13th AIAA/ISSMO Multidisciplinary Analysis Optimization Conference AIAA 2010-9050

13 - 15 September 2010, Fort Worth, Texas

A Comparison of Airfoil Shape Parameterization Techniques

for Early Design Optimization

V. Sripawadkul1, M. Padulo2, and M. Guenov3

Cranfield University, Cranfield, MK43 0AL, UK

Parameterization has a significant role in airfoil design and optimization where, especially in the
early design stages, the aim is to be able to explore as many design alternatives as possible while
keeping the number of design parameters to a minimum. Several desirable characteristics, namely,
Parsimony, Intuitiveness, Orthogonality, Completeness and Flawlessness are considered. These are
applied as comparison metrics to five airfoil parameterization techniques: Ferguson’s curves, Hicks-
Downloaded by PURDUE UNIVERSITY on January 18, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2010-9050

Henne bump functions, B-Splines, PARSEC, and Class/Shape function Transformation. The method
allows each parameterization technique to be ranked according to each criterion considered, thus
providing a basis for objective multiple criteria decision making.

Nomenclature
rle = Leading edge radius
S = Shape Function
t = Curvilinear coordinate
x = Chord wise position
xup, xlo = Upper and lower crest location
xte = Trailing edge position
zup, zlo = Upper and lower crest value
zxx,up, zxx,lo = Upper and lower curvature at crest location
zte = Trailing edge vertical coordinate
Δzte = Trailing edge thickness
αte = Trailing edge direction
βte = Trailing edge wedge angle

I. Introduction

A t present, the parameterization techniques used in aerospace design can be separated into two main categories:
computer aided design (CAD) based techniques, and analytical techniques. Modern CAD systems have become
more sophisticated and are able to represent complex geometry using Non-uniform Rational B-Splines (NURBS)
curves and surfaces. However, using control points as design variables in CAD is considered to be computationally
expensive. Furthermore, such an approach does not provide physical meaning so that the designer could understand
the change one is trying to make to the design. Analytical techniques have been developed in parallel to overcome
this problem, but challenges associated with practical application still remain. In this respect, efficiency in design
parameterization is seen as a compromise between achieving sufficient level of detail and local control on the one
hand, and minimizing complexity of the design task on the other. Within this context, the objective of this paper is to
explore the applicability of a number of criteria for the selection of an airfoil parameterization for the needs of
multidisciplinary design optimization at early design stage.
The next section gives a background on airfoil shape parameterization methods. Section III presents the
methodology for the comparison, followed by the results from airfoil fitting tests reported in Section IV. Finally,
conclusions are drawn and future work is outlined in section V.

1
PhD Student, Aerospace Engineering Department, Bldg. 83.
2
Research Officer, Aerospace Engineering Department, Bldg. 83.
3
Professor, Aerospace Engineering Department, Bldg. 83.

1
American Institute of Aeronautics and Astronautics

Copyright © 2010 by the American Institute of Aeronautics and Astronautics, Inc. All rights reserved.
 II. Background
A. Airfoil Shape Parameterization Methods
Five shape parameterization methods are considered in this paper. These are methods which have been widely
cited in the literature and which are considered to have potential practical application. Ferguson’s curves and Splines
are used to generate curves in standard Computational Aided Design (CAD) tools. Hicks-Henne bump functions and
Parametric Section (PARSEC) were both specifically developed to generate airfoil profiles. Last, but not least,
Class/Shape function Transformations (CST) have been introduced to generate a wide range of shapes used in
aircraft geometry representation.

1. Ferguson’s Curves
James Ferguson1 first introduced the curve definition in Computer Aided Design (CAD) in 1964. The curve z(t)
is defined as Hermite polynomial on parametric points t as follows:

⎡ 1 0 0 0 ⎤⎡ A⎤
⎢ 0 1 0 ⎥⎥ ⎢⎢ B ⎥⎥
Downloaded by PURDUE UNIVERSITY on January 18, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2010-9050

3 ⎢ 0
z (t ) = [1 t t 2 t ] (1)
⎢ −3 3 −2 −1⎥ ⎢TA ⎥
⎢ ⎥⎢ ⎥
⎣ 2 −2 1 1 ⎦ ⎣TB ⎦
Six parameters are required to construct a complete airfoil as shown in Figure 1. The curves start from leading
edge (A) to trailing edge (B) given also their corresponding tangent vectors TA and TB. Camber angle, αc and boat
tail angle, αb define the orientation of tangent vectors, TBlower and TBupper, respectively. The trailing edge position on
upper and lower surface, xte ,up and xte ,lo can be considered as two additional variables for fitting existing airfoils.

TAlower
zu (t) αb TBupper
A αc B
TBlower
zl (t)
TAupper

Figure 1. Airfoil described by 2 Ferguson’s Curves.

2. Hicks-Henne Bump Functions

Hicks and Henne2 introduced a compact formulation for parameterization of airfoil sections. The perturbation is
a linear superposition of “bumps” or analytical shape functions on baseline airfoil geometry.
The shape functions are defined as:

bi ( x) = sin t (π x ln(0.5) ln( xMi ) ) , (2)

where xMi is the position of maximum point of the bump, and t controls the width of the bump.
The contribution of each parameter is determined by the value of the participating coefficients, αi associated
with a shape function. The airfoil curve z is therefore described by:
N
z ( x) = zbasis ( x) + ∑ α i bi ( x) . (3)
i =1
3. B-Splines
A Bezier curve of order n is described as:
n
z (t ) = ∑ Bin (t ) ⋅ Pi , (4)
t =0
n
where Bi is Bernstein Polynomials, Pi represents the set of n+1 control points, and t is curvilinear coordinate.

2
American Institute of Aeronautics and Astronautics
 The Bernstein polynomials are described by:

⎛n⎞
Bin (t ) = ⎜ ⎟ t i (1 − t ) n −i , i = 0,1,..., n , t ∈ [0,1] . (5)
⎝i⎠
A Bezier curve is accurate when representing simple curves. As the curve complexity increases, the degree of the
polynomials must be increased, which results in a larger error. In order to represent a complex curve, it is more
efficient to separate the curve into segments and use a set of low-order Bezier curves instead.
The B-splines is the generalization of the Bezier curve, and is defined as:
m
z (t ) = ∑ Pi ⋅ N i , p (t ) , (6)
i =0

where p is the order of the polynomial, Pi is a set of control points, N i , p is B-splines basis function.
Downloaded by PURDUE UNIVERSITY on January 18, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2010-9050

A detailed description of the basis function can be found in3. Figure 2 shows a B-spline airfoil generated with 13
control points.

Figure 2. B-Splines curve representing airfoil by means of 13 control points.

4. Parametric Section (PARSEC)

The PARSEC method, proposed by Sobieczky4, uses eleven geometric parameters which directly manipulate the
shape of the airfoil as summarized in Table 1.

Parameters Descriptions
rle leading edge radius
xup, xlo upper and lower crest location
zup, zlo upper and lower crest value
zxx,up, zxx,lo upper and lower curvature at crest location
Δzte trailing edge thickness
zte trailing edge vertical coordinate
βte trailing edge wedge angle
αte trailing edge direction
Table 1. Descriptions of PARSEC-11 parameters.

The airfoil shape for each upper and lower curve is described by a linear combination of suitable shape
functions:
6 1
n−
z ( x) = ∑ an x 2
, (7)
n =1

3
American Institute of Aeronautics and Astronautics
 The coefficients an are determined from solving system of equations for the upper profile as follow:
⎡ 1 0 0 0 0 0 ⎤
⎢ 1 3 5 7 9 11⎥
⎢ x2 x 2
x 2
x 2
x 2
x ⎥ a2 ⎡ 2rle ⎤
⎢ te te te te te te
⎥⎡ 1⎤ ⎢ ⎥
⎢ 12 3 5 7 9 11
⎥ ⎢ a ⎥ ⎢ z + Δzte ⎥
⎢ xup x2
x2
x2
xup2 xup ⎥ ⎢ 2 ⎥ ⎢ te
2
up up up
2 ⎥ (8)
⎢ 1 −1 ⎢ ⎥
3 12 5 32 7 52 9 72 11 92 ⎥ ⎢ a3 ⎥ = ⎢ zup ⎥
⎢ xte 2 xte xte xte xte xte ⎥ a ⎢ ⎥
⎢ 2 2 2 2 2 2 ⎥ ⎢ 4 ⎥ ⎢ tan(α te − βte ) ⎥
⎢ 1 −1 3 12 5 32 7 52 9 72 11 92 ⎥ ⎢⎢ a5 ⎥⎥ ⎢ ⎥
⎢ xup2 xup xup xup xup xup ⎥ ⎢ 0 ⎥
⎢ 2 2 2 2 2 2 ⎥ ⎣⎢ a6 ⎦⎥ ⎢ z ⎥
⎢ 1 −3 ⎣ ⎦
99 72 ⎥
xx ,up
3 − 12 15 12 35 23 53 25
⎢ − xup2 xup xup xup xup xup ⎥
⎣ 4 4 4 4 4 4 ⎦
Downloaded by PURDUE UNIVERSITY on January 18, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2010-9050

5. Class/Shape function Transformation (CST)

One of the recently developed methods for geometry representation was proposed by Kulfan5 in 2006. Two-
dimensional geometry is represented as the product of the class function C(x), and a shape function S(x) as follows:

z ( x) = CNN21 ( x) ⋅ S ( x) . (9)

The Class function is given in a generic form by:

CNN21 ( x) ≡ ( x) N1 (1 − x) N2 , for 0 ≤ x ≤ 1 . (10)

The exponents N1 and N2 range from 0 to 1 which yields an arbitrary shape. In order to generate a general NACA
symmetric airfoil, the exponents N1 and N2 are equal to 0.5 and 1.0, respectively. The first term, x , produces
round leading edge, while the second term, (1 − x) , ensures a sharp trailing edge. The selection of these exponents
makes the basis shape in the “airfoil” class. Other airfoils can be derived from this class function.
With additional term defining trailing edge thickness, Eq. 9 becomes:

z ( x) = C1.0
0.5
( x) ⋅ S ( x) + ( x) ⋅ Δzte , (11)

To generate arbitrary airfoil shape, Bernstein polynomial is also chosen as the shape function:

⎛n⎞
S ( x) = Bin ( x) = ⎜ ⎟ ⋅ xi (1 − x ) .
n −i
(12)
⎝i⎠
The complete equations to represent the upper and lower surface of CST airfoils then become:

n ⎡ ⎤
⎛n⎞
z = x ⋅ (1 − x ) ⋅ ∑ ⎢ Ai ⋅ ⎜ ⎟ ⋅ x i ⋅ (1 − x) n −i ⎥ + x ⋅ Δzte . (13)
i =0 ⎣ ⎝i⎠ ⎦
B. Previous Comparison Studies
There have been several attempts at comparison of parameterization methods. Samareh6 rated nine
parameterization approaches with ten criteria focusing on the efficiency, effectiveness, ease of implementation, and
analytical sensitivities for geometry and grids availability. The rating was given in 3 levels: good, fair or poor. This
provides the basic comparisons on parameterization, which in this paper will be strengthened to reflect the selection
criteria taking into account also the recently developed parameterization methods as outlined above. Examples of
pair-wise numerical comparison between parameterization methods considered in this paper can be found in:
Sobester and Barrett7 (Ferguson’s Curve vs. B-Splines), Wu et al.8 and Kumano et al.9 (PARSEC vs B-Splines), and
Azamatov et al.10 (Non-Uniform Rational B-Splines vs CST)

4
American Institute of Aeronautics and Astronautics
 III. Methodology
The comparisons of the five airfoil parameterisation methods outlined in the previous section are based on five
desirable properties summarized by Padulo et al.11 and made more precise in this work. These are: Parsimony,
Orthogonality, Completeness, Flawlessness and Intuitiveness. In each property, the parameterisation methods are
tested and given a rating according to their performance.

A. Parsimony
Parsimony defines the preference for the parameterization method which can induce significant changes in the
main geometric features of the airfoil by using the smallest number of parameters. Ferguson’s curves and PARSEC,
adopt a fixed number of parameters, while the other considered methods can use more or fewer parameters
depending on the application. In such cases, the parsimony of the parameterization method has been assessed by
deriving from the literature the suggested minimal number of design parameters for each method.
Analysis of the sensitivity of the flow solver to the geometrical definition performed by Trepanier et al12
suggested that for NACA 2412 airfoil, 9 control points are sufficient for a precision of 9x10-5. Various types of
Downloaded by PURDUE UNIVERSITY on January 18, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2010-9050

airfoil were fitted with the specified tolerance and it was showed that the selected airfoils could be represented with
13 control points or less.
For Class/Shape function Transformation, the number of design parameters depends on the order of the
polynomials. Kulfan5 used the standard wind tunnel tolerances (±3.5x10-4 from leading edge to 20% units of chord
and ±7x10-4 elsewhere) to determine the lowest order of the polynomials and concluded that Bernstein polynomial
of order 5 is sufficient for most of airfoil approximations. This yields 6 coefficients as design variables for each
surface of the airfoil. And since the first coefficients of each polynomial are set to be equal to ensure constant
leading edge radius, the total number of design variable becomes 11. The wind tunnel tolerances are used to
determine number of bumps required for Hicks-Henne bump functions.

B. Completeness
Completeness assesses whether the parameterization can describe any airfoil, up to a specific degree of accuracy.
In general, the ability to describe a larger number of shapes is expected to directly correlate with the
parameterization degrees of freedom. In this paper, however, the completeness score is obtained by assessing the
capability of the methods to fit a significant number of existing airfoils, by fixing the considered number of
parameters to the parsimonious value, as described in the previous section. The existing airfoils coordinates are
taken from the University of Illinois at Urbana-Champaign (UIUC) airfoil database13 which contains over 1,550
airfoils. The airfoils of interests in this paper are conventional types. Additionally, only airfoils which are described
with more than 60 coordinates were selected from the database in order to ensure that they can be directly used in
the fitting procedure without any pre-smoothing. This brings down the number of airfoils to 250, covering a wide
range used in conventional aircrafts.
To perform fitting test, the selected airfoil coordinates is imported into the MATLAB14 environment where the
design parameters and parameterized airfoil coordinates are generated. The root mean square of the residuals is then
used as the objective function for the MATLAB optimizer, fmincon, while the constraints are taken to represent the
standard wind tunnel tolerances15.

C. Orthogonality
Orthogonality guarantees that each airfoil shape corresponds to a unique set of input parameters. This property is
particularly relevant to the parameterization methods which construct an airfoil by combining existing sets of airfoils
or other types of base functions. This in turn can lead to spurious modality which arises where similar airfoil
geometries are created by different combinations of the base functions16 and is considered undesirable for
optimization processes.
Reuther et al.17 pointed out that in the Hicks-Henne formulation, the sine bumps which are used as base
functions are non-orthogonal, and thus the solutions are not guaranteed to be attained. Keane and Nair16also proved
that Ferguson’s curve and B-Splines are non-orthogonal.
Ceze et al18 have tested the orthogonality of the CST and found that at high degree of the polynomials, there
exists another set of coefficients which yields the same profile as a set of unitary coefficients. This behavior starts
occurring at the order of polynomial of 30 which is far beyond the order of the polynomials considered in this paper.
The design spaces for PARSEC were defined from the bounds of each variable for each parameterization
methods in section B. One thousand samples from this design space were generated through Latin Hypercube and
the condition number is checked for each sample to prove orthogonality.

5
American Institute of Aeronautics and Astronautics
 D. Flawlessness
Flawlessness guarantees that the parameterization technique will not generate an “ill-behaved” shape. B-Splines
uses splines interpolation while Ferguson’s Curve, PARSEC and CST use polynomial interpolation and therefore are
subject to fitting oscillations.
Previous research by Padulo et al11 on PARSEC formulation found an undesirable bump which occurs at another
position of the profile in addition to the defined maximum crest location. Another flaw mentioned in the same paper
was compenetration. These flaws are checked on the generated samples used in section C.

E. Intuitiveness
Intuitiveness considers whether the method relates to the physical design meaning of the parameters, which in
turn simplifies the choice of input bounds or design judgement. The design parameters are assessed according to the
following criteria:
- The design parameters are directly related to the geometrical parameters
- The designer gets the physical meaning of the parameter change which he/she is trying to impose
Downloaded by PURDUE UNIVERSITY on January 18, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2010-9050

IV. Results
Five parameterization methods: Ferguson Curves, Hicks-Henne bump functions, B-Splines, PARSEC, and CST,
were compared against five desirable criteria as defined above. The score is given according to their respective
performance in each criterion, ranging from 1 (worst) to the full score of 4 (best).

A. Parsimony
The minimum numbers of design variables required for Ferguson’s Curve and PARSEC are fixed due to their
formulation, while the parsimonious number of B-Splines and CST are taken from the literature5,12. For Hicks-
Henne bumps, the number of required design variables is determined from fitting airfoil with typical wind tunnel
tolerance. The effect of number of bumps on representation of the RAE2822 airfoil is presented in Figure 3. The
maximum residuals from the leading edge to 20% chord (lower line) are stay within the tolerance of 3.5x10-4, while
the residuals on the rest of the airfoil fall within the tolerance of 7x10-4 when using 16 bumps or more. This is the
number of bumps required for each upper and lower profile, and therefore resulting in total of 32.

Maximum residuals (1/10,000) RAE2822 with RAE100 base

24.5
21 0.2<x<1.0
17.5 0<x<0.2
14
10.5
7
3.5
0 number
10 11 12 13 14 15 16 17 18 19 20 of bumps

Figure 3. Effect of number of bumps on representation of RAE2822 airfoil.

Amongst five methods, Ferguson’s curve is the most parsimonious with 8 variables and is therefore given the
full score of 4. The other methods were linearly scaled according to this as presented in Table 2.

Method Number of variables Converted Score Remarks

Ferguson’s Curves 8 4.0 Fixed variables
Hicks-Henne 32 1.0 Depends on number of bumps
B-Splines 9 3.5 Depends on number of control points
PARSEC 11 2.9 Fixed variables
CST 11 2.9 Depends on order of polynomials
Table 2. Parsimonious number and converted score.

6
American Institute of Aeronautics and Astronautics
 B. Completeness
Fitting performances were evaluated against the wind tunnel tolerance. Examples of a parameterized RAE2822
airfoil with Ferguson’s curve and PARSEC are presented in Figure 4. As can be seen, Ferguson’s curve shows
limitations in capturing the sharp curvature on the lower profile, whereas PARSEC, as well as the other three
methods (which perform in the same level of accuracy), achieve good approximation of the profile.
Downloaded by PURDUE UNIVERSITY on January 18, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2010-9050

Figure 4. Parameterized RAE2822 airfoil by Ferguson’s curve and PARSEC.

The parameterized airfoil which falls within acceptable tolerance after performing a fitting test is considered as a
good approximation. The number of good approximation airfoils out of the selected 250 airfoils is summarized in
Table 3. The results show that Hicks-Henne bump functions, which is described with the highest number of design
variables is able to capture all airfoil features considered in this paper. B-splines capture 98.4% of all airfoils,
followed by PARSEC and CST. The Ferguson’s curve captures the lowest number of airfoils, since it describes
airfoil geometry with the lowest number of design variables.

Methods Number of Good Approximation [/250] Percentage Converted Score

Ferguson’s Curve 150 60.0 2.4
Hicks-Henne 250 100.0 4.0
B-Splines 246 98.4 3.9
PARSEC 235 94.0 3.8
CST 234 93.6 3.7
Table 3. Percentage of airfoils with satisfactory fitting approximation.

The minimum and maximum of design variables for PARSEC and CST are shown in Table 4. These values are
used as a boundary to construct the design space for the orthogonality test in the following section.

Variables rLE xup xlo zup, zlo zxx,up zxx,lo zte Δzte βte αte
min 0.003 0.25 0.191 0.040 - 0.061 - 0.726 0.14 0.0037 0.0001 - 0.28 0.032
max 0.018 0.46 0.521 0.067 - 0.020 - 0.197 1.00 0.0052 0.0008 - 0.05 0.131
Table 4. Minimum and maximum values of PARSEC design variables.

C. Orthogonality
Orthogonality with regard to Ferguson’s Curve, Hicks-Henne, B-Splines and CST can be concluded from the
published literature. The first three methods are non-orthogonal and therefore receive no point. On the other hand,
CST is orthogonal but this holds for the polynomials order below 30 which is quite sufficient for practical purpose.
For PARSEC, orthogonality is assessed by examination of the condition numbers of 1,000 samples. The condition
numbers of these samples range from 3,945 to 22,137 which are sufficient to prove orthogonality and therefore
receive full score.

Methods Orthogonality Score

Ferguson’s Curve
Hicks-Henne Non-orthogonal 0
B-Splines
PARSEC Orthogonal 4
CST Orthogonal, for polynomials order under 30 4
Table 5. Orthogonality of five methods.

7
American Institute of Aeronautics and Astronautics
 D. Flawlessness
The flaws considered in this paper are additional bump at another location apart from at maximum crest position,
and compenetration. The number of flaws is summarized in Table 6. The total number of non-flaw airfoils is then
counted and converted to the score of 2.9 out of 4.0. The design space could be rendered robust through a procedure
based on self-organizing maps, as can be found in the work by Padulo et al11.

Flaws Number of Cases [/1,000]

Additional bump on upper profile 9
Additional bump on lower profile 132
Compenetration 131
Total 272
Table 6. Number of flaws in PARSEC formulation.

At this stage, no flaw mentioned above is found in other four methods and therefore still keep a full score of 4.
However, this does not guarantee flawlessness and still require further investigation for the future work.
Downloaded by PURDUE UNIVERSITY on January 18, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2010-9050

E. Intuitiveness
The PARSEC method appears to be the most intuitive amongst all methods compared in this paper since it
directly describes the main geometrical features of the airfoil. All the design parameters such as: leading edge
radius, trailing edge thickness, boat-tail angle, etc., carry geometrical meaning. Designers can understand the effect
on geometry when manipulating one or more of these design parameters, for instance, larger leading edge radius,
thicker airfoils, or changing the location of maximum crest position. This is useful for optimization processes,
especially when the designer understands how to improve the aerodynamic performance of airfoils.
Regarding CST, Kulfan19 managed to incorporate a set of geometric parameters into the shape functions in the
same manner as PARSEC. This allows the designers directly to manipulate the shape of the airfoil using these
geometric parameters instead of polynomials coefficients. This is considered very practical and therefore both
methods received the highest score of 4.
The Hicks-Henne bump functions provide information about the influence of parameter change on profile by
manipulating contributions of each base function, while B-Splines enable the designers to manipulate the shape by
moving control points. These design variables are not directly related to geometry, but still provide the designer with
a physical meaning of the design change; therefore these two methods are ranked one unit below the maximum
score of PARSEC and CST. Lastly, Ferguson’s curve which provides only information about the positions and
tangent vector directions at the starting and ending positions of each profile, but does not provide control over the
curve, received lowest score of 2.
All the score are summarized in Table 7 as shown below.

Methods Parsimony Completeness Orthogonality Flawlessness Intuitiveness

Ferguson’s Curve 4.0 2.4 0.0 4.0 2.0
Hicks-Henne 1.0 4.0 0.0 4.0 3.0
B-Splines 3.5 3.9 0.0 4.0 3.0
PARSEC 2.9 3.8 4.0 2.9 4.0
CST 2.9 3.7 4.0 4.0 4.0
Table 7. Scoring of Airfoil Parameterization Methods.

V. Conclusions
Presented in this paper are five objective metrics, which are intended to help the designer to choose airfoil
parameterization methods according to his/her preferences regarding multidisciplinary design optimization at early
design stage. These include Parsimony, Completeness, Intuitiveness, Orthogonality, and Flawlessness. The metrics
are demonstrated on a comparison of five popular parameterization methods, including Ferguson’s Curves, Hicks-
Henne bump functions, B-Splines, PARSEC, and CST. The results suggest that the metrics can provide a basis for
objective comparison, thus allowing the designer to select the proper parameterization according to the problem at
hand.

8
American Institute of Aeronautics and Astronautics
 References
1
Ferguson, J., “Multivariable Curve Interpolation”, Journal of the Association for Computing Machinery, Vol.11, No.2, pp.
221-228, 1964.
2
Hicks, R. and Henne, P., “Wing Design by Numerical Optimization”, Journal of Aircraft, Vol. 15, 7, pp. 407-413, 1978.
3
Farin, G., “Curves and Surfaces for Computer Aided Geometric Design”, Academic Press, New York, 1990.
4
Sobieczky, H., “Parametric Airfoils and Wings”, Notes on Numerical Fluid Mechanics, Vol.16, pp.71-88, 1998.
5
Kulfan, B.M. and Bussoletti, J.E., “Fundamental Parametric Geometry Representations for Aircraft Component Shapes”,
AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, Portsmouth, Virginia, AIAA paper 2006-6948, 2006.
6
Samareh, J.A., “A Survey of Shape Parameterization Techniques”, CAES/AIAA/ICASE/NASA Langley International
Forum on Aeroelasticity and Structural Dynamics, Williamsburg, VA, 1999.
7
Sobester, A. and Barrett, T., “The Quest for Truly Parsimonious Airfoil Parameterisation Scheme”, 8th AIAA 2008 ATIO
Conference, Anchorage, USA, AIAA paper 2008-8879, 2008.
8
Wu, H.Y., Yang, S., Liu, F., and Tsai, H.M., “Comparison of Three Geometric Representations of Airfoils for Aerodynamic
Optimization”, 16th AIAA Computational Fluid Dynamics Conference, Orlando, FL, AIAA paper 2003-4095, 2003.
Downloaded by PURDUE UNIVERSITY on January 18, 2015 | http://arc.aiaa.org | DOI: 10.2514/6.2010-9050

9
Kumano, T., Jeong, S., Obayashi, S., Ito, Y., Hatanaka, K., and Morino, H., (2006) “Multidisciplinary Design Optimization
of Wing Shape for a Small Jet Aircraft Using Kriging Model,” AIAA paper 2006-0932.
10
Azamatov, A.I., Lee, J.W., Byun, Y.H., Kim, S.H., “Advanced Configuration Generation Technique for the Complex
Aircraft Geometry”, IEEE/ASME International Conference on Advanced Intelligent Mechatronics, China, 2008.
11
Padulo, M.., Maginot, J., Guenov, M. and Holden, C., “Airfoil Design under Uncertainty with Robust Geometric
Parameterization”, AIAA paper 2009-2270, 2009.
12
Trepanier, J.Y., Lepine, J., and Pepin, F., “An Optimized Geometric Representation for Wing Profile Using B-SPLINE”,
CASI Journal, Vol.46, No.1, pp.12-19, 2000.
13
University of Illenois at Urbana-Champaign, UIUC Applied Aerodynamics Group, UIUC Airfoil Coordinates Database,
[online database], URL: http://www.ae.illinois.edu/m-selig/ [cited 22 JUL 2010]
14
Matlab, MATLAB User’s Guide, The Mathworks, 2004
15
Sobester, A., “Exploiting Patterns in the Kulfan Transformations of Supercritical Airfoils”, 9th AIAA Aviation Technology,
Integration, and Operation Conference, South Carolina, AIAA paper 2009-6951, 2009.
16
Keane, A.J., and Nair, P.B., “Computational Approaches for Aerospace Design”, John Wiley & Sons, Sussex, England,
2005.
17
Reuther, J., Jameson, A., Farmer, J., Martinelli, L., and Saunders, D., “Aerodynamic Shape Optimization of Complex
Aircraft Configurations via an Adjoint Formulation”, AIAA paper 1996-94, 1996.
18
Ceze, M., Hayashi, M., and Volpe, E., “A Study of the CST Parameterization Characteristics”, 27th AIAA Applied
Aerodynamics Conference, San Antonio, Texas, AIAA paper 2009-3767, 2009.
19
Kulfan, B.M., “Universal Parametric Geometry Representation Method – CST”, 45th AIAA Aerospace Science Meeting and
Exhibition, Reno, Nevada. AIAA paper 2007-62, 2007.

9
American Institute of Aeronautics and Astronautics