Download as pdf or txt
Download as pdf or txt
You are on page 1of 138

Study and design of a monocoque wing

structure with composite materials

BACHELOR’S DEGREE THESIS

Bachelor’s degree in Aerospace Vehicles Engineering

Miguel Alejandro Pareja Muñoz


Director: Lluı́s Gil Espert

June 2016

DOCUMENT 1.- REPORT

Universitat Politècnica de Catalunya

Escola Superior d’Enginyeries Industrial, Aeroespacial i Audiovisual de Terrassa


”Inspiration unlocks the future.”

The Wind Rises (2013)

i
Acknowledgements

Firstly, I would like to express my sincere gratitude to my advisor Prof. Lluı́s Gil for
the continuous support during the making of my Bachelor’s degree thesis, for his
patience, motivation, and immense knowledge. His guidance has helped me in all the
time of research and design, and also in the writing of this thesis.

My most sincere thanks also goes to my team mate and friend, Roger Serra. He has
selflessly dedicated his precious time to help me so many times, not only allowing me to
reach the solution of my problems, but also making me have a great time with him.

I also want to acknowledge my team mate Oriol Chandre for his great company at
the Trencalòs Team workshop, in which we both have spent several hours working on our
thesis. He always being in a great mood and his energy have always motivated me and
influenced me very positively.

Lastly, my highest appreciation goes to my family and friends, without whom I would
not be who I am today. Special thanks goes to my mom, as she has supported me
wholeheartedly throughout the making of this thesis - both in the good and bad moments
I have experienced.

ii
Abstract

Trencalòs Team - an aeromodelling research team, is looking to enlarge its know-how


about composite-made structures. For their own experience, the best option to win the
competitions they attend is with a monocoque wing structure made with composite
materials.

Me, as a member of Trencalos Team, I have taken this oportunity to help in the
study of that matter. The thesis covers the study of the main the knowledge necessary
to approach a design of such a structre in depth. This type of structure is more complex
to characterize than a classical truss structure - its structural behaviour is influenced by
several parameters, i.e. is multi-parametric.

This parameters/features have been studied, as well as the governing laws that
describe the structural beeahviour of tgis structures. Additionally, the design of a
structure of this kind has been performed following a design methodology that is based
in optimisation of two objectives. The design of such a structure represents a problem
with multiple possible objectives, and for the sake of simplicity, just two of them have
been addressed.

The design methodology has been implemented in the following way: making a set of
initial cases, analyse them and then apply an optimisation procedure to discard cases
and find the best ones. After doing that, more cases have been made from the best cases
of the last iteration. The modelling of the geometry in this thesis has been done with
CATIA V5R20. The structural analysis has been performed with AbaqusTM .

From the methodology premises that have been set in order to delimit such a
multi-parametric study, results have been obtained. These mainly have been useful to
observe the trends/effects of each of these design parameters in the final structure
behaviour. Besides, an optimal solution has been found following an optimisation
procedure, which is interesting from the perpspective of familiarising with this kind of
tool, as any complex engineering problem depends on a lot of parameters and they are
required to solve it.

With all this accomplished, conclusions have been extracted about which are the best
parameters combinations to include in a monocoque wing structure made with composite
materials.

iii
Summary

The thesis is divided in 8 chapters: Introduction, State of the Art, Methodology,


Results, Conclusions, Future Work, Environmental Impact, Budget and Annxes.

In the Introdcution chapter, a brief and concise summary of the aim, scope and
objectives of the thesis is delivered. Also, the justification and a planning of the same is
attached.

In the State of the Art chapter, all the bibliographical research is done. All the
infomation necessary to build a base on how monocoque wing structures made of
composite materials behave is given there. Also, basic information about the design tools
used for the design process is given here.

In the Methodology chapter, the premises on which the design is based on are
detailed. Then, the design procedure - set of steps - followed to obtain an optimal
solution is explained. And finally, the process to prepare, execute and post-process an
structural analysis is explained.

In the Results chapter, the results obtained are thoroughly explained with all kinds
of details and different supports - i.e. tables, plots, figures, etc.

In the Conclusions chapter, all the information collected in the thesis is analysed
from a general point of view and some conclusions are extracted.

In the Future Work chapter, a suggestion of how the work perfomed in this thesis
could be continuedç, and in what direction, is explained.

In the Environmental Impact, an insight of the impact of what has been done to
accomplish this thesis on the environment is explained.

In the Budget chapter, a budget of how much a project like this one would cost if it
was undertaknen professionally is given. The budget is done separately in a different
document.

And for last, Annexes are attached providing complementary information that is of
interest. The annexes are done separately in a different document.

iv
Contents

Acknowledgements ii

Abstract iii

Summary iv

List of Figures vii

List of Tables xi

List of Abbreviations and Symbols xiii

1 Introduction 1
1.1 Aim . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Scope . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.3 Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.4 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5 Justification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5.1 Identification of the Need . . . . . . . . . . . . . . . . . . . . . . . . 3
1.5.2 Usefulness of the Study . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.6 Study’s Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.7 Planning . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.7.1 Brief Description of Tasks . . . . . . . . . . . . . . . . . . . . . . . . 6
1.7.2 Detailed Relation Between Tasks . . . . . . . . . . . . . . . . . . . . 8
1.7.3 Scheduling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2 State of the Art 10


2.1 Historical Preamble . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.2 Structural Concepts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.3 Composite Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.3.1 Main Properties of Composites . . . . . . . . . . . . . . . . . . . . . 20
2.3.2 Classification of Composites . . . . . . . . . . . . . . . . . . . . . . . 22
2.3.3 Laminae and Laminates . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.3.3.1 Laminae . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.3.3.2 Laminates . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.4 Laminae Mechanical and Physical Properties . . . . . . . . . . . . . . . . . 31
2.5 Classical Laminated Plate Theory . . . . . . . . . . . . . . . . . . . . . . . 35
2.5.1 Kinematics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.5.2 The Material Law . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
2.5.3 Resultant Forces and Moments . . . . . . . . . . . . . . . . . . . . . 38
2.6 Buckling of Laminated Plates . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.7 Finite Element Analysis of Laminates . . . . . . . . . . . . . . . . . . . . . 43

v
2.7.1 Basic FEM Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2.7.2 General FEM Procedure . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.7.3 AbaqusTM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.7.4 Shell Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
2.8 Failure Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.8.1 Hashin Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
2.8.2 Tsai-Hill Criterion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
2.9 Pareto Optimality . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

3 Methodology 54
3.1 Premises of the Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.1.1 Type of Structural Testing . . . . . . . . . . . . . . . . . . . . . . . 54
3.1.2 Structural Requirements . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.1.3 Geometrical Features of the Wing . . . . . . . . . . . . . . . . . . . 55
3.1.4 Load-bearing Elements Employed . . . . . . . . . . . . . . . . . . . . 56
3.1.5 Materials Employed and its Properties . . . . . . . . . . . . . . . . . 57
3.1.6 Maximum Weight of the Structure . . . . . . . . . . . . . . . . . . . 59
3.1.7 Minimum Buckling Constant . . . . . . . . . . . . . . . . . . . . . . 59
3.1.8 Variation of Parameters Criterion . . . . . . . . . . . . . . . . . . . . 60
3.1.9 Parameters of the Design . . . . . . . . . . . . . . . . . . . . . . . . 61
3.2 Design Process Description . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.1 Initial Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.1.1 Mesh Convergence . . . . . . . . . . . . . . . . . . . . . . . 62
3.2.1.2 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . 63
3.2.1.3 Load Definition . . . . . . . . . . . . . . . . . . . . . . . . 64
3.2.1.4 Buckling Modes Check . . . . . . . . . . . . . . . . . . . . 67
3.2.1.5 First Tendencies Observed . . . . . . . . . . . . . . . . . . 67
3.2.2 Starting Point . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.2.2.1 Decision Making . . . . . . . . . . . . . . . . . . . . . . . . 68
3.2.3 Improvements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69
3.2.4 Final Decision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
3.2.5 Flowchart of the Design Process . . . . . . . . . . . . . . . . . . . . 70
3.3 Analysis Process Description . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3.1 Pre-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.3.1.1 Geometry Preparation . . . . . . . . . . . . . . . . . . . . . 72
3.3.1.2 AbaqusTM Model Making . . . . . . . . . . . . . . . . . . . 73
3.3.2 Simulation of the Analysis . . . . . . . . . . . . . . . . . . . . . . . . 76
3.3.3 Post-processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.3.4 Flowchart of the Analysis Process . . . . . . . . . . . . . . . . . . . 78

4 Results 80
4.1 Initial Tests Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.1.1 Converged Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
4.1.2 Final Encastre . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
4.1.3 Final Load Distribution . . . . . . . . . . . . . . . . . . . . . . . . . 84
4.1.4 Buckling Modes Verification . . . . . . . . . . . . . . . . . . . . . . . 87
4.1.5 Trends and Structural Behaviour . . . . . . . . . . . . . . . . . . . . 95
4.2 First Batch of Cases - Starting Point . . . . . . . . . . . . . . . . . . . . . . 96
4.3 Second Batch of Cases - Improvement 1 . . . . . . . . . . . . . . . . . . . . 100
4.4 Third Batch of Cases - Improvement 2 . . . . . . . . . . . . . . . . . . . . . 101
4.5 Fourth Batch of Cases - Improvement 3 . . . . . . . . . . . . . . . . . . . . 102
4.6 Decision after Improvements 1, 2 and 3 . . . . . . . . . . . . . . . . . . . . 106

vi
4.7 Fifth Batch of Cases - Improvement 4 . . . . . . . . . . . . . . . . . . . . . 108
4.8 Sixth Batch of Cases - Improvement 5 . . . . . . . . . . . . . . . . . . . . . 110
4.9 Final Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

5 Conclusions 114
5.1 Conclusions about the Results . . . . . . . . . . . . . . . . . . . . . . . . . . 114
5.2 Conclusions about the Numerical Model . . . . . . . . . . . . . . . . . . . . 116
5.3 Conclusions about the Design Process . . . . . . . . . . . . . . . . . . . . . 116
5.4 Project Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

6 Future Work 118

7 Environmental Impact 119

Bibliography 121

vii
List of Figures

1.1 Lancair IV. Retrieved from [1]. . . . . . . . . . . . . . . . . . . . . . . . . . 4


1.2 The b.ONE, Trencalòs Team’s aircraft that competed in the Air Cargo
Challenge 2015. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 Gantt chart of the organisation of tasks in time. . . . . . . . . . . . . . . . 9

2.1 Chanute’s biplane glider made of truss construction. Retrieved from [2]. . . 10
2.2 The Wright flyer. Retrieved from [3]. . . . . . . . . . . . . . . . . . . . . . . 11
2.3 Bleriot’s airplane. The first monoplane. Retrieved from [4]. . . . . . . . . . 11
2.4 The Deperdussin, the first airplane with monocoque structure in the
fuselage. Retrieved from [5]. . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.5 The Junkers J1. Retrieved from [6]. . . . . . . . . . . . . . . . . . . . . . . 13
2.6 An American Airlines Boeing 707. Retrieved from [7]. . . . . . . . . . . . . 13
2.7 Research aircraft X-29. Retrieved from [8]. . . . . . . . . . . . . . . . . . . 14
2.8 Rate of materials employed in the Boeing 787. Retrieved from [9]. . . . . . 14
2.9 Example of a truss structure of an aircraft fuselage. Retrieved from [10]. . . 15
2.10 Monocoque vs. Semi-monocoque structure of a fuselage tail boom.
Retrieved from [11]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.11 Fatigue curve of an alluminium alloy (a) and a unidirectional composite
(b). Retrieved from [12]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
2.12 Comparison of different composites and an aluminium alloy specific
properties. Retrieved from [12]. . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.13 Stress-strain curves of a metal and a composite. Retrieved from [12]. . . . . 21
2.14 Comparison of characteristics of different materials: a) density (kg/m3 ) b)
tenisle fracture strength (MPa) c) modulus of elasticity (MPa) d) price per
unit mass. Retrieved from [12]. . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.15 Composites classification depending on the reinforcement type. Retrieved
from [13]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.16 Anisotropic material. Retrieved from [14]. . . . . . . . . . . . . . . . . . . . 25
2.17 Rotation of axes. Retrieved from [14]. . . . . . . . . . . . . . . . . . . . . . 26
2.18 Laminae representation in space. Retrieved from [12]. . . . . . . . . . . . . 26
2.19 Variation of the elasticity modulus with θ from the longitudinal El (fibre
mudulus) to the transverse Et (matrix modulus). Retrieved from [12]. . . . 27
2.20 Laminate example with 0◦ , -θ◦ , θ◦ , 90◦ layer arrangement. Retrieved from
[15]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.21 Variation of properties with volume fraction modification of the different
layers. Retrieved from [15] . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.22 Deformation due to the anisotropy of composites.Retrieved from [12]. . . . . 29
2.23 Laminate representation in space. Retrieved from [12]. . . . . . . . . . . . . 30
2.24 Idealisation of monolayer of fibre and matrix with longitudinal force applied.
Retrieved from [12]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.25 Idealisation of monolayer of fibre and matrix with transverse force applied.
Retrieved from [12]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33

viii
2.26 Idealisation of monolayer of fibre and matrix with shear force applied.
Retrieved from [12]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
2.27 Applied tractions and moments in plate. Retrieved from [16]. . . . . . . . . 36
2.28 Undeformed and deformed geometries of an edge of a plate under the CLPT
assumptions. Retrieved from [17]. . . . . . . . . . . . . . . . . . . . . . . . . 37
2.29 Bifurcation problem scheme. Retrieved from [18]. . . . . . . . . . . . . . . . 41
2.30 Plate of dimensions a x b. Retrieved from [19]. . . . . . . . . . . . . . . . . 42
2.31 Physical (a) and mathematical (b) model. Retrieved from [20]. . . . . . . . 43
2.32 Discretisation into elements. Retrieved from [20]. . . . . . . . . . . . . . . . 44
2.33 Linear interpolation functions for a two-node element bar. Retrieved from
[20]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.34 Three two-node element bar discretisation. Retrieved from [20]. . . . . . . . 46
2.35 Flow chart of the process to complete an AbaqusTM analysis. Retrieved
from [21]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.36 Examples illustrating the Pareto optimality. Retrieved from [22]. . . . . . . 52
2.37 Example of a Pareto front. Retrieved from [23]. . . . . . . . . . . . . . . . . 53

3.1 Schematic representation of the structural testing that is going to be


performed on the wing. Retrieved from [24]. . . . . . . . . . . . . . . . . . . 55
3.2 Clarky airfoil with a cut at 0.17 m of the chord line. . . . . . . . . . . . . . 56
3.3 Structural elements employed. . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.4 Skin layer sequence depiction and differentiation between high and low load-
bearing regions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
3.5 (a) Pareto front of a two minimum objectives (b) Pareto front of a maximum
objetive and a minimum objective. . . . . . . . . . . . . . . . . . . . . . . . 60
3.6 First encastre. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.7 Second encastre. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.8 Third encastre. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
3.9 (a) Lift distribution (b) Cl distribution. . . . . . . . . . . . . . . . . . . . . 64
3.10 Induced angle along the span. . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.11 Lift distribution and tip vortex effect downstream. . . . . . . . . . . . . . . 65
3.12 Discretisation in elements of the airfoil. . . . . . . . . . . . . . . . . . . . . 65
3.13 Pressure distribution along extrados and intrados. Retrieved from [25]. . . . 66
3.14 First load distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
3.15 Second load distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.16 Third load distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.17 Decision making for each point of the set of points analysed after each step. 69
3.18 Flowchart of the design process. . . . . . . . . . . . . . . . . . . . . . . . . 71
3.19 (a) Wrong geometry due to lack of partitioning (b) Good geometry with
skin partitioning. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72
3.20 Definition of a composite layup in AbaqusTM . . . . . . . . . . . . . . . . . . 74
3.21 Stack-up plot of a 5 layer region, where no CE is employed. . . . . . . . . . 74
3.22 Stack-up plot of a 7 layer region, where CE is employed. . . . . . . . . . . . 75
3.23 Close-up of the stack-up plot 7 layer region. . . . . . . . . . . . . . . . . . . 75
3.24 Definition of a tie interaction in AbaqusTM . . . . . . . . . . . . . . . . . . . 76
3.25 Flowchart of the analysis process. . . . . . . . . . . . . . . . . . . . . . . . 79

4.1 Displacement vs. semi-span for the 5 meshes. . . . . . . . . . . . . . . . . . 81


4.2 Close-up of the Displacement vs. semi-span for the 5 meshes. . . . . . . . . 82
4.3 Stress distribution with the encastre 1. . . . . . . . . . . . . . . . . . . . . . 83
4.4 Stress distribution with the encastre 2. . . . . . . . . . . . . . . . . . . . . . 83
4.5 Stress distribution with the encastre 3. . . . . . . . . . . . . . . . . . . . . . 84

ix
4.6 Displacement field exhibit by the three load distributions, this one belonging
to the first one. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.7 First buckling mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
4.8 Second buckling mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.9 Third buckling mode. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
4.10 Intrados σ11 distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87
4.11 Intrados σ22 distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.12 Intrados τ12 distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88
4.13 Local coordinate systems at the intrados. . . . . . . . . . . . . . . . . . . . 89
4.14 Stresses in an intrados element. . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.15 Stress at the intrados and lamina equivalence of a pair of laminae at ±45
and 0/90◦ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
4.16 First buckling mode of the 0/90◦ GE skin configuration. . . . . . . . . . . . 90
4.17 Second buckling mode of the 0/90◦ GE skin configuration. . . . . . . . . . . 90
4.18 Third buckling mode of the 0/90◦ GE skin configuration. . . . . . . . . . . . 91
4.19 Deformed shape of the composite-made structure. . . . . . . . . . . . . . . . 92
4.20 Deformed shape of the equivalent structure. . . . . . . . . . . . . . . . . . . 92
4.21 (a) Mode 1 of the composite-made structure (b) Mode 1 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.22 (a) Mode 2 of the composite-made structure (b) Mode 2 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.23 (a) Mode 3 of the composite-made structure (b) Mode 3 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.24 (a) Mode 4 of the composite-made structure (b) Mode 4 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.25 (a) Mode 5 of the composite-made structure (b) Mode 5 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.26 (a) Mode 6 of the composite-made structure (b) Mode 6 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
4.27 (a) Mode 7 of the composite-made structure (b) Mode 7 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.28 (a) Mode 8 of the composite-made structure (b) Mode 8 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.29 (a) Mode 9 of the composite-made structure (b) Mode 9 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.30 (a) Mode 10 of the composite-made structure (b) Mode 10 of the equivalent
structure. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.31 150 points of Batch 1. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96
4.32 Batch 1 with limiting lines m = 0.45 and 1/λ=1. . . . . . . . . . . . . . . . 97
4.33 Batch 1 with the first step of the decision making procedure applied. . . . . 97
4.34 Batch 1 with the second step of the decision making procedure applied. . . 98
4.35 Batch 1 valid points and the Pareto front. . . . . . . . . . . . . . . . . . . . 98
4.36 Close-up of the Batch 1 valid points and the Pareto front. . . . . . . . . . . 99
4.37 Batch 1 valid points, the Pareto front, and the two selected points to
continue the design process. . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
4.38 Batch 1 valid points, old and new Pareto front and Batch 2 valid points. . . 101
4.39 Batch 1 valid points, old and new Pareto front and Batch 3 valid points. . . 102
4.40 Stringers improvement implemented in Bacth 4. . . . . . . . . . . . . . . . . 103
4.41 Batch 1 valid points, old and new Pareto front and Batch 4 valid points. . . 104
4.42 Batch 1 valid points, old and new Pareto front and extra (valid) points for
Batch 4. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

x
4.43 Batch 1 valid points, old and new Pareto front and final Batch 4 points. . . 105
4.44 Close-up of Batch 1 valid points, old and new Pareto front and final Batch
4 points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106
4.45 Batches 1 to 4 valid points, and the old and new Pareto front. . . . . . . . . 107
4.46 Batches 1 to 4 valid points, and the old and new Pareto front, and the new
optimal points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
4.47 Batches 1 to 4 valid points, old and new Pareto front and Batch 5 valid
points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.48 Batches 1 to 4 valid points, old and new Pareto front, Batch 5 valid points
and the 6 optimal points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
4.49 Batches 1 to 5 valid points, old and new Pareto front and Batch 6 valid
points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
4.50 Batches 1 to 6 valid points, old and new Pareto front and final optimal points.111
4.51 Close-up of the final optimal points. . . . . . . . . . . . . . . . . . . . . . . 112

xi
List of Tables

1.1 Relation between tasks . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

2.1 Comparison between structural concepts of important features. . . . . . . . 17


2.2 First example of laminate notation. . . . . . . . . . . . . . . . . . . . . . . . 30
2.3 Second example of laminate notation. . . . . . . . . . . . . . . . . . . . . . 30
2.4 AbaqusTM units logic. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

3.1 Summary of the materials employed . . . . . . . . . . . . . . . . . . . . . . 57


3.2 Different types of skin considered for the study. . . . . . . . . . . . . . . . . 58
3.3 Balsa and Rohacell R
110 wf properties. . . . . . . . . . . . . . . . . . . . . 58
3.4 CE and GE properties. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
3.5 Set of parameters analysed in the first family of points. . . . . . . . . . . . 68
3.6 Core combinations for the stringers improvement. . . . . . . . . . . . . . . . 69
3.7 Steps followed after the first family of points are obtained. . . . . . . . . . . 70

4.1 Parameters used to perform the mesh convergence tests. . . . . . . . . . . . 81


4.2 Results of the convergence mesh tests. . . . . . . . . . . . . . . . . . . . . . 81
4.3 Parameters used to perform the buckling modes verification. . . . . . . . . . 87
4.4 Parameters of the equivalent structure. . . . . . . . . . . . . . . . . . . . . . 92
4.5 Trends observed in the initial tests. . . . . . . . . . . . . . . . . . . . . . . . 95
4.6 Main features of the two selected points of Batch 1. . . . . . . . . . . . . . 100
4.7 Core combinations for the stringers of improvement 3. . . . . . . . . . . . . 103
4.8 Stringer configurations for improvement 3 points. . . . . . . . . . . . . . . . 103
4.9 Main features of the five selected points of Batches 1 to 4. . . . . . . . . . . 108
4.10 Main features of the six selected points of Batches 1 to 5. . . . . . . . . . . 110
4.11 Main features of the final optimal points. . . . . . . . . . . . . . . . . . . . 112
4.12 Final Pareto front points. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113
4.13 Main features of the points 117, 222 and 246. . . . . . . . . . . . . . . . . . 113

7.1 Typical composites fabrication processes, products used and wastes.


Retrieved from [27]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

xii
List of Abbreviations and Symbols

Abbreviations

UPC Universitat Politècnica de Catalunya


ESEIAAT Escola Superior d’Enginyeries Industrial, Aeroespacial i Audiovisual de Terrassa
FE Finite Element
FEM Finite Element Method
FEA Finitie Element Analysis
CE Carbon fibre/ Epoxy
GE Glass fibre/ Epoxy
UAV Unmanned Aerial Vehicle
MTOW Maximum Take-Off Weight
ACC Air Cargo Challenge
DBF Design/ Build/ Fly
NACA National Advisory Committee on Aeronautics
MMC Metal Matrix Composite
CMC Ceramic Matrix Composite
PMC Polymeric Matrix Composite
CLPT Classical Laminated Plate Theory
FSDT First-order Shear Deformation laminated plate Theory
ODE Ordinary Differential Equation
PVW Principle of Virtual Work

xiii
Symbols

ρ Density
µ Poisson coefficient
E Young’s modulus
G Transverse modulus
U Displacement
UR Angular displacement
m Mass
I Second moment of area/ moment of inertia of plane area
ε Strain
γ Angular strain
σ Normal stress
τ Shear stress
F Force
A Area
N Traction
M Moment
λ Buckling constant
IF Failure index
Lsw Lift per semi-wing

xiv
Chapter 1

Introduction

1.1 Aim
The purpose of this thesis is the design of a moncoque structure for a wing with
composite materials. To achieve such a goal, thorough research on monocoque structures,
composite materials, and Finite Element Method (FEM) software will be necessary. Some
performance requirements for this structure will also have to be set.

1.2 Scope
The scope of the project, divided in groups, is the following:

Information research

• Historical precedents: what wing monocoque structures made of composite


materials have been made, are being made, and are intended to be developed.
What are the main trends and characteristics of this specific type of structures.

• Composite materials: get to know their main features, i.e. anisotropy, failure
mechanisms, constitutive models, etc. Be able to predict which combination of their
wide range of parameters will be best for the approach of the structure design, i.e.
know how to overcome buckling issues or how to reinforce an under shear stress
region by changing the laying sequence or the thickness of the layers they are made
of. Get an insight of the theories that explain their mechanical behaviour.

• Monocoque structures: get to know their pros and cons, what are the design
parameters to take into account and be able to use them when approaching the
structure of the wing.

• FEM software: with a commercial FEM software chosen, get the most accurate
result of the desired structure. Thus, a thorough examination of the solvers, models
(static, dynamic, buckling, fatigue, etc), and other features will be necessary.

Design process

• Structural requirements: set the requirements for the wing’s structure. The
maximum weight, the ultimate load (flight performance) and others, must be defined.
Also how the wing is going to be sized is important to be defined: static load, dynamic
load, buckling or fatigue.

1
Chapter 1. Introduction

• Ease of use of the FEM software: get used to work with the software selected
and be able to confront the design process of the wing monocoque structure made
of composite materials.

• Design results: obtain proof of a structure that fulfils the requirements settled. As
for any engineering problem, there will be more than one solution possible. Therefore
those multiple valid structures must be discussed and classified for its pros and cons,
function that may be able to perform or other important aspects that must be kept
in mind when developing an engineering project.

• Optimisation of the process: finally, think of a way in which the whole process
undergone could be improved in as much ways as possible, i.e. reduce the time
required, obtain more accurate results, automation of the process, etc.

Project overview

• Conclusions: extract conclusions of the work done and examine the final point of
the project and how it could be improved or developed from that point on.

• Impact: do a brief study of how the work done influences the field of study, as well
as the economical or environmental effect of the same.

Deliverables

Elaboration of all the deliverables of the project, which are:

• A report.

• If necessary, annexes of the extra information/work done.

• A budget.

1.3 Requirements
To accomplish the goal of the thesis the following requirements must be fulfilled:

• The wing structure dimensioning will just take into account static loads and
buckling phenomena. No dynamic loads like fatigue loads or aeroelastic loads will
be considered.

• The monocoque structure must have buckling limiting elements that reduce the
buckling length. Also it must have shear transmitting elements.

• The composite materials used must be fibrous based.

• The matrix used must be epoxy.

• The dimensions of the wing that will be designed will be: 120 cm of wingspan x 19
cm of chord (rectangular wing). The airfoil used is a CLARKY.

• Computer-based: improve technical characteristics or use a well-equipped computer


in order to run the simulations in the most optimal way.

2
Chapter 1. Introduction

1.4 Objectives
The main objectives of this thesis are:

• Obtain the best solution among all possible solutions that the set of designing
variables can yield. It is important to find at least the trends of these variables
towards the best solution, as the best solution can only be sought - and proved to
be the best one - by means of optimisation methods.

The secondary objectives of this thesis are:

• Achieve good knowledge about a FEM software.

• Get used to the design process of an engineering project.

1.5 Justification
The motivation behind the selection of this matter of study lies on two main aspects:
firstly, the subjects that have enthralled me the most throughout the degree have been
Aerospace Structures and Science of Materials. I have always been passionate about how
the right combination of materials and the proper geometry allow a structural system to
withstand loads that may even surpass by large its own weight.

And secondly, and the most decisive one for the making of this project, I have been
enrolled in an aeromodelling research team - Trencalòs Team - for two years now. The
experience of being part of a group of people that dedicates its precious time to explore
new technologies has motivated me a lot. Also, applying the knowledge gained during
the degree to compete against other universities in international competitions, has
pushed me to go deeper into the fields of study that I love the most.

Adding up to those rock solid reasons, this thesis will serve as a leap forward in
Trencalòs’ know-how, as the kind of structure that I am going to study and design has
never been developed by the team. That fact adds to my motivation and willingness to
pursue the goal of this thesis in a substantial way.

Next is detailed the identification of the need that this thesis will cover - from a wider
point of view to a more specific one, and the usefulness of the study that I am going to
perform.

1.5.1 Identification of the Need


When designing an aircraft’s structure, its whole is divided in substructures of the
different parts of the craft. Later, elements or other substructures are designed to join
together all the parts. Usually different groups of people are in charge of developing each
substructure. Likewise, each part’s structure is confronted at different times, following a
sequence.

Addressing the design of a wing’s structure is an essential task needed when an


aircraft is being designed, for it is the main lifting device of the plane as well as the main
limiting element in terms of its flight performance. Being so, the goal of the project is
justified as it can be used as a guideline for anyone who is willing to design a monocoque
wing structure made of composite materials.

3
Chapter 1. Introduction

A perfectly fitting example of that need is Trencalòs Team, whose main purpose is
designing unmanned aerial vehicles (UAV) to compete in aerospace engineering-related
international competitions. For that very reason, the design of a wing must be tackled
and, for major facts that will be exposed further in the text, a monocoque composite
material structure seems to be the most effective one to choose.

1.5.2 Usefulness of the Study


In terms of wing structures made of composite materials, big airliners do not use
full-made composite materials wing structures but instead they combine these materials
with historically and well-known metals that have been long used in the aerospace
industry. A full-made composite material wing structure in an aircraft is more common
in private/training, two/four-seat, single-engine aircrafts (e.g. the Cirrus SR20), as well
as in drones or radio-controlled aircrafts that usually are scaled up or simply have
smaller dimensions.

Monocoque structures - structures designed so all the loads are taken by the skin and
there is no internal framework to assist - are not the most used approach in the history
of aviation. Other techniques like the semi-monocoque (or stressed skin) have been more
employed as, for example, they tackle better the distribution of loads in the structure,
avoiding thus a probable stress concentration that may lead to fatal failure.

Despite that, monocoque structures do have their perks and have also been used in
aircrafts (e.g., the Lancair IV). Up to a certain dimension of the aircraft, monocoque
structures can be lighter than a semi-monocoque structure composed by an internal
skeleton plus the outer skin. Furthermore, as it is a continuous shell structure there is
plenty of space available that can be used for the possible necessity (e.g. fuel, baggage,
control surfaces deployment system storage, etc).

Figure 1.1: Lancair IV. Retrieved from [1].

So, aside from the important usefulness for Trencalòs Team that I have already
mentioned before, the the kind of aircraft that would best fulfil the wing’s monocoque
composite material structure intended to be studied and designed in this thesis, is a
small airplane: single-engine, two or four seats, maximum wingspan of 10-11 m
approximately. Therefore, this project can be useful for a company or a group of people
that want to undertake the design of an aircraft of such features and their inclined to
this kind of structure for the wing.

4
Chapter 1. Introduction

Coming back to Trencalòs Team’s interests, it is important to stress how the UAVs
that the team designs should look like, i.e, what pilar features must it have to maximize
the score to be obtained. To date, Trencalòs has only taken part in the biannual
international competition Air Cargo Challenge (ACC), but is currently seeking new
horizons as the SAE Aero Design (i.e, Society of Aerospace Engineers) or the DBF (i.e,
Design/Build/Fly) competitions. In these competitions, a UAV has to be designed under
some regulations in order to undergone a series of tests and the lightest and most
resistant structure can make the difference to win the contest.

It has to be taken into account that the aircrafts in the ACC have to carry as many
payload as possible and be as fast as possible in order to score the highest points. For
that very reason the weight is a crucial aspect. Likewise, the manoeuvrability of the craft
is also addressed meticulously. That’s why these UAVs have no bulky unnecessary parts
and the main structure is the wing (3-3.5 m wingspan). Figure 1.2 depicts the essence of
the airplanes of these competitions.

Figure 1.2: The b.ONE, Trencalòs Team’s aircraft that competed in the Air Cargo
Challenge 2015.

Historically, Trencalòs Team has always used plywood and balsa wood as the main
construction materials. In the last edition the team took a step up and changed to a
high-performance foam reinforced with carbon fibre. But still, monocoque composite
materials wing structures are the best for the ACC as the final results of 2015 edition
proved: the first and second teams in the final ranking used this technique and stood out
from the rest of teams.

For that very reason, and with the next contest in mind - Air Cargo Challenge 2017 –
the team has taken the opportunity of doing research projects about this subject.
During this semester a wing with this characteristics will be built by the team, advised
and powered by myself and another member who is doing his bachelor’s thesis of the
construction process.

To conclude, this project can take the team a step up and let it expand its know-how,
making Trencalòs Team more competitive for the next contest(s). Moreover, it will serve
me to gain more knowledge in the field of aerospace engineering that I am keener on.

5
Chapter 1. Introduction

1.6 Study’s Approach


To address this project I am going to follow three main steps or parts, as stated in
the scope section: information research or education in the field of study, design
process or evolution with the designing tools towards the final solution (with the prior
knowledge basis set), and a final project overview to give closure to the project.

In the information research I want to study the main features of each distinctive
characteristic of the wing structure: what parameters characterize a monocoque
structure and a composite materials structure and what are the trends they follow to
maximize the structures behaviour.

Doing so I will not tackle the design process with the FEM software completely blind
but I will be able to know which combination of parameters may be more promising than
others. As I will need time to learn how to use the software, a good information research
will be very useful in managing my time. In this part of the project, some requirements
will be necessary to be settled and the degree of difficulty of the final structure will rely
on them. So, when the time comes to define them, a good decision based on time left for
delivery of the project will have to be made.

Once the requirements are set, a model in the program will have to be made and
then work on it until some results that agree with the requirements and the project goals
are obtained. After that, a closer look to the process followed will be given and analysed
in order to propose improvements of any kind.

Finally, some conclusions will be extracted as well as the impact of the project at
different levels: economical, environmental or others.

The critical elements of this approach are the following: taking the exact necessary
time to accomplish the first part so there’s enough time to perform the design through
the FEM software; choosing adequate requirements for the structure in accordance to
the time left when that point of the project is reached; and being able to obtain a good
quality result with a FEM commercial software on time as this programs are very
sophisticated and complicated to use.

I feel this is the best way I can confront this project. The only disadvantage of the
approach of my study is not having enough certainty of the exact timing that every little
part is going to take me. That situation only leaves with the one possible way of action:
perseverance and hard work.

1.7 Planning
1.7.1 Brief Description of Tasks
The main tasks or work packages that must be accomplished in order to achieve the
goal of the thesis are the following:

• Project charter (DE): development of the project charter, document used to


outline the project objectives and constraints, the directions concerning the solution,
the in-scope and out-of-scope items, the reasons for undertaking the project and an

6
Chapter 1. Introduction

approximate planning of the same. It is very useful to define and make clear how
exactly the project is going to be done.

• Follow-ups (DE): during the time doing the project, some follow-up reports will be
necessary to be done. These will come handy to think of which point of the project
you are at and will help in the whole planning of the project.

• Information research (IR): prior to any designing task, whatever the type or
field of it, a foundation knowledge is necessary to be achieved or consolidated. For
that very reason, thorough research and study of the information more relevant to
the attainment of the project will be undergone. That is: composite materials,
monocoque structures, general theory of structures, FEM main theoretical basis and
how the method is developed (i.e. solvers, models, etc), structural regulations and
naturally the state of the art in order to grasp the trends and reasons in the history
of monocoque wing structures made of composite materials.

• Set the structural requirements (TD): before tackling the structures’ design,
a set of requirements that the structure must fulfil must be defined. The maximum
weight of the structure, the ultimate load it has to withstand or the maximum
displacement of the wing tip has to be defined. Also how the wing is going to be
sized is important to be defined: static load, dynamic load, buckling or fatigue.

• FEM software learning (TD): in order to achieve the goal of this project, an
analysis of the structure will have to be undergone in a commercial software. To do
so I will have to learn how the program works and get used to it. Accordingly, I
will have to read from beginning to end the program’s manual and extract the most
valuable information for my purpose.

• Wing structure design (TD): once I handle the program with ease I will develop
a model of the structure and work on it, through the iterations that are necessary,
and obtain the set of possible results, among the many combinations of parameters,
that fulfil the requirements.

• Design process optimization (TD): when the design process has been completed
and results have been accomplished, the whole process will be submitted to a close
examination seeking for ways of improving it: reduce the time needed, obtain more
accurate results, automation of the process, etc.

• Report writing (DE): as long as information is being read it will be summarized


in documents. Properly scheduled dates will be set to have certain parts of the final
report already written with the desired cohesion and style.

• Annexes and budget writing (DE): at the end of the project the extra
information that has been used must be written in the annexes and the budget of
the project must be elaborated.

• Conclusions and impact of the project (PO): at the end of the project some
conclusions will have to be extracted. Likewise, the environmental impact, as well
as the economical or technical impact of the project will have to be evaluated.

DE: deliverable; IR: information research; TD: technical design; PO: project overview

7
Chapter 1. Introduction

1.7.2 Detailed Relation Between Tasks


The relation between the tasks explained in the previous section will be depicted with
the following table:

Code of task Task identification Preceding task


IR Information research
IR-1 Composite materials
IR-2 Monocoque structures
IR-3 Theory of structures
IR-4 FEM
IR-5 Structural regulations
IR-6 Precedents
TD Technical Design
TD-1 Structural requirements settlement IR
TD-2 FEM software learning IR
TD-3 Wing structure design IR, TD-1, TD-2
TD-4 Design process optimisation IR, TD-1, TD-2, TD-3
PO Project overview
PO-1 Conclusions and impact of the project IR, TD
DE Deliverables
DE-1 Project charter
DE-2 Follow-up 1
DE-3 Follow-up 2
DE-4 Follow-up 3
DE-5 Follow-up 4
DE-6 Report writing-1 IR
DE-7 Report writing-2 IR, TD
DE-8 Annexes and budget IR, TD, DE-6, DE-7

Table 1.1: Relation between tasks

1.7.3 Scheduling
Finally, an allocation of the tasks in time has been done. To do so, a Gantt chart has
been used.

8
9
Figure 1.3: Gantt chart of the organisation of tasks in time.
Chapter 1. Introduction
Chapter 2

State of the Art

2.1 Historical Preamble


During the early years, aircraft structural design relied more on intuition and
empirical rules than theoretical principles or analysis. As time passed, aviation moved
beyond the isolated, individual inventor and attracted the attention of engineers. The
fields of elasticity, strength of materials, and structural mechanics were harnessed to
tackle flight structures problems. The information and knowledge derived from
structural theory enabled impressive improvements in airplane performance and
eventually established the subject of airframe structures as an engineering specialty.

Early airplane builders saw structural design and construction as a minor problem,
easily solved using commonly available materials and trial-and-error procedures. Of the
conventional materials available for construction of bridges and other stationary
structures, only steel and a few varieties of wood were considered suitable for flight
structures. Familiarity with woodworking and wood construction made wood the natural
choice of aviation pioneers. Wood was readily available, inexpensive, required few
specialized tools and could be easily worked by those with limited construction skills.

Early aviation construction could be called the


era of wood and piano wire since joining wooden
trusses and other components together to form a
truss-like structural framework was the easiest
solution to early aircraft structural problems.

Octave Chanute
(1832-1910) is the father of airplane structures.
Chanute was a successful civil engineer who, after
his retirement from a successful career as a railway
engineer, became interested in flight. Chanute used
his knowledge of structural engineering – he had
designed and built award winning truss railway Figure 2.1: Chanute’s biplane
bridges – to create the first biplane surfaces. glider made of truss construction.
This type of aerodynamic and structural design Retrieved from [2].
was adopted quickly by the Wright Brothers.

Truss construction, fabricated from simple “two-force” (traction-compression)


members that could be easily replaced when damaged, creates a stiff, lightweight
structure. Reliable methods for analysis and design of beams and trusses had been

10
Chapter 2. State of the art

available for decades, which made truss structures ideal for bi-plane and tri-plane
aerodynamic designs. Covering the truss-work with canvas or linen created an
aerodynamic surface that flew relatively well, given the low speeds possible with low
power engines. Bi-plane design created high drag, but at low speeds, no one cared.

The Wright Brothers were the first


true aeronautical engineers. They built
on the success of others such as Chanute.
Using other people’s data wasn’t easy
on that age, for both the reliability and
disponibility of the same. For instance,
some of Chanute’s aerodynamic data was
incorrect, and the Wrights, driven by the
uncertainty of believing or not the data,
built a wind tunnel to create their own
data set. This way of approaching the
design was very forward-thinking for their
time, and that led them to their famous
first ever successful controlled, powered
and sustained heavier-than-air human
Figure 2.2: The Wright flyer. Retrieved from flight, which was on December 17, 1903.
[3].
The Wright Brothers understood
structures so well that they arranged the fabric cover for their wings so that the weave
was oriented plus or minus 45 degrees with respect to the spars. This had the effect of
resisting in-plane shear forces. Their patent stated “These spars, bows, and ribs are
preferably constructed of wood having the necessary strength, combined with lightness and
flexibility. Upon this framework the cloth which forms the supporting surface of the
aeroplane is secured, the frame being enclosed in the cloth. The cloth for each aeroplane
previous to its attachment to its frame is cut on the bias and made up into a single piece
approximately the size and shape of the aeroplane, having the threads of the fabric
arranged diagonally to the transverse spars and longitudinal ribs .... Thus the diagonal
threads of the cloth form truss systems with the spars and ribs, the threads constituting
the diagonal members.”[28]

Besides,
the Wright’s structural engineering
tasks included flight safety. Two entries
in Wilbur Wright’s diary are typical: ”I am
constructing my machine to sustain about
five times my weight and am testing every
piece.” (1900) and ”We also hung it by the
tips and ran the engine screws with the man
also on board. The strength of the machine
seems ’OK’.” (December 2, 1903)[28] Figure 2.3: Bleriot’s airplane. The first
monoplane. Retrieved from [4].
About 1907 monoplane designs began to
appear. The pioneer of this type of design was Louis Bleriot. These new monoplane
designs had external wire supports attached to the wing and ending in a central post
above the fuselage and the landing gear below. Bleriot’s designs added an additional
post, half-way along the wing span. This gave his designs increased strength and

11
Chapter 2. State of the art

permitted of a greater wing span and thus reduced drag.

In the beginning aircraft structures were little more than cloth covered trusses and
frames. However, in 1911, Louis Bechereau built on an idea from Eugene Ruchonnet, a
Swiss enginner who had worked at a shop as foreman for the Antoinette Airplane
Company in France. Rather than simply mounting fabric onto a truss, Ruchonnet’s idea
was to use the airplane’s skin carry the structural load and eliminate the heavy truss. He
formed the fuselage out of multiple layers of wood to create a streamlined shape.

The wood layers were glued


together with their grains running
in different directions to strengthen
the skin. Ruchonnet called the new
technique monocoque construction,
or single-shell construction. Bechereau
used the monocoque fuselage,
together with a monoplane design
that was a streamlined, externally
braced, mid-wing monoplane. This
airplane, named the Deperdussin, won
the Gordon Bennett Cup in 1913 with
a speed of 125 mile per hour – only
Figure 2.4: The Deperdussin, the first airplane a decade after the Wright’s first flight!
with monocoque structure in the fuselage.
Retrieved from [5]. Later, when airplane size grew the
monocoque design was not suitable because it buckled under high loading. A variation of
the concept, a reinforced or semi-monocoque structure was created to create a
lightweight durable structure. This type structure is still used today, both for fuselage
design and wing design.

A box spar was created by constructing the main supporting member of the wing as
a long, narrow box beam that provided the strength and also torsional stiffness. While
cantilevered wings first were covered with fabric in the same way as biplane wing
structures, they began to use skins made of thin wooden veneers. These wooden-skinned
wings became known as stressed-skin wings because the veneers added strength while
they shaped the wing surface itself.

Three men contributed to the development of revolutionary aircraft structures and


airplanes with these structures before and after World War I. All three were German.
And the most distinguisehd one was Hugo Junkers, who was the first to choose metal
rather than wood for construction.

He recognized that streamlining airplane shapes was the key to increasing speeds
with limited engine power. And with metals, he was able to produce low-drag
monoplanes with high durability. Steel structures were too heavy - he realised that in his
first semi-monocoque wing structures, so he used aluminum alloys developed for Zeppelin
airships. His first all-metal, aluminum, cantilever wing airplane was the Junkers J.1.

By 1920 all of the essential structural concepts were in place so that structural design
could enable better, faster aircraft. In particular, semi-monocoque structural design had
enabled streamlined fuselages and low-drag cantilever wings. Metallic designs created

12
Chapter 2. State of the art

strong, robust, durable structures. The recognition of the importance of streamlining


began to take hold, as did the development of new, more powerful piston engines.

Science entered the process as speeds


increased and airplane designs improved.
German designers in particular profited
from the close relationship between technical
university faculty members and the aircraft
industry. The German rigid, lighter-than-air
airship industry (e.g, blimps, zeppelins, etc)
allowed structural designers to learn valuable
lessons, particularly in metal construction,
that were applied to airplane design.

The United States was slower to recognize


the needs for aeronautical research and
information transfer. Aviation enthusiasts in Figure 2.5: The Junkers J1. Retrieved
the United States were successful in launching from [6].
the National Advisory Committee on
Aeronautics, the NACA, in 1915. The NACA played a significant role in airplane
structures development in the 1920’s and 1930’s. They developed new structural analysis
methods and provided valuable test results to American industry. As a result, American
propeller driven commercial transports were the best in the world.

Jet propulsion was developed by British and German engineers in the decade before
World War II. After the war, a team of American engineers was sent to Germany to
collect aeronautical data. Some of this data related to German swept wing research and
development. This data was instrumental in creating the Boeing B-47 jet bomber. After
the development of the B-47 and its successor, the B-52, Boeing took a tremendous
financial risk and developed the Boeing 707. This airplane revolutionized air travel for
the last half of the 20th century and Boeing became the premier air transport company.

The next big step in aircraft


history came by the hand of NACA,
as they were the major participant in the
development of supersonic vehicles. On
October 14, 1947 Captain Charles Yeager
broke the sound barrier in the rocket
propelled Bell X-1. This kind of airplanes,
with small thickness to chord ratio,
high in-service temperatures, complex
and demanding load distributions
or even variable in-flight geometry (e.g,
the XB-70 supersonic bomber), supposed
a new challenge for structural engineers.
As a result of the great developments Figure 2.6: An American Airlines Boeing
on the field, in the early 1960’s manmade 707. Retrieved from [7].
fibrous materials such as fiber glass, boron
and carbon/graphite began to appear as part of a material referred to as “advanced
composites” to differentiate them from wood. One of the first uses was fiberglass
filament wound solid rocket casings for missiles such as the Navy Posideon Fleet Ballistic

13
Chapter 2. State of the art

Missile. These materials were used as woven cloth, tape or wound fibers embedded in a
matrix that held them together as a unit. These matrices were polymers, metals or
resins.

The individual fibers are very


strong due to the fact that there are few
imperfections in their structure. They are
also made of materials such as carbon that
has a very low molecular weight compared
to aluminum. Composites are stronger
and stiffer compared to metals, and also
the manufacturing processes are different.

An extraordinary feature of these


manufactured materials is that they can
be tailored to efficiently resist high loads
in one or more directions. This tailoring
feature was central to the proposal to
use laminated composite wing covers for a
Figure 2.7: Research aircraft X-29. Retrieved swept forward research aircraft, the X-29.
from [8]. The composites were constructed so that
when the wing bent upward it also twisted nose down to reduce the airloads and avoid
an aeroelastic instability called wing divergence. The X-29 was a remarkable aircraft
whose existence depended upon the ability to tailor the wing stiffness with advanced
composites.

In the 1970’s it became standard practice to use advanced composites, particularly


graphite fiber composites, in military aircraft structures. The reason most often given
was not tailoring, but weight savings. This weight savings often came with a large price
tag.

The manufacturers of large


commercial transports have generally lagged
their counterparts in the military and business
aviation areas. For instance, the Boeing 767,
a dominant part of the civil transport industry
has advanced composites, but only for items
that are not subject to high loading, such
as ailerons, elevators and landing gear doors.

That reticence on the part of commercial


transport manufacturers is changing
rapidly and becuase of the last airliner
models of Boeing and Airbus. The Boeing Figure 2.8: Rate of materials employed
787 and Airbus A380, make use of substantial in the Boeing 787. Retrieved from [9].
advanced composite materials - adding up to
almost 50 % of the materials employed - in the fuselage and wings, areas that are highly
loaded and flight critical.

Summing up, it is to say that the history of aviation has evolved from truss structures
mainly made of wood, to monocoque and semi-monocoque structures made of different

14
Chapter 2. State of the art

kinds of materials: at the beginnings of this structural concepts wood was the top-choice,
but through time it has been replaced first by metals and then by a combination of the
latter and composite materials.

2.2 Structural Concepts


During the 2.1.Historical Preamble, the three most used structural concepts in
aicrafts, as in any other field of engineering in which a structure is required, have been
named. These concepts or techniques are the following:

• Truss structures[29, ch. 2]: these structures are made up of individual straight
members which are connected at joints. The members are assumed to be connected
to the joints in a manner that allow rotation, and thereby it follows from equilibrium
considerations that the individual structural members act as bars, i.e. structural
members that can only carry an axial force in either tension or compression. Often
the joints do not really permit free rotation, and the assumption of a truss structure
then is an approximation. Even if this is the case the layout of a truss structure
implies that it can carry its loads under the assumption that the individual members
act as bars supporting only an axial force.

Figure 2.9: Example of a truss structure of an aircraft fuselage. Retrieved from [10].

• Monocoque1 structures: in this kind of structures, loads are supported only by


the skin of the object. No auxiliary elements are used to help in the withstanding of
loads. For that reason, this technique is also known as structural skin. Therefore,
the design of the skin must fulfill the load requirements. To do so, anisotropic
materials (e.g, composite materials) are employed and are properly oriented and
stacked in order to achieve such a goal. Their most common mode of failure is
buckling.

• Semi-monocoque structures: this technique is a half-way approach of the last


two. Instead of building a load-bearing internal skeleton or a load-bearing outer
skin, the idea behind is the following: using a partial skeleton that reinforces the
skin in the critical areas. The shell is still carrying much of the load, but failure is
best prevented. Monocoque structures tend to be lighter than semi-monocoque up
to a certain size, then this fact reverses.
1
The word monocoque is a French term for ”single shell”.

15
Chapter 2. State of the art

Figure 2.10: Monocoque vs. Semi-monocoque structure of a fuselage tail boom. Retrieved
from [11].

Truss structures are a good approach as the calculations and design methods are easier
and faster than the other two. The main reasons behind that fact are, first, the
”two-force” members composition of the structure, which simplifies a lot the load
transmission; and second, the huge know-how on this subject for historical reasons - i.e,
this technique has been widely used in a lot of fields as civil engineering for way before
the first succesful flight of history of the Wright brothers in 1903.

Truss structures resemble and are also known as a skeleton structures, in which the
skin has only a protective function or in the case of aircrafts, an aerodynamic function.
A way to make structures lighter than a skeleton structure is taking profit of the skin
that has to be employed either way, and stiffen it up so it carry the load by itself. In
that way we can get rid of the inner ”bone” system - and use a moncoque configuration.

Structural skin or monocoque structures can be incredibly light, but they also have a
few aspects that must be tackled very carefully: the least imperfection in the construction
of such a strucutre can easily be a source of mechanical failure; and the most important
one for this thesis, as the construction of the structure will be assumed perfectly delivered,
is that this structures suffer from buckling as their main mode of failure. To avoid this
phenomena, elements that limit the buckling lenght are required. As it can be seen in
Figure 2.10, the monocoque structure of the tail boom uses frames, even though the whole
of the loads is carried by the skin. The placement of these elements has the function of
reducing the buckling length between frames [30].
Another thing that must be considered is that monocoque structures are also
sensitive to stress concentrations. The reason behind that fact is the lack of
load-transmission members, which complicates the transmission of loads and, depending
on the geometry of the strucutre, concentration of the stress can easily occur.

Depending on the object or craft that is going to employ a monocoque structure,


sometimes the skin thickness needs to be too high leading to an increase of the weight,
which counteracts the great feature of lightness that these structures have.
Semi-monocoque structures solve this problem by using auxiliary elements that
reduce the load carried by the skin, stiffening it up without having to increase its
thickness, or in comparison with a monocoque structure, having a thinner thickness than
the latter. This elements obviously add weight but their load-bearing contribution is way
more noticeable, as they increase the moment of inertia of plane area (I) of the section
they are placed at [31]. In Figure 2.10 these elements are called stringers and they help
in the the withstanding of the beding moment and also reduce the buckling lengths along

16
Chapter 2. State of the art

the perimeters of the transversal sections.

Below we have a table with some of the more relevant features discussed up to this
point:

Truss Monocoque Semi-monocoque


Ease of design xxx x x
Lightness xx xxx xx
Number of elements xxx x xx
Sensitivity to buckling x xxx xx
Load transmission xxx x xx

Table 2.1: Comparison between structural concepts of important features.

Monocoque structures, the one kind that really matters for this thesis, have lightness
as their most remarkable feature. A clever design, that is, the best possible arrangement
of the skin - varying its composition, thickness and geometry - along the element, leads
to a light and resilient configuration without the need of any kind of support, i.e,
internal skeleton or external bracing with struts or wires.

Despite that, under severe loading states or due to big dimensions, the skin of this
type of structures needs to be thickened to overcome the issues that come with those
conditions. Doing so, the lightweightness is partly lost. Adding to that is the fact that
the buckling event of a portion of the skin can lead to a fatal failure of the whole
structure, as the skin is the one and only structural element. Sudden loss of strength in a
region due to failure effects results in a redistribution of loads in the rest of the skin.
This two characteristics of monocoque structures make them less appealing in front of
semi-monocoque structures. Firstly, because the increase of thickness will imply an
increase of weight; and secondly because semi-monocoque structures deal better with
buckling and with load distribution, as they add the extra elements in its configuration.

A great solution to the main mode of failure of monocoque structures, that is,
buckling, is the incorporation of elements to the structural skin. This can be seen as a
contradiction, as the essence of this sort of structure is that the only load-bearing
element is the skin and it doesn’t need any extra support. The addition of this elements
has the purpose of reducing the buckling length at the skin and also transmit and
distribute stresses from and to other parts of the skin so these do not concentrate at any
specific region, which may cause a breakage of the same - i.e, failure of the skin.

For the case that will be undertaken in this thesis - the design of a monocoque wing,
these elements can be two: ribs or spar(s). On the one hand, ribs help in the
transmission of loads between the intrados and extrados skin of the wing. They mainly
transmit shear forces, avoiding thus a possible stress concentration at either the leading
edge or the trailing edge of the section. They also help in giving torsional strength to the
section at which they are placed, even though the skin has a good torsional resitance - as
it is a closed profile. And the placing of ribs will reduce the buckling length in the
longitudinal direction (along the span), helping the under-compression part of the skin,
the extrados, due to the bending moment that is alaways generated in a wing caused by
the lift distribution.

Another way of enhancing the buckling resistance of the extrados is by enlarging the

17
Chapter 2. State of the art

I and the stiffness of the section with a stringer-like making of the skin. That is, design
the skin with different layer sequences along the extrados, providing stiffened regions -
i.e. by putting high stiffness materials like carbon or using a soft material as a core to
increase the I, known as sandwich skins.

Then, a spar or multiple spars - placed preferentially at the highest thickness region
of the profile, will help mainly in the transmission of the shear force between the
extrados and intrados skins. It has to be taken into account that neither the spar(s) nor
the ribs have a load-bearing function. Their only purpose is the transmission of loads
and the limitation of the buckling lenght. Dealing with the dissipation of the stresses
generated on the wing is undertaken only by the skin.

Summing up, monocoque structures may use extra elements and still be monocoque
structures. The skin is the only part that will bear the loads to which is exposed the
wing, and the arrangement of ribs or spars it is only done with the purpose of reducing
the chances of failure - they limit buckling and avoid stress concentrations. For this very
reason, and despite being persisten on this, the study of the compostion of the skin is
very important for this kind of structures. This subject will be tackled in the next
section, 2.3.Composite materials.

2.3 Composite Materials


The term composite2 could mean almost anything if taken at face value, since all
materials are composed of dissimilar subunits if examined at close enough detail. But in
modern materials engineering, the term usually refers to a matrix material that is
reinforced with particles or fibres (reinforcement).

The definition of Zagainov [15] of this type of materials is the following:


”volumetrically formed special combinations of two or more components dissimilar in
form and properties, exhibiting clear boundaries between components, using the
advantages of each component. Whatever the material, a composite is formed by a
matrix (binder) or low-modulus material and reinforcing elements with strength and
stiffness properties 10 to 1000 times higher than those of the matrix.”

Many composites used today are at the leading edge of materials technology -
fibre-reinforced ones being the most widely used, with performance and costs appropriate
to ultrademanding applications such as aircrafts. But heterogeneous materials combining
the best aspects of dissimilar constituents have been used by nature for millions of years.
Ancient society, imitating nature, used this approach as well: straw was employed to
reinforce mud in brickmaking, without which the bricks would have almost no strength.

The fibres used in modern composites have strengths and stiffnesses far above those
of traditional bulk materials (e.g, wood, aluminium, and even some steels). For instance,
the high strengths of the glass fibres are due to processing that avoids the internal or
surface flaws which normally weaken glass; and the strength and stiffness of the
polymeric aramid fibre is a consequence of the nearly perfect alignment of the molecular
chains with the fibre axis.

2
The word composite is used as a synonym of composite materials.

18
Chapter 2. State of the art

Of course, these materials are not generally usable as fibres alone, and typically they
are impregnated by a matrix material that acts to transfer loads to the fibres, and also
to protect the fibres from abrasion and environmental attack. The matrix dilutes the
properties to some degree, but even so very high specific (weight-adjusted) properties are
available from these materials.

The fibres may be oriented randomly within the material, but it is also possible to
arrange for them to be oriented preferentially in the direction expected to have the
highest stresses. Such a material is said to be anisotropic - different properties in
different directions, and control of the anisotropy is an important means of optimizing
the material for specific applications.

Composite materials have a wide range of other great properties besides anisotropy,
despite the latter being the most important. In the following sections, a discussion of the
features of composites and also some of the different classifications of them will be
delivered. But prior to that, it is worth mentioning how composite materials influence
the structure design, and specially an aircraft’s structue.

A great feature associated with the development of composite structures is the


possibility to design concurrently both the structure and the material for the latter. As a
rule, composite materials do not exist separately from the structure and specific
fabrication practice. Manufacturing processes possibilities are broad but limited.
Conversely, the restrictions of those processes are the ones that play a much more
important role in the design of composite structures than in that of metal structures.

Designers that use isotropic materials must select an existing material and once taken
that decision, they can jump into the sizing of the part or element. On the other hand,
when composite materials are employed, the designer builds the material according to
the needs defined by the functional element.

Experience of composite application shows that, provided the design and fabrication
problems are adequately taken into account, the composite structure usually has much
lower number of parts, units and, especially, connecting elements. A high material
utilization factor, high potentialities for automation of the production process and
robotization, decrease the labour expenditures and the cost of production.

Their application in flying vehicles structures has led to an increase of weight


efficiency and a noticeable improvement of their flight performance. This increase
directly affects its fuel efficiency. Hence, in assessing the total expenses associated with
aircraft service time, the use of the composite structures may be more economically
profitable as compared to an aircraft made of conventional metallic alloys despite the
existing high cost of composites.

Employing composites causes the so-called “cascade effect”: smaller weight → smaller
lift → smaller wing → drag reduction → required thrust reduction → smaller engine weight
→ fuel reserve reduction → ultimate load reduction. “Investigations indicate that 1 kg
of weight saved during design results in reduction of the take-off weight by 4-5 kg, being
able to develop an aircraft with take-off weight of about 14 tons instead of 18 tons for an
all-metal aircraft.” [15, p. 6]

19
Chapter 2. State of the art

2.3.1 Main Properties of Composites


The most outstanding and worth mentioning properties of composites are the following:

• Weight efficient: as it has been emphasised previously, composites are very


lightweight (see Figure 2.14 a)) - low densities - and their application always entails
less elements to build the same structure as it would take to with traditional
materials like metals or wood - thus, reducing the overall weigth.

• Less subelemetns required: composites have the advantage of needing less joints
and fixations systems, thus reducing the number of parts (simplicity of the final
product) and less machining time (reduction of manufacturing costs).

• Complex geometries: any type of geometry can be achieved by moulding


(formation process).

• Anisotropy: physical properties are different depending on the direction of the


applied force.

• Tailor-made mechanical properties: thanks to the control of the anisotropy,


enhancement of the mechanical properties along a specific direction can be achieved
by means of proper fibre orientation.

• Good fatigue resistance: composites withstand high strengths under fatigue loads
and up to higher number of cylces than metals. This enhanced life that entails
savings in the long-term costs of the product. Their specific fatigue strength (σ/ρ)
is three times higher than that of an aluminium alloy and two times higher than
that of a high-strength steel or titanium alloy. Besides, up to a great number of
cycles (∼ 106 ), composite materials fatigue strength is equal to 90% of their static
tensile strength while for aluminium alloys and steel it is just equal to 35% and 50%,
respectively. See Figure 2.11.

Figure 2.11: Fatigue curve of an alluminium alloy (a) and a unidirectional composite (b).
Retrieved from [12].

• Mechanical properties: high tenisle fracture strength and modulus of elasticity.


Even though their mechanical properties are of the same order of magnitude as some
metals, their specific mechancial properties are way better that that of metals thanks
to their lightweightness. See Figure 2.12 and Figure 2.14 b) and c).

20
Chapter 2. State of the art

Figure 2.12: Comparison of different composites and an aluminium alloy specific


properties. Retrieved from [12].

• Do not yield: unlike metals, the elastic limit corresponds to the rupture limit. See
Figure 2.13.

Figure 2.13: Stress-strain curves of a metal and a composite. Retrieved from [12].

• Good corrosion resistance: less inspections required that leads to savings on


maintenance costs. It is fair to say that composite materials do not corrode, except
in the case of aluminium with carbon fibre in which galvanic phenomenon creates
rapid corrosion.

• Ageing: the main originators are heat and moisture.

• Sensitivity to chemicals: no sensitivity to chemicals used in engines like grease,


oils, hydraulic liquids, paints and solvents, petroleum, etc. However, they are sensible
to cleaners for paint, which attack epoxy resins.

• Impact resistance: they have medium-to low-level impact resistance (inferior to


that of metallic materials). Aramid fibres are an exception to that as they exhibit
great impact resistance. For that reason they are widely used to protect regions on
the aircraft that are susceptible to receive an impact - e.g. leading edge of the wings,
cockpit, etc.

21
Chapter 2. State of the art

• Fire resistance: excellent fire resistance as compared with light alloys of identical
thicknesses. However, the smokes emitted from the combustion of certain matrices
can be toxic.

• Price: comoposites are very expensive materials. See Figure 2.14 d).

Figure 2.14: Comparison of characteristics of different materials: a) density (kg/m3 ) b)


tenisle fracture strength (MPa) c) modulus of elasticity (MPa) d) price per unit mass.
Retrieved from [12].

2.3.2 Classification of Composites


Composite materials are commonly classified at two distinct levels: one first level of
classification is made with respect to the matrix constituent, and the second level is
made with respect to the reinforcement form.

In the classfication by matrix we find three main types of composites:

• Metal matrix composites (MMCs): materials consisting of a metallic matrix


combined with a ceramic (oxides, carbides) or metallic (lead, tungsten, molybdenum)
reinforcement.

• Ceramic matrix composites (CMCs): materials consisting of a ceramic matrix


combined with a ceramic (oxides, carbides) reinforcement.

• Polimer matrix composites (PMCs): materials consisting of a polymer (resin)


matrix combined with a polymeric (aramid fibre, glass fibre), ceramic (oxides,
carbides) or metallic (lead, tungsten, molybdenum) reinforcement.

It must be noted the following: as ceramics are materials with a higher elasticity
modulus, or Young’s modulus (E) - less deformation by an increase in the stress in the
elastic region - than metals and polymers, the latter cannot be used as reinforcement for
a CMC. The same happens with MMCs and polymers. And lastly, PMCs, having the

22
Chapter 2. State of the art

lowest E of the three families, can be reinforced with any type of material.

Typical value of E for a ceramic like silicon carbide (SiC) is E = 300 GPa; for a
metal like an steel alloy is E = 207 GPa; and for a polymer like epoxy is E = 2.41 GPa.

Then, in the classification by reinforcement we find three main types of composites as


well:

• Fibre-reinforced composites: are composed of fibres embedded in matrix


material. Such a composite is considered to be a discontinuous fibre or short
fibre composite if its properties vary with fibre length. This type or fibre reinforced
composites can have the fibres aligned or randomly oriented. On the other hand,
when the length of the fibre is such that any further increase in length does not
increase the elastic modulus of the composite, the latter is considered to be
continuous fibre reinforced. Fibres are small in diameter and when pushed
axially, they bend easily although they have very good tensile properties. These
fibres must be supported to keep individual fibres from bending and buckling.

• Structural composites: materials comprised of layers bonded together - also


known as laminatively or in a determined order for many times as required. Sandwich
structures fall under this category as well.

• Particulate-reinforced composites: are composed of particles distributed or


embedded in a matrix body. The particles may be flakes or in powder form.

Figure 2.15: Composites classification depending on the reinforcement type. Retrieved


from [13].

The most commonly used composite materials in the aerospace industry are the ones
with polymeric matrix and fibrous reinforcements. They are almost always in the form of
laminates, that is, monolayers with the fibres oriented in different directions, that are
then stacked up following a sequence .

Polymers are ideal materials as they can be processed easily, are very lightweight,
and have desirable mechanical properties - obviously, metalic and ceramic materilas have
higher mechanical properties but they are not as lightweight as polymers.
Two main kinds of polymers are thermosets and thermoplastics. Thermosets are the
most popular among the fibre composite matrices without which, a lot of research and
development in structural engineering field could have not been possible. Aerospace
components, automobile parts, defense systems etc., use a great deal of this type of fibre
composites. Epoxy matrix, among other thermosets like phenolic polyamide resins or
polyester, is the most used and well-known resin as its properties stand oud from the rest

23
Chapter 2. State of the art

of thermoset resins.

And among the wide variety of reinforcements that a polymeric matrix can use, the
ones that are more used in modern aircrafts and have the highest mechanical properties,
are continuous polymeric fibres - e.g. carbon, glass or aramid fibres being the
most employed. A set of long continuous fibres embedded in a matrix in one direction
constitute the basic structural element of composite construction: the unidirectional
layers or monolayers.

This basic structural element can be formed with fabrics instead of unidirectional
layers. Fabrics are two unidirectional sets of fibres weaved together perpendicular to
each other - i.e. are bidirectional elements. Being so, the main structural element in
composite designing is more known as laminae. When a laminae is formed from a fabric
embedded in a matrix material, it can be considered like two laminae formed with
unidirectional layers of half the thickness each and stacked up at 0/90◦ .

A laminae by itself can hardly form a functional material that has to fit some
strength and stifness requirements. That is why, a combination of several monolayers -
with different thicknesses and properties - is employed, making up laminates. If you
add a soft material in between a laminate - to gain second moment of inertia of the
section or other purposes, you obtain a so-called sandwich panel.

Layers of different materials may be used, resulting in a hybrid laminate. The


individual layers generally are orthotropic - that is, with principal properties in
orthogonal directions, or transversely isotropic - with isotropic properties in the
transverse plane. The laminate then will exhibit anisotropic, orthotropic, or
quasi-isotropic properties (depending on the sequence of stacking and orientation of the
monolayers). Quasi-isotropic laminates exhibit isotropic - that is, independent of
direction - inplane response but are not restricted to isotropic out-of-plane (bending)
response.

2.3.3 Laminae and Laminates


As it has been mentioned in the previous section, laminates of polymeric matrixes with
fibrous reinforcements are the composites that are mostly employed in the aerospace sector
- and thus, are the ones that must be studied in this thesis. As a laminate is made up by
several laminae, in this section a discussion of laminates and laminae main characteristics
will be delivered. Recall that in Section 2.3.2. Classification of Composites we have
explained that laminaes can be either made by unidirectional fibres or fabric, but in either
way the laminae can be treated as a composition of laminae.
Next to this section, in 2.4. Laminae Mechanical and Physical Properties, the
deduction of the calculation of the properties of laminae will be delivered.

Prior to the discussion on laminae and laminates fundamental knowledge, it is


important to know which elastic properties are needed to be known to characterize an
anisotropic material.

The relation between strains and stresses for an isotropic material under plane stress -
plane stress state is very important for the development of the Classical Laminated Plate
Theory (see Section 2.5. Classical Laminated Plate Theory) - only counts with two
independent variables.

24
Chapter 2. State of the art

    
 εx  1/E −ν/E 0  σx 
εy = −ν/E 1/E 0  σy (2.1)
γxy 0 0 1/G τxy
   

This two independent variables are E and the Poisson coefficient (ν), as it is important
to recall that the shear modulus (G) is related with those two variables in the following
E
way: G = 2(1+ν) (2.2).
In plane stress there is also a strain in the z direction due to the Poisson effect. This
strain component will be ignored in the sections to follow.

Figure 2.16: Anisotropic material. Retrieved from [14].


When E1 6= E2 6= E3 , the material is orthotropic. It is common, however, for
the properties in the plane transverse to the fiber direction to be isotropic to a good
approximation (E2 = E3 ); such a material is transversely isotropic - laminae, as it will be
discussed later on, will be considered transversely isotropic. The elastic constitutive laws
must be modified to account for this anisotropy, and the following form is an extension of
the usual equations of isotropic elasticity to transversely isotropic materials:

    
 ε1  1/E1 −ν21 /E1 0  σ1 
ε2 = −ν12 /E2 1/E2 0  σ2 (2.3)
γ12 0 0 1/G12 τ12
   

The parameter ν12 is the principal Poisson’s ratio; it is the ratio of the strain induced
in the 2-direction by a strain applied in the 1-direction. This parameter is not limited to
values less than 0.5 as in isotropic materials. Conversely, ν21 gives the strain induced in
the 1-direction by a strain applied in the 2-direction.
Since the 2-direction (transverse to the fibres) usually has much less stiffness than the
1-direction, a given strain in the 1-direction will usually develop a much larger strain in the
2-direction than will the same strain in the 2-direction induce a strain in the 1-direction.
Hence we will usually have ν12 >ν21 . There are five constants in equation (2.3) (E1 , E2 ,
ν12 , ν21 and G12 ). However, only four of them are independent; since the matrix, named
S, is symmetric, we have ν21 /E2 = ν12 /E1 .

ε=S·σ (2.4)

σ =D·ε (2.5)
where D = S −1 is the stiffness matrix. The simple form of equation (2.3), with
zeroes in the terms representing coupling between normal and shearing components, is
obtained only when the axes are aligned along the principal material directions; i.e.
along and transverse to the fibre axes. If the axes are oriented along some other
direction, all terms of the compliance matrix will be populated, and the symmetry of the

25
Chapter 2. State of the art

material will not be evident. If for instance the fibre direction is off-axis from the loading
direction, the material will develop shear strain as the fibres try to orient along the
loading direction. There will therefore be a coupling between a normal stress and a
shearing strain, which does not occur in an isotropic material.

Figure 2.17: Rotation of axes. Retrieved from [14].


The transformation when the fibre direction is off-axis from the loading direction is
the follwing [14]:

σ1 = σx cos2 θ + σy sin2 θ + 2τxy sin θ cos θ




σ2 = σx sin2 θ + σy cos2 θ − 2τxy sin θ cos θ (2.6)
τ12 = (σy − σx ) sin θ cos θ + τxy (cos2 θ − sin2 θ)

where θ is the angle from the x axis to the 1 (fibre) axis. These relations can be written
in matrix form as

   2
s2
 
 σ1  c 2sc  σx 
σ2 = s
 2 c2 −2sc  σy (2.7)
τ12 −sc sc c2 − s2 τxy
   

where c = cos θ and s = sin θ. This can be abbreviated as σ 0 = T σ, where T is the


transformation matrix in brackets above.

2.3.3.1 Laminae

Figure 2.18: Laminae representation in space. Retrieved from [12].

They are composed by the reinforcing fibres, the binder (matrix) and the surface of
contact between the two. Each part has clearly differentiated functions:
• The stiff fibres bear the major stresses in loading, providing strength and stiffness in
the direction of fibre orientation. The total strength and stiffness of the composite
material are determined mainly by the properties of the fibres, their dimensions,
orientation and content by volume.

26
Chapter 2. State of the art

• The yielding matrix filling the interfibre space transmits stresses to the individual
fibres owing to the tangential stresses applied along the fibre-matrix interface, and
takes up the stresses acting in.

• The structural unity of the composite is ensured by bonds at the fibre-matrix


interface. This interface is an important item of a fibre composite, influencing the
physical and mechanical properties of the material. The importance of the interface
is ensuring structural unity. When poor adhesion occurs and/or cavities are created,
the result is separation of the components in loading and drastic reduction of the
mechanical properties.

The level of strength and stiffness properties is determined by the volume content of
fibres in the composite, i.e. the higher the volume fraction, the greater are the loads that
can be withstood. Actually, this relationship only holds true to a certain limit. The
maximum volume fraction of cylindrical fibres that can be packed in a composite
amounts to about 90%. However, at a volume fraction of fibres over 20%, the composite
properties experience drastic degradation. This happens due to the fact that the matrix
is no longer able to soak and impregnate the fibre bundles, leading to deteriorating
adhesion of the fibres with the matrix and resulting in formation of cavities in the
composite. The optimal volume content of fibres in most of fibre composites is 50-60%.
In the same way, the laminae elastic and strength properties in the longitudinal direction
increases monotonically as the stiffness and strength of the reinforcing fibres increase.

The matrix mechanical properties are


decisive in shear, compression along the fibres and loading
with the normal stresses in different directions from fibre
orientation. The adhesion at the interface also matters
for this properties. Then, with respect to resistance
to fatigue failure, it is the matrix that primarily
determines this property. In addition, the matrix protects
the fibre against detrimental environmental effects
and prevents strength reduction due to fibre abrasion.

There are two effective methods to accomplish better


shear and compression properties. The first consists
on an improvement of the stiffness and strength of
the matrix by means of whiskerization, i.e. introduction
of filamentary crystals into the matrix occupying Figure 2.19: Variation of
part of the interfibre space. The degree of whiskerization the elasticity modulus with
has an optimum value that depends on the matrix type. θ from the longitudinal
And secondly, to strengthen the bonds at the interface, El (fibre mudulus) to
the fibre can be subjected to surface treatment in order the transverse Et (matrix
to increase its surface area and its roughness leading to a modulus). Retrieved from
higher molecular interactions at the interface. “Treatment [12].
of the fibres substantially increases the shear strength and
simultaneously increases the strength by 20-30% in the case of longitudinal tension owing
to the strengthening of the reinforcing fibres”.[15, p. 26]

Laminae are anisotropic materials. Applying a biased load (with an angle θ with
respect to the fibres) will mean that the properties that will exhibit the laminae will be
in between the ones of the fibres and the matrix. Figure 2.19 depicts this phenomena.

27
Chapter 2. State of the art

So, before discussing the main characteristics of laminates, it is important to highlight


the following: a laminae can be treated as a transversely isotropic material. In the
direction of the fibres, the laminae is highly influenced by the fibre properties; and in the
directions perpendicular to the fibres - where the matrix abounds, the properties are more
E3
influenced by the matrix properties. That is traduced as E2 = E3 and G23 = 2(1+µ 23 )
[20].

2.3.3.2 Laminates
Laminates are composed by two
or more unidirectional layers with different
thicknesses and orientations in space in a
certain sequence of laying. The same concepts
explained in the unidirectional layer section
apply here for the matrix, the reinforcement
and the interface between the two.

These type of
composite materials are the most used in the
aerospace sector as aircrafts are subjected to
a wide range of loads and operate in combined
stress-strain conditions. Unlike unidirectional
layers, laminates can display great properties
in more than one direction, even though there
is always one direction properties standing
Figure 2.20: Laminate example with
out over the rest. The latter is due to the
0◦ , -θ◦ , θ◦ , 90◦ layer arrangement.
optimised-aimed design of composites in which
Retrieved from [15].
the main loads direction is most taken care of.

Complex fibre arrangement with a large number of differently oriented layers may
fulfil the structure requirements for which the composite laminate has been designed.
However, investigations indicate that in many cases, only four directions, i.e. 0◦ , 90◦ ,
+45◦ , -45◦ need to be employed [15]. With the proper variation of the volume fraction of
those 4 layers, the sequence of laying and thickness selection, almost every loading state
can be overcome. A clear example is shown below in Figure 2.21.

(a) Stiffness characteristics of a laminate (b) Strength characteristics of a laminate


varying the volume fraction of 0◦ , 90◦ and ±45◦ varying the volume fraction of 0◦ , 90◦ and ±45◦
layers. layers.

Figure 2.21: Variation of properties with volume fraction modification of the different
layers. Retrieved from [15]

28
Chapter 2. State of the art

In laminated composites, when


part of the fibres are aligned with the direction of
the load, there is a linear dependence of the strain
on load up to break and creep is non-existent.
However, when the load is not aligned with any
fibre direction, creep occurs intensely, being the
dependence completely nonlinear at high strains.

Accounting for the above mentioned


(see Figure 2.22 ), it is almost mandatory to use a
certain number of layer aligned with the direction
of the acting forces in the structure that uses
multilayer composites. When the load field has
a wide range of variation, multilayer composites
must at least have three different oriented types
of layers. Doing so, the fibres form geometrically
rigid triangular elements, unloading the
matrix exhibiting high viscosity in the interlayers.

The integrity of fibres under tension and


compression depends on the matrix stiffness,
deformability and strength. To achieve that, a Figure 2.22: Deformation due to the
rational combination of fibre and binder is necessary. anisotropy of composites.Retrieved
Laminae integrity means that laminated from [12].
composite materials, being composed of multiple
layers, have also integrity granted [12]. A proper design results in low deformability of
the fibres perpendicular to the load thanks to the ones that are aligned with it. Besides,
the deformability of unidirectional layers is actually pretty low, adding to the robustness
of the laminated composites.

The last thing to comment about laminates is the the denotation of the stacking
order. It is described as follows:

Supposing that the laminate is placed in space (x,y,z) with its midplane in the (x,y)
plane (value of z = 0), the first ply to be named is the lowest ply on the side z < 0.
Then the stacking sequence is named moving towards the uppermost ply of the side z >
0. In doing so each ply is noted by its orientation. The successive plies are separated by
a slash (/). And the grouping of too many plies of the same orientation is noted with an
index indicating the number of identical plies of that orientation.

29
Chapter 2. State of the art

Figure 2.23: Laminate representation in space. Retrieved from [12].

A laminate is symmetric (midplane symmetry property) when the stacking of plies on


both sides starting from the middle plane is the same. Symmetry is indicated with a
subscript ”s”. If the number of layers of the laminate is odd, the layer which falls at the
midplane is noted with a line over the orientation of the same (xx).

• Example:

Ply number Orientation (◦ ) Conventional notation


1 90
2 0
3 0
4 -45
5 45
midplane (90/02 / − 45/45)s
6 45
7 -45
8 0
9 0
10 90

Table 2.2: First example of laminate notation.

• Example:

Ply number Orientation (◦ ) Conventional notation


1 0
2 45
3 -45
4 (midplane) 90 (0/45/ − 45/90)s
5 -45
6 45
7 0

Table 2.3: Second example of laminate notation.

30
Chapter 2. State of the art

Having a midplane is an interesting feature for a laminate in manufacturing terms.


After plies have been cured at high temperatures, plies have a trend to contract differently
depending on the fibre direction. That leads to residual stresses after complete cooling.
The same thing happens in a lot of loading cases due to the orthotropy of composite
materials, i.e. coupling of forces due to different properties in each direction. But this
phenomenon is counteracted with midplane symmetry, as residual stresses of both sides
of the laminate cancel each other out.

2.4 Laminae Mechanical and Physical Properties


As unidirectional layers are the basic structure of laminated composite materials, it is
mandatory to know the physical and mechanical characteristics of them.

• Mass fraction

mass of f ibres
Mf = (F ibre mass f raction) (2.8)
total mass
mass of matrix
Mm = (M atrix mass f raction) (2.9)
total mass
Mm = 1 − M f (2.10)

• Volume fraction

volume of f ibres
Vf = (F ibre volume f raction) (2.11)
total volume
volume of matrix
Vm = (M atrix volume f raction) (2.12)
total volume
Vm = 1 − Vf (2.13)

Note that mass fraction can be obtained from volume fraction. If ρf and ρm are the
densities of the fibre and matrix, we have

Vf ρf
Mf = (2.14)
Vf ρf + Vm ρm

Vm ρm
Mm = (2.15)
Vf ρf + Vm ρm
• Density of a monolayer

ρ = Vm ρm + Vf ρf (2.16)

• Ply thickness: taking into account the grammage or weight per unit area of the
monolayer mof

mof
h · 1(m2 ) = T otal volume = T otal volume ·
F ibre volume · ρf
mof
h= (2.17)
Vf ρf

31
Chapter 2. State of the art

The thickness can be also expressed in terms of mass fraction of fibres:

  
1 1 1 − Mf
h = mof + (2.18)
ρf ρf Mf

The monolayer mechanical constants (elasticity constants) of most significance are


El , Et , Glt and νlt , where l stands for longitudinal and t stands for transverse. It is
important to account that, following the coordinate system of Figure 2.24, Et = Ez as
monolayers can be considered transversely isotropic materials. This constants are
going to be deduced as a function of the matrix (m) and fibre (f ) moduli, as well as their
respective volume fractions. [12]

• Longitudinal modulus El (along the fibre direction)

Figure 2.24: Idealisation of monolayer of fibre and matrix with longitudinal force applied.
Retrieved from [12].

A force in the longitudinal direction is applied. The following hypothesis are taken:

(a) Both matrix and fibre have the same longitudinal strain value εl .

(b) There is a freedom of movement in the direction perpendicular to the monolayer


(direction z) plane on the interface between the matrix and fibres εzm 6= εzf .
ε
We can then say that σlm+f = FA = F
(εm +εf )·1 = σlm (εmε+ε
m
f)
f
+ σlf (εm +εf)
(2.19), which
can be traduced into the following:

σlm+f = σlf Vf + σlm Vm (2.20)

where Vf and Vm are the volume fraction of the fibre and matrix, respectively. Now,
considering that each material – matrix and fibre – are isotropic materials, and taken into
account their behaviour law (the Hooke’s law), we can state the following:

El εl = Ef εl Vf + Em εl Vm (2.21)

then,

32
Chapter 2. State of the art

El = Ef Vf + Em Vm (2.22)

• Poisson coefficient νlt

Considering again the load applied to deduce the longitudinal modulus El , the
transverse strain for the composite can be written as

νlt
εtm+f = − σl = νlt εl (2.23)
El m+f
The transverse length variations can be written as

∆(εm + εf )
εtm+f = = εtm Vm + εtf Vf (2.24)
εm + εf

And applying equation (2.23), and doing the same for both the transverse strain of
the fibre and matrix, we get to

−νlt εl = −νm εl Vm − νf εl Vf

νlt = νm Vm + νf Vf (2.25)

• Transverse modulus Et (perpendicular to the fibre direction)

Figure 2.25: Idealisation of monolayer of fibre and matrix with transverse force applied.
Retrieved from [12].

A force in the transverse direction is applied. At the interface between the two
materials, the following hypothesis are taken:
(a) Freedom of movement in the l direction εlm 6= εlf .

(b) Freedom of movement in the z direction εzm 6= εzf .

(c) Both materials are under the same state of stress σt .

33
Chapter 2. State of the art

As stated previously:

σt ∆(εm + εf )
εtm+f = = = εtm Vm + εtf Vf
Et εm + εf

where Vf and Vm are the volume fraction of the fibre and matrix, respectively. Now,
considering that each material – matrix and fibre – are isotropic materials, and taken into
account their behaviour law (the Hooke’s law), we can state the following:

σt σt σt
= Vm + Vf (2.26)
Et Em Ef

then,

1 Vm Vf
= + (2.27)
Et Em Ef
or

" #
1
Et = Em Em
(2.28)
(1 − Vf ) + Ef Vf

• Shear modulus Glt

Figure 2.26: Idealisation of monolayer of fibre and matrix with shear force applied.
Retrieved from [12].

A shear stress τlt is applied. The following hypothesis are taken:

(a) The angular strains of the matrix and fibre are different γltm 6= γltf .

(b) Both materials are under the same state of stress τlt .
τlt
Keeping in mind that γlt = Glt (2.29), we can then say that:

γltm+f (em + ef ) = γltm em + γltf ef (2.30)

34
Chapter 2. State of the art

where em and ef are the thickness of the matrix and fibre respectively. Equation (2.30)
can be rewritten as

γltm+f = γltm Vm + γltf Vf (2.31)

where Vf and Vm are the volume fraction of the fibre and matrix, respectively. Now,
considering that each material – matrix and fibre – are isotropic materials, and taken into
account the previous stated relation between the angular strain and the shear modulus,
we can state the following:

τlt τlt τlt


= Vm + Vf (2.32)
Glt Gm Gf

then,

1 Vm Vf
= + (2.33)
Glt Gm Gf
or

" #
1
Glt = Gm Gm
(2.34)
(1 − Vf ) + Gf V f

For last, it is important to remark that all this formulas apply to a laminae of a certain
thickness made from unidirectional fibres. If the laminae is made from fabric, the resultant
laminae is a two monolayer laminae stacked up at 0/90◦ with half the thickness each.

2.5 Classical Laminated Plate Theory


Analysis of composite plates is based on the following approaches: equivalent
single-layer theories (2-D) or three dimensional elasticity theory (3-D). For the
goal of this thesis, focusing on 2-D theories to explain how composite plates behave is
enough.

The two theories that lay inside the 2-D theories are the Classical Laminated Plate
Theory (CLPT) and the First-order Shear Deformation laminated plate Theory (FSDT).
The simplest theory is the CLPT and requires that the Kirchhoff hypothesis holds,
which assumes that plane cross sections remain plane and normal to the middle-plane
during deformations. This implies that the transverse shear strains vanish. The FSDT is
a bit more complicated and is build on the Reissner-Mindlin hypothesis, where plane
cross sections remain plane after deformation, but not necessarily normal to the
reference plane. This results inclusion of out-of-plane shear deformation.

As the study performed in this thesis will be executed with a commercial software,
the usefulness of this section is to be brief and concise providing a general understanding
of composite plates structural behaviour: distribution of stresses and deformations.
Being so, the CLPT is the only theory that is going to be explained. ([19] and [32])

The CLPT is a direct extension of the classical plate theory for isotropic and
homogeneous material as proposed by Kirchhoff – Love [33]. However, the extension of

35
Chapter 2. State of the art

this theory to laminates requires some modifications to take into account the
inhomogeneity in thickness direction. In the following, the assumptions made in this
theory along with the assumptions made for classical plate theory are given.

Assumptions of Classical Laminated Plate Theory

1. The laminate consists of perfectly bonded layers. There is no slip between the
adjacent layers. In other words, it is equivalent to saying that the displacement
components are continuous through the thickness.

2. Each lamina is considered to be a homogeneous layer such that its effective properties
are known.

3. Each lamina is in a state of plane stress.

4. The strain in the direction perpendicular to the layer due to the Poisson effect is
ignored.

5. The individual lamina can be isotropic, orthotropic or transversely isotropic.

6. The laminate deforms according to the Kirchhoff - Love assumptions for bending and
stretching of thin plates (as assumed in classical plate theory). The assumptions are:

(a) The normals to the midplane remain straight and normal to the midplane even
after deformation.
(b) The normals to the midplane do not change their lengths.

CLPT is a theory for bending of homogeneous plates, but accounting for in-plane
tractions in addition to bending moments. Also, the stiffness variation through layers (or
plies) is taken into consideration. In general cases, the determination of the tractions
and moments at a given location will require a solution of the general equations for
equilibrium and displacement compatibility of plates. This theory is treated in a number
of standard texts, and will not be discussed here.

We begin by assuming a knowledge of the tractions N and moments M applied to a


plate at a position x,y, as shown in Figure 2.27. It will be convenient to normalize these
tractions and moments by the width of the plate, so they have units of N/m and N-m/m,
or simply N, respectively. Coordinates x and y are the directions in the plane of the plate,
and z is customarily taken as positive downward.

Figure 2.27: Applied tractions and moments in plate. Retrieved from [16].

36
Chapter 2. State of the art

   
 Nx   Mx 
N = Ny M = My (2.35)
Nxy Mxy
   

2.5.1 Kinematics
When this tractions and moments are applied, displacements occur in the plate. The
plate bends, as shown in Figure 2.28 and the different in-plane displacements can be
described as follows:

Figure 2.28: Undeformed and deformed geometries of an edge of a plate under the CLPT
assumptions. Retrieved from [17].

∂w ∂w
u(x, y, z) = u0 − z , v(x, y, z) = v0 − z , w(x, y, z) = w0 (x, y) (2.36)
∂x ∂y

So u is displacement in x direction, v is displacement in y direction and w is


displacement in z direction, while u0 , v0 and w0 are displacements of the midplane in x,
y and z directions, respectively. It is important to recall that w is constant as it was
stated on assumption 4: the strain in z direction due to Poissson effects is neglected.
Being so, equation (2.36) can be reqritten as

∂w0 ∂w0
u(x, y, z) = u0 − z , v(x, y, z) = v0 − z (2.37)
∂x ∂y
Based on the displacement field above, we can find the strains as follows:

37
Chapter 2. State of the art

  ∂     ∂2w 
 εx  0  ∂u ∂x
0
 − ∂x
0
2 
 ∂x
    
∂  u
 
∂v0 ∂ 2 w0
ε = εy =  0 ∂y  = ∂y +z − ∂y 2
=
∂ ∂ v  ∂u0 + ∂v0   ∂ 2 w0 
γxy
   
−2 ∂x∂y
 
∂y ∂x ∂y ∂x
 0   (2.38)
 εx   κx 
= εy 0 + z κy = ε0 + zκ
 0
γxy κxy
 

where ε0 is the midplane strain and k is the vector of second derivatives of the
displacement, called the curvature. The component kxy is a twisting curvature, stating
how the x -direction midplane slope changes with y (or equivalently how the y-direction
slope changes with x ).

2.5.2 The Material Law


The stresses relative to the x-y axes are now determined from the strains, and this
must take consideration that each ply will in general have a different stiffness, depending
on its own properties and also its orientation with respect to the x-y axes. This is
accounted for by computing the transformed stiffness matrix D, which is the same
matrix D as in equation (2.5) but multiplied by the trasnformation matrix T .

Recall that the ply stiffnesses as given by equation (2.3), are those along the fiber and
transverse directions of that particular ply. The properties of each ply must be transformed
to a common x-y axes, chosen arbitrarily for the entire laminate. The stresses at any
vertical position are then:

σ = Dε = Dε0 + zDκ (2.39)

where here D is the transfomed stiffness of the ply at which the stresses are being
computed.

2.5.3 Resultant Forces and Moments


Each of these ply stresses must add to balance the traction per unit width N:

Z +h N Z zk+1
2 X
N= σdz = σk dz (2.40)
−h
2 k=1 zk

where σk is the stress in the kth ply and zk is the distance from the laminate midplane
to the bottom of the kth ply. Using equation (2.39) to write the stresses in terms of the
midplane strains and curvatures:

N Z zk+1 Z zk+1 
X k 0 k
N= D ε dz + D κzdz (2.41)
k=1 zk zk

The curvature κ and midplane strain ε0 are constant throughout z, and the transformed
k
stiffness D does not change within a given ply. Removing these quantities from within
the integrals:

38
Chapter 2. State of the art

N  Z zk+1 Z zk+1 
X k 0 k
N= D ε dz + D κ zdz (2.42)
k=1 zk zk

After evaluating the integrals, this expression can be written in the compact form:

N = Aε0 + Bκ (2.43)

where A is an ”extensional stiffness matrix” defined as:

N
X k
A= D (zk+1 − zk ) (2.44)
k=1

and B is a ”coupling stiffness matrix” defined as:

N
1X k
B= D (zk+1 2 − zk 2 ) (2.45)
2
k=1

The A matrix gives the influence of an extensional midplane strain (ε0 ) on the
in-plane traction N, and the B matrix gives the contribution of a curvature (κ) to the
traction (coupling). ”It may not be obvious why bending the plate will require an in-plane
traction, or conversely why pulling the plate in its plane will cause it to bend. But
visualize the plate containing plies all of the same stiffness, except for some very
low-modulus plies somewhere above its midplane. When the plate is pulled, the more
compliant plies above the midplane will tend to stretch more than the stiffer plies below
the midplane. The top half of the laminate stretches more than the bottom half, so it
takes on a concave-downward curvature.”[14, p. 9]

Similarly, the moment resultants per unit width must be balanced by the moments
contributed by the internal stresses. Following the same procedure as with the tractions:

Z +h N Z zk+1
2 X
M= σzdz = σk zdz (2.46)
−h
2 k=1 zk

N Z zk+1 Z zk+1  N  Z zk+1 Z zk+1 


X k 0 k 2
X k 0 k 2
M= D ε zdz + D κz dz = D ε zdz + D κ z dz
k=1 zk zk k=1 zk zk
(2.47)

M = Bε0 + Dκ (2.48)

where D is a ”bending stiffness matrix” defined as:

N
1X k
D= D (zk+1 3 − zk 3 ) (2.49)
3
k=1

The complete set of relations between applied forces and moments, and the resulting
midplane strains and curvatures, can be summarized as a single matrix equation:

39
Chapter 2. State of the art

     0
N A B ε
= (2.50)
M B D κ

The A/B/B/D matrix in brackets is the laminate stiffness matrix, and its inverse
will be the laminate compliance matrix.

The presence of nonzero elements in the coupling matrix B indicates that the
application of an in-plane traction will lead to a curvature or warping of the plate, or
that an applied bending moment will also generate an extensional strain. These effects
are usually undesirable. However, they can be avoided by making the laminate
symmetric about the midplane, as examination of equation (2.45) can reveal.

The above relations provide a straightforward - although tedious, unless a computer is


used (see Section 2.7. Finite Element Analysis of Laminates) - means of determining
stresses and displacements in laminated composites subjected to in-plane traction or
bending loads.

2.6 Buckling of Laminated Plates


From Annex A.1. Governing Equations for Plates, the equations A.1, A.2 and
A.7 are the equilibrium equations for a laminated thin plate:
∂Nx ∂Nxy
+ =0
∂x ∂y
∂Ny ∂Nxy
+ =0
∂y ∂x
∂ 2 Mx ∂ 2 M x y ∂ 2 My
+ 2 + + p∗ = 0
∂x2 ∂x∂y ∂y 2
2 2 2 2
where p∗ = p + Nx ∂∂xw2 + Ny ∂∂yw2 + 2Nxy ∂x∂y
∂ w
+ ρ∗ ∂∂tw
2 .

The last equation solves a buckling problem [19]. Buckling is the loss of stability
due to geometric effects rather than material failure. But it can lead to material failure
and collapse if the ensuing deformations are not restrained. It is characterized by a
sudden sideways failure of a structural member subjected to high compressive stress -
and/or in-plane shear stress for composites, where the stress at the point of failure is less
than the ultimate stress that the material is capable of withstanding.

Buckling is commonly referred as a bifurcation problem. That is, there is a


bifurcation point where the primary (without lateral or out-of-plane deformations w = 0)
and secondary (when the instability occurs) loading paths intersect - point A in Figure
2.29. At this point, more than one equilibrium position is possible. The primary path is
not usually followed after loading exceeds this point and the structure is in the
post-buckling state. The slope of the secondary path at the bifurcation point determines
the nature of the post-buckling. A positive slope indicates that the structure will have
post buckling strength whilst a negative slope means that the structure will snap
through or simply collapse.

40
Chapter 2. State of the art

Figure 2.29: Bifurcation problem scheme. Retrieved from [18].

Let’s keep developing equation ??.7. Insert equation (2.48), we obtain:

∂4w ∂4w ∂4w ∂4w


−D11 − 4D 16 − (2D 12 + 4D 66 ) − 4D 26
∂x4 ∂x3 ∂y ∂x2 ∂y 2 ∂x∂y 3
4
∂ w 3
∂ u0 3
∂ u0 3
∂ u0 ∂ 3 u0
−D22 4 + B11 + 3B 16 + (B 12 + 2B 66 ) + B 26 (2.51)
∂y ∂x3 ∂x2 ∂y ∂x∂y 2 ∂y 3
∂ 3 v0 ∂ 3 u0 ∂ 3 u0 ∂ 3 v0
+B16 3 + (B12 + 2B66 ) 2 + 3B26 + B 22 + p∗ = 0
∂x ∂x ∂y ∂x∂y 2 ∂y 3

If we particularise equation (2.51) for an orthotropic laminate - with principal


properties in orthogonal directions, its constitutive equations satisfy the following
conditions [26]:

A12 = A26 = 0
Bij = 0
D12 = D26 = 0

Incorporation of conditions above into equation (2.51) simplifies the equilibrium


equation (buckling problem) as follows:

∂4w ∂4w ∂4w


D11 + (2D 12 + 4D 66 ) + D 22 = p∗ (2.52)
∂x4 ∂x2 ∂y 2 ∂y 4

The solution of equation (2.52) depends on the boundary conditions to which the
laminated plate is subjected to and the force p∗ that is applied to it. Here are some
examples of typical forces that can make a plate buckle:
2
U niaxial compressive load p∗ = Nx ∂∂xw2
2 2
Biaxial compressive load p∗ = Nx ∂∂xw2 + Ny ∂∂yw2
2
In − plane shear load p∗ = Nx y ∂x∂y
∂ w

There also can be combined load states - e.g. uniaxial compressive load plus in-plane
shear load. And here are some examples of boundary conditions that can be applied to a
plate like the one in Figure 2.30:

41
Chapter 2. State of the art

Figure 2.30: Plate of dimensions a x b. Retrieved from [19].

x=0: w(0, y) = 0 Mx (0, y) = 0

x=a: w(a, y) = 0 Mx (a, y) = 0


Simply supported plate
y=0: w(x, 0) = 0 My (x, 0) = 0

y=b: w(x, b) = 0 My (x, b) = 0

∂w(0,y)
x=0: w(0, y) = 0 ∂x =0

x=a: w(a, y) = 0 , ∂w(a,y)


∂x =0
Clamped plate
∂w(x,0)
y=0: w(x, 0) = 0 ∂y =0

∂w(x,b)
y=b: w(x, b) = 0 ∂y =0

Each case - a set of boundary conditions and forces applied - has a particular
solution, which also depends on the theory adopted for the problem (not the same
solution for CLPT than for FSDT).

Each particular solution yields a set of critical loads, each of them causing failure by
buckling to the plate when applied. The number of possible solutions (critical loads) to a
particular case are called buckling modes.

Every critical load is a product of the original load - being this a distribution of an
out-of-plane load like aerodynamic lift or a concentrated force, with the possibility of
being lower or higher. If it is lower it means that the plate fails with the original load
multiplied by a factor less than 1. Conversely, if the critical load is higher than the
original load, it means that buckling occurs when the plate is subjected to the original
load multiplied by a factor greater than 1.

As recalled prior in this section, the study performed in this thesis is going to be
tackled with a commercial software. This kind of software solve this type of analysis as
an eigenvalue analysis because the critical loads are the eigenvalues λi of the discretised
system of equations

42
Chapter 2. State of the art

([K] − λ[Kg ]){v} = 0 (2.53)

where K and Kg are the stiffness and geometric stiffness matrix, respectively, and v
is the column of eigenvectors (buckling modes). Equation 2.53 represents the linearised
buckling problem.

The geometric stiffness matrix, also known as the initial stress stiffness matrix, is a
symmetric matrix dependent on the element stress level. It reflects the effect of
geometric change on the element force vector from a known stress state. For beam and
plate bending structures, the geometric stiffness matrix represents the stiffening effect of
the tensile axial/membrane stresses - i.e. the in-plane stresses.

The linear buckling solver performs the following steps:

• Calculates and assembles the element stiffness and geometric stiffness matrices to
form the global stiffness and geometric stiffness matrices. Constraints are assembled
in this process.

• Modifies the stiffness matrix when a shift value is applied. A shift may be used to
determine modes near to a desired value.

• Solves the eigenvalue problem to get buckling load factors and the corresponding
buckling modes.

2.7 Finite Element Analysis of Laminates


In this section an introduction to the FEM is going to be delivered [20]. After, the
most important features of FE analysis of laminates are going to be explained.

2.7.1 Basic FEM Procedure


Consider a bar - see Figure 2.31(a) - with a clamped end, an axial distributed force
over its length (L) and a punctual force applied over its other end. The deformation of
the bar can be described with the ordinary differential equation (ODE)

 
d du
− EA −f =0 ; 0≤x≤L (2.54)
dx dx

Figure 2.31: Physical (a) and mathematical (b) model. Retrieved from [20].

where E, A are the modulus and cross section area of the bar, respectively, and f is
the distributed force. The boundary conditions for this case are:

43
Chapter 2. State of the art

u(0) = 0 (2.55)
 
du
EA =P (2.56)
dx x=L

This bar is mathematically modeled as a line in 2.31(b).

• Discretisation

Figure 2.32: Discretisation into elements. Retrieved from [20].

The first step is divide the domain of the bar (x ∈ [0, L]) into discrete elements, as
shown in Figure 2.32.

• Element Equations

To derive the element equations, an integral form of the ODE is used. That is obtained
by integrating the product of the ODE time a weight function v as follow

Z xB    
d du
0= v − EA − f dx (2.57)
xA dx dx

This is called a weak form because the solution u(x) does not have to satisfy the ODE
(2.54) for all and every of the infinite values of x in [0,L], in a strong sense. Instead, the
solution u(x) only has to satisfy the ODE in (2.57) in a weighted average sense. It is
therefore easier to find a weak solution than a stronger one. Altough for the case of the
bar, the strong (exact) solution is known, most problems of composite mechanics do not
have a exact solution. The governing equation is obtained by integrating equation (2.54)
by parts as follows

xB xB
du xB
Z Z  
dv du
0= EA dx − vf dx − EA (2.58)
xA dx dx xA dx xA

where v(x) is a weight function, which is usually set equal to the primary variable u(x).
Now, boundary conditions must be applied. There are two kinds: essential boundary
conditions (v(x) at xA or xB ), or natural boundary conditions (EA du dx at either end of the
element). Applying u(xA ) = u1 , u(xB ) = u2 , − EA dx x = P1 and − EA du
e e du e
   
dx xB
=
A
P2e , the governing equation becomes

Z xB  
dv du
0= EA − vf dx − P1e v(xA ) − P2e v(xB ) (2.59)
xA dx dx

44
Chapter 2. State of the art

• Approximation over an Element


Now, the unknown u(x) is approximated as a linear combination (series expansion) of
known functions Nie (x) and unknown coefficients aej , as

n
X
ue (x) = aej Nie (x) (2.60)
j=1

where aej are the coefficients to be found and Nie (x) are the interpolation functions.
For the weight function v(x) the Ritz method can be used [34], in which v(x) = Nje (x).
Substituting in the goberning equation /2.58) we get

n Z xB
dNie dNje
X  Z xB
e
EA dx aj = Nie f dx + P1e Nie (xA ) + P2e Nie (xB ) (2.61)
xA dx dx xA
j=1

which can be rewritten as

n
X
e e
Kij aj = Fie (2.62)
j=1

or in matrix form

[K e ]{ae } = {F e } (2.63)

where [K e ] is the element stiffness matrix, {F e } is the element vector equivalent


force and {ae } are the element unknown parameters, which must represent the nodal
displacements.
• Interpolation Functions

Figure 2.33: Linear interpolation functions for a two-node element bar. Retrieved from
[20].
The interpolation functions chosen in such a way that the unknown parameters
represent the nodal displacements. For a two-node element with xe ≤ x ≤ xe+1 , the
following linear interpolation functions (Figure 2.33) can be used

xe+1 − x
N1e = (2.64)
xe+1 − xe
x − xe
N2e = (2.65)
xe+1 − xe
(2.66)

45
Chapter 2. State of the art

which satisfies the following conditions

 
0 if i 6= j
Nie (xj ) = (2.67)
0 if i=j

2
X
Nie (x) = 1 (2.68)
i=1

Many other interpolation functions can be used - e.g. two-dimensional interpolation


functions. Broadly speaking, mode nodes per element imply more accuracy and less need
for a fine mesh, but also imply higher cost of computer time.

• Assembly of Element Equations

Since a node must have the same displacement on both adjacent elements, the value
is unique. To assemble the element equations we must define the connectivity of
elements- i.e. the contribution of each element to each nodal displacement.

Close examination of the definition of the element stiffness matrix in equation (2.61),
making use of the interpolation functions of equation (2.64), allow us to get to

k e −k e
   
EAe 1 −1
[K e ] = = (2.69)
−k e k e he −1 1

Figure 2.34: Three two-node element bar discretisation. Retrieved from [20].

which is the element stiffness matrix of a two-node linear element. If we assemble a


three two-node elements discretisation of the bar problem exposed earlier on (see Figure
2.34, we have u11 = U1 ; u12 = U2 = u21 ; u21 = U3 = u31 ; u32 = U4 (superscript: element
number; subscript: nodal number). If we assemble the element equations (2.63) taking
into account this connectivity and the element stiffness matrix of (2.69), we get to

k1 −k 1
  
0 0  U1 
−k 1 k 1 + k 2

 
0 0  U2

= {F e }

 (2.70)
 0 −k 2 k 2 + k 3 −k 3  
U3 
 
0 0 −k 3 k3 U4

• Solution of Equations

Having this equation, the only thing left to do is apply the boundary conditions and
then it can be resolved. The boundary conditions are U1 = 0 (essential), and the
contribution of the distributed force of each element to each node plus the punctual force
applied at the second node of the third element (F23 ). Thus, we have 4 equations and 4
unknowns (U2 , U3 , U4 and F11 ).

This is known as an implicit method of resolving the problem.

46
Chapter 2. State of the art

2.7.2 General FEM Procedure


The derivation of the element equations, assembly, and solution for any type of elements
is similar to that of the one-dimensional bar element described in Section 2.7.1. Basice
FEM Procedure, with the exception that the principle of virtual work (PVW) is used
instead of the governing equation (2.54). The PVW provides a weak form similar to that
in (2.57). The PVW reads
Z Z Z
σij δεij dV − ti δui dS − fi δui dV = 0 (2.71)
V S V

which represents the work performed by the internal stresses (first term) against the
work made by the external forces (second and third term). The parameters ti and fi are
the force per unit area and force per unit volume, respectively, made on the body - thus,
the negative sign.

Expanding equation (2.71) for full 3D state of deformation, the internal virtual work
is
Z
δWI = (σxx δεxx + σyy δεyy + σzz δεzz + τyz δγyz + τxz δγxz + τxy δγxy )dV (2.72)
Z
= σ T δεdV (2.73)
V

Next, the external work is


Z Z
T
δWE = t δudS + f T δudV (2.74)
S V

The virtual strains are strains that would be produced by virtual displacements δu(x).
Therefore, we can relate each other by

ε=δ u (2.75)
δε = δ δu (2.76)

where

 δ δ δ

δx 0 0 δy 0 δz
δ δ δ
δ=0 0 0 (2.77)
 
δy δx δz
δ δ δ
0 0 δz δy 0 δx

Then, the PVW is written in matrix notation as


Z Z Z
T T
σ δ δudV = f δudV + tT δudS (2.78)
V V S

Now, putting the stress in the first term as a function of the virtual displacements (as
it is a function of the strains, i.e. σ = C ε), and developing all the procedure explained in
Section 2.7.1. Basic FEM Procedure, we get to

K a=P (2.79)

47
Chapter 2. State of the art

2.7.3 AbaqusTM
The commercial software with the FEM incorporated that has been selected to
undergone the study of this thesis is AbaqusTM . A complete AbaqusTM analysis
usually consists of three distinct stages: pre-processing, simulation, and post-processing.
These three stages are linked together by files as shown in Figure 2.35 below:

Figure 2.35: Flow chart of the process to complete an AbaqusTM analysis. Retrieved from
[21].

An AbaqusTM model is composed of several different components that together


describe the physical problem to be analysed and the results to be obtained. At a
minimum, the analysis model consists of the following information: discretised geometry,
element section properties, material data, loads and boundary conditions, analysis type,
and output requests.

• Discretised geometry: each element in the model represents a discrete portion


of the physical structure, which is, in turn, represented by many interconnected
elements. Elements are connected to one another by shared nodes. The coordinates
of the nodes and the connectivity of the elements comprise the model geometry. The
collection of all the elements and nodes in a model is called the mesh. Generally,
the mesh will be only an approximation of the actual geometry of the structure.

• Element section properties: AbaqusTM has a wide range of elements, many of


which have geometry not defined completely by the coordinates of their nodes. Such
additional geometric data are defined as physical properties of the element and are
necessary to define the model geometry completely.

• Material data: material properties for all elements must be specified.

• Loads and boundary conditions: loads and boundary conditions must be set
according to the analysis desired to perform.

48
Chapter 2. State of the art

• Analysis type: the type of analysis must be defined depending on the study that
the user is performing. The types are many, and the ones of interest for this thesis
are the static and buckle types.

• Output requests: the output data to be generated must be selected.

For last, just to say that AbaqusTM doesn’t use units. The units you use to model the
geometry are the ones that determine the rest of the units of the output variables - see
Table 2.4.

Table 2.4: AbaqusTM units logic.

2.7.4 Shell Elements


Shell elements are used to model structures in which one dimension (the thickness) is
significantly smaller than the other dimensions and in which the stresses in the thickness
direction are negligible. Plate theories like the one explained in Section 2.5. Classical
Laminated Plate Theory are implemented and differentiated by element types.

AbaqusTM has two types of shell elements, which are described in the user’s manual
[21] as: ”Two types of shell elements are available in AbaqusTM : conventional shell
elements and continuum shell elements. Conventional shell elements discretise a
reference surface by defining the element’s planar dimensions, its surface normal, and its
initial curvature. The nodes of a conventional shell element, however, do not define the
shell thickness; the thickness is defined through section properties. Continuum shell
elements, on the other hand, resemble three-dimensional solid elements in that they
discretise an entire three-dimensional body yet are formulated so that their kinematic and
constitutive behaviour is similar to conventional shell elements. Continuum shell
elements are more accurate in contact modelling than conventional shell elements, since
they employ two-sided contact taking into account changes in thickness. For thin shell
applications, however, conventional shell elements provide superior performance.”

Study of composites can be done at different level of detail: michromechanics, lamina


level or laminate level. For the purposes of the thesis the lamina level or meso-scale level,
is going to be used. Shell elements have the ability of computing the laminate properties
with the laminate stacking sequence and the unidirectional layers (known as laminae)
properties.

Shell elements, unlike continuum elements, use material directions local to each
element. Anisotropic material data, such as the one of laminates, and element output

49
Chapter 2. State of the art

variables, such as stress and strain, are defined in terms of these local material
directions. In large-displacement analyses the local material axes on a shell surface
rotate with the average motion of the material at each integration point.

2.8 Failure Criteria


Failure criteria [35] are curve fits of experimental data that attempt to predict failure
under multiaxial stress based on experimental data obtained under uniaxial stress. All
failure criteria described in this section predict the first occurrence of failure in one of
the laminae but are unable to track failure propagation until complete laminate failure.
Continuum damage mechanics that track damage evolution up to laminate failure are
out of the scope of this thesis.

Failure criteria are presented using the notion of failure index, which is defined as

stress
IF = (2.80)
strength
Failure is predcited when IF ≥ 1.

2.8.1 Hashin Criterion


Hashin criterion [36] is implemented for elements with a plane stress formulation, thus
being a 2-D failure criterion for unidirectional laminae. Failure indices for Hashin criteria
are related to fibre and matrix failures and involve four failure modes: tensile fibre failure,
compressive fibre failure, tensile matrix failure, and compressive matrix failure. Prior to
announce the criterion, let’s define the failure properties:
+
S11 value of σ11 at longitudinal tensile failure

S11 value of σ11 at longitudinal compressive failure
+
S22 value of σ22 at transverse tensile failure

S22 value of σ22 at transverse compressive failure
S12 absolute value of σ12 at longitudinal shear failure
S23 absolute value of σ23 at transverse shear failure

If σ11 ≥ 0 the tensile fibre failure criterion is:

 2  2
+ σ11 σ12
IF,f ≡ + +α ≥1 (2.81)
S11 S12

If σ11 < 0 the compressive fibre failure criterion is:

 2
− σ11
IF,f ≡ − ≥1 (2.82)
S11

If σ22 ≥ 0 the tensile matrix failure criterion is:

 2  2
+ σ22 σ12
IF,m ≡ + + ≥1 (2.83)
S22 S12

50
Chapter 2. State of the art

If σ22 < 0 the compressive matrix failure criterion is:

 2 "
− 2 #  2
− σ22 S22 σ22 σ12
IF,m ≡ + −1 − + ≥1 (2.84)
2S23 2S23 S22 S12

Note that subscript f stands for fibre, and m for matrix. Then, there is the parameter
α, which is a coefficient that determines the contribution of the longitudinal shear stress
to fibre tensile failure. Allowable range is 0 ≤ α ≤ 1. AbaqusTM only has this criterion
implemented and the default value of α is α = 0.

2.8.2 Tsai-Hill Criterion


The Tsai-Hill criterion [37] is a quadratic, interactive stress-based criterion that
identifies failure, but does not distinguish between different modes of failure. Failure
occurs whenever the following condition is satisfied:

2
σ11 2
σ22 σ11 σ22 2
σ12
IF ≡ 2 + 2 − 2 + 2 ≥1 (2.85)
S11 S22 S11 S12

where the Sij coefficients are the same as the ones in the Hashin criterion and if σ11 ≥ 0,
+ − +
then S11 = S11 ; if σ11 < 0, then S11 = S11 ; if σ22 ≥ 0, then S22 = S22 and if σ22 < 0, then

S22 = S22 .

2.9 Pareto Optimality


For a nontrivial multi-objective optimization problem, there does not exist a single
solution that simultaneously optimizes each objective. In that case, the objective
functions are said to be conflicting, and there exists a (possibly infinite) number of
Pareto optimal solutions [38]. A solution is called nondominated, Pareto optimal, Pareto
efficient or noninferior, if none of the objective functions can be improved in value
without degrading some of the other objective values. The term is named after Vilfredo
Pareto (1848–1923), an Italian engineer and economist who used the concept in his
studies of economic efficiency and income distribution.

Figure 2.36 shows four geometric examples of Pareto optimality. In these figures, the
circles represent objectives that are satisfied best when the area of the circle is maximized.
The constraints are that the circles may not overlap and must fit within the triangle.

51
Chapter 2. State of the art

Figure 2.36: Examples illustrating the Pareto optimality. Retrieved from [22].

Given a set of choices and a way of valuing them, the Pareto frontier or Pareto front
is the set of choices that are Pareto efficient. The Pareto frontier, P(Y) is described as
follows. Consider a system with function f : Rn → Rm , where X is a compact set of
feasible decisions in the metric space Rn , and Y is the feasible set of criterion vectors in
Rm , such that Y = {y ∈ Rm : y = f (x), x ∈ X }.

We assume that the preferred directions of criteria values are known. A point y 00 ∈ Rm
is preferred to (strictly dominates) another point y 0 ∈ Rm , written as y 00  y 0 . The Pareto
frontier is thus written as:

P (Y ) = {y 0 ∈ Y : {y 00 ∈ Y : y 00  y 0 , y 00 6= y 0 } = Ø}.

An example of a Pareto front is depicted in Figure 2.37, where we can see that point
C does not belong to the front as it is dominated by points A and B. The latter points
are non-dominated points or Pareto efficient points.

52
Chapter 2. State of the art

Figure 2.37: Example of a Pareto front. Retrieved from [23].

53
Chapter 3

Methodology

In this chapter, the process undergone to obtain the design of a moncoque wing
structure with composite materials is explained.

The first aspect of the methodology that is detailed are the premises from which the
design process has been addressed. Following to that, a thorough description of the whole
design process - steps and decision making along the procedure - is given - supported by a
flowchart. And for last, a brief and schematic description, also supported with a flowchart,
of the process of the making of an AbaqusTM simulation is delivered.

3.1 Premises of the Design


In the design of a monocoque composite structure there are several parameters to
take into account, as each of them vary the overall behaviour of the structure. As for
that, the iteration over all of those parameters has been discarded for being too
complicated to be tackled - highly time-consuming and expert programming required.
For that reason, it has been mandatory to select the possible ranges of the parameters of
the study - some of them have been fixed. In other words, the design of the wing’s
structure has been delimited.

The main parameters that have been delimited are geometrical and material-related.
The criterion of varying parameters has been delimited as well. Then, the structural
requirements that the structure must withstand, the structure’s maximum weight, the
structure’s minimum buckling constant value, and the type of structural testing to which
the structure is going to be subjected to in order to be sized, have been set.

Following, each of these premises are individually explained. After that, a list of all
the parameters used for the design is going to be presented.

3.1.1 Type of Structural Testing


The type of structural testing performed to the structure is the following: symmetrical
testing of a cantilevered semi-wing with a clamped end (restricted displacements and
rotations), and a distributed force applied over the semi-wing (the lift generated by the
semi-wing) - see Figure 3.1.

54
Chapter 3. Methodology

Figure 3.1: Schematic representation of the structural testing that is going to be performed
on the wing. Retrieved from [24].

The clamped end is assumed to emulate the union between the wing and the fuselage of
the aircraft, which is ideally placed at the right center of the wing, allowing the maximum
wing surface unaltered by the union.

3.1.2 Structural Requirements


As it has been explained in Section 1.5. Justification, this thesis has the goal to be
useful for Trencalòs Team in developing a know-how about monocoque composite
structures. The type of competitions in which the team participates require an aircraft
capable of turning fast and carrying the highest amount of pay-load.

Accordingly, the structural requirements have been set keeping those two factors in
mind. The maximum take-off weight(MTOW) has been set to 10 kg and the load factor
n, which is related with making a steep turn and going very fast, is set to 4. Summing up:

• MTOW = 10 kg.

• n = 4.

• Lift per semi-wing = 200 N (approximating ~g to 10 m/s2 ).

3.1.3 Geometrical Features of the Wing


The geometrical characteristics of the wing chosen to develop this design are the
following:

• Wingspan = 1.2 m.

• Planform: rectangular.

• Chord = 0.19 m.

• Airfoil: CLARKY.

• Airfoil’s maximum thickness and position = 0.0222 m (11.71% of the chord)


at 0.0532 m (28% of the chord).

55
Chapter 3. Methodology

As I have mentioned in Section 1.5. Justification, another member of the team has
studied the construction technique of monocoque composite structures for his bachelor’s
thesis. He has picked a 0.6 m wingspan wing, with CLARKY airfoil, to make his tests.
For that reason, I have chosen this geometrical characteristics so I can be of any help
with the numerical results that I have yield - mainly, the tendencies that I have
observed. However, I have decided to change the wingspan.

I have doubled the wingspan for two reasons: first, the sizing of a wing lies mainly on
the bending moment that is created along its wingspan. Taking into account that, I have
made the design more challenging by adding more lever arm for the lift, which will
generate a higher bending moment. And secondly, the aircrafts employed in the
competitions in which Trencalòs Team takes part in have wingspans greater than 3
metres. Being so, doubling the wingspan up to 1.2 m makes the wing more alike to the
one that would be used in such a competition.

For last, I want to mention that the airfoil has been cut at the 0.17 m point of the
chord - see Figure 3.2. That is because the ailerons are always built separately and do not
have to withstand high loads. Thus, I have not included them in the wing structure.

Figure 3.2: Clarky airfoil with a cut at 0.17 m of the chord line.

3.1.4 Load-bearing Elements Employed


As explained in Section 2.2. Structural Concepts, monocoque structures can be
complemented with auxiliary elements that help in the transmission of loads and
increase the buckling resistance of the skin - e.g. by enlarging the I of a certain region or
by limiting the buckling length.

The main load-bearing element though, which takes the greatest part of the loads, is
the skin. The auxiliary members that are going to be tested in the wing’s structure are:
spar or spars (transmission of the shear stress and torsional resistance) and ribs (limit
the buckling length in the main buckling direction , i.e. the span-wise direction).

The main variable in relation with this elements is its position along the chord line.
For a spar of a certain width, there are infinite possible positions along the 0.17 m of
chord available. For that reason, the positions that will be tested are going to be few but
representative. Intuitively, the most appropriate position to put the spar is at the
maximum thickness point of the profile, as it has the highest I and thus helps more with
the bearing of shear loads. So, the spar will be placed at the maximum thickness point
for one of the study cases, and placed before and after that point for some other cases.

Putting a second spar will be based on the optimal position of the first spar.
Regarding ribs, there is the same problem as with the spar: there are infinite
possibilities. Being so, a finite number of combinations will be tested, which are: 3, 4
and 5 ribs per semi-wing.

56
Chapter 3. Methodology

Lastly, in relation with the skin, there is the possibility of enhancing the I of the
extrados by using different layer sequences - e.g. using different materials and/or
different core thickness. Doing so, buckling resistance is enlarged. This stiffened regions -
known as stringers - are going to be used. The main variables regarding this ”elements”
(as they actually belong to the skin) are: materials applied, core thickness and region of
appliance - position and width.

Summing up, we have:

Figure 3.3: Structural elements employed.

3.1.5 Materials Employed and its Properties


The reinforcing materials employed for the making of the skin have been decided to
be fibre-based. Unidirectional fibres and fabrics will both be considered. Glass fibre
and carbon fibre will be the two types of fibres used for being the most commonly used
for this type of structures. And the matrix material for the making of the skin will be a
polymeric resin, specifically, epoxy.

A set of fibres plus the matrix form a laminae. The skin will be made up with
laminae stacked up in different orientations, forming a laminate or sandwich panel, if
a soft core is added. If a sandwich panel is used, the core employed will be Rohacell
R

110 wf, a high performance foam.

Secondary elements added to the structure, like ribs or spar(s), will be made out of
balsa wood. Summing up:

Materials employed
fibre Glass fibre / Carbon fibre
Skin matrix Epoxy
core (if employed) Rohacell R
110 wf
Auxiliary elements Balsa wood

Table 3.1: Summary of the materials employed

Based on the discussion on laminates in Section 2.3.3.2. Laminates, the orientations


that are going to be used for the skin are 0, 90 and ±45◦ . The possible combinations for
the laminate with this 4 different orientations is very large. For that reason, the possible
skins considered are the following:

57
Chapter 3. Methodology

Skins employed
Type of fibre Skin code Region Layer stacking sequence
Low load-bearing region (+45/ − 45/core)s or (+45/−45)s
A
High load-bearing region (+45/ − 45/0/core)s or (+45/ − 45/0)s
Fabric
Low load-bearing region (0/90/core)s or (0/90)s
B
High load-bearing region (0/90/0/core)s or (0/90/0)s
Low load-bearing region (+45/core/ − 45) or (+45/ − 45)
Unidirectional C
High load-bearing region (+45/0/core/0/ − 45) or (+45/02 / − 45)

Table 3.2: Different types of skin considered for the study.

where the plies that are in bold are laminae made of carbon fibre/epoxy and the ones
that are not are made of glass fibre/epoxy. Each skin is considered with and without
core. There are two made with fabrics and one made with unidirectional layers - that
refers to the glass fibre/epoxy laminaes, the carbon fibre/epoxy ones are unidirectional in
the 3 skins. Note that each skin is mainly differentiated by this. Furthermore, the skin is
not continuous along the wing, making a distinction between high and low load-bearing
regions - see Figure 3.4, where these regions are defined (where carbon fibre/epoxy is
applied and where is not) and the skin layer sequence defined in Table 3.2 is depicted.

Figure 3.4: Skin layer sequence depiction and differentiation between high and low load-
bearing regions.

For last, I will present the mechanical and physical properties of the 4 materials
employed for the study. First of all, I am going to expose balsa wood and Rohacell R

110 wf, which have been considered isotropic materials.

Balsa and Rohacell properties


Material ρ(kg/m3 ) E (MPa) ν Source
Balsa 160 3000 0.38 [39]
Rohacell 110 180 0.33 [40]

Table 3.3: Balsa and Rohacell


R
110 wf properties.

Regarding carbon fibre/epoxy(CE) and glass fibre/epoxy(GE) composites, there are


a wide range of them. Only by changing the grammage - i.e. weight per unit of area - of
the fibre, the properties of the resulting composite can vary a lot. Additionally, as seen
in Section 2.4. Laminae Mechanical and Physical Properties, the volume fraction
of each individual that makes up the composite, also influences a lot the final properties
of the same. Besides those facts, composite manufacturers do not provide the whole
range of mechanical properties necessary to characterize a composite material.

For that reason, the properties of this two composite monolayers have been extracted
both from research articles, which also provided the failure properties necessary to
implement the Hashin an Tsai-Hill failure criteria - see Section 2.8. Failure Criteria.

58
Chapter 3. Methodology

Recall that the monolayers are going to be characterized as transversely isotropic


materials - see Section 2.3.3.1. Laminae.

CE and GE properties
Property Carbon fibre/epoxy Glass fibre/epoxy
ρ(kg/m3 ) 1536 1881
E1 (MPa) 175320 52200
E2 (MPa) 8111.5 21932
ν12 0.31 0.27
G12 (MPa) 4991 5947
G23 (MPa) 5450 4503
+
S11 (MPa) 2840 1020

S11 (MPa) 1551 620
+
S22 (MPa) 60 40

S22 (MPa) 167 120
S12 (MPa) 95 60
S23 (MPa) 30 20
Source [41] [42]

Table 3.4: CE and GE properties.

The CE used in [41] is T800/3900-2, and the GE used in [42] is Fiberite HyE
9082Af. Both have a fibre volume of 60%.

To end this section, the thickness of a laminae and the core must be defined, as it is
of very important for the final weight of the structure. The thickness of a laminae can be
computed with [43], which is a free service that Hexcel R
offers for free. You can choose
the type of resin, type of fibre, the grammage of the latter and the fibre volume and it
computes the cured ply thickness.

Using the fibre grammages that my team member who does his thesis on the
construction, the Hexcel R
calculator yields between 0.09 and 0.12 mm of ply thickness
for both CE and GE. As the fibres that I have selected cannot be found in Hexcel R

catalogue, I have decided to fix the thickness of each laminae to 0.1 mm. Recall
that if the laminae is made from a fabric - see Section 2.3.2. Classification of
Composites, then it is characterized as 2 laminae of half the thickness (0.05 mm) and
oriented at 90◦ of difference one from the other.

Regarding the thickness of the core, it has been set to a possible range of 0, 1, 2, 3,
4 and 5 mm. These thickness have been selected based on Rohacell R
’s catalogue on its
website.

3.1.6 Maximum Weight of the Structure


Not only is important for an aircraft to be able to withstand its requested load states,
but to be as lightweight as possible too. For that very reason I have set a maximum weight
that the structure must no overpass. That maximum weight is 0.450 kg.

3.1.7 Minimum Buckling Constant


The indicator of failure by buckling is a buckling constant λ < 1 - see Section 2.6.
Buckling of Laminated Plates. Therefore, λmin > 1.

59
Chapter 3. Methodology

3.1.8 Variation of Parameters Criterion


In a multi-parametric study like the one performed in this thesis, the decision of how
all the variables must be varied is a non-trivial thing. For that reason, it is interesting to
approach it using some must-reach-objectives as the main drivers of the iteration through
the parameters. In other words, setting this objectives will allow to make the changes in
parameters more adequately and to avoid a misleading path of variation of the latter.

The different objectives of interest of a monocoque structure are: achieving the


minimum deflection at the wing tip δtip min , achieving the minimum weight
possible mmin , achieving the maximum buckling constant λmax or achieving the
maximum factor of safety F.O.Smax . For the study of this thesis the following two
objectives have been selected: highest buckling constant λmax and minimum weight
mmin .

Evidently, for each case analysed the failure indicators of the structure will have to
be checked. If there is a case - i.e. a certain type of skin, spar position, etc - in which the
indicators tell that the structure is failing, that case will be discarded. To do so, the
criteria explained in Section 2.8. Failure Criteria, both Hashin and Tsai-Hill will
have to be fulfilled.

The reason behind this decision is the following: Hashin criterion appreciates failure
at the fibre and matrix level, while Tsai-Hill criterion is a quadratic interactive criterion
which considers the failure of the laminae as an homogeneous material with the
equivalent transversely isotropic properties. Thus, ensuring the compliance of both
criteria is a way of taking into account macroscopic and constituent safety of the
structure.

The variation of the parameters will be driven by the Pareto optimality - see Section
2.9. Pareto Optimality. Starting from a set of cases - with different types of skins
and spar positions, each case will be treated as a point of coordinates (m,1/λ). All
points will be represented in a graphic (x axis corresponding to the mass m(kg), and the
y axis corresponding to the inverse of the buckling constant 1/λ) in order to seek the
Pareto front: non-dominated or Pareto optimal points.

Note that it is easier to analyse these points with two objective functions that aim for
the minimum - see Figure 3.5. Using 1/λ as the objective function instead of λ makes it
easier to define an optimal region to aim at - i.e. the region where I want all the points
to be. In Figure Figure 3.5 (b) the optimal region is not defined as λmax tends to infinity.

Figure 3.5: (a) Pareto front of a two minimum objectives (b) Pareto front of a maximum
objetive and a minimum objective.

60
Chapter 3. Methodology

The procedure of variation of parameters is then, the following: for the first cases
analysed, once the Pareto front is found, all dominated points can be discarded. This
enable the design process to be less time and computationally-consuming. By proper
examination of the non-dominated points, it is possible to observe the effect that each
parameter has on the structure behaviour. Looking at that effect it is easier to aim (vary
the parameters) at the direction depicted in Figure 3.5 (a).

3.1.9 Parameters of the Design


The design parameters that have been named until now and its range of variation, if
they are not fixed values, are:

• MTOW = 10 kg.

• n = 4.

• Lift per semi-wing = 200 N.

• Wingspan = 1.2 m.

• Modified chord = 0.17 m.

• Airfoil’s maximum thickness and position = 0.0222 m and 0.0532 m.

• Maximum weight mmax = 0.45 kg.

• Material properties: fixed. See Tables 3.3 and 3.4.

• Laminae thickness= 0.1 mm if it is made of unidirectional fibres and 0.05 mm


each if it is made of fabric.

• Core thickness= 0, 1, 2, 3, 4 and 5 mm.

• Skins: 3 main types of skin. As defined in Section 3.1.5. Materials Employed


and its Properties, each type has a with and without core version with its
differentiation of region within the skin - high and low load-bearing regions.

• Spar position: a few positions ahead of the maximum thickness point, a few
positions behind the maximum thickness point and one at the maximum thickness
point - i.e. 25, 35, 53.2, 65 and 75 mm.

• Spar width: fixed until a skin is selected as optimal and a position is selected as
optimal. Then vary its width and appreciate the effects.

• Second spar: its width and position will be decided based on the result of prior
analysis.

• Stringers: materials, region and core of them will be decided once a skin is selected
as optimal and a position is selected as optimal.

• Ribs: 3, 4 and 5 ribs per semi-wing will be tested. Width and shape will be decided
depending on the results obtained on prior analysis.

Spar(s), stringers and ribs parameters depend on the first set of analysis done. That
is explained in detail in Section 3.2. Design Process Description.

61
Chapter 3. Methodology

3.2 Design Process Description


In this section, the procedure followed - i.e. steps and decision making - to obtain
a monocoque composite wing structure is going to be explained. For that purpose, a
thorough discussion of each step and procedure followed will be delivered, and it will be
supported with a flowchart - as a summary of this section.

3.2.1 Initial Tests


The first step taken has been getting used to work with AbaqusTM and define a solid
model with which analyse all the cases of the study. The main aspects that have been
mandatory to be settled before beginning to do a lot of simulations are the following:

• Achieve a mesh convergence of the model with this tests, and then apply the same
mesh for all the cases run.

• Define proper boundary conditions that generate the least numerical inaccuracy.

• Define the lift distribution - load definition, to which the wing is going to be
subjected to.

• Check that the buckling modes yielded by the composite-made structure are
logical, i.e. ensure they are not a mathematically possible solution of the buckling
equation.

Besides all those 4 points, some trends of the structural behaviour can be
obtained in order to set in the best way the starting point of the design - see Section
3.2.2. Starting Point.

Now, each of these 5 aspects is going to be individually explained.

3.2.1.1 Mesh Convergence


In this initial tests, some cases have been simulated with and without spar. The ones
with the spar, though, have been the ones considered to obtain the mesh convergence of
the model. That is because the vast majority of cases will need of supporting elements
for the skin.

Keeping that in mind, the option selected when creating the mesh of each part in
AbaqusTM has been independent mesh, which has allowed to mesh all parts together
at the same time. With this option, the selection of the seeds per edge, from which
AbaqusTM generates the elements, can be done to all parts at the same time.

Using that feature, the same case has been run repeatedly with increasing density of
seeds per edge, i.e. increasing number of elements. For each repetition, the displacement
of the leading edge of the wing has been obtained. And comparing all the simulations
run up after each repetition, this process has been finished when the variation between
the difference in the displacement at the tip has been less than 1% with the last mesh
displacement at the tip, namely:

|δtip new − δtip old |


Dif f erence(%) = · 100 ≤ 1% (3.1)
δtip old

62
Chapter 3. Methodology

3.2.1.2 Boundary Conditions


The boundary conditions that must be defined are the restrictions of displacement
and rotation of the nodes at the end of the semi-wing, i.e.
U1 = U2 = U3 = U R1 = U R2 = U R3 = 0. That applies to the points at the skin and the
points at the spar(s) and/or rib. To do so in a software like AbaqusTM that has a
user-friendly interface, it’s nothing more than selecting the edges and/or surfaces to be
restricted and then apply the boundary condition.

When applying boundary conditions in FEA software, stres concentrators appear.


The latter, also known as stress singularities, are caused by the restriction movement of
points. Although stress at these singularities is very high, this does not mean that the
model results are incorrect overall.

First of all, the displacements are correct even at the singularity point. On the other
hand, the stress at the singularity will pollute the stress results near the singularity.
However, some distance away from the singularity the stress results will be fine. This is
an immediate consequence of St. Venant’s Principle1 , perhaps one of the most
important principle in FEA as it validates FEA results even with the presence of
singularities.

To try and avoid as much as possible this stress singularities, different ways of modelling
the geometry at the clamped end - known as encastre - of the semi-wing have been tested.
The geometrical models tested are the following:

• Encastre 1: restriction of the points at the end of the semi-wing.

Figure 3.6: First encastre.

• Encastre 2: thick panel completely restricted added to the end of the wing, and a
perfect contact interaction between the skin and/or the spar/rib with it defined.
1
St. Venant’s Principle is a principle stating that a system of forces in equilibrium applied to some
segment of a solid body produces stresses in that body that rapidly diminish with increasing distance
from the segment. Thus, at distances greater than the maximum linear dimensions of the region of load
application, the stresses and deformations are negligibly small.

63
Chapter 3. Methodology

Figure 3.7: Second encastre.

• Encastre 3: extension of 100 mm of each part (skin, spar, rib) which is fully
restricted.

Figure 3.8: Third encastre.

3.2.1.3 Load Definition


The lift that is generated in the semi-wing must be modelled in AbaqusTM . A few
non-viscous analysis have been run in the free-source software XFLR5 to obtain the lift
distribution of the wing of this study. The following lift and Cl distributions have been
obtained.

Figure 3.9: (a) Lift distribution (b) Cl distribution.

The lift decreases towards the tips due to the increase of the induce velocity towards
the tips - i.e. increase of the induced angle, which reduces the effective angle of attack -
thus, reducing lift. The distribution of the induced angle can be seen in Figure 3.10, and

64
Chapter 3. Methodology

in Figure 3.11 the effect of it its observed with the tip vortex generated which adds to the
total drag force generated by the wing.

Figure 3.10: Induced angle along the span.

Figure 3.11: Lift distribution and tip vortex effect downstream.

Modelling such a load distribution in AbaqusTM is not an easy task. For that reason,
some simplifications of this load distribution which yield the equivalent lift per
semi-wing defined in Section 3.1.9. Parameters of the Design, Lsw = 200N , have
been considered.

I have discretised the airfoil (see Figure 3.12) in elements and I have treated it with
an Excel
R
file - loaddefinition.xlsx (see Annex B.4. Loaddefinition.xlsx). In that file I
have applied different pressure distributions over the extrados and the intrados such that
the Lsw is accomplished.

65
Chapter 3. Methodology

Figure 3.12: Discretisation in elements of the airfoil.

To do so I have used Excel R


’s solver, which allows you to compute the value of a cell
that depend on other cells by changing this dependent cells values aiming for an
objective - i.e. in my case reducing the difference with the goal lift (Lsw ) and the
calculated lift with the input pressure (pressure cells are the one that are changed).

Using this file I have modelled the force needed to be generated in three ways. I have
considered that the percentage of the force generated by extrados and intrados is a 75 and
25%, respectively - see Figure 3.13.

Figure 3.13: Pressure distribution along extrados and intrados. Retrieved from [25].

The three load distributions tested in the initial tests have been the following:

• Distribution 1: uniformly distributed pressure over the extrados and the intrados
in order to generate 150 N and 50 N in each, respectively.

Figure 3.14: First load distribution.

• Distribution 2: uniformly distributed pressure over a band comprised between the


25 and 40% of the chord - where the resulting force is commonly found - at the
extrados and the intrados in order to generate 150 N and 50 N in each, respectively.

66
Chapter 3. Methodology

Figure 3.15: Second load distribution.

• Distribution 3: non-uniformly distributed pressure over the extrados and the


intrados - such that the suction peak can be modelled (Figure 3.13)- in order to
generate 150 N and 50 N in each, respectively. There have been 5 regions defined -
3 in the extrados and 2 in the intrados - to which 4 different uniformly distributed
pressure values have been applied.

Figure 3.16: Third load distribution.

3.2.1.4 Buckling Modes Check


The bifurcation problem is a non-linear problem and its solution can be
mathematically possible but without physical meaning. Adding to that, as it has been
remarked in Section 2.6 Buckling of Laminated Plates, boundary conditions
determine the buckling mode of failure - shape and critical load. To ensure that a
composite-made skin structure (higher complexity in the matrix Kg ) does not yield a
non-logical solution in a FEA model, I have made some tests to check it out.

The procedure taken has been the following: analyse the buckling modes of the same
structure but made with an isotropic equivalent material. To do so, I have preserved the
structural behaviour of the composite-made structure under static loads. I have varied
the thickness of the skin of the new structure and the E and ν until I have obtained the
same deformed shape under the same static loads. Once done that, I have run in both
composite and equivalent isotropic material structures a buckling analysis and compared
them.

3.2.1.5 First Tendencies Observed


For last, from all the initials tests performed to specify the mesh, the boundary
conditions, the load definition and some extra tests, some trends have been observed.
This trends will be noted and taken into account to define the starting point of the
design - see Section 3.2.2. Starting Point.

67
Chapter 3. Methodology

3.2.2 Starting Point


Having performed the initial tests, a starting point of the design must be defined -
i.e. an initial family of points (m, 1/λ) from which an optimum is going to be sought.
Taking into account all the design parameters seen in Section 3.1.9. Parameters of the
Design, the following set of cases have been decided to be analysed:

• 3 types of skin. The 3 skins defined in Section 3.1.5. Materials Employed and
its Properties, with its differentiation of region within the skin - high and low
load-bearing regions.

• The high load-bearing region position has been placed at the maximum thickness
of the airfoil, taking profit of this geometric condition that allows the highest I(x)
- where x axis is pointed chord-wise towards the leading edge - of the wing’s section.
The width of this region that will be studied is 20 or 30 mm.

• 6 types of core: from 0 to 5 mm, with 1 mm increments.

• 5 positions of the spar. One at the maximum thickness of the airfoil (53.2 mm
from the leading edge), two ahead of that position (25 and 35 mm from the leading
edge), and two behind that position (65 and 75 mm from the leading edge).

• The spar width is going to be fixed with 10 mm.

That adds to 3 x 2 x 6 x 5 = 180 different cases, which traduces to 360 AbaqusTM


simulations, as each ase requires 2 simulations: one for the static condition in which static
failure indices are obtained, and the other to analyse the resistance to failure by buckling.
Summing up this step of the process, the cases run are:

Starting Point
Parameter Value(s)
Skin 3 types
Width of reinforced region 20 / 30 mm
Core thickness 0 / 1 / 2 / 3 / 4 / 5 mm
Spar position 25.5 / 35.5 / 53.2 / 65.5 / 75.5 mm
Spar width 10 mm
Total = 180 cases

Table 3.5: Set of parameters analysed in the first family of points.

3.2.2.1 Decision Making


Once this family of points is obtained, the points that do not fulfil the conditions
specified for the Hashin and Tsai-Hill criteria (see Section 2.8. Failure Criteria)
will be discarded. Also, the points which have a mass m > 0.450 kg or a buckling
constant λ < 1 will also be discarded.

With only the points that do not fail statically or by buckling remaining, the Pareto
optimality is applied - see Section 2.9. Pareto Optimality. The Pareto front is obtained,
and the next iterations (variation of parameters) will be done only from this non-dominated
points. The rest of valid points, from the failure point of view, are now discarded too.

68
Chapter 3. Methodology

Figure 3.17: Decision making for each point of the set of points analysed after each step.

This procedure has been implemented with a Matlab code (Decisionmaking.m) - see
Annex B.1. Decisionmaking.m.

3.2.3 Improvements
Once the Pareto optimal points are achieved, the step that follows is: select the best
points of the Pareto front and introduce improvements to those points. The best points
of the Pareto front will be the ones closer to the optimal region, as Figure 3.5 (a)
depicts.

The improvements that are going to be tested on the non-dominated points are the
following. First, to each Pareto optimal point each of these improvements are going to be
tested separately:

• Improvement 1: vary the spar width to 5, 15, 20 and 25 mm.

• Improvement 2: add a second spar of the same width (10 mm) placed at 110,
120, 130 and 140 mm from the leading edge.

• Improvement 3: add stringers with 3 different widths. The core combinations


for the stringers region (stiffened region), and the no-stringer region (non-stiffened
region) are shown in Table 3.6.

Core combinations
No-stringer region thickness (mm) Stringer region thickness (mm)
0 1/2/3/4/5
1 2/3/4/5
2 3/4/5
3 4/5
4 5

Table 3.6: Core combinations for the stringers improvement.

That adds to (Z x 4) + (Z x 4) + (Z x 45) = 53Z possible new different cases,


where Z is the number of non-dominated points selected in the previous step.

After obtaining these new points, the same process of decision making explain in the
previous Section 3.2.2.1. Decision Making is followed. See Figure 3.17 to recall the
procedure to which all the points are subjected to in order to determine their optimality
for the study of this thesis.

69
Chapter 3. Methodology

Once applied this decision making, next step is apply some of the new points’
improvements at the same time. Then, we have:

• Improvement 4: combination of a spar position, spar width, set of stringers or


second spar, based on the best points (Pareto optimal points) selected from the set
of points obtained from the 3 prior improvements. For instance, if there is a very
good point with a certain spar width, and another very good point with a certain set
of stringers, it will be interesting to try a combination of those two improvements
and see what is the outcome of that.

The total amount of different cases up to this point is very complicated to estimate.
The process to follow after this new cases are run is the same as the one explained
before. The decision making is applied and new non-dominated points are obtained.

Finally, with the best remaining points, a fifth and final improvement is implemented
in such points.

• Improvement 5: add ribs to the non-dominated points. The tests that will be
performed will be with 3, 4, and 5 ribs per semi-wing.

Decision making procedure is applied again and the final Pareto front is obtained.
Summing up, the sets of steps explained in this section are the following:

Steps followed after the first family of points are obtained


Improvement Important parameter Possible values
1 Spar width 5 / 15 / 20 / 25 mm
2 Second spar position 110 / 120 / 130 / 140 mm
3 Stringers 3 different widths
Decision making
4 Combination Combination or improvements
Decision making
5 Ribs 3 / 4 / 5 per semi-wing

Table 3.7: Steps followed after the first family of points are obtained.

3.2.4 Final Decision


With the final Pareto front, there are a certain number of points - i.e.
materially-constituted and geometrically different moncoque composite structures, which
are optimal points. These points are each of them better than the rest of the Pareto
front points in one of the two objectives of my interest: m or 1/λ.

With this cases on the table, a proper consideration of all of them will be done, and
by thorough and detailed reasoning some of the will be chosen as the most optimal for the
goal of this thesis.

3.2.5 Flowchart of the Design Process


Before ending this section, a flowchart of the whole design process is attached. The
Analysis procedure flowchart and the Decision making procedure flowchart are shown
in Figure 3.25 and 3.17, respectively.

70
71
Figure 3.18: Flowchart of the design process.
Chapter 3. Methodology
Chapter 3. Methodology

3.3 Analysis Process Description


In this section, the process followed to perform an analysis - in order to obtain the
information necessary for every case - is explained.

The process is divided in 4 sub-processes: geometry preparation, AbaqusTM


model making, simulation of the analysis and post-processing. In relation with
Section 2.7.3. AbaqusTM , the pre-processing part is the geometry preparation and
the AbaqusTM model making, the simulation part is the simulation of the analysis and
the post-processing part is the equally-defined post-processing.

3.3.1 Pre-processing
3.3.1.1 Geometry Preparation
Each case has its own geometrical parameters. Being so, the first thing to do in order
to analyse the case in AbaqusTM is to adjust the geometry to the case’s geometric
features. Let’s recall which are the main geometrical parts that comprise this study:
wing skin, main spar, secondary spar and ribs.

In order to quicken the process, the geometrical parts have been modelled (with
CATIA V5R20) using parametric formulas. These geometric parameters have been
stored in an Excel R
file - parameters.xlsx (See Annex B.6. Parameters.xlsx), from
which the update of any of the values has a modification effect of the part. The wing
skin has been modelled as a surface - allowing me to define different material sequences
at different regions of it in AbaqusTM ; and the main spar, secondary spar and ribs as
solid parts.

The parameters that have been defined to model the geometry are the following:
• spar w and spar o, which represent the main spar width and offset from the leading
edge of the wing, respectively. With these two parameters the main spar has been
modelled. Also, the skin surface has been partitioned with these two parameters -
creating edges on it. The latter has been done in order to avoid wrong imported
geometry in AbaqusTM . If the skin is partitioned as mentioned, as both the main spar
and the surface share the extrados and intrados edges, perfect geometric coupling
between the two parts is achieved - see Figure 3.19.

Figure 3.19: (a) Wrong geometry due to lack of partitioning (b) Good geometry with skin
partitioning.

• spar2 w and spar2 o, which represent the secondary spar width and offset from
the leading edge of the wing, respectively. With these two parameters the secondary
spar has been modelled. Also, for the same reasons as I have already explained, the
skin surface has been partitioned with these two parameters - creating the reference
edges on it.

72
Chapter 3. Methodology

• S1 w, S1 o, S2 w, S2 o, S3 w and S3 o, which are the width and offset from the


leading edge of each stringer region on the skin. With these parameters, the skin
extrados has been partitioned to define the stringer regions. The partitioning process
can be also done in AbaqusTM , where you can define datum planes to partition the
imported geometry. But doing it with CATIA formulas assignation feature has
allowed me to do it in a more time-efficient way.

• rib4 o and rib5 o, which are the offset of ribs 4 and 5 from the plane of symmetry
of the semi-wing. As detailed in Section 3.1.9. Parameters of the Design, three
combinations of ribs are going to be tested: 3, 4 and 5 ribs per semi-wing. The 5 ribs
have been modelled as follows: rib 1 and rib 3 are the wing symmetry plane and tip
rib, respectively. Rib 2 is the rib fixed at the symmetry plane of the semi-wing, for
3 and 5 ribs combinations. And the last two ribs, rib 4 and 5, can be placed either
at an offset of ±100 mm - 4 ribs combination - or ±150 mm - 5 ribs combination -
from the symmetry plane of the semi-wing.

For last, as it has been detailed in Section 3.2.1.3. Load Definition, there are 3
different ways of defining the sizing load of the structure. If either load distribution 2 or
3 detailed in that section is the one that results the most appropriate for the study, then
partitioning of the wing skin - to define the different uniformly distributed pressure
values - must be done.

To end this section, a summary of the geometry preparation is delivered:

Case geometry −→ parameters.xlsx modification −→ open CATIA V5R20 −→


update each part modified −→ save modified parts −→ Geometry is ready

3.3.1.2 AbaqusTM Model Making


With the geometry ready, the AbaqusTM model can be built. The set of steps that
takes to do so are the following:

1. Import the geometry. AbaqusTM has compatibility with CATIA V5R20.

2. Assign surfaces and sets. For the sake of quickening the process, I have seen that
it is convenient to assign surfaces and sets in each part if they are going to be used
to define a load, boundary condition or an interaction. So, skin surfaces must be
defined - recall that different pressure values may be assigned, as long as surfaces in
skin and spar/ribs which are going to have a perfect contact interaction. Then, the
sets of points to define a boundary condition must be selected too.

3. Assign materials. For the non-composite material part, I have created a solid
homogeneous section with balsa wood. This section is assigned to the spar and
ribs. For the skin, a composite layup is created, in which the following has to be
defined: local coordinate system (1 direction is the fibre direction, 2 direction is the
matrix direction, and 3 direction is the matrix and out-of plane direction), number
of plies, orientation and thickness of each ply, the number of integration points per
ply and the region of the part for each ply - see Figure 3.20. When stringers are
defined, two composite layups must be created - as different core thickness are to be
assigned.

73
Chapter 3. Methodology

Figure 3.20: Definition of a composite layup in AbaqusTM .

The great thing about the composite layup feature of AbaqusTM is that you define
7 layers, as in Figure 3.20, but defining the regions of each layer the skin can have
different layer sequences - high and low load-bearing regions (see Section 3.1.5. Material
Employed and its Properties. The next thre figures, Figure 3.21, Figure 3.22 and
Figure 3.23 depict the composite layup defined in Figure 3.20.

Figure 3.21: Stack-up plot of a 5 layer region, where no CE is employed.

74
Chapter 3. Methodology

Figure 3.22: Stack-up plot of a 7 layer region, where CE is employed.

Figure 3.23: Close-up of the stack-up plot 7 layer region.

4. Create instances of the different parts. The instance module allows the AbaqusTM
user to instantiate one part more than once if it is necessary for the model. Then
the relations of this instances are defined in the interaction module. Also, when
the instance is defined the mesh type is selected: dependent or independent mesh. I
have selected independent mesh, as so has allowed me to mesh all instances at the
same time. A dependent mesh is related directly with the part while an independent
one is related with the instance.

5. Step and field output requests. A step must be defined, in which the load
and boundary conditions are applied. The step type used for this thesis are static
and buckle ones. Regarding the field output requests, the output variables to be
calculated must be defined. For this thesis I have extracted the following variables:
stress, displacement and Hashin indicators for the static test; buckling constant for
the buckling test.

6. Interactions. Interactions between parts are defined in this module. I have defined
a tie interaction between the skin and spar(s)/ribs, which simulates a perfect bonding
between parts - see an example in Figure 3.24. For this module the surfaces that
have been defined earlier on are used.

75
Chapter 3. Methodology

Figure 3.24: Definition of a tie interaction in AbaqusTM .

7. Loads and boundary conditions. I have defined uniformly distributed pressure


in the skin, with the load option: Pressure. And I have defined the encastre
boundary condition using the option: Symmety/Antisymmetry/Encastre. For
this module the surfaces and sets that have been defined earlier on are used.

8. Mesh. I have meshed the parts assigning the parts the convergence mesh obtained
in the initial tests - see Section 3.2.1.1. Mesh Convergence.

9. Job input. For last, a job must be defined in order for the model to be simulated.
Recall that for every case definition, two models must be defined - static and buckling
tests - and so, two jobs must be defined as well. When the job is created, I have
written the job into a .inp file. This file is the file that AbaqusTM uses to solve the
model.

To end this section, a summary of the making of the AbaqusTM model is delivered:

Import geometry −→ assign surfaces and sets −→ assign materials −→ create


instances −→ step and field output request definition −→ interactions definition
−→ loads and boundary conditions definition −→ mesh the parts −→ job input

3.3.2 Simulation of the Analysis


Once the job has been defined and it has been written the input file (.inp), the
static/buckling case that the job defines can be simulated.

Running a simulation in AbaquTM s is a very computer-demanding task, and makes


almost impossible to work on another tasks at the same time. For that reason, the
submittal of these analysis has been done in batches. The batches have been organised as
follows:

• Batch 1: set of 180 initial cases, which are 360 abaqusTM models and thus, jobs.

• Batch 2: set of cases after the first improvement.

76
Chapter 3. Methodology

• Batch 3: set of cases after the second improvement.

• Batch 4: set of cases after the third improvement.

• Batch 5: set of cases after the fourth improvement.

• Batch 6: set of cases after the fifth improvement.

The simulation of the batch of jobs has been done by implementing a python script
(python is the programming language of AbaqusTM ) named BatchAnalysis.py -
BatchAnalysisStringers.py for the cases that contain stringers.

This script automates the chain simulation of the batch of jobs by just introducing
the job names before submitting it - see Annex B.2. BacthAnalysis.py and B.3.
BacthAnalysisStringers.py. The script executes a job when the previous one has
ended, and keeps going after it has executed all of them. The command line to simulate
the case associated with the job is the following: ’abaqus job=job name interactive
cpus=# of cpus (parallellization) ’.

The interactive command is the command that makes the jobs run one at a time. For
the type of jobs I have had to run in this thesis, it is more convenient to run jobs one at
a time than all at the same time. Running jobs at the same time is more time-efficient
when the jobs are simple and with coarse meshes.

Besides executing the job files, the script also performs a task, which is part of the
post-processing process - see next Section 3.3.3. Post-processing.

3.3.3 Post-processing
Once the simulations have been run, AbaqusTM generates an output database file
(.odb) for each case in which the results can be displayed. Opening each .odb file to
obtain the values of the variables of interest is an arduous and slow process. For that
reason, the python script used to run all the batch of analysis is also used for
post-processing the information required for the decision making process explained in
Section 3.2. Design Process Description.

In the post-processing process the Hashin indices are collected, as long as the
maximum displacement and the stresses at every element. As commented in Section 2.8.
Failure Criteria, the only criterion that AbaqusTM incorporates is the Hashin one. For
that reason, the calculus of the Tsai-Hill criterion is implemented in the script.

To calculate the Tsai-Hill index, the stresses of every element are used (at the
centroid of the element). As the criterion must be validated for every composite material
employed - CE and GE, there must be a differentiation of different material elements.
Luckily, AbaqusTM creates material-sorted sets of elements, allowing to make this easy.

That is all for the static analysis. For the buckling analysis, I have not found a way
of writing the buckling constant as an output variable. What I have done instead is
adding code to the script so once it has finished post-processing the static cases, it opens
one by one the buckling cases. Doing so, I have noted manually the buckling constant.

All the information of the batch of analysis has been printed into .txt file -
output.txt. To see how post-processing is implemented more in detail, see Annex B.2.

77
Chapter 3. Methodology

BacthAnalysis.py.

For last, the only variable which is left to know is the mass of every case/specimen.
That has been calculated with an Excel R
file - mass.xlsx. That file has been done such
that by introducing the following parameters, it gives you the mass of each specimen:
width of the reinforced section, width of the stringers, thickness of the core (stringer and
no-stringer regions) and balsa wood volume. See Annex B.5. Mass.xlsx.

With the mass calculated, each mass has been add to its belonging case in the
output.txt. Once done that, the ouput.txt can be loaded into Decisionmaking.m (see
Section 3.2.2.1. Decision Making). Running this Matlab file, the post-processing
process is ended with the valid (non-dominated) and dispensable (dominated) points
(cases).

Finally, the next scheme sums up this section and the previous, Section 3.3.2.
Simulation of the Analysis.

Python script (BatchAnalysis.py) −→job execution −→ post-processing of


data −→ mass calculation −→ unification of data −→ Matlab script
(decisionmaking.m) −→ decision making

3.3.4 Flowchart of the Analysis Process


Before ending this section, a flowchart of the whole analysis process is attached.

78
79
Chapter 3. Methodology

Figure 3.25: Flowchart of the analysis process.


Chapter 4

Results

In this chapter, the results obtained in this thesis are depicted and explained. The
chapter is organised as the Section 3.2. Design Process Description. The results will
be delivered as follows:

1. Initial tests.

(a) Mesh convergence.


(b) Boundary conditions.
(c) Load definition.
(d) Buckling modes check.
(e) Main trends and structural behaviour.

2. First batch of cases - starting point.

3. Second batch of cases - improvement 1.

4. Third batch of cases - improvement 2.

5. Fourth batch of cases - improvement 3.

6. Fifth batch of cases - improvement 4.

7. Sixth batch of cases - improvement 5.

4.1 Initial Tests Results


In this section, the following results are going to be presented: first, the converged
mesh that has then been used for the rest of cases; secondly, from the different ways of
defining the boundary condition considered in Section 3.2.1.2. Boundary Conditions,
one has been picked as the most optimal; third, the best way of defining the sizing load
from the options presented in Section 3.2.1.3. Load Definition is selected; then, the
buckling modes are verified between the composite-made skin and a isotropic material-
made skin; and finally, an insight of the trends observed during the initial tests and the
basic features of the structural behaviour noticed are given.

4.1.1 Converged Mesh


The tests performed to obtain the converged mesh have been done with the following
settings:

80
Chapter 4. Results

Model settings
Parameter Value
Skin A
Spar width 10 mm
Spar position 53.2 mm
Core thickness 2 mm
CE reinforcement width 10 mm

Table 4.1: Parameters used to perform the mesh convergence tests.

Regarding the load, the load distribution that has been applied is the first one
proposed in Section 3.2.1.3. Load Definition. And regarding the boundary condition
application, the encastre used has been the first one proposed in Section 3.2.1.2.
Boundary Conditions.

As explained in Section 3.2.1.1. Mesh Convergence, a static analysis has been


performed with different meshes. For each analysis the curve displacement (δ) vs.
semi-span (b/2) has been obtained. The final mesh has been picked when the difference
between the δtip new and δtip old represents less than a 1% of the δtip old - see equation 3.1.

Next, the results are presented - recall that seeds are the points that AbaqusTM places
at the edges to generate the mesh:

Mesh Skin elements Spar elements Seeds (seed/mm) δtip (mm) Difference (%)
1 362 20 30 25.541 -
2 1122 40 15 25.244 1.16
3 4082 480 7.5 22.674 10.18
4 17444 3078 3.5 22.125 2.42
5 52802 16500 2 22.013 0.51

Table 4.2: Results of the convergence mesh tests.

Figure 4.1: Displacement vs. semi-span for the 5 meshes.

81
Chapter 4. Results

Figure 4.2: Close-up of the Displacement vs. semi-span for the 5 meshes.

The mesh that has been used for the rest of simulations of the thesis is mesh 5, which
has the following characteristics: 52802 elements for the skin and 16500 elements for the
main spar. For other parts that may be included in the model (e.g. second spar, ribs), a
seeding of 2 seeds/mm will be used.

4.1.2 Final Encastre


The tests performed to determine which encastre
(U1 = U2 = U3 = U R1 = U R2 = U R3 = 0) is the best to minimise as much as possible
stress concentrators have been done with the settings and load distributions as in the
previous Section 4.1.1. Converged Mesh, except for the encastre which has been
varied.

The results obtained for the three encastres proposed in Section 3.2.1.2. Boundary
Conditions are presented next. The main output variable that is going to be compared
between the three options is the stress σ11 , as it is the highest stress generated in the
structure - due to the bending moment generated by the lift distribution.

82
Chapter 4. Results

• Encastre 1. Maximum |σ11 | is 490.5 MPa (compression stress).

Figure 4.3: Stress distribution with the encastre 1.

• Encastre 2. Maximum |σ11 | is 484.0 MPa (compression stress).

Figure 4.4: Stress distribution with the encastre 2.

83
Chapter 4. Results

• Encastre 3. Maximum |σ11 | is 479.3 MPa (compression stress).

Figure 4.5: Stress distribution with the encastre 3.

Note that in the last three Figures, the spar has been hidden in the .odb fie in order
to visualize better the stress distribution at the skin. Also, the stress field shown is the
envelope one (all plies together) with the option Max absolute value of the display
options of AbaqusTM , which shows the max absolute value at each element.

As the stress distributions with the three encastres are very similar, the final encastre
selected is the third one (encastre 3) it has the less stress σ11 at the symmetry plane of
the wing.

4.1.3 Final Load Distribution


The tests performed to determine which load distribution is the best to resemble
the real lift distribution over the wing have been done with the settings as in Section
4.1.1. Converged Mesh, but with the final encastre already incorporated and varying
the load distribution.

The criterion to select one load distribution over another is difficult to set as I don’t
know how the structure should precisely behave - i.e. how much should it displace at the
tip or how much the stress values should rise. So, firstly, I am going to compare both the
stress and displacement field of the three configurations.

The stress field obtained for the three load distributions is the same as in Figure 4.5.
That Figure corresponds to the first proposed load distribution - the three possible
distributions are explained in Section 3.2.1.3. Load Definition. That stress field
seems reasonable, being the region where carbon fibre/epoxy is placed the high
load-bearing region and leaving the glass fibre/region with less stress concentration.

As I said, the three distributions yield the same stress field with minor differences.
The maximum stress σ11 yielded in each case, from distribution 1 to 3, are 479.3 MPa,
466.7 MPa and 473.6 MPa, respectively. Regarding the displacement field, the three also

84
Chapter 4. Results

exhibit the the same displacement distribution (see Figure 4.6) with the only difference
only laying on the maximum displacement at the tip. In each case, from distribution 1 to
3, the latter is 22.99 mm, 22.32 mm and 22.67 mm, respectively.

Figure 4.6: Displacement field exhibit by the three load distributions, this one belonging
to the first one.

As the static analysis doesn’t help to decide which of the distributions is better, I have
performed a buckling analysis. I have only analysed the three first buckling modes for each
distribution, and the deformed shape of the latter is the same for the three distributions
(see Figures 4.7, 4.8 and 4.9). Furthermore, the bucking constants (λ) do not vary much
from one load distribution to the other. In each case, from distribution 1 to 3, the first
mode λ is 3.35, 3.45 and 3.37, respectively.

85
Chapter 4. Results

Figure 4.7: First buckling mode.

Figure 4.8: Second buckling mode.

Figure 4.9: Third buckling mode.

The first and second modes happen in the extrados - due to the compression stress
caused by the bending moment. The third mode happens in the intrados, which seems a
bit odd as that is the tension region of the wing. But due to the net force produced by
the wing (the lift), there is a shear stress distribution along the wing, which causes a
compression force in the intrados - this phenomena is known as shear buckling. The
latter will be explained more in detail in next Section 4.1.4. Buckling Modes
Verification. Lastly, the laminae at which the section fails is the CE laminae, as CE
regions are the ones that bear the highest stresses.

86
Chapter 4. Results

Coming back to the matter of this section, that is choosing which is the best load
distribution to mimic the lift distribution, I have seen so far: stress and displacement
distribution, as well as the buckling modes, are the same for the three possibilities - so
are the characteristic values of each field. Being so, the decision I have taken is not based
on numerical results but on a theoretical basis. The load distribution selected is the load
distribution 3 - see Figure 3.16 - for being the one that best resembles the characteristic
suction peak of an airfoil, and thus the lift distribution along the wing.

4.1.4 Buckling Modes Verification


In this section is intended to ensure that a composite-made skin structure (higher
complexity in the matrix Kg ) does not yield a non-logical buckling solution in a FEA
model. To do so I have followed the following settings: load distribution 3, encastre 3 and
the material configuration shown in Table 4.3.

Model settings
Parameter Value
Skin A
Core thickness 0 mm
CE reinforcement width 0 mm

Table 4.3: Parameters used to perform the buckling modes verification.

Note that no spar nor CE has been used. But prior to jumping to this tests, I
want to explain the buckling phenomena depicted in Figure 4.9, which happened in the
intrados - as it is going to be one of the modes obtained in this section. That buckling
phenomena is mainly caused by the shear stress at the wing. let me show the only three
non-zero stresses distribution at the intrados (σ11 , σ22 and τ12 ) - model settings of Section
4.1.1. Converged Mesh, encastre 3 and load distribution 3.

Figure 4.10: Intrados σ11 distribution.

87
Chapter 4. Results

Figure 4.11: Intrados σ22 distribution.

Figure 4.12: Intrados τ12 distribution.

It is important to recall that shell elements are employed in this thesis - in the modelling
of the skin, and that this elements use a local coordinate system as a reference (see Section
2.7.4. Shell Elements). Being so, the results depicted in Figures 4.10, 4.11 and 4.12 are
referenced to that local coordinate system. The latter is different for every shell element,
as it follows the normal direction of each element. The 1 direction is in the span-wise
direction, the 3 direction is the normal of the element, and the 2 direction is the one that
forms an orthogonal coordinate system with the 1 and 3 directions. Some local coordinate
systems of the intrados are depicted in Figure 4.13:

88
Chapter 4. Results

Figure 4.13: Local coordinate systems at the intrados.


Following that local coordinate systems, and the stress fields depicted in Figures 4.10,
4.11 and 4.12, the stresses (at the intrados) per element are:

Figure 4.14: Stresses in an intrados element.


Rotate now the element in Figure 4.14 ±45◦ and in you have compression stress in the
intrados - some in the span-wise derection and other in the chord-wise direction. Having
understood that, I made a prediction which was the following: the buckling modes of the
Figures 4.7, 4.8 and 4.9 with a fabric skin oriented at ±45◦ , and the buckling modes of
a fabric skin oriented at 0/90◦ should be way different. Figure 4.15 depicts what I have
explained in this paragraph:

Figure 4.15: Stress at the intrados and lamina equivalence of a pair of laminae at ±45 and
0/90◦ .
The last Figure shows the equivalence of the two laminae to a laminae of equivalent
longitudinal El0 and transverse Et0 modulus, rotated a certain θ. The 0/90◦ is equivalent
to ±45◦ laminae of equivalent modulus, and vice versa with the two laminae at ±45◦ .

89
Chapter 4. Results

That fact makes the 0/90◦ set of laminae weaker in the directions where compression
stresses appear at the intrados. Being so, this configuration will be much more susceptible
to have buckling modes at the intrados than the ±45◦ set. I have extracted the first three
modes of the same structure as for Figures 4.7, 4.8 and 4.9, just changing the skin to
0/90◦ . The results are the following:

Figure 4.16: First buckling mode of the 0/90◦ GE skin configuration.

Figure 4.17: Second buckling mode of the 0/90◦ GE skin configuration.

90
Chapter 4. Results

Figure 4.18: Third buckling mode of the 0/90◦ GE skin configuration.

Figures 4.16, 4.17 and 4.18 depict what I have predicted: 0/90◦ configuration is
weaker at the intrados and thus, the first two modes happen there. Also, the buckling
mode is more agressive. In comparison with the buckling mode depicted at Figure 4.9,
which is only a wrinkle, the buckling mode at the intrados makes almost the whole
section buckle - mainly the trail edge region.

Nevertheless, the buckling modes happen at λ higher than the ±45◦ configuration.

Now that this intrados buckling phenomena has been explained, it is time to check if
the buckling modes of the composite-made structure yield a physically logical solution.
As it has been explained in Section 3.2.1.4. Buckling Modes Check, the procedure
that I have taken is the following:

Analyse the buckling modes of the same structure but made with an isotropic
equivalent material. To do so, I have preserved the structural behaviour of the
composite-made structure under static loads. I have varied the thickness of the skin of
the new structure and the E and ν until I have obtained the same deformed shape under
the same static loads. Once done that, I have run in both composite and equivalent
isotropic material structures a buckling analysis and compared them.

Doing this, I have two structures that behave in the same way statically, and thus I can
check if they also do the same in a buckling situation. Composite structure characteristics
can be found in Table 4.3. The properties of the equivalent isotropic material and the
thickness of the skin are the following:

91
Chapter 4. Results

Equivalent structure
Parameter Value
E 3127.5 MPa
µ 0.3
Skin thickness 1.5 mm

Table 4.4: Parameters of the equivalent structure.

Here are two Figures with the deformed shape of the two structures under the
static analysis:

Figure 4.19: Deformed shape of the composite-made structure.

Figure 4.20: Deformed shape of the equivalent structure.


With the equivalent structure on the hand, I have obtained the first 10 buckling modes
of each structure. Those are presented next.

92
Chapter 4. Results

Figure 4.21: (a) Mode 1 of the composite-made structure (b) Mode 1 of the equivalent
structure.

Figure 4.22: (a) Mode 2 of the composite-made structure (b) Mode 2 of the equivalent
structure.

Figure 4.23: (a) Mode 3 of the composite-made structure (b) Mode 3 of the equivalent
structure.

Figure 4.24: (a) Mode 4 of the composite-made structure (b) Mode 4 of the equivalent
structure.

Figure 4.25: (a) Mode 5 of the composite-made structure (b) Mode 5 of the equivalent
structure.

93
Chapter 4. Results

Figure 4.26: (a) Mode 6 of the composite-made structure (b) Mode 6 of the equivalent
structure.

Figure 4.27: (a) Mode 7 of the composite-made structure (b) Mode 7 of the equivalent
structure.

Figure 4.28: (a) Mode 8 of the composite-made structure (b) Mode 8 of the equivalent
structure.

Figure 4.29: (a) Mode 9 of the composite-made structure (b) Mode 9 of the equivalent
structure.

Figure 4.30: (a) Mode 10 of the composite-made structure (b) Mode 10 of the equivalent
structure.

As it can be seen in Figures 4.21 to 4.30, the modes of both structures are very
similar. Despite that for the same mode some deformed shapes do not coincide, that is
not important as the order of the modes depend on the stiffness (the different modulus
of each direction) of the structure - as I have explained earlier with the ±45 and
0/90◦ GE skin configurations.

Evidently, the buckling constants λ for the equivalent structure are way lower than
the composite-made one. That is due to the same reason: the difference in stiffness causes
this phenomena. But what is important is that the deformed shapes of the modes really
look alike, and by that I can assure that the results I’m yielding can be believed from a
physical standpoint.

94
Chapter 4. Results

4.1.5 Trends and Structural Behaviour


Finally, besides running the tests that I have already explained, I have also have
tried different parameter configurations to observe which tendencies does the structure
behaviour follow.

Parameter Value Observation


Core 0 mm λ >1. Even when adding stringers to the structure, the
lack of I make the regions without core very vulnerable to
buckling. Failure indices are sometimes higher than 1.
Core 1 mm Sometimes λ >1. The lack of I make the regions with 1 mm
of core vulnerable to buckling.
Spar position - For the same skin configuration and fixed spar width, it
seems that the best results (λmax , Hashin and Tsai-Hill
indices minimum) are yielded when the spar is placed
towards the maximum thickness point of the airfoil.
Spar width - For a certain configuration, increasing the spar width means
increasing the total mass but also increasing λ and reducing
the Failure indices. That is mainly due to the amount of
shear stress that is now transmitted by the spar, that is
greater than with a lower width. In the other hand, reducing
the width of the spar has the opposite effect.
Stringers - Adding stringers to a certain configuration means adding
weight to the structure but also increasing λ. It is hard to
tell what is the trend: whether is better to have a great
reinforced surface region or not.
Second spar - Adding a second spar adds weight but reduces the Failure
indices as more shear stress is transmitted from extrados to
intrados. Also λ increases.
Reinforced region - Test with 10 mm of width of CE reinforcement have proved
that the structure fails statically (Failure inidices >1). Also,
trying to remove one laminae of CE of the reinforce section
(1 at the extrados and 1 at the intrados) has result in static
fail - when doing this the width was 20 or 30 mm. Despite
that, this seems a good idea for the stringers.

Table 4.5: Trends observed in the initial tests.


From all this trends observed, no further conclusion can be extracted without using
the Pareto optimality (see Section 2.9. Pareto Optimality) as a tool to distinguish
optimal points from non-optimal points - being a point a case with a mass and a λ,
having the Failure indices lower than 1.

However, as all cases I have tested without core (0 mm) both fail statically and by
buckling, I have discarded the 0 mm core cases for the first batch of cases - the starting
point (see Section 3.2.2. Starting Point), reducing the amount of cases from 180 to
150 cases.

To end this section and jump to the optimisation of the design, I just want to remark
that the behaviour of the structure - statically and buckling, has been already shown
throughout the Section 4.1. Initial Tests Results. That behaviour - stress and
displacement field, buckling modes - is the one that all structures that will be tested
in the thesis are going to follow.

95
Chapter 4. Results

4.2 First Batch of Cases - Starting Point


As described in Section 3.2.2. Starting Point, a set of initial cases have been run
to start the design process of the monocoque wing structure with composite materials.
This batch of cases have been made from the combination of the parameters described in
Table 3.5, which have been selected trying to cover a wide varied range.

But taking into account the first conclusion - sections will have to be made of
sandwich panels, i.e. core thickness ≥1 mm - extracted in last Section 4.1.5. Trends
and Structural Behaviour, the number of cases reduces to 150.

Next, the 150 cases are represented. Each is characterized by its mass m (kg) and the
inverse of the first buckling mode constant 1/λ in an XY plot - x axis representing the
mass and y axis representing the inverse of the first buckling mode constant.

Each point also has its failures indices (Hashin and Tsai-Hill criteria indices) and two
tags: a numerical one (order) and a code made by letters, numbers and colours. This
information can be consulted in Annex C. Numerical Results. Recall that the failure
indices are used to determine if a point is valid or not - see Section 3.2.2.1. Decision
Making.

Next, all the points (m,1/λ) run for this starting point are presented:
1.5
1.4
Batch 1 points
1.3
1.2
1.1
1
0.9
0.8
1/λ
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
mass (kg)

Figure 4.31: 150 points of Batch 1.

The first step of the decision making procedure is reject the points that satisfy either
m > 0.45 kg or 1/λ >1. Drawing two lines, m = 0.45 and 1/λ = 1 as limiting lines, a
rectangle delimited by the latter lines and the x and y axes is created. The points that
lay outside that rectangle are rejected.

96
Chapter 4. Results

1.5
1.4
Batch 1 points
m = 0.45 kg
1.3
1/λ = 1
1.2
1.1
1
0.9
0.8
1/λ
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
mass (kg)

Figure 4.32: Batch 1 with limiting lines m = 0.45 and 1/λ=1.

1
Batch 1 points
0.9

0.8

0.7

0.6

1/λ 0.5

0.4

0.3

0.2

0.1

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.33: Batch 1 with the first step of the decision making procedure applied.

After this step, 19 points have been rejected. Hence, points that fail by buckling or
exceed the maximum mass set are discarded.

The second step of the decision making procedure is reject any point that has at
least one failure index which satisfies IF ≥ 1. By inspection of every point failure indices,
invalid points are discarded, leaving just the ones that satisfy the failure (statical and
buckling) and mass premises.

97
Chapter 4. Results

1
Batch 1 points
0.9

0.8

0.7

0.6

1/λ 0.5

0.4

0.3

0.2

0.1

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.34: Batch 1 with the second step of the decision making procedure applied.

After this step, 102 points have been rejected, with only 29 points remaining. All this
points failed statically. As seen in Figures 4.10, 4.11 and 4.12, the stress perturbations of
the encastre take a great portion of the semi-wing. That concentration - that is for sure
increased by the numerical simulation - has caused stress peaks that make the structure
suffer statically.

It must be recalled that the load factor n tested is 4, so it is a very demanding flight
performance. Despite that, failure indices IF are very influenced by this stress
concentration, resulting in this large amount of points rejected in this step.

The last step of the decision making procedure, which recall is implemented with
a Matlab code named Decisionmaking.m (see Annex B.1. Decisionmaking.m.), is find
the Pareto optimal points or the also-known Pareto front. The points belonging to this
front are the non-dominated points, and thus are the ones from which the design process
is continued - see Section 2.9. Pareto Optimality.
1
Batch 1 points
0.9 Pareto points
Pareto front
0.8 Axes bisector

0.7

0.6

1/λ 0.5

0.4

0.3

0.2

0.1

0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.35: Batch 1 valid points and the Pareto front.

As all the cases I have worked with in this thesis have a mass >0.25 kg, from now on

98
Chapter 4. Results

the plots that are going to be presented will have m = 0.25 as the initial value. Also, it
will help depicting the points on the plots, as there will be a lot of them.
1
Batch 1 points
0.9 Pareto points
Pareto front
0.8 Axes bisector

0.7

0.6

1/λ 0.5

0.4

0.3

0.2

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.36: Close-up of the Batch 1 valid points and the Pareto front.
With the Pareto front achieved - made of 17 points out of the 29 - it is time to consider
which are the most optimal within the Pareto optimal. As depicted in Figure 4.35, the
axes bisector is indicated with an arrow. The arrow points towards the optimal region
where I want all the points to be.

Points near the bisector are the most optimal points I can get, as they are points
which are both strong in the two objective functions - minimizing the mass and minimizing
the inverse of the first buckling mode constant (i.e. maximizing the first buckling mode
constant). Following that reasoning, I have focused on the two columns of non-dominated
points near the bisector - see Figure 4.35. From those points, the most optimal ones are
the following two:
1
Batch 1 valid points
0.9 Pareto points
Pareto front
0.8 Axes bisector
Optimal points
0.7

0.6

1/λ 0.5

0.4
117
0.3

0.2 88

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.37: Batch 1 valid points, the Pareto front, and the two selected points to continue
the design process.
The points/cases 88 and 117 are the best options to continue with the design process.
The criterion to choose both points is the following: besides being close to the bisector,

99
Chapter 4. Results

both belong to a column of points, which all have almost the same mass, but they two
are the ones with the best 1/λ in their respective groups.

The trends that have been observed with the Pareto front points are the following:
• 7 points have skin A, 8 points have skin B and 2 points have skin C. Both selected
points, 88 and 117, have skin B. Conclusion: it is way better to use a symmetric
skin layer sequence - see the skins in Section 3.1.5. Materials Employed and its
Properties.

• 4 points have 1 mm core thickness, 4 points have 2 mm core thickness, 4 points have
3 mm core thickness and 5 points have 4 mm core thickness. None Pareto optimal
point has 5 mm core thickness. Also, the two columns of points mentioned before
(8 points, exactly), are the 2 and 3 mm core thickness points. The selected points,
thus, have a core thickness of 2 mm, one, and 3 mm, the other.

• 2 points have the spar positioned at 35 mm from the leading edge, 3 at 53.2 mm, 10 at
65 mm and 2 at 75 mm. This shows a clear trend towards having the spar positioned
towards the 34.21% of the chord (that’s the percentage that 65 mm represents). This
is logical, as the suction peak modelled in Section 3.2.1.3. Load Definition for
option 3 was centred in between the 14% and the 54% of the chord - being the
resultant of the force approximately centred towards the 33% of the chord. The spar
position of the points selected is 53.2 mm (maximum thickness) and 65 mm.

• The region where CE is applied (reinforced region placed at the maximum thickness
point) has had a width of 30 mm in 16 of the 17 points, whereas only 1 point has a
20 mm width. Therefore, for the weight that add 10 mm more of reinforced section,
it is clearly optimal to choose this option.
A summary of the two selected points is detailed in the next table:

Features 88 117
Mass (kg) 0.362 0.3157
1/λ 0.2304 0.3401
Skin B B
Core thickness (mm) 3 2
Spar position (mm) 53.2 65
CE reinforcement width (mm) 30 30

Table 4.6: Main features of the two selected points of Batch 1.

From now one, the improvements explained in Section 3.2.3. Improvements have
been applied to this points first, and then two some of their children - i.e. optimal points
which have been obtained from points 88 and 117.

4.3 Second Batch of Cases - Improvement 1


The improvement 1 detailed in Section 3.2.3. Improvements is now applied to
points 88 and 117. Recall that this improvement consists on changing the spar width
and maintain the rest of the parameters unaltered. The spar widths that have been
tested are 5, 15, 20 and 25 mm.

Being so, 8 new cases have been tested: 4 with point 88 and 4 with point 117.
Applying the decision making procedure, of this 8 cases, 7 are valid - see Annex C.

100
Chapter 4. Results

Numerical Results to consult the parameters of this points.

Next, a plot with the Batch 1 valid points and the valid points of this new Batch 2 are
represented:
1
Batch 1 valid points
0.9 Pareto points
Batch 2 valid points
0.8 Old Pareto front
New Pareto front
0.7 Axes bisector
Optimal points
0.6

1/λ 0.5

0.4

117
0.3

0.2 88

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.38: Batch 1 valid points, old and new Pareto front and Batch 2 valid points.
As it can be seen in Figure 4.38, the bacth 2 adds 4 points to the new Pareto front,
which has been reduced from 17 to 14 points. In other words, 7 old non-dominated points
are now dominated points - i.e. are no longer optimal points.
Finally, the trends that have been observed with the Batch 2 points are the following:
• Increasing the spar width leads to more total mass of the structure. But
increasing the section of the spar allows it to transmit more stresses unloading
the skin, and thus statical failure indices are lower and the first buckling mode
constant increases.
• Decreasing the spar width leads to less total mass of the structure. But
decreasing the section of the spar reduces the ability of the spar to unload the skin,
and thus statical failure indices are higher and the first buckling mode constant
decreases.
As explained in Section 3.2.3. Improvements, a decision of the next points to
optimise from the new Pareto front will not be taken until the 3 improvements and tested
and all children points are compared all together.

4.4 Third Batch of Cases - Improvement 2


The improvement 2 detailed in Section 3.2.3. Improvements is now applied to
points 88 and 117. Recall that this improvement consists on adding a second spar, of the
same width as the main spar, a certain distance away and maintain the rest of the
parameters unaltered. The second spar positions that have been tested are 105, 115,
125 and 135 mm from the leading edge.

Being so, 8 new cases have been tested: 4 with point 88 and 4 with point 117.
Applying the decision making procedure, of this 8 cases, 8 are valid - see Annex C.
Numerical Results to consult the parameters of this points.

101
Chapter 4. Results

Next, a plot with the Batch 1 valid points and the valid points of this new Batch 3 are
represented:
1
Batch 1 valid points
0.9 Pareto points
Batch 3 valid points
0.8 Old Pareto front
New Pareto front
0.7 Axes bisector
Optimal points
0.6

1/λ 0.5

0.4

117
0.3

0.2 88

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.39: Batch 1 valid points, old and new Pareto front and Batch 3 valid points.

As it can be seen in Figure 4.39, the batch 3 adds 8 points to the new Pareto front,
which has been increased from 17 to 24 points. As the 8 points of the Batch 3 belong to
the new Pareto front, 1 old non-dominated point is now a dominated point - i.e. is no
longer an optimal point.

To conclude, the trends that have been observed with the Batch 3 points are the
following:

• Adding a second spar increases the total mass of the structure. It also increases
the section of stress transmission between extrados and intrados, which unloads the
skin. Thus, statical failure indices are lower and the first buckling mode constant
increases.

• Decreasing or increasing the distance of the second spar will result in more or less
mass addition. As the distance is increased, the thickness of the airfoil decreases,
and thus the second spar has less volume -i.e. less mass. Accordingly, the I decreases
and so does the unloading ability of the spar, resulting in higher failure indices and
lower first buckling mode constant than a structure with a second spar closer to the
main spar.

As explained in Section 3.2.3. Improvements, a decision of the next points to


optimise from the new Pareto front will not be taken until the 3 improvements and tested
and all children points are compared all together.

4.5 Fourth Batch of Cases - Improvement 3


The improvement 3 detailed in Section 3.2.3. Improvements is now applied to
points 88 and 117. Recall that this improvement consists on adding stringers of different
widths and maintain the rest of the parameters unaltered.

Points 88 and 177 have 3 and 2 mm of core thickness, respectively. Thus, the core
combinations that are going to be tested on each are:

102
Chapter 4. Results

Features 88 117
Original core thickness (mm) 3 2
Core thickness no-stringer region (mm) 2 3 1 2
Core thickness stringer region (mm) 3/4/5 4/5 2/3/4/5 3/4/5

Table 4.7: Core combinations for the stringers of improvement 3.

Core thickness thicker than the original core thickness have not been considered as the
increase of weight would be very substantial. Besides, the stringer regions are going to
have the same skin composition as the CE reinforced region already applied up to now -
and that adds weight too.

Figure 4.40: Stringers improvement implemented in Bacth 4.

As shown by Figure 4.40, one stringer has been placed before the spar and two behind.
Stringer one is centred at 16.6 mm from he leading edge, stringer two is centred at 96.25
mm from the leading edge, and stringer three is centred at 143.75 mm from the leading
edge. Their width has been varied equally for the two optimal points (88 and 117) - see
Table 4.8.

Stringer configuration S1 w (mm) S2 w (mm) S3 w (mm)


1 10 20 20
2 15 30 30
3 24.2 42.5 42.5

Table 4.8: Stringer configurations for improvement 3 points.

Being so, 15 new cases have been tested for point 88, and 21 new cases for point 117.
Applying the decision making procedure, of this 36 cases, 26 are valid - see Annex C.
Numerical Results to consult the parameters of this points.

Next, a plot with the Batch 1 valid points and the valid points of the Batch 4 are
represented:

103
Chapter 4. Results

1
Batch 1 valid points
Pareto points
0.9
Batch 4 valid points
Old Pareto front
0.8 New Pareto front
Axes bisector
0.7 Optimal points

0.6

1/λ 0.5

0.4
117
0.3

0.2 88

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.41: Batch 1 valid points, old and new Pareto front and Batch 4 valid points.

Figure 4.41 shows the new Pareto front, which has 9 points added by the batch 4 -
the Pareto front has been increased from 17 to 25 points. Subtracting the 9 front points
of the Batch 4, there are 16 points left- that means that 1 old non-dominated point is
now a dominated point.

After carrying out this improvement, I thought that it would be smart to test another
skin sequence at the stringers. As the skin already has the CE reinforced region, I
have removed one laminae of CE in the skin that is applied at the stringers regions.
Besides, I have moved this laminae to gain I. I have done the following: (0/90/0/core)s →
(0/90/core/90/0/0), changing from a symmetric layer sequence to a unsymmetrical one
- note that the bold 0 denotes the CE laminae. This removal of the laminae has lowered
the total weight and as I am going to show you next, given I to the section - as the λ have
increased.
Again, 15 new cases have been tested for point 88, and 21 new cases for point 117.
Applying the decision making procedure, of this 36 cases, 15 are valid - see Annex C.
Numerical Results to consult the parameters of this points. This is logical, the
reduction in CE - a high stiffness material - has increased the maximum stresses and
thus more specimens have failed.

Next, a plot with the Batch 1 valid points and the valid points of this extra points for
Batch 4 are represented:

104
Chapter 4. Results

1
Batch 1 valid points
Pareto points
0.9
Batch 4 extra valid points
Old Pareto front
0.8 New Pareto front
Axes bisector
0.7 Optimal points

0.6

1/λ 0.5

0.4
117
0.3

0.2 88

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.42: Batch 1 valid points, old and new Pareto front and extra (valid) points for
Batch 4.

Figure 4.42 shows the new Pareto front, which has 10 points added by the batch 4 -
the Pareto front has been increased from 17 to 26 points. Subtracting the 10 front points
of the Batch 4, there are 16 points left- that means that 1 old non-dominated point is
now a dominated point.

Now, it is convenient to compare the 26 valid points of the first stringer skin
configuration and the 15 valid points of the second. To do that, a plot with the Batch 1
valid points and the 41 stringer valid points of Batch 4 is represented next:
1
Batch 1 valid points
Pareto points
0.9
Stringer configuration 1
Stringer configuration 2
0.8 Old Pareto front
New Pareto front
0.7 Axes bisector
Optimal points
0.6

1/λ 0.5

0.4
117
0.3

0.2 88

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.43: Batch 1 valid points, old and new Pareto front and final Batch 4 points.

Figure 4.43 shows the new Pareto front, which has 11 points added by the batch 4 -
the Pareto front has been increased from 17 to 27 points. Subtracting the 11 front points
of the Batch 4, there are 16 points left- that means that 1 old non-dominated point is
now a dominated point.

What is important to remark is the following: 10 of the 11 points added to the Pareto
front belong to the second stringer configuration. Even though the valid points of this

105
Chapter 4. Results

second configuration were less than the first - 15 versus 26, they have proven to be more
optimal. See Figure 4.44 to visualise with clarity that almost all new front points belong
to the second stringer configuration.
0.5
Batch 1 valid points
Pareto points
Stringer configuration 1
Stringer configuration 2
Old Pareto front
0.4 New Pareto front
Axes bisector
Optimal points

117

1/λ 0.3

88
0.2

0.1
0.3 0.35 0.4
mass (kg)

Figure 4.44: Close-up of Batch 1 valid points, old and new Pareto front and final Batch 4
points.

Now, the trends observed with Batch 4 points are the following:

• The mass trend is difficult to predict because there are two main influencing
variables: the core combination and the stringer width combination. A general
trend can be described as: for low thickness core combinations and low stringer
widths, the mass gets reduced; for low thickness core combinations and mid-high
stringer widths, the mass can either be reduced or increased; and for high
thickness core combinations and whatever the stringer width, the mass gets
increased.

• Following the same reasoning for the first buckling mode constant: or low thickness
core combinations and low stringer widths, λ gets reduced; for low thickness core
combinations and mid-high stringer widths, λ can either be reduced or increased;
and for high thickness core combinations and whatever the stringer width, λ gets
increased.

Now that the 3 improvements have been tested, it is time to compare the first batch
of points, Batch 1, and the Batches 2 to 4 - created from the improvements - all together.
After that, new optimal points will be extracted and the design will be followed from those
points on.

4.6 Decision after Improvements 1, 2 and 3


To continue with the design process, new points must be selected. Let’s see if new
Pareto optimal points are obtained from the Batches 2 to 4. Next is the representation of
the valid points of Batches 1 to 4, and the new Pareto front:

106
Chapter 4. Results

1
Batch 1 valid points
Pareto points
0.9 Batch 2 valid points
Batch 3 valid points
Stringer configuration 1
0.8
Stringer configuration 2
Old Pareto front
0.7 New Pareto front
Axes bisector
Optimal points
0.6

1/λ 0.5

0.4
117
0.3

0.2 88

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.45: Batches 1 to 4 valid points, and the old and new Pareto front.

As it can be seen in Figure 4.45, new points from Batches 2 and 4 have been added to
the Pareto front. Batch 3, in the other hand, hasn’t add any new non-dominated point.

Following the same reasoning as with Batch 1, the points that I want to be the
new optimal points within the Pareto front are the ones that are closer to the bisector.
Therefore, there are 3 points which are very interesting - 1 from Batch 4 and the other 2
from Batch 2. Then, points 88 and 117 still are non-dominated points and for that reason,
they are optimal points of the design process.
1
Batch 1 valid points
Pareto points
0.9 Batch 2 valid points
Batch 3 valid points
Stringer configuration 1
0.8
Stringer configuration 2
Old Pareto front
0.7 New Pareto front
Axes bisector
Optimal points
0.6

1/λ 0.5

0.4

117
0.3
222
156
0.2 151 88
185

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.46: Batches 1 to 4 valid points, and the old and new Pareto front, and the new
optimal points.

The new optimal points are the 151, 156 and 222, plus the already known 88 and 117
- see Figure 4.46. It is important to remark that point 151 and 156 come from the
improvement 1, and one has a spar width of 5 mm - 151 - and the other of 15 mm -
156. Point 151 is a child of point 88 and point 156 is a child of 117.

Regarding point 222, it comes from improvement 3 and its configuration is the
following: stringer 2 mm of core in the no-stringer region, 3 mm core in the stringer

107
Chapter 4. Results

region, stringer width configuration 1 and stringer skin configuration 2. Point 222 is a
child of point 117. The point 88 has also a child that has made it to the Pareto front,
point 185 (remarked in Figure 4.46), but I’m no going to consider it as it is very far from
the bisector. Even though, point 185 is useful because it allows me to see an optimal
pattern with the stringer improvement. Point 185 features are the following: stringer 3
mm of core in the no-stringer region, 4 mm core in the stringer region, stringer width
configuration 1 and stringer skin configuration 2.

Stringer improvement is optimal if the original core thickness is kept, if the stringer
core thickness is 1 mm higher than the original, if the stringer width configuration is
used (the one with lower widths) and the skin configuration 2 (the one with less CE
laminae in each stringer). This is the way to optimise the mass by getting a good
buckling resistance behaviour. This conclusion is going to be useful for next Section 4.7.
Fifth Batch of Cases - Improvement 4.

Finally, Table 4.9 is attached with the main features of the 5 optimal points: 88, 117,
151, 156 and 222.

Features 88 117 151 156 222


mass (kg) 0.3620 0.3157 0.3408 0.3371 0.3321
1/λ 0.2304 0.3401 0.2447 0.2808 0.2954
Skin B B B B B
Core thickness (mm) 3 2 3 2 2
Core thickness stringer region (mm) - - - - 3
Stringer skin configuration - - - - 2
Main spar position (mm) 53.2 65 53.2 65 65
Main spar width (mm) 10 10 5 15 10
CE reinforcement width (mm) 30 30 30 30 30

Table 4.9: Main features of the five selected points of Batches 1 to 4.

4.7 Fifth Batch of Cases - Improvement 4


As explained in Sectiom 3.2.3. Improvements, after running the cases belonging to
the first three improvements and obtaining the optimal points, a set of tests is going to
be done in which the different improvements are going to be tested together - i.e.
combination of the improvements.

From the last step, I have obtained 3 new optimal points - two belonging to Batch 2
and one to Batch 4. Being so, the 2 points of Batch 2 - 151 and 156 - are going to be
tested with the improvement 3 features, and the point of Batch 4 - 222 - is going to be
tested with the improvement 1 features.

Points 151 and 156 are going to have stringers now. But instead of doing the whole
set of combinations done with points 88 and 117, I have just applied the parameters
configuration that have been the best for the latter points - as by doing this I have
obtained two non-dominated points: 222 and 185. So, as point 151 is a child of point 88,
the stringer features that will be applied to it will be the ones of point 185. And
regarding point 156, as is a child of point 117, the stringer features that will be applied
to it will be the ones of point 222.

108
Chapter 4. Results

And point 222 feature that is going to be changed is the spar width, as this was what
it was done in improvement 1. As seen with points 151 and 156, the best spar widths to
try are 5 and 15 mm.

That adds up to 4 new cases. Applying the decision making procedure, of this 4
cases, 3 are valid - see Annex C. Numerical Results to consult the parameters of this
points.

Next, a plot with the Batch 1 to 4 valid points and the valid points of this new Batch
5 are represented:
1
Batch 1 valid points
Pareto points
0.9 Batch 2 valid points
Batch 3 valid points
0.8 Stringer configuration 1
Stringer configuration 2
Batch 5 valid points
0.7 Old Pareto front
New Pareto front
0.6 Axes bisector
Optimal points

1/λ 0.5

0.4
117
0.3 222
156
0.2 151 88 185

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.47: Batches 1 to 4 valid points, old and new Pareto front and Batch 5 valid
points.

As it can be seen in Figure 4.47, only one point is now part of the new Pareto front.
That point is a child of point 151. No point has been removed from the old Pareto front.
This new point is now a new optimal point, to which improvement 5 will be performed.
1
Batch 1 valid points
Pareto points
0.9 Batch 2 valid points
Batch 3 valid points
Stringer configuration 1
0.8 Stringer configuration 2
Batch 5 valid points
0.7 Old Pareto front
New Pareto front
Axes bisector
0.6 Optimal points

1/λ 0.5

0.4
117
0.3
222
156
0.2 151 239 88

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.48: Batches 1 to 4 valid points, old and new Pareto front, Batch 5 valid points
and the 6 optimal points.

109
Chapter 4. Results

This points is number 239. Note that points 88, 117, 151, 156 and 222 still are
non-dominated points and thus can still be taken as optimal points. Now,
improvement 5 will be performed in all 6 points.

Prior to that, let’s actualise Table 4.9 with the new point 239:

Features 88 117 151 156 222 239


mass (kg) 0.3620 0.3157 0.3408 0.3371 0.3321 0.3535
1/λ 0.2304 0.3401 0.2447 0.2808 0.2954 0.2316
Skin B B B B B B
Core thickness (mm) 3 2 3 2 2 3
Core thickness stringer region (mm) - - - - 3 4
Stringer skin configuration - - - - 2 2
Main spar position (mm) 53.2 65 53.2 65 65 53.2
Main spar width (mm) 10 10 5 15 10 5
CE reinforcement width (mm) 30 30 30 30 30 30

Table 4.10: Main features of the six selected points of Batches 1 to 5.

4.8 Sixth Batch of Cases - Improvement 5


For last, the improvement 5 explained in Section 3.2.3. Improvements is going
to be implemented in the 6 optimal points found so far: 88, 117, 151, 156, 222 and 239.

This improvement consists on adding ribs to the wing. 3 tests are going to be performed
on each optimal point:

• 3 ribs per semi-wing, equally spaced.

• 4 ribs per semi-wing, equally spaced.

• 5 ribs per semi-wing, equally spaced.

The rib thickness is 2 mm and as its mass without emptying it just adds up to 1.15 g,
hollowing it has not been considered.

The amount of new cases tested in this final step have been: 6 x 3 = 18 new cases.
Of this 18 cases, 2 are not valid - see Annex C. Numerical Results to consult the
parameters of this points.

Now, a plot with all the valid points seen up to now is going to be presented:

110
Chapter 4. Results

1 Batch 1 valid points


Pareto points
0.9 Batch 2 valid points
Batch 3 valid points
Stringer configuration 1
0.8
Stringer configuration 2
Batch 5 valid points
0.7 Batch 6 valid points
Old Pareto front
0.6 New Pareto front
Axes bisector
Optimal points
1/λ 0.5

0.4
117
0.3
222
156
0.2 151

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

Figure 4.49: Batches 1 to 5 valid points, old and new Pareto front and Batch 6 valid
points.

First thing that stands out is that 2 of the optimal points I had so far do not belong
to the Pareto front any more. Those points are 239 and 88. On the other hand, Batch 6
points add 4 points to the final Pareto front. This 4 points are points number 243, 246,
249 and 255. All 4 points have one common feature: all have 3 ribs per semi-wing.

The trend of this last improvement is the following: adding more number of ribs
increases the first buckling mode constant as it reduces the buckling length.
Likewise, adding more number of ribs increases the total mass of the structure.

From the 3 configurations tested - 3, 4 and 5 ribs equally spaced per semi-wing - the
best option seems to be 3 ribs per semi-wing, as all points of Batch 6 in the final Pareto
front have this configuration.

Of this 4 points, the less appealing one is 243, as despite it is a non-dominated point,
its distance to the bisector is too much. For that reason, the final optimal points selected
by myself from the final Pareto front are the following: 117, 151, 156, 222 - both 4 have
maintained themselves non-dominated - and 246, 249 and 255 (see Figure 4.50).
1
Batch 1 valid points
Pareto points
0.9 Batch 2 valid points
Batch 3 valid points
0.8 Stringer configuration 1
Stringer configuration 2
Batch 5 valid points
0.7 Batch 6 valid points
Old Pareto front
0.6 New Pareto front
Axes bisector
1/λ Optimal points
0.5

0.4
117 255
0.3 246 222
156
0.2 151
249

0.1

0
0.25 0.3 0.35 0.4 0.45
mass (kg)

111
Chapter 4. Results

Figure 4.50: Batches 1 to 6 valid points, old and new Pareto front and final optimal points.

0.5
Batch 1 valid points
Pareto points
Batch 2 valid points
Batch 3 valid points
Stringer configuration 1
Stringer configuration 2
0.4
Batch 5 valid points
Batch 6 valid points
Old Pareto front
117 New Pareto front
Axes bisector
246 Optimal points
1/λ 0.3
222
156 255

151
249
0.2

0.1
0.3 0.35 0.4
mass (kg)

Figure 4.51: Close-up of the final optimal points.

Finally, Table 4.11 includes the main features of the 7 final optimal points:

Features 117 151 156 222 246 249 255


mass (kg) 0.3157 0.3408 0.3371 0.3321 0.3228 0.3483 0.3392
1/λ 0.3401 0.2447 0.2808 0.2954 0.3215 0.2272 0.2793
Skin B B B B B B B
Core thickness (mm) 2 3 2 2 2 3 2
Core thickness stringer region (mm) - - - 3 - - 3
Stringer skin configuration - - - 2 - - 2
Main spar position (mm) 65 53.2 65 65 65 53.2 65
Main spar width (mm) 10 5 15 10 10 5 10
CE reinforcement width (mm) 30 30 30 30 30 30 30
Ribs per semi-wing - - - - 3 3 3

Table 4.11: Main features of the final optimal points.

4.9 Final Results


From a pure optimisation point of view, all the points from the final Pareto front
represent a structure which is Pareto optimal - as there is no other point which is better
in both objective functions at the same time. From all the points, there are some that
allow you to have a great low mass but a low λ; and there are other points which allow
you to have a high λ but a high mass instead - see Table 4.12.

In terms of minimisation of the two objective functions at the same time, the points
that are closer to the bisector of the axes, are the ones that really do best in the two
fields. For that reason, the points selected during the design process have been the ones
closer to that area. Following that reasoning, the best appealing points to be the final
structure of the wing are: 117, 246 and 222 - see Figure 4.51.

112
Chapter 4. Results

Of those 3 points, the best Point mass (kg) 1/λ


one in optimisation terms is point 246, as 131 0.2691 0.7077
it almost lays in the bisector. It is the point 106 0.2697 0.6897
that best minimises both objective functions. 111 0.2697 0.6667
116 0.2697 0.5952
Besides, point 246 and 117 198 0.3026 0.5732
have another key feature that I think it is 221 0.3067 0.5668
a great advantage over point 222: simplicity 132 0.3150 0.4705
of construction. Recall from Table 4.11 that 107 0.3157 0.3846
this two points do not have stringers. Thus, 112 0.3157 0.3717
the building process of such a structure 117 0.3157 0.3401
is easier as stringers incorporate geometrical 246 0.3229 0.3215
features that are difficult to tackle. 222 0.3321 0.2955
156 0.3371 0.2809
Taking a closer 255 0.3393 0.2793
look at the building point of view, point 151 0.3408 0.2447
117 is really the most desirable one, as unlike 249 0.3483 0.2272
246, doesn’t have ribs. Thus, it eliminates 243 0.3692 0.2186
a tedious building part, as the ribs have 185 0.3784 0.2071
to fit the extrados and intrados shape. The 158 0.3790 0.1580
latter are not going to have the theoretical 154 0.4256 0.1553
shape after the build of the skin, and the ribs
will have to adapt to its shape by a manual Table 4.12: Final Pareto front points.
process - e.g. progressive sanding until it fits.

In conclusion, point 246 is the best point from an optimisation point of view. And
from a building point of view, point 117 is the best one as it is the simplest one to
construct. And leaving aside those points of view, points 246, 117 and 222 are the best
points of the total mount of 260 points tested for this design process. Tabel 4.13 sums up
the features of this 3 points.

Features 117 222 246


mass (kg) 0.3157 0.3321 0.3228
1/λ 0.3401 0.2954 0.3215
Skin B B B
Core thickness (mm) 2 2 2
Core thickness stringer region (mm) - 3 -
Stringer skin configuration - 2 -
Main spar position (mm) 65 65 65
Main spar width (mm) 10 10 10
CE reinforcement width (mm) 30 30 30
Ribs per semi-wing - - 3

Table 4.13: Main features of the points 117, 222 and 246.

113
Chapter 5

Conclusions

In this chapter, some conclusions are extracted to put an end to this thesis. The
conclusions that have been obtained are of different nature. These are:

• Conclusions about the results obtained.

• Conclusions about the numerical model employed in AbaqusTM .

• Conclusions about the design procedure followed.

• A project overview about the organisation, goals achieved, and self-valuation of


the thesis.

Next, each of this sections are going to be covered.

5.1 Conclusions about the Results


First of all, the results obtained mainly help me see the trends and effects that every
parameter of design has on the structural behaviour. These results are not a definite
optimised solution to the monocoque structure made with composite materials. They
just represent a possible iteration from a starting point of parameter values. To obtain
such final solution, a full iteration loop over a wider range of parameters and with
different starting points should be done.

Being so, the conclusions I want to subtract from the results are the trends that each
parameter has shown in modifying the structural behaviour. The trends are:

• Skin: of the 3 types of skin proposed for this thesis - which are actually 6 depending
on the thickness of the core - I have concluded the following: first, sandwich panels
(with non-zero core thickness) are better than laminates (zero core thickness) for the
expected performance of the wing; secondly, skins with layers stacked symmetrically
are way better than if the sequence is unsymmetrical. This was predicted in Section
2.3.3.2. Laminates and proved in Section 4.2. First Batch of Cases - Starting
Point.

• Core thickness: as commented with the skin, sandwich panels have proven to be
the best option for the wing and for that reason, zero core thickness is not an option
- a high section second moment of inertia I is necessary to have a good behaviour
against the bending moment. Then, regarding the resting 5 options I have tested,
the optimum is found with 2/3 mm core thickness, feature that have had the most
promising points tested throughout the whole design process. Despite that, there are
also Pareto optimal points from the final Pareto front with the 5 core thicknesses.

114
Chapter 5. Conclusions

So, it just depends on how much you want to strengthen one objective function or
the other.

• Spar width: different spar widths have been tested. Increasing the spar width
leads to an increment of the total mass but an unloading of the skin stresses. The
latter results in higher resistance to statical and buckling failure. In the other hand,
reducing the spar width leads to a lower total mass but has the contrary effect with
statical and buckling failure. The final selected points have all 10 mm spar width,
but within the final Pareto front there are also cases with 5 and 15 mm spar widths.

• Spar position: 5 positions have been tested in the design process. Of this 5, two
of them have proven to be the best of them: spar placed at the maximum thickness
point of the airfoil (53.2 mm) and placed at 65 mm from the leading edge. The
front-chord positions tested have had a lack of I, as there the thickness is reduced,
and thus the skin has suffered more both statically and by buckling. Although
53.2 and 65 mm options have add more mass, as the spar has more height in those
positions, the I gained has compensated with the good structural behaviour shown.
Plus, as the airfoil suction peak has been modelled between the 14% and 54% of
the chord - to simulate the resultant lift force at about a 33% of the chord, those
two configurations are more optimal to transmit the stresses more directly without
having to deal with a noticeable torsion moment.

• Second spar: adding a second spar is an option that adds mass but improves the
statical and buckling behaviour of the structure - it adds more surface of stress
transmission, thus unloading the skin. Then, the position of this second spar adds
more or less mass depending on how far from the main spar it is placed: the further it
is placed, the less mass it adds, and vice versa. Accordingly, the further it is placed,
the less it improves the structural behaviour of the structure, and vice versa. I have
nothing more to say about the second spar, as it has not thrived in the optimisation
process.

• Stringers: 3 widths of stringers and different core combinations have been tested in
the design process - see Section 4.5. Fourth Batch of Cases - Improvement 3.
Also two stringers skin combinations have been tested. The mass trend is difficult
to predict because there are two main influencing variables: the core combination
and the stringer width combination. A general trend can be described as: for low
thickness core combinations and low stringer widths, the mass gets reduced; for low
thickness core combinations and mid-high stringer widths, the mass can either be
reduced or increased; and for high thickness core combinations and whatever the
stringer width, the mass gets increased. If the skin with more CE is used, it will add
more mass than the other. Following the same reasoning for the structural behaviour
of the structure: for low thickness core combinations and low stringer widths, the
structural behaviour worsens; for low thickness core combinations and mid-high
stringer widths, the structural behaviour can either worsen or get better; and for
high thickness core combinations and whatever the stringer width, the structural
behaviour get better. The optimal combinations, though, have been the ones in
which the stringer has had the following combination of parameters:

– Original core thickness in the no-stringer region.


– Stringer core thickness 1 mm higher than the original.
– Width configuration 1: the one with the less width (10, 20 and 20 mm).
– Skin configuration 2 at the stringers: less CE but similar I.

115
Chapter 5. Conclusions

• Ribs: the use of ribs have proven that the structural behaviour against buckling
improves. That is due to the reduction of the buckling length. Despite that, adding
ribs also adds to the total structure mass. 3 possible combinations have been tested
- see Section 4.8. Sixth Batch of Cases - Improvement 5. From those 3, the
one that has turned to be the most optimal one has been the 3 ribs per semi-wing
configuration - it has add 4 non-dominated points to the final Pareto front.

Finally, I want to extract some conclusions about the failure criteria employed in this
thesis. Regardless of the values obtained for each criterion indices (as numerical parameters
influence the latter), I want to state the following: Hashin indices have always been lower
than Tsai-Hill ones, which is completely logical: the latter criterion takes into account the
influence of different type of stresses - σ11 , σ22 and τ12 - at the same time. The Hashin
criterion, conversely, doesn’t consider the influence of different type of stresses.

5.2 Conclusions about the Numerical Model


Regarding the numerical model made with AbaqusTM , I feel that the way I have
approached it is a good one - see Section 3.3.1. Pre-processing - but I want to state
the following: FEA is a complicated method of solving an structural problem such as
mine, and for that reason AbaqusTM has a very long set of possibilities (different solvers,
mesh types, material types, etc) that I haven’t had time to master. Being so, I cannot be
sure that there is not a better way of tackling the thesis problem - i.e. modelling the
geometry differently, modelling and assignation of materials, meshing of parts,
interaction properties between parts, etc.

I wanted to state the latter as I feel that, as remarked in previous Section 5.1.
Conclusions about the Results, the numerical results obtained cannot be trusted
with a 100% sureness. For instance, with the ribs models I have noticed a strange
behaviour in some points, of which I don’t understand its reasons. Despite re-doing the
model, it has kept happening. That behaviour is the following: even though the case to
which ribs were added was statically correct (IF < 1), when adding the ribs, this case
now failed statically. As perfect bonding between rib and skin and rib and spar is
simulated, I don’t understand why the ribs can create the stresses to rise, when they
should unload the skin.

Accordingly, other rare event is the fact that out 150 starting cases, 102 failed
statically. That is mainly due to the boundary conditions effect - singularities (see
Section 3.2.1.2. Boundary Conditions), which cannot be completely solved as it is a
price that must be paid when using FEA software. Maybe, the most important thing
about this, is not how to prevent it from happening, but to know how to treat it - see
next Section 5.3. Conclusions about the Design Process.

To conclude, I’m satisfied with the numerical model but I am sure it can be improved
as some flaws have shown to exist.

5.3 Conclusions about the Design Process


Regarding the design process, I have learned a precious lesson. When addressing
multi-objective and multi-parametric problem as the one in this thesis, the preparation of
the design process so it is as automatised as possible is crucial. If the latter is achieved,
then more tests can be run - more parameters can be tested and the range of the design

116
Chapter 5. Conclusions

gets wider - and less time (and patience/energy) is lost getting models ready to run.

One of the main obstacles has been the modelling of the new geometry for every case
to pass it to AbaqusTM . Not for the complexity of that process - which was
quasi-automatised in CATIA V5R20 - but for the time lost in the making of every
model. As I haven’t been able to parametrize the geometry in AbaqusTM , the latter has
had to come from outside the program. AbaqusTM models can be entirely made by a
script: the geometry, materials definition and assignation, mesh, loads and boundary
conditions definition, etc. But I haven’t had enough time to learn how to do so, as
learning how to use the program itself has already been complex.

If I had been able to achieve such a modelling (geometry and numerical model)
technique, the number of iterations over the parameters could have been way larger than
it has been. Throughout the design process I have picked the most promising points
after each step. The reasoning employed to do so is completely correct but if I had had
this means implemented, I could have followed the design process after each step with all
the points in the Pareto front instead of just a few, as I have done.

Also, more parameters could have been tested. Instead of limiting the spar positions,
or the types of skin, or even the material properties, a wider range of values could have
been tested for every parameter. But to do so the process has to be automatised or else
you will need three times (or more) the time I have employed to accomplish this thesis.

In other words, I am satisfied with the design process I have followed: do some initial
tests to be able to define a not-so-imprecise starting point; define a set of simple cases as
the starting point, and from those cases select the best points and keep adding
improvements to see how they change. But the process has had to be delimited due to
the time expense of every model-making and running of the latter with the tools I have
been able to develop.

Lastly, I want to comment something about the post-processing of the output variables.
As occurred to me when runnign the last cases for the thesis, the failure indices could have
been collected at a certain distance from the encastre to avoid the singularities that are
always generated in a FEA software. Therefore, following the St. Venant’s Principle
the results obtained regarding statical failure would have been more logical - composite
materials are high stiffness materials which are hard to break statically and low thickness
composite-made structures almost always fail by buckling, not statically.

5.4 Project Overview


Overall, I am satisfied with the output of the thesis. The scope and objectives set at
the beginning of the project have been fully achieved. The main issue I have had is time.
Getting to know what I needed to know to be able to jump onto the next step has
always been slower than I expected. For that reason, I have needed to employ more
hours than the 600 that I first thought I would need.

Finally, just say that this thesis has encouraged me to keep up with my education in
the aerospace engineering field, as it has been very useful to make me see what I already
know and that I have still a lot new things to know.

117
Chapter 6

Future Work

The future work for this thesis is the automation of the whole design process.
Doing so, a wide range of parameters could be tested and there would be no need for
delimitation of the study.

AbaqusTM has the option of creating a parametric study with a script. In the
AbaqusTM Scripting User’s Guide there are some examples and some guidelines on how
to create such a script. In order to develop such a designing tool, it is necessary to know
Python language of programming, the basic commands to communicate with AbaqusTM ,
and as the problem of this thesis is a multi-objective one, also knowledge of
optimisation algorithms.

A parametric study allows you to create just one time the model and then in the
script you assign the variables that are going to have different parameters. With a loop
iteration, each parameter is assigned and the case is run each time a change is performed
- for each case the geometry, assembly, mesh, loads, and others, are actualised.

Being able to automatise that would save a lot of time as the main loss of time I have
suffered in this thesis has been in the making of the models and jobs definition. But in
addition to that, an optimisation algorithm is necessary to be employed. Otherwise, the
only part that would be automatised of the process would be the analysis process - from
the making of the model to the execution of the job.

With an optimisation algorithm, the decision making procedure - that has been
performed after each set of new cases has been obtained - would be included in the
script. By doing so, once the script is ready, the whole design process can be run without
interruptions. To be able to do so, some optimisation algorithms - that can make the
appropriate decisions based on the objective functions defined - are necessary. The most
well-known optimisation algorithms are the genetic algorithms, which generate solutions
using techniques inspired by natural evolution, such as inheritance, mutation, selection
and crossover.

118
Chapter 7

Environmental Impact

As a part of every project, the environmental impact that this thesis has must be
covered. Next I will explain the advantages and disadvantages that the making of this
thesis has in the environment.

• Advantages

– Achievement of a multi-objective and multi-parametric design faster. This can


eventually increase the development of more composite-made structures, as
the biggest brake for this structures against conventional-material-made ones
is the complexity of design and thus, time required. Finally, if more composite-
made structures are build for aircraft wings, or any other moving vehicle, this
will result in more fuel efficient and less polluting vehicles - as this type of
structure has proven to be more lightweight than conventional structures.
– Saving of raw materials. Thanks to the optimisation process carried out in
the thesis, the results obtained reduce a lot the possibilities of such a
multi-parametric problem. This allows less need of prototyping and,
consequently, less tests required - with all it entails: test-benches, special
tooling and qualified personnel. All this means a great money saving. And
the environmental advantage is:
∗ Reduction of wastes: as less materials are required, the production of
the latter will be lower as well. That implies that the wastes generated
throughout the process are also cut down. [27] has a very interesting table
that sums up which are the wastes (and the effects of those) generated in
composite materials production, which now can be reduced - see Table 7.1.
– Savings of paper: as the computer is the main working tool for this thesis -
and in nowadays society, paper has not barely used.

• Disadvantages

– Energy consumption:
∗ Electricity: the main tool of design is the computer. Throughout the
time spent to do this thesis, the latter has needed power to be functioning.
Also, it must be taken into account that the working place to do the thesis
needs eletricity to be suitable: air-conditioning, lights or other office/indoor
devices also need electricity. This electricity must be generated by either
hydroelectric or nuclear power station or others means like fossil fuels. The
point is that the production of that electricity pollutes the environment by
emitting hazardous emissions to the air, contaminating the water of rivers
or adulterating the food that we eat as a result of the first two.

119
Chapter 7. Environmental Impact

∗ Natural gas: in regard with the needs of the place of development of the
thesis, some heating infrastructure is needed in any housing, and most of
this heating systems work on natural gas. As explained with electricity,
natural gas production also harms the environment.

Table 7.1: Typical composites fabrication processes, products used and wastes. Retrieved
from [27].

120
Bibliography

[1] “Lancair IV.” [Online]. Available: http://www.aerodynamicsllc.com/planes we


build.html

[2] “Chanute’s biplane glider.” [Online]. Available: http://media-2.web.britannica.com/


eb-media/13/99713-004-3D59FB9D.jpg

[3] “The Wright flyer.” [Online]. Available: http://robustdesignconcepts.com/files/


spaceshipone/www.nasm.si.edu/exhibitions/gal100/wright flight.jpg

[4] “Bleriot’s airplane.” [Online]. Available: http://i.dailymail.co.uk/i/pix/2009/07/17/


article-1200178-03125B93000005DC-448 634x286.jpg

[5] “The Deperdussin.” [Online]. Available: http://www.ctie.monash.edu.au/hargrave/


images/dep gb provost 041013 500.jpg

[6] “The Junkers J1.” [Online]. Available: http://www.junkers.de/sites/default/files/


media/J1 THF 024 1175 2.jpg

[7] “The Boeing 707.” [Online]. Available: http://cdn-www.airliners.net/


aviation-photos/photos/1/7/4/1948471.jpg

[8] “X-29.” [Online]. Available: https://www.nasa.gov/sites/default/files/images/


299396main EC87-0182-14 full.jpg

[9] “Materials employed in the Boeing 787.” [Online]. Available: http://www.1001crash.


com/dossier/composite/B787 composite.jpg

[10] “Truss structure.” [Online]. Available: hhttp://legendaryaircraft.hellobit.hu/gallery/


yak-3-2/fuselage frame 1.jpg

[11] “Monocoque vs. Semi-monocoque.” [Online]. Available: http://source-http//www.


aero-mechanic.com/wp-content/uploads/2011/03/2-14.gif

[12] D. Gray, Composite Materials: Design and Applications. CRC Press, 2014.

[13] W. Callister and D. Rethwisch, Materials science and engineering: an introduction.


Wiley, 2007.

[14] D. Roylance, “Laminated composite plates,” 2000.

[15] G. Zagainov and G. Lozino-Lozinski, Composite Materials in Aerospace Design.


Chapman & Hall, 1996.

[16] “Tractions and moments in a plate.” [Online].


Available: http://www.mathworks.com/matlabcentral/mlc-downloads/downloads/
submissions/44717/versions/6/screenshot.jpg

121
Chapter 7. Environmental Impact

[17] J. N. Reddy, Mechanics of Laminated Composite Plates - Theory and Analysis. CRC
Press, 1997.

[18] “Force-displacement bifurcation diagram.” [Online]. Available: http://www.strand7.


com/images/equations/image19.gif

[19] Q. J. Yang, “Simplified Approaches to Buckling of Composite Plates,” Master’s thesis,


2009.

[20] E. J. Barbero, Finite Element Analysis of Composite Materials. CRC Press, 2007.

[21] “Abaqus 6.14 user manual,” DS Simulia, 2014.

[22] “Pareto optimality example.” [Online]. Available: http://www-cdr.stanford.edu/


NextLink/images/pareto.gif

[23] “Pareto front.” [Online]. Available: https://upload.wikimedia.org/wikipedia/


commons/thumb/b/b7/Front pareto.svg/640px-Front pareto.svg.png

[24] “Cantilever beam.” [Online]. Available: http://www.codecogs.com/users/23287/


Cantilever-Beams-101.png

[25] “Pressure distribution over an airofil.” [Online]. Available: http://www.efm.leeds.ac.


uk/CIVE/CIVE1400/Section4/img00048.gif

[26] B. D. Agarwal, L. J. Broutman, and K. Chandrashekara, Analysis and Performance


of Fiber Composites. Wiley, 2006.

[27] “A Guide for Composites Fabrication,” Washington State Department of Ecology,


2004.

[28] T. A. Weisshaar, Aerospace Structures - an Introduction to Fundamental Problems,


2011.

[29] S. Krenk and J. Høgsberg, Statics and Mechanics of Structures. Springer


Science+Business Media, 2013.

[30] M. C.-Y. Niu, Airframe Structural Design: Practical Design Information and Data
on Aircraft Structures. Conmilit Press, 1988.

[31] T. H. G. Megson, Aircraft Structures for engineering students, 2007.

[32] “Introduction to Classical Laminated Plate Theory,” MHRD, Govt. of India.

[33] C. R. Steele and C. D. Balch, Introduction to the Theory of Plates.

[34] J. N. Reddy, Energy and Variational Methods in Applied Mechanics. Wiley, 1984.

[35] S. Sádaba, “High fidelity simulations of failure in fiber-reinforced composites,” Ph.D.


dissertation, Universidad Politécnica de Madrid, 2014.

[36] “Hashin Criterion,” AUTODESK. [Online]. Available:


http://help.autodesk.com/view/ACMPAN/2016/ENU/?guid=
GUID-A2597BFC-606B-4C21-ABCA-8F34D37FBC41

[37] “Tsai-Hill Criterion,” AUTODESK. [Online]. Available:


http://help.autodesk.com/view/ACMPAN/2016/ENU/?guid=
GUID-01A4F20B-09D0-4DED-91F7-967277BEDBD6

122
Chapter 7. Environmental Impact

[38] “Pareto Efficiency.” [Online]. Available: https://en.wikipedia.org/wiki/Pareto


efficiency

[39] “Balsa properties.” [Online]. Available: http://www.makeitfrom.com/


material-properties/Balsa/

[40] “Rohacell properties.” [Online]. Available: http://www.rohacell.com/sites/lists/


PP-HP/Documents/ROHACELL-WF-mechanical-properties-EN.pdf

[41] E. J. Barbero, F. A. Cosso, R. Roman, and Weadon, “Determination of Material


Parameters for Abaqus Progressive Damage Analysis of E-Glass Epoxy Laminates,”
2013.

[42] P. B. Bogert, A. Satyanarayana, and P. B. Chunchu, “Comparison of Damage


Path Predictions for Composite Laminates by Explicit and Standard Finite Element
Analysis Tools.”

[43] “Composite calculator.” [Online]. Available: http://www.hexcel.com:82/calculators/


src/CompositeProps2WithBleed.aspx

[44] “Junior engineer salary.” [Online]. Available: http://www.payscale.com/research/


US/Job=Junior Mechanical Engineer/Salary

123

You might also like