Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

SAE TECHNICAL
SERIES961317
PAPER

Math Modeling of Airfoil Geometry

J. Philip Barnes
Northrop Grumman Corp.

Aerospace AtlanticDayton,
Conference
Ohio
May 22-23, 1996
400 Commonwealth Drive, Warrendale, PA 15096-0001 U.S.A. Tel:(412)776-4841 Fax:(412)776-5760
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

The appearance ofthe ISSN code at the bottom of this page indicates SAE's consent
that copies of the paper may be made for personal or internal use of specific clients.
This consent is given on the condition however, that the copier pay a $7.00 per article
copy fee through the Copyright Clearance Center, Inc. Operations Center, 222
Rosewood Drive, Danvers, MA 01923 for copying beyond that permitted by Sections
107 or 108 of U.S. Copyright Law. This consent does not extend to other kinds of
copying such as copying for general distribution, for advertising or promotional
purposes, for creating new collective works, or for resale.

SAE routinely stocks printed papers for a period of three years following date of
publication. Direct your orders to SAE Customer Sales and Satisfaction
Department.
Quantity reprint rates can be obtained from the Customer Sales and Satisfaction
Department.

To request permission to reprint a technical paper or permission to use copyrighted

SAE publications in other works, contact the SAE Publications Group.

No part of this publication may be reproduced in any form, in an electronic retrieval

system or otherwise, without the prior written permission of the publisher.
ISSN 0148-7191
Copyright 1996 Society of Automotive Engineers, Inc.
Positions and opinions advanced in this paper are those of the author(s) and not
necessarily those of SAE. The author is solely responsible for the content of the
paper. A process is available by which discussions will be printed with the paper if
it is published in SAE Transactions. For permission to publish this paper in full or in
part, contact the SAE Publications Group.
Persons wishing to submit papers to be considered for presentation or publication
through SAE should send the manuscript or a 300 word abstract of a proposed
manuscript to: Secretary, Engineering Meetings Board, SAE.
Printed in USA96-0049
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

961317

Math Modeling of Airfoil Geometry

J. Philip Barnes
Northrop Grumman Corp.

ABSTRACT
New and powerful methods of characterizing existing and available today, we can promote maximum laminar flow,
new airfoil geometries with mathematical equations are thereby also promoting maximum aerodynamic efficiency.
presented. The methods are applicable to a wide range of
airfoil shapes representing traditional, cusped, reflexed, In addition to the aerodynamic advantages, airfoil
flat-bottom, laminar, transonic, and supersonic designs. geometry math modeling provides an efficient interface
With the emphasis on low-speed airfoils, several existing with the wide range of software tools related to CAD
airfoils are first closely matched with the math-modeling geometry definition, computational fluid dynamics,
methods. Then, to support the design of new airfoil structural analysis, and manufacturing. For example, by
geometries, a new interpretation of Theodorsen's potential extending the math modeling into three dimensions,
flow method is outlined for the calculation and presenta numerical machining of wing molds can be efficiently
tion of surface velocity in inviscid flow. Also, a vector programmed for prescribed variations in airfoil shape,
approach is introduced for the calculation of pitching chord, and twist across the span. In another example,
moment. Finally, new math-modeled airfoils are proposed nodes for wing panel finite element analysis are readily
for conventional and unique aircraft configurations. defined for a math-modeled surface.

Math modeling of airfoil geometry is not new. However, a

INTRODUCTION method with independent control of all major geometric
parameters (Figure 1) appears to have been previously
"The velocity distribution on the surface ofa profile is unavailable. Relative to previous math-modeling methods,
primarily determined by the distribution of curvature. the techniques herein offer control over camber distribu
...The questions arise, what effects discontinuities tion, leading edge radius, forward thickness buildup,
have on the velocity and pressure distributions, and maximum thickness, thickness location, cusp effects, and
how the resultant changes affect the boundary layer... finite trailing edge thickness. Moreover, largely indepen
The answers are now clear... A jump in the curvature dent control of each geometry feature is demonstrated.
leads to a point of inflexion in the velocity
distribution, and even a kink in the curvature
manifests itself in a slight concavity in the pressure
distribution. "
F. W. Riegels
In his book, "Aerofoil Sections," Riegels (1) makes it clear
that an airfoil which appears smooth to the eye may not
appear smooth to the air flowing over it. Recent studies
by Selig (2) have observed significant "zigzag"
discontinuities in the calculated surface velocity
distributions when using tabulated airfoil coordinates.
Math modeling of airfoil geometry, the subject of this
paper, eliminates discontinuities in the specified
geometry. By combining accurate geometry definition
with the accurate manufacturing tolerances which are
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

To enhance aerodynamic efficiency, the proposed In terms of math modeling, only the thickness and
equations are continuous from the leading to the trailing camber can be controlled in (EQs 1.1 and 1.2). However,
edge (except for the second derivative on the lower the extraordinary property of the Zhukovsky airfoil is its
surface of flat-bottomed airfoils). The equations are easy application to surface velocity calculation. Theodorsen
to use with scientific software which has integrated (5, 6) showed that an airfoil of arbitrary thickness and
spreadsheet, programming, and graphics capabilities. camber can be transformed into an equivalent Zhukovsky
airfoil which, as shown by Zhukovsky, can be transformed
After briefly reviewing previous math-modeling methods into an equivalent rotating cylinder to obtain the surface
(Section 1.0), the new methods are outlined in Section 2.0 velocity distribution with potential flow analysis. Theodor
and applied to closely match the existing geometry of a sen's calculations are outlined and applied later herein.
range of contemporary low speed airfoils (Section 3.0).
Then, fundamental low-speed airfoil design tools are
covered. Building upon the work of Abbott and von 1.2 NACA Airfoil Math Model
Doenhoff, a previously-unpublished relationship of airfoil Abbott and von Doenhoff (7) describe the math model
pitching moment coefficient to camber distribution is of the NACA 4- and 5-Series airfoils. The thickness
presented (Section 4.0). Then, Theodorsen's classic coordinate (ît=Zt/C) is added normal to the meanline
potential flow velocity calculation is outlined and applied coordinate (îc=Zc/C). The thickness, defined by NACA in
to show the ultra-smooth velocity profiles obtained with (EQ. 1.3), adds a 4th-order polynomial to a square root
airfoil geometry math modeling (Section 5.0). term for the leading edge. The maximum thickness,
designated (t), is easily controlled, but a change in the
Finally, having summarized the basic aerodynamic and maximum thickness location and/or the leading edge
geometric design tools, the methods herein are applied radius would require a new set of polynomial coefficients.
toward the design of new airfoils for conventional, tailless,
and canard-configured aircraft (Section 6.0). The camber for the NACA 5-Series is reproduced as
EQ. (1.4), where (m) and (k) control the maximum camber
and its location. Two-piece characterizations are used,
1.0 HISTORICAL NOTES one forward and the other aft of the location of maximum
camber. A well-known 5-series airfoil is the NACA 23015,
1.1 The Zhukovsky Airfoil shown in Figure 1.2.
One of the earliest math-modeled airfoils (Figure 1.1) was
introduced by Zhukovsky in 1910. As shown by Glauert
(3) the thickness shape of the Zhukovsky airfoil can be
expressed in parametric form with a polar angle (?),
counter-clockwise from the trailing edge, and a thickness
parameter (a), where the maximum dimensionless
thickness is approximately (1.3a). Houghton and
Carruthers (4) add a camber (mean line) parameter (b),
where the maximum dimensionless camber is (b/2).

The NACA method of adding thickness normal to the

mean line maintains constant upper and lower radii of
EQs (1.1) and (1.2) below, from References 3 and 4, have curvature at the nose. However, the upper portion of the
been simplified, non-dimensionalized, and converted to nose extends "microscopically" to the left of the z-axis,
modern sign convention. thereby rotating the effective chord line to a point where
the true leading edge lies above the x-z origin. From a
"geometry" point of view, this condition is not convenient.
From an "aerodynamic" point of view, this feature may be
important for aerodynamic efficiency. Nevertheless,
contemporary airfoils regularly exhibit different upper and
lower curvatures at the nose. For all of the new methods
herein, thickness is added vertically to the camber.
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

2.0 NEW AIRFOIL GEOMETRY MATH MODELS

As with the NACA method, the methods herein generally

add two equations, one for camber and the other for
thickness. However, a new building block for both
thickness and camber geometry is introduced -- the
modified half sine wave.

2.1 Modified Sine Wave for Thickness

The sinusoidal function sin(è) has a maximum value of
unity, and over the range from (0 to ð), has the appro
priate boundary conditions (z=0) for basic modeling of The skew transformation is illustrated in Figure 2.3 and
the upper half of an airfoil thickness distribution. Defining the resulting half-thickness shape is shown in Figure 2.4.
the airfoil maximum thickness as (t = t / c ), the half sine Note that the leading edge radius remains essentially
wave can be implemented by applying the factor constant as the thickness location is moved aft to 43 %.
(+/-) (t /2) and then replacing the argument (è) with the
argument (Xð), where the term (X) is the dimensionless
chordwise coordinate (x/c) from the leading edge.
To obtain the desired thickness shape, the X-coordinate
is transformed or "skewed" with dedicated operations for
largely independent control of the leading edge radius,
maximum thickness location, cusp depth, and cusp
extent. The skewed coordinate is then designated (Xs ).
Therefore, the local half-thickness takes the
form of (EQ 2.1):

The first, and most basic, coordinate transform is accom

plished with a leading edge exponent (?). Here, Xs = X?,
as shown in Figure 2.1 (with ?=0.6). This "square root"
operation obtains an appropriate leading edge shape and
shifts the sine wave maximum forward, thereby (initially)
locating the maximum thickness at about 30 % from the
leading edge as shown in Figure 2.2. The last transformation adds a cusp. An auxiliary
"inchworm" half sine wave of amplitude (ä1) and exponent
(ä2) is added to reduce the thickness of the aft region. The
depth of the cusp is controlled by (ä1) while the extent of
the cusp is controlled by (ä2). The "inchworm" sine wave,
illustrated in Figure 2.5, is represented by the last term of
(EQ 2.3) which defines the final configuration of the
transformed X-coordinate.

By varying the leading-edge exponent (?), typically

between 0.45 and 0.55, the leading edge shape can be
controlled. An exponent of unity obtains a sharp point, as
exhibited by the dashed curve in Figure 2.2.
Next, by adding an auxiliary half sine wave of amplitude
(µ) to "skew" the X-coordinate, the maximum thickness
location can be shifted forward or aft. The transformed
X-coordinate then takes the form of (EQ 2.2):
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

The cusp-related coordinate transformation using the math model for a skewed sine wave camber line. The
inchworm sine wave is illustrated in Figure 2.6 and its skewing effect is shown in Figure 2.9. A negative value for
effect on the half thickness shape is shown in Figure 2.7. (µc) would skew the maximum aft of 50 % chord.
Note that the location of maximum thickness moves
forward slightly during this operation. However, the
thickness shift can be reversed by updating (µ). Finally,
taking a mirror ¡mage and assuming 15 % thickness, we
obtain Figure 2.8, which is the final result of this example.

2.3 Varabola Math Model

For dimensionless x- and z-coordinates (X, Z/Zmax) both
ranging between zero and unity, a parabola would take
the form Z/Zmax= X2. For airfoil math modeling, we need
the flexibility to vary the exponent. Hence we introduce the
name "varabola," where the exponent is designated (v)
and has any value greater than unity. Increasing the
exponent affects the curve shape as seen in Figure 2.10.

Summarizing thus far, EQs (2.1) and (2.3) constitute the

modified sine wave thickness math model. The geometry
is controlled by the parameters ( t, ë, µ, ä1, and ä2 ). The
first term on the right-hand-side of EQ (2.3) controls the The simplest application of the varabola is in modeling
leading edge shape. The second term controls the the finite thickness at the trailing edge. No airfoil can, or
location of maximum thickness. The last term controls the should, be manufactured with zero edge thickness.
cusp depth and extent. Consequently, an edge thickness (te) should be included
when modeling new airfoils. One method is to add a
In the next sub-section, applicability of the skewed sine varabolic term to the thickness of (EQ 2.1) as indicated by
wave toward math modeling of the camber line is the last term of (EQ. 2.6) below. By selecting a large
demonstrated. Then, the camber and thickness are exponent, such as 4, the added term will have very little
added to obtain the final airfoil shape. However, before impact on the airfoil shape upstream.
leaving the current topic, we note that the upper and lower
surfaces can be modeled with separate "half-thickness"
models. In this case, a dedicated camber math-model is
not used. However, the resultant formula for camber can
be determined by taking half the sum of the equations Another important application of the varabola is camber
representing the upper and lower surface math models. line modeling. Assuming for the moment that the
maximum camber (K) occurs at 50 % chord, an inverted
parabolic arc with its vertex at ( X, î = 0.5, K ) would take
2.2 Skewed Sine Wave for Airfoil Camber the form of (EQ 2.7), where the exponent (v) has a value
Designating (K) as the maximum camber and (µc) as the of (2.0). The absolute value makes the characterization
forward skew parameter, (EQs. 2.4 and 2.5) constitute the applicable at any chordwise position.
4
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

In another example of a function product, an airfoil half

thickness shape (Figure 2.13) can be similarly derived by
multiplying the functions Xë and (1-Xv), where the
Given a larger exponent, say 4 to 6, the inverted varabola parameter (ë, typically 0.5) controls the leading edge.
takes on a plateau shape, characteristic of laminar airfoil Both (ë) and the varabola exponent (v) affect the
camber lines. The X-coordinate can be skewed if neces maximum thickness and its location.
sary to shift the location of the maximum camber. This is
accomplished with (EQ 2.8), which uses the same
camber skew parameter which was previously designated
as (µc). An inverted varabola camber line, with and
without skew, is illustrated in Figure 2.11.

Of the three new methods for thickness modeling herein,

the function product approach is uniquely suited to
laminar airfoils. Such airfoils are characterized by
maximum thickness aft of about 45 % chord. With a
coordinate transformation subsequently to be described,
2.4 Function Product Method the function product approach avoids the undesirable
Working within a graphical "box" where any two functions narrow forebody condition illustrated in Figure 2.14.
f1(X) and f2(X) extend from corner to corner, a wide range
of functions can be multiplied together to obtain airfoil
camber or half-thickness lines. Imagine that two flexible
bamboo sticks originally arranged in an "X-shape" are
constrained by sleeves which are free to rotate at the
corners of the box (Figures 2.12 and 2.13). Either function
can be inverted by using 1-f(x) and/or reflected by using
f(1-x) to achieve the desired effect.
Gale (8) presents Birnbaum's equation for an airfoil
camber line which was used by the Horten brothers for To buildup the thickness in the region of 25 % chord, the
their tailless designs of the 1940s. Here, the product of X-coordinate will be transformed by a "square root"
(X) and (1-X)3 yields an inchworm shape with a maxi operation and then distorted by adding a bell-shaped sine
mum of 0.1055 at X=0.25, as il ustrated in Figure
2.12. wave. This modified sine wave (Ref. 8) can have any
exponent greater than unity, but 2.0 is used in Figure 2.15.

In modifying the pre-transformed coordinate (Xë), the

addition of a bell-shaped sine wave adds a smooth
"bump" peaking at (X0.5=0.5) and also at (X=0.25). This
has the desired effect of building up the forward
thickness.
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

The two necessary transformations and their integration

with the function product technique are outlined in
(EQs 2.9-2.11). The parameter (ó) represents the ampli
tude of the bell-shaped sine wave and the parameter (A)
controls the maximum thickness. The final result, in
Figure 2.16, shows a significant improvement when
compared back to Figure 2.14. Note that an increase in
the varabola exponent (v) was required to maintain the 2.6 Inchworm and Reflexed Camber Lines
maximum thickness location at 58 % during this
operation. Also, a cusp could have been added in this An "inchworm" camber shape can be obtained by taking
example, but it was omitted for clarity. the reflected coordinate (1-X) to the power (w) in the
modified sine wave of (EQ 2.14). This is illustrated in
Figure 2.18 (Return to Figure 2.5 for the reflected version).
The derivative (dî/dX) is zero at the airfoil trailing edge.
Also, the location (XK) of the maximum camber (K) is
related to the exponent (w) as shown by (EQ 2.15). The
Birnbaum quadratic (return to Figure 2.12) obtains a
similar result, but (EQs 2.14 and 2.15) offer precise
control of the maximum and its location.

2.5 Flat-bottom Math Model

Modeling the geometry of a flat-bottom airfoil can be
accomplished with separate upper and lower equations.
Consequently, dedicated thickness and camber curves
are not defined, although they can be calculated. A Reflex (rear camber falling below the chord line) can be
discontinuity in the second derivative appears unavoid provided in many ways, one of which is a cubic polyno
able on the lower surface at the junction of the lower mial (EQ 2.16). The coefficients (a,b,c) are obtained by
leading edge and flat-bottom segments. (EQs 2.17-2.19) given the maximum camber point
(XK,K), where (dî/dX=0), and given the third boundary
A quarter ellipse of height (?o) and width (Xo) can be used condition represented by (î=0 at X=1).
to model the lower leading edge up to the junction with
the flat bottom. A function product with (1-X) is then
applied to ensure continuity of the first derivative at the
junction. Suitably modified to account for the lower half of
the finite trailing edge thickness, the lower-surface math
model is described by (EQs 2.12-2.13) and is illustrated
for a representative flat-bottom airfoil in Figure 2.17.
Another approach is to transform the X-coordinate by
adding an inchworm sine wave of sufficient amplitude to
increase the argument of the sine function beyond ð,
thereby obtaining negative î-coordinates in the aft region.
The transformation and reflexed camber math model are
shown in (EQs 2.20 and 2.21). This approach provides
added flexibility in controlling the camber, whereby the
maximum camber (K) is fixed while the shape is varied by
the choices of (ä1, ä2, and ä3). The last term of (EQ 2.20)
is needed to avoid a dip near the leading edge. A repre
sentative reflexed camberline is presented in Figure 2.19.
6
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

Figure 2.21 shows the airfoil thickness corresponding to

Figure 2.20 and identifies all input and output parameters.
For this example, a Newton-Raphson scheme is used
with a scientific spreadsheet. The guess for (m), along
with the resulting dimensionless error, E= [t / (2î r)] -1,
and the corresponding error derivative (dE/dm) are placed
in a row. Then, for the calculations of each new row, the
previous row is interrogated. Approximately six iterations
are required, depending on the initial guess. As with the
modified sine wave method, the varinomial is best applied
when the maximum thickness is forward of 45 % chord.

2.7 Varinomial Thickness Math Model

The math models discussed so far involve non-iterative
calculations, but typically require several design
iterations. An alternative approach, requiring iterative
computation and some risk, is to directly specify the
leading edge exponent (ë), maximum thickness (t),
thickness location (Xt), trailing edge included half angle
(e), and trailing edge thickness (te). Then, an equation is
solved to meet these specifications. One such formula
proposed here is designated a "varinomial."
If we plot a typical airfoil half-thickness versus the square
root of X, we find a shape similar to Figure 2.20. A very
high-order polynomial (perhaps up to 10) in î vs. Xë
would be required to accommodate the aft region.
However, the lower-order terms would introduce
unwanted oscillations in the forward thickness. A three
term "varinomial" as defined by EQ. 2.22 avoids this
problem.
3.0 APPLICATION TO EXISTING AIRFOILS
In this section, existing airfoils will be modeled as closely
as possible with the techniques of Section 2.0. The
original tabulated coordinates will be shown for
comparison with the math models.
The process of math modeling an existing airfoil takes
several attempts. First, thickness is isolated. Then the
camber is separately modeled. With practice, most airfoils
can be closely matched within ten or fifteen iterations,
including both thickness and camber. In some cases,
only one iteration on thickness is sufficient when using the
varinomial approach. An essential tool is a scientific
spreadsheet software (9) which has integrated
As in a cubic polynomial, the third-term exponent (n) is programming and plotting capabilities so that one or
50 % greater than the second-term exponent (m). The more geometry parameters can be changed and the
problem is to solve for the coefficients (A,B,C) and effect on airfoil shape observed within seconds.
exponent (m) from the stated boundary conditions. The
equations are non-linear and the solution is iterative, For our first example we take Fauvel's F-2 reflexed airfoil,
whereby (m) is estimated, and C, B, and A are calculated which was designed for the AV-36 tailless sailplane of
from (EQs 2.23 - 2.25). Then, a resultant half-thickness Figure 3.1. As shown in Figure 3.2, (EQs 2.1-2.3) are used
(î r) is found by applying (EQ 2.22) at (X=Xt). With a to model the thickness. Remarkably, no skew or cusp
suitable iterative scheme, (m) is varied until the true half modifiers (µ, d1) are required. The reflexed camber of
thickness (t / 2) matches (î r). Convergence is reliable (EQs 2.20 - 2.21) is included in Figure 3.3. The reflex is
below a critical (e), typically 12 °, depending on (t and Xt). required for pitch trim, as discussed in Section 4.0.
7
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

An airfoil suitable for a low speed aircraft with a conven

tional tail, such as the Discus sailplane of Figure 3.4, is
the FX-67. The thickness is modeled in Figure 3.5 and the
thickness with camber in Figure 3.6. A varabola model
has been used for the camber.

For both lifting surfaces of a general aviation canard

configured aircraft such as that of Figure 3.7 (Ref. 14),
or for the swept tailless SB-13 sailplane of Figure 3.8, an
airfoil such as the Quabeck HQ-36N is applicable due to
its zero pitching moment. The thickness and cubic mean
line math models for the HQ-36N are shown in Figure 3.9.
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

4.0 AERODYNAMIC EFFECTS OF CAMBER (µ0 and ßo by their terminology) which operate on the
Before designing new airfoils, the fundamental effects of camber shape. The related formulas are reproduced as
the camber line on the airfoil pitching moment coefficient (EQs 4.1 - 4.3) using the nomenclature of this paper.
(cm) and zero-lift incidence (ß) should be understood. (EQ 4.4) combines these to directly relate (cm) to the
This is particularly important for the pitch trim of unique camber distribution of (î=Z/C) versus (x=x/c). Although
configurations. fundamental, this last equation appears to have been
previously unpublished.
Figure 4.1 defines the sign convention used herein for the
various aerodynamic parameters, including the angle of
attack (a) and lift coefficient (c1). The sign convention for
the zero-lift angle varies in the literature. The convention
used here is perhaps the most straight forward, whereby
the lift coefficient is, in theory, simply 2ð(a + ß). For most
airfoils, then, (ß) is positive.

Combining these, we have:

4.1Pitch Trim Considerations

Whereas the airfoil section pitching moment coefficient is In (EQ 4.4), the combination of terms in [brackets]
designated (cm), the pitching moment coefficient for the represents a weighting factor on the local camber (î)
wing is designated (cM). The latter is affected by the in terms of its contribution to (cm). Figure 4.2 plots this
spanwise distribution of (cm), chord, twist, and sweep. weighting factor, designated [F 1] versus chordwise
An aircraft with conventional empennage trims (cM) with a position. Note that camber near the trailing edge
lifting force on the horizontal tail. Typically, cM<0, where dominates (cm). To calculate (cm) with EQ 4.4, the
by the tail develops negative lift for a typical statically product [F1] î is numerically integrated while avoiding
stable center of gravity location. However, the tail can trim the singularities which occur at (X=0 and X=1).
a wide range of positive or negative (cM).
In a statically-stable canard (wing + foreplane) configu
ration, a conventional airfoil (cm<0) is not recommended
for the wing. Even an aggressive combination of wing
sweep and twist may not be sufficient to prevent canard
overload and wing "underload," whereby both perfor
mance and flight safety are adversely affected. The wing
(lifting surface #2) of a canard configuration should have
cM2 at least zero, or even positive, to improve lift sharing.
A further constraint on (cM) applies to a statically-stable
tailless configuration, whereby only a positive wing
pitching moment is permitted. The pitching moment is
established with airfoil reflex and/or sweep with twist. If the It is interesting to note that positive camber forward of
wing is unswept, an airfoil with (cm) on the order of +0.05 30 % chord actually contributes toward positive pitching
is representative, and this is obtained with reflex. moment. Consequently, an "inchworm" camber line
However, a swept and twisted wing can use an airfoil with (returning to Figure 2.18) can develop high lift while
(cm) typically ranging between zero and +0.025. maintaining a near-zero (cm).
4.2Thin-Airfoil Calculation of cm and ß The function of Figure 4.2 has been numerically
Abbott and von Doenhoff (7) show, based on thin-airfoil integrated over the range X=0.0001 to 0.9999 to obtain
theory, that the pitching moment coefficient (cm) about the thin-airfoil prediction of (cm) for most of the airfoils
the quarter-chord point is the sum of two integrals
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

herein. It appears, however, that the thin-airfoil theory can The outline herein relates to the complete method and
significantly overpredict the magnitude of (cm) for cusped adds important details while simplifying and interpreting
airfoils when compared to test data (Table 4.1). Conse Thwaite's terminology. In addition, due to the extensive
quently, pitching moment will be revisited with the thick changes made by Thwaites relative to Theodorsen's
airfoil methods of Section 5.0. Nonetheless, the thin-airfoil original equations, two versions of the BASIC computer
theory provides valuable insight into the pitching moment program were prepared, one faithfully following Theodor
trends due to the camber shape. sen's original sign convention and method. These
programs obtained identical results. The method will be
Turning now to the zero-lift incidence, the integral (ßo) of applied here to predict the aerodynamic characteristics of
(EQ 4.2) is identical to (ß) as defined in Figure 4.1. Figure the Somers NLF(1)-041 6 airfoil (11), which is shown
4.3 shows this weighting factor along the chord. Again, together with its geometry math model in Figure 5.1.
trailing edge camber dominates the result. However, in
contrast with Figure 4.2, positive camber at any chord
wise position contributes toward positive (ß). Table 4.1
compares thin-airfoil predictions of (ß) with test data.

In his method, Theodorsen introduces two geometry

parameters (ø and e). These are analyzed in terms of the
polar angle (è), counter-clockwise from the trailing edge.
Calculations at (è=0, ð, 2ð) are avoided. Also, since
these parameters and their derivatives are periodic
around the circle, all related arrays should be conditioned
and extrapolated both left and right for continuity at
(ð=0 and 2ð). A small (è) increment, such as 6-degrees,
is required to properly observe leading edge pressure
spike effects, while a larger increment, say 12 degrees, is
best near the trailing edge.
The calculation begins by estimating the leading and
trailing edge radii (pn and pe), normalized to the chord. For
the trailing edge, (pe) is half the edge thickness, typically
a very small number. For the leading edge, a short
iterative calculation is shown below (as Algorithm 5.0),
where Theodorsen's chord parameter (4a/c) represents
the chord, less half the leading and trailing edge radii. The
algorithm includes Theodorsen's approximate formula for
5.0 THEODORSEN'S METHOD FOR VELOCITY
(pn). The angle (èn) is just beyond (ð), say 190 degrees.
In his two classic papers (5, 6) of 1931 and 1932, First, (Xn) is calculated from (èn) as shown. Then the airfoil
Theodorsen extends Zhukovsky's airfoil-to-circle confor math model of (î vs X) is used to calculate the lower
mai transformation so that the surface velocity on an surface coordinate (în). The result of the algorithm for the
airfoil of arbitrary shape can be analyzed. In addition, he Somers airfoil is (pn)=0.0109.
provides valuable insight and supporting formulas with
regard to the stagnation point, ideal angle of attack, zero
lift incidence, and cusped trailing edge velocity.
When data is unavailable, Theodorsen's method Calculate ?n from airfoil math model
frequently serves as the standard by which the more
modern vortex panel methods are tested. In contrast to
many contemporary methods, Theodorsen's method is
highly robust and computationally efficient. However, his
terminology is perhaps hard to follow. Fortunately,
Thwaites and Goldstein (10) provide outlines of both
complete and approximate versions of the calculations.
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

Next, the (X, ø, e, and î) arrays versus (è), and arrays Defining (ao) as the derivative of lift coefficient with angle
representing the derivatives (dø/dè, de/dè) are
calculated. of attack, all of Thwaites' references to (c1/a0) have been
Here, the (ø) parameter is defined by (EQs 5.1 and 5.2), replaced with (a+ß) for the interpretation here. (EQ 5.3)
the latter representing an update for maximum accuracy. has been significantly clarified and transformed from
Theodorsen's original, but it obtains the same results.
Translating Theodorsen's conclusions to modern sign
convention, we find that (ß) is equal to (-e) evaluated at
the trailing edge (è=0). This result is used in (EQ 5.3).
Theodorsen provides additional insight by calculating the
An average (øm) is also calculated. The parameter (e) is ideal angle of attack (a1), where the stagnation point
numerically evaluated from the distribution of (ø vs è) by a coincides with the leading edge. Also, he calculates a
straight-forward method outlined in the Appendix of theoretical cusped trailing edge velocity (ve). These are
Theodorsen's paper (6). For maximum accuracy, (øm) identified in (EQs 5.4 and 5.5). Thwaites observes that
and (e) can be updated based on an auxiliary "updated" Theodorsen's lift slope exceeds, typically by 10 %, the
angle (v=è+e). However, this update was not performed thin-airfoil prediction of ( 2ð = 0.1097 / deg ) by the ratio
for the calculations herein. Finally, we briefly review the (eøm
4a/c) as shown in (EQ. 5.6).
concept of vortex density to better interpret and imple
ment Theodorsen's velocity formula. Early attempts to adjust Theodorsen's formula to repre
sent an empirical lift slope (ao) led to infinite velocity at the
Anderson (12) shows that the local velocity (V) is equal in trailing edge. However, this problem is not encountered if
magnitude to the vortex density (y), or vortex strength per both (a) and (ß) are adjusted by an arbitrary and identical
unit surface length. Anderson explains this in terms of the factor when using (EQ 5.3). A reasonable factor could be
boundary layer, where at the surface, the velocity is zero. ao/ [2ð eøm 4a/c]. However, unity was used herein.
The local velocity (V) resides just outside of the boundary
layer. Downstream and clockwise from the stagnation
point, (y) is positive. Counter-clockwise downstream, (y) is
negative, as shown in Figure 5.2. As stated by Theodor
sen, both (V) and (y) are non-zero at the leading edge,
except when the airfoil is at its ideal angle of attack.

It is important to recognize that much of the pre-war

NACA lift and pressure coefficient data was incorrect.
Consequently, the data at the time verified only the
velocity trend predicted by Theodorsen. Lift slopes
measured in the 1930's were routinely up to 20 % lower
Since (y) has the units of velocity, it can be non-dimen than those published for the same airfoils after the war by
sionalized with flight velocity (Vo). Theodorsen's formula Abbott and von Doenhoff. In Theodorsen's paper, data for
(EQ 5.3) can then be interpreted to relate (yNo) to the the NACA M-6 airfoil was used to check his theory, but its
airfoil geometry and angle of attack. lift slope at the time was quoted at 0.08 per degree. Post
war NACA data showed that low-speed airfoils exhibit a
lift slope from 0.10 - 0.12 per degree with rare exceptions.
EQ (5.3):
We will compare the results of Theodorsen's method to
contemporary test data (11) for the Somers NLF airfoil.
First, the geometry parameters are shown in Figure 5.3.
Note the relationship between the angle (è) and (X), as
well as the periodicity of the geometry parameters (ø, e ).
At the trailing edge, (e) is approximately (-0.077) radians,
corresponding to (ß= +4.4°). The test data shows 3.8°.
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

Following the presentation of Liebeck (13), the vortex

density can be integrated along the normalized surface
distance (s/c), clockwise around the airfoil perimeter,
beginning at the trailing edge. This, according to the
Kutta-Zhukovsky theorem, yields the lift coefficient, which
is twice the area under the curve of (y/Vo) versus (s/c) as
shown for two angles of attack in Figure 5.5. Notice that
with increasing angle of attack, the stagnation point
moves to the left and the curve shifts upward while the
ends remain fixed. The predicted lift coefficient agrees
favorably with test data for the Somers airfoil, as tabulated
in the lower right-hand corner.

Figure 5.4 shows the surface velocity (vortex density) for

the Somers airfoil, from Theodorsen's method, at two
angles of attack. The test data, which includes viscous
effects, compares favorably with Theodorsen's inviscid
flow prediction. At both angles of attack, the X-coordinate
of the stagnation point (y/Vo=0) was predicted with
remarkable accuracy. The case for (a=0) is near the ideal
angle of attack (0.2°). Note the smoothness of the velocity
curves, as they are based on math-modeled geometry.

Next, to calculate the moment coefficient (cm4) about the

quarter-chord point, a vector approach is proposed. First,
velocities are converted to pressure coefficients (cp),
which are given by [1-(y/Vo)2]· Then, working counter
clockwise from the trailing edge as shown in Figure 5.6,
a tangential vector element (dt) of surface length (dt) is
directed along the airfoil perimeter. The local pressure
force vector (dF) of magnitude (cpqdt) is normal to the
surface element and is defined by the cross product of
(cpq dt) with the unit vector (j), which is directed into the
page. The corresponding vector radius (r) from the
quarter-chord position is also shown.

At the leading edge, (y/Vo) is approximately 1.6 for

(a=8.16°). At the trailing edge, the vortex densities are of
opposite sign and equal magnitude, thereby satisfying the
Kutta condition.

With confidence now in Theodorsen's method, we turn to

calculating the lift coefficient, moment coefficient, and
aerodynamic center, all based on the calculated vortex
density disribution.
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

Finally, (EQs 5.7-5.9) non-dimensionalize all variables By thin-airfoil theory, the aerodynamic center (ac) of the
and lead to the result of (EQ 5.10), which can be airfoil resides at X=0.25. Whether the airfoil is thick or thin,
interpreted as the counter-clockwise sum of surface the moment coefficient about the (ac) is invariant with lift
element pressure and coordinate properties. coefficient. Consequently, Table 5.1 suggests that the
(ac) of the Somers airfoil does not reside at (X=0.25).
We can calculate the location (Xac) of the (ac) and the
moment coefficient (cm) about the (ac) with the aid of
Figure 5.7, which shows the case where the aerodynamic
center resides behind the quarter-chord point. Here, an
increase in angle of attack leads to a lift coefficient
increase and a corresponding nose-down pitching
moment increment at the quarter-chord location. The
derivative (dcm/dc1) allows us to isolate the aerodynamic
center properties with (EQs 5.11-5.12).

The last term [in brackets] of (EQ 5.10) constitutes

approximately 3 % of the moment coefficient. Its sign is
opposite to that used in (Ref. 11). Changing the sign of a
3 % term has a 6 % effect on the calculation.

Returning now to Figure 5.6 and selecting an upper

surface point around 10 % chord, we note that (cp) is
negative. Also, (AX) is negative for a counter-clockwise
integration. Isolating the z-component of the local suction
force, a nose-up (positive) contribution to (cm) is found.
To attain a positive (cm) increment at this location, the
integration must subtract [cp (x-0.25) AX], which is a
negative quantity. Such is the case in (EQ 5.10).
Next we isolate the x-component of the local suction Applying these equations to Table 5.1, (Xac) for the
force, which is seen to develop a nose-down contribution Somers airfoil is predicted to reside at 27.4 % chord and
to (cm). Since (î) is positive while (cp and Aî) are negative, (cm) is predicted to be (-0.084). Now to approximately
correct the (cm) test data, the sign of the last term in
a counter-clockwise integration must subtract(cp
î Aî), (EQ 5.10) is temporarily reversed and the (cm) calculation
which is a positive quantity, to obtain the nose-down
effect. Such is the case in (EQ 5.10). It is concluded that is repeated to assess the effect. For the corrected test
the (cm) data of (11), which has been calculated from data at a Reynolds number of 4 million, we then apply
surface pressures, should be re-processed. (EQs 5.11-5.12), and find that (Xac) resides at 27.9 %
chord. Also, (cm) is found to be (-0.089). These agree well
Applying now (EQ 5.10) to the Somers airfoil, and using with the predictions of Theodorsen's method.
the pressure coefficients from Theodorsen's method and
math-modeled geometry, the quarter-chord pitching Finally, we have outlined, and verified, the basic geometry
moment coefficient is found to vary with lift coefficient as definition and low-speed aerodynamic design tools. We
shown in Table 5.1 below: turn in Section 6.0 to the math-modeling of new airfoils.

6.0 NEW AIRFOILS

Since maximum laminar flow is one of our primary goals,
we begin this last section by proposing the pmc-L15
airfoil of Figure 6.1 for application to a laminar aircraft
design such as that of Figure 6.2 (Ref. 14). Notice the
pusher-propeller feature which ensures that the flow over
the wing is undisturbed.
13
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

Calculated velocity distributions for the pmc-L15 are

shown in Figure 6.3. Again, the curves are remarkably
smooth, owing in part to the math-modeling of the
geometry. At the ideal angle of attack (0.44°), corres
ponding to (c1=0.53), the velocity gradient is favorable up
to approximately 45 % chord. Carmichael (15) states that
natural laminar flow, under the best of conditions, can be
maintained up to 70 % chord. However, he shows that the
chordwise location of transition from laminar to turbulent
flow depends on Reynolds number and angle of attack.
The pmc-L15 geometry math model uses the function Whereas the pmc-L15 develops a negative pitching
product thickness of (EQs 2.9-2.11). The maximum moment trimmed by the horizontal tail, the pmc-Z15 airfoil
thickness is 15%, residing at 50% chord. An inchworm of Figure 6.4 has zero pitching moment. This airfoil could
cusp and finite trailing edge thickness have been added be used for the full range of low-speed aircraft configura
as defined by the thickness increment of (EQ 6.1). For the tions. However, its primary applications would be (1)
camber, the inverted varabola of (EQ 2.8) is used, and the a tailless aircraft with sweep and twist, (2) an unswept
maximum camber is 4.5 %.
tailless aircraft with a reflexed, blended body, or (3) both
the foreplane and wing of a canard arrangement.
Although the pmc-z15 edge thickness is non-zero, the
trailing edge included angle is nearly zero. As shown in
Figure 6.5, the trailing edge velocity ratio is nearly unity.
This agrees well with Theodorsen's cusped trailing edge
prediction of 1.03. For the leading edge, note that (a) can
increase up to 5° without a leading edge velocity "spike."

14
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

Moving now to transonic airfoil applications, the pmc-T11 Finally, math modeling of a representative supersonic
of Figure 6.6 is very similar to that used on the Harrier airfoil is shown for the pmc-s05 in Figure 6.10. This is
aircraft of Figure 6.7. Although transonic airfoil design is similar to the airfoil used on the F-20 lightweight fighter of
beyond the scope of this paper, the pmc-T11 has been Figure 6.11. Since the primary parameter affecting
introduced to demonstrate the math modeling of a blunt supersonic airfoil drag is its thickness, this airfoil "spreads
leading edge and also to demonstrate the separate out" the thickness, much like a laminar airfoil, so that
modeling of the upper and lower airfoil surfaces. actuators and other equipment can be more readily
installed inside. A detail of the leading edge is shown. An
increase of the parameter (ë) could have been selected to
obtain a sharper leading edge.

Another transonic airfoil is the pmc-T18 of Figure 6.8. This

is applicable to the central portion of the blended wing
body transport (16) of Figure 6.9. As shown by
Woersching (17), transonic airfoils exhibit significantly
better lift-to-drag ratio if the camber is inverted. This has
the effect of lowering the strength of the shock on the
upper surface. An essential feature of the pmc-T18 for this
application is its pitching moment coefficient of +0.03.

SUMMARY
Math modeling tools for airfoil geometry have been
presented and applied toward the characterization of
existing and new airfoil geometries. The advantages of
math modeling, including maximum laminar flow, have
been stated. Three thickness models and numerous
camber math models have been proposed.
The qualitative effect of camber distribution on pitching
moment coefficient has been shown based on thin-airfoil
theory. Guidelines have been proposed for the selection
of airfoil camber as it relates to the pitch trim of conven
tional and unique aircraft configurations. Theodorsen's
method was outlined and applied toward a new presen
tation of surface velocity. A vector approach to calculating
the pitching moment coefficient was introduced.
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

RECOMMENDATIONS I> trailing edge included half angle (rad)

y vortex strength per unit surface length (m/s)
1. Math modeling in some form is proposed for the next K dimensionless maximum camber, z / c
generation of airfoils. Such modeling could be applied to A. leading edge exponent
specify new geometry or to closely match the geometry j..l skew parameter
which emerges from a computational fluid dynamics v varabola exponent
analysis which iteratively shapes the airfoil. e polar angle, counter-clockwise fror T.E. (rad)
p air density of free stream (kg/m )
2. An adjustable finite trailing edge thickness should 0' bell-shaped sine wave amplitude
always be included in the airfoil characterization.
't dimensionless maximum thickness, tic
(i) inchworm camberline exponent
3. Math-modeled airfoils, perhaps similar to those
proposed herein, should be wind-tunnel tested if possible.
s dimensionless z-coordinate, zlc

4. Math modeling should not stop at the two-dimensional

ACKNOWLEDGEMENT
level. Wings of arbitrary planform and twist, and blended
The author is grateful to Northrop Grumman Corporation
wing-body arrangements, can be completely modeled
for providing the technical review, sponsorship of this
with (x,y,z) equations. This is the topic of a future paper.
paper, and outstanding library support. He is also g~ateful
for the opportunity to present his paper with SAE. ThIS
document was prepared on Microsoft Word. All of the
NOMENCLATURE
graphics were imported. SigmaPlot was used for the
airfoil geometry design and performance plots. Mlcrografx
a cubic camber coefficient, first term
Designer was used for the sketches.
A varinomial coefficient, first term
A thickness factor for function product
b wing span (m)
REFERENCES
b cubic camber coefficient, second term
B varinomial coefficient, second term
1. Riegels, FW., "Aerofoil Sections," London
c airfoil chord (m)
Butterworths, 1961
C wing mean geometric chord (m)
C varinomial coefficient, third term 2 2. Selig, M.S., "The DeSign of Airfoils at Low Reynolds
cm airfoil pitching moment coefficient, dM / (q c dy)
Numbers," p. 60-61, 1984 AIM Student Paper
cM wing pitching moment coefficient, M/qSC
Conference
c1 airfoil lift coefficient, dL I (q c dy)
cd airfoil drag coefficient, dO / (q c dy)
3. Glauert,H., "Aerofoil and Airscrew Theory," p.75,
D drag force (N)
Cambridge University Press, 1943
L lift force (N)
m varinomial exponent, second term
4. Houghton,E.L. and Carruthers, N.B., "Aerodynamics
M wing pitching moment (N-m)
for Engineering Students," p. 172-175, Edward Arnold,
n varinomial exponent, third term 2
1982
q freestream dynamic pressure, (1/2) p Vo (N/m2)
s perimeter distance, clockwise from T.E. (m)
5. Theodorsen, T., "Theory of Wing Sections of Arbitrary
Swing planform area (m 2)
Shape," NACA Report No. 411, 1931
t airfoil maximum thickness (m)
t perimeter distance, counter clockwise from T.E. (m)
6. Theodorsen, T., and Garrick, I.E., "General
V local velocity outside of boundary layer (m/s)
Potential Theory of Arbitrary Wing Sections,"
Vu velocity of free stream (m/s)
NACA Report No. 452, 1932
x chordwise coordinate from leading edge (m)
y spanwise coordinate from wing-body axis (m)
7. Abbott, I.H. and von Doenhoff, A.E.,
z vertical coordinate from chord line (m) "Theory of Wing Sections," Dover, 1949
a angle of attack (rad)
8. Gale, F., "Tailless Tale," p. 71-84, B2 Streamlines,
aO angle of attack (deg) 1991, P.O. Box 976, Olalla, WA 98359
incidence of zero-lift line (rad)
~o incidence of zero-lift line (deg)
9. Jandel Scientific, "SigmaPlot for Windows,"
X dimensionless x-coordinate, X/c P.O. Box 7005, San Rafael, CA 94912
8J cusp depth parameter
82 cusp extent exponent
 Downloaded from SAE International by Imperial College London, Monday, August 20, 2018

10.Thwaites, B., "Incompressible Aerodynamics,"

Dover. 1960

11. Somers, D.M., "Design and Experimental Results for

a Natural Laminar Flow Airfoil for General Aviation
Applications," NASA TP 1861, 1981
12.Anderson, J. D., "Fundamentals of Aerodynamics,"
McGrawHill, 1984

13.Liebeck, R.H., "A Class of Airfoils Designed for High

Lift in Incompressible Flow," J. Aircraft, Vol.10, No.10,
Oct 1973

14.Selberg, B.P. and K. Rokhsaz, "Aerodynamic Tradeoff

Study of Conventional, Canard, and Trisurface Aircraft
Systems," J. Aircraft, Vol. 23, No. 10, Oct. 1986
15.Carmichael, B., "Personal Aircraft Drag Reduction,"
San Clemente, CA, 1995

16.Liebeck, R.H., et.al., "Concepts for Advanced

Subsonic Transports," NASA CR-4624, 1993
17.Woersching, T.B., "Negative Camber Airfoils for
Transonic Flight," J. Aero. Sci., Vol. 18, No.6, 1951

18.Althaus, D. and Wortmann, F.X., "Stuttgarter

Profilkataiog 1," Friedr. Vieweg & Sohn, 1981

ABOUT THE AUTHOR

Phil holds a BSME from the University of Arizona and
MSAE from Cal Poly Pomona. He is a Senior Technical
Specialist and Lead Engineer at Northrop Grumman
where he and his team predict air vehicle propulsion and
mission performance. Phil has a previous publication,
AAS 87-537, titled "Relating Conic-Section Orbital
Position, Velocity, and Time from Periapsis as
Dimensionless Quantities."