Professional Documents
Culture Documents
AIAA JA Nikbay Kuru 2013
AIAA JA Nikbay Kuru 2013
AIAA JA Nikbay Kuru 2013
net/publication/253650894
CITATIONS READS
16 520
2 authors:
Some of the authors of this publication are also working on these related projects:
Computational Aeroelasticity Investigations For NATO-RTO-AVT-203 and Aeroelastic Prediction Workshop View project
All content following this page was uploaded by Melike Nikbay on 21 May 2014.
aerodynamic uncertainties, such as in yield strength, Mach number, and angle of attack. To evaluate the probability
of failure for the probabilistic constraints a first-order reliability analysis method, the Hasofer–Lind iteration
method, is implemented in MATLAB to compute the most probable failure point solution. The integrated framework
is validated with structural problems and then extended to more realistic wing configurations with aeroelastic criteria.
The presented reliability-based multidisciplinary optimization process is proven to be fully automatic, modular, and
practical, which could find potential applications in industrial problems.
model-based approaches in [6–8]. Among those, MPP-based ap- η. Additionally, probabilistic constraint (gprob
i ≥ 0) can be written as
proaches have attracted more attention, as they require relatively less given below:
computational effort while still producing results with acceptable
accuracy compared to the other three approaches [9,10]. Another Pallowi ≥ Pi (2)
research topic for RBDO is on the integration of reliability analysis
and optimization, using nested double-loop strategy or decoupled where Pi is the probability of failure due to ith failure mode at the
double-loop strategy. Nested double-loop methods treat the reli- given design. On the other hand, Pallowi is the allowable probability of
ability analysis as the inner loop analyzing the probabilistic con- failure for this failure mode. The probability of failure Pi is given by
straint satisfaction of which solutions are given to the outer optimizer Z
to locate the optimal solution iteratively. As a result, nested double-loop Pi fXR x dx (3)
methods are computationally expensive for a complex engineering gprob
i <0
design [9,11,12]. Therefore, decoupled double-loop methods have
been developed to address the computational challenges as in [9, where fXR x denotes the joint probability density function of XR ,
12–17]. and gprob
i < 0 represents the failure domain.
Integration of RBDO methodologies to aerospace engineering Equation (3) can not be evaluated analytically. When g has a
applications has been a challenging research subject recently. Pettit normal distribution, the probability of failure can be written as
[18] presented general sources of uncertainty on aeroelastic response,
such as flutter flight testing, prediction of limit cycle oscillations, and P 1 − Φβ (4)
Downloaded by ISTANBUL TEKNIK UNIIVERSTESI on May 28, 2013 | http://arc.aiaa.org | DOI: 10.2514/1.C031693
min kUk s.t. gU; η 0 (7) If the limit-state function is nonlinear the approximate limit-state
surface is obtained by linearizing the original limit-state function at
where reliability analysis in PMA is formulated as the inverse of the the mean value point. Therefore, this method is called the mean-value
reliability analysis in RIA: method, and the β0 given in Eq. (13) is called a MVFOSM reli-
ability index.
min gU s.t. kUk βt (8) 4) After βk is found its direction cosine is computed with the
following equation:
Here, βt represents the target reliability index. Although the Hasofer–
Lind (HL) [5] method can be used for solving the optimization ∂gXk
∂XRi σ XRi
R
problem in Eq. (7), advanced mean value and conjugate mean value
γ ki cos θkXRi cos θkUi − s
2 (14)
methods presented in [27] are available to solve the optimization Pn ∂gXkR
problem in Eq. (8). i1 ∂XRi σ XRi
In this work, probabilistic constraint is evaluated with RIA. The
HL [5] iteration method is employed to solve the optimization
problem in RIA.
5) Next, a new design point Xk1 R in original space and its
A. Hasofer–Lind Iteration Method corresponding set Uk1 in standard normal space are found as
The HL method for normally distributed variables was proposed
XkRi − μXRi
by Hasofer and Lind and then extended by Rackwitz and Fiessler to Uk1 βk cos θkXRi (15)
include random variable distribution information, calling their ex-
i
σ XRi
tended method the Hasofer/Lind–Rackwitz/Fiessler (HL–RF) meth-
od [28]. Assuming that the limit-state surface with n-dimensional
Ri μXRi β σ XRi cos θXRi
Xk1
normally distributed and independent random variables XR is k k (16)
∂XRi σ XRi U i
k1
reliability analysis implemented in this work is as follows: βk1 s
2 (17)
1) Assuming that the variables XR are statistically independent the Pn ∂gUk1
approximate limit-state function at the mean is written as i1 ∂XRi σ XRi
In this work, the partial derivatives needed for direction cosine and
μg~ ≈ EgμXR gμXR (11) β calculations will be computed by finite differencing, because
commercial codes are used to solve for structural and aerodynamic
where E⋅ is the expectation operator. The standard deviation of the responses. These commercial codes will be treated as black box and
approximate limit-state function is run again with a small perturbation parameter from batch mode, so
that the numerical derivative needed in reliability analysis will be
s
2
X n automatically computed by forward differencing in HL iterations in
∂gμXR
σ g~ σ (12) reliability analysis.
i1
∂XRi XRi
B. Validation of Reliability-Based Design Optimization Methodology
3) The initial reliability index β for iteration k 0 is computed as An in-house RBDO code is developed in MATLAB and validated
by a cantilever beam problem [29]. Figure 3 shows the cantilever
μg~
β0 (13) beam with a rectangular cross section subjected to a vertical load, PY ,
σ g~ and a lateral load, PX , at the tip. It is assumed that the beam has a fixed
NIKBAY AND KURU 711
length of L 100.0 in: and that the cross-sectional dimensions of Table 1 Statistics of loads and material properties
the beam remain constant along the length of the beam. The design for the cantilever beam problem
objective is to prevent yielding due to normal stress while keeping the Random variable R, psi PX , lbf PY , lbf
weight of the beam low or, equivalently, the cross-sectional area,
Mean 40,000 500 1,000
w t. We consider yielding at a corner of the fixed end of the beam, Standard deviation 2,000 100 100
where the normal stress (S) is maximum as a limit state, so that
600PY 600PX
S (18) min F w t s.t. β ≥ βt 3.0
wt2 w2 t
600PY 600PX
g R−S R− (20)
wt2 w2 t
600PY 600PX
g R−S R− (19) In the first step, stress constraint is solved analytically as given in
wt2 w2 t
Eq. (19) and in the second step, Abaqus 6.7.1 is used to compute
where R is the random yield strength, S is the maximum normal stress solution with the purpose of building the high-fidelity optimi-
stress, and PX and PY are mutually independent random loads. zation framework for further studies. An in-house RBDO code is
Assume that there are scatters in PX , PY , and R, and they follow developed in MATLAB and integrated to handle the probabilistic
normal distributions with the following parameters as shown in constraint. An optimization workflow is prepared in Modefrontier to
Downloaded by ISTANBUL TEKNIK UNIIVERSTESI on May 28, 2013 | http://arc.aiaa.org | DOI: 10.2514/1.C031693
Table 2 Optimization results for the cantilever beam where ρ is the fluid density; v1 , v2 , v3 are three velocities; and E~ is the
problem total internal energy per mass. The flux, F, ~ has three components, F~ 1 ,
w, in. t, in. β F w t; in:2 F~ 2 , and F~ 3 as follows:
Present study 2.40 3.96 3.00 9.52 0 1 0 1
Reference study [29] 2.45 3.89 3.00 9.52 ρv1 ρv2
B C B C
B ρv21 p C B ρv2 v1 C
B C B C
B C B C
F~ 1 B ρv1 v2 C F~ 2 B ρv22 p C
optimization problem. As a result, 67 unfeasible designs and 283 B C B C
B ρv v C B ρv v C
feasible designs are found. The results shown in Table 2 supplied by @ 1 3 A @ 2 3 A
the current framework agree well with the analytical stress solution E~ pv1 E~ pv2
and RBDO benchmarks in literature. 0 1
ρv3
B C
B ρv3 v1 C
III. Aeroelastic Analysis B C
B C
In this study, the MPCCI is used as an aeroelastic coupling F~ 3 B ρv3 v2 C (25)
B C
interface. The advantage of using MPCCI is that it facilitates the B ρv2 p C
@ 3 A
exchange of data between nonmatching mesh interfaces of CFD and
Downloaded by ISTANBUL TEKNIK UNIIVERSTESI on May 28, 2013 | http://arc.aiaa.org | DOI: 10.2514/1.C031693
CSD codes. The staggered algorithm used in this study is given in E~ pv3
Fig. 5. The first arrow shows that CFD code sends the pressure data to
CSD code. The second arrow shows that CSD code calculates the where p is the fluid pressure.
deformation of the structure under this pressure load. Then, the third
arrow shows that structural displacement values are sent to CFD
code, and the mesh is updated. Finally, by the fourth arrow, CFD code
starts a new flow analysis with the new boundary conditions. IV. Reliability-Based Aeroelastic Optimization
Abaqus 6.7.1, a finite element analysis software, was used as a of AGARD 445.6 Wing
linear, static structural solver. Equations of motion for a linear In the final step of this study, RBDO is applied to an aeroelastic
structure can be written as follows: optimization problem. The well-known AGARD 445.6 wing is
chosen as the wing structure. This wing is the first aeroelastic
Mfug
Dfug
_ Kfug fFa g fFe g (21) configuration that is tested in [31] in the “transonic dynamics tunnel”
at the NASA Langley Research Center. The AGARD 445.6 wing is a
The time derivatives of Eq. (21) are neglected, because a static swept-back wing with a quarter-chord sweep angle of 45 deg. Cross
aeroelastic analysis of a cantilever wing will be performed. Also, sections of the wing are NACA 65A004 airfoils. The wing has a
gravity is neglected, and only the aerodynamic forces are considered. taper ratio of 0.66 and an aspect ratio of 1.65. Moreover, it is a
Therefore, the system of linear equations can be written as wall-mounted model made with laminated mahogany. The wings
parametric CAD model prepared with Catia V5 is given in Fig. 6.
Kfug fFa g (22) There are two models of the AGARD 445.6 wing: solid and weak-
ened. In this study, the weakened model of the wing is used. The finite
Abaqus 6.7.1 calculates displacements by using the aerodynamic loads element model in Abaqus is composed of 19,610 linear hexahedral
calculated from the flow solver Fluent. Fluent 6.3.26 can be employed structural elements. The computational grid of the flow domain was
for modeling fluid flow both for structured and unstructured grids by constructed in Gambit with 691,000 tetrahedral elements and 1.35 ×
using the Navier–Stokes/Euler equations. A finite volume-based 108 faces. The flow is modeled with the Euler equations.
approach is used to define the discrete equations. The fluid solver There are two objective functions in the problem: maximizing
solves the governing equations of continuity, momentum, and energy the lift/drag ratio “L∕D” and minimizing wing mass “m”. For multi-
simultaneously. In this study, the flow was assumed to be inviscid, and objective optimization problems the optimization driver will try to
the Euler equations in Eq. (23) were used. This is a valid approximation find the Pareto optimal set. Deterministic aeroelastic optimization of
for the high Reynolds number flows according to the Prandtl’s
the AGARD 445.6 wing problem was presented formerly in [24] for
boundary-layer analysis. Moreover, in [30], inviscid flow models
the freestream Mach number of 0.85 and the angle of attack of 5 deg.
give acceptable results for maximizing the lift/drag optimization
In the present work, we consider a set of random variables for fluid
problems in transonic cruise conditions. In conservative form, the
and structural domain XR σ yield M α , whereas uncertainties
three-dimensional Euler equations are written as
in yield strength, Mach number, and angle of attack will be accounted
∂w for in the reliability analysis. The probabilistic optimization problem
∇ · Fw
~ 0 (23) has two objectives, z1 s and z2 XR ; s, and can be formulated as
∂t
The conservative fluid state variable, w, is defined as L
minfz1 sg; maxfz2 XR ; sg minfmsg; max XR ; s
D
0 ρ 1
(26)
B ρv1 C
B C
w B ρv2 C (24)
@ A
ρv3 76 mm
gprob
1 XR ; s − 1.0 ≥ 0 gprob
1 XR ; s ∈ R (27)
E umax XR ; s
σ yield XR
gprob
2 XR ; s − 1.0 ≥ 0 gprob
2 XR ; s ∈ R (28)
σ max XR ; s
Fig. 6 AGARD 445.6 wing geometry and finite element method model.
Downloaded by ISTANBUL TEKNIK UNIIVERSTESI on May 28, 2013 | http://arc.aiaa.org | DOI: 10.2514/1.C031693
βgprob will be used as N8.0; 0.4 MPa. For the RBDO problem, the
2 XR ; s ≥ 0 − βs ≥ 0 (30)
freestream Mach number was set to 0.85, and the angle of attack to
5 deg. The Mach number is treated with a normal distribution
S fs ∈ R2 jsL ≤ s ≤ sU g N0.85; 0.03, and the angle of attack is treated as N5; 0.25 deg.
Here, βs and βd are the target reliability indices for stress and dis-
s λ; Λ4c 0.1 ≤ λ ≤ 0.5 0° ≤ Λ4c ≤ 50° (31) placement constraints and chosen as to be 5.1993 for a probability of
failure of 10E-7. The actual reliability index values for the current
where ms is the total mass of the wing, DL XR ; s is the lift over drag design at each optimization iteration is calculated according to Eq. (17)
value for the wing, σ max XR ; s is the maximum Von Mises stress on and passed to the outer optimization loop as a constraint. The workflow
ctip
the wing, λ is the taper ratio defined as λ croot , and Λc4 is the sweep for the above probabilistic optimization problem is given in Fig. 7.
value at the quarter-chord. In this study, the yield strength σ yield will Several commercial software codes were coupled during the
be treated as a random variable with a normal distribution, and σ yield optimization process in this problem. Fluent 6.3.26 is used to solve
inviscid three-dimensional Euler equations, Gambit to generate the shown in mass vs L∕D ratio space in Fig. 8, where the regression line,
fluid domain mesh generator, and Catia V5-R16 to model the param- which shows the relationship between objective functions for the
etric three-dimensional solid. Abaqus 6.7.1 was driven to compute feasible design points, is demonstrated. As it is seen from the
the structural response of the aeroelastic system. MPCCI 3.0.6 was regression line in Fig. 8 the L∕D values are increasing while mass
employed to exchange the pressure and displacement information values are increasing.
between Fluent and Abaqus. Modefrontier 4.0 was treated as a multi-
objective and multidisciplinary optimization software to automate V. Conclusions
the workflow. In Fig. 7, the optimization variables (their limits and
increments), optimization algorithm (scheduler), DOE, objectives, In this work, a reliability index approach (RIA) is implemented
constraints, output variables, and the software are defined. At every in an in-house-developed reliability-based design optimization
(RBDO) code and integrated into a multidisciplinary optimization
optimization iteration, the workflow is run from the beginning with
framework composed of high-fidelity commercial software. In this
the changes in design.
computational framework, a finite volume-based flow solver Fluent
The workflow shown in Fig. 7 is run with 12 DOEs with “Sobol
is used to solve inviscid three-dimensional Euler equations, and Catia
Sequence,” where the multi objective genetic algorithm II (MOGA II)
is treated as a parametric three-dimensional solid modeler. Abaqus,
is used for attaining optimum result. Finally, a total number of 43
a structural finite element method solver, is driven to compute the
designs are generated for the optimization problem. A solution of the
structural response of the aeroelastic system. A mesh-based parallel
problem took about 215 h 40 min on a workstation with an Intel Core code coupling interface (MPCCI), is used to exchange the pressure
2 Quad CPU 6700 at 2.40 GHz processor with 2 GB of RAM on a and displacement information between Fluent and Abaqus to
Microsoft Windows XP operating system. Twenty-three feasible perform a loosely coupled aeroelastic analysis. Modefrontier is
designs and 20 unfeasible designs were found. As a result, six designs employed as a multi-objective and multidisciplinary optimization
are found in the Pareto front set for this optimization problem. These driver to control the optimization workflow. The optimization criteria
Paretos are demonstrated in Table 3. The design, which corresponds include probabilistic constraints with both structural and aero-
to Pareto 3 in Table 3, is chosen as optimum design because of its dynamic uncertainties. A first-order reliability analysis method, the
minimum mass value, satisfactory L∕D ratio, while still satisfying HL iteration method, is implemented in MATLAB to compute the
the target reliability index constraints. most probable failure point (MPP) solution. The RBDO meth-
The selected Pareto design 3 gives approximately a 40.7% odology is validated with an example from the literature and then
decrease in mass and 0.7% increase in L∕D ratio with respect to the applied to an aeroelastic optimization problem for the AGARD 445.6
reference values of 1.831 kg and 11.4589. All feasible designs are wing. In the final application, random variation in structural and
aerodynamic parameters, such as in yield strength, Mach number,
and angle of attack are considered. The presented reliability-based
multidisciplinary optimization process is proven to be fully auto-
matic, modular, and practical, which could find potential applications
in industrial problems.
Acknowledgments
This work is supported by the Scientific and Technological
Research Council of Turkey (TUBITAK) through the 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”
under grant No. 105M235. The writers would like to thank to Istanbul
Technical University Informatics Institute High Performance Com-
puting Laboratory for providing mesh-based parallel code coupling
interface (MPCCI) licenses.
References
[1] Eggert, R. J., “Quantifying Design Feasibility Using Probabilistic
Feasibility Analysis,” 1991 ASME Advances in Design Automation,
Vol. 32, New York, NY, 1991, pp. 235–240.
[2] Parkinson, A., Sorensen, C., and Pourhassan, N., “General Approach for
Fig. 8 Mass vs L∕D ratio space. Robust Optimal Design,” Journal of Mechanical Design, Vol. 115,
NIKBAY AND KURU 715
No. 1, 1993, pp. 74–80. [17] Youn, B., Choi, K., and Park, Y., “Hybrid Analysis Method for
doi:10.1115/1.2919328 Reliability-Based Design Optimization,” Journal of Mechanical
[3] Hohenbichler, M., Gollwitzer, S., and Kruse, W., “New Light on First- Design, Vol. 125, No. 2, 2003, pp. 221–232.
and Second-Order Reliability Methods,” Structural Safety, Vol. 4, No. 4, doi:10.1115/1.1561042
1987, pp. 267–284. [18] Pettit, C. L., “Uncertainty Quantification in Aeroelasticity: Recent
doi:10.1016/0167-4730(87)90002-6 Results and Research Challenges,” Journal of Aircraft, Vol. 41, No. 5,
[4] Koyluoglu, H. U., and Nielsen, S. R. K., “New Approximations for 2004, pp. 1217–1229.
SORM Integrals,” Structural Safety, Vol. 13, No. 4, 1994, pp. 235–246. doi:10.2514/1.3961
doi:10.1016/0167-4730(94)90031-0 [19] Allen, M., and Maute, K., “Reliability Based Design Optimization of
[5] Hasofer, A. M., and Lind, N. C., “Exact and Invariant Second-Moment Aeroelastic Structures,” Structural and Multidisciplinary Optimization,
Code Format,” Journal of the Engineering Mechanics Division, Vol. 27, No. 4, 2004, pp. 228–242.
Vol. 100, No. 1, 1974, pp. 111–121. doi:10.1007/s00158-004-0384-1
[6] Chen, W. J., Allen, K., and Tsui, K., “Procedure for Robust Design: [20] Hosder, S., and Maddalena, L., “Non-Intrusive Polynomial Chaos for
Minimizing Variations Caused by Noise Factors and Control Factors,” the Stochastic CFD Study of a Supersonic Pressure Probe,” AIAA Paper
Journal of Mechanical Design, Vol. 118, No. 4, 1996, pp. 478–485. 2009-1129, 2009.
doi:10.1115/1.2826915 [21] Lee, J., Kang, H. Y., Kwon, J. H., and Kwak, B. M., “Reliability of
[7] Sues, R. H., Oakley, D. R., and Rhodes, G. S., “Multidisciplinary Aerodynamic Analysis Using a Moment Method,” International
Stochastic Optimization,” 10th Conference on Engineering Mechanics, Journal of Computational Fluid Dynamics, Vol. 23, No. 6, 2009,
Boulder, CO, 21–24 May 1995. pp. 495–502.
[8] Koch, P. N., Simpson, T. W., and Allen, J. K., “Statistical Approxi- doi:10.1080/10618560902984422
Downloaded by ISTANBUL TEKNIK UNIIVERSTESI on May 28, 2013 | http://arc.aiaa.org | DOI: 10.2514/1.C031693
mations for Multidisciplinary Design Optimization: The Problem of [22] Chassaing, J.-C., and Lucor, D., “Stochastic Investigation of Flows
Size,” Journal of Aircraft, Vol. 36, No. 1, 1999, pp. 275–286. About Airfoils at Transonic Speeds,” AIAA Journal, Vol. 48, No. 5,
doi:10.2514/2.2435 2010, pp. 938–950.
[9] Zou, T., and Mahadevan, S., “A Direct Decoupling Approach for doi:10.2514/1.42637
Efficient Reliability-Based Design Optimization,” Structural and [23] Missoum, S., Dribusch, C., and Beran, P., “Reliability Based Design
Multidisciplinary Optimization, Vol. 31, No. 3, 2006, pp. 190–200. Optimization of Nonlinear Aeroelasticity Problems,” Journal of
doi:10.1007/s00158-005-0572-7 Aircraft, Vol. 47, No. 5, 2010, pp. 992–998.
[10] Liu, H., Chen, W., and Sheng, J., “Application of the Sequential doi:10.2514/1.46665
Optimization and Reliability Assessment Method to Structural [24] Nikbay, M., Oncu, L., and Aysan, A., “Multidisciplinary Code Coupling
Design Problems,” Proceedings of DETC’03 ASME 2003 Design for Analysis and Optimization of Aeroelastic Systems,” Journal of
Engineering Technical Conferences and Computers and Information in Aircraft, Vol. 46, No. 6, 2009, pp. 1938–1944.
Engineering Conference, Paper DETC2003/DAC-48710Chicago, IL, doi:10.2514/1.41491
2–6 Sept. 2003. [25] Nikbay, M., Yanangonul, A., Oncu, L., and Kocas, M., “Multi-Objective
[11] Yang, R., and Gu, L., “Experience with Approximate Reliability-Based and Gradient Based Structural Design Optimization of an Aircraft
Optimization Methods,” Structural and Multidisciplinary Optimization, Wing,” Second International Conference on Multidisciplinary Design
Vol. 26, No. 1, 2004, pp. 152–159. Optimization and Applications, ASMDO, Gijon, Spain, 2–5 Sept. 2008.
doi:10.1007/s00158-003-0319-2 [26] Rosenblatt, M., “Remarks on a Multivariate Transformation,” Annals of
[12] Du, X., and Chen, W., “Sequential Optimization and Reliability Mathematical Statistics, Vol. 23, No. 3, 1952, pp. 470–472.
Assessment Method for Efficient Probabilistic Design,” Journal of doi:10.1214/aoms/1177729394
Mechanical Design, Vol. 126, No. 2, 2004, pp. 225–233. [27] Youn, B., and Choi, K., “A New Response Surface Methodology for
doi:10.1115/1.1649968 Reliability-Based Design Optimization,” Computers and Structures,
[13] Thanedar, P. B., and Kodiyalam, S., “Structural Optimization Using Vol. 82, Nos. 2–3, 2004, pp. 241–256.
Probabilistic Constraints,” Proceedings of the 32nd AIAA/ASME/ASCE/ doi:10.1016/j.compstruc.2003.09.002
AHS/ASC Structures, Structural Dynamics and Materials Conference, [28] Choi, S. K., Grandhi, R. V., and Canfield, R. A., Reliability-Based
AIAA Paper 1991-0922-CP, New York, NY, 8–10 April 1991. Structural Optimization, Springer, 2007.
[14] Tu, J., Choi, K. K., and Park, Y. H., “Design Potential Method for Robust [29] Sues, R., Aminpour, M., and Shin, Y., “Reliability-Based Multi-
System Parameter Design,” AIAA Journal, Vol. 39, No. 4, 2001, disciplinary Optimization for Aerospace Systems,” Proceedings of the
pp. 667–677. 42nd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics,
doi:10.2514/2.1360 and Materials Conference, AIAA Paper 2001-1521, Seattle, WA,
[15] Chen, X., Hasselman, T. K., and Neill, D. J., “Reliability Based 16–19 April 2001.
Structural Design Optimization for Practical Applications,” 38th AIAA/ [30] Barcelos, M., and Maute, K., “Aeroelastic Design Optimization for
ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materi- Laminar and Turbulent Flows,” Computer Methods in Applied
als Conference and Exhibit, and AIAA/ASME/AHS Adaptive Structures Mechanics and Engineering, Vol. 197, Nos. 19–20, 2008, pp. 1813–
Forum, AIAA Paper 1997-1403, Kissimmee, FL, 1997. 1832.
[16] Royset, J. O., Der Kiureghian, A., and Polak, E., “Reliability-Based doi:10.1016/j.cma.2007.03.009
Optimal Structural Design by the Decoupling Approach,” Reliability [31] Yates, E., “AGARD Standard Aeroelastic Configurations for Dynamic
Engineering and System Safety, Vol. 73, No. 3, 2001, pp. 213–221. Response I-Wing 445.6,” AGARD Rept. No. 765, 1985.
doi:10.1016/S0951-8320(01)00048-5