SCAT 1st Latin-American Summer School January 9th, 2007 Part II:
Large-eddy-simulation (LES) and filtering: N-S equations:
!u j !t + !uk u j !x k =" !p + #$2 u j !x j
!u j =0 !x j
! G(x): Filter
Large Eddy Simulation, Dynamic Model, and Applications
Charles Meneveau
Department of Mechanical Engineering Center for Environmental and Applied Fluid Mechanics Johns Hopkins University
Filtered N-S equations:
!u j !uk u j !p + =" + #$2 u j !t !x k !x j
!u j !u j !p ! + uk =" + #$2 u j " % !t !x k !x j !x k jk
where SGS stress tensor is: !
! ij = uiu j " ui u j
Mechanical Engineering
Effects of !ij upon resolved motions: Energetics (kinetic energy):
! 1 ujuj !1 u u ! 2 2 j j =" (....) " 2#S jk S jk " "$ jk S jk + uk !t !x k !x j
SGS energy dissipation:
! " = # $ jk S jk
)
G(x): Filter !
! ! 1 " !u !u j % ! Sij = $ i + 2 # !x j !xi ' &
If we wish to control dissipation of energy we can set !ij proportional to -Sij E.g. Smagorinsky-Lilly model:
!=
u" 3 L
! " = # $ jk S jk
Inertial-range flux (most attention)
! "
E(k)
!"
% $u $u j ( ! ! d ! ! ij = "# sgs ' i + * = "2# sgs Sij & $x j $xi )
LES resolved SGS
! sgs = ?? = (velocity " scale) # (length " scale)
Theoretical calibration of cs (D.K. Lilly, 1967):
% $u $u j ( ! ! d ! ! ij = "# sgs ' i + * = "2# sgs Sij & $x j $xi )
Length-scale: ~ " (instead of L), Velocity-scale ~ " |S|
! ! " = # = $ % ij Sij ! ! ! # = cs2 " 2 2 | S | Sij Sij ! ! # & cs2 " 2 2 3/2 Sij Sij
3/2
! ! ! ij = "2(cs #) | S | Sij
2
!=
u" 3 L
E(k) = cK ! 2 / 3 k "5 / 3
! "
E(k)
!"
! ! ! 1 'ui ( 'ui 'u j + ! ! Sij Sij = + = 2 'x j * 'x j 'xi ) , =
LES resolved SGS
! ! sgs ~ " 2 | S |
1 1 k2 2 3 2 E(k) 3 000/ " [k j .ii (k) + ki k j .ij (k)]d k = 2 |k|</ / " [k ( 4/ k 2 (1 ii $ k 2 )) + 0]d k 000 2 |k|< / 1 2
/ /"
! sgs = (c s")
|S |
= cK # 2 / 3
0
0
k $5 / 3+ 2
3$1 4/ k 2 dk = cK # 2 / 3 4/ k 2
3/2
/ /"
0
0
3(/ + k 1/ 3dk = cK # 2 / 3 * 4 ) ",
4/3
cs: Smagorinsky constant
#&c " 2
2 s 2
3/2
4/3 ( + 2/3 3 ( / + * cK # 4 * " - ) , , )
$ 3c ' ! 1 " cs2# 2 & K ) % 2 (
3/2
$ 3c ' ! cs = & K ) % 2 (
*3/ 4
# *1
cK = 1.6 ! cs " 0.16
cs=0.16 works well for isotropic, high Reynolds number turbulence
But in practice (complex flows)
How does cs vary under realistic conditions? Interrogate data: Measure: Measure:
! " = # $ jk S jk
! Smag " = 2"2 | S | Sij Sij 2 cs
= f(x,conditions):
c s = c s (x,t)
Ad-hoc tuning? Examples: Transitional pipe flow: from 0 to 0.16 Near wall damping for wall boundary layers (Piomelli et al 1989)
Obtainempirical Smagorinsky coefficient
$ !" S jk jk & cs = & 2 % 2# | S | Sij Sij
'1/ 2 ) ) (
y
cs = 0.16
An example result from atmospheric turbulence:
cs = cs (y)
Measure empirical Smagorinsky coefficient for atmospheric surface layer as function of height and stability (thermal forcing or damping):
HATS - 2000 (with NCAR)
Kettleman City (Central Valley, CA)
$ !" S jk jk cs = & 2 & 2# | S | S S ij ij %
'1/ 2 ) ) (
How to avoid tuning and case-by-case adjustments of model coefficient in LES? The Dynamic Model (Germano et al. Physics of Fluids, 1991)
c s = c s (x,t)
Germano identity and dynamic model
(Germano et al. 1991): Exact (rare in turbulence):
Neutral stratification
Example result: effect of atmospheric stability on coefficient from sonic anemometer measurements in atmospheric surface layer (Kleissl et al., J. Atmos. Sci. 2003)
Stable stratification
Germano identity and dynamic model
(Germano et al. 1991): Exact (rare in turbulence):
ui u j ! ui u j = uiu j
! uiu j
E(k)
ui u j ! ui u j = uiu j
- uiu j + uiu j
! uiu j
E(k)
L T
"
L T
"
LES resolved
SGS
LES resolved
SGS
Germano identity and dynamic model
(Germano et al. 1991): Exact (rare in turbulence):
Germano identity and dynamic model
(Germano et al. 1991): Exact (rare in turbulence):
ui u j ! ui u j = uiu j
- uiu j + uiu j
! uiu j
E(k)
ui u j ! ui u j = uiu j
- uiu j + uiu j
! uiu j
E(k)
Tij =
! ij
+ Lij
L T "
Tij =
k
!2 (c s 2" ) | S | Sij
2
! ij
+ Lij
L T "
Lij ! (Tij ! " ij ) = 0
Lij ! (Tij ! " ij ) = 0
2 !2 (c s" ) | S | Sij
Assumes scale-invariance:
LES resolved
SGS
Lij ! c Mij = 0
2 s
LES resolved
SGS
where
2 M ij = 2! | S | Sij " 4 | S | Sij
Germano identity and dynamic model
(Germano et al. 1991):
Similarity, tensor eddy-viscosity, and mixed models
mnl $ ij = Cnl !2
Lij ! c 2 Mij = 0 s
Over-determined system: solve in some average sense (minimize error, Lilly 1992):
~ ~ #ui #u j ~~ " 2(CS !) 2 S Sij #xk #xk
~~ ~~ Lij " Tij ! # ij = uiu j ! ui u j
Two-parameter dynamic mixed model
! = (Lij " c 2 Mij ) s
Minimized when:
c =
2 s
Lij Mij Mij Mij
Averaging over regions of statistical homogeneity or fluid trajectories
Mixed tensor Eddy Viscosity Model:
~ ~ ~~ 2 !u !u j Tij = #2(CS 2") 2 S Sij + Cnl (2" ) i !xk !xk
Lagrangian dynamic model (CM, Tom Lund & Bill Cabot, JFM 1996): Average in time, following fluid particles for Galilean invariance:
A =
!"
# A(t') T e
(t -t' ) T
dt
Taylor-series expansion of similarity (Bardina 1980) model (Clark 1980, Liu, Katz & Meneveau (1994), ) Deconvolution: (Leonard 1997, Geurts et al, Stolz & Adams, Winckelmans etc..) Significant direct empirical evidence, experiments: Liu et al. (JFM 1999, 2-D PIV) Tao, Katz & CM (J. Fluid Mech. 2002): tensor alignments from 3-D HPIV data Higgins, Parlange & CM (Bound Layer Met. 2003): tensor alignments from ABL data From DNS: Horiuti 2002, Vreman et al (LES), etc
!" d ij
Sij
Reality check: Do simulations with these closures produce realistic statistics of ui(x,t)?
Contractions
Remake of Comte-Bellot & Corrsin (1967) decaying isotropic turbulence experiment at high Reynolds number
Active Grid M=6" Test Section
Need good data
Need good simulations Next: Summary of results from Kang et al. (JFM 2003) Smagorinsky model, Dynamic Smagorinsky model, Dynamic 2-parameter mixed model
u2 x2
u1 x1
X-wire probe array Flow
!1 = 0.01m !2 = 0.02m !3 = 0.04m !4 = 0.08m
0.91m
20M
h1 = 0.005m !1 = 0.01m
30M
!
40M
48M
h3 = 0.02m !3 = 0.04m
Corrsin Wind-tunnel & active grid:
X-Wire Probe Array and Automatic Calibration System
* Four X-wire Probes ! L/d = 200, d = 2.5 m ! Measurement volume = 0.5 mm
Randomly rotating winglets Makita (1991) Mydlarski & Warhaft (1996)
Grid Size M: 6 inches Winglet Speed # [208, 417] rpm selected with uniform probability randomly in both directions.
* " = 0.01, 0.02, 0.04, 0.08 m * 36,000,000 data points / probe * with a sampling rate of 40 kHz
u2 x2
u1 Flow x1
20 M
30 M
40 M
48M
Initial condition for LES (m3s-2)
100
Results
10-1 x /M=20 10-2 10-3 x /M=48
1 1
E(!)
10-4 10
-5
10-6 10-7 0 10
10
10
10
10
(m-1)
Results: Dynamic Model Coefficients
LES of Temporally Decaying Turbulence
" Dynamic Smagorinsky
Experiment Spatially Decaying Grid Turbulence 2-D Box Filter
" " "
" Dynamic Mixed tensor eddy visc:
mnnl $ ij = Cnl !2
Taylor Hypothesis
x1 = u1 t
LES Temporally Decaying Turbulence 3-D Filter
! dyn "Smag = "2 ij
0.2
Lij M ij M ij M ij
#2 S Sij
~ ~ #ui #u j ~~ " 2(CS !) 2 S Sij #xk #xk
0.2
Pseudo-spectral code: 1283 nodes, carefully dealiased (3/2N) All parameters are equivalent to those of experiments. Initial energy distribution: 3-D energy spectrum at x1/M = 20 LES Models: standard Smagorinsky-Lilly model, dynamic Smagorinsky and dynamic mixed tensor eddy-visc. model
C
0.15
0.15
nl
nl
C
0.1
0.1
nl 0.05
C
1/12 from the first order approximation of "ij
0.05
"
0 0 10 20
x /M
1
30
40
50
0 0 10 20
x /M
1
30
40
50
3-D Energy Spectra (LES vs experiment)
" Dynamic Smagorinsky
10
0
3-D Energy Spectra (LES vs experiment)
" Dynamic Mixed tensor eddy-visc. model
10
0
10
-1
x /M=20
1
10
-1
x /M=20
1
10
-2
x /M=48
1
Exp.
10
-2
E(!)
10
-3
x /M=48
1
E(!)
cutoff grid filter (! = 0.04 m) cutoff test filter at 2!
0
10
-3
10
-4
10
-4
10
-5
cutoff grid filter (! = 0.04 m) cutoff test filter at 2!
0
10
10
10
10
10
-5
10
10
10
10
PDF of SGS Stress (LES vs experiment)
"SGS Stress
25
Lagrangian dynamic model (M, Lund & Cabot, JFM 1996): Average in time, following fluid particles for Galilean invariance:
~~ # 12 " u1u2 ! u1u2
Smagorisnky Dynamic Smagorisnky
E =
!"
# (L
ij
! Cs2 M ij
t
-t') 1 !(t T e dt' T
x /M=48
1 20
! E =0 "
Dynamic mixed tensor eddyvisc model predicts PDF of the SGS stress accurately.
15
PDF
10
1 # (t -t' ) % Lij M ij T e T dt' Cs2 = #$ t (t -t' ) 1 # % Mij Mij T e T dt' #$
!LM =
"#
$L M
ij
ij
-t') 1 " (tT e dt' T
!MM =
Dynamic Mixed Nonlinear
"#
$M
ij
Mij
1 "(t -t' ) e T dt' T
With exponential weight-function, equivalent to relaxation forward equations:
Exp.
0 -0.4 -0.3 -0.2 -0.1
~ ! 12 / u12 rms
0.1
0.2
0.3
0.4
!"LM !" 1 + uk LM = (Lij M ij # "LM ) !t !xk T !"MM !" 1 + uk MM = (Mij Mij # "MM ) !t !x k T
c s2 =
!LM (x,t) !MM (x,t)
Lagrangian dynamic model has allowed applying the Germanoidentity to a number of complex-geometry engineering problems
LES of flows in internal combustion engines: Haworth & Jansen (2000) Computers & Fluids 29.
Examples:
LES of flow over wavy walls Armenio & Piomelli (2000) Flow, Turb. & Combustion. LES of structure of impinging jets: Tsubokura et al. (2003) Int Heat Fluid Flow 24.
Examples:
LES of flow in thrust-reversers Blin, Hadjadi & Vervisch (2002) J. of Turbulence.
Examples:
LES of convective atmospheric boundary layer: Kumar, M. & Parlange (Water Resources Research, 2006) Transport equation for temperature Boussinesq approximation Coriolis forcing Lagrangian dynamic model with assumed #=Cs(2 ")/Cs(") Constant (non-dynamic) SGS Prandtl number Prsgs=0.4 Imposed surface flux of sensible heat on ground Diurnal cycle: start stably stratified, then heating.
Imposed ground heat flux during day:
Examples:
Diurnal cycle: start stably stratified, then heating. Resulting dynamic coefficient (averaged):
Large-eddy-Simulation of atmospheric flow over fractal trees:
Consistent with HATS field measurements:
Downtown Baltimore:
wind
URBAN CONTAMINATION AND TRANSPORT
Useful references on LES and SGS modeling:
P. Sagaut: Large Eddy Simulation of Incompressible Flow (Springer, 3rd ed., 2006)
Momentum and scalar transport equations solved using LES and Lagrangian dynamic subgrid model. Buildings are simulated using immersed boundary method.
U. Piomelli, Progr. Aerospace Sci., 1999 C. Meneveau & J. Katz, Annu Rev. Fluid Mech., 2000
Yu-Heng Tseng, C. Meneveau & M. Parlange, 2006 (Env. Sci & Tech. 40, 2653-2662)