You are on page 1of 30

N81 -70009

NASA

Technical

Memorandum

IIIIIIIIIIIHIMllllllll

81230

Effect of Winglets on the Induced


Drag of Ideal Wing Shapes
R. T. Jones and T. A. Lasinski

September

1980

National Aeronautics
Space Administration

and
REPRODUCED BY
U.S. DEPARTMENT
OF COMMERCE
NATIONAL TECHNICAL
INFORMATION SERVICE
SPRINGFIELD, VA 22161

....
/i

1.

Report

No.

NASA
4.

Title

2. Government

Accretion

No.

and

Subtitle

EFFECTS
OF WINGLETS
WING SHAPES

ON

THE

INDUCED

DRAG

OF

9.

Recipilmt's

Catalog

T.

Jones

Plrformi_

Report

6.

Performing

Orglnizltion

Code

8.

Performiog

OrgBnizlltion

Report

and

_gani_ti_

T.

Name

A.

aM

Date

Research

Moffett

Laslnski
10.

Work

Unit

No.

11.

Contract

or

13.

Type

Report

14.

Sponsoring

_m

12.

Spomori_

Center,

Field,

Agency

_rne

CA

and

NASA

SUpl;:dem_'ttary

16.

Abstract::

I-OR
Grant

No.

94035

Addrms

of

and

Technical

National
Aeronautics
and
Washington,
D.C. 20546
15.

No.

A-8329

505-'_1-I
Ames

No.

IDEAL

7. Author(,)
R.

3.

TM-81230

Space

Administration

Period

Covered

Memorandum
Agency

Code

Note,.

A conventional

wing

having

a given

lift

and

limited

span

achieves

minimum
induced
drag when the spanwlse
distribution
of llft is elliptical.
However,
if the limitation
on span is removed
and replaced
by a structural
constraint
on the integrated
bending
moments,
Prandtl
found
that a
10-percent
reduction
of induced
drag can be _chleved
by a 10-percent
increase
of wing span accompanied
by a more highly
tapered
loading.
In
the present
report,
we have extended
such calculations
to wings
having
vertical
tip extensions
or wlnglets.
It is found
that essentially
the
same result
can be obtained
by a 15-percent
vertical
extension.
Thus,
it
appears
that with ideal wing shapes
similar
reductions
of induced
drag can
be achieved
by either
horizontal
or vertical
tip extensions.

17.

Key

W_

(Sugg_ted

by

Auth.(s))

18.

Distri_tion

Statement

Unlimited

Aerodynamics
Wings
Winglets
Tip
19.

Security

STAR

extensions
Oa=if.

(of this re_rt)

20.

Unclassified

Security

Cla_if.

(of

this

Category

_)

-- 02
21.

Unclassified
"For

sale by the National

Ter.hPic_l

Information

No,

of

29
Service,

Springfield,

Virginia

22161

Pages

??

Price

NASA

Technical

Memorandum

81230

Effect of Winglets on the Induced


Drag of Ideal Wing Shapes
R. T. Jones
T. A. Lasinski,

Ames

Research

N/ A
National Aeronautics
Space Administration

and

Ames Research Center


Moffett Field. California 94035

Center,

Moffett

Field,

California

EFFECTS

OF WINGLETS

ON THE

R. T.

Jones
Ames

INDUCED
and

DRAG

T.

A.

Research

OF

IDEAL

WING

SHAPES

Lasinski

Center

It has been known


for many years
that vertical
fins or end plates
at the
tips of a wing can significantly
reduce
vortex
drag.
Recent
work (ref. i)
has shown
considerable
improvement
over earlier
designs
and has raised
the
question
of whether
such vertical
extensions
should
be a part of the basic
design
of a wing intended
for maximum
efficiency.

tantly

The answer
to such
on the weight
of

a question
depends
the wing structure.

on

several
factors,
As is well known,

but
the

most imporrequire-

ments
of minimum
vortex
drag and minimum
structure
weight
are almost
directly
opposed.
To minimize
the vortex
drag,
the wing system
must have either
large
lateral
or large vertical
dimensions
-- leading
to a heavy
structure.
If we
independent

assume
that
of details

the wing weight


depends
of the lift distribution,

only
we

on the dimensions
and
are led to the problem

is

originally
solved
by Munk
(ref. 2), that is, minimum
vortex
drag for a given
total lift and given
span.
Munk's
problem
was later extended
by Hemke
(ref. 3) to include
the effect
of vertical
tip fins.
In references
4 and 5,
Faulkner
and Darwin
determine
the distribution
of lift for wings with fins
having

minimum
In addition

drag
to

with
their

given

dimensions.

dependence

on

the

weight
and vortex
drag will depend
to a
lift or side force over the wing system.
this aspect
of the
bution
for minimum

absolute

dimensions,

certain
extent
on
In 1933,
Prandtl

the wing

the distribution
of
(ref. 6) considered

problem
and sought
to determine
the spanwise
lift distridrag with a fixed wing weight
as well as a given
total lift

and given
span.
Prandtl
assumed
that a fraction
of the weight
of the wing
structure
is proportional
to the bending
moment
integrated
over the whole
span.
For a planar
wing the average
or integrated
value
of the bending
moments
turns
out to be just the second moment
or "moment
of inertia"
of the load curve.
Prandtl's
criterion
the present
writers
root was considered.

seems more appropriate


(see ref. 7) in which

than the criterion


used by one of
only the bending
moment
at the wing

If we relax
the restriction
on the absolute
dimensions
of the wing and
consider
a family
of wings
having
the same total lift and the same integrated
bending
moment,
but varying
in span, Prandtl's
solution
shows
that the vortex
drag can be reduced
by about i0 percent
when compared
with that of the elliptic wing.
Figure
I shows one of the wing shapes
obtained
by Prandtl's
method
compared
with an elliptic
wing.
We have assumed
that each section
operates
at the same lift coefficient,
so that the chord
distribution
is proportional
to the load distribution.
It is interesting
that such a narrow
tip extension
can reduce
the vortex
drag by I0 percent.

To extend Prandtl's analysis to nonplanar wings, we have to consider not


only the spanwlse lift distribution
but also the varying inclination
of the
lift vectors along the length of the wing. Figure 2 shows a nonplanar wing
in front view and illustrates
the notation.
Positions along the span are
denoted by arc length s and angle 8 or by rectangular coordinates y and z.
Using
customary
force normal to the curved wing
surfacea will
be notation, the local "lift"

where

is

the

circulation

in potential

these definitions

jump
Wlth

(1)

= _

(2)

and

r
is the
wlng.

= vr(s)

across
the
the total

L =

wake
llft

+s t oVF

in
L

cos

the Trefftz
will be

plane

behind

the

8 ds

(3)

-s t
where

st

denotes

the

wing

tip;

the

Di =

where

is now

the

component

normal
to the curved
wing
wash at the wing is taken
For the bending
moment
at

and
be

the

bending

moment

of

+st
st

the

plane.
In
as one-half
a station

pVr[(y

integrated

drag,

DI,

induced

is

given

by

(4)

ds

"downwash"

the integration
for
the final value
in
So, Yo, Zo
we have

- Yo)COS

e +

(z - Zo)Sln

over

arc

length

the

fo st M(So)dS

of

in

the

the
the

direction

drag the downTrefftz


plane.

e]ds

the wing

(5)

spars

will

(6)

We are now in a position


to find the condition
for minimum
drag with a given
structural
constraint
such as
M.
This is, of course,
a problem
in the calculus of variations,
but since
the drag
Di
is simply
a quadratic
function
in
the space
of load distributions
it
techniques.
Assume
that the total
constraint

is

given

as well.

is not necessary
to employ
lift
L
is given
and that
Assume

also

that

a lift

sophisticated
the structural

distribution

(s)

which

satisfies

given.
figure

these

conditions

Now
consider
3.
We have

6L

three

6_ = o ;

6D i

Here

the

trace;
the

m n

they

the

are

(Prandtl's

or

second

In

the

course,

of

Munk's

and

we

the

added

_i,

42,

the

geometric

M.

of

equations

A's

we

will

drag

employed,

the

element

41 .

the

wn

drag

added
the

refer

to

by

the

original

must
for

first
but
not

hold

for

simply

the

all

general

the

equal

addition,
the

and

is

that

is,

bending
[(I/2)yl]

by

etc.,

However,
terms

to

root

produced

the
will,

because
are

equal,

the

drag

square

first-order

variations

wing
n,

becomes

interference

the

Yl,

A41,

to

the

integrated

m I

in

wing

to

the

"downwash"

is,

of

position

distributions

to

in

shape

at

if

proportional

contribute

the

moment

There
is

at

function

two

one.
this

the

element

distribution.

the

also

Similarly,

is

the

does

is

83

is

depicted

(7)

llft

lift

of

alone,

(7)

m I

drag
as

cos

involving
a

bending

41 .

theorem

only

write

then

element

the

hence

the

minimum

643

w 3
_
=

643

of

case

planar,

The
the

Since

642

functions

In

equation,

and

the_contribution,

distribution
etc.,

3 =

w 2
"_-+

is

consider

643m

642

Wl
_

mutual-interference

need

2 +

lift

contain

82

62m

641

distribution.

of

cos

I +

criterion)

third

variation,

641m

the

moment

the

of

81

wing

of

in

cos

purely

the

arm

moment

original

results
elements

641

quantity

and

moment

the

represent

structural

specified

and

discrete

of

of

variation.

all

positions

of

solution:

w
2-V = A

In
we

the
have

case

of

planar

wing

with

cos

the

(8)

Bm(s,e)

8 +

integrated

bending

moments

specified,

0=0
(9)
m

y2

hence,

w
2-V = A

and

the

Prandtl's

In
but

no

induced

downwash

must

vary

(io)

By 2

parabolically

along

the

span.

This

is

result.

case

the

additional

dimensions
structural

and

shape

constraint

of

the
is

wing

imposed,

in

front
we

have

view

are

given,

2,

2"-V= A

cos

(11)

Problems
of this type are treated
in reference
8.
For a wing with vertical
fins we find that
w = 0 (zero
"sidewash")
over the fins in agreement
with
the classical
treatments
of this problem.
The solution
of the variational
problem
does not give the load distribution
for minimum
drag directly,
but gives
instead
the variation
of downwash
in the Trefftz
plane
in terms
of the various
constraint
functions.
It is
"then necessary
to calculate
the load distribution
that is consistent
with
this

downwash

distribution

-- a standard

problem

in airfoil

theory.

Although
the theory
can be applied
to curved
wing shapes,
it is probably
sufficient
for practical
reasons
to consider
wings
composed
of straight
line
segments,
that is, a main wing panel
and fins at each tip.
Assume
at first
that the fins are vertical
so that
8 = 90
at the fins.
To

derive

the

functions

ml,

the bending
moment
all along
the
tion
Yl
on the horizontal
part

m2,

6M1

Integrating

this

over

the

within

the

semispan

fM
so that

horizontal

etc.,

(Yl

side-force

case

we

first

element
gives

%1

calculate
at

a posi-

(12)

gives

1 yI2_AI

(13)

= M = _

portion

of

For

this

- Y) 6AI

m = _

as before.
moment
is

in

span due to a lift


of the wing.
This

elements

6A 2

the wing

y2

(14)

on

the

vertical

fin

the

bending

6M = (z2 - z)_ 2

Since
the moment
at
wing spar, we have

the

base

of

the

fin

(z =

_o st _Mds--(z22

where

Yt

is

the

semispan.

0)

(15)

is

transmitted

+ YtZ2)_2

Collecting

terms

to

the

main

(16)

gives

m =_

(y2 + z2) + Yt z
4

(17)

so

that

the

downwash

for

minimum

w--=
B2V

The constants
A and B
will
fied average
bending
moment,
and the structural
criterion,

drag

A +

[i

becomes

(y2 +

the

trace

of

the

YtZ ]

(18)

depend
on the specified
total
lift
M.
Because
of the relation
between
the induced
drag can be written
D i = AL

Given

z2)+

wing

plus

and the specithe downwash

+ BM

fins

in

(19)
the

Trefftz

plane

and

the

ity component
w
everywhere
normal
to the trace,
we now have to obtain
solution
of Laplace's
equation
in two dimensions
with
w
as a boundary
dition.
The
equal
to the
is

given

in

solution
will give
circulation
around
the

the
the

velocthe
con-

potential
jump
A_
across
the wake,
wing.
The method
of solving
this problem

appendix.

Our calculations,
based
on lifting-llne
theory,
do not give details
of
chordwise
pressure
or load distribution.
It is important
to realize,
however,
that calculations
of induced
dragby
lifting-line
theory
are fully
equivalent
to those made by linear
lifting-surface
theory
and are usually
more accurate.
In cases
of wings
having
square
or blunt
tips, vortex
roll-up
will occur
at
high angles
of attack
and may lead to significant
increases
of drag over that
given by the theory.
For wings
of high aspect
ratio having
rounded
or _lliptical
tips,
such nonlinear
effects
are negligible
in the normal
flight
range.
In

the

case

of

unswept

by our formulas,
w/2, can
of twist angle
consistent
wings,
however,
additional

wings

of

high

aspect

ratio

the

downwash

obtained

be applied
at the wing and will give a distribution
with the optimum
loading.
In the case of swept
twist angles
are introduced
by the sweep,
and the

wing twist
required
to produce
the optimum
loadings
cannot
be determined
simple
lifting-line
theory.
The optimum
span load distributions
are not
altered
by sweep,
although
the appropriate
structural
criterion
may be
affected.
The

induced

drag

is

in principle

not

affected

by

the

fore

or

aft

by

posi-

tioning
of the fin or winglet,
although
the camber
and twist required
to produce the ideal
loadings
will be altered.
In the designs
proposed
by Whitcomb
(ref. i), the winglet
is given
a rearward
position
to avoid
interference
with
the

region

having

of peak

velocity

on wing.

Figure
4 shows
span load
winglets
whose
heights

distributions
are 20 percent

calculated
by our
of the semispan.

method
for wings
The load on the

winglet
is shown
projected
horizontally
beyond
the tip of the main wing.
Assuming
that each section
operates
at the same lift coefficient,
the load
curves
are proportional
to the chord length
distribution
for both wing and
winglet.
Calculations
of induced
drag,
load distribution,
and structural
parameters
remain
unchanged
if the wing system
is inverted.
The load on the
downward
projecting
fin will then, of course,
be directed
outward.

Referring to figure 4, it appears that wing A, which has 0.9 the span
of a comparable elliptic
wing (i.e., an elliptic
wing having the samelift
and the samespar weight), will have the sameinduced drag
Di

= Die

wing,
leads
of the
extreme
further

= gp/2V2b2

but
keeping
to a more
elliptic
and
the
reduction

Extending

the

wing

to

equal

the

span

the
total
spar
weight
(including
winglet
highly
tapered
loading
and
an induced
drag
wing.
With
further
Increases
of span,
the
load,
together
with
the
winglet,
tends
to
in drag.

of

the

elliptic

spar)
the
same,
that
is
0.89
that
taper
becomes
disappear
with
no

Figure
5 shows
the results
of such calculations,
for 10-percent
and
20-percent
winglets
compared
with those for flat wings,
as a function
of span
ratio,
using
the elliptic
wing as a basis.
For span ratios
less than i, the
wings
(or load curves)
become
shorter
and more blunt
and the addition
of winglets is surprisinglyeffective.
However,
the minimum
drag that can be achieved
by extending
the span while
keeping
the spar weight
constant
seems
to be about
0.89 Die , or the same as that obtained
by Prandtl
for flat wings.
Hence,
if
our criterion
of spar weight
is adopted
it appears
that the addition
of winglets cannot
reduce
the drag significantly
below
that of a flat wing having
an
ideal
tapered
planform
and a span i0 percent
greater
than the ellipse
(wing C, b/b e = i.i).
However,
the same drag value
can be achieved
without
an increase
of span by adding
15-percent
winglets
(a point
slightly
above
wing B).
For wings
having
planforms
shorter
than the ellipse
and more blunt,
the induced
drag increases
rapidly.
In this region,
however,
the benefits
of
winglets
are most pronounced.
Similar
can be
moment

calculations,

performed.
at a single

using

different

criteria

Figure
6 shows
the results
point -- the wing root -- is

that a 15-percent
reduction
of induced
drag
can be achieved
either
by a flat wing or by
no decisive
advantage
for either
type.

for

the

structure

weight,

obtained
when the bending
used.
In this case,
it appears

below
wings

that
with

of the elliptic
winglets,
again

wing
with

The foregoing
results
appear
to be in disagreement
with
those of reference 9, which
show a decisive
advantage
for winglets
if the root bending
moment
is used as a criterion.
The differences
are probably
attributable
to
our use of idealized
wing shapes,
which
result
in relatively
narrow
tip
extensions,
and the optimization
of planform
shape
for each case.

APPENDIX
I.

Introduction

We record here the more formal analysis upon which the results and discussion of the main text are based. In section II we derive the llft,
drag,
and integral bending momentexpressions for wings with winglets at 90. With
these relations, a discussion of the variational problem for minimuminduced
drag is presented. Once the downwashand sidewash are specified by the variational problem, we determine the load distribution
by considering the problem
in the Trefftz plane. This is done in section III.
Relations that lead to
the comparison with the induced drag of elliptic
wings (figs. 5, 6) are given
in section IV. Although we restrict
the present discussion to winglets at
90, it should be emphasized that the techniques sketched below may be
extended to treat winglets at any angle.
II.

MinimumInduced Drag for Wings with 90 Winglets

Weproceed by finding the lift,


drag, and integral bending momentin
terms of induced downwashand sidewash for the lifting-line
geometry shown
in figure 7. Using the Biot-Savart law, we find for the induced downwash
Vz(y) and sidewash Vy(Z)

Vz(y)

dF

(Y - Yt)
-

- Vz(X = 0,y,z = 0) = 4-_


Z% dzz [(y

i
Yt

dY'dy
y dF y'

- Yt)2

+ z2

+ Yt)+
(y +(Y Yt)2

dz
z 2-]

Yt

1
Vy(Z)

- Vy(X

= O,y

z
+_

Yt

= yt,z)

= _-_

dy
(y + yt)
2 +
d__[F

Iz' [z +

z2

' -

(z'

(z - z')
- z) 2 +

4Yt2Jaz'

dy

Yt
(AI)

Notice that the load distribution


(AI)

we

refer

to

F(y

y
integrations.
induced
drag are

= yt,z)

By symmetry
given
by

L = pV

is

for

Vy(X

a function

of

integrations,
0, y =

both

and

-yt,z)

y and
r(y,z
The

to

= -Vy(Z).

z.
In
= 0) for
lift and

(y)dy

Yt

(A2)

-Yt

D = p

To

compute

at points
moment
at

the

Yt Vz(y)r(y)dy
-Yt

integral

bending

along
the wing-winglet
an arbitrary
point
s

2p

moment,

trace.
In
is written

_(s) -- pv

fz
o

G'

[ st

we

first

vector

7) x

(A3)

Vy(Z)r(z)dz

find

the

notation,

bending

the

_(s')ds'

moment

bending

(A4)

where

st

denotes

the

wing

tip.

I+

For

the

F(y')_

for

F(z')_

for

90 winglet,

we

0 < s'

= y'

< Yt

-- Z I

--< Z

take

(A5)
F(s')

In

the

M(y)

region

= oV

0 _

fYt
Y

= xpVIJ

=_-

s = y _ Yt,

we

have

(y'-y)_x_(y')_
^

dy'+pV

(y'

- y)F(y')dy'

L-Y
Similarly,

in

the

integral

fz
o

[z'_+(yt-y)_
^

zF(z)d

- M(y)_

o
region

_(z)

The

-< S w

bending

0 _ s =

- M(z)_=

moment

z _ z,

find

_pV ._zz (z' - z)r(z')dz'

is

(s) Ids =
IO

we

M(y)dy

M(z)dz
O

^l x [-(z')gldz'
^

where
M(y)
for example,

and

M(z)

_o yt

are

given

M(y)dy

above.

= yM(y)

These

I Yt

integrals

_oYt

are

easily

evaluated;

y_ydy
dM

= YtM(Yt

) + pV

y
_oYt

= pVy t _oZ_,

A similar

calculation

M=

as

found

an

We may
extremum

F(y,z),
moment

in

for

pV

equation

so

Yt y2
2

(17)

of

now formulate
in the induced

subject
(A6) be

_o z

F(y)dy

the

%L and
their

then

main

our minimum
drag (A3)

= pV

_oYt y2
2

r(y)dy

yields

yt z +

r(z)dz

induced
drag problem.
as a functional
of the

- %L_L

the lift (A2)


require
that
-

%M_M

%M
are Lagrangian
multipliers.
variations
are simply

_L

pV

Ioz (

pV

dy
']

(A6)

text.

to the constraint
that
constant.
Formally,
we
6D

where
F(y,z),

zr(z)dz

M(z)dz

F(y')dy
[_y yt

and

We wish to find
load distribution

integral

bending

= 0

Since

(A7)

L and

are

"linear"

in

'Yt _r(y)dy
-Yt
(A8)
2

6r(y)dy

+ oV

_oZ_

(YtZ+_)_r(z)dz

-Yt
The variation
of the induced
drag is a more complicated
sidewash
and downwash
are themselves
functionals
of the
straightforward,
albeit
tedious,
calculation
yields

Vz(y)6r(y)dy
Yt

2p

calculation
since
load distribution.

z_ Vy(z)_r(z)d

the
A

(A9)

Notice that this result

is simply twice what

of
D
were
Vz(y)
and Vy(Z)
of Munk's
reciprocal
theorem
equations
(A7)-(A9),
we have

we

would

pVz(y ) _

XLPV

oV
2

-Yt

XM

r(y,z)

for

the

variation

6r(y)dy

Assuming

find

not functionals
of
r(y,z);
this is the essence
(ref. 2) for wings
with 90 wlnglets.
Combining

z2

to be piecewise

continuous,

vz(y)=

we

conclude

that

Y --A+BY2
2

2
(AI0)

Vy(Z)=

/xMv /

The parameters
A and B
of the text are
multipliers
of our variational
problem.
with equation
(18) of the main text.

Although
the above
results
readily
generalized
to winglets

III.
The

problem

wash

and

sidewash

(A!0)

i.
The problem
specifies
the

the
the

of

discontinuity
in
llft distribution.

2.

We

next

determining

map

the

ure

In
of
_

of

the

load

(A10)

is

is first formulated
stream
function
_
potential

the

the

Load

are

Distribution

distribution
attacked

in

such

three

that

the

down-

stages:

in the Trefftz
plane.
Notice
that
at the wing-winglet
surface.
We seek

across

wing-winglet

Christoffel
transformation.
and _
are reversed,
that
and
_
is now continuous.

3.
behavior
ity in

form

effectively
just the Lagrangian
Equation
(AI0) should
be compared

are for 90 winglets,


we note that they
at any angle
or to curved
wing shapes.

Determination

have

B(YtZ+

that

into

surface;

a straight

it

llne

is

proportional

with

to

a Schwarz-

Themapping
is such that the analytic
roles
of
is, _
is discontinuous
across
the wing-winglet,

the transformed
plane,
a Cauchy
_.
Mapping
back to the Trefftz
yields
the load distribution.

Our geometry
for the Trefftz
8.
Notice
that we have stood

integral
relation
plane
and taking

and transformed
the wing on its

i0

yields
the
the discontinu-

planes
is presented
side and changed
to

in figthe

notation commonlyused in the complex plane. The symmetry of the wing about
the x-axis (y = 0), both geometrically and analytically,
should also be noted.
The semispan is normalized to one (z = 0 to z = i) and the winglet has length
x (z = i to z = -x + i).
Expressed in the Trefftz
reads

or z-plane of figure 8, the constraint


Vx = A + B

Y2
2

(A10)

= _-_
8y
(All)

Vy

With

the

introduction

of

the

= B(-x

complex

W(z)
it

is

implied

by

(All)

the

lifting

line.

=-_-_x

potential

(x,y)

i_(x,y)

that

_(x,y)

on

+_)

The

= Ay + B (6_

desired

load

+ x22

distribution

(AI2)

x63 )
is

given

by

r(x,y) =_
where
line.
have

the subscripts
(+,-) refer
to the
The factor
of 2 is a normalization
r = i
at the root.

We next map
Schwarz-Christoffel

the

z-plane
into
transformation

z(_) = A

subject

(+-

to

the

the
is

(AI3)

_)

top (+) and bottom


(-) of
convenience;
an elliptic

_-plane

(see

fig.

8).

_1 (t + a)I/2(t
(t - c)+l/2(t
+ b) - i) I/2 dt
l)l/2(t

The

the lifting
wing would

appropriate

(AI4)

constraints

z(-l) = o
z(-a) = i
z(-b)

= -x

z(+c) = i
z(1) = 0

ii

(AI5)

By convention, we take the parameters A, a, b, c to be positive.


The
integral (A14) is computedusing Simpson's rule.
In the vicinity
of points
where the integrand is singular, the following procedure is used. Consider
the integral
'_ (tg(t)
+ i)= dt

z(_) -where

g(-l)

is

finite

-l

and

lim [g(t)- _(-I)]


t+-l t (t+l)= ] = 0
This

latter

0 < _ < i;
We evaluate

relation

will

generally

in particular,
z(_) with

we

hold

could

for

consider

[g(t)

any
wings

- g(-i)]

transformation
with

where

winglets

at

any

angle.

dt

(t + i)_

i
+--_ g(-l)_

The remaining
larity
having

linear

integrals
may then be
been removed
from the

The parameters
equations

A,

a, b,

and

_
-I+_

(AI6)

g(t)
(t + i) _ dt

evaluated
using
Simpson's
regions
of integration.

Re

are

z(-b)

Im z(-b)

found

by

numerically

rule,

the

solving

singu-

the

non-

= -x A
=

i
(A17)

Re

z(1)

= 0

Im

z(1)

= 0

The resulting
mappings
were in general
accurate
to about one part in a million.
A further
check
was provided
by comparison
with the results
of Faulkner
and
Darwin
(ref. 4) who evaluated
(Al4) using
elliptic
functions.
As

noted

earlier,

-i _ _ E i
is given
_(_) then yields

the
by

discontinuity

_(_,0).

of

simple

=--

_(_)

dt

12

across

application

the
of

_ : 0

Cauchy's

line

for

theorem

(AI8)

to

where
_(_) and _(t) indicate
values
of
# and _
on the
n = 0 line.
This
expression
is valid
up to the usual polynomial
ambiguity.
We define
two
potentials
corresponding
to the Lagrangian
multipliers
A and B (cf. (AI0)
and (AI2))

_A(_)

= _

_B(_)

: P-_
2_

(AI9)

_i-

-i

2(t)

In (AI9) and (A20),


the values
the potential
_A
corresponds
constraint;
this is the problem
winglets
constant

at 90 .
integral

x3(t)

+ _

(t)

(A20)

of
x and y
are given
by (AI4).
Notice
that
to minimum
drag with only the constant
lift
solved
by Faulkner
and Darwin
(ref. 4) for

The potential
bending
moment.

_B

reflects

the

additional

constraint

The numerical
evaluation
of the principal
valued
integrals
(eqs.
(A20))
is reasonably
straightforward.
The behavior
of
y(_) and x($)
20-percent
winglet
is shown
in figure
9.
Besides
subtracting
off the
mic variation
due to the principal
value,
for example,

#A(_)

we also
behavior

for

For
_A

used
of

a method
y(_).

unit constant
is given by

ifl
: _
-I

[y(t) - y(_)]
t -

similar

to that

downwash

in

CA(_)

of

the

dt +

(AI6)

Trefftz

Re

z(_)

y(_)

to handle

plane,

the

1-

of

(AI9),
for a
logarith-

"cusp

points"

analytic

in

the

solution

- A_

(A21)

where
A
is the mapping
parameter
in (AI4).
In figure
l0 we compare
this
result
with our numerical
evaluation
of
_A"
Notice
that the abscissa
corresponds
to the wing and winglet
as denoted
by
"y" and "x" (see, also,
fig. 7).
The two curves
agree up to an offset
of -0.041.
This value
is just the limit
of (A2I) as
_ _ _.
Recall
that the Cauchy
relation
(AI8) must be modified
by

adding

a polynomial

Pn(_)

Thus,
the offset
in figure
this constant
will have no
tribution.
its absolute

The behavior
of
value
is offset

when

lO(a) is expected
effect
on thevalue
cB
as

is
for

shown
_A"

13

in

when
of
figure

we use (AI8).
Notice
that
the corresponding
load dis10(b).

We

suspect

that

The

load

distributions

are

found

from

_A

and

_B

with

1
FA

=_

($A+

- iA -)

1
rB =2" (B+where

the

subscripts

(+,-)

denote

values

CB-)
of

above

and

below

the

abscissa

in figure
i0.
These
load distributions
are shown
in figure
ii.
In practice,
only discrete
points
for
_A and _B, not continuous
curves,
are obtained.
The calculation
of
rA and F B thus requires
the interpolation
of values
for
_A/ and _B
at corresponding
abscissa
points.

IV.

Induced-Drag

Comparison

We may now derive


the expressions
that give the
Di/Die
shown
in figures
5 and 6.
By requiring
that
have the same llft and integral
bending
moment
as an
for the parameters
A and B (recall
(AI0))

[MB - -_- _--_-) J/(LAMB

L-HA
the

factor

_/4

corresponds

to

the

elliptic

quantities
in section

L A,
III-

wing
L B,

LA

with

M A,

the

MB

semlspan

are

be/2

simply

A(Y)dy

moments

LB =

normalized

to

y = i.

of

rB

as

FA and

fl

rA(Y)

y2
2

dy +

FB(Y)

y2
2

dy +

14

The

determined

FB(Y)dy
o

MB

LBMA)

(A22)

(i - y2) I/2

f
=

LBMA)

taking

MA

]/(LAMB -

+ "-'_-_'_')

Fe(Y)
for

induced-drag
ratio
the wing with winglets
elliptic
wing,
we find

(x
+
+

(A23)

rA(x) dx 1
rE(x) dx

Once
tions

rA and
of

the

rB are known, the parameters


ratio
be/b.
The induced-drag

Die

In deriving
(A22)-(A24),
integral
bending
moment

= _

[A2LA

A and
ratio

B
may be considered
is then given by

AB(L A + M B)

we note that lift


scales
as
b 4.

and

drag

func-

(A24)

+ B2MB]

scale

as

b 2, whereas

the

In the case of a straight


wing,
the formalism
discussed
in this appendix
was used to solve
the Prandtl
problem
(ref. 6) as well as that posed
in reference
7.
All the calculations
can be done in "closed
form";
they yielded
the same results
as in references
6 and 7.
To check
our accuracy,
this exercise was also done using
the numerical
techniques
described
above
for evaluating the various
integrals
involved.
The two methods
of calculation
agreed
to
better
than a tenth
of a percent.

15

REFERENCES

le

Whitcomb,
Richard
T.:
A Design
Approach
at High Subsonic
Speeds
for Wing Tip
1976.

and Selected
Wind
Mounted
Winglets.

Tunnel
Results
NASA TN D-8260

Munk,
Max M.:
The Minimum
(Reprinted
in -- Jones,
NASA RP-I050,
1979.)
Paul

E.:

Hemke,

4.

Faulkner,
A.R.C.

C.:
The Design
of Minimum
R. & M. No. 2279,
1945.

Drag

Tip

5.

Faulkner,
V. M.; and Darwin,
C.:
The Design
of Minimum
Br. A.R.C.
Technical
Report,
R. & M. No. 2280,
1945.

Drag

Inboard

6.

Prandtl,

Widerstandes.

V. M.; and Darwin,


Technical
Report,

L.:

Zeltschrlft
(Reprinted
Gesammelte
Jones,

of Wings

Robert

Uber

TragflHgel

T.:

The

Induced
Drag of Wings
vol. 24, Dec. 1973.)

Cone,
of

Clarence
Nonplanar

des

End

Plates.

Kleinsten

NACA

D., Jr.:
Lifting

Spanwlse

Distribution

TR-267,

Induzierten

f_r Flugtechnlk
und Motorluftschiffahrt
in -- Tollmlen,
W.; Schlichtlng,
H.; and
Abhandlungen.
Sprlnger-Verlag,
1961.)

Drag of Wings
Having
TN-2249,
1950.
(See

8.

with

TR-121,
1921.
Theory.

3.

7.

Drag

Induced
Drag of Aerofoils.
NACA
R. T., ed.:
Classical
Aerodynamics

of

Lift

24 Jg.
Gortler,

for

a Given
Lift and a Given
Bending
also Klein,
A.; and Viswanathan,
with

Given

Theory
Systems.

of

Lift

and

Root

Induced
Lift
NASA TR-RI39,

Heyson,
H. H.; Riebe,
G. D.; and Fulton,
C. L.:
Study
of the Relative
Advantages
of Winglets
NASA TM X-74003,
1977.

16

1927.
Fins.

Fins.

1933.
H., eds:

Minimum

Induced

Moment.
S. P.:

NACA
Minimum

Bending

Moment.

and Minimum
1962.
Theoretical
and Wing-Tip

Induced

ZAMP,

Drag

Parametric
Extensions.

ELLIPTIC

WING
L2

Die lrq b2
I

Figure

.2

I.- Planar wings

.4

having

.6

WING C (FIG. 5)

.8

equal spar weight


criterion.

17

1.0

according

1.2

to Prandtl's

Figure

2.- Coordinates

18

o_ nonplanar

wing.

_3

"

_1
f

_1 cos 01+ _2

8;=_1ml

Wl

+ 6i2 m2

__1_ +_2

Figure

cos 02 + _3

w2
2V

+ _'_3 m3

_,o,w3

+v_3

3.- Variational

19

cos 03 = 0

2V

= 0

_
-

problem.

WING A
D i = 1.0 Die

Figure

4.- Load curves

20

for wings with wlnglets.

1.4

1.2

_ /

\\

FLAT WINGS
10%WINGLETS

L2
_'q oe

1.0

6-

WING_

.8

.6

.4

e_
w

.2--

I
.8

.9

1.0

1.1

1.2

1.3

1.4

SPAN RATIO b/b e

Figure

5.- Induced

drag of wlngs

having

21

the same integrated

moment

M.

FLAT WINGS
1"4 F

1 2_

__

/IO%WINGLETS

__20%

WINGLETS

,_'1.o1-- \\
_

.8-

,6--

"_

.4-

.2--

I
0.8

.9

1.0

1.1

1.2

_
1.3

]
1.4

SPAN RATIO b/b e

Figure

6.- Induced

drag of wings

having
root.

22

the same bending

moment, at the wlng

.....

t
z_
Y
0

-Yt

Figure

7.- Lifting-line

geometry

for wing

+Yt

with 90 winglets

23

(x = 0 plane).

11

-a

c
z=x+iy

+1

_'=_+i_

x
-1

Figure

-a

-b

+1

8., Transformation
of wing-winglet in the Trefftz plane into a straight
llne; corresponding points are labeled
-i, -a, -b, c, +i.

24

-a

1.0
.8
.6

mz(_)

.4
.2
0

-1.0

Figure
c(x=

9.-

Behavior

=
0.2);
0.2739.

the

of
mapping

-.5

.5

1.0

Im

z(_)

constants

and
are

x
A

25

=
=

Re

z(_)

1.1195,

for
a

a
=

20-percent

0.7110,

winglet
=

0.4924

(a)
_A

4_

:4

L
-1.2

I_'_
0

.2

I
.4

I
.6

i
.8

I
1.0

I
1.2

.3
_B

(b)

.2

.1

-.1

-.2

-.3

I
.2

IFigure

I
.4

I
.6

I
.8

,y,,

i0.-

Potential

I
1.0
-I ....
-r:

distributions

for wing

with

I
1.2
x

,_,J
_"I'

winglets.

(a) Comparison
of numerical
evaluation
of
_A
with the analytic
solution
a 20-percent
winglet;
"y" denotes wing and
"x" denotes winglet.
The
constant
offset of 0.041 is just the limit of
z(_) - A_
as
_ + _.
(b) The

potential

_B

corresponding

to the integral

26

bending

moment

for

constraint.

1.2

1.0-

.6

.4

--

.2

.2
'_

.4

.6

.8

"Y"

1.0
_I_

1.2
"x"-_

Figure ii.- Load distributons


for 20-percent winglets; rA corresponds
constant downwash, rB to the integral bending moment constraint.

27

to

You might also like