You are on page 1of 42

(1/9)

Turbulence
In his second edition of the classic book "Hydrodynamics", published
in 1895, Sir Hoarse Lamb wrote (which remained in the sixth edition
published in 1932, and undoubtedly appropriate for today also)
" It remains to call attention to the chief outstanding difficulty of our
subject…."
"…It has already been remarked that the neglect of the terms of the second
order (u∂u/∂x, &c.) seriously limits the application of many of the
preceding results to fluids possessed of ordinary degrees of mobility.
Unless the vortices, or the linear dimensions involved, are very small the
actual motion in such cases, so far as it admits of being observed, is found
to be very different from that represented by our formulae. For example,
when a solid of 'easy' shape moves through a liquid, an irregular eddying
motion may be produced in a layer of the fluid next to the solid, and a trail
of eddies left behind, whilst the motion at a distance laterally may be
comparatively smooth and uniform"
(2/9)
Turbulence
The generation of "different scales" by the non-linearity can be well illustrated
by the simple non-linear differential equation:
θ θ
θ 0
t x

If we choose  x,0   cos kx, then it is evident that the development of


higher harmonics appears as:
2
 x, t    cos kx  ktsin 2kx
2

which obviously contain higher harmonics.


(6/9)
Turbulence
It is difficult to propose a precise definition for turbulence, and it is more
appropriate to present the properties of turbulent flows rather then defining it.
Some definitions given in the past are:
Continuous state of instability (Monin and Yaglom – 1961)
This was first proposed by Lev Landau; accordingly the flow undergoes an infinite
sequence of bifurcations (instabilities) before it becomes unpredictable and chaotic.
(However, recent work on dynamical systems approach shows that such behavior can
arise only after a finite number of bifurcations.)
Irregular motion in space and time, which is not connected with the Brownian
motion of the particles, so that average statistical values can be discerned (Hinze
1956 and Corrsin 1961).
Irregular motion of fluid that makes appearance during flow over boundaries or in
the presence of velocity shear (Karman 1937; Taylor 1937) ---- Vortex dynamics
are important.
Unpredictable flow in the sense that, for given initial conditions, motions are
amplified as to give impossible a precise definition of the evolution (Lesieur 1986).
“Sinuous” motion (Reynolds 1883).
(7/9)
Turbulence
Some of the properties and traits of turbulent motion are as
follows:
General randomness (although some special cases show some
periodic large structure – irregular and random, no deterministic
approach)
Rationality – i.e. lots of three dimensional vorticity fluctuations; if
no vorticity, the flow cannot be called turbulent
Straininess – i.e. lots of strain-rate fluctuations, three-
dimensionality (although some special cases show some two-
dimensional, large structures – Brown and Roshko 1973 JFM,
Chandrasuda et al 1978 JFM)
Continuous range of “eddy” sizes
(9/9)
Turbulence
Viscous thin sub-layer at a smooth, solid boundary
Thin viscous super-layer (Corrsin) at a boundary with non-
turbulent fluid, (entrainment of non turbulent fluid into
turbulent layers)
Mixing of statistically homogeneous scalar containment
(highly diffusive)
Dispersion/transport of statistically inhomogeneous
containment (in addition to mixing)
(1/8)
Momentum Equations
 ,   1, 2; i, j, k  1, 2, 3
u u u 1 p   2 u  2 u 
 u w   jk f j u k       

t x  z  x 
  
x x z 2 

 LH  u*  * u* * u   fLH 


 
*

 
 *   u  w * 
    jk  j u k
 H 
U T t  x
*
z   UH 
2
po p *   2 u*   LH   2 u*
    
 oU H 2 x* LH U H x * x *p U H LH  Lv  z
2

f  2   Coriolis Parameter
~ ~
(2/8)

Momentum Equations
w w w 1 p  2w  2 w
 u w   b    2
t x  z  0 z  
   z 
x x

2
 Lv   LH w* * w
*
* w 
*
p0 p *  b0 Lv  *
    u *  w *
   b
 LH   H
U T t *
x  z.   U
0 H
2
z * 2
 H 
U

 Lv  2w
2
   2w
    2
 LH U H  LH  x x z 
  

U H LH UH 
Re  Reynolds Number Ro  ~ Rossby Number
 fLH LH
Lv p0
 Length scale Ratio Eu  Euler Number
LH  0U H2

b0 LH Bulk Richardson UH Internal Froude


Ri B  Fri 
U H2 Number b0 LH . Number
Definition – 3D Turbulence

Chaotic Motions, where the inertial vortex forces


dominates
Viscous, Coriolis and Other Body forces
(4/9)
Turbulence
Lewis Fry Richardson (1920)
"Big whorls have little whorls, which feed their velocity; Little
whorls have smaller whorls, and so unto viscosity."

Kolmogorov Scales

U, L U1 , L1 U 2 , L2 U k , Lk

U3
 UL U 1 L1 U 2 L2 U k Lk
L  1  1 1 ~1
v v v v
(3/9)
Turbulence
Inertia
Viscous Inertia
Viscous

 2u ~  U UL
2
2 u  u ~ U  Re
L L 
(5/9)
Turbulence
The dimensionless parameters governing the dissipating scales are and
(rate of dissipation), and hence the Kolmogorov scales can by written as:

E k 
14
v 3
 Spectrum
 k    (length)
 
14
v k  () (velocity)
v
and Tk  ( )1 2 (time)
 k  2
2 lk
L
L
It was at a meeting of the British Association in London in 1932
that I remember Hoarce Lamb remarked

“I am an old man now, and when I die and go to Heaven there


are two matters on which I hope for enlightenment. One is
quantum electrodynamics, and the other is the turbulent motion
of fluids. And about the former I am really rather optimistic.”

– Sir Sydney Goldstein


Annual Reviews of Fluid Mechanics, Vol. 1, 1
(8/9)

In Heavens……

From Physics Today, May 2000.


Statistical Averaging
Ensemble Averaging:
Conduct a large number (N) of identical experiments (realizations)

Ensemble Average  
U x ,t 
~

N

n =1
u~n ( x ,t )
N
`

Fluctuation   ~
u' x , t  u x , t   U x , t 
~

 
12

 u~ ( x, t) - U(x, t) 


1 1 N
2
u x , t   u  2 
2
r.m.s. n
N 1
Statistically Stationary Flows
U
(x,t) = 0  U(x,t) = U(x)
t ~ ~


U(x, t) = 0  U(x, t) = U(t)
x j ~ ~

T
1
U(x) = lim
~ T  T
 d ˜u(x,t + )
~
-T Ergodicity – all converge to
x
ensemble averaging at sufficiently
1 large averaging periods
U(t) = lim
x 0 2x 
-x
dru˜ (x + r, t)
Reynold’s (1896) Averaging
f g f g

af  a f If a is a constant

fg fg

f  f
 If s  x1 , x2 ,...
 s s
Examples:

u~iu~j  (U i  u'i )(U j  u' j )  U i U j  uiu j for U i u 'i  U i u 'i  0


Reynolds Averaged Equations
Continuity Equation:
u˜ j
0
x j

(U j  u j )  U j
 (U j  u' j )  0
x j x j x j
Balance of a Transferable
Quantity (a scalar)
P˜ P˜ 2P˜
 u˜ j D  F˜ p
t x j x jx j

P P  P
or U j  (D  p' u' j )  F p
t  x j x j x j
(1/4)

Fluctuations
p    2 p
 ( Pu j  U j p  u j p  pu j )  D
' '

t x j x jx j


 p2 2 
U j
 p2 2  
 u 'j p
P 
  
p2u 'j 2  D p
p
t x j x j x j x jx j

(A) (B) (C) (D) (E)

(A) represents the rate of p'2 change of at a given point and (B) represents the advection
of p'2 following a fluid particle moving at U . (A) and (B) together represent the time
variation of p'2 following a fluid particle, moving at U .
p
The third term  u ' j p ' represents the production of scalar fluctuations due to the
x j
interaction of the mean gradient and the turbulent flux u ' j p'.
(2/4)
p'2
p'  2 p'  2

D D 2  D p  p
The term (E) is of special interest  x j x j  x j x j x j  x j
(E1 (E2
) )

(a) Binary distribution in space. Schematic sketch of turbulent mixing of


a contaminant, both without and with
molecular diffusion. These represent a
small sample from a large statistically
homogeneous field.
t t1 > t2 >
0 t0 t1
(b) Amplitude on a cut. (c) Effect of molecular transport for t > t0.

t t1 > t2 > t t1 > t2 >


0 t0 t1 0 t0 t1
(3/4)

Separation Distance 


Diffusion Time 2 D

u  v 
1
4
or  
1
3 Separation Time  u

 v 
3
1
4
or 
(4/4)

Spectrum
u   , l  ~ l  3
1

t~  
l
1
3

 
l~ v
3
1
2

 
t~ v
1
2

Sc  v
D

 1 ;   vD
2


1
4
Batchelor Scale

Sc  1 ;   D 
1
3 4
Obukhov-Corrsin Scale

Momentum Equation
u˜ i ˜ ˜ui 1 p˜ ˜  2 u˜ i
 Uj   ij k j uk   b i3  
t x j 0 x i x jx j

~ ~
~
ui  Ui  u' i , b  b  b and P  p  p

Ui  u i  1 P  p'  2 U  ui 


t


x j
 
Ui  ui  U j  u j   ijkj Uk    x  b  b' i3  x x
o i j j
Turbulent Kinetic Energy Equation
u' 2i u' 2i
   '2 ' ' 
2 U 2  u u
' ' U i  ' p uj 2 u' i
j i j  u j ui 
2
 b' u' 3  u' i
t x j x j x j   0  x j x j

(A) (B) (C) (D) (E) (F)

The terms of the above kinetic energy equation can


be interpreted as follows:
(A) rate of increase of TKE a given point
(B) gain of TKE by the advection caused by mean flow
(C) production of turbulent kinetic energy by the working of mean
flow on the Reynolds stresses
(D) transport of turbulent kinetic energy by the advection by the
turbulent fluctuations u 'j u 'j 2  and turbulent pressure
(1/2)

TKE Dissipation
  2s'ij s'ij

1  u'i u' j 
Where s'ij  
2  x j x i 

 2 u' i2 
u' i u' i    u' i u' j 
2

ν  ν 2    
x x j  x jx j x ix j  F3
j       
F  F1 F2 
(2/2)
  2s' ij s' ij   i  i

u1 2
  15  For isotropic turbulence
x1 

u1 2 u12


  
x1  2

Microscale Reynolds Number


u
R 

Isotropic Turbulence
A flow in which any relationship
between turbulent quantities (in averaged
sense) is invariant under rotation of the
coordinate system, translation of the
coordinate system and under the reflection
with respect to the coordinate system.
(1/2)
Spectral Analysis of Turbulent Flows
In spectral description:
turbulent signals are decomposed into various
spatial length and time scales
it is possible to obtain some understanding of the
behavior of various scales
the essential property of turbulence is the exchange of
energy between various scales.
Different wave lengths or frequencies resulting
from Fourier analysis are usually interpreted as
"eddies", although some might think that an eddy
is an isolated blob of fluid.
(2/2)
Spectral Analysis of Turbulent Flows
 i 2 nx
u x   a ne
L , n  1,2 ,3....


i 2nx
L/2
ux e L dx
1
which is the Fourier representation of the signal. Further,an  
L L / 2
2n 
Now define  k and an L  U k 
L
  L/2
u x   a Uk    

ikx ikx
Then ne and u x e dx
 L / 2

2 4 6
Now that k changes as , , ... as n takes n  1,2,3.......
L L L
2
So that dk 
L
G.I. Taylor
ui (t )  ui (t ) for -T<t<T

and = 0 if t >T


Then, ui (t )   uˆi ( )eit d




T
1
 i
it
and uˆi ( )  u (t ) e dt
2 T

  i
Hence, ui (t )u j (t   )   ( ) e d
*
ij

The Fluid Dynamics group in the Cavendish Laboratory, April
1955. Front row: Ellison, Townsend, Taylor, Batchelor, Ursell,
van Dyke. Middle row: Barua, Thomas, Morton, Thompson,
Philips, Bartholomeusz, Thorne. Back row: Nisbet, Grant, Hawk,
Saffman, Wood, Hutson, Turner. From Huppert (2000).
(1/2)
One and Three
Dimensional Spectra

i k r
u 'i ( x)u ' j ( x  r )  Rij r )     ij
 ( k ) e ~ ~
dk
~ ~ ~ ~ ~ ~
 ~

where r  (r1, r2 ,r 3) and k  (k1 , k 2 , k3 )


~ ~


Rij (r ,0,0)1 r     ij
 ( k ) e ik1r
dk1dk 2 dk3
~



Fij (k1 )    ij (k )dk 2 dk3 one-dimensional spectra
~

(2/2)
One and Three Dimensional Spectra

The spectra in the inertial subrange can be written


Ek   E , k , or on dimensional grounds E k    2 3k 5 3
where  is called the Kolmogorov constant, in recognition
of Kolmogorov’s early work.

F11 (k1 ) ~  2 3k15 3


(1/2)
Dispersion of Particles in
Turbulent Fluids
If a gas particle moves in successive displacements of yi ,
N
then the resultant displacement after N steps is y   yi.
1

If the field is homogeneous, the average of y, y  0,

(over a large number of displacements), then the mean square value is

 N
 N  N
y 2    yi   y j    yi   N y 2 .
2

 i 1  j 1  i 1
(2/2)

Dispersion

 v2 t , and hence the distance yt   y0   v2 t dt .


dy t

dt 0

1  
y t     dt v2 t0  t dt    dt v2 t0  t dt  
t
2

T 0  0 
r.m.s. Dispersion
y  2v
2 '2
2  t  R  d   t  R  d 
t*

0
L
t

t*
L

at large times t > t*, RL() is small and negligible

2 '2
2  0

y  2v t  RL  d  2v2'2t L
t*
Eddy Diffusivity

y 2  2v2'2t L  2v2'2 L t 
y    2v   t
2
1
2
'2
2 L
1
2 1
2

K  2v2'2 L
Stratified Turbulence
y t   2v
2 '2
2  tR  d  2 R  d 
t

0
L
0
t
L

Can we use this?


 z 
(1/2)

Effects of Stratification
t t

y
ye ye  ye
t  t0
A
B t  t0

 d 
   g dv  d y2 d
  ye  ye  y
2
 dz  dt dt
1    d  g  d  g 2 N 2 2 d
  ye  y
d '2

2
 g   g dvd      
 0 dv 0
 dz  0  dz  2 2 dt dt

small
(2/2)
Effects of Stratification

1 2 1 2 2
  w ~ N  max
 2 2

2 w
t  y   max ~
N
(1/2)
Diffusion of a Substance
 
   y  y 2
e cdydx
ys2  
 
cdydx

ys s 
d ys2

  yF dydx

s

dt   cdydx
(2/2)
Diffusion of a Substance d ys2
 2k c
dt
60
400

40

300

Second moment (m2)


20
Height (m)

0 200

-20

100
-40

-60 0
0.0 0.2 0.4 0.6 0.8 1.0 0 50 100 150 200
Concentration (scaled) Time (days)

You might also like