See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228519310

A Mathematical Formulation and Solution of the Optimized Wing Aero-

structural Problem

Article · January 2008

CITATION READS

1 1,747

3 authors:

Stephen Akwaboa Frederick Ferguson

Southern University and A&M College North Carolina Agricultural and Technical State University
27 PUBLICATIONS   51 CITATIONS    94 PUBLICATIONS   421 CITATIONS   

SEE PROFILE SEE PROFILE

Patrick F Mensah
Southern University and A&M College
65 PUBLICATIONS   461 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Heat Transfer Simulator View project

Waverider View project

All content following this page was uploaded by Patrick F Mensah on 31 May 2014.

The user has requested enhancement of the downloaded file.

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

A MATHEMATICAL FORMULATION AND SOLUTION OF

THE OPTIMIZED WING AERO-STRUCTURAL PROBLEM
Stephen Akwaboa1, Frederick Ferguson2, Patrick Mensah3
1
Post Doc Researcher, Southern University and A&M College, [email protected]
2
Professor, North Carolina A&T State University, [email protected]
3
Professor, Southern University and A&M College, [email protected]

Abstract: - In this research the design of a class of aircraft wing with a constant taper angle is formulated
as a constrained optimization problem. The design variables for the optimization problem are the
placements of the internal components and geometric dimensions of the wing. The cost function is the
weight of the wing consisting of the weights of the front spar, the rear spar and the reinforced skin. The
design constraints imposed on the optimization process are the bending stresses in the spars, the shear
stresses and the angle of twist in the ribs. All the design calculations are non-dimensionalized with a
corresponding calculation of a rectangular planform wing, which is a member of the class of wings
studied. The optimization of the dimensionless weight of the wing, defined in terms of ten design
variables, is implemented using Nelder and Mead technique. Simplified versions of the optimization
problem for 1D and 2D cases are presented to serve as the basis for the validation of the solution process
for the multivariate (10 variables) optimization problem. For the 1D case where only one (xdp = tip ratio)
of the ten variables was allowed to vary, at a tip ratio of xdp = 0.4, the lift coefficient of the finite wing is
maximum at 1.48 for the NACA 4412 wing section used for the construction of the wing. The total
percentage weight saving at that tip ratio is about 32% compared to the worst weight. For the simplified
2D case, the minimum weight of the NACA 4412 section wing with a constant taper ratio was sought in
the domain 0.1 ≤ x1 ≤ 0.3 and 0.5 ≤ x 2 ≤ 0.7 . Utilizing Solver, an Excel Optimization routine, the
optimized weight of the wing occurred at x1 = 0.3 and x2 = 0.65 with a total weight savings of about 20%
compared to the worst weight in the solution domain. With respect to x1 alone the weight saving was
about 13%, and with respect to x2 alone the weight saving was about 10%. It is expected that allowing the
three variables, xdp, x1 and x2, to vary, the total weight saving will be about 55%, a remarkable weight
reduction that requires an independent validation effort.

Keywords: - Optimization, Lift Coefficient, Airfoil, Spars, Ribs, Reinforced skin

1 Introduction coefficient of lift constraints [2, 3]. The other

Due to the dominating effects on the vehicle approach involves selecting a desirable lift
overall performance, aircraft weight plays a distribution and then computing the twist, taper,
significant role in its design. Statistically it’s and thickness distributions that are required to
been established that the amplification factor of achieve this distribution [4, 5]. The latter
any weight-carrying component is about 4.525, approach is generally used to obtain analytic
indicating a 1.0 Ibs reduction in the structural solutions and insight into the important aspects
weight signifies a 4.525 Ibs reduction in the of the design problem, but it is difficult to
gross aircraft takeoff weight [1]. The wing is a incorporate certain constraints and off-design
key component of an airplane. In fact the entire considerations. The former approach, often
performance of an airplane depends heavily on combined with numerical optimization, is used
the performance of the wing. Essentially, there more extensively in the design of the wing. The
are two standard approaches to the design of an direct method is the approach which is used in
aircraft wing. In the direct approach, one finds this research work.
the planform and the placement of the internal Past researchers have done some work
members of the wing that optimize some in the area of wing design. Most of these
combination of structural weight, drag, and researchers have designed the wing with the aim

ISSN: 1790-5095 74 ISBN: 978-960-474-106-9

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

of optimizing a figure of merit. For instance, a optimization solution method that uses the
multidisciplinary design optimization (MDO) objective function values to obtain the minimum
system for a minimum weight wing with nacelle of the objective function. The method makes no
and pylon was developed by Takayasu Kumanu assumption about the differentiability of the
et al. [6]. The MDO system was based on the objective functions and the constraint functions.
integration of Computational Fluid Dynamics
(CFD) codes and NASTRAN aeroelastic- 2 Design of Lightweight Wing
structural interface code. The multidisciplinary
design optimization of a strut-braced aircraft
Structures
The wing is a framework composed mainly of
wing (SBW) and its benefit relative to
spars, ribs and skin reinforced with stringers.
conventional cantilever wing configuration were
The spars, which run from the root to the tip of
studied by Gern et al. [7]. The study included
the wing, are the main members of the wing.
aerodynamics, structures, aeroelasticity and
The entire load carried by the wing is
synthesis of various disciplines. Shijun Guo [8]
predominantly borne by the spars. In flight, the
presented an investigation into a minimum
force of the air acts against the skin. From the
weight optimal design and aeroelastic tailoring
skin, this force is transmitted to the ribs and then
of an aerobatic aircraft wing structure. Lerner
to the spars. Most wing structures employ two
and Markowitz [9] and Lerner [10] presented
spars – the front spar and the rear spar. Since the
composite wing design studies with strength and
spars are designed to carry the bulk of normal
both static and dynamic aeroelastic constraints
stress due to bending, it follows that the two-
using control surface effectiveness, flexible
spar wing box employing stiffened wing skins
surface lift-curve slope and divergence speed.
will generally provide a low cost, easily
The studies included both forward and backward
manufactured, light-weight structure. The front
sweeping wings. A distinguishing feature of the
spar is located near the leading edge while the
studies was that they used finite elements of
rear spar is located about two-thirds of the chord
build-up wings with laminated composite skins.
distance. The ribs are the component of the wing
Bohlman et al. [11] made extensive strength and
which support the covering and provide the
aeroelastic design studies on full and scale
airfoil shape. These ribs are called forming ribs.
models of validation of aeroelastic tailoring
The externally applied load on the wing gives
(VAT) wings. Lockheed-Martin’s ELAPS
rise to internally applied bending moments and
program and ASTROS were used in these design
shear loads. This requires the internal
studies. An ELAPS structures model is a RITZ
components of the wing to be sized such that the
flat plate representation while the ASTROS
components can withstand the external loads
models were based on finite element methods.
imposed on them in flight.
In the present work, the design of an
Shown in Fig. 1 are the components of a
aircraft wing structure is posed as a constrained
typical wing box of Airbus A380 wing. From
optimization problem which can be solved by
the figure it can be seen that the main
any available optimization solution techniques.
components of the total weight of the wing are
Posing the design in this manner brings the
the weights of the ribs, reinforced skin, and the
preliminary wing design process to the forefront
spars. It must also be pointed out that the
of existing capabilities of computers. The weight
material type as well as the wing cross-section
of the aircraft wing which mainly consists of the
and the placements of the internal members of
weights of the spars, the ribs and the reinforced
the wing all have tremendous effects on the total
skin is the cost function of the optimization
weight of the wing.
problem. The constraints which must be satisfied
are the bending stresses in the spars, the shear
stresses in the webs and the skin, the angle of 2.1 Wing
twist in the ribs and the angle of deflection at For the design of the wing (Fig. 2), two spars
any section of the wing. The method of solution (front and rear spars) and five ribs spaced along
of the optimization problem is the Nelder and the spanwise direction of the wing and a class of
Mead solution technique, a non- gradient-based

ISSN: 1790-5095 75 ISBN: 978-960-474-106-9

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

wing with constant taper ratio having no angle indicating that the choice of wing section is also
of sweep is employed to perform the analysis. a design parameter. However in this work, a
constant wing section, NACA 4412, is used
which eliminates the dependence of the wing
section on the weight of the wing. The
minimization of the dimensionless weight of the
wing is an optimization problem which is posed
mathematically as:

Fig. 1 Main Components of a Wing Box

Structure

The design analysis of the wing is performed in

non-dimensional form by normalizing all
calculations of the constant taper wing (the
target wing) with the corresponding calculation
Fig. 2 Design Variables for the Target and
of a rectangular wing (the reference wing). The
rectangular wing is a member of the target wings NACA 4412

corresponding to specific values of the design

variable vector x. x1
x1
As shown in Figure 2, for a given material, the NACA 0010

weight of the wing depends on the airfoil section

x1
used for the construction of the wing and also on
the design variable vector x = [x1, x2,…, x10]t (a) (b)
where the superscript, ‘t’, denotes the transpose
of the design variable vector x. The elements of
Fig. 3 (a) Effect of Design Parameters on the
the design variable vector, x, represent the
Weight of the Wing (b) Effect of NACA
locations and the dimensions of the internal
Airfoil Section on the Weight of the Wing
members of the wing. These dimensions are
normalized with the root chord, CR of the wing. Minimize:
Thus, xi = x C R for i = 1, 2…, 10. Figure 3 . Wwing = Wwing (x) Wspars( x) + Wribs ( x) + Wskin (x) (1)
=
shows the dependence of the weight of the wing Wwing,ref ( xref ) Wspars( xref ) + Wribs ( xref ) + Wskin( xref )
on a design variable element, x1, and the type of Subject to:
wing section. From Figure 3a, it can be seen that
altering, for instance, the design variable x1, t ≤ t critical
which defines the location of the front spar of
. Max (σ k , τ k , θ k ) ≤ (σ , τ , θ ) critical (2)
the wing clearly affects the weight of the wing.
Thus the overall weight of the wing can be C l ≤ C l , critical
affected if all the elements of the design variable
vector, x = [x1, x2…, x10]t, are allowed to change In the design constraints, σ is the bending stress
within certain specified values. The elements of in the spars, τ is the shear stress in the ribs and
the design variable however have different skin, θ is the angle of twist per unit length at a
sensitivities to the overall weight of the wing.
section on the wing, t is the thickness of the
Similarly, using the cambered wing section,
web and Cl is the section lift coefficient. The
NACA 4412, instead of the symmetric wing
optimization is carried out with the explicit
section, NACA 0010, for the construction of the
constraints on the design variable vector as
wing, different wing cross-sections are produced

ISSN: 1790-5095 76 ISBN: 978-960-474-106-9

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

xl ≤ x ≤ xu where xl and xu are the lower and assumptions must be made about its geometry,
upper bounds of the design variables x and the material properties, and the way the wing is
vector xref corresponds to the design variable loaded and supported, so that established
vector x for the reference rectangular wing. In methods of engineering mechanics can be used
the process of optimizing the weight of the wing, to study the wing in the preliminary design
whilst satisfying the design constraints specified stage. For subsonic aircraft the wings, tail planes
in Equation 2, it is ensured that the structural and control surfaces consist of thin-walled
performance parameters (strength-to-weight cellular structure which may be idealized into a
ratio for bending and torsion which measure the combination of direct stress carrying booms and
efficiency of the wing structure) are also shear stress carrying skins and webs [12]. The
measured. These performance parameters are actual wing and the idealized representation of
defined as the wing consisting of these thin webs and
concentrated flange areas A1, A2, A3 and A4 are
shown in Figures 4a and 4b. For a completely
. S / W bending = M (3) idealized section the flange areas (booms) are
W
assumed to carry all the direct stresses and the
Non-dimensionalizing these structural skin and webs carry all the shear stresses and so
parameters with the corresponding parameters of the centroid of the wing section is thus the
the reference wing results in the following centroid of the boom areas. The coordinates of
equations for the non-dimensional structural the centroid, G(xG,yG), of the idealized wing are
performance parameters of the wing: thus calculated as:

i=4 i=4
. S ⎛ W wing , ref ⎞ ⎛ M wing
=⎜ ⎟⎜
⎞ M .
⎟= (4) . xG = ∑ Ai xi ∑A (7)
bending ⎜ W ⎟⎜ M ⎟ W i
⎝ wing ⎠⎝ wing , ref ⎠ i =1 i =1
i =4 i =4

2.2 Estimation of Wing Weight . y G = ∑ Ai y i ∑A i (8)

i =1 i =1
The total weight of the wing which is made up
of the weights of the ribs, the two spars and the
reinforced skin, can be expressed as

. W wing = W ribs + W spars + W skin (5)

where Wribs, Wspars and Wskin represent the total

weights of the ribs, the spars and the skin
respectively. Each component of the weight can Fig. 4 (a) Cross-section of Actual Wing
also be expressed as showing the Front and the Rear Spars (b) Cross-
section of an idealized wing showing virtual
. W i = ρ i L i Ai g (6) areas of spars

where i denotes the ribs, spars or skin, ρi is the where Ai (for i = 1,2,3, and 4) are the virtual
density of the material of component i of the areas of the spar caps attuned to account for the
wing, Li is a characteristic length, Ai is the bending areas of the stringers in the skin. These
cross-sectional area and g is the acceleration due virtual areas are given by [12]
to gravity. . A1 = A1,actual + t12s12 ⎛⎜ 2 + σ 2 ⎞⎟ + t14s14 ⎛⎜ 2 + σ 4 ⎞⎟ (9)
6 ⎜⎝ σ1 ⎟⎠ 6 ⎜⎝ σ1 ⎟⎠
2.3 Structural Idealization of the Wing .
The aircraft wing is an extremely complex t s ⎛ σ ⎞ t s ⎛ σ ⎞
redundant structure. In order to analyze the A2 = A2,actual + 12 12 ⎜⎜ 2 + 1 ⎟⎟ + 23 23 ⎜⎜ 2 + 3 ⎟⎟ (10)
6 ⎝ σ2 ⎠ 6 ⎝ σ2 ⎠
performance of the wing, simplifying

ISSN: 1790-5095 77 ISBN: 978-960-474-106-9

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

bending moments are also produced by the

t s ⎛ σ ⎞ t s ⎛ σ ⎞ aerodynamic forces due to pressure and shear
. A3 = A3,actual + 23 23 ⎜⎜ 2 + 2 ⎟⎟ + 34 34 ⎜⎜ 2 + 4 ⎟⎟ (11) stress acting on the surface of the wing. For
6 ⎝ σ3 ⎠ 6 ⎝ σ3 ⎠
subsonic flow, the contribution of shear stress to
.
the overall loading on the wing is negligible in
t34s34 ⎛ σ3 ⎞ t14s14 ⎛ σ1 ⎞
A4 = A4,actual + ⎜2 + ⎟ + ⎜2 + ⎟ . (12) comparison with the contribution of the loading
6 ⎜⎝ σ 4 ⎟⎠ 6 ⎜⎝ σ 4 ⎟⎠ due to pressure. Hence the shear stress is
neglected in the analysis of the external loading
The stress ratio σi/ σj (i,j = 1,2,3,4) defined in on the wing. In general, the axial bending stress,
Equations 9 to 12 is equal to -1 for bending σz, for the wing, under bi-directional bending
loads and +1 for axial loads. For example, in due to the internal resultant bending moments
calculating the virtual area A1, σ2/σ1 = -1, but Mx and My can be evaluated using the
the ratio σ4/σ1 = h4/h1, where h1 and h4 are the expression [12]
distances between the centroid of the boom areas
and the neutral axis (N-N) in the y-direction; t12 . σ = ⎛⎜ IGxMy − IGxyMx ⎞⎟(x − x ) +⎛⎜ IGyMx − IGxyMy ⎞⎟( y − y ) . (17)
is the thickness of the web or skin between z ⎜ I I −I2 ⎟ G ⎜ I I −I2 ⎟ G
⎝ Gx Gy Gxy ⎠ ⎝ Gx Gy Gxy⎠
boom areas A1 and A2 whilst s12 is the arc length
between the boom areas A1 and A2; A1,actual is The coupling of aerodynamics and structures is
the actual cross sectional area of a spar cap etc. clearly evident in equation 17. Whilst the
The second moment of areas of the spar caps moments Mx and My are determined primarily
with reference to a set of orthogonal axes from the pressure distribution, the section
through the centroid G(xG,yG) at any cross- properties Ix, Iy and Ixy are determined from the
section of the wing are also given by geometry of the structure.
4
. I Gx = ∑ Ai ( yi − y G ) 2 (13) 2.5 Shear Stress and Angle of twist
1 The shear stress in the skin and the webs of the
4
. I Gy = ∑ Ai ( x i − xG ) .
2
(14) idealized wing section are determined using the
1 theory of thin shells [12]. The idealized wing
4
I Gxy = ∑ Ai ( xi − xG )( yi − yG ) . section consisting of three closed cells is
. (15)
1 subjected to an external torque T0, obtainable
from the pressure distribution on the wing. By
The orientation of the neutral axis, NN, at any their nature, the ribs of the wing are designed to
section along the span of the idealized wing is withstand torsional loads. If the ribs are assumed
defined by the angle α measured in the to be thin shells, then the shear flow in the skin
clockwise direction about the centroid of the of the ribs can be taken to be constant. For a thin
cross section, G, and is given by shell of thickness t, τt = q, where q is a constant
. called the shear flow and τ is the shear stress in
⎛ I Gx M y − I Gxy M x ⎞ the skin of the shell. Furthermore, the algebraic
α = Tan −1 ⎜ ⎟. (16) sum of shear flows at the location where two or
⎜ M x I Gy − M y I Gxy ⎟
⎝ ⎠ more thin shells converge is equal to zero
(conservation of shear flow). Hence from Fig. 5,
2.4 Direct Stress in the Spars at the location of the virtual area A1 (designated
The wing, being a non-uniform beam, should, in by 1),
motion, have the ability to twist and bend at the
same time under the action of an externally
applied aerodynamic load. The main members in
the wing responsible for its flexural rigidity
(bending resistance) are the spanwise spars. The
direct bending stress distribution in the wing is
produced by the moments Mx and My. These

ISSN: 1790-5095 78 ISBN: 978-960-474-106-9

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

τ1… τ5 and θ at every section along the span of

the wing.

2.6 Estimation of Mx and My

To determine the quantities Mx and My which
are crucial to the structural analysis, the shear
forces Sx(z), Sy(z) and the bending moments
Mx(z) and My(z) must first be evaluated from the
Fig. 5 An Idealized Rib of the Wing of unit pressure distribution, assumed known. The shear
thickness force at any z location from the root of the wing
is obtained by the integration of the pressure
.. τ 1t1 = τ 2 t 2 + τ 4 t 4 . (18) distribution over the area ABCD (Fig. 6). Thus,
in the y-direction, the shear force at a distance z
Similarly, at the location of virtual area A4 from the root of the wing is given by

. τ 2t 2 = τ 3t3 + τ 5t5 . (19) S y ( z) = ∫ p( x, z)dxdz = p ave, ABCD AABCD . (23)

Area ABCD
According to Newton’s third law of motion, the
torque associated with the internally generated The bending moment in the x-direction at the
shearing stresses must resist the externally same section of the wing can also be expressed
applied torque, T0. Hence, as

. T0 = 2(A1τ 1t1 + A2τ 2 t 2 + A3τ 3 t 3 ) , (20) . M x ( z ) = p ave , ABCD A ABCD z ABCD , (24)

where A1 , A 2 and A 3 are the medianal areas of where z ABCD is a moment arm which, for the
the three cells. The area, A1234, in Fig. 5 wing planform shown in Fig. 6, is defined as
represents the medianal area ( A2 ) of cell2 of the
wing. The three cells at any rib location must 1 (C + 2Ct )(Span − z )
. z ABCD = . (25)
undergo the same angular rotation if the whole 3 C + Ct
wing must rotate as a unit about a reference line.
This requirement (compatibility requirement), The local chord of the wing, C(z), and the area
according to Breth’s equation [13], can be AABCD are respectively given , by
expressed mathematically as

1 . C = CR −
z
(C R − C t ) (26)
θi = ∫ τ ds = constant . (21) Span
2G A Cell i
. AABCD =
1
(Span − z )(C + Ct ) . (27)
2
Performing the line integral in Equation 21 for
the three cells results in the following set of The projected area of the wing in the x-direction
equations: is negligible compared to the planform area.
Hence the shear force Sx(z) can be ignored.
τ 1 s1 + τ 4 s 4 = 2 G θ A1 Therefore, to a very reasonable degree of
− τ 4 s 4 + 2τ 2 s 2 + τ 5 s 5 = 2 G θ A2 (22) accuracy, the bending moment My(z) can also be
neglected.
− τ 5 s 5 + 2τ 3 s 3 = 2 G θ A3

Equations 18, 19, 20 and 22 form a system of six

linear equations in six unknowns which can be
solved simultaneously for the unknown variables

ISSN: 1790-5095 79 ISBN: 978-960-474-106-9

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

2.8 Simplified 2D model of the Wing

Shown in Fig. 8 is the cross-section of the
simplified 2D engineering model of the wing. In
this model, the following assumptions are made:
• All but two of the ten design variables
(i.e., x1 which represents the location of
the front spar and x2 which represents the
location of the rear spar) are fixed.
Fig. 6 Wing Planform Showing the Moment • The materials for the front spar and rear
Arm for Mx and My Evaluations spars are assumed to be the same.
• The widths of the front and rear spars are
2.7 Simplified 1D model of the Wing maintained the same.
In this model (Fig. 7), among the ten
x3
components of the design variable vector, x, x4
1
only the component x6 is allowed to vary. To y 4
tskin
further simplify the model, the leading edge tFS tRS

taper angle of the wing is kept constant at zero 3

x
degrees, whilst the trailing edge angle is allowed x1
2

to vary through the parameter Ct, the tip chord x2

x5 = 1.0
length. The dimensionless design parameter is
xdp = Ct/CR. The front and rear spars are placed
at 19% and 70% from the leading edge Fig. 8 Typical Wing Section for the Simplified
respectively. For this 1D engineering model, the 2D Engineering Model of the Wing
general aero-structural problem as described in
Equation 1 reduces to the following: Now, the weight of the wing can be written as

Minimize: . Wwing = Wribs + Wspars + Wskin . (30)

Wwing ( xdp )
Wwing = = w( x xp ) (28) Under the assumptions listed above, the weight
Wwing,ref ( xdp = xref ) of the wing can be approximated as

Subject to: . Wwing = ρ sparsgLsparsAspars + W0 , (31)

⎧Geometric : t ≤t where W0 is the constant total weight of the ribs

. ⎪⎨Structural : max[(σ k , τ k )i ] ≤ (σ , τ )critical ,i . (29)
web critical
and reinforced skin, Aspars is the total cross-
⎪ sectional area of the front and rear spars, Lspars is
⎩ Aerodynamic : C l ≤ C l ,critical
the span of the wing, ρspars is the material density
X
of the spars and g is the acceleration due to
P
TRAILING
EDGE
gravity. From Equation 31, since ρspars, g, Lspars
Ct x dp
A
D
and W0 are constants, to minimize the weight of
CR C(Z)
TIP
the wing, the total cross-sectional area of the
ROOT
spars can as well be minimized. Hence the
B
LEADING EDGE
C Z
minimization of the weight of the wing is posed
P
Z as:
Semi Span = b

Minimize:
Fig. 7 Wing Planform of the Simplified 1D
Engineering Model of the Wing

ISSN: 1790-5095 80 ISBN: 978-960-474-106-9

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

A spars ( x 1 , x 2 ) solution for the optimization of the 1D

. A (x) = = A ( x1 , x 2 ) (32) engineering model of the wing design occurs at
A spars ( x = x ref )
xdp = 0.4, since this is the value of the design
variable for which the lift coefficient for the
Subject to: wing section is maximum (i.e., Clmax= 1.48).
t FS = α FS ( y FS ,U − y FS , L ) ≥ t FS ,critical
According to [14], for NACA 4412 wing
section, the maximum coefficient of lift is about
. t RS = α RS ( y RS ,U − y RS , L ) ≥ t RS ,critical (33) 1.52. This value is very close to the Clmax value
t RS obtained from the graph of CL against xdp which
tˆ = ≥ tˆcritical in a way validates the 1D problem results.
t FS
1.60
The ordinates of the upper and lower surfaces of 1.40

Coefficient of Lift (CL)

the airfoil section are denoted by yU and yL in 1.20

equation 33. The subscripts FS and RS indicate 1.00

0.80
the front spar and rear spar respectively. The
0.60
parameters for the web thickness of the front
0.40
spar, αFS, and the rear spar, αRS, were chosen to 0.20
be the same. Furthermore, the web thicknesses 0.00
were defined in a fashion specified in Equation 0.00 0.20 0.40 0.60 0.80 1.00
xdp
33 to allow for the influence of the design
variables on the thicknesses of the web. The
thickness ratio, tˆ , was included in the Fig. 9 A Plot of CL as a Function of xdp
constraints equations to ensure that the front spar  
web thickness, tFS, and the rear spar web Fig. 10 shows the plot of the strength-to-weight
thickness, tRS, could be controlled by the ratio and the dimensionless weight of the wing
designer with a single design parameter. The as functions of the design variable. It can be
following design parameters were used for the deduced from the plot that as the design variable
solution of the optimization problem defined in xdp increases, the mass drops. However the
Equations 32 and 33: strength-to-weight ratio increases, though
marginally. The saving in mass and the gain in
α FS = α RS = 0 . 15 strength-to-weight (S/W) ratio at xdp = 0.4 are
approximately 32% and 13% respectively.
. t FS , cri = t RS , cri = 0 . 01 . ( 34)
tˆcri = 0 .5

The optimization problem was solved within the

following explicit values of the design
parameters: 0 . 10 ≤ x 1 ≤ 0 . 30 and
0 . 50 ≤ x 2 ≤ 0 . 70 . These nondimensional
numbers for the locations of the front and rear
spars were chosen to produce a more realistic
geometry for the wing.  

Fig. 10 A Plot of Mass ratio and Strength-to-

3 Results and Discussion Weight Ratio as a Function of xdp
3.1 Simplified 1D Engineering Model of
the Wing 3.2 Simplified 2D Engineering Model of
Fig. 9 shows the plot of the lift coefficient, CL, the Wing
of the finite wing as a function of the design When the optimization problem for the 2D
parameter xdp using lifting line theory. The engineering model was solved, the minimum

ISSN: 1790-5095 81 ISBN: 978-960-474-106-9

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

weight of the wing was minimum at Amin = solution occurred at x1 = 0.3 and x 2 = 0.65 for
0.813 when x1 = 0.30 and x2 = 0.65. To validate which the total non-dimensional total cross-
the, solution, the field for the cost function was sectional area of the spars was 0.0141. All the
manually constructed in the domain of the entries in Table 1 were obtained by normalizing
design variables ( 0.0 ≤ x1 ≤ 0.3 and the cost function values by the worst solution of
0.5 ≤ x 2 ≤ 0.7 ) using Excel. Table 1 shows the cost function. By allowing both x1 and x2 to
the cost function values for various values of the vary, the total influence on the weight of the
design variables within the solution domain. wing was about 20% (Table 2). Thus the
combined influence of
Table 1 Dimensionless Total Areas of the
Spars as a Function of x1 and x2
x1 x2 = 0.50 0.55 0.60 0.65 0.70

0.10 1.000 0.984 0.961 0.933 0.901

0.15 0.956 0.939 0.917 0.888 0.857

0.20 0.922 0.906 0.883 0.855 0.824

0.25 0.895 0.879 0.856 0.828 0.796

0.30 0.869 0.852 0.829 0.813 0.770

Fig. 11 Contour Plot of Dimensionless Total

Fig. 11 is the contour plot of the total area of the Cross-sectional Areas of the Spars
spars, as functions of x1 and x2. From the
contour plot it can be seen that the solution to x1 and x2 on the cost function is more than the
the optimization problem is not single-valued as individual influence of x1 or x2 on the cost
Excel depicted. The minimum value of the cost function.
function ( Amin = 0.813) occurred for various
combinations of x1 and x2 values. These various Table 2 Influence of x1 and x2 on the Total
combinations of x1 and x2 for which the total Cross-Sectional Areas of the Spars
cross-sectional area of the spars is a minimum at Relative %
x1 and x2
Relative %
x2 = fixed Savings in x1 = fixed Savings in
0.813 gives an indication of the non-linearity of Mass
vary
Mass

the problem that was solved. The area ratios 0.50 13.14 0.10 9.87

0.796 and 0.770 (highlighted red in Table 1) are 0.55 13.36 0.15 10.32
each less than the optimized area ratio 0.813.
0.60 13.67 0.20 10.70
However, the optimizer did not accept any of
these numbers as the minimum because they are 0.65 14.09 0.25 11.02

non-feasible solutions for the optimization 0.70 14.58 0.30 11.36

problem. The effects of the design variables x1 Average

Savings in 13.77 19.88 10.65
and x2 on the total cross-sectional area of the Mass

spars and hence the total weight of the wing are

presented in Table 2. From the table it can be 4 Conclusions
seen that the influence of x1 alone on the weight A mathematical formulation and solution for the
of the wing is about 14% whilst that of x2 alone preliminary design of wing structures for
is about 11% compared to the worst solution. minimum weight was presented. The design
The worst solution is defined as the set of values process was posed as an optimization problem
of the design variables for which the total cross- which lends itself to any available optimization
sectional area of the front and rear spars is a solution technique. Ten design variables were
maximum. Within the rectangular domain used to describe the family of wings that was
0.0 ≤ x1 ≤ 0.3 and 0.5 ≤ x2 ≤ 0.7 , the worst designed. The solution of the 1D simplified

ISSN: 1790-5095 82 ISBN: 978-960-474-106-9

 Proceedings of the 7th IASME/WSEAS International Conference on FLUID MECHANICS and AERODYNAMICS

engineering model of the wing indicated that a [6] Michimasa Fujino, Design and Development
constant taper wing, with straight leading edge of the HondaJet, International Air and Space
and having a tip ratio of 0.4, gives the best Symposium and Exposition: The Next 100
aerodynamic performance. At this tip ratio, the Years, Dayton, OH, July 14-17, 2003.
saving in mass of the wing was about 32 % [7] Takayasu Kumano , Shinkyu Jeong, Shigeru
whilst the strength-to-weight ratio, a parameter Obayashi, Yasushi Ito, Keita Hatanaka, and
used to measure the efficiency of beam Hiroyuki Morino, Multidisciplinary Design
structures, was about 13%. For the solution of Optimization Of Wing Shape With Nacelle And
the 2D engineering model, the design parameters Pylon, European Conference on Computational
used were the locations of the front spar, x1, and Fluid Dynamics, 2006.
the location of the rear spar, x2. Studies [8] Gern F. H., Gundlach J. F., Tetrault P. A.,
performed on the singular effects of these design Nagshineh-Pour A., Multidisciplinary Design
parameters on the mass of the wing showed that, Optimizationof a Strut-Braced WingTransonic
with x1 fixed, saving in mass due to x2 was about Transport, AIAA, Jan 2000.
11% and with x2 fixed, the saving in mass due to [9] Shijun Guo, Aeroelastic Optimization of an
x1 was about 14%. The combined effect of x1 Aerobatic Wing Structure, Aerospace Science
and x2 on the saving in weight of the wing was and Technology, Jan 2007.
about 20%. It is expected that allowing the three [10] Lerner E. and Markowitz J., An Efficient
variables xdp, x1 and x2 to vary, the total weight Structural Resizing Procedure for Meeting Staic
saving will be about 55%, an impressive weight Aeroelastic Design Objective, AIAA Journal of
reduction that requires an independent validation Aircraft, Volume 16, pages 65-71, 1979.
effort. The weight saving could be used to fetch [11] Lerner E., The Application of Practical
more payload. The weight saving also means Optimization Techniques in the Preliminary
that less amount of fuel is needed to power the Structural Design of a Forward-Sept Wing,
vehicle in flight and hence a reduction in the Proceedings: 2nd. International Symposium On
cost of running the aircraft. Aeroelasticity and Structural Dynamics,
Technology University of Aachen, Germany,
References: 1985.
[1] Ampofo J., F. Ferguson, “Optimal design of [12] Bruce K. Donaldson, Analysis of Aircraft
aircraft wing structures: a computer aided design Structures: An Introductions, McGraw-Hill
method”, World Automation Congress, 2002. Inc.,1993.
[2] Jeffrey C. Lagarias, James A. Reeds, [13] Yurkovich R., The use of Taguchi
Margaret H. Wright and Paul E. Wright, Techniques with the ASTROS Code for
Convergence Properties of the Nelder and Mead Optimum Wing Structural Design, Proceeding:
Simplex Method in Low Dimensions, Society of 35th. AIAA SDM Conference, 1994.
Industrial and Applied Mathematics (SIAM) [14] Ira H. Abbot and Albert E. Von Doenhoff,
Journal, Volume 9, No. 1, pp 112-147, 1998. Theory of Wing Sections, General Publishing
[3] Michimasa Fujino, Natural Laminar Flow Company Limited, 1959.
Airfoil Development for a Lightweight Business
Jet, AIAA 20th Applied Aerodynamics
Conference , St. Louis, Missouri, June 24, 2002.
[4] Wakayama. S.R., Lifting Surface Design
Using Multidisciplinary Optimization, PhD
Dissertation, Stanford University, Stanford, CA,
December 1994.
[5] Tischler V. A., and Venkayya V. B., Design
Optimization of Airfram Structures, Journal of
Aerospace Engineering, Jul1997.

ISSN: 1790-5095 83 ISBN: 978-960-474-106-9

View publication stats