Turbulence

Fluid Mechanics
Dρ ∂u ⎛ D ∂ ∂ ⎞
mass +ρ i =0 ⎜⎜ where = +uj material derivative ⎟⎟
Dt ∂xi ⎝ Dt ∂t ∂xi ⎠
Du i ∂τ ij
momentum ρ = i = 1,2,3
Dt ∂x j

cf. Einstein Summation convention - Repeated index is summed

∂ui 3 ∂u ∂u ∂u ∂u
=∑ i = 1+ 2 + 3
∂xi i =1 ∂xi ∂x1 ∂x2 ∂x3

∂τ ij ∂τ ∂τ ∂τ
= i1 + i 2 + i3
∂xi ∂x1 ∂x2 ∂x3

Define pressure to be mean normal stress

p = − 1 (τ 11 +τ 22 +τ 33 )= − 1τ kk
3 3

τ ij = − pδ ij + τ ij(v ) ← viscous stress

(v )
Dui ∂p ∂τ ij 4 eqn.s + 1eqn. of state(& maybe new unknowns)
ρ =− +
Dt ∂xi ∂x j ui (3)
Dρ ∂u ρ (1) 11unknowns
+ρ i =0
Dt ∂xi p (1)
τ ij(v ) (9) (6 since
(6, i symmetrical)
y t i l)

Turbulence & Turbomachinery Lab. Hanyang Univ. 1

 Turbulence

We need equation’s
equation s to relate viscous stress to the motion

We require that this relationship depend only on the properties of the material, not of the motion

e.g. : Newtonian fluid Relation between stress & strain rate is linear

⎛ 1 ⎞ 1 ⎛ ∂u ∂u ⎞
τ ij( v ) = 2μ ⎜ eij − ekk δ ij ⎟ ; eij = ⎜⎜ i + j ⎟⎟
⎝ 3 ⎠ 2 ⎝ ∂x j ∂xi ⎠

Constitutive equation
τ ij =F (eij )

(v ) ⎛ 1 ⎞
Newtonian : τ ij = 2μ ⎜ eij − ekk δ ij ⎟
⎝ 3 ⎠

Viscosity (independent of the flow,dependent, on the fluid)

τ ij : 9 EA (6EA)
eij : 9 EA (6EA)

Most general linear relation

τ ij = cijklkl ekl

Symmetry reduces unknowns from 81 to 27

All but (2) (1 in practice) can be eliminate for isotropic fluid Newtonian

Turbulence & Turbomachinery Lab. Hanyang Univ. 2

 Turbulence

Incompressible
p
Dρ
Dt =0
Derivative of density following fluid material pt. is zero
r ∂ui
⇒ ∇ ⋅ u = 0 or =0
∂xi
⎛ 1 ⎞
τ ij = − pδ ij + 2μ ⎜ eij − ekk δ ij ⎟
⎝ 3 ⎠
0 0 incompressilbe.
1 ⎛ ∂u ∂u ⎞
ekk = ⎜⎜ k + k ⎟⎟
2 ⎝ ∂xk ∂xk ⎠

⎛ ∂u ∂u j ⎞
⇒ τ ij = − p δ ij + μ ⎜ i + ⎟ for incompressible.
⎜ ∂x ⎟
⎝ j ∂ xi ⎠
∂∂P ∂
ρ
Dui
=−
P
+ (2μeij )
Dt ∂xi ∂x j
∂P ∂ ⎡ ⎛⎜ ∂ui ∂u j ⎞⎟⎤
=− + ⎢μ + ⎥
∂xi ∂x j ⎢⎣ ⎜⎝ ∂x j ∂xi ⎟⎠⎦⎥
∂ui
=0
∂xi

Dui 1 ∂P ∂ 2 ui ∂ui
if μ=constant throughout flow , = − + ν , =0
Dt ρ ∂xi ∂x j ∂x j ∂xi

Turbulence & Turbomachinery Lab. Hanyang Univ. 3

 Turbulence

Reynolds
y decomposition
p
Impossible to approach turbulence deterministically ?
split flow into a mean and a random part (fluctuation part)

u~i = U i + ui

Instantaneous mean fluctuation

~p = P+ p
~
τ (V) = T (V) + τ (V)
ij ij ij

deal with only constant density case ρ~ = R

ρ =0 : no density fluctuation

∂
(U i + ui ) = 0
∂xi

ρ
∂
(U i + ui ) + ρ (U j + u j ) ∂(U i + ui ) = − ∂ (P + p ) + ∂ Tij(V) + τ ij(V)
( )
∂t ∂x j ∂xi ∂x j

Average these instantaneous equations to get equations governing the mean flow
0 by definition

ρ
∂
(U i + ui ) = ρ ∂ (U i + ui ) = ρ ∂ (U i + ui ) = ρ ∂U i
∂t ∂t ∂t ∂t

Turbulence & Turbomachinery Lab. Hanyang Univ. 4

 Turbulence

∂ (P + p ) ∂P
= : gradient of mean pressure
∂xi ∂xi

∂ ∂ (V )
∂x j
( )
Tij(V ) + τ ij(V ) =
∂x j
Tij : divergence of mean viscous stress

0 0
(U j + u j ) ∂(U i + ui ) = U j ∂U i + U j ∂ui + u j ∂U i + u j ∂ui = U j ∂U i + U j ∂ ui + u j ∂U j + u j ∂ui
∂x j ∂x j ∂x j ∂x j ∂x j ∂x j ∂x j ∂x j ∂x j

e.g.) u ≠ 0
2
Product of fluctuating quantities can not be assumed zero, they may or may not be zero.

Look at continuity equation

0
∂
0= (U i + u i ) = ∂ U j + ∂ u i
∂xi ∂x j ∂xi
∂U i
⇒ 0= ← mean satisfies same continuity equation as does instantaneous
∂xi

Wh t equation
What ti d
does fl
fluctuation
t ti satisfy
ti f ?
∂ ~
ui = 0
∂xi ∂ ~
∂ (ui − U i ) = 0
Ui = 0 ∂xi
∂xi ∂ui
⇒ = 0 ← even fluctuation satisfy same continuity equation
∂xi

Turbulence & Turbomachinery Lab. Hanyang Univ. 5

 Turbulence

∂u i ∂ui ∂u j
B k to u j
Back uj =0 ⇒ ui =0
∂x j ∂xi ∂x j
∂u
uj
∂ui
∂x j
∂u
+ 0 = u j i + ui j =
∂x j
∂
∂x j ∂x j
ui u j ( )
∂ 2 ∂ ∂
i =1 u1 + u1u2 + u1u3
∂x1 ∂x2 ∂x3
∂ ∂ 2 ∂
i=2 u1u2 + u2 + u 2 u3
∂x1 ∂x2 ∂x3
∂ ∂ ∂ 2
i=3 u1u3 + u 2 u3 + u3
∂x1 ∂x2 ∂x3

Since turbulence is 3
3-dimension
dimension u12 , u22 , u32 ≠ 0

In general the other velocity conditions do not vanish either

Equations for mean motion unknowns

∂U i Ui − 3
=0 9 (6) new unknowns P − 1
∂xi
Tij(V ) − 6 (symmetric)
ρ
∂
∂t
U i + ρU j
∂U i
∂x j
=−
∂P
+
∂
∂x j ∂x j
(
− ρ ui u j + Tij(V ) ) ui u j − 6 (symmetric)
total 16
(
Tij(V ) = 2μEij )
The 6 new unknown u i u j mean the problem is no longer closed

closed
l d ; {# off unknown
k = # off equations}
ti } ; solvable
l bl

Turbulence & Turbomachinery Lab. Hanyang Univ. 6

 Turbulence

These u i u j are not stresses (in spite of what you’ve heard). They are only acceleration term rewritten for
convenience on R.H.S. when written this way they appear to play the role of a stress.
For that reason we often refer to them as the “Reynolds stress” ← fictitious stresses

Define Reynolds stresses

Tij( R ) ≡ − ρ u i u j
∂U i
∂t
+U j
∂U i
∂x j
=−
1 ∂P 1 ∂
+
ρ ∂xi ρ ∂x j
[
Tij( R ) + Tij(V ) ]
For equations to be solvable we must have constitutive equations relating Reynolds stress to mean motion.
This is the TURBULENCE PROBLEM.

( R)
Try to write constitutive equation for Tij
Stress should be some functional of history of motion

Tij( R ) = F ( Eij ) ← local assumption

stress at point depends only on history of stream near point
(assumes turbulence at point is determined by strain rate at point)

Ignoring this local problem, let’s postulate a linear relation between Reynolds stress and mean strain rate

Tij( R ) = eijkl E kl

81 new unknowns

Turbulence & Turbomachinery Lab. Hanyang Univ. 7

 Turbulence

We could postulate an isotropic medium and write

TijR = 2 μ t Eij (i ≠ j ) incompressible

turbulent viscosity
(Eddy)

The “medium” in this case is not the fluid, but the flow itself
Turbulent flow are flows !!

Keep going in spite of obvious inconsistence

Turbulence & Turbomachinery Lab. Hanyang Univ. 8

 Turbulence

Can we derive a dynamical equation for Reynolds stress ?

Can we get equation for velocity fluctuation itself ?

Instantaneous
~ ~ (V )
∂u~i ~ ∂u~i 1 ∂P 1 ∂τ ij
+uj =− + ( Navier − Stokes equation ) where τ~ijV = 2 μ~
eij
∂t ∂x j ρ ∂xi ρ ∂x j
123
∂U i ∂U i ∂u ∂u
Uj +uj +U j i + u j i
∂x j ∂x j ∂x j ∂x j
Mean
∂U i ∂U j 1 ∂P 1 ∂Tij
(V )
∂
+U j =− + − ui u j
∂t ∂x j ρ ∂xi ρ ∂x j ∂x j
1424 3

∂u i
uj
∂x j
Recall
u~i = U i + ui
~
p = P+ p

Subtract mean from instantaneous to get equation for fluctuation

Fluctuation

∂u i ∂u ⎡ ∂U i ⎛⎜ ∂u i ∂u ⎞⎤ ∂τ (V )
⎟⎥ = − 1 ∂P + 1 ij
+ U j i + ⎢u j + uj −uj i
∂t ∂x j ⎢⎣ ∂x j ⎜⎝ ∂x j ∂x j ⎟⎥
⎠⎦ ρ ∂xi ρ ∂x j

Reynolds stress

Turbulence & Turbomachinery Lab. Hanyang Univ. 9

 Turbulence

NOTE ① Average of equation for fluctuation is identically zero

② Suppose infinitesimal disturbance
Remaining equation for fluctuation is linear & Reynolds stress is gone (2nd order)

∂u i ∂u i ∂U i 1 ∂P 1 ∂τ ij
(V )
Fluctuation : +U j +uj =− + linear
∂t ∂x j ∂x j ρ ∂xi ρ ∂x j
No Reynolds stress
∂U i ∂U i 1 ∂P 1 ∂Tij
(V )
Mean : +U j =− +
∂t ∂x j ρ ∂xi ρ ∂x j

Equations for infinitesimal disturbance are linear & closed (well-posed)

Equations for mean flow is just the unchanged original viscous equations
- presence of infinitesimal disturbances does not change base flow (laminar solution)
This is so called stability problem (Fluid dynamic stability)
Fluctuation equation can easily be transformed into Orr-Sommerfeld equation

Superimpose disturbance
Dose it grow, decay, remain the same ( neutral ... disturbance )

If infinitesimal disturbance grows, where does additional energy come from?

Develop energy equation for fluctuation
mean fluctuation
fl t ti Ki
Kinetic
ti E it mass =
Energy per unit
1
2
1
2
(
ui ui = u12 + u22 + u32 )
To 2nd order ( multiply linearized equation by u i )
∂ ⎛1 ∂ ⎛1 ∂U i 1 ∂P 1 ∂τ ij
(V )
⎞ ⎞
⎜ ui ui ⎟ + U j ⎜ ui ui ⎟ + ui u j = − ui + ui Kinetic Energy of infinitesimal disturbance
∂t ⎝ 2 ⎠ ∂x j ⎝ 2 ⎠ ∂x j ρ ∂x j ρ ∂x j

Turbulence & Turbomachinery Lab. Hanyang Univ. 10

 Turbulence

Energy equation for turbulent fluctuation

( momentum of fluctuation ) • ( velocity ) = ( kinetic energy of fluctuation )

ρ u~ ⋅ u~ = ρ u~ ⋅ u~ = 2 × Kinetic Energy per unit volume

{
momentum
per unit
it
volume
⎧ ⎫
⎪⎪ ∂u ∂u 1 ∂P ∂u 2 ⎛ ∂u ∂u ⎞ ∂U i ⎪⎪
ui × ⎨ i + U j i = − + ν 2i − ⎜ u j i − ui i ⎟ − u j ⎬
⎪ ∂t ∂x j ρ ∂xi ∂x j ⎜⎝ ∂x j ∂x j ⎟⎠ ∂x j ⎪
123 0 continuity
⎪⎩ || ⎪⎭
∂eij ∂ ⎛⎜ ∂ui ∂u j ⎞⎟ ∂ 2u ∂ ⎛⎜ ∂ui ⎞⎟
2ν =ν + = ν 2i + ν
∂x j ⎜
∂x j ⎝ ∂x j ∂xi ⎠⎟ ∂x j ∂xi ⎜⎝ ∂x j ⎟⎠

q2
Let q 2 = ui ui = u12 + u 22 + u 22 , k =
2
⎧ ∂ 2u ⎫ Reynolds stress
⎪ν ui 2i ⎪
∂x j ⎪ ⎡
∂ q2 2 ∂ q2 2 ui ∂p ⎪⎪ ⎪ ∂ui ⎤ ∂U i
+U j =− + ⎨or ⎬ − ⎢ui u j + 0 ⎥ − ui u j
∂t ∂x j ρ ∂xi ⎪ ∂x j ∂x j
∂e ⎪ ⎢⎣ ⎥⎦
⎪2ν ui ij2 ⎪
⎪⎩ ∂x j ⎭⎪
0 continuity

u i ∂P ∂ ⎛ pu i ⎞ P ∂u i ∂ ⎛1 ⎞
= ⎜ ⎟ ⎜⎜ pu j ⎟⎟
⎜ ρ ⎟ − ρ ∂x = ∂x
come back to this later since it is an interchange
ρ ∂xi ∂x j ⎝ ⎠ i j ⎝ρ ⎠ term where net effect on energy is zero

∂u i ∂ q2 2 ∂ ⎛1 2 ⎞
ui u j = uj = ⎜ q uj ⎟ since continuity
∂x j 2x j ∂x j ⎝2 ⎠

Turbulence & Turbomachinery Lab. Hanyang Univ. 11

 Turbulence

⎡ 2 ⎤ 2
∂ 2u ∂ ⎜⎛ ∂u j ⎞
⎟ =ν ⎢ ∂
⎛ ∂u i
⎜ ui
⎞ ⎛
⎟ − ⎜ ∂u i
⎞
⎟ ⎥ = ν ∂ q 2 − ν ⎛⎜ ∂u i
2 2 ⎞
⎟
ν u i 2i = ν u i
∂x j ∂x j ⎜⎝ ∂x j ⎟
⎠
⎢ ∂x j ⎜
⎝ ∂x j
⎟ ⎜ ∂x
⎠ ⎝ j
⎟
⎠
⎥ ∂x 2j ⎜
⎝ ∂x j
⎟
⎠
⎢⎣ ⎥⎦
⎡ ∂ ∂u ⎤ ⎡ ∂
2ν u i
∂eij
∂x j
= 2ν ⎢ ( )
u i eij − eij i ⎥ = 2ν ⎢ ( )⎤⎥ − 2ν (e )
u i eij
2

⎢⎣ ∂x j ∂x j ⎥⎦ ⎢⎣ ∂x j
ij
⎥⎦

eij + ω ij ( eij ω ij = 0 )

Summary

f
form 1

∂ q2 2 ∂ q2 2 ∂ ⎡ 1 1 ∂ ⎛ q2
⎜
⎞⎤ ⎛
⎟⎥ − u i u j ∂U i − ν ⎜ ∂u i ∂u i
⎞
⎟
+U j = ⎢− pu j − q 2 u j + ν
∂t ∂x j ∂x j ⎢ ρ 2 ∂x j ⎜ 2 ⎟⎥ ∂x j ⎜ ∂x ∂x ⎟
⎣ ⎝ ⎠⎦ ⎝ j j ⎠

form 1

∂ q2 2 ∂ q2 2 ∂ ⎡ 1 1 2 ⎤ ∂U i
∂t
+U j
∂x j
=
∂x j ⎢− ρ pu j − 2 q u j + 2ν u i eij ⎥ − u i u j ∂x − 2ν eij eij
⎣ ⎦ j

Turbulence & Turbomachinery Lab. Hanyang Univ. 12

 Turbulence

True dissipation
p

ε ≡ 2ν eij eij
⎡ 1 ⎛ ∂u ∂u j ⎞ 1 ⎛ ∂u i ∂u j ⎞⎤ 1 ⎡⎛ ∂u1 ⎞ ⎛ ∂u1 ⎞ ⎤
2 2

ε = 2ν ⎢ ⎜⎜ i + ⎟ ⎜
⎟ 2 ⎜ ∂x
+ ⎟⎥ = ν ⎢⎜ ⎟ + ⎜
⎟⎥ 4 ⎢⎜ ∂x ⎟ ⎜ ∂x ⎟
⎟ + L⎥≥0 Can only remove kinetic energy
⎢⎣ 2 ⎝ ∂x j ∂xi ⎠ ⎝ j ∂xi ⎠⎦ ⎣⎝ 1 ⎠ ⎝ 2 ⎠ ⎦⎥

Dissipation removes fluctuation kinetic energy and sends it to internal energy.

From the entropy equation it can be seen that this is irreversible
irreversible.
- due to deformation work

Note difference with 2ν u ij eij which acts to accelerate (increase Kinetic Energy) of neighboring
fluid by action of viscous stress
stress.

Turbulence & Turbomachinery Lab. Hanyang Univ. 13

 Turbulence

Pseudo – dissipation
p
Commonly the dissipation in the sense that synthetic

∂ui ∂ui 0
D ≡ν
∂x j ∂x j
( )
D = ν (eij + ω ij )(eij + ω ij ) = ν eij eij + 2eijω ij + ω ijω ij
⎧ ⎡ 2 ⎤ ⎡⎛ 2 ⎤⎫
( ) ⎪ 1 ⎢⎛ ∂u1 ⎞
∴ D = ν eij eij + ω ijω ij = ν ⎨ ⎜⎜ ⎟⎟ + L⎥ + ⎢⎜⎜
1 ∂u1
⎞ ⎪
⎟⎟ + L⎥ ⎬
⎪⎩ 4 ⎢⎣⎝ ∂x1 ⎠ ⎥ 4 ⎢⎝ ∂x2 ⎠
⎦ ⎣
⎥⎪
⎦⎭
in general D≠ε

Note : D ≥0 always

Why is ε the real dissipation and not D when both clearly remove Kinetic Energy from flow ?

eg)
e.g.)
u 1 u
eij ≠ 0 r eij = 0
r
ω ij = 0 ω ij ≠ 0
distortion rigid body rotation
r but no rotation r fluid particle has no shape change

Turbulence & Turbomachinery Lab. Hanyang Univ. 14

 Turbulence

ε ≡ 2ν eij eij cannot depend on rotation directly

D =ν
∂ui ∂ui
∂x j ∂x j
[
= ν eij eij + ω ijω ij ]

Dissipation of energy is due only to distortion ( i. e. strain rates) not rotation

∴ ε is real dissipation of energy
If turbulent is isotropic (almost never is) then D = ε
and 2
⎛ ∂u1 ⎞
ε = 15ν ⎜⎜ ⎟⎟
⎝ ∂xi ⎠
because all derivatives are interrelated

But at very high turbulent Re, the small scale turbulence is nearly isotropic – local isotropy !
At high turbulent Re, we shall see that dissipation occurs only at smallest scales of motion

Therefore in all turbulent flow at high turbulent Re ⎛⎜ ul > 10 3 ~ 10 4 ⎞⎟

⎝ν ⎠
2
⎛ ∂u ⎞
ε = 15ν ⎜⎜ 1 ⎟⎟ and D ≈ ε
⎝ ∂xi ⎠
∴ At high turbulent Re, both forms of energy equation are equivalent

For turbulence modeling usually use form of Kinetic Energy equation involving D and
∂
⎢ν
( )⎤⎥
⎡ ∂ q2 2
and
∂x j ⎢⎣ ∂x j ⎥⎦
treat D as if it were the dissipation.
((O.K. at high
g turbulent Re))

Turbulence & Turbomachinery Lab. Hanyang Univ. 15

 Turbulence

Task
measure ε in an expt.
measure 2ν e ij eij
2 2 2
⎛ ∂u ⎞ ⎛ ∂u ⎞ ⎛ ∂u ⎞
measure ⎜⎜ 1 ⎟⎟ , ⎜⎜ 1 ⎟⎟ , ⎜⎜ 1 ⎟⎟ etc
⎝ ∂x 1 ⎠ ⎝ ∂x 2 ⎠ ⎝ ∂x 3 ⎠
∂u1 ∂u ∂u
∴ be able to measure , 1 , 1 etc
∂x1 ∂x 2 ∂x3

need hot wire smaller than the smallest eddy

never been done ( almost impossible )

Turbulence & Turbomachinery Lab. Hanyang Univ. 16

 Turbulence

Taylor’s
y frozen field hypothesis
yp
suppose I can imagine turbulence being swept a fixed probe at velocity U c in such away
that the turbulence does not change as it passes the probe
Uc

x1
λ
spatial disturbance
λ λ : wavelength
Uc
UC frequency : f =
λ

If disturbance is convected along x1

then
∂ 1 ∂
= Taylor’s frozen field hypothesis
∂ x1 U c ∂t

- spatial changes are measured temporal changes

⎛ urms ⎞
- only good when turbulence intencity ⎜⎜ ⎟⎟ is less then 10%
⎝ U c ⎠

S
Suppose Ih
have probe
b tto measure u1 (t ) att fi
fixed
d point
i t
⎛ ∂ u1 ⎞
- differentiate to get ⎜ ⎟
⎝ ∂t ⎠

- square & ave. to get ⎛⎜ ∂u 1 ⎞⎟

2

⎝ ∂t ⎠

Turbulence & Turbomachinery Lab. Hanyang Univ. 17

 Turbulence

Using Taylor frozen field hypothesis (if applicable)

2
⎛ ∂u1 ⎞ 1 ⎛ ∂u ⎞
2

⎜⎜ ⎟⎟ = 2 ⎜ 1 ⎟
⎝ ∂x1 ⎠ U c ⎝ ∂t ⎠

If isotropy applies or locally isotropy ( high turb. Re )

1 ⎛ ∂u ⎞
2

ε = 15ν 2 ⎜ 1 ⎟
U c ⎝ ∂t ⎠

turb Intensity ( < 10 % )

Need : ① low turb.
ul
② high Rl = ( > 10 4 )
ν

Turbulence & Turbomachinery Lab. Hanyang Univ. 18

 Turbulence

Transport
p terms ((diffusion term))

Role of transport terms

advection (swept by)

2 2
∂q ∂q
2 +u 2 = ∂ ⎡⎢ − pu j
−
1 2 ⎤
q u j + 2υ u i e ij ⎥ − u i u j
∂ui
−ε
∂t ∂x j ∂ x j ⎣⎢ ρ ∂x j
j
2 ⎦⎥
dissipation
time rate of change often referred to as prod of turb.E
of KE per unit mass transport terms from mean flows
following the mean motion or diffusion term
: convective derivative

Consider special case – turbulence bounded by a rigid surface ( turbulence in a box )

Integrate energy eq. over entire volume

∫∫∫ dVρ[eq]
V0

since constant density assumed, multiply by ρ

⎡ ⎛q2 ⎞⎤
⎢ ∂⎜ 2 ⎟ ⎥ ⎡ ⎛ q 2 ⎞⎤ d
⎢ρ ⎝ ⎠⎥ = ∂
∫∫∫ dV
⎢ ∂t ⎥ ∂ t ∫∫∫
dV ⎢ ρ ⎜ 2 ⎟⎥ = ( KE in V 0 )
V0 V0 ⎣⎢ ⎝ ⎠ ⎦⎥ dt
⎢ ⎥
⎣ ⎦
V0 is time independent since assumed rigid

Turbulence & Turbomachinery Lab. Hanyang Univ. 19

 Turbulence
Production
⎡ ∂ui ⎤
∫∫∫ dV ρ ⎢ − u i u j ⎥ =
∂ u j ⎥⎦ ∫∫∫ ργ dV = rate at which turb. KE is produced in V0
V0 ⎣⎢ V0

γ = production per unit mass

Dissipation

∫∫∫ ρε dV
V0
= rate at which energy is dissipated in V0

Transport
⎧⎪ ∂ ⎡ ρ 2 ⎤ ⎫⎪
∫∫∫ dV ⎨⎪ ∂x ⎢− pu j − 2 q u j + 2 μ ui eij ⎥ ⎬⎪
V0 ⎩ j⎣ ⎦⎭
⎧ ⎡ ρ r ⎤⎫
∫∫∫
V0
dV ⎨∇ ⋅ ⎢− pu~ − q 2u + u~ ⋅τ~ ( v ) ⎥ ⎬
⎩ ⎣ 2 ⎦⎭

Since integrand is a divergence, we can apply the divergence them (since simply-connected region)
r r r
A = A( x )
vector field
∂ ⎡ ρ ⎤ ⎡ ρ ⎤
n~ = n~ (~
x) ∫∫∫ dV ⎢ − pu j − q 2 u j + 2 μ u i eij ⎥ = ∫∫ ⎢ − pu j − q 2 u j + 2 μ u i eij ⎥ n j ds
V0
∂x j ⎣ 2 ⎦ A0 ⎣ 2 ⎦
⎡ ~ ρ 2 r ~ ~ v ⎤ ~d
⎢⎣ − p u − 2 q u + u ⋅ τ ⎥⎦ ⋅ n ds

[ ] is evaluated on surface A0 enclosing V0

r r r
∫∫∫ ∇ ⋅ AdV = ∫∫ A ⋅ nds
V0 A0

Turbulence & Turbomachinery Lab. Hanyang Univ. 20

 Turbulence
r r r
− pu ⋅ n = − pu j n j = − pun
un = 0 if surface is at rest ( Kinematic B.C )
DF ∂F r
K.B.C = 0 + u ⋅∇F = 0
Dt ∂t
− ppu j n j = 0 on surface
ρ
− q 2u j n j ≡ 0 on surface by kinematic B.C
2
2 μ u i e ij n j = τ ij
(V )
uin j
r
u is identically zero on surface
- normal component due to K.B.C
- tangential component due to no-slip

∴ ui zij n j ≡ 0 on surface
(V )

∴ entire integral is exactly zero !

Therefore net effect of turbulent transport term (divergence) is zero :

Th d
They don’t increase
i or d
decrease KEKE.
They only move energy around.

pu j - moves energy by action of pressure forces H →L←H

ρ
q 2u j - carries energy from region of higher energy to region of lower energy
2
2 μ ui eij = τ ij ui
(V )
- accelerate adjacent region of flow due to viscous stress

υ
∂q 2 ( ) 2

- similar
i il tto above
b
∂x j

Turbulence & Turbomachinery Lab. Hanyang Univ. 21

 Turbulence

Digression
g on k − ε model for turbulence
- forget whole Reynolds stress eq.
- use eddy viscosity 1
k ≡ q2 KE L2 L2 1 L2 L2
k2 2 [ε ] ~ ⋅ 2 ~ 3 [ν ] ~
ν e ∝ ul = [k ] ~
T2 T T T T
ε ε = 2γ eij eij dissipation

− ui u j = 2ν e Eij
⎡ ∂U ∂V ⎤
⇒ −uυ = ν e ⎢ + ⎥
⎣ ∂y ∂x ⎦
write & close eqn.s for k & ε
∂k ∂k ∂ ⎡ ∂k ⎤
∂t
+U j =
∂ x j ∂ x j ⎢⎣
1 2
⎢ − pu j − q u j + ν
2
⎥−
∂ x j ⎥⎦
P −D
Model transport term using a gradient-diffusion approach
- assume energy is moved down gradients in energy by these terms i.e. from high to low

Simplest model

⎡ 1 2 ⎤ ∂k
⎢⎣ − pu j − 2 q u j ⎥⎦ ∝ ∂ x
j

∂k
[ transport term ] ∝ ν e
∂x j
− ui u j = e1ν e Eij
∂k ∂k ∂ ⎡ ∂k ⎤
+U j = ⎢C2ν e ⎥ + C1ν e Eij Eij − D
∂t ∂x j ∂x j ⎢⎣ ∂x j ⎥⎦

Turbulence & Turbomachinery Lab. Hanyang Univ. 22

 Turbulence

Production Term
Derive KE eq. for mean motion
⎛ ∂u ∂U i 1 ∂P ∂ ∂ ⎞
Ui × ⎜ i +U j =− + 2νEij − ui u j ⎟
⎜ ∂t ∂x j ρ ∂x j ∂x j ∂x j ⎟
⎝ ⎠
KE eq. for mean flow
∂ ⎛1 2⎞ ∂ ⎛1 2⎞ ∂ ⎛ 1 ⎞ ∂U i
⎜ Q ⎟ +U j ⎜ Q ⎟= ⎜⎜ − PU j − ui u jU i + 2νU i Eij ⎟⎟ + ui u j − 2νEij Eij
∂t ⎝ 2 ⎠ ∂x j ⎝ 2 ⎠ ∂x j ⎝ ρ ⎠ ∂x j

(when Q = U iU j = U1 + U 2 + U 3 )
2 2 2

2νEij Eij = E : dissipation of KE to heat due to mean motion ( deformation )

∂U i
ui u j is only term which occurs in both equ’s for mean and fluctuation , but with opposite sign!
∂x j
Effect of this term is take energy from mean and put it into flucuation , or vice versa
∂U i
Al
Almost always
l (i
(in engineering
i i flow)) ui u j h
fl has apposite
i sign
i as ∂x
j
∂U i
∴ ui u j ≤0 almost always P ≥0
∂x j

mean ∂U i fluctuation
ui u j
∂x j
∂U i
− ρ ui u j ≡P is called turbulent production term
∂x j

Turbulence & Turbomachinery Lab. Hanyang Univ. 23

 Turbulence

Pressure-Strainrate
Pressure Strainrate Term ( pressure interaction term )

KE eq. For fluctuation

∂ q2
∂q 2
2
∂ ⎡ pu 1 2 ⎤ ∂ui
+uj = ⎢− − q u j + 2ν u i e ij ⎥ − u i u j −ε
2 j

∂t ∂x j ∂x j ⎢⎣ ρ 2 ⎥⎦ ∂x j

3-dimensionality ! ( tendency to isotropic )

u12 ≅ u22 + u32

u22 ≅ u32
0.7u1′ = u2′ = u3′
Fluctuation mom.

∂u i ∂u 1 ∂p ∂ ⎡ ∂u ∂u ⎤ ∂U i
+U j i = − + ( 2νeij ) − ⎢u j i − u j i ⎥ − u j
∂t ∂x j ρ ∂ xi ∂ x j ⎢⎣ ∂x j ∂u j ⎥⎦ ∂x j
for i=1
⎛ ∂u ∂u 1 ∂p ∂ ⎡ ∂u ∂u ⎤ ∂U 1 ⎞⎟
u1 ⋅ ⎜ 1 + U j 1 = − + ( 2νeij ) − ⎢u j 1 − u j 1 ⎥ − u j
⎜ ∂t ∂x j ρ ∂x1 ∂x j ∂x j ⎟⎠
⎝ ⎣⎢ ∂x j ∂u j ⎦⎥

or div. (no net effect on closed sys.)

∂ ⎛1 2⎞ ∂ ⎛1 2⎞ ∂ ⎛1 ⎞ ∂ ⎡ 1 2 ⎤ 1 ∂u1 ∂U1 ∂u1

⎜ u1 ⎟ + U j ⎜ u1 ⎟ = − ⎜⎜ pu1 ⎟⎟ + −
⎢ 2 u u + 2ν u e + p
⎥ ρ ∂x − u u − 2ν e
∂t ⎝ 2 ⎠ ∂x j ⎝ 2 ⎠ ∂x1 ⎝ ρ ⎠ ∂x j ∂x j ∂x j
1 j 1 1 j 1 j 1 j
⎣ ⎦ 1

P1
⎛ ∂u ∂u j ⎞⎛ ∂u1 ⎞
− v⎜ 1 + ⎟⎜ ⎟
⎜ ∂x ∂x ⎟⎜ ∂x ⎟
⎝ j 1 ⎠⎝ j ⎠

Turbulence & Turbomachinery Lab. Hanyang Univ. 24

 Turbulence

i=2 : ∂ ⎛1 2⎞ ∂ ⎛1 2⎞ ∂ ⎛1 ⎞ ∂ ⎡ 1 2 ⎤ 1 ∂u 2 ∂U 2 ∂u 2
⎜ u2 ⎟ + U j
∂t ⎝ 2 ⎠
⎜ u2 ⎟ = −
∂x j ⎝ 2 ⎠
⎜⎜ pu 2 ⎟⎟ +
∂x1 ⎝ ρ ⎢− 2 u 2 u j + 2ν u 2 e2 j ⎥ + ρ p ∂x − u 2u j ∂x − 2ν e2 j ∂x
⎠ ∂x j ⎣ ⎦ 2 j j

i=3 : ∂ ⎛1 2⎞ ∂ ⎛1 2⎞ ∂ ⎛1 ⎞ ∂ ⎡ 1 2 ⎤ 1 ∂u3 ∂U 3 ∂u3

⎜ u3 ⎟ + U j
∂t ⎝ 2 ⎠
⎜ u3 ⎟ = −
∂x j ⎝ 2 ⎠
⎜⎜ pu3 ⎟⎟ +
∂x3 ⎝ ρ ⎢− 2 u3 u j + 2ν u3e3 j ⎥ + ρ p ∂x − u3u j ∂x − 2ν e3 j ∂x
⎠ ∂x j ⎣ ⎦ 3 j j

continuity

∂u1 ⎡ ∂u ∂u ⎤
= −⎢ 2 + 3 ⎥
∂x1 ⎣ ∂x 2 ∂x3 ⎦
1 ∂u1 1 ⎡ ∂u ∂u ⎤
p = − p⎢ 2 + 3 ⎥
ρ ∂x1 ρ ⎣ ∂x 2 ∂x3 ⎦

The rote of the pressure-strain rate term in the eq.’s for the individual compents of KE is to attempt to distribute
the energy among them, more or less equally.

This course a tendency of turbulence to “return-to-isotropy” when left alone (isotropy : u1 = u 2 = u 3 )

2 2 2

Turbulence & Turbomachinery Lab. Hanyang Univ. 25

 Turbulence

e.q grid turbulence

M
∂U 1
≠0
∂x 2
∂u
u1u 2 1
∂x 2

~ u12
The effect of pressure-strain rate terms on this flow makes it an excellent approx to isotropy turbulence
(after about 40 mash lengths)

cf. smooth out the gradient (can use honeycomb also)

after
~ 100M

e.g. Homogeneous turbulence with uniform mean shear (homogeneous shear flow)

∂U1 ∂U i i =1
= const (≠ 0) = const (≠ 0) all but
∂x2 ∂x j j=2
r ∂U 1
U = [U ( y ),0,0] U ( y ) = Ky = K = const
∂x 2

homogeneous : statistics are independence of position

u12 = const u 22 = const u1u 2 = const

∂
=0
∂t
∂ ∂ ∂ ∂
Uj = U1 +U 2 +U3
∂x j ∂x1 ∂x 2 ∂x3

Turbulence & Turbomachinery Lab. Hanyang Univ. 26

 Turbulence

∂U 1
all derivatives of averaged quantities (except
( ) are identically zero
∂x 2
i.e.
( ),
∂ u1 2
2
∂ ⎛1 2 ⎞
⎜ u1 u3 ⎟ = 0 etc
∂x2 ∂x3 ⎝ 2 ⎠

∂u1
p
∂x1 averages of products of fluctuations are not, in general ,
∂u zero even if flow is homogeneous !
e1 j 1
∂x j

write comp. of KE equ’s for the fluctuation.

0 =P −ε
∂U
0 = − uv − 2ν eij eij
∂y
1 ∂u ∂U ∂u ①
0 = + p 1 − uv − 2ν e1 j 1
ρ ∂x1 ∂y ∂x j
1 ∂u ∂u
0=+ p 2 − 2ν e2 j 2 ②
ρ ∂x 2 ∂x j
1 ∂u ∂u
0=+ p 3 − 2ν e3 j 3
ρ ∂ x3 ∂x j ③
1 ⎛ ∂u ∂u ⎞ ∂U ∂u
0 = − ⎜⎜ p 2 + p 3 ⎟⎟ − uv − 2ν e1 j 1
①/ ρ ⎝ ∂x 2 ∂ x3 ⎠ ∂y ∂x j

Turbulence & Turbomachinery Lab. Hanyang Univ. 27

 Turbulence

At very high turb. Re , the dissipation tends to be equally distributed between components
( local isotropy of small scale )

1 ⎛ ∂u 2 ∂u ⎞ ∂U 1
0=− ⎜p + p 3 ⎟⎟ − uv − ε
⎜
ρ ⎝ ∂x 2 ∂ x3 ⎠ ∂y 3
1 ∂u 2 1
0=+ p − ε
ρ ∂x 2 3
1 ∂u 3 1
0=+ p − ε
ρ ∂x3 3

In fact , it is the action of the pressure-strain

pressure strain rate terms which is responsible for the local isotropy of
the small scales where the dissipation occurs

Production ≈ Dissipation in many shear flows, even where other terms are present.

Turbulence & Turbomachinery Lab. Hanyang Univ. 28