You are on page 1of 81

MASTER'S THESIS

Modeling and Experimental Evaluation of


a Tilt Wing UAV

Adrian Lindqvist
2015

Master of Science in Engineering Technology


Space Engineering

Luleå University of Technology


Department of Computer Science, Electrical and Space Engineering
Master Thesis

Modeling and Experimental


Evaluation of a Tilt Wing UAV

Adrian Lindqvist

Luleå University of Technology


Department of Computer Science, Electrical and Space Engineering
Division of Systems and Interaction

10th December 2014


A BSTRACT

The aim of this thesis is to design an optimal model for describing the behavior of a tilt
wing UAV during its transitional state from a helicopter to an airplane and to perform
an experimental evaluation of its validity. A tilt wing UAV can be described as a hybrid
between a helicopter and a plane, which has the capabilities of both aircrafts, meaning it
can stand still and hover, and by tilting its wing 90 degrees, act like a regular plane. This
would imply that the aircraft can have vertical take offs and landings, stand still and
hover in mid air, while being able to travel long distances efficiently. The novelty of this
thesis stems from: a) the simulations on the main wing during the transition between
helicopter and fixed-wing plane mode for angles between 0-90 degrees, b) comparison
between the model extracted from the simulations for the main wing and tests from a
wind tunnel and, c) reaching a sufficiently good design for a tilt wing UAV with, d) an
overall experimental evaluation.

ii
P REFACE

Almost since the start of this master thesis I felt like I would be done long before the
initial deadline. With time as I started digging more in how I were supposed to design
an aircraft the workload slowly increased, with more simulations that were needed to be
done, parameters that would need to be set, realising that some simulations would need
to be redone because of a change somewhere. I do not think I can tell how many times
that I said that ”now I have finally done my last simulation”, two days later I surely had
to do another one.

I would gladly like to thank my supervisor George Nikolakopoulos for his advices during
this time. His constructive feedback and the motivation to continue working, enlightening
me about the work I do can actually have an impact in the world of science. Encourag-
ing me to strive for more which have resulted in writing an article in this area that will
hopefully be accepted.

I would as well like to thank my supervisor and friend Emil Fresk who have always
believed in me even when the times were hard. He have helped explain so much to me,
evaluated my ideas and point out critical parts. I would surely not have been able to
complete this thesis without his help.

For supplies and access to equipment I would like to thank Luleåa University of Tech-
nology and the Department of Computer Science, Electrical and Space Engineering. For
access to the wind tunnel and feedback for simulations I would also like to thank the
Department of Engineering Sciences and Mathematics.

And thanks to you, the reader who is still reading this and might have an interest of
the work that I have conducted during this last year.

iii
A BBREVIATION L IST

v : Freestream velocity
c : Chord length
ν : Kinematic viscosity
µ : Dynamic viscosity
ρ : Density of fluid
Re : Reynolds number
AoA : Angle of attack
W CL : Wing Cubic Loading
AR : Aspect Ratio
b : Wing span
S : Main wing area
L : Lift force
D : Drag force
CL : Lift coefficient
CD : Drag coefficient
CG : Centre of Gravity
AC : Aerodynamic Center
lH : Length from CG to AC of horizontal stabilizer
cH : Chord of horizontal stabilizer
lV : Length from CG to AC of vertical stabilizer
dh : Wing span of horizontal stabilizer
AV : Area of vertical stabilizer
AH : Area of Horizontal stabilizer
NP : Neutral Point
RM SE : Root-Mean-Square-Error
N RM SE : Normalized-Root-Mean-Square-Error

iv
C ONTENTS
Chapter 1 – Introduction 4
1.1 Background to aircraft parts . . . . . . . . . . . . . . . . . . . . . . . . . 5

Chapter 2 – Design and requirements 8


2.1 Requirements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.2.1 Choosing an airfoil . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.2 Wing model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.3 Aspect ratio and Wing Cubic Loading . . . . . . . . . . . . . . . 14
2.2.4 Size of Aileron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.5 Dihedral angle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.2.6 Tail design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.2.7 Center of gravity . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.2.8 Summarize of design . . . . . . . . . . . . . . . . . . . . . . . . . 19

Chapter 3 – Simulations 21
3.1 Main wing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.1 Designmodeler . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.1.2 Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.1.3 FLUENT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.1.4 Wind tunnel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.1.5 Finding the lift model . . . . . . . . . . . . . . . . . . . . . . . . 36
3.1.6 Finding the drag model . . . . . . . . . . . . . . . . . . . . . . . 41
3.1.7 Simulation results . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.2 Aileron . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2.1 Designmodeler . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.2.2 Mesh . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
3.2.3 Fluent . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2.4 Simulation results and model of aileron . . . . . . . . . . . . . . . 50
3.3 Rudder . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

Chapter 4 – Experimental Setup 57


4.1 Equipment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.2 Building . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.1 Main wing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
4.2.2 Horizontal and vertical stabilizers . . . . . . . . . . . . . . . . . . 59
4.2.3 EDF clamp . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.2.4 Worm drive . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62
4.2.5 UAV body . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
4.3 Final design parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . 64

Chapter 5 – Experimental evaluation 65


5.1 Testing the UAV’s hovering characteristics . . . . . . . . . . . . . . . . . 65
5.1.1 Percentage of power while hovering . . . . . . . . . . . . . . . . . 66
5.1.2 Yaw behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.1.3 Pitch behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
5.1.4 Roll behavior . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

Chapter 6 – Conclusions 67

Chapter 7 – Future work 68

Appendix A – FindKneeLift.m 69

Appendix B – LiftEqn.m 70

Appendix C – LiftEqnDeriavative.m 71

vi
List of Figures

1.1 Overview of standard parts of an aircraft. . . . . . . . . . . . . . . . . . 6


1.2 Structure of a general airfoil. . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.1 Lift and drag coefficients at different Reynold numbers, highlighting the
Re numbers of importance, NACA 4412. Source: Airfoil section charac-
teristics as affected by variations of the Reynolds number [1]. . . . . . . . 11
2.2 From left to right: rectangular, tapered, elliptical. . . . . . . . . . . . . . 12
2.3 Stall characteristics of wing models. . . . . . . . . . . . . . . . . . . . . . 13
2.4 Aileron sizing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.5 Dihedral angle of wing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.6 Aircraft during a roll, low wing. . . . . . . . . . . . . . . . . . . . . . . . 16
2.7 Aircraft during a roll, high wing. . . . . . . . . . . . . . . . . . . . . . . 17
2.8 Overview of tail design with horizontal stabilizer. . . . . . . . . . . . . . 17
2.9 Overview of tail design with vertical stabilizer. . . . . . . . . . . . . . . . 18
2.10 Display distance between CG and NP. . . . . . . . . . . . . . . . . . . . 19
3.1 Overview of the fluid volume. . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2 Overview of the fluid volume. . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3 Edge Sizing around the back. . . . . . . . . . . . . . . . . . . . . . . . . 23
3.4 Edge Sizing around the back. . . . . . . . . . . . . . . . . . . . . . . . . 24
3.5 Edge Sizing for the front of the wing. . . . . . . . . . . . . . . . . . . . . 24
3.6 Edge Sizing around the wing. . . . . . . . . . . . . . . . . . . . . . . . . 25
3.7 Edge Sizing of the inlet. . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.8 Overview of the finished mesh. . . . . . . . . . . . . . . . . . . . . . . . . 27
3.9 Close up of the mesh at the front. . . . . . . . . . . . . . . . . . . . . . . 27
3.10 Close up of the mesh at the rear. . . . . . . . . . . . . . . . . . . . . . . 28
3.11 Process of solver for Fluent where all residuals have to pass the 10−6 limit,
which is the black horizontal line. . . . . . . . . . . . . . . . . . . . . . . 29
3.12 Velocity distribution at a 19 degree angle and a velocity of 18 m/s. . . . 29
3.13 Rotating the fluid volume to get a more correct view of the simulation. . 30
3.14 Black coordinate system representing the sought after force vectors for
drag and lift, while red is the solvers coordinate system. . . . . . . . . . . 30
3.15 Lift distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.16 Drag distribution. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

1
3.17 Overview of wing positioned in the wind tunnel, AoA is zero. . . . . . . . 33
3.18 Wing is tilted, creating the angle θ (AoA) and a difference between d1 and
d2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.19 Wing model that were used for the wind tunnel. . . . . . . . . . . . . . . 34
3.20 Wing model mounted in the wind tunnel. . . . . . . . . . . . . . . . . . . 35
3.21 Difference in lift force between the simulated values from ANSYS and the
wind tunnel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
3.22 Lift distribution at 18 m/s with improved number of data points between
0-20 degrees. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.23 Splitting the curve into three parts. . . . . . . . . . . . . . . . . . . . . . 37
3.24 Comparison between Lift of simulated and modeled values at 18 m/s. . . 39
3.25 A full comparison between simulated and modeled lift distribution. . . . 39
3.26 Error in lift force at 18 m/s. . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.27 3D plot of error in lift generated between simulated and model at 18 m/s. 40
3.28 Drag distribution at 18 m/s. . . . . . . . . . . . . . . . . . . . . . . . . . 42
3.29 Dividing the model of the drag. . . . . . . . . . . . . . . . . . . . . . . . 42
3.30 Comparison between Drag of simulated and modeled values at 18 m/s. . 44
3.31 A full comparison between simulated and modeled lift distribution. . . . 44
3.32 Error in drag force. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.33 3D plot of error in lift generated between simulated and model at 18 m/s. 45
3.34 Ratio of lift/drag at all angles. . . . . . . . . . . . . . . . . . . . . . . . . 46
3.35 Forces generated from lift and drag at all angles, at 18 m/s. . . . . . . . 47
3.36 Aileron at 15 degrees angle downwards. . . . . . . . . . . . . . . . . . . . 47
3.37 Overview of the finished mesh for the new airfoil. . . . . . . . . . . . . . 48
3.38 Generated lift from aileron at different angles. . . . . . . . . . . . . . . . 50
3.39 Error in lift force of aileron. . . . . . . . . . . . . . . . . . . . . . . . . . 51
3.40 Generated drag from aileron at different angles. . . . . . . . . . . . . . . 52
3.41 Error in drag force of aileron. . . . . . . . . . . . . . . . . . . . . . . . . 53
3.42 Generated yaw force from rudder at different angles. . . . . . . . . . . . . 54
3.43 Generated drag from rudder at different angles. . . . . . . . . . . . . . . 55
3.44 Generated drag from rudder at different angles. . . . . . . . . . . . . . . 55
3.45 Generated drag from rudder at different angles. . . . . . . . . . . . . . . 56
4.1 Insertion of mounting rod for aileron. . . . . . . . . . . . . . . . . . . . . 58
4.2 Aileron mounted with Highlighting of key parts for the aileron. . . . . . . 59
4.3 Highlighting of key parts for the horizontal stabilizer and elevator. . . . . 60
4.4 Highlighting of key parts for the vertical stabilizer and elevator. . . . . . 60
4.5 Tail section finalized. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.6 Highlighting of key parts for the EDF clamp. . . . . . . . . . . . . . . . . 62
4.7 Overview of the worm drive. . . . . . . . . . . . . . . . . . . . . . . . . . 63
4.8 Overview of the attached worm drive motor. . . . . . . . . . . . . . . . . 63
5.1 Coordinate system of UAV with origin in the center of the worm drive. . 65

2
List of Tables

2.1 WCL table . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14


3.1 Mesh results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.2 Evaluation of skewness . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
3.3 Simulations vs wind tunnel results . . . . . . . . . . . . . . . . . . . . . . 36
3.4 Lift coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
3.5 Drag coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.6 Mesh results for aileron . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.7 Aileron lift coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.8 Aileron drag coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.9 Rudder coefficients . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
4.1 Equipment for the UAV . . . . . . . . . . . . . . . . . . . . . . . . . . . 57

3
C HAPTER 1
Introduction

The area of Unmanned Aerial Vehicles (UAV) and especially the area of the one of having
Vertical Take-Off and Landing (VTOL) such as a quadrotor, has been in focus of research
and development efforts, mainly due to their efficiency in accomplishing complex missions
[2], [3].
A drawback with using quadrotors are that they are not efficient in covering great
distances and that their top speed are not near the velocities of UAVs that are designed
as a plane due to the fact that they have different working areas. By creating a hybrid
aircraft that can act both as a helicopter and as a plane it can take advantage of both their
superior sides, getting great hovering capabilities and VTOL and by tilting the wings be
able achieve high cruise velocities. There has not been much development in this kind of
area, only a few aircrafts have had these properties, LTV XC-142 and Canadair CL-84
are a few of the newest aircrafts with a tilt wing mechanic and they are from the years of
1964 and 1965. The tilt wing aircraft was replaced later on by tilt rotor aircrafts which
only tilted the rotors. In the area of tilt rotors there have been more progress and are
still developing [4], [5]. Though the tilt rotor aircraft will create a down wash on the wing
while hovering, lowering the lift efficiency and creating more stress on the wings. This is
where the tilt wing excel, when a tilt wing aircraft hovers there will be minimum down
wash on the wings because it will always be aligned with its engine. When the UAV
starts tilting the wings they will start generating lift and drag because there will be an
increase in speed at the same time. Another advantage that the tilt wing has compared
to the tilt rotor are that during the transition state when going from hover to plane mode
are that the wing will gradually increase in lift generated. An ordinary wing will stall in
the area of 15-20 degrees and the lift generated will drop, though at even higher angles
the generated lift will increase again which will be beneficial during the UAVs transitions
state. Even though the lift to drag ratio are low under the transition state, the wing will
still generate a fair amount of lift, meaning the UAV can use smaller engines and become
more fuel efficient as long as it can overcome the drag that will occur.
At the moment there has been a few works in the area of tilt wing aircrafts [6], [7], [8],

4
1.1. Background to aircraft parts 5

but none that especially handles how the transition model looks like with wind tunnel
effects and experimental data. Large aircrafts that would have a tilt wing mechanic
would suffer from a number of problems that a small lightweight UAV would not have.
Often large aircraft store their fuel in the wings, meaning they will be heavy to rotate
and there will also be a change in their center of gravity. With a UAV that have a battery
as a power supply which is positioned between the wings, the only weight that needs to
be rotated are the structure of the wing and the engine in itself, making UAVs highly
advantageous for this kind of mechanic. This will enable the UAV to fly as a plane when
the angle of attack is below 15 degrees and be able to cover great horizontal distances
and hover as a helicopter when the angle is 90 degrees in a single spot.
The proposed idea for this article is to create a tilt wing UAV, a hybrid between a
helicopter and a plane. To achieve this behavior a tilt wing mechanic will be needed
which will allow the wings to change the angle that they are in reference to the body
of the UAV. With an angle of 0 degrees the UAV would be able to cruise efficiently at
high speeds and by tilting the wings 90 degrees so that they are perpendicular to the
body of the UAV it would enable it to hover steadily. To achieve a steady transition a
model will be needed to accurately calculate the lift and drag generated which can change
drastically. The model shall be deviated from simulated runs and then be confirmed by
testing with a wind tunnel. An optimal design will then be based on the simulations to
produce a UAV with good properties both in the area for hovering but also have a high
lift to drag ratio when cruising as a plane. With the use of a tilt wing UAV it could
efficiently cover great areas as mentioned in [2] but also work in other areas where it
would need a more stationary position [9].

1.1 Background to aircraft parts


This will be a short introduction to the different parts of a regular aircraft, highlighting
the important areas which this thesis will handle.

Main wing: These can have different shapes, rectangular, tapered, ellipse and so on,
this will be the main contributor to both lift and drag that will be generated. Each wing
will also contain an aileron.

Aileron: These small ”flaps” are positioned at the end of the main wing, by turning
these up or down it will create a difference in lift and drag generated for each wing,
making the aircraft to start rolling.

Horizontal stabilizer: This small wing that is positioned at the tail of the aircraft
will create a small lift just to ensure that the tail of the aircraft does not drop. Each
small wing will contain an elevator.
1.1. Background to aircraft parts 6

Elevator: These will control the pitch of the aircraft, by changing the pitch the air-
craft can either start pointing up or down.

Vertical stabilizer: This is the fin of the tail. Ensuring that the aircraft does not
start slipping, i.e making sure it will fly straight. This stabilizer contain the rudder.

Rudder: By changing the angle of the rudder the aircraft can change its yaw accord-
ingly. This will ensure that the aircraft can fly in a straight line even though there might
be different drag forces on the main wing, trying to make the aircraft turn.

Figure 1.1: Overview of standard parts of an aircraft.

Another important aspect to mention is the wing profile, the airfoil. Depending on
different airfoil the aircraft can handle very different, gaining more lift at lower angles or
can have a higher angle before stalling, drag generated, lift to drag ratio and much more.
The shape of the airfoil also depends on what speeds the aircraft will be susceptible to,
subsonic or supersonic, there are even airfoils that are shaped to be optimal during the
transition to supersonic. Figure 1.2 below shows how an unsymmetrical airfoil is struc-
tured which this thesis will use and be discussed later on. In the early 1930s, NACA -
forerunner to NASA, started constructing airfoils rationally and systematically. Many of
these NACA airfoils are in common use even today. This thesis will handle an airfoil from
the NACA 4-digit series which is from the first family of airfoils that they developed,
1.1. Background to aircraft parts 7

more about this will be discussed later in the thesis.

Chord: The straight line connecting the leading edge and the trailing edge, giving the
precise distance between the front and rear of the wing.

Camber: The camber is the maximum distance between the mean camber line and
the chord line.

Thickness: Measures the distance between the upper and lower surfaces, also perpen-
dicular to the chord line.

Mean Camber: This is the points that is exactly halfway between the upper and
lower surface.

Example of a NACA 4-digit airfoil, NACA 4412.


Different numbers in the name change different properties of the airfoil.

1-digit, 4: This is the maximum camber in hundredths of chords, meaning maximum


camber is 0.04 · c.

2-digit, 4: Gives the location of the maximum camber along the chord from the leading
edge in tenths of chord, meaning maximum camber is located at 0.4 · c.

3 & 4-digit, 12: Gives the maximum thickness of the airfoil in hundredths of the chord,
which is 0.12 · c.

Figure 1.2: Structure of a general airfoil.


C HAPTER 2
Design and requirements

When designing a tiltrotor aircraft one will have several requirements that needs to be
fulfilled, some due to the mission and some in order to be able to fly efficiently. Usually
an aircraft have a very specific mission (transport, trainer, military, aerobatics) and from
there one can set up the requirements and then plan the design. The mission for this
thesis were to create a modeled transition between helicopter to plane mode and back.

2.1 Requirements
The specific requirements for the presented design have been the following ones:
• The UAV should be able to lift its own weight while hovering and also generate at
least the same amount when flying at an efficient angle of attack (AoA)
• Be able to tilt wings and rotors 90 degrees for switching between plane and heli-
copter mode
• Max weight of 6 kg (new regulations regarding aircrafts over 6 kg)
• The UAV shall have a top speed of 20 m/s
• The UAV shall have ailerons, elevators, rudder with horizontal and vertical stabi-
lizers to control roll, pitch and yaw
• Create a model which follows the transition between helicopter and plane mode
with change in AoA and velocity

2.2 Design
Due to the design need to both work for a helicopter mode and a plane, some consider-
ations need to be taken into thought, because one thing can work very well in one state

8
2.2. Design 9

but not work in the other. As of this this thesis will mostly focus on the design of the
plane When designing a plane, one have a myriad of choices on what to focus, often
there are parts that are interconnected to each other. For example, by increasing the
wing span one will also change the properties of (drag, lift, wing loading, aspect ratio,
hovering stability), but at the same time also lowering the maneuverability when flying
as an plane. The following points are the key aspects to think of when designing an
aircraft.
• Airfoil (lift vs drag)
• Wing model
• Aspect ratio and Wing cubic loading
• Size of aileron
• Dihedral angle
• Size of the vertical and horizontal stabilizers
• Size of rudder and elevator
• Center of gravity
• Large propellers to enable a good and strong hovering mode.
• Wide wingspan for the rotors to enable the aircraft to have good properties when
hovering.
For the last two items it is more of a directional move and something to aim for than
proven with fact or scientific background. The remaining items will all be discussed in
this thesis in the order as they are presented in the list.

2.2.1 Choosing an airfoil


The airfoil is the wing profile, mentioned earlier in section 1.1 and seen in Figure 1.2.
It controls at what AoA it will generate the most lift, how much lift and the properties
of how much drag will be induced. The first thing to do when choosing an airfoil is
to determine in what range of Reynold number the mission will be executed at. The
Reynolds number is a dimensionless quantity that help predict similar flow patterns in
different fluid flow systems. For an airfoil it is dependent on the fluid velocity when
traversing the wing, the chord of the wing and the kinematic viscosity of the fluid [10].

v∗c
Re = (2.1)
ν
µ
ν = (2.2)
ρ
2.2. Design 10

Where v is the freestream velocity (m · s−1 ), c is the chord length (m), ν the kinematic
viscosity (m2 · s−1 ), µ the dynamic viscosity (N s · m−2 ) and ρ is the density of the fluid
(kg · m−3 ).

With the mission mentioned earlier, the UAV was proposed to have a top speed of 20
m/s with the lowest speed of 2 m/s, this will be the freestream velocity for the Reynolds
number. The kinematic viscosity is calculated by using the properties of air at 273 K
from Table T-1.5 in [11],

ρair = 1.2929 kg · m−3 (2.3)

µair = 16.7 · 10−6 N s · m−1 (2.4)

With the constants of ρ and µ for air given, it is now possible to use Equation 2.2
which results in:

ν = 1.292 · 10−5 m2 · s−1 (2.5)

At the moment the chord length of the wing have to be approximated due to respect
of aspect ratio and wing loading. With an iterative process the chord length were put to
0.15 m, this result will be explained later. With the given values the Reynold numbers
will be:

Re2m/s = 0.2 · 105 (2.6)

Re20m/s = 2.3 · 105 (2.7)

By inspecting [1] one will get an overview of different standard airfoils at low Reynolds
number. An important factor here is that the airfoil has a good lift coefficient at really
low Reynolds number (20k-80k), but also has a low drag coefficient when increasing the
Reynold number. The result ended in choosing the NACA 4412 airfoil as it fulfilled the
criteria better than other documented profiles at low Reynolds number. An estimation
on how it performs can be observed in Figure 2.1.
2.2. Design 11

Figure 2.1: Lift and drag coefficients at different Reynold numbers, highlighting the Re
numbers of importance, NACA 4412. Source: Airfoil section characteristics as affected
by variations of the Reynolds number [1].

Inspecting Figure 2.1 one can determine at what AoA the airfoil will generate the
most lift with the given Reynolds number. The diagrams are shown in units of the lift
coefficient (CL ) and drag coefficient (CD ) which governs the equations of lift and drag.
L0
CL = (2.8)
q∞ c
D0
CD = (2.9)
q∞ c
1 2
q∞ = ρV (2.10)
2 ∞
L0 and D0 is the lift and drag per unit span (N/m), ρ is the density of the fluid, V∞ is
the freestream velocity and c is the chord of the airfoil. An early estimation of how much
lift this airfoil can produce is now possible. By extruding a value for the lift coefficient
2.2. Design 12

for Reynold number of 164k at an angle of 6 degrees and with previously estimation of
the chord, a temporary summarize can be done.

CL = 1.0 (2.11)

AoA = 6 degrees (2.12)

ρ = 1.2929 kg · m−3 (2.13)

V∞,Re=164k = 14 m · s−1 (2.14)

c = 0.15 m (2.15)
With the given values, Equation 2.8 can be determined and an early estimation of the
wingspan can be determined.

L0 = 19.0 N/m (2.16)


This result would point in the direction of having a wingspan around 1.6 m for the
UAV to retain its altitude at a speed of 14 m/s. This will later on be confirmed with
simulations but for now its an early approximation on the size of the UAV.

2.2.2 Wing model


Picking the right wing model can change the performance and stability of the aircraft
drastically. Some standard wing models are the rectangular, tapered and elliptical pic-
tured in Figure 2.2. The wings different advantages and drawbacks will be discussed to
evaluate which would suit best on the UAV.

Figure 2.2: From left to right: rectangular, tapered, elliptical.

This thesis is based on using the rectangular wing model due to the fact that it is
easy to work with and its drawbacks (that will be discussed later) for this mission are
2.2. Design 13

not critical. Though to increase the performance (i.e better lifting properties) from the
UAV one should consider using either swept back tapered or elliptical wings to reduce
the drag. Figure 2.3 displays how the stall develops over each wing.

Figure 2.3: Stall characteristics of wing models.

A) Rectangular wing model.


This is a wing widely used on slower aircrafts as it induce more drag than other wing
models. When the stall start developing over the wing, the aircraft will have the benefit
over others that it can still use the ailerons to make maneuvers. The drawback in this
case is that the wing will loose much of its lifting force because of the stalling occur in
the center, which can lead to a drastic fall of the aircraft, this can be countered by either
drop the nose to increase the flow of air over the wings. The other option would be to
increase the angle of attack of the wings/motors to convert the airplane to an helicopter,
so that the rotors will generate the main lifting force.

B) Tapered wing model.


A tapered wing is more utilized in faster aircrafts, often in a combination with a swept
back form, creating more lifting force close to the body of the aircraft. The tapered wing
is a compromise between the rectangular and elliptical wing, decreasing the complexity
of the elliptical wing but significantly increasing the lifting efficiency of the rectangular.
Other than that, it also has some beneficial stall characteristics, compared to the rect-
angular. Instead of the development of the stall start from the center of the aircraft it
will begin closer to the tip of the wing, meaning that the rest of the wing can generate
lift. Drawback here is that it will loose much of its maneuverability because the ailerons
will be in the stall area and wont respond correctly any more.

C) Elliptical wing model.


The elliptical wing will perform best of these three wing models, though it is also more
2.2. Design 14

complex to work with, demanding more effort when making a 3D-model, simulating and
manufacturing. Because of this, most manufacturers use the tapered wing model.

2.2.3 Aspect ratio and Wing Cubic Loading


When designing a UAV that will work as a plane, it can in some areas be compared with
RC-planes, this means that the definition called Wing Cubic Loading (WCL) [12, 13] can
be used, which is defined as:
mass in gram m
W CL = 1.5 = (2.17)
(Wing area in dm) S 1.5

The WCL value will give the pilot an assumption on how the plane will perform in the
air. Table 2.1 show each type of plane with their general WCL.

Table 2.1: WCL table


Type WCL
Gliders under 4
Trainers 6
Aerobatics 9
Scale 10 to 15
Racers 12 and over

For those that are unfamiliar with how these type of planes handle, one can say that
gliders are easy and racers and above are hard to handle. Because the WCL is dependent
on the wing area, it will decide how much lifting force the plane would lift it it were
decided now, which is impossible. Though an estimation can be made, given that the
UAV will weigh in around 3000 g, with a wing span of 1.6 m will result in a wing area of
24 dm2 , applying this to Equation 2.17.

3000
W CLU AV = = 25.5 (2.18)
241.5
This would put the UAV in the class of racers and above. A key point to note here are
that rc-planes usually are very light weight, around 1 kg. Rather than the UAV will be
hard to handle which would be indicated by Table 2.1 it will probably feel very heavy
when flying, creating a slow movement and response from the control. Defining ’hard
to handle’ even further for the aerobatics and racers classes at least are that they are
unstable as per design to create high maneuverability, to achieve this state these planes
often have very short and stubby wings which will increase the WCL.
2.2. Design 15

With respect to the WCL, another variable to consider as well is the aspect ratio,
defined as:
b2 S2 1 S
AR = = 2 · = 2 (2.19)
S c S c
Which is the ratio between the wing area and its chord, a high value indicates long and
narrow wings, whereas a low aspect ratio indicates short and stubby wings. A typical
glider will have a high aspect ratio (>15), while an aerobatic plane might have a ratio
of 5. Using the estimated values so far that has been made for the wing area and the
chord, applying these to Equation 2.19 will result in the UAVs aspect ratio.
24
AR = = 10.7 (2.20)
1.52
This value indicate that the plane will maintain good maneuverability and still have some
good lifting properties of a glider. As this mission is not dependent of having a high ma-
neuverability one can safely increase the wing span to gain even more lifting force, which
means that the UAV can fly at a lower velocity and therefore save some power on the
battery.

2.2.4 Size of Aileron


The ailerons control the roll-rate of the plane. By standard, aerobatic planes use larger
area because the need of high roll rates, whereas gliders use relatively small and therefore
create slow movements. The general size of ailerons on small airplanes are around 1/4 of
the chord and 1/8 of the wingspan in total, which is what will be used and is presented
in Figure 2.4.

Figure 2.4: Aileron sizing.


2.2. Design 16

2.2.5 Dihedral angle


When an aircraft fly in a straight horizontal line there will be disturbance from air gusts
and its like, such disturbances will cause the aircraft to diverge from its original path.
With an aircraft that has a mounted low wing it is possible to avert this process by
increasing the angle of the wing relative to the aircraft body, Figure 2.5.

Figure 2.5: Dihedral angle of wing.

During a roll or a strong gust that have positioned the aircraft like in Figure 2.6 it
will naturally try and get back to its original state in Figure 2.5. This is because the
left wing 2.6 will have a higher AoA than the right wing and therefore will generate less
lift compared to it. Hence the right will rise, while the left wing will drop down and the
aircraft will return to its neutral state.

Figure 2.6: Aircraft during a roll, low wing.

Usually the dihedral angle is very small, around 3-4 degrees, this will create stability
at the expense of lift. As it is only the vertical component that generate lift, it will be
relative to the cosine of the angle.

Lloss = cos(0) − cos(θ) (2.21)

By using a dihedral angle of 3 degrees for an aircraft, there will be a loss of 0.13% of
the lift. Because the UAV in this thesis will have a VTOL it is advisable to have the
lifting force when hovering above the center of mass than below. Instead a high wing
design will be used. Mounting a high wing on the aircraft will as well create a lateral
2.2. Design 17

stability as the low wing did. When the aircraft is banking, or because of a side wing
has hit the aircraft, it will naturally go back to its primary state. Due to the air flowing
laterally around the body in a side slip, the AoA of the low wing will be higher than that
of the high wing. This will cause a restoring rolling moment, Figure 2.7.

Figure 2.7: Aircraft during a roll, high wing.

2.2.6 Tail design


The tail design is built up by two parts, its horizontal- and vertical stabilizer. At the
moment it would be too early to discuss the tail design because it is highly dependent
on the area of the main wing and where the center of gravity is located.

Figure 2.8: Overview of tail design with horizontal stabilizer.


2.2. Design 18

Figure 2.9: Overview of tail design with vertical stabilizer.

lH AH
VH = (2.22)
cS
lV AV
VV = (2.23)
bS
By applying the estimated values of S, c and b the unknown variables would be lH , AV .
Given table [6.4] in [14] and the chapter of tail design in [15] gives that the value for VH
and VV should be in the ranges of 0.5 − 1.0 and 0.04 − 0.08 to create a stable aircraft.
This would mean that it is possible to balance lH , lV , d, CH and y almost freely to create
a tail that will work well to the given aircraft as long as it is in the specific range.

Elevator and rudder


To enable the aircraft to elevate and yaw it is needed to install elevators and a rudder.
The elevator as it sounds will change the inclination of the aircraft, pitching it to go up
or down while the rudder will turn it slowly left or right. A rule of thumb here is to
create an elevator that at least have an area of 1/4 of the horizontal stabilizers chord and
the rudder should be around 2/5 of the vertical stabilizers.

2.2.7 Center of gravity


The UAV will have a relatively light weight, therefore its main weight will come from the
battery that it will use as a main power supply. Because of this, the Center of Gravity
(CG) will be almost at the same point where the battery is located. While the UAV
is hovering the CG should be between the front propellers and the EDF to get a good
stability. Though this configuration is not desired as much while flying as a plane. More
often then not it is advisable to have the CG in front of the main lifting force, in this
case in front of the aerodynamic center of the main wing. Another case to inspect are the
distance between the CG and the Neutral Point (NP), which is the aerodynamic center
of the whole aircraft. The aircraft will be stable when the CG is ahead of the NP and
2.2. Design 19

would become unstable if the CG is behind the NP. The distance between CG and NP
is defined as λ, presented in Figure 2.10 and Equation 2.24

Figure 2.10: Display distance between CG and NP.

AH
λ = X (2.24)
S + AH
Where X is the distance between the main lifting point of the main wing and the
horizontal stabilizer (m), S as mentioned before are the main wing area (m2 ) and AH is
the area of the horizontal stabilizer (m2 ).

2.2.8 Summarize of design


Estimated values so far:

• Wing span : 16 dm

• Wing chord : 1.5 dm

• Total mass : 3000 g

• Aileron size : 1/8 of wing span and 1/4 of wing chord

Obtained values so far:

• Highest lift coefficient : ∼ 1.2

• WCL : 25.5
2.2. Design 20

• AR : 10.7

Selected properties:

• Airfoil : NACA 4412

• Wing model : Rectangular

• Top speed : 20 m/s

• High wing design

Some properties can yet not be decided, like the horizontal- and vertical stabilizers due
to they are being dependent on the size of the main wing. This will be decided after the
simulations are completed.

The following chapter will handle the setup for how to use a correct model in ANSYS
which is the program that performs the simulations. The parts that will be simulated
are the main wing, aileron, elevator and the rudder. Each part will have its own specific
interval of angles that it will be simulated on, but every part will be subject to the
velocities between 2 m/s and 20 m/s. After the simulations are done for the main wing,
a complementary test will be performed on it in a wind tunnel to compare with the
simulations to verify its authenticity. The chapter will also go into finding the model for
the transition between helicopter and plane modes for the main wing and the effect of
changing the ailerons,elevators and rudder.
C HAPTER 3
Simulations

The simulations were done in ANSYS, using the FLUENT tool which uses three parts;
designmodeler, mesh and setup/solver. The steps done in this chapter were a combination
of the online instructions from references [16, 17].

3.1 Main wing


3.1.1 Designmodeler
Using the NACA 4 Digits Series Profile Generator [18], a set of data values will be given
that can be imported to Designmodeler which will represent the specified wing profile.
Because the wing will be similar over the whole wing span it is possible to model it in
2D, which will save a lot of time when simulating, this means that the z-coordinate which
represents the depth will be put to zero. By using this method, the wing would get the
properties being infinite, the program will express the final result in N/m and therefore
the wing span can easily be extended or shortened to the length that is sought after.
The data points gathered from the generator will need to be multiplied with 0.15 dm
to get the desired length presented in meters. By importing the data points from the
generator to Designmodeler one will get the correct model of the airfoil. The second task
will be to create a fluid volume where the air will flow. This volume has to be large
enough so that the air can flow freely around the wing and not bounce against walls and
back to the wing, which would give inaccurate data.

21
3.1. Main wing 22

Figure 3.1: Overview of the fluid volume.

The fluid volume that were created was a half circle with a radius of 1.5 m and a
rectangle of 1.5*3 m, about 10 times as big as the airfoil. The center of the volume will
be positioned at the tail of the wing. By positioning the wing in the half circle one can
change the angle at which the air will hit the wing, between −90 to +90 degrees.

3.1.2 Mesh
There are several ways to create a good mesh, this thesis will handle how to set up a
structured mesh with good mesh values.
First thing to do is to enable Mapped Face Meshing for the four visible faces of the fluid
volume from Figure 3.1. Now each face will have their own mesh properties and only
connect at the borders to the other faces, this mean that an increase of the intensity of
the mesh grid for specific areas can be done, this is done by applying Edge Sizing.

Figure 3.2: Overview of the fluid volume.


3.1. Main wing 23

The intensity of the Edge Sizing depends on several options, as it has been indicated
in the left of Figure 3.2. By defining an Edge Sizing it will create a structured mesh but
also allows the user to tweak the mesh to its liking, increasing or decreasing the density
of the mesh in specific areas. The options that were used in this thesis are as follows:

Number of divisions are how many nodes are to be used.


Behaviour handles how much the nodes are ”set in stone”, if they can move around
some while meshing or if they are locked to a given position.
Bias Type/Factor is in what interval the nodes should be placed, a high value indicate
a higher density of nodes at the end of the line which depends on the bias type.

Following figures are the setup that were used to create a structured mesh.

Figure 3.3: Edge Sizing around the back.

Figure 3.3 and 3.4 will handle the outlet of the mesh mostly, some of it will also handle
how the mesh will behave close to the end of the wing. These areas will have relatively
small amount of nodes as these are not as important as the front and middle section of
the wing are where the most significant lift are generated.
Figure 3.5, 3.6 and 3.7 will be more important. This is due to the fact that the fluid
that hit the wing will need to have a high density of layers to enable good and accurate
data in the end.
3.1. Main wing 24

Figure 3.4: Edge Sizing around the back.

Figure 3.5: Edge Sizing for the front of the wing.


3.1. Main wing 25

Figure 3.6: Edge Sizing around the wing.

Figure 3.7: Edge Sizing of the inlet.

The results of the mesh can be seen in table 3.1, with Figures 3.8, 3.9 and 3.10.
3.1. Main wing 26

Table 3.1: Mesh results


Nodes 50666
Orthogonal Quality (min) 0.325
Orthogonal Quality (max) 0.999
Orthogonal Quality (average) 0.985
Skewness (min) 5 · 10−6
Skewness (max) 0.78
Skewness (average) 0.05

There are some requirements that need to be met to achieve a mesh that is good and
will present accurate results in the end, for this thesis there are two things that will be
of importance to check.
The first thing is the orthogonal quality of the mesh, it is a sum of how good the mesh
is in overall, it should at all points in the mesh have a value above 0.05 to at least be
acceptable.
The second rule is the skewness of each element. The closer the element is to zero the
better. Quads should have a maximum skewness of 0.85. Elements with a value above
0.95 is to be considered as a sliver, which should be avoided in the mesh, [19].

Table 3.2: Evaluation of skewness


Quality Value of Skewness
Excellent 0-0.25
Good 0.25-0.50
Acceptable 0.50-0.80
Poor 0.80-0.95
Sliver 0.95-0.99
Degenerate 0.99-1.00
3.1. Main wing 27

Figure 3.8: Overview of the finished mesh.

As mentioned earlier, it is good to create a high density of layers close to the wing,
often refereed to as boundary layer. Critical points are the front and rear of the wing as
it has been indicated in Figure 3.9 and 3.10. The first boundary layer around the wing
is about 1 mm or less, which will give a fairly accurate result for the velocities that will
be simulated for.

Figure 3.9: Close up of the mesh at the front.


3.1. Main wing 28

Figure 3.10: Close up of the mesh at the rear.

3.1.3 FLUENT
With the mesh finished, it is now possible to import it to FLUENT which will handle
the simulations of flow.
The configuration used were a pressure based system, and more often than not using the
transient time option than steady state, steady state were used for approximately 15 % of
the simulations that were done. With regard to article [20] where different solver models
were discussed, the Spalart-Allmaras (SA) equation were chosen. It is a one-equation
model which solves an empirical transport equation for the eddy viscosity. This equation
handles the production, transport, diffusion and destruction of the turbulent kinematic
viscosity. An advantage here against other turbulence models is its simplicity, that it only
contains of one equation which in turn will increase the rate of simulations to be solved.
SA is often used as a first try model, in the meaning it will give a fairly good result
but if a more precise result are sought after one should consider using the Shear Stress
Transport (SST) model or other similar models which can handle turbulence better.
The simulations have been done by changing the angle of which the air will be streaming
from rather than changing the angle of the wing. This helped in reducing the workload
of doing a new mesh for each new angle. The work to be done is simulating for a velocity
between 2-20 m/s with a step increase of 2 m/s at a time, over the angles of 0-85 degrees
with a step increase of 5 degrees. The result will be a grid of 10*18 data points, or 180
simulations. The residual criteria value were set to 10−5 .
The following values would often change, depending on at what angle and at what ve-
locity that would be simulated, but this is the general values that were used during most
of the simulations.

Time Step Size: 0.01 s.


Number of Time Steps: 60.
Max Iterations/Time Step: 50.

For the most part if the solver could not reach a convergence, improving/decreasing
the max iterations per time step often solve the problem, else an increase in the number
3.1. Main wing 29

of time steps can be a solution. If none of these work, a solution could be to change the
time step size and then repeat the process. As it is 180 simulations and there are bound
to be numerous of cases that do not converge, this is a very time consuming part.

Figure 3.11: Process of solver for Fluent where all residuals have to pass the 10−6 limit,
which is the black horizontal line.

Because the inlet air will be released at an angle, it is important to remember that
the lift force is no longer straight up as in Figure 3.12. The wing detailed is subjected
to an inlet air at a velocity of 18 m/s, and angle of 19 degrees. This would result in a
x-component of 17 m/s and y-component 5.9 m/s.

Figure 3.12: Velocity distribution at a 19 degree angle and a velocity of 18 m/s.


3.1. Main wing 30

By rotating the fluid volume in the top part of Figure 3.13, a representation of how it
actually would be in reality is displayed.

Figure 3.13: Rotating the fluid volume to get a more correct view of the simulation.

Figure 3.14: Black coordinate system representing the sought after force vectors for drag
and lift, while red is the solvers coordinate system.
3.1. Main wing 31

Figure 3.14 shows the new coordinate system that would present the real values. The
force vector for drag and lift would result in:

   
cos α − sin α cos α sin α
FA = FW ⇒ FW = FA (3.1)
sin α cos α − sin α cos α

 
Drag = FW,drag = cos(α) sin(α) F A (3.2)

 
Lif t = FW,lif t = −sin(α) cos(α) F A (3.3)

Were FA is the force that ANSYS see and FW is the force that reflects the reality in this
case.
In total there were 23 simulations that could never converge, these were at high AoA
with low velocities. These points had to be interpolated from existing known data points.
By combining all the simulations that were done into a full 3D plot in Matlab one can
get an overview of how the lift force is distributed over the span of 0-90 degrees and 2-20
m/s, Figure 3.15.

50

40
Lifting Force (N/m)

30

20
20
10

10
0
100 80 60 40 20 0 Speed (m/s)
0
Angle of Attack (Degrees)

Figure 3.15: Lift distribution

The same were done for the drag, Figure 3.16.


3.1. Main wing 32

80

60

Drag Force (N/m)


40

20

0
100
20
50 15
10
5
Angle of Attack (Degrees) 0 0
Speed (m/s)

Figure 3.16: Drag distribution.

3.1.4 Wind tunnel


The wind tunnel that were used was a AF100 Subsonic Wind Tunnel with a section
volume of 600x305x305 mm (w · d · h). A smaller wing were created with a chord of 150
mm and a span of 300 mm. By using the whole depth of the tunnel, it will create the
same behavior as in the simulations, acting like it were in 2D.
The wind tunnel can handle velocities between 0-36 m/s, which are measured by a pitot
tube which in turn gives an output value in kPa. This is not as accurate as desired but
it will give a fair estimate on the real values. The velocity V is then calculated by using
the pressure difference 4 p and the density ρ in the room.

s
4p
V = 2 (3.4)
ρ

P
ρ = (3.5)
RT
R = 287.05 J/(kgK) (3.6)

The pressure P and temperature T in the room were measured by an analogue barom-
eter and R is the universal gas constant.

p = 1014 mbar (3.7)

T = 294, 15 K (3.8)

ρ = 1.2009 kg/m3 (3.9)


3.1. Main wing 33

With the given density of the room, the velocity of the wind in the tunnel can be
calculated.
The AoA for the wing are calculated by the distance from the bottom of the tunnel to
the front and rear of the wing, as indicated in Figure 3.17. The AoA will then be zero
when the distances (d1, d2) are of the same length.

Figure 3.17: Overview of wing positioned in the wind tunnel, AoA is zero.

By tilting the wing by θ degrees, as displayed in Figure 3.18, there will be a difference
in length between the front and rear of the wing to the floor.
This angle can be calculated by using the difference in length and divide it by the chord
of the wing, Equation 3.10.

   
Diff d1 - d2
θ = arcsin = arcsin (3.10)
Chord c

To use the wind tunnel a model of the real wing had to be created. This was done using
a 3D printer that printed the wing with plastic as material. As the wing were to big to
print as a whole part it had to be done in three separate sections of 10 cm each. Each
section had an infill of 10 percent, meaning they contain 90 percent air and will therefore
be very light. The sections were then mounted on a carbon fiber tube and glued together
using epoxy with the final result in Figure 3.19.
3.1. Main wing 34

Figure 3.18: Wing is tilted, creating the angle θ (AoA) and a difference between d1 and
d2 .

Figure 3.19: Wing model that were used for the wind tunnel.

The wing were then mounted in the wind tunnel section as seen in Figure 3.20 and con-
figured according to the procedure mentioned with Figure 3.17, Figure 3.18 and Equation
3.10.
3.1. Main wing 35

Figure 3.20: Wing model mounted in the wind tunnel.

Comparison between wind tunnel data and simulations


The results between the simulated values from ANSYS and the wind tunnel are visually
presented in Table ?? and Figure 3.21. A key note to notice is the increase in difference
of lift when the angle is increasing after an AoA of 10 degrees, also note that the values
of the wind tunnel will return a higher lift force at 15 degrees than the simulated. A
possible explanation behind this can be due to the small size of the flow section where
the wing is placed, meaning that the air can stabilize and create a false upward push on
the wing. By increasing the AoA more resulted in even higher differences and giving in
the end non-realistic lift force values.

30
14 m/s
16 m/s
18 m/s
25 20 m/s
Difference in percent (%)

20

15

10

5
0 5 10 15
Angle of attack (θ)

Figure 3.21: Difference in lift force between the simulated values from ANSYS and the
wind tunnel.
3.1. Main wing 36

Table 3.3: Simulations vs wind tunnel results


AoA, 14 m/s Simulated (N/m) Wind tunnel (N/m) Difference (%)
0 6.25 4.83 23
5 14.8 12.2 18
10 21.2 19.7 7.1
15 19.9 23 16

AoA, 16 m/s
0 8.29 6.17 26
5 19.4 15.4 21
10 27.9 26.2 6.1
15 26.2 29.8 14

AoA, 18 m/s
0 10.6 8.07 24
5 24.7 19.7 20
10 35.6 33.3 6.5
15 33.5 36.9 10

AoA, 20 m/s
0 13.2 9.73 26
5 30.7 25.3 18
10 44.2 40.7 7.9
15 41.8 46.3 11

3.1.5 Finding the lift model


By extruding a slice from the 3D image of lift, a 2D model which have constant speed
and varies only with the AoA from 0 to 90 degrees could be created. Figure 3.22 gives
a more detailed overview what happens at the different angles, whereof a critical area
can be seen between the angles 15 and 20 degrees where there are big changes in the lift
generated. This is due to the wing will stall, generating less lift and creating a lot more
drag which can be seen in Figure 3.28.
3.1. Main wing 37

40

35

30

Lift (N/m) 25

20

15

10

0
−10 0 10 20 30 40 50 60 70 80 90
Angle of Attack (Degrees)

Figure 3.22: Lift distribution at 18 m/s with improved number of data points between
0-20 degrees.

To create this smooth curve additional simulations were used, simulating the lift force
at a 1 degree step between 0-20 degrees. The goal here is to find the model used to
express this curve. First step is to split the curve into three parts and find each separate
solution for that specific part, Figure 3.23.

Figure 3.23: Splitting the curve into three parts.


3.1. Main wing 38

Inspecting Figure 3.23 shows that there is a ”knee” at 18 degrees, which will be im-
portant later.

1) For the first equation, it will use all the values up to 17 degrees and plot these with
the curve fitting tool (cftool) in Matlab, which leads to a polynomial equation of order
3.
3) Skipping equation 2 for the moment and go directly to equation 3. Same as earlier
but using the values at the angles between 20 and 90 degrees instead, which leads to a
polynomial equation of order two.
2) With equation 1 and 3 now solved, equation two can now be handled. To create a
smooth transition between equation 1 and 3, a polynomial of order 3 will be used. Using
the Matlab code found in Appendices A, B and C will return four constants that will be
used for the second equation. The final result for the model of the lift distribution are
presented with Equations 3.11 and 3.12.

L(α, V ) = CL (α) · V 2 (3.11)


3 2
A1 α + B1 α + C1 α + D1 , α ≤ 17

CL (α) = A2 α3 + B2 α2 + C2 α + D2 , 17 < α < 19 (3.12)

A3 α2 + B3 α + C3 , α ≥ 19

Table 3.4: Lift coefficients


Prefix\Constant A B C D
1 −5.772 0.6895 0.4923 0.0326
2 −28.89 37.22 −15.08 2.041
3 −0.1096 0.1476 0.0428

By comparing the result of the lift Equation 3.12 to the simulated values from Figure
3.22, it can be seen that the modeled lift equation fits very good to the simulated values.
By subtracting the difference between the simulated values and the model, a graph for
the error in lift force is displayed with Figure 3.26.
3.1. Main wing 39

Lift
40
Simulated
35 Modeled

30

25
Lift (N/m)

20

15

10

−5
−10 0 10 20 30 40 50 60 70 80 90
Angle of Attack (Degrees)

Figure 3.24: Comparison between Lift of simulated and modeled values at 18 m/s.

Verifying first that the model is correct for the whole area of interest, a comparison
between the simulated and model are done in a 3D plot, as displayed in Figure 3.25. As
the figures look very much alike, a more precise inspection can be done again by looking
into how big the error between the simulations and model. This is done by the same
method as earlier and are displayed in Figure 3.27.

3D Modeled lift plot

50 50

40 40
Lifting Force (N/m)

Drag Force (N/m)

30 30

20 20
20
10 20
10
15
10
0 10
0
100 80 100 5
60 40 80 60
20 0 Speed (m/s) 40 20 0
0 0 Speed (m/s)
Angle of Attack (Degrees) Angle of Attack (Degrees)

Figure 3.25: A full comparison between simulated and modeled lift distribution.

Error estimation
An error estimation gives the user an indication of how well the model actually is. As
what is acceptable are as well up to the user, as it can change depending on how important
3.1. Main wing 40

it that the correct value is given. In this Thesis an error of 5 % will be accepted, due to
the uncertainty of how well simulations actually reflect reality.

0.6

0.4

0.2
Force (N/m)

−0.2

−0.4

−0.6

−0.8
0 10 20 30 40 50 60 70 80 90
Angle of Attack (Degrees)

Figure 3.26: Error in lift force at 18 m/s.

1.5
Error in Lift Force (N/m)

0.5

0
100
20
50 15
10
5
Angle of Attack (Degrees) 0 0
Speed (m/s)

Figure 3.27: 3D plot of error in lift generated between simulated and model at 18 m/s.

To evaluate how well the result actually is in Figure 3.26, an evaluation with the use of
to the Root-Mean-Square-Error (RMSE) has to be done. This will decide of how good
3.1. Main wing 41

the model is in relation to the simulations.

r Pn
t=1 (ym,t − ys,t )2
RM SE = (3.13)
n
Equation 3.13 takes the lift value at a given angle for the model and subtract with the
simulated value at the same angle, and then divide by the number of points used. This
will give an estimate of the mean error for the model. Another useful point to observe
that is important is the error in a percentage, the Normalized-Root-Mean-Square-Error
(NRMSE), where a lower values indicate less residual variance.

RM SE
N RM SE = (3.14)
ymax − ymin
Where ymax and ymin are the highest and lowest observed values in the range used.
First calculate RMSE from Equation 3.13. This will result in an expression that is
given in N/m, applying this value to Equation 3.14 will hence lead to an expression in
percentage.

RM SELif t = 0.2419N/m

N RM SELif t = 0.69%

This mean that the general error in the model will deviate about 0.2097 N/m at all angles
from the simulated values, or ∼ 0.6 % from what it should be when flying at 18 m/s.
Keep in mind that this is only simulated values and therefore may deviate more from an
actual flight in real life.

3.1.6 Finding the drag model


The procedure for determining the model of drag will be very similar to that of the model
for lift. By inspecting Figure 3.28 there is a knee located at 18 degrees again.
3.1. Main wing 42

Drag
60

50

40
Drag (N/m)

30

20

10

0
0 10 20 30 40 50 60 70 80 90
Angle of Attack (Degrees)

Figure 3.28: Drag distribution at 18 m/s.

By using the same method as with the lift model, start with dividing the curve into
three parts as in Figure 3.29 which will be the three equations; with the knee be the
equation that will be solved last.

Figure 3.29: Dividing the model of the drag.

1) Solve equations 1 by using the drag values of θ < 17 and import this to cftool. The
best approximation that was found were an exponential function.
3.1. Main wing 43

3) Equation 3 will hold the interval of angles θ > 20. Again by using cftool, the solution
were found to be a polynomial of order 3.
2) The second equation (at the knee) will also be a polynomial of order 3. Switching the
equations in the previously mentioned Matlab code in Appendices A, B and C to those of
drag will generate the constants for the second equation. The final result for the model
for the drag distribution are presented with Equations 3.15 and 3.16.

D(α, V ) = CD (α) · V 2 (3.15)


B1 α
A1 e
 + C1 eD1 α , α ≤ 17
3 2
CD (α) = A2 α + B2 α + C2 α + D2 , 17 < α < 19 (3.16)

A3 α3 + B3 α2 + C3 α + D3 , α ≥ 19

Table 3.5: Drag coefficients


Prefix\Constant A B C D
1 0.0016 0.0022 9.173 · 10−5 0.0526
2 −55.57 51.10 −15.4 1.539
3 −0.0745 0.1662 0.0226 0.0015

The result of the model compared to the simulated can be seen in Figures 3.30 and
3.31.
3.1. Main wing 44

Drag
60
Simulated
Modeled
50

40
Drag (N/m)

30

20

10

0
0 10 20 30 40 50 60 70 80 90
Angle of Attack (Degrees)

Figure 3.30: Comparison between Drag of simulated and modeled values at 18 m/s.

3D Modeled Drag plot

80

80
60
Drag Force (N/m)

60
Drag Force (N/m)

40

40
20

20
0
100 20

20 0 15
100
50 15 80 10
10 60
40 5
5 20
Angle of Attack (Degrees) 0 0 0 0 Speed (m/s)
Speed (m/s) Angle of Attack (Degrees)

Figure 3.31: A full comparison between simulated and modeled lift distribution.

Error estimation
Start with calculating the RMSE and NRMSE for drag with the same method as were
done for lift to get an evaluation of the model.

RM SEDrag = 0.1228N/m

N RM SEDrag = 0.24%
3.1. Main wing 45

With an error percentage that is ∼ 0.25 % it is clear to say that it is an acceptable model
of the simulations. The error in drag force can be viewed in Figures 3.32 and 3.33. Figure
3.33 also verifies the low value of NRMSE for the whole area of interest. This indicate
that the model fit very well even at other velocities than it were extracted from.

0.3

0.2

0.1

0
Force (N/m)

−0.1

−0.2

−0.3

−0.4

−0.5

−0.6
0 10 20 30 40 50 60 70 80 90
Angle of Attack (Degrees)

Figure 3.32: Error in drag force.

1.5
Error in Drag Force (N/m)

0.5

0
100
20
50 15
10
5
Angle of Attack (Degrees) 0 0
Speed (m/s)

Figure 3.33: 3D plot of error in lift generated between simulated and model at 18 m/s.
3.1. Main wing 46

3.1.7 Simulation results


By knowing both the lift and drag distribution a comparison plot in which area to fly
most efficiently can be created. Figure 3.34 displays the ratio between lift and drag,
because lift and drag both are functions of V 2 only a 2D plot is necessary as it would
look the same when increasing the velocity in the 3D plot. With this it is now possible
to locate at what angles are most efficient to fly at, i.e gaining most lift force at the cost
of drag, which is at ∼ 7 degrees.

Lift to drag ratio


40

35

30

25
Lift/Drag ratio

20

15

10

0
0 10 20 30 40 50 60 70 80 90
Angle of Attack (Degrees)

Figure 3.34: Ratio of lift/drag at all angles.

It is also of importance to see how the lift and drag relate to each other, as displayed
in Figure 3.35. Even tough the lift/drag ratio is very poor at all angles except for 4-10
degrees it does not point out how great the forces are. By inspecting Figure 3.24 it can
be seen that the lift force generated can be equal over different set of angles. An example
of this is that the lift generated at 5 degrees are equal to that of when the AoA is 60
degrees. With the scenario of flying at 18 m/s these angles would have generated about
25 N/m, though it would not have been suitable to fly with an AoA of 60 degrees because
of the high drag generated which would most probably tear the wings off the UAV. The
forces from the lift and drag generated are displayed in Figure 3.35.
The most energy effective way for the UAV, when going from helicopter to plane mode, is
to keep the velocity low while decreasing the AoA, this method will decrease the impact
of drag upon the UAV. Though this process has to be carefully performed because a
reminder must be done to always generate enough lift to ensure the UAV will not drop
drastically in height or stall out.
3.2. Aileron 47

60
Lift
Drag
50

40

Force (N/m)
30

20

10

0
0 10 20 30 40 50 60 70 80 90
Angle of Attack (Degrees)

Figure 3.35: Forces generated from lift and drag at all angles, at 18 m/s.

3.2 Aileron
The ailerons will be designed to have a max angle of 25 degrees down- and upwards.
By lowering the an aileron the wing will generate additional lifting force while rising the
aileron will create a negative lift force on its wing. By alternating two ailerons on the
aircraft one can create large differences in lift force over the wings, this can be used when
turning, restore the aircraft to its original state if a strong side wind hits. By lowering
both ailerons the aircraft can greatly increase its lifting force which can be used when
the aircraft is flying at low speed and wants to maintain its height above ground.

3.2.1 Designmodeler
The best solution here would be to use the old model from the main wing, but because
of limitations to the geometry which could not change the angle of its tail, (where the
aileron would be) a new geometry were created which could.

Figure 3.36: Aileron at 15 degrees angle downwards.

Note: Because a new model of the wing were being used, a new model for the simu-
lations area would also be needed. This mean that the meshing results will be different
3.2. Aileron 48

from the first part and can therefore create slight changes in lift and drag properties
when doing the simulations later on.

3.2.2 Mesh
The mesh will be created as before, a structured mesh with center around the airfoil with
the results in Figure 3.37. Because the aileron does not need to be tested at different
AoA’s, a small change in the design were made that involved that the inlet area on the
left is no longer rounded.

Figure 3.37: Overview of the finished mesh for the new airfoil.

With the meshing done for the aileron, all important data are collected in Table 3.6.
3.2. Aileron 49

Table 3.6: Mesh results for aileron


Nodes 84213
Orthogonal Quality (min) 0.103
Orthogonal Quality (max) 1.00
Orthogonal Quality (average) 0.964
Skewness (max) 0.998
Skewness (average) 0.117

Inspecting Table 3.6, it can be seen that the max value of skewness exceeds the recom-
mended value of 0.95 which is not desired. By determining how many nodes that have
a high skewness and where they are localized, a decision have to be made if they will
impact the simulations or not. In this case there were around 10 nodes which all were
in non-critical areas, which mean it call still be used for the final mesh. Another value
to note are the increase of the number of nodes that are ∼84000, which is an increase
of ∼34000 from the mesh of the main wing. This increase were made to get a better
orthogonal value which otherwise would have been to low. By altering the aileron at
which angle it will work on, the model will change which in turn alters the mesh and its
end result. This mean that the number of nodes, quality and skewness might have small
differences between the different angles.

3.2.3 Fluent
As done before with the main wing, the aileron simulations will be done using the same
method. Though the aileron will first go between 0 - 25 degrees with a step size of 5
degree at a time and velocities between 2 - 20 m/s with steps of 2 m/s to get a general
view of how the aileron behaves. After that a more thorough inspection will be done
at 20 m/s by simulating with a 1 degree step size, this is what will be used to create a
model later on.
The important part with these simulations is to extract the lift and drag forces the
come only from the aileron and not from the wing in itself. As the aileron that has an
angle will create a greater force versus the original wing with the aileron at 0 degrees,
one can use the difference between the two separate states to locate the ailerons forces.

FAileronDrag (angle) = FAileronDrag (angle) − FAileronDrag (0) (3.17)

FAileronLif t (angle) = FAileronLif t (angle) − FAileronLif t (0) (3.18)

With these two equations located and all the data values collected for the aileron, it is
now possible to create a model for the aileron in Matlab.
3.2. Aileron 50

3.2.4 Simulation results and model of aileron


The collected data from the simulations for lift and drag of the aileron are seen in Figure
3.38 and Figure 3.40. By observing Figure 3.38 it can be seen that the lift generated
from an aileron can dramatically change the properties of the wing. It could increase or
decrease the lift generated by a total of ±30 N/m. This will be a very useful property
for the UAV when it rolls, though this value is per meter and the size of each aileron in
the end might be around 2 dm, meaning that each aileron will generate around ±6 N/m
at 20 m/s. The model that describes the behavior of the aileron are represented in
Equations 3.19 and 3.20 where θ the AoA expressed in radians and V is the velocity
in m/s. A negative angle in this case means that the aileron is tilted downward and a
positive angle that the aileron is tilted upward.
Laileron (θ) = CL (θ) · V 2 (3.19)
(
A1 θ3 + B1 θ2 + C1 θ, θ ≤ 0
CL (θ) = (3.20)
A2 θ3 + B2 θ2 + C2 θ, θ > 0

Table 3.7: Aileron lift coefficients


Prefix\Constant A B C
1 −0.5760 −0.5373 0.2908
2 −1.247 1.118 −0.4245

Lift from aileron


30
Simulated
Modeled
20

10
Drag (N/m)

−10

−20

−30

−40
−25 −20 −15 −10 −5 0 5 10 15 20 25
Angle of Attack (Degrees)

Figure 3.38: Generated lift from aileron at different angles.


3.2. Aileron 51

Calculating the lift error from the aileron with the same method as before with Equation
3.13 and 3.14 yield the following values. A visual display of the error is seen in Figure
3.39 with a max peak at +16 degrees.

RM SEAilLif t = 0.9601 N/m

N RM SEAilLif t = 1.61 %

1
Force (N/m)

−1

−2

−3
−25 −20 −15 −10 −5 0 5 10 15 20 25
Angle of Attack (Degrees)

Figure 3.39: Error in lift force of aileron.

With an understanding of the lift generated, it is also very important to inspect the
drag generated when tilting the ailerons. This is important due to the fact that if not
the ailerons generate the same amount of drag, a yaw force will occur on the UAV which
can divert it from its original flight path if nothing correct this. Fortunately the rudder
will prevent this kind of effect and will be examined later. The drag from an aileron is
explained with Equations 3.21 and 3.22.
At first glance when inspecting Figure 3.40 it can look like the behavior is wrong because
the drag is negative in some angles and that would indicate that the drag were pushing
the aileron forward. The drag from the aileron is indeed negative between the angles of
0 to 12 degrees, but, it is also just for the aileron which is just a small part of the wing.
Summing up the drag from the wing and aileron will in the end land on a positive value,
thus creating a force that will impact the UAV in a negative way.

Daileron (θ) = CD (θ) · V 2 (3.21)


3.2. Aileron 52


2
A1 θ + B1 θ,
 θ≤0
3 2
CD (θ) = A2 θ + B2 θ + C2 θ, 0 < θ < 12 (3.22)

A3 θ + B3 , θ ≥ 12

Table 3.8: Aileron drag coefficients


Prefix\Constant A B C
1 0.0159 −2.202 · 10−4 −
2 0.0270 15.18 · 10−4 −19.70 · 10−4
3 0.0118 −25.90 · 10−4 −

Drag from aileron


1.4
Simulated
Modeled
1.2

0.8
Drag (N/m)

0.6

0.4

0.2

−0.2
−25 −20 −15 −10 −5 0 5 10 15 20 25
Angle of Attack (Degrees)

Figure 3.40: Generated drag from aileron at different angles.

Calculating RMSE and NRMSE again but now for drag. The NRMSE value is still
fairly small, with a peak value at +2 degrees as observed in Figure 3.41.
3.3. Rudder 53

RM SEAilDrag = 0.032 N/m

N RM SEAilDrag = 2.31 %

0.14

0.12

0.1

0.08
Force (N/m)

0.06

0.04

0.02

−0.02

−0.04
−25 −20 −15 −10 −5 0 5 10 15 20 25
Angle of Attack (Degrees)

Figure 3.41: Error in drag force of aileron.

3.3 Rudder
The mission for the rudder is to prevent any yawing motions that will occur. As men-
tioned earlier with the aileron, if there is a difference in the drag generated on any side
of the UAV (to the left or right of CG) there will be a yawing force. This is countered
by applying a rudder to the vertical stabilizer of the UAV.
The vertical stabilizer for the UAV need to have a symmetric geometry to ensure no
unnecessary yawing occurs. This leads to a new wing profile, the NACA0008 airfoil.
This airfoil is pretty thin so it will decrease drag generated from it compared to other
thicker models, though the rudder will create less force when the aircraft is trying to
yaw. This means a new model will be needed for Designmodeler and more meshing and
simulations. This process are done identical to the aileron as these function the same but
have different geometries. The results gathered are seen in Figure 3.42 and Figure 3.43.
As the profile is symmetric the simulations were done only for positive angles. Because
3.3. Rudder 54

it is symmetric, negative angles will give exactly the same values.

The generated yaw force is described with Equation 3.23 and the induced drag with
Equation 3.24. These models will be able to be used between the angles of ±15 degrees,
and as before, θ is expressed in radians and V stands for the velocity given in m/s.

Yrudder (θ) = (A1 θ2 + B1 θ) · V 2 (3.23)

Drudder (θ) = (A2 eB2 θ − A2 ) · V 2 (3.24)

Table 3.9: Rudder coefficients


Prefix\Constant A B
1 0.3565 −0.3418
2 1.626 · 10−5 0.0432

Rudder, yaw force


0

−5

−10
Yaw force (N/m)

−15

−20

−25

−30
0 5 10 15
Angle (degrees)

Figure 3.42: Generated yaw force from rudder at different angles.


3.3. Rudder 55

Rudder, drag
0.7

0.6

0.5
Drag force (N/m)

0.4

0.3

0.2

0.1

0
0 5 10 15
Angle (degrees)

Figure 3.43: Generated drag from rudder at different angles.

Extracting the error from each plot results in Figure 3.44 and Figure 3.45. Both these
plots are within reasonable levels, which can be confirmed again by calculating the RMSE
and NRMSE values.

Error in lift generated between simulated and modeled values


0.6

0.4

0.2

0
Force (N/m)

−0.2

−0.4

−0.6

−0.8

−1
0 5 10 15
Angle of Attack (Degrees)

Figure 3.44: Generated drag from rudder at different angles.


3.3. Rudder 56

Error in drag generated between simulated and modeled values


0.025

0.02

0.015

0.01
Force (N/m)

0.005

−0.005

−0.01

−0.015

−0.02
0 5 10 15
Angle of Attack (Degrees)

Figure 3.45: Generated drag from rudder at different angles.

The calculated values for RMSE and NRMSE are low for the rudder as well and hence,
the model seem to fit very good to the simulated values.

RM SERudderY aw = 0.4477 N/m

N RM SERudderY aw = 1.76 %

RM SERudderDrag = 0.0132 N/m

N RM SERudderDrag = 2.14 %
C HAPTER 4
Experimental Setup

4.1 Equipment
Table 4.1 contains all the items that were either purchased or used for the final build of
the UAV.

Table 4.1: Equipment for the UAV


Description Pieces
NTM Prop Drive Series 42-58 500kv / 1300w Motor 2
APC Style Propeller 15x8 1
APC Style Propeller 15x8R (Right Hand Rotation) 1
Dr. Mad Thrust 70mm 11-Blade Alloy EDF 1900kv Motor 1
Turnigy Plush 60amp Speed Controller 3
Turnigy XGD-11MB Digital Mini Servo 2.2kg / 11g / 0.12 4
Turnigy nano-tech 6000mah 6S 25 50C Lipo Battery Pack 1
10mm*8mm Carbon Fiber Tube 3K Twill 1000mm 3
Worm drive 40:1 ratio 1
DC motor 133 rpm 1
Kfly flight system 1

57
4.2. Building 58

4.2 Building
4.2.1 Main wing
The wings were printed in plastics with a 3D printer. Because the printer was not big
enough to print the whole wing in one go, it had to be divided into smaller parts. Each
wing consists of seven 10 cm parts and two 1 cm parts which were used in the process to
construct the aileron. All parts were then mounted on a 10 mm in diameter carbon fiber
tube and glued together using epoxy, this process results in one wing.

Aileron
These parts were as well created with plastics in the 3D printer. In Figure 4.1 it is
possible to see the rod which the aileron will be mounted on. The aileron had to be
created in two parts due to be able to fit a lever in the middle; which are used to change
the angle on the aileron. A servo will be placed in front of the aileron which is seen in
4.2. It will have an arm that is connected to the lever and with this it is now possible to
control the angle of the aileron with the servo.

Figure 4.1: Insertion of mounting rod for aileron.


4.2. Building 59

Figure 4.2: Aileron mounted with Highlighting of key parts for the aileron.

4.2.2 Horizontal and vertical stabilizers


The horizontal stabilizer has a couple of key parts of interest that are mentioned in Figure
4.3. As seen in the figure, it will have two servos, one that will control the elevator and
the other which will control the rudder on the vertical stabilizer. Both will work in the
same way as the servo for the aileron. A carbon fiber tube will be placed in the mounting
area as noted and will be fastened with epoxy. Parallel to the mounting rod are two
carbon fiber tubes that is located about 25 mm and 50 mm from the front, this will build
the structural strength for this stabilizer. Inspecting Figure 4.4 it can be seen that the
vertical stabilizer has a rod that sticks out at the bottom of it, this rod will be mounted
through the carbon tube that is on top of the horizontal stabilizer and into the insertion
point found on the stabilizer. These parts will then be glued together with epoxy with a
final result seen in Figure 4.5.
4.2. Building 60

Figure 4.3: Highlighting of key parts for the horizontal stabilizer and elevator.

Figure 4.4: Highlighting of key parts for the vertical stabilizer and elevator.
4.2. Building 61

Figure 4.5: Tail section finalized.

4.2.3 EDF clamp


The holder for the EDF has a basic design of a clamp. It contains of two parts as seen
in Figure 4.6 and by fastening the screws on the side, it will squeeze around the EDF
and hold it in place. This clamp will be mounted behind the horizontal and vertical
stabilizer, at the end of the carbon fiber tube.
4.2. Building 62

Figure 4.6: Highlighting of key parts for the EDF clamp.

4.2.4 Worm drive


The two wings will be connected in the upper part of the worm drive, in the middle of
the black wheel in Figure 4.7 and in practice in Figure 4.8. These will then as well be
fastened with screws to the wheel to ensure that they do not move outwards while flying.
To mount the worm drive additional parts where added. The whole configuration needed
to be fastened to the body of the UAV while still having room beneath to get the drive
in with the motor for it. Part 1 will be fastened in the cut outs in the top layer of the
UAV body, while part 2 will connect to the parts below the top layer. Each part will
hold bearings. As a rod that is inserted in the drive, the lower part of the worm drive,
when start rotating it will induce a rotation on the wings with a ratio of 40:1, meaning
it will need 40 turns before the wings have done one. To power the worm drive a small
DC motor will be used with a max rpm of 133. This motor will be fastened to the tail
rod with a similar construction as used with the EDF.
4.2. Building 63

Figure 4.7: Overview of the worm drive.

Figure 4.8: Overview of the attached worm drive motor.


4.3. Final design parameters 64

4.2.5 UAV body


The body for the UAV consists of two layer and were milled from a 2 mm thick aluminum
sheet. The top layer will hold the worm drive as previously mentioned as well as the
carbon fiber tube going to the tail. The top layer is connected to the bottom layer by
eight spacers of length 50 mm. Slightly in front and behind of where the worm drive
is fastened to the top layer is a square Hollow Structural Section (HSS) to increase the
strength of the UAV body, this structure can be seen briefly in Figure 4.8. At the front
of the bottom layer is where the battery will be located. It will be fastened with zip ties
to the body.

4.3 Final design parameters


Main wing dimensions: 1.6 m · 0.15 m
Aileron dimensions: 0.2 m · 0.05 m
Vertical stabilizer, Av : 29.1·10−3 m2
Rudder dimensions: 0.165 m · 0.045 m
Horizontal stabilizer, Ah : 33·10−3 m2
Elevator dimensions: 0.2 m · 0.05 m
Tail length, lH , lV : 0.81 m

Total weight: ∼ 3.5 kg

Center of gravity, CG: 145 mm behind the center of the worm drive
Wing Cubic Loading, WCL: 30
Aspect ratio, AR: 10.7
VH : 0.74
VV : 0.06
Neutral Point, λ: 72mm
Rotational speed of tilt mechanism: ∼ 20 degrees per second
C HAPTER 5
Experimental evaluation

Due to problems that were discovered during flight test (discussed in Section 5.1.2), the
UAV were only tested for its hovering characteristics.

5.1 Testing the UAV’s hovering characteristics


In this area there are four important tests to consider; at what percentage can the UAV
hover in mid air by itself, yaw behavior, pitch behavior and roll behavior. Each test were
done by one person holding the UAV in a non-dangerous area, while the pilot would start
the UAV up and adjust the velocity accordingly.

Figure 5.1: Coordinate system of UAV with origin in the center of the worm drive.

65
5.1. Testing the UAV’s hovering characteristics 66

5.1.1 Percentage of power while hovering


This is a pretty self explanatory test. The percentage of power of the engines are deter-
mined when the UAV can sustain itself in mid air without a person holding it. For the
engines of the wings this value were at 50 percent, while the EDF were at 70 percent.

5.1.2 Yaw behavior


To determine its behavior of yaw, it would need to be rotated around its Z-axis. An idea
here were that the UAV were supposed to use the downwash from the propellers to create
an airstream over the ailerons. With this, the ailerons could be alternated depending on
which way the UAV were supposed to go.
During this test a problem occurred though. Due to the propellers are rotating in different
direction they will always counter the rotational moment that would otherwise occur
from each other. The problem here are the EDF, nothing will counteract the rotational
movement that occurs. This led to a counter clockwise rotation for the UAV around its
Z-axis. This problem also indicate that there can’t be any free flying before it has been
solved. More about this problem will be discussed in Chapter 7.

5.1.3 Pitch behavior


The change in pitch during hovering would alter the UAV’s velocity in the Y-axis, en-
abling it either to move forward or backwards. Tests done here were to see if the UAV
could get back to its original state, i.e be in a horizontal state. A change in pitch means
that the UAV has a rotation around its X-axis. By testing extreme values here, which
meant a rotation of ± 40 degrees around the X-axis, indicated that it would always be
able to get back to its original state. This means that it is stable for its pitch angles.

5.1.4 Roll behavior


The roll behavior will determine the UAV’s ability to counter crosswinds and return to its
original state. The UAV will control the roll moment with a change in velocity between
the propellers, referred to as 4V . The roll will have a rotation around the Y-axis. Again
the UAV were tested for extreme value, approximately ± 40 degrees. As soon as the UAV
started to deviate from its original state it changed 4V , the wing that were high would
have an decrease in velocity and the wing that were on the low side had an increase.
If the UAV would not have encountered a problem during the yaw test it would now have
been ready for a real flight test.
C HAPTER 6
Conclusions

The novelty of this master thesis were to extract a flight model of a tilt wing aircraft
during its transition from a helicopter to an airplane and perform experimental evalu-
ation of its validity. A complete model for the transition state have been derived from
simulations on a specific wing profile which then were evaluated using a wind tunnel.
From the testing phase it was clear that the wind tunnel were too small and it affected
the test runs and therefore a validation could not be drawn. Simulations were also done
to find the flight characteristics with models derived for the ailerons, elevator and the
rudder. A design for a full scaled tilt wing UAV were designed and constructed with a
working tilt mechanism. This UAV were supposed to evaluate the flight model derived
and prove its validity, though a design flaw regarding the EDF ended with the hovering
test for the yaw behavior failed. An efficient method will need to be found to counter
the momentum generated from the EDF before more advanced tests can be done.

67
C HAPTER 7
Future work

This area can be widely expanded into different areas depending what the aim is. The
following is a list of points that could in different ways improve the UAV flight charac-
teristics and its efficiency.

• Investigate and see how the lift force will change in relative to the chord of the wing.
This could help in shorten the wingspan of the UAV and increase the structural
strength of the wings while still maintaining the same lift force.

• Simulate and find how the change in lift and drag change relative to the angle
of tapered and swept back wings to a rectangular wing. And by this information
decide if it is worth to change the wing model.

• Look into how to counter the momentum that is built up from the EDF when
hovering. An idea to this problem would be to install a servo which could alter
the angle of the EDF around the Y-axis. Another less efficient idea is to install a
second EDF at the tail which rotate counter wise to the primary EDF.

• Increase the stability for the tail of the UAV. An idea might be to use two carbon
fiber tubes with some smaller adjustment on the UAV body.

• Incorporating the flight models that were derived from simulations into the UAV.

• Increase the range of angles simulated upon, which would enable more maneuver-
ability.

• Do experimental tests in a wind tunnel to better confirm the model derived.

68
A PPENDIX A
FindKneeLift.m

function [ Theta ] = FindKneeLift( x1, x2 )


% This function will find and return constants for a polynomial of
% order 3. This in turn will be the line between two given x−positions.

A = [x1ˆ3 x1ˆ2 x1 1;
x2ˆ3 x2ˆ2 x2 1;
3*x1ˆ2 2*x1 1 0;
3*x2ˆ2 2*x2 1 0];

B = [LiftEqn(x1); % The lift equation that has been derived


LiftEqn(x2);
LiftEqnDeriavative(x1); % The derivative of the lift equations
LiftEqnDeriavative(x2)];

theta = A\B; % The constants for the polynomial

end

69
A PPENDIX B
LiftEqn.m

function [ L ] = LiftEqn( Angle, V)


% Return the lift force for the given angles and velocities

L = zeros(length(Angle), length(V));

%Constants for the equations


A1 = −5.772; B1 = 0.6895; C1 = 0.4923; D1 = 0.326;
A2 = −28.89; B2 = 37.22; C2 = −15.08; D2 = 2.041;
A3 = −0.1096; B3 = 0.1476; C3 = 0.0428;

% Run the equations for all given angles and velocites


for j=1:length(V)
for i=1:length(Angle)
x = Angle(i);

if (x ≤ 17*pi/180) % Equation 1
L(i,j) = (A1*x.ˆ3 + B1*x.ˆ2 + C1*x + D1).*V(j)ˆ2;

elseif (x ≥ 19*pi/180) % Equation 3


L(i,j) = (A3*x.ˆ2 + B3*x + C3).*V(j)ˆ2;

else % Equation 2
L(i,j) = (A2*x.ˆ3 + B2*x.ˆ2 − C2*x + D2).*V(j)ˆ2;
end
end
end

end

70
A PPENDIX C
LiftEqnDeriavative.m

function [ L ] = LiftEqnDeriavative(Angle)
% Will return the derivative at a given angle.

x = Angle;

A1 = −5.772; B1 = 0.6895; C1 = 0.4923; D1 = 0.326;


A3 = −0.1096; B3 = 0.1476; C3 = 0.0428;

if (x ≤ 17*pi/180) % Derivative of Equation 1


L = (3*A1*x.ˆ2 + 2*B1*x + C1).*V(j)ˆ2;

else % Derivative of Equation 3


L = (2*A3*x + B3).*V(j)ˆ2;
end
end

71
R EFERENCES

[1] N. Eastman, J. Sherman, and A. Sherman, “Airfoil section char-


acteristics as affected by variations of the reynolds number.”
http://naca.central.cranfield.ac.uk/reports/1937/naca-report-586.pdf, 1937.

[2] K. Alexis, G. Nikolakopoulos, A. Tzes, and L. Dritsas, “Coordination of helicopter


uavs for aerial forest-fire surveillance,” Applications of Intelligent Control to Engi-
neering Systems, vol. 39, pp. 169–193, June 2009.

[3] S. Siebert and J.Teizer, “Mobile 3d mapping for surveying earthwork projects using
an unmanned aerial vehicle (uav) system,” Automation in Construction, vol. 41,
pp. 1–14, May 2014.

[4] C. Papachristos, K. Alexis, and A. Tzes, “Towards a high-end unmanned tri-


TiltRotor: design, modeling and hover control,” in Mediterranean Conference on
Control & Automation, 2012.

[5] C. Papachristos, E. . C. Dept., U. of Patras, K. Alexis, and A. Tzes, “Model predic-


tive hovering-translation control of an unmanned Tri-TiltRotor,” IEEE International
Conference on Robotics and Automation, 2013.

[6] J. Dickeson, D. Mix, J. Koenig, K. Linda, O. Cifdaloz, V. Wells, and A. Rodriguez,


“H∞ Hover-to-Cruise Conversion for a Tilt-Wing Rotorcraft,” pp. 6486–6491, De-
cember 2005.

[7] E. Cetinsoy, S. Dikyar, C. Hancer, K. Oner, E. Sirimoglu, M. Unel, and M. Aksit,


“Design and construction of a novel quad tilt-wing UAV,” Mechatronics, pp. 723–
745, September 2012.

[8] K. Muraoka, N. Okada, and D. Kubo, “Quad Tilt Wing VTOL UAV: Aerodynamic
Characteristics and Prototype Flight,” April 2009.

[9] N. Metni and T. Hamel, “A uav for bridge inspection: Visual servoing control law
with orientation limits,” Automation In Construction, vol. 17, pp. 3–10, 2007.

72
73

[10] J. Andersson, Fundamentals of Aerodynamics. No. ISBN: 978-007-128908-5,


McGraw-Hill, 5th ed., 2011.

[11] C. Nordling and J. Österman, Physics handbook for science and engineering.
No. ISBN: 978-91-44-04453-8, 8:3 ed., 2007.

[12] K. Myers, “Wing cube loading (wcl).” http://www.theampeer.org/M1-


outrunners/M1-outrunners.htm, 2014.

[13] T. Wilkerson, “Cubic wing loading.” http://www.eastbayrc.org/index.php/tims-


tips/81-cubic-wing-loading.

[14] M. H. Sadraey, Aircraft Design: A Systems Engineering Approach. No. ISBN: 978-
1-119-95340-1, 2012.

[15] I. Kroo, “Aircraft design: Synthesis and analysis.”


adg.stanford.edu/aa241/AircraftDesign.html.

[16] B. Mullen and S. lachance Barett, “Fluent - flow over an air-


foil.” https://confluence.cornell.edu/display/SIMULATION/FLUENT+-
+Flow+over+an+Airfoil, 2014.

[17] S. Richards, K. Martin, and J. Cimbala, “Ansys workbench tutorial - flow over an
airfoil.” http://www.mne.psu.edu/cimbala/, 2011.

[18] J. Trapp and R. Zores, “Naca 4 digits series profile generator.”


http://www.ppart.de/aerodynamics/profiles/NACA4.html, 2011.

[19] A. Bakker, “Applied computational fluid dynamics.”


http://www.bakker.org/dartmouth06/engs150/07-mesh.pdf, 2006.

[20] F. Villalpando, M. Reggio, and A. Ilinca, “Assessment of turbulence models for flow
simulation around a wind turbine airfoil,” Modelling and Simulation in Engineering,
p. 8, 2011.

[21] J.-P. Renaud, “Advanced technologies and new roles for vtol aircraft (part ii/ii),”
Air & Space Europe, vol. 2, pp. 53–60, May 2000.

[22] G. Varela, P. Caamaño, F. Orjales, A. Deibe, F. López-Peña, and J. Duro, “Au-


tonomous uav based search operations using constrained sampling evolutionary al-
gorithms,” Neurocomputing, vol. 132, pp. 54–67, May 2014.

[23] A. Verger, N. Vigneau, C. Chéron, J.-M. Gilliot, A. Comar, and F. Baret, “Green
area index from an unmanned aerial system over wheat and rapeseed crops,” Remote
Sensing of Environment, vol. 152, pp. 654–664, September 2014.
74

[24] Principles of Flight. No. ISBN: 82-8107-054-4, Nordian AS and London Metropolitan
University, 2nd ed.

You might also like