Page 1
The LIniversity of Adelaide
NUMERICAL SOLUTION OFPRANDTL'S LIFTING-LINE trQUATION
Boedi Koerniawan
Ir. (Tanjungpura), Grad. Dip. Math. Sc.(Adelaide)
Thesis submitted for the degree of
Master of Sciencein the
University o{ Adelaide
Department of Applied Mathematics
May 1992
Page 2
This thesis is dedicated to
my rnother and rr. y late father,
whorr-^ l love very much.
Page 3
The fear of the Lord is the beginning of knowledge
but fools despise wisdom and instruction.
For the Lord giveth wisdom :
out of his mouth cometh knowledge and understanding.
Proverbs I:7 ;2:6
Page 4
-l '.
Contents
List of Figures
List of Tables
Summary
Signed Statement
Acknowledgements
1 General Introduction
1.1 Lifting Surface Theory
L.2 Lifting-line Theory
2.3 Listing Program
2.4 Numerical Results
vl
vlt
vul
lx
x
1
1
2 stewartsonts solution of Prandtlts Lifting-line Equation L3
2.L Introduction to Stewa¡tson's Method .
10
13
L7
27
2.2 Numerical Evaluation of Stewartson's Double Integral 15
1V
Page 5
3 Semi-inflnite Wing with an Asymptotically Constant
Chord
3.1 Numerical Method
3.2 Listing Program
3.3 Numerical Results
4 Semi-inffnite \ü'ing -
Chord c(y) Proportional to yå ¿s tTends to fnffnity
4.7 Numerical Method
4.2 Listing Program
4.3 Numerical Results
5 Application of Prandtlts Lifting-line Equation to the Corn-
pliant Layer Problem
5.1 Varley-'Walker Solution .
23
23
27
31
39
39
4L
45
51
51
5.2 Numerical Evaluation of the Varley-Walker Solution 52
5.3 Direct Solution of the Compliant Layer Equation 59
6 Conclusion
Appendices
A DOlAHF - NAG Fortran Routine Summary
B FO4JGF - NAG Fortran Routine Summary
Bibliography
69
72
72
lÐ
79
v
Page 6
List of Figures
1.1 An aeroplane wing with airfoil section.
L.2 A closed curve C in flow field.
1.3 Uniform flow past a thin wing
4.1 The curves ø : +"(g) are touching the ellipse at one end. 40
5.1 Comparisonbetween our method and Varley-Walker solution
for the compliant layer equation.
2
4
b
68
vl
Page 7
List of Tables
2.1 Numerical Results for f (x) Provided by Stewartson .
2,2 Numerical Results for /(r) Obtained from Stewartson's Dou-
ble Integral
3.1 Numerical Results for Prandtl's Lifting-line Equation with
Constant Chord
15
22
34
34
35
Ðt
3.2
3.3
3.4
Optimum Balance between L and n
Numerical Results for /(c) - a Comparison
Numerical Results for Prandtl's Lifting-line Equation with
Chord Defined by (3.2a)
4.1 Numerical Results for Prandtl's Lifting-line Equation with
Chord c(fi : yL
4.2 Numerical Results for Prandtl's Lifting-line Equation with
Chord Defined by (4.10) . . .
b.r Numerical Results for /(r) Obtained from Varley-Walker So-
lution
5.2 Numerical Results of the Compliant Layer Equation
47
49
58
67
vrl
Page 8
Surnmary
Prandtl's lifting-line equation is an integro-differential equation which is
used for calculating the span-wise distribution of circulation around wings
which are three-dimensional quasi-planar, with negligible sweep and large
aspect ratio, placed at a small angle of attack to a stream of incompressible
fluid. Since Prandtl introduced this famous equation, there have been many
who have tried to solve the equation. Stewartson (1g00) has solved the
equation analytically for a semi-infinite wing of constant chord, the final
result being a function involving a double integral. To calculate the value of
this function numerically, we need a computer routine which is, in general,
very time consuming. In this thesis, we shall develop an efficient numerical
method for solving Prandtl's lifting-line equation directly, using a non-
uniform grid which is concentrated at the wing-tip. The results agree to
nearly three significant figures with computation, based on Stewartson's
double integral, for the case of semi-infinite wings of constant chord. Our
method is useful for semi-infinite wings in a more general case, not only of
constant chord. The same equation but with negative constant chord, which
applies to a non-aerodynamic problem, e.g. a compliant layer problem, is
also studied. However, our method is less accurate for this problem, because
wave-like behaviour occurs.
vllr
Page 9
Signed Statement
The contents of this thesis have not been submitted to any university
for the purpose of obtaining any other degree or diploma. Also, to the
best of my knowledge and belief, the thesis contains no material previously
published or written by another person, except where due reference is made
in the text of the thesis.
I give consent to this copy of my thesis, when deposited in the
University Library, being available for loan and photocopying.
Boedi Koerniawan
lx
Page 10
.A'cknowledgements
I would like to express my sincere thanks to my supervisor, Profes-
sor Ernest Oliver Tuck, for his guidance during my Master's research and
during the preparation of this thesis. I am very much indebted to him for
his interest, ideas and advice during the discussion of all problems.
I am also very much grateful to Professor Ren B. Potts for his guid-
ance prior to my Master's research and during the finishing touches of this
thesis.
I am indebted indeed to the International Development program
of Australian Universities and Colleges (IDP) for its financial supports.
\Mithout the scholarships from IDP, this work could not have been done.
The invaluable help I have received from Dr. David L. clements -the Dean of the Faculty of Mathematical and Computer Sciences, Dr. peter
M. Gill - the Head of the Department of Applied Mathematics, Ms. Lorna
Read - the Secretary of the Department of Applied Mathematics, Ms. Vivien
Hope - overseas Student Adviser and Dr. F. Salzborn are all much ap-
preciated.
x
Page 11
Chapter 1
General fntroduction
1-.1 Lifting Surface Theory
A lifting surface is a thin streamlined body that supports the flight of birds,
aircraft etc., on which a hydrodynamic lift force is generated when it moves
through the surrounding fluid.
A typical example of lifting surface is a wing of an aeroplane. Let
us consider the wing of an aeroplane as sketched in Figure 1.1. The cross-
sectional shape obtained from the intersection of the wing with a vertical
plane parallel to the centre-line of the aeroplane is called an airfoil or aero-
foil. The most forward and rearward edges of the wing are called lhe leailing
edge and the trailing ed,ge of the wing, respectively. The straight line con-
necting the leading and trailing edges on an airfoil is the chord line of. the
airfoil, and the distance between the leading and trailing edges, measured.
along the chord line, is its choril. The angle between the chord line and the
direction of the free-stream is defined as the geometricøI angle of attøclc, ot
inciilence, of the airfoil. The distance between the two wing-tips is called
lhe span. The mean chord of a wing is the average of the chord of all its
1
Page 12
I
l
leading trailing edge
T -lI
I
JI
L--J
_II
i.-chordJ
Figure 1.1: An aeropla,ne wing with airfoil section.
section-wise airfoils, namely the wing planform area divided by the span.
The ratio of the span to the mean chord is called tlne aspect ratio of. lhe
wing. For a rectangular wing, the aspect ratio A, is
sA, tc
where s is the span and c is the chord of the wing. For a non-rectangular
wing, the aspect ratio is
Ar:"'.,A,
where s is the span and ,4. is the planform area of the wing. If the aspect
ratio is very large, then two-dimensional theory is valid.
In fluid dynamics, we have a basic equation, which is widely appli-
cable to any fluid, compressible or incompressible, viscous or inviscid, i.e.
lhe conseruation of mass equation, ot continuity equati,on
0oAt + div (Ps-) : g, (1.1)
2
)
Page 13
where p is the density of fluid and g- is the velocity of fluid particle. For
steady flow, we have
and hence from (1.1) we obtain
div (Pq-) : 9. (t.2)
For an incompressible fluid, where temperature effects are not significant,
the density p remains constant. Hence from (1.2), we obtain
divf:g' (1.3)
In subsonic flow, for instance, the fluid is incompressible.
T}:e uorticity õ at any point in a flow field is defined as
ui : curl g". (1.4)
If al : õ', th" flow is called to be irrotationøl. otherwise, the flow is called.
to be rotational. For irrotational flow, we may define lhe oelocity potential
ó(r,V,z,ú) such that
i: i ó. (1.5)
For an incompressible, irrotational flow, we have
div g'- ¡
and
í:íó,
v.V/ : g
0pat
0
3
and hence
Page 14
v
curve c
Figure 1.2: A closed cr¡rve C in flow ffeld.
qIntegrationproceeds sothat enclosedarea remainson left
x
or
Y'ë:0. (1.6)
This is the well-kno\¡/n Laplace equation, which is a linear, second-order
partial differential equation.
T}:e circulation is defined as the line integral of the velocity around
any closed curve. Referring to the closed curve C of FigureI.2, the circu-
lation I is given by
I t' íar (1.7)
Kelvin's theorem of the conservation of circulation states that for an invis-
cid fluid acted upon by conservative forces only, the circulation is constant
throughout the fluid.
Now consider an incompressible uniform flow with free-stream veloc-
ity U in the positive r-direction, past a thin wing with small geometrical
angle of attack, as sketched in Figure 1.3. The wing spans along the y-axis.
Since the free-stream velocity is constant, the vorticity there is zero. The
4
Page 15
I z v
wake
Figure 1.3: Uniform flow past a thin wing
vorticity will remain zero throughout the surrounding fluid, except in the
regions adjacent to wing surface and behind the wing. The region adjacent
to the wing surface, in which non-zero vorticity exists, is called the bounilary
Iayer. The thickness of the boundary layer depends upon the viscosity of
the fluid; thinner layer for smaller viscosity. The existence of the non-zero
vorticity in the boundary layer is due to shear force acted upon the fluid
particles by the surface of the wing, as the effect of viscosity. In most flows
of engineering interest, the viscosity of the fluid is extremely small, and
hence the boundary layer is negligibly thin. The region behind the wing,
where the vorticity is also non-zero, is called lhe walce. Since it forms a
thin sheet behind the trailing edge, it is also called the trailing aorter sheet.
Furthermore, rüe assume the surface of the wing is impermeable. Then
the normal velocity component of the flow field adjacent to wing surface is
equal to that of the wing surface. In any uniform flow, the wing is assumed
to be in the state of rest. So we have
ñ.í6: s (1.8)
x
Ð
Page 16
on the wing surface, where / is the velocity potential of the flow field and
ñ is the unit vector normal to wing surface.
The velocity g-of the undisturbed free-stream is
where /i" th" unit vector in the ø-direction in Cartesian coordinates system.
This implies
Ó:Uæ
in the free-stream. If Q(r, y,z) is the perturbation velocity potential due
to the wing, then the velocity potential in the neighbourhood of the wing
i: tli,
IS
IS
ó : Ux + Q. (1.9)
Hence
íó: ui +iø. (1.10)
Let z: f+(x,y) and ": l-@,y) be the equations of the upper and lower
surfaces of the wing. The unit vector d on the upper surface " : f+(*,y)
n: !_? - r_*þ,ù) . : ffi, (1.11)l V("-f+@,ù)l W' \¿'r
where i,i and, k- are the unit vectors in the x,y, z-directions of the Cartesian
coordinates, respectively. Substituting (1.10) and (1.11) into (1.g) yields
#:rfftrv+#)+rfftrff>, G12)
which applies on the upper surface , : f+(*,a). A similar boundary
condition applies on the lower surface ,: f-@,y).Since the wing is thin
6
Page 17
and the geometrical angle of attack is small, the slopes #,W,A{} and,
ff ur. small, except at the leading edge. Hence, as we can see from (1.12),
the vertical perturbation velocity component ff is small.
If both I Í+@,y) | and I f-@,y) | are very small, the exact wing
may be replaced by a cut on the r, gr-plane, with upper surface z : 0+ and
lower surface z :0-. Using Taylor series expansion we have
(ff),=,* : (ff),=o.*#r#),=oa * #r#),=01 * ... = (#),=01.
(1.13)
Similarly, we have.aø. ao( ¿r)"=t-
x ( *)"=o-. (1.14)
The perturbation velocity componentr å3 .rd ffi, which are tangential to
the cut, are negligible since the viscosity of the fluid is small. Hence the
boundary condition (1.12) can be approximated by
Uar+
(1.15)
orL z :0..,", for (*rA) on the cut. Similarly, we have
UaÍ-
(1.16)
oÍt z :0-, for (rry) on the cut.
Another boundary condition, which is known as Kutta-condition,
requires the velocity at the trailing edge to be finite, i.e.
tÕ<oo
at the trailing edge. At infinity, the flow is undisturbed; hence the veloc-
ity at infinity is equal to the free-stream velocity, or in other words, the
0r
ôr
aøOz
aoOz
t
Page 18
perturbation velocity potential tends to zero at infinity. Since there can be
no pressure jump except through a solid surface, the pressure is continuolls
at trailing edge and throughout the fluid. Consequently, the pressure is
continuous through z :0 within the wake. The pressure coeffi.cient at any
point in the flow field, including the upper and lower wing surfaces z : 0*,
is proportional to ffi. H"n"" we have
.aa, .ôo.(
6*-),=o+ : ( 6*),=o- (1.17)
within the wake.
So, now we have a linear, second-order partial differential equation
V2o :0
throughout the surrounding fluid, subject to the boundary conditions
A6 trãfl-ã-:u+-ozôrlAO ¡rð1-Ø-" ur'VQ < oo,
üÕ --' o,zâOr ¡âÕr\ gs- )z=o1 : \fi )z=o-1
oî z:0..,. for (rra) on the cut,o'n z:0- for (*,a) on the cut,at the trailing edge,at infinity,within the wake.
(1.18)
The perturbation velocity potential Õ(*,A,2) can be decomposed
into even and odd functions of z, i.e.
Q(*, A, z) : Q "(r, y , z) I Q o(n, y, z),
where
(Þ" (", y, z) : e "(*,, u, - z) :
|W {*, a, z) + e (*, y, - ")],
Q o(*, a, z) : - e o(æ, a, - z) : f,[o {r, a, z) - e (r, y, - z))
8
Page 19
Hence
?Q.(x,A,z) _0z
õQ.(r,A,z) _0z
aÕ v z) 7Q(x,a, -z)ôz 0z
)Q(r,A, -z)0z
+0z
âoir
|r
(r
( at )
Then the boundary conditions on the cut z : O+ are
49 = ++U(%!- eJ:1. ott z : o+3i: =1,û(4'* #j," ;; ; : o;
The even function Q"(*ry,z) represents the perturbation velocity
potential of the flow past a wing which is symmetrical with respect to the
plane z : 0, of thickness /+(ø, y) - l-@,U), at zero geometrical angle of
attack. This potential does not produce lift force and therefore it holds
Iittle interest for the present study. The odd function Qo(r,y,z) repre-
sents the perturbation velocity potential of the flow past curved surface of
zero thickness, with equation z : f,f¡+(r,y) + f-@,y)], namely the n"teo,n-
camber surface of the original wing. This represents the effect of camber
and geometrical angle of attack. This is the important part of the pertur-
bation velocity potential since a lift force will be produced. The lift force
depends merely on the function |lr*@,,ù + r- @,y)l : r(n,,9r). From now
on,\¡/e shall ignore (Þ", writing Õ - Oo.
In order to solve the boundary value problem (1.18), one technique
(see Thwaites, [2] 1960) is to replace the wing and wake by a vortex sheet of
strength l@,y). At any fixed span-wise coordinate y, there is a circulation
l(y) around the chord, as defined by (1.2), and related to 7(ø, y) by
rbf(y): I t@,a)dn, (1.19)Ja
I
Page 20
where ø and ó are the ø-coordinates of the leading and trailing edges, re-
spectively. After some manipulation, we obtain
r-(ld€dn, (1.20)
@-Ð'+(a-q)'where u(*,A,0) : åÕ( rrU,0) is the vertical component of the perturbation
velocity oÍL z : 0 and the integration is over the wing surface ^9. This
is called the lifting surface equation. The vertical perturbation velocity
component t¿ is induced by the trailing vortex sheet. In general, the induced
velocity is in downward direction and therefore we call it llne downwash
aelocity, and the above boundary condition states that it is a given quantity,
, : UH. Here ff is the local geometrical angle of attack of the mean-
camber surface " : f (*) of the wing. There exists a singularity at rl : y
for all æ ) (, which represents physically the trailing vortex sheet.
L.2 Lifting-line Theory
The first three dimensional-wing theory, which is the most prominent one
in this century for calculating aerodynamic forces acting on the wing is
the lifting- line theory, introduced by L. Prandtl in 1g18. prandtl,s lifting-
line theory leads to a famous linear integro-differential equation which is
singular in the Cauchy principal value sense. \Me call this Prandtl's lifting-
line equation. This equation can be derived from the lifting surface equa-
tion (1.20) by further approximation for very large aspect ratio, using the
method of matched asymptotic expansions (Ashley and Landahl, [B] 1g6b).
Prandtl's lifting-line equation is used for calculating the distribution of cir-
culation round a wing which is three-dimensional quasi-planar, with neg-
w(x,a,,o):¿; lrlffin*
10
Page 21
ligible s\Meep and large aspect ratio, placed at a small geometrical angle
of attack to a steady, subsonic stream of incompressible fluid. The wing
is considered as a straight line, parallel to its leading edge. An additional
piece of information required for solving Prandtl's lifting-line equation con-
cerns the boundary values at the wing-tips. \Me expect the circulation to
fall to zero al, the wing tips. According to Robinson and Laurmann ([4]
1956), the circulation must drop to zeto at the tips of the wing, otherwise
there will exist trailing vortices of finite strength behind the tips, which
would give rise to infinite downwash velocities at the wing.
Once the distribution of circulation l(y) as defined in (1.19) has
been found, we can subsequently calculate the lift force and induced drag
acting on the wing. Bera ([5] 1991) found that a dragless lifting solution
to the equation exists. However, Bera did not force the circulation at the
wing-tips to be zero.
If we construct a y-axis in the plane of the wing parallel to the leading
edge and the wing spans from y : 0 to A : s, then Prandtl's lifting-line
equation is of the form
f(v) : rc(v)l(J.,s* * l":Hl, 0 ( v { s, (r.2r)
where l(y) is the circulation around the y-axis,, "(v) is the chord of the
wing, U is the velocity of the free-stream and ae is the geometrical angle
of attack of the wing. The integral is a Cauchy principal value.
In this thesis, we emphasize the aerodynamics of the wing-tip. So,
\r¡e are going to use a serni-infinite wing rather than a finite wing. The wing
now spans from y : 0 at the wing-tip to y : oo and the circulation l(y)
11
Page 22
around the y-axis satisfies (1.21) with s : oo, i.e.
f(y) : nc(y)luas+ * l"*'l!lrl, 0 <y. -., (1.22)
Prandtl's lifting-line equation can also be used for some non-aero-
dynamic applications. For example, if we take c(y) in (1.22) as a negative
constant, say c(y) - -C, where C is apositive constant and define
r(y): -rcuaslr- l@)1,rC
9 - , e)+
(1.23)
then the integro-differential equation (L.22) becomes an integro-differential
equation in /(ø), i.c.
"f("):-!¡*ry, o(r(oo, (1.24)\ / rJo {-nwhich is the integro-differential equation in a compliant layer problem (see
Varley and Walker, [6] 1989). Similar equations can be found in other
physical applications, such as waves near ship bow or stern (Tuck, [z] 1gg1),
\I/ave near a floating dock (Varley and Walker, [6] 1989) and heat conduction
(Varley and Walker, [6] 1939).
Stewartso" ([8] 1960) was able to solve equation (r.22) analytically
for a wing of constant chord. Varley and \Malket (16] 1g8g) have also red-
erived and generalised that solution. A generalised lifting-line theory for
curved and swept wings has been recently introduced by Guermond ([9]
1ee0).
72
Page 23
Chapter 2
Stewartsonts Solution ofPrandtlts Lifting-line Equation
2.L Introduction to Stewartsonts Method
Stewartso" ([8] 1960) solved Prandtl's lifting-line equation (1.22) for a wing
of constant chord "(A)
: c subject to the boundary conditions
l(0) : 9, l(-) : ¡rc(Jo¿o. (2.1)
He defined a nelv function /(ø) such that
l(y):rcUaslt-f@)1, y:rcr/ . (2.2)
By substituting (2.2) into equation (7.22), he obtained an integro-differential
equation in /(r), i.e.
t(,):+1"*# (,>o), (2.s)
subject to the boundary conditions
/(o) : r, /(-) : o.
13
Page 24
Instead of solving equation (2.3) directly, Stewartson solved the more gen-
eral integro-differential equation
l(o,r):-i lo* orr(",t)Kr(ol€-*l)ssn((-n) (" >0), (2.4)
subject to the boundary conditions
/(4,0) : 1, /(a, oo) : 6,
where a is real and positive, and .I(r is the Bessel function of order one,
of the second kind and with imaginary argument. In order to solve this
equation, he wrote l@,r) as the inverse Fourier transform of. F¡(u), viz
f(a,r):1r-_t
2r J-- F*(u:)e'i" dt:, (2.5)
where F+(r) is a regular function of ø in the upper half plane, Im c.r ) 0,and
used the \Miener-Hopf technique to determine F+(.). The final result for
the case c : 0 when (2.4) reduces to (2.8) is
1¡æf@):' ¡
7t Jo ""pr-* 1""##ro,e-tt
(1 + úr)(2.6)
once the function /(r) has been found, rve can calculate l(y) using (2.2).
Stewartson did not show how to evaluate the double integral (2.6).
Instead, he provided series expansions of /(z) for large and small values of
r as the approximation of the solutions. For large r,he found that
î,\ 1 I 1 4f @) : ;+ *z*z(z+l"g ")+#12(log r)2 +47log r-1 -i"1*. . ., (2.7)
where 7 is Euler's constant 0.5772...For small r, he found the following
series expansion
r@) : t - zç!¡î - lfiltOosux*, - *74
) + o,(!)*,lf
(2.8)
Page 25
0.2370.2670.3080.3610.4380.5641r)r7.21.00.80.60.40.20T
Table 2.1: Numerical Reeulte for /(o) Providedby Stewa¡tgon
0.0860.1130.1610.1750.L920.272f(r)4.03.02.01.81.61.4fr
where O¿ means that the order includes an unspecified power of log r. Ste',rr/-
artson also provided a table of the numerical results of (2.6), which is shown
in Table 2.1.
2.2 Numerical Evaluation of Stewartson's Dou-ble Integral
In order to obtain accurate numerical results for small values of r, and to
reduce cpu time, our method for evaluating the double integral (2.6) is as
follows. Let 0 : tan u and hence
larctant
fo*"'^'
Now lettingu: I for the first integralin (2.9) and z : i- i for the second
integral in (2.9), we obtain
l"' '# : T fo'*u"
' log(2 "inf,)a, * i I:-'
arctanú rog(2
"inf,)au: -Tifrarctan t) * o(tr - 2 arctanú)1, (2.10)
where g(d) i. Clausen's integral defined by
s(ó): - loÓ øsçz"inf,)d,, (0 < d < r). (2.11)
log(tan z)du
log(2 sinu)du - log(2 cosu)du. (2.9)
15
Page 26
According to Abramowitz and Stegun ([t0] L964, page 1005) , Clausen's
integral (2.11) can be expanded as follows
s(ó): -órosó+ó.å1#",rffi: -dlos ó + ó + 0.013888889d3 + 0.000069444ó5 + 0.000000787ó7
+0.000000011de+0(10-10) (0=ø.i), Q.12)
where B2¡ are Bernoulli numbers. Slightly greater accuracy is obtainable
by economising the series (2.72), and we use a set of economised coeffi.-
cients supplied by Prof. J.N. Newman which give at least 14 decimal place
accuracy for g($), 0 < d 1rf2, the economised series being
s(ó) + ór"eöó
A similar economised series is used for the range r12 < þ < r. Hence,the
double integral (2.6) reduces to a single integral
f (*) : + lr* #"*pt#{s( 2arctant)+ s(r -2arctant)}ldt. (2.18)
To evaluate this single integral numerically, we write the whole integral as
Io* o(*,t)dt :
lo' o(*,,t)dt + f,* hçr,,t¡dt, (2.14)
where h(r,t) is the integrand in (2.13). Since h(n,,t) tends Lo zero for 1arge
values of ú , by choosing appropriate values of r, the second integral on the
right hand side of (2.L4) is negligible. The values of r are chosen through
: 1 + 0.0342 6s45s7260fi(+),
+ 0 . 0 0 0 422T 82 5 r2s 6 s f?*f + 0 . 0 0 0 0 1 r B2T BT r s7 s (+f+ 0 . 0 0 0 0 0 0 42 5 5T 8z t 5 f7*f + 0 . 0 0 0 0 0 0 0 1 z s 6 6 6 0, (*)' "
+ 0 . 0 0 0 0 0 0 0 0 0 z b 2 6 6 4 f?*l' + 0 . 0 0 0 0 0 0 0 0 0 0 427 4a (1* )' ^ .
16
Page 27
experiments to give acceptable results, and r is inversely proportional to
r. The value of r for 6 figure accuracy has been chosen as r: 200f r for
r ) 0 and r :1.441x 105 for z :0. Then the single integral (2.13) can be
evaluated by any integration method.
2.3 Listing Program
The program in Fortran 77 for evaluating the integral (2.13), using the
method described above, is listed below, where DO1AHF is a NAG routine,
available from the computer system, for evaluating the definite integral of
a function FUN over a finite range. For a description of this NAG routine,
see Appendix A.
IMPLICIT REAL*8 (A-H,O-Z)
COMMON X,PI
EXTERNAL FUN
PI : 3. 1415926535897932384626D0
DO 10 J:0,L2
rF (J.LE.10) THEN
X : 0.2*DFLOAT(J)
ELSE
X:DFLOAT(J)-8.0D0
END IF
rF (X.LT.5.0D-7) THEN
R : 1.441d5
T7
Page 28
i)
'i
,l
ELSE
R : 200.0d0/x
END IF
RESULT : O.ODO
A : 0.0d0
B:REPS : 1.0D-7
NLIMIT: 10000
IFAIL : O
Z : DOL AHF(A,B,EPS,NPTS,REL,FUN,NLIMIT,IFAIL)
RESULT:RESULT +ZIPT
WRITE(7,20)X,RESULT
10 CONTINUE
20 FORMAT(F5.2,5X,F9.6)
PRINT *,"SEE FILE 'FORT.7' FOR THE RESULTS."
STOP
END
,<**{<t(*'F***{<***,t<t(*{<***t<****'F***{<***'t*:ft{<>t*{<t<*{<******{(**{<**
REAL*8 FUNCTTON FUN(T)
IMPLICIT REAL*8 (A_H,O_Z)
REAL*8 B(8),BB(12)
coMMoN x,PI
B(0) : 1.0D0
18
Page 29
B(1) : 0.034269459726017D0
B(2) : 0.000422782512969D0
B(3) : 0.00001182737r379D0
B(4) : 0.000000425578715D0
B(5) : 0.000000017366602D0
B(6) : 0.000000000752664D0
B(7) : 0.000000000042746D0
BB(0) : 0.693147180559945D0
BB(1) : -0.102808379178034D0
BB(2) : -0.0063 41737696915D0
BB(3) : -0.0007 45724327 445D0
BB(4) : -0.000108523615117D0
BB(5) : -0.0000 77757036264D0
BB(6) : -0.000003726027195D0
BB(7) : -0.000000590910020D0
BB(8) : -0.000000097125621D0
BB(9) : -0.0000 00037 414793D0
BB(10) : 0.000000004587491D0
BB(11) : -0.000000003754394D0
THETA : 2.0D0*DATAN(T)
SUM : 0.0D0
rF (T.LE.1.0) THEN
DO 30 L:0,7
19
Page 30
30
40
50
60
SUM : SUM + B(L)*(2.0D0*THETA/PI)**(2*L)
CONTINUE
ELSE
DO 40 L:0,11
SUM : SUM + BB(L)+(2,ODO_2.ODO*THETA/PI)**(2*L)
CONTINUE
END IF
rF (T.LE.1.0) THEN
FIRST : SUM*THETA _ THETA*DLOG(THETA)
ELSE
FIRST : SUM*(PI-THETA)
END IF
THETA_PI_THETA
SUM : 0.0D0
rF (T.LE.1.0) THEN
DO 50 L:0,7
SUM : SUM + B(L)*(2.0D0*THETA/PI)**(2*L)
CONTINUE
ELSE
DO 60 L:0,11
SUM : SUM + BB(L)*(2.0D0-2.0D0*THETA/Pr)+*(2*L)
CONTINUE
END IF
rF (T.LE.1.0) THEN
20
Page 31
SECOND : SUM*THETA - THETA*DLOG(THETA)
ELSE
SECOND : SUM*(PI-THETA)
END IF
ZZ : -0.5D0*(FIRST+SECOND)
FUN : DEXP (-T*X) *DEXP (-ZZ / PI) / ( 1. 0D 0+T* *2) * *0. 7b
RETURN
END
2.4 Numerical Results
The results obtained from the single integral (2.19) are displayed in Ta-
bre 2.2, compared with the 3-figure table given by Stewartson ([g] 1960).
The Stewartson solution method has been generalised by Varley and
Walker ([6] 1989) to a method for solving the integro-differential equation
u(r) :: [^* :(")^a" (o < , ( -),TtJo s-r
where u(ø) and u(ø) can be expressed in terms of the unknown function
/(ø) and its derivatives as
d" f dn_r fu : d"ñ * an-tã,,-, + ... + crof
and
' : b*# t bn-t#+ ... + óof.
However, this generalisation still does not allow us to treat the case c f con-
stant. Instead, we use a direct method, which we are going to describe in
the following chapters.
27
Page 32
r)
T¿ble 2.2: Nurnerical Reeults for J(ø) Obtained fro¡n Stewa¡teon'e Double T:rtegral
1.0000.5640.4380.3610.3080.2670.2370.2120.1920.1750.1610.1130.086
1.0000000.5639120.4379150.3610560.3075910.2677980.2368830.2121260.1918420.t749200.1605930.1129380.086303
0.000.200.400.600.801.001.201.401.601.802.003.004.00
from StewartsonTs tableour methodr fr
22
Page 33
Chapter 3
Serni-infinite Wing with anAsyrnptot ically ConstantChord
3.1- Nurnerical Method
In this chapter, we shall assume that c is a function of g which tends
to a constant coo as y tends to infinity; this includes as a special case
Stewartson's case c : constant. For this class of c(y), the appropriate
boundary conditions on I are
|(0) : g, l(-) : ¡rcao(Jao.
Let us write equation (1.22) as
f(y) : :"c(y)l(rc,s* * 1""1r-
++nJ"r'þt)dnq-a l, (3.1)
where .t is a positive constant. The first integral in (3.1) is the part of
the integral equation which is going to be evaluated numerically, while the
second integral will be evaluated analytically in advance, to give a correction
to the numerical process. We have to choose a value ror L later, to give a
23
Page 34
certain accuracy. Motivated by Stewartson, let us write
f(y) = rrc*tlas(t - 2) (3.2)a'
as the approximation of l(y) for U ) L, where rc is a constant and is yet
unknown. Hence
f'(y) = nc*Uaorc
.y2 (3'3)
Now the second integral in (3.1) becomes
+ f"r'Ø)dn =! f coouaorcdq. (3.4)4rJt q-y -4lt n,Ø-y)
\Me know that this is equal to f,rc*Uo¡rc times the Hilbert transform of
U-2, so we obtain
* l:W =f,"*u*""(h",st ht -h), v +0,a + L (85)
Using the result in (3.5), we rewrite (3.1) as
f(y) : nc(y)l(ras + * 1""'T44, * I"*u,""(hrcr I h I -hn(3.6)
Since the aerodynamics at the wing-tip is crucial,'we are going to use a non-
uniform grid which is concentrated at the wing-tip to solve the Prandtl's
lifting-line equation. The process will be described below.
Define
y:Lt2, o<¿<1, (B.z)
n:Lu2, o(z(1, (g.s)
and hence
drl :2Ludu.
24
(3.e)
Page 35
Now if we define
then we obtain the relationships
r(v)
and
h(u):l'(Lu2)2Lu
f,(Lu2)2Ludu
(r)
u
f'
h(
: l,': 1",: 1",
dr¡
)du,
(3.10)
(3.11)
(3.12)
l"
l"
I t'(Lu2)2LuduL(uz - tz)h(u)d,u
L(uz - tz¡
1
By substituting (3.11) and (3.12) into (3.6) we obtain
|o,nçu¡au+f,"{r'*>lI"'m|trcoo(Jag"(#_#'"*l#zlll: rc(Lt2)uao. (9.18)
Now we use a uniform grid in u for equation (3.18), i.e. let
ui : j ln, h¡ : h(u¡), j :0,7)2,... )n. (8.14)
Applying the trapezoidal rule of integration to equation (8.18) with this
grid, is equivalent to applying the trapezoidal rule with a non-uniforrn grid
in 4 to equation (3.6), since q : Lu2. Again we have to choose a value of
n in (3.14) such that, together with choosing ,t in (8.1), we obtain a certain
accuracy. Now if we replace f in (8.13) by
t' t:( 1
-t-i - ,i - ¡)1", (3.15)
25
Page 36
where 'i : Ir2r3r... , r? respectively, then by applying the numerical method
we have chosen above, we obtain a system of rz linear algebraic equations
with n * 2 unknowns i.e. horhtrhzr. . . ,hn and ¡c. The i-th equation of this
system is of the form
1I-'4n
_c(Lt?-)r ho _ hn _ts 2h¡ .- 8', L@;:6-fçti_,-r- kf@j-q)r1.L1
*]trc*uasc(Lt2¿-)n(øT - ¡ra ,t"g1x-;i x-;l) : rc(Lt?_;)uoo
ó(i) (h¿-, *
1
L-t?,"2
h¿-t I h¿
2
r r?'
(3.16)
where
0 ifi:l#(t , a lr') in i, :2h(no* h¿-t +Di-:r2hi) in i:3,4,5,,
Note that rtye use t¿-r here instead of t¿: if nri:0,1,2r...rn in order to
avoid the singularities.
Using (3.11) we obtain
t(¿) : lo' rrçu¡au. (s.12)
Then using (3.2) and (3.17) we obtain
¡l
Jo' nç"¡a": rcoo(Jaort - i). (3.1S)
Therefore numerically we obtain the extra (n*l)'st equation
*ro, t h* + .i.ro,, * ry3n, : ircoo(J .,o. (s.1e)j=t
So far we have n + 7 equations with n * 2 unknowns.
26
Page 37
Now using (3.3) and (3.10), we obtain a relationship between rc and
hn as followsLt*: *"*r1^h'' (3'20)
where hn : h(u) l"=t. This reduces the number of unknowns to n*1.
Substituting (3.20) into (3.16) and (3.19) yields
1 .-' h¿-, * h¿,ó(i)+
^(h,-'+ï)Lc(Lt?-r)¡ ho - hn -S 2h¡ t' s" tfç6, - tç,r, - kfej -q)t
+!"e,tl-t)h,( Lt? . Ltl ,t-; x-il"s
I
1 1 1 l): n"(Lt?-¡)U'o,I-t? ,2-;
'i :7r2r3r, . . ,n (3.21)
and1Tr1
b(h" + h* + Lzn¡) * ;n": Tcco(Jo¿o. (3.22);tt, j=t
So we have a system of.nl1 linear algebraic equations with n*1 unknowns
i.e. hsrhtrhrr. . . ,hn which can be solved by any available method.
Once horht,,hrr...rhn have been found, v¡e can calculate
l@o) for
i : 0,1,,2, . .. , n where a¿ : Lt? : L(*)'.This can be done numerically i.e.
from (3.11) we obtain
r(v¿) : 01
2n1
2n
ifi:0(åo*hr) ifi:1(åo * h +D'j--lzhj) if i : 2,,9,4,...,n.
(3.23)
3.2 Listing Program
The program in Fortran 77 for solving Prandtl's lifting-tine equation (I.22)
for asymptotically constant chord, using the method described above, is
27
Page 38
listed below, taking U : ao: 1, where FO4JGF is a NAG routine, available
from the computer system, for finding the solution of a linear algebraic
equations system. For a description of this NAG routine, see Appendix B.
IMPLTCTT REAL*8 (A-H,O-Z)
PARAMETER (LEN:400)
REAL*8 AA(LEN,LEN),F(LEN),WK(4*LEN)
coMMoN IPAP"I ALPHA,YU
LOGICAL SVD
ALPHA : I.ODO
YU :1.0D0
PI : 3. 1415926535897932384626D0
CINF E ......... (Here is the value of c*)
NRA : LEN
TOL :5.0D-16
LW : 4*LEN
IFAIL : O
PRINT *,"READ FROM FILE 'FORT.l':"
READ(1,*)RL,N
PRINT 'kr)rL:t),RL," N :",N
PRINT *,"'WAIT ......'
IG : N*l
JG:IG
DO 20 I:l,N
28
Page 39
T : (DFLOAT(r)-0.5D0)/DFLOAT(N)
Y - RL*T>k>I,2
DO 10 J:I,JG
U : DFLOAT(J-1)/DFLOAT(N)
DENOM : RL*(T**2_U**2)
rF (J.EQ.1.OR.J.EQ.Jc) THEN
W : 1.0D0
ELSE
\M : 2.0D0
END IF
AA(I,J) : \M*C(Y)/(8.0D0*DFLOAT(N)*DENOM)
rF (J.EQ.r) THEN
AA(I'J) : AA(I,J) + 3.0D0/(8.0D0*DFLOAT(N))
END IF
rF (J.EQ.r+1) THEN
AA(I'J) : AA(I,J) + 1.0D0/(8.0D0*DFLOAT(N))
END IF
rF (r.NE.1.AND.J.EQ.1) THEN
AA(I'J) : AA(I,J) + 1.0D0/(2.0D0*DFLOAT(N))
END IF
rF (r.NE.1.AND.J.EQ.r) THEN
AA(I,J) : AA(I,J) + 1.0D0/(2.0D0*DFLOAT(N))
END IF
29
Page 40
10
20
30
rF (r.GT.2) THEN
rF (J.LT.r.AND.J.GT.1) THEN
AA(I,J) : AA(I,J) + 1.0D0/DFLOAT(N)
END IF
END IF
CONTINUE
TEMP : DLOG(DABS(1.0D0/(1.0D0-T**2))) / (Rr_,*t**+¡
TEMP : 1.0D0/(RL*T**2) - TEMP
AA(I,JG) : AA(I,JG) + C(v¡*TEMP/8.0D0
F(I) : PI*C(Y)*YU*ALPIIA
CONTINUE
AA(IG,1) : 1.0D0/DFLOAT(2*N)
AA(IG,JG) : AA(rG,1)
DO 30 J:2,N
AA(rG,J) - 1.0D0/DFLOAT(N)
CONTINUE
AA(IG,JG) : AA(IG,JG) + 0.bD0
F(IG) : PI*CINF*YU*ALPHA
CALL FO4J GF(IG, J G,AA,NRA,F,TOL,SVD,SIG,IR,\MK,LW,IFAIL)
Y : 0.0D0
GAMMA : O.ODO
WRITE(2,40)Y,GAMMA
DO 50 I:2,IG
30
Page 41
Y : RL*(DFLOAT(I-1)/DFLOAT(N))**2
GAMMA : GAMMA + (F(I-l)+F(I))/(2.0D0*DFLOAr(N))
WRITE(2,40)Y,cAMMA
40 FORMAT(F7.4,5X,F11.8)
50 CONTINUE
PRINT X,"SEE FILE 'FORT.2' FOR THE RESULTS.''
STOP
END
+'F* * + * * * * * * * * * * + * * * * + * * * * * x {< {< * * + * * {< * * *{< ** * * {< * * * ** * * * * * *
REAL*8 FUNCTTON C(Y)
IMPLICTT REAL*8 (A_H,O_Z)
coMMoN IPAP"/ ALPHA,YU
C: ( Here is the function c: c(y) )
RETURN
END
3.3 Numerical Results
As a first example, we turn our attention to a wing with constant chord,
as studied by Stewartson ([8]1g60). Let c : fJ : do : 1. If we choose
L - 20rn : 68, for instance, then we obtain the results as shown in
Table 3.1. Further, if we transform the results from l(y) to /(ø) using
(2.2), then we obtain the results as shown in Table 8.3, compared with the
results obtained by Stewartson's method, i.e. from formula (2.1g). From
Table 3.3, we can see that relative error peaks at y/L:0.0002 with error
31
Page 42
€.1 : 0.00054348, at A /L : 0.0106 with error ez : 0.00053283 and at
AIL:0.2500 with error es: 0.00002712. Relative error increases again
at the upper end of the range 0 < y < L. At y/L: 1: the relative error
is large, which is ea - 0.00362680. In general, if we choose other values
of .t and n, most results have peak errors et¡ê2¡€3 and ea similar those
we mentioned above, with e1,e2 arLd e3 slightly shifting in their positions.
The error e3 is always much less than the others and therefore we shall not
consider it. Both €1 and e2 ãîê sensitive to n, wher€âs €4 is sensitivelo L.
If we keep .t constant, then e1 increases as ?? increases. on the other hand,
e2 decreases as r? increases. Optimum balance between .t and n is when
€1 and €.2 ãîe about equal, and Table 3.2 shows the best such choice of n
for each L,,e.g.rz:68 is the optimal choice for n when L:20. In these
circumtances, maximum relative-errors for all pairs of .t and rL aiÍe equal
to 0.00054, in the range specified in Table 3.2, for L : 10,20,80,40 and
50. For all values of .t and n, maximum relative errors are never less than
0.00054. However, in the case where -t and n are balanced, absolute errors
are less than 0.0005 for all values of y in the whole range 0 < y 1 L.
As a second example, we consider the chord
( 0 if y:Qclv):\ ify>o. G.24)
The solution of equation (L22) is therefore
r(v) : n t*vAgain, we take U : ao : 7. If we choose L : I0 and n - 60, then we
obtain the results as shown in Table 3.4, compared with the exact solution.
v
32
Page 43
Relative error is large for small value of y. The error decreases as n in-
creases. In order to obtain relative error less than 0.0005 for all values of
y inthe range 01A (.t, wemay chooser¿:60for L:L0, n:85forL :20 and r¿ : 100 far L: 30. For larger value of .t, we need larger value
of n.
33
Page 44
Table 3.1: Numerical Reeults for Prandtl's Lifting-line Equation with Constant Chord
(L-20,n-68)
Table 3.2: Optimum Bala¡rce between L and n
3.049322583.053336663.057098523.060628323.063944343.067063133.069999713.072767723.075379583.077846603.080179113.082386533.084477463.086459773.088340653.090126643.091823683.093437043.094971243.096429573.097812453.099105503.10033294
9.L5229.55459.9654
10.384910.813111.250011.695572.749772.612513.083913.564014.0528L4.550215.056215.570916.094316.6263L7.L67077.7L6318.274218.840819.416120.0000
2.777798432.804429772.8290L9722.851260302.87L472462.889704t42.906336842.92L488282.935315302.947956362.959533842.970155982.979918602.988906682.9977956L3.004852433.011936783.018501823.024595003.030258753.035531033.040445893.04503391
2.28812.49132.70332.92393.15313.39103.63753.89274.15664.429r4.71025.00005.29845.60555.92L36.24576.57876.92047.27087.62987.99748.37378.7587
0.000000000.260236230.511670150.749567L60.97L547321.776547691.364363061.535358221.690269171.830061001.955825382.068707302.169853542.260377622.34t336952.473719702.478434662.536314882.588112552.634505022.676098702.773434272.74699227
0.00000.00430.01730.03890.06920.10810.15570.21190.27680.35030.43250.52340.62280.73100.84780.97321.10731.25001.40L41.56141.73017.90742.0934
f(v)avIraf(s)v
0.000540.000540.000540.000540.00054
0<alL<0.660<ylL<0.830<ylL<0.910<alL<0.940<alL<0.96
48688395L07
1020304050
Maximum relative errorRangenL
34
Page 45
Table 3.3: Numerical Resulte for J(o) - a Comparison
(L=2O,n=68)
0.00000.00020.00090.00190.00350.00540.00780.01060.01380.01750.02160.02620.03110.03650.04240.04870.05540.06250.07010.07810.08650.09540.70470.77440.72460.13520.L4620.L5770.16950.18190.19460.20780.22750.2355
0.000000070.000543480.000012520.000262810.000410670.000488190.000523790.000532830.000524330.000503880.000475350.000447440.000404280.000365480.000326300.000287670.000250380.000214930.000181780.000151130.000123320.000098040.000075640.000055870.000038760.000023920.000011470.000001050.000007730.000014290.000019270.000023150.000025470.00002692
0.000000070.000498730.000010480.000200050.000283550.000305210.000296160.000272280.000242100.000210250.000179330.000150690.000125000.000102480.000083090.000066630.000052840.000041400.000032020.000024390.000018270.000013360.000009500.000006480.000004160.000002380.000001060.000000090.000000620.000001070.000001350.000001520.000001570.00000156
1.000000070.917662960.837140810.761205310.690463330.625188030.565413590.511008020.461728510.477263240.377262770.341359330.309189170.280396980.254646270.237622720.211036910.192624500.776146770.161386620.148153060.136273680.725595720.115983800.107318120,099492870.092414600.086000940.080179220.074885320.070062750.065661640.061637920.05795268
1.000000000.977764230.837130330.761405360.690746880.625493240.565709750.511280300.461970610.4L7473490.377447440.341510020.309314170.280499460.254729300.231689350.211089750.192665900.176178190.161411010.148171330.136287040.L25605220.115990280.L07322280.099495250.092415660.086001030.080178600.074884250.070061400.065660120.061636350.05795112
0.00000.00550.02200.04960.08810.73770.19830.26980.35250.44670.55070.66640.79300.93071.0794L.23971.40981.59157.78431.98812.20282.42862.66542.91333.t7273.44793.72284.01.474.3L764.63154.95645.29235.63935.9972
alLrelativeerror
absoluteerrorfrom (2.6)our method
T It
35
See next page -)
Page 46
Table 3.3 (continued):
0.25000.26490.28030.29610.31230.32890.34600.36350.38150.39990.4t870.43790.4ó760.47770.49830.51920.54070.56250.58480.60750.63060.65420.67820.70260.72750.75280.77850.80470.83130.85840.88580.91370.94200.97081.0000
0.000027120.000026230.000024690.0000224L0.000019300.000015510.000011230.000006440.000000850.000005330.000012110.000019200.000027580.000036310.000046110.000056650.000067980.000080520.000094790.000110930.000128600.000148860.000L72420.000199550.000231070.000269030.000316310.000374940.000450910.000554180.000701300.000930900.001340810.002386620.00362680
0.000001480.000001350.000001200.000001030.000000840.000000640.000000440.000000240.000000030.000000180.000000390.000000590.000000810.000001020.000001240.000001460.000001680.000001910.000002160.000002430.000002710.000003020.000003370.000003760.000004200.000004720.000005360.000006140.000007140.000008490.000010400.000013370.000018660.000032200.00004746
0.054577470.051463800.048602660.045964040.043526610.04L271390.039181470.037241750.035438710.033760280.032195620.030735010.029369670.028091730.026894080.025770290.024714540.02372L570.022786580.021905230.021073570.020287980.019545170.018842140.018176130.017544620.016945280.016376000.015834820.015319920.014829660.014362490.013917010.013491880.01308591
0.054569990.05L462450.048601460.045963010.043525770.041270750.039181030.03724t5r0.035438680.033760460.032196010.030735600.029370480.028092750.026895320.025777750.024776220.023723480.022788740.021907660.021076280.020291000.019548540.018845900.018180330.017549340.016950640.016382140.015841960.015328410.014840060.014375860.013935670.013524080.01313337
6.36626.74627.13727.53927.95228.37638.81139.25749.7L4510.182610.661711.151911.653012.L65212.688313.2225L3.767774.323974.897215.469416.058716.659017.270277.892518.525979.t70219.825520.49L921.169327.857622.557023.267523.988924.727325.4648
alLrelativeerror
absoluteerrorfrom (2.6)our method
fr T
36
Page 47
0.00000.00030.00110.00250.00440.00690.01000.01360.01780.02250.02780.03360.04000.04690.05440.06250.07110.08030.09000.10030.11110.L2250.73440.14690.16000.17360.18780.20250.27780.2336
0.000000000.000432000.000421860.000405600.000384070.000358380.000329700.000299210.000268050.000237180.000207370.000179220.000153120.000129300.000107880.000088830.000072080.000057480.000044840.000033980.000024690.000016790.000010100.000004440.000000330.000004350.000007730.000010580.000012990.00001504
0.000000000.000071430.000138930.000199000.000248900.000286900.000312300.000325360.000327170.000319340.000303750.000282390.000257120.000229590.000201230.000173080.000145990.000120500.000096950.000075530.000056280.000039150.000024020.000010760.000000810.000010880.000019620.000027200.000033790.00003953
0.000000000.165346980.329328390.490634340.648061080.800551940.947225831.087392311.220553811.346396851.464775r3r.575687251.679251911.775682801.865264881.948333052.025254002.096411082.162L92352.222987462.279750862.331057092.379037732.423409622.464468042.502486632.537717902.570393992.600727752.62891393
0.000000000.165275550.329189460.490435340.647812180.800265040.946913531.08706695L.22022664r.3460775L1.46447r381.575404867.678994797.7754532L1.865063651.948159972.025108012.096290582.L62095402.222905932.279094582.331017942.3790L3772.423398862.464468852.502497572.537737522.570421792.60076t542.62895346
0.00000.00280.01110.02500.04440.06940.10000.13610.17780.22500.27780.33610.40000.46940.54440.62500.71110.80280.90001.00281.11117.22507.34447.46941.60001.73611.87782.02502.L7782.3361
alLrelativeerror
absoluteerrorour exact u
Iv
T¿ble 3.4: Nurnerical Reeulte for Prandtl'e Lifting-line Equation with Chord Defined by (3.2a)
(.L = 1O, zr = 60)
See next page ->
37
Page 48
Table 3.4 (continued):
0.25000.26690.28440.30250.32110.34030.36000.38030.40110.42250.44440.46690.49000.51360.53780.56250.58780.61360.64000.66690.69440.72250.75110.78030.81000.84030.87110.90250.93440.96691.0000
0.000016780.000018290.000019600.000020750.000021790.000022750.000023630.000024480.000025310.000026130.000026970.000027840.000028750.000029720.000030760.000031890.000033120.000034470.000035960.000037610.000039430.000041460.000043710.000046170.000048840.000051590.000054100.000055410.000052270.000027150.00006045
0.000044560.000049010.000052960.000056520.000059780.000062820.000065670.000068430.000071130.000073820.000076560.000079370.000082320.000085420.000088750.000092310.000096190.000100420.000105060.000110180.000115820.000122080.000128990.000136570.000144750.000153210.000160970.000165170.000156070.000081210.00018107
2.655130402.679539462.702289082.723574052.7433371.82.761870322.779275402.795465262.810704592.825010602.838453792.851098542.863003682.874223072.884805962.894797542.904239272.913168992.92762L822.929629832.937222602.944427362.9ó7269272.95777L522.963955552.969841192.975446792.980789362.985884682.990747382.99539107
2.655174962.679588472.702342042.723570572.743396962.76L933L42.779287072.795533692.810775722.825084422.838530352.851177972.863086002.874308492.884894772.894889852.904335402.9L3269412.92L726882.929740072.937338422.944549442.951398262.957908092.964100302.969994402.975607762.980954532.986040752.990828592.99557274
2.50002.66942.84443.02503.2L773.40283.60003.80284.01114.22504.44444.66944.90005.13615.37785.62505.87786.13616.40006.66946.94447.22507.51117.80288.10008.40288.71119.02509.34449.6694
10.0000
ylLrelativeerror
absoluteerrorexact solutionour method
v I v
38
Page 49
.1
Chapter 4
Serni-infinite Wing Chord.(V) Proportional to yt as yTends to Infinity
4.L Numerical Method
Another class of c(y), which is of practical importance, is the class where
c(y) is proportion al to yI as y tends to infinity. For example, this is the
case when r4/e are studying flow near the tip of a wing of elliptic planform,
which is very common. In this case, the curves , : t.(A) are touching the
elliptic curvature near the wing-tip as sketched in Figure 4.!.
For this class of c(g), we use the same method as that we have used
in chapter 3, except a different correction term for replacing the second
integral in equation (3.1). Let us write
t(y) x ny* (4.1)
as the approximation of l(y) for a > L, where rc is an unknown constant.
Hence
r'(y) = T"o-r. (4.2)
39
Page 50
xx:c(y)
x:-c(y)
Figrrre 4.1: The curves o - Ic(y) are touching the ellipse at one end.
Now the second integral in (3.1) can be evaluated analytically, i.e.
Iì/ n)dnn-a
becomes
v
l- /-4" J" = * l:'#f : #^rt'# t (v #o,v * L)
(4.3)
Hence the integro-differential equation for the circulation l(y) becomes
r(y) : nc(y)l(Ias + * 1""';!1, + #"^*l ,," _ r+W ll. @.4)
Performing the same substitutions as those in chapter 3, equation @.a)
lo'nçu¡au+f,"@t ¡tl,'#!h- #".e ri= ll
: rc(Lt2)(Jao. (4.b)
Then from (3.L7) and (4.1) we obtain
lo' nçu¡ar: LL n. (4.6)
Using (3.10) and (4.2), we obtain a relationship between rc and hn asfollows
1
L;
40
hn, (4.7)
Page 51
where h^: h(u) 1.,=r. Hence using trapezoidal rule with the same grid as
the previous section yields
t h¿-t * h¿,,ó(i)+*(h-a+ 2 )
c( Lt?\ ¿-
8Lt¿-t) 1+Í' 1
lz,, los | å l: ""(Lt?_t)(J ao,
'2i:Lr2r3r...rn (4.8)
and1
(åo * h" + Dzh¡) - hn :0.
4.2 Listing Program
The program in Fortran 77 for solving Prandtl's lifting-line equation (L.22)
when the Chord c(y) is proportional b y* as y tend.s to infinity, using
the method described above, is listed below, taking (J : o,o: 1, where
FO4JGF is a NAG routine, available from the computer system, for finding
the solution of a linear algebraic equations system. For a description of this
NAG routine, see Appendix B.
n-l
2n(4.e)
j=l
Note that
I o if i:rö(i):l *(n"¡h) if i:2
|. #(nr r h¿t +D'j-:zhj) if i :3,4,5,. . . ,n.
Now we have a system of. n 11 linear algebraic equations with n * 1 un-
knowns, which is easily solved. Then we can calculate l(g) bv means of
(3.23) as before.
47
Page 52
IMPLICIT REAL*8 (A-H,O-Z)
PARAMETER (LEN:400)
REAL*8 AA(LEN,LEN),F(LEN),WK(4*LEN)
LOGICAL SVD
coMMoN IPAP"LI ALPHA,YU
coMMoN /PAP.2/ Pr
ALPHA : I.ODO
YU: 1.0D0
PI : 3.1415926535897932384626D0
NRA : LEN
TOL : 5.0D-16
LW : 4*LEN
IFAIL : O
PRINT *,"READ FROM FILE 'FORT.l':"
READ(1,*)RL,N
PRINT *,ttL :tt,RL," N:",N
PRINT *,''WAIT ......"
IG : NtlJG:IG
DO 20 I:l,Nr - (DFLoAr(r)-o.5Do)/DFLoAr(N)
Y : ¡¡*1**2
DO 10 J:I,JG')
42
Page 53
U : DFLOAT(J-1)/DFLOAT(N)
DENOM : RL*(T**2_U+*2)
rF (J.EQ. 1.OR.J.EQ.JG) THEN
W : 1.0D0
ELSE
W : 2.0D0
END IF
AA(I,J) : W*C(Y)/(s.0D0*DFLOAT(N)*DENOM)
rF (J.EQ.r) THEN
AA(I,J) : AA(I,J) + 3.ODO/(8.ODO*DFLOAT(N))
END IF
rF (J.EQ.r+1) rHEN
AA(I'J) : AA(I,J) + 1.0D0/(8.0D0*DFLOAT(N))
END IF
rF (r.NE.1.AND.J.EQ.1) THEN
AA(I,J) : AA(I,J) + 1.0D0/(2.0D0*DFLOAT(N))
END IF
rF (r.NE.1.AND.J.EQ.r) THEN
AA(I,J) : AA(I,J) + 1.0D0/(2.0D0*DFLOAT(N))
END IF
rF (r.GT.2) THEN
rF (J.LT.r.AND.J.cT.1) THEN
AA(I,J) : AA(I,J) + 1.0D0/DFLOAT(N)
43
Page 54
10
20
30
END IF
END IF
CONTINUE
TEMP : DLOG(DABS((1.ODO+T)/(1.ODO-T)))
TEMP : C(Y)*TEMP/ (8.0D0*RL*T)
AA(r,JG) : AA(I,JG) - TEMP
F(I) : PI*C(Y)*ALPHA*YU
CONTINUE
AA(IG,1) : 1.0D0/DFLOAT(2*N)
AA(IG,JG) : AA(IG,1)
DO 30 J:2,N
AA(IG,J) : 1.0D0/DFLOAT(N)
CONTINUE
AA(IG,JG) : AA(rG,Jc) - 1.0D0
F(IG): 0.0D0
C ALL FO4J G F( IG, J G, AA, NRA, F,T O L,SVD, S IG, IR,\ryK, LW, IFAIL )
Y : 0.0D0
GAMMA : O.ODO
WRITE(2,40)Y,cAMMA
DO 50 I:2,IG
Y : RL*(DFLOAT(I_1)/DFLOAT(N))**2
GAMMA : GAMMA + (F(I-1)+F(I))/(2.0D0*DFL9AT(N))
WRITE(2,40)Y,cAMMA
FORMAT(F8.4,5X,F1 2. 8)
44
40
Page 55
50 CONTINUE
PRINT *,"SEE FILE 'FORT.2' FOR THE RESULTS."
STOP
END
********************x**********************************
REAL*8 FUNCTTON C(Y)
IMPLICIT REAL*8 (A-H,O-Z)
coMMoN /PAR1/ALPHA,YU
coMMoN lPLPú2lPr
C - ............. ( Here is the function c:c(y) )
RETURN
END
4.3 Nurnerical Results
As a first example, we consider
/\1cla) : y' .
The exact solution for equatior' (1.22) is therefore
f(y) : nuasyT
Now let us compare the results obtained from our numerical method with
the exact solution. Let U - do:1. If we choose L:20 and n:70, then
we obtain the numerical results as shown in Table 4.1, compared with the
45
Page 56
exact solution. As a second example, we consider the chord
.(v) :
The exact solution for (1.22) is
(ol+( taSr[,IooI no!-rl)(t-r)(@
ify:Qif. y :1if0<y<7or1<y<æ.
(4.10)
r(v) :Again, let [/ : oo : 1 and choose L :20rn:70. We obtain the numerical
results as shown in Table 4.2, contpared with the exact solution. In both
these examples, the errors increase monotonously as y increases. The results
are good only for small values of y. However, for a particular value of y,
the error decreases if we take larger values of. n ot L.
46
Page 57
0.00000.00020.00080.00180.00330.00510.00730.01000.01310.01650.02040.02470.02940.03450.04000.04590.05220.05900.06610.07370.08160.09000.09880.10800.1 1750.72760.13800.14880.16000.17160.18370.19610.20900.22220.2359
0.000000000.000390220.000390370.000390670.000391050.000391540.000392140.000392880.000393710.000394670.000395740.000396940.000398260.000399710.000401300.000403010.000404870.000406870.000409010.000411320.000413780.000416410.0004192L0.000422200.000425380.000428750.000432340.000436150.000440190.000444480.000449040.000453870.000459000.000464440.00047022
0.000000000.000078320.000156700.000235230.000313950.000392930.000472240.000551980.000632170.000712920.000794290.000876370.000959220.001042940.001127610.001213320.001300170.001388250.001477670.001568540.001660990.001755120.001851080.001949000.002049050.002151370.002256150.002363560.002473820.002587150.002703770.002823960.002947990.003076180.00320886
0.000000000.200708990.401417980.602126980.802835971.003544967.204253951.404962951.605671941.806380932.007089922.207798922.408507912.609216902.809925893.010634883.211343883.4\2052873.612761863.813470854.014179854.214888844.475597834.616306824.817015825.0L77248L5.218433805.479742795.619851785.820560786.027269776.227978766.422687756.623396756.824L0574
0.000000000.200630670.401261280.601891750.802522021.00315203r.20378r7L7.4044L0971.605039771.805668012.006295632.206922552.407548692.608173962.808798283.009421563.210043713.410664623.611284193.811902314.012518864.2L3733724.413746754.614357824.874966775.0L5573445.216177655.476779235.617377965.817973636.018566006.219154806.419739766.620320576.82089688
0.00000.00410.01630.03670.06530.10200.14690.20000.26720.33060.40820.49390.58780.68980.80000.91847.04491.17967.32241.4735r.63271.80001.97552.75922.35102.55102.75922.97553.20003.43273.67353.92244.17964.44494.7784
a/Lrelativeerror
absoluteerrorexact solutionour method
v
Table 4.1: Numerical Reeults for Pra¡rdtl's Lifting-line Equation with Chord c({ = gtr
(L=2O,n=70)
47
See next page ->
Page 58
Table 4.1 (continued):
0.25000.26450.27940.29470.31040.32650.34310.36000.37730.39510.41330.43180.45080.47020.49000.51020.53080.55180.57330.59510.61730.64000.66310.68650.71040.73470.75940.78450.81000.83590.86220.88900.91610.94370.97161.0000
0.000476370.000482900.000489860.000497260.000505150.000513580.000522580.000532210.000542530.000553610.000565520.000578360.000592220.000607220.000623490.000641210.000660550.000681740.000705050.000730790.000759370.000791250.000827050.000867520.000913640.000966680.001028360.001101030.001188020.001294300.001427650.001601230.001840030.002201360.002883090.00333873
0.003346400.003489220.003637780.003792570.003954170.004123190.004300340.004486430.004682330.004889050.005107750.005339760.005586570.005849940.006131900.006434830.006761510.007115250.007500010.007920550.008382650.008893450.009461810.010098920.010819180.011641330.012590490.013701120.015022110.016625790.078625220.0272Lr760.024743840.030044560.039927690.04690797
7.024814737.225523727.426232727.626947777.827650708.028359698.229068688.429777688.630486678.831195669.031904659.2326L3659.433322649.634031639.83474062
10.0354496210.2361586110.4368676010.6375765910.8382855911.0389945811.23970357L7.44047256L7.6411215511.8418305512.04253954t2.2432485312.44395752L2.6446665212.84537ó5713.0460845013.2467934973.4475024913.6482114873.8489204774.04962946
7.021468337.222034507.422594947.623749147.823696538.024236508.224768348.425297258.625804348.826306619.026796909.227273899.427736079.628181699.82860872
I0.0290747910.2293977070.4297523510.6300765810.8303650411.0306119311.2308101211.4309507511.6310226311.8310113712.0308982112.2306580472.43025640L2.62964441t2.8287497213.0274592873.2255823373.4227586513.6181669213.8089927814.00272L49
5.00005.28985.58785.89396.20826.53066.86127.20007.54697.90208.26538.63679.01639.40479.8000
70.204710.616311.036711.465311.902012.346912.800073.267213.730674.208214.693915.187815.689816.200076.7184L7.244977.779678.322418.87351.9.4327
20.0000
ylLrelativeerror
absoluteerrorexact solutionour method
a
48
Page 59
0.00000.00020.00080.00180.00330.00510.00730.01000.01310.01650.02040.02470.02940.03450.04000.04590.05220.05900.06610.07370.08160.09000.09880.10800.11750.72760.13800.14880.16000.17160.18370.19610.20900.22220.2359
0.000000000.004373740.007545240.002592580.002953630.003110560.003189850.003234520.003262100.003280530.003294110.003304940.003314240.003322780.003331020.003339210.003347550.003356190.003365230.003374700.003384700.003395240.003406360.003418090.003430490.003443560.003457330.003471850.003487170.003503280.003520250.003538100.003556910.003576700.00359752
0.000000000.000016780.000022370.000079920.000153630.000240560.000338830.000447000.000563920.000688630.000820400.000958580.001102630.001252110.001406660.001565950.00L729720.001897780.002069950.002246080.002426090.002609890.002797430.002988680.003183650.003382340.003584790.003791070.004001260.004215430.004433720.004656250.004883200.005L74720.00535103
0.000000000.003836530.014476760.030826410.052013930.077336560.70622L370.138196600.172870280.209914010.249050640.290044590.332694350.376826560.422291240.468957970.576712820.565455830.615098900.665564100.776782270.768691770.82L237520.874370130.928045210.982222721.036866471.091943681.t47424561.203287971.259491181.316029577.372876421.430012697.48742097
0.000000000.003853310.014454390.030746490.051860300.077096000.105882540.137749600.172306360.20922ó380.248230240.289086010.331591720.375574450.420884580.467392020.514983100.563558050.613028950.663318020.714356180.766081880.818440090.871381450.924861560.978840381.033281681.088152617.143423301.19906654L.255057461.311373327.36799322r.424897971.48206988
0.00000.00410.01630.03670.06530.10200.14690.20000.26120.33060.40820.49390.58780.68980.80000.91847.04491.1 7967.3224L.47357.63271.80001.97552.75922.35102.55102.75922.97553.20003.43273.67353.92244.L7964.44494.7784
ylLrelativeerror
absoluteerrorexact solutionour method
a
"a)
Ta,ble 4.2: Numerical Results for Pra¡rdtl's Lifting-Iine Equation with Chord Defined by (+.f O)
(L=2O,n=70)
49
See next page ->
Page 60
Table 4.2 (continued):
0.25000.26450.27940.29470.31040.32650.34310.36000.37730.39510.41330.43180.45080.47020.49000.51020.53080.55180.57330.59510.61730.64000.66310.68650.71040.73470.75940.78450.81000.83590.86220.88900.91610.94370.97161.0000
0.003619450.003642540.003666860.003692490.003719520.003748020.003778130.003809950.003843610.003879260.003917060.003957210.003999920.004045420.004094010.004146010.004201800.004261820.004326600.004396760.004473080.004556460.004648060.004749290.004862010.004988560.005132120.005296990.005489280.005717990.005997120.006350150.006820830.007506120.008734760.01084271
0.005592360.005838950.006091100.006349120.006613370.006884220.007162140.007447610.00774L170.008043450.008355140.008677040.009010010.009355090.009713440.010086400.010475530.010882650.011309900.011759820.0L2235490.012740580.013279680.013858450.0t4484140.015166030.015916480.076752290.017697050.018785410.020070980.021643090.023667250.026507670.031385310.03962879
r.545084971.60299001L.667122301.719469131.778018731.836760161.895683271.954778602.014037352.073457292.133012752.192714562.252549982.3L2572732.372596892.432796922.493L07672.553524042.614041592.6746559r2.735362902.796158662.857039542.918002062.979042963.040159113.101347573.162605563.2239304L3.285319613.346770773.408281613.469849963.537473763.593151053.65487995
1.539492611.597151061.655031201.713120017.771405361.829875941.888521131.947330992.006296182.065407842.724657672.184037522.243539972.303757642.362883452.4227t0522.482632082.542641392.602731692.662896092.7231274t2.783418082.843759862.904L43672.964558823.024993083.085431093.t45853273.206233363.266534203.326699793.386638523.446L82713.504966093.56L765743.61525116
5.00005.28985.58785.89396.20826.53066.86127.20007.54697.90208.26538.63679.01639.404L9.8000
70.204L10.616311.036711.465311.902012.346912.800073.26L213.730674.208214.693915.187815.689816.200076.718477.244977.779678.322418.8735L9.432720.0000
alLrelativeerror
absoluteerrorexact solutionour method
v vlt
50
Page 61
ì',,.
'ì :l r\
ìjl)
., .1
¡'. .. ¿l
{
\
Chapter 5
Application of Prandtl'sLiftir,g-line Equation to theCornpliant Layer Problern
5.1- Varley-Walker Solution
Let us now let c(y) in (1.22) be c(y) - -C, where C is a positive constant
Then equation (L.22) becomes
* l'(n)dnq-vf(y): -ncltras**1" I (0(y(oo), (5.1)
The boundary condition is again l(0) :0. Define
f(y):-rcuasll-l(')1, rCu:
-1."4 (5.2)
By substituting (5.2) into equation (5.1), we obtain an integro-differential
equation in /(r), i.e.
f(*): -t7t l"* (, > o), (5.3)
which is the integro-differential equation in a compliant layer problem with
boundary condition /(0) : 1. Varley and Walker ([O] 1g8g) obtained the
51
Page 62
exact solution for (5.3) as follows
r(*) :,/i cos(æ* ä) - + l"* #"*pt+ 1"" ffioao"
,,E cos(æ * ä)
(5.4)
5.2 Nurnerical Evaluation of the Varley-'WalkerSolution
Our method for evaluating the function (5.a) is similar to that for evaluating
the function (2.6).
\ /ith this method, formula (5.4) can be rewritten as
(r + "')i ""pt-#{ s(2arctans) + sþr - 2arc!,ar.s)}lds,
(5.5)
where g(ó) i" Clausen's integral defined by (2.11). To evaluate (5.b) nu-
merically, we rewrite it as
t@)
f(")
e-"tl,*1
1f
,,E cos(r * ä)
(1 + "')""pt-#{ s(2arctar.s) + oQr - 2arctans)}lds
SEI7f
I7f
e
l.*e-",
""pt-#{ s(2arctans) + oQr - 2arctans)}lds.
(5.6)
By choosing appropriate values of r, the second integral on the right hand
side of (5.6) is negligible. The values of r are chosen through experiments
to give acceptable results, and r is inversely proportional to z. The value
of r for 6 figure accuracy has been chosen as about r : r00 I r for all values
52
Page 63
of æ, except for x : 0. Then (5.6) can be evaluated by any integration
method.
The program in Fortran 77 Íor evaluating (5.6), using the method
described in section 2.2, is listed below, where DO1AHF is a NAG routine,
available from the computer system, for evaluating the definite integral of
a function FUN over a finite range. For a description of this NAG routine,
see Appendix A.
IMPLICIT REAL*8 (A-H,O-Z)
coMMoN x,PI
EXTERNAL FUN
PI : 3. 1415926535897932384626D0
PRINT *,"READ FROM FILE 'FOFùT.l':"
READ(1,*)RL,N
PRINT '*r))L :)),RL," N :",N
PRINT *,"WAIT ..."
DO 10 J:O,N
x : RL*(DFLOAT(J)/DFLOAT(N))**2
rF (X.LT.5.0D-7) THEN
R : 1.0d5
ELSE
R: 1.0D2/X
END IF
RESULT : DSQRJT(2.ODO)*DCOS(X+PI/8.ODO)
oó
Page 64
A : 0.0D0
EPS: 1.0D-7
NLIMIT : 10000
IFAIL : O
Z : DOl AHF(A,B,EPS,NPTS,REL,FUN,NLIMIT,IFAIL)
RESULT: RESULT _ZIPI
\4/RITE( 7, 20 )X, RES ULT
10 CONTINUE
20 FORMAT(f8.4,5X,F10.6)
PRINT *,"SEE FILE 'FORT.7'FOR THE RESULTS."
STOP
END
*>ß*rk*********+>t>k*****'1.>lc*,ß**{<>t<*****{<rt<>ß**{<***rkrF{<*t<>t<***rkrßrt<*
REAL*8 FUNCTTON FUN(S)
IMPLICIT REAL*8 (A-H,O-Z)
REAL*8 B(8),BB(12)
COMMON X,PI
B(0) : 1.0D0
B(1) : 0.034269459726077D0
B(2) : 0.000422782512969D0
B(3) : 0.00001 1827371'379D0
B(4) : 0.000000425578715D0
B:R
54
Page 65
i-t-:1
.I
I
1.l
'l
I1l
I
30
B(5) : 0.000000017366602D0
B(6) : 0.0000000007õ2664D0
B(7) : 0.000000000042746D0
BB(0) : 0.693147180559945D0
BB(1) : -0.102808379178034D0
BB(2) : -0.0063 4L737696915D0
BB(3) : -0.0007 45L24327 445D0
BB(4) : -0.000108523615117D0
BB(5) : -0.0000t7757036264D0
88(6) : -0.0000 03L26027 L95D0
BB(7) : -0.000000590910020D0
BB(8) : -0.000000097125621D0
BB(9) : -0.000000037414793D0
BB(10) : 0.000000004587491D0
BB(11) : -0.000000003754394D0
THETA : 2.0D0*DATAN(S)
SUM : 0.0D0
rF (s.LE.1.0) THEN
DO 30 L:0,7
SUM : SUM + B(L)*(2.0D0*THETA/PI)** (2*L)
CONTINUE
ELSE
DO 40 L:0,11
bb
Page 66
40
50
60
SUM : SUM + BB(L)*(2.0D0-2.0D0*THETA/PI)**(2*L)
CONTINUE
ENDIF
rF (s.r8.1.0) THEN
FIRST : SUM*THETA _ THETA*DLOG(THETA)
ELSE
FIRST : SUM*(PI-THETA)
END IF
THETA:PI_THETA
SUM : 0.0D0
rF (s.LE.1.0) THEN
DO 50 L:0,7
suM : suM + B(L)+(2.0D0*THETA/PI)** (2*L)
CONTINUE
ELSE
DO 60 L:0,11
SUM - SUM + BB(L)*(2.0D0-2.0D0*THETA/er)**(2*L)
CONTINUE
END IF
rF (s.LE.l.o) THEN
SECOND : SUM*THETA - THETA*DLOG(THETA)
ELSE
SECOND : SUM*(PI-THETA)
56
Page 67
END IF
ZZ : -0.5D0*(FIRST+SECOND)
FUN : DEXP (-X*S) *DEXP (ZZ I PD I (1. 0D0+ S * *2)** 7.25
RETURN
END
The numerical results obtained from the above program for r :L(*)', L :20, n :70, j : 0,I,2,. .. , n is shown in Table 5.1.
57
Page 68
Table 5,1: Numerical Results for J(ø) Obtained from Varley-Walker Solution
1.1169131.3023951.3687891.3000007.0926240.7587280.327093
-0.157670-0.638565-1.052301-7.337677-1.445358-r.347640-1.046378-0.577158-0.007995
0.5684251.0488091.3380301.369516]-7230230.635102
-0.001723-0.653165-1.169869-L.422L35-1.334999-0.914159-0.2537740.4794321.0853511.3851971.2774370.7772990.024136
5.28985.58785.89396.20826.53066.86127.20007.54697.90208.26538.63679.01639.40479.8000
10.204110.616311.036711.465311.9020L2.346912.8000L3.267213.730674.208214.693915.187815.689816.200016.718477.244977.779678.322418.873519.432720.0000
1.0000250.9998240.9984250.9945790.9869440.9740770.9544290.9263560.8881460.8380580.7743830.6955260.6001090.4870950.3559430.2067700.040531
-0.L40797-0.334045-0.534718-0.736901-0.933258-1.115133-7.272789-1.395809-1.473674-7.496525-1.456087-7.346772-1.166480-0.918239-0.610456-0.2577220.1192580.4943520.837477
0.00000.00410.01630.03670.06530.10200.14690.20000.26720.33060.40820.49390.58780.68980.80000.91847.04491.17967.3224t.47351.63271.80001.97552.L5922.35102.55102.75922.97553.20003.43273.67353.92244.r7964.44494.71845.0000
r@)tr)ftr
58
Page 69
5.3 Direct Solution of the Cornpliant LayerEquation
Our numerical method for solving the integro-differential equation (5.3)
directly is again using a non-uniform grid which is concentrated at the
wing-tip, as described in chapter 3. We rewrite equation (5.3) as
l@): _! [" f'G)de _! f f'G)d€ (, > o), (5.2)rJo (-æ nJt (-xwhere ,t is a positive constant. For ø > L, f @) is approximated by
f @) * tÆcos(x+ ii - að' rrr(5.8)
Hence
t: l'G)dc(-r
: t/llcos(n + f,).i{z - r) + sin(ø + f,)ci{z -')l.Ï,#bet*t-*t, (5.e)
where
and
si(ø) : - l,*sin ú
dt (5.10)t
ci(r) : - l: Yor. (5.11)
Now equation (5.7) becomes
r@) ++ 1"" # : -fl.o"(*+ f)si(r - u) + sin(z r ä)ci(¿ - r)l
å[år"s l*l-hl. (5.12)
The functions si(ø) and Ci(ø) can be evaluated using series expan-
sions (Abramowitz and Stegun, [10] L964, pages 2J2-2JJ), i.e.
59
Page 70
si(r) Ën=O (2n * 1)(2n + 1)! 2
(-t¡"*2"+r 7f
Ci(r)
si(r)
ci(r)
7rrog'+Ëffia r -1¡"*zn7*ros'+tffi,
-f @) cos 0 - 9(r) sin c,
/(o) sin r - g(r) cos 0,
for r ( 1, with 12 figure accuracy,
for u ( 1, with 14 figure accuracy,
foræ)1,
forr)1,
40.02L433322.624977570.236280L57.705423
48.796927482.485984
1114.978885449.690326
where 7 is Euler's constant 0.5772156649. . . and
r@) :38.027264
26õ.187033335.67732038.102495
Iæ
h:b2:bs:b4:
A1 :
Q2:c$:A4:
h:b2:h:b4:
A1 :A2:A3:A4:
42.242855302.757865352.01849821.821899
So equation (5.12) together with the boundary condition /(0) : 1
can be solved using the method described in chapter B. Unfortunately, the
method is less accurate for this equation. This is a more diffi.cult problem
which needs furthur work in order to obtain better results. The problem is
due to the wave-like behaviour that occurs in the solution.
60
Page 71
)
The programin Fortran 77 fot solving (5.3) directly, using the method
described above, is listed below, where FO4JGF is a NAG routine, avail-
able from the computer system, for ffnding the solution of a linear algebraic
equations system. For a description of this NAG routine, see Appendix B.
IMPLICIT REAL*8 (A-H,O-Z)
PARAMETER (LEN:400)
REAL+8 AA(LEN,LEN),F(LEN),WK(4+LEN)
LOGICAL SVD
COMMON PI
PI : 3. 1415926535897932384626D0
NRA : LEN
TOL :5.0D-32
LW : 4*LEN
IFAIL : O
PRINT *,"READ FROM FORT.l:''
READ(1,*)RL,N
PRINT *,"L:",RL," N :",N
PRINT *,"WAIT ......"
IG : Nf1
JG:IG
DO 20 I:l,NT : (DFLOAT(r)-0.5D0)/DFLOAT(N)
X - RL*T¡F{<2
61
Page 72
DO 10 J:I,JG
u : DFLOAT(J-1)/DFLOAT(N)
DENOM : RL*(U**2_T**2)
rF (J.EQ.1.OR.J.EQ.Jc) THEN
W : 1.0D0
ELSE
W : 2.0D0
END IF
AA(I,J) : W / (2.0D0*DFLOAT(N)*DENOM*Pr)
rF (J.EQ.r) THEN
AA(I,J) : AA(I,J) + 3.ODO/(8.ODO*DFLOAT(N))
END IF
IF (J.EQ.I+l) THEN
AA(I'J) : AA(I,J) + 1.0D0/(8.0D0*DFLOAT(N))
END IF
rF (r.NE.1.AND.J.EQ.1) THEN
AA(r,J) - AA(r,J) + 1.0D0/(2.0D0*DFLOAT(N))
END IF
rF (r.NE.1.AND.J.EQ.r) THEN
AA(I,J) : AA(I,J) + 1.0D0/(2.0D0*DFLOAT(N))
END IF
rF (r.cT.2) THEN
rF (J.LT.r.AND.J.cT.1) THEN
62
Page 73
')
10
20
30
AA(I,J) : AA(I,J) + 1.0D0/DFLOAT(N)
END IF
END IF
CONTINUE
P : DLOG(DABS(RL/(RL-X)))/X**2 - 1.0D0/(RL*X)
R:X+PI/8.ODO
Z:RL_XF(r) : -DSQRT(2.0D0)*(DCOS(R)*Sr(z)+DSrN(R)*Cr(z))/pr
1 - 1.0D0
F(I) : F(I)-PIPI**2
CONTINUE
AA(IG,1) : 1.0D0/DFLOAT(2*N)
AA(IG,JG) : AA(rG,1)
DO 30 J:2,N
AA(IG,J) : 1.0D0/DFLOAT(N)
CONTINUE
PL : PI*RL
F(IG) : DSQRiT(2.0D0)*Dcos(Rl-.,+prl8.0D0) - 1.0D0 - 1.0D0/pL
CALL FO4JGF(IG,JG,AA,NRA,F,TOL,SVD,SIG,IR,\A/K,L'W,IFAIL)
X : 0.0D0
FX : 1.0D0
WRITE(2,50)X,FX
DO 40 I-2,IG
x : RL*(DFLOAT(I-1)/DFLOAT(N))**2
63
Page 74
FX : FX + (F(r-l)+F(r))/(2.0D0*DFLOAT(N))
WRITE(2,50)X,FX
40 CONTINUE
50 FORMAT(F8.4,5X,F10.6)
PRINT *,"SEE FILE 'FORT.2'FOR THE RESULTS."
STOP
END
{<******rF*rF***t<{<*<{<***t<>t<****>t(*t<*****'t<t<*(*{<*rk*t<*<***rk{<{<t<'Èt r<*
REAL*8 FUNCTTON Sr(Z)
IMPLICIT REAL*8 (A-H,O_Z)
COMMON PI
rF (2.LE.1.0) THEN
SI : g.¡¡g
DO 60 N:0,7
D : DFLOAT(2*N+1)
SI : SI + (-1.0D0)*<'r,N:tz,t<*(2*N+1)/(D*H(D))
60 CONTINUE
SI:SI-PI/2.0D0
ELSE
SI : -DCOS(Z)*FF (z) - DSIN(z)*c(z)
END IF
RETURN
END
* * * * * * * {< * + * * t<,k *,F * * *,t< {< * * * * * * * * * * * * * * *,k * * * {< * * {< t( )t< *,Fr< X X *
64
Page 75
REAL*8 FUNCTTON CI(Z)
IMPLTCTT REAL*8 (A-H,O-Z)
rF (2.LE,7.0) THEN
cI : 0.5772756649 + DLOG(Z)
DO 70 N:1,8
D - DFLOAT(2*N)
CI : CI + (-1.0D0)**19*7**(2*N)/(D*H(D))
70 CONTINUE
ELSE
CI : DSIN(Z)*FF(Z) - Dcos(z)*c(z)
END IF
RETURN
END
,<***+>k*>1.+*{<*t **'ßt**{<****t<rt(***r<********'F*>t<**<*'ß*t *{<***
REAL*8 FUNCTTON FF(Z)
IMPLICIT REAL*8 (A_H,O_Z)
L - Z**8 + 38.027264*Z**6 + 265.L87099*Z**4
A : A + 335.677320*Z**2 + 39.102495
B :7**3 + 40.021433*Z**6 + 322.62497L*fZ**4
B : B + 570.236290*Z**2 + r57.L05423
FF : A/(B*Z)
RETURN
END
65
Page 76
REAL*S FUNCTTON G(Z)
IMPLICIT REAL*8 (A-H,O-Z)
A - Z**8 + 42.24285ó*Z**6 + 302.757865*Z**4
A : A + 352.018498*2**2 + 21.821899
B : 2**3 + 48.196927*Z**6 + 482.485984*Z**4
B : B + 11L4.978885*Z**2 + 449.690326
G : A/(B*Z**2)
RETURN
END
*,k * >k * * * * * * *< * * * + * * * * X >k * * t< * * * *,F * * {< {< * X * * {< * * * {< * * r. * * * t< * >k *
REAL*8 FUNCTTON H(D)
IMPLTCTT REAL*8 (A-H,O-Z)
H : 1.0D0
N : INT(D)
DO 80 I:2,N
H : H*DFLOAT(I)
80 CONTINUE
RETURN
END
Table 5.2 shows the numerical results, using our method with L :20and r¿ - 70, compared with that obtained from Varley-Walker solution; and
Figure 5.1 shows the results in graphs.
** t< * * * * * * ** * * * * * * ** ** * * * * * ** * * * * * * * * * ** * * * * t< * * * ** * r. *
66
Page 77
1.1 169131.3023951.3687891.300000r.0926240.7587280.327093
-0.157670-0.638565-1.052301-L.337677-1.445358-7.347640-1.046378-0.577158-0.007995
0.5684251.0488091.3380301.369516L.7230230.635102
-0.001723-0.653165-1.169869-1.422735-1.334999-0.914159-0.2537740.4794321.0853511.3851977.2774370.7772990.024136
1.241600t.3620541.358165r.2206500.9534470.5756260.1.21644
-0.360516-0.813585-r.L77529-1.397836-L.434597-L.270927-0.919189-0.4237440.1450700.6938351.7262451.3578181.3346977.0484790.543672
-0.085381-0.709439-1.189788-1.409809-7.305746-0.888480-0.2488770.4576281.0464061.3520937.2780070.8317630.023187
5.28985.58785.89396.20826.53066.86127.20007.54697.90208.26538.63679.01639.404L9.8000
70.204L10.616311.036711.465311.902072.346912.800073.267213.730614.208214.693915.187815.689816.200076.778417.244977.779678.322418.873579.432720.0000
1.0000250.9998240.9984250.9945790.9869440.9740770.9544290.9263560.8881460.8380580.7743830.6955260.6001090.4870950.3559430.2067700.040531
-0.140797-0.334045-0.534718-0.736901-0.933258-1.115133-1.272789-1.395809-r.473674-7.496525-1.456087-7.346772-1.166480-0.918239-0.610456-0.2577220.1192580.4943520.837477
1.0000000.9862520.9709430.9526070.9297790.9009540.8645810.8190770.7628570.6943880.6L22480.5L52280.4024280.2733960.728269
-0.032072-0.205836-0.390179-0.581069-0.773208-0.960012-1.133684-7.285407-1.405669-r.484746-1.513352-7.483429-1.389073-7.227573-1.000083-0.713059-0.378244-0.013137
0.3594510.7724461.016316
0.00000.00410.01630.03670.06530.10200.14690.20000.26120.33060.40820.49390.58780.68980.80000.91841.04491.1796t.3224L.47357.63271.80001.97552.75922.35102.55102.75922.97553.20003.43273.67353.92244.77964.44494.77845.0000
from(5.4)our methodt
5.4our methodr ftr
Table 5.2: Numerical Results of the Compliant Layer Equation
(L=2o'n-70)
67
Page 78
r(")our methodVarley-\Malker sulution
2.O
1.5
1.0
0.5
.0"9OO
c
c
o
o5 x-o.5
-1.O
-1.5
-2,O
.o 10
o¿
15 20
'?9o.0å
Figure 5.1: Comparison between ou¡ method and Varley-Walker solution for the cornpliant layer
equation.
68
Page 79
Chapter 6
Conclusron
The lifting-line equation was developed by Prandtl (1918) in order to calcu-
late the span-wise distribution of circulation around wings of large aspect
ratio. This is an aerodynamic problem, but with some modification, the
lifting-line equation can also be used in other applications, such as the com-
pliant layer problem. Solving Prandtl's lifting-line equation is a difficult
task, since it is a singular integro-differential equation. In the aerodynamic
problem, the degree of difficulty depends on the wing chord distribution,
i.e. the chord function c(y).
An analytic solution of Prandtl's lifting-line equation for a semi-
infinite wing of constant chord has been found by Stewartson (1g60) and
confirmed by Varley and Walker (1989). However, it is still difficult to
obtain accurate numerical results from the Stewartson's solution, because
of the double integral and semi-infinite range involved in the function ap-
pearing in the solution. It is therefore convenient to simplify the function
before evaluating it numerically. This can be done by some manipulation of
the inner integral of Stewartson's double integral, leading to the so-called
Clausen's integral. By evaluating Clausen's integral beforehand, it becomes
69
Page 80
easier to calculate the whole function of Stewartson's solution, which now
reduces to a single integral. The same technique can also apply to the
Varley-\Malker solution of the compliant layer equation.
Direct numerical solution of Prandtl's lifting-line equation for a wing
of finite span 0 I A 1 s can be obtained by transforming the integro-
differential equation into an equivalent integral equation, and then using
an appropriate grid and integration algorithm. A suitable grid for the
equation is a non-uniform grid which is concentrated at the wing tips, the
grid spacing increasing quadratically as the distance from the wing tips
increases. The integration algorithm used is the trapezoidal rule. The
use of this grid and integration method is especially accurate for a wing
of constant chord, where the circulation l(y) near the wing tip g : 0 is
approximately proportional b y*.
The numerical task becomes more difficult for the case treated in
this thesis - a semi-infinite wing as s tends to infinity, because rü/e norv
need to devise a satisfactory procedure for truncating the infinite range
of integration. This can be achieved by separating the whole integral in
Prandtl's lifting-line equation into two parts, one orì. a finite range and
the other on a semi-infinite range. The unknown function on the semi-
infinite range is approximated with a predicted function which is valid for
large values of the span-wise coordinate. It is not easy to predict such a
function. It has to be treated case by case, as we vary the chord function
c(y). In other words, we have to choose different functions in the truncation
correction for different classes of c(y).
For the compliant layer equation, the source of difficulty is the tvr¡ave-
70
Page 81
like behaviour that occurs in the solution of the equation. In my opinion,
in order to obtain more accurate results in solving this equation in future
work, we should consider other grids and integration methods. The results
are very sensitive to the angular frequency and the phase of the wave-
like function which we choose for the truncation correction. Therefore the
angular frequency and the phase should be correctly chosen, as well as the
actual form of this function.
7t
Page 82
Appendix A
DO]-AHF - NAG Fortrarr.Routine Summary
Important: For a complete specification of the use of this routine see the
NAG FORTRAN Library Manual. Terms marked ll...ll may be imple-
mentation dependent.
A. Purpose:
I /D}LAHF I / cornputes a definite integral over a finite range to a
specified relative accuracy.
B. Speciffcation:
I lreall I FUNCTION //D01AHF I I (4, B, EPSR, NPTS, RELERR,
F, NLIMIT, IFAIL)
C IîITEGER NPTS, NLIMIT, IFAIL
C l/reaIl/ A, B, EPSR, RELERR, F
C EXTERNAL F
C. Parameters:
A,- llrcatl/ .
on entr¡ A must specify the lower limit of integration. unchanged
72
Page 83
on exit.
B-llrcatl l.On entry, B must specify the upper limit of integration. Unchanged
on exit.
EPSR - l/rcarll.On entry, EPSR must specify the relative accuracy required. Un-
changed on exit.
NPTS _ INTEGER.
On exit, NPTS contains the number of function evaluations used in
the calculation of the integral.
RELERR-llrcarll.
On exit, RELERR contains a rough estimate of the relative error
achieved.
F - l/rcal// FUNCTION, supplied by the user.
It is called by I ID}LLHF / I to evaluate the integrand at the point
X. Its specification is:
llrcatll FUNCTION F(x)
I lrcatll xx- llreat/l .
on entry, x specifies the point at which the value of the integrand
is required by I /D}LAHF / / . X must not be reset by F.
F must be declared as EXTERNAL in the (sub)program from which
//D01AHF // is called.
NLIMIT _ INTEGER.
73
Page 84
On entry, NLIMIT must specify a limit to the number of function
evaluations. If NLIMIT.LE.O, the routine uses a default limit of 10,000.
Unchanged on exit.
IFAIL _ INTEGER.
Before entry, IFAIL must be assigned a value. For users not familiar
with this parameter, the recommended value is 0. Unless the routine detects
an error (see next section), IFAIL contains 0 on exit.
D. Error fndicators and 'Warnings:
Errors detected by the routine:
IFAIL: 1
The integral has not converged to the accuracy requested. It may
be worthwhile to try increasing NLIMIT.
IFAIL : 2
Too many unsuccessful levels of subdivision have been invoked.
IFAIL : 3
Invalid accuracy request (i.e. on entry EPSR.LE.0.0).
when IFAIL : 1 or 2 a result is obtained by continuing without
further subdivision, but this is likely to be *+inaccurate** .
74
Page 85
Appendix B
FO4JGF NAG FortranRoutine Surnrnary
Important: For a complete specification of the use of this routine see the
NAG FORTRAN Library Manual. Terms marked /1...1/ may be imple-
mentation dependent.
A. Purpose:
I |î} JGF / I frnds the solution of a linear least squares problem,
Ax:b, where A is a real m*n (m.GE.n) matrix and b is an m element
vector. If the matrix of observations is not of full rank, then the minimal
least squares solution is returned.
B. Specification:
suBRourINE //F04JGF// (M, N, A, NRA, B, TOL, SVD,
SIGMA, IRANK, WORK, L\ /ORK, IFAIL)
C LOGICAL SVD
C INTEGER M, N, NRA, IRANK, L\MORK, IFAIL
C llrealll A(NRA,N), B(M), TOL, STGMA, \MORK(L\^/ORK)
tÐ
Page 86
C. Parameters:
M _ INTEGER.
On entry, M must specify the number of rows of A, where M.GE.N.
Unchanged on exit.
N _ INTEGER.
On entry, N must specify the number of columns of A, where
1.LE.N.LE.M. Unchanged on exit.
A - llrcalf f anay of DIMENSION (NRA,t), where t.GE.N. Before entry,
the leading M*N part of A must contain the matrix to be factorised.
On successful exit, if SVD is returned as .FALSE., the leading M*N
part of A, together with the first N elements of the vector WORK, contains
details of the Householder QU factorisation of A. See NAG Library routine
llF}zwDFll for further details. If SVD is returned as .TRUE., then the
top N*N part of A contains the right hand singular vectors, stored byt<'Frows*{< . The rest of the first N columns of A is used for workspace.
NRA _ INTEGER.
on entry, NRA must specify the first dimension of A as declared in
the calling (sub)program, where NRA.GE.M. Unchanged on exit.
B - I lrealf f artay of DIMENSION at least (M).
Before entry, B must contain the M element vector b. on successful
exit, the first N elements of B contain the minimal least squares solution
vector The remaining M - N elements are used for workspace.
TOL- //reaIl l.on entry, ToL must specify a relative tolerance to be used to de-
to
Page 87
termine the rank of A. TOL should be chosen as approximately the largest
relative error in the elements of A. For example, if the elements of A are
correct to about 4 significant figures then TOL should be set to about
5*19**(-4). See Section 11 of the Library Manual routine document for
a description of how TOL is used to determine rank. If TOL is outside
the range ( "pr,
L.0), where eps is the machine accuracy (see NAG Library
routine I /XlzAAF I / ), then the value eps is used in place of TOL. For
most problems this is unreasonably small. Unchanged on exit.
SVD _ LOGICAL.
on successful exit, SVD is returned as .FALSE. if the least squares
solution has been obtained from the QU factorisation of A. In this case A
is of full rank. SVD is returned as .TRUE. if the least squares solution has
been obtained from the singular value decomposition of A.
SIGMA - l/reat/l .
on successful exit, SIGMA returns the value seRT(r(transpose)r/(M-
IRANK)) when M.GT.IRANK, and returns the value zero when M -IRANK.
IRANK _ INTEGER.
on successful exit, IRANK returns the rank of the matrix A. Itshould be noted that it is possible for IRANK to be returned as N and.
SVD to be returned as .TRUE.. This means that the matrix U only just
failed the test for non-singularity.
WORK - I lrcal// array of DIMENSION (L\ /ORK).
on successful exit, if svD is returned as .FALSE., then the first
77
Page 88
N elements of WORK contain information on the QU factorisation of A,
(see parameter A above and NAG Library routine I |F}2WDF I I ) and
WORK(Nf 1) contains the condition number norm(U)*norm(U(inverse)) ,
where norm denotes the Euclidean norm, of the upper triangular matrix U.
If SVD is returned as .TRUE., then the first N elements of \MORK
contain the singular values of A arranged in descending order and
\MORK(N*l) contains the total number of iterations taken by the QR
algorithm. Otherwise WORK is used as workspace.
L\MORK _ INTEGER.
On entry, LWORK must specify the dimension of the array WORK
as declared in the calling (sub)program, where L\MORK.GE.4 *N. Un-
changed on exit.
IFAIL _ INTEGER.
Before entry, IFAIL must be assigned a value. For users not familiar
with this parameter, the recommended value is 0. Unless the routine detects
an error (see next section), IFAIL contains 0 on exit.
D. Error Indicators and Warnings:
Errors detected by the routine:
IFAIL : 1
On entry, N.LT.l, or M.LT.N, or NRA.LT.M, or LWORK.LT.4*N.
IFAIL : 2
The QR algorithm has failed to converge to the singular values in
50*N iterations. This failure can only happen when the singular value
decomposition is employed, but even then it is not likely to occur.
78
Page 89
Bibliography
[1] Prandtl, L. 1918 "tagflügeltheorie" Nachr. Ges. Wiss. Göttingen,
Math. Phys. K1., 1st part, 451-477;2nd part, 107-137.
[2] Thwaites, B. 1960 "Incompressible Aerodynamics" (Oxford University
Press, Oxford).
[3] Ashley, H and Landahl, M. 1965 "Aerodynamics of \Mings and Bodies',
(Addison-\Mesley, Massachusetts).
[4] Robinson, A and Laurmann, J.A. 1g56 "\Ming Theory,, (Cambridge
University Press, Cambridge).
[5] Bera, R.K. 1991 "The Lifting Line Equation - revisited", fnterna-
tional Journal of Mathematical Educøtion in Science and Technology
22,341-349.
[6] Varley, E. and \Malker, J.D.A. 1g8g "A Method for Solving Singular
Integrodifferential Equations", IMA Journal of Apptied, Mathematics
43,77-45.
[7] Tuck, E.o. 1991 "Ship-Hydrodynamic Free-Surface problems \Mithout
Waves" ,Journal of Ship Research, 95, 2ZT-287.
79
Page 90
I
I
I
I
I
1
I
II
I
,]
ì:J
J
I
I
.;
t,l
'.tl
'lI.i
[8] Stewartson, K. 1960 "A Note on Lifting-line Theory" , Quørterly Jour-
nal of Mecha,ni,cs ønd, Applieil MøthematicsL3,49-56.
[9] Guermond, J.-L. 1990 "A Generalized Lifting-line Theory for Curved
and Swept \Mings", Journal of Fluid Mechanics zLL,497-6L3.
[10] Abramowilz, M. and Stegun, I. A. 1964 "Handbook of Mathematical
Functions with Formulas, Graphs, and Mathematical Tables".
jl
:]
'1
80