SCAT 1st Latin-American Summer School

January 9th, 2007

Introduction to Turbulence and

its Numerical Simulation
Charles Meneveau
Department of Mechanical Engineering
Center for Environmental and Applied Fluid Mechanics
Johns Hopkins University

Objectives of course (I & II):

Give an introduction to turbulent flow
characteristics and physics,
Provide a first, basic understanding of standard
turbulence models used in Computational
Fluid Dynamics (CFD)
Provide basic understanding of Direct Numerical (DNS)
and Large Eddy Simulation (LES)
Illustrate state-of-the-art subgrid model for LES
and applications.

Prerequisites:
Basic Fluid Mechanics,
Tensors and Index Notation

Mechanical Engineering

Outline (I):

Turbulent flows:

Overview of turbulent flow characteristics,

Reynolds decomposition,
Turbulence physics and energy cascade,
Turbulence modeling for CFD:
Eddy-viscosity and k-! model
Filtering, Large Eddy Simulation (LES)
Direct Numerical Simulation (DNS)

multiscale,
mixing,
dissipative,
chaotic,
vortical
well-defined statistics,
important in practice

Outline (II):
Smagorinsky model and coefficient calibration,
Non-universality and problems in complex flows,
Dynamic model and applications

From: Multimedia Fluid Mechanics, Cambridge Univ. Press

Turbulent flows:

multiscale,
mixing,
dissipative,
chaotic,
vortical
well-defined statistics,
important in practice

Turbulence in atmospheric boundary layer

e.g. 50 m

From: Multimedia Fluid Mechanics, Cambridge Univ. Press

Turbulence in reacting flows:

Premixed flame in I.C.
engine, combustion

Turbulence in aerospace systems:

LES of flow in thrust-reversers

Blin, Hadjadi & Vervisch (2002)
J. of Turbulence.

Numerical simulation of flame

propagation in decaying
isotropic turbulence

Turbulence in thermofluid equipment:

Turbulence in turbomachines:
Optical velocity field measurements in index-matched axial pump
Phase-averaged turbulent kinetic energy distribution
(Uzol, Katz & Meneveau, J. Turbomach. 2003)

Lattice of wakes:

Superimposed Image of 10 Different Rotor Phases

at 500 rpm and at the Mid-span location. x=0 is at
rotor leading edge. The stage length is Ls = 203
mm.

From: Multimedia Fluid Mechanics, Cambridge Univ. Press

Simplest turbulence: Isotropic decaying turbulence

JHU Corrsin wind tunnel
Active Grid
M=6"

Physical quantities describing fluid flow

Density field
Velocity vector field
Pressure field
Temperature field (or internal energy, or enthalpy etc..)

Physical laws governing fluid flow

Active Grid
M=6"
Contractions

u2
x2

Test Section

u1
x1

Conservation of mass
Newtons second law
First law of thermodynamics
Equation of state
Some constraints in closure relations from second law of TD

Flow

Navier Stokes equations for a Newtonian, incompressible fluid

Navier-Stokes equations, incompressible, Newtonian

Turbulence: Reynolds decomposition

!u j
=0
!x j

!u j
=0
!x j
!u j
!t

!uk u j
!xk

!u j

1 !p
="
+ $% 2u j + g j
# !x j

aj =

!t

!uk u j
!xk

="

1 !p
+ $% 2u j + g j
# !x j

Fj
m
t

Turbulence: Reynolds decomposition

Reynolds stress

!u j
=0
!x j
!u j
!t

!uk u j
!xk

Energy cascade

1 !p
="
+ $% 2u j + g j
# !x j

Spectral energy tensor

Isotropic turbulence

Reynolds equations:
!u j
!x j
!u j
!t

!u j uk
!xk

Kolmogorov spectrum (1941)

=0
t

="

1 !p
!
+ $% 2u j + g j "
u j uk " u j uk
# !x j
!xk

 Kinematic Reynolds stress (minus):

!u j
!x j
!u j

!t

!u j uk
!xk

Kinematic Reynolds stress (minus):

!u j
!x j

=0
="

1 !p
!
+ $% 2u j + g j "
u j uk " u j uk
# !x j
!xk

! Rjk = u j uk " u j uk
Written as velocity co-variance tensor:

Kinematic Reynolds stress (minus):

!u j
!t

!u j uk
!xk

=0
="

!u j uk
!xk

="

!& Rjk
1 !p
+ $% 2u j + g j "
# !x j
!xk

! Rjk " u #j uk#

!x j

!t

Deviatoric (anisotropic part):

! Rjk = (u j + u "j )(uk + uk" ) # u j uk = u j uk + u j uk" + uk u "j + u "j uk" # u j uk = u "j uk"

!u j

!u j

=0

!u j
!t

!u j uk
!xk

!& Rjk
1 !p*
2
="
+ $% u j + g j "
# !x j
!xk

Spectral representation of co-variance tensor:

u !j uk! =

!& Rjk
1 !p *
+ $% 2u j + g j "
# !x j
!xk

1 R
! Rjk " # Rjk $ # mm
% jk
3

1
(2" )3

$$$ #

jk

(k1 , k2 , k3 ) d 3k

! jk (k1 , k2 , k3 ) : Spectral tensor of turbulence

(how much energy there is in each wave vector k)

Unknowns: mean velocity and
pressure field

u1 (x1 , x2 , x3 ,t),
u2 (x1 , x2 , x3 ,t),
u3 (x1 , x2 , x3 ,t),
p(x1 , x2 , x3 ,t)

Closure required for Reynolds

stress tensor: express stress
in terms of mean velocity field

! Rjk = func(u)

In homogeneous isotropic turbulence (simplest case, with no preferred

directions) the spectral tensor function of a vector can be expressed based on
a single scalar function of magnitude of wavenumber, E(k):

! jk (k) =

k j kk (
1 %
# $ 2 * E(k)
2 ' jk
4" k &
k )

 What is the dependence of energy density E(k)

with wavenumber?

Turbulence Physics: the energy cascade

(Richardson 1922, Kolmogorov 1941)

We now discuss the energy cascade

and Kolmogorov theory of turbulence

From: Multimedia Fluid Mechanics, Cambridge Univ. Press

Turbulence Physics: the energy cascade

(Richardson 1922, Kolmogorov 1941)

!K

Turbulence Physics: the energy cascade

(Richardson 1922, Kolmogorov 1941)

which feed on their velocity,

L
Constant Flux of energy
across scales: !

and little whorls have lesser whorls,

and so on to viscosity (in the molecular sense)

!K

1 ! !
ui ui
3

u!2
u!2
u! 3
!
!
Time L / u !
L

Injection of kinetic energy into

turbulence (from mean flow)

Big whorls have little whorls,

u! "

!!

u" 3
L

!!

u "(r1 )3
r1

!!

u "(r2 )3
r2

! = 2"

#ui$ % #ui$ #u $j (
+
#x j '& #x j #xi *)

Dissipation of kinetic energy into

heat (due to molecular friction)

Solid line is equivalent of a 3D radial spectrum equal to

Turbulence Physics: the energy cascade

(Richardson 1922, Kolmogorov 1941)

E(k) = cK ! 2/ 3k "5/ 3 , cK ~ 1.6

Solid experimental
support for K-41:

Injection of kinetic energy into

turbulence (from mean flow)

1
L

Constant Flux of energy

across scales: !

kL =

E = f (k, ! )
Dimensional Analysis (Pi-theorem: 3-2=1):

k
k
1
k! =
!K

E k5/3
= const
! 2/3

E(k) = cK ! 2 / 3 k "5 / 3

Supports (approximately)
the notion that ! is the only
relevant physical scale in
the inertial range,
+ L at large scales
+ " at small scales

Dissipation of kinetic energy into

heat (due to molecular friction)

Back to Closure problem for Reynolds equations:

!u j
!t

!u j uk
!xk

="

!& Rjk
1 !p *
+ $% 2u j + g j "
# !x j
!xk

! = func(u)
R
jk

Kolmogorov:
Characterization of turbulence: minimum 2 variables
Turbulent kinetic energy:

!!

u" 3
L

!!

K 3/2
L

Need (K,L) or (!,L) or (K,!)

In analogy with molecular friction:
% $u j $u (
k
! Rjk = "# T '
+
*
$x j )
& $xk

Eddy-viscosity

! T ! (velocity scale) " (length scale)

K=

1
1
ui!ui! = u ! 2
2
2

Transport equations for K(x,t) and !(x,t)

Eddy-viscosity scaling:

Launder & Spalding (1972) - but earlier (1940s) Kolmogorov K-# model

! T ! (velocity scale) " (length scale)

~ u!
~K

~L
1/2

) !K )
!u
!K
!K
! & & $T
+ uk
= "# ijR i +
+ $+
((
+ ",
!t
!xk
!x j !x j ' ' % k
* !x j *

u!

K 3/2
~
!

K2
! T = C
"

K 2 (x,t)
! T (x,t) = C
" (x,t)

K-! model for turbulence mean flow predictions:

!u j
!x j
!u j
!t

!u j uk
!xk

=0
="

% !u j !u ( .
1 !p
! +
k
-($ + $ ) '
+ gj +
+
*0
# !x j
!xk - T
!x j ) 0
& !xk
,
/

) !K )
!u
!K
!K
! & & $T
+ uk
= "# ijR i +
+ $+
((
+ ",
!t
!xk
!x j !x j ' ' % k
* !x j *

&
) !, )
!u ) ,
!,
!,
! & & $T
,
+ uk
= C, 1 ( "# ijR i + +
+ $+
((
+ " C, 2,
!t
!xk
!x j * K !x j ' ' % ,
K
* !x j *
'
+ boundary (& initial) conditions

&
) !, )
!u ) ,
!,
!,
! & & $T
,
+ uk
= C, 1 ( "# ijR i + +
+ $+
((
+ " C, 2,
!t
!xk
!x j * K !x j ' ' % ,
K
* !x j *
'

C = 0.09

Empirical calibrations:
5 adjustable coefficients
!K

C! 1 = 1.44
C! 2 = 1.92

" K = 1.0
" ! = 1.3

Regions of large vorticity in isotropic turbulence

Direct Numerical Simulation:

N-S equations:
!u j
!t

!u k u j
!xk

="

!p
+ #$ 2 u j + g j
!x j

World-record DNS (Nagoya

group in Japan):

!u j
=0
!x j
Moderate Re
(~ 103),
DNS possible

High Re
(~ 107),
DNS impossible

On Earth Simulator (2003)

4,0963 grid points
~ 2 Terabytes at each time-step
Thousands of time-steps

Source: Kaneda & Ishihara,

J. of Turbulence, 2006 Taylor & Francis

From: Multimedia Fluid Mechanics, Cambridge Univ. Press

Large-eddy-simulation (LES) and filtering:

!u
j
!t

k
+u

!u
j
!x k

="

!p
!
j "
+ #$2 u
%
!x j
!x k jk

G(x): Filter
$

Large-eddy-simulation (LES) and filtering:

G(x): Filter

N-S equations:
!u j
!uk u j
!p
+
="
+ #$2 u j
!t
!x k
!x j

!u j
=0
!x j

Filtered N-S equations:

j
!u
!uk u j
!p

+
="
+ #$2 u j
!t
!x k
!x j
!u
j
!t

k
+u

!u
j
!x k

="

!p
!
j "
+ #$2 u
%
!x j
!x k jk

where SGS stress tensor is:

! ij = uiu j " ui u j

Useful references:

S. Pope: Turbulent Flow

(Cambridge Univ. Press, 2000)
J. Ferziger & M. Peric: Computational Methods for
Fluid Dynamics (Springer, 1996)