy ˆy 2 fl ˆy 2
Turbulent motion in the vicinity of the wall The Reynolds stress equations 34 / 46
a0 = b0 = c0 = 0
u ≥ a1 y , v ≥ b2 y 2 , w ≥ c1 y
D y æ0 y æ0 y æ0
Turbulent motion in the vicinity of the wall Specific features of near-wall turbulence 35 / 46
Asymptotic analysis close to the wall
Reynolds stresses :
a12 + c12
k ≥ y2
É in
y æ0 2
Dissipation :
1 2
1
‘= (‘11 + ‘22 + ‘33 ) ≥ ‹ a12 + c12
2 y æ0
Dissipation is not nil at the wall ! It even reaches its maximal value
Turbulent motion in the vicinity of the wall Specific features of near-wall turbulence 36 / 46
Anisotropy
u2 ∫ w 2 ∫ v 2
v2
+
6 w2
4
1. Hierarchy of normal
stresses 3
2. Production is local 2
3. Blocking distance 1
0
0 50 100 150 200 250 300 350 400
y+
Turbulent motion in the vicinity of the wall Specific features of near-wall turbulence 37 / 46
Anisotropy
u2 v2
uv/k
¥1 ¥ 0.4
k k
1
w2 ≠uv 0.5
¥ 0.6 ¥ 0.3
k k
0
-0.5
0 50 100 150 200 250 300 350 400
y+
Turbulent motion in the vicinity of the wall Specific features of near-wall turbulence 38 / 46
Influence of viscosity
y + = 3270
Viscous friction dominates the 104
near-wall region
y + = 121
itielle
103
Je
2
10
1/4
101
quantities y + = 1,21
5/3
/ 1
100
Low turbulence Reynolds number
a12 + c12
1
10
Re t ≥ y4
y æ0 4‹ 2
10
3
10
4 3 2 1
10 10 10 10 100 101
1 ⇥ ⌫ 3/4 /✏1/4
ø”
Energy drain from the mean motion (production) in the buffer layer
¿0
Turbulent transfer towards the wall
Transfer towards the wall by molecular diffusion
Dissipation of the transferred energy
Turbulent motion in the vicinity of the wall Specific features of near-wall turbulence 43 / 46
pour
lethéorème d'E turbulent pressure-strain
ajesoin diffusion correlation
¿ ¿
ˆ‘ ˆ‘
+ Uk = P‘1 + P‘2 + P‘3 + P‘4 + D‘t + D‘‹ + ‘ ≠
ˆt ˆxk
¸ ˚˙ ˝ ø ø
production viscous destruction
dedissipation diffusion deladissipation
3 4 3 4
ˆ ˆui ˆui ‹ ˆ ˆui ˆp
D‘t = ≠‹ uk ‘ = ≠2
ˆxk ˆxj ˆxj fl ˆxi ˆxj ˆxj
ˆ2‘ ˆ 2 ui ˆ 2 ui
D‘‹ =‹ = ≠2‹ 2
ˆxj ˆxj ˆxj ˆxk ˆxj ˆxk
Turbulent motion in the vicinity of the wall Specific features of near-wall turbulence 45 / 46
Pseudo-dissipation budget aide à
sortirdestendances
0.03
DNS of channel flow (Kim, Moin & Moser, 1987) blablablablablabla faire
dans la couchelimite
P✏1 + P✏2
0.02 P✏3
(⌥ P✏4 )
D✏t
0.01
⇧✏
D✏⌫
0
-0.01
-0.02
-0.03
0 10 20 30 40 50
+
y
Turbulent motion in the vicinity of the wall Specific features of near-wall turbulence 46 / 46
L 3/4
n1D = ⇠ Re`
x
What for ?
DNS Build comprehensive flow data bases for physical
analysis of turbulent flows; design and validation of
LES and RANS models.
1. Average specifications
I Global specifications :
lift, load losses, heat exchange rate, efficiency;
I Design at the nominal poin:
shaped for averaged loads.
RANS
2. Local specifications time/space/frequency
I unsteady modes at large scales (vibrations, fluid-structure
interactions, (aero-)acoustics);
I local stresses (local temperature/load maxima);
I Lagrangian transport: droplets, bubbles, interfaces;
I unsteady flow control.
URANS,LES, DES, Hybrid URANS-LES
Navier-Stokes Equations
Incompressible flow
@Ui
= 0,
@xi
@Ui @Ui 1 @P @ 2 Ui
+ Uj = +⌫ .
@t @xj ⇢ @xi @xj xj
Constitutive law of the Newtonian fluid
✓ ◆
1 @Ui @Uj
⌧ij = 2µSij avec Sij = +
2 @xj @xi
Averaged Navier-Stokes Equations
@ Ūi
= 0,
@xi
@ Ūi @ Ūi 1 @ P̄ @ 2 Ūi @ui uj
+ Ūj = +⌫ .
@t @xj ⇢ @xi @xj xj @xj
Reynolds stress tensor
⌧¯ij = 2µS̄ij
I Reynolds stresses
2
⇢Rij = 2µt S̄ij ⇢k ij
3
An Eddy-viscosity model
k2
I Dimensionally correct scheme: ⌫t ⇠ u ` ⇠
"
k2
I Operational scheme: ⌫t = Cµ
"
I Mean continuity
@ Ūi
=0
@xi
I Mean momentum
@ Ūi @ Ūi 1 @ P̄ @ 2 Ūi @ k2 2
+ Ūj = +⌫ + 2Cµ S̄ij k ij
@t @xj ⇢ @xi @xj @xj @xj " 3
@k @k
Deke + Ūj = Pk + Dkp + Dkt + Dk⌫ "
@t @xj
@ Ūi
Production Pk = ui uj
@xj
@puj
Pressure transport Dkp = ⇢1
@xj
@ui ui uj
Turbulent transport Dkt =
@xj
@2k
Molecular di↵usion Dk⌫ = ⌫
@xj @xj
@ui @ui
Dissipation "=⌫
@xj @xj
Gradient-type model q ⇠ r
⌫t
Operational scheme qk = rk
k
@ui @ui
A transport equation for " = ⌫
@xj @xj
@" @"
DEE + Ūj = P" + D" + D"⌫ d"
@t @xj
Production P"
@"uj
Turbulent transport D" =
@xj
@2"
Molecular di↵usion D"⌫ =⌫
@xj @xj
Destruction d"
Gradient-type model q ⇠ r
⌫t
Operational scheme q" = r"
"
Advancement of the closure procedure
I 2 transport equations
@k @k @ ⌫t @k
+ Ūj = (⌫ + ) + Pk "
@t @xj @xj k @xj
@" @" @ ⌫t @"
+ Ūj = (⌫ + ) + P" d"
@t @xj @xj " @xj
dk d2 k
= " = d" P " > 0 ) P " 6 d"
dt dt 2
d" d2 " dd" dd"
= P" d" ' > 0 ) 60
dt dt 2 dt dt
" "
Dimensionally correct scheme P" ⇠ Pk and d" ⇠ "
k k
" "
Operational scheme P" = C"1 Pk and d" = C"2 "
k k
"
P" d" = (C"1 Pk C"2 ")
k
I 2 transport equations
@k @k @ ⌫t @k
+ Ūj = (⌫ + ) + Pk "
@t @xj @xj k @xj
@" @" @ ⌫t @" "
+ Ūj = (⌫ + ) + (C"1 Pk C"2 ")
@t @xj @xj " @xj k
Couchetonite
Test case: the turbulent boundary layer turbulente
Hence Cµ = 0.09
Prescription of C"2
Calibration the ratio between " and k destruction rates
@ Ūi
=
@xi
@ Ūi @ Ūi 1 @ P̄ @ 2 Ūi @ 2
+ Ūj = +⌫ + 2⌫t S̄ij k ij
@t @xj ⇢ @xi @xj @xj @xj 3
@k @k @ ⌫t @k
+ Ūj = (⌫ + ) + Pk "
@t @xj @xj k @xj
@" @" @ ⌫t @" "
+ Ūj = (⌫ + ) + (C"1 Pk C"2 ")
@t @xj @xj " @xj k
2
Rij = 2⌫t S̄ij k ij
3
I Linear constitutive law
I The principal axes of Rij aligned with those of S̄ij
I Isotropic eddy-viscosity,
I Cµ prescribed under the assumption of local equilibrium
u3
I (k, ") (Jones & Launder 1972) ! p = 3 , q = 1 then z ⌘ " ⇠
2 `
u
I (k, !) (Willcox 1980) ! p = 1 , q = 1 then z ⌘ ! ⇠
2 `
k ` "2 1
I (', ✓) (Lewandowski 1991) ! ✓ = ⇠ et ' = 3 ⇠ 2
" u k `
2. Turbulence-viscosity
k2 k2 ⌫t
⌫t ⇠ we take ⌫t⇤ = =
" " Cµ
⌫t⇤ Pk @ 2 ⌫t⇤
⌫t Cµ
Dt ⌫t⇤ = (2 C"1 ) + (C"2 2)k+(⌫ + ) (r⌫t⇤ )2
k " @xj @xj "
@U Le mélange delangue
I Velocity scale: U = | |L tourbillonnaire
@y
@U 2 @U 2
I Mixing length model: ⌫t = | |L , uv = L2 | |
@y @y
5 / 33
Algebraic Modelling
È L
I Logarithmic layer
ngi L
IEF
L
y
9
Inner layer Overlapping layer >
>
⌧t >> ⌧l >
>
>
>
>
>
@U >
>
⌫ uv = u⌧2 uv = u⌧2 >
> Model
@y =
L = y
>
@U 2 >
> ⌫t = yu⌧
Constitutive law uv = L2 | | >
>
>
>
@y >
> M
@U u⌧ >
>
Logarithmic profile = >
; modèle
@y y linéaire
7 / 33
Algebraic Modelling
Test of the fully linear model
✓ ◆2
@U @U @U
⌫ uv = ⌫ + L2 = u⌧2
@y @y @y
!2
@U + @U +
L+ + 1=0
@y + @y +
8 / 33
Algebraic Modelling
Test of the fully linear model
Z p
y+
1 + 4L+ 2 1
U+ = dy +
0 2L+ 2
with U+ = 0 at y = 0
30
Log law
linear model
25 Hoyas et. al (2008) - channel
Schlatter et. al ( 2012) - BL
peutpas
20
on ne
intégrer jusquà
U+
15
15
10
g
5
0 1
10 100 101 102 103
y+
9 / 33
Algebraic Modelling
The Van Driest model
Need to add a damping function in the vicinity of the wall
15
10
0 1
10 100 101 102 103
+
y
11 / 33
Algebraic Modelling
The wake layer
I U, uv knowns 0.06
0.04
I ⌫t : max then tends
towards 0 0.02
0.00
0.0 0.2 0.4 0.6 0.8 1.0 1.2
y/
12 / 33
Algebraic Modelling
Final formulation
⌫t = (min(Le , Li ))2 |
@U @U
|
@y @y
D
With Li = y (1 exp y + /A+ ) and Le = ↵
Model performance
I Lack of generality
I Definition of a normal to the wall
I y + , evaluation
I Le and pressure gradient dependency
I Outer layer intermittency not taken into account
13 / 33
Algebraic Modelling
Baldwin & Lomax, Boundary layer scalel
6
I Inner layer: ⌫ti = (Li )2 k ⌦ k
5
I Outer layer: A more
general function F 4
F (y)/u
3
+ +
F (y ) = y k ⌦ k(1 exp y /A )
2
! ymax = 0.646 0
0.0 0.2 0.4 0.6 0.8 1.0 1.2
⇧ pressure-gradient dependent y/
16 / 33
Algebraic Modelling
Baldwin & Lomax algebraic model
Final formulation
18 / 33
Algebraic Modelling
Baldwin & Lomax algebraic model
19 / 33
Two equations Modelling
Two Equations k " formulation
21 / 33
k2
⌫t = Cµ
" ! ✓ ◆
DK @Ui @Uj @Ui @ ⌫t @k
= ⌫t + + "
Dt @xj @xi @xj @xj k @xj
! ✓ ◆
D" " @Ui @Uj @Ui @ ⌫t @" "2
= C"1 ⌫t + + C"2
Dt k @xj @xi @xj @xj " @xj k
22 / 33
Two equations Modelling
Two Equations k " formulation - Large Reynolds number
u⌧3 u⌧2 k2
I "= and k = p with ⌫t = Cµ
y Cµ "
p p
I uv /k = Cµ , 2 = (C"1 C"2 ) " Cµ - Class C6
! = 0.43 ' 0.4
23 / 33
u⌧3
"=
y Lage Ree
u⌧2
k=p
Cµ
U 1 ⇣ yu ⌘
⌧
= log +C
u⌧ ⌫
24 / 33
Two equations Modelling
Two Equations k " formulation - Low Reynolds number
25 / 33
II
@ k
"e = " 2⌫
@y
26 / 33
Two equations Modelling
Two Equations k " formulation - Low Reynolds number
3/2
k
T = = O(1) at the wall
"e
k
L = = O(y ) at the wall ! not regularised
"e
27 / 33
@ui @ 2 Ui
P"3 = 2⌫uk
@xj @xj @xk
Non negligible in the bu↵er layer
Additional modeling for a first order closure
@ui @ 2 Ui
uk ⇠ ui uk T
@xj @xj @xk
✓ 2 ◆2
@ U i
⇠ ⌫v 2 T
@xj @xk
✓ 2 ◆2
@ Ui
⇠ 2⌫⌫t
@y 2
28 / 33
Two equations Modelling
Two Equations k " formulation - Low Reynolds number
I The Launder & Sharma model lowRe
k2
⌫t = Cµ fµ
"e ! ✓ ◆
DK @Ui @Uj @Ui @ ⌫t @k
= ⌫t + + ⌫+ "
Dt @xj @xi @xj @xj k @xj
! ✓ ◆
D "e "e @Ui @Uj @Ui @ ⌫t @e"
= C"e1 ⌫t + + ⌫+
Dt k @xj @xi @xj @xj "e @xj
✓ ◆2
"e2 @ 2 Ui
C"e2 f2 + 2⌫⌫t
k @y 2
( 3.4)
With fµ = exp , f2 = 1 0.3 exp Ret2
(1 + Ret /50)2
p 2
@ k
and "e = " 2⌫
@y 29 / 33
33 / 33
Introduction
LES: the right accuracy vs cost compromise?
Direct numerical simulation cost:
!3
L3 k2
3 ⇣ " ⌘3 9 k2
3 4
N = 3 ⇠ ⇠ Ret (with Ret =
4
) (1)
⌘ ✏ ⌫3 ⌫✏
où 2 k et v34
Motivations e y_ E
I Direct Numerical Simulation is too expensive (1000 4 = 5.6e6)
9
Problems
I Numerical schemes
I Wall bounded flows
I Require high computational power (parallel computers)
I Filter definition
8 / 57
12 / 57
Large Eddy Simulation
LES: at which cost
15 / 57
In a CFD code
(For eddy viscosity models) just switch µt to µsgs , and you are
done!
However...
RANS/LES problems have their own and di↵erent problems!
LES formalism
Although simple to implement, the LES formalism is not
straightforward.
Filtering operation
Try to a priori estimate the e↵ect of an under-resolved grid,
formalized with a filtering operation
16 / 57
LES filtering
Scale separation and cut-o↵ length
18 / 57
LES filtering
A very simple space filtering for LES
I The box-filter with cut-o↵ length scale ¯
8
< 1 if |x x 0 | 6 ¯
G (x x 0 ) = ¯ 2
:
0 otherwise
25 / 57
LES filtering
Decomposition of the non-linear term
@ ūi @ui uj 1 @ p̄ @ 2 ūi
Filtered Navier Stokes equation + = +⌫
@t @xj ⇢ @xi @xj @xj
Final formulation
@ ūi @ ūi ūj 1 @ p̄ @ 2 ūi @⌧ij
+ = +⌫
@t @xj ⇢ @xi @xj @xj @xj
28 / 57
31 / 57
Sub-grid Scale Modelling
The closure issue
Final formulation
Physical issues
I The model is necessary if sub-grid scale activity,
I Has to yield the right dissipative or dispersive e↵ect,
Numerical issues
I Acceptable cost hence local in time and space,
I Do not generate numerical instabilities,
I Has a distinct e↵ect from numerical dissipation
32 / 57
Smagorinski model
Construction of ⌫sg
34 / 57
Smagorinski model
Limitations
35 / 57
Smagorinski model
Pro/cons
In practice
I Implemented with Cs ⇡ 0.1 0.12 in practice, not following
HIT results
Problems
I A model referenced against HIT, misleading in flows with
strong inhomogeous and/or anisotropy,
I Fail in a simple laminar boundary layer
I Purely dissipative model, over-dissipative in transition zones or
inverse cascade,
Alternatives
I Dynamical model: adjustment of the constant to the smallest
resolved scales (srs),
36 / 57
Smagorinski model
Dynamic evaluation of the constant
Idea
I Use a test filter (with lengthscale e )
I Usually, e = 2
I Uses the Germano identity
37 / 57
Smagorinski model
Dynamic evaluation of the constant
For the model to work well, we must chose the dynamic constant
in order to statisfy the best we can the Germano’s identity :
Ldij Tijd ⌧f d
ij minimum
Ldij Cd (2Mij ) minimum
38 / 57
Smagorinski model
Dynamic evaluation of the constant
@ ⇣ d ⌘
Lij Cd (2Mij ) = 0 (5)
@Cd
Finally:
d
1 Lij Mij
Cd =
2 Mkl Mkl
39 / 57
44 / 57
Wall bounded flows
Grid requirements-Outer layer
48 / 57
49 / 57
Wall bounded flows
Detached Eddy Simulation (Spalart et al., 97)
Original idea
I RANS has be tuned to work reasonnably well in turbulent
boundary layers
I Large eddy simulation works well in detached shear layers
(wakes, flow separation)
I Turbulent boundary layers are very expensive with LES
pro/cons
I on a practical point of view, very dependant to the grid used
I Mixing of resolved fluctuations and mean quantities
53 / 57
Wall bounded flows
Wall modeling
Exchange
Locations
RANS model
I RANS model: Simple mixing length model
I To some extent, “shift the boundary condition”
I Cost decreased by 2 orders of magnitude!
I Shear layers and outer layer of the TBL are solved using LES
55 / 57
Conclusion
LES limitations
I LES pros
- Part of the fluctuating flow field
- Aeroacoustic feedback
- Fluid Structure interaction
- Transitional flows
I LES cons
- Computational cost
- Inflow conditions
- Grid convergence
57 / 57