0% found this document useful (0 votes)
27 views10 pages

Week 10

This document discusses turbulent free shear flows and self-preserving wakes. It provides scale estimates and analysis of the Reynolds-averaged Navier-Stokes equations for such flows. It also examines the assumptions and accuracy of a simple eddy viscosity turbulence model for predicting mean velocity profiles in self-similar wakes.

Uploaded by

Gautham Giri
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
27 views10 pages

Week 10

This document discusses turbulent free shear flows and self-preserving wakes. It provides scale estimates and analysis of the Reynolds-averaged Navier-Stokes equations for such flows. It also examines the assumptions and accuracy of a simple eddy viscosity turbulence model for predicting mean velocity profiles in self-similar wakes.

Uploaded by

Gautham Giri
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 10

TURBULENT FREE SHEAR FLOWS

Free — far away from solid boundaries


Shear — a cross-stream (mean) velocity gradient is present

Thin (or “almost parallel”): Cross-stream dimensions (ℓ) much smaller


than length scale in streamwise direction (L). Some type of boundary
layer approximation should apply.

Question: How do profiles of U , uv, 12 uiui etc, behave?

Two velocity scales for the mean flow. Velocity variation along the
flow (Ũ ) versus across the flow (US ). Jets, Wakes, Mixing layers (or
Free shear layers): Fig. 4.1 of T&L.

Jets: a source of momentum. US (x) = centerline velocity. Ũ = US .

Wakes: momentum deficit (max US (x)), behind a body. Ũ = U∞


(U 0 in T&L)

1
Basic scale estimates:
 
∂U US
In common with Laminar BL: ∂x ∼O L , V ∼ U S Lℓ

Use 2D RANS equations: Let u2, v 2, uv all be of order U 2

Cross-stream mean momentum equation


 2
∂ 2V

∂V ∂V ∂uv ∂v 2 1 ∂P ∂ V
U +V + + =− +ν +
∂x ∂y ∂x ∂y ρ ∂y ∂x2 ∂y 2
Estimate O.M of each term. Let Ũ = US for jet, U∞ for wake.
US ℓ 1 US ℓ US ℓ 1 U2 U2 US ℓ 1 US ℓ 1
Ũ ? ν ν
L L L L ℓ L ℓ L L2 L ℓ2
Use ℓ ≪ L, while expecting U 2/ℓ (turbulence) to be important.
Compare O.M of other terms with this:
 2
ℓ 1 .U 2 Ũ US ℓ
1st conv. term Ũ US ∼
LL ℓ U U L
.U 2  2  2
2 ℓ US ℓ
2nd conv. term US 2 ∼
L ℓ U L
US . U 2 US ℓ ν
Viscous term ν ∼
ℓL ℓ U L Uℓ
If Rl ≫ 1 and ℓ/L ≪ 1, while US and U are of same O.M, the
pressure gradient may be the only one left to provide the balance.
Thus
∂v 2 1 ∂P
≈− ⇒ P /ρ + v 2 = constant
∂y ρ ∂y
i.e. pressure varies cross-stream in the flow. This property is in
contrast to laminar flow in the same geometry.
2
Streamwise mean momentum equation

∂ 2U ∂ 2U
 
∂U ∂U ∂u2 ∂uv 1 ∂P
U +V + + =− +ν +
∂x ∂y ∂x ∂y ρ ∂x ∂x2 ∂y 2
Estimate O.M of each term, then compare with U 2/ℓ:
US US ℓ US U2 U2 US US
Ũ ? ν ν
L L ℓ L ℓ L2 ℓ2
US . U 2 Ũ US ℓ
1st convective term Ũ ∼
L ℓ U UL
2  2
US . U 2 US ℓ
2nd convective term ∼
L ℓ U L
US . U 2 US ν
Viscous term ν 2 ∼
ℓ ℓ U Uℓ
In a laminar free shear flow there would be no mechanism to sustain
a (mean) pressure gradient in x. But, here from the cross-stream
RANS equation of the form P /ρ + v 2 = constant we obtain
−(1/ρ)∂P /∂x ≈ ∂v 2/∂x
which is O(U 2/L), and negligible compared to U 2/ℓ.

If Rl is high, viscous transport is negligible. Hence we get


∂U ∂U ∂uv
U +V + =0
∂x ∂y ∂y
For wakes, US ≪ Ũ = U∞. This further simplifies to (T&L Eq. 4.19):
∂U ∂uv
U∞ + =0
∂x ∂y

3
ANALYSIS OF SELF-PRESERVING WAKES

If shape of mean velocity profile is to be preserved downstream, then


we would have as self-similar solution, via a dimensionless variable
ξ = y/ℓ(x) where l is a cross-stream length scale to be determined.
∂U ∂uv
U∞ + =0
∂x ∂y
(T&L Eq. 4.19): (Two unknowns in 1 equation. A closure problem.)
2
Assume (U∞ − U )/US = f (y/ℓ); −uv/US = g(y/ℓ).
∂ξ 1 ∂ξ y dl ξ dl
Chain rule : = ; =− 2 =−
∂y ℓ ∂x ℓ dx ℓ dx
Hence
∂U dUS US dl ′
=− f+ f
∂x dx ℓ dx
2
∂uv 2 ∂g 2 ′ ∂ξ US ′
= −US = −US g =− g
∂y ∂y ∂y ℓ
HW 7, Q.2: show that conditions for self-similarity (such that coef-
ficients in ODE for f are all dimensionless constants) are
U∞ℓ dUS U∞ dl
2 = const ; = const .
US dx US dx

Suppose power law variations: US ∝ xm. ℓ ∝ xn. Both conditions


above give n − m − 1 = 0. Need one more equation: to come from
something that is conserved, i.e. independent of x.

Consider momentum integral: no external force acts on the fluid


in wake, downstream of the solid body. (Momentum deficit is only
being spread laterally, from y = −∞ to y = ∞.)
4
Integrate the RANS streamwise momentum equation:
ˆ ∞ h i∞

U∞ U dy + uv =0
∂x −∞ −∞
ˆ ˆ
∂ ∂
⇒ U dy = (U∞ − US f ) dy = 0
∂x ∂x
ˆ
US f dy = const ⇒ US ℓ = const, m = −n

We may now take US = Ax−1/2, ℓ = Bx1/2. Substitute back into


∂U ∂uv
U∞ + =0
∂x ∂y
∂ ∂ 2
U∞ (U∞ − US f ) + (−US g) = 0
∂x ∂y
Using chain rule results on last page, should get (HW 7, Q.3)
U∞B U∞B
(ξf )′ = g ′ ⇒ (ξf ) = g
2A 2A
upon integration and application of boundary conditions in the freestream.

A very simple turbulence model

“eddy viscosity”: hypothesis: assume turbulent momentum flux be-


haves analogously as viscous shear stress: down the mean velocity
gradient with length scale ℓ
∂U
−uv = νT ,
∂y
with νT ∝ US ℓ. [It turns out νT = const in this flow.]

5
Let’s define a “turbulence Reynolds number” RT = (US ℓ)/νT . We
can write νT = (US ℓ)/RT . Hence,
US ℓ 1
uv = (−US f ′ )
RT ℓ
which gives g = −f ′/RT . With this closure for g, now we need to
solve
U∞ B f′
(ξf ) = −
2A RT
f′ U∞ B
= −αξ where α = RT .
f 2A
The solution is
f = exp − 12 αξ 2


A smooth “Gaussian” curve, with maximum at centerline.

How accurate is this result?

We used the simplest turbulence model (a constant νT ) possible.


Comparison with experiments shows predicted profile is (a) quite
accurate near centerline, but (b) too high at the edges of the flow.

What are the reasons for agreement/disagreement?

Near centerline: flow is well-mixed, neither −uv nor ∂U /∂y vary


strongly with position. Not surprising that νT also varies little there.
6
Near edges of the flow: “external intermittency” at the interface,
where conditions are turbulent only part of the time. Actual νT is
smaller near the edges. Smaller νT implies larger RT , thus larger α.
The latter explains steeper decrease of U∞ − U towards the edges.

Expts suggest RT ∼ 12.5, which sounds small for a “Reynolds num-


ber”. That’s because we used US , which is much smaller than U∞.

To complete the solution (for mean velocity) ...

Still need to fix values of ℓ, A, B, etc.

Set α = 1: then f = exp − 12 ξ 2 . Then ℓ becomes the value of y




where f = exp(−1/2) = 0.6. With α = 1, RT = 12.5,


U∞B B 1 2 0.16
= 1/RT ⇒ = = (i)
2A A U∞ RT U∞

Momentum deficit, in terms of momentum thickness (θ):


ˆ ˆ ∞
2
ρU∞ θ = ρU∞ (U∞ − U ) dy = ρU∞US ℓ f dξ
−∞
With the form of f known from above, we get
√ √
θ = (US /U∞)ℓ 2π = AB 2π/U∞ (ii)

Multiplying (ii) by (i) gives B = 0.2526√ θ.
Dividing (ii) by (i) gives A = 1.579 U∞ θ.

Finally:

US = Ax−1/2 ⇒ US /U∞ = 1.579 (θ/x)1/2. (Eq. 4.2.18, T&L).

ℓ = Bx1/2 ⇒ ℓ/θ = 0.2526 (x/θ)1/2. (Eq. 4.2.19, T&L).


7
TKE BUDGET FOR TURBULENT PLANE WAKE

Recall from HW 3 the general form of the TKE equation is


 
D1 ∂ 1 1
uiui = − uj p + uiuiuj − 2νuisij −uiuj Sij −2νsij sij ,
Dt 2 ∂xj ρ 2
where D/Dt ≡ ∂/∂t + U j ∂/∂xj .

Assume stationarity in time, U 1 ≫ U 2, homogeneity in x1 (approx)


and x3, and that the flow is “thin”. We obtain
∂ 1 ∂ ∂U 1
U∞ ( 2 uiui) = − (u2p/ρ + 21 uiuiu2) − u1u2 −ϵ
∂x1 ∂x2 ∂x2
Or, with q 2 as shorthand for uiui, and re-arranging as in in Eq. 4.2.30
in T&L:
∂ 1 2 ∂ ∂U
0 = −U∞ (2q ) − (vp/ρ + 12 q 2v) − uv −ϵ
∂x ∂y ∂y
Advection by mean flow, cross-stream transport by pressure and ve-
locity fluctuations, production by mean shear, and dissipation.

Fig. 4.5 of T&L: How do the relative strengths of these terms vary
with position, across the width of the wake?

Production is zero at y = 0 (symmetry) and y = ±∞ (freestream)

Dissipation is max and nearly constant (well mixed) at y = 0

Turb. Transport: lateral spreading: negative inside, positive outside

Advection: expect ∂ 21 uiui/∂x < 0 near centerline (where turbulence


decays with x), but the sign may reverse far from centerline (due to
lateral spreading).
8
Plane wake: T&L, Fig. 4.5, p. 123.

Axisymmetric wake: Pope, Fig. 5.29, p. 153.

9
WALL-BOUNDED TURBULENT SHEAR FLOWS

Influence of a solid boundary:

No matter how small, viscosity enforces no slip boundary condition.


Wall shear stress is dependent on the mean velocity gradient.

Impermeability causes “normal” velocity fluctuations to be attenu-


ated in near-wall region.

Multiple cross-stream length scales

1. A viscous length scale: ν/w where w is a velocity scale that


characterizes mean velocity gradient due to velocity fluctuations.
This gradient drives the turbulence: hence w may be comparable
to the turbulent velocity scale U.
2. Surface roughness: zero if surface is “aerodynamically smooth”.
Rough surfaces can trigger turbulence but are sometimes also
useful for drag reduction (by “riblets”).
3. Width of the flow: a BL thickness, or half of height of a channel

Clearly, ν/w is important near the wall, unimportant far from the
wall. While for δ (or h) it is just the reverse.

If δ ≫ ν/w (which may occur at high Reynolds number) there may


be a range of values of distance from the wall such that
ν/w ≪ y ≪ δ
neither ν/w nor δ are important. This intermediate region is called
the “inertial sublayer”.
10

You might also like