You are on page 1of 7

Conference of Global Chinese Scholars on Hydrodynamics

DESIGN AND OPTIMIZATION METHOD FOR


A TWO-DIMENSIONAL HYDROFOIL

Ching-Yeh Hsin, Jia-Lin Wu and Sheng-Fong Chang


Department of Systems Engineering and Naval Architecture, National Taiwan Ocean University, Keelung, China
E-mail: hsin@mail.ntou.edu.tw

ABSTRACT: In this paper, methods for the design and the the Lagrange multiplier method is presented for the
optimization of two-dimensional hydrofoils are presented. performance optimization.
First a previously developed potential flow design method is
The foil design problem was first studied by
extended to the viscous flow computations. In this design
method, the foil geometry is defined by B-splines, and the Lighthill[1], and he solved it for the case of the
RANS method is utilized for the viscous flow computations. incompressible flow by using the conformal mapping
The foil is designed based on a prescribed pressure distribution, technique. Since then, many methods have been de-
and the Newton-Raphson method is used to achieve a foil veloped, such as Tranen[2] , Henne[3] and Volpe[4] ,
geometry generating the prescribed pressure distribution. In and all these methods mapped the foil shape to a
the design method described above, although designers can
mapping plane. Eppler[5] and Drela[6][7] developed
specify a pressure distribution based on the desired local
characteristics, it is hard to know the forces on the desired foil methods extensively used by the industry. Epplers
in advance. Therefore, an optimization method by using the sections are famous for its advantages in the pre-
Lagrange multiplier method is then developed. The optimi- vention of cavitations, and Drelas method based on
zation problem described in the proposed paper is to satisfy the the boundary element method with the incorporation
lift requirement by minimizing the drag force. The design of the boundary layer computations, and this method
variables are the angle of attack and the geometry parameters.
is then developed as the widely used foil design tool
In the presented design examples, the camber distribution and
the pressure distribution shape function are selected as the XFOIL. Recently, the Genetic Algorithms and the
geometry parameter. Both the design and the optimi- Evolutionary Strategy are also utsed for the foil de-
zation methods can be used as a design tool individually; signs, such as De Falco[8][9][10] and Obayash[11].
however, the combination of two methods provides a useful In this paper, we will present two methods, and
design tool that can not only optimize the global performance, the first method is referred as the design method,
but also design the detailed geometry according to designers
and the second method is referred as the optimization
requirements. Design examples are illustrated in the paper,
and this design method is proved to be practical and effective. method. The design method is used to design a foil
based on a given pressure distribution, and the opti-
KEYWORDS: foil design, lagrange multiplier method, RANS mization method is used to optimize the foil per-
formance. In the design method, a pressure distri-
1. Introduction bution is given, and the foil geometry is directly
obtained in the Cartesian coordinate system by solving
Two-dimensional foil designs have been widely the Laplace's equation. Note that the pressure distri-
applied to different applications, such as marine pro- bution has to satisfy certain physical constraints such
pellers, rudders, pumps, wind turbine blades, cooling as existence of the stagnation point, satisfaction of the
fans, etc., and many newly-designed foil geometries Kutta condition, etc. A detail description of the direct
have been successfully applied to the performance method can be seen in Hsin[12]. This direct method is
improvements for different purposes. In the marine found to have difficulties to resolve the leading edge
propeller applications, two-dimensional sections, such geometry of the foil due to the abrupt changes of the
as NACA series, Newton-Rader sections, Eppler pressure distribution near the leading edge. There-
section etc. have been used for different design con- fore, an improved method is then developed. In the
siderations. In this paper, a design method for the improved method, the foil geometry is defined by
two-dimensional hydrofoil is first presented. In this B-splines (or NURBS), and the pressure distribution is
method, the foil geometry is designed based on a a function of B-spline control points. A nonlinear
given pressure distribution. Then, a method based on problem is then solved to perturb the B-spline control
points to match the pressure distribution to the given
323
pressure distribution. The improved method will be Results of a computational example are shown in
described in the section 2. Notice that the designed Fig. 2 and Fig. 3. Fig. 2 shows the designed pressure
foil geometry from the direct method is the initial distributions by the potential flow and by the viscous
solution of the improved method. This design method flow computations, and Fig. 3 shows the designed
has been successfully developed based on the po- geometries by the potential flow and viscous flow
tential flow boundary element method[13][14], and it computations. The viscous effect seems to be over
then has been extended to the application of viscous predicted in this example.
flow RANS code[15].
We will then present an optimization method 3. Optimization Method
using the Lagrange Multiplier method. The presented
optimization method can be used to optimize the per- In the design method described above, although
formance of the hydrofoil based on different criteria. designers can specify a pressure distribution based on
the desired local characteristics, it is hard to know the
2. Design Method forces on the desired foil in advance. Therefore, an
optimization method by using the Lagrange Multiplier
In this design method, the foil geometry is first method is then developed. The Lagrange Multiplier
de- fined by B-splines (Fig. 1): method can transfer a constrained problem to a non-
Nv Nv constrained problem by introducing the Lagrange
x(t) = (xv )i Bi4 (t) y (t ) = ( yv )i Bi4 (t ) multiplier . The constrained optimization problem
i =1 i =1
of the foil design problem in this paper is to find a
where t is the B-spline parameter, xv and yv two-dimensional foil geometry which provides the
are B- spline control points, Nv is the number of minimum drag coefficient ( CD ) with a given lift
4
B-spline control points, and Bi is the B-spine basis coefficient ( C L ). The constraint is therefore, CL = CL* ,
functions of order 4 (degree 3). The foil is designed
and CL* is the objective lift coefficient. The design
based on a prescribed pressure distribution, and the
Newton- Raphson method is used to achieve a foil problem thus can be stated as:
geometry generating the prescribed pressure min CD
distribution. Dif- ferent kinds of computational
(1)
methods can be used for the flow computations, and *
the RANS method is utilized for the viscous flow subject toCL CL = 0
computations in this paper. The design procedure is We also assume that the lift coefficient and the drag
described as follows: coefficient are functions of angle of attack and
The foil geometry is defined by B-splines, and geometric parameters. The geometric parameters
the pressure distribution on the foil can be ex- pressed can be defined by different ways such as the camber
as a function of B-spline control points; ratio, the thickness, NURB control points that
JJG
CP = C P ( X ) defined the foil geometry, etc.
JJG
X = ( xv , yv )1 ,( xv , yv )2 ,( xv , yv )3 ,......( xv , yv ) N v CL = CL ( , i )
i = 1,2,3...m (2)
A Jacobian is established by relating the geo- CD = CD ( , i )
metry changes and the pressure changes In equation (2), is the angle of attack, and is
i
(Cp )
[ J ] = [ JJG i ] the geometric parameters.
( X ) j We thus can define the Lagrangian of this opti-
Where (Cp )i is the pressure distribution on the mization problem as:
JJG
i collocation point, and ( X ) j is the j th B-spline
th
L = C D + (C L C L* ) (3)
control point.
By defining (Cp ) as the difference between and the derivatives of equation (3) with respect to
angle of attack, geometric parameters and Lagrange
the designed pressure distribution and the desired multiplier can be obtained:
pressure distribution, we can then solve a nonlinear
equation L = C D + CL = 0
(4)
[ (Cp )] = 0 CL C L* = 0
by the Newton-Raphson method L can also be expressed as
JJG JJG
[ X ]k +1 = [ X ]k [ J ]1[ Cp ]k

324
C D C I = 0 . The solution can be obtained by Newtons
+ L method:
i = 1,2,..., m (5)
G
C D C Gj I (Z j )
+ L Z =
i i I j (12)
G j +1 G j Gj
This is a non-linear optimization problem, and we Z = Z + Z
define:
In equation (12), the update of Z is to update

the angle of attack and the geometric parameters.
Once the angle of attack and the geometric parameters
L 1
G= Z = (6) are updated, CD and C L can be calculated.
* 2
C L C L
#

After an iterative procedure, Z can be solved, and
the design is reached.
m
If we only have one geometric parameter, and the
and the solution of this optimization problem is to above equations can be derived as follows:
solve:
G CL = CY cos C X sin
G( Z ) = 0 (7) CD = CY sin + C X cos
The Newtons method is used to solve this non-linear where C X and CY are X-component force and Y-
problem, and we can get component force of the foil. Equations (4) and (5)
G thus become:
Gj G(Z j )
Z = (8)
G j L CD CL (a)
= + = 0
Since the solutions are obtained through iterations, we
can express the Lagrange multiplier, , as a function L CD C
= + L = 0 (b)
of C L , CD and . From equation (5), we have
L
C D C = CL CL* = 0
+ L =0 (c)

The above equation can be further simplified by
and thus can be obtained: introducing equation (9):
CD CD CL
= (9)
C L = /

The number of variables to solve thus can be reduced, CD C
+ L =0
and we can further derive:
CD C C C = 0 *

+ L L L

i i

CD CD CL
(10) 4. Design Examples
i CL i
From equation (6) to (10), we first define From the theory introduced in the last section, we
can conclude the design procedure as the following
steps:
= =
(1) Under the condition of CL = CL* , computing
i
CD CL CD CL
And we have: , , , and ;

C (2) Computing elements of I and I in
I C D C (11)
D C L equation (11);
L (3) Solving equation (11) to obtain and ;
In order to solve this design problem, we have to solve (4) Updating the geometric parameter and angle

325
of attack ; NACA 0015 foil. Table 2 shows that the NACA
(5) Repeating steps (1) to (4), until the equation (4) is 0015 section has the lift coefficient 1.2 at the angle of
satisfied. attack 10.904 degrees, and the drag coefficient is
We will demonstrate several design examples in this 0.01076. The drag is almost 30% higher than the
section. The flow field computation is by a com- designed foil. Fig. 10 shows the comparison between
mercial RANS solver FLUENT. the designed foil and NACA 0015 section, and Fig. 11
shows the comparison of the pressure distributions
4.1 Design example 1 between the designed foil and NACA 0015 section at
In the first design example, we define the the same lift coefficient.
geometric parameter as the camber ratio, and the
Table 2 Verification of the design example 2
design goal is to obtain the lift coefficient CL = 0.7
Camber CL CD
with the minimum drag coefficient. ratio
The initial foil geometry is a NACA 1215 section Design 6.846 0.0396 1.2 0.00753
(thickness ratio is 15%, and the maximum camber is
Verification 10.904 0.0 1.2 0.01076
located at 20% of the foil from leading edge). Fig. 4
shows the convergence history of the angle of attack
4.3 Design example 3
(AOA) and the camber ratio (f/C), and Fig. 5 shows
In the design example 3, we try to combine both
the variations of the lift coefficients ( C L ) and drag
the design method and the optimization method
coefficients ( CD ) during the iterations. It is obvious described in this paper. In this example, we define the
that the solutions are converged at the 3rd iteration. geometric parameter as a shape function of the
In order to verify this design, we compare our design pressure distribution. That is, we maintain the shape
with the performance of the NACA 0015 foil (without of the pressure distribution, and then scale it according
the camber). Table 1 shows that the NACA 0015 to the geometric parameter . The design procedure
section has the lift coefficient 0.7 at the angle of attack is separated into the following steps:
6.249 degrees, and the drag coefficient is 0.00725. (1) Giving a range of the geometric parameters to
The drag is 15% higher than the designed foil. Fig. 6
limit the designs, and in this case, 1.0 1.1 ;
shows the comparison between the designed foil and
NACA 0015 section, and Fig. 7 shows the comparison (2) Using the optimization method to proceed to
of the pressure distributions between the designed foil the design. That is, to update the angle of attack
and NACA 0015 section at the same lift coefficient. and the geometric parameter ;
Notice that the pressure distribution look like a (3) For each geometric parameter -- a pressure
bump near the trailing edge is due to that the distribution, the design method is used to obtain
maximum camber distribution is located at the 20% of the geometry generating that pressure distribution;
the foil. (4) Repeat (2) and (3) until a convergent solution is
obtained.
Table 1 Verification of the design example 1 The given pressure distribution shape function
Camber CL CD of this design example is shown in Fig. 12, along with
ratio the pressure distribution of the designed foil based on
Design 4.328 0.02 0.7 0.00612 that pressure distribution. Note that this pressure
Verification 6.249 0.00 0.7 0.00725 distribution is equivalent to the pressure distribution
corresponding to = 1.0 . The objective lift coefficient
4.2 Design example 2 is set to be 0.7631. When we adjust the geometric
In the design example 2, we still define the parameter, the suction side of the pressure distribution
geometric parameter as the camber ratio; however, the will be scaled (the lift is increased as the geometric
objective lift coefficient is higher such that CL = 1.2 . parameter increases). The design results are shown in
The NACA 1215 section is still used as the initial foil Table 3. By comparing to the lower and upper limit of
geometry. the geometric parameters, we can find that the
Fig. 8 shows the convergence history of the angle designed foil gives a lower drag coefficient. Fig. 13
of attack (AOA) and the camber ratio (f/C), and Fig. 9 shows the designed foil geometry and geometries
shows the convergences of the lift coefficients ( C L ) corresponding to = 1.0 and = 1.1 , and Fig. 14
and the drag coefficients ( CD ). 8 iterations are shows the pressure distributions corresponding to the
needed in this example, and it is because the design designed foil and those corresponding to the geo-
lift coefficient is higher. metric parameters = 1.0 and = 1.1 .
Similarly, in order to verify this design, we
compare our design with the performance of the

326
Table 3 Design results of the design example 3
CL CD
Lower limit 1.000 2.9414 0.7631 0.00595
Upper limit 1.100 3.2054 0.7631 0.00845
Design 1.096 2.7796 0.7631 0.00474

5. Conclusions

In this paper, we present a design method that


can design a foil generating the given pressure dis- Fig. 3 The foil geometries designed by the viscous flow
tribution, and an optimization method to optimize computations and the viscous flow computations
the performance of the hydrofoil. Both methods can
be used as a design tool individually; however, the
combination of two methods provides a useful design
tool that can not only optimize the global performance,
but also design the detailed geometry according to the
designers requirements. Several design examples are
illustrated in the paper, and this design method is
proved to be practical and effective. This method can
also be used for other design purposes, for example, a
hydrofoil with a better cavitaion performance.

Fig. 4 The convergence history of the angle of attack (AOA)


and the camber ratio (f/C) of the design example 1

Fig. 1 A two-dimensional hydrofoil geometry defined by


B-splines

Fig. 5 The variations of the lift coefficients ( C L ) and drag


coefficients ( CD ) during the iterations of the design
example 1

Fig. 2 The desired pressure distribution and the designed


pressure distributions by the viscous flow computation
and the potential flow computation respectively

327
Fig. 6 The comparison between the designed foil (design 1) of
the design example 1 and NACA 0015 section (check)

Fig. 9 The variations of the lift coefficients ( C L ) and drag


coefficients ( CD ) during the iterations of the design
example 2

Fig. 10 The comparison between the designed foil (design 2)


Fig. 7 The comparison between the pressure distributions of of the design example 2 and NACA 0015 section
the designed foil (design 1) of the design example 1 (check)
and NACA 0015 section at the same lift coefficient

Fig. 8 The convergence history of the angle of attack (AOA)


and the camber ratio (f/C) of the design example 2 Fig. 11 The comparison between the pressure distributions of
the designed foil (design 2) of the design example 2
and NACA 0015 section at the same lift coefficient

328
References

[1] Lighthill, M.J. A new method of two-dimensional


aerodynamic design, RAND Technical Report M2112,
ARC. 1945.
[2] Tranen, J.L. (1974), A rapid computer aided transonic
airfoil design method, AIAA paper 74-501.
[3] Henne, P.A., An inverse transonic wing design method,
AIAA paper 85-0330., 1980
[4] Volpe, G and Melnik, R.E., The design of transonic
airfoils by a well posed inverse method, Int. J.
Numerical Methods in Engineering, Vol. 22, 1986
[5] Eppler, R. and Somers, D.M., A Computer Program for
the Design and Analysis of Low-Speed Aitfoils, Tech.
Rep., NASA TM 80210, 1980
[6] M. Giles, M. Drela, A two-dimensional transonic
aerodynamic design method, AIAA Journal, Vol.25,
Fig. 12 The given pressure distribution shape function No.9, 1986.
( = 1.0 ) of the design example 3, and the pressure [7] M. Drela, XFOIL: An Analysis and Design System for
distribution from the designed foil Low Reynolds Number Airfoils, in Low Reynolds
Number Aerodynamics, Vol. 54, 1989, Springer-Verlag
Lecture Notes in Eng.
[8] De Falco, R. Del Balio, A. Della Cioppa and E.Tarantina,
A Parallel Genetic Algorithm for Transonic Airfoil
Optimisation, Evolutionary Computation, IEEE
International Conference, Volume: 1, PP.429-434, 29
Nov-1 Dec 1996.
[9] De Falco., An introduction to Evolutionary Algorithms
and their application to the Aerofoil Design Problem Part
IThe Algorithms, Inverse Design and Optimisation
Method, von Karman Institute for Fluid Dynamics,
Lecture Series, 1997-05.
[10] De Falco. An introduction to Evolutionary Algorithms
and their application to the Aerofoil Design Problem Part
II The Results, Inverse Design and Optimisation
Method. von Karman Institute for Fluid Dynamics,
Lecture Series 1997-05.
[11] Shigeru Obayashi, Susumu Takanashi, Genetic
Algorithm for Aerodynamic Inverse Optimization
Fig. 13 The designed foil geometry and foil geometries Problems, Genetic Algorithms in Engineering Systems:
corresponding to the geometric parameters = 1.0 Innovations and Applications, pp.7-12, 12-14, Sep 1995.
[12] Hsin, C.-Y., "Application of the panel method to the
and = 1.1 in design example 3 design of two-dimensional foil sections", J. of Chinese
Society of Naval Architecture and Marine Engineers,
Vol.13,No.2,pp.1-11, 1994
[13] Hsin, C.-Y. and Chang, Y.-L., Solving a Hydrodynamic
Design Problem by a Distributed Computing System,
Proceedings of the 3rd International Conference on
Hydro-dynamics, Seoul Korea, Oct. 1998.
[14] Hsin, C.-Y. and Chang, Y.-L., A Hydrodynamic Design
Method Developed on a Distributed Computing System,
Transcations of the Aeronautical and Astronautical
Society, Vol. 32, No.1, pp.89-95, 2000
[15] Wu, Jia-Lin, Foil Design by Viscous Flow
Computation, Master Thesis, National Taiwan Ocean
University, 2005 (in Chinese)

Fig. 14 The pressure distributions corresponding to the


designed foil and geometric parameters = 1.0 and
= 1.1 in design example 3

329

You might also like