EM_11_008_-_Ditten-van_-_MSc_
Department of Precision and Microsystems Engineering
Torque Sensing for E-bike Applications
Name:
Marien van Ditten
Report no:
EM 11.008
Coach:
Ir. E.J.H. de Vries
Professor:
Prof. Dr. Ir. E.G.M. Holweg
Specialisation:
Automotive
Type of report:
Date:
Master Thesis
June 27, 2011
TORQUE SENSING FOR E-BIKE APPLICATIONS
MASTER THESIS
Marien van Ditten
June 27, 2011
Delft University of Technology
Faculty of Mechanical, Maritime and Materials Engineering
CONFIDENTIAL
The information in this report is property of SKF ADC-SI.
It cannot be divulged or reproduced without company's authorization.
The work in this thesis was supported by SKF.
Their cooperation is hereby gratefully acknowledged.
Abstract
The Automotive Development Center of SKF was interested in the possibility to develop a torque sensor, either for e-bike or power steering application.
A master thesis project was initiated to explore the possibilities. The results
of that study are presented in this thesis.
In the first phase of the study, a benchmark was made of the existing
technology and the requirements for the product. Market research showed
that a sensor for e-bike applications would be the most promising application
in the short term. The analysis of the state of the art showed that the
prevailing technologies are magnetic field measurement and magnetic phase
shift measurement. One of the existing products, a torque sensor integrated
in a bottom bracket from Thun, was analyzed. To gain more insight into the
requirements of the product, the Quality Function Deployment method was
used. To identify the most important needs and functions of the product
a paired comparison matrix and a House of Quality were used respectively.
The House of Quality showed that, next to torque measurement, price and
safety are also important functions.
The second phase concerned the actual design of the sensor. First of all
the general concept of the sensor was developed by answering three basic
questions: where should the sensor be located, what should it measure and
how should it measure. The best concept proved to be a sensor located in the
bicycle’s bottom bracket (the part that connects the pedals to the frame and
contains the pedal spindle). The applied torque will cause a torsion angle
between the chain wheel and the spindle. Angled surfaces between the chain
wheel and a sleeve fitted around the spindle will cause a translation of that
sleeve. On the sleeve an aluminium target ring is fitted, which moves in a
time-depended magnetic field. By measuring the influence on the magnetic
field caused by the induced eddy current in the ring, the position of the
ring can be determined and thereby the applied torque can be calculated.
All relevant parameters of the concept were analyzed by simulations of the
TU Delft
SKF ADC-SI
viii
sensor setup in Ansoft Maxwell, a finite element electromagnetic field simulation tool. With the results of these simulations a torque sensing bottom
bracket was designed.
In the third phase, the concept was analyzed and tested. To make the
Maxwell simulation more usable, an interface was made using MATLAB and
Visual Basic scripts. With the interface the user is able to change parameters quickly and perform batch analyses. With the help of this interface, a
Design of Experiment analysis was performed in order to identify the most
influential parameters and optimize the geometry. It showed there is an
important connection between the width of the target ring and the space
between the coils. A full size prototype was made and tests were performed
on the measurement principle. These test showed a linear relation between
the displacement of the target ring and the output of the sensor. The sensor
has a sensitivity of 120 mV/mm. The tests also showed that the set-up is
sensitive for errors in the position of the target ring. Static tests on the
prototype showed that the sensor concept functions. There was no time to
perform dynamic tests. These are scheduled in the near future.
The concept torque sensing bottom bracket shows promising first test
results. But without sufficient test data, it is not yet possible to give a final
verdict. When it is decided to continue the project, several points should be
addressed. These are: the design of the intermediate part; the mechanical
assembly; the cost price; the sensation to the cyclist and the integration of
the electronics.
TU Delft
SKF ADC-SI
Contents
Preface
xvii
Introduction
xix
Part I
Benchmarking
1
Market Research
1.1 The Market of E-bikes . . . . . . . . . . . . . . . . . . .
1.2 The Market of Power Steering Systems . . . . . . . . . .
1.3 Conclusion Market Research . . . . . . . . . . . . . . . .
2
Competitor Analysis
2.1 Magnetic Rotor-Stator Phase Shift Measurement
2.2 Magnetic Rings Phase Shift Measurement . . . .
2.3 Mechanical Phase Shift Measurement . . . . . .
2.4 Magnetic Field Measurement . . . . . . . . . . .
2.5 Magnetic Translation Measurement . . . . . . .
2.6 Optical Phase Shift Measurement . . . . . . . .
2.7 Displacement of Rear Drop Out . . . . . . . . .
2.8 Forces on the Bottom Bracket Bearing . . . . . .
2.9 Summary of Technologies . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
3
3
6
7
9
9
10
11
12
14
15
16
17
18
3
Analysis of Purchased Parts
21
3.1 Thun Bottom Bracket Sensor . . . . . . . . . . . . . . . . 21
3.2 Ergomo Bottom Bracket Sensor . . . . . . . . . . . . . . 26
3.3 Integrative Motor Kit . . . . . . . . . . . . . . . . . . . . 27
TU Delft
SKF ADC-SI
x
Contents
4
Part II
Quality Function Deployment
4.1 Needs . . . . . . . . . . .
4.2 Paired Comparison Matrix
4.3 Functions . . . . . . . . .
4.4 House of Quality #1 . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Concepts for a Torque Sensor for
5.1 Where - Location of Sensor . .
5.2 What - Physical Quantity . . .
5.3 How - Measurement Principle .
6
Design of Sensing Elements
6.1 Eddy Current & Variable Reluctance . . . . .
6.2 Concept of Sensor . . . . . . . . . . . . . . .
6.3 Modelling of Sensor Using Ansoft Maxwell . .
6.4 Results of Electromagnetic Field Simulations
6.5 Conclusions of Simulations . . . . . . . . . .
Part III
8
9
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Sensor Design
5
7
.
.
.
.
29
29
31
32
32
E-bike Application
39
. . . . . . . . . . . . . . . 39
. . . . . . . . . . . . . . . 40
. . . . . . . . . . . . . . . 44
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
49
49
52
53
56
65
Detailed Design of Torque Sensing Bottom Bracket
67
7.1 Mechanical Design . . . . . . . . . . . . . . . . . . . . . 67
7.2 Electrical Design . . . . . . . . . . . . . . . . . . . . . . . 72
Simulation & Testing
BoB SimControl for Torque
8.1 Refining the Mesh . . . .
8.2 Visual Basic script . . . .
8.3 Structure of Program . .
8.4 Results . . . . . . . . . .
Sensing Bottom Bracket
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
. . . . . . . . . . . . . . .
Design of Experiment Analysis
9.1 Design of Experiment . . . . . . . . . . . .
9.2 Build Up of Simulation Model . . . . . . .
9.3 Phase 1: Parameter Screening . . . . . . .
9.4 Phase 2: Parameter Optimization . . . . .
9.5 Conclusions Design of Experiment Analysis
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
75
75
76
77
81
.
.
.
.
.
83
83
84
85
86
89
10 Assembly & Test of Prototype Bottom Bracket
91
10.1 Assembly of Prototype . . . . . . . . . . . . . . . . . . . 91
10.2 Test of Measurement Principle . . . . . . . . . . . . . . . 93
10.3 Static Tests . . . . . . . . . . . . . . . . . . . . . . . . . 95
TU Delft
SKF ADC-SI
CONTENTS
xi
10.4 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . .
97
11 Conclusions and Discussion
99
11.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 99
11.2 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 100
Part IV
Appendices
A
Competitor Analysis Matrix
105
B
Thun Bottom Bracket Sensor Data Sheet
109
C
Paired Comparison Matrix
113
D
House of Quality #1
115
E
SVP BoB SimControl
117
F
Design of Experiment
F.1 DoE Screening 1 . . . . . . . . . . . . . . . . . . . . . .
F.2 DoE Screening 2 . . . . . . . . . . . . . . . . . . . . . .
F.3 DoE Full Factorial . . . . . . . . . . . . . . . . . . . . . .
121
121
122
122
Bibliography
TU Delft
123
SKF ADC-SI
xii
TU Delft
Contents
SKF ADC-SI
List of Figures
1.1
1.2
Products from competitors . . . . . . . . . . . . . . . . . . . .
ION technology e-bikes . . . . . . . . . . . . . . . . . . . . . .
2.1
2.2
2.3
2.4
2.5
2.6
2.7
2.8
2.9
Torque sensor using a magnetic rotor and stator assembly
Drawing of torque sensor using magnetic rings . . . . . .
Drawing of mechanical phase shift torque sensor . . . . .
Inverse magnetostrictive effect torque sensor . . . . . . .
Examples of application of NCTE torque sensor . . . . .
Honda Accord EPAS system torque sensor . . . . . . . .
Optical phase shift measurement . . . . . . . . . . . . .
Displacement sensor in rear drop out . . . . . . . . . . .
Force measurement in bottom bracket . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
11
11
12
14
15
15
16
17
18
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
Thun bottom bracket sensor . . . . . . . . . . . . .
Output of Thun BB sensor during measurement . . .
Exploded view of the disassembled Thun BB sensor .
Magnetic ring from the Thun BB sensor . . . . . . .
Measurement of the magnetized region on the spindle
PCB from Thun BB sensor . . . . . . . . . . . . . .
Ergomo bottom bracket sensor . . . . . . . . . . . .
Integrative motor kit . . . . . . . . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
22
23
24
24
25
25
26
28
4.1
4.2
4.3
4.4
4.5
Work flow of Quality Function Deployment . . . . . . . . . .
Example of paired comparison matrix . . . . . . . . . . . . . .
Results of paired comparison matrix analysis for torque sensor
Example of a House of Quality . . . . . . . . . . . . . . . . .
Results of House of Quality #1 for torque sensor . . . . . . .
.
.
.
.
.
30
32
34
35
36
5.1
Schematic mapping of bicycle . . . . . . . . . . . . . . . . . . .
41
TU Delft
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
5
5
SKF ADC-SI
xiv
List of Figures
5.2
5.3
5.4
Concept Bottom Bracket with flexible connection . . . . . . . .
Schematic drawing of concept magnetic rings & duty cycle . . .
Schematic drawing of concept variable reluctance . . . . . . . .
42
44
46
6.1
6.2
6.3
6.4
6.5
6.6
6.7
6.8
6.9
6.10
6.11
6.12
6.13
6.14
6.15
Eddy current proximity sensor. . . . . . . . . . . . . . . . . . .
Magnetic circuit caused by a coil with an air gap . . . . . . . .
Differential variable reluctance displacement sensor . . . . . . .
Cross section of sensing concept . . . . . . . . . . . . . . . . .
Modelling of sensor in Ansoft Maxwell . . . . . . . . . . . . . .
External circuit imported in Maxwell simulation . . . . . . . . .
Magnetic Flux density using different materials . . . . . . . . .
Induced eddy current in the target ring in steel and aluminium .
Effect of thickness (axial) of target ring on the sensitivity . . . .
Effect of coil spacing on the sensitivity . . . . . . . . . . . . . .
Magnetic flux density of sensor with ferrite core . . . . . . . . .
Influence of different frame tubes on the sensor . . . . . . . . .
New model of torque sensing bottom bracket . . . . . . . . . .
Output of the Wheatstone bridge versus displacement of the sleeve
Target ring and coils with current magnitude and direction . . .
50
51
52
53
54
55
57
57
58
59
59
61
63
64
64
7.1
7.2
7.3
7.4
7.5
CAD design of the SKF Torque Sensing Bottom Bracket . . .
Connection of chain wheel bracket to the spindle . . . . . . .
Intermediate piece connecting spindle and chain wheel bracket
Mechanism for transforming the rotation to translation . . . .
Electronic circuit for the prototype . . . . . . . . . . . . . . .
.
.
.
.
.
68
69
70
71
72
8.1
8.2
8.3
8.4
8.5
8.6
2-D Maxwell model of torque sensing bottom bracket
Flow chart of the main structure of the program . . .
Example of recorded Visual Basic script . . . . . . .
Architecture of the MATLAB program . . . . . . . .
Screen shot of MATLAB program (BoB SimControl)
Result of a batch simulation . . . . . . . . . . . . .
.
.
.
.
.
.
76
78
78
80
81
82
9.1
9.2
9.3
9.4
9.5
Bold approach of parameter variation in screening phase of DoE
Build up of analysis model showing the four main parts . . . . .
Result of screening phase of the DoE . . . . . . . . . . . . . . .
Results of the second phase of the DoE, varying two parameters
Results from Minitab of the second phase of the DoE . . . . . .
84
85
87
88
90
10.1
10.2
10.3
10.4
10.5
10.6
Mold for fabricating the intermediate part . . . . . . . . . . . .
Complete assembled prototype . . . . . . . . . . . . . . . . . .
Test set-up for static test of measurement principle. . . . . . . .
Output of sensor when moved around the center between the coils
Influence of eccentricity of the target ring on the output voltage
Output of sensor during manipulation of the prototype . . . . .
92
92
93
95
96
97
TU Delft
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
SKF ADC-SI
List of Tables
2.1
Summary of technologies . . . . . . . . . . . . . . . . . . . . .
19
4.1
4.2
Critical to Satisfaction for torque sensor product . . . . . . . . .
Critical to Quality . . . . . . . . . . . . . . . . . . . . . . . . .
31
33
5.1
5.2
5.3
Locations and its possible physical quantities . . . . . . . . . . . 43
Advantages and disadvantages of different measurement methods 46
Multi criteria table . . . . . . . . . . . . . . . . . . . . . . . . . 47
6.1
6.2
6.3
6.4
Sensitivity
Sensitivity
Sensitivity
Sensitivity
8.1
Sensitivity of sensor calculated using 3-D and 2-D Maxwell model 77
9.1
9.2
Parameters for the screening phase of design of experiment analysis 85
Parameter range and interval for DoE phase 2 . . . . . . . . . . 87
of
of
of
of
sensor
sensor
sensor
sensor
using different materials as target ring
with ferrite core . . . . . . . . . . . .
with different kind of frame tubes . .
with reduced air gap . . . . . . . . . .
.
.
.
.
.
.
.
.
.
.
.
.
10.1 Oscillation of measurement signal caused by mounting of target
ring. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
TU Delft
56
58
60
60
94
SKF ADC-SI
xvi
TU Delft
List of Tables
SKF ADC-SI
Preface
When I look back to September of 2010 I did not know what to expect.
People said that it would be difficult and tiring to work abroad and speak
a different language, and they were right. They also said that all the effort
you would invest in the beginning would pay out at the end, and they were
right again.
I was going to live and work in France. The country where, at least to
the Dutch clichés, everybody is always on strike, nobody speaks English,
everybody is lazy and nobody can drive a car. After 10 months I can report
that most of it is not true.
But more importantly, it showed me the subtle difference between two
cultures which are only a few hundred kilometers apart. A valuable lesson
learned in a profession which is more and more global.
I would like to thank Alberto Carlevaris, for giving me the opportunity to
come to the ADC-SI team, Olivier Joubert, my Maı̂tre de Stage, who had
confidence in me and gave me the space to follow my own path, and Mathieu
Hubert, my day-to-day mentor, for pointing me in the right direction and
answering all my silly questions.
The ‘Torque’ team further consisted of Vincent Sausset (mechanical engineer), Matthieu Rioteau (electrical engineer), Alan Roué (quality) and
Alexis Gatesoupe (project manager). I would like to thank them all for
their work and help.
And I would like to thank all the other people in ADC-SI, the team
within SKF where I spend 10 months on this project. They showed me that
the persistent cliché that that French people are lazy is certainly untrue.
Merci à tous, pour votre accueil chaleureux, votre patience quand j’étais
aux prises avec la langue française, et tous les jeudis soirs !
I would also like to thank Edward Holweg, who made it possible for me to
do my assignment at SKF in France, and Edwin de Vries, who made sure
that everything was running smoothly in Delft.
TU Delft
SKF ADC-SI
xviii
TU Delft
Preface
SKF ADC-SI
Introduction
Torque Sensing for E-bike Applications
Thousands of people face the same dilemma every day: do I go by car, or do
I take the bike? And too often people choose the car. Whether the argument
is that there is a strong head wind in the Dutch polder, the groceries that
have to be carried are too heavy or there is a steep incline on the way to
work, it almost always can be attributed to the fact that it takes too much
effort. For this reason, a lot of small journeys are made by car, leading to
the emission of harmful gases into the atmosphere.
In the last ten years a strong increase in the sales of electrical assisted
bicycles has taken place. These e-bikes (sometimes called pedelec) will assist
the cyclist with an electric motor to overcome the arguments as stated above.
People will use the bike more frequent and over longer distances than a
traditional bike. Therefore, urban mobility can be increased and emissions
can be reduced. Not to mention the advantages to public health.
To be able to control the assistance provided by the electrical motor, it
is important to measure the cyclist’s effort. One way of doing that is by
measuring the torque applied to the pedals. SKF sees opportunities in the
e-bike market and has launched a project to investigate the possibilities.
Structure of the Thesis
This master thesis is the result of a nine month project at the Automotive
Development Center - Senor Integration department of SKF. The thesis is
split up in three parts: Benchmarking, Sensor Design and Simulation &
Testing.
In Part I, the market of e-bikes and torque sensing is investigated. A
market research is done in chapter 1. The existing competition is analyzed
in chapter 2. Several products that seemed interesting are purchased for
analysis in chapter 3. At the end of the benchmarking phase a Quality
TU Delft
SKF ADC-SI
xx
Introduction
Function Deployment is done in chapter 4 to translate the voice of the
customer into clear design requirements.
Part II covers the path from brainstorming to the final concept. First,
in chapter 5, several ways of torque sensing are considered, and finally one
concept is chosen for further analysis. In the following chapter (chapter
6) this concept is simulated using finite element calculations to investigate
the influence of several design parameters in order to determine the optimal
geometry. At the end of part II, the final mechanical and electrical concept
are presented in chapter 7.
In Part III, the design is evaluated using simulations and testing. Before the simulations, a MATLAB program is written as an interface for the
simulations. This program is described in chapter 8. With the help of this
program the concept can be analyzed to find the best parameter combination. A Design of Experiment was performed, the analysis can be found in
chapter 9. In chapter 10 the prototype of this concept will be tested. After
that, the thesis will be concluded with the conclusions and recommendations
in chapter 11.
SKF and the Automotive Development Center
SKF Group is the leading global supplier of products, solutions and services
within rolling bearings systems. SKF stands for Svenska Kullagerfabriken
(Swedish ball bearing factory). The company started in 1907 and grew
quickly into a global player. Nowadays, the company has over hundred
manufacturing sites and is represented in more than 130 countries. The
company specialized in more than just bearings. As shown, SKF is built on
five platforms: bearings, seals, lubrication, mechatronics and services. SKF
employs over 44,000 people worldwide.
The company consists of thee main divisions: Industrial, Services and Automotive. The first two focus on servicing industrial original equipment manuTU Delft
SKF ADC-SI
xxi
facturers (OEMs) and aftermarket customers respectively. The automotive
division focuses on the automotive OEMs and aftermarket customers.
Part of the automotive division is the Automotive Development Center
(ADC). This department develops products, not only bearings, that can be
used in the automotive sector. Within the department there is a high level
of mechatronics engineering. The unit Sensor Integration (ADC-SI), located
at the production facility in Saint-Cyr-sur-Loire, France, mainly focuses on
the integration of sensors in bearings and other products. The ADC-SI
team consists of 25 people, containing mechatronic engineers, mechanical
engineers, electrical engineers and project managers.
TU Delft
SKF ADC-SI
xxii
TU Delft
Introduction
SKF ADC-SI
Part I
Benchmarking
TU Delft
SKF ADC-SI
CHAPTER
1
Market Research
The market research is split in two parts, e-bikes (section 1.1) and EPS
systems (section 1.2), because the two fields are very different. Afterwards,
in section 1.3, a conclusion is drawn to decide which field has the most
possibilities and opportunities for SKF and thus what the focus of the project
should be.
1.1 The Market of E-bikes
Trends
With the ban of motorcycles in almost ninety Chinese cities, late 1990s,
the annual sales of electric bikes grew from 56,000 in 1998 to 20 million in
2008 [1]. In Europe, with the introduction of inner city congestion charges,
growing delay due to city traffic and a growing environmental lobby, the sale
of e-bikes is also on the rise. In the Netherlands, for instance, the sales of
e-bike has quadrupled to 153,000 between 2006 and 2009. Nowadays one in
eight bikes sold is an e-bike, even if these bikes are on average three times
as expensive as a normal bike [2].
E-bikes are sold in two ways: as a complete bike, designed from the
ground up as an e-bike, or as a retrofit kit. Both consist of the same basic
parts:
• A motor, either a hub motor in the front or the rear wheel or a frame
mounted motor driving the chain;
• A battery pack, providing the power for the bike. Most battery
packs can be detached from the bike to charge them with a separate
charger;
TU Delft
SKF ADC-SI
4
Market Research
• A handle bar mounted control panel, to operate the system,
choose assistance mode, etc.;
• A controller, to control the operation of the system;
• A sensor, to monitor movement of the pedals (in European systems).
Contrary to Asia, where e-bikes are sometimes more like electric scooters
than a bike, in Europe e-bikes have to obey some strict rules. One of the most
important rules is that the bike is not allowed to be electrically propelled if
the cyclist is not pedaling or pedaling backwards.
To control this, a simple cadence sensor could be sufficient. But to
deliver a superior mode of electric assistance control, a torque sensor can be
used. The sensor measures the effort the cyclist is putting into the bike, and
with that information the amount of assistance from the electric motor can
be controlled. The market for intelligent e-bikes is not yet mature. There
are several systems on the market to measure torque, but there are still
opportunities to conquer a part of the market.
At the moment, the prevailing technology used is magnetic field measurement, using the reversed magnetostrictive effect (section 2.4). This technology can be easily integrated into the bottom bracket, and does not require
any modification to the bike frame.
Competitors
At the moment most e-bikes are either produced in high volume in China, or
by European or North American bike manufacturers as high-end products.
The torque sensor is more applicable for the high-end products. The bike
manufacturers do not develop the sensor themselves because they lack the
know-how or the experience. In the last year, a lot of new companies entered
the e-bike market. The German company Schaeffler Technologies introduced
through its bearing manufacturer FAG a torque sensing bottom bracket.
The same was done by Thun, a bottom bracket manufacturer for more
than forty years. Shimano, a global manufacturer of high quality bicycle
parts, announced its Shimano Total Electric Power System (STEPS, figure
1.1a), containing all the necessary parts to construct an e-bike. During the
summer automotive giant Bosch introduced its vision on the e-bike market
with their eBike system (figure 1.1b). Bike manufacturers themselves are
also busy developing technology. The Accell group, which contains bicycle
brands like LaPierre, Sparta and Koga Miyata, has its own technology, called
ION (figure 1.2). These companies can be seen as competitors to SKF.
TU Delft
SKF ADC-SI
1.1 The Market of E-bikes
(a)
5
(b)
Figure 1.1: (a) Shimano STEPS [3] (b) Bosch eBike. [4]
(a)
(b)
Figure 1.2: (a) Example of an e-bike with ION technology from Sparta [Sparta] (b)
schematic view of the ION technology with a rear wheel hub motor and battery pack
integrated into the frame. [5]
TU Delft
SKF ADC-SI
6
Market Research
Possible Market Gap
At the moment, the e-bike market is still wide open. Not a lot of products
are available. If SKF gets in the market of e-bikes before the market is
saturated, there is a chance that SKF can take a leading role.
1.2 The Market of Power Steering Systems
Trends
More and more cars are equipped with electrical power steering systems
(EPS). EPS systems have several advantages over traditional hydraulic systems:
• More efficient: The hydraulic pump has to run all the time, even
when there is no steering action. This drains power from the engine
constantly. Eliminating the constant power demand of a hydraulic
system will reduce the fuel consumption up to 5%. [7]
• Less space: The EPS system is more compact than a traditional
system, making packaging easier and also makes it a suitable option
for smaller cars.
• Less weight: The EPS system is lighter than a traditional system,
which will boost efficiency even more.
• Better steering action: Because the system is assisted by electric
motors, it is able to be tuned to the wishes of the manufacturer and/or
consumer.
To control the EPS system, measurement of steering angle and steering
effort (torque applied by the driver) should be measured. Most vehicles
are already equipped with steering angle sensors. It would be beneficial to
integrate the position sensing and torque sensing into one sensor.
The market is in a mature state, as can be seen in chapter 2. Several
manufactures offer torque sensors, using various technology. The technologies that are mostly used are contactless and based on magnetics. The two
most used technologies are magnetic rotor-stator phase shift measurement
(section 2.1) and magnetic field measurement (section 2.4). Both technologies require modifications to the steering shaft. The first technology requires
a torsion bar, the second requires a magnetic region applied into or onto the
shaft.
Competitors
Most car manufacturers do not produce their steering systems themselves.
They buy complete steering racks from automotive suppliers like ZF and
TU Delft
SKF ADC-SI
1.3 Conclusion Market Research
7
Valeo. These systems contain the torque sensor. Some suppliers develop
their sensors themselves, some use sensors from companies like Bosch or
Bourns. These companies can be seen as competitors to SKF.
Possible partners in the development of a torque sensor for EPS applications can be companies that develop only the technology. Afterwards they
license the patents to other companies. Companies that develop technology
for steering sensors are MMT and NCTEngineering. Another possibility is
to work together with a manufacturer of steering systems that does not develop its own sensor, like ZF. SKF and ZF already had an earlier project on
this subject.
Possible Market Gap
The market for steering sensors is already mature. This means that it will
be hard to convince car manufacturers to change their existing sensor for
the new sensor from SKF. Unless SKF comes up with a design that is better
in either performance or price.
1.3 Conclusion Market Research
Because the two fields of application of the torque sensor are quite different
with regards to the specification, a choice has to be made. The duration of
the project does not allow to develop a torque sensor for both applications in
parallel. Therefore, a strategical decision was made to focus the project on
the development of a torque sensor for e-bike applications first. Afterwards,
a study can be done to determine whether the technology can be used for
EPS applications, if desired.
TU Delft
SKF ADC-SI
8
TU Delft
Market Research
SKF ADC-SI
CHAPTER
2
Competitor Analysis
The competitor analysis was done by using the internet, patent databases
and internal knowledge. During the analysis, it turned out that there is a lot
of overlap in the technologies used by different companies. Therefore, this
section is arranged according to the technologies used. All the specifications
(if provided) of the sensors are collected in the competitor analysis matrix
in appendix A.
Some sensors are actually not real torque sensors. These sensors measure
the phase shift between an input and an output shaft which are connected
with a torsion bar. The measured phase shift between the two shafts is
equal to the torsion angle of the torsion bar. With the material properties,
the dimension of the bar and the torsion angle the applied torque can be
calculated. But for the fate of simplicity, these phase shift sensors are also
called torque sensors in this report, because they are designed specifically
to serve that purpose.
Although the project is now focused on e-bike applications, it is still useful to look at solutions used in the EPS field. Some measurement principals
for EPS sensors might be adaptable to use on an e-bike application.
2.1 Magnetic Rotor-Stator Phase Shift Measurement
The first and probably most used measurement principal in power steering
applications is the measurement of the phase shift between the input and
the output shaft by using a magnetic rotor and a ferro-magnetic stator assembly. The input shaft (connected to the steering wheel) and the output
shaft (connected to the steering rack) are connected through a torsion bar.
To keep the assembly compact, the torsion bar is usually situated within
TU Delft
SKF ADC-SI
10
Competitor Analysis
either the (hollow) input or output shaft. This way, the ends of the input and output shaft can be situated close to each other, through which a
considerable phase shift can occur, as can be seen in figure 2.1a.
On one of the shafts (input or output shaft, there is no difference in
application) a multi polar magnetic ring is fitted. This ring contains a
plurality of alternating north and south poles. On the other shaft, a stator
assembly is fitted. This stator assembly usually contains two rings of ferromagnetic material. Both rings are situated around the shaft and have teeth
that cover the magnetic ring. The teeth of both rings lock in to each other,
but don not touch, and cover the magnetic rotor. There is the same number
of teeth per stator ring as there are pole pairs on the rotor. The teeth of
one ring cover the north poles of the rotor, the teeth of the other ring cover
the south poles.
In this setup, the magnetic flux of the magnets on the rotor is guided
into the stators. If a sensor that is able to measure the magnetic flux, for
instance a Hall Effect sensor, is placed between the two stator rings, the
magnetic flux can be measured, as is illustrated in figure 2.1b. When a
torque is applied between the input and output shaft, the rotor and stator
rotate relative to each other and so the positions of the magnets change.
Therefore, the magnetic flux between the stator rings changes, and can be
measured by the Hall sensor. To amplify the signal at the location of the
Hall sensor, flux concentrators are used. These are ferro-magnetic parts
which guide the magnetic flux in the stator rings to one specific place, close
to the Hall sensor.
This way, the phase shift between the input and output shaft can be
determined. If the stiffness of the torsion bar is known, the applied torque
can be calculated.
The technology is only used in EPS applications. A possible explanation
is that this kind of assembly requires a considerable amount of space in the
radial direction. Therefore, it might be hard to implement it on an e-bike
application.
Companies that use this technology are: Moving Magnet Technologies
(MMT), Bosch, Continental, NSK and Valeo.
2.2 Magnetic Rings Phase Shift Measurement
At the end of the input shaft and the output shaft a ring with a plurality
of magnets is fixed (figure 2.2). Both rings have the same number of pole
pairs and are aligned in a way that on both rings the north and south poles
are parallel. The input and the output shaft are connected with a torque
coupling. The torque coupling is designed to flex when torque is applied.
This results in a phase shift and a change in direction of the magnetic fields
between the two rings. In between the two rings, an AMR sensor is placed
TU Delft
SKF ADC-SI
2.3 Mechanical Phase Shift Measurement
(a)
11
(b)
Figure 2.1: Drawing of a torque sensor using a magnetic rotor and stator assembly, (a)
with input and output shaft (1,3), torsion bar (2), stator assembly (10) and Hall sensor
(15). (b) Detail of sensor with multi polar magnetic ring (9), ferro magnetic stators
(11, 12) with flux concentrators (13, 14) and air gap for Hall sensor (16). [8]
Figure 2.2: Drawing of torque sensor using magnetic rings. Magnetic rings (510, 515)
and AMR sensor (525). [9]
which measures this direction. The direction of the magnetic field is thus the
measure for the phase shift, which is used to calculate the applied torque.
A patent application for this measurement principle, to be used in a
EPS system, was done by SSI Technologies, which was acquired by Bourns
inc., an automotive sensor manufacturer. The patent also describes a third
magnetic ring, which can be used to measure the rotor position.
2.3 Mechanical Phase Shift Measurement
The phase shift between the input and output shaft can also be measured
mechanically with planetary gear sets and the use of rotational sensors. To
the end of both input and output shaft a sun gear is fitted. These sun gears
are connected to their own planetary gears. These planetary gears are both
connected to a common planet carrier. Around the two planetary gears, two
separate ring gears are fitted. One of the ring gears is fixed to the sensor
housing and is stationary. The other ring gear is also toothed at the external
TU Delft
SKF ADC-SI
12
Competitor Analysis
Figure 2.3: Drawing of mechanical phase shift torque sensor with input and output
shaft (3, 4), torsion bar (5), sun gear (8, 9), planets (10, 11), common planet carrier
(12) stationary ring gear (13), rotating ring gear (14) and rotation sensor (16). [10]
circumference. Their teeth are connected to one or more rotation sensors.
This sensor is illustrated in figure 2.3.
When the two shafts rotate without a phase difference (i.e. no torque is
applied), the two sun gears will spin, also moving the planets. But because
they spin at the same speed, the ring gears remain stationary. When a torque
is applied, and thus a phase shift occurs, the sun gears will rotate relative
to each other. This will cause the two sets of planets to move differently.
Because the planets share the same planet carrier, the ring gear that is
not fixed to the body will have to move to compensate the difference of
movement of the planets. This rotation is measured by the rotation sensor,
and the applied torque can be calculated.
Bourns inc. has a patent application for this technology, but as far as
known, this technology is not available commercially.
2.4 Magnetic Field Measurement
Instead of measuring the phase shift of two shafts, the stress induced by the
application of torque can also be measured directly. If stress is applied to a
magnetic structure, the properties of the magnetic field change, and can be
measured.
This truly contactless measurement principal starts with a magnetic
shaft. This shaft can be magnetized itself, covered with a magnetic material or fitted with a magnetic sleeve.
The shaft is placed through the sensor body, without causing any friction
during the rotation of the shaft. In the vicinity of the shaft one or more
measurement coils are placed to pick up the magnetic field emanating from
TU Delft
SKF ADC-SI
2.4 Magnetic Field Measurement
13
the magnetic region of the shaft. The chance current in the coil can be the
measure for torque applied on the shaft. There are two different methods of
measurement: fluxgate technology and the inverse magnetostrictive effect.
Fluxgate
The sensing element contains an assembly of two coils, situated circumferentially, with a magnetic saturable material in between. An alternating current
is passed through one of the two coils, causing a time-dependent magnetic
field in the material between the coils. This time-dependent field induces
a current in the second coil, the measurement coil. When no other fields
are present, the output of the measurement coil should be the same as the
input of the first coil. The magnetic field from the magneto-elastic material
of the shaft is superimposed on this magnetic field. The changes in the field
emanating from the shaft (caused by stress) will change the induced current
in the measurement coil, which is a measure for the applied stress on the
shaft.
The Inverse Magnetostrictive Effect
This effect is the change of magnetic susceptibility when the material is
subjected to mechanical stress. On the load bearing shaft, a layer of magnetostrictive material is applied. The shaft is placed inside the sensor body,
which contains two coils. Between the coils there is no material, unlike
the fluxgate sensor where a saturable material is placed between the coils.
Through one coil, an alternating current is passed. The coil magnetizes the
magnetostrictive material with a time dependent magnetic field. That magnetic field in turn induces a current in the second coil. Again, when no load
is applied, the output of the measurement coil should be equal to the input
of the first coil. When a torque is applied to the shaft, the susceptibility of
the magnetostrictive material changes. This changes the magnetic field in
the layer, and therefore also the current induced in the second coil. Hence,
the applied torque can be measured.
This technology can be widely used for almost all torque transmitting shafts.
Because the torque transmitting shaft is directly the primary sensor part,
the only modification that has to be made to the shaft is the application
or attachment of a magnetic surface. The company NCTEngineering has
patented a method of direct magnetization of a shaft, which is probably
used by Thun to create a torque sensor for e-bike applications which is
analyzed in section 3.1. The German company Schaeffler also introduced a
e-bike torque sensor (figure 2.4), which is also used in the new Bosch eBike
product [6]. NCTEngineering also offers torque sensors themselves, for all
TU Delft
SKF ADC-SI
14
Competitor Analysis
Figure 2.4: Inverse magnetostrictive effect torque sensor integrated in a bottom bracket
from Schaeffler/FAG, used in the Bosch eBike system. With cranks (3, 5), chain wheel
(28), magnetic regions (11, 30) and coils (13). [11]
kinds of automotive applications as can be seen in figure 2.5. AAB and
Siemens VDO also offer this technology for automotive applications.
2.5 Magnetic Translation Measurement
The input and output shaft are connected through a torsion shaft. At the
location where the two ends of the shafts meet, a plastic sleeve is loosely
fitted around the shaft. On one shaft-end, the sleeve is connected with a
pin and is allowed to translate along the axial direction of the shaft through
a slit. The other shaft end has also a pin, but this pin moves through a
slit that is tilted 45◦ with respect to the axis of the shaft. Therefore, when
the two shafts rotate compared to each other the tilted slit will cause the
rotational movement to be converted to a translation of the plastic sleeve.
The sleeve contains an aluminium ring. The housing around the plastic
sleeve contains two coils. The input coil causes a magnetic field in the ring.
This field in turn will induce a current in the other coil. The magnitude of
the current of the output coil depends on the location of the ring, and thus
the amount of phase difference between the input and output shaft.
TU Delft
SKF ADC-SI
2.6 Optical Phase Shift Measurement
(a)
15
(b)
Figure 2.5: Examples of application of NCTEngineering torque sensor (a) in a gearbox
and (b) in a drive shaft. [12]
Figure 2.6: Honda Accord EPAS system torque sensor. [13]
This type of torque sensor was found during the analysis of the steering
rack of a Honda Accord. A picture of the analysis can be seen in figure 2.6.
2.6 Optical Phase Shift Measurement
A solid shaft runs through the sensor body. At the extremities within the
sensor body, two radially slotted disks are fitted. At one side of a disk a
light source is placed, on the other side a phototransistor. Upon rotation
of the shaft, the phototransistors of both disks give a square wave signal as
output, because the slotted disk moves in between the light source and transistor. In unloaded rotation, the two signals have a fixed phase difference.
When a torque is applied to the shaft, this phase difference will change due
to torsional deformation of the shaft. By measuring this change in phase
difference, the applied torque can be calculated, as illustrated in figure 2.7.
TU Delft
SKF ADC-SI
16
Competitor Analysis
Figure 2.7: Optical phase shift measurement, using two slotted disks with light source
and phototransistor. Torque is calculated by measuring the change of phase difference
between the two square signals. [14]
Ergomo uses this measurement principle in their bicycle power meter
product which is aimed at professional athletes and enthusiastic amateurs
to track their training efforts and progress. See also section 3.2 for a detailed
analysis of this product.
2.7 Displacement of Rear Drop Out
All the previous measurement principals measure the torque on the shaft it
is applied to. But that is not the only place where torque can be measured.
When torque is applied through the pedals and crank assembly on a bike,
the energy is transmitted through several components to the rear wheel.
Everywhere along this route, the torque can be measured one way or another.
Sometimes this means that the torque is measured indirectly.
An example of an indirect method is the measurement of the displacement of the rear drop out. The rear drop out is the place on a bicycle
frame where the axle of the rear wheel is connected to the frame. This
piece of frame is replaced with a sensorized frame attachment. The rear
axle is connected to a bracket, which is in turn connected to the frame. The
bracket is designed to be less stiff than the frame in the direction of the
forces transmitted through the chain.
When a torque is applied to the bike, it will be transmitted through the
chain wheel and the chain to the rear sprocket. There, the bracket of the
new rear drop out will deflect. This deflection can be measured with a load
cell. One side of the load cell is attached to the bracket, the other side is
TU Delft
SKF ADC-SI
2.8 Forces on the Bottom Bracket Bearing
17
Figure 2.8: Displacement sensor in rear drop out using a load cell between sensor
bracket and frame. [15]
connected to the frame with a spring-loaded metal ball. If the rear sprockets
radius is known, the torque can be calculated.
The Dutch company IDbike is the inventor of this sensor. It is used
by different manufacturers of e-bikes, like Dutch manufacturers Sparta and
Gazelle.
2.8 Forces on the Bottom Bracket Bearing
Another example of measurement without a torsion bar is the measurement
of the forces on the bottom bracket bearing. While pedaling, the cyclist
exerts a force directed mostly downwards. The torsion applied to the chain
wheel leads to a tension in the chain, which causes a horizontal force between
the frame and the bottom bracket. Both these forces are concentrated in
the bearings of the bottom bracket. The chain tension is mostly present in
the right hand bearing.
The bearings are placed in a cup which is fastened externally to the
bottom bearing tube. An exploded view of such a bottom bracket can be
seen in figure 2.9. The right hand bearing cup has a special insert, in which
the bearing is fitted. The insert is machined in such a way that all the
forces on the bearing are concentrated in a small area. On the surface of
this area, a strain gauge is placed. The forces on the bearing will deform
the insert, which are measured by the strain gauge. With this measurement,
the chain tension can be derived. When the used gear ratio is also known or
measured, the applied torque can be calculated. The external placement of
the bearings has the additional advantage that bigger bearings can be used,
and therefore a bigger spindle. This leads to a stiffer bottom bracket.
TU Delft
SKF ADC-SI
18
Competitor Analysis
Figure 2.9: Force measurement in bottom bracket, with bearings (54, 56) mounted in
cups (48, 80) externally to the bottom bracket tube (74). Right hand bearing mounted
in special insert (82) to concentrate forces to area with strain gauge (in circle). [16]
Shimano has applied for a patent on this system, and is using it in their
new Shimano Total Electric Power System (STEPS, see figure 1.1a).
2.9 Summary of Technologies
All the technologies that are discussed in this chapter are summarized in
table 2.1, together with their advantages and disadvantages. Some technologies use a very accurate method, but rely on magnetic rings, which are
made of rare earth materials. These materials are only exported by China.
This is not a desirable situation, because it will make the manufacturer
dependent of one supplier.
TU Delft
SKF ADC-SI
TU Delft
SKF ADC-SI
2.3
2.4
2.5
2.6
2.7
Mechanical
Magnetic field
Magnetic translation
Optical
Displacement drop out
Simple method
Simple method
Cheap components
Simple method
No signal post-processing
Non-contacting
No torsion bar
Proven sensor technology
Compact construction
High precision
Advantage
Table 2.1: Summary of technologies.
2.2
Magnetic ring
2.8
2.1
Magnetic rotor-stator
Force in bottom bracket
Section
Technology
Indirect measurement
Needs gear ratio measurement
Needs modification of
bicycle frame
High mounting precision
Requires calculation
with microprocessor
Mechanical components
Need to magnetize shaft
Contacting method
Mechanical components
Torsion bar
Torsion bar
Magnetic rings
Torsion bar
Magnetic ring
Disadvantage
2.9 Summary of Technologies
19
20
TU Delft
Competitor Analysis
SKF ADC-SI
CHAPTER
3
Analysis of Purchased Parts
To gain more understanding and get more hands-on experience with some
of the available technology, several existing products were purchased. Two
of the products were torque sensors integrated in a bottom bracket. The
products were both made in Germany and at a price point aimed for the
high-end market. The first one is a sensor from the German bottom bracket
manufacturer Thun and uses magnetic field measurement as its measurement
principal (section 3.1). The second bottom bracket sensor is from Ergomo, a
company specialized in training equipment for professional athletes. It uses
an optical measurement system (section 3.2).
The last product is made in China and at a considerable lower price
point. The interesting part of this product was to analyze its working principal, and to find out how it was possible to produce it for a low cost.
The product is a ‘integrative motor kit’, which contains a hub motor with
integrated controller and torque sensor (section 3.3).
3.1 Thun Bottom Bracket Sensor
Alfred Thun GmbH & Co. is a German company that specializes in bottom
brackets for more than forty years. Therefore, it is interesting to see such
a company makes a step towards sensorized products. The products are
professionally packaged, and look like quality items. The products are accompanied by detailed documentation about assembly and sensor properties
(see appendix B).
The Thun X-Cell RT series consists of two different models. Both models
have a torque sensor and a speed sensor. The difference is that one model
has an analog measurement signal output for the speed sensor, while the
TU Delft
SKF ADC-SI
22
Analysis of Purchased Parts
(a)
(b)
Figure 3.1: Thun bottom bracket sensor (a) product as received [Thun] (b) Illustration
from the Thun website showing the magnetization on the shaft. On the left side a 16
pole pair ring and on the right side a magnetized region. [18]
other has a digital output. Both models were purchased for 100. Documentation on the website of the company reveals the working principle of
the sensors, illustrated in figure 3.1b . The speed measurement is done with
a 16 pole pair magnetic ring and two Hall effect sensors. The use of two
Hall sensors, placed at an electrical phase difference of 90◦ , provides two
outputs. A sine and cosine for the analog output, two square waves for
the digital output. Both types signals have a 90◦ phase difference. This
way, the direction of rotation can also be measured. Torque measurement is
done using the magnetostrictive principle, as described in section 2.4. The
shaft is permanently magnetized using Pulsed Current Modulated Encoding (PCME) technology. This technology is patented by NCTEngineering,
another German company [17].
Before disassembly, the product was first connected to an oscilloscope,
to analyze the signals. Figure 3.2a shows a screen shot of the oscilloscope
showing the speed measurements, while turning the spindle by hand. Two
square signals, with a 90◦ phase difference enable the user to measure the
pedaling speed and the direction. The torque signal was also analyzed. One
end of the spindle was clamped in a workbench clamp and to the other end
a torque was applied, with a wrench. The output, shown in figure 3.2b,
shows the response to the applied torsion. A torsion wrench was also used,
but because it was too hard to apply a steady torque, no real measurement
could be done to determine the linearity or sensitivity.
After this analysis, the sensor was disassembled. Figure 3.3 shows an
exploded view of the sensor. From left to right it shows: the end cap used to
screw the bottom bracket (BB) into the frame tube; a standard sealed deep
groove ball bearing (DGBB) with no manufacturers markings; the sensor
body; the spindle (greased) with second DGBB and the second end cap.
The assembly also contained a magnetic impulse ring, which was mounted
TU Delft
SKF ADC-SI
3.1 Thun Bottom Bracket Sensor
23
(a)
(b)
Figure 3.2: Screen shot of oscilloscope of the output of the Thun Bottom Bracket
sensor measuring (a) the speed and direction of rotation, (b) the applied torque to the
spindle by a wrench.
TU Delft
SKF ADC-SI
24
Analysis of Purchased Parts
Figure 3.3: Exploded view of the disassembled Thun Bottom Bracket sensor.
(a)
(b)
Figure 3.4: (a) Magnetic ring from the Thun BB sensor, used for speed and direction
measurement. (b) Magnetic ring under magnetic field foil, showing the magnetization
of the ring with 16 pole pairs.
on the spindle.
The impulse ring (figure 3.4a) is made of a flexible magnetic material and
had 16 pole pairs (figure 3.4b). The magnetic field of the spindle was very
weak, and could not be detected with the foil, and was hardly measurable
with a Gauss-meter. The spindle was further analyzed with a Metrolab
three-axis Hall effect sensor. This measurement device is able to measure
magnetic fields as small as 1 Gauss. The sensing element was held at one
end of the magnetized region of the spindle and thereafter moved to the
other end of the region. The measurement is plotted in figure 3.5. The
plot shows that the magnetic region is very weak and therefore, a lot of
noise is disturbing the measurement. But there is an offset visible when the
sensor is moved from one side of the region to the other side (approximately
half way during the measurement time). It shows that the spindle is indeed
magnetized, although very weak.
The sensor body was fitted over the spindle, and contained an over molded PCB as shown in figure 3.6a. The PBC (figure 3.6b) contained two Hall
effect sensors for the speed measurement, two coils for the torque measureTU Delft
SKF ADC-SI
3.1 Thun Bottom Bracket Sensor
25
ment and several components for amplification and signal conditioning.
Figure 3.5: Measurement of the magnetized region on the spindle using a Metrolab
three-axis Hall magnetometer. The horizontal axis denotes the measurement time, the
vertical the magnetic field strength in Gauss.
(a)
(b)
Figure 3.6: PCB from Thun BB sensor. (a) Stripped sensor body showing the over
molded PCB. (b) PCB of the Thun sensor with two Hall effect sensors (far right), two
coils for torque measurement (left and center) and several components for amplification
and signal conditioning.
TU Delft
SKF ADC-SI
26
Analysis of Purchased Parts
3.2 Ergomo Bottom Bracket Sensor
The bottom bracket sensor of Ergomo is supposed to be part of a power
measurement setup, targeted on professional athletes and enthusiastic amateur cyclists. It is normally combined with a microcontroller which converts
the measured torque signal together with the measured speed to a power
value. This measurement is used for training purposes and post race analysis.
The sensor uses the optical phase shift measurement principle as explained in section 2.6. The sensor consists of two radially slotted disks and
two optical sensors, as can be seen in figure 3.2. Both sensors give a square
wave output. When a torque is applied, the phase shift between the two
signals is the measure for the amount of applied torque.
Upon delivery of the sensor, the first impression of the product was that
it was somewhere between prototype and production state. This is probably
due to the low volume of production of the sensor. The sensor body was
sealed on several places with some kind of white putty. The price of the
sensor was $ 700.
The interior of the bottom bracket is visible on figure 3.2. It shows that
the sensor uses five roller bearings. Normally there are only two bearings
in a bottom bracket. The reason for this construction with five bearings
is probably that the optical system used requires a high level of precision,
which cannot be obtained with a single bearing on either side.
Figure 3.7: Ergomo bottom bracket sensor with slotted disks for optical measurement.
[14]
TU Delft
SKF ADC-SI
3.3 Integrative Motor Kit
27
3.3 Integrative Motor Kit
The integrative motor kit promises to be a complete kit of an electric hub
motor, a built-in sensor and controller and a control panel (figure 3.8a). Next
to this kit, only a battery pack (24 volt) should be necessary to retrofit a
bicycle to make it an e-bike. The price of the purchased product was $316.
Before disassembly, the product was briefly analyzed. The whole assembly weighted around 4.5 kg. It contained a hub motor with two cables.
To one cable a connector was fitted, to connect the system to a battery pack.
The other cable was connected, via a rather crude interconnection, to the
panel.
As a preliminary analysis, the battery connection of the motor kit was
connected to a power supply at 24 volt. When the system was turned on, the
low battery light was blinking on the control panel. Increasing the supply
voltage to 27 volt solved this problem.
The location of the torque sensor was thought to be at the mounting
position of the rear sprocket(s). Manually applying torque to this connection did not yield any result. When the motor was rotated in the forward
direction around the axle, a noise was heard. Turning the motor in the other
direction did not result in the same noise.
Opening of the hub motor revealed a brushless DC motor (figure 3.8b),
with the control electronics in the center of the motor. After the stator was
removed from its housing, the torque sensor was visible. It consists of two
magnetic rings (both with 20 poles, figure 3.8c), a torsional spring and a PCB
with four Hall effect sensors (mounted on the stator, figure 3.8d). When a
torque is applied to the pedals of the bicycle, it is transmitted through
the chain to the rear sprocket. The sprocket mounting is connected to the
motor housing via the torsional spring, which is the rotor of the hub motor.
The connection with a torsional spring will cause a phase shift between the
sprocket mounting and the rotor. This phase shift can be measured by
measuring the phase shift of the output of the Hall cells. Of the four Hall
effect cell, probably only two are used. The three cells on the outer ring are
probably used for the control of the motor, and only one of these three is
used in combination with the cell on the inner ring to measure the applied
torque.
TU Delft
SKF ADC-SI
28
Analysis of Purchased Parts
(a)
(b)
(c)
(d)
Figure 3.8: Integrative motor kit (a) complete kit with a hub motor with integrated
controller and a control panel, (b) brushless DC motor, (c) torque sensor containing
two magnetic rings and torsion spring, (d) PCB with four Hall sensors.
TU Delft
SKF ADC-SI
CHAPTER
4
Quality Function Deployment
Quality Function Deployment (QFD) is a method used by SKF to translate
the fuzzy voice of the customer into a clear list of design requirements of the
product that has to be developed. The developer of QFD described it as a
method to transform user demands into design quality, to deploy
the functions forming quality, and to deploy methods for achieving the design quality into subsystems and component parts, and
ultimately to specific elements of the manufacturing process. [19]
After listening to the voice of the customer, by taking meetings, receiving
technical specifications and hearing the customer requirements, all the needs
of the customer are formulated. A paired comparison matrix is used to
identify the most important needs by comparing all the needs to each other.
When the most important needs are identified, functions are selected
that are needed to fulfill these needs. The connection between the needs
and the functions is quantified in a House of Quality (HoQ).
After this first House of Quality, a clear view is obtained about which
functions are important in the product. With this list, the concept phase can
be started. After several concepts are worked out, a Pugh Matrix is used to
rate the different concepts in order to select the best. The second House of
Quality is used to couple the functions to real quantified design requirements.
Because this project ends with a proof of concept prototype, the Pugh matrix
and HoQ#2 should be done if the project would be continued.
4.1 Needs
First, all the needs are formulated. Needs are aspects of the products that
are needed to be present in order to satisfy the customer. Usually the term
TU Delft
SKF ADC-SI
30
Quality Function Deployment
Figure 4.1: Work flow of Quality Function Deployment from voice of customer to design
requirements.
TU Delft
SKF ADC-SI
4.2 Paired Comparison Matrix
31
critical to satisfaction (CTS) is used. For the torque sensor eleven needs
were identified and summarized in table 4.1. They are divided into three
categories: three CTSs for the sensing part of the product, four CTSs for
the application of the products and four CTSs regarding the life time of the
product.
4.2 Paired Comparison Matrix
To determine which needs should be focused on, the Pareto principle is used.
The Pareto principle (also known as the 80-20 rule) states that 80% of the
result can be obtained by 20% of the work. Or in this case: 80% of the total
satisfaction of the customer can be reached by fullfulling 20% of the CTS.
But to know which CTS to focus on, the relative importance of the CTS has
to be derived. This can be done by using a paired comparison matrix. In this
matrix, each CTS is compared to all the other needs and it is determined
which need is more important. This rating is done on a scale from 1 to 10,
and to avoid long discussions between team members only the grades 1, 4, 7
and 10 are used. So if CTS1 is compared to CTS2 and the rating is 4 it means
that CTS1 is 4 times more important than CTS2. When the rating between
CTS2 and CTS4 is 17 , it means that CTS4 is 7 times more important than
CTS2. To aid the process, a matrix is used as pictured in figure 4.2. The
fields above the diagonal are filled in the reading direction. The values of
the fields under the diagonal are the inverse of the fields above the diagonal.
When the complete matrix is filled, the values in each row are summed.
These totals are then normalized so that the sum of all CTSs is 100%.
Sensor
Needs to
measure torque
measure position
measure speed
Application
Needs to
be mechanically integrated in customer application
communicate with customer application
give no perception of presence to cyclist
be safe
Life Time
Critical to Satisfaction
Needs to
be maintainable
withstand disturbances
be reliable
be of reasonable cost
Table 4.1: Critical to Satisfaction for torque sensor product.
TU Delft
SKF ADC-SI
32
Quality Function Deployment
Reading direction
S
CT
1, 4, 7, 10
1
S
CT
CTS 1
1
4
CTS 2
1/4
1
CTS 3
CTS 4
2
S
CT
3
S
CT
4
l
ta
To
ed
iz ht
al eig
m
w
No
1/7
1
7
1
Figure 4.2: Example of paired comparison matrix.
For the torque sensor, the analysis was done and the results are displayed
in figure 4.3. The full matrix can be found in appendix C. The main conclusions are that both the need to be maintainable and the need to measure
position are of very little importance for the torque sensor for e-bike application. The former because the product should be designed such that it does
not need any maintenance during its life time. The latter because position
information is of no interest for the e-bike manufacturer. Therefore, the
need for position is dropped in the next steps of the analysis.
4.3 Functions
The next step in the analysis is to formulate the functions that are needed
in the product in order to fulfill the needs. These functions are needed to
ensure the quality of the product. Therefore, the term Critical to Quality,
or CTQ, is regularly used. For example to satisfy the need of measuring
torque, a certain range of torque measurement is needed. To satisfy the
need of communicating with the customer application, it might be useful to
use some kind of standardized connector or a certain protocol. For every
CTS, several CTQs can be thought of. All these CTQs are summarized in
table 4.2.
4.4 House of Quality #1
To identify the importance of a function, one has to look at the connection
between a function and the needs. The importance of the needs is already
quantified in the previous section. By quantifying the connection between a
need and its functions, the importance of the functions can be determined.
TU Delft
SKF ADC-SI
4.4 House of Quality #1
33
Critical to Quality
CTS
CTQ
Torque & Speed
Range
Accuracy
Repeatability
Amplitude
Offset
Resolution
Input voltage
Refresh rate
Weight
Communication
Protocol
Standardized connector
Integration
Presence
Safety
Maintenance
Reliability
Disturbances
Cost
Aesthetics
Assembly time duration
No modification to customer application
Extra friction
Time lag
Fail safe function
Over torque test
No part destroyed during maintenance
Time duration to repair
Endurance test
Vibration
Moisture
Temperature
Shock
Price
Table 4.2: Critical to Quality for torque sensor product.
TU Delft
SKF ADC-SI
20%
100%
15%
75%
10%
50%
5%
25%
0%
M
Co
m
e
as
u
m
e
ur
a
ni c
rq
To
t
Total weight
Quality Function Deployment
Normalized weight
34
0%
ue
i th
ew
s
cu
t. a
p
pl .
Be
sa
fe
Be
li
re
a
M
M
bl e
e
ec
as
e
ur
Of
ha
e
Sp
a
re
al l
ni c
ed
so
n
yi
b
na
le
t
ra
te g
ve
Gi
co
ed
no
st
in
p
a
st.
cu
ce
er
p ti
p
pl .
on
o
s
re
fp
i th
W
s ta
e
e
nc
nd
tu
di s
a
rb
nc
Be
es
ai n
m
ta i
M
na
ea
bl e
re
su
s
Po
n
itio
Figure 4.3: Results of paired comparison matrix analysis for torque sensor.
This analysis is done by using a House of Quality-diagram (HoQ). A simple
example of the HoQ is shown in figure 4.4, to help explain the concept of
the diagram.
On the left of the matrix, the CTSs are written, on the top of the matrix
the functions. Colors are used to indicate the connections between certain
needs and functions, according to table 4.2. But the connections are not
limited to these color links. It is quite possible that a function that is
thought of by a certain need also has a connection with another need. For
example, the CTQ price has a connection with almost all CTSs, not only
with the CTS cost.
The functions can be divided into two groups. Quantifiable CTQs and
non-quantifiable attributes. The quantification can be done at the bottom
of the diagram. Above the CTQs, the desired property of the quantification
can be written. An arrow upwards means that a higher value is better, an
arrow down indicates that a lower value is desired. An O means that a value
closest to the indicated value is desired.
In the body of the matrix, the rating is again done on a scale from 0 to
10, using only 0, 1, 4, 7 and 10 for simplicity. A rating of 10 means that
there is a really strong connection between the function and the need, a 1
means that there exists a connection but it is really small. A rating of 0
means that there is no connection. Usually, if the rating is 0, the cell is left
blank to aid readability of the diagram.
TU Delft
SKF ADC-SI
4.4 House of Quality #1
35
+
CTQ
Attributes
We
igh
t
No
rma
lize
d
CTQ7
CTQ6
CTQ5
CTQ4
CTQ3
CTQ2
CTQ1
O
CTS1
CTS2
25%
1
CTS3
10
17%
CTS4
total
0,25
1,7
weight
13%
87%
mean
LSL
USL
sigma
Figure 4.4: Example of a House of Quality.
The rating of every connection is multiplied by the normalized weight
of the CTS from the paired comparison matrix and is summed per CTQ.
This total is again normalized to a total of 100%. This value indicates the
relative importance of the functions for the product.
In the roof of the House of Quality, relations between the different CTQs
can be indicated. In the example, there is a positive relation between CTQ1
and CTQ2. This means that by improving CTQ1, CTQ2 will also improve.
A minus indicates a negative relation. Most of the time the CTQ cost has a
negative relation to a lot of other CTQs, because there is usually a trade-off
between cost and quality.
The results of the first House of Quality for the torque sensor for e-bike
application are plotted in figure 4.5. The complete HoQ#1 of this project
can be found in appendix D. It shows that the most important functions,
next to torque measurement, are: cost, safety and reliability.
TU Delft
SKF ADC-SI
TU Delft
Page 1
l e d
t
t
t
t
t
t
t
)
.
r
r
)
s
e
e e
ic ion u u es es se ed es ue es co u e pl ate es ed to ed ue ed lag ion es tic ue ed gh rd ce ai
Pr nct orq orq ce t n t noi Spe re t orq ck t oto orq Spe . Ap h R e t Spe nec Spe orq Spe e frict re t the orq Spe ei nda nan rep
W ta te g
fu e T ty T an atio e ( ity stu n T ho Pr y T e ust es rqu n on cy e T e Tim ra atu es t T et
s in rin
g
i
e
r
r
g l
i
A fse ffs
xt er
a
ac an o c efr r to lutio d C ura itud litud
af ang bil du Vib lta abi Mo lutio S
r
f
E
p
s
(to m du
t
l
O
u
e
R v
a n
t
e
o
c
R
o
O
p
l
t
c
R
o
i
p
m
z
n
v
a
c
s
g ion
n
E
i
a
c
s
a
m
e
o
t
e
O
n
e
o
m
A
d
i
e
e
A
r
F
t
T
A
ti
ri
R er
A
R
pu ep
ra du ura
ep
d
ca
In R
R
ifi
d
du ed
an
d
t
e
e y
o
S
m
tim ro Tim
o
y est
l
N
b
m l. d
se pp
s
a
A r
e
om
st
u
c
of
tr s
pa
o
N
0%
1%
2%
3%
4%
5%
6%
7%
8%
edit copy(2) Chart 88
0%
20%
40%
60%
80%
100%
36
Quality Function Deployment
Figure 4.5: Results of House of Quality #1 for torque sensor.
SKF ADC-SI
Part II
Sensor Design
TU Delft
SKF ADC-SI
CHAPTER
5
Concepts for a Torque Sensor for E-bike Application
The generation of concepts for the torque sensor is done in three phases:
where, what and how. The where phase (section 5.1) looks at the location
on the bike where the sensor should be situated. The what phase (section
5.2) determines the physical quantity that is going to be measured. Finally,
the objective of the how phase (section 5.3) is to find a suitable measurement
principle.
5.1 Where - Location of Sensor
The first thing that has to be determined is the location of the sensor. On
a bicycle, there are a lot of possible locations for a torque sensor, all with
different advantages and disadvantages. Therefore, a schematic mapping is
made of a bicycle. In figure 5.1a this mapping is shown. The blue connections indicate the known products, as analyzed in part I.
For SKF, two major requirements are that the sensor can be sold as
an individual item and that the sensor is usable on every type of e-bike.
Therefore, the sensor should be located on such a location that an individual
item can be created. So a torque sensor integrated in a hub motor is not
interesting for SKF, because SKF does not produce these motors nor is it
certain that every bike is equipped with a rear wheel hub motor.
bike equipped with such a motor. Figure 5.1b shows the same mapping
of the bike, but now with all the possible locations of the sensor.
An additional advantage is that there is some prior knowledge available
within the company about the product. Therefore, two locations seam interesting: the bottom bracket (SKF produced a bottom bracket, although it
TU Delft
SKF ADC-SI
40
Concepts for a Torque Sensor for E-bike Application
is now discontinued) and the freewheel (SKF produces freewheels for automotive applications).
Looking only at the criteria for the location, the best locations for the
torque sensor for SKF are:
• Between the bottom bracket spindle and the chain wheel;
• On the bottom bracket spindle;
• In the bottom bracket bearing;
• In the freewheel.
The next step will focus on what physical quantity can be measured on the
different locations. The combination of the location and physical quantity
that can be measured on that location will determine the final choice.
5.2 What - Physical Quantity
The question what should be measured seems simple: torque. But torque
can be measured in many different ways, either direct or indirect. In chapter
2, several technologies were already discussed. The possibilities are (but not
limited to):
• Effects of stress on the material
Magnetostrictive effect
Magnetic susceptibility
Resonance frequency (surface acoustic waves)
• Torsion
Strain gauge
Torsion angle
• Indirect measurement
Reaction forces
Chain tension
For every identified possible sensor location, as illustrated in figure 5.1b,
a measurable physical quantity is thought of. These concepts, with their
advantages and disadvantages, are summarized in table 5.1. It shows that
the best location for the sensor will be to integrate it in the bottom bracket.
This way, a part can be developed that is suitable for retrofitting, is within
the scope of existing knowledge and products of SKF and is has enough
space for a sensor.
TU Delft
SKF ADC-SI
5.2 What - Physical Quantity
41
(a)
(b)
Figure 5.1: Schematic mapping of bicycle with (a) location of sensors of competitors
and (b) possible locations for torque sensor and final concept.
TU Delft
SKF ADC-SI
42
Concepts for a Torque Sensor for E-bike Application
(a)
(b)
(c)
(d)
Figure 5.2: (a) A normal right pedal crank, attached rigidly to the chain wheel. (b)
Front view of concept geometry with flexible connection between BB spindle and chain
wheel. (c) Cross section of concept BB spindle with measurement faces for measurement of torsion angle. (d) Visualization of torques on concept BB spindle with flexible
connection between spindle and chain wheel.
The concept is to measure a torsion angle between the BB spindle and
the chain wheel. To achieve this, some structural changes have to be made
to the assembly. A traditional chain wheel is normally connected to the
right pedal crank as illustrated in figure 5.2a. This way, the chain wheel
and the crank are connected rigidly to the spindle, and no torsion angle
can occur. This concept uses a new geometry, where the pedal cranks are
connected rigidly to the spindle and the chain wheel is connected to the
spindle through a flexible connection (figure 5.2b). This way, the torque
applied to the pedals will be transferred through the spindle to the chain
wheel (figure 5.2d). Because the connection between the spindle and the
chain wheel is flexible, the applied torque will cause a rotation between the
two parts, which can be measured to determine the torque (figure 5.2c).
TU Delft
SKF ADC-SI
TU Delft
Stress on axle
Torsion angle
Stresses on spindle
Deformation of inner ring
Reaction forces
Chain tension
Stresses on axle
?
Pedal ↔ crank
BB spindle ↔ Chain wheel
BB spindle
BB bearing
BB spindle ↔ Frame
Frame ↔ Chain tensioner
Axle
Freewheel
Universal item
Simple
Universal item
Direct measure of torsion
Technology reusable
Universal item
Space for sensor
Universal item
Space for sensor
Universal item
Advantage
Table 5.1: Locations and its possible physical quantities
What
Where
Complex geometry
Small space
Rotating
Not compatible with hub motors
Indirect measurement
Gear information needed
Indirect measurement
Rotating
Only one side measured
Needs modification of right crank
Rotating
Disadvantage
5.2 What - Physical Quantity
43
SKF ADC-SI
44
Concepts for a Torque Sensor for E-bike Application
(a)
(b)
Figure 5.3: (a) Measurement concept with magnetic rings with two Hall sensors which
measure the magnetic field of the rings. The phase difference between the two sine
waves is the measure for applied torque. (b) Measurement concept with slotted disks.
Duty cycle of measurement signal is measured to determine applied torque.
5.3 How - Measurement Principle
Now that is known what is going to be measured, several concepts were
made for a measurement principle. These concepts are shortly explained in
the following sections (5.3.1 - 5.3.3). Thereafter the best concept is chosen
to be further analyzed in section 5.3.4.
5.3.1 Concept Magnetic Rings
The first concept will focus on the measurement of the magnetic field of two
magnetic rings. The rings will be fitted to the two measurement faces. The
rings will have a plurality of magnetic pole pairs around the perimeter of
the ring. The magnetic field will be measured by one Hall sensor per ring.
The Hall sensor will measure the amplitude of the magnetic field, thus upon
rotation both Hall sensors will output a sine wave. When no load is applied,
the two sine waves will have a fixed phase difference. When a torque is
applied, the rings will rotate relative to each other. Either embedded or
external software will measure this phase difference, and will be able to
calculate the torsion angle and therefore the applied torque. A schematic
drawing of the concept is shown in figure 5.3a.
5.3.2 Concept Duty Cycle
The second concept has as goal to measure the duty cycle of the measurement signal. Instead of magnetic rings, now two slotted disks are used.
The disks are placed perpendicular to the measurement faces. On one side a
TU Delft
SKF ADC-SI
5.3 How - Measurement Principle
45
source (magnetic ring, LED) is placed, and on the other side a detector (Hall
sensor, phototransistor). When no load is applied to the sensor, the slots of
the two disks will be in phase. Therefore, when the disks are rotating, 50%
of the time the path between the source and the detector is unobstructed.
A duty cycle of 50% will be measured. When a load is applied, the disks
will rotate relative to each other thereby reducing the opening between the
source and sensor. The duty cycle of the measurement signal will decrease.
The duty cylce of the measurement signal will be a measure for the applied
torque. This measurement principle is illustrated in figure 5.3b.
5.3.3 Concept Rotation to Displacement
The last concept will measure the torsion angle indirectly by measuring the
displacement of an intermediate piece that will transform the rotation into a
translation. The previous two concepts both are based on the measurement
of rotating parts. This can be quite challenging. When the rotation is
transformed to a translation, more conventional measurement principles can
be used.
A ring will be placed between the two measurement faces. On the inside,
the side of the spindle, the ring will be constrained to move only in the axial
direction with a key. On the outside, a pin will be fitted in a helical slot.
If a rotation occurs between the spindle and the chain wheel, the pin in the
helical slot will force the ring to displace axially. The displacement can be
measured with several different measurement techniques:
• Variable reluctance. By modifying the path of a magnetic circuit
(i.e. modifying the air gap between the coil core and the armature),
the inductance of a coil will change. By using two coils in a differential,
or push/pull arrangement, a linear relation can be measured between
the displacement and output voltage.
• Eddy current. When a magnetic field comes close to a conductor,
the magnetic flux will induce eddy currents in the conductor. These
currents will produce a magnetic field in the opposite direction. This
will lower the inductance of the coil, which again can be measured in
a differential set-up.
• Hall sensing (with or without bias magnet). When a magnetic field
comes close to an electrical conductor with its current perpendicular to the magnetic field, a voltage difference can be measured over
the conductor. A bias magnet can be used to enhance the magnetic
field. Hall sensors are widely used in the field of speed and position
measurement.
• Capacitive sensing. When two plates move in relation to each other,
the change in capacitive coupling can be measured.
TU Delft
SKF ADC-SI
46
Concepts for a Torque Sensor for E-bike Application
Figure 5.4: Schematic drawing of concept with measurement of variable reluctance.
Ring will move in axial direction when torque is applied. The variation of the air gap
between the ring will be measured to determine torsion angle.
The concept is illustrated in figure 5.4, with the variable reluctance measurement technique. The other techniques all work in the same way, with a
translating sleeve which will have a ring on it and sensing elements (coils,
Hall cells, capacitor plates) in the direct neighborhood. The advantages
and disadvantages of the different measurement methods are summarized in
table 5.2.
Method
Advantage
Disadvantage
Variable
reluctance
Linear measurement
Concentrated measurement field
Ferrite target material
Eddy current
PCB based coil
Aluminium target
Contamination no problem
Temperature
Range
Broad measurement field
Hall
sensing
Temperature compensated
Small component
Non-linear
Magnetic target
Hall +
Bias magnet
Temperature compensated
Small component
Non-linear
Ferromagnetic target
Capacitive
High sensitivity
Concentrated measurement field
Temperature sensitive
Sensitive to contamination
Range
Table 5.2: Advantages and disadvantages of different measurement methods.
TU Delft
SKF ADC-SI
5.3 How - Measurement Principle
47
5.3.4 Conclusion Measurement Principle
To select a final concept, the three concepts are judged on several criteria.
Some of these criteria are based on the analysis during the benchmarking
phase in chapter 4. Other criteria came up during the brainstorm sessions.
The criteria accuracy, cost of components, simplicity of mechanical design
and necessity for signal post processing are all related to the measurement
quality, price and compatibility of the final product; factors that were identified as Critical to Satisfy.
Next to that, during the concept phase it came apparent that some
concepts were able to measure torque at zero rotation speed while others
were not. This was also identified as an important property. Finally, the
availability of available knowledge would help to speed up the design. In
table 5.3, the three concepts are qualitatively rated to gain insight in the
viability of the concept. The rating is done on a scale [++, + , ◦, −, −−],
where ++ means that there is a positive relation with the criterion (e.g. it
would be favorable to choose that concept according to that criterion).
Based on the analysis, the concept Rotation to Displacement was chosen.
As measurement principle, both eddy current and variable reluctance based
measurement would be applicable. This measurement concept is further
worked out in chapter 6.
Criterion
Accuracy
Component Costs
Simplicity of Mechanical Design
Necessity of Post Processing
Zero Speed Measurement
Available Knowledge
Magnetic
Rings
Duty
Cycle
Rotation to
Displacement
−
−
+
−
+
++
+
−
−
−
−−
−−
+
+
−
+
++
◦
Table 5.3: Multi criteria table to compare the three measurement principle concepts.
TU Delft
SKF ADC-SI
48
TU Delft
Concepts for a Torque Sensor for E-bike Application
SKF ADC-SI
CHAPTER
6
Design of Sensing Elements
Before a complete sensorized bottom bracket can be designed, first the sensor
system has to be developed. First, the principles of eddy current and variable
reluctance are further explained (section 6.1). In section 6.2 the geometry of
the sensor system is introduced. Thereafter, a model is built and simulated
using Ansoft Maxwell in section 6.3. The results are presented in section
6.4. Finally, conclusions were drawn in section 6.5.
6.1 Eddy Current & Variable Reluctance
From the several measurement techniques introduced in section 5.3.3, variable reluctance and eddy current were selected as the most suitable for
integration in the torque sensing bottom bracket. Both techniques would
require roughly the same design, with two coils with a target ring in between
in order to create a differential set-up.
6.1.1 Eddy Current
Eddy currents are currents that are induced in a conductor by a changing
magnetic field. This can occur either when a conductive material is moved
in a magnetic field, or the conductive material is exposed to a magnetic field
that is varying in time. The change in magnetic flux through the conductor
will cause a circulating flow of electrons in the material. This current in turn
will cause a magnetic field of its own, and opposes the existing magnetic field.
The magnetic field can be created with a coil. When a coil is supplied
with an alternating current, the magnetic field will change in time. The
induced voltage in a coil, Uind , is dependent on the inductance of the coil,
TU Delft
SKF ADC-SI
50
Design of Sensing Elements
Figure 6.1: Eddy current proximity sensor. Eddy currents are induced in the target
when an alternating current is applied to the coil. [20]
L and the change of current I. Or when a magnetic field induces a current
in the coil, the induced voltage depends on the number of turns, N , and the
change of magnetic flux, Φb .
Uind = −L
dI
dΦb
= −N
dt
dt
(6.1)
When a conductor is in the neighborhood, the coil will cause the induction
of eddy currents in the conductor. The opposing field caused by the eddy
currents will then in turn influence the inductance of the coil. This principle
is often used in proximity sensors, as illustrated in figure 6.1. The induced
voltage in the coil will be measured and when a conductive target will come
into range, this voltage will change.
6.1.2 Variable Reluctance
The working principle of a variable reluctance displacement sensor is to
measure the variation of the inductance of a coil by variating the air gap
in the magnetic circuit as shown in figure 6.2. The inductance of a coil is
described by:
L=
N2
R
(6.2)
where N is the number of windings of the coil and R the reluctance of the
magnetic circuit. The reluctance can be seen as a magnetic resistance. It
limits the magnetic flux just like the resistance limits the electric current in
an electrical circuit. The reluctance is defined as:
R=
TU Delft
l
µ0 µA
(6.3)
SKF ADC-SI
6.1 Eddy Current & Variable Reluctance
51
Figure 6.2: Simple representation of a magnetic circuit caused by a coil with an air
gap. [21]
where l is the path length of the magnetic circuit, µ0 the permeability of the
free space (µ0 = 4π · 10−7 Hm−1 ), µ the permeability of the material and
A the surface perpendicular to the magnetic path. To simplify the analysis
of displacement sensor it is possible calculate the reluctance of the core,
armature and air gap separately.
Rtot = Rcore + Rarmature + Rairgap
(6.4)
To obtain a better, and more linear measurement, a differential displacement
sensor can be used. Here, the movable armature is placed between two coils,
as illustrated in figure 6.3a.
Rtot = R0 + Rairgap
2(d − x)
= R0 +
µ0 µAairgap
= R0 + k(d − x)
(6.5)
Where R0 is the reluctance at zero air gap. The inductance of the coils will
be influenced by the size of the air gap.
N2
L0
=
R0 + k(d − x)
1 + α(d − x)
2
N
L0
L2 =
=
R0 + k(d + x)
1 + α(d + x)
L1 =
With L0 =
TU Delft
N2
R0 ,
and α =
(6.6)
(6.7)
k
R0 .
SKF ADC-SI
52
Design of Sensing Elements
(a)
(b)
Figure 6.3: (a) Schematic setup of a differential variable reluctance displacement sensor.
[21] (b) Wheatstone bridge
So the self-inductance of both coils is a function of the self-inductance at
zero air gap, L0 , and the length of the air gap, (d − x).
The self-inductance that is measured has a non-linear relation to the
displacement. This is of course, not desirable for a sensor. To overcome
this problem, the two measurement coils are placed in a Wheatstone bridge.
The output voltage of the bridge will be
VL
L1
R1
= VS
−
L1 + L2 R1 + R2
!
1
1
= VS
−
L1
2
1+ L
2
αx
= VS
2 (1 + αd)
(6.8)
So, the output of the bridge will be linear with the displacement of the
armature.
6.2 Concept of Sensor
To make sure that manufacturing or machining errors on the target ring
would not influence the measurement (errors in thickness would introduce
harmonics due to the rotation of the ring) it was decided that the coils
should be placed around the full circumference of the target ring. This
would also improve the inductance of the coils, which in turn would improve
the accuracy of the measurement. This decision has a big influence on the
available space inside the bottom bracket. The coils that have to be used
TU Delft
SKF ADC-SI
6.3 Modelling of Sensor Using Ansoft Maxwell
53
Figure 6.4: Cross-section of sensing concept. A sleeve will move axially according to
the applied torque. A target ring is placed on the sleeve. The target ring is centered
between coils, which are wound oppositely. The movement of the ring will be measured
by measuring the change of the induced voltages of the coils.
when the variable reluctance is chosen need a (ferrite) core surrounding the
coils and should face the target ring directly, as in figure 6.3a. The eddy
current principle is less demanding on geometry. The only requirement is
that the target ring is well within the magnetic field created by the coil.
The following concept geometry was developed: Two coils, wound oppositely, are placed around the spindle. On the spindle, a sleeve is fitted which
translates in axial direction according the amount of applied torque. On
this sleeve a target ring is fitted. A cross section of the proposed geometry
is shown in figure 6.4.
The movement of the target ring will influence the induced voltage of
the coils. When there is no load on the bottom bracket, the ring will be
in the center of the two coils, and the voltage will be the same for the two
coils. When torque is applied, the ring will translate axially and the induced
voltages will change. One will go up, while the other will go down. This
change can be measured with a Wheatstone bridge.
6.3 Modelling of Sensor Using Ansoft Maxwell
To be able to investigate variations in parameters, a simulation was made using Ansoft Maxwell. This program allows the user to make electromagnetic
field simulations using finite element analysis. In this section, the construction of the analysis is described. Results of the analysis are given in the next
section. This simulation will be used to investigate the influence of several
parameters. It will be hard to validate the results of the simulation before
a prototype can be tested. But the trend that will be found by varying a
parameter will be presumed to be reliable.
TU Delft
SKF ADC-SI
54
Design of Sensing Elements
(a)
(b)
Figure 6.5: Modelling of sensor in Ansoft Maxwell, (a) Simplified geometry of the
torque sensing bottom bracket consisting of two coils (wound oppositely) and a body
that represents both spindle, sleeve and target ring, (b) Meshed model of the simplified
model of the torque sensing bottom bracket.
6.3.1 Modelling & Meshing
In the previous section, a concept geometry was proposed for the sensor.
This geometry was built in Maxwell, using estimated sizes for all the components. The goal was to build representative simulation of the torque
sensing bottom bracket. This model will be used to evaluate the effect of
changes made to the geometry or design parameters on the sensitivity of
the sensor. The model used for the simulations is shown in figure 6.5a. The
geometry was simplified to reduce calculation time. The spindle, sleeve and
target ring were merged into one part. All the parts that would not interfere
with the measurement (due to distance or magnetic properties) were left out
of the model.
The model was meshed using the standard mesh generator available in
Maxwell, and was refined in places which were expected to be important
to the analysis. This included the coils and the target ring. In total, the
model consisted of 35,374 tetrahedra. The complete geometry was evaluated
within a region of air, 200% the size of the model. Later on, the influence
of other objects in the vicinity of the sensor is also analyzed. The meshed
geometry is shown in figure 6.5b.
6.3.2 External Circuit
The Maxwell software is also able to simulate the electronics used for the
sensor in an external circuit. In this model, both the input signal and the
output signal where modeled in the external circuit editor. This enabled a
TU Delft
SKF ADC-SI
6.3 Modelling of Sensor Using Ansoft Maxwell
55
Figure 6.6: External circuit imported in Maxwell simulation used to measure the change
of inductance of the coils.
direct read-out of the signal over the Wheatstone bridge. The circuit used is
shown in figure 6.6. The circuit consists of a voltage source, with an input
of 5 V AC at 20 kHz. The two coils from the model, called FirstCoil and
SecondCoil respectively, are placed in a bridge circuit. The two resistors of
the bridge circuit were chosen arbitrary and have a resistance of 10 kΩ. An
additional resistor of 1 Ω was included in the circuit to avoid a short circuit
through the branch of the bridge with the coils, because Maxwell treats the
coils as ideal inductances an have therefore no resistance. A voltmeter was
placed over the bridge to measure the change in inductance. The output of
the voltmeter can be accessed directly from the Maxwell interface.
6.3.3 Analysis Parameters
Because the input signal is an AC voltage, a transient analysis type is chosen.
This type of analysis will also be able to simulate movement of the spindle
in a later stage. The output of interest will be the output of the Wheatstone
bridge. This will be a sinusoidal signal with the same frequency as the input
voltage. For each parameter, two simulations will be done: the first with
the target ring centered between the coils, the second with the target ring
axially displaced by 1 mm. The output signal will be demodulated and the
magnitude will be used to calculate the sensitivity.
To obtain a good read-out of the Wheatstone bridge, two full periods
of the input are simulated, to eliminate errors caused by initial overshoot
of the signal caused by the initial current (0 Ampere) in the coils. For a
supply voltage with a 20 kHz frequency, this leads to a simulation time of
1 · 10−4 seconds. For each period, ten simulation steps are taken to make
sure the output sine wave has sufficient detail. Therefore, the time step is
5 · 10−6 seconds. Each simulation took about 6 minutes on a single 3.2 GHz
processor.
TU Delft
SKF ADC-SI
56
Design of Sensing Elements
6.4 Results of Electromagnetic Field Simulations
In total, more than forty variations in parameters were done. The most
significant results are shown in this section.
6.4.1 Materials
First subject of investigation was the material of the target ring. Four possible materials were used: steel (1008), ferrite, aluminium and copper. Figure 6.7 shows the magnetic flux density on the YZ-plane of the model. The
calculated sensitivity is shown in table 6.1. It shows a significant difference
between the two magnetic materials (steel and ferrite) and the two nonmagnetic materials (aluminium and copper). But figure 6.7b and 6.7d show
little to no flux density in the two materials (as can be expected with these
materials). Therefore, it was assumed that not variable reluctance, but eddy
current would be the dominating physical property in these two materials.
Figure 6.8 shows the induced current in the target ring. It is clear that in
the aluminium substantially more current is induced than in the steel ring.
These eddy currents are in the opposite direction of the current direction in
the coils. They influence the inductance of the coils, which can be measured
in the Wheatstone bridge circuit. To verify this hypothesis, the simulation
with the aluminium target ring was done again, but now the option which
takes into account the eddy current was disabled. The sensitivity dropped
with nearly 95%.
Material
Sensitivity
[mV/mm]
Improvement
[%]
90
150
100
160
67
11
78
Steel 1008
Aluminium
Ferrite
Copper
Table 6.1: Sensitivity of sensor using different materials as target ring.
6.4.2 Target Ring Thickness
The second parameter that was investigated was the thickness (in axial
direction) of the target ring. The effect on the sensitivity was investigated
in both steel and aluminium target rings. The results are summarized in
figure 6.9. It shows that the thickness of the target ring has different effects
on the two materials. It is presumed that this is caused by the fact that
the prevailing physical phenomenon is different in the two materials, as
explained in the previous section.
TU Delft
SKF ADC-SI
6.4 Results of Electromagnetic Field Simulations
57
(a)
(b)
(c)
(d)
Figure 6.7: Magnetic Flux density using different materials, (a) steel 1008, (b) aluminium, (c) ferrite and (d) copper. All surface plots are scaled on a color scale from 0
to 2.5 · 10−3 Tesla (blue to red).
(a)
(b)
Figure 6.8: Induced eddy current in the target ring from the same simulations as above
(a) in a steel target ring (scale 2.5 · 104 A/m2 ), and (b) an aluminium target ring (scale
2.5 · 105 A/m2 ). In the (non magnetic) aluminium ring, the induced eddy currents are
more than 10 times stronger.
TU Delft
SKF ADC-SI
58
Design of Sensing Elements
Figure 6.9: Effect of thickness (axial) of target ring on the sensitivity for both metal
and aluminium rings.
6.4.3 Coil Spacing
The next parameter that was tested was the space between the coils. In the
initial design, the coils were arbitrarily spaced 2 mm apart. This distance
was varied, again for both the steel and aluminium target ring. The results
of these simulations are shown in figure 6.10. It shows that for both target
rings, the sensitivity increases when the coils are closer together.
6.4.4 Ferrite Core
Another method to improve the sensitivity might be to use a ferrite core
to guide the magnetic field. Two L-shaped and one I-shaped ferrite rings
are placed in between and around the coils. The resulting magnetic field
is shown in figure 6.11. When used with a steel target ring, the sensitivity
increased 26%. With an aluminium target ring it improved around 3%, as
shown in table 6.2.
Target Ring
Steel 1008
Aluminium
Sensitivity
[mV/mm]
Improvement
[%]
113
155
26
3
Table 6.2: Sensitivity of sensor with ferrite core and improvement over sensor without
the core (table 6.1).
TU Delft
SKF ADC-SI
6.4 Results of Electromagnetic Field Simulations
59
Figure 6.10: Effect of coil spacing on the sensitivity for both metal and aluminium
rings.
Figure 6.11: Magnetic flux density of sensor with ferrite core. This result shows that
the magnetic flux is guided by the ferrite core towards the target ring, thereby improving
the sensitivity.
TU Delft
SKF ADC-SI
60
Design of Sensing Elements
6.4.5 Frame Tube
The last parameter that was tested was the influence of the surroundings,
especially the frame tube the bottom bracket will be placed in. Therefore, an
optimized geometry (2 mm aluminium target ring, 0.1 mm coil spacing) was
used as a reference. Three simulations were done. One with a steel frame
tube, one with an aluminium frame tube and a last one with an aluminium
frame tube and ferrite shielding in between the coils and the tube. In figure
6.12 the results of these four simulations are plotted. The illustrations show
the current density on the YZ-plane cross-section. The resulting sensitivities
are summarized in table 6.3. It is clear that the use of an aluminium frame
tube will deteriorate the sensitivity and shielding should be used.
Frame Tube
Sensitivity
[mV/mm]
Improvement
[%]
None
Steel 1008
Aluminium
Al. with shield
193
218
103
180
+13
-47
-7
Table 6.3: Sensitivity of sensor with different kind of frame tubes.
6.4.6 Air Gap
An additional parameter that was tested was the influence of the air gap
between the target ring and the coils. These simulations were in some way
trivial, because the impact of reducing the air gap could be predicted with
a great amount of certainty. The simulation was done for both steel and
aluminium target ring as the air gap was reduced from 1.5 mm to 0.5 mm. As
expected, the sensitivity increased significantly. The results are summarized
in table 6.4.
Target Ring
Steel 1008
Steel 1008
Aluminium
Aluminium
Air Gap
[mm]
Sensitivity
[mV/mm]
Improvement
[%]
1.5
0.5
1.5
0.5
90
145
150
295
61
97
Table 6.4: Sensitivity of sensor with reduced air gap.
TU Delft
SKF ADC-SI
6.4 Results of Electromagnetic Field Simulations
61
(a)
(b)
(c)
(d)
(e)
(f)
(g)
(h)
Figure 6.12: Influence of different frame tubes on the sensor. Magnetic flux (left)
and current density (right) on YZ plane of different frame tubes. (a),(b) optimized
geometry of the model without a frame tube present, (c),(d) a steel frame tube, (e),(f)
an aluminium frame tube and (g),(h) an aluminium frame tube with a thin ferrite ring
for shielding. All flux plots are scaled to 1 · 10−3 Tesla, all current plots are scaled to
1 · 105 A/m2 .
TU Delft
SKF ADC-SI
62
Design of Sensing Elements
6.4.7 Preliminary Conclusions
From the results of these simulations, conclusions can be drawn which can
be used as guidelines and goals in the mechanical design. The following list
of conclusions is compiled in descending order of importance.
• Air gap. The air gap is by far the most important parameter in
the sensor setup. Reducing the air gap will have the greatest impact
on the sensitivity. However, the mechanical design will constrain this
reduction.
• Target ring material. The material that is chosen for the target ring
defines the dominating physical principle. When a magnetic material
is used, variable reluctance is dominating. Eddy current is dominating
when a highly conductive material is used. The simulation showed that
eddy currents lead to the highest sensitivity. Therefore aluminium or
copper should be used. The difference in sensitivity is only around
7%.
• Target ring thickness. As shown in figure 6.9, there is an optimum
in the axial thickness of the target ring. When an aluminium ring is
used, the thickness should be 2 mm.
• Coil spacing. The sensitivity of the sensor increases when the coils
get closer together. Therefore, the coils should be placed as close as
possible to each other.
• Ferrite core. Section 6.4.4 showed that the use of a ferrite core
around the coils could increase the sensitivity. However, the effect is
small, when used in combination with an aluminium target ring. Also,
because space within the bottom bracket is limited, the incorporation
of a core will be problematic.
• Frame tube. Although not of influence on the mechanical design,
attention should be paid to the possible influences of the immediate
surroundings of the bottom bracket. It should be noted that an aluminium frame tube will deteriorate the sensitivity of the sensor and
therefore some shielding should be included.
6.4.8 Simulation of Moving Spindle
With the conclusions from the simulations as described in the previous section, a more detailed mechanical design was made. In this design, all the
important constraints (i.e. spindle diameter, maximum bottom bracket diameter, etc.) were taken into account. Also factors as producibility and
possible ways to assemble the product were considered. This led to the
TU Delft
SKF ADC-SI
6.4 Results of Electromagnetic Field Simulations
63
Figure 6.13: New model of torque sensing bottom bracket designed according to the
design constraints of the bottom bracket and the guidelines laid out in the preliminary
conclusions.
design shown in figure 6.13. With this model a transient simulation was
done, in which the sleeve with the target ring was moving. The model was
meshed the same way as the earlier model and contained 37,819 tetrahedra.
To control the simulation time, the spindle moved with a speed of 200
m/s through the sensor, from -1 mm to +1 mm. This lead to a simulation
time of 0.01 seconds. This equates to 200 periods of the supply voltage.
The time step was chosen to be 5 · 10−6 seconds again. These parameters
resulted in 2,000 data points. The simulation lasted around 39 hours, again
on a single 3.2 GHz processor.
The resulting output of the voltmeter over the bridge circuit is shown
in figure 6.14. After demodulation, it would give a linear relation between
the displacement and the output of the bridge. A small offset of the minimum output was observed. Theoretically the minimum should be at zero
displacement, but in this simulation it was at 0.05 mm. This small offset is
presumed to be caused by the fact that the sleeve is not completely symmetrical, and therefore will influence the magnetic field. It functions like a
core.
Figure 6.15 shows a close up of the target ring and the coils. The arrows
represent the magnitude and direction of the induced current in the coils and
the target ring in the YZ-plane. It shows that the induced eddy currents in
the target ring are in opposite direction to the coil currents.
TU Delft
SKF ADC-SI
64
Design of Sensing Elements
Figure 6.14: Output voltage of the Wheatstone bridge versus the displacement of the
sleeve.
Figure 6.15: Close-up of target ring and coils with current magnitude and direction. It
shows the induced eddy currents are opposing the direction of the current in the coils.
Coils are represented as cross-sectional planes only for clarity.
TU Delft
SKF ADC-SI
6.5 Conclusions of Simulations
65
6.5 Conclusions of Simulations
The simulations of the sensor concept in Ansoft Maxwell showed that the
concept is feasible, and will produce a linear output. Clear design requirements are presented which can be used as design guidance. The next step
will be to build a functional prototype to proof the concept and validate the
simulations done in Maxwell.
TU Delft
SKF ADC-SI
66
TU Delft
Design of Sensing Elements
SKF ADC-SI
CHAPTER
7
Detailed Design of Torque Sensing Bottom Bracket
After defining the critical parameters for the sensor setup in the previous
chapter, both the mechanical and the electrical design could be completed.
This work was done by the mechanical engineer and the electrical engineer
respectively, both member of the project team. Their work is included in
this report for the sake of completeness and clarity.
7.1 Mechanical Design
To be able to comply with the CTS that were defined in chapter 4, several guidelines should be followed during the detailed design of the bottom
bracket.
• Maxwell simulations. The results of the simulations that were done
in Maxwell were used as guidelines for the final design of the torque
sensing bottom bracket. These guidelines were already presented in
section 6.4.7.
• ISO standards for the bottom bracket. For the design of bottom
brackets, there are two ISO standards. The first, ISO 6695:1991, describes the size and shape of the square ends of the bottom bracket
spindle, in order to be compatible with pedal cranks which are designed
following the same standard. The second standard, ISO 6696:1989, describes the size (diameter and width) and the screw thread of the frame
tube in which the bottom bracket will be fitted.
• Standard bearings. The last constraint of the design is the size
of bearings. The goal is to use standard deep groove ball bearings
(DGBBs).
TU Delft
SKF ADC-SI
68
Detailed Design of Torque Sensing Bottom Bracket
Figure 7.1: CAD design of the SKF Torque Sensing Bottom Bracket.
Following these constraints, the mechanical designer designed a torque sensing bottom bracket with the sensor concept as described in the previous
chapters as shown in figure 7.1. All the important design details will be
explained in the following sections.
7.1.1 Interface Spindle and Chain Wheel Bracket
As described in previous chapters, the chain wheel bracket will be integrated
in the bottom bracket assembly. The bracket will be connected to the spindle
through an intermediate piece, as shown in figure 7.2. The intermediate piece
will transfer the torque from the spindle to the chain wheel bracket. Because
the intermediate piece will be made of flexible material (either a rubber or a
plastic), a torsion angle between the spindle and the chain wheel bracket will
occur. The shape of the piece can be modified, for example by adding holes
(figure 7.3a), to obtain a better relationship between the applied torque and
the deformation of the piece. The deformation of the piece was only analyzed
TU Delft
SKF ADC-SI
7.1 Mechanical Design
69
Figure 7.2: Connection of chain wheel bracket to the spindle through a flexible intermediate piece, which will cause a torsion angle between the bracket and the spindle
when torque is applied.
with a preliminary finite element analysis, as shown in figure 7.3b. Because
the material behavior of plastics is very hard to model, the analysis of such
a piece can be a complete separate project. Therefore, for the first (proofof-concept) prototype, an injection mold will be fabricated to make several
pieces. When the proof-of-concept phase will turn out to be successful, a
more in-depth analysis will be necessary. This does however mean that the
relation between the applied torque and the translation of the sleeve will
remain a question until the proof-of-concept prototype will be tested.
7.1.2 Bearings
The torque sensing bottom bracket uses three deep groove ball bearings, as
shown in figure 7.1 in light blue. A normal bottom bracket uses only two
DGBBs (or sometimes needle bearings). Because the chain wheel bracket
is integrated into the bottom bracket, a third DGBB is needed to transfer
the vertical forces from the right pedal to the frame. When the design is
further improved, the inner of the two right DGBBs could be replaced by
a thinner needle bearing or a bushing. The outer of the two right DGBBs
is fitted outside the frame tube (so called external bearing), to be able to
accommodate a spindle of sufficient diameter.
TU Delft
SKF ADC-SI
70
Detailed Design of Torque Sensing Bottom Bracket
(a)
(b)
Figure 7.3: Intermediate piece connecting spindle and chain wheel bracket (a) with
holes to influence the behavior of deformation, (b) preliminary finite element analysis
using ANSYS, showing displacement under load.
7.1.3 Transformation from Rotation to Translation
After the applied torque has created a torsion angle, this torsion angle has to
be transformed into a displacement. A transformation ring is fitted around
the spindle and attached to the chain wheel bracket. This transformation
ring is shown in figure 7.4 in red. The transformation ring is connected to
the chain wheel through several slots (figure 7.4a). The transformation ring
and the sleeve (dark grey) have interlocking teeth with helical interlocking
surfaces. Because the sleeve is fitted around the spindle with three keykey way connections (figure 7.4b), the sleeve is blocked in the rotational
direction. When the transformation ring rotates, the sleeve will translate
away from the ring. A spring (light green, figure 7.1) will ensure that the
sleeve will return to its initial position when the applied torque is reduced.
The stiffness of this spring should be taken into account when calculating
the applied torque together with the material properties of the intermediate
part.
To protect the intermediate piece from a deformation beyond its elastic
limits, the transformation ring is also limited in rotation to an angle of 5
degrees, by using a notch, interlocking in a recess on the flange of the spindle.
7.1.4 Target Ring and Coils
The target ring will be fitted on the sleeve (light blue, figure 7.4a). For
the first prototype, the exact location of the ring will not be determined
mechanically, but the ring will be glued into place after measurement of the
complete assembly to account for fabrication and mounting tolerances. The
TU Delft
SKF ADC-SI
7.1 Mechanical Design
71
(a)
(b)
Figure 7.4: Mechanism for transforming the rotation to translation with (a) transformation ring (red) fitted between sleeve (dark gray, left) and chain wheel bracket (right),
connected to chain wheel bracket through several slots and interlocking teeth between
transformation ring and sleeve with helical sliding surfaces. (b) The rotation of the
transformation ring is limited to 5 degrees with notches interlocking in a recess of the
flange of the spindle. The sleeve is blocked in the rotational direction with three key-key
way connections.
coils will be wound into the two trenches in the outer body (see figure 7.1).
The coils will be wound manually for the first prototype.
TU Delft
SKF ADC-SI
72
Detailed Design of Torque Sensing Bottom Bracket
Figure 7.5: Electronic circuit for the prototype, containing the power supply, resonator
circuit, Wheatstone bridge and filtering.
7.2 Electrical Design
The electronics for the prototype will be located outside of the prototype
on a normal circuit board. When the proof-of-concept prototype turns out
to be successful, the electronics can be integrated into the product in a new
design iteration. The electronics can be miniaturized into a PCB. The PCB
will be located inside the housing or over molded in the housing.
The electronic circuit that will be used is shown in figure 7.5. The circuit
consists of four major parts.
• The power supply: supplying the circuit with 5 volt DC.
• A resonator circuit: to transform the direct current into an alternating current with a frequency of 12 kHz.
• The Wheatstone bridge: here, the coils will be connected to the
circuit board.
• Filtering: the sinusoidal output from the Wheatstone bridge will be
demodulated in order to get a single voltage as output.
TU Delft
SKF ADC-SI
Part III
Simulation & Testing
TU Delft
SKF ADC-SI
CHAPTER
8
BoB SimControl for Torque Sensing Bottom Bracket
This chapter will describe the BoB SimControl program. It is a MATLAB
interface to control the Maxwell simulation. All relevant parameters can be
changed in a simple graphical user interface (GUI). Batch simulations can
be made, and results are automatically generated and plotted.
8.1 Refining the Mesh
Because the measurement principle of the sensor is based on inducing alternating eddy currents into a metallic target ring, special care should be taken
when applying a mesh to the finite element analysis model. The current
density will be greatest at the surface of the conductor and will decrease
deeper into the material. This is called the skin effect. The skin depth, a
measure used to describe the distribution of the current density, will decrease
with increasing frequency. The equation for the skin depth is
r
δ=
2ρ
ωµ
(8.1)
With ρ being the resistivity of the conductor, ω the angular frequency and
µ the absolute magnetic permeability of the conductor. For the aluminium
target ring, at a sensor frequency of 12 kHz, the skin depth will be 0.7 mm.
That means that the meshing used in the simulation should be at least as
detailed to be able to simulate this effect correctly.
The simulations that were presented in chapter 6 were done on a three
dimensional model. A single simulation took around six minutes to calculate
the results. But to account for the skin effect, the meshing should be refined.
This would have a dramatic effect on the simulation time. Therefore, the
TU Delft
SKF ADC-SI
76
BoB SimControl for Torque Sensing Bottom Bracket
(a)
(b)
Figure 8.1: (a) 2-D Maxwell model of torque sensing bottom bracket, (b) close up view
of the refined meshing of the target ring and the coils.
Maxwell model was changed from a 3-D model to an axisymmetrical 2-D
model. To make sure the simulation was not excessively meshed in areas
that are less important, several mesh parameters were used. For the target
ring, a maximum element length of 0.25 mm was used. The coils were
meshed with a maximum element length of 0.5 mm and the sleeve with a
maximum element length of 0.9 mm. The spindle was meshed with the
default parameters. In this 2-D model, the simulation time was cut to a
mere twenty seconds. The results from the 3-D and the 2-D model, using
the same parameters (except for meshing) differed by 4.0 % (table 8.1), and
therefore it was concluded that the 2-D model could replace the 3-D model.
Figure 8.1 shows the 2-D model and the new meshing.
8.2 Visual Basic script
Up until this point, all modifications to the model, like changing of design
parameters as done in chapter 6, were done manually. This requires an
extensive knowledge of the Maxwell software and the way the model is conTU Delft
SKF ADC-SI
8.3 Structure of Program
77
Model
Sensitivity
[mV/mm]
Difference
[%]
3-D
2-D
177
184
4.0
Table 8.1: Sensitivity of torque sensor calculated using 3-D and 2-D Maxwell model.
structed. Sometimes the change of one parameter would mean that parameters of several parts should be changed. It is possible to parameterize the
model in Maxwell, but this still needs knowledge of the complete program
and model. A dedicated program was created in MATLAB to function as
an interface for this simulation.
Maxwell permits the use of dedicated Visual Basic (VB) scripts to control
the build of the model, the execution of the analysis and the post-processing
of the results. Almost all actions possible in the graphical user interface
(GUI) of Maxwell can be scripted to be executed automatically [22]. Because
writing a complete VB code can be a long and difficult process, it is also
possible to record a script. This way, all actions that are done in the GUI
are recorded in a VB script.
This option was used to create the VB script to create the model of the
torque sensing bottom bracket. The script should be modified to be able to
read the script in MATLAB, change the parameters and create a new VB
script. Figure 8.2 shows how this is done. A (default) model of the sensor
is built in Maxwell and all the actions are recorded in a VB script. This
script is then modified by hand once, so that MATLAB can work with it.
This new, parametrized script will be used as an input for the MATLAB
program. The other input will be the list of parameters of the analysis
the user wants to run next. The user will input these parameters into the
GUI of the MATLAB program. When the user executes the program, the
parameters will be inserted into the modified script to write a new VB script.
This new VB script will then be executed in Maxwell.
To read the code in MATLAB, the values of the parameters in the script
should be replaced by variables corresponding with the variables in the
MATLAB program. Figure 8.3a shows an example of VB code to create
a rectangle in Maxwell. Figure 8.3b shows the modifications necessary to
the script to be able to use it as an input for the MATLAB program.
8.3 Structure of Program
The program will use, as explained in figure 8.2, two inputs: the modified
script and all relevant parameters. Using the GUI (shown in figure 8.5), the
user will be able to change all the parameters concerning the dimensions of
TU Delft
SKF ADC-SI
78
BoB SimControl for Torque Sensing Bottom Bracket
Figure 8.2: Flow chart of the main structure of the program. A VB script is recorded
in Maxwell and modified to be used as input for the program. Together with the new
parameters a new VB script will be written and executed in Maxwell.
(a)
(b)
Figure 8.3: Example of recorded Visual Basic script for creating a rectangle (a) as
recorded in Maxwell, (b) modified to use as input in MATLAB.
TU Delft
SKF ADC-SI
8.3 Structure of Program
79
the model, the mesh-settings and the parameters for the analysis. There are
two main ways to execute the program.
• Execute a single run. This option enables the user to run a single
analysis of a certain parameter-set. This can be used for quick analysis
of a certain configuration. When the single run is finished, Maxwell
will remain open so that the user can directly analyze the results.
• Execute a batch of analyses. When the influence of a certain
parameter has to be analyzed, a batch can be run. This batch will
contain a plurality of analyses. The number of analyses will depend
on the range of the parameters. The user can choose to variate one
or two parameters. Next to these chosen parameters, the offset of the
sleeve and target ring will always by varied, in order to calculate the
sensitivity. The sensitivity will be calculated between the two outer
positions of offset per parameter variation. After all runs of the batch
are done, the program will process the results:
◦ It will plot and save the output of the sensor for each parameter
combination;
◦ It will calculate the sensitivity of the sensor for each parameter
combination;
◦ It will plot and save the sensitivity of the sensor versus the range
of the chosen parameter(s) in a 2-D or 3-D plot (for when 1 or 2
parameters are chosen to be varied respectively).
Figure 8.4 shows the architecture of the program as programmed in MATLAB. It uses several (sub)GUIs and m-files which will be briefly explained.
• Maxwell Interface GUI. Main GUI, used to input necessary parameters and launch single run analysis.
• GUI AnalysisSettings Sub GUI for changing the parameters associated with the analysis.
• GUI Batch Settings. Sub GUI for running a batch of analyses.
• RetrieveParameters.m. Script to recalculate parameters that are
dependent on other parameters. After a change of a parameter, all
dependent parameters are recalculated.
• WriteNewVBScript.m. Writes Visual Basic script according to syntax used by Maxwell to build the model and run the analysis in Maxwell.
• RunScriptInMaxwell.m. Executes the script in Maxwell by executing a command in the Linux command line.
TU Delft
SKF ADC-SI
80
BoB SimControl for Torque Sensing Bottom Bracket
Figure 8.4: Architecture of the MATLAB program. (top) The architecture of the single
run mode, and (bottom) the architecture of the batch mode. Orange boxes represent
(sub)GUIs, green boxes represent m-files.
TU Delft
SKF ADC-SI
8.4 Results
81
• RunBatchScript1Para.m Runs batch of analyses with one changing
parameter (+ parameter offset to be able to calculate the sensitivity).
Afterwards, the results will be plotted and saved.
• RunBatchScript2Para.m Runs batch of analyses with two changing
parameters (+ parameter offset to be able to calculate the sensitivity).
Afterwards, the results will be plotted and saved.
• WriteBatchInfoFile.m. Writes a .txt file with the parameters for
each run of the batch.
• CalcOutput.m. Calculates output using the results from the Maxwell simulation (results are exported from Maxwell in .csv format and
imported in MATLAB).
Figure 8.5: Screen shot of MATLAB program (BoB SimControl) to control simulation
in Maxwell.
8.4 Results
After all the code was written, a design validation plan (SVP) was written
and used to validate all separate functions. For each function, step by step
its functioning and correctness was checked, to make sure the program was
running correctly. The main document of the SVP can be found in appendix
E.
TU Delft
SKF ADC-SI
82
BoB SimControl for Torque Sensing Bottom Bracket
When a batch was executed, the time per run was around forty seconds.
Figure 8.6 shows the result of a batch analysis with two parameters: coil
spacing and the axial thickness of the target ring. The coil spacing was varied
between 0.5 and 4.5 mm, with 0.5 mm intervals. The target ring thickness
was varied between 1 and 5 mm, also with 0.5 mm intervals. So there were
9 · 9 = 81 parameter combinations. To calculate the sensitivity, there are at
least two points needed. All parameter combinations were simulated with
an offset of 0 and 1 mm. Therefore, a total of 162 runs were analyzed, with
a total run time of 74 minutes.
Figure 8.6: Result of a batch simulation with coil spacing and axial target ring thickness
as chosen parameters.
TU Delft
SKF ADC-SI
CHAPTER
9
Design of Experiment Analysis
In chapter 6 the influence of several parameters was investigated on a preliminary simulation model. The selection of the parameters that were varied
was mostly based on estimated guesswork. Also, the influence of several
parameters variating at the same time was not investigated. Therefore, it
was not possible to discover correlations between parameters. An example
of such a correlation can be seen in figure 8.6 in the previous chapter. It
shows that the combination of the coil spacing and the target ring thickness
has a specific effect on the sensitivity of the sensor.
9.1 Design of Experiment
Now that the final geometry of the concept is known, it is interesting to
analyze possible modifications of the concept. This is done by simulating
all possible design variations. However, with over ten variable parameters,
there will be hundreds of possible designs.
So, a more structured way of analyzing is the Design of Experiment
(DoE) method. This method has as goal to reach a conclusion with a small
but carefully planned set of experiments. The method consists of two phases:
screening and optimization.
9.1.1 Screening
Before one begins with the DoE analysis, there is probably little knowledge
of the system. Some influential parameters might be known (in this concept
the air gap will probably be of major importance), but there might be some
major effect present in the system that is not as easy to see. There are also
a lot of parameters to consider.
TU Delft
SKF ADC-SI
84
Design of Experiment Analysis
Figure 9.1: Bold approach of parameter variation in screening phase of DoE. [24]
The goal is to identify parameters and low order interactions that have
a significant effect on the response of the system. This way, all the other
parameters that are not significant can be eliminated.
To make this first selection, a bold approach is taken. The effects of the
variation of the parameters is presumed to be linear, as illustrated in figure
9.1. Therefore, only the outer edges of the range of the parameter have to
be taken into account. But when there are ten parameters, all with two
levels, to be analyzed, a full analysis would still consist of 210 = 1, 024 tests.
To reduce the number of tests, a fractional factorial experiment is designed. This means that several parameters are grouped into one, reducing
the number of experiments. By careful planning of these groups, the analysis can still yield a reliable result. The outcome of the analysis will not
have a sufficient resolution to directly select the optimal configuration, but
is detailed enough to separate the influential parameters. This planning can
be done using standardized tables. In this project, the program Minitab
was used to perform all data processing. [23]
9.1.2 Optimization
In the optimization phase, another DoE is done, but now limited to the
parameters that were found to be of great influence on the output, in this
case the sensitivity of the sensor. The goal of the optimization phase is
to gain more understanding of the mathematical relationship between the
parameters, and so be able to select the optimal combination of parameters.
9.2 Build Up of Simulation Model
To minimize the calculation time, the Maxwell model of the torque sensing
bottom bracket as described in chapter 7 is simplified into a 2-D model, as
explained in chapter 8. The model consists of four main parts: the spindle,
the sleeve, the target ring and the coils. The sleeve is initially modeled as
three parts, and thereafter merged into one. The schematic build up of the
TU Delft
SKF ADC-SI
9.3 Phase 1: Parameter Screening
85
Figure 9.2: Build up of analysis model showing the four main parts. The sleeve part is
built with three parts and merged thereafter.
model is shown in figure 9.2. All parts are modeled with a measurement in
the Z (axial) and X (radial) direction. Next to the physical measurement,
there are some other parameters that can be varied. All the parameters that
will be used in this DoE are summarized in table 9.1.
Parameter
unit
Default
Min
Max
Air Gap
Coil spacing
Spindle X
Sleeve A X
Sleeve A Z
Sleeve B X
Target ring X
Target ring Z
Coil X
Coil Z
Number of turns
Supply voltage
Supply frequency
mm
mm
mm
mm
mm
mm
mm
mm
mm
mm
V
Hz
1.25
0.5
8.5
3.25
32
0.5
1.5
2.5
2
2.1
80
5
12k
0.5
0.5
5
1
15
0.5
0.5
0.5
0.5
0.5
10
1
10k
5
10
10
5
35
5
5
10
2
5
500
10
1M
Table 9.1: Parameters, with range, for the screening phase of Design of Experiment
analysis.
9.3 Phase 1: Parameter Screening
At first, all the variables from table 9.1 were used for the first phase of
the DoE. Minitab designed an experiment with sixteen simulations. Some
simulations had the problem that they were not possible to run due to
interference problems. Two parts in the model would overlap, making the
geometry not possible. These simulations were then modified to make the
TU Delft
SKF ADC-SI
86
Design of Experiment Analysis
runs possible. The results of the experiment can be found in appendix F.1.
After this first screening phase, it could be concluded that five parameters
were depended on other parameters:
• Spindle X
• SleeveA X
• SleeveA Z
• Coil X
• Coil Z
It was decided to redo the screening phase only with the independent parameters. The data of the second screening phase can be found in appendix
F.2. The Pareto diagram of the second screening phase is shown in figure
9.3. The first thing that can be noticed is that a lot of parameters are influential when combined with the supply voltage. But it is presumed that the
supply voltage will influence the output just on a global scale: increasing
the supply voltage will always increase the output. Therefore, it is decided
to limit the second phase only to mechanical parameters. The four most
influential parameters are:
• Air gap
• SleeveB X
• Coil spacing
• Target ring Z
9.4 Phase 2: Parameter Optimization
The goal of the optimization phase is to find the optimal combination of
the four parameters mentioned above. In order to limit the number of
simulations, the air gap parameter is omitted in the simulations, because
the effect of this parameter is already clear. Reducing the air gap will alway
improve the sensitivity of the sensor. For the other three parameters, three
batches of simulations are done, varying two parameters while keeping one
on its default value. The range and interval of the parameters is summarized
in table 9.2. The results of these simulations can be found in figure 9.4.
Figure 9.4a shows that when the coil spacing is kept constant, there is
a clear optimum in the target ring thickness, at 2.5 mm. The influence of
sleeveB is far less important but has its optimum at 0.5 mm.
Figure 9.4b shows that when the target ring thickness is kept constant,
the coil spacing shows an optimum at its minimum value of 0.5 mm. The
TU Delft
SKF ADC-SI
9.4 Phase 2: Parameter Optimization
87
Figure 9.3: Result of screening phase of the DoE, showing that the parameters air
gap, sleeveB X, the coil spacing and target ring Z are the most influential mechanical
parameters.
Parameter
Min
[mm]
Max
[mm]
Interval
[mm]
Default
[mm]
SleeveB X
Target ring Z
Coil spacing
0.5
0.5
0.5
1.5
9.5
9.5
0.2
1
1
0.5
2.5
0.5
Table 9.2: Parameter range and interval for DoE phase 2.
influence on the sensitivity of the thickness of sleeveB is the same as in the
previous figure.
Figure 9.4c shows a clear coupling between the target ring thickness and
the coil spacing. The optimal combination is with a target ring thickness of
2.5 mm and a coil spacing of 0.5 mm. But when one parameter is changed,
the other parameter should changed as well, in order to keep the optimal
configuration.
These results show that the concept geometry as presented in chapter
7 was already close to the optimal geometry. When designing a new iteration of the sensor, reducing the air gap should be of the utmost importance.
Thereafter, care should be taken when changing either the target ring thickness or the coil spacing, because these parameters show a clear coupling.
This analysis with fixing one parameter while varying the other two
TU Delft
SKF ADC-SI
88
Design of Experiment Analysis
(a)
(b)
(c)
Figure 9.4: Results of the second phase of the DoE, varying two parameters while
keeping the third fixed on its default value (see table 9.2), (a) sleeveB X vs. target ring
Z, (b) sleeveB X vs. coil spacing, (c) target ring Z vs. coil spacing
TU Delft
SKF ADC-SI
9.5 Conclusions Design of Experiment Analysis
89
can be dangerous, because it will exclude a possible influence of the third
parameter. To make sure that this was not the case, a full factorial analysis
was done on these three parameters. To limit the number of simulations, the
step size of the parameters coil spacing and target ring Z was increased to 2
mm, instead of 1 mm. The step of parameter sleeveB X was increased to 0.5
mm. The parameters coil spacing and target ring Z had five levels and the
third parameter, sleeveB X, three. The full factorial analysis consisted thus
of 5 · 5 · 3 = 75 simulations. The results of these simulations can be found
in appendix F.3. On first inspection, these results point also to a maximum
around sleeveB X = 0.5 mm; coil spacing = 0.5 mm; target ring Z = 2.5
mm.
The results were also put into the Minitab software. The mathematical
model modeled the results in such a way that its conclusion was that the
optimum configuration was at: sleeveB X = 0.5 mm; coil spacing = 0.5
mm; target ring Z = 4.5 mm. As seen in figure 9.4 and in the results of
the full factorial, this seems not likely. In order to force Minitab to find
the maximum around a target ring thickness of 2.5 mm, the range of the
full factorial was reduced to a range between 0.5 and 4.5 mm for both coil
spacing and target ring Z. The results of the Minitab analysis can be seen
in figure 9.5. Minitab advises an optimum at: sleeveB X = 0.5 mm; coil
spacing = 0.5 mm; target ring Z = 2.88 mm. This result is more in line with
the empirical conclusions.
9.5 Conclusions Design of Experiment Analysis
The DoE analysis shows that it is dangerous to rely solely on one analysis,
either empirical or using software like Minitab. When only the results of
Minitab would have been followed, the advised optimal geometry would not
reflect the true optimal geometry. In order to optimize the geometry, the
knowledge gained by empirical analyses helped to guide the analysis into
the right direction. Why the initial result from Minitab was so far from
the empirical result is unknown, but it is suspected that the mathematical
fitting used in Minitab did not recognize the shape of the data, and therefore
fitted the data wrong.
The parameter sleeveB X was presumed to be of significant influence
at the end of the first phase of the DoE analysis. But the results shown
in the previous section indicate that the parameter has less influence than
expected. This is probably caused by the interference problems encountered
in the simulations. In the first phase, because of the big range chosen for
sleeveB X, in some simulations sleeveB X was bigger than the target ring.
In the second phase, the range was reduced to overcome the interference
problems and therefore it did not surpass the height of the target ring.
Presumably it caused the decrease of the influence of parameter sleeveB X.
TU Delft
SKF ADC-SI
90
Design of Experiment Analysis
(a)
(b)
(c)
Figure 9.5: Results from Minitab of the second phase of the DoE, using the results of
a full factorial analysis (see appendix F.3), (a) sleeveB X vs. target ring Z (coil spacing
= 0.5), (b) sleeveB X vs. coil spacing (target ring Z = 0.5), (c) target ring Z vs. coil
spacing (sleeveB X = 0.5).
TU Delft
SKF ADC-SI
CHAPTER
10
Assembly & Test of Prototype Bottom Bracket
A prototype according to the design presented in chapter 7 was fabricated.
The goal of this prototype was to function as a proof-of-concept. So it should
be fully functioning, but the actual performance was less important. The
fabrication of the parts, both metal and plastic parts, was done by a third
party. The assembly was done in-house, as described in section 10.1. The
sensor concept was tested outside of the assembly to evaluate its performance
in section 10.2. Unfortunately, there was no time to perform dynamic tests
before the print deadline of this thesis, but tests are planned in the near
future.
10.1 Assembly of Prototype
A prototype of the torque sensing bottom bracket was built to function as
a proof-of-concept. The results of the first test on this prototype would be
used to judge the validity of the chosen sensor concept and will be a basis
for further development.
As described in section 7.1.1, the intermediate piece would be made
using a mold. This mold is shown, with the molded part, in figure 10.1.
The material used for the molding was Macromelt OM 678, a thermoplastic
molding compound based on polyamide that is normally used as an overmolding material.
The coils were wound by hand, with copper wire with a diameter of 0.2
mm. The coils have sixty turns each and an inductance of around 210 µH.
The complete assembly of the internal components and of the finished
prototype is shown in figure 10.2.
TU Delft
SKF ADC-SI
92
Assembly & Test of Prototype Bottom Bracket
(a)
(b)
Figure 10.1: Mold for fabricating the intermediate part between the spindle and the
chain wheel bracket. (a) Mold, (b) intermediate part.
Figure 10.2: Complete assembled prototype.
TU Delft
SKF ADC-SI
10.2 Test of Measurement Principle
93
Figure 10.3: Test set-up for static test of measurement principle.
10.2 Test of Measurement Principle
The goal of these tests was to obtain the relationship between the displacement of the target ring and the output of the sensor. This was tested without
the complete assembly.
10.2.1 Test Set-Up
The sensor body containing the two coils was fixed on a aluminium bar
which is attached to an arm that can be moved in x,y and z-direction with a
precision of ±10µm. The target ring was fixed on a plastic tube which was
clamped onto the axis of rotation. The body was lowered over the target
ring. In the neighborhood of the coils, the output of the sensor was measured
as a function of the position. The output of the sensor was measured with an
oscilloscope. The set-up is shown in figure 10.3. With these measurements,
conclusions can be drawn about the theoretical sensitivity, the linearity and
the correlation with the Maxwell simulations. A Maxwell model with only
the target ring and the two coils gave a sensitivity of 560 mV/mm.
TU Delft
SKF ADC-SI
94
Assembly & Test of Prototype Bottom Bracket
10.2.2 Test Results
During the test the overall sensitivity of the sensor was measured and the
influence of eccentricity and mounting of the target ring was investigated.
Sensitivity
To measure the sensitivity, first the target ring was placed in the center of the
tube by moving the tube in the x and y plane until the ring was positioned
in the center of the opening of the tube. After the target ring was centered,
the tube was lowered over the target ring. Empirically, the central position
of the ring between the two coils was found by slowly lowering the tube over
the ring until the maximum output voltage was found. This position on the
z axis was set as relative zero point. Thereafter, the output of the sensor
was observed within a region of ±3 mm with a resolution of 0.1 mm. The
results of the measurements are plotted in figure 10.4.
The results shows an output that is not symmetric around the relative
zero point. This is caused by the influence of the aluminium arm that is
used to connect the tube to the xyz-table. The left half of the plot, between
−3 mm and 0 mm, represents the measurements that are not (or at least
less) influenced by the arm. In that region a linear output of the sensor is
visible. The sensitivity in that region is 120 mV/mm.
The measurements were repeated with a rotating target ring with a speed
of 5 rpm and 50 rmp. For both rotational speeds, the output and thus the
sensitivity remained the same.
Oscillation
During the measurement of the output of the sensor with a rotating target
ring, some oscillation in the output signal was noticed. Presumably, this
was caused by skewed assembly of the target ring on the plastic tube. To
investigate this oscillation, the target ring was fitted extremely askew on
purpose, and the output voltage was measured again on a specific position.
The average output remained the same as with a straight target ring. The
amplitude of oscillation however rose significantly to an amplitude of almost
four times greater than a straight target ring, as can be seen in table 10.1.
Target ring
Straight
Skewed
Amplitude
[mV]
6.5
23
Table 10.1: Oscillation of measurement signal caused by mounting of target ring.
TU Delft
SKF ADC-SI
10.3 Static Tests
95
800
700
Output [mV]
600
500
400
300
200
100
0
-3,00
-2,00
-1,00
0,00
1,00
2,00
3,00
Relative Z position [mm]
Figure 10.4: Output of sensor when moved ±3 mm around the center between the coils.
The plot shows a linear relation between the displacement and the output between −3
mm and the center. The other side of the plot is influenced by the aluminium arm of
the test set-up.
Eccentricity
Another influence on the output of the sensor can be the position of the
target ring in the plane of the coils. The eccentricity from the axis of the
spindle and the coils that can occur during assembly can cause a change in
the output voltage. To investigate this influence, the target ring was moved
±0.8 mm in both x and y direction. The results of these measurements are
plotted in figure 10.5. The initial x position of the target ring was 54 mm,
the initial y position was 23 mm. Both plots show roughly the same shape.
The increase of output voltage due to the eccentricity can be up to 3%.
10.3 Static Tests
The goal of the static test was the analyze whether the prototype was functioning. The spindle of the prototype was clamped in a workbench clamp.
The chain wheel bracket was rotated by hand. The output of the sensor
was measured with an oscilloscope. The output is shown in figure 10.6. The
prototype suffered from a lot of play in the connection between the spindle
and chain wheel bracket. After the initial play, it was hard to move the chain
wheel bracket by hand. The output varied between 200 mV and 350 mV.
The measurements show that the sensor concept works and has a sensitivity
of at least 150 mV/mm.
TU Delft
SKF ADC-SI
96
Assembly & Test of Prototype Bottom Bracket
685
Output [mV]
680
675
670
665
660
655
53,00
53,20
53,40
53,60
53,80
54,00
54,20
54,40
54,60
54,80
55,00
23,60
23,80
24,00
Absolute X position [mm]
(a)
685
Output [mV]
680
675
670
665
660
655
22,00
22,20
22,40
22,60
22,80
23,00
23,20
23,40
Absolute Y position [mm]
(b)
Figure 10.5: Influence of eccentricity of the target ring on the output voltage (a) varying
the absolute X position, (b) varying the absolute Y position.
TU Delft
SKF ADC-SI
10.4 Conclusions
97
400
Output [mV]
300
200
100
0
0
5
10
15
20
Time [s]
Figure 10.6: Output of the sensor during manipulation of the prototype which was
clamped to a work bench.
10.4 Conclusions
The test of the measurement principle showed that the output has a linear
relation to the displacement, as was expected. The sensitivity was lower than
simulated. The oscillation and eccentricity tests showed that the sensor is
sensitive to assembly errors. This is against the expectations. The reason
for the choice of this measurement concept was that the full circumferential
coils would eliminate these errors, because it would integrate the effect of
the full circumference of the target ring. A possible explanation is that hand
winding of the coils caused unevenly wound coils and therefore a not fully
homogeneous magnetic field.
TU Delft
SKF ADC-SI
98
TU Delft
Assembly & Test of Prototype Bottom Bracket
SKF ADC-SI
CHAPTER
11
Conclusions and Discussion
The objective of this thesis was to investigate the possibilities for developing
a torque sensing for SKF. This report showed which technologies already
exists, what the functions of the sensor should be and how this could be
realized.
11.1 Conclusions
The benchmarking phase showed that there are several existing technologies that are used to measure torque in either a power steering or e-bike
application. The two most prevailing technologies are magnetic field measurement and magnetic phase shift measurement. The former makes use of
the magnetostrictive effect and measures the stress applied to the sensing
element directly by measuring the properties of the magnetic field. The
latter measures a relative rotational displacement between two shafts which
are connected through a torsion bar.
A structured analysis of the requirements was made to investigate the
relative importance of all the functions of the product. By using the quality
function deployment (QFD) method, needs and functions were rated against
each other to find out which needs and functions are the most important. By
focusing on these topics first, a design can be made that will quickly satisfy
the costumer. The analysis showed that that, next to torque measurement,
price and safety are the most important functions. These results gave a good
inside in the priorities for the design, although some functions are not yet
realized in the prototype that was presented. Because this first design and
prototype had as goal to function as a proof-of-concept, less attention was
paid to the price for example. Although it has to be said that the chosen
TU Delft
SKF ADC-SI
100
Conclusions and Discussion
sensor concept, which consists only of a simple aluminium ring, two coils
and simple electronics, will probably be less expensive than other systems
that use Hall effect sensors and magnetized rings or shafts.
The design phase showed a structured approach to developing a sensing concept. After answering the questions where, what and how, a sensor
concept was chosen. It will be placed in the bicycle bottom bracket. It
will measure a torsion angle. And it will measure that angle by sensing a
displacement proportional to that torsion angle using an aluminium target
ring and two coils. This measurement concept was mainly chosen because
of its integrative properties. Because it is placed around the complete circumference of the spindle, harmonics were supposed to be eliminated. By
choosing for a highly conductive target ring, the leading physical principle
in the sensor would be the influence of induced eddy currents on the coils.
Finite element electromagnetic field simulations were made to analyze
the influence of varying design parameters. In the initial tests, the parameters were varied one by one on a simplified model. The results of this
analysis were used to design the torque sensing bottom bracket. In a second
analysis, a design of experiment (DoE) was set up to make a more structured
analysis, and to find interactions between several parameters. To speed up
the analyses and the variation of parameters, an interface was made using
MATLAB and modified Visual Basic scripts. The DoE was performed in
two phases. The first phase had as goal to find the most influential parameters, the second to optimize the geometry. The DoE showed that the initial
design was already close to the optimal geometry. It also showed a clear
connection between the width of the target ring and the spacing between
the coils.
At the end of the project, a fully functional prototype was made. The
initial test on the measurement principle and the prototype showed promising results. The measurement principle shows a linear relation between
the displacement of the target ring and the electronic output. Static tests
on the prototype indicate that the sensor concept is working. But the lack
of dynamic testing, due to lead time and test bench availability issues, a
final decision on the validity of the concept still has to be made.
11.2 Discussion
This thesis ends with a proof-of-concept prototype. But the product is far
from finished. To continue the project, several recommendations are given.
At the end, a small discussion is given on the usability of the concept for
other applications.
TU Delft
SKF ADC-SI
11.2 Discussion
101
11.2.1 Torque Sensing for E-bike Applications
As said at the end of the conclusion, performing the dynamic tests is of the
utmost importance to continue with the project. The dynamic tests will give
an insight in the dynamic behavior of the sensor. With the results of these
tests, the conclusions already drawn from the static test can be supported.
If the overall conclusion is that the chosen measurement concept is a suitable
concept, a start can be made with optimizing the design towards a product
that can be produced and sold to customers. The next phase should focus
on:
• Design of intermediate part. The plastic intermediate part that
transfers the torque from the spindle to the chain wheel, and deforms
under load, is at the moment not yet functioning as it should. A study
should be done to find a suitable material that will fulfill a number of
criteria. It should:
◦ deform linear over the defined range of applied torque;
◦ remain in the domain of elastic deformation;
◦ sustain constant varying deformation.
The development of this piece will require extensive knowledge of
plastics and rubbers and therefore will be a job for an expert, either
in-house or from a third party. It might even be interesting to investigate the possibility of eliminating the piece completely and use the
spring to define the relation between the applied torque the torsion
angle. However, this might introduce harmonics into the system because the spring system will not have the damping properties of the
plastic part.
• Simplify mechanical assembly. Although the sensor concept is
elegantly simple, the mechanical assembly is quite complex. The assembly consists of more than ten parts and some of them are quite
complicated. To simplify the assembly, it will be wise to review the
current design and look for improvement.
• Reduce costs. As already mentioned in the previous point, the complicated assembly will also contribute to the cost price. To cut the
cost, each part should be reviewed to cut cost. The current design
uses three ball bearing, as opposed to two in a traditional bottom
bracket. One of the ball bearings will only rotate up to six degrees, so
its main function is to support the vertical loading of the spindle. If
this ball bearing can be replaced by another part, considerable costs
could be saved.
TU Delft
SKF ADC-SI
102
Conclusions and Discussion
• Sensation to the cyclist. One aspect that has not yet been discussed is the sensation to the cyclist. The relative rotation between
the spindle and the chain wheel might give a weak sensation or the
sensation of power loss to the cyclist. At the moment, a maximum
rotation of six degrees was chosen, but this choice is not yet supported
by any research. More research should be done into the sensation of
the cyclist, and the way the cyclist applies the torque to the bike.
• Integration of electronics. The electronics have to be integrated in
to the bottom bracket assembly. This is not expected to be a problem,
but it will need attention for the next version.
11.2.2 Torque Sensing Concept for Other Applications
At the start of this thesis, it was not yet clear for what application the torque
sensor should be designed. After the market research and competitor analysis was done, it was decided to design a sensor for e-bike applications. But
now that a promising sensor concept is developed, it is time to look again
at other possible markets. The two most interesting markets are electronic
power steering (EPS) and drive train torque measurement. The latter one
involves the measurement of torque on drive shafts and gearboxes. The
presented sensor concept will be also suitable for these applications. The
design would even be simpler. The problem that occurred in the e-bike
sensor, measurement of two inputs (two pedals), does not exist in these
applications. Therefore, the intermediate piece could be replaced by a torsion bar or a spring. However, care should be taken not to breach existing
patents.
TU Delft
SKF ADC-SI
Part IV
Appendices
TU Delft
SKF ADC-SI
APPENDIX
A
Competitor Analysis Matrix
TU Delft
SKF ADC-SI
Espacenet.com
Espacenet.com
Espacenet.com
High precision
Torsion bar
Magnetic ring
High precision
Torsion bar
Magnetic ring
Advantage
Disadvantage
Conclusions
Document
Granted,
Application (12/2008)
Granted
Status
-
DE102005031086A1
EP2006650A2
-
∞
US2004011138A1
∞
-40°C to +120°C
-
∞
Stability
No torque ±0.03%
Max torque
±0.3%
Torsion bar
Magnetic ring
High precision
US2010139419A1,
WO2008068334A1
WO2010007068A1
Appl. (06-2010)
Appl. (06-2008)
Appl. (01-2010)
Espacenet.com
Espacenet.com
Espacenet.com
-
-
0.002°
-
±0.3% to ±0.05% FS
±0.2% FS
-
Magnetic RotorStator Phase Shift
Measurement
±4°
Magnetic RotorStator Phase Shift
Measurement
Magnetic RotorStator Phase Shift
Measurement
Unknown
Continental
±1° to ±8°
Announced
Bosch
Licensing technology
EPS
MMT
Number
Patents
Other specs
Angular range
Temperature range
Resolution
Accuracy
Hysteresis
Non-linearity
Range
Technical data
Measurement principle
Status of product
Picture
Field
Company
Torsion bar
Magnetic rings
Compact construction
Espacenet.com
Granted
Noncontacting
No torsion bar
Needs magnetized
shaft
Contacting method
Mech. components
Torsion bar
Espacenet.com
Granted
US2007034004A1
Total error: ±1%
∞
-40°C to +150°C
<0.1%
<0.1%
~0.1%
Costumer defined
Magnetic Field
Measurement
Production?
Siemens VDO
Proven sensor tech
Espacenet.com
Granted
DE102009011352B3
-
Total error
±0.066°
US7339370B2
?
-
-
-
Mechanical
Phase Shift
Measurement
Unknown
Bourns
∞
-
±4° to ±10°
2.5Nm to 4 Nm
0.0083°
1% FS
0.5%
-
Magnetic Rings
Phase Shift
Measurement
Production?
Bourns
Needs magnetized
shaft
Noncontacting
No torsion bar
Espacenet.com
Granted
US2002189372A1
Measuring error <2%,
Repeatability error
<0.1%
-40°C to +200°C
250°C peak
∞
10Nm to 5000Nm
Magnetic Field
Measurement
Production?
ABB
Needs magnetized
shaft
Noncontacting
No torsion bar
Espacenet.com
Granted
US2008115591A1
Repeatability: 0.05%
FS
∞
-
0.01% FS
-
Depends on design
Magnetic Field
Measurement
Licensing technology,
Production?
NTCE
Torsion bar
Magnetic ring
High precision
Espacenet.com
Appl. (11-2003)
EP1424541A2
No specs: patent for
production method
∞
-
-
-
Magnetic RotorStator Phase Shift
Measurement
Valeo
Torsion bar
Magnetic ring
High precision
Espacenet.com
Appl (09-2008)
WO2008105541A1
∞
Magnetic RotorStator Phase Shift
Measurement
NSK
Simple method
No signal post
processing
Mechanical
components
-
-
-
Magnetic Translation
Measurement
Production
"Honda Accord”
Noncontacting
No torsion bar
Needs magnetized shaft
Noncontacting
No torsion bar
Needs magnetized shaft
Remarks
Advantages
Disadvantages
€102 (sample)
WO2009079980A1
Appl. 07-2009
Espacenet.com
Purchased, to be
analyzed (test bench)
and disassembled
-
Speed measurement
-
Needs magnetized shaft
Noncontacting
No torsion bar
EP1978343A2
granted
Espacenet.com
-
-
-
2.5%
-
-
-
±200Nm
Magnetic Field
Measurement
Unknown
Shimano
0-90Nm (300Nm
optional)
Up to 1%
-
2 pedal separate
torque measurement
Resolution
Accuracy
Hysteresis
Non-linearity
Temperature
range
Other specs
Patents
Number
Status
Document
Price
Comments
Range
X-Cell RT
Magnetic Field
Measurement
Magnetic Field
Measurement
Measurement
principle
Technical data
Production
Thun
Announced
E-bike
Schaeffler/FAG
Status of
product
Picture
Field
Company
Simple method
Needs modification of
frame
High mounting
precision
Requires calculation
with microprocessor
Indirect measurement
Needs gear ratio
information
Used mainly on Dutch
e-bikes.
WO2006091089A2
Granted
Espacenet.com
10 mV/µm ± 10 %
-10°C to +40°C
-
0.3mm
TMM
Displacement of Rear
Drop-Out
Production
Cheap components
Purchased, to be
analyzed (test bench)
and disassembled
US6356847B1
Granted
Espacenet.com
$700 + $500 for MC
-
-
0.0025°
±0.5%
-
0 Nm to 450 Nm
Optical Phase Shift
Measurement
Production
IDbike
Simple method
Included in Shimano
STEPS package
US20100282001 A1
Appl. 11-2010
Espacenet.com
-
-
-
-
STEPS
Force in bottom
bracket
Production
Ergomo
Needs magnetized shaft
Needs magnetized shaft
Noncontacting
No torsion bar
Used on several
American bikes, like
Kalkhoff and Giant
Complete system of
motor, sensor, battery
and controller, uses
Schaffler system
Noncontacting
No torsion bar
-
-
-
-
-
Magnetic Field
Measurement
Production
Panasonic
-
-
-
-
-
eBike
-
Announced
Bosch
Not a torque sensor
Purchased, to be
analyzed and
disassembled
$ 23
-
-
-
-
Chain wheel torque (?)
sensor
Production
Chinese product
Magnetic rings
Integrated in motor
Compact construction
Integrated in motor
Purchased, to be
analyzed and
disassembled
$316
-
-
-
-
Torque sensor
integrated in hubmotor
Production
Chinese product
108
TU Delft
Competitor Analysis Matrix
SKF ADC-SI
APPENDIX
B
Thun Bottom Bracket Sensor Data Sheet
TU Delft
SKF ADC-SI
Sensory BB-Cartridges X-CELL R and X-CELL RT
Specifications
X-CELL R
X-CELL RT
Performance 1
Cadence: rotation/min.
Cadence: rotation/min.
Performance 2
Rotational direction
Rotational direction
Performance 3
-
Torque [Nm]
Length of spindles
120K; 120L; 128K; 128L; 136L
120K; 120L; 128K; 128L; 136L
Certification: EN 14764
(City-Trekking)
Yes
Yes
Certification: EN 14766 (MTB)
Yes
TBC
Cup threads
BS 1.375x24
BS 1.375x24
Right-hand cup
Low profile
Low profile
Material of cups
PA 6.6 Gf 30 %
PA 6.6 Gf 30 %
Material of sensor shell
Macromelt
Macromelt
Ball Bearings
2 x 61902 2RS
2 x 61902 2RS
Square
12.73 mm
12.73 mm
Surface of spindles
A2B
A2B
Assembly tool
Shimano® compatible
Shimano® compatible
Sensory system
2 x Hall-sensors
2 x Hall-sensors,
PCME-sensor
Impulse transmitter 1
Poled ring - 32 impulses/rotation
Poled ring - 32 impulses/rotation
Impulse transmitter 2
-
Magnetized spindle
Voltage feed
Analogue: +7…16 V DC,
Digital: +4…16 V DC
+7…+16 V DC
White cable (input)
Power supply
Power supply
Brown cable (output)
Sine signal
Sine signal
Blue cable (output)
Cosine signal
Cosine signal
Black cable (ground-connection)
Ground
Ground
Grey cable (output)
No connection
Torque signal
Length of cable
1100 mm
1100 mm
Different lengths optional:
surcharge applies
Signal output: sine
Analogue or
digital (open collector)
Analogue or
digital (open collector)
Analogue: offset +2.5 V
Amplitude max. 4.5 Vss
Digital: 0 V/Open collector
Signal output: cosine
Analogue or
digital (open collector)
Analogue or
digital (open collector)
Analogue: offset +2.5 V
Amplitude max. 4.5 Vss
Digital: 0 V/Open collector
Signal output torque: attribute 1
-
Offset +2500 mV at 0 Nm
Signal output torque: attribute 2
-
Analogue: ±10 mV/Nm
Signal output torque: attribute 3
-
Bandwidth: 250 Hz at -3 dB
Accuracy of signals: sine/cosine
± 3° (± 0,8 %)
± 3° (± 0,8 %)
Accuracy of signals: torque
-
Effective range ± 200 Nm
Accuracy of signals: torque
-
± 2.5 %
IP level
IP 56 as per EN 60529
IP 56 as per EN 60529
Technical Support:
Dipl.-Ing. Toni Valente
0049-2333-836-170
[email protected]
Dipl.-Ing. Tareq Higleh
0049-2333-836-115
[email protected]
Remarks
See drawing
For lower voltage feed
please contact technical
support.
Per turn of crank (360°)
Of effective range
WIRING DIAGRAM X-CELL R
AND X-CELL RT
X-CELL RT Digital (open collector)
Wire color
White
Black
Blue
Brown
Grey
Description
Power supply
Ground
Output
Output
Output
Signal
+7…16 V DC
0V
Cosine
Sine
Torque
Signal range
max. 20 mA
0 V/Open collector
0 V/Open collector
Offset +2,5 V bei 0 Nm
Remark
16 Impulses/Rotation
16 Impulses/Rotation
+/- 10 mV/Nm
Signal
+4…16 V DC
0V
Cosine
Sine
-
Signal range
max. 10 mA
0 V/Open collector
0 V/Open collector
-
Remark
16 Impulses/Rotation
16 Impulses/Rotation
-
Signal
+7…16 V DC
0V
Cosine
Sine
Torque
Signal range
max. 20 mA
Offset +2,5 V amplitude max. 4,5 Vss
Offset +2,5 V amplitude max. 4,5 Vss
Offset +2,5 V bei 0 Nm
Remark
16 Impulses/Rotation
16 Impulses/Rotation
+/- 10 mV/Nm
Signal
+7…16 V DC
0V
Cosine
Sine
-
Signal range
max. 15 mA
Offset +2,5 V amplitude max. 4,5 Vss
Offset +2,5 V amplitude max. 4,5 Vss
-
Remark
16 Impulses/Rotation
16 Impulses/Rotation
-
X-CELL R Digital (open collector)
Wire color
White
Black
Blue
Brown
Grey
Description
Power supply
Ground
Output
Output
No connection
X-CELL RT Analogue
Wire color
White
Black
Blue
Brown
Grey
Description
Power supply
Ground
Output
Output
Output
X-CELL R Analogue
Wire color
White
Black
Blue
Brown
Grey
Description
Power supply
Ground
Output
Output
No connection
SIGNAL CHARACTERISTICS OF X-CELL R AND X-CELL RT
TABLE 1
TABLE 2
Signal characteristics angle sensor analogue version
sinus digital [V]
sinus analogue [V]
cosinus analogue [V]
4,000
voltage [V]
5,000
3,500
0,000
4,500
0
3,000
50
100
2,500
150
200
angle [°] of the crank
250
300
350
300
350
signal characteristics angle sensor digital version
2,000
cosinus digital [V]
5,000
voltage [V]
1,500
1,000
0,500
0,000
0
50
100
150
200
angle [°] of the crank
250
300
350
5,000
100
150
200
angle [°] of the crank
250
TABLE 2:
X-CELL R and RT digital version
signal torque sensor [V]
4,000
3,500
TABLE 3:
X-CELL RT torque
3,000
2,500
2,000
Thun Sensor
1,500
VC ~3…16V
Aquivalent Output Stage
Digital Hall-sensor
Open collector
1,000
0,500
0,000
-250
50
TABLE 1:
X-CELL R and RT analogue version
Signal characteristics torque sensor
4,500
0,000
0
TABLE 3
voltage [V]
voltage [V]
Signal characteristics angle sensor digital version
5,000
Cable brown/blue
-200
-150
-100
-50
0
torque [Nm]
50
100
150
200
250
Hall-sensor
220R
2n2
Cable black
0V
0V
Thun Sensor
VC ~3…16V
Äquivalenter Ausgang
Digitaler Hall-Sensor
„Open collector“
220R
2n2
Kabel schwarz
0V
Controller
„Pull-Up“
Widerstand erforderlich
~4…20kOhm
zum Controller-Eingang
Kabel braun/blau
Hall-Sensor
Controller
Pull-Up
Resistor required
~4…20kOhm
to Controller input
0V
112
TU Delft
Thun Bottom Bracket Sensor Data Sheet
SKF ADC-SI
APPENDIX
C
Paired Comparison Matrix
TU Delft
SKF ADC-SI
n
tio
of
To
Measure Torque
Measure Position
1
0,1
10
4
4
1 0,143 0,143
4
7
4
7
0,1 0,143 0,143 0,143
4
4
0,1 0,143
7
ht
es
nc
t
a
n
os
te
rb
e
o
e
c
d
w
i
l
n
u
p
u
t
i
le
st
si
ee
te
rq
ab
ce
ly
di
in
ab
ca
er
al
e
To
Po
Sp
i
a
l
d
n
c
p
t
i
n
b
n
re
re
re
n
u
an
so
fe
lia
no
ai
m
st
su
su
su
ha
ea
e
sa
m
re
h
r
m
t
v
i
ea
ea
ea
ec
i
f
M
M
M
M
Co
G
Be
Be
Be
W
O
i th
ei
g
.
ce
w
st
cu
en
es
r
p
al
iz
ed
in
.
pl
Ap
ta
l
1, 4, 7, 10
ed
at
gr
.
N
Reading direction
st
cu
l
pp
A
.
or
m
E-bike, Torque
56
18%
0,1 2,257
1%
Measure Speed
0,25
7
1
4
1
4
1
7
0,25
1
1
27,5
9%
Mechanically integrated in cust. Appl.
0,25
7
0,25
1
0,25
1
0,25
7
0,25
1
4 22,25
7%
Communicate with cust. Appl.
0,25
10
1
4
1
4
1
7
4
7
4 43,25
14%
0,143
7
0,25
1
0,25
1
0,25
7
0,25
1
4 22,14
7%
0,25
7
1
4
1
4
1
10
4
7
4 43,25
14%
7 0,143 0,143 0,143 0,143
0,1
Give no perception of presence
Be safe
Be maintainable
0,143
Be reliable
0,25
10
4
4
Withstand disturbances
0,25
7
1
1 0,143
0,143
10
1
Of reasonable cost
0,25
0,25
0,25
4
1 0,143 0,143
0,1
9,2
3%
0,25
7
1
4
4 38,75
13%
1 0,143
7
0,25
1
1 19,79
6%
10
0,25
1
1 24,39
8%
0,25
0,25
APPENDIX
D
House of Quality #1
TU Delft
SKF ADC-SI
-
-
--
- -
+
-
-
+
+
+
+
+
4
7
d Ra
18%
Measure Position
CTS
w We
ight
4
Norm
alize
Shock test
Temperature test
Moisture test
Vibration test
Fail safe function
4
No modification to cust.
appl.
4
Aesthetics
7
Standerdized Connector
4
Price
1
Protocol
Time durarion during
repair
Assembly time duration (to
standard)
Weight
Resolution Speed
Resolution Torque
O
Offset Speed
Amplitude Speed
Attributes
O
Offset Torque
Amplitude Torque
4
Refresh Rate
Input voltage (noise)
Repeatability Speed
Repeatability Torque
7
O
Time lag
7
O
Extra friction
10
Measure Torque
Accuracy Speed
0, 1, 4, 7, 10
Accuracy Torque
Range Torque
E-bike, Torque
Range Speed
CTQ
O
Endurance test
+
++
+
+
Over torque test
+
+
+
+
+
+
No parts of customer appl.
destroyed during maintenance
-
1%
10
Measure Speed
Mechanically integrated in
cust. Appl.
Communicate with cust. Appl.
4
Give no perception of presence
4
7
7
4
7
1
4
10
7
1
7
4
7
7
4
7
7
1
1
4
4
1
4
7
7
1
4
9%
4
10
10
7
7
10
4
4
1
4
Be safe
10
Be maintainable
7
Be reliable
Withstand disturbances
4
Of reasonable cost
1
7
7
7
7
7
1
4
1
4
4
4
4
4
7
4
4
4
4
4
1
4
1
7
10
1
4
4
1
7
4
4
4
1
4
7
7
1
1
4
4
7%
4
14%
7%
1
1
14%
4
7
7
1
4
1
1
10
4
4
1
7
4
4
3%
10
7
7
4
4
13%
4
10
4
1
10
6%
7
1
4
4
8%
3,7 2,0 2,2 1,2 3,5 2,6 2,6 1,2 1,2 1,9 1,0 1,0 2,5 1,6 0,9 0,8 1,2 1,2 0,3 4,5 2,4 1,5 1,1 1,9 4,0 1,7 0,7 3,0 2,9 2,5 1,1 2,4
Weights
Mean
LSL
USL
Sigma
6% 3% 4% 2% 6% 4% 4% 2% 2% 3% 2% 2% 4% 3% 1% 1% 2% 2% 0% 7% 4% 2% 2% 3% 6% 3% 1% 5% 5% 4% 2% 4%
APPENDIX
E
SVP BoB SimControl
TU Delft
SKF ADC-SI
BoB SimControl
Simulation Verification/Validation Plan & Report
Simulated Application :
Matlab interface for BB testing in Maxwell
Mechatronic Engineer : Mr. Marien van Ditten
Test #
Assessme
nt
References
Test Name/Purpose
Acceptance Criteria
Revision: 1
Date : 18-04-2011
Quality Engineer :
Project Manager:
MANAGER:
Test
Respons
ible
Inital Tests
Initial
Testing
Test
Test
completio
Scheduled
n Date
Tests results
Report #
Block Simulations Verification
TITLE:
11-1
Maxwell_interface_GUI.m
Launch Maxwell_interface_GUI by typing
Maxwell_Interface_GUI
OK
Change value in GUI and
observe if the value in the Change variable SpindleZ, obsever change in workspace
workspace is changed
to make variables visible in
SpindleX
workspace, use
GetVarsInWS.m
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-2
OK
11-3
OK
11-4
OK
SleeveAZ
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-5
OK
SleeveAX
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-6
OK
SleeveBX
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-7
OK
SleeveCZ
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-8
OK
SleeveCX
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-9
OK
TargetRingZ
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-10
OK
TargetRingX
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-11
OK
CoilZ
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-12
OK
CoilX
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-13
OK
CoilSpacing
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-14
OK
Airgap
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-15
OK
Offset
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-16
OK
Mesh Coils
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-17
OK
Mesh Target Ring
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-18
OK
Mesh Sleeve
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-19
OK
Number of Turns
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-20
OK
Circuit Voltage
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-21
OK
Circuit Frequency
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-22
OK
File Name
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
11-23
OK
Reads DefaultValuesModel.m
Push 'Return to Default', all values in GUI should return to
their default value
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_11.doc
TITLE:
GUI_AnalysisSettings.m
12-1
OK
Push Analyis Settings: open Analysis Settings GUI
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_12.doc
12-2
OK
Change Simulation Time, push save: Value in workspace
changes
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_12.doc
12-3
OK
Analysis Step
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_12.doc
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_12.doc
12-4
OK
Change value in GUI and
observe if the value in the Results Save start
workspace is changed
12-5
OK
Results Save stop
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_12.doc
12-6
OK
Results Save step
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_12.doc
TITLE:
WriteNewVBScript.m
13-1
OK
New VB file is written to build model
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_13.doc
13-2
OK
New VB file is written for the external circuit
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_13.doc
13-3
OK
New VB file is written for the analysis
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_13.doc
Simulation Validation Plan_BOB.XLS
Page 1 of 6
BoB SimControl
Simulation Verification/Validation Plan & Report
Simulated Application :
Matlab interface for BB testing in Maxwell
Mechatronic Engineer : Mr. Marien van Ditten
Test #
Assessme
nt
References
Test Name/Purpose
13-4
OK
Change parameters and
run WriteNewVBScript.m
13-5
Acceptance Criteria
Revision: 1
Date : 18-04-2011
Quality Engineer :
Project Manager:
MANAGER:
Test
Respons
ible
Inital Tests
Initial
Testing
Test
Test
completio
Scheduled
n Date
Tests results
Report #
Change dimensions in GUI, observe change in script
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_13.doc
OK
Change mesh parameters in GUI, observe change in script
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_13.doc
13-6
OK
Change electrical parameters in GUI, observe change in
script
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_13.doc
13-7
OK
Change analysis parameters in GUI, observe change in
script
MvD
18/04/2011
18/04/2011
18/04/2011
SVP_BoB_13.doc
Run RunScriptInMaxwell.m, Maxwell should open and
execute the script
MvD
19/04/2011
19/04/2011
19/04/2011
SVP_BoB_14.doc
The script VB_BuildModel.VBS should build the model
MvD
19/04/2011
19/04/2011
19/04/2011
SVP_BoB_14.doc
All model parameters should correspond with inputted values
in the WS
MvD
19/04/2011
19/04/2011
19/04/2011
SVP_BoB_14.doc
All mesh parameters should correspond with inputted values
in the WS
MvD
19/04/2011
19/04/2011
19/04/2011
SVP_BoB_14.doc
13-8
OK
TITLE:
14-1
OK
14-2
OK
RunScriptInMaxwell.m
When
RunScriptInMaxwell.m is
run, all the parameters
controlled by the program
should be in the Maxwell
simulation
14-3
OK
14-4
OK
14-5
OK
All analysis parameters should correspond with inputted
values in the WS
MvD
19/04/2011
19/04/2011
19/04/2011
SVP_BoB_14.doc
14-6
OK
The script VB_Analysis.VBS Should launch the analysis, and
plot the output of the measurement of the voltmeter
MvD
19/04/2011
19/04/2011
19/04/2011
SVP_BoB_14.doc
MvD
19/04/2011
19/04/2011
19/04/2011
SVP_BoB_15.doc
TITLE:
15-1
OK
TITLE:
RetrieveParameters.m
Check calculations after
Compare parameters in WS to results in excel sheet
parameter change in
RetrieveParameters.m, by
using calculation in excelsheet SVP_BoB_15.xls
GUI_BatchSettings.m
16-1
OK
Create vector OffsetBatch, with values in corresponding
boxes
MvD
20/04/2011
20/04/2011
20/04/2011
SVP_BoB_16.doc
16-2
OK
Create vector Parameters, filled with ones
MvD
20/04/2011
20/04/2011
20/04/2011
SVP_BoB_16.doc
16-3
OK
Modify Parameters, insert the length of OffsetBatch as the
first element
20/04/2011
20/04/2011
20/04/2011
16-4
OK
MvD
20/04/2011
20/04/2011
20/04/2011
SVP_BoB_16.doc
16-5
OK
MvD
20/04/2011
20/04/2011
20/04/2011
SVP_BoB_16.doc
16-6
OK
MvD
20/04/2011
20/04/2011
20/04/2011
SVP_BoB_16.doc
16-7
OK
Change value BatchVar2 corresponding to chosen parameter
in second pull down menu
MvD
20/04/2011
20/04/2011
20/04/2011
SVP_BoB_16.doc
16-8
OK
Create BatchVar2Vecotr, with values in corresponding boxes
MvD
20/04/2011
20/04/2011
20/04/2011
SVP_BoB_16.doc
16-9
OK
Modify Parameters for the chosen parameter
MvD
20/04/2011
20/04/2011
20/04/2011
SVP_BoB_16.doc
TITLE:
Change value BatchVar1 corresponding to chosen parameter
When either Run batch – 1 in first pull down menu
parameter of Run batch –
2 Parameters is pushed,
Create BatchVar1Vecotr, with values in corresponding boxes
the information in the GUI
is put in to the appropriate
variables
Modify Parameters for the chosen parameter
RunBatchScript2Para.m
17-1
OK
Create new *_BatchInfo.txt file
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
17-2
OK
Create BatchIteration parameter
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
17-3
OK
For each batch iteration: Increase BatchIteration parameter,
update FileName, change chosen parameter for batch, write
new line in ParameterVector
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
17-4
OK
Save output voltmeter in Maxwell sim as .csv file
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
17-5
OK
Calculate Output of voltmeter
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
write ParametersString
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
Save vectors: ParameterVector, Result, Parameters and
ParametersString
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
Calculate sensitivity of each configuration, plot and save the
results as .fig and .png, and save in Sensitivity parameter
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
In GUI_BatchSettings, the
parameters for the offset
are chosen and 2
parameters are selected.
The button Run Batch -2
parameters is pushed, and
the batch of simulations is
launched
17-6
OK
17-7
OK
17-8
OK
17-9
OK
Save vectors: Sensitivity, BatchVar1Vector,
BatchVar2Vector
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
17-10
OK
Make surface and contour plot of results (Sensitivity) with
the right parameters on the X and Y axis.
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
Simulation Validation Plan_BOB.XLS
Page 2 of 6
BoB SimControl
Simulation Verification/Validation Plan & Report
Matlab interface for BB testing in Maxwell
Simulated Application :
Mechatronic Engineer : Mr. Marien van Ditten
Revision: 1
Date : 18-04-2011
Quality Engineer :
Project Manager:
MANAGER:
Test
Respons
ible
Inital Tests
Test #
Assessme
nt
17-11
OK
save vector: SensMatrix
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
17-12
OK
Save surface and 3D plot as .fig and .png
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
17-13
OK
If chosen batch variables are switched in GUI, plots will
change accordingly
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
Create new *_BatchInfo.txt file
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_17.doc
TITLE:
References
Test Name/Purpose
Acceptance Criteria
Initial
Testing
Test
Test
completio
Scheduled
n Date
Tests results
Report #
RunBatchScript1Para.m
18-1
OK
18-2
OK
Create BatchIteration parameter
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
18-3
OK
For each batch iteration: Increase BatchIteration parameter,
update FileName, change chosen parameter for batch, write
new line in ParameterVector
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
18-4
OK
Save output voltmeter in Maxwell sim as .csv file
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
Calculate Output of voltmeter
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
write ParametersString
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
Save vectors: ParameterVector, Result, Parameters and
ParametersString
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
In GUI_BatchSettings, the
parameters for the offset
are chosen and 1
parameters are selected.
The button Run Batch -1
parameters is pushed, and
the batch of simulations is
launched
18-5
OK
18-6
OK
18-7
OK
18-8
OK
Calculate sensitivity of each configuration, plot and save the
results as .fig and .png, and save in Sensitivity parameter
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
18-9
OK
Save vectors: Sensitivity, BatchVar1Vector,
BatchVar2Vector
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
18-10
OK
Make plot of results (Sensitivity) with the right parameters
on the X.
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
18-11
OK
Save plot as .fig and .png
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_18.doc
Write new line for each iteration containing all parameters
that can be used in a batch
MvD
20/04/2011
20/04/2011
20/04/2011
SVP_BoB_19.doc
MvD
21/04/2011
21/04/2011
21/04/2011
SVP_BoB_20.doc
19-1
TITLE:
WriteBatchInfoFile.m
OK
While running a batch, a
file is written which
contains all values of the
TITLE:
20-1
OK
Simulation Validation Plan_BOB.XLS
CalcOutput.m
CalcOutput will caluculate
the average between the
minimum and maximum of
Calculate the output
the second half of the
graph of the output of the
voltmeter.
Page 3 of 6
APPENDIX
F
Design of Experiment
F.1 DoE Screening 1
TU Delft
SKF ADC-SI
122
Design of Experiment
F.2 DoE Screening 2
F.3 DoE Full Factorial
TU Delft
SKF ADC-SI
Bibliography
[1] C.-J. Yang 2010 Launching strategy for electric vehicles: Lessons from
China and Taiwan, Technological Forecasting & Social Change 77,
p.831-834
[2] Bike Europe 2010 In Holland One out of Eight Bikes Is Electric,
http://www.bike-eu.com/news/4030/in-holland-one-outof-eight-bikes-is-electric.html (15-12-2010)
[3] http://road.cc/content/news/
18907-shimano-announce-steps-e-bike-groupset, (12-05-2011)
[4] http://www.bosch-ebike.de
[5] EBSC Manuel de service version 4.0, E-bike service manual for IonTechnology E-bikes
[6] Schaeffler (UK) Ltd. 2010 Precision bearings make electric bikes more
reliable and energy efficient, Press release 000-002-672 GB-EN, 30-112010, Sutton Coldfield U.K.
[7] Akira Noguchi et al. 2004, Development of a Steering Angle and Torque
Sensor of Contact-type, Furukawa Review, No.25
[8] Moving Magnet Technologies 2004 Position sensor, designed in particular for detecting a steering column torsion, patent US7,028,545
[9] SSI Technologies Inc. 2007 Position and torque sensor,
patent application WO2007/067196 A1
[10] Bourns Inc 2010 Torsion angle sensor,
patent application US2010/0224011 A1
TU Delft
SKF ADC-SI
124
Bibliography
[11] Schaeffler KG 2009 Tretlager mit Drehmomentsensorik,
patent application WO2009/079980A1
[12] http://www.ncte.de/ncte/cms/upload/downloads/
motorsport/Handout.pdf , (12-05-2011)
[13] Loı̈c La Pierre Honda Accord EPAS Systems, Application note, SKF
ACD-SI
[14] http://www.ergomousa.com/index.cfm/ergomo-products/
?event=store.item&itemGUID=b2ca8fbb-8bf8-4e7c-a9531b7f4dd0ed23, (12-05-2011)
[15] http://www.idbike.com/tmm-powermanagement.htm, (12-05-2011)
[16] Shimano Inc 2010 Bicycle bottom bracket force sensor,
patent application US2010/0282001 A1
[17] NTCEngineering GmbH 2008 Torque Sensor,
patent application US2008/0115591 A1
[18] http://www.thun.de/thun eng/sensor technology video.html,
(14-04-2011)
[19] Dr. Yoji Akao 1994 QFD: The Customer Driven Approach to Quality
Planning and Deployment, Asian Productivity Organization, Tokyo,
Japan
[20] http://www.lionprecision.com/tech-library/technotes/
tech-pdfs/article-0011-cve.pdf, (12-05-2011)
[21] John P. Bentley 2005 Principles of Measurement Systems, 4th edition,
Pearson Education Limited, Essex, England
[22] Introduction to Scripting in Maxwell, 5th edition September 2010,
ANSYS Inc., Canonsburg, Pensylvania, USA
[23] http://www.minitab.com
[24] Design of Experiments Basics, Six Sigma Black Belt training, SKF
TU Delft
SKF ADC-SI
TU Delft
SKF ADC-SI
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement