MATHEMATISCH INSTITUUT
RIJKSUNIVERSITEIT
GRONINGEÑ
REPORT TW-111
ROTATING AIRFOILS
IN
STEADY FLOW
by Felix Hessda
ARCHL
Téchñische HogeschoI
Onderaf deli nische Hogeschoo, DJCUMENTATIE DATUM: oo C J N £ T A Il
t
ti
air sus
in
y
REPORT TW-1 11 by Felix Hess Completed March1972
This work as supported by the
Netherlands organization for the advancement of pure research (z.w.o.)
Contents.
§ 1 Introduction. 3
Chapter I. Motion of the system parallel to its plane.
§ 2
Winglet structures.4
§ 3 An example of a winglet structure. 11
§ 4
Linearized aerodynamic theory. 13§ 5
General character of the inteal equation and its solution. 17§ 6
The collocation method. 25§ 7
Tackling the singular integral. 28§ 8
Derivation of the expressions (7.12), (7.18), (7.19).34
§ 9
Some elementary "spanwise" integrals.39
§10 The resulting forces and torques.
Chapter II. Motion of the system under an angle r with its plane.
§11 Linearized aerodynamics.
45
§12 Derivation of the expressions
(11.12)
and (11.13).49
§13 Application to wingl.et structures. 53
§14 General character of the integral equation and its solution. 56
§15 Tackling the singular integral (first method). 58
§16
Derivation of the expressions(15.12)
and(15.16).
62
§17 Tackling the singular integral (second method).
65
§18 Derivation of the expressions
(17.28)
and(17.32).
71Chapter III. Numerical calculations.
§19 The method of nunieri2al integration. 71#
§20
The distribution of pivotal points.79
§21 A numerical comparison with Van Spiegel's theory for circular wings.
82
§22
The winglet structures used in numerical examples. 84§23 Influence of the accuracy of the numerical integrations. 88
§24 Influence of the distribution of pivotal points. 90
§25
A numerical comparison between programs B, CB and C.. 95§26
An experiinent. 99§27 Comparison between experiment and theory. 105
§28 Pictures of some lift distributions. 122
§29
References. 141This report presents a method of tackling the aerod.ynamic problem associated with devices such as boomerangs or helicopter rotors, where a number of airfoils perform a rotational and a translational motion at the same time, while remaining approximately in one plane.
The problem in which we are particularly interested concerns the case where
the shape and the motion of the airfoils are given and. the aerodynamic forces are
to be found. It is clear that, since the interaction between such devices andthe
medium through which they move is very complicated, a simplified model has to be
chosen, which can be taken as a basis for numerical calculations, in order to be
of practical use.
We have chosen the following model:
The medium is an ideal, inviscid., incompressible fluid. The theory is linearized, since in general non-linear theories present nearly insurmountable difficulties.
The airfoils are infinite in number, each having a vanishing chordlength; they
form a continuum filling a plane region S. We name such infinitesimal airfoils:
winglets.
In this way a continuous distribution of lift forces is produced. The motion
of the wingletsis such that it gives rise to a steady flow. Hencé the structure
of winglets as a whole can be looked upon as a sort of pervious lifting surface,
fixed in a steady flow.
A strip theory applied to the winglets gives the aerodynamic forces in terms
of the motion of the winglets and the induced velocity field. To this steady flow
problem lifting surface theory is applied, yielding the induced velocity in terms of the lift distribution on S. The resulting integral equation can be numerically solved by a collocation method.
In chapter I we treat the case where the motion of the undisturbed fluid
is directed parallel to the plane of winglets, and in chapter II we deal with the
case where this motion is directed under an angle with the plane of winglets. Numerical results are given in chapter III. The accuracy of the numerical
results is influenced by: i2 the number and distribution of pivotal points, and
22 the accuracy of the numerical integrations. In our method the second factor
is completely separated from the first one. The number of steps in the numerical integrations is determined by a preset tolerance requirement.
In the last part of chapter III the theory is compared with the results of experimental measurements.
§2 Winglet structures.
The systems in which we are interested consist of a number of airfoils, not necessarily straight, lying approximately in one plane, named the B-plane. (e.g. boomerang, rotor of a helicopter). Such a system moves through a fluid (e.g. air, water), which is originally at rest with respect to an inertial frame I.F. and which, for our purposes, can be assumed to extend to infinity.
We introduce a right-handed cartesian coore.iate system (x,y,z), moving with the structure of airfoils in such a way that this structure occupies a fixed region
fig. 2.1. Structureof airfoils.
with respect to the (x,y,z)-system. The B-plane is taken to be the (x,y)-plane. The airfoils thusl.ie approximately in the (x,y)-plane, their projections on this plane occupy a region s. (see fig. 2.1). The (x,y,z)-system, and with it the region S, moves with a uniform velocity U in the negative x-direction with respect to I.F. through
the fluid. (Hence the (x,y,z)-system itself is an inertial frame). In addition to
this motion the airfoils may have velocities with respect to the (x,y,z)-system,
which differ from place to place. The airfoils may also change their shapes and their angles of incidence.
/Y
)Xan airfoil-'
its projection on S 3 31 Z 1 2fig. 2.2. The local, coordinate system.
We also introduce a local right-handed cartesian coordinate system (1,2,3), which is related to the orientation of the airfoils (which, for the present, be a function not only of position but also of time). We do this in two steps
(see fig. 2.2). First we define the (1',23')-system. The 3'-direction is parallel
to the z-direction. The 1 '-direction is parallel to the (x,y)-plane, pointing from
the local leading edge towards the local trailing edge of an airfoil. The
2'-direction is parallel to the local spanwise 2'-direction of the airfoilts projection on the (x,y)-plane. The angle t3(x,y) is defined by identifying the 2'-direction with the direction of the vector (cosç, sine, o) as notated in the (x,y,z)-systeni. The spanwise direction of the airfoils deviates from the (x,y)-plane by a small angle y(x,y). Now we take the 2-direction parallel to the airfoilYs local spanwise
direction. Thus the 2-direction deviates from the 2'-direction by the angle y,
also the 3-direction deviates from the 3'-direction by the same angle. The 1-direction
is parallel to the 1 '-direction. The transformation matrix for the rotation from
the (x,y,z)-system to the (1,2,3)-system is:
Next we introduce the Ttrelative velocity" 'r(xY) which is the local velocity
of the fluid relative to the airfoils:
_,
-,
-,
-4Vr=V+V_W
where
x-direction
(2.2)
y = (vv,v) is the local velocity of the fluid with respect to I.F.,
-v = -(u,o,o) is the velocity of the (x,y,z)-system with respect to I.F.,
W= (WW11) is the local velocity of the airfoils with respect to the
(X, y, z)-system.
The components of these velocities are notated in the (x,y,z)-system. The components sine cosç3cósy - cossiny - cos sin3cosy - sin3siny o -sin7 cosy (2.1)
V = (U - W)sinJ3 W cosf3 . + y
sin3 -
V CO5V = [(u - W)cos
- W sin]cosy
(-w + y )siny+ cose +
+ vsin?cosy
V
1'3 - [(u - W )cosç3 + W sin]siny
+ (-w + V )cos
z z
The meaning of the dotted Une will be explained later on.
By definition of the 1-direction we have V O always. The function (x,y) may
have jumps of magnitude ir at points where Vn = O, and at these points an individual
airfoil may suddenly have its leading edge turned into a trailing edge and reversely.
Further we introduce the local "effective angle of incidence" of the airfoils,
cL(x,y). By a. is meant the angle between the direction the vector (VrIOV
), as
notated in the (1,2,3)-system, should have in order for the local lift force to
fig.2.3. Angle of incidence.
direction of zero lift
vanish (direction of zero lift) and. the actual direction of this vector. The effective angle of incidence cx. is the sum of the geometrical angle of incidence a0(x,y) and the angle:
-. V
a.1(X,y) =
arctan ...!
(2.11.)V
ri
(see fig. 2.3).
-We assume that the interaction between fluid and airfoils is determined by
the
following conditions:The airfoils cannot locally exert or experience forces parallel to their local
spanwise direction, i.e. the 2-direction.
The airfoils experience a local force in the (1,3)-plane, normal to the direction
of , i.e. parallel to the vector (-V,O,V). The magnitude of this force,
per unit of length along an airfoil, is given by
l(x,y)
=
i(V + y2), b(x,y) CL(x,y) (2.5)where p. is the density of the fluid, b is the local chordlength of the airfoils
and
CL their local lift coefficient.e) CL is taken to be proportional to sins:
Remark: for a thin airfoil in two dimensional flow the factor of proportionality theoretically is 2ic,see [i].
These conditions mean that we apply a strip theory to the airfoils. The assumptions AB,C appear to be valid for a reál straight, airfoil of large aspect
ratio placed in a uniform flow, provided that the angle between the spanwise
direction and the flow direction is not too small, and that effects due to viscosity may be neglected. (see .11]). We expect that the assumptions will hold approximately in cases where the radiüs of curvature of an airfoil is large compared to its
chordl.ength. Generally, the conditions will not be satified everhere on S.
Exceptions are for instance the regions close to the wingtips of the airfoils and the -region where the airfoils are attached to an axle or a central disk.
The aerodynamic prOblem presented by the sems as they are described so
far would be rather difficult to solve mathematically. A time-independent case
would be considerably simpler to deal with. This is our motive for approximating the structures of airfoils by the föllowing.mod.el.
.
Devide-the reg-ion S into subregions s, each centered around apoint (x,y).
The number of s should not be too small, and in the.neÌghboiirng s. the physical
situationshould. not betoo different. Hence we assume that thè number of air-foil,s is rather large and that thé local properties of the airfoils, determined
-
by,,y,Ç7
do not differ much at neighbouring points (x,y).. We define the local"filling factor" d by naming the fraction of s1 which is occupied by airfoils
d(x,y)... We shall, replace d, ,,y,bycontinuous'fuictions of (x,y) by replacing
every airfoil by a large number of similar airfoils with smaller chordlengths, in such a way that the local filling factor d remains unchanged. In the limit we have the region S occupied by a continuum of airfoils with vanishing chordlengths. We
shall, neme such infinitesimal airfoils winglets Such a system of winglets exerts
forces upon the fluid which are distributed continuously over S (to be moreexact,
over a two dimensional region close to s): (x,y) per unit of area. In general this
field of forces will be time-dependent. However, we shall. òonfine' our attention
to those cases, only where is independent of time. Hence wé only consider cases
where d(x,y), a(x,y), p(x,y) 7(x,y), t(xy) do not, have an explicit time-dependence.
From a somewhat different -and perhaps more rewarding- point of view we could say: the original airfoils have been "smeared out", so that at anr poit the
original situation has been replaced by a kind of time-average. The wiñglet model thus would correspond to a sort of time exposure of the original. airfoils.. If the original' system was periodic in time, its periòd shoúld be small cpared .o the
characteristic time D/U, where D is the diameter of S Thus even a two-armed
boomerang could be simulated b. a winglet model, provided that it would spin fast enough. A theory based.on the winglet model, of course, would at most yield
and on average velocities of the fluid.
The wingletz move (approximately) within the region S, but as far as the
interaction with the fluid is concerned, only the stationary field of forces -,
f(x,y) is of importance, and a steady flow results. For the magnitude of the
forces per unit of area acting on the fluid we have, taking into account (2.6):
a(x,y) (V2
+V ):
2ri r
where c depends on the local density and profile shape of the winglets. Comparing (2.7) with (2.5) we find:
f(x,y) = -.1c(x,y)sin
CL(x,y) c(x,y) = d(x,y)
2sin a.(x,y)
For thin airfoils we would obtain theoretically (see remark under C):
c(x,y) it . d(x,y)
We remark that the condition
(2.7)
(2.8)
(2.9)
(2.10)
which is part of condition B, follows from a general argument, which is valid if
viscosity is absent. Locally at the point (x,y,O) the winglets, per unit, of area
of S and per unit of time, do the work (x,y) . (?(x,y) - on the fluid. On the
other hand the fluid there wins, per unit of area and per unit of time, the energy '(x,y) . '(x,y). Therefore: '.
(i'-')
=
'. and(+ V-
).
The components of in the (1,2,3)-system are:
fi = - f3
f2 = O
2 2
f3=-IcsinV
Jv
+v
ri ri r3
After substitution of (2.3) and (2.ti.), (2.1 1) shows how the forces ? depend on the
fluid's velocity v. (2.11) gives a description of the interaction between winglets and. fluid which is a property of the winglet structure. On the other hand in section
a relation between. and will be derived which is aproperty of the fluid. This
relation will be a linear one, since we will linearize the equations of motion
for the fluid. An exact, non-linear theory would. be extremely difficult to work
out. *)
Under what conditions is the application of the linearized theory to a winglet structure justified? The angle of incidence a.(x,y) should. be small, this means:
wz/J( 2 2
u-w) +v
«1
Xy
«1
«1
The conditions can be forniulated more exactly this way: Take a system with U,
a(x,y), f3(x,y), y(x,y), d(x,y), (x,y) given. Now reduce
'
y and w by a small
factor e, and let e tend to zero. Then y, w are of O(e) and so will be
V ,
y , y f . However, f , f will be of O(e2). The essence of the linearizedx y z z x y
theory is that velocities and forces of O(e2) are neglected with respect to those
of o(e).
In (2.2) the velocity components of
o(i)
are denoted by capital letters,those of 0(c) by small letters. In (2.3) the ternis at the right-hand 'side of the
dotted line are of O(e). We shall now linearize. The sign between two expressions
means that the relative difference between these expressions is of O(e). (2.3) becomes:
y
(u
-
w)sin +
w coseri Y
V
(u
-
W)cos
- w siner2
YV
-,
[(u
-
Wx)cos- W sin]y - w
+We define the local "effective velocity" V(x,y) by:
.Jv 2 V 2 (U - W )sin
+ w
cosev
(2.1li.)ri r x
y
eAnd the angle
e by:
[(UW)cosWsin]yw+v
V +Sifl.+
(UW)s
o ein+W cos3
' eThe linearized version of (2.7) becomes:
T(x,y) - c(c,y) [0(x,y)
v(x,y,o)
V(x,y)
(2.16)V(x,?T)
fi - 3
e +
= o
f3 f
-4
The linearized components of f in the (x,y,z)-system are
As far as the linearized theory is concerned the fluid is acted upon by the force field
-4
f(x,y) (0,0,f(x,y))
situated in the region S of the (x,y)-plane. Substitution for ie and V yields: fz - ic{a. +
{(U-W)cos-W sinj3]y-w +v
Xy
z z(u -
W )sin + W cose) Xy
2x [(u -
W)sin
+ WY cosJX
(2.20)
This represents one relation betwe,en
Ç
and v, which is a property of the wingletstructure. The linearized aerodynamics in section i. will lead to an integra1 relation
of the form
v(x,y)
ffK(x,y,,i-)
f(,r1)ddr1 (2.21)S
The equations (2.20) and (2.21) together constitute an integral equation for the
function f (x,y).
In some cases be known before hand, but in other cases some
of these quantities may depend on the forces which the winglet experience from
the fluid. (for instance cases with elastic bending). In those cases one or more additional equations may be needed to determine the behaviour of the system.
(Bemeniber that we only consider systems giving rise to a steady flow). A siinple
example of such a system is discussed in section
3.
f X f y f z + .tc(a - tc(a. o - ic(a + a. e + a. e + a. e 2
+ )v
(a. + _!)sin e e e e 2+ _)v
(a. + _)cos e e e e 2 + e e (2.18)In this section a simple example is described of a system where Wz depends on the forces exerted by the fluid on the mechanism.
Up
fluid
fig. 4.1. Winglet structure as viewed
from
z-direction.The wingletsin this case are straight, have length R and are directed radially
butward from a centre 0. They rotate with a constant angular velocity w around the z-axis. The winglets are hinged at O so that they can move in a direction
perpendicular to the (x,y)-plane. (There is a resemblance to the rotor of a
helicopter). The angle by which the winglets deviate from the (x,y)-plane towards
the positive z-axis as viewed from O is supposed to be small (0Cc)). We call it 5.
The region S is a circle with radius R.
We introduce the polar coordinates (r,cp):
x
r coscp , y= r sinq (3.1)We thèn have:
?= (-
wr sincp , wr coscp, w(r,cp)) (3.2)
At points where wr + U sincp = O individual winglets interchange their leading and
trailing edges. Hence:
f3
=
p . si(wr
¡J sincp) (3.3)y =
. sign(wr + U sincp) (3.4)Ve
=
¡wr + U sinp (3.5)(For a definition of W,f3,y,V see section 2).
dS
w(r,cp) =
r -
=
wr
-dt dcp
Substituting (3.2) through (3.6) into (2.20), we obtain:
which can be ritten as:
cIS
f
-U8coscp-wr+v
dcp zf LC +
L o + }twr + U
smp]2
which together with the expression for v of the form
(2.21)
constitutes an equationfor f and. 8.
z
The ang1e 8(p) is determined by the condition that the angular acceleration of a wingLet times its moment of intertia with respect to its hinge at O must equal the torques exerted upon the winglet. f(r,p) being the force in z-direction exerted upon the fluid per unit of area, the torque per unit of angle cp exerted on the winglets by the fluid is given by:
M(p)
= - f
(3 .8)
The torque per unit of angle p due to the centriftgal forces is:
2
- 8(1)1
where I denotes the moment of intertia of the winglets per unit of angle cp. We
now obtain as the equation of motion for 8: 2 I -_ = M(cp) dt2 2 R
r.
dp 1w2 o(3.7)
(3.9)
(3.10) (3.11)Thus we bave a set of coupled equations (3.7) and (3.11) for the two unknown functions f(r,p) and 8(p).
It may be remarked that for a stationary solution of
(3.11)
the Fourierexpansion o± M(cp) cannot contain terms with sincp or coscp. This means that the load. distribution f(r,p) can have no first moments with respect to any line through the centre 0. This, of course, could be expected beforehand, since the hinges at O cannot transfer such moments.
§li. Linearized aerodynamic theory.
The behaviour of an ideal inviscid. incompressible fluid is determined by the following two equations (see [2]), the equation of motion:
dv-'
p = F - grad p
and the equation of continuity:
div '= O (Ii..2)
The used s,mbo1s have the following meaning: -9
y: velocity of the fluid, p: pressure in the fluid,
p: density of the fluid (constant), -3
F: external, forces per unit of volume acting on the fluid.
d
-The operator stands for + y
y
We shall, use a right handed cartesian coordinate system (x,y,z).
The-left-hand side of (1..1) is not linear in which makes an exact
treat-ment of the f'uid's behaviour very difficult. Therefore we shall work with a linear
-3 -3
approximation.. We assume F and y to be small of O(e) (for an explanation see
section
2),
and we neglect quantities o± 0(e) with respect to those of O(e). Thelinearized theory tends to be exact as e tends to zero.
Taking the divergence of both sides of (li..1), we obtain: -3
div = div - div grad p
(li..3)
As to the left-hand side:
with -9 dv d. . -div = (div V) a,b= x;y,z.
Here the first term vanishes because of (4.2), and the second term is of O(e2) and can be negl.ected. in the linearized theory. Thus we obtain, correct to o(e):
-3
div F = div grad. p J
(4.5)
This Poisson's equation yields for an infinite region (see
[31, p. 323):
va Vb
p(x,y,z,t)
=
fff -
div F(,,t)ddd
G
-4
We assumed that F vanishes outside and on the boundary of a finite region G. A more convenient expression for the pressure field can be obtained. Since
i -) 1_ -,
div - = - div F +
)-i-,tr itr.r
"ItT and-,
¡fi div
ddid =
= OG
where the last integration is performed over the boundary of G, we have
-, div F
-«J
1tr dr,d1d = fffddrd
G
and instead of (ii..6) we obtain:
with
p(x,y,z,t)
=
jtjjddd
G irr
'= (x - , y - , z
-From (ti..8) it can be seen that the pressure field can be regarded as a superposition of pressure dipoles.
From now on we confine our attention to cases where a field of external
-,
forces F moves uniformly through a fluid which was originally at rest. The cartesian coordinate system (x,y,z) is now chosen in such a way that its origin moves with the field of forces and the negative x-axis points in the d,irection of motion. With respect to this inertial frame we have a stationary field of forces in a
uni-form flow with a constant velocity U in the positive x-direction. This presents a
steady flow problem.
If the velocity of the fluid with respeet to thi system iE with
'= (u,o,o) of o(i) and
= (v,v1,v)
of o(e), the equation of motion (i.i) becomes-) -, _,
1-4
1(V.v)v + (v.v)v= - F - - grad p
and the linearized equation of motion is:
-+ -, i
(V.v)v = - F - - grad p
(11.9)
We are particularly interested in fields of forces acting on a finite region S in the (x,y)-plane, with the forces directed parallel to the z-axis. In these cases the field of forces can be represented by:
with
(E) denotes Dirac's delta function). The pressure
field. (!..8)
now becomes:-) -,
F(x,y,z,t) = f(x,y)5(z)
-4
f
=
(O,O,f(x,y))p(x,y,z
The first of the equations
i..iIi.) yields
v=--X
where we have used. the condition v - O for x -'
-.
An equation of the typev(x,y,z) with ! -*0 for x
-,-has the solution(x,y,z)
=
f
,y,z)dwhere again we have used the condition
-*
O for x -, -.
In this way we
obtainfrcmi (li..1!i.) and (ij..13):
f (,T)z
3 S4itr
with 2 221
r
= [(x-e)
(y-y)
+ z ]2
Differentiating under the integral
sign,
we obtain from (l..12):- - .... 3z(x-) - ! r5 f(E,,1Ì)ddr1 .2
__Ljj3z(-iì)
1r5
22
- -
j._jj
3z-r
(, )
ddî1z_ 1S
r5
Outside S i..io) reduces to:
ddr1
}4.11)
The equations 1..i6) give the velocity field outside S if the field of
external forces on S is known. For our calculations we need, however, in particular
an expression for v on S. From (li-.16) it can be seen that, while v and. v are
odd functions of z y is an even function of z. We define:
z
v(x,y,O)
= hm y (x,y,z) (li-.17)z-40 Z
From
the third equation
(ii.. 16)we obtain after perfonning the integration
with respect to :
def
K(x,y,,i)
=
i -v(x,y,z) 4-3TLV IIK(x,y,z,,i)f(,i)ddi
S with + 2 2 = 22 2
L [z + (y-r1) I [(x-e)+ () +
2-2z (x-a)
2 2 [z +(i) ][(x-)2
+z2]3/2
Substitution of z=O into (1.I..19) yields:
-1
2
{
2(.2o)
(i)
+ [(x)2
(Y-i)
]2 JObviously this expression has a second order singularity for r=y. Because of this
singular behaviour it is not permitted. to
simply
substitute z0 into.i8)
inorder to obtain an expression for the induced velocity v on S. A careful analysis
yields
(see for instance R-I):
i
v(x,y,O)
=
him y (x,y,z)=
z-,Q Z
where K is given by (1I..2O) andJ denotes the Hadamard/ principal value of the
integral with respect to i defined by
b
RPJ}
ft:ji)
di -f()]
¿I
f(n)
jthj=lim
2 2 ß-40 (r-î)a
ta
(Ii..18) (ii.. 19)(li..21)
v(x,y,z) =v(x,z)
-v(x,y,z)=
-1 ,,. zf(,i)dd
.16) 2 223/2
'S'
[(x-)
+ (y-n)
+ z ] X3z(y-)
If ! 2 2 25/2
d
S-
[w +(y'-r1)+z I
x-E 2- (3rT)j
fI ! 2 2 25/2
dw S-
[-r+(r-r)
+z]
In this section we shall investigate the general behaviour of the solution of the integral equation resulting froni (ii..21) and (2.20). It can be written in
the forni:
with
L(x,y) = P(x,y) - Q(x,y)
-f L
= -i
itU i L(x)=
P(x) - Q(x) -Swhere K(x,y,,i) is given by (i..2o) and where we have used the abbreviations:
2
V
P =c( -I-ct
U V eQ,=
c-fl-The meaning of the symbols has been explained in section 2. Note that Q, 0.
(5.1) is an integral equation (with singular kernel) of the second kind, and it may be expected that the behaviour of its solution near the boundary of S will
differ from the solution of the integral equation for an ordinary lifting surface, which is of the first kind.
First we shall, investigate the behaviour of the solution in x-direction. Some essential information can already be obtained by studying the siinpl,e
two-dimensional problem where the region S is a strip in y-direction: -1 x 1,
and where P and Q, do not depend on y. In this case (5.1), after integration with respect to r, reduces to:
+1
k(x,)
=
The solution of this integral equation can be explicitly found (see 15]), but for our purpose it is sufficient to consider the case where P and Q are
constants. We then have:
+ (x,y,,i1) L(F,,,1)ddi (5.1)
K(x,) L()d
(5.2) (5.3) (5.5)Its general solution is:
[5]
1 +1 1L(x)=
E' [f
(\-arctanQ
1-a)
.(1-).r-
+c'}]
l-I-jQ2 -1 (5.6)where C' is an arbitrary constant. Computing the integral in (5.6) yields:
1 L(x) P (i-x ' arctan-Q +
Ji2+x)
.{i 1-xJwhere C is an arbitrary constant. Thus L(x) has a singularity at x-1 ('pleading
-s,
edgett) of 0[(1x) U] with
i i
s1
= -
arctanQand a singularity at x.1
("trailing edge") of O[(i-x)5]
with1 -' arctan-Q
If Q tends to zero with P/Q remaining finite, we obtain the solution
2P fi-x C
L(x)=----.
(5.8)
(5.9)
(5.10)
which corresponds to an
ordinary two dimensional flat plate placed in a homogeneousfLow under an
angle P/Q.For each value of C
(5.7)
represents a different mathematical solution to theintegral
equation (5.5). A definite choice for C has to be made to obtain the unique solution corresponding to the physical system to which we want to applythe theory. In the case of ordinary lifting surfaces the choice
C
=
0
(5.11)is generally made. It corresponds to the Kutta condition: the trailing
edge
singularity vanishes. This condition is generally
found
to bein
agreement w:ithexperiments.
Further down in this section it is shown that porous liftingsur-faces obey
an
integral equation similar to ours. It seems plausible that theKutta condition would apply to these pervious lifting surfaces as well. We decide to make the choice (5.11) also for our winglet systems, but we recognize
We then have:
L(x) P
(i-x
arctanQ
1+x}
This solution has the leading edge singularity of order (5.10), but near the trailing
edge it tends to zero of OE(l_x)zt] with
zt = - arctan-Q, (5.13)
(5.12)
resultant flow through surface
fig.
5.2.
Porous lifting surface.integral equation
(x,y)
L(x,y) = a(x,y) - i L(,i1)ddy (5.15)s
where K(x,y,,i) is the same as in (5.1). The two-dimensional case is treated in [6].
Thus a winglet structure can be exactly simulated by a porous lifting
sur-face with angle of incidence
V
a(x,y) = = o+OEe)
u
(5.16)
and porosity coefficient
o-(x,y) =
=
(5.17)
a z
We see that the limit Q, - corresponds to a vanishing porosity coefficient, Figure 5.1 gives an impression of the general behaviour of L(x) for different values
of Q.
Returning to the original three-dimensional equation (5.1), we see that the behaviour of L(x,y) near the boundary of S varies from place to place with Q(x,y) hence with the local density of the winglets and their local effective velocity V.
Thare exists a remarkable analogy between winglet structures and porous
lifting surfaces. With a normal non-porous lifting surface placed in a homogeneous flow the induced velocity must be such that the resulting flow is tangential to the lifting surface. With a porous lifting surface, howevér, there may be a leakage
through it. If this leakage is taken proportionalto the pressure difference between the two sides of the lifting surface, the induced velocity satisfies the equation:
v(x,y)
= -a(x,y) + cr(x,y) L(x,y) (5.11i.)
Here a denotes the local angle of incidence (see fig.
5.2)
and o the local porosityAfter substibition of -i-1 z
v(x)l
f
L()
-1 (5.19) can be written as 4-1 +1P P
L(x) L() 2tJJ
x-Fddx
-1 -1which is easily seen to vanish. We stress that these integrals should include possible singularities at the leading and trailing edges.
The energy taken from the fluid at the leading edge thus must equal the work done on the fluid between leading and trailing edge. This per unit of time and per unit of length in y-direction, is given by:
=
-u3
f
L(x)(5.20)
(5.21)
and the integral equation (5.1) tends to one of the first kind. On the other hand, in the limit Q - O the integral equation tends to the algebraic equation:
L(x,y) = P(x,y) (5.18)
In this case the induced velocity is of no importance. (In fact essentially (5.18) was used in a simple theory for boomerangs [7:1).
Because of the leading edge singularity a concentrated "suction" force of 2
o(e )
acting on the leading edge will exist. We assume this to be a local phenomenon,depending only on the local singularity in the lift distribution. Therefore the energy associated with "suctioiiY forces taken locally from the fluid at the leading edge can easily be calculated by considering again the simple two-dimensional case. The total work done on the fluid in this ('ase is zero. This follows from a
general arg.mtent: the total work done on the fluid, per unit of time and per unit
of length in T-direction is given, correct to
O(E2)
by:+1
s
v(x)
1.LU
f
L(x) dx (5.19)where the integral does not include singularities at the leading edge itself. L(x) is given by (5.12) and v(x)/U follows from
With a two-dimensional porous lifting surface the suction force is simply
directed in the negative x-direction; its strength per unit of length in y-direction is given by:
Table 5.1
L
=
P
+ QThis leads to:
+1 ,.., 2 2 = -.IU3
{
(i-X 'Ç
arctanQ
(!
)
arctanQ -1 ç.,..i..Ji j2
T,)
- i +which, after integration, yields:
122
-
i22tQ
(1)
2 1 arctan-W
22
2o--f=2U.
2 2 (Ii.o +1)where we have used the relations (5.16) and (5.17). The factor
2 1
arctan
-A_It
2 2
(1,.o. +1)
contains the porosity dependence of the suction force (A1 for the non-porous case). That the strength of the suction force decreases rapidly with increasing porosity cari be seen from this table:
(5.
25)(5.
211.) (5.25) (5.26) a= -
arctanr
) o Ao
o O .1 0.65 1.5!i. 0.002.2
1.11.50.69
0.011.8.3
2.75
0.36
0.256
.11.6.16
o.i6
0.655
.5
o
TFor a three-dimensional winglet structure we have to use a modification of (5.2!i). Suppose that in the "chordwise" expansion of the lift flincton L(x,y) the local strength of the singularity is represented by a term
x (y)-x arctan {Q(x1(y),
ï)1
L*(x,y)
=
c*(y) (
t
)
(5.27)XXl
where x1(y) and x(y) are the x-coordinates at the leading and. trailing edge respectively. We thenmust replace the factor
2
P
12
in (5.21i), which originates from (5.12), by c*2(y). Now we obtain for the local value of W:
3 arctan-Q
c*2
W
= iU(xt-x1)
1
where aU quantities have their local values.
Denoting by -f, -f1 the cartesian components of the local suction force per unit of length in y-direction, we have:
r (w-u)r
W =-W
Ex
x
sy y
with W from (5.28). We also have, because of condition A in section 2, or (2.18):
f cose + f sine = 0
(5.30)
sx
sy
Thus, once c*(y) is
known,
f and f51 are determined by(5.28)
through (5.30).It is perhaps somewhat surprising that the lift distribution over a winglet structure should have leading edge singularities at all, the boundary of S being the locus of the wingtips of the individual winglets. It is true that systems
consisting of a finite (though possibly large) number of airfoils have a lift
distribution which vànishes towards the wingtips. Our theoretical winglet
structure, however, is a limit case, and it might be misleading to think of it as
being composed of individual winglets. It resembles a porous lifting surface.
Let us now consider the behaviour of the solution of the integral equation
(5.1) in y-direction. Since we are particularly interested in cases where the
region S is circular, we shall consider the case where S. is a circle with radius 1. obably we can obtain some essential information as to the behaviour of the
lift ftnction near the "wingtipst' y =
ti,
by considering the simple case where(5.28)
=
L(î),
independent of . The integral in (5.1) then becomes:L(i1)
2f
(y)
2 ..(x_)2
()2}
dd
We choose x=O. Then the integral with respect to yields simply 2.J1_q2 and. (5.31)
becomes:
where we have put
a/i2 L(i1) =
r()
(5.33)
and made use of the assumption that r(1)
= r(-i) =
0. (5.1) now becomes (for x=O):2
(y-.q)
dr(r1) dr=
-dr1yi
r(y) i ' dr(n) di P(0,y)+ Q,(0,y)
-2J1_y2 = !l.it dr y-yThis integral is of the type which is examined in [5], chapter 17. By using the
results given there, it is easy to show that, in the special case in which P(0,y)
and Q(0,y) are contants,
r(y)
=2..Ji-y2.L(y)
=0(..J1-y2)
(5.35)
for y close to ±1. We assume that
(5.35)
will generally hold if P(x,y) and Q(x,y)are functions corresponding to actual winglet structures.
Finally it may be remarked that the method for solving the integral equation
(5.1) outlined in the following sections obviously can also be applied to porous lifting surfaces satisfying equation (5.1!,.).
(5.31)
(5.32)
(.3".)
+§6 The collocation method.
From now on we shall, consider systems where the region S is circular. The
structure of winglets performs a translational motion and a rotational motion
around an axis through the centre of the circle parallel to the z-axis. A system
like this is described by the equations (Ii-.21) and (2.20). In our problem the
motion of the winglets is given and the forces on the winglets are to be found.
Equations (2i..21) and
(2.20)
together constitute an integral equation for theload distribution f(x,y), which we have to solve. This integral equation can be
written in the form (5.1)
L(x,y) = P(x,y) - Q(x,y)
-
K(x,y,,i1)L()ddi1 (6.1)where the meaning of the snnbols is given by (5.2). In this equation, the kernel
function, given by ()4.2o), is singular. The functions P and are known, they
depend on the structure of winglets and the way it moves.
In the ordinary lifting surface theory the integral quation is of the first
kind; (6.1) however is an integral equation of the second kind. Yet our method
for finding an approximate solution of (6.1) will not be essentially different from Multhopp's collocation method [ti.].
Suppose that L(x,y) can be expanded in a series of the te
with
where G.() and H.() are known functions and. a.. unknown coefficients. Then the integral equation (6.1) reduces to
a. .G.H. = P -
131J
This presents an equation for an infinite number of a.., which must be satisfied every where on S.
If we approximate L(x,y) by taking only a finite number, say n, of terms in the
expression (6.2), (6.3) presents an equation for the n unknown a.., which in
general. cannot be satisfied in every point of S. The a.. can then be determined
by a least squares method. The expression
G.(x)H. (y)
a.. JJ'KG.H.dd.7
J
(6.2)
where g(x,y) is a chosen (positive definite) weight function, should. be minimal as a function of the n a... In general this method requires an excessive amount
of numerical work, but a relatively simple way to determine the a1 is possible
if g is chosen to be a sum of n 3-peaks. The weight
function
then vanishes everywhere on S except for n distinct points where it is positive. With this special
choice for g(x,y) (6.Li.)
is minimal
ifs
rpr n
Jjlzza
[GH
J
sij
n a. .[G.H.j
13 J. 3at each of the n pits.
This method is called the collocation method and the n points are called
pivotal points. For each pivotal point (6.5) gives one linear equation for the a..;
we thus have in stead of (6.3) a set of n linear algebraic equations for the n unknown coefficients a... These can be solved numerically on a computer. The
2 13
computation of the n integrations (one for each pivotal point for each term in
the expansion of L):
if
KG.H.ddr1
s
will present the bulk of the work required for this method.
It is obvious that the distribution of the pivotal points over S should be
ttreasonablett (e.g. not all, of them clustered together in a small part of s), but
it seems very difficult to give definite rules for an ttLptimal" distribution.
A convenient choice for the coordinate system in our case is indicated
in fig. 6.1. The lines of constant are parallel to the x-axis and. the lines of
constant 3 are lines of constant "chord
fraction".
Some of these lines are drawnin the figure, which resembles a picture of a world globe with meridians and
latitude circles drawn upon it. The pivotal points are chosen on the intersections
of the lines. A definite choice for the loatioi of the piVotal points shall be made later in section 20.
çrKGHdd) -
p]
g(x,y)ddi (6.1k)JJ
13
S
(6.5)
fig. 6.1. Possible distribution of pivotal points.
§7
Tackling the singular integral.
We start from equation ¿)-i-.21), which for a circular region
S with a radius a
takes the form
or:
with:
lxU..v(x,y,O)
=
lI.ic v'(x',y',o
where
=
x/a
,
y'=
y/a
'
=
/a
,
ti'
=
ti/a,f 2 Vf
V/j ft
= z +a -fsja2-112-('ti)
2-
('-ti)
+1 -f..i1-TI'2-.
If
('ti')
-1 _V.tit2
are
dimensionless quantities. For convenience weshall drop
the primes from nowon and write: X,Y,,ti,V,f.
We introduce the following symbols:
qJl i12
X=
X X = sJ1_ti2 Y (7.1) + 2 1[(x-c)
+(rti)22
x?-,
+(7.2)
2 ('-ti')}
7.3)
17.
X'
and X
will be usedas
new variables instead ofand X, while X and Y are mere
abbreviations. In (X',ti)-space the circular region S is transformed into a square.
We expandf (,ti) in
a series in ("chordwise") direction with coefficientsdepending on the ("spanwise") coordinate
ti:
M-1
=
Ea (ti)H (X')
(7.5)
The function -(it-cp) replaces the customary cotgç) term of the Birnbaum series. The cotgp guarantees the correct behaviour of the load function near the leading edge of an ordinary lifting surface, while each of the terms yield the correct behaviour near the trailing edge in that case. Unfortunately, we cannot make such a claim for the series expansion (7.6) inouï case. On the other hand, since the behaviour of the load function generally varies from place
to place near the boundary of s (as explained in section 5), it is impossible
to obtain a series expansion of the type (7.5) which would yield the correct behaviour everwhere near the boundary of S. We found by actual numerical calculations that the customary seriès with the cotg-cp term gave satisfactory results only in cases with large values of Q, (as could be expected, see section
5).
Omitting the cotgp term, however, gave rise to solutions with a tendencyto build up a peak near the"leading edge". This peak, in particular for higher values of Q, could only be formed at the cost of a some times strongly oscillating
solution. The choice of the function (ic-cp) is a compromise which seems to give
better results than the two possibilities mentioned above. As to the "trailing
edge", our series (7.6) yields the correct behaviour only for large values of Q,.
This is not so serious, since the load fuction vanishes there any way.
Note that, since the function (it-cp) remains finite at the "leading edge"
(x'= -1),
we have no suction force to take care of explicitely. More about this is said in section 10.H(X') = sinpcp , p 1,2,...,M-1
p
H (X') = 1()
=o
X' = -coscp
The functions a(i) are expanded as N
a(r)
=
a1 G1(i) with sirJ G1(î) = sine l 1,...,N=
COS,, This expansion can also be written as:with
N
a*(i)
a(t1)..J1_iì2 E a1.h1(i1)11
= sinli}7.6)
(7.7)}7.8)
J(7.9)
which is in accord.ance with (5.35). Substitution of (7.7) into (7.5) yields the
series expansion for f(,r):
N M-1
E E a1G1(i1)H(X')
1=1 p0
which corresponds to (6.2).
After substitution of (7.l,) and (7.5) into (7.2) we obtain
-c-1 *( \ -i-1 M-1 a rj X_X? i
v(X,y)
=
E 2 L [(x-x')21}Hx?'d
(7.11)The expansion of the integral with respect to X? in (7.11) near o (or iy)
contains a term
-Y21Y1
(7.12)(for a derivation see section 8), which should be taken care of. Therefore we
write
(7.11.) in the form+1 1'. M-1 a.*(y)
r
rr
1+
11-H
(x')dx' +2 L
[(x-x')
+]2
J P 1 JL \Á., y / = '-' 2 2 +1 2 -M-lfdH\
( a \
(y-)lflIv-it
Idi1 - E
Ç .)
1 2
j
a*(i1) lny-,1d1 \dx')
2 X l-y J ç?Xi-y1
Here we have replaced the factôr ?tnIYf from (7.12)
by
lTl 2
2 1nI-ìI
l-y
(7.13)
which behaves essentiaUy the same near Y=0. The expression between rectangular
brackets in (7.13) is twice differentiable with respect to
i1 at ipy.
Introducing the functions:
=
2 H(X')dX'(dli)
2lflIi1I
[(x-x')
+ y J2 x l-y g(x,y) wecan
write (7h13) as (7.10)M-1
4t
v(X,y) = E
p=O
Next we define the function by:
f (X;y,r)
= f(x2)
+ (n-y) [
s
]
+Near =y R(XTÌ) contains a factor
(_)2
and.:!! R(X,y,,1) S(XY1-I) 2
(y-n)
with:T(X)
=¡[(
Further we have: =2f
H(X')dx'(7.19)
f.(X,y,r) -1 2y L JII=
l-y2xH(x)
PThe first expression follows directly from
(7.14)
and the second one is derivedin section
8.
Now, using (7.9), we can write
(7.15)
as:a* () 2 + g(X,y) (y-n)
_(\\
lax'
jr'\
dx'IX] X'-X
¡1
for Oa*(1))1nIy_d
(7.15)is continuous at iy. In section 8 an expression is derived for in the Um:t
'1-i y: 2
¡dH
2S(XY,)
i p x'il
2 X + +lflí2(1_X1_} :
=
2 (X) + 1_2 '
y
2 l-y 1- y 'XL 1-y for p1,...,M-1(7.16)
(7.17)
7.18)
4ic v(X,y)
= Then we have: N M-1Va
LP
1c1 Q h1(,1) 2 th1 +()
[
JI
The integrand. of (7.21) vanishes for
0 andit.
As a final expression for the induced velocity v(x,y,O) we now have:
NM-i
= ap1[b1(I) c1(y)I
10
L + (y) 5(X,y) +{
f
+f
}
S(X,y, -cos)sinlsind
o ewhere b1(y), c1(y) and d(y) are given by (9.9), (9.10) and (9.il).
h1 ()
d,' +
Finally we have to substitute the expressions (7.6) for the functions H(X') into our formulas.
We put X = -coscl (7.20) + 5(X,y)
f
h1()
!y-ld
+f
h1()
The first three integrals in (7.20) can be evaluated analrticalLy. This is done in
section 9. The last integral has to be evaluated numerically.
In order to avoid being forced to compute its integrand for too close to y,
we choose y itself as an integration point and write the integral as:
f
={
f }
(7.21)
with
COSi ,y
=
-cose(7.22)
f
H(X')dX' 2 (cossinp - psincos)
p2. .
.M-1 -1p-1
lt
= pf
[(+coscp)2 cospcp dcp olt
ltr
-'
221
=
+ J
[(X-i-coscp) + y J2 coscp dcp o=
-
sin -(lt -
)cos)The integral in the expression (7.lLt.) for f has to be evaluated numerically,
but it can be simplified by integration by parts:
+1
f
{12 1
(X')dX' =
+ YlJ
lt -' 221
=
i [(x+i) +Y ]) -
f[(i-cosp)2
dq,
oBy the method outlined in the sections 7 through 9 the induced velocity at
each of the N.M pivotal points cam be expressed in
terms
of a linear conbinationof the N.M unknown coefficients a1. This reduces the integral equation to a set of N.M linear algebraic equations fòr the N.M unknowns, which can be solved easily with the help of a computer.
(7.25)
7.26)
=
-
sin2) p=1p=2...M-1
§8 Derivation of the ecpressions(7.12, (7.18), (7.19).
We want to investigate the behaviour of the "chordwise" integral in (7.11) in the limit ri -*y. Consider the expression:
Using X X
fH(X')dX'
=
f
H(x')dx'
+ (i-x) H(X)w-x) ()
+ o(y3) (8.4) -1 -1and
+1I{1
X-X,-1
+ [_xt)2
21}H
(x')'
p it can be written as +1 +1X-X? -
X-X' X-X't'f{
22'
-iIx-x'I
y
[(x-x')
+ II-x'I
The first term of (8.2) equals
X
-f H(X')dX'
x
x
LV X 2 3X-X=
-
-
y+.
y 0(Y)
..Jiri2
.Ji-y
J-y2
i-y2
we obtain for (8.3): r., X X 2 P
=
J
%(X')dX'
2yXH (X)
+f
H(X')dX'
Y-1
i-y2
P22
H(X)
yX
i-y2
+i2()X+0
The second terni of (8.2) equals
ra
r
-
sign(X-X')H(X')dX'
J
"
2 2 r.' 2 (x-x') + + x-x'J [(x-x') +y
]2 (8.1)H(X')dX
(8.2)
(8.3)
(8.5)
(8.6)
(8.7)
We expand H(X?) in a Taylor series:
Making use of (8.5) we can write
[(i-x)
+ (x'-))
E (x'-)3(
x\fl-j flj flj+1\
L
j)
1-y2)
Y
+0Y
(8.9)We thus have to consider expressions like:
which,
+1
rp0,l,2,...
Y-40 '- (x'-')2 + y2 IX'-I [(x'-)2
q= 0,1,2,...
The integrand in (8.10) is bounded for p 2, hence the expressions (8.10) vanish
for p 2, q. 1. For p 2, q. = O finite contributions are obtained. These
contributions to (8.7) can be taken together in the form:
him {H (x')-H (x)-(xT-x)(
Y-40_1
p
pTaking the limit yields
+1
I
{H(X') - H
(X) - (x'-x) (
)}
2(x'-X)Ix'-xIdx'
p p
Next the case p
=
1:yq-2
f{
Ix'-Integration yields
"t
since X = X
+ 0(Y),
can be written asV
-
Y2 + Y21n 1.i.(1-X2Y-40
03 H (X') = H (X) + E [(i-x)+ (x'-))
(d
p pr1
n!tflJ
"im Y-40 (x'-)sign(X'-')dx'
r-, sign(X'-X)dx')}
(x'-)2Ix'-I[(x'-)
2 2J
+y
(x'-x)
2 í(x'-x)+ y
Ji2
2-
Y hziY + o(Y (8.8) (8.10) (8.11) (8.12)(8.13)
"t
2 2 2x'-x)[(x'-x) y iy
ln,X?_X{(XT_)2Y2]l}
(8.i)(8.15)
This vanishes for q 1. The only contribution to (8.7) originates from q = O, it is given by
(Ofl\
)
f
lY
-+ ln
(1x2)}
(8.16)
Of the expressions (8.10) we finally have to consider those for which p = 0:
+1 2 (8.17) him q-2
r
{si(Xt_)
-[(x'-)2
+ y ]2J Y-40 -1 Integration yieldslun
q-2
Y-, Owhich
can
be written as1yq_2i.[ Y
2Y-40
This vanishes for q 1. Again the only contribution to (8.7) originates from
q
=
O, it is given by-x
H(X)
2p
Taking together (8.6), (8.12), (8.16) and (8.20), we obtain for the expression (8.1) in the limit Y -,0: X i
2y XH(X)(
f
H(x')dX'
y1_
p - 1-X2I y2
2 1 Li-y
V)
dV
)I
IXt-XI(X'-x))
H(X) +
(8.18) (8.19)(8.20)
+o(nnIYt)
(8.21)From
(8.21) it can be seen that indeed the expansion of theintegral
in (7.11)contains
a term (7.12).
We now obtain for f(XY,ri), as defined by (7.11i.) in thelimit r -by:
{çxi)
- H(X) -(V-x
2 2
(y-i')
(Y-'l)
=
2
+ dX'2'
[nI
+lnil_y2)
+ O((y-)3lnIY-ri1)
X
l-ri
(8. 22) where [...) stands for the expression (8.21).
. Only the second term in (8.21) contributes to
23r (X,y,r)
=
2 X H(X) O((y-ri)1.n l-y If p 1 this equals: 1 +1(p
1Pdx'
(ip
-()j
x-x'
X'-X+
j
'yand (7.19) follows immediately.
Substituting
(8.23)
with (8.22) into (7.16) and taking the limit -,y, we obtain(dJ{p
S(XYY)
2 F(1+2y2 1H
(x){2 X2-4ln[1(1 _x2)(1_y2)]}
i-y L i-y
i_2)
+1
dx'
+
f
-
H(X) -(x'-x)
(x'-x)Ix'-xIl
-1
The integral in (8.21i.) can be simplified as follows. Integration by parts leads to: 1 +1
r,'
(dHp
dx' X+H(X)
2 p X (dJI1+H(X)
2dJ('
p
+ dx' 1 sinpcI cospp dq(
)f
= sine +Ç
with X = -cose.The second term of (8.26) has to be evaluated numerically.
If p = 0 (8.25) equals
The first term of (8.26) can be evaluated analytically, it equals:
H(-1)
H(1)
2(1+x) 2(1-X) (8.23) (8.25))
(8.26) + (8. 2h.))ln (
(8.27)The first term of (8.28)
equals
1 1 2___
lnhi.(i-xrdH
X (o\
ir + -1 idx'
(X) j +i-x2
riJ
'-j
°
dx' Ix
li-(1x)
With these results (8.2ti.) can be written in the
form (7.18).
(8.
28)
§9 Some elementary "spanwise" integrals.
In this section we shall evaluate the first three integrals with respect to q occurring in equation (7.20), and some other integrals as well.
We have b1(y) = +1
-4 (Try)2
j
Try
dr1 =/
ILt (ri) = sinl
= -cose
Further we put
y
= -cosePresently we shall obtain expressions for some integrals which may be used
both in chapter I and in chapter II.
sinl sin
(cose - cos®)
-sinl sin
cose - cos®
h1(ri) lnIy-dri
=f
sinl sine in! cose - cos®ld(9.2)
(9.3)
(9.1.i.)
h(ri)dri
=f
sin1 sine d (9.6)f
=
f
d=f_sini
sine cose d (9.7)g1
f
(ri) df
sinl sin2 d (9.8)-1 0
The first three of these integrals can be reduced to Glauert's integrals
(see for instance [8],
p. 173),
(9.14.) directly and (9.3) and (9.5) after integrationby parts. We obtain
sinle
bL(Y) = -in
s in®
1f
1+1 l-
ìifl2
if i
(9.6), (9.7)
and (9.8)
yield easily:e =0
=1t
= O=:1c
= (1,-2)
1(12)
=
-{ cose - 1n2} if i i if i = if i 2 if i=
2 if 1 is even if i IS odd J(9.1i) J(9.12)§10 The resulting forces and torques.
In this section expressions are derived for the resulting forces and torques
fig. 10.1. Forces and torques acting on system.
acting on the structure of winglets. We are particularly interested in the
foUowing six quantities (see fig. 10.1): the forces in x-, y- and z-directions and the torques around the x-, y- and z-axes. The dimensionless equivalents of
these will be evaluated. In order to obtain the actual forces ¡nd torques one
should multiply the dimensionless quantities by 1.iU2a2 and 1U2a3 respectively. (a is the radius of the winglet structure).
The lift:
LFZ = ¡f - f (x,y)dxdy (10.1)
s
The torque due to the lift distribution, with components:
LTX = ¡f - f(x,y)ydxdy (10.2)
and
LTY = ¡f + f(x,y)xdxdy (10.3)
s
The force in the (x,y)-plane, with components:
DC
=f!
-
ç (x, y) dxdyand
D'IZ
=
If
(+ f(x,y)y - f (x,y)x)dxdy (10.6)s
y
In the first three of these integrals we substitute the series expansion for f(x,y) as given by (7.5) through (7.10):
N M-1 +1 +1 LFZ =
a1
f h.1,(y) f
H(X)dXd3r N M-1 +1 +1 LTX =a1
f h1(y)y f
H(X)dXdy
N M-1 +1 +1LTY =
-a1
f
h1(y) ..J1_y2f
H(X)XdXdYAfter evaluation of the integrals with respect to X we obtain:
L2 =
LTX=
LTY=
L
DFY= If -
f (x, y) dxdy ywhere f and f are defined by (2.18).
X Y
And the torque around the z-axis:
-(a i-a 2 ol It'
(- a1
12
L=-It(a
i-a4
01 11 LDC=-It(a
12
i-a ) 02 12 LTY = 7tç,
(a
+ h1 (y) dy h1, (y) dy+ al)
f
(y)1y2dy
The integrals with respect to y are evaluated in section
9,
and. we have(l-2) 1(1+2)
11
i odd. (10.5) (10.7) (10.8)(10.9)
(10.10) (10.12)+ a1) j
N M-1 -f(x,y)
=
EE b1G1(Y) H(X)
1c1p0
NM-i
-f7(x,y)= E
Z c1G,,(7) H (X) p 1=1 p=0 N M-11
1
DFX=
b1
f
ktcsr)f
H(X)dXdy
N M-1 +1 +1DFY = c
f
h(y)
f
H(X)dXdy
N M-1
-i-1 +1IXIZ = {_b.1, fh1(7)ï. fH(x)dxd +
The integrals can be evaluated, just like those for the lift distribution, and ve obtain:
When the integral equation for the load function f(x,y) is solved by the
method outlined in the section.s6 through 9, the induced velocity v can be calculated
either according to the algebraic equation (2.20) or according to the integral
equation (Li..21). The results will differ, since the solution for
Ç
is not exact.Only at the pivotal points there is agreement, and only there we can apply (5.18).
In order to evaluate the integrals in (10.4), (10.5) and (10.6) we shall expand
the functions f(x,y) and fy(XY) in a series which is quite analogous to the
expansion (7.10) for f(x,y). The coefficients are determined by the requirement
that the series have the calculated values at the N.M pivotal points. Thus we put:
The equations (10.16) and (10.17) have to be satisfied at each of the N.M pivotal points, where
f and f are
known.
There result two sets of N.M linear algebraicequations for the unknown coefficients b1 and. c1 which can be solved by straight forward methòds.
The expansions (10.16) and (10.17) can be expected to correspond to a y
calculated froni (2.20) rather that to a V calculated from (1i.21). (1i..21) would
yield a v having a logarithmic singularity at the leading edge. The "SUCtIOn
forces" would then be shifted into the interior of S, so that they would be
incorporated. in the forces and f7. By using (10.16) and (10.17), however, strong
fluctuations in f and f near the leading edge cannot be taken care of very well,
and we may expect that "suction forces" will not be taken into account properly
If we accept these series expansions for f(x,y) and f(x,y) we obtain:
(10.18)
(10.19)
DTZ = -
02+b
12) N(c1
c2)
Ei=1
(i-2) 1(12)
1, odd
(1 0.22)
(lo. 23)
Ak'Ii1' II. Motion of the system under an angle r with its plane.
§11 Linearized aerodynamics.
In this chapter we will consider the case where a field of external forces pointing in the z-direction is moving through an ideal fluid with a velocity having a component in z-direction. The field of forces is the same as the one considered in section 1i-. The conditions under which this model can be applied to winglet structures are discussed in section 13.
Again we choose a cartesian coordinate system (x,y,z) in which the field cf forces is at rest, and we have a steady flow problem.
fig. Ti.!. The system considered in this chapter.
Let the undisturbed velocity of the fluid with respect to the (x,y,z)-system be:
V = V.(cosjr,O,sinjr)
i with
v>o
,
o<J
it/2J
and the disturbance velocity field:
y = (y ,v ,v )
X
yz
The meaning of the symbols is the same as in section 2. As before, we are particularly interested in v. An equation of the type
A + B
= !(x,y,z)
with cp-*O, -4O for (x,y,z) .-,-c(A,O,B) has the solution:
(x,y,z)
=
f
y(x + AT,y,z + BT)dTHence we obtain from (11.2)
v(x,y,z)
=
F(x
+ V cos.T,y,z +V sin.)d
with
-co
Again we consider the case where (to O(s)):
(x,y,z) = (O,O,f(x,y).6(z))
is given by (k.13). Now we
can
write (11.3) as:o
v(x,y,z) =
f
+ V cosllI.T,y) 6(z + V sinilr.r)d'r +-co 0 2
3(z
+ V sinljr..r) - dT R -+ V cosjr.,y,z V sin'ir.í)d.r (11.3)The first
term
of(71.5)
equalsR
=
[(x - + Vcostr..r)
+ (y -q) + (z + V sin*..r)2] J.1p
----1p
1p
(11.2)dz
it V cosijr + Vsinir
=
F iV cosr-+V sinc-=-F
i____________
ivIsinivjf (x
COS1J( z siriir z,y)if
z.sinir> O and O ifz.sin'jr<QUsing the abbreviations:
X
=
- x,
Y=
-
y, Z=
-z, T=
-VT, C = CO5%fr,s
sinir,
2 2 2
R
=
[(XTc) + y+ (Z+Ts) ]2
we obtain for the second
term
of (11.5):1LV
So
[2(Z+Ts)2
-
(X+Tc)2
-
Y2]R5
dTThe integral with respect to T yields
2-
2 2Z Z
Zs-Xc
cK(X,Y,Z)=
3+
22+
2 2(p+o)p
(p+á) p
(p+a) p
(p+o)
with
1(11.9)
2 2
21
p={X Y
+Z
P,
o=Xc+Zs
We need an expression for v(x,y,O). 'Substitution of zO
in (11.9) leads to:
1
12
oi i{2
XcK(X,Y,0)
=
(p+o)2L
-
j
[[x2Y ]
2+ xc)
2 [X +Y2 2]2J
(11.8)
This kernelfunction, however,
has a. singularity
for X=0, =0 i.e. for E=x ,iy.
Therefore we shall consider v(x,y,z) for z0 and approaòh the plane z=0 from the upstream side, where the kernelfunction has no singularities. If sinr> O we must
approach from the side with z < O and conversely. This means that we have Zs > 0.
We only need to evaluate the integral (11.8) over a region G of sufficiently
small diameter containing the point x, ,=y (or X=o,
Y=o).
For the rest of theregion S we
can
simply substitute z0. Itturns out that the limit
Um
fi
K(X,Y,Z)Z-,OG
Zs>0
over a smaU region G containing the point (x,y) depends on the shape of G. In section 12 derivations are given for two cases:
If G is a strip in e-direction with vanishing width, (11.11) yields: 2t
¡sinI
ç(x,y)
If G is a strip in ri-direction with vanishing width, (11.11) yields:
2ltIsin4r1 f(x,y)
Hence we obtain as expressions for v(x,y,O):
where (11.12) is used, and
f(x,y) lsin*I
v(x,y,o) = 2jtV +
where (11.13) is used. K stands for the kernelí'unction:
2 (x-E)cos*
K(x,y,,ri)
1
[{()22())2
{
coz (11.16)It can easily be checked that for
rQ
(11.16) reduces to the kernel of the integralequation in chapter I: (1..2o).
In the special case If* K(x,y,,ri) vanishes and both (11.1ti) and (11.15) reduce to v(x,y,o) = 2i.tV v(x,y,o) = 2.1VIsinjrI + i ¡f K x,y,,r1) f(,ri)ddr (11.11,.)
¡f
K x,y,,ri) )dridZ (11.15)§12 Derivation of the expressions (11.12) and (11.13).
In this section we evaluate the expression (11.11):
hin ¡f
K(X,Y,Z) f(,i)dd11 (12.1)Z-30 G Z s>O
where K(X,Y,Z) is given by (11.9). For the small region G we choose two different shapes succesiv&Ly.
G is a part of a strip in e-direction of width 26, confined by a circle of radius e with e > 6. The point (x,y) is situated in the centre of this small region. Later on we shall take the limit 6
G is a region similar to the one described under A, but the strip is in
ii-direction.
A
2.-x,y)
8 ) Bfig. 12.1. The two shapes chosen for the region G.
We assume a Taylor expansion to be valid for
00
pq
=i
p!q!(
îci)(,)
p=0 q=0 1-dire ction >-direction (12.2)It turns out that only the term with p=0, q=0 i.e. the term f(x,y) gives a finite contribution to (12.1) in the limit 8 -,0. We shall evaluate this contribution; it is not difficult to see that the contributions of the other terms in (12.2) indeed vanish.
Introducing the polar coordinates R,cp by: