Passive Aeroelastic Control in Truss-Braced Wings Using Vibration Suppression
Passive Aeroelastic Control in Truss-Braced Wings Using Vibration Suppression
Passive Aeroelastic Control in Truss-Braced Wings Using Vibration Suppression
Author:
Szczyglowski, Christopher Patrick
Title:
Passive Aeroelastic Control In Truss-Braced Wings Using Vibration Suppression
General rights
Access to the thesis is subject to the Creative Commons Attribution - NonCommercial-No Derivatives 4.0 International Public License. A
copy of this may be found at https://creativecommons.org/licenses/by-nc-nd/4.0/legalcode This license sets out your rights and the
restrictions that apply to your access to the thesis so it is important you read this before proceeding.
Take down policy
Some pages of this thesis may have been removed for copyright restrictions prior to having it been deposited in Explore Bristol Research.
However, if you have discovered material within the thesis that you consider to be unlawful e.g. breaches of copyright (either yours or that of
a third party) or any other law, including but not limited to those relating to patent, trademark, confidentiality, data protection, obscenity,
defamation, libel, then please contact [email protected] and include the following information in your message:
Your claim will be investigated and, where appropriate, the item in question will be removed from public view as soon as possible.
Passive Aeroelastic Control In
Truss-Braced Wings Using
Vibration Suppression
By
S EPTEMBER 2019
In recent years a significant effort has been devoted to the study of more energy efficient aircraft
that will meet the environmental goals set out in initiatives such as Vision 2020 and Flight Path
2050. Some of these studies have considered the implementation of new aircraft concepts that can
provide enhanced performance in terms of overall aircraft efficiency and noise. One such concept
is the truss-braced wing aircraft, which has been shown to provide an overall benefit to aircraft
mass and fuel-burn by virtue of its high-aspect ratio wing and efficient structural design. However
aeroelastic phenomena such as flutter and gust loads place a limit on the practical efficiency
savings that can be achieved by adopting a braced-wing design, therefore any mechanism by
which these negative aeroelastic effects can be mitigated will be a key enabler for the success
of this concept. This thesis investigates the possibility of achieving gust loads alleviation and
flutter suppression in a truss-braced wing via passive vibration control. A full-scale aircraft
model based on the NASA/Boeing SUGAR concept aircraft is used to run a series of studies
where a vibration suppression device is included in the truss structure and the device properties
are optimised in order to suppress flutter and minimise gust loads. Different device layouts
are considered including devices that can be frequency-tuned to target specific modes of the
structure. The results for the SUGAR-inspired model demonstrate that improvements in flutter
speed between 1 - 6% and reductions in gust loads of approximately 4% are achievable with an
almost negligible mass penalty from the device. It is also noted that further benefits could be
realised if the design of the device was included as part of a wider structural optimisation. Finally,
the methods used in this thesis can be used to model a generic vibration absorber attached to
any generic finite element model, fundamentally enhancing the scope for vibration suppression
devices to be considered in the design and optimisation of large and complex systems.
i
D EDICATION AND A CKNOWLEDGEMENTS
There are many people I would like to thank for their support over the years: Firstly, I would
like to thank my supervisors Professor Simon Neild, Dr Branislav Titurus and Dr Jason Zheng
Jiang for their guidance throughout my PhD. Despite my apparent desire to do everything but
my PhD work I always knew I had your support and without your valuable insights and patience
this work would not have been possible. I would also like to thank all of the researchers and
PhD students from the Agile Wing Integration Project for the countless coffee breaks and pizza
socials that were instrumental in getting me through the process. I would also like to thank
the UK Aerospace Technology Institute for funding my PhD and Professor Jonathan Cooper for
managing the AWI-related research at the University of Bristol. I would especially like to thank
Dr Etienne Coetzee for all of his advice, for his time spent reviewing my publications and for
arranging my industrial visits; I would not have got nearly as much out of this PhD without your
input, so thank you.
Finally, to my friends and family...I can’t begin to describe how instrumental you all were
in getting me over the line. Whether it was pot-luck dinner evenings, weekends away in the
mountains or a post-conference road-trip in America, you all helped to make the experience as
enjoyable as possible. Thank you.
"Nothing in this world is worth having or worth doing unless it means effort, pain, difficulty....I
have never in my life envied a human being who led an easy life; I have envied a great many
people who led difficult lives and led them well."
— Theodore Roosevelt, "American ideals: And other essays, social and political", 1903
iii
A UTHOR ’ S DECLARATION
declare that the work in this dissertation was carried out in accordance with the
v
TABLE OF C ONTENTS
Page
List of Tables xi
1 Introduction 1
1.1 Research Motivation and Themes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Novel Contribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Publications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.4 Thesis Outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
vii
TABLE OF CONTENTS
viii
TABLE OF CONTENTS
ix
TABLE OF CONTENTS
Bibliography 223
x
L IST OF TABLES
TABLE Page
4.1 Parameter values for the spring, damper and inerter elements used in the BUG-
T flutter study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
4.2 Design variable bounds for the flutter optimisation . . . . . . . . . . . . . . . . . . . . . 103
4.3 Device parameter values for the single absorber optimisation . . . . . . . . . . . . . . . 106
4.4 Device parameter values for the multiple absorber optimisation . . . . . . . . . . . . . 112
5.1 Design variable upper and lower bounds for the gust load alleviation optimisation study128
5.2 Cost function values and device parameters for the gust load alleviation optimisation
with a single damper at locations A, B or C. . . . . . . . . . . . . . . . . . . . . . . . . . . 130
5.3 Maximum damper force as a function of gust gradient . . . . . . . . . . . . . . . . . . . 132
5.4 Cost function and device parameters for a selection of single device configurations . . 134
5.5 Cost function and device parameters for the two device configuration . . . . . . . . . . 136
xi
LIST OF TABLES
B.1 Control surface deflection limits for the BUG-T NeoCASS model . . . . . . . . . . . . . 187
B.2 Fuel tank parameters for the BUG-T NeoCASS model . . . . . . . . . . . . . . . . . . . 188
B.3 Fuel tank volumes for the BUG-T NeoCASS model . . . . . . . . . . . . . . . . . . . . . 189
B.4 Mass distribution of the BUG-T NeoCASS model . . . . . . . . . . . . . . . . . . . . . . 192
E.1 Nastran transfer function coefficients for an inerter and TID device . . . . . . . . . . . 220
xii
L IST OF F IGURES
F IGURE Page
xiii
LIST OF FIGURES
3.7 Relative velocity at the proposed vibration suppression device locations as a function
of frequency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.8 Example panel geometry for the Nastran doublet lattice method . . . . . . . . . . . . . 69
3.9 Static aeroelastic loads envelope for the BUG-T FEM . . . . . . . . . . . . . . . . . . . 74
3.10 Static aeroelastic deflections for the BUG-T FEM . . . . . . . . . . . . . . . . . . . . . . 75
3.11 Pratt gust load factor for different values of C L α . . . . . . . . . . . . . . . . . . . . . . . 77
3.12 Change in beam loads associated with Pratt gusts for different values of C L α . . . . . . 78
3.13 Flutter speed vs. Mach number for the BUG-T model . . . . . . . . . . . . . . . . . . . 82
3.14 Reduced frequency/damping plots for the flutter response of the BUG-T model . . . . 82
3.15 Flutter modeshape and complex modal coordinates for branch seven, Mach 0.726,
U∞ = 212m/s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83
3.16 Flutter modeshape and complex modal coordinates for branch 10, Mach 0.726, U∞ =
228m/s . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
3.17 Relative modal displacement at the proposed vibration suppression device locations
for flutter branches 7 and 10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
5.1 Aeroelastic finite element model of the BUG-T half-wing model . . . . . . . . . . . . . 116
5.2 1-cosine gust family . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
5.3 Incremental gust loads envelope for the BUG-T half-model . . . . . . . . . . . . . . . . 119
5.4 Example of time-correlated incremental gust loads for the BUG-T half-wing model . 120
5.5 Maximum modal coordinates for the BUG-T half-model during a discrete gust encounter121
5.6 Gust bandwidth and energy density for an altitude of 36000ft as a function of gust
gradient . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
xiv
LIST OF FIGURES
6.1 Diagram showing line of action of the device force with respect to a rotational joint . 144
6.2 Pressure vessel diameter and available device stroke as a function of lever arm and
joint rotation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 145
6.3 Diagram of a typical hydraulic damper . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
6.4 Range of viscous damping coefficients for the hydraulic damper concept . . . . . . . . 148
6.5 Equivalent viscous damping coefficient and orifice diameter for a range of piston and
normalised orifice diameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
6.6 Network diagram for a damper including fluid compressibility effects . . . . . . . . . . 150
6.7 Fluid stiffness as a function of pressure vessel length and diameter . . . . . . . . . . . 150
6.8 Damper mass as a function of pressure vessel length and diameter . . . . . . . . . . . 152
6.9 Diagram of a conceptual fluid inerter device . . . . . . . . . . . . . . . . . . . . . . . . . 153
6.10 Network diagram of the fluid inerter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
6.11 Parameter study of the fluid inerter device . . . . . . . . . . . . . . . . . . . . . . . . . . 155
6.12 Variation in inertance values as a function of helical channel dimensions . . . . . . . . 156
6.13 Fluid inerter mass as a function of pressure vessel length and diameter . . . . . . . . 157
6.14 Physical parameters for a viscous damper with c equiv = 1 × 105 . . . . . . . . . . . . . . 159
6.15 Physical parameters for the final TID and TID-D fluid spring . . . . . . . . . . . . . . . 160
6.16 Physical parameters for the final TID and TID-D fluid inerter . . . . . . . . . . . . . . 161
6.17 GLA performance for a single device at the strut-root joint . . . . . . . . . . . . . . . . 163
xv
LIST OF FIGURES
xvi
HAPTER
1
C
I NTRODUCTION
The global aviation community is under increasing pressure to reduce the environmental impact
of air travel. In 2010, the International Air Transport Association (IATA) established a series of
environmental goals for aviation 1 . These included:
• An average improvement in fuel efficiency of 1.5% per year from 2009 to 2020.
• A cap on net aviation CO2 emissions from 2020 (carbon-neutral growth)
• A reduction in net aviation CO2 emissions of 50% by 2050, relative to 2005 levels.
Various other international authorities have defined similar environmental goals, including
the International Civil Aviation Authority 2 (ICAO), the European Union 3 and NASA 4 . Several
studies are investigating engineering solutions to achieve these goals, including NASA’s Subsonic
Ultra Green Aircraft Research (SUGAR) project 5 and the Clean Sky initiative 6 .
The IATA goals are principally concerned with reducing the carbon-footprint of commercial
aviation. This can be achieved by making aircraft more efficient, however, in order to make
sure the aircraft is commercially viable any improvement in efficiency must not come at the
expense of compromising safety or damaging the airline’s business model. For example, the
aircraft must meet the demanding requirements on turnaround time, route time and affordability
that modern, low-cost airlines rely on to make a profit. These considerations make the design of
the next-generation commercial aircraft a challenging task which is forcing manufacturers to
consider radical concepts such as hybrid-electric propulsion 7 , new aircraft configurations 8 and
formation flying 9 .
1
CHAPTER 1. INTRODUCTION
Concerning aircraft efficiency, the main contributing factors can be deduced by considering
the Brequet Range Equation 10 ,
V L/D Wf
µ ¶ µ ¶
R= ln 1 + (1.1)
g SFC Wp + W0
where R is the aircraft range, V is the aircraft forward velocity, g is the acceleration due to
gravity, SFC is the Specific-Fuel-Consumption of the engine, (L/D) is the lift to drag ratio and
W f ,Wp ,W0 are the weights of the fuel, payload and airframe respectively. Note that this equation
only applies for the case where the fuel-burn, velocity and aerodynamic performance are constant
with respect to time, i.e. during the cruise phase of the mission. Examining the Brequet Range
Equation it is clear that for a fixed cruise speed the range is dependent on three factors:
• Specific-Fuel-Consumption (SFC) - This is typically a function of cycle and propulsive
efficiency, which, for a jet engine, is driven by the core temperature of the combustion
chamber and the by-pass ratio.
• Lift-to-drag ratio - Dependent on the aircraft geometry.
• Aircraft empty weight - Mostly driven by the aircraft mission profile and overall design
but is strongly affected by the material performance, specifically the density.
• Fuel weight - Closely linked to the SFC of the engine as well as the route length.
• Payload Weight - Dictated by the route analysis and business case of the various airlines
and as such is fixed very early on in the design process.
A wide range of technological improvements have been implemented over the past 50 years
targeting these areas, including but not limited to: higher by-pass-ratio engines, more aerody-
namically efficient designs and high-performance materials such as composites. Figure 1.1 shows
that whilst these innovations have led to a steady improvement in overall efficiency the latest
generation of aircraft have seen smaller incremental improvements in performance than their
predecessors. During this time development costs and unit prices have increased 11,12 and airlines
are beginning to see a smaller and smaller benefit from ordering newer aircraft models, preferring
instead to extend the life of their existing fleet. Therefore, the next-generation of aircraft will
need to provide significant performance benefits over current models which may not be possible
with the traditional ‘tube-and-wing’ configuration 4 .
Regarding aerodynamic efficiency, the dominant metric of interest is the aircraft lift-to-drag
ratio (L/D), which must be maximised to obtain the best possible performance. As the aircraft
must generate enough lift to sustain level flight during cruise the only option to improve (L/D) is
to minimise the drag force, which can be broken down into three main components 13 :
• Parasitic Drag - Resulting from shear and pressure forces acting on any exposed surface.
• Induced Drag - Additional shear and pressure forces due to changes in the flow-field
around the aircraft from lift forces.
• Wave Drag - Associated with the pressure drop across shock waves.
Figure ?? shows the drag breakdown for a typical commercial airliner operating in the transonic
region 14,15 . Noting that the exact composition of the drag force is dependent on a number of
2
1.1. RESEARCH MOTIVATION AND THEMES
factors, it is clear that the parasitic and induced drag terms are dominant whilst the wave-drag
contributes only a small fraction to the total drag1 . Moreover, the drag force varies as a function
of aircraft speed. For example, Figure ?? shows that at low speeds the induced drag constitutes
the majority of the drag force and the parasitic drag is minimal, however, as the speed increases
there is a quadratic decrease/increase in the induced-/profile-drag respectively and the point
where the two lines intersect is the point of minimum drag and the corresponding speed is the
minimum drag speed (VMD )2 . According to Torenbeek 16 , p. 158 an aircraft will typically fly
10-20% faster than VMD to maintain stability and avoid buffet as well as improve the block time,
hence the so-called "best utilisation speed" in Fig. ??.
Reducing lift-induced drag is important as it enables cruise efficiency to be maximised and
limits drag during the take-off and landing phases. The later allows a greater maximum take-
off/landing weight to be achieved which provides an increase in performance greater than the
cruise benefits alone 15 . According to Prandtl-Glauert theory 13,17 , which is valid for describing
the aerodynamics of finite unswept wings in steady incompressible non-viscous flow, lift-induced
1 For commercial aircraft the cruise Mach number is typically chosen such that it is less than the Drag-Divergence
Mach number, thereby minimising wave drag. For aircraft operating further in the transonic region or at supersonic
speeds the contribution from wave drag is significantly higher.
2 This approach is consistent with the one presented in Torenbeek 16 where wave drag is neglected because the
complex flow behaviour at transonic speeds cannot be accurately described by analytical formulae.
3
CHAPTER 1. INTRODUCTION
F IGURE 1.2. Drag components for a typical transonic commercial aircraft 14,15 .
C 2L
CD i = (1.2)
e 0 π AR
where, C D i is the lift-induced drag coefficient, C L is the lift coefficient, AR is the wing aspect ratio
and e 0 is the Oswald Efficiency Factor; a factor that defines how close the lift distribution is to the
theoretical elliptical lift distribution. For a fixed C L , such as during cruise, there are two options
for reducing the induced drag - increasing the aspect ratio or increasing e 0 . In recent years, a
significant effort has been made to reduce lift-induced drag by using ‘winglets’ - devices which
carefully manage the formation and strength of wingtip vortices in order to improve the Oswald
Efficiency Factor. Several commercial aircraft utilise winglets as they allow for a reduction in
drag with only a small increase in the total wingspan, something that is highly beneficial for
airlines as ground-handling costs at airports are linked to the overall size of the aircraft. Indeed,
numerous studies have shown that for a fixed span a winglet provides a reduction in total drag of
approximately 5% with minimal increases in root bending moment, although, it remains unclear
whether a winglet is categorically better than an equivalent span extension 18–21 .
As well as using winglets to control lift-induced drag, there has been a trend in recent aircraft
to increase the aspect ratio in order to restrict the influence of wingtip vortices to the outer-portion
of the wing and reduce induced drag. For example, Fig. 1.3 shows a steady increase in aspect ratio
in recent decades, specifically for twin-engine, wide-body aircraft, such as the A350 and B787,
which have begun operating on longer routes with significantly higher ETOPS3 ratings. Whilst
a higher aspect ratio can improve performance during flight, when operating on the ground
(i.e. at airports) the wingspan is limited by the ICAO gate codes 22 , thereby placing a physical
limit on the overall dimensions of the aircraft when it is at the gate. To counter this, a new idea
3 ETOPS refers to Extended Operations, which is the amount of time an aircraft is rated to fly at the one-engine
4
1.1. RESEARCH MOTIVATION AND THEMES
20
Sugar Volt
18 Concept
Aircraft
D8
14
Future Concepts
12 A330-800neo D8 ‘Double-Bubble’
737 MAX
AR
ing
c reas 777-X
10 In
A350-900 XWB
A330-800neo
8 Airbus
747-8
Boeing
A380-800 Conceptual
6
1975 1980 1985 1991 1996 2002 2007 2013 2018 2024 2029 2035
First Flight (Year)
F IGURE 1.3. Wing aspect ratio vs. aircraft entry into service. The lines between data
points denote aircraft families.
has emerged that involves incorporating a ‘fold’ into the structure of the wing. This concept is
currently being investigated for the new B777X 23 where it would allow the full wingspan to be
achieved during flight and then for ground operations a section of the wing will fold up in order to
meet the requirements on airport gate size. Seperately, Airbus are investigating the possibility of
using a semi-aeroelastic folding wingtip to provide loads alleviation during turbulence encounters
and have produced several papers and patents concerning this application 24–26 , as well as a
scale-model flight demonstrator named AlbatrossOne 27 .
The possibility of a wingspan that can exceed the normal gate size has led many researchers
to investigate high aspect ratio wing (HARW) designs as a means of reducing lift-induced drag.
Notable examples of HARW concept aircraft include the Double Bubble D8 28 , Airbus Concept
Plane 29 and NASA/Boeing SUGAR Volt 30 , images of which are shown in Figure 1.4. The aspect
ratio of these concept aircraft is in range of 17-25, whereas a typical commercial aircraft has an
aspect ratio of less than 10. Such designs represent a departure from the typical ‘tube-and-wing’
5
CHAPTER 1. INTRODUCTION
configuration and as such none of these concepts have progressed beyond the initial design stage,
although they are the subject of intense and on-going research .
Whilst high-aspect ratio wings provide an aerodynamic benefit there is a considerable penalty
for the wing structural weight. The increased span leads to a larger moment arm, causing
increased bending loads along the wing which then require additional structure in order to
prevent the components from exceeding their failure stresses/strains, hence the increase in
weight. Various methods have been proposed to mitigate for this, including: aeroelastic tailoring
using composite materials 33–36 , active loads alleviation using aerodynamic control surfaces 37–40
and morphing wings 41,42 . Another possible solution is to use an external bracing structure to
support the wing, thus providing additional stiffness and reducing internal loads. This external
structure can be comprised of any number of ‘strut’ elements which are collectively described as
the ‘truss’, as in Figure 1.5, hence the name for this concept is the Truss-Braced Wing (TBW).
The concept of a truss-braced wing aircraft was first popularised during the Second World
War but quickly lost out to the traditional monoplane ‘tube-and-wing’ configuration which became
dominant during the jet era and beyond. However in recent years there has been renewed interest
in this concept, thanks in part to the SUGAR project which has identified the TBW as one of the
concepts that can achieve the N+3 emissions goals set out in Collier et al. 4 and more recently
it has been announced as one of Boeing’s concept aircraft 43 . The addition of the truss leads to
6
1.1. RESEARCH MOTIVATION AND THEMES
Jury-Struts
Wing
Strut
a structurally efficient design with reduced weight and lower induced drag from the increased
aspect ratio, however, the truss structure generates interference drag that must be traded as part
of a comprehensive multidisciplinary optimisation study 44–46 . Despite this, it has been shown
that the TBW concept can provide improvements in maximum take-off weight and fuel burn
over a traditional cantilevered-wing design if the interface between the truss structure and the
wing is properly designed. Recently, Malik et al. 47 identified that aeroelastic phenomena such
as flutter are the main constraints which drive the overall design of the TBW aircraft, thereby
placing a limit on the practical efficiency savings that can be achieved when considering a braced
wing design. Furthermore, initial sizing studies conducted as part of the SUGAR project 5 have
shown that gust loads are responsible for sizing many of the structural components in the wing
box 30 . Hence, a primary requirement for the success of the TBW concept is to reduce the effect of
these aeroelastic phenomena, thus enabling further reductions in wing weight which leads to a
more efficient design.
In this thesis an alternative approach to aeroelastic control in a truss-braced wing is presented,
which is to use a two-terminal vibration suppression device embedded within the truss structure
to passively provide gust loads alleviation and flutter suppression. Such an approach is not
feasible in a cantilever wing as a location does not exist where the relative motion of the wing
structure can be exploited by a two terminal device, however the addition of the truss-structure
means there are now several candidate locations where a device could be placed. For example,
if a hinge connection between the strut and the wing is employed then the rotation of the strut
about that joint could be utilised or the jury-strut may experience significant relative motion
due to the combined bending of the wing and the primary strut. This application of a vibration
absorber to control aeroelastic effects in truss-braced wings is highly novel and has not, to the
best of the author’s knowledge, been investigated previously. As such, there are several questions
that this thesis will attempt to answer:
7
CHAPTER 1. INTRODUCTION
1. Can a vibration suppression device reduce structural loads during a turbulence encounter?
2. Can a vibration suppression device provide flutter suppression?
3. Assuming a vibration suppression device yields performance improvements, are the mass
and geometry of the device feasible within the context of an aerospace application?
4. How can a vibration suppression device be integrated within a complex structure and then
designed and optimised in a computationally efficient manner?
1.3 Publications
The following conference papers have been presented during the course of this PhD:
1. Christopher P. Szczyglowski, Christopher P. Howcroft, Simon A. Neild, Branislav Titurus,
Jason Z. Jiang, Jonathan. E. Cooper and Etienne Coetzee, "Strut-Braced Wing Modelling
with a Reduced Order Beam Model.", Proceedings of the Royal Aeronautical Engineering
Society 5th Aircraft Structural Design Conference, 2016.
2. Christopher P. Szczyglowski, Simon A. Neild, Branislav Titurus, Jason Z. Jiang, and
Etienne Coetzee, "Passive Gust Loads Alleviation in a Truss-Braced Wing Using Integrated
Dampers", Proceedings of the 17th International Forum on Aeroelasticity and Structural
Dynamics, International Forum on Aeroelasticity and Structural Dynamics, 2017.
3. Christopher P. Szczyglowski, Simon A. Neild, Branislav Titurus, Jason Z. Jiang, and
Etienne Coetzee, "Passive gust load alleviation in a truss-braced wing using an inerter-
based device", Proceedings of the AIAA/ASCE/AHS/ASC Structures, Structural Dynamics,
and Materials Conference, American Institute of Aeronautics and Astronautics, 2018. DOI:
10.2514/6.2018-1958
Furthermore, the following journal papers have been published:
1. Christopher P. Szczyglowski, Simon A. Neild, Branislav Titurus, Jason Z. Jiang, and Etienne
Coetzee, "Passive Gust Loads Alleviation in a Truss-Braced Wing Using an Inerter-Based
Device", Journal of Aircraft (2019), accessed September 01 2019. DOI: 10.2514/1.C035452
8
1.4. THESIS OUTLINE
This thesis consists of seven chapters and five appendices which are organised as follows:
• Chapter 1 provides an introduction to the general research themes and outlines the novelty.
• Chapter 2 is a comprehensive review of the research literature which highlights some
important gaps in the field of truss-braced wing design, some of which are addressed here.
• Chapter 3 introduces the full-span BUG-T aeroelastic finite element model which is derived
from the NASA/Boeing SUGAR 765-095 Rev. D model. The fundamental dynamic and aeroe-
lastic behaviour is characterised and three potential locations for the vibration absorber
are identified. These insights are used in Chapters 4 and 5 to inform the optimisation of
vibration suppression devices for flutter suppression and gust load alleviation.
• Chapter 4 investigates the potential of a vibration suppression device to alleviate flutter in
truss-braced wings. The techniques for modelling and optimising a vibration absorber in
Nastran are discussed and the three candidate device layouts are introduced. A combined
MATLAB-Nastran optimisation scheme is used to optimise the device parameters in order
to maximise the flutter speed of the BUG-T model.
• Chapter 5 evaluates the potential for vibration suppression devices to provide gust loads
alleviation in a truss-braced wing. Here, a half-wing version of the BUG-T is used and the
response to a family of discrete 1-cosine gusts is evaluated. Next, a device optimisation
framework is introduced where the optimisation is formulated as a frequency response
problem with the device parameters optimised to target specific structural modes in the
primary system. Finally, this framework is used to evaluate both single and multiple
absorber configurations for gust loads alleviation.
• Chapter 6 presents the preliminary design of a vibration suppression device that can
provide the required linear force coefficient values identified by the flutter suppression and
gust load alleviation studies. The objective of this chapter is to understand whether the
device properties are realisable within the context of a truss-braced wing aircraft.
• Chapter 7 is the final chapter and provides a summary of the technical work presented in
this thesis as well as recommendations for future work.
• Appendix A provides the reference data for the SUGAR 765-095 Rev. D aircraft model.
• Appendix B details the NeoCASS model which provides the BUG-T mass distribution.
• Appendix C provides an overview of the BUG-T model and discusses the parametric aircraft
model generator O2 MeGA which was used to generate the BUG-T model.
• Appendix D provides additional results related to the analysis in Chapter 3.
• Appendix E discusses three different methods for modelling vibration absorbers in the
Nastran finite element software and details the pros and cons of the different approaches.
9
HAPTER
2
C
B ACKGROUND T HEORY AND L ITERATURE R EVIEW
his chapter provides an overview of the research literature that is pertinent to the studies
T presented in this thesis. Three main topic areas are discussed: In Section 2.1 a brief
overview of the field of aeroelasticity is provided, including the basic equations of motion,
the various phenomena associated with aerolastic interactions and a discussion on some of the
methods used to simulate these systems. In Section 2.2, a comprehensive review of the truss-
braced wing literature is provided, with specific emphasis placed on the the rich history of the
truss-braced wing aircraft, including its origins during the pioneering days of powered flight.
Finally, in Section 2.3 the concept of vibration suppression is introduced, with consideration
given to the fundamental physical principles, the preliminary design of vibration suppression
devices and their application to engineering structures. Through this review it is shown that the
use of vibration suppression devices in truss-braced wings has not been covered in the existing
literature, which only serves to highlight the novelty of the work presented in this thesis.
2.1 Aeroelasticity
Classical aeroelasticity is the study of the interaction between aerodynamic, elastic and iner-
tial forces, shown pictorially by Collar’s Aeroelastic Triangle in Figure 2.1. Collar’s Triangle
demonstrates that a variety of interactions are possible between the three forces, for instance the
interaction between aerodynamic and inertial forces is related to the discipline of Stability and
Control whilst the interplay between inertia and elastic forces is the domain of Vibrations. In
general, aeroelasticity is divided into two parts: static aeroelasticity and dynamic aeroelasticity -
the former is concerned with non-oscillatory aerodynamic forces acting on a flexible structure
whereas the latter is the interaction of all three forces and includes unsteady aerodynamic terms.
In his review paper, Friedmann 48 argues that the Collar’s Triangle can be expanded into an
11
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
Inertial Forces
Dynamic Aeroelasticity
Gust and Turbulence Loads
Elastic Aerodynamic
Forces Forces
Static Aeroelasticity
Equilibrium/Steady
Manoeuvre Loads
F IGURE 2.1. Collar’s Aeroelastic Triangle and associated loads, from Wright and
Cooper 49
Controls
C
I Inertial
Elastic E
A
Aerodynamic
T
Thermal
12
2.1. AEROELASTICITY
There are several practical consequences of including aeroelasticity in the design of fixed-wing
aircraft. Firstly, the aerodynamic forces acting on the wing - and other exposed surfaces - cause
the structure to deflect, and as the local aerodynamic forces are dependent on the local angle-
of-attack this deflection causes a change in the the aerodynamic forces. During the modelling
process this interaction can be iterated until an equilibrium position is reached where the external
aerodynamic and inertial loads are equal to the internal reaction loads from the elasticity of the
structure, resulting in a deflected shape known as the static aeroelastic shape or 1g jig-shape. As
the drag force acting on the aircraft is a function of its shape it is important that static aeroelastic
effects are accounted for otherwise a significant drag penalty will be incurred, which will have
a detrimental effect on performance. Furthermore, it is possible for the aerodynamic loads to
exceed the internal reaction loads resulting in the catastrophic failure of the airframe. This
phenomena is a type of static instability known as divergence and is strongly dependent on the
stiffness of structure. Another stiffness-related static-aeroelastic phenomenon is control reversal
which occurs when the aerodynamic moment due to wing twist exceeds the nose-down pitching
moment of the control surface (typically an aileron) causing the opposite response to the one the
pilot has demanded.
Regarding dynamic aeroelasticity, the main concerns are due to flutter - an instability caused
by interactions between unsteady aerodynamic forces and the flexible modes of the structure
- and increased airframe loads due to turbulence encounters. The latter can be explained by
considering atmospheric turbulence as unsteady air (‘gusts’) which have velocity components
in all three directions. Any component of the gust velocity which is normal to the aircraft’s
flight path will cause an induced angle-of-attack and hence a change in the aerodynamic and
internal loads. For cases of severe turbulence these perturbations in the gust velocity can cause
an increase in loads that would otherwise exceed the limit loads of the structure if they were not
included in the design process. For aircraft operating the in transonic region the effects of shock
buffet1 must also be accounted for, both in the definition of the flight envelope and the design of
aerofoil cross-sections. All of the considerations listed in the previous paragraphs are well-known
to aerospace engineers and are the subject of several textbooks 49–52 and review papers 48,53–56 .
Clearly the effects of aeroelasticity must be accounted for during the design process. Histori-
cally, this has been achieved by increasing the stiffness of the structure such that the frequencies
of the flexible structural modes are outside of the range of excitation frequencies typically as-
sociated with unsteady aerodynamics, or similarly by adding mass to the structure in order to
decouple the structural modes from the aerodynamic forcing. Such an approach inevitably leads
to an increase in structural mass, hence the term ‘flutter penalty’ used to describe any increase
in weight due to aeroelastic effects. Another means of alleviating negative aeroealstic effects is
1 Shock buffet refers to high frequency instabilities causes by flow separations, shock wave oscillations and wake
13
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
to use a technique known as active loads control, where active control methods are employed to
actuate control surfaces in such a way that unsteady aerodynamic loads are minimised 37 . Whilst
this approach has been widely adopted by the aerospace communitity active control methods are
complex and notoriously difficult to certify 57–60 . In recent years researchers have investigated
ways to take advantage of aeroelasticity in order to provide a benefit to the overall aircraft design.
Potential avenues of approach include curvilinear structural elements 61 , composite tailoring 33,62
and morphing structures 41,63 .
To properly account for aeroelastic effects as well as make them beneficial to aircraft perfor-
mance aeroelasticity must be inherent in the aircraft design process, see Fig. 2.3. The classical
approach to aeroelastic design is use low-fidelity aeroelastic models during the conceptual design
phase in order to generate global mass and stiffness data for the aircraft. This data is typically
corrected using historical data based on aircraft-family type, which often has only a loose cor-
relation to the physical attributes of the aircraft. As the design progresses, medium and high
fidelity modelling becomes more common, however, due to limitations surrounding computational
time (and cost) these models still require corrections using data from wind tunnel and flight
tests. If a problem is identified during testing then a redesign is incredibly costly and can delay
an aircraft programme by months if not years. A more favourable approach is to incorporate
higher fidelity tools early in the design process using a concept known as the virtual aircraft 67 .
Here the objective is to offload much of the physical testing to an earlier stage in the design,
allowing issues to be identified much earlier and reducing the potential for costly redesigns. Such
an approach is further necessitated by the recent interest in unconventional aircraft types for
which no historical data exists, making it even more important that at the conceptual design
stage the aircraft is evaluated using physics-based models. To this end, several researchers have
developed multidisciplinary optimisation (MDO) tools which allow for the analysis and parametic
design of fully-flexible aircraft; notable contributions include: the CPACS schema 68 , NeoCASS 69 ,
14
2.1. AEROELASTICITY
Fidelity
Virtual Aircraft
Flight tests
Higher
fidelity Classical Approach
earlier
Wind tunnel tests
- Design improvements
- Dataset generation
- Design verifications
Design Maturity
Conceptual Preliminary Detailed Design
Design Design
Manufacture
GeoMACH 70 and PyGFEM 71 . These tools encompass different levels of fidelity and each one is
aimed at improving the understanding of aeroelasticity during the initial design stage as well as
increasing modelling fidelity at all stages of the design process, as shown qualitatively in Fig. 2.3.
of a finite element (FE) representation of the structure. In structural modelling the FE approach is most common but
it is by no means the only method for representing the structure.
15
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
model various levels of fidelity are available, from simple analytical 2D formulations (e.g.
strip theory) to complex 3D numerical models that utilise computational fluid dynamics
(CFD) 73 .
• Aero-Structural Coupling - A means for the aerodynamic loads to be applied to the
structure and for structural deformations to be translated to the aerodynamic states which
in turn generate new aerodynamic loads. Basic models utilise a fixed reference point for the
application of aerodynamic forces and moments, such as the aerofoil quarter-chord, however
it is more typical to use mathematical splines to generate a matrix relating the structural
states (deflections, twists, global geometry) to the aerodynamic states (angle-of-attack,
downwash velocity, outer-mold-line). The coupling of CSD and CFD methods is referred to
as Computational Aeroelasticity (CAE).
Bearing these three core components in mind, the basic equation of motion (EoM) governing
aeroelastic systems is given in Wright and Cooper 49 as
2
A q̈ + ρ f U∞ B + D q̇ + ρ f U∞
¡ ¢ ¡ ¢
C+E q =0 (2.1)
16
2.1. AEROELASTICITY
described the frequency dependent nature of the aerodynamic forces and related this to a reduced
frequency, k = ω c/2U∞ , where ω is the circular frequency of the aerofoil motion and c is the
aerofoil chord. This dimensionless parameter indicates whether the aerofoil will extract energy
from the fluid and hence whether flutter will occur.
As the community’s understanding of the physical principles of aeroelasticity grew the com-
plexity of the models increased, leading to the development of matrix methods to efficiently handle
the larger EoM 80 . This early work would lay the foundations for the wide-scale adoption of finite
element analysis (FEA) as the basis for most medium-to-high fidelity aeroelastic modelling 81 .
One of the most widely adopted codes was developed by NASA during the 1960s to allow the
modelling of complex aerospace flight vehicles which could not be accurately represented using
existing methods. This code, NASTRAN (NASA Structural Analysis), has since become ubiquitous
across a number of engineering disciplines and is considered to be industry-standard in aerospace
engineering. However, despite this accolade, the aeroelastic capabilities of NASTRAN are in fact
quite limited. For example, the subsonic aerodynamic formulation available in NASTRAN is
the Doublet-Lattice Method (DLM) 82 . This formulation is based on linear potential flow theory,
meaning that the aerodynamic forces are only valid for inviscid, irrotational, incompressible
and attached flow, subject to small angles-of-attack or side-slip 13 . Therefore, DLM cannot give
accurate predictions of aerodynamic forces at transonic speeds (which most commercial aircraft
operate at), estimate drag associated with viscous and compressible effects or model shock buffet.
Despite these limitations it is generally understood by the industry that the DLM approach can
give a reasonable understanding of the general aeroelastic characteristics of an aircraft, as well
as provide predictions of the aircraft response during turbulence encounters. Furthermore, the
pressure distribution from the DLM can be ‘corrected’ using results from CFD analysis, allowing
an interaction between the aerodynamic and structures disciplines that was previously difficult
to realise. Indeed, the strength of NASTRAN lies in its ability to model the structural properties
of any generic model and to couple this to a broadly accurate (on an aircraft-level) representation
of the aerodynamics. This vehicle level approach to aeroelasticity is what makes NASTRAN so
useful for the analysis of aircraft structures, even if the fidelity of the aerodynamic model is
limited to what is typically expected at the preliminary design stage.
In the period between the 1960s and the late 1990s the trend was towards increasing the
fidelity of analysis models to provide improved estimates of the structural and aerodynamic
aspects of the aircraft. However as the models increased in size the computational cost of
dynamic aeroelastic analysis became prohibitive, leading to the investigation of dynamic reduction
techniques 83,84 . The aim of this process is to retain the key dynamic characteristics of the full-
model but using a reduced number of states, typically the normal modes of the structure. This
method is wide-spread and remains a common approach for many modern CAE tools. Other
important developments also took place in the fields of time domain aeroelastic modelling and
transonic flutter prediction 48 . With regards to the former, the introduction of the Rational
17
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
Fluid-Structure-Interaction (FSI) 91 , however as this involves generating aerodynamic loads outside of the NASTRAN
environment the prior statement on NASTRAN aeroelastic capabilities remains valid.
18
2.1. AEROELASTICITY
3. Fully nonlinear - All aspects of the system are modelled using nonlinear methods.
There are several sources of nonlinearity in aeroelastic systems, including: control surface
freeplay, aerodynamic nonlinearity - both from transonic aerodynamics and high angle-of-attack
flow (e.g. stall flutter) and geometric nonlinearity - nonlinearirty due to large-scale structural
deflections. A typical hallmark of a nonlinear system is the presence of a Limit Cycle Oscillation
(LCO) which is a bounded oscillation where energy is traded between the fluid and the structure.
It has been shown that LCOs in aircraft structures can occur due to the presence of both
structural and aerodynamic nonlinearities 99 , as well as internal resonances within the system 55
and more recently LCOs were observed during wind-tunnel testing of the NASA SUGAR Volt
aircraft 100 . Concerning the modelling of nonlinear aeroelastic systems, as aerospace structures
are typically slender in nature there has been a concerted effort to develop rapid, yet accurate,
1D beam formulations that encompass the global stiffness and inertia properties of the aircraft
whilst capturing geometrically nonlinear deflections 101–103 . The nonlinear aerodynamics can
then by incorporated by developing reduced order models (ROM) of the sectional lift, drag and
moment coefficients based on CFD data, which are then incorporated into the low order structural
formulations. These techniques have been applied to a wide variety of aircraft types, including:
high altitude long-endurance (HALE) 104 , joined wings 105 , strut-braced wings 106 and blended-
wing-body aircraft 107 . The full nonlinear modelling of dynamic aeroelasticity using high-fidelity
CSD-CFD models is still yet to be realised, however it remains the ‘holy-grail’ for much of the
aeroelastic community.
Below is a brief summary of the main points that have been identified from the papers reviewed
in this section:
• Aeroelasticity is a well-developed field with applications across various industries and a
broad range of modelling techniques.
• Research state-of-the-art is concerned with high-fidelity nonlinear aeroelastic modelling
however traditional aeroelastic tools such as NASTRAN are still used to model vehicle-level
aeroelastics of both conventional and unconventional aircraft.
• The principal problems arising from aeroelastic effects in fixed wing aircraft are increased
structural loads due to atmospheric turbulence, loss of control due to aileron reversal and
catastrophic failure due to flutter instability. These effects must be included at the early
stages of the design process in order to derive additional performance benefits as well as
properly understand their impact on the aircraft design.
• Numerous researchers are investigating methods to mitigate aeroelastic effects, notable
technologies include: active control via aerodynamic control surfaces 37,57–60 , composite
tailoring 33,62 and morphing structures 24,26,39,40,64–66 .
19
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
In this section a complete overview of the braced wing concept is provided in order to familiarise
the reader with the configuration and also highlight areas which have not been sufficiently
explored in the existing research. To avoid ambiguity the term "braced wing" encompasses both
strut-braced wing (SBW) and truss-braced wing aircraft, examples of which are shown in Fig. 2.4.
Before progressing it is necessary to clarify the vocabulary specific to braced-wings:
• Strut - A structural component which forms a connection between the fuselage and the
wing, sometimes referred to as a "lifting-strut". The main function of the strut is to limit
the deflection of the wing, which has the effect of reducing the loads inboard of the strut
attachment point.
• Jury-Strut - A structural component which connects a strut to the wing. The purpose of
a jury-strut is to reduce the effective buckling length of the strut.
• Truss Structure - A collection of strut and jury-strut elements.
• Strut-Braced Wing (SBW) - A wing that is supported by strut elements only.
• Truss-Braced Wing (TBW) - A wing that is supported by strut and jury-strut elements.
Jury-Struts
Wing
Pinned/Ball-Joints
Strut
20
2.2. THE BRACED-WING AIRCRAFT
On the connectivity between the truss-structure and the wing/fuselage components various
joint types have been proposed, including:
• Pinned - Allows rotation about a single axis. Moments in the two fixed axes are non-zero
at the joint location and these loads are transferred across the joint.
• Ball - Allows rotation about all three axes. All bending moments/torques have a zero-value
at the joint location and only forces are transferred across the joint.
• Fully-Fixed - No rotation is allowed and all moments/torques/forces are transferred
across the joint. A fully-fixed joint is usually accompanied by an aerodynamic fairing which
limits interference drag from the strut.
In terms of numerical modelling, accounting for different joint types is as simple as releasing
specific degrees of freedom in the model, however the physical design of the joints is more
complicated and has not been addressed in any detail in the literature. Figure. 2.5 provides a
colour-coded breakdown of a TBW with the various structural members and joint types annotated.
The braced-wing concept offers an attractive alternative to the cantilever wing design as
the addition of the truss structure creates a structurally efficient design which reduces the
spanwise bending moments and torques inboard of the strut 108,109 , as in Fig. 2.6. This enables
the inboard portion of the wing to be designed with a reduced thickness-to-chord ratio providing
a reduction in wing weight as well as lower wave and profile drag. Furthermore, the reduced
bending moment allows a larger wingspan to be achieved resulting in an increased aspect ratio
and a reduction in induced drag. Multidisciplinary optimisation studies 44,45 have shown that
a truss-braced wing design has a lower sweep angle than a cantilever wing. This leads to a
reduction in spanwise crossflow 110 thus promoting more laminar flow and hence lower friction
drag, however it should be noted that the interference drag from the truss-structure can have a
Spanwise Bending Moment
Cantilever Wing
∆BM
strut attachment
Braced Wing
Span
F IGURE 2.6. Qualitative comparison of the spanwise bending moment distribution for
a braced and cantilevered wing
21
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
detrimental effect on performance if it is not properly designed 111,112 . Even with the effects of the
truss interference drag taken into account, numerous studies have shown that the truss-braced
wing concept can provide an overall improvement in maximum take-off weight and fuel burn
compared to a traditional cantilever design 44–46 .
This section is ordered more-or-less chronologically, starting with the early adopters of the
TBW concept during the initial days of powered flight and ending with the substantial body of
work carried out as part of the NASA SUGAR project. As the braced wing aircraft has historically
received less attention than the cantilever configuration the quantity of published research on
it is limited. For this reason the papers reviewed here are discussed in more detail than other
sections in order to provide the reader with a complete understanding of the work that has been
completed to date. Furthermore, Section III of the review paper by Cavallaro and Demasi 113
is an excellent resource for understanding the braced-wing concept within the scope of other
joined-wing research.
The braced-wing aircraft can be thought of as an simplified version of traditional bi- and tri-planes
from the early days of powered-flight. In these aircraft, wires or other supporting structures
were used to brace the wings in order to maintain structural integrity and ensure the upper
wing(s) remained attached, as in Figure 2.7(a). However, as aircraft structures became more
advanced and the community’s understanding of aerodynamics improved, the need for multi-wing
configurations diminished and the industry settled on the mono-plane concept with a conventional
tube-and-wing design. Despite the prevalence of the tube-and-wing configuration some early
aircraft designers continued to investigate braced-wings but for a mono-plane design. As the size
of the aircraft increased designers moved away from wire-bracing and instead utilised metallic
structures to provide support for the main wing, usually in a truss-like structure similar to
many bridges of the era, hence the name truss-braced wing. Several such aircraft were developed
during the Second World War, including: the Piper Pawnee, Consolidated PBY Catalina and
the Westland Lysander. Many of these were light-aircraft or flying-boats, with the braced-wing
favoured for its high-span which provided excellent short take-off and landing performance.
One of the early pioneers of the TBW concept was Maurice Hurel, who founded the Hurel-
Dubois aircraft company in 1947 and flew his first high aspect-ratio wing (HARW) demonstrator
aircraft in 1948. Hurel was one of the first designers to realise the benefits of a high-aspect
ratio design in reducing induced drag 120 and he was quick to identify that a truss-structure
supporting the wing would enable larger wing spans to be achieved as a result of the loads
alleviation provided by the strut 121 . His demonstrator aircraft, the HD-10 (see Fig 2.7(b)), had a
32.5 aspect ratio wing supported by a strut and jury-strut truss-structure and was so successful
it led to the development of two medium range prototype aircraft designated the HD-31 and
HD-32, Fig. 2.7(c) and Fig. 2.7(d) respectively. These aircraft (and their successor the HD-34 - Fig.
22
2.2. THE BRACED-WING AIRCRAFT
23
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
2.7(e)) were of a conventional design apart from the braced wing which featured a flat stub-wing
section followed by a primary strut and two supporting jury-struts 121,122 . Dubois was aware of
the increased interference drag that resulted from the truss structure and he made a special
effort to design the struts to be as thin as possible, as well as incorporating pre-twist into the
struts to ensure minimum drag at cruise 121 . Managing the interference drag due to the truss is
still a major concern for modern TBW designs and is the subject of intense research .
The Hurel-Dubois company did not produce any further aircraft after the HD-34, although
a jet-powered successor to the HD-34, the HD-45, was proposed but never materialised. A
collaboration with the Miles Aircraft Company led to the construction of a HARW variant of
the Miles Aerovan designated HDM.105 (Fig. 2.7(g)), which combined the fuselage of the Miles
Aerovan with the wings from the HD-34 123 . The design of the HDM.105 was then incorporated
into the SC.7 Skvan (Fig. 2.7(h)) which was manufactured by the Short Brothers aircraft company
based in Belfast. Two variants of this aircraft, the Short 330 and Short 360, were in service from
1947-1992 and 1981-1991 respectively and operated as regional airliners 124 . Perhaps the most
famous SBW aircraft is the Cessna 172, an early version of which is shown in Fig. 2.7(f). This
light aircraft has many similarities with the early HD-10 design, however, the wing span and
aspect ratio are much smaller; this is perhaps due to concerns about the handling qualities of a
large and flexible wing . The reduced wingspan means the length of the supporting strut is much
shorter and therefore additional jury-struts are not required to prevent buckling in the primary
strut. Another well-known SBW aircraft is the de Havilland Canada DHC-6 Twin Otter (Fig.
2.7(i)), which is similar in design to the Cessna 172 but has a higher payload capacity.
Table 2.1 shows a summary of the performance metrics for recent and historic braced-
wing aircraft. In the interest of brevity, the Cessna 172 and DH6 Twin Otter are taken to be
representative of the numerous light-/utility-aircraft that have adopted the braced-wing design,
also, the HDM.105 has been neglected as it used the same wings as the HD-34. Examining the
data in Table 2.1 shows that, with the exception of the early HD- series, most braced-wing aircraft
have used a SBW design. This is a direct consequence of the smaller wingspan which results
in a reduced strut-length with a higher critical buckling load. As the primary purpose of the
24
2.2. THE BRACED-WING AIRCRAFT
jury-struts is to manage buckling in the primary strut there is no need to provide additional
bracing at the expense of all-up mass. Also, more recent braced-wing aircraft have had much
lower aspect ratios than the HD- variants, indicating that the full potential of the braced-wing
design is still to be realised. Furthermore, the data in Table 2.1 shows without exception all
previous braced-wing aircraft have been low-speed, low-to-mid range variants with propeller
propulsion systems. This would likely explain why this concept has not seen application in
the traditional single- and multi-isle markets, which involve longer routes flown at transonic
speeds. It would certainly have been possible for a braced-wing aircraft to have been designed
for this mission profile, however, it is likely that the concept was ruled out at an early stage
due to uncertainties regarding the interference drag from the truss and structural complexity of
integrating the truss-structure with the wing and fuselage.
Despite the fact that the braced-wing concept was not widely adopted during the jet-era there
was still continued interest from the research community. Several design studies took place
during the second half of the 20th century that investigated whether a braced-wing could offer
performance benefits over the traditional cantilever design.
The first of these studies was conducted by Werner Pfenninger while he was working at
Northrop during the 1950s 130 . Pfenninger’s research was primarily aimed at the design of a fully-
laminar flow aircraft, nonetheless, he recognised that supporting the wing with a truss-structure
would enable his designs to achieve a higher aspect ratio, which would have the dual benefit of
extending the laminar region along the aerofoil and also reducing drag 131 . Figure 2.8 shows one
of his concepts for a fully-laminar flow aircraft with a truss-braced wing.
In the 1960s Boeing conducted a study comparing the initial design of a large-span military
25
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
transport aircraft in a cantilever and SBW configuration 132 . It was found that the strut-braced
wing configuration could "offer the potential of lower gross weight, lower empty weight and reduced
fuel consumption" 132 , however, this was assuming that further improvements in the aerodynamic
design of the strut-joint regions could be made. A follow up study in 1978 by Jobe et al. 133 found
that the SBW should not be ruled out based on the result of a conceptual design study. In their
report, they highlight the usefulness of the SBW in avoiding a wing-tip ground strike during a
taxi-bump load case as well as the reduction in loads at locations inboard of the strut. However,
in this study the strut was attached to the wing load reference axis and so there is no reduction
in the wing torque due to the strut reaction force. Such an effect is possible if the strut-wing
attachment point is moved towards the leading edge, a benefit that later studies would go on to
highlight 134 .
Turriziani et al. 135 studied the fuel efficiency of a business jet with an aspect ratio 25 SBW
compared to a conventional cantilevered wing. It was found that a SBW could offer fuel savings
of up to 20% and that for the same wing planform the strut-braced wing reduced the total wing
weight. Some drawbacks were also highlighted, including uncertainties about the manufacturing
cost of the SBW and a concern that the strut-braced variant would be less productive as a result
of flying at a lower cruise Mach number. Smith et al. 136 considered the application of the SBW
concept to a high-altitude research aircraft. They showed that the SBW configuration had a
31% increase in range compared to the cantilever baseline aircraft and that this was a direct
consequence of a reduction in wing weight due to the loads alleviation provided by the strut.
Moreover, both of these studies mention the benefit of designing the struts to produce lift, thereby
increasing the total L/D ratio. Although, the aerodynamic loads were calculated using simple
methods and so it is likely that the effectiveness of using lifting struts has been overstated.
Finally, Park 137 evaluated the possibility of block fuel saving for the preliminary design of a
short-haul strut-braced wing aircraft. It was found that although the SBW did offer a reduction
in wing weight and induced drag the overall block fuel savings were not significant due to the
increased parasitic drag from the strut. The increase in strut drag was a result of sizing the strut
for buckling considerations, therefore it is possible that if the strut chord length could be reduced
by alleviating the buckling then further benefits could be possible. Park also noted that strut
flutter should factor into the design considerations, something which had not been identified
before.
Through all of these studies it was gradually understood that a braced-wing design could
offer many benefits over a cantilever configuration. The increased wing-span leads to a higher
aspect ratio design which provides a reduction in lift-induced drag and the loads alleviation
from the truss enables the inboard wing section to be designed with a reduced wing profile and
chord length, which in turn leads to a reduction in parasitic and wave drag. Although various
individual benefits were emphasised none of these preliminary studies had considered a truly
multidisciplinary approach to their designs, nor had flutter or gust loads been considered.
26
2.2. THE BRACED-WING AIRCRAFT
The initial design studies that took place throughout the second half of the 20th century indicated
the need for a multidisciplinary approach to the design of a strut- or truss-braced wing. With
this in mind NASA commissioned Virginia Tech University to undertake a major study with the
aim of applying MDO techniques to the design of a SBW aircraft 138 . The core focus of the project
was to investigate mutli-disciplinary design and optimisation 139 , structural optimisation 140
and aeroelastic stability 141 of strut-braced wings, as well as the numerical prediction of strut
interference drag 142 . The study used a mission profile with a range of 7500nm at Mach 0.85
with 325 passengers in a three-class configuration to size both the strut-braced wing aircraft
and the baseline cantilever design 140 . NASA Langley’s Flight Optimization System (FLOPS) 143
was used to size the overall aircraft and determine performance metrics. Additionally, a Boeing
777-200IGW was used for overall performance comparisons. The SBW configuration included a
single supporting strut for each wing and two engine configurations were considered: tip-mounted
and under-wing. This study was the first to use a MDO approach for the design of a SBW aircraft
and the use of Computational Fluid Dynamics (CFD) tools to analysis the interference drag from
the strut represented a major milestone in the research of braced-wing aircraft. For this reason,
a brief summary of each thesis is provided along with some key deductions.
The MDO focussed on the wing only, with the fuselage and tail geometry fixed for each
design. The objective function was the aircraft take-off gross weight (TOGW), however, the TOGW
optimised SBW was then re-optimised for minimum total cost, maximum seat-miles per gallon
and maximum L/D, with the results showing little variation in the optimised values of the
design variables for the different cost function. Whilst the MDO framework was comprehensive
in its scope, the individual modules relied upon relatively simple analytical models and design
rules which were loosely physics-based. Also, aeroelastic constraints were not included in the
optimisation problem but calculations carried out offline verified that the optimum solution
was free from flutter. Overall, the MDO study determined that the optimum configuration for
all cost functions considered was a strut-braced wing with under-wing engines, which largely
corroborated the findings of earlier studies 130,132,133,135–137 .
The structural optimisation considered three load cases: a 2,5g and -1g trim case and a 2g taxi
bump load case, no gust load cases were considered. Naghshineh-Pour used a double-plate model
to calculate the bending weight of the wing whilst structural loads were determined using linear,
Euler-Bernoulli beam theory and the aerodynamic loading used a piecewise representation of
the loads generated by the MDO process. The structural components were sized using the fully-
stressed design criterion. The structural optimisation predominantly focussed on the wing/truss
structure and relied on FLOPS to determine the parameters for the rest of the aircraft. It was
found that the optimised SBW aircraft had a 9.2% saving in TOGW and a 15.4% reduction in fuel
weight compared to the fully optimised cantilever wing 140 .
The drag investigation carried out by Tétrault used the FUN2D & FUN3D 144,145 flow solvers
27
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
28
2.2. THE BRACED-WING AIRCRAFT
associated with this concept, such as the effects of free-play, impact loading and other forms of
nonlinearity, it represents a possible solution to the problem of strut buckling. It is interesting to
note that no further details are offered on this strut device and so it is highly likely that it did
not move past the conceptual stage. Nonetheless this is the first mention of a loads alleviation
device being used in a SBW and therefore it is important to highlight this concept within the
context of applying vibration absorbers to a braced-wing design.
In summary, the NASA/Virginia Tech study established some of the key design challenges
facing the braced-wing concept, including strut buckling, aeroelastic stability and reliable predic-
tion of strut-wing interference drag. Whilst it was one of the first MDO studies of a SBW it still
had a few drawbacks:
1. The studies only considered a single-strut configuration which led to an overemphasis on
the problem of strut-buckling. In more recent MDO studies it is generally accepted that
it is simpler to add additional truss-members (jury-struts) to prevent the primary strut
buckling, rather than increase the strut dimensions or design a novel solution.
2. Many of the MDO modules relied on simplified analytical models. For instance, the struc-
tural model used a piecewise loads assumption and linear beam theory and the strut
axial force was included as a design variable, as opposed to being calculated by solution
of the structural equations. A more refined analysis in each of the disciplines (structures,
aerodynamics, aeroelasticity) would represent a significant improvement.
3. Aeroelastic constraints were not considered at all during the MDO process. More recent
studies 47,109 have shown the flutter speed to be highly sensitive to the truss properties
which implies any MDO process must incorporate aeroelastic considerations.
The Subsonic Ultra Green Aircraft Research project (SUGAR) is a collaboration between NASA,
Boeing and several American universities, including Virginia Tech and Georgia Tech, it be-
gan in 2005 and continues to this day. The principal aim is to explore novel aircraft concepts
and develop innovative technologies that will improve aircraft performance and reduce noise
emissions in-line with the goals set out by the global aviation community. A significant body of
research has been conducted including: route analysis, trade studies, high fidelity simulations
and wind tunnel experiments, the details of which are contained within several NASA technical
reports 4,5,28,30,146–149 .
Using the results of the Boeing Commercial Market Outlook 150 , the SUGAR team investigated
aircraft concepts that could target the single-aisle, short/medium range market with a minimum
cruise speed of Mach 0.7. A reference aircraft, ‘N’, was defined which is loosely based on a
B737NG, and the next-generation concepts had a potential entry into service of 2030-2035, hence
the name for this group of concepts is SUGAR N+3. An overview of the main phases of the project
is provided below:
29
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
1. Phase I - This phase focussed on: initial market and route analysis, definition of the
initial concepts, trade-off studies, initial sizing and aerodynamic design and technology
ranking. Three novel N+3 concepts were identified at the end of this phase:
• SUGAR High - A high-span, braced wing aircraft with conventional turbofan engines.
• SUGAR Volt - A high-span, braced wing aircraft with a hybrid gas turbine-battery
electric propulsion system.
• SUGAR Ray - A hybrid-wing body6 aircraft.
Most notably, all three N+3 concepts have large L/D ratios in an attempt to reduce induced
drag during cruise. Also, a wing-fold mechanism was proposed for the SUGAR High and
SUGAR Volt concepts to allow the aircraft to fit within the airport gate limits. Further
details can be found in Bradley and Droney 5 and Greitzer et al. 28 .
2. Phase II - The second phase involved a more detailed investigation into several aspects
of the N+3 concepts from Phase I.
• Volume I - An exploration of the TBW concept including: MDO of the TBW concept
using Virginia Tech’s framework, detailed sizing and aerodynamic design of the
final concept and wind-tunnel testing incorporating flutter suppression and gust
load alleviation technologies. Several important lessons were identified, including
the prevalence of gust loads in the critical loads envelope and the importance of
using pre-stressed modeshapes in the calculation of the flutter speed, as proposed by
Sulaeman 141 . This SUGAR study represents the most comprehensive investigation so
far into a modern TBW aircraft and provides the basis for much of the work in this
thesis. Further details can be found in Bradley et al. 30 .
• Volume II - An investigation of the hybrid-electric power system including noise
analysis, preliminary design of the power system and sizing of the gas turbine engine.
Further details can be found in Bradley and Droney 147 .
• Volume III - A detailed discussion of the aeroelastic testing and novel techniques
used to simulate the flutter behaviour of the TBW model. The aeroelastic testing was
a significant undertaking and revealed many interesting behaviours, including the
presence of LCOs due to both structural and aerodynamic nonlinearities 100 as well as
evidence that the tension/compression of the strut strongly influences the aeroelastic
behaviour. Flutter suppression was achieved through the use of control laws which
also provided some gust load alleviation, although the controllers were not designed
for this purpose. Further details can be found in Bradley et al. 148 .
3. Phase III - The third phase of the SUGAR project is currently ongoing and as such
there is not a significant amount of detail available. A recent technical report by Kapania
et al. 149 presents a MDO and cruise Mach number study where a comparison is made
between a SBW and TBW aircraft optimised for Mach 0.7 and Mach 0.8. It is found that
30
2.2. THE BRACED-WING AIRCRAFT
at Mach 0.7 the SBW is the more efficient aircraft, whereas at Mach 0.8 the TBW is more
efficient. This observation is attributed to flutter, which for a SBW at Mach 0.8 restricts
the wingspan leading to an increase in induced drag, however, at Mach 0.7 a single strut
can provide sufficient stiffness to alleviate flutter as well as manage loads. This is slightly
in contradiction to the findings presented in Bradley et al. 30 , however as the MDO code
has been updated since the earlier studies it is likely that the most recent results are more
accurate. Also, it is noted that for longer range mission profiles the TBW is always superior
to a SBW. Details are also provided of a transonic flutter prediction method that uses a
ROM based on RANS simulations to correct the lift and moment coefficients for transonic
behaviour 98 . It is shown that this method provides good correlation with the observed
experimental behaviour from the SUGAR wind tunnel tests, however, it should be noted
that the transonic flutter method was not implemented into the MDO routine although this
is identified as future work.
Clearly the SUGAR project is a significant research undertaking and represents the current
state-of-the-art in terms of next generation aircraft research. It is testament to the significant
potential of the truss-braced wing concept that almost all N+3 concepts use a TBW to achieve
optimum performance.
Several researchers have continued to investigate the multidisciplinary design and optimisation
of TBW and SBW aircraft. With the advent of high-powered computing researchers have been
able to include higher-fidelity analyses in the initial MDO process, which has led to an increased
focus on the validity of some of the performance claims made by previous researchers. Many
research groups now have their own MDO processes, including groups at Virginia Tech, Stanford,
ONERA, Georgia Tech, Delft and Michigan, and an overview of their recent work is provided in
this section.
Virginia Tech have continued to build on their optimisation expertise, thanks in part to their
involvement in the SUGAR project. The MDO framework that is used in their TBW studies is
detailed in Gur et al. 151 and is an upgrade on the initial work carried out by Gern et al. 138 . The
upgrade included the implementation of a finite element solver to determine the loads in the
truss structure and also a global buckling check for all of the parts of the wing structure that
are subject to compressive loads. The truss members are modelled as simple rod elements and
the wing uses symmetric beam elements (CBAR). The torque box is a simple rectangular design
and can only model symmetric sections. The optimisation framework includes aerodynamics,
propulsion and weights modules. The aerodynamics module accounts for various elements of the
drag build-up, including friction, interference and form drag. Most of the calculations are based
on empirical methods such as the Korn equation, or in the case of the strut interference drag,
use response surfaces in order to save time during the optimisation. As with the earlier work,
31
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
aeroelastic constraints are not present in the optimisation procedure and gust load cases are
modelled using the Pratt method for gusts. Further details can be found in Gur et al. 151 . Bhatia
et al. 109 used the MDO process described in Gur et al. 151 to conduct a parametric study on the
structural and aeroelastic characteristics of TBWs. Several wing configurations are considered,
including a SBW and TBWs with one, two and three jury struts. In the report only linear flutter
analysis is used and the geometric stiffening effect due to the strut reaction force is included. The
authors present several important observations, including that a difference in sweep angles for
the primary strut and the wing can have a favourable effect on the flutter speed. On the whole
it is shown that the flutter performance of the TBW configurations is significantly better than
the SBW configurations. The SBW concept has also been applied to a medium-range transonic
aircraft with a mission profile similar to the Boeing 737-800NG in Meadows et al. 45 , whereas
previous optimisation studies by Virginia Tech had focussed on a ‘777-200ER-like’ aircraft with a
longer mission profile. Meadows et al. 45 found that the SBW and TBW configurations provided
a similar benefit in terms of reduced TOGW and fuel burn, however there was little difference
between the optimised SBW and TBW designs. For both the SBW and TBW the configurations
with under-wing engines slightly outperformed the fuselage-mounted engines. It is also noted in
this paper that most of the previous studies that use the Virginia Tech MDO framework tend
towards designs that incorporate lifting struts. It is suggested that this mechanism acts as a
form of load alleviation and helps to drive the decrease in sized wing weight.
As an extension to this work Mallik et al. 47 considered the effect of including flutter con-
straints in the MDO process for both a medium and long range mission profile. The flutter
analysis used the method presented by Sulaeman 141 and considers only linear flutter analysis
in the MDO process, although nonlinear flutter analysis was conducted ‘offline’ to verify the
results of the MDO study. It was found for the medium-range case adding the flutter constraint
meant that the TBW did not offer a reduction in TOGW compared to the cantilever configuration,
however, it still offered a 6% reduction in fuel consumption. For the long-range mission the
TBW design offered significant benefits in term of both TOGW and fuel-burn over the optimised
cantilever configuration. The flutter constraint was found to be the determining factor in most
of the TBW designs however it was not active for the cantilever design. The nonlinear flutter
analysis returned a flutter velocity that was 4.5% lower than the linear analysis, however, the
linear pre-stressed flutter analysis was deemed acceptable for the MDO process as actual flutter
mechanism was unchanged by the nonlinear effects. Recently Wells 152 presented an MDO study
which uses the Virginia Tech framework to examine four aircraft configurations: a low wing
cantilever, a high wing cantilever, a strut-braced wing and a single jury TBW. Only gross wing
parameters were included in the optimisation and no technology factors were assumed. Wells
found that for the current technology level the cantilever configurations had the lowest TOGW
followed by the SBW and then the TBW. This result is in stark contrast to previous findings,
although this is perhaps not surprising given the top-level approach that was adopted. As with
32
2.2. THE BRACED-WING AIRCRAFT
the Boeing study 133 , it is apparent that the benefits of the TBW only become evident when a
more sophisticated analysis is considered.
In addition to the ongoing MDO work at Virginia Tech there are several other research groups
taking a similar approach. The ALBATROSS project at ONERA 46 is aimed at evaluating the
benefits of the SBW concept with respect to a reduction in TOGW and also the potential to enable
natural laminar flow. The project is utilising high fidelity CFD and detailed structural design as
well as MDO to conduct parametric studies of the SBW concept. It is interesting to note that so far
the ALBATROSS project has only considered a SBW despite its known tendency to be subject to
strut buckling. In their report 46 a curved strut is proposed in order to induce ‘controlled-buckling’
which will help alleviate some of the strut loads and therefore reduce the weight penalties
associated with the buckling constraint. Variyar et al. 153 from Stanford University presents a
multi-fidelity optimisation framework for a SBW that uses finite element-based structural sizing
coupled with vortex-lattice and CFD aerodynamics to perform design and optimisation studies.
Their optimisation of a SBW did not include buckling or aeroelastic constraints however this is
identified as future work. In comparison to the work carried out by the MDO group at Virginia
Tech this study uses much higher-fidelity tools however it does not encompass all of the aircraft
disciplines so its application is not truly multidisciplinary. Recently Rajpal and De Breuker 154
presented a paper where the Delft University MDO tool PROTEUS was used to optimise the
wing mass of a fixed-planform strut-braced wing using composite materials and including fatigue
effects. In this study over 5300 aeroelastic loads cases where considered, including dynamic gust
encounters across a number of flight points. This is considerably more than in the SUGAR project
and revealed that the wing structure is sized by both static and dynamic loads, reinforcing the
need to consider dynamic gust loads at the conceptual design stage.
An important point to note is that in all of the MDO studies to date no effort has been made
to consider the design of the joint between the truss-structure and the wing/fuselage. This is
understandable as many of the MDO studies have utilised low-fidelity beam models where this
level of detail is not necessary to facilitate modelling a joint. However, given the sensitivity of the
flutter speed to the precise stiffness and mass distribution of the model it is important to account
for any additional structural mass resulting from the presence of a joint. Where high-fidelity
models have been used, such as in the work by Michigan University (see Section 2.2.6), an elegant
blend region is created between the truss and the wing volumes which is then populated by
stringer and skin panels during the meshing stage. However, this approach is more akin to a
fully-fixed joint, whereas the use of pinned-pinned joints has been shown to be more advantageous
in MDO studies 30 .
33
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
The presence of the truss-structure makes accurately modelling the flow field an extremely
difficult task and one that is ill-suited towards rapid, approximate techniques such as strip-
theory or panel methods. Most of the MDO studies outlined in the previous sections have used
the vortex-lattice method with additional corrections to include the effects of the different drag
components acting on the wing 155 . Whilst this approach is acceptable for an MDO study the
aerodynamic loads acting on the wing have been shown to be vastly different 156 . Over the years
various attempts have been made to determine the optimum level of aerodynamic modelling for a
TBW. Seber et al. 157 detailed an upgrade to the aerodynamics module used by the Virginia Tech
MDO group which enables a more accurate prediction of the transition point by incorporating
CFD data via a response surface. This upgrade allowed for a better estimate of the friction
drag acting on the SBW and led to a reduction in TOGW and fuel weight. Duggirala et al. 112
conducted a study to analyse the interference drag due to strut-strut interaction. The aim was to
improve upon the previous work by Tetrault 142 which stated that the interference drag reaches
a minimum when the strut intersects at 90°. This had led to the adoption of the engineering
rule: "if the angle of the strut is less than 45°add a vertical offset" 142 . This approach was well
established and widely implemented in several SBW designs, however, Duggirala et al. 112 found
that there is an unexpected drag rise when the strut intersects the wing at 90°. Updated response
surfaces were generated as part of this study and have since been used in further MDO studies.
A study of the transonic aerodynamics of a SBW by Ko et al. 158 revealed that under certain
conditions the area enclosed by the strut, strut-offset and wing can act as a nozzle and lead
to increased drag. It was shown that careful design of the strut upper surface and wing lower
surface can lessen the strength of the shocks and even eliminate them. It is also noted that
this analysis used inviscid flow solvers and that including viscous effects is likely to reduce the
nozzle effect. Finally, Ting et al. 156 proposed a method of aerodynamic analysis for a TBW that
used a vortex-lattice superposition approach. They found that using the vortex-lattice method
without applying superposition led to an underestimating of both the C L − C D and C L − α slopes,
however improvements could be made if a superposition approach was adopted. Ting observed
good agreement between the VLM-superposition results and CFD results and suggests that
the superposition approach could still be used in the conceptual analysis as it is based on the
vortex-lattice method.
Recently, the optimisation techniques developed at Michigan Univesity 159 have been applied
to the aerodynamic shape optimisation of TBW and SBW 160–162 . Here, the shape optimization
solved a lift-constrained drag minimization problem using a RANS (Reynolds-Averaged-Navier-
Stokes) CFD solver for the aerodynamic calculations. It was found that a drag reduction of 28%
and final a lift-to-drag ratio of 25.3 is achievable compared to the baseline SUGAR 765-095-
TS configuration, however in this study aerolastic effects were neglected, this is a significant
drawback as aeroelasticity would have a considerable impact on the 1g jig-shape and resulting
34
2.2. THE BRACED-WING AIRCRAFT
aerodynamic drag. Work has also been ongoing at NASA to investigate the aero-structural
modelling of a TBW travelling at transonic speeds. Xiong et al. 163 studied wing-strut interference
effects using the FUN3D flow solver and found that the presence of a strut leads to a suction
peak on the lower wing surface which influences the aerodynamic forces and moments and the
location of the aerodynamic centre. A parametric study of Mach and Reynolds number was also
carried out and it was found these interference effects are more pronounced as the Mach number
increases with the Reynolds number having less effect. This work led to the development of
aerodynamic correction factors which were then used in the work by Nguyen et al. 164 . Here, an
aero-structural model was developed which includes the corrections for transonic and viscous
flow, as well as geometric nonlinearities resulting from the strut tension force. A dynamic finite
element model was also generated using a beam-reduction technique to reduce the 3D NASTRAN
GFEM to a 1D VSPAERO model which was used for flutter calculations. A jig-twist optimization
was conducted in order to generate a model that matched wind tunnel data from the recent TBW
aeroelastic wind tunnel testing 148 with excellent agreement found between the VSPAERO model
and the experimental data.
The increased level of interest shown in TBW aircraft by the SUGAR project has inspired
other research groups to begin investigating this concept. Groups at research centres such as
DLR 165 (Germany) and ONERA 46 (France) have developed tools for investigating the conceptual
design of braced-wing aircraft, their aim being to assess whether the braced-wing is in fact
favourable compared to the classic cantilever design. Many of these tools have focussed on using
physics-based methods to generate accurate estimates for the strucutral weight of the braced
wing, in contrast to a non-physics-based method which would rely on regression analysis based
on historical data 166 . Chiozzotto 165 developed a wing weight estimation method that uses a
low-order structural representation, VLM aerodynamics and analytical models of the wing load-
carrying structure. A mass-optimization was carried out and it was found that the SBW has a
18% lower wing weight than a cantilever for the same planform, however this does not account
for the fact that a fuel-burn optimized cantilever wing would likely have a different planform
than a SBW 30 . A similar study is carried out by Locatelli et al. 61 but using a slightly different
formulations for the structure, aerodynamics and loads-carrying structure. In both studies the
tools are benchmarked against historical wing-weight data and good correlation is found, which
is then used as justification for these tools to be utilised for the design of braced-wing aircraft.
The reduced structural weight of the braced-wing concept inevitably leads to a more flexible wing
due to the reduced thickness of the structural components. The flexibility of the wing gives rise to
nonlinear phenomena such as LCOs 100 and the transonic flutter dip 98 . Recently, several papers
35
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
have been published dedicated to the nonlinear aeroelastic analysis of truss-braced wings, their
aim being to model these effects and assess their impact on the overall aircraft design. The early
work by Sulaeman 141 identified that the truss-structure generates additional in-plane loads
in the wing structure which affects the flutter behaviour of SBW and TBW. The precise cause
of this is modified natural frequencies due to the effects of pre-load and geometric stiffness in
the structural stiffness matrix resulting from in-plane loads. Sulaeman’s investigations showed
that including these effects leads to a reduction in the flutter speed of anywhere between 0-
10% depending on the strut spanwise location. This technique has since been incorporated into
the aeroelastic analysis conducted during the SUGAR project. It was found that this method,
combined with corrected pressure data from offline CFD analysis, could provide an accurate
estimation of the flutter speed and was validated as part of the SUGAR aeroelastic wind tunnel
test campaign 148,167 . Details of the NASTRAN implementation of this method are provided in
Zhao et al. 168 .
What is clear from all of these studies is that flutter is dominant in the design of braced-wing
aircraft and must be mitigated for if further performance benefits are to be realised. Regarding
flutter suppression in TBW, Butt et al. 169 investigated a method for alleviating flutter by using a
ballast mass of between 2% and 8% of the wing mass placed at 98% of the wing span. It was found
that the flutter speed was relatively insensitive to the addition of a ballast mass, with the flutter
speed being extended by only 1% for the largest mass. Active control has been trialled during the
SUGAR aeroelastic wing tunnel tests where it was found to extend the flutter boundary by up to
8%, although the exact amount of flutter suppression was dependent on the angle-of-attack. In
these same tests the flutter controller was tested for gust load alleviation where it provided a
25% decrease in the magnitude of the wing tip acceleration for frequencies between 9.6-10Hz 167 .
In similar tests on a conventional aircraft configuration, Marchetti et al. 170 were able to extend
36
2.2. THE BRACED-WING AIRCRAFT
the flutter point by up to 6% using a control law for the port and starboard ailerons. It is
important to note that to-date neither of these methods have been included in the optimisation
of a braced-wing aircraft, therefore their benefits to the aircraft design in terms of weight
cannot be quantified. Regarding gust load alleviation via active control, potential reductions
in wing root bending moment of up to 70% have been demonstrated for conventional aircraft
configurations 171–173 although the precise benefit of an active control scheme is dependent on the
model behaviour and the controller design. For example, bending moment reductions between
7-40% were demonstrated for a LQG controller with a variety of design parameters 174,175 , and
for the case where the control surface effectiveness is taken into account the performance of the
GLA system can be significantly reduced 176 .
Below is a brief summary of the main points that have been identified from the papers reviewed
in this section:
• The braced-wing aircraft is not a new concept, however existing examples of braced-wing
aircraft are all subsonic and propeller-driven. Hence why the main focus in the literature is
the design of a jet-powered braced-wing aircraft that can compete with a cantilever design
when flying at transonic speeds.
• The general state of MDO analysis for TBW aircraft is well advanced with several research
groups dedicated to this task. Numerous MDO studies have shown that a TBW can provide
a benefit in terms of aircraft mass and fuel-burn over the traditional cantilever design.
• Aeroelastic phenomena such as gust loading and flutter have been identified as critical
design drivers in SBW and TBW, thus placing a limit on any practical efficiency savings
that can be realised by adopting a braced-wing design.
• Accurate flutter prediction for a TBW operating at transonic speeds requires a model
capable of accounting for both structural and aerodynamic nonlinearities. Structural non-
linearities can be readily accounted for using a modified stiffness matrix based on pre-load
and large-deflection analyses, however nonlinear aerodynamics requires correction factors
derived from RANS CFD analysis which incurs a significant offline time penalty.
• Flutter suppression of the SUGAR TBW via active loads control was achieved during
wind-tunnel testing, where it was also found to provide some gust loads alleviation. Flutter
suppression in a SBW using ballast masses has also been investigated but this was not
very effective. Neither technology has been included in the optimisation of the aircraft
structure and so the benefits of these technologies in terms of mass and fuel-burn cannot
be quantified.
37
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
Vibration suppression devices have been used for decades to control systems which experience
high levels of vibration, including but not limited to: vehicle suspension systems 177 , earthquake-
resistent structures 178 , suspension bridges 179–181 and power transmission lines 182,183 . Particu-
larly high-profile examples of vibration suppression devices in engineering structures include
the tuned mass damper in the Tapei 101 tower 184 and the fluid viscous dampers retrofitted to
the London Millennium Footbridge 181 . These devices can take a wide variety of forms, from the
classical mass-spring-damper formulation 178 to magnetorheological 185–188 and other fluid-based
devices 189–191 , and can involve passive 191,192 , semi-active 193,194 or active controls 195–197 . A com-
prehensive review of existing vibration suppression technologies is beyond the scope of this thesis
and instead the reader is directed to a number of review articles 177,178,185–189,193,195,197 .
The rest of this section is structured as follows: first a brief overview of the theory of vibration
suppression is given in order to familiarise the reader with the basic physical principles. Next a
description of mechanical network design and network synthesis is provided, with a specific focus
on the use of inerter-based devices. Finally, some examples of vibration suppression devices in
aerospace structures are detailed and examples relating to truss-braced wings are highlighted.
Vibration suppression devices are common in many structures which undergo large motions
or are subject to vibrations which are detrimental to the overall performance or health of the
structure. The principal function of these devices is to remove energy from a structure, or parts
of a structure, at a set of frequencies that are critical to the safe functioning of the system, i.e.
resonant frequencies. In general, there are two mechanisms for achieving this:
1. Transfer energy away from the system - Energy can be removed from the system by
converting kinetic energy into thermal energy which is then dissipated. This is typically
achieved by introducing friction forces between mechanical components, deforming an
elastomeric material, or, in the case of a fluid-based device, compression via a piston and
generating turbulent flow by channelling fluid through an orifice.
2. Transfer energy within the system - Another option is to transfer kinetic energy away
from the parts of the structure where vibrations need to be minimised. This can be achieved
by modifying the properties of the primary system in order to tailor the modal response,
such as by adding mass or changing the stiffness of a particular component. Alternatively,
the primary system can be augmented by a secondary system which is precisely tuned
to vibrate at a set of pre-defined frequencies, thus transferring kinetic energy from the
primary to the secondary system. The first approach can be described as frequency tuning 198
whereas the second is the classic dynamic vibration absorber 199 , also referred to as the
Tuned Vibration Absorber (TVA).
38
2.3. VIBRATION SUPPRESSION DEVICES
x1
k
c
P(t) = P0 sin (ω t)
The fundamental theory of vibration absorbers is described in some detail in Den Hartog 200
and so only a brief overview is provided in this work to help familiarise the reader with the
basic concepts. First, consider the classical single degree-of-freedom (DOF) system shown in
Fig. 2.10. For a system comprising of a lumped mass connected to a fixed point by a spring
and dashpot-damper7 with linear stiffness and damping coefficient respectively the equation of
motion is
m ẍ1 + c ẋ1 + kx1 = P(t) (2.2)
where m is the mass, c is the damping coefficient, k is the spring stiffness, P is some applied
force, x is the translational displacement of the mass and t is time. Table 2.2 shows the four
principal cases for the 1DOF system and their respective solutions8 and Fig. 2.11 shows some
qualitative results for this system. Examining Figure 2.11(a) shows that for Case 1 the system will
p
vibrate without decay at its undamped natural frequency, ωn = k/m . Clearly, this is not realistic
as damping is present in all physical systems, hence the inclusion of a damper which causes the
system to decay (Case 2). Here, the rate of decay is a function of the damping ratio, ζ, which is
defined as the ratio between the damping coefficient and the critical damping value (ζ = c/2mωn ),
p
and the period of the decay is related to the damped natural frequency ωd = ωn 1 − ζ2 .
Concerning harmonic forced vibrations of the 1DOF system at a discrete frequency, Fig.
2.11(b) shows that when damping is excluded (Case 3) there exists a forcing frequency where the
response of the system tends sharply to infinity - denoted mathematically as the denominator
term 1 − (ω/ωn )2 tending to zero which occurs when the forcing frequency is equal to the natural
frequency of the system (ω = ωn ). Clearly an infinite response is undesirable and so efforts must
be made to avoid these frequencies or manage the response of system such that resonance does
not occur. Introducing a damper (Case 4) shows that as the damping ratio is increased the
magnitude of the resonant peak decreases until it is almost completely minimised, whilst the
phase angle, shown in Fig. 2.11(c), has a shallower slope as it transitions through the resonant
point, φ = π/2, and a nonzero value at nonresonant frequencies. Most importantly, the addition of
7 A dashpot damper is one where the damping force is proportional to the relative velocity across the element.
Other types of damping, such as friction, are possible but will not be considered here.
8 Note - C and C are constants that define the initial conditions of the system.
1 2
39
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
a damper means the response must be finite for a nonzero damping ratio, thus demonstrating
the first mechanism of vibration suppression: transfer energy away from the system.
TABLE 2.2. Analytical solutions for the single degree-of-freedom system 200 .
π
10−1
ζ>1
10−2
Response
ζ=1 π
| x0 |
ζ<1 −3
φ 2
10
ζ=0
10−4 0
0 1 2 0 1 2
Time ω /ω n ω /ω n
(a) Time domain response
(b) Response magnitude as a function (c) Phase angle φ as a function of
of forcing frequency forcing frequency
F IGURE 2.11. Example response of the 1DOF system for the undamped (ζ = 0), under-
damped (ζ < 1), critically-damped (ζ = 1) and over-damped (ζ > 1) cases. System
parameters: m = 10kg, ωn = 6πrads-1 , ζ = [0, 0.1, 1, 2], P0 = 5N, C 1 = 1, C 2 = -2.
x1 x2
ka
k
m ma
ca
P(t) = P0 sin (ω t)
40
2.3. VIBRATION SUPPRESSION DEVICES
where the symbols m, c, k, P and x have their previous meaning and subscript a denotes
properties belonging to the dynamic absorber. In order to demonstrate the second mechanism of
vibration suppression we consider the case where the absorber damping coefficient is set to zero
(c a = 0). Assuming a solution of the form x i = a i sin (ω t) and using the dimensionless form of the
EoM as defined in Den Hartog 200 , p. 88 yields
x1 1 − ω2 /ω2a
=¡ sin (ω t)
1 − ω2 /ω2a 1 + k a /k − ω2 /Ω2n − k a /k
¢ ¡ ¢
xst
(2.4)
x2 1
=¡ sin (ω t)
1 − ω /ωa 1 + k a /k − ω2 /Ω2n − k a /k
2 2
¢¡ ¢
xst
where xst is the static deflection of the primary system (xst = P0 /k), ωa is the natural frequency
of the absorber ω2a = k a /m a and Ωn is the natural frequency of the primary system Ω2n = k/m .
¡ ¢ ¡ ¢
The solution for the primary system (x1 ) shows that the amplitude will equal zero when the
numerator 1 − ω2 /ω2a is equal to zero, which occurs when the natural frequency of the absorber
is the same as the forcing frequency. This occurs because the absorber spring generates a force
which is equal and opposite to the force applied to the primary system, as shown by the absorber
having motion x2 = −P0 /k a sin (ω t) when ω = ωa . Note, this result is achieved without the addition
of any damping elements, therefore clearly demonstrating the second mechanism of vibration
suppression: transfer energy within the system.
Whilst it is advantageous to know that a system can be augmented with a vibration absorber
in order to reduce vibrations at resonance conditions, for this knowledge to be of practical use
the parameters of the absorber must be better understood; specifically we are interested in the
values of absorber stiffness and mass which give optimum performance in the primary system.
For this study we consider the case where the absorber natural frequency is equal to the natural
frequency of the primary system, (ωa = Ωn ), and we introduce the concept of the absorber mass
ratio (µ) - which is the mass of the absorber expressed as a fraction of the primary system mass,
41
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
x1 1 − ω2 /ω2a
=¡ sin (ω t)
1 − ω2 /ω2a 1 + µ − ω2 /ω2a − µ
¢¡ ¢
xst
(2.5)
x2 1
=¡ sin (ω t)
1 − ω /ωa 1 + µ − ω2 /ω2a − µ
2 2
¢¡ ¢
xst
The resonant frequencies of the 2DOF system can be found by setting the denominator equal to
zero yielding a quadratic equation in ω2 /ω2a which has roots
µ µ´
r ³
2 2
ωr /ωa = 1 + ± µ 1 + (2.6)
2 4
where ωr are the resonant frequencies of the 2DOF system. Figure 2.13(a) shows an example
frequency-amplitude relationship for the case where the 1DOF system is augmented by a dynamic
vibration absorber. The two resonant peaks are shown and it is clear that the absorber causes
the primary system to experience negligible motion at the previous resonant frequency with the
absorber mass containing most of the kinetic energy. Figure 2.13(b) is the graphical representation
of Eqn. 2.6 showing the variation in the 2DOF resonant frequencies as a function of absorber
mass ratio. Clearly the 2DOF resonant peaks diverge away from the original resonant frequency
as the absorber mass ratio increases, this could be viewed as beneficial as it provides a wider
frequency-band around the original resonant frequency which is useful if the optimum operating
frequency of the system is near the 1DOF resonant point, such as in helicoter rotors 201,202 .
The disadvantage of the TVA is that the system now contains two resonant peaks with infinite
magnitude where before there only existed one - so the issue of resonance has not been solved,
merely shifted to a different part of the frequency spectrum.
101
1.2
10−1
ωr /ωa
1
| x|
10−3
2DOF - x2 0.8
2DOF - x1
10−5 1DOF
0 0.2 0.4 0.6 0.8 1
0 0.5 1 1.5 2 µ
ω/Ωn
(b) Variation in 2DOF resonant frequencies as a func-
(a) Undamped 2DOF amplitude response tion of absorber mass ratio
F IGURE 2.13. Example response for a 1DOF system augmented with a dynamic vibra-
tion absorber. System parameters: m = 10kg, ωn = ωa = 6πrads-1 , µ = 0.3, P0 = 5N.
42
2.3. VIBRATION SUPPRESSION DEVICES
To mitigate this, Den Hartog 200 suggested the addition of a damper element to the TVA,
as in Fig. 2.12, thus combining the two mechanisms of vibration suppression. This new device
is termed the Tuned Mass Damper (TMD) and its use is widespread throughout a variety of
engineering disciplines 178 . The derivation of the EoM of this system are slightly more involved
and so for brevity the reader is directed to Den Hartog 200 , p. 93 for the full derivation and
resulting solutions. Using Hartog’s approach it can be shown that the motion of the primary
system is dependent on only four parameters: the absorber damping ratio ζa = c a /2mΩn , the ratio
between the natural frequency of the absorber and the natural frequency of the primary system
f = ωa /Ωn , the absorber mass ratio µ and the forcing frequency as a fraction of the primary
system natural frequency g = ω/Ωn .
Figure 2.14 shows that for small values of ζa the amplitude of the resonant peaks decreases
as the work done by the damper replaces kinetic energy, however as the damping ratio increases
the two peaks coalesce and the amplitude increases again. This is because the relative motion
between the two masses decreases as the damping coefficient increases leading to a reduction
in work done by the damper. In the limiting case, i.e. as the damping ratio tends to infinity, the
damper element will approximate a rigid joint and the 2DOF system will acts as a single DOF
with total mass (m + m a ). Hartog hypothesised that there must exist some "optimum damping"
102
101 P
x1 /xst
100
1 2 3µ
2DOF - f = 1+ µ , ζa = 8(1+µ)3
2DOF - f = 1, ζa = 0.3
2DOF - f = 1, ζa = 0.1
1DOF - f = 1, ζ = 0
2DOF - ζa = 0
10−1
0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2 1.25
g
F IGURE 2.14. Example response for a 2DOF system augmented with a tuned mass
damper. System parameters: ζa = [0.1, 0.3], f = 1, µ = 0.05.
43
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
where the work done by the damper is maximised and the motion of the primary system is
minimised. To obtain this, we note the existence of two fixed-points, P and Q, which for a given
value of f all amplitude curves pass through irrespective of damping ratio. Hence the optimum
absorber mass ratio will be the one that minimises the amplitude of the points P and Q and the
optimum damping ratio will define a curve that is tangential to the higher of the fixed points, as
in Fig. 2.14. Deriving this relationship analytically yields 200
1 3µ
f= , ζ2a = ¡ (2.7)
1+µ
¢3
8 1+µ
which are the optimum absorber parameters for minimising the displacement of the primary sys-
tem. These closed-form equations are known as tuning rules and provide a convenient analytical
means by which the absorber parameters can be estimated. This remarkably simple approach
reveals a very important consideration in the design of a vibration absorber, which is that the
optimum absorber frequency does not necessarily equal the resonant frequency of the primary
system - although they are close. This is purely due to the definition of optimality that drives
the absorber design. For instance, minimising the motion of the primary system does not always
represent the most favourable (i.e. optimimum) operating conditions, in which case optimum
absorber parameters can be derived using a similar approach but for a different measure of
optimality - examples of which can be found in Warburton 203 and Krenk 204 . Furthermore, tuning
rules have also been proposed for absorbers attached to flexible structures 205 and for devices
which utilise a mechanical element known as an inerter 206,207 .
Finally, it must be noted that there are a number of disadvantages to the TVA and TMD
vibration suppression devices. Firstly, they require mass to be added to the system which in
general engineering applications is viewed as a negative. Secondly, a passive TVA or TMD can
only be tuned to a specific frequency of interest, which is not useful for structures that have
multiple resonant frequencies that must be targeted. Limitations such as these are reasons
why vibration suppression devices have been developed that can target multiple frequencies
simultaneously, either through active control or the use of more complicated devices which
comprise of multiple elements connected together to form a mechanical network.
Designing vibration absorbers using mechanical networks is an emerging research area which
caught increasing attention in recent years. Here, a vibration suppression device is idealised as
an abstract interconnection of mechanical elements arranged in a similar fashion to an electrical
circuit diagram, which has a transfer function Y representing the force-velocity relationship
between the terminals of the device9 . Figure 2.15 shows an example mechanical network with
various parts labelled for clarity. This representation is deliberate as a well known analogy exists
9 Using the terminology from Smith 208 , a one port element (such as a spring, damper or inerter) has two terminals
which are "a pair of nodes in a mechanical system to which an equal and opposite force is applied and can experience a
44
2.3. VIBRATION SUPPRESSION DEVICES
Y = F/∆v
c1 k1
b1
T1 T2
F F
k2 k3
v1 v2
k4
between electrical and mechanical networks which has been exploited for many years in order to
develop and analyse mechanical networks comprising of mass, spring and damper elements 184,210 .
In this analogy mechanical force is equivalent to electrical current and a mechanical spring
and damper are equivalent to an electrical inductor and resistor respectively. However, the
mass element causes this analogy to breakdown as it by definition has one of its terminals
grounded, that is, it can only generate a force according to the acceleration of its centre of gravity
(F inertial = − m ẍ), which is incompatible with the electrical capacitor which has no requirement
for one of its terminals to be grounded 10 . In 2002 Smith 208 introduced a mechanical element
known as an inerter which has the property that the force generated across the two terminals is
proportional to the relative acceleration across the terminals. This relationship is demonstrated
by the equation
d (v2 − v1 )
F inerter = b (2.8)
dt
where F inerter is the force generated by the inerter, b is a quantity termed inertance which is
measured in units of kg, d/dt represents differentiation with respect to time and (v2 , v1 ) are the
velocities at the inerter terminals. The formal introduction of the inerter element completed the
analogy between electrical and mechanical systems and fundamentally enhanced the design
methodologies and types of devices available to the mechanical network designer. Figure 2.16
shows the full analogy between electrical and mechanical elements.
The discipline of mechanical network design is concerned with establishing the network
configuration11 which provides optimimum performance. A review of the literature shows there
relative velocity".
10 Here, the term "grounded" does not imply a physical connection to the ground but instead refers to the mathe-
45
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
Mechanical Electrical
F F i i
k 1
Y (s) = s
Y (s) = Ls
v1 v2 v1 v2
dF
= k (v2 − v1 ) Spring di
= 1
(v2 − v1 ) Inductor
dt dt L
F F i i
Y (s) = bs Y (s) = Cs
v1 v2 v1 v2
F = b d (vdt
2 − v1 ) Inerter i = C d (vdt
2 − v1 ) Capacitor
F F i i
1
Y (s) = c Y (s) = R
v1 v2 v1 v2
Damper i= 1
(v2 − v1 ) Resistor
F = c (v2 − v1 ) R
are three possible approaches to the design of the network: structure-based, immittance-based
and structure-immittance 209 .
• Structure-based - The structure-based approach is the simplest of the three. In this
method the layout of the device is initially defined then the parameter values for each
element are selected based on certain performance criteria. This allows the complexity of
the device (i.e. how the elements are connected) and the parameter values to be tightly
controlled, however, the selected layouts are usually sub-optimum compared to those that
can be realised by the other two methods.
• Immittance-based - The immittance-based approach is concerned with finding the opti-
mum device transfer function before the corresponding network layout and element values
are determined using network synthesis theory12 . This is very much the opposite of the
structure-based approach where the device layout and the related transfer function are
predetermined but can lead to complicated layouts. The immittance-based approach allows
a full range of device layouts with a fixed number of each component type to be explored
although it is not possible to place constraints on the parameter values. This can lead to
the generation of devices which have an excessive number of elements or parameter values
12 Network synthesis is a specific method for designing a network, its principle being to determine the components
required to realise a given rational function that represents the network transfer function 211,212 . It was first proposed
as a means of determining the layouts of electrical networks however, due to the complete mechanical-electrical
analogy, the network synthesis algorithms can readily be used for the design of mechanical networks 213,214 .
46
2.3. VIBRATION SUPPRESSION DEVICES
Concerning the use of vibration suppression devices in aerospace structures, for decades they have
been used as shock absorbers in aircraft landing gear 233 , to suppress flutter in fixed-wing aircraft
control surfaces13 and to ensure lead-lag stability of helicopter rotor blades 192,194 - more recently
13 See Section 5.1.4.3 of CS-25 234 for a discussion of the certification requirements of these systems.
47
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
embedded devices have also been proposed for the same application 235–237 . In general, vibration
suppression research in the rotorcraft industry is more advanced than for fixed-wing aircraft given
the need to limit coupling between the blades, hub and fuselage as much as possible. Konstanzer
et al. 238 details a wide variety of devices have been proposed for this application, including mass-
spring systems and electromagnetic piezoelectric actuators. Some of these systems are in use on
commercially available rotorcraft, such as the EC225 238 and EC130T2 239 . The use of vibration
absorbers in wind-turbine blades has also been investigated 240,241 and there are several patents
related to the use of spring/damper devices to alleviate wing-pylon instability 242–244 and wing-
stores instability 245 , as well as a novel wing-to-fuselage joint with active suspension 246 , however,
in contrast to the rotorcraft industry, none of these patented technologies have progressed
beyond the conceptual research stage. Tuned mass dampers and other vibration suppression
devices have also been investigated as a means of controlling excessive vibrations in multi-stage
launch vehicles 191,247 and for improving landing performance in reusable rocket systems 248 .
Furthermore, inerter-based devices have been proposed to suppress shimmy 225 and improve
touch down performance 226 in aircraft landing gear. Inerter-based devices are particularly useful
in aerospace applications as the apparent mass effect means the inertance b can be much larger
than the actual mass of the device; an obvious benefit considering the industry’s fixation on
aircraft weight and efficiency.
An early study by Karpel 249 discussed the possibility of using traditional mass/spring/damper
vibration absorbers to passively alleviate flutter in aircraft wings, however this approach has
not been widely adopted outside of the research community, perhaps due to the lack of guidance
regarding the design of passive flutter suppression technologies in the certification documents.
Recently, Verstraelen et al. conducted a numerical 250 and experimental 251 study of a passive
vibration absorber attached to an aircraft wing for the purpose of flutter suppression. Their
results show that the optimum absorber can provide a 36% improvement in flutter speed and
reduce LCO amplitude, however when the absorber is detuned the flutter speed is decreased,
resulting in a significantly worse performance than the original unmodified system. This work
uses the classical 2DOF binary flutter model as the basis for their numerical investigations, with
much of the focus on predicting the nonlinear characteristics via a process known as numerical
continuation. In this respect, there is no conflict with the work presented in this thesis as here
the focus is on designing an absorber that is compatible with a full-scale aircraft model and
developing a method that can be integrated into the aircraft design process.
Research into exploiting such devices on braced wings is limited to a single application - a
telescopic strut design that was proposed by Haftka et al. 139,140 as part of an early MDO study at
Virginia Tech. The concept was barely investigated and no consideration was given to whether
additional mechanical elements such as dampers or inerters could be used to improve the dynamic
characteristics of the aircraft, possibly because the purpose of the telescoping mechanism was to
prevent buckling in the primary strut which was considered to be a static aeroelastic problem,
48
2.4. CHAPTER SUMMARY
not a dynamic one. Finally, a recent patent by Boeing titled Active Strut Apparatus for use with
Aircraft and Related Methods 252 details a method for controlling wing deformation by adjusting
the tension force and effective length of a pair of struts joining the fuselage to the wing. Whilst
this proposal is highly novel, it does not contain any mention of vibration suppression devices
and is more akin to a morphing wing technology.
Below is a brief summary of the main points that have been identified from the papers reviewed
in this section:
• Vibration suppression is a well-established discipline and its use throughout engineering
systems is extensive.
• Vibration suppression devices are available in a number of configurations and include both
passive, semi-active and active devices.
• Vibration suppression is routinely used in aerospace applications, such as landing gear,
rotor blades and vibration isolation in helicopters, however this technology has not been
widely applied to aircraft wings.
• The early work by Karpel 249 and more recently by Verstraelen et al. 250 considered the
possibility of using tuned-mass dampers to provide flutter suppression, however their work
was limited to a simple 2DOF model and did not consider the physical design of the device
or how it would be integrated into the airframe.
• The concept of using vibration suppression devices to provide flutter suppression and gust
load alleviation in a braced-wing aircraft has not been previously investigated, therefore
this thesis represents the first study of its kind.
This chapter has provided a comprehensive overview of the research literature that is pertinent to
the novel research presented in this thesis. Three broad topic areas were considered: aeroelasticity,
braced-wing aircraft and vibration suppression, with individual summaries of the literature
provided for each of these areas.
This literature review has highlighted that there is a substantial body of work related to
the design and modelling of truss-braced wing aircraft and that several research groups are
investigating this concept. It is well understood within the research community that a braced
wing can offer a reduction in aircraft mass and fuel-burn compared to an equivalent cantilever
wing aircraft, however aeroelastic effects such as flutter and gust loads have been identified as
limiting factors. Therefore any technology that can limit these effects will be a key enabler for the
success of the braced-wing concept. Apart from two preliminary studies using a heave-pitch flutter
model, the use of vibration suppression devices to achieve aeroelastic control in a fixed-wing
49
CHAPTER 2. BACKGROUND THEORY AND LITERATURE REVIEW
aircraft has not been widely researched. Furthermore, this technology has not been previously
applied to the braced wing concept before and so the work presented in this thesis represents a
novel contribution.
In addition to the findings highlighted in the previous sections the following research oppor-
tunities have been identified:
• Flutter Suppression in TBW - Flutter suppression has been investigated for SBW and
TBW however its effects have not been included in the design and optimisation of the
aircraft. It is possible that an improved design can be achieved by including these effects in
an optimisation routine.
• Gust Load Alleviation in TBW - Gust loads have been identified as being critical to
the loads envelope that sizes the structural components in a TBW. If the the magnitude
of gust/turbulence loads can be limited, either by active control methods or some novel
approach, then it is possible that a lower structural mass could be achieved.
• Nonlinear Aeroelastic Modelling of TBW - Several research groups have investigated
modelling techniques which capture the various nonlinear aeroelastic effects inherent to
transonic truss-braced wing aircraft. So far none of these techniques have been applied
within a MDO environment however this has been identified as future-work by a number
of key researchers in the field.
• Truss Joint Modelling - The physical design of the mechanism joining the truss-structure
to the wing and/or fuselage has not been addressed in the literature. Including the design
of this joint into an optimisation procedure would lead to an improved understanding of the
loads transferred between the truss-structure and the wing, which would impact the mass
and stiffness of the model as well as the aeroelastic response. Furthermore, it is possible
that the the joint might prove to be a limiting factor in the design of a commercially viable
braced-wing aircraft, if this is the case then it is vital that the research community dedicate
time and resources to tackling this potential problem.
In the following chapters an aeroelastic model of a truss-braced wing aircraft will be introduced
and its fundamental dynamic and aeroelastic characteristics will be discussed. This model is
used to perform two studies, one where vibration suppression devices are optimised to provide
flutter suppression and another where devices are optimised for gust loads alleviation. In the
final chapter the optimised device properties are used to inform the design of a physical vibration
suppression device using simple design formulae.
50
HAPTER
3
C
R EFERENCE M ODEL AND C HARACTERISTIC B EHAVIOUR
his chapter introduces the Nastran finite element model used for all analysis in this
T thesis and discusses the main dynamic and aeroelastic features of the model. A number
of simulations are conducted in order to understand the characteristic behaviour of the
model and identify any adverse phenomena which could be alleviated by including a vibration
suppression device in the structure. Studies performed include: eigenanalysis, frequency response
analysis, static and dynamic aeroelastic simulations and flutter analysis. The observations from
these studies are used to determine potential locations in the model geometry where a vibration
suppression device could be included. Therefore this chapter provides the context and motivation
for the novel work carried out in Chapters 4 and 5. This chapter is formatted as follows: In
Section 3.1 the BUG-T finite element model is discussed and key features and assumptions
behind the model development are described. In Section 3.2 the flexible modes and corresponding
modeshapes of the structure are calculated using the commercial finite element software Nastran.
These modes are classified using the energy distribution between the various components and
degrees of freedom in the model. This information is used to identify locations in the truss
structure where a significant amount of relative motion is experienced, making them viable
candidates for a vibration suppression device. In Section 3.3, the aeroelastic response of the
BUG-T model is calculated for steady-state trim manoeuvres and flutter. The flutter mechanism
is identified and related to the flexible modes identified in Section 3.2 and comments are made
regarding the ability of a vibration suppression device to affect the identified flutter modes.
51
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
Before any investigation into the use of vibration suppression in a truss-braced wing context can
begin it is necessary to establish the reference model and its fundamental dynamic and aeroelastic
characteristics. However, unlike for the cantilever aircraft configuration, there exists no formal
equivalent of the NASA Common Research Model 253 for a braced-wing aircraft. Therefore, to
enable the work of this thesis a suitable reference model must be developed that is broadly
representative of a truss-braced wing aircraft designed for a commercial mission.
The NASA/Boeing SUGAR project represents the largest publically available data set on truss-
braced wing aircraft, so it is appropriate to use the SUGAR High 765-095-Rev. D model presented
in Bradley et al. 30 as a starting point. This model was developed following a comprehensive
multidisciplinary optimisation study where the the wing planform and truss topology1 were
optimised to minimise the aircraft fuel-burn and direct operating costs for a mission of 3500nm.
However, given the commercial nature of the research carried out during the SUGAR project,
the raw modelling data is not available meaning that it is not possible to recreate the exact
model that was used in the SUGAR TBW studies. Instead, a substantial effort has been made to
generate a model that closely matches the overall geometry, mass and stiffness of the SUGAR
High 765-095-Rev. D model with engineering judgement and aircraft design textbooks 16,254
used to fill in any gaps in the reference data. This model is termed the Bristol Ultra Green -
Truss-Braced (BUG-T) and its development is detailed in Appendix B and C. A Nastran finite
element model (FEM) of the BUG-T has been generated using the bespoke parametric aircraft
modelling software described in Appendix C. As discussed in Chapter 2, Nastran has been chosen
given its prominence in the aerospace industry, especially in terms of defining the certification
standards for commercial aircraft.
Figure 3.1 shows the structural and aerodynamic components of the BUG-T aeroelastic
‘beam-stick’ finite element model derived from the SUGAR High 765-095-Rev. D. A brief overview
of the model is provided to familiarise the reader with some of the modelling aspects however a
full description can be found in Appendix C.
• Model assembly - The aircraft is divided into several sub-assemblies which are connected
by stiff spring elements, with the major sub-assembly components detailed in Table 3.1.
Breaking down the structure in a logical manner makes it easier to understand how
different components interact as well as making model management significantly easier.
• Structure - The structure of each component is modelled using nodes connected by flexible
beam elements with the element density determined from a mesh convergence study which
sought convergence for natural frequencies up to 50Hz. For those components which
represent lifting-surfaces, additional nodes are defined along the leading and trailing edge
which are connected to the beam nodes using rigid bar elements. These nodes provide the
1 Here, "truss topology" refers to the number of strut and jury-strut elements used to construct the truss-structure
as well as the type of joints which connect the truss elements to the wing and fuselage.
52
3.1. BUG-T FINITE ELEMENT MODEL
basis for the aero-structural spline and the use of rigid elements implies there is no warping
of the beam cross-section due to applied aerodynamic loads. This is a typical level of fidelity
adopted for static and dynamic aeroelastic analysis at the conceptual design stage and
similar models have been used in several studies 26,69,255–257 .
• Aerodynamic modelling - Aerodynamic elements are generated for each aeroelastic
component in the assembly and in keeping with the modelling approach adopted during the
SUGAR aeroleasitc analysis 30 the jury-strut is modelled as a structural component only
with no aerodynamic elements. Each aerodynamic panel has an aspect ratio of one and the
chordwise panel length is set to 0.1m to capture reduced aerodynamic frequencies up to
50Hz 82 . An aerodynamic control surface is defined on the port and starboard horizontal
stabilizer to allow the aircraft lift force and pitching moment to be balanced during a
steady-state manoeuvre. Downwash corrections due to twist or camber are not included in
the model.
• Truss connectivity - The connectivity of the truss structure is idealised as a pinned-joint
at the root and tip of every truss element (i.e. the port and starboard strut and jury-struts).
53
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
This means the local beam out-of-plane bending moment will have zero value at the root
and tip of each truss element, however all other forces and moments are transferred across
the joint. See Appendix C for a discussion of the local beam coordinate system for the
various components in the model.
• Component stiffness - The stiffness properties for the wing, strut and jury-strut are
derived from the data provided in Bradley et al. 30 , however there is no stiffness infor-
mation for the fuselage and empennage components. Instead, the beam stiffness of these
components has been set to a high value in order to provide ‘quasi-stiff’ behaviour. This
assumes that the fuselage and tail dynamics will not impact the wing dynamics, which
is not necessarily the case but this approach is consistent with the modelling carried out
during the SUGAR aeroelastic analysis and so is considered acceptable for this thesis.
• Mass distribution - A detailed mass breakdown is provided for the wing and truss
components in the SUGAR reports, however this information includes contributions from
the fuel which means that it is not possible to analyse mass cases which have different
fuel-fractions. As with the beam stiffness there is no detailed mass data for the fuselage
and empennage components and so the data in Bradley et al. 30 is used to estimate the
gross mass properties. Further details are provided in Appendix A.
For all analysis in this thesis the structural and aerodynamic properties of the BUG-T model are
fixed, with the main focus on how the aeroelastic behaviour changes as the vibration suppression
device properties vary. These device properties augment the structural mass, damping and
stiffness matrices depending on the layout of the various of mechanical elements in the device(s).
As the truss topology is fixed this approach places a restriction on the potential locations where
a vibration suppression device can be incorporated into the structure. However, given that the
SUGAR reference model was the result of a comprehensive optimisation process, the assumption
is made that it would be unacceptable to adjust the planform and/or truss topology to obtain a
more favourable device performance if it meant an increase in fuel burn. Instead the device is
considered as an augmentation to the existing structure in order to limit any adverse dynamic
behaviour. Any change to the planform to improve the device performance would need to be
considered within the scope of a holistic optimisation process similar to the study carried out
during the SUGAR project. As this is the first investigation into the use of vibration suppression
to control aeroelastic phenomena in truss-braced wings this approach is deemed acceptable.
When considering the design of a vibration suppression device the first step is to understand the
operating environment and identify the frequencies where unfavourable vibrations take place.
Typically, catastrophic failure of a structure can occur when a structure is excited at its so-called
‘resonant frequency’ resulting in the maximum transfer of energy to the structural modes. It
54
3.2. BUG-T DYNAMIC CHARACTERISTICS
follows that a key step in designing a vibration suppression device is calculating these modes as
it is likely that a device will need to influence one (or several) of them to limit the response during
resonance conditions. To that end, this section provides an overview of the baseline dynamic
behaviour of the BUG-T model and a description of the modelling techniques and assumptions
used to solve the structural equations of motion. The purpose of this section is to understand
the fundamental dynamic behaviour of the BUG-T model in order to identify locations within
the existing model geometry where a vibration suppression device could be placed in order to
influence the system dynamics.
Considering only the structural parts of the model, the equations of motion can be expressed
using the classical second-order, ordinary differential equation for a linear dynamic mechanical
system
M ẍs + C ẋs + K xs = P(t), (3.1)
where M, C and K are the mass, damping and stiffness matrices, P is some generalized force, t
is time and xs are the structural degrees of freedom (DOF) of the model, which for Nastran are
the physical displacements of each node in three orthogonal directions in both translational and
rotational senses. Removing damping and forcing terms and assuming a harmonic solution of the
form xs i = φ i sin (ω i t) allows Eqn. 3.1 to be reduced to
K − ω2i M φ i = 0,
¡ ¢
(3.2)
where ω i is the circular natural frequency of the ith mode and φ i is the corresponding eigenvector.
The non-trivial solution to Eqn. 3.2 can be found by solving the equation det K − ω2i M = 0
¡ ¢
which yields the natural frequencies and corresponding eigenvectors of the system. Note that
for a dynamic mechanical system the eigenvectors are commonly referred to as vibration modes,
modeshapes and/or normal modes - all three are used interchangeably throughout this thesis.
Furthermore, for a linear system, such as the one considered in this thesis, the displacement of
the physical coordinates can be recovered as the sum of the normal modes
Nmodes
X
xs = φi ξi , (3.3)
i =1
where i is the mode number, ξ i is coordinate of the ith mode2 and Nmodes is the number of modes
included in the summation - for the case where Nmodes is equal to the total number of degrees of
¡ ¢
freedom Ndo f the system response is recovered exactly. For models containing a large number
of DOFs the normal modes are typically used to map the structural matrices onto a reduced set in
order to limit the computational cost of carrying out the analysis. The number of modes is usually
2 Sometimes referred to as the modal participation factor.
55
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
chosen such that key dynamic characteristics of the model are captured for all frequencies of
interest, meaning that the solution recovered from the modal coordinates and modeshapes is
only an approximation of the exact solution of Eqn. 3.1. This is commonly referred to as modal
truncation and is fundamental to the dynamic aeroelastic analysis conducted by Nastran and so
a brief description of this process and its implications is necessary.
Firstly, to transform Eqn. 3.1 from the physical to the modal domain a solution of the form
xs (t) = Φξ(ω)e jω t is assumed3 and each term is pre-multiplied by ΦT , where Φ is the modeshape
matrix4 .
− ω2 ΦT M Φξ(ω) + j ωΦT C Φξ(ω) + ΦT K Φξ(ω) = P(ω) (3.4)
The pre and post-multiplication of the system matrices by ΦT and Φ has the effect of reducing the
problem size from Ndo f × Ndo f to Nmodes × Nmodes . Clearly, if Nmodes < Ndo f then the solution
of the equations of motion becomes less computationally expensive. Furthermore, the following
terms can be introduced to simplify the equations of motion in the modal domain
M hh = ΦT M Φ, C hh = ΦT C Φ, K hh = ΦT K Φ, P hh = ΦT P, (3.5)
where, using the Nastran notation, M hh , C hh and K hh are the modal mass, damping and stiffness
matrices respectively and P hh is the modal load vector. Finally, if the system matrices M, C
and K are real and symmetric and the normal modes represent an orthogonal set of vectors,
i.e. φ i × φ j = 0 for i 6= j, then the matrices M hh and K hh will be diagonal. Here, the diagonal
terms of M hh and K hh are referred to as the modal or generalised mass and stiffness terms,
denoted as m i and k i respectively. Furthermore, for a given mode the circular frequency can
be expressed as ω2i = k i /m i . Also, for the special cases of stiffness and/or mass proportional
damping the diagonal terms of the damping matrix (c i ) are related to the modal damping ratios
(ζ i ) via c i = 2m i ω i ζ i . Therefore if orthogonal eigenvectors are used and the damping matrix is
stiffness/mass proportional then Eqn. 3.4 becomes uncoupled and the solution for each frequency
can be calculated using the modal load, mass, damping and stiffness terms.
p i (ω )
− ω2 m i ξ i (ω) + j ω c i ξ i (ω) + k i ξ i (ω) = p i (ω), =⇒ ξ i (ω ) = , (3.6)
− m i ω + jb i ω + k i
2
Decoupling the equations of motion by transforming the system into the modal domain greatly
reduces the complexity and computational cost of any analysis, however, in general the damping
matrix is non-symmetric and so an uncoupled system cannot be obtained. In this case the modal
coordinates can be found using the so-called direct method by collecting ξ terms and inverting
the resulting system
¤−1
ξ = −ω2 M hh + j ωC hh + K hh P hh (ω), ξ = H hh P hh (ω),
£
=⇒ (3.7)
3 This assumption implies that the response of a system excited at a ω will contain only that single frequency and
i
no others.
p
4 j is defined as the imaginary number −1 in order to avoid any ambiguity when i is used as the mode number.
56
3.2. BUG-T DYNAMIC CHARACTERISTICS
where H hh is termed the modal frequency response function (FRF) matrix, which has an equiva-
lent in the physical domain H which is similarly constructed from the physical system matrices
M, C and K and yields the physical displacements xs when multiplied by the load vector P(ω).
¤−1
xs = −ω2 M + j ωC + K P(ω), xs = HP(ω),
£
=⇒ (3.8)
This discussion on coupled vs. uncoupled EoM is relevant because when vibration suppression
devices containing viscous damping elements are augmented to the BUG-T model in Chapters 4
and 5 the damping matrix becomes non-symmetric and so the direct solution method must used
to obtain the forced frequency response of the system.
As the truss-braced wing aircraft is a novel and unusual configuration its modal properties are
not as straightforward as a typical cantilever wing. It is also important to classify the dynamics
of the model to understand how vibration suppression can be used to influence any adverse
dynamic behaviour. Therefore, this section presents a study of the so-called ’normal modes’ of
the BUG-T model - which are the eigenvalues and corresponding eigenvectors resulting from
the solution of the eigenvalue problem posed by Eqn 3.2. A method for classifying these modes is
presented which uses the Modal Assurance Criterion (MAC) coupled with the modal strain and
kinetic energy distributions as detailed in Towner and Band 258 to determine which parts of the
BUG-T model sub-assembly are active across the different modes. This is relevant for vibration
suppression considerations as if a certain mode is identified as having a strong participation
during aeroelastic flutter or turbulence encounters then an understanding of that mode will help
guide the placement of a vibration suppression device.
Concerning the calculation of the normal modes of the BUG-T model, the standard Lanczos
method 259 is employed to determine the first 50 structural modes including the 6 rigid body
modes of the model. Each calculated mode is mass-normalised, i.e. each mode has a modal mass
of unity, which is the standard approach for a normal modes analysis in Nastran. Figure 3.2
shows the modeshapes of the mass-normalised, air-off normal modes for a selection of the flexible
modes of the BUG-T model and additional modeshape plots are provided in Appendix D. These
modes are termed the ‘air-off ’ modes to distinguish them from the flutter modeshapes shown in
Section 3.3.3. The modes are numbered in order of increasing natural frequency, meaning that
the first flexible mode (i.e. non-rigid body mode) of the structure is numbered as mode seven and
not mode one. Note, there is no ‘engine-mode’ as the engine is included in the model as a lumped
mass and there is no structural representation of the pylon. Upon inspection of Fig. 3.2, one thing
that is immediately obvious is the appearance of modes which are localised to specific parts of
the model. For example, mode 21 in Fig. 3.2(c) is clearly a symmetric out-of-plane bending mode
for the port and starboard strut elements.
57
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
(a) Mode 7 - 1.498Hz
(b) Mode 10 - 3.027Hz
58
F IGURE 3.2. Linear, mass-normalised, air-off modeshapes for the BUG-T model. The translation terms of each modeshape
are scaled by a factor of 100 to emphasise the shape whilst the beam twist is visualised by forming a plane between the
leading and trailing edge nodes of each lifting surface - the shading of the plane represents the magnitude of the beam
twist. Finally, the grid point kinetic energy and element strain energy for each sub-assembly component is presented as
a fraction of the total strain and kinetic energy for each mode.
3.2. BUG-T DYNAMIC CHARACTERISTICS
Generally a mode is not so easily identified by visual inspection of the modeshape, therefore,
to aid in the classification of the normal modes the Grid Point Kinetic Energy (GPKE) and
Element Strain Energy (ESE) distribution are output for each mode. These terms are defined in
the Nastran dynamics user guide 259 as
ESE = 1/2 xT
e K e xe , GPK E = ΦT
mass M Φ mass , (3.9)
where x e are the DOFs connected to any flexible elements, K e is the element stiffness matrix and
Φmass denotes the matrix of mass normalised modeshapes. Note that the 1/2 term is dropped from
the GPKE equation as Nastran normalises the GPKE such that the total kinetic energy in the
model is equal to one 259 . Also, the GPKE terms have been transformed into the global coordinate
system5 (X G , YG , ZG ) from the local component coordinate systems during the post-processing
stage as this allows the energy terms across different components to be evaluated in a like-for-like
manner. Furthermore, the energy distributions have been grouped according to the component
sub-assemblies in Table 3.1 and then normalised with respect to the total strain and kinetic
energy in the model. These normalised terms are referred to as ESE and GPK E and are shown
as inset bar charts in Fig. 3.2. Finally, as the fuselage and empennage components are modelled
as quasi-stiff beams they contain less than 0.001% of the total model strain and kinetic energy
and so have been neglected from the energy distribution charts.
Using this additional information it is straightforward to quantitatively assess which compo-
nent and degree of freedom is most active in each normal mode. For example, mode 7 (Fig. 3.2())
is the first wing out-of-plane bending mode, mode 10 (Fig. 3.2(a) is the first wing twist mode
and mode 21 is the first strut out-of-plane bending mode. That being said, not all modes exhibit
component-localised behaviour - for example mode 11 is mostly dominated by energy terms in
the wings although it is spread across the different DOFs and a further 20% of the strain energy
is contained in the strut elements. Such modes are termed global models and are more difficult
to classify as the energy is spread throughout the structural components and degrees of freedom.
Figures 3.3(a) and 3.3(b) show the ESE and GPK E quantities for the first 50 normal modes.
Examining the normalised strain energy distribution in Fig. 3.3(a) the presence of the six rigid
body modes is immediately obvious as the strain energy for these modes is numerically zero - this
is a strong indicator that the FEM and resulting structural matrices are well-posed. From both
energy distributions it is clear that the majority of the model energy is concentrated in the wing
components during the low frequency modes and as the natural frequency increases there are
modes which are strut dominant with increased strain energy in the jury-strut as well. Although
one point to note regarding the GPK E distribution is that it is biased towards components that
have a much larger mass, for example the wing vs. the jury-strut. Whilst this may appear to be an
obvious observation it has important implications when classifying the modeshapes. For instance,
the GPKE distribution in Fig. 3.3(b) implies that the jury-strut components barely participate in
5 The global coordinate system considered here is the same as the MSC Nastran basic coordinate system.
59
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
F IGURE 3.3. Element strain energy (a) and grid point kinetic energy distribution (b)
by sub-assembly component for the first 50 normal modes of the BUG-T model.
60
3.2. BUG-T DYNAMIC CHARACTERISTICS
F IGURE 3.4. Modal Assurance Criterion for the BUG-T model normal modes using
the conventional MAC (a), the Element Strain Energy (b), the Grid Point Kinetic
Energy (c) and the average of all three measures (d).
the first 50 modes, whereas the ESE distribution in Fig. 3.3(a) shows that these components are
much more active. Finally, there is a tendency for the mode energy terms to be grouped together
in sets of two, this is due to the appearance of symmetrical and asymmetrical modes which is
symptomatic of the model symmetry about the X G ZG plane.
The final discussion on the BUG-T normal modes is focussed on the measure of similarity
between each modeshape. There exists a well-known scalar quantity called the Modal Assurance
61
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
Criterion 260 (MAC) which is used to assess the degree of similarity between two modeshapes,
defined as ¯ T ¯2
¯φ φ t ¯
r
M AC = ¡ T ¢ ¡ T ¢ , ∀ r, t ∈ [1, Nmodes ] (3.10)
φr φr φ t φ t
Mathematically speaking, this measure corresponds to the magnitude of the cosine of the angle
between the two modeshape vectors φr and φ t squared, therefore, the value of the MAC is
bounded by 0 and 1, where a value of 1 means the two vectors are the same (i.e. ∀ r = t) and
a value of 0 means the two vectors are geometrically orthogonal. When constructing the MAC
matrix for a set of numerically generated modeshapes it is expected that each modeshape will
be mass orthogonal with respect to the rest of the generated set. Figure 3.4(a) shows the MAC
matrix for the first 50 normal modes of the BUG-T model. Here, the orthogonality between the
different normal modes is clearly demonstrated by the dominant diagonal terms and the lack
of coupling on the off-diagonals. There is a small amount of coupling towards the higher mode
numbers but this is attributed to the model symmetry as discussed in Gockel 259 . In this instance,
using the MAC as the only measure of similarity between the modes implies that the modeshapes
are distinct. Whilst this may be true in a mathematical sense, it is clear by visual inspection that
certain aspects are shared across the different modes. To help quantify this, a similar measure to
the MAC can be constructed using the ESE and GPKE distributions as described in Towner and
Band 258 . These terms are denoted as M AC ESE and M ACGPK E and their corresponding matrices
are shown in Fig. 3.4(b) and Fig. 3.4(c) respectively. These new terms show that there is a higher
degree of similarity between the different modes than previously indicated by using the MAC
alone. The M AC ESE matrix especially shows a significant degree of coupling, although this may
be because Nastran groups together the element strains from all six DOFs into a single metric,
unlike the GPKE which is split amongst all six DOFs. In both the ESE and GPKE MAC matrices
the symmetric-asymmetric mode pairs are clearly identified by sets of 2x2 entries centred around
the matrix diagonal. Following the methodology of Towner and Band 258 these three metrics can
be combined into a scalar quantity with a value between 0 and 1 by taking the average of the
M AC, M AC ESE and M ACGPK E terms. This combined quantity is referred to as M AC and is
shown in Fig. 3.4(d) 6 . The main observation here is that BUG-T normal modes show some degree
of similarity when additional metrics such as the element strain energy and grid point kinetic
energy are taken into account. Therefore, if a vibration suppression device can be designed to
influence a single mode then there is a possibility it will also affect other modes which share
similar characteristics with the targeted mode.
As the truss-topology of the BUG-T model is fixed the potential device locations are con-
strained to existing connections within the wing-truss structure. These locations are shown in Fig.
3.5 and are termed Locations A, B and C - which correspond to a rotational device at the strut
root joint, a translational device which spans the jury-strut and a further rotational device at the
6 Note that this method of classifying modeshapes is not particular to truss-braced wings and can be used for any
generic finite element model which can be divided into a logical breakdown of components 258
62
3.2. BUG-T DYNAMIC CHARACTERISTICS
Wing LocationWC
Jury-Strut
FlexibleWBeams
LocationWB PinnedWJoints
Strut Location Label DeviceWType
LocationWA A Strut-Root Rotational
B Jury-Strut Translational
C Strut-Tip Rotational
F IGURE 3.5. Proposed vibration suppression device locations. Devices will be placed at
these locations in both the starboard and port components.
strut tip joint7 . Note that rotation about the hinge connections at the root and tip of the jury-strut
is not considered as the primary load path for the jury-strut is tension/compression for which a
translational device is best suited. In order for a two-terminal vibration suppression device to
target a particular structural mode it must experience some relative motion across the device
terminals at the frequencies of interest. To determine how ’active’ these devices will be at the
various normal modes frequencies the relative modal displacements at the three device locations
have been extracted and are presented in Fig. 3.6 for the first 50 modes up to 30Hz. These results
show that all three candidate locations experience very little relative motion at frequencies less
than 5Hz. This is slightly concerning from an aeroelastic control point of view as these modes tend
to be critical for determining the dynamic aeroelastic response of the aircraft, such as for flutter
or during turbulence encounters. The modal relative displacement is more significant in the
10-15Hz and 25-30Hz regions of the frequency domain, which corresponds to a series of localised
modes in the truss-structure. For modes in these regions the strut experiences significant rotation
about its root and tip due to the pinned connection between the truss elements and the wing and
fuselage, also, the jury strut experiences some extension/compression motion due to the combined
bending of the strut and the wing - hence the higher modal relative displacement. These results
could indicate that a passive vibration absorbed placed at one or all of these locations would be
more effective at targeting the localised (high-frequency) truss modes instead of the fundamental
modes which are more relevant for dynamic aeroelasticity 52 . This implies that the layout and
parameter values of the vibration suppression devices will need to be carefully tailored in order
to target the low frequency modes and provide a more favourable dynamic aeroelastic response.
This section has demonstrated a method for quantitatively classifying the normal modes of
the BUG-T model using three metrics: the visualised modeshapes, the distribution of strain and
kinetic energy and a modified Modal Assurance Criterion which accounts for the orthogonal-
ity between the eigenvectors, the kinetic energy vectors and the strain energy vectors. Three
candidate device locations have been identified based on the existing connection points in the
7 Even though Fig. 3.5 shows device locations for a half-wing devices will be placed at these locations in both the
63
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
F IGURE 3.6. Relative modal displacement at the proposed vibration suppression device
locations. The left and right y-axis show the relative velocity for the rotational and
translational devices respectively.
BUG-T truss-structure. Examining the relative modal displacements at these locations revealed
that the low-frequency modes do not exhibit much relative modal displacement at the proposed
device locations, however, for modes in the range 10-15Hz and 20-25Hz there is an increase
in relative modal displacement which is attributed to localised modes in the truss-structure.
It is hypothesised that if the structure is excited at these natural frequencies, such as during
aeroelastic flutter or turbulence encounters, then a vibration suppression device placed at the
joint locations of the strut or across the ends of the jury-strut could be used to influence the
dynamics of the wing. Although it is likely that the device properties will need to be tailored to
target the fundamental modes which are most critical for aeroelastics 52 .
In this section the candidate device locations will be evaluated by examining the amount of
relative velocity experienced across the two DOFs that will act as the device terminals. To achieve
this a frequency response analysis will be conducted where the structure is excited by a harmonic
load applied at the wing tips and the velocities at three device locations will be extracted and
presented as a function of forcing frequency. This will allow device effectiveness to be assessed
at the normal mode frequencies identified in Section 3.2.2. The relative velocity is used because
within the field of mechanical network design the device admittance functions are commonly
formulated in terms of the velocity at the two terminals of the device 208 .
64
3.2. BUG-T DYNAMIC CHARACTERISTICS
First, the physical FRF matrix H is formulated using the M, C, K matrices and frequencies in
the range 0 : dF : F max , where dF = 0.01Hz and F max = 30Hz. The structure is excited by a unit
harmonic load at the wing tip leading edge of the port and starboard wings as this will activate
most of the structural bending modes within the frequency range of interest. Furthermore, 3%
structural damping is applied at all excitation frequencies in accordance with the guidance in the
certification documents 234 . Next, the physical displacements are calculated at each frequency
using Eqn. 3.8. With the displacements calculated at each forcing frequency the velocities can
be recovered using ẋ = j ω x, which holds true as long as the applied force and assumed solution
are harmonic, i.e. x = x0 e jω t when P = P0 e jω t . Each device location has two velocity terms per
frequency corresponding to the two terminals of the vibration suppression devices. In the case of a
rotational device these are the dependent and independent nodes of the pinned-joints at Location
A and C and for the translational device at Location B the root and tip node of the jury-strut are
used. For each location the relative velocity can be calculated as ∆ ẋ = ẋ1 − ẋ2 , where subscript
1 and 2 denote the node at the terminal of the device. The frequency response function of the
magnitude of the relative velocity at each device location (|∆ ẋ|) is shown in Fig. 3.7, with inset
plots of modeshapes where the device locations experience large magnitudes of relative velocity.
The first point to note is that in general the velocities at Locations A and C are larger than at
Location B which could imply that a rotational device will be more effective than a translational
F IGURE 3.7. Unit frequency response function of the relative velocity at the proposed
device locations with 3% structural damping. The left and right y-axis show the
relative velocity for the rotational and translational devices respectively and inset
axes show the modeshapes at frequencies where one or more of the device locations
experience a large relative velocity.
65
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
device. However there is not one location which consistantly experiences a larger magnitude
across all excitation frequencies, so it could be that one device location is better suited to targeting
a particular mode, which introduces the possibility of using multiple devices to target the response
of several modes. Next, it is clear that the magnitude of all three locations reaches a maximum
in the 10-15Hz range. This corresponds to the group of strut/jury-strut dominated modes that
are present between modes 20-30 in Fig. 3.3(a), suggesting that a device located at any of the
three locations will be able to influence this group of modes. Either side of the 10-15Hz range the
magnitude of ∆ ẋ is substantially less, especially for the translation device at Location B which
has almost zero value for modes in the range 0-5Hz8 . In contrast, the rotational devices have a
much larger magnitude in this frequency range, with Location A generally having a higher value
of |∆ ẋ| because of the lower beam stiffness in the inboard section of the strut9 . These results
confirm the observations made for the normal modes in Fig. 3.6.
Finally, it is important to acknowledge that the proper characterisation of the frequency
response of a model requires a more detailed study than has been presented here. For example a
distributed harmonic load could have been applied which is more representative of the aerody-
namic loads experienced during flight, or the contribution of different components of loads applied
at different points in the model could have been assessed to determine which load conditions
will induce relative motion at the proposed device locations. However the purpose of this section
was to identify trends in the amount of relative motion at each proposed device location in order
to gain a preliminary understanding of the model dynamics, and so for this reason a simple
frequency response analysis was preferred over a more comprehensive study.
This section has provided an overview of the normal modes of the BUG-T FEM and has used
this information to determine three locations where a vibration suppression device could be
incorporated into the structure. Next, a frequency response study was conducted to determine
the magnitude of the relative velocity at these three locations as a function of frequency, with this
information used to hypothesize about the effectiveness of a given device location at influencing
certain structural modes. Based on these two studies the following conclusions have been made:
• Localised Modes - The BUG-T model is broken down into a sub-assembly of components
with different stiffness and mass distributions which leads to the presence of localised
modes within the structure. This is further exaggerated by the pinned-joint connectivity
between the truss elements and the wing/fuselage components which promotes a series
of out-of-plane bending modes in the 10-15Hz range. Given the large amount of relative
motion that occurs during these ‘truss modes’ it is hypothesized that a vibration suppression
8 It is especially important that a device is active at these low frequencies as structural modes in this range tend
66
3.3. BUG-T AEROELASTIC ANALYSIS
device located at the strut joints or across the jury-strut could be used to influence the
dynamics of the wing and truss components.
• Device Locations - Three potential device locations are identified: a rotational device at
the strut root joint (Location A in Fig. 3.5), a translational device which spans the jury-strut
(Location B in Fig. 3.5) and a rotational device at the strut tip joint (Location C in Fig. 3.5).
The viability of these locations is heavily dependent on the use of pinned-joints to connect
the truss elements to the wing and fuselage. However, numerous trade studies have shown
that pinned-joints provide an overall benefit in terms of aircraft weight 30,44–46 and so it is
fair to proceed with the assumption that pinned-joints will be used.
• Device Effectiveness - The frequency response study in Section 3.2.3 showed that for
all three device locations the relative velocity reaches a maximum in the 10-15Hz range,
which corresponds to the localised truss modes identified in Section 3.2.2. It was noted
that low-frequency modes are more critical during aeroelastic phenomena 52 , therefore it is
hypothesized that rotational devices at the strut pinned-joints will be more effective than a
translational device across the jury-strut based on the larger relative velocities at these
locations in the 0-5Hz range.
During the aircraft design process, a considerable effort is made to simulate all potential operating
conditions in order to establish the critical loads envelope, that is, the maximum and minimum
loads that every component can experience across all possible operating conditions. Such a
process often involves evaluating tens of thousands of flight points, load factors and mass cases
in order to determine the critical loads which size the airframe. Often, these loads are attributed
to both static and dynamic aeroelastic load cases across multiple altitudes and aircraft velocities.
Furthermore, it is also necessary to certify that the stability of the airframe is not compromised
at any point during the mission. Here, stability can relate to static stability such as local or
global buckling as well as dynamic instability resulting from aeroelastic flutter. In the case of a
truss-braced wing, work conducted during the SUGAR project 30 and more recently by Rajpal and
De Breuker 154 identified that gust loads are critical to sizing many of the airframe components.
Also, a multidisciplinary optimisation study conducted by Mallik et al. 47 identified that satisfying
the flutter constraint leads to a 7.5% increase in the mission fuel burn compared to the case
where the flutter constraint was relaxed. These results highlight the importance of investigating
methods which can provide gust loads alleviation and suppress aeroelastic flutter in order to
enable more efficient aircraft designs.
This section is formatted as follows: First, Section 3.3.1 provides a brief overview of the
theory and limitations of the aeroelastic formulation used by Nastran. In Section 3.3.2 the static
aeroelastic response of the BUG-T model is evaluated using the load cases from the SUGAR TBW
67
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
sizing study. Using these results the critical loads envelope is generated for the wing, strut and
jury-strut components and comparisons are made against the data from the SUGAR reports 30
to test the validity of the BUG-T model. Finally, in Section 3.3.3 the flutter behaviour of the
BUG-T is determined and this is related to the observations from the dynamic analysis in Section
3.2.
As the aeroelastic methods in Nastran are considered industry-standard a full derivation of the
equations of motion is not necessary, instead this section will discuss the underlying assumptions
that govern the Doublet Lattice Method and their relevance to the work presented in this thesis.
A brief overview of the working principles of DLM is also given to provide context for the static
and dynamic equations of motion presented in the remainder of this chapter.
The DLM theory used by Nastran is an extension to the classical Vortex Lattice Method
(VLM) for unsteady flow conditions and is described in detail by Albano and Rodden 261 . DLM
is based on linearised velocity potential theory, meaning that the following assumptions are
inherent to the formulation 13 :
• Inviscid flow - The fluid viscosity is equal to zero. Therefore there is no development of
a boundary layer on any of the external surfaces and aerodynamic forces are a function
of pressure only, i.e. the shear force at the surface boundary is zero and there is no
parastic/form drag. For this condition to be true the Reynolds number of the flow would
have to be infinite, however it is typical for flow conditions with high Reynolds numbers to
be modelled using inviscid formulations.
• Irrotational flow - The fluid particles do not have angular velocity and can be described
by translational components only. Irrotational flow implies that the flow cannot become
turbulent and is laminar at all points, i.e. the streamlines of the flow remain parallel.
Turbulent flow typically occurs as a result of boundary layer growth or at high angle of
attack, so assuming that the flow is inviscid and the aircraft restricts itself to low angles of
attack this assumption is valid.
• Small perturbations - The linearised potential theory is based on the concept of velocity
perturbations. These perturbations must be small with respect to the freestream velocity in
order to linearise the flow equations, which requires that the aerodynamic body is slender
and operating at small angles of attack. In Nastran’s DLM this is facilitated by using thin
aerofoil theory to model the lifting surface as a collection of flat panel elements with their
chord orientated parallel to the freestream. Thin aerofoil theory implies that the velocity
induced by these panels acts normal to the panel plane.
• Incompressible flow - The fluid density and mass flow rate remains constant meaning
that shock waves cannot form and there is no wave drag, although compressible effects
can still be accounted for using corrections such as the Prandtl-Glauert transformation 13 .
68
3.3. BUG-T AEROELASTIC ANALYSIS
Strictly speaking such an assumption is only broadly correct at Mach numbers less than 0.3,
however, the SUGAR project assumed a cruise Mach of 0.7 with a maximum operating Mach
of 0.82. This implies that compressible effects are likely to be important to the aerodynamic
performance of the aircraft, however as this thesis focusses on the dynamic aspects of the
model and does not consider any aerodynamic performance this assumption is valid within
the scope of this research.
• Uniform flow - The freestream must be uniform or varying harmonically. This restricts
Nastran to the evaluation of steady flow conditions or flow that is varying harmonically,
such as during a 1-cosine gust or at the point of flutter 49 .
In addition to these aerodynamic considerations, Nastran’s aeroelastic solution sequences assume
that the structural response to the applied aerodynamic loading is linear. Meaning that typical
nonlinear aeroelastic effects such as ‘tip-shortening’ 262 and other large rotation effects are
neglected. Also, the coordinate systems and structural mass, damping and stiffness matrices
are considered fixed for the duration of the analysis. It is possible to include these effects using
a fluid-structure-interaction approach 168,262 however this is considered outside of the scope of
this thesis. Taking these assumptions together it may seem like Nastran is quite limited in
its scope. However as mentioned in Chapter 2, the strength of Nastran is its ability to handle
aircraft-level aeroelastics using aerodynamic theories that are broadly accurate, making it ideal
for trade studies at the preliminary design stage or conceptual research. What follows is a brief
discussion of the subsonic aerodynamic modelling in Nastran and how the aerodynamic loads are
transferred to the structure.
In Nastran each aerodynamic surface is represented by a collection of trapezoidal panels, as
in Fig. 3.8. Each panel is parallel to the freestream and contains a line of acceleration potential
U∞
Line of doublet elements
ci
F IGURE 3.8. Example panel geometry for the Nastran doublet lattice method.
69
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
doublets on the quarter chord line of the panel, with uniform but unknown strength. The central
problem of the doublet lattice method is to determine the strength of every doublet element such
that the normal velocity at each panel control point is zero and the Kutta condition is enforced 10 .
It can be shown 261 that once the doublet strengths are determined the aerodynamic pressure
acting on each panel can be recovered from the panel downwash velocity, which itself is a function
of the panel geometry and the influence of all doublet elements on an individual panel. Next, the
aerodynamic pressure distribution can be converted to a force on each panel by multiplying the
pressure by the panel area, yielding a force which is normal to the panel plane11 . In Nastran, this
process is generalised by forming a matrix of Aerodynamic Influence Coefficients (AICs) which
can be expressed as
Q kk = S ki A −ii1 D 1ik + jkD 2ik
¡ ¢
(3.11)
where subscripts kk and i denote the aerodynamic set and ith aerodynamic panel respectively.
Q kk is the overall AIC matrix, S ki is the integration matrix and A ii is the AIC matrix of each
panel which is a function of Mach number and reduced frequency k only. D 1ik and D 2ik are the
real and imaginary parts of the substantial differentiation matrix which relates the deflections of
the aerodynamic DOFs to the downwash on each panel via
g
w i = D 1ik + jkD 2ik xk + w i
¡ ¢
(3.12)
where w i is the downwash velocity on each panel, xk are the displacements of the aerodynamic
g
degrees of freedom12 , and w i is a ‘static’ downwash term which includes contributions from
some distribution of initial incidence on each panel, such as from angle of attack or aerofoil cam-
ber/twist 82 . The aerodynamic DOFs are related to the structural DOFs by the use of mathematical
splines
xk = G ks xs (3.13)
where G ks is the spline matrix and xs are the deflections of the structural DOFs. It is important
to note that the spline matrix only maps structural deflections which contribute to the deflection
of xk in the direction normal to the panel plane, and any additional incidence due to the twisting
of the structure is transferred to the aerodynamic set via additional downwash terms. This
is a result of the small perturbation assumption which requires the aerodynamic body to be
represented by a series of potential elements located on the aerofoil zero-camber line.
10 The Kutta condition requires that the circulation distribution yields a stagnation point at the trailing edge of
calculated by Nastran.
12 The aerodynamic degrees of freedom are defined as the deflections of the quarter chord point of each panel in the
70
3.3. BUG-T AEROELASTIC ANALYSIS
In this section the static aeroelastic response of the BUG-T model will be quantified using the
mass configurations and aeroelastic load cases from the SUGAR TBW design study 30 . The purpose
of this study is to understand the magnitude of loads that the structure will experience during a
typical steady-state manouevre and determine which load cases form the critical loads envelope.
This is important in the context of vibration suppression as any vibration absorbers included in
the structure will need to be designed to withstand these static loads during normal operation.
In Nastran calculating the static aeroelastic response requires certain degrees of freedom to be
‘released’ so that the external forces and moments due to aerodynamic and inertial loads can be
balanced by the reaction loads from the structural flexibility and additional aerodynamic loads
from control surfaces. Here, an additional set of DOFs (x e ) are introduced which comprise the
released degrees of freedom and any additional aerodynamic degrees of freedom 13 . Hence, the
static aeroelastic equation of motion for a flexible aircraft is defined as 82
(K − qQ ss ) xs + M ẍs = qQ se x e + P s (3.14)
where Q ss is the AIC matrix which provides aerodynamic forces at the structural DOFs due
to structural deformations, q is the dynamic pressure, Q se is the AIC matrix which provides
aerodynamic forces at the structural DOFs due to change in x e and P s is a vector of additional
applied loads. Note that the complex (oscillatory) terms of Eqn. 3.12 are neglected during static
aeroelasitc analysis. This equation is solved directly using the process described in Rodden and
Johnson 82 to find the values of x e which satisfy the aircraft trim. In this work, the aircraft is
trimmed by balancing vertical acceleration and pitching moments using the angle of attack and
elevator.
In this section the static aeroelastic response of the BUG-T model is calculated using the load
cases from the SUGAR sizing study 30 . The load cases used in this section are given in Table
3.2 and are a subset of the 17 load cases and two mass configurations that were used for the
preliminary sizing study in Bradley et al. 30 . The full 17 load cases could not be considered as
the wing mass data that was used to develop the BUG-T model includes contributions from the
fuel mass. Meaning it is not possible to evaluate mass cases with different levels of fuel loading
as the structural mass and fuel mass are defined as a single lumped mass - further details are
provided in Appendix A. With this in mind the load cases used in this section assume an all-up
mass (AUM) of 68038kg, which is the same as the Maximum Take-off Weight (MTOW) of the
SUGAR 765-095 Rev. D. Table 3.2 defines each load case as a combination of altitude, Mach
number, aircraft velocity and vertical load factor, with the title of each load case taken from
13 These additional aerodynamic degrees of freedom are typically associated with control surfaces.
71
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
TABLE 3.2. BUG-T static aeroelastic load cases for an AUM of MTOW, taken from
Bradley et al. 30 Table 2.9. The values of U∞ are found by interpolating the flight
envelope data in Appendix A.
No. U∞ M Z nz
Title
[−] [m/s] [−] [ft] [g]
1 206.8 0.7 36,000 2.5 2.5g manoeuvre at cruise, M 0.7 (MTOW)
2 206.8 0.7 36,000 -1 -1g manoeuvre at cruise, M 0.7 (MTOW)
3 159.6 0.2 0 2.75 Pratt Gust at sea level, M 0.2 (MTOW)
4 159.6 0.4 0 2.84 Pratt Gust at sea level, M 0.4 (MTOW)
5 185.6 0.5 10,000 2.68 Pratt Gust at 10K ft, M 0.5 (MTOW)
6 218.5 0.6 20,000 2.57 Pratt Gust at 20K ft, M 0.6 (MTOW)
7 212.1 0.7 30,000 2.23 Pratt Gust at 30K ft, M 0.7 (MTOW)
8 202.7 0.7 40,000 1.85 Pratt Gust at 40K ft, M 0.7 (MTOW)
Bradley et al. 30 . For each load case the resulting loads envelope and structural deflections are
compared against the SUGAR results study 30 in order to check the validity of the BUG-T model.
For load cases three to eight, the vertical load factor is calculated using the quasi-static Pratt
Gust formulation 263 . This is an approximate method which allows an equivalent static vertical
load factor to be derived for a given altitude and aircraft mass based on the response of a rigid
aircraft to a 1-cosine gust. Using the method detailed in Pratt 263 the equivalent static load factor
is defined as 1 + ∆ n z g . Where ∆ n z g is the incremental gust load factor, defined as
K g C L α ρ S re f U∞Ure f
∆n z g = (3.15)
2 × AUW
where K g is the gust load alleviation factor that accounts for the aircraft motion and the build-up
of unsteady aerodynamic forces due to aerodynamic lag 49,52 , C L α is the rigid aircraft lift-curve
slope, ρ is the air density, S re f is the reference wing area, U∞ is the aircraft forward velocity and
Ure f is the gust vertical velocity, which itself is a function of altitude 234 . Reference quantities for
the BUG-T model are provided in Table 3.3. Precise data for C L α is usually not available at the
preliminary design stage so it is often estimated as 2π. In Pratt 263 the value of K g is a function
of the mass ratio µ g , which is approximated by the curve
¡ ¢
0.88µ g
Kg = ¡ (3.16)
5.3 + µ g
¢
where µ g is defined as
2 × AUW
µg = (3.17)
C L α ρ cS re f g
with c denoting the mean geometric chord (wing area / wing span) and g is the acceleration due
to gravity. Values for U∞ and AUW are taken from Table 3.2 and the variation of Ure f with
respect to altitude is defined in Section 25.341 of CS-25 234 . Finally, for this study the value of
C L α is assumed to be equal to the aircraft vertical force stability derivative (with respect to angle
72
3.3. BUG-T AEROELASTIC ANALYSIS
of attack) ignoring elastic effects, C Z ri gid . This quantity is part of the standard output for the
Nastran static aeroelastic solution and is calculated at the 1g flight condition for each Pratt gust
in Table 3.2.
Figure 3.9 shows the loads envelope for a selection of beam loads and components whilst the
full loads envelopes comprising all six beam loads for the wing, strut and jury-strut components
are given in Appendix D. The loads envelope is calculated by taking the maximum and minimum
values of the beam loads across each component and the plots in Fig. 3.9 are colour-coded to
denote which load case has the maximum/minimum load at that point. Note that where two
different beam elements are attached to the same finite element node the average value of the two
sets of beam loads has been used, which is consistent with the presentation in Bradley et al. 30 .
For example, the wing axial force distribution in Fig. 3.9(b) should show an instantaneous jump
in force at the strut attachment point due to the additional loads transferred from the strut to
the wing. However as the average load is used the variation in loads at the attachment point is
gradual instead of a step change. Furthermore, the x-axis for each plot is normalised beam axis
of the component (η r ), which is the straight line distance along the line of nodes which the finite
element beams are attached to. This allows consistent comparison of components which have
different orientations with respect to the global coordinate system.
Concerning the loads envelopes, the wing spanwise bending moment envelope in Fig. 3.9(a)
shows that the loads reach a maximum at the wing strut attachment point and then reduce
towards the wing-fuselage joint - which is the typical behaviour for a braced-wing aircraft 108,109 .
The wing axial force envelope in Fig. 3.9(b) shows good qualitative agreement with the SUGAR
data with the exception of the behaviour inboard of the root position. This is due to the constraints
applied to the BUG-T FEM to model the wing-fuselage connection which are not necessarily
the same as the SUGAR model. Regarding the strut spanwise bending moment in Fig. 3.9(c),
the main observation is that the load is equal to zero at the root and tip and that there is no
step change in bending moment at the jury-strut attachment point. Both these points are a
direct consequence of the pinned connections used to connect the truss elements to the wing and
fuselage. The critical sizing cases for the three loads envelopes are the two sea level static-gust
cases and the -1g and 2.5g manoeuvre cases. This is precisely what was found in the SUGAR
study and gives confidence that the BUG-T model is a reasonable approximation of the SUGAR
model. These load cases have the highest values of dynamic pressure out of the cases in Table 3.2
73
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
·106
1
Root Engine Jury-Strut
[Nm]
0. 5 Strut
My
·106
0
Fx
−1
·104
2
[Nm]
0
My
−2 Jury-Strut
F IGURE 3.9. Static aeroelastic loads envelope for the wing spanwise bending moment
(a), wing axial force (b) and strut spanwise bending moment (c).
74
3.3. BUG-T AEROELASTIC ANALYSIS
Figure 3.10 shows the static aeroelastic deformed shape of the starboard wing and truss
components for all eight load cases and Table 3.4 provides the wing tip deflections for the
BUG-T model compared against the SUGAR data in absolute values and as a percentage of
the semi-span. The 2.5g and -1g deflections show reasonable agreement with the SUGAR data,
however there is a significant error in the results for the Pratt gust cases. This is because during
the SUGAR analysis the Pratt gust loads for wing sections outboard of the strut attachment point
were scaled using a linear scale factor that varied from 1 at the wing-strut attachment point, to
1+δ at the wing tip. As the value of δ is not provided in the SUGAR reports it has not been applied
in this study which is the reason for the large discrepancy in the deflection results. Regarding
the magnitude of the wing tip deflection, a rough design rule in the aeroelastic community is that
nonlinear effects start to become important when the wing tip deflection reaches ±10% of the
semi-span 255 . Therefore from the results in Table 3.4 it is likely that the BUG-T and SUGAR
models will begin to exhibit nonlinear aeroelastic effects, which is consistent with observations
from the SUGAR wind tunnel tests 148 . A thorough study of nonlinear aeroelastic effects on the
6
Z [m]
2
0 2 4 6 8 10 12 14 16 18 20 22 24 26
Y [m]
F IGURE 3.10. Static aeroelastic deflections for the starboard wing, strut and jury-strut
components for each of the load cases in Table 3.2.
75
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
TABLE 3.4. Comparison of static aeroelastic wing tip deflection for the BUG-T and
SUGAR models.
critical sizing loads of a braced wing configuration has not been conducted within the research
community although it has started to be addressed by some research groups 100,106 , including a
preliminary study by this author 264 .
The study in the previous section showed that quasi-static gusts form a major component of the
critical loads envelopes. However, examining Eqns. 3.15 and 3.17 it is clear that the Pratt gust
load factor is heavily dependent on the aircraft lift-curve slope which is commonly estimated as
2π. To test this assumption, this section presents a study where the value of C L α is varied using
four simple approximations and the corresponding effect on the gust load factor and resulting
critical loads envelope is discussed. The four different values of C L α considered are:
• Case 1 - The theoretical lift-curve slope as predicted by thin aerofoil theory - 2π.
p
• Case 2 - The theoretical lift-curve slope with the Prandtl-Glauert correction - 2π/ 1 − M 2 .
• Case 3 - The derivative of the aircraft vertical force with respect to AoA for a rigid wing -
C Z ri gid . Calculated using Nastran for the 1g condition of each Pratt gust load case.
• Case 4 - The derivative of the aircraft vertical force with respect to AoA with elastic effects
- C Z f l exibl e . Calculated using Nastran for the 1g condition of each Pratt gust load case.
Note that C Z f l exibl e is calculated by Nastran in a similar way as C Z ri gid but includes the effects of
structural deformations on the stability derivative. Of these four methods the one which assumes
a rigid aircraft response (i.e. Case 3 C Z ri gid ) will be used as the reference case as this is consistent
with the original derivation of the Pratt gust load factor 263 .
Figure 3.11 shows the value of ∆ n z g calculated for load cases three to eight using the four
different values of C L α . At low Mach numbers (load case 3 & 4) the largest difference in load factor
¡ ¢
is 0.09, which is 3.5% of the C Z ri gid value, however, as the Mach number increases (increasing
load case number) the maximum difference grows to 0.36, or 16% of the reference value. This is
because the assumption that C L α = 2π becomes increasingly incorrect as Mach number increases.
76
3.3. BUG-T AEROELASTIC ANALYSIS
2
p
C L α = 2π / 1 − M 2
C L α = 2π
1.8
C L α = C Z f l exibl e
C L α = C Z ri gid
1.6
1.4
∆n z g
1.2
0.8
0.6
3 4 5 6 7 8
Load case number [-]
Discounting this result leads to a the maximum error at load case eight of 7.3%. In general, the
Prandtl-Glauert corrected C L α overestimates the load factor, except at low Mach numbers where
¡ ¢
C Z ri gid gives a slightly larger value. Furthermore the load factor predicted by the flexible aircraft
is consistently lower both the rigid and Prandtl-Glauert values. The impact of the different values
of ∆ n z gust on the critical loads envelope is assessed using the following process:
1. The loads envelope is calculated for all six beam loads in the wing, strut and jury-strut
components for each value of C L α . The loads envelope for each case is denoted P i , where i
is the case number.
¡ ¢
2. Each loads envelope is normalised with respect to the reference case C L α = C Z ri gid and the
change in loads is calculated. The normalisation for each case is P i = P i /P3 where i = 1, 2, 4.
The normalised
³ difference
´ between the loads envelope and the reference loads envelope is
given as ∆P i = P i − 1
3. The data from the different beam loads and components is then combined into a single
vector and the MATLAB programme histfit is used to calculate the distribution of the
change in loads assuming a normal distribution.
The resulting normal distribution plots are shown in Fig. 3.12. By examining these distributions
it is clear that the case where C L α = C Z f l exibl e shows the least variance which is expected as the
load factors for these two cases in Fig. 3.11 provide the most consistent match, however, the
77
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
maximum change in loads is still ±8%. The remaining two cases show a maximum difference
of between -12% and 10% which is alarming considering these loads are being used to size
the wing at the preliminary design stage. These distributions have been generated assuming
a normal distribution, which is not necessarily the case, however a more thorough discussion
on uncertainty modelling in the design process is beyond the scope of this thesis. Instead, this
study has demonstrated the scale of differences in critical beam loads that can be introduced by
uncertainty in the aircraft C L α value, which should be factored into the evaluation of a critical
loads envelope which uses quasi-static gust loads. This discussion is relevant as any vibration
suppression devices that are included in the truss structure must designed to withstand both
static and dynamic loads, therefore it is necessary for these loads to be accurately estimated
in order for the device to be correctly designed and its mass accurately accounted for. This is
especially important at the preliminary design stage as a situation could arise where a vibration
suppression scheme was rejected as a result of over-estimating the required design loads leading
to a device with an unacceptable mass penalty. Further discussions on the practical design of a
vibration suppression device are provided Chapter 6.
250
C L α = C Z f l exibl e
C L α = 2π
p
200 C L α = 2π/ 1 − M 2
No. occurences
150
100
50
0
−15 −10 −5 0 5 10
∆P beam [%]
F IGURE 3.12. Approximate normal distribution of change in critical beam loads with
respect to the case where C L α = C Z ri gid for quasi-static gust load cases 3 - 8.
78
3.3. BUG-T AEROELASTIC ANALYSIS
Aeroelastic flutter has been identified as one of the main limiting factors in the design of a
truss-braced wing aircraft 47,98,149,265,266 , therefore, to understand how a vibration suppression
device can be used to influence flutter this study will determine the complex flutter modeshapes
and critical flutter speed of the BUG-T model. These flutter modeshapes will be linked to the
normal modes discussed in Section 3.2.2 and this information will be used in Chapter 4 to inform
the optimisation of a device that can suppress aeroelastic flutter.
In this work the Nastran PK flutter method 82 is used to calculate the aeroelastic stability
of the model. In the PK method the unsteady aerodynamic terms are introduced as additional
frequency-dependent stiffness and damping terms, as in the ‘British’ flutter method, in contrast
to the K method which considers the aerodynamic terms as additional inertia terms 74 . Both
methods yield the same flutter speed, however as noted in Hassig 267 the two methods can
identify different flutter branches that have substantially different pre-flutter behaviour. As the
PK method is favoured within the industry 267 it will be used for all flutter analysis in this thesis.
The Nastran PK method is a slight variation of the original PK method proposed by Hassig 267
and has a flutter equation of the form
I 2 R
ρ c re f U∞ Q hh ρ U∞
" Ã ! Ã !#
2
Q hh
− M hh p + B hh − p + K hh − ξ = 0, (3.18)
4k 2
where M hh , C hh and K hh are the modal mass, damping and stiffness matrices as defined in Eqn.
I
3.5, ξ are the modal coordinates, p is the complex eigenvalue, Q hh is the aerodynamic damping
R I R
matrix and Q hh is the aerodynamic stiffness matrix. Note that Q hh and Q hh are calculated at
a set of user-defined Mach numbers m and reduced frequencies k at the start of the analysis
and an interpolation is performed14 for each new value of k which is used during the solution of
Eqn. 3.18. In Nastran the complex eigenvalue has the form p = ωγ + j ω, where ω is the circular
frequency and γ is the transient rate of decay - equal to g/2 where g is the fictitious structural
damping term from the K flutter method. The objective of the flutter analysis is to find the
complex eigenvalue p which simultaneously satisfies Eqn. 3.18 and the necessary condition
k = ω c re f /2U∞ . The solution is conducted iteratively for each mode in ξ by varying the value k
using the process described in Rodden and Johnson 82 and Hassig 267 for a given value of ρ and
U∞ . The flutter velocity (VF ) is then identified as the lowest speed where a complex eigenvalue
becomes purely imaginary, i.e. γ = g = 0 - note the flutter velocity is not an output of the solution
of Eqn. 3.18 and must be calculated during post-processing. It is worth noting that use of a
decay rate in the complex eigenvalue assumes that the amplitude of the oscillations is changing
with respect to time, however, this directly contradicts the assumption inherent to Theodorson’s
unsteady aerodynamic theory which is that the motion of the aerofoil is harmonic with constant
14 In this work both linear and spline interpolation schemes were traded using the interpolation algorithms within
Nastran. It was found that the flutter behaviour of the BUG-T model was consistent regardless of interpolation
scheme so the linear interpolation scheme was favoured for its lower computational cost.
79
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
TABLE 3.5. Flight point data for BUG-T flutter analysis. A range of velocities between
0.8VC and 1.15VD are used with a velocity increment dV . For each velocity a linear
variation in Mach number is assumed between MC and M MO and the density is
calculated using the International Standard Atmosphere 269 .
amplitude. In this regard the behaviour predicted by the PK method is only strictly true at
the point of flutter and at all other values it is only an approximation of the true damping and
frequency of the system. The rationale used by Hassig 267 is that the amplitude of the oscillations
is varying slowly with time therefore aerodynamics based on constant amplitude oscillations are
a fair approximation.
According to CS-25 an aircraft must demonstrate freedom from flutter at speeds up to 15%
above the dive speed (VD ) at all flight points of interest. However, in this study only the cruise
altitude is considered so as to limit the computational burden and preserve the focus on vibration
suppression as opposed to a detailed aircraft design process. Using the flight envelope data in
Appendix A a series of flight points have been generated at the SUGAR cruise altitude of 36000ft
which are summarised in in Table 3.5. Using these flight points the flutter equation was solved
for 1584 combinations of Mach number, density and velocity using the BUG-T normal modes
up to 50Hz with fully-fixed boundary conditions. Applying full-fixed boundary conditions means
additional aerodynamic forces from the rigid body modes are not accounted for in the solution,
however this approach is standard for a linear flutter analysis 268 so is acceptable for the work
in this thesis. The aerodynamic reduced frequencies and Mach number used to define the AIC
matrices for flutter are shown in Table 3.6. Here, the Mach numbers are the same as the flutter
flight points and the reduced frequency values are defined to satisfy a maximum value of 3.11 for
an aircraft velocity of 165m/s at an excitation frequency of 50Hz. Results using the M/k pairs in
Table 3.6 were checked against a refined distribution which used reduced frequency values with
increments of 0.01 up to 3.5 and good agreement was found for the flutter modeshapes and the
V-g/V-k plots, therefore all subsequent flutter analysis uses the M/k values in Table 3.6.
The results of the flutter sweep were grouped by Mach number and the structural damping
term g was extracted for every branch in the solution. Any branch which had a positive value
of g at any point was deemed an unstable branch and the complete velocity-damping (V-g) and
velocity-reduced frequency (V-k) data was retained for plotting. Two branches were identified
80
3.3. BUG-T AEROELASTIC ANALYSIS
TABLE 3.6. Reduced frequencies and Mach numbers for calculating the AIC matrices
during the Nastran PK flutter analysis.
as having unstable components across a range of Mach numbers, branches seven and ten. For
both branches the flutter velocity was determined for each Mach number by performing a linear
interpolation of the V-g flutter data either side of g = 0, note as the velocity increment was set to
1m/s any error introduced by the interpolation should be small. Figure 3.13 shows the variation
of flutter speed vs. Mach for these two unstable branches. It should be noted that this variation
does not match expected trends. As discussed by Jonsson et al. 268 , the flutter velocity should
have a slight dip as the Mach number increases before increasing sharply as the Mach number
approaches unity. This is known as the transonic flutter dip and whilst it is not accurately
captured by DLM aerodynamic theory is should still be present. The reason for this discrepancy
is unclear and requires further investigation. However, as the focus of this section is simply to
identify a baseline flutter behaviour it is acceptable to proceed with these results so long as
flutter analysis in the next Chapter uses the same flight point data and reduced frequency/Mach
pairs, thus allowing a fair comparison to be made.
Figure 3.13 shows that the critical Mach number is 0.726, with a critical flutter velocity
of 212m/s for branch seven and a corresponding flutter velocity of 228m/s for branch ten. As
branch seven has the lowest flutter velocity it is the critical flutter branch however both branches
are evaluated in this section to understand their underlying characteristics. V-g and V-k plots
of these two branches at Mach = 0.726 are shown in Fig. 3.14. Here, both modes exhibit soft
flutter at speeds below the required stability margin of 1.15VD , although this is not surprising
given that the BUG-T model is an approximation of the SUGAR 765-095 Rev. D and does not
include additional modelling details such as downwash corrections due to aerofoil twist and
camber or pressure corrections from CFD analysis. Again, as the focus is on identifying the
BUG-T flutter mechanism it is acceptable to proceed with the current results. The two complex
aeroelastic flutter modes are shown in Figs. 3.15 and 3.16 with an inset panel showing a summary
of the strain energy across the different component sub-assemblies. A breakdown of the modal
coordinates15 for each normal mode that contributes to the complex mode is also provided in Figs
3.15 and 3.16. Values for the first 25 modes are shown as the contribution from higher modes (26
- 50) was found to be negligible. Note that as only the flexible modes were included in the analysis
the mode numbers in Figs. 3.15 and 3.16 have been adjusted during post-processing to match the
normal mode numbers from Section 3.2.2. By correlating the real translational displacements
15 Also referred to as the "modal participation factor" in the Nastran aeroelastic user guide 82 .
81
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
230
Branch 7
Branch 10
225
[m/s]
220
VF
215
210
0.7 0.72 0.74 0.76 0.78 0.8 0.82
Mach Number [−]
0.2
Branch 7
10
Branch 10
[%]
g = 3%
[−]
0.15
0
g
k
0.1
VC VD 1.15VD
−10
160 180 200 220 240 260 280 160 180 200 220 240 260 280
U∞ [m/s] U∞ [m/s]
(a) V-k plot at Mach 0.726 (b) V-g plot at Mach 0.726
F IGURE 3.14. V-k (a) and V-g (b) plots for Mach = 0.726.
of the flutter modeshapes with the modal coordinates, branch seven can be identified as a
global wing bending mode and branch ten as a global bend-twist mode. Examining the modal
coordinates shows that branch seven has a strong contribution from normal mode numbers seven
and eleven, which are the first wing out-of-plane bending and global bending modes shown in
Fig. 3.2. Furthermore normal modes eleven and thirteen have a large component of in-plane
bending which is not properly accounted for by Nastran’s DLM aerodynamic theory, implying
that these flutter modes may differ substantially if tested in a wind tunnel or flight test. Branch
ten is dominated by twist motion and its modal coordinates are more evenly spread than for
branch seven. The flutter modes of the BUG-T model do not match the SUGAR results presented
in Bradley et al. 30 , which is unsurprising as the SUGAR flutter analysis was conducted using
the full 3D model and not a simplified 1D beam model. Also, the Nastran AIC matrices were
augmented by additional pressure information from CFD analysis. Given the somewhat simpler
approach adopted in this thesis it is acceptable to proceed with the identified flutter mechanism.
82
3.3. BUG-T AEROELASTIC ANALYSIS
(a) Real translational terms of complex flutter modeshape. The translational displacements have been
multiplied by a scale factor of 100 and the shading of the plane represents the magnitude of the beam twist.
0.2
[−]
[−]
0.5
ξR
ξI
−0.2
0
5 10 15 20 5 10 15 20
Normal Mode Number [−] Normal Mode Number [−]
(b) Modal coordinates - Real. (c) Modal coordinates - Imaginary.
F IGURE 3.15. Flutter modeshape and complex modal coordinates for branch seven -
Mach 0.726, U∞ = 212m/s, k = 0.1361
83
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
(a) Real translational terms of complex flutter modeshape. The translational displacements have been
multiplied by a scale factor of 100 and the shading of the plane represents the magnitude of the beam twist.
0.2
1 Mode 14 Mode 12
Mode 12 Mode 8
Mode 8 0
0.5
[−]
[−]
−0.2
ξR
ξI
0 −0.4
−0.6
−0.5
5 10 15 20 5 10 15 20
Normal Mode Number [−] Normal Mode Number [−]
(b) Modal coordinates - Real. (c) Modal coordinates - Imaginary.
F IGURE 3.16. Flutter modeshape and complex modal coordinates for branch 10 - Mach
0.726, U∞ = 228m/s, k = 0.1345
84
3.3. BUG-T AEROELASTIC ANALYSIS
Regarding the locations of vibration suppression devices, the modeshapes of both unstable
branches show some rotation at the proposed device locations as well as a small amount of
compression across the jury-strut due to the combined bending of the wing and strut. Which
could imply that a vibration suppression device at one of the three locations could have a positive
influence on these unstable flutter modes if the correct device parameters can be determined.
This observation is based on visual inspection of Figs. 3.15(a) and 3.16(a) which only show the
real translation terms of the flutter modeshape and do not provide a complete picture of the
flutter mechanism. Examining the modal participation factors shows that branch ten has a large
participation from normal mode fourteen which was specifically identified in Section 3.2.3 as
having a large amount of rotation at location A, therefore it is likely that a device at one of
the candidate locations will be able to influence branch ten. However, branch seven is mostly
comprised of mode seven, eleven and thirteen, which are the first wing spanwise bending mode,
global wing bending mode and an in-plane bending mode respectively. These modes do not exhibit
significant relative motion at any of the device locations so it is likely that branch seven will be
less-sensitive to the effects of a device.
Figure 3.17 shows the relative modal displacement at the three device locations for the
flutter modeshapes in Fig. 3.15(a) and 3.16(a). These quantities have been calculated in a similar
manner as for Fig. 3.6, however as the flutter modeshape is complex the magnitude of the complex
85
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
displacements has been used as opposed to just the real or imaginary parts. These results shows
that the translational device at location B has a fairly consistent response for both unstable
branches, indicating that a device at this location may be effective in suppressing flutter if the
appropriate parameter values are chosen. For the rotational devices at locations A and C, it
is clear that Location A has a greater relative motion for both flutter branches, which could
mean that a device at the strut-root as opposed to the strut-tip is more favourable . As with the
forced-frequency response results in Fig. 3.7, it is likely that this increased motion at the strut
root is attributed to the lower stiffness value in the region inboard of the jury-strut. The use
of vibration suppression devices to influence the flutter behaviour of the BUG-T model will be
investigated in the next chapter.
Studies have been presented which determine the static aeroelastic and flutter response of the
BUG-T model and the following observations have been made:
• Static Aeroelastic Response - The static aeroelastic response of the BUG-T model
showed good agreement with the results from the SUGAR sizing process. Some discrepan-
cies were identified but these were associated with differences in the modelling processes
as opposed to an error in the model. Also, the magnitude of the tip displacements calculated
during the static aeroelastic analysis indicated that nonlinear aeroelastic effects should be
accounted for during the aeroelastic analysis of truss-braced wings.
• Static Aeroelastic Loads Envelope - The loads envelope was generated using the load
cases from the SUGAR sizing study. It was found that the dominant load cases where the
2.5g and -1g manoeuvre load cases at cruise altitude and the Mach 0.2 and 0.4 Pratt gusts
at sea level, which matches observations made during the SUGAR sizing process.
• Pratt Gust Loads Sensitivity - A sensitivity analysis of the effect of the C L α value on
the load factors and resulting loads envelopes of the Pratt gust load cases was carried
out. The aircraft rigid body vertical force coefficient was used as the baseline and it was
identified that a difference in loads of up to 12% is possible when assuming a value of
2π, falling to 8% when flexibility is accounted for in the aircraft vertical force coefficient.
As parts of the critical loads envelope are formed by Pratt gust loads it is important to
understand the limitations of the quasi-static assumption on the equivalent static loads.
Furthermore, as a vibration suppression device will need to withstand both static and
dynamic design loads it is important that these loads are accurately estimated at the
preliminary design stage.
• Flutter Modes - A sweep of Mach number and velocity has been carried out at the SUGAR
cruise altitude of 36,000ft in order to identify the flutter mechanism of the BUG-T model.
Two flutter modes have been identified, a global wing bending mode at 212m/s and a global
bend-twist mode at 228m/s for a Mach 0.726. Decomposing the complex flutter modes
86
3.4. CHAPTER SUMMARY
into the normal modes revealed that they are mostly dominated by flexible modes in the
frequency range 0-5Hz, with some rotation about the strut joint locations due to strut
bending. Based on the modal participation factors it was hypothesized that branch ten will
be more sensitive to the effects of a vibration suppression device due to its large component
from normal mode fourteen, whereas, branch seven is dominated by low frequency bending
modes which do not have significant relative motion at any of the proposed device locations.
This chapter has introduced the BUG-T aeroelastic finite element model which is based on the
SUGAR 765-095 Rev. D model from Bradley et al. 30 and provided an overview of the key dynamic
and aeroelastic behaviours that will be discussed in this thesis.
The dynamic analysis in Section 3.2 included a study of the normal modes of the structure
where different techniques were used to classify the structural modes. This process identified
several localised truss modes from which three possible locations for a vibration absorber were
identified - a rotational device at the strut-fuselage joint, a translational device at across the jury-
strut and a further rotational device at the strut-wing joint. Next, a frequency response analysis
was performed where the relative velocity at each potential device location was calculated based
on a unit harmonic load applied at the wing tip. It was found that the relative velocity for
each device location reaches a maximum in the 10-15Hz range, which corresponds to a series
of localised truss bending modes. At low frequencies the rotational devices experienced more
relative motion, with the strut root joint showing the largest amount owing to the increased
flexibility of the strut root region. As the low frequency modes tend to be more important during
aeroelastic interactions, it was hypothesised that the rotational devices will be more effective
than the translational device at alleviating aeroelastic effects. Finally, as the device locations
favoured different frequencies, it is possible that multiple devices could be used to target the
structural response across different frequency ranges.
In Section 3.3 the static aeroelastic response of the BUG-T model was calculated using the
aeroelastic load cases from Bradley et al. 30 and the results were compared against the data from
the SUGAR TBW sizing study. Good agreement was found for the manoeuvre load cases, however
the Pratt gust load cases showed differences of up to 30% in the calculated tip displacement.
This was attributed to the fact that the SUGAR sizing process applied additional scaling factors
to the Pratt gust loads in order to emulate dynamic gust loads, however these scaling factors
where not applied during the analysis presented in Section 3.3.2. The magnitude of the tip
displacements approached 10% of the semi-span, which implies that nonlinear aeroelastic effects
due to large displacements are likely to be important in the aeroelastic analysis of braced-wing
aircraft. The sensitivity of the calculated quasi-static load factor and resulting loads envelope to
the aircraft C L α value was investigated. It was found that differences in loads of up to 12% and
87
CHAPTER 3. REFERENCE MODEL AND CHARACTERISTIC BEHAVIOUR
10% are possible when using the compressibility corrected and uncorrected thin-aerofoil value of
2π respectively, falling to 8% when the Nastran calculated value was used. As parts of the critical
loads envelope are formed by Pratt gust loads it is important to understand the limitations of
the quasi-static assumption on the equivalent static loads, especially at the preliminary design
stage where accurate predictions of the airframe loads would be required to perform trade studies
on the potential benefit of novel concepts, such as aeroelastic control via vibration suppression
devices.
The flutter mechanism of the BUG-T model was calculated in Section 3.3.3 by performing
a sweep of Mach number and velocity at the SUGAR cruise altitude of 36,000ft. Two flutter
modes were identified, a global wing bending mode at 212m/s and a bend-twist mode at 228m/s
for a critical Mach of 0.726. Decomposing the complex flutter modes into the the normal modes
revealed that they are mostly dominated by flexible modes in the frequency range 0-5Hz, with
some rotation about the strut joint locations due to strut bending. Based on the observations in
Section 3.2 it is likely that a rotational device at the strut root or tip joint location will be able to
influence the response of these flutter modes.
In the next chapter the methods and observations made in this chapter will be used to
implement an optimisation approach where vibration suppression devices are included in the
model and their properties are optimised to improve the flutter speed of the BUG-T model. The
devices will be placed at the three candidate locations shown in Fig. 3.5 and different layouts will
be investigated, including layouts that can be frequency tuned to match the flexible modes of the
structure.
88
HAPTER
4
C
PASSIVE F LUTTER S UPPRESSION U SING V IBRATION A BSORBERS
his chapter investigates the potential for passive vibration absorbers to provide flutter
T suppression in truss-braced wings. Devices are placed at the candidate locations identified
in the previous chapter and three different device concepts are considered, including
devices that can be frequency-tuned to a flexible mode in the primary system. A MATLAB-
based optimisation scheme is adopted which uses Nastran to calculate the flutter response and
associated sensitivities in order to determine device parameters that increase the flutter speed.
Using this method it is shown that improvements in flutter speed between 1 - 6% are possible
for a range of device layouts and locations, although the effectiveness of the device is heavily
dependent on the flutter mechanism. In general a viscous damper at the strut-root joint has the
most effect on the flutter speed, however it is noted that a tunable device can provide comparable
performance to the viscous damper for the same viscous damping coefficient.
This chapter is formatted as follows: In Section 4.1 the different methods for modelling a
generic vibration suppression device are discussed and the candidate device layouts and their
associated admittance functions are introduced. In Section 4.2 a parameter study is presented
where the stiffness, damping coefficient and inertance at each device location is varied to simulate
the effect of a linear spring, damper or inerter at one of the device locations. In Section 4.3 a
combined MATLAB-Nastran optimisation scheme is introduced and the force coefficients of three
absorber layouts are optimised in order to increase the flutter speed. Cases with both single and
multiple devices are considered to determine the range of performance improvements available.
89
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
This section outlines two methods for modelling a generic vibration absorber attached to any
primary system, with additional detail provided for the Nastran-specific method adopted in
this work1 . Also, the candidate device layouts for flutter suppression are introduced and their
characteristics are discussed with reference to the ability of some of the layouts to be frequency-
tuned to match a resonant frequency in the primary system. It is noted that the modelling
methods and device layouts discussed in this section are not specific to aerospace and could be
applied to any mechanical system.
In the field of mechanical network design a generic vibration suppression device can be repre-
sented as a collection of mass, spring, damper and inerter elements which are joined in series
of parallel in a similar fashion to an electrical network diagram 208 . As discussed in Chapter 2,
the purpose of mechanical network design is to determine the optimum layout of a collection of
mechanical elements, and their corresponding force coefficient values, such that the vibration
suppression device limits some adverse dynamic behaviour in the primary system to which it is
attached. Two approaches for modelling the device are prevalent in the literature:
• Discrete mechanical elements - Any generic device can be catered for by connecting
the individual mechanical elements to the necessary degrees of freedom. If required by the
device layout the state-vector can be agumented by additional degrees of freedom to account
for any ‘internal’ device DOFs, see Figs. 4.1(b) and 4.1(c). Using this approach means the
mechanical elements generate additional terms in the system matrices, as in Eqn. 4.1.
• Admittance function - The transfer function relating the device force at each terminal
to the relative velocity acting across the device can be constructed for a generic layout of
mechanical elements using the process introduced by Firestone 210 . In keeping with the
mechanical-electrical analogy discussed in Chapter 2, this transfer function is referred
to as the device admittance function 208 and is typically expressed in the Laplace domain
as Y (s) = F/v. Examples of the magnitude of this quantity as a function of frequency are
shown in Fig. 4.2 for the three device layouts in Fig. 4.1. When modelled in this way the
vibration suppression device enters the equations of motion as state-dependent forcing
terms on the right-hand side.
The method of device modelling is dependent on the chosen method of device optimisation, i.e
immittance vs. structure-based. For example, if the immittance-based 209 approach is adopted then
the device must be represented by its transfer function. This is because the individual mechanical
elements are not available until after the optimised admittance functions are converted to
viable device layouts using network synthesis algorithms 211,212 . Whereas if the structure-based
1 A more detailed discussion on modelling vibration absorbers in Nastran is available in Appendix E
90
4.1. DEVICE MODELLING AND CANDIDATE LAYOUTS
approach is chosen then either of the modelling methods can be used, however for large and
complex networks the derivation of the device transfer function can become unwieldy and it may
be preferable to model the device using individual elements.
In this work the approach is taken to model the device using discrete mechanical elements.
This is because the ability to model, and more importantly optimise, individual spring, damper
and inerter elements is readily available within Nastran 270 . Despite this capability, the technique
of modelling a generic vibration suppression device using a commercial FE tool such as Nastran
has not been exploited by the mechanical network community. In effect, the device can be thought
of as a sub-structure within the larger model assembly with, for the case where one device is
used, the existing structural mass, damping and stiffness matrices, M s , C s and K s respectively,
augmented as 206
" # " # " #
M s + m d wwT −m d w C s + c d wwT −cd w K s + k d wwT −k d w
M= ,C = ,K = (4.1)
−m d wT md − c d wT cd −k d wT kd
where subscript s and d denote the terms belonging to the structure and the device respectively
and the term w is a column vector of zeros and ones describing the connectivity between the
device and the host structure. Note that multiple devices can be catered for by using Equation
4.1 iteratively. For the case where the elements are connected in series, new DOFs are introduced
to the system via internal device-DOFs. In this instance the displacement-vector is expanded to
include the new device DOF, xT = [xs , xd ]T where x contains the independent DOFs of the system
which for Nastran are the displacement of the nodes in the local coordinate system. The internal
DOF introduces a new mode into the system which can be be tuned to match the frequency of one
of the flexible modes of the host structure.
In terms of Nastran-specific implementations of a generic mechanical network, the Nastran
equivalents of the spring, damper and inerter are the CELAS, CDAMP and CMASS elements respec-
tively. Using this approach it is possible to model any generic mechanical network by joining
together these elements and include this network in a FE model of arbitrary size, fundamentally
enhancing the potential for mechanical network design methods to be applied to large and
complex models. It is also possible to model the network admittance function within Nastran
using the TF element. An example of this approach was demonstrated in the context of gust loads
alleviation in a TBW by this author 271 , however it should be noted that there are some caveats
when this approach is adopted in Nastran, further details can be found in Appendix E.
As this thesis represents a preliminary study into the use of vibration suppression devices in
TBWs it is favourable to adopt the structure-based approach for the design of the device. This
will allow the complexity of the device to be tightly controlled and also enables key aspects of the
device topology to be investigated. Regarding the device layout, the main areas of interest are:
91
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
1. Damping vs. frequency-tuning - A linear viscous damper is usually the first vibration
suppression concept that is considered, however, as mentioned in Chapter 2 this represents
only one possible mechanism for removing energy from the primary system. To take
advantage of the vibration absorber effect 198,199 it is necessary to consider layouts which
have internal ‘device-DOFs’ which can be frequency-tuned to one of the natural frequencies
of the primary system. Hence a key outcome of this chapter is whether a tunable device
can offer superior performance over the pure damping case.
2. Physical device model vs. conceptual - Without a proper understanding of how the
devices proposed by mechanical network design algorithms are physically realised it is not
possible to understand their impact on the aircraft response or ascertain how they will be
included in the design process. Therefore, as well as considering a variety of conceptual
device layouts, physical models based on examples from the literature 191,208,228 will be
investigated. This will allow the viability of the proposed devices to be determined via the
use of simple design rules to estimate the required physical parameters which will yield
the optimised force coefficients. The results of this study are presented in Chapter 6.
Bearing these points in mind, the following conceptual devices are considered in this thesis:
The layouts and associated admittance functions for these three candidate devices are provided
in Fig. 4.1. The process of deriving the device admittances is well documented throughout the
mechanical network literature and so only a brief overview is provided here. First, the admittance
of each individual element is defined as shown in Fig. 2.16. Next, the admittances of each branch
in the network are found by combining the elements that are in series or parallel following the
process discussed in Firestone 210 . Finally the admittance of each branch can be combined to
create the overall admittance function of the device. For a specific derivation of the admittance of
the TID element the reader is directed to the paper by Lazar et al. 220 .
92
4.1. DEVICE MODELLING AND CANDIDATE LAYOUTS
b T ID ( c T ID s+ k T ID ) s b T ID ( c T ID s+ k T ID ) s
Y (s) = c Y (s) = b T ID s2 + c T ID s+ k T ID
Y (s) = b T ID s2 + c T ID s+ k T ID
+c
T2 T2 T2
c T ID k T ID c T ID k T ID
c c
b T ID b T ID
T1 T1 T1
(a) Damper (b) TID (c) TID-D
F IGURE 4.1. Candidate device layouts and associated admittance functions for the
damper (a), Tuned-Inerter-Damper (TID) (b) and the Tuned-Inerter-Damper-
Damper (TID-D) (c).
Figure 4.2 shows the admittances for the three candidate device layouts as a function of
frequency. In contrast to the damper which has a constant value across all frequencies, the TID
and TID-D devices can target a particular frequency by selecting the inertance and stiffness
values appropriately. The exact tuning condition of the device is dependent on the background
effects of the host structure 206 , however for a TID device in isolation the formula ω2 = k/b will
yield the pole of the admittance function. The advantages of a TID device over the conventional
YT ID −D
YT ID
YDamper
¯F ¯
¯ ¯
v
ω2T ID = k T ID /b T ID
Frequency
93
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
TMD are twofold: Firstly, the use of the inerter element allows a force to be imparted at both
terminals of the device, something that is particularly useful for the TBW configuration as it
allows for the possibility of a device that can influence multiple parts of the structure. Secondly,
due to internal gearing within the inerter it can provide an inertance which is far greater than
the mass of the device, this has obvious benefits in an aerospace application as the weight of the
aircraft should be kept to a minimum.
Before proceeding with the optimisation of the vibration suppression device layouts detailed in
Section 4.3 it is sensible to conduct a simple parameter study to understand the variation in
flutter behaviour resulting from the addition of a device. Here the effect of separately increasing
the stiffness, damping or inertance at the three candidate device locations on the critical flutter
speed is investigated - note, this is analogous to placing a single spring, damper or inerter at one
of the three candidate locations and increasing the force coefficient value. Table 4.1 shows the
upper and lower bounds of the parameters for the mechanical elements considered in this study.
Here the inertance value is expressed as a fraction of the primary system mass which for this
study is 3781kg, i.e. the half-wing mass of the SUGAR 765-095 Rev. D model2 . The half-wing
mass is used for the primary system mass as devices are placed in both the port and starboard
strut/jury-strut components of the full BUG-T model. Note that as an inerter is used the actual
mass of the device will be far less than 10% of the wing mass. For instance, a typical value of
the inertance-to-mass ratio is around 40-80, see for example the commercially available device
tested in Gonzalez-Buelga et al. 223 , although devices have been produced with ratios as high
as 300 208 . The TID or TID-D devices are not considered in this parameter study as previous
research indicated that the tuning of a vibration absorber is heavily dependent on background
flexibility effects 206 , which would likely have necessitated a very fine distribution of stiffness and
inertance values to capture the correct tuning. The output of the parameter study is the flutter
velocity of the two unstable branches identified in Chapter 3 for each parameter value3 . This
2 See Appendix A for further details on the mass properties of the SUGAR 765-095 Rev. D model.
3 In the post-processing step the first 30 flutter branches were checked for instability but only the two flutter
TABLE 4.1. Parameter values for the spring, damper and inerter elements used in the
BUG-T flutter study.
94
4.2. PARAMETER STUDY OF BUG-T FLUTTER RESPONSE
Wing LocationWC
Jury-Strut
FlexibleWBeams
LocationWB PinnedWJoints
Strut Location Label DeviceWType
LocationWA A Strut-Root Rotational
B Jury-Strut Translational
C Strut-Tip Rotational
F IGURE 4.3. Location of vibration suppression devices used for flutter suppression and
gust load alleviation. Devices will be placed at these locations in both the starboard
and port components.
velocity is found by selecting the critical flutter branches and then interpolating the structural
damping terms either side of the stability boundary to find the velocity where the damping
has a zero value. The flight points from the flutter study in Chapter 3 are used with a velocity
increment of 1m/s to minimise mode-switching 272 and a Mach number of 0.726 is used as this
was determined to be the critical Mach number for the BUG-T model.
Figure 4.4 shows the change in flutter speed as a function of spring stiffness, damping
coefficient and inertance at the three candidate locations identified in Chapter 3, shown here in
Fig. 4.3. Results are shown for unstable branches seven and ten to test whether the effectiveness
of a device is dependent on the flutter mechanism, although seven is the critical branch. The first
observation is that varying the stiffness, damping and inertance at locations B and C has an
almost negligible effect on the flutter velocity. Specifically, for branch seven there is a maximum
change of 0.24%, 0.23% and 0.4% for the spring, damper and inerter cases respectively, whereas
for branch ten there is a small decrease in flutter velocity of approximately 0.3%. Also, for both
locations the flutter speed decreases as the parameter value is increased. The scale of these
changes imply that a device at location B or C will have little impact on the flutter behaviour of the
BUG-T model. The results for location A are much more promising. Considering both rotational
spring and damper devices there is a clear increase in flutter velocity of approximately 3.5%
for branch seven and 7.5% for branch ten. For the pure damper case there is a clear maximum
at a viscous damping coefficient value of approximately 2 × 106 whilst the pure stiffness case
reaches a plateau for spring constants greater than 1 × 108 . Interestingly the scale of change
in flutter speed is the same regardless of whether the joint stiffness or damping is increased,
which implies that they are modifying the flutter mechanism in the same way. In fact, the effect
of increasing the stiffness or damping at the joint is analogous to locking up the joint, as shown
by the dashed line in Figs. 4.4(a) to 4.4(d), labelled as "Fixed at A". Here, it is clear that for
large values of stiffness and damping the flutter behaviour tends towards the case of a fully-fixed
joint at location A, which occurs for damping values greater than 105 and spring stiffness values
greater than 108 . This is viewed as unfavourable as the SUGAR MDO study identified pinned
95
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
8 8
Location A Location A
6 Location B 6 Location B
[%]
[%]
Location C Location C
4 4
Fixed at A Fixed at A
∆VF
∆VF
2 2
0 0
−2 −2
101 103 105 107 109 1011 101 103 105 107 109 1011
Spring Stiffness Spring Stiffness
8 8
Location A Location A
6 Location B 6 Location B
[%]
[%]
Location C Location C
4 4
Fixed at A Fixed at A
∆VF
∆VF
2 2
0 0
−2 −2
101 103 105 107 101 103 105 107
Damping Coefficient Damping Coefficient
0 0
[%]
[%]
∆VF
∆VF
−1 −1
Location A Location A
Location B Location B
Location C Location C
−2 −2
0 5 10 15 20 25 0 5 10 15 20 25
Inertance-to-Mass Ratio [%] Inertance-to-Mass Ratio [%]
F IGURE 4.4. Change in flutter velocity for unstable branches seven and ten as a
function of joint stiffness, damping and inertance. The original flutter velocity of
branches seven and ten are 212m/s and 228m/s respectively.
96
4.2. PARAMETER STUDY OF BUG-T FLUTTER RESPONSE
joints as providing a lower wing structural mass so it would be advantageous to preserve the
pinned behaviour. That being said, a pure pinned connection is overly optimistic as in a real
design there will be additional forces acting on the joint, such as friction. Therefore, if the pinned
behaviour is considered as the upper bound of realisable flutter benefits then a damping value
between 103 - 105 or a spring stiffness between 105 - 107 will yield an improvement of 1 - 6% in
the flutter speed depending on the flutter mechanism. Finally, introducing an inerter at location
A had a negative effect on the flutter speed for both branches and for all mass ratios considered,
however, the mechanism for this is unclear and requires further investigation.
• Flutter modes that exhibit more relative motion at the device locations will be more
susceptible to the influence of the device. This may appear to be obvious however it is
important to recognise that the correct type of device could have a positive effect on the
flutter speed. Of the three device locations considered location A has the most influence on
the flutter behaviour which matches the predictions from Chapter 3.
• Increasing the stiffness or damping at location A yields similar improvements in the
flutter speed - approximately 4% for branch seven and 7.5% for branch 10. This level of
improvement is quite significant in the context of the overall design of the TBW aircraft.
For example, Mallik et al. 47 demonstrated that a 15% increase in the flutter speed equated
to a 5% decrease in fuel burn. Therefore it follows that if the critical flutter mode of a
TBW exhibited large relative motion at one of the joint locations then a device could
be designed to influence that flutter mode and likely improve the mission performance.
However, additional factors such as the mass of the device and the effect of increased loads
97
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
This section presents an optimisation study of the candidate device layouts presented in Section
4.1.2. The objective of this study is to identify whether an improvement in the flutter speed is
possible by including one or more of these devices in the structure. The design variables are the
force coefficient values of the various mechanical elements in each layout and the optimisation
is conducted using a gradient-based algorithm available within MATLAB. The Nastran design
sensitivity and optimisation solution sequence, Solution 200, is used in this chapter to perform
the flutter analysis and compute the sensitivity of the flutter damping values with respect to the
device properties. It is shown that improvements in flutter speed between 1 - 6% are possible
depending on the device layout and location, however it is noted that this is highly dependent of
the flutter mechanism and the amount of relative motion available at the device locations.
Structural optimisation methods which consider aeroelastic flutter have been commonplace in
the aerospace industry since the early 1980s 273 , with ASTROS 274 and Nastran Solution 200 270
being examples of programmes that are widely accepted throughout the industry. In a standard
aircraft optimisation process flutter is typically regarded as a constraint 268 , although some
authors have considered the flutter speed as the objective function - see for example the paper by
Guo et al. 275 . As this chapter is investigating the potential of vibration suppression devices to
provide flutter suppression it is sensible to pose the optimisation problem in such a way that the
objective function leads to the flutter speed being increased. However the exact calculation of
the flutter speed can require substantial resources 274 so it is advantageous to select an objective
function that does not require additional post-processing and can use the output from a standard
Nastran flutter analysis. Here this is achieved by minimising the maximum value of the artificial
structural damping term (g max ) needed to ensure that the damping values across the modes are
less than the required damping value for all excitations considered4 . This can be expressed as
the following optimisation statement
¡ ¢
min (g(x)max ) subject to g(x) − g req ≤ 0, (4.2)
4 Using the Nastran convention a negative value of g is considered stable and a positive value is unstable.
98
4.3. OPTIMISING FOR FLUTTER SUPPRESSION
g g max
VF
g = g req
U∞
Branch 1 Branch 2
Branch 3
(a) Example flutter response showing the flutter speed VF and the maximum
structural damping value g max .
g
g max
VF
g = g req
∆VF U∞
Branch 1
Branch 2
Branch 3
(b) Sketch of an optimised flutter response by minimising the value of g max .
F IGURE 4.5. Example of the original (a) and optimised (b) flutter response. In this
example an increase in flutter speed ∆VF has been achieved by minimising the
maximum structural damping term (g max ). Once this point ceases to be critical the
optimiser will target the next critical point which in this example lies on branch
two.
where x are the design variables, which for this study are the force coefficients of the various
spring, damper and inerter elements in the chosen device layout, g max is the maximum value of
the artificial structural damping response, g is a vector of artificial structural damping responses
for all flight points considered and g req is the required damping value - defined as 0.03 in the
certification documents 234 . As noted by Neill et al. 274 , an optimisation approach that considers
flutter in terms of the structural damping values is advantageous as it does not require the
flutter speed to be calculated and is also insensitive to mode switching. Furthermore, this method
can be extended to include an arbitrary number of flight points with different combinations of
Mach number, velocity and density. Meaning it is possible to evaluate the entire flight envelope
in a single optimisation and design a device that provides flutter suppression at the critical
flight points. A graphical explanation of this approach is provided in Fig. 4.5 and a description is
provided below:
1. At each iteration of the optimisation the damping value g is calculated for a set of discrete
aircraft velocities, Mach numbers and densities, yielding the standard V-g flutter response.
The damping values can joined to form the branches of each flutter mode, as in Fig. 4.5(a),
99
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
There exists within Nastran a mutlti-disciplinary design and optimisation solution known as
solution sequence (SOL) 200 270 . This is a mutli-disciplinary tool that allows optimisation of a
set of model-based design variables subject to the response from a variety of analysis types such
as: statics, normal modes and frequency response as well as static aeroelastic and aeroelastic
stability (flutter). The main advantage of using SOL 200 is that it computes the gradients
of response quantities with respect to the model properties analytically, meaning that the
optimisation requires fewer function evaluations and consequently is much quicker than an
algorithm that uses finite differences to estimate the gradients/design sensitivities. This is an
important consideration when applying the principles of mechanical network design to complex
mechanical structures that can contain thousands or even millions of DOFs. However, the caveat
is that sensitivity information can only be calculated when the linear form of the structural and
aerodynamic formulations are used, meaning nonlinearites cannot be included in the model.
This could be seen as a drawback by researchers within the field of mechanical network design
as many of device models contain nonlinearities 228,230 , indeed even a simple damper element
100
4.3. OPTIMISING FOR FLUTTER SUPPRESSION
quickly becomes complex once the physical damping mechanism is considered 189 . The rational
adopted here is that at the initial design stage the precise nonlinear behaviour of the device
is less important than determining whether there is a benefit to including a device in the first
place. Therefore, it is acceptable to only consider linear force-displacement/velocity/acceleration
relationships for the various mechanical elements with the condition that the effect of device
nonlinearities on the force coefficients is understood and evaluated at the next stage of the design
process, as in Rittweger et al. 191 .
The optimisation process described in Section 4.3.1 requires the flutter structural damping
values to be considered as an objective in the design optimisation, however such an approach is
not possible within SOL 200 as the flutter damping terms can only be considered as contraints 270 .
Therefore, to bypass this problem an alternative optimisation algorithm must be used which
invokes Nastran externally and extracts the flutter response at each iteration of the optimisation
before constructing the relevant objective and constraint terms. Furthermore, to take advantage
of the analytical gradient capability of SOL 200 it is sensible to consider a gradient-based
optimisation algorithm provided that the design space is convex5 or can be traversed in such a
way that all possible local minima are found. To achieve this, the MATLAB optimisation function
fmincon is chosen along with the standard interior-point algorithm 276 , and multiple starting
points are used to test the global vs. local nature of the design space. An overview of this process
is provided in Fig. 4.6 which shows the transfer of data between the optimiser and Nastran and a
description of each of the block is provided below:
1. Initial conditions - The starting values of all the design variables in the optimisation
problem are referred to as the initial conditions and are termed x0 . In this study the effect
of the initial conditions on the final solution is evaluated by starting the optimisation from
ten different initial conditions for each device layout.
2. Optimiser - The fmincon optimisation function using the standard interior-point
algorithm in MATLAB is used to conduct the optimisation. The design variables are
normalised in the numerical domain, meaning that each variable has an equal weighting
in the problem. This removes any bias towards properties that have large magnitudes, for
example the device spring constant O (8) vs. the inertance O (2). In this study the design
variables are normalised using x = (v − vlb ) / (vub − vlb ), where x are the numerical values
of the design variables (the values which the optimiser has access to), v are the physical
values of the design variables (the values which are passed to Nastran) and vub and vub
are the upper and lower bounds of the physical design variables as given in Table 4.2.
3. Obtain physical design variables - The design variables are transformed into the
physical domain by performing the inverse normalisation, v = x (vub − vlb ) + vlb , and the
necessary Nastran files are written using the current value of the physical design variables.
4. Calculate flutter response and sensitivities - The flutter response (g) and the sensi-
5 A convex design space is one which has only one local minimum point which is also the global minimum.
101
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
tivities of flutter structural damping terms with respect to the device parameters (∂ g/∂v)
are calculated analytically by SOL 200. The flutter analysis uses an altitude of 36,000ft
which yields a density ratio of 0.298, as in Table 3.5. The Mach number is set to 0.726 as
this was determined to be the critical Mach number and a velocity increment of 10m/s is
used in order to limit the computational burden of the optimisation but still capture the key
behaviour of the flutter branches as explained in Neill et al. 274 . Once the flutter analysis is
complete the response data is extracted and passed back to the MATLAB environment.
5a. Convert cost function and gradient to numerical domain - The sensitivity infor-
mation provided by SOL 200 is calculated with respect to the physical design variables,
however the optimiser requires this information in the numerical domain. So to trans-
form these sensitivities back into the numerical domain the chain rule is used where
∂v/∂ x = vub − vlb , although no such transformation is required for the cost function.
5b. Convert constraint function and gradients to numerical domain - The same as
block 5a but for the constraint sensitivities.
J ( x) = J ( v)
∂ J ( x)/∂ x = ∂ J (v)/∂v × ∂v/∂ x
5a.
J ( x), ∂ J ( x)/∂ x Physical to J (v), d J (v)/ dv
numerical
v = f ( x, vlb , vub )
2. 4.
3. SOL 200
1.
x0 fmincon
x Numerical v Responses &
Start Optimisation
to physical sensitivities
(MATLAB)
(Nastran)
5b.
Physical to
C ( x), ∂C ( x)/∂ x C (v), ∂C (v)/∂v
numerical
C ( x) = C ( v)
∂C ( x)/∂ x = ∂C (v)/∂v × ∂v/∂ x
F IGURE 4.6. Block diagram for the combined MATLAB-Nastran optimisation process
used to optimise the flutter response. The data flow for each branch is annotated
to emphasise the division between the numerical optimisation domain and the
physical analysis domain.
102
4.3. OPTIMISING FOR FLUTTER SUPPRESSION
TABLE 4.2. Design variable bounds for the single absorber flutter optimisation.
Once the optimisation has terminated an additional flutter analysis is performed using the device
parameters from the last iteration and a refined velocity increment of 1m/s, which provides an
improved estimate of the flutter velocity. With the exception of this final flutter analysis, the
optimisation scheme is generic and sufficiently modularised such that different optimisation
problems could be considered so long as they fall within the scope of Nastran SOL 200. For
instance, using this approach it would be possible to conduct a structural optimisation of the
airframe and incorporate the design of the device in a combined structural-device optimisation.
Or a different primary system could be considered (such as a car) and a suspension system could
be designed within the scope of a larger structural optimisation problem.
Regarding the initial conditions for the optimisation process, the upper and lower bounds for
the various mechanical elements are shown in Table 4.2. Note that as the inerter element did not
have much influence on the flutter behaviour during the initial parameter study it was decided
to limit the inertance-to-mass ratio to 10% of the primary system mass to force the optimiser
to consider devices with a lower mass. Ten different initial conditions at regular intervals of
between 10% and 100% of the bounds are used and any design variables which span more than
two orders of magnitude, such as the spring stiffness and damping coefficient, are mapped to an
exponential distribution to ensure good coverage of the design space.
In this section a single vibration absorber is incorporated into the structure and the optimisation
scheme detailed in Section 4.3.1 and 4.3.2 is used to obtain the force coefficients of the mechanical
elements which yield an increase in the flutter speed. The three device layouts are the damper,
TID and TID-D devices detailed in Section 4.1.2 and the device locations are the same as those
used in the parameter study. In total, three device layouts were tested at three different locations
for ten sets of initial conditions which resulted in 90 different optimisations. In all these cases
the optimiser terminated because the relative change in all design variables was less than
the tolerance value of 1 × 10−6 . This means that flutter was not completely suppressed for all
velocities of interest but some change in flutter speed was achieved by the optimisation process.
The resulting V-g curves of unstable branches seven and ten are shown in Fig. 4.7 for a selection
of device configurations. Here, the change in flutter speed due to the addition of a vibration
absorber is clearly shown by a right shift of the V-g curve, as described in Fig. 4.5. A summary of
the maximum change in flutter speed is given in Fig. 4.8 for unstable branches seven and ten for
103
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
(a) Branch 7
(b) Branch 10
F IGURE 4.7. Comparison of baseline and optimised V-g curves at Mach 0.726 for flutter
branch seven (a) and branch ten (b).
104
4.3. OPTIMISING FOR FLUTTER SUPPRESSION
all combinations of devices and locations and the corresponding parameter values are provided in
Table 4.3 - note that these results represent the best possible performance out of the ten different
starting positions.
From Fig. 4.8(a), the results for branch seven show that the damper and TID-D layouts achieve
the best performance resulting in an increase in flutter speed of approximately 3.5%, which agrees
well with the predictions from the parameter study presented in Section 4.2. Examining the
parameter values for the damper and TID-D at location A reveals that the parallel damper has
opted for the maximum permitted damping coefficient and the TID element of the TID-D has also
opted for a high internal damping value. The tuning frequency of the inerter and spring in the
TID-D is a low enough frequency to imply that the device has attempted to tune itself to one of the
normal modes of the structure, however the two closest normal modes are modes 15 and 16 which
do not feature heavily in the flutter mode. Interestingly, the TID device at Location A has achieved
a tuning frequency of 2.27Hz which is comparable with one of the low frequency modes that have
large participations in the flutter mode. Regardless of device layout the maximum change in
flutter speed for a device at location B or C is approximately 1.75%, which is better than any of
the results from the parameter study but falls short of the performance improvements possible for
a device at location A. Furthermore, for a given viscous damping coefficient the tunable devices
have marginally better performance than the pure damper case, which shows that there is some
benefit to incorporating tunable elements in the device design, although the TID damping values
at locations B and C are larger than for location A. Also the tuning frequencies of the device
at location B are much higher than location A or C, which could be an attempt to maximise
the relative motion across the device terminals as location B is not particularly active at low
frequencies. All tunable devices opted for large values of inertance which implies the device is
seeking the largest possible separation between the two modes generated by tuning the device,
as in Section 2.3. Finally, Fig. 4.8(b) shows the change in flutter speed for branch ten for the
parameter values in Table 4.3. As with branch seven the damper and TID-D devices show the
most favourable performance and the TID at location A shows only minor improvements in the
flutter speed. All other devices at locations B and C have yielded a slight decrease in the flutter
speed, however as branch ten is not the critical branch this is acceptable so long as it does not
lead to an overall reduction in flutter speed.
To understand how the vibration suppression devices affect the flutter mechanism the complex
flutter modeshape for branch seven has been extracted for the case where a damper is placed at
location A with viscous damping coefficient 1 × 105 . The resulting change in modal coordinates is
shown in Figure 4.9. Here, the increase in flutter speed has manifested itself as small changes
in the modal coordinates that make-up the complex flutter modeshape with the largest changes
occurring for normal modes seven, nine and thirteen6 . Each of these modes contain out-of-plane
bending motion in the wing and strut components that generates aerodynamic stiffness and
105
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
4 TID
3.68%
3.55% TID-D
Damper
[%]
3
∆VF
2.34%
8
6.43% 6.18% TID
6 TID-D
Damper
[%]
4
∆VF
2
0.41%
0
−0.48% −0.45% −0.45% −0.46% −0.47% −0.47%
−2
Location A Location B Location C
(b) Change in flutter speed for branch 10
F IGURE 4.8. Maximum change in flutter speed for the single absorber optimisation.
damping, so it makes sense for the device to influence these modes in order to obtain a more
favourable flutter performance. A similar analysis cannot be performed for the TID and TID-D
devices as the intermediate DOF in these devices increases the number of modes in the system
meaning that a straightforward comparison cannot be made between the modified and unmodified
TABLE 4.3. Device parameter values for the single absorber optimisation.
106
4.3. OPTIMISING FOR FLUTTER SUPPRESSION
·10−2 ·10−2
6 1
4 0
[−]
[−]
2 −1
∆ξR
∆ξ I
0
−2
−2
−3
5 10 15 20 25 5 10 15 20 25
Normal Mode Number [−] Normal Mode Number [−]
(a) Change in modal coordinates - Real (b) Change in modal coordinates - Imaginary
F IGURE 4.9. Change in flutter modal coordinates when a damper is located at the
strut root joint with viscous damping coefficient 1 × 105 .
modes. An alternative approach could be to perform a frequency response analysis and identify
the points in the frequency domain where new modes have been introduced. If it were found that
the device tunes to a mode that contributes to the out-of-plane bending or twist motion of the
lifting surfaces then that would be an indicator of how a tunable device influences the flutter
mechanism.
The flutter behaviour of a 2DOF aerofoil section with a TMD attached at various points along
the aerofoil chord was investigated by Karpel 249 and more recently by Verstraelen et al. 251 . In
both of these works it was identified that the TMD was highly sensitive to the tuning condition of
the absorber and that the flutter speed deteriorated when the absorber operated away from its
optimum tuning frequency. To understand whether a similar problem will be encountered with
the tunable devices considered in this section a sensitivity study has been run where the viscous
damping coefficient and inertance of the TID at location A are varied by ±50% and the change in
flutter speed is calculated. The TID stiffness is not included in the sensitivity study as varying the
inertance has the effect of detuning the TID as shown by ω2T ID = k T ID /b T ID . Figure 4.10 shows
the change in flutter velocity for two seperate sensitivity studies. Examining Fig. 4.10(a) it is clear
that the TID device at location A has not been correctly tuned to maximise the flutter velocity
as there is a clear peak in the flutter speed for branch seven and branch ten at approximately
-34% and -42% of the reference inertance value. This is textbook behaviour for the case where
a vibration absorber is detuned from the primary system response and matches the findings of
previous research 249,251 . Branch ten is more susceptible to the TID tuning, which makes sense
as this flutter mode has been shown to be more sensitive to all of the device parameters as a
consequence of the normal modes participating in the complex flutter modeshape. Furthermore,
Fig. 4.10(b) shows that there is negligible change in the flutter velocity as the TID damping
107
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
3
2
[%]
[%]
2
∆VF
∆VF
1
1
0 0
−40 −20 0 20 40 −40 −20 0 20 40
∆ b T ID [%] ∆ c T ID [%]
F IGURE 4.10. Sensitivity of flutter speed to TID damping and tuning frequency for a
single TID at location A.
coefficient is varied. However, given that the fundamental theory of vibration absorbers states
that the tuning condition is a function of both absorber frequency and damping ratio, it is likely
that these results are a consequence of the datum TID parameters starting from an ‘untuned’
condition. These results demonstrate the importance of considering the robustness of tunable
devices such as the TID layout to changes in their parameter values and to off-design conditions.
Figure 4.11 shows the change in flutter speed for a damper at location A for each initial start
point. Here, there are two clear configurations that have been returned by the optimisation,
these are a damping coefficient of 9.1 × 103 and 1 × 105 yielding a change in flutter speed of
3.5 105
[ Nm/rads ]
-1
[%]
3
∆VF
2.5
c
104
2
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Initial Condition Case Number [-] Initial Condition Case Number [-]
(a) Change in flutter speed (b) Final value of damping coefficient
F IGURE 4.11. Change in flutter speed and corresponding damping coefficient for a
single damper at location A with multiple start points.
108
4.3. OPTIMISING FOR FLUTTER SUPPRESSION
105
1.74
[Nm/rads-1 ]
104
[%]
1.73
∆VF
103
1.72
c
1.71 102
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Initial Condition Case Number [-] Initial Condition Case Number [-]
(a) Change in flutter speed (b) Damping coefficient of parallel damper
0.1
[Nm/rads-1 ]
8 · 10−2
104
[-]
µ
6 · 10−2
c T ID
103
4 · 10−2
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Initial Condition Case Number [-] Initial Condition Case Number [-]
(c) TID damping coefficient (d) Inertance-to-mass ratio
107
20
106
[Nm/rad]
[Hz]
105
10
f T ID
4
k T ID
10
103
0
1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10
Initial Condition Case Number [-] Initial Condition Case Number [-]
(e) TID stiffness (f) TID frequency
F IGURE 4.12. Change in flutter speed and corresponding device parameters for a
single TID-D device at location B with multiple start points.
109
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
2.12% and 3.55% respectively. The fact that there are two different configurations is a strong
indication that the design space is non-convex and that multiple devices configurations could
exist which satisfy some notional increase in the flutter speed. This also suggests that a different
optimisation algorithm would be required to determine the device parameters that provide the
largest possible flutter speed, for example a stochastic method as opposed to a gradient-based
algorithm. Interestingly the TID at location A has a slightly better performance than the pure
damper with a damping coefficient of 9.1 × 103 , suggesting that for certain damping values a
tunable device can offer better performance than the damper alone. Similar results were found
for the devices at locations B and C however in the interest of brevity only the results for a TID-D
device at location B are shown in Fig. 4.12. Here, every start point has converged on roughly
the same device performance but with vastly different parameter values. This is good from a
robustness point of view as it provides the designer with multiple options for achieving a given
improvement in flutter speed, however the mechanism by which a tunable device influences the
flutter mechanism needs to be better understood.
Finally, to check the results of the optimisation process are appropriate the design space has
been visualised for the case where a TID is placed at location A. Here, the objective function value
(g max ) and the change in flutter speed (∆VF ) have been calculated for an inertance-to-mass ratio of
0.1 (i.e. the optimised value), with TID frequencies between 0.25-10Hz and damping coefficients
between 50-500Nm/rads-1 . A flutter analysis was conducted for each unique combination of
parameter values, yielding over 2000 separate analyses7 . The results of this parameter study
are shown in Fig. 4.13, where the MATLAB triangulation algorithm trisurf has been used to
visualise the scattered dataset and a contour plot of the surface has been constructed using the
tricontour function from the MATLAB File Exchange 277 . Examining these results, it is clear
that the current optimisation process (shown by the red marker in Fig. 4.13) has failed to identify
the optimum absorber parameters. As mentioned at the start of this section, this is because the
optimisation process terminated prematurely due to the design variable step tolerance being
exceeded. Clearly if the optimiser is allowed to proceed beyond its current value an improved
device design is possible, as show by the peak in Fig. 4.13(b) which yields ∆VF = 5.36% for a TID
tuning frequency of 2.85Hz and a viscous damping coefficient of 50Nm/rads-1 . This is a greater
improvement than that identified by the pure damper case, although at operating frequencies
away from the tuning condition the TID performance deteriorates rapidly and ends up being
worse than the pure damper case. These results show a similar trend to the sensitivity study
shown in Fig. 4.10, although the magnitude of the change in flutter speed is slightly different as
here the inertance is fixed at µ = 0.1 whereas in Fig. 4.10 the inertance was varied and the TID
spring stiffness was constant.
7 Such a large number of runs was required in order to capture the tuning behaviour of the TID device
110
4.3. OPTIMISING FOR FLUTTER SUPPRESSION
F IGURE 4.13. Variation in maximum flutter damping ratio and flutter speed for a TID
at location A with a constant inertance of µ = 0.1.
111
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
In the frequency response analysis in Chapter 3 it was identified that different device locations
exhibit maximum relative motion at different excitation frequencies. It was hypothesized that
this could be utilised by including multiple devices in the structure that target the structural
response across a range of frequencies. Therefore this section presents a study where a TID
device is included simultaneously at the three candidate device locations and the force coefficients
are optimised in order to extend the flutter speed. TID devices are chosen for this study to test
whether the tuning capabilities of this layout can be used to target the response of different
structural modes, also a TID device at location A showed promising results for the single device
optimisation. Finally, as many of the TID and TID-D layouts in the single device optimisation
had inertance-to-mass ratio values at the upper limit of 10% it was decided to extend the upper
bound to µ = 25% to investigate whether higher inertance values could yield improved results.
As before, 10 different initial conditions were used for the gradient-based optimisation process
described in Sections 4.3.1 and 4.3.2.
The ten optimisation runs yielded an increase in flutter speed between 1.7% and 4.1%, with
the results of the ∆VF = 4.1% case shown in Table 4.4. For this case there was also in increases
of 1.8% in the flutter speed of the non-critical branch. Examining the data in Table 4.4 shows
that all three TID devices have tuning frequencies around the 2Hz region. This indicates that
the devices have tuned to the low frequency structural modes which participate in the flutter
response, however this would need to be verified via a separate analysis. As with the single
device case, the TID has opted for large inertance-to-mass ratios, possibly to maximise the device
bandwidth as discussed in the previous section. The damping coefficients are also small compared
to the pure damper case, indicating that the tuning effects are providing most of the performance
improvement. Most important is that the performance of multiple TID devices is almost double
that of a single TID device but still only approaches the benefits of a single damper device.
However, given that the damping coefficient required to deliver more than 4% improvement in
flutter speed using a damper alone also locks the joint, a distributed tunable-device approach
could be a viable alternative to a single viscous damper. Although a trade-off would need to be
considered in terms of device and layout complexity, as from an cost/integration/maintenance
point of view it may be preferable to have a single device in the structure.
TABLE 4.4. Device parameter values when three TID devices are present at locations
A, B and C for ∆VF = 4.1%.
Location Device c T ID b T ID k T ID f T ID µ
3 5
Loc. A TID 6.23 × 10 745.0 1.95 × 10 2.57Hz 0.20
Loc. B TID 7.37 × 103 577.0 9.82 × 104 2.08Hz 0.15
Loc. C TID 6.24 × 103 647.0 8.08 × 104 1.78Hz 0.17
112
4.4. CHAPTER SUMMARY
This chapter has presented a methodology for optimising the flutter response of a truss-braced
wing aircraft using passive vibration absorbers embedded in the truss-structure. The optimisation
scheme used a MATLAB gradient-based optimiser which invokes Nastran Solution 200 to compute
the flutter response and sensitivities with respect to the device properties. It is noted that the
implementation of the optimisation is generic and could be used to optimise any mechanical
network attached to any primary system - with the caveat that only linear device parameters
are used and the analysis falls within the scope of Nastran Solution 200. This method has not
been exploited by the mechanical network community and so the work presented in this chapter
fundamentally enhances the potential for mechanical network design methods to be applied to
the design and optimisation of large and complex systems.
The results have demonstrated that a small amount of flutter suppression is available from
including vibration absorbers in the BUG-T model, with improvements in flutter speed between
1.7% and 5.36% depending on the chosen device layout and location, rising to 6.4% when the
non-critical flutter branch was considered. These results are similar to the levels of flutter
suppression provided by active control methods 167,170 which is strong evidence that passive
vibration absorbers should be considered as a viable method for providing flutter suppression
in truss-braced wings. For the single device case the damper or TID-D layout at location A
offered the best performance and the TID device at location A also exhibited promising tuning
characteristics, however the tuning behaviour could not be achieved by the current optimisation
process. Devices at location B and C provided a small amount of flutter suppression and there was
little correlation between the parameter values and the expected tuning behaviour. It is suggested
that this could lead to a robust design for the device however the mechanism by which flutter
suppression is achieved at location B and C requires further investigation, especially for tunable
devices. For the case where multiple TID devices were included flutter speed improvements up to
4.1% were observed, however the inertance values required by the TID devices were between 15 -
20% of the SUGAR half wing mass which may not be practical for this application. However, the
damping coefficients for the TID elements were all O (3), which is less than the values required
by the pure damper layout, indicating that a tunable device can provide better performance
than the pure damper for the same viscous damping coefficient. It is recommended that a more
comprehensive flutter study should be run where multiple combinations of Mach number, aircraft
velocity and altitude are considered to better understand the robustness of the device parameters
to changes in their operating conditions.
Regarding the tuning of the TID and TID-D layouts, the poor tuning of these devices is
likely related to the fact that the flutter velocity is not considered directly by the optimisation
process and is only calculated during post-processing. Therefore, if the optimiser was instructed to
maximise the critical flutter velocity it is likely that an improved performance could be obtained
for all of the device layouts considered in this chapter. However, this is difficult to achieve in
113
CHAPTER 4. PASSIVE FLUTTER SUPPRESSION USING VIBRATION ABSORBERS
a standard Nastran flutter analysis as there is no built-in mode tracking capability, so the
calculation of the flutter velocity is not robust enough to handle changes in the design. One option
could be to incorporate an automatic mode-tracking algorithm, as in Hang et al. 272 , however this
would be a significant undertaking and is considered beyond the current scope.
This study considered three candidate device layouts in the flutter optimisation but in a
typical structure-based mechanical network study it is usual for ten or more candidate layouts
to be investigated. It is possible that a different layout could have achieved a better flutter
performance, however the focus of this was study was to demonstrate flutter suppression by
including a vibration absorber in the structure of a TBW. In this sense the study was a success
as it was shown that improvements in flutter speed between 1 - 6% are achievable, although
the effectiveness of vibration absorber is heavily dependent on the flutter mechanism as this
dictates the level of relative motion at the device locations. In the case of the BUG-T model
the flutter mechanism did not show much relative motion at the device locations, which is the
main reasons for the small improvements in flutter speed that have been obtained. To achieve
better performance from the use of vibration absorbers it is suggested that the design of the
device properties should be considered as part of a wider structural optimisation. This would
allow the structural properties of the airframe to be tailored to promote improved performance of
the device whilst simultaneously satisfying the necessary design constraints. Given the generic
implementation via Nastran such an approach is readily available provided that the device
properties are assumed to be linear and frequency invariant. If this is not the case then different
analysis methods and optimisation schemes must be considered.
Finally, both studies identified parameter values in the range k = 105 − 106 , c = 103 − 105 ,
b = 102 − 103 as having a beneficial impact on the flutter speed. These values will be used in
Chapter 6 to investigate the design of a physical device that can provide the required magnitude of
force coefficients whilst maintaining a geometry and mass that is conducive to being incorporated
into a truss-braced wing.
114
HAPTER
5
C
PASSIVE G UST L OADS A LLEVIATION USING V IBRATION
A BSORBERS
his chapter investigates the potential for passive vibration absorbers to provide gust
T loads alleviation (GLA) in truss-braced wings. The candidate device layouts and locations
introduced in the previous chapters are used to minimise the response during a discrete ‘1-
cosine’ gust encounter. The device parameters are optimised using a MATLAB-based optimisation
framework which modifies the frequency response of the model in order to tailor the device
properties to target specific structural modes. It is demonstrated that either a single damper
or a combination of inerter-based devices can be used to achieve a reduction in gust loads of
approximately 4% for spanwise locations inboard of the strut attachment point and that this
reduction is consistent across the full range of gust gradients. Furthermore, it is noted that the
inerter-based device has a significantly smaller damping coefficient than the case where just a
damper is used and that the device parameter values are viable within the scope of an aerospace
application. This chapter is organised as follows: In Section 5.1, the BUG-T half-wing model and
the equations of motion for discrete gust analysis in Nastran are introduced, then the baseline
response of the half-wing model is calculated for a family of 1-cosine gusts. In Section 5.2 a
software framework for optimising vibration absorbers for gust loads alleviation is introduced
and in Section 5.3 and 5.4 the three candidate vibration suppression devices are included in the
model and their affect on the gust loads envelope is observed for a range of parameter values and
device configurations. The work in this chapter is based on the following published work:
Christopher P. Szczyglowski, Simon A. Neild, Branislav Titurus, Jason Z. Jiang, and Etienne
Coetzee, "Passive Gust Loads Alleviation in a Truss-Braced Wing Using an Inerter-Based Device",
Journal of Aircraft (2019), accessed September 01 2019. DOI: 10.2514/1.C035452
115
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
In this section the baseline gust response of the BUG-T half-wing model is defined to provide
a suitable reference for the gust load optimisation studies in Sections 5.3 and 5.4. First a brief
overview of the BUG-T half-wing model is provided and the equations of motion used by Nastran
for the solution of discrete gust encounters are detailed. Next, the model response to a family of
1-cosine gusts is evaluated, with specific focus given to the structural modes which participate in
the gust response and gust gradients which form the gust loads envelope.
The work presented in this chapter uses a half-wing version of the BUG-T model for all dynamic
and aeroelastic analysis. This model was a precursor to the full-span BUG-T model presented
in previous chapters and as such the two models have many common features. For example,
the wing planform and mass and stiffness distribution are the same as for the full-span model,
however the strut is assumed to have a constant chord length of 0.8m instead of the distinctive
‘bow-tie’ shape of the SUGAR 765-095 Rev. D. The beam axis is positioned at the midchord for
all flexible members and rigid bars are connected to the beam nodes to represent the planform
shape and provide the basis for the spline between the structural and aerodynamic DOFs. The
connection between the truss members and the wing are modelled as pinned-joints which transfer
both in-plane and torque moments across the joint. The wing and strut are joined to a reference
node which is fully-fixed for all subsequent analysis, meaning that the rigid body modes will
not participate in the gust response. This will have an effect on the participation of the flexible
modes, however, as this is a preliminary investigation this simple approach is deemed satisfactory.
Finally, an aerodynamic symmetry condition is enforced in the X G ZG plane to ensure the correct
downwash distribution for the DLM aerodynamic theory. The structural and aerodynamic model
is shown in Fig. 5.1 and additional details can be found in Appendix C.
116
5.1. GUST RESPONSE OF THE BUG-T HALF-WING MODEL
The gust loads calculated in Chapter 3 were based on a quasi-static load factor derived by
Pratt 263 using the assumption of a rigid aircraft response. This is a gross oversimplification
of the aircraft dynamics during a turbulence encounter, and is particulary inaccurate for a
TBW model given its well-documented nonlinear aeroelastic characteristics reslulting from the
increased flexibility of the structure 100,106,264 . In an attempt to improve upon the quasi-static
Pratt gust, the certification guidelines specify atmospheric turbulence as a time-varying function
with two common formulations - discrete gusts and continuous turbulence 234 . In this chapter only
discrete gust encounters are considered, however the concepts and modelling methods discussed
here could be readily applied to reduce loads from continuous turbulence.
In the standard Nastran process the gust loads represent an incremental load acting on the
structure and, by virtue of assuming a linear structural response, these additional loads can
be added to the 1g steady-state loads to generate the gust loads envelope. In this chapter the
beam loads will be prefixed by ∆ to denote the fact that they are incremental loads resulting from
a gust analysis, as opposed to the loads which comprise the full loads envelope. The simplest
approach for modelling a discrete gust is to idealise it as a one-dimensional 1-cosine waveform
with equation 49,234
Uds πU∞ t
· µ ¶¸
w g (t) = 1 − cos , (5.1)
2 H
where w g is the gust vertical velocity, H is the gust gradient (distance to reach the peak gust
velocity), U∞ is the aircraft forward velocity and Uds is the gust design velocity, defined as
¶1
H
µ
6
Uds = Ure f F g , (5.2)
106.17
where F g is the flight load alleviation factor and Ure f is the reference gust velocity, varied linearly
from 13.4m/s EAS at 15,000ft to 7.9m/s EAS at 50,000ft as specified in CS-25 234 . The value of F g
is altitude and mass dependent and is calculated using the method in CS-25 for the MTOW and
MZFW in Appendix A. Note that for the purposes of the analysis all velocities must be converted
to units of true air speed from their definition in CS-25 which is in equivalent air speed (EAS).
The gust family used for the simulations in this chapter is shown in Fig. 5.2.
In Nastran the response of an aircraft to discrete gusts is termed dynamic aeroelasticity and
the solution sequence associated with this analysis is referred to as SOL 146 82 . In this analysis
the gust response is represented by the following equation of motion
1
· ¸
− M hh ω2 + jC hh ω + (1 + j g)K hh − ρU∞
2
Q hh (m, k) ξ = P hh (ω), (5.3)
2
where M hh , C hh and K hh are the modal mass, damping and stiffness matrices, g is the structural
damping term, Q hh is the AIC matrix in the modal domain which itself is a function of Mach
number m and reduced frequency k, ξ are the modal coordinates and P hh is the modal load
vector of aerodynamic loads. The modal loads vector is a function of dynamic pressure q, gust
117
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
velocity w g , the AIC matrix Q hi which relates the downwash velocity on every panel to the modal
coordinates and the gust downwash matrix w i . This matrix is a function of both the excitation
frequency and the geometry of the aerodynamic panels
where subscript p and i represent the pth excitation frequency and the ith aerodynamic panel
respectively, ω p is the excitation frequency, γ i is the dihedral angle of the panel, x i is the x-
coordinate of the panel quarter-chord and x0 is an offset distance used to delay the application of
the gust load. As the unsteady aerodynamic forces used by Nastran are frequency dependent the
solution of Eqn. 5.3 is formulated as a frequency response problem, with any time varying gust
loads transformed into the frequency domain using the Fourier transform. Once the frequency
response analysis is complete Nastran transforms any output quantities into the time-domain by
performing an inverse Fourier transform.
Before a device optimisation study can take place it is necessary to determine the baseline gust
response of the aeroelastic model so that the effect of the vibration suppression device(s) can
be quantified. The simulation is carried out at a flight point of 36,000ft at Mach = 0.75 30 (U∞ =
221.57m/s) which yields a gust reference velocity and flight load alleviation factor of 10.12m/s EAS
and 0.98 respectively. Solution frequencies up to 30Hz are included and a frequency increment of
0.005Hz is used.
118
5.1. GUST RESPONSE OF THE BUG-T HALF-WING MODEL
To generate the loads envelope ten equally-spaced gust gradients are considered in the range
9m to 107m and for each gust response the maximum beam loads are calculated across all
time steps for every beam node along the wing, strut and jury-strut. Examples of this quantity
are shown in Fig. 5.3 for the three wing bending moments and axial force with colour-coded
markers indicating which gust gradients are driving the loads at the different spanwise locations.
Inspecting Fig. 5.3 it is clear that a number of gusts are responsible for generating the incremental
(a) Wing in-plane/chordwise bending moment (b) Wing out-of-plane/spanwise bending moment
F IGURE 5.3. Incremental gust loads envelope for the wing in-plane bending moment
(a), out-of-plane bending moment (b), torque (c) and axial force (d).
119
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
gust loads envelope, implying that any device that seeks to minimise gust loads must be effective
across the entire gust spectrum. Also, in general there are two spanwise locations where the wing
beam loads reaches a maxima, these are the strut-attachment point and the wing-fuselage joint.
This suggests that a device that can reduce the loads at one or both of these points will provide a
net benefit to the wing gust response and may also minimise the wing weight via reduction of the
critical loads envelope. Therefore during the device optimisation the loads at these two locations
will be used to indicate the amount of loads alleviation that the devices are providing.
When the loads envelope is evaluated by extracting the maximum loads across all timesteps
there is no consideration given to the fact that different loads experience their maximum value at
different points in time, i.e. the loads are uncorrelated. It is important to consider the correlated
time histories of the loads as the resulting principal stresses can depend upon 2D or even
3D loads 49,278 . Figure 5.4 shows an example of a correlated loads plot for the wing spanwise
bending moment and torque loads at the wing-fuselage location and the strut-wing joint. Here
the complete time histories for the ten gust cases have been plotted and the extreme points
have been identified using the MATLAB convexHull algorithm 279 . These results show that the
torque and spanwise bending loads at these two positions are predominantly driven by gust
gradients H = 20m, 31m and 64m, again indicating that the optimised vibration suppression
device must be effective across multiple gust gradients. It is possible to extend this method to
form a six-dimensional convex hull which represents the time-correlated loads of all six beam
sectional loads. This has not been demonstrated here but is mentioned to highlight the scale (and
F IGURE 5.4. Time correlated incremental gust loads for the spanwise bending moment
and the torque at the wing-fuselage joint (a) and wing-strut attachment point (b).
120
5.1. GUST RESPONSE OF THE BUG-T HALF-WING MODEL
complexity) of the challenge facing any technology that seeks to minimise gust loads. Finally, it is
important to note that these results represent the loads from ten discrete gusts at a single flight
point, whereas a typical loads cycle for a commercial aircraft will consider hundreds of thousands,
or even millions of load cases in order to determine the critical design loads 49 . Therefore, the
fact that short-to-medium gust gradients are prevalent in these results should not be taken as
indicative of the general gust loads requirements of a truss-braced wing aircraft.
One of the key requirements of the device optimisation process is to understand which modes
the device should influence in order to minimise the chosen cost function. Figure 5.5 shows the
maximum modal coordinate for the first seven structural modes as a function of gust gradient and
plots of the corresponding modeshapes can be found inset in Fig. 5.9. It demonstrates that the
dominant mode in the gust response is the fundamental one, which corresponds to an outboard
wing ‘flapping’ mode. The other modes follow, almost in mode order, with some fluctuation at the
lower and higher ends of the gust spectrum, although there is a clear gap between modes two
and three and the rest of the modes. Therefore any vibration suppression devices included in
the model must target these low frequency modes in order to minimise the gust response. The
dominance of the low frequency modes is due to the frequency and energy density of the gust
input signal. Here, the gust frequency content is evaluated by applying the Fourier transform to
the time-domain 1-cosine gust signal, i.e. ŵ g (ω) = F w g (t) . This operation is performed using
¡ ¢
the MATLAB fft algorithm and the corresponding signal bandwidth and energy density are
calculated for the gust profiles used in the discrete gust analysis. The signal bandwidth is defined
p
as the frequency where the amplitude equals 1/ 2 of the corner frequency value and the signal
¯2
energy density is defined as ¯ŵ g (ω)¯ . Figure 5.6 shows the bandwidth and energy density for the
¯
F IGURE 5.5. Maximum modal coordinate for the first seven modes as a function of
gust gradient.
121
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
ten discrete gusts considered in this analysis and annotations detailing the natural frequencies of
the first eight modes are included to provide context for the magnitude of the bandwidth values.
Examining these results it is clear that the gust bandwidth drops sharply as the gust gradient
increases whilst the gust energy density increases in a logarithmic fashion. The effect of this is
that the low frequency modes receive a much higher input power than the high frequency modes,
meaning that these modes have a greater participation in the gust response. Also, Fig. 5.6 shows
that the maximum gust bandwidth within the CS-25 mandated limits is approximately 8.7Hz,
however, that is not to say that higher frequencies should not be accounted for when simulating
the gust response. The gust bandwidth simply gives an indication of which parts of the frequency
spectrum are most important during a gust encounter, thereby focusing the efforts of a vibration
suppression device.
In order to understand the magnitude of the of the relative motion across the device terminals
the displacement at each of the device terminals has been extracted for all time steps in the
solution1 . Figure 5.7(a) shows an example of the relative displacement at location A for three
different gust gradients as a function of time and the maximum joint rotation/device stroke is
annotated with a dashed line for each gust gradient. For this location it is clear that the maximum
rotation/stroke is achieved at different times for the three gust gradients considered and also that
the number of device cycles varies. This implies that the performance of a vibration suppression
device will be inconsistent across different gust gradients and it may be necessary to pick a
specific gust as the design case in order to tailor the properties of the device. Fig. 5.7(b) shows the
1 Such an approach was not possible in the previous chapter as the magnitude of the calculated flutter eigenvectors
F IGURE 5.6. Gust bandwidth and energy density as a function of gust gradient.
122
5.1. GUST RESPONSE OF THE BUG-T HALF-WING MODEL
F IGURE 5.7. Relative displacement at the device locations as a function of gust gradi-
ent. Provided as an example time history (a) and the maximum value across all
gust gradients (b).
maximum relative displacement at all three locations for all gust gradients considered2 . These
results show that the magnitude of the relative motion is not directly proportional to the gust
gradient, rather it depends on a combination of the gust bandwidth and energy density. Each
location has a similar trend where the relative motion reaches a maximum and then decreases
for the longer gust gradients. As with the flutter analysis in the previous chapter, this trend
is a consequence of the modes which participate in the solution and how much relative motion
each mode has at the various device locations. Therefore the results in Fig. 5.7(b) suggest that a
device at location A will be more effective for gust gradients up to 42m and that a device at either
location B or C will have more influence during the longer gusts. As noted previously, this opens
the possibility of using multiple devices to target the response of different gusts and such an
approach is demonstrated in Section 5.4. Finally, the rotational joints at locations A and C show
a maximum relative rotation of approximately 1.8°. This information will be used in next chapter
to guide the design of a physical vibration suppression device that can provide the required linear
force coefficients identified by the flutter suppression and gust load alleviation studies.
In summary, the incremental gust loads envelopes and correlated loads plots showed that
the maximum structural loads are attributed to a variety of gust gradients, meaning that the
vibration suppression devices must be designed to be effective across a range of gust gradients.
Also, the incremental spanwise bending moment envelope revealed two locations along the wing
2 These quantities are the same as the dashed lines in Fig. 5.7(a) but now calculated for every device location and
gust gradient.
123
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
where the loads reach a maximum; these are the wing-strut joint and the wing-fuselage joint.
The loads at these locations will be used to formulate a cost function for the device optimisation
and will provide a reference from which the GLA of the different device configurations can be
evaluated. Figure 5.5 has shown that the lower frequency modes are most active throughout the
gust spectrum and constitute the main components of the structural response. Also, from Fig. 5.6
it is clear that for the chosen flight point all one-minus cosine gusts have a bandwidth lower than
8.7Hz, therefore it is appropriate for the vibration suppression device to target the low frequency
modes in order to influence the gust response and reduce loads.
A typical strategy to reduce gust loads is to enable loads control via aerodynamic control sur-
faces 37 . This approach is commonplace across the aerospace industry and has been applied to the
SUGAR TBW aircraft during aeroelastic wind tunnel tests where it was observed to suppress
flutter and alleviate gust loads, although the controller was not expressly designed for gust load
alleviation 148 . Recently, Bartels et al. 280 performed a study of GLA and flutter suppression in a
TBW aircraft using the the Variable Camber Continuous Trailing Edge Flap (VCCTEF) concept
coupled with a static output feedback controller. It was found that the controller was effective in
suppressing flutter and stabilizing the aircraft during a gust encounter, however only a single
gust gradient was considered during the controller design and the overall level of loads reduction
is not quantified. Whilst active control schemes have been shown to suppress flutter and provide
GLA, the certification of these systems and their integration into the design process does present
a number of challenges 60 , hence alternative (non-active) methods of loads alleviation have been
proposed such as folding wing tips 26 , aeroelastic tailoring 33 and morphing structures 41 .
In this section an alternative approach to loads alleviation in a truss-braced wing is presented,
which is to use a two-terminal vibration absorber embedded within the truss structure to reduce
the gust response. Here, the same candidate device layouts and locations that were used to
provide flutter suppression in Chapter 4 will be trialled for gust loads alleviation. Note that all
discussions on modelling methods and device layouts from Chapter 4 apply equally to the work
in this chapter and no further explanation of the device modelling will be given here.
Figure 5.8 shows a flowchart of a software framework for optimising a vibration suppression
device which alleviates gust loads. As before the optimisation of the device is conducted using
Nastran SOL 200 270 , however, there are three main factors to consider when using SOL 200 to
optimise a vibration absorber to alleviate gust loads:
• Continuous and discrete gust analysis are types of dynamic aeroelasticity, therefore, the
optimisation of the aircraft response to turbulence encounters constitutes a dynamic
response optimisation. Problems of this type will typically have global design spaces 281
and as SOL 200 uses a gradient-based optimiser it is not suitable for solving this class of
124
5.2. OPTIMISING A VIBRATION ABSORBER FOR GUST LOADS ALLEVIATION
problem, therefore additional measures will be required to ensure that the design space is
successfully traversed and a global optimum is found.
• The dynamic aeroelastic response solution, SOL 146, cannot be invoked within SOL 200,
hence it is not possible to directly optimise the device parameters to minimise the gust
response. Instead, an alternative cost function which is independent of the gust response
must be used. Although it is possible to optimise the gust response of an aircraft using
Fluid-Structure-Interaction methods 92 such an approach is not always favourable given
the increased computational cost of dynamic aeroelastic analysis.
• SOL 200 can only cater for linear finite element models and linear analysis types, therefore,
nonlinearities in the structure, aerodynamics or device properties will not be considered.
In order to mitigate for these points, the optimisation framework uses a Multi-Start Optimisation
approach 282 to sample the multi-modal design space before running a local gradient-based
optimisation at each sample point. The cost function is formulated in terms of a frequency
response problem with the device parameters optimised to reduce the influence of certain modes
of the structure that are identified as having a significant participation during the gust response.
This approach is particularly useful for two reasons: Firstly, the optimisation of the device is
decoupled from the gust response which removes the need to run a computationally expensive gust
analysis for each function evaluation. Secondly, the lower computational cost means the approach
can easily be applied to high-fidelity finite element models given the generic implementation via
Start
Definecdesigncvariableclimits Evaluateccostcfunctions
C SOLcY46
B
Choosecdeviceclocation1s4 GradientxBasedcOptimiser
A
SOLcN?? Evaluatecdevicecperformance
Choosecdeviceclayout1s4
Tunecdevice
Yes End
IntermediatecDOF?
No
Samplecdesigncvariablesc
Choosectargetcmode
c
ChoosecnoIcofcsamplesc1ns4
F IGURE 5.8. Flow-chart of a process for optimising a generic vibration absorber for
gust load alleviation.
125
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
Nastran. This is an important consideration as the device properties will be tightly coupled to
the mass and stiffness of the model which will vary significantly between a 3D and 1D model,
even when a model reduction technique is employed. Further details on each module of the
optimisation framework are provided in the following sections.
There are several stages involved in the set-up of the optimisation problem:
1. Cost function. The purpose of the device is to alleviate gust loads by reducing the par-
ticipation of modes that have a significant contribution to the gust response. This can be
achieved by adding damping to that mode or by tuning the device frequency to match the
modal frequency, therefore, the cost function must be able to observe changes in a desired
response quantity as a function of frequency, X (ω) = H(ω)F(ω) where,
¤−1
H(ω) = K − ω2 M + j ωC .
£
(5.5)
and X is some output response from the system, F is a vector of applied forces, H is the
frequency response function (FRF) matrix, ω is the forcing frequency and the M, C, K
matrices include the terms from the device. As the device will target gust loads it is sensible
to evaluate the transfer function relating the beam loads and responses. More specifically,
the spanwise bending moment will be evaluated as it typically sizes the wing covers which
constitute a significant portion of the wing weight. An example of this quantity is shown
in Fig. 5.9 for the bending moment at the Wing-Fuselage (WF) and Wing-Strut (WS) joint
for the case where no devices are included in the model; the first eight flexible modes are
shown to highlight which mode contributes to the bending moment at these points in the
wing. Note, there is no ‘engine-mode’ as the engine is included in the model as a lumped
mass and there is no structural representation of the pylon. From Fig. 5.9 it is clear that
the fundamental mode dominates the bending moment at the WS joint as the first mode
corresponds to a ‘flapping’ motion of the outboard wing. However, the bending moment at
the WF location includes contributions from the first five modes, so any device configuration
that wishes to minimise loads at this point must be effective across multiple modes. Finally,
the cost function is defined as the area under the transfer function, which for a discrete
response and a suitably small frequency increment, is approximated by,
f2
X
Ja (v) = δf × | H BM (v, f )| , (5.6)
f1
where, v are the design variables, δf is the frequency increment, f is the frequency, f 1
and f 2 are the upper and lower limits of the frequency band and HBM is the FRF matrix
corresponding to the spanwise bending moment. This cost function has been chosen as it
allows multiple modes to be targeted by careful selection of the upper and lower frequency
limits which will be critical if there are multiple modes that the device must influence.
126
5.2. OPTIMISING A VIBRATION ABSORBER FOR GUST LOADS ALLEVIATION
F IGURE 5.9. Spanwise bending moment FRF due to a vertical harmonic force at the
wing tip with 3% modal damping. Plots of the translational terms of the first
eight normal modes are shown in inset panels and the the shading of the plane
represents the magnitude of the beam twist.
2. Design variables. As with the flutter response optimisation in the previous chapter, the
design variables in this study are the spring stiffness, viscous damping coefficient and
inerterance-to-mass-ratio for the various spring, damper and inerter elements in the chosen
device layout. The mass-ratio is defined as a fraction of the primary structure mass, which
for this study is the wing, including truss structure, with a total mass (M wing ) of 10467kg
accounting for all primary and secondary structural masses. An inertance-to-mass-ratio of
10% has been chosen, however, as an inerter element is used the actual mass of the device
will be far less than 10% of the wing mass. An upper limit of 1 × 105 is chosen for the viscous
damping coefficient in order to prevent the rotational joints from locking up and an upper
limit of 1 × 108 is defined for the spring stiffness as this covers device tuning frequencies up
to 50Hz when considering the maximum permitted inertance value 3 . A summary of the
design variable upper and lower bounds is provided in Table 5.1.
3. Device location(s). As the truss topology is fixed the device position is constrained to the
three candidate locations identified in the previous chapters, these are the hinge joints at
the root and tip of the strut as well as a device in parallel with the jury-strut. The study in
3 This spring stiffness was also identified in Chapter 4 as the upper limit in order to preserve the pinned-joint
behaviour.
127
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
TABLE 5.1. Design variable upper and lower bounds for the gust load alleviation
optimisation study.
Design Variable Lower Bound Upper Bound Translational Units Rotational Units
Spring Stiffness 1 1 × 108 N/m Nm/rad
Damping Coefficient 1 1 × 105 N/ms−1 Nm/rads−1
Mass Ratio 1 × 10−4 0.1 − −
Sections 5.3 and 5.4 will consider both a single device or a combination of devices at any of
these three locations.
4. Device layout(s). To maintain consistency the three candidate layouts introduced in the
previous chapter will be used here for the purposes of gust loads alleviation. These layouts
are: a viscous damper, the Tuned-Inerter-Damper and the Tuned-Inerter-Damper-Damper.
The multi-start method involves sampling the design space at an appropriate number of intervals
so as to ensure the many local-minima are captured. Next, a gradient-based optimiser, in this
case the Modified Method of Feasible Directions algorithm belonging to Nastran’s MSCADS suite
of optimisers within SOL 200, is used to search the local design space for an optimal solution
at every sample point. Once all samples have been evaluated the minimum cost function and
its associated design variables are selected from the individual optimisation results. Whilst
this approach allows global design spaces to be traversed it is important to note that it does
not guarantee a global optimum. Also, the robustness of the solution is heavily dependent on
the initial sampling of the design space 283 . In this study the Latin Hypercube Sampling (LHS)
algorithm lhsdesign is used to randomly sample each design variable. A parameter study was
carried out to determine the number of samples (n S ) and a value of 100 was chosen as this was
found to give consistent results. As some of the design variables have allowable values that span
several orders of magnitude the sample points from lhsdesign are mapped to an exponential
distribution to ensure the design space is sufficiently covered. Next, the optimisation process
is called n s times with each run using a new sample point as the initial conditions, however,
before SOL 200 is executed a check is made on the device layout. If a TID or TID-D layout has
been selected and a tuning mode specified then the initial device stiffness is adjusted using the
approximate tuning rule
k = µ × M wing × ω2i , (5.7)
where ω i is the natural frequency of the primary system mode that the device will target. This
step is not necessary for the optimisation to proceed but it does yield slightly better performance
as the device will start the optimisation in the correct region of the frequency domain. Once the
optimisation loop has finished the best design associated with the minimum cost function value
128
5.3. PASSIVE GUST LOADS ALLEVIATION USING A SINGLE DEVICE
across all sample points is extracted and the necessary bulk data entries are passed into the gust
analysis module.
The discrete gust analysis described in Section 5.1.2 is used to calculate the gust response for
the best device configuration identified by the multi-start optimisation procedure. Note that only
one gust analysis is performed compared to n S frequency response optimisations. This yields a
significant benefit in terms of computational cost as an aeroelastic analysis is more expensive
than a (structural) frequency response analysis as the calculation of the AIC matrices incurs a
significant overhead.
The purpose of this section is to determine which device layout is most appropriate for reducing
gust loads when only a single device is considered. This is achieved by running two optimisation
studies: the first considers a single damper placed at one of the candidate locations and the
second extends this study to all three device layouts. This allows the performance of the TID and
TID-D layouts to be benchmarked against the GLA benefits of a pure damper.
Employing a single damper, the viscous damping coefficient of the device is optimised in order
to reduce the area under the wing-root bending moment transfer function, Ja (v), at the wing-
fuselage joint location. The wing-fuselage location is chosen as it captures the response of multiple
modes, as opposed to the wing-strut FRF which is mostly the fundamental mode. Figure 5.10a
shows the baseline and optimised transfer function for the three different damper locations as
well as the change in FRF amplitude at the original natural frequencies. Figures. 5.10b and 5.10c
show the percentage change in the actual gust-induced maximum spanwise bending moment
as a function of gust gradient for the wing-fuselage joint and wing-strut attachment points
respectively. Note that as the difference is presented with respect to the baseline aeroelastic
model a negative value indicates that the device has a beneficial effect, whereas a positive value
is detrimental. Table 5.2 shows a summary of the device performance and damping coefficients
for each of the three locations.
The results show that placing a damper at Location A achieves the lowest cost function value.
In terms of the frequency domain performance, the damper has successfully reduced the FRF
response amplitude around modes two and three as well as a reduction around mode one. This
has translated into a consistent reduction in spanwise bending moment at the wing-fuselage joint
across all gust gradients considered with a maximum change of 3.84% for the 42m gust, however,
this is at the expense of a very large viscous damping coefficient. Interestingly, the small amount
129
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
B C
F IGURE 5.10. Gust loads alleviation results for the case with a damper at locations A,
B or C.
TABLE 5.2. Cost function values and device parameters for the case with a damper at
locations A, B or C.
of additional damping that has been provided by the damper around mode one has a negligible
effect on the loads at the wing-strut joint, despite mode one having the largest contribution to
the loads at this wing position.
The damper at location B has had a negligible effect on the amplitude of the first three modes
and has increased the spanwise bending moment at both monitoring points for all gust gradients
considered. It is hypothesized that this is because the force generated by the damper acts in the
same plane as the lift force acting on the wing, which leads to an increase in sectional shear force
in the region around the jury-strut and a subsequent increase in spanwise bending moment at
nodes away from the jury-strut attachment point. For this reason, a device in parallel with a
vertically-orientated jury-strut may be inappropriate for reducing the wing bending moment.
Conversely, the damper at location C has provided additional damping to the fundamental mode
but has had very little effect elsewhere and has only had a minor effect on the bending moment
values at the monitoring points. For the wing-fuselage joint there is a net reduction in bending
130
5.3. PASSIVE GUST LOADS ALLEVIATION USING A SINGLE DEVICE
F IGURE 5.11. Relative velocity across device terminals for each candidate location in
response to a unit harmonic load applied in the positive z-direction at the wing-tip.
moment, although it is modest - the maximum reduction is 0.65%. There is a net increase in
bending moment at the wing-strut joint but it does not exceed 0.2%, which may be acceptable
once static aeroelastic load cases are considered. This suggests that modes two and three are
driving the value of the wing-fuselage bending moment and that targeting these modes will lead
to a much larger reduction in loads.
As with the full-span BUG-T model, the tendency for a device at a particular location to
influence the response of certain modes is suggested by considering the relative velocity across
the device terminals as a function of frequency. Examining Fig. 5.11, it is clear that location A is
best suited to targeting the response of the first three modes due to the increased relative angular
velocity at low frequencies, whereas location B is more appropriate for the higher frequency modes
as these are dominated by truss bending and twist modes. Interestingly, location C experiences
significant relative velocity for the fundamental mode. This is a direct consequence of the pinned
connection between the primary strut and the wing which allows the outboard section of the wing
to ‘flap’ freely about this joint and generate large bending moments at the joint location.
When considering the design of a vibration suppression device it is necessary to understand
the magnitude of the force that is transmitted into the primary system from the device as this
will influence the design of the vibration suppression device4 . Here, the damper force has been
extracted for the case where a damper is positioned at the strut root joint (location A) with
a viscous damping coefficient of 1 × 105 Nm/rads-1 . Figure 5.12 shows the time history of the
damper force for the ten gust gradients considered with a marker placed at the location where
the damper force reaches a maximum and these values are provided in Table 5.3. Clearly the
4 See Chapter 6 for further details of the preliminary design of a vibration absorber for aeroelastic control.
131
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
F IGURE 5.12. Time history of damper force for a damper at location A with viscous
damping coefficient 1 × 105 Nm/rads-1 .
TABLE 5.3. Maximum magnitude of the damper force as a function of gust gradient for
the case where a damper is positioned at the strut-root joint with viscous damping
coefficient 1 × 105 Nm/rads-1 .
9 20 31 42 53
Gust Gradient m
64 75 86 97 107
10,008 19,636 25,184 25,433 21.206
Max. Damper Force N
16.861 11,958 9493 8579 7743
amount of force generated by the damper (or any device) will be a function of the relative motion
across the terminals, therefore a comprehensive study of the flight envelope must be conducted
to determine the range of loads that the device must provide.
In summary, this study has shown that a damper positioned at location A using the highest
permitted viscous damping coefficient can achieve a maximum reduction of 3.84% in spanwise
bending moment at the wing-fuselage joint. Furthermore, the favourable influence of the damper
is due to the increased modal damping of the first three structural modes, indicating that for a
device to successfully reduce gust loads it must target each of these modes. For the other damper
locations, location B results in a net increase in spanwise bending moment whilst a damper at
location C provides only a minor benefit via increased damping for the fundamental mode. In
all instances there was a negligible change in loads at the wing-strut joint with the change in
bending moment restricted to ±0.5%. All dampers required a large viscous damping coefficient
to have an effect on the gust response but without considering the design of the damper it is
not possible to say whether these values are realistic or not. Finally, given the poor gust load
132
5.3. PASSIVE GUST LOADS ALLEVIATION USING A SINGLE DEVICE
alleviation results as well as the reduced relative velocity across the device terminals for the first
three modes, location B will not be considered in further studies.
In this section the TID and TID-D layouts are compared for the case where a single device is used
in either location A or location C and the results in this section are benchmarked against the
damper performance from Section 5.3.1. Note, as the study in the previous section has shown
that reducing the response of the first three modes is important for alleviating gust loads the
cut-off frequency f 2 is set to 5Hz to prevent the device targeting the higher frequency modes that
are less critical to the gust response.
A preliminary study was conducted to determine the effect of initially tuning each device to
one of the structural modes before running the optimisation. It was found that for a TID device
at location A setting the initial tuning mode, m dev , to the third structural mode yielded the best
possible GLA for both the TID and TID-D devices, whilst at location C the best GLA was obtained
when the device was tuned to the fundamental mode; although in both cases the performance was
inferior to the pure damper case at the respective locations. The case with a TID-D at location A
was found to be insensitive to the initial device tuning frequency as the optimiser converged to
the same optimum solution regardless of initial conditions, furthermore the performance of the
TID-D was superior to the pure damper. A selection of these results are shown in Fig. 5.13 and
their respective performance metrics and device parameters are given in Table 5.4.
Considering the devices at location A, the TID has successfully reduced the FRF amplitude
for modes two and three with a small increase in the response of mode one. This has translated
into a reduction of almost 4% in BMW F for the shorter gust lengths but during longer gusts the
benefits are lessened as these gusts are dominated by the response of the fundamental mode. This
is a clear consequence of the TID targeting mode three which resulted in an increased response
of mode one. The TID-D is able to mitigate against this by providing damping to those modes
which are not targeted by the TID, yielding a more consistent reduction in loads across all gust
gradients which serves to highlight the importance of using a cost function and device layout
that can target multiple structural modes. For location C, the optimiser has tuned the TID to
match the frequency of the fundamental mode, clearly demonstrated by the ‘split-peak’, enabling
the device to significantly reduce the amplitude of mode one, although modes two and three are
unaffected. The TID-D device has had a less pronounced effect on the fundamental mode and
the split peak is no longer present due to the increased damping from the parallel damper. In
terms of GLA, the TID-D has achieved a better reduction in gust loads despite having a lower
cost function value, which is clear evidence that the mapping between the optimised FRF and
the gust response is not direct. Furthermore, the GLA performance of both devices at location
C is inferior to the devices at location A because the response of modes two and three are not
influenced by a device at location C due to the reduced relative velocity at these locations/modes.
133
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
B C
F IGURE 5.13. Gust loads alleviation results for a selection of single device configura-
tions.
TABLE 5.4. Cost function and device parameters for the configurations in Fig. 5.13.
For all device locations and layouts there is a negligible effect on the spanwise bending
moment at the wing-strut joint. This may be because the gust input power to this mode is simply
too great for the device to have a significant impact on the participation of this mode without
generating an extremely large force in order to influence the outboard motion of the wing. Even
so, as the change in loads is ±0.5% this is an acceptable outcome. There is a significant variation
in the optimum device parameter values for the different device layouts. When the TID is used
the damping coefficient is much lower than the pure damper case, which shows the benefit of
implementing a tunable device over the pure damper case, however the TID-D device has the
opposite configuration. That is, the parallel damper element has a very large damping coefficient
whilst the damper within the TID is effectively zero as it is at the lower bound of the design
variable. In this case the addition of the TID yields only a small reduction in gust loads, although,
134
5.4. GUST LOADS ALLEVIATION USING TWO DEVICES
it does improve the performance during the longer, low-frequency, gusts which is achieved by the
TID element targeting the response of mode one. In terms of inertance values both the TID and
TID-D have opted for the maximum possible mass ratio. This is consistent with the fundamental
theory of vibration absorbers which states that a higher mass ratio increases the separation
between the split-peak frequencies and provides a broader bandwidth for the device which then
manifests itself as a reduction in the overall area under the FRF curve.
In summary, the studies presented in this section have shown that it is possible to achieve a
maximum reduction in wing-fuselage bending moment of 3.9% using a single device configuration.
The best performance was achieved by the TID-D layout but in general either a damper, TID or
TID-D device can provide a reasonable reduction in gust loads when it is positioned at location A.
For device location C, a small reduction in loads of the order of 1% is achievable and location B
was found to be unsuitable due to the reduced relative velocity across the jury-strut terminals for
the lower frequency modes.
In the previous section it was shown that it is possible to reduce gust loads by using a device to
target specific structural modes, however, as the tunable device layouts considered in this paper
only have one internal degree-of-freedom it is not possible to target more than one mode with
a single device. To mitigate this, a two-device configuration will be investigated which allows
two devices to simultaneously target different structural modes. Based on the relative velocity
FRF in Fig. 5.11, a pair of devices located at the strut-root joint and at the strut-tip joint should
be sufficiently active during the first three modes. Also, the results in the previous section have
shown that a tunable device is the best option for influencing a specific mode, therefore, a pair
of TID-D devices will be placed at location A and location C, tuned to modes three and one
respectively. As with the previous study f 2 is set to 5Hz to minimise the response of the first
three modes.
Figure 5.14 shows the gust loads alleviation results for the two device configuration and Table
5.5 details the parameter values and the performance metrics. Examining Fig. 5.14b, it is clear
that the two device configuration has achieved a greater reduction in spanwise bending moment
at the wing-fuselage joint, with a maximum change in bending moment of -4.19% and an average
change of -3.37%. While still relatively small, both metrics are better than any of the single
device configurations considered in Section 5.3.2. Also, despite the increased damping at mode
one there is still a negligible change in the loads at the wing-strut joint which is consistent with
the results from Section 5.3. Also shown in light grey on Fig. 5.14 is the undamped FRF for the
two device configuration, i.e. using the parameter values in Table 5.5 but with all damping terms
set to zero. This highlights the two new modes that have been introduced by the tunable devices
and provides a clear indication of the classical split-peak behaviour of the vibration absorbers.
135
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
B C
F IGURE 5.14. Gust loads alleviation results for the two device configuration.
TABLE 5.5. Cost function and device parameters for the two device configuration
The TID-D at location A has tuned to the fundamental mode whilst the TID-D at location C has
targeted mode three, both of which are understandable when considering the increased relative
velocity at the respective device locations. Furthermore, this result clearly demonstrates the
ability of a combined FE/optimisation approach to tune an inerter-based device attached to a
complex host structure without the need to adopt tuning rules, such as those proposed by Krenk
and Høgsberg 206 , allowing for the possibility of more complex device layouts to be investigated
which could yield greater GLA benefits. Concerning the parameter values, the TID-D at location
A has opted for the maximum damping coefficient value for the parallel damper whilst the TID
element of the TID-D has tuned itself to mode three with a minimum damping coefficient. The
TID-D at location C has an almost negligible damping coefficient for the parallel damper which
implies that the vibration absorber effect is more beneficial than the modal damping. The TID
element has a moderate value of damping coefficient and has tuned itself to the first structural
mode. These results were largely expected given the increased relative velocity at the respective
device locations, as shown in Fig. 5.11.
Finally, Fig. 5.15 shows a comparison between the baseline wing incremental gust loads
136
5.4. GUST LOADS ALLEVIATION USING TWO DEVICES
envelope and the case where the structure is augmented with a TID-D at locations A and C. All
six beam loads are presented to show that even though the optimisation process only considered
the spanwise bending moment at the wing-fuselage joint, as the targeted modes are global
quantities, the device configuration has successfully reduced the loads across the span of the wing.
Specifically, for all six beam loads there is a reduction of approximately 4% for spanwise stations
inboard of the strut-attachment point. Clearly, if gust loads are critical for the components in the
wing then a 4% reduction in loads will translate to a reduction in the wing weight, this indicates
the scale of performance benefits that might be realised from passive devices installed on a high
aspect-ratio braced wing.
Designing a device that could fulfill multiple functions would help mitigate for the fact that
the device is a fixed mass that would need to be carried for the duration of the mission. For
example, another potential application of the device is alleviating wing flutter. Analysis carried
out by the NASA/Boeing SUGAR project indicated that wing flutter was a combination of wing
and strut bending, implying that a passive vibration suppression device could be used to alleviate
flutter so long as the flutter modeshapes have translation or rotational components at the device
locations, as discussed in Chapter 4. Furthermore, the loads reduction provided by the device
could be translated into an improvement in fatigue life by using an analytical fatigue model such
as the one proposed by Rajpal and De Breuker 154 . However, as the purpose of this study is to
introduce the concept of using vibration suppression devices to provide gust loads alleviation,
any benefit in terms of fatigue life extension is considered beyond the current scope.
137
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
(a) Wing in-plane/chordwise bending moment (b) Wing out-of-plane/spanwise bending moment (c) Wing torque
138
(d) Wing in-plane shear force (e) Wing out-of-plane shear force (f) Wing axial force
F IGURE 5.15. Incremental gust loads envelope for the wing beam loads, comparing the baselin response, the pure damper
and the two device configuration.
5.5. CHAPTER SUMMARY
This chapter has presented a novel method for gust loads alleviation in a truss-braced wing based
on using vibration suppression devices to target the specific modes of the structure in order to
reduce their participation in the gust response. Three candidate locations were considered, the
two hinge locations at the root and tip of the primary strut and the jury-strut, and three device
layouts have been tested, a pure damper, the Tuned-Inerter-Damper and the Tuned-Inerter-
Damper-Damper. The gradient-based optimiser within Nastran Solution 200 was used to optimise
the device parameters and a multi-start optimisation approach was used to traverse the global
design space and find a suitably global solution.
The results indicated that the frequency-response optimisation approach introduced in this
chapter allows a vibration suppression device that is capable of providing moderate loads relief
across the gust spectrum to be designed. Furthermore, for the half-wing model considered it
was found that the strut-root joint was the best location for a device regardless of the device
layout, whilst the strut-tip joint provided only minor loads relief and a device located parallel to
the jury-strut was unsuitable for gust loads alleviation. Considering the device layout, for the
case where a single device was considered a damper at the strut-root joint achieved a maximum
reduction of 3.8% in spanwise bending moment at the wing-fuselage joint, however this was
at the expense of a very large viscous damping coefficient. In contrast, the use of an inerter-
based device enabled similar reductions in incremental gust loads with only a fraction of the
viscous damping coefficient. When two inerter-based devices were considered there was a slight
increase in gust load alleviation with a maximum reduction of 4.2% in spanwise bending at the
wing-fuselage joint and similar reductions inboard of the strut-wing joint for the other five beam
loads. Furthermore, the results have shown that for both the single and multiple device cases
the optimised inerter-based devices have large mass ratios, although this could be mitigated by
designing the device to have a large inertance-to-mass ratio to limit the device mass.
The use of a frequency-response optimisation was necessitated by the inability of Nastran
Solution 200 to provide gradient information for a gust analysis. Whilst a frequency response
optimisation favours computational efficiency it has been shown that this approach does not
always lead to an improved gust response, therefore it would be preferable to formulate the
cost function in terms of the gust loads. Such an approach is possible using FSI or adopting
a low-order aeroelastic modelling technique, such as the nonlinear formulations described in
Chapter 2 - this is identified as future work.
Based on the results in this chapter, the level of GLA provided by a passive vibration absorber
scheme is significantly less than that which can be achieved by an active control scheme 171–176 .
However, as the device optimisation was conducted ’by-proxy’ it is possible that greater per-
formance could be obtained if the gust loads were considered directly as the optimisation cost
function. Something that could be readily achieved with any one of the reduced-order aeroelastic
models that are prevalent throughout the research literature 104–107 .
139
CHAPTER 5. PASSIVE GUST LOADS ALLEVIATION USING VIBRATION ABSORBERS
In the next chapter the damper force and device inputs identified in Section 5.1.3 will be
used to inform the design of a device that can achieve the magnitude of linear force coefficients
identified by the flutter suppression and gust load alleviation studies. The findings of this study
will allow the concept of using vibration suppression devices for aeroelastic control to be evaluated
in terms of the device mass, as well as determine whether the required dimensions are feasible
in the context of a truss-braced wing aircraft.
140
HAPTER
6
C
P RELIMINARY D ESIGN OF A V IBRATION S UPPRESSION D EVICE
FOR A EROELASTIC C ONTROL
his chapter is concerned with the preliminary design of a vibration suppression device
T that can provide the required linear force coefficient values identified by the flutter
suppression and gust load alleviation studies from Chapters 4 and 5. The objective of
this chapter is to understand whether the magnitudes of the identified stiffness, damping and
inertance are realisable within the context of a truss-braced wing aircraft, therefore the focus is
on using approximate design formulae and not the detailed analysis of the device characteristics.
Only fluid-based devices are considered given their prevalence in the literature and the fact
that they are already widely applied in the aerospace industry, for example in landing gear 233
and rotor stability applications 192 . Before proceeding it is important to note that this section
relies heavily on reference material from three papers 191,228,229 . Therefore this chapter does
not represent novel work, although it does provide context for the previous device optimisation
studies so it is still important and relevant to this thesis.
This chapter is formatted as follows: In Section 6.1 some practical considerations for designing
a physical vibration absorber are discussed including issues regarding the certification of these
devices for use in a commercial aircraft. In Section 6.2 the required geometry and available
stroke for the vibration suppression device is considered with a view to incorporating the device
within the profile of the primary strut of the BUG-T model. In Section 6.3 a fluid-based hydraulic
damper based on the work by Rittweger et al. 191 is used to determine the range of equivalent
viscous damping coefficients that are realisable by a device that fits inside the strut profile. In
Section 6.4 a fluid-inerter is introduced and various parameter studies are run to determine the
range of device inertance values that can be achieved. Estimates of the device mass are made in
Section 6.3 for the hydraulic damper and 6.4 for the fluid-inerter.
141
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
As a vibration suppression device is a novel and unusual design feature in a fixed wing aircraft
there are likely to be many technical challenges to the certification of this technology, however,
because of this novelty, the requirements are not explicitly defined in any of the existing certifica-
tion documents. Discussion on the use of "Passive Flutter Dampers" is provided in Section 5.1.4.3
of CS-25 234 , although such devices are concerned with suppressing control surface flutter and
not providing gust loads alleviation - however these requirements can still be used as guidelines.
For example, the probability of device failure would need to be assessed, especially for the case
where multiple devices are included, and a fail-safe design methodology would need to be adopted,
similar to the certification of the Boeing 747-8 flutter suppression technology 284 . Here, the
strength and stability of the airframe must be guaranteed even for conditions where the device
had failed. Furthermore, the capabilities of the passive system would have to be demonstrated
in wind tunnel tests and associated analysis models validated. Inspiration could also be taken
from the rotorcraft industry where blade lead-lag dampers have been designed and certified
for a number of helicopters 192 and anti-vibration control systems have been installed on the
EuroCopter EC225 238 and EC130T2 239 . A summary of some of the other key considerations are
detailed below:
• Operating Frequency - The frequencies at which the device reach peak efficiency must
be identified before beginning the detailed design of the device. In this study the flutter
frequencies are used to estimate the desired operating frequency of the device and a
sensitivity study is also conducted for a fixed device design to determine the effect of
frequency on the equivalent viscous damping coefficient.
• Available Stroke - The required stroke1 must be defined as this drives the geometry of
the device and defines the available fluid volume.
• Geometry - The available space and the attachment locations must be specified so that
the device can be designed such that it meets all geometrical constraints. This is especially
important in the context of a TBW as the device must fit within the internal volume of the
aircraft otherwise additional drag penalties will be incurred.
• Device Mass - Depending on the application the device mass may be one of the main
design constraints. In an aerospace context it is critical that the device mass is minimised
whilst still providing the correct response as unnecessary mass leads to a decrease in
mission performance.
• Attachment Loads - As the device is transmitting a force into the structure the local
loads in the vicinity of the attachment points will be larger than those predicted from
an analysis which does not consider the effects of the device, such as a static aeroelastic
design case. Therefore additional factors may have to be defined in the design of structures
which include these devices so that the required stiffness and mass properties are not
1 The device stroke is the distance the piston head travels in half a cycle.
142
6.2. GEOMETRIC CONSTRAINTS AND ESTIMATING DEVICE STROKE
underestimated.
• External Loads - The device must be designed to withstand externally applied loads
which are transmitted to the device from the primary system. In the case of a TBW these will
be loads resulting from static and dynamic conditions such as: steady-state manoeuvres,
gust and turbulence encounters and landing. Pressurisation loads will also need to be
considered to ensure the pressure vessel does not fail as the atmospheric pressure varies
across the mission flight profile.
As noted in Chapter 2, the use of vibration absorbers is widespread throughout several industries
and there is a wealth of knowledge related to the design and characterisation of these devices. In
this work particular attention is paid to the work by Rittweger et al. 191 , Swift et al. 228 and Liu
et al. 229 : Rittweger et al. 191 provides an excellent reference for the design of passive damping
devices in an aerospace context and his equations on a hydraulic damping concept are used in
Section 6.3 to design a translational damping device. Swift et al. 228 and Liu et al. 229 detail
the design and modelling of a fluid inerter device and their methods are used in Section 6.4 to
estimate the range of available inertance values. Note a key assumption in the following sections
is that the design rules presented in these papers can be applied freely without considering the
complex flow behaviour that is inherent to a fluid-based device. This is a gross simplification
however as the focus is on simply understanding the range of parameter values that can be
realised this approach is acceptable.
Finally, as a fluid-based device is being investigated the term pressure vessel will be used
interchangeably with device and/or damper to describe the compartment which houses the fluid
and other associated components such as the piston rod and piston head. An example is shown in
Fig. 6.3 with the various parts annotated for clarity.
Before proceeding with the preliminary design of the vibration absorber it is necessary to
understand the available space for housing the device and also what magnitude of stroke is
available as the device input. Furthermore, as location A is the strut root pinned joint the relative
motion is in fact a rotation, therefore the translational stroke must be generated via a lever arm
which is based on the available geometry. Vibration absorbers that respond to relative rotations
are also available, such as the motorcycle steering compensator detailed in Jiang et al. 216 or the
shock absorber patented by Peo and Lautz 285 , however it was decided to focus on translational
devices given the larger volume of research available on this topic.
Regarding geometric constraints, Fig. 6.1 shows how the moment arm, maximum thickness
and pressure vessel diameter are related. Here, the assumption is made that the device lies
within the profile of the strut, which represents a conservative estimate of the available geometry
for housing a vibration suppression device in a truss-braced wing aircraft. Other integration
143
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
φ pv
Line of action
t max /2 of device force
∆θ , ∆ω r
Hinge line
F IGURE 6.1. Diagram showing line of action of the device force with respect to a
rotational joint and the associated lever arm.
concepts could be considered, for example as location A is collocated with the strut-fuselage joint
it is possible that the device could be included in the belly fairing of the fuselage. This would
offer substantially more space however the geometry requirements of the device would need to be
traded against the design of the landing gear, which for high-wing aircraft is typically situated in
the fuselage belly-fairing. In this chapter the device integration scheme in Fig. 6.1 will be used,
the rationale being that if a device can be designed that fits into this space and provides the
necessary linear force coefficients then it is highly likely that a device located in the belly-fairing
would also be appropriate. Furthermore, as the device is confined to the strut profile the following
discussions are also applicable for a device at location B or C. To calculate the available geometry
for the pressure-vessel diameter a NACA 0018 aerofoil is used with a chord length of 521mm
yielding a maximum possible section thickness (t max ) of 93.76mm2 . Assuming that the hinge lies
along the centreline of the section this leaves a maximum height of 45mm for the lever arm r and
the pressure vessel diameter φ pv , meaning the following relationship must hold φ pv + 2r ≤ t max .
As with the device location, the line of action of the device force does not necessarily have to be
above the hinge line but as this represents a conservative estimate it is appropriate to proceed.
Estimating the device stroke is not straightforward as the flutter modeshapes are relative
quantities and the gust analysis in Chapter 5 showed that the relative motion at the device
locations is a function of gust gradient. Instead a parameter study will be performed to determine
the range of device strokes that can achieve the required viscous damping coefficient using values
between 1 - 10° for the relative rotation about the hinge3 (∆θ ). Once the lever arm has been
chosen the translational device stroke can be calculated from X 0 = r ∆θ . Figure 6.2 shows the
relationship between the pressure vessel diameter and the normalised lever arm r = 2r/t max .
Here normalised lever arm values between 5% and 25% of the section thickness have been trialled
and the corresponding pressure vessel diameter has been calculated. Figure 6.2(a) shows that
2 This is the same profile and strut geoemtry as the SUGAR High 765-095 Rev. D.
3 These values are of similar magnitude to the gust analysis results shown in Chapter 5.
144
6.2. GEOMETRIC CONSTRAINTS AND ESTIMATING DEVICE STROKE
F IGURE 6.2. Pressure vessel diameter and available device stroke as a function of
lever arm and joint rotation.
for the strut dimensions considered the pressure vessel diameter is between 22mm and 40mm
depending on the chosen lever arm value. Next, the relative rotation is included in the trade
study and the translational device stroke is calculated for all values of ∆θ and r. As shown in Fig.
6.2(b) translational device strokes between 0.2mm and 1.8mm are available based on the section
geometry and candidate hinge rotation. These values are in line with the design data presented
by Rittweger et al. 191 and Panda et al. 192 so it is likely that a viable device can be obtained
despite the seemingly small value of device stroke. However, it should be noted that for such
small values of device stroke the effects of freeplay and/or friction in the pressure vessel seals
and piston rod/cylinder will have a significant effect on the device performance 230 . Therefore, one
option could be to increase the device stroke via additional leverage or gearing mechanisms 189 ,
however this would require a more detailed design which is beyond the current scope. In the
following sections the identified values of device stroke will be used to calculate approximate
dimensions for three different device concepts in order to determine whether the damping and
inertia values identified by the optimisation process are feasible.
145
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
In this section the hydraulic damper concept is introduced and a series of parameters studies are
performed to understand the range of viscous damping coefficients and spring stiffness values
that are obtainable for the pressure vessel geometries described in Section 6.2. This section is
split into three parts: In Section 6.3.1 the equivalent viscous damping coefficient is estimated
using the method described in Rittweger et al. 191 . In Section 6.3.2 the equivalent spring stiffness
resulting from the fluid compressibility is calculated for a range of pressure vessel geometries.
Finally, in Section 6.3.3 a component breakdown of the hydraulic damper is used to estimate the
mass of the damping device.
Figure 6.3 shows a diagram of a simple hydraulic damper as described by Rittweger et al. 191 .
The damping force for this concept is generated via losses from turbulent flow which occur when
the fluid is forced through an orifice in the piston head. In this model it is assumed that the
only mechanism for fluid to be transferred between the two chambers is via this orifice, however
in a real-life viscous damper there will be some leakage around the sides of the piston head
as discussed in Liu et al. 229 . Using this model the equation for the damper force is given in
Milwitzky and Cook 286 as
A 3net ρ f
FD = Ẋ 2 , (6.1)
2α2 A 2o
where FD is the damper force, A net is the net area of the piston head π φ2 − φ2r /4 , ρ f is the
¡ ¡ ¢ ¢
fluid density, α is a loss factor which has a value between 0.6 and 0.64 191 , A o is the area of the
L pv
φo
φr
FD , Ẋ , X˙0
φ pv φ
Piston rod
Piston head
ρf Pressure Vessel
F IGURE 6.3. Diagram of a simple hydraulic damper showing the principal dimensions,
derived from Rittweger et al. 191 .
146
6.3. HYDRAULIC DAMPER CONCEPT
orifice πφ2o /4 and Ẋ is the velocity of the piston head. In Rittweger et al. 191 the coefficient terms
¡ ¢
A 3net ρ f
c h ydraul ic = , (6.2)
2α2 A 2o
Note, the symbols φ, φr , φ o are the same as those used in Fig. 6.3 and refer to the diameter
of the piston head, the diameter of the piston rod and the diameter of the orifice respectively.
Examining Eqn. 6.2 it is obvious that a larger hydraulic damping coefficient can be achieved by
maximising the ratio of the net piston area to the orifice area, which in physical terms corresponds
to generating a higher flow velocity through the orifice which leads to increased losses from the
turbulent flow. Clearly the damping force is nonlinear with respect to the piston velocity and
so the coefficient terms cannot be used directly as the linear viscous damping coefficient. In
such cases it is common to derive an equivalent viscous damping coefficient (c equiv ) based on
the energy dissipated in a single cycle. Here, the energy dissipated in a single cycle WD of an
equivalent viscous damper with damping force FD = c equiv Ẋ is given by the integral
Z Z
WD = FD dx =⇒ WD = c equiv Ẋ 2 dt, (6.3)
where the substitution dx = Ẋ dt has been used to transform the integral into the time domain.
Furthermore, assuming cyclic loading the substitution X = X 0 sin (ω t) can be made which yields
an expression for piston velocity Ẋ = X 0 ωcos (ω t) allowing the integral to be evaluated with the
cycle frequency as the upper limit
Z 2π/ω
2
WD = c equiv ω X 02 cos2 (ω t) dt, (6.4)
0
where X 0 is the stroke and ω is the cycle frequency. Finally, using the trigonometric identity
cos (2ω t) = cos2 (ω t) − sin2 (ω t) the cos2 (ω t) term can be simplified to 1 + cos (2ω t) /2 allowing the
work done by an equivalent viscous damper to be defined as
WD = ω X 02 c equiv π, (6.5)
A similar process can be used to derive the work done by the hydraulic damper, however in the
interest of brevity this term is given by Rittweger et al. 191 as
8
WD = c h ydraul ic ω2 X 03 . (6.6)
3
Assuming that the amount of work done by the hydraulic damper and the equivalent viscous
damper is equal, it is possible to define the equivalent viscous damping coefficient in terms of the
hydraulic damping coefficient
8 8 A 3net ρ f
c equiv = ω X 0 c h ydraul ic =⇒ c equiv = ω X 0 . (6.7)
3π 3π 2α2 A 2o
147
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
This equation indicates that the equivalent viscous damping coefficient is a function of the
damper geometry, the choice of fluid, the device stroke and the excitation frequency, implying
that the desired damping coefficient value could be achieved in a number of ways. For example
the device stroke, operating frequency or fluid density could be adjusted to achieve a linear
variation in c equiv , or the ratio of the piston head and orifice areas could be tailored to promote
the desired flow behaviour through the orifice. Using Eqn. 6.7 it is possible to determine the
range of available viscous damping coefficients for the hydraulic damper concept by considering a
range of device strokes, frequencies and geometry combinations. Figure 6.4 shows the minimum
and maximum values of equivalent viscous damping coefficient that are achievable when the
following parameter study is performed: First, values of excitation frequency in the range 0Hz
to 10Hz and device stroke in the range 0.1mm to 1.8mm are used to cover the frequency and
(a) Minimum viscous damping coefficient - N/ms-1 (b) Maximum viscous damping coefficient - N/ms-1
F IGURE 6.4. Range of viscous damping coefficients for the hydraulic damper concept
using the damper properties in Table 6.1.
148
6.3. HYDRAULIC DAMPER CONCEPT
F IGURE 6.5. Equivalent viscous damping coefficient and orifice diameter for a range of
piston and normalised orifice diameters using X 0 = 0.2mm f = 2.8Hz. Remaining
geometrical properties are given in Table 6.1.
stroke values of interest. Next, for each combination of device stroke and excitation frequency the
equivalent viscous damping coefficient is calculated for a range of piston and orifice diameters
using the values in Table 6.1 - Note, a factor of 0.8 is used for the normalised piston diameter
to account for the thickness of the pressure vessel and values for the loss factor and the oil
density are taken from Rittweger et al. 191 . Finally, the maximum and minimum values of c equiv
are retained for each value of X 0 and ω. These results show that equivalent viscous damping
coefficients up to 1 × 108 are possible, which gives confidence that the viscous damping coefficient
values considered within the optimisation process are viable in the context of a truss-braced
wing.
A further parameter study is performed to understand the effect of orifice diameter on the
viscous damping coefficient. Here, a conservative estimate of 0.2mm for the device stroke is used
and the excitation frequency is defined as 2.8Hz based on a flutter reduced frequency of 0.135
and a velocity of 212m/s. Next, the equivalent viscous damping coefficient is calculated for a
range of piston and orifice diameters using the values in Table 6.1. Figure 6.5 shows that for this
set of parameters values of c equiv up to 1 × 106 are achievable. Furthermore, for a target viscous
damping coefficient in the range 104 - 105 a normalised orifice diameter of between 1% and 2.5%
of the piston head should be used, which corresponds to an orifice diameter of approximately
0.2mm to 0.6mm based on the contour plot in Fig. 6.5(b).
149
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
k f luid c equiv
T1 T2
F IGURE 6.6. Network diagram for a damper including fluid compressibility effects.
So far the discussion of the damper properties has focussed on calculating an equivalent viscous
damping coefficient based on the key parameters of the device. However, it is well known that
in this type of damper idealisation additional effects are present that are attributed to the
compressibility and mass of the fluid. These effects can be idealised as additional spring and
mass/inerter elements connected in series or parallel with the damper, and examples of such
networks are provided in throughout the literature 191,228,229 . In the simplest of cases these
additional effects are ignored however it is common to include a spring in series with the
equivalent viscous damper in order to model the compressibility of the fluid, as in Fig. 6.6. The
series spring has the effect of limiting the effectiveness of the damper by reducing the relative
motion across the nodes of the damper, as shown in Rittweger et al. 191 . Therefore it is desirable
to have the highest possible stiffness for the series spring in order to maximise the work done by
the viscous damper. It is likely that there will be numerous design drivers which determine the
length of the pressure vessel however, in lieu of more detailed information, Rittweger et al. 191
provides the following analytical expression for the effective stiffness of the fluid
E e f f A net
k f luid = (6.8)
L pv
where L pv is the length of the pressure vessel and E e f f is the effective modulus of the fluid
and the pressure vessel, which itself is a function of the pressure vessel geometry, the modulus
F IGURE 6.7. Fluid stiffness for varying pressure vessel length and diameter - N/m.
150
6.3. HYDRAULIC DAMPER CONCEPT
of the fluid and the modulus of the pressure vessel material. Based on this equation it would
be possible to use a fluid-spring as the spring element of the TID or TID-D layouts by simply
removing the orifice in the piston head, although, it should be noted that there will still be a small
amount of damping due to leakage of fluid around the piston head 229 . In this study, E e f f is set
to 2280N/mm2 using the value from Rittweger et al. 191 and a parameter sweep is performed of
A net and L pv to gain an appreciation for the magnitude of the series spring stiffness. The results
of this study are shown in Fig. 6.7 for vessel lengths between 10mm and 100mm and values of
A net corresponding to the data in Table 6.1. Here spring stiffness values in the range 5.3 × 106
and 1.8 × 108 have been achieved, again providing evidence that the stiffness values considered
in the optimisation process are appropriate for this application.
Given that much of the design of aircraft structure is driven by mass considerations it is prudent
to estimate the mass of the damper in order to understand the potential mass penalty that could
arise from using these devices. Unfortunately detailed analytical estimates of the mass of a
fluid-based shock absorber are not readily available given the commercially-sensitive nature
of this technology. This makes the analytical estimation of the mass of a vibration suppression
device a difficult task as very little reference material is available. To bypass this issue the device
is divided into its subcomponents and the mass is calculated from the geometry of the component
and the material/fluid density. Therefore, considering mass contributions from the fluid, pressure
vessel casing and the piston rod, the mass of the damper can be defined as
πn h³ ´ ¢ i o
L pv φ2pv − φ2 ρ pv + φ2r ρ pv + φ2 − φ2r ρ f + 2φ2pv t pv ρ pv ,
¡
M damper = (6.9)
4
where the first term inside the brackets defines the contribution from pressure vessel casing4 ,
the second term is the mass of the piston rod and the third term is the mass of the fluid. The
term 2ρ pv φ2pv t pv accounts for the mass of the two end-plates that seal the pressure vessel,
with the thickness of the plate (t pv ) assumed to be equal to the pressure vessel thickness, i.e.
t pv = φ pv − φ /2. In a typical civil engineering application the thickness of the pressure vessel
¡ ¢
will be driven by the hoop stress resulting from the pressure differential across the internal and
external diameters. In an aircraft landing gear shock absorber, an example of a similar technology
application, fluid pressures up to 5000psi (345bar) are typical 288 . Using this as a reference point
for the damper fluid pressure and assuming the largest possible pressure vessel diameter of
40mm and a yield strength of 350MPa yields a wall thickness of 1.3mm at ultimate loading5 ,
which is in-line with the 0.8 knock-down factor assumed by Rittweger et al. 191 . As the calculated
thickness to resist hoop stress is quite low it is unlikely this will be the critical design case for
4 A steel alloy with density ρ -3
pv = 8050kgm has been chosen as a first estimate for the pressure vessel material
based on the comments in Bickell 287 .
5 Assuming a safety factor of 1.5 the pressure vessel thickness to prevent failure at ultimate loading is given by
t pv = P f φ/3σ y , where P f is the fluid pressure and σ y is the yield stress of the pressure vessel material.
151
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
the pressure vessel casing, instead external attachment loads are likely to drive the sizing of the
device components. For example, the maximum damper force from the gust analysis in Chapter 5
can be used to estimate the buckling length of the piston rod. Using a critical buckling load of
25,433N and a rod diameter of 6.4mm yields a buckling column length of 40mm6 . A piston rod
length of 40mm may not be sufficient for transmitting the device force into primary structure,
however these dimensions are a consequence of requiring the device to fit inside the cross-section
of the strut. Therefore it is likely that more appropriate device dimensions could be obtained if the
a different device integration scheme was considered. This simple example clearly demonstrates
the need to consider all aspects of the device design when trading the use of vibration absorbers
for aeroelastic control.
Figure 6.8 shows the variation in the damper mass as a function of the pressure vessel length
and diameter, with the remaining parameters defined in Table 6.1. Here masses in the range
0.05kg to 0.5kg have been demonstrated however due to the simplicity of the calculations involved
this is clearly overly optimistic. A better estimate could be gained by including components such
as seals, fittings and the piston head, as well as calculating the pressure vessel thickness based
on the required stress and fatigue performance. It is likely that once these considerations are
included the device mass will be of the order of tens of kilograms, therefore the addition of a
damper to provide aeroelastic control is unlikely to have a significant mass penalty in the context
of aerospace wing structures.
6 Assuming a Youngs modulus for steel of 200GPa and fully fixed boundary conditions for the Euler buckling
p
formula gives, L = π2 EI/4P cr . These dimensions are for a device with an outer pressure vessel diameter of 40mm
and a rod diameter which is 20% of the piston diameter, as in Rittweger et al. 191 .
F IGURE 6.8. Damper mass as a function of pressure vessel length and diameter for
the geometry data in Table 6.1.
152
6.4. FLUID INERTER CONCEPT
In addition to using a fluid-based device to generate damping forces it is also possible to use the
mass of the fluid to generate additional inertia effects. In this section, the conceptual fluid inerter
proposed by Swift et al. 228 and more recently investigated by Liu et al. 229 will be assessed to
understand the scale of inertance values that can be realised by using a fluid-based device. In this
concept the inertance is generated by forcing the fluid around a helical channel that surrounds
the pressure vessel, as shown in Fig. 6.9. In addition to generating inertance, the movement of
the fluid along the helical channel also leads to energy losses (i.e. damping) as a result of pressure
drops and turbulent flow at the inlet and outlet of the channel 228 . These additional terms enter
the network layout as a parallel inerter and damper attached in series to the viscous damper
network from Section 6.3, yielding the network shown in Fig. 6.10. As discussed by Swift, the
advantage of a fluid inerter over the conventional mechanical inerter is the relative simplicity of
the design as well as improved durability due to the reduction in the number of moving parts.
Furthermore, the amount of damping and inertance can be tailored by varying the diameters
of the various orifices and channels in the device, presenting the possibility of a semi-active or
active device whose properties can be tailored to the desired operating conditions. Further details
are provided in Liu et al. 229 . The inerter described in this section can be thought of as the pure
inerter element in the TID and TID-D layouts, except with additional components resulting from
the damping and stiffness effects of the fluid.
In rest of this section the inertance from the helical channel is estimated using the geometry
of the hydraulic damper presented in Section 6.3. Damping contributions from the helical channel
are not addressed but could be readily incorporated using the methods described by Swift et al. 228
and Liu et al. 229 . The inertance value for the fluid in the helical channel is given as 228
A 2net
b hc = ρ f l hc , (6.10)
A hc
where b hc is the inertance from the helical channel, l hc is the length of the helical channel and
Helical channel
φhc
FD , Ẋ , X˙0
F IGURE 6.9. Diagram of a conceptual fluid inerter device, derived from Swift et al. 228 .
153
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
c hc
k f luid c equiv
T1 T2
b hc
F IGURE 6.10. Network diagram of the fluid inerter, taken from Swift et al. 228 . Terms
related to the helical channel are denoted by the subscript hc.
A hc is the cross-sectional area of the helical channel. Assuming that both the pressure vessel and
the helical channel have circular cross-sections this equation has the form
πφ4net
b hc = ρ f l hc , (6.11)
4φ2hc
where φhc is the diameter of the helical channel which can be calculated for known values of
A net and A net /A hc . Using the geometry data in Table 6.1 the potential values of inertance are
determined by sweeping through values of l hc and A net /A hc . In this study two configurations
are considered for the helical channel. In the first configuration the channel is assumed to be
wrapped around the exterior of the pressure vessel as shown in Fig. 6.9. This means that the
length of the helical channel is coupled with the length of the pressure vessel and the diameter
of the channel, as shown in Equation 6.13. In the second configuration the helical channel is
located externally to the pressure vessel and the available length is not linked to the pressure
vessel geometry, as in Liu et al. 229 . To enable the first configuration to be considered a geometric
expression for the helical channel length must be developed in order to link the dimensions of
the pressure vessel to the length of the channel. Assuming the channel forms a perfect helix the
length of the fluid path is given by
p
l hc = N h2 + c 2 , (6.12)
where N is the number of revolutions, h is the rise of the helix per revolution and c is the
circumference of the helix, c = πφ path . For the configuration where the helix wraps around the
exterior of the pressure vessel the diameter of the helix is simply the sum of the pressure vessel
and channel diameters, i.e. φ path = φ pv + φhc . Also, assuming the channel tubing is perfectly
aligned with no gaps and ignoring the thickness of the tubing, the rise of the helix is given by the
diameter of helical path. Finally, the number of revolutions can be determined from the number
of channel diameters that will fit along one side of the pressure vessel, i.e. N = L pv /φhc . Therefore
the helical channel length in terms of the pressure vessel dimensions and channel diameter is
given by,
L pv q 2 ¢2
φhc + π2 φ pv + φhc .
¡
l hc = (6.13)
φhc
154
6.4. FLUID INERTER CONCEPT
Using this equation the variation in the length of the helical path can be found by performing
a two level parameter study. At the first level the pressure vessel length (L pv ) and area ratio
(A net /A hc ) are defined, then the corresponding helix length and fluid inertance are calculated
using the geometry in Table 6.1. The maximum values of l hc and b hc obtained from this study
are presented in Fig. 6.11. Figure 6.11(a) shows that path lengths up to 13 times the vessel
length can obtained and the magnitude of the path length (i.e. in thousands of mm) matches the
results presented by Swift et al. 228 . Regarding the fluid inertance, Fig. 6.11 shows that values
up to 500kg are obtainable however this is only for the most extreme case of L pv = 500mm and
A net /A hc = 1%. This makes physical sense as a longer vessel allows for more revolutions of the
helix and hence a longer channel length, also, a higher area ratio yields a larger flow velocity
in the channel which provides a higher inertance. However, it should be recognised that these
values represent a theoretical maximum and for this size of device inertance values in the range
of 100kg are more typical 228 . Relating these results to the data used in the optimisation process,
an inertance of 100kg is roughly equivalent to a mass ratio of 2.64%7 which is certainly not
the same magnitude as the mass ratios used in the optimisation process. This highlights the
importance of relating the bounds of the design variables to the physical properties of the device
in order to ensure that only feasible devices configurations are considered.
If a larger inertance value is desired then one option could be to increase the length or
diameter of the pressure vessel, however this would impact the performance of the damper by
reducing the equivalent fluid stiffness. An alternative is for the helical channel to be placed
7 Calculated using the half-wing mass of the SUGAR 765-095-Rev. D of 3781kg.
F IGURE 6.11. Variation in helical channel length and fluid inertance as a function of
pressure vessel length and channel area ratio. Caclulated for the case where the
helical channel length is coupled to the pressure vessel length.
155
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
externally from pressure vessel and connected via additional tubing, as demonstrated by Liu
et al. 229 . This would decouple the helical channel length from the overall length of the pressure
vessel and allow longer channel lengths to be obtained. For example Liu demonstrated that helical
channel lengths up to 12m are possible, however this was in a laboratory setting and in a practical
application this would be subject to the available volume at the device location. To demonstrate
the range of available inertance values that can be realised by decoupling the channel length
and pressure vessel geometry an additional parameter study is performed. Here, the helical
channel length is varied between 0.01m and 10m and the inertance value is calculated for a range
of channel diameters corresponding to area ratios between 1% and 10% of the net piston area
A net . Figure 6.12 shows the resulting inertance values for the cases where the pressure vessel
diameters are equal to 22mm and 40mm, which represent the maximum and minimum values for
a device at location A. These results show that a much larger range of inertance can be achieved
with a maximum of 700kg and a minimum of 0.11kg, corresponding to a channel with the longest
length/smallest diameter and the shortest length/largest diameter respectively. These values are
in line with the inertance considered during the optimisation process, however it must be noted
that these results represent theoretical values based on simple design formulae. Meaning that
the actual realisable values of inertance may vary significantly once a more detailed model is
adopted 229 .
Finally, as the fluid inerter considered in this section is a simple extension of the hydraulic
damper in Section 6.3 the mass of the inerter can be estimated using a similar approach but with
(a) Inertance values (kg) for φ pv = 22mm (b) Inertance values (kg) for φ pv = 40mm
156
6.4. FLUID INERTER CONCEPT
an extra term to account for the mass of the fluid in the helical channel (M hc ).
As before, this represents an optimistic estimate of the device mass and requires further
refinement to provide an accurate estimate. Figure 6.13(a) shows the variation in the mass of
the fluid inerter as a function of pressure vessel length and diameter and Fig. 6.13(b) shows the
variation in the mass of the fluid in the helical channel, note the range of pressure vessel lengths
has been increased compared to the damper to accommodate the required helical channel lengths.
As with the hydraulic damper concept, the estimated mass of the fluid inerter is small in the
context of aerospace structures, implying that the device mass will not be barrier for further
implementation of this concept. These results also show that the additional mass due to the fluid
in the helical channel is small and accounts for approximately 7% of the overall mass of the device.
Therefore for a given length and diameter the mass of the fluid inerter is roughly equivalent to
the mass of the hydraulic damper. For the case where the helical channel length is coupled with
the length of the device, achieving a large inertance value requires a longer pressure vessel and
hence a larger mass. This approach would also deteriorate the damping performance due to a
reduction in the equivalent fluid stiffness 191 resulting from a longer vessel length. Therefore
it is clearly beneficial to decouple the length of the helical channel and the length of the device
by adopting the configuration presented by Liu et al. 229 . This would allow the device length to
be designed to maximise damping performance whilst the length of the helical path could be
tailored to provide the correct inertance.
(a) Mass of fluid inerter device - kg (b) Mass of fluid in helical channel - kg
F IGURE 6.13. Fluid inerter mass as a function of pressure vessel length and diameter
for the geometry data in Table 6.1.
157
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
In this section device layouts from the single absorber flutter optimisation are used to define a
set of target parameter values in terms of equivalent viscous damping coefficient, spring stiffness
and inertance - as in Table 6.2. Note that when designing these devices it is assumed that the
normalised rod diameter (φr ), loss factor (α), fluid density (φ f ) and normalised piston diameter
(φ) have the values specified in Table 6.1, therefore the only parameters that need to be chosen are
the operating frequency ( f ), device stroke (X 0 ), pressure vessel diameter (φ pv ) and the normalised
orifice diameter (φ o ), as well as the helical channel properties φhc and l hc .
This section will consider the design of a hydraulic damper that can achieve the viscous damping
coefficients for the pure damper and TID-D layouts in Table 6.2. The first step is to choose the
design operating frequency and device stroke as these are required to determine the equivalent
viscous damping coefficient.
• When designing a device for flutter suppression the device operating frequency can be
derived from the flutter frequency, which for the BUG-T model is 2.8Hz. For GLA the
operating frequency is not as obvious, although it could be based on the natural frequency
of the mode which the device is most likely to influence. For the purposes of this study the
the design operating frequency will be set to 2.8Hz.
• The device stroke will usually be determined by geometric constraints. For the case where
the device is located inside the cross-section of the strut the available device stroke is a
function of the pressure vessel diameter and the assumed rotation about the strut joint, as
shown by Fig. 6.2. In this example a conservative device stroke of 1mm is specified which
allows the pressure vessel diameter to be calculated by extrapolating the data in Fig. 6.2.
Here, a device stroke of 1mm is possible for values of r and ∆θ in the range 0.13-0.25 and
5-10° respectively. Assuming a conservative value of 6° for ∆θ yields a pressure vessel
diameter of 25mm.
Now that f , X 0 and φ pv have been calculated the only remaining parameter to be determined
is the normalised orifice diameter. Figures 6.14(a) and 6.14(b) shows the results of a sweep in
φ pv and φ o using the values in Table 6.1. The contour relating to a target viscous damping
TABLE 6.2. Target device parameters from the single absorber flutter optimisation.
158
6.5. INITIAL DESIGN OF CANDIDATE VIBRATION ABSORBERS FOR AEROELASTIC
CONTROL
coefficient of 1 × 105 is highlighted in red on Fig. 6.14(a), the next contour above this corresponds
to a viscous damping coefficient of 3.16 × 104 which is very close to the desired viscous damping
coefficient of the TID-D device in Table 6.2. Interpolating these two contours at φ pv = 25mm
yields a normalised orifice diameter of 0.9% and 1.2%, corresponding to orifice diameter of
approximately 0.2mm for both cases. For completeness it is possible to determine the equivalent
spring stiffness of the fluid and the damper mass by choosing an appropriate value of L pv . Using
values of pressure vessel length in the range 10-100mm yields a k f luid between 6.8 × 106 and
6.8 × 105 N/m and a damper mass between 30-175g. At this point is important to remember that
these design formulae are highly idealised and do not include additional considerations such as
attachments and additional material to resist external loading.
(c) Equivalent spring stiffness of the fluid - N/m (d) Damper mass - kg
F IGURE 6.14. Physical parameters for a viscous damper with c equiv = 1 × 105 .
159
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
F IGURE 6.15. Physical parameters for the final TID and TID-D fluid springs.
This section will consider the design of a hydraulic spring that can achieve the spring stiffness
values for TID and TID-D layouts in Table 6.2. Fixing the effective modulus of the fluid at
2280N/mm2 (i.e. the value from Rittweger et al. 191 ) and conducting a simple trade in φ pv 8 and
L pv yields a set of possible spring stiffness values which are calculated using Eqn. 6.8. Figure
6.15 shows the results of this trade study for pressure vessel lengths between 50-1000mm and
diameters between 5-40mm, with the contours relating to the target k f luid values highlighted
in red. The first point to note is that in order to achieve the comparatively low spring stiffness
values (compared to the damper designs in Section 6.3.2) the pressure vessel aspect ratio must be
increased, i.e. the length increases and the diameter decreases. This is because the required value
of k f luid is between two and three orders of magnitude lower than the values from Section 6.3.2. In
order to keep the mass to a minimum the smallest possible diameters are chosen which satisfy the
required stiffness values, yielding a values of φ pv = [5mm, 6.2mm] and L pv = [227mm, 50mm] for
k T ID and k T ID −D respectively. Inserting these values into Eqn. 6.9 gives a mass of approximately
100g for each fluid spring.
This section will consider the design of a fluid inerter that can achieve the inertance values for
TID and TID-D layouts in Table 6.2. Here the target inertance for both the TID and TID-D is
set to 378.1kg, which is equivalent to an inertance-to-mass ratio of 10%. of the BUG-T wing. As
8 Only φ
pv needs to be traded as the rod diameter is fixed at 20% of the piston diameter and the piston diameter is
fixed at 80% of the pressure vessel diameter
160
6.5. INITIAL DESIGN OF CANDIDATE VIBRATION ABSORBERS FOR AEROELASTIC
CONTROL
F IGURE 6.16. Physical parameters for the final TID and TID-D fluid inerter.
shown by Eqn. 6.11, the inertance of the fluid inerter is governed by the length of the helical
channel, the net piston head diameter and the helical channel diameter. Assuming circular
cross-sections the inertance scales to the fourth power of φnet and is inversely proportional to φ2hc ,
meaning that the easiest way to generate a large inertance value is to have a large net diameter
and a small diameter helical channel. To obtain an estimate for the available inertance a trade
study is performed for the helical channel length and diameter with the pressure vessel diameter
set to the maximum value of 40mm and the pressure vessel length fixed at 100m. The results of
this parameter sweep are shown in Fig. 6.16 with the contour for the target value of inertance
highlighted in red. It is clear that only a small corner of the parameter space is able to yield
this high value of inertance, with valid path lengths between 5-10m and an area ratio between
1-2% yielding a mass for the fluid inerter between 580-700g. Furthermore, the large values of
l hc indicate that the helical channel must be decoupled from the length of the pressure vessel
as discussed in Section 6.4. In the interest of minimising the device mass the minimum viable
dimensions are chosen, which are l hc = 5000mm and A net /A hc = 1%.
This section presents a summary of the initial device design for a damper, TID and TID-D located
at the strut-root joint which can obtain the target parameter values from the single absorber
flutter optimisation in Chapter 4. Table 6.3 shows the final values of the physical parameters that
were traded in the previous sections in order to obtain the target parameter values from Table 6.2.
These results show that the target parameter values are easily achievable by devices which will
fit inside the cross-section of the strut and that there is a negligible mass penalty associated with
161
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
TABLE 6.3. Final physical parameters for a damper, TID and TID-D device at the
strut-root joint for the target parameter values from the single absorber flutter
optimisation.
TABLE 6.4. Summary of flutter suppression and GLA performance for a single device
at the strut-root joint.
these devices. Once additional material is accounted for (i.e. to resist external loads) it is likely
that these devices will have a mass in the tens of kilograms, giving confidence that a passive
vibration suppression scheme can be used in a TBW with a negligible mass penalty.
Finally, the GLA performance of these devices have been evaluated in order to understand
how effective a device designed to suppress flutter can be at reducing gust loads. The gust analysis
described in Chapter 5 has been used to determine the gust response of the BUG-T model for the
three single device cases detailed in Table 6.2. The results in Fig. 6.17 and Table 6.4 show that for
the spanwise bending moment at the wing-fuselage location a small amount of GLA is available
from the TID and pure damper devices, although the TID-D device actually increases the loads
by an average of 1%. For the wing-strut loads all of the devices have an average increase in the
spanwise bending moment which matches the trends from Chapter 5. These results demonstrate
that a small amount of GLA is available for a device that is optimised to extend the flutter
boundary of a TBW aircraft, although the amount of GLA is far less than the 60-80% reductions
that have been reported in the literature for active control methods 171–173 . Even so, the flutter
suppression achievable by a passive vibration absorber is of a similar magnitude to active control
results 167,170 therefore there is still a clear benefit to including this concept as part of a strategy
to alleviate aeroelastic effects in TBW aircraft.
162
6.6. CHAPTER SUMMARY
(a) Change in wing-fuselage (WF) loads (b) Change in wing-strut (WS) loads
F IGURE 6.17. GLA performance for a single device at the strut-root joint using the
device parameters from the single absorber flutter optimisation.
This chapter has presented a preliminary study into the design of a physical vibration suppression
device that can provide the level of stiffness, damping and inertance values that were identified as
being beneficial for flutter suppression and gust load alleviation. Estimates for the device stroke
and pressure vessel diameter were derived from the depth of the strut profile at location A and the
device frequency was calculated from the flutter reduced frequency. Two concepts were considered
for the design of the device: a hydraulic damper and a fluid-inerter. Using simple design rules
a series of parameter studies were performed for both concepts to understand the influence of
the device stroke, frequency and geometry on the range of equivalent stiffness, damping and
inertance values. Based on these studies the following observations were made:
• The device diameter can vary between 22 - 40mm for a lever arm that varies between
5% and 25% of the strut cross-section depth at location A. Investigation of alternative
device integration schemes that place the device inside the belly-fairing of the fuselage are
recommended as future work.
• Assuming a hinge rotation of between 1°- 10°, this moment arms corresponds to device
strokes in the range 0.2 - 1.8mm, which are of the same scale as examples in the litera-
ture 191,228,229 .
• The hydraulic damper can be represented as an equivalent viscous damper with a spring in
series. For a given fluid density, the equivalent viscous damping is dependent on the area
ratio between the piston head and the orifice - with a larger ratio yielding a larger viscous
163
CHAPTER 6. PRELIMINARY DESIGN OF A VIBRATION SUPPRESSION DEVICE FOR
AEROELASTIC CONTROL
damping coefficient.
• The equivalent spring stiffness of the damper fluid is a function of the fluid and pressure
vessel modulus as well as the area of the piston head and the length of the pressure vessel.
For a given effective modulus the vessel length is the dominant parameter and it was noted
in Rittweger et al. 191 that a lower stiffness value reduces the effectiveness of the equivalent
viscous damper.
• For a hydraulic damper with an outer diameter between 22 - 40mm, equivalent viscous
damping coefficient and spring stiffness values up to 1 × 108 can be realised9 , although the
exact value is dependent on the absorber frequency and stroke value.
• The mass of the damper was estimated as less than 1kg however this was calculated using
a highly simplified formula. Despite this and given the scale of the dimensions involved it
is highly unlikely the mass of the damper would exceed tens of kilograms, meaning there
would not be a significant mass penalty from including these concepts in a truss-braced
wing.
• For the fluid inerter device, inertance values up to 700kg (µ = 0.18) are available however
this represents the theoretical upper limit of what is possible when the most favourable
device properties are considered, i.e. maximum pressure diameter, minimum channel
diameter, longest channel length.
• Two concepts for the helical channel were traded, one where the channel length is coupled
with the pressure vessel length and one where it was independent of the pressure vessel
geometry. When the length of the helical channel was linked to the length of the pressure
vessel inertance values of approximately 100kg were realisable, although this was at the
expense of increasing the mass and reducing the damping effectiveness. When the channel
length was decoupled from the vessel length values of inertance up to 700kg were generated,
corresponding to a channel length of 10m. However although the amount of volume required
by the helical channel was not considered and so it is not certain whether there is enough
space at the device location to accomodate the volume of additional tubing required.
• The additional mass of the fluid in the helical channel accounted for approximately 7% of
the mass of the fluid inerter, indicating that there is only a small weight penalty associated
with adopting a fluid inerter concept over the hydraulic damper.
• Using these simplified design formulae it is possible to design a collection of fluid-based
springs, dampers and inerters that can yield the parameter values identified by the single
device flutter optimisation in Chapter 4 with a minimal mass penalty. Furthermore, these
devices are capable of providing a small amount of GLA as well as extend the flutter points
which means they are a viable options for reducing aeroelastic effects in TBW.
A key source of uncertainty in this process is the exact stroke and operating frequency of the
device. Given that the device stroke cannot be extracted from the flutter modeshapes the value of
164
6.6. CHAPTER SUMMARY
the stroke may have to be defined as a fixed parameter which is traded in the design of the device.
Or, the stroke could be identified from a different type of analysis, for example a ground vibration
test or a simulated frequency response analysis. The dependency on frequency is less problematic
as the spring, damper and inerter force coefficients could be defined using a look-up table which
is referred to during the analysis10 . It is important that these stroke and operating frequency
are defined early on in the design as they can have a significant impact on the achievable device
performance. For example, Liu et al. 229 observed that stroke amplitudes between 5mm and 20mm
were required to obtain a good match between experimental results and theoretical predictions
for absorber frequencies between 0.2Hz and 20Hz, which was attributed to inaccuracies in the
modelling of the fluid stiffness and friction effects which are dominant at low frequencies.
Finally, if design equations can be constructed such that the equivalent spring stiffness,
viscous damping coefficient and inertance are functions of the device geometry it would be
possible to pose an optimisation problem where the design variables are the physical dimensions
of the device. This would allow geometry constraints to be directly integrated into the design of the
device and would also prevent infeasible force coefficients from being analysed. The generalisable
model of a fluid-inerter integrated damping (FID) device proposed by Liu et al. 229 provides
several useful equations which could be used to achieve this, however many of the parameters in
these devices have nonlinear terms which are not conducive to an optimisation within Nastran
Solution 200. Therefore it may be necessary to investigate alternative analysis and optimisation
methods which can include these nonlinear device terms in the optimisation.
10 Such an approach is readily available within Nastran using the PELAST and PDAMPT bulk data entries, however
165
HAPTER
7
C
C ONCLUSIONS AND F UTURE W ORK
his chapter presents a summary of the thesis and details the main conclusions along with
T some recommendations for future work. As each of the preceding chapters has included
a comprehensive conclusion section only the main findings of the thesis are provided and
the reader is directed to the specific chapters if they require further detail.
This thesis has presented an investigation into the design and optimisation of vibration sup-
pression devices in a truss-braced wing aircraft in order to achieve passive flutter suppression
and gust loads alleviation. Two aeroelastic finite element models were used for the analysis, a
full-span aircraft model and a half-wing model - both were referred to as the BUG-T model and
were based on the NASA/Boeing SUGAR 765-095 Rev. D aircraft.
The preliminary dynamic analysis in Chapter 3 indicated that the truss pinned-joints expe-
rienced relative motion across a range of excitation frequencies, allowing for the possibility of
vibration absorbers to be included at these locations to influence any adverse dynamic behaviour.
Furthermore, at higher excitation frequencies the combined bending of the wing and strut compo-
nents induced some relative motion across the terminals of the jury-strut, implying that a device
parallel to the jury-strut could also be viable. The static aeroelastic response of the full-span
BUG-T was compared against the results from the SUGAR sizing study and good agreement
was found between the BUG-T and SUGAR data, with errors of approximately 5% in the tip
deflection for the 2.5g and -1g manoeuvre cases. Given the approximations required to replicate
the SUGAR model these results indicated that the BUG-T model is a suitable test model for
conducting further studies into aeroelastic control via passive vibration suppression.
167
CHAPTER 7. CONCLUSIONS AND FUTURE WORK
Chapter 5 presented an investigation into the use of vibration absorbers to provide gust loads
alleviation in a truss-braced wing. The optimisation was formulated as a frequency response
problem with the device parameters optimised to target specific structural modes in the primary
system in order to reduce their participation in the gust response. This approach decoupled the
device design from the computationally expensive gust analysis and allowed for the possibility of
vibration suppression devices to be optimised as part of a large-scale finite element model more
typical of detailed stress analysis. Vibration absorbers were included at the three candidate device
locations and it was found that the wing gust loads envelope could be reduced by approximately
4% at spanwise locations inboard of the strut attachment point and that these improvements
were achievable across a range of gust gradients. Configurations with multiple absorbers were
investigated and it was shown that different absorbers could be tuned to target the response
of different structural modes. This is an important consideration for gust loads alleviation as
different gust gradients will activate different structural modes as a result of the gust signal
bandwidth.
Chapter 6 presented a preliminary study into the design of a physical absorber which could
achieve the linear force coefficients identified by the flutter suppression and gust load alleviation
studies. Using simple design rules for a hydraulic damper and a fluid-inerter concept it was
found that a hydraulic damper device with a diameter in the range 22 - 40mm could provide the
required equivalent viscous damping coefficient and spring stiffness using a conservative device
stroke estimate of 0.2mm at the flutter frequency of 2.8Hz. Parameter studies of the fluid inerter
concept showed that inertance values up to 700kg are achievable for a helical channel length of
10m and that it is favourable to de-couple of the length of the helical channel from the length of
the pressure vessel in order to maximise the efficiency of the damper. Considering the mass of
the fluid, piston rod and pressure vessel casing the device mass was estimated as approximately
0.45 - 2.5kg; this is a highly optimistic estimate but given the scale of the dimensions considered
the device is unlikely to have a significant mass penalty.
In general, the following device parameter values were identified as providing beneficial
flutter suppression and gust loads alleviation: spring stiffness 105 − 106 N/m or N/rad, viscous
damping coefficient 103 − 105 N/ms-1 or N/rads-1 , inertance 102 − 103 kg. Although the devices
were optimised separately for gust and flutter performance it was found that a damper, TID
168
7.2. CONCLUSIONS
or TID-D device at the strut-root was capable of providing a small amount of GLA (1-6%) as
well as increasing the flutter speed by 1-5% whilst have a negligible mass penalty associated
with the device. These results are a strong indication that passive vibration absorbers should
be considered as part of a strategy for reducing the negative effects of aeroelastic phenomena in
truss-braced wings.
7.2 Conclusions
169
CHAPTER 7. CONCLUSIONS AND FUTURE WORK
9. The modelling and optimisation techniques adopted in this thesis are generic enough
that they could be applied to the design of any generic device layout attached to a finite
element model - with the condition that the device force coefficients are linear and frequency
invariant.
This thesis has demonstrated that improvements in the flutter speed between 1-6% and reductions
in gust loads of approximately 4% are achievable by a vibration suppression device that can
be readily incorporated into a truss-braced wing aircraft. These performance improvements
are significant as no effort has been made to promote the device performance within the BUG-
T model, such as by tailoring the mass or stiffness of the airframe to yield greater relative
motion at the device locations. Therefore it is likely that further benefits could be realised by
incorporating the design of the device within a holistic design process that considers all aspects
of the aircraft performance. The recommendation of this thesis is that further research activities
should take place which can advance the understanding of this technology and enable it to be
included in a viable aircraft design.
170
APPENDIX
A
SUGAR 765-095-R EV. D D ESIGN D ATA
he purpose of this appendix is to provide the design data that is used to conduct the
T aeroelastic analysis presented in this thesis. In order to maintain equivalence with the
sizing and optimisation work carried out as part of the NASA/Boeing Subsonic Ultra
Green Aircraft Research (SUGAR) programme the design data is taken directly from the principal
truss-braced wing report titled "Subsonic Ultra Green Aircraft Research: Phase 2 - Volume 1 -
Truss Braced Wing Design Exploration" 30 for the 765-095-Rev. D aircraft variant.
Much of the data presented in the SUGAR report is in graphical format making it difficult to
access the raw data and use this in an analysis. In order to enable the SUGAR design data to
be used the relevant data must be extracted from these graphs and saved in an appropriate
format. As much of the pre and post-processing for the analysis in this thesis is conducted using
MATLAB it makes sense to use the MATLAB File Exchange programme GRABIT 289 to extract
the data. The workflow for GRABIT is as follows:
1. Import an image into the MATLAB environment.
2. Specify an origin location and the (x, y) axes system by selecting points in the image.
3. Select data points on a pixel-by-pixel basis until the desired dataset has been compiled.
As the data must be selected manually from an image the accuracy of the dataset is dependent on
the resolution of the image and the precision by which each data point is selected. The following
data sets have been extracted using GRABIT:
• Design Speeds 234 - The variation of aircraft speed vs. altitude has been extracted for the
Design Dive Speed, Max Operating Speed and the Operational Speed.
171
APPENDIX A. SUGAR 765-095-REV. D DESIGN DATA
• Beam Stiffness - The beam stiffness data has been extracted for the wing, strut and
jury-strut for the in-plane, out-of-plane and torsional stiffness parameters.
• Beam Axis Location - The location of the beam shear centre in the wing is extracted
from the data in the SUGAR reports.
Whilst every effort has been made to ensure the data has been extracted accurately it is possible
that the extracted data will deviate from the design data used during the SUGAR study.
The mass of the aircraft under different loading conditions must be established in order to define
the correct mass cases during the structural optimisation process. Table A.1 shows a summary
of the principal mass characteristics for the aircraft and Table A.2 provides a breakdown of the
aircraft mass by component part. The MTOW1 , MLW, MZFW and OEW are provided in Table
2.29 of Bradley et al. 30 and the design payload and the design fuel are found in Table 2.37 of the
same report. The calculations used for payload and fuel masses are provided below:
• Design Payload - In Bradley et al. 30 the design payload is calculated assuming that
each passenger has an average mass of 200lbs (including baggage), so for 154 passengers
the design payload is 30800lbs or 13970kg.
• Maximum Payload - The maximum payload is defined as the design payload with an
additional 15200lbs, yielding 20865kg.
• Usable/Design Fuel - The usable fuel is calculated as the mass required to reach MTOW
for an aircraft carrying the design payload, i.e. M f desi gn = MTOW − OEW − M pl desi gn
• Maximum Fuel - The maximum fuel is calculated as the fuel capacity multiplied by the
density of aviation fuel. The SUGAR 765-095-Rev. D fuel capacity is given in Table 2.29 of
Bradley et al. 30 as 5416USG, which equates to 20503m3 , assuming a standard density of
aviation fuel of 804kg/m3 yields a maximum fuel mass of 16484kg.
172
A.3. FLIGHT ENVELOPE
TABLE A.2. SUGAR 765-095-Rev. D group mass statement, taken from Bradley et al. 30 .
Mass
Mass Group
[kg]
Wing 7561.38
Tail 1433.35
Fuselage 7679.31
Wing Strut & MLG support installation 1669.22
Landing Gear 2304.25
Nacelle and Pylon 2190.85
Propulsion - Engine 3819.24
Propulsion - Fuel System 748.43
Flight Controls 1202.02
Power Systems - APU 458.13
Power Systems - Hydraulics 344.73
Power Systems - Electrical 1043.26
Instruments 349.27
Avionics & Autopilot 680.39
Furnishings & Equipment 4136.76
Air Conditioning 653.17
Anti-icing 54.43
Manufacturer’s Empty Weight (MEW) 36328.18
Operational Items 3270.40
Operational Empty Weight (OEW) 39598.58
Usable Fuel 14469.58
Design Payload 13970.63
Takeoff Gross Weight (TOGW) 68038.80
Proper definition of the flight envelope is crucial when designing the aerodynamic and structural
properties of the aircraft as these help define the aeroelastic load cases which will size the aircraft
components. The variation in aircraft design speeds vs. altitude are presented in Fig. A.3 and
Table A.5 in terms of true air speed for a range of altitudes. In addition to these design speeds
there are numerous other metrics which are used to characterise the flight envelope of a given
aircraft design. Many of these parameters are listed in Bradley et al. 30 however those specific
parameters required to define gust load cases (Z MO ) and aeroelastic stability requirements
(M MO ) are provided in Table A.3.
173
APPENDIX A. SUGAR 765-095-REV. D DESIGN DATA
F IGURE A.1. Flight envelope for the SUGAR 765-095-Rev.D aircraft, taken directly
from Bradley et al. 30 .
Figure A.1 shows the SUGAR project design speeds 234 as a function of altitude in units of
calibrated airspeed (CAS)2 :
• Design Cruise Speed (VC ) - The normal operating speed of the aircraft, used for perfor-
mance calculations and calculating cruise operating conditions.
• Maximum Operating Speed (VMO ) - The maximum expected speed of the aircraft.
• Design Dive Speed (VD ) - The highest speed planned during testing. Used for defining
the aeroelastic stability envelope, e.g. determining flutter and buffet margins.
These design speeds are used for all performance calculations, aerodynamic analysis and struc-
tural sizing studies conducted in Bradley et al. 30 and so it is appropriate to use the same data in
this thesis. The data in Fig. A.1 was manually extracted using the process described in Section
A.1 and is shown in Fig. A.2(a). However, the aeroelastic analysis requires the aircraft speed to
2 Calibrated airspeed is the speed of the aircraft as given by the the airspeed indicator with a correction for static
pressure 269 .
174
A.3. FLIGHT ENVELOPE
be defined in true airspeed (TAS) therefore the flight envelope data must be converted from CAS
to TAS using the equation provided in Gracey 269
s
f ρ0
VT AS = VC AS (A.1)
f0 ρ
where VT AS is the true air speed, VC AS is the calibrated airspeed, f is a compressibility factor, ρ
is the air density and the subscript 0 indicates a parameter calculated at sea level or using sea
level properties.
The compressibility factor f is a function of the impact pressure (q c ), a term relating the
compression of the air as it enters the pitot-static tube to the total pressure outside the aircraft,
which for subsonic speeds is given by
"µ γ #
γ − 1 ρ0
¶
γ−1
q c = P0 1+ VC2 AS −1 (A.2)
2γ P 0
where P0 is the air pressure at sea level and γ is the adiabatic index for air. Once the impact
pressure is known the compressibility factor can be calculated using
s
γP ³ qc
· ´ γ−1 ¸
γ
f= +1 −1 (A.3)
γ − 1 qc P
where P is the pressure at a given altitude - calculated using the International Standard Atmo-
sphere (ISA) model 290,291 . Once the compressibility factor has been calculated at the altitudes of
interest the CAS data can be converted to TAS using Eqn. A.1 and the data in Table A.4. The
resulting flight envelope data in TAS is shown in Figure A.2(b).
The final flight envelope data is presented for a uniform altitude distribution in Fig. A.3 and
tabulated in Table A.5. Altitudes in the range [0 : 500 : 50,000]ft are considered and a piecewise
linear interpolation is used to calculate the design speeds based on the data in Fig. A.2(b).
TABLE A.4. Conversion factors and ISA mean sea level conditions 291 .
175
APPENDIX A. SUGAR 765-095-REV. D DESIGN DATA
(a) Flight envelope raw data from GRABIT. (b) Flight envelope data in TAS.
F IGURE A.2. Discrete Dataset for the SUGAR Flight Envelope in CAS and TAS
176
TABLE A.5. Final flight envelope data at altitude intervals of 500ft.
7500 178.50 184.26 213.01 24500 217.51 232.76 254.12 41500 201.73 215.95 235.58
8000 179.87 185.78 214.69 25000 217.16 232.10 253.83 42000 201.32 215.36 235.16
8500 181.34 187.21 216.31 25500 216.51 231.91 253.30 42500 200.78 214.79 234.80
9000 182.73 188.64 217.99 26000 216.29 231.19 252.59 43000 200.44 214.33 234.17
9500 184.18 190.18 219.74 26500 215.72 230.92 252.39 43500 199.58 213.83 233.65
10000 185.57 191.63 221.42 27000 214.99 230.14 251.64 44000 199.15 213.16 232.99
10500 187.00 193.15 223.21 27500 214.62 229.84 251.18 44500 198.62 212.56 232.29
11000 188.52 194.73 224.94 28000 213.95 229.02 250.43 45000 198.02 212.09 231.88
11500 190.02 196.27 226.81 28500 213.55 228.66 249.97 45500 197.60 211.59 231.11
12000 191.60 197.87 228.64 29000 212.96 228.24 249.54 46000 197.21 210.90 230.57
This section provides an overview of the aeroelastic load cases and associated mass cases which
are used in the aeroelastic analysis throughout this thesis. Steady-state manoeuvres, landing
or dynamic gust cases are not considered however they should be included in a future study in
order to correctly size the airframe. Due to the low-fidelity of the structural modelling the mass
of the aircraft is considered in a gross sense only, with the payload and fuel masses for each load
case assigned to fuel tanks and cargo bays as described in Appendix C.
Table A.6 shows the aeroelastic load cases used to size the airframe during the initial multidisci-
plinary optimisation in Bradley et al. 30 . There are seventeen load cases in total, including four
manoeuvre cases, a ‘taxi-bump’ case and twelve gust cases. The gust cases use the Pratt gust
formulation 263 to calculate a revised load factor based on a gust velocity which is a function of
altitude. The load cases in Table A.6 are consistent with initial SUGAR MDO study carried out by
Virginia Tech, however for the refined sizing analysis, which was conducted by Boeing, a different
set of load cases was used which is not available in Bradley et al. 30 . Therefore, it is appropriate
to proceed with the load cases defined in Table A.6 with the caveat that a more refined set of load
cases is required to correctly size the airframe.
TABLE A.6. SUGAR aeroelastic load cases, taken from Bradley et al. 30 Table 2.9.
178
A.4. AEROELASTIC LOAD CASES AND MASS CONFIGURATIONS
Table A.6 shows that there are several combinations of fuel and payload masses across the
different load cases with only two distinct All-Up Mass (AUM) configurations - MTOW and
MZFW. This means that regardless of the fuel or payload fraction the total mass of the aircraft
should equal the AUM that is defined for each load case. Upon examination it is clear that there
is a mismatch between the payload/fuel masses and the AUM for some of the load cases. For
example, in load cases 2 and 4 (LC2 & LC4) the payload mass should be at its maximum and
the fuel mass should be at 50% for an AUM of MZFW, however this cannot be true as the MZFW
is defined as the OEW plus the maximum payload mass. To mitigate for this the mass cases in
Table A.6 are modified to provide the correct AUM for each load case. Specifically, the fuel fraction
for LC2 & LC4 is set to 0% to make the AUM equal to MZFW and the 100% fuel fraction for
the remaining load cases is redefined as the design fuel value from Table A.1, not the maximum
possible fuel as defined by the fuel capacity. Table A.7 shows a summary of these mass cases and
Table A.8 shows a slightly modified version of the SUGAR load cases.
TABLE A.8. Aeroelastic load cases including the modified mass cases.
179
APPENDIX A. SUGAR 765-095-REV. D DESIGN DATA
The stiffness distribution of the full-scale3 SUGAR aeroelastic model is given in Figs. 3.1, 3.2 and
3.3 of Bradley et al. 30 for the wing, strut and jury-strut respectively. As with the flight envelope
data, the beam stiffness data is extracted from the SUGAR report using the GRABIT MATLAB
programme. For each wing component (i.e wing/strut/jury-strut) the in-plane, out-of-plane and
torsional stiffness are provided, however data for the axial stiffness is not given anywhere in
the report. As has been noted by many studies, the axial stiffness has a large influence on the
aeroelastic response of a truss-braced wing because the truss elements generate significant axial
loads for wing locations inboard of the truss attachment points. Therefore, without accurate data
for the beam axial stiffness it is almost impossible to make any meaningful comparisons between
results for the SUGAR 756-095 Rev. D and any results obtained in this thesis. As a workaround,
the modelling assumption adopted by Su 106 is used to estimate the axial stiffness for each wing
component. Here, the axial stiffness is assumed to be one order of magnitude greater than the
in-plane bending stiffness; which is an appropriate assumption given that the cross-sectional
area of a typical aerospace box-beam is usually larger than the geometric moments. The final
extracted stiffness data for the wing, strut and jury-strut is shown in Fig. A.4. Also, the beam
shear centre location has been extracted from the data in Figure 3.2 of Bradley et al. 30 using
GRABIT and the corresponding data is provided in Table A.9.
A detailed mass distribution for the full-scale SUGAR aeroelastic model is provided in Table 3.2
and Table of the SUGAR aeroelastic test report 148 for the wing and truss elements respectively.
The mass distribution includes additional masses that were added to the strut and jury-strut to
allow the finite element model (FEM) to meet the target scaled-mass of the wind tunnel model.
The wing mass distribution was generated from the full-scale detailed finite element model and is
split into 29 discrete masses which are derived from the load-carrying structure and the inboard
and outboard ailerons - a detailed breakdown is given in Table A.10. The strut and jury-strut
3 The term "full-scale" refers to the sized SUGAR aircraft model and not the wind-tunnel model used during
aeroelastic tests.
TABLE A.9. SUGAR 765-095 Rev. D wing shear centre data. The value of x sc is the
distance from the leading edge as a fraction of the chord value at each spanwise η
position.
0.024 0.049 0.096 0.143 0.190 0.239 0.286 0.333 0.384 0.431 0.478
η
0.525 0.576 0.624 0.670 0.717 0.765 0.811 0.858 0.906 0.953 1
0.248 0.273 0.249 0.228 0.209 0.191 0.178 0.167 0.159 0.154 0.151
x sc
0.153 0.158 0.152 0.137 0.125 0.114 0.106 0.104 0.105 0.114 0.131
180
A.6. WING MASS DISTRIBUTION
1010
108
107
106
0 0.2 0.4 0.6 0.8 1
η r [-]
Strut Beam Stiffness [N/m2 ]
107
106
G J - Torsional
EI 22 - Out-of-plane
EI 11 - In-plane
105
F IGURE A.4. SUGAR 765-095-Rev. D beam stiffness distribution for the wing, strut
and jury-strut, taken from Bradley et al. 30 .
181
APPENDIX A. SUGAR 765-095-REV. D DESIGN DATA
components have a significantly less detailed mass distribution as even in the detailed SUGAR
FEM these components were modelled as beam elements, as opposed to the 2D shell elements in
the wing. For the strut, there are 33 point masses of 23.66kg distributed evenly along the length
of the strut, whereas for the jury-strut, there is a point mass of 2.27kg at the root and tip with
four additional masses of 0.68kg along the jury-strut. For both the strut and jury-strut masses
the inertia tensor is a null matrix.
TABLE A.10. SUGAR 765-095-Rev. D detailed wing mass, taken from Bradley et al. 148 .
182
APPENDIX
B
BUG-T N EO CASS M ODEL
T Bristol Ultra Green Truss-Braced Wing (BUG-T) model and discuss how the open-source
aircraft sizing software NeoCASS can be used to generate a model of this configuration.
To gain the most insight from this appendix the reader should be familiar with the contents of
the NASA/Boeing report titled "Subsonic Ultra Green Aircraft Research: Phase 2 - Volume 1 -
Truss Braced Wing Design Exploration" 30 .
This appendix is organised as follows: First, a brief overview of NeoCASS and the geometry
pre-processor AcBuilder is provided to familiarise the reader with the relevance of the software.
Next, the aircraft geometry and mass distribution is defined, with specific emphasis given to the
origin of the reference material used to generate the model. Finally, some of the limitations of the
AcBuilder software are discussed with regards to the modelling of truss-braced wing aircraft. The
output of this appendix is a NeoCASS geometry and mass-distribution file, as well as a Nastran
finite element model which is used to generate the final version of the BUG-T model in Appendix
C.
183
APPENDIX B. BUG-T NEOCASS MODEL
F IGURE B.1. Final ’AcBuilder’ geometry for the BUG-T aircraft model.
The final revision of the 965-095 aircraft model, 965-095-Revision D, is used to define the
AcBuilder geometry as this variant is described in the most detail in the report. Much of the data
for generating the NeoCASS model is contained in the SUGAR reports but in some instances
engineering judgement has been used or dimensions have been estimated from the General
Arrangement (GA) drawing of the 965-095-Revision D model in Figure 2.152 of Bradley et al. 30 .
Also, the parameterisation scheme used in AcBuilder differs from the data presented in Bradley
et al. 30 and so for some quantities a conversion has taken place to define the NeoCASS geometry.
In the following sections the various aspects of the BUG-T NeoCASS model are described and
related to the data presented in the SUGAR report. Note, where tables and figures are referenced
without the prefix ‘B’ they refer to the data found in Bradley et al. 30 and not to tables and figures
in the main body of the thesis, also, for the sake of brevity 765-095-Revision D will be shortened
to 756-095-RD for the remainder of the appendix.
Examining the origin coordinates of the various aircraft components described in Table 2.30 and
the GA drawing in Figure 2.152 it is clear that there is a mismatch between the two datasets.
Primarily, the convention for describing the origin of a component is not explained in Bradley
et al. 30 and instead it is left to the reader to infer exactly what is meant by "origin", meaning it
is possible to define an aircraft that does not match the one shown in the GA.
It is important to establish the correct position of all the major components of the model as
this will have significant effect on any subsequent analysis. Therefore, before the geometry of
the BUG-T can be defined in AcBuilder it is necessary to convert between the parameterisation
scheme used in the SUGAR report and the AcBuilder programme. Corroborating the data in
184
B.2. AIRCRAFT GEOMETRY
Table 2.30 with the GA in Figure 2.152 and the component drawings in Figures 2.155, 2.158
and 2.159 reveals that for lifting surfaces the component origin is closely related to position of
the Mean Aerodynamic Chord (MAC), therefore, to convert between the SUGAR convention and
NeoCASS convention one must perform the following operation,
xSUG AR − x M AC LE
x̄ N eo = (B.1)
L f uselage
where x̄ N eo is the normalised x-location of the leading edge of the root chord, xSUG AR is the
x-location of the MAC in the global coordinate system, x M AC LE is the x-location of the MAC with
respect to the component leading edge and L f uselage is the length of the fuselage. Furthermore,
to obtain the z-location of each lifting surface the following conversion is necessary,
zSUG AR − z f uselage
z̄ N eo = (B.2)
h f uselage
where z N¯ eo is the normalised z-location of the leading edge of the root chord, zSUG AR is the z-
coordinate of the origin of the component, z f uselage is the z-coordinate of the origin of the fuselage
and h f uselage is the height of the fuselage. Note, the standard aircraft body axis coordinate system
is adopted by NeoCASS, which is the x-axis along the fuselage, the y-axis along the starboard
wing and the z-axis in the positive lift direction.
185
APPENDIX B. BUG-T NEOCASS MODEL
aircraft geometry is described in turn and the various parameters are detailed - beginning with
the fuselage, then the lifting surfaces and concluding with the engine and landing gear.
B.2.1 Fuselage
The fuselage length is calculated from the GA drawing in Figure 2.152 and the dimensions of the
cross-section are taken from Figure 2.160. There is not sufficient detail in the SUGAR report to
reproduce the fuselage geometry to a high level of accuracy, therefore, the parameters controlling
the geometry of the nose and tail sections were tweaked until the fuselage "looked right". This
means that the subsequent geometry is not an accurate representation of the original 965-095-RD
model and any comparison between the aerodynamic aspects of the fuselage and the SUGAR
model will be invalid. As this thesis does not consider any high-fidelity aerodynamic analysis this
is considered to be an acceptable compromise.
The model has three sets of lifting surfaces, the wing, horizontal tail plane (HTP) and the vertical
tail plane (VTP). For each lifting surface the classical wing parameterisation scheme based
on the component’s area, span, taper, sweep, etc is used to define the planform shape and the
aerodynamic cross-section is defined by lofting between a series of cross-sections at each kink
point along the lifting surface reference line. For further details the reader is directed to the
documentation found on the NeoCASS website.
The wing geometry is derived from several sources. The gross wing properties such as the
span, dihedral angle and the Aspect Ratio (AR) are taken from Table 2.31 and the corner point
diagram in Figure 2.156 is used to derive the location of the spanwise kinks, leading edge sweep
angles and taper ratios of each wing segment. The location and size of the various control surfaces
are also derived from Figure 2.156 and the coordinates of the wing origin are derived from the
data in Table 2.30 using Equations B.1 and B.2. Also, the Angle-of-Attack (AoA) is estimated
from the GA in Figure 2.156 and a linear variation from root-to-tip has been assumed. A fairing
is included at the wing root but due to the lack of detail in the SUGAR reports the parameters
were estimated by comparing the NeoCASS model with the GA in Figure 2.152. In NeoCASS
the fairing is principally used to define the ’belly-tanks’ and so any discrepancy between the
765-095-RD and the NeoCASS models in terms of the fairing geometry will likely manifest itself
as a slight shift in the fuel centre-of-gravity which should have a negligible effect on the overall
sizing of the aircraft.
The VTP planform is taken from Figure 2.158 and Table 2.32 and the HTP is derived using
Figure 2.159 and Table 2.33. As with the wing, Equations B.1 and B.2 are used to derive the
origin of the HTP and VTP based on the data in Table 2.30 and the AoA is estimated from the GA
in Figure 2.156.
186
B.2. AIRCRAFT GEOMETRY
TABLE B.1. Maximum deflection limits for the BUG-T control surfaces 16,254 .
The aerofoil cross-sections for the lifting surfaces are not provided in the SUGAR reports
and so engineering judgement must be used to assign an appropriate aerofoil. Given that the
cruise Mach number is 0.7 it is likely that a supercritical aerofoil will be required for the wing,
therefore, the FOIL31 aerofoil is selected from NeoCASS’s pre-packaged library of aerofoils. The
HTP and VTP are assumed to have a NACA0012 profile.
The 765-095-RD model has a more-or-less conventional control surface layout. The leading edge
of the wing has a kreuger flap along the entirety of its span with breaks for the engine mount,
outboard kink and wing fold. In terms of trailing edge devices, the 765-095-RD has a series
of single-slotted flaps extending from the fuselage-joint to the wing fold as well as a high and
low-speed aileron and spoilers. A diagram of the wing control surface layout is shown in Figure
2.155 and the HTP and VTP control surfaces are shown in Figures 2.158 and 2.159 respectively.
Trim tabs are not defined for the aileron, elevator or rudder. Table B.1 defines the control surface
deflection limits based on the guidance found in Raymer 254 and Torenbeek 16 .
B.2.4 Engine
The gFan+ ducted fan (DF) is selected for this model as NeoCASS does not allow unducted
fan engine types. The data for the gFan+DF is found in Figure 2.164 and Table 2.34 and the
placement of the engines relative to the wing, as well as the fineness ratio of the nacelle, has been
estimated using the GA in Figure 2.156. It is important to note that the finite element model
generated by AcBuilder and used in NeoCASS only models the engine as a lumped mass. This
means no consideration is given to the design of the engine pylon and its interaction with the
wing and that there will no combined pylon-wing modes present in the structure.
The landing gear track geometry is taken from Figure 2.163 and the origin of both the main and
auxiliary landing gear is estimated from the GA in Figure 2.156. It is especially important to get
the location of the MLG correct as this has a major effect on the take-off and landing performance
187
APPENDIX B. BUG-T NEOCASS MODEL
of the aircraft. Furthermore, the attachment point for the strut is co-located with the MLG pylon
and the orientation and position of the strut will have a significant effect on the overall response
of the aircraft.
In this section the main components of the aircraft mass are defined, including the layout of the
fuel tanks and the overall dimensions the cabin and cargo hold. The output of this section is a
NeoCASS mass distribution file containing a breakdown of the principal masses, including the
coordinates of the centres of gravity.
The SUGAR reports do not provide any information on the fuel tank locations and so they must
be estimated based on the maximum fuel volume defined in Table 2.29 and the available volume
of the wing. The Virginia Tech MDO code, which is the basis for the SUGAR TBW optimisation
studies in Bradley et al. 30 , uses Raymer’s method for estimating the cross-sectional area of a
wing fuel tank 254 ,
S f uel = 0.85 × c f uel × t (B.3)
where S f uel is the cross-sectional area of the fuel tank, c f uel is the chord length of the fuel tank
and t is the maximum thickness of the aerofoil cross-section. Values for these parameters can
be found in Table B.2 of this appendix. Also, note that the fuel tank is assumed to fill the space
between the front and rear spar, the locations of which are defined in Table B.2.
Next, the spanwise location of the fuel tanks must be defined so that the total wing fuel tank
volume can be calculated. Torenbreek recommends that the final metre of the wing should be
free from fuel 16 p. 261 which gives an spanwise extent of 95.8% for the outboard fuel tank based
on the 51.798m span of the 765-095-RD. Finally, using the formula for a trapezoidal fuel tank
in Torenbreek 16 p. 448 and the data in Tables B.2 and B.3 gives a total fuel volume for both
wings of 20.206m3 . This is slightly less than the fuel capacity of the 765-095-RD which has a
fuel volume of 20.502m3 . As the formula in Torenbreek are for preliminary design purposes only,
it was decided to restrict the amount of fuel in the wings to 20m3 and keep the remaining fuel
TABLE B.2. Parameters for calculating the fuel tank volume in the wing.
188
B.3. MASS DISTRIBUTION
TABLE B.3. Dimensions and volume for each fuel tank in the wing.
in the ‘belly’ tank in the wing fairing, this should mitigate for any inaccuracies in the design
formula and ensure that the wings have approximately the right volume for the required fuel
loading with some additional capacity in the fairing tanks.
The cabin layout module in NeoCASS is primarily used for defining the volume of the cabin
and cargo hold so that a sensible estimate can be made of the payload CoG. Figure B.2 shows
the cabin layout and notional cargo hold for the SUGAR-inspired aircraft model where the
main cabin dimensions and passenger layout are taken from Table 2.29 and Figure 2.160 of
the SUGAR report. The baggage volume has been estimated using Torenbreek’s formula for a
‘volume-limited-payload’ 16 p. 79,
where M pa yload is the maximum payload mass, ρ pa yload is the payload density, Vpa yload is the
cargo volume, m passenger is the mass per passenger and N passenger is the number of passengers.
The data in Table 2.37 gives a design payload value of 30,800lb (13971kg) which, using the values
in Torenbreek for the payload density and passenger mass, gives a payload volume of 12.140m3
for the 154 passenger capacity of the 765-095-RD.
Passenger aircraft typically have a fore and aft cargo hold in order to balance the payload
mass and enable a faster turnaround, however, NeoCASS only allows for a single cargo hold to be
defined. As the principal purpose of the fore and aft cargo holds is to ensure that moment arm of
the payload CoG is minimised it was decided to position the single NeoCASS cargo hold close to
the wing origin to mimic the approximate CoG of a two-hold layout.
Table A.2 shows a copy of the ‘Group Weight Statement’ data in Table 2.37 of the SUGAR report.
This data is used as the starting point for defining the mass distribution of the BUG-T model.
Unfortunately, no information is provided about the location of the masses within the SUGAR 765-
095-Rev. D, and so it is not possible to generate a mass distribution using this data alone. Instead,
the Weights and Balance module of AcBuilder is used to compute the CG (Centre of Gravity)
189
APPENDIX B. BUG-T NEOCASS MODEL
F IGURE B.2. Cabin and cargo hold layout for the BUG-T NeoCASS model.
locations for each mass in the aircraft, however as the mass breakdown used in AcBuilder differs
from the SUGAR 765-095-Rev. D it is not possible to map the data in Table A.2 directly to the
NeoCASS mass data. Therefore, the following assumptions have been made:
1. The "Tail" mass is divided equally between the HTP and the VTP.
2. The "Wing Strut & MLG support installation" mass is split equally between the wing and
the main landing gear (MLG).
3. The "Landing Gear" mass is split between the MLG and auxilliary landing gear (ALG)
using a 3:1 ratio.
4. The "Nacelle and Pylon" and "Propulsion - Engine" are combined to give the "Powerplant 1
with nacelle and pylon" mass.
5. The "Pilots" mass accounts for two pilots and five crew assuming the average person weighs
80kg.
6. The "Passengers" mass accounts for 154 passengers with an average mass of 80kg per
passenger.
7. The "Baggage & Cargo" mass is equal to the "Design Payload" minus the passenger mass.
8. The "Fuel - wing" and "Interior" mass are equal to the "Usable Fuel" and "Furnishings &
Equipment" mass groups respectively.
9. The "Total systems or miscellaneous" mass is a combination of the following mass groups:
• Propulsion - Fuel System
• Flight Controls
• Power Systems - APU
• Power Systems - Hydraulics
• Power Systems - Electrical
190
B.4. LIMITATIONS OF THE NEOCASS GEOMETRY PRE-PROCESSER ACBUILDER
• Instruments
• Avionics & Autopilot
• Air Conditioning
• Anti-Icing
• Operational Items less the weight of the crew (560kg)
Figure B.3 shows the location of the different mass groups and the centre of gravity at Maximum
Empty Weight (MEW) and Maximum Take-off Weight (MTOW) respectively and the data is
tabulated in Table B.4.
The following items have been identified as limitations of the AcBuilder module:
• Incompatible with braced/joined-wings - The parameterisation scheme used by AcBuilder
does not allow a truss structure to be attached to an existing wing structure. This is seen
as a major limitation as far as this thesis is concerned because the interaction between the
truss and wing structure has a significant impact on the static and dynamic response of the
aircraft. Furthermore, this precludes the design and modelling of joined-wing configurations,
F IGURE B.3. Locations of component masses and CG location for MTOW and MEW
for the BUG-T NeoCASS model
191
APPENDIX B. BUG-T NEOCASS MODEL
which have been the subject of many design studies in recent years 113,296–299 .
• Limited control surface layout - The control surface layout for the wing is restricted
to a single leading/trailing edge control surface and for the HTP and VTP components only
a single trailing edge flap is allowed. This prevents the user from modelling realistic control
surface layouts, such as the one shown in the SUGAR corner point diagram in Figure 2.156
• No engine-pylon modelling - The finite element model generated by AcBuilder for use
by NeoCASS only considers the engine mass and does not account for the pylon which
connects the engine to the wing. This has serious implications for the dynamics of the wing
as the transfer of energy between the engine and the wing is strongly dependent on the
properties of the engine pylon.
For these reasons the sizing and modelling capabilities within NeoCASS are not appropriate
for analysing a model derived from the SUGAR 765-095-Rev. D aircraft. Instead, an alternative
approach must be adopted which allows the modelling of truss-braced wings using a more
‘physics-based’ formulation.
192
PPENDIX
C
A
BUG-T F INITE E LEMENT M ODEL D EFINITION
he purpose of this appendix is detail the final versions of the SUGAR Volt-inspired,
T Bristol Ultra Green Truss-Braced Wing (BUG-T) model as well as introduce O2 MeGA, an
object-based parametric aircraft model generator that was developed as part of this thesis.
The output of this appendix is a Nastran finite element model used for all analysis in this thesis.
This appendix is organised as follows: First, a brief overview of the BUG-T model variants
is provided, with reference to the previous work carried out during the Agile Wing Integration
project. Next, a brief overview of the O2 MeGA software is given, with specific focus given
to its object-orientated code structure and generic aircraft parameterisation schemes. Lastly,
O2 MeGA is used to generate the final geometry of the BUG-T model ready for the dynamic and
aeroelastic analysis carried out in Chapters 3, 4 and 5.
The BUG-T model is a truss-braced wing (TBW) version of the Bristol Ultra Green (BUG)
model 300 , which itself is based on the wing geometry of the NASA/Boeing SUGAR High aircraft.
The BUG model was generated to act as a test structure for various nonlinear aeroelastic
simulations and 3D GFEM (Global Finite Element Model) optimisation studies that took place as
part of the Agile Wing Integration (AWI) project1 at the University of Bristol. Figure C.1 shows
the planform of the BUG model and the cross-section of the wing GFEM. The planform geometry
matches the planform of the SUGAR Volt aircraft however the stiffness and mass distribution
has been tailored to promote nonlinear behaviour during aeroelastic simulations. Furthermore,
the truss-structure is not included in the model as the focus of the AWI project was the simulation
193
APPENDIX C. BUG-T FINITE ELEMENT MODEL DEFINITION
F IGURE C.1. BUG model planform and GFEM cross-section, taken directly from
Stodieck et al. 300
of fixed-wing aircraft with cantilever wings. For these reasons it is not desirable to use the BUG
model for the analysis carried out in this thesis and so an alternative model must be created. The
BUG-T model is a full-aircraft model based on the SUGAR Volt 765-095-Revision D aircraft from
Bradley et al. 30 . The truss-structure is included however and stiffness and mass distribution of
the wing and truss elements is defined using the SUGAR data presented in Appendix A.
The unconventional configuration of the truss-braced concept means that many existing aircraft
parameterisation schemes and sizing tools are not appropriate for modelling these structures.
Indeed, a number of recent publications have focussed on developing "physics-based" sizing
methodologies in order to specifically address the problems posed by a braced wing configura-
tion 61,151,165 . In this thesis, the author has developed their own parameterisation and geometry
modelling tool, O2 MeGA , which is loosely based on the CPACS2 data model 68 . This tool does
not seek to replace or even replicate the capabilities of CPACS, instead it draws inspiration
from the CPACS parameterisation scheme and applies this to the modelling a truss-braced wing
configuration. A brief explanation is provided for each aspect of the tool:
1. Object-Orientated - The codebase is written using the MATLAB implementation of a
class-based, object-orientated (OO) programming paradigm. This approach allows classes
to be defined with properties and methods specific to that class or subclass. In this instance,
adopting an OO approach means it is possible to very quickly define a model as an expand-
able collection of objects, with some user-specified hierarchy which allows co-dependencies
to be established between them. Such an approach is ideally suited to the design and
2 Common Parametric Aircraft Configuration Schema
194
C.2. OBJECT-ORIENTATED MODEL GENERATION FOR AIRCRAFT (O 2 MEGA)
modelling of unconventional aircraft as there is no limit to the number of objects that can
be added to the model. Further details are provided in Section C.2.1.
2. Model Generation - The aim of the tool is to enable the generation of unconventional
aircraft geometries and then convert these geometries into a model which is suitable for
analysis. Low-fidelity, beam-stick models are considered but given the availability of the
3D aircraft geometry there is no reason this could not be extended to include automatic
generation of global finite element models (GFEMs).
3. Aircraft - The codebase is specific to modelling fixed-wing aircraft geometries although
given the generic definition of the model geometry it would be possible to create geometries
of other slender structures, such as missiles or submarines, with little effort.
At this point it is worth mentioning that the concept behind O2 MeGA is not novel and that there
are several other research groups which have defined their own tools for modelling unconventional
aircraft geometries, notable efforts include: GeoMACH 70 , PyGFEM 71 and the CPACS schema 68 .
The object design in O2 MeGA relies on class inheritance to develop a minimum viable represen-
tation of the various elements required to build an aircraft model. Figure C.2 shows the basic
hierarchy of the objects and some of the principal properties.
The base-class is the Entity class which handles the basic parent/child relationship and some
background functionality such as naming the objects and managing the model hierachy. The
Entity object inherits much of its functionality from a separate package that was provided by a
Mathworks consultant as part of a wider modelling and simulation effort during the AWI project.
This package provides the functionality for a generic Model-View-Controller (MVC) software
architecture and has no specific methods for modelling aircraft; it simply handles the background
functions which enable the O2 MeGA framework.
All geometry definitions are handled by the Component class or one of its-subclasses. These
classes handle the relative positioning of components with respect to their parent objects and also
the orientation of these components with respect to the global coordinate system. For example,
Fig. C.3 shows the starboard truss components of the BUG-T model, with the local coordinate
systems shown at the root and tip of each component. When the model is converted to a finite
element model these coordinate systems become the local beam coordinate system. Detailed mass
configurations can be defined using Point Mass or Volume objects (e.g. Compartment, Fuel Tank,
Cargo Hold, Cabin). Volume objects can be defined for any beam object which has a cross-section
associated with it. The 3D volume of these beams is generated by lofting between the various
cross-sections and this volume can then be subdivided into any number of compartments to
form cargo holds, fuel tanks or cabins. Each Volume class has a method which defines its mass
distribution, thus allowing mass configurations to be defined with very overhead from the user.
A generic aircraft model, such as the one shown in Fig. C.4, can be constructed by creating
195
APPENDIX C. BUG-T FINITE ELEMENT MODEL DEFINITION
Entity
Abstract Object
Properties
Name Beam Property
Parent Property Object
Child
Properties
Beam handle
Component
Point Object Orientation
Property Object
Properties
3D position (×1) Properties
Stick Compartment
Line Object Total mass/inertia Rotation matrix
Volume Object
Centre of Gravity Euler angles
Properties Properties Rotation order
3D positions (× N) Quaternions
Geometry
Vectors
Payload Fraction
Point Mass
Payload Density
Point Object Material
Connector Property Object
Properties Fuel Tank
Line Object Volume Object
Mass Properties
Properties Inertia tensor Properties Material type
Root/tip joints Elastic moduli
Fuel Density
Density
Poisson’s Ratio
Control Surface Cargo Hold Failure criteria
Beam Point Object Volume Object
Line Object Cross-Section
Properties Properties Property Object
Properties 2D coordinates Cargo Density
Hinge line Properties
Stiffness
- Shear centre Max deflection Cabin 2D coordinates
- Neutral Axis Max deflection Volume Object Orientation
Inertia (variable) rate Cross-section
- Centre of Mass Properties name
Orientation(s) No. passengers
Material(s) Seat layout
Cross-Section(s) Seat dimensions
Compartment(s) Passenger mass
Baggage mass
Properties Properties
Chord (× N c ) Cant angle
AoA (× Nα ) Pitch angle
Dihedral (× Nγ )
Sweep (× NΛ )
196
C.3. BUG-T GEOMETRY MODEL
Z
Y
X
Z
Y
Z
X
Y
Y
a custom hierarchy of any of the objects in the O2 MeGA package. There is no limit on the
number of components in the hierarchy and only a limited set of rules regarding the parent/child
relationship of the different objects3 , allowing a wide-variety of aircraft geometries to be defined.
Once a valid geometry model has been generated it is possible to automatically create Nastran
finite element models which are suitable for aeroelastic analysis and design optimisation.
Figure C.4 shows the O2 MeGA representation of the BUG-T geometry. This model has been
generated using the data from the SUGAR reports as discussed in Appendix A and B.In the
following sections additional information is provided for each major component in the model,
including the fuselage, wing, strut, jury-strut and empennage sub-assemblies.
C.3.1 Fuselage
The fuselage is modelled as a Beam object aligned with the global X-axis. The cross-section is
derived from the data in Figure 2.170 of Bradley et al. 30 and is modelled as an ellipse with major
and minor axis lengths of 4.23m and 3.78m respectively. The ellipse geoemtry has been tapered
3 Examples of these rules include: "a Control Surface object can only be added to a Lifting Surface object" or
197
APPENDIX C. BUG-T FINITE ELEMENT MODEL DEFINITION
X Y X
ZZ
Z
X YY Z
X
Y
X
Y
Z X
Y Z
X
Z X
Y
Y
Z
X Y X
Y
Z
X
Z
ZZ
Y Z
Z X Y Y
X
Y X X
X
Y
Z Z
X Y Y Z
X
Z
Z X
Y Z
X Y
Y
X
Y
Z
X
Y
X
Y
at the front and rear of the fuselage, however this is purely for cosmetic reasons and has no
influence on the generation of the finite element model and subsequent aeroelastic analysis. The
passenger configuration is modelled using the Cabin object which subclasses the Compartment
baseclass. The Cargo Hold object is used to model the two payload compartments which are
located beneath the cabin, the dimensions of which have been approximated using the data for a
typical Boeing 737 aircraft.
C.3.2 Wing
The port and starboard wing are modelled using a Lifting Surface object and the wing geometry
is based on the general arrangement (GA) drawing of the 765-095-Rev. D and is the same data
used to generate the NeoCASS wing geometry in Appendix B. The control surface configuration
is taken from the corner point diagram in Fig. 2.156 of Bradley et al. 30 . Table C.1 provides the
key geometry information for the port and starboard wing.
C.3.3 Strut
The port and starboard strut is modelled using a Lifting Surface object and the geometry
information is taken from the general arrangement (GA) drawing of the 765-095-Rev. D as well
as the corner point diagram in Fig. 2.157 of Bradley et al. 30 . Table C.2 provides the key geometry
information for the port and starboard strut.
198
C.3. BUG-T GEOMETRY MODEL
C.3.4 Jury-Strut
The port and starboard jury-strut is modelled using a Lifting Surface object and the geometry
information is taken from the general arrangement (GA) drawing of the 765-095-Rev. D. Table
C.3 provides the key geometry information for the port and starboard jury-strut.
The vertical tail plane is modelled using a Lifting Surface object and the geometry information
is taken from the general arrangement (GA) drawing of the 765-095-Rev. D and the corner point
diagram in Fig. 2.158 of Bradley et al. 30 . Table C.4 provides the key geometry information for
the vertical stabilizer.
199
APPENDIX C. BUG-T FINITE ELEMENT MODEL DEFINITION
The port and starboard horizontal tail plane is modelled using a Lifting Surface object and the
geometry information is taken from the general arrangement (GA) drawing of the 765-095-Rev.
D and the corner point diagram in Fig. 2.159 of Bradley et al. 30 . Table C.5 provides the key
geometry information for the vertical stabilizer.
The BUG-T finite element model (FEM) is a ‘beam-stick’ representation of the SUGAR 765-095-
Rev. D aircraft suitable for use with the commercial finite element package Nastran. Nastran is
chosen given its widespread use throughout the aerospace industry and its industry-standard
capabilities in aircraft level aeroelastic analysis. A beam-stick representation is appropriate as
it is typical of aircraft models used for dynamic aeroelastic analysis, also, given the conceptual
nature of the research presented in this thesis it is preferable to keep the model as simple as
possible. The FEM was generated in three stages:
1. The NeoCASS AcBuilder program was used to generate a preliminary FEM that provided
the gross mass distribution and origin locations of the various aircraft components. This
process is described in Appendix B.
2. The O2 MeGA tool was used to model the aircraft geometry using the data from Bradley
200
C.4. BUG-T AEROELASTIC FINITE ELEMENT MODEL
et al. 30 . The component root nodes in the NeoCASS FEM were used to position the different
aircraft components and the mass distribution from Table B.4 was used to generate a series
of point mass objects which were associated with the relevant components.
3. The O2 MeGA geometry was converted to a valid finite element model using an automated
object-specific meshing scheme contained within the O2 MeGA tool. The the detailed mass
and stiffness data for the BUG-T model are provided in Appendix A. The beam element
density was determined following a mesh convergence study shown in Fig. C.5 and the
aerodynamic mesh was defined to capture reduced aerodynamic frequencies up to 50Hz.
The beam element lengths for the wing, strut and jury-strut components is given in Chapter
3.
Figure C.6 shows the Nastran finite element representation of the BUG-T model and Fig. C.7
shows the normalised and un-scaled mass distribution. Note that the differences in detail between
the wing/truss mass distribution and the mass data for the rest of the aircraft reflects the data
provided by Bradley et al. 30 .
F IGURE C.5. Mesh convergence study for the BUG-T model. Subfigure C.5(a) shows
the variation in natural frequency as a function of beam element length for all
normal modes up to 50Hz, with the right-hand y-axis showing the total number
of degrees of freedom in the model. Subfigure C.5(b) shows the number of beam
elements in the wing, strut and jury-strut components as a function of beam
element length.
201
APPENDIX C. BUG-T FINITE ELEMENT MODEL DEFINITION
202
APPENDIX
D
BUG-T R EFERENCE D ATA
he purpose of this appendix is to provide additional results for the dynamic and aeroelastic
T analysis of the BUG-T model detailed in Chapter 3. In the interest of brevity these results
were not shown in Chapter 3 however they are included here in order to provide a complete
description of the results obtained in the initial dynamic and aeroelastic analysis. This appendix
is organised in two sections: In Section D.1 the BUG-T normal modes and associated energy
contributions for flexible modes 1 - 24 are shown in Figs. D.1 - D.6. Then in Section D.2 the static
aeroelastic loads envelopes are shown for the wing, strut and jury-strut for all six beam loads.
The BUG-T normal modes were calculated using the standard Lanczos method 259 . Here, the
first 24 flexible modes are presented and each mode is mass-normalised, i.e. each mode has a
modal mass of unity, which is the standard approach for a normal modes analysis in Nastran.
The modes are numbered in order of increasing natural frequency, meaning that the first flexible
mode (i.e. non-rigid body mode) of the structure is numbered as mode seven and not mode one.
203
APPENDIX D. BUG-T REFERENCE DATA
(a) Mode 7 - 1.498Hz (b) Mode 8 - 1.858Hz
204
In this section the static aeroelastic loads envelopes are presented for all six beam loads for the
wing, strut and jury-strut components. Note that each load cases was calculated for an aircraft
AUM of 68038kg (the MTOW of the SUGAR 765-095 Rev. D) and the aircraft angle of attack
and elevator control surface deflection were used to trim the vertical acceleration and pitching
moment. The loads envelopes are calculated by taking the maximum and minimum values of
the beam loads across each component and the plots are colour-coded to denote which load case
has the maximum/minimum load at that point. Note that where two different beam elements
are attached to the same finite element node the average value of the two sets of beam loads has
been used, which is consistent with the presentation in Bradley et al. 30 . For example, the wing
axial force distribution in Fig. D.8(a) should show an instantaneous jump in force at the strut
attachment point due to the additional loads transferred from the strut to the wing. However as
the average load is used the variation in loads at the attachment point is gradual instead of a
step change. Furthermore, the x-axis for each plot is normalised beam axis of the component (η r ),
which is the straight line distance along the line of nodes which the finite element beams are
attached to. This allows consistent comparison of components which have different orientations
with respect to the global coordinate system.
210
D.2. BUG-T STATIC AEROELASTIC LOADS ENVELOPE
·105
Root Engine Jury-Strut Strut
1
[Nm]
−1
Mx
−2
·106
1
Root Engine Jury-Strut
[Nm]
0. 5 Strut
My
·106
[Nm]
Root Engine
Mz
−2 Jury-Strut Strut
211
APPENDIX D. BUG-T REFERENCE DATA
·106
0
Fx
−1
·106
1
Root Engine Jury-Strut Strut
[N]
0.5
Fy
·105
2
0
[N]
Fz
−2
Root Engine Jury-Strut Strut
−4
0 0. 1 0.2 0.3 0. 4 0.5 0. 6 0.7 0. 8 0.9 1
η r [-]
Envelope 2.5g trim @ cruise (MTOW)
-1g trim @ cruise (MTOW) Pratt Gust @ sea level, M 0.2 (MTOW)
Pratt Gust @ sea level, M 0.4 (MTOW)
212
D.2. BUG-T STATIC AEROELASTIC LOADS ENVELOPE
·104
1
[Nm]
0
Mx
−1
Jury-Strut
−2
0 0.1 0.2 0.3 0.4 0. 5 0.6 0. 7 0.8 0. 9 1
η r [-]
Envelope 2.5g trim @ cruise (MTOW)
-1g trim @ cruise (MTOW) Pratt Gust @ sea level, M 0.2 (MTOW)
Pratt Gust @ sea level, M 0.4 (MTOW) Pratt Gust @ 40K ft, M 0.7 (MTOW)
·104
2
[Nm]
0
My
−2 Jury-Strut
·105
0
[Nm]
Mz
−0.5 Jury-Strut
−1
0 0. 1 0.2 0. 3 0.4 0. 5 0.6 0.7 0. 8 0.9 1
η r [-]
Envelope 2.5g trim @ cruise (MTOW)
-1g trim @ cruise (MTOW) Pratt Gust @ sea level, M 0.2 (MTOW)
Pratt Gust @ sea level, M 0.4 (MTOW) Pratt Gust @ 40K ft, M 0.7 (MTOW)
213
APPENDIX D. BUG-T REFERENCE DATA
·106
2
1 Jury-Strut
[N]
Fx
−1
0 0. 1 0.2 0. 3 0.4 0.5 0. 6 0.7 0. 8 0.9 1
η r [-]
Envelope -1g trim @ cruise (MTOW)
Pratt Gust @ sea level, M 0.4 (MTOW)
·104
0. 5
0
[N]
−0.5 Jury-Strut
Fy
−1
·104
2
[N]
0
Fz
−2 Jury-Strut
0 0. 1 0.2 0. 3 0.4 0.5 0. 6 0.7 0. 8 0.9 1
η r [-]
Envelope 2.5g trim @ cruise (MTOW)
-1g trim @ cruise (MTOW) Pratt Gust @ sea level, M 0.2 (MTOW)
Pratt Gust @ sea level, M 0.4 (MTOW)
214
D.2. BUG-T STATIC AEROELASTIC LOADS ENVELOPE
20
[Nm]
Mx
4
[Nm]
2
My
−2
0 0. 1 0.2 0. 3 0.4 0.5 0. 6 0.7 0. 8 0.9 1
η r [-]
Envelope 2.5g trim @ cruise (MTOW)
-1g trim @ cruise (MTOW) Pratt Gust @ sea level, M 0.4 (MTOW)
·104
2
[Nm]
1
Mz
−1
0 0.1 0. 2 0.3 0.4 0. 5 0.6 0. 7 0.8 0. 9 1
η r [-]
Envelope 2.5g trim @ cruise (MTOW)
-1g trim @ cruise (MTOW) Pratt Gust @ sea level, M 0.2 (MTOW)
Pratt Gust @ 40K ft, M 0.7 (MTOW)
215
APPENDIX D. BUG-T REFERENCE DATA
·104
2
0
[N]
−2
Fx
−4
·104
0. 5
0
[N]
−0.5
Fy
−1
10
[N]
0
Fz
−10
216
APPENDIX
E
M ODELLING A G ENERIC V IBRATION S UPPRESSION D EVICE IN
N ASTRAN
This appendix provides an overview of the various method available for modelling vibration
suppression devices in Nastran. This should serve as a useful resource for any researcher
interested in adopting the same simulation strategy that was used in this thesis. This appendix
is arranged in two parts: In Section E.1 the merits of using a commercial software package to
simulate a vibration suppression device are discussed and in Section E.2 the three methods for
modelling a generic vibration suppression device in Nastran are detailed and their pros and cons
are discussed.
Prior to the work by this author there existed very little research concerning the simulation
of inerter-based devices using the Nastran software environment, or indeed any other commer-
cial finite element analysis package. Instead, previous researchers have chosen to derive the
equations of motion from first principles before running the appropriate analysis in the fre-
quency/time/Laplace domain to obtain the system response, with the simulation typically carried
out using the MatLab/Simulink software due to its prevalence amongst academic institutions.
Such an approach is advantageous because the formulation and solution of the system can be
tightly controlled, however difficulties arise as the complexity of problem increases. For example,
a vibration absorber may have to be applied to a different primary system but continue to use
the same algorithm for designing the device, such as switching from vehicle suspension design
to shimmy suppression in aircraft landing gear using a network synthesis algorithm. Using
the previous approach will require a bespoke set of EoM to be derived for the new problem and
217
APPENDIX E. MODELLING A GENERIC VIBRATION SUPPRESSION DEVICE IN NASTRAN
requires the researcher to develop a deep understanding of the new system before any analysis
can begin. However, utilising a commercial analysis tool such as Nastran means any overhead
from switching design task is associated with generating a new analysis model, which given the
prevalence of CAD tools with built-in meshing capabilities (e.g. CATIA, AUTODESK etc.) is not
always a huge undertaking. If the system of interest contains a large number of degrees of free-
dom then deriving the EoM becomes a lengthy process which can easily introduce errors. Using a
commercial simulation package helps reduce the likelihood of the these errors as the formulation
of the system matrices (mass/damping/stiffness) is done in a systematic fashion and relies only
on the correctness of the analysis model. A further advantage of using a commercial simulation
package is the vast catalogue of element types and analyses which are available. This enables
vibration suppression devices to be included in the analysis of a wide-variety of large and complex
engineering structures. Clearly, there will be applications which favour one approach over the
other, however, in general the commercial software approach favours large-scale problems typical
of industry-standard analysis models, whereas the first principles approach is better-suited to
simple models which are primarily used for research purposes.
There are three methods for modelling a generic vibration suppression device in Nastran:
1. Direct Matrix Input (DMI) - The DMI approach allows additional terms to be added
directly to the mass/damping/stiffness matrices. This approach allows the user complete
control over the device topology but the implementation can become complex when large
models are considered. This method is the most generic but also the most counter-intuitive
if the user is unfamiliar with the Nastran notation.
2. Scalar Elements - As described in Chapter 4, any generic vibration suppression device
can be modelled by joining together the necessary scalar elements (CELAS, CDAMP, CMASS) to
the required degrees of freedom. This is by far the simplest method as it is essentially the
same process as developing any finite element model.
3. Transfer Functions (TF) - Using the TF element a second order transfer function can be
defined which creates set of direct input matrices, M2PP, B2PP, K2PP. This method provides
a user-friendly method for generating direct input matrices, however these matrices are
restricted to the p-set instead of the g-set which are used in the DMI method (see below).
Modelling a device using the Nastran transfer function approach has a major flaw. When a
device is modelled by the TF element and has an intermediate DOF there will be no new natural
frequency in the system due to the new DOF introduced by the device. This is because the
stiffness and mass terms associated with the K2PP and M2PP matrices are not included in the
analysis set which is used for the eigenanalysis. All M2PP, B2PP and K2PP terms are included
after the eigenanalysis has been completed and before the requested analysis begins. Therefore,
218
E.2. MODELLING METHODS
the effects of the device will be present in the chosen analysis1 but the device mode will not be
included. This is particularly concerning when designing a device for aeroelastic control as the
gust and flutter solution sequences both use the modal form of the system matrices. Therefore,
the TF approach is not recommended for any device which has an intermediate DOF.
The Nastran TF element uses the following equation to define a generic second-order transfer
function between a single dependent DOF u d and an arbitrary number of independent DOFs u i
(B0 + B1 p + B2 p2 )u d + (A 0i + A 1i p + A 2i p2 )u i = 0,
X
(E.1)
i
where p is the differential operator d/dt, and the A and B terms are coefficients which, together
with p, define the zeros and poles of the transfer function. This capability can also be used to
define more complex device layouts by considering the device as a series of transfer functions
chained together, an approach that is made necessary by the fact that the TF element is limited
to a second order function. Furthermore, the inputs to the device, such as relative displacement,
velocity and acceleration, are themselves defined as Extra Points (EPOINT) or Scalar Points
(SPOINT) using Equation E.1. Further information can be found in the Nastran Dynamic Analysis
User’s Guide 259 . To demonstrate the use of this element the transfer function for the inerter
and the Tuned-Inerter-Damper (TID) will be derived and related to the terms in Equation E.1.
Beginning with the inerter element, the force generated by a linear inerter at each terminal of
the device is given by
f 1 = f 2 = b( ẍ2 − ẍ1 ) = b × ë, (E.2)
where f i is the force at each terminal, b is the inertance, ẍ i is the acceleration at each terminal
and e is the EPOINT representing the relative displacement between terminals. To implement
this relationship in Nastran the following steps must be taken:
1. Define an EPOINT.
2. Define the device input referencing the EPOINT as the dependent DOF.
3. Define the force at each terminal. Note, when defining a force or moment acting on a
dependent degree of freedom the B terms of the transfer function must be zero.
It is important to note that a transfer function is required to define the force at each terminal of
the device, that is to say, a device with only two terminals requires three transfer functions, one
for each terminal of the device and one for the EPOINT.2 Considering the TID, the force at the two
terminals of the device is
f 1 = b( ÿ − ẍ1 ), (E.3)
1 As long as the analysis is not a normal or complex modes analysis.
2 It is possible to remove the EPOINT from the transfer function definition and instead use two independent
degrees of freedom for each TF element. However, defining the input to the device as an EPOINT allows the relative
displacements, velocities and accelerations to be requested as part of the results which is useful for understanding the
behaviour of the vibration suppression device.
219
APPENDIX E. MODELLING A GENERIC VIBRATION SUPPRESSION DEVICE IN NASTRAN
The equation of motion for the intermediate y-DOF can be found by setting the two device forces
equal to one another. Converting to the Laplace domain yields
here, upper case letters denote the Laplace response and p is the Laplace variable in keeping
with MSC.Nastran convention for the TF element. Remaining in the Laplace domain, the forces
at each terminal of the device are given as
The implementation of the TID element in Nastran is slightly different due to the presence of the
intermediate DOF:
1. Define a SPOINT and two EPOINT.
2. Define the intermediate DOF as a SPOINT using Equation E.1.
3. Using the EPOINT to define the relative quantities for the forces at the device terminals.
4. Define the force at each terminal.
Table E.1 shows the transfer function coefficients for the inerter and tuned-inerter-damper as
well as the various input quantities. Note that e and s represent EPOINT and SPOINT respectively
and subscript 1 and 2 denote the terminals of the device.
TABLE E.1. Transfer function coefficients for the inerter, Tuned-Inerter-Damper and
the device input/output quantities.
Label Equation B0 B1 B2 ud A0 A1 A2 ui
Relative quantity 1 0 0 e1 1 0 0 x1
e 1 = x2 − x1
between x1 and x2 -1 0 0 x2
Inerter
Force on terminal 1 f 1 = b p2 × e 1 0 0 0 f1 0 0 −b e1
Force on terminal 2 f 2 = b p2 × e 1 0 0 0 f2 0 0 −b e1
b p2 x1 +( c p+ k) x2 k c b s 0 0 −b x1
Intermediate DOF s= b p2 + c p+ k −k −c 0 x2
Relative quantity 1 0 0 e1 1 0 0 x1
e 1 = s − x1
between s and x1 -1 0 0 s
TID
Relative quantity 1 0 0 e2 1 0 0 s
e 2 = x2 − s
between x2 and s -1 0 0 x2
Force on terminal 1 f 1 = b p2 × e 1 0 0 0 f1 0 0 −b e1
Force on terminal 2 f 2 = ( c p + k) × e 2 0 0 0 f2 −k −c 0 e2
220
E.2. MODELLING METHODS
Below are a few key considerations when modelling a vibration absorber in Nastran:
• Sensitivity information for any terms related to the DMI and TF methods cannot be
generated when performing a SOL 200 analysis. Therefore, any device modelled using the
DMI or TF methods cannot be optimised using the optimisation methods in this thesis. This
is not a problem when the vibration suppression device is modelled using scalar elements.
• The TF and scalar element method can be used to model spring and damper elements with
nonlinear force-displacement/velocity relationships, although this is only possible in the
relevant solution sequences3 .
• Frequency dependent force coefficients can be readily modelled by commercial finite element
software, however in the case of Nastran this is limited to frequency response analysis and
it not available for gust analysis, flutter or design optimisation.
• Many of the standard Nastran solution sequencies cannot handle large rotations and so
the relative motion across the device terminals is not accurately captured. Therefore it may
be necessary to use the nonlinear solution sequence (SOL 400) or integrate the analysis
with a multi-body simulation tool.
3 See the Nastran Quick Reference Guide entries for TF, PELAST and PDAMPT.
221
B IBLIOGRAPHY
[1] International Air Transport Association, “IATA Climate Change Policy,” https://www.iata.org/policy/
environment/Pages/climate-change.aspx, ????
[Online; accessed 14-January-2019].
[3] Kallas, S., Geoghegan-Quinn, M., Darecki, M., Edelstenne, C., Enders, T., Fernandez, E., and Hartman, P.,
“Flightpath 2050 Europe’s Vision for Aviation,” Tech. rep., 2011.
[4] Collier, F., Zavala, E., and Huff, D., “Subsonic Fixed Wing Project: Reference Document,” National Aeronautics
and Space Administration (NASA), http://www.aeronautics.nasa.gov/nra_pdf/sfw_proposal_c1.pdf , 2008.
[5] Bradley, M. K., and Droney, C. K., “Subsonic Ultra Green Aircraft Research: Phase I Final Report,” Tech. rep.,
NASA, CR-2011-216847, 2011.
URL https://ntrs.nasa.gov/search.jsp?R=20110011321.
[7] Bowman, C. L., “Visions of the Future: Hybrid Electric Aircraft Propulsion,” Tech. rep., 2016.
[8] Frota, J., “NACRE Novel Aircraft Concepts,” The Aeronautical Journal, Vol. 114, No. 1156, 2010, pp. 399–404.
[11] Lee, J. J., Lukachko, S. P., Waitz, I. A., and Schafer, A., “Historical and future trends in aircraft performance,
cost, and emissions,” Annual Review of Energy and the Environment, Vol. 26, No. 1, 2001, pp. 167–200.
[12] International Air Transport Association and others, “Vision 2050,” Available in: https://www.iata.org/
pressroom/facts_figures/Documents/vision-2050.pdf , 2011.
[13] Anderson Jr, J. D., Fundamentals of aerodynamics, Tata McGraw-Hill Education, 2010.
223
BIBLIOGRAPHY
[14] Thomas, A., “Aircraft drag reduction technology,” Tech. rep., LOCKHEED-GEORGIA CO MARIETTA FLIGHT
SCIENCES DIV, 1984.
[15] Kroo, I., “Drag due to lift: concepts for prediction and reduction,” Annual review of fluid mechanics, Vol. 33,
No. 1, 2001, pp. 587–617.
[16] Torenbeek, E., Synthesis of subsonic airplane design: an introduction to the preliminary design of subsonic
general aviation and transport aircraft, with emphasis on layout, aerodynamic design, propulsion and
performance, Springer Science & Business Media, 2013.
[17] Glauert, H., The elements of aerofoil and airscrew theory, Cambridge University Press, 1983.
[18] Jansen, P. W., Perez, R. E., and RA Martins, J. R., “Aerostructural optimization of nonplanar lifting surfaces,”
Journal of Aircraft, Vol. 47, No. 5, 2010, pp. 1490–1503.
[19] Takenaka, K., Hatanaka, K., Yamazaki, W., and Nakahashi, K., “Multidisciplinary design exploration for a
winglet,” Journal of Aircraft, Vol. 45, No. 5, 2008, pp. 1601–1611.
[20] Verstraeten, J. G., and Slingerland, R., “Drag characteristics for optimally span-loaded planar, wingletted, and
C wings,” Journal of Aircraft, Vol. 46, No. 3, 2009, pp. 962–971.
[21] Ning, A., and Kroo, I., “Multidisciplinary considerations in the design of wings and wing tip devices,” Journal of
Aircraft, Vol. 47, No. 2, 2010, pp. 534–543.
[24] Wilson, T., Herring, M., Pattinson, J., Cooper, J., Castrichini, A., Rafic, A., and Dhoru, H., “An aircraft wing with
a moveable wing tip device for load alleviation,” , Jan. 3 2019.
US Patent App. 16/067,221.
[25] Castrichini, A., Hodigere Siddaramaiah, V., Calderon, D., Cooper, J. E., Wilson, T., and Lemmens, Y., “Nonlinear
folding wing tips for gust loads alleviation,” Journal of Aircraft, Vol. 53, No. 5, 2016, pp. 1391–1399.
[26] Castrichini, A., Siddaramaiah, V. H., Calderon, D., Cooper, J., Wilson, T., and Lemmens, Y., “Preliminary
investigation of use of flexible folding wing tips for static and dynamic load alleviation,” The Aeronautical
Journal, Vol. 121, No. 1235, 2017, pp. 73–94.
doi:10.1017/aer.2016.108.
[27] Airbus, “How the albatross is inspiring the next generation of aircraft wings,” https://www.airbus.com/
newsroom/press-releases/en/2019/06/how-the-albatross-is-inspiring-next-generation-of-aircraft-wings.
html, ????.
[Online; accessed 26-June-2019].
[28] Greitzer, E., Bonnefoy, P., De la Rosa Blanco, E., Dorbian, C., Drela, M., Hall, D., Hansman, R., Hileman, J.,
Liebeck, R., and Lovegren, J., “N+ 3 Aircraft Concept Designs and Trade Studies, Final Report Volume 1,”
Tech. rep., NASA, CR-2010-216794, 2010.
224
BIBLIOGRAPHY
[29] Airbus, “Airbus Concept Plane offers glimpse into the future of flight,” https://www.airbus.com/newsroom/
press-releases/en/2010/07/airbus-concept-plane-offers-glimpse-into-the-future-of-flight.html, ????.
[Online; accessed 16-January-2019].
[30] Bradley, M. K., Droney, C. K., and Allen, T. J., “Subsonic Ultra Green Aircraft Research. Phase II-Volume I;
Truss Braced Wing Design Exploration,” Tech. rep., NASA, CR-2015-218704, 2015.
URL https://ntrs.nasa.gov/search.jsp?R=20150017036.
[33] Stodieck, O., Cooper, J. E., Weaver, P., and Kealy, P., “Optimization of tow-steered composite wing laminates for
aeroelastic tailoring,” AIAA Journal, Vol. 53, No. 8, 2015, pp. 2203–2215.
doi:https://doi.org/10.2514/1.J053599.
[34] Stodieck, O., Cooper, J., Weaver, P., and Kealy, P., “Aeroelastic tailoring of a representative wing box using
tow-steered composites,” AIAA journal, 2016, pp. 1425–1439.
[35] Stanford, B. K., and Jutte, C. V., “Comparison of curvilinear stiffeners and tow steered composites for aeroelastic
tailoring of aircraft wings,” Computers & Structures, Vol. 183, 2017, pp. 48–60.
[36] Stanford, B. K., Jutte, C. V., and Wieseman, C. D., “Trim and structural optimization of subsonic transport
wings using nonconventional aeroelastic tailoring,” AIAA Journal, Vol. 54, No. 1, 2015, pp. 293–309.
[37] Karpel, M., “Design for active flutter suppression and gust alleviation using state-space aeroelastic modeling,”
Journal of Aircraft, Vol. 19, No. 3, 1982, pp. 221–227.
doi:https://doi.org/10.2514/3.57379.
[38] Roger, K. L., Hodges, G. E., and Felt, L., “Active flutter suppression-a flight test demonstration,” Journal of
Aircraft, Vol. 12, No. 6, 1975, pp. 551–556.
[39] Nguyen, N., Kaul, U., Lebofsky, S., Ting, E., Chaparro, D., and Urnes, J., “Development of variable camber
continuous trailing edge flap for performance adaptive aeroelastic wing,” Tech. Rep. 15ATC-0250, NASA,
2015.
[40] Ting, E., Chaparro, D., Nguyen, N., and Fujiwara, G. E., “Optimization of Variable-Camber Continuous Trailing-
Edge Flap Configuration for Drag Reduction,” Journal of Aircraft, Vol. 55, No. 6, 2018, pp. 2217–2239.
[41] Wlezien, R., Horner, G., McGowan, A., Padula, S., Scott, M., Silcox, R., and Simpson, J., “The aircraft morphing
program,” 39th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference
and Exhibit, 1998, p. 1927.
URL https://ntrs.nasa.gov/search.jsp?R=19980053567.
[42] Lyu, Z., and Martins, J. R., “Aerodynamic shape optimization of an adaptive morphing trailing-edge wing,”
Journal of Aircraft, Vol. 52, No. 6, 2015, pp. 1951–1970.
[43] Boeing, “Spreading our wings: Boeing unveils new Transonic Truss-Braced Wing,” https://www.boeing.com/
features/2019/01/spreading-our-wings-01-19.page, ????
[Online; accessed 17-January-2019].
225
BIBLIOGRAPHY
[44] Gundlach, J. F., Philippe-André, Tétrault, Gern, F. H., Nagshineh-Pour, A. H., Ko, A., Schetz, J. A., Mason,
W. H., Kapania, R. K., Grossman, B., and Haftka, R. T., “Conceptual design studies of a strut-braced wing
transonic transport,” Journal of aircraft, Vol. 37, No. 6, 2000, pp. 976–983.
doi:https://doi.org/10.2514/2.2724.
[45] Meadows, N. A., Schetz, J. A., Kapania, R. K., Bhatia, M., and Seber, G., “Multidisciplinary design optimization
of medium-range transonic truss-braced wing transport aircraft,” Journal of Aircraft, Vol. 49, No. 6, 2012,
pp. 1844–1856.
doi:https://doi.org/10.2514/1.C031695.
[46] Carrier, G., Atinault, O., Dequand, S., Hantrais-Gervois, J., Liauzun, C., Paluch, B., Rodde, A., and Toussaint,
C., “Investigation of a strut-braced wing configuration for future commercial transport,” 28th Congress of
the International Council of the Aeronautical Sciences, ICAS Bonn, 2012, pp. 2012–1.
[47] Mallik, W., Kapania, R. K., and Schetz, J. A., “Effect of flutter on the multidisciplinary design optimization of
truss-braced-wing aircraft,” Journal of Aircraft, Vol. 52, No. 6, 2015, pp. 1858–1872.
doi:https://doi.org/10.2514/1.C033096.
[48] Friedmann, P. P., “Renaissance of aeroelasticity and its future,” Journal of Aircraft, Vol. 36, No. 1, 1999, pp.
105–121.
[49] Wright, J. R., and Cooper, J. E., Introduction to aircraft aeroelasticity and loads, Vol. 20, John Wiley & Sons,
2008.
[50] Fung, Y. C., An introduction to the theory of aeroelasticity, Courier Dover Publications, 2008.
[51] Hodges, D. H., and Pierce, G. A., Introduction to structural dynamics and aeroelasticity, Vol. 15, cambridge
university press, 2011.
[52] Hoblit, F. M., Gust loads on aircraft: concepts and applications, American Institute of Aeronautics and Astro-
nautics, 1988.
[53] Livne, E., “Future of airplane aeroelasticity,” Journal of Aircraft, Vol. 40, No. 6, 2003, pp. 1066–1092.
[54] Bendiksen, O., “Modern developments in computational aeroelasticity,” Proceedings of the Institution of Me-
chanical Engineers, Part G: Journal of Aerospace Engineering, Vol. 218, No. 3, 2004, pp. 157–177.
[55] Dowell, E., Edwards, J., and Strganac, T., “Nonlinear aeroelasticity,” Journal of aircraft, Vol. 40, No. 5, 2003, pp.
857–874.
[56] Dowell, E. H., and Tang, D., “Nonlinear aeroelasticity and unsteady aerodynamics,” AIAA journal, Vol. 40, No. 9,
2002, pp. 1697–1707.
[57] Favre, C., “Fly-by-wire for commercial aircraft: the Airbus experience,” International Journal of Control, Vol. 59,
No. 1, 1994, pp. 139–157.
[58] Liddle, S., and Crowther, W., “Systems and certification issues for active flow control systems for seperation
control on civil transport aircraft,” 46th AIAA Aerospace Sciences Meeting and Exhibit, 2008, p. 158.
[59] Abbas, A., De Vicente, J., and Valero, E., “Aerodynamic technologies to improve aircraft performance,” Aerospace
Science and Technology, Vol. 28, No. 1, 2013, pp. 100–132.
226
BIBLIOGRAPHY
[60] Livne, E., “Aircraft active flutter suppression: State of the art and technology maturation needs,” Journal of
Aircraft, Vol. 55, No. 1, 2017, pp. 410–452.
doi:https://doi.org/10.2514/1.C034442.
[61] Locatelli, D., Mulani, S. B., and Kapania, R. K., “Wing-box weight optimization using curvilinear spars and ribs
(SpaRibs),” Journal of Aircraft, Vol. 48, No. 5, 2011, pp. 1671–1684.
[62] Shirk, M. H., Hertz, T. J., and Weisshaar, T. A., “Aeroelastic tailoring-theory, practice, and promise,” Journal of
Aircraft, Vol. 23, No. 1, 1986, pp. 6–18.
[63] Barbarino, S., Bilgen, O., Ajaj, R. M., Friswell, M. I., and Inman, D. J., “A review of morphing aircraft,” Journal
of intelligent material systems and structures, Vol. 22, No. 9, 2011, pp. 823–877.
[64] Mallik, W., Kapania, R. K., and Schetz, J. A., “Aeroelastic Applications of a Variable-Geometry Raked Wingtip,”
Journal of Aircraft, 2016, pp. 1–13.
[65] Woods, B. K., Dayyani, I., and Friswell, M. I., “Fluid/structure-interaction analysis of the fish-bone-active-camber
morphing concept,” Journal of Aircraft, Vol. 52, No. 1, 2014, pp. 307–319.
[66] Woods, B. K., Bilgen, O., and Friswell, M. I., “Wind tunnel testing of the fish bone active camber morphing
concept,” Journal of Intelligent Material Systems and Structures, Vol. 25, No. 7, 2014, pp. 772–785.
[67] Rizzi, A., “Modeling and simulating aircraft stability and control the SimSAC project,” Progress in Aerospace
Sciences, Vol. 47, No. 8, 2011, pp. 573–588.
[68] Rizzi, A., Zhang, M., Nagel, B., Boehnke, D., and Saquet, P., “Towards a unified framework using CPACS for
geometry management in aircraft design,” 50th AIAA aerospace sciences meeting including the new horizons
forum and aerospace exposition, 2012, p. 549.
[69] Cavagna, L., Ricci, S., and Travaglini, L., “NeoCASS: an integrated tool for structural sizing, aeroelastic analysis
and MDO at conceptual design level,” Progress in Aerospace Sciences, Vol. 47, No. 8, 2011, pp. 621–635.
[70] Hwang, J., and Martins, J., “GeoMACH: geometry-centric MDAO of aircraft configurations with high fidelity,”
12th AIAA Aviation Technology, Integration, and Operations (ATIO) Conference and 14th AIAA/ISSMO
Multidisciplinary Analysis and Optimization Conference, ????, p. 5605.
[71] Travaglini, L., Ricci, S., and Bindolino, G., “PyPAD: a multidisciplinary framework for preliminary airframe
design,” 56th AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2015, p.
0186.
[72] Ostergaard, M. G., Ibbotson, A. R., Le Roux, O., and Prior, A. M., “Virtual testing of aircraft structures,” CEAS
Aeronautical Journal, Vol. 1, No. 1-4, 2011, p. 83.
[73] Bendiksen, O. O., “Review of unsteady transonic aerodynamics: theory and applications,” Progress in Aerospace
Sciences, Vol. 47, No. 2, 2011, pp. 135–167.
[74] Jocelyn Lawrence, A., and Jackson, P., “Comparison of Different Methods of Assessing the Free Oscillatory
Characteristics of Aeroelastic Systems,” British Aeronautical Research Council, CP, , No. l084, 1968, p.
l970.
[75] Garrick, I., and Reed III, W. H., “Historical development of aircraft flutter,” Journal of Aircraft, Vol. 18, No. 11,
1981, pp. 897–912.
227
BIBLIOGRAPHY
[76] Cassel, K. W., Variational methods with applications in science and engineering, Cambridge University Press,
2013.
[77] Wagner, H., “The production of the dynamic lift on wings,” Zeitschrift fur Angewandte Mathematik und Mechanik,
Vol. 5, 1925, pp. 17–35.
[78] Theodorsen, T., and Garrick, I., “Mechanism of flutter a theoretical and experimental investigation of the flutter
problem,” Tech. rep., NACA, TR-685, 1940.
[79] Leishman, J., and Nguyen, K., “State-space representation of unsteady airfoil behavior,” AIAA journal, Vol. 28,
No. 5, 1990, pp. 836–844.
[80] Frazer, R. A., Duncan, W. J., Collar, A. R., et al., Elementary matrices and some applications to dynamics and
differential equations, Vol. 1963, Cambridge University Press Cambridge, 1938.
[81] Bartels, R. E., and Sayma, A., “Computational aeroelastic modelling of airframes and turbomachinery: progress
and challenges,” Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering
Sciences, Vol. 365, No. 1859, 2007, pp. 2469–2499.
[82] Rodden, W. P., and Johnson, E. H., MSC/NASTRAN aeroelastic analysis: user’s guide; Version 68, MacNeal-
Schwendler Corporation, MSC Software Corporation, 4675 MacArthur Court, Suite 900, Newport Beach,
CA 92660, 1994.
[83] Guyan, R. J., “Reduction of stiffness and mass matrices,” AIAA journal, Vol. 3, No. 2, 1965, pp. 380–380.
[84] Klerk, D. d., Rixen, D. J., and Voormeeren, S., “General framework for dynamic substructuring: history, review
and classification of techniques,” AIAA journal, Vol. 46, No. 5, 2008, pp. 1169–1181.
[85] Roger, K., “Airplane math modelling and active aeroelastic control design,” AGARD Structures and Materials
Panel, 1977.
[86] Vepa, R., “Finite state modeling of aeroelastic systems,” Tech. rep., NASA, CR-2779, 1977.
[87] Edwards, J. W., “Unsteady aerodynamic modeling for arbitrary motions,” AIAA Journal, Vol. 17, No. 4, 1979, pp.
365–374.
[88] Leishman, J. G., “Unsteady lift of a flapped airfoil by indicial concepts,” Journal of Aircraft, Vol. 31, No. 2, 1994,
pp. 288–297.
[89] Isogai, K., “On the transonic-dip mechanism of flutter of a sweptback wing,” AIAA journal, Vol. 17, No. 7, 1979,
pp. 793–795.
[90] Isogai, K., “Transonic dip mechanism of flutter of a sweptback wing. II,” AIAA Journal, Vol. 19, No. 9, 1981, pp.
1240–1242.
[91] Chargin, M., and Gartmeier, O., “A finite element procedure for calculating fluid-structure interaction using
MSC/NASTRAN,” Computational Flight Testing, 1990.
[92] Kamakoti, R., and Shyy, W., “Fluid–structure interaction for aeroelastic applications,” Progress in Aerospace
Sciences, Vol. 40, No. 8, 2004, pp. 535–558.
doi:https://doi.org/10.1016/j.paerosci.2005.01.001.
228
BIBLIOGRAPHY
[93] Romanelli, G., Castellani, M., Mantegazza, P., and Ricci, S., “Coupled CSD/CFD non-linear aeroelastic trim of
free-flying flexible aircraft,” 53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and
Materials Conference 20th AIAA/ASME/AHS Adaptive Structures Conference 14th AIAA, ????, p. 1562.
[94] Huebner, A., and Reimer, L., “Gust Encounter Simulations of a Generic Transport Aircraft and Analysis of
Load Alleviation Potentials by Control Surface Deflections Using a RANS-CFD-based Multidisciplinary
Simulation Environment,” AIAA Aviation 2019 Forum, 2019, p. 3198.
[95] Yang, G., Zheng, G., and Li, G., “Computational methods and engineering applications of static/dynamic
aeroelasticity based on CFD/CSD coupling solution,” Science China Technological Sciences, Vol. 55, No. 9,
2012, pp. 2453–2461.
[96] Förster, M., and Breitsamter, C., “Aeroelastic prediction of discrete gust loads using nonlinear and time-
linearized CFD-methods,” Journal of Aeroelasticity and Structural Dynamics, Vol. 3, No. 3, 2015.
[97] Reimer, L., Ritter, M., Heinrich, R., and Krüger, W., “CFD-based gust load analysis for a free-flying flexible
passenger aircraft in comparison to a DLM-based approach,” 22nd AIAA computational fluid dynamics
conference, 2015, p. 2455.
[98] Mallik, W., Schetz, J. A., and Kapania, R. K., “Rapid transonic flutter analysis for aircraft conceptual design
applications,” AIAA Journal, Vol. 56, No. 6, 2018, pp. 2389–2402.
[99] Eaton, A., Howcroft, C., Coetzee, E., Neild, S., Lowenberg, M., and Cooper, J., “Numerical Continuation of Limit
Cycle Oscillations and Bifurcations in High-Aspect-Ratio Wings,” Aerospace, Vol. 5, No. 3, 2018, p. 78.
[100] Bartels, R. E., Funk, C. J., and Scott, R. C., “Limit-Cycle Oscillation of the Subsonic Ultra-Green Aircraft
Research Truss-Braced Wing Aeroelastic Model,” Journal of Aircraft, 2017.
[101] Hodges, D. H., “A mixed variational formulation based on exact intrinsic equations for dynamics of moving
beams,” International journal of solids and structures, Vol. 26, No. 11, 1990, pp. 1253–1273.
[102] Hodges, D. H., Atilgan, A. R., Cesnik, C. E., and Fulton, M. V., “On a simplified strain energy function for
geometrically nonlinear behaviour of anisotropic beams,” Composites Engineering, Vol. 2, No. 5-7, 1992, pp.
513–526.
[103] Howcroft, C., Neild, S., Lowenberg, M., and Cooper, J., “Efficient aeroelastic beam modelling and the selection of
a structural shape basis,” International Journal of Non-Linear Mechanics, Vol. 112, 2019, pp. 73–84.
[104] Patil, M. J., Hodges, D. H., and S. Cesnik, C. E., “Nonlinear aeroelasticity and flight dynamics of high-altitude
long-endurance aircraft,” Journal of Aircraft, Vol. 38, No. 1, 2001, pp. 88–94.
[105] Shearer, C. M., and Cesnik, C. E., “Nonlinear flight dynamics of very flexible aircraft,” Journal of Aircraft,
Vol. 44, No. 5, 2007, pp. 1528–1545.
[106] Su, W., “Nonlinear Aeroelastic Analysis of Aircraft with Strut-Braced Highly Flexible Wings,” 58th
AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2017, p. 1351.
doi:https://doi.org/10.2514/MSDM17.
[107] Su, W., and S. Cesnik, C. E., “Nonlinear aeroelasticity of a very flexible blended-wing-body aircraft,” Journal of
Aircraft, Vol. 47, No. 5, 2010, pp. 1539–1553.
[108] Sulaeman, E., Kapania, R., and Haftka, R., “Parametric studies of flutter speed in a strut-braced wing,” 43rd
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2002, p. 1487.
229
BIBLIOGRAPHY
[109] Bhatia, M., Kapania, R. K., and Haftka, R. T., “Structural and aeroelastic characteristics of truss-braced wings:
A parametric study,” Journal of Aircraft, Vol. 49, No. 1, 2012, pp. 302–310.
[110] Green, J., “Laminar flow control-back to the future?” 38th Fluid Dynamics Conference and Exhibit, 2008, p.
3738.
[111] Philippe-André, Té, trault, Schetz, J. A., and Grossman, B., “Numerical prediction of interference drag of
strut-surface intersection in transonic flow,” AIAA journal, Vol. 39, No. 5, 2001, pp. 857–864.
[112] Duggirala, R. K., Roy, C. J., and Schetz, J. A., “Analysis of interference drag for strut-strut interaction in
transonic flow,” AIAA journal, Vol. 49, No. 3, 2011, pp. 449–462.
[113] Cavallaro, R., and Demasi, L., “Challenges, ideas, and innovations of joined-wing configurations: a concept from
the past, an opportunity for the future,” Progress in Aerospace Sciences, Vol. 87, 2016, pp. 1–93.
[114] Nico Braas, “1000 Aircraft Photos - Nico Braas Collection,” http://1000aircraftphotos.com/Contributions/
Braas/, ????
[Online; accessed 17-January-2019].
[115] Jacques Tremp, “1000 Aircraft Photos - Jacques Trempe Collection,” http://1000aircraftphotos.com/
Contributions/Trempe/, ????
[Online; accessed 17-January-2019].
[116] Walter Van Tilborg, “1000 Aircraft Photos - Walter Van Tilborg Collection,” http://1000aircraftphotos.com/
Contributions/VanTilborg/, ????
[Online; accessed 17-January-2019].
[117] Fred Kullas, “1000 Aircraft Photos - Fred Kullas Collection,” http://1000aircraftphotos.com/Contributions/
KullasFred/, ????
[Online; accessed 17-January-2019].
[118] Lars Opland, “1000 Aircraft Photos - Lars Opland Collection,” http://1000aircraftphotos.com/Contributions/
Opland/, ????
[Online; accessed 17-January-2019].
[119] Detlef Billig, “1000 Aircraft Photos - Detlef Billig Collection,” http://1000aircraftphotos.com/Contributions/
Billig/, ????
[Online; accessed 17-January-2019].
[120] Hurel, M., “The Advantages of High Aspect Ratios,” Interavia, Vol. VII, No. 12, 1952, pp. 695–699.
[121] “Hurel-Dubois H.D. 32: An Appraisal of the Unorthodox yet Promising Type Ordered by Air France,” Flight and
Aircraft Engineer: Official Organ of the Royal Aero Club, Vol. 2369, No. 65, 1954, pp. 793–796.
[122] “Hurel Dubois Transports: Progress with the HD-31 and 32: The HD-45 Jet Project,” Flight and Aircraft
Engineer: Official Organ of the Royal Aero Club, Vol. 2288, No. 62, 1952, pp. 676–677.
[123] “H.D.M. 105–Hurel’s Aero with Mile’s Van,” Flight, 1957, pp. 504–505.
[124] Donald, D., The Encyclopedia of Civil Aircraft, Aurum Press, 1999.
[125] Bridgman, L., Janes’ All The World’s Aircraft 1951–1952, Sampson Low, Marston & Compant, Ltd., 1951.
230
BIBLIOGRAPHY
[126] Michell, S., “Civil and Military Aircraft Upgrades 1994–95,” , 1994.
[127] Lambert, M., Janes’ All The World’s Aircraft 1994–1995, Jane’s Information Group, 1994.
[129] Frawley, G., The International Directory of Civil Aircraft 2003/2004, MBI Publishing Company, 2003.
[130] Pfenninger, W., “Design considerations of large subsonic long range transport airplanes with low drag boundary
layer suction,” Tech. rep., Northrop Aircraft, Inc., Report NAI-54-800 (BLC-67), 1954.
[131] Pfenninger, W., “Special course on concepts for drag reduction, laminar flow control,” Tech. rep., laminarization.
Tech. Rep. AGARD Report, 1977.
[132] Kulfani, R. M., and Vachal, J. D., “Wing Planform Geometry Effects on Large Subsonic Military Transport
Airplanes.” Tech. rep., DTIC Document, 1978.
[133] Jobe, C. E., Kulfani, R. M., and Vachal, J. D., “Wing planforms for large military transports,” Journal of Aircraft,
Vol. 16, No. 7, 1979, pp. 425–432.
[134] Gern, F., Ko, A., Sulaeman, E., Kapania, R., Mason, W., Grossman, B., and Haftka, R., “Passive load alleviation
in the design of a strut-braced wing transonic transport aircraft,” 8th Symposium on Multidisciplinary
Analysis and Optimization, 2000, p. 4826.
doi:https://doi.org/10.2514/MMAO00.
[135] Turriziani, R., Lovell, W., Martin, G., Price, J., Swanson, E., and Washburn, G., “Preliminary design charac-
teristics of a subsonic business jet concept employing an aspect ratio 25 strut braced wing,” Tech. rep.,
1981-0002505, 1980.
[136] Smith, P. M., DeYoung, J., Lovell, W. A., Price, J. E., and Washburn, G. F., “A study of high-altitude manned
research aircraft employing strut-braced wings of high-aspect-ratio,” 1981.
[137] Park, P. H., “Fuel consumption of a strutted vs cantilever-winged short-haul transport with aeroelastic consider-
ations,” Journal of Aircraft, Vol. 17, No. 12, 1980, pp. 856–860.
[138] Gern, F. H., Ko, A., Sulaeman, E., Gundlach, J. F., Kapania, R. K., and Haftka, R. T., “Multidisciplinary design
optimization of a transonic commercial transport with strut-braced wing,” Journal of aircraft, Vol. 38, No. 6,
2001, pp. 1006–1014.
[139] Grasmeyer III, J. M., “Multidisciplinary design optimization of a strut-braced wing aircraft,” Ph.D. thesis,
Virginia Tech, Blacksburg, VA 24061, USA, 1998.
URL http://hdl.handle.net/10919/36729.
[140] Naghshineh-Pour, A. H., “Structural optimization and design of a strut-braced wing aircraft,” Ph.D. thesis,
Virginia Tech, Blacksburg, VA 24061, USA, 1998.
URL http://hdl.handle.net/10919/36142.
[141] Sulaeman, E., “Effect of compressive force on aeroelastic stability of a strut-braced wing,” Ph.D. thesis, Virginia
Tech, Blacksburg, VA 24061, USA, 2001.
231
BIBLIOGRAPHY
[142] Tetrault, P.-A., “Numerical Prediction of the Interference Drag of a Streamlined Strut Intersecting a Surface in
Transonic Flow,” Ph.D. thesis, Virginia Tech, Blacksburg, VA 24061, USA, 2000.
[143] McCullers, L., “FLOPS: Aircraft Configuration Optimization,” Tech. rep., NASA, CP-2327, 1984.
[144] WK, A., and Daryl L, B., “An implicit upwind algorithm for computing turbulent flows on unstructured grids,”
Computer Fluids, Vol. 23, No. 1, 1994, pp. 1–21.
[145] Anderson, W. K., Rausch, R. D., and Bonhaus, D. L., “Implicit/multigrid algorithms for incompressible turbulent
flows on unstructured grids,” Journal of Computational Physics, Vol. 128, No. 2, 1996, pp. 391–408.
[146] Bradley, M. K., and Droney, C. K., “Subsonic ultra green aircraft research phase II: N+ 4 advanced concept
development,” Tech. rep., NASA, CR-2012-217556, 2012.
[147] Bradley, M. K., and Droney, C. K., “Subsonic Ultra Green Aircraft Research: Phase II-Volume II; Hybrid Electric
Design Exploration,” Tech. rep., NASA, CR-2015-218704, 2015.
[148] Bradley, M. K., Allen, T. J., and Droney, C., “Subsonic Ultra Green Aircraft Research: Phase II-Volume III; Truss
Braced Wing Aeroelastic Test Report,” Tech. rep., NASA, CR-2015-218704, 2014.
URL https://ntrs.nasa.gov/search.jsp?R=20150017040.
[149] Kapania, R. K., Schetz, J. A., Mallik, W., Segee, M. C., and Gupta, R., “Multidisciplinary Design Optimization
and Cruise Mach Number Study of Truss-Braced Wing Aircraft,” Tech. rep., NASA, CR-2018-219836, 2018.
[151] Gur, O., Bhatia, M., Mason, W. H., Schetz, J. A., Kapania, R. K., and Nam, T., “Development of a framework for
truss-braced wing conceptual MDO,” Structural and Multidisciplinary Optimization, Vol. 44, No. 2, 2011,
pp. 277–298.
[152] Wells, D. P., “Wing Configuration Impact on Design Optimums for a Subsonic Passenger Transport,” 52nd
Aerospace Sciences Meeting, 2014, p. 0185.
[153] Variyar, A., Economon, T. D., and Alonso, J. J., “Multifidelity conceptual design and optimization of strut-braced
wing aircraft using physics based methods,” 54th AIAA Aerospace Sciences Meeting, 2016, p. 2000.
[154] Rajpal, D., and De Breuker, R., “Dynamic Aeroelastic Tailoring of a Strut Braced Wing Including Fatigue Loads,”
Proceedings of the 18th International Forum on Aeroelasticity and Structural Dynamics, Savannah, Georgia,
USA, 2019.
[155] Gur, O., Schetz, J. A., and Mason, W. H., “Aerodynamic considerations in the design of truss-braced-wing
aircraft,” Journal of Aircraft, Vol. 48, No. 3, 2011, pp. 919–939.
[156] Ting, E., Reynolds, K. W., Nguyen, N. T., and Totah, J., “Aerodynamic Analysis of the Truss-Braced Wing
Aircraft Using Vortex-Lattice Superposition Approach,” 32nd AIAA Applied Aerodynamics Conference, 2014,
p. 2597.
[157] Seber, G., Ran, H., Nam, T., Schetz, J., and Mavris, D., “Multidisciplinary Design Optimization of a Truss Braced
Wing Aircraft with Upgraded Aerodynamic Analyses,” 29th AIAA Applied Aerodynamics Conference, 2011,
p. 3179.
232
BIBLIOGRAPHY
[158] Ko, A., Mason, W., and Grossman, B., “Transonic aerodynamics of a wing/pylon/strut juncture,” 21st AIAA
Applied Aerodynamics Conference, 2003, p. 4062.
[159] Gray, J. S., Hwang, J. T., Martins, J. R., Moore, K. T., and Naylor, B. A., “OpenMDAO: An open-source framework
for multidisciplinary design, analysis, and optimization,” Structural and Multidisciplinary Optimization,
Vol. 59, No. 4, 2019, pp. 1075–1104.
[160] Hwang, J., Kenway, G., and Martins, J., “Geometry and structural modeling for high-fidelity aircraft conceptual
design optimization,” 15th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, 2014, p.
2041.
[161] Ivaldi, D., Secco, N. R., Chen, S., Hwang, J. T., and Martins, J., “Aerodynamic Shape Optimization of a Truss-
Braced-Wing Aircraft,” 16th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference, 2015,
p. 3436.
[162] Secco, N. R., and Martins, J., “RANS-based aerodynamic shape optimization of a strut-braced wing with overset
meshes,” 2018 AIAA/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2018,
p. 0413.
[163] Xiong, J., Fugate, J., and Nguyen, N. T., “Investigation of Truss-Braced Wing Aircraft Transonic Wing-Strut
Interference Effects Using FUN3D,” AIAA Aviation 2019 Forum, 2019, p. 3026.
[164] Nguyen, N. T., Fugate, J., and Xiong, J., “Aero-Structural Modeling of the Truss-Braced Wing Aircraft Using
Potential Method with Correction Methods for Transonic Viscous Flow and Wing-Strut Interference
Aerodynamics,” AIAA Aviation 2019 Forum, 2019, p. 3028.
[165] Chiozzotto, G. P., “Wing weight estimation in conceptual design: a method for strut-braced wings considering
static aeroelastic effects,” CEAS Aeronautical Journal, Vol. 7, No. 3, 2016, pp. 499–519.
[166] Rocha, H., Li, W., and Hahn, A., “Principal component regression for fitting wing weight data of subsonic
transports,” Journal of aircraft, Vol. 43, No. 6, 2006, pp. 1925–1936.
[167] Scott, R. C., Bartels, R. E., Funk, C. J., Allen, T. J., Sexton, B. W., Dykman, J. R., and Coulson, D. A., “Aeroser-
voelastic test of the subsonic ultra-green aircraft research truss-braced wing model,” Journal of Guidance,
Control, and Dynamics, 2016, pp. 1820–1833.
[168] Zhao, W., Kapania, R. K., Schetz, J. A., and Coggin, J. M., “Nonlinear aeroelastic analysis of SUGAR truss-braced
wing (TBW) wind-tunnel model (WTM) under in-plane loads,” 56th AIAA/ASCE/AHS/ASC Structures,
Structural Dynamics, and Materials Conference, 2015, p. 1173.
[169] Butt, L., Bhatia, M., and Kapania, R., “Flutter Modeling and Suppression for a Strut-Braced Wing,”
53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference 20th
AIAA/ASME/AHS Adaptive Structures Conference 14th AIAA, 2015, p. 1794.
[170] Marchetti, L., De Gaspari, A., Riccobene, L., Toffol, F., Fonte, F., Ricci, S., Mantegazza, P., Livne, E., and
Hinson, K. A., “Active Flutter Suppression Analysis and Wind Tunnel Studies of a Commercial Transport
Configuration,” AIAA Scitech 2020 Forum, 2020, p. 1677.
[171] Hashemi, K. E., Nguyen, N. T., Drew, M. C., Chaparro, D., and Ting, E., “Performance optimizing gust load
alleviation control of flexible wing aircraft,” 2018 AIAA Guidance, Navigation, and Control Conference,
2018, p. 0623.
233
BIBLIOGRAPHY
[172] Zhao, Y., Yue, C., and Hu, H., “Gust load alleviation on a large transport airplane,” Journal of Aircraft, Vol. 53,
No. 6, 2016, pp. 1932–1946.
[173] Chen, G., Zhou, Q., Ronch, A. D., and Li, Y., “Computational-Fluid-Dynamics-Based Aeroservoelastic Analysis
for Gust Load Alleviation,” Journal of Aircraft, Vol. 55, No. 4, 2018, pp. 1619–1628.
[174] Liu, X., Sun, Q., and Cooper, J., “LQG based model predictive control for gust load alleviation,” Aerospace science
and Technology, Vol. 71, 2017, pp. 499–509.
[175] Dillsaver, M., Cesnik, C., and Kolmanovsky, I., “Gust load alleviation control for very flexible aircraft,” AIAA
Atmospheric Flight Mechanics Conference, 2011, p. 6368.
[176] Haghighat, S., Liu, H. H., and Martins, J. R., “Model-predictive gust load alleviation controller for a highly
flexible aircraft,” Journal of Guidance, Control, and Dynamics, Vol. 35, No. 6, 2012, pp. 1751–1766.
[177] Sharp, R., and Crolla, D., “Road vehicle suspension system design-a review,” Vehicle system dynamics, Vol. 16,
No. 3, 1987, pp. 167–192.
[178] Soto, M. G., and Adeli, H., “Tuned mass dampers,” Archives of Computational Methods in Engineering, Vol. 20,
No. 4, 2013, pp. 419–431.
[179] Casalotti, A., Arena, A., and Lacarbonara, W., “Mitigation of post-flutter oscillations in suspension bridges by
hysteretic tuned mass dampers,” Engineering Structures, Vol. 69, 2014, pp. 62–71.
[180] Carpineto, N., Lacarbonara, W., and Vestroni, F., “Mitigation of pedestrian-induced vibrations in suspension
footbridges via multiple tuned mass dampers,” Journal of Vibration and Control, Vol. 16, No. 5, 2010, pp.
749–776.
[181] Dallard, P., Fitzpatrick, A., Flint, A., Le Bourva, S., Low, A., Ridsdill Smith, R., and Willford, M., “The London
millennium footbridge,” Structural Engineer, Vol. 79, No. 22, 2001, pp. 17–21.
[182] Havard, D., and Pohlman, J., “Field testing of detuning pendulums for controlling galloping of single and bundle
conductors,” IEE Paper A, Vol. 79, 1979, pp. 499–5.
[183] Rowbottom, M., “The optimization of mechanical dampers to control self-excited galloping oscillations,” Journal
of Sound and Vibration, Vol. 75, No. 4, 1981, pp. 559–576.
[184] Haskett, T., Breukelman, B., Robinson, J., and Kottelenberg, J., “Tuned-Mass Damper Under Excessive Struc-
tural Excitation,” Report of the Motioneering Inc., Guelph, Ontario, Canada, 2004.
[185] Wang, D., and Liao, W. H., “Magnetorheological fluid dampers: a review of parametric modelling,” Smart
materials and structures, Vol. 20, No. 2, 2011, p. 023001.
[186] Zhu, X., Jing, X., and Cheng, L., “Magnetorheological fluid dampers: a review on structure design and analysis,”
Journal of intelligent material systems and structures, Vol. 23, No. 8, 2012, pp. 839–873.
[187] Imaduddin, F., Mazlan, S. A., and Zamzuri, H., “A design and modelling review of rotary magnetorheological
damper,” Materials & Design, Vol. 51, 2013, pp. 575–591.
[188] Muhammad, A., Yao, X.-l., and Deng, Z.-c., “Review of magnetorheological (MR) fluids and its applications in
vibration control,” Journal of Marine Science and Application, Vol. 5, No. 3, 2006, pp. 17–29.
[189] Lee, D., and Taylor, D. P., “Viscous damper development and future trends,” The Structural Design of Tall
Buildings, Vol. 10, No. 5, 2001, pp. 311–320.
234
BIBLIOGRAPHY
[190] Titurus, B., “Generalized liquid-based damping device for passive vibration control,” AIAA Journal, Vol. 56,
No. 10, 2018, pp. 4134–4145.
[191] Rittweger, A., Albus, J., Hornung, E., Öry, H., and Mourey, P., “Passive damping devices for aerospace structures,”
Acta Astronautica, Vol. 50, No. 10, 2002, pp. 597–608.
[192] Panda, B., Mychalowycz, E., and Tarzanin, F. J., “Application of passive dampers to modern helicopters,” Smart
Materials and Structures, Vol. 5, No. 5, 1996, p. 509.
[193] Symans, M. D., and Constantinou, M. C., “Semi-active control systems for seismic protection of structures: a
state-of-the-art review,” Engineering structures, Vol. 21, No. 6, 1999, pp. 469–487.
[194] Titurus, B., “Vibration control in a helicopter with semi-active hydraulic lag dampers,” Journal of Guidance,
Control, and Dynamics, Vol. 36, No. 2, 2013, pp. 577–588.
[195] Fisco, N., and Adeli, H., “Smart structures: part I active and semi-active control,” Scientia Iranica, Vol. 18,
No. 3, 2011, pp. 275–284.
[196] Hrovat, D., Barak, P., and Rabins, M., “Semi-active versus passive or active tuned mass dampers for structural
control,” Journal of Engineering Mechanics, Vol. 109, No. 3, 1983, pp. 691–705.
[197] Casciati, F., Rodellar, J., and Yildirim, U., “Active and semi-active control of structures–theory and applications:
A review of recent advances,” Journal of Intelligent Material Systems and Structures, Vol. 23, No. 11, 2012,
pp. 1181–1195.
[198] Cheney, N., Ritz, E., and Lipson, H., “Automated vibrational design and natural frequency tuning of multi-
material structures,” Proceedings of the 2014 Annual Conference on Genetic and Evolutionary Computation,
ACM, 2014, pp. 1079–1086.
[199] Frahm, H., “Device for damping vibrations of bodies.” , Apr. 18 1911.
US Patent 989,958.
[201] Leishman, G. J., Principles of helicopter aerodynamics, Cambridge university press, 2006.
[202] Bramwell, A. R. S., Balmford, D., and Done, G., Bramwell’s helicopter dynamics, Elsevier, 2001.
[203] Warburton, G., “Optimum absorber parameters for various combinations of response and excitation parameters,”
Earthquake Engineering & Structural Dynamics, Vol. 10, No. 3, 1982, pp. 381–401.
[204] Krenk, S., “Frequency analysis of the tuned mass damper,” Journal of applied mechanics, Vol. 72, No. 6, 2005,
pp. 936–942.
[205] Krenk, S., and Høgsberg, J., “Tuned mass absorber on a flexible structure,” Journal of Sound and Vibration, Vol.
333, No. 6, 2014, pp. 1577–1595.
[206] Krenk, S., and Høgsberg, J., “Tuned resonant mass or inerter-based absorbers: unified calibration with quasi-
dynamic flexibility and inertia correction,” Proc. R. Soc. A, Vol. 472, No. 2185, 2016, p. 20150718.
doi:https://doi.org/10.1098/rspa.2015.0718.
[207] Barredo, E., Blanco, A., Colín, J., Penagos, V. M., Abúndez, A., Vela, L. G., Meza, V., Cruz, R. H., and Mayén, J.,
“Closed-form solutions for the optimal design of inerter-based dynamic vibration absorbers,” International
Journal of Mechanical Sciences, Vol. 144, 2018, pp. 41–53.
235
BIBLIOGRAPHY
[208] Smith, M. C., “Synthesis of mechanical networks: The inerter,” IEEE Transactions on Automatic Control, Vol. 47,
No. 10, 2002, pp. 1648–1662.
doi:10.1109/TAC.2002.803532.
[209] Zhang, S. Y., Jiang, J. Z., and Neild, S. A., “Passive vibration control: a structure–immittance approach,”
Proceedings of the Royal Society of London A: Mathematical, Physical and Engineering Sciences, Vol. 473,
No. 2201, 2017.
doi:https://doi.org/10.1098/rspa.2017.0011.
[210] Firestone, F. A., “A new analogy between mechanical and electrical systems,” The Journal of the Acoustical
Society of America, Vol. 4, No. 3, 1933, pp. 249–267.
[211] Bott, R., and Duffin, R. J., “Impedance synthesis without use of transformers,” Journal of Applied Physics,
Vol. 20, No. 8, 1949, pp. 816–816.
[212] Guillemin, E. A., Synthesis of passive networks: theory and methods appropriate to the realization and approxi-
mation problems, Wiley, 1957.
[213] Jiang, J. Z., and Smith, M. C., “Regular positive-real functions and five-element network synthesis for electrical
and mechanical networks,” IEEE Transactions on Automatic Control, Vol. 56, No. 6, 2011, pp. 1275–1290.
[214] Zhang, S. Y., Jiang, J. Z., Wang, H. L., and Neild, S., “Synthesis of essential-regular bicubic impedances,”
International Journal of Circuit Theory and Applications, Vol. 45, No. 11, 2017, pp. 1482–1496.
[215] Smith, M. C., and Wang, F.-C., “Performance benefits in passive vehicle suspensions employing inerters,” Vehicle
System Dynamics, Vol. 42, No. 4, 2004, pp. 235–257.
doi:https://doi.org/10.1080/00423110412331289871.
[216] Jiang, J. Z., Smith, M. C., and Houghton, N. E., “Experimental testing and modelling of a mechanical steering
compensator,” 2008 3rd International Symposium on Communications, Control and Signal Processing,
IEEE, 2008, pp. 249–254.
[217] Jiang, J. Z., Matamoros-Sanchez, A. Z., Goodall, R. M., and Smith, M. C., “Passive suspensions incorporating
inerters for railway vehicles,” Vehicle System Dynamics, Vol. 50, No. sup1, 2012, pp. 263–276.
doi:https://doi.org/10.1080/00423114.2012.665166.
[218] Jiang, J., Matamoros-Sanchez, A., Zolotas, A., Goodall, R., and Smith, M., “Passive suspensions for ride quality
improvement of two-axle railway vehicles,” Proceedings of Mechanical Engineering Part F: Journal of Rail
and Rapid Transit, Vol. 229, No. 3, 2015, pp. 315–329.
doi:https://doi.org/10.1177/0954409713511592.
[219] Chen, M. Z., Hu, Y., Li, C., and Chen, G., “Application of semi-active inerter in semi-active suspensions via force
tracking,” Journal of Vibration and Acoustics, Vol. 138, No. 4, 2016, p. 041014.
[220] Lazar, I., Neild, S., and Wagg, D., “Using an inerter-based device for structural vibration suppression,” Earth-
quake Engineering & Structural Dynamics, Vol. 43, No. 8, 2014, pp. 1129–1147.
[221] Marian, L., and Giaralis, A., “Optimal design of a novel tuned mass-damper–inerter (TMDI) passive vibration
control configuration for stochastically support-excited structural systems,” Probabilistic Engineering
Mechanics, Vol. 38, 2014, pp. 156–164.
236
BIBLIOGRAPHY
[222] Zhang, S. Y., Jiang, J. Z., and Neild, S., “Optimal configurations for a linear vibration suppression device in a
multi-storey building,” Structural Control and Health Monitoring, 2016.
doi:https://doi.org/10.1002/stc.1887.
[223] Gonzalez-Buelga, A., Lazar, I. F., Jiang, J. Z., Neild, S. A., and Inman, D. J., “Assessing the effect of nonlinearities
on the performance of a tuned inerter damper,” Structural Control and Health Monitoring, Vol. 24, No. 3,
2017.
doi:https://doi.org/10.1002/stc.1879.
[224] Luo, J., Jiang, J. Z., and Macdonald, J. H., “Cable Vibration Suppression with Inerter-Based Absorbers,” Journal
of Engineering Mechanics, Vol. 145, No. 2, 2018, p. 04018134.
[225] Li, Y., Jiang, J. Z., and Neild, S., “Inerter-Based Configurations for Main-Landing-Gear Shimmy Suppression,”
Journal of Aircraft, 2016, pp. 1–10.
doi:https://doi.org/10.2514/1.C033964.
[226] Li, Y., Jiang, J. Z., Neild, S. A., and Wang, H., “Optimal Inerter-Based Shock–Strut Configurations for Landing-
Gear Touchdown Performance,” Journal of Aircraft, 2017.
doi:https://doi.org/10.2514/1.C034276.
[227] Brzeski, P., Kapitaniak, T., and Perlikowski, P., “Novel type of tuned mass damper with inerter which enables
changes of inertance,” Journal of Sound and Vibration, Vol. 349, 2015, pp. 56–66.
[228] Swift, S., Smith, M. C., Glover, A., Papageorgiou, C., Gartner, B., and Houghton, N. E., “Design and modelling of
a fluid inerter,” International Journal of Control, Vol. 86, No. 11, 2013, pp. 2035–2051.
[229] Liu, X., Jiang, J. Z., Titurus, B., and Harrison, A., “Model identification methodology for fluid-based inerters,”
Mechanical Systems and Signal Processing, Vol. 106, 2018, pp. 479–494.
[230] Liu, X., Titurus, B., and Jiang, J. Z., “Generalisable model development for fluid-inerter integrated damping
devices,” Mechanism and Machine Theory, Vol. 137, 2019, pp. 1–22.
[231] Gonzalez-Buelga, A., Clare, L., Neild, S., Jiang, J., and Inman, D., “An electromagnetic inerter-based vibration
suppression device,” Smart Materials and Structures, Vol. 24, No. 5, 2015, p. 055015.
[233] Walls, J. H., “An experimental study of orifice coefficients, internal strut pressures, and loads on a small
oleo-pneumatic shock strut,” Tech. rep., NACA, TN 3426, 1955.
[234] CS-25 Certification Specifications for Large Aeroplanes, European Aviation Safety Agency (EASA), 2003.
URL https://www.easa.europa.eu/document-library/certification-specifications/cs-25-initial-issue.
[235] Kang, H., Smith, E. C., and Lesieutre, G. A., “Experimental and analytical study of blade lag damping
augmentation using chordwise absorbers,” Journal of aircraft, Vol. 43, No. 1, 2006, pp. 194–200.
[236] Byers, L., and Gandhi, F., “Embedded absorbers for helicopter rotor lag damping,” Journal of Sound and
Vibration, Vol. 325, No. 4-5, 2009, pp. 705–721.
[237] Han, D., Wang, J., Smith, E. C., and Lesieutre, G. A., “Transient loads control of a variable speed rotor during
lagwise resonance crossing,” AIAA journal, Vol. 51, No. 1, 2012, pp. 20–29.
237
BIBLIOGRAPHY
[238] Konstanzer, P., Enenkl, B., Aubourg, P., and Cranga, P., “Recent advances in Eurocopter’s passive and active
vibration control,” Annual Forum Proceedings-American Helicopter Society, Vol. 64, AMERICAN HELI-
COPTER SOCIETY, INC, The Vertical Flight Society, 2701 Prosperity Avenue, Suite 210, Fairfax, VA 22031,
USA, 2008, p. 854.
[239] Kerdreux, B., Jouve, J., Marrot, F., Reymond, M., Priems, M., and Dreher, S., “Vibration comfort improvement
through active cabin vibration control and its certification on EC130T2,” 38th European Rotorcraft Fo-
rum, Netherlands Association of Aeronautical Engineers, Anthony Fokkerweg 2, Amsterdam, 1059 CM,
Netherlands, 2012.
[240] Arrigan, J., Pakrashi, V., Basu, B., and Nagarajaiah, S., “Control of flapwise vibrations in wind turbine blades
using semi-active tuned mass dampers,” Structural Control and Health Monitoring, Vol. 18, No. 8, 2011, pp.
840–851.
[241] Chen, B., Zhang, Z., Hua, X., Nielsen, S. R., and Basu, B., “Enhancement of flutter stability in wind turbines
with a new type of passive damper of torsional rotation of blades,” Journal of Wind Engineering and
Industrial Aerodynamics, Vol. 173, 2018, pp. 171–179.
[242] Schnelz, J. R., and Powell, D. T., “Tuned engine mounting system for jet aircraft,” , Feb. 20 2001.
US Patent 6,189,830.
[244] Hey, K. E., and Polk, J. H., “Adjustable length brace for cyclic loads,” , Oct. 17 2000.
US Patent 6,131,850.
[245] Reed III, W. H., “Decoupler pylon: wing/store flutter suppressor,” , Aug. 10 1982.
US Patent 4,343,447.
[246] Lakic, B., and Vance, J. B., “Aircraft wing-to-fuselage joint with active suspension and method,” , Jul. 26 2016.
[247] Rodden, J., Dougherty, H., Reschke, L., Hasha, M., and Davis, L., “Line-of-sight performance improvement with
reaction-wheel isolation,” Proceedings of the Annual Rocky Mountain Guidance and Control Conference,
Keystone, CO, USA, 1986, pp. 71–84.
[248] Yue, S., Titurus, B., Nie, H., and Zhang, M., “Liquid spring damper for vertical landing Reusable Launch Vehicle
under impact conditions,” Mechanical Systems and Signal Processing, Vol. 121, 2019, pp. 579–599.
[249] Karpel, M., “Design for active and passive flutter suppression and gust alleviation,” Tech. rep., NASA, CR-3482,
1981.
[250] Verstraelen, E., Kerschen, G., and Dimitriadis, G., “Flutter and limit cycle oscillation suppression using linear
and nonlinear tuned vibration absorbers,” Shock & Vibration, Aircraft/Aerospace, Energy Harvesting,
Acoustics & Optics, Volume 9, Springer, 2017, pp. 301–313.
[251] Verstraelen, E., Habib, G., Kerschen, G., and Dimitriadis, G., “Experimental passive flutter suppression using a
linear tuned vibration absorber,” AIAA Journal, 2016, pp. 1707–1722.
[252] Pitt, D. M., and Haudrich, D. P., “Active strut apparatus for use with aircraft and related methods,” , Nov. 1
2016.
US Patent 9,481,450.
238
BIBLIOGRAPHY
[253] Vassberg, J., Dehaan, M., Rivers, M., and Wahls, R., “Development of a common research model for applied
CFD validation studies,” 26th AIAA Applied Aerodynamics Conference, 2008, p. 6919.
[254] Raymer, D., Aircraft Design: A Conceptual Approach 5e and RDSWin STUDENT, American Institute of
Aeronautics and Astronautics, Inc., 2012.
[255] Calderon, D., Cooper, J., Lowenberg, M., Neild, S., and Coetzee, E., “Sizing High-Aspect-Ratio Wings with a
Geometrically Nonlinear Beam Model,” Journal of Aircraft, 2019, pp. 1–16.
[256] Prananta, B., Kanakis, T., Vankan, J., and van Houten, R., “Model updating of finite element model using
optimisation routine,” Aircraft Engineering and Aerospace Technology, Vol. 88, No. 5, 2016, pp. 665–675.
[257] Pototzky, A. S., “Roll damping derivatives from generalized lifting-surface theory and wind tunnel forced-
oscillation tests,” AIAA Atmospheric Flight Mechanics Conference, 2014, p. 0731.
[258] Towner, R., and Band, J., “An analysis technique/automated tool for comparing and tracking analysis modes of
different finite element models,” 53rd AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics
and Materials Conference 20th AIAA/ASME/AHS Adaptive Structures Conference 14th AIAA, 2012, p.
1987.
URL https://ntrs.nasa.gov/search.jsp?R=20120014370.
[259] Gockel, M., MSC/NASTRAN Handbook for Dynamic Analysis: MSC/NASTRAN Version 63, MacNeal-
Schwendler Corporation, MSC Software Corporation, 4675 MacArthur Court, Suite 900, Newport Beach,
CA 92660, 1983.
[260] Allemang, R. J., “The modal assurance criterion–twenty years of use and abuse,” Sound and vibration, Vol. 37,
No. 8, 2003, pp. 14–23.
[261] Albano, E., and Rodden, W. P., “A doublet-lattice method for calculating lift distributions on oscillating surfaces
in subsonic flows.” AIAA journal, Vol. 7, No. 2, 1969, pp. 279–285.
[262] Howcroft, C., Cook, R., Calderon, D., Lambert, L., Castellani, M., Cooper, J., Lowenberg, M., Neild, S., and
Coetzee, E., “Aeroelastic modelling of highly flexible wings,” 15th Dynamics Specialists Conference, 2016, p.
1798.
[263] Pratt, K. G., “A revised formula for the calculation of gust loads,” Tech. rep., NACA TN-2964, Washington, 1953.
[264] Szczyglowski, C. P., Howcroft, C., Neild, S. A., Titurus, B., Jiang, J. Z., Cooper, J. E., and Coetzee, E., “Strut-
Braced Wing Modelling with a Reduced Order Beam Model,” Proceedings of the Royal Aeronautical
Engineering Society 5th Aircraft Structural Design Conference, Manchester, United Kingdom, 2016.
[265] Mallik, W., Kapania, R. K., and Schetz, J. A., “Multidisciplinary Design Optimization of Medium-Range
Transonic Truss-Braced Wing Aircraft with Flutter Constraint,” 54th AIAA/ASME/ASCE/AHS/ASC
Structures, Structural Dynamics, and Materials Conference, 2013, p. 1454.
[266] Mallik, W., “Aeroelastic Analysis of Truss-Braced Wing Aircraft: Applications for Multidisciplinary Design
Optimization,” Ph.D. thesis, Blacksburg, VA 24061, USA, 2016.
[267] Hassig, H. J., “An approximate true damping solution of the flutter equation by determinant iteration.” Journal
of Aircraft, Vol. 8, No. 11, 1971, pp. 885–889.
[268] Jonsson, E., Riso, C., Lupp, C. A., Cesnik, C. E., Martins, J. R., and Epureanu, B. I., “Flutter and post-flutter
constraints in aircraft design optimization,” Progress in Aerospace Sciences, 2019.
239
BIBLIOGRAPHY
[269] Gracey, W., “Measurement of aircraft speed and altitude,” Tech. rep., NATIONAL AERONAUTICS AND SPACE
ADMINISTRATION HAMPTON VA LANGLEY RESEARCH CENTER, 1980.
[270] Moore, G. J., MSC/NASTRAN design sensitivity and optimization: user’s guide, version 68, MacNeal-Schwendler
Corporation, MSC Software Corporation, 4675 MacArthur Court, Suite 900, Newport Beach, CA 92660,
1994.
[271] Szczyglowski, C. P., Neild, S., Titurus, B., Jiang, J. Z., and Coetzee, E., “Passive Gust Load Alleviation In a
Truss-Braced Wing Using an Inerter-Based Device,” 2018 AIAA/ASCE/AHS/ASC Structures, Structural
Dynamics, and Materials Conference, 2018, p. 1958.
doi:https://doi.org/10.2514/6.2018-1958.
[272] Hang, X., Fei, Q., and Su, W., “On Tracking Aeroelastic Modes in Stability Analysis Using Left and Right
Eigenvectors,” AIAA Journal, 2019, pp. 1–11.
[273] Haftka, R., “Structural optimization with aeroelastic constraints: a survey of US applications,” International
Journal of Vehicle Design, Vol. 7, No. 3-4, 1986, pp. 381–392.
[274] Neill, D., Johnson, E., and Canfield, R., “ASTROS-a multidisciplinary automated structural design tool,” Journal
of Aircraft, Vol. 27, No. 12, 1990, pp. 1021–1027.
[275] Guo, S., Cheng, W., and Cui, D., “Optimization of composite wing structures for maximum flutter speed,” 46th
AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics and Materials Conference, ????, p. 2132.
[276] Coleman, T. F., and Li, Y., “An interior trust region approach for nonlinear minimization subject to bounds,”
SIAM Journal on optimization, Vol. 6, No. 2, 1996, pp. 418–445.
[278] Tartaruga, I., Cooper, J. E., Sartor, P., Lowenberg, M. H., and Lemmens, Y., “Geometrical based method for the
uncertainty quantification of correlated aircraft loads,” Journal of Aeroelasticity and Structural Dynamics,
Vol. 4, No. 1, 2016.
[279] Barber, C. B., Dobkin, D. P., Dobkin, D. P., and Huhdanpaa, H., “The quickhull algorithm for convex hulls,” ACM
Transactions on Mathematical Software (TOMS), Vol. 22, No. 4, 1996, pp. 469–483.
[280] Bartels, R. E., Stanford, B., and Waite, J., “Performance Enhancement of the Flexible Transonic Truss-Braced
Wing Aircraft Using Variable-Camber Continuous Trailing-Edge Flaps,” AIAA Aviation 2019 Forum, 2019,
p. 3160.
doi:https://doi.org/10.2514/MAVIAT19.
[281] Kang, B.-S., Park, G.-J., and Arora, J. S., “A review of optimization of structures subjected to transient loads,”
Structural and Multidisciplinary Optimization, Vol. 31, No. 2, 2006, pp. 81–95.
doi:https://doi.org/10.1007/s00158-005-0575-4.
[282] Martí, R., Lozano, J. A., Mendiburu, A., and Hernando, L., Multi-start Methods, Springer International
Publishing, Cham, 2016, pp. 1–21.
doi:10.1007/978-3-319-07153-4_1-1, URL https://doi.org/10.1007/978-3-319-07153-4_1-1.
240
BIBLIOGRAPHY
[283] Hu, X., Shonkwiler, R., and Spruill, M. C., “Random restarts in global optimization,” Tech. rep., Georgia Institute
of Technology, 2009.
URL http://hdl.handle.net/1853/31310.
[284] “Proposed Special Condition for Installation of Flutter Suppression System: Applicable to Boeing 747-8/-8F,”
Tech. Rep. Special Condition C-18, European Aviation Safety Agency, May 2011.
URL https://www.easa.europa.eu/document-library/product-certification-consultations/
proposed-special-condition-installation-flutter, [Online; accessed 12-July-2019].
[285] Peo, R. F., and Lautz, C. F., “Hydraulic Shock Absorber,” , ????
URL https://patents.google.com/patent/US2004902A/en, uS Patent 2,004,902.
[286] Milwitzky, B., and Cook, F. E., “Analysis of landing-gear behavior,” Tech. rep., NATIONAL AERONAUTICS
AND SPACE ADMINISTRATION WASHINGTON DC, 1953.
[287] Bickell, M. B., Pressure vessel design and analysis, Macmillan International Higher Education, 1967.
[288] Currey, N. S., Aircraft landing gear design: principles and practices, American Institute of Aeronautics and
Astronautics, 1988.
[289] jiro (Mathworks File Exchange Online Profile), “GRABIT - Mathworks File Exchange,” https://uk.mathworks.
com/matlabcentral/fileexchange/7173-grabit, ????
[Online; accessed 17-June-2019].
[290] Talay, T. A., Introduction to the Aerodynamics of Flight, Vol. 367, Scientific and Technical Information Office,
National Aeronautics and Space, 1975.
[291] Cavcar, M., “The international standard atmosphere (ISA),” Anadolu University, Turkey, Vol. 30, 2000, p. 9.
[292] Afsar, M. R., Banna, M. A. H., Uddin, M. J., and Salam, M. A., “CEASIOM: An open source multi module
conceptual aircraft design tool,” International Journal of Engineering, Vol. 2, No. 7, 2013.
[293] Von Kaenel, R., Rizzi, A., Oppelstrup, J., Goetzendorf-Grabowski, T., Ghoreyshi, M., Cavagna, L., Berard,
A., et al., “CEASIOM: simulating stability & control with CFD/CSM in aircraft conceptual design,” 26th
International Congress of the Aeronautical Sciences, ICAS, Vol. 1, 2008, pp. 14–19.
[294] Rizzi, A., Eliasson, P., Goetzendorf-Grabowski, T., Vos, J. B., Zhang, M., and Richardson, T. S., “Design of a
canard configured TransCruiser using CEASIOM,” Progress in Aerospace Sciences, Vol. 47, No. 8, 2011, pp.
695–705.
[295] Richardson, T. S., McFarlane, C., Isikveren, A., Badcock, K., and Da Ronch, A., “Analysis of conventional and
asymmetric aircraft configurations using CEASIOM,” Progress in Aerospace Sciences, Vol. 47, No. 8, 2011,
pp. 647–659.
[296] Wolkovitch, J., “The joined wing-An overview,” Journal of Aircraft, Vol. 23, No. 3, 1986, pp. 161–178.
[297] Kroo, I., Smith, S., and Gallman, J., “Aerodynamic and structural studies of joined-wing aircraft,” Journal of
aircraft, Vol. 28, No. 1, 1991, pp. 74–81.
[298] Samuels, M. F., “Structural weight comparison of a joined wing and a conventional wing,” Journal of Aircraft,
Vol. 19, No. 6, 1982, pp. 485–491.
241
BIBLIOGRAPHY
[299] Gallman, J. W., and Kroo, I. M., “Structural optimization for joined-wing synthesis,” Journal of Aircraft, Vol. 33,
No. 1, 1996, pp. 214–223.
[300] Stodieck, O., Cooper, J., Neild, S., Lowenberg, M., and Iorga, L., “Slender-Wing Beam Reduction Method for
Gradient-Based Aeroelastic Design Optimization,” AIAA Journal, Vol. 56, No. 11, 2018, pp. 4529–4545.
242