Design and prototyping of PCB stator electric motor

with FDM 3D printed parts

Word count: 25440

Ewoud Collijs
Student number: 0170 4499

Ries Vereecken
Student number: 0150 6230

Supervisor: Prof. dr. ir. Peter Sergeant

Counsellor: Ing. Jelle De Decker

Master's dissertation submitted in order to obtain the academic degree of

Master of Science in de industriële wetenschappen: elektrotechniek

Academic year 2020-2021

Design and prototyping of PCB stator electric motor
with FDM 3D printed parts
Word count: 25440

Ewoud Collijs
Student number: 0170 4499

Ries Vereecken
Student number: 0150 6230

Supervisor: Prof. dr. ir. Peter Sergeant

Counsellor: Ing. Jelle De Decker

Master's dissertation submitted in order to obtain the academic degree of

Master of Science in de industriële wetenschappen: elektrotechniek

Academic year 2020-2021

 Preface
To bring this thesis to a successful ending we have several people to thank. This research could not
have been accomplished without their supervision and vision to approach certain problems.

First, we express our special thanks to our promoter prof. dr. ir. Peter Sergeant and our supervisor
ing. Jelle De Decker. They assisted us throughout the process and provided us with their insights on
the problem that needed to be solved to accomplish our goal. The insights presented by prof. dr. ir
Peter Sergeant and ing. Jelle De Decker were extremely helpful and will provide us with a way of
thinking for future problems that need to be solved.

Secondly, we would like to show our gratitude to the Ghent University for providing the required
materials and equipment to conduct our research. Also, our sincere thanks to the 3D-print lab of
Ghent University and ing. Jelle De Decker for guiding us in the 3D printing discipline and printing all
our different parts with great precision.

Last but not least we would like to thank our families for giving us the proper motivation and support
throughout the process of this research thesis. We are very grateful for everyone who helped us
throughout the process.

II
 Abstract
The axial flux motor with permanent magnets already has a small axial length. This research will
place the stator windings on a PCB for further reduction of the axial length and weight. The goal is to
make a working motor with 3D-printed components as base material. First, the electromagnetic
analytical model to calculate the different variables of the motor will be described. The results of this
model will be verified in two ways. The first verification is the comparison with the finite element
method program, the second verification will be testing the motor itself. The mechanical model used
for this motor is based on the results of the electromagnetic analytical model. Afterwards, the motor
is assembled, at this point a few problems arise. The 3D prints are not strong enough to withstand
the forces produced by the magnets. This problem was easily solved by enlarging the backplate and
using steel spacers instead of 3D-printed ones. At the end, the motor is tested on a 3D-printed
testbench. The comparison of the test results with the electromagnetic analytical model results are
remarkably close to each other. Therefore, it can be assumed that the model is a particularly
adequate estimation of the reality. This thesis provides a comprehensive explanation of the methods
and models used.

III
Contents
Preface..................................................................................................................................................... II
Abstract .................................................................................................................................................. III
Contents .................................................................................................................................................. 1
List of Figures........................................................................................................................................... 4
List of Tables ............................................................................................................................................ 7
List of Abbreviations ................................................................................................................................ 8
List of Symbols ......................................................................................................................................... 9
1 Introduction ................................................................................................................................... 12
1.1 Importance of Research ........................................................................................................ 12
1.2 Goal ....................................................................................................................................... 12
1.3 Outline ................................................................................................................................... 13
2 Literature Review .......................................................................................................................... 14
2.1 Axial-flux Machine ................................................................................................................. 14
2.1.1 Structure ........................................................................................................................ 14
2.1.2 Working principle .......................................................................................................... 15
2.2 PCB Winding Design .............................................................................................................. 17
2.2.1 Manufacturing process of a PCB ................................................................................... 17
2.2.2 Concentric Windings...................................................................................................... 17
2.2.3 Wave Windings .............................................................................................................. 19
2.3 Electromagnetic Finite Element Analysis .............................................................................. 20
2.3.1 FEMM ............................................................................................................................ 20
2.4 Application possibilities ......................................................................................................... 23
2.4.1 Electro-mobility ............................................................................................................. 23
2.4.2 Electric devices .............................................................................................................. 24
2.4.3 Automation.................................................................................................................... 24
3 Electromagnetic analytical model ................................................................................................. 25
3.1 Back EMF ............................................................................................................................... 25
3.1.1 Magnetic flux density .................................................................................................... 25
3.1.2 Flux linkage per pole...................................................................................................... 27
3.1.3 Back EMF ....................................................................................................................... 31
3.2 Torque ................................................................................................................................... 33
3.3 Maximum turns per winding ................................................................................................. 38
3.4 Electrical Wire Resistance ..................................................................................................... 39
3.5 Magnetic flux density 3D visualization .................................................................................. 41

1
4 Electromagnetic Design Using Analytical Model ........................................................................... 43
4.1 Magnet Choice and magnet topology ................................................................................... 43
4.1.1 Configurations ............................................................................................................... 44
4.1.2 Magnet Test Setup 1 ..................................................................................................... 45
4.1.3 Magnet Test Setup 2 ..................................................................................................... 45
4.1.4 Magnet Final Configuration ........................................................................................... 47
4.2 PCB Winding Design .............................................................................................................. 48
4.3 Backplate analysis.................................................................................................................. 53
4.3.1 Electromagnetic design ................................................................................................. 53
4.3.2 Computation of axial electromagnetic forces ............................................................... 57
5 Mechanical design ......................................................................................................................... 61
5.1 Back plate deflection ............................................................................................................. 61
5.2 Motor topology ..................................................................................................................... 64
5.3 Rotor design .......................................................................................................................... 65
5.4 Stator design.......................................................................................................................... 68
5.5 Assembled motor .................................................................................................................. 70
5.6 Test Bench ............................................................................................................................. 73
6 Motor Controller ........................................................................................................................... 74
6.1 General .................................................................................................................................. 74
6.2 Specifications ......................................................................................................................... 74
7 Thermal Analyses .......................................................................................................................... 76
7.1 Topology ................................................................................................................................ 76
7.2 Calculations ........................................................................................................................... 77
7.3 Conclusion ............................................................................................................................. 79
8 Motor assembly............................................................................................................................. 80
8.1 Assembly Rotor...................................................................................................................... 80
8.2 Assembly Stator ..................................................................................................................... 83
8.3 Assembly process .................................................................................................................. 86
8.4 Test setup .............................................................................................................................. 88
8.5 Failures and improvements ................................................................................................... 91
9 Prototype testing ........................................................................................................................... 92
9.1 Back EMF testing ................................................................................................................... 92
9.2 Torque and efficiency testing ................................................................................................ 95
9.2.1 Extra losses beside motor.............................................................................................. 98
10 Conclusion ............................................................................................................................... 104
11 Future work ............................................................................................................................. 106

2
 11.1 Wave versus concentric windings ....................................................................................... 106
11.1.1 Electromechanical performance ................................................................................. 106
11.1.2 Manufacturability ........................................................................................................ 106
11.1.3 Milling .......................................................................................................................... 106
11.1.4 Punching ...................................................................................................................... 106
11.2 Wedge shaped magnets ...................................................................................................... 107
11.3 Back iron saturation ............................................................................................................ 107
11.4 Magnet modelling ............................................................................................................... 107
11.5 Modular implementation .................................................................................................... 108
11.6 Mechanical analysis ............................................................................................................. 109
11.7 Thermal analysis .................................................................................................................. 109
11.8 Motor losses ........................................................................................................................ 109
11.9 Test setup ............................................................................................................................ 110
Reference List ...................................................................................................................................... 111
Appendix 1: Drafts ............................................................................................................................... 114
Appendix 2: Bearing losses .................................................................................................................. 123

3
List of Figures
Figure 2.1: Basic structure of a single rotor axial flux motor with PM; Stator (1); rotor with permanent
magnets (2). Adapted from [3] .............................................................................................................. 14
Figure 2.2: Structure used for this thesis. Adapted from [3] ................................................................ 15
Figure 2.3: Magnetic field induced by permanent magnets. ................................................................ 16
Figure 2.4: Area of Utilization comparison; (a) Trapezoidal winding; (b) Rhomboidal winding ........... 18
Figure 2.5: Different kinds of wave windings; (a) Parallel windings, (b) Arc windings and (c) Unequal
width parallel windings. Adapted from [6] ........................................................................................... 19
Figure 2.6: topology of the motor structure in FEMM .......................................................................... 21
Figure 2.7: Structure divided into triangles by the triangle executable of FEMM. ............................... 21
Figure 2.8: Results calculated and plotted by FEMM (Inductance B in tesla [T]).................................. 22
Figure 3.1: Rotor Magnet Configuration with calculation variables ..................................................... 25
Figure 3.2: Bar magnet dimensions ....................................................................................................... 26
Figure 3.3: Comparison of analytical and FEM magnetic flux density field B ....................................... 27
Figure 3.4: Concentrated spiral windings approximated by concentric loops with calculation
parameters ............................................................................................................................................ 28
Figure 3.5: Coil absolute position .......................................................................................................... 29
Figure 3.6: Flux linkage per pole............................................................................................................ 30
Figure 3.7: Bavaria tool, winding connection diagram. Adapted from [10] .......................................... 31
Figure 3.8: Generated back EMFs in phase 1 at 6000 𝑟𝑝𝑚 .................................................................. 32
Figure 3.9: Torque production calculation using concentric loops with calculation variables ............. 34
Figure 3.10: Torque production from phase 1 and its individual coils .................................................. 35
Figure 3.11: Torque production of each phase and full motor torque ................................................. 36
Figure 3.12: Influence of load angle ς on motor torque ....................................................................... 37
Figure 3.13: Track length calculation setup with calculation variables................................................. 39
Figure 3.14: 3D magnetic flux density plot............................................................................................ 41
Figure 3.15: 3D magnetic flux density plot, top view ............................................................................ 42
Figure 4.1: Different magnet topologies ............................................................................................... 44
Figure 4.2: Magnetic flux density field for different magnet topologies .............................................. 46
Figure 4.3: Motor power in function of track radii and track width ..................................................... 49
Figure 4.4: Side view of motor power in function of track radii and track width ................................. 50
Figure 4.5: (Left) Motor power in function of track width; (Right) Maximum number of turns per
winding in function of track width ........................................................................................................ 51
Figure 4.6: (Left) Back EMF in function of track width; (Right) Maximum phase current in function of
track width............................................................................................................................................. 51
Figure 4.7: (Left) Motor torque in function of track width; (Right) Phase resistance in function of track
width...................................................................................................................................................... 51
Figure 4.8: PCB coil track design; (a) Front; (b) Back............................................................................. 52
Figure 4.9: Magnet saturation modelling with backplate thickness = 5 mm; (Top) FEM magnet flux
density plot; (Bottom) Comparison between analytical and FEM magnetic field graph ...................... 54
Figure 4.10: Magnet saturation modelling with backplate thickness = 0.5mm; (Top) FEM magnet flux
density plot; (Bottom) Comparison between analytical and FEM magnetic field graph ...................... 55
Figure 4.11: Motor power in function of back plate thickness ............................................................. 56
Figure 4.12: Motor power density in function of back plate thickness ................................................ 56
Figure 4.13: FEM analysis with minimal fringing effects ....................................................................... 57
Figure 4.14: Magnetic flux density plot along magnet surface ............................................................. 58
Figure 4.15: Magnetic attraction force calculation setup ..................................................................... 58

4
Figure 5.1: Representation of back plate deflection due to magnetic forces ....................................... 61
Figure 5.2: Back plate deflection in function of back plate thickness ................................................... 63
Figure 5.3: Rotor design with exploded view; (a) Front; (c) Side; (1) Magnet retainer; (2) Magnets; (3)
Steel back plate; (4) Bearing; (5) Bearing mounting plate; (6) Mounting screws; (7) Washers and
standoff spacers. ................................................................................................................................... 65
Figure 5.4: Backplate design with magnets; (a) Radial stick out; (b) Assembly lugs; (c) Clamping lugs.
............................................................................................................................................................... 67
Figure 5.5: Stator design with exploded view; (a) Front; (c) Side; (1) Bearing; (2) Aluminium spacer; (3)
Aluminium flange clamp; (4) Electrical insulator; (5) PCB stator; (6) Shaft with soldered aluminium
flange; (7) Motor wiring ........................................................................................................................ 68
Figure 5.6: PCB connection in detail...................................................................................................... 69
Figure 5.7: Assembled motor with exploded view; (a) Front; (c) Side .................................................. 70
Figure 5.8: Sliced side view of assembled motor .................................................................................. 71
Figure 5.9: Motor assembly dimensions ............................................................................................... 72
Figure 5.10: Exploded view test bench; (1) Torque measurement arm; (2) Bearing clamps; (3)
Bearings; (4) Bearing holding brackets; (5) DC motor. .......................................................................... 73
Figure 5.11: Full motor test bench ........................................................................................................ 73
Figure 6.1: Motor controller needed specifications with a track width = 2,06 mm ............................. 75
Figure 7.1: Simplified thermal structure of the motor .......................................................................... 76
Figure 7.2: Thermal resistance network ................................................................................................ 78
Figure 8.1: Rotor halve; (a) Back plate backside; (b) Backplate with glued magnets; (c) M3 clamp hole;
(d) M3 tapped assembly hole; (e) Magnet; (f) Magnet retainer; (g) M2,5 nut ..................................... 80
Figure 8.2: Rotor parts; (a) Rotor halve, (b) Bearing; (c) Bearing bracket; (d) M2,5 bolts .................... 81
Figure 8.3: Assembled rotor halves; (a) Back; (b) Front ........................................................................ 81
Figure 8.4: Rotor angled view................................................................................................................ 82
Figure 8.5: (a) PCB stator front; (b) PCB stator back; (c) Solder connection point; (d) Mounting hole 83
Figure 8.6: (a) Shaft; (b) Aluminium clamping plate; (c) Wire; (d) Mounting M2 nut; (e) Mounting M2
bolt; (f) Backside of solder joint; (g) Soldered flange; (h) Flange solder connection ............................ 83
Figure 8.7: (a) PCB stator front view; (b) Stator with spacers top view; (c) PCB stator with spacers
back view; (d) Spacer wire side; (e) Spacer non-wire side; (f) Soldered flange .................................... 84
Figure 8.8: (a) Bearing installed wire side; (b) Bearing installed non-wire side; (c) Bearing installed top
view ....................................................................................................................................................... 85
Figure 8.9: Assembly process; (a) Assembly bolts; (b) Clamping bolt; (c) Rotor spacer; (d) Washer fine
adjustment spacer; (e) PCB standoff spacer ......................................................................................... 86
Figure 8.10: Final assembled motor; (a) Non-wire side; (b) Wire side; (c) Side view; (d) Holes for M2,5
bearing bracket bolts; (e) PCB stator .................................................................................................... 87
Figure 8.11: (a) Motor non-wire side; (b) Test setup bracket; (c) Motor wire side; (d) Test setup
bracket; (e) M2,5 clamping bolt; (f) Test setup bearing; (g) Press fitted shaft ..................................... 88
Figure 8.12: (a) Brace; (b) Motor; (c) Holding bracket .......................................................................... 89
Figure 8.13: (a) Torque weight; (b) Torque measurement arm; (c) Counterbalance plate; (d) Rotating
shaft; (e) Holding bracket for secondary DC motor; (f) Stationary shaft .............................................. 90
Figure 9.1: Back EMF test setup; (a) Secondary drive motor; (b) Motor under test; (c) Motor coupling;
(d) Oscilloscope measuring back EMF line-line voltage. ....................................................................... 92
Figure 9.2: Back EMF test setup speed measurement; (a) Tachometer ............................................... 92
Figure 9.3: Comparison of measured back EMF with analytical model at different speeds................. 94
Figure 9.4: Torque/efficiency test setup; (a) Motor power supply; (b) Power supply output current
measurement; (c) Motor under test; (d) Secondary motor used as generator/brake; (e) Adjustable

5
torque measurement arm; (f) Tachometer; (g) Generator load; (h) Load current measurement; (i)
Motor controller ESC. ............................................................................................................................ 95
Figure 9.5: Motor torque measurement setup; (a) Counterweight for torque measurement arm; (b)
Torque measurement arm; (c) Torque measurement weight .............................................................. 95
Figure 9.6: Schematic for an adjustable load for the torque/efficiency test [23] ................................. 96
Figure 9.7: Schematic representation with extra losses; PSU 3D model adapted from [24]; ESC 3D
model adapted from [25] ...................................................................................................................... 98
Figure 9.8: Efficiency in function of input current and DC bus voltage of Aerostar 50A ESC [26] ........ 99
Figure 9.9: Exploded view test bench; (1) Torque measurement arm; (2) Bearing clamps; (3) Bearings;
(4) Bearing holding brackets; (5) DC motor......................................................................................... 100
Figure 11.1: Magnetic flux density field with approximation error .................................................... 108
Figure 11.2: Modular motor system with 3 individual elements ........................................................ 108

6
List of Tables
Table 3.1: Parameters for flux density field comparison ...................................................................... 27
Table 3.2: Parameters used in flux linkage analysis .............................................................................. 30
Table 4.1: Magnet topology parameters ............................................................................................... 44
Table 4.2: Coverage ratio of different magnet topologies .................................................................... 45
Table 4.3: Magnet test setup 2 parameters .......................................................................................... 45
Table 4.4: Generated back EMF per magnet topology ......................................................................... 46
Table 4.5: Motor parameters for testing optimal winding configuration. ............................................ 49
Table 4.6: Final motor design characteristics ........................................................................................ 52
Table 4.7: Back plate test motor parameters ........................................................................................ 53
Table 4.8: Analysis results of the magnet saturation modelling ........................................................... 55
Table 4.9: Magnetic attraction forces comparison ............................................................................... 59
Table 5.1: Deflection comparison between analytical and FEM with a p𝑙𝑎𝑡𝑒 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 = 3 𝑚𝑚 ..... 62
Table 5.2: Motor assembly dimensions ................................................................................................ 72
Table 7.1: Parameters for each layer .................................................................................................... 76
Table 7.2: Calculated variables for the thermal analysis....................................................................... 78
Table 9.1: Back EMF and velocity constant ........................................................................................... 93
Table 9.2: Torque/efficiency measurement data using resistors.......................................................... 97
Table 9.3: Torque/efficiency measurement data using PWM adjustable motor load .......................... 97
Table 9.4: ESC efficiencies at different working states ....................................................................... 100
Table 9.5: Bearing power loss and true motor efficiency. .................................................................. 102
Table 0.1: Geometric constants 𝐾𝑍 and 𝐾𝐿 ....................................................................................... 124
Table 0.2: Geometric and load dependent variables for rolling and sliding frictional moments - Deep
groove ball bearings ............................................................................................................................ 125
Table 0.3: Geometric constants for rolling and sliding frictional moments of deep groove ball bearings
............................................................................................................................................................. 125

7
List of Abbreviations
AF Axial Flux
AFPM Axial Flux motor with Permanent Magnets
Back EMF Back Electromotive Force
CA Cyanoacrylate
EV Electric Vehicle
ESC Electronic Speed Controller
FEA Finite Element Analysis
FEM Finite Element Method
FEMM Finite Element Method Magnetics
IGBT Insulated-Gate Bipolar Transistor
NdFeB Neodymium Iron Boron
NRMSE Normalized Root Mean Square Error
PCB Printed Circuit Board
PM Permanent Magnet
PWM Pulse Width Modulation
TD Transverse Direction
UBEC Universal Battery Eliminator Circuit
VSI Voltage Source Inverter
2D Two-dimensional space
3D Three-dimensional space

8
List of Symbols
Electromagnetic modelling

𝐵 = 𝑀𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑙𝑢𝑥 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 [𝑇]

𝐵̅ = 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑙𝑢𝑥 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 [T]
𝐵𝑟𝑒𝑚 = 𝑅𝑒𝑚𝑎𝑛𝑒𝑛𝑡 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑠𝑎𝑡𝑖𝑜𝑛 𝑜𝑓 𝑡ℎ𝑒 𝑝𝑒𝑟𝑚𝑎𝑛𝑒𝑛𝑡 𝑚𝑎𝑔𝑛𝑒𝑡𝑠 [T]
𝑧 = 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑓𝑟𝑜𝑚 𝑏𝑎𝑐𝑘 𝑝𝑙𝑎𝑡𝑒 𝑡𝑜 𝑡ℎ𝑒 𝑑𝑒𝑠𝑖𝑟𝑒𝑑 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑝𝑙𝑎𝑛𝑒 [m]
𝑧2 = 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑓𝑟𝑜𝑚 𝑏𝑎𝑐𝑘 𝑝𝑙𝑎𝑡𝑒 𝑡𝑜 𝑡ℎ𝑒 𝑚𝑖𝑑𝑑𝑙𝑒 𝑙𝑖𝑛𝑒 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑡ℎ𝑒 2 𝑟𝑜𝑡𝑜𝑟𝑠 [m]
𝑟 = 𝑅𝑎𝑑𝑖𝑢𝑠 [m]
𝜇0 = 𝑉𝑎𝑐𝑢𝑢𝑚 𝑝𝑒𝑟𝑚𝑒𝑎𝑏𝑖𝑙𝑖𝑡𝑦 [H⁄m]
𝜇𝑃𝑀 = 𝑅𝑒𝑙𝑎𝑡𝑖𝑣𝑒 𝑝𝑒𝑟𝑚𝑒𝑎𝑏𝑖𝑙𝑖𝑡𝑦 𝑜𝑓 𝑝𝑒𝑟𝑚𝑎𝑛𝑒𝑛𝑡 𝑚𝑎𝑔𝑛𝑒𝑡
ℎ𝑃𝑀 = 𝐻𝑒𝑖𝑔ℎ𝑡 𝑜𝑓 𝑡ℎ𝑒 𝑝𝑒𝑟𝑚𝑎𝑛𝑒𝑛𝑡 𝑚𝑎𝑔𝑛𝑒𝑡 [m]
𝑙𝑃𝑀 = 𝐿𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑡ℎ𝑒 𝑝𝑒𝑟𝑚𝑎𝑛𝑒𝑛𝑡 𝑚𝑎𝑔𝑛𝑒𝑡 [m]
𝑤𝑃𝑀 = 𝑊𝑖𝑑𝑡ℎ 𝑜𝑓 𝑡ℎ𝑒 𝑝𝑒𝑟𝑚𝑎𝑛𝑒𝑛𝑡 𝑚𝑎𝑔𝑛𝑒𝑡 [m]
𝑝 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑝𝑜𝑙𝑒 𝑝𝑎𝑖𝑟𝑠
𝜃𝑐𝑠 = 𝐶𝑜𝑖𝑙 𝑠𝑝𝑎𝑛 [rad]
𝜏𝑝 = 𝑀𝑎𝑔𝑛𝑒𝑡 𝑝𝑜𝑙𝑒 𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑤𝑖𝑑𝑡ℎ [m]
𝜏𝑚 = 𝑀𝑎𝑔𝑛𝑒𝑡 𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑤𝑖𝑑𝑡ℎ [m]
𝜑 = 𝑃𝑜𝑙𝑎𝑟 𝑐𝑜𝑜𝑟𝑑𝑖𝑛𝑎𝑡𝑒 𝑜𝑛 𝑟𝑜𝑡𝑜𝑟 [m]
𝜃𝑚 = 𝑀𝑒𝑐ℎ𝑎𝑛𝑖𝑐𝑎𝑙 𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛 𝑜𝑓 𝑟𝑜𝑡𝑜𝑟 [rad]
𝜃𝑒 = 𝐸𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑎𝑙 𝑑𝑒𝑔𝑟𝑒𝑒𝑠 [rad]
𝑡𝑠𝑙 = 𝑅𝑎𝑑𝑖𝑎𝑙 𝑠𝑙𝑖𝑐𝑒 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 [m]
𝑆𝑠𝑙 = 𝑆𝑢𝑟𝑓𝑎𝑐𝑒 𝑎𝑟𝑒𝑎 𝑜𝑓 𝑠𝑙𝑖𝑐𝑒 [m2 ]
𝜔 = 𝑅𝑜𝑡𝑎𝑡𝑖𝑜𝑛𝑎𝑙 𝑠𝑝𝑒𝑒𝑑 [rad⁄s]
𝑤𝑡𝑟 = 𝑊𝑖𝑑𝑡ℎ 𝑜𝑓 𝑃𝐶𝐵 𝑡𝑟𝑎𝑐𝑘 [m]
𝑤𝑠 = 𝑊𝑖𝑑𝑡ℎ 𝑜𝑓 𝑃𝐶𝐵 𝑡𝑟𝑎𝑐𝑘 [m]
𝑤𝑑 = 𝐷𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑡ℎ𝑒 𝑐𝑒𝑛𝑡𝑒𝑟𝑙𝑖𝑛𝑒𝑠 𝑜𝑓 2 𝑎𝑑𝑗𝑎𝑐𝑒𝑛𝑡 𝑃𝐶𝐵 𝑡𝑟𝑎𝑐𝑘𝑠 [m]
𝑡𝑡𝑟 = 𝑃𝐶𝐵 𝑡𝑟𝑎𝑐𝑘 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 [µm]
𝜙𝐼𝑛,𝑖 = 𝑆𝑡𝑎𝑟𝑡𝑖𝑛𝑔 𝑎𝑛𝑔𝑙𝑒 𝑜𝑓 𝑖 𝑡ℎ 𝑖𝑛𝑛𝑒𝑟 𝑙𝑜𝑜𝑝 [rad]
𝜙𝑂𝑢𝑡,𝑖 = 𝑆𝑡𝑎𝑟𝑡𝑖𝑛𝑔 𝑎𝑛𝑔𝑙𝑒 𝑜𝑓 𝑖 𝑡ℎ 𝑜𝑢𝑡𝑒𝑟 𝑙𝑜𝑜𝑝 [rad]
β𝐼𝑛,𝑖 = 𝐴𝑟𝑐 𝑙𝑒𝑛𝑔ℎ𝑡 𝑖𝑛𝑛𝑒𝑟 𝑎𝑛𝑔𝑙𝑒 𝑜𝑓 𝑖 𝑡ℎ 𝑙𝑜𝑜𝑝 [rad]
β𝑂𝑢𝑡,𝑖 = 𝐴𝑟𝑐 𝑙𝑒𝑛𝑔ℎ𝑡 𝑜𝑢𝑡𝑒𝑟 𝑎𝑛𝑔𝑙𝑒 𝑜𝑓 𝑖 𝑡ℎ 𝑙𝑜𝑜𝑝 [rad]
𝐿𝑂𝑢𝑡,𝑖 = 𝐿𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑡ℎ𝑒 𝑜𝑢𝑡𝑒𝑟 𝑎𝑟𝑐 𝑜𝑓 𝑡ℎ𝑒 𝑖 𝑡ℎ 𝑙𝑜𝑜𝑝 [m]
𝐿𝐼𝑛,𝑖 = 𝐿𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑡ℎ𝑒 𝑖𝑛𝑛𝑒𝑟 𝑎𝑟𝑐 𝑜𝑓 𝑡ℎ𝑒 𝑖 𝑡ℎ 𝑙𝑜𝑜𝑝 [m]
𝐿𝑠,𝑖 = 𝐿𝑒𝑛𝑔𝑡ℎ 𝑜𝑓 𝑏𝑜𝑡ℎ 𝑠𝑡𝑟𝑎𝑖𝑔ℎ𝑡 𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 𝑤𝑖𝑟𝑒𝑠 𝑜𝑓 𝑡ℎ𝑒 𝑖 𝑡ℎ 𝑙𝑜𝑜𝑝 [m]
𝑟𝑜 = 𝑂𝑢𝑡𝑒𝑟 𝑤𝑖𝑛𝑑𝑖𝑛𝑔 𝑟𝑎𝑑𝑖𝑢𝑠 [m]
𝑟𝑖 = 𝐼𝑛𝑛𝑒𝑟 𝑤𝑖𝑛𝑑𝑖𝑛𝑔 𝑟𝑎𝑑𝑖𝑢𝑠 [m]
𝑟𝑠𝑜 = 𝑂𝑢𝑡𝑒𝑟 𝑃𝐶𝐵 𝑟𝑎𝑑𝑖𝑢𝑠 [m]
𝑟𝑠𝑖 = 𝐼𝑛𝑛𝑒𝑟 𝑃𝐶𝐵 𝑟𝑎𝑑𝑖𝑢𝑠 [m]
𝑟𝑜,𝑃𝑀 = 𝑂𝑢𝑡𝑒𝑟 𝑚𝑎𝑔𝑛𝑒𝑡 𝑟𝑎𝑑𝑖𝑢𝑠 [m]
𝑟𝑖,𝑃𝑀 = 𝐼𝑛𝑛𝑒𝑟 𝑚𝑎𝑔𝑛𝑒𝑡 𝑟𝑎𝑑𝑖𝑢𝑠 [m]
𝛼𝑐 = 𝑇𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 𝑜𝑓 𝑢𝑠𝑒𝑑 𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙 [1⁄K]
𝜌20°𝐶 = 𝑅𝑒𝑠𝑖𝑠𝑡𝑖𝑣𝑖𝑡𝑦 𝑜𝑓 𝑢𝑠𝑒𝑑 𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙 𝑎𝑡 20°𝐶 [Ωm]
𝜌𝑇 = 𝑅𝑒𝑠𝑖𝑠𝑡𝑖𝑣𝑖𝑡𝑦 𝑜𝑓 𝑢𝑠𝑒𝑑 𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑜𝑟 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙 𝑎𝑡 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑇 [Ωm]
𝑇 = 𝑇𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 [°C]
𝑅𝑝ℎ𝑎𝑠𝑒 = 𝑇𝑜𝑡𝑎𝑙 𝑝ℎ𝑎𝑠𝑒 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 [Ω]
𝑡𝐵𝑝 = 𝑆𝑡𝑒𝑒𝑙 𝑏𝑎𝑐𝑘 𝑝𝑙𝑎𝑡𝑒 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 [m]
𝜑𝑀𝐼𝑁,𝑙,𝑞 = 𝐼𝑛𝑛𝑒𝑟 𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑖𝑜𝑛 𝑙𝑖𝑚𝑖𝑡 𝑓𝑜𝑟 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑙𝑢𝑥 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 𝑓𝑖𝑒𝑙𝑑 𝑜𝑓 𝑠𝑙𝑖𝑐𝑒 𝑞 𝑓𝑟𝑜𝑚 𝑙𝑜𝑜𝑝 𝑙
[rad]
𝜑𝑀𝐴𝑋,𝑙,𝑞 = 𝑈𝑝𝑝𝑒𝑟 𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑖𝑜𝑛 𝑙𝑖𝑚𝑖𝑡 𝑓𝑜𝑟 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑙𝑢𝑥 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 𝑓𝑖𝑒𝑙𝑑 𝑜𝑓 𝑠𝑙𝑖𝑐𝑒 𝑞 𝑓𝑟𝑜𝑚 𝑙𝑜𝑜𝑝 𝑙
[rad]

9
𝜑𝑀𝐼𝑁,𝑐 = 𝐴𝑏𝑠𝑜𝑙𝑢𝑡 𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛 𝑜𝑓 𝑡ℎ𝑒 𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑒𝑑 𝑐𝑜𝑖𝑙 𝑖𝑛 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑡𝑜 𝑎 𝑧𝑒𝑟𝑜 𝑝𝑜𝑖𝑛𝑡
𝑑𝑒𝑓𝑖𝑛𝑒𝑑 𝑜𝑛 𝑡ℎ𝑒 𝑠𝑡𝑎𝑡𝑜𝑟 [rad]
𝜑𝑀𝐴𝑋,𝑐 = 𝐴𝑏𝑠𝑜𝑙𝑢𝑡 𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛 𝑜𝑓 𝑡ℎ𝑒 𝑠𝑝𝑒𝑐𝑖𝑓𝑖𝑒𝑑 𝑐𝑜𝑖𝑙 𝑖𝑛 𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒 𝑡𝑜 𝑎 𝑧𝑒𝑟𝑜 𝑝𝑜𝑖𝑛𝑡
𝑑𝑒𝑓𝑖𝑛𝑒𝑑 𝑜𝑛 𝑡ℎ𝑒 𝑠𝑡𝑎𝑡𝑜𝑟 [rad]
𝜍 = 𝑇𝑜𝑟𝑞𝑢𝑒 𝑎𝑛𝑔𝑙𝑒 [rad]
∅𝑙 = 𝑇𝑜𝑡𝑎𝑙 𝑓𝑙𝑢𝑥 𝑓𝑜𝑟 𝑙𝑜𝑜𝑝 𝑙 [Wb]
𝜓𝑐𝑜𝑖𝑙 = 𝑇𝑜𝑡𝑎𝑙 𝑓𝑙𝑢𝑥 𝑝𝑒𝑟 𝑐𝑜𝑖𝑙 [Wb]
Î = 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝑎𝑚𝑝𝑙𝑖𝑡𝑢𝑑𝑒 [A]
𝑓 = 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 [Hz]
𝐹 = 𝐿𝑜𝑟𝑒𝑛𝑡𝑧 𝑓𝑜𝑟𝑐𝑒 [N]
𝐹𝑛 = 𝑃𝑎𝑟𝑡 𝑜𝑓 𝑡ℎ𝑒 𝐿𝑜𝑟𝑒𝑛𝑡𝑧 𝑓𝑜𝑟𝑐𝑒 𝑝𝑒𝑟𝑝𝑒𝑛𝑑𝑖𝑐𝑢𝑙𝑎𝑟 𝑡𝑜 𝑡ℎ𝑒 𝑟𝑎𝑑𝑖𝑢𝑠 [N]
𝑇𝑙 = 𝑇𝑜𝑟𝑞𝑢𝑒 𝑝𝑟𝑜𝑑𝑢𝑐𝑒𝑑 𝑏𝑦 2 𝑤𝑖𝑟𝑒 𝑠𝑒𝑐𝑡𝑖𝑜𝑛𝑠 𝑓𝑟𝑜𝑚 1 𝑙𝑜𝑜𝑝 [Nm]
𝑇𝑐𝑜𝑖𝑙 = 𝑇𝑜𝑡𝑎𝑙 𝑡𝑜𝑟𝑞𝑢𝑒 𝑝𝑒𝑟 𝑐𝑜𝑖𝑙 [Nm]
𝑁𝑐 = 𝑇𝑜𝑡𝑎𝑙 𝑎𝑚𝑜𝑢𝑛𝑡 𝑜𝑓 𝑝𝑜𝑠𝑠𝑖𝑏𝑙𝑒 𝑡𝑢𝑟𝑛𝑠 𝑝𝑒𝑟 𝑐𝑜𝑖𝑙
𝑁𝑐1 = 𝑇𝑜𝑡𝑎𝑙 𝑎𝑚𝑜𝑢𝑛𝑡 𝑜𝑓 𝑝𝑢𝑟𝑒𝑙𝑦 𝑡𝑟𝑎𝑝𝑒𝑧𝑜𝑖𝑑𝑎𝑙 𝑡𝑢𝑟𝑛𝑠
𝑁𝑐2 = 𝑇𝑜𝑡𝑎𝑙 𝑎𝑚𝑜𝑢𝑛𝑡 𝑜𝑓 𝑝𝑢𝑟𝑒𝑙𝑦 𝑡𝑟𝑖𝑎𝑛𝑔𝑢𝑙𝑎𝑟 𝑡𝑢𝑟𝑛𝑠

Electromagnetic design

𝜆 = 𝑃𝑟𝑜𝑑𝑢𝑐𝑒𝑑 𝑏𝑎𝑐𝑘 𝐸𝑀𝐹 𝑝𝑒𝑟 𝑤𝑒𝑖𝑔ℎ𝑡 𝑜𝑓 𝑚𝑎𝑔𝑛𝑒𝑡 [V⁄kg]

𝑑𝑤 = 𝑊𝑖𝑛𝑑𝑖𝑛𝑔 𝑟𝑎𝑑𝑖𝑖 𝑐ℎ𝑎𝑛𝑔𝑖𝑛𝑔 𝑓𝑎𝑐𝑡𝑜𝑟 [m]
𝐼𝑝ℎ𝑎𝑠𝑒 = 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝑝𝑒𝑟 𝑝ℎ𝑎𝑠𝑒 [A]
𝑅𝑡𝑜𝑡,𝑠𝑒𝑟𝑖𝑒𝑠 = 𝑇𝑜𝑡𝑎𝑙 𝑠𝑒𝑟𝑖𝑒𝑠 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 𝑜𝑓 𝑎 𝑠𝑖𝑛𝑔𝑙𝑒 𝑝ℎ𝑎𝑠𝑒 [Ω]
𝐹𝑚 = 𝑀𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑜𝑟𝑐𝑒 𝑎𝑐𝑡𝑖𝑛𝑔 𝑜𝑛 𝑎 𝑠𝑢𝑟𝑓𝑎𝑐𝑒 [N]
𝐴 = 𝑆𝑢𝑟𝑓𝑎𝑐𝑒 𝐴𝑟𝑒𝑎 [m2 ]
𝑟𝑑 = 𝑅𝑎𝑑𝑖𝑢𝑠 𝑜𝑓 𝑟𝑎𝑑𝑖𝑎𝑙 𝑠𝑒𝑐𝑡𝑖𝑜𝑛 𝑑 𝑓𝑜𝑟 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑜𝑟𝑐𝑒 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑖𝑜𝑛 [m]
𝑆𝑠𝑙 = 𝑆𝑢𝑟𝑓𝑎𝑐𝑒 𝑎𝑟𝑒𝑎 𝑜𝑓 𝑟𝑎𝑑𝑖𝑎𝑙 𝑠𝑙𝑖𝑐𝑒 [m2 ]
𝐹𝑑 = 𝑇𝑜𝑡𝑎𝑙 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑜𝑟𝑐𝑒 𝑒𝑥𝑒𝑟𝑡𝑒𝑑 𝑜𝑛 𝑠𝑙𝑖𝑐𝑒 𝑑 [N]
̅̅
𝐵̅̅
𝑑 = 𝐴𝑣𝑒𝑟𝑎𝑔𝑒 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑖𝑒𝑙𝑑 𝑓𝑜𝑟 𝑠𝑙𝑖𝑐𝑒 𝑑 [T]
𝐹𝑝 = 𝑇𝑜𝑡𝑎𝑙 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑜𝑟𝑐𝑒 𝑒𝑥𝑒𝑟𝑡𝑒𝑑 𝑜𝑛 1 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑝𝑜𝑙𝑒 [N]
𝐹𝑎𝑡𝑡 = 𝑇𝑜𝑡𝑎𝑙 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑎𝑡𝑡𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑓𝑜𝑟𝑐𝑒 𝑒𝑥𝑒𝑟𝑡𝑒𝑐 𝑜𝑛 𝑡ℎ𝑒 𝑟𝑜𝑡𝑜𝑟 [N]
𝛽𝑃𝑀,𝑑 = 𝐴𝑛𝑔𝑙𝑒 𝑓𝑜𝑟 𝑎𝑟𝑐 𝑙𝑒𝑛𝑔𝑡ℎ 𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑜𝑓 𝑠𝑙𝑖𝑐𝑒 𝑑 [ °]

Mechanical design

𝑦𝑏 = 𝑇𝑜𝑡𝑎𝑙 𝑑𝑒𝑓𝑙𝑒𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑎𝑛𝑛𝑢𝑙𝑎𝑟 𝑝𝑙𝑎𝑡𝑒 𝑎𝑡 𝑖𝑛𝑛𝑒𝑟 𝑟𝑎𝑑𝑖𝑢𝑠 [m]

𝑊 = 𝑈𝑛𝑖𝑡 𝑙𝑖𝑛𝑒 𝑙𝑜𝑎𝑑 [N⁄m]
𝐷 = 𝑃𝑙𝑎𝑡𝑒 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡
𝐸 = 𝑌𝑜𝑢𝑛𝑔′ 𝑠 𝑚𝑜𝑑𝑢𝑙𝑢𝑠 [GPa]
𝜈 = 𝑃𝑜𝑖𝑠𝑠𝑜𝑖𝑛𝑠 𝑟𝑎𝑡𝑖𝑜
𝑎 = 𝑂𝑢𝑡𝑒𝑟 𝑟𝑎𝑑𝑖𝑢𝑠 𝑜𝑓 𝑡ℎ𝑒 𝑎𝑛𝑛𝑢𝑙𝑎𝑟 𝑟𝑖𝑛𝑔 [m]
𝑏 = 𝐼𝑛𝑛𝑒𝑟 𝑟𝑎𝑑𝑖𝑢𝑠 𝑜𝑓 𝑡ℎ𝑒 𝑎𝑛𝑛𝑢𝑙𝑎𝑟 𝑜𝑝𝑒𝑛𝑖𝑛𝑔 [m]
𝜑𝑀𝐼𝑁,𝑓,𝑑 = 𝐼𝑛𝑛𝑒𝑟 𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑖𝑜𝑛 𝑙𝑖𝑚𝑖𝑡 𝑓𝑜𝑟 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑎𝑡𝑡𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑓𝑜𝑟𝑐𝑒 [rad]
𝜑𝑀𝐴𝑋,𝑓,𝑑 = 𝑈𝑝𝑝𝑒𝑟 𝑖𝑛𝑡𝑒𝑔𝑟𝑎𝑡𝑖𝑜𝑛 𝑙𝑖𝑚𝑖𝑡 𝑓𝑜𝑟 𝑚𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑎𝑡𝑡𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑓𝑜𝑟𝑐𝑒 [rad]

10
 Thermal design

𝑙 = 𝐿𝑎𝑦𝑒𝑟 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 [m]

𝑘 = 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑐𝑜𝑛𝑑𝑢𝑐𝑡𝑖𝑣𝑖𝑡𝑦 [W⁄mK]
𝐴 = 𝐶𝑜𝑛𝑡𝑎𝑐𝑡 𝑎𝑟𝑒𝑎 [m²]
ℎ = 𝐻𝑒𝑎𝑡 𝑡𝑟𝑎𝑛𝑠𝑓𝑒𝑟 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡[W⁄mK]
𝑈 = 𝑆𝑝𝑒𝑒𝑑 𝑜𝑓 𝑡ℎ𝑒 𝑎𝑖𝑟 [m⁄s]
𝜈 = 𝐾𝑖𝑛𝑒𝑚𝑎𝑡𝑖𝑐 𝑣𝑖𝑠𝑐𝑜𝑠𝑖𝑡𝑦 𝑜𝑓 𝑎𝑖𝑟 [m⁄ ]
s²
𝐷 = 𝐷𝑖𝑎𝑚𝑒𝑡𝑒𝑟 𝑜𝑓 𝑒𝑎𝑐ℎ 𝑙𝑎𝑦𝑒𝑟 [m]
𝑁𝑟𝑝𝑚 = 𝑅𝑜𝑡𝑎𝑡𝑖𝑜𝑛 𝑠𝑝𝑒𝑒𝑑 𝑜𝑓 𝑡ℎ𝑒 𝑚𝑜𝑡𝑜𝑟 [rpm]
𝑅 = 𝑇ℎ𝑒𝑟𝑚𝑎𝑙 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 [K⁄W]
𝑄̇ = 𝐻𝑒𝑎𝑡 𝑝𝑟𝑜𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑜𝑓 𝑡ℎ𝑒 𝑃𝐶𝐵 [W]
𝑇𝑖𝑛 = 𝐼𝑛𝑛𝑒𝑟 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑚𝑜𝑡𝑜𝑟 [°C]
𝑇𝑜𝑢𝑡 = 𝑇𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 𝑜𝑓 𝑡ℎ𝑒 𝑠𝑢𝑟𝑟𝑜𝑢𝑛𝑑𝑖𝑛𝑔𝑠 [°C]

Motor testing

𝐾𝑣 = 𝑉𝑒𝑙𝑜𝑐𝑖𝑡𝑦 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 [rad⁄V s]

𝐾𝑒 = 𝐵𝑎𝑐𝑘 𝐸𝑀𝐹 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡 [V s⁄rad]
𝑃𝑖𝑛,𝐸𝑆𝐶 = 𝐸𝑆𝐶 𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑎𝑙 𝑖𝑛𝑝𝑢𝑡 𝑝𝑜𝑤𝑒𝑟 [W]
𝑃𝑀 = 𝑀𝑒𝑐ℎ𝑎𝑛𝑖𝑐𝑎𝑙 𝑚𝑜𝑡𝑜𝑟 𝑜𝑢𝑡𝑝𝑢𝑡 𝑝𝑜𝑤𝑒𝑟 [W]
𝜂𝑠 = 𝑇𝑒𝑠𝑡 𝑠𝑒𝑡𝑢𝑝 𝑠𝑦𝑠𝑡𝑒𝑚 𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 [%]
𝜂𝐸𝑆𝐶 = 𝐸𝑆𝐶 𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 [%]
𝜂𝑚 = 𝑀𝑜𝑡𝑜𝑟 𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑐𝑦 [%]
𝑃𝐸𝑆𝐶,𝑙𝑜𝑠𝑠 = 𝐸𝑆𝐶 𝑝𝑜𝑤𝑒𝑟 𝑙𝑜𝑠𝑠𝑒𝑠 [W]
𝑃𝑏𝑒𝑎𝑟𝑖𝑛𝑔 = 𝑇𝑒𝑠𝑡 𝑠𝑒𝑡𝑢𝑝 𝑏𝑒𝑎𝑟𝑖𝑛𝑔 𝑝𝑜𝑤𝑒𝑟 𝑙𝑜𝑠𝑠𝑒𝑠 [W]
𝑃𝑖𝑛,𝑚𝑜𝑡𝑜𝑟 = 𝑇𝑟𝑢𝑒 𝑒𝑙𝑒𝑐𝑡𝑟𝑖𝑐𝑎𝑙 𝑖𝑛𝑝𝑢𝑡 𝑝𝑜𝑤𝑒𝑟 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑚𝑜𝑡𝑜𝑟 [W]

Bearing friction losses

𝑀𝑟𝑟 = 𝑅𝑜𝑙𝑙𝑖𝑛𝑔 𝑓𝑟𝑖𝑐𝑡𝑖𝑜𝑛𝑎𝑙 𝑚𝑜𝑚𝑒𝑛𝑡 [Nmm]

𝑀𝑠𝑙 = 𝑆𝑙𝑖𝑑𝑖𝑛𝑔 𝑓𝑟𝑖𝑐𝑡𝑖𝑜𝑛𝑎𝑙 𝑚𝑜𝑚𝑒𝑛𝑡 [Nmm]
𝑀𝑠𝑒𝑎𝑙 = 𝐹𝑟𝑖𝑐𝑡𝑖𝑜𝑛𝑎𝑙 𝑚𝑜𝑚𝑒𝑛𝑡 𝑑𝑢𝑒 𝑡𝑜 𝑠𝑒𝑎𝑙𝑠 [Nmm]
𝑀𝑑𝑟𝑎𝑔 = 𝐹𝑟𝑖𝑐𝑡𝑖𝑜𝑛𝑎𝑙 𝑚𝑜𝑚𝑒𝑛𝑡 𝑑𝑢𝑒 𝑡𝑜 𝑜𝑖𝑙 𝑏𝑎𝑡ℎ [Nmm]
Φ𝑖𝑠ℎ = 𝑆ℎ𝑒𝑎𝑟 ℎ𝑒𝑎𝑡𝑖𝑛𝑔 𝑟𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟
Φ𝑟𝑠 = 𝐾𝑖𝑛𝑒𝑚𝑎𝑡𝑖𝑐 𝑟𝑒𝑝𝑙𝑒𝑛𝑖𝑠ℎ𝑚𝑒𝑛𝑡⁄𝑠𝑡𝑎𝑟𝑣𝑎𝑡𝑖𝑜𝑛 𝑟𝑒𝑑𝑢𝑐𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟
𝐺𝑟𝑟 = 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒 𝑑𝑒𝑝𝑒𝑛𝑑𝑖𝑛𝑔 𝑜𝑛 𝑡ℎ𝑒 𝑏𝑒𝑎𝑟𝑖𝑛𝑔 𝑡𝑦𝑝𝑒 𝑎𝑛𝑑 𝑙𝑜𝑎𝑑
𝜈𝑏 = 𝐴𝑐𝑡𝑢𝑎𝑙 𝑜𝑝𝑒𝑟𝑎𝑡𝑖𝑛𝑔 𝑣𝑖𝑠𝑐𝑜𝑠𝑖𝑡𝑦 𝑜𝑓 𝑡ℎ𝑒 𝑏𝑎𝑠𝑒 𝑜𝑖𝑙 𝑜𝑓 𝑡ℎ𝑒 𝑔𝑟𝑒𝑎𝑠𝑒 [mm2⁄s]
𝑑 = 𝐵𝑒𝑎𝑟𝑖𝑛𝑔 𝑏𝑜𝑟𝑒 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 [mm]
𝐷 = 𝐵𝑒𝑎𝑟𝑖𝑛𝑔 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 [mm]
𝑑𝑚 = 𝐵𝑒𝑎𝑟𝑖𝑛𝑔 𝑚𝑒𝑎𝑛 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 [mm]
𝐾𝑟𝑠 = 𝑅𝑒𝑝𝑙𝑒𝑛𝑖𝑠ℎ𝑚𝑒𝑛𝑡/𝑠𝑡𝑎𝑟𝑣𝑎𝑡𝑖𝑜𝑛 𝑓𝑎𝑐𝑡𝑜𝑟
𝐾𝑧 = 𝐵𝑒𝑎𝑟𝑖𝑛𝑔 𝑡𝑦𝑝𝑒 𝑟𝑒𝑙𝑎𝑡𝑒𝑑 𝑔𝑒𝑜𝑚𝑒𝑡𝑟𝑖𝑐 𝑐𝑜𝑛𝑠𝑡𝑎𝑛𝑡
𝐺𝑠𝑙 = 𝑉𝑎𝑟𝑖𝑎𝑏𝑙𝑒 𝑑𝑒𝑝𝑒𝑛𝑑𝑖𝑛𝑔 𝑜𝑛 𝑡ℎ𝑒 𝑏𝑒𝑎𝑟𝑖𝑛𝑔 𝑡𝑦𝑝𝑒 𝑎𝑛𝑑 𝑙𝑜𝑎𝑑
µ𝑠𝑙 = 𝑆𝑙𝑖𝑑𝑖𝑛𝑔 𝑓𝑟𝑖𝑐𝑡𝑖𝑜𝑛 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡
µ𝑏𝑙 = 𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡 𝑑𝑒𝑝𝑒𝑛𝑑𝑖𝑛𝑔 𝑜𝑛 𝑡ℎ𝑒 𝑡𝑦𝑝𝑒 𝑜𝑓 𝑚𝑜𝑣𝑒𝑚𝑒𝑛𝑡
µ𝐸𝐻𝐿 = 𝑆𝑙𝑖𝑑𝑖𝑛𝑔 𝑓𝑟𝑖𝑐𝑡𝑖𝑜𝑛 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 𝑖𝑛 𝑓𝑢𝑙𝑙 − 𝑓𝑖𝑙𝑚 𝑐𝑜𝑛𝑑𝑖𝑡𝑖𝑜𝑛𝑠
Φ𝑏𝑙 = 𝑊𝑒𝑖𝑔ℎ𝑖𝑛𝑔 𝑓𝑎𝑐𝑡𝑜𝑟 𝑓𝑜𝑟 𝑡ℎ𝑒 𝑠𝑙𝑖𝑑𝑖𝑛𝑔 𝑓𝑟𝑖𝑐𝑡𝑖𝑜𝑛 𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡

11
 1 Introduction
1.1 Importance of Research

The world population has a growing need of electricity. Not only is the world population a growing
factor, there are also lots of new technologies that use electricity. It is important to use every unit of
energy as efficiently as possible. On top of these factors, global warming has reached a critical level
[1]. This issue demands inventive solutions to use electricity in an as efficient and climate neutral way
as possible. One way of reaching this goal is redefining our means of transport. A good step in that
direction is the growing interest and sales of electric vehicles (EV). In 2019 there were over 2,2
million electric vehicles sold which brings the total number of EV’s in use to 7,2 million according to a
study conducted by Vitra [2].

Whilst the growing population is impactful for the energy market, it also results in the shortage of
metals and other resources. Therefore, it is important to make the new technologies with as little
material as possible. There are many ways to reduce the consumption of resources. A possibility for
ensuring that nothing gets wasted can be attained by using 3D-printing. With 3D printing there are
parameters that can be set to use as little as possible filament for making the desired structure.
Therefore, 3D-printing is a good step to reducing wastage of resources. Our research’s ambition is to
combine a sustainable and economic use of material and energy.

This research thesis focusses on an axial flux permanent magnet motor with printed circuit board
stator windings. This topology is suitable for applications that require light weight and compact
motors. Thanks to these characteristics, these motors are suitable for use in a variety of applications.
It is possible to place these kinds of motor in a modular structure. Because of this, it is possible to
create the torque and power according to the required levels. The shape of this motor is ideal for
installing them as in-wheel motors of an electric vehicle. Given all these considerations, this research
is therefore impactful with regards to global warming rehabilitation and reducing the use of
resources.

1.2 Goal

The primary goal of this thesis is to develop a fully functioning and working axial flux permanent
magnet motor with printed circuit board stator windings. The primary objective is obtained by using
the following secondary goals:

• First, an analytical model will be created to calculate the possibilities of this new way of axial
flux motor design. Testing this model’s parameters will reveal the best way to build the
prototype.
• Second, the parameters will be defined and calculated within the analytical model. The
motor will be mechanically designed together with the PCB.
• The final steps will be the assembly of the motor and testing by comparing the results with
the analytical model. This comparison will clarify whether the model is correct or not.

12
1.3 Outline

The structure of this thesis encloses a few major parts. In the first part of this thesis the working of an
axial flux motor will be described and explained. This is important to comprehend the
electromagnetic analytical model and used formulas.

The second section will consider the electromagnetic analytical model. The fundamental formulas
used in the model will be explained. The code behind this model and the interpretation of the results
will be discussed and verified. The electromagnetic analytical model is one of the most interesting
aspects of the thesis, this section is essential for further employment of the model.

Thirdly, the electromagnetic design is determined by using the analytical model. First, the
configuration of the magnets is determined. The second part of the motor that will be defined is the
winding configuration. Defining the windings consists of the following characteristics: the number of
turns per winding, width of the tracks and inner and outer radius of the tracks. The final part that has
an influence on the electromagnetic design is the backplate. The thickness and radius of the
backplate will be defined. These three parts define the electromagnetic design that will be used to
prototype the motor.

The next chapter and the fourth part of the thesis is the mechanical design. The mechanical design is
done with the goal of making the assembly process easier. Next, the used motor controller will
briefly be discussed. Not much attention goes into this part because it is not of much meaning for
this thesis.

To estimate the inner motor temperature, a simplified thermal analysis will be conducted. This
estimation can be used to look for extra cooling solutions or if extra ventilation is needed. A good
thermal control of the motor is needed to avoid dangerous situations.

If the aforementioned design goals are reached and the individual parts are manufactured, the
assembling process can be started and will be explained in chapter 8 step by step. This part is
important for reconstruction of the prototype by others. Afterwards, the testing setup and
procedure are explained. First, the back-EMF tests are carried out and the results are discussed. The
second part of the testing is the torque and efficiency test.

The next-to-last chapter of this thesis is the conclusion. In this part the results will be reflected and
explained. This section gives a clear image of the possibilities of the performances of an axial flux
motor with permanent magnets and windings placed on a PCB.

To finalize the thesis, possible future work is discussed and further needed research is defined.

13
 2 Literature Review
The literature review consists of four sections. The first section will give insights into the working
principles of an axial-flux machine. The second section explains the different possibilities of windings
on a printed circuit board. To understand the possibilities, the characteristics and production process
of a PCB is explained. Thirdly, the use of finite element method in this thesis is explained. After these
insights that are needed to comprehend the thesis, a few possibilities of applications are stated.

2.1 Axial-flux Machine

Axial flux machines are already very commonly used in different kinds of applications. Due to its
small axial length and high electrical performance, it can easily be used in electrical vehicles. To
understand the analytical design, it is necessary to know how axial flux machines with permanent
magnets work.

2.1.1 Structure

The basic structure of an axial flux motor is shown in Figure 2.1. This structure has three main parts:

• Two stators with copper windings

• One rotor with permanent magnets

The airgap should be as small as possible in order to keep the reluctance low and thus obtain the
strongest magnetic flux possible. The ideal airgap would be as close as possible to zero. Because of
mechanical limitations like vibrations and stator thickness this is not possible since the rotor and
stator could touch causing short-circuits.

2
1 1

Figure 2.1: Basic structure of a single rotor axial flux motor with PM; Stator (1); rotor with permanent magnets (2). Adapted
from [3]

Figure 2.1 shows that these axial flux machines have a small axial length, this makes them easy to use
in a various number of applications. A second advantage of this structure is that they are very light,
and therefore the power density is very high.

14
In this thesis a similar structure will be used. Firstly, the magnets will be on the outside and the stator
will be the middle of the structure, this is shown in Figure 2.2.

Figure 2.2: Structure used for this thesis. Adapted from [3]

Secondly, the stator with its copper winding that weigh a lot will be replaced by a PCB with windings
etched on top of it. This will cause a further weight reduction. In the next part the working principle
will be explained.

2.1.2 Working principle

The permanent magnets induce a magnetic field that is perpendicular to the airgap. This way the
magnetic field makes a bridge between the stator and rotor. The result of this magnetic field is an
induced magnetic flux in the stator. As shown in Figure 2.3, the magnets next to each other are
alternated magnetized which causes an alternating magnetic flux in the stator. The reason why the
two rotors contain a backplate at the outside is because they will establish a low reluctance return
path for the fluxes. The flux path is shown in Figure 2.3.

To explain the operating principle, the motor is watched as a generating entity. The rotation of the
rotor due to the external force will lead to an alternating flux through the stator windings. This
alternating flux will induce a back electromotive force (EMF) in these windings due to Faraday’s law
of induction. When the windings are connected to a load, this will lead to a current flow through the
stator windings.

The motor could also be approached as an entity providing mechanical power. In that case the stator
windings must be supplied with three phase current flow. The imposed current will induce a
magnetic field in the windings, this magnetic field will interact with the magnetic field induced by the
permanent magnets. The magnetic field induced by the current alternates in time due to the
different phases. The interaction results in rotation of the rotors, this way the electrical power is
transferred into mechanical power [4].

15
 Φ 𝑀𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝑓𝑙𝑢𝑥

𝑠 𝑛 𝑠 𝑛 𝑠
𝑛 𝑠 𝑛 𝑠 𝑛

𝑠 𝑛 𝑠 𝑛 𝑠
𝑛 𝑠 𝑛 𝑠 𝑛

Figure 2.3: Magnetic field induced by permanent magnets

16
2.2 PCB Winding Design

There are different winding shapes that can be placed on a PCB. They all have different advantages
and disadvantages. When placing windings on a PCB, there are boundaries that need to be followed.
These boundaries are a result of the machine printing the PCBs. In general, there are two kinds of
windings. First, there are the turns that are placed concentric to each other, therefore, this type of
winding is called a concentric winding. The second kind of winding is called the wave winding, here,
the turns are not placed concentric to each other. To understand the limitations of the PCBs a short
introduction is given hereafter.

2.2.1 Manufacturing process of a PCB

Initially printed circuit boards were developed to aid in electronic designs. These boards can contain
the paths between components that otherwise would have to be a physical wire. Copper sheets are
places on a nonconductive substrate. Afterwards the pathways are etched from the copper sheets.
Nowadays, it is impossible to imagine a society without printed circuit boards. The making of a PCB
consists of 6 steps [5]:

1. Layout design, this step is carried out in CAD software

2. Negative film development for the circuit
3. Copper clad preparation
4. UV-exposure of copper clad and its developing
5. Drilling
6. Etching

The etching defines the boundaries to PCB use and its pathways placed on it. Therefore, the
following can be concluded. The paths must have a minimum distance between the tracks because of
the etching. The etching needs some space to etch out the pathways. The thicker the copper gets the
more etching space is needed because etching will cause chamfered sides.

Hereafter the different kind of ways to place windings on the PCBs are explained.

2.2.2 Concentric Windings

Concentric windings are windings that are placed concentric to each other. A way to recognize
concentric windings is that the start and end of the winding are placed on the same side of the PCB.
These windings can have different kinds of shapes but in general have the same disadvantages and
advantages.

The advantages of using concentric windings are:

• High induced voltage

• High induced torque
• High flux linkage area

The determining factor within the PCB windings for the induced voltage is the flux linkage area and
for torque this is the radial length of the pathways. The turns of the concentric windings are placed in
that way that most parts of the pathways are radial.

17
Nevertheless, there are some disadvantages to concentric windings also:

• The radial part of the winding is the only part of the copper traces that induce torque.
• Harmonic levels of the induced voltages are high.

There are many concentric winding configurations that can be used. The two most common are the
rhomboidal windings and the trapezoidal windings. The trapezoidal windings are, as the name
indicates, trapeziums that are placed concentric to each other. Rhomboidal windings are concentric
placed diamond shaped windings. Both kind of windings are displayed in Figure 2.4.

For further analysis, the two configurations are compared with their area of utilization. Figure 2.4
shows that rhomboidal shaped windings have a much lower Area of Utilization coefficient. These
findings are publicised in the research paper of Togöz et al. This research uses the next coefficient
formula to calculate the Area of utilization [6].
(𝑤𝑡𝑟 . 𝑙 𝑡𝑜𝑡𝑎𝑙 )
𝐴𝑟𝑒𝑎 𝑈𝑡𝑖𝑙𝑖𝑧𝑎𝑡𝑖𝑜𝑛 = (2.1)
𝜋(𝑟𝑠𝑖 + 𝑟𝑠𝑜 )

Where 𝑤𝑡𝑟 is the width of the tracks, 𝑙𝑡𝑜𝑡𝑎𝑙 is the total length of the winding tracks and 𝑟𝑠𝑖 and 𝑟𝑠𝑜 are
the radii of the inner- and outermost tracks on the PCB, respectively.

Figure 2.4: Area of Utilization comparison; (a) Trapezoidal winding; (b) Rhomboidal winding

The second difference between these two configurations is the amount of torque produced by the
two. Due to the high fill factor and large radial conductor sections of the trapezoidal configuration,
the induced torque will be higher. In rhomboidal windings, every part of the path induces a torque.
But because the windings are not perpendicular to the radius, not al the induced forces will
contribute to useful torque production. The trapezoidal windings produce, contrary to rhomboidal
windings, only torque in the radial part of the paths.
This short analysis brings to the conclusion that the most efficient kind of windings within concentric
windings are trapezoidal. Therefore, this configuration will be used in the first prototype of this thesis
research.

18
2.2.3 Wave Windings

Wave windings are characterised by the configuration of their windings. The start and end of the
winding are not on the same side of the PCB. Just as concentric windings, wave windings also have
lots of variations. Most common are parallel windings, arc windings and unequal width parallel
windings.

Figure 2.5: Different kinds of wave windings; (a) Parallel windings, (b) Arc windings and (c) Unequal width parallel windings.
Adapted from [6]

These wave windings are more complex than the concentric windings. It is possible to reach more
filling factor using this method. In the next section of this paragraph the advantages and
disadvantages will be listed.

Parallel windings

The first winding that will be discussed and explained is the parallel winding. The advantages of this
kind of winding are that they have a short path which implies low resistance. Because of these radial
placed paths, the filling factor will be higher than using concentric windings. Nevertheless, these
advantages cause also disadvantages. Because of the short paths, the flux linkage area is lower than
with concentric winding. A second disadvantage is that the paths are not exactly radial, therefore,
the output torque must be multiplied by a factor cos(α) to take into account the angular deviation. In
this factor, 𝛼 is the angle between the radial direction and the direction of the paths [6].

Arc windings

The advantages of these windings are increased copper path lengths and their bigger flux linkage
area. Because of these, this kind of winding will induce more torque [6].

Unequal Width Parallel Winding

The name of this kind of windings explains the layout design. At the inner circle of the PCB, the traces
are smaller than the traces at the outer radius of the PCB. So, the width of the traces increases when
the radius increases. This method will require more copper and therefore will cause more induced
torque [6].

19
2.3 Electromagnetic Finite Element Analysis

Finite element analysis (FEA) is commonly used in the industry and academies. Programs that
implement this method are often easier to comprehend than analytical modelling because of the
graphic approach. Finite element analysis is the simulation of a finite element method model, this is a
mathematical technique. Nevertheless, it can be used in a variety of disciplines due to its great
calculation method.

The finite element analysis divides the structure into a finite number of elements. Each element gets
its own preconditions and is connected to the other elements with nodes. All elements together form
a grid.

The elements used in this method can be one-dimensional, 2D or 3D, according to its dimensions
each element gets its own parameters. When these are defined, a second choice needs to be made
for each element. In a finite element analysis, there are two kinds of elements:

• H-elements: these elements have fixed shapes.

• P-elements: these elements have a variable polynomial and more complex shape.

The H-elements are the most used because they are faster to calculate, by using these elements
more complex structures can be modelled without needing much time to calculate.

Nowadays there al lots of software that apply this method. In this thesis FEMM [7] is used. FEMM
stands for finite element method magnetics. The advantage of this software is that it is very graphic
and therefore easier to interpret [8].

2.3.1 FEMM

FEMM is, as written here above, a finite element method software specialized in magnetic, linear
electrostatic and steady-state heat problems. This software is suited for solving low frequency
electromagnetic problems on 2D planes. The FEMM software consists of three major sections [8]:

• The interactive shell: Firstly, this is the interface to interact with the user to model the
structure that needs to be analysed. Secondly, this section is also used to show figures,
graphs and other results that came out of the analysation of the structure.
• Triangle executable: This section of the software divides the structure into little triangles
with nodes. These triangles are needed to use the finite element method.
• Solver: This section solves the given structure divided into little triangles. The result of these
solvers are matrixes, these send back to the interactive shell to display the results of the
structure.

For this thesis FEMM calculations are used to compare the results of the programmed model in
MATLAB. The different stadiums of FEMM software analysis are shown in the figures below, these
figures are used to show the different stages but is not the actual motor. Therefore, the used
variables are not of any use and will not be discussed in this paragraph. The first step in creating a
FEMM analysis is creating the structure. In Figure 2.6 the structure is displayed. The structure
consists of two halves that are the mirror image of each other. The upper half consists of the
following, seen from top to bottom:

20
 a) This rectangle represents the air arround the motor.
b) This rectangle is a representative for the back plates.
c) This points to a line of rectangles with arrows within, these represent the magnets with their
magnetic polarisations.
d) This part consits of a rectangle with different rectangles ontop, these are not visible on the
figure. These two are representing the PCB and its windings.

In Figure 2.6 there are also green dots visible within the rectangles, these dots represent the material
characteristics given to the rectangles. These characteristics can contain a various amount of
information to define the rectangles.

a
b
c
d

Figure 2.6: Topology of the motor structure in FEMM

Figure 2.7 is the result when the structure is run through the triangle executable of the FEMM
software. This section divides the structure into triangles. In this step, the structure is checked for
holes. Holes can occur when the structure is not closed or has ambiguous characteristics. When this
occurs, the compiler will not be able to calculate the structure.

Figure 2.7: Structure divided into triangles by the triangle executable of FEMM.

After implementing the two previous steps correctly, the software moves on to the next step. In this
step the results are provided by a schematic drawing (plotted) and displayed for the user. The result
of the motor structure is displayed in Figure 2.8.

21
 Figure 2.8: Results calculated and plotted by FEMM (Inductance B in tesla [T]).

As announced in the beginning of this chapter, in this thesis the FEMM analysis will not be further
discussed and will only be used to compare the analytical results with.

22
2.4 Application possibilities

As described in the first paragraphs of the literature review an axial flux permanent magnet motor
with windings printed on a printed circuit board has small axial dimensions. Due to these dimensions
the motor has various number of possibilities. In this paragraph three groups of possibilities will be
discussed. The first group are the electro-mobility applications. The second group are the electrical
devices used in households. The last group that will be briefly discussed are the automation
applications.

2.4.1 Electro-mobility

In-wheel motors

In-wheel motors are as the name indicates motors that are placed within the wheels. Lots of car
manufacturers are looking for motors that can be placed inside the wheel because it has lots of
advantages. Two of the biggest advantages are:

• Transmission components are not longer needed which results in a weight loss.
• The transmission is directly located at the wheels this cause a faster reaction time.

E-bikes

E-bikes and S-bikes are very popular nowadays. E-bikes are now an important element in the modal
shift for personal transport for commuters. It is often seen as a good alternative for car use for
commuters and therefore companies offer E-bike leasing for their personnel as this is a good way of
sustainable mobility. The motor in these bicycles is very small, but also the battery capacity is limited.
Therefore, it would be useful to integrate energy efficient and lightweight motors just like the AFPM
motors.

Electric steps

Electric steps are just like E-bikes a fairly new way of mobility that is popular nowadays. These steps
are used for small distances and can be easily taken on the train for example. This way, these little
vehicles are part of the multi-modal way of mobility for commuters also.

Drones

Some say that drones are the future for mail delivery and take away. The motor studied in this thesis
can be easily further downscaled and used in drones. Drones are easy to fly with and can be used for
lots of applications. These drones could be programmed to fly automatically, this would be great for
fast deliveries.

23
2.4.2 Electric devices

There are lots of electric devices to ease the work that sustaining a household demands. A great deal
of these devices requires motors. A characteristic to create good householding equipment is the
price in total. The motor studied in this thesis is noticeably cheap. This characteristic makes it ideal to
use in this group of devices. An example of these devices is a vacuum cleaner.

Vacuum cleaner

Vacuum cleaners are commonly used in almost every household. This would be a great sales market
to enter a new kind of motor in. Just like in-wheel motors, vacuum cleaners request small motors
with high rpm. A second requirement for vacuum cleaners could be weight, especially upright
vacuum cleaners where the motor weight must be carried by the person using it.

2.4.3 Automation

Automation is a growing industry nowadays. Lots of processes are preferred to be processed by

machine infrastructure without the interventions of humans. An example of this is an automated
warehouse.

Automated warehouse

Automated warehouses are obtained by a communication network of little transport vehicles that
are automatically driven. These little vehicles require small, light weighted motors that can easily be
controlled. AFPM motors are an ideal solution for this application.

24
 3 Electromagnetic analytical model
To design a motor that is as optimal as possible, an analytical model is made in MATLAB [9]. This
model will be able to predict certain characteristics with a certain amount of accuracy. This is an
electromagnetic model that will calculate various parameters and graphs. The most important
parameters will be the generated back EMF and torque production. Furthermore, the electrical
resistance of the phases and maximum possible number of turns per stator winding will be modelled.

3.1 Back EMF

3.1.1 Magnetic flux density

The first step in modelling the motor is to calculate the magnetic flux density that is present in the
airgap. For this calculation it is assumed that the back plate has an infinite magnetic permeability.
The magnetic flux density field is not sinusoidal. Therefore, it is necessary to also include the
harmonic components.

The rotor configuration can be seen in Figure 3.1.

𝑧2
𝑡𝐵𝑝
𝜃𝑚 ℎ𝑃𝑀

𝑧
𝜏𝑝
𝜏𝑚

𝜑
𝑟

Figure 3.1: Rotor Magnet Configuration with calculation variables

25
This can be done by using the following formula [4].
∞ 𝑛𝑝
4𝐵𝑟𝑒𝑚 𝜏𝑀 sinh ( 𝑟 ℎ𝑃𝑀 ) 𝑛𝑝
𝐵 = µ0 ∑ sin (𝑛𝜋 )∗ 𝑛𝑝 ∗ cosh ( (𝑧2 − 𝑧)) ∗ cos(𝑛𝑝(𝜑 + 𝜃𝑚 )) (3.1)
𝑛𝜋 ∗ µ0 µ𝑃𝑀 2𝜏𝑃 sinh ( 𝑟 𝑧2 ) 𝑟
𝑛=1

Where 𝑛 is the number of harmonics that is considered, 𝑝 is the number of pole pairs, 𝐵𝑟𝑒𝑚 is the
magnet remanence, ℎ𝑃𝑀 is the height of the magnets, 𝑧2 is the distance from the backplate to the
middle line between the two rotor halves, 𝑧 is the distance from the back plate to the desired
calculation plane, 𝑟 is the radius of the polar coordinate system, 𝜑 is the angular position of the polar
coordinate system, µ0 is the permeability of vacuum, µ𝑃𝑀 is the relative permeability of the
permanent magnet and 𝜃𝑚 is the relative mechanical angular position of the rotor in reference to the
stator. Some of the used variables can also be seen in Figure 3.1.

𝜏𝑃 is the arc length of one magnet pole and is function of the radius 𝑟. This can be calculated using
the following formula
𝜋
𝜏𝑃 (𝑟) = 𝑟 ∗ (3.2)
𝑝
Furthermore, because rectangular magnets are used, the magnet angular width τ𝑚 is not a constant
but is a function of the radius 𝑟, this can be seen in Figure 3.1, the dimensions of the bar magnet can
be seen in Figure 3.2.

ℎ𝑃𝑀
𝑙𝑃𝑀

𝑤𝑃𝑀

Figure 3.2: Bar magnet dimensions

The magnet angular width at radius 𝑟 is determined by.

𝑤𝑃𝑀
𝜏𝑀 (𝑟) = 2 sin−1 ( )∗𝑟 (3.3)
2𝑟
The magnetic field density is plotted in Figure 3.3 for both the analytical and FEM method. This FEM
analysis was done using a very thick backplate of 10 mm with the material properties of 1010 steel.
The higher thickness makes sure almost no saturation takes place. This way, a fair comparison
between the FEM model and analytical model can be done. For the analytical model, this test was
done at a radius 𝑟 = 40 mm. Further properties and dimensions that were used for this analysis are
listed in Table 3.1.

26
Table 3.1: Parameters for flux density field comparison

FEM model Analytical model

𝑀𝑎𝑔𝑛𝑒𝑡 𝑤𝑖𝑑𝑡ℎ 𝑤𝑃𝑀 [mm] 18 18
𝑀𝑎𝑔𝑛𝑒𝑡 ℎ𝑒𝑖𝑔ℎ𝑡 ℎ𝑃𝑀 [mm] 5 5
𝑀𝑎𝑔𝑛𝑒𝑡 𝑙𝑒𝑛𝑔𝑡ℎ 𝑙𝑃𝑀 [mm] 10 10
𝑀𝑎𝑔𝑛𝑒𝑡 𝑔𝑟𝑎𝑑𝑒 𝑁45 𝑁45 (𝐵𝑟𝑒𝑚 = 1,38 T)
𝐵𝑎𝑐𝑘𝑝𝑙𝑎𝑡𝑒 𝑚𝑎𝑡𝑒𝑟𝑖𝑎𝑙 1010 𝑠𝑡𝑒𝑒𝑙 𝐼𝑛𝑓𝑖𝑛𝑖𝑡𝑒 𝑝𝑒𝑟𝑚𝑒𝑎𝑏𝑖𝑙𝑖𝑡𝑦
𝐵𝑎𝑐𝑘𝑝𝑙𝑎𝑡𝑒 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 [mm] 10 /
𝐴𝑖𝑟𝑔𝑎𝑝 [mm] 2,8 2,8

Figure 3.3: Comparison of analytical and FEM magnetic flux density field B

It is clear from Figure 3.3 that the analytical model generates very accurate results.

In the following calculations it will be necessary to calculate the average magnetic flux density, this
can be done by integrating formula (3.1) to the variable 𝜑. After integrating, the function is
∞ 𝑛𝑝
4𝐵𝑟𝑒𝑚 𝜏𝑀 sinh ( 𝑟 ℎ𝑃𝑀 ) 𝑛𝑝 sin(𝑛𝑝(𝜑 + 𝜃𝑚 ))
∫ 𝐵𝑑𝜑 = µ0 ∑ sin (𝑛𝜋 ) 𝑛𝑝 cosh ( (𝑧2 − 𝑧)) (3.4)
𝑛𝜋 ∗ µ0 µ𝑃𝑀 2𝜏𝑃 sinh ( 𝑧 ) 𝑟 𝑛𝑝
𝑛=1 𝑟 2

3.1.2 Flux linkage per pole

To calculate the generated back emf, the flux linkage per pole should be determined. In this model,
concentrated spiral trapezoidal windings are used. To speed up and simplify the calculations, these
windings are approximated by multiple closed concentric loops. The magnetic flux is then calculated
for every loop. Such loops are shown in Figure 3.4.

27
 𝑤𝑠

𝑤𝑡𝑟
𝑟𝑜

𝑡𝑠𝑙

𝑟𝑖

𝜑𝑀𝐼𝑁,𝑙,𝑞
𝜑𝑀𝐴𝑋,𝑙,𝑞

Figure 3.4: Concentrated spiral windings approximated by concentric loops with calculation parameters

Here, the width of the tracks is denoted by 𝑤𝑡𝑟 and the spacing between two adjacent tracks is
denoted by 𝑤𝑠 . Furthermore 𝑤𝑑 is defined by:

𝑤𝑑 = 𝑤𝑡𝑟 + 𝑤𝑠 (3.5)

Because the magnetic flux density is variable along the radius, it is best to divide the loops in several
radial sections with equal thickness 𝑡𝑠𝑙 . It should be noted that the more sections used, the more
accurate the result will be, but this will increase calculation time.

The total magnetic flux ∅𝑆 through a surface 𝑆 is per definition.

∅𝑆 = ∬ 𝐵 𝑑𝑆 (3.6)

In this calculation, the average of the magnetic flux density per slice is used. This average is obtained
using the previous determined integral of the magnetic flux density, where the radius and integration
limits (𝜑𝑀𝐴𝑋,𝑙,𝑞 , 𝜑𝑀𝐼𝑁,𝑙,𝑞 ) are variable for every loop.
𝜑𝑀𝐴𝑋,𝑙,𝑞
1
𝐵̅ = ∫ 𝐵𝑑𝜑 (3.7)
𝜑𝑀𝐴𝑋,𝑙,𝑞 − 𝜑𝑀𝐼𝑁,𝑙,𝑞
𝜑𝑀𝐼𝑁,𝑙,𝑞

28
Because of the winding shape, the integration limits for the calculation of the average magnetic flux
density and the angles for the arc length for the calculation of the surface area are not constant.
Therefore, these parameters should be calculated for every loop and slice individually. This is done
using the following equations
𝑡𝑠𝑙
𝑟𝑙,𝑞 = (𝑟𝑖 + (𝑙 ∗ 𝑤𝑑 )) + + (𝑞 ∗ 𝑡𝑠𝑙 ) (3.8)
2
𝑤𝑑
(𝑙 ∗ 𝑤𝑑 ) +
𝜑𝑀𝐼𝑁,𝑙,𝑞 = 𝜑𝑀𝐼𝑁,𝑐 + sin −1
( 2) (3.9)
𝑟𝑙,𝑞

𝑤𝑑
(𝑙 ∗ 𝑤𝑑 ) +
𝜑𝑀𝐴𝑋,𝑙,𝑞 = 𝜑𝑀𝐴𝑋,𝑐 − sin −1
( 2) (3.10)
𝑟𝑙,𝑞

Where 𝑟𝑙,𝑞 is the radius of the middle line of the slice that is being considered. 𝑙 is the loop number
with the outermost loop being 𝑙 = 0. In Figure 3.4, the orange loop is then 𝑙 = 1. Furthermore, 𝑞 is
the radial slice number where the innermost slice is 𝑞 = 0. In Figure 3.4 the highlighted slice is then
𝑞 = 3.

Here 𝜑𝑀𝐼𝑁,𝑐 and 𝜑𝑀𝐴𝑋,𝑐 denote the absolute angular position of the coils in reference to a zero
point. They simply account for the physical position of the coil on the stator. This is necessary in
order to obtain the correct magnetic flux density field in reference to the chosen coil. An example
can be seen in Figure 3.5 where the middle line of the coloured coil is chosen as the absolute zero
point.

𝜑𝑀𝐼𝑁,𝑐

𝜑𝑀𝐴𝑋,𝑐

Figure 3.5: Coil absolute position

The flux for slice 𝑞 with a thickness 𝑡𝑠𝑙 from loop 𝑙 is then:

∅𝑙,𝑞 = 𝐵̅𝑙,𝑞 𝑡𝑠𝑙 𝑟𝑙,𝑞 (𝜑𝑀𝐴𝑋 − 𝜑𝑀𝐼𝑁 )𝑙,𝑞 (3.11)

29
The total flux per concentric loop is the sum of the fluxes from all the radial sections. If one loop is
divided in 𝑘 sections, for example 𝑘 = 7 in Figure 3.4, the total flux per loop is then:
𝑘−1

∅𝑙 = ∑ ∅𝑙,𝑞 (3.12)
𝑞=0

This calculation should be done for each loop in this coil. The flux linkage per coil 𝜓𝑐𝑜𝑖𝑙 with 𝑗 turns,
for example 𝑗 = 3 in Figure 3.4, is then given by:
𝑗−1

𝜓𝑐𝑜𝑖𝑙 = ∑ ∅𝑙 (3.13)
𝑙=0

The generated flux linkage per pole can be seen in Figure 3.6. The parameters that were used for this
analysis are listed in Table 3.2.
Table 3.2: Parameters used in flux linkage analysis

𝑃𝑎𝑟𝑎𝑚𝑒𝑡𝑒𝑟 𝑉𝑎𝑙𝑢𝑒
𝑀𝑎𝑔𝑛𝑒𝑡 𝑟𝑒𝑚𝑎𝑛𝑒𝑛𝑐𝑒 𝐵𝑟𝑒𝑚 [T] 1,38
𝐴𝑖𝑟𝑔𝑎𝑝 [mm] 2,8
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑢𝑟𝑛𝑠 3
𝑊𝑖𝑑𝑡ℎ 𝑜𝑓 𝑡ℎ𝑒 𝑡𝑟𝑎𝑐𝑘𝑠 𝑤𝑡𝑟 [mm] 2,06
𝑆𝑝𝑎𝑐𝑖𝑛𝑔 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑡𝑟𝑎𝑐𝑘𝑠 𝑤𝑠 [mm] 0,8
𝑀𝑎𝑔𝑛𝑒𝑡 𝑤𝑖𝑑𝑡ℎ [mm] 10
𝑀𝑎𝑔𝑛𝑒𝑡 𝑙𝑒𝑛𝑔𝑡ℎ [mm] 18
𝑀𝑎𝑔𝑛𝑒𝑡 ℎ𝑒𝑖𝑔ℎ𝑡 [mm] 5
𝑀𝑎𝑔𝑛𝑒𝑡 𝑖𝑛𝑛𝑒𝑟 𝑟𝑎𝑑𝑖𝑢𝑠 𝑟𝑖,𝑃𝑀 [mm] 15,4
𝑀𝑎𝑔𝑛𝑒𝑡 𝑜𝑢𝑡𝑒𝑟 𝑟𝑎𝑑𝑖𝑢𝑠 𝑟𝑜,𝑃𝑀 [mm] 43,4

Figure 3.6: Flux linkage per pole

30
3.1.3 Back EMF

The back EMF that one coil produces can now be determined using Faraday’s law of induction:
𝑑𝜓𝑐𝑜𝑖𝑙 (𝜃𝑚 ) 𝑑𝜓𝑐𝑜𝑖𝑙 (𝜃𝑚 ) 𝑑𝜃𝑚 𝑑𝜓𝑐𝑜𝑖𝑙 (𝜃𝑚 )
𝐸𝑙 = − =− =− 𝜔 (3.14)
𝑑𝑡 𝑑𝜃𝑚 𝑑𝑡 𝑑𝜃𝑚
Where 𝜔 is the mechanical rotational speed of the motor.

The motor consists of multiple coils per phase, the generated back EMFs should thus be summed
according to how they are connected in order to get the back EMF that each full phase produces. The
connection diagram is automatically generated using the Bavaria online winding scheme calculator
tool [10]. This is a 12 slot, 10 pole motor. The connection diagram is shown in Figure 3.7.

Figure 3.7: Bavaria tool, winding connection diagram. Adapted from [10]

The generated back EMF from phase one rotating with a speed of 6000 rpm can be seen in Figure
3.8. Also, the induced back EMFs from the different coils of phase 1 can be seen. In this plot the
connection polarity is very clear where coil 2 and coil 7 should be connected with an opposite
polarity than coil 1 and coil 8. This way the generated back EMFs are in phase and generate the
largest possible phase voltage.

31
Figure 3.8: Generated back EMFs in phase 1 at 6000 𝑟𝑝𝑚

32
3.2 Torque

The electromagnetic force can be calculated using the Lorentz force, in vector form this is:

𝐹⃗ = 𝐼𝑙⃗ × 𝐵
⃗⃗ (3.15)
Because of the model that is used, it is quite easy to determine the magnetic flux density using a
polar coordinate system, the point is given by the radius 𝑟 and the angle 𝜑. When the motor is
spinning, it is also important to factor in the actual rotor position 𝜃𝑚 .

To calculate the produced electromagnetic torque, the coils are divided in concentric trapezoidal
loops. Because the magnetic flux density is variable along the radius it is best to divide the loop wires
in small increments. Furthermore, not all sections of the coil wire will contribute to a production of
torque because of the geometrical orientation, only the straight conductor wires will contribute to a
torque production. These are the active portions of the windings and can be seen in Figure 3.9.

In order to calculate the torque, a 3-phase sinusoidal current is modelled for each phase.

𝐼𝐴 (𝑡) = Î sin(2𝜋𝑓𝑡) (3.16)

2𝜋
𝐼𝐵 (𝑡) = Î sin (2𝜋𝑓𝑡 − ) (3.17)
3
4𝜋
𝐼𝐶 (𝑡) = Î sin (2𝜋𝑓𝑡 − ) (3.18)
3

33
 𝐼 𝐼

𝐹𝑛,𝑙,𝑞,1
1 2
𝐹

𝑟𝑙,𝑞

𝛼𝑙,𝑞

Figure 3.9: Torque production calculation using concentric loops with calculation variables

Using the definition off the Lorentz force and the correct orientation of the current and magnetic flux
density, the direction of the produced force can be determined.

The coils are approximated by concentric loops, the same method is used as in chapter 3.1.2. For this
calculation 𝑙 denotes the loop number with the outermost loop being 𝑙 = 0. In Figure 3.9 the
analysed loop is 𝑙 = 1. Furthermore, 𝑞 is the radial wire section number where the innermost section
is 𝑞 = 0. In Figure 3.9 the dark grey wire section is then 𝑞 = 1.

The torque 𝑇𝑙,𝑞,1 produced by the dark grey section of the wire in Figure 3.9 is then given by:

𝑇𝑙,𝑞,1 = 𝐹𝑛,𝑙,𝑞,1 ∗ 𝑟𝑙,𝑞 = (𝐼 ∗ 𝐵𝑙,𝑞,1 ∗ 𝐿 ∗ cos(𝛼𝑙,𝑞 )) ∗ 𝑟𝑙,𝑞 (3.19)

Where 𝐵𝑙,𝑞,1 is the magnetic flux density in the middle of the wire section, 𝐼 is the current running
through it and 𝐿 is the length of this wire section. It is also important to include a correction factor
cos(𝛼𝑙,𝑞 ) because only the part of the produced Lorentz force that is perpendicular to the radius of
the rotor will contribute to a torque production. As shown in Figure 3.9, 𝛼𝑙,𝑞 is the angle between the
tangential line of the conductor section and the radius at the centre point of the conductor section.

In order to obtain the angular polar position for determining the magnetic field density 𝐵𝑙,𝑞,1 , the
parameters 𝑟𝑙,𝑞 and 𝜑𝑀𝐼𝑁,𝑙,𝑞 can be used, these can be calculated using equation (3.8) and equation
(3.9) respectively. For 𝐵𝑙,𝑞,2 , 𝑟𝑙,𝑞 and 𝜑𝑀𝐴𝑋,𝑙,𝑞 should be determined using equation (3.8) and
equation (3.10) respectively. Furthermore, 𝛼𝑙,𝑞 can be determined using the following equation

34
 𝑤𝑑
(𝑙 ∗ 𝑤𝑑 ) +
𝛼𝑙,𝑞 = sin −1
( 2) (3.20)
𝑟𝑙,𝑞

The torque 𝑇𝑙,𝑞,2 produced by the opposite wire section can be calculated using formula (3.21).

𝑇𝑙,𝑞,2 = (−𝐼 ∗ 𝐵𝑙,𝑞,2 ∗ 𝐿 ∗ cos(𝛼𝑙,𝑞 )) ∗ 𝑟𝑙,𝑞 (3.21)

The only difference with 𝑇𝑙,𝑞,1 is the magnetic flux density 𝐵𝑙,𝑞,2 and that the direction of current flow
is now in the opposite direction.

In order to get the full torque produced by these two wires, the calculation should be repeated for
every section and then summed. With 𝑛 equally sized sections per loop, for example nine radial
conductor segments of the second winding turn shown in Figure 3.9, the total torque produced by
the concentric loop 𝑙 is given by:
𝑛−1

𝑇𝑙 = ∑ 𝑇𝑙,𝑞,1 + 𝑇𝑙,𝑞,2 (3.22)

𝑞=0

The total torque per coil, where 𝑔 is the number of concentric turns per winding, for example 𝑔 = 3
in Figure 3.9, is then:
𝑔−1

𝑇𝑐𝑜𝑖𝑙 = ∑ 𝑇𝑙 (3.23)
𝑙=0

The motor consists of 12 concentric coils for each PCB side that are grouped in three phases. Using
the online Bavaria tool, the coils should be connected with the right polarity, this can be seen in
Figure 3.7.

The torque produced by phase one and its individual coils can be seen in Figure 3.10. The analysis
was done with three turns per winding, an airgap of 2,8 mm, at 6000 rpm and a phase current of
1,0 A. For the magnet setup, the same values as in Table 3.2 were used.

Figure 3.10: Torque production from phase 1 and its individual coils

35
The produced torque of the 3 phases and the full motor torque can be seen in Figure 3.11.

Figure 3.11: Torque production of each phase and full motor torque

Because this is a synchronous motor, one thing to consider is the load angle 𝜍 for this motor. If the
motor is running with no load attached, the load angle will be zero. If, however the motor is used at
its maximum torque, the load angle will be 90°, this can also be seen as a 90° phase difference
between the terminal voltage and generated back EMF.

In order to model the maximum torque this motor can produce, the rotor is mechanically shifted by
changing the mechanical angular rotor position 𝜃𝑚 so that it lags the stator field until the torque is
maximum. For this model, if the mechanical rotor position is not shifted, the load angle and also
torque will be 0. If the mechanical rotor position 𝜃𝑚 is shifted by 18°, the torque will be maximum.
This 18° mechanical degree shift will contribute to a 90° electrical shift due to the number of pole
pairs 𝑝 = 5.

𝜃𝑚 − 0° → 𝜍 = 0 → 𝑇 = 0
𝜃𝑚 − 18° → 𝜍 = 90° → 𝑇 = 𝑀𝐴𝑋
The change in torque in function of the load angle 𝜍 can be seen in Figure 3.12.

36
Figure 3.12: Influence of load angle ς on motor torque

37
3.3 Maximum turns per winding

In order to optimize the motor, a function is written that will determine the maximum number of
turns per winding possible with the given winding parameters [11].

The total amount of possible turns 𝑁𝑐 is limited by the inequality

1 𝑟𝑜 − 𝑟𝑖
𝑁𝑐 < ( + 1) (3.24)
2 𝑤𝑑
It is possible that besides trapezoidal shaped turns there are also triangular shaped turns. The
number of purely trapezoidal shaped turns 𝑁𝑐1 is given by
𝜃
𝑟𝑖 sin ( 2𝑐𝑠 )
𝑁𝑐1 < +1 (3.25)
𝜃
𝑤𝑑 (1 − sin ( 2𝑐𝑠 ))

If 𝑁𝑐 > 𝑁𝑐1 , this means that there are also triangular shaped turns, the number of triangular shaped
turns 𝑁𝑐2 is then
1 𝑟𝑜 − 𝑟𝑖 𝜃𝑐𝑠
𝑁𝑐2 < ( − csc ( ) − 2𝑁𝑐1 + 2) (3.26)
𝜃 𝑤𝑑 2
1 + csc ( 𝑐𝑠 )
2

38
3.4 Electrical Wire Resistance

A third component that is important for the full electromagnetic analysis is the wire resistance of
each phase. This will be important to calculate the maximum allowable current per phase and thus
limit the power of the motor.

Again, the coil is divided in concentric loops as can be seen in Figure 3.13. The coil resistance can
then be calculated using simple trigonometry where the individual loops are divided in two radial line
segments and two concentric arc lengths. Because of the design, it is possible that there is no inner
arc so that the coil just comes to a point.

𝑟𝑜

𝛽𝐼𝑛,𝑖
𝜃𝑐𝑠
2 𝑟𝑖
𝜙𝐼𝑛,𝑖

Figure 3.13: Track length calculation setup with calculation variables

These calculations use the following formulas.

The coil span is simply

𝜋
𝜃𝑐𝑠 = (3.27)
𝑝
It is necessary to calculate the inner starting angle β𝐼𝑛 of the arc for every loop. This inner angle of
the 𝑖 𝑡ℎ loop is given by the following formula
𝑤
(𝑖 ∗ 𝑤𝑑 ) + ( 𝑑 )
𝜙𝐼𝑛,𝑖 = sin −1
( 2 ) (3.28)
𝑟𝑖 + (𝑖 ∗ 𝑤𝑑 )

𝜃𝑐𝑠
𝛽𝐼𝑛,𝑖 = − 𝜙𝐼𝑛,𝑖 (3.29)
2

39
This formula is nearly identical for the outer starting angle
𝑤
(𝑖 ∗ 𝑤𝑑 ) + ( 𝑑 )
𝜙𝑂𝑢𝑡,𝑖 = sin −1
( 2 ) (3.30)
𝑟𝑜 − (𝑖 ∗ 𝑤𝑑 )

𝜃𝑐𝑠
𝛽𝑂𝑢𝑡,𝑖 = − 𝜙𝑂𝑢𝑡,𝑖 (3.31)
2
The length of the 𝑖 𝑡ℎ loop is then given by the sum of the outer arc length 𝐿𝑂𝑢𝑡,𝑖 , the inner arc length
𝐿𝐼𝑛,𝑖 and the 2 straight conductor wires 𝐿𝑠,𝑖

𝐿𝑂𝑢𝑡,𝑖 = 2 ∗ (𝑟𝑜𝑢𝑡 − (𝑖 ∗ 𝑤𝑑 )) ∗ 𝛽𝑂𝑢𝑡,𝑖 (3.32)

𝐿𝐼𝑛,𝑖 = 2 ∗ (𝑟𝑖𝑛 + (𝑖 ∗ 𝑤𝑑 )) ∗ 𝛽𝐼𝑛,𝑖 (3.33)

𝑤 𝑤
((𝑖 ∗ 𝑤𝑑 ) + ( 2𝑑 )) ((𝑖 ∗ 𝑤𝑑 ) + ( 2𝑑 ))
𝐿𝑠,𝑖 =2∗( − ) (3.34)
tan(𝜙𝑂𝑢𝑡,𝑖 ) tan(𝜙𝐼𝑛,𝑖 )

However, for the inner starting angle an additional check should be done in order to make sure that
there is in fact an inner arc. This is done by checking if

𝛽𝐼𝑛,𝑖 > 0

If this is not the case, there is no inner arc length and the straight conductor segments will be
shorter. The length of the straight lines is then
𝑤 𝑤
((𝑖 ∗ 𝑤𝑑 ) + ( 𝑑 )) ((𝑖 ∗ 𝑤𝑑 ) + ( 𝑑 ))
2 2
𝐿𝑠,𝑖 =2∗( − ) (3.35)
tan(𝜙𝑂𝑢𝑡,𝑖 ) 𝜃
tan ( 2𝑐𝑠 )

The total length of the 𝑖 𝑡ℎ loop is then simply

𝐿𝑡𝑜𝑡,𝑖 = 𝐿𝑂𝑢𝑡,𝑖 + 𝐿𝐼𝑛,𝑖 + 2𝐿𝑠,𝑖 (3.36)

This calculation should be repeated for every loop, the total length 𝐿𝑡𝑜𝑡 of the coil is then the sum of
the different loop lengths.

The resistivity at temperature 𝑇 can be calculated using the following formula

𝜌𝑇 = 𝜌20°𝐶 ∗ (1 + 𝛼𝑐 ∗ (𝑇 − 20°)) (3.37)

Here, 𝛼𝑐 is the temperature coefficient and 𝜌20°𝐶 is the resistivity of the used conductor material at
20 °C.

The total resistance of the phase is then

𝜌𝑇 ∗ 𝐿𝑡𝑜𝑡
𝑅𝑝ℎ𝑎𝑠𝑒 = (3.38)
𝑤𝑡𝑟 ∗ 𝑡𝑡𝑟

40
3.5 Magnetic flux density 3D visualization

In order to get a better understanding of the magnetic flux density field in its entirety, a function is
written that will visualise the magnetic field in a 3D manner. For this to work, a matrix must be made.
The rows of the matrix denote a constant radius 𝑟 along the rotor and the columns denote a constant
polar coordinate angle 𝜑. For every specific point with variable 𝑟 and 𝜑, the magnetic flux density
field should be calculated. To get an accurate representation, the rotor should be divided in many
concentric circles with constant 𝑟, these circles should then be divided in many points with each its
own polar coordinate angle 𝜑.

Furthermore, because this is a polar representation, an additional conversion should be made. The
built-in function from MATLAB pol2cart will convert the matrix into a cartesian representation. After
this, the matrix can be visualised with a surf plot. An example of such a plot can be seen in Figure
3.14 and Figure 3.15.

Figure 3.14: 3D magnetic flux density plot

41
Figure 3.15: 3D magnetic flux density plot, top view

42
 4 Electromagnetic Design Using Analytical Model
The motor consists of three main critical parts being the rotor magnets, the stator windings, which is
in this case a printed circuit board, and the steel back plate supporting the magnets. It is therefore
important to make good decisions about the design and choice of these components in order to
optimize the motor as much as possible. The analytical model will be used to examine and
substantiate these choices.

4.1 Magnet Choice and magnet topology

It is important to get a strong magnetic field in between the two rotor halves. Therefore, neodymium
magnets were chosen with a grade that is as high as possible.

There are a few limitations when it comes to choosing the magnets.

1. Cost: The motor should partially be optimized for cost, therefore only standard shaped
magnets can be used. These include rectangular and square shaped magnets, custom wedge-
shaped magnets are not an option.
2. Temperature rating: When the motor is running at peak power it will get hot. If the
temperature exceeds the maximum temperature of the magnets, they can demagnetize and
be permanently damaged. This demagnetization will of course decrease motor power or
worse, render it completely useless. Therefore, special grade magnets were chosen, these
grades include:
a. H: Max working temperature = 120 °C
b. SH: Max working temperature = 150 °C
c. UH: Max working temperature = 180 °C
3. Manufacturability: In order to get a maximum coverage of magnets on the rotor, multiple
shapes and sizes of magnets could be used. However, the motor should be easy to
manufacture and assemble, because of this a maximum of two magnets per pole was
chosen.

A few options were available and will be examined and discussed in the following chapter.

43
4.1.1 Configurations

A visual representation of the different magnet configurations can be seen in Figure 4.1. The green
plate is the maximum size of the stator PCB.

(a) (c)

(b) (d)

Figure 4.1: Different magnet topologies

The different parameters of these magnets are listed in Table 4.1.

Table 4.1: Magnet topology parameters

Configuration Grade Dimensions # Magnets per Total price [€]

[mm] rotor
a N45SH 18 × 10 × 5 20 15,30
b N45SH 18 × 10 × 5 40 44,40
c N45SH 17,5 × 7,3 × 4,5 40 135,12
d N48H 20 × 10 × 5 40 137,13

44
4.1.2 Magnet Test Setup 1

A simple parameter to compare the magnets is the coverage ratio of the magnets, this is the surface
area of the magnets in comparison to the surface area of the back plate. This will give a fast measure
of performance indicating how much of the available surface area is covered with magnetic material.
The calculated coverage ratios are listed in Table 4.2.
Table 4.2: Coverage ratio of different magnet topologies

Configuration Surface area back Surface area magnet Coverage ratio [%]
plate [cm2 ] [cm2 ]
a 28,4 18,0 63,4
b 54,3 36,0 66,3
c 39,6 25,6 64,6
d 60,4 40,0 66,2

From this simple analysis, it is clear that configuration b utilizes the available surface area most
optimal.

4.1.3 Magnet Test Setup 2

To make a better decision, a new parameter is defined so that these configurations can be compared
with only one single measure of performance. This parameter is the produced back EMF per weight
V
of the magnets 𝜆 [ ].
kg

To get a uniform analysis, the analytical model will be run with the parameters shown in Table 4.3 for
each magnet setup.
Table 4.3: Magnet test setup 2 parameters

Parameter Value
𝑅𝑜𝑡𝑜𝑟 𝑠𝑝𝑒𝑒𝑑 𝜔 1500 rpm
𝐴𝑖𝑟𝑔𝑎𝑝 3 mm
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑢𝑟𝑛𝑠 𝑝𝑒𝑟 𝑤𝑖𝑛𝑑𝑖𝑛𝑔 1

The absolute values of these chosen parameters are not very crucial because only the relative
difference between the 𝜆’s is important for the comparison.

The winding is modelled so that it includes one full pole.

First, the magnets were modelled in MATLAB in order to obtain the magnetic flux density field. These
visual representations were generated using the method described in chapter 3.5 and can be seen in
Figure 4.2. In these figures, the green circle also represents the maximum allowed PCB diameter of
100 mm.

45
 𝑀𝑎𝑔𝑛𝑒𝑡𝑖𝑐 𝐹𝑙𝑢𝑥 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 𝑓𝑖𝑒𝑙𝑑 𝑠𝑡𝑟𝑒𝑛𝑔𝑡ℎ𝑡 [𝑇]
(a) (b)

(c) (d)

Figure 4.2: Magnetic flux density field for different magnet topologies

Then the analytical model is run with the given parameters for these 4 configurations, the data
obtained from this analysis is listed in Table 4.4.
Table 4.4: Generated back EMF per magnet topology

Configuration RMS back EMF [V] Full weight of the magnets 𝜆 [V⁄kg]
[kg]*
a 0,256 0,135 1,89
b 0,520 0,270 1,93
c 0,325 0,172 1,88
d 0,603 0,300 2,01
*Density of magnets: 7,5 g/cm³

Configuration d has the highest produced back EMF per weight.

46
4.1.4 Magnet Final Configuration

Configuration b and d are the best options. While b has the highest coverage ratio, d still produces a
stronger magnetic field and thus more back EMF. This is because configuration d uses a stronger
grade magnet, N48 as opposed to N45. These magnets have a stronger remanent field. The increase
in back EMF per weight from configuration b to d is about 4%.

However, for the final prototype configuration b was chosen because of the price difference.
Configuration d is 3 times more expensive for only about 4% increase in back EMF per weight.

A second advantage of configuration b is that these magnets are of the SH grade type as opposed to
H grade for configuration d. Because of this, the motor will be less prone to thermal damage.

47
4.2 PCB Winding Design

The second critical component of this motor is the PCB stator. The design of these windings is very
important and because many parameters can be changed, quite complex. For this motor
concentrated trapezoidal windings are used. The design for this stator was made with the goal of
milling the tracks straight from a bare copper plate. This way thick copper could be used to increase
the current carrying capabilities of the motor. Because of this, the minimum spacing between tracks
𝑤𝑠 is 0,8 mm, this is the diameter of a fine mill. However, due to time constraints and technical
difficulties a standard PCB will be made and tested instead with the same track spacing. Ideally this
track spacing should be as small as possible to optimally make use of the PCB surface area.

A few parameters should be determined in order to get an optimal design, these include the track
width 𝑤𝑡𝑟 , number of turns per winding, the radius of the innermost winding track 𝑟𝑖 and the radius
of the outermost winding track 𝑟𝑜 . These parameters should be balanced between each other to
simultaneously keep the length and thus resistance low while maintaining enough surface area in
order to generate more back EMF and torque.

To do so, the analytical model calculates the produced motor power while varying 𝑟𝑖 , 𝑟𝑜 and 𝑤𝑡𝑟 . In
order to store this data in a matrix, the inner and outer radii are reduced to one variable using a
calculation factor 𝑑𝑤 . For this optimization, the starting radii are chosen to be equal to the inner and
outer radius of the magnets, these are

𝑟𝑖 = 15,4 mm
𝑟𝑜 = 43,4 mm
Using 𝑑𝑤 , the radii of the winding is then given by

𝑟𝑖 = 15,4 − 𝑑𝑤 (4.1)

𝑟𝑜 = 43,4 + 𝑑𝑤 (4.2)

This way the calculation radii are symmetrically being moved along the radial axis using only 1
parameter.

Furthermore, the motor power is calculated for different track widths.

With a particular 𝑟𝑖 , 𝑟𝑜 and 𝑤𝑡𝑟 , the maximum number of possible turns per winding is determined
using the function described in chapter 3.3. After that, the coil resistances are determined. For this
analysis, a maximum of 4 W heat radiation from copper losses for the entire PCB is chosen. This way
the maximum current per phase can be determined using ohms law. The two sides of the PCB are in
series in order to get the highest possible back EMF. The maximum current per phase, with a phase
resistance 𝑅𝑡𝑜𝑡,𝑠𝑒𝑟𝑖𝑒𝑠 , is then given by

4
(3 W)
𝐼𝑝ℎ𝑎𝑠𝑒 =√ (4.3)
𝑅𝑡𝑜𝑡,𝑠𝑒𝑟𝑖𝑒𝑠

After this it is only a matter of calculating the produced back EMF and torque to get the motor
power.

A few parameters that were used for this optimization can be seen in Table 4.5.

48
Table 4.5: Motor parameters for testing optimal winding configuration.

Parameter Value
𝑀𝑜𝑡𝑜𝑟 𝑠𝑝𝑒𝑒𝑑 6000 rpm
𝑇𝑟𝑎𝑐𝑘 𝑠𝑝𝑎𝑐𝑖𝑛𝑔 𝑤𝑠 0,8 mm
𝑇𝑟𝑎𝑐𝑘 𝑡𝑒𝑚𝑝𝑒𝑟𝑎𝑡𝑢𝑟𝑒 50 °C
𝐴𝑖𝑟𝑔𝑎𝑝 1,4 mm
𝑇𝑟𝑎𝑐𝑘 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 𝑡𝑡𝑟 34,8 µm
𝑀𝑎𝑥𝑖𝑚𝑢𝑚 𝑝𝑜𝑤𝑒𝑟 𝑑𝑖𝑠𝑠𝑖𝑝𝑎𝑡𝑖𝑜𝑛 4W

This function will run for track widths ranging from 0,16mm to 3,5mm with a step size of 0,1mm. For
the radii, 𝑑𝑤 will go from −10 mm to 10 mm with a step size of 1 mm.

The obtained 3D plot can be seen in Figure 4.3.

Figure 4.3: Motor power in function of track radii and track width

It is clear that a winding design with larger track widths is better. This is to be expected because as
the track thickness increases, the number of tracks decreases. And because the track spacing 𝑤𝑠 is
quite big, with less tracks there are less spacing areas with no copper. The copper surface area to
total PCB surface area ratio is thus bigger.

For the track radii, there seems to be a maximum for 𝑑𝑤 ranging from 0 to 3 mm, this can clearly be
seen in Figure 4.4.

49
First of all, the track radii are chosen, here a 𝑑𝑤 of about 3 mm seems to give optimal results. For the
final design, the following values will be used.

𝑟𝑖 = 12,5 mm
𝑟𝑜 = 44,0 mm
One thing to note is that 𝑟𝑜 does not have an offset of 3 mm in reference to the magnet outer radius,
this is done on purpose because the maximum permissible PCB diameter is 100 mm. Choosing an
offset of just 1 mm has the advantage that there is enough room on the outside diameter to route
individual coil connections and make the star connection as well.

Figure 4.4: Side view of motor power in function of track radii and track width

After the track radii are chosen, the model is run again where 𝑟𝑖 and 𝑟𝑜 are now set. This way a plot
can be made where only the track width is variable. Alongside this curve other characteristics can
then easily be displayed like number of turns per winding, current draw and produced back EMF in
order to make a more informed design decision. The motor power in function of the track width
alongside the number of turns per winding can be seen in Figure 4.5. Also, the produced back EMF
and maximum current in function of the track width are shown in Figure 4.6. The produced motor
torque and phase resistances can be seen in Figure 4.7.

50
Figure 4.5: (Left) Motor power in function of track width; (Right) Maximum number of turns per winding in function of track
width

Figure 4.6: (Left) Back EMF in function of track width; (Right) Maximum phase current in function of track width

Figure 4.7: (Left) Motor torque in function of track width; (Right) Phase resistance in function of track width

51
It is clear from Figure 4.5 that thicker tracks offer more motor power. However, there are practical
limits and difficulties that influence this decision. First of all, the track thickness should not be chosen
too large because when for example only one turn per winding is possible, the design of the PCB
would get hard and inefficient. Furthermore, the resistance would get very low leading to a high
current draw, but small generated back EMF as can be seen in Figure 4.6, this is undesirable and
would make controlling the motor unnecessarily hard.

For this design, a track width of 2,06 mm is chosen, this choice offers a good balance between motor
power and generates a practical back EMF. The characteristics of this motor following the design
choices are listed in Table 4.6.
Table 4.6: Final motor design characteristics

Parameter Value
𝑇𝑟𝑎𝑐𝑘 𝑤𝑖𝑑𝑡ℎ 2,06 mm
𝑀𝑜𝑡𝑜𝑟 𝑝𝑜𝑤𝑒𝑟 40,12 W
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑙𝑜𝑜𝑝𝑠 𝑝𝑒𝑟 𝑤𝑖𝑛𝑑𝑖𝑛𝑔 3
𝐺𝑒𝑛𝑒𝑟𝑎𝑡𝑒𝑑 𝑝ℎ𝑎𝑠𝑒 𝑏𝑎𝑐𝑘 𝐸𝑀𝐹 7,31 V
𝑃ℎ𝑎𝑠𝑒 𝑐𝑢𝑟𝑟𝑒𝑛𝑡 1,84 A
𝑀𝑜𝑡𝑜𝑟 𝑠𝑝𝑒𝑒𝑑 6000 rpm
𝑀𝑜𝑡𝑜𝑟 𝑇𝑜𝑟𝑞𝑢𝑒 63,85 mNm
𝑃ℎ𝑎𝑠𝑒 𝑟𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 0,39 Ω

The PCB was designed using Fusion 360 and Eagle and can be seen in Figure 4.8.

(𝑎) (𝑏)

Figure 4.8: PCB coil track design; (a) Front; (b) Back

52
4.3 Backplate analysis

4.3.1 Electromagnetic design

The steel back plate is a vital component that is necessary for mechanical strength and most
importantly acts as a low reluctance return path for the magnetic flux from the rotor magnets. The
lower the reluctance in this part, the stronger the magnetic field in the air gap will be.

For the material, a low carbon steel is used. Because the material is set, there is only 1 parameter
that can be altered to change the reluctance. This is the thickness of the back plate 𝑡𝐵𝑝 . However,
because it is also a design goal to optimize the power density, the thickness should be chosen
carefully to keep the weight and reluctance as low as possible.

Because the analytical model in MATLAB does not account for saturation in the back plate, FEM
software is used, and an approximation is made to account for saturation in the model. In order to
model this saturation, the FEM software runs the model with the desired back plate thickness first.
This gives a magnetic field density plot along a given radius that also accounts for saturation. After
this, the analytical model is run in an iterative manner where the remanence of the magnet 𝐵𝑟𝑒𝑚 is
changed in MATLAB until the magnetic field density plots are as close to each other as possible. This
way the attenuation due to saturation is modelled. Changing the 𝐵𝑟𝑒𝑚 however is an approximation
and is not the best method for modelling saturation. This is because saturation will not only
attenuate the magnetic flux density but will also distort the waveform. Yet this method was used
because it is easy with quite accurate results when the level of saturation is not too high.

In order to make sure that both plots are similar within a given limit, the normalised root mean
square error (NRMSE) method was used. For normalization, the peak-to-peak values of the FEM
magnetic density were used. This can be seen in formula (4.4).

𝑛 2
√∑𝑖=1(𝐵𝐹𝐸𝑀,𝑖 − 𝐵𝐴𝑛,𝑖 )
𝑛
𝑁𝑅𝑀𝑆𝐸 = (4.4)
𝐵𝐹𝐸𝑀,𝑀𝑎𝑥 − 𝐵𝐹𝐸𝑀,𝑀𝑖𝑛

When the magnetic flux density field from the analytical model is within specifications, the generated
motor power can be calculated using the determined magnet remanence.

Furthermore, when the thickness of the back plate changes, so does the weight.

For this analysis, a few motor parameters were chosen so that the results could be compared, these
can be seen in Table 4.7.
Table 4.7: Back plate test motor parameters

Parameter Value
𝐹𝑙𝑢𝑥 𝑑𝑒𝑛𝑠𝑖𝑡𝑦 𝑝𝑙𝑜𝑡 𝑟𝑎𝑑𝑖𝑢𝑠 40 mm
𝐴𝑖𝑟𝑔𝑎𝑝 4 mm
𝑆𝑝𝑒𝑒𝑑 6000 rpm
𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑤𝑖𝑛𝑑𝑖𝑛𝑔𝑠 3
𝐶𝑜𝑝𝑝𝑒𝑟 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 0,0348 mm
𝑇𝑟𝑎𝑐𝑒 𝑤𝑖𝑑𝑡ℎ 2,06 mm

53
In Figure 4.9 the magnetic flux density from the FEM analysis can be seen with a back plate thickness
of 5 mm. The magnetic flux density can also be seen in the graph where the black curve is the flux
density from the FEM analysis that accounts for saturation. The green curve is the flux density from
the analytical model where the magnet remanence was attenuated. Furthermore, the name in the
legend “matlab_1.38” refers to the analytical model where a magnet remanence 𝐵𝑟𝑒𝑚 = 1,38 T was
chosen. In this case because the back plate was thick enough and almost no saturation occurred, no
attenuated was needed.

1
0.8
0.6
Magnetic Flux Density [T]

0.4
0.2
0 FEMM
0 90 180 matlab_1.38
-0.2
-0.4
-0.6
-0.8
-1
Angular Position [ °]

Figure 4.9: Magnet saturation modelling with backplate thickness = 5 mm; (Top) FEM magnet flux density plot; (Bottom)
Comparison between analytical and FEM magnetic field graph

In Figure 4.10 the magnetic flux density from the FEM analysis can be seen with a back plate
thickness of 0,5 mm. The magnetic flux density can also be seen in the graph where the black curve is
the flux density from the FEM analysis that accounts for saturation. The green curve is the flux
density from the analytical model where the magnet remanence was attenuated and a magnetic
remanence 𝐵𝑟𝑒𝑚 = 0,727 T was chosen, the red curve is the unattenuated curve. It is clear that
there is a big error when back plate saturation is ignored.

As can be seen in the graphs, this method gives a very accurate approximation. However, when the
back plate gets thinner, the flux density field not only attenuates but also distorts due to excessive
saturation.

54
 1
0.8
0.6
Magnetic Flux Density [T]

0.4
0.2
FEMM
0
0 90 180 Matlab_1.38
-0.2
Matlab_0.727
-0.4
-0.6
-0.8
-1
Angular Position [ °]

Figure 4.10: Magnet saturation modelling with backplate thickness = 0.5mm; (Top) FEM magnet flux density plot; (Bottom)
Comparison between analytical and FEM magnetic field graph

The study was done for back plates ranging from a thickness of 0,5 mm to 6,0 mm. The results can
be seen in Table 4.8.
Table 4.8: Analysis results of the magnet saturation modelling

Backplate Magnet NRMSE Power [W] Backplate Full Motor Power

thickness Remanence Weight Weight W
Density [kg]
[mm] 𝐵𝑟𝑒𝑚 [T] [kg] [kg]

0,5 0,727 0,0237 22,28 0,055 0,426 52,25

1 0,85 0,0202 26,05 0,113 0,483 53,96
1,5 0,95 0,0179 29,12 0,169 0,539 54,00
2 1,038 0,0164 31,81 0,226 0,596 53,41
2,5 1,122 0,0154 34,39 0,282 0,652 52,74
3 1,2 0,0149 36,78 0,338 0,708 51,91
3,5 1,269 0,0146 38,89 0,395 0,765 50,85
4 1,328 0,0144 40,70 0,451 0,821 49,56
5 1,38 0,0144 42,30 0,564 0,934 45,28
6 1,385 0,0143 42,45 0,677 1,047 40,54

For the calculation of the full motor weight, the following weights were used: the weight of the
backplate, 270 g for the magnets and 100 g extra for prints and bearings.

The motor power in function of the back plate thickness can be seen in Figure 4.11.

55
 Figure 4.11: Motor power in function of back plate thickness

From this graph it is clear that the motor power flattens out for back plates thicker than about 4 mm.
This is to be expected because the reluctance gets so low that there is almost no saturation
happening in the steel part. Increasing the thickness will thus not strengthen the magnetic field
density in the airgap and consequently bring no further power gain.

A second important factor is the power density. Figure 4.12 show the power density in function of
the back plate thickness.

Figure 4.12: Motor power density in function of back plate thickness

This graph shows that there is a maximum power density achievable with a back plate of 1 mm. This
power density however is rather constant up until a plate thickness of about 3 mm. The steep
decline with higher thicknesses is normal because the power generated evens out but the weight of
the motor goes up drastically.

A motor with a backplate of 1 mm would have an optimal power density, however a backplate of
3 mm was chosen to balance the motor power and motor power density. Because of this extra

56
thickness, the motor power increases with almost 50% while the power density only decreases with
less than 4%.

Furthermore, this thicker back plate will increase the mechanical strength of the motor. These
mechanical strength requirements should not be underestimated. Due to the small airgap, very large
axial forces act on the two rotor halves and the components separating them. The mechanical design
considerations will be discussed in the following section.

4.3.2 Computation of axial electromagnetic forces

In order to calculate the mechanical requirements, the attraction forces between the magnets should
first be determined. This can be done using formula (4.5) [12].

𝐵2 𝐴
𝐹𝑚 = (4.5)
2µ0
Where 𝐹𝑚 and 𝐴 denote the force on the magnetized surface and the surface area of this magnetized
surface, respectively. 𝐵 is the magnetic flux density at the surface of the magnet.

Formula (4.5) is only valid if fringing effects are minimal and if the volume of the airgap is smaller
than the volume of the magnets. In this case, both requirements are met. This can clearly be seen in
Figure 4.13 where almost no fringing occurs. Also, in Figure 4.14 the magnetic flux density field plot is
shown taken from the surface of the magnets along the red line in Figure 4.13. It is clear that the 𝐵
field is nearly 0 outside the magnet perimeters and almost constant right above the magnet surfaces.
Furthermore, the analytical simulation gives accurate results compared with the FEM analysis.

Figure 4.13: FEM analysis with minimal fringing effects

57
 Figure 4.14: Magnetic flux density plot along magnet surface

In order to calculate the average magnetic flux for the force calculation, a similar method as in
chapter 3.1.2 will be used. The only difference is that the magnets will be used as integration limits
instead of the windings. Because the magnetic flux density field is function of the radius 𝑟, 1 magnet
pole is divided in radial sections. For this example, the pole is divided in 14 equally sized radial
sections as can be seen in Figure 4.15 along with the other calculation parameters. Only the force on
1 magnet pole should be determined because the force will be equal for all poles due to symmetry.

𝑡𝑠𝑙

𝑤𝑃𝑀

𝑟𝑑

𝛽𝑃𝑀,𝑑
𝑟𝑖,𝑃𝑀

Figure 4.15: Magnetic attraction force calculation setup

58
For each radial section formula (4.5) is calculated. The radius for radial section 𝑑 is given by
𝑡𝑠𝑙
𝑟𝑑 = 𝑟𝑖,𝑃𝑀 + ( ) + (𝑡𝑠𝑙 ∗ 𝑑) (4.6)
2
Where 𝑡𝑠𝑙 is the thickness off one radial section.

The angle 𝛽𝑃𝑀,𝑑 for calculating the arc length of slice 𝑑 is given by
𝑤𝑃𝑀
𝛽𝑃𝑀,𝑑 = sin−1 ( ) (4.7)
2𝑟𝑑
The surface area 𝑆𝑠𝑙 of slice 𝑑 is then
𝑆𝑠𝑙,𝑑 = 2𝛽𝑃𝑀,𝑑 ∗ 𝑟𝑑 ∗ 𝑡𝑠𝑙 (4.8)

The average 𝐵-field for slice 𝑑 is given by

𝛽𝑃𝑀,𝑑
1
̅̅
𝐵̅̅
𝑑 = ∫ 𝐵(𝜑) 𝑑𝜑 (4.9)
2𝛽𝑃𝑀,𝑑 −𝛽𝑃𝑀,𝑑

The total force exerted on the radial section is then

̅̅̅̅
(𝐵 2
𝑑 ) 𝑆𝑠𝑙,𝑑
𝐹𝑑 = (4.10)
2µ0
This calculation should be done for the entire rotor pole, the total force 𝐹𝑝 on one pole with 𝑙
number of radial slices is then
𝑙−1

𝐹𝑝 = ∑ 𝐹𝑑 (4.11)
𝑑=0

For a motor with 𝑝 pole pairs, the total attraction force on 1 rotor halve is then
𝐹𝑎𝑡𝑡 = 2𝑝 ∗ 𝐹𝑝 (4.12)

This calculation is carried out for the motor with a total airgap of 2,8 mm. The attraction force on
one rotor halve can be seen in Table 4.9 along with the force from a FEM analysis that uses the force
via weighted stress tensor method.
Table 4.9: Magnetic attraction forces comparison

Magnetic attraction force comparison

Analytical FEM
846,80 N 1083,92 N

There is a difference between these two forces, this could be due to a few factors. Firstly, in the
analytical model, an error is made by assuming that there is no fringing happening. Furthermore,
there is a small deviation in the magnetic density field between the analytical and the FEM model.

Secondly, an error in the FEM model is made due to the way this force is calculated. In the FEM
model, this motor is represented by a flat 2D sheet along a constant radius. To calculate the force,
the motor is represented by two FEM analysis, one along a radius going through the middle of the
first row of magnets, the inner most magnets. The second is represented by a section with a radius
going through the middle of the outer most magnets. The force is then calculated by using the

59
average flux density field along these sections and simply multiplying this with the surface area of the
inner or outer magnets of its respective average B field radius. This however is not completely
correct because the magnetic field is not constant along the radius of the motor.

That said, the forces are still of the same order of magnitude and further analysis can be done with
acceptable accuracy.

60
 5 Mechanical design
5.1 Back plate deflection

The forces on these magnets are transmitted through to the steel back plate, this plate is only
supported on its outermost diameter and thus deflection of the plate will occur. Young & Budynas
propose a few simple formulas to obtain the deflection in an annular plate, these formulas account
for geometry and loading parameters [13]. In order to use the proposed method, a few
simplifications to the 3D model must be made.

There are four conditions that must be met:

1. The plate is flat of uniform thickness and of homogenous isotropic material.

2. The thickness is not more than one quarter of the least transverse dimension, the
maximum deflection is not more than one-half of the thickness.
3. All forces are normal to the plain.
4. The plate is nowhere stressed beyond its elastic limit.

It is also assumed that there is no support on the inner edge of the annular steel plate. In the final
motor assembly, there are of course bearings in place that generate a counteracting force and thus
partially support the inner edge. This however is not the purpose of the bearings; therefore, they are
left out of this calculation and only the deflection without inner support is determined. In future
designs this deflection could be used to get an accurately determined preload on the bearings. Lastly
in the real model, the two rotor halves are separated by only three support spacers resting on lugs,
for this analysis however it is assumed that the steel annular plate is simply supported along the full
outer edge of the plate as can be seen in Figure 5.1.

Because the loading of this plate is nonuniform, these loads should be treated as several ring loads,
the total deflection is then the sum of the individual deflections due to their respective ring load.

The setup along with some calculation parameters is shown in Figure 5.1 where the green area is the
full magnetic load due to attraction forces, the red section is a single line load for one deflection
calculation.

𝑟0

𝑦𝑏
𝑏
𝑎

Figure 5.1: Representation of back plate deflection due to magnetic forces

The deflection 𝑦𝑏 [m] is given by

−𝑊𝑎3 𝐶1 𝐿9
𝑦𝑏 = ( − 𝐿3 ) (5.1)
𝐷 𝐶7
Where 𝑊 is the unit line load along a given radius 𝑟0 . 𝐷 is the plate constant defined by
3
𝐸 ∗ 𝑡𝑏𝑝
𝐷= (5.2)
12(1 − 𝜈 2 )

61
With

𝐸 = 𝑌𝑜𝑢𝑛𝑔’𝑠 𝑚𝑜𝑑𝑢𝑙𝑢𝑠 𝑜𝑓 𝑡ℎ𝑒 𝑠𝑡𝑒𝑒𝑙 𝑝𝑙𝑎𝑡𝑒 = 190 Gpa

𝑡𝑏𝑝 = 𝑇ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 𝑜𝑓 𝑡ℎ𝑒 𝑏𝑎𝑐𝑘 𝑝𝑙𝑎𝑡𝑒 [m]

𝜈 = 𝑃𝑜𝑖𝑠𝑠𝑜𝑛𝑠 𝑟𝑎𝑡𝑖𝑜 = 0,27

Furthermore 𝐶1 , 𝐿9 , 𝐶7 and 𝐿3 are constants defined by the following formulas
1+𝜈𝑏 𝑎 1−𝜈 𝑎 𝑏
𝐶1 = ln ( ) + ( − ) (5.3)
2 𝑎 𝑏 4 𝑏 𝑎
1 𝑎 𝑏
𝐶7 = (1 − 𝜈 2 ) ( − ) (5.4)
2 𝑏 𝑎
𝑟0 𝑟0 2 𝑎 𝑟0 2
𝐿3 = {[( ) + 1] ln ( ) + ( ) − 1} (5.5)
4𝑎 𝑎 𝑟0 𝑎
𝑟0 1 + 𝜈 𝑎 1−𝜈 𝑟0 2
𝐿9 = { ln ( ) + [1 − ( ) ]} (5.6)
𝑎 2 𝑟0 4 𝑎
In order to obtain the full deflection, only the radius of the applied force 𝑟0 and the applied unit line
load 𝑊 are variable.

The model discussed before can easily be adapted to calculate this ring load per specified radius.

The analytical predicted deflection and the actual deflection are listed in Table 5.1.
Table 5.1: Deflection comparison between analytical and FEM with a p𝑙𝑎𝑡𝑒 𝑡ℎ𝑖𝑐𝑘𝑛𝑒𝑠𝑠 = 3 𝑚𝑚

Back plate deflection

Analytical Measured
0,19 mm ~0,24 mm

As can be seen, there is a small deviation. This is normal because the practical case was simplified
due to some assumptions that had to be made. However, these numbers are still useful because it
gives a rudimentary idea of the deflection that will occur. For a better and more complete analysis,
3D FEM should be used.

For future designs the deflection was calculated for several plate thicknesses, the graph can be seen
in Figure 5.2.

First it should be noted that the first section of the curve, highlighted in red, should not be used. The
numbers obtained for these calculations are not correct because the deflection is bigger than one
half of the plate thickness. The backplates are simply not strong enough, there is a very high
possibility that plastic deformation occurs.

From this curve it can be seen that a plate thickness of about 2 mm would be the absolute minimal
thickness required with a deflection of about 0,65 mm. Going thinner would result in excessive
deformation. For the final design, a back plate of 3 mm is chosen, this will make for a strong
mechanical rotor.

62
Figure 5.2: Back plate deflection in function of back plate thickness

63
5.2 Motor topology

This motor is a double-sided machine with an internal PCB stator. For the motor, an outer rotor
PMSM topology was chosen. With this topology the shaft is stationary while the housing spins. This
design decision will probably lead to a motor with a smaller axial length. This is because the rotor
back plate itself, that is of course also needed in an inrunner topology, serves partly as the motor
housing, therefore a large piece of the housing can be eliminated.

However, there are also some challenges to overcome. The two rotor halves are held apart at the
outer diameter, consequently the stator can only be held on the inside diameter at the motor shaft.
This small support surface area leads to a large unsupported radial stick out of the PCB stator, this
could lead to excessive deflection. If too much deflection occurs, due to for example vibrations, the
airgap will have to be made bigger than necessary to avoid collision of the stator and rotor leading to
a weakened magnetic field.

Furthermore, the wiring cannot be routed via the outside of the motor but must go through the
hollow shaft.

64
5.3 Rotor design

The motor was designed using Siemens NX [17]. All the motor part drafts can be found in Appendix 1.

The full rotor design can be seen in Figure 5.3.

(1) (2) (4) (5) (6)

(7)
(3)

(𝑎) (𝑏) (𝑐)

Figure 5.3: Rotor design with exploded view; (a) Front; (c) Side; (1) Magnet retainer; (2) Magnets; (3) Steel back plate; (4)
Bearing; (5) Bearing mounting plate; (6) Mounting screws; (7) Washers and standoff spacers.

As discussed in chapter 4.3, the rotor will consist of a steel back plate that not only acts as a low
reluctance return path for the magnetic flux but will also be the mechanical support for the magnets.
The steel back plate has three lugs with holes at the outer diameter that are used for bolting the two
rotor halves together. The three remaining lugs have tapped holes to accept M3 bolts and are used
for assembling the motor, this will be discussed in chapter 8.3.

The two rotor halves should be placed as close to each other as possible without touching the stator
in order to get the strongest possible magnetic field. Because of the small axial distance between the
magnets, there is a strong attraction force between the two rotor halves. This exerts a large force on

65
the components separating them. Furthermore, for prototyping it is beneficial to make the airgap
easily adjustable.

To make the airgap adjustable and get a solid mechanical design, the rotor halves are held apart by
standard metal PCB standoffs. These are cheap and are available in many sizes. In order to make fine
adjustments to the airgap thickness, thin washers can be placed between the rotor halve and the
standoff.

Furthermore, because the steel back plate is only 3mm thick, there is not enough material to support
the bearings properly. Therefore, an additional part is 3D-printed with its main functionality being to
hold the bearings in place. For cooling purposes, ventilation holes were provided in the limited space
between the bearings and the steel back plate. These holes will substantially improve the cooling
ability of the motor [14].

The magnet retainer has 2 purposes. One is to make sure the magnets are properly aligned during
the gluing stage. A second one is to try and prevent the magnets from flying outwards due to large
centrifugal forces at higher rotational speeds if the glue should fail. It therefore serves as a secondary
safety feature.

One point to note is the importance of designing the back plate as such that there is as little as
overhang, i.e., unnecessary weight, as possible. This is very important because the back plate is one
of the components with the biggest impacts on weight alongside the magnets, therefore it has a
great impact on the power density of the motor. A detailed view of the final design of the back plate
can be seen in Figure 5.4. It is clear that the backplate has an unnecessary radial stick out that will
only partially contribute to the low reluctance return path for the magnetic flux. Still this outer ring
contributes to a weight gain in the full motor of 43,0 g.

Furthermore, there are three clamping lugs, these are very important to make for a strong and stable
mechanical design. The assembly lugs on the other hand serve no other purpose than making the
assembly easier. These assembly lugs for the full motor contribute to a weight gain of approximately
20,0 g.

These weights might seem unsubstantial, but for a motor that only weighs about 700 g, this means a
decrease in power density of about 8,5%. It is therefore important to try and reduce the amount of
material present in the back plate in order to optimise the power density.

66
 (𝑏)
(𝑐)

(𝑎)

(𝑐)

(𝑏)

(𝑏) (𝑐)

Figure 5.4: Backplate design with magnets; (a) Radial stick out; (b) Assembly lugs; (c) Clamping lugs.

67
5.4 Stator design

The stator is made up of a few components and can be seen in Figure 5.5.

(1) (2) (3) (4) (4) (6) (7) (1)

(5)

(𝑎) (𝑏) (𝑐)

Figure 5.5: Stator design with exploded view; (a) Front; (c) Side; (1) Bearing; (2) Aluminium spacer; (3) Aluminium flange
clamp; (4) Electrical insulator; (5) PCB stator; (6) Shaft with soldered aluminium flange; (7) Motor wiring

Because the full motor torque will be transmitted through the PCB stator, it is important that the PCB
is fixed well to the stationary shaft. Furthermore, it is needed that the clamping assembly has a small
axial length because it must fit in between the bearings and PCB, this limited space can be seen in
Figure 5.8. To do so, an aluminium shaft and flange were chosen. This way these two components
can be soldered in place to form a strong compact mechanical bond.

This flange is equipped with three holes for securing the PCB in place with M2 screws. Furthermore,
three holes are made for the electrical connection between the motor wiring and the PCB. This
connection can be seen in detail in Figure 5.6.

68
 Figure 5.6: PCB connection in detail

Because bare copper PCBs are used with no soldering mask and conductive aluminium flanges to
clamp it in place, it is necessary to use an insulator between the two components. This insulator
should be able to withstand higher temperatures because it is in direct contact with the PCB stator.
For the final design standard electrical tape was used because of its simplicity, compactness and
relatively high temperature rating.

The aluminium spacers make sure the PCB is positioned in the centre between the two bearings and
thus rotors. Furthermore, these spacers exert an outwards force on the inner race of the bearings
effectively providing a preload. In a new design this preload should ideally be done by using, for
example, wave washers. One of these spacers has cut outs for the electrical wiring to go through as
can be seen in Figure 5.6.

69
5.5 Assembled motor
The whole motor assembly can be seen in Figure 5.7.

(𝑎) (𝑏) (𝑐)

Figure 5.7: Assembled motor with exploded view; (a) Front; (c) Side

The two rotor halves and stator can be assembled separately. After this the motor can easily be
assembled by simply sliding the two rotor halves closer to each other along the stator shaft. This can
be done safely and gradually using the method mentioned in chapter 8.3.

This whole assembly is very compact with an axial length of about 29 mm. This length could be
reduced even more by bringing the bearings further inward. This can be seen in the sliced view in
Figure 5.8. This however was not done because if the distance between the two bearings would
decrease further, tilting stability would also decrease due to play and tolerances. Because of this, it
could be possible that the rotor tilts too much and partially collides with the PCB stator. Therefore, a
wider supporting area was chosen.

70
In this sliced view, the cooling path can also be seen. The air is sucked in via the holes in the bearing
mounting plates and runs along the PCB and magnets effectively cooling them. The air is then ejected
out via the outer diameter of the motor.

Figure 5.8: Sliced side view of assembled motor

In Figure 5.9 and Table 5.2 a few of the most important final dimensions of the motor are shown.

71
 (𝑒)

(𝑎)

(𝑓)

(𝑐)

(𝑑)
(𝑏)

Figure 5.9: Motor assembly dimensions

Table 5.2: Motor assembly dimensions

(𝑎) 𝐵𝑎𝑐𝑘 𝑝𝑙𝑎𝑡𝑒 𝑜𝑢𝑡𝑒𝑟 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 95,0 mm

(𝑏) 𝑀𝑜𝑡𝑜𝑟 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 118,0 mm
(𝑐) 𝐵𝑎𝑐𝑘 𝑝𝑙𝑎𝑡𝑒 𝑎𝑥𝑖𝑎𝑙 𝑙𝑒𝑛𝑔𝑡ℎ 18,8 mm
(𝑑) 𝑀𝑜𝑡𝑜𝑟 𝑎𝑥𝑖𝑎𝑙 𝑙𝑒𝑛𝑔ℎ𝑡 28,8 mm
(𝑒) 𝑇𝑜𝑡𝑎𝑙 𝑎𝑖𝑟𝑔𝑎𝑝 2,8 mm
(𝑓) 𝑆ℎ𝑎𝑓𝑡 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 10,0 mm

72
5.6 Test Bench

To test the motor properly, a test bench is designed. This bench will be used to measure the
generated back EMF and torque characteristic of the motor. A few extra components were made in
order to make sure the stator was floating and free to spin, this is needed for torque measurements.
The torque measurement arm is equipped with a clip that can slide along the length of the arm.
Furthermore, weights can be attached to this clip to adjust the torque accurately.

These parts can be seen in the exploded view in Figure 5.10, the full assembled test setup can be
seen in Figure 5.11.

(2) (3)
(1) (4) (4) (3) (2)
(5)

Figure 5.10: Exploded view test bench; (1) Torque measurement arm; (2) Bearing clamps; (3) Bearings; (4) Bearing holding
brackets; (5) DC motor.

Figure 5.11: Full motor test bench

73
 6 Motor Controller
The motor controller used in this thesis research is the “Hobbywing Skywalker 50A ESC UBEC”
controller. This controller is used to test the motor, it is easy to use and can be programmed for the
needed requirements. In this paragraph this motor controller will be further explained.

6.1 General

A motor controller is a device that can coordinate and guide the performance of an electric motor. It
does so by regulating the frequency of the supplied current to the windings. The name of the
controller encloses a few definitions that will be further explained. First of all, the abbreviation ESC
stands for electronic speed control, this means this controller is capable of controlling the motor
speed. Secondly, the term UBEC can be noticed, this abbreviation stands for universal battery
elimination circuit. This part of the controller can reduce battery voltages to the voltage needed for
the electronic circuit. In this thesis, this was not necessary because the controller was connected to a
5 V DC source. This would be interesting if this motor was used in, for example, an electric step, then
the battery of the step could feed the ESC directly [15].

The ESC motor controller converts a given PWM signal to the AC voltage needed by the motor. If the
PWM frequency changes the motor speed changes. This PWM was supplied by a PWM module that
uses a potentiometer to adjust the duty cycle of the PWM [16].

6.2 Specifications

The ‘Hobbywing skywalker ESC 50A UBEC’ has the following specifications:

• Continuous output: 50 A
• Burst output: 65 A for maximum 10 seconds.
• Input: 2-4S Lipo, 5-12 cells NIMH
• BEC: 5 A⁄5 V
• Max speed: 210000 rpm for 2 Poles BLM, 70000 rpm for 6 poles BLM, 35000 rpm for 12
poles BLM.

These specifications needed for these tests were calculated by the analytical model and shown in
Figure 6.1. The required current per phase is much lower than the maximum possible continuous
current by the controller.

74
Figure 6.1: Motor controller needed specifications with a track width = 2,06 mm

75
 7 Thermal Analyses
7.1 Topology

The motor can be represented as a composition of different layers with a cylindrical form. This opens
the possibility to thermally examine this structure like a wall. It is possible to calculate the resistances
related to the layers displayed in Figure 7.1, this figure is the simplified layer structure of the right
side of the motor. Table 7.1 contains the used parameters for each part. The contact area and
surfaces that are used, are extracted from the mechanical model made in Siemens NX [17]. The part
Air rad makes a bridge from the airgap between the PCB and the magnets to the surrounding air. This
part represents the ventilation due to air flowing through the airgap. Due to this extra ventilation the
motor will be cooler than when the airflow would be zero. The contact area for this region is the
cylindric surface that is wrapper around the layer air 1.

𝐴𝑖𝑟 rad

𝑃𝐿𝐴 1 𝐴𝑖𝑟 2

𝐵𝑎𝑐𝑘
A𝑖𝑟 1

𝐴𝑖𝑟 3
𝐼𝑟𝑜𝑛
𝑀𝑎𝑔 𝑃𝐿𝐴 2
𝑃𝐶𝐵

Figure 7.1: Simplified thermal structure of the motor

Table 7.1: Parameters for each layer

Part Thickness Contact surface [mm²] Diameter

[mm] [mm]
PCB 0,6000 7667,000 0,1000
Air 1 1,3500 7667,000 0,1000
Air rad 1,3500 424,1150 0,1000
PLA 1 5,0000 2442,0000 0,0950
Magnets 5,0000 3600,0000 0,0950
Back iron 3,0000 7186,0000 0,0950
PLA 2 5,0000 2570,0000 0,0615
Air 2 5,0000 4616,0000 0,0950
Air 3 𝑖𝑛𝑓𝑖𝑛𝑖𝑡𝑦 7186,0000 0,0950

76
7.2 Calculations

To calculate the thermic resistance of the layers there are two principles used. The first principle
used is the principle of conduction, this principle is used when the two layers are made of solid
materials. The second principle is the principle of convection, this principle is used when one of the
layer materials is a fluid.

The first principle uses this formula:

𝑙 K
𝑅= [ ] (7.1)
𝑘∗𝐴 W
Where 𝑙 is the thickness of the layer, 𝑘 is the thermal conductivity coefficient and 𝐴 is the contact
area with the previous layer.

The second principle uses a similar formula:

1 K
𝑅= [ ] (7.2)
ℎ∗𝐴 W
Where ℎ is the heat transfer coefficient for that layer and 𝐴 is the same as before mentioned, the
contact area with the previous layer.

The ℎ value depends on the air flow. In this case the speed is assumed to be equal to the rotational
speed of the motor. This will differ in each part because of the differentiation in diameters between
the parts. It is also possible that natural convection has an influence on the air movement, but these
factors are difficult to calculate analytically. According to Osczevski, the heat transfer coefficient can
be approximated by the next formula [18].

𝑈 0,6 𝑘𝑎𝑖𝑟 W
ℎ = 0,23 ∗ ( ) ∗ 0,4 [ ] (7.3)
𝜈 𝐷 m2 K
m2
𝜈 = 1,516 ∗ 10−5
s
W
𝑘𝑎𝑖𝑟 = 0,02514
mK
Where 𝑈 is the estimated speed of the air calculated by formula (7.4), 𝜈 is the kinematic viscosity of
air, 𝑘𝑎𝑖𝑟 is the thermal conductivity of air and 𝐷 is the diameter of that layer.

The speed of the air is equal to the next formula:

2𝜋 m
𝑈=𝑟∗ ∗ 𝑁𝑟𝑝𝑚 [ ] (7.4)
60 s
Where 𝑟 is the radius of each layer and 𝑁𝑟𝑝𝑚 is the rotation speed of the motor.

These variables are calculated for each layer and shown in Table 7.2 [19], [20], [21].

77
Table 7.2: Calculated variables for the thermal analysis

Part Air velocity Conductivity Heat transfer Resistance Heat

[m/s] coefficient coefficient [K/W] production
[W/mK] [W/mK] [W]
PCB 4,0000
Air 1 31,4159 89,5290 1,4569
Air rad 31,4159 89,5290 26,3362
PLA 1 0,1300 15,7500
Magnets 9,0000 0,1543
Back iron 54,0000 0,0077
PLA 2 0,1300 14,9656
Air 2 29,8451 88,6152 2,4447
Air 3 29,8451 88,6152 1,5704

To calculate the total resistance, the layers are represented with their different resistances in Figure
7.2.

Figure 7.2: Thermal resistance network

Out of this resistance network the total thermal resistance can be estimated. This network is a
simplification of the real thermal analysis. The resistances need to be added together the same as an
electrical network.
−1

1 1
𝑅𝑡𝑜𝑡𝑎𝑙 = 𝑅𝑎𝑖𝑟 1 + −1 −1 + + 𝑅𝑎𝑖𝑟 3 (7.5)
1 1 1 1 𝑅𝐴𝑖𝑟 𝑟𝑎𝑑
(𝑅 +𝑅 ) + 𝑅𝐵𝑎𝑐𝑘 𝑖𝑟𝑜𝑛 + (𝑅 +𝑅 )
𝑃𝐿𝐴 1 𝑀𝑎𝑔𝑛𝑒𝑡𝑠 𝐴𝑖𝑟 2 𝑃𝐿𝐴 2
( )
K
𝑅𝑡𝑜𝑡𝑎𝑙 = 5,11
W

The heat radiation coming from the PCB is assumed to be 4 W for each side. A second assumption is
made for the surrounding temperature this temperature is 20 °C. With this assumption it is possible
to determine the temperature in the centre of the motor applying formula (7.5).

78
 𝑇𝑖𝑛 − 𝑇𝑜𝑢𝑡
𝑄̇ = ⟺ 𝑇𝑖𝑛 = 𝑄̇ ∗ 𝑅𝑡𝑜𝑡𝑎𝑙 + 𝑇𝑜𝑢𝑡 (7.5)
𝑅𝑡𝑜𝑡𝑎𝑙
𝑇𝑖𝑛 = 40,44 °C

Where 𝑇𝑜𝑢𝑡 is the temperature of the surrounding air, 𝑇𝑖𝑛 is the temperature inside the motor and 𝑄̇
is the heat transmitted by one side of the PCB.

7.3 Conclusion

This analysis reassures that the heating will not lead to the deformation of the motor.

From analysing the model, some observations can be made:

• The temperature is significantly higher when the motor runs at low speed. This is normal
because the speed of the air is directly dependent of the running speed of the motor.
Therefore, the heat convection coefficient rises when the speed descends.
• The external influences have not been considered. If these were considered, this would only
result in better cooling. Thus, the temperature calculated by this model is the maximum
temperature possible.

A side note that needs to be taken is that this calculation is a strong simplification of the thermal
convections that really are happening within this motor. This temperature is not exact, but it gives a
good indication.

79
 8 Motor assembly
8.1 Assembly Rotor

The rotor consists of a steel back plate and can be seen in Figure 8.1. This plate has six lugs with holes
around its outer circumference. Three of these holes are for clamping the two halves together after
assembly. The three other holes are tapped for M3 bolts and are used for safely assembling the
motor, this will be discussed in chapter 8.3.

The magnets are glued to the backplate using cyanoacrylate (CA) glue. Furthermore, a magnet
retainer was used to accurately position the magnets and make the gluing easier. The magnet
retainer also contains recesses for a press fit of five 𝑀2,5 nuts. These nuts along with the bolts will
be used to clamp the bearing bracket, backplate and magnet retainer in place. A rotor halve with
glued magnets and magnet retainer can be seen in Figure 8.1.

The bearing brackets along with the bearings and bolts can be seen in Figure 8.2. Furthermore, the
bearing brackets are designed to be a press fit for the bearings. Venting holes are also provided along
the inner circumference.

One fully assembled rotor halve can be seen in Figure 8.3 and Figure 8.4.

(𝑓)
(𝑔)

(𝑑)

(𝑒)

(𝑐)
(𝑎) (𝑏)

Figure 8.1: Rotor halve; (a) Back plate backside; (b) Backplate with glued magnets; (c) M3 clamp hole; (d) M3 tapped
assembly hole; (e) Magnet; (f) Magnet retainer; (g) M2,5 nut

80
 (𝑏)
(𝑑)
(𝑐)
(𝑎)

(𝑏)
(𝑑)
(𝑐) (𝑎)

Figure 8.2: Rotor parts; (a) Rotor halve, (b) Bearing; (c) Bearing bracket; (d) M2,5 bolts

(𝑎) (𝑏)

Figure 8.3: Assembled rotor halves; (a) Back; (b) Front

81
Figure 8.4: Rotor angled view

82
8.2 Assembly Stator

The stator is a PCB, no solder mask is used to imitate a milled stator from bare copper as good as
possible. In the centre there are three through-holes for soldering the motor wires to. Furthermore,
four holes are made in order to clamp the PCB to the shaft flange with 𝑀2 bolts. The stator can be
seen in Figure 8.5.

(𝑐)

(𝑑)

(𝑎) (𝑏)

Figure 8.5: (a) PCB stator front; (b) PCB stator back; (c) Solder connection point; (d) Mounting hole

An aluminium flange is soldered to the motor shaft at the short end. This flange has four holes for
mounting 𝑀2 screws and three oversized holes in order not to make contact with the soldered
motor wires. From the other side, an aluminium clamping plate with corresponding holes is used to
clamp the PCB to the soldered flange. The motor wires go through the hollow shaft and through
holes lining up with the PCB connection holes. The shaft along with the flange, clamping plate and
wires can be seen in Figure 8.6.

(𝑒)
(𝑐)
(𝑓)

(𝑑)
(𝑎)

(ℎ) (𝑔)
(𝑏)

Figure 8.6: (a) Shaft; (b) Aluminium clamping plate; (c) Wire; (d) Mounting M2 nut; (e) Mounting M2 bolt; (f) Backside of
solder joint; (g) Soldered flange; (h) Flange solder connection

83
Furthermore, the PCB should be positioned exactly in the middle between both bearings, for this,
spacers are used. For the side with the wire connections, cut-outs are provided. The stator along with
the spacers can be seen in Figure 8.7.

The installed bearings can be seen in Figure 8.8.

(𝑓)

(𝑒)
(𝑑)
(𝑎)

(𝑏) (𝑐)

Figure 8.7: (a) PCB stator front view; (b) Stator with spacers top view; (c) PCB stator with spacers back view; (d) Spacer wire
side; (e) Spacer non-wire side; (f) Soldered flange

84
(𝑎)

(𝑏) (𝑐)

Figure 8.8: (a) Bearing installed wire side; (b) Bearing installed non-wire side; (c) Bearing installed top view

85
8.3 Assembly process

The strong attraction forces between the magnets makes the assembly process difficult. The distance
between the two rotor halves should gradually be reduced. This is done by fully inserting 𝑀3 bolts in
the tapped holes on the corresponding lugs at the outer circumference of the rotor halves. After this,
the rotor halves can safely be placed, by hand, in the right position until the 𝑀3 bolts rest on each
other. The rotor spacers, consisting of PCB standoffs, can be installed on one side. After this, the
bolts should gradually and evenly be retracted until the airgap gets close to its final dimension. Then
the standoffs can be connected from the other side with bolts and thin washers can be added for fine
spacing adjustments. The setup and assembly process can be seen in Figure 8.9.

(𝑎)
(𝑏)
(𝑏)
(𝑒)

(𝑏)

(𝑑)

(𝑎)
(𝑎)

(𝑏)

(𝑐)

Figure 8.9: Assembly process; (a) Assembly bolts; (b) Clamping bolt; (c) Rotor spacer; (d) Washer fine adjustment spacer; (e)
PCB standoff spacer

86
The fully assembled motor can be seen in Figure 8.10. One thing to note is that the 𝑀2,5 bolts, that
clamp the bearing brackets to the backplate, are not yet installed. This is because there are still two
more pieces that will have to be attached to those bolt holes for the torque test setup. If the motor
should not be used in the test setup, the 𝑀2,5 bolts can be installed. It is important that these bolts
are the correct length otherwise, if they are too long, they can interfere with the PCB stator. The
airgap can also be seen clearly in Figure 8.10. Because of the method that is used, the airgap can be
easily adjusted and the whole rotor assembly is very stable. The PCB stator however is the limiting
factor because it is very thin, flexible and has a large stick-out, this compromises the ability to make
the airgap even smaller.

Lastly, the final weight of the assembled motor is 708 g.

(𝑒)

(𝑑)
(𝑎)

(𝑏) (𝑐)

Figure 8.10: Final assembled motor; (a) Non-wire side; (b) Wire side; (c) Side view; (d) Holes for M2,5 bearing bracket bolts;
(e) PCB stator

87
8.4 Test setup

To test the motor, a setup is made. This setup makes sure the motor and secondary DC motor are
aligned properly and held firmly in place. For measuring the torque, the stator should be able to
rotate. For this purpose, two extra components and bearings are needed. These components directly
bolt onto the bearing brackets on the existing motor and can be seen in Figure 8.11.

In order to connect the motor with the secondary DC motor, a rotating shaft is needed. The motor
shaft in this design however is stationary. Therefore, a secondary shaft is press fitted in one of the
test setups brackets. This shaft will then rotate along with the motor rotors.

(𝑔)
(𝑓)

(𝑒)

(𝑎) (𝑏)

(𝑑)

(𝑐)

Figure 8.11: (a) Motor non-wire side; (b) Test setup bracket; (c) Motor wire side; (d) Test setup bracket; (e) M2,5 clamping
bolt; (f) Test setup bearing; (g) Press fitted shaft

88
With the test setup brackets installed, the motor can be held in place in a large brace. This can be
done by using an extra holding bracket on both sides that fits in the inside of the large bearing and
can be bolted to the large brace. The holding bracket and brace can be seen in Figure 8.12.

(𝑎)

(𝑏)

(𝑐)

Figure 8.12: (a) Brace; (b) Motor; (c) Holding bracket

Furthermore, a torque arm is clamped to the stationary shaft that is connected to the motor stator.
This arm is equipped with a counterbalance plate and torque measurement weight. The
counterbalance weight balances the torque measurement arm. When the torque measurement
weight is removed, weight is added to the counterbalance plate until the arm is perfectly horizontal.
This way only the torque weight should be accounted for in the torque calculations. The torque
measurement weight can move along the arm to adjust the torque required to lift the arm to a
horizontal position. The whole testing setup can be seen in Figure 8.13.

89
 (𝑐)

(𝑏)
(𝑑)

(𝑎)

(𝑒)

(𝑐)
(𝑏)

(𝑓)
(𝑎)

Figure 8.13: (a) Torque weight; (b) Torque measurement arm; (c) Counterbalance plate; (d) Rotating shaft; (e) Holding
bracket for secondary DC motor; (f) Stationary shaft

90
8.5 Failures and improvements

The first model used a plastic housing that held the two rotor halves apart. As discussed in chapter
4.3.2, very large magnetic attraction forces act on the rotor halves. The plastic parts were not strong
enough and broke. The metal design with the PCB standoff discussed in the previous sections was
simple yet highly effective in obtaining an accurate airgap with consistent thickness.

For newer designs, a better alignment of the bearings and making sure the rotor halves are
perpendicular to the shaft should drastically reduce the vibrations of the motor. Also, a thicker and
stiffer PCB should be used. Because the used PCB is so thin and flexible, the airgap still must be
relatively large. If a slightly thicker PCB with more copper is used, the phase resistance should
decrease causing les copper losses without having to make to airgap larger.

Furthermore, a better design for connecting the phase wires to the PCB should be designed. The
motor wires had to make quite sharp 90° bends without touching the aluminium. After this the wires
had to be soldered in place. This was a difficult process that could use a lot of improvement.

Another major improvement would be to use better glue for the magnets. During testing a magnet
came loose because the glue failed, probably due to overheating. Furthermore, the magnet came
loose from the backplate and clamped the PCB stator in between the opposite rotor magnet. The
motor abruptly stopped. A better glue that is more heat resistant should be used in newer designs.

91
 9 Prototype testing
9.1 Back EMF testing

The test setup can be seen in Figure 9.1 and Figure 9.2.

Figure 9.1: Back EMF test setup; (a) Secondary drive motor; (b) Motor under test; (c) Motor coupling; (d) Oscilloscope
measuring back EMF line-line voltage.

Figure 9.2: Back EMF test setup speed measurement; (a) Tachometer

92
To compare the generated back EMF, the motor is driven by a secondary DC motor. Two of the
motor phases from the motor under test are connected to an oscilloscope in order to measure the
generated line-line voltage back EMF. The motor is tested at different speeds going from 454 rpm up
to 2877 rpm with intervals of roughly 500 rpm. The speed is measured with a tachometer. The
highest testing speed was limited by the secondary DC motor. Higher speeds are not necessary
because there is a linear relation between the generated back EMF and motor speed. This is given by
the velocity constant 𝐾𝑣 , which is the inverse of the back EMF constant 𝐾𝑒 . The back EMF constant
for PM AC motors is defined as the ratio of the peak line-line voltage to the rotor speed [rad⁄s] [22].
The back EMF constant and velocity constant are determined for every measured speed and are
listed in Table 9.1. It is clear that there is a linear relation, the motor velocity constant 𝐾𝑣 is
rad
determined as the mean of these values and is 32,32 Vs
.

Table 9.1: Back EMF and velocity constant

454 rpm 1009 rpm 1559 rpm 2045 rpm 2514 rpm 2877 rpm
𝑉𝑝𝑒𝑎𝑘 [V] 1.47 3,26 5,08 6,56 8,18 9,35
𝑆𝑝𝑒𝑒𝑑 [rad⁄s] 47,54 105,66 163,26 214,15 263,27 301,28
𝐾𝑒 [V s⁄rad] 0,031 0,031 0,031 0,031 0,031 0,031
𝐾𝑣 [rad⁄V s] 32,34 32,41 32,14 32,65 32,17 32,22

A comparison between the generated back EMF line-line voltage and the results from the analytical
model can be seen in Figure 9.3. It is clear that the analytical model is accurate, only a small deviation
of maximum 8% is present. This deviation can be the cause of several factors like influence of
temperature on various parameters, difference in airgap, inaccuracy of the magnet remanence or
saturation of the back plate.

93
Figure 9.3: Comparison of measured back EMF with analytical model at different speeds

94
9.2 Torque and efficiency testing

The test setup for the torque and efficiency testing can be seen in Figure 9.4 and Figure 9.5.

Figure 9.4: Torque/efficiency test setup; (a) Motor power supply; (b) Power supply output current measurement; (c) Motor
under test; (d) Secondary motor used as generator/brake; (e) Adjustable torque measurement arm; (f) Tachometer; (g)
Generator load; (h) Load current measurement; (i) Motor controller ESC.

(𝑎) (𝑏)

(𝑐)

Figure 9.5: Motor torque measurement setup; (a) Counterweight for torque measurement arm; (b) Torque measurement
arm; (c) Torque measurement weight

95
To measure the torque, a measurement arm is connected to the stator shaft, this component can be
seen in Figure 9.5. In order to make an accurate measurement, the weight of the measurement arm
is counterbalanced using counterweights. Washers are added to the counterweight plate until the
measurement arm is balanced and could maintain a horizontal position without support. Because of
this, only the torque measurement weight contributes to the torque. The torque weight is 38,0 g.

Furthermore, the motor is controlled by an ESC, this ESC is powered by a power supply. To calculate
the motor efficiency, the ESC input power 𝑃𝑖𝑛,𝐸𝑆𝐶 is measured by measuring the output current and
voltage of the power supply. The mechanical motor output power 𝑃𝑀 is calculated by determining
the motor speed using a tachometer and measuring the motor torque by adjusting the torque weight
along the torque measurement arm until the arm is perfectly horizontal. The first method for raising
or lowering the output power of the motor was by simply adjusting the resistance and thus current
at the generator. More resistance means less current and less motor power, less resistance means
more current and thus more motor power. The system efficiency is then simply
𝑃𝑀
𝜂𝑠 = (9.1)
𝑃𝑖𝑛,𝐸𝑆𝐶

Because there are only a few resistors, the motor torque must be adjusted by moving the torque
weight along the torque measurement arm. This method is quite cumbersome and difficult to get an
accurate alignment. Therefore, for the second method an adjustable load for the generator was
made. The second methods schematics is shown in Figure 9.6. This consist of an IGBT that is
connected between the generator terminal clamps. The IGBT gate is driven by a PWM signal from an
Arduino. Because of this, the generator load can be adjusted from 0% (no gate signal, generator
connectors are open) to 100% (constant gate signal, generator is shorted). Using this method, the
motor torque can be kept constant by adjusting the PWM signal so that the torque measurement
arm is aligned at its calibrated height. This way the motor can be tested quite fast at a constant
torque at different speeds.

Figure 9.6: Schematic for an adjustable load for the torque/efficiency test [23]

96
The measurement speed range is limited. The lower limit is determined by the IGBT. If the speed of
the motor is low, the generated back EMF is too small to be able to generate enough current and
thus torque. This could be due to a high resistance of the circuit. The upper limit is determined by the
motor itself, at speeds exceeding 4000 rpm the motor vibrations get too high and the risk of
breaking the motor increases. The motor torque is also limited by the generator’s torque abilities.

It should be noted that this setup and way of measuring is not entirely correct. Because the input
power is determined at the power supply, also the ESC’s efficiency is measured. To measure the
efficiency of the motor only, a 3-phase power analyser should be used at the motor connection.
Furthermore, large bearings were used (as can be seen in Figure 9.9) so that the motor stator can
rotate freely, these bearings however are not part of the actual motor and contribute to an extra
power loss which will affect the motor efficiency negatively. The aforementioned points will be
discussed in chapter 9.2.1.

The measured values at the different operating states for both methods can be seen in Table 9.2 and
Table 9.3.
Table 9.2: Torque/efficiency measurement data using resistors

𝑇𝑜𝑟𝑞𝑢𝑒 𝜔 [rpm] 𝑃𝑀 [W] 𝑉𝑜𝑙𝑡𝑎𝑔𝑒 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝑃𝑖𝑛,𝐸𝑆𝐶 [W] 𝐸𝑓𝑓𝑐𝑖𝑒𝑛𝑐𝑦

[Nmm] [V] [A] 𝜂𝑠 [%]
𝑇𝑒𝑠𝑡 1 47,12 1915 9,45 11,0 2,26 24,92 37,9
𝑇𝑒𝑠𝑡 2 49,15 2154 11,09 11,0 2,47 27,17 40,8
𝑇𝑒𝑠𝑡 3 42,96 2465 11,09 11,0 2,31 25,41 43,6
𝑇𝑒𝑠𝑡 4 43,79 2633 12,07 11,0 2,39 26,29 45,9
𝑇𝑒𝑠𝑡 5 46,23 2909 14,08 13,2 2,52 33,26 42,3
𝑇𝑒𝑠𝑡 6 51,20 3180 17,05 13,2 2,76 36,43 46,8
𝑇𝑒𝑠𝑡 7 61,79 3397 21,98 14,6 3,05 44,53 49,4
𝑇𝑒𝑠𝑡 8 57,24 3811 22,85 17 2,94 49,98 45,7

Table 9.3: Torque/efficiency measurement data using PWM adjustable motor load

𝑇𝑜𝑟𝑞𝑢𝑒 𝜔 [rpm] 𝑃𝑀 [W] 𝑉𝑜𝑙𝑡𝑎𝑔𝑒 𝐶𝑢𝑟𝑟𝑒𝑛𝑡 𝑃𝑒 [W] 𝐸𝑓𝑓𝑐𝑖𝑒𝑛𝑐𝑦

[Nmm] [V] [A] 𝜂 [%]
𝑇𝑒𝑠𝑡 9 43,41 2490 11,32 13,4 2,11 28,27 40,0
𝑇𝑒𝑠𝑡 10 43,41 2779 12,63 13,4 2,16 28,94 43,7
𝑇𝑒𝑠𝑡 11 43,41 3001 13,64 13,4 2,22 29,75 45,9
𝑇𝑒𝑠𝑡 12 43,41 3197 14,53 13,4 2,28 30,55 47,6
𝑇𝑒𝑠𝑡 13 51,44 3125 16,83 13,4 2,58 34,57 48,7
𝑇𝑒𝑠𝑡 14 51,44 3327 17,92 13,5 2,64 35,64 50,3
𝑇𝑒𝑠𝑡 15 51,44 3407 18,35 14,4 2,66 38,30 47,9
𝑇𝑒𝑠𝑡 16 34,46 2303 8,31 12,7 1,79 22,73 36,6
𝑇𝑒𝑠𝑡 17 34,46 2333 8,42 12,7 1,79 22,73 37,0
𝑇𝑒𝑠𝑡 18 61,44 3998 25,72 16,3 3,12 50,86 50,6

The motor efficiency is quite low at about maximum 50%. This is partly because of the ESC and
losses due to the large bearings, these losses should not be accounted for in the efficiency calculation
and will be approximated in the following chapter to get a more accurate result.

97
9.2.1 Extra losses beside motor

As discusses previously there are a few losses that are also measured but are not due to the motor.
An approximation of these losses will be discussed in this chapter. In Figure 9.7, a schematic
representation with the extra losses that will be accounted for is given.

𝑃𝑖𝑛,𝐸𝑆𝐶 𝑃𝑖𝑛,𝑚𝑜𝑡𝑜𝑟

𝑃𝐸𝑆𝐶,𝑙𝑜𝑠𝑠 𝑃𝑏𝑒𝑎𝑟𝑖𝑛𝑔

Figure 9.7: Schematic representation with extra losses; PSU 3D model adapted from [24]; ESC 3D model adapted from [25]

9.2.1.1 ESC efficiency

An ESC is the link between the DC power supply and 3-phase AC motor phases. This conversion is
achieved with high frequency switching leading to considerable power losses. The motor is driven by
a “Hobbywing Skywalker 50A ESC”. A datasheet of this model is available, but no information about
the losses can be found, therefore an approximation should be made. Some research has been done
regarding ESC efficiency in function of input current, DC bus voltage and ESC brand [26]. An efficiency
plot in function of input current from an ESC with similar parameters can be seen in Figure 9.8. From
this graph it is possible to extract the approximated ESC efficiencies 𝜂𝐸𝑆𝐶 at different working states.

98
 Figure 9.8: Efficiency in function of input current and DC bus voltage of Aerostar 50A ESC [26]

The approximated efficiency of the ESC for every working state (in function of input voltage and input
current) is determined using Figure 9.8 and is listed in Table 9.4. Furthermore, the approximated
input power that the motor receives after the ESC power loss is determined using the following
formula

(𝑃𝑖𝑛,𝑚𝑜𝑡𝑜𝑟 + 𝑃𝑏𝑒𝑎𝑟𝑖𝑛𝑔 ) = 𝜂𝐸𝑆𝐶 ∗ 𝑃𝑖𝑛,𝐸𝑆𝐶 (9.2)

Where 𝑃𝑖𝑛,𝑚𝑜𝑡𝑜𝑟 is the true electrical motor input power, 𝑃𝑏𝑒𝑎𝑟𝑖𝑛𝑔 are the power losses due to the
large bearings and 𝑃𝑖𝑛,𝐸𝑆𝐶 is the electrical power input at the ESC.

99
Table 9.4: ESC efficiencies at different working states

𝑃𝑖𝑛,𝐸𝑆𝐶 [W] 𝜂𝐸𝑆𝐶 [%] 𝑃𝑖𝑛,𝑚𝑜𝑡𝑜𝑟 + 𝑃𝑏𝑒𝑎𝑟𝑖𝑛𝑔 [W]

𝑇𝑒𝑠𝑡 1 24,92 79,0 19,69
𝑇𝑒𝑠𝑡 2 27,17 80,0 21,74
𝑇𝑒𝑠𝑡 3 25,41 79,5 20,20
𝑇𝑒𝑠𝑡 4 26,29 79,5 20,90
𝑇𝑒𝑠𝑡 5 33,26 79,0 26,28
𝑇𝑒𝑠𝑡 6 36,43 79,5 28,96
𝑇𝑒𝑠𝑡 7 44,53 81,0 36,07
𝑇𝑒𝑠𝑡 8 49,98 78,0 38,98
𝑇𝑒𝑠𝑡 9 28,27 77,5 21,91
𝑇𝑒𝑠𝑡 10 28,94 78,0 22,57
𝑇𝑒𝑠𝑡 11 29,75 78,0 23,21
𝑇𝑒𝑠𝑡 12 30,55 78,5 23,98
𝑇𝑒𝑠𝑡 13 34,57 79,0 27,31
𝑇𝑒𝑠𝑡 14 35,64 79,0 28,16
𝑇𝑒𝑠𝑡 15 38,30 78,5 30,07
𝑇𝑒𝑠𝑡 16 22,73 76,5 17,39
𝑇𝑒𝑠𝑡 17 22,73 76,5 17,39
𝑇𝑒𝑠𝑡 18 50,86 79,0 40,18

9.2.1.2 Large motor setup bearings

As discussed in chapter 5.6, for torque testing the whole motor will be placed on 2 large bearings
attached to the rotor in order to make the stator floating. The bearings can be seen in Figure 9.9
denoted with the number (3).

(2) (3)
(1) (4) (4) (3) (2)
(5)

Figure 9.9: Exploded view test bench; (1) Torque measurement arm; (2) Bearing clamps; (3) Bearings; (4) Bearing holding
brackets; (5) DC motor.

These bearings are not part of the actual motor and should thus not be considered in the efficiency
calculations. Therefore, an approximation of the bearing losses is made. However, because bearing
losses are very complex due to the many factors influencing the frictional moment, it should be

100
noted that these calculations only give a very rough estimation. Still this method is used in order to
get an idea of the magnitude and thus influence of these components. Calculations for the bearing
losses are made using SKF’s frictional moment calculator [27]. The total frictional moment consists of
a few frictional causes, these are

• Rolling frictional moment 𝑀𝑟𝑟

• Sliding frictional moment 𝑀𝑠𝑙
• Frictional moment from seals 𝑀𝑠𝑒𝑎𝑙
• Frictional moment from oil bath effects 𝑀𝑑𝑟𝑎𝑔

For this application only the rolling and sliding frictional moments are of importance because the
bearings do not have seals, only shields and no oil bath is used.

In the following section only the method for calculating the losses is discussed, however because this
is quite an extensive calculation, the calculations itself along with the used tables and parameters are
not shown in this chapter but are added in Appendix 2.

Rolling frictional moment 𝑴𝒓𝒓

The rolling frictional moment can be calculated using

𝑀𝑟𝑟 = Φ𝑖𝑠ℎ Φ𝑟𝑠 𝐺𝑟𝑟 (𝜈𝑛)0,6 (9.3)

Where 𝑀𝑟𝑟 is the rolling frictional moment, Φ𝑖𝑠ℎ is the shear heating reduction factor, Φ𝑟𝑠 is the
kinematic replenishment/starvation reduction factor, 𝐺𝑟𝑟 is a variable depending on the bearing type
and load, 𝜈 is the actual operating viscosity of the base oil of the grease and 𝑛 is the rotational speed.

Furthermore Φ𝑖𝑠ℎ can be determined using the following formula

1
Φ𝑖𝑠ℎ = (9.4)
1 + 1,84 ∗ 10−9 (𝑛𝑑𝑚 )1,28 𝜈 0,64
Where 𝑑𝑚 is the bearings mean diameter.

The kinematic replenishment/starvation reduction factor Φ𝑟𝑠 can be calculated using

1
Φ𝑟𝑠 = (9.5)
𝐾𝑧
(𝐾𝑟𝑠 𝜈 𝑛 (𝑑+𝐷)√ )
2(𝐷−𝑑)
𝑒
Where 𝐾𝑟𝑠 is the replenishment/starvation constant, 𝐾𝑧 is a bearing type related geometry constant,
𝑑 is the bearing bore diameter and 𝐷 is the bearing outside diameter.

Sliding frictional moment 𝑴𝒔𝒍

The sliding frictional moment can be calculated using

𝑀𝑠𝑙 = G𝑠𝑙 µ𝑠𝑙 (9.6)

Where 𝑀𝑠𝑙 is the sliding frictional moment, 𝐺𝑠𝑙 is a variable depending on the bearing type and load
and µ𝑠𝑙 is the sliding friction coefficient.

Furthermore, µ𝑠𝑙 can be determined using

101
 µ𝑠𝑙 = Φ𝑏𝑙 µ𝑏𝑙 + (1 − Φ𝑏𝑙 ) µ𝐸𝐻𝐿 (9.7)

Where µ𝑏𝑙 is a coefficient that depends on the type of movement, µ𝐸𝐻𝐿 is the sliding friction
coefficient in full-film conditions and Φ𝑏𝑙 is the weighting factor for the sliding friction coefficient.

Φ𝑏𝑙 can be determined using the following formula

1
Φ𝑏𝑙 = −8 1,4 (9.8)
𝑒 2,6∗10 (𝑛 𝜈) 𝑑𝑚
The total frictional moment 𝑀𝑏 is then given by
𝑀𝑏 = 𝑀𝑟𝑟 + 𝑀𝑠𝑙 + 𝑀𝑠𝑒𝑎𝑙 + 𝑀𝑑𝑟𝑎𝑔 = 𝑀𝑟𝑟 + 𝑀𝑠𝑙 (9.10)

At a working temperature of 40 °C and a rotational speed of 4000 rpm, the bearing frictional
moment is merely 0,44 Nmm. For both bearings this attains to a power loss of 0,37 W. This is almost
negligible, still this loss is accounted for in the efficiency calculations. The motor input power
𝑃𝑖𝑛,𝑚𝑜𝑡𝑜𝑟 is listed in Table 9.5 along with the bearing losses 𝑃𝑏𝑒𝑎𝑟𝑖𝑛𝑔 and the true motor efficiency
𝜂𝑚 for the different testing states. 𝜂𝑚 is determined by
𝑃𝑀
𝜂𝑚 = (9.11)
𝑃𝑖𝑛,𝑚𝑜𝑡𝑜𝑟
Table 9.5: Bearing power loss and true motor efficiency.

𝑃𝑏𝑒𝑎𝑟𝑖𝑛𝑔 [mW] 𝑃𝑖𝑛,𝑚𝑜𝑡𝑜𝑟 [W] 𝑃𝑀 [W] 𝜂𝑚 [%]

𝑇𝑒𝑠𝑡 1 123,4 19,57 9,45 48,3
𝑇𝑒𝑠𝑡 2 147,7 21,59 11,09 51,4
𝑇𝑒𝑠𝑡 3 181,1 20,02 11,09 55,4
𝑇𝑒𝑠𝑡 4 200,2 20,70 12,07 58,3
𝑇𝑒𝑠𝑡 5 232,5 26,05 14,08 54,0
𝑇𝑒𝑠𝑡 6 265,5 28,69 17,05 59,4
𝑇𝑒𝑠𝑡 7 292,8 35,78 21,98 61,4
𝑇𝑒𝑠𝑡 8 346,7 38,63 22,85 59,2
𝑇𝑒𝑠𝑡 9 184,0 21,73 11,32 52,1
𝑇𝑒𝑠𝑡 10 217,1 22,35 12,63 56,5
𝑇𝑒𝑠𝑡 11 243,6 22,97 13,64 59,4
𝑇𝑒𝑠𝑡 12 267,6 23,71 14,53 61,3
𝑇𝑒𝑠𝑡 13 258,7 27,05 16,83 62,2
𝑇𝑒𝑠𝑡 14 283,9 27,88 17,92 64,3
𝑇𝑒𝑠𝑡 15 294,0 29,78 18,35 61,6
𝑇𝑒𝑠𝑡 16 163,5 17,23 8,31 48,2
𝑇𝑒𝑠𝑡 17 166,7 17,22 8,42 48,9
𝑇𝑒𝑠𝑡 18 371,8 39,81 25,72 64,6

As can be seen from the calculations, the motor efficiency is quite low. Some losses occur in the
motor like copper losses, windage losses and some high frequency losses. Furthermore, while
testing, the motor was vibrating a lot, this could be due to the less-than-optimal alignment of the
motor. The inner bearings of the motor itself were not perfectly aligned with the larger outer
bearings and the rotor plates are not perfectly perpendicular to the shaft causing more vibrations. An
optimization of the mechanical build could reduce these effects and increase motor efficiency.
Getting a perfect alignment with 3D-prints alone is extremely hard as the slightest deviation has
major impact on vibration, to make the motor run more true, a lathe should be used. If the vibrations

102
could be reduced, the motor should be able to reach much higher speeds. Not many testing points
were taken due to some limitations in the testing rig. Furthermore, the torque measurement arm
was shaking a lot due to the motor vibrations and large stick out, because of this it was difficult to get
an accurate torque reading. Another factor to consider is the motor connection wires, these are
directly connected to the stator and the ESC, partly retaining the torque measurement arm. This has
a small effect on the motor torque measurement but will undoubtedly influence it. This however is
very hard to account for in the calculations.

A motor efficiency map gives great insight in the capabilities of the motor. However, because the test
setup had many limitations, it was not possible to obtain enough data points in order to generate a
useful efficiency map.

To get a better understanding of the motor efficiency, a more in-depth analysis of the possible motor
losses should be done. This way weak points can be determined.

103
 10 Conclusion
The design and build of an axial flux machine with PCB stator were done in this thesis. First, an
introduction to the axial flux machine was given and why researching this type of motor is important.
Furthermore, a short summary and overview of the possible PCB winding designs was given along
with a few possible applications for this motor.

In the second part of this thesis, an electromagnetic analytical model was made in MATLAB. This part
explains in detail how the magnetic flux density field, flux linkage and thus back EMF can be
determined along with the produced torque for a given set of input parameters. Functions for
calculating the electrical phase resistance and the maximum possible number of turns per winding
were also made. Using these functions and models, the influence of certain design choices can be
predicted leading to a more optimal design and thus better motor performance.

In the following part of the thesis the analytical model is used to determine the most optimal
magnets and magnet configuration. Four main topologies and magnet types were considered by
comparing the surface area coverage ratio and produced back EMF per weight of the magnets. The
2nd configuration had slightly lower performance than the 4th configuration due to the lower
magnetic grade but was still chosen because of the large price difference.

Furthermore, the PCB winding design was analysed in depth. For the winding shape a concentrated
trapezoidal concentric winding was chosen due to its large generated back EMF and good torque
production characteristics compared to the wave winding type. Also, the influence of the geometrical
parameters was studied. This analysis led to the conclusion that the inner and outer winding radii
should be equal to or slightly larger than the magnet inner and outer radii in order to get the highest
possible motor power. Furthermore, the track width was considered. From the analysis it is clear that
wider tracks are beneficial. This is because with decreasing track width and thus increasing number
of turns, the total amount of winding spacing also increases leading to less copper on the PCB. But
due to some practical reasons, a compromise between number of turns and manufacturability was
made. Lastly, the influence of the backplate thickness was studied. A thicker backplate results in less
saturation and thus a stronger magnetic field. However, because this plate is made of steel, weight
also increases drastically with thickness. A backplate thickness of 3 mm was chosen leading to an
optimal motor power density.

In the following chapter, the mechanical design was discussed. First, the strength requirements for
the backplate were determined. According to this analysis, very large magnetic attraction forces are
present. Because of this, a strong mechanical design for separating the rotor halves is needed. Using
PCB standoffs with washers for fine height adjustment yielded very good results. Furthermore, a
minimum thickness of the back plate is necessary in order to keep the deflection of the plate within
limits. Also, a full 3D-model of this motor was made using Siemens NX. The practical design choices,
regarding for example manufacturability, were also discussed.

In the next chapter, a thermal analysis of the motor was done. A lumped parameter model was used
in order to simplify the calculations. From this study, it is clear that the motor will stay relatively cool
when a heat radiation due to copper loss of 4 W for each side of the PCB is present.

The motor was also built. A first model used plastic parts to separate the rotor halves. However,
these plastic parts could not withstand the large magnetic forces. It is also particularly important to
develop a good method for assembling and disassembling the rotor halves. For this design, six M3
bolts were used along with tapped holes in the rotor. This proved to be an effective method. The

104
magnets were glued using regular CA glue. During testing some magnets came loose, probably due to
overheating of the glue. For further prototypes, a glue with better thermal resilience should be used.

In the following chapter, the motor was tested, and the results were compared with the results from
the analytical model. The generated back EMF was measured at different speeds. These were
compared with the determined generated back EMFs from the analytical model. The comparison
showed that the model was very accurate. During these tests, the motor vibrations were quite high.
Because of this, the maximum testing speed was limited. Also, the torque production and efficiency
of the motor is determined. The test setup was not ideal because also the efficiency of the ESC was
measured. Because of this, the ESC efficiency had to be approximated in order to get more accurate
results. The motor efficiency is relatively low due to several factors. Firstly, because several parts had
to be 3D printed, alignment of the bearings and rotor halves was less than optimal causing a lot of
vibrations. Also, the very thin copper tracks cause a lot of copper losses. Furthermore, windage
losses occur. Besides the losses, also the test setup was not ideal. Measuring the torque was very
difficult and inaccurate. In order to get a better idea about the true characteristics of the motor, a
better test setup should be designed.

105
 11 Future work
11.1 Wave versus concentric windings

11.1.1 Electromechanical performance

Many winding topologies are possible, for this motor a concentric trapezoidal shape was chosen. The
concentric windings offer a high back-EMF with good torque characteristics but also have a large
phase resistance. This leads to higher copper losses and thus excessive heating in the PCB occurs. A
few different winding topologies have been suggested and studied leading to a lower phase
resistance and thus less heating of the PCB [6]. An in-depth comparison between the wave type
windings and concentric windings regarding back EMF, electromagnetic torque and copper losses
should lead to a more optimal winding topology choice, possibly increasing motor performance.

11.1.2 Manufacturability

It could also be beneficial to study the possibility of milling the concentric stator coils from a blank
copper sheet. This way the stator could be manufactured with thicker copper and thus effectively
decreasing copper losses. This however has its own difficulties. In order to optimize the available
stator surface area, the copper tracks itself should be as wide as possible in relation to the track
spacing. This means that a very fine mill should be used which in turn makes manufacturing harder.
New manufacturing methods and winding topologies could be explored to improve the
manufacturability. Two of these methods could be punching and milling. The punching method could
be easily used for mass production.

11.1.3 Milling

Milling is a very commonly known process in the metal industry. This method is used to shape out
metal forms of high precision. It could also be used to mill out paths out of metal sheets. The mills
used on these milling machines can be very small so the metal waste would be as little as possible.
The advantage of milling the copper paths would be that the copper paths can be a lot thicker. When
the thickness of the copper paths increases so does the torque. However, because the copper paths
are thicker the weight and airgap of the motor increases accordingly. Therefore, a careful
examination of the power density should be done. This process could be further investigated to
determine if this is advantageous for the motor’s efficiencies.

11.1.4 Punching

Punching is a way to fabricate metal tools or forms. This process uses a punching press that removes
the surplus by punching a hardened steel tip through the metal sheet. It is a quick process that could
be used for mass production. This process can be used for each winding, nevertheless there are some
windings more plausible for this production process. The wave windings are the easiest to make from
a metal sheet, it is always the same shape that needs to be punched out. Next to the advantage of
mass production possibilities there is a second advantage. The second advantage is that the paths

106
could be made thicker than using PCB printed paths. This production method must be further
investigated, to define the possibilities within axial flux motors.

11.2 Wedge shaped magnets

Rectangular shaped magnets are an economical option compared to wedge type magnets for
prototyping. However, for large quantities when manufacturing, the price will only depend on the
magnet weight. The rectangular magnets lead to a decrease in the magnetic flux in the airgap due to
the non-optimal use of the rotor surface. Therefore, the analytical model could be adapted to
generate the magnetic field of these wedge type magnets and afterwards compare the power
densities generated by both types of rotors.

11.3 Back iron saturation

As stated in chapter 4.3.1 the steel back plate saturates lowering the magnetic field strength. This
effect is only accounted for in the analytical model using an arbitrary method. This method is time
consuming and rather cumbersome. A better analysis and modelling of the back iron saturation will
make for a more accurate model and allow for a more in-depth optimisation of the motor.

11.4 Magnet modelling

The analytical model does not give an accurate representation of the magnetic flux density field in
the radial direction. As can be seen in Figure 11.1 the magnetic field is non-existent just outside the
magnetic boundaries in the radial direction. This is not correct as the magnetic field flares out along
the radial direction. This inaccuracy could influence the winding dimension choice and performance
predictions. Therefore, a correction function could be used to get a quasi-3D electromagnetic model
[28]. However, for this method a 3D FEM analysis must be made in order to determine a few
parameters for this correction function.

Furthermore, because a polar system is used, the rectangular magnets are rounded off. This leads to
an approximated magnetic field that is not an entirely correct representation of the real magnetic
field. Further improvement of the analytical model could try and factor in this inaccuracy.

107
 Figure 11.1: Magnetic flux density field with approximation error

11.5 Modular implementation

The motor geometry and PCB stator lend itself to a modular system where different components
could be stacked and implemented for various purposes. A motor with 𝑚 PCB stators could be made
using 𝑚 + 1 rotors. When stacking a few rotors together, the axial length would still be relatively
small due to the axial compactness of the PCB stator and rotor combination. Furthermore, efficiency
would increase because only the losses of 2 bearings are present for multiple rotors. This system
could effectively make for a versatile motor that can be used in various situations. The possibilities
should be explored, and further motor design should be adapted to these possibilities. One such
example of a modular system is shown in Figure 11.2. This motor is made up of 3 stators. Because of
this, the motor power triples whilst maintaining a relatively small axial length. Individual elements
can easily be added or subtracted from this system. This makes for a very versatile system. It should
be noted that this is just a simplified example.

Figure 11.2: Modular motor system with 3 individual elements

108
11.6 Mechanical analysis

To further optimize the motor, an in-depth mechanical analysis should be carried out. This is
especially necessary because of the strong attraction forces between the magnets. If the motor
torque were to increase, extra research should be done regarding the motor shaft flange and PCB
strength. This is because the full electromagnetic motor torque is transmitted through to these
components.

Furthermore, extra care should be taken in the alignment and balancing of the motor. During testing
the vibrations were very high. Because of this the maximum speed was limited.

11.7 Thermal analysis

A method was proposed by Rasekh for determining the convective heat transfer coefficients using a
few parameters that are linked to the airgap, air speed and magnet dimensions [14]. A thermal
analysis using this study should be carried out to get a better understanding of the thermal
weaknesses and strengths of this design. This way a better cooling system could be designed to
improve the cooling characteristics and thus get more power out of the motor without causing
thermal damage. Furthermore, an optimal balance between the windage losses and cooling of the
motor via airflow could be found.

11.8 Motor losses

A further modelling of the motor losses should be done besides the copper losses, this will give more
insight in the weaknesses and strengths of this motor. This way the parts that must be optimised can
be determined. A few of the most important losses are listed below.

High frequency losses

Because this is a synchronous motor the rotational speed is proportional to the applied supply
voltage frequency. Running the motor with 5 pole pairs at 6000 rpm requires a frequency of 500 Hz.
Running the motor faster would require higher frequencies. At higher frequencies skin effect and
proximity effect get more and more important and increase the effective electrical resistance. To
fully analyse the losses and characteristics of this motor, a study of these effects should be made.

Bearing losses

The bearings will contribute to a mechanical power loss. Knowledge about these losses could
contribute to a better bearing choice. However, these losses are generally almost negligible.

109
Windage losses

The rotor halves act as a fan sucking in air via the annular opening at the motor shaft and flinging the
air outwards along the outer radius. This generates an air flow that effectively cools the magnets and
stator but also contributes to a windage power loss. A better understanding of this effect could help
with the design. Research about windage losses in axial flux motors has been carried out by Rasekh,
where the windage losses are determined using a few parameters that are linked to the airgap, air
speed and magnet dimensions [14].

11.9 Test setup

Because the test setup was limited, only a limited amount of data points was taken. For future work
it could be useful to improve the setup and get a better understanding of the true motor
characteristics.

110
 Reference List
[1] C. Nunez, What is global warming, explained, National Geographic, Jan. 22, 2019. Accessed
on: Mar. 3, 2021. [Online]. Available:
https://www.nationalgeographic.com/environment/article/global-warming-overview

[2] Vitra, The global electric vehicle market in 2021: statistics & forecasts, Vitra, Mar. 10, 2021.
Accessed on: Mar. 3, 2021. [online]. Available: https://www.virta.global/global-electric-
vehicle-market

[3] P. Sergeant. (2020). Elektrische aandrijftechniek, Hoofdstuk 5: Synchrone machines

[PowerPoint slides]

[4] H. Vansompel, “Design of an energy efficient axial flux permanent magnet machine,”
Department of Electrical energy, systems and automation, Ghent, Belgium, 2013. [Online].
Available: http://hdl.handle.net/1854/LU-4121450

[5] V. Soni, M. Patel, and R. S. Narde, A Laboratory Method for Printed Circuit Board
manufacturing., Apr. 2013. Accessed on: Mar. 3, 2021. [Online]. Available doi:
10.13140/2.1.2221.3124

[6] F. Tokgöz, G. Çakal and O. Keysan. “Comparison of PCB winding topologies for axial-flux
permanent magnet synchronous machines,” IET Electric Power Applications, vol. 14, no. 13,
pp. 2577-2586, Feb. 8, 2021. Accessed on: Mar. 3, 2021. [Online]. Available doi:
https://doi.org/10.1049/iet-epa.2020.0622

[7] FEMM 4.2. (2021). Accessed: 2021. [Online]. Available:

https://www.femm.info/wiki/Download

[8] D. Meeker. Finite Element Method Magnetics Version 4.2 User’s Manual, (2015). Accessed:
May 2, 2021. [Online]. Available: https://www.femm.info/wiki/Documentation/

[9] MATLAB 9.9. (2021), MathWorks. Accessed: 2021. [Online]. Available:

https://nl.mathworks.com/downloads/

[10] F. Niessen, Winding Scheme Calculator, Bavaria direct, 2013. Accessed on: Nov. 2020.
[Online]. Available: https://www.bavaria-direct.co.za/scheme/calculator/

[11] M. D. Noh, J. Kim and Y. Park, "Comparisons of Concentrated Printed-Circuit Stator Windings
for Axial Flux Permanent Magnet Machines," 2019 IEEE/ASME International Conference on
Advanced Intelligent Mechatronics (AIM), 2019, pp. 229-234, doi:
10.1109/AIM.2019.8868912.

[12] R. Clarke, The force produced by a magnetic field, The University of Surrey, July 8, 2010.
Accessed on: Apr. 30, 2021. [Online]. Available:
http://info.ee.surrey.ac.uk/Workshop/advice/coils/force.html#MPF

111
[13] W. C. Young and R. G. Budynas, “Flat Plates,” in Roark’s Formulas for Stress and Strain, 7th
ed. USA: McGraw-Hill, 2001, ch. 11, sec. 11.4, pp. 455-520.

[14] A. Rasekh, “Computational study of convective cooling and windage losses of axial flux
permanent magnet synchronous machines,” Ghent University. Faculty of Engineering and
Architecture, Ghent, Belgium, 2017. [Online]. Available: http://hdl.handle.net/1854/LU-
8527952

[15] Hobbywing, Skywalker ESC (50A UBEC), Hobbywing, 2021. Accessed on: May 20, 2021.
[Online]. Available: https://www.hobbywingdirect.com/products/skywalker-esc-50a-
ubec?variant=3471292932

[16] S. Mukhopadhyay, S. Fernandes, M. Shihab and D. Waleed, “Using Small Capacity Fuel Cells
Onboard Drones for Battery Cooling: An Experimental Study,” Applied Sciences: Battery
Management and State Estimation, June 6, 2018. Accessed on: Mar. 3, 2021 . [Online].
Available doi: https://doi.org/10.3390/app8060942

[17] Siemens NX 12.0, Siemens. Accessed: 2021. [Online]. Available:

https://www.plm.automation.siemens.com/global/en/products/nx/

[18] R. J. Osczevski, “The Basis of Wind Chill,” Arctic, vol. 48, no. 4, pp. 372-382, Jan. 1, 1995.
Accessed on: Mar. 3, 2021. [Online]. Available doi: https://doi.org/10.14430/arctic1262

[19] Alliance, “Alliance N-45 Neodymium Iron Boron Magnetic Material,” Neodymium Iron Boron
Magnetic Material datasheet, n.d. Accessed on: Mar. 3, 2021. [Online]. Available:
http://www.matweb.com/search/datasheet_print.aspx?matguid=e7eda0f5b11243a6a9ea46
f21d667dc1

[20] P. Keane, Thermally Conductive Polymer Materials for 3D Printing, 3Dprinting.com, July 29,
2020. Accessed on: Mar. 12, 2021. [Online]. Available: https://3dprinting.com/3d-printing-
use-cases/thermally-conductive-polymer-materials/

[21] Engineering toolbox, “Thermal Conductivity of Metals, Metallic Elements and Alloys,”
Thermal conductivity datasheet, n.d. Accessed on: Mar. 3, 2021. [Online]. Available:
https://www.engineeringtoolbox.com/thermal-conductivity-metals-d_858.html

[22] D. Lin, P. Zhou and Z. J. Cendes, "In-Depth Study of the Torque Constant for Permanent-
Magnet Machines," in IEEE Transactions on Magnetics, vol. 45, no. 12, pp. 5383-5387, Dec.
2009, doi: 10.1109/TMAG.2009.2026043.

[23] Arduino, Arduino Uno rev3, Arduino Official Store, n.d. Accessed on: May 20, 2021. [Online].
Available: https://store.arduino.cc/arduino-uno-rev3

[24] CHARLIE, lab bench power supply, GrabCad, Jan. 11, 2017. Accessed on May 14, 2021.
[Online]. Available: https://grabcad.com/library/lab-bench-power-supply-1

[25] J. Olavides, HGLRC 30A Brushless ESC, GrabCad, Aug. 1, 2020. Accessed on May 14, 2021.
[Online]. Available: https://grabcad.com/library/hglrc-30a-brushless-esc-1

112
[26] A. Gong and D. Verstraete, “Experimental Testing of Electronic Speed Controllers for UAVs,”
in 53rd AIAA/SAE/ASEE Joint Propulsion Conference, 2017, doi: 10.2514/6.2017-4955

[27] SKF, The SKF model for calculating the frictional moment, SKF, n.d. Accessed on: May 15,
2021. [Online]. Available: https://www.skf.com/binaries/pub12/Images/0901d1968065e9e7-
The-SKF-model-for-calculating-the-frictional-movement_tcm_12-299767.pdf

[28] J. Azzouzi, G. Barakat and B. Dakyo, "Quasi-3D analytical modeling of the magnetic field of an
axial flux permanent magnet synchronous machine," IEEE International Electric Machines
and Drives Conference, 2003. IEMDC'03., 2003, pp. 1941-1947 vol.3, doi:
10.1109/IEMDC.2003.1210717.

113
Appendix 1: Drafts
The drafts for all the motor components are listed below with a scale of 1: 1. The drafts were made using Siemens NX [17].

114
115
116
117
118
119
120
121
122
 Appendix 2: Bearing losses
Calculations for the bearing losses are made using SKF’s frictional moment calculator [27]. The total
frictional moment consists of a few frictional causes, for this application only the rolling and sliding
frictional moments are of importance because the bearings do not have seals, only shields and no oil
bath is used.

The bearing frictional moment will be calculated in this example using a rotational speed of 𝑛 =
4000 rpm at a temperature of 40 °C. Furthermore, the bearing grease LTH23 is used and has an
operating viscosity of 𝜈 = 22,3 mm2⁄s, this value was retrieved from the SKF online bearing
selection tool at https://www.skfbearingselect.com/#/bearing-selection-start. The exact bearing that
will be used for this example is the 61804-2Z bearing from SKF. The dimensions are the following:

𝐵𝑒𝑎𝑟𝑖𝑛𝑔 𝑏𝑜𝑟𝑒 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 𝑑 = 20 mm

𝐵𝑒𝑎𝑟𝑖𝑛𝑔 𝑜𝑢𝑡𝑠𝑖𝑑𝑒 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 = 32 mm
𝐵𝑒𝑎𝑟𝑖𝑛𝑔 𝑚𝑒𝑎𝑛 𝑑𝑖𝑎𝑚𝑒𝑡𝑒𝑟 = 26 mm
For this example, all the tables, data and examples are taken from SKF’s frictional moment calculator
[27].

Rolling frictional moment 𝑴𝒓𝒓

The rolling frictional moment can be calculated using

𝑀𝑟𝑟 = Φ𝑖𝑠ℎ Φ𝑟𝑠 𝐺𝑟𝑟 (𝜈𝑛)0,6

Where

𝑀𝑟𝑟 = Rolling frictional moment [Nmm]

Φ𝑖𝑠ℎ = Shear heating reduction factor
Φ𝑟𝑠 = Kinematic replenishment/starvation reduction factor
𝐺𝑟𝑟 = Variable depending on the bearing type and load
𝜈 = Actual operating viscosity of the base oil of the grease [mm2⁄s]
𝑛 = Rotational speed [rpm]

Furthermore Φ𝑖𝑠ℎ can be determined using the following formula

1 1
Φ𝑖𝑠ℎ = = = 0,965
1 + 1,84 ∗ 10−9 (𝑛𝑑𝑚 )1,28 𝜈 0,64 1 + 1,84 ∗ 10−9 (4000 ∗ 26)1,28 22,80,64
The kinematic replenishment/starvation reduction factor Φ𝑟𝑠 can be calculated using
1
Φ𝑟𝑠 =
𝐾𝑍
(𝐾𝑟𝑠 𝜈 𝑛 (𝑑+𝐷)√ )
2(𝐷−𝑑)
𝑒
Where

𝐾𝑟𝑠 = Replenishment/starvation constant

𝐾𝑍 = Bearing type related geometric constant
𝑑 = Bearing bore diameter [mm]

123
 𝐷 = Bearing outside diameter [mm]

The replenishment/starvation factor is equal to

• 𝐾𝑟𝑠 = 3 ∗ 10−8 for low level oil bath and oil jet lubrication
• 𝐾𝑟𝑠 = 6 ∗ 10−8 for grease and oil- air lubrication

The bearings in this example use grease, thus 𝐾𝑟𝑠 = 6 ∗ 10−8 should be used.

Furthermore, 𝐾𝑍 can be retrieved from Table 0.1 and is equal to 𝐾𝑍 = 3,1.

Table 0.1: Geometric constants 𝐾𝑍 and 𝐾𝐿

The kinematic replenishment/starvation reduction factor Φ𝑟𝑠 is then

1 1
Φ𝑟𝑠 = = = 0,905
𝐾𝑍 3,1
(𝐾𝑟𝑠 𝜈 𝑛 (𝑑+𝐷)√ ) (6∗10−8 ∗ 22,3∗4000(20+32)√ )
2(𝐷−𝑑) 2(32−20)
𝑒 𝑒
Lastly, 𝐺𝑟𝑟 is a variable depending on the bearing type and load and can be retrieved from Table 0.2.

124
Table 0.2: Geometric and load dependent variables for rolling and sliding frictional moments - Deep groove ball bearings

In the setup for the bearings, no axial loads are present and thus 𝐹𝑎 = 0. 𝐺𝑟𝑟 is then
1,96 0,54
𝐺𝑟𝑟 = 𝑅1 𝑑𝑚 𝐹𝑟
The weight of the motor is approximately 700 g, this radial load is shared by 2 bearings.
𝑚𝑔 (0,700 ∗ 9,81)
𝐹𝑟 = = = 3,43 N
2 2
Furthermore, 𝑅1 is a geometric constant dependant on the bearing type. This parameter can be
retrieved from Table 0.3.
Table 0.3: Geometric constants for rolling and sliding frictional moments of deep groove ball bearings

This bearing is of the 61804-2Z type, thus 𝑅1 = 4,7 ∗ 10−7 .

𝐺𝑟𝑟 is then
1,96 0,54
𝐺𝑟𝑟 = 𝑅1 𝑑𝑚 𝐹𝑟 = (4,7 ∗ 10−7 ) ∗ 261,96 ∗ 3,430,54 = 54,26 ∗ 10−5
Now that all the parameters are known, the rolling frictional moment 𝑀𝑟𝑟 can be calculated

𝑀𝑟𝑟 = Φ𝑖𝑠ℎ Φ𝑟𝑠 𝐺𝑟𝑟 (𝜈𝑛)0,6 = 0,965 ∗ 0,905 ∗ (54,26 ∗ 10−5 ) ∗ (22,3 ∗ 4000)0,6 = 0,442 Nmm

Sliding frictional moment 𝑴𝒔𝒍

The sliding frictional moment can be determined using

𝑀𝑠𝑙 = G𝑠𝑙 µ𝑠𝑙

125
Where

𝑀𝑠𝑙 = Sliding frictional moment [Nmm]

𝐺𝑠𝑙 = Variable depending on the bearing type and load
µ𝑠𝑙 = sliding friction coefficient

𝐺𝑠𝑙 can be determined using Table 0.2. No axial loads are present thus
−0,26 5 ⁄3
𝐺𝑠𝑙 = 𝑆1 𝑑𝑚 𝐹𝑟
𝑆1 is a constant that can be determined using Table 0.3. This bearing is of the 61804-2Z type, thus
𝑆1 = 6,5 ∗ 10−3 . 𝐺𝑠𝑙 is then
−0,26 5 ⁄3
𝐺𝑠𝑙 = 𝑆1 𝑑𝑚 𝐹𝑟 = (6,5 ∗ 10−3 ) ∗ 26−0,26 ∗ 3,435⁄3 = 21,74 ∗ 10−3

After this µ𝑠𝑙 can be determined using

µ𝑠𝑙 = Φ𝑏𝑙 µ𝑏𝑙 + (1 − Φ𝑏𝑙 ) µ𝐸𝐻𝐿

Where

µ𝑠𝑙 = Sliding friction coefficient

µ𝑏𝑙 = Constant depending on the movement where
µ𝐸𝐻𝐿 = Sliding friction coefficient in full-film conditions
Φ𝑏𝑙 = Weighting factor for the sliding friction coefficient

µ𝑏𝑙 is dependant on the movement and

• µ𝑏𝑙 = 0,12 for 𝑛 > 0

• µ𝑏𝑙 = 0,15 for 𝑛 = 0

The rotational speed is nonzero and thus µ𝑏𝑙 = 0,15.

Furthermore, µ𝐸𝐻𝐿 is a coefficient dependant on the type of lubrication and is

• µ𝐸𝐻𝐿 = 0,02 for cylindrical roller bearings

• µ𝐸𝐻𝐿 = 0,002 for tapered roller bearings
• µ𝐸𝐻𝐿 = 0,05 for lubrication with mineral oils
• µ𝐸𝐻𝐿 = 0,04 for lubrication with synthetic oils
• µ𝐸𝐻𝐿 = 0,01 for lubrication with transmission fluids

The bearing lubrication uses mineral oils, thus µ𝐸𝐻𝐿 = 0,05.

Φ𝑏𝑙 can be determined using the following formula

1 1
Φ𝑏𝑙 = 2,6∗10−8 (𝑛 𝜈)1,4 𝑑𝑚
= 2,6∗10−8 (4000∗22,3)1,4 ∗26
= 31,50 ∗ 10−4
𝑒 𝑒
µ𝑠𝑙 is then

µ𝑠𝑙 = Φ𝑏𝑙 µ𝑏𝑙 + (1 − Φ𝑏𝑙 ) µ𝐸𝐻𝐿 = (31,50 ∗ 10−4 ) ∗ 0,12 + (1 − 31,50 ∗ 10−4 ) ∗ 0,05
= 50,22 ∗ 10−3
With all the parameters determined, the total sliding frictional moment 𝑀𝑠𝑙 is then

126
 𝑀𝑠𝑙 = G𝑠𝑙 µ𝑠𝑙 = (21,74 ∗ 10−3 ) ∗ (50,22 ∗ 10−3 ) = 10,92 ∗ 10−4 Nmm
The total frictional moment 𝑀𝑏 is then given by

𝑀𝑏 = 𝑀𝑟𝑟 + 𝑀𝑠𝑙 + 𝑀𝑠𝑒𝑎𝑙 + 𝑀𝑑𝑟𝑎𝑔 = 𝑀𝑟𝑟 + 𝑀𝑠𝑙 = 0,442 + (10,92 ∗ 10−4 ) = 0,443 Nmm

Bearing power loss 𝑷𝒃𝒆𝒂𝒓𝒊𝒏𝒈

The bearing power los 𝑃𝑏𝑒𝑎𝑟𝑖𝑛𝑔 for both bearings at a speed of 4000 rpm is then
2𝜋
𝑃𝑏𝑒𝑎𝑟𝑖𝑛𝑔 = 2𝑀𝑏 ∗ 𝑛 = 2(0,443 ∗ 10−3 ) (4000 ∗ ) = 0,369 W
60

127