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

net/publication/268685154

Reliability Based Multi-objective Design Optimization of an Aircraft Wing

Structure with Abstract Optimization Variables

Conference Paper · August 2009

CITATIONS READS

0 545

4 authors, including:

Melike Nikbay Ahmet Aysan

Istanbul Technical University Istanbul Technical University
91 PUBLICATIONS   644 CITATIONS    5 PUBLICATIONS   43 CITATIONS   

SEE PROFILE SEE PROFILE

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

Development of a Nonlinear Sonic Boom Prediction Software View project

Commercial Supersonic Technology View project

All content following this page was uploaded by Melike Nikbay on 29 December 2014.

The user has requested enhancement of the downloaded file.

5. ANKARA INTERNATIONAL AEROSPACE CONFERENCE AIAC-2009-115
17-19 August, 2009 - METU, Ankara - TURKEY

RELIABILITY BASED MULTI-OBJECTIVE DESIGN OPTIMIZATION OF AN AIRCRAFT WING

STRUCTURE WITH ABSTRACT OPTIMIZATION VARIABLES

Melike Nikbay∗, Çağrı Ulucenk†,

Arda Yanangönül‡ and Ahmet Aysan§
Istanbul Technical University
Istanbul, Turkey

ABSTRACT
We consider reliability-based multi-objective design optimization of a simple aircraft wing structure with
stochastic parameters. The wing has a NACA0012 airfoil profile and consists of isotropic skin panels, ribs,
spars, stringers, and stiffeners made of aluminium. The aim of this study is to optimize the thicknesses of
these structural members and positioning of selected ribs and spars by employing both commercial engineer-
ing software and an in-house code developed for reliability analysis. The optimization criteria will include
both deterministic and probabilistic constraints. To compute the probability of failure for the probabilistic
constraints, one of the well known MPP (Most Probable Point) based reliability analysis methods, FORM
(First Order Reliability Method) will be employed. For MPP search, Performance Measure Approach (PMA)
is implemented. Catia V5-R16 is used as a parametric 3D solid modeler whereas Abaqus 6.7.1, structural
finite element method solver, is used to compute the structural response of the wing system under specified
loading. The finite element model is constructed by using shell and beam elements. As a multi-objective and
multi-disciplinary optimization driver, a commercial software, Modefrontier 4.1, is used for its gradient based
optimization algorithm called NLPQL which is based on sequential quadratic programming method for solving
nonlinear constrained optimization problems with differentiable objective and constraint functions. Pareto
optimal solution of the multiobjective problem is computed by the vector optimization methods of Modefron-
tier software. The core of the problem leans on the structural optimization of a statically loaded wing and
optimization of modal frequencies and deflections of that wing. The optimization criteria will make use of
mass, fundamental frequency, maximum deflection and maximum stress of the structure. The uncertainties
concerning both the yield strength and Young’s Modulus of the aluminium material will be analyzed by the
reliability analysis code developed for this study. The design variables are chosen as the thicknesses of all
structural members and geometric positions of selected rib and spar members. Abstract optimization variables
are introduced to reduce the number of optimization variables which will still be enough to relate the full set
of design variables to the optimization criteria and update the geometry.

INTRODUCTION
In recent years, design optimization of complex aircraft structures for maximum performance and minimum
cost has been a challenging research area for aircraft manufacturer companies. The multi-objective task of
attaining minimum weight and cost with maximum reliability of structure is one of the most time-consuming
phases of an aircraft design project. Therefore, robust computational methodologies are strongly required in
order to increase the efficiency and success of this design phase. A strong and easy-to-apply methodology can
∗ AssistantProfessor in Faculty of Aeronautics and Astronautics, Email: nikbay@itu.edu.tr
† Graduate Student in Informatics Institute, Email: cagriulu@be.itu.edu.tr
‡ Graduate Student in Faculty of Aeronautics and Astronautics, Email: yanangonul@itu.edu.tr
§ Graduate Student in Faculty of Aeronautics and Astronautics, Email: aysanah@itu.edu.tr
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

be developed by implementing the numerical optimization techniques directly into the everyday used analysis
tools that have been well and commonly employed in aerospace engineering. Numerical optimization is an
iterative scheme to reach the most desired design within a design space bounded by the lower and upper
limits of optimization variables. The design criteria defined as functions of optimization variables have to be
evaluated at each optimization iteration as variables are updated. Thus, optimization studies require a high
number of sequential analyses automatically and needs longer computational time as compared to only analysis
studies. For that reason, a serious research is focusing on developing more efficient optimization algorithms
for problems with large analysis size. For only optimization purposes simpler analysis models can be preferred
in the iterative process and parametric geometries can be used to reduce the number of optimization variables
that can sufficiently describe a problem.

In literature, for the structural analysis component of aircraft wing multi-disiplinary optimization, mostly
simple structural models are employed. Bowman et al. [31] used finite element analysis with bars, panels, and
membranes to model the wing as a rectangular prism with a constant chord. In order to model a helicopter
rotor blade, Friedmann [28] used finite element analysis for thin walled, rectangular box sections representing
the structural member at each span-wise station. Barthelemy et al. [25] and Dovi et al. [22] used an equiv-
alent laminated plate formulation to model aircraft wings by the trapezoidal plates. Jha and Chattopadhyay
[18] modeled the wing configuration of a high speed business jet with a rectangular box beam with taper
and sweep and used a panel code as the flow solver. Meanwhile, the validity of the results obtained by us-
ing these simple models have been increasingly questioned as advanced structural analysis methods have been
developed. Tzong et al. [26] and Maute et al. [13] used high fidelity structural codes for coupled optimization.

Nevertheless, most of the designs which are designed following a deterministic optimization process have
generally a high probability of failure due to the production phase. Since that process does not take uncer-
tainties into account, the results often lead to unreliable designs. There are two categories of methodologies
handling uncertainties in engineering design: reliability based design and robust design. An optimization
process that accounts for feasibility under uncertainty is commonly referred to as reliability based design opti-
mization (RBDO). RBDO ensures the design is feasible regardless of the variations of the design variables and
parameters. Robust design focuses on minimizing the variance of the design outcome under the variations of
design variables and parameters. In this work, RBDO is used instead of deterministic optimization to be able
to model uncertainties.

Reliability analysis and optimization are two essential components of RBDO: (1) Reliability Analysis focuses
on analyzing the probabilistic constraints to ensure the reliability levels are satisfied; (2) Optimization is seek-
ing for the optimal performance subject to the probabilistic constraints. Extensive research has been done
to explore various efficient reliability analysis techniques including sensitivity based approximation approaches
[29, 27], MPP (most probable point) based approaches [32, 24, 33], Monte Carlo Simulations (MCS) and
Response Surface Model based approaches [21, 23, 17]. Among those, MPP based approaches have attracted
more attention as they require relatively less computational effort while still producing results with acceptable
accuracy compared to the other three approaches [3, 12]. Another research issue in RBDO is to investigate the
integration of reliability analysis and optimization, using nested double-loop strategy or decoupled double-loop
strategy. Nested double-loop methods treat the reliability analysis as the inner loop analyzing the probabilistic
constraint satisfaction given the solutions provided by the outer optimizer which locates the optimal solution
iteratively. As a result, nested double-loop methods are computationally expensive for a complex engineering
design [10, 3, 5]. Therefore, decoupled double-loop methods have been developed to address the computa-
tional challenges [3, 5, 30, 15, 20, 16, 11].

With the advances in computer technology, the numerical optimization tools are well developed and ap-
plied to both single discipline and multi-disiplinary engineering problems by academic researchers in the last
decades. However, the numerical optimization applications in the industry do not emerge very fast since opti-
mization requires a deep understanding of the original model and the analysis tools and practices on validated
problems in order to combine analysis tools with an optimization driver. Moreover, an insight in the field of
optimization theory is required to formulate the optimization problem correctly and efficiently and choose the
most appropriate solution algorithm. So, coupling of commercial codes efficiently for optimization purposes
still remains a challenging task for product development engineers in the industry. In a previous work [1],
a single discipline optimization problem on a generic three dimensional wing geometry by employing Catia
and Abaqus as two of the most commonly used structural engineering tools for computer aided engineering
in aerospace industry was evaluated. A practical optimization methodology was created as a commercial op-
2
Ankara International Aerospace Conference
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

timization software, Modefrontier was coupled by this finite element based framework for its gradient-based
optimization algorithm options. Three similar but distinct optimization problems were investigated. The first
case leant on the structural optimization of a statically loaded wing where as the second case leant on the
optimization of modal frequencies and deflections of that wing. Finally, third case was a combination of both
the first and the second cases previously mentioned. The optimization criteria made use of mass, fundamental
frequency, maximum deflection and maximum stress of the structure. The design variables were chosen as
the thicknesses of all structural members and geometric positions of selected rib and spar members. Abstract
optimization variables were introduced to reduce the number of optimization variables which were still enough
to relate the full set of design variables to the optimization criteria to update the geometry. Additionally, in
this paper, the aforementioned third case is solved with variables Young Modulus E and yield strength σyield
of the material as treated to be random. Consequently, the constraints concerning stress, displacement and
frequency become probabilistic constraints.

AIRCRAFT WING DESIGN MODEL

A simple aircraft wing which has a NACA0012 airfoil profile is modeled parametrically in Catia V5-R16. The
wing’s three dimensional geometric model consists of 90 skin panels, 10 ribs and 4 spars while some of the
skin panels are stiffened by stringers along the wing span. The wing has a rectangular planform with 6m
semi-span and 1.6m chord length. Finite element model of the wing is prepared at Abaqus 6.7.1 and is
composed of linear shell and beam elements. The model is shown in figure 1, and consists of 17,070 linear
quadrilateral elements of shell type, 1264 linear line elements of beam type, total element number of 18,334
and 16,024 nodes, thus 96,144 degrees of freedom. In all members of the structure, aluminium is employed
with Young’s modulus E = 70000M P a, Poisson ratio ν = 0.33, density ρ = 2700kg/m3 , yield strength
σyield = 400M P a. As a cantilevered boundary condition, all of the degrees of freedom at the root of the
wing are set to zero. The aerodynamic load that will be applied to the wing is supplied from a computational
fluid dynamics (CFD) analysis performed for the initial design. An Euler inviscid flow analysis by using Fluent
commercial software was performed for M ach = 0.3 at sea level. For the sake of simplicity, the obtained total
lift force of approximately 25000N is then expressed as an elliptic lift function which changes along the wing
span but assumed to be constant along the chord [1].

Figure 1: Computational model of the wing structure

DETERMINISTIC OPTIMIZATION OF THE AIRCRAFT WING

A generic optimization problem associated with a given system can be formulated as:

min z(s) = min{z1 (s), z2 (s), ......, znz (s)} (1)

sǫS sǫS

h(s) = 0 h(s) ∈ IRnh (2)

ng
g(s) ≥ 0 g(s) ∈ IR (3)
S = {s ∈ IRns |sL ≤ s ≤ sU } (4)

3
Ankara International Aerospace Conference
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

where s is a set of abstract parameters restricted by lower and upper bounds sL and sU , z(s) is a cost function
of interest, h(s) denotes a set of equality constraints, g(s) is a set of inequality constraints. For compu-
tational simplicity, the physical design parameters can be defined as independent functions of some abstract
optimization variables. In this work, the optimization variables will cover structural parameters such as cross
sectional and thickness dimensions of the structural elements and some shape parameters. The optimization
criteria can cover the structural behavior descriptors such as mass, displacements, stresses, strains, and modal
frequencies.

The deterministic optimization problem that will be solved has two objectives as minimization of weight and
maximization of the first modal frequency of the structure while constraining maximum Von Mises stress with
the yield strength of the material. A factor of safety is not used on the stress constraint in the deterministic
optimization.
min M (s), max f1 (s) (5)
sǫS sǫS
σyield
g1 (s) = σmax (s) −1 ≥0 g1 (s) ∈ IR (6)
u0
g2 (s) = umax (s) −1 ≥0 g2 (s) ∈ IR (7)
0
f1
g3 (s) = 1 − f (s)
≥0 g3 (s) ∈ IR (8)
M0
g4 (s) = M (s) −1 ≥0 g4 (s) ∈ IR (9)
where M (s) is the total mass, umax (s) and σmax (s) are the maximum displacement and maximum Von
Mises stress of the wing structure. u0 = 187mm and M0 = 330kg are chosen as reference values from the
reference wing to constrain the displacement and mass. f1 (s) is the first natural frequency of the structure,
while f1 0 = 4.35Hz is the first natural frequency of the reference wing.

DEFINITION OF OPTIMIZATION VARIABLES

Since ribs, spars and skin panels are modeled as shell elements, the thicknesses of these elements and the
diameter of the stringers are chosen as design parameters. The thicknesses of spars, ribs and skin panels are
divided into three groups along the wing span, introducing 9 design variables. The outer diameter of all the
stringers are kept constant along the span and expressed as only one design parameter while the wall thickness
of the stringers are taken as one over third of the outer diameter. In figure 1, the structural components of
the wing and the thickness parameters related to these components are presented so that each different color
shows a different design parameter.

The computational time that will be spent for optimization will be shortened if the number of optimization
variables that will be used in the optimization loop can be reduced by using abstract optimization variables.
Therefore, four abstract optimization variables k1 ,k2 ,k3 ,k4 are used to describe 9 design variables related to the
thicknesses of all spars, ribs and stringers. The relation between design parameters and abstract optimization
variables are as follows;

tA1 = k1 k2 k3 t̃A1 tA2 = k2 k3 k4 t̃A2 tA3 = k3 k4 k1 t̃A3 (10)

where, tA1 , tA2 , tA3 are the physical design variables describing the skin panel thicknesses for the three
partitions along the span. tA1 is chosen to be on the cantilevered side. t̃A1 , t̃A2 , t̃A3 are the reference values
for the thicknesses of the three type skin panels which are dictated in the initial wing design. Similarly;

tB1 = k1 k2 k3 t̃B1 tB2 = k2 k3 k4 t̃B2 tB3 = k3 k4 k1 t̃B3 (11)

where, tB1 , tB2 , tB3 are the physical design variables describing the spar thicknesses for the three partitions
along the span. tB1 is chosen to be on the cantilevered side. t̃B1 , t̃B2 , t̃B3 are the reference values for the
thicknesses of the three spar partitions which are dictated in the initial wing design. Finally;

tC1 = k1 k2 k3 t̃C1 tC2 = k2 k3 k4 t̃C2 tC1 = k3 k4 k1 t̃C3 (12)

where, tC1 , tC2 , tC3 are the physical design variables describing the rib thicknesses for the three partitions
along the span. tC1 is chosen for the first rib on the cantilevered side. t̃C1 , t̃C2 , t̃C3 are the reference values
for the thicknesses of the three different rib groups which are dictated in the initial wing design. In addition,

4
Ankara International Aerospace Conference
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

two more design variables, the stringer outer diameter d0 and the inner wall thickness of the stringer beam tw
are:
d0 = k4 d˜0 tw = d30 (13)
where d0 is the reference diameter value of the initial wing design. The abstract optimization variables are
chosen to be less than one so that the inital rough structure will be forced to get lighter. The lower and upper
limits of the abstract optimization variables are determined as:

0.8 ≤ k1 ≤ 1.0
0.6 ≤ k2 ≤ 1.0
(14)
0.4 ≤ k3 ≤ 1.0
0.2 ≤ k4 ≤ 1.0

In addition, the location of the first four ribs which is the group on the wing root side and also the location
of the middle two spars are chosen to be variable. The absolute distance from the root to each of the first
four ribs are chosen as four optimization variables y1 ,y2 ,y3 ,y4 . For two middle spars, the ratio of the distance
between the leading edge of the wing to the spar divided by the chord length is chosen as two dimensionless
optimization variables c1 ,c2 . Thus, 16 independent design variables are introduced at all.

500mm ≤ y1 ≤ 800mm 2150mm ≤ y4 ≤ 2800mm

900mm ≤ y2 ≤ 1300mm 0.25 ≤ c1 ≤ 0.45 (15)
1400mm ≤ y3 ≤ 1950mm 0.55 ≤ c2 ≤ 0.75

A rather bulk wing initial design will be given for the optimization problem since abstract variables are chosen
as such to reduce the thicknesses in any ways. At the initial configuration, t̃A1 = t̃A2 = t̃A3 = 5mm,
t̃B1 = t̃B2 = t̃B3 = 20mm, t̃C1 = t̃C2 = t̃C3 = 16mm, y1 = 600mm, y2 = 1100mm,y3 = 1600mm,
y4 = 2250mm, c1 = 0.35, c2 = 0.65.

PROBABILISTIC OPTIMIZATION OF THE AIRCRAFT WING

Reliability-Based Design Optimization Formulation
A reliability-based design optimization problem can be formulated as follows:

min f (d,p,y(d,p)) (16)

s.t. gprob
i (X, η) ≥ 0, i = 1, · · · , Nprob ,
gjdet (d,p,y(d,p)) ≥ 0, j = 1, · · · , Ndet ,
l u
d ≤d≤d

where d denotes the design variables and p denotes the fixed parameters. y(d,p) is a function which is defined
to predict performance characteristics of the designed product. Probabilistic constraints are represented with
the superscript ”prob” while deterministic constraints are represented with the superscript ”det”. Moreover,
X denotes the vector of continuous random variables with known (or assumed) joint cumulative distribution
function (CDF), FX (x). Deterministic parameters, also called the limit state parameters are denoted by η.
Additionally, gprob
i can be written as given below:

giprob = Pallowi − Pi or βi − βreqi (17)

where Pi and βi are the probability of failure and reliability index respectively due to ith failure mode at the
given design. On the other hand, Pallowi and βreqi are the allowable probability of failure and required (target)
reliability index for this failure mode. The equation regarding the relationship between the probability of failure
and reliability index is
Pf ≈ Φ(−β) (18)
where Φ is the standard normal cumulative distribution function (CDF). The probability of failure Pi is given
by Z
Pi = fX (x)dx, (19)
gi (x,η)≤0

where fX (x) denotes the joint probability density function (PDF) of X and g(x, η) ≤ 0 represents the failure
domain.
5
Ankara International Aerospace Conference
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

RBDO Formulation for the Aircraft Wing

A structural optimization problem similar to the one in equations (5)-(9) will be solved with two random vari-
ables which are Young Modulus E and yield strength σyield of the material. Thus, the constraints concerning
stress (g1 ), displacement (g2 ) and frequency (g3 ) in those equations become probabilistic constraints due to
their dependencies on the random variables vector X = [E σyield ]

E and σyield are modeled with normal distributions using N(70000, 350) MPa and N(400, 20) MPa. Thus,
the optimization problem can be formulated as;

min M (s), max f1 (X, s) (20)

sǫS sǫS

σyield (X)
g1prob (X) = σmax (X,s) −1 ≥0 g1prob (X, s) ∈ IR (21)
u0
g2prob (X, s) = umax (X,s) −1 ≥0 g2prob (X, s) ∈ IR (22)
f1 0
g3prob (X, s) = 1 − f (X,s)
≥0 g3prob (X, s) ∈ IR (23)
M0
g4det (s) = M (s) −1 ≥0 g4det (s) ∈ IR (24)

Methodology
Since equation (19) cannot be evaluated analytically in most cases, one of the well known numerical methods
is employed in this work. First order reliability method (FORM) approximate the reliability index (β) by
searching for the most probable (failure) point (MPP) on the limit state surface g(x, η) = 0 in the standard
normal space. FORM employs a linear approximation of the limit state function at the MPP and is considered
accurate as long as the curvature is not too large. FORM also requires the transformation of the random
variables vector X into the standard normal space:

U = T (X) (25)

After the transformation, the components of U are normally distributed with zero means and unit variance
and are statistically independent. Rosenblatt transformation [34] is preferred in this work among possible
approaches. In FORM, MPP search can be performed in two different ways; either using Reliability Index
Approach (RIA) or Performance Measure Approach (PMA). Reliability analysis in RIA can be formulated
as follows:

min kUk
s.t. G(U) = 0 (26)

where reliability analysis in PMA is formulated as the inverse of the reliability analysis in RIA:

min G(U)
s.t. kUk = βt (27)

Here, βt represents the target reliability index. While Hasofer Lind and Rackwitz Fiessler (HLRF) method
can be used for solving the optimization problem in equation (26), advanced mean value and conjugate
mean value methods are available to solve the optimization problem in equation (27). In this work, reliability
analysis is done with PMA because it can be guaranteed that the equality constraints in (27) can be satisfied in
contrast to the standard formulation in (26). Advanced Mean Value, which is preferred in this work to solve the
inverse problem in PMA is a method in which the normalized steepest descent direction n(u) of a performance
function is introduced to minimize the performance function G(U). The AMV method iteratively updates
(k)
the steepest descent direction vector at the probable point uAM V . The AMV method can be formulated as:
(0) (k+1) (k)
uAM V = 0 and uAM V = βt n(uAM V ) (28)
(k)
(k) ∇U G(uAM V )
where n(uAM V ) = − (k)
k∇U G(uAM V )k

In order to perform a reliability analysis for the wing problem, a MATLAB code is written. Target reliability
index is chosen as β = 3. The implemented algorithm was verified for the sample problems in [11] and the
benchmark problem in [14], which involves a simple uniform cantilever beam.
6
Ankara International Aerospace Conference
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

Optimization Framework
During the optimization process Abaqus-6.7.1 is used to compute the structural response of the structural
system and AMV method code written in MATLAB is used to evaluate the random variables. In order to
perform an optimization study, a workflow should be prepared in Modefrontier to govern the optimization
process. In this workflow the optimization variables (with their upper and lower bounds and incrementations),
scheduler, design of experiments, objectives, constraints, output variables and the softwares are defined.
Optimization workflow is prepared to automate the multiobjective multidisciplinary optimization problem.
Once the workflow is run, it controls the optimization process automatically by using the well prepared script
files and models. Figure 3 shows the workflow of this optimization problem. In this workflow, Modefrontier’s
script files drive Abaqus node in batch mode. In each optimization iteration, Modefrontier updates the
thickness parameters of the wing and creates a new input file for Abaqus and MATLAB. Basically, Abaqus
computes mass, frequency, stress and displacement values which are needed in the evaluation of both objective
functions as well as deterministic and probabilistic constraints. MATLAB node performs an inner loop for
reliability analysis to compute the values of random variables according to the current values of optimization
variables at this optimization iteration. The reliability analysis code developed in MATLAB leans on PMA
(performance measure approach) where an inner optimization problem is solved to minimize the violation of
probabilistic constraints while subjected to a given reliability index value in the standard normal space. In
this MATLAB code, the steepest descent direction of probabilistic constraints need to be computed thus the
gradient of probabilistic constraint in standard normal space needs to be computed. The partial derivatives
of probabilistic constraints with respect to the transformed random variables in standard normal space have
to be evaluated. Since Abaqus is a commercial code, we treat it as a blackbox in the reliability analysis,
and calculate the derivatives by central finite differencing, so Abaqus is called twice for each derivative of
probabilistic constraint with respect to one random variable: Once for yplus (design point plus step size h)
and once for yminus (design point minus step size h). The output of the MATLAB reliability code provides
updated values of random variables which are namely Young’s Modulus and yield stress. Finally, this output
file is used in the outer deterministic optimization loop of Modefrontier to calculate the desired values of
displacement, frequency, mass and maximum stress by Abaqus. For reliability analysis, target reliability index
is chosen as βt = 3. Concerning probabistic constraints, a parallel type of system is assumed for the stress
and displacement constraints. For the sake of simplicity, stress constraint is considered to be dominant for
reliability analysis. Figure 2 is a simple representation of the implemented workflow.

Figure 2: Workflow of the optimization problem

OPTIMIZATION RESULTS
In order to confirm that the integration of MATLAB code and the commercial softwares are working well, a
similar workflow for the beam problem defined in [14] was prepared and the results are close enough to the
ones found only from MATLAB code. While final values of design variables were (w, t) = (2.45, 3.88) and
minimum value of objective function wt was 9.52 using only MATLAB, the prepared workflow in ModeFrontier
7
Ankara International Aerospace Conference
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

Figure 3: ModeFrontier Workflow

gives (w, t) = (2.42, 3.83) and wt = 9.27 as probabilistic optimization outputs.

In the aircraft wing study, 24 design of experiments (DOE) with ”Sobol sequence” are used and 300 maximum
number of iterations per subiterations for the NLPQLP (an algorithm based on SQP) are defined. Sobol
sequence distributes the experiments uniformly in the design space. Finally, a total number of 171 designs
are generated for the optimization problem. Solution of the problem took about 85 hours on a workstation
with Intel(R) Core(TM)2 Quad CPU 6700@2.40 GHz processor and 2GB of RAM on Microsoft Windows XP
operating system. 65 designs were found to be feasible that satisfy the constraint condition. Furthermore,
there are 10 error designs. As a result, 3 designs are found in the pareto front set for this optimization problem.
These paretos are demonstrated in table 1 and in figure 4. The first design in table 1 is selected as the

Table 1: Paretos from RBDO

Pareto Mass (kg) Frequency (Hz)

1 291.47 5.7084
2 291.51 5.7087
3 291.61 5.7075

Table 2: Paretos from Deterministic Optimization

Pareto Mass (kg) Frequency (Hz)

1 282.23 5.2682
2 282.65 5.2691
3 282.77 5.2658

optimum design due to its cumulative improvement concerning both objectives. Table 4 demonstrates those
improvements after reliability based design optimization with respect to the reference values of M0 = 330kg
and f10 = 4.35 Hz. Here, both objectives are evaluated with equal weight, i.e. %50 importance for each is
taken into account. Other design engineers may choose among these three designs according to their needs.
Moreover, the designs which are found previously in the deterministic optimization process in [1] are given
in table 2. Briefly, table 3 shows the difference between deterministic and probabilistic optimization analysis.
The results demonstrate that the mass of the reliability based designed wing has to be greater compared to
the wing of the deterministic optimization in order to obtain a safer design. Furthermore, it can be observed
that the first natural frequency also increased after RBDO when compared to the deterministic optimization.
As can be seen in the figures 5 and 6, thicknesses of the structural members decrease from the edge to the
root of the wing in order to increase the inertia of the wing structure. Therefore, depicted results explain the
increase in frequency clearly and compromise with the expectations.
8
Ankara International Aerospace Conference
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

Table 3: Comparison of Deterministic and Probabilistic Optimization Results

Deterministic Optimization Probabilistic Optimization

Mass (kg) 282 291
Frequency (Hz) 5.26 5.70

Table 4: Cumulative Improvements for Multicriteria Decision

Mass (kg) Imp. in M. (%) Frequency (Hz) Imp. in F. (%) Cumu. Imp. (%)
291.47 11.6750 5.7084 31.2275 21.4512
291.51 11.6636 5.7087 31.2344 21.4490
291.61 11.6333 5.7075 31.2068 21.4200

Figure 4: Pareto Optimal Solution of Aircraft Wing Problem

CONCLUDING REMARKS
In this work, reliability-based design optimization is investigated and some of the related solution methodolo-
gies are explained, coded and implemented in MATLAB. Finally, MATLAB code and Abaqus are integrated in
a ModeFrontier framework to solve an optimization problem concerning an aircraft wing. Reliability analysis
can be performed much more effectively using Performance Measure Approach compared to Reliability Index
Approach. Due to its efficiency, PMA is preferred in solving the problems. Relevant methods to PMA are
introduced and validated using some example problems from the literature [11, 14] before proceeding to the
multi-objective design optimization problem of an aircraft wing which is reevaluated following the developed

9
Ankara International Aerospace Conference
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

22
Spar
Rib
20 Skin

18

16

14
Thickness (mm)

12

10

0
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500
Distance from root of the wing (mm)

Figure 5: Locations and Thicknesses of Wing Structure Members Before Optimization

22
Spar
Skin
20 Rib

18

16

14
Thickness (mm)

12

10

0
2000 2500 3000 3500 4000 4500 5000 5500 6000 6500
Distance from root of the wing (mm)

Figure 6: Locations and Thicknesses of Wing Structure Members After RBDO

approach. RBDO results are proved to be acceptable when the results of first evaluation (i.e. deterministic
optimization) of the same problem are considered.

To the best of the author’s knowledge, commercial finite element analysis tools such as Abaqus and Nas-
tran do not include reliability analysis modules. Therefore, the integration of the reliability analysis code
written in MATLAB and Abaqus is the most remarkable novelty of this work. As the main inference, it is ob-
served that RBDO method exhibits superiorities to deterministic optimization when dealing with uncertainties
10
Ankara International Aerospace Conference
AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

in the design process. Future work for this study may include using more complicated models and improving
the capabilities of reliability analysis code: First, the system integration of multiple probabilistic constraints
has to be implemented. Next, more effective methods such as second-order reliability method (SORM, based
on MPP) or Latin Hypercube Sampling (LHS, one of the most efficient sampling methods) may be imple-
mented and performed. Furthermore, reliability analysis algorithms based on a sequential single-loop can be
preferred to overcome the potential computational burden. Parallel computing is another effective solution
where applicable.

ACKNOWLEDGEMENTS
This work is funded by TUBITAK (The Scientific and Technological Research Council of Turkey) through 3501
National Young Researchers Career Development Program for the project titled ”Analysis and Reliability
Based Design Optimization of Fluid-Structure Interaction Problems Subject to Instability Phenomena”
with grant number 105M235.

References
[1] Nikbay, M., Yanangönül, A., Öncü, L. and Koçaş, M., Multi-objective and gradient based structural
design optimization of an aircraft wing, Second International Conference on Multidisciplinary Design
Optimization and Applications, ASMDO, 2-5 September, Gijon, Spain, 2008

[2] Li, F. and Wu, T., d-RBDO: A Deterministic Approach for Reliability Based Design Optimization,
Department of Industrial Engineering, Arizona State University, Tempe, Arizona 85287-5906, USA, 2008

[3] Zou, T. and Mahadevan, S., A Direct Decoupling Approach for Efficient Reliability-Based Design
Optimization, Structural and Multidisciplinary Optimization, 31 (3): 190–200, 2006
[4] (Edited by) Nikolaidis, E., Ghiocel, D. M. and Singhal, S., Engineering Design Reliability Handbook,
CRC Press, 2005

[5] Du, X. and Chen, W., Sequential Optimization and Reliability Assessment Method for Efficient
Probabilistic Design, Journal of Mechanical Design, 126 (2): 225–233, 2004

[6] Arora, J. S., Introduction to Optimum Design, 2nd Edition, Elsevier Academic Press, 2004

[7] Allen, M., Reliability-Based Design Optimization of Multiphysics, Aerospace Systems, Department of
Aerospace Engineering Sciences, University of Colorado, 2004

[8] Agarwal, H., Reliability-Based Design Optimization: Formulations and Methodologies, Department
of Aerospace and Mechanical Engineering, University of Notre Dame, 2004
[9] Youn, B. D. and Choi, K. K., A New Response Surface Methodology for Reliability-Based Design
Optimization, Computers and Structures, 82: 241–256, 2004

[10] Yang, R. J. and Gu, L., Experience with Approximate Reliability-Based Optimization Methods, Struc-
tural and Multidisciplinary Optimization, 26: 152–159, 2004

[11] Youn, B. D., Choi, K. K. and Park, Y. H., Hybrid Analysis Method for Reliability-Based Design
Optimization, Journal of Mechanical Design, 125 (2): 221–232, 2003

[12] Liu, H., Chen, W. and Sheng, J., Application of the Sequential Optimization and Reliability Assessment
Method to Structural Design Problems, ASME Design Engineering Technical Conferences and Computers
and Information in Engineering Conference, Anonymous American Society of Mechanical Engineers, New
York, NY 10016-5990, United States, Chicago, IL, United States, 2 A: 63–72, 2003

[13] Maute, K., Nikbay, M. and Farhat, C., Coupled Analytical Sensitivity Analysis and Optimization of
Three-Dimensional Nonlinear Aeroelastic Systems, AIAA, 39 (11): 2051–2061, 2001

[14] Sues, R., Aminpour, M. and Shin, Y., Reliability-Based Multidisciplinary Optimization for Aerospace
Systems, Proc. 42nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials
Conference, AIAA, 16-19 April, Seattle, WA, Paper No 2001-1521, 2001

11
Ankara International Aerospace Conference
 AIAC-2009-115 Nikbay, Ulucenk, Yanangönül & Aysan

[15] Tu, J., Choi, K. K. and Park, Y. H., Design Potential Method for Robust System Parameter Design,
AIAA Journal, 39 (4): 667–677, 2001
[16] Royset, J. O., Der Kiureghian, A. and Polak, E., Reliability-Based Optimal Structural Design by the
Decoupling Approach, Reliability Engineering and System Safety, 73 (3): 213–221, 2001
[17] Koch, P. N., Simpson, T. W. and Allen, J. K., Statistical Approximations for Multidisciplinary Design
Optimization: The Problem of Size, Journal of Aircraft, 36: 275–286, 1999
[18] Jha, R. and Chattopadhyay, A., Multidisciplinary Optimization of Composite Wings Using Refined
Structural and Aeroelastic Analysis Methodologies, Engineering Optimization, 32: 59–78, 1999
[19] Robinson, D. G., A Survey of Probabilistic Methods Used in Reliability, Risk and Uncertainty Analysis:
Analytical Techniques I, Sandia National Laboratories, SAND98-1189, 1998
[20] Chen, X., Hasselman, T. K. and Neill, D. J., Reliability Based Structural Design Optimization for
Practical Applications, Anonymous AIAA, New York, NY, USA, Kissimmee, FL, USA, 4: 2724–2732,
1997
[21] Chen, W., Allen, J. K. and Tsui, K., Procedure for Robust Design: Minimizing Variations Caused by
Noise Factors and Control Factors, Journal of Mechanical Design, 118: 478–485, 1996
[22] Dovi, A. R., Wrenn, G. A., Barthelemy, J.-F. M., Coen, P. G. and Hall, L. E., Multidisciplinary Design
Integration Methodology for a Supersonic Transport Aircraft, Journal of Aircraft, 32 (2): 290–296,
1995
[23] Sues, R. H., Oakley, D. R. and Rhodes, G. S., Multidisciplinary Stochastic Optimization, Anonymous
ASCE, New York, NY, USA, Boulder, CO, USA, 2: 934–937, 1995
[24] Koyluoglu, H. U. and Nielsen, S. R. K., New Approximations for SORM Integrals, Structural Safety,
13: 235–246, 1994
[25] Barthelemy, J.-F. M., Wrenn, G., Dovi, A. and Hall, L., Supersonic Transport Wing Minimum Design
Integrating Aerodynamics and Structures, Journal of Aircraft, 31: 330–338, 1994
[26] Tzong, G., Baker, M., Vari, R. D. and Giesing, J., Aeroelastic Loads and Structural Optimization of
a High Speed Civil Transport Model, 5th AIAA/NASA/USAF/ISSMO Symposium on Multidisciplinary
Analysis and Optimization, AIAA, Panama City Beach, FL, Paper No 94-4378, 1994
[27] Parkinson, A., Sorensen, C. and Pourhassan, N., General Approach for Robust Optimal Design, Journal
of Mechanical Design, 115: 74–80, 1993
[28] Friedmann, P., Helicopter Vibration Reduction Using Structural Optimization with Aeroelas-
tic/Multidisciplinary Constraints - A Survey, Journal of Aircraft, 28: 8–21, 1991.
[29] Eggert, R. J., Quantifying Design Feasibility Using Probabilistic Feasibility Analysis, 17th Design
Automation Conference presented at the 1991 ASME Design Technical Conferences, ASME, New York,
NY, USA, 32: 235–240, 1991
[30] Thanedar, P. B. and Kodiyalam, S., Structural Optimization Using Probabilistic Constraints, Pro-
ceedings of the 32nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials
Conference, AIAA, 8-10 April, New York, NY, USA, Baltimore, MD, USA, pp. 205, 1991
[31] Bowman, K., Granddhi, R. and Eastep, F., Structural Optimization of Lifting Surfaces with Divergence
and Control Reversal Constraints, Structural Optimization, 1: 153–161, 1989.
[32] Hohenbichler, M., Gollwitzer, S. and Kruse, W., New Light on First- and Second Order Reliability
Methods, Structural Safety, 4: 267–284, 1987
[33] Hasofer, A. M. and Lind, N. C., Exact and Invariant Second-Moment Code Format, Journal of the
Engineering Mechanics, 100: 111–121, 1974
[34] Rosenblatt, M., Remarks on a Multivariate Transformation, Annals of Mathematical Statistics, 23
(3): 470–472, 1952
[35] Abaqus 6.7 Version Documentation, Simulia.
[36] Modefrontier V4 Version Documentation, Esteco.
12
Ankara International Aerospace Conference

View publication stats