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

net/publication/371718622

Multidisciplinary structural optimization of novel high‐aspect ratio

composite aircraft wings

Article in Structural and Multidisciplinary Optimization · June 2023

DOI: 10.1007/s00158-023-03600-1

CITATIONS READS

0 175

2 authors:

Spyridon Kilimtzidis Vassilis Kostopoulos

University of Patras University of Patras
9 PUBLICATIONS 16 CITATIONS 370 PUBLICATIONS 5,760 CITATIONS

SEE PROFILE SEE PROFILE

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

DECODE - 956470 (Marie Skłodowska-Curie Actions) : Drug-coated balloon simulation and optimization system for the improved treatment of peripheral artery disease
View project

PhD Research - ESA NPI Programme View project

All content following this page was uploaded by Spyridon Kilimtzidis on 20 June 2023.

The user has requested enhancement of the downloaded file.

Structural and Multidisciplinary Optimization (2023) 66:150
https://doi.org/10.1007/s00158-023-03600-1

INDUSTRIAL APPLICATION PAPER

Multidisciplinary structural optimization of novel high‑aspect ratio

composite aircraft wings
Spyridon Kilimtzidis1 · Vassilis Kostopoulos1

Received: 2 May 2022 / Revised: 26 December 2022 / Accepted: 19 May 2023

© The Author(s) 2023

Abstract
Novel high-aspect ratio airframe designs pave the way for a more sustainable aviation future. Such configurations enhance
the aerodynamic efficiency of an aircraft through induced drag reduction mechanisms. Further performance gains, mainly
in terms of structural mass, are accomplished via composite materials airframes. Nevertheless, undesired phenomena such
as geometric nonlinearities and aeroelastic couplings due to elevated flexibility may often rise, rendering the design and
optimization of such airframes extremely intricate and prohibitive in terms of computational cost. Low-fidelity tools, often
preferred on the early design stages, accelerate the design process, albeit suffering from reduced accuracy and ability to
capture higher-order phenomena. Contrastingly, high-fidelity computational methods incur excessive computational cost and
are therefore utilized at the later, detailed design stages. There arises, therefore, the need for a combination of the various
fidelities involved in a cost-effective manner, in order to drive the design towards optimal configurations without significant
performance losses. In our approach, variable fidelity analyses are initially conducted in order to shed light on their effect
on the structural response of a high-aspect ratio composite materials reference wing. An optimization framework combin-
ing low and high-fidelity tools in a sequential manner is then proposed, aiming at attaining a minimum mass configuration
subject to multidisciplinary design constraints. As demonstrated, reasonable mass reduction was obtained for a future aircraft
wing configuration.

Keyword Aircraft wings, Structural optimization, Composite materials, Computational aerodynamics

1 Introduction increase of CO2 and NOx concentration in the atmosphere

that constitutes a severe public threat. On that end, the
1.1 Background European Commission envisages through the definition
of ACARE 2050 (2011), among others, a 75% reduc-
Ever since its inception at the dawn of the twentieth cen- tion in CO2 emissions for future commercial transport
tury, the airline industry has been dramatically expanding, aircrafts. A key enabler for this effort is the introduction
with the latest predictions (Boeing 2018; Airbus 2018) of novel, more efficient airframe designs with improved
foreseeing a 4.3% annual traffic growth over the next 20 aerodynamic as well as structural efficiency. In the field of
years. This forthcoming rise in fuel consumption raises aerodynamics, the aerodynamic efficiency of an aircraft is
environmental concerns, mainly due to the associated closely related to the aspect ratio of the wings, with higher
aspect ratios reducing the induced drag and increasing the
lift-to-drag ratio, which in turn improves fuel efficiency.
Spyridon Kilimtzidis and Vassilis Kostopoulos have contributed
equally to this work. Therefore, and as clearly depicted in Fig. 1, an increas-
ing trend in the aspect ratio of commercial aircraft wings
Responsible Editor: Graeme James Kennedy over the past decades is observed. On the other hand, the
advent of advanced composite materials in the aeronautics
* Spyridon Kilimtzidis
s.kilimtzidis@upnet.gr scene, replacing conventional metallic alloys, has enabled
the development of lighter yet stiffer airframe configura-
1
Mechanical Engineering & Aeronautics Department, tions with aeroelastic tailoring and active load alleviation
University of Patras, Patras University Campus, capabilities. Nevertheless, the aforementioned aircraft
26504 Patras, Greece

13
Vol.:(0123456789)
150 Page 2 of 23 S. Kilimtzidis, V. Kostopoulos

design space exploration is reinforced by alleviating the

computational burden, steering the design process towards
optimality.
Early efforts, confined by modest computational
resources, focused primarily on low-fidelity representations
of the structure as well as of the aerodynamic disciplines.
Pioneering research was conducted by Triplett (1980) and
Love and Bohlman (1989), where TSO, one of the earliest
multidisciplinary design and optimization tools, was devel-
oped. An equivalent plate model of the structure coupled
with the doublet lattice method (DLM) for aerodynamics
were combined to optimize the thickness distribution and
Fig. 1  Commercial aircraft wings aspect ratio trend laminate orientations of a fighter wing subject to strength
and flutter velocity constraints. Haftka (1973, 1977) was also
among the first to combine computational tools of similar
fidelity along with optimization algorithms to perform aer-
oelastic analysis and optimize aircraft wings under stress
and drag constraints. The superiority of a composite wing
design against the up-to-date traditional aluminum one
was demonstrated in terms of mass and induced drag. The
discontinuity of the flutter constraint was also pinpointed.
In a similar fashion, Grossman et al. (1988) conducted an
aerostructural optimization study of a sailplane wing using
lifting-line aerodynamics along with beam equations and
went on to optimize a transport aircraft wing, while devel-
Fig. 2  Design freedom and knowledge trend in aircraft design oping methodologies for calculating sensitivity derivatives
(Mavris and DeLaurentis 2000) as well as a sequential approximate optimization module
(Grossman et al. 1990).
With the advent of the computational prowess, higher
design novelties are accompanied by certain shortcom- fidelity computational tools and mostly computational fluid
ings, the most prominent being the elevated flexibility of dynamics (CFD) methods started to emerge. Traditional
the wing that on one hand induces a material and/or geo- panel methods were gradually getting replaced by nonlinear
metric nonlinear behavior and on the other hand a closer aerodynamic solvers and on the other hand beam or plate
coupling between the structure and the surrounding fluid, structural representation techniques were substituted by
aggravating static and dynamic aeroelastic phenomena. global finite element method (FEM) models. Aeroelastic
Furthermore, the use of composite materials perplexes effects were also introduced in the formal coupled analyses,
the design optimisation procedure by vastly increasing the due to their criticallity in the aerodynamic and structural
design space via the introduction of new variables such as design. Maute et al. (2001) presented a methodology for
ply thickness and ply orientation. As a result, the develop- optimizing the aerodynamic and structural parameters of a
ment of novel design methodologies and computational steady-state aeroelastic system, using 3D Euler finite volume
tools capable of predicting the whole spectrum of phe- method (FVM) and FEM models. A plethora of novel solu-
nomena inherent to this new generation of structures as tion methodologies was developed and applied, aiming at
well as producing optimum configurations is of paramount the single and multi-objective optimization of various wing
importance. test cases. A highly efficient and robust Schur–Newton–Kry-
Unfortunately, contemporary high-fidelity tools are lov solution method to the coupled, steady-state aeroelastic
commonly associated with elevated and often prohibi- problem was given in Barcelos et al. (2006), with the meth-
tive computational time, hence their use is limited to the odology being extended to also include Reynolds averaged
later, more detailed, design stages. Since in these stages the Navier–Stokes (RANS) CFD (Barcelos and Maute 2008).
design freedom is bounded and any design alteration could Moreover, coupled aerostructural optimization studies
induce severe cost penalties, as illustrated in Fig. 2, low and started to draw the attention of the scientific community,
medium fidelity tools, capable of representing a portion of with Reuther et al. (1999) presenting an integrated, high-
the phenomena that might rise, need to be exploited for the fidelity aerostructural optimization framework of complete
conceptual and preliminary design stages. In this manner, aircraft configurations. Several optimization studies were

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 3 of 23 150

conducted and optimal configurations of complete configu- managed to effectively balance aerodynamic and structural
ration flight and wind-tunnel models were obtained. Later performance, despite the vastness of the design space and
on, Martins et al. (2004) proposed a coupled sensitivity the constraints posed at each discipline. In particular, a 4.2%
analysis of aerostructural systems, demonstrating the accu- reduction in takeoff gross weight with a corresponding 6.6%
racy and efficiency of the coupled-adjoint method by opti- fuel burn reduction were accomplished for the first of the
mizing the aerostructural design of a supersonic business jet. two optimization problems, whereas for the second one a
Nearly concurrently, exploratory efforts on unconventional 11.2% fuel burn reduction with no significant change in the
configurations commenced, with emphasis given on high- takeoff gross weight was observed. A similar study for the
aspect ratio wings due to their increased aerodynamic per- single and multiple point aerodynamic optimization of the
formance. Smith et al. (2001) underlined the necessity of CRM wing was also conducted in Lyu et al. (2015). The
high-fidelity CFD in the aeroelastic modeling of high-aspect geometry of the wing was paremetrized using the free-form
ratio wings, since significant differences emerged in com- deformation (FFD) technique (Kenway et al. 2010), resulting
parison to linearized methods. Particularly, linear methods into 720 variables. The RANS equations constituted the
tend to underestimate the tip bending and twist, leading to aerodynamics model, with the resulting aerodynamic shape
overly conservative divergence and flutter speeds predic- optimization problem being handled via the MACH frame-
tions. Furthermore, in their work Patil et al. (2001) demon- work. The minimization of the drag coefficient subject to lift,
strated a dramatic reduction in flutter speed due to the non- pitching moment, and geometric constraints formed the
linear coupling among edgewise bending and torsion as well objective function and constraints of the optimization prob-
as in the flight dynamic characteristics of an aircraft due to lem respectively. For the single point optimization, a drag
the wing flexibility, vastly affecting the trim solution. The reduction of 8.5% was achieved, while a more robust design
differences between a linear and a nonlinear beam model of was obtained through a multipoint optimization. Various
a high-aspect ratio wing coupled with RANS CFD at the randomly generated starting points were also examined in
transonic regime were also presented in Garcia (2005). The order to assess the multimodality of the design space, with
static aeroelastic response of unswept and swept configura- the resulting geometries pertaining similar geometrical char-
tions was investigated, indicating a reversal in wing twist acteristics. The effect of varying the number of shape design
due to nonlinear torsion-bending coupling effects for the variables was also investigated. Various aeroelastic mode-
former and a reduction in the amount of washout, in com- ling techniques of high-aspect ratio wings are demonstrated
parison with the linear solution, for the latter. Recently, Cog- in Howcroft et al. (2016) and Castellani et al. (2017). Spe-
gin et al. (2014) studied the aeroelastic response of a truss- cifically, the linear models tend to overestimate the wing
braced wing configuration. The NASTRAN solver was deflection, bending moment and lift force. The large influ-
modified in such manner that the nonlinear aeroelastic ence of the drag force, often disregarded in panel aerody-
response of the wing was accounted for, with the resulting namics, on the increase of the resulting twisting moment is
pre-stressed stiffness matrix being used to calculate the pre- also highlighted. The research efforts of Calderon et al.
stressed mode shapes which in turn were fed to the flutter (2018, 2019) focused on the ramifications of including the
analysis module. In Bartels et al. (2015), the flutter and limit geometrically nonlinear behavior of the structure on the
cycle oscillation (LCO) of a similar configuration was inves- structural optimization. Based on a coupled geometrically
tigated using low-fidelity panel methods as well as CFD nonlinear VLM and beam models, a series of studies were
simulations. Both studies indicate a discrepancy in the flutter conducted, obtaining optimum panels and stiffeners configu-
speed calculation and LCO onset between the linear and rations under strength, skin buckling and skin-stringer fail-
nonlinear approaches, stressing the necessity for the inclu- ure constraints. For a baseline wing with an aspect-ratio of
sion of nonlinear structure and aerodynamic effects for the 18, a lighter configuration of 5% emerged for the geometri-
accurate prediction of the aeroelastic behavior of the truss- cally nonlinear model, with the root bending and torsional
braced wing. In parallel, the development of high-fidelity moment being reduced by 8% and 50% respectively. A mod-
aerostructural optimization tools, employing hundreds of est 1.3% improvement in the Breguet range was also noted.
variables and constraints was also of great interest to the In the latter study, the effect of the aspect-ratio using linear
scientific community. An approach towards the resolution of and nonlinear models on the structural sizing was also inves-
such vast optimization problems was conducted in Kenway tigated. Wings optimised using geometrically nonlinear
and Martins (2014), where a multipoint high-fidelity aero- analysis emerged lighter with similar aerodynamic effi-
structural optimization of the common research model ciency, by means of lift-to-drag ratio, and therefore greater
(CRM) was presented. The optimization problem was Breguet ranges than those optimized using a linear approach.
divided into two problems, aiming at the takeoff gross Of particular interest is the fact that due to the overestima-
weight and fuel burn minimization under five cruise and two tion of loads in the linear analysis the optimum aspect ratio
maneuver conditions respectively. The proposed framework increases from 18 to 19, when accounting for geometric

13
150 Page 4 of 23 S. Kilimtzidis, V. Kostopoulos

nonlinearities. The inclusion of geometrically nonlinear flut- a common way of optimizing composite structures involves
ter constraints was accomplished in Lupp and Cesnik (2019), the declaration of ply angles and thicknesses as design varia-
where a blended wing body (BWB) aircraft was optimized bles, resulting in a more complex and irregular design space,
under a bending curvature (static aeroelastic) constraint, a but with facilitated algorithmic applicability. On the other
linearized about the jig shape as well as a geometrically hand, optimization via lamination parameters, extensively
nonlinear flutter constraint. The addition of the flutter con- studied over the past years, is presented as an alternative.
straint in the aerostructural optimization showed a reduced Miki and Sugiyama (1993) was among the first to generate
wing span and aspect ratio when compared to the bending optimum designs of laminated composite plates via lamina-
curvature constraint. Introducing the geometrically nonlinear tion parameters for maximum in-plane and bending stiffness,
flutter constraint yielded a more conservative design with a buckling strength and natural frequency. Fukunaga et al.
higher fuel burn. The authors conclude that the necessity for (1994) explored the effect of bending–twisting coupling on
a geometrically nonlinear flutter constraint is certainly prob- the fundamental frequency of symmetric laminated plates
lem and configuration dependent. Straying off the vastly cast in the lamination parameters space, indicating that this
used beam models and raising the computational fidelity in type of coupling reduces the fundamental frequencies. An
terms of structural representation, Verri et al. (2020) coupled optimal laminate configuration for maximum fundamental
a full-order geometrically nonlinear wing box FEM model frequencies was also generated. The buckling load of a com-
of an Embraer regional jet with a wing aspect ratio of 12 to posite panel was maximized using flexural parameters in Liu
high-fidelity CFD. A 2.5 g pull-up maneuver limit load was et al. (2004). The resulting stacking sequence was compared
considered and the differences between the linear and non- to one generated via a genetic algorithm, indicating a close
linear structural behavior were underlined. In particular, and correlation between the two methods. In more recent works,
from the aerodynamics point of view, a modified pressure efforts were directed towards the implementation of lamina-
coefficient distribution acting on the wing due to the tip dis- tion parameters optimization frameworks for the aeroelastic
placement difference of 16% was obtained. Furthermore, as tailoring of regular and variable stiffness composite materi-
also observed in the majority of the aforementioned studies, als wings (Thuwis et al. 2009; IJsselmuiden et al. 2010), as
the lift force was shifted towards the wing tip direction due well as the stiffness optimization of composite wings subject
to the outboard rotation of the wing. The resulting internal to aeroelastic constraints (Dillinger et al. 2013). Blending
shear loads remained unaltered in the root, but higher in the constraints, guaranteeing a certain degree of ply continu-
outboard portion by 14% for the nonlinear case. Regarding ity inside a variable stiffness and thickness composite wing
the bending moment, a slightly higher at the root but much were studied in Macquart et al. (2018) as well as Bordogna
elevated in the outboard portion was observed. A complete et al. (2020), extending the capabilities of the state-of-the art
integrated aerostructural design and optimization effort was lamination parameters optimization algorithms.
performed in Gray (2021), where two modified versions of
the original CRM wing model were compared and opti- 1.2 Motivation
mized. Initially, and for a similar loading scenario, the non-
linear analysis resulted in greater bending stresses in the Despite the existence of a wide variety of low-fidelity aeroe-
upper and lower skins which in turn increased the Von Mises lastic optimization frameworks, studies towards high-fidelity
and buckling failure criteria by around 10% for the high- optimization frameworks including geometric nonlinearities
aspect ratio wing. The presence of the Brazier loads (1927) for composite materials high-aspect ratio wings are rarely
was also demonstrated, resulting in substantially greater documented in the literature. This knowledge gap is aggra-
compressive axial stresses in the ribs. Several structural and vated when examining sequential optimization frameworks,
aeroelastic design optimisation studies were subsequently combining low and high-fidelity computational tools. Aim-
executed, with the increase in bending stresses leading to a ing to address this issue, this research study bridges the
6% and 4% increase in mass for the optimised high and mod- various fidelities of the numerical tools involved in order to
erate aspect ratio wingbox respectively. A computational provide an efficient sequential structural optimization frame-
cost analysis was also included, indicating an order of mag- work of a high-aspect ratio composite materials aircraft
nitude increase when geometric nonlinearities are wing subject to local panel buckling, strength and flutter
considered. constraints, combining low and high-fidelity computational
The majority of the aforementioned studies steered its tools. The geometric nonlinear behavior of the structure as
attention mainly towards the identification of the influence well as the follower forces nature of the aerodynamic forces
of the geometrically nonlinear phenomena in the aerody- is accounted for at each stage of the optimization framework,
namic and structural disciplines, with any effort towards the since as demonstrated in the literature as well as in later
implementation of composite materials in the analysis and sections they alter the deformation and stress field, which
optimization frameworks being nearly absent. On that front, eventually translates into possible structural mass reduction.

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 5 of 23 150

Reference analyses are initially conducted in order to assess relevant geometric data are summarized in the following
the effect of fidelity of the associated computational tools Table 1.
on the structural response of the reference wing, guiding Regarding the internal configuration, two spars located
the development of the proposed optimization approach. at 10% and 60% of the local chord are present along with 54
Low-fidelity tools form the backbone of this optimiza- evenly distributed ribs and upper and lower skin stiffeners.
tion framework and are initially tasked with guiding the Spar and rib caps are also included withing the framework of
structural design towards promising regions of the design this study, in contrast to the internal configuration presented
space. Higher fidelity methods are then employed, aiming at in Brooks et al. (2018). These primary load-carrying compo-
exploring possible further gains in performance. The present nents are commonly found in a commercial airliner’s wing-
approach intends to contribute to sequential structural opti- box, hence their effect and influence in the overall stiffness
mization frameworks for composite materials future wing and strength of the structure were deemed necessary to be
configurations, which are less prominently covered in the investigated, as demonstrated in later sections. An exploded
literature. We demonstrate that, through the use of this opti- view of the internal structure is illustrated in the following
mization approach, reasonable reduction in structural mass Fig. 3. For the sake of clarity, the spar and rib caps as well as
of the test case wing can be realized. the stiffeners are depicted via spanwise straight lines.
For the subsequent analyses a 2.5 g pull-up maneuver
limit load, assumed to be exerted at the conditions indicated
2 Methodology in Table 2, is considered, with the angle of attack (AoA)
being modified accordingly at each level of fidelity in order
In this section, the various variable fidelity computational to generate the appropriate aerodynamic loading (with
tools used for the aerodynamics and structural disciplines MTOW being the maximum take-off weight of the aircraft,
are rigorously described. The methodology starts with the set to 268e3 kg Gray 2021).
particulars of the high-fidelity CFD analysis and proceeds
with the low-fidelity 3D panel and DLM method. Aspects 2.2 High‑Fidelity CFD
of the nonlinear FEM analysis are also manifested, which
will eventually lead to the core of this research work, the For the high-fidelity CFD analysis the compressible RANS
proposed sequential structural optimization framework. equations (Crovato et al. 2020; Economon et al. 2016) are
discretized and solved via the FVM in a C-grid shaped
2.1 Reference wing geometry

Within the context of this work, the undeflected CRM,

namely uCRM−13.5, has been chosen as the reference wing
model. Constituting a modified, high aspect ratio deriva-
tive of the original CRM wing (Vassberg et al. 2008), the
uCRM−13.5 model (Brooks et al. 2018) serves as a bench-
mark configuration for CFD and aerostructural optimiza-
tion studies of realistic, contemporary as well as future air-
craft configurations operating at the transonic regime. The

Table 1  uCRM−13.5 geometric data

Entity Value

Wingspan 72 m
Root chord 11.07 m
Tip chord 2.06 m
Fig. 3  uCRM−13.5 internal configuration and OML
Reference area 383.78 m 2
Taper ratio 0.25
Aspect ratio 13.5 Table 2  Critical aerodynamic loading summary
Quarter chord sweep angle 35◦
Condition Lift constraint Mach Altitude (m)
Yehudi chord 7.56 m
MAC 5.77 m 2.5 g maneuver 2.5 ⋅ MTOW 0.64 0

13
 150 Page 6 of 23 S. Kilimtzidis, V. Kostopoulos

domain, as presented in Fig. 4 along with the relevant has been chosen for the evaluation of gradients. The resulting
dimensions and boundary conditions. Turbulence in the structured hexahedral CFD mesh is shown in Fig. 5. For verifi-
form of RANS methods is also modeled and introduced cation purposes, a mesh convergence study for the major aero-
via the Spalart–Allmaras one-equation turbulence model dynamic lift, drag and pitching moment coefficients, CL, CD
(1992) along with wall functions providing near wall treat- and CM respectively, has also been conducted and is presented
ment. A first cell wall distance, y+, necessary for capturing in the subsequent Table 3. The resulting aerodynamic loads are
the evolution of the boundary layer using RANS turbulence transferred to the structural FEM mesh via interpolation tech-
models, in the region of 50–70 was targeted due to the size niques, as discussed in Kilimtzidis (2022) at a greater extent.
of the wing, constituting the use of lower values prohibitive
in terms of computational cost. Given the target y+ value, as 2.3 3D panel method
well as the Reynolds number of the analysis, which in our
case is equal to 35.5e6 , the first layer height is calculated On the medium to low-fidelity aerodynamic analysis front, an
based on equations derived from analyses of the bound- in-house 3D panel method framework has been developed in
ary layer development in flat plates, as described in White MATLAB. This module accepts a discretized geometry of an
(2010). Moving on, and to account for temperature induced aircraft wing along with the reference flight conditions and
changes in viscosity, Sutherland’s law, based on the kinetic calculates several aerodynamic related quantities. In particular,
theory of ideal gases and an idealized intermolecular-force and under the assumption of an inviscid and irrotational flow,
potential, is used. there exists a potential function 𝛷 that satisfies the Laplacian
The spatial terms of the Navier–Stokes equations are discre- equation (Katz and Plotkin 2001):
tized on the computational finite volume mesh with appropri-
ate discretization schemes. In particular, the spatial convection ∇2 𝛷 = 0. (1)
terms are discretized using second order upwind schemes, mit- Applying Green’s third identity and the impermeability
igating numerical oscillations while maintaining good accu- boundary condition on the wing’s surface, while introduc-
racy, while spatial diffusion and source terms are discretized ing sources and doublets as elementary solutions, denoted as
using central differences. Moreover, a Green–Gauss method 𝜎 and 𝜇p respectively, Eq. 1 is recast in the following form:

∫ ∫
1 1 1 1
( ) ( )
Table 3  CFD analysis mesh convergence study 𝛷= 𝜇p n ⋅ ∇ dS − 𝜎 dS + 𝛷∞ . (2)
4𝜋 wing+wake r 4𝜋 wing r
Mesh CL CD CM
With n being the surface normal vector, r the distance vector
Coarse (8.7e 6 cells) 0.549 0.0277 1.61 and 𝛷∞ the free-stream potential function. For a discretized
Medium (12e 6 cells) 0.5536 0.027 1.622 surface consisting of N wing and Nw wake panels, integration
Fine (15.6e 6 cells) 0.5668 0.02689 1.6713

Fig. 4  CFD C-shaped domain

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 7 of 23 150

Fig. 5  CFD structured FVM

mesh

of Eq. 2 along with application of the Kutta condition to the transfer to the structural FEM mesh and for computational
trailing edge panels can be further simplified into: efficiency purposes, the resulting aerodynamic panels and
Npanels Npanels
structural meshes are coincident, allowing for a direct load
transfer between the two disciplines. The results of the 3D
(3)
∑ ∑
A1k 𝜇k + B1k 𝜎k = 0,
k=1 k=1
panel method developed within the framework of this study
were also verified and validated against the ONERA M6
where A1k and B1k doublet and source influence coefficient wing case study (Schmitt and Charpin 1979) and are pre-
for the kth panel respectively. Setting the source terms on sented in Kilimtzidis (2022). The resulting aerodynamic
the right-hand-side of Eq. 3 equal to 𝜎k = nk ⋅ U∞, a set of mesh is displayed in the following Fig. 6.
linear equations that can be solved for the unknown doublet
distribution is obtained. Making use of spatial interpolation 2.4 DLM aerodynamics
schemes one can calculate the induced velocities and via the
Bernoulli equation the pressure coefficients on each panel. Three-dimensional panel methods are particularly advan-
To account for possible compressibility effects in the flow, tageous in problems where thick airfoil configurations
the Prandtl–Glauert correction factor has been also imple- exist. When the airfoil thickness is adequately small (less
mented, as in Katz and Plotkin (2001). Additionally, the than 12%), further simplifications can be applied, reducing
wake of the wing has been considered to be fixed and extend- the dimensionality of the aerodynamic surface to planar
ing in the chordwise direction, since wake shape calculations representations. Among of the existing panel methods,
are beyond the scope of this research study. In terms of load the DLM constitutes the cornerstone for aerodynamic

Fig. 6  3D panels mesh

13
150 Page 8 of 23 S. Kilimtzidis, V. Kostopoulos

calculations at this level of fidelity. A comprehensive

work on the mathematical derivation of the method is pro-
vided in Blair (1992). Aerodynamic surfaces are typically
divided into small trapezoidal panels and a constant pres-
sure distribution is assumed. The DLM is capable of mod-
eling unsteady flows in the frequency domain and is based
on the linearized small disturbance potential flow equa-
tion, hence neglecting large perturbations and shockwaves:
( ) ( )
2 2
2
M∞ M∞
(1 − M∞ )𝛷xx + 𝛷yy + 𝛷zz − 2 𝛷xt − 𝛷tt = 0,
U∞ U∞
(4)
where M∞ and U∞ the free-stream Mach number and veloc-
ity respectively and 𝛷xx , 𝛷yy , 𝛷zz , 𝛷xt , 𝛷tt second deriva-
tives of the potential function. Assuming small amplitude,
harmonic motion, solution to Eq. 4 is given by the so-called
acceleration potential and by introduction of doublets across
the discretized planar lifting surface. The doublet line of
constant value is placed at the 1/4th of the chord of each
panel, while the normalwash collocation point, where the
impermeability boundary condition is applied, is placed at
the 3/4th of the chord. The resulting pressure coefficient for
a panel i, denoted as p̄ i , is expressed in terms of a complex
matrix Aij (Albano and Rodden 1969), including the contri-
bution of the generated downwash from panel j to panel i, Fig. 7  uCRM−13.5 wing DLM mesh and corresponding W2GJ
and the downwash value w̄ i at the current panel of interest: entries

Npanels
∑
p̄ i = Aij w̄ j = [AIC(M∞ , k)ij ]wj . (5) • For each section and panel of the aerodynamic mesh:
j=1

Complex valued matrix Aij , widely termed as the aerody-

– Stage 1
namic influence coefficient (AIC) matrix, is now subject to
evaluation given the discretized aerodynamic surface, the
operating Mach number as well as the reduced frequency k.
• Extract upper and lower airfoil profiles for the current
The DLM aerodynamic surfaces of the uCRM−13.5
section.
wing are modeled in NASTRAN, with the resulting aero-
• Initially estimate the camber line equation as the mid-line
dynamic mesh being presented in Fig. 7 along with the
between the upper and lower airfoil sections.
calculated collocation points. Despite the attractive char-
• For each panel of the current section calculate the col-
acteristics of the DLM aerodynamic modeling, certain
location point and the derivative of the camber line.
corrections need to be applied in order to replicate the
actual airfoil geometry to a greater extent. Inclusion of the
– Stage 2
realistic camber line and twist angle of the corresponding
airfoil, allowing for a proper calculation of the aerody-
namic coefficients of the wing at various angles of attack
• Given the initial camber line points draw tangent and
is often sought. This is accomplished by populating the
perpendicular secant lines.
W2GJ entry vector in NASTRAN with the values of the
• Calculate intersecting points between the aforementioned
slope of the camber line of each airfoil of the wing under
lines and the upper and lower airfoil sections.
consideration calculated at the collocation points of the
• For each intersecting set of points, equidistant points
aerodynamic panels. For the calculation of the relevant
define the new camber line.
correction factors for the uCRM−13.5 wing, a two-stage
procedure, similar to the one presented in Demirer (2021)
The resulting W2GJ distribution for the current aerody-
has been implemented herein:
namic mesh is shown in the subsequent Fig. 7. Regarding

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 9 of 23 150

the coupling between the aerodynamic and structural mesh,

Ply thickness (m)

infinite plate splines have been used in this study to trans-
fer the aerodynamic loads to the structure. For the splining

1.8288e−4
1.9281e−4
procedure, nodes close to the ribs and spars as well as to the
DLM mesh have been chosen, allowing for a proper load
distribution and avoiding possible local loading scenarios.
For the sake of completeness, a verification and validation

𝜌 (kg/m 3)
analysis of the DLM aerodynamic model as well of the flut-

1590
1570
ter analysis has been conducted (Kilimtzidis et al. 2018).

2.5 FEM model development

(MPa)

53.5
56
S
The well-established and reliable FEM analysis has been
vastly used among researchers and the aerospace industry

YC (MPa)
to model structures and to predict their behavior under the

285.7
application of loads. The accuracy of the method for wing

752
structures has been validated against experimental data in
a plethora of literature studies (Ritter et al. 2021; Dessena

YT (MPa)
et al. 2022; Keimer et al. 2022). In our case and based on the

640.5
742
external geometry and internal configuration of the reference
wing, a FEM model has been developed. Specifically, the
upper and lower skins along with the spar and rib webs are

XC (MPa)
modeled via 4-noded quadrilateral shell elements, denoted

793.8
1531
as CQUAD4. On the other hand, the spar and rib caps as
well as the skin stiffeners, are assumed, as per common prac-
tise, to be modeled via beam elements (denoted as CBEAM)

XT (MPa)
along with the relevant offset values. As a datum design

751.6
2500
point, a reference balanced and symmetric layup and specifi-
cally the [(45∕0∕ − 45∕90)2s ]s , has been considered for all
the relevant wing parts. Furthermore, similarly to Brooks
0.315
0.04
et al. (2019, 2020), the upper and lower skins are assumed
𝜈12

to be manufactured of the Hexcel IM7/8552 UD (Marlett

2011a) composite material, while the spars and ribs of the
G13 (GPa)

PW Fabric AS4/8552 (Marlett 2011b), with the respective

4.68
4.96

material properties, cured ply thickness, density as well as

strength values listed in Table 4. Simple rectangular cross-
sections are chosen for all the relevant beam elements, with
G12 (GPa)

the corresponding height values being prescribed. The thick-

4.68
4.96

ness, however, is calculated based on the aforementioned

lay-up. Since only isotropic materials are allowed for the
definition of the CBEAM elements and the baseline lay-up
E2 (GPa)

is symmetric, equivalent laminate axial and shear moduli,

65.57
8.96

Eeq and Geq respectively, can be calculated based on the fol-

Table 4  Composite materials properties

lowing equations (Kassapoglou 2013):

E1 (GPa)

( )
158.51

A212 A
64.53

1
Eeq = A11 − Geq = 66 , (6)
t A22 t

where t the thickness, A11, A12 , A2 and A66 the correspond-

Hexcel IM7/8552

ing terms of the extensional stiffness matrix of a laminate.

PW AS4/8552

Regarding the boundary conditions, the wing is assumed to

Material

be clamped at its root, thus fixing all relative nodal degrees

of freedom (DoF). For the sake of completeness, external

13
150 Page 10 of 23 S. Kilimtzidis, V. Kostopoulos

mass as well as gravitational loads have been accounted for.

In particular, a fuel load of 56.000 kg (Brooks et al. 2018)
has been considered as traction loads and introduced in the
model at each wingbox bay via RBE2 elements acting on the
lower skin. Furthermore, the engine mass, equal to 2000 kg,
is also modeled and connected to the wingbox via a combi-
nation of concentrated mass and RBE2 and RBE3 elements.
The resulting FEM model of the uCRM−13.5 wing is shown
in the subsequent Fig. 8.
The NASTRAN solvers SOL 101 and SOL 400 (Lee
1992) were chosen to carry out the linear and nonlinear
static analysis respectively. In order to avoid any possible
divergence issues mainly during the optimization procedure,
a full Newton–Raphson stiffness matrix update technique
has been preferred for the nonlinear analyses.
Fig. 9  Global–local modeling approach

2.6 Global–local panel buckling analysis

and at this level, blade stiffeners are considered and modeled
Due to the criticality of the buckling phenomenon and its explicitly via CQUAD4 elements. Buckling analyses are exe-
impact on the design of aircraft wings, a global–local FEM cuted for each of the 106 panels present in the wing via the
modeling technique was implemented for the panels of the SOL 105 solver implemented in NASTRAN, with the first
upper and lower skins within the framework of this study. eigenvalue, namely the critical buckling load expressed as a
A panel is considered as the intersection between the front percentage of the originally applied load, being the quantity
and rear spar as well as two adjacent ribs, as indicated in of interest (NASTRAN 2021).
Fig. 9. As far as the global–local modeling technique is con-
cerned, the resulting nodal displacements at each of the four
edges of each panel of the global FEM act as input boundary 3 Reference analyses
conditions to the local level FEM analysis. Since dissimilar
meshes are considered, interpolation schemes are employed Prior to the optimization study, reference analyses are con-
at each edge of each panel in order to calculate the local ducted in order to ascertain the differences among each
displacement field at intermediate positions. Furthermore, level of fidelity on the structural response of the uCRM−
13.5 wing. Various levels of fidelity have been considered
from an aerodynamics and structural point of view, with the
relevant analyses conducted being listed in Table 5.
In all types of analyses, the AoA is modified accord-
ingly in order to attain the required loading conditions, as
described in Table 2 and listed in Table 6.
Significant changes in AoA are required for the various
aerodynamic analysis tools in order achieve a similar lift
load. For the 3D panel method a smaller angle is required
since pressure recirculation as well as flow separation
near the trailing edge phenomena are not captured in such

Table 5  Reference analyses summary

Analysis ID Aerodynamic fidelity Structural fidelity

1 CFD Linear
2 CFD Nonlinear
3 3D panel method Linear
4 3D panel method Nonlinear
5 DLM rigid Linear
6 DLM elastic Linear
Fig. 8  uCRM−13.5 wing FEM mesh and fuel distribution detail

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 11 of 23 150

Table 6  AoA change Substantial differences in the results can be observed

Aerodynamic fidelity CFD 3D panel method DLM depending on the level of fidelity of the aerodynamics as
well as of the structural representation of the uCRM−13.5
AoA ( ◦) 2 1.8 3.5 wing configuration. For the former, and despite acquiring
similar loading conditions, the 3D panel method seems to
be overestimating the maximum deflection and tip torsion
methods, resulting in higher pressure difference at the outer induced at the wing. Pressure recirculation as well as flow
regions of the wing. It should be point out, nevertheless, that separation near the trailing edge effects are not captured in
a flat wake has been assumed throughout the analysis. Wake such methods as stated earlier, resulting in higher pressure
shape calculation techniques are reported in Katz and Plot- difference and eventually loads at the outer regions of the
kin (2001) and influence the calculation of the aerodynamic wing, which in turn are responsible for this increase. Higher
coefficients. For the DLM, the camber correction method bending moments induce a higher stress field and lower the
presented earlier is not sufficient to obtain the required lift critical buckling eigenvalues, as seen in Table 7. On the con-
load at the high-fidelity AoA, the main culprit being thick- trary, the DLM seems to be more accurately predicting the
ness effects, which are not captured in this method, that displacement and stress field when compared to the high-
allow for a greater pressure difference between the upper fidelity CFD linear analysis, as indicated in Fig. 10a and
and lower surfaces of each wing section. b. Of particular interest is the aeroelastic behavior of the
The induced displacement field in terms of deflection and uCRM−13.5 wing, allowing for aerodynamic load redis-
tip torsion angle, the static strength as well as the buckling tribution and overall lower displacement and stress fields
critical loads for the upper and lower skins constitute the due to the bending–torsion coupling phenomenon. From a
quantities of interest for the analyses. On the static strength structural point of view, inclusion of nonlinearities affect
front, a first-ply failure (FPF) via the Tsai–Wu criterion the solution and result in lower displacements and stresses
(Jones 2018; Tsai and Hahn 2018) is used for the strength as expected mainly due to the geometric stiffening of the
prediction in terms of the failure index (FI) of the composite structure. This effect is on the contrary less pronounced for
skins, spar and rib webs. For the rest of the parts, a direct the torsion angle, as depicted in Fig. 10c. Other nonlinear
comparison between the maximum stress and the corre- phenomena also rise, with one of the most prominent being
sponding material strength value is made in order to obtain the tip shortening effect, as illustrated in Fig. 10b. Moreo-
the FI for the wing component under consideration. As far ver, the presence of Brazier loads drastically alters the stress
as the buckling critical load is concerned, buckling analyses field in the rib caps, as demonstrated in Table 7, resulting in
are executed for each panel of the upper and lower skin with highly elevated stresses by nearly 50% in comparison with
the first buckling eigenvalue (BE) being extracted. The mini- the linear models.
mum eigenvalue should then be greater than unity in order to The distribution of the maximum FI per ply for the linear
avoid any buckling failure. To avoid local maximum stress and nonlinear structural analyses under the CFD loading,
driven designs, while simultaneously keeping the number along with the relative differences is presented in Fig. 11a.
of constraints for the optimization problem to a minimum, The resulting distributions are quite similar, with the major-
the constraint aggregation technique of Kreisselmeier–Stein- ity of the differences lying near the Yehudi break of the
hauser (KS) has been employed (Kreisselmeier and Stein- wing. In particular, the reduction of the effective span of
hauser 1979; Poon and Martins 2006) for the static strength the wing as well as the rotation of the outboard sections of
as well as buckling constraints. For a problem consisting of the wing reduce the moment arm and the bending moment,
Nc quantities of interest, g, having a maximum value of gmax, which cascade in lower stress values near this spanwise
the KS functions are of the following form: region. A similar situation for the 3D panels method is por-
[N ] trayed in Fig. 11b. In terms of the effect of the aerodynamic
loading on the distribution of the FI, the 3D panel method
c
1 ∑
(7)
( )
KS gj = gmax + ln exp(𝜌(gj − gmax ))
𝜌 j
overpredicts the induced root bending moment and as a
result the developed stress field, as indicated in Fig. 11c.
and are formed separately for the calculated FI of each of Regarding the computational time, a wall-clock time
the components involved in the static strength as well as for comparison for the linear and nonlinear solvers has been
each upper panel buckling critical load evaluation procedure. conducted for the reference analyses presented. In particular,
The aggregation parameter 𝜌 is set to 100 for all KS func- the wall-clock time was 5.68 min and 17.8 min for the linear
tions. This value has been reported in structural optimiza- and nonlinear solver respectively, indicating the superiority
tion studies of the CRM wing (Lambe et al. 2016; Lambe of the linear solver in terms of computational efficiency. In
and Martins 2015) and is shown to provide accuracy of the terms of convergence, nonlinear analyses at specific points
optimal solution. of the design space indicated divergence issues for the spar

13
150 Page 12 of 23 S. Kilimtzidis, V. Kostopoulos

Fig. 10  Reference analyses overall comparison

caps CBEAM elements near the yehudi break of the wing variables of the optimization problem. Setting the num-
at certain points of the design space, even for a full New- ber of 45◦ and −45◦ plies equal implies generation of sym-
ton–Raphson stiffness matrix update technique. As a result, metric and balanced lay-ups, thus achieving conformity
the lower bounds of those elements were increased in order to composite materials design guidelines (Kassapoglou
to avoid divergence issues during the optimization process. 2013). Within the framework of this optimization study,
the wing is also divided into eight spanwise and evenly
spaced zones, as illustrated in Fig. 12, allowing for a wider
design space as well as increased structural design free-
4 Optimization framework dom. A summary of the variables present along with the
respective lower and upper bounds as well as their type is
The particulars of the optimization framework are discussed provided in Table 8. It should be pointed out that the two
in detailed fashion in this section, starting from the defini- materials presented in Table 4 are also used as variables
tion of the variables and proceeding with the formulation of and can be assigned during the optimization process to the
the optimization problem. rib webs and caps, spar webs and caps, upper and lower
skin as well as the stringers. This particular parametriza-
tion allows for a rapid cost estimation for each part of the
4.1 Variables definition wingbox and provides the designed with greater freedom
with respect to the material of each component.
The ply count of the 0◦ , 90◦ and 45◦ ∕ − 45◦ plies of each
laminate and for each wing component constitute the

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 13 of 23 150

Fig. 11  FI distribution and rela-

tive differences for the reference
analyses

13
150 Page 14 of 23 S. Kilimtzidis, V. Kostopoulos

Table 7  Reference analyses results summary

Value CFD linear CFD nonlinear 3D panels linear 3D panels DLM rigid DLM elastic
nonlinear

KS(FI), upper skin 0.3758 0.3554 0.5315 0.5395 0.4096 0.0869

KS(FI), lower skin 0.4398 0.4108 0.6089 0.5897 0.4683 0.1036
KS(FI), spar webs 0.1774 0.176 0.2407 0.2421 0.1866 0.0444
KS(FI), spar caps 0.2516 0.2447 0.3088 0.3115 0.2638 0.108
KS(FI), rib webs 0.5389 0.5182 0.6454 0.6287 0.5578 0.2374
KS(FI), rib caps 0.019 0.0366 0.0217 0.0442 0.0209 0.011
KS(FI), stringers 0.2058 0.2082 0.2449 0.2585 0.2079 0.0901
Maximum deflection (m) 8.861 8.39 11.152 10.794 9.745 4.42
Tip torsion ( ◦) 6.3441 5.8798 9.6973 9.422 6.2631 0.0247
KS(BE), upper skin 2.7280 1.8893 2.6316 1.8542 2.754 4.5364
KS(BE), lower skin 2.4414 1.8506 2.3975 1.8218 2.4798 3.6822

Fig. 12  uCRM wing optimiza-

tion zones

Table 8  Optimization variables bounds and type Furthermore, the static strength and critical buckling load
Variable Lower bound Upper bound Type
of the upper and lower skin panels for each candidate design
are also included and expressed in terms of the KS func-
Ply count, 0◦ 2 5 Integer tions, as described in earlier sections. Completing the set
Ply count, 90◦ 2 5 Integer of constraints, the dynamic aeroelastic instability by means
Ply count, (45◦, −45◦) 2 5 Integer of flutter velocity of the candidate design solutions is also
Material ID 1 2 Integer investigated. In particular, the p−k method implemented in
NASTRAN SOL 145 is used to identify any possible diver-
gence and flutter instability that might be present. For the
4.2 Objective function and constraints purpose of the analysis, a set of 20 reduced frequencies are
also used to calculate and interpolate in user-defined veloci-
In our study the minimization of the mass of the wing rep- ties the reduced-frequency-dependent aerodynamic loads.
resents the objective function. Design constraints in terms Possible flutter instabilities are investigated for the first ten
of static strength, buckling and flutter velocity, formulate structural modes via the velocity–damping (V−g) and veloc-
the optimization problem. Since no explicit static stiff- ity–frequency (V−f ) plots. In particular, the trends of each
ness requirements do exist in the literature for this genera- mode are monitored regarding the damping values, with
tion of aircraft wings, such constraints are not imposed. positive damping indicating a possible flutter instability. The

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 15 of 23 150

corresponding flutter velocity is then calculated via linear sequential approach with multiple runs has been adopted.
interpolation between the previous and subsequent velocities On that end, the predefined number of executions is
and damping values. In case of no flutter point, the flutter divided into runs pertaining different algorithmic param-
velocity is set to a value outside of the velocity range of the eters as well as level of fidelity. Particularly, initial runs
analysis. Classical infinite plate splines, similar to the ones are executed via the 3D panel method and mainly focus
described earlier, have been used to transfer the aerodynamic on extensive design space exploration, bounding the
loads to the structure. The objective function as well as the structural design towards promising regions of attraction.
constraints are summarized in the following Table 9. From a structural point of view, and based on the results
presented earlier, the inclusion of nonlinearities even at
4.3 Optimization algorithm parameters this design stage strongly affects the structural behavior
of the reference wing and are therefore deemed necessary
Within the framework of our study, the MIDACO solver throughout the design stages. Early runs are also accom-
(Schlüter et al. 2013) has been chosen to carry out the panied by a relaxed constraint satisfaction tolerance. As
optimization problem. MIDACO adopts a combination of the solution advances, higher-fidelity CFD solutions are
an extended ant colony optimization algorithm (ACO; employed, while the search becomes increasingly local.
Schlüter et al. 2009) along with the Oracle Penalty This is accomplished by modifying accordingly the inter-
Method (Schlüter and Gerdts 2009), an advanced method nal FOCUS parameter that forces the MIDACO solver to
developed for metaheuristic search algorithms for con- focus mostly on the current best solution. In particular,
straint handling of the solution process. Despite the abun- the ACO algorithm implemented in MIDACO generates
dance of optimization algorithms in the literature, the samples of iterates based on multi-kernel Gaussian prob-
nature of the optimization problem, featuring black-box ability density functions (PDF). For a generic variable k
objective function and constraints along with discrete with upper and lower bounds xu and xl respectively, the
variables, as described in Sect. 4.2, limit the applicability FOCUS parameter applies an upper bound for the stand-
of many of the optimization algorithms present, such as ard deviation of a Gaussian PDF given by uFOCUS and
x (k)−xl (k)
� �
gradient-based ones. Despite their reduced efficiency xu (k)−xl (k)
max FOCUS , √ 1
for continuous variables and inte-
with increased number of variables, gradient-free algo-
FOCUS
ger variables respectively. As a result, smaller values of
rithms (e.g. simulated annealing, genetic algorithms) tend
the FOCUS parameter is recommended for the initial
to make less assumptions about the modality and smooth-
runs, with larger ones used for refinement purposes. In
ness of the design space, thus present increased robust-
parallel, the constraint satisfaction tolerance is tightened.
ness characteristics (Martins and Ning 2021). Among
At each succeeding run, the previous best solution
these algorithms, the MIDACO algorithm has proven its
obtained serves as the starting point for the current run,
efficiency and accuracy in a plethora of mathematical
with this procedure being repeated for a predefined num-
benchmark optimization problems as well as in engineer-
ber of iterations, satisfying user-defined stopping criteria.
ing applications (Schlueter 2014; Kontogiannis and Savill
The number of iterations was defined based on the com-
2020). Regarding the solution procedure and in order to
putational resources available, while the associated
increase the effectiveness of the optimization problem, a
parameters adopted for the current optimization problem
were selected based on a similar study (Kilimtzidis et al.
Table 9  Optimization problem setup 2023). The cascading runs approach along with certain
Objective function Minimize structural mass
algorithmic values is also suggested in MIDACO (2018).
Under the constraints
Overall, the optimization algorithm parameters are sum-
marized in Table 10:
Constraint type Limit value A flowchart of the proposed optimization framework is
KS(FI), upper skin ≤1 also provided in Fig. 13.
KS(FI), lower skin ≤1
KS(FI), spar caps ≤1
KS(FI), spar webs ≤1
KS(FI), rib webs ≤1 5 Results
KS(FI), rib caps ≤1
KS(FI), stringers ≤1 The convergence of the objective function for the sequential
KS(BE), upper skin ≥1 optimization approach is presented in Fig. 14. Convergence
KS(BE), lower skin ≥1 to a minimum mass, along with non-violating design con-
Flutter speed ≥ 1.2⋅ Dive speed straints, has been achieved. In particular, significant mass

13
150 Page 16 of 23 S. Kilimtzidis, V. Kostopoulos

Table 10  Optimization Run Iterations Tolerance FOCUS Starting point Aerodynamic fidelity Structural fidelity
algorithm parameters
1 200 0.001 0 From scratch 3D panel method Nonlinear
2 100 0.001 10 Previous solution 3D panel method Nonlinear
3 100 0.0001 100 Previous solution CFD RANS Nonlinear
4 50 0.0001 −1000 Previous solution CFD RANS Nonlinear

Fig. 13  Sequential optimization framework flowchart

Table 11  Optimization results summary

Mass (kg) 9.2068e4

KS(FI), upper skin 0.2065
KS(FI), lower skin 0.4986
KS(FI), spar caps 0.187
KS(FI), spar webs 0.0987
KS(FI), rib webs 0.5079
KS(FI), rib caps 0.049
KS(FI), stringers 0.1865
KS(BE), upper skin 1.0375
KS(BE), lower skin 1.0262
Flutter speed (m/s) Flutter-free
Fig. 14  Wing mass convergence

reduction is obtained by employing high-fidelity CFD solu-

tions, since the corresponding loading condition results into
lower stress levels, which eventually translate into reduced

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 17 of 23 150

Fig. 15  Optimal design—component-wise mass

Fig. 17  Optimal design—maximum failure indices

Fig. 16  Optimal design—critical panel buckling load factor distribu-

tion

structural mass. The corresponding optimal mass, as well as

design constraints at this solution are presented in Table 11. Fig. 18  Optimal design—deflected shape
The resulting mass distribution for each wing compo-
nent, along with the corresponding material, is provided in
Fig. 15. The PW Fabric material is assigned to the majority The thickness distribution among the wing parts is pro-
of the parts, resulting in lower wing mass due to its lower vided in Table 12 and also visualized in Fig. 19. As a general
density value. On the other hand, and despite the associ- trend, a spanwise thickness decrease can be observed for
ated higher density, the Hexcel IM7 UD is assigned to the the majority of the wing components, specifically occur-
lower skin due its superior performance in tension in com- ring at the outboard section of the wing, since the bend-
parison with the PW Fabric. On the constraints front, the ing moment is decreased towards the wing tip. The region
resulting optimal design is clearly stiffness driven, since the around the yehudi break, (Zones 3 and 4) appears to be the
aggregate minimum buckling eigenvalue for the upper and highest stressed one in contrast with the root of wing, and
lower skin closely approximate their limit value. As illus- is therefore is accommodated with elevated thickness val-
trated in Fig. 16, the critical panel buckling load decreases ues. On a component-wise analysis, the spar caps and skins
in the spanwise direction, indicating that the outboard pan- obtain the highest thickness values, increasing the second-
els are more prone to buckling. On the other hand, strength ary moment of inertia of the cross-sections, thus allowing
constraints are less tight, indicating possible further mass for greater stiffness of the wing. On the contrary, the front
reduction. However, additional strength safety factors are and rear spar webs are assigned with lower thickness values,
commonly introduced to account for various stress risers, since they are closer to the neutral axis of bending of the
such as manufacturing defects, holes, impact phenomena, wing resulting in lower stresses. Shear forces and induced
etc. The developed stress field, by means of the maximum FI torsional moments, however, highlight the need for elevated
as well as the deflected wing are depicted in Figs. 17 and 18. thicknesses even for these components. Of particular interest

13
150 Page 18 of 23 S. Kilimtzidis, V. Kostopoulos

Table 12  Optimization Component Zone 1 Zone 2 Zone 3 Zone 4 Zone 5 Zone 6 Zone 7 Zone 8
results summary—thickness
distribution (m) Upper skin 0.0176 0.0205 0.0263 0.0219 0.0176 0.0176 0.0146 0.0146
Lower skin 0.0254 0.0269 0.0317 0.0254 0.0206 0.0206 0.019 0.019
Front spar webs 0.0176 0.0205 0.0219 0.019 0.0176 0.0161 0.0146 0.0146
Front spar caps 0.0263 0.0278 0.0322 0.0263 0.0249 0.0249 0.0234 0.0205
Rear spar webs 0.0205 0.0249 0.0249 0.0234 0.019 0.019 0.0161 0.0146
Rear spar caps 0.0278 0.0322 0.0336 0.0263 0.0234 0.0234 0.0205 0.0176
Rib webs 0.0205 0.0205 0.0249 0.0249 0.0205 0.0176 0.0146 0.0146
Rib caps 0.0102 0.0102 0.0117 0.0102 0.0102 0.0073 0.0088 0.0059
Stringers 0.019 0.0205 0.0205 0.019 0.0176 0.0176 0.0176 0.0176

Fig. 19  Optimal design—thickness distribution per component

are the thickness values obtained by the rib webs and caps, concerned, relatively high thickness values throughout the
which are mainly driven by the Brazier loading. The for- span of the wing are deemed necessary in order to sustain
mer obtain the majority of the aforementioned loading and the critical buckling load.
therefore an increase in their thickness is required, while The percentages of the 0◦, 90◦ and 45◦ ∕ − 45◦ plies of
the latter appear to obtain low thickness values, transferring each laminate and zone are also presented in Fig. 20. The
the developed loads to the webs. As far as the stringers is majority of the components, and specifically the front and

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 19 of 23 150

Fig. 20  Optimal design—ply percentages distribution per component

rear spar webs are dominated by 45◦ ∕ − 45◦ plies mainly due aforementioned studies differs to the one presented herein in
to their initial percentage in the baseline lay-up. 0◦ plies are the sense that combined aerostructural tools and measures such
present in the skins and caps to accommodate for the pres- as fuel burn were considered for the optimization cases. The
ence of normal stresses in these components, while on the geometry of the wing was also parametrized in terms of twist,
other hand 90◦ plies, enhance the strength of the components which produced an optimized lift loading distribution, decreas-
at the normal to the fiber direction. ing the bending moment by decreasing the lift produced in
Regarding the flutter analysis, the V−g and V−f plots are the outer sections of the wing. The results presented in the
presented in the following Fig. 21. Clearly, no indication optimization study conducted of Gray (2021) in terms of struc-
of instabilities at each of the first ten structural modes is tural mass can be compared more accurately to the ones of the
observed for the optimal design and the velocities range of present study, since no aerostructural performance optimiza-
the flutter analysis. tion was conducted and the nonlinear response of the structure
The comparison of the obtained optimal mass of the wing was also accounted for. The resulting 2.5 g mass was around
with similar results present in the literature is also discussed. 14.000 kg, with the major difference lying in the mass of the
In particular, upon removal of the external as well as the lead- stringers. In our case, buckling was deemed critical in the cur-
ing and trailing edge skin masses, which are not included rent configuration. As a result, and in order to support the
in the literature, the optimized mass of the wing is 29.247 developed buckling loads, the thickness of the stringers was
kg, while a mass of 30.032 kg is obtained in Brooks et al. directed towards greater thickness values, as also demonstrated
(2018). In Brooks et al. (2020), the mass of the optimized in Fig. 19. This phenomenon along with the predefined width
composite materials wing ranges between 20.000 and 40.000 and spacing of the stiffeners led to their elevated total mass.
kg, indicating good agreement between the optimized mass Another important issue is the fact that the approach towards
values. A similar study was conducted in Brooks et al. (2019), buckling analysis is different in the two studies, with the study
with the optimized mass being 19.796 kg. The results of the in Gray (2021) following a smeared modeling approach along

13
150 Page 20 of 23 S. Kilimtzidis, V. Kostopoulos

Fig. 21  Optimal design—V−g

and V−f plots

with a global buckling analysis. In our case, nevertheless, the uCRM−13.5 wing, based on the MIDACO algorithm has
stiffeners have been modeled explicitly and buckling was also been presented herein. The wing structure was represented
examined through a more detailed, global–local modeling using a 3D shell and beam elements FEM model, while
approach, which could have eventually shed light into fail- aerodynamics were treated via RANS, 3D panel as well
ure mechanisms that cannot be predicted by a combination of as DLM solutions, depending on level of fidelity. The lat-
a smeared modeling and global buckling analysis approach. ter provided also insight on the static aeroelastic behavior
Nevertheless, different spacing and width values should be of the reference wing. Comparison between the different
explored in future frameworks in order to assess their effect fidelities was initially performed with the aim of highlight-
at a greater extent. ing their effect on the structural response of the test case
wing. For the aerodynamics, significant changes to the
AoA to attain the target critical aerodynamic loading sce-
6 Conclusion nario compared to the high-fidelity CFD were observed.
Furthermore, the developed 3D panel method resulted into
A sequential optimization framework combining variable higher vertical deflection and torsional angles in compari-
fidelity computational tools for future high-aspect ratio son with the high-fidelity CFD, while the corrected DLM
composite aircraft wings, and specifically the undeflected, provided a closer approximation. From a structural point of

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 21 of 23 150

view, the inclusion of follower forces and geometric non-

linearities greatly altered the displacements field, induc-
ing lower vertical deflections yet higher tip shortening.
The aforementioned models were combined and cast in a
sequential optimization framework aiming at obtaining a
minimum mass configuration subject to strength, stiffness
in terms of panel buckling as well as flutter constraints.
The framework was initially guided by low-fidelity aero- Funding Open access funding provided by HEAL-Link Greece.
dynamics into promising basins of the design space, with
high-fidelity CFD runs providing refinement of the optimal Declarations
solution. The resulting wing structure was critical to panel Conflict of interest The authors declare that they have no conflict of
buckling, with strength criteria being more relaxed. Zones interest.
near the yehudi break, and specifically skins and spar caps,
Replication of results The numerical models as well as the optimi-
obtain higher thickness values in order to provide the suf-
zation framework referred herein are available from the author upon
ficient stiffness and strength for the structure. request.
In terms of future work, the present framework could
be benefited by the addition of aerodynamic shape opti- Open Access This article is licensed under a Creative Commons Attri-
bution 4.0 International License, which permits use, sharing, adapta-
mization procedures, which due to the lack of computa-
tion, distribution and reproduction in any medium or format, as long
tional tools and limited pre-processing capabilities were as you give appropriate credit to the original author(s) and the source,
not accounted for in this study. Geometry parametrization provide a link to the Creative Commons licence, and indicate if changes
techniques, such as the FFD are capable of altering geo- were made. The images or other third party material in this article are
included in the article's Creative Commons licence, unless indicated
metric variables of the wing, such as the twist, sweep, span
otherwise in a credit line to the material. If material is not included in
etc. during the optimization stages, thus enabling the opti- the article's Creative Commons licence and your intended use is not
mization of combined measures of aerostructural perfor- permitted by statutory regulation or exceeds the permitted use, you will
mance. When cast into a sequential optimization approach, need to obtain permission directly from the copyright holder. To view a
copy of this licence, visit http://​creat​iveco​mmons.​org/​licen​ses/​by/4.​0/.
additional effort needs to be directed towards analyzing the
effect of low-fidelity tools on drag predictions, since they
can significantly affect the overall performance. Turning
our attention towards composite materials, the inclusion
of a library of materials and lay-ups in order to support References
decision making in manufacturing and cost analysis at a
ACARE (2011) ACARE flightpath 2050—Europe’s vision for aviation.
greater extent as well as a wider variety of global–local Technical report. Advisory Council for Aviation Research and
analysis techniques (spars buckling, presence of holes and Innovation in Europe
rivets, etc.) could further expand the capabilities of the Airbus (2018) Global market forecast-cities, airports and aircraft 2019–
2038. Technical report. Airbus S.A.S.
present framework. In order to demonstrate the robust-
Albano E, Rodden WP (1969) A doublet-lattice method for calculating
ness of the optimization algorithm, several starting points lift distributions on oscillating surfaces in subsonic flows. AIAA
could be provided in subsequent studies, comparing the J 7(2):279–285. https://​doi.​org/​10.​2514/3.​5086
optimal solutions obtained. The algorithmic parameters, Barcelos M, Maute K (2008) Aeroelastic design optimization for
laminar and turbulent flows. Comput Methods Appl Mech Eng
as presented in Table 10, should also be further evaluated
197(19–20):1813–1832. https://​doi.​org/​10.​1016/j.​cma.​2007.​03.​
in order to shed more light on their effect in the sequen- 009
tial optimization framework. The ratio of the low to high- Barcelos M, Bavestrello H, Maute K (2006) A Schur–Newton–Krylov
fidelity runs as well as the effect and potential benefit from solver for steady-state aeroelastic analysis and design sensitivity
analysis. Comput Methods Appl Mech Eng 195(17–18):2050–
early on high-fidelity CFD samples in the optimization
2069. https://​doi.​org/​10.​1016/j.​cma.​2004.​09.​013
framework should be also investigated. Bartels RE, Scott RC, Allen T et al (2015) Aeroelastic analysis of
SUGAR truss-braced wing wind-tunnel model using FUN3d and
Acknowledgements This research was financially supported by the a nonlinear structural model. In: 56th AIAA/ASCE/AHS/ASC
State Scholarships Foundation of Greece (IKY), which granted a structures, structural dynamics, and materials conference, 2015.
scholarship to Spyridon Kilimtzidis. The scholarship is co-financed by American Institute of Aeronautics and Astronautics. https://​doi.​
Greece and the European Union (European Social Fund, ESF) through org/​10.​2514/6.​2015-​1174
the Operational Programme << Human Resources Development, Blair M (1992) A compilation of the mathematics leading to the dou-
Education and Lifelong Learning >> in the context of the Project blet lattice method. Technical report. Airforce Wright Laboratory
“Strengthening Human Resources Research Potential via Doctorate Boeing (2018) Boeing market outlook; Boeing commercial airplanes.
Research” (MIS-5000432), implemented by the State Scholarships Technical report. The Boeing Company
Foundation (IKY). The support is greatly appreciated by the author. Bordogna MT, Lancelot P, Bettebghor D et al (2020) Static and
dynamic aeroelastic tailoring with composite blending and

13
 150 Page 22 of 23 S. Kilimtzidis, V. Kostopoulos

manoeuvre load alleviation. Struct Multidisc Optim 61(5):2193– Haftka RT (1977) Optimization of flexible wing structures subject
2216. https://​doi.​org/​10.​1007/​s00158-​019-​02446-w to strength and induced drag constraints. AIAA J 15(8):1101–
Brazier LG (1927) On the flexure of thin cylindrical shells and other 1106. https://​doi.​org/​10.​2514/3.​7400
“thin’’ sections. Proc R Soc A 116(773):104–114. https://​doi.​org/​ Howcroft C, Calderon D, Lambert L et al (2016) Aeroelastic mod-
10.​1098/​rspa.​1927.​0125 elling of highly flexible wings. In: 15th Dynamics specialists
Brooks TR, Kenway GKW, Martins JRRA (2018) Benchmark aero- conference, 2016. American Institute of Aeronautics and Astro-
structural models for the study of transonic aircraft wings. AIAA nautics. https://​doi.​org/​10.​2514/6.​2016-​1798
J 56(7):2840–2855. https://​doi.​org/​10.​2514/1.​j0566​03 IJsselmuiden ST, Abdalla MM, Gurdal Z (2010) Optimization of
Brooks TR, Martins JR, Kennedy GJ (2019) High-fidelity aerostruc- variable-stiffness panels for maximum buckling load using lami-
tural optimization of tow-steered composite wings. J Fluids Struct nation parameters. AIAA J 48(1):134–143. https://​doi.​org/​10.​
88:122–147. https://​doi.​org/​10.​1016/j.​jflui​dstru​cts.​2019.​04.​005 2514/1.​42490
Brooks TR, Martins JRRA, Kennedy GJ (2020) Aerostructural trade- Jones RM (2018) Mechanics of composite materials. CRC Press,
offs for tow-steered composite wings. J Aircr 57(5):787–799. Boca Raton. https://​doi.​org/​10.​1201/​97814​98711​067
https://​doi.​org/​10.​2514/1.​c0356​99 Kassapoglou C (2013) Design and analysis of composite structures.
Calderon D, Cooper JE, Lowenberg MH (2018) On the effect of Wiley, Hoboken. https://​doi.​org/​10.​1002/​97811​18536​933
including geometric nonlinearity in the sizing of a wing. In: Katz J, Plotkin A (2001) Low-speed aerodynamics. Cambridge Uni-
2018 AIAA/ASCE/AHS/ASC structures, structural dynamics, versity Press, Cambridge. https://​doi.​org/​10.​1017/​cbo97​80511​
and materials conference, 2018. American Institute of Aero- 810329
nautics and Astronautics. https://​doi.​org/​10.​2514/6.​2018-​1680 Keimer R, Giannaros E, Kilimtzidis S et al (2022) Instrumentation
Calderon DE, Cooper JE, Lowenberg M, Neild SA, Coetzee EB and ground testing of a composite elastic wing wind tunnel
(2019) Sizing high-aspect-ratio wings with a geometrically model. IOP Conf Ser Mater Sci Eng 1226(1):012051. https://​
nonlinear beam model. J Aircr 56(4):1455–1470. https://​doi.​ doi.​org/​10.​1088/​1757-​899x/​1226/1/​012051
org/​10.​2514/1.​c0352​96 Kenway GKW, Martins JRRA (2014) Multipoint high-fidelity aero-
Castellani M, Cooper JE, Lemmens Y (2017) Nonlinear static aer- structural optimization of a transport aircraft configuration. J
oelasticity of high-aspect-ratio-wing aircraft by finite element Aircr 51(1):144–160. https://​doi.​org/​10.​2514/1.​c0321​50
and multibody methods. J Aircr 54(2):548–560. https://​doi.​org/​ Kenway G, Kennedy G, Martins JRRA (2010) A CAD-free approach
10.​2514/1.​c0338​25 to high-fidelity aerostructural optimization. In: 13th AIAA/
Coggin JM, Kapania RK, Schetz JA, Vijayaumari H, Zhao W. (2014) ISSMO multidisciplinary analysis optimization conference,
Nonlinear aeroelastic analysis of a truss based wing aircraft. 2010. American Institute of Aeronautics and Astronautics.
In: 55th AIAA/ASME/ASCE/AHS/ASC structures, structural https://​doi.​org/​10.​2514/6.​2010-​9231
dynamics, and materials conference, 2014. American Institute Kilimtzidis S (2022) Multi-fidelity multidisciplinary design and opti-
of Aeronautics and Astronautics. https://​d oi.​o rg/​1 0.​2 514/6.​ mization of composite materials aircraft wings. PhD Thesis,
2014-​0335 University of Patras
Crovato A, Almeida HS, Vio G et al (2020) Effect of levels of fidel- Kilimtzidis S, Mazarakos DE, Kostopoulos V (2018) Aeroelastic-
ity on steady aerodynamic and static aeroelastic computations. flutter analysis of AGARD wing from composite materials. Int
Aerospace. https://​doi.​org/​10.​3390/​aeros​pace7​040042 J Recent Adv Eng Res 4:16
Demirer HG (2021) Static and dynamic aeroelastic analysis of a very Kilimtzidis S, Kotzakolios A, Kostopoulos V (2023) Efficient struc-
light aircraft. Master’s Thesis, Middle East Technical University tural optimisation of composite materials aircraft wings. Com-
Dessena G, Ignatyev DI, Whidborne JF et al (2022) Ground vibra- pos Struct 303(116):268. https://​doi.​org/​10.​1016/j.​comps​truct.​
tion testing of a flexible wing: a benchmark and case study. 2022.​116268
Aerospace 9(8):438. https://​doi.​org/​10.​3390/​aeros​pace9​080438 Kontogiannis SG, Savill MA (2020) A generalized methodology for
Dillinger JKS, Klimmek T, Abdalla MM et al (2013) Stiffness opti- multidisciplinary design optimization using surrogate modelling
mization of composite wings with aeroelastic constraints. J and multifidelity analysis. Optim Eng 21(3):723–759. https://​doi.​
Aircr 50(4):1159–1168. https://​doi.​org/​10.​2514/1.​c0320​84 org/​10.​1007/​s11081-​020-​09504-z
Economon TD, Palacios F, Copeland SR et al (2016) SU2: an open- Kreisselmeier G, Steinhauser R (1979) Systematic control design by
source suite for multiphysics simulation and design. AIAA J optimizing a vector performance index. IFAC Proc Vol 12(7):113–
54(3):828–846. https://​doi.​org/​10.​2514/1.​j0538​13 117. https://​doi.​org/​10.​1016/​s1474-​6670(17)​65584-8
Fukunaga H, Sekine H, Sato M (1994) Optimal design of symmet- Lambe AB, Martins JRRA (2015) Matrix-free aerostructural optimi-
ric laminated plates for fundamental frequency. J Sound Vib zation of aircraft wings. Struct Multidisc Optim 53(3):589–603.
171(2):219–229. https://​doi.​org/​10.​1006/​jsvi.​1994.​1115 https://​doi.​org/​10.​1007/​s00158-​015-​1349-2
Garcia JA (2005) Numerical investigation of nonlinear aeroelastic Lambe AB, Kennedy GJ, Martins JRRA (2016) An evaluation of con-
effects on flexible high-aspect-ratio wings. J Aircr 42(4):1025– straint aggregation strategies for wing box mass minimization.
1036. https://​doi.​org/​10.​2514/1.​6544 Struct Multidisc Optim 55(1):257–277. https://​doi.​org/​10.​1007/​
Gray AC (2021) Geometrically nonlinear high fidelity aerostructural s00158-​016-​1495-1
optimisation for highly flexible wings. Master’s Thesis, Delft Lee SH (1992) MSC/NASTRAN nonlinear analysis handbook version
University of Technology 67. MacNeal-Schwendler Corporation, Newport Beach
Grossman B, Gurdal Z, Strauch GJ et al (1988) Integrated aerody- Liu B, Haftka R, Trompette P (2004) Maximization of buckling
namic/structural design of a sailplane wing. J Aircr 25(9):855– loads of composite panels using flexural lamination parameters.
860. https://​doi.​org/​10.​2514/3.​45670 Struct Multidisc Optim 26(1–2):28–36. https://​doi.​org/​10.​1007/​
Grossman B, Haftka RT, Kao PJ et al (1990) Integrated aero- s00158-​003-​0314-7
dynamic–structural design of a transport wing. J Aircr Love M, Bohlman J (1989) Aeroelastic tailoring studies in fighter air-
27(12):1050–1056. https://​doi.​org/​10.​2514/3.​45980 craft design. Technical report. NASA Langley Research Center
Haftka RT (1973) Automated procedure for design of wing structures Lupp CA, Cesnik CE (2019) A gradient-based flutter constraint includ-
to satisfy strength and flutter requirements. Technical report. ing geometrically nonlinear deformations. In: AIAA Scitech 2019
NASA Langley Research Center Forum, 2019. American Institute of Aeronautics and Astronautics.
https://​doi.​org/​10.​2514/6.​2019-​1212

13
Multidisciplinary structural optimization of novel high‑aspect ratio composite aircraft… Page 23 of 23 150

Lyu Z, Kenway GKW, Martins JRRA (2015) Aerodynamic shape Forum, 2021. American Institute of Aeronautics and Astronautics.
optimization investigations of the common research model wing https://​doi.​org/​10.​2514/6.​2021-​1713
benchmark. AIAA J 53(4):968–985. https://​doi.​org/​10.​2514/1.​ Schlueter M (2014) MIDACO software performance on interplanetary
j0533​18 trajectory benchmarks. Adv Space Res 54(4):744–754. https://d​ oi.​
Macquart T, Maes V, Bordogna MT et al (2018) Optimisation of com- org/​10.​1016/j.​asr.​2014.​05.​002
posite structures—enforcing the feasibility of lamination param- Schlüter M, Gerdts M (2009) The Oracle penalty method. J Glob Optim
eter constraints with computationally-efficient maps. Compos 47(2):293–325. https://​doi.​org/​10.​1007/​s10898-​009-​9477-0
Struct 192:605–615. https://​doi.​org/​10.​1016/j.​comps​truct.​2018.​ Schlüter M, Egea JA, Banga JR (2009) Extended ant colony optimi-
03.​049 zation for non-convex mixed integer nonlinear programming.
Marlett K (2011a) Hexcel 8552 IM7 unidirectional Prepreg 190 gsm Comput Oper Res 36(7):2217–2229. https://​doi.​org/​10.​1016/j.​
35% RC qualification statistical analysis report, Report NCP- cor.​2008.​08.​015
RP-2009-028 Rev B. Technical report. National Institute for Schlüter M, Erb SO, Gerdts M, Kemble S et al (2013) MIDACO on
Aviation Research MINLP space applications. Adv Space Res 51(7):1116–1131.
Marlett K (2011b) Hexcel 8552S AS4 plain weave fabric 193 gsm https://​doi.​org/​10.​1016/j.​asr.​2012.​11.​006
38% RC qualification material property data report, Report. CAM- Schmitt V, Charpin F (1979) Pressure distributions on the ONERA-
RP-2010-006 N/C. Technical report. National Institute for Avia- M6-wing at transonic Mach numbers. Technical report. Advisory
tion Research Group for Aerospace Research and Development, North Atlantic
Martins JRRA, Ning A (2021) Engineering design optimization. Cam- Treaty Organization
bridge University Press, Cambridge. https://​doi.​org/​10.​1017/​ Smith M, Patil M, Hodges D (2001) CFD-based analysis of nonlinear
97811​08980​647 aeroelastic behavior of high-aspect ratio wings. In: 19th AIAA
Martins JRRA, Alonso JJ, Reuther JJ (2004) High-fidelity aerostruc- applied aerodynamics conference, 2001. https://d​ oi.o​ rg/1​ 0.2​ 514/6.​
tural design optimization of a supersonic business jet. J Aircr 2001-​1582
41(3):523–530. https://​doi.​org/​10.​2514/1.​11478 Spalart P, Allmaras S (1992) A one-equation turbulence model for aer-
Maute K, Nikbay M, Farhat C (2001) Coupled analytical sensitivity odynamic flows. In: 30th Aerospace sciences meeting and exhibit,
analysis and optimization of three-dimensional nonlinear aeroe- 1992. American Institute of Aeronautics and Astronautics. https://​
lastic systems. AIAA J 39(11):2051–2061. https://​doi.​org/​10.​ doi.​org/​10.​2514/6.​1992-​439
2514/2.​1227 Thuwis GAA, Breuker RD, Abdalla MM et al (2009) Aeroelastic
Mavris DN, DeLaurentis DA (2000) Methodology for examining the tailoring using lamination parameters. Struct Multidisc Optim
simultaneous impact of requirements, vehicle characteristics, and 41(4):637–646. https://​doi.​org/​10.​1007/​s00158-​009-​0437-6
technologies on military aircraft design. In: 22nd ICAS congress, Triplett WE (1980) Aeroelastic tailoring studies in fighter aircraft
2000 design. J Aircr 17(7):508–513. https://​doi.​org/​10.​2514/3.​57932
MIDACO (2018) MIDACO user manual Tsai SW, Hahn HT (2018) Introduction to composite materials. Rout-
Miki M, Sugiyama Y (1993) Optimum design of laminated compos- ledge, New York. https://​doi.​org/​10.​1201/​97802​03750​148
ite plates using lamination parameters. AIAA J 31(5):921–922. Vassberg J, Dehaan M, Rivers M et al (2008) Development of a com-
https://​doi.​org/​10.​2514/3.​49033 mon research model for applied CFD validation studies. In: 26th
NASTRAN M (2021) MSC/NASTRAN 2021.4 linear static analysis AIAA applied aerodynamics conference, 2008. American Insti-
user’s guide tute of Aeronautics and Astronautics. https://​doi.​org/​10.​2514/6.​
Patil MJ, Hodges DH, Cesnik CES (2001) Nonlinear aeroelasticity and 2008-​6919
flight dynamics of high-altitude long-endurance aircraft. J Aircr Verri AA, Bussamra FLS, de Morais KC et al (2020) Static loads evalu-
38(1):88–94. https://​doi.​org/​10.​2514/2.​2738 ation in a flexible aircraft using high-fidelity fluid–structure itera-
Poon NMK, Martins JRRA (2006) An adaptive approach to constraint tion tool (e2-FSI): extended version. J Braz Soc Mech Sci Eng.
aggregation using adjoint sensitivity analysis. Struct Multidisc https://​doi.​org/​10.​1007/​s40430-​019-​2154-4
Optim 34(1):61–73. https://​doi.​org/​10.​1007/​s00158-​006-​0061-7 White FM (2010) Fluid mechanics, 7th edn. McGraw Hill Higher Edu-
Reuther J, Alonso J, Martins JRRA et al (1999) A coupled aero-struc- cation, Maidenhead
tural optimization method for complete aircraft configurations.
In: 37th Aerospace sciences meeting and exhibit, 1999. Ameri- Publisher's Note Springer Nature remains neutral with regard to
can Institute of Aeronautics and Astronautics. https://​doi.​org/​10.​ jurisdictional claims in published maps and institutional affiliations.
2514/6.​1999-​187
Ritter M, Hilger J, Zimmer M (2021) Static and dynamic simulations
of the Pazy wing aeroelastic benchmark by nonlinear potential
aerodynamics and detailed FE model. In: AIAA Scitech 2021

13

View publication stats