3D Wings

Aero 301: Spring 2011
III.5 3D Vortices & BiotSavart
Page 1
III.5 Vortices in 3D & The BiotSavart Law

We are finally ready to start thinking about the aerodynamics of 3D objects: wings that
do not extend to into and out of the page. To do this we start by thinking about
what would happen if we were to impulsively start a motionless airfoil in a 2D world.
We know 2 (seemingly) contradictory things
1. There will be vorticity/circulation associated with
the lift the moving airfoil generates but,
t < 0, No Motion
=0
2. In the initial motionless state = 0 everywhere and

D /Dt = 0 so we should have = 0 everywhere for all time.
What gives?
D /Dt doesnt hold at the trailing edge (the
Kutta condition arises because of viscosity) so
we do generate the vorticity and circulation
required to produce the lift.
t > 0, Airfoil moving right
But, besides that point, the flow remains inviscid

so no torque is applied to a large control volume
enclosing the wing and lots of space around it so the
circulation about that volumes perimeter remains zero.
>0
=0
<0
bound vortices
So, for some region close to the wing to have positive circulation, some
negative circulation path must exist around a vortex not bound to the airfoil.
This other negative- vortex is called a starting vortex. It has equal but
opposite strength to the net vortex strength thats bound to the moving airfoil.
starting vortex
Does this really happen? Yes!

Set aside this 2D picture for a few moments
and lets move gingerly into 3D. . .
Imagine that we make the simplest extension
from our 2D airfoil picture into 3D. This
would mean that the point vortices would
become lines that extend to in
the y direction, into and out of the page.
Next, imagine that instead of just vortex lines,
the 3D world can have vortex filaments, twisty
vortex strings that produce infinite v -type velocities
as distance from the filament goes to zero.
(You know these twisty vortices as tornados.)
Working from the 3D incompressible Euler equations, we
could prove three vortex theorems developed by Helmholtz.
(But we wont.)
H1 A vortex filament has constant strength, along its length.
H2 A vortex filament cannot begin or end in a fluid. It must
end at a boundary, form a closed loop or extend to infinity.
H3 An inviscid fluid that is initially irrotational will remain
irrotational for all time.
H1 and H2 do not apply in 2D
~ /Dt = 0.
H3 is nothing more than D
Page 2
Page 3
The closed loop option of H2 explains the starting vortex thats

observed in 2D flows. The two vorticies one bound to the airfoil, the
other behind the airfoil are just two bits of the same vortex loop.
pa
gs
in
w
=
b
As a finite wing (a 3D shape that

doesnt extend to in y) begins to
move, the Kutta condition generates a
vortex loop with one part of the vortex
bound to the wing, a starting vortex that
remains more or less at its starting
position and two legs called wingtip
vortices that connect the bound vortex
to the starting vortex.
What about the rest of the loop?
We can observe the wingtip vortices and the starting

vortex. We know that the bound vortex exists
because we can measure the lift on the wing: L = U b
(although this ignores some nasty details). So, the
theoretical picture is in good agreement with observation.
So, when a finite wing starts to move, it generates a starting vortex
and this remains pretty much where it originated (unless its near the
ground). As the wing moves, the bound vortex moves with it (so the
Kutta condition is satisfied) and the wingtip vortices get longer and
longer.
n
rti
x
te
r
vo
a
St
= constant
around loop
Top View
Page 4
Rear Views
Close to
trailing
edge
Side View
Far from
trailing
edge
III.6 Prandtls Lifting Line
Page 5
How do we analyze the effect of 3D vortex filaments?

The 3D extension of v = /2 r is known as the
BiotSavart Law:
~v =
d~s r
4 |~r|3
d~s points along the vortex filament indicating the sense

of rotation (right-hand rule);
~r is the vector pointing from the point of interest (i.e.,
where the velocity induced by the vortex filament is
evaluated) to the point s along the vortex; and
vortex
filament
the integral is evaluated along the length of the vortex

which is usually or a closed loop
The velocity potential of this field satisfies Laplaces
Equation in 3D as it must.
ds
When the BiotSavart Law is applied to a straight vortex

filament that extends from , the velocity reduces to the
correct 2D behavior.
Use the setup to the right to verify this.
(Hints: r = h/ cos and s = h tan .)
As another example, what is the velocity at the
center of a circular vortex ring?
du
Page 6
III.6 Prandtls Lifting Line Theory, Downwash and Induced Drag

What are the implications of H1, H2 and the BiotSavart
Law on finite wings? How do we model finite wings?
= constant
around loop
Imagine that all the little ds vortices arrayed along the

camber line of a 2D airfoil are collected into a single
thats placed at the quarter-chord point.
The starting vortex is so far aft, it does not induce any

velocity near the wing.
The wingtip vortices generate downwash: w(y) < 0
between themselves, including at the location of the
bound vortex.
What is the downwash velocity induced by the pair of
wingtip vorices along the bound vortex? Calculate using
BiotSavart. . .
in
gs
In 3D, this vortex cannot end so wingtip vortices

connect the bound vortex back to the starting
vortex that is still sitting on the runway, 1000
miles behind the wing.
The bound vortex generates lift.
pa
n
w(y) < 0
w(y) =
Page 7
b [1 (2y/b)2]
The preceding expression is a bit scary. It implies we have infinite

downwash velocities at the wing tips. Even disregarding the
impossibility of infinite velocities, we expect our small-angle
approximations do not work and that the wingtips are probably
stalled.
What do we do about this?
The next level of approximation allows to vary
along the wingspan, despite the fact that must
be constant along a vortex filament. If we let
0 at the wing tips, maybe we can avoid infinite
downwash.
How do we have our cake and eat it too?
Lifting Line
1
Because a single vortex filament must have

constant , we simply stack a number of bound
vortices along the lifting line but allow them to
turn back into trailing vortices at different points
along the span. This allows for a varying (y).
3
2
1
Page 8
Why might vary along the span? All the reasons that L can
vary for an airfoil: chord, angle of attack and zero-lift angle
can all vary along the span.
Chord variations are called taper: c = c(y)
Angle of attack variations are called twist: = (y)
Zero-lift angle variations are called
aerodynamic twist: L=0 = L=0(y)
And, as we will see, there is an induced angle of attack,
i (y ) that decreases the effective angle of attack to less
than the geometrical angle of attack.
Now we need to consider what multiple vortices mean for the
downwash velocity w(y) along the lifting line.
Consider the diagram shown to the right that shows
the lifting line from directly upstream. The
downwash at the point y0 due to each of the
half-infinite vortices is
wn(y0 ) =
n
1
2 2 (y0 yn )
where n is an index for each of the half vortices and

yn is the y location of each of these. Each of the A
vortices is positive, each of the B vortices is the
negative of the corresponding A.
z
1,A
b/2
2,A
y
y0
3,A
3,B
2,B
1,B
b/2
Page 9
In the picture the vortices 1A, 2A, 3B, 2B, and 1B induce
velocities down; 3A induces a velocity up. The different
directions is given by the signs of y0 yn and the signs of n
If we have a large number of very weak vortices we can do a
little calculus and say

d
1
d (y)
dy
=
dw(y0 , y) =
4 (y0 y)
4 (y0 y) dy y
and, with this,
1
w(y0 ) =
4

1 d
dy.
b/2 y0 y dy y
Z b/2
This is the net downwash at y due to all the vortices.

The net effect of this downwash is to tilt the
incoming flow vector down the tilt angle is called
an induced angle of attack, i and, like w(y), this
angle depends on y.
We define the induced angle of attack to be positive if
w < 0 (as it usually is) so
i (y0 )
w(y0)
U
if
w(y) U
U
w(y0)
i(y0)
What are the effects of this tilt?

1. The effective angle of attack is less than the geometrical angle
of attack (the angle between the chord line and U and this
leads to less lift at each station along the blade that you would
expect based on the geometrical angle of attack.
2. The aerodynamic force perpendicular to U at any y0 is L cos(i )
(i.e., it is decreased slightly because it is tilted back)
3. There is now an aerodynamic force in the direction of U
called induced drag: Di = L sin i .
To sort out the implications of all this, we need a way to determine
(y). This function determines L and w at each station along the span
and needs to correctly reflect all the geometrical features of the wing.
The strategy for finding (y) is to equate two separate expressions for
L for each 2D airfoil section along the wing. First, the
KuttaJoukowski Theorem gives at y0
L (y) = U (y0 ).
Second, the 2D airfoil characteristics give
1
L (y) = 2 [ (y0 ) i (y0 ) L=0(y0 )] U2 c(y0 )
2
So, we have two different expressions for L at any y0. If we set these
equal to each other we can solve for the one thing we do not know for
a wing we have built, the (y) distribution.
Page 10
III.7 Elliptical Wings
Setting the two L espressions equal results in the

Fundamental Equation of Finite Wing Theory
"
2 (y0 )
1
= 2 (y0 ) L=0(y0 )
U c(y0 )
4 U
#
1 d
dy
b/2 y0 y dy y
Z b/2
If we solve this equation for (y0 ) we know the the lift at each
section and, from this, the lift on the wing.
This whole process is very similar to the development of
thin-airfoil theory. Equating two expressions for L gives (y0 )
and integrating (y0 ) over the wingspan gives the overall lift.
It is difficult to solve this equation (it is another integral

equation) so our solution procedure will again include a sine series
(this is nice because we would like to have = 0 at the wingtips
to avoid infinite downwash.
What is unfortunate about all this is that for a given wing at a
given U more lift requires more (via an increased geometrical
angle of attack). We see that increasing also increases the
drag. However, increasing also increases i so, actually, Di
increases like L.
Page 11
Page 12
III.7 The Elliptical Lift Distribution

Solving the Fundamental Equation of Finite Wing Theory requires us to guess at a (y) distribution and show
its a correct guess. (The same approach we used for the (x) distribution for thin airfoils.)
As a first guess we consider an elliptic distribution:
"
2y
(y) = 0 1
b
2 #1/2
This distribution has circulation 0 at the root (y = 0) and = 0 at the wingtips (which avoids the infinite
downwash problem).
First, lets compute the downwash by taking the derivative d /dy and performing the variable transformations:
2y/b = cos . With this we obtain
w=
0
2b
An elliptic distribution produces uniform downwash.
Is such a distribution possible? Put it into the

fundamental equation of finite wing theory to verify. . .
Yep! It works for an elliptic c(y) distribution if there is no

twist and no aerodynamic twist (i.e., and L=0 are
constants along the span):
"
2#1/2
2y
c(y) = c0 1
b
We can integrate this chord distribution to find the
planform area, S and the aspect ratio, AR= b /S
S=
c0b
4
and
AR =
4b
c0
We prefer to cite results in terms of b and AR rather than b

and c0 because not all wings are elliptical but all have an
unambiguious wingspan and a (nearly) unambiguous
planform area.
Page 13
Because we know (y), then we can integrate across the span to

find the lift:
0 =
4U b ( L=0)
AR + 2
And with w << U such that cosi 1 then i can be written as:
L=
U2 b2( L=0)
AR + 2
CL = 2 ( L=0)
AR
AR + 2
Page 14
What about the induced drag? Integrating across the span gives
Di =
02
8
The induced drag coefficient is
CDi =
CL2
AR
The induced drag coefficient depends on the lift coefficient squared.
Page 15
III.8 Non-Elliptic Lift Distributions
Page 16

In general, any combination of chord distribution, twist distribution, and aerodynamic twist distribution will
produce lift and induced drag.
To accommodate all the possible variations, a Fourier series approach is used that maintains = 0 at y = b/2:
= 2bU
An sin(n )
where
cos = 2y/b
n=1
This series approach is nothing more than a generalization of the elliptic distribution because the elliptic
distribution has
(
0 /2bU n = 1
An =
0
n>1
Using this approach the odd coefficients, A1 , A3 , . . . , represent the symmetric variations in that one typically
imagines for a normal wing.
The even coefficients represent asymmetric variations (i.e., when the left wing is different than the right
wing). This doesnt seem to be very common at first but these are the terms that are used to model aileron
displacements that induce rolling moments.
Page 17
As before, we can think about the results we would get with a given set of An s without actually computing the
coefficients. The approximate lift and induced drag are given by integrations similar to those used for the
elliptic distribution.
Begin by integrating for the lift (watch for helpful orthogonality!)
CL = A1 AR
The lift coefficient only depends on the first coefficient in the series. However, unlike thin airfoil theory, this
single-term result becomes a better approximation to the lift as the number of Ans computed is increased.
(See below.)
Similarly, for drag,
CL2
CDi =
(1 + )
AR
where
Page 18
An
= n
A1
n=2
2
is always positive and, for a reasonably well designed wing, 1

so the preceding expression is often written
CL2
CDi =
e AR
where
e = (1 + )1 1
The symbol e is selected because this number is an efficiency that is never greater than 100%.
Note that an elliptic lift distribution gives the minimum possible induced drag for a particular AR because it is
the only distribution that gives e = 1 because all of its An terms equal zero for n > 1.
To find the An s choose some number, N, of terms to keep in the sine series.
Page 19
Then, choose N points along the span and evaluate the fundamental equation at those N points using the
values of c, , and L=0 for each point. This gives N equations for the N unknown Ans.

3D Wings

Uploaded by

Copyright:

Available Formats

3D Wings

Uploaded by

Document Information

Original Description:

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

3D Wings

Uploaded by

Copyright:

Available Formats

Aero 301: Spring 2011

III.5 3D Vortices & BiotSavart

III.5 Vortices in 3D & The BiotSavart Law

2. In the initial motionless state = 0 everywhere and

t > 0, Airfoil moving right

But, besides that point, the flow remains inviscid

Aero 301: Spring 2011

III.5 3D Vortices & BiotSavart

Does this really happen? Yes!

Aero 301: Spring 2011

III.5 3D Vortices & BiotSavart

The closed loop option of H2 explains the starting vortex thats

As a finite wing (a 3D shape that

What about the rest of the loop?

We can observe the wingtip vortices and the starting

Aero 301: Spring 2011

III.5 3D Vortices & BiotSavart

Aero 301: Spring 2011

III.6 Prandtls Lifting Line

How do we analyze the effect of 3D vortex filaments?

d~s points along the vortex filament indicating the sense

the integral is evaluated along the length of the vortex

When the BiotSavart Law is applied to a straight vortex

Aero 301: Spring 2011

III.6 Prandtls Lifting Line

III.6 Prandtls Lifting Line Theory, Downwash and Induced Drag

Imagine that all the little ds vortices arrayed along the

The starting vortex is so far aft, it does not induce any

In 3D, this vortex cannot end so wingtip vortices

Aero 301: Spring 2011

III.6 Prandtls Lifting Line

The preceding expression is a bit scary. It implies we have infinite

Because a single vortex filament must have

Aero 301: Spring 2011

III.6 Prandtls Lifting Line

where n is an index for each of the half vortices and

Aero 301: Spring 2011

III.6 Prandtls Lifting Line

This is the net downwash at y due to all the vortices.

Aero 301: Spring 2011

III.6 Prandtls Lifting Line

What are the effects of this tilt?

Aero 301: Spring 2011

III.7 Elliptical Wings

Setting the two L espressions equal results in the

It is difficult to solve this equation (it is another integral

Aero 301: Spring 2011

III.7 Elliptical Wings

III.7 The Elliptical Lift Distribution

An elliptic distribution produces uniform downwash.

Aero 301: Spring 2011

III.7 Elliptical Wings

Is such a distribution possible? Put it into the

Yep! It works for an elliptic c(y) distribution if there is no

We prefer to cite results in terms of b and AR rather than b

Aero 301: Spring 2011

III.7 Elliptical Wings

Because we know (y), then we can integrate across the span to

Aero 301: Spring 2011