JOURNAL OF AIRCRAFT

Vol. 37, No. 4, July– August 2000

Modern Adaptation of Prandtl’s Classic Lifting-Line Theory

W. F. Phillips¤ and D. O. Snyder†

Utah State University, Logan, Utah 84322-4130

The classical solution to Prandtl’s well-known lifting-line theory applies only to a single lifting surface with no
sweep and no dihedral. However, Prandtl’s original model of a Ž nite lifting surface has much broader applicability.
A general numerical lifting-line method based on Prandtl’s model is presented. Whereas classical lifting-line theory
is based on applying the two-dimensional Kutta– Joukowski law to a three-dimensional  ow, the present method
is based on a fully three-dimensional vortex lifting law. The method can be used for systems of lifting surfaces
with arbitrary camber, sweep, and dihedral. The accuracy realized from this method is shown to be comparable to
that obtained from numerical panel methods and inviscid computational  uid dynamics solutions, but at a small
fraction of the computational cost.
Downloaded by UNIVERSITY OF ARIZONA on July 1, 2014 | [Link] | DOI: 10.2514/2.2649

Nomenclature V1 = magnitude of V1
Ar = global reference area V = local  uid velocity
b = twice the lifting surface semispan V1 = velocity of the uniform  ow or freestream
C `i = section lift coefŽ cient for wing section i =  uid volume
C `a = airfoil section lift slope vi j = dimensionless velocity induced at control point j by
C `a i = section lift slope for wing section i vortex i , having a unit strength
C `1 = section lift coefŽ cient for an inŽ nite wing v1 = unit vector in direction of freestream
C mi = section moment coefŽ cient for wing section i a i = local angle of attack for wing section i
c = local section chord length a L 0i = local zero-lift angle of attack with no  ap de ection
c̄ = overall aerodynamic mean chord length C = vortex strength in the direction of r0
c̄i = aerodynamic mean chord length for wing section i C i = strength of horseshoe vortex i
d Ai = differential planform area at control point i D G = dimensionless strength correction vector
dF = differential aerodynamic force vector d Ai = planform area of wing section i
d` = directed differential vortex length vector d i =  ap de ection for wing section i
F = net force exerted by  uid on the surroundings d `i = spatial vector along the bound segment i
f di = local section induced drag per unit span d Mi = quarter-chord moment for wing section i
f` = local section lift per unit span e i =  ap effectiveness for wing section i
G = dimensionless vortex strength vector ³i = dimensionless spanwise length vector
Gi = dimensionless vortex strength for section i h = angle from r1 to r2
[ J] = N by N matrix of partial derivatives q =  uid density
Lr = global reference length X = relaxation factor
M = net moment about the c.g. exerted by the  uid ! = local  uid vorticity
N = total number of horseshoe vortices
n = number of horseshoe vortices per semispan
R = residual vector Introduction
Ra
ri
r i1 j
=
=
=
aspect ratio
vector from c.g. to control point i
vector from node i 1 to control j
T HE developmentof Prandtl’s lifting-linetheory,1,2 providedthe
Ž rst analyticalmethod for accuratelypredictinglift and induced
drag on a Ž nite lifting surface. In this theory, Prandtl hypothesized
r i2 j = vector from node i 2 to control j that each spanwise section of a Ž nite wing has a section lift equiva-
r 1 , r2 = magnitudes of r1 and r2 lent to that acting on a similar section of an inŽ nite wing having the
r0 = vector from beginning to end of vortex segment same section [Link] this hypothesis,the two-dimensional
r1 = vector from beginning of vortex segment to arbitrary vortex lifting law of Kutta3 and Joukowski4 was applied at each sec-
point in space tion of the three-dimensionalwing, to relate the local aerodynamic
r2 = vector from end of vortex segment to arbitrary point in force to the local circulation. However, to Ž x the direction of the
space aerodynamicforce vector, the undisturbedfreestream velocity in the
s = spanwise coordinate Kutta– Joukowski law was, intuitively and without proof, replaced
uai = chordwise unit vector at control point i with the vector sum of the freestream velocity and the velocity in-
uni = normal unit vector at control point i duced by the trailing vortex sheet. This theory gives good agreement
usi = spanwise unit vector at control point i with experimental data for straight wings of aspect ratio greater than
u1 = unit vector in direction of the freestream about four. Prandtl’s lifting-line theory has had a profound impact
on the development of modern aerodynamics and hydrodynamics
and is still widely used today. However, conventional lifting-line
Received 8 October 1999; presented as Paper 00-0653 at the AIAA 38th theory applies only to a single lifting surface with no sweep and no
Aerospace Sciences Meeting and Exhibit (Oct), Reno, NV, 12 – 14 January dihedral.
2000; revision received 2 February 2000; accepted for publication 10 Febru-
In most modern textbooks (e.g., Bertin and Smith5 ), Prandtl’s hy-
ary 2000. Copyright ° c 2000 by the American Institute of Aeronautics and
Astronautics, Inc. All rights reserved. pothesis is justiŽ ed based on the provision that  ow in the spanwise
¤ Professor, Mechanical and Aerospace Engineering Department. Member direction is small. The failure of conventional lifting-line theory to
AIAA. accurately predict the aerodynamic forces acting on a swept wing
† Graduate Student, Mechanical and Aerospace Engineering Department. is usually blamed on the violation of this provision. However, it
Member AIAA. can be shown from the three-dimensionalvortex lifting law that the
662
 PHILLIPS AND SNYDER 663

relationship between section lift and section circulation is not af- Glauert.7 The most popular method, based on Gaussian quadrature,
fected by  ow parallel to the bound vorticity (see Saffman6 ). The was originally presented by Multhopp.8 Most recently, Rasmussen
three-dimensional vortex lifting law requires that, for any volume and Smith9 have presented a more rigorous and more rapidly con-
enclosed by a stream surface in an inviscid, incompressible, steady verging method, based on a Fourier series expansion similar to that
 ow, a force must be exertedon the surroundingsequal to the product Ž rst used by Lotz10 and Karamcheti.11
of the  uid density and the cross product of the local  uid velocity Purely numerical methods for solving the lifting-line equation
with the local  uid vorticity, integrated over the volume. Saffman6 have also been proposed. McCormick12 has presented a numerical
presents a proof of this vortex lifting law. method that can be used for a single lifting surface having a straight
Only in the limiting case of two-dimensional potential  ow can lifting line. This method is based on applying the two-dimensional
the local  uid velocity vector in the vortex lifting law be replaced Kutta– Joukowski law to the three-dimensional ow and neglects the
with the freestream velocity. For this special case, the general vor- downwash generated by the bound vorticity. Results obtained from
tex lifting law reduces to the two-dimensional Kutta– Joukowski this method are essentiallyidenticalto those obtainedfrom the series
law. Strictly speaking, the two-dimensional Kutta– Joukowski law solution. A numerical lifting-line method has also been developed
cannot be used to relate section lift to section circulation in a three- by Anderson et al.13 that relaxes the assumptionof linearitybetween
dimensional potential  ow. However, the three-dimensional vortex section lift and section angle of attack. For a single straight lifting
lifting law, applied to Prandtl’s model of the Ž nite wing, requires a surface, this method gives good agreement with experimental data
local section lift equal to the cross product of the local  uid veloc- at angles of attack both below and above stall. However, the method
ity vector with the local circulation vector, multiplied by the  uid still assumes a straight lifting line and ignores the downwash pro-
density, q (V £ C ). This cross product results in a section lift that is duced by the bound vorticity. Thus, as is the case with all methods
independentof the component of  uid velocity that is parallel to the used to obtain solutions to the classical lifting-line equation, this
bound vorticity. numerical method applies only to a single lifting surface with no
Downloaded by UNIVERSITY OF ARIZONA on July 1, 2014 | [Link] | DOI: 10.2514/2.2649

If the  uid  ow component parallel to the bound vorticity has sweep and no dihedral.
no effect on the section lift, why does conventional lifting-line the- Here, a numerical lifting-line method is presented that can be
ory fail to predict the performance of swept wings? The answer is used to obtain the forces and moments acting on a system of lift-
also provided by the general vortex lifting law. When computing ing surfaces with arbitrary position and orientation. This method,
section lift from the general vortex lifting law, the local velocity based on Prandtl’s original model of a Ž nite wing, accurately pre-
induced on each vortex segment must include the velocity induced dicts the effects of both sweep and dihedral as well as the effects of
by the remainder of that same vortex as well as that induced by aspect ratio, camber, and planform [Link] obtainedfrom this
all other potentials included within the  owŽ eld. This means that, method are compared with experimental data and with results from
when computing the lift on any wing section, we must include the other numerical methods. The accuracy realized from the present
velocity induced by all other vortex segments, free or bound, that method is shown to be comparable to that obtained from numerical
are contained within the  owŽ eld. In the development of classical panel methods and inviscid computational  uid dynamics (CFD)
lifting-line theory, Prandtl intuitively added the velocity induced solutions, but at a small fraction of the computational cost. In ad-
by the trailing vortex sheet to the undisturbed freestream velocity dition to the obvious applications to aeronautics, this method has
speciŽ ed by the two-dimensional Kutta– Joukowski law. However, broad application to the Ž eld of hydrodynamics, including hydro-
he did not suggest including the velocity induced by one bound foils, marine propellers, and control surfaces. Unlike the classical
vortex segment on another. lifting-line solution, the present method is not based on a linear re-
For  ow over a straight lifting surface, the bound vortex Ž laments lationship between section lift and section angle of attack. Thus,
are all reasonably parallel. For any two parallel vortex Ž laments, the the method could conceivably be applied, with caution, to account
force resulting from the velocity induced on the Ž rst Ž lament by the approximately for the effects of stall.
second is equal, opposite,and collinearwith the force resultingfrom
the velocity induced on the second Ž lament by the Ž rst. Thus, for Formulation
straight lifting surfaces, the parallel nature of the bound vorticity
In what is commonly referred to as the numerical lifting-line
makes it reasonable to neglect the interaction between bound vortex
method (e.g., Katz and Plotkin14 ), a Ž nite wing is synthesizedusing
Ž laments and compute the section lift based only on the velocity
a composite of horseshoe shaped vortices. The continuous distribu-
induced by the trailing vortex sheet and the undisturbed uniform
tion of bound vorticity over the surface of the wing, as well as the
 ow.
continuous distribution of free vorticity in the trailing vortex sheet,
For  ow over a swept wing, the bound vortex Ž laments on each
is approximated by a Ž nite number of discrete horseshoe vortices,
side of the wing are roughly parallel to each other and to the local
as shown in Fig. 1.
wing quarter chord. However, the bound vortex Ž laments on one
The bound portion of each horseshoe vortex is placed coincident
side of the wing are not parallel to the bound vortex Ž laments on
with the wing quarter-chord line and is, thus, aligned with the local
the other side. Thus, for a lifting swept wing, the bound vorticity
sweep and dihedral. The trailing portion of each horseshoe vortex
generated on one side of the wing produces downwash on the other
is aligned with the trailing vortex sheet. The left-hand corner of one
side of the wing. This downwash reducesthe net lift and increasesthe
total induced drag for the wing. The downwash resulting from the
bound vorticity is greatest near the center of the wing, whereas
the downwash resultingfrom the trailingvorticityis greatestnear the
wing tips. Thus, for a swept wing, the lift is reduced both near the
center of the wing and near the tips.
Prandtl’s classical lifting-line theory is based on a linear relation-
ship between section lift and section angle of attack. With this linear
assumption, and with the assumption of a straight lifting line, the
theory provides an analytical solution for the spanwise distribution
of lift and induced drag acting on a Ž nite lifting surface. The solution
is in the form of an inŽ nite sine series for the circulation distribu-
tion. Historically, the coefŽ cients in this sine series have usually
been evaluated from collocation methods. Typically, the series is
truncated to a Ž nite series, and the coefŽ cients in the Ž nite series
are evaluated by requiring the lifting-line equation to be satisŽ ed
at a number of spanwise locations equal to the number of terms Fig. 1 Horseshoe vortices distributed along the quarter chord of a
in the series. A very straightforward method was Ž rst presented by Ž nite wing with sweep and dihedral.
 664 PHILLIPS AND SNYDER

induced at an arbitrary point in space, by a complete horseshoe

vortex, is

C u1 £ r 2 (r + r )(r £ r ) u1 £ r1
V= + 1 2 1 2 ¡
4p r2 (r2 ¡ u1 ¢ r2 ) r 1r 2 (r 1 r2 + r1¢ r2 ) r1 (r1 ¡ u1 ¢ r1 )
(3)

As is the case with panel methods, the user must specify the orien-
tation of the trailing vortex sheet. In obtaining the classical lifting-
line solution for a single lifting surface with no sweep or dihedral,
Prandtl assumed the trailing vortex sheet to be aligned with the wing
Fig. 2 Position vectors describing the geometry for a horseshoe vortex. chord. This was done to facilitate obtaining an analytic solution. In
obtaining a numerical solution, there is little advantage in aligning
the trailing vortex sheet with a vehicle axis such as the chord line.
horseshoe and the right-hand corner of the next are placed on the More correctly, the trailing vortex sheet should be aligned with the
same nodal point. Thus, except at the wing tips, each trailing vortex freestream. This is done easily in the numerical solution by setting
segment is coincident with another trailing segment from the adja- u1 equal to the unit vector in the direction of the freestream. Al-
cent vortex. If two adjacent vortices have exactly the same strength, though it is intuitively more appealing to align the trailing vortex
then the two coincidenttrailingsegments exactly cancelbecause one sheet with the freestream, in reality, this makes very little difference
has clockwise rotation and the other has counterclockwiserotation. in the Ž nal result. For typical wings, aligning the trailing vortex
sheet with the chord line rather than the freestream produces errors
Downloaded by UNIVERSITY OF ARIZONA on July 1, 2014 | [Link] | DOI: 10.2514/2.2649

The net vorticity that is shed from the wing at any internal node is
simply the difference in the vorticity of the two adjacent vortices in the resulting forces and moments of less than 1%. Still, because
that share that node. the method allows the trailing vortex sheet to be aligned easily with
Each horseshoe vortex is composed of three straight vortex seg- the freestream, this should always be done.
ments. From the Biot – Savart law and the nomenclature deŽ ned in When a system of lifting surfaces is synthesized using N horse-
Fig. 2, the velocity vector induced at an arbitrary point in space, by shoe vortices,in a manner similar to that shown in Fig. 1, Eq. (3) can
any straight vortex segment, is readily found to be, for example, see be used to determine the resultant velocity induced at any point in
Bertin and Smith5 or Katz and Plotkin,15 space, if the strength of each horseshoe vortex is known. However,
these strengths are not known a priori. To compute the strengths of
the N vortices, we must have a system of N equations relating these
C r1 £ r2 r1 r2 N strengthsto some known propertiesof the wing. In what has com-
V = r0 ¢ ¡ (1)
4p j r1 £ r2 j 2 r2 r2 monly been referred to as the numerical lifting-line method,14 these
N equations are provided by forcing a Neumann condition, which
speciŽ es zero normal velocity at the three-quarterchord of the wing
Although Eq. (1) is in the form commonly found in modern text- section midway between the trailing legs of each horseshoe vortex.
books, it is not the most useful form for numerical [Link] This method works remarkably well for planar swept wings with no
induced velocity computed from Eq. (1) is indeterminate whenever camber. However, not surprisingly, this method does not work well
r1 and r2 are collinear, even for points that lie outside the vortex for cambered wings or for wings including de ected  aps and/or
segment. To eliminate this division by zero for points that are not on control surfaces.
the vortex segment, we can make use of the trigonometric relations In reality, this numerical lifting-line method is simply the vortex
lattice method16,17 applied using only a single lattice element, in the
chordwise direction, for each spanwise subdivision of the wing.
r0 = r1 ¡ r2 , r1 ¢ r2 = r1 r2 cosh , j r1 £ r2 j = r1 r2 sin h
Applying the Neumann condition at only one point in the chordwise
direction is clearly not adequate for wing sections with camber or
Using these relations we have  ap de ection. This method gives a result that depends only on the
position and slope of the camber line at the three-quarterchord. The
r0 r1 r2 r1 ¡ r2 r1 r2 predicted performance is completely independent of camber line
¢ ¡ = ¢ ¡ shape at any other chordwise location. This is clearly not realistic.
j r1 £ r2 j 2 r1 r2 j r1 £ r2 j 2 r1 r2 For a more pragmatic approach, we turn to the general three-
dimensional vortex lifting law,6
1 r1 ¢ r2 r1 ¢ r2
= r1 + r 2 ¡ ¡
r12 r22 sin2 h r2 r1
F= q (V £ !) d
(r1 + r 2 )(1 ¡ cosh ) r1 + r2
= 2 2 = 2 2
r 1 r 2 (1 ¡ cos2 h ) r1 r2 (1 + cosh )
Using Prandtl’s hypothesis, we assume that each spanwise wing
r 1 + r2 section has a section lift equivalentto that acting on a similar section
= of an inŽ nite wing with the same localangleof [Link], applying
r1 r2 (r1 r2 + r1 ¢ r2 )
the vortex lifting law to a differential segment of the lifting line, we
have
and Eq. (1) can be more conveniently written as
dF = q C V £ d` (4)
C (r 1 + r 2 )(r1 £ r2 )
V = (2) If  ow over a Ž nite lifting surface is synthesizedfrom a uniform  ow
4p r1 r2 (r1 r2 + r1 ¢ r2 )
combined with horseshoe vortices placed along the quarter-chord
line, from Eq. (3), the local velocityinduced at a control point placed
Notice that, unlike the result from Eq. (1), the induced velocity anywhere along the bound segment of horseshoe vortex j is
computed from Eq. (2) is not singular when the angle from r1 to r2
is zero. It is, however, still singular when this angle is §p . N
C i vi j
When we use Eq. (2) for the Ž nite bound segment and the two V j = V1 + (5)
semi-inŽ nite trailing segments shown in Fig. 2, the velocity vector c̄i
i =1
 PHILLIPS AND SNYDER 665

where vi j is the dimensionless induced velocity where

u1 £ ri 2 j V1 d`i C i
v1 ´ , ³i ´ c̄i , Gi ´
ri 2 j ri 2 j ¡ u1 ¢ ri 2 j V1 d Ai c̄i V1
c̄i ri 1 j + ri 2 j ri 1 j £ ri 2 j and, from Eqs. (5) and (9), the local angle of attack written in terms
+ , i 6= j of the dimensionless variables is given by
4p r i1 j ri 2 j ri 1 j ri 2 j + ri 1 j ¢ ri 2 j
u1 £ ri 1 j N
vi j ´ ¡ v1 + v ji G j ¢ uni
r i1 j r i1 j ¡ u 1 ¢ r i 1 j ¡ 1
j =1
a i = tan (12)
u1 £ ri 2 j v1 + N
v ji G j ¢ uai
j =1
c̄i ri 2 j ri 2 j ¡ u1 ¢ ri 2 j
, i = j (6)
4p u1 £ ri 1 j Equation (11) can be written for N different control points, one
¡
r i1 j r i1 j ¡ u 1 ¢ r i 1 j associated with each of the N horseshoe vortices used to synthesize
the lifting surface or system of lifting surfaces. This provides a sys-
At this point, c̄i could be any characteristiclength associatedwith tem of N nonlinearequationsrelatingthe N unknown dimensionless
the wing section aligned with horseshoevortex i . This characteristic vortex strengths G i to known properties of the wing. At angles of
length is simply used to nondimensionalizeEq. (6) and has no effect attack below stall, this system of nonlinear equations surrenders
on the induced velocity. The choice of characteristic length will be quickly to Newton’s method.
addressed later. The bound vortex segment is excluded from Eq. (6), To apply Newton’s method, the system of equations is written in
when i = j , because a straight vortex segmentinduces no downwash the vector form:
Downloaded by UNIVERSITY OF ARIZONA on July 1, 2014 | [Link] | DOI: 10.2514/2.2649

along its own length.

From Eqs. (4) and (5), the aerodynamicforce acting on a spanwise (G) = R (13)
differential section of the lifting surface located at control point i is
given by where
N
N
C j i (G) =2 v1 + v ji G j £ ³ i G i ¡ C `i (a i , d i ) (14)
dFi = q C i V1 + v ji £ d`i (7)
c̄ j j =1
j =1

When we allow for the possibilityof  ap de ection, the local section We wish to Ž nd the vector of dimensionless vortex strengths G that
lift coefŽ cient for the airfoil section located at control point i is a makes all components of the residual vector R go to zero. Thus, we
function of local angle of attack and local  ap de ection, want the change in the residual vector to be ¡ R. We start with an
initial estimate for the G vector and iteratively reŽ ne the estimate
C `i = C `i (a i , d i ) (8) by applying the Newton corrector equation
The local angle of attack at control point i is [J ]D G = ¡ R (15)

1 Vi ¢ uni where [J ] is the N by N matrix of partial derivatives

a i = tan ¡ (9)
Vi ¢ uai
2wi ¢ (v j i £ ³ i )
where uai and uni are, respectively,the unit vectors in the chordwise Gi
j wi j
direction and the direction normal to the chord, both in the plane of , j 6= i
the local airfoil section as shown in Fig. 3. If the relation implied @C `i v ai (v ji ¢ uni ) ¡ v ni (v ji ¢ uai )
by Eq. (8) is known at each section of the wing, the magnitude of ¡
@a i
2
v ai + v ni2
the aerodynamic force acting on a spanwise differential section of @ i
Ji j = =
the wing located at control point i can be written as @G j 2wi ¢ (v ji £ ³ i )
2j wi j + Gi
j dFi j = 1
q 2
V1 C `i (a i , d i ) d Ai (10) j wi j
2
, j =i
@C `i v ai (v ji ¢ uni ) ¡ v ni (v j i ¢ uai )
Setting the magnitudeof the force from Eq. (7) equal to that obtained ¡
from Eq. (10) and rearranging, we can write @a i
2
v ai + v ni2
N (16)
2 v1 + v ji G j £ ³ i G i ¡ C `i (a i , d i ) = 0 (11) N
j =1 wi ´ v1 + v ji G j £ ³i (17)
j =1

N
v ni ´ v1 + v j i G j ¢ uni (18)
j =1

and
N
v ai ´ v1 + v ji G j ¢ uai (19)
j =1

Using Eq. (16) in Eq. (15), we compute the correction vector D G.

This correction vector is used to obtain an improved estimate for
the dimensionless vortex strength vector G according to
Fig. 3 Unit vectors describing the orientation of the local airfoil sec-
tion. G =G +X D G (20)
 666 PHILLIPS AND SNYDER

This process is repeated until the magnitude of the largest residual The aerodynamic moment generated about the center of gravity
is less than some convergence criteria. For angles of attack below is
stall, this method converges very rapidly using almost any initial
estimate for G and a relaxation factor X of unity. At angles of attack N N
C iC j
beyond stall, the method must be highly under relaxed and is very M =q ri £ C i V1 + vji £ d ` i + d Mi (27)
sensitive to the initial estimate for G. c̄ j
i =1 j =1
For the fastest possible convergence of Newton’s method, we
require an accurate initial estimate for the dimensionless vortex If we assume a constant section moment coefŽ cient C mi over each
strength vector. For this purpose, a linearized version of Eq. (11) is spanwise increment, then
useful. For a straight lifting surface of inŽ nite aspect ratio at small
angles of attack, the downwash is zero, the section lift is a linear s2
1
functionof angle of attack,and all nonlinearterms in Eq. (11) vanish. d Mi = ¡ q V 2 Cm i c2 ds usi (28)
For a lifting surface of high aspect ratio with no sweep or dihedral, 2 1 s = s1
at small angles of attack, we can still ignore the nonlinear terms and
compute an approximate dimensionlessvortex strength vector from where usi is the local spanwise unit vector shown in Fig. 3,
this linear system.
At small angles of attack, the local section lift coefŽ cient can be usi = uai £ uni (29)
approximated as
C `i (a i , d i ) » = C `a i (a i ¡ a L 0i + e id i) (21) When we use Eq. (28) in Eq. (27) and nondimensionalize
N N
Using the small angle approximation for both the geometric angle M
Downloaded by UNIVERSITY OF ARIZONA on July 1, 2014 | [Link] | DOI: 10.2514/2.2649

of attack and the induced angle of attack, after applying Eq. (12) to 1
= 2ri £ G i v1 + G i G j v ji £ ³i
q2 V1 2 Ar L r
Eq. (21), we have i =1 j =1

N s2
C mi d Ai
C `i » = C `a i v1 ¢ uni + v j i ¢ uni G j ¡ a L0i + e id i (22) ¡ c2 ds usi (30)
d Ai s = s1
Ar L r
j =1

Applying Eq. (22) to Eq. (11) and ignoring second-order terms, we To this point, the local characteristic length c̄i has not been de-
obtain the linear system Ž ned. It could be any characteristiclength associated with the span-
wise incrementof the liftingsurfacethatis associatedwith horseshoe
N vortex i . From Eq. (30), we see that the natural choice for this lo-
2j v1 £ ³ i j G i ¡ C `a i v ji ¢ uni G j = C `a i (v1 ¢ uni ¡ a L 0i + e id i) cal characteristic length is the integral of the chord length squared,
j =1 with respect to the spanwise coordinate, divided by the incremental
(23) area. If we also assume a linear variation in chord length over each
spanwise increment, we have
Equation (23) gives good results, at small angles of attack, for a
s2
single lifting surface of high aspect ratio with no sweep or dihedral. ci 1 + ci 2
For larger angles of attack, highly swept wings, or for interacting d Ai = c ds = si 2 ¡ si 1 (31)
s = s1 2
systems of lifting surfaces, the nonlinear system given by Eq. (11)
should be used. However, Eq. (23) provides a reasonable initial
estimate for the dimensionless vortex strength vector, to be used and
with Newton’s method for obtaining a solution to this nonlinear
2 ci 1 + ci1 ci 2 + ci 2
s2 2 2
system. 1
c̄i ´ c2 ds = (32)
d Ai s = s1 3 ci 1 + ci 2
Aerodynamic Forces and Moments
Once the vortex strengths have been determined, the total aero- When we use these deŽ nitions in Eq. (30), the dimensionless aero-
dynamic force vector can be determined from Eq. (7). If the lifting dynamic moment about the c.g. is
surface or surfacesare synthesizedfrom a large number of horseshoe
vortices, each covering a small spanwise increment of one lifting N N
M
surface, we can approximate the aerodynamic force as being con- 1
= 2ri £ G i v1 + G i G j v ji £ ³i
stant over each spanwise increment. Then, from Eq. (7), the total 2
q V1 2 Ar L r i =1 j =1
aerodynamic force is given by
N N d Ai
C iC j ¡ C mi c̄i usi (33)
F =q C i V1 + v ji £ d `i (24) Ar L r
c̄ j
i =1 j =1
Once the N dimensionlessvortex strengths G i are known, Eqs. (25)
where d `i is the spatialvector along the bound segment of horseshoe and (33) are used to evaluate the aerodynamic forces and moments.
vortex i from node 1 to node 2, in the direction of segment vorticity. Like panel methods, lifting-line theory provides only a potential
When we nondimensionalize Eq. (24), the total nondimensional  ow solution. Thus, the forces and moments computed from this
aerodynamic force is method do not include viscous effects, so that the parasitic drag is
[Link] addition to this restriction,that also applies to panel
N N
F d Ai methods, lifting-line theory imposes an additional restriction that
=2 G i v1 + G i G j v ji £ ³i (25) does not apply to panel [Link] lifting surfaces with low aspect
1
q2 V1 2 Ar Ar
i =1 j =1 ratio, Prandtl’s hypothesis breaks down, and the usual relationship
between local section lift and local section angle of attack no longer
where d A i is the planform area of segment i , applies. It has long been established that lifting-line theory gives
s2 good agreement with experimental data for lifting surfaces of aspect
d Ai = c ds (26) ratio greater than about four.2 For lifting surfaces of lower aspect
s = s1 ratio, panel methods or CFD solutions should be used.
 PHILLIPS AND SNYDER 667

Fig. 4 Lifting-line grid with cosine clustering and 20 elements per

semispan.

Grid Generation and Control Points

Each lifting surface must, of course, be divided into spanwise
elements, in a manner similar to that shown symbolically in Fig. 1.
In Fig. 1, the wing is divided into elements of equal spanwise in-
crement. However, this is not the most efŽ cient way in which to
grid a lifting surface. Because vorticity is shed from a lifting sur-
face more rapidly in the region near the tips, the nodal points should
be clustered more tightly in this region for best computational efŽ -
Fig. 5 Section lift distribution for three elliptic wings.
ciency. The authors have found conventionalcosine clustering to be
quite efŽ cient. For straight lifting surfaces,clusteringis only needed
near the tips and the cosine distribution can be applied across the
entire span. However, for a lifting surface with sweep and/or dihe-
Downloaded by UNIVERSITY OF ARIZONA on July 1, 2014 | [Link] | DOI: 10.2514/2.2649

dral, there is a step change in the slope of the quarter chord at the
root. This step change causes the downwash to change very rapidly
in the region near the root. Thus, in general, the authors recom-
mend applying cosine clustering independentlyover each semispan
of each lifting surface, as shown in Fig. 4. This clusters the nodes
more tightly at both the tip and the root. This clustering is based on
the change of variables,
s / b = [1 ¡ cos(u ) ]/ 4 (34)

Over each semispan, u varies from zero to p as s varies from zero to

b / 2. Distributing the nodes uniformly in u will provide the desired
clustering in s. If the total number of horseshoe elements desired on
each semispan is n, the spanwise nodal coordinates are computed
from,
si / b = 14 [1 ¡ cos(i p / n) ], 0 ·i ·n (35) Fig. 6 Section induced drag distribution for three elliptic wings.

where the bound segment of horseshoe vortex i extends from node This same solution gives the local section induced drag
i to node i ¡ 1 on a left semispan and from node i ¡ 1 to node i
on a right semispan. The authors have found that using this nodal f di 4Ra 1
distribution with about 40 horseshoe elements per semispan gives 1
= 2
[1 ¡ (2s / b) 2 ] 2 (38)
q V1 2 c̄C`12 p Ra + C `a
the best compromise between speed and accuracy. Figure 4 shows 2

a system of lifting surfaces overlaid with a grid of this type using

20 elements per semispan. As a Ž rst-order test, the numerical method presented here should
Because singularities occur at the junctures of adjacent bound- agree with Eqs. (37) and (38), at small angles of attack, for a straight
vortex segments, for maximum accuracy and computational efŽ - elliptic wing with no geometric or aerodynamic twist. Figure 5
ciency, the location of control points must be critically assessed. At shows a comparison between the section lift distribution predicted
Ž rst thought, it would seem most reasonable to place control points by Eq. (37) and that predicted by the present numerical method
on the bound segment of each vortex, midway between the two trail- for three straight elliptic wings of different aspect ratio. Figure 6
ing legs. However, the authors have found that this does not give the compares the induced drag distribution predicted by Eq. (38) to
best numerical accuracy. A signiŽ cant improvementin accuracy,for that predicted from the present method for the same three elliptic
a given number of elements, can be achieved by placing the control wings. A similar comparison for tapered wings shows very similar
points midway in u rather than midway in s. Thus, the spanwise results. The numerical results shown in Figs. 5 and 6 were all gen-
control point coordinates are computed from erated using 40 horseshoe elements per semispan for wings having
a NACA 2412 airfoil section. Similar comparisons were made for
si / b = 14 {1 ¡ cos[(i p / n) ¡ (p / 2n)]}, 1 ·i ·n (36) other straight wings having varying amounts of camber, taper, and
washout. Camber was varied from 0 to 8%, taper ratio was varied
This distributionplaces control points very near the spatial midpoint from 0.1 to 1.0, and washout was varied from 0 to 5 deg. In all cases,
of each bound vortex segment, over most of the wing. However, near the numerical solution using 40 horseshoe elements per semispan
the root and the tip, these control points are signiŽ cantly offset from agreed with the classical lifting-line solution to within two-tenths
the spatial midpoint. of 1% for the induced drag and to within Ž ve-hundredths of 1% for
the lift. Thus, for all practical purposes, results obtained from this
Results numerical lifting-line method are identical to those obtained from
For straight elliptic wings with no geometric twist and no aero- the classical lifting-line solution for a single lifting surface with
dynamic twist, Prandtl’s classical lifting-line theory gives a very no sweep or dihedral. Because, for straight wings of aspect ratios
simple and well-known closed-form solution. From this solution, greater than about four, the classicalsolution is known to adequately
the section lift distribution in the spanwise direction is given by predict the effects of aspect ratio, camber, and planform shape, the
same can be said for the present numerical solution.
f` 4Ra 1
= [1 ¡ (2s / b) 2 ] 2 (37) Unlike the analytical solution to Prandtl’s classical lifting-line
1
2
q V1 2 c̄C`1 p Ra + C `a theory, the present numerical method can be applied to wings with
 668 PHILLIPS AND SNYDER

sweep and/or dihedral. To examine how well the present method

predicts the effects of sweep and dihedral, results obtained from this
method were compared with resultsobtainedfrom a numerical panel
method and from an inviscid CFD [Link] results predictedfor
the effects of sweep were also compared with limited experimental
data.
For the numericalpanel method comparison,the commercialcode
PMARC 18,19 was [Link] code was developedat NASA Ames
Research Center and is one of the most efŽ cient numerical panel
codes available. PMARC uses  at quadrilateral panels with uni-
form source and doublet distributions on each panel. This code also
accounts for the effects of wake rollup, using an unsteady wake
development approach.
The commercial code WIND,20 developed by the NPARC Al-
liance, was chosen for the inviscid CFD comparison. This code uses
a node-centered Ž nite volume approach to solve the Euler equa-
tions on a structured grid. The available Ž fth-order upwind-biased
discretization scheme was used to reduce the effects of numerical Fig. 8 Comparison between the induced drag coefŽ cient predicted by
viscosity. the numerical lifting-line method, PMARC, and WIND with that ob-
tained from experimental data, for a straight wing and a wing with
For all three numerical methods the grid size was repeatedly re-
45 deg of sweep.
Ž ned until the solutionwas no longersigniŽ cantly affectedby further
Downloaded by UNIVERSITY OF ARIZONA on July 1, 2014 | [Link] | DOI: 10.2514/2.2649

grid reŽ nement. By this procedure, an optimum grid for each of the
three methods was selected as that which gave the best combina- From the results shown in Fig. 7, we see that the lift coefŽ cient
tion of accuracy and computationalspeed. With this method of grid predicted by all three methods is in good agreement with experi-
resolution, the numerical lifting-line method required 40 section el- mental observations for both wings. From Fig. 8 we see that, for
ements per semispan. Grid-resolved solutions for PMARC required the straight wing, the induced drag predicted by both the numeri-
4500 panels per semispan. The selected grid was partitioned to give cal lifting-line method and by PMARC is in good agreement with
45 segments along the semispan and 90 around the circumference, experimental data, whereas WIND gives an induced drag that is
with panelsclusterednear the leadingedge, the trailing edge, and the somewhat higher. The predictions for induced drag on the swept
wingtips. For WIND, a two-block H – H type mesh with 1.25 £ 106 wing are not nearly as good. For this wing, the induced drag pre-
grid points per semispan gave the best compromise between accu- dicted by the panel code is about 40% less than that observed exper-
racy and computational efŽ ciency. The computational domain for imentally. However, both the numerical lifting-line method and the
the WIND solutions extended approximately 30 chords from the CFD solution give induced drag values that are about 25% above
wing in all directions, with grid points clustered near the wing and the experimental values. Still, the values of induced drag predicted
the trailing-tip vortices. When these resolved grids were used for by the numerical lifting-line method are as good as or better than
the three methods, the computationaltime required to obtain a solu- those predicted by the other two methods, even for this highly swept
tion from PMARC was about 2 £ 104 times that required using the wing. Furthermore, the induced drag on a highly swept wing, as pre-
presentnumericallifting-linemethod. The WIND solutionsrequired dicted by the present method, appears to be somewhat conservative,
approximately 2 £ 106 times as long as the lifting-line solutions. whereas the panel code predicts an induced drag that is too low.
A comparison between results obtained from these three numer- Though none of the numerical methods tested seemed to do very
ical methods and previously published experimental data is shown well at predicting the effects of sweep on induced drag, all three of
in Figs. 7 and 8. The solid lines and Ž lled symbols correspond to the methods appeared to do a good job of predicting the effects of
a straight wing of aspect ratio 6.57, with experimental data ob- sweep on lift.
tained from McAlister and Takahashi.21 The dashed lines and open To obtain the numerical lifting-linesolutionsshown in Figs. 7 and
symbols are for a 45-deg swept wing of aspect ratio 5.0, having ex- 8, we have used the linear relationship between a two-dimensional
perimental data reported by Weber and Brebner.22 Both wings have section lift coefŽ cient and a section angle of attack that is predicted
symmetrical airfoil sections with no geometric twist and constant by thin airfoil theory. Because the downwash and induced angle of
chord throughoutthe span. The straight wing has a thicknessof 15% attack are obtained from the solution to the nonlinear system given
and the swept wing has a thickness of 12%. by Eq. (11), we have no reason to expect that the lift coefŽ cient
predicted for the complete wing should be a linear function of geo-
metric angle of attack. However, the results in Fig. 7 show that this
relationship is, in fact, very nearly linear for both wings. The result
predicted for the straight wing is almost exactly linear, as is also
predicted by the classical lifting-line solution. Even for the highly
swept wing, the deviation from linearity is quite small. Both the
panel code and the experimental data conŽ rm this result.
The overpredictionof induced drag by the CFD code for both the
straight wing and the swept wing is likely due to numerical viscosity
and is expected. However, the reason that the numerical lifting-line
method overpredictsinduced drag for the swept wing, but not for the
straight wing, may be less obvious. The foundation of lifting-line
theory requires the bound vorticity to follow the chordwise aerody-
namic center of the wing. The numerical lifting-line solution shown
in Fig. 8 was obtained by assuming that the lifting-line follows the
wing quarter chord, which is the theoretical aerodynamic center
from thin airfoil theory. With this assumption, the discontinuity in
the slope of the quarter chord at the spanwise midpoint of a highly
swept wing produces a rather strong singularity (see, for example,
Fig. 7 Comparison between the lift coefŽ cient predicted by the numer- Cheng and Meng23 ). This results in the prediction of strong down-
ical lifting-line method, PMARC, and WIND with that obtained from wash and large induced drag in the region near this singularity. In
experimental data, for a straight wing and a wing with 45 deg of sweep. reality, the experimental results of Weber and Brebner22 show that,
 PHILLIPS AND SNYDER 669

Fig. 9 Deviation of the aerodynamic center from the quarter chord in

the region near the spanwise midpoint of a highly swept wing.
Fig. 11 Pressure forces and shed vorticity for a system of lifting sur-
faces as predicted by the numerical lifting-line method.

are combined and forced to satisfy Eq. (11) as a single system of

coupled equations, the interactions between lifting surfaces are di-
rectly accounted for. An example of such interactionscan be seen in
Fig. 11. The lifting surfaces shown in Fig. 11 all have symmetric air-
foil sectionsand are orientedto give a zero geometricangle of attack.
Downloaded by UNIVERSITY OF ARIZONA on July 1, 2014 | [Link] | DOI: 10.2514/2.2649

The only direct production of lift in this conŽ guration comes from
a de ection of the ailerons. The de ected ailerons produce lift and
vorticity that in turn induce lift and vorticity on all other [Link]
Fig. 11 the arrows indicate the magnitude and direction of the local
section force, and the streamwise lines at the trailing edge of each
lifting surface indicate the magnitude of the local shed vorticity. As
can be seen from Fig. 11, the lifting surface interactions predicted
by the present method are at least qualitatively correct. However,
further study is needed to determine the quantitativeaccuracy of the
predicted lifting surface interactions.
Fig. 10 Comparison between the rolling moment coefŽ cient predicted The present numerical method contains no inherent requirement
by the numerical lifting-line method, PMARC, and WIND for wings for a linear relationshipbetween section lift and section angle of at-
with 10 and 20 deg of dihedral. tack. Thus, the method could conceivably be applied, with caution,
to account approximately for the effects of stall. The lifting-line
method requires a known relationship for the section lift coefŽ cient
near the spanwise midpoint of a highly swept wing, the aerodynamic as a function of section angle of attack. Because such relationships
center moves considerably aft of the wing quarter chord, as shown must be obtained experimentally beyond stall, the present method
in Fig. 9. This removes the discontinuity in slope at the spanwise would predict stall by using a semi-empirical correction to an other-
midpoint of the lifting line and signiŽ cantly reduces the predicted wise potential  ow solution. For this reason, the method should be
induced drag. Unfortunately, there is currently no simple means to used with extreme caution for angles of attack beyond stall. How-
predict the true chordwise aerodynamic center of a highly swept ever, the method may be able to predict the onset of stall. Further
wing in the region near the spanwise midpoint. Thus, the present study is needed to determine to what extent the method can be used
method cannot easily be corrected to remove this error. at angles of attack near or beyond stall.
The accuracy of the present numerical lifting-line method for
predictingdihedral effectswas also tested by comparing this method
with both PMARC and WIND. Figure 10 shows the variation in
Conclusions
rolling moment coefŽ cient with sideslip angle for two nonplanar The insight of Ludwig Prandtl (1875– 1953) was nothing short of
rectangularwings having differentdihedral,as predictedby all three [Link] was never more dramatically demonstratedthan in
methods. Both wings have symmetric NACA 0015 airfoil sections the developmentof his classicallifting-linetheory, during the period
with no sweep or twist and an aspect ratio of 6.57. The solid line and from 1911– 1918. The utility of this simple and elegant theory is so
Ž lled symbols correspond to a wing with 20 deg dihedral, whereas great that it is still widely used today. Furthermore, with a few minor
the dashedline and open symbolsare for a wing with 10 deg dihedral. alterations and the use of a modern computer, the model proposed
From Fig. 10, we see that all three methods agree very closely in by Prandtl can be used to predict the inviscid forces and moments
their predictions for the wing with 10 deg dihedral. For this wing acting on lifting surfaces of aspect ratio greater than about four with
the agreement between the lifting-line method and the panel code is an accuracy as good as that obtained from modern panel codes or
within 2%, and the CFD solutionagrees with the lifting-linesolution CFD, but at a small fraction of the computational cost.
to within 7%. The three methods do, however, diverge somewhat in
their predictionsfor the wing with 20-deg dihedral. This divergence References
becomes more pronounced at very large sideslip angles. For the 1 Prandtl, L., “Trag ügel Theorie,” Nachrichten von der Gesellschaft der
wing with 20-deg dihedral at sideslip angles in the range of 12 deg, Wisseschaften zu Göttingen, Gesch äeftliche Mitteilungen, Klasse, 1918,
the panel code predicts a rolling moment coefŽ cient that is about 9% pp. 451– 477.
2 Prandtl, L., “Applications of Modern Hydrodynamics to Aeronautics,”
above that predicted by the lifting-line method, whereas the CFD
solution predicts a result that is nearly 10% below the lifting-line NACA 116, June 1921.
3 Kutta, M. W., “Auftriebskräfte in Strömenden Flüssigkeiten,” Illustrierte
solution.
Aeronautische Mitteilungen, Vol. 6, 1902, p. 133.
The present numerical method could be used to predict the aero- 4 Joukowski, N. E., “Sur les Tourbillons Adjionts,” Traraux de la Section
dynamic forces and moments acting on a system of lifting surfaces Physique de la Societe Imperiale des Amis des Sciences Naturales, Vol. 13,
with arbitrary position and orientation. Each lifting surface would No. 2, 1906.
be synthesized by distributing horseshoe vortices on a grid struc- 5 Bertin, J. J., and Smith, M. L., “Incompressible Flow About Wings of
tured similarly to that shown in Fig. 4. Because all of the horseshoe Finite Span,” Aerodynamics for Engineers, 3rd ed., Prentice– Hall, Upper
vortices used to synthesize the complete system of lifting surfaces Saddle River, NJ, 1998, pp. 261– 336.
 670 PHILLIPS AND SNYDER

6 Saffman, P. G., “Vortex Force and Bound Vorticity,” Vortex Dynamics, Hill, New York, 1991, pp. 291– 294.
Cambridge Univ. Press, Cambridge, England, U.K., 1992, pp. 46 – 48. 16 Falkner, V. M., “The Calculation of Aerodynamic Loading on Sur-
7 Glauert, H., The Elements of Aerofoil and Airscrew Theory, 2nd ed., faces of Any Shape,” Reports and Memoranda 1910, Aeronautical Research
Cambridge Univ. Press, Cambridge, England, U.K., 1959, pp. 142– 145. Council, London, Aug. 1943.
8 Multhopp, H., “Die Berechnung der Auftriebs Verteilung von Trag- 17 Multhopp, H., “Method for Calculating the Lift Distribution of Wings

 ugeln,” Luftfahrtforschung, Vol. 15, No. 14, 1938, pp. 153– 169. (Subsonic Lifting Surface Theory),” Reports and Memoranda 2884, Aero-
9 Rasmussen, M. L., and Smith, D. E., “Lifting-Line Theory for Arbitrary nautical Research Council, London, Jan. 1950.
Shaped Wings,” Journal of Aircraft, Vol. 36, No. 2, 1999, pp. 340– 348. 18 Katz, J., and Maskew, B., “Unsteady Low-Speed Aerodynamics Model
10 Lotz, I., “Berechnung der Auftriebsverteilug beliebig geformter Flugel,” for Complete Aircraft ConŽ gurations,” Journal of Aircraft, Vol. 25, No. 4,
Zeitschrift für Flugtechnik und Motorluftschiffahrt, Vol. 22, No. 7, 1931, 1988, pp. 302– 310.
pp. 189– 195. 19 Ashby, D. L., Dudley, M. R., and Iguchi, S. K., “Development and
11 Karamcheti, K., “Elements of Finite Wing Theory,” Ideal-Fluid Aero- Validation of an Advanced Low-Order Panel Method,” NASA TN-101024,
dynamics, Wiley, New York, 1966, pp. 535– 567. Oct. 1988.
12 McCormick, B. W., “The Lifting Line Model,” Aerodynamics, Aeronau- 20 Bush, R. H., Power, G. D., and Towne, C. E., “WIND: The Production

tics and Flight Mechanics, 2nd ed., Wiley, New York, 1995, pp. 112– 119. Flow Solver of the NPARC Alliance,” AIAA Paper 98-0935, Jan. 1998.
13 Anderson, J. D., Jr., Corda, S., and Van Wie, D. M., “Numerical Lifting- 21 McAlister, K. W., and Takahashi, R. K., “NACA 0015 Wing Pressure

Line Theory Applied to Drooped Leading-Edge Wings Below and Above and Trailing Vortex Measurements,” NASA TP-3151, Nov. 1991.
Stall,” Journal of Aircraft, Vol. 17, No. 12, 1980, pp. 898– 904. 22 Weber, J., and Brebner, G. G., “Low-Speed Tests on 45-deg Swept-Back
14 Katz, J., and Plotkin, A., “Lifting-Line Solution by Horseshoe Ele- Wings, Part I: Pressure Measurements on Wings of Aspect Ratio 5,” Reports
ments,” Low-Speed Aerodynamics, from Wing Theory to Panel Methods, and Memoranda 2882, Aeronautical Research Council, London, May 1958.
23
McGraw – Hill, New York, 1991, pp. 379– 386. Cheng, H. K., and Meng, S. Y., “The Oblique Wing as a Lifting-Line
15 Katz, J., and Plotkin, A., “Constant-Strength Vortex Line Segment,” Problem in Transonic Flow,” Journal of Fluid Mechanics, Vol. 97, No. 3,
Low-Speed Aerodynamics, from Wing Theory to Panel Methods, McGraw – 1980, pp. 531– 556.
Downloaded by UNIVERSITY OF ARIZONA on July 1, 2014 | [Link] | DOI: 10.2514/2.2649