A Fast Tool For Structural Sizing, Aeroelastic Analysis and Optimization in Aircraft Conceptual Design

50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference<br>17th AIAA 2009-2571
4 - 7 May 2009, Palm Springs, California
A Fast Tool for Structural Sizing, Aeroelastic Analysis

and Optimization in Aircraft Conceptual Design
L. Cavagna∗†, S. Ricci‡, L. Riccobene§,

Dipartimento di Ingegneria Aerospaziale, Politecnico di Milano, Milano, 20156, Italy
Most of procedures available in conceptual design estimate airframe weight by statistical

methods and do not take into account for aeroelastic requirements, postponing them to
successive more detailed design phases. If the former task is not based on sound struc-
tural principles, the estimation will be likely to be unreasonable and results in serious
difficulties, especially when non-conventional layouts are considered. Indeed, a large error
in weight estimation will usually have far-reaching effects in the whole project; such errors
have sometimes been the main factors in the unsuccessful designs. Furthermore, the late
discovery of adverse aeroelastic issues may result in significant re-design costs, consider-
able changes in the structural design, limitations in flight envelope, weight penalties, and
in some cases, it may even require to actually close the project. Thus, in order to over-
come the insurgence of these issues, the influence of deformability on flight and handling
performances, on structural weight and on design costs needs to be taken into account as
early as possible in the design process.
NeoCASS (Next generation Conceptual Aero-Structural Sizing Suite) enables the creation
of efficient low-order, medium fidelity models particularly suitable for structural sizing,
aeroelastic analysis and optimization at the conceptual design level. It provides the total
structural weight on physical basis, reducing to the minimum the adoption of statistics, and
includes aeroelastic requirements and performances from the very beginning of the con-
ceptual design phase. The whole methodology is based upon the integration of geometry
construction, aerodynamic and structural analysis codes that combine depictive, computa-
tional, analytical, and semiempirical methods, validated in an aircraft design environment.
I. Introduction
The estimation of the structural weight is an important part of the conceptual design of an airplane.
Obviously, it is impossible to wait until the airplane is completely designed, yet, it is necessary to know,
with considerably accuracy, how much the airplane will weigh and what is the stiffness distribution to
preclude failures and to satisfy strength, stiffness and, more generally speaking, aeroelastic constraints. The
structural design of an airframe is indeed determined by multidisciplinary criteria such as stress, fatigue,
buckling, flutter, control surface effectiveness, manufacturing and costs to mention a few.
Being most of the life-cycle cost of an aircraft incurred during the conceptual design phase, the earlier
an appropriate conceptual configuration can be found, the more economical the whole design process will
be, avoiding costly later redesign and corrections. Analytical formulas for structural weight prediction can
be adopted (see the books by Raymer1 and Torenbeek2 ), usually relating the weight of existing aircraft
to various parameters known to affect the structural weight, i.e. gross weight, span, maximum load factor.
Weight formulas of this type have serious disadvantages since extrapolation of the curves considerably beyond
the range of the basic data may be misleading. Radical changes in configuration such as innovative layouts,
may also introduce serious errors; it appears as quite unreliable to adopt statistical-based approaches where
no enough knowledge is available, as in case of unconventional configurations and new technologies such as
Sensorcraft3, 4 and Blended Wing Body aircraft.5 The use of statistical based approaches for the structural
∗ Ph.D. fellow, Dipartimento di Ingegneria Aerospaziale, Politecnico di Milano, Milano, 20156, Italy.
† Now at FOI, Swedish Defence Research Agency, 164 90 Stockholm, Sweden.
‡ Associate Professor, M.Eng. Ph.D., Dipartimento di Ingegneria Aerospaziale, Milano, 20156, Italy. AIAA Member.
§ M.Eng., Ph.D. candidate, Dipartimento di Ingegneria Aerospaziale, Politecnico di Milano, Milano, 20156, Italy.
1 of 25
American
Copyright © 2009 by the American Institute of Aeronautics and Institute
Astronautics, Inc. All of Aeronautics
rights reserved. and Astronautics
weight estimation means that the aircraft structure is practically absent till the preliminary design phase.
Due to this choice, it is almost impossible to take into consideration aeroelastic requirements that in fact
appear later during the design loop. New transport aircraft are very flexible and aeroelastic effects must be
taken into consideration right from the beginning of the design phase so to avoid very expensive redesign
during preliminary design phase or resulting weight penalties needed to satisfy aeroelastic requirements not
previously taken into account.
The capability of including stiffness and aeroelastic constraints starting from the conceptual design is
particularly challenging and important. Fluid-structure simulations are not very established in conceptual
phase and optimization is even a big step further. As a matter of fact, aeroelastic requirements are usually not
considered by the stress engineers determining structural sizes. In most cases there are significant time-delays
until the design determined by the stress-group is available to aeroelasticians. Shortfalls in the aeroelastic
behavior then demand additional efforts in order to find feasible solutions which may be non-optimal and
expensive. Being aeroelastic synthesis particularly troublesome to determine, remedies to negative impacts
of aeroelasticity on the design can be recovered by optimization strategies to ensure for example flutter-free
structures, excellent multipoint performace characteristics, satisfying different classes of constraints.
Due to program requirements, the development cycles shrink continuously, while the technical demands
grow. These contradictory requirements cannot be fulfilled by traditional sequential engineering. Because
of typical sizes and complexity, there is a clear need for advanced tools integrating and accelerating the
design process. Recently, new software systems specifically tailored for aircraft conceptual design have been
proposed by Antoine et al.6 but while they include specific tools for taking into account of different aspects
and requirements, such as ones coming from environmental impact, the capabilities of considering more
realistic structural models are still missing. In some cases statistical-based approaches are substituted by
deriving structural weight prediction based on single loading parameter, like the root wing bending (see Kroo
and Shevell,7 Ning and Kroo8 ). On the other hand, specific methods based on semi-analytical approaches
have been developed mainly to enhance the weight prediction capabilities (see for example Macci9 and
Bindolino et al.10 ) but in many cases they are specific modules not included into a more general aircraft
conceptual design framework, including aircraft performances and stab! ility and control.
II. NeoCASS: aeroelasticity in conceptual design

Present trends in aircraft design towards augmented-stability and expanded flight envelopes call for an
accurate description of the flight-dynamic behavior of the aircraft in order to properly design the Flight
Control System (FCS). Hence the need to increase knowledge about stability and control (S&C) as early as
possible in the aircraft development process in order to be first-time-right with the FCS design architecture.
The starting point and inspiration for the development of a software named CEASIOM (Computerized
Environment for Aircraft Synthesis and Integrated Optimization Methods), is the QCARD package for
aircraft conceptual design with quasi-analytical shape definitions, aero-data correlations, and performance
predictions developed by Isikveren.11 The CEASIOM code is developed within the frame of the SimSAC12
(Simulating Aircraft Stability And Control Characteristics for Use in Conceptual Design) Specific Targeted
Research Project (STREP) approved for funding by the European Commission 6th Framework Programme
on Research, Technological Development and Demonstration. The SimSAC project aims at significantly en-
hancing CEASIOM functionality by introducing software that initially focuses on rapid low fidelity analysis,
and as appropriate, resorts to higher fidelity numerical simulations. Moreover, the code involves stability and
control driven sizing and optimization earlier in the design cycle than is standard practice today. Engineers
are supported in the conceptual design process of the aircraft, with emphasis on the improved prediction
of stability and control properties achieved by higher-fidelity methods than found in contemporary aircraft
design tools. CEASIOM integrates into one application the main design disciplines, i.e. aerodynamics, struc-
tures, and flight dynamics, impacting on the aircraft performance (see Figure 1). It is thus a tri-disciplinary
analysis brought to bear on the design of the aero-servoelastic aircraft. However, the entire conceptual design
process is not carried out. The framework requires as input an initial layout as the baseline configuration
that is then refined and output as the revised layout. In doing this, CEASIOM, through its simulation mod-
ules, generates significant knowledge about the design in the performance, loads, and stability and control
databases. An overview of the whole framewok and its architecture is reported in Von Kaenel et al.13
In particular, the need of aeroelastic analysis within SimSAC has led to the development of a completely
new module called NeoCASS (Next Generation Conceptual Aero-Structural Sizing Suite) to perform struc-
2 of 25
American Institute of Aeronautics and Astronautics

Figure 1. Geometry centric conceptual design.
tural sizing, aeroelastic analysis and optimization.

NeoCASS is compounded by three different tools:
• Weight and Balance (WB) to have a prediction of non-structural masses and their location mainly
based on statistical handbooks (see Section III);
• GUESS (Generic Unknowns Estimator in Structural Sizing) to have a first guess analytical sizing
of the airframe based on ultimate loads estimated on simple structural principles, one-dimensional
aerodynamics and inertia distribution predicted by WB (see Section IV);
• SMARTCAD (Simplified Models for Aeroelasticity in Conceptual Aircraft Design) which is the numeric
kernel for aero-structural analysis and optimization and can be used to enhance the solution provided
by GUESS (see Section V).
All the sub-modules can be combined sequentially or used as stand-alone applications, providing data to
different applications. For example, running through WB, GUESS and SMARTCAD it is possible to give
a low to high fidelity estimation of inertia to the flight mechanics solvers. On the other hand, once an
input structural model is available, SMARTCAD can enhance the aerodynamic database of the aircraft with
trimmed polars and corrections to aerodynamic derivatives for different flight regimes.
Geometry handling is managed by the CADac module, developed by The Royal Institute of Technology
(KTH) and outlined in Bérard et al.14 CADac adopts a set of geometrical parameters which are general
enough to ensure that a wide array of aircraft morphologies can be represented and analyzed. It also allows
to estimate the room available for the airframe to be designed and all the internal-external volumes and
surfaces.
Figure 2 shows a sample of some geometries that can be modeled ranging from large transport to small
regional aircraft and illustrates the versatility of the geometry module.
An aircraft design geometry is fully described in a unique parametrization based on the Extensible Markup
Language (XML) format to which all different analysis modules refer. This format allows storing each
component of the aircraft and its parameter in a hierarchical and sorted way. It facilitates the sharing of
data as well as the expansion of the dataset, i.e. the number of components of the aircraft can be expanded
at will and it is possible to introduce new components as well as new parameters, thus increasing the range
of morphologies that can be modeled. The XML format is used by most modules compounding CEASIOM,
NeoCASS included, to store and manage data-transfer. This allows a high level of flexibility in updating and
expanding each local database and guarantees a common format is used within the whole design framework.
Furthermore, dedicated pieces of software (filter, converters, parsers) can be adopted to ease the management
of such databases.
As depicted in Figure 3, geometry parameters together with geometry derived variables, e.g. area, mean
chord, leading edge, quarter chord sweep, are given in the aircraft.xml file to:
3 of 25

(a) Boeing 747-100. (b) Transonic cruiser. (c) Regional turbofan.
Figure 2. Sample of parametric geometry models.
Figure 3. NeoCASS layout.
• WB which directly applies statistical formulas relying on common geometry variables;

• GUESS which generates a structural stick model, performs the sizing and generates a Finite-Element
(FE) mesh;
• all other modules within the framework requiring geometry information, e.g. CAD engines, automatic
mesh generators for CFD calculations.
Specific flight points chosen by the designer are given in the file states.xml and directly used as ultimated load
conditions by GUESS to perform the first sizing. Structural concepts (see Section IV) and material properties
adopted for modelling fuselage and lifting surfaces components when performing the sizing are given in the
techno.xml file. After the sizing process, a first estimate of the structural weight and stiffness distribution
is available. A stick mesh model with all elements connectivity, material properties, non-structural lumped
masses coming from WB such as painting, furniture, payload, is exported to the ASCII file stick.dat to be used
by the solver SMARTCAD for aero-structural analyses. Further parameters are provided as setting cards to
the solver to determine and rule the analysis to be carried out. When the parameters are combined to the
mesh file, a smartcad.dat file is available which can be independently re-used and modified by the designer
to carry out the simulation required using SMARTCAD as a stand-alone application. Several outputs are
finally available to the structural engineer for post-processing purposes, e.g. stresses, displacements, and to
the other modules, e.g. stability derivatives, trimmed polars, vibration modes for flight-dynamic simulations
4 of 25

including aeroelastic effects, under the hypothesis of small structural displacements.
III. Weight and balance semi-empirical estimation (WB)

Once an appropriate geometry description has been defined, the next step is to be able to get a first
estimation of weight and balance. At conceptual stage, aircraft weight estimation is not a trivial task,
especially if new unconventional design are investigated: lack of data and continuous design changes are two
main issues a designer has to face. Many aeronautical common approaches like Raymer’s1 or Torenbeek’s2
rely on semi-empirical formulas and need extensive table to search and set correction coefficients, making
process automating uneasy.
The weights of the different aircraft components can be classified in three principal groups:
1. those directly related to Maximum Take-Off Weight (MTOW), i.e. wing, tail and fuselage weight;
2. fixed equipment, function only of passengers accommodation and held constant;
3. fuel and payload weights.
These three groups concur to the final MTOW by means of an iterative method: a user’s defined fuel
amount acts as control variable on the process leading to the minimum MTOW that fulfills airframe strength
requirements at a fixed payload. Of course several payload configurations can be analyzed together with
estimating the Green Manufacturer’s Empty Weight (MEW). MEW and MTOW configurations define cen-
ter of gravity range limits, since the former is the airframe weight comprising also propulsion, furnishing,
miscellaneous contributions and any manufacturer’s tolerances but without fuel and payload, while the latter
is the maximum take-off weight at brake release.
Following this classification, the input set is restricted to external geometry data, such as span or fuselage
length, and a few internal layout information, i.e. cabin length, fuel and payload, which includes passengers,
crew and baggage (design specifications). It should be noted that statistical formula validity is not universal
and their accuracy may vary, depending on the design under consideration. Nevertheless, the main target
of WB is a first reasonable estimate of structural and non-structural items which will be later refined by
the structural sizing and eventually by the optimization framework. Arms calculation, either for external
or internal geometry, is made using volume information relying on the analytical approximation by CADac.
Concentrated masses are also considered separately: if available, additional data such as such APU’s (Aux-
iliary Power Unit) weight or auxiliary fuel tank capacity and position can be specified directly, improving
the accuracy and realism of the weight prediction. A key factor for weight calculation is the maximum fuel
value (and its distribution in the different tanks) that is a required user input; as for internal layout, with
the exception of cabin extent (which although can be roughly estimated in 70%-80% of fuselage length), the
required data can be automatically estimated by WB from statistical and semi-empirical methods.
A dedicated routine calculates inertia matrix using either a coarse approximation, which treats the whole
aircraft as solid with constant density (thus using simple formulas with some correction coefficient, an
example can be found in Raymer1 ), or a refined method based on the Mitchell code.15 Both methods result
in approximations, giving a first estimate of the inertia matrix that will be necessarily refined by GUESS
and SMARTCAD.
IV. Analytical structural sizing (GUESS)

A method based on fundamental structural principles for estimating the load-bearing airframe for fuselage
and lifting surfaces is adopted. The same approach is used within the conceptual design framework presented
in Perez et al.16 in order to have an estimate of the structural weight for a multidisciplinary optimization
of an isolated wing. This method is particularly useful in the preliminary weight estimation of aircraft since
it represents a compromise between the rapid assessment of component weight using empirical methods,
based on actual weights of existing aircraft, and detailed but time-consuming finite-element analysis. Both
methods have particular advantages but also limitations which make them not completely suitable for the
conceptual design phase. The empirical approach is the simplest weight estimation tool, which requires the
knowledge of fuselage and wing weights from a number of similar existing aircraft in order to produce a
linear regression useful to derive the data required for the aircraft to be designed. Obviously, the accuracy
of this method depends upon the quality and quantity of available data for existing aircraft and how much
5 of 25

closely the aircraft under investigation matches them. Thus, this approach is inappropriate for studies of
unconventional aircraft concepts for two reasons:
• since the weight estimating formulas are based on existing aircraft, their application to unconventional
configuration (i.e., canard aircraft) is suspect;
• the impact of advanced technologies and materials (i.e., advanced composite laminates) cannot be
assessed in a straightforward way.
Continuum structures, as fuselages or wings, which would use some combination of solid, flat plate or
shell finite elements, are not easily discretizable and the solution of even a moderately complex model like
an aircraft is computationally intensive and can become a bottleneck in the vehicle synthesis.
Finite-element methods, commonly used in aircraft detailed design, are not appropriate for conceptual design,
as the idealized structure model must be built off-line and many details are missing in such a premature
phase. The following two approaches which may simplify the finite-element model also have drawbacks.
The first aims at creating detailed analysis models at a few critical locations on the fuselage and wing, to
successively extrapolate the results to the entire aircraft. This approach can be misleading because of the
great variety of structural, load and geometric characteristics in a typical design.
The second approach aims instead to creating a coarse model of the aircraft, but this scheme may miss key
loading and stress concentrations.
An alternative intermediate approach exists and it is based on beam theory. The work was originally
performed by Ardema et al.17 and presented as structural sizing tool for fuselage and wing. The method
results in a weight estimate which is directly driven by material properties, load conditions, and vehicle size
and shape, considering classic failure modes of each structural component involved as outlined below, thus
being not confined to an existing data base.
The present work starts from this approach, extending it to the sizing of horizontal and vertical tail planes
to have a complete view of the whole airframe. The distribution of loads and vehicle geometry is accounted
for, since the analysis is done station-by-station along the vehicle longitudinal axis and along the lifting
surface structural chord, giving an integrated weight which depends on local conditions. Classic hypothesis
on inertia distribution according to local amount of volume are considered combined with one-dimensional
aerodynamic analytical load distribution.
Figure 4 briefly outlines the steps involved to analytically determine the overall weight of the airframe.
Figure 4. Layout of the analytical structural sizing tool.
The Loads and Sizing modules are the main components in the process. They respectively determine the
ultimate loads and load-bearing material distribution starting from the geometry description of the airframe,
maximum load factors allowed, structural concepts adopted (see Tables 1 and 2) and constraints on minimum
gage.
Nevertheless, an analysis based solely on fundamental principles will give an accurate estimate of struc-
tural weight only. Thus, weights for fuselage and lifting surfaces secondary structure (including control
surfaces and leading and trailing edges) and items from primary structure (such as doublers, cutouts and
6 of 25

Kcon Structural concept Best performance/
compromise
1 Simply stiffened shell, frames minimum weight in buckling
2 Z-stiffened shell, frames best buckling Kcon Covers Webs
3 Z-stiffened shell, frames buckling, min. gage 1 Unstiffened Truss
4 Z-stiffened shell, frames buckling, pressure 2 Unstiffened Unflanged
5 Truss-core sandwich, frames best buckling 3 Unstiffened Z-stiffened
6 Truss-core sandwich, no frames best buckling 4 Truss Truss
7 Truss-core sandwich, no frames buckling, minimum gage, 5 Truss Unflanged
pressure 6 Truss Z-stiffened
Table 1. Fuselage structural concepts Table 2. Wing-box structural concepts
fasteners) must be estimated by a correlation to existing aircraft. The initial predictions of airframe weight
by WB are overwritten with the ones coming from GUESS, supposing them as more accurate. Finally, an
iterative loop is performed until convergence on the structural weight.
IV.A. Sizing process

GUESS determines the distribution of stiffness and non-structural mass of the airframe subjected to dif-
ferent types of maneuvers such as pull-up at prescribed normal load, landing, bump on irregular runway,
and rudder maximum deflection. These maneuvers can be easily prescribed by the user depending on the
design specifications (for example maximum load factor, landing sink velocity, differential pressure inside the
fuselage). Specific lateral maneuvers according to FAR-25 criteria are considered to improve the accuracy
of vertical tail sizing, i.e. abrupt rudder maneuvers and engine-out sideslip flight. To estimate some of
the key lateral directional analysis, including stability and control derivatives for use in estimating aircraft
characteristics, the approach proposed by Mason18 is adopted.
Once the loads are defined, the sizing is performed for each station under the constraints of ultimate com-
pressive and tensile strength, local and global buckling and minimum gage.
Fuselage is idealized as slender beam with large fineness ratios, whose sizing is predominantly governed
by bending loads. Usually, typical vehicles considered feature integrally stiffened shells stabilized by ring
frames. In the buckling analysis of these structures, the shell is usually analyzed as a wide column and the
frames are sized by the Shanley criterion19 as depicted in Figure 5(a). This is the approach reported also in
the book by Niu.20 The general instability criterion is:
(E J)f L
Cf = (1)
Mu D 2
where Cf is the Shanley’s constant, (E J)f is the effective frame stiffness, D the diameter, M the ultimate
bending moment and L the frame spacing. Experimental tests show that general type of instability when Cf
is relatively low and that panel type of instability occurs when Cf is relatively high. The typical threshold
1
value is 16.000 where the occurrence of the two failure has the same probability, neglecting imperfections.
Usually, Cf is nearly constant for any given cross-section shape and it does not contain any information
regarding the sheet-stringer combination, thus allowing sizing the frame independently. In order to easily
introduce the weight of each frame, the efficiency factor Kf given by the ratio of the moment of inertia Jf
J
and the cross area of the frame Af , is introduced, i.e. Kf = Aff . This parameter depends on the chosen type
of section for the frame and clearly affects the proportion of the stiffened frame in terms of optimal frame
spacing and frame overall weight (see the book by Shanley,19 Chapter 4 for an example).
As for lifting surfaces, a multi web box beam (see Figure 5(b)) with the webs running in the direction of
structural semispan is adopted.
As outlined in Shanley,19 when the loading intensity is high, as in very thin or highly-loaded wings, the sheet
thickness may become great enough to permit the elimination of the stringers, provided that the unsupported
width of the plate is reduced by means of spanwise members. This type of construction is not only the most
efficient under such conditions, but also has many advantages such as simplicity, ease of production, surface
smoothness and increased stiffness. Nevertheless in this stage, a sensible distribution of stiffness and inertia
is required, being the details on the section of secondary importance.
The typical structural box has width B, given a section of chord-length C. The spanwise separators have
a uniform chord-spacing b and thickness tw while the sheet thickness is ts . Both thicknesses tw and ts are
7 of 25

D
Buckled frame
Bending moment Stiffener ts
L
b tw
L Axial load
Neutral axis B
Spring analogy for frames
Frame C
(a) General instability for framed fuselage (b) Multi-web concept
Figure 5. Structural concepts used by GUESS.
assumed as constant over the chord but vary spanwise. The same happens for the spacing b which varies
spanwise, usually becoming smaller when the tip is approached. Although no ribs are needed for support
of the compression surface, a minimum of two end ribs are required to maintain the airfoil contour and to
complete the torsion box. Nevertheless, ribs are indeed not taken into account in this model, despite they
may be needed where large concentrated loads are applied. Moreover, their contribution to the primary
overall structural weight is neglected. The beam is designed by spanwise bending and shear. The cover
and the webs are sized in order to prevent the critical instability mode for multi-web box beams which is
characterized by simultaneous buckling of the covers and of the webs due respectively to local instability
and flexure induced crushing.
ts
h tc
θ
bf
(a) Truss-core sandwich
ts
tw
bw
tw
bw bf
ts
ts bs bs
(b) Unflanged, integrally stiffened shell (c) Z-stiffened shell
Figure 6. Structural concepts adopted.
Tables 1 and 2 reports the different concepts of stiffened shells available whose primary components are
depicted in Figure 6. At this stage, the goal of the structures engineer is to have available a mean of evaluating
many structurally optimized cross sections together with their respective cost-producibility fabricability
values. This would greatly facilitate early selection of a minimum number of candidate configurations in
preliminary design situations, and narrow the focus of follow-on efforts to only those concepts that can
reasonably be expected to result in an optimum design.
Principles of minimum weight can then be used such that, given an applied load and the limitations
on the outside dimensions, the most efficient type of construction, its geometry and material are directly
determined. Usually, these criteria are applied to the optimum design of classic structural elements such
as columns, plates and composite structures, i.e. box beams and cylinders. The work by Gerard21 is an
historical review and assessment worth being mentioned, where a detailed and comprehensive bibliography
on optimum structural design is critically reviewed. Significant advances in the literature of those years
concerning analytical optimum design for various components, materials and thermal protection systems is
8 of 25

also given.
In order to afford a sound basis for weight prediction, the factors controlling the minimum weight, e.g.
material adopted, cross-section geometry and its proportions, must be identified and combined into formulas
or charts to be directly used as guidelines in sizing without actually designing the structure. For optimum
structural design, the required strength is determined through structural indexes. Generally speaking, global
and local buckling equations are combined for each element depicted in Figure 6, resulting in an efficiency
equation, relating the loading material index, expressing the loading intensity divided by a specified dimension
and Young modulus, to the weight index, given by the ratio of the equivalent thickness and the previously
introduced dimension, through the efficiency index. This last is usually a function of geometric proportions
(usually thickness ratios) and has to be maximized to have minimum weight. Having the dimensions of
a stress, it directly measures the loading intensity, allowing for the application of the principles of dimen-
sional similarity, shrinking and enlarging a-priori optimized proportions. Indeed, design proportions that are
optimum for a particular structure are also optimum for structure of any size, provided they all have the
same structural index. Such indexes have been developed for many basic structural components, allowing
analytical minimum weight design to be carried out for rather complex stiffened structures. Examples can
be found in Shanley19 and in the report by Crawford and Burns22 which is the basis of the work by Ardema
et al.17 and hence of the present work.
Previous considerations lead to the optimum weight, representing the lowest value achievable. This will
hardly represent the real value because of the the simplifications introduced in the modelling and secondary
issues (extra stiffeners, cut-outs, joints, nontaper) which cannot be introduced in this early design phase.
It is then necessary to determine the sources on non-optimum weight and ad-hoc formulas to consider such
weight in a rational manner. Thus, to conclude the process and determine a new prediction of structural
weight, a regression analysis is carried out. This phase enables to have an estimation of the primary and
total airframe weight starting from the load-bearing weight determined analytically by GUESS. Different
expressions for the relation between bearing and total structural weight are carried out by Ardema et al.17
for classic passenger-transport aircraft.
IV.B. Finite-element aeroelastic model

At the end of the sizing process, GUESS automatically generates a stick beam model for SMARTCAD by
means of a semi-monocoque method. Figure 7 shows some details of the typical model generated.
c/4 singularity point
Stress recovery points

Spatial coupling points
Aerodynamic panel
Three−node beam element 3/4 c collocation point
Control surface
offset
Node Lumped mass
Beam collocation points

(a) Structural model (b) Structural and aerodynamic coupled models
Figure 7. Nomenclature for the aeroelastic model.
As long as the strength requirements are fulfilled, structural flexibility is not necessarily objectionable. Aeroe-
lastic deformations, however, may not only strongly influence the structural dynamics and flight stability,
but also the overall performance and controllability of the aircraft. As introduced above, the airframe design
traditionally starts with pure strength requirements since formulating and including aeroelastic requirements
in the process is impractical.
Following this strategy, the first stiffness and inertial distribution determined by GUESS is used as a
starting point for more detailed and specific analyses during which the airframe is assessed for several flight
points and configurations by means of more refined numerical models, e.g. panel methods, finite elements.
9 of 25

Should deficiencies be detected, an optimal concurrent stress and repair process can be carried out through
the optimization process in SMARTCAD to identify and modify areas with the most beneficial impact on
stress, stiffness and aeroelastic requirements.
The main issues rely in how to transfer stiffness, inertial and aerodynamic properties. As previously men-
tioned, stiffness distribution are determined using simple analytical formulas starting from the bearing ma-
terial estimated on each section. The two meshes may have a different discretization, thus an interpolation
process of mechanical properties is carried out. This allows the designer to use a refined mesh during the
analytical sizing which is a very cheap process. The discretization of the finite-element model which is cer-
tainly more time-consuming to be analyzed, is then ruled by the analyst at his own taste. Grid nodes are
simply laid to form a cruciform shape and maintain the closed path of load transfer between lifting surfaces
and fuselage. Thus, offsets are introduced to correctly position the elastic axis of the beam connecting the
nodes which are laid so that a close connected circuit is determined, e.g. wing and vertical tail intersecting
the fuselage. Besides exporting mechanical properties and the basic stick model, extra information are given:
• stress-recovery points along the sides of the wing-box and fuselage where stresses are calculated;
• extra-nodes perpendicular to the beam axis and simply connected to beam nodes through rigid elements
(hyphotesis of beam model implying rigid section); they instantaneously enable to visualize the twist
rotational motion of the boundaries of the wing-box and are mainly be used for the coupling with the
aerodynamic model to transfer forces and displacements.
Major efforts are required to describe inertial distribution which is particularly important when the trim
condition for the free-flying aircraft is sought, since a detailed description of mass values and their location is
of primary importance to correctly define inertial loads. Thus, when the beam stick model is automatically
generated, all non-structural masses are correctly introduced in the structural mesh as:
• lumped non-structural masses on mesh nodes (engines, landing gears, auxiliary tanks, systems);
• non-structural densities per unit length along the beams (passengers in fuselage, fuel in wings, paint,
furniture).
As introduced above, these data are either provided by statistical methods or directly by the user if available.
Lumped masses are easily introduced in the model by means of rigid offsets from a reference node. Beams
can also be used for this purpose, but an estimation of their stiffness is required (this may happen in the
case of engine pylons). For example, the non-optimum structural weight estimated by the regression method
introduced above is introduced as lumped nodal masses being its chordwise position difficult to estimate in
this stage. As far as distributed masses are concerned, for example fuel, the availability of an estimate of the
fuel volume available in the wing-box allows to determine the mass stored for each beam along the wing-span,
and thus to estimate the mass per unit length. The same approach can be applied to any distributed mass,
e.g. passengers, furniture, painting and so on.
Aerodynamic mesh is represented by a a series of flat lifting boxes collecting quadrilateral panels and
defined by own span, twist, dihedral and sweep. These simple tangible parameters are particularly suitable
when it concerns aerodynamic shape optimization as showed in Perez et al.16 The aerodynamic mesh created
allows to be indistinguishably used for classic lifting surface panel methods such as Vortex Lattice,23 Vortex
Ring,24 Doublet Lattice25 and Harmonic Gradient.26 Wing-box dimensions, internal volume of fuel available
and chord-wise camber distribution are simply determined once the airfoil used at different spanwise control
sections is defined. To make this feature more flexible, a user-defined airfoil library allows to include whatever
airfoil the designer wants to adopt which make the tool particular suitable for simple shape-optimization
once an external parametrization of the airfoil is adopted.
Considering the early design phase the framework is intended for, control surfaces are currently repre-
sented by their aerodynamic contributions, neglecting their inertia, dynamics and actuation systems. When
static aeroelastic trim is sought, the analyst can specify arbitrary constraints among the control surfaces,
with different gains. For example, antisymmetric ailerons deflection, symmetric elevators deflection, or wing
flaps deflection can be imposed as needed. As for flutter analysis, for example, a further step would consist
in the structural sizing of the control surfaces, and include a lumped static impedance to model a mechanic
control-chain or to approximate the impedance of the actuators. This is currently out of the target the code
is developed for and it is left to future developments.
10 of 25

V. Aeroelasticity in conceptual aircraft design (SMARTCAD)
SMARTCAD is the numerical module dedicated to aero-structural analysis. It can be used as a stand-
alone application once the structural and aerodynamic meshes are provided, together with solver parameters.
Different kind of analysis can be carried out:
• static analysis, linear buckling.
• vibration modes calculations
• linearized flutter analysis;
• linear/non-linear static aeroelastic analysis, trimmed calculation for a free-flying rigid or deformable
aircraft;
• steady and unsteady aerodynamic analysis to extract derivatives for flight mechanics applications;
• structural optimization.
At the moment structural beam models and classic lifting aerodynamic surfaces are heavily used, despite
the code can be coupled to high fidelity aerodynamic models belonging to the class of Computational Fluid
Dynamics (CFD). Complex aeroelastic phenomena. mainly due to aerodynamic non-linearities such as the
transonic dip, can then be investigated as outlined in Cavagna et al.27
The time for the preparation of the aeroelastic model is eased by the CAD-centric concept introduced
above: a change in a geometry parameter is automatically reflected to the numerical model by means of
automatic grid-generation which allows the whole process to be re-started for another analysis on the updated
model.
In the following pages some of the components used in this work are outlined: the beam element and
aerodynamic method adopted, static aeroelastic trim solution and structural optimization. Finally, an
application of the whole process for a problem of static aeroelasticity is presented in Section VI.
V.A. Beam model

SMARTCAD adopts a three-node linear/non linear finite-volume beam, whose formulation was originally
proposed in Ghiringhelli et al.,28 which proved to be intrinsically shear-lock free. In this setion the main fea-
tures of the formulation are outlined, focusing on the linear model which be used in the application section.
The finite-volume approach leads to the collocated evaluation of internal forces and moments, as opposed
to usual variational principles which require numerical integration on a one-dimensional domain. The kine-
matic description of the generalized deformations, the strains and the curvatures, is based on an intrinsic
(kinematically exact) formulation of the beam.
As sketched in Fig. 8, each beam element is divided in three parts. Each part is related to a reference
point Gi : the mid- and the two endpoints. They are referred to geometrical nodes Ni by means of offsets si .
This allows the elastic axis of the beam to be offset from the center of mass.
Every node is characterized by a position vector xi and a rotation matrix R(g) through Gibbs-Rodrigues
rotation parameters g. A reference line p describes the position of an arbitrary point p(ǫi ) on the beam
section.
Independent parabolic shape functions are used to interpolate displacements and rotation parameters of
the generic point p(xi ) as functions of those of the reference nodes. The derivatives of the displacements and
the rotation parameters at the two collocation points Cj are used to evaluate the strains ε and the elastic
curvatures k which are then used to compute the internal forces√and moments, balancing the external forces
ti and moments mi . Collocation points are laid at ǫi = ± 1/ 3 to recover the exact static solution for a
beam loaded at the end points, as showed in Masarati and Mantegazza.29
Being the position of the ith node pi = xi + Rsi , the position of an arbitrary point of the reference line is
interpolated by means of parabolic shape functions N(ǫ) = 1/2ǫ(ǫ − 1) 1 − ǫ2 1/2ǫ(ǫ + 1) as:

p(ǫ) = Ni (ǫ) (xi + Ri (g(ǫ)) si ) with g(ǫ) = Ni gi (2)
11 of 25

mII ǫ
tII G3
s3
CII N3
mI G2
tI s2
CI N2
p
G1
s1
N1
Figure 8. Finite Volume three-node beam coupled to a classic lifting surface method.
Equilibrium for each node, considering external nodal forces and internal forces at each collocation point,
leads to AΘ = F with:
   
−I 0 0 0  


 f1 


 −(pI − p1 )× −I 0 0  f m + s f
   
1 1× 1
 I  
 

  
 
 
 

I 0 −I 0 m f
   
I 2
 
A=   , Θ = , F = (3)
 (pI − p1 )× I (pII − p2 )× −I   

 fII 



 m2 + s2× f2  

0 0 −I 0  mII f3
    
 

 
 

 
0 0 (pII − p3 )× I m3 + s3× f3
 
The matrix A is referred as moment arm matrix and depends on the current deformed shape, Θ is the
vector of internal generalized forces and F is the vector of external generalized nodal forces. In the previous
expression, the term ( · )× represent the vector product matrix, i.e. if a and b are two vectors, a× is the
matrix that multiplied by b gives a × b, and I is the identity matrix.
Internal generalized forces Θi at each collocation point are recovered from the local reference material frame
as: ( ) " #( )
fi Ri 0 f̃i
= with i = I, II (4)
mi 0 Ri m̃i
with Ri = Ri (gi ) function of the rotation parameters interpolated from the three nodes of the element. and
(˜· ) indicating that an entity is referred to the material frame. The previous expression can be extended to
both collocation points as:
Θ = RΘ̃ (5)
with Θ = (f I , mI , f II , mII ), R = diag (RI , RI , RII , RII ).
Internal forces are related to generalized strains and curvatures through an arbitrary sectional stiffness matrix
at the collocation point:
( ) " # ( )
f̃i D̃εε D̃εk ε̃i
= with i = I, II (6)
m̃i D̃kε D̃kk k̃i
i
summarized as:
Θ̃ = D̃ Ψ̃ (7)

with D̃ = diag D̃I , D̃II , , Ψ̃ = ε̃I , k̃I , ε̃II , k̃II .
12 of 25

The generalized strains are defined as the difference between the current and the initial derivatives of the
reference line p(ǫ) that describes the position of the beam:
( ) ( )
ε̃ RT p′ − p̃′0
= (8)
k̃ ρ̃ − ρ˜0
where p̃′0 is the initial configuration of the reference line and ρ̃0 is the initial geometric curvature of the
undeformed beam. The geometric curvature ρ̃ of the beam reference line is defined as the axial derivative
of the reference frame of the beam section:
ρ̃× = RT R′ (9)
Considering the linear case which will be applied in Section VI, strains and curvatures are linearized as:
ε = p′ + p′0× ϕ = N′i xi + p′0× Ni ϕ − N′i si× ϕi (10)

k = ϕ′ = N′i ϕi (11)
Substituting back the previous expression, the stiffness matrix is recovered:

 
NIi p′0× NIi ϕ − N′Ii si× ϕi ( )
0 N′Ii xi
 
T 
A R D̃ R  =F (12)

p′0× NIIi ϕ − N′IIi si× ϕi  ϕi

 NIIi
0 N′IIi
All the other matrices are evaluated in the reference configuration and do not participate to the linearization.
The formulation leads to a loss of symmetry of the stiffness matrix. This is not a major issue considering
that:
• usually the models are quite small and the computational cost is very limited; thus the adoption of
algorithms specifically suited for symmetric problems are not of primary importance despite algorithms
working with sparse matrices are recommended;
• when dealing with linearized aeroelastic problems the symmetry is always lost since the resulting
aerodynamic matrices summing up to structural terms are never symmetric (see Section V.C).
Finally lumped mass matrix is created for each of the three sectors along the beam.
V.B. Low-fidelity aerodynamic models

Two low-fidelity aerodynamic methods are available in SMARTCAD, depending on whether steady or un-
steady analysis is carried out:
• Vortex Lattice Method (VLM)23 with camber contribution on normalwash once the airfoil description
is provided;
• Doublet Lattice Method (DLM)25 for the prediction of the generalized forces due to harmonic motion
in the subsonic regime.
Both methods are based on potential flow theory leading, under the hypothesis of irrotational, isoentropic and
inviscid flow to a Laplace’s equation respectively for the linearized velocity or acceleration potential. They
are certainly the most common tools adopted in aerodynamic loads prediction in the conceptual/preliminary
phase (see Kier30 for some examples in flight-mechanics applications).
By virtue of the model-linearity allowing superimposition, approximate solutions are obtained by ide-
alizing the surface as a set of interfering elementary singularities, e.g. vortex lines and doublets, each of
them satisfying Laplace’s equation. Because the loading is approximated by discrete lifting elements, the
method permits straightforward analyses of nonplanar and interfering configurations. The same geometric
discretization of the aerodynamic surfaces as a flat lifting surface with singularities and collocation points
located respectively at 41 and 43 of each panel chord so that Kutta condition is satisfied a priori. Thus, the
same mesh is used (see Fig. 7(b)), with the only exception that the VLM takes into account camber contri-
butions through boundary conditions and requires trailing vortexes for wake modelling. Figure 9 shows two
13 of 25

0.2
0
Updated trailing vortexes
−0.2
Fuselage −0.4
−0.6
−0.8
Updated VLM mesh
Deformed structrural mesh

−1
Undeformed structural mesh
Undeformed VLM mesh
(a) Vortex Lattice (b) Clamped wing, lattice updated
Figure 9. Application of the Vortex lattice method.
example concerning the VLM; a rigid case and a case of non-linear trim where the whole lattice is updated
and combined to the non-linear beam model to consider geometric non-linear effects.
Finally, an interface scheme to exchange displacements and loads between the structural grid and the
aerodynamic boundary surface is required. In this specific case, the beam model represent a one-dimensional
domain where the real structural geometry is hidden. On the other hand, the aerodynamic lifting surface is
reduced to a two-dimensional domain. As a consequence, two radically different representations of the same
aircraft geometry must be made compatible in order to transfer information between them. This is a well
known problem, deeply investigated in the literature; for further reference, see Smith et al.31, 32 Two class of
methods are available for this purpose within SMARTCAD: an innovative scheme, based on Moving Least
Square (MLS) method (see Quaranta et al.33 ) and the Radial Basis Function (RBF) method (see Beckert and
Wendland,34 Schaback and Wendland35 ). Both methods ensure the conservation energy transfer between
the fluid and the structure and they are suitable for the treatment of complex configurations. To avoid
interpolating rotations and using the same algorithms in a straightforward way also for CFD meshes, extra
points are added along the wingbox through rigid-arms. Aerodynamic loads are distributed along both
master beam nodes and these additional nodes. Loads are then reduced to lumped forces and moments on
the former set of nodes.
V.C. Static linear aeroelasticity

Considering a generic point can undergo a motion given by the sum of rigid and elastic deformable shape
functions, i.e. u = ΦR qR + ΦE qE , the general equation governing the dynamic of the flying deformable
aircraft is:
Mq̈ + Cq̇ + Kq = Q0 − (Mc q̈c + Cc q̇c + Kc qc ) (13)
where M, C, K are respectively the mass, damping and stiffness matrix of the overall system, q the general-
ized coordinates given by rigid and deformable contributions, i.e. q = (qR , qE ), Q0 the vector of generalized
external forces including reference aerodynamic load and qc are the generalized coordinates for controls
deflections, whose effects are introduced through the generalized matrices with subscript c.
The common approach in aeroelasticity considers these equations in the reference Cartesian frame with a
selected origin in a point O; rigid dynamic is simply given by three translations and rotations along its axes.
In this work, the elastic coordinates are directly represented by nodal generalized displacements uE , i.e. ΦE
is the identity matrix and qE = uE . Linearized rigid body modes ΦR enable to recover displacements and
rotations for every node from the motion of the reference point representing the gross motion of the aircraft,
as:
uRi = ΦRi qR = I (xP − x0 )× qR (14)
with xP − x0 the distance between the generic ith node and the refence point O.
14 of 25

Adopting a quasi-steady approximation for aerodynamic forces and neglecting control surface mechanical
stiffness and damping, the system of equations governing the overall motion of the aircraft and its structural
response, is:
" #( ) " # " # !( )
MRR MRE q̈R 06×6 06×NE q∞ La CRR CRE q̇R
+ − +
MER MEE q̈E 0NE ×6 CEE V∞ CER CEE q̇E
S A
" # " # !( ) ( ) " #
06×6 06×NE KRR KRE qR QR0 MRc
− q∞ = − q̈c +
0NE ×6 KEE KER KEE qE QE0 MEc
S A
" # " #
q∞ La CRc KRc
q̇c + q∞ qc (15)
V∞ CEc KEc
A A
being La the reference aerodynamic chord and q∞ and V∞ respectively the flight dynamic pressure and
velocity. Structural terms are indicated with ( · )S subscript while aerodynamic influence matrices with
( · )A . The inertial coupling between qE and qR through MER and MRE is highlighted; MRR is the rigid
body mass matrix with respect to point O, while MEE is the mass matrix associated to elastic coordinates.
Six reference conditions have to be defined in order to correctly couple the reference motion and the structural
motion referred to the former (see Canavin and Likins36 for further information). Considering a mean axes
formulation, it is possible to nullify such inertial coupling, imposing the orthogonality of the structural
displacement with the rigid modes through the mass matrix MEE of the model:
ΦTR MEE ΦE qE = 0 (16)
Neglecting contributions due to q̈E , q̇E and control derivatives (the definition of static aeroelasticity), it
is possible to determine the deformed shape from the second system of equations, representing structural
steady equilibrium:

q∞ La
qE = K−1AE C ERA q̇R + q∞ K ERA qR + Q E0 + q∞ K Ec q
A c (17)
V∞
with KAE = KEE S − q∞ KEEA . Finally the equation governing the gross motion of the deformable aircraft
is:
q∞ La
CRRA + q∞ KREA K−1 −1

MRR q̈R = AE CERA q̇R + q∞ KRRA + q∞ KREA KAE KERA qR
V∞
+ q∞ KREA K−1 −1

q∞ KRcA AE KEcA qc + QR0 + q∞ KREA KAE KERA QE0 (18)
from which it is possible to see extra-aerodynamic terms sum to the ones of the rigid aircraft, referred as
corrections to aerodynamic derivatives and the intercept is affected by structural deformability.
The final equation including aeroelastic correction in the ( ˜· ) terms is:
q∞ La
MRR q̈R = C̃RRA q̇R + q∞ K̃RRA qR + q∞ K̃RcA qc + Q̃R0
V∞
which in the most general case leads to six equations of equilibrium from which six unknowns can be
determined once the remaining are defined in order to specify the maneuver to perform.
V.D. Optimization method

The optimal solution is not necessarily sought for. The minimum weight structural concepts are adopted
and geometric constraints on minimum gage are imposed to avoid unfeasible optimal solutions. This fact
together with stiffness and aeroelastic constraint does not necessarily lead to the global optimum solution.
Nevertheless the first structural sizing on minimum weight principles according to strength criteria on ulti-
mate loads is considered to be a good starting point for stiffness and inertia distribution. Thus, a simple
optimization process based on Gradient-based Optimization Methods can be adopted to slightly modify the
airframe so that aeroelastic requirements are satisfied as well (see Figure 10).
Computational design techniques are based on numerical analysis methods evaluating the merit, com-
monly referred as objective function, of a set of feasible designs. Traditionally, the process of design using
15 of 25

Figure 10. Optimization process within SMARTCAD.
numerical simulation has been carried out by trial and error, mainly relying on the experience and intuition
of the designer in selecting the combination of the design variables. When the number of design variables
increases, making correct choice may become very hard, thus to efficiently span the design space of large
dimension, numerical simulations need to be combined with automatic procedures.
The classic statement for the problem of constrained optimization reads:
minimize : I(dj )
with respect to : dj , j = 1, 2, . . . , Nd
subject to : gm (dj ) ≥ 0, m = 1, 2, . . . , Ng
where I is a non-linear function of the design variables dj and gm are the non-linear constraints to be satisfied.
The parameters are allowed to vary in a limited design space to guarantee a feasible and realistic solution
such that the design features the best figure of merit. Optimization algorithms are supposed to perform
such task in a rigorous way. Many methods are available and the literature on this topic is large. A good
textbook to refer is for example the one by Hatfka.37
A well known category is the gradient-based one which uses the value of the objective function and its
gradients with respect to the design variables. The sensitivity information is used to determine iteratively
the design variables d:
dn+1 = dn + αn Sn (19)
where n if the iteration number and αn is the step length along the search direction S which assures the
furthest reduction of the objective function. In the present work, the direction is defined determining the
gradients and sensitivities of constraints numerically. During the optimization process, the objective and
the constraints must be repeatedly evaluated, raising the overall computational cost since each evaluation
demands for a numerical simulation. A common approach consists in creating an approximated linearized
problem and feed the optimizer with such simplified model. This results in a very relatively inexpensive
process.
Gradient-based methods converge to the optimum with significantly fewer functional evaluations than
the zeroth order methods like the evolutionary ones. The disadvantage is that these methods work well with
smooth objective functions, convergence to local minima is guaranteed but the optimal design could depend
on the initial point. Zeroth order methods strictly rely only on the value of the objective function. When the
number of design variables increases, many evaluations are required resulting in an extremely computation
cost. On the other hand, in problems with a limited number of design variables with multiple local minima
or discontinuities, zeroth order methods are more suitable.
To allow efficient and fast optimizations, the approach adopted here directly relyes on simplified aero-
dynamic and structural models, i.e. combination of beam models and lifting surfaces methods for static
aeroelastic analysis. A similar approach is adopted in Osterheld et al.38 where beam and plate models are
combined with source/doublet panel methods to carry out the first structural sizing. Due to limited amount
of detail within global aircraft models, the strength and buckling analysis cannot be performed based purely
on finite elements methods. Thus, semi-analytical procedures are usually adopted and included in the opti-
mization process as pursued in the present work. The same efficiency equation for minimal weight sections
16 of 25

are included as buckling constraints, which, generally speaking, depends by a combination of skin thickness
and web thickness and spacing.
Requirements definition and quantification is for sure one of the most important phases allowing the
designer to drive the optimization process and the final result. Interactions among different aeroelastic
requirements exist and are not obvious to quantify. Also, since the design process is performed with a few
dominating load cases, there is a risk of not meeting the design criteria for the complete set of design driving
load cases. Engineering experience is extremely valuable in such task.
VI. Application to the Transonic Cruiser TCR aircraft

The whole process developed is applied to the target aircraft within the SimSAC project: the TranSonic
Cruiser (TCR) whose main design specifications are summarized in Table 3. Figure 11 shows the CAD
model coming from the parametric geometry module CADac. In particular, Figure 12 shows respectively
the wing and the vertical fin, while Figure 13 the fuselage from two different perspectives. As it can be seen,
the parametes used in the geometry module can describe successfully its complex geometry, especially when
it concerns the fuselage fore and aft extremes.
Cruise Mach 0.97 at altitude≥ 37.000 ft
Range 5.500 nm +
250 nm to alternate airport +
0.5 hour loiter at 1.500 ft
Max payload 22.000 Kg
Passengers 200
Crew 2 pilots, 6 cabin attendants
Take-off distance 2.700 m at max WT O
at h=2.000 ft
Landing distance 2.000 m at max WL
at h=2.000 ft max payload, normal reserves
Powerplant 2 turbofans
Certification JAR25
Maneuvering
load factors 2.5, -1
Max
load factors 3.1, -1.7
Figure 11. CAD model for the TCR Table 3. Design specifications for TCR
(a) Wing (b) Vertical fin

Figure 12. Wing and vertical tail CAD representation.
VI.A. Weight estimates

The first estimates determined through the WB module gives a maximum empty empty weight MMEW of
128.034 Kg and a maximum take off weight MT OW of 220.034 Kg (considering a total fuel weight of 92.000
Kg distributed respectively for 75% in wings and 25% in the fuselage central box). The longitudinal position
17 of 25

(a) Front view (b) Rear view
Figure 13. Fuselage CAD representation.
of the center of gravity with respect to the fuselage nose for the first case is XMEW = [36.562, 0, −0.033]
m, while for the second case is: XMT OW = [36.389, 0, −0.30] m. The principal moments of inertia are
Jp = [Jxx , Jyy , Jzz ] = [9.768, 20.407, 28.242] · 106 Kg/m2 .
The weight of the main components are briefly summarized in Table 4 where comparisons with a similar
approach by SAAB are reported. Fairly good agreement is found by both methods despite the MT OW is
slightly higher, the wing is slightly lighter and the tail heavier.
Component Mass WB Mass SAAB

[kg] [kg]
Wing 24.904 21.534
Horizontal tail 1.033 1.929
Vertical tail 2.890 3.678
Landing gear 7.227 8.102
Fuselage 20.931 21.353
Powerplant 35.545 31.127
Total MT OW 220.034 226.764
Table 4. TCR weight comparison with SAAB data.
VI.B. First structural sizing

The GUESS module determines the first structural sizing and generates the aeroelastic model for further
structural and aeroelastic analyses. One of the main issues relies in faithfully distributing payload and non-
structural masses. Representing concentrated items like engines, auxiliary tanks and landing gears by means
of lumped masses rigidly attached to the main structure is straightforward. Masses are simply introduced
by means of rigid arms connected to the nodes of the mesh. Figure 14 shows the stick model and the
aerodynamic mesh generated. The black dots represent the lumped contributions for landing gears, engines
and secondary structural masses coming from the regression analysis.
On the other hand, introducing distributed masses estimated by the WB module, e.g. fuel in the wing-
box, paint, passengers, furniture weight, is more complicated because the overall weight of each item has
to be properly distributed. Figure 15(a) shows the distribution of such masses along fuselage axis, whose
components are smeared over a specific portion of the total fuselage length on the basis of the position of
their center of gravity:
• interior, over fuselage length in proportion to beam volume;
• furniture, over fuselage length in proportion to beam volume;
18 of 25

Figure 14. Aeroelastic model for the TCR aircraft.
• baggage, over passenger compartment length in proportion to beam volume;

• crew, over passenger compartment length in proportion to beam volume;
• passengers, over passenger compartment length in proportion to beam volume;
• pilots, over cockpit length in proportion to beam volume;
• paint, over fuselage length in proportion to beam wetted area.
Non-structural mass for all lifting surfaces, wing and tails, include paint weight which is distributed in
proportion to the beam wetted area. No additional masses are considered for the tail planes.
For the main lifting surface, wing fuel tanks and central fuel tank are accounted for, recovering the estimations
performed by WB module:
• fuel stored in wing tanks is distributed in proportion to beam volume over the cantilever wing, from
wing-fuselage intersection to wing tip;
• fuel stored in central tank is uniformly distributed over the wing carrythrough structure.
Figure 15(b) represents the distribution of non-structural masses over the structural beams for the semiwing,
from body center-line to wing tip.
600
Interior 7000 Fuel tanks
Furniture
Paint
Baggage
500 Crew 6000 Central fuel tank
Passengers
Mass density [kg/m]
Mass density [kg/m]
Pilots
Paint 5000
400
4000
300
3000
200 2000
1000
100
0
0
0 10 20 30 40 50 60 0 5 10 15 20 25 30 35
Spanwise distance y [m] Spanwise distance y [m]
(a) Fuselage (b) Semi-wing
Figure 15. Distribution of mass linear densities for non-structural components.
The estimated weights for the fuselage are briefly summarized in Table 5. All the structural concepts available
are investigated together with different values for the efficiency factor Kf concurring to the bending stiffness
of the frames (only for the concepts between 1 and 5). Also, a pressure gradient ∆P = 13.65 psi is considered
19 of 25

to show the benefits in stress relieving. The concepts 6 and 7 lead to a lower structural weight, thanks to
their better buckling performance. For the concepts between 1 and 5, it is possible to note the weight
estimates are fairly close, thus showing a weak dependence on the structural concept adopted. However, the
discrepancy with the last two concepts can be reduced considering better performing frame section, raising
the frame efficiency Kf . Finally, the values determined also have a good correspondence with the statistical
prediction by WB when poor frame sections are considered (Kf = 1 or 2).
Kcon Kf = 5.24 Kf = 2 Kf = 1 Kf = 2, ∆p Kf = 5.24 Kf = 2 Kf = 1

[Kg] [Kg] [Kg] [Kg] [Kg] [Kg] [Kg]
1 5.937 6.659 7.277 16.555 15.173 17.106 18.690
2 5.356 5.924 6.241 14.761 13.692 15.215 16.494
3 5.622 6.283 6.850 15.614 14.372 16.138 17.596
4 5.671 6.318 6.888 15.750 14.497 16.229 17.693
5 6.332 7.055 7.675 17.648 16.214 18.124 19.714
6 5.207 - - 13.274 13.376 - -
7 5.139 - - 12.854 13.200 - -
Table 5. Fuselage load bearing (left) and total material (right) estimated for different parameters.
The structural weights for the main wing are reported in Table 6. Again, all the available structural
concepts are used. It is possible to note the first three concepts featuring a simple unstiffened skin, lead to
high structural weights if compared to the last three concepts. This is basically due to the buckling critical
loads. Indeed, the last three concepts feature a better behavior in this respect since the cover is stiffened,
allowing for a higher critical load. This results in a section with thinner cover (which has the major role in
contributing to the weight of bearing material), thicker webs but with a higher spacing. Again, the results
for the last three concepts agree fairly well with the statistical prediction by WB.
Kcon Bear. material Tot. weight Kcon VT Tot. weight HT Tot. weight
[Kg] [kg] [Kg] [kg]
1 27.610 44.280 1 7.375 2.675
2 26.670 40.470 2 6.337 2.498
3 25.710 47.170 3 8.119 3.013
4 17.540 25.820 4 2.454 1.632
5 17.240 24.980 5 2.188 1.593
6 16.460 25.920 6 2.618 1.568
Table 6. GUESS prediction for TCR wing. Table 7. GUESS prediction for TCR tail.
The same trend is found when it concerns tail empennages. The stiffened cover concepts outperforms the
unstiffened ones, as reported in Table 7. Again, good agreement with statistical methods is determined.
Finally, three combinations of fuselage and lifting surfaces are showed in Table 8, ranging from the two
extrema of lightest and heaviest solution. They are simply determined considering the best and worst case
for each component. Moreover, an intermediate solution is also reported to prove for this case the result is
very close to the minimum value determined. The longitudinal position of the center of gravity is reported,
together with the principal moments of inertia. Again, acceptable agreement with the statistical methods
by WB is found.
VI.C. Structural analysis and optimization

In this section a simple example of optimization is considered starting from the first sizing provided by
GUESS. One of the heaviest predicted cases is analyzed, considering a model with the following features:
• 11 beams for the fuselage (Kcon = 4), with three design variables each: cover thickness tsF , frame
spacing dF and frame area AF ;
20 of 25

Kcon Kcon MTOW XCG Jxx · 106 Jyy · 106 Jzz · 106
fuselage lifting surfaces [Kg] [m] Kg/m2 Kg/m2 Kg/m2
7 6 211.420 37.215 7.827 18.962 25.447
4 4 212.143 37.156 7.798 19.144 25.570
5 1 240.024 37.899 9.482 23.240 30.820
Table 8. TCR MT OW for different combinations of structural concepts.
• 32 beams for the wing (Kcon = 1), with three design variables each: cover thickness tsW , web spacing
dwW and web thickness twW ;
• 12 beams for the horizontal tail (Kcon = 1), with three design variables each: cover thickness tsHT ,
web spacing dwHT and web thickness twHT ;
• 8 beams for the vertical tail (Kcon = 2), with three design variables each: cover thickness tsV T , web
spacing dwV T and web thickness twV T .
Considering the symmetry of the model, the total number of design variables is 123. Three simple frozen
maneuvers at altitude 5.000 m and M∞ = 0.5 are considered:
• steady pull-up at load factor nz = 2.5;
• cruise condition with a sideslip angle of 20 deg;
• snap-roll maneuver from a cruise condition with abrupt rudder and elevator deflection of 20 deg.
These maneuvers are chosen with the objective of bringing each component, i.e. wing, fuselage and tail
empennages, to the limit case for bending and torsional loads.
For each beam of the model, the following constraints are considered for all the maneuvers:
• positive safety margin for normal and shear stresses;
• positive safety margin for fuselage buckling using Shanley’s method;
• positive safety margin for lifting using their efficiency equation which combines cover thickness, web
spacing and thickness;
resulting in 567 constraints.
1.05 8
1.04
7
1.03
6
1.02
Objective function
5
Design variable
1.01
1 4
0.99
3
0.98
2
0.97
1
0.96
0.95 0
0 10 20 30 40 50 60 70 80 90 100 10 20 30 40 50 60 70 80 90
Iteration Iteration
(a) Objective function (b) Design variables
Figure 16. Optimization history (all quantities normalized respect to nominal value).
The guess solution is unfeasible; due to the simplifications adopted in performing the first sizing some sections
indeed result in a slightly negative safety margin.
21 of 25

Furthermore, design variables are have an inferior limit to determine a real feasible solutions, i.e. at least
two webs for each section and minimum gage thicknesses allows for covers and webs. The objective function
is rapresented by the total weight which has to be minimized.
The optimization process leads to a decrement of the total bearing material weight of approximately
3.2% of WMT OW , starting from a solution of WMT OW = 236.608 Kg.
Figure 16(a) shows the history of the objective function while Figure 16(b) gives a qualitative trend of the
design variables. It can be seen the solution approaches the final value after approximately 40 iterations. All
the constraints are satisfied, i.e. each section has a positive safety margin respect to yielding and buckling.
An overall lightening is determined, adjusting at the same time the critical sections.
Figure 17 shows the deformed shape for the steady pull-up before and after the optimization process. The
displacements at the wing-tip are higher because of the lightening, while the downward displacement for the
horizontal tail is reduced thanks to a localized stiffening for the backward sections of the fuselage and the
vertical-tail.
(a) Reference case (b) After optimization

Figure 17. Deformed shapes for a steady pull-up, nz = 2.5 before and after the optimization.
0.08 0.07
Guess solution Guess solution
After Optim. After Optim.
0.07 0.06
0.06
0.05
0.05
0.04
y
t
0.04
J
0.03
0.03
0.02
0.02
0.01
0.01
0 0
1 2 3 4 5 6 7 8 9 10 11 1 2 3 4 5 6 7 8 9 10 11
Beam index Beam index
(a) Torsional constant (b) Moment of inertia (circular section)

Figure 18. Fuselage updated solution.
Indeed, Figure 18 shows the differences between the torsional constant and moment of inertia of the new
solution. The sections from the carrythrough up to the intersection with the vertical tail are stiffened,
resulting in raise of the aforementioned properties. The maximum peak is shifted back with respect to the
reference solution simply because the GUESS module uses the aerodynamic quarter-chords as reference axis,
while the stick model is correctly laid along the elastic axis.
Figure 19 outlines the new distribution of stiffness for the vertical tail. The torsional constant and out-of-
plane moment of inertia are governed by the asymmetric maneuvers. The downward displacements of the
horizontal tail are lowered consequently to a raise in the in-plane bending stiffness which is, on the other
hand, governed by the symmetric pull-up.
Figure 20 outlines the new distribution for the wing and the carrythrough (first three beams). For this
22 of 25

0.05 0.14 0.014
Guess solution Guess solution Guess solution
0.045 After Optim. After Optim. After Optim.
0.12 0.012
0.04
0.035 0.1 0.01
0.03
0.08 0.008
Jt
0.025
Iz
Iy
0.06 0.006
0.02
0.015 0.04 0.004
0.01
0.02 0.002
0.005
0 0 0
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8
Beam index Beam index Beam index
(a) Torsional constant (b) In plane moment of inertia (c) Out of plane moment of inertia
Figure 19. Vertical tail updated solution.
component the overall stiffness is reduced, which explains the overall weight saving achieved. The same
happens for the horizontal tail where minor changes take place.
0.7 1.4 0.2

Guess solution Guess solution Guess solution
After Optim. After Optim. 0.18 After Optim.
0.6 1.2
0.16
0.5 1 0.14
0.12
0.4 0.8
Jt
0.1
y
z
I
I
0.3 0.6
0.08
0.2 0.4 0.06
0.04
0.1 0.2
0.02
0 0 0
0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16 0 2 4 6 8 10 12 14 16
Beam index Beam index Beam index
(a) Torsional constant (b) In plane moment of inertia (c) Out of plane moment of inertia
Figure 20. Wing updated solution.
The changes in some of the aerodynamic derivatives are briefly reported in Table 9. In particular, there
is a reduction in the derivative associated to the aileron on the roll moment due to the lightening of the wing
and the associated reduction in the torsional stiffness. On the other hand, due to a torsional stiffening of
the vertical tail, the side aerodynamic derivative associated to the rudder is increased.
Rigid Reference After opt.

Cz /α 3.761 3.774 3.740
Cm /α -7.214 -7.160 -7.057
Cl /δaileron 0.079 0.059 0.049
Cy /δrudder 0.158 0.079 0.101
Cl /p -0.304 -0.297 -0.286
Table 9. Stability derivatives, M∞ = 0.5, z = 5000 m.
In order to raise the aforementioned stability derivatives to the 75% of the rigid value, a second optimization
is carried out, adding two extra constraints on Cl /δaileron and Cy /δrudder . In this case, the overall weight
saving is raised respect to the previous case, resulting in an overall weight reduction of 2.62% of the WMT OW .
As expected, Figure 21 shows the torsional constants for the wing and the vertical tail are incremented to
comply with the new constraints.
VII. Conclusion
A procedure for aero-structural conceptual design has been presented. It allows to design the airframe
once the aerodynamic shape is defined relying on numerical methods and reducing to the minimum the
adoption of statistics typical of this design phase. All the steps to carry out these kind of simulation have
23 of 25

0.35 0.35
#1 #1
#2 #2
0.3 0.3
0.25 0.25
0.2 0.2
t
t
J
J
0.15 0.15
0.1 0.1
0.05 0.05
0 0
0 2 4 6 8 10 12 14 16 1 2 3 4 5 6 7 8
Beam index Beam index
(a) Wing (b) Vertical tail

Figure 21. Updated torsional constants with constraint on aerodynamic derivatives.
been presented together with an overview of the methods adopted for linear static aeroelasticity.
Future works will also enhance the optimization process with flutter constraint in terms of minimum damping
allowed.
Acknowledgments
The financial support by the European Commission through co-funding of the FP6 project SimSAC is
acknowledged.
References
1 Raymer, D. P., Aircraft Design: A Conceptual Approach, Fourth Edition, AIAA Education Series, New York, NY, 2006.
2 Torenbeek, E., Synthesis of Subsonic Airplane Design, Kluwer Academic Pub, Delft University Press, 1982.
3 Lucia, D. J., “The SensorCraft Configurations: A Non-Linear AeroServoElastic Challenge for Aviation,” Proceedings of
the 46th AIAA/ASME/ASCE/AHS/ASC SDM Conference, Austin, Texas, 18 - 21 April 2005.

4 Silva, W. A., Vartio, E., Shimko, A., Kvaternik, R. G. and, E. K. W., and Scott, R., “Development of Aeroservoelastic
Analytical Models and Gust Load Alleviation Control Laws of a SensorCraft Wind-Tunnel Model Using Measured Data,”
Proceedings of the IFASD International Forum on Aeroelasticity, Stockholm, June 17 - 20 2007.
5 Liebeck, R. H., “Design of the BlendedWing Body Subsonic Transport,” Journal of Aircraft, Vol. 41, No. 1, January-
February 2004, pp. 10–25.

6 Antoine, N., Kroo, I., Willcox, K., and Barter, G., “A Framework for Aircraft Conceptual Design and Environmental
Performance Studies,” Proceedings of the 10th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, Albany,
New York, 30 August - 1 September 2004 2004.
7 Kroo, I. and Shevell, R., “Aircraft Design: Synthesis and Analysis,” Tech. rep.,
http://adg.stanford.edu/aa241/AircraftDesign.html, September 2006.
8 Ning, S. A. and Kroo, I., “Tip Extensions, Winglets, and C-wings: Optimization in Conceptual Wing Design,” Tech.
rep., http: //aero.stanford.edu/adgreports.html, August 2008.

9 Macci, S. H., “Semi-analytical method for predicting wing structural mass,” Tech. rep., SAWE Paper No.2282, May 1995.
10 Bindolino, G., Ghiringhelli, G., and Ricci, S., “Preliminary Sizing Of The Wing-Box Structure By Multi-Level Approach,”
Int. Forum on Aeroelasticity and Structural Dynamics IFASD 2005, - Munich, Germany, June 28 - July 01 2005.
11 Isikveren, A., Quasi-analytical modelling and optimization techniques for transport aircraft design, Ph.D. thesis, Royal
Institute of Technology (KTH), 2002.

12 SimSAC, “http://www.simsacdesign.eu/,” last accessed 2009.
13
Von Kaenel, R., Rizzi, A., Grabowski, T., Ghoreyshi, M., Cavagna, L., Ricci, S., and Bérard, A., “Bring-
ing Adaptive-Fidelity CFD to Aircraft Conceptual Design: CEASIOM.” Proceedings of the 26th ICAS Congress,
Anchorage, Alaska, 14–19 September 2008.
14
Bérard, A., Rizzi, A., and Isikveren, A., “CADac: A New Geometry Construction Tool for Aerospace Vehicle
Pre-Design and Conceptual Design,” Proceedings of the 26th AIAA Applied Aerodynamics Conference, Honolulu,
Hawaii, 18–21 August 2008.
15
Mitchell, C., “A Computer Programme to Predict the Stability and Control Characteristics of Subsonic Air-
24 of 25

craft,” Report TR 73079, Royal Aircraft Establishment, Procurement Executive, Ministry of Defence, September
1973.
16
Perez, R. E., Jansen, P. W., and Martins, J. R. R. A., “Aero-Structural Optimization of non-planar lifting sur-
face configurations,” Proceedings of the 12th AIAA-ISSMO Multidisciplinary Analysis and Optimization Conference,
Victoria, British Columbia, 10–12 September 2008.
17
Ardema, M., Chambers, A., Hahn, A., Miura, H., and Moore, M., “Analytical Fuselage and Wing Weight
Estimation of Transport Aircraft,” Tech. Rep. 110392, NASA, Ames Research Center, Moffett Field, California, May
1996.
18
Mason, W. H., “http://www.aoe.vt.edu/~mason/Mason f/MRsoft.html/,” LDSTAB software, last accessed
2009.
19
Shanley, F. R., Weight-strength analysis of aircraft structures, Dover Publications, Inc., New York, 2nd ed.,
1960.
20
Niu, M. C., Airframe Structural Design, Hong Kong Conmilit Press Ltd, Honk Kong, 1st ed., 1988.
21
Gerard, G., “Optimum Structural Design Concepts for Aerospace Vehicles,” Journal of Spacecraft and Rockets,
Vol. 3, No. 1, January 1966, pp. 5–18.
22
Crawford, R. F. and Burns, A. B., “Minimum weight potentials for stiffened plates and shells,” AIAA Journal ,
Vol. 1, No. 4, 1963, pp. 879–886.
23
Moran, J., An Introduction to Theoretical and Computational Aerodynamics, John Wiley & Sons, 1984.
24
Katz, J. and Plotkin, A., Low-Speed Aerodynamics From Wing Theory to Panel Methods, McGraw-Hill series
in Aeronautical and Aerospace Engineering, 1991.
25
Albano, E. and Rodden, W. P., “A Doublet–Lattice Method for Calculating the Lift Distributions on Oscillating
Surfaces in Subsonic Flow,” AIAA Journal , Vol. 7, No. 2, 1969, pp. 279–285.
26
Chen, P. C. and Liu, D. D., “A Harmonic Gradient Method for Unsteady Supersonic Flow Calculations,”
Journal of Aircraft, Vol. 22, No. 15, 1985, pp. 371–379.
27
Cavagna, L., Quaranta, G., and Mantegazza, P., “Application of Navier-Stokes simulations for aeroelastic
assessment in transonic regime,” Computers & Structures, Vol. 85, No. 11-14, 2007, pp. 818–832, Fourth MIT
Conference on Computational Fluid and Solid Mechanics.
28
Ghiringhelli, G. L., Masarati, P., and Mantegazza, P., “A Multi-Body Implementation of Finite Volume Beams,”
AIAA Journal , Vol. 38, No. 1, January 2000, pp. 131–138.
29
Masarati, P. and Mantegazza, P., “On the C 0 Discretisation of Beams by Finite Elements and Finite Volumes,”
l’Aerotecnica Missili e Spazio, Vol. 75, 1997, pp. 77–86.
30
Kier, T. M., “Comparison of Unsteady Aerodynamic Modelling Methodologies with respect to Flight Loads
Analysis,” Proceedings of the AIAA Atmospheric Flight Mechanics Conference and Exhibit, San Francisco, California,
15–18 August 2005.
31
Smith, M. J., Hodges, D. H., and Cesnik, C. E., “Evaluation of Computational Algorithms for Suitable Fluid-
Structure Interactions,” Journal of Aircraft, Vol. 37, No. 2, 2000, pp. 282–294.
32
Bathe, K.-J. and Zhang, H., “Finite Element Developements for General Fluid Flows with Structural Interac-
tions,” International Journal for Numerical methods in Engineering, Vol. 60, 2004, pp. 213–232.
33
Quaranta, G., Masarati, P., and Mantegazza, P., “A Conservative Mesh-free Approach for Fluid-Structure
Interface Problems,” International Conference on Computational Methods for Coupled Problems in Science and En-
gineering, edited by M. Papadrakakis, E. Oñate, and B. Schrefler, CIMNE, Santorini, Greece, 2005.
34
Beckert, A. and Wendland, H., “Multivariate Interpolation for Fluid-Structure-Interaction Problems Using
Radial Basis Functions,” Aerospace Science and Technology, Vol. 5, 2001, pp. 125–134.
35
Schaback, R. and Wendland, H., “Characterization and Construction of Radial Basis Functions,” EILAT Pro-
ceedings, edited by N. Dyn, D. Leviatan, and D. Levin, Cambridge University Press, 2000.
36
Canavin, J. R. and Likins, P. W., “Floating reference frames for flexible spacecraft,” Journal of Spacecraft,
Vol. 14, No. 12, December 1977, pp. 724–732.
37
Haftka, R. T. and Gürdal, Z., Elements of structural optimization, Dordrecht, 3rd ed., 1992.
38
Österheld, C. M., Heinze, W., and Horst, P., “Influence of Aeroelastic Effects on Preliminary Aircraft Design,”
Proceedings of the 22th ICAS Congress, Harrogate, UK, 27 August – 1 September 2000.
25 of 25

A Fast Tool For Structural Sizing, Aeroelastic Analysis and Optimization in Aircraft Conceptual Design

Uploaded by

Copyright:

Available Formats

A Fast Tool For Structural Sizing, Aeroelastic Analysis and Optimization in Aircraft Conceptual Design

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

A Fast Tool For Structural Sizing, Aeroelastic Analysis and Optimization in Aircraft Conceptual Design

Uploaded by

Copyright:

Available Formats

50th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference<br>17th AIAA 2009-2571

4 - 7 May 2009, Palm Springs, California

A Fast Tool for Structural Sizing, Aeroelastic Analysis

L. Cavagna∗†, S. Ricci‡, L. Riccobene§,

Most of procedures available in conceptual design estimate airframe weight by statistical

II. NeoCASS: aeroelasticity in conceptual design

American Institute of Aeronautics and Astronautics

tural sizing, aeroelastic analysis and optimization.

American Institute of Aeronautics and Astronautics

Figure 3. NeoCASS layout.

• WB which directly applies statistical formulas relying on common geometry variables;

American Institute of Aeronautics and Astronautics

III. Weight and balance semi-empirical estimation (WB)

IV. Analytical structural sizing (GUESS)

American Institute of Aeronautics and Astronautics

Figure 4. Layout of the analytical structural sizing tool.

American Institute of Aeronautics and Astronautics

IV.A. Sizing process

American Institute of Aeronautics and Astronautics

Bending moment Stiffener ts

American Institute of Aeronautics and Astronautics

IV.B. Finite-element aeroelastic model

c/4 singularity point

Stress recovery points

Node Lumped mass

Beam collocation points

American Institute of Aeronautics and Astronautics

American Institute of Aeronautics and Astronautics

V.A. Beam model

p(ǫ) = Ni (ǫ) (xi + Ri (g(ǫ)) si ) with g(ǫ) = Ni gi (2)

American Institute of Aeronautics and Astronautics

American Institute of Aeronautics and Astronautics

ε = p′ + p′0× ϕ = N′i xi + p′0× Ni ϕ − N′i si× ϕi (10)

Substituting back the previous expression, the stiffness matrix is recovered:

V.B. Low-fidelity aerodynamic models

American Institute of Aeronautics and Astronautics

Deformed structrural mesh

V.C. Static linear aeroelasticity

American Institute of Aeronautics and Astronautics

ΦTR MEE ΦE qE = 0 (16)

V.D. Optimization method

American Institute of Aeronautics and Astronautics

American Institute of Aeronautics and Astronautics

VI. Application to the Transonic Cruiser TCR aircraft

(a) Wing (b) Vertical fin

VI.A. Weight estimates

American Institute of Aeronautics and Astronautics

Component Mass WB Mass SAAB

VI.B. First structural sizing

American Institute of Aeronautics and Astronautics

• baggage, over passenger compartment length in proportion to beam volume;

Mass density [kg/m]

American Institute of Aeronautics and Astronautics

Kcon Kf = 5.24 Kf = 2 Kf = 1 Kf = 2, ∆p Kf = 5.24 Kf = 2 Kf = 1

VI.C. Structural analysis and optimization

American Institute of Aeronautics and Astronautics

(a) Objective function (b) Design variables

American Institute of Aeronautics and Astronautics

(a) Reference case (b) After optimization