Appl. Math. Mech. -Engl. Ed.

, 2008, 29(5):603–610
DOI 10.1007/s10483-008-0505-y Applied Mathematics
Editorial
c Committee of Appl. Math. Mech. and and Mechanics
Springer-Verlag 2008 (English Edition)

∗
Aerodynamic optimization of 3D wing based on iSIGHT

YIN Bo ( )1 , XU Dian ( 
)2 , AN Yi-ran ( Ë) , 2

CHEN Yao-song ( í)2

(1. Department of Engineering Mechanics, Tsinghua University,
Beijing 100084, P. R. China;
2. College of Engineering, Peking University, Beijing 100871, P. R. China)

(Communicated by DAI Shi-qiang)

Abstract A method for combining the CFD software, Fluent, with the iSIGHT design
platform is presented to optimize a three-dimensional wing to ameliorate its aerodynamics
performance. In the optimization design, two kinds of genetic algorithms, the Neighbor-
hood Cultivation Genetic Algorithm (NCGA) and the Non-dominated Sorting Genetic
Algorithm (NSGAII), are employed and the Navier-Stoke (N-S) equations are adopted to
derive the aerodynamics functions of the 3D wing. The aerodynamic performance of the
optimized wing has been significantly improved, which shows that the approach can be
extended and employed in other cases.
Key words aerodynamics computation, multi-objective optimization, genetic algo-
rithm, three-dimensional wing
Chinese Library Classification O354.1
2000 Mathematics Subject Classification 76G25

Introduction

With the booming development of computer science and computational mathematics, the
cost and design period in the aerodynamic design has been hugely diminished by employing
numerical optimization. The aerodynamic design[1,2] counts on the unification of numerical op-
timization strategy and aerodynamic computation, and the reliability, precision and efficiency
of the methods are consequentially dominant factors[3]. Traditional deterministic optimization
methods can only optimize locally, which is determined by its intrinsic characteristics. By con-
trast, the random optimization method with the genetic algorithm exhibits the representative
features not only in global optimization but also in robustness, reliability and portability, which
results in the extensive utilization of genetic algorithm in the field of engineering optimization[4] .
Wing design plays a pivotal role in the aerodynamic performance of the aircraft. Based on
the iSIGHT platform, the solver Fluent and preprocessor Gambit are combined to fulfill this
optimization.

∗ Received Sep. 3, 2007 / Revised Mar. 28, 2008

Corresponding author CHEN Yao-song, Professor, E-mail: chenys@pku.edu.cn
604 YIN Bo, XU Dian, AN Yi-ran, et al.

1 Optimization algorithm

Aerodynamic design based on the genetic algorithm[5] is an approach for aircraft design,
which is performed by a reproduction, crossover and mutation operator. Multi-objective op-
timizations are implemented based on the iSIGHT, while two types of genetic algorithms are
employed in the progress: the NCGA[6] and the NSGAII[7,8] . The iSIGHT provides an open
platform for integration, of which the progress interface is able to integrate various tools, such
as the commercial CAD software, manifold finite element analysis software and user-developed
procedures.

2 Computational progress
The computational progress is depicted as Fig. 1[1] .

Generate airfoil

Generate mesh by gambit

Compute by fluent
Controlled by
iSIGHT
No Renew the
Achieved?
parameters
Yes
Optimized wing

Fig. 1 Process of the optimization

3 Airfoil geometry

The airfoil geometry can be expressed by polynomial fitting or linear superposition of an-
alytic functions and so on. Using linear superposition of analytic functions, the shape of the
airfoil is represented according to the base airfoil, shape functions and the design variables[9] :
n

y(x) = y0 (x) + αi · fi (x), (1)
i=1

where x is position along the chord, y0 (x) is the upper and lower position of initial airfoil,
fi (x) is the shape bump function series, αi is the design variable, and then y(x) denotes the
coordinate vertical to the chord. In this work, the function series representation proposed by
Hichs and Hennes[9] is employed:
 ln 0.5 
fi (x, m, n) = sinn πx ln m . (2)

Then, y(x) is parameterized with variables αi to create serials of smooth airfoils. Five
parameters (α1 , α2 , α3 , α4 , α5 ) are considered to adjust the upper and lower surfaces, of which
the values from α1 to α4 correspond to m = 0.1, 0.3, 0.5, 0.7 to adjust the front and middle of
the airfoil with n = 3 and the value of α5 matches m = 0.9 for the rear part with n = 1. The
parameters αi (i = 1, 2, · · · , 10) are the design variables in the process of generating a random
smooth airfoil.
 Aerodynamic optimization of 3D wing based on iSIGHT 605

4 Fluid computation

4.1 Governing equations[10,11]

The N-S equations are taken as the leading governing equations to calculate the flow field,
of which the conservative form in the Cartesian coordinate is

∂U ∂F (U ) ∂G(U )
+ + = 0. (3)
∂t ∂x ∂y

The Sparlart-Allmara (S-A) model is employed as the turbulence model, with its empirical
constants as default in Fluent. Far-field pressure condition is applied as boundary condition
and a second-order upwind scheme is selected to discretize the governing equation.
4.2 Computation confirmation
In this subsection, a computation result concerning wing ONERA M6 is compared with the
experiment data under such circumstances: the Mach number M a = 0.84, the angle of attack
α = 3.06, the total grid number is 182 970 (Fig. 2).

Fig. 2 Contour of static pressure of M6

At 44% and 65% along the wingspan, the static pressure coefficient distributions of compu-
tation and experiment are presented in Fig. 3.

Experiment on upper wing Experiment on upper wing

Computation on upper wing Computation on upper wing
1.2 Experiment on lower wing 1.2 Experiment on lower wing
1.0 Computation on lower wing 1.0 Computation on lower wing
0.8 0.8
0.6 0.6
0.4 0.4
− Cp

− Cp

0.2 0.2
0.0 0.0
− 0.2 − 0.2
− 0.4 − 0.4
− 0.6 − 0.6
− 0.8 − 0.8
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
X X
(a) η = 44% (b) η = 65%

Fig. 3 Static pressure coefficients of computation and experiment

The computational result accords well to the experimental data, corroborating the selected
model and parameters.
606 YIN Bo, XU Dian, AN Yi-ran, et al.

5 Optimization example

Under such circumstances as M a = 0.6, α = 2, the wing with NACA65006 as its airfoil
is to be optimized constrained in the condition that the maximum thickness of the optimized
wing should be no less than the initial one and the genetic algorithm dominates the process to
ameliorate the Cj (the ratio of lift coefficient to drag coefficient), Cl (the lift coefficient) and
minify the Cd (the drag coefficient) as expected.
5.1 Generate grid
The initial front sweep is 35◦ , the aspect ratio 3.5, the taper ratio 0.17, and the airfoil section
NACA65006. The preprocessor gambit generates the grid with a total number of 172 940 (shown
in Fig. 4).

Fig. 4 Mesh of the wing

5.2 Optimization with neighborhood cultivation GA

Every generation has 50 individuals and 20 generations are developed. Table 1 shows the
enhancement. Figures 5 and 6 illustrate the pressure and Mach number distributions of the
initial wing shape, respectively.
An airfoil section of optimized wing I is represented in Fig. 7.

Table 1 Optimization of NCGA

Cl Δ/% Cd Δ/% Cl /Cd Δ/%

Initial wing 0.136 533 0.008 804 57 15.51

Optimized wing I 0.196 183 +43.69 0.010 306 +17.05 19.04 +22.76

Fig. 5 Contour of static pressure of the initial

 Aerodynamic optimization of 3D wing based on iSIGHT 607

Fig. 6 Contour of Mach number of the initial

0.08
0.07 Optimized
0.06 Initial
0.05
0.04
0.03
0.02
0.01
Y 0.00
− 0.01
− 0.02
− 0.03
− 0.04
− 0.05
− 0.06
− 0.07
− 0.08
0.0 0.2 0.4 0.6 0.8 1.0
X

Fig. 7 Optimized NCGA

Fig. 8 Contour of static pressure of optimized NCGA

As was illustrated in Figs. 8 and 9, the head of the optimized wing tends to be obtuse,
the middle becomes smoother with an insignificant thickness change and the upper face of
the tail has a subtle concave with a slight warp at the tip. The lower surface becomes flat
with smoothened curve and shock wave near the upper surface is inconspicuous. The overall
arching shape actually augments the effective angle of attack. The increase in Cl counteracts
the increase in Cd , which results in a significant enhancement of Cj . The optimization target
for the 3D wing is basically achieved.
608 YIN Bo, XU Dian, AN Yi-ran, et al.

Fig. 9 Contour of Mach number of optimized NCGA

5.3 Optimization with non-dominated sorting GA

Every generation has 50 individuals and 20 generations are developed. Table 2 displays the
improvement. A section airfoil of optimized wing II is shown in Fig. 10.

Table 2 Optimization of NCGA of NSGA-II

Cl Δ/% Cd Δ/% Cl /Cd Δ/%

Initial wing 0.136 533 0.008 804 57 15.51

Optimized wing II 0.215 111 6 +57.55 0.011 025 3 +25.22 19.510 7 +25.79

0.08
0.07 Initial
0.06 Optimized
0.05
0.04
0.03
0.02
0.01
Y 0.00
− 0.01
− 0.02
− 0.03
− 0.04
− 0.05
− 0.06
− 0.07
− 0.08
0.0 0.2 0.4 0.6 0.8 1.0
X

Fig. 10 Optimized NSGA-II

As showed in Figs. 11 and 12, the head becomes more acute after optimizing the NSGA-
II. The middle arches more with the thickest part locating slightly forward. The lower surface
almost reaches the level. The rear part spindles obviously with the tip almost unchanged. Shock
wave near the upper surface stays unobvious. Because the overall shape arches more than the
previous optimized wing, which results in greater increase of the effective angle of attack, and
the lift force enhances more. Besides the influence brought by the airfoil section the entire wing
geometry impacts the aerodynamics significantly, therefore the Cd hardly drops, however, Cj is
still improved.
 Aerodynamic optimization of 3D wing based on iSIGHT 609

Fig. 11 Contour of static pressure of optimized NSGA-II

Fig. 12 Contour of Mach number of optimized NSGA-II

6 Conclusions

It can be effectively feasible to improve the aerodynamic performance of the wing by using
the genetic algorithms to optimize the wing shape. Because of the global optimization nature
of genetic algorithm, the design can reach an optimal one within the search range, but com-
pared with the traditional optimization methods, the genetic algorithm suffers from excessive
computation and time consumption. The calculation time can be curtailed by modifying the
genetic algorithm or implementing similar engineering methods.
Thus, the optimization of fluid machinery design has focused on CFD solver and efficient
numerical methods, which are the two advanced research directions. Compared with fluid
numerical simulation, numerical optimization solver is relatively less sophisticated. For different
issues, the choice of optimization algorithms may greatly affect the result, for instance, two
genetic algorithms are chosen in this computation, and the CFD solver is used to determine
the comprehensive computation. With the iSIGHT a highly efficient optimization platform
combining with the fluent can be extended to a wider field of fluid equipment design based on
three-dimensional wing optimization.

References
[1] Holland J H. Genetic algorithms[J]. Scientific American, 1992, 267(1):44–50.
[2] Shinya W, Tomolyuki H, Mitsunori M. Neighborhood cultivation genetic algorithm for multi-
objective optimization problems[C]. In: Proceedings of the 4th Asia-Pacific Conference on SEAL,
2002:198–202.
[3] Besnard E, Schmitz A, Boscher E, et al. Two-dimensional aircraft high lift system design and
optimization[R]. AIAA 98-0123, 1998.
[4] Yang Xudong, Qiao Zhide, Zhu Bing. Subsonic and transonic wing inverse design using control
theory[J]. Acta Aerodynamics Sinica, 2003, 21(1):13–19 (in Chinese).
610 YIN Bo, XU Dian, AN Yi-ran, et al.

[5] Zhu Ziqiang, Fu Hongyan, Yu Rixin, et al. Multi-objective optimization design of airfoil and
wing[J]. Science in China (Series E), 2004, 33(11):999–1006 (in Chinese).
[6] Deb K, Geol T. Multi-objective evolutionary algorithms for engineering shape design[R]. KanGAL
Report No. 200003.
[7] Deb K, Pratap A, Agarwal S, et al. A fast and elitist multi-objective genetic algorithm: NSGA-
II[J]. IEEE Transactions on Evolutionary Computation, 2002, 6(2):182–197.
[8] Srinivas N, Deb K. Multi-objective function optimization using non-dominated sorting genetical
algorithms[J]. Evolutionary Computation, 1995, 2(3):221–248.
[9] Hicks R and Henne P. Wing design by numerical optimization[J]. Aircraft, 1978, 15(7):407–413.
[10] Anderson J D. Computational fluid dynamics the basics with applications[M]. Beijing: Tsinghua
University Press, 2002.
[11] Patanka S V. Numerical heat transfer and fluid flow[M]. Zhang Zheng (transl). Beijing: Science
Press, 1984 (Chinese version).