You are on page 1of 70

Autonomous Systems Lab Prof.

Roland Siegwart

Master-Thesis

Design and Development of a Flapping Wing Micro Air Vehicle


Oct 2009 - Apr 2010

Supervised by: Stefan Leutenegger Samir Bouabdallah

Author: Atilla Yilmaz

Contents
Abstract Acknowledgment 1 Introduction 1.1 1.2 2 Motivation & Background . . . . . . . . . . . . . . . . . . . . . . . . . . Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii v 1 1 1 3 3 5 7 9 9

State of the Art 2.1 2.2 2.3 Flapping Wings Aerodynamics . . . . . . . . . . . . . . . . . . . . . . . . Flapping Wings Worldwide . . . . . . . . . . . . . . . . . . . . . . . . . . Flapping Wings at ETH Zurich . . . . . . . . . . . . . . . . . . . . . . . .

Conceptual Design 3.1 3.2 3.3 Aerodynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Mechanism . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 Simulation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11 3.3.1 3.3.2 3.3.3 Mechanisms (Simulink) . . . . . . . . . . . . . . . . . . . . . . . 12 Actuators & Gears . . . . . . . . . . . . . . . . . . . . . . . . . . 14 Amplitude Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 14

3.4 4

Evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 19

Development 4.1 4.2 4.3

Wing Concept . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Mechanical Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 Tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 4.3.1 4.3.2 Test Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 Result Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24 27

Results 5.1 5.2

Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Controllability . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28 33 i

Conclusion

A Measurements B Matlab/Simulink List of Figures Bibliography

35 57 60 61

ii

Abstract
Flapping wing micro air vehicles (MAV) are a very interesting and young research area. Nature is a big antetype and still not reached by todays mechanisms. At the ASL1 , a apping wing MAV for hovering ight was designed and developed. Due to the fact that the highly unsteady aerodynamics are difcult to capture, for a order of magnitude estimation steady airfoil theory was taken for simulating different concepts of mechanism. A 4-bar mechanism was used to turn the rotation of the applied DC-motor into a apping of the wings. The variation of the angle of attack for keeping positive lift force generation in up and down stroke was realized by a passive rotation of the wings along their span axis. In order to achieve three degrees of freedom, roll and pitch and variation in altitude, a four wing design was implemented. The wing design was inspired by ies on one hand and hummingbirds on the other hand. After having tested both designs the hummingbird design showed his advantages over the y design. These advantages were mainly the light weight construction and the clear dened angles of attack. After improving the wings with over 300 test the optimal shape was found which led to 7.6g0.3g lift force generation with two wings with a input of 4V. The MAV has a weight of approximately 14.5g. Through a motor controller and a joypad the four motors of the device could be reached and calibrated independently. It could be shown that there is a linear dependency between input and lift production. The unsteady dynamics in torque generation were quantied and a free ight test with a guideway from the bottom of the MAV showed the capability for hovering.

1 Autonomous

Systems Lab, ETH Zurich

iii

iv

Acknowledgment
First of all I would like to thank my supervisor, Stefan Leutenegger, for guiding me through my thesis. He was always a big help and inspiration with his ideas and thoughts. And I also would like to thank Manuel Naef for his co-work in the rst part of my thesis. And of course I have to thank my family and my friends for their support over my whole studies.

Zurich, 7. May 2010 Atilla Yilmaz

vi

Chapter 1

Introduction
1.1 Motivation & Background

Of all locomotions nature created ying was always the most mystic one. From Daedalus to Leonardo da Vinci, all tried to imitate the apping of birds wings without having knowledge about the aerodynamics that are present. The lift force generation of insects in hovering ight cant be explained by steady airfoil theory. Thus, unsteady aerodynamics must be attendant [2]. Most fascinating in apping wing ight is hovering of insects and hummingbirds. Flapping wing mechanisms are in one way able to perform hovering ight and are also used as xed wings when thermal lift is present for long distant ights. The capability of combining this two skills with one apparatus contains a huge potential. Also the extreme maneuverability [3] that underlies in apping wings is still unreached by todays MAVs.

1.2

Objectives

The aim of this master thesis was to design and built a apping wing MAV capable of hovering. In the rst part there was a co-work with an ongoing semester thesis which analyzed apping wing mechanism and amplitudes. First, an overview about existing platforms is given. Some preliminary studies are shown and the unsteady aerodynamics are explained. In chapter 3 the conceptual design is presented with the classication into actuation and mechanism. The equations of motion are given and it is looked into the simulation. Also a general view over the amplitude analysis and rst results are presented. In the subsequent chapter 4 the hardware development such as wings and structure is shown. The last chapter covers the whole testing part and the results and also the controllability of the developed MAV.

Chapter 1. Introduction

Chapter 2

State of the Art


2.1 Flapping Wings Aerodynamics

The air ow eld in apping wings can not be assumed as steady. A large angle of attack would lead to ow separation and turbulences too. Obviously, there must be phenomena producing extra lift. These unsteady aerodynamics cant be explained by common airfoil theory and are illustrated below. For deeper comprehension it is referred to previous works by O. Breitenstein [10], S. Gisler [9] and M. Naef [11].

Unsteady Aerodynamics The highly difcult fact that lift production of apping wing mechanisms are explained with unsteady aerodynamic effects is the reason for the emerging and so far not yet fully explored eld of FWMAVs 1 . In the following the most important effects are listed. Leading edge vortex Clap and ing mechanism Rotational lift Wing-wake interactions A leading edge vortex (LEV) is created at a high angle of attack. A high angle of attack would normally lead to ow separation. But the LEV is responsible for the ow to stay attached to the wing. At the beginning of the down stroke, a strong vortex owing from the base to the tip is formed at the leading edge and helps to generate high lift force. This is best seen in the right part of the next gure.
1 Flapping

Wing Micro Air Vehicle

Chapter 2. State of the Art

Figure 2.1: Leading edge vortex, [6] In the left part another effect of the LEV is shown. The vortex generated in the down stroke is separated after it reaches the tip of the wing. This vortex works as a support for the upstroke and helps to produce even more lift force. In the clap and ing mechanism the wings come together dorsally at the end of the upstroke to perform a clap (A).

Figure 2.2: Clap and ing mechanism, [7] After the clap the wings ing apart (B). Air is sucked in (C) as the wings start to move downwards creating a bound vortex on each of the wings which produces an instantaneous lift force (D).

2.2. Flapping Wings Worldwide

At the end of each half stroke, the wing performs a rotation around a span wise axis which allows an insect to maintain a positive angle of attack during both down stroke and upstroke.

Figure 2.3: Wing rotation over one apping cycle, [8] This rotation guarantees continuous lift production over the whole apping cycle. The wings operate with the surrounding ow in specic ways. Used air interacts positively into the apping process. This phenomena is also referred to as wake capture.

2.2

Flapping Wings Worldwide

During the last years there were huge advancements in micro actuation and ultra light electronic devices. This improvements opened the ways to micro aerial vehicles in different shapes and dimensions. Also apping wing MAVs are considered in the wide range of their application area. In the following just some examples are introduced and their principles of lift production is explained. DelFly The DelFly was developed at TU Delft. As seen in the next gure it consists of four wings.

Figure 2.4: DelFly, TU Delft, www.photoscala.de

Chapter 2. State of the Art

The wings are connected in pairs. The right upper wing is coupled with the left lower wing and vice versa. Thus lift force is mainly generated with the clap and ing mechanism as described in 2.1. Actuation is realized by one motor. The DelFly has capability for hovering ight but due to the structure it performs only forward ight. i-Bird The i-Bird can be seen as a direct offspring of the DelFly. The apping mechanism is the same thus lift force is generated with the same clap and ing mechanism, as seen in gure 2.5.

Figure 2.5: i-Bird, Silverlit, http://www.woti.de/images/ibird2.jpg Also the i-Bird generates enough lift force to perform hovering but it is constructed the way that it bends to forward ight. Robotic Insect As described in [5] the Robotic Insect (gure 2.6) is a ultra micro apping wing aerial vehicle developed at the Harvard Microrobotics Laboratory.

Figure 2.6: Robotic Insect, Harvard University, http://wyss.harvard.edu/viewmedia/82/roboticy With a wing span of 3cm it imitates the wing concept of small insects. Lift force is generated mainly by the leading edge vortex and wing rotation as described in 2.1. The wing rotation is realized by a passive rotation around the span axis. For actuation a piezoelectric cantilever is used. At resonance frequency high amplitudes and remarkable lift forces

2.3. Flapping Wings at ETH Zurich

are produced. A drawback is that piezoelectric devices need high voltage and therefore no power supply could be integrated in this vehicle so far.

2.3

Flapping Wings at ETH Zurich

At the Autonomous Systems Lab (ASL) at ETH Zurich there were already a semester thesis [10] and a studies on mechatronics [9] completed at the start of this thesis. The result of the semester thesis is seen in the next gure.

Figure 2.7: First design of a apping wing mechanism at ASL, ETH Zurich The mechanism for converting the rotation of the chosen actuator, a DC-motor, into a apping features relatively much friction. Also the whole structure was just built for measuring lift forces and no actual hovering can be performed with this device nor was there any improvements on the wings. In the studies on mechatronics the unsteady aerodynamics in apping wings are described. The different effects are deeply explained and comparisons are made with insects and existing apping wing MAVs. The whole thesis is theoretical and no hardware was designed nor were there any suggestions for wing design. Thus, the goal of this thesis was to develop a MAV actually capable of hovering ight and to come up with a device to augment the knowledge of the mechanisms of lift generation in practice. The focus was also in the improvement in wing design and controllability.

Chapter 2. State of the Art

Chapter 3

Conceptual Design
Design of efcient apping mechanisms for apping wing micro air vehicles is mainly constrained by micro actuators. However, the energetics of apping wing motion is not well understood. Insects and hummingbirds wing apping at high frequency is fascinating and still unreached by todays articial wing mechanisms. There also have been studies exploring the possibility of using springs to store the kinetic energy [1].

3.1

Aerodynamics

To capture an order of magnitude estimation of the aerodynamics forces steady airfoil theory is applied. This approach is limited mainly by the velocity of the wings and the unknown lift coefcients. The lift force generating unsteady aerodynamic effects as described in 2.1 can not be explained with this theory. For estimation reasons it is applied anyhow. Thus, the following equations are valid only for steady conditions.

Lift & drag forces in steady airow

The main formulas are given in this section and for more insight it is referred to the semester thesis of M. Naef [11]. F = c 2 A 2 (3.1)

This is the very general formula to start with. is the air density, is the airspeed, A is the wing area and c is a coefcient that will be dened later. This formula must now be applied to apping wings (gure 3.1). 9

Chapter 3. Conceptual Design

10

Figure 3.1: Flapping wing mechanism with dened angle for equation The wing angle is marked as . With (t) being the angular velocity of the wing the velocity in equation 3.1 can be derived with the following formula. = l l is a characteristic length of the wing. Therefore the general formula for apping wings is given below with: F(t) = c A l2 2 2 (3.3) (3.2)

Both, lift and drag forces are calculated with the same formula. The difference is in the so called lift and drag coefcients, marked as cl and cd . These coefcients are mostly dependent on the angle of attack and the wing parameters, such as chord and wing length. For the simulations (3.3.1) cl and cd are taken as constant in the rst approximation and their dependencies are considered later in the tests.

3.2

Mechanism

The basic question is how to realize a apping mechanism with given actuators. The list of possible actuation is indeed rather small. There are linear actuators and DC motors. Other kind of actuation, such as piezo elements e.g., was not considered due to small deection. Linear motors have smaller force density than DC motors. Together with the wide range of availability of DC motors considering the dimension they were selected as actuators for the mechanism. There were three different concepts that were considered and simulated. Concept A: Direct actuation In this conguration the wing is directly attached to the motor (gure 3.2). The apping would be realized by alternating the input to the actuator.

Figure 3.2: Flapping wing mechanism: Direct actuation

11

3.3. Simulation

Both the frequency and the amplitude can be adjusted and are not predened. This would make it much more convenient to test a wide range of different setups.

Concept B: Actuation with mechanism In this arrangement the mechanism is a four-bar linkage (gure 3.3).

Figure 3.3: Flapping wing mechanism: Actuation with mechanism The amplitude is predened and is adjusted with the lengths of the links. This design allows the motor to rotate continuously and hence doesnt need a alternating input signal.

Concept C: Mechanism combined with torsional spring This assembly is basically the same as concept B just with the difference that there is a torsional spring in the wing (gure 3.4).

Figure 3.4: Flapping wing mechanism: Combination with torsional spring In a situation with resonance the amplitude of the wing is amplied much more than the amplitude of the bar-mechanism. So it could probably support the motor and allow higher amplitude and frequency at smaller motor input.

3.3

Simulation

In the next sections the simulation parameters of the developed mechanisms for MATLAB/Simulink are given. Actuators and gears are analyzed in order to fulll the requirements of the results from the simulation and a brief overview is given to the amplitude analysis.

Chapter 3. Conceptual Design

12

3.3.1

Mechanisms (Simulink)

In the following the equations of motion for the different concepts are formulated. The detailed structure of the Simulink les are given in appendix B.

Concept A The relevant forces are the motor torque Mm and the torque from the aerodynamic drag force Fd (t). The main relation between the motor angle and the wing angle is given as: (t) = (t) The aerodynamic drag force is given below. Fd (t) = cd A 2 l 2 sgn() 2 (3.5) (3.4)

with cd as the drag coefcient and l as a characteristic length of the wing (depending on the shape of the wing). The equation of motion for the direct actuated apping wing mechanism is given as: w = Mm Fd l with w as the moment of inertia of the wing. (3.6)

Concept B The relevant forces for concept B are seen in the next gure.

Figure 3.5: Actuation with mechanism: Relevant forces and parameters

The equation of motion for the motor angle is as follows (for c m = Mm b cos() F

b): (3.7)

with m as the moment of inertia of the motor and F as the force in the joint c. For the wing angle follows: w = Fd l a cos() F (3.8)

13

3.3. Simulation

The relationship between the motor angle and the wing angle is given as: (t) = 0 sin((t)) with 0 being the half amplitude of a apping. It follows for the angular velocity (t) = 0 cos((t)) and for the angular acceleration (t) = 0 ( sin() 2 + cos() ) The equation of motion for the four-bar mechanism is given therefore as: = m 1
b cos()w 0 cos() a cos(0 sin())

(3.9)

(3.10)

(3.11)

Mm

b cos() [Fd l w 0 sin() 2 ] a cos(0 sin()) (3.12)

Concept C The schematic design for concept C is seen in the next gure.

Figure 3.6: Actuation with torsional spring: Relevant forces and parameters The rst equation of motion for the wing is given as follows: w = k( ) + Fd l with k as the spring constant and as a new angle between the wing and the bar. With F a cos( ) = k( ) and m = Mm b cos() F the second equation of motion for the four-bar mechanism follows as: m = Mm b cos() k 0 sin() + a cos(0 sin()) (3.16) (3.15) (3.14) (3.13)

Chapter 3. Conceptual Design

14

3.3.2

Actuators & Gears

For a detailed analysis of the different actuators see chapter 3.1 in Design of a Flapping Wing Mechanism by Manuel Naef [11]. As mentioned there, DC motors have a high force density when torque to weight relation is compared. Together with the exibility in the availability DC motors were chosen as actuators. Different congurations were simulated. There was a trade off between the weight, the power and the supplied torque. There was an overall constraint in the weight of the MAV too. So for the simulation only lightweight motors were used. For a typical frequency of 250Hz in a DC motor the gear ratio was chosen to be about 25 so a apping around 10Hz could be achieved.

3.3.3

Amplitude Analysis

As in the formula 3.3 seen a high amplitude leads to high lift force generation for 0 as the amplitude is squared. Therefore it was chosen to have a high amplitude for the simulation and for the rst mechanical design too. The effects of the four-bar mechanism to the DC motor were analysed in [11] in chapter 4.2. The amplitude was chosen to be 135 .

3.4

Evaluation

Mechanisms & Structure The three different concepts were implemented in Simulink and analyzed by their energetic usage for the same apping amplitude. The results for lift forces were just taken for comparison and not as absolute values. Concept A combines two degrees of freedom as the frequency and the amplitude are adjustable separately. But the simulation showed that for the motor to ap with high frequency very high voltages are needed such as 100V and more. On top of that the oscillation has a drift from the initial value as seen in the next gure. This is due to the fact that the oscillating input frequency is too high for the wing inertia to break and accelerate at the same frequency which leads to delays and to drift from a set point.

Figure 3.7: Simulation of concept A: drift of initial value.

15

3.4. Evaluation

In any case the oscillation is not obtainable with common energy source and therefore this concept was dropped. Concept B and concept C allow the DC motor to run continuously and achieve higher efciency due to this fact. Also a adjustable gear ratio is possible. In concept B there is a steady apping and the amplitude is x (gure 3.8).

Figure 3.8: Simulation of concept B: steady apping with predened angle. The difculty with concept C is that high amplitudes are achieved in resonance frequency. The related spring constant was calculated with the next formula. k = w (2 f )2 (3.17)

It it seen that the spring constant denes the apping frequency at resonance. In the simulation an unsteady apping occurred as seen in the next gure.

Figure 3.9: Simulation of concept B: unsteady apping with undened angle. Besides, the results for energy consumption showed that concept B was more efcient. Thus it was chosen as the apping wing mechanism for the MAV. In order to obtain at least three degrees of freedom without having the disadvantage of

Chapter 3. Conceptual Design

16

an additional actuator that doesnt support the lift production the mechanism was designed with four wings. For detailed information see chapter 4.1 in [11].

Figure 3.10: Four-Wing Concept.

Ordered in a 90 rotated position two wing pairs should permit at least three degrees of freedom. These are roll, pitch and altitude. All four wings generate lift force and are thus responsible for the changing in the height. For roll or pitch two wings are sufcient. A wing pair is for that purpose actuated in different frequencies which leads to a higher force generation on one side and so to a induced torque.

Actuators

Four DC motors will outbalance the overall weight of the MAV. The wing span of 13cm was set as a constraint in order to keep the dimensions of the MAV close to the ones of a hummingbird. Thus micro motors of very small dimensions were considered. With a diameter of 6mm and a length of 15mm the DC motors from Precision Microdrives (Model 106-002) weigh only 1.7 grams.
1

1 www.precisionmicrodrives.com

17

3.4. Evaluation

Figure 3.11: Micro DC motor with 6mm diameter and 15mm length

Chapter 3. Conceptual Design

18

Chapter 4

Development
In this chapter the design and development of the hardware is given. A seperation is made between the wings on one hand and the structure on the other hand and it is completed with the test section.

4.1

Wing Concept

For hovering ight the apping of the wings are basically described by two movements. As seen in the next gure there is a horizontal motion indicated as up and down stroke.

Figure 4.1: Wing motion of a y in hovering ight, [8]

The wings have a second motion around the span axis. This rotation changes the angle of attack and allows the y to generate positive lift force at up and down stroke. In order to achieve both movements the horizontal motion is controlled activly by the design of the mechanism and the motion of the wing is accomplished by a passiv rotation around the pivot. There were mainly two different shapes for the wing design. 19

Chapter 4. Development

20

Figure 4.2: Different shapes of wings, source: http://www.news.wisc.edu/; naturefocused.com One inspiration was taken from insects, such as ies. The shape of their wings is seen in gure 4.2. The second possible shape for the wings was inspired by hummingbirds. The implementation of these shapes into the design of the wings is seen in gure 4.4.

Figure 4.3: Implementation of y wings

Figure 4.4: Implementation of hummingbird wings Both wings are allowed to rotate freely around the wingspan axis. The y shaped design had to be extended by additional mechanics to keep the passive rotation dened. These stoppers were placed on the wing axis and adjusted to the desired angle. This angle of attack was constant over the whole wing span. Also two supporting rods were placed on the wings in order to keep the structural integrity (gure 4.3). This supplementary structures resulted in heavy wings and slowed down the passiv rotation which is crucial for imitating the movements of apping wings. In the hummingbird design two wings are made of one piece (gure 4.5).

21

4.2. Mechanical Design

Figure 4.5: Two wings made of one piece Although the wings rotate around the wingspan axis no stoppers are needed as in the y shaped design for the wings are xed to the rod (gure 4.1). This makes it also possible to get rid of the supporting rods as in the y shaped design and even to use a lighter material like boPET 1 which affects positivly the ability for passiv rotation. Additional, the angle of attack is not constant over the wing span which comes close to imitating a hummingbird wing.

4.2

Mechanical Design

After the conceptual design of the amplitude the geometry of the structure was already predened. Mainly the distances between the axis were set in advance. This is seen in the next gure.

Figure 4.6: The main structure with predened axis The parameters a, b, c and d correspond to the parameters as explained in chapter 3. In order to achieve a apping frequency of 10Hz with the chosen motors, which have a speed of 15000rpm, a ratio of 25:1 for the gears were necessary. The available gears 2 led to the idea to try three different gear ratios. These were 27:1, 20:1 and 16:1, performed with different combinations of teeth as listed below.

1 Biaxially-oriented 2 www.didel.ch

polyethylene terephthalate (e.g. Mylar)

Chapter 4. Development

22

27:1 : 9 to 36/12 to 81 20:1 : 9 to 36/12 to 60 16:1 : 9 to 36/12 to 48 Together with the additional xed distances between the gear axis the geometry of the structure was set. The motors though were still to be placed and were chosen to be close to eachother and to the wing rotation axis allowing to keep the center of gravity concentrated in the middle of the MAV (gure 4.7).

Figure 4.7: The main structure with the distances between the gear axis The nal structure of the MAV with two wings is seen in gure 4.8.

Figure 4.8: The main structure with all components All the structure components were manufactured with a 3D printer.

4.3

Tests

Due to the fact that the unsteady aerodynamics could not be simulated the important parameters of the MAV had to be found in tests. The results of the rst tests for the y inspired design were as follows. 4.5g lift at 4Hz with four wings for a MAV of 14.5g weight. This poor result was effected by the heavy wings. No higher frequency could be achieved and thus no deeper improvements on this design was done.

23

4.3. Tests

The hummingbird inspired design produced 5.8g lift at 8Hz with two wings. This promising design was improved with further testing. The interesting parts of the search space are listed below. Flapping amplitude: 65 100 135 Gear ratio: 16 - 20 - 27 Angle of attack Chord length There was no variation on shape of the wings nor in the motors. The variation in amplitude and gear ratio made it necessary to design the structure in 9 different setups. These setups were then combined with the variations in the wing parameters. The angle of attack was changed by altering the wing foil geometry. The wing and the structure form a 90 angle to eachother. By adding more degrees to this angle the supplementary material will dene the angle of attack in the wings. The additional degrees are taken as values to be compared. For the chord length a characteristic length was taken as seen in the next gure.

Figure 4.9: Wing parameters: Chord length and angle In the rst iteration the angle was tested with ve different congurations and the chord lenght with seven. Therefore over 300 tests were performed to cover the whole search space. Some results made it necessary to extend the tests and a second iteration with a wider search space was accomplished.

Chapter 4. Development

24

4.3.1

Test Setup
3

The rst design with y shaped wings were tested directly in a 4-wing conguration. The tests with the hummingbird shaped wings were performed with two wings. The sensor was capable of measuring the three forces with an accuracy of 0.3 Newton. The measurements of torque were used to derive the apping frequency. In the next gure the two setups with the different wings are shown.

Figure 4.10: Testing of the y inspired design

Figure 4.11: Testing of the hummingbird inspired design All motors were connected to a DC voltage supply and were given the same voltage, namely 4V. The current supply was restricted to 0.34A but was never reached.

4.3.2

Result Analysis

The sensor is reached by an existing GUI written in LabVIEW. The data is sampled with 10kHz. The generated le was analysed with Matlab. The data was ltered and the lift force was calculated over 4 seconds in order to get mean values. An example plot is seen in the next gure.
3 Nano17

6-Axis force/torque sensor, Ati Industrial Automation (www.ati-ia.com)

25

4.3. Tests

Figure 4.12: Plot of a lift force measurement Although the sensor was biased befor the tests there was still an offset. The lift force in this example is about 0.07 Newton. The torque of the yaw axis was plotted against time so the frequency of the apping was visible (gure 4.13).

Figure 4.13: Plot of a torque measurement (yaw) to derive apping frequency As seen the counting is indicated. At the end of each second there was an estimation of the last apping cycle. The accuracy was 0.1Hz. In this example the apping frequency is about 10.8Hz.

Chapter 4. Development

26

Chapter 5

Results
All the test for lift generation were accomplished with two wings. It was assumed that the results could be adopted to the four wing design without issues. For controllability however the MAV with actually four wings was tested.

5.1

Measurements

The detailed results are given in the Appendix. In order to nd the optimal wing all the results from the measurements were compared to eachother. In the next gure the wing with all the wing parameters is shown.

Figure 5.1: Shape of the optimal wing

The wing design was improved in chord length and wing angle. In the left part of the next gure, given as an example, the angle was x and the chord length was varied and vice versa in the right part. 27

Chapter 5. Results

28

Figure 5.2: Results from the measurements

Also the gear ratio and the apping amplitude were not varied. It can be seen that sharp optima occurred. After all the tests for the hummingbird inspired design the optima in the search space was found.

Flapping amplitude: 100

Gear ratio: 27

Wing angle: 12.5

Chord length: 55mm

With these congurations the generated lift force with two wings is 7.6g 0.3g with a input of 4V. The MAV has a weight of approximately 14.5g. At 4.5V the lift force increases to 8.7g 0.3g.

5.2

Controllability

After having found the optimal conguration with two wings a four wing MAV was designed in order to test further behavior such as controllability in roll and pitch axis. Also the lift generation was evaluated. The four wing design needed a new connection between the upper and lower part, which are the same but just turned by 90 and upside down as seen in the next gure.

29

5.2. Controllability

Figure 5.3: Connection between upper and lower part

A redesign of the upper and lower parts resulted in 40% loss of weight. For advanced testing the four DC-motors were connected to a motor controller with inputs by a joypad.
1

and fed

Figure 5.4: The motor controller MotorBee

Figure 5.5: The joypad Rumblepad, sysprole.de


1 MotorBee,

www.pc-control.co.uk

Chapter 5. Results

30

Mapping The inputs from the joypad are mapped into the motor signals with a c++ application. The joypad has 2 analog joysticks and 10 digital buttons. The left joystick is used to give the collective input the each of the motors. The right joystick moved left and right causes a variation in the apping frequency of the wings of the upper part and provokes a changing in the roll movement. The joystick moved up and down is causing with the same effect in the lower part the pitch movement. For testing the changes in the input were set to 20% of the maximum input. With the digital buttons each motor can be reached independently. So buttons 1 to 4 increase the input to the respective motor by one percent. Buttons 5 to 8 decrease them by one percent.

Tests & Results The tests for lift force generation showed a linear dependency between the input voltage and the lift force. This is seen in the next gure.

Figure 5.6: Linear dependency for input and lift force generation

The analysis of the torques showed that these dynamics are highly unsteady. The next gure shows one result from the tests.

31

5.2. Controllability

Figure 5.7: Unsteady dynamics in torque generation

The measured torques were not reduced to the center of gravity of the MAV thus no conclusion can be made for the controllability. The only statement is that by varying the input the torque can be varied also. The indications left, right, up, down in the gure above refer to the joystick inputs through the gamepad. The nal tests to prove capability for hovering ight were therefore arranged with a guideway from the bottom of the MAV. The next gure shows a screenshot from the demonstration video.

Figure 5.8: Proof of capability of hovering ight

Chapter 5. Results

32

The guideway is seen on the bottom of the gure and also the proof that the apping wing MAV is capable of hovering ight.

Chapter 6

Conclusion
The goal of this thesis was to develop a MAV capable of hovering ight. The focus was also in the improvement in wing design and controllability. This goals were, except of the controllability, fully reached. The chosen mechanism for turning rotation of the DC-motors into a apping movement of the wings is accurate and leads to error free functioning. By decreasing the amplitude even smoother behavior of the motor can be expected. The 3D printer material proved its usability in all design steps and is recommended for further use. Through more than 300 tests the wings were optimized in angle and chord length and yield to enough lift generation for hover ight. The improvement according the rst design is an increase by a factor of 4. The four wings produce about 0.16N to 0.2N lift force depending on the input voltage. The weight of the MAV is about 14.5g and therefore could be extended by additional hardware such as electronic or batteries or sensors. The static behavior with respect to controllability was not fully characterized. It was shown that by changing the inputs to the motors the applied torques can be changed also but there was no reduction of the torques to the center of gravity of the MAV. The motor controller that allows independent inputs to each motor is essential for control issues. These inputs can easily be mapped to the wished behavior through a coupled gamepad. Future Work In order to achieve more lift generation the search space for nding the optimal wing and mechanical conguration can be extended by a smoother discretization and including also the wing shape. There were no improvements on the motors so far. Many suppliers have the same size of DC-Motors but with different parameters, such as velocity constant. Also there should be a betterment in the controllability by calibrating the motors and reducing play in the structure. 33

Chapter 6. Conclusion

34

Concluding it can be said that the apping wing MAV designed and developed in this master thesis is capable of hovering ight. Yet it is not controlled and yields to torque in the pitch and roll axis. The gained knowledge can be used for further improvements and opens the way toward an autonomous apping wing MAV.

Appendix A

Measurements
Promising results are marked and retested with a wider search space. Shape the best means that the optimal conguration was extended by a search over the wing shape. Some results were varied in the voltage input to the motors.

Figure A.1: Results: Gear-ratio 27, Amplitude 135 35

Appendix A. Measurements

36

Figure A.2: Results: Gear-ratio 27, Amplitude 135

37

Figure A.3: Results: Gear-ratio 27, Amplitude 100

Appendix A. Measurements

38

Figure A.4: Results: Gear-ratio 27, Amplitude 100

39

Figure A.5: Results: Gear-ratio 27, Amplitude 100

Figure A.6: Results: Gear-ratio 27, Amplitude 100 As seen in gure A.6 there was a extended testing on the optimal wing by varying the wing span.

Appendix A. Measurements

40

Figure A.7: Results: Gear-ratio 27, Amplitude 65

41

Figure A.8: Results: Gear-ratio 27, Amplitude 65

Appendix A. Measurements

42

Figure A.9: Results: Gear-ratio 27, Amplitude 655

43

Figure A.10: Results: Gear-ratio 20, Amplitude 135

Appendix A. Measurements

44

Figure A.11: Results: Gear-ratio 20, Amplitude 135

45

Figure A.12: Results: Gear-ratio 20, Amplitude 100

Appendix A. Measurements

46

Figure A.13: Results: Gear-ratio 20, Amplitude 100

47

Figure A.14: Results: Gear-ratio 20, Amplitude 100

Appendix A. Measurements

48

Figure A.15: Results: Gear-ratio 20, Amplitude 100

49

Figure A.16: Results: Gear-ratio 20, Amplitude 65

Appendix A. Measurements

50

Figure A.17: Results: Gear-ratio 20, Amplitude 65

51

Figure A.18: Results: Gear-ratio 16, Amplitude 135

Appendix A. Measurements

52

Figure A.19: Results: Gear-ratio 16, Amplitude 135

53

Figure A.20: Results: Gear-ratio 16, Amplitude 100

Appendix A. Measurements

54

Figure A.21: Results: Gear-ratio 16, Amplitude 100

55

Figure A.22: Results: Gear-ratio 16, Amplitude 65

Appendix A. Measurements

56

Figure A.23: Results: Gear-ratio 16, Amplitude 65

Appendix B

Matlab/Simulink

The block diagrams are given for the different concepts.

Figure B.1: Simulink block diagram for concept A 57

Appendix B. Matlab/Simulink

58

Figure B.2: Simulink block diagram for concept B

Figure B.3: Simulink block diagram for concept C

Figure B.4: Simulink block diagram for concept C

List of Figures
2.1 2.2 2.3 2.4 2.5 2.6 2.7 3.1 3.2 3.3 3.4 3.5 3.6 3.7 3.8 3.9 Leading edge vortex, [6] . . . . . . . . . . . . . . . . . . . . . . . . . . . Clap and ing mechanism, [7] . . . . . . . . . . . . . . . . . . . . . . . . Wing rotation over one apping cycle, [8] . . . . . . . . . . . . . . . . . . DelFly, TU Delft, www.photoscala.de . . . . . . . . . . . . . . . . . . . . i-Bird, Silverlit, http://www.woti.de/images/ibird2.jpg . . . . . . . . . . . . First design of a apping wing mechanism at ASL, ETH Zurich . . . . . . 4 4 5 5 6 6 7

Robotic Insect, Harvard University, http://wyss.harvard.edu/viewmedia/82/roboticy

Flapping wing mechanism with dened angle for equation . . . . . . . . . 10 Flapping wing mechanism: Direct actuation . . . . . . . . . . . . . . . . . 10 Flapping wing mechanism: Actuation with mechanism . . . . . . . . . . . 11 Flapping wing mechanism: Combination with torsional spring . . . . . . . 11 Actuation with mechanism: Relevant forces and parameters . . . . . . . . . 12 Actuation with torsional spring: Relevant forces and parameters . . . . . . 13 Simulation of concept A: drift of initial value. . . . . . . . . . . . . . . . . 14 Simulation of concept B: steady apping with predened angle. . . . . . . 15 Simulation of concept B: unsteady apping with undened angle. . . . . . 15

3.10 Four-Wing Concept. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 3.11 Micro DC motor with 6mm diameter and 15mm length . . . . . . . . . . . 17 4.1 4.2 4.3 4.4 4.5 4.6 4.7 4.8 4.9 Wing motion of a y in hovering ight, [8] . . . . . . . . . . . . . . . . . . 19 Different shapes of wings, source: http://www.news.wisc.edu/; naturefocused.com . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 Implementation of y wings . . . . . . . . . . . . . . . . . . . . . . . . . 20 Implementation of hummingbird wings . . . . . . . . . . . . . . . . . . . 20 Two wings made of one piece . . . . . . . . . . . . . . . . . . . . . . . . . 21 The main structure with predened axis . . . . . . . . . . . . . . . . . . . 21 The main structure with the distances between the gear axis . . . . . . . . . 22 The main structure with all components . . . . . . . . . . . . . . . . . . . 22 Wing parameters: Chord length and angle . . . . . . . . . . . . . . . . . . 23

4.10 Testing of the y inspired design . . . . . . . . . . . . . . . . . . . . . . . 24 4.11 Testing of the hummingbird inspired design . . . . . . . . . . . . . . . . . 24 4.12 Plot of a lift force measurement . . . . . . . . . . . . . . . . . . . . . . . . 25 59

List of Figures

60

4.13 Plot of a torque measurement (yaw) to derive apping frequency . . . . . . 25 5.1 5.2 5.3 5.4 5.5 5.6 5.7 5.8 Shape of the optimal wing . . . . . . . . . . . . . . . . . . . . . . . . . . 27 Results from the measurements . . . . . . . . . . . . . . . . . . . . . . . . 28 Connection between upper and lower part . . . . . . . . . . . . . . . . . . 29 The motor controller MotorBee . . . . . . . . . . . . . . . . . . . . . . . . 29 The joypad Rumblepad, sysprole.de . . . . . . . . . . . . . . . . . . . . . 29 Linear dependency for input and lift force generation . . . . . . . . . . . . 30 Unsteady dynamics in torque generation . . . . . . . . . . . . . . . . . . . 31 Proof of capability of hovering ight . . . . . . . . . . . . . . . . . . . . . 31

A.1 Results: Gear-ratio 27, Amplitude 135 . . . . . . . . . . . . . . . . . . . 35 A.2 Results: Gear-ratio 27, Amplitude 135 . . . . . . . . . . . . . . . . . . . 36 A.3 Results: Gear-ratio 27, Amplitude 100 . . . . . . . . . . . . . . . . . . . 37 A.4 Results: Gear-ratio 27, Amplitude 100 . . . . . . . . . . . . . . . . . . . 38 A.5 Results: Gear-ratio 27, Amplitude 100 . . . . . . . . . . . . . . . . . . . 39 A.6 Results: Gear-ratio 27, Amplitude 100 . . . . . . . . . . . . . . . . . . . 39 A.7 Results: Gear-ratio 27, Amplitude 65 . . . . . . . . . . . . . . . . . . . . 40 A.8 Results: Gear-ratio 27, Amplitude 65 . . . . . . . . . . . . . . . . . . . . 41 A.9 Results: Gear-ratio 27, Amplitude 655 . . . . . . . . . . . . . . . . . . . 42 A.10 Results: Gear-ratio 20, Amplitude 135 . . . . . . . . . . . . . . . . . . . 43 A.11 Results: Gear-ratio 20, Amplitude 135 . . . . . . . . . . . . . . . . . . . 44 A.12 Results: Gear-ratio 20, Amplitude 100 . . . . . . . . . . . . . . . . . . . 45 A.13 Results: Gear-ratio 20, Amplitude 100 . . . . . . . . . . . . . . . . . . . 46 A.14 Results: Gear-ratio 20, Amplitude 100 . . . . . . . . . . . . . . . . . . . 47 A.15 Results: Gear-ratio 20, Amplitude 100 . . . . . . . . . . . . . . . . . . . 48 A.16 Results: Gear-ratio 20, Amplitude 65 . . . . . . . . . . . . . . . . . . . . 49 A.17 Results: Gear-ratio 20, Amplitude 65 . . . . . . . . . . . . . . . . . . . . 50 A.18 Results: Gear-ratio 16, Amplitude 135 . . . . . . . . . . . . . . . . . . . 51 A.19 Results: Gear-ratio 16, Amplitude 135 . . . . . . . . . . . . . . . . . . . 52 A.20 Results: Gear-ratio 16, Amplitude 100 . . . . . . . . . . . . . . . . . . . 53 A.21 Results: Gear-ratio 16, Amplitude 100 . . . . . . . . . . . . . . . . . . . 54 A.22 Results: Gear-ratio 16, Amplitude 65 . . . . . . . . . . . . . . . . . . . . 55 A.23 Results: Gear-ratio 16, Amplitude 65 . . . . . . . . . . . . . . . . . . . . 56 B.1 Simulink block diagram for concept A . . . . . . . . . . . . . . . . . . . . 57 B.2 Simulink block diagram for concept B . . . . . . . . . . . . . . . . . . . . 58 B.3 Simulink block diagram for concept C . . . . . . . . . . . . . . . . . . . . 58 B.4 Simulink block diagram for concept C . . . . . . . . . . . . . . . . . . . . 58

Bibliography
[1] M ADANGOPAL , R., K HAN , Z. A., AGRAWAL , S. K.: Biologically inspired design of small apping wing air vehicles using four-bar mechanisms and quasi-steady aerodynamics, Journal of Mechanical Design, Vol. 127, No.4, 2005, 809-816.

[2] J IANG H AO W U , M AO S UN: Unsteady aerodynamic forces of a apping wing, Journal of Experimental Biology 207, 2004, 1137-1150.

[3] G HEORGHE B UNGET: A Bio-inspired Micro-Air Vehicle for Flapping Flight, saarbr cken, Germany, 2008. u

[4] M ICHAEL H. D ICKINSON: The effects of wing rotation on unsteady aerodynamic performance at low reynolds numbers, Journal of Experimental Biology 192, 179-206, 1994.

[5] R. J. W OOD: Design, fabrication, and analysis of a 3DOF, 3cm apping-wing MAV, Proceedings of the 2007 IEEE/RSJ International Conference on Intelligent Robots and Systems San Diego, CA, USA, Oct 29 - Nov 2, 2007.

[6] S ANJAY P. S ANE: The aerodynamics of insect ight, The Journal of Experimental Biology 206, 4191-4208, 2003.

[7] T. W EIS -F OGH: Unusual mechanisms for the generation of lift in ying animals, Scienti c American, 233:80 87, 1975.

[8] F RITZ -O LAF L EHMANN , S IMON P ICK: The aerodynamic beneft of wing wing interaction depends on stroke trajectory in apping insect wings, The Journal of Experimental Biology 210, 1362-1377, 2007.

61

Bibliography

62

[9] S TEFAN G ISLER: Unsteady aerodynamics and control issues in apping ght. Studies on mechatronics, Autonomous Systems Lab (ASL), Swiss Federal Institute of Technology Zurich (ETH), 2008.

[10] O LIVER B REITENSTEIN: Development of a Flapping Wing Mechanism. Semester Project, Autonomous Systems Lab (ASL), Swiss Federal Institute of Technology Zurich (ETH), 2009.

[11] M ANUEL NAEF: Design of a Flapping Wing Mechanism. Semester Project, Autonomous Systems Lab (ASL), Swiss Federal Institute of Technology Zurich (ETH), 2009.

You might also like