Professional Documents
Culture Documents
Charles Meneveau
Johns Hopkins University
Baltimore, USA
Additional reference:
http://www.scholarpedia.org/article/Turbulence:_Subgrid-Scale_Modeling
OVERVIEW:
Mercoled:
Venerd:
Test Section
Contractions
Flow
Decaying turbulence
Johns Hopkins University: Baltimore, Maryland USA
JHU: Latrobe Hall
(Mechanical Engineering)
Downtown
Baltimore
multiscale,
mixing,
dissipative,
chaotic,
vortical
well-defined statistics,
important in practice
Physical quantities describing fluid flow
Density field
Velocity vector field
Pressure field
Temperature field (or internal energy, or enthalpy etc..)
u j
=0
x j
u j uk u j 1 p
+ = + 2u j + g j
t xk x j
Fj
aj =
m
Traditional approach: Reynolds decomposition
u j
=0
x j
u j uk u j 1 p
+ = + 2u j + g j
t xk x j
t
Traditional approach: Reynolds decomposition
u j
=0
x j
u j uk u j 1 p
+ = + 2u j + g j
t xk x j
Reynolds equations:
u j
=0
x j t
( )
u j u j uk 1 p
+ = + u j + g j
2
u j uk u j uk
t xk x j xk
E(k) k 5 / 3
K
Turbulence problem:
last unsolved problem in classical physics
(e.g. Feynman 1979)
First-principles derivation of
Kolmogorovsk-5/3 spectrum
N-S equations:
u j u k u j p u j
+ = + 2 u j + g j =0
t xk x j x j
u(k,t)
k u(k,t) = 0, = P(k) (u
)(k) k 2 u(k,t) + P(k) f(k,t)
t
(u
)(k) = FFT [u(x,t) u(x,t)]
x2 FFT k2
u(k,t)
u(x,t)
x1 k1
DNS - pseudo-spectral calculation method
(Orszag 1971: for isotropic turbulence - triply periodic boundary conditions)
x2 FFT k2
u(k,t)
u(x,t)
x1 k1
Computational state-of-the-art:
~ 40 grid
points
1,024 grid
points
1,024 grid
points
1,024 grid points
Take 10243 DNS of forced isotropic turbulence
(standard pseudo-spectral Navier-Stokes simulation, dealiased):
http://turbulence.pha.jhu.edu
http://turbulence.pha.jhu.edu
we adapted Fortran,
C and Matlab to
surf the web
for data
64
4x109 105
d.o.f. d.o.f.
u j uk u j p u j
t
+
x k
=
x j
+ 2 u j
x j
=0
u1 (x, y, z0 ,t 0 )
ui (x,t) = G * ui = G (x x ')ui (x ')d 3x '
D
Large-eddy-simulation (LES) and filtering:
u j uk u j p u j
t
+
x k
=
x j
+ 2 u j
x j
=0
Filtered N-S equations: u1 (x, y, z0 ,t 0 )
u j
u p
ku j
+ = + 2u j
t xk x j
D
Large-eddy-simulation (LES) and filtering:
u j uk u j p u j
t
+
x k
=
x j
+ 2 u j
x j
=0
Filtered N-S equations: u1 (x, y, z0 ,t 0 )
u j
u p
ku j
+ = + 2u j
t xk x j
u u D
j j p
+ uk = + u j
2
t x k x j x k jk
ij = uiu j ui u j
Most common modeling approach: eddy-viscosity
ui u j
= sgs
d
+ = 2 sgs Sij
x j xi
ij
sgs = (c s) | S |
2
ui u j
= sgs
d
+ = 2 sgs Sij
x j xi
ij
ui u j
= sgs
d
+ = 2 sgs Sij
x j xi
ij
ui u j
= sgs
d
+ = 2 sgs Sij
x j xi
ij
ui u j
= sgs
d
+ = 2 sgs Sij
x j xi
ij
( ) ( )
d d
u
d
ij i u j ui u j
u
iu
A fluid-mechanical rationale for basic eddy-viscosity:
( ) ( )
d d
u
d
ij i u j ui u j
u
iu
A
A fluid-mechanical rationale for basic eddy-viscosity:
( ) ( )
d d
u
d
ij i u j ui u j
u
iu
( ) ( )
d d
u
d
ij i u j ui u j
u
iu
du u
Ai
dt
u(0)
u(t)
A
A fluid-mechanical rationale for basic eddy-viscosity:
i u(0)
u(t) = exp( At)
du u
Ai
dt
u(0)
u(t)
A
A fluid-mechanical rationale for basic eddy-viscosity:
i u(0)
u(t) = exp( At)
du u 1 2
Ai u(t) I At + (At) ... iu(0)
dt 2
u(0)
u(t)
A
A fluid-mechanical rationale for basic eddy-viscosity:
du u 1 2
Ai u(t) I At + (At) ... iu(0)
dt 2
( ) (
i u(0) uT (0) i I A
u u = uuT I At )
Tt
u(0)
u(t)
A
A fluid-mechanical rationale for basic eddy-viscosity:
du u 1 2
Ai u(t) I At + (At) ... iu(0)
dt 2
( ) (
i u(0) uT (0) i I A
u u = uuT I At
Tt
)
u(0)
u(t)
u u T A ( )
I At
i u u T A
t =0
( Tt
i IA )
A
A fluid-mechanical rationale for basic eddy-viscosity:
du u 1 2
Ai u(t) I At + (At) ... iu(0)
dt 2
( )
i u(0) uT (0) i I A
u u = uuT I At
Tt
( )
u(0)
u(t)
u u T A (
I At )
i u u T A
t =0
( Tt
i IA )
A
( )I
2
isotropy : ce S
A fluid-mechanical rationale for basic eddy-viscosity:
du u 1 2
Ai u(t) I At + (At) ... iu(0)
dt 2
( )
i u(0) uT (0) i I A
u u = uuT I At
Tt
( )
u(0)
u(t)
u u T A (
I At )
i u u T A
t =0
( Tt
i IA )
A
( )I
2
isotropy : ce S
( ) ( ) I (A + A ) t
2 2
u u A
T
ce S A )i(I A
(I At T t A ) ce S T
+ O(t 2 )
tA A
2 S
A fluid-mechanical rationale for basic eddy-viscosity:
du u 1 2
Ai u(t) I At + (At) ... iu(0)
dt 2
( )
i u(0) uT (0) i I A
u u = uuT I At
Tt
( )
u(0)
u(t)
u u T A (
I At )
i u u T A
t =0
( Tt
i IA )
A
( )I
2
isotropy : ce S
( ) ( ) I (A + A ) t
2 2
u u A
T
ce S A )i(I A
(I At T t A ) ce S T
+ O(t 2 )
tA A
2 S
( )t
d 2
= uu A
d
ij
T
2 ce S A S
tA
sgs
A fluid-mechanical rationale for basic eddy-viscosity:
( )t
2
= u u
d
ij
T d
tA
2 ce S A S
sgs
1
choosing tA
S
u1 (x, y, z0 ,t 0 )
= jk S jk
u 3
= Inertial-range flux
L
E(k)
k
LES resolved SGS
Theoretical calibration of cs (D.K. Lilly, 1967): u 3 E(k) = cK 2 / 3 k 5 / 3
=
L
= = ij Sij ij = 2(cs )2 | S | Sij
= cs2 2 2 | S | Sij Sij E(k)
3/2
c 2 2
s
2 3/2
Sij Sij
1 ui ui u j k
Sij Sij = + = LES resolved SGS
2 x j x j xi
1 1 2 E(k) k2
= [k j ii (k) + ki k j ij (k)]d k = [k (
2 3
( )) + 0]d 3
k
4 k 2
ii
2 |k|< / 2 |k|< / k2
/ /
31 2/3 3
4/3
2/3 1
= cK k 5 / 3+ 2 4 k 2 dk = cK 2 / 3 k dk = cK
1/ 3
2 0
4 k 2
0
4
3/2
3
4/3
3/2 3/ 4
cs 2 cK
2 2 3/2 2/3
3c 3c
4 1 cs2 2 K cs = K 1
2 2
cK = 1.6 cs 0.16
Theoretical calibration of cs (D.K. Lilly, 1967): u 3 E(k) = cK 2 / 3 k 5 / 3
=
L
= = ij Sij ij = 2(cs )2 | S | Sij
= cs2 2 2 | S | Sij Sij E(k)
3/2
c 2 2
s
2 3/2
Sij Sij
1 ui ui u j k
Sij Sij = + = LES resolved SGS
2 x j x j xi
1 1 2 E(k) k2
= [k j ii (k) + ki k j ij (k)]d k = [k (
2 3
( )) + 0]d 3
k
4 k 2
ii
2 |k|< / 2 |k|< / k2
/ /
31 2/3 3
4/3
2/3 1
= cK k 5 / 3+ 2 4 k 2 dk = cK 2 / 3 k dk = cK
1/ 3
2 0
4 k 2
0
4
3/2
3
4/3
3/2 3/ 4
cs 2 cK
2 2 3/2 2/3
3c 3c
4 1 cs2 2 K cs = K 1
2 2
cK = 1.6 cs 0.16
Theoretical calibration of cs (D.K. Lilly, 1967): u 3 E(k) = cK 2 / 3 k 5 / 3
=
L
= = ij Sij ij = 2(cs )2 | S | Sij
= cs2 2 2 | S | Sij Sij E(k)
3/2
c 2 2
s
2 3/2
Sij Sij
1 ui ui u j k
Sij Sij = + = LES resolved SGS
2 x j x j xi
1 1 2 E(k) k2
= [k j ii (k) + ki k j ij (k)]d k = [k (
2 3
( )) + 0]d 3
k
4 k 2
ii
2 |k|< / 2 |k|< / k2
/ /
31 2/3 3
4/3
2/3 1
= cK k 5 / 3+ 2 4 k 2 dk = cK 2 / 3 k dk = cK
1/ 3
2 0
4 k 2
0
4
3/2
3
4/3
3/2 3/ 4
cs 2 cK
2 2 3/2 2/3
3c 3c
4 1 cs2 2 K cs = K 1
2 2
cK = 1.6 cs 0.16
Theoretical calibration of cs (D.K. Lilly, 1967): u 3 E(k) = cK 2 / 3 k 5 / 3
=
L
= = ij Sij ij = 2(cs )2 | S | Sij
= cs2 2 2 | S | Sij Sij E(k)
3/2
c 2 2
s
2 3/2
Sij Sij
1 ui ui u j k
Sij Sij = + = LES resolved SGS
2 x j x j xi
1 1 2 E(k) k2
= [k j ii (k) + ki k j ij (k)]d k = [k (
2 3
( )) + 0]d 3
k
4 k 2
ii
2 |k|< / 2 |k|< / k2
/ /
31 2/3 3
4/3
2/3 1
= cK k 5 / 3+ 2 4 k 2 dk = cK 2 / 3 k dk = cK
1/ 3
2 0
4 k 2
0
4
3/2
3
4/3
3/2 3/ 4
cs 2 cK
2 2 3/2 2/3
3c 3c
4 1 cs2 2 K cs = K 1
2 2
cK = 1.6 cs 0.16
cs=0.16 works well for isotropic,
high Reynolds number turbulence
But in practice
(complex flows)
c s = c s (x,t)
Ad-hoc tuning?
cs = cs (y)
How does cs vary under realistic conditions?
Interrogate data:
Measure: = jk S jk
Measure: Smag
2 = 22
| S | Sij Sij
cs
S 1/ 2
jk jk
cs = 2
2 | S | Sij Sij
HATS - 2000
(with NCAR
researchers:
Horst, Sullivan) S 1/ 2
Kettleman City
(Central Valley, CA) cs = jk jk
22 | S | S S
ij ij
Neutral stratification
(Kleissl et al., J. Atmos. Sci. 2003)
c s = c s (x,t)
How to avoid tuning and case-by-case
adjustments of model coefficient in LES?
ui u j ui u j = uiu j u iu j E(k)
! k
ui u j ui u j = uiu j - u iu j + u iu j u iu j E(k)
! k
ui u j ui u j = uiu j - u iu j + u iu j u iu j E(k)
Tij = ij + Lij
Lij (Tij ij ) = 0
L ! k
T
ui u j ui u j = uiu j - u iu j + u iu j u iu j E(k)
Tij = ij + Lij
Lij (Tij ij ) = 0
2 (c s 2 ) | S | Sij
2
2 (c s ) | S | Sij
2
L ! k
T
Assumes scale-invariance:
LES resolved SGS
Lij c Mij = 0
2
s
(
where M ij = 2 | S | Sij 4 | S | Sij
2
)
Germano identity and dynamic model
(Germano et al. 1991):
Lij c s Mij = 0
2
Over-determined system:
solve in some average sense
(minimize error, Lilly 1992):
= (Lij c Mij )
2 2
s
Averaging over regions of
Minimized when: statistical homogeneity
or fluid trajectories
Lij Mij
c =
2
s
Mij Mij
Germano identity and dynamic model
(Germano et al. 1991):
Lij c s Mij = 0
2
Over-determined system:
solve in some average sense
(minimize error, Lilly 1992):
= (Lij c Mij )
2 2
s
Averaging over regions of
Minimized when: statistical homogeneity
or fluid trajectories
Lij Mij
c =
2
s
Mij Mij
Homework:
(a) Prove the above equation, and
qi = u
i T ui T
(b) repeat entire formulation for the SGS heat flux vector
T
qi = Cscalar | S |
2
modeled using an SGS diffusivity (find Cscalar) xi
Similarity, tensor eddy-viscosity, and mixed models
u~ u~ ~~
ijmnl = Cnl 2 i
j
2(CS ) 2 S Sij
xk xk
xk xk
Mixed tensor Eddy Viscosity Model:
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) dij
Tao, Katz & CM (J. Fluid Mech. 2002): Sij
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
Reality check:
h1 = 0.005m
D1 = 0.01m h3 = 0.02m D
D3 = 0.04m
u2
x2 u1 Flow
x1
20 M 30 M 40 M 48M
x /M=20
-2 1
10
x /M=48
10-3 1
E()
10-4
10-5
10-6
10-7
100 101 102 103 104
(m-1)
LES of Temporally Decaying Turbulence
0.15 C 0.15
s C C
nl nl
C
s
C 0.1 0.1 C
s C C s
nl s
1/12 from the first order
0.05 0.05 approximation of Lij
0 0
0 10 20 30 40 50 0 10 20 30 40 50
x /M x /M
1 1
3-D Energy Spectra (LES vs experiment)
Dynamic Smagorinsky
0
10
-1
10 x /M=20
1
-2
10 x /M=48 Exp.
1
E()
-3
10
cutoff grid filter
-4
(D = 0.04 m)
10
cutoff test filter
at 2D
-5
10
0 1 2 3
10 10 10 10
3-D Energy Spectra (LES vs experiment)
Dynamic Mixed tensor eddy-visc. model
0
10
-1
10 x /M=20
1
-2
10 x /M=48
1
E()
-3
10
cutoff grid filter
(D = 0.04 m)
10
-4
cutoff test filter
at 2D
-5
10
0 1 2 3
10 10 10 10
PDF of SGS Stress (LES vs experiment)
SGS Stress ~ ~
12 u1u2 u1u2
25
x /M=48
1
20
Smagorisnky
15 Dynamic
Smagorisnky
PDF
Dynamic Mixed Dynamic mixed tensor
10
Nonlinear eddy-visc model predicts
PDF of the SGS stress
5
accurately.
Exp.
0
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4
12 / u~12rms
Germano identity and dynamic model
(Germano et al. 1991):
ui u j ui u j = uiu j - u iu j + u iu j u iu j E(k)
Tij = ij + Lij
Lij (Tij ij ) = 0
2 (c s 2 ) | S | Sij
2
2 (c s ) | S | Sij
2
L ! k
T
Assumes scale-invariance:
LES resolved SGS
Lij c Mij = 0
2
s
(
where M ij = 2 | S | Sij 4 | S | Sij
2
)
Germano identity and dynamic model
(Germano et al. 1991):
Lij c s Mij = 0
2
Over-determined system:
solve in some average sense
(minimize error, Lilly 1992):
= (Lij c Mij )
2 2
s
Averaging over regions of
Minimized when: statistical homogeneity
or fluid trajectories
Lij Mij
c =
2
s
Mij Mij
t
1 (t T-t')
(L )
2
E = ij C M ij
2
s e dt'
T
Lagrangian dynamic model (M, Lund & Cabot, JFM 1996):
Average in time, following fluid particles for Galilean invariance:
t
1 (t T-t')
(L )
2
E = ij C M ij
2
s e dt'
T
t
1 (tT-t')
LM = Lij M ij e
t (t -t' )
1 dt'
Lij M ij
T
e dt'T
T
E =0 Cs =
2
t (t -t' )
1 T
Mij Mij T e dt'
t
1 (t -t'
)
t
1 (t T-t')
(L )
2
E = ij C M ij
2
s e dt'
T
t
1 (tT-t')
LM = Lij M ij e
t (t -t' )
1 dt'
Lij M ij
T
e dt'T
T
E =0 Cs =
2
t (t -t' )
1 T
Mij Mij T e dt'
t
1 (t -t'
)
MM 1
s
MM (x,t)
+ uk MM = (Mij Mij MM )
t x k T
Lagrangian dynamic model has allowed applying the Germano-
identity to a number of complex-geometry engineering problems
Michael Phelps
local Baltimore boy
Ciclo di seminari
Universita degli studi di Roma, Tor Vergata
May, 2013
OVERVIEW:
Mercoled:
Venerd:
Imposed ground
heat flux during day:
Examples:
Downtown Baltimore:
wind
Momentum and scalar transport equations solved using LES and Lagrangian dynamic
subgrid model. Buildings are simulated using immersed boundary method.
Agriculture: What is the isolation distance to avoid cross-polination?
Deposition flux?
Liso
LES:
Eulerian approach C(x,y,z,t)
Vertical settling velocity ws
C
t (( )
+ ( u ws e 3 ) i C = i Csdyn Scsgs
2 1
| S | C )
Scale-dep dynamic eddy-viscosity eddy-diffusivity SGS
Log-law type boundary condition
For C, log-law, corrected by settling velocity
Low ws
emitter field
High ws
Deposition flux?
Liso
LES results:
Downstream evolution of concentration profiles
Downstream evolution of
deposition flux
1-D reduction (back to early 1900s) - vertical profile
Similarity solution
for any eta (Rouse #)
Deposition downstream of field:
Sc ws
=
u*
Over field
Downstream
Scaling of deposition flux: field edge delta instead of L
Sc ws
= = 0.313
u* Traditional approach
(current rules)
Sc ws
= =0
u*
deltaL
Different L
L
Sc ws
Scaling with deltaL
= = 0.313
u*
Sc ws
= =0
u*
Collapse for
different L
Isolation distance as function of flux threshold:
Sc ws
E.g., ws=0.06 m/s (d=40 micro-m), u*=0.5 m/s, Sc=1, k=0.4, = = 0.3
u*
ID = 4.5km !!
Useful references on LES and SGS modeling: