Week 12

FULLY-DEVELOPED TURBULENT CHANNEL FLOW
Wall-bounded shear flows can be bounded on one side (as in bound-

ary layer) or both sides (as in plane channel), or even fully enclosed
(as in a pipe of uniform or variable cross-section).
Case of a channel formed by two “large” parallel plates is proba-

bly best understood, and most accessible to numerical simulation.
Similar to TBL in the near-wall regions; different if far from the wall.
“Fully-developed”: plates are long, so that inlet and exit effects may
be neglected. This also means length scales do not grow downstream
(a major contrast with boundary layers).
Mean velocity field independent of x, but a uniform streamwise mean

pressure gradient exists (to sustain the motion).
Assume: (i) mean flow is 2D (in x, y), (ii) spanwise homogeneity (in
z), (iii) statistical stationarity in time.
Continuity
∂U /∂x + ∂V /∂y = 0
“Fully dev.” leads to V = 0 everywhere, if walls are impermeable.
1
Momentum: Simplify from the general form
∂U i ∂U i ∂uiuj 1 ∂P ∂ 2U i
+ Uj + =− +ν
∂t ∂xj ∂xj ρ ∂xi ∂xj ∂xj
Cross-stream: For i = 2, equation reduces to

∂v 2 1 ∂P
=−
∂y ρ ∂y
(Had a similar result for free shear flows, but there it was only ap-
prox.) Integrate in y, with boundary conditions at impermeable wall:
v 2 + P /ρ = P0/ρ = f (x) only
where P0(x) is the pressure measured at the wall. Further, since
under fully-developed conditions v 2 cannot depend on x, we get
∂P /∂x = dP0/dx
Streamwise: i = 1 with V = 0, etc, as noted above, and making

use of fully-developed conditions again:
duv 1 dP0 d2U
=− +ν 2
dy ρ dx dy
Integrate in y:
y y dP0 dU y
uv = − +ν
0 ρ dx dy 0
Since no-slip and impermeability still apply, uv = 0 at the wall.
If the fluid is Newtonian the wall shear stress is given by
τw = µ(dU /dy)y=0
p
i.e. ν(dU /dy)y=0 = u2∗ where the friction velocity u∗ ≡ τw /ρ.
2
With these results, the streamwise mean momentum equation gives
y dP0 dU
uv = − +ν − u2∗
ρ dx dy
On the centerline (y = h in our setup), both uv and dU /dy vanish.
Thus
h dP0 h dP0
0=− + 0 − u2∗ , ⇒ u2∗ = −
ρ dx ρ dx
which is expected, since a force balance requires the pressure gradient
to balance viscous shear stress at the wall. The momentum equation
can now be re-cast as
dU h yi
2
−uv + ν = u∗ 1 −
dy h
This is Eq. 5.2.6 of T&L. It says the total shear stress (collected
on LHS) has a linear variation in y — which is well reproduced in
numerical simulations, as cited in Pope (Fig 7.3).
It is also not surprising that viscous shear stress dominates near the
wall, while Reynolds stress dominates away from it.
3
INNER versus OUTER SCALING
Near the wall, key velocity and length scales are u∗ and ν/u∗. Nor-
malized mean velocity and Re-stress profiles may be of the form
U /u∗ = f (y +) ; uv/u2∗ = g(y +) ;
where y+ ≡ yu∗/ν is “distance from the wall, in wall units”.
Far from the wall, the scales would be u∗ and h. We can use
(U − U 0)/u∗ = F (η)
where U 0 is the centerline velocity, and η ≡ y/h.
“Close to the wall” may mean y/h 1. “Far from the wall” may
mean y+ 1. If the “friction Reynolds number” hu∗/ν is very high,
there we may be range of y where both inequalities apply. That is,
ν/u∗ y h
If there is a smooth transition between inner and outer regions, the
velocity gradients deduced from expressions above must agree: i.e.,
u∗ dF u2∗ df
=
h dη ν dy+
Multiply by y/u∗:
ηdF/dη = y+df /dy+
LHS is independent of y+; RHS is independent of η. They can both
hold and agree only if both sides reduce to a constant:
ηdF/dη = y+df /dy+ = 1/κ
where κ is known as the von-Karman constant, which value can only
come from experiments or DNS. Its accepted value is close to 0.4.
HW 8: will show that these arguments lead to a logarithmic profile
for U vs y.
4
“Inertial sublayer; a.k.a Constant-stress layer”
Consider again the equation

dU h yi
2
−uv + ν = u∗ 1 −
dy h
At y+ 1 viscous effects are weak: (2 frames in Pope Fig 7.3)
|ν dU /dy| | − uv|
But if y/h 1 the geometry becomes unimportant
1 − y/h ≈ 1
If both of these conditions hold we obtain
−uv ≈ u2∗
The region ν/u∗ y h called the inertial sublayer since there
neither viscosity nor external boundary conditions are important.
It is also a region of approximately constant Reynolds shear stress.
Another (more common) name is “log-layer” because (see HW 8)
U /u∗ = (1/κ) ln y+ + C
where C ∼ 5 − 5.5 increases slightly with Reynolds number.
This result is so common, that it is often called the “log-law”.

Experiments: y+ at least 30, and far from centerline.
Viscous sublayer, near the wall
Close to the wall: y/h ∼ 0, while −uv is small. Hence

dU
ν ≈ u2∗ .
dy
5
A simple integration, assuming the wall is fixed, gives
U yu∗
= . i.e. U+ = y+
u∗ ν
Same result from a Taylor-series expansion for U (y) for small y (or
T.S for the instantaneous velocity, followed by averaging).
Experimental evidence suggests the linear variation that character-

izes the viscous sublayer extends up to y+ ≈ 5.
In this region, although viscous effects are strong and r.m.s velocities
are small, turbulence is still present, in the velocity gradients.
Buffer layer, 5 < y+ < 30
Both uv and dU /dy significant. No simple formula for U (y) here,

but contributes strongly to production of TKE (P ≈ −uv dU /dy).
Mean velocity profile from a high-Re DNS
DNS of turbulent channel flow up to Reτ ≈ 5200 401
30
25
ject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms.
20
15
10
5
0
10 0 10 1 10 2 10 3 10 4
F IGURE 2. (Colour online) Mean streamwise velocity profile for all the cases listed in
table 1, where the legend of line styles and symbols is also given.
(a) (b) (c) 2.70

3.4 3.4
to ∼ 5200 in Lee &

hu∗/ν up 3.2
3.0
3.2
3.0
Moser JFM Vol 774, 2.65 395-415 (2015)
Reported 2.8
κ = 0.384. Perhaps2.8differs slightly from2.60BL or pipe flow.
2.6 2.6
2.4 2.4 6 2.55
2.2 2.2
2.0 2.0 2.50
10 1 10 2 10 3 10 4 0 1000 2000 3000 4000 200 400 600 800 1000
Mean-square velocities and Reynolds shear stress
• Strong anisotropy in Reynolds stress tensor “near the wall”

• hu2i, hv 2i, hw2i show different growth rates in viscous sublayer
Production and dissipation
• P/ = O(1) in inertial sublayer (short here due to lower Re)

• High SK/ (peak ≈ 15 associated with stronger production
7
The TKE budget
• Dissipation has maximum value at the wall (large velocity gra-

dient fluctuations, due to boundary conditions on the velocity)
• Viscous diffusion is important in inertial and buffer layers
TURBULENT BOUNDARY LAYERS (Pope. Sec. 7.3)
Differences with channel flow include the following:

1. thickness of the flow grows downstream (no “fully-dev.” state)
2. wall shear stress τw (x) not known in advance (unlike fully-developed
channel where u2∗ = −(h/ρ) dP0/dx)
3. external intermittency in outer parts of the boundary layer
(a turbulent/non-turbulent interface, similar to free shear flows)
4. greater deviation from log-law near the edge of the flow
TKE budget for the TBL (Pope Eq. 7.177, LHS=0 for channel):
∂K ∂K ∂ 1 2 1 ∂ ∂ 2K
U +V = − ( vq ) − (vp ) + ν 2 + P − ∗
0
∂x ∂y ∂y 2 ρ ∂y ∂y
where ∗ is the pseudo-dissipation (denoted by ˜ in Pope)
8

Week 12

Uploaded by

Copyright:

Available Formats

Week 12

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Week 12

Uploaded by

Copyright:

Available Formats

FULLY-DEVELOPED TURBULENT CHANNEL FLOW

Wall-bounded shear flows can be bounded on one side (as in bound-

Case of a channel formed by two “large” parallel plates is proba-

Mean velocity field independent of x, but a uniform streamwise mean

Cross-stream: For i = 2, equation reduces to

Streamwise: i = 1 with V = 0, etc, as noted above, and making

Consider again the equation

This result is so common, that it is often called the “log-law”.

Viscous sublayer, near the wall

Close to the wall: y/h ∼ 0, while −uv is small. Hence

Experimental evidence suggests the linear variation that character-

Buffer layer, 5 < y+ < 30

Both uv and dU /dy significant. No simple formula for U (y) here,

Mean velocity profile from a high-Re DNS

DNS of turbulent channel flow up to Reτ ≈ 5200 401

(a) (b) (c) 2.70

to ∼ 5200 in Lee &

• Strong anisotropy in Reynolds stress tensor “near the wall”

Production and dissipation

• P/ = O(1) in inertial sublayer (short here due to lower Re)

• Dissipation has maximum value at the wall (large velocity gra-

Differences with channel flow include the following:

You might also like

• P/ = O(1) in inertial sublayer (short here due to lower Re)