邊界層理論1).pdf · 2018. 10. 26. · 2-D Flow on Positive Curved Surface . Effect of Pressure...

Preview:

Citation preview

邊界層理論 〈Theory of Incompressible Boundary Layers〉

方富民 教授 編著 〈Dr. Fuh-Min Fang〉

Department of Civil Engineering

National Chung Hsing University

September 2018

Outline Textbook: BOUNDARY LAYER THEORY – by Hermann Schlichting Subjects:

1. INTRODUCTION

Mathematical Example Physical Example Simple Successive Approximation

2. BOUNDARY LAYER EQUATION

Order Analysis Falkaner–Skan Equation (Flat-Plate Boundary Layer) 2-D Flow on Positive Curved Surface Effect of Pressure Gradient on the Fluid Motion Flow Past A Wedge Free Shear Flows – Jet, Mixing Layer, Wake

3.Turbulent Boundary Layer

Karman–Polhausen Method Separation Depth – Averaged in Open Channel Flows Smooth / Rough Flow Inner – Outer layer Joint Condition Law of the Wake Power Law Developed Boundary-Layer Flows in Open Channel

*** Initiation & Participation in discussions are highly encouraged.

1

A Mathematical Example

Given a 2nd-order linear ordinary differential equation (O.D.E.) as

0yxdyd

xdyd2

2

=++ε , where ε > 0 and ε << O[1]

with boundary conditions (B.C.’s) as 0)0(y = and 1)1(y = It is plausible to approximate the O.D.E. as

0yxdyd

=+

As the solution is in a form as xey λ= , substitution yields

0ee xx =+λ λλ

or 0e)1( x =+λ λ

Since 0e x ≠λ , 1−=λ . The general solution is x-eAy = .

B.C. @ 1=)1(y x-1ey eA =⇒=∴

B.C. @ 0=)0(y (N.G.) 0ey 0-1 ≠=

y

3

2

1

actual soluation x

approximate solution

Region that approximate solution does not work (B.L.)

2

Comments:

* Sometimes, even ε is small, a negligence might cause a serious problem (near x = 0) because of a loss of information.

* In some flow problems, viscosity may be neglected so as to simplify flow analyses. However, if the viscosity is neglected recklessly, one might lose the essence of the problem.

Physical Example

In a one-dimensional (1-D) flow with a heated fluid in a pipe (flow moves slowly, only molecular problem is of interest. That is, the flow is laminar.)

Within △, 2

2

xdyd

ε is important. (Although △ is small, gradient is very large.)

Beyond △, 2

2

xdyd

ε can be neglected.

x

Light

x=00θ=θ x x+dx

dx x=L01 2

1θ=θ=θ

Temp is free to vary in the region

3

t∂∂

(heat in box)=(heat inflow rate)-(heat outflow rate)+(rate of internal production)

Flux = rate of transfer across boundaries /time/ boundary area

= (diffusive flux)+(convective flux) ………….……………(1) � Diffusive flux:

xkFD ∂

θ∂−=

k:molecular diffusivity of heat (joules / m‧sec) θ:absolute temperature x:spatial coordinate

� Convective flux= volumequantity

‧(velocity along direction of interest)

Define specific heat (at constant pressure) = Cp (joule / kg‧°K)

3oo3p mjoule � K K kg

joule mkg � C θρ

uCF pC θρ=

Therefore, flux x

kuCFFF pDC ∂θ∂

−θρ=+= ……………………..(2)

* Assumption for heat produced by reaction

( )0-timevolume

heatθθβ=

⋅ ( in

K secmjoules

o3 ⋅ )

From (1), therefore, one gets

( ) dxA-FA-FAdxA]C[t 0dxxxp θθβ+=θρ

∂∂

+ ( ∀= ddxA )

For a constant cross-sectional area (A) of the pipe

( ) ( )0dxxx

p -xdFF

Ct

θθβ+−

=θρ∂∂ +

4

Assume Cp is not a function of time and for constant density (ρ),

( )0p xF

tC θ−θβ+

∂∂

−=∂θ∂

ρ

Substitution of F gives (equation 2 on pp. 3)

( )02

2

pp xk

xuC

tC θ−θβ+

∂θ∂

+∂θ∂

ρ−=∂θ∂

ρ

Assume steady ( )0t=

∂∂ , setting Uu −= leads to

( ) 0x

UCx

k 0p2

2

=θ−θβ+∂θ∂

ρ+∂θ∂

Normalization:

Let Lxx = and

01

0

θ−θθ−θ

=θ ( )( )θθ−θ+θ=θ 010

*

Noted that as 0x = , 0x = , 0θ=θ , 0=θ

Lx = , 1x = , 1θ=θ , 1=θ

By substitution,

( )( ) ( )( ) ( )01010p0102

2

2 xdd

L1UC

xdd

Lk

θ−θθβ+θθ−θ+θρ+θθ−θ+θ

( )

θ−θρ×

01pUCL gives 0

UCL

xdd

xdd

LUCk

p2

2

p

=θρβ

ρ …(3)

Chain rule gives ( ) xdd

L1

L1

dd

xdd

Lx

==

( ) 2222Lx2

2

xdd

L1

L1

dd

xdd

==

xFC

∂∂

− x

FD

∂∂

5

Define kinematic heat diffusivity as

pCk

ρ=α ( ) ν~ sec

m2

which is independent of the substance transfer.

Hence ULLUC

kp

α=

ρ (similar to

Re1 )

For simplicity, set 1* =β . Equation (3) on pp. 4 becomes

0xd

dxd

d2

2

=θ+θ

ε

with boundary conditions as 0)0( =θ and 1)1( =θ .

For large ULα , 1<<ε . As it is compared to the mathematical example as

0y'y"y =++ε 1<<ε

@ ( ) 10y = ( ) 11y =

⇒ One gets a similar problem.

Simple Successive Approximation (Perturbation Method)

Purpose: to overcome the result of unsatisfactory boundary condition.

To get the zeroth-order solution, set 0=ε then solve

0y'y 00 =+ ( ) 11y0 = x10 ey −=→

To get the 1st-order solution of 0y'y"y 110 =++ε

"yy'y 011 ε−=+ with B.C. as ( ) 11y1 =

To get the 2nd-order solution of 0y'y"y 221 =++ε

………………………………

6

Formalize in terms of a regular (outer) asymptotic expansion, assume

∑ ε=∞

=0i

ii0yy

+ε+ε+= 2020100 yyy

By substitution, it yields

0..)yyy(..)'y'y'y(..)"y"y"y( 022

0100022

0100022

0100 =+ε+ε+++ε+ε+++ε+ε+ε

Re-arranging according to the exponential of ε leads to

( ) ( ) ( ) ( ) 0O''yy'y''yy'yy'y 3010202

20001010000 =ε+++ε+++ε++

For any 1<<ε , which makes the above exist, all in ( ) should be equal to zero.

Accordingly, for ( ) ( ) ( ) ( ) 00y0y0y0y 022

0100 =+ε+ε+=

( ) 00y00 =∴ ( ) 00y01 =

Proof :

If 0A0i

ii =∑ ε∞

= for all 1* <<ε<ε , then 0AAA 210 ====

0AAA 22

10 =+ε+ε+⇒

As 0→ε , 0A0 = then ( ) 0AA 21 =+ε+ε

0AA0 21 =+ε+→≠ε

As 0→ε , 0A1 =

Similarly, one can show that ∞A,A,A 32 are all zero.

( ) 00y = ( ) 11y =

O[0] 0y'y 0000 =+ ( ) 00y00 = ( ) 11y00 =

O[1] ''yy'y 000101 −=+ ( ) 00y01 = ( ) 01y01 =

O[2] ''yy'y 010202 −=+ ( ) 00y02 = ( ) 01y02 =

7

Supposed one uses the boundary condition at 0x =

0y'y 0000 =+

Let x00 ey λ= , substitution gives ( ) 0e1 x =+λ λ

The general solution is x00 eCy −= . For 0e x ≠λ 1−=λ∴ .

@ ( ) 00y00 = then 0C = ( ) 0xy00 =∴

For 0y'y 0101 =+ , the general solution is x01 eDy −=

@ ( ) 0D0y01 == thus ( ) x101 ex1y −−=

Similarly,

Now, use the boundary condition at 1x =

0y'y 0000 =+ @ ( ) 10y00 = x100 ey −=⇒

Then x100 e'y −−= x1

00 e''y −=

x10101 ey'y −−=+ @ ( ) x1

01 e1y −= Homogeneous solution ( ) x

ey h01

=

Particular solution ( ) x1xey p01

−=

x

e

0

1

1

( ) ( ) ( ) ( ) 11y1y1y1y 022

0100 =+ε+ε+=

( )[ ] ( ) ( ) 01y1y11y 022

01000 =+ε+ε+−ε∴

( ) 11y00 =∴ ( ) 01y01 = ( ) 01y02 =

⇒ x101 e)x1(y −−=

( ) 0xy01 =

( ) 0xy02 = ⇒ trivial solutions

( ) 0xy03 =

8

Now, go back to the beginning

])(O)x1(1[eyyyy 2x102

20100 ε+−ε+=+ε+ε+= −

⇒ Now, look at this small region using a “microscope”.

Define ε

=ζx

e

x0 1

ε

y

ζ0 1

y

0 1ε x

x

0 ε

inner outer

Check : x1x

01 execy −− −=

x1x1x01 exeec'y −−− +−−=

x10101 ey'y −−=+∴ (O.K.)

The solution obtained by using perturbation method is even worse!

Why? ⇒ due to bias at the start.

9

Substituting into 0y'y"y =++ε leads to

0yd

yd1d

yd12

2

2 =+ζε

+ζε

ε

or 0yd

ydd

yd2

2

=ε+ζ

< diffusion > < convection > < reaction >

The boundary condition applies ( ) 00y ==ζ .

Since ε is small, the equation can be approximated as

0d

ydd

yd2

2

then Cyd

yd=+

ζ1

The general solution is CeDy += ζ− (inner solution)

○aE

A ( ) CD00y +== , thus one has ( )1eDy −= ζ−

To find out the behavior at the “outer edge” of the boundary, take limit as ∞→ζ

( ) Dy −=∞→ζ

Matching the solution at the edge as 0x

lim→

(outer region) = ∞→ζ

lim(inner region)

One has De −=

e

0

1

1 x

outer approx

inner approx

combined solution

Chain rule: ζε

ζ=

dd1

dd

xdd

xdd , 2

2

22

2

dd1)

dd1(

xdd

xdd

ζε=

ζε=

Finally,

inner solution: ( )ζ−−= e1ey

outer solution: x1ey −=

10

Review

ii

33

22

11 x

ex

ex

ex

e∂∂

=∂∂

+∂∂

+∂∂

=∇

Gravitational force / mass of fluid:

3xgG −= (force potential, 22

tL )

gkGF −=∇= (downward)

3ii gF δ−=

Divergence:

jji

i eux

eu ⋅∂∂

=⋅∇

zw

yv

xu

xu

xu

j

j

i

jji ∂

∂+

∂∂

+∂∂

⇒∂∂

=∂∂

δ= (3-D)

Laplacian:

jj

ii x

ex

e∂∂

⋅∂∂

=∇⋅∇

2

jjjiji xxxx

∇⇒∂∂

∂∂

=∂∂

∂∂

δ=

Continuity:

0u =⋅∇

=

∂∂

0xu

j

j (valid for incompressible flows)

Momentum (Navier-Stokes) equation:

Guxp1uu

tu 2

i

∇+∇ν+∂∂

ρ−=∇⋅+

∂∂

ij

i

jij

ij

i

xG

xu

xxp1

xuu

tu

∂∂

+∂∂

∂∂

ν+∂∂

ρ−=

∂∂

+∂∂ ( )3xgG −=

11

Scaling: t~ULt = , p~Up 2ρ= , G~UG 2=

∇=∇ ~L1 u~Uu =

Normalization yields

Continuity: 0u =⋅∇ 0u~~LU

=⋅∇ or 0u~~ =⋅∇

Momentum:(x-direction)

u~U~L1u~U

tu~

LU2

∇⋅+

∂∂ G~~

LUu~~U

Lp~~

LU1 2

22

2

∇+∇ν

+∇ρ

ρ−=

Dropping “~” and multiplying

2U

L gives

Gupuutu 2 ∇+∇ε+−∇=∇⋅+∂∂

where LURe

1 ν==ε

U

iUu =

airfoil

L

* When the Reynolds number (Re) is small, ε can be large.

* For a large Re, if u2∇ε is neglected, the solution will not be satisfactory within a thin region where u2∇ε is not small.

12

In the region where Re >> 1: ( )1<<ε

Droping u2∇ε terms yields 0=φ∇×∇ (irrotational flow, φ exists)

Also, as it can be proved that

( )uuu21uu

2×∇×−∇=∇⋅

Navier-Stokes (momentum) equation then becomes

( ) Gpuuu21

tu 2

∇+−∇=×∇×−∇+∂∂

For a potential flow (incompressible and irrotational),

φ∇=u where φ is the potential function

then

∂φ∂

∇=φ∇∂∂

=∂∂

tttu

also ( )[ ] 0u)u(u ≡φ∇×∇×=×∇×

One gets

0Gpu21

t2

=

−++

∂φ∂

∇ …………………….(∆)

Integrating along a streamline yields

( )tFGpu21

t2

=−++∂φ∂

(Bernoulli’s equation)

The continuity equation

0u =⋅∇

can become ( ) 02 =φ∇=φ∇⋅∇ (Laplace equation)

13

Two-Dimensional Boundary Layer on a Flat Plate

The momentum equations are (2-D laminar; gravity effect neglected)

x-direction:

∂∂

+∂∂

ν+∂∂

ρ−=

∂∂

+∂∂

+∂∂

2

2

2

2

yu

xu

xp1

yuv

xuu

tu

y-direction:

∂∂

+∂∂

ν+∂∂

ρ−=

∂∂

+∂∂

+∂∂

2

2

2

2

yv

xv

yp1

yvv

xvu

tv

The continuity equation is 0yv

xu

=∂∂

+∂∂

As the Prandtl’s slender flow assumption is taken as 1L<<

δ

0yv

xu

=∂∂

+∂∂

0VOLUO =

δ

+

δ∴

V~LU or U

L~V δ or 1

L~

UV

<<δ

x-momentum: (UL~t )

2

2

2

2

yu

xu

xp1

yuv

xuu

tu

∂∂

ν+∂∂

ν+∂∂

ρ−=

∂∂

+∂∂

+∂∂

LU2

L

U2

δ

δ UUL

? 2LU

ν << 2

ν

( 2UL

× ) [ ]1O [ ]1O [ ]1O ? LUν 2

2LLU δν

ε << 2L

δ

ε

U U U

δy

xL

∆1 ∆2

Blasius Layer

14

Suppose one considers a limiting case as 0→ε and drops 1∆ and 2∆ , the

case returns to inviscid flows again since viscosity is no longer involved.

Accordingly, one must have

[ ]1O~L 2

δ

ε

or LRe

1ε~Lδ

= where νLUReL =

y-momentum:

ν=εε

δLU

; U~UL

~V

∂∂

+∂∂

ν+∂∂

ρ−=

∂∂

+∂∂

+∂∂

2

2

2

2

yv

xv

yp1

yvv

xvu

tv

UL

L Uδ

L

UU Lδ ( )

δ

δ 22L U A∆ 2

L

LUδν 2

L Uδ

ν δ

× 2U

L

δ

L

δ

L

δ

L B∆

ULLν

δ

δν L

UL

ε ε ε ? 23ε ε

Therefore, for a large LRe , 0yp1→

∂∂

ρ−

Then )x(pp ≅ . That is, p is not a function of y .

( ) ( )xpy,xp limp psolution

flow-potential0y=≅

* 2

2

xu

∂∂

ν is negligible compared to 2

2

yuν

∂∂ .

* εδ ~L

15

Outside the boundary layer: ( )smallis u2∇ν

xp1

yuv

xuu

tu

∂∂

ρ−=

∂∂

+∂∂

+∂∂

yp1

yvv

xvu

tv

∂∂

ρ−=

∂∂

+∂∂

+∂∂

0yv

xu

=∂∂

+∂∂

Within the boundary layer:

2

2p

yu

xp1

yuv

xuu

tu

∂∂

ν+∂∂

ρ−=

∂∂

+∂∂

+∂∂

0yv

xu

=∂∂

+∂∂

* 1) The above are valid for incompressible flows.

* 2) Both forms are approximated outcomes.

21 QQ > , flow in section <2> will transport momentum upwards a little bit.

Generally, however, it will not significantly affect the free stream velocity since the thickness of boundary layer is small.

d 1Q 2Q

< 1 > < 2 >

some fluid are thrown out of b .l.

3 unknowns:u, v, p

3 equations

2 unknowns:u, v

2 equations

Euler equations

“entrainment”

16

Normalized form of Navier-Stokes equations:

∂∂

+∂∂

ε+∂∂

−=∂∂

+∂∂

+∂∂

2

2

2

2

yu

xu

xp

yuv

xuu

tu

LURe1 ν

==ε

∂∂

+∂∂

ε+∂∂

−=∂∂

+∂∂

+∂∂

2

2

2

2

yv

xv

yp

yvv

xvu

tv

Let ε

=ξy , vv ε=

then ξ∂

∂ε

=∂ξ∂

ξ∂∂

=∂∂ v1

yv

yv

The continuity equation becomes 0v1xu

=ξ∂

∂ε

+∂∂

or 0vxu

=ξ∂

∂+

∂∂

x-momentum equation:

2

2

2

2 uxu

xpu1v

xuu

tu

ξ∂∂

εε

+∂∂

ε+∂∂

−=ξ∂

∂ε

⋅ε+∂∂

+∂∂

or 2

2uxpuv

xuu

tu

ξ∂∂

+∂∂

−=ξ∂

∂+

∂∂

+∂∂ (neglect 2

2

xuε

∂∂ )

y-momentum equation:

2

2

2

2 vxvp1vv

xvu

tv

ξ∂∂

εε

ε+∂∂

εε+ξ∂

∂ε

−=ξ∂

∂ε

εε

+∂∂

ε+∂∂

ε

( )ε× 2

2

2

22 v

xvpvv

xvu

tv

ξ∂∂

ε+∂∂

ε+ξ∂

∂−=

ξ∂∂

ε+∂∂

ε+∂∂

ε

0 0 0 0 0

For smallRe1ε →= 0p

→ξ∂

17

Laminar Boundary Layer Equation for Blasius Layer (steady 0

t=

∂∂ ; zero-pressure gradient 0

xdpd= )

The governing equations are

2

2

yu

yuv

xuu

∂∂

ν=∂∂

+∂∂ ……………. (1)

0yv

xu

=∂∂

+∂∂ …………….. (2)

with B.C.’s: 0)0,x(v)0,x(u == (no-slip, impermeable condition)

U),x(u =∞

Prandtl: 1) The larger the L, boundary layer thickness will grow.

2) The larger the U, the thinner the boundary layer thickness.

3) The larger the ν, the thicker the boundary layer thickness.

Due to dimensional homogeneity, one has

UL~ ν

δ or LU

~L

νδ

Accordingly, one has xU

~x

νδ or Ux~ ν

δ

Question: Are all )x(

y~U

)y,x(uδ

the same? (δ is “some” thickness)

( )xδy

U UU

18

For a steady, laminar boundary layer with 0xdpd p = (U = constant), Prandtl

assumes: (similarity assumption)

δ

=yg

Uu where g is a similarity function of

δy

For Ux~ ν

δ , η=νδ xUy~y

and x

Uy ν=

∂η∂ ,

x2Uyx

21]xUy[

xx2/32/1 η

−=ν

−=ν∂

∂=

∂η∂ −−

One looks for a solution of the form (steady, 0xdpd= for Blasius layer)

)(gUu

η=

and tries to make it satisfy the continuity equation.

Let’s define a stream function, ψ , which satisfies

yyu ψ−=

∂ψ∂

−= (Some defines y

u∂ψ∂

= , x

v∂ψ∂

−= )

and xxv ψ=

∂ψ∂

=

Choose ( )ηfxUνψ −=

then 'fUx

Ud

fdxUy

u =νη

ν=∂ψ∂

−= or g'fUu

≡=

Suppose one chooses ( )η−=ψ f

( )ην

=∂η∂

η+=

∂ψ∂

−= 'fx

Uyd

fdy

u

then ( )ην

= 'fxU

1Uu (N.G.)

since 'f and LHS are dimensionless and xUν is dimensional.

19

yxx xuu ψ−=

∂∂

=⇒ xyy yvv ψ=

∂∂

=

yyy yuu ψ−=

∂∂

= yyyyyu ψ−=

Substituting into equations (1) and (2) on pp. 17 gives

0xyyx ≡ψ+ψ− (O.K.)

yyyyyxyxy ψν−=ψψ−ψψ+ ………………….. ( )∆

For ( )ην−=ψ fxU where x

Uyν

xU

y ν=

∂η∂

x2x1

xUy

21x

21

xUy

x2

3 η−=

ν

−=

ν=

∂η∂ −

( )ην∂∂

−=∂ψ∂

=ψ fxUyyy yd

fdxU∂η∂

ην−= 'fU

xνU'fxUν −=−=

( )'fUxx

yyx −

∂∂

=∂ψ∂

=ψ "fx2

Ux

"fU η+=

∂η∂

−=

( )]fUx[xxx ην

∂∂

−=∂ψ∂

=ψx

'fxUfxU

21

∂η∂

ν−ν

−=

η−ν−

ν−=

x2'fxUf

xU

21 ( )f'f

xU

21

−ην

=

"fx

UUyy ν−=ψ

'''fx

UUyyy ν−=ψ

* 'fUu y =ψ−= ,

δ

=y'f

Uu

20

Substitution into equation )(∆ on pp. 19 leads to

( ) '''fx

UUf'f"fxU

21

xUU

x"ff

2U2

νν=−η

ν⋅

ν+

η−

or ( ) '''fxUf'f"f

x1

2U

x"ff

2U 222

=−η+η

Finally, 0f"f'''f2 =+ → 3rd order non-linear equation

with 3 boundary conditions as

( ) 00f0v0y

==η⇒==

( ) 00'f0u0y

==η⇒==

( ) 1η'fUuy

=∞=⇒=∞=

The solution techniques:Frohenius Method

Finite difference (numerical) Galerkin-finite element

Noted that for 0.5≅η %99Uu

=

and ( )ηf is linear as 0.5>η (free-stream portion)

( ) )(f)5(d'fd'fd'fd'ff 505

500 η=−η+∫ η=∫ η+η∫=η∫=η ηη

Nominal boundary layer thickness (Some chooses 95δ ):

At 99y δ= U99.0u = (see Table on pp. 21)

xU0.5

x99 ν=

δ → Laminar (Blasius) boundary layer

( )0x0y

f'fxU

21v

=η=−η

ν=ψ=

1'f = when η > 5

Falkner-Skan Equation

(#)

21

Table of the function )(f η for the boundary layer along a flat plate at zero incidence after L. Howarth

xUyν

f Uuf =′

f ′′

0 0.2 0.4 0.6 0.8 1.0

1.2 1.4 1.6 1.8 2.0

2.2 2.4 2.6 2.8 3.0

3.2 3.4 3.6 3.8 4.0

4.2 4.4 4.6 4.8 5.0

5.2 5.4 5.6 5.8 6.0

6.2 6.4 6.6 6.8 7.0

7.2 7.4 7.6 7.8 8.0

8.2 8.4 8.6 8.8

0 0.00664 0.02656 0.05974 0.10611 0.16557

0.23795 0.32298 0.42032 0.52952 0.65003

0.78120 0.92230 1.07252 1.23099 1.39682

1.56911 1.74696 1.92954 2.11605 2.30576

2.49806 2.69238 2.88826 3.08534 3.28329

3.48189 3.68094 3.88031 4.07990 4.27964

4.47948 4.67938 4.87931 5.07928 5.27926

5.47925 5.67924 5.87924 6.07923 6.27923

6.47923 6.67923 6.87923 7.07923

0 0.06641 0.13277 0.19894 0.26471 0.32979

0.39378 0.45627 0.51676 0.57477 0.62977

0.68132 0.72899 0.77246 0.81152 0.84605

0.87609 0.90177 0.92333 0.94112 0.95552

0.96696 0.97587 0.98269 0.98779 0.99155

0.99425 0.99616 0.99748 0.99838 0.99898

0.99937 0.99961 0.99977 0.99987 0.99992

0.99996 0.99998 0.99999 1.00000 1.00000

1.00000 1.00000 1.00000 1.00000

0.33206 0.33199 0.33147 0.33008 0.32739 0.32301

0.31659 0.30787 0.29667 0.28293 0.26675

0.24835 0.22809 0.20646 0.18401 0.16136

0.13913 0.11788 0.09809 0.08013 0.06424

0.05052 0.03897 0.02948 0.02187 0.01591

0.01134 0.00793 0.00543 0.00365 0.00240

0.00155 0.00098 0.00061 0.00037 0.00022

0.00013 0.00007 0.00004 0.00002 0.00001

0.00001 0.00000 0.00000 0.00000

22

For 0.5>η

( )ηf )ttancons0.5( −−η= ]ηd)η('fηd1[η5

0

5

0∫ ∫−−= ( )∫ η−−η=5

0d'f1

area of deficit ∆

( )η= 'fUu

xUyν

=η , dyx

Udν

( ) ( ) η∫ η=η η d'ff 0

( )η∫

η= η d

Uu

0

( )dyyuU1

xU y

0∫ν=

q

Then ( )ην= fxUq

23

Summary: (for Blasius layer)

1) Laminar boundary layer

2) Zero-pressure gradient; UUp = , 0xdpd p =

3) Similarity profile: ( )η= 'fUu where

xUyyν

5=η , ( ) 99.0'fUu

≅η=

dyuq y0∫= ( ) ηη

ν= ∫

ηd'f

UxU

0

( ) ( )ην=ηην= ∫η

fxUd'fxU0

[t

L2

] [0] → [Lt

L3

⋅]

Entrainment (transverse) velocity at the edge of boundary layer:

( ) ( )∞ψ=∞ xv ( )f'flimUx

21

−ην

=∞→η

(see # on pp. 20)

As ∞→η , 1'f = ( )[ ]∫ η−η−+⋅ην

⇒∞→η

5

0d)'f1(1lim

xU

21

( )∫ η−ν

=5

0d'f1

xU

21

area of deficit

[ ])028.3()05(xU

21

−−−ν

= xU86.0 ν

Or ( )xRe

86.0xU

86.0U

v=

ν=

See ∆ on pp. 22

24

Generally, 0v y ⇒∞→

Since we are looking at the problem in a “micro” scale, the error of x-velocity will happen far away from the area of interest.

Drag on a Flat Plate

In a 2-D case, ( )0or0yw xv

yu

=η=

∂∂

+∂∂

µ=τ

For ( )η= 'fUu ,

xUyν

( )y'fU

yu

∂η∂

=∂∂

xU"fU

yd'fdU

ν=

∂η∂

η=

thus ( )0η

w xνUη"fUμτ

=

= ( )x

U0"fUν

µ=x

UU332.0ν

µ=

0.332

U

( )↑v

L

bx

>>

25

Local friction coefficient: (normalized form of wall shear stress)

( )xU

1332.02U

)x('C

22

wf νρ

µ=

τ= ρ

Another definition:

xU664.0 ν

= ⇒ 'C21C f

*f =

xRe1664.0=

x2

w

Re1332.0

Uρτ

==

Total drag on one side of plate:

dxbτDL

0 w∫ ⋅= where 0y

w yuμτ

=∂∂

=

∫ −⋅ν

µ=L

0dxxbUU)332.0(D 2

1 ( ) LbUU)664.0(ν

νρ=

LLU

bU664.0 2 νρ=

x

0τ singularity (solution is not satisfactory )

Hopefully, for small sx , dxτsx

0 w∫ →small << dxτL

x ws∫

∼ 2/1x −

26

Global Friction Factor: (flat-plate, Length = L ; ν

=LUReL )

( )Lb2UD2C 2

21f ρ

= LbU

LULbU)664.0(2

2

2

ρ

νρ

= LRe

328.1= ()

Displacement Thickness: ( )1δ

Total discharge deficit:

Udy)uU( 10δ≡−∫

therefore dyUu1

01 ∫∞

−=δ

Recall that ( )η= 'fUu where

xUyν

dyx

Udν

=η , ην

= dUxdy

For Blasius layer

[ ] ηη−ν

=δ ∫∞

d)('f1Ux

01 ( ))(flimUx

η−ην

=∞→η U

x72.1 ν=

x

1

Re72.1

x=

δ⇒ , where

ν=

xURex

( )uU −

u

U

A

U

B1δ

Some use *δ .

e.g. pick up 5>η 72.1)08.78.8(8.8 =−⇒=η

27

Momentum thickness: ( )2δ

uρ → smmkg

3 ⋅ → volumemomentumx −

Flux of momentum deficit )uU(u −ρ⋅=

Discharge of momentum deficit dy1)uU(u0

⋅⋅−ρ= ∫∞

22U δρ≡

Thus dyUu1

Uu

02 ∫∞

−=δ

For Blasius layer [ ] ηη−ην

= ∫∞

d)('f1)('fUx

0

x

2

Re664.0

x=

δ⇒

Integral Mass Balance

For incompressible flows, the continuity equation is

0yv

xu

=∂∂

+∂∂

If one tries to calculate ∫∞

∂∂

0dy

xu , it leads to an infinite number. However, if

one replaces xu

∂∂ by )uU(

x−

∂∂

∂∂

∫∞

00dy

xU , the result is

finite.

Noted that U is neither a function of x nor y for Blasius layer.

some uses θ

numerical result=0.664

28

dyyvdy)uU(

x 00 ∫∫∞∞

∂∂

+−∂∂

( ) ∞∞ +∫ −∂∂

−= 00 vdyuUx

[ ( ) 00yv == → non-penetrable ]

( ) 0vxd

Ud 1 ⇒+δ

−= ∞ U is a spatial constant.

Entrainment velocity x~1δ

( )↑δ∞v~

x1~

xdd 1

0xd

vd<∞ (decreases with x)

* The discharge deficit will result in upward entrainment.

Since area of velocity deficit increases as x increases, ( )1Uxd

dδ is positive.

Therefore, ∞v is also positive ( )↑ . Accordingly, within the Blasius layer, the

growth of velocity deficit (due to long distance of friction effect) tends to produce upward entrainment.

Shape Factor:

2

1Hδδ

=

For Blasius Layer (laminar)

==664.072.1H 2.59

Noted that the shape factor of a turbulent boundary layer is about 1.3 .

29

Summary: for Blasius Layer

(Laminar boundary layer with 0xdpd p = or U = constant)

xUy ν

xU0.5

x99. ν=

δ

)('fUu

η=

)(fxUq ην=

xνUUν332.0τw =

xU664.0'Cf

ν=

x

1

Re72.1

x=

δ

x

2

Re664.0

x=

δ

xRe860.0

Uv

=∞

59.2H =

30

Two-Dimensional Flow on a Convex Curved Surface

L

R(x)

y

x

dxdpp

U

The momentum equation in the normal direction (y-direction) is

+∇ν+∂∂

ρ−=−

∂∂

+∂∂

+∂∂ v

yp1

Ru

yvv

xvu

tv 2

2

(body force terms)

UL

L Uδ

L

UU Lδ

( )δ

δ 2L U

RU2

δ

2U

δ

× 2U

2

L

δ

2

L

δ

2

L

δ

Rδ 1

νLU

O ⇒ For large LRe , A○1 E

A, A○2 E

A, A○3 E

A→0

If [ ]1OR

<<δ , 0

ydpd→ or p=p(x only)

same as that in the flat plate case

⇒ Boundary layer approximations still holds for large LRe and small Rδ .

For L

~Uv δ ,

UL~t ,

2U~p ρ

31

Boundary Layer Equation on a 2-D Body with Outer Potential Flow

with 0xdpd p ≠

( or )x(UU = ) and R <<δ

0yv

xu

=∂∂

+∂∂

2

2p

yu

xdpd1

yuv

xuu

∂∂

ν+ρ

−=∂∂

+∂∂

In the potential flow, along a streamline (boundary layer edge), Bernoulli’s equation yields

=ρ++ρ zgpU2 p

2p constant

By neglecting z effect (assume on a horizontal plane)

=+ρ

p2p pU

2 constant

Then 0xdpd

xdUd

U ppp =+ρ or

xdUd

Uxdpd1 p

pp =

ρ−

where )x(Up is the potential-flow (tangential) velocity (inviscid)

extrapolated to y = 0 (surface).

y

x

y

x

x:arc axis

y:axis normal to x (dependent on x)

32

(I) 0xdpd<

→↑

smalleargl ppU

0xdUd> earglsmall UU → → Favorable (negative) pressure gradient

(II) 0dx

pd>

→↓

earglsmall ppU

0xdUd< smalleargl UU → → Adverse (positive) pressure gradient

Recall:Potential-flow solution of flow around a circular cylinder

R

2U

2U

u=0u=0

xU U

( )

=

Rxsin2

Uxu

AdverseFavorableAdverse Favorable

0dxdp

> 0dxdp

< 0dxdp

> 0dxdp

<

possible region of separation

33

Von Karman’s Momentum Integral Equation For incompressible boundary layer can be expanded to TBL

The boundary layer equations are:(unsteady form)

0yv

xu

=∂∂

+∂∂ …………………….. (1)

2

2p

yu

xdpd1

yuv

xuu

tu

∂∂

ν+ρ

−=∂∂

+∂∂

+∂∂ …………………….. (2)

where pp is the pressure at the boundary layer edge.

Noted that, in the outer (potential flow) region, local pressure along the edge of boundary layer is related to inviscid flow velocity along the surface (y = 0)

based on the Navier-Stokes equation

∂∂

ν∂∂

ν dropped y

,x 2

2

2

2

.

xdpd1

yUV

xUU

tU p

y ρ−=

∂∂

+∂∂

+∂∂

δ=

at boundary layer edge xdpd p⇒

(2)⇒ 2

2

yu

xUU

tU

yuv

xuu

tu

∂∂

ν+∂∂

+∂∂

=∂∂

+∂∂

+∂∂ ……….. (2a)

where U is the edge velocity. Re-arranging equation (2a) gives

2

2

yu

xUU

yuv

xuu)Uu(

t ∂∂

ν+∂∂

=∂∂

+∂∂

+−∂∂ ………………. (2b)

1∆ 0 (for incompressible flows)

∂∂

+∂∂

+∂∂

+∂∂

=∆yv

xuu

yuv

xuu1 )vu(

yxuu2

∂∂

+∂∂

= )vu(y

)u(x

2

∂∂

+∂∂

=

2∆

34

[ ] )Uv(y

)Uu(vy2 ∂

∂+−

∂∂

=∆

[ ]yUv

yvU)Uu(v

y ∂∂

+∂∂

+−∂∂

= since U is function of x only

[ ]

∂∂

−−∂∂

=xuU)Uu(v

y[ ]

xUu)Uu(

x)Uu(v

y ∂∂

+∂∂

−−∂∂

=

Equation (2b) then becomes

[ ] [ ] 2

2

yu)Uu(v

yxU)Uu()Uu(u

x)Uu(

t ∂∂

ν=−∂∂

+∂∂

−+−∂∂

+−∂∂ … (2c)

Integrating equation (2c) with respect to y from 0 to ∞ (or δ ) gives

[ ] [ ]∞∞

∞∞∞

∂∂

ν=−+∂∂

−+−∂∂

+−∂∂

∫∫∫00

000 yu)Uu(vdy

xU)Uu(dy)Uu(u

xdy)Uu(

t

0y00

2

0 yu00dy1

UuU

xUdy1

Uu

UuU

xdy1

UuU

t=

∞∞∞

∂∂

ν−=+

∂∂

+

∂∂

+

∂∂

∫∫∫ Changing sign yields

ρτ

=δ∂∂

+δ∂∂

+δ∂∂ w

122

1 xUU)U(

x)U(

t

von Karman’s integral equation Suppose velocity profiles are of a similar type (not always)

( )η= GUu , ( )x

=η , 10 ≤η≤

* 0)Uu( =− ∞ and 0v0 → (impermeable boundary)

0yu

=

∂∂

∫∞

−=δ

01 dyUu1

δ

−δ=

1

0

ydUu1

[ ] ηη−δ= ∫ d)(G11

0

∫∞

−=δ

02 dyUu1

Uu

δ

−δ=

1

0

ydUu1

Uu

[ ] ηη−ηδ= ∫ d)(G1)(G1

0

35

Example: (a steady case)

≤η≤η>η

=η=10 ,

1 ,1)(f

Uu

and U = constant (independent of x & t ) The von Karman’s integrate equation is

ρτ

=δ∂∂

+δ∂∂

+δ∂∂ w

122

1 xUUU

xU

t

One has ρτ

=δ∂∂ w

22

xU

Substitution leads to U6

xdd ν

δ and further to get U

12xd

d 2 ν=δ

B.C. Supposed at the leading edge (x = 0), 0=δ , then

Ux122 ν

=δ ⇒ Ux12 ν=δ ⇒

xRe46.3

x=

δ

For 21δ

62δ

U

0

δ

y

u

η−δ=δ ∫ d)f1(f1

02 η−δ=δ ∫ d)f1(1

01

ηη−ηδ= ∫ d)1(1

0 ηη−δ= ∫ d)1(

1

0

632

1

0

32 δ=

η−

ηδ=

22

1

0

2 δ=

η−ηδ=

Also, δν

=δρ

µ=

ρτ UUw

δν

=

δ⇒

U6xd

dU2

3H2

1 =δδ

= ( Blasius layer: 2.59 , pp. 29)

36

x

1

Re73.1

x=

δ (Blasius layer: 1.72)

x

2

Re576.0

x=

δ (Blasius layer: 0.664) → more sensitive

For δν

=ρτ Uw

Ux12

ν=

xU

121 3ν

= ,

the local friction coefficient is

x2

2

wf Re

58.0U

'C =τ

= ρ (Blasius Layer: x

f Re664.0'C = )

Homework:

Assume

1) 22Uu

η−η= , δ

=ηy

2)

π

⋅η=2

sinUu

Redo the calculation. Tabulate the results and compare to the Blasius solution.

37

Effect of Pressure Gradient on the Fluid Motion (Approximated Analysis)

For a small fluid parcel with 1-D

motion (xdpd

= constant)

0xdpd< 21 pp > (favorable)

0xdpd> 21 pp < (adverse)

pΔpp 12 +=

Newton’s second law gives

tdudmFx =∑ xA∆ρ

then ( )tdudxΔAρ

tdudmΔApp 21 ==−

and xdpd1

tdud

ρ−= λ−=

Assume for a small t∆ , tΔUxΔ = , where U is the local velocity independent of x (actually not quite) then tdUxd =

One gets ( ) λ−=U

xdud or

Uxdud λ

−=

Then cxU

u +λ

−= with I.C. as 0x = ( ) 0U0u =

By substitution, one gets xU

Uu 0λ

−= xxdpd

U1U0

ρ

−=

11Ap 22Ap

x∆

cross-sectionalarea=A

)(xΔxdpd

±

38

(I) 0xdpd< (favorable), and

xdpd = constant (neither a function of x nor y)

2Bu

2Au

y

u0

2xx =

1Au

0Au

y

u0

X=0

1Bu

1Au

y

u0

1xx =

** Profile becomes fuller as x increases.

By

Ay

0Bu

0Au

y

u0

Initial velocity profile (x=0)

0Bu

u

x0

0Au

1x 2x

xxdpd

U1Uu 0

ρ

+=

39

(II)xdpd = constant > 0 (adverse pressure gradient)

* Separation would occur near the bottom since this area contains the least

momentum.

0Bu

x0

0Au

1x 2x

y

u0

0xx = y

u0

1xx = y

u0

2xx =

xxdpd

U1Uu 0

ρ−=

40

Steady Flow Past A Wedge

The potential-flow (analytical) solution at 0y = is (for 02 >β> )

m1 xu)x(U = (Note: 0x = 0U = → stagnation point)

where 1u is “some” velocity and β−

β=

2m

(** For 02 >β> , ↑β , ↑m )

(i) 0m = ( )0=β → flow along a flat plate

(ii) 1m = ( )1=β → stagnation-point flow

(iii) 0m < ( )0<β

For xd

)x(Ud)x(Uxdpd1 p =

ρ− ( ) 1m22

11m

1m

1 xumxum)xu( −− == ,

the boundary layer equations (steady) become

0yv

xu

=∂∂

+∂∂

2

21m22

1 yuxum

yuv

xuu

∂∂

ν+=∂∂

+∂∂ − ……………………… ( )∆

∞U

βπ

yx

U=0

*1uU =

(singularity , separation maybe occur )

(Imagined) → physically does not exist

41

To get a similarity solution,

Let xν

)x(U2

1myη += so as to get a simpler form of O.D.E.

For m1 xu)x(U =

21m

xu2

1my 1−

ν+

=η 21m

xu2

1myd

d 1−

ν+

Introduce )(f1m

2)x(U η+

ν−=ψ

)(fxu1m

22

1m

1 ην+

−=+

(Compared to that on pp. 18 .)

yu ψ−=y∂ψ∂

−= 21m

21m

xu2

1m)('fxu1m

2 11

−+

ν+

ην+

+=

)('fxu m1 η= )('fU η=

Or )('fUu

η=

By substituting into ( )∆ on pp. 40, one gets

0)'f1("ff'''f 2 =−β++ where 1m

m2+

=β (see pp. 20)

B.C.’s: (3)

0)0(f = 0v 0y =← = (same reason as # on pp. 20)

0)0('f = 0u 0y =← = (no-slip)

1)('f =∞ )x(Uu y =← δ=

** For a flat plate case ( )0m = , xν

Uy21η =

which is a little bit different from the previous one on pp. 18.

42

As 091.0m −< ( )199.0−<β , separation occurs.

When separation occurs, ( )( )

0d

d

0y0

Uu

==η

p velocity

A 0dxdpp <

Favorable (negative)pressure gradient

p velocity

A 0dxdpp >

Adverse (positive)pressure gradient

0yu

00

=∂∂=τ

* If 0=β (flat-plate case), one gets

0f"f'''f =+ → (Previously on pp. 20, the equation is 0f"f'''f2 =+ . The difference is due to the inconsistency of η definition.)

Blasius

43

Incipient separation occurs when τw = 0 (for 2-D boundary layers)

3-D B.L. separation criterion is different.

For wedge flows,

0yw y

u)x(=

∂∂

µ=τ )('fUu

η= , x

)x(U2

1myν

+=η

0yydud

=∂η∂

ηµ= m

1 xuU =

21m

xu2

1m)0(''fU 1−

ν+

µ=

For the flow around a circular cylinder: (according to the inviscid solution)

Favorable Adverse pressure gradient

2U

u=0 u=0

If separation occurs, if must be within this region

(where or )0τ 0yu→

∂∂

Nominal boundary layer thickness (δ) is evaluated at )('f99.0Uu

η== .

?99 =η

For wedge flows, 21m

xνu

21myη 1

−+=

21m

xνu

21mδη 1

99

−+=∴

If separation occurs, it must be within this region.

44

Since 99η , 1u , ν are constant, 21m

x~−−δ .

Growth rate of boundary layer along the wedge surface:

21m

x2

1m~xd

d +−−−

δ , 23m

x4

1m~xd

dxd

d 2+−−

δ

(#)

⇒ 02 ≥β> 02

m ≥β−

β=

m1 xu)x(U =

1m1 xum

xdUd −=

21m

x2

1m~xd

d +−−−

δ

23m

x4

1m~xd

d 2

2

2+−−δ

< I > m > 0,

0xdUd> → Favorable

(i) m > 1

0xd

d<

δ

δ decreases with x .

Strength of favorable pressure gradient overcomes natural growth of the boundary layer.

U=0βπ

U=0 (stagnation point)

1>m=0

m=0

m>1

Reasons of increase of δ: 1) natural growth 2) adverse pressure gradient

* Favorable pressure gradient tends to make boundary layer thinner.

45

(ii) 1 > m > 0

0xd

d>

δ ,

δ increases with x

Strength of favorable pressure gradient is not strong enough to overcome the natural growth rate of boundary layer.

Also, since 0xd

d2

2

<δ , the growth rate of

xdd δ decreases with x .

(iii) m = 1 ( 1=β , angle =π ) → stagnation-point flow

< II > m < 0 (Imagined, 0<β ) → Stagnation point does not exist.

0xdUd< → Adverse pressure gradient

0xd

d>

δ , δ grows with x .

δ

U=0

At large x, b.l. thickness asymptotically equals to d

xuU 1=

1uxdUd=

0xd

d=

δ , δ does not grow.

46

Free Shear Flows (for which boundary layer approximation holds)

Free stream velocity is constant, 0xdpd=

(1) Wake

(2) Jet

(3) Mixing Layer

velocity deficit

47

Laminar Wake Far Downstream of A Flat Plate ( 0xdpd= )

* The momentum deficit is due to the action of viscous drag on the plate.

Continuity equation:

0yv

xu

=∂∂

+∂∂

Integration yields 0dyyv

xu

0≡

∂∂

+∂∂

∫∞

(upper half)

Then 0)0(vVdyuxd

d0

≡−+∫∞

∫∞

∞ −=∴0

dyuxd

dV ………..……………. (1)

Momentum equation: (boundary-layer equation)

2

22

yu

yvu

xu

∂∂

ν=∂∂

+∂∂

Integrating with respect to y from 0 to ∞ leads to (upper half)

U

y

1u

L

width = ”b ” δ

x

impermeable (x < L) and symmetric (x > L)

For yuv

xuu

∂∂

+∂∂

yuv

xuu

yv

xuu

∂∂

+∂∂

+

∂∂

+∂∂

= vuy

ux

2

∂∂

+∂∂

=

48

∞∞∞

∂∂

ν=+∫0

00

2

yuvudyu

xdd

or

∂∂

−∂∂

ν=−+∞

∫0

000

2

yu

yu)vuVU(dyu

xdd ……............... (2)

At plate, u = v = 0 (no-slip, impermeable) After plate, v = 0 (symmetric)

Substituting (1) into (2) gives

ρτ

−=− ∫∫∞∞ 000

2 dyuxd

dUdyuxd

d

or ρτ

=−∫∞ 00

dy)uU(uxd

d

* ])x(δU[xd

dLHS 22=

Integrating with respect to x from 0 to x ( > L) leads to

( ) ∫∫ τρ

=δL

0 0

x

0 22 dx)x(bdx)x(U

xddb ( 0τ0 = for x > L)

( ) dx)L(Uxd

dbLHSL

0 22∫ δ=∴

Let 22 )L( δ=δ (at the trailing edge of the plate) ⇒ bUD 22 δρ=

One tries to find a similarity solution at far wake ( Lx >> , Uu1 << ).

Define the velocity deficit as uUu1 −= , or 1uUu −=

Continuity:

0yv

x)uU( 1 =

∂∂

+∂−∂ or 0

yv

xu1 =

∂∂

+∂∂

δV~

Lu1

Note:For x > L, ρ

=τρ

=τρ ∫∫

Ddx)x(bdx)x(b L

0 0

x

0 0

49

Momentum:

)uU(y

)uU(y

v)uU(x

)uU( 12

2

111 −∂∂

ν=−∂∂

+−∂∂

21

211

11

yu

yuv

xuu

xuU

∂∂

ν=∂∂

+∂∂

−∂∂

LuU 1

Luu 1

1 δ

δ 11

uL

u 21u

δν

×

UuL1

1 Uu1

Uu1

2LLU

δ

ν

○AE

A A○B E

A A○C E

For 1Uu1 << ( far wake ), A○AE

A >> A○B E

A , A○AE

A >> A○C E

Since 21

2

yu

∂∂

ν cannot be neglected (for viscous flows)

1~LLU

2

δ

ν LRe

1~Lδ

⇒ For a large LRe , the approximate boundary layer equation (wake) is

21

21

yu

xuU

∂∂

ν=∂∂ ………………………… (3)

B.C.’s:

A○1 E

A For 00y=τ

=, 0

y)uU(

0y

1 =∂−∂

=

or 0yu

0y

1 =∂∂

=

at x > L

A○2 E

A 0uy1 =

∞→

Note: ( ) *)t(uuLHS 1

Ux

1

∂∂

→∂∂

⇒ . The form is similar to a heat equation.

(symmetric)

50

Since 1u must decay with x due to diffusion,

Let Uu1 ~ m)

Lx( − , where m > 0

Guess )(g)Lx(C

Uu m1 η= − → similarity solution

where x

Uyν

At far wake Uu1 << 1 ,

Then D≒ ∫∞

ρ0

12 dyUuUb ∫ η

νρ= ∞

012 d

Uu

UxUb ……….…. (4)

Or D≒ ∫∞− ηη

ν

ρ0

m2 d)(g)Lx(C

UxUb

Since RHS becomes independent of x as x>L, thus 21m = .

⇒ )(g)Lx(CUu 2

1

1 η= − , where x

Uyν

ν=

∂η∂

η−=

∂η∂

xU

y

x2x

Examine equation (3) on pp. 49

21

21

yu

xuU

∂∂

ν=∂∂

Recalled that (pp. 48)

22UbD δρ= → dose not change with x as x>L

∫ =−ρ ∞0

112 Ddy)Uu1(

UuUb

51

[ ]'ggx1)

Lx(CU

21)

x2)((g)

Lx(CU)(gLxCU

21

xu

21

21

21

231 η+−=

η−η′+η−=

∂∂ −−−−

xU)(g)

Lx(CU

yu

211

νη′=

∂∂ −

xUg)

Lx(CU

xU

xU)(g)

Lx(CU

yu

21

21

21

2

ν′′=

ννη′′=

∂∂ −−

Substitution gives

ν

η′′ν=

′η+− −−

xU)(g)

Lx(UC)gg(

x1)

Lx(CU

21U 2

12

1

Or g)gg(21 ′′=′η+− then g)g(

dd

21 ′′=η

η−

Thus 0)g(dd

21

dgd2

2

=ηη

(O. D. E.)

Or 0)g(21

dgd

dd

=

η+

ηη

Integration leads to =η+η

g21

dgd C1

○aE

A At 0y = , 0yu1 =∂∂ ( x > L; far wake; symmetric)

Or for 0=η , 0d

gd→

η thus C1 = 0

ηη−= d21

ggd then

241

eCgη−∗=

⇒ Take 1C =∗ (this will be absorbed into C )

Noted that from a mathematic table, π=η∫ η∞ d)(g0

52

From (4) on pp.50

η∫ ην

ρ= ∞− d)(g)Lx(C

UxUbD 0

2 21

πν

ρ= CLU

Ub 2

Since in the flat plate case, LU

LUb664.0D 2 νρ= (see pp. 25)

Comparison leads to π

=664.0C

Finally, the solution is

ν

−π

=−

xUy

41exp)

Lx(664.0

Uu 22

11

53

Laminar Plane Jet (Ambient fluid is still)

* Mostly, jets are turbulent due to high Reynolds number (ν

= 00j

bURe > 40)

Continuity:

0yv

xu

=∂∂

+∂∂

Integrating with respect to y from ∞− to ∞+ leads to

0vdyuxd

d≡+ ∞

∞−

∞−∫

Define entrainment velocity eV (positive downwards)

then eee V2)V(Vv =−−=∞∞−

Momentum: (boundary-layer equation)

2

22

yu

yvu

xu

∂∂

ν=∂∂

+∂∂

Integration gives ∞∞−

∞∞−

∞∞−∫

∂∂

ν=+yuvudyu

xdd 2

0-0 0-0

Thus 0dyuxd

d 2 =∫∞∞− or =∫

∞∞− dyu2 constant = 0

20 bU = K ….(A)

Similarity:

0b

x

bcu

0b

y0U

eV

Note: For the flat plate case, the entrainment is positive outwards.

At 0x = , 0Uu = , width 0b=

αx~uc )0( <α ; βx~b )0( >β

Uo

54

Assume a similarity profile ))x(b

y(fun~)x(u

uc

then )xy(funx~u β

α ⋅

Try a stream function )xy(fx~ q

pψ , ( qxy

=η , qx1

y=

∂η∂ , 1qx

yqx +−=

∂η∂ )

For qqp

x1)

xy(fx~

yu ′

∂ψ∂

−= , then qp −=α , q=β .

Therefore, )xy(fx~u q

qp ′− [ b(x) 〜 xq ]

As qxdyd =η or η= dxdy q , substituting into (A) on pp. 53 yields

=η∫ η′⋅ ∞∞−

− d)(fxx 2q)qp(2 constant independent of x

∴ 0qp2 =− ………………………………… (B)

Now, examine the momentum equation as

2

2

yu

yuv

xuu

∂∂

ν=∂∂

+∂∂

For )(fx~u qp η′−

1qp1q

qp1qp x~)x

yq(fxfx)qp(~xu −−

+−−− −′′+′−

∂∂

)("fxy

)(''fx~yu q2pqp η=

∂η∂

η∂∂ −− )('''fx~

yu q3p2

2

η∂∂ −

2

2

yu

yuv

xuu

∂∂

ν=∂∂

+∂∂

)1qp()qp(x −−+− q3px −

After comparisons, one gets q3p1q2p2 −=−− ……………….. (C)

From (B) and(C), therefore, one gets 31p = ,

32q =

55

Or 3

2

31

x~x~b

x~x~uq

qpC

−−

Now, go back to the stream function.

Recall that )xy(gx~)

xy(gx~

32

31

qp−ψ ,

32x

y~η

32x~b )x~

bb( 3

2

0

Take )(g)bx(c 3

1

01 η−=ψ + where

32

0

0

)(c

bx

by

2=η

There are 2 unknowns to be determined ( 1c & 2c ) and the condition that has

to be satisfied is

020

2 bUconstdyuK ∫ === ∞∞− (pp. 53)

The solutions are

)tanh1()bx()(Re4543.0

uu 2

0j

0

31

31

ξ−= −

[ ]ξ−ξ−ξ= −− tanh)tanh1(2)bx()(Re5503.0

uv 2

0j

0

32

31

where ν

= 00j

bURe and 3

2

0

032

)()(Re2752.0

bx

by

j=ξ

Noted that 32

31 )

bx()(Re5503.0

UV

UV

lim0

j00y

−−∞

∞→−==

( Cu ∼ u )

56

Potential Core of Jets

pcx

( )xuc

x

x

31x − (Laminar)

The analysis is valid at x >> pcx

The equations are suggested to work by starting x from the end of the potential core.

57

Laminar Boundary Layer with A Non-Zero Pressure Gradient (steady case)

Assume a similarity solution as 0xdpd≠ , )x(UU =

432 dcba)x(U)y,x(u

η+η+η+η+α= where δ

=ηy , )1

y(

η∂∂

δ=

∂∂

Similar to the derivation of von Karman integral equation, the profile is an approximation.

Matching conditions:

○1 E

A 0u 0y ==

A○2 E

A Uu y =δ=

A○3 E

A boundary layer equation along wall (y = 0)

0y2

2

yu

xdUdU

yuv

xuu =∂

∂ν+=

∂∂

+∂∂

or xdUdU

yu

0y2

2

−=∂∂

ν =

A○4 E

A 0yu

y =∂∂

δ=

A○5 E

A 0yu

y2

2

=∂∂

δ= (smooth matching)

A○1 E

A → 0Uu

0==η , 0=α

A○2 E

A → 1Uu

1==η , 1dcba =+++

A○3 E

A → xdUdU

ddU

02U

u2

2 ν−=

ηδ =η

58

xdUdU)d12c6b2(U

02

2 ν−=η+η+

δ =η

Λ−=νδ

−=21

xdUd

21b

2

, ν

δ=Λ xd

Ud2

normalized velocity gradient

A○4 E

A → 0dd

1U

u=

η =η , 0d4c3b2a =+++

A○5 E

A → 0dd

12U

u2

=η =η , 0d12c6b2 =++

Finally , 0=α Λ+=612a Λ−=

21b Λ+−=

212c Λ−=

611d

And 432 )611()

212(

21)

612(

)x(U)y,x(u

ηΛ−+ηΛ+−+ηΛ−ηΛ+=

)33(6

)22( 43243 η−η+η−ηΛ

+η+η−η= )(G)(F ηΛ+η=

A○AE

A: Approximated profile without pressure gradient

A○B E

A: Due to non-zero pressure gradient 3)1(6

η−ηΛ

=

Favorable pressure gradient ( 0xdpd< , 0

xdUd> ) , 0>Λ

Adverse pressure gradient ( 0xdpd> , 0

xdUd< ) , 0<Λ

Zero pressure gradient 0=Λ

○A ○B

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

η

F

50G

u/VS

-0.2 0 0.2 0.4 0.6 0.8 1 1.20

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

η

-12

-30

300.7052-6 0

12

59

At the point of separation, 0yu

0y =∂∂

= ( valid for 2-D cases only)

For 0612a

d)U/u(d

0 =Λ+==η =η , then 12

xdUd2

−=νδ

∴ It happens when 0xdUd< (or 0

xdpd> , adverse pressure gradient)

Displacement thickness:

∫ η−δ=δ 101 d)

Uu1( )

120103( Λ−δ=

Momentum thickness:

η∫ −δ=δ d)Uu1(

Uu1

02 )144155

37(63

2Λ−

Λ−

δ=

Wall shear stress:

0y0 yu

=∂∂

µ=τ 0d)U/u(dU

=ηηδµ

= )6

2(U Λ+

δµ

= or 6

2U

0 Λ+=

µδτ

Define Λδ

δ=

ν′δ

= 22

2 )(Uk 2 = ΛΛ

−Λ

− 22

2 )144155

37()631( (still dimensionless)

Shape factor:

)k(f)

14415537(

)12010

3(6312

2

1 =Λ

−Λ

Λ−

=δδ

Also, δδ

µδτ

=µδτ 2020

UU)

14415537()

631()

62(

2Λ−

Λ−

Λ+= )k(f2=

Let νδ

=2

2Z (dimensional) , xd

d2xdZd 22 δ

νδ

=

60

then UZUk2

2 ′=ν

′δ=

Substitution into von Karman integral equation (steady form) as (see pp. 34)

ρτ

=δ+δ+δ∂∂ 0

122

1 xdUdUU

dxdU

t

ρτ

=′δ+δ+δ′ 012

22 UU'UUU2

U2

νδ

× UU

U'UU2 20202122

22

µδτ

=νρδτ

=νδδ′

+δδν

+νδ′

A○1 E

A A○2 E

A A○3 E

For A○2 E

A xdZdU

21

= and A○1 E

A+ A○3 E

A )2(k2

1

δδ

+=

⇒ [ ] )k(f)k(f2kxdZdU

21

21 =++

Let [ ])k(f2k2)k(f2)k(F 12 +−= , then U

)k(FxdZd=

Karman-Polhausen Method →Predict ( )x2δ growth under action of pressure gradient by extrapolation.

Based on ( )UkF

xdZd= , ( ) x

UkFZZ

xxxx

∆+=∆+

Typically, the calculation starts from the stagnation point where ( ) 0y,xU =

However, problem occurs since locally ∞→==

0

)k(FU

)k(FxdZd

Thus, )k(F is forced to be zero to assure not being blown out.

That is 00

U)k(Flim

xdZdlim

0x0x→=

→→

61

The limit can be achieved by the application of the I’Hopital’s rule.

Auxiliary function for the approximate calculation of laminar boundary layers after Holstein and Bohlen (1940)

Λ k ( )kF ( ) 122

11 Hkf =

δδ

= ( )U

kf 022 µ

τδ=

15 0.0884 -0.00658 2.279 0.346

14 0.0928 -0.0885 2.262 0.351

13 0.0941 -0.00914 2.253 0.354

12 0.0948 -0.00948 2.25 0.356

11 0.0941 -0.00912 2.253 0.355

10 0.0919 -0.008 2.26 0.351

9 0.0882 -0.00608 2.273 0.347

8 0.0831 -0.00338 2.289 0.34

7.8 0.0819 -0.00271 2.293 0.338

7.6 0.0807 -0.00203 2.297 0.337

7.4 0.0794 -0.00132 2.301 0.335

7.2 0.0781 -0.00051 2.305 0.333

7.052 0.077 0 2.308 0.332

7 0.0767 0.0021 2.309 0.331

6.8 0.0752 0.0102 2.314 0.33

6.6 0.0737 0.0186 2.318 0.328

6.4 0.0721 0.0274 2.323 0.326

6.2 0.0706 0.0363 2.328 0.324

6 0.0689 0.0459 2.333 0.321

5 0.0599 0.0979 2.361 0.31

4 0.0497 0.1579 2.392 0.297

3 0.0385 0.2255 2.427 0.283

2 0.0264 0.3004 2.466 0.268

1 0.0135 0.382 2.508 0.252

0 0 0.4698 2.554 0.235

-1 -0.014 0.5633 2.604 0.217

-2 -0.0284 0.6609 2.647 0.199

-3 -0.0429 0.764 2.716 0.179

-4 -0.0575 0.8698 2.779 0.16

-5 -0.072 0.978 2.847 0.14

-6 -0.0862 1.0877 2.921 0.12

-7 -0.0999 1.1981 2.999 0.1

-8 -0.113 1.308 3.085 0.079

-9 -0.1254 1.4167 3.176 0.059

-10 -0.1369 1.529 3.276 0.039

-11 -0.1474 1.6257 3.383 0.019

-12 -0.1567 1.7241 3.5 0

-13 -0.1648 1.8169 3.627 -0.019

-14 -0.1715 1.9033 3.765 -0.037

-15 -0.1767 1.982 3.916 -0.054

Example:. ( )

( )0

1x2lim

xxd

d

xxd

d

limxxlim

0x

2

0x

2

0x===

→→→

* Note:

At the stagnation point,

052.7=Λ

077.0k =

( ) 0kF =

> favorable,0

xdUd

62

At a location with known velocity profile (to get 2δ ) and free-stream velocity distribution along wall ( ))x('U ,)x(U , the procedure is as follows:

Firstly, start from the stagnation point ( )0x =

0U = 0)k(F = , 077.0k =

( )UkFlim

xdZd

0x0x

→=

= where 'UZk =

I’Hopital Rule dxdUdxdF

0xlim→

⇒( )

+=

→ UkF

'U''UK

dkdFlim 20x

( )20x0x 'U''U

kdFdklim

kdFd1

UkFlim

→→=

( )0x

2077.0kdK

dFdKdF

0x0x 'U

''U1k

UkFlim

xdZd

==→

=

==

0x2'U''U0652.0

=

−= at stagnation point

Procedure:

(1) Grid generation (determine ∆x)

(2) Must know U , 'U as functions of x, also ''U and 2δ at 0x =

(3) At next location ( )x0x ∆+= , x'U''U0652.0ZZ

020x0

∆⋅−=∆+

then for ( )kF'UZk →= ( ) xUkFZZ

xxxx

∆⋅+=∆+

( )kf1→

Separation would occur at 12−=Λ or 1567.0k −= ( 00 =τ )

( )kf2

x

( )'U'Z''ZUdkdF

dxdk

dkdF

dxdF

+==

+=

U)k(F''U

'Uk

dkdF

2

* At the stagnation point, 'U

077.0Z 2 =νδ

=

20 'U

''U0652.0xdZd

−=

63

Improved Method: (Dimensionless form of Karman-Polhausen method)

Let 0U and L be the velocity and length scales

Lx

=ξ ( ) ( )ξ== wU

xUw0

and the normalized “Z” as p

22 R

Lh

δ

= where ν

=LUR 0

p

Now, one has ξν

δ=

νδ

=d

wdL

UxdUdk 0

22

22

ξ

δ

ν=

dwd

LLU 2

20 wh =

By substitution, ( )U

whFxdZd =

Then ( ) ( )wUwhF

dLd

0

22

νδ

or w

)wh(FL

LUdd 2

20 =

δ

νξ

Thus w

)k(Fw

)wh(Fd

hd⇒=

ξ

* For a stagnation-point flow

y

x

xu α=yv α−=

const=α(potential flow)

'U077.0Z

22

0x=

νδ

==

→ calculate 2δ , ν

At x = 0 (stagnation point) 0'U''U0652.0

xdZd

0

== 0xdxx

ZZZ ==+

⇒ Nothing would change with x (δ , H, 0τ , U)

→ Boundary layer remains the same shape throughout.

At the wall ( )0y = :

( ) xxU α= ( )0v =

α='U

0''U =

64

Homework:

Use Karman-Polhausen method to calculate the growth of 2δ on a 2-D cylinder in a uniform velocity field. Find out the location )(θ of separation.

0U

x

( )

=

RxsinU2xU 0

Also, plot

H

θ=Rx

θ=R

x

Zf

2f

65

66

Turbulent Boundary Layer

0xu

i

i =∂∂

jj

i2

ij

ij

i

xxu

xp1

xuu

tu

∂∂∂

ν+∂∂

ρ−=

∂∂

+∂∂

iu

Time

'ui

Time Average:

'dt)'t(uT1u

2/Tt

2/Tt ii ∫+

−=

where T is a sufficiently long period.

* Note: ii uu = , 0'ui = (white noise)

Reynold’s Averaging Conditions: ( 'fff += , 'ggg += )

(1) gfgf +=+ (f and g are stationary process.)

'gg'ff)'gg()'ff( +++=+++

(2) fafa = where “a” is a constant independent of t .

(3) sf

sf

∂∂

=∂∂ where s can be a function of x, y, z and t .

( )sf'f

ssf'f

sf

s'ff

ssf

∂∂

=∂∂

+∂∂

=∂∂

+∂∂

=+∂∂

=∂∂

(4) gfgf =

O(x, y, z)

)t('uuu iii +=

)t('ppp +=

ensemble average fluctuation

67

* Note: )'gg()'ff(gf ++= 'g'fg'f'gfgf +++= 'g'fg'f'gfgf +++=

'g'fgf +=

Continuity:

0)'uu(x ii

i

=+∂∂ or 0

x'u

xu

i

i

i

i =∂∂

+∂∂

Taking a time average gives

i

i

i

i

i

i

i

i

x'u

xu

x'u

xu

∂∂

+∂∂

=∂∂

+∂∂ ⇒ 0

xu

i

i =∂∂ …..…………. (1)

Momentum:

jj

i2

ij

ij

i

xxu

xp1

xuu

tu

∂∂∂

ν+∂∂

ρ−=

∂∂

+∂∂

* Note: j

ji

j

ji

j

ij x

uu

xuu

xuu

∂−

∂=

∂∂

Taking a time average gives

jj

i2

ij

jii

xxu

xp1

xuu

tu

∂∂∂

ν+∂∂

ρ−=

∂∂

+∂∂

Since 'u'uuuuu jijiji +=

'u'uxx

uuxu

uxuu

jijj

ij

j

ji

j

ji

∂∂

+∂∂

+∂∂

=∂∂

Assume µ is a constant independent of jx Reynold’s Averaged equation

ρ−

∂∂

µ∂∂

ρ+

∂∂

ρ−=

∂∂

+∂∂ 'u'u

xu

x1

xp1

xuu

tu

jij

i

jij

ij

i

1∆ 2∆

1∆ :laminar shear stress ; 2∆ :turbulent shear stress (Reynold’s stress)

(continuity)

68

O(x, y, z)Lam.Trans.

Turb.

500U0 ≅νδ (transition)

* Mostly, 22 'v~'u~'v'u

○1 E

A: If laminar stress term is important,

LRe1~

A○2 E

A If turbulent stress term is important,

L2

2

Re1~

U'u

Generally, for a 2-D turbulent boundary layer

0yv

xu

=∂∂

+∂∂

y1

xdpd1

yuv

xuu p

∂τ∂

ρ+

ρ−=

∂∂

+∂∂ where 'v'u

yu

ρ−∂∂

µ=τ

* It can be shown that 0yp=

∂∂ (same as before).

For 2-D steady (stationary) flows

= 0

xdpd

0yv

xu

=∂∂

+∂∂

0VLU

+ UL

~V δ∴ ( )UV <<

'v'uy

'uxy

uxu

xp1

yuv

xuu 2

2

2

2

2

∂∂

−∂∂

∂∂

+∂∂

ν+∂∂

ρ−=

∂∂

+∂∂

LU2

δ

δ UU

L

LU2

22

ULU

δν<<ν

δ<<

22 'uL'u

(2U

L× ) 1 1 1

2LULUL

δ

ν

<<ν

δ<< 2

2

2

2

U'uL

U'u

69

Two-Dimensional Turbulent Jet → ambient fluid is still

Momentum: All variables denote averaged quantities.

ρ1

yvu

xu 2

∂∂

=∂∂

+∂∂ ……………………………. A○1 E

A

where 'v'uyu

ρ−∂∂

µ=τ

Continuity:

0yv

xu

=∂∂

+∂∂ ……………………………………. A○2 E

For turbulent jets, integrate A○1 E

A with respect to y from 0 to ∞ gives

∂τ∂

ρ=

∂∂

+∂∂∞

0

2

y1

yvu

xu

or ∞∞∞ τρ

=+∫ 0002 1vudyu

xdd

0y = , 0)0(v = , 0)0( =τ ( 0'v'u → , 0yu

0

=∂∂ due to symmetry )

∞=y , 0)(u =∞ , 0)( =∞τ ( 0'v'u → in free stream, 0yu

=∂∂

)

Therefore, 0dyuxd

d0

2 =∫∞

or kdyu0

2 =∫∞

a constant independent of x (see also pp. 53)

Integration of A○2 E

A leads to 0vdyuxd

d00 =+∫∞∞

( )0Ve −− (due to symmetry)

Thus ∫= ∞0e dyu

xddV (see also pp. 53 for a laminar jet)

negative (upwards)

70

Experimentally, similarity is satisfied well in the turbulent cases.

2uc

cub

( )( ) ( )

=

xbyf

xuy,xu

c

Experimental results show that both Ve and u c decrease with respect to x . Accordingly, guess the entrainment coefficient (dilution factor) as

'uV

c

e α= (about 0.07 experimentally) or ce u'V α=

Then c0u'dyu

xdd

α=∫∞

………….. (∆)

Recall that kdyu0

2 =∫∞

(pp. 69)

By taking )x(b

y=η , η= dbdy

Substitution gives kd)(f)x(b)x(u0

22c =ηη∫

Then =)x(b)x(u2c constant independent of x

Also, from (∆) c0c u'αηd)η(fbuxd

d=∫

Then [ ] =)x(b)x(uxd

d)x(u

1c

c

constant independent of x

=bu2c constant

( ) =buxd

du1

cc

constant

71

Assume αx~uc , βx~b

=β+α+−α−=β+α

0102

or 01=−β

One gets 21x~uc

1x~b

Experimental results for plane turbulent jests are as follows:

21

00

c

bx5.3

Uu

= (3.2 ~ 3.6)

00 bx10.0

bb

= (0.09 ~ 0.115) * x starts from pcxx =

053.0UV

0

e = (0.05 ~ 0.07)

Recall that for laminar plane jets (pp. 55),

31x~uc

− , 32x~b

72

73

Turbulent Plane Channel Flows (Fully-Developed)

x

Hu

2H

c

When the flow is fully developed, )y(u , 0τ and other variables (except p)

are constant (independent of x ).

Choose a velocity scale as *0 u=ρτ (friction/shear velocity)

Related length scales are δ , ks and *uν

ν≅δ

*L u

6.11

thickness of viscous sublayer

If 1ks

L >>δ , the turbulent boundary layer roughness elements are lubricated

(smooth turbulent boundary layer).

Then the possible length scales leave δ and *uν .

It is noted that ks*s

L

R6.11

ku6.11

k=

ν≅

δ where ν

= s*k

kuR .

For smooth turbulent boundary layers, 1R k <<

In practice, =kR 3 70

smooth transitional rough

k s (roughness height)

74

Define a cross-sectional average velocity (bulk velocity) as

∫δ

δ=

0dyu1U

(i) Plane (2-D) channel flows ( )H=δ

500U>

νδ turbulent

(ii) Circular pipe flows

==δ

2DR

2000DU>

ν

or 1000U>

νδ turbulent

Hydraulic Radius: Hr = area wettedofperimeter

area

(1) Open channel (plane pipe) flows (2-D)

δ=⋅δ

=2

12rH

(2) Circular pipe flows

22r

2

=δπδπ

=

⇒ Pipe flows 1000rU H >ν

turbulent

2H

75

Friction Reynolds number: (for a 2-D turbulent boundary layer)

6.11sublayer viscousof thicknesscknesslayer thiboundary u

u/R *

** ×≅

νδ

=νδ

=

δ

Core region (outer region)ν not important

wall regionν important

viscous sublayer

In the core region: (ν is relatively not important.)

Flow is almost independent of ν (or *uν ). *u and δ are important.

In the wall region:

Flow is affected by ν ; *u and *uν are important.

(A) In the core region, use “vorticity” to represent the flow (2-D)

( )y , ,ufunctionydud

yu

xv

*z δ=−≈∂∂

−∂∂

Due to dimensional homogeneity, by Rayleigh’s method one has

δ

=δ yg

ydud

u*

Let δ

=ηy (outer variable), one has

)(gd

)u/u(d * η=η

<< (for a boundary layer)

76

Integration yields 'd)'(guu

*

ηη= ∫η

( 'η is a dummy variable),

)η(F~

which leads to the Outer Law for turbulent boundary layers as

)(Fu

uu*

0 η=− [ At y = δ, u = u0 ]

)(F η is a universal function and is dependent on domain geometry and pressure gradient. (See pp. 81 for experimental results.)

(B) In the wall region,

= y,

uν ,ufun

ydud

**

Dimensional analysis yields let ν

=+ *

*

uyu/

yy (inner variable)

ν

**

*

u/yf~

ydud

uu/ or )y(f~

yd)u/u(d * +

+ =

Integration yields )y(fu

)y(u

*

++

= [ At y (or y+) = 0, u = 0 ]

which is the Inner (Wall) Law for turbulent boundary layers.

(C) In the viscous sublayer, ( )6.11y <+

δy is so small that local shear stress ≈ wall shear stress = constant

Thus, for a very small y

ρτ

=ν 0

ydud = 2

*u ( *u is a function of x only.)

Integration gives cyuu2* +ν

=

(Turbulent stress is nearly zero.)

77

B.C.: @ 0y = , 0u = (no-slip)

Then yuu2*

ν= or

νuy

uu *

*

= or ++ = yu .

Summary:

Outer Law: )(Fu

uu

*

0 η=− where

δ=η

y

Inner Law: )y(fu ++ = where *u

uu =+ and ν

=+ *uyy

Viscous sublayer: ++ = yu

78

2-D Turbulent Channel Flows

y=2h

y=0

y=h

yx

dxdp0−⇒ (const, developed)

Assume nothing varies with x except for po

≠= 0const

xd)x(pd 0

Continuity: (u, v and p denote time-averaged quantities)

0yv

xu

=∂∂

+∂∂

Because 0xu=

∂∂ thus 0

yv=

∂∂ (v is independent of y)

Since at 0y = , 0v = therefore 0v ≡ (everywhere)

x-momentum:

2

22

2

2

xu'u

xyu'v'u

yxp1

yuv

xuu

∂∂

ν+∂∂

−∂∂

ν+∂∂

−∂∂

ρ−=

∂∂

+∂∂

y-momentum:

'v'ux

)yv

xv(ν'v

yyp

ρ1

yvv

xvu 2

2

2

22

∂∂

−∂∂

+∂∂

+∂∂

−∂∂

−=∂∂

+∂∂

Integration with respect to y yields )x(F'vp1

2 =+ρ

→ independent of y

Taking derivative w.r.t. x yields )x(Fxp1

2=∂∂

ρ → independent of y and x

=∂∂

ρ⇒

xp1 constant

xdpd1 0

ρ=

(fully- developed)

( = constant; fully-developed )

0 0 fu l ly-developed

0 0 0 0 fu l ly-deve loped

79

Integrating the x-momentum equation with respect to y gives

cydud'v'uy

xdpd10 0 +ν+−

ρ−= ……………………. (1)

@ at 0y = , 0'v'u = (lower wall) then 2*

0

0

uydudc −=

ρτ

−=ν−=

@ at h2y = (upper wall) 0'v'u = and 2*

0

h2

uydud

−=ρτ

−=ν

Thus, 2*

2*

0 u)u(0)h2(xdpd10 −−−

ρ−=

then 2*

0 uhxdpd1

− , or hu

xdpd1 2

*0 =ρ

Substituting into (1) yields

2*

2* u

ydud'v'u

hyu0 −ν+−=

or

−=ν+−

hy1u

ydud'v'u 2

* …………………..(#)

Normalized Form I ⇒ for outer (core) region, νδ

= **

uR

η1ηd

)u/u(dR1

u'v'u *

*2*

−=+− , where hy

=η , h=δ

Normalized Form II ⇒ for inner (wall) region

*2* R

y1ydud

u'v'u +

+

+

−=+−

80

⇒ In the limit →νδ

= **

uR large, where )hy(=η is finite (but →+y large)

Form I becomes (outer region)

η−=− 1u

'v'u2*

(valid in outer region)

⇒ In the limit →*R large, [ ]1Oy ≡+

Form II becomes (inner region)

1ydud

u'v'u

2*

=+− +

+

(valid in inner region)

For a smooth turbulent boundary layer ( )sL kδ >> ,

Inner layer: (wall region)

Since )y(fuu

*

+= (see pp. 77)

then )y(gu

'v'u2*

+=−

Note: η=ν

=+*

** Rhyhuyuy ++

+

+ ν=

ν==

ydfduu

ydfdu

ydyd

ydfdu

ydud 2

****

Outer layer: (core region)

)(Fu

uu

*

0 η=− (see pp. 77)→ Velocity Defect Law or Friction Law

Note: yd

dd

Fduydud

η=

η=

dFd

hu* F is still unknown.

Experimentally, η=η ln5.2)(F in the core region.

Law of the Wall

81

Universal Velocity Distribution Laws for Smooth and Rough Pipes (δ = R: radius)

(Schlichting’s book, pp. 607)

Ryln5.2

uuu

*

0 −=−

]Ry1)

Ry11(ln[1

uuu

*

0 −+−−κ

−=−

2/3

*

0 )Ry(08.5

uuu=

(The case for a circular pipe is similar to the 2-flat-plate case; R → h or δ)

In the buffer zone or inertia sublayer (the transition part between the outer and inner regions), matching the vorticity (- du/dy) as (see pp. 80)

∞→+y (inner)

0→η (outer)

one has ηη

=ν +

+

d)(Fd

hu

yd)y(fdu *

2*

η=

dFd

yu

hy * where

hy

Then κ1

ηd)η(Fdη

yd)y(fdy ⇒=+

+

+ where κ is Karman constant.

)y(fun + )(fun η

aln1)(F +ηκ

=η⇒

cylnκ1)y(f += ++

Ry

*

0

uuu −

valid in the inertia sublayer

0.4

82

Noted that

Outer: η−=− 1u

'v'u2*

As 0→η 2*u'v'u ≅−⇒

Inner: 1dy

)y(dfu

'v'u2*

=+− +

+

As ∞→+ y 0y1

yd)y(fd →

κ=⇒ ++

+

Therefore, in the inertia sublayer, 2*u'v'u ≅−

The profile:(for a pipe flow or flow between two flat plates)

cuy

ln1uu *

*

+νκ

=

which is called the logarithmic law (or log law).

Also, since aln1u

uu)(F*

0 +ηκ

=−

=η …...……...…………….

cyln1uu)y(f

*

== ++ .……....…….………….

As both relationship are valid within the inertial sublayer, - yields

( )acyln1uu

*

0 −+ηκ

=+

( )acRln1* −+

κ= *

*y* Ru/uyy

=νδ

=η δ

+

(κ, c) = (0.4 , 5.0)

(0.41, 5.5) more popular

83

Coles Law of the Wake

Deviation of the velocity profile/maximum deviation at the boundary layer edge:

δ

=−δ−−−

++

++ yW

21

5.5ln5.2u5.5yln5.2u

e

Note: at δ=y , 2W =

+ =

Experiential result shows that

δπ

=

δ

y2

sin2yW 2 in the outer region

Then, the composite law is

δκ

π++

κ= ++ yWcyln1u

where 55.0=π for flat plate case

)5.0(8.0 +β≈ , where β is a pressure parameter

Law of the Wake

y/R

pipe center

84

For a flow in a circular pipe:(stationary, fully-developed)

R

y

R y

x

Momentum: (in an index form; averaged form)

'u'uxxx

uxp1

xu

u jijjj

i2

ij

ij ∂

∂−

∂∂∂

ν+∂∂

ρ−=

∂∂

For )y(uu = , 0wv ≡= (everywhere)

0yuv

xuu ⇒

∂∂

+∂∂

x-momentum: xdpd1 0

ρ−

'u'ux

uxp10 ji

ji

2

∂∂

−∇ν+∂∂

ρ−= (see also pp. 78)

Converting into the cylindrical coordinate system:

Recall that

∂∂

∂∂

=∂∂

+∂∂

=∇rur

rr1

ru

r1

ru)r(u 2

22

θ∂∂

+∂∂

+∂∂

=•∇ θAr1

xA

rA

r1A xr

where 'u'uA xrr = , 'u'uA xxx = , 'u'uA xθθ =

One gets

( )'u'urrd

dr1

rdudr

rdd

r1

xd)x(pd1

xro −

ν=

ρ

symmetry

85

Multiplying by r and integrating leads to

*xr

02

c'u'urrdudr

xd)x(pd1

2r

+−ν=ρ

(at centerline r = 0, c* = 0)

or 'u'urdud

xdpd1

2r

xr0 −ν=

ρ

At Rr = (wall), 0'u'u xr =

then one has Rr

0

rdud

xdpd1

2R

=

ν=ρ 0yyd

ud

=

ν−= 2*

0 u−=ρτ

−=

Or R2u

xdpd1 2

*0 −=

ρ

R

p

C.V.x∆

p xdxdp0 ∆+

Then 'u'urdud

Rru xr

2* −ν=−

Changing variables: rRy −=

v'u'u yr −=−= 'u'ux =

'v'uydud

RyRu2

* +ν−=−

−⇒

or

−=ν+−

Ry1u

ydud'v'u 2

* (compare to (#) on pp. 79)

dydr −=

For a fully developed flow,

( )2o0 Rx

xdpdx)R2( π

∆−=∆πτ

2*uρ

then R2u

xdpd1 2

*o −=

ρ

2 R

τ0

86

Eddy Viscosity near the Wall

(I) Inertia sublayer (Logarithmic region)

yu

ydudu'v'u *

tt2* κ

ν=ν==− (based on log-law; see pp. 82)

yu*t κ=ν∴

From Prandtl’s mixing-length concept

ydud2

t =ν yuy

u*

*2 κ=κ

= yκ=⇒

How does 'v'u− behave in the log-region (inertia sublayer)?

From 0y'v

x'u

=∂∂

+∂∂ dy

x'u'v

y

0∫ ∂∂

−=

For a small y, expand 'u in power series as

+++= 2210 y)t,x(fy)t,x(f)t,x(f'u (note: 0y = , 0'u = , 0f0 = )

then dy)yxfy

xf('v

y

0

221∫ +∂∂

+∂∂

−= −∂∂

−∂∂

−=xfy

31

xfy

21 2312

Thus

∂∂

−∂∂

−++= xfy

31

xfy

21)yfyf('v'u 23122

21

)y(Oxffy

21 41

13 +

∂∂

−=

Note: 0f21

xxff 2

11

1 ⇒∂∂

=∂∂ ⇒ Fully developed, average is independent of x.

4y~'v'u⇒ in the logarithmic region

87

(II) In the viscous sublayer ( ++ = yu )

1)/uy(d

)u/u(du

'v'u*

*2*

+− (see pp. 82)

As 0y → (at wall) , 0'v'u →

One has 1)/uy(d

)u/u(d*

* =ν

For ++ = yu or ν

= *

*

uyuu , one has yuu

2*

ν=

Form Prandtl’s mixing-length concept,

22*2

22 u

ydud'v'u

ν

=

=−

one has 42 y~ or 2y~

88

Van Driest Correction

From Van Driest “On the turbulent flow near a wall”, Journal of Aeronautical Science, 23, 1007(1956)

For an equilibrium turbulent boundary layer

−κ=

+− A

y

e1y

where A = 26 (originally, 25) to give 1.5yln1uu

*

= +

Define ν

=+ *u

+y

+l

Traditional profile for an equilibrium (smooth) turbulent boundary layer:

+yln

+u

11.6

+

89

Power Law (An approximation valid in most regions of smooth turbulent boundary layers)

10u

u

δy

Experimentally, 7

1

*

*

uy74.8uu

ν= for 510Re ≈

Since at δ=y , 7

1

*

*

0 u74.8uu

νδ

= ……………………(*)

Thus 7

1

yuu

0

δ

= (power-law)

Also, define a (local) friction coefficient (normalized wall shear stress) as 2

0

*202

1

2*

202

1w

f uu2

uu

uC

=

ρρ

=ρτ

=

From (*), 7

2

0

0

*2f

uuu

)74.8(2C

νδ

= 41

0f

u045.0C−

νδ

=⇒

2Cf

n1

yuu

0

δ

= for flat plate

n1

Ry

= for circular pipe ry −δ= , R=δ

where n depends on Reynolds numbers.

90

Turbulent Boundary Layer Development on a Smooth Flat Plate 0u

x x’(virtual origin of f.b.l.)

500u≅

νδ

TransL

Turb

From the von Karman integral equation

ρτ

=δ+δ w0012

20 xd

udu)u(xd

d (see pp. 34)

2*

w220 u

xddu =

ρτ

For 2/u

u2/u

C 20

2*

20

wf ρ

ρ=

ρτ

= , 20

f2* u

2Cu =

one has 20

f220 u

2C

xddu =δ …………………………... (1)

Take the (1/7) power law assumption as 7

1

yuu

0

δ

= (pp. 89)

η

−δ=δ ∫ d

uu1

uu1

000

2 , where δ

=ηy

∫ ηη−ηδ=1

0d)1( 7

17

1δ=δ

−= 0972.0

97

87

For 4

1

0f

u045.0C−

νδ

= (pp. 89), substituting into (1) gives

41

0u0225.0xd

d0972.0−

νδ

dxu0972.00225.0d

41

41 0

ν

=δδ Cxu0972.00225.0

54 4

1

45 0 +

ν

=δ−

51

xu371.0x

0−

ν=

δ⇒

At virtual origin (x = 0), δ = 0,

thus C = 0

Turbulent boundary layer : 54x~δ

Laminar boundary layer : 21x~δ

( 0xdpd= )

500uo ≈νδ

(virtual origin of turbulent boundary layer)

91

Turbulent drag on the plate (2 sides)

∫ τ=L

0 w dxb2D

∫ρ=L

0 f20 dxC)u

21(b2

dxu045.0ub L0

020

41

νδ

ρ=−

(pp. 89)

dxx

xu0.045ub L0

020

41

41

δ

νρ=

−−

4/15/10 xu0371.0

−−

ν (pp. 90)

where

( )20

1

41

41

xu0371.0xu045.0C 0of

ν

ν= −

− 51

xu0577.0 0−

ν=

* For large L, Total Drag ≒ Turbulent Drag

5/4020

L

002

0 L)45(u)0577.0(ubdxxu)0577.0(ubD

51

515

1 −−

ν

ρ=

ν

ρ= ∫

Global friction coefficient: (normalized drag)

( )5

1

Lu0721.0Lbu

DC 020

gf

ν=

ρ= with

ν=

LuRe 0L

which is valid for 7L

5 101Re105 ×<<× . (see also pp. 26 )

b L

92

x2

1x

54x

Blasius Transition Fully Developed Turbulent

510U•

νδ∞

Growing wave

rough transition smooth

( ) 51 102.3xU×=

ν∞

( ) 500U2 =νδ∞

Blasius layer (laminar):

21

x~δ 21

xRe5x

−≅δ where

ν= ∞ xURex

21

x1 Re72.1

x−=

δ 21

x2 Re664.0

x−=

δ 6.2H2

1 ≅δδ

=

Turbulent boundary layer: ( for 66x 1018~ 101.7 Re ××= )

71 1315.0

10 δy737.0

uu

= from Nikuradge

This (power) law is valid in the log-region. However, it is not valid to predict the shear stress at wall (tends to ∞).

81δ

≅δ δ≅δ727

2 30.1H ≅

139.0xf Re02296.0C −=

93

Example: Air-steam ( 4108.1 −×=ν ft2/s) at 50 ft/s passes a smooth flat

plate. It is assumed that as 50 102.3xu×=

ν the boundary layer changes

suddenly from laminar to turbulent. Calculate δ at x = 5 ft .

【Solution】

Formulae: TBL:5

xRe371.0

x=

δ ν

xuRe 0x = ( 0u = 50 ft/s)

LBL:xRe

0.5x=

δ Transition occurs at 5x 102.3Re ×=

(A) Check the nature of boundary layer at x = 5 ft

( ) 564x 102.31039.1

108.1550Re ×>×=

×= − (turbulent)

(B) Find the location of transition

For 4tr5

108.1x50102.3 −×

=× 152.1x tr = ft

(C) Calculate boundary layer thickness (laminar) at xtr

For xRe

0.5x=

δ ( )( )( ) 01018.0102.30.5152.1 215 =×=δ

− ft

(D) Find the virtual origin of turbulent boundary layer

5 xu0

371.0x

ν

5108.1

50 x371.0

x01018.0

4−×

=

371.001018.0

108.150x 5

45/4

−×= , x = 0.256 ft

(E) Calculate turbulent boundary layer thickness at x = 5 ft

104.4256.0)152.15(x =+−= ft,

64x 1014.1

108.1)104.4(50Re ×=

×= −

in123.1ft0936.0)1014.1)(371.0)(104.4( 516 ==×=δ −

94

Power Law of Rough Flows

The profile 5.8kyln1

uu

s*

+

κ

=

κ

=sk

y30ln1 (see also pp. 82)

can be approximated as

61

s* ky34.9

uu

= for 2006 <<

skδ

Power Friction Law:

dyky34.91uV

61

0s

* ∫

δ

= δ where V:average of u within δ

61

s* k8

uV

δ= for 200

k6

s

< (error is within 3%)......()

Rough Turbulent Boundary Layer:

In case that 70kuu/

k s*

*

s >ν

, define sk

yy =+ and δ

=ηy

For 1ks

>>δ , the log-law is wake correction

δκ

π++

κ=

yW5.8kyln1

uu

s*

…….……… (1)

For channel flows, π is small ( )3.0< and

δ

π=

δ

y2

sin2yW 2

(see pp. 83)

@ as δ=y )2(5.8k

ln1uu

s*

0

κπ

++δ

κ= ….…..…….. (2)

(2)-(1) leads to

δ

−κπ

+δκ

−=− yW2yln1

uuu

*

0

95

Recall the Darcy-Weisbach Equation (valid for fully developed pipe flows)

p

L

p p∆+(-)

cross sectional area A

π 2D

4

Average vel=V

or

ρ=∆ 2V

2DLfp

Force balance yields ( )LDD4

p w2 π⋅τ=

π⋅∆

Therefore, 22

4w V

2DLf

LDD ρ

π=τ

π

2*

2 uV8f

ρ⇒ρ=

One gets the Darcy friction coefficient 2

*

Vu8f

=

Since ∫δ

= δ0 dyu1V ∫

δ≅ δ

sk dyu1 ( )sk>>δ

∫δ

= δ

sk**

dyuu1

uV

η∫

+

δδκ

d5.8k

yln11

ssk (neglect wake correction term)

η

+

δκ

+ηκ

= ∫δ

d5.8k

ln1

ln11

ssk

1

0s

5.8k

lnln

η+

δκη

+

κη−ηη

( ) 5.8k

ln1101

s

κ+

κ−⋅

= 6k

ln1s

κ=

δκ

=sk

11ln1or

DL

g2Vfph

2

L =g∆

=

96

< Stanton-Moody Diagram >

70νku s* ≅

ReD = V D / ν

Fully rough

Laminar

f

ks/D

64/ReD

97

Fully Developed Boundary Layer Flows in Open Channel

δ

y

1S

0xp

=

∂∂

At the free surface ( )δ=y , since 0yu→

∂∂ , 0→τ

δ−δρ=τ∴

y1Sg

At the bottom ( )0y = , since 2*w uSg ρ=δρ=τ

δ= Sgu*

61

skV

81

δ= (from on pp. 94)_

Then *s

uk

8V6

1

δ= ( V is the average velocity )

δ

δ= Sg

k8

61

s

61

32

21

skSg8 δ

=

21

32 S

n1δ= (from Manning’s formula)

The Manning’s coefficient 61

skg8

1n = 61

sk04.0≅ (in SI unit system)

The momentum equation is (fully developed)

Sgy

1xp1

yuv

xuu +

∂τ∂

ρ+

∂∂

ρ−=

∂∂

+∂∂

Or 0Sgy

1=+

∂τ∂

ρ

Integration gives ySgc ρ−=τ

98

General Turbulent Boundary Layer (2-D)

The law of the wall (for an equilibrium smooth turbulent boundary layer) is

cuyln1uyFunuu **

*

+νκ

=

ν

=

where ( )ρ

τ=

)x(xu w*

Noted that before separation, 0w ≥τ

after separation, 0w <τ . u* becomes meaningless.

Velocity defect law is generally in the form

δ

=− x,

)x(yF

u)x(Uu

*

e due to non-equilibrium nature

In order to be consistent with law of the wall

)x(ayln1x,yF +δκ

δ

Noted that )x(a , c, )x(Ue and )x(u* are related by the friction law

cauln1u

)x(U *

*

e +−νδ

κ=

∞→∞→

νδ

*

e*

u)x(U ,u

99

Von Karman Integral Equation (Steady, Turbulent B.L.)

y)/(

yu

xUu

yuv

xuu t

2

2e

e ∂ρτ∂

+∂∂

ν+∂∂

=∂∂

+∂∂ where 'v'ut ρ−=τ

)Uu(yv)Uu(v

yxUU

x)Uu(uLHS ee

ee −∂∂

−−∂∂

+∂∂

+∂−∂

=

)Uu(

xu

e−∂∂

ρτ

∂∂

+∂∂

ν=−∂∂

+−+−∂∂ t

2

2

ee

ee yyu)Uu(v

yxdUd)Uu()Uu(u

x

Integrating with respect to y from 0 to ∞ leads to ( 2*

w0yy u

ρτ

yuν

yuν −=−=

∂∂

−∂∂

=∞=)

0dy)Uu(xd

Uddy)Uu(uxd

d0 e

e0 e +−+− ∫∫

∞∞∞

ρτ

+−=0

t2*u

or 2*1e

e2

2e uU

xdUdU

xdd

=δ+δ

Clauser (1984) introduced

∫∞ −

=∆0

*

e dyu

uU ( )1e* Uu δ=∆

1*

e

uU

δ= where ∆ is comparable to δ

Noted that ∫∞

−0 e dy)Uu(u

∫∫∞∞

−+−=0

2e0 ee dy)Uu(dy)Uu(U

( )∫ −+∆−= ∞0

2e*e dyUuuU

(see pp.33 for laminar cases)

100

Estimate

ν+⋅∆≈−∫

∞ 2e

*

2*0

2e U

uOudy)Uu(

−∞small

2

*

e2* dy

uUuu ∫ −

ν small

0

2e

*

dy)Uu(u

∆ν

+∆= 2*

2e

*

2* u

Uu

O1u

Since 2

*

*

2

*

e

*

cauln1uu

Uu

+−

νδ

κ∆ν

∆ν

( ) 0RlnR1~ *R2

**

→ ∞→

For large *R , ∆−≅−∫∞

*e0 e uUdy)Uu(u

The approximated form is

2*

e*e* u

xdUdu)Uu(

xdd

=∆+∆

Recommended