Vous êtes sur la page 1sur 6

Chapter 4

Circulation and lift


4.1 Circulation and outow
4.1.1 Curve-integrals in plane ideal ow
We have seen in 3.4 how both Cartesian components u = 'w and v = w can be recovered
from either the real ( = 'W )
u =

x
, v =

y
(4.1)
or imaginary ( = W )
u = +

y
, v =

x
(4.2)
parts of the complex potential W dened by dW/dz = w (3.45). We now investigate recovering
the scalar potential and the stream-function from the velocity. This leads to important and
useful physical interpretations of and .
Integrating (3.45) between two points in the uid uid gives
W(z
2
) W(z
1
) =

z2
z1
w dz . (4.3)
Expanding (4.3) into real and imaginary parts gives
(z
2
) (z
1
) + i(z
2
) (z
1
) =

z2
z1
(u iv) (dx + idy) (4.4)
=

z2
z1
(u dx + v dy) + i

z2
z1
(u dy v dx) . (4.5)
Now, as in the discussion of circulation in 2.3.3,
u dx + v dy = q ds ; (4.6)
i.e., the tangential component of velocity along an innitesimal curve segment. And, as in the
discussion of conservation of mass in 2.2.1,
u dy v dx = q n ds ; (4.7)
41
42 AERODYNAMICS I COURSE NOTES, 2005
i.e., the component of velocity at right-angles to the same segment (reckoned positive when the
velocity crosses the segment from left to right). Thus
(z
2
) (z
1
) = '

z2
z1
w dz =

z2
z1
q ds (4.8)
(z
2
) (z
1
) =

z2
z1
w dz =

z2
z1
q nds . (4.9)
In words, change in scalar potential along a curve through the ow eld gives minus the circulation
along that curve, and the dierence in the stream-function between two ends of the curve gives
the volumetric ow rate (per unit span) through the curve from left to right.
Also, formulae (4.8)(4.9) can be used to determine or measure either or , up to an
arbitrary additive constant: the value at some reference point z
1
.
4.1.2 Closed circuits
Now say we take the nal point, z
2
, of the curve to coincide with rst point, z
1
; i.e. close the
curve to form a circuit C . Then
[W]
C
=

C
w dz (4.10)
(4.11)
where [W]
C
means the jump in W as it moves around the circuit C . Of course, if W is a
single-valued function of z , [W]
C
is necessarily zero for any circuit C ; however, we will see that
multivalued complex potentials are also of interest in aerodynamics.
The real and imaginary parts involve the circulation
C
around and outow Q
C
through the
circuit C,
[]
C
+ i[]
C
=

C
(u dx + v dy) + i

C
(u dy v dx) (4.12)
=

C
q ds + i

C
q n ds (4.13)

C
+ iQ
C
. (4.14)
If the circuit C is lled with uid, the circuit integrals can be replaced with integrals of
vorticity and divergence over the contained region R (see 2.3.3 for the circulation and 2.2.1
for the outow):
[]
C
=
C
=

v
x

u
y

dx dy (4.15)
[]
C
= Q
C
=

u
x
+
v
y

dx dy . (4.16)
If the velocity eld in the region R is irrotational, the integrand in the integral for the circulation

C
vanishes, and if its divergence-free, the integrand in the integral for the outow Q
C
vanishes.
If the circuit is not lled with uid in plane ideal ow, say, for example, because it encloses
an obstacle such as an aerofoil, then the integrals over R are meaningless, the circulation and
outow need not vanish, and the complex potential might be multivalued. The same conclusions
apply if the circuit contains a singular point at which the velocity is undened.
Circulation and lift 43
4.1.3 Example: powers of z and circles around the origin
Say we take w = A
k
z
k
= A
k
r
k
e
ik
, where A
k
is a complex constant, and let the circuit C be a
circle of radius a, centred on the origin; i.e. on C , z = ae
i
. Then around the circuit,
dz = d(ae
i
) = aie
i
d (4.17)
with going from 0 to 2, and the integral around the circuit is

C
w dz = A
k

2
0
a
k
e
ik
iae
i
d (4.18)
= ia
k+1
A
k

2
0
e
i(k+1)
d . (4.19)
Now if k ,= 1, this is

C
A
k
z
k
dz = ia
k+1
A
k

e
i(k+1)
i(k + 1)
2
0
(4.20)
= 0 , (4.21)
since e
i
= cos + i sin is periodic in with period 2.
For the special case k = 1, though, we have

C
A
1
z
dz = ia
0
A
1

2
0
e
i(0)
d (4.22)
= iA
1

2
0
d (4.23)
= 2iA
1
. (4.24)
Notice that this result is independent of the radius of the circle used for integration. In fact,
though not proven here, the circulation and outow Q are both independent of the shape of
the circuit too, so long as the circuit encloses the same obstacles and singularities.
In summary, for w = A
k
z
k
, = Q = 0 for k ,= 1 . For w = A
1
/z , = 2A
1
and
Q = 2'A
1
. For this reason we usually represent a point source at the origin as
w
source
=
Q
2z
(4.25)
and a point vortex as
w
vortex
= i

2z
, (4.26)
and call Q and the strengths of the source and vortex. For Q < 0 , we call the source a sink.
The results obtained here correspond to the fact that the complex potentials for w = A
k
z
k
are, as in 3.5.2,
W =

A
k
z
k
dz =

A
k
(k + 1)
1
z
k+1
, (k ,= 1) ;
A
1
ln z , (k = 1) ,
(4.27)
i.e. all single-valued except for when k = 1 .
44 AERODYNAMICS I COURSE NOTES, 2005
4.2 More on the scalar potential and stream-function
4.2.1 The scalar potential and irrotational ow
We have seen that in an ideal ow, the dierence in the scalar potential between two points is
equal to minus the circulation along a path though the uid joining them. In general, any plane
velocity eld derived from a scalar function by the operations
u =

x
, v =

y
(4.28)
will be irrotational, since

v
x

u
y
=

2

xy


2

yx
= 0 . (4.29)
On the other hand, if the scalar potential isnt the real part of an analytic function, the derived
velocity eld neednt be divergence-free; the condition q = 0 (2.3) becomes

x
2
+

2

y
2
= 0 . (4.30)
This second-order linear partial dierential equation is called Laplaces equation, and is often
written concisely as

2
= 0 . (4.31)
4.2.2 The stream-function and divergence-free ow
If a velocity eld is derived from a scalar function by
u = +

y
, v =

x
, (4.32)
it will automatically be divergence-free since
u
x
+
v
y
=

2

xy


2

yx
= 0 . (4.33)
On the other hand, it neednt be irrotational: the condition that it is is
=
v
x

u
y
=

x
2


2

y
2

2
= 0 . (4.34)
Thus, velocity elds derived from an analytic complex velocity are automatically divergence-
free and irrotational, those derived from a scalar potential are irrotational, and those from a
stream-function divergence-free. Scalar potentials or stream-functions can be used for ideal ow,
but only if they satisfy Laplaces equation (4.31).
For more general, nonideal, kinds of ow, there is no complex velocity, and these representa-
tions can be useful. Thus, the scalar potential can be used for irrotational ow of compressible
uids (for which generally the divergence is nonzero), and the stream-function can be used for
viscous ow of incompressible uids (for which vorticity is generated at the ow boundaries).
4.3 Lift
In this section, we derive an expression for the aerodynamic force per unit span on an impermeable
object in a plane ideal ow in terms of the complex velocity. We then show that the drag always
vanishes while the lift is proportional to the circulation around the object.
Circulation and lift 45
4.3.1 Blasiuss theorem
The Cartesian components of the aerodynamic force per unit span on an obstacle with contour
C are given by a
x
=

C
p dy and a
y
= +

C
p dx (see 2.3.1) . From them, form the complex
quantity
a
x
ia
y
=

C
p (dy + i dx) = i

C
p dz

(4.35)
and use Bernoullis equation for ideal ow (2.26), dropping the constant terms which integrate
to zero,
a
x
ia
y
= i

+
q
2

2

q
2
2

dz

(4.36)
=
i
2

C
q
2
dz

(4.37)
=
i
2

C
ww

dz

. (4.38)
Now
w

dz

= (w dz)

= dW

= d i d (4.39)
but d = 0 along the outline of an impermeable obstacle, so we can replace dW

with its complex


conjugate d + i d = dW w dz and
a
x
ia
y
=
i
2

C
w
2
dz (4.40)
which is Blasiuss Theorem.
References: Glauert (1947, p. 81), Milne-Thomson (1973, p. 89).
4.3.2 The KuttaJoukowsky theorem
Now say, for example, the ow is composed of a uniform stream with speed q

and incidence
and a vortex of strength ; i.e.
w = q

e
i
+ i

2z
. (4.41)
Then
w
2
= q
2

e
i2
+ i
q

z
e
i


2
4
2
z
2
, (4.42)
and using the results on the circuit integrals of powers of z from 4.1.3

C
w
2
dz = 2i

i
q

e
i

(4.43)
= 2q

e
i
. (4.44)
Blasiuss theorem (4.40) then gives
a
x
ia
y
= iq

e
i
. (4.45)
Since the drag is measured in the direction of the airstream () and the lift is measured a right-
angle anticlockwise from the lift, we have (as in 3.3.4 for the rotated orthogonal components of
velocity)
d i = e
i
(a
x
ia
y
) (4.46)
= iq

; (4.47)
46 AERODYNAMICS I COURSE NOTES, 2005
i.e.
d = 0 , = q

. (4.48)
If circulation is positive ( > 0 ), the vortex is clockwise and the lift is positive ( and q

are
always positive).
These two results constitute the KuttaJoukowsky Theorem, which actually holds for any ow
over any aerofoil and is perhaps the most important result in two-dimensional aerodynamics. It
means that instead of having to calculate the aerodynamic force by integrating contributions
from stresses around the surface, we can just calculate the circulation and use (4.48).
The KuttaJoukowsky Theorem is easily veried for plane ideal ow over a circular cylinder.
More general ows than (4.41) could be considered by adding more terms to the series. Terms
of the form z
k
with positive k should be excluded, however, as they imply speeds increasing
without bound in the far eld (q [z[
k
). Terms of the form z
k
with k = 2, 3, . . . on the other
hand all only produce terms like z
m
with m 2 in w
2
, and none of these contribute to the
contour integral: all z
m
terms vanish except those for m = 1 .
The interpretation of the theorem is straightforward. We know that the pressure is related to
the kinetic energy of the uid (i.e. the square of the velocity) by Bernoullis equation, and so it
is pairwise products of velocities that will be of interest. Of all possible pairs of velocities of the
type z
k
, the only combination giving an average speed dierence between the upper and lower
surfaces is one varying like cos or sin or e
i
, not e
in
for any n ,= 1 . If the velocities are
restricted to have k 0 (by the requirement of bounded velocity in the far eld) this leaves only
the possibility 01 = 1 or 1+0 = 1, the uniform streamvortex pairing. In this pairing, the
vortex on one side reinforces the stream and on the other retards it, and it is this that generates
the net dierence in speeds, kinetic energies, and pressures, and therefore generates the net lift.
On the other hand, there exists no such mechanism for the generation of drag by plane ideal
ow.
References: Glauert (1947, p. 82), Prandtl and Tietjens (1957, pp. 161166), Abbott and von
Doenho (1959, ch. 3), Milne-Thomson (1973, p. 91), Kuethe and Chow (1998, p. 107), Anderson
(2001, pp. 238239), Bertin (2002, p. 90).