POLITECNICO DI MILANO

Scuola di Ingegneria Industriale e dell'Informazione

Corso di Laurea Magistrale in Ingegneria Aeronautica

An optimization procedure to refine the

optimal design of morphing devices

Relatore: Prof. Sergio RICCI

Correlatore: Ing. Alessandro DE GASPARI

Tesi di Laurea di:

Vittorio CAVALIERI Matr. 863275

Anno Accademico 2017 - 2018

Abstract

Morphing devices for aircraft wings represent a promising technology to

improve the performances, enabling the achievement of more efficient
aircraft. However the design of morphing wings is a complex problem
due to the conflicting requirements that must be handled. Therefore
specific procedures must be developed in order to adequately face the
challenging morphing problem. Optimization analysis is the main tool
that can help the designer in finding solutions that are optimal from both
the aerodynamic and the structural point of view. In the present work an
optimization procedure is proposed for the improvement of the optimal
design of morphing devices based on the distributed compliance concept.
The integration of a mathematical toolbox with a finite element solver
allows to refine morphing solutions but also to adapt existing topologies
to different materials and geometries. Another tool is employed to get
the numerical solution closer to the manufacturing process. The new
procedure is applied to the design of a morphing droop nose to be installed
on a reference regional aircraft. The same steps are repeated using a
superelastic material as an alternative to the aluminium alloy. The results
show a significant improvement of shape quality, but only the superelastic
material allows to completely satisfy stress requirements. Moreover the
preliminary design of a scaled wind tunnel model aimed at operational
tests is performed. All the results validate each aspect of this optimization
procedure which can assist the engineer up to the definition of a virtual
prototype.
Keywords: morphing, optimization procedure, compliant mechanisms,
droop nose, superelastic material

iii
Sommario

I dispositivi morphing per le ali dei velivoli rappresentano una tecnologia

promettente per migliorare le prestazioni, consentendo di realizzare velivoli
più efficienti. Tuttavia il progetto di ali morphing è un problema complesso
a causa dei requisiti conflittuali che occorre gestire. Pertanto apposite
procedure devono essere sviluppate per affrontare in maniera adeguata
l’impegnativo problema del morphing. L’ottimizzazione è il principale
strumento che può aiutare il progettista nel trovare soluzioni che siano
ottime sia dal punto di vista aerodinamico che strutturale. In questo
lavoro viene proposta una procedura di ottimizzazione per il miglioramento
del progetto ottimo di dispositivi morphing basati sul’idea di flessibilità
distribuita. L’integrazione di strumenti matematici con un solutore a
elementi finiti permette di perfezionare soluzioni morphing ma anche di
adattare topologie esistenti a materiali e geometrie diversi. Un altro
strumento è utilizzato per avvicinare la soluzione numerica al processo
produttivo. La nuova procedura è applicata al progetto di un droop nose
morphing da installare su un velivolo regionale di riferimento. Gli stessi
passi sono ripetuti usando un materiale superelastico come alternativa alla
lega di alluminio. I risultati mostrano un miglioramento significativo della
qualità della forma, ma solo il materiale superelastico permette di soddisfare
completamente i vincoli sugli sforzi. Inoltre è stato eseguito il progetto
preliminare di un modello in scala destinato a prove di funzionalità in
galleria del vento. Tutti i risultati validano ogni aspetto di questa procedura
di ottimizzazione che può assistere l’ingegnere fino alla definizione di
prototipi virtuali.
Parole chiave: morphing, procedura di ottimizzazione, meccanismi com-
pliant, droop nose, materiale superelastico

iv
Ringraziamenti

Desidero ringraziare il Professor Ricci per avermi dato la possibilità di

intraprendere questo lavoro di tesi e per la sua supervisione durante tutto
il percorso.
Vorrei poi ringraziare Alessandro De Gaspari per il suo costante supporto,
per avermi seguito con la sua esperienza e con la grande passione che mette
nel suo lavoro.
Ringrazio inoltre tutti quei docenti che nel corso degli anni hanno con-
tribuito ad accrescere la mia preparazione e le mie competenze.
Ringrazio tutta la mia famiglia, i miei genitori per il loro affetto e per
avere sempre appoggiato le mie scelte di vita anche quando non potevano
comprenderle a fondo, mio fratello Riccardo per il suo essere continuamente
fonte di ispirazione e motivazione e per ricordarmi sempre di analizzare
ogni situazione da punti di vista diversi. Ringrazio inoltre tutti gli altri
parenti che mi sono stati vicini.
Ringrazio Emanuele per tutte le esperienze che abbiamo condiviso sin da
quando eravamo piccoli, per riempire i miei silenzi nelle nostre conversazioni
e per aiutarmi a trovare una direzione quando ne ho bisogno.
Ringrazio tutti quei colleghi universitari con cui ho trascorso le giornate di
lezione, affrontato lo studio per gli esami e le stimolanti collaborazioni nei
progetti grazie ai quali ho imparato divertendomi.
Ringrazio tutti gli amici che mi sono stati vicini nel corso degli anni, anche
quelli con cui le strade ora si sono divise. Tutti hanno contribuito in un
modo o nell’altro a farmi diventare la persona che sono e al raggiungimento
dei miei traguardi.
Infine ringrazio Ornella per essermi vicina nei momenti di gioia e in quelli
di difficoltà, per tutto ciò che amiamo condividere, per i suoi stimoli a
ricercare sempre nuove sfide e per sognare ogni giorno insieme a me ad
occhi aperti.

Milano, Dicembre 2018 V. C.

v
Contents

1 Introduction 1
1.1 General overview of morphing aircraft . . . . . . . . . . . . 1
1.2 Active camber morphing . . . . . . . . . . . . . . . . . . . 4
1.3 Compliant structures . . . . . . . . . . . . . . . . . . . . . 7
1.4 Aims of the work . . . . . . . . . . . . . . . . . . . . . . . 9
1.4.1 Thesis outline . . . . . . . . . . . . . . . . . . . . . 11

2 Design procedure 13
2.1 Morphing shape optimization . . . . . . . . . . . . . . . . 15
2.1.1 Airfoil geometry representation . . . . . . . . . . . 16
2.1.2 Optimization problem definition . . . . . . . . . . . 19
2.2 Multi-objective Genetic Algorithm . . . . . . . . . . . . . 20
2.2.1 Load path representation . . . . . . . . . . . . . . . 21
2.2.2 Genetic Algorithms based on Multi-objective approach 22
2.2.3 Optimization Problem . . . . . . . . . . . . . . . . 23
2.3 Solution refinement . . . . . . . . . . . . . . . . . . . . . . 25

3 Droop nose 27
3.1 Reference wing . . . . . . . . . . . . . . . . . . . . . . . . 27
3.2 Topological synthesis results . . . . . . . . . . . . . . . . . 29
3.3 Shape memory alloys . . . . . . . . . . . . . . . . . . . . . 32
3.3.1 Superelastic behaviour . . . . . . . . . . . . . . . . 35
3.3.2 Modeling and numerical simulations . . . . . . . . . 35
3.3.3 Selection of material properties . . . . . . . . . . . 39

4 Solution refinement tools 43

4.1 Gradient-based optimization . . . . . . . . . . . . . . . . . 44
4.1.1 Optimization problem definition . . . . . . . . . . . 44
4.1.2 Gradient-based methods . . . . . . . . . . . . . . . 45
4.1.3 Automatic model generation . . . . . . . . . . . . . 46
4.1.4 Optimizer-solver interface and problem set-up . . . 50

vii
viii Contents

4.2 Compliant mechanism shape optimization . . . . . . . . . 51

5 Results 57
5.1 Gradient-based optimization . . . . . . . . . . . . . . . . . 57
5.1.1 Isotropic material . . . . . . . . . . . . . . . . . . . 62
5.1.2 Nitinol material . . . . . . . . . . . . . . . . . . . . 71
5.2 Compliant mechanism shape optimization . . . . . . . . . 83
5.2.1 Isotropic material . . . . . . . . . . . . . . . . . . . 83
5.2.2 Nitinol material . . . . . . . . . . . . . . . . . . . . 88
5.3 Considerations about the use of Nitinol . . . . . . . . . . . 95
5.4 Considerations about the optimization procedure . . . . . 97

6 Applications 99
6.1 Preliminary design of a scaled wind tunnel model . . . . . 99
6.1.1 Results . . . . . . . . . . . . . . . . . . . . . . . . . 100

7 Conclusions 111

Bibliography 115
List of Figures

1.1 Classification of shape morphing wings . . . . . . . . . . . 2

1.2 F-111 Advanced Fighter Technology Integration . . . . . . 5
1.3 Mission Adaptive Wing: variable camber leading edge . . . 5
1.4 Dornier patent (1979) . . . . . . . . . . . . . . . . . . . . . 6
1.5 Two-dimensional compliant gripper . . . . . . . . . . . . . 8

2.1 Optimization procedure for the design of morphing wings . 14

2.2 Two-levels approach . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Parametric framework . . . . . . . . . . . . . . . . . . . . 16
2.4 CST geometric parameters for airfoil sections . . . . . . . . 17
2.5 A demonstrative connected load path representation of a
SIMO compliant mechanism . . . . . . . . . . . . . . . . . 22

3.1 NLF optimized wing . . . . . . . . . . . . . . . . . . . . . 28

3.2 Reference Aircraft with morphing leading edge . . . . . . . 28
3.3 Aerodynamic loads for the structural (left) and kinematic
(right) requirements . . . . . . . . . . . . . . . . . . . . . . 30
3.4 Pareto Front from the genetic algorithm . . . . . . . . . . 30
3.5 Optimal selected solution . . . . . . . . . . . . . . . . . . . 30
3.6 Updated selected solution . . . . . . . . . . . . . . . . . . 31
3.7 Shape-memory effect (left) and superelastic behaviour (right) 33
3.8 Representation for the appearance of superelasticity in
temperature-stress space . . . . . . . . . . . . . . . . . . . 34
3.9 Uniaxial stress-strain curve for the superelastic behaviour . 36
3.10 Stress-temperature diagram for the superelastic behaviour 37
3.11 Experimental data for the uniaxial tension stress-strain
response of a Ni-Ti alloy [2] . . . . . . . . . . . . . . . . . 37
3.12 Reference and limit temperatures in the stress-temperature
diagram for the superelastic behaviour . . . . . . . . . . . 41

4.1 Abaqus models for the gradient-based optimization . . . . 47

ix
x List of Figures

4.2 Steps from the beam elements model to the 2D solid elements
model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.3 Shape optimization: example of design area . . . . . . . . 54
4.4 Kinematic coupling for the connection of reference points
to boundary edges . . . . . . . . . . . . . . . . . . . . . . 55

5.1 Pressure coefficient distributions . . . . . . . . . . . . . . . 59

5.2 Abaqus model corresponding to the initial variables . . . . 60
5.3 Pressure loads on the model . . . . . . . . . . . . . . . . . 61
5.4 Isotropic material: deformed shape of the initial solution . 63
5.5 Isotropic material: initial solution undeformed/deformed
shape and target shape . . . . . . . . . . . . . . . . . . . . 64
5.6 Isotropic material: LSE vs iteration . . . . . . . . . . . . . 64
5.7 Isotropic material: initial solution and optimal solution
deformed shapes . . . . . . . . . . . . . . . . . . . . . . . . 65
5.8 Isotropic material: Von Mises stress in the rib . . . . . . . 66
5.9 Isotropic material: optimal solution deformed shape and
target shape . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5.10 Isotropic material: thicknesses . . . . . . . . . . . . . . . . 68
5.11 Isotropic material: LSE vs iteration (sizing variables only) 70
5.12 Nitinol: deformed shape of the initial solution . . . . . . . 72
5.13 Nitinol: initial solution undeformed/deformed shape and
target shape . . . . . . . . . . . . . . . . . . . . . . . . . . 73
5.14 Nitinol: LSE vs iteration . . . . . . . . . . . . . . . . . . . 73
5.15 Nitinol: initial solution and optimal solution deformed shapes 74
5.16 Nitinol: Von Mises stress in the rib . . . . . . . . . . . . . 75
5.17 Nitinol: optimal solution deformed shape and target shape 76
5.18 Nitinol: thicknesses . . . . . . . . . . . . . . . . . . . . . . 77
5.19 Nitinol: LSE vs iteration (sizing variables only) . . . . . . 79
5.20 Optimal solutions: thicknesses . . . . . . . . . . . . . . . . 80
5.21 Optimal solutions: isotropic material and nitinol deformed
shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.22 Optimal solutions: isotropic material and nitinol unde-
formed shapes . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.23 Isotropic material: 2D model of the compliant mechanism . 84
5.24 Isotropic material: static analysis of the compliant mechanism 85
5.25 Isotropic material: optimization region . . . . . . . . . . . 86
5.26 Isotropic material: shape optimization . . . . . . . . . . . 86
5.27 Nitinol: 2D model of the compliant mechanism . . . . . . . 88
5.28 Nitinol: static analysis of the compliant mechanism . . . . 89
5.29 Nitinol: shape optimization for the right region . . . . . . 90
List of Figures xi

5.30 Nitinol: stress-strain curve in the right region . . . . . . . 92

5.31 Nitinol: shape optimization for the left region . . . . . . . 93
5.32 Nitinol: stress-strain curve in the left region . . . . . . . . 94
5.33 Nitinol: deformed shape of optimal solution at different
temperatures . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.34 Nitinol: stress-strain relationship at different temperatures 97

6.1 Scaled model and full scale model . . . . . . . . . . . . . . 101

6.2 Scaled model: deformed shape of the initial solution . . . . 103
6.3 Scaled model: LSE vs iteration . . . . . . . . . . . . . . . 104
6.4 Scaled model: initial solution and optimal solution deformed
shapes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.5 Scaled model: Von Mises stress in the rib . . . . . . . . . . 106
6.6 Scaled model: optimal solution deformed shape and target
shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.7 Scaled model: thicknesses . . . . . . . . . . . . . . . . . . 107
6.8 Scaled model: initial solution and optimal solution unde-
formed shapes . . . . . . . . . . . . . . . . . . . . . . . . . 108
List of Tables

3.1 Superelasticity UMAT: list of parameters . . . . . . . . . . 38

3.2 Nitinol characteristics from [46] . . . . . . . . . . . . . . . 40
3.3 Critical stresses at T0 = 293 K . . . . . . . . . . . . . . . . 40

5.1 Nitinol: optimal solution at different temperatures . . . . . 95

6.1 Mechanical parameters of PA 2200 . . . . . . . . . . . . . 100

6.2 Mechanical parameters of glass . . . . . . . . . . . . . . . 100

xiii
Chapter 1

Introduction

1.1 General overview of morphing aircraft

The design of conventional aircraft wings is the result of an optimization
process based on a single flight condition, typically the cruise, at a given
speed and altitude. Regarding the other conditions the performances of
the aircraft are sub-optimal. For example, low cambered thin airfoils are
suitable for high speed flight while highly cambered thick airfoils are better
for low speed high angle of attack flights such as take-off and landing.This
specific problem is tipically addressed with the adoption of high lift devices.
If there was the possibility of changing on demand the wing shape during
the mission, near-optimal performances for different conditions could be
achieved. This potentiality is the idea behind the morphing concept.

Nowadays, increased awareness of environmental issues like those in the

ACARE Flightpath 2050, concerning emissions and noise, in combination
with growing passenger volumes determine the request for more efficient
aircraft. For these reasons aircraft manufacturers have to take into account
novel configurations and technologies in order to achieve substancial im-
provements in aircraft efficiency, in terms of both fuel consumption and
acoustic emissions. Morphing aircraft can be the solution to overcome the
future challenges in aviation.
Morphing is a bio-inspired concept. Looking at the wings of birds and
insects it is evident their ability of adaptation depending on the required
aerodynamic characteristics in a specific flight condition [39]. It can be
stated that in nature mission adaptive and flight control capabilities are
naturally implemented. These lessons learned push the research in the di-
rection of trying to emulate such adaptability features, aiming at achieving

1
2 1.1. General overview of morphing aircraft

the related benefits.

Many definitions of morphing aircraft there exist. According to the NATO

RTO Technical Team on Morphing Vehicles morphing is "real-time adap-
tation to enable multi-point optimized performance" [37]. Weisshaar [63]
suggested that "morphing aircraft are multi-role aircraft that change their
external shape substantially to adapt to a changing mission environment
during flight".
Although morphing has been acquiring great visibility in the last years, it
is far from being a recent idea. Even the Wright brothers’ first biplane was
provided with morphing flight control. Indeed, roll control was performed
by changing the twist of the wing using cables actuated by the pilot [5].
Moreover the usual conventional aircraft are equipped with many classical
examples of morphing, like slats, flaps and retractable landing gear.
According to the wing parameters involved a classification of morphing
is possible: planform alteration, out-of-plane transformation and airfoil
adjustment [5]. Each transformation can be due to different parameters,
as shown in Figure 1.1. Wing planform morphing can concern span, sweep
or chord changes. Wing out-of-plane transformation is mainly affected
by twist, dihedral, and span-wise bending. Airfoil adjustment is tipically
achieved via camber variation, although thickness change has also been
considered. Airfoil modification by means of camber variation will be the

Figure 1.1: Classification of shape morphing wings

morphing strategy discussed in this thesis work. An in-depth overview of

the related solutions considered over the years is reported in section 1.2.

Morphing solutions can provide many benefits. The possibility, in subsonic

regime, to modify the airfoil geometry depending on the flight conditions,
allows to increase the lift/drag ratio. [5] Regarding the replacement of
Chapter 1. Introduction 3

classical high lift devices with smart ones, the major advantage lies in the
resulting aerodynamic efficiency, the low drag and also the airframe noise
reduction. Other advantages of morphing concern the integration between
structure and functionality, the possibility of embedding smart material
actuation into the structure with a reduced mechanical complexity [39],
the natural redundancy provided by distributed actuation [5].
Unfortunately the design of morphing structures is not an easy task.
Specific tools are needed to assist the engineers in the many design phases
encountered when dealing with this kind of structures. The reason behind
this complexity lies in the morphing paradox: the same structure has to be
stiff enough to withstand the external aerodynamic loads without suffering
appreciable deformations and it has to be fairly flexible to change its
shape into the required one [3]. In this sense many authors talk about the
morphing challenge. In general lots of key disciplines may be involved in
the aforementioned design process. According to [64], they are: advanced
materials, smart structures, flow control, acoustics, controls, integration,
multidisciplinary design optimization.
Careful attention must be devoted to skin design. Indeed, in order to
properly apply the morphing wing concept, flexible or sliding aerodynamic
surfaces are needed. The design of flexible skins requires trade-off studies
since they are subjected to the usual conflicting requirements of morphing
structures [5]. The main problem is related to the high strains in the skin.
Moreover the external surface must be smooth in order to not compromise
the aerodynamic efficiency. Many skin solutions have been proposed, for
instance the use of a composite corrugated structure [61] and elastomeric
matrix composites [41, 45]. Being part of a semi-monocoque structure, the
skin plays a role in sustaining the aerodynamic loads, hence it is subjected
to tensile, compression and shear stresses. This explains the need for
compromise between its stiffness and its flexibility [3]. A suitable morphing
skin must possess a high curvature at rupture. Moreover, the stiffness
has to be maximized in the spanwise direction in order to withstand the
aerodynamic loads without requiring an excessive number of kinematic
ribs. The typical solution to satisfy these requirements is the adoption of
extremely anisotropic skin structures [51].
In addition to the design complexity, many other issues are related to
morphing technology and they must be carefully assessed before a wider
application of morphing structures could take place.
First of all "large shape change concepts usually have associated design
penalties such as added weight or complexity" [5]. Therefore sound esti-
mations have to state if overall system-level benefits overcome the above-
mentioned drawback. Another problem concerns the unconventional nature
4 1.2. Active camber morphing

of morphing devices that makes their certification an open issue. This

is mainly due to the typically low Technology Readiness Level (TRL)
of morphing solutions. With this in mind it is clear that a transition
period to bring the current technology to an high maturity level is required.
In order to do that, technology programs shoud flank the research activities.

To sum up, morphing wings have the potential to revolutionize future

designs of next-generation aircraft, improving significantly their perfor-
mances [4, 5, 8]. However, a lot of work is needed to adequately face the
challenging morphing problem and to overcome the widespread skepticism
about morphing so that morphing aircraft can be seen filling the skies in
the near future.

1.2 Active camber morphing

One of the most promising morphing concepts is the active camber morph-
ing, whose aim is the variation of the airfoil camber in order to increase
the aerodynamic performances, especially in take-off and landing phases.
Different implementing solutions of this concept have been thought and
evaluated during the years. In 1973, at the NASA Ames 14-ft transonic
wind tunnel, Boeing tested an advanced variable camber wing, having some
smooth and curved variable camber flaps in addition to the usual hinged
leading and trailing-edge flaps [22]. The performance improvement was
meaningful, however the required internal design was too complex. Later
Boeing research [7, 17] focused on wing camber control for military aircraft
performed through an automated control system. The first significant
program was the Advanced Fighter Technology Integration (AFTI)/F-111
(Figure 1.2 [12]), promoted by NASA in collaboration with the USAF. The
adopted solution for the wing modification consisted in sliding panels and
flexible glass fiber panels for the trailing edge surfaces, while the leading
edge surfaces were made of flexible composite panels (Figure 1.3). Camber
control was provided by an internal rods and linkages arrangement driven
by electro-hydraulic actuators. The flight tests assessed valuable perfor-
mance increase in terms of range, efficiency and loads withstanding [56,
57].
In the field of civil transport aircraft, Szodruch and Hilbig [59, 60] demon-
strated the benefits provided by a variable camber wing concept.
In the framework of the ADIF project carried out by EADS-Airbus, Daim-
lerChrysler F&T and DLR, Monner [40] worked at the replacement of
the usual rigid ribs with a flexible solution able to guarantee at the same
Chapter 1. Introduction 5

Figure 1.2: F-111 Advanced Fighter Technology Integration

Figure 1.3: Mission Adaptive Wing: variable camber leading edge

6 1.2. Active camber morphing

time the required high stiffness. The proposed finger concept consisted
of ribs made of several plates linked by revolute joints. The skin relative
movement was allowed by including slide joints. The focus of the research
was an optimization aimed at the minimization of the joint stress while
achieving the desired camber. The leading idea of the developed concept
is that introducing a chordwise and spanwise differential camber variation
with a smooth contour and no gaps can provide many advantages from
both the aerodynamic and the structural points of view. For instance
higher aerodynamic efficiency with the related saving in fuel, higher opera-
tional flexibility and reduction of bending moment at the wing root can
be achieved.
The subsequent work of Monner within the SmartLED project, based on a
collaboration between DLR and EADS, concerned the development of a
smart leading edge as an alternative to the conventional droop nose device
used for the A380. The starting point of the new smart device was the
Dornier Patent DE 2907912 (Figure 1.4) and an optimization process for
the design of the overall system was performed. The main purpose of the
device was to achieve a smooth leading surface to be deflected for high
lift application. The seamless and gapless design for the high lift leading
egde device aimed at the reduction of airframe noise and drag and also to
promote the introduction of laminar wing technology.

Figure 1.4: Dornier patent (1979)

Many other EU projects included the design of smart leading-edge devices.

SADE project focused on the design, manufacturing, and ground testing
of a seamless smart leading-edge device, which was enabler for laminar
wings and offered great benefits in reducing the acoustic emissions [25, 32].
The proposed concept for the internal structure design were the eccentric
Chapter 1. Introduction 7

beam and a kinematic chain. Subsequently the SARISTU project dealt

with the integration of typical leading-edge functions like de-icing, erosion
protection, impact protection, bird strike protection and lightning strike
protection [27]. The major challenge was due to the limited design space
and the large curvature at the wing leading edge. In both cases, wind
tunnel tests assessed the performances.
The described concepts of droop nose device are based on a flexible skin
coupled to a rigid mechanism having rigid levers and kinematic joints. An
alternative to the rigid kinematics approach for the realization of variable
camber wings has been developed by Flexsys Inc. [30] based on compliant
structures, which are discussed in section 1.3.
In this thesis an optimization procedure for the design of morphing wings
is proposed following the active camber concept, but limited to leading
and trailing-edge deflections, with an internal structure designed with
compliant structures.

1.3 Compliant structures

"Compliant mechanism are mechanical devices that achieve motion via
elastic deformation". They are designed to be flexible and this flexibility,
efficiently implemented into the structure, allows the desired motion [20,
42]. A force acts as input, the structure deforms and a desired motion
or shape is obtained as output. The input energy is stored as strain
energy in the structure, and most of that strain energy is transformed
into controlled displacements of the output points in order to obtain the
requested shape [52]. This concept of distributed compliance was originally
proposed by Kota [30] as an alternative to the distributed actuation one.
Compliant structures are optimized to deform in such a way to produce the
desired shape change. Instead of using hinges and rigid links, an efficient
use of the material provides the achievement of the required shape. An
example of compliant structure is shown in Figure 1.5 [29].

The major issue when dealing with the design of compliant structures is
that conflicting requirements must be simultaneously satisfied: the struc-
ture must be flexible enough to accomplish the desired motion (kinematic
requirement), moreover it must be stiff enough to withstand the exter-
nal loads (structural requirement). In order to achieve a feasible design
according to these requirements, dedicated design procedures have to be
developed and specific tools like multiobjective optimization are required
for them. Due to the peculiar characteristic of compliant structures, they
8 1.3. Compliant structures

Figure 1.5: Two-dimensional compliant gripper

can be considered having the features of both structures and mechanisms:

they are structures with imposed deformation, they are mechanisms with-
out kinematic joints [36].

The great interest in compliant structures lies in the numerous advan-

tages they can give. Indeed, unlike rigid mechanisms, they are built in a
single-piece without joints, hence they don’t require assembly and are less
subjected to wear, noise and backlash [38]. The absence of flexural joints
reduces high stress concentration. Moreover compliant structures exhibit
higher fatigue life and reliability. The easier manufacture can also lead to
time and cost savings [29].
In addition to the above-mentioned advantages, compliant mechanisms are
suitable to be used in aircraft wings in order to realize seamless and gapless
high lift devices. Indeed, replacing flaps and slats with morphing leading
and trailing–edge surfaces eliminates the discontinuities. The result is that
these camber variations allow the optimization of low speed performances
without the drawback of conventional high lift systems.

FlexSys Inc., the company founded by Professor Sridhar Kota in 2001,

exploits material elasticity to design variable geometry trailing edge control
surfaces, based on the distributed compliance concept [24]. Moreover, in
collaboration with the Air Force Research Laboratory and NASA, FlexSys
team has successfully flight tested the developed structure on a Glufstream
III aircraft. The flight tests have demonstrated the structural feasibil-
ity and robustness of the FlexFoilTM Adaptive Compliant Trailing Edge
(ACTE) [28]. This technology can improve aerodynamic efficiency and
reduce noise during takeoffs and landings. Estimated fuel savings between
2% to 11% have encouraged further research, also related to different
Chapter 1. Introduction 9

variable geometry technologies, like trim tabs, leading–edges, winglets and

engine inlets.

As far as the design of compliant mechanisms is concerned, according

to [23], the synthesis procedure must consider different criteria: required
kinematic motion, available design space, required stiffness under external
loads, material properties, stress limits, buckling, weight limitations. It is
evident that the design is a complex problem, and must be the result of
many trade-offs between different objective functions, subjected to many
constraints.
Possible objective functions are: structural weight, required energy to ac-
complish the shape change, aerodynamic characteristics, control effort [32].
The design variables can be classified in shape variables and structural
variables. The optimization methods can be classified in topology optimiza-
tion algorithms and parameter optimization (gradient-based optimization
algorithms).
In this context, topology synthesis assumes a fundamental role for the
definition of the overall structural configuration able to give the desired
behaviour. Genetic algorithms can be adopted for the initial searching over
the optimization region to determine a suitable topology, but they must
be followed by an additional search, at a local level, performed through a
gradient-based optimization, to determine a more accurate sizing of the
mechanism in order to improve the previously found optimal solution [36].
Therefore suitable tools must be developed in order to face the challenging
design of compliant structures. Among the others, attention should be
paid to avoid permanent deformation or fatigue failure of the material [53].

1.4 Aims of the work

In this thesis an approach for the optimal design of morphing structures
is presented. The main focus of the work is on the internal structure
design by means of compliant mechanisms. The choice of the distributed
compliance concept is led by the need to avoid the high stresses due to the
adoption of rigid mechanisms.
The proposed procedure is general for the application to leading and
trailing–edge morphing devices. However, since the criticalities concerning
the leading–edge morphing solutions are higher, only application to droop
nose will be shown. The main difference between leading and trailing edge
is the different chord extension that is lower in the first case, hence it
requires a more complex effort to achieve the desired deflection.
10 1.4. Aims of the work

Many difficulties are encountered when dealing with the design of a droop
nose and also compliant structures can be characterized by an high stress
level. In this thesis the problem is overcome by the adoption of Nitinol as
the material for the internal mechanism. Nitinol is here used in a passive
way, by exploiting its superelastic behaviour instead of the shape memory
effect.

The working environment for the examples of application of the procedure

is the contribution of POLIMI in the framework of the EU funded Clean
Sky 2 REG-IADP AG2 project [55].

Some important preliminary considerations about the selected morphing

solution follow. The idea of design addressed in this thesis is based on the
concept of a wing composed by a central wing-box with morphing leading
and trailing edges attached to it. This choice is based on the belief that the
traditional central wing-box must be left unchanged by morphing, both
for strength-to-weight ratio considerations and for functionality issues,
such as the need for fuel storage in the wing. As a consequence, the
concept here considered is that of hybrid wings. The traditional wing-box
is the same of conventional fixed-wing aircraft and also the approach for
its design doesn’t change. Leading and trailing edges become morphing
devices, with the role of conformable and gapless control surfaces. This
partially morphing solution have the potential to be worth of consideration
from industries and certification bodies and can represent the intermediate
step in the transition from conventional aircraft wings to more futuristic
fully-morphing wings.
Another issue to be considered is the impact of aeroelasticity on the
functionality of morphing devices, in terms of deployment capability and
accuracy of the target deformed shape. Concerning this point, the idea
is that a compliant solution instead of a rigid one has the advantage to
introduce a degree of freedom into the structure. This should avoid further
stress concentrations and also the appearance of unwanted coupling effects
between the morphing device and the rest of the structure. Indeed, a
problem encountered when using rigid kinematics is that the resulting
structure is overdetermined. This may represent a problem, in particular
in these times of more and more flexible aircraft. Moreover no one has
tested so far rigid kinematics-based mechanisms attached to a wing-box
structure. Conversely, as an activity related to the design of the Clean Sky
2 droop nose, FEM simulations have virtually assessed the suitability of the
coupling between wing-box and morphing devices realized with compliant
mechanisms: the presence of morphing device doesn’t modify the wing
Chapter 1. Introduction 11

deflection under loads neither the wing deflection compromises droop nose
deployment.

1.4.1 Thesis outline

An overview of the chapters content is reported:

Chapter 1 has described the concept of morphing aircraft, focusing on

the potential benefits but also on the intrinsic complexity. Particular
attention has been given to the active camber morphing technology
and its application to the design of high lift devices based on compliant
structures. Finally the aims of the work have been reported.

Chapter 2 introduces the proposed procedure for the refinement of the

design of morphing devices, its potentialities and the background on
which it is located.

Chapter 3 describes the optimal solution that will be improved by means

of the optimization procedure. Moreover it contains a review of
the material superelastic behaviour that will be exploited in the
applications.

Chapter 4 focuses on the new optimization tools, showing their working

principles as well as the reasons that have led to their development.

Chapter 5 reports the results of the application of the procedure and

analyzes the obtained improvement.

Chapter 6 is devoted to additional examples and their results.

Chapter 7 draws the conclusions of the whole thesis and outlines possible
future works.
Chapter 2

Design procedure

The aim of this thesis is the development of a procedure for the optimal
design of morphing wings. A schematic representation of this procedure is
shown in Figure 2.1.
The general framework can be outlined in four phases:
1. An aerodynamic shape optimization with structural constraints. The
objective is the definition of the optimal morphing shape according
to the requirements in terms of performances. At the same time, a
constraint on the strain computed as a function of curvature assures
the feasibility of the skin solution from a structural point of view.
2. A multi-objective genetic algorithm optimization for the definition
of the topology of the internal compliant structure. In this phase a
fictitious material is used for the compliant structure.
3. A gradient-based optimization to optimize the sizing variables of the
selected topology when the actual material of the compliant structure
is included. The objective is the improvement of the morphing shape
obtained when the mechanism is actuated, while limiting the stress
inside the structure.
4. A shape optimization of the compliant mechanism to reduce local
peaks of stress and finalize the design details.
At the end of the procedure, the designed solution has to be verified from
both the morphing shape quality and the structural points of view. If the
result is acceptable, the mechanism is ready for the manufacturing.

The background of the suggested procedure is the two-levels optimization

approach proposed by De Gaspari [13], whose capabilities are expanded
here developing dedicated tools. Their main potentialities are:

13
14

• the possibility of postponing the choice of the material and the

geometric scaling factor, after the definition of an optimal topology;

• the versatility of the FEM simulations, in terms of material constitu-

tive laws and modeling;

• the high-level refinement of the final solution, which brings it nearer

to the step of manufacturing process.

Figure 2.1: Optimization procedure for the design of morphing wings

The following sections focus on the description of the existing optimization

procedure, also introducing the main differences for its use in the new
approach. The detailed description of the new tools is reported in chapter 4.

The scheme of the two-levels approach for the design of compliant structure
is shown in Figure 2.2.
The first level of the procedure consists in an aero–structural optimization
of the morphing shape. The objective is to obtain the most efficient
aerodynamic shape, but minimizing the axial and bending strains in the
skin. This aspect is very relevant since the skin plays an important role in
Chapter 2. Design procedure 15

Figure 2.2: Two-levels approach

the whole morphing solution. Therefore taking into account its structural
behaviour from the begininning helps in finding feasible structural solutions
at the later stage.
The aerodynamic shape needs a suitable geometric representation to be
described. The approach here adopted is the compact airfoil representation
technique (CST), based on the method proposed by Kulfan in [31] and
extended to leading and trailing edge continuous control surfaces. The
advantages of this fully analytical description of the airfoil shape are the
reduced number of parameters needed and the straightforward calculation
of the stresses into the skin.
After the completion of the first level, the topological synthesis of the
internal compliant mechanism takes place. This design is based on the
Load Path Representation method [34] and aims at obtaining the best
internal structural configuration able to achieve the target optimal shape
previously found.

2.1 Morphing shape optimization

The morphing shape optimization is performed by means of a Knowledge-
Based Engineering (KBE) framework that revolves around an object ori-
ented code named PHORMA (Parametric sHapes for aerOdynamic and
stRuctural Modelling of Aircraft) [15]. At this stage, the objective is the
design of morphing shapes which are optimal from the aerodynamic point
of view, under skin structural requirements. The aerodynamic optimality
can concern the efficiency, in order to reduce the fuel consumption, but also
the improvement of the performances over a wide range of different flight
conditions. An overview of the framework is shown in Figure 2.3. It is able
16 2.1. Morphing shape optimization

CSTv3
Parametric Geometry Representation XML File
for Aircraft Section Shapes

PHORMA
Main Exchange Class

2D CFD Mesh 3D Aerodynamic

Interface 3D FEM Modeler CAD Interface
(Structured grids) Interface

IGES STEP
ICEM 3D Aerodynamic Model File File

Solver
PFEM ANSYS MSC
Class APDL Nastran
(In-house Model Input File
Post-processing FEM code)
Class

Figure 2.3: Parametric framework

to provide advanced parametric capabilities for the aircraft description.

Moreover it can couple the parametric geometry representation with aero-
dynamic and structural solvers, enabling the computation of aerodynamic
performances and the prediction of the skin structural response. Thanks
to the second aspect, there is the guarantee that only feasible shapes are
considered in the optimization problem; these shapes must satisfy wing–box
volume constraints and morphing skin structural requirements.

2.1.1 Airfoil geometry representation

In order to describe aerodynamic shapes to be efficiently included in an
optimization algorithm, it is essential to represent adequately the admissible
continuous geometry changes. The adopted parameterization technique
has a direct effect on the number of design variables and the accuracy of
the optimal shape.
According to the morphing application, the chosen technique must allow
to deform the global shape properties without, however, affecting the
local regularity of the airfoil. Therefore the analytical approach is used
to introduce a compact formulation that combines a baseline shape and a
shape function.
The starting point of the presented technique is the analytically func-
tion transformation technique, proposed by Kulfan. However, in order
to properly represent leading and trailing-edges morphing, a particular
Chapter 2. Design procedure 17

β
δ
ζ(ψ) ing T.E. Mo
ph rphing
. M or Skin
s
L.E ψ
Structural Box

δ LE ZTE δ TE
ZLE RLE
CFS ∆ZTE
CRS
xLE
c

Figure 2.4: CST geometric parameters for airfoil sections

modification of the basic technique has been performed.

Bezier curves can be used to model smooth curves. The Class func-
tion/Shape function Transformation (CST) geometry representation method
is an application of the generalized Bezier curve definition to represent
aircraft geometries. The class function defines fundamental classes and
the shape function defines unique geometric shapes within each class. The
shape function is represented with a Bernstein polynomial, thus eliminating
slope or curvature discontinuities.
The CST parameterization is based on merging four terms: a Shape
Function, a Class Function and two additional terms related to the air-
foil leading–edge and trailing–edge shapes. The general mathematical
expression representing the airfoil geometry is:

ζ(ψ) = CNN21 (ψ)a · bTn (ψ) + ψζT E + (1 − ψ)ζLE (2.1)

where, ψ = x/c, ζ = z/c are the non–dimensional coordinates with respect
to airfoil chord c, ζT E = ZT E /c and ζLE = ZLE /c. Other significant
parameters are the airfoil leading–edge nose radius RLE , the trailing–edge
boat–tail angle β, the thickness ∆ZT E and the angular thickness δ. All
these “CST parameters” are shown in Figure 2.4. Starting from these
parameters, more definitions follow. The leading edge deflection and the
trailing edge equivalent deflection are respectively:

δLE = arctan (ZLE /(CF S − xLE )) (2.2)

δT E = arctan (ZT E /(c − CRS ))

where CF S and CRS are the front and rear spar positions.
18 2.1. Morphing shape optimization

The first term of Equation 2.1 is the Class function, that affects both the
leading and the trailing edge shape. Its general form is:

CNN21 (ψ) , ψ N 1 (1 − ψ)N 2 (2.3)

By varying the exponents (N 1 and N 2) in the class function, it is possible
to obtain different basic general shapes.
The second term of Equation 2.1 is the Shape function. It is defined
using a set of (n + 1) Bernstein polynomial components Bi (ψ), of selected
order n, arranged in the function array bn (ψ) and scaled by the unknown
extra–coefficients Ai in vector a.
The i–th shape function component is defined as the product between the
i–th Bernstein polynomial component and the i–th associated binomial
coefficient. Each Bernstein polynomial component is multiplied by the
Class function defining the systematic decomposition of the airfoil shapes
into corresponding scalable airfoil components. Consequently, the constant
coefficient Ai can be used to scale the entire airfoil function which is a
smooth function that eliminates local discontinuities.
The first and last terms of vector a are related to the leading and trailing
edge boundary conditions and are equal to the Shape function evaluated
in x/c = 0 and x/c = 1. In particular:

p
a(1) = 2RLE /c, a(n + 1) = tan β + ZT E /c − ZLE /c (2.4)

The other n − 1 terms have no effect neither on the leading–edge radius

nor on the trailing–edge angle and can be used to scale the individual
Bernstein components.
Thanks to the analytical nature of the CST parameterization technique
it is also possible to describe the structural behavior of the skin. The
calculation of the first and second order derivatives can be used to compute
the length and curvature of airfoil upper and lower surfaces:
Z c s  2
dζ
L(x) = 1+ dx (2.5)
0
dx

ζ 00 (x)
κ(x) = , (2.6)
(1 + ζ 02 )3/2
These geometrical quantities allow the computation of the stress (or strain)
inside the skin. This stress is the sum of two contributes: the axial stress
σaxial and the bending stress σbend . The former is related to the variation
Chapter 2. Design procedure 19

of length between the undeformed shape and final morphing shape. The
bending stress is related to the curvature difference between the initial and
the final airfoil shape:

∆κ(l) = κm (x(l)) − κu (x(l)) (2.7)

where κu and κm are the curvature functions of the undeformed and
morphing airfoil, x(l) is the inverse of normalized arc length function
nal(x). According to Euler–Bernoulli beam theory, the maximum bending
stress along the skin can be computed from the curvature difference function
as:

Et
σbend = ∆κ(l) (2.8)
2
where E is the Young’s Modulus and t is the minimum skin thickness to
be assigned to the morphing skin.
The stress in the skin calculated in this way can be used as constraint in
the morphing shape optimization procedure.

A parametric identification is required to provide a CST identification of

preexisting CAD models. The detailed mathematical description of this
identification process is beyond the scope of this discussion on CST and it
can be found in [15]. The identified models are then ready to be modified
in order to introduce the morphing shape changes.

2.1.2 Optimization problem definition

This first level shape optimization aims at defining the best airfoil change
that satisfies specific mission requirements. As already mentioned, in
order to take into account the structural limitations of the skin, the
optimization must obtain the best morphing shape that is optimal according
to the aerodynamic performances, but at the same time assuring limited
deformation energy of the skin itself and limited actuation power required
for the shape change. Moreover, a Constant Cross–section Length (CCL)
strategy is adopted in order to limit the axial stresses in the skin. It
consists in minimizing the maximum axial stress along the skin in the
morphing configuration.
The aerodynamic models of the physically acceptable shapes are generated
by means of parametric meshing capabilities. The wing structural box con-
straint is implemented in an implicit way into the the shape optimization.
A possible formulation of the optimization problem is:
20 2.2. Multi-objective Genetic Algorithm

Minimize Cd (2.9)
such that ∆ψwing−box ≡ (CF S , CRS )
Cl ≥ C̄l
| ∆LLE,skin | = 0
max (∆κ(x)) ≤ ∆κ̄
The objective function to be minimized is the drag coefficient Cd . The
first constraint is the wing box constraint, which imposes that the wing
region between the front spar position CF S and the rear spar position CRS
isn’t affected by morphing changes. The second constraint assures that
the lift coeffient Cl is at least equal to C̄l . The third constraint doesn’t
allow any length variation of the morphing leading edge skin. The fourth
constraint limits the maximum curvature variation ∆κ(x) to a value ∆κ̄.
The optimization variables are the geometric parameters used in the CST
formulation to affect the camber morphing. A Genetic algorithm is used
to solve the optimization problem.
It is important to note that this optimization problem does not depend on
the material of the compliant mechanism that will be designed in the next
level.

What described until now represents an older version of the first level
optimization. After that the geometric representation tools have been
updated and today the state of the art consists in a tool able to directly
deals with three dimensional aerodynamic shapes. In other words, instead
of focusing on few airfoils and then interpolating between them in order
to obtain the 3D solution, it is possible to generate three dimensional
parametric shapes and use them in the optimization loop.
This upgrade represents a noteworthy improvement at this level of the
process, but doesn’t change the general meaning of the design procedure.

2.2 Multi-objective Genetic Algorithm

The second level optimization makes use of the target aerodynamic shape
that is the result of the morphing shape optimization problem described
in section 2.1. Given this shape, the current step aims at designing a
suitable internal compliant structure able to match once actuated the
desired morphing shape according to PHORMA.
The already described CST tool is useful also in the second level. It
provides the target curve, analytically imposes the structure internal
Chapter 2. Design procedure 21

point boundaries for the the optimization constraints, allows to find the
optimal position of the active output points and to compute the external
aerodynamic loads.
The main tool used in the second level is SPHERA (Synthesis of comPliant
mecHanisms for EngineeRing Applications) [13, 14]. Its purpose is the
design of morphing airfoils, taking into account the presence of the skin,
a structure and a mechanism, in such a way to satisfy the conflicting
requirements of deformability, load–carrying capability and low weight.
As already said, the approach here adopted for the synthesis of compliant
mechanisms is based on the distributed compliance concept.

2.2.1 Load path representation

SPHERA is based on the load path representation. It is a design parame-
terization of the stiffness tensor that makes use of load paths in order to
represent different structural topologies, cross–sectional beam areas and
the position of the points of interest for the shape problem. The presented
approach is based on beam element models and it is able to provide solu-
tions according to the distributed compliance concept. The approach is
intended for the synthesis of Single–Input Single–Output (SISO) compliant
mechanisms [19, 54] but it can be also used for the design of Single–Input
Multi–Output (SIMO) compliant mechanisms for the structural shape
control [35].
Load path representation application to morphing airfoils is suitable to
optimize a compliant structure and make it able to transfer the input
actuator force to a set of active output points along the boundary in order
to obtain a target shape.
Load paths are physical connection sequences between points. The three
types of points are input actuation, structure constraints and active output
points. The possible load paths connect respectively load input and active
output points (InOut paths), load input and constraint points (InSpc
paths), and constraint points and active output points (SpcOut paths).
Another type of characteristic points is that of the structure internal points
at the intermediate connections between the load paths.
An example of load path representation is shown in Figure 2.5
In the application of the load path representation to shape control problems,
the design variables are: path sequence (Seq), binary path existence variable
(Top), internal point coordinates (InterLoc) and cross sectional load path
22 2.2. Multi-objective Genetic Algorithm

Figure 2.5: A demonstrative connected load path representation of a SIMO

compliant mechanism

sizes (Dim), load path output destinations (Dest) and structure boundary
sizes (hBound).
Some points placed along the airfoil skin contour, more or equal than those
used as active output points, are used to minimize the Least Square Error
(LSE) between the deformed shape and the target shape obtained as result
of the morphing shape optimization level. The minimization problem is
subjected to many constraints: size constraints for the load path beam
elements and structure boundary elements, internal point boundaries, two
global connectivity equations, stress and buckling constraints and the
elastic equilibrium equation. For the analysis of the actuated mechanism,
each set of load path is transformed into a sequence of Finite Volume
Beams [21].
Non–linear analyses are performed to correctly describe the behavior of
mechanisms subject to large displacements; the equilibrium equation is
represented by the residual convergence. Moreover, the structural solver
includes modal, buckling, static linear, non–linear analysis modules.

2.2.2 Genetic Algorithms based on Multi-objective ap-

proach
The deformation of the airfoil skin depends on the topology of the compliant
mechanism and also on its dimensions, therefore topology and size design
must be simultaneously faced. SPHERA includes a customized Genetic
Optimizer where the individuals of the population are composed by mixed–
type design variables. The generation of each new population is produced
by selection, crossover and mutation strategies, that combine the topology
synthesis and the sizing optimization into the same process.
Chapter 2. Design procedure 23

A compliant mechanism must be able to satisfy the kinematic and structural

requirements, for all the load conditions corresponding to the considered
flight conditions. This is a multi–objective design problem that can be
efficiently incorporated into the genetic algorithm.
When multiple objectives are considered, the optimal solution is a trade–off
between the different objectives.
The selected approach to solve the multi–objective optimazion problem is
the so called Elitist Non-Dominated Sorting Genetic Algorithm (NSGA–
II) [16].

2.2.3 Optimization Problem

The design of morphing airfoil consists in the definition of a structure able

to transfer the deformation work from an input point (actuator) to the
output points placed along the skins. This is the kinematic requirement, but
also structural requirements are needed, in order to assure the fulfillment
of the undeformed or the deformed shape under the external loads. In
general, more than one load conditions corresponding to the undeformed
shape and more than one target shapes can be considered.
When SPHERA is applied to the design of morphing leading or trailing
edge wing section, two main objectives are considered:

• the minimization of the SE (Strain Energy) in a problem where the

input point is fixed and the external aerodynamic loads are those of
the structural requirement;

• the minimization of the LSE (Least Square Error) between the

deformed and the target shape, under the input actuation load
and the external aerodynamic loads corresponding to the kinematic
requirement.

The definition of the multi–objective optimization problem consists in two

sub–problems. The former is the kinematic design:
24 2.2. Multi-objective Genetic Algorithm

Minimize:
n
1X
q
LSE = (xd,i − xm,i )2 + (yd,i − ym,i )2 + (zd,i − zm,i )2 (2.10)
n i=1

such that:
Dimmin ≤ Dimi,j ≤ Dimmax
hBoundmin ≤ hBoundk ≤ hBoundmax
(xlow , ylow , zlow ) ≤ interCoord ≤ (xup , yup , zup )
X
T opi ≥ 1
i∈InOut
X
T opi ≥ 1
i∈InSpc∪pathSpcOut

σ ≤ σallowable
r=0

where n is the number of control points along the boundary, (xd , yd , zd )

are the grid positions in the deformed condition computed by the struc-
tural analysis and (xm , ym , zm ) the corresponding target shape points.
interCoord are the coordinates of the internal points. Dimi,j is the cross
sectional thickness of the j–th beam of the i–th load path. hBoundk is
the thickness of the k–th structure boundary. The optimization prob-
lem is subjected to size constraints for the load path beam elements
(Dimmin and Dimmax ) and the structure boundary elements (hBoundmin
and hBoundmax ) and to constraints for the internal point boundaries
((xlow , ylow , zlow ) and (xup , yup , zup )). Moreover, constraints on the T opi
variables, assure the connectivity of the structural topology. Finally there
are a stress limitation and the imposition of the elastic equilibrium equation.

The other problem is the structural design:

min SE = min uTout K(Ee )uout (2.11)
u, Ee u, Ee

such that:

K(Ee )uout = −fout , with SP Cin

Ee ∈ Eadmissible

where SE is the strain energy, uout is the nodal output displacement vector
due to the force −fout and K is the symmetric global stiffness matrix.
Chapter 2. Design procedure 25

It is important to underline that, according to the idea of the new optimiza-

tion procedure, the stress constraint is optional: it can be included in the
constraints during the optimization or it can be considered at the end of
the optimization by descarding from the Pareto front all the solutions sub-
jected to a maximum stress higher than a chosen threshold. Anyway, the
presence of a stress limitation doesn’t invalidate the material-independent
character which the new procedure assigns to the solution of the genetic
optimization. Indeed, the demand of a stress restriction can be thought
as the desire to avoid stress peaks and to uniformly distribute the stress
inside the entire structure, whatever is the adopted material. Later, the
actual admissible stress value will represent an effective constraint when
the actual material will be considered.

2.3 Solution refinement

Starting from the results of the two-level optimization procedure, an
improvement of the solution is aimed by implementing two further opti-
mization analyses, with an increasing level of detail. The description of
the refinement tools is the subject of chapter 4. Here only one matter
is discussed, a clarification about the term refinement. The developed
procedure is born with the purpose of further enhancing the optimal de-
sign of morphing devices previously performed by means of the existing
two-levels procedure. However it is capable of doing more than that and
the application examples prove this. Indeed, although the design phase has
been carried out using the usual isotropic aluminium material, after the
definition of the topology the adoption of a different material is possible
and this material can reveal itself even more promising. For these reasons
in the general framework the material of the first two phases has been
defined as fictitious. This fact opens a huge range of opportunities in
the applications. Moreover, as it will be shown, the adaptation of the
existing topology to a different scale is also possible successfully. This brief
foreword suggests the interpretation for the whole work.
Chapter 3

Droop nose

3.1 Reference wing

In the framework of EU funded Clean Sky 2 REG-IADP AG2 project,
some morphing concepts have been being developed. One of this concerns
the conceptual design of a morphing Leading Edge able to guarantee high
lift requirements as well as Natural Laminar Wing (NLF) flow.
The first design phase has consisted in the definition of the optimal shape
able to satisfy the aerodynamic requirements. Then a possible structural
solution for the skin, the internal compliant structure and an adequate ac-
tuation mechanism have been assessed. Thanks to the experience acquired
participating into SARISTU and NOVEMOR projects, the work carried
out inside CS2-AG2 consortium is trying to adopt the concept of a fully
compliant structure for the morphing leading–edge structure [50].
The reference aircraft addressed in the project is a 90 pax, twin prop
Regional Aircraft. The reference wing is the NLF optimized wing shown
in Figure 3.1, whose aerodynamic design has been provided by ONERA.
According to the initial requirements on the layout, the morphing Leading
Edge is composed by an inboard and an outboard region, as depicted in
Figure 3.2. The adoption of a morphing leading edge is forced by the
need to delay the stall of the wing during take-off and landing. In order
to achieve this goal a seamless and smoothed surface is required so that
no anticipated loss of laminarity occurs. This explains the reason why
classical slat can’t be employed.

The performance requirements consists in the enhacement of maximum lift

coefficient CLmax : 2.4% in take-off condition and 1.7% in landing condition.
These values must be obtained thanks to the considered leading edge

27
28 3.1. Reference wing

Figure 3.1: NLF optimized wing

Figure 3.2: Reference Aircraft with morphing leading edge

Chapter 3. Droop nose 29

morphing in cooperation with multi-functional flap designed by University

of Naples.
The work already done by Polimi has achieved a structural solution for
the leading edge compliant structure. The obtained results represents
the starting point for an example of application of the tools specifically
developed in this thesis work. The major aim of this application is the
validation of the optimization procedure, as well as the demonstration of its
potential benefits in terms of solution improvement. Further applications
will also show its versatility.

3.2 Topological synthesis results

The morphing shape obtained with PHORMA is used as target shape for
the topological synthesis of the compliant ribs. This synthesis is performed
by means of the multi-objective genetic algorithm optimization. In principle
the most critical flight conditions should be concurrently considered for
the design of the internal structure. However, in the preliminary design,
only three objective functions are considered:
1. Structural requirement: minimizing the Strain Energy (SE) to pre-
serve NLF wing shape. The morphing mechanism is kept fixed under
the aerodynamic loads, at M ach = 0.48, corresponding to the Dive
Speed at Sea Level;
2. Kinematic requirement: minimizing the Least Square Error (LSE)
between Target and Deformed LE shape. The DN mechanism is
deployed, under the external aerodynamic loads, at M ach = 0.197,
corresponding to the Landing flight condition at α = 10 ◦ , at Sea
Level.
3. Minimizing the maximum value of stress in the mechanism, when it
is deployed, under the aerodynamic loads of landing condition.
The aerodynamic loads corresponding to these requirements are shown
in Figure 3.3. The results of the genetic algorithm optimization were
evaluated looking at the Pareto Front shown in Figure 3.4, in order to
select a good compromise between the conflicting kinematic, structural
ans stress requirements. The red circle represents the chosen solution.
Concerning the kinematic requirement it is characterized by a LSE value
of 0.0065 m. The internal structure and its deformation once actuated are
shown in Figure 3.5.
Then some manual modifications have been applied, driven by the need of
30 3.2. Topological synthesis results

Figure 3.3: Aerodynamic loads for the structural (left) and kinematic (right)
requirements

109 Pareto front Pareto front

Maximum Internal Structure Stress [N/m ]

Strain Energy for NLF shape requirement [J]

3.0 0.025

2.5 0.02

2.0
0.015

1.5
0.01
1.0
0.005
0.5

0 0.003 0.004 0.005 0.006 0.007 0.008 0.009 0.01 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018
Least Square Error w.r.t. the Optimal Target Morphing Shape [m] Least Square Error w.r.t. the Optimal Target Morphing Shape [m]

(a) Maximum stress vs LSE (b) Strain energy vs LSE

Figure 3.4: Pareto Front from the genetic algorithm

0.2

0.15

0.1

0.05

-0.05

-0.1

-0.15

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45

Figure 3.5: Optimal selected solution

Chapter 3. Droop nose 31

less occupied space. The resulting solution is the starting point for all the
applications discussed hereafter and it is shown in Figure 3.6.

Figure 3.6: Updated selected solution

This model is made of beam elements, both for the skin and the mechanism.
In order to account for the effect of the skin stiffness due to its spanwise ex-
tension, the link between the presented two-dimensional model and the real
three-dimensional configuration is given by the following expression [13]:
r
wrib
tskin = 3 · tbeam (3.1)
prib
where tskin and tbeam are the thickness of 3D wing skins and the thickness
of the equivalent beams representing the skins in the 2D beam model used
during the genetic optimization, respectively; wrib and prib are the rib
thickness, in the span-wise direction, and the rib pitch.

The analysis of the model in the droop nose configuration shows a maxi-
mum stress in the mechanism equal to 550 MPa, exactly what expected
from the selection operated in the Pareto front of Figure 3.4. The choice of
accepting a level of stress so high, was dictated by the need of achieving the
16◦ droop angle, required to be close to the aerodynamic target. Obviously
32 3.3. Shape memory alloys

it is unthinkable to presume the availability of an aluminium alloy with

such a yield strength. This aspect is confirmed by the maximum principal
strain provided by the finite element analysis, equal to 0.0076, higher than
the elongation at yield of a typical aluminium alloy, which can be 0.0065.
This result establishes the unfeasibility of the aluminium alloy as material
of the compliant mechanism in the considered droop nose device. Moreover
it suggests the need to adopt materials able to guarantee high recoverable
strains, greater than 1%. This potentiality can be found in the superelas-
tic behaviour of the shape memory alloys, described in section 3.3, and
employed in the applications of chapter 5 as material of the mechanism, as
an alternative to the aluminium alloy.
Although preliminary computations have assessed the impossibility to
satisfactorily use the isotropic elastic material, in any case it will be
subjected to the proposed optimization procedure in order to evaluate the
result about it.

3.3 Shape memory alloys

Shape memory alloys are materials that exhibit the so-called shape memory
effect: once deformed, they are able to recover their original shape thanks
to a thermal cycle. This effect was firstly discovered by the swedish
researcher Arne Olander which observed it in a gold-cadmium alloy [26].
In the 1960s shape memory materials gained interest with the work of
Buehler and Wiley at the U.S. Naval Ordnance Laboratory [10]. The
research dealt with nickel titanium alloys, which took the name NiTiNOL
(Nickel Titanium Naval Ordnance Laboratory). Also other materials were
discovered to present the shape memory behaviour, for instance Cu-Al-Ni
and Cu-Zn-Al-Mn alloys, but until today Nitinol is the shape memory alloy
most adopted in the applications.
In addition to the shape memory effect, another peculiar property that
shape memory alloys do show is the so-called superelasticity. At relatively
high temperatures it can happen the recovery of large deformations as a
consequence of mechanical stress induced transformation. A representation
of the two effects is shown in Figure 3.7 [2].

A brief description of the phase transformations as well as the shape-

memory alloy micromechanics behind the material macroscopic behaviour
follows.
Shape memory alloys may exist in two state phases: austenite (parent
phase) and martensite (product phase). The transformation process be-
Chapter 3. Droop nose 33

Figure 3.7: Shape-memory effect (left) and superelastic behaviour (right)

tween the two phases is a solid-solid transformation. This kind of process is

non-diffusive, it is typical for those materials that present crystalline struc-
ture, and consists in small-length ordered displacements of the atoms [33].
In order to transform austenite in martensite, two possibilities there exist.
If the transformation is induced by a temperature decrease, twinned (or
multi-variant) martensite is obtained. Instead if the transformation is
induced by stress increase, detwinned (or single-variant) martensite is the
result. The difference between them concerns the crystallographic geometry
of the martensite.
The macroscopic behaviour of the shape memory alloys can be described in
terms of superelasticity and shape memory effect. Superelastic effect refers
to the recovery of very large strains upon unloading, without permanent
deformations. Shape memory effect is related to the ability of returning to
an initial shape after a temperature variation.
In order to understand the behaviour of shape memory alloys it is essential
to refer to their four characteristic transition temperatures, that are Mf ,
Ms , As and Af , in ascending order. M stands for martensite and A for
austenite. Subscripts f and s specify finish and start temperatures of the
transformation processes. Martensitic phase is stable at temperatures
below Mf . Austenitic phase is stable at temperatures higher than Af . For
temperatures between As and Ms , both phases are contemporary present.
At a temperature below Mf the application of stress induces the reorienta-
tion of the crystals forcing martensite twinned to transform into martensite
detwinned. Associated with this transformation there are residuals strains
that still remains after the removal of the load. In order to recover them,
shape memory effect is exploited: temperature is brought above Af , the ma-
terial transforms in austenite, then lowering the temperature single-variant
martensite in the original shape is obtained.
34 3.3. Shape memory alloys

The superelastic effect is due to the dependence of the transition tem-

peratures on the stress level. Superelasticity occurs when starting from
austenite, at a temperature above Af and kept constant, a load is applied
up to reach a critical stress value, that induces martensite transformation.
Continuing to increase the stress level the transformation finishes and then
the elastic loading of martensite starts. If the plastic limit is not reached,
the described transformation is reversible upon unloading and will return
to a zero strain condition. An initial austenite phase is therefore necessary
condition of the appearance of superelasticity. However if the tempera-
ture is higher than a value indicated as Md , the material will show an
elastic-plastic behaviour. This can be explained looking at Figure 3.8 [44].
Slip is an irreversible process. If the critical stress for slip at a certain
temperature is lower than the critical stress to induce martensite at the
same temperature, the superelastic behaviour won’t occur and irreversible
strains will remain.

Figure 3.8: Representation for the appearance of superelasticity in temperature-

stress space

A review of the kinematics and the kinetics approaches for the description
of the martensitic phase transformations occurring in shape-memory alloys
can be found in [2].
Chapter 3. Droop nose 35

3.3.1 Superelastic behaviour

Superelastic effect finds application in different fields, such as aerospace,
civil, mechanical and biomedical engineering. For instance it can be em-
ployed in medical guidewires, cardiovascular stents, orthodontic wires,
eyeglass frames. In the aerospace field many examples of shape memory
effect applications are available [4, 58]. Superelastic effect is less common,
but some applications can be found, as in [62] where nickel titanium su-
perelastic material is used in the manufacturing of a compliant mechanism.
The selected manufacturing process was the wire electrical discharge ma-
chining (wire-EDM). The choice of using nitinol was lead by its high strain
capability, that was confirmed by the strains measured in the experimental
tests.
In general, superelasticity-based applications exploit the possibility of
recovering large deformations and the existence of the transformation
stress plateau, which guarantees near constant stress over large strain
intervals [2].

3.3.2 Modeling and numerical simulations

The complexity of shape memory alloys make their modeling very difficult.
The fundamental characteristics that numerical simulation should be able
to reproduce are those of a very flexible material able to achieve very
large strains that are reversible upon unloading. At Hibbitt, Karlsson &
Sorensen (West) [47] a user material routine was developed on the basis
of the model proposed by Auricchio and Taylor [2]. Starting from the
generalized plasticity concept, the mathematical model is based on an
additive strain decomposition, where the strain is written as the sum of
two terms: a linear elastic component and a transformation component:

∆ = ∆el + ∆tr (3.2)

Concerning the transformation strain:

∂F
∆tr = a ∆ζ (3.3)
∂σ
where F is a transformation potential and ζ is the fraction of martensite.
They are related by a stress potential law:

∆ζ = f (σ, ζ)∆F (3.4)

If the stress overcomes the yielding of the material, also plastic strains
develop. The transformation strain is of the order of 6%. The elastic strain
36 3.3. Shape memory alloys

is smaller, usually below the 2%. Since the transformation strains are
greater than typical elastic strains in a metal, the material behaviour is
indicated as superelastic. This constitutive model is available in Abaqus as
UMAT user material. It is suitable to simulate the superelastic behavior
of shape memory alloys such as Nitinol.
The material data required to set up the numerical model are obtained
from uniaxial tests. These data characterize the start and the end of
the phase transformation during loading, unloading and reverse loading.
The different elastic constants for the austenite and martensite phases are
accounted for and temperature dependence is also included. The σ− curve
along with the significant parameters is shown in Figure 3.9. Figure 3.10

Figure 3.9: Uniaxial stress-strain curve for the superelastic behaviour

shows the dependence of critical stresses on temperature. An example of

σ −  curve coming from an experimental test is shown in Figure 3.11.
Concerning the user interface, in order to define the material properties, the
*USER MATERIAL option is required. Moreover, the name parameter of
the *MATERIAL must start with ABQ_SUPER_ELASTIC. The number
of solution-dependent state variables (SDVs) is provided via the *DEPVAR
option. The formulation requires 24 SDVs in the elastic case and 31 when
also plasticity is considered.
S
Different behavior in tension and compression can be obtained if σCL
S
is assigned different from σL . This is usually the case and has many
Chapter 3. Droop nose 37

Figure 3.10: Stress-temperature diagram for the superelastic behaviour

Figure 3.11: Experimental data for the uniaxial tension stress-strain response of
a Ni-Ti alloy [2]
38 3.3. Shape memory alloys

consequences, as widely depicted in [48]. For instance, the neutral axis may
deviate from the centerline of a symmetric section beam and plane sections
may not remain plane. However this asymmetric behaviour, though easily
intelligible, won’t be taken into account for the sake of simplicity.
In order to define the material model it is necessary to provide 15 + NA
material constants on the data lines of the *USER MATERIAL option,
where NA is the number of anneals that are performed. When plastic
behaviour is also considered, 16 + NA + 2NP material constants must be
specified, where NP is the number of stress-strain pairs needed for the
definition of the yield curve. Due to the aeronautical application, since no
permanent deformations are admitted in the operational cases, superelastic-
only behaviour will be considered while plasticity won’t be modelled.
Neither the possibility to simulate an annealing process will be used; it
consists in the resetting of the state variables to zero during an analysis
in order to provide a new unloaded configuration. This functionality
is of no interest in the present dissertation and will not be considered.
Table 3.1 describes the parameters required by the material routine for a
superelastic-only use.

Table 3.1: Superelasticity UMAT: list of parameters

Symbol Definition
EA Austenite Young’s modulus
νA Austenite Poisson’s ratio
EM Martensite Young’s modulus
νm Martensite Poisson’s ratio
L
Transformation strain
δσ
δT L
δσ/δT loading
σLS Start of transformation loading
σLE End of transformation loading
T0
Reference temperature
δσ
δT U
δσ/δT unloading
σUS Start of transformation unloading
σUE End of transformation unloading
S
σCL Start of transformation stress in compression
L
V Volumetric transformation strain

Many elements are supported to be used in conjunction with the Nitinol

material: 3D solids, plane strain, axisymmetric, plane stress, 3D shells,
3D membranes, 3D beams. A simple uniaxial loading and unloading of
Chapter 3. Droop nose 39

a beam-like body has been performed in order to compare the results

obtained using beam, shell and solid elements. The results do not show
significant differences among them, therefore in what follows the more
appropriate modeling stategy according to the situation will be preferred.
Concerning the analysis procedure, the material model can be used with
all the analysis procedures related to mechanical behavior, hence it is well
suitable for the non linear static analysis that will be performed by means
of the keyword *STATIC.
Many outputs are provided at the end of the analysis, among which
the linear elastic strains, the transformation strains and the fraction of
martensite.

3.3.3 Selection of material properties

In absence of detailed information about the material data for superelastic
applications, in this section a simple numerical design of the material
is carried out in the respect of what explained in the aforementioned
discussion on Nitinol. The purpose is to establish a set of parameters that
is able to give a material suitable for the specific application here addressed.
In particular, the dependence on temperature has to be taken into account
and the desired superelastic behaviour must be valid for the wide range of
aeronautical operative temperatures. The material coming from this study
will be used in the applications as the material of the internal compliant
mechanism, in alternative to the usual isotropic aluminum material, and
the comparison between them will be shown.
The phase transformations occur in temperature ranges depending on the
material composition. Transformation stresses increase linearly with the
temperature. The proportionality is described by means of the Clausius-
Clapeyron coefficients, typically supposed to be equal for both martensitic
and austenitic transformation and constant in the entire temperature
range [9]. These coefficients can assume values in the range between
2.5 MPa/K and 15 MPa/K [2]. According to [11], phase transformations
don’t depend on temperature rates and stress rates.
Some material parameters (elastic modulus, Poisson’s ratio, stress influence
coefficient, maximum transformation strain) are borrowed from Qidwai
and Lagoudas [46], and reported in Table 3.2.
Concerning the transformation temperatures, they can be widely shifted
after manufacturing by performing suitable solution annealing and aging
treatments [18]. Therefore all the temperatures are chosen according to
the desired useful operative range, and it is supposed that the selected
values are achievable. Further investigation, prototype fabrication and
40 3.3. Shape memory alloys

Table 3.2: Nitinol characteristics from [46]

Parameter Value
EA 72 GPa
νA 0.33
EM 30 GPa
νm 0.33
L
0.05
δσ
δT
L
7 MPa/K
δσ
δT U
7 MPa/K

testing are intended as future work.

It is expected that the material would show the superelastic behaviour
for a wide range of temperatures, here selected between −10 ◦ C and 40 ◦ C
(263 K and 313 K). This means that austenitic finish temperature must be
below the previous minimum temperature (263 K), hence a value of 260 K
is chosen. Another fundamental requirement is that up to the maximum
temperature (313 K) the stress at the end of the martensite transformation
is below the critical stress for slip, here assumed in the order of 800 MPa;
therefore a suitable martensite finish temperature may be 220 K.
The reference temperature considered in the analysis is 293 K. According
to this, the stress values characterizing the plateaus are shown in Table 3.3.
Figure 3.12 shows the relationship between the selected transformation
Table 3.3: Critical stresses at T0 = 293 K

Parameter Value
σLS 441 MPa
σLE 511 MPa
σUS 301 MPa
σUE 231 MPa
S
σCL 441 MPa

temperatures and stress level. In particular, the reference temperature and

the minimum and maximum temperatures are highlighted.
All the computations for the design involving Nitinol will be performed at
the reference temperature T0 = 293 K. After the design will be completed,
verification in terms of deformed shape and internal stress will be performed
also using the limit temperatures.
Chapter 3. Droop nose 41

108
8

5
(Pa)

0
220 240 260 280 300 320 340
T (K)

Figure 3.12: Reference and limit temperatures in the stress-temperature diagram

for the superelastic behaviour
Chapter 4

Solution refinement tools

This chapter describes the tools specifically developed in this thesis work
in order to improve the existing design procedure. These tools are not
simply intended for the refinement of the previously found solution. They
are conceived to specialize the field of application of the current solution,
to enhance its performance in terms of desired requirements and also to
get close to the manufacturing.
The main capabilities provided are:
• The possibility to adapt the internal mechanism synthesized topology
to:

– a different wing;
– different positions in the span of the same wing in case of
tapering;
– a scaled model of the same wing.

• The possibility to change the material both of skin and mechanism

even if the topology has already been defined. Several types of
material can be implemented, among which isotropic, composite,
superelastic.

• The availability of:

– different strategies of finite element modeling and levels of detail;

– different types of static analysis;
– different types of optimization analysis.

• The achievement of a solution almost ready for manufacturing process

and experimental testing.

43
44 4.1. Gradient-based optimization

All these potentialities can be allowed only by means of the combined use
of a numerical computing environment and a finite element analysis soft-
ware. The presented approach is very general and various implementative
techniques may be used. The detailed description of the above-mentioned
features will take place referring to the specific softwares employed in
this work, that are Matlab and Abaqus, both of them selected for their
respective powerful capabilities.

Section 4.1 describes the specialization of the topology to the actual

structural configuration. At this phase, a sizing and geometric optimization
is performed. Section 4.2 faces the problem of the transition from a beam-
elements model to the actual drawing of the designed device, performing a
shape optimization of the compliant mechanism aimed at the minimization
of local stresses.

4.1 Gradient-based optimization

4.1.1 Optimization problem definition
The problem that is intended to be solved in this phase can be formulated
as:

min LSE (4.1)

VM
such that ≤ σ̄
σmax
LSEN LF ≤ tolN LF
trib rib rib
LB ≤ ti ≤ tU B ∀i = 1, . . . , nrib
tskin skin
LB ≤ tj ≤ tskin
UB ∀j = 1, . . . , nskin
xLBk ≤ xk ≤ xU Bk ∀k = 1, . . . , nintP oints
yLBk ≤ yk ≤ yU Bk ∀k = 1, . . . , nintP oints

The scalar objective function is the Least Square Error (LSE) between
the deformed shape and the target shape when the mechanism is actuated,
computed as:
ncp q
1 X
LSE = (xd,i − xm,i )2 + (yd,i − ym,i )2 (4.2)
ncp i=1
where ncp is the number of control points along the skin, (xd , yd ) are the
positions of the control points in the deformed condition obtained by the
Chapter 4. Solution refinement tools 45

structural finite element analysis and (xm , ym ) the corresponding points

on the target shape.
The first constraint is the structural limitation related to the stress inside
VM
the mechanism. σmax is the maximum Von Mises stress in the compliant
rib and σ̄ is its admissible value. The evaluation of this constraint must
be done in the morphing configuration.
LSEN LF is the least square error between the undeformed shape and
the deformed shape computed under the aerodynamic loads of a critical
condition such as the Dive Speed, while keeping the mechanism fixed. This
function represents an index of the deformability of the morphing structure
when the device is not actuated. It must be restricted below a selected
tolerance. This requirement is needed to guarantee that the structure
is able to withstand the aerodynamic loads and that the aerodynamic
performances of the conventional configuration are not affected by the
presence of an internal compliant structure; the last point is especially
critical when the morphing device must be adopted in combination with a
Natural Laminar Flow wing.

The variables of the problem are:

• the in-plane dimensions trib

i of the beam segments of the compliant
mechanism;

• the thickness tskin

j of the sectors in which the skin is divided;

• the positions (xk , yk ) of the internal points of the compliant mecha-

nism.

For each variable, properly lower and upper bounds must be selected.
Subsection 4.1.3 discusses in details the models and the analyses used to
evaluate the objective function and the constraints of the optimization
problem, whose solution method is explained in subsection 4.1.2.

4.1.2 Gradient-based methods

Gradient-based methods are a family of methods that try to minimize or
maximize an objective function with searching directions suggested by the
gradient of the function at the current solution point. The optimization
problem of interest is set up in Matlab by means of the fmincon function,
able to deal with the minimization of constrained nonlinear multivariable
function [43].
46 4.1. Gradient-based optimization

The general formulation of the problem is:

min f (x) (4.3)

x

such that c(x) ≤ 0

ceq(x) = 0
A·x≤b
Aeq · x = beq
lb ≤ x ≤ ub
Among the available optimization algorithms, Sequential Quadratic Pro-
gramming (SQP) is chosen. At each major iteration of this algorithm, the
Hessian of the Lagrangian function is approximated using a quasi-Newton
updating method. Then a QP subproblem is generated and solved, giving
the search direction of a line search procedure [6].

4.1.3 Automatic model generation

The objective function and the constraints of the optimization problem are
evaluated as results of finite element analyses performed by Abaqus. The
genetic algorithm optimization discussed in section 2.2 used the PFEM
code as finite element solver. The model used to design the compliant
mechanism, whose results have been reported in 3.2, employed beam
elements both for the mechanism and the skin. In particular, the thickness
of the skin was modified in order to take into account its extension in the
wing-span direction.
Here a more accurate model for the skin is adopted. The length of the
skin in the span-wise direction is chosen equal to the rib pitch and a single
rib is modeled at the middle of this length. Concerning the finite element
modeling, shell elements are used for the skin and beam elements for the
compliant rib. Illustrative examples of the described model, both in case
of leading and trailing–edge devices, are shown in Figure 4.1.
The reference system has:
• x-axis parallel to the relative wind, directed from the leading to the
trailing edge;
• z-axis along the span, from the root to the tip of the left wing;
• y-axis such that a right-handed frame is obtained.
Concerning the analysis type, since large deformations are expected, static
non linear analysis is mandatory. Two different analyses are required:
Chapter 4. Solution refinement tools 47

Y
X
Z

X
Z

ZX

Figure 4.1: Abaqus models for the gradient-based optimization

48 4.1. Gradient-based optimization

1. analysis of the model when the compliant mechanism is actuated,

while it is subjected to the aerodynamic loads of the morphing
condition, for instance take-off or landing;
2. analysis of the model when the compliant mechanism is fixed, under
the aerodynamic loads of a critical flight condition, such as Cruise or
Dive Speed condition.
The model is the same in the two cases, apart from the loads and the
boundary conditions, as better described later.
As already explained, the variables of the optimization problem are the
section parameters of skin and rib, but also the position of the internal
points of the rib. Therefore each analysis required by the optimization
algorithm to evaluate the finite differences needed to compute the gradient
has to be performed on a different finite element model. This fact forces
the automatization of the model generation as well as its interaction with
the optimization solver. Abaqus script capabilities are exploited in order
to automatize the generation of the model. The Abaqus Scripting Interface
is an application programming interface (API) to the models and data
used by Abaqus [1]. It is an extension of the Python object-oriented
programming language and it works through Python scripts. The following
potentialities of the Interface will be used:
• create the components of an Abaqus model, such as parts, materials,
loads, and steps;
• create and submit Abaqus analysis jobs.
These operations are the counterpart of what is usually done by hand using
the CAE environment.
A dedicated Python script is written to realize the above-mentioned tasks.
It requires the variables of the optimization problem, but also some param-
eters and data that remain constant throughout the optimization process
and that depend on the particular morphing device that is intended to be
designed. The required parameters and data are:
• The position of the skin points that define the leading (or trailing)
edge airfoil; these points are used to create the skin model and also
as control points in which the target shape is known.
• The position of the input actuation point of the mechanism.
• A connectivity matrix to indicate the topology of the compliant
mechanism, starting from the identifiers of the internal points and
the active output points.
Chapter 4. Solution refinement tools 49

• The rib pitch prib .

• The rib thickness wrib .

• The characteristics of the materials of the rib and the skin.

• The dimensions of the stringers.

• The actuation force.

• The pressure coefficient distribution and the dynamic pressure for

the two aerodynamic configurations.

The sequence of actions performed by the script to implement the modeling

strategy is:

1. Create the Sketch of the rib and the Sketch of the skin.

2. Create a Wire Part for the rib and an Extruded Shell Part (with
depth equal to prib ) for the skin.

3. Create the sets required to assign the dimensions given by the sizing
variables of the problem.

4. Define the materials according to the provided data.

5. Define the rib sections by means of the related variables as well as

the rib thickness wrib ; define the skin sections by means of the related
thickness variables.

6. Create the stringers in the airfoil arc-length position corresponding

to the active output points, with an extension equal to the span-wise
length of the model.

7. Create a Tie Constraint to connect the rib Part to the skin Part at
the points in common.

8. Create a Non-linear analysis Static Step.

9. If the current model is intended for the morphing configuration, apply

the actuation force at the actuation point.

10. Apply the aerodynamic loads as Pressure Load providing the pressure
coefficient distribution and the dynamic pressure referred to the
considered aerodynamic condition.
50 4.1. Gradient-based optimization

11. Apply an encastre boundary condition to ground the skin where it

would connect to the structural box in correspondence of the spar.
If a morphing trailing–edge is under consideration, the lower surface
constraint is replaced with an horizontal slider.

12. If the current model is intended for the morphing configuration, apply
a planar constraint at the actuation point.

13. If the current model is intended for the configuration with fixed
mechanism, apply a pinned constraint at the actuation point.

14. Mesh the rib Part using three-dimensional Timoshenko beam linear
elements (B31); mesh the skin Part using 4-node, quadrilateral,
stress/displacement shell elements with large-strain formulation (S4).

15. Submit the Job and run the analysis.

This process is carried out every time an analysis run is invoked by the
optimization algorithm.
A simplified version of the optimization problem, and as a consequence
of the model generation, consists in the exclusion of the internal points
position from the optimization variables. This implies a lower computional
time at each iteration of the optimization: the search space is smaller and
also the required computations. In addition the model generation can be
speeded up by observing that now the geometry is kept fixed while only the
sizing quantities do change. Therefore the model can be generated once
for all, and the operation of changing the section parameters can be simply
performed modifying the INP file before each analysis. A consideration
about if it is worth implementing the more expensive approach will be
discussed in the results of chapter 5.

4.1.4 Optimizer-solver interface and problem set-up

In order to link the optimization process and the finite element solver an
appropriate interface between Matlab and Abaqus has been developed. It
mainly consists of Matlab functions in charge of writing and reading text
files. The fundamental tasks required are:

• pass the current values of the variables to the Python script;

• specify whether the morphing configuration or the conventional one

has to be analysed;
Chapter 4. Solution refinement tools 51

• read the outputs from the finite element analysis results. They
are the maximum Von Mises stress inside the mechanism and the
displacement of the skin points.
In order to set-up the optimization problem and run it in Matlab, the
following input data must be supplied:
• maximum admissible stress in the mechanism;

• tolerance for the assessment of deformability;

• bounds for the sizing variables;

• bounds for the internal points position;

• initial solution.
At the end of each iteration, the current solution is saved as well as the
corresponding model and the objective function value.

4.2 Compliant mechanism shape optimization

At the end of the previous step, the numerical design phase may have been
considered finished since an optimized model has been obtained such that
it well satisfies the aerodynamic performances and at same time doesn’t
violate the structural requirements. However something is still missing,
since another step is required to be ready for the subsequent manufac-
turing process of the designed device. Indeed, the compliant mechanism
model addressed until now consists of beam elements. They guarantee a
sufficiently accurate estimate of the stress far enough from the element
extremities, but they lack of reliability concerning what happens at the
constrained regions. Moreover in this zones, the manufacturing detail of
the fillets is missing, since it can’t be thought to be a simple intersection
in a single point. These premises constitute the reason behind the idea of
performing the optimization analysis described in the following.

In this section, only the internal compliant mechanism is dealt with. First
of all, starting from the beam elements model of the optimal solution,
a CAD model of the compliant mechanism is created. In order to do
that, the internal and the boundary points of the mechanism of the beam
elements model are sketched in the 2D drawing and are connected with
lines according to the topology. Then, each line is used as mid line of
the corresponding load path, whose in-plane thickness is inherited from
52 4.2. Compliant mechanism shape optimization

the beam elements and is represented in the drawing by means of two

symmetric offsets. Finally, suitable fillets link the lines at their intersection.
Later, starting from the CAD model a plane sketch is generated, that
exhibits the actual internal points positions and the load path thicknesses
of the optimal solution.

In this optimization phase, the selected modeling software is Abaqus again.

The created sketch is imported in Abaqus and it is finely meshed with
two-dimensional, 4-node bilinear, plane stress solid elements (CPS4). The
described sequence is shown in Figure 4.2.

The optimization analysis is implemented by means of a suite of opti-

mization tools, named Tosca Structure. This is a modular system for
non-parametric structural optimization able to perform different kinds of
optimization analyses coupled with finite element solvers. The integration
with CAE environments is straightforward. The optimization process is
simulation-based and allows to improve the design in an iterative manner.
Among the different tools of optimization offered, Tosca.shape is the one
that best fits to the desired purpose. Indeed, the shape optimization
consists in finding the optimal position of the nodes for the minimization
or maximization of a chosen objective function, under some constraints.
Shape optimization is typically adopted to modify the component surface
in order to reduce local stress: Tosca.shape is a tool specifically developed
to improve designs for more reliability and durability. The setup of shape
optimization problem doesn’t require model parameterization; the design
area is defined by specifying groups.

According to the previosly described aim of stress reduction, a suitable

objective function is the minimization of a stress index inside the structure.
In this work the selected objective function is the minimization of the
maximum principal stress. Concering the constraints, the volume is not
allowed to change. This is a suggested equality constraint that must be
satisfied within a tolerance of 10−3 . Another constraint that is needed in
order to properly lead the analysis is the minimum member size, belonging
to the family of the geometric restrictions; the minimum distance between
two geometric edges is not allowed to go below a chosen value. The
design region, which is modified by the optimization process, includes the
surface nodes of the whole model or of a subset of it. In this case of two-
dimensional model, the considered nodes are those that lie on the contour
of the mechanism. However, also the internal nodes are allowed to move;
indeed, a powerful capability provided by Tosca is the mesh smoothing
Chapter 4. Solution refinement tools 53

(a) Beam elements model (b) 2D drawing

(c) Plane sketch (d) 2D solid elements model

Figure 4.2: Steps from the beam elements model to the 2D solid elements model
54 4.2. Compliant mechanism shape optimization

in each design cycle in order to keep high quality meshes throughout the
optimization process. In other words, the internal nodes displacements
are not design variables but they are moved as a consequence of the outer
nodes displacement, with the only aim of avoiding the creation of too
distorted elements.
An example of nodes belonging to the design area is shown in Figure 4.3.

Figure 4.3: Shape optimization: example of design area

The static analysis that is performed in order to evaluate the objective

function and the constraints for the optimization analysis consists in
imposing the displacements and the in-plane rotation at the points of the
rib attached to the skin and applying the actuation force at the actuation
point. The displacement history to be imposed is extracted from the
outputs of the beams model analysis for the optimum solution of the
gradient-based optimization. The major issue faced in performing this
kind of analysis comes from the nature of the model, since there are no
simple lines and points, but a mesh of solid elements. This means that
in order to correctly apply the boundary conditions and the load, some
reference points have to be defined. In particular, they are created in
correspondence of the external points of the rib and also in the actuation
point location. Then they are linked to the rest of the structure by means
of the Kinematic Coupling constraint provided by Abaqus. Each reference
Chapter 4. Solution refinement tools 55

point is selected as master point in the connection with the contour of

the mechanism in the outer region. An illustrative representation of this
interface is shown in Figure 4.4.

RP−4

RP−3

RP−2
Y

Z X

Figure 4.4: Kinematic coupling for the connection of reference points to boundary
edges
Chapter 5

Results

In this chapter the refinement process is applied to the results coming from
the genetic algorithm optimization shown in section 3.2. Starting from
this model, an optimized topology and sizing solution is found by means
of the gradient-based optimization performed with the combined used of
Matlab and Abaqus. Then, the new optimal solution is transposed in a
CAD model used to generate the input for the shape optimization of the
compliant mechanism performed by Tosca.
The whole procedure is applied twice, for different materials of the mecha-
nism: the aluminium alloy isotropic material exactly as in the GA topology
optimization, and the Nitinol whose characteristics have been discussed in
section 3.3.3. Then a comparison between the two sets of results is shown.

5.1 Gradient-based optimization

First of all, the common features of the model and the set-up of the
optimization problem are discussed.
• The rib pitch prib is equal to 130 mm and it is used as span-wise
length of the skin.
• The rib thickness wrib is equal to 35 mm.
• The chord of the airfoil is c = 2.911 m. Front spar is located at the
16% of the chord, hence the chord extension of the model is 467 mm.
• The actuation force F in the input point has vector components
(−468.21 N, −40.96 N).
• The aerodynamic conditions used for the evaluation of the objective
function and the NLF constraint are respectively:

57
58 5.1. Gradient-based optimization

– Landing condition at sea level: angle of attack α = 10◦ , dynamic

pressure qD = 2561 Pa;
– Dive speed condition at sea level: angle of attack α = 0◦ ,
dynamic pressure qD = 5000 Pa.

The pressure coefficient distribution is shown in Figure 5.1.

The model corresponding to the inital solution is depicted in Figure 3.6.

It shows the starting design variables, that are the result of the genetic
algorithm. Figure 5.3 shows the application of the pressure loads on the
model in the two different configurations.

The other required data to set-up the problem are the limit values for the
constraints and the bounds. For the NLF LSE tolerance a value of 10−3 m
is selected.
Concerning the admissible values in terms of stress, they are 550 MPa for
the isotropic material and 500 MPa for Nitinol. The value of 550 MPa for
the aluminium alloy has already been discussed in section 3.2. The value of
500 MPa for Nitinol is just below the stress level at the end of the loading
plateau, for the selected material parameters and reference temperature,
as discussed in section 3.3.3.
The bounds for the sizing variables are 1 mm ≤ trib
i ≤ 7 mm and 0.5 mm ≤
tskin
j ≤ 7 mm. The internal points position can vary of ±5 mm both in x
and y direction.
Chapter 5. Results 59

-3
cp upper side
c lower side
-2 p

-1

2
0 0.2 0.4 0.6 0.8 1
x/c
(a) Landing condition (α = 10◦ )

-1
cp upper side
c lower side
p
-0.5

0.5

1.5
0 0.2 0.4 0.6 0.8 1
x/c
(b) Dive speed condition (α = 0◦ )

Figure 5.1: Pressure coefficient distributions

60 5.1. Gradient-based optimization

Z X

(a)

Y
X
Z

(b)

Figure 5.2: Abaqus model corresponding to the initial variables

Chapter 5. Results 61

Z X

Z X

(a) Landing condition (α = 10◦ )

Z X

Z X

(b) Dive speed condition (α = 0◦ )

Figure 5.3: Pressure loads on the model

62 5.1. Gradient-based optimization

5.1.1 Isotropic material

The isotropic material is a typical aluminium alloy, characterized by a
Young’s modulus E = 72 GPa and a Poisson’s coefficient ν = 0.33.
Concerning the skin, it is made of composite materials, but here an
equivalent isotropic material is selected, as already done with the genetic
optimizer. Its characteristics are Eskin = 40 GPa and ν = 0.12.
At first, a static analysis in the droop nose configuration is performed
to evaluate the LSE and the maximum internal stress for the starting
solution. The deformed shape is shown in Figure 5.4. Its comparison with
the undeformed shape and the target shape is represented in Figure 5.5.
The obtained morphed shape is not so good compared to the target one,
especially in the lower skin region. The deformed shape is characterized by
an LSE value of 10.19 mm. The different value with respect to that given
by the genetic algorithm optimization can be explained with the different
modeling strategy. The maximum Von Mises stress in the mechanism
is 541 MPa, while LSEN LF = 0.4 mm. Therefore the initial solution is
feasible.

The gradient-based optimization is run and it converges to an optimal

solution in 50 iterations. The evolution of the objective function with the
iteration number is shown in Figure 5.6. The LSE for the optimal solution
is 2.23 mm. Maximum stress in the rib is 549.9 MPa.
The comparison between the initial deformed shape and the optimal de-
formed shape is shown in Figure 5.7. The stress level inside the mechanism
is shown in Figure 5.8.
Figure 5.9 illustrates the comparison between the optimal deformed shape
and the target one.
Figure 5.10 compares the optimization variables between the starting
solution and the optimal solution. For a more clear visualization, thickness
values are magnified by a factor 3.
Chapter 5. Results 63

Z X

Z X

Figure 5.4: Isotropic material: deformed shape of the initial solution

64 5.1. Gradient-based optimization

Z X

Figure 5.5: Isotropic material: initial solution undeformed/deformed shape and

target shape

-3
10
12

10
LSE (m)

2
0 10 20 30 40 50
Iteration
Figure 5.6: Isotropic material: LSE vs iteration
Chapter 5. Results 65

Z X

Figure 5.7: Isotropic material: initial solution and optimal solution deformed
shapes
66 5.1. Gradient-based optimization

S, Mises
Bottom Left Corner
(Avg: 75%)
+5.499e+08
+5.041e+08
+4.582e+08
+4.124e+08
+3.666e+08
+3.208e+08
+2.750e+08
+2.292e+08
+1.834e+08
+1.376e+08
+9.175e+07
+4.594e+07
+1.261e+05

Z X

Figure 5.8: Isotropic material: Von Mises stress in the rib

Chapter 5. Results 67

0.25
Deformed
0.2 Target

0.15

0.1

0.05
y (m)

-0.05

-0.1

-0.15

-0.2
0 0.1 0.2 0.3 0.4 0.5
x (m)

Figure 5.9: Isotropic material: optimal solution deformed shape and target shape
68 5.1. Gradient-based optimization

Z X

(a) Initial solution

Z X

(b) Optimal solution

Figure 5.10: Isotropic material: thicknesses

Chapter 5. Results 69

The results show that the gradient-based optimization, coupling Matlab

optimizer and Abaqus finite element solver, is effective in performing the
improvement of the kinematic requirement. Indeed, a meaningful reduction
of the LSE between the optimal deformed shape and the target shape,
with respect to the initial deformed shape, has been achieved. The LSE
has been decreased from 10.19 mm to 2.23 mm. The comparison between
the initial deformed shape and the optimal deformed shape shows the
greater droop angle that characterizes the optimal solution and also its
smoother skin surface. Right now, the achieved deformed solution shows a
better agreement with the target morphing shape.
Concerning the comparison related to the sizing variables, it can be noted
that in order to minimize the objective function with respect to the initial
solution, a different arrangement of the internal mechanism in-plane thick-
nesses is required to allow the desired improvement. Moreover, also the
internal points positions are slightly changed.
Concerning the stress limitation, a maximum stress value equal to 549.9 MPa
indicates that stress constraint is satisfied within the optimization analysis.
However it must be remembered that the selected admissible value is
overestimated.
Therefore, the presented results have a purely numerical relevance. They
demonstrate the strength of the developed optimization procedure, however
the obtained optimal solution can’t be considered feasible, in the absence
of elastic materials with a sufficiently large elastic range.

In addition to the reported analysis, still from the numerical validation

point of view, an optimization analysis without including the position of
the internal points has also been performed. This corresponds to a classical
sizing optimization. It has got rapidly to convergence, with a minimum
LSE value equal to 2.73 mm, slightly higher than the optimum previously
obtained. This means that the increased computational cost due to the
higher number of variables when the position variation is allowed, may not
assure an advantage and hence it would be worthless, at least when the
adopted material is the same of the genetic algorithm optimization.
The LSE optimization history is shown in Figure 5.11.
70 5.1. Gradient-based optimization

10-3
12

10
LSE (m)

2
0 10 20 30 40 50
Iteration
Figure 5.11: Isotropic material: LSE vs iteration (sizing variables only)
Chapter 5. Results 71

5.1.2 Nitinol material

Nitinol behaviour and its properties have been widely described in sec-
tion 3.3. Here this material is adopted in the same optimization analysis
previously performed with the isotropic material.
At first, a static analysis in the droop nose configuration is considered to
evaluate the LSE and the maximum internal stress for the starting solution.
The deformed shape is shown in Figure 5.12. Its comparison with the
undeformed shape and the target shape is represented in Figure 5.13. The
deformed shape is characterized by an LSE value of 10.22 mm. Therefore
the adoption of a different material with the topology designed using the
isotropic material gives no appreciable difference concerning the initial solu-
tion. The maximum Von Mises stress in the mechanism is 448 MPa, while
LSEN LF = 0.4 mm. Therefore the initial solution is feasible. Moreover,
the maximum stress is less than that obtained for the isotropic material.
This stress reduction, with the same global deformation, is one of the
reasons that push in the direction of the use of superelastic materials in
the design of compliant structures.
The gradient-based optimization is run and it converges to an optimal
solution in 49 iterations. The evolution of the objective function with the
iteration number is shown in Figure 5.14. The LSE for the optimal solution
is 1.44 mm. Maximum stress in the rib is 462 MPa. The comparison
between the initial deformed shape and the optimal deformed shape is
shown in Figure 5.15. The stress level inside the mechanism is shown in
Figure 5.16.
Figure 5.17 illustrates the comparison between the optimal deformed shape
and the target one.
Figure 5.18 compares the optimization variables between the starting
solution and the optimal solution. For a more clear visualization, thickness
values are magnified by a factor 3.
72 5.1. Gradient-based optimization

Z X

Z X

Figure 5.12: Nitinol: deformed shape of the initial solution

Chapter 5. Results 73

Z X

Figure 5.13: Nitinol: initial solution undeformed/deformed shape and target

shape

10-3

10

7
LSE (m)

1
0 10 20 30 40 50
Iteration

Figure 5.14: Nitinol: LSE vs iteration

74 5.1. Gradient-based optimization

Z X

Figure 5.15: Nitinol: initial solution and optimal solution deformed shapes
Chapter 5. Results 75

S, Mises
Bottom Left Corner
(Avg: 75%)
+4.620e+08
+4.236e+08
+3.851e+08
+3.467e+08
+3.082e+08
+2.698e+08
+2.313e+08
+1.929e+08
+1.544e+08
+1.160e+08
+7.752e+07
+3.907e+07
+6.192e+05

Z X

Figure 5.16: Nitinol: Von Mises stress in the rib

76 5.1. Gradient-based optimization

0.25
Deformed
0.2 Target

0.15

0.1

0.05
y (m)

-0.05

-0.1

-0.15

-0.2
0 0.1 0.2 0.3 0.4 0.5
x (m)

Figure 5.17: Nitinol: optimal solution deformed shape and target shape
Chapter 5. Results 77

Z X

(a) Initial solution

Z X

(b) Optimal solution

Figure 5.18: Nitinol: thicknesses

78 5.1. Gradient-based optimization

The results show a significative LSE reduction, from 10.22 mm to 1.44 mm,
even better than that obtained with the isotropic material (2.23 mm).
Now the agreement between the optimal deformed shape and the target
shape is almost perfect. It must be observed that the LSE reduction is
again associated with a rearrangement of the thickness distribution inside
the compliant mechanism.
The maximum stress in the internal mechanism is 462 MPa. The corre-
sponding maximum strain is 0.018, which is in the range of the recoverable
strains. These results reveal that:

• when dealing with compliant structures large strains are expected,

hence materials able to provide them are preferred;
• the existence of the stress plateau allows to limit the stress values in
the structure;
• the possibility to exploit a large range of deformations allows to
achieve a meaningful improvement of the kinematic requirement,
more than that a conventional elastic material can provide.
Also in combination with the adoption of Nitinol, an optimization analysis
without including the position of the internal points among the optimiza-
tion variables has been performed. The LSE optimization history is shown
in Figure 5.19. The optimization analysis rapidly gets to convergence,
however the minimum LSE value is equal to 4.18 mm, significantly worse
than the previously obtained optimum value. This means that adapting
an existing topology to a different material requires the possibility of
slightly moving the internal points of the mechanism, otherwise a sufficient
improvement of the objective can’t be achieve.

To sum up, the use of Nitinol in combination with the more advance version
of the gradient-based optimization here discussed, guarantees a meaningful
enhancement of the existing design that uses isotropic material.
Moreover, also the results of section 5.2 will show that Nitinol not only
improves the morphing solution from the aerodynamic performance point
of view, it may assure the structural feasibility that the isotropic material
isn’t able to guarantee from the strength requirement point of view.

Finally, the comparison between the optimal isotropic material solution

and the optimal Nitinol solution in terms of optimization variables and
deformed shape is shown in Figure 5.20 and Figure 5.21, while Figure 5.22
shows the topological differences between the two optimal solutions.
Chapter 5. Results 79

10-3
11

10

9
LSE (m)

4
0 5 10 15 20
Iteration

Figure 5.19: Nitinol: LSE vs iteration (sizing variables only)

Apart from the considerations about the kinematic and structural require-
ments that have been already discussed, here the focus is on the values
which the optimization variables assume for the two materials. It is evident,
as expected, that different materials are characterized by different values
of optimization variables that minimize the objective function.
80 5.1. Gradient-based optimization

Z X

(a) Isotropic material

Z X

(b) Nitinol

Figure 5.20: Optimal solutions: thicknesses

Chapter 5. Results 81

Z X

Figure 5.21: Optimal solutions: isotropic material and nitinol deformed shapes
82 5.1. Gradient-based optimization

Z X

Figure 5.22: Optimal solutions: isotropic material and nitinol undeformed shapes
Chapter 5. Results 83

5.2 Compliant mechanism shape optimization

In this section the results of the shape optimization of the compliant
mechanism are presented and discussed. The procedure is applied to
the optimal solutions previously found by means of the gradient-based
optimization, both in case of isotropic material and Nitinol, leading to
different results and conclusions.
Starting from the beam model of the optimal solution, a CAD model
is generated. It represents the optimal internal compliant mechanism,
featuring the actual internal points position and the thicknesses of the
optimal solution. Moreover adequate hypothesis for the fillets radiuses are
done. Their improvement is one of the aim of the current optimization
step.
The sketch is imported in Abaqus and a two-dimensional solid mesh is
generated. At first a static analysis is performed in order to observe the
behaviour of the mechanism and to identify the most critical regions. Then
the same static analysis is included in the optimization analysis in order
to evaluate the objective function and the constraints.

5.2.1 Isotropic material

The two-dimensional solid elements model of the compliant mechanism
made of isotropic material is shown in Figure 5.23.
The result of the static analysis with imposed displacements at the bound-
ary is shown in Figure 5.24. In addition to the global deformation, the
most stressed regions are highlighted. It can be noted a maximum Von
Mises stress of 1200 MPa, higher than the value obtained from the beam
model. This can be explained by the fact that a beam model can’t give an
accurate estimate of the stress in singular points like its extremities or the
intersections. The aim of the shape optimization is to try to reduce these
localized stress peaks.
Shape optimization is firstly applied to the region depicted in Figure 5.24(c)
where the maximum stress is 877 MPa. The design area is shown in Fig-
ure 5.25. The objective function is the minimization of the maximum
principal stress evaluated in the surroundings of the optimization region.
The optimization performs ten iterations, providing a redistribution of the
stress and a reduction of its maximum peak. The comparison between the
initial solution and the optimized solution is shown in Figure 5.26. The
stress is decreased at a value of 598 MPa, however it is still too high for an
aluminium material. A similar optimization procedure is carried out on
the region of Figure 5.24(b), but failing.
84 5.2. Compliant mechanism shape optimization

Z X

Figure 5.23: Isotropic material: 2D model of the compliant mechanism

Chapter 5. Results 85

Z X

(a)

S, Mises
(Avg: 75%)
+8.772e+02
+8.041e+02
+7.310e+02
S, Mises +6.579e+02
(Avg: 75%) +5.848e+02
+5.117e+02
+1.202e+03 +4.386e+02
+1.102e+03 +3.656e+02
+1.002e+03 +2.925e+02
+9.016e+02 +2.194e+02
+8.014e+02 +1.463e+02
+7.013e+02 +7.321e+01
+6.011e+02 +1.269e−01
+5.009e+02
+4.007e+02
+3.005e+02
+2.004e+02
+1.002e+02
+1.202e−07

Y Y

Z X Z X

(b) (c)

Figure 5.24: Isotropic material: static analysis of the compliant mechanism

86 5.2. Compliant mechanism shape optimization

Z X

Figure 5.25: Isotropic material: optimization region

S, Max. Principal (Abs) S, Max. Principal (Abs)

(Avg: 75%) (Avg: 75%)
+9.249e+02 +5.895e+02
+7.812e+02 +4.951e+02
+6.374e+02 +4.006e+02
+4.937e+02 +3.061e+02
+3.500e+02 +2.116e+02
+2.063e+02 +1.172e+02
+6.258e+01 +2.268e+01
−8.114e+01 −7.180e+01
−2.249e+02 −1.663e+02
−3.686e+02 −2.607e+02
−5.123e+02 −3.552e+02
−6.560e+02 −4.497e+02
−7.997e+02 −5.442e+02

Y Y

Z X Z X

(a) Initial solution (b) Optimum solution

Figure 5.26: Isotropic material: shape optimization

Chapter 5. Results 87

Therefore the first application demonstrates that the procedure is able to

modify the junction region in such a way to reduce the stress, however
it can’t overcome the intrinsic limitation of using a conventional linear
material for compliant mechanism devices.
88 5.2. Compliant mechanism shape optimization

5.2.2 Nitinol material

The two-dimensional solid elements model of the compliant mechanism
made of Nitinol is shown in Figure 5.27.

Z X

Figure 5.27: Nitinol: 2D model of the compliant mechanism

The result of the static analysis with imposed displacements at the bound-
ary is shown in Figure 5.28. In addition to the global deformation, the
most stressed regions are highlighted, the same of the previous case. Also
in case of Nitinol the adoption of a more refined finite element model shows
stress concentrations, hence shape optimization is needed.
Shape optimization is firstly applied to the region depicted in Figure 5.28(c)
where the maximum stress is 549 MPa. The objective function is the
minimization of the maximum principal stress evaluated in the surroundings
of the optimization region. After eight iterations the maximum stress value
is reduced to 525 MPa. The comparison between the initial solution and
the optimized solution is shown in Figure 5.29.
Chapter 5. Results 89

Z X

(a)

S, Max. In−Plane Principal (Abs) S, Max. Principal (Abs)

(Avg: 75%) (Avg: 75%)
+1.052e+03 +5.485e+02
+8.760e+02 +4.583e+02
+6.999e+02 +3.681e+02
+5.237e+02 +2.779e+02
+3.476e+02 +1.878e+02
+1.714e+02 +9.760e+01
−4.714e+00 +7.430e+00
−1.809e+02 −8.274e+01
−3.570e+02 −1.729e+02
−5.332e+02 −2.631e+02
−7.093e+02 −3.533e+02
−8.855e+02 −4.434e+02
−1.062e+03 −5.336e+02

Y Y

Z X Z X

(b) (c)

Figure 5.28: Nitinol: static analysis of the compliant mechanism

90 5.2. Compliant mechanism shape optimization

S, Max. Principal (Abs)

(Avg: 75%)
+5.485e+02
+4.583e+02
+3.681e+02
+2.779e+02
+1.878e+02
+9.760e+01
+7.430e+00
−8.274e+01
−1.729e+02
−2.631e+02
−3.533e+02
−4.434e+02
−5.336e+02

Z X

(a) Initial solution

S, Max. Principal (Abs)

(Avg: 75%)
+5.249e+02
+4.378e+02
+3.506e+02
+2.635e+02
+1.763e+02
+8.916e+01
+2.004e+00
−8.515e+01
−1.723e+02
−2.595e+02
−3.466e+02
−4.338e+02
−5.209e+02

Z X

(b) Optimum solution

Figure 5.29: Nitinol: shape optimization for the right region

Chapter 5. Results 91

It can be seen that the redistribution of stress is more effective when a

superelastic material is used. Moreover, the transformation strains are
significantly reduced, from a maximum value of 0.0427 to a maximum
value of 0.0157. The stress-strain relationship for the loading history in
the two cases is reported in Figure 5.30.
A similar optimization procedure is carried out on the region of Fig-
ure 5.28(b). After seven iterations the maximum stress value is reduced to
541 MPa. The comparison between the initial solution and the optimized
solution is shown in Figure 5.31.
Also in this case the shape optimization achieves the stress reduction.
Concerning the strains, Figure 5.32 shows that the initial solution is not
acceptable since the plateau is fully exploited and then the stress reaches the
critical value for slip: permanent deformations will remain upon unloading.
Differently, the optimized solution shows an acceptable behaviour: the
plateau is not fully exploited and deformations can be recovered. Therefore
shape optimization is crucial in the achievement of a feasible structural
solution, as well as the adoption of a superelastic material such as Nitinol.
Obviously the situation must be examined more deeply, manufacturing
and testing are mandatory, however these preliminary numerical results
give confidence to the developed optimization procedure and its tools, as
well as the introduction of the superelastic material.
Moreover, also from a numerical point of view a verification is needed, in
order to assess if the shape optimization phase decreases the performance
in terms of LSE with respect to what has been estimated by the gradient-
based optimization.
92 5.2. Compliant mechanism shape optimization

600

500

400
(MPa)

300

200

100

0
0 0.01 0.02 0.03 0.04 0.05

(a) Initial solution

600

500

400
(MPa)

300

200

100

0
0 0.01 0.02 0.03 0.04 0.05

(b) Optimum solution

Figure 5.30: Nitinol: stress-strain curve in the right region

Chapter 5. Results 93

S, Max. Principal (Abs)

(Avg: 75%)
+1.052e+03
+8.760e+02
+6.999e+02
+5.237e+02
+3.476e+02
+1.714e+02
−4.714e+00
−1.809e+02
−3.570e+02
−5.332e+02
−7.093e+02
−8.855e+02
−1.062e+03
Max: +1.052e+03
Elem: PART−1−1.28430
Node: 14162

Y
Max: +1.052e+003

Z X

(a) Initial solution

S, Max. Principal (Abs)

(Avg: 75%)
+5.413e+02
+4.440e+02
+3.467e+02
+2.495e+02
+1.522e+02
+5.495e+01
−4.231e+01
−1.396e+02
−2.368e+02
−3.341e+02
−4.314e+02
−5.286e+02
−6.259e+02
Max: +5.413e+02
Elem: PART−1−1.13664
Node: 29493

Y Max: +5.413e+002

Z X

(b) Optimum solution

Figure 5.31: Nitinol: shape optimization for the left region

94 5.2. Compliant mechanism shape optimization

1000

800
(MPa)

600

400

200

0
0 0.02 0.04 0.06 0.08

(a) Initial solution

1000

800
(MPa)

600

400

200

0
0 0.02 0.04 0.06 0.08

(b) Optimum solution

Figure 5.32: Nitinol: stress-strain curve in the left region

Chapter 5. Results 95

5.3 Considerations about the use of Nitinol

The results of sections 5.1 and 5.2 have shown the comparison between
Nitinol and the isotropic aluminum alloy as material for the internal
compliant mechanism, highlighting the advantages of the adoption of the
superelastic material, both in terms of aerodynamic requirements (LSE)
and structural feasibility. Anyway, it must be remembered the dependence
of superelastic behaviour from the transition temperatures in relation with
the application temperature. The previously shown results were referred
to a design temperature T0 = 293 K, but the choice of the parameters was
done aiming to achieve a wide range of possible application temperatures.
In this section the optimal solution of the gradient-based optimization
using Nitinol, is verified at the limit temperatures, Tmin = 263 K and
Tmax = 313 K. Once the optimal solution has been obtained, a kinematic
chain able to guarantee the desired displacement of the actuation point
must be designed. Therefore, analyses with imposed displacements at
the actuation point are performed. The displacement history is extracted
from the results obtained at the design temperature. The results of these
analyses in terms of deformed shape are shown in Figure 5.33. Table 5.1
compares the achieved LSE values at the different temperatures. Moreover
it reports the magnitude of the actuation force required to accomplish the
imposed displacement in each case.

Table 5.1: Nitinol: optimal solution at different temperatures

Temperature (K) LSE (mm) Actuation Force (N)

263 1.77 422
293 1.44 470
313 1.57 498

As expected, the smaller LSE is obtained at the design temperature, but

the values in the off-design conditions do not worsen considerably. The
different actuation force required is due to the different stress level at which
the plateau starts changing the operative temperature. The comparison
of the maximum principal stress-strain relationships in the most critical
point of the structure is shown in Figure 5.34. In addition to the usual
loading phase, it reports the unloading of the structure and the recovery
of the transformation strains. Lower the temperature, lower the plateau
stress and higher the transformation strain.
96 5.3. Considerations about the use of Nitinol

Z X

0.25
T=263K
0.2 T=293K
T=313K
0.15 Target

0.1

0.05
y (m)

-0.05

-0.1

-0.15

-0.2
0 0.1 0.2 0.3 0.4 0.5
x (m)

Figure 5.33: Nitinol: deformed shape of optimal solution at different temperatures

Chapter 5. Results 97

108
6
T=263K
T=293K
5 T=313K

4
(Pa)

3
I

0
0 0.005 0.01 0.015 0.02 0.025 0.03
I

Figure 5.34: Nitinol: stress-strain relationship at different temperatures

5.4 Considerations about the optimization pro-

cedure
The results of the present chapter have shown the application of the
optimization procedure and its benefits in terms of fulfillment of the
requirements. A local refinement of the solution obtained from the genetic
algorithm optimization allows a substancial improvement of the morphing
device. Moreover, even adopting the same topology, the application of
a different material is possible and the results demonstrate a further
enhancement if Nitinol is selected. Indeed, it has permitted the achievement
of a better kinematic requirement, as well as the withstanding of the
loads within the material strength limits. This confirms the suitability
of superelastic materials for compliant mechanisms application in the
morphing field. The overall procedure can be considered validated since
it is able to improve the starting solution, decreasing the LSE at the first
step and reducing the stresses at the second step.
The proposed procedure is general for the application to morphing wings
devices. Even though only an example of droop nose has been analyzed, the
procedure can be straightforwardly extended to the optimization refinement
of a trailing-edge morphing device.
98 5.4. Considerations about the optimization procedure

Moreover, the optimization procedure allows to adapt a given topology to

different airfoil shapes and also to different geometric scale of the same
airfoil. An example of this is the scaling process performed when a scaled
wind tunnel model has to be designed. This usually requires a complete
redesign of the structure. However the application of the procedure to the
1:3 scaled model of the droop nose suggests the possibility to scale down
the topology provided that the sizing variables and also the internal points
position are re-optimized according to the requirements of the new model.
The results are discussed in chapter 6.
Chapter 6

Applications

6.1 Preliminary design of a scaled wind tun-

nel model
In the framework of EU funded Clean Sky 2 REG-IADP AG2 project, one
of the activity is related to the construction and testing of a wind tunnel
model at scale 1:3 (WTT2). Its main purpose is the characterization and
validation of NLF technology, but also the study of high lift conditions.
The model is equipped with many features, among which the deployable
morphing leading–edge. In this section a preliminary approach to the design
of the scaled droop nose is presented, aimed at a further confirmation of
the validity of the optimization procedure.
The design of scaled wind tunnel models represents a challenging problem
since the scaling law for the stiffness is different from the geometric one.
Therefore there is usually the need of changing the materials and the
topology of the structure in order to reproduce the required characteristics.
As already explained, the proposed optimization procedure is here applied
to readapt the full scale topology to the scaled model, changing the sizing
variables and the internal points position according to the optimization
analysis in order to fulfill the requirements. Concerning the material of the
mechanism, it must be inevitably changed. In order to identify material
data for the design of the internal structure, based on previous experience at
POLIMI [49], 3D printing technology can be selected for the manufacturing
of the scaled model; a possibile technique is the Stereolithography (SLA)
in combination with a fine powder on the basis of polyamide 12 (PA 2200),
whose mechanical characteristics are reported in Table 6.1. This technique
allows to realize morphing devices in a single piece with good accuracy in
terms of thickness distribution (0.5 mm).

99
100 6.1. Preliminary design of a scaled wind tunnel model

Table 6.1: Mechanical parameters of PA 2200

Parameter Value
Flexural Modulus 1.4 GPa
Poisson’s ratio 0.35
Flexural Strength 76 MPa

Concerning the skin, glass fibers composite material can be selected, whose
mechanical properties are shown in Table 6.2.

Table 6.2: Mechanical parameters of glass

Parameter Value
E11 25 GPa
E22 25 GPa
ν 0.20
G12 4 GPa
G13 4 GPa
G23 4 GPa

6.1.1 Results
First of all, the geometry of the model and the set-up of the optimization
problem are discussed. The airfoil shape, the internal mechanism topology,
rib pitch and rib thickness are scaled down of a factor 3.

• The rib pitch prib is equal to 43.3 mm and it is used as span-wise

length of the skin.

• The rib thickness wrib is equal to 11.7 mm.

• The chord extension of the model is 155.7 mm.

• The actuation force F in the input point has vector components

(−31 N, −2 N).

• The aerodynamic conditions taken into account are both at M ach =

0.148, Sea Level, corresponding to a dynamic pressure qD = 1550 Pa;
Chapter 6. Applications 101

an angle of attack α = 10◦ is considered for the evaluation of the

objective function while α = 0◦ for the deformability constraint. The
pressure coefficient distributions are the same shown in Figure 5.1.

The scaled model corresponding to the inital solution is depicted in Fig-

ure 6.1 together with the original model.
Concerning the starting design variables, the thicknesses of the skin are

Z X

Figure 6.1: Scaled model and full scale model

reduced by the scaling factor, while the dimensions of the rib paths are
preserved.
The other required data to set-up the problem are the limit values for
the constraints and the bounds. For the deformability LSE tolerance
a value of 10−4 m is selected. The admissible value in terms of stress is
76 MPa. The bounds for the sizing variables are 1 mm ≤ trib
i ≤ 6 mm and
skin
0.2 mm ≤ tj ≤ 4 mm. The internal points position can vary of ±5 mm
both in x and y direction.
102 6.1. Preliminary design of a scaled wind tunnel model

At first, a static analysis in the droop nose configuration is performed

to evaluate the LSE and the maximum internal stress for the starting
solution. The deformed shape is shown in Figure 6.2. It is characterized
by an LSE value of 3.56 mm. The maximum Von Mises stress in the
mechanism is 31 MPa, while LSEN LF = 0.06 mm. Therefore the initial
solution is feasible.
The gradient-based optimization is run and it converges to an optimal
solution in 21 iterations. The evolution of the objective function with
the iteration number is shown in Figure 6.3. The LSE for the optimal
solution is 0.84 mm. Maximum stress in the rib is 31 MPa. The comparison
between the initial deformed shape and the optimal deformed shape is
shown in Figure 6.4. The stress level inside the mechanism is shown in
Figure 6.5.
Figure 6.6 illustrates the comparison between the optimal deformed shape
and the target one.
Figure 6.7 compares the sizing variables between the starting solution
and the optimal solution. Moreover, Figure 6.8 shows the topological
differences between the two solutions.
Chapter 6. Applications 103

Z X

X
Z

Figure 6.2: Scaled model: deformed shape of the initial solution

104 6.1. Preliminary design of a scaled wind tunnel model

-3
10
4

3.5

3
LSE (m)

2.5

1.5

0.5
0 5 10 15 20
Iteration
Figure 6.3: Scaled model: LSE vs iteration
Chapter 6. Applications 105

Z X

Figure 6.4: Scaled model: initial solution and optimal solution deformed shapes
106 6.1. Preliminary design of a scaled wind tunnel model

S, Mises
Bottom Left Corner
(Avg: 75%)
+6.710e+07
+6.152e+07
+5.594e+07
+5.035e+07
+4.477e+07
+3.919e+07
+3.360e+07
+2.802e+07
+2.244e+07
+1.685e+07
+1.127e+07
+5.684e+06
+1.005e+05

Z X

Figure 6.5: Scaled model: Von Mises stress in the rib

Chapter 6. Applications 107

0.08
Deformed
Target
0.06

0.04

0.02
y (m)

-0.02

-0.04

-0.06

-0.08
0 0.05 0.1 0.15 0.2
x (m)

Figure 6.6: Scaled model: optimal solution deformed shape and target shape

Y Y

Z X Z X

(a) Initial solution (b) Optimal solution

Figure 6.7: Scaled model: thicknesses

108 6.1. Preliminary design of a scaled wind tunnel model

Z X

Figure 6.8: Scaled model: initial solution and optimal solution undeformed
shapes
Chapter 6. Applications 109

The preliminary design of the scaled droop nose has been performed. The
results can be considered acceptable, since the optimal deformed shape
shows a good enough adherence with the desired target shape, both in
terms of droop nose angle and smoothness of the skin surface. This
achievement has been allowed by an appropriate distribution of thicknesses,
quite different from that of the inital guess solution, whose performance
was not very satisfactory. Also a considerable displacement of the internal
points with respect to the initial solution contributes to the enhancement
of the result.
Once again, the developed gradient-based optimization turns out to be
efficient for the improvement of the kinematic requirement of a morphing
design problem. Moreover, the optimal solution satisfies the structural
requirements related to the stress level in the actuated mechanism and the
sufficient stiffness under the aerodynamic loads of a conventional unmor-
phed configuration.

It must be remembered that the topology of the internal mechanism had

been designed as the result of an optimization process related to the full
scale geometry of the reference wing. The presented preliminary results
prove the validity of the proposed approach, also when dealing with a geo-
metric scaling. In order to further improve the solution, different materials
of the compliant mechanism could be taken into account, looking for the
one that is able to guarantee the best performances. Moreover, additional
constraints can be included in the optimization analysis, depending on
specific requirements needed for the wind tunnel tests.
Once an adequate solution in terms of beam elements model has been
obtained, shape optimization can be performed in order to finalize the
design in view of the manufacturing process.
Finally, manufacturing and testing are mandatory for the assessment of
the numerical solution and to study the overall behaviour of the morphing
device.
Chapter 7

Conclusions

In this work an optimization procedure to improve the design of morphing

devices has been proposed. The field of morphing in which it is located is
the active camber morphing, limited to leading and trailing edges. The
resulting concept is that of hybrid wings, where the wing-box structure is
not affected by morphing while the performance enhancement is due to
the leading and trailing edge conformable devices.
The selected approach for the design of the internal structure is based on the
distributed compliance concept. Instead of using rigid levers connected by
joints, the desired shape change is accomplished by efficiently distributing
the flexibility into the structure. This choice should avoid the high stress
concentrations encountered in mechanisms based on rigid kinematics.
In order to satisfactorily deal with the design problem, conflicting require-
ments must be simultaneously fulfilled, therefore specific optimization tools
are needed to assist the engineer in the definition of morphing solutions
representing an optimal compromise between the expected performances
and the imposed restrictions.

An existing optimization procedure based on a two-levels approach is

the starting point for the whole work. The first level is an aerodynamic
shape optimization aimed at obtaining the best morphing shape able to
guarantee the aerodynamic requirements under skin structural constraints.
The second level regards the design of an internal compliant structure able
to realize, once actuated, the desired target shape previously found.
In this thesis, the existing tools have been included in a wider procedure,
whose main purpose is the improvement of the second level solution, from
both the perfomance and the structural feasibility points of view. This
last point is especially critical, since it has been noted that for droop
nose applications a conventional aeronautical aluminium alloy isn’t able to

111
112

provide, limited to the elastic range, the high strains required to achieve
the target aerodynamic shape. Therefore the need of large recoverable
strains has suggested to exploit the superelastic behaviour, characteristic
of shape memory alloys such as Nitinol.
The use of a different material with respect to that employed in the opti-
mization for the definition of the internal mechanism topology represents
the main added value of the proposed procedure. A previously obtained
topology can be later readapted including various materials. To do this, a
dedicated optimization tool has been specifically developed in this work
in order to find the optimal design variables according to the selected
material. It is a gradient-based optimization whose variables are the sizing
in-plane dimensions of both the mechanism and the skin and also the local
position of the internal points of the mechanism. The objective of the
optimization analysis is the improvement of the kinematic requirement
under structural constraints. The beam elements model of the mechanism
used in this analysis is later translated into a more detailed two-dimensional
solid elements model in order to assess the stress concentrations at the
load path intersections. Then a shape optimization is performed on this
model to reduce the stress peaks by modifying the design of the critical
regions, bringing to a final solution almost ready for the manufacturing
process.

The entire procedure has been applied to the droop nose device of a
regional aircraft in the framework of an European project. First of all the
internal compliant structure made of isotropic aluminum alloy has been
assessed. The results have shown an improvement of the aerodynamic
requirement in terms of deformed target shape, however the violation of
the structural constraints establishes the unfeasibility of the solution. As
already explained, the unsuitability of the traditional aeronautical material
for the considered application leads to its replacement with the Nitinol
superelastic material. The same procedure has been repeated with the
new material giving a satisfactory outcome. Indeed, the exploitation of
the stress plateau typical of the superelastic behaviour has allowed the
limitation of the maximum stress and as a consequence the achievement of
better performances in obtaining the desired morphed shape. Very large
strains can appear in the morphing configuration but the superelastic effect
assures their recovery. Later, the dependence of the material behaviour on
the operative temperature is assessed by comparing the optimal solution
at the design temperature and at the selected limit temperatures. The
results show good performances also in the off-design conditions.
All these results demonstrate the validity of the procedure according to its
Chapter 7. Conclusions 113

purposes of performance improvement, structural feasibility, versatility of

applications.
In addition to the refinement of existing solutions and the possibility to
adopt new materials, another potentiality of the procedure concerns the
application of available optimal topologies to a different geometric scaling
of the wing. This has been practically shown by facing the preliminary
design of a 1:3 scaled wind tunnel model of the same droop nose device
previously addressed. The results have well confirmed the eligibility of
the procedure in dealing with this particular task. The topology has
been reduced according to the scaling factor and the starting optimiza-
tion variables have been selected in a suitable manner. Concerning the
material, three dimensional printing technology has been selected for its
manufacturing. The results have proved once again the effectiveness of the
optimization in the achievement of a solution characterized by valuable
performances in terms of deformed shape and also the structural feasibility
of the device.

Concerning the future work needed to further improve the proposed proce-
dure, the next required step is the numerical verification of the designed
optimal morphing device. This is needed since the last step of the pro-
cedure optimizes the internal mechanism only, with boundary conditions
inherited from the deformed shape. Therefore it is essential to verify that
the final result doesn’t exhibit a worsening of the performances. The
last step from the numerical point of view is the realization of complete
three-dimensional models of the morphing devices and the assessment
of their overall behaviour, as well as the interaction with the wing-box
structure. This is essential in view of the future integration of morphing
devices inside a real wing structure.
Once the numerical design phase is completed, manufacturing and testing
are required to establish the real feasibility of the conceived solution.
This includes in-depth studies of characterisation for the materials and
also the investigation of the related manufacturing technologies. This
is especially crucial in case of Nitinol. Due to the dependence of its
mechanical characteristics on the alloy composition, the manufacturing
process, the thermal treatments and the operative temperatures, a lot of
work is required to better understand what is needed to achieve the actual
application of Nitinol in the aeronautical structures.
Finally, the acquisition of a strong experience in the design of morphing
devices by means of dedicated procedures is fundamental to undertake the
difficult certification process of unconventional aircraft.
Bibliography

[1] Abaqus Scripting User’s Guide. Dassault Systemes Simulia Corpora-

tion. 2016.
[2] Ferdinando Auricchio, Robert L. Taylor, and Jacob Lubliner. “Shape-
memory alloys: macromodelling and numerical simulations of the
superelastic behavior”. In: Computer methods in applied mechanics
and engineering 146.3-4 (1997), pp. 281–312.
[3] S. Barbarino et al. “Airfoil structural morphing based on SMA ac-
tuator series: numerical and experimental studies”. In: Journal of
Intelligent Material Systems and Structures 22.10 (2011), pp. 987–
1004.
[4] Silvestro Barbarino, Wulf G. Dettmer, and Michael I. Friswell. “Mor-
phing trailing edges with shape memory alloy rods”. In: Proceeding
of. 2010.
[5] Silvestro Barbarino et al. “A review of morphing aircraft”. In: Journal
of intelligent material systems and structures 22.9 (2011), pp. 823–
877.
[6] M.C. Biggs. “Constrained minimization using recursive quadratic
programming”. In: Towards global optimization (1975).
[7] Kenneth Bonnema and Stephen Smith. “AFTI/F-111 mission adap-
tive wing flight research program”. In: 4th Flight Test Conference.
1988, p. 2118.
[8] Jason Bowman et al. “Development of next generation morphing
aircraft structures”. In: 48th AIAA/ASME/ASCE/AHS/ASC Struc-
tures, Structural Dynamics, and Materials Conference. 2007, p. 1730.
[9] L. Catherine Brinson. “One-dimensional constitutive behavior of
shape memory alloys: thermomechanical derivation with non-constant
material functions and redefined martensite internal variable”. In:
Journal of intelligent material systems and structures 4.2 (1993),
pp. 229–242.

115
116 Bibliography

[10] W.J. Buehler and R.C. Wiley. “Nickel-based alloys Technical Report”.
In: (1965).
[11] A. Chrysochoos, M. Lobel, and O. Maisonneuve. “Thermomechanical
coupling of pseudoelastic behavior of CuZnAl and NiTi alloys”. In:
Comptes Rendus De L Academie des sciences serie II 320.5 (1995),
pp. 217–223.
[12] Monroe Conner. F-111 Advanced Fighter Technology Integration.
2017. url: https://www.nasa.gov/centers/dryden/multimedia/
imagegallery/F-111AFTI/EC85-33205-07.html.
[13] Alessandro De Gaspari and Sergio Ricci. “A two-level approach for
the optimal design of morphing wings based on compliant structures”.
In: Journal of Intelligent Material Systems and Structures 22.10
(2011), pp. 1091–1111.
[14] Alessandro De Gaspari and Sergio Ricci. “Application of the active
camber morphing concept based on compliant structures to a re-
gional aircraft”. In: Industrial and Commercial Applications of Smart
Structures Technologies 2014. Vol. 9059. International Society for
Optics and Photonics. 2014, p. 90590D.
[15] Alessandro De Gaspari and Sergio Ricci. “Knowledge-based shape
optimization of morphing wing for more efficient aircraft”. In: Inter-
national Journal of Aerospace Engineering 2015 (2015).
[16] Kalyanmoy Deb. Multi-objective optimization using evolutionary
algorithms. Wiley, 2005.
[17] R. Decamp and Richard Hardy. “Mission adaptive wing advanced re-
search concepts”. In: 11th Atmospheric Flight Mechanics Conference.
1984, p. 2088.
[18] Mohammad Elahinia et al. “Fabrication of NiTi through additive
manufacturing: A review”. In: Progress in Materials Science 83 (2016),
pp. 630–663.
[19] M. Frecker, Noboru Kikuchi, and S. Kota. “Topology optimization of
compliant mechanisms with multiple outputs”. In: Structural Opti-
mization 17.4 (1999), pp. 269–278.
[20] M.I. Frecker et al. “Topological synthesis of compliant mechanisms
using multi-criteria optimization”. In: Journal of Mechanical design
119.2 (1997), pp. 238–245.
Bibliography 117

[21] Gian Luca Ghiringhelli, Pierangelo Masarati, and Paolo Mantegazza.

“Multibody implementation of finite volume C 0 beams”. In: AIAA
journal 38.1 (2000), pp. 131–138.
[22] Douglas K. Gould. Final Report, Variable Camber Wing| Phase I.
Tech. rep. D180-17606-1 (Contract No. N00014-73-C-0244), Research
& Engineering Div., Boeing Aerospace Co, 1973.
[23] J.A. Hetrick and S. Kota. “An energy formulation for parametric size
and shape optimization of compliant mechanisms”. In: Journal of
Mechanical Design 121.2 (1999), pp. 229–234.
[24] FlexSys Inc. 2018. url: https://www.flxsys.com/.
[25] Markus Kintscher, Hans Peter Monner, and Olaf Heintze. “Experi-
mental testing of a smart leading edge high lift device for commercial
transportation aircrafts”. In: 27th International Congress of the Aero-
nautical Sciences (ICAS), Nice, France. 2010.
[26] Kyle Kirkland. Chemistry: Notable Research and Discoveries. Vol. 2.
Infobase Publishing, 2010.
[27] Johannes Kirn and Stefan Storm. “Kinematic solution for a highly
adaptive droop nose”. In: ICAST2014: 25th international conference
on adaptive structures and technologies. 2014, pp. 6–8.
[28] Sridhar Kota, Peter Flick, and Fayette S. Collier. “Flight Testing
of FlexFloilTM Adaptive Compliant Trailing Edge”. In: 54th AIAA
Aerospace Sciences Meeting. 2016, p. 0036.
[29] Sridhar Kota et al. “Design of compliant mechanisms: applications to
MEMS”. In: Analog integrated circuits and signal processing 29.1-2
(2001), pp. 7–15.
[30] Sridhar Kota et al. “Mission adaptive compliant wing–design, fabri-
cation and flight test”. In: RTO Applied Vehicle Technology Panel
(AVT) Symposium. 2009.
[31] Brenda M. Kulfan. “Universal parametric geometry representation
method”. In: Journal of Aircraft 45.1 (2008), pp. 142–158.
[32] Daochun Li et al. “A review of modelling and analysis of morphing
wings”. In: Progress in Aerospace Sciences (2018).
[33] Paulo Silva Lobo, João Almeida, and Lúıs Guerreiro. “Shape memory
alloys behaviour: A review”. In: Procedia Engineering 114 (2015),
pp. 776–783.
118 Bibliography

[34] Kerr-Jia Lu and Sridhar Kota. “An effective method of synthesizing

compliant adaptive structures using load path representation”. In:
Journal of Intelligent Material Systems and Structures 16.4 (2005),
pp. 307–317.
[35] Kerr-Jia Lu and Sridhar Kota. “Parameterization strategy for opti-
mization of shape morphing compliant mechanisms using load path
representation”. In: ASME 2003 International Design Engineering
Technical Conferences and Computers and Information in Engineer-
ing Conference. American Society of Mechanical Engineers. 2003,
pp. 693–702.
[36] Kerr-Jia Lu and Sridhar Kota. “Topology and dimensional synthesis
of compliant mechanisms using discrete optimization”. In: Journal
of Mechanical Design 128.5 (2006), pp. 1080–1091.
[37] Anna-Maria R. McGowan et al. “Perspectives on Highly adaptive or
morphing aircraft”. In: (2009).
[38] Ashok Midha, Tony W. Norton, and Larry L. Howell. “On the Nomen-
clature and Classification of Compliant Mechanisms. the Components
of Mechanisms”. In: (1992).
[39] Giulio Molinari, Andres F. Arrieta, and Paolo Ermanni. “Aero-
structural optimization of 3-D adaptive wings with embedded smart
actuators”. In: 54th AIAA/ASME/ASCE/AHS/ASC Structures, Struc-
tural Dynamics, and Materials Conference. 2013, p. 1528.
[40] H.P. Monner et al. “Design aspects of the adaptive wing—the elastic
trailing edge and the local spoiler bump”. In: The Aeronautical
Journal 104.1032 (2000), pp. 89–95.
[41] Gabriel Murray, Farhan Gandhi, and Charles Bakis. “Flexible matrix
composite skins for one-dimensional wing morphing”. In: Journal of
Intelligent Material Systems and Structures 21.17 (2010), pp. 1771–
1781.
[42] Shinji Nishiwaki et al. “Topology optimization of compliant mecha-
nisms using the homogenization method”. In: International journal
for numerical methods in engineering 42.3 (1998), pp. 535–559.
[43] Optimization Toolbox User’s Guide. Matlab The Mathworks Inc.
2017.
[44] Kazuhiro Otsuka and Xiabing Ren. “Physical metallurgy of Ti–Ni-
based shape memory alloys”. In: Progress in materials science 50.5
(2005), pp. 511–678.
Bibliography 119

[45] Larry D. Peel et al. “Development of a simple morphing wing us-

ing elastomeric composites as skins and actuators”. In: Journal of
Mechanical Design 131.9 (2009), p. 091003.
[46] M.A. Qidwai and D.C. Lagoudas. “On thermomechanics and transfor-
mation surfaces of polycrystalline NiTi shape memory alloy material”.
In: International journal of plasticity 16.10-11 (2000), pp. 1309–1343.
[47] Nuno Rebelo, N. Walker, and H. Foadian. “Simulation of implantable
nitinol stents”. In: Abaqus user’s conference. 2001, pp. 1–14.
[48] Benjamin Reedlunn et al. “Tension, compression, and bending of
superelastic shape memory alloy tubes”. In: Journal of the Mechanics
and Physics of Solids 63 (2014), pp. 506–537.
[49] Sergio Ricci, Alessandro De Gaspari, and Luca Riccobene. “Design,
manufacturing and wind tunnel test of a morphing wing based
on compliant structures”. In: 24th AIAA/AHS Adaptive Structures
Conference. 2016, p. 1316.
[50] Sergio Ricci et al. “Design of a Leading Edge Morphing Based on
Compliant Structures for a Twin-Prop Regional Aircraft”. In: 2018
AIAA/AHS Adaptive Structures Conference. 2018, p. 1063.
[51] Anton Rudenko et al. “Extremely deformable morphing leading edge:
Optimization, design and structural testing”. In: Journal of Intelligent
Material Systems and Structures 29.5 (2018), pp. 764–773.
[52] Laxminarayana Saggere and Sridhar Kota. “Static shape control of
smart structures using compliant mechanisms”. In: AIAA journal
37.5 (1999), pp. 572–578.
[53] M. Santer and S. Pellegrino. “Topological optimization of compliant
adaptive wing structure”. In: AIAA journal 47.3 (2009), pp. 523–534.
[54] Ole Sigmund. “On the design of compliant mechanisms using topol-
ogy optimization”. In: Journal of Structural Mechanics 25.4 (1997),
pp. 493–524.
[55] Clean Sky. 2018. url: http://www.cleansky.eu/.
[56] John W. Smith, Wilton P. Lock, and Gordon A. Payne. “Variable-
camber systems integration and operational performance of the
AFTI/F-111 mission adaptive wing”. In: (1992).
[57] Stephen B. Smith and David W. Nelson. “Determination of the
aerodynamic characteristics of the mission adaptive wing”. In: Journal
of Aircraft 27.11 (1990), pp. 950–958.
120 Bibliography

[58] A.Y.N. Sofla, D.M. Elzey, and H.N.G. Wadley. “Two-way antagonistic
shape actuation based on the one-way shape memory effect”. In:
Journal of Intelligent Material Systems and Structures 19.9 (2008),
pp. 1017–1027.
[59] J. Szodruch and R. Hilbig. “Variable wing camber for transport
aircraft”. In: Progress in Aerospace Sciences 25.3 (1988), pp. 297–
328.
[60] Joachim Szodruch. “The influence of camber variation on the aero-
dynamics of civil transport aircraft”. In: 23rd Aerospace Sciences
Meeting. 1985, p. 353.
[61] C. Thill et al. “Composite corrugated structures for morphing wing
skin applications”. In: Smart Materials and Structures 19.12 (2010),
p. 124009.
[62] Srinivas Vasista et al. “Evaluation of a compliant droop-nose mor-
phing wing tip via experimental tests”. In: Journal of Aircraft 54.2
(2016), pp. 519–534.
[63] Terrence A. Weisshaar. Morphing aircraft technology-new shapes for
aircraft design. Tech. rep. PURDUE UNIV LAFAYETTE IN, 2006.
[64] R. Wlezien et al. “The aircraft morphing program”. In: 39th AIAA/AS-
ME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materi-
als Conference and Exhibit. 1998, p. 1927.