Multi-scale Simulation of Wall-bounded Flows

Mechanics

Feb 22, 2014 (4 years and 4 months ago)

98 views

Computational Combustion Lab
Aerospace Engineering

Multi
-
scale Simulation of Wall
-
bounded Flows

Ayse G. Gungor and Suresh Menon

Georgia Institute of Technology

Atlanta, GA, USA

Supported by Office of Naval Research

WALL BOUNDED SHEAR FLOWS: TRANSITION AND TURBULENCE

Isaac Newton Institute for Mathematical Sciences

Cambridge, UK

September 11
th
, 2008

Computational Combustion Lab
Aerospace Engineering

Motivation

Flows of engineering relevance is at high Re

Wall bounded flows, wake and shear flows

The cost of simulations that resolve all the scales of motion is of the
order of Re
3

Almost 90% of this cost is a result of attempting to explicitly resolve
near
-
wall boundary layers

Near
-
wall turbulence contains many small, energy containing,
anisotropic scales that should be resolved

DNS Computations of channel flows

18 B grid points, Re
t

= 2003 (Hoyas
et al
., 2006)

DNS Computations of turbulent separated flows

151 M grid points, Re
t

= 395 (Marquillie
et al.
, 2008)

DNS at lower Reynolds number (Experiment at Re
t
= 6500)

Computational Combustion Lab
Aerospace Engineering

Motivation

Conventional LES requires very high near
-
wall resolution

Near
-
wall Models

Use algebraic relationships to compute wall stresses

Resolution requirement reduced significantly

Additional source of errors due to the modeling the dynamics in
the near
-
wall region

Zonal Approaches

Two Layer Approach

Solves boundary layer equations and/or
employ local grid refinement

RANS
-
LES Approach

Uses RANS near the wall and LES in
the core region

Most of the cost
-
effective approaches do not properly resolve the
turbulent velocity fluctuations near the wall

Here, a two
-
scale approach for high
-
Re flows is discussed

attempts to resolve near
-
wall fluctuations

Computational Combustion Lab
Aerospace Engineering

Multi
-
Scale Simulation Approaches

Multi
-
scale approaches:

Dynamic multilevel method (Dubois, Temam et al.)

Rapid Distortion Theory SS model (Laval, Dubrulle et al.)

Variational multiscale method (Hughes et al.)

Two
-
level simulation (TLS*) (Kemenov & Menon),
extended for compressible flows (Gungor & Menon)

Simulate both LS and SS fields explicitly

Computed SS field provides closure for LS motion

All use simplified forms of SS equations

Some invoke eddy viscosity concept for SS motions

TLS simulates the SS explicitly inside the LS domain

*Kemenov and Menon, J. Comp. Phys., Vol. 220 (2006), Vol. 222 (2007)

Gungor and Menon, AIAA
-
2006
-
3538

Computational Combustion Lab
Aerospace Engineering

Two
-
Level Simulation: Key Features

Simulate both large
-

and small
-
scale fields simultaneously

large
-
scales
(LS) evolve on the 3D grid

small
-
scales

(SS) evolve on 1D lines embedded in 3D domain

3D SS equations
collapsed
to 3x1D equations with closure

Scale Separation approach employed

No grid or test filtering invoked

No eddy viscosity assumption invoked

High
-
Re flows simulated using a “relatively” coarse grid

Efficient parallel implementation needed

Cost becomes acceptable for very high
-
Re flow

Potential application to complex flows

Computational Combustion Lab
Aerospace Engineering

Two Level Grid in the TLS
-
LES Approach

Small scale equations are solved on three 1D lines embedded in the 3D domain

D
x
LES

D
z
LES

D
y
LES

D
y
SS

D
x
SS

D
z
SS

x

y

z

Large
-
Scale Grid

Small
-
Scale Grid

Resolution requirements

Number of LES control volumes:

N
LES
3

N
LES
<N
SS

Grid points for TLS
-
LES:

N
LS
3

+ 3N
LS
2
N
SS

N
LS
<N
LES
, N
LS
<<N
SS

Grid points for DNS:

N
SS
3

Computational Combustion Lab
Aerospace Engineering

TLS v/s LES

Two degrees of freedom in Conventional LES

Filter Width and Filter Type

Two degrees of freedom in TLS:

Sampling/Averaging Operator (SS <=> LS)

Interpolation Operator (LS <=> SS)

TLS does not require commutativity to derive LS Eqns.

Full TLS approach described earlier

isotropic turbulence, free shear and wall
-
bounded flows*

Here, a new hybrid TLS
-
LES approach demonstrated**

Application to wall bounded flows with separation

* Kemenov and Menon, J. Comp. Phys. (2006, 2007)

** Gungor et al., Advances in Turbulence XI (2007)

Computational Combustion Lab
Aerospace Engineering

TLS

Scale Separation Operator
L

Exact Field is split into LS and SS fields:

Continuous large scale field is defined

(,) (,)
(,)
(,)
L
L
k
u x t L u x t
F L u x t
F u x t
D D
D

 

 
(,) (,) (,)
L S
u x t u x t u x t
 
SS field is defined based on LS
field from decomposition:

(,) (,) (,)
S L
u x t u x t u x t
 
Sampling at LS grid nodes

:
L
D
Interpolation to the SS nodes

:
F
D
Computational Combustion Lab
Aerospace Engineering

Fully resolved signal (black) from a 128
3

DNS of
isotropic turbulence study.

The resolved field is represented with a 16 grid point.

The top hat filtered LES field (red) obtained by taking a
moving average of the fully resolved field over 8 points.

The TLS LS field (green) truncated from the fully
resolved signal.

The TLS SS field (blue) obtained by subtracting the LS
field from the fully resolved field.

The longitudinal energy spectra of a fully resolved signal (black) and (a) LES
energy spectra (red), (b) The TLS LS (green) and SS (blue) energy spectra.

TLS has higher
spectral support

A priori analysis of scale separation operators (LES and TLS)

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES Wall Model

The TLS equations are used in the near
-
wall region

The LES equations are used in the outer flow

All zonal approaches (Hybrid RANS
-
LES) use some form of
domain decomposition

Hybrid TLS
-
LES uses functional decomposition

No need for interface boundary conditions

Need to determine the transition region dynamically

RANS

LES

prescribed
y interface

Hybrid RANS
-
LES Strategy

TLS

LES

prescribed y interface
for wall
-
normal lines

TLS
-
LES Strategy

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES Formulation

Scale Separation

Hybrid TLS
-
LES scale separating operator
R

operator that blends the LES operator
F

with the TLS operator
L

R

= k
F

+ (1
-

k)
L

LES operator F is the standard filtering operator

k is a transition function relating TLS and LES domains

1
if
0
TLS
TLS
y Y
k
y Y

1 2
1
2 2
1
1 tanh tanh
1 2
2
c y d c
k c
c y d c
 
 

 
 
 
 
 
 
 
 
Step Function

Tanh

Function

Computational Combustion Lab
Aerospace Engineering

Application of additive scale separation operator

Velocity components

Turbulent stress

Hybrid TLS
-
LES Equations

Hybrid Terms

(,) (,) (,) 1 (,)
R F L
i i i
u x t R u x t ku x t k u x t
   

1
1
R F L
ij ij ij
F L
F F R R L L R R
i j i j i j i j
k k
k u u u u k u u u u
t t t
  
   
    
   

Hybrid terms also in RANS
-
LES formulation (Germano, 2004)

Combination of time and space operation

Here, the hybrid terms appear due to LES and TLS combination

both are space operators !!

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES Equations

Resolved / Large Scale Equations

0,
R
i
i
u
x

Continuity:

Momentum:

2
2
R
R R
R
R
ij
R R
i i
i j
i i j i
u u
p
u u
t x x x x
t

 
 
 
    
 
    
( ) ( )
R R R R
ij i j i j
uu u u
t
 
The unresolved term in the momentum equation

1/2
2
2
3
  D 
F
ij t sgs ij ij sgs
C k S k
t 

  
R R R
L R S S R S S
ij i j i j i j
u u u u u u
t
TLS:

LES
:

1 1
F L
R F L F F R R L L R R
ij ij ij i j i j i j i j
k k k u u u u k u u u u
t t t
   
       
   
Specific closures
for each model

The scale interaction terms are closed if the small scale field is known

any SGS model

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES Equations

Small Scale Equations

Represents the smallest scales of motion

“Hybrid TLS
-
LES SS domain”

discrete set of points along 3
-

1D lines

3D evolution of small
-
scales in each line

Full 3D SS equations “collapsed” on to these 1D lines

Cross
-
derivatives modeled based on
a priori

DNS analysis

Channel and forced isotropic turbulence (Kemenov & Menon, 2006, 2007)

Explicit forcing by the large scales on these 1D equations

0,
S
i
i
u
x

2
2
S S
S
R S R S R
i i
i i j j i
j i j
u u
p
u u u u F
t x x x

 
 
      
   
Continuity:

Momentum:

Computational Combustion Lab
Aerospace Engineering

Numerical Implementation of SS Equations

1) Approximate LS field on each 1D SS line by linear interpolation

2) Evolve SS field from zero initial condition until the SS energy
matches with the LS energy near the cut off

3) Calculate the unclosed terms in the LS equation

Time evolution of the SS velocity and SS spectral energy

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES of Channel Flow

Mean velocity profiles demonstrates the
capability of the model

Wall skin friction coefficient provides

good agreement with DNS

well comparison with Dean’s correlation

Computational Combustion Lab
Aerospace Engineering

Re
t
=‱㈰〠

Re
t
=‵90

3
-
D energy spectra

Red line : Instantaneous energy spectra

Blue line : Volume average spectra

Black line: k
-
5/3

slope

Re
t
=′400

Hybrid TLS
-
LES approach recovers both
LS and SS spectra near the wall

Hybrid TLS
-
LES of Channel Flow

Computational Combustion Lab
Aerospace Engineering

Numerical Solver

Incompressible Multi
-
domain Parallel Solver

4
th

order accurate kinetic energy conservative form (used here)

5
th

order accurate upwind
-
biasing for convective terms

4
th

order accurate central differencing for the viscous terms

Pseudo
-
compressibility with five
-
stage Runge

Kutta time stepping

Implicit time stepping in physical time with dual time stepping

DNS, LES (LDKM), TLS
-
LES

Computational Combustion Lab
Aerospace Engineering

Turbulent Channel Flow

Computational Combustion Lab
Aerospace Engineering

Turbulent Channel Flow

Coarse DNS

192 x 151 x 128

Well prediction of the mean velocity,
turbulent velocity fluctuations and
turbulent kinetic energy budget.

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES of Separated Channel Flow

Hybrid TLS
-
LES(~0.18M) and LES(~1.6M) at Re
t

㴠=㤵

DNS(~151 M) at Re
t

= 395 by Marquillie
et al.
, J. of Turb., Vol. 9, 2008

Experiment at Re
t

= 6500 by Bernard
et al.
, AIAA J., Vol. 41, 2003

Spatial resolution (%75 coarser than DNS)

TLS
-
LES
-
LS (64 x 46 x 64) :
D
x
+LS

= 77.4,
D
z
+LS

= 19.2,
D
y
+LS

= 5.4

TLS
-
LES
-
SS (8 SS points/LS):
D
x
+SS

= 9.6,
D
z
+SS

= 2.4,
D
y
+SS

= 0.68

DNS (1536 x 257 x 384) :
D
x
+

= 3,
D
z
+

= 3,
D
y
+
|
max

= 4.8

Inflow turbulence from a separate LES channel study at Re
t

㴠=㤵

Total vorticity on a spanwise plane (LES)

Streamwise vorticity on a horizontal plane (LES)

Computational Combustion Lab
Aerospace Engineering

Spanwise line in the
separation region

Time evolution of the SS velocity

Computational Combustion Lab
Aerospace Engineering

SS iterations: 20

SS vorticity magnitude isosurfaces colored with SS streamwise velocity

SS iterations: 100

SS iterations: 300

SS evolution effect on the instantaneous flow

simulations on each line

optimal parallel approach

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES of Separated Channel Flow

Hybrid TLS
-
LES grid is chosen very coarse
deliberately

Hybrid TLS
-
LES C
p

shows good agreement with DNS

~%30 off from experiments (higher Re) for all studies

Hybrid TLS
-
LES C
f

in reasonable agreement with DNS and LES

Separation is not properly predicted due to coarse LS resolution

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES of Separated Channel Flow

Streamwise velocity fluctuation

Wall
-
normal velocity fluctuation

The authors would like to thank Dr. J.
-
P. Laval for providing the DNS data

DNS
-
151M (circles and shaded contours), TLSLES (red), LES (green)

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES of Asymmetric Diffuser Flow

Hybrid TLS
-
LES(~0.25M) and LES(~1.8M) at Re
t

= 500

LES(~6.5 M) by Kaltenbach
et al.
, J. of Fluid Mech., Vol. 390, 1999

Experiment by Buice and Eaton, J. of Fluids Eng., Vol. 122, 2000

The main features of this flow

A large unsteady separation due to the APG

A sharp variation in streamwise pressure gradient

A slow developing internal layer

Inflow turbulence from a separate LES channel study at Re
t

= 500

Inclination angle: 10
0

Computational Combustion Lab
Aerospace Engineering

Spatial resolution

TLS
-
LES
-
LS (110 x 56 x 40) :
D
x
+LS

= 54,
D
z
+LS

= 50,
D
y
+LS

= 5.72

TLS
-
LES
-
SS (8 SS points/LS) :
D
x
+SS

= 6.7,
D
z
+SS

= 6.2,
D
y
+SS

= 0.72

LES (278 x 80 x 80) :
D
x
+

= 25,
D
z
+

= 25,
D
y
+

= 0.98

LES by Kaltenbach et al., 1999 (590 x 100 x 110)

Step function (

, F:
LES
, L:
TLS operator)

pre
-
defined interface,
Y+
TLS

= 152

1
if
0

TLS
TLS
y Y
k
y Y

1
  
R F L
ij ij ij
k k
t t t
Hybrid TLS
-
LES of Asymmetric Diffuser Flow

Isosurfaces of the second invariant of the velocity gradient
tensor colored with local streamwise velocity predicted
with LES model

TLS

TLS

LES

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES of Asymmetric Diffuser Flow

C
p

along the lower and upper wall predicted reasonably well

Hybrid TLS
-
LES shows reasonable agreement with the experiment

Skin friction coefficient over the upper flat wall displays a strong drop and a long
plateau starting near the separation region in the bottom wall, and a more

Overall, TLS
-
LES shows ability to predict separation regions without any model
changes

Computational Combustion Lab
Aerospace Engineering

Hybrid TLS
-
LES of Asymmetric Diffuser Flow

The total pressure decreases 30% in the streamwise direction due
to frictional losses.

Mean velocity predicted reasonably with the hybrid TLS
-
LES
model

Separation location agrees well

But reattachment is observed further downstream

Exp. (symbols), TLS
-
LES (dashed lines)

LES (solid lines)

Computational Combustion Lab
Aerospace Engineering

Conclusion and Future Plans

A generalized hybrid formulation developed to couple TLS
-
LES

New hybrid terms identified but they still need closure

TLS as a “near
-
wall” model for high
-
Re flows used in a TLS
-
LES
approach without the hybrid terms

Reasonable accuracy using “relatively” coarse LS grid

Potential application to complex flows with separation

Efficient parallel implementation can reduce overall cost

Next Step

Analyze the hybrid terms in the TLS
-
LES equations and develop
models for hybrid terms

A priori analysis of SS derivatives for arbitrarily positioned SS lines