058:0160 Chapter 7

Professor Fred Stern Fall 2010 1

Chapter 7: Boundary Layer Theory

7.1. Introduction:

Boundary layer flows: External flows around streamlined bodies at

high Re have viscous (shear and no-slip) effects confined close to
the body surfaces and its wake, but are nearly inviscid far from the
body.
Applications of BL theory: aerodynamics (airplanes, rockets,
projectiles), hydrodynamics (ships, submarines, torpedoes),
transportation (automobiles, trucks, cycles), wind engineering
(buildings, bridges, water towers), and ocean engineering (buoys,
breakwaters, cables).

7.2 Flat-Plate Momentum Integral Analysis & Laminar approximate

solution

Consider flow of a viscous fluid at high Re past a flat plate, i.e., flat
plate fixed in a uniform stream of velocity Uiˆ .

Boundary-layer thickness arbitrarily defined by y =  99% (where,  99% is

the value of y at u = 0.99U). Streamlines outside  99% will deflect an
amount  * (the displacement thickness). Thus the streamlines move
outward from y  H at x  0 to y  Y    H   * at x  x1 .
058:0160 Chapter 7
Professor Fred Stern Fall 2010 2

Conservation of mass:
H H  
 V ndA=0= 0 Udy  0
CS
udy
Assuming incompressible flow (constant density), this relation
simplifies to
UH   udy   U  u  U dy  UY   u  U dy
Y Y Y

0 0 0

Note: Y  H   * , we get the definition of displacement thickness:

 *  0Y 1  dy
u
 U
 * ( a function of x only) is an important measure of effect of BL on
external flow. To see this more clearly, consider an alternate derivation
based on an equivalent discharge/flow rate argument:
δ

δ* Lam=/3
δ* Turb=/8

 

 Udy  udy
* 0

Inviscid flow about δ* body

Flowrate between  * and  of inviscid flow=actual flowrate, i.e.,

inviscid flow rate about displacement body = equivalent viscous flow
rate about actual body
 *  
 u
0  0  0    0  U dy
 *
Udy Udy udy 1

w/o BL - displacement effect=actual discharge

For 3D flow, in addition it must also be explicitly required that  * is a
stream surface of the inviscid flow continued from outside of the BL.
058:0160 Chapter 7
Professor Fred Stern Fall 2010 3

Conservation of x-momentum:
 Fx   D 
H Y
  uV  ndA    U Udy     u  udy 
0 0
CS

Y
Drag  D  U 2 H  0 u 2 dy = Fluid force on plate = - Plate
force on CV (fluid)
Y u
Again assuming constant density and using continuity: H  0 U dy
Y
2 Y x
D  U 0 u / Udy   u 2 dy  0 w dx
0
D Y u u 
2
  0 1  dy
U U U  
where,  is the momentum thickness (a function of x only), an
important measure of the drag.
2 1
x
2D
U 2 x x x 0 f
CD    C dx Per unit span

w
Cf 
1
 Cf 
d
xCD   2 d
U 2 dx dx Special case 2D
2 momentum integral
d C f d equation for px = 0
  w  U 2
dx 2 dx
058:0160 Chapter 7
Professor Fred Stern Fall 2010 4

Simple velocity profile approximations:

u  U (2 y /   y 2 /  2 )

u(0) = 0 no slip
u(δ) = U matching with outer flow
uy(δ)=0

Use velocity profile to get Cf() and () and then integrate momentum
integral equation to get (Rex)

δ* = δ/3
θ = 2δ/15
H= δ*/θ= 5/2
 w  2U / 
2U /  d d
 Cf  2  2 (2 / 15);
1 / 2 U 2
dx dx
15dx
d 
U
30dx
2 
U
 / x  5.5 / Re1x/ 2
Re x  Ux /  ;
 * / x  1.83 / Re1x/ 2 10% error, cf. Blasius
 / x  0.73 / Re1x/ 2
C D  1.46 / Re1L/ 2  2C f ( L)
058:0160 Chapter 7
Professor Fred Stern Fall 2010 5

7.3. Boundary layer approximations, equations and comments

U, , y
x

u=v=0

2D NS, =constant, neglect g

ux  vy  0
1 p
ut  uu x  vu y     (u xx  u yy )
 x
1 p
vt  uv x  vv y     (v xx  v yy )
 y

Introduce non-dimensional variables that includes scales such that all

variables are of O(1):
x*  x / L
y
y*  Re
L
t *  tU / L
u*  u /U

v*  Re
U
p  p0
p* 
U 2
Re  UL / 
058:0160 Chapter 7
Professor Fred Stern Fall 2010 6

The NS equations become (drop *)

ux  vy  0
1
ut  uu x  vu y   px  u xx  u yy
Re
1 1 1
(vt  uvx  vv y )   p y  2 vxx  v yy
Re Re Re

For large Re (BL assumptions) the underlined terms drop out and the BL
equations are obtained.

Therefore, y-momentum equation reduces to

py  0
i.e. p  p ( x, t )
 px    (U t  UU x ) From Euler/Bernoulli equation for
external flow
2D BL equations:
u x  v y  0;
ut  uu x  vu y  (U t  UU x )  u yy

Note:

(1) U(x,t), p(x,t) impressed on BL by the external flow.

2
(2) 2
 0 : i.e. longitudinal (or stream-wise) diffusion is
x
neglected.
(3) Due to (2), the equations are parabolic in x. Physically, this
means all downstream influences are lost other than that
contained in external flow. A marching solution is possible.

(4) Boundary conditions

058:0160 Chapter 7
Professor Fred Stern Fall 2010 7

matching
inlet
δ

Solution by
marching
y

x
X0
No slip

No slip: u  x,0, t   v x,0, t   0

Initial condition: u  x, y,0 known
Inlet condition: u  x0 , y, t  given at x0
Matching with outer flow: u  x, , t   U  x, t 

(5) When applying the boundary layer equations one must keep in
mind the restrictions imposed on them due to the basic BL
assumptions
→ not applicable for thick BL or separated flows (although
they can be used to estimate occurrence of separation).

(6) Curvilinear coordinates

058:0160 Chapter 7
Professor Fred Stern Fall 2010 8

Although BL equations have been written in Cartesian

Coordinates, they apply to curved surfaces provided δ << R and
x, y are curvilinear coordinates measured along and normal to
the surface, respectively. In such a system we would find under
the BL assumptions
u 2
py 
R
Assume u is a linear function of y: u  Uy 

dp U 2 y 2

dy R 2
U 2
p( )  p(0) 
3R
Or

p 
 ;
U 2 3R therefore, we require δ << R
058:0160 Chapter 7
Professor Fred Stern Fall 2010 9

(7) Practical use of the BL theory

For a given body geometry:
(a) Inviscid theory gives p(x) → integration gives L,D = 0
(b) BL theory gives → δ*(x), τw(x), θ(x),etc. and predicts
separation if any
(c) If separation present then no further information → must
use inviscid models, BL equation in inverse mode, or NS
equation.
(d) If separation is absent, integration of τw(x) → frictional
resistance body + δ* , inviscid theory gives → p(x), can go
back to (2) for more accurate BL calculation including
viscous – inviscid interaction

(8) Separation and shear stress

1
At the wall, u = v = 0 → u yy  px

1st derivative u gives τw →  w  u y
w

τw = 0 separation

2nd derivative u depends on px

058:0160 Chapter 7
Professor Fred Stern Fall 2010 10

Inflection point

7.4. Laminar Boundary Layer - Similarity solutions (2D, steady,

incompressible): method of reducing PDE to ODE by appropriate
similarity transformation
ux  vy  0
uu x  vu y  UU x  u yy

BCs: u  x,0  v x,0  0

u  x,    U  x 
+ inlet condition
058:0160 Chapter 7
Professor Fred Stern Fall 2010 11

u  x, y   y 
 F   g  x  related to   x 
For Similarity U  x 
 g  x   expect
Or in terms of stream function  : u   y v   x
For similarity   U  x g  x  f     y g x 
u   y  Uf ' v   x  (U x gf  Ug x f  Ug x f ' )
BC:
u  x,0  0  U ( x) f (0)  0  f (0)  0
v x,0  0  U x ( x) g ( x) f (0)  U ( x) g x ( x) f (0)
 U ( x) g x ( x)  0  f (0)  0
 U x ( x) g ( x)  U ( x) g x ( x)  f (0)  0
 f (0)  0
u  x,    U  x   U ( x) f ()  U  x   f ()  1
Write boundary layer equations in terms of 
 y yx   x yy  UU x   yyy
Substitute
 yy  Uf '' g
2
 yyy  Uf '''
g
 xy  U x f '  Uf ''g x / g
Assemble them together:

Uf ' U x f '  Uf '' ggx   U x gf  Ug x f  Ug xf ' Uf '' g 
 
 UU x   U f ''' g 2 
058:0160 Chapter 7
Professor Fred Stern Fall 2010 12

U
UU x f '2  UU x ff ''  U 2 g x g ff ''  UU x   2 f '''
g
U U
UU x f '2  Ug x ff ''  UU x   2 f '''
g g

f  '''


g
Ug x ff  '' g2

U x 1 f '2
 0
C1 C2

Where for similarity C1 and C2 are constant or function  only

 i.e. for a chosen pair of C1 and C2 U x  , g x  can be found

(Potential flow is NOT known a priori)
 Then solution of 
f '''  C1 ff ''  C2 1  f '2  0 gives 
f    u x, y  , w  
u

Uf ''  0 
, , *,, H, Cf, CD
y w
g

The Blasius Solution for Flat-Plate Flow

U=constant U x  0  C2  0
U
Then C1  gg x

d 2 2C1
dx
 
g 
U g  x   2C1x U 1 2

Let C1  1 , then g  x   2x y

U
U 2x
058:0160 Chapter 7
Professor Fred Stern Fall 2010 13

Blasius equations
f '''  ff ''  0 for Flat Plate
f 0  f ' 0  0, f '    1 Boundary Layer

Solutions by series technique or numerical

u  5 Ux
 0.99 when   3.5   Re x 
U x Re x 
*
   u
0 1  U




 '
dy  0 1  f d
2x
U
 
*
x

1.7208
Re x

 
0 

1 
u
U
u 

' ' 2x
 dy  0 1  f f
U U
d  

x

0.664
Re x
*
So,  H  2.59

u Uf '' 0  w 0.664 
Cf   
w     1 Re x x
y w 2x U U 2
2
D L dx 1.328
CD    Cf  Re L 
UL
1
U 2 L
0 L Re L ; 
;
2
v f '  f
  1 for Re x  1
U 2 Re x
 058:0160 Chapter 7
Professor Fred Stern Fall 2010 14

CD ReL
Oseen 3-226 (3rd <1
edition,vicous
flows)
Blasius 100<Re<Retr~3
×106

LE Higher
order Similar breakdown occurs at Trailing edge.
correction From triple – deck theory the correction is
C D  1.328 / Re L  2.3 / Re L +2.661/ Re7L/ 8

Rex small therefore local breakdown

of BL approximation
058:0160 Chapter 7
Professor Fred Stern Fall 2010 15
058:0160 Chapter 7
Professor Fred Stern Fall 2010 16

Falkner-Skan Wedge Flows

f  f  

f '''  C1 ff ''  C2 1  f '2  0    y g x 
Similarity
form of BL

f 0  f ' 0  0, f '    1

eq.

u U  f '  
g g2
C1  Ug x C2  Ux (Blasius Solution: C2=0, C1=1)
 

Consider Ug 
2
x  2Ugg x  g 2U x
 2Ugg x  2 g 2U x  g 2U x
 2 g Ug x  g 2U x
 2C1  C2
Hence  Ug x   2C1  C2 ,
2
C2 
g2

Ux
Choose C1=1 and C2 arbitrary=C,

Integrate Ug 2   2  C x
Ux C 1
Combine 
U 2C x
2
C  g Ux 
C
ln U  ln x  k
2C

Then U  x   kxC 2 C 
1C
 2  C 
g x  x 2 C
k
058:0160 Chapter 7
Professor Fred Stern Fall 2010 17

Change constants
U  x   kx m
y m 1 U
 y
g 2 x
2m 
f  ff   1  f
''' ''
 '2
 0 , 
m 1 ,
m 
2
f 0  f ' 0  0 f '    1
Solutions for  0.19884    1.0

Separation (  w  0 )
Solutions show many commonly observed characteristics of BL flow:
 The parameter  is a measure of the pressure gradient, dp dx .
For   0 , dp dx  0 and the pressure gradient is favorable. For
  0 , the dp dx  0 and the pressure gradient is adverse.
 Negative  solutions drop away from Blasius profiles as separation
approached
 Positive  solutions squeeze closer to wall due to flow acceleration
 Accelerated flow:  max near wall
 Decelerated flow:  max moves toward  2
058:0160 Chapter 7
Professor Fred Stern Fall 2010 18
058:0160 Chapter 7
Professor Fred Stern Fall 2010 19

7.5. Momentum Integral Equation

Historically similarity and other AFD methods used for idealized flows
and momentum integral methods for practical applications, including
pressure gradients.

Momentum integral equation, which is valid for both laminar and

turbulent flow:

 BL form of momentum equation  u  U continuity dy
y 0
w 1 d  dU
 C f   2  H 
U 2 2 dx U dx
dU
For flat plate equation 0
dx
 u
 u
 1  dy;
0U  U
*
H ;


*  u
   1  dy
0 U
  p  1 
Momentum: uu x  vu y    
x     y
The pressure gradient is evaluated form the outer potential flow using
Bernoulli equation
1
p U 2  constant
2
1
px   2UU x  0
2
 p x  UU x
058:0160 Chapter 7
Professor Fred Stern Fall 2010 20

u  U u x  v y   uu x  uv y  Uu x  Uv y ,
Continuity

1
uu x  vu y  UU x   y  uu x  uv y  Uu x  Uv y  0

0 0
1
  y  2uu x  vu y  UU x  uv y  Uu x  Uv y 


x
  
uU  u 2  U  u U x  vU  vu 
y
0
 0
1   
   y dy  (    w ) /    u U  u dy  U x  U  u dy  vU  vu  0

0  x 0 0

w   2 u  u   
 U  1  dy  U x  U  u dy  
 x  0 U  U  0 
U 2 x  2UU x  U x *

Cf
2

d
dx

 2   *
1 dU
U dx

Cf d  dU *
  2  H  ,H
2 dx U dx 
w 1 
2
 C f   x   2  H  Ux
U 2 U
058:0160 Chapter 7
Professor Fred Stern Fall 2010 21

Historically two approaches for solving the momentum integral equation

for specified potential flow U(x):

1. Guessed Profiles
2. Empirical Correlations

Best approach is to use empirical correlations to get integral parameters

(, *,, H, Cf, CD) after which use these to get velocity profile u/U

Thwaites Method


Multiply momentum integral equation by
U
 w U d  2 dU
  2  H 
U  dx  dx

LHS and H are dimensionless and can be correlated with pressure

 2 dU
gradient parameter   as shear and shape-factor correlations
 dx
 w
 S     (  0.09)0.62
U
5
H   * /   H      ai (0.25   )i
i 0

ai = (2, 4.14, -83.5, 854, -3337, 4576)

Note

U d 1 d   2 
 U  
 dx 2 dx   
058:0160 Chapter 7
Professor Fred Stern Fall 2010 22

Substitute above into momentum integral equation

1 d  2 
S ( )  U     2  H 
2 dx   
d  / U x 
U  2S   2  H    F  
dx

F     0.45  6 based on AFD and EFD

2 dU
Define z  so that   z
 dx
dz dU
U  0.45  6  0.45  6 z
dx dx
dz dU
U  6z  0.45
dx dx
i.e.
1 d
5 dx
zU6

 0.45
U
x
zU  0.45  U 5 dx  C
6
0
0.45
x
    
2 2 5
 0
6
U dx
U 0
 0 ( x  0)  0 and U(x) known from potential flow solution
Complete solution:

 2 dU
     
 dx
058:0160 Chapter 7
Professor Fred Stern Fall 2010 23

 w
 S  
U
 *  H  

Accuracy: mild px  5% and strong adverse px (w near 0)  15%

i. Pohlhausen Velocity Profile:

u
 f    a  b 2  c 3  d 4 with   y
U 
a, b, c, d determined from boundary conditions
U
1) y  0  u = 0, u yy   Ux

2) y    u  U , u y  0 , u yy  0
No slip is automatically satisfied.
separation
F    2  2 3   4
u
   F    G  ,  12    12
G    1   3 U
6 (experiment:  separation = -5)
 2 dU 2
   px
 dx U
pressure gradient parameter related to

 37  2 
         
 315 945 9072 
 

Profiles are fairly realistic, except near separation. In guessed profile

methods u/U directly used to solve momentum integral equation
numerically, but accuracy not as good as empirical correlation methods;
therefore, use Thwaites method to get  etc., and then use  to get and
plot u/U.
058:0160 Chapter 7
Professor Fred Stern Fall 2010 24

ii. Howarth linearly decelerating flow (example of exact solution

of steady state 2D boundary layer)

Howarth proposed a linearly decelerating external velocity distribution

 x
U ( x)  U 0 1   as a theoretical model for laminar boundary layer study.
 L
Use Thwaites’s method to compute:
a) Xsep
b) C f   0.1
x
L 
Note Ux = -U0/L

Solution
6
2 0.45 x
5 x
5
L  x  
  6  U 0 1   dx  0 .075 1    1
 x 0  L U 0  L  
U 06 1  
 L

can be evaluated for given L, ReL

  0  x  0,
(Note: )
 xL

 2 dU  x   6 
  0.0751    1
 dx  L  
058:0160 Chapter 7
Professor Fred Stern Fall 2010 25

X sep
 sep  0.09   0.123
L

3% higher than exact solution =0.1199

x 
C f   0.1 i.e. just before separation
L 
  0.0661
1
S     0.099  C f Re
2
2(0.099)
Cf 
Re

Compute Re in terms if ReL

 2  0.075
L
U0
1  0.1 6

 1  0.0661
L
U0
2 L 0.0661
2
 0 . 0661 
L U 0 Re L
 0.257

L Re 12 To complete
L
solution must
 1 specify ReL
Re  Re L  0.257 Re L 2
L
20.099  1
Cf  Re L 2  0.77 Re L1/ 2
0.257
058:0160 Chapter 7
Professor Fred Stern Fall 2010 26

Consider the complex potential

a a
F  z   z 2  r 2 e 2 i
2 2
a
  ReF  z   r 2 cos 2
2
a
  ImF  z   r 2 sin 2
2
Orthogonal rectangular hyperbolas

 : asymptotes y = ± x

 : asymptotes x=0, y=0

1
V     r eˆr   eˆ
r

v r  ar cos 2    0 (flow direction as shown)
2
v  ar sin 2

V  vr cos iˆ  sin ˆj   v  sin iˆ  cos ˆj  

vr cos  v sin  iˆ  vr sin   v cos  ˆj
Potential flow slips along surface: (consider   90  )

1) determine a such that vr  U 0 at r=L,   90 

U0
vr  aL cos(2  90)  U 0  aL  U 0 , i.e. a  
L
2) let U  x   vr at x=L-r:
 vr  aL  x cos(2  90)  U ( x)
U x
Or : U ( x)  a( L  x)  0 ( L  x)  U 0 (1  )
L L
058:0160 Chapter 7
Professor Fred Stern Fall 2010 27
058:0160 Chapter 7
Professor Fred Stern Fall 2010 28
058:0160 Chapter 7
Professor Fred Stern Fall 2010 29

7.6. Turbulent Boundary Layer

1. Introduction: Transition to Turbulence

Chapter 6 described the transition process as a succession of Tollmien-
Schlichting waves, development of Λ - structures, vortex decay and
formation of turbulent spots as preliminary stages to fully turbulent
boundary-layer flow.
The phenomena observed during the transition process are similar for
the flat plate boundary layer and for the plane channel flow, as shown in
the following figure based on measurements by M. Nishioka et al.
(1975). Periodic initial perturbations were generated in the BL using an
oscillating cord.
For typical commercial surfaces transition occurs at Re x ,tr  5  10 5 .
However, the transition can be delayed to Re x ,tr  3 10 6 by different ways
such as having very smooth walls and/or very low turbulent wind tunnel.

2. Reynolds Average of 2D boundary layer equations

u  u  u ; v  v  v; w  w  w; p  p  p ;

Substituting u, v and w into continuity equation and taking the time

average we obtain,
058:0160 Chapter 7
Professor Fred Stern Fall 2010 30

u v w u ' v ' w'

  0   0
x y z x y z
Similarly for the momentum equations and using continuity (neglecting
g),
DV
  p    ij
Dt
Where
 ui u j 
 ij        ui' u 'j
 x j xi 
 
Laminar Turbulent

Assume
 
  x   x which means v  u , 
a. x y

b. mean flow structure is two-dimensional: w  0 , z  0
'2
Note the mean lateral turbulence is actually not zero, w  0 , but its z
derivative is assumed to vanish.
Then, we get the following Reynolds averaged BL equations for 2D
incompressible steady flow:

u v
 0 Continuity
x y
u u dU e 1 
u v  Ue  x-momentum
x y dx  y
p v '2
  y-momentum
y y
058:0160 Chapter 7
Professor Fred Stern Fall 2010 31

Where U e is the free-stream velocity and:

u
   u 'v '
y
Note:
 The equations are solved for the time averages u and v
 The shear stress now consists of two parts: 1. first part is due to
the molecular exchange and is computed from the time-averaged
field as in the laminar case; 2. The second part appears
additionally and is due to turbulent motions.
 The additional term is new unknown for which a relation with
the average field of the velocity must be constructed via a
turbulence model.

Integrate y- momentum equation across the boundary layer

p  p e  x   v ' 2
So, unlike laminar BL, there is a slight variation of pressure across the
turbulent BL due to velocity fluctuations normal to the wall, which is no
more than 4% of the stream velocity and thus can be neglected. The
Bernoulli relation is assumed to hold in the inviscid free-stream:
dpe / dx   U e dU e / dx
Assume the free stream conditions, U e  x  is known, the boundary
conditions:
No slip: u  x ,0   v  x , 0   0
Free stream matching: u  x,    U e  x 
3. Momentum Integral Equations valid for BL solutions

The momentum integral equation has the identical form as the

laminar-flow relation:
d  dU e  C
 2  H   w2  f
dx U e dx U e 2
058:0160 Chapter 7
Professor Fred Stern Fall 2010 32

For laminar flow:

 2 dU e
( C f , H , ) are correlated in terms of simple parameter 
 dx

For Turbulent flow:

( C f , H , ) cannot be correlated in terms of a single parameter.
Additional parameters and relationships are required that model the
influence of the turbulent fluctuations. There are many possibilities all of
which require a certain amount of empirical data. As an example we will
review the  method.

4. Flat plate boundary layer (zero pressure gradient)

a. Log law analysis of Smooth flat plate

Assume log-law can be used to approximate turbulent velocity profile
and use to get Cf=Cf() relationship
1 yu *
u /u  *
ln  B where  =0.41 and B = 5
 
At y=δ (edge of boundary layer)
u * 1
U e / u  ln
*
B
 
However:
1/ 2 1/ 2
 
1/ 2
 1 / 2 U e 2 C f   2 
U e / u *  U e /  w   U e /   
     C 
   f 
1/ 2
u * U e u * C 
   Re  f 
  Ue  2 
1/ 2
 2  1  Cf 
1/ 2

   ln Re    B
C   
 f    2  
058:0160 Chapter 7
Professor Fred Stern Fall 2010 33

Following a suggestion of Prandtl, we can forget the complex log law

and simply use a power-law approximation:

C f  0.02 Re 1 / 6

b. Use u / U e profile to get , Cf, , *, and H for smooth plate

d 1
 w  U 2  C f U 2
dx 2
d
or : C f  2
dx
1 / 6
LHS: From Log law or C f  0.02 Re
d
RHS: Use u / U e to get
dx

Example:


u u 7
u /Ue  ( y /  ) 1/ 7
   (1  )dy  
0
Ue Ue 72
d 7 2 d
Cf  2  0.02 Re1/ 6  2  U e
dx 72 dx
d (Re ) d (Re ) 1
 Re1/ 6  9.72  1 / 6
 d (Re x )
d (Re x ) Re 9.72
Assuming that: =0 at x=0 or Re=0 at Rex=0:

 / x  0.16 / Re1x/ 7 or :   x 6 / 7
Turbulent BL has almost linear growth rate which is much faster than
laminar BL which is proportional to x1/2.
058:0160 Chapter 7
Professor Fred Stern Fall 2010 34

Other properties:
C f  0.027 / Re1/x 7
0.0135 1/ 7  6 / 7U 13/ 7
 w,turb 
x1/ 7
7
CD  0.031/ Re1/L 7  C f ( L)
6
1
*  
8
H   * /   1.3
w,turb decreases slowly with x, increases with  and U2 and insensitive to


c. Influence of roughness
The influence of roughness can be analyzed in an exactly analogous
manner as done for pipe flow i.e.
1 yu *

u  ln  B  B (  )
  ×
1
B (  )   ln(1  0.3  )

i.e. rough wall velocity profile shifts downward by a constant amount
B(  ) which, increases with    u* / 
A complete rough-wall analysis can be done using the composite log-
law in a similar manner as done for a smooth wall i.e. determine Cf(δ)
and θ(δ) from × and equate using momentum integral equation
d
C f ( )  2  ( )
dx
Then eliminate δ to get C f ( x,  / x)
However, analysis is complicated: solution is Fig. 7.6. For fully rough-
flow a curve fit to the Cf and CD equations is given by,
058:0160 Chapter 7
Professor Fred Stern Fall 2010 35

Fig. 7.6 Drag coefficient of laminar and turbulent boundary layers on

smooth and rough flat plates.

x
C f  (2.87  1.58 log )  2.5
 Fully rough flow
L
CD  (1.89  1.62 log )  2.5

Again, shown on Fig. 7.6. along with transition region curves developed
by Schlichting which depend on Ret = 5×105
3×106
058:0160 Chapter 7
Professor Fred Stern Fall 2010 36

5. Boundary layer with pressure gradient

ux  vy  0
 1 
uu x  vu y   ( p /  ) 
x  y
u
     u v
y
The pressure gradient term has a large influence on the solution. In
particular, adverse pressure gradient (i.e. increasing pressure) can cause
flow separation. Recall that the y momentum equation subject to the
boundary layer assumptions reduced to

py= 0 i.e. p = pe = constant across BL.

That is, pressure (which drives BL equations) is given by external

inviscid flow solution which in many cases is also irrotational. Consider
a typical inviscid flow solution (chapter 8)
058:0160 Chapter 7
Professor Fred Stern Fall 2010 37

Even without solving the BL equations we can deduce information about

the shape of the velocity profiles just by evaluating the BL equations at
the wall (y = 0)
 2u pe
 
y 2 x
pe dU e
where  - Ue
x dx

which, shows that the curvature of the velocity profile at the wall is
related to the pressure gradient.

Effect of Pressure Gradient on Velocity Profiles

Point of inflection: a point where a graph changes between concave
upward and concave downward.
The point of inflection is basically the location where second derivative
 2u
of u is zero, i.e. y 2  0

(a) favorable gradient ( px<0, Ux>0)

at y ≥ 0: uy>0, uyy<0

No point of inflection i.e. curvature is negative all across the BL and BL

is very resistant to separation. Note uyy()<0 in order for u to merge
smoothly with U.
058:0160 Chapter 7
Professor Fred Stern Fall 2010 38

(b) zero gradient (px = 0, Ux =0)

at y = yPI=0: uy>0, uyy=0
at y>0: uy>0, uyy<0

(c) weak adverse gradient ( px>0, Ux<0)

at y < yPI: uy>0, uyy>0
at y = yPI: uy>0, uyy=0
at y > yPI: uy>0, uyy<0
058:0160 Chapter 7
Professor Fred Stern Fall 2010 39

(d) critical adverse gradient ( px>0, Ux<0)

at y =0 : uy=0, uyy>0 Note that τw = 0 since uy=0.
at y < yPI: uy>0, uyy>0
at y = yPI: uy>0, uyy=0
at y > yPI: uy>0, uyy<0

(e) excessive adverse gradient ( px>0, Ux<0)

at y =0 : uy<0, uyy>0
at y < yPI: uy increases gradually to positive value, uyy>0
at y = yPI: uy>0, uyy=0
at y > yPI: uy>0, uyy<0

PI in flow, backflow near wall i.e. separated flow region

i.e. main flow breaks away or separates from the wall: large increase in
drag and loss of performance:
Hseparation = 3.5 laminar
= 2.4 turbulent
058:0160 Chapter 7
Professor Fred Stern Fall 2010 40

6. -Method

As mentioned earlier, the momentum integral equation for turbulent

flow has the identical form as the laminar-flow relation:
d C f  dU e
  2  H  (I)
dx 2 U e dx
With U(x) assumed known, there are three unknown C f , H , for
turbulent flow. Thus, at least two additional relations are needed to find
unknowns. There are many possibilities for additional relations all of
which require a certain amount of empirical data. As an example we will
review the  method.

Cole’s law of the wake:

By adding the wake to the log-law, the velocity profile for both overlap
and outer layers can be written as:
1 2
u  ln y   B  f ( )
 
where

  y /

f ( )  sin 2 (  )  3 2  2 3
2
 A/ 2

The quantity  is called Coles' wake parameter.

By integrating wall-wake law across the boundary layer:

H
  a ( )
H 1
2  3.179  1.5 2
a ( ) 
 (1   )
U 1  
Re   exp(   B  2 )
 H
058:0160 Chapter 7
Professor Fred Stern Fall 2010 41

If we eliminate  between these formulas, we obtain a unique relation

among C f  2 /  , H and  :
2

 H 2
C
 f  2 /  2
 2 / [ a (  ) ]
H 1

 2  3.179  1.5 2
 a ( ) 
  (1   ) (II)
 U 1  
Re   exp(   B  2 )
  H

Clauser's equilibrium parameter :

For outer layer,

dp
U e  u  f ( w ,  , y,  , )
dx
Using dimensional analysis:
Ue  u y  dp
 g( , )
( w /  )1/ 2
  w dx

Clauser (1954) replaced by displacement thickness :

Ue  u y
 g( ,  )
( w /  )1/ 2

 * dp  dU e
   2 H
 w dx U e dx
 is called Clauser's equilibrium parameter.

Das (1987) showed that EFD data points fit into the following
polynomial correlation:
  0.4  0.76  0.42 2

Therefore:
 dU e
 2 H  0.4  0.76  0.42 2 (III)
U e dx
058:0160 Chapter 7
Professor Fred Stern Fall 2010 42

U 1 
If we eliminate  using that Re   exp(   B  2 ) , we obtain
 H
another relation among C f  2 /  , H and  .
2

Equations (I), (II), and (III) can be solved simultaneously using say a
Runge-Kutta method to find C f , H , . Equations are solved with initial
condition for (x0) and integrated to x=x0+x iteratively. Estimated 
gives Re and ,  gives H. Lastly Cf is evaluated using Re and H.
Iterations required until all relations satisfied and then proceed to next
x.

7. 3-D Integral methods

Momentum integral methods perform well (i.e. compare well with

experimental data) for a large class of both laminar and turbulent 2D
flows. However, for 3D flows they do not, primarily due to the inability
of correlating the cross flow velocity components.

p
The cross flow is driven by , which is imposed on BL from the
z
outer potential flow U(x,z).
058:0160 Chapter 7
Professor Fred Stern Fall 2010 43

3-D boundary layer equations

 
uu x  vu y  wu z   ( p /  )  u yy  (u v);
x y
 
uwx  vw y  wwz   ( p /  )  w yy  (vw);
z y
u x  v y  wz  0;
 closure equations

Differential methods have been developed for this reason as well as for
extensions to more complex and non-thin boundary layer flows.
7.7 Separation
What causes separation?
The increasing downstream pressure slows down the wall flow and
can make it go backward-flow separation.
dp dx  0 adverse pressure gradient, flow separation may occur.
dp dx  0 favorable gradient, flow is very resistant to separation.

Previous analysis of BL was valid before separation.

Separation Condition
 u 
 w      0
 y  y 0
058:0160 Chapter 7
Professor Fred Stern Fall 2010 44

Note: 1. Due to backflow close to the wall, a strong thickening of the

BL takes place and BL mass is transported away into the
outer flow
2. At the point of separation, the streamlines leave the wall at a
certain angle.
Separation of Boundary Layer

Notes:
1. D to E, pressure drop, pressure is transformed into kinetic energy.
2. From E to F, kinetic energy is transformed into pressure.
3. A fluid particle directly at the wall in the boundary layer is also
acted upon by the same pressure distribution as in the outer flow
(inviscid).
4. Due to the strong friction forces in the BL, a BL particle loses so
much of its kinetic energy that is cannot manage to get over the
“pressure gradient” from E to F.
5. The following figure shows the time sequence of this process:
058:0160 Chapter 7
Professor Fred Stern Fall 2010 45

a. reversed motion begun at the trailing edge

b. boundary layer has been thickened, and start of the reversed
motion has moved forward considerably.
c. and d. a large vortex formed from the backflow and then soon
separates from the body.

Examples of BL Separations (two-dimensional)

Features: The entire boundary layer flow breaks away at the point of
zero wall shear stress and, having no way to diverge left or right, has to
go up and over the resulting separation bubble or wake.
Thin wall
1. Plane wall(s)

(a). Plane stagnation-point flow: no separation on the streamlines of

symmetry (no wall friction present), and no separation at the wall
(favorable pressure gradient)
(b).Flat wall with right angle to the wall: flow separate, why?
058:0160 Chapter 7
Professor Fred Stern Fall 2010 46

2. Diffuser flow:

3. Turbulent Boundary Layer

(a)

(b)

Influence of a strong pressure gradient on a turbulent flow

(a) a strong negative pressure gradient may re-laminarize a flow
(b) a strong positive pressure gradient causes a strong boundary
layer top thicken. (Photograph by R.E. Falco)
058:0160 Chapter 7
Professor Fred Stern Fall 2010 47

Examples of BL Separations (three-dimensional)

Features: unlike 2D separations, 3D separations allow many more
options.
There are four different special points in separation:
(1). A nodal Point, where an infinite number of surface streamlines
merged tangentially to the separation line
(2). A saddle point, where only two surface streamlines intersect and
all others divert to either side
(3). A focus, or spiral node, which forms near a saddle point and
around which an infinite number of surface streamlines swirl
(4). A three-dimensional singular point, not on the wall, generally
serving as the center for a horseshoe vortex.

1. Boundary layer separations induced by free surface (animation)

CFDSHIP-IOWA
058:0160 Chapter 7
Professor Fred Stern Fall 2010 48

2. Separation regions in corner flow

3. 3D separations on a round-nosed body at angle of attack

058:0160 Chapter 7
Professor Fred Stern Fall 2010 49

Video Library (animations from “Multi-media Fluid Mechanics”,

Homsy, G. M., etc.)

Conditions Producing Separation Separations on airfoil (different attack angles)

Leading edge separation Separations in diffuser

058:0160 Chapter 7
Professor Fred Stern Fall 2010 50

Effect of body shape on separation Laminar and Turbulent separation

Flow over cylinders: effect of Re Flow over spheres: effect of Re

058:0160 Chapter 7
Professor Fred Stern Fall 2010 51

Flow over edges and blunt bodies Flow over a truck

Effect of separation on sports balls