1008 J. AIRCRAFT VOL. 22, NO.

11

Doublet Strip Method for Oscillating

Swept Tapered Wings in Incompressible Flow
A. Ichikawa*
Civil Aviation College, Miyazaki, Japan

An improved numerical method is developed for calculating the load distributions on oscillating swept tapered
wings in incompressible flow. The integration domain is transformed into a rectangular domain, and the domain
is divided into many chordwise strips. In the strip containing the control point, the proposed method properly
accounts for Cauchy and logarithmic singularities. The solutions generally compared well with other lifting-
Downloaded by UNIVERSITY OF CALIFORNIA - DAVIS on February 11, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.45239

surface theories, but with much smaller computational times.

Nomenclature k,l = chordwise and spanwise loading points,

= aspect ratio respectively
A,B = seeEq. (12) or (52) LE,TE = leading and trailing edges, respectively, of wing
B = K/eikxo (~ ) = complex amplitude or coordinate on left wing
b = wing span panel
c(y) = wing semichord
= wing root semichord
= wing root chord, =2c0 Introduction
CT = wing-tip chord A ERODYNAMIC characteristics of oscillating wings at
Q = sectional lift coefficient ./^subsonic speeds are generally calculated by doublet-lattice
CL = wing lift coefficient or lifting-surface methods. Doublet-lattice methods1 are
= lifting pressure coefficient, simpler and easier to apply to complex configurations than
= (p--p+)/(pU2/2) lifting-surface methods, but they require a relatively large
= influence coefficient of RS, see Eqs. (10) and (54) number of unknowns ("boxes") for convergence.
K = kernel function, see Eq. (3) To improve the convergence, the author has proposed an
Vk = influence coefficient of SS, see Eq. (50) improved numerical method (doublet strip method2) for rec-
k = reduced frequency, =coc0/(7 tangular wings in incompressible flow. A key point of the
MLE = tanALE method is to use an expansion series of the kernel3 together
mM with proper treatment of the chordwise logarithmic singular-
= tanATE ity.4 Unlike conventional doublet-lattice methods, the pro-
= tanATE tanAM posed method uses many chordwise strips, and spanwise in-
NC = number of chordwise loading points tegration is performed first to avoid Mangier 's principal
NS = number of spanwise loading points (semispan) values.
RS = regular strip (strip without control point) The purpose of the present paper is to extend the idea of the
SS = singular strip (strip with control point) doublet strip method to involve swept tapered wings. The in-
s = tip inset, see Eq. (8) tegration domain is transformed into a rectangular domain in
U = freestream velocity order to apply a technique that has been developed previously
w = down wash velocity, positive upward in Ref. 2. The solutions generally compared well with other
= downwash velocity induced by left and right wing lifting-surface theories, but with much shorter computational
panels, respectively times.
x,y = Cartesian coordinates, see Fig. 1
xM(y) = chordwise coordinate of mean chord line Doublet Strip Method
- x x j The coordinate system is illustrated in Fig. 1, where
=y-y\ trapezoidal geometry is defined by
= nondimensional chordwise variable
(9 = cos- 1 (-D X = C/C CR = 2b/&(l + X)
X = wing taper ratio, = cr/cR
ALE = leading-edge sweep, deg
AM mean chord line sweep, deg
ATE = trailing-edge sweep, deg A/H = tanA T E -tanA O T =2(X-l)Afc(X+l) (1)
P = air density
The downwash equation in incompressible flow is
Subscripts and Superscripts
ij = chordwise and spanwise control points,
respectively
U Sir J-6/2
(2)
Received July 2, 1984; revision received June 6, 1985. Copyright
American Institute of Aeronautics and Astronautics, Inc., 1985. All
rights reserved. (3)
'Associate Professor. Member AIAA.
 NOVEMBER 1985 DOUBLET STRIP METHOD 1009

A swept tapered domain can be transformed into a rectangular The spanwise integration of Eq. (10) is performed by the
domain by introducing a new variable , where same method as in the Ref . 1 which uses a parabolic approx-
imation. It must be noted that JCG in Eq. (10) becomes
= (x-xM(y)}/c(y) - (4)

is the nondimensional chordwise variable.

Since the forms of xM(y) and c(y) on the left wing panel (12)
are different from those on the right wing panel, it is conven- by transforming the independent variable from x,y to ,y.
ient to treat each wing panel separately. The total downwash
at
( y) on the swept wing is composed of the contribution Singular Strip
from both wing panels and is defined by Since the kernel function in the singular strip has several
y\ 1 r b/2 r> 1
singularities, careful treatment is necessary. Using
K(x^y^k)=e-ikxoB(k, \yQ\,xQ)=e-ik^y^B^B(k, l y 0 U o ) >
Uf =
~87T~ JO J-l the singular strip term of Eq. (7) becomes
Downloaded by UNIVERSITY OF CALIFORNIA - DAVIS on February 11, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.45239

(5)
where
x e~ikBCp(^,yj)c(yj)d^ (13)
c(yl) = - (6)
and the j>0-integration portion of the above equation is
Note that yl and j^ are the spanwise integration points on the
right and left wing panels, respectively. Due to symmetry, con- (14)
trol points are needed on the right wing panel only. / =/,+/,

Downwash Contribution from the Right Wing Panel B(k,\y0\,Xo)dy0 (15)

The wing panel is divided into many chordwise strips by us-
ing uniform spacing with inset.5 The strip containing the con-
trol point is named the singular strip (SS) and the other strips (16)
are named regular strips (RS). It is assumed that the con-
tinuous pressure distribution over the wing is replaced with the /! and 72 are now evaluated.
one that is constant in the spanwise direction on each strip. A. Il: Using an expansion series of the kernel,3
Then, the first integration of Eq. (5) can be written as becomes
(17)
+I
Crf) =T--ET'
\ U / jj O7T ^j Jo/ f
J "

2
+ 1 p f 1 Kte
07T J>'/ ~?i J - 1

(7)
where ' implies the exception of the singular strip, and
27 2w + :
(18)

yl(j)=b(\-s)/2.(2I(j)-\)/2NS, l,j=l,29...,NS

s=l/4NS (8) Y ! (j^\ 2ne2{n +l+e22^

(19)
;~n (n+\)(n!)2 \2/ 2+l
Regular Strip
Since the kernel function in the regular strip contains no
singularity; the treatment of the integral is straightforward. 70 ~ C/3 are given as
Using the Gauss formula, the first integral of Eq. (7) is
discretized to be U0=-[l/el-(l+Rl/B) + l/e2-(l+R2/B)] (20)
NS NC
(21)
/T! ' X
(9)
4(2NC+1)
U2= -k2/2- [AJl+el^i(Rl-Ael-B)
**/+
(10)
+ e2&i(R2+Ae2-B)-e{-e2] (22)
where the positions of the control points and loading points (23)
are
where
/=1,2,...,7VC

/b= 1,2,..., JVC

(11)
respectively. (24)
 1010 A. ICHIKAWA J. AIRCRAFT

IF = 1M42 , I (A2 (43)

(25)
Since /^ is the same form as the term appearing in Ref. 2, it
can be discretized accurately in the same manner. The result is
(26)
27T
Cpkjc(yj)sinOk (44)
27VC+1
7VC-1
/2 - l/2(A2 + l ) 2 - ( [ ( A 2 + l)el- 3AB]Rl + [ (A2 + 1 )e2 l)6k]Fr(di) (45)

+ 3AB]R2+B2(2A2-1)IF] (28)
(r = 0)
The70 integrations of the series Um in Eqs. (18) and (19) are
performed analytically with a parabolic approximation for the
integrand.
As Eq. (21) has a chordwise logarithmic singularity, it is The remaining term IR can be reduced to a finite sum
Downloaded by UNIVERSITY OF CALIFORNIA - DAVIS on February 11, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.45239

possible to write BI = BI + 2k/^jA2 + 1 &, l 0 I . Thus, Eq. (17) through the Stark6 formula because the integrand has a
can be written as Cauchy singularity and other regular terms. Thus,

/! - BR (29)
(47)
B. 72: The term 72 can be evaluated approximately by ex-
panding the integrand in powers of yQ, retaining terms up to
G(yQ), and carrying out the^ 0 integration, which yields (48)

72 - CR (30) Substituting Eqs. (44) and (47) into Eq. (41),

1 NC
(31)
[SS] = (49)
4(2NC+\)

-e-ikB/( 1 (50)
(32)
Regular Strip Plus Singular Strip
where Combining Eqs. (9) and (49), the downwash contribution
from the right wing panel is
-Rl +Ae{ I
NS NC

(33)
(51)
(34)
where 6/. is the Kronecker delta.
(35)
Downwash Contribution from the Left Wing Panel
The downwash contribution from the left wing panel is
given by the last term of Eq. (5). On the left wing panel,
(36)
x0=Ay0
= -KJ2 (37)
A =-(A7
(38)
B = (52)
Since Eq. (33) has a chordwise logarithmic singularity, one
can write C7 = C7 4- 2^ 2 /V^ 2 + 1 &, If 0 I . Thus, Eq. (30)
becomes

I2 = CR + iCj + 2/fc4 2 A/>4 2 + l-&i l? 0 I (39)

C. /j +/ 2 : Substituting Eqs. (29) and (39) into Eq. (14),

Fig. 1 Geometry and
2 coordinate system.
1= BR + CR+ i ( B j + C,) + 2/Wy4 + 1 -^1^1 (40)

Using this expression, the term SS becomes

[SS] - I/STT- [IR +2/^V^42Tr-e-/A:c(^4] (41)

C p (f 1 o^.)c(>v)d{i (42)
 NOVEMBER 1985 DOUBLET STRIP METHOD 1011

Since only the control points on the right wing panel are used, of 0.5, a reduced frequency of 1.0, and downwash
all of the strips on the left wing panel have no singularity. Ac- w/U= -1.0. In this figure, the number of loading points is
cordingly, the last term of Eq. (5) can be discretized in the 6x 10 and the dots on the pressure curves represent the solu-
same manner as a regular strip. The result is tion of Eq. (56).
Figure 3 compares the spanload distribution between the
i present method and the doublet-lattice method (DLM) of Ref.
4(27VC+1) E E (53) 1 for the same wing as in Fig. 1. With as few as 4 x 5 loading
points, the distribution from the present method compares
/._ f rf/4
:/ (54)
very well with that from the DLM using 10x40 loading
points.
~L, Figures 4 and 5 show CL convergence of the present
method. It is to be noted that the method greatly improves the
(55)

Formulation for the Entire Swept Wing

Combining Eqs. (51) and (53), the downwash equation for REAL
the entire swept wing is
Downloaded by UNIVERSITY OF CALIFORNIA - DAVIS on February 11, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.45239

NS

/=!
NC NS
(56) A,A PRESENT A 5
o.cr PRESENT A 10
rrrr DLM1 10 40
where Cpkl-= Cpkl and c(yl-) = c(y() are used because the wing
is symmetrical. 0. .5 1.5
Fig. 3 Comparison of spanload distribution between present method
Numerical Results and doublet lattice method (ALE = 45 deg, ^R = 2, A = 0.5, & = 1.0,
The solution of Eq. (56) yields the pressure distribution on w/U= -1.0).
the right wing panel. The sectional lift coefficient and the total
lift coefficient are given as
3.0
1j
(57) <
LJ
cr
27T
NS NC \d
I] c(^) X) ^/sin^ (58) 2.5 k=0.
Cod NS(2NC+\)

Figure 2 shows the pressure distribution for a wing with a

45-deg leading-edge sweep, an aspect ratio of 2, a taper ratio
2.0-

~l k=2.

0 10 NS(semi-span) 50
Fig. 4 Effect of the number of loading points on CL (real part).

3.8

3.5
v
NC =6 MS =10
CLR=2.1538
CLI=1.5218

3.0

k=i.
1.5-

0 10 NS(semi-span) 50

Fig. 2 Example of pressure distribution on swept tapered wing Fig. 5 Effect of the number of loading points on CL (imaginary
fA LE - 45 deg, jfc = 2, A = 0.5, k = 1.0, w/U= -1.0,7VC = 6, NS= W). part).
 1012 A. ICHIKAWA J. AIRCRAFT

2.8
AR=2 k = 1.0
*-=0.5
2.5 w/U=~lO I
pGC
_j
LU
\r-*
i3; ' pSs
2.3
^ <s>9o
-f^
2.0
O A D X ~^""T~"~
NC A 6 8 10 jyeo0
2.0

/r^
Downloaded by UNIVERSITY OF CALIFORNIA - DAVIS on February 11, 2015 | http://arc.aiaa.org | DOI: 10.2514/3.45239

i i i i i
I!1III
1 50 100 NCxNS 500
0 0 NS(semi-span) 50

Fig. 6 Effect of leading-edge sweep on CL (real) convergence, Fig. 9 Comparison of computational time between present method
and doublet-lattice method (with FACOM M-382 computer).

3.0| A = 0.5 convergence for moderate reduced frequency as well as for

^J k=1.0 relatively high reduced frequency.
LU It can be seen that the method with NC = 4 and NS= 10
results in a converged solution at = 2.0.
Figures 6-8 show the effect _of wing geometry on con-
2.5 vergence. Only the real part of CL is given in these figures. It
should be noted that the method gives excellent convergence in
any case.
2.6 Figure 9 compares the computational time between the pres-
ent method and the DLM. The examples computed herein,
were performed on a FACOM-M382 computer in Nagoya
2.3 University. It can be seen that the computational time for the
present method is almost the same as for the DLM with the
same number of unknowns.
2.3

Conclusions
2.0
An improved numerical method (doublet strip method) for
calculating the lift distributions on unsteady tapered wings in
0 10 NS(semi-span) 50 incompressible flow has been developed. For convenience in
the analysis, the swept tapered wing is transformed into a rec-
Fig. 7 Effect of aspect ratio on CL (real) convergence. tangular domain. In the strip containing the control point, the
kernel function is treated in expansion series form and many
singularities are properly accounted for.
The present method shows excellent convergence of solution
for moderate reduced frequency as well as for relatively high
reduced frequency. The computational time for the method is
almost the same as for the doublet-lattice method with the
same number of unknowns.

References
, E. and Rodden, W. P., "A Doublet Lattice Method for
Calculating Lift Distributions on Oscillating Surface in Subsonic
Flow," AIAA Journal, Vol. 7, Feb. 1969, pp. 279-285.
2
Ichikawa, A. and Ando, S., "Improved Numerical Method for
Unsteady Lifting Surfaces in Incompressible Flow," Journal of Air-
craft, Vol. 20, July 1983, pp. 612-616.
3
Ueda, T., "Expansion Series of Integral Functions Occurring in
Unsteady Aerodynamics," Journal of Aircraft, Vol. 19, April 1982,
pp. 345-347.
4
Ando, S. and Ichikawa, A., 'The Use of an Error Index to Im-
prove Numerical Solutions for Unsteady Lifting Airfoils," AIAA
2.0- Journal, Vol. 21, Jan. 1983, pp. 47-54.
5
Hough, G. R., "Remarks on Vortex-Lattice Methods," Journal
of Aircraft, Vol. 10, May 1973, pp. 314-317.
0 10 NS(semi-span) 50 6
Stark, V. J. E., "A Generalized Quadrature Formula for Cauchy
Fig. 8 Effect of taper ratio on CL (real) convergence. Integrals," AIAA Journal, Vol. 9, Sept. 1971, pp. 1854-1855.