You are on page 1of 64

Chapter 4:

Incompressible Flow over Airfoils

M   0 .3
1912-1918 Prandtl (Göttingen)
• Airfoil theory
• Finite-wing theory

1875-1953
2D: Inviscid  zero drag + lift = d’Alembert’s Paradox

3D: Inviscid  induced drag + lift

2D In experiment:
 straight (zero-sweep) wing, spanning the wind tunnel
from wall to wall  no effects of wing tips: infinite wing span

1
Roadmap chapter 4

2
Airfoils

Definition of airfoils
i.e. radius ≈ 0.02c

Trailing edge

1884 Phillips: first patent on airfoil shape wind tunnel experiments


1902 Wright Brothers: Wind tunnel experiments  Lilienthal
1930’s NACA

• NACA 4-digit series NACA i j k l


Abbott and von Doenhoff (1949)
i: maximum camber i 100 c Theory of Wing Sections
j: location max camber  j 10 c Dover Publications
kl: maximum thickness kl 100 c 3
Airfoils

Definition of airfoils

• NACA 4-digit series NACA i j k l Trailing edge

i: maximum camber i 100 c


j: location max camber  j 10 c
kl: maximum thickness kl 100 c
• NACA 5-digit series NACA i j k l m
i: design lift coefficient 3 2  i 10 
Laminar flow profiles (WW-II) jk: location max camber 1 2  jk 100 c
• NACA “6-series” NACA 6j - k l m lm: maximum thickness lm 100 c
i=6: series identification Abbott and von Doenhoff (1949)
j: location minimum Cp j 10 c Theory of Wing Sections
Dover Publications
k: design lift coefficient k 10
lm 100 c
4
lm: maximum thickness At zero lift for symmetric profile
Airfoils

Definition of airfoils
Examples
Trailing edge

Company-designed airfoils:
Boeing 727, 737, 747, 757, etc.
Designed by computational methods
NLR airfoil sections  super-critical
5
DUT airfoil sections  sailplanes, wind turbines
Recommended airfoils for the project

NACA 5 series: Pretty good, available in XFoil

NACA 6 series: Good, search for the geometry yourself, e.g.


http://m-selig.ae.illinois.edu/ads/coord_database.html

Make sure to use a relatively thick airfoil, t/c > 0.15

6
Airfoil characteristics

Characteristics:
Trailing edge

cl , max : determines minimum speed of aircraft


W S
L W  Vmin 
2   C L , max
1

↑ Very important research topic


  0 : zero-lift angle of attack
symmetric airfoil (zero camber):   0  0

Theory (classical) provides:

  0 Fully separated flow


= post-stall
dcl
 2 Attached flow
d cl  
(per radian)
7
Airfoil characteristics

(max camber =0.02c at 0.4c


Experimental data NACA 2412 max thickness =0.12c)

Friction friction drag


dcl
Re , cl , max   constant cm  constant Drag
d Flow separation  form drag
Strong function of Re
M  0

dcl
 2 per radian
d
 0.1 per degree

Aerodynamic
 cl ,max  16
center:

mac  m1 c  n( xac  x 1 c )
4 4

n   cos   d sin  8
Example 4.1: NACA 2412

Airfoil Properties: c  0.64m


V  70 m s
   1.225 kg m 3 
Question:   ? 
p  1.01 105 N m 2 
cd  ? sea level conditions
T  288.15K 
d ?
   1.789 10 5 kg ms 

l  1254 N m span
Answer: l 1254
cl    0.65
1
2
V2c 2
0.5  1.225  70  0.64

From cl ( )    4 irrespective Re

 V c 1.225  70  0.65
cd  cd ( ; Re) Re      3.08 106
 1.789 105
For Re  3.1  10 6 cd for   4  cd  0.0068
d  cd 12  V2 c  0.0068  0.5 1.225  70 2  0.64  13 .06 N m span
9
Example 4.1: NACA 2412, What does XFoil give?

Re  3.110 6 , cl  0.65
Free transition
  3.75 , cd  0.00555

Forced transition at the LE

  3.78 , cd  0.01006

Abbott and Von Doenhoff

  4, cd  0.0068

10
Exam like question

What does this tell you and how should you deal with this when
designing an airplane?

11
Airfoil modeling 
n

Simulation of flow
around airfoil with lift
s: monotonically increasing
 parameter between 0 and
Discrete vortex at xv of strength  the arc length of the airfoil.

  y  yv
 
  
 e z  ( x  xv ( s)) u ( x, y )   2
u ( x)   ( x  xv ) 2  ( y  yv ) 2

2 x  xv ( s) 2  v ( x, y ) 
 x  xv
 2 ( x  xv ) 2  ( y  yv ) 2


Vortex distribution on airfoil surface x (s )

d( s )   ( s )ds     m 2 s ;    m s
  
  1  ( s)ez  ( x0  x ( s))
u ( x0 )  
2 C  
x0  x ( s)
2
ds

12
Airfoil modeling 
n

Simulation of the flow


around an airfoil with lift
  
  1  ( s)ez  ( x0  x ( s))
u ( x0 )  
2 C  
x0  x ( s)
2
ds

 
In case x0  x ( s  s0 ) on “vortex sheet”:
 
    1   1  ( s )e z  r   
u ( x0  x ( s0 )  n )    ( s0 )ez  n ( s0 ) 
2 
2 C 
r
2
ds r  x0  x ( s )

So we satisfy boundary condition on C if:


 
1   ( s )e z  r    
n ( s0 )    2 ds  U   n ( s0 ), s0   (s ) U   V e x
2 C r
Integral equation for  (s )
    ( s)ds Circulation
C l   V  d 0 13
Kutta-Joukowski d’Alembert
Kutta Condition

Kutta Condition
• For circle-cylinder we could add vortex of arbitrary
strength  and
 still satisfy the boundary
condition u  n  0 on c.c.

• For airfoil we have a similar situation: potential flow


is determined but for arbitrary circulation.

• There is only one  that results in a physically


realistic flow, namely flow that smoothly
leaves airfoil at T.E. i.e. a flow solution that is
similar to the one of a more complete flow
model such as Navier Stokes

 Kutta-(Joukowski) Condition

• In inviscid flow, Kutta Condition “models” effects due to viscosity that


cannot be neglected, i.e. flow separation at sharp edges!!

14
Kutta Condition

15
Kutta Condition

Finite TE-angle (2D)

• Flow must be tangential toboth


lower and upper surface  u  0 at TE
  p1  0  p  12  V2
• C p (a )  C p (a )
p2  12 V22  p  12  V2

p1  p2  12 V22  0 only if V2  0

Trailing Edge is Stagnation Point!


Zero TE-angle (cusp) (2D)

• Now V1 and V2 are parallel


 TE  0
• Condition p1  p2 leads to V1  V2 equal velocity in
direction and magnitude

16
Exam like question

Can you explain why the Kutta condition for a sharp trailing edge
leads to  TE  0 ?

17
Kutta Condition

In 3D?
• Flow leaves TE either along upper surface, or along lower surface

18
Airfoil modeling

Thin airfoil theory = approximation

Neglect thickness
     Position vortex sheet on camber line
U ( x0 )  V ex  u ( x0 )

Induced velocity field:


 
  1  ( s )e z  r   
u ( x0 )  
2 C 
r
2
ds r0  x0  x ( s )

Approximation: satisfy b.c. on camberline:


  
 ( s)ez  ( x0  x ( s))  
sTE
1 
n ( s0 )     ds  U   n ( s0 ), s0   (s)
2 s 0 x0  x ( s)
2

19
Airfoil modeling
 
    1   1  ( s )e z  r
Thin airfoil theory u ( x0  x ( s0 )  n )    ( s0 )ez  n ( s0 ) 
2 
2 C 
r
2
ds

2  u  2  u  2 
 C p  C p  C p    
u
Cp  1
U 2 U 2

20
Thin Airfoil Theory

Thin airfoil theory

 
   dyc
U   V cos ex  sin e y yc 
dx

n  ez  xs xs   yc ( x)ex  e y 
     
1  yc ( x) 2

Assume:  small  cos   1  ( 2 ); sin     ( 3 )   (1  ( 2 ))


yc (x) small  1  yc ( x) 2  1  ( 2 ); n  ( yc ( x)ex  ey )(1  ( 2 ))

21
Thin Airfoil Theory

Thin airfoil theory

  
 ( s)ez  ( x0  x ( s))  
sTE
1 
n ( s0 )     ds  U   n ( s0 ), s0
2 s 0 x0  x ( s)
2

Applying the “small” assumptions gives

x c
1  ( x)
2 
x 0
x0  x
dx  V [  yc ( x0 )] for x0  [0, c]

22
Thin Airfoil Theory

Result thin-airfoil theory for flat plate (=thin symmetric airfoil)


yc ( x )  0
1
c
 ( x)
2 x
0 0 x
dx  V , for x0  [0, c]

cx  0
Solution:  ( x)  2V
x

cx
c c
    ( x)dx  2V  dx  cV
0 0
x

Kutta-Joukowski: l   V    V2 c

dcl
cl  2  2 d 0
d
23
Thin Airfoil Theory

Result thin-airfoil theory for flat plate (=thin symmetric airfoil)

cx
 ( x)  2V
x

dcl
 2
d

24
Thin Airfoil Theory

Classical thin-airfoil theory


dyc
 1
dx
y  yc (x ) camberline   1

 (x) Vortex distribution ony  0 for x0  0, c 


Boundary condition on y  0 for x0  0, c 

1
c
 ( x)
2 
x 0
x0  x
dx  V [  yc ( x0 )] for x0  [0, c] Kutta Condition:  ( x  c)  0

  0,  
c
Transformation: x  (1  cos  )
2
LE :   0 : x  0

1  ( ) sin  TE :    : x  c

2 
 0 cos  cos 0
d  V (  yc ( 0 )) with  ( )  0
25
Thin Airfoil Theory
Classical thin-airfoil theory y  yc (x )

1
c
 ( ) sin 
2 x0 cos  cos0 d  V [  yc (0 )] for 0 [0,  ]  (x)

Choose form vortex  ( )  2 V [ A 1  cos 


distribution:  0
sin 
 A sin n ]
n
n 1 Regular at   0, 

And try to Singular at   0( LE ) 1  cos cx


   flatplate
find A0 , A1,.... Regular at    (TE ) sin  x
 
(1  cos ) 
sin  sin n
 A0  d   An  d   [  yc ( 0 )]
 0 cos  cos 0 n 1 0
cos  cos 0
 
cos n  sin n 0 sin  sin n
Integrals:
0
cos  cos 0
d 
sin 0
;

0
cos  cos 0
d   cos n 0


Substitution: A0   An cos n 0    yc ( 0 )
n 1

dyc
Or: ( 0 )    A0   An cos n 0
dx n 1
26
Thin Airfoil Theory
y  yc (x )

Classical thin-airfoil theory


  (x)
dyc
( 0 )    A0   An cos n 0
dx n 1
Fourier cosine series:

f    B0   B cos n
Develop in Fourier cosine series
n

 1 dyc n 1
 A0     ( 0 )d 0 
  0 dx
 f  d
1
 with B0 
 
A  2 dyc
 ( 0 ) cos n 0 d 0 0
 n
 0 dx 

 f  cos n d
2
Bn 
c 

c 0
Circulation:     ( x)dx    ( ) sin  d
0
20
  
   c V A0  1  cos d  c V  An  sin n sin  d
n 1

0
   
0
 27
 
2
for n 1
0 for n 1
Thin Airfoil Theory
y  yc (x )

Classical thin-airfoil theory


    (x)
  c V A0  1  cos d  c V  An  sin n sin  d
n 1

0
   
0

 
2
for n 1
0 for n 1

  c V [ A0  12 A1 ]

l
Kutta-Joukowski: l   V  cl  1
 V 2
c
  [2 A0  A1 ]
2  


1 dy dcl
cl  2 [   c ( 0 )(cos 0  1)d 0 ]  2 cl  2 (   l  0 )
 0 dx d


 
1 dyc
cm , c 4   A2  A1   l 0  
 dx
( 0 )(cos 0  1)d 0
4 0

Zero-lift angle of attack 28


Exam like question

The design of the NACA 5 digit series of profiles is such that at


design lift condition the slope of the camber line is aligned with
the free stream direction. Do you have an explanation for this?

29
The vortex panel method

Thin airfoil theory has several problems that limits its application

 Pressure distributions are inaccurate near the leading edge (except when
the slope of the camberline is aligned with the free-stream)

 Airfoils with high camber or large thickness violate the assumptions of the
theory, and, therefore, the prediction accuracy degrades in these situations
even away from the leading edge

 It does not include the effect of the thickness distribution on Cl, although this
effect is rarely large

We would like to do better!!!!

30
The vortex panel method

Let’s go back to the original formulation


  
  1  ( s)ez  ( x0  x ( s))
u ( x0 )  
2 C  
x0  x ( s)
2
ds


n

Impossible to obtain closed form solutions.


Therefore discrete approximation => panel method
31
Panel methods (general)
 Approximate the geometry by a certain number of (possibly curved) panels.

 Put a discrete distribution of elementary solutions (sources, dipoles,


vortices) on these panels. Vortices need to be present in case a lifting body
is computed.

 Solve for the strengths of these elementary solutions by obeying the Kutta-
condition and by requiring that the normal velocities at certain control points
are zero. The more panels the more control points can be used => more
accurate, but also more expensive to solve

32
Panel methods (general)
 They extend naturally to 3D, in contrast to the stream function formulation

 They are very fast

 Mildly compressible flow via Prandtl-Glauert. Transonic effects cannot be


modelled.

 Viscous effects via boundary layer correction (XFoil). Difficult for 3D (XFLR5
fails very often in 3D viscous mode).

 Typical tool for conceptual design.

33
Vortex panel method

 The simplest of all panel methods (also least accurate). Accuracy issues can
be overcome by using more panels.

 Only linear vortex panel methods are used. Higher order versions result in
ill-conditioned matrices (Fredholm integral equation of the first kind).

 If higher accuracy is needed, typically combine it with sources, dipoles,


quadrupoles, etc.). The panels must also be curved!!

 Used by XFoil.

34
Prandtl-Glauert equation

Compressible, small disturbance potential equation

 2  2
1  M 
2

x 2
 2  0, x is the streamwise direction
y

Introduce ~
x   x,   1  M 2

 2  2
This results in ~ 2  2  0, i.e. the standard Laplace equation
x y

Panel method can be used on a modified geometry

Always check the maximum isentropic Mach number.


This should be less than 0.9 for a valid solution!!!!!!
35
Maximum isentropic Mach number

Occurs where the pressure coefficient Cp is minimum

p
1
p  p p
Cp  1 1 ,  : specific heat ratio (1.4 for air)
 V
2  
2
2  M 2

Total pressure equation for compressible flow



   1 2   1
p t  p 1  M 
 2 

Use these two equations to determine the Mach number


that corresponds to the minimum value of Cp.

36
Exam like question

If the Prandtl-Glauert equation is used to account for


compressibility, why should you check that the maximum
isentropic Mach number < 0.9?

37
Airfoil Behavior

Wright Flyer (1903)

38
Airfoil Behavior

Thick airfoil sections  Internal structure, no wires


 C D , 0  0.032
 high C L , max  better performance

Fokker Dr-I

WW-I

Thick airfoil wing

39
Airfoil Behavior

Effects due to viscosity: friction drag + flow separation


Leading edge stall, NACA 4412
Appearance of LE separation
bubble

NACA 4412
Abrupt massive flow separation: LE Stall

 Re  2.1 105
air 
V  8 m s

 Constructed from smoke flow pictures

LE Stall: Rapid decrease of cl 40


Typical for thin airfoils (t c  (0.1,0.15 )) 40
Airfoil Behavior

Trailing edge stall,


NACA 4421

 Typical for thicker airfoils (here t c  21 % )


 Gradually movement of separation from trailing
edge in upstream direction towards
leading edge with increasing 

“soft” stall, but lower cl , max


4412
NACA  same camber line, different t c
4421 41
Airfoil Behavior

Thin airfoil stall


t c  0.02

 : separation bubble becomes


larger, until it bursts.

NACA 63-2XX

cl ,max
Re  than cl ,max 

100  t c Wind tunnel Re-number range


42
Airfoil Behavior

Airfoil performance parameters:



: Lift-to-drag ratio
d Important for performance such as range

W S Minimum speed (in horizontal


• cl , max : Vmin 
2   C L , max
1 steady flight = stalling speed

 Turning rate and turning radius: maneuverability

43
Airfoil Behavior

Effect of trailing-edge flap deflection (high lift device)

 f  or :
 shift of cl   curve
 increased/decreased
camber
 cl , max  as  f 
 cl ,max  same

44
Airfoil Behavior

Leading-edge high lift devices

NACA 4412+LE Slat

Generation of secondary
flow through gap between
 stall increases from 15 to 30   stall  slat, flap and main wing
45
  l  0  same
Airfoil Behavior

Combination of leading-edge and trailing-edge devices:

No gaps

Gaps
highest cl , max

46
Boundary Layer

Viscous Flow – Airfoil Drag



U 
1) Pressure drag = Form drag d pressure     [  ( p  p )ndC ]
 U C
n
p(s) C
Separated flow


U
Stagnation point

Reduced pressure recovery on aft part airfoil due to flow separation

If pressure recovery perfect: d pressure  0

In steady, inviscid, incompressible


attached flow drag of 2D finite Paradox of d’Alembert
body equals zero
47
Exam like question

For inviscid flow around an airfoil zero drag is found (paradox of


d’Alembert). However, for viscous flow there is a (small)
contribution from the pressure to the drag, while the same theory
is used as for inviscid flow to compute the pressure. How is this
possible?

48
Boundary Layer

Viscous Flow – Airfoil Drag



U   
2) Skin friction drag due to wall d friction    [   w t sign (u  t )dC ]
shear stress  w ( s ) N m2 U C

 
d (u  t )
w  
dn  : dynamic viscosity
 
(Newton)  u  t : tangential velocity
n : normal to wall

Skin friction drag coefficient usually No-slip condition


small: ~ 0.01
49
Boundary Layer

Estimating skin friction drag


First approximation: Replace airfoil by flat plate

Flat plate at zero incidence   w (s )


Thin airfoil   w (s )
Note: Flat plate has upper and lower surface.

Boundary Layer on flat plate at   0


U e ( x)  U   constant

Symmetric
with respect
to y  0

50
Boundary Layer

Estimating skin friction drag

Boundary-layer approximation (Prandtl 1903)


 ( x)  
 1 and 
c x y

Navier-Stokes Equations reduce to boundary-layer equations


 U c Reynolds
Re c    Simpler (parabolic)
 number for flow in thin layer along wall
driven by external velocity U e (x )

Boundary-layer flow is laminar: steady, regular  low Re


turbulent unsteady, highly irregular  high Re
transitional forward part laminar, aft part turbulent

51
Boundary Layer

Laminar flow in boundary layer incompressible

 Boundary-layer equations can be solved analytically for case of flow


along a flat plate at zero incidence   0 (Blasius)
5.0 x  U x
 ( x)  With Re x   
Re x 

 ( x)  x : Laminar bo.la. thickness increases parabolically with distance from leading edge

 Local skin friction coefficient c f (x) :


 ( x) 0.664 1
c f ( x)  1 w 2  
2  U  Re x x

 Net skin friction flat plate: C f  2Cˆ f ,with


c
1
Cˆ f   c f ( x)dx 
1.328  U c
Re c   
c x 0 Rec 
52
Example

NACA 2412
With
c  1.5m  U   30 .2 m s Sea level:
6   1.225 kg m3
Re c  3.1  10
  1.789 105 kg ms
Experiment: C f  0.0068
5.0c
 (c )   0.00426m  0.5cm very thin !!!
Rec

1.328
Cˆ f   7.54 104  C f  0.00151
Rec

Conclusion: Assumption of fully laminar flow


not valid for this high Re
53
Boundary Layer

Turbulent flow in boundary layer incompressible, time-averaged

 No analytical solution

 Empirically found expressions


0.37 x
 ( x)  15
 x4 5 Grows faster than laminar
Re x boundary layer

0.0592
c f ( x)   x 1 5
Re1x 5
c
1 0.074
Cˆ f   c f ( x)dx 
c x 0 Re1c 5

54
Exam like question

Consider a wind tunnel experiment to measure skin friction for an


airfoil. For which type of flow, laminar or turbulent, is it more
important to exactly match the Reynolds number?

55
Example

Sea level:
  1.225 kg m3
NACA 2412   1.789 105 kg ms
With
c  1.5m  U   30 .2 m s
Re c  3.1  106

Experiment: C f  0.0068

0.37c
 (c )  15
 0.0279m  2.8cm 5   (c) laminar !
Rec
0.074
Cˆ f   0.00372  C f  0.00744 5C f , laminar !
Re1c 5

Conclusion: Overprediction when assuming flow in boundary layer is


turbulent from leading edge to trailing edge.
 Boundary-layer flow is transitional 56
Boundary Layer

Laminar versus turbulent boundary layer flow

57
57
Boundary Layer

Transitional flow in boundary layer


Instabilities in boundary layer grow if Re x  Re crit → turbulence

 turb ~ ( x  x0 ) 4 5

 U  xcr
 Recr    (5 105 )  depends on
 - surface roughness
- inflow turbulence
- etc.
 xcr Follows from  experiment
 flow stability criteria 58
 prediction is very difficult
Example
Sea level:
  1.225 kg m3
  1.789 105 kg ms
Transition position
Re xcr   500000 1.789 105
U   50 m s  xcr  
 U  1.225  50
 0.146 m  14.6 cm from LE

U   100 m s  xcr  0.0730 m  7.3 cm from LE

U   30 m s  xcr  0.2434 m  24 .34 cm from LE

NACA 2412
c  1 .5 m
Re c  3.1  106 59
Boundary Layer

Procedure for computing skin


friction transitional boundary layer
If flow fully turbulent; 1)+2):
k
dˆ f ,t ,c  12  U 2 cCˆ f ,t ,c  q c t nt
Rec kt  0.0740
k
Turbulent drag in region 1): dˆ f ,t , x1  q x1 t nt  1 5
nt
Re x
1
c x
Turbulent skin friction region 2): dˆ f ,t ,c  dˆ f ,t , x1  q kt [ nt  1nt ]
Rec Re x1
k kl  1.328
Region 1), laminar: dˆ f ,l , x  q x1 l
1
Re l
n nl  1 2
x1
Total: dˆ f  dˆ f ,l , x1  dˆ f ,t ,c  dˆ f ,t , x1 dˆ f kl x1 kt kt x1
Cˆ f    
dˆ f 1 nl 1 nt q c Re nx1l c Re cnt Re nx1t c
k x  k x 
Cˆ f   l nl  1   t nt [1   1  ]
q c Rec  c  Rec c 60
Boundary Layer

Procedure for computing skin friction


transitional boundary layer
NACA 2412:
Recr  5 105
Rec  3.1106

x 5 105
  0.16129
c 3.110 6

12 45
1.328  x  0.074  x1 
Cˆ f  1 2  1   [1    ]  0.00030  0.00285  0.00316
Rec  c  Re1c 5 c

C f  2Cˆ f  0.00632 Experiment: C f  0.0068

Conclusion: Reasonably OK. Also some pressure drag, maybe. 61


Boundary Layer

Effect of variation of Re xcr


Re xcr ,i.e. transition, depends on many parameters, so
does C f in the case of transitional boundary layers

 U x 
Re xcr    cr 
  xcr Re x cr
 
 U  c  c Re c
Re c 
 

Linear dependence
on Re x
cr

Transition point moves


downstream
Lower drag

62
Boundary Layer

For given Re c Ĉ
, f decreases
as transition point moves
downstream

63
Boundary Layer
Flow Separation  Pressure Drag
 Large contribution in c d

Separated Flow:
Loss of pressure on aft
Attached Flow:
Balance in horizontal component pressure force on forward
part of airfoil
(+drag) and aft (-drag) part of airfoil
 Unbalance

 Pressure drag
Computed, inviscid flow
Pressure on upper surface

cp 

64

You might also like