You are on page 1of 27

Pro0. Aerospace Sci. Vol. 24, pp. 1-27, 1987 0376-0421/87 $0.00+.

50
Printed in Great Britain. All rights reserved. Copyright C) 1987 Pergamon Journals Ltd.

AERODYNAMIC THEORY FOR MEMBRANES AND SAILS

BARRY G . NEWMAN

Professor of Aerodynamics, McGill University, Montreal, Canada

(Received 13 May 1986)

CONTENTS

NOTATION 1
1. INTRODUCTION 2
2. THE NORMAL BOUNDARY CONDITION FOR A MEMBRANE 3
3. TWO-DIMENSIONAL MEMBRANES PRODUCING LIFT 4
3.1. Some immediate results 5
3.1.1. Two-dimensional membranes with small camber and incidence in inviscid flow 6
3.1.2. Solutions of the impervious equations 7
3.1.3. Extensions of inviscid theory 10
4. LUFFING OF PLANAR SAILS 11
5. DOUBLE MEMBRANE AEROFOILS 14
5.1. Thin inflated lenticular aerofoil at zero incidence 15
5.2. Inflated aerofoil at incidence 18
6. BLUFF MEMBRANES 19
7. THREE-DIMENSIONAL SAILS AND SAILWINGS 21
7.1. Slender parawings and sails of low aspect ratio 21
7.2. Three-dimensional sails of moderate aspect ratio 23
8. CONCLUDING REMARKS 25
REFERENCES 26
APPENDIX 27

NOTATION

a - ( i ) amplitude of an oscillating membrane


(ii) inverse measure of vorticity in oncoming flow, Eq. (39)
(iii) thickness of idealized battened sail
b--thickness of membrane
B--measure of the size of a sail
c--phase velocity, sometimes complex for a travelling wave
cl--aerofoil chord
Ca--drag coefficient D/(qcl )
CI--skin friction coefficient
Ce--internal pressure coefficient (P - p)/q
Cpa--base pressure coefficient (Pa- P~)/q
Cs--Suction force coefficient S/qc~
CT--tension coefficient T/qc~
e---E/4nq
el, ez--amplitudes of the harmonics of a standing wave
E--Young's modulus for a Hookean material
h--half the thickness of an inflated aerofoil
/--moment of the plane area of the cross section of an idealized batten
k--(i) local circulation per unit length in two-dimensional flow
(ii) wave number
kl--porosity constant [Eq. (49)]
K--non-dimensional pressure drop for a porous membrane
~'--length of a membrane
/o--unloaded length of a membrane
L--lift per unit width of membrane
m--source strength per unit x
n--exponent in formula for porosity
N--number of elements in the numerical analysis of a sail
p~pressure
pa--base pressure for a bluff membrane

JPAS 34~ 1-A l


B. G. Newman
P--pressure inside an inflated aerofoil
q--~pU 2, the dynamic pressure of oncoming flow
R--radius of curvature
Re--Reynolds number Uct/v
s--{i} half the semispan at the trailing edge of a low aspect-ratio wing
(ii) arc length
S--leading edge suction force per unit span
t--time
T--tension per unit width of membrane
U--free stream velocity
U~, U2--velocities on either side of an oscillating membrane
v,--velocity normal to a porous membrane
w--velocity on the suction side of a lifting membrane
VL---velocityon the pressure side of a lifting membrane
x--chordwise axis
xL~entre of lift
y--axis perpendicular to the chord in two dimensional flow
(replaced by z in three-dimensional flow, Fig. 19)
or--incidence of a wing
fl--angle between S and the chord
),~amber ratio
A~'difference of'
n--(~-c~)/Cl, excess-length ratio
~ 0 - ( : 0 -- CI )If1
ez--a small number << l
e2,e3 -small displacements
q--(i) shape of an oscillating membrane
(ii) shape of a low aspect-ratio wing sail
0--(i) angle subtended at a centre of curvature
(ii) membrane angle
01--leading edge or luff membrane angle
02--trailing edge or leech membrane angle
2--(Tk)/(pU 2)
22--proportional increase of length due to tension
/t- (i) half the leading edge angle of a conical wing
(ii) (ak)/p
v--kinematic viscosity
p--fluid density
a - ( i ) coefficient of porosity
(ii) mass per unit area of membrane
~b--perturbation velocity potential
to--radian frequency of an oscillating membrane, sometimes complex for a standing wave
92~to/(k U)

1. I N T R O D U C T I O N

M a r c h a j (1964, 1976) has written two b o o k s on the science o f sailing a n d in the first o f
these he r e m a r k e d ' T h e t h e o r y a n d p r a c t i c e o f sailing a r e u n d o u b t e d l y uneasy bedfellows'.
It is h a r d to disagree. T h a t p a r t o f t h e o r y which is the a e r o d y n a m i c s o f sails has in the
p a s t p r o v i d e d general r a t h e r than specific i n f o r m a t i o n for designers. T h e usefulness o f
h i g h - a s p e c t - r a t i o sails, s t r e a m l i n e d masts a n d battens, for example, is c e r t a i n l y better
u n d e r s t o o d in terms o f the science o f a e r o d y n a m i c s but it must be a c k n o w l e d g e d that
engineering d e v e l o p m e n t a n d testing r e m a i n s a d o m i n a n t p a r t o f a n y useful i n n o v a t i o n s .
Sails a r e usually classified as either "square sails' as used on brigs, barques a n d c l i p p e r
ships o r fore-and-aft sails as d e v e l o p e d for N i l e n a g g a r s ( t r a p e z i u m - s h a p e d lug sails),
Chinese j u n k s (the s a m e but fully battened), P o l y n e s i a n o u t r i g g e r s a n d E g y p t i a n G y a s s a s
( t r i a n g u l a r lateen sails). All are very ancient (Phillips-Birt, 1962).*
S q u a r e sails seem s i m p l e enough a n d were d e v e l o p e d as early as 3000 B.C. by the
E g y p t i a n s a n d Phoenicians. T h e y p r o p e l the b o a t p r e d o m i n a n t l y in the w i n d direction,
relying on a d r a g force to d o so. In function they resemble a p a r a c h u t e a n d survive on
m o d e m pleasure sailing craft as spinnakers.* By swinging the yards, square sails can also
sail b r o a d s i d e to the wind, but their a b i l i t y to sail u p w i n d is very limited. A specific

*Sketches of the various sails are shown in the appendix.


Aerodynamictheory for membranes and sails 3

feature of square sails is that the windward concave side of the sail is always the same.
This is in contrast to fore-and-aft sails which generate a significant force at right angles to
the apparent wind, a force which pulls the boat obliquely upwind, alternately using either
side of the sail as they tack. These sails may be thought of as upright wings generating
more 'lift', (cross-wind force) than drag. A modern catamaran with high aspect ratio jib
and main, streamlined rotating mast, and heavily battened main with elliptic planform,
can sail as close to 35 to the wind which is equivalent to less than 20 to the apparent
wind (Howard-Williams, 1976).
The use of sails is not limited to sailboats, landsailers and iceboats. They have also been
applied to hang-wing gliders and ultra-light aircraft (Shaw, 1984). Early European wind
mills also used sails for easy furling, and designs based on loose-footed jib sails are still to
be found in Portugal and Greece. More recently the double sided type of wing sail has
been applied to both horizontal and vertical axis wind turbines (Dekker, 1977; Newman
and Ngabo, 1978).
This review summarizes the aerodynamics of membranes which produce mainly lift at
low incidence and mainly drag as a bluff body at high incidence. This is done in a classical
manner starting with two-dimensional aerofoils and proceeding to the more complicated
three-dimensional sails of both low and high aspect ratio, even though the connection
between the two- and three-dimensional situations is often more tenuous for sails than it
is for rigid wings. The instabilities of sails at low and high incidence are also discussed.

2. T H E N O R M A L B O U N D A R Y C O N D I T I O N F O R A M E M B R A N E

In two-dimensional static conditions the normal force associated with the difference of
pressure on either side of a membrane is balanced by the component of tension T in that
direction (see Fig. 1).
For a small curved arc ds on a membrane of unit width which subtends an angle dO at
the centre of curvature of the membrane, the excess pressure Ap on the concave or inside of
the membrane exerts an outward normal force Ap ds which is balanced by the two tension
components T(dO/2) and (T + dT) (d0/2).
In the limit the change of tension d T brought about by forces parallel to the membrane
is usually unimportant (Newman, 1981)
Ap ds = T d0

0 ds

T T

Fro. 1. Equilibrium of a two-dimensional membrane.


4 B. G. Newman

and thus
T
Ap -- (1)
R
where R is the local radius of curvature and the tension T is a constant.
For a general three-dimensional membrane containing orthogonal curvilinear co-
ordinates s~ and s2 the generalization of the result would seem to be

T~ T2 (2)
Ap: ~-1 + R ~

where RI is the local radius of curvature of the Sl coordinate and T1 is the tension in the 1
direction per unit width with similar definitions applying to suffix 2. In contrast to the
two-dimensional case, the tensions T~ and T2 are not usually constant.
Equation (2) is only strictly correct for one of two conditions. The shearing stresses on
the sides of the rectangular element dslds 2 have in general a component normal to the
membrane if the four comers of the element do not lie in one plane and, in such cases, the
equation is more complicated (Novozhilov, 1959; Flugge, 1973; Irvine, 1981). The two
conditions for which Eq. (2) is correct are:
(i) when the shearing stresses are zero. In other words when the axes sl and s2 are chosen
so that the tensions T1 and T2 are principal. They are then the maximum and minimum
local tensions.
(ii) when the comers of ds~ and d s 2 a r e coplanar. This occurs when the comers, which lie
on a conic section, are such that s~ and s2 are parallel to the major and minor axes of the
conic section, and the radii R~ and R 2 a r e the maximum and minimum values locally.
R~ and R2 are then the principal radii of curvature.
Thus Eq. (2) applies when either the local tensions or the local radii are principal
(Newman, 1981).

3. T W O - D I M E N S I O N A L M E M B R A N E S P R O D U C I N G L I F T

It is assumed that the membrane is an aerofoil supported without interference at x = 0


and x = 1, and has a length o f : (see Fig. 2).
A streaming flow U is inclined at incidence ~ to the x axis which is the aerofoil chord,
and the flow is steady, inviscid and incompressible. The tension T in the membrane is
taken as constant. It changes due to skin friction and is thus constant in inviscid flow. An
estimate of the increase of T from trailing edge to leading edge is 2C:f times the dynamic
pressure q =pU z. Thus the increase of tension coefficient C r = T/qC~ is 2C:(:/cl) and this
is usually small enough to be neglected compared with Cr itself, which is 0(1).

....]Z--- o
LI,p

FIG. 2. Two-dimensional membrane in a flow at incidence


Aerodynamictheory for membranes and sails 5

Yn

...as.~ ~ .... /'kds,

u
FIG. 3. Boundary condition for a two-dimensional membrane.

The membrane is assumed to be:


(a) extensible; ~=~o(1 +21 T) for a Hookean membrane,
(b) slightly porous; the local velocity normal to the membrane vn is related empirically
to the pressure difference (PL - Pu).
At ds, Eq. (1) gives
PL--Pv = T/R(x) (3)
where R(x) is the local radius of curvature of the membrane. By Bernouilli's equation
PL -- PU = p(v~: -- V~) = p(v U+ VL)k (4)
where k = v v - vL is the local circulation per unit length.
At ds the vectorial combination of U, and the sum of the velocities 'induced' by the
vortices kdsl at every part of the membrane must be such that the flow at ds is parallel to
the vector combination of vn and (vv+ vL) there (see Fig. 3). The length of the membrane,

where y(x) is the shape of the membrane.


With the Kutta condition applied at the trailing edge these requirements are sufficient
to determine the six unknowns p~., Pv, vL, vu, y(x) and T. The relevant non-dimensional
determining parameters for the flow are a, eo=(#o-Cl)/c~, 2~qc~*, (pUc~)/g and
parameters defining the porosity of the membrane.
CT = T/(qcl) is the tension coefficient and is usually taken as a dependent parameter
determined by the above parameters. However, it is often more convenient to view CT as a
defining parameter and e = ( # - c~)/cl as dependent on it.

3.1. SOMEIMMEDIATERESULTS

Some results can be obtained solely from the fact that the aerofoil is a membrane.
(Jackson, 1983; Newman and Low, 1984). The forces on the membrane supports are T at
the trailing edge and T plus the leading-edge section S at the leading edge 0 (see Fig. 4).

T T

FIG.4. Resultant forceson a two-dimensionalmembrane.

*Jackson (1985)used the reciprocal[E(membranethickness)]/(qcl) where E is Young's modulus.


6 B.G. Newman
It may be assumed that the suction-force angle f is small like ~, 01 and 0z.
Resolving in the streaming flow direction
S cos(~t - fl) = T[cos(01 - ~t)- cos(02 + 0t)]
C s = (01 +02)[02-01 +2~] Cr+0(04) Cr. (6)

Resolving in the lift direction


L = T[sin (01 - at)+ sin(02 + ~t)] + S sin(or- fl)
CL = (01 + 02)Cr +O(03)Cr.

Since Cs also equals CL sin(0t--fl)= CL(ct--fl)--O(O 3) CL, Eqs (6) and (7) show that

fl= 01 ~2 +0(03). (8)


2
Taking moments about the leading edge to obtain the centre of lift xL

Lx L cos~t= Tc 1 sin02

XL 02
----- I-0(02). (9)
cj 01 +02

Jackson (1983) used Eqs (7) and (9) in conjunction with a cubic shape for the sail to
obtain an approximate solution for the impervious membrane. This solution is
particularly useful when the sail stretches, i.e. 21 is finite. One interesting result is that
when the sail is very stretchy, (2~qcl >> 1) and is not initially cambered, the lift varies as
U 3.

3.1.1. Two-dimensional Membranes with Small Camber and Incidence in Inviscid Flow

Most of the solutions have been obtained using the linearization associated with first-
order thin aerofoil theory for which camber, or its equivalent, eo = ~o - l, and incidence 0t
are small. It turns out that this approximation is sufficient for most practical purposes,
since the neglect of viscous effects, in particular leading edge separation bubbles, is of
overriding importance (Chapleo, 1968; Nielsen, 1963; Newman and Low, 1984).
Assume that the membrane is both impervious and inextensible. It is easily shown that
the flow now depends on one combined parameter ~e -t. Equations (3) and (4) applied at
ds, become
d2yl
PL--Pu = -- T ~ - w = pUkl. (10)
ax~

The boundary condition that the flow is locally parallel to ds becomes


1 ~1 ktdx I dy
0~+-~ J O 21r(Xl __X) -- dx

and thus

1 C-r f o I d2(yl/a)
dx 2 d(y/a)
- 2- 2rt(xi - x ) dxl - dx (11)

Thus
y/u = f{x, Cr).
Aerodynamictheory for membranesand sails 7

Equation (5) becomes

~2f1(~)2
8 = ~'- 1 = ~ - o ox. (12)

Combining with Eq. (10) 8= ~2f(Cr)


Thus
C T =f(~ -*) and y/~ =f(x,~8- ). (13)
Similarly the lift coefficient CL, the membrane slopes 0 t and 02 are such that CU~, xL,
0~/~ and 02/~ are all determined uniquely by either C T or 0~ -.

3.1.2. Solutions of the Impervious Equations

Following the original work of Voelz (1950) Eqs (11) to (13) have been solved many
times. Thwaites (1961) solved an inverted form of Eq. (11) using matrix methods. He was
particularly interested in the eigenvalue solutions which correspond to cases for which the
leading edge has a finite velocity, usually called the optimum or ideal condition. Unlike a
rigid aerofoil, there are many of these for a sail.
Nielsen (1963) used a more conventional approach expanding the slope of the
membrane as a Fourier sine series with a lead term to represent the leading edge
singularity at other than the optimum incidences. The results are conveniently tabulated in
his paper for ~>0.
Greenhalgh et al. (1984) obtained a direct numerical solution by dividing the membrane
into n~ equal linear segments located conceptually on the x axis, concentrating the vortices
at the point and then applying the boundary conditions at the point of each element. In
this way the overall Kutta condition was satisfied for n~ large (James, 1972).
Mateescu and Newman (1985) obtained a simpler solution using singularities in the
complex physical plane to represent directly the complex conjugate velocity there. This
method is convenient when slopes and changes of slope are prescribed as they are in both
membrane and jet-flap problems.
The results for CLe- and CT are plotted as functions of ~t8- in Figs 5 and 6.
For large 0~ -, Thwaites (1961) has shown that
dCL
= 2re (14)

as might be expected, and


~CT
_~---
d(0 ~ = 0.983. (15)

For ~>0, the sail is wholly concave to the wind in the normal manner of sailing. For
~t=0, CT = 1.727 is the first eigenvalue or optimum incidence at which the velocity at the
leading edge or luff is finite. The sail is symmetrical fore and aft and the centre of lift is at
the half chord.
As ~te-* is reduced to negative values, the membrane becomes S shaped and a second
optimum incidence occurs at ~ - = - 0 . 6 with CT = 0.726 and CL--~0+ (see Fig. 7). Other
eigenvalues appear for smaller ~t8- with wavier and increasingly unrealistic shapes.
The predictions have been compared with experiment in Figs 5 and 6. The results are
found to be affected by the presence of separation at the leading edge which at low ~ is
followed by reattachment (Newman and Low 1984). In general the results which are for
Reynolds numbers greater than 105 do not collapse in terms of 0~-* and CL and C r are
8 B.G. Newman

Z(- / i I

i S
%
t ~ ^ 0~OIl
,,~ A = "0.017
' ,'7 "o "
o A 0 0 0

~'(~o04"=2O O o a oe,0.03

i=&~ , e0.057
o o o 00.05

1 I
-i o i z
a/,/'6"
FIG. 5. Lift coefficient as a function of incidence. Theory compared with measurements: Open symbols, Newman
and Low (1984), Re--- 1.2 x 10s; Closed symbols, Greenhalgh et a/(1984), Rex 1.3 106. Numbers give the excess-
length ratio 8.

,.,..) 0.017

..00.03
e=O.OS

-I 0 I 2
o//~"

FIG. 6. T e n s i o n coefficient as a f u n c t i o n o f incidence. T h e o r y c o m p a r e d w i t h m e a s u r e m e n t s o f N e w m a n a n d L o w


(1984), R e = 1.2 x 105.
Aerodynamic theory for membranes and sails

FIG. 7. Sketch of the flow at the second optimum incidence.

significantly less than the theory, particularly for moderate to large e, say > 0.02. The main
effect on lift is probably due to the thick boundary layers near the trailing edge rather than
the presence of leading edge separation bubbles.
The theoretical leading edge section Cs I-Eq. (7)] cannot be realized due to the presence
of leading edge separation bubbles on one or other side of the sail.

From Eqs (6) and (7)


Cs = ~1/02 -a01 )
[ ~ - + 2/~-
CLa~
and is a function of ~-.
The camber ratio ~ when normalized to 7~- is also a unique function of as -. Thus

= f (16)

as shown in Fig. 8.
The drag coefficient must necessarily exceed the values of C s i n Fig. 8 and it is by no
means insignificant. Taking typical values a = 0.2 = 11.5 , ? = 0.10 (e = 0.3)

Cs 0.09.

The magnitude of this typical result for a sail beating into wind is noteworthy for the
designer. Clearly jib sails in particular would be improved by any device which seeks to
recapture Cs.
Figure 5 shows that e or sail camber is a useful control for lift, playing the same
role as the trailing edge flap on a rigid aerofoil. For example in high winds when beating
into wind, relatively flat jib sails are normally used to limit the lift which would otherwise
excessively heel or even turn over the boat.

(J

5 IO 15

FIG. 8. Theoretical ratio of leading-edge suction to lift: dependence on incidence and camber.

JP/di 2 4 : 1 - B
10 B.G. Newman
3.1.3. Extensions of lnviscid Theory
Solutions for an extensible membrane are readily obtained by incorporating an
equation similar to g = f o ( l + 2 1 T ) and by iterating the previous solution (see also
Jackson, 1983). Since at ~=0, Cr = 1.727 and is independent of e, the change of e due to
stretch at this incidence is readily calculated leading directly to the value of CL (Jackson,
1985).
Thwaites (1961) formulated the original problem for a slightly porous membrane
adopting a linear relationship
tr(pL--Pv) = pvnU (17)
where a is the coefficient of porosity which may vary with x.
Barakat (1968) obtained solutions of Thwaites equation by expanding the slope of the
sail as a series of Jacobi polynomials rather than the sine series used by Nielsen. As
expected, lift is reduced by porosity. The value of Cr at ~ = 0 decreases with increasing tr to
about half the impervious value when tr = 0.3. However, practical values of a are typically
an order of magnitude less than this and for them the effect is very small.
Tuck and Haselgrove (1972) have extended the previous solution to cases when the
trailing edge is held by a straight sheet which itself experiences no aerodynamic force.
In unpublished work, Newman (1983) has examined a thin idealized two-dimensional
battened sail which obeys the simple beam equation with a flexural rigidity EI. The beam
shear forces now help to take some of the pressure load and Eq. (10) becomes

d2 ( d2yl "~ T d2yx (18)


Ap = P L - - P u = ~ E1 dxYa] - dx-~l.
The case of constant loading, Ap, is of interest: it was historically the basis for the
NACA 16 series of aerofoils (Stack, 1944)

y = ~-App 2 [ - ( 1 - x ) l n ( 1 - x ) - x l n x ] (19)

is the shape of the camber line and e = 0.


For a batten of rectangular cross section and thickness a, Eqs (18) and (19) can be
readily integrated to give

(ael/3)3 = 12x(1 - x ) + ~ - ((1 -x)ln(1 - x ) +xlnx) (20)

where e= E/(2npU 2) an elasticity number for bending. Shapes for three values of Cr are
shown in Fig. 9. For very small distances from the leading and trailing edges, a becomes
negative. For example at 0.001Cl from these edges, Cr must be less than 0.794 for a to
remain positive.
Presumably a gradation of batten thickness according to Eq. (20) might improve the
performance of a battened sail for small values of e.
Non-linear solutions for unstiffened membranes have been obtained by Vanden-Broeck
and Keller (1981) for large Cr, and more generally by Vanden-Broeck (1982). In this case
the results do not collapse in terms of the combined parameter ee -, although a
corresponding non-linear form (sin2~t/2)e-~r is used in the numerical analysis. The results
now depend separately on e and e, but unfortunately do not show any better agreement
with experiment (Newman and Low, 1984).
Jackson (1984) has used the surface vortex method iteratively and applied it to both
single and twin sails. The latter type of problem, which represents an idealised jib and
main, is similar to the problem of multi-element rigid aerofoils, such as aerofoils with
slotted flaps.
Dugan (1970) has used hodograph methods to solve for separated flow about a two-
dimensional membrane when the line joining the supports is oblique to the flow. This is a
Helmholtz solution with ambient pressure in the wake.
Aerodynamictheory for membranes and sails 11
0.7

/// c
'1 /// \\\

I I I I
0 0.2 0.4 0.6 0.8
ItlC t

FIG. 9. Thickness distribution for a uniformity-loaded battened membrane.

4. L U F F I N G O F P L A N A R SAILS

When a sail boat is close-hauled, the sails are required to generate high 'lift' with
minimum drag. They therefore operate, as do those on hang-wing gliders, at incidences
below the stall. However, unlike solid wings they must also avoid negative incidences for
then a sail begins to lose its concave form to become S shaped and finally, as the incidence
is further reduced, to oscillate. This behaviour is known as luffing.
The classical static theory described in Section 2 showed that at ~t=0 the sail was
symmetrical fore and aft and hence was at its 'optimum incidence' with the Kutta
condition applying at the leading edge as well as the trailing edge. The sail is therefore
bistable being just as likely to snap through and set itself with opposite camber. The
tension coefficient CT corresponding to ~e-=0 is 1.73 and this condition was proposed
by Thwaites (1961), Nielsen (1963) and Irvine (1979) as a criterion for luffing.
However laboratory experiments, and even observations on full scale boats, indicate
that sails may be stable at slightly negative incidence. Both theory and experiment show
that the sail is slightly S shaped with a point of inflection (Fig. 7) near the leading edge.
Greenhalgh et al. (1984) suggest that the criterion for luffing is when
~Cr
------~'00.
~(~-)
Theoretically this occurs when ~te-=-0.93 and C~=0.94 (Fig. 6). The idea is as
follows. If for a given membrane (i.e. e) ~t is reduced to negative values until it reaches
-0.93e , the solution becomes impossible and a snap through to the opposite camber is
inevitable in order to accommodate more negative values (which have now become
theoretically more positive values) of ct.
Both the above criteria for luffing, ~t= 0 and ~t= -0.93e , are based on considerations of
static stability and would only apply to the extent that the static theory itself is correct.
The dynamic stability of a planar membrane has been analysed by Newman (1982) for a
travelling wave and is an extension of the linearized theory of Rayleigh (1929) and Binnie
(1970).
The membrane is identified by
= a exp [ik(x -ct)] (21)
which is a travelling wave of wave length 2~/k and wave speed equal to the real part of c.
The parameter c is the only complex parameter and it is understood that the real part of
the right hand side of Eq. (21) is taken.
12 B.G. Newman
y,r/

uz + V~z
P2
r _- ,~ (xot)

0 Ci X
131
U,+V~,
FIG. 10. Oscillation of a two-dimensional membrane.

The velocities on either side of the membrane are U2 for y---~ + oo and U~ for y---* - ~ .
The flow is irrotational and represented anywhere by U+V~b with suffix 1 below the
membrane and suffix 2 above. The kinematic boundary condition at the membrane
expresses the fact that the velocity perpendicular to the membrane equals the velocity of
the membrane itself in that direction. For small slope OqlOx and small IV4,1/U this
condition becomes
04~ _ oq oq
c3y dt ~- u Oxx at y = o. (22)

Equilibrium of the element ds = dx of membrane requires that, with y = 0


t32r/
dx -(p~ - p2) d x - T ~x2 dX = 0 (23)

where tr is the mass per unit area of membrane and T is the tension per unit width of
membrane and is constant.
The pressure is obtained from the unsteady form of Bernoulli's equation

+ U~x+ ~- = 0 (24)
P
and may be inserted with suitable suffixes in Eq. (23).
Appropriate forms for q~ come from a complex potential (Milne Thomson, 1965)
~bl = C1 exp(ky) expik(x -ct) which correctly----~0 as y---. - ~ (25)
tP2 = C2 e x p ( - k y ) expik(x-ct) which correctly---.0 as y---~ + ~ . (26)

Substituting (25) and (26) in (22)


Ci = ai (U1 -c) (27)
C2 = - ai (U2 - c) (28)
and Eq. (23) becomes

(~+ 2 ) c 2 _ 2 ( U 1 + U2)c + U2 + U2 ___Tkp= 0. (29)

Instability occurs if in this equation there is a negative imaginary part in the solution for
c. That is if

- ( U , - U2)2-ak (u~ + u ~ ) + ~ (~kp + (30)

Since the original static theory for the membrane is linear we may write as an
approximation U1 = U(1 - e l ) , U2 = U(1 +el) where el is ~t: 1
Aerodynamic theory for membranes and sails 13

Thus

Cr < 4[a/(PCl)] (31)


[(~k)/p] + 2
for instability. For a travelling wave the wave length would have to be less than about one
quarter of the chord ct, that is k > (8n)/cl.
For full scale sails tr/(pca) is usually <0.05. Thus Cr <0.06. for (32) instability. Even for
large values of ~/(pc~) which are typical of model sails, instability occurs when CT<
1/(2r0= 0.16.
These predictions are much smaller than observed values which range from 0.5 to 1.
Thus a travelling wave does not appear to be the appropriate solution for this particular
problem. A standing wave of larger wave length seems to be more plausible and Newman
(1985) has developed a similar, although approximate, analysis for such a wave.
A single harmonic for r/ does not satisfy Eq. (23) and hence two terms were used to
obtain an approximate solution using the Galerkin method. (21) is replaced by
q = (el sin kx + e2 sin 2kx) exp(itot) (33)
where to is the frequency which may be complex.
The corresponding expressions for 4) are
q~l = [(al dkx + ba e-ikx)ekY + (1 e2ikx +dl e- 2ikx)e2kjr]~tat (34)
tip2 = [(a2~kx + b2 e- ikx)e- ky + (c2e2ikx + d2e- 2ikx)e- 2ky]~ t . (35)

(34) and (35) combine with (22) to give values of a to d in terms of el and e2.
Since Eq. (23) cannot be satisfied exactly at all x the left-hand side of the equation is
written as E and the weighted averages

fC~ E sinkx dx and fC2 E sin2kx dx


o o
are equated to zero to give the values of el and e 2 and an equation corresponding to (29).
Choosing U1 = U(1 - e l ) , U 2 = U(1 +el) as before
[512 ]
(# + 1) (/* + 2)f~4- [~n2 + 4 ( 2 - 1 X / , + 2) + ( 2 - 2X# + 1) f~2 + 4 ( 2 - 1 X 2 - 2 ) = 0 (36)

where/~ = (ak)/p, 2 = (Tk)/(p U 2) and f~ = to/(k U).


The most unstable solution is one of pure divergence when
(2-1X2-2)<0. (37)
The smallest value of k is n/cl
2 = *t/2Cr
and a divergent instability occurs when

C r < 4 = 1.27, (38)


n

a criterion which naturally does not involve the inertia of the membrane.
A comparison with some measured values of luffing incidence are shown in Table 1.
It is clear that luffing does not occur at a unique value of the linearized parameter ~-.
Dimensional analysis would show that ct for luffing depends, in general, on e, cr/(pc) and
(Uc)/v. The results are plotted in Fig. 11 and confirm that at the start of luffing at is indeed
mainly a function of e. Curves for linearized theory with Cr = 1.27 or C r = 0.93 are shown
and clearly give completely the wrong trend. As with the errors in Cr and CL in the static
analysis, the discrepancy is attributed to flow separation as well as thick boundary layers.
Flow visualization in a water tunnel (Fig. 12) shows how extreme this can be on the
underside of the membrane. Indeed, it appears that the flow there might be treated as
14 B. G. Newman

TABLE 1. LUFFING ANGLES

Luffing Measured ,r
Source C~ pc,

Newman (1984) 0.03 -2.5 0.77


air 0.05 - 1.Y' ~0.5 0.41
Re 1.2 x 10 5 0.10 - 1o ~0.4

Greenhalgh (1983) 0.0114 -3.3 1.25


0.0077 -3.5': ~ 1.15
air 0.0569 - 1.7" 0.23
Re 1.3 x 10 ~ 0.0422 -2.0

Newman (1983) 0.025 - 2.5 ~'


water 0.001
Re 105 0.080 - 1.5

Nielsen (1963j 0.026 - 2.0


air
Re 8 105

I
NeL --010

.f
?
w
G~
/ NoL 005
G=C

CT ~',,b
1.27 /
^ ^-.G~C
.

-4 -3 -Z -I
a (d~l

FIG. 11. Angle of luffing as function of excess-length ratio. Dotted lines are theory for two values of Cr. Solid line
passes through experimental points, N.L. - N e w m a n and Low (1984), N . - - N e w m a n (1983), G . C . - 4 9 r e e n h a l g h
and Curtiss [1983), as described in Table 1.

approximately stagnant so that U2 = U and U1 = 0 in the equation corresponding to (29)


and the first divergent solution would then occur when 2 < 1 or Cr <0.64. This prediction
is seen to be consistent with the measurements for moderate to large values of e in Table 1.
It is concluded that luffing occurs at a slightly negative incidence due to divergence of a
standing wave for C,r < 1.27 if there is negligible separation at small e or for C-r <0.64 if
there is significant separation on the 'underside' of the membrane for e>0.025 say. There
is no dependence on tr/(pc), and scale effects are probably small for Reynolds numbers
> l 0 s.

5. D O U B L E M E M B R A N E A E R O F O I L S

In order to improve the aerodynamic characteristics of single membrane aerofoils, and


in particular to avoid leading-edge separation, double sided membranes wrapped over a
rigid cylindrical leading edge have been used (Sweeney, 1961; Fink, 1967; Robert and
Aerodynamic theory for membranes and sails 15

FIG. 12. Water flow past a two-dimensional membrane just before luffing. ~=0.08, ct= - 1 , Re= 105, flow from
left to right.

Newman, 1979). A thin lenticular aerofoil with a sharp leading edge at zero incidence has
also been investigated since it represents a reflection-plane model of a shallow inflated
building (Newman and Tse, 1980). This is perhaps the easiest case to analyse and the
theory closely resembles that for single membrane aerofoils at incidence. As will be seen
the internal pressure coefficient Ce plays the same role as the angle of incidence ~ in the
previous analysis.

5.1. THIN INFLATED LENTICULAR AEROFOIL AT ZERO INCIDENCE

The aerofoil has unit chord and an internal pressure P (see Fig. 13). The analysis is not
much more complicated if the upstream flow is assumed to have constant vorticity

u=U(l+ I~_]l) (39,

and it then more realistically represents an Earth boundary-layer flow over an inflated
building. Since the vorticity is step uniform and the flow is symmetrical about the x axis,

_/
U x x~ I

FIG. 13. Inflated lenticular aerofoil at zero incidence.


16 B.G. Newman

the aerofoil shape may be generated by distributing sources of strength m per unit length
along the x axis (Milne Thomson, 1968).
The source at (xl,yl) on the aerofoil is

m d x l = 2U(1 + [ ~ ) d y l (40)

since the aerofoil is thin and the flow 'within' the aerofoil is to first order the same as that
upstream, Eq. (39).
The incremental velocity at x due to this source is
-/,ndx 1
du - - - (41)
2~(Xl - x )
and the total velocity at (x, y) for positive y is

u = U 1+
( ~) ff2U(l+[y,/a])
- 2n(x l - x ) dxl" (42)
o
The pressure p at (x, y) on the aerofoil arises from the flow along the central streamline
p~o- p = pU(u- U). (43)

For equilibrium of a membrane of small slope and y positive


d2y
-T~ = P-p (44)

where P is the pressure within the aerofoil.


Equations (42) to (44) combine to give
CT d2(y/Ce) 1 (y/Ce) t " [1 +(y,/a) d(y,/Ce)]
2 dx 2 = 2 q a J o ~(x, - x ) (45)

where
T P-Poo
Cr=pU2~1- and Cp= l p U 2 .
If a is 0(1) or more, lyll/a << 1. Thus
y/Ce is a function of x, Cr and a. (46)
The length of each half of the membrane g is

-- (iX.
c~ 2 o
Thus
Cr is a function of Cee - and a/q (47)
where cl is the aerofoil chord, previously taken as unity and
l-c~
e=E-1 -
Cl

The result (47) is seen to be similar to that obtained for single membranes but with
replaced by Cp.
A solution for a ~ m has been obtained by Newman and Tse (1980) by expanding the
shape as a Fourier sine series. The results are compared with experiment in Fig. 14 for
moderately high Reynolds numbers and thickness to chord ratios (2h)/q, less than 0.28.
For these conditions there was no observable separation of the flow ahead of the trailing
edge.
Aerodynamic theory for membranes and sails 17

//
0 O.S I 15 2 2.5 3 3.5 4
v

FIG. 14. Relation between internal pressure and tension. Solid line is theory.
A, 0.0056 to 0.0082; V, 0.0078 to 0.0122; 17 0.0140 to 0.0171; O, 0.0228 to 0.0256; +, 0.0267 to 0.0312; x, 0.0480
to 0.0518. Experiment Newman and Tse (1980) Re=3 x 105 to l0 s.

Solutions for a/c1> 0.5 were obtained by Newman and Goland (1981) using the method
of Hess and Smith (1967). An empirical formula representing these results for positive Ce
is the straight line

Cr = 0.5 +0.25 C P ~ -1 a
C1+(--~
The (24) factor comes directly from a simple analysis for U = 0 (Newman and Tse,
1980). For values of Cee-< -0.1 (i.e. Cr<0.48) there are multiple solutions and they
correspond to membranes with wavy shapes. For Cr less than about 0.3, it was found
experimentally that the membranes became unstable.
The stability analysis of Section 4 may be applied to this case also by putting U2 = U
and U~ =0.
It predicts instability for a travelling wave when
2 l+(trk/p)
Cr< kcl 2 +(ak/p)"
With a wave length
2rt
-y =
1

and
o"
-- = 0.05
pcl
for Newman and Tse's measurements
CT < 0.06.
JPU 24 : ].-C
18 B.G. Newman
This value is less than observed values which indicated some instability for Cr<0.3
approximately. It is also much less than a static instability criterion of the
Greenhalgh-Curtiss kind which, from Fig. 14, occurs at Cee-= -0.95 and Cr =0.28 and
agrees well with the observed collapse.
With the velocity on one side of the membrane zero, a standing wave is unstable when

Tk
2 = pu2<l.

In this case the longest wave length


2~
---~- c1
k
Thus
1
Cr < - = 0.32

which is close to the observed values.

5.2. INFLATEDAEROFOILAT INCIDENCE

Vanden-Broeck (1982a) has generalized the solution for uniform upstream flow by
considering an inflated membrane again with line supports at leading and trailing edges.
The solution was obtained in terms of Ce and CT by expanding the velocity potential
function in powers of C~ 1: it is limited to large values of Cr, of 0(10) or more. For values
of Cp<0 the top and bottom membranes touch and become a single membrane with
modified boundary conditions. A particular solution is sketched in Fig. 15. The conditions
are for ~=45 , C p = - 2 / 3 , Cr=2. Such cases have not yet been investigated
experimentally but their shapes seem to be very attractive and might avoid leading edge
separation at sufficiently high Reynolds numbers.
Inflatable aerofoils with solid cylindrical leading edges have been investigated
experimentally by Sweeney (1961), Fink (1967) and Robert and Newman (1979).
Theoretical work has been done by Murai and Maruyama (1980) for impermeable
membranes up to and including the conditions for which the top and bottom membranes
begin to touch near the trailing edge. The boundary condition for pressure is linearized as
in Eq. (44) although the flow is calculated iteratively using a numerical method based on
sources distributed over the membrane surfaces themselves. The pressure distributions for
circular, elliptic and D shaped leading edges were obtained and confirmed that the latter
avoided sharp suction peaks which accounted for the improved characteristics measured
by Sweeney and by Fink.

/
FIG. 15. Inflatedaerofoilat incidence,Vanden-Broeck(1982).
Aerodynamic theory for membranes and sails 19

6. B L U F F MEMBRANES

When ~t= 90 for a single-membrane aerofoil, the situation becomes an idealization of


the ftow past square sails, spinnaker sails and parachutes. Sails are preferably impervious
whereas parachutes are usually slight porous to improve stability and to reduce squidding.
Both impervious and slightly porous two dimensional bluff membranes have been studied
by Newman and Low (1981) and Low and Newman (1986).
The flow is sketched in Fig. 16. The supports are small compared with Cl the distance
between them. The length of the membrane is . For Reynolds numbers (Ucl)/v greater
than about 10+ and with fixed separation points at the supports, the drag coefficient Co,
tension coefficient Cr and the base pressure coefficient CuB (which is constant over the rear
of the membranes to within 0.1) are uniquely determined by the geometry /cl and, if the
membrane oscillates, possibly by the non-dimensional membrane density a/pc~. Although
it is not possible to predict these particular relationships, the dependent variables Co and
Cr can be related approximately to C ~ and the geometry.
Following Taylor (1919) it is assumed that the pressure on the upstream concave side of
the membrane is constant. For an impervious membrane it is the stagnation pressure, for a
porous membrane slightly less than this.
If the upstream pressure at 0 is P0 where the velocity is U 0

P- o-- P B _ K
1 2
where Po +~PUo
1 2 = poo +pU 2. (48)
~pUo
Equation (48) is the conventional way of expressing the pressure drop through a screen or
porous cloth. K is not constant for a cloth of low porosity and is usually expressed as

k,(~) n (49)

where b is the thickness of the cloth and V~ is the velocity perpendicular to it which is
equal to Uo in this case.
Measured values of n (Low, 1982) are very nearly - so that the pressure drop is
proportional to F~/z implying a flow within the cloth which is transitional between
laminar and turbulent.
From Fig. 16
D = 2T sin0 (50)
and also
D = (p0-~)c, (51)

U
P
cll %] Po

FIG. 16. Flow past a two-dimensionalbluffmembrane.


20 B. G. Newman

From (48), (50) and (51)

Co = K(I - C~) (52)


K+I
CD
C r - 2 sin0 (53)

and

- (54)
cl sin 0
if the shape of the membrane is assumed to be a circular arc.
The values of K for parachute materials under normal operating conditions range from
just under 104 for personnel chutes to 5 x 102 for cargo and flare chutes. Thus Eq. (52)
approximates to
Co
- - = 1. (55)
1 - Cps
This prediction is compared with measurements in Fig. 17 and is seen to be fairly
accurate for a wide range of//c~. Average measured values of 0 are also in good agreement
with Eq. (54) for kl > 1000 (Low and Newman, 1986).
Newman and Low (1981, 1986) obtained slightly more accurate predictions for the
pressure on the upstream face by replacing the membrane with a sheet of vortices in the
manner of Parkinson and Jandali (1970) and Bearman and Fackrell (1974). The membrane
shape was iterated from a circle until the membrane boundary condition was satisfied
everywhere. For a porous membrane this boundary condition remains P - P B = T/R since
the 'refraction' of the flow is usually sufficiently small.
The use of Eq. (53) and (54) for medium-to-large values of #/c I is complicated by the fact
that the membrane oscillates significantly although the oscillation may be reduced by
increasing the porosity (Low and Newman, 1986). The oscillation is in harmony with
regular vortex shedding and the Strouhal number (nc~)/U is a function of d/c~, a/(pcl) and
porosity as represented mainly by kt.
Average results for g/cj = 1.6 and 2.6 are shown in Fig. 18 where the Strouhal number (St)
is plotted against k? for convenience. It is seen that St is insensitive to a/pc~ for the range
of values which were tested, and is also independent of f/ci for kl < 1000. This may be

[ o

?
: Q5
o

i I I
1.5 2 2.5

FIG. 17. Co~(1-Cp a) as a f u n c t i o n m e m b r a n e length.


Measurements for ( Ucl )/v = 4 x 10" kt a / ( p c ~)
0 Impervious 0.41
A Personnel-parachutematerial 5800 0.55
[] Flare-parachutematerial 960 0.52
Aerodynamic theory for membranes and sails 21

O.IS

0.1(

0.0~

I I
0 0.02 004 0,6
k;,~t

FIG. 18. Strouhal number as a function of porosity and membrane length (Uct)/v = 2 x 105 e/(pc,) between 0.41
and 2.59.

explained in the following way. Membranes which are nearly impervious oscillate but on
the average are circular in shape even when d/c1 exceeds n/2. St therefore varies with g/Cl.
More porous membranes on the other hand trail downstream with parallel sides
culminating in a hemicylindrical end so that the wake and the Strouhal number tend to
become independent of d/cl. The drag and base pressure coefficients also show the same
tendency.
Some specific conclusions on two-dimensional bluff membranes are of practical interest.
For impervious membranes the drag is a maximum when e ---*0 i.e. d/cl = 1 and the drag
coefficient is 1.9 being close to that of a flat plate. The tension is then very high. The
tension coefficient is a minimum however when d/ct is only 1.4 (Newman and Low, 1981)
and the drag coefficient Co is then 2.4 representing an actual drag which is 10 % less than
that of the same membrane if it were held flat. This suggests that the loss of driving force
for a spinnaker of light material which is designed and deployed so that the tension is
minimal is not so great that the use of heavier and therefore stronger materials would
normally be justified. The drag coefficient Co and the base pressure coefficient Cpn are
however very sensitive to even small amounts of porosity. For example empirical
expressions for the drag coefficient for kt greater than about 300 and Re = 2 x 105 are

C o = 2 . 4 5 - 2 1 k i - f o r - - = 1.6
Cl

and
d
Co = 2 . 0 - 1 5 ki- f o r - - = 2.6
Cl

7. T H R E E - D I M E N S I O N A L SAILS AND S A I L W l N G S

7.1. SLENDER PARAWINGS AND SAILS OF L o w ASPECT RATIO


LOW aspect ratio sails were introduced by Rogallo (1954) for gliding parachutes or
parawings and they were used on early hang-wing gliders. Nielsen et al. (1963) analysed a
22 B.G. Newman

1
z

a~ Y
\

T~

-s x

FIG. 19. Single-lobed conical parawing.

single-lobed conical parawing of such low aspect ratio that the flow might be assumed to
be two-dimensional in planes perpendicular to the oncoming flow, the usual assumption of
slender-body theory. The leading edges are rigid (see Fig. 19).
The tensions in the spanwise y and chordwise x directions are Tr and T~. If/~ and ct are
small these tensions are very nearly principal so that Eq. (2) applies. For a conical surface
the changes of Tx are small and mainly due to skin friction. Being zero at the trailing edge
T~ is effectively zero everywhere. Ty is also constant. It is replaced by T and directly related
to the local pressure difference Ap across the membrane by Eq. (1). Thus for small camber

- T ~ = Ap (56)

where r/= rl(x,y ) is the surface of the wing.


The flow is two-dimensional in the cross planes perpendicular to 0x.
By expressing the perturbation velocity due to the wing in terms of a potential function
satisfying the boundary conditions at the wing surface and Eq. (56), it is easily shown
(Newman, 1981) that the collapsed parameter, corresponding to ~te- in planar flow, is the
'reduced incidence'

tan p \ ~ /
Nielsen et al. (1963) obtained the shape of the membrane (~/tan I~)/(sot) as a function ofy/s
and the reduced incidence. Cr/(tan#) 2, CL/(ottanp) and the trailing vortex drag expressed as
Coj~2tanla) were all determined as unique functions of the reduced incidence. C-r is
defined as T/[p U2(2s)].
As with the two-dimensional problem there are eigen-value solutions corresponding to
'ideal' incidences with vanishing leading-edge suction. The first of these is for a concave
shape and corresponds to Cr/(tan2p)= 0.42 and
ct / ' f - 2 s ' ~ -
tan/~ \ 2s ,] = 1.84.

Below this incidence the shape contains two or more inflection points and is presumably
very close to an unstable condition (Section 4).
Nielsen et al. (1964) have extended the analysis to double-lobed parawings with rigid
keel and have expressed the results in terms of a similar reduced incidence. When
compared with experiment (Burnell and Nielsen, 1965), the theory adequately predicts the
membrane shapes and the angle of zero lift but the lift over the linear range is significantly
Aerodynamictheory for membranesand sails 23
lower than the predicted value particularly for aspect ratios >2, and trailing edge
oscillation is also encountered at low ~. A summary review of the work on parawings
which includes further references to the work of Nielsen and his group has been given by
Mendenhall et al. (1968). Parawings with both rigid and fully flexible leading edges are
considered in one-lobe and two-lobe, conical and cylindrical configurations.

7.2. THREE-DIMENSIONALSAILSOF MODERATEASPECT RATIO

The combined aerodynamic and structural analysis of an elastic three-dimensional sail


of moderate aspect ratio is complicated by a number of factors (Warner and Shatswell,
1925; Marchaj, 1976; Jackson, 1985):
(i) The aerodynamic pressure on the sail produces an elastic distortion of the sail which is
non-linear. This can be seen from the two-dimensional equations. If the unstretched
membrane has a length
d = do(1 +21 T)
~=e0 + doq'~l Cr.
But C r = 1.727 +0.983 ~e-t theoretically for large Cr. Therefore
= eo + d0q~.l(1.727 + 0.983 ~e-) (57)
so that the excess-length ratio e is a non-linear function of the dynamic pressure q
(except when ~ = 0).
(ii) A woven sail itself stretches in a non-isotropic and possibly a non-linear manner
(Milgram, 1971; Satchwell, 1982). It is often locally reinforced with patches and rope. It
is also provided with stiffening battens let into narrow chordwise pockets at various
heights.
(iii) The boundary conditions defining the edges of the sail are forestays, masts, booms and
ropes with respect to which the sail may or may not be allowed to slide. Furthermore
these bounding members are themselves not fully rigid. Indeed mast bending is often
exploited as a means of tensioning the leech of a sail (e.g. Finn and Laser I).
(iv) The change in apparent wind due to the combination of the wind boundary layer and
the boat velocity increases the angle of attack with height.
(v) Finally, the regions of boundary-layer separation and reattachment which were found
in the two-dimensional studies to cause significant differences between experiment and
inviscid theory, are also a complicating factor in the general three-dimensional case.
Nevertheless combined structural and aerodynamic solutions have been obtained for a
number of idealized situations. In all cases complications (iv) and (v) have been neglected.
The developments are summarized in Table 2.
The early work before 1980 made assumptions which enabled lifting line theory to be
combined with the two-dimensional theory of Section 3. However, sail stretch is usually
an essential feature of three-dimensional sails and then the tension in the spanwise or
nearly vertical direction becomes important (Eq. 2). Moreover the sail twists due to the
elasticity of the leech.
The most sophisticated theory to date is probably that developed by Jackson (1985) (see
Fig. 20). The sail plan is divided into small plane triangular dements in preparation for a
finite element method of structural analysis. The same elements are used in the
determination of the aerodynamic pressures using inviscid theory. The sides of each
element then represent a lattice of vortices. The vortex strength of each side is constant
and is eventually determined by satisfying the boundary condition at the centroid of each
element. The wind flow and the effect of all the vortices are combined to give zero normal
velocity at every such point. The Kutta condition is satisfied at the trailing edge by
shedding trailing vortices conventionally in the streaming flow direction. Thus separation
is assumed to occur only at the trailing edge. The direct effect of skin friction on structural
24 B. G . N e w m a n

TABLE 2. THEORIES FOR A THREE-DIMENSIONALSAIL

Sail

Aerodynamic Structural
Author Geometry Elastic properties Boundaries analysis analysis

Nielsen (1963) ~ rectangular inelastic Curved inelastic Chordwise C~ = No lateral


cables at luff constant. 2D tension in
and leech theory in strips the sail

Sneyd et al. (1971) triangular inelastic rigid luff and Weissinger No lateral
inelastic cable lifting line: tension in
along leech C~. constant the sail.
Out-of-plane
movement in
the leech

Ormiston (1971 ) ~ rectangular isotropic rigid luff and Lifting line and Large pretensions
and Hookean elastic cable 2D theory with and thus small
with zero along leech Cr constant displacements.
Poisson's ratio rigid foot Lateral tension
and tip in the sail

Murai et al. (1982) rectangular: same same Lifting line or No lateral


double-sided Hess and Smith tension in
distributed the sail
source method

Kroo (1983) general general non- flexible luff Weissinger or Virtual work
isotropic spar: Vortex Lattice at nodes
with flexible battens based on planar
buckling geometry

Holla et aL (1984) rectangular isotropic and fixed spars at potential flow Uniform biaxial
Hookean with all edges theory tension
zero Poisson's
ratio

loading is shown to be unimportant. The water surface is represented by a simple


reflection of the vortex array so that any flow round the foot of the sail is neglected.
Once the elemental vortex strengths have been determined the velocities parallel to the
elements on the leeward and windward surfaces may also be computed and hence the local
pressure differences across each triangular element.
Under these loads the sail distorts. Each triangular dement is assumed to have a
uniform strain and to remain plane under the loads applied at each corner or node. As
usual a constitutive law (using perhaps an isotropic Youngs modulus and Poisson's ratio)

t
t
t
t

Luff tL~ch

g '
I
I
I

FI6. 20. J a c k s o n ' s d i s c r e t i z a t i o n o f a n elastic sail.


Aerodynamictheoryfor membranesand sails 25

relates the stress and strain in each element. However there is one important difference. If
the minimum principal stress becomes negative in the calculation, so that the sail would
buckle and form a wrinkle, this stress is formally put equal to zero and the calculation
repeated in terms of uniaxial stress in these regions.
The strains in each element are related geometrically to the displacement of each node.
As usual the method of virtual work is used to relate force and displacement at each node
to the strain energy stored in each element. Thus there are three groups of structural
equations which relate the unknown nodal forces, nodal displacement, element stresses
and element strains. The final set of equations comes from equating the nodal forces to the
aerodynamic loads.
The iterative numerical solution of these equations is not easy and convergence is not
always assured. However, Jackson (1985) points out that the change of aerodynamic load
is a maximum for out-of-plane displacements 52 and of the order of q52g2 where g is the
typical size of an element. The change in an elastic nodal force on the other hand is
greatest for in-plane displacements 53 and is proportional to Esatg where t is the thickness
of the sail. If B is the size of the sail and there are N elements, gocB/~/N. The ratio of
elastic to aerodynamic forces is thus proportional to (Et/qB)x/N. Et/qB is usually very
large for Mylar sails and is even of the order of 100 for Dacron sails. Thus the change of
aerodynamic load with nodal displacement does not require the recalculation of
aerodynamic influence coefficients for every iteration. This simplification helps to speed
up convergence of the solution. Kroo (1983) adopted a similar strategy in developing his
method for heavily battened sailwings on hang gliders.
Jackson (1985) applied his method to flat triangular sails of reflected aspect ratio 8 with
fixed or sliding luff, free leech and with the foot free to slide. He used 36 triangular
elements. The initially fiat sail stretched under wind load and became cambered. Due to
the unrestrained leech it also twisted which reduced the local angle of attack particularly
halfway up the sail. This effect outweighed the increase of lift due to camber until the leech
was so curved at higher incidences that it became effectively stiffer. The lift then increased
more rapidly with incidence as the camber effect took over. A similar behaviour was found
by Ormiston (1971) with his model.
Jackson searched for other generalizations in his computed results. He found that
boundary conditions, in particular whether the luff is free to slide, were of overriding
importance in determining the stress trajectories or isostatics. Also much of the sail is
wrinkled and the wrinkles usually fan out from the clew at the bottom of the luff. As he
pointed out, since the divergence of the stress is zero, the major stress is inversely
proportional to the distance between the wrinkles. Thus the tension is often a maximum
near the clew where it is customarily reinforced.
Further details of this combined structural and aerodynamic analysis will be found in a
recent unpublished paper by Jackson and Christie (1986).

8. CONCLUDING REMARKS

Much of the classical work on flexible membranes has been two-dimensional. This is
because such problems are more tractable than general three-dimensional cases and also
because the approach follows the traditional development of aerodynamics from aerofoils
to wings. However the boundary condition for a three-dimensional membrane [Eq. (2)]
clearly shows that the shape of the sail is not solely determined by the chordwise tension.
Indeed the spanwise tension of a sail may only be neglected for certain very specific
situations (Table 1). Nevertheless the study of two-dimensional membranes may be useful
in increasing our understanding of the luffing of foresails, in determining the effects of
excess length ratio and porosity on bluff sails such as spinnakers, in alerting us to the
importance of leading edge separation bubbles and separation ahead of the trailing edge
and finally, in suggesting improvements for the chordwise stiffening of sails by the use of
tapered battens.
JP&S 24: I - D
26 B.G. Newman

T h e recent a d v a n c e s o f K r o o (1983) a n d J a c k s o n (1985) in t h e n u m e r i c a l a n a l y s i s o f


e x t e n s i b l e t h r e e - d i m e n s i o n a l sails h a v i n g b o t h l o c a l stiffness a n d v a r i o u s b o u n d a r y
c o n d i t i o n s , m a k e it v e r y d e s i r a b l e to c o m p a r e t h e s e p r e d i c t i o n s w i t h e x i s t i n g e x p e r i m e n t s
such as t h o s e o f M i l g r a m (1972) a n d G r e e n h a l g h a n d C u r t i s s (1984), a n d also w i t h n e w
experiments.
Sail design is c u r r e n t l y b e i n g i m p r o v e d by m a k i n g t h e u n s t r a i n e d s h a p e a n d t h e
o r i e n t a t i o n o f the w a r p a n d w o o f o f t h e f a b r i c m o r e intricate. I n c r e a s i n g use is b e i n g m a d e
o f l a m i n a t e d m a t e r i a l s w i t h g r a d e d stiffness p a r t i c u l a r l y in b e n d i n g . M o d i f i c a t i o n s to t h e
l e a d i n g e d g e o f f o r e s a i l s are b e i n g m a d e to r e d u c e s e p a r a t i o n a n d in this c o n n e c t i o n
inflated l e a d i n g edges m a y be useful.
It is w o r t h r e i t e r a t i n g t h a t this t e c h n o l o g y is n o t o n l y a p p l i c a b l e to s a i l b o a t s a n d
s i m i l a r s a i l i n g c r a f t but also, for e x a m p l e , to the lifting surfaces o f u l t r a - l i g h t a i r c r a f t a n d
h a n g gliders, a n d to the b l a d e s o f l o w - p o w e r e d w i n d turbines.

REFERENCES

BARAKAT,R. (1968) Incompressible flow around porous two-dimensional sails and wings, J. Math. Phyg 47,
327-349.
B ~ , P. W. and FACKRELL,J. E. (1974) Calculation of two-dimensional and axisymmetric bluff body
potential flow. I.C. Aero. Rept. 74-07.
Bn~E, A. M. (1970) Air-generated waves on a moving medium, J. Mech. En 0. Sci. 12, 230-231.
Bt~a~.L, J. A. and NmLS~, J. A. (1965) Theoretical aerodynamics of flexible wings at low speeds IV--
Experimental programme and comparison with theory. Vidye report 172.
C-~A~IBERS,L. I. (1966) A variational formulation of the Thwaites sail equation, Q. J. Mech. Appl. Math. 19,
221-231.
CHAPLEO,A. Q. (1968) A review of two-dimensional sails. University of Southampton, S.U.Y.R. Rept. 23.
DEKKER, TH. A. H. (1977) Performance characteristics of some sail- and steel-bladed wind rotors S.W.D.
Gronigen U. Public. 77-5.
DUGAN,J. P. (1970) Free streamline model of the two dimensional sail, J. Fluid Mech. 42, 433-46.
I~NK, M. P. (1967) Full-scale investigation of the aerodynamic characteristics of a model employing a sailwing
concept. NASA, TN D-4062.
FLUGGE,W. (1973) Stresses in Shells, 2nd Edn, Springer Verlag, Berlin.
GREE~nALGH, S. (1983) The two-dimensional inextensible lifting membrane airfoil--theory and experiment.
NADC Report 83096~0.
GREENHALGH,S. and CURnSS, H. C. (1984) Aerodynamic characteristics of a membrane wing. AIAA 2nd Applied
Aerodynamics Conference, Seattle, Wash., AIAA-84-2168.
GREENHALGH,S., CURTISS,H. C. and SMITH,B. (1984) Aerodynamic properties of a two-dimensional inextensible
flexible airfoil, AIAA J. 22, 865-876.
HESS, J. L. and SM~,~, A. M. O. (1967) Calculation of potential flow about arbitrary bodies, Pro#. Aerosp. Sci. 8,
1-138.
HOLLA, V. S., ROA, K. P., ASTHANA,C. B. and AROKraASWAMY,A. (1984) Aerodynamic characteristics of
pretensioned elastic membrane rectangular sailwings. Comp. Methods Appl. Mech. Eng. 4, 1-16.
HOWARD-WILLIAMS,J. (1976) Sails, 4th Edn, John de Graft.
IRVINE,H. M. (1979) A note on luffing in sails. Proc. R. Soe. A 365, 345 347.
IRVINE,H. M. (1981) Cable Structures, M.I.T. Press.
JACKSON,P. S. (ed.) (1982) The science of sail design. Proc. oJConJerence, University of Western Ontario
JACKSON,P. S. (1983) A simple model of elastic two-dimensional sails, AAIA J. 21,153 155.
JACKSON,P. S. (1984) Two-dimensional sails in inviscid flow, S N A M E J. Ship. Res. 28, 11 k 17.
JACKSON,P. S. (1985) The analysis of three-dimensional sails. Proc. C A N C A M , University of Western Ontario.
JACKSON, P. S. and CHRISTIE,G. W. (1986) Numerical analysis of three-dimensional elastic membrane wings.
(unpublished).
JAMES,R. M. (1972) On the remarkable accuracy of the vortex lattice method, Comp. Methods Appl. Mech. Eng. !,
75-79.
KRoo, I. 0983) Aerodynamics, aeroelasticity and stability of hang gliders. Ph.D. thesis, Stanford University.
Low, H. T. (1982) Planar two-dimensional flow past membranes. Ph.D. thesis, McGill University.
Low, H. T. and NEWMAN,B. G. (1986) Two-dimensional flow past bluff flexible membranes of low porosity,
J. R. Aeronaut. Soc. (in press).
MARCHAJ,C. A. (1964) Sailino Theory and Practice, Dodd-Mead.
MARCnAJ, C. A. (1976) Aero-hydrodynamics oJ Sailin O, Dodd- Mead.
MATEESCU,D. and NEWMAN, B. G. (1985) Analysis of flexible membrane aerofoils by a method of velocity
singularities Proc. C A N C A M , University of Western Ontario.
MENDENHALL, M. R., SPANGLER, S. B. and NIELSEN, J. N. (1968) Review of methods for predicting the
aerodynamic characteristics of parawings. AIAA 6th Aerospace Science Meeting, New York. Paper 68 10.
MILGRAM.J. H. (1971) The distortion of sails due to fabric deformation. Proc. HIS WA Syrup. Yacht Arch. 2nd, pp.
50~9.
MILGRAM,J. H. (1972) Sailing vessels and sails, Annu. Rev. Fluid Mech. 4, 397-430.
Aerodynamic theory for membranes and sails 27

MILLWARD, A. (1962) The aerodynamics of sails, J. R. Aeronaut. Soc. 66, 760-764.


MILNE THOMPSON, L. M. (1968) Theoretical Hydrodynamics, 5th Edn, Macmillan, London.
MUR~a, H. and MARUV^MA, S. (1980) Theoretical investigation of the aerodynamics of double membrane
sailwing aerofoil sections, AIAA J. Aircraft, 17, 294-299.
Mua~a, H. and MARUVAMA,S. (1982) Theoretical investigation of sailwing airfoils taking account of elasticities,
AI AA. J. Aircraft 19, 385-389.
MURAl, H. and MARUYAMA,S, (1984) Possibility of various airfoil shape modes and their steady state stability of
single membrane sailwing, Bull. JSME 2% Paper 225-13.
MVALL, J. and BERGER, S. (1969) Interaction between a pair of two-dimensional sails for the case of smoothly
attached flow, Proc. R. Soc. A 310, 373-391.
MYALL, J. O. and BEGGER,S. A. (1970) Recent progress in the analytical study of sails, Second AIAA Symposium
on Sailing, 9, 9-19.
NEWMAN, B. G. (1981) The aerodynamics of flexible membrane~ Proc. CANCAM, Moncton. (See also Proc.
Indian Acad. of Sci. 5, 107-129, 1982)
NEWMAN, B. G. (1985) Luffing of planar sails. Proc. CANCAM, University of Western Ontario.
NEWMAN, B. G. and GOLAND, D. (1981) Two-dimensional inflated buildings in a cross wind, J. Fluid Mech. 117,
507-530.
NEWMAN, B. G. and LOW, H. T. (1981) Two-dimensional flow at right angles to a flexible membrane, R. Aeronaut.
Soc.: Aeronaut. Q. 32, 243-269.
NEWMAN, B. G. and Low, H. T. (1984) Two-dimensional impervious sails; experimental results compared with
theory, J. Fluid Mech. 144, 445-462.
NEWMAN, B. G. and NGABO,T. M. (1978) The design and testing of a vertical axis wind turbine using sails, Energy
Conversion, 18, 141-154.
NEWMAN, B. G. and TSE, M.-C. (1980) Flow past a thin, inflated lenticular aerofoil, J. Fluid Mech. 100, 673-689.
NIELSEN, J. N. (1963) Theory of flexible aerodynamic surfaces. ASME. J. Appl. Mech. 30, 435-422.
NIELSEN, J. N. and BURNELL,J. A. 0963) Flexible-rotor design. Vidya, Paper 6.
NIELSEN, J. N., KRIEBEL, A. R. and GOODWlN, F. K. (1963) Theoretical aerodynamics of flexible wings at low
speeds l--one-lobed parawing~ Vidya report 84.
NIELSEN, J. N., BARAKAT,R., GOODWIN, F. K. and RtJt)lN, M. (1964) Theoretical aerodynamics of flexible wings at
low speeds II--two-lobed parawings. Vidya report 133.
NOVOZHILOV, V. V. (1959) The Theory of Thin Shells, Noordhoff, Groningen.
OR~S'rON, R. A. (1971) Theoretical and experimental aerodynamics of the sailwing, AlAA. J. Aircraft 8, 77-84.
PARKINSON,G. V. and JANDALI,T. (1970) A wake source model for bluff body potential flow, J. Fluid Mech. 40,
577-594.
PrnLLIPS-BIRT, D. (1962) Fore and Aft Sailiny Craft, Seeley, Service, London.
RAYLEIGH, J. W- S. (1929) The Theory of Sound, !, p. 306, Dover, N.Y.
ROaERT, J. and NEWMAN. B. G. (1979) Lift and drag of a sail aerofoil, Wind En0. 3, 1-22.
ROGALLO, F. M. (1954) Introduction to aero-flexibility. ARDC Unit, Langley, Va. Paper presented April 21, 1954.
SATCHWELL, C. J. (1982) Advances in lifting membrane analysis. Proc. 12th AIAA Symp. on Sailino, 28, pp.
206-223.
SHAW, S. B. (1984) Canadian ultralights take off. Industry, Trade and Commerce, Canada, Canadian Commerce,
March.
SNEYD, A. D., BUNDOCH, M. S. and REID, D. (1982) Possible effects of wing flexibility on the aerodynamics of
Pteranodon Am. Nat. 120, 455-477.
STACK, J. (1944) Tests of airfoils designed to delay the compressibility burble. NACA TN-976.
SWEENEY, T. E. (1961) Exploratory sailwing research at Princeton. Princeton Aero. Eng. Rept. 578.
TANNER, T. (1962) The application of lifting line theory to an upright Bermudan mainsail, J. R. Aeronaut. Soc. 71,
553-558.
TAYLOR, G. I. (1919) On the shapes of parachutes. Scientific papers of G.I.T., Vol. 11I, pp. 26-37.
THWARTS, B. (1961) The aerodynamic theory of sails. 1 Two-dimensional sails. Proc. R. Soc. A 261,402-422.
TUCK, E. O. and I-IASELGROVE,M. K. (1972) An extension of two-dimensional sail theory, J. Ship. Res. 16, 148.
VANDEN-BROECK, J.-M. (1982a) Nonlinear two-dimensional sail theory, Phys. Fluids 25, 420-423.
VANDEN-BROECK, J.-M. (1982b) Contact problems involving the flow past an inflated aerofoil, ASME J. Appl.
Mech. 49, 263-265.
VANDEN-BROECK, J-M. and KELLER, J. M. (1981) Shape of a sail in a flow, Phys. Fluids 24, 552-553.
VOELZ VON, K. (1950) Hauptaufs/itze, Profil und Auftrieb eines Segels. Zeitschfrift for Angewandte Mathematik
und Mechanik, Inyenieurwissenschqltliche Forschungsarbeiten 30, 22-39.
WARNER, E. P. and SHATSWELL,O. (1925) The aerodynamics of yacht sails, Trans. Soc. Naval Architects Marine
Eng. 33, 207-232.

APPENDIX

Lug lateen Square Spinnaker

You might also like