You are on page 1of 18

AIAA-83-1867

Airfoil Generation with a Desktop Computer


Using Lighthill's Exact Inverse Method
M.R. Schopper
David W. Taylor ~ a v aShip
l Research and
Development Center
Bethesda, MD

AlAA Applied Aerodynamics


Conference
Danvers, Massachusetts f

For permission to copy or republish, contact the American Institute of Aeronautios and Astronautics
1633 Broadway, New York, NY 10019
AIRFOIL GENERATION WITH A DESKTOP COMPUTER
USING LIGHTHILL'S EXACT INVERSE METHOD

M. R. Schopper flf3-39O 96 *
Aviation and Surface Effects Department
David Taylor Naval Ship Research and Development Center
Bethesda, Maryland 20084

Abstract The primary purpose of the paper is to demonstrate the types


of airfoils which can be generated with Lighthill's method using
Lighthill's 1945 exact inverse conformal-mapping method for the a desktop computer. Some desktop computers now have megabytes
incompressible potential flow design of two-dimensional airfoils has of memory, but the present program was developed on a machine
been reviewed and numerically implemented on a desktop computer with a working memory of approximately 55 kilobytes. The design
having 55 kilobytes of working memory. Numerous examples are method is reviewed and its numerical implementation is discussed
presented of airfoils and their velocity distributions. The designs in some detail. Numerous examples of symmetric and cambered air-
include both symmetric and cambered airfoils encompassing a wide foils with cusped trailing edges are shown which encompass a wide
range of lift coefficients and thicknesses. M. B. Glauert's methods variety of lift coefficients and thicknesses. The capability of employ-
of c&trolling the leading edge radius and applying rear loading are ing different leadingedge radii and varying the degree of rear loading
also demonstrated. Examples are given of various types of velocity are illustrated. Different types of upper-surface pressure recovery
decays over the rear portion of airfoils. Very good control over the distributions are shown to illustrate various degrees of velocity con-
velocity in the airfoil plane is demonstrated using piecewise con- trol in this region. An example of precise control using piecewise-
tinuous linear velocity segments. continuous linear velocity segments is included. The computer code
was written in the standard BASIC programming language. Pro-
Introduction grams were run on a Tektronix 4052 computer.

In 1945, ~ighthill'presented an exact inverse conformal-mapping Review of the Method


method for the incompressible potential flow design of two-dimen-
sional airfoils. With this method the shape of the airfoil is deter- The material presented is liberally taken from Refs. 1,4,7, and
mined in terms of the velocity distribution prescribed as a function 8. It is expected that the interested reader will study Refs. 1 and
of the angular coordinate around the circle into which the airfoil 4. The nomenclature will follow that of Lighthill and Glauert.
may be transformed. Lighthill presented examples of airfoils de-
signed using the method, and shortly thereafter applied the method Basic Theory
to airfoils with leading-edge suction.2 That same year, M. B. Gla-
uert presented four suction airfoils designed with Lighthill's For two-dimensional incompressibleflow a complex potential w(z)
method3 Glauert employed additional techniques to obtain suitable exists which is an analytic function defined by
velocity distributions. These techniques, which he documented in
1947,4 contained some very attractive design procedures including
the capabilities to rear load the airfoil and to allow some control
over the leading-edge radius. Most important for future users of
the method was an appendix to the article containing analytic ex-
+
where z = x + iy; is the velocity potential, and y is the stream func-
tion. The complex velocity is the analytic function defined by dw/dz,
pessions which greatly facilitate the application of the method. and dw/dz = u - iv, where u and v are the velocity components
Although no such claim was made, Glauert's paper provided those in the direction of the axes. In polar form dw/dz = q e - h, where
with no special gift the tools and techniques for a straightforward q is the magnitude of the fluid velocity, and X is the direction of
implementation of Lighthill's inverse design method. the flow with respect to the x-axis. With unit velocity at infinity
and at an angle of attack of a with respect to the real axis, the com-
Although Lighthill's method has been a basis for later inverse plex potential of the flow in the 5 -plane around the unit circle
computational design codes,S the method itself does not seem to 141 = 1 is given by the superposition of the complex potentials of
have been widely employed in airfoil research or design; at least, uniform flow, a doublet, and a vortex, as
it has not been as widely referenced as might be expected. Sat06
has utilized the method, and no doubt other articles and papers exist.
Books devoted to potential flow generally do not discuss inverse
methods, and few of the books on airfoils or aerodynamics men-
tion Lighthill's method (or inverse methods). Two notable excep-
tions are the books by Robinson and Laurmann7 and Thwaites.8 where r is the circulation. The complex velocity is
Both of these British books review Lighthill's method. The situa-
tion is similar with airplane design books; in the chapters dealing
with airfoil design methods few contain discussions of inverse
methods or mention Lighthill's method. It thus appears that it is
not generally recognized that Lighthill's method can be implemented
without a major research effort, and that it can be programmed on On the unit circle 5 = eie and here
a small computer. The method has been and is being used to design
single-element airfoils in an R&D environment; it should be emi-
nently suitable in an academic environment where it could be used
as a teaching and research aid for potential flow and boundary layer
studies.
This pawr is declared a work of the U.S.
The flow speed over the surface is

Therefore, if qo is prescribed in the circle plane as a function of


With the trailing edge stipulated to be at 5 = 1 on the circle, the 8, the airfoil coordinates can be found if X(8) is known.
Kutta-Joukowsky condition requires the velocity to vanish at 8 =
0. Hence, from Eq. (5) the circulation is 4n sina and, therefore, The matter of the sign in Eq. (14) may be resolved as follows.
To determine X(O), use is made of the equation for the complex
velocity dw/dz = q e- ix. Here X indicates the direction of motion.
This is consistent with the definition of x by dz = ds eix at the air-
foil surface, since the flow must be tangential to the surface, pro-
vided s is measured in the direction of flow. Thus, ds must be mea-
sured in the local direction of the flow, and in Eqs. (12) through
Suppose now the region outside the circle is conformally trans- (14) the + sign applies if the direction of flow is in the direction
formed into the region outside an airfoil in the z-plane by means of increasing cos8. Since the front stagnation point occurs at 19 = n,
of an analytic function z(5), such that the airfoil trailing edge cor- the direction of the flow always coincides with the direction of in-
responds to 5 = 1 on the circle and the velocities at infinity are creasing cos8. Equation (14) may thus be taken as
equal. Now the velocity in the airfoil plane is dw/dz = (dw/dt)
(dC/dz); thus for equal velocities at infinity, dC/dz-* 1 as 5 -,0
(or z = 5 as 5 -, 03). On the airfoil at incidence a, the modulus of x = - $&cas X sin 8 d8
the velocity, q, is now given by
y = - J L90 sin X sin 8 dB

Returning to theX(8) problem, dwo/dz is given by qo e-ix. Thus


On the airfoil surface dz = ds eix, where ds is the element of
length of the airfoil, and X is its local incidence or slope. Correspond-
ingly, on the circle ds = d(ei8) = ieie dB, and so q, becomes
Now dwo/dz is an analytic function of z in the region outside the
q, = 4 1 sin-cos
8
2
(--
I9
2
a)-/dB
ds
airfoil, including the point at infinity and, therefore, is an analytic
function of 5 outside the unit circle. The term dwo/dz does not
vanish in the region, and becomes equal to one at infinity. Hence,
The velocity modulus at no lift (a = 0) is thus ln(dwo/dz) is also an analytic function of 5 outside the unit circle.
Thus In qo and x are harmonic conjugate functions. They can be
expanded in conjugate Fourier series in 9 on the circle and are related
qo = 2 ( sin 8-1dB by Poisson's integral. Poisson's integral relating the conjugate func-
ds tions is

hence, X(8) = 1
-So 2"
In qo(t) cot -
I9 -t dt
2n 2
8
cos ( 3 - a)
40 = qo (10) where the Cauchy principal value is taken at the singularity 19 = t.
COS
I9
- No constant appears since X is zero at infinity. If In q, is an even
2 or odd function of 8 the integral is simplified.
Equation (10) shows that once qo(8) is known the value of q, (8) Because there is not always an airfoil corresponding to any qo(8),
at any other incidence can easily be determined. the q0(8) term cannot be prescribed arbitrarily. Three integral con-
~ ~ u a t i o(9)
n can be written as ditions exist, as derived in Eqs. (18) through (23). The mapping of
the z- and <-planes must be of the form

The equation dz = ds eix may also be expressed as


where the coefficients a, are constants. Hence

From Eq. (3)


Substituting Eq. (11) into Eq. (12) yields

2 .
dz = 2 I-e~Xd(cos%)
qo Dividing Eqs. (20) by (19) yields

Taking dz = dx + idy and eix = cos X + i sin x and separating


the real and imaginary parts gives
To find the first term in the expansion for ln(dw,/dz), use is made The procedure employed in task 1 is the method of "direct design
of the formula ln(1 -t) = t - t2/2 + t3/3 - . . . Then, . at incidence." In this method the upper-surface velocity is prescribed
at an incidence a l , and the lower-surface velocity at a lower inci-
dence a2. The incidence angles a1 and a2 are design parameters and
the range of incidence from a1. to a2- is referred to as the incidence
range i f the airfoil.

From Eq. (10) it follows that on the upper surface


Equation (22) shows that because n starts at 2 the conjugate Fourier
series for In q, and - X on the circle have no constant terms or 8
terms in cost3 and sine. This implies that In q, satisfies the follow- cos -
ing three conditions: In qo = In +Inqa1
8
cos (? - all

and on the lower surface

The first condition is a consequence of requiring unit velocity at


infinity; the other two conditions ensure that the airfoil defined by
In q, is a closed contour. Where the upper and lower surfaces join
Once the airfoil shape has been determined, the lift coefficient
can be found quite easily. By the Kutta-Joukowski law the lift co-
efficient of an airfoil is 2 r k , where c is the chord length. For
cos (8- a1 =
2
+ cos (--28 ad
the present case r is 4n sina; thus,

Thus,

The chord length is found from the resultant airfoil shape. The air-
foil is computed in its no-lift attitude and, therefore, the no-lift angle
can be determined from the orientation of the computed airfoil.
The angle a in Eqs. (10) and (24) must be measured from the no- The joining at 8 = n + a1 + a 2 is near the nose.
lift angle.
For a cambered airfoil, therefore, the general expression for the
Expressions for the moment coeffieient and other aerodynamic velocity distribution may be written as
properties are given by Lighthill' and Glauert.4

Design Procedure

lcos~~
As envisioned during the present effort, the design procedure con- L
sists of the following six basic tasks: In , a1 + a:! <B<n + a1 +a:!
1. Prescribe In q, over the airfoil as a function of 8 in the circle
plane; the expression for In q, should have three arbitrary
constants.
- 2. Determine the three constraints using the three integral condi-
tions of Eq. (23).

3. Find X(8) from Poisson's integral, Eq. (17).


16
4. Compute the airfoil shape by evaluating the integrals in Eq. f43.

5. Measure the chord length and the orientation of the airfoil; deter-
mine the lift coefficient using Eq. (24).

6. Obtain the velocity distribution using Eq. (10) and the prescribed
velocities.
where S(0) is the value of In qa, on the upper surface and of In
The art is to choose a distribution of In q, which enables the design qaz on the lower surface. When the airfoil is flown at an angle of
requirements to be met, yet is simple enough to allow the integrals attack of al, the velocity distribution on the upper surface is es@),
for X(8) and the three constraint conditions to be evaluated ana- where S(8) is the prescribed distribution on the upper surface.
lytically. Then X(8) becomes a matter of evaluating expressions in- Similarly, when flown at a2, the velocity distribution on the lower
stead of computing an integral, and the three integral conditions surface is es(e); where S(8) is the prescribed distribution on the
reduce to a system of three algebraic equations to solve. The inte- lower surface. Many examples are considered in the Results and
grals involved in the computation of the airfoil shape are evaluated Discussion which provide good insight into appropriate quantities
numerically. to prescribe for S.
Glauert4 determined the conjugate of the first two terms in Eq. erty. The symmetric airfoils generated for the present report, in fact,
(29) as well as the contribution of the terms to the three integrals were produced with codes which were among the first developed
of Eq. (23). He also determined the conjugate as well as the values for the airfoil design task by the author.
of the three integrals for a number of simple trigonometric func-
tions prescribed between two arbitrary angles. These simple func- For a symmetrical airfoil a1 = - a2 = a , and Eq. (29) takes
tions are suitable for velocity prescriptions around the circle. The the simpler form
results are conveniently displayed in Appendix I of Glauert's report.4
This appendix was a vital document in the present effort.

As an example of the conjugate function and of the contribu-


tion to the three integrals, consider as part of S the constant func-
tion k prescribed over the range K O - + . From Glauert,4 the con-
jugate function is(k/n)lnl sin% (8 - A)/sin% (8 - p)l. The three
integrals of Eq. (23) are k(p - A), k(sin p - sin A), and k(cos A
- cos p), respectively. Other functions, such as the k cos8 prescribed where S = In q,, and In q, is an even function of 8 . The tasks
over an interval, give corresponding expressions which are only outlined above are still followed (but are again simpler; e.g., the
slightly more complicated. In this example, k would be an unknown no-lift angle is zero).
constant to be determined by satisfaction of the three integral
conditions. An example discussed by both Lighthilll and Glauert4 is the case
where
It should now be clear that if simple trigonometric functions are
used in a piecewise continuous manner to define S completely around
the circle, each of the three integral conditions reduces to a condi-
tion that the sum of algebraic quantities representing the pieces be
zero. By using simple functions for S containing three unknowns,
where the unknowns are often the functions amplitude coefficients,
the three integral conditions of task 2 can be reduce to the solving
of a system of three algebraic equations for the three unknown The velocity at the design incidence is constant from the nose to
constants. 8 = /3 and thereafter decays steadily. There are only two unknown
constants here, 1 and k, but these are sufficient because the third
Using Glauert's Appendix I,4 task 3 can be reduced to the sum- integral condition of Eq. (23) is satisfied automatically for sym-
ming of algebraic quantities corresponding to the pieces of S. The metric airfoils. After the appropriate expressions are determined
quantities involve the three constants determined in task 2. The air- from Glauert's report,4 the two integral conditions are reduced to
foil shape in task 4 can now be computed with a straightforward a linear system of two equations. This is one of the nice advan-
numerical integration using Simpson's rule. Here, it is important tages of simple symmetric airfoils. A linear system solver is not even
to keep track of the developing X,8 relationship because this is needed needed beacuse the system can be manipulated by hand to give
in task 6. Once the airfoil shape is computed, the chord and no-lift algebraic expressions for the two unknowns. At the design incidence
angle can be measured and the lift coefficient computation of task the upper-surface constant velocity is el; therefore as soon as 1 is
5 is trivial for any angle of attack. For the velocity distribution task determined the velocity can be determined. This velocity informa-
Eq. (10) is used. By taking ratios of velocities at two different angles tion-which comes early in the computer run-is nice to know (es-
of attack, q, can be eliminated to give, 'for incidences a and ai, pecially for complicated cambered airfoils), because it can some-
times provide a hint as to whether the upcoming airfoil is likely to
be satisfactory (e.g., too thick or too thin). This early velocity in-
formation was especially important in the precomputer days; it could
save days instead of seconds of wasted time.

Letting i take on the values 1 and 2 to represent the upper- and lower-
surface design angles of attack, the desired result is
Cambered Airfoils

With the move to cambered airfoils the symmetry and simplicity


is lost. Lighthilll found that if an unsymmetrical airfoil is designed
so that it has the same maximum velocity on the upper surface at
incidence a1 as on the lower surface at a2, the resultant airfoil has
an undesirable shape. This was stated in connection with airfoils
having a suction slot; in fact, all of Lighthill's cambered airfoils
were suction airfoils. To avoid the problem, Lighthill stated that
where is is necessary to have the maximum velocities at the two ends of
the incidence range different, with that at the upper end being
greater. Lighthill employed a velocity distribution which was con-
tinuous at the nose (although the derivative was not) and had the
desirable upper- and lower-velocity difference. The conjugate func-
tion was obtained in an approximate manner. Glauaert4 chose a
different form of the velocity distribution, one which did not re-
By keeping track of the relation between 8 and x, (i.e., developing quire approximations. Clauert simply added a term to In q, over
x as a function of O), Eq. (31) can be used to express the velocity a portion of the upper surface, but this was done in a manner which
distribution over the airfoil for any angle of attack. ensured continuity of In q, and d(ln qo)/d8 at the locations
8 = n + a1 + a2. The significance of the continuity of the derivative
Svmmetric Airfoils is discussed in the next section of the paper. (It may be noted that
all of Glauert's cambered airfoils were suction airfoils also.)
In the development of an airfoil design computer code, it is natural
to consider symmetric airfoils first. Symmetrical airfoils are simplest In Glauert's method, S is now taken to include the upper-surface
to compute because advantage can be taken of the symmetry prop- terms
ing the conjugate function for a constant k prescribed over the in-
terval A<O<p. Examination of the conjugate, given earlier, shows
that as 8 A the conjugate behaves as (k/n)ln(8 - A). The con-
+

jugate of d(ln qo)/d8 is dX/d8, as shown by differentiating the


conjugate Fourier series. There, dX/dO becomes logarithmically in-
finite at 8 = n + a1 + a2, because in the method of direct design
at incidence d( In qo)/d8 is discontinuous at this point. Now the
curvature of the airfoil surface is dX/ds = (dX/dB)(dO/ds), and
since dO/ds remains finite and non-zero, the curvature has a
where a, = (al - a2)/2. The E term provides the velocity incre- logarithmic infinity. This implies that at 8 = n + a1 + a2 the sur-
ment on the upper surface. The quantity E is unknown and is one face has a zero radius of curvature. As a result of the zero radius
of the three constants to be determined by satisfaction of the three of curvature, this airfoil would have a velocity peak at the nose
integral conditions. With the terms provided in Eq. (35) there is con- for angles of attack above a]. Boundary layer separation in the im-
tinuity of In q, at both 8 = n + a1 + a2 and 8 = n + a1 + mediate recovery downstream of the peak would seriously limit the
a2 - E. The choice of E cot a, stems from consideration of performance of the airfoil. Lighthill recognized the problem and
d( ln qo)/d8. Examination of Eq. (29) shows that at 8 = n + a1 presented a leading-edge term to add to In q, to eliminate the
+ a2 there is continuity of In q,, but d(ln qo)/d8 has a discon- derivative discontinuity. An asymptotic formula was used for the
tinuity by the amount -cots,. This is a consequence of using the conjugate expression. Glauert used a different term which had more
method of direct design at incidence. With the addition of the terms deisrable characteristics.
in Eq. (35) d(ln qo)/d8 has been rendered continuous at 8 = n
+ a , + a2. However, there is now a discontinuity in the derivative Glauert's leading edge term eliminates the discontinuity in d(ln
by the same amount of -cots, at the point 8 = n + a1 + a2 - r. qo)d8. The value of the discontinuity is -cot a, and the E techni-
Thus, the r technique did not cure the derivative discontinuity prob- que for cambered airfoils merely shifted the location of the discon-
lem-but that was not its purpose. Its function was to provide the tinuity from 8 = n + a1 + a2 to 8 = n + a1 + a2 - E. Because
upper-surface velocity increment. Glauert solved the derivative prob- the magnitude of the discontinuity is the same for symmetrical or
lem with a leading-edge term. cambered airfoils, Glauert's leading-edge term is valid for both types
(for symmetrical airfoils, E = 0, a 2 = - a l , a, = a1 = a).
A simple cambered airfoil could be designed using Glauert's r
technique in the following manner. With cambered airfoils the
velocity must be specified around the whole airfoil, and, of course,
-
Glauert's leading-edge- term eliminates the discontinuitv in the
derivative by, respectively, adding and subtracting one-half the
the velocity must be continuous. Glauert's velocity terms in Eq. (35) discontinuitv as it is avvroached from each side. Letting 4 = 8 -
did satisfy velocity continuity on the,upper surface. To complete (n + a1 + i2 - E ) , ihe leading-edge term to be added'to S is
the design problem, consider a cambered airfoil design that is akin
to the symmetric airfoil example presented earlier. The velocity
prescription S(8) contains the term 1, which is now prescribed from
0 to 2n instead of 0 to n. Just as in the previous example, assume
that at the prescribed angle /3 the upper-surface velocity begins a
decay described by - k (cos 8 - cos /3). With the addition of the
+
At = 0, dK,/d+ = 2 %cots,, as required There is no discon-
terms of Eq. (33, the velocity on the upper surface at incidence
a1 will be constant from 8 = rr + a1 + a2 - r to 8 = /3. The velo- +
tinuity in d2Kn/d+2 anywhere. Furthermore, at = + n/Y2n),Kn
city here will be el +Ecotao. There must be a velocity decay region and its first two derivatives equal zero, so the function blends
on the lower surface; assume it begins at the prescribed angle d . smoothly at the limits of its domain. The velocity is reduced by K,,
At incidence a2 the lower surface will then have a velocity of el from with the largest reduction occurring at 4 = 0 where K, = - (n - 2)/
8 = n + a1 + a 2 to 8 = d . The decay must be prescribed so that (4n)cotao. As n decreases, the reduction increases and the reduction
S is continuous at the trailing edge, 8 = n. On the upper surface occurs over a wider range of +.One effect of this is that with small
at 8 = 0, d has the value E cot a, - k (1 - cosp). Thus, an ap- n the velocity peaks formed for incidences above a1 are reduced
propriate decay for the lower surface would be [r cot a, - k(l - and rounded. The effect on the airfoil shape is a rounding of the
cosp)] (cos8 - cosd )/(I -cosd ). Examples of airfoils with velocities nose.
prescribed in this manner are shown in Results and Discussion.

The three unknowns in the above example are 1, k, and E. A com- There is a restriction on n which depends on a,. In essence, a
plication arises in the determination of these values. When the con- large radius cannot be put on a thin airfoil. If an attempt is made
tributions from the new upper-surface S terms given in Eq. (35) are to do this, a negative radius of curvature will occur. The thickness
included in the algebraic system of equations comprising the three
of an airfoil depends on a,. Glauert's analysis showed, for exam-
integral conditions, the system becomes nonlinear. Thus, in going ple, that for n = 6 the incidence range (2a0) should be specified
from symmetrical to nonsymmetrical airfoils, the problem of sat- to be over 4 degrees; for n = 3 the range has to be greater than
isfying the three integral restraints jumps from one of needing no 8 degrees.
system solver to one of needing a nonlinear system solver. Once
the hurdle of incorporating a nonlinear system solver into the com-
puter program is cleared, the complication ends. The equations to
solve are well conditioned, and no difficulty has ever been en- Glauert4 presented tables for the conjugate function of Kg, but
countered in solving the system of equations. t'here appear to be some problems with the analytic expressions as
presented. For example, expressions for the three integrals in Eq.
It should be noted that it is possible to prescribe values for a l , (23) are presented for Kg (Appendix 111); however E is not involved
a2, /3, and d which willproducenegative values for r. In this case, and this cannot be correct. If Glauert was actually considering the
the whole problem should be restructured so that rcota, is added symmetrical airfoil case (E = O), the results appear to be of the wrong
to the lower surface, and Eq. (35) restructured in an appropriate sign. Sign discrepancies also appear to be present in the conjugate
manner. Then E will be positive, and the discontinuity in d(ln q,)/d function. Use of Glauert's conjugate and integral expressions
8 will be moved to n + a1 + a2 + r. These airfoils are usually resulted in unsuccessful results. Thus in the present effort, the con-
of little practical interest. jugate and the three integral expressions have been rederived, and
the results are presented in a general form for all n. Glauert often
Leading-Edge Term used the letters G, A, B, and C to represent, respectively, the con-
jugate function and the values of the three integrals in Eq. (23).
If a function has a discontinuity, its conjugate has a logarithmic For convenience, this practice has been followed here. Letting
infinity at the same point. This can be seen, for example, by examin- R = n + a l + u ~ - E ,
The procedure, then, is that in the specification of S the velocity
on the lower surface is defined to be greater than the velocity on
the upper surface at the trailing edge by the amount, for example.
b, where b is a quantity input by the designer and the larger the
value the greater the rear loading. If b = 0, then no rear loading
+ 2)
+$ kos n4 ln
sin 'h

sin 'h
(+

(+ - ?n;)
+ 2n
n sin n+
occurs. The quantity % b P , is then added to the S terms. A
negative value for b will produce an airfoil with the upper surface
having the greater velocity near the trailing edge. The addition of
the rear loading term produces no discontinuities in either In qo or
d (In qo)/d8 at 8 = + n/(2m).

"-2 sin [(n -j - 1) L] Glauert presented the conjugate function and the three integral
+ 2 ,=o
,I 2n sin [(j + l)+l expressions for Pg (and the Pg conjugate function is tabulated). The
n-j-1 more eneral expressions for P,, derived by the present author, are

1 n2
A =~ ( - - 11 cot a.
8

B = (- n2 c o IIs - l ) c o s R c o t o o
- (1 + sin me) In +-
2m
cos me
n2 -1 2n

C= (-
n2
n2
-
cos %- 1I sin R cot a. m-j

The function X (8) is defined by

1 8 In sin-dt
X(8) = --J t
n o 2

and is tabulated in Glauert's report. The function will be discussed


later.
[
C = 2 I--
2,
q
- 1 cos 21-17
Suction Airfoils
Rear Loading
The majority of airfoils presented by Lighthill and Glauert were
On an airfoil with a cusped trailing edge the velocities on the up- suction airfoils. This type of airfoil apparently was originally sug-
per and lower surfaces usually approach the common trailing-edge gested by A. A. Griffith, whose idea was to concentrate all of the
velocity with deceleration gradients which are nearly constant in pressure recovery on a surface into a single point. At this point a
magnitude-although not equal-over about the last 20% of the suction slot was to be located so that the boundary layer could be
chord. Most often, the upper-surface velocity is greater than the removed to prevent boundary layer separation. Symmetrical air-
lower-surface velocity over the rear region of the airfoil. The pressure foils of this type, called Griffith airfoils, were designed and tested9
difference between the upper and lower surfaces drops steadily, and (although they were originally designed using an approximate
the rear portion of the airfoil develops little lift. A rear-loaded air- method of Goldstein). Glauert concentrated on cambered airfoils
foil is one in which the pressures on the upper and lower surfaces with a single slot on the upper surface3; such airfoils have also been
are kept different until the very end of the airfoil. The velocity gra- tested. l o
dients on these airfoils thus become very steep near the trailing edge
of the airfoil. The upper-surface velocity near the trailing edge can A suction airfoil can be designed with the inverse method by sim-
be either greater or less than the lower-surface velocity. Rear loading ply requiring the velocity to have a discontinuity at the desired loca-
can be employed for such purposes as boundary layer control, pitch- tion. For example, with a slot to be located at 8 = 0 on a cambered,
ing moment control, or adding lift to the airfoil. single-slot airfoil, the terms prescribed for S upstream of the slot
could be the usual constant velocity 1 + E cota,; downstream of
With Lighthill's inverse design method, rear loading can be em- the s b t a constant velocity of 1 - k would be used, instead of 1
ployed in a rather direct fashion. In Glauert's approach the velocities - k (cos8 - cosp). The magnitude of the velocity drop is not pre-
on the upper and lower surfaces must be specified so that there is scribed but comes from the solution (k is an unknown constant).
a velocity difference near the trailing edge. The trailing-edge term At incidence a , the magnitude of the velocity drop is el + m t a o
blends the two velocities together at the trailing edge by adding half - el - k. A positive or negative velocity gradient can also be
of the velocity difference to the slower velocity and subtracting half prescribed for the region downstream of the slot by simply adding
from the faster velocity. The trailing-edge term to be added to S is _t re, &&/I. The value of r is a prescribed quantity. Of course,
rear loading can also be applied.

- 1 - sinm8 , - - ! ! - < @ < ( I At 8 = /3 there is a singularity inX(0) and, therefore, the airfoil
2m shape must be generated in two segments. The region downstream
p m =( (39) of the slot on the upper surface is created via an integration of Eq.
1 - sin m e , 0<8<& (14) from 8 = 0 to 8 /3. The lower surface, nose region and up-
-+

per surface ahead of the slot come from integrating from 8 = 2n


to 8 + /I. No actual slot opening is generated by this method.
Lighthill and Glauert discuss the fact that the surfaces ahead of and
The blending occurs over the region of 8 = rt n/(2m) in the circle behind the slot come together in a logarithmic spiral. The slot open-
plane, so large values of m produce a sudden blending with very ing must be created by altering the surface shape after the airfoil
steep velocity gradients. is generated. The region over which the surface shapes curl together
is very small and would be removed with any practical slot open- The first 30 nonzero Bernoulli numbers may be found in Ref. 12.
ing. A slot opening would be created if a sink were placed at the
velocity discontinuity location, as discussed by Lighthill1 and
Glauert .4 Evaluation of X(8) is the next task. As indicated, this requires
the summation of the G expressions obtained from Appendix I of
Numerical Implementation Glauert's report for all of the terms in Eq. (29). The resulting ex-
pression is long and is evaluated in a subroutine. The expression
The alogrithm developed in the present effort essentially follows contains a number of logarithmic terms involving trigonometric ex-
the six tasks listed earlier. It is a relatively straightforward implernen- pressions which become singular at certain angles (e.g.,
tation of Lighthill's method as applied by Glauert. To date, no time In lsin l/2 (0 - A)/sin ?h(0 - p)l , where A and p are given angles such
has been taken to review the algorithm to see if significant improve- as the velocity decay angles p and 6). If done properly these log
ments in speed and storage requirements could be made. The com- terms will either cancel or be multiplied by expressons which go to
puter programs were written in standard BASIC. zero faster than the logarithms approach infinity (suction airfoils
will retain a singularity at the slot).
Because of limited storage capacity no attempt was made to
develop a general code that would allow the user to specify the type As given by Glauert, the G function for the first two terms of
of decay law at run time. Instead, individual programs were writ- Eq. (29) contains the integral F function defined by
ten for each type of decay law. Differences in the programs exist
primarily in two subroutines: one evaluatingX(8) and the other con-
taining the algebraic expressions associated with the three integral
equations. Implementing the first task, that is, prescribing In qo(8),
thus rests largely on the writing of these two routines. Appendix
I of Glauert's report4 is used to obtain the G,A,B, and C algebraic
expressions for each term prescribed in S as well as the first two
terms in Eq. (29). The expressions are then combined and inserted This function is tabulated in Table 1 of Glauert's report. The tabular
values were used in a spline-fitted manner as described earlier. If
in the routines. The algebra is not tricky, but it is a little long; the
desired, this integral may be evaluated as follows. The integrand
expression for X ( B ) can take over 40 lines of code. The developer
is singular at t = 0 and, therefore, is broken into two parts as before.
should be prepared to go over-and over-the algebra and the cod-
ing during the debugging stage. As part of the first task, provisions The second part can be accurately evaluated using, for example,
Romberg integration. The first part may be integrated using the
must also be made for user input at run time of the values of the
approach given in the text by ~ t k i n s o n !for
~ evaluating the integral
design parameters (such as a l , a2, the velocity decay angles /I and
of g(x) In x from 0 to A. The g(x) term is expanded in a Taylor
d, and the leading- and trailing-edge fhctors n,m,and b).
series, and each term of the integral is then obtained by integrating
by parts. The result in the present case can be shown to be
The second task is the solution of the nonlinear system of equa-
tions fj(xl,x2,x3) = 0, i = 1,2,3 associated with the three integral
conditions. For example, f l is simply the sum of the A expressions
for each of the terms describing In q,; xl, x2, and x3 represent the
three unknown quantities. For the cambered airfoil programs written
so far, the unknowns have always been the 1, e , and k quantities
discussed in the examples. Newton's method has been used to solve
the system. Newton's method, which requires a linear system solver, The fourth task, which is the last major task, is the integration
also requires evaluation of afi/axj.The derivatives were approx- of Eq. (14) to obtain the airfoil shape. Using Eq. (lo), the 2/ qo
imated by [fi(xj + h) - fi(xj)]/h, where h was a small number. term in Eq. (14) can be replaced by
Newton's method is an iterative procedure and requires an initial
solution guess. The values 1 = 0.8, E = 0.02, and k = 0.5 were
often used. As stated, no problems have been encountered with the
solution. The time for convergence is only a small fraction of the
total problem computation time.

The fi for the terms prescribed in S involve simple trigonometric


expressions. The A expression for the first two terms in Eq. (29), To reduce the number of cosine evaluations Eq. (45) can be writ-
however, involves the X integral function defined in Eq. (38). This ten as
function is tabulated in Table 2 of Glauert's reportd, and the
tabulated values were used in the present effort. This was done by 2 = -
2 sina (cot a
first fitting a cubic spline through about 80 of 180 points given in -
qo qo
+ tan-)20
the table. The computed spline coefficients, which allow the spline
to be evaluated at any point, were then saved and used in all airfoil
design programs. If desired, the table may be generated using the
following procedure. The integrand is singular at t = 0, but the provided a f 0. Now q, is simply eW), and Eq. (14) finally can
integral may be evaluated by breaking it into two parts, viz., be written as
t dt =
In sin -
2
5 A In sin f- dt
2
+ A
t
In sin 2 dt (41) 0
(cot ai + tan -2t ) sin t
x(8) = 2 sinai lo ,S(t) cos X(t) dt
where A is appropriately small. The second integral, from A to 8 , (47)
may be integrated accurately using Romberg integration. The in-
tegral from 0 to A may be handled by expanding the integrand in a
Taylor series (see formula 20.48 of Ref. 11) and integrating term
by term. The result contains Bernoulli numbers and may be writ-
ten as
This is the form used in the present programs. For a1 + a2<0<n
+ a1 + a2, ai = al; for the other half of the circle ai = a2. As
8 is increased from 0 to 2n, the airfoil is generated in its no-lift
orientation.
The integration in Eq. (47) may be performed using Simpson's Symmetric Airfoils
rule. The circle is partitioned into sectors, and partitions are located
at the angles where the prescription of S changes or where ai Many of the properties of the design method are best illustrated
changes. For the typical cambered airfoil design with a rear-loading with symmetric airfoils. The effects of changes in design parameters
capability, there will be 10 sectors. Partitions will be located at 8 are readily apparent with symmetric airfoils in that the changes af-
equal to 0, + n/(2m), a1 + a2, /3, n + a1 + a2 - r , f n/(2n) fect both the upper and lower surfaces in a similar fashion. For
n + a1 + a2 - r , n + a1 + a2 and d, where (as in the examples) symmetric airfoils, a2 = - a1 and d = - 0. Because of the sym-
/3 and d are the angles on the upper and lower surfaces, respectively, metry the calculation need be performed for only one surface.
where the velocity decay begins. At run time, the user inputs for
each sector the number of surface coordinate points to be defined The first results illustrate the effect of the velocity decay angle
in the sector. The greater the number of points, the greater is the /3 on the airfoil and velocity distribution. The design angles of at-
accuracy of the integration. The x and y arrays have presently been tack are + 5 degrees, and the results for the three decay angles of
dimensioned to 300. /3 = 45,90, and 135 degrees are shown in Fig. 1. The cosine decay
law is used. Important observations are: (1) The chordwise loca-
To determine the lift coefficient using Eq. (24), the chord must tion of the velocity decay point is approximated by the projection
be measured. This task is done by simply scanning the points to onto the horizontal axis in the circle plane of the point on the cir-
see which one has the greatest distance from the origin, which is cle at the angle /3. Thus, x/c of the velocity decay point is approx-
the tail. This distance is saved as the chord length, and the no-lift imately given by (1 + cos8)/2. The correspondence between points
angle is given by tan- l(y/x) of the maximum-distant point. For on the airfoil and on the unit circle is described by x/c = (1 +
plotting purposes the airfoil is rotated so that the chord is horizon- cos8)/2. (2) The smaller the decay angle, the longer the rooftop
tal, the lengths are scaled by the chord, and the nose is placed at region and the thicker the airfoil. (3) The cosine decay law does
the origin. indeed produce velocity decays which are the steepest at 8 = /3. (4)
These airfoils were generated without the leading-edge term, and
Because x has been determined as a function of 0 , the last task- the rooftop velocities begin abruptly. This is due to the zero radius
that of computing the velocity distribution over the airfoil at any of curvature at the nose. (5) The rooftop velocities do not differ
particular angle of attack-may easily be done using Eqs. (31) and significantly, nor do the lift coefficients. A long rooftop region is
(32). The values of the quantity es@)/cos(0/2 - ai) in Eq. (31) are generally conductive to high lift performance, however, for sym-
actually stored in an array at the time the integrands in Eq. (47) metrical airfoils, the long rooftop specification also gives a lower-
are being evaluated. Thus, this array is available for repeated use surface velocity with its maximum velocity at the decay point nearly
in Eq. (3 1) for velocity distributions at a variety of angles of attack. matching the rooftop velocity. The lift coefficients for all of the
symmetric airfoils presented in the paper agree to within 2 percent
Results and Discussion of the standard formula CL = 2n (1 + 0.77 t/c)sina.
The remainder of the paper presents illustrations of some of the The effect of the design angle of attack is shown in Fig. 2. Here
types of airfoil which can be generated. The results are typically d = - /3 = 90 degrees; airfoils with design angles of attack of 2.5,
displayed with figures showing both airfoil shapes and the surface 5.0 and 10.0 degrees have been plotted. Again, the cosine decay
velocity distributions. In most figures, a single parameter is varied, law is used. The figure shows rather dramatic changes in lift coef-
and two or three designs are superimposed to show the effects of ficient and thickness among the three airfoils. These differences il-
the parameter variation. With one exception, the velocity distribu- lustrate two general observations concerning the design method
tions are shown for the airfoil at its upper-surface design angle of which are also valid for nonsymmetric airfoils. (1) The most domi-
attack. At this angle the upper-surface velocity has always been pre- nant parameter affecting the lift coefficient is a l , the upper sur-
scribed to be constant for some length, and the velocity is called the face design angle of attack; and (2) the most dominant factor af-
rooftop velocity. The expression "design angle of attack" stems fecting the thickness is a1 - a2, the difference between the upper-
from the method of direct design at incidence, and use of the ex- and lower-surface design angles of attack (the incidence range of
pression is not meant to imply that the airfoil is designed to be flown the airfoil). These observations are reinforced later in connection
at the "design" angle. In fact, flight at this angle generally should with a cambered airfoil. Figure 2 also shows the expected result that
be avoided, because a slight increase in incidence can result in an as the lift (and circulation) increases, the stagnation point on the
adverse pressure gradient near the nose and possibly a laminar lower surface moves further downstream.
separation bubble and/or early boundary layer transition. Note that
all angles are with respect to the no-lift position. The no-lift angles In Fig. 3, three simple forms of velocity decay law are compared.
are given in the figure captions. The order in which the angles are For all three airfoils, a2 = - a1 = 5 degrees and d = - /3 = 90
given corresponds to the order of the airfoils as ranked by lift co- degrees. The three decay laws illustrated are the cos8, sine, and 8.
efficient; the no-lift angle of the airfoil with the highest lift coef- The results show that there is little change in lift coefficient and
ficient is listed first. that the velocity distributions over the front half of the airfoils are
nearly identical. The shape of the aft part of the airfoils to some
extent mimics the shape of the velocity decay. The sin8 decay law
The terms and nomenclature used in this section are as follows. gives a velocity distribution whose decay becomes steeper as the trail-
The upper- and lower-surface design angles of attack are q , and ing edge is approached-a feature which is unattractive from a
02, respectively. The angle in the circle plane where the upper- boundary layer standpoint. The 0 decay law produces a velocity
surface rooftop velocity ends is called /3. The angle /3 is measured distribution which is nearly linear until very close to the trailing edge.
in the same manner as is 0 (counter-clockwise from trailing-edge Both the sin0 and 8 decay laws have a disconcertingly large velo-
region); /3 will also be referred to as the velocity decay angle, since city gradient very close to the trailing edge. These abrupt decelera-
the velocity begins its decay at 0 = /3. The decay angle for the lower tions would cause boundary layer separation. Unlike cos8 decay
surface is d. In the velocity decay region, the velocity is prescribed law airfoils, which have surfaces with a natural concavity leading
as being proportional to f(8) - f(/l), and the form of f(8) is refer- nicely into cupsed trailing edges, the sin0 and 8 decay law airfoils
red to as the type of velocity decay law. For example, if the pre- have surfaces which appear to be more natural for trailing edges
scribed velocity is taken as being proportional to c o d - cos/3, the with finite trailingedge angles. The trailing-edge compression
airfoil will be termed as one having a c o d decay law. Because the regions appear to be associated with the rather sudden concavity
cos8 decay law is a simple one having a reasonable behavior from enforced by the cusped trailing-edge requirement.
a boundary layer viewpoint, this law has been used for many of
the airfoils presented. In the decay region the boundary layer is tur- Other simple decay laws are available. A prescribed cos0/2 decay
bulent, and a turbulent boundary layer can withstand a steeper velo- produces results very similar to the cos0 decay (a slight increase
city decay earlier in its life time than it can when it is older and in concavity occurs). The cos 20 decay produces initial velocity ac-
thicker. The cosine decay starts out at 8 = /3 with its steepest velocity celerations if 0 is greater than 90 degrees. There is a corresponding
gradient, and the deceleration decreases as the trailing edge is local increase in airfoil thickness in the acceleration regions. One
approached.
can learn a lot beforehand about how the velocity distributions are I varied, the lift coefficients vary significantly. There is an appreciable
going to look by simply plotting and comparing the decay func- 1 change in thickness due to the variation in a1 - a2 (5 to 11
tions themselves (e.g., cos 8 - cos /3,181 - /3, suitably scaled so degrees). Figure 9 illustrates a case where both a1 and a2 are varied,
that they go through zero at ' 81 = /3 and one a t , 81 = 0). All other but in such a way that a1 - a2 remains constant at 5 degrees. The
airfoils presented have a cos I Bl decay law on the lower surface. thickness of each of these airfoils is essentially identical at 21 per-
cent. There is a significant change in the lift coefficients since a1
All of the airfoils so far have had a zero radius of curvature nose. has been varied from 7 to 11 degrees.
The following airfoils all have the leading-edge radius of curvature
terms incorporated into the solution and, therefore, have finite The airfoils thus far have all been "smooth" inasmuch as they
leading-edge radii. The leading-edge radius capability is illustrated were defined using many points, typically 250 to 300 points. A large
in Fig. 4. The decay law is the d81; again a2 = - a1 = 5 degrees number of points is desirable for final design purposes, but for other
qnd d = - /3 = 90 degrees. As stated, the leading-edge factor n purposes it is more efficient to work with far fewer points. The
spreads the leading-edge region over an angle of *90/n degrees. tlesign method allows this in that it generally provides a reasonable
In the figure the values of n are 3, 5. and 20, and it is clear that representation of an airfoil with only 20 to 30 points. These air-
significant changes in the nose radius have occurred. The largest foils, appropriately called "stick" models, are quite useful. Crude
radius is for n = 3; the results for n = 20 differ very little from stick models with 20 to 30 points are suitable for program debugg-
the zero radius case. Despite the large changes in the thickness of ing purposes. More refined stick models with 30 to 40 points are
the airfoils, the lift coefficients are comparable (a1 is the same for often adequate for parametric studies or for boundary layer analyses.
all). The increase in thickness allows for greater structural efficiency. A comparison of a stick model with 28 points and a smooth model
To keep the airfoil thickness approximately unchanged while the withi 292 points is shown in Fig. 10. The stick models appears to
nose radius is increased requires a reduction in the design angles be quite satisfactory. The comparison, however, is not strictly cor-
of attack (and consequent reduction in lift). If too large a nose radius rect because the present algorithm scales the airfoil so that its chord
is requested for given design angles of attack, unsatisfactory sur- length is one. A direct measure of the differencein the actual chord
faces develop. lengths can be made, however, by taking the ratio of the life coef-
ficients. Equation (24) shows that for equal design angles of attack
The effect of a finite nose radius is also quite apparent in the cI/c2 = CL2(CLI. For the two airfoils shown in Fig. 10, the lift
velocity distributions. The rooftop velocities no longer begin in an coefficients differed by only one percent. Thus the stick model was
abrupt manner. The rounding of the nose and the associated round- is quite satisfactory. Note that the number of points only affects
ing of the velocity distribution near the beginning of the rooftop the accuracy of the numerical integrations giving x and y; the
are important conditions at angles of attack above the upper-surface magnitude of the surface velocities do not depend on the intergra-
design angle of attack. The greater the nose radius the less the velo- tions. Experience to date has shown that occasionally a single sec-
city overshoot for angles above the design angle. This is illustrated tor in a stick model will require many more points than do the other
in Fig. 5 where the velocity distributions for the n = 3 and n = sectors before the airfoil will satisfactorily close.
20 airfoils in Fig. 4 are shown for angles 2 degrees above and below
the design angle of attack. The absence of the overshoot spike is With unsymmetrical airfoils the design repertory can include rear
quite noticeable for the n = 3 airfoil. loading. Two examples of rear loading are shown in Fig. 11. The
two airfoils have identical input parameters except for the rear-load-
Cambered Airfoils ing amplitude coefficients, which differ only in sign. The magnitude
of the coefficient if 0.5. The trailing-edge factor m is 6; thus the
With cambered airfoils, at, a2, /3, and d can be varied in- rear-loading is relieved within the angular region of 8 = 15
dependently, and different series of airfoils can be generated by vary- degrees in the circle plane. A higher factor would tend to square
ing one parameter at a time. The leadingedge radius factor n can off the tail velocity distribution because the reduction would be con-
also be varied. While it is instructive to go through such a parametric fined to an even smaller region.
variation, this is not really necessary. With the knowledge gained
from the symmetric airfoil study, ii is not difficult to understand The most obvious effect of rear loading on the shape of an air-
and anticipate the effects of various parameter variations. In fact, foil is the "wagging" of the tail of the airfoil. The tail is turned
only four figures are used to illustrate these effects. All of the air- down or up depending on whether the upper-surface velocity near
foils in these figures have a cos 0 decay on both surfaces. the trailing edge is greater than or less than the lower-surface velo-
city. Rear loading significantly affects the velocity distribution and
In Fig. 6, a2 = - a1 = 5 degrees (just as in Figs. 3, 4, and 5); thus would have a large effect on boundary layer development. The
however, /3 and d are varied independently. Examination of the upper lift coefficients did not change appreciably with the rear loading.
and lower surfaces gives no surprises in airfoil shapes or in the loca- The airfoil with no real loading was previously shown in fig. 8.
tions of the velocity decays points. Although a1 - a2 is constant,
the thicknesses of the airfoils differ. This shows (as did Fig. 1) that All of the basic features of the airfoil design method have now
large variations in the decay angles can appreciably affect the been illustrated. They have been illustrated using simple velocity
thickness. The lift coefficients of the airfoils are comparable; a1 decay laws, primarily the cos8 decay law. While the cos8 law is at-
is still dominating the lift determination. tractive, it is desirable to have greater flexibility and greater con-
trol over the decay. By using more complicated decay laws, it is
In Fig. 7 the lower-surface design angle of attack is varied while possible to achieve this. Figs. 12 through 15 show airfoils designed
everything else is held constant. For these airfoils a1 = 10 degrees, with decay laws offering progressively more control over the decay.
/3 = 85 degrees, d = 260 degrees, and n = 8. The parameter a2
varies in 4 degree increments from - 2 to 6 degrees. The airfoils Figure 12 shows three airfoils designed with an upper-surface velo-
exhibit a considerable change in thickness, which reflects that a1 city decay proportional to cos(8 - q) - cos(l3 - q). Values of q
- a2 varies from 4 to 12 degrees. The lift coefficients are within are 10,0, and -20 degrees, respectively, and it is shown that q of-
6 percent of each other, and agaip this is attributed to a1 being held fers some control over the velocity decay. With q = 0 this law
constant. The lift coefficients of these airfoils-in fact, of all of reduces to the c o d law. To illustrate the effect of q more clearly,
the cambered airfoils shown in the report-are within 6 percent of a1 has been adjusted from 11.6 to 12.1 degrees so that all three air-
the values predicted using the simple lift coefficient formula pre- foils have the same rooftop velocity (d must be varied slightly to
sented earlier. keev a constant initial decay station). With n = 10 degrees,
- - the
velocity gradient approache: zero at'the trailing edge; q = - 20
In Fig. 8, the lower-surface design angle of attack is held at 2 degrees
- vroduces
- a more linear-like decay. The distributionsn shown
degrees, while the upper-surface design angle of attack varies from are reasonable, but the velocity distributions can be made to go
7 to 13 degrees in increments of 3 degrees. The values of P, d, and through a minimum or a maximum between the decay point and
n are the same as in the previous figure. Because a1 is the quantity the trailing edge. In fact, q can be varied so that the velocity
distributioin curve at the velocity decay point ranges from shooting I j = 50 degrees for this airfoil, and although (1 + cos 8)/2 is 0.82,
nearly straight up to shooting nearly straight down. The upper and the slot is located near 0.73. For this type of airfoil the chordwise
lower surfaces can cross if a velocity distribution with a minimum position of the slot is usually overestimated by the preceding rela-
is produced. tion. No sink has been incorporated into the design of this airfoil;
therefore, no slot opening appears in the design. So far only a hint
Figure 13 shows an airfoil with a two-tiered velocity decay. The of the logarithmic spiral has shown up in the calculations of these
purpose of the two tiers is boundary layer control. One applica- airfoils (the spiral is too small).
tion, illustrated in a rather exaggerated fashion in the figure, is pat-
terned after airfoils designed by Wortmann.I4 The first tier pro- Interesting airfoil shapes occur when /3 and d are varied. With
vides a gentle adverse pressure gradient region to induce boundary small values of /3, such as /3 = 20 degree, airfoils can be generated
layer transition before the flow enters the second tier which con- having slot locations on the underneath side of an overhang region
tains the primary pressure recovery region. Each of the tiers has (the upper-surface trailing edge region is C-shaped). With /I= 80
a decay law similar to the type in Fig. 12. The chordwise location degree, typical slot locations fall within the range of x/c = 0.41
of the initial and final decay points may be controlled with I j 2 and to 0.45 (depending on 6); hence these airfoils have long thin tail
PI, respectively. The velocity magnitude at which the final decay regions (these airfoils have a thickness of 20 percent). Airfoils hav-
begins can be controlled with k,. Finally, the slope of the velocity ing velocity distributions in which no deceleration occurs anywhere
decay just downstream of each decay point can be controlled with other than at the slot can be designed. Such airfoils would appear
the q values. to be a good candidates for an all-laminar flow airfoils, but expe-
rimental evidence 9.10 indicates that the concave surfaces downs-
Nearly complete control of the velocity decay distribution can stream of the slot produces Gortler vorticies which induce bound-
be achieved by simply approximating the desired velocity distribu- ary layer transition.
tion with piecewise continuous linear velocity segments. The 8 decay
law is appropriate for each segment. Although this law does not Accuracy and Run Times
yield truly linear velocity segments, for small segments the velocities
are very close to being linear. Each segment requires a velocity and A measure of the accuracy of the computations can be obtained
an angular coordinate specification for each of its two end points. by seeing how well the airfoil closes upon itself at the trailing edge.
The velocities can conveniently be specified as a fraction of the roof- Typical values for the magnitude of the nondimensional distance
top velocity. The rooftop velocity is largely controlled by the design of the last point from the trailing edge are 1 x 10- 3 to 6 x 10- 4
anbles of attack and thus can be kept rather stable during the design for the cambered airfoils in the present effort. This accuracy is quite
iteration process. The problem can be set up such that the number acceptable. The accuracy, of course, is machine dependent, but the
of segment end points and their relative velocity amplitudes can results show that spline-fitting through the tabulated X and F func-
be input a run time. Considering the problem encountered earlier tions of Glauert will not significantly reduce the accuracy of the
with the 0 law near the trailing edge, a c o d decay for the trailing- computations.
edge segment seemed appropriate.
The accuracy has also been checked by comparing velocity
An example airfoil is shown in Fig. 14. Rather than develop an distributions with those obtained by running generated airfoils
airfoil tailored for optimum boundary layer control-such as air- through the potential-flow panel-method code contained in the Tran-
foil having a Stratford-like velocity rec0very,l5~16-for demonstra- sition Analysis Program System (TAPS) code.17 The results were
tion purposes it was easier (and more fun) to develop an airfoil with equally satisfactory.
a monogrammed potential-flow velocity distribution. Thus, the
lumpy airfoil shown is the author's MRS airfoil (if his initials had The primary factor affecting the run time is the number of points
been BXQ he would have opted for the Stratford airfoil). The air- specified to define the airfoil. Because the leading-edge and rear
foil has 17 linear segments and took only three attemps to make loading terms involve summations in the computation of X ( B ) , the
the initials legible (at least the M is clear). More segments require values of n and m in these terms also affect the run times. Sym-
only more input values and longer run times. For n segments there metrical airfoils having the equivalent of 300 points (only 150 are
are n inputs of angular coordinate, relative velocity amplitude, and used in the computation) with no leading-edge terms can be
number of integation points along each segment. For the airfoil generated in less than a minute. A 300-point cambered airfoil takes
shown, there are eight coordinate points along each linear segment. about seven minutes with n = m = 6. Lighthill mentioned that
for airfoils of high accuracy (use of seven-figure tables) the com-
The facetiousness of the airfoil in Fig. 14 may undermine the im- putation time was 4 to 5 days. Someday people will smile at the
portance of the linear segment contouring capability. To seriously present runtimes.
test the claim of nearly complete control of the velocity distribu-
tion, a velocity decay was specified in the airfoil plane as a func- Viscous Effects
tion of the chordwise poistion. The challenge was to generate an
airfoil with a reasonably matching velocity distribution. To add a
little realism to the problem, the decay was taken as an expotential
Viscous effects reduce the lift of an airfoil. The reduction is most
dramatic when boundary layer separation occurs, but a reduction -
one, but the form was rather arbitrarily specified as is present when the boundary layer remains attached. The interest
here is in the attached flow case. The boundary layer tends to reduce
the effective camber of an airfoil. Furthermore, the circulation is
affected by the wake thickness and curvature. To obtain an idea
of the magnitude of the lift coefficient decrement, a very brief survey
of measured lift coefficients was conducted.

Reasonable rooftop and trailing-edge velocities of 1.8 urn and 0.8 Only six airfoils were included in the survey and all had cusped
urn, respectively, were used. The rooftop was to end at 0 . 5 ~ .For trailing edges.'0*1s-22Except for the Glauert suction airfoiPO, all
these conditions, A = 1.8, B = - 1.143, and x, = 0.5. A com- of the airfoils were of recent vintage. The thicknesses ranged from
parison between the specified and resulting airfoil velocities is shown 13 to 28 percent for the conventional airfoils; the Glauert airfoil
in Fig. 15(a). Twenty-four segments were used, and the two curves was 31 percent. The Mach numbers were low (M 5 0.1) and the
are nearly coincident. The airfoil and complete velocity distribu- chord Reynolds numbers were moderately high (Re, 2 106). The
tion are shown in Fig. 15(b). The procedure took many more than experimental lift coefficients were taken from the published curves
three attemps; it could be automated. of lift coefficient vs. angle of attack. The points selected were well
away from the stall condition. The potential flow results were ob-
Boundary Laver Control Airfoils tained using the published airfoil coordinates and the panel-method
code contained in TAPS.17. The code models the Kutta condition
A Glauert-type suction airfoil is shown in Fig. 16. The slot is at by equating the velocities at the center of the last panels on the up-
per and lower surfaces. It is thus important that the two panels be
small and of the same size. T o satisfy this requirement, and to en-
sure that the panel size increased in a smooth manner with distance
from the trailing edge, new coordinate points were established in
the trailing-edge region by fitting the existing points with cubic
splines.

The comparisons between the observed lift coefficients and those


obtained from potential flow theory are shown in Fig. 17. The or-
dinate is the experimental lift coefficient and the abscissa is the
potential flow value; along the 45-degree line shown in the figure
the two values are equal. The other line on the figure is a least-
squares straight line fitted through the data from five of the six air-
foils, and the scatter about the line is not large. The slope of the
line is 0.855, implying that the actual life coefficients are about 86
percent of the potential flow values. Before accepting such a crude
rule-of-thumb estimate more data need to be examined.

The data from the Kennedy and Marsden airfoil22 were not
included in the least-squares fit because the data do not follow the
trend shown by the other airfoils. This airfoil is a very thick airfoil
and is significantly rear loaded. With this type of airfoil wake cur-
vature and displacement thickness effects may become important.

The drag of the airfoil can be estimated using the Squire-Young


formula. This formula requires the boundary layer properties at
the trailing edge, and a boundary layer analysis can conveniently
follow the airfoil inverse design program. The design code was writ-
ten with the option of storing the coordinates and velocities for later
use by a boundary layer codes. An intermediate program was writ-
ten to locate the stagnation point and determine the velocity-surface
distance information. The program could also be used to delete any
of the points.
Concluding Remarks 4. Glauert, M. B., "The Application of the Exact method of
Aerofoil Design," A.R.C. R&M 2683, pp 733-797 (Oct 1947).
Lighthill's 1945 exact inverse conformal-mapping method for the
incompressible potential flow design of two-dimensional airfoils has 5. Eppler, Richard and Dan M. Somers, "A Computer Program
been reviewed and numerically implemented on a desktop computer. For the Design and Analysis of Low-Speed Airfoils," NASA TM
With this method the shape of the airfoil is determined in terms 80210 (1980).
of the velocity distribution prescribed as a function of the angular
coordinate around the circle into which the airfoil may be trans- 6. Santo, Junzo, "An Exact Two-Dimensional Incompressible
formed. M. B. Glauert presented may useful analytic results in 1947 Potential Flow Theory of Aerofoil Design with Specified Velocity
which enable the Lighthill method to be numerically implemented Distributions," Transactions of the Japan Society for Aeronautical
in a straightforward manner, and in the present effort considerable and Space Sciences, Vol. 9, No. 14, pp 11-18 (1%6).
use has been made of Glauert's work.
7. Robinson, A , and J. A. Laurmann, Wing Theory, Cambridge
Lighthill's method was programmed using the BASIC language University Press (1950).
on a desktop computer having 55 kilobytes of working memory.
The numerical implementation has been discussed in some detail. 8. Thwaites, Bryan, ed., Incompressible Aerodynamics, Clarendon
Use has been made of some of Glauert's tables by fitting them with Press, Oxford (1960).
splines; however, a numerical procedure is presented for genera-
tion of the tables. The spline routine, a nonlinear system solver using 9. Richards, E. J., Walker, W. S. and C. R. Taylor, "Wind-tunnel
Newton's method, and Simpson's integration method comprise the Tests on a 30 per cent. Suction Wing," A.R.C. R&M 2149, pp
basic numerical repertory of the inverse design method. 465-486 (July 1945).

The primary purpose of the paper is to demonstrate some of the 10. Glauert, M. B., Walker, W. S., Raymer, W. G. and N. Gregory,
types of airfoils which may be generated using the method on a small "Wind Tunnel Tests on a Thick Suction Aerofoil with a Single
computer. Numerous examples are presented of airfoils and their Slot," A.R C. R&M 2646, pp 261-275 (Oct 1948).
velocity distributions. The designs include both symmetric and
cambered airfoils encompassing a wide range of life coefficients and 11. Spiegel, Murray R., Mathematical Handbook of Formulas and
thicknesses. The primary factors affecting lift and thickness are iden- Tables, Schaum's Outline Series, McGraw-Hill Book Co. (1968).
tified. Glauert's methods of controlling the leading edge radius and
applying rear loading are also demonstrated (the latter technique 12. Beyer, William H., ed., CRC Standard Mathematical Tables,
has been corrected and generalized somewhat). Examples are given 25th Ed., CRC Press Inc. (1974).
of various types of velocity decays over the rear portion of the air-
foils. Run times for 300 point airfoils varied from under one minute 13. Atkinson, Kendall E., An Introduction to Numerical analysis,
for symmetric airfoils to seven minutes for cambered airfoils with John Wiley & Sons (1978).
rear loading.
14. Wortmann,, F. X., "Progress in the Design of Low Drag
Although seemingly restricted by the use of simple velocity Aerofoils," in Boundary Layer and Flow Control, Vol. 2, Ed. by
description terms in the circle plane, very good control over the G. V. Lachmann, Pergamon Press, pp 748-770 (1961).
velocity in the airfoil plane is possible. The capability was tested
and demonstrated by specifying a velocity decay as a function of 15. Stratford, B. S., "The Prediction of Separation of the Turbulent
the chordwise distance in the airfoil plane and then designing and Boundary Layer," J. of Fluid Mech., Vol. 5, pp 1-16 (1959).
airfoil having an upper-surface velocity distribution which matched
the specified distribution. The design was completed in a trial-and- 16. Smith, A. M. O., "High-Lift Aerodynamics," J. Aircraft, Vol.
error fashion by approximating the desired velocity decay with 24 12, No. 6, pp 501-530 (June 1975).
piecewise-continuous linear velocity segments. The program was set
up so that the number of segment points, their location in the cir- 17. Gentry, Arvel E., "The Transition Analysis Program System,"
cle plane, and their relative velocity magnitudes could be input at Vol. 1, Rept. No. MDCJ7255/01, McDonnell Douglas Astronautics
run time. Co. (June 1976).

Lighthill's method has been shown to be quite suitable for use 18. Bingham, G. J . and A. W. Chen, "Low-Speed Aerodynamic
on a small computer. In the interactive user-friendly environment Characteristics of an Airfoil Optimized for Maximum Lift Coeffi-
of the desktop computer the method is very flexible, enlightening, cient," NASA TN D-7071 (Dec 1972).
and powerful. The method has been and is being used to design
single-element airfoils in an R&D environment; it should be emi- 19. Somers, D. M., "Experimental and Theoretical Low-Speed
nently suitable in an academic environment where it could be used Aerodynamic Characteristics of a Wortman Airfoil as Manufac-
as a teaching and research aid for potential flow and boundary layer tured on a Fiberglass Sailplane," NASA TN D-8324 (Feb 1977).
studies.

References 20. Somers, D. M., "Design and Experimental Results for a Natural
Laminar Flow Airfoil for General Aviation Applications," NASA
1. Lighthill; M. J., "A New Method of Two-dimensional TP-1861 (June 1981).
Aerodynamic Design," A. R. C. R&M 21 12, pp 105-157 (Apr 1945).
21. Somers, D. M., "Design and Experimental Results for a Flap-
2. Lighthill, M. J., "A Theoretical Discussion of Wings with ped Natural-Laminar-Flow Airfoil for General Aviation Applica-
Leading-Edge Suction," A. R. C. R&M 2162, pp 556-564 (May tions," NASA TP-1865 (June 1981).
1945).
22. Kennedy, J . L. and D. J. Marsden, "The Development of High
3. Glauert, M. B., "The Design of Suction Aerofoils with a Very Lift, Single-Component Airfoil Sections," Aeronautical Quarterly,
Large CL-Range," A.R.C. R&M 21 11, pp408-415 (Nov 1945). pp 343-359 (Feb 1979).
0.2 f(8) = Sin 8

0.1 0.1

ylc 0 ylc 0

-0.1 -0.1

"0 0.5 1 .O -0 0.5 1 .O


xlc xlc
Fig. 1 - Effect of Varying Velocity Decay Angle Fig. 3 - Effect of Type of Velocity Decay Law, f(8) - f(0)

0.2

0.1

ylc 0

-0.1

0 I I

0 0.5 1.O 0.5 1.O


xlc xlc
Fig. 2 - Effect of Varying Upper-Surface Design Angle of Fig. 4 - Effect of Varying Leading-Edge Factor
Attack
nL. a , = - a 2 = 5 O , 8 = - B = 9 0 , C o s 8 - C o s B DECAY

-
0 I 8 8 . P 1

n = 8 , a, = 1 0 , B = 8 5 0 , 8=260,
Cos 8 - Cos B DECAY
0
OO 0.5 1.O 0 0.5 1 .O
xlc xlc
Fig. 5 - Velocity Distributions for Angles of Attack 2 Degrees Fig. 7 - Effect of Varying Lower-Surface Design Angle of
Above and Below Design Angle of Attack Attack, No-lift Angles; -2.85, -3.69, -4.61 Deg.

0.2
/B = 90
0.1

ylc 0 ylc 0
0.5
-1 .o

I I
I
ulu, .

n = 8, a , = 2O, B = as0, 8 = 260, C o s 8 - C o s B DECAY

0.5 0.5 1 .O
xlc xlc
Fig. 6 - Cambered Airfoils With Equal Magnitude Design Fig. 8 - Effect of Varying Upper-Surface Design Angle of
Angles of Attack, No-lift Angles; 3.39, 4.53 Deg. Attack, No-lift Angles; -4.56; -3.69, -2.79 Deg.
b =-0.5

ylc 0 1
I ylc 0
0.5

, I I
I

n = 8, 8 = 8s0, 8 = 260, COS8 - COS8 DECAY


HY/ COS8 - COS8 DECAY
I I

0.5 1.O 0.5


xlc xlc
Fig. 9 - Effect of Varying a , and a, While Keeping a , - a, Fig. 1 1 - Effect of Rear Loading, No-lift Angles; -0.37,
Constant, No-Lift Angles; -4.95, -2.79 Deg. -8.75 Deg.

ylc 0 j
lo

0
0.5 1.O
xlc xlc

Fig. 10 - Comparison of Stick Model (28 Pts) With Smooth Fig. 12 - Illustration of Cos ( 9 - q) - Cos (13 - r ~ Decay,
)
Model (292 Pts) No-lift Angles; -4.07, -3.35, -2.67 Deg.
1.21. \ ,ulu, FROM AIRFOIL

n = 6 , a , = 10,a2=50,fi1=800,@2=1000,'l)l=loO,
I,=-20
kl = - 0.85 xlc
1 1

0.5 0.1 Fig. 15(a) - Comparison of Specified and Airfoil Upper-Surface


xlc Velocities
Fig. 13 - Illustration of Two-Tiered Decay, No-lift Angle,
-3.38 Deg.

0.2 -

MONOGRAMMED POTENTIAL-FLOW VELOCITY


-0.2L DISTRIBUTION AIRFOIL
2.0 r

n=6,a1=12O, a2=1.2S0,@= , 8=280


I
T"; 1

0.5 1.O
xlc
Fig. 15(b) - Airfoil Having Specified Velocity Decay
xlc
Fig. 14 - Illustration of Custom Velocity Contouring With Fig. 15 - Test Case Airfoil Generated With Linear Velocity
"Linear" Segments Segments in Upper-surface Velocity Decay
0' 0

5 1.5
W
Ref. Re, x
1 .o 0 18 9.0
0 10 0.96
0 19 1.5
A 20 1.0
0.5 a 21 3.0
0 22 1.o
I I I I I I I
0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0
POT. FLOW CL
Fig. 17 - Comparison of Potential Flow and Experimental Lift
Coefficients for a Variety of Airfoils
xlc
Fig. 16 - Glauert (Griffith) Suction Airfoil, No-lift Angle,
-3.75 Deg.

You might also like