JOURNAL OF AIRCRAFT

Effect of Flutter on the Multidisciplinary Design Optimization

of Truss-Braced-Wing Aircraft

Wrik Mallik,∗ Rakesh K. Kapania,† and Joseph A. Schetz‡

Virginia Polytechnic Institute and State University, Blacksburg, Virginia 24061
DOI: 10.2514/1.C033096
This study highlights the effects of a flutter constraint on the multidisciplinary design optimization (MDO) of a
truss-braced-wing transport aircraft for both medium-range and long-range missions. Previous MDO studies for
both of these missions were performed without considering the effect of flutter. Hence, the flutter constraint has now
been added to the other design constraints in this MDO study. Minimizing the takeoff gross weight and the fuel burn
are selected as the objective functions. The results show that, for the medium-range mission, the flutter constraint
applied at 1.15 times the dive speed imposes a 1.5% penalty on the takeoff weight and a 5% penalty on the fuel
consumption while minimizing these two objective functions. The penalties imposed on the minimum-takeoff-gross-
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096

weight and minimum-fuel-burn designs for the long-range mission due to the similar constraint are 3.5 and 7.5%,
respectively. Importantly, the resulting truss-braced-wing designs are still superior to equivalent cantilever designs
for both of the missions, as they have both lower takeoff gross weight and fuel burn. However, a relaxed flutter
constraint applied at 1.05 times the dive speed can restrict the penalty on the takeoff gross weight to only 0.3%, and
that on the fuel burn to 2% for minimizing both the objectives, respectively, for the medium-range mission. For the
long-range mission, a similar relaxed constraint can reduce the penalty on fuel burn to 2.9% when that objective
function is minimized. These observations suggest the need for a cost-benefit study to determine whether active-
flutter-suppression mechanisms with their added weight and complexities can be used for the truss-braced-wing
aircraft to further reduce either the takeoff gross weight or the fuel burn.

Nomenclature θf = time-independent part of the pitching degree of

AC = part of the aerodynamic forces analogous to damping freedom
AK = part of the aerodynamic forces analogous to stiffness Λ = sweep angle
AM = part of the aerodynamic forces analogous to mass λF = eigenvalues of the aeroelastic system
Akref  = aerodynamic forces as a function of reference reduced Ξ = matrix of the elastic normal modes of the structure
frequency ξ = generalized degrees of freedom of the aeroelastic
a = distance between the reference point and the center of system
mass of an airfoil nondimensionalized by semichord ρ = density of air
bref = reference semichord ω = frequency of oscillation of the aeroelastic system
g = damping of the aeroelastic system · = first derivative with respect to time
h = plunging degree of freedom ·· = second derivative with respect to time
hf = time-independent part of the plunging degree
of freedom Subscripts
kref = reference reduced frequency
c = circulatory part of the aerodynamic forces
L = lift at the elastic axis of the airfoil
NC = noncirculatory part of the aerodynamic forces
M = pitching moment at the elastic axis of the airfoil
Mbound = Mach-number values representing the flutter
boundary I. Introduction
Mflut = Mach-number values representing the flutter velocity
M∞
QA
=
=
freestream Mach number
aerodynamic forces, bold represents a matrix T HE history of commercial aircraft shows that the cantilever wing
aircraft (the so-called tube and wing aircraft) configuration has
been most widely pursued by the aircraft industry. As a result, many
t = time
V = velocity of the air advanced technologies have been applied to the current cantilever
θ = pitching degree of freedom design over the past few decades to make it ever more efficient.
However, further large improvements are considered improbable
without a radical change in the whole aircraft configuration. In recent
years, the Virginia Polytechnic Institute and State University’s
multidisciplinary aircraft design group has focused on studying
Presented as Paper 2013-1454 at the 54th AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference, Boston, MA, 8–
Pfenninger’s [1] vision of a truss-braced-wing (TBW) configuration
11 April 2013; received 15 July 2014; revision received 14 December 2014; to understand its true potential, and the feasibility of incorporating
accepted for publication 9 February 2015; published online 20 April 2015. such unconventional configurations in future transport aircraft. A
Copyright © 2014 by the American Institute of Aeronautics and Astronautics, strut-braced-wing (SBW) configuration is a TBW without any
Inc. All rights reserved. Copies of this paper may be made for personal additional truss members(s) connecting the main wing to the strut.
or internal use, on condition that the copier pay the $10.00 per-copy fee The TBW concept has been studied extensively at Virginia
to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA Polytechnic Institute and State University first for a Mach 0.85
01923; include the code 1533-3868/15 and $10.00 in correspondence with the long-range transport aircraft having a flight mission similar to that
CCC. of a Boeing 777-200 Longer Range (LR), a range of 7730 n mile,
*Graduate Research Assistant, Department of Aerospace and Ocean
Engineering. Student Member AIAA. and carrying 305 passengers [2–5]. The subsequent studies focused
†
Mitchell Professor, Department of Aerospace and Ocean Engineering. on using the multidisciplinary design optimization (MDO) for
Lifetime Associate Fellow AIAA. investigating TBW configurations for a medium-range transport
‡ aircraft that has a flight mission similar to the Boeing 737-800 Next
Holder of Fred D. Durham Chair, Department of Aerospace and Ocean
Engineering. Lifetime Fellow AIAA. Generation (NG) with a cruise Mach of 0.78, for a 3115 n mile range
AIAA Early Edition / 1
 2 AIAA Early Edition
6 Bending moment of TBW and cantilever designs
7 x 10
6
Bending moment (lbs−ft)
5
4
3
2
1
0 number (T/O, takeoff; LDG, landing).
−1
0 10 20 30 40 50 60
Distance along span of wing (ft.)
Fig. 1 Bending-moment diagram for 737-800-like mission aircraft.
and 162 passengers [6]. The results obtained from both series of
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096
studies show that the TBW designs have a reduced takeoff gross
weight (TOGW) and fuel consumption compared to cantilever
configurations. The additional truss members help in reducing the
spanwise bending moment, as shown in Fig. 1, thereby reducing the
t∕c as well as the chord length. The lighter wing structure is
accompanied by an increased wing semispan increasing the aspect
ratio. Aerodynamically, low airfoil thickness reduces the form and
wave drag, whereas a higher aspect ratio decreases the induced drag.
A shorter chord length reduces the Reynolds number and the
unsweeping of the wing reduces the spanwise crossflow [7], both of
which, in combination with the thinner wings, allow more laminar
flow. These benefits, discussed in some previous studies [8,9],
combine together to reduce the drag and fuel consumption as well.
However, these studies were conducted without considering the
effect of a flutter constraint on the aircraft designs during
optimization. An analysis of flutter of the TBW configurations has
been undertaken previously at Virginia Polytechnic Institute and
State University, and a limited parametric study was conducted
toward understanding the aeroelastic behavior of the TBWs [10–12].
However, these parametric studies were not performed by including
the flutter as a constraint within the MDO framework, and were
focused only on the structural aspects of the design. Thus, while these
studies do provide a significant insight into the structural aspects of
TBW configurations’ aeroelastic behavior, they do not show the true
effect of flutter on the designs obtained from an MDO framework.
The present study focuses on adding a flutter constraint in the VT
MDO computational-design environment to study its impact on
the design of TBW aircraft. The flight missions are similar to those of
the previous studies, simulating a Boeing 737-800NG for the
medium-range mission, except that the cruise Mach for the medium-
range mission here is 0.7, and a Boeing 777-200LR for the long-range
mission. The schematics of the long-range and medium-range
missions are illustrated in Figs. 2 and 3, respectively. The develop-
ment of the flutter constraint for the optimization problem is
discussed in the Methodology section. The two objectives selected
for the MDO studies are minimizing the TOGW and fuel
consumption. The results shown here involve a comparison of the
TBW designs with wing-mounted engines for the medium-range
missions, and fuselage-mounted engines for the long-range missions,
Fig. 2 Boeing 777-200LR-like mission profile (T/O, takeoff; LDG,
landing).
/ MALLIK, KAPANIA, AND SCHETZ
TBW 1−jury
Cantilever
Fig. 3 Boeing 737-800NG-like mission profile with reduced cruise Mach
70 80 90
with and without the flutter constraint. The authors intentionally
chose wing-mounted TBW designs for medium-range missions, as a
careful inspection of results from a previous study [6] revealed that
wing-mounted TBW designs outperform their fuselage-mounted
versions in minimizing both TOGW and fuel weight. On the other
hand, for the long-range mission, the much larger span of TBW
configurations over the conventional cantilever designs demands
much larger counteryawing moment from the rudder in the case of an
engine-out condition. Because rudder design is beyond the scope of
this paper, wing-mounted TBW configurations have not been
considered. Finally, the sensitivity of the flutter of the TBWs to some
of the key design variables for the given design problems is discussed.
Designs close to the flutter boundary, but violating it by only a small
margin were also investigated to determine whether they merit a cost-
benefit study of using a relaxed flutter constraint with active-flutter-
suppression mechanisms, which may add some weight, but may also
result in much smaller penalties on fuel burn or TOGW due to the
flutter constraint.
II. Methodology
A. Development of an Extended Structural Model
The structural analysis in all the previous MDO studies [4,6,13]
was carried out using the double-plate idealization of the aircraft
wing box, as shown in Fig. 4. It was assumed that the wing box would
act as a beam, with the major contribution to the bending stiffness
provided by the upper and lower skins. The total load-bearing weight
of the wing is accurately predicted by this simple model [3], and that
was adequate for the initial MDO studies. However, assuming the
skins to provide all of the bending stiffness leads to much thicker
upper and lower skins than in an actual aircraft wing box where the
spar caps and the stiffeners provide most of the bending stiffness, and
the skins just form very thin panels for carrying shear. The thicker
skins result in an overestimation of the effective torsional stiffness of
the wing sections, thereby leading to an inaccurate aeroelastic
analysis. This problem has been solved in this study by an updated,
higher-fidelity wing-box model extension, as shown in Fig. 5. In the
extended model, the wing box is designed as a combination of spars,
skin panels, and stiffeners. The skin panels formed by the network of
spars, stiffeners, and ribs are designed against panel buckling stresses
due to shear loads assuming simply supported boundary conditions
for the panel. The shear stresses developed in the skin panels due to
external transverse forces and torques are compared against critical
shear buckling stresses of the skin panels. Thus, the skin panels are
sized iteratively by increasing their thickness, starting from a
minimum value of 0.1 in. until the stresses meet the buckling
constraint. The spar webs are designed against shear stresses
generated from shear flow due to transverse lift forces and those
generated due to torsional loads. The spar webs are sized by
Fig. 4 Previous wing-box structural model, as presented in [3].
 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ 3

Fig. 5 Extended wing-box structural model.

comparing the shear stresses against yielding due to shear. The old designs obtained from the MDO code in the current study. This is in
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096

model remains giving an estimation of the material weight and of the contrast to the 11 constraints used for the previous studies, the flutter
normal bending stresses and the buckling stresses for the TBWs that constraint being the new addition. The first 11 constraints listed in
need to be satisfied by the new model. The spar caps and the stiffeners Appendix A have been detailed in previous publications [2–4]. The
are sized for these bending and buckling stresses developed at the new flutter constraint checks if the critical flutter speed of the
various sections. For the present study, the extended wing-box model designed aircraft satisfies the flutter boundary. The development of
results in large spar-cap and stringer areas, and very thin skin and spar the flutter constraint will be explained as follows.
webs, which would produce very little contribution to the section’s The optimization algorithm used for this study is DARWIN, a
in-plane or out-of-plane second moment of area. It should be stated genetic algorithm (GA) available in ModelCenter. Such optimization
that the present model would underpredict bending stiffness if the algorithms do not use gradient-based methods; hence, they are useful
thickness of the skins and spar webs is significant as it neglects their for objective functions with discrete variables, and they can reach a
contribution, and future analysis will take care of the thicker skin global minimum [16]. To briefly explain the process, the optimizer
panels. However, validation studies [14] show that the extended starts with a user-specified number of random designs, referred to as
model evaluates similar bending stiffness compared to the old model, the initial population. A crossover between two parents takes place to
but measures torsional stiffness much more accurately. develop a pair of children members. The design variables may also
undergo mutation. Designs that do not undergo crossover or mutation
B. Virginia Polytechnic Institute and State University’s TBW MDO pass on unchanged to add to the subsequent generation’s population
Design Environment list. The crossover and mutation probabilities, as well as the
The MDO tool used for this TBW study has been built within the population size, are user-specified inputs to the optimization scheme.
integration and optimization software ModelCenter [15]. This has For this study, a multiple elitist-selection scheme was used, in which
been used for the VT TBW studies over the past few years [3–6], and both the parent and the children are ranked together on the basis of
it allows full integration of several in-house analysis codes and their fitness. A specified number of designs are chosen from the
software packages into a single environment. The tool has a number combined members according to their fitness value. The rest of the
of analysis nodes and optimization nodes, and these are connected via population is made up of the best of the remaining children. Further
links. The analysis codes were developed in FORTRAN and C++, details on how the fitness function is evaluated, as well as the multiple
and connected to the various analysis nodes via Python scripts called elitist and other selection schemes, can be obtained from the
Wrappers. As a result, the various analysis nodes are referred to as the DARWIN Help Manual [15]. DARWIN is used to minimize the final
various Wrappers. A detailed explanation of the ModelCenter TOGW or fuel consumption, whichever is chosen as the objective,
environment is available in the manual [15]. and it is believed to have converged to a global minimum if a specified
The preprocessing wrapper initializes the various analysis nodes number of generations are produced without any significant
with information, like the aircraft geometry, the lifting surface, improvement in the objective function. It must be stated that the
structural member definitions, and other specifications that are
subsequently required by the other analysis modules. The aircraft
configurations are calculated based on a set of 17 load cases that are Table 1 Design load cases
shown in Table 1. These same load cases are also used for the
prestressed flutter analysis once the final aircraft configuration is Load-case Flutter-analysis
obtained. The first four load cases are the principal load cases number Load factor Fuel, % Altitude, kft altitudes, kft
required for aircraft maneuvers. The fifth load case is used to 1 2.5g 100 — — 0, 10, 20, 30, 40
calculate maximum deflection during landing, whereas the final 12 2 2.5g 20 — — 0, 10, 20, 30, 40
load cases are gust load cases, specified by the Federal Aviation 3 −1.0g 100 — — 0, 10, 20, 30, 40
Regulation (FAR) structural loading requirements. These load cases 4 −1.0g 20 — — 0, 10, 20, 30, 40
5 2.0g Taxi bump 100 0 — —
are assumed to be satisfactory at the aircraft conceptual design stage, 6 Gust (V app ) 100 0 0
as they are representatives of some of the main load cases among 7 Gust (V app ) 10 0 0
hundreds of load cases used in the final design and manufacture of a 8 Gust 100 0 0
transport aircraft. 9 Gust 10 0 0
The analyses used in the various modules have been extensively 10 Gust 100 10 10
discussed previously [2,3,6]. The design variables used in the 11 Gust 10 10 10
analyses include a combination of those used to parameterize the 12 Gust 100 20 20
geometry, like the sectional chords, thickness-to-chord ratios, and 13 Gust 10 20 20
14 Gust 100 30 30
nongeometric variables, like fuel weight, TOGW, maximum required
15 Gust 10 30 30
thrust and altitude. These design variables are used to form the 16 Gust 100 40 40
optimization problem that has been discussed in detail in previous 17 Gust 10 40 40
papers [2,3]. There are a total of 12 constraints applied to the aircraft
 4 AIAA Early Edition
method used to check the convergence of the GA cannot provide
unique solutions because the GA cannot guarantee convergence to a
single global optimal design. However, it has been proved that, for a
GA used with the multiple elitist scheme, there is always an upper
bound on the number of iterations that the GA will take to reach a
global optimum with a probability of less than 1 [17]. Thus, in other
words, the results coming from DARWIN can always reach the
vicinity of a global optimum. Therefore, although one cannot state
that the results coming out of DARWIN for the various designs will
always be unique and repeatable, the behavior shown by the flutter
constraint should still be similar in a different MDO simulation,
because it is believed that DARWIN reaches the vicinity of the global
optimum for each simulation.
C. Flutter Boundary and Flutter Constraint
1. Development of Flutter Boundary
Flutter is a dynamic aeroelastic instability that causes oscillations
of exponentially increasing amplitude when analyzed as a linearized
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096
eigenvalue problem. Therefore, it is imperative that, during the
design of aircraft, one knows both the flight envelope and the flutter
boundary, based on the flight mission. The flight envelope and design
dive speeds were evaluated by using a polynomial fit that represents
the flight data for medium-range and long-range transport aircraft,
keeping the cruise Mach to be 0.7 and 0.85, respectively, for the
current medium-range and long-range missions. For the medium-
range mission, the Mach number at sea level is 0.47, and increases to
0.7 at about 21,000 ft; thereafter, it stays constant. The dive speed is
selected to be 1.2 times the operational speed up to a maximum Mach
number of 0.82. The flutter boundary is selected as 1.15 times the
dive speed or dive Mach number, as per Federal Aviation
Administration regulations. For the long-range mission, the cruise
flight envelope starts at Mach number 0.5 at sea level, increases to
0.85 at about 30,000 ft, and remains constant thereafter. The dive
speed is evaluated at 1.1 times the cruise speed, and the flutter
boundary is evaluated at 1.15 times the dive speed. However, going
by the assumption that supersonic flight is not expected for these
aircraft, the flutter boundary is restricted to Mach 0.92 for these
conditions. A detailed transonic flutter analysis required to detect the
transonic dip phenomena expected in Mach numbers ranging from
0.9 to 1.1 is beyond the scope of this simplified flutter analysis. The
variations of the operational Mach number MO and the flutter-
boundary Mach number MF are shown in Fig. 6 for the medium-
range mission, and in Fig. 7 for the long-range mission.
2. Analysis of Flutter
The flutter analysis was conducted using the k method or the V-g
method. The unsteady aerodynamic analysis was performed using
Theodorsen’s method [18], and the compressibility effect was
incorporated using the Prandtl–Glauert correction factor for swept
wings [19]. Although Theodorsen’s method is only strictly applicable
for incompressible flow, compressibility corrections applied to
Flight envelope and Flutter boundary for medium−range mission
4
x 10
4
MO
3 MF
Altitude (ft.)
2 analyzed using the doublet-lattice method in a previous paper [11].
1 from a linear prestressed flutter analysis performed in NASTRAN
0
0.4 0.5 0.6 0.7 0.8 0.9
Mach no.
Fig. 6 Flight envelope and flutter boundary for the medium-range
mission.
/ MALLIK, KAPANIA, AND SCHETZ
Flight envelope and Flutter boundary for long−range mission
x 10 4
5
MO
4 MF
Altitude (ft.)
3
2
1
0
0.5 0.6 0.7 0.8 0.9 1
Mach no.
Fig. 7 Flight envelope and flutter boundary for the long-range mission.
incompressible, quasi-steady pressure distributions provide a good
steady-state condition about which a small perturbation can be
provided to the system to observe its stability. As a result, these
corrections are applied only to the circulatory portion of the
aerodynamic forces, and the transient noncirculatory forces are
measured from incompressible-flow theory. The other important
thing to consider is the aft movement of the center of pressure
observed in high subsonic compressible flow, which has been
observed from both computational and experimental results [20].
Such transonic effects caused by an acceleration of subsonic flow to
locally supersonic flow followed by the formation of shock waves
can only be rigorously incorporated in an aeroelastic study by
performing a computational fluid dynamics simulation. While that is
not feasible within the scope of the present MDO study, such effects
can be represented to some extent by allowing the location of the
aerodynamic center to shift from the quarter chord at Mach 0.4 to
40% of the chord at Mach 0.9. Such an approach has been pursued by
Beddoes [21] and Leishman [22] who developed modified indicial
functions based on Wagner’s functions [23] to perform transonic
aeroelastic analysis [24] along with other authors [25]. The details of
the strip-theory-based aerodynamics and the flutter calculations
using the V-g method or k method are provided in Appendix B.
The flutter analysis is relatively simple for the cantilever configu-
rations, as there are no nonplanar sections involved. However, for the
TBW configurations, the strut and the jury develop in-plane stresses
in the wing–truss system. These in-plane stresses developed from
the various load cases have to be accounted for in the modal analysis
to obtain the modes, which would actually take part in flutter. A
similar analysis for TBW has been explained in detail in another
article [26]. For brevity, we just explain the method followed here.
The combination of the aerodynamic forces, the fuel weight, the
engine weight (if it is wing mounted), and the self-weight of the
structure is modified by the respective load factor for a particular load
case. Then, a linear static analysis is performed, and the in-plane
stresses are evaluated to obtain the geometric stiffness matrix. The
modal analysis is performed about these prestressed conditions by
using an equivalent stiffness matrix obtained by combining the actual
stiffness and the geometric stiffness matrix. These prestressed mode
shapes and frequencies are then used for the flutter analysis.
3. Flutter Benchmarking
The flutter analysis performed for cantilever and SBW with the in-
house method using the strip theory was validated against those
Here, the authors intend to perform an evaluation of the current
prestressed analysis by comparing the results with those obtained
using the doublet-lattice unsteady aerodynamics. The TBW design
chosen was a long-range aircraft, optimized for minimum-fuel
1 weight without applying any constraint on the flutter, and the 2.5 g
maneuver load case with full fuel was used to perform the linear static
analysis. First, a grid-convergence study was performed for the
doublet-lattice method in NASTRAN, and the result was found to
 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ 5

V−g diagram with extended model

Mode 3
0.5 Mode 5

Damping g
315.38 KEAS
0

−0.5

0 100 200 300 400 500 600

Velocity (KEAS)
V−f diagram with new model
Mode 3
Frequency (Hz.)

4 Mode 5

0
0 50 100 150 200 250 300 350 400 450 500
Velocity (KEAS)
Fig. 8 V-g and V-f plots with the extended structural model (KEAS, equivalent airspeed in knots).
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096

Fig. 9 Modes 3 and 5 from linear prestressed modal analysis in NASTRAN.

converge satisfactorily using 1980 panels for the half-wing. The wing
and the strut were considered as the only lifting members. The V-g
and V-f plots obtained from the in-house code in Fig. 8 show the
flutter crossing and the main modes participating in the flutter. The
same modes were also found to participate in the flutter from
NASTRAN. The mode shapes obtained from NASTRAN are
illustrated in Fig. 9. It indicates that modes 3 and 5, the second
primarily out-of-plane bending mode and the first primarily torsion
mode, respectively, participate in the flutter. The results for flutter
speed and flutter frequency are compared in Table 2.
The table also shows the results for the flutter speed and frequency
using the prestressed modes from a nonlinear analysis. Other authors
have shown the importance of a nonlinear static analysis for static
[27] and aeroelastic analyses [28,29] of flexible nonplanar structure
similar to the TBW studied here. Thus, it was decided to perform an
off-line nonlinear aeroelastic analysis for each of the optimal designs.
Such a nonlinear static analysis is not available in the MDO
framework mainly because of its very high computational cost.
However, the designs obtained from the MDO are checked for
prestressed flutter analysis with an off-line nonlinear static analysis
performed via NASTRAN, as shown in Table 2. The difference in the
flutter velocities and frequencies of the two NASTRAN results
compared to the in-house ones is expressed as a percentage. For this
benchmarking case, the results obtained from the nonlinear
aeroelastic analysis show a small deviation from the linear results
with the flutter velocity only 4.5% lower than that obtained from the
static analysis obtained from doublet lattice. The modes that
participated in the flutter for both analyses are the same, and the
nonlinear prestressed mode shapes, shown in Fig. 10, look similar to
their linear counterparts, indicating that no significantly different
physical behavior is observed between the two flutter results.
However, the doublet-lattice linear prestressed results from
NASTRAN show a slightly higher flutter velocity than the in-house
code, which means the difference in the flutter speed between the
linear and nonlinear prestressed aeroelastic analysis performed using
NASTRAN is larger (7.5%). A nonlinear prestressed flutter analysis
was performed in another paper [30]. However, these results cannot
be compared directly to those results, because an iterative method
was used by Coggin et al. in [30], which cannot be incorporated into
an MDO framework because of its huge computational cost. Also, the
results were obtained for a wind-tunnel model with different load
cases and boundary conditions than what is simulated in the MDO.
The authors would also like to mention that, from the perspective
of analysis techniques, the strip theory used for the aerodynamic
analysis is a lower-fidelity method for the TBWs than similar
analyses performed in the past [10] using panel methods. Although
near-vertical members, like the jury and the strut–wing vertical
intersections, are not considered as lifting surfaces, the strip theory
cannot be considered as accurate as doublet lattice for nonplanar
members, like the strut, as the three-dimensional and finite span
effects for the swept sections become more important. However,
unlike the past studies done outside the MDO environment, the
current goal is always to select simple methods, like the strip theory,
which can quickly perform a relatively accurate analysis over more
complicated and computationally expensive methods so that one can
rapidly compare several configurations. This idea would have been

Table 2 Prestressed flutter-analysis benchmarking

In-house k method, strip theory, NASTRAN PKNL, doublet lattice, NASTRAN PKNL, doublet lattice,
linear prestressed linear prestressed nonlinear prestressed
Flutter speed, KEAS 315.38 327.7 (3.5%) 301.34 (−4.5%)
Flutter frequency, Hz 2.55 2.48 (−2.7%) 2.34 (−8.2%)
Participating modes 3 and 5 3 and 5 3 and 5
 6 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ

Fig. 10 Modes 3 and 5 from linear prestressed modal analysis in NASTRAN.

proven ill-conceived if the strip theory had provided significantly
different results from the doublet-lattice theory. But, the flutter
benchmarking performed here indicates that the results shown later
using the strip theory can be trusted at the conceptual design stage.
4. Development of the Flutter Constraint
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096
Looking at Fig. 11, we observe that, moving from DESIGNS1 to
DESIGNS2, a gradual increase in flutter margin occurs until a point is
reached where the constraint is satisfied. However, this is associated
with a gradual increase in the takeoff weight, indicating that the
flutter constraint is active. To obtain a better understanding of this
effect, two designs, POINT 1 and POINT 2, have been chosen, which

At the various altitudes mentioned in Table 1, a flutter check is
performed by the iterative Mach-number matching procedure, which
starts with a test Mach number and, after each iteration, the final
flutter Mach number is observed. The test Mach number is replaced
by the flutter Mach number after each iteration until the solution
converges after some finite number of iterations. The operating Mach
number at cruise is chosen as the initial test Mach number, and the
onset of flutter is checked iteratively up to the flutter boundary, shown
in Figs. 6 and 7. If, for a particular load case at a particular altitude, no
flutter is observed within the search Mach-number envelope, one can
conclude that the aircraft design will not suffer from flutter.
The flutter constraint is evaluated as follows:
Mflut
Flutter constraint  −1
Mbound
in which Mflut means the flutter Mach number, and Mbound means the
Mach number of the flutter boundary. A negative value of the flutter
constraint means that the flutter speed lies within the flight envelope,
and the constraint has been violated. There are several different flutter
represent the best designs from those obtained without the flutter
constraint (DESIGNS1) and those obtained after satisfying the
constraint (DESIGNS2), respectively. These configurations are
shown in Fig. 12 and compared against each other in Table 3. We
observe that, going from POINT 1 to POINT 2, there is a 1.5%
increase in the TOGW associated with satisfying the flutter
constraint. This increase is mainly caused by a 12% increase in the
bending weight of the wing system. The bending weight is partly
increased by an increase in the span of the main wing and the jury, and
partly increased by the stiffening of the wing at the junction of the
wing and the strut. The increase in the span is not a usual strategy to
suppress flutter, but mainly occurs because the GA traverses the
design space selecting discrete design points. Thus, although it
comes up with a design that has a slightly longer span, the major
(1)
5 TOGW vs Flutter margin
x 10
1.44
Designs1
Designs2
1.43
1.42
TOGW (lbs.)

checks in total, conducted at selected altitudes for the various load

cases shown in Table 1. The critical flutter margin is the one where the 1.41 POINT 2
margin between the flutter speed and the flutter boundary is a
minimum. This may or may not be the lowest flutter speed, because at 1.4
1.5%
penalty
lower altitudes, the operational Mach number is much lower than it is on
TOGW
at higher altitudes. 1.39 1.15 VD
POINT 1
1.38
−20 −15 −10 −5 0 5 10
III. Results and Discussion Flutter margin (%)
The objectives of this study were to include the flutter constraint Fig. 11 Designs obtained by minimizing TOGW for the medium-range
mission.
inside the VT MDO tool, and to observe its effects on the final
optimized TOGWor the fuel consumption of the TBW designs for the
medium-range and long-range missions.
4 Fuel consumption vs Flutter margin
A. Medium-Range-Mission Designs 2.7 x 10
Designs1
1. Optimization Results Designs2
2.65
First, designs were obtained without applying the flutter constraint
to serve as the baseline for comparisons. The GA, DARWIN, was 2.6
Fuel weight (lbs.)

allowed to run for a sufficient amount of time until no significant

improvements were obtained. Only those designs that did not violate 2.55
any of the other 11 constraints at the end of the optimization 2.5
POINT 2
procedure were saved. These designs have been termed as 5% penalty
DESIGNS1, represented by dark dots in Figs. 11 and 13 for the 2.45 on fuel
burn
minimum-TOGW and minimum-fuel-consumption objectives.
Then, in the subsequent runs, the saved designs were chosen as 2.4 1.15 VD

initial points, and the flutter constraint was included within the POINT 1
2.35
optimization procedure. These designs have been termed as −25 −20 −15 −10 −5 0 5
DESIGNS2, represented by squares in Figs. 11 and 13. The Flutter margin (%)
procedure undertaken will show the true effect of the flutter constraint Fig. 12 Minimum-TOGW optimized designs for medium-range
on the designs. mission: POINT 1 (left) and POINT 2 (right).
 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ 7

Table 3 Minimum-TOGW optimized TBW designs consumption can be attributed to the reduced aerodynamic efficiency
for medium-range mission of POINT 2 compared to POINT 1. Lowered aerodynamic efficiency
POINT 1 POINT 2 results from the reduced span and aspect ratio of POINT 2 compared
Minimum-TOGW TBW (no flutter (with 1.15V D to POINT 1. These changes are brought about by the MDO to satisfy
one jury constraint) flutter constraint) the flutter constraint. Although there is no significant addition of
TOGW, lb 138,400 140,600 (1.5%) bending material, the decrease of the span results in the stiffening of
Fuel weight, lb 26,600 26,500 the structure, thus leading to higher frequencies. There is also a
Structural bending weight, lb 24,500 27,500 redistribution of mass and stiffness along the outboard main wing as
Design L∕D 31.9 32.2 structural material is moved from the strut–wing junction toward the
Wing/strut semispan, ft 75.8∕46.7 78.8∕50.9 tip when going from POINT 1 to POINT 2, but without any structural
Wing one-fourth chord sweep, deg 9.80 8.31 mass added. A thinner jury and a thicker strut are also observed for the
Root chord, ft 13.0 9.6
POINT 2 design as compared to the POINT 1 design. The combined
Strut–wing junction chord, ft 8.8 12.0
Tip chord, ft 4.0 5.8 effect of these changes is a 15% increase in flutter speed for a 5%
Strut chord, ft 5.9 5.3 increase in fuel consumption. If one allows a relaxed flutter
Jury chord, ft 2.4 3.1 constraint, then POINT 3 can be chosen as a feasible design, which
Root t∕c 0.111 0.051 will only undergo a 2% penalty on the fuel burn. The cost of the
Strut–wing junction t∕c 0.100 0.118 reduced fuel burn against the weight penalties and efficiency of an
Tip t∕c 0.107 0.050 active-flutter-suppression mechanism to increase flutter speed by
Strut t∕c 0.100 0.090 10% may be worth investigating.
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096

Jury t∕c 0.090 0.095

Flutter margin, % −17.60 1.30
Flutter speed, KEAS 367.2 427.6
Flutter frequency, Hz 4.25 5.76
increase in the bending stiffness is what suppresses flutter. Some
major redistribution of wing stiffness and mass is observed as the
MDO tool moves structural material from outboard and inboard
sections toward the wing–strut junctions. The overall stiffening of the
wing increases the frequencies of the modes participating in flutter, as
well as modifying the mode shapes. This modification of the bending
and torsional mode shapes and their frequencies causes a significant
increase in the flutter margin. If one allows a relaxed flutter
constraint, and then picks a design POINT3 from the results, the
penalty on TOGW is restricted to only 0.3%, provided some active-
flutter-suppression mechanism can be used to increase flutter speed
by 7%.
Looking now at Fig. 13 for the minimum-fuel objective, we
observe that, moving from DESIGNS1 to DESIGNS2, a gradual
increase in flutter margin is observed as in the last case until a point is
reached where the constraint is satisfied. Here, this is associated with
a gradual increase in the fuel consumption of the designs, indicating
again that the flutter constraint is active. To obtain a clearer
understanding of this effect, two designs, POINT 1 and POINT 2,
have been chosen, which represent the best designs from those
obtained without the flutter constraint and those obtained after
satisfying the constraint, respectively. These are illustrated in Fig. 14
and compared against each other in Table 4. Going from POINT 1 to
POINT 2, there is a 5% increase in the fuel consumption associated
with satisfying the flutter constraint. The increase in fuel
It is interesting to note that the flutter constraint brought the best
designs from minimizing TOGW and from minimizing fuel
consumption closer to each other. Thus, without any flutter constraint
added, the best design from minimizing TOGW weighs 1.9% less and
has 12.2% more fuel consumption than the best design from
minimizing fuel consumption. After application of the flutter
constraint, the best design from minimizing TOGW weighs 1.4% less
than and has 6.4% more fuel consumption than the best design from
minimizing fuel consumption. This can be easily explained by the
fact that the flutter constraint increases the bending weight of the best
minimum-TOGW design by 12% by thickening the wing and a small
increase in the span. The overall effect is added TOGW and a slightly
better aerodynamic performance. The flutter constraint acting on the
best minimum-fuel-consumption design reduces wingspan by 12%,
thereby bringing about a reduction in aerodynamic performance,
which leads to higher fuel consumption.
It was a critical part of the study to compare the unconventional
TBW configurations against the conventional tube and wing, or
cantilever aircraft configurations. To perform such a comparison, the
same optimization procedure, as used for the TBWs, was followed
with the same flight envelope. It was observed that the best cantilever
designs obtained using optimization do not flutter within the
provided flutter boundary; thus, the flutter constraint does not play
any part in changing the feasible designs. The TOGW and the fuel
consumption for the optimized cantilever and TBW configurations
for both the objectives have been listed in Table 5. It can be observed
that, while minimizing for TOGW, the best TBW design obtained
does not show significant benefits in weight reduction over the
conventional cantilever designs. However, the TBW designs show a
3.5% reduction in fuel consumption. On the other hand, while
minimizing fuel consumption, the TBW designs show a 6%

Fig. 13 Designs obtained by minimizing fuel burn for the medium-range mission.
 8 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ

Fig. 14 Minimum-fuel optimized designs for medium-range mission: POINT 1 (left) and POINT 2 (right).
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096

reduction in fuel consumption. Thus, although applying the flutter constraint is applied, the MDO design environment pushes the strut-
constraint takes away some of the benefits of the high-aspect-ratio span intersection outboard to 60–65% of the main wingspan. The
TBW configurations, significant fuel efficiency of the TBW over the strut stiffens the wing as it moves outboard, because it provides
cantilever is retained. kinematic boundary conditions analogous to a very stiff rotational
and translational spring connected to the wing. However, if the strut is
2. Sensitivity Studies too long it becomes flexible itself, as bending stiffness decreases by
Some design variables were investigated for the minimum-fuel- the third power of the length of the member. Thus, the constraint
consumption case to observe the sensitivity to the flutter speed. chooses an optimum strut location via the MDO tool.
Figure 15a shows the geometry of the wing–strut intersection, and Figure 17 shows the sensitivity of the flutter margin to the fore–aft
Fig. 15b shows the midchord of the fuselage–strut intersection location of the strut–fuselage intersection relative to the intersection
relative to the midchord of the wing–fuselage intersection. Figure 16 of the wing and the fuselage. The MDO without the flutter constraint
shows the sensitivity of flutter margin to the strut-span intersection. It selects a strut that intersects the fuselage about 1 ft behind the
can be observed that the MDO without the flutter constraint selects a midchord location of the wing–root. However, once the constraint is
strut intersecting the wing at 50% span. However, once the flutter applied, the MDO tool pushes the strut–fuselage intersection forward
to about 2 ft ahead of the midchord location of the main wing. Several
papers [19,31] explain how shifting the c.g. of the wing forward
increases the flutter velocity. The forward movement of the strut
Table 4 Minimum-fuel optimized TBW designs for intersection serves a similar purpose and, hence, increases the flutter
the medium-range mission
velocity. Thus, the constraint pushes it forward once it is applied in
POINT 1 POINT 2 the MDO tool.
Minimum-fuel TBW (no flutter (with 1.15V D
one jury constraint) flutter constraint) B. Long-Range-Mission Designs
Fuel weight, lb 23,700 24,900 (5%) 1. Optimization Results
TOGW, lb 141,000 142,500
Structural bending weight, lb 30.400 29,800 The procedure pursued for the medium-range mission was
Design L∕D 37.5 35.1 reemployed for the long-range mission. Designs obtained without
Wing/strut semispan, ft 97.4∕49.6 85.6∕48.6 applying the flutter constraint have been termed as DESIGNS1,
Wing one-fourth chord sweep, deg 5.7 8.2 represented by dark dots in Figs. 18 and 20. The designs obtained in
Root chord, ft 14.4 14.6 the subsequent step, starting with the saved designs from the previous
Strut–wing junction chord, ft 8.9 8.2 runs and applying the flutter constraint, are termed as DESIGNS2,
Tip chord, ft 3.4 4.1
Strut chord, ft 3.6 4.0
represented by squares in Figs. 18 and 20.
Jury chord, ft 3.0 3.2 Figure 18 shows that, as one moves from DESIGNS1 to
Root t∕c 0.107 0.111 DESIGNS2, a gradual increase in flutter margin is observed
Strut–wing junction t∕c 0.136 0.122 associated with a gradual increase in the takeoff weight, indicating
Tip t∕c 0.063 0.092 that the flutter constraint is active. Two designs, POINT 1 and POINT
Strut t∕c 0.083 0.115 2, representing the best designs from those obtained without the
Jury t∕c 0.098 0.083 flutter constraint and those obtained after satisfying the constraint
Flutter margin, % −15.30 0.01
respectively, will be discussed in detail. These are shown in Fig. 19
Flutter speed, KEAS 372.6 417.2
Flutter frequency, Hz 3.25 4.26 and compared against each other in Table 6. Observe that, as we go
from POINT 1 to POINT 2, there is a 3.2% increase in the TOGW

Table 5 Comparison of TBW with cantilever configurations for the medium-range mission
Minimum-TOGW Minimum-TOGW Minimum-fuel Minimum-fuel
optimized cantilever optimized TBW optimized cantilever optimized TBW
TOGW, lb 141,500 140,600 (−0.6%) 142,500 142,500
Fuel weight, lb 27,500 26,500 (−3.5%) 26,400 24,900 (−6%)
 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ 9

Fig. 15 Connection of strut with the wing and the fuselage.

Sensitivity of Flutter margin to strut−span intersection 5 TOGW vs Flutter margin

Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096

5 4.85 x 10
Designs1
0 Designs2
1.15 V 4.8 Point 2
Flutter margin (%)

D
−5

TOGW (lbs.)
3.5 % penalty
−10 4.75 on TOGW to
satisfy flutter
constraint
−15
4.7
−20 Flutter boundary

−25 4.65
Point 1
0.45 0.5 0.55 0.6 0.65
Strut span / wingspan
4.6
Fig. 16 Sensitivity of flutter margin to strut–wing intersection location. −50 −40 −30 −20 −10 0 10
Flutter margin (%)
Fig. 18 Designs obtained by minimizing the TOGW for the long-range
mission.

Sensitivity of Flutter margin to strut−fuselage intersection

10
1.15 VD 5 Fuel burn vs Flutter margin
1.5 x 10
0 Designs1
Flutter margin (%)

midchord of Designs2
wing−fuselage
Point 2
connection
−10
Fuel burn (lbs.)

1.45
7.5 %
increase in
fuel burn to
−20 satisfy
flutter
constraint
1.4
−30
−5 −4 −3 −2 −1 0 1 Point 3
Distance along the flow (ft.)
Flutter boundary
Fig. 17 Sensitivity of flutter to location of strut–fuselage connection Point 1
along the flow. 1.35
−25 −20 −15 −10 −5 0 5
Flutter margin (%)
Fig. 19 Minimum-TOGW optimized designs for long-range mission:
POINT 1 (left) and POINT 2 (right).
associated with satisfying the flutter constraint. This increase is
caused by a 21.9% increase in the bending weight of the wing–truss
system. The bending-weight increase is partly caused by an increase
in the span of the main wing and the jury, but mainly by the thicker boundary. Two designs, POINT 1 and POINT 2, which represent the
wing. A significant increase in the stiffness of the inboard sections of best designs from those obtained without the flutter constraint and
the wing is observed. The stiffening of the wing increases the those obtained after satisfying the constraint, respectively, were
frequencies of the modes participating in flutter, as well as modifying selected to observe the changes imposed by the MDO tool on the
the mode shapes. This modification of the bending and torsional design variables and parameters to satisfy the flutter constraint. These
mode shapes and their frequencies causes an increase in the flutter designs are shown in Fig. 21 and compared against each other in
margin. Table 7. Note that, going from POINT 1 to POINT 2, there is a 7.5%
Consider now the minimum-fuel-burn results. Figure 20 shows an increase in fuel consumption associated with satisfying the flutter
increase in flutter margin moving from DESIGNS1 to DESIGNS2 constraint. The increase in fuel consumption can be attributed to the
associated with an increase in fuel consumption. However, in this reduced aerodynamic efficiency of POINT 2 compared to POINT 1,
case, the rise in fuel burn is quite steep close to the flutter boundary, indicated by their L∕D values. Lowered aerodynamic efficiency can
indicating that the flutter constraint is very active close to its be attributed to the reduced span and aspect ratio of POINT 2
 10 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ

Table 6 Minimum-TOGW optimized TBW designs Table 7 Minimum-fuel optimized TBW designs
for the long-range mission for the long-range mission
POINT 1 POINT 2 POINT 1 POINT 2
Minimum-TOGW (no flutter (with 1.15V D Minimum-fuel (no flutter (with 1.15V D
TBW one jury constraint) flutter constraint) TBW one jury constraint) flutter constraint)
TOGW, lb 465,300 479,700 (3.2%) TOGW, lb 481,400 482,000
Fuel weight, lb 147,600 148,000 Fuel weight, lb 136,000 146,400 (7.5%)
Structural bending weight, lb 37,000 45,100 Structural bending weight, lb 53,900 47,900
Design L∕D 28.0 28.9 Design L∕D 32.7 29.0
Wing/strut semispan, ft 122.3∕68.4 125.5∕69.1 Wing/strut semispan, ft 144.8∕67.81 130.8∕71.1
Wing one-fourth chord sweep, deg 29.0 28.9 Wing one-fourth chord sweep, deg 28.0 27.9
Root chord, ft 19.6 20.1 Root chord, ft 20.2 19.3
Strut–wing junction chord, ft 14.2 15.0 Strut–wing junction chord, ft 14.5 13.6
Tip chord, ft 11.1 11.2 Tip chord, ft 11.8 10.5
Strut chord, ft 10.7 9.0 Strut chord, ft 10.3 10.8
Jury chord, ft 3.0 3.0 Jury chord, ft 3.1 3.0
Root t∕c 0.090 0.113 Root t∕c 0.112 0.127
Strut–wing junction t∕c 0.073 0.098 Strut–wing junction t∕c 0.096 0.090
Tip t∕c 0.083 0.110 Tip t∕c 0.117 0.113
Strut t∕c 0.098 0.094 Strut t∕c 0.104 0.106
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096

Jury t∕c 0.075 0.086 Jury t∕c 0.086 0.066

Flutter margin, % −28.38 0.406 Flutter margin, % −18.74 0.789
Flutter speed, KEAS 260.31 390.73 Flutter speed, KEAS 315.38 392.22
Flutter frequency, Hz 1.83 4.16 Flutter frequency, Hz 2.55 3.46

compared to POINT 1. Table 7 shows a reduction in the bending
weight primarily driven by the reduction in the span of the wing. The
structure is stiffened by the reduction in span. The inboard sections
are stiffened by an increase in thickness ratios going from POINT1 to
POINT2. This is indicated by an increase in the flutter frequency. The
overall effect in the redistribution of the structural stiffness and its
overall increase is a 20% increase in the critical flutter speed of
POINT2. The results also indicate that, if one allows a relaxed flutter
constraint with an active-flutter-suppression mechanism to increase
flutter speed by 7%, then a design, like POINT 3, can be accepted as a

Fig. 20 Designs obtained by minimizing fuel burn for the long-range mission.

Fig. 21 Minimum-fuel optimized designs for long-range mission: POINT 1 (left) and POINT 2 (right).
 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ
Table 8 Comparison of TBW with cantilever configurations for the long-range mission
Minimum-TOGW optimized Minimum-TOGW optimized Minimum-fuel optimized Minimum-fuel optimized
cantilever TBW
TOGW, lb 488,700 479,700 (−1.8%)
Fuel weight, lb 163,500 148,000 (−9.5%)
feasible design. Such a design will only suffer a 2.9% increase in
fuel burn.
As observed for the medium-range missions, the flutter constraint
brought the best designs from minimizing TOGW and from
minimizing fuel consumption closer to each other. Thus, the
minimum-fuel-burn design only has a 4% larger semispan, 1% lower
fuel weight, and a 0.5% larger takeoff weight than its minimum-
TOGW counterpart. Comparing the right-hand sides of Figs. 19 and
21, even the orientation of the jury members and the strut/wingspan
ratios are similar.
The TBW configurations were compared against the cantilever
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096
aircraft configurations for the long-range mission using the same
procedure pursued for the medium-range mission. It was again found
that the best cantilever designs obtained using the optimization tool
do not flutter within the provided flutter boundary. Thus, the flutter
constraint does not play any part in changing the feasible designs. The
TOGW and the fuel burn for the optimized cantilever and TBW
configurations for both the objectives are listed in Table 8. Unlike the
medium-range missions, it was observed that both the minimum-
TOGW and minimum-fuel-burn TBW designs were superior to their
cantilever counterpart in both objectives. The results show that the
minimum-TOGW TBW design has a 1.8% lower takeoff weight and
9.5% lower fuel burn than its minimum-TOGW-cantilever
counterpart. On the other hand, the minimum-fuel-burn TBW design
shows a 4.0% lower takeoff weight and a 6.0% lower fuel burn than
its minimum-fuel-cantilever counterpart.
To explain why the cantilever designs do not flutter, the authors
took a look at the mode shapes and frequencies of the cantilever
design mentioned in Table 8 and shown in Fig. 22a, and compared
them to the TBW design POINT 1 shown in Table 7. The mode
shapes and the frequencies of the TBW POINT 1 optimized for
minimum-fuel weight without any constraint for flutter have been
shown before in Fig. 9. The first out-of-plane bending and torsion
modes, represented by modes 1 and 2, respectively, for the cantilever
design are expected to take part in conventional bending–torsion
flutter for cantilever designs. These modes are shown in Fig. 22b.
Bisplinghoff et al.’s text on aeroelasticity [19] provides nondimen-
sional flutter sensitivity plots for a variety of parameters, which
indicate that, for bending–torsion flutter, the flutter velocity is
expected to go down as the frequencies of the participating modes
come closer to each other, thereby increasing the ease of coupling of
Fig. 22 Long-range minimum-fuel optimized cantilever design.
11
cantilever TBW
502,000 482,000 (−4.0%)
156,000 146,400 (−6.0%)
the two modes with each other. The flutter speed is shown to reach a
minimum when this ratio approaches unity. For the TBW designs,
this ratio of frequencies between the second out-of-plane bending and
the torsion modes, represented by modes 3 and 5, respectively, is
0.511. For the cantilever design, this ratio between the first out-of-
plane bending mode and the torsion modes, which usually participate
in flutter, represented by modes 1 and 2, respectively, is 0.27. This
indicates that, for the TBWs, the ratio of the modal frequencies of the
coupling modes in flutter comes closer to unity than the probable
flutter modes for the cantilever design. This probably is the prime
reason why cantilever designs do not flutter, whereas TBWs do. The
same behavior is expected from the medium-range designs, as the
TBWs show similar changes in structural, aerodynamic, and
aeroelastic behavior from cantilever configurations for both the long-
range and medium-range missions. As one can observe, the bending
frequency of the first out-of-plane bending is relatively low and may
take part in body freedom flutter if it is allowed to go down
uninhibited. However, such an analysis would require an actual free–
free vehicle modeling, which is beyond the scope of the present study.
2. Nonlinear Effects
Although a nonlinear prestressed aeroelastic analysis showed only
4.5% lower flutter velocity than the existing results during the
benchmarking studies, the authors performed a similar off-line
aeroelastic analysis on several other TBW designs. We choose the
TBW designs POINT 1 and POINT 2 for the minimum-fuel long-
range and medium-range missions. These designs, detailed in
Tables 4 and 7, respectively, are expected to be the more flexible ones
for each of these missions and, hence, are expected to show the
maximum effects of a nonlinear prestressed aeroelastic analysis.
Table 9 compares the linear and nonlinear flutter speeds and
frequencies. The acronyms MR and LR stand for medium-range and
long-range missions, respectively. The LR POINT 1 is the same
design as used in the flutter-benchmarking study. The POINT 1
designs were the optima obtained before applying the flutter
constraint, and the POINT 2 designs were the optima obtained after
applying the constraint. The results indicate that the nonlinear results
do show somewhat different flutter speeds and frequencies for the
POINT 1 designs. For the POINT 2 designs, which are expected to
be less flexible, the difference in results is definitely reduced once the
designs are stiffened by the constraint to prevent flutter. Thus, the
 12 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ
Table 9 Comparison of linear and nonlinear flutter velocities and frequencies
Velocity, KEAS
Linear Nonlinear % Change Linear Nonlinear % Change
MR minimum-fuel POINT 1 372.60 355.30
MR minimum-fuel POINT 2 417.20 412.30
LR minimum-fuel POINT 1 315.38 301.34
LR minimum-fuel POINT 2 377.30 393.22
final results for flutter shown in this study can be assumed to be stiff
enough to obtain reasonably accurate results by employing linear
prestressed flutter analysis.
IV. Conclusions
The present study shows the effects of the application of a flutter
constraint on TBW transport aircraft designs obtained from the MDO
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096
design environment that has been developed at the Virginia
Polytechnic Institute and State University. The two objective
functions considered for the optimization were minimization of the
takeoff weight and the fuel consumption. Previously, this MDO
design environment has been used both for long-range and medium-
range flight missions, in which the studies have shown significant
benefits of the TBW configuration over the cantilever wing designs
without considering flutter. Now, a medium-range mission with a
lower cruise Mach number and the previous long-range mission have
been chosen to observe the effect of a new flutter constraint on the
designs obtained from the MDO framework. The flutter boundary
was developed based on the flight envelope and the dive speed of the
aircraft at various altitudes. Then, a flutter constraint was used to
calculate the margin between the critical flutter speed and the flutter
boundary.
The results obtained from the MDO study show that, when the
TBW configurations are optimized for the minimum TOGW for the
medium-range mission, applying the usual 1.15V D flutter constraint
would provide a best design, which has 1.5% higher takeoff weight
than the best TBW obtained without that constraint. The major
contribution to this increase in takeoff weight is from a 12% increase
in the bending weight of the wing structure. The flutter constraint
applied to the minimum-TOGW designs for the long-range mission
resulted in a best design having a 3.2% higher takeoff weight than the
best TBW obtained without the flutter constraint. Again, the major
contributor is the 18.5% increase in the bending weight. The MDO
tool brings about an addition of bending material weight and a
redistribution of stiffness and mass, as well as a reduction of wing
thickness at the wing root and tip, and a significant increase in t∕c at
the wing–strut junction. An increase in the wingspan is observed in
both cases. However, for the medium-range case, the MDO stiffened
the wing–strut junction, whereas for the long-range mission inboard
stiffening of the wing was observed. These bring about sufficient
changes in the mode shapes participating in flutter while increasing
the frequency of the bending modes. The overall effect is to increase
the flutter speed so as to satisfy the flutter constraint.
Imposing the 1.15V D flutter constraint to the designs obtained
while minimizing fuel consumption for the medium-range mission
provides the best design having 5% higher fuel consumption than the
best design obtained previously without the constraint. For the long-
range mission, the increase in fuel burn observed is 7.5%. However,
the steep gradient of increase in fuel burn with increase in flutter
margin close to the flutter boundary may render active flutter-control
mechanisms attractive. The flutter constraint directs the MDO tool to
designs having significantly shorter spans. Some redistribution of
stiffness and mass is observed. A slightly thicker strut, a slightly
thicker outboard section, and thinner wing–strut sections for the main
wing are found for the medium-range case. For the long-range case,
thicker inboard sections are observed. The significant stiffening in
bending produced by the shortening of the wingspan and the thicker
inboard sections for the long-range-mission design can be thought of
as the major contributor to the increased flutter margin. This,
Frequency, Hz
−4.6 3.25 3.04 −6.5
−1.2 4.26 4.10 −3.8
−4.5 2.55 2.32 −8.2
−3.8 3.46 3.34 −3.5
however, is accompanied by lower L∕D and aspect ratios, which
reduce aerodynamic efficiency.
The best TBW designs obtained with the new MDO tool with
flutter were compared to the best cantilever designs obtained with the
same tool. It was found that the flutter constraint was a passive
constraint for the cantilever wing designs, as the best designs do not
experience flutter within the flight envelope. The results show that,
while minimizing TOGW for the medium-range mission, the best
TBW design shows a minimal reduction in takeoff weight over the
best minimum-TOGW cantilever, but it has about 3.5% lower fuel
consumption. While minimizing fuel consumption, it was found that
the TBW design shows a 6% reduction in fuel consumption over the
best minimum-fuel-cantilever design. For the long-range mission, it
was observed that the best TBW design shows a 1.8% reduction in
takeoff weight over the best minimum-TOGW-cantilever-wing
cases, and it has about 9.5% lower fuel consumption over the
cantilever. While minimizing the fuel consumption, it was observed
that the TBW design shows a 4% reduction in TOGW and a 6%
reduction in fuel consumption over the best minimum-fuel-
cantilever-wing design. Thus, it can be concluded that minimizing
fuel consumption would be the more productive objective function
for the TBW designs for the current medium-range mission to
achieve the full benefits of these large-aspect-ratio truss-braced wing
aircraft. On the other hand, for the long-range mission, both of the
objectives render TBW designs superior to the cantilever both in
takeoff weight and fuel burn.
The sensitivities of the flutter margin of the TBW designs to
various relevant design variables were studied for the medium-range
mission. The results show that the MDO environment considers the
optimum location of the wing-strut intersection to be at 60% of the
span to prevent flutter. Also, it was observed that a strut–fuselage
connection forward of the midchord–fuselage junction of the main
wing is beneficial for flutter. These were some of the passive
measures for flutter control that have been investigated.
The results for both the medium-range-mission objectives as well
as the minimum-fuel-burn case for the long-range mission show a
steep increase in the objective function close to the flutter boundary.
Thus, if one applies a relaxed flutter constraint, some other designs
become available, which face much smaller penalties on their takeoff
weight or fuel burn. These designs become feasible by applying
active or passive flutter-suppression mechanisms, which can bring
about up to 10% increase in the flutter speed. The weight penalties
associated with such flutter-suppression mechanism as well as their
efficiency will be investigated in a future study.
To understand why the cantilever designs do not flutter whereas the
TBWs do, the authors took a look at the ratios of the frequencies of the
main modes participating in bending–torsion flutter for both
configurations. The results show that the ratio of the first bending and
torsion modes for the cantilever is further apart than the frequencies
of the second out-of-plane bending and the first torsion mode of the
TBW. Based on the notion that the flutter velocity drops to its lowest
when the frequencies of the modes come closest to each other, it is
believed that this is the prime reason why cantilever wings do not
flutter, whereas the TBW configurations do. However, it is interesting
to note that the cantilever designs optimized for minimum fuel may
have very low first-bending frequencies, which may lead to body
freedom flutter. Although such an analysis performed for a free–free
vehicle is beyond the scope of the existing analysis, such an event
would also constrain the performance of cantilever designs, and make
the TBW configurations even more superior.
 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ 13

The authors also compare the results of a nonlinear prestressed M  MNC  MC

flutter analysis with the existing linear prestressed flutter analysis,      
and found that, while the flutter frequencies and velocities differ 1 1
 πρbref bref ah  Vbref − a θ − bref  a aθ
2  _ 2 2
slightly for the unconstrained TBW designs, the basic physical 2 8
behavior is the same. Also, once the constraint is applied and the      
2πρVbref 1 1
TBW structure is stiffened to prevent flutter, the difference in the   a Ckref  h_  Vθ  bref − a θ_
 p
results between the linear and nonlinear prestressed flutter analyses is 1 − M∞ cosΛ2 2 2
very small at least for simplified modeling used in the present (B3)
analysis.
 
Appendix A: Optimization Constraints L
 QA 
The following 11 constraints, other than the flutter constraint, have M
been used in the MDO studies:     _  
b2ref h bref h h
1) The cruise-range constraint ensures that the designs have  πρV 2 2
AM   AC   AK 
enough fuel to complete the flight-mission range of 3315 n mile with V θ V θ_ θ
an additional 200 n mile (reserve). (B4)
2) The initial-cruise rate-of-climb (ROC) constraint ensures that
the ROC at cruise flight conditions and initial-cruise weight is higher
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096

than 300 ft∕ min. Here, M∞ is the freestream Mach number used to develop the
3) The maximum cl constraint ensures that the local two- quasi-steady Prandtl–Glauert compressibility correction for the slope
dimensional (2-D) lift coefficient cl at cruise flight conditions is not of the coefficient of lift. If we assume that harmonic oscillations take
greater than 0.8. This value represents a typical 2-D operating limit place at a frequency ω, then the local sectional model’s degrees of
for transonic airfoils. freedom can be written in terms of the assumed modal coordinates hf
4) The fuel-capacity constraint ensures that the available fuel and θf as
volume in the fuel tanks is higher than the required fuel for mission
accomplishment.    
h hf iωt
5) The wingtip-deflection constraint ensures that, at the taxi-bump  e (B5)
θ θf
condition, 2.0g acceleration, the wingtip deflection does not exceed
the fuselage diameter, which is 12.4 ft.
6) The second-segment-climb constraint ensures that, during Substituting Eq. (B5) into Eq. (B4) and using Eq. (B1), the
takeoff at the second-segment conditions (1.2 stall speed), the climb expression for the aerodynamic forces in frequency domain can be
gradient should be higher than 2.4% (the requirement for a two- written as
engine vehicle according to the FAR).
7) The approach-velocity constraint ensures that, during the  
h
approach condition (Cl  1.52, landing-gear down), the vehicle QA   πρV 2
−k2ref AM   ikref AC   AK  f eiωt
can maintain true airspeed, which is less than 132.5 kt. θf
8) The missed-approach constraint ensures that, during landing,  πρV 2 Akref  (B6)
missed-approach conditions (Cl  1.52, landing-gear up), the
climb gradient should be higher than 2.1% (the requirement for a two- in which the aerodynamic matrices are written as
engine vehicle according to the FAR).
9) The balanced field constraint ensures that the takeoff and  
landing balanced field length is less than 8700 ft. The balanced field −1 bref a
AM   (B7)
constraint is calculated according to the Roskam Model. bref a −b2 0.125  a2 
10) The landing-distance constraint ensures that the landing
distance satisfies the FAR regulations.
11) The geometric constraint penalizes configurations that are 2 3
either infeasible or not supported by the current analysis methods. −2Ckref 
p
 −bref
p12Ck ref 0.5−a


6 1−M∞ cos Λ 2 1−M∞ cos Λ2 7
AC   4 2bref Ckref 0.5a b2ref 0.25−a2 −12Ckref  5 (B8)
p2 p2
Appendix B: Mathematical Formulation of the Flutter 1−M∞ cos Λ a 1−M∞ cos Λ
Analysis
The unsteady aerodynamics uses Theodorsen’s method [18] to
evaluate aerodynamic lag functions. The nondimensional reduced
2 3
frequency kref is based on the reference airfoil semichord bref , the 0 p−2b ref Ckref  

flight velocity V, and the frequency of oscillation ω as AK   4
1−M∞ cos Λ 2
5 (B9)
2b2ref Ckref 0.5a
0 p2
ωbref 1−M∞ cos Λ
kref  (B1)
V

The complex aerodynamic lag function or the lift deficiency
function Ckref  is in terms of the reduced frequency kref , as explained
by Theodorsen [18]. Finally, for a 2-D airfoil section, the lift L and the
pitching moment about the elastic axis M of a swept wing are defined
in terms of the circulatory and noncirculatory parts based on the
airfoil’s pitch h and plunge θ degrees of freedom as
Once the mass matrix M and the stiffness matrix K are evaluated
in terms of the beam finite element degrees of freedom, and the
normal modes obtained from the modal analysis [11] are grouped
together in the Ξ matrix, one can use the orthonormal property of the
normal modes to obtain a mass-normalized reduced-order system as
follows:

L  LNC  LC  πρb2ref h  V θ_ − bref aθ  Mξ   ΞMΞT

   
2πρVbref 1 (B2) Kξ   ΞMΞT
 Ckref  h_  Vθ  bref
p − a θ_
1 − M∞ cosΛ 2 2 Akref ξ   ΞAkref ΞT (B10)
 14 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ

We can obtain the equations of motion of the aeroelastic system References

using Lagrange’s equations. For more details, please see [11]. It can
be written as
1 2π
Mξ fξtg
  Kξ fξtg  ρV 2 p Akref ξ fξtg
2 1 − M∞ cos Λ2
in which ξt represents the generalized coordinates as a function of
time. Because the aeroelastic system is assumed to follow harmonic
oscillations, the generalized coordinates can be represented in terms
of the mode shapes fξF g and the frequency of oscillation ω as
fξtg  fξF geiωt
Substituting Eq. (B12) into Eq. (B11), we get an eigenvalue
problem as shown in the following:
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096
[1] Pfenninger, W., “Laminar Flow Control Laminarization, Special
Course on Concepts for Drag Reduction,” AGARD Rept. 654, von
Karman Institute for Fluid Dynamics, Rhode-St-Genese, Belgium,
1977.
[2] Gundlach, J. F., Tétrault, P. A., Gern, F. H., Nagshineh-Pour, A. H., Ko,
A., Schetz, J. A., Mason, W. H., and Kapania, R. K., “Conceptual Design
(B11) Studies of a Strut-Braced Wing Transonic Transport,” Journal of
Aircraft, Vol. 37, No. 6, 2000, pp. 976–983.
doi:10.2514/2.2724
[3] Gur, O., Bhatia, M., Mason, W. H., Schetz, J. A., Kapania, R. K., and
Nam, T., “Development of a Framework for Truss-Braced Wing
Conceptual MDO,” Structural and Multidisciplinary Optimization,
Vol. 44, No. 2, 2011, pp. 277–298.
doi:10.1007/s00158-010-0612-9
(B12) [4] Gur, O., Bhatia, M., Schetz, J. A., Mason, W. H., Kapania, R. K., and
Mavris, D. N., “Design Optimization of a Truss-Braced-Wing Transonic
Transport Aircraft,” Journal of Aircraft, Vol. 47, No. 6, 2010, pp. 1907–
1917.
doi:10.2514/1.47546
[5] Seber, G., Ran, H., Schetz, J. A., and Mavris, D. N., “Multidisciplinary

  Design Optimization of a Truss Braced Wing Aircraft with Upgraded

1 2π Aerodynamic Analyses,” 29th AIAA Applied Aerodynamics
−ω2 Mξ   Kξ  − ρV 2 p Akref  fξF g  0 Conference, AIAA Paper 2011-3179, June 2011.
2 1 − M∞ cos Λ2 [6] Meadows, N. A., Schetz, J. A., Kapania, R. K., Bhatia, M., and Seber,
(B13) G., “Multidisciplinary Design Optimization of Medium-Range
Transonic Truss-Braced Wing Transport Aircraft,” Journal of Aircraft,
Vol. 49, No. 6, 2012, pp. 1844–1856.
For nontrivial solutions, one must have doi:10.2514/1.C031695
[7] Green, J. E., “Laminar Flow Control—Back to the Future?” 38th Fluid
  Dynamics Conference and Exhibit, AIAA Paper 2008-3738, June 2008.
1 2π [8] Gur, O., Mason, W. H., and Schetz, J. A., “Full-Configuration
Det −ω2 Mξ   Kξ  − ρV 2 p Akref   0
2 1 − M∞ cos Λ2 Drag Estimation,” Journal of Aircraft, Vol. 47, No. 4, 2010, pp. 1356–
1367.
(B14) doi:10.2514/1.47557
[9] Gur, O., Schetz, J. A., and Mason, W. H., “Aerodynamic Considerations
in the Design of Truss-Braced-Wing Aircraft,” Journal of Aircraft,
In the k method, it is assumed that a fictitious damping g similar to Vol. 48, No. 3, 2011, pp. 919–939.
structural damping is added to the system. Thus, using Eq. (B1), doi:10.2514/1.C031171
Eq. (B14) can be rearranged as [10] Sulaeman, E., Kapania, R. K., and Haftka, R. T., “Parametric Studies of
Flutter Speed in a Strut-Braced Wing,” 43rd AIAA/ASME/ASCE/AHS/
 2  ASC Structures, Structural Dynamics, and Materials Conference,
kref 1 2π
Det Mξ   ρ p Akref  − λF Kξ   0 AIAA Paper 2002-1487, April 2002.
bref 2 1 − M∞ cosΛ2 [11] Bhatia, M., Gur, O., Kapania, R. K., Mason, W. H., Schetz, J. A., and
(B15) Haftka, R. T., “Progress Towards Multidisciplinary Design
Optimization of Truss Braced Wing Aircraft with Flutter Constraints,”
13th AIAA/ISSMO Multidisciplinary Analysis Optimization
in which for a given reduced frequency kref, the complex eigenvalue Conference, AIAA Paper 2010-9077, Sept. 2010.
λF is defined as [12] Bhatia, M., Kapania, R. K., and Haftka, R. T., “Structural and
Aeroelastic Characteristics of Truss-Braced Wings: A Parametric
1  ig Study,” Journal of Aircraft, Vol. 49, No. 1, 2012, pp. 302–310.
λF  (B16) doi:10.2514/1.C031556
V2
[13] Gupta, R., Mallik, W., Kapania, R. K., and Schetz, J., A.,
“Multidisciplinary Design Optimization of Subsonic Truss-Braced
The real and imaginary parts of λF are related to the frequency and Wing Cargo Aircraft,” 52nd Aerospace Sciences Meeting, AIAA Paper
damping of the system as 2014-0186, 2014.
[14] Mallik, W., Kapania, R. K., and Schetz, J. A., “Aeroelastic Analysis and
1 ImagλF  Optimization of Truss-Braced Wing Aircraft with Novel Control
V  p g (B17) Effectors,” 56th AIAA/ASCE/ASME/AHS/ASC Structures, Structural
RealλF  RealλF 
Dynamics, and Materials Conference, AIAA Paper 2015-1175,
Jan. 2015.
Because flutter represents a dynamic instability, we can say that, [15] Anon., Phoenix Integration, “ModelCenter 10.2,” http://www.phoenix-
when the value of g becomes zero, the system is neutrally stable. It is int.com/software/phx-modelcenter.php [retrieved 26 Feb. 2013].
also the only condition at which the whole assumption of harmonic [16] Martins, J. R., Alonso, J. J., and Reuther, J. J., “A Coupled-Adjoint
oscillation is valid. Thus, Eq. (B15) is solved for various values of kref Sensitivity Analysis Method for High-Fidelity Aero-Structural Design,”
Optimization and Engineering, Vol. 6, No. 1, 2005, pp. 33–62.
with an initial guess of Mach number until the value g becomes zero.
doi:10.1023/B:OPTE.0000048536.47956.62
The flutter velocity is evaluated and Mach number is calculated [17] Greenhalgh, D., and Marshall, S., “Convergence Criteria for Genetic
again. This process is repeated until the initial guessed Mach number Algorithms,” SIAM Journal on Computing, Vol. 30, No. 1, 2000,
converges to the final evaluated Mach number. At this point, the pp. 269–282.
evaluated flight speed for a zero g is considered as the flutter speed. doi:10.1137/S009753979732565X
By substituting the value of kref in Eq. (B1), the corresponding flutter [18] Theodorsen, T., “General Theory of Aerodynamic Instability and the
frequency is obtained. Mechanism of Flutter,” NACA Rept. 496, 1949.
[19] Bisplinghoff, R. L., Ashley, H., and Halfman, R. L., Aeroelasticity,
Dover, New York, 1996, pp. 532–555, Chap. 9.
[20] Hoerner, S. F., and Borst, H. V., “Fluid-Dynamic Lift: Practical
Acknowledgment Information on Aerodynamic and Hydrodynamic Lift,” 2nd ed., 1985,
pp. 7-1–7-12.
The work was funded by NASA Langley Research Center.
 AIAA Early Edition / MALLIK, KAPANIA, AND SCHETZ
[21] Beddoes, T., “Practical Computation of Unsteady Lift,” Vertica, Vol. 8,
No. 1, 1984, pp. 55–71.
[22] Leishman, J., “Validation of Approximate Indicial Aerodynamic
Functions for Two-Dimensional Subsonic Flow,” Journal of Aircraft,
Vol. 25, No. 10, 1988, pp. 914–922.
doi:10.2514/3.45680
[23] Wagner, H., “Uber die Einstehung des Dynamischen Auftriebes von
Tragfugeln,” Zeitschrift fur Angewandte Mathematik und Mechanik,
Vol. 5, No. 1, Feb. 1925, pp. 17–35.
[24] Crouse, G., and Leishman, J., “Transonic Aeroelasticity Analysis Using
State-Space Unsteady Aerodynamic Modeling,” Journal of Aircraft,
Vol. 29, No. 1, 1992, pp. 153–160.
doi:10.2514/3.46139
[25] Issac, J. C., and Kapania, R. K., “Sensitivity Analysis of Aeroelastic
Response of a Wing in Transonic Flow,” AIAA Journal, Vol. 32, No. 2,
1994, pp. 350–356.
doi:10.2514/3.59995
[26] Sulaeman, E., “Effect of Compressive Force on Aeroelastic Stability of a
Strut-Braced Wing,” Ph.D. Thesis, Virginia Polytechnic Institute and
State Univ., Blacksburg, VA, 2001.
[27] Demasi, L., Cavallaro, R., and Márquez Razón, A., “Postcritical
Downloaded by UNIVERSITY OF MISSISSIPPI on June 21, 2015 | http://arc.aiaa.org | DOI: 10.2514/1.C033096
Analysis of PrandtlPlane Joined-Wing Configurations,” AIAA Journal,
15
Vol. 51, No. 1, 2012, pp. 161–177.
doi:10.2514/1.J051700
[28] Blair, M., Canfield, R. A., and Roberts, R. W., “Joined-Wing Aeroelastic
Design with Geometric Nonlinearity,” Journal of Aircraft, Vol. 42,
No. 4, 2005, pp. 832–848.
doi:10.2514/1.2199
[29] Cavallaro, R., Demasi, L., and Bertuccelli, F., “Risks of Linear Design of
Joined Wings: A Nonlinear Dynamic Perspective in the Presence of
Follower Forces,” 54th AIAA/ASME/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference, AIAA Paper 2013-
1558, 2013.
[30] Coggin, J. M., Kapania, R. K., Zhao, W., Schetz, J. A., and
Siddaramaiah, V., “Nonlinear Aeroelastic Analysis of a Truss Based
Wing Wind Tunnel Model,” 55th AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference, AIAA
Paper 2014-0335, 2014.
[31] Cox, D., Curtiss, H. C., Jr., Edwards, J. W., Hall, K. C., Peters, D. A.,
Scanlan, R. H., Simiu, E., Sisto, F., Strgnac, T. W., and Dowell, E. H., “A
Modern Course in Aeroelasticity,” edited by Dowell, E. H., Solid
Mechanics and Its Applications, 4th ed., Vol. 116, Springer, New York,
2004, pp. 81–87, Chap. 3.