Design, Optimization and Comparison of Permanent Magnet Motors For A Low-Speed Direct-Driven Mixer

Design, Optimization and Comparison
of Permanent Magnet Motors for a

Low-Speed Direct-Driven Mixer
Florence Libert
Royal Institute of Technology

Department of Electrical Engineering
Electrical Machines and Power Electronics
Stockholm 2004
Submitted to the School of Computer Science, Electrical Engineering and
Engineering Physics, KTH, in partial fulfilment of the requirements for the
degree of Technical Licentiate.
Copyright c Florence Libert, Sweden 2004

Printed in Sweden
Universitetsservice US AB
TRITA–ETS–2004–12
ISSN 1650–674X
ISRN KTH/R-0412-SE
ISBN 91-7283-901-5
Abstract
Induction motors are commonly coupled to a gearbox when an application

requires a low speed and a high torque. The gearbox is costly, needs
maintenance and decreases the efficiency of the drive. Therefore, taking
away the gearbox is very advantageous. This can be made possible by
using a permanent magnet (PM) synchronous motor running directly at low
speed, so called direct-drive PM synchronous motor.
In this thesis, low-speed direct-drive PM motors are investigated for a

4.4 kW, 50 rpm mixer used for wastewater treatment. The thesis begins with
an overview of different existing low-speed direct-driven applications. It
concentrates then on a few topologies that are promising for the considered
application: five different radial-flux PM machine configurations with
surface-mounted or buried permanent magnets and two Torus axial-flux PM
machine configurations.
The designs of the PM machines are conducted by solving an opti-

mization problem. The goal is to minimize the active weight of the machine,
while fulfilling the requirements on the outer dimensions, the efficiency and
the cost. The design process, based on analytical calculations as well as
finite-element simulations, is described taking the particularities of each
topology into account.
Designs of low-speed radial-flux PM machines with high pole numbers

are first calculated with conventional distributed windings and one slot per
pole per phase. The results obtained for different PM machine configurations
are compared. To improve further the designs, concentrated windings are
investigated for PM machines with high pole numbers. Concentrated
winding layouts offering high winding factors and limiting the rotor losses
i
as well as the noise and vibrations of the machine are identified and sorted
out. Some interesting concentrated winding layouts are combined with the
different radial-flux PM machine configurations and further investigated
with FEM simulations.
The design of axial-flux PM motors is treated separately in the thesis.

It was found that the Torus axial-flux PM machines are not satisfying
the requirements of the application. Therefore multi-disk axial-flux PM
machines are investigated.
Finally, the investigated theoretical designs suitable for the application are
compared and their advantages and drawbacks are emphasized.
ii
Acknowledgments
The work presented in this thesis has been carried out at the division of
Electrical Machines and Power Electronics (EME), Department of Electrical
Engineering (ETS), Royal Institute of Technology (KTH). It is an ongoing
project within the Permanent Magnet Drives (PMD) research program of
the Competence Centre in Electric Power Engineering. As one of the pilot
projects within PMD, the main objective is to investigate and develop new
concepts of electrical drives in close cooperation with industrial partners.
The main industrial partner involved in this project is ITT Flygt.
Since I started to work on my Ph.D., several people have been in-

volved in the project. Hereby, I kindly acknowledge them.
I thank first of all my project leader, Dr. Juliette Soulard, for her
guidance, support, enthusiasm and trust. I thank also to Dr. Jörgen
Engström from ITT Flygt, for following the project, helping and making
useful suggestions. I am also very grateful to my supervisor, Prof. Chandur
Sadarangani for his friendly approach in our discussions throughout the
project.
Besides my leaders and supervisor, there are many people that helped
me with my work:
I thank the staff of ITT Flygt, especially Jürgen Mökander, Ulf Carlsson,
Dr. Lars Uby and Gert Hallgren for their help on questions dealing with the
mixer.
I thank Dr. Thomas Bäckström, Freddy Magnussen, Prof. Hans-Peter Nee,
Tech. Lic. Mats Leksell, Tech. Lic. Maddalena Cirani and Dr. Peter Thelin,
for interesting discussions.
iii
Two persons were of great help during this project. The first one is Stephan
Meier. I began my project using some parts of his master thesis work.
Stephan also critically read the thesis and made it easier to read for non
French speakers. The second is Tech. Lic. Robert Chin, who always succeeds
giving an interesting answer to my sometimes tricky questions. Many thanks!
I am grateful to Peter Lönn, who was helping me with my various

computer problems, so stupid they could be sometimes.
I might not have started this work if the working environment at EME were
not so great. Therefore, I would like to thank the whole personal at EME
as well as the former employees for contributing to make EME as it is.
Special thanks to Maddalena, Ruslan, Stefan and Sylvain for their friendship.
Very special thanks to my office-mate Robert, my boyfriend’s office-mate
Karsten, and to my across-the-corridor office-neighbor Eva. They are always
friendly, cheerful and ”cool”...
Thanks a lot to my parents for your support and care. Sorry if you
do not always understand or know what I am doing here (mais je travaille,
recherche, et trouve... un petit-ami ;o)
Finally, I would like to thank again my ”sambo” Stephan for his help,
support and love...
Stockholm,
November 2004
iv
Contents
1 Introduction 1
1.1 Permanent magnet synchronous machines . . . . . . . . . . . 1
1.2 Low-speed direct-drive applications . . . . . . . . . . . . . . . 1
1.3 The project application: The banana-blade mixer . . . . . . . 2
1.4 Goal of the project . . . . . . . . . . . . . . . . . . . . . . . . 4
1.5 Content of the thesis . . . . . . . . . . . . . . . . . . . . . . . 5
2 PM machines for low-speed direct drives 7

2.1 Radial-flux PM synchronous machine . . . . . . . . . . . . . . 7
2.1.1 Surface-mounted PM machines with inner rotor . . . . 8
2.1.2 Surface-mounted PM machines with outer rotor . . . . 9
2.1.3 Inset PM machines . . . . . . . . . . . . . . . . . . . . 10
2.1.4 Buried PM machines . . . . . . . . . . . . . . . . . . . 10
2.2 Axial-flux PM machines . . . . . . . . . . . . . . . . . . . . . 13
2.3 Transverse-flux PM machines . . . . . . . . . . . . . . . . . . 15
2.4 Line-Start PM, modular PM and induction PM machines . . 16
2.4.1 Line-Start PM machines . . . . . . . . . . . . . . . . . 16
2.4.2 Modular PM machines for variable speed operation . . 17
2.4.3 Induction PM machines . . . . . . . . . . . . . . . . . 18
2.5 The investigated motor configurations . . . . . . . . . . . . . 19
3 Properties of low-speed radial-flux PM motors 21

3.1 Geometrical properties: Dimensions of the motors . . . . . . 21
3.1.1 Inner-rotor SMPM motors and inset PM motors . . . 22
3.1.2 Outer-rotor SMPM motors . . . . . . . . . . . . . . . 23
3.1.3 Buried PM motors . . . . . . . . . . . . . . . . . . . . 24
3.2 Magnetic properties . . . . . . . . . . . . . . . . . . . . . . . 24
3.2.1 Analytical calculation of the flux density in the airgap 24
3.2.2 Analytical calculation of the flux density in the teeth . 33
3.3 Electrical properties . . . . . . . . . . . . . . . . . . . . . . . 34
v
Contents
3.3.1 Inductances . . . . . . . . . . . . . . . . . . . . . . . . 35
3.3.2 Resistance of one phase of the stator winding . . . . . 36
3.3.3 The external voltage . . . . . . . . . . . . . . . . . . . 37
3.3.4 Induced voltage . . . . . . . . . . . . . . . . . . . . . . 37
3.3.5 Ampere-turns per slot . . . . . . . . . . . . . . . . . . 38
3.3.6 Angle β . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3.7 Number of conductors per slot . . . . . . . . . . . . . 41
3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
4 Optimization procedure and comparison of designs with dis-

tributed windings 43
4.1 The design procedure . . . . . . . . . . . . . . . . . . . . . . 43
4.1.1 Objective function . . . . . . . . . . . . . . . . . . . . 45
4.1.2 Design variables and their ranges . . . . . . . . . . . . 45
4.1.3 Given constants . . . . . . . . . . . . . . . . . . . . . . 47
4.1.4 Constraints . . . . . . . . . . . . . . . . . . . . . . . . 48
4.2 Results and comparison of designs with distributed windings 50
4.2.1 Influence of some of the constraints . . . . . . . . . . . 50
4.3 Comparison of the configurations . . . . . . . . . . . . . . . . 52
4.3.1 Weight and pole number . . . . . . . . . . . . . . . . . 53
4.3.2 Torque ripple . . . . . . . . . . . . . . . . . . . . . . . 56
4.3.3 Iron losses . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . 60
5 Concentrated windings 61
5.1 Irregularly distributed teeth . . . . . . . . . . . . . . . . . . . 62
5.1.1 Winding factor . . . . . . . . . . . . . . . . . . . . . . 62
5.1.2 Influence on the torque ripple . . . . . . . . . . . . . . 64
5.2 Combinations of pole and slot numbers . . . . . . . . . . . . . 64
5.2.1 Winding layouts and winding factors . . . . . . . . . . 65
5.2.2 MMF and harmonics . . . . . . . . . . . . . . . . . . . 73
5.2.3 Torque ripple . . . . . . . . . . . . . . . . . . . . . . . 74
5.2.4 Magnetic noise . . . . . . . . . . . . . . . . . . . . . . 77
5.3 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
vi
Contents
6 Results and comparison of designs with concentrated wind-

ings 81
6.1 SMPM motors . . . . . . . . . . . . . . . . . . . . . . . . . . 82
6.2 Outer-rotor SMPM motors . . . . . . . . . . . . . . . . . . . 85
6.3 Tangentially-magnetized PM motors . . . . . . . . . . . . . . 88
6.4 Choice of a design . . . . . . . . . . . . . . . . . . . . . . . . 91
6.5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91
7 Design and results of axial-flux PM motors 93

7.1 Main features of the Torus AFPM motor . . . . . . . . . . . . 93
7.1.1 Dimensions of the motor . . . . . . . . . . . . . . . . . 93
7.1.2 Flux densities . . . . . . . . . . . . . . . . . . . . . . . 95
7.1.3 Windings . . . . . . . . . . . . . . . . . . . . . . . . . 97
7.2 Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
7.2.1 The design procedure . . . . . . . . . . . . . . . . . . 99
7.2.2 The design constraints . . . . . . . . . . . . . . . . . . 101
7.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
7.3.1 Torus AFPM machine . . . . . . . . . . . . . . . . . . 102
7.3.2 Multi-disk Torus AFPM machine . . . . . . . . . . . . 103
7.3.3 Comparison with the investigated RFPM motors . . . 106
7.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
8 Conclusions and future work 109

8.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
8.2 Suggestions for future work . . . . . . . . . . . . . . . . . . . 110
Bibliography 111
Glossary of symbols and acronyms 117
A Direct-driven machines manufacturers 121

A.1 Wind turbines . . . . . . . . . . . . . . . . . . . . . . . . . . 121
A.2 Other applications . . . . . . . . . . . . . . . . . . . . . . . . 121
B Finite element methods simulations 123

B.1 2D-FEM simulations . . . . . . . . . . . . . . . . . . . . . . . 123
B.2 3D-FEM simulations . . . . . . . . . . . . . . . . . . . . . . . 125
vii
Contents
C Optimization 127
C.1 A simple example . . . . . . . . . . . . . . . . . . . . . . . . . 127
C.2 Optimization program for the motor design . . . . . . . . . . 127
C.2.1 How to deal with the integer variables . . . . . . . . . 128
C.2.2 How to deal with the non-linearities . . . . . . . . . . 129
D Thermal study 131
viii
1 Introduction
1.1 Permanent magnet synchronous machines
Nowadays, electrical motors account for 65 % of the worldwide energy

consumption. As environmental concern increases, electrical drives with
higher efficiency are desirable. Thus, replacing conventional induction
machines with Permanent Magnet (PM) synchronous machines has recently
gained great interest, as the price of PM materials decreases. Indeed,
PM machines have no rotor winding resulting in lower copper losses, and
therefore they feature a higher efficiency than the induction machines.
For low-speed applications, below 500 rpm, PM machines may further

eliminate the need of a gearbox. To adapt the speed and torque of the
machine, a gearbox is traditionally coupled to a standard induction machine.
It is advantageous to take away this mechanical element, because it is
costly, decreases the drive efficiency and needs maintenance. Low-speed
(or high-speed) drives without gearbox are called direct drives, since the
machines are directly coupled to the load.
1.2 Low-speed direct-drive applications
Today, low-speed PM direct drives are mostly used for two applications,
wind turbines and boat propulsion (figure 1.1). Direct drives for wind tur-
bine generators are widely studied [1], [2], [3], [4], [5], [6]. For wind turbines,
a direct-drive generator makes it possible to avoid the noisy gearbox and to
reduce the travels aimed to oil the gearbox of remote turbines. Some man-
ufacturers already produce wind turbines with direct-drive generators, such
as Lagerwey, Jeumont Industry or Enercon (appendix A).
1
1 Introduction
Figure 1.1: Some direct-drive applications: Wind turbines from Jeumont

Industry, ABB’s Azipod for boat propulsion (appendix A) and
Kone’s elevator without machine room [7].
For boat propulsion, direct drives offer the possibility of light pods that con-
tain the motor and are easy to maneuver [8], [9]. Direct drives are also
successfully used for elevators [7], [10], where they make it possible to get rid
of the whole machine room. More detailed examples of direct-drive applica-
tions are given in chapter 2, and some low-speed direct-drive manufacturers
are listed in appendix A.
1.3 The project application: The banana-blade mixer
In this thesis, the work will mainly focus on a mixer from ITT Flygt. Their
banana-blade mixer is used in wastewater treatment plants to stir big volumes
of liquid. The mixer is driven by an induction motor coupled to a gearbox
(figure 1.2). It runs at 50 rpm, 4.4 kW. Its total weight is 245 kg, where the
propeller accounts for 92 kg and the motor for 22 kg. The characteristics of
the two-pole motor are given in table 1.1. The gearbox has an efficiency of
around 90 % at 4 kW, which decreases to 80-85 % at lower power (1.5 kW).
2
1.3 The project application: The banana-blade mixer
1: Cable entry
2: Gearbox
3: Seal housing
4: Induction motor
5: Propeller
1 2 3
4
5
Figure 1.2: ITT Flygt’s banana-blade mixer [11].
Table 1.1: Characteristics of the 2-pole induction motor.

Length 95 mm
Outer diameter 180 mm
Bore diameter 94 mm
Weight 22 kg
Power factor 0.92
Efficiency 81 %
Nominal current 8.6 A
Voltage 400 V
Frequency 50 Hz
3
1 Introduction
1.4 Goal of the project
The goal of the project is to define suitable PM drives for low-speed appli-
cations to replace induction drives with gearbox (figure 1.3). The focus will
be on the banana-blade mixer described previously. The PM solution should
compete with the induction motor and its gearbox. It should also fulfill the
specifications described in table 1.2 and have a competitive price. Most PM
motors are synchronous so they require to be inverter-fed to be able to start
when connected. The presence of the inverter allows variable speed that may
lead to large improvements of the total efficiency of the mixer.
1500 rpm
50 rpm 50 rpm
~
~
Induction
Motor Gearbox
PM motor
Figure 1.3: Description of the targeted PM drive.
Table 1.2: Design specifications of the direct-drive system.

Maximum outer diameter 500 mm
Maximum length 500 mm
Maximum weight 150 kg
Minimum efficiency 73 %
Nominal torque 840 Nm
Nominal speed 50 rpm
The maximum outer stator diameter is chosen so that it does not exceed 20 %
of the mixer’s propeller diameter. Thus, the motor does not disturb the fluid
behavior near the mixer’s blades. The maximum length should not exceed
the length of the induction motor and the gearbox.
4
1.5 Content of the thesis
1.5 Content of the thesis
A low-speed PM motor features many poles, which makes its design

unconventional. Already the choice of the number of poles is a challenge.
The design procedure should then make it possible to find the number of
poles that gives the best design. The choice of the winding type is also an
important issue. Therefore a whole chapter is dedicated to concentrated
windings for low-speed PM machines. Finally, the design of axial-flux PM
motors is treated separately due to the important differences with the
designs of radial-flux machines.
Different topologies of PM machines are adapted for low-speed direct

drives. They are described in chapter 2 with their advantages and draw-
backs. An overview of investigations, prototypes, and existing products that
use direct drives is presented as well.
Chapter 3 deals with the geometrical, magnetic and electric properties

of low-speed PM motors. The geometrical structures of the different motor
designs are first described. Then the focus is on the calculation of the flux
density in the airgap and teeth, which depends on the rotor configuration.
Finally some details on the calculations of motor inductances and number
of conductors are given. The influence of the saliency of the motor is
emphasized.
In chapter 4, the design procedure for the radial-flux PM machines is

described. It is based on an optimization program that is first explained.
Then some results obtained by running the design procedure are provided.
They allow a comparison between the different investigated configurations.
The concentrated windings for low-speed high-pole number PM ma-

chines are investigated in chapter 5. Indeed, these windings could improve
the designs. The drawbacks of concentrated windings are also studied.
In chapter 6, the optimized designs with concentrated windings are

presented and compared to each other and to the designs with distributed
windings.
5
1 Introduction
Chapter 7 deals with axial-flux PM machines. The design procedure and the
results are presented and compared with the characteristics of the radial-flux
PM machines from chapter 6.
The main results from this design study of PM direct drives for the
banana-blade mixer are given in chapter 8. At the end of the thesis, the
possible future work is presented.
6
2 PM machines for low-speed direct
drives
In this chapter, the different PM machines that are suitable for low-speed
direct-drive applications are presented. Each configuration is shortly de-
scribed and its advantages and drawbacks are emphasized. Examples of both
academic work and industrial applications are provided.
2.1 Radial-flux PM synchronous machine
Radial-flux PM synchronous machines (RFPM) are the most conventional

PM machines. They are widely used for direct-drive applications.
Figure 2.1 shows two views of a RFPM machine with the direction of the flux
and current flows. The flux flows radially in the machine while the current
flows in the axial direction.
flux direction Rotor Shaft

current direction PM
Figure 2.1: RFPM machine with flux and current flows.
7
2 PM machines for low-speed direct drives
RFPM machines are the easiest and cheapest to manufacture among the PM
machines, as they are the most commonly used and their stators are similar
to the ones of the induction machines. However, they are much bigger than
the axial-flux and transverse flux machines in terms of active weight and axial
length [12]. Different possible configurations of RFPM are described below.
2.1.1 Surface-mounted PM machines with inner rotor
For these machines, the permanent magnets are placed on the rotor surface,
as shown in figure 2.2. This is the most commonly used configuration.
The main advantage of the Surface-Mounted PM (SMPM) machine is

its simplicity and consequently its lower construction cost compared to
other PM machines. The main drawback is the exposition of the permanent
magnets to demagnetisation fields. Furthermore, the magnets are subject to
centrifugal forces that can cause their detachment from the rotor. However,
as these forces increase with the rotational speed, they are low in the studied
low-speed applications.
The main direct-drive application of the inner-rotor SMPM machine is

ship propulsion, as the shape of the machine (small diameter, large length)
fits to the pod requirements [13], [14]. SMPM generators have also been
used in the first direct-driven wind turbines [15], [16].
Stator iron
Rotor iron
Permanent magnet
Direction of magnetisation
Figure 2.2: Cross-section of a surface-mounted PM motor with inner rotor

(one pole pair).
8
2.1 Radial-flux PM synchronous machine
2.1.2 Surface-mounted PM machines with outer rotor
The machine consists of a stationary wound stator located in the centre of

the machine while the magnets are mounted along the inner circumference
of the rotor (figure 2.3).
Stator iron
Rotor iron
Permanent magnet
Figure 2.3: Cross-section of a surface-mounted PM motor with outer rotor

(one pole pair).
Some advantages of this configuration are:
• The rotor diameter is larger than for conventional radial-flux machines,

which allows a higher number of poles.
• During the rotation of the outer rotor, the centrifugal forces exert a
pressure on the permanent magnets making their detachment more
improbable.
• The structure is well adapted to wind turbines, as the hub carrying the
blades can be fixed directly to the outer rotor.
This type of machine has been studied by J. Chen and W. Wu and two
prototypes of 20 kW, with 36 and 48 poles respectively were constructed [6],
[17], [18]. Outer-rotor generators are commonly used in small wind turbines
(up to 30 kW), like for example the 7.5 kW turbine of Bergey, the turbines
from Westwind or from Genesys (refer to appendix A).
9
2.1.3 Inset PM machines
As the SMPM machines, the inset PM machines have permanent magnets

mounted on the rotor surface. However, the gaps between the permanent
magnets are partially filled with iron, as shown in figure 2.4. This configu-
ration is referred to as inset PM machine. The iron between the permanent
magnets creates a saliency and gives a reluctance torque in addition to the
torque from the magnets. No example of inset PM machine for low-speed
direct-drive applications was found. Therefore it will be of interest to study
the influence of the reluctance torque on the performance of the machine
and to compare it to the SMPM machine designs.
Stator iron
Rotor iron
Permanent magnet
Figure 2.4: Cross-section of an inset PM motor (one pole pair).
2.1.4 Buried PM machines
Another way to place the permanent magnets in the rotor is to bury them
inside the rotor lamination. An advantage of the buried PM machine
configurations compared to the surface PM machines is the possibility to
concentrate the flux generated by the permanent magnets in the rotor and
thus achieve high airgap flux densities. Moreover, the buried permanent
magnets are well protected against demagnetisation and mechanical stress.
Many different possibilities for the placement of the permanent mag-

nets in the rotor are conceivable, but this thesis concentrates on the two
configurations described below.
10
2.1 Radial-flux PM synchronous machine
V-shaped permanent magnets
In this configuration (figure 2.5), two permanent magnets per pole are
placed in a certain angle taking the form of a ”V”. Between the ends of the
V-shaped permanent magnets and the airgap, there are two iron bridges.
Stator iron
Rotor iron
Permanent magnet
Figure 2.5: Cross-section of a V-shaped buried PM motor (one pole pair).
The main drawback of rotors with V-shaped PMs is the presence of

the iron bridges. The minimal thickness of the bridges being fixed by
mechanical constraints, a large part of the permanent magnet flux leaks
through these bridges instead of crossing the airgap and contributing to the
torque. Besides, the V-shaped rotor is not very easily adapted for high pole
numbers: the higher the pole number, the smaller the place available for the
permanent magnets and the smaller the angle between the two V-shaped
magnets. Therefore, the iron between the permanent magnets can easily get
saturated if the angle is too little. Another drawback of the V-shaped PM
configuration is the high number of permanent magnets, which increases the
production cost.
Such machines were studied in [19], where a 45 kW, 600 rpm synchronous
motor with V-shaped PMs was constructed for an application in the pulp
and paper industry.
11
Tangentially-magnetized permanent magnets
In this second buried PM configuration (figure 2.6), referred to as

tangentially-magnetized PM, the rotor consists of different pieces of iron and
permanent magnets that are fixed together on a non-ferromagnetic shaft.
With a ferromagnetic shaft, a large portion of the flux generated by the
permanent magnets would leak through the shaft.
Stator iron
Rotor iron
Permanent magnet
Figure 2.6: Cross-section of a tangentially-magnetized PM motor (one pole

pair).
The tangentially-magnetised PM motor has the drawback that many pieces

of iron and permanent magnets have to be manipulated if the number of
poles is high. Therefore, some production difficulties can arise. However,
this motor does not contain any iron bridges compared to the configuration
with V-shaped PM and the flux leakage is thus very low.
A design of a tangentially-magnetized PM machine was presented in

[16]. However, the study emphasizes more the choice of the type of windings
than the advantages of this rotor configuration.
12
2.2 Axial-flux PM machines
2.2 Axial-flux PM machines
The axial-flux PM (AFPM) machine is another possible solution for low-

speed direct-drive applications. This machine features a large diameter and
a relatively short axial length compared to a radial-flux PM machine. As
suggested by its name and figure 2.7, the flux from the PMs flows axially
while the current flows in the radial direction.
Windings
x x Stator iron
Rotor iron
PM
x x Shaft
x
Direction of magnetisation of the PM
Current direction
Figure 2.7: AFPM machine with flux and current flows.
Different kinds of AFPM exist, but for low-speed applications the most
commonly studied topology is the Torus machine [20]. Therein, the stator
is placed between the two external rotors that are rigidly connected to the
mechanical shaft. The permanent magnets are placed opposite to each other
on the two rotors and the stator windings are toroidal (figure 2.8), [21]. The
stator can be slotless. The main benefits of the Torus machine are:
• The machine is compact and lightweight with a short axial length.
• Good ventilation and cooling of the stator windings.
• Absence of slots and large effective airgap, which leads to a negligible

cogging torque, reduced high-frequency losses and also a low acoustic
noise.
• Possibility to stack together many stator and rotor discs, which gives
a higher effective airgap surface.
13
Permanent magnets
Stator core
Rotor discs
Figure 2.8: Torus machine [20].
The main drawback of this kind of machine is its complex assembly due to
axial forces.
Other AFPM machine configurations are the double-sided with internal PM

disk machines, single-sided machines, ironless double-sided machines, and
multidisk machines. These are described in [22]. The different configurations
are compared in [12] for a nominal speed of 1000 rpm.
AFPM machines were studied using prototypes in [3], [8], [10], [23],
[24], [25]. The applications are small wind turbines, ship propulsion and
elevators. Different companies already produce AFPM machines as the
wind turbine producer Jeumont Industry, the motor fabricant Lynx Motion
Technology Corporation, and the elevator producer Kone that established
the concept of an elevator without a machine room thanks to the use of
axial-flux PM motors [7].
14
2.3 Transverse-flux PM machines
2.3 Transverse-flux PM machines
The basic arrangement of a transverse-flux PM machine (TFM) is shown

in figure 2.9. This type of machine suits well for low-speed direct-drive
applications because of a small pole pitch. Nevertheless, it is not very
common yet.
Stator Stator hoops

hoops
N
S Winding Magnets
N Rotor
N Rotor
S Windings
Windings
Magnet Stator hoops
Figure 2.9: Transverse-flux PM machine [26].
The advantage of the TFM is the high specific torque that allows the
compactness of the machine. The major drawbacks are the poor power
factor at high specific torque and the complexity of the structure which
leads to high manufacturing costs.
A prototype of a TFM for an electrical vehicle was built and tested

in Germany [27] in 1997. The same year, Rolls-Royce Internatinal Research
and Development was planning a prototype of a 20 MW, 180 rpm TFM for
ship propulsion [26].
15
2.4 Line-Start PM, modular PM and induction PM

machines
2.4.1 Line-Start PM machines
A Line-Start PM motor (LSPM) is a synchronous PM motor with a squir-

rel cage in the rotor (figure 2.10). The squirrel cage, as it is the case for
an induction machines, allows the motor to start without any inverter. The
LSPM motor is thus directly connected to the mains. It starts asynchronously
thanks to the squirrel cage and operates at steady state synchronously.
Stator iron
Rotor iron
Permanent magnet
Squirrel cage
Figure 2.10: Line-Start PM machine (one pole pair).
LSPM motors are intended for replacing induction motors, having the advan-
tage of a much higher efficiency and power factor, but a decreased starting
capability and higher costs. These motors are not very common yet, since
their design is tricky and they are more expensive than induction machines.
They have been investigated for replacing induction motors in pumps [28].
No example of a low-speed direct-drive LSPM machine has been found.
16
2.4 Line-Start PM, modular PM and induction PM machines
2.4.2 Modular PM machines for variable speed operation
The modular PM machine is used for variable speed wind turbine generators.
As its name points out, the machine consists of different modules. Every
rotor module comprises one pole. The stator modules are composed of
an E-shaped core, whose two slots contain a single coil of the winding
(figure 2.11). Each coil is then connected to a single-phase rectifier bridge
and operates independently.
Stator module
with core and coil
Steel pole side Ferrite magnet
Figure 2.11: Stator and rotor arrangement of a modular PM machine [29].
The advantages of this configuration are mostly its easy assembly, the simple
cooling, and the simple coil shape. A drawback is the high reactance of
the coils, causing the need for several capacitors to compensate the reactive
voltage drop. Some additional eddy-current losses may also arise due to the
design structure.
This kind of machine was studied by E. Spooner with the conception

of a prototype with 26 poles [29], [30], [31]. O. Carlson studied a 40 kW,
48-pole prototype [2]. The goal was to apply this modular concept for wind
turbine generators in the power range of several megawatt.
17
2.4.3 Induction PM machines
The induction PM machine was studied at the university of Darmstadt

for wind turbine applications [32]. No prototype was built. The work was
mostly based on Finite Element Methods (FEM) simulations. The generator
consists of two rotating parts, the rotor with a squirrel cage in the center
and a ring that supports the permanent magnets (see figure 2.12).
Stator with Rotating ring with

3 phase winding permanent magnets
Rotor with
squirrel cage
Figure 2.12: Induction PM generator [32].
An advantage of the induction PM generator for wind turbines is the

possibility to connect the generator via a transformer directly to the grid,
without any power electronics. The slip of the induction machine puts
less stress on the blades than with a synchronous generator. However, the
mechanical construction is complex due to the two rotating parts.
18
2.5 The investigated motor configurations
2.5 The investigated motor configurations
In this thesis, designs of SMPM motors with inner or outer rotors, inset
PM motors, buried V-shaped PM motors and tangentially-magnetized PM
motors, as well as Torus AFPM motors are investigated (figure 2.13). These
motors are not self-starting and need a converter for start-up.
Radial-flux PM Axial-flux PM
machines machines
Surface mounted Multi-disk

Buried PM Torus
PM Torus
Inner Outer Inset V-shaped Tangentially-

rotor rotor PM PM magnetized PM
Figure 2.13: The investigated motor configurations.
Transverse-flux motors are not studied because their technology is too

different from the one of the radial-flux machines. It is investigated in
another project of the department for a direct-driven wind generator.
The modular PM and induction PM machines (such as those described

previously) are intended for the special use in wind turbines and therefore
they are not considered.
The possibility of a Line-Start PM motor was studied at the begin-

ning of the project and rapidly abandoned because it was impossible to
fulfill the requirements. In order to achieve the required nominal speed with
a supplied frequency of 50 Hz, 120 poles are needed. These are difficult to
fit in a stator with an outer diameter of 500 mm. The results of the studied
V-shaped PM motors presented in chapter 4 confirm the impossibility of a
LSPM motor design for the mixer.
19
20
3 Properties of low-speed radial-flux
PM motors
This chapter focuses on the description and modelling of the geometrical,

magnetic and electric properties needed for the design of low-speed PM mo-
tors. The design process is described subsequently in chapter 4.
• The geometric properties are the different dimensions and areas of the
motors.
• The magnetic properties are the flux densities in the airgap and in the
teeth.
• The electric properties are the inductances, resistances, ampere-turns

and number of conductors per slot.
Most of these properties depend on the motor configuration.
3.1 Geometrical properties: Dimensions of the motors
The geometrical parameters of the different motor configurations are pre-

sented in this section. All radial-flux PM motors with inner rotors have the
same stator structure as described in section 3.1.1.
21
3 Properties of low-speed radial-flux PM motors
3.1.1 Inner-rotor SMPM motors and inset PM motors
Figure 3.1 shows the geometry of a SMPM motor including the parameters
of the geometrical dimensions. These dimensions are expressed in equations
(3.1) to (3.6), where Qs is the number of stator slots. The parameter kopen
is the ratio of the stator slot opening to the slot width (3.6). The teeth are
straight, which means that the tooth width bts is constant all along the tooth.
2α
Do
δ
bss2
Do
hsy
hsw bss1
bts
hss
bso τs
D
Drc
lm
Figure 3.1: Definition of the geometrical parameters for the SMPM designs.
1
D = Drc + 2lm + 2δ (3.1) hsy = (Do − D − 2hss ) (3.4)
2
D D + 2hss
τs = π (3.2) bss2 = π − bts (3.5)
Qs Qs
D + 2hsw kopen = bso /bss1 (3.6)
bss1 = π − bts (3.3)
Qs
The inner stator diameter D is very large compared to the slot pitch τs .
Therefore, bss1 , bss2 and bts , which actually are arcs of circle, are approxi-
mated as straight lines in equations (3.3) and (3.6). The slot area Asl is given
by equation (3.7).
1
Asl = (bss1 + bss2 ) ∗ (hss − hsw ) (3.7)
2
22
3.1 Geometrical properties: Dimensions of the motors
According to the equations, the two-dimensional geometrical structure of the

stator can be described entirely with the following parameters: Drc , lm , δ,
hsw , bts , Do , hss , kopen and Qs . Adding the number of poles p, the half pole
angle α and the active length L, the whole three-dimensional geometry can
be described (without considering the end-windings).
3.1.2 Outer-rotor SMPM motors
The geometrical parameters of the outer-rotor designs are defined in

figure 3.2. The equations of the inner-rotor SMPM motor are adapted to
this particular configuration and become equations (3.8) to (3.13).
hss hry
hsy
bss2
Di D
Drc
Do
Figure 3.2: Definition of the geometrical parameters for the outer-rotor

SMPM motors.
1
D = Drc − 2lm − 2δ (3.8) hsy = (D − Di − 2hss ) (3.11)
2
D − 2hsw 1
bss1 = π − bts (3.9) hry = (Do − Drc ) (3.12)
Qs 2
D − 2hss bso
bss2 = π − bts (3.10) kopen = (3.13)
Qs bss1
23
3.1.3 Buried PM motors
For the buried designs, the geometrical parameters of the stator are ex-
pressed as for the inner-rotor SMPM motors. The rotor parameters for the
V-shaped PM and tangentially-magnetized PM motor designs are described
in figure 3.3.
2α
one pole
wFe Drc lFe
rotor
wm PM PM
iron
lm
2α 2liron αD
l m= prc
li wm 2αv
tangentially-magnetized
PM rotor
V-shape PM rotor
Figure 3.3: Definition of the parameters for the buried PM rotor

geometries.
The inner stator diameter D is not dependent on the thickness of the per-
manent magnets. The position of the permanent magnets can be deduced
from the magnet thickness lm , the half pole angle α and the angle in the V
αv . The airgap around the permanent magnets and the thickness of the iron
bridges are set to li = 0.2 mm and wF e = 0.54 mm respectively.
3.2 Magnetic properties
3.2.1 Analytical calculation of the flux density in the airgap
The amplitude of the fundamental airgap flux density B̂δ , has to be calculated
with accuracy since the design procedure relies on it. The calculation of B̂δ
varies for the different investigated configurations.
24
Surface-mounted PM motors
For the SMPM motor designs, the airgap flux density is assumed to have a
rectangular shape as wide as the permanent magnet width and a maximum
value Bm . Figure 3.4 shows the flux density in the airgap simulated with
FEM and the analytical estimation.
1
Bm with kleak =1, B δ =0.974 T
0.8
0.6
Airgap flux density [T]
with kleak =1, Bδ =0.910 T

0.4
0.2
FEM, Bδ =0.911 T
0
-0.2 2α
-0.4
-0.6
-0.8
1
0 1 2 3 4 5 6 7
Electrical angle [rad]
Geometry
AIRGAP AIRGAP
PM AIRGAP PM
IRON
Figure 3.4: Analytically calculated and FEM simulated airgap flux densi-
ties, example of a 60-pole SMPM motor.
Bm is calculated as follows:
Br kleak
Bm = μr δkC
(3.14)
1+ lm
where Br is the remanence flux density of the magnet, μr the relative magnet
permeability and kC the Carter factor [33].
τs
kC = (kopen bss1 )2
(3.15)
τs − bss1 kopen +5δ
The factor kleak takes the magnetic leakage between two neighboring per-
manent magnets into account. Indeed, assuming that the flux density has
25
a rectangular shape and neglecting the leakage, can cause a difference of

around 10 % between the analytically calculated and FEM simulated funda-
mental airgap flux densities. This can be seen in figure 3.4.
M
A
G
N H
E TOO T SLOT
T MAGNET
A SLOT
R I TOO TH
O R R
T G O
AIRGAP SLOT
O A TOO TH T
R P O
R
M TOO TH
A SLOT
G MAGNET SLOT
N
E
T
leakage
Figure 3.5: Flux lines for SMPM motors with p = 40 and p = 70.
Figure 3.5 shows the leakage between two permanent magnets for two dif-
ferent pole numbers (100 flux lines are plotted). The higher the number of
poles, the lower the distance between the permanent magnets and conse-
quently the reluctance gets lower, thus increasing the leakage. The magnetic
leakage between the permanent magnets is also notably dependent on the
magnet height and the airgap length, but this has been neglected in order
to keep a simple model. Magneto-static FEM simulations are used to de-
fine the factor kleak as a function that depends linearly on the pole number
(figure 3.6). The leakage factor is defined as the percentage of the flux lines
that pass through the airgap (3.16). Appendix B describes how the FEM
simulations were conducted.
100 − number of leakage f lux lines
kleak = (3.16)
100
⎧
⎪ 100−(7p/60−0.5)
⎪
⎨ 100 for SMPM motors
100−(7p/60−3)
kleak = for outer-rotor SMPM motors
⎪
⎪
100
⎩ 100−(p/10) for inset PM motors
100
26
Number of leaking flux lines from the magnet side [%]

points used to plot the line
verification (simulations with kleak)
8 linear interpolation
1
20 30 40 50 60 70 80
Pole number
Figure 3.6: Flux lines leaking from the permanent magnets as a function of
the number of poles, for SMPM motors.
Furthermore, the influence of the slot opening on the airgap flux is taken into
account, as can be seen in figure 3.4. The width of the drop is approximated
as the slot opening width, whereas the depth is set proportional to the width.
Table 3.1 presents the analytical results obtained for the amplitude of
the fundamental airgap flux density compared to the FEM results for
different geometries. The difference between analytical and FEM values
does not exceed 2 %.
Table 3.1: Comparison of analytical and FEM values of the flux density in
the airgap of SMPM motor designs.
Pole number 20 30 40 50 60 70 80
B̂δ,an [mT] 990.1 956.5 946.7 942.9 945.7 948.2 880.7
B̂δ,F EM [mT] 974.2 943.9 937.2 931.1 935.3 930.1 872.9
Relative error [%] 1.6 1.3 1.0 1.3 1.1 1.9 0.9
27
Inset PM motors
For the inset PM motor designs, the method used for the SMPM motors to
calculate the flux density has to be adapted since there is even more flux
leakage between two permanent magnets. Instead of crossing the airgap, a
significant part of the magnetic flux leaks through the iron pieces (figure 3.7).
Figure 3.7: Flux leakage through the iron piece in an inset PM motor.
The maximum value of the airgap flux density Bm is calculated with

equation (3.14). kleak is obtained by using FEM simulations as in the
SMPM motor designs. According to [34], the flux leakage through the iron
pieces is negligible if the gap between the magnet and the iron piece is more
than twice the airgap length. The waveform of the airgap flux density in
figure 3.8 is based on this assumption. The influence of the slot opening is
also taken into account.
The difference between the FEM simulated and analytically calculated
values of the fundamental airgap flux density is less than 3 % for pole
numbers between 20 and 70, as shown in table 3.2.
Table 3.2: Comparison of analytical and FEM values of the fundamental

airgap flux density for inset PM motor designs.
Pole number 20 30 40 50 56 60 70
B̂δ,an [mT] 943.4 877.5 887.4 891.1 954.1 890.8 885.5
B̂δ,F EM [mT] 928.4 873.9 891.2 901.0 958.4 906.6 906.2
Relative error [%] 1.6 0.4 -0.4 -1.1 -0.4 -1.7 -2.3
28
2δ 2α' 2δ
0.8 SMPM
0.6 inset PM
Airgap flux density [T] 0.4
0.2
-0.2
2α
-0.4 FEM
-0.6
-0.8
-3 -2 -1 0 1 2 3
AIRGAP AIRGAP δ
PM PM
IRON IRON
Figure 3.8: FEM simulated and analytically calculated airgap flux densities
of an inset PM motor.
V-shaped buried PM motors
The maximum of the airgap flux density Bm for the V-shaped buried PM
motor is calculated using equation (3.17), which is derived in [35]. Figure 3.3
describes the different notations used in the formula.

Br − Bsat wwFme 1 + μr llmi
Bm = (3.17)
2αDr
pwm + 2 lF e ·
wF e k C δ
wm 1 + μ li
r lm + μ kC δ
r lm
where Bsat is the flux density in the saturated iron bridges.

The fundamental airgap flux density is then calculated as for the SMPM
motor designs.
29

airgap flux density for V-shaped PM motor designs.
Pole number 20 30 40 50 60
B̂δ,an [T] 1.10 1.11 1.10 1.12 0.864
B̂δ,F EM [T] 1.06 1.08 1.09 1.10 0.877
Relative error [%] 3.8 2.7 0.9 1.8 -0.1
Tangentially-magnetized PM motors
The tangentially-magnetized PM motor configuration is not very conven-

tional. Thus, the calculation of Bm has been determined from the Ampere’s
law and the equivalent lumped circuit of the motor (figure 3.9). The con-
sidered part of the permanent magnet is seen as a constant magneto-motive
force source Hm lm , in series with a constant internal reluctance m . Hm is
the coercive magnetic field intensity of the magnet.
MMF
ℜm+ ℜgap
Hml m
Figure 3.9: Equivalent magnetic circuit for half a pole (half a PM and air-
gap for half iron width).
Br
Hm lm = lm (3.18)
μ0 μr
and
lm
m = (3.19)
μ0 μr Lwm
30
The airgap reluctance is equal to:

δkC
gap = (3.20)
μ0 Lliron
with π D
rc
liron = −α (3.21)
2 p
Assuming that the permeability of iron is infinite, the following equation can
be obtained from the equivalent lumped circuit:
Hm lm = (m + gap )Φ (3.22)
where the flux Φ is:

Φ = Bm Lliron (3.23)
Finally Bm is given by:
Br wm
Bm = μr δkC wm
. (3.24)
1+ liron
lm liron
The approximation used to calculate the fundamental airgap flux density

is described in figure 3.10. As illustrated in the figure, a factor kleakα is
introduced to correct the pole angle. It is determined from FEM simulations:
• kleakα = 1 for Bm ≤ 0.9 T
• kleakα = 0.9 for 0.9 T< Bm ≤ 1 T
• kleakα = 0.8 for 1 T< Bm
Table 3.4 shows how accurate the analytical calculations are, compared with
the respective FEM values.
31
2αkleakα
Bm
Airgap flux density [T] 1
0.5 Analytic
0 2liron
FEM
-0.5
-1
0 1 2 3 4 5 6
AIRGAP AIRGAP
PM IRON PM IRON PM
Figure 3.10: FEM simulated and analytically calculated flux densities in

the airgap for a tangentially-magnetized PM motor.

airgap flux density for tangentially-magnetized PM motors.
Pole number 30 40 50 60 70
B̂δ,an [T] 1.06 1.10 1.10 1.09 1.03
kleakα = 0.8
B̂δ,F EM [T] 1.09 1.06 1.12 1.11 0.99
1 T< Bm
Relative error [%] -2.7 3.8 -1.8 -1.8 4.0
B̂δ,an [T] 1.10 1.03 0.98 0.99 0.96
kleakα = 0.9
B̂δ,F EM [T] 1.08 1.03 0.93 0.97 0.96
0.9 T< Bm ≤ 1 T
Relative error [%] 2.0 0 5.3 2.0 0
B̂δ,an [T] 1.05 1.00 0.93 0.89 0.92
kleakα = 1
B̂δ,F EM [T] 1.01 1.00 0.91 0.84 0.88
Bm ≤ 0.9 T
Relative error [%] 4.0 0 2.2 5.9 4.5
32
3.2.2 Analytical calculation of the flux density in the teeth
It is important to obtain a correct analytical value of the flux density created

by the permanent magnets in the teeth. If the value turns out to be higher
than expected, the teeth can be very saturated which means a high magnetic
leakage.
The flux density in the teeth Bts is calculated from the PM flux flow-
ing through the airgap, the width of a tooth bts and the number of teeth
through which the flux is flowing (2 for q = 1). Equation (3.25) gives the
flux density in a tooth for a SMPM motor. The factor kleaktooth is used to
take into account the part of leakage flux, flowing through the tooth shoe only.
D
Bm 2α p2 2 − δ · (1 − kleaktooth )
Bts = (3.25)
2bts
17p/56 − 13/14
kleaktooth = (3.26)
100
Flux lines not going

through the teeth
Figure 3.11: Flux lines for SMPM motors with p = 30 and p = 70.
Figure 3.11 reveals the flux lines that are used for the calculation of the
leakage factor (the total number of flux lines is 100). For the 30-pole motor,
eight flux lines are not flowing through the teeth, whereas it is twenty flux
33
lines for the 70-pole motor. The factor kleaktooth , given in equation (3.26), is
then determined as a function of the pole number by linear interpolation,
using the FEM simulation results as those of figure 3.11. Table 3.5 shows
the obtained results for different geometries of SMPM motors.
Table 3.5: Comparison of analytical and FEM values of the flux density in
the teeth for SMPM motors.
Pole number 20 30 40 50 60 70 80
B̂ts,an [T] 1.6 1.6 1.6 1.6 1.6 1.6 1.6
B̂ts,F EM [T] 1.55 1.59 1.59 1.59 1.59 1.59 1.62
Relative error [%] 3.1 0.6 0.6 0.6 0.6 0.6 -1.3
3.3 Electrical properties
jLq ωIq RId
jLd ωId
jLq ωIq RI q
RIq
δ E
E
V V I
γ Iq
Iq
β β
Ψm Id Ψm
Non-salient Salient
Figure 3.12: Phasor diagrams for salient and non-salient PM machines at

rated speed.
The phasor diagrams for non-salient and salient PM machines at rated speed
are represented in figure 3.12. SMPM motors are non-salient. The others,
buried and inset PM designs, are salient. In this section, it will be shown
how the different parameters of the phasor diagram are calculated, which
comprise:
34
• The d- and q- axis inductances, Ld and Lq
• The resistance of one phase of the stator winding R
• The external voltage V
• The induced voltage E
• The ampere-turns, ns I
• The angle β between the flux from the permanent magnets Ψm or the
d- axis, and the current I
From the phasor diagram, the number of conductors per slot ns can then be
calculated as shown in the following.
3.3.1 Inductances
SMPM motors
For a non-salient synchronous motor the d- and q- axis synchronous induc-

tances are equal and
Ld = Lq = Ll + Lmd = Ll + Lmq (3.27)
where Ll is the leakage inductance and Lmd and Lmq are the d- and q- axis
magnetizing inductances respectively. The calculation of these inductances
are detailed in [36].
Ll = p q ns 2 Lμ0 λ1 (3.28)
3 μ0
Lmd = (q ns kw1 )2 lm
(D − δ) L (3.29)
π δkC + μr
where λ1 is the specific permeance coefficient of the slot opening and depends
on the slot geometry [33]; q is the number of slots per pole per phase and kw1
is the fundamental winding factor. If the number of slots per pole per phase
is equal to 1, then the winding factor kw1 is 1. Other winding configurations
are discussed in chapter 5.
35
Inset PM motors
For the inset PM motor designs, the d-axis magnetizing inductance is cal-
culated with the same equation as for the SMPM motor (3.29). The q-axis
magnetizing inductance is calculated with:
3 μ0
Lmq = (q ns kw1 )2 (D − δ) L (3.30)
π δkC
Tangentially-magnetized PM motors
Since there could not be found any analytical expression of the q-axis induc-
tance, the saliency is checked with FEM simulations. The d-axis inductance
is calculated as for the SMPM motor. Two FEM simulations at different
load conditions are run, one with only q-axis current and one with only d-
axis current. The ratio of the simulated voltages in phase A gives the saliency
(Lq /Ld ). This method has been first checked on inset PM motor designs. For
tangentially-magnetized PM motors, the saliency (Lq /Ld ) was found to be
around 4 for two different designs. Therefore, the q-axis inductance has been
roughly approximated as 4Ld . Although this approximation might not be
very accurate for other tangentially-magnetized PM motor designs, the cal-
culation of the q-axis inductance was not improved. Indeed, an inaccurate
value of the q-axis inductance causes a reluctance torque that is not opti-
mized. The torque might then be underestimated. However, this does not
penalize the tangentially-magnetized PM motor configuration, as will be seen
in the next chapter.
3.3.2 Resistance of one phase of the stator winding
It is assumed that all the coils in one phase are coupled in series. The phase
resistance is calculated as:
(p L + (D + hss )πkcoil ) n2s q
R = ρcu (3.31)
fs Asl
fs is the slot fill factor. The end-windings are taken into account by intro-
ducing the term Dπkcoil , [37].
36
3.3.3 The external voltage
Since all the investigated motor configurations are not self-starting, power
electronic devices are essential to start the motor and make it rotate at the
right speed.
Figure 3.13 shows a diagram of a simple drive that suits the application. A
diode rectifier is coupled to a 3-phase inverter via a filter capacitor. The out-
put line-to-line voltage VLL of the inverter, which is also the voltage applied
to the motor, can be estimated as [38]:
VLL = 0.612ma Vd (3.32)
where ma is the amplitude modulation ratio of the inverter and Vd the

rectified DC voltage across the filter capacitor (Vd = 400 V). From equation
(3.32), a phase voltage V of around 70 V is obtained assuming a low
amplitude modulation ratio (ma = 0.5). This is the worst case as the current
and the losses in the inverter increase with the decreasing voltage. This
voltage is used to calculate the number of conductors per slot of the motor.
Vd
50 Hz PM
AC VLL motor
Diode Filter Inverter

rectifier capacitor
Figure 3.13: Possible drive for PM motors.
3.3.4 Induced voltage
The induced phase voltage E is deduced from Faraday’s law:

dΨm dφm
E(t) = = Ns (3.33)
dt dt
The maximum fundamental of the magnet flux φ̂m linked to one turn of the
coil is:
2 π
φ̂m = B̂δ L(D − δ) (3.34)
π p
37
Ns is the number of turns per phase which are in series:

p
Ns = q n s (3.35)
2
Finally the rms-value of the induced voltage can be calculated as:
1
E = √ ω kw1 q ns B̂δ L(D − δ) (3.36)
2
3.3.5 Ampere-turns per slot
Current loading
The peak value of the fundamental current loading Sˆ1 is calculated from the
torque equation:
4T
Sˆ1 = 2
(3.37)
π (D − δ) LB̂δ kw1 kcor sinβ
where β is the angle between Ψm and I. β depends on the saliency. For a
non-salient geometry as the SMPM motor, β is π2 . For a salient geometry,
the method applied to calculate β is described in section 3.3.6.
The correction factor kcor is used to compensate the losses and leakages that
are not analytically calculated, such as the flux leakage through the slots.
This factor is obtained from FEM simulations at load conditions, for the dif-
ferent motor configurations. It assures that the demanded nominal torque is
reached in the FEM simulations and thus enables a fair comparison between
the different motor configurations.
Table 3.6: Correction factor kcor

Motor Outer-rotor Tangentially-
SMPM Inset PM
design SMPM magnetized PM
kcor 0.95 0.94 1 0.99
Table 3.6 provides the factor for the studied motor configurations. For the
inset PM motors and tangentially-magnetized PM motors, the current load-
ing is calculated with equation (3.37), that does not take the contribution of
the reluctance torque into account. Therefore, kcor is equal or almost equal
to one. The reluctance torque, which is around 5 % of the nominal torque,
compensates already the leakages that kcor was introduced for.
38
Ampere turns and current density
From the fundamental peak current loading and the ampere-turns (3.38), the
current density J can be calculated.
ns Iˆ = Sˆ1 τs (3.38)
ns I
J= (3.39)
Asl fs
3.3.6 Angle β
YES NO
Salient?
current I current I
Choose β=π/2
β β
βn > π/2, n=1
d- axis d- axis
equation
Current loading (3.37) Current loading
equation
Current density (3.39) Current density
recursion
n=n+1
Find β that gives equations
the highest torque (3.42) and (3.44)
βn+1−βn
small?
phasor diagram
Conductor number (figure 3.12) Conductor number
Figure 3.14: Chart description of how to determine β depending on the

saliency.
39
For a salient machine, the d-axis current generates a reluctance torque in

addition to the torque created by the permanent magnets. Thus, the torque
equation becomes:
3p
T = [Ψm Iq + (Ld − Lq ) Id Iq ] (3.40)
22
The contribution of the reluctance torque to the total torque can be adjusted
with the angle β. Indeed, the d- and q- axis currents depend on β:
ˆ
Id = Icosβ
(3.41)
ˆ
Iq = Isinβ
For non-salient machines, the d-axis current does not contribute to the
torque, therefore β is chosen as π2 .
For a salient machine, the angle β is chosen so that it gives the maximum
torque for a given current.
dT
=0 (3.42)
dβ
By inserting equations (3.40) and (3.41) into equation (3.42), a second order
equation is obtained:
2 (Ld − Lq ) Iˆ2 cos2 β + Ψ̂m Icosβ

ˆ − (Ld − Lq ) Iˆ2 = 0 (3.43)
In this equation, the inductances depend on the square of the number of

conductors per slot. These are unknown as is also the nominal current I.
However, the value of the ampere-turns per slot is known from the required
torque (3.38). The second order equation can therefore be transform into
equation (3.44), which can be solved for β.

ˆ 2 cos2 β + p q φ̂m (ns I)cosβ
2 Ld − Lq (ns I) ˆ − Ld − Lq (ns I)
ˆ 2 = 0 (3.44)
2
where Ld , Lq and φ̂m are independent of the number of conductors per slots
and depend only on the dimensions:
Ld = Ld n2s Lq = Lq n2s (3.45)
The determination of the angle β needs a recursive process, since the current
loading depends on it (3.37). The recursion is described in figure 3.14. Differ-
ent values of β are tested in the recursive process in order to find the minimum
current together with the angle β that provides the required torque. Know-
ing the angle β and the d- and q- axis currents, the number of conductors
per slot can be calculated using the phasor diagram at nominal speed.
40
3.3.7 Number of conductors per slot
As equations (3.29), (3.30) and (3.31) imply, the inductances and resistances
depend on the square of the number of conductors per slot, whereas the
induced voltage is linearly dependent on the number of conductors per slot
(3.33):
Ld = n2s Ld Lq = n2s Lq
R = n2s R E = ns E (3.46)
Non-salient motor designs
For the non-salient motors, the vector diagram in figure 3.12 gives the fol-
lowing equation:
2
V̂ 2 = Ê + RIq + (Ld ωIq )2 (3.47)
Using the same method as for calculating β, the equation becomes:
2
2
V̂ 2 = ns Ê + n2s R Iq + n2s Ld ωIq (3.48)
2 2
V̂ 2 = n2s Ê + R (ns Iq ) + n2s Ld ω (ns Iq ) (3.49)
Finally, as Iq = Iˆ (β = π/2), the number of conductors per slot can be

calculated as:
V̂
ns = 2 2 (3.50)

Ê + R ns Iˆ + Ld ωns Iˆ
Salient motor designs
For the salient motors, the vector diagram of figure 3.12 gives the following
equation:
2
V̂ 2 = Ê + RIq + Ld ωId + (−Lq ωIq + RId )2 (3.51)
41
Therefore, the number of conductors per slot can be derived as:

V̂
ns = 2 2
2
Ê + R ns Iˆ sinβ + Ld ω ns Iˆ cosβ + ns Iˆ −Lq ωsinβ + R cosβ
(3.52)
3.4 Summary
Analytical models to design low-speed radial-flux PM synchronous motors

have been presented. These models were validated with FEM simulations.
Improvements were conducted by introducing correction factors when re-
quired. In the next chapter, the method to include these models in an opti-
mization process is described.
42
4 Optimization procedure and
comparison of designs with
distributed windings
In the following, the procedure for designing radial-flux PM motors for low-
speed applications is presented. The goal is to find a design for a 4.4 kW,
50 rpm motor with a limited size and a low weight. The resulting designs
with q=1 are then compared to each other.
4.1 The design procedure
The geometry of a SMPM motor is completely described with 12 parameters,

as seen in section 3.1.1. To find these unknown parameters, an optimization
problem has been defined. Hence, the design of radial-flux PM motors is
conducted by solving an optimization problem, applying sequential quadratic
programming methods [39]. The goal of the optimization is to minimize
the weight of the motor and fulfill the requirements from table 1.2 and the
constraints that guarantee the required mechanical, thermal, and magnetic
behaviors. Appendix C gives some definitions related to optimization and
some explanations about how the optimization problem was defined.
The chart in figure 4.1 shows the different steps followed during the design
procedure, from the choice of the initial design variables to the calculation of
the active weight. It includes the influence of the different constraints as well
as appropriate references (to sections and equations). The active weight and
the constraints need to be expressed as a function of the design variables.
The objective function along with the variables and constraints are given in
the following sections.
43
4 Optimization procedure and comparison of designs with distributed windings
table 4.4
table 4.1
Subject to
constraints on:
Vary the parameters: Outer

dimensions
number of poles, length,
airgap length, slot and Rigidity of
magnet dimensions the structure
Express stator parameters

as a function of the variables
section 3.2.1
Maximum of flux density
in the air-gap
section 3.2.2
Flux density in:
stator yoke and Magnetic
saturation
stator teeth
equation (3.37)
Current loading
equation (3.39)
Current density Thermal

behaviour
Conductor number section 3.3.7
Copper losses Efficiency
Active weight
NO Optimal
Result?
YES
Verification with
FEM
Figure 4.1: Followed procedure to optimize the design of low-speed PM

motors.
44
4.1.1 Objective function
The objective function is the expression, depending on the design variables,

that is to be maximized or minimized. Different objective functions are
possible, for example:
• The permanent magnet weight, contributing considerably to the total

motor cost.
• The copper losses corresponding to the efficiency.
• The active motor weight, being important for the practicability of the
motor. For example, a light motor is easy to transport and install. The
active motor weight contributes also to the cost of the machine.
• A weighted combination of the three previous functions.
The chosen objective function is the active weight of the motor. It is to be

minimized. The copper losses and the permanent magnet weight are limited
by means of constraints.
The reasons for this choice are as follow. Primarily, it was considered more
challenging to fulfill the motor weight requirements than to achieve a high
efficiency, as the induction motor and the gearbox together have a quite low
efficiency. Besides, a low motor weight is a better sales argument than a
high motor efficiency. The permanent magnet weight is on the other hand
not convenient as objective function because the machine weight is then not
minimized.
4.1.2 Design variables and their ranges
The design variables of the optimization problem are the number of poles,
the number of slots per pole per phase, geometrical parameters that define
the stator teeth and the magnets, the rotor diameter, the airgap length and
the machine length. The variables and their ranges are given in table 4.1.
45
The followings remarks complete the table:
• The expression for the minimum length of the airgap was found in [40].
• An upper limit is set for the thickness of the permanent magnets to

avoid too high flux leakage between two neighboring permanent mag-
nets.
• The expression of the lower limit of the half pole angle was taken from
[19].
• The lower limits for the stator tooth width and the ratio of the stator
slot opening to the slot width (kopen ) are set with the intention to
avoid too thin teeth and slot openings that would be impossible to
manufacture.
• For the tangentially-magnetized PM motor, the design variables that

lm p
describe the permanent magnets are the half pole angle α = D rc
and
the magnet thickness wm .
Table 4.1: Design variables and their ranges

Variable Symbol Range
Number of poles p p ≥ 20 and p is even
Number of slots per pole per phase q q≤1
Rotor diameter Drc Drc ≤ Do
Airgap length δ 0.0002 + 0.003 DL
2 ≤ δ [m]
Magnet thickness: surface, V-shaped lm 2 mm ≤ lm ≤ 8 mm
tangentially-magnetized wm 2 mm ≤ wm
Half pole angle in electrical degrees α 0.7π/2 ≥ α
Outer stator diameter Do Do ≤ 500 mm
Stator tooth width bts bts ≥ 2.5 mm
Stator slot height hss hss ≥ 0
Stator slot opening over slot width kopen 0.2 ≤ kopen
Slot wedge height hsw hsw ≥ 1 mm
Machine length L 100 mm ≤ L ≤ 500 mm
V-angle between magnets αv αv ≤ π
46
4.1.3 Given constants
The constants given in table 4.2 are used in the design procedure. The
considered permanent magnet material is Neodymium Iron Boron (NdFeB).
Table 4.2: Constants.

Rated speed 50 rpm
Motor Rated torque 840 Nm
Rated power 4.4 kW
Permanent Remanence flux density 1.08 T
magnet Relative permeability 1.03
Iron Stacking factor of the stator iron laminations 1
Stator slot fill factor fs [table 4.3]
Winding End-winding factor kcoil [table 4.3]
Copper temperature 80 ◦ C
Magnet density 7500 kg/m3
Material Iron density 7750 kg/m3
properties Copper density 8920 kg/m3
Copper resistivity at 20 ◦ C 1.72e − 8 Ω/m
Flux density in the rotor yoke 1.4 T
The end-winding factor kcoil used in equation (3.31) and the slot fill
factor fs depend on the type of winding. These values are given in table 4.3
for distributed windings as well as concentrated windings with single or
double layer [37], [41].
Table 4.3: Winding constants.

Distributed Concentrated Concentrated
winding 1-layer winding 2-layer winding
kcoil 1.6Qs /p 1.46 0.93
fs 0.45 0.60 0.60
The magnetization curve of the iron material (LOSIL500) used in the

FEM simulations is given in figure 4.2.
47
1.8
1.6
1.4
Flux density [T]
1.2
0.8
0.6
0.4
0.2
0
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
Magnetic field [A/m]
Figure 4.2: Magnetization curve of the iron material (Losil500).
For the SMPM and inset PM motors, the flux density in the rotor
yoke Bry is set constant in order to limit the number of design variables. In
this way, Bry determines the inner rotor diameter Di (4.1), which thus does
not need to be a design variable.
Bm 2α p2 (D − δ)
Di = Drc − (4.1)
Bry
4.1.4 Constraints
The design constraints are described in table 4.4. They guarantee the re-
quired mechanical, thermal and magnetic behaviors. A constraint is set on
the maximum weight of the permanent magnets and on the maximum copper
losses. In addition to these constraints, it is checked for each design that the
permanent magnets are not subject to a too high negative magnetic field that
reduces their flux density and irreversibly demagnetize them. The followed
method is described in [36] and does not take the influence of the tempera-
ture into account.
In the next section, the optimization results obtained with these constraints
are presented.
48
Table 4.4: Constraints.
Objective of the constraints Description of the constraint Equation Reference
stator yoke height at least half the slot height hsy ≥ hss /2 [19]
slot width between 0.15 and 0.5 times 0.15hss ≤ bss2 ≤ 0.5hss [19]
guarantee the rigidity the slot height
of the mechanical structure tooth width at least 30 % of the slot pitch bts ≥ 0.3τs [19]
slot opening width at least 2 mm bss1 kopen ≥ 2 mm
slot opening height at least 2 mm hso ≥ 2 mm
flux density in stator teeth under 1.6 T Bts ≤ 1.6 T from the iron
flux density in stator yoke under 1.4 T Bsy ≤ 1.4 T magnetization
avoid magnetic saturation
flux density in rotor yoke under 1.4 T Bry ≤ 1.4 T curve
fundamental airgap flux density under 1.1 T B̂δ ≤ 1.1 T
prevent high temperature and winding temperature set to 80 ◦ C Tcu = 80 ◦ C Appendix D
guarantee a minimal efficiency copper losses under 700 W Pcu ≤ 700 W figure 4.4
limit the price magnet weight under 5.5 kg mmag ≤ 5.5 kg figure 4.3
limit the machine weight machine weight under 150 kg mtot ≤ 150 kg table 1.2
49
4.2 Results and comparison of designs with distributed

windings
In this section, the results obtained by following the design optimization pro-
cedure are presented. The designs that will be compared have a distributed
stator winding with either q = 1 or q = 2. In the first part, explanations
regarding the choice of some constraints will be given. In the second part,
the different designs will be compared.
4.2.1 Influence of some of the constraints
Maximum permanent magnet weight
Since the price of PM material is high compared to the rest of the machine
materials, the weight of the permanent magnets should be limited as much as
possible. However, the lower their weight, the higher will be the total active
weight (figure 4.3). A lower flux from the permanent magnets is indeed
compensated by a higher armature reaction, resulting in bigger slots and
consequently a bigger stator. As a compromise, the maximum permanent
magnet weight is chosen as 5.5 kg.
125
120
115
Active weight [kg]
110
105
100
95
90
85
3 4 5 6 7 8 9 10
Magnet weight [kg]
Figure 4.3: Active weight as a function of the permanent magnet weight

for optimized 60-pole SMPM motors.
50
4.2 Results and comparison of designs with distributed windings
Maximum copper losses
The copper losses should be as low as possible to assure a good efficiency.

However, the lower the copper losses, the higher will be the active weight of
the machine (figure 4.4). A compromise has to be found between efficiency
and active weight of the machine. The limit for the copper losses is set to
700 W.
240
220
p = 60
200 p = 30
180
Active weight [kg]
160
140
120
100
80
60
40
400 600 700 800 1000 1200 1400 1600
Copper losses [W]
Figure 4.4: Active weight of optimized SMPM motor designs as a function

of the copper losses.
Constraints on the rigidity of the structure
In [19], a 45 kW, 600 rpm buried V-shaped PM motor was investigated. The
same constraints on the rigidity of the structure as in [19] were considered.
These constraints are thus not optimized for the specific design requirements.
Figure 4.5 gives an idea of what happens without constraints on the rigidity
of the structure. It shows two designs of a 50-pole SMPM machine obtained
by following the optimization procedure. The first one respects all the con-
straints described previously. The second one does not respect the constraints
on the rigidity of the structure. The teeth are very long and thin and the
stator back is very thin. In this case, the flux leakage through the teeth is
slightly higher. The second design is also much shorter.
51
[mm]
40 40
35 L=191 mm 35 L=134 mm
30 30
25 25
20 20
15 15
10 10
5 5
0 0
-5 -5
180 190 200 210 220 230 240 250 180 190 200 210 220 230 240 250
[mm]
Figure 4.5: Two optimized designs of a 50-pole SMPM motor, with and
without the mechanical constraints on the structure.
Copper temperature
The copper temperature is determined by the copper losses, the environ-

mental temperature and the cooling of the motor. The copper temperature
was chosen equal to 80 ◦ C as described in appendix D. Figure 4.6 shows the
active weight as a function of the copper temperature for optimized 60-pole
SMPM motors. The higher the copper temperature, the higher the resistiv-
ity of the copper and thus the copper resistance. This implies that the slot
area must become bigger for constant copper losses. Therefore the active
weight increases with an increasing copper temperature. If the estimation of
the copper temperature were incorrect, the curve of figure 4.6 shows how it
would affect the active weight of the motor.
4.3 Comparison of the configurations
In this section, the obtained designs from the optimization program for dif-
ferent radial-flux PM motors are compared to each other. First, the active
weight as a function of the pole number is provided for the different configu-
rations. Then, the torques of several selected machines, computed with FEM
simulations, are analyzed.
52
4.3 Comparison of the configurations
140
130
Active weight [kg]

120
110
100
90
80
60 80 100 120 140 160 180
Copper temperature [°C]
Figure 4.6: Active weight as a function of the copper temperature for op-
timized 60-pole SMPM motor designs.
4.3.1 Weight and pole number
Figure 4.7 shows the dependency of the total active weight on the number of
poles for different optimized designs. In the following, the resulting curves
are discussed.
Influence of the constraints
The total active weight of the inset PM and tangentially-magnetized PM

motors reaches a minimum at a pole number of 70 and 60 respectively. This is
due to the constraints that guarantee the rigidity of the structure (table 4.4).
As soon as the limit on one or more of these constraints is reached, the active
weight increases. For the inset PM designs for example, the tooth width
reaches its lower limit. Consequently, the length of the motor is increased to
fulfill the torque requirement and the active weight increases.
53
170
160 SMPM motor

Outer-rotor SMPM motor
150
Inset PM motor
Active weight [kg]
140 V-shaped PM motor with q=2

Tangentially-magnetized PM motor
130
120
110
100
90
80
25 30 35 40 45 50 55 60 65 70 75
Pole number
Figure 4.7: Active weight as a function of the number of poles for different
optimized designs.
Flux concentration
The topology that is the lightest for a pole number over 34 is the tangentially-
magnetized PM motor. This is due to the fact that the flux concentration in
the rotor allows a higher flux density in the airgap and thus a lower machine
length (table 4.5). Therefore, the flux concentration is a non-negligible ad-
vantage.
Table 4.5: Comparison of different motor configurations with p = 50.

Outer rotor Tangentially-
SMPM Inset PM
SMPM magnetized PM
Bδ [T] 0.911 0.857 0.906 1.12
L [mm] 191 179 199 144
54
V-shaped PM motors
Optimized designs of V-shaped PM motors with q = 1 were investigated

with FEM simulations. The obtained torque ripples were extremely high
(over 100 % of the nominal torque). This may be explained by the high
harmonic content in the airgap flux density, notably the amplitude of the
fifth harmonic that is over 25 % of the fundamental. By choosing q = 2,
the torque ripple becomes more reasonable. However there were no solutions
found for pole numbers higher than 30, as the teeth become too numerous,
too narrow and saturated (figure 4.7). Thus, V-shaped PM motor designs
are not suitable for low-speed direct drives.
Comparison inner/outer rotor
As can be seen in figure 4.7, the outer-rotor SMPM motor configuration is

lighter than the inner-rotor SMPM motor. Indeed, an outer-rotor geometry
allows a larger inner stator diameter and thus decreases the current loading
needed to obtain the required torque (3.37).
Comparison of SMPM and inset PM motors
Figure 4.7 also reveals that the SMPM and the inset PM motor configurations
have almost the same weight for equal permanent magnet weight and copper
losses. The inset PM rotor is heavier than the SMPM rotor due to the iron
pieces placed between the permanent magnets. However, the inset PM motor
is slightly lighter than the SMPM motor for pole numbers over 50. This is
due to the reluctance torque that contributes around 5 % to the required
nominal torque. This allows a lighter stator, and explains why the inset PM
motor can be lighter than the SMPM motor.
55
Choice of the pole number
Figure 4.7 can be used to choose an optimal pole number. However, it should
be kept in mind that the number of permanent magnets and the production
cost increase with a higher pole number. Therefore a compromise should be
found between the weight and the number of magnets.
For pole numbers higher than 50, the active weight decreases slower due to
the constraints on the structure. A minimum weight under 80 kg is reached
for the 60-pole tangentially-magnetized PM motor design.
4.3.2 Torque ripple
The torque is pulsating because of:
• The variation of permeance in the airgap, generating the cogging

torque.
• The interaction between the airgap flux and the space harmonics in the
magneto-motive force (MMF) depending on the winding layout.
• The interaction of the airgap flux with the current harmonics. The
current harmonics generated by voltage harmonics are neglected here,
as the supplied voltage is assumed to be ideally sinusoidal.
The torque ripple is obtained from a time-step constant-speed FEM simu-

lation (Appendix B). The simulated designs have the same winding layout
(distributed winding, q = 1) and have almost the same airgap length. Ta-
ble 4.6 provides the ratio between the torque ripple and the mean torque
for three different pole numbers and different motor topologies. The corre-
sponding curves are displayed in figure 4.8. The ratio between the cogging
torque and the mean torque for different 50-pole motor designs are provided
in table 4.7.
56
Table 4.6: Ratio between the torque ripple and the mean torque in %.
Outer-rotor Tangentially-
SMPM Inset PM
SMPM magnetized PM
p = 30 19 28 24 70
p = 50 16 27 20 46
p = 70 10 26 22 47
1.25
1.2
1.15
1.1
Torque [p.u.]
1.05
0.95
0.9
0.85
0.8
0.75
0 0.002 0.004 0.006 0.008 0.01 0.012 0.014 0.016 0.018 0.02
Time [s]
SMPM motor Inset PM motor
Outer-rotor SMPM motor Tangentially magnetized PM motor
Figure 4.8: FEM simulated torque for different configurations of 50-pole

PM motors, at nominal load.
Table 4.7: Ratio between the cogging torque and the mean torque in %.
SMPM Inset PM
SMPM magnetized PM
p = 50 3.9 4.2 2.6 12.9
57
The torque ripple varies between different configurations and pole numbers.
The SMPM motor designs have the lowest torque ripple. The very high
torque ripple of the tangentially-magnetized PM motors can partly be
explained by their higher cogging torque (table 4.7). As stated before, the
airgap flux plays also an important role in the torque pulsation. Therefore,
the relevant low-order harmonics in the airgap flux density are provided in
table 4.8. The difference of harmonic contents reflects well the difference in
the torque ripple if the contribution of the cogging torque is kept in mind.
For example, the torque ripple of the SMPM motor with 30 poles is higher
than the ripples of the 50- and 70-pole SMPM motors. This is reflected in
the amplitudes of their 3rd , 5th and 7th harmonics in the airgap flux density,
which are also higher.
Table 4.8: Relative values of the 3rd , 5th and 7th harmonics in the airgap
flux density in %.
SMPM Inset PM
SMPM magnetized PM
p = 30 5.0 0 1.5 16.1
rd
3 harmonic p = 50 5.0 0 0.8 13.1
p = 70 4.5 0 1.5 14.6
p = 30 15.2 18.9 17.9 8.7
5th harmonic p = 50 13.6 17.0 17.9 12.7
p = 70 13.4 16.0 18.1 12.2
p = 30 9.0 9.0 10.9 4.5
7th harmonic p = 50 4.5 5.6 7.2 2.3
p = 70 1.0 1.9 3.6 0.8
Table 4.9: Torque ripple and low-order harmonics in the airgap flux density
of two 50-pole SMPM motors with different pole angles
Torque 2α Harmonic no. [%] Active
ripple [%] [el. ◦ ] 3rd
5th 7th 9th 11th weight [kg]
22.7 120 0 16.1 3.5 0 6.4 101.2
16.3 126 5 13.6 4.5 2.2 4.9 100.7
58
The harmonic contents in the airgap flux-density can be influenced by

varying the pole angle. A pole angle of 120 electrical degrees allows to
eliminate the third harmonic, as it was done for the outer-rotor SMPM
motors in table 4.8. Table 4.9 shows the influence of the pole angle (2α)
on the harmonics in the airgap flux-density and on the torque ripple of
50-pole SMPM motors. The active motor weight is given to indicate which
design was the calculated optimal solution. As can be seen, the motor with
a pole angle of 120 electrical degrees has higher 5th and 11th harmonics than
the other simulated design and also a higher torque ripple. Therefore, the
torque ripple of SMPM motors can be optimized by finding a pole angle
that gives a compromise between the amplitudes of the 3rd and 5th harmonics.
For the outer-rotor SMPM motor, the calculated optimal design has a
pole angle that is lower than 120 electrical degrees, which should be avoided
according to table 4.10. Indeed, both the 3rd and the 5th harmonics are
very high for low pole angles, causing a high torque ripple. For outer-rotor
SMPM motors, the lower limit of the pole angle in the optimization program
is therefore changed to 120 electrical degrees to avoid very high torque ripples.
Table 4.10: Torque ripple and low-order harmonics in the airgap flux den-
sity of three 50-pole outer-rotor SMPM motors with different
pole angles
Torque 2α Harmonic no. [%] Active
ripple [%] [el. ◦ ] 3rd 5th 7th 9th 11th weight [kg]
49.7 110.4 8.8 19.6 13.2 4.0 6.6 89.7
26.7 120 0 17 5.6 0 6.6 90.5
27.0 126 5.3 15 7.2 3.0 5.6 91.8
For the tangentially-magnetized PM motors, the pole angle is also influencing

the torque ripple. However, no torque ripple lower than 40 % of the nominal
torque was found by trying different values of pole angles and decreasing the
cogging torque as much as possible.
For all investigated motor configurations, the torque ripple was found to be
relatively high (over 10 %), even though the pole angle was optimized for a
low torque ripple.
59
The torque ripple can be decreased by skewing the machine, which was not
investigated in this thesis. Another solution that will be investigated in
chapter 5 is to use a concentrated winding with a good combination between
the pole number and the slot number. The influence of different stator tooth
widths on the torque ripple is also considered in chapter 5.
4.3.3 Iron losses
The iron losses are calculated using FEM simulations as described in ap-
pendix B. No big difference between the different designs can be noticed and
iron losses of about 120 W have been found, which is around one sixth of the
value of the copper losses.
4.3.4 Summary
Different motor configurations were compared for stators with distributed

windings with q = 1 and q = 2. Several interesting aspects have been pointed
out.
The V-shaped PM motors are not adapted for high pole numbers (over 30)
because of a too high torque ripple.
Inner- and outer-rotor SMPM motor, inset PM motor and tangentially-
magnetized PM motor designs fulfill the given constraints. The calculated
active weights are between 80 and 100 kg, which is far below the limit of
150 kg. Since the inset PM motor is nearly equivalent to the SMPM motor
in terms of active weight and performance, it will not be studied further.
However, all designs have the same drawback: a very high torque ripple.
The torque ripple was not taken into account when defining the constraints,
because it is difficult to calculate analytically. Nevertheless, it should be as
small as possible to guarantee the motor a maximum life time. Therefore,
the motor designs need to be improved. A good way to decrease the torque
ripple, besides skewing, is the use of concentrated windings.
Concentrated windings for inner- and outer-rotor SMPM motors as well as
tangentially-magnetized PM motors are studied in the next chapters. V-
shaped PM motor and inset PM motor designs are not considered further.
60
5 Concentrated windings
In this thesis, a concentrated winding is a winding that is concentrated around

the teeth. It has no overlapping end-windings. Several advantages make the
concentrated windings interesting:
• Their end-windings are much shorter than those of distributed wind-

ings. As a consequence, the copper losses are lower.
• They can be mounted very easily around the teeth since the end-
windings are not overlapping. This simplifies the production and re-
duces the cost.
• Some configurations enable a low torque ripple.
However, concentrated windings have also some drawbacks. The number of

poles and the number of slots have to be chosen carefully. Otherwise, the
winding factor may be low, the torque ripple and the rotor losses high and
the machine may be very noisy.
Concentrated windings are not a recent discovery; they are first men-
tioned in a patent from the year 1895, [42]. In the nineties they were used in
sub-fractional power machines with a reduced number of phases. The studies
for larger three-phase machines started in 2002 to become very trendy. In
2002, J. Cros and P. Viarouge showed how to find winding layouts that have
high winding factors [43]. The applied methods were to use an irregular
teeth distribution or to find a good combination of pole and slot numbers.
Winding factors for motors with up to 22 poles were presented. The same
year, Th. Koch and A. Binder compared two designs of PM machines for
electric traction, one with p = 28, q = 0.25 and irregularly distributed
teeth, one with p = 28, q = 0.5 and double-layer windings [44]. In 2003,
61
F. Magnussen calculated the winding factors for one-layer and double-layer

windings up to 16 poles and compared some designs [37]. In [45], three
designs were compared with 14 poles and 42, 21 and 15 slots. Torque ripple,
losses and thermal behavior were investigated. In [46], D. Ishak looked at
the combinations of poles and stator slots linked by the relation p = Qs ± 2.
Concentrated windings were also studied in combination with soft magnetic
composite (SMC) materials by A.G Jack [41], [47]. Different pole-slot
number combinations were compared for an 8-pole motor in [47]. Some
multi-phase concentrated windings were investigated in [48] and [49].
Concentrated windings are appropriate for PM machines with high

pole numbers and therefore highly interesting for low-speed direct drives. As
mentioned previously, some pole-slot number combinations were studied for
pole numbers lower than 20. In this chapter, the main goal is to find winding
configurations that give high winding factors for pole numbers higher than
20. Winding configurations with high torque ripples, excessive rotor losses
or noise should be avoided. The method with irregularly distributed teeth
described in [43] is tested first and then pole-slot number combinations for
pole numbers over 20 are investigated. Some hints are given to avoid bad
combinations.
5.1 Irregularly distributed teeth
A first possibility to obtain a high winding factor with a concentrated winding

is to have an irregular distribution of teeth [43]. The windings investigated
in this part have a number of slots per pole per phase equal to 0.5, which is
easier to handle because of the symmetries in these windings.
5.1.1 Winding factor
The fundamental winding factor can be calculated using the MMF waves.
Figure 5.1 shows the MMF wave created by the stator currents at the instant
where the current in phase A is maximal and equal to the rated peak current
ˆ The currents in phases B and C are equal to −I/2.
I. ˆ In the first case, the
62
5.1 Irregularly distributed teeth
teeth are of equal size, in the second case the teeth are irregularly distributed.
The fundamental winding factor kw1 is equal to the ratio of the amplitude
of the MMF harmonic corresponding to the number of poles, divided by the
fundamental of the MMF of the distributed winding with q = 1. The winding
factor has been calculated for different cases of irregularly distributed teeth
with q = 0.5. Table 5.1 presents the obtained results for different ratios
between the smaller tooth width bt1 and the wider one bt2 .
Case 1: B- A+ A- C+ C- B+
bt1 /bt2 = 1
N S N S
Case 2:
B- A+ A- C+ C- B+
bt1 /bt2= 0.4
N S N S
MMF
nsI case 1
case 2
ns I/2
el. angle
2π 4π
Figure 5.1: Calculation of the winding factor for two different cases of ir-
regularly distributed teeth with q = 0.5.
Table 5.1: Winding factors for different cases of irregularly distributed teeth
with q = 0.5.
Tooth ratio bt1 /bt2 1 0.8 0.7 0.6 0.5 0.4
Winding factor 0.866 0.894 0.909 0.924 0.940 0.956
63
5.1.2 Influence on the torque ripple
The load torques of a 64-pole SMPM motor were computed for different
teeth distributions at the same current loading. The results are given in
table 5.2. As can be seen, the special teeth distribution allows not only a
better winding factor but also a lower torque ripple. However, the lowest
torque ripple for a tooth ratio of 0.4 is still quite high with 13 % of the
nominal torque. The tooth ratio was not checked under 0.4 as the teeth are
already very saturated and thin for this value. As will be shown in the next
section, lower torque ripples can be achieved with a good slot-pole number
combination, and regularly distributed teeth.
Table 5.2: Torque values for different teeth distribution.

Tooth Winding Torque Torque
ratio bt1 /bt2 factor kw1 [p.u.] ripple [%]
1 0.866 0.90 49
0.8 0.894 0.94 46
0.6 0.924 0.98 32
0.4 0.956 1.00 13
5.2 Combinations of pole and slot numbers
The aim of this section is to sort out the different winding layouts and possible
slot-pole number combinations in order to simplify the choice of a layout. For
this purpose, the winding factors of the investigated concentrated windings,
the harmonic contents of the respective MMFs, their torque ripples and radial
magnetic forces are calculated and analyzed for some promising slot-pole
number combinations.
64
5.2.1 Winding layouts and winding factors
Determination of the winding layout for a double-layer winding
Double-layer windings have been investigated since they have better proper-
ties such as shorter end-windings and more sinusoidal back-EMF waveforms
compared to single-layer windings [37], [46]. For different slot-pole number
combinations, the winding layout, i.e. the placement of the conductors
of each phase in the slots, is determined. The method presented in [43]
describes how to obtain the layout that gives the highest winding factor for
a given pole number p and slot number Qs . The method is based on the
decomposition of the number of slots per pole per phase q. It is similar to
the method used for the large synchronous machines with a fractional value
of q [40]. The method is described in figure 5.2 using Qs = 24 and p = 26 as
an example.
4 4 x "1"
a) q= 9 x "0"
13
b) 1001001001000
3p x ("1" or "0")
c) 1 0 0 1 0 0 1 0 0 1 0 0 0 1 0 0 1 0 0 1 0 0 1 0 0 0 1 0 0 1 0 0 ......
A C' B A' C B' A C' B A' C B' A C' B A' C B' A C' B A' C B' A C' B A' C B' A C' ......
d) A A' A A' C' C C' .... A' A A' A C .....
slot
B' A A' A' A A A' A' A C' C C C' C' .... B A' A A A' A' A A A' C.....
number 1 2 3 4 5 6 7 13 14 15 16 17
e) S = [1 -2 -2 3 3 -4 -4 5 -13 14 14 -15 -15 16 16 -17]
Figure 5.2: Determination of the winding layout for Qs =24 and p = 26.
a) The number of slots per pole per phase q is written as a fraction which
is cancelled down to its lowest terms: q = n/d = 4/13 where n and d
are integers.
b) A sequence of d − n = 9 zeros (”0”) and n = 4 ones (”1”) is found, the

ones being distributed in the sequence as regularly as possible.
65
c) The found sequence is repeated 3p/d = Qs /n = 6 times. It is compared

to the layout of the distributed winding with 3p slots and q = 1.
d) Those conductors from the distributed winding that correspond to the

”1” are kept and form one layer of the double-layer concentrated wind-
ing. The second winding layer is obtained by writing the corresponding
return conductor on the other side of the tooth for every obtained con-
ductor, i.e. A’ for A.
e) A vector S is written to describe the layout of phase A. It will be used

to calculate the winding factor. Thereby, the slots are numbered from
1 to Qs . The vector S consists of the numbers corresponding to these
slots that contain conductors of the phase A. If both layers of one slot
contain conductors of phase A, the number of the slot is written twice
in the vector. S has thus 2Qs /3 elements. For conductors A’, a minus
is added to the corresponding slot number.
Method used to calculate the winding factor
The winding factor is calculated using the EMF phasors [37]. The corre-
−
→
sponding EMF phasor Ei of conductor i is:
−
→ j πp |S(i)|
Ei = sign(S(i))e Qs (5.1)
The fundamental winding factor kw1 can then be calculated:

2Qs /3
1 − →
kw1 = Ei (5.2)
nl Qs /3
i=1
where i is an element of S and nl is the number of layers (nl = 2).

For the example with Qs = 24 and p = 26, the sum of the EMF phasors is:
16
−
→ j πp −2j πp 3j πp −4j πp 5j πp πp
−13j Q
Ei = e Qs + 2e Qs + 2e Qs + 2e Qs + e Qs + e s
i=1
πp πp πp πp
14j Q −15j Q 16j Q −17j Q
+2e s + 2e s + 2e s +e s (5.3)
This results in a winding factor of 0.95.
66
Calculation of the winding factor
Winding factors for machines from 4 to 80 poles and 6 to 90 slots are cal-
culated. This represents a total of 935 pole-slot number combinations. Fig-
ures 5.4 and 5.5 show the fundamental winding factors calculated for double-
layer concentrated windings with p between 20 and 80, and Qs between 15
and 90, with Qs being a multiple of 3. The method presented above to find
the winding layout with the highest winding factors does not need to be ap-
plied for every single combination. Indeed, winding factors of some pole-slot
number combinations can be found directly:
• Some combinations do not give a balanced three-phase winding. It

is the combinations where the denominator d (q = n/d) is a multi-
ple of the number of phases. They are colored red in figures 5.4 and 5.5.
• Combinations with the same number of slots per pole per phase have
equal winding factors. These winding layouts consist of the same basic
sequence reproduced a certain number of times to fill the required
number of slots (see table 5.3)
• For each Qs , i.e. each line of the table, there is a periodicity of 2Qs :
The winding layout and factors of the pole-slot number combinations
with Qs slots and p + 2kQs poles (k=0,1,2...) are identical. This is
easily shown with equations (5.2) and (5.4):
pπ pπ π(p+2Qs )
jQ S(i) jQ S(i) j2πS(i) j S(i)
∀i : e s =e s e = e Qs (5.4)
=1
• For each Qs , i.e. each line of the table, there is a symmetry around
kQs (k = 1,2,3...). Pole-slot number combinations with p = Qs − k and
p = Qs + k, (k = 1,2,3...) have an identical layout and winding factor.
67
This is proved below:
j
p πS(i) p πS(i) p πS(i)
e Qs = cos +j sin
Qs Qs
i i i
⎡ 2 2 ⎤1/2
p πS(i) p πS(i)
=⎣ cos + sin ⎦ (5.5)
Qs Qs
i i
For p = Qs − k, the sum of cosinus and sinus terms can be rewritten

as:
⎛ ⎞
π(Qs − k)S(i) ⎜ ⎟
cos = ⎜cos(πS(i)) cos πkS(i) + sin(πi) sin πkS(i) ⎟
Qs ⎝ Qs Qs ⎠
i i
=0
π(Qs + k)S(i)
= cos (5.6)
Qs
i
⎛ ⎞
π(Qs − k)S(i) ⎜ ⎟
sin = ⎜sin(πS(i)) cos πkS(i) − sin πkS(i) cos(πS(i))⎟
Qs ⎝ Qs Qs ⎠
i i
=0
π(Qs + k)S(i)
=− sin (5.7)
Qs
i
This leads to:

j
π(Qs −k)
S(i)
j
π(Qs +k)
S(i)
e Qs = e Qs (5.8)
i i
which means that the winding factors for p = Qs − k and p = Qs + k,

with k = 1,2,3... are identical.
68
• It has been noticed that the winding factor increases and decreases as
shown in figure 5.3. Figure 5.3 shows also the symmetries and period-
icity described previously.
Qs even
2 4
=Qs Q + Q + 2Q
s
3Q
s
s
s
p p = p = p= p=
Qs kw1 kw1
Qs odd
-1 +1 -1 +1
s 3Q s
Q =Q 2Q s 3 Q
s
p = p p= p= p=
Qs kw1 kw1
Figure 5.3: Evolution of the winding factor kw1 for even and odd slot num-
bers Qs and different pole numbers p
(Boxes filled with the same pattern have the same winding factor and
layout. Black boxes are combinations where concentrated windings are not
possible.)
Figures 5.4 and 5.5 show that winding factors up to 0.954 can be reached.
Some interesting winding layouts are described in table 5.3. In the following,
the windings presenting high winding factors where p and Qs are approxi-
mately equal will be investigated further.
69
70
Qs\p 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40
6 0.866 0.866 0.5 0.5 0.866 0.866 0.5 0.5 0.866 0.866 0.5 0.5 0.866
9 0.617 0.866 0.945 0.945 0.866 0.617 0.328 0.328 0.617 0.866 0.945 0.945 0.866 0.617 0.328 0.328 0.617
12 q=1 0.866 0.933 0.933 0.866 0.866 0.933 0.933 0.866
15 0.621 0.866 0.951 0.951 0.866 0.621 0.621 0.866
18 q=1 0.647 0.866 0.902 0.945 0.945 0.902 0.866 0.647
21 0.866 0.89 0.953 0.953 0.89 0.866
24 q=1 0.76 0.866 0.933 0.95 0.95 0.933 0.866 0.76
27 0.866 0.877 0.915 0.945 0.954 0.954 0.945 0.915 0.877 0.866
30 q=1 0.866 0.874 0.936 0.951 0.951 0.936 0.874 0.866

33 0.866 0.903 0.928 0.954 0.954 0.928 0.903
36 q=1 0.866 0.867 0.902 0.933 0.945 0.953 0.953 0.945
39 0.866 0.863 0.918 0.936 0.954 0.954
42 q=1 0.866 0.89 0.913 0.945 0.953
45 0.866 0.859 0.886 0.927 0.945
48 q=1 0.866 0.857 0.905 0.933
51 0.866 0.88 0.901
54 q=1 0.866 0.854 0.877
57 0.866 0.852
60 q=1 0.866
q=1/2, 1/4 q=3/8, 3/10 Qs=21+6k, p=Qs ±1, k = 0, 1, 2 …

q=3/7, 3/11 q=5/14, 5/16 Qs=24+6k, p=Qs ±2, k = 0, 1, 2 …
q=2/5, 2/7 not appropriate kw1<0.866
Figure 5.4: Winding factors for different pole-slot number combinations.

Qs\p 42 44 46 48 50 52 54 56 58 60 62 64 66 68 70 72 74 76 78 80
33 0.866
36 0.933 0.902 0.867 0.866
39 0.936 0.918 0.863 0.866
42 0.953 0.945 0.913 0.89 0.866
45 0.951 0.955 0.955 0.951 0.945 0.927 0.886 0.859 0.866
48 0.95 0.954 0.954 0.95 0.933 0.905 0.857 0.866
51 0.933 0.944 0.955 0.955 0.944 0.933 0.901 0.88 0.866
54 0.902 0.915 0.93 0.945 0.949 0.954 0.954 0.949 0.945 0.93 0.915 0.902 0.877 0.854 0.866
57 0.932 0.912 0.937 0.946 0.955 0.955 0.946 0.937 0.912 0.932 0.852 0.866
60 0.874 0.892 0.933 0.936 0.951 0.954 0.954 0.951 0.936 0.933 0.892 0.874 0.866
63 0.866 0.85 0.871 0.89 0.905 0.919 0.945 0.948 0.953 0.955 0.955 0.953 0.948 0.945 0.919 0.905 0.89 0.871
66 0.866 0.849 0.887 0.903 0.928 0.938 0.951 0.954 0.954 0.951 0.938 0.928 0.903
69 0.866 0.867 0.884 0.914 0.925 0.943 0.949 0.955 0.955 0.949 0.943 0.925
72 0.866 0.847 0.867 0.902 0.911 0.933 0.933 0.945 0.95 0.953 0.954 0.954 0.953 0.95 0.945
75 0.866 0.846 0.88 0.895 0.92 0.93 0.945 0.951 0.955 0.955 0.951
78 0.866 0.863 0.879 0.906 0.918 0.936 0.943 0.952 0.954 0.954
81 0.866 0.845 0.86 0.877 0.89 0.904 0.915 0.925 0.933 0.945 0.946 0.951 0.954 0.955
84 0.866 0.845 0.876 0.89 0.913 0.933 0.939 0.945 0.953
87 0.866 0.859 0.874 0.899 0.91 0.929 0.936 0.947
90 0.866 0.843 0.859 0.874 0.886 0.902 0.918 0.927 0.936 0.945
Figure 5.5: Winding factors for different pole-slot number combinations, continued.
71
72
Table 5.3: Winding layouts for some pole-slot number combinations
(A’ stands for the return conductor corresponding to conductor A).
Qs , p or q Winding layout
q = 2/5, 2/7 ...|C A|A A |AB |BB|B C|C C |CA |AA|A B|B B |BC |CC|...
q = 3/8, 3/10 ...|C A|A A |AA|A B|B B |BB|B C|C C |CC|...
q = 3/7, 3/11 ...|C A|AB |BB|B C |CA|A A |AB |BC |CC|C A|A B|B B |BC |CA |AA|A B|B C|C C |...
q = 5/14, 5/16 ...|C A|A A |AA|A A |AA|A B|B B |BB|B B |BB|B C|C C |CC|C C |CC|...
Qs = 12 + 6k, p = Qs ± 2
if p/2 even ... A|A A |A...A|A A |A B |BB|B ...B |BB|B C|C C |C...C|C C |C ...

Qs /6 Qs /6 Qs /6
if p/2 odd ...C
A|A A |A...A |AA|A B |BB|B ...B|B B |B C|C C |C...C |CC|C A |...|A B|...|B C

Qs /6 Qs /6 Qs /6 Qs /6 Qs /6 Qs /6

Qs = 9 + 6k, p = Qs ± 1 A|A A |A...A |AA|A B|B B |B...B |BB|B C|C C |C...C |CC|C

Qs /3 Qs /3 Qs /3
5.2.2 MMF and harmonics
Analysing the MMF and its harmonics is of interest since it can cause
extra iron losses in the rotor compared to distributed windings. The MMF
is calculated analytically with the method fully described in [37]. The
harmonics are calculated by taking the periodicity of the MMF waveform
into account, which corresponds to the number of symmetries in the winding.
Independently of the number of poles, one period of the MMF waveform is
taken as the base time frame. This has the advantage that the harmonic
orders are integers. The harmonic interacting with the permanent magnet
flux and producing the average value of the torque is then the harmonic,
whose number is equal to the ratio between the number of poles and the
number of periods in the MMF.
Figure 5.6 shows the harmonic content in the MMF for two machines
with double-layer concentrated windings with 48 slots/40 poles (q = 2/5)
and 51 slots/50 poles. The torque harmonics are the 10th and the 50th
harmonics, for the 40-pole and 50-pole motor designs respectively. As can be
seen in the figures, the MMFs contain many harmonics with high amplitude.
The winding without symmetries also contains many more harmonics than
the one with symmetries (figure 5.6).
0.9
torque
0.9 torque
0.8 harmonic
harmonic
0.7 0.8
0.7
0.6
0.6
0.5
0.5
0.4
0.4
0.3
0.3
0.2 0.2
0.1 0.1
0 0
10 20 30 40 50 60 70 80 90 100 20 40 60 80 100 120 140 160 180 200
Harmonic order Harmonic order
Figure 5.6: MMF harmonics in double-layer concentrated windings with 48

slots and 40 poles (q = 2/5) on the left and 51 slots and 50 poles
on the right.
73
0.8
Normal component of flux density [T]

0.78
0.76
0.74
0.72
0.7
0.68
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Time [s]
Figure 5.7: Time variation of the normal component of the total flux den-
sity in the middle of the rotor iron between two permanent
magnets, for a motor with Qs = 72 and p = 64.
Due to MMF harmonics, alternating magnetic fields appear in the rotor. As

can be seen in figure 5.7, the flux density in the rotor is not constant. This
gives rise to eddy currents in the permanent magnets as well as iron losses in
the rotor iron. Therefore, these losses should be estimated during the design
process when concentrated windings are applied. If they are too high, a
laminated rotor and permanent magnets in smaller pieces can decrease the
losses.
5.2.3 Torque ripple
Cogging torque
A very low cogging torque can be obtained if the slot and pole numbers are
chosen so that the least common multiple (LCM) between them is large [37].
Windings with q = 2/5 and q = 3/8 have a lower LCM than other windings
and consequently have a higher cogging torque (table 5.4). The closer the
number of slots to the number of poles, the higher the LCM. Even though
the combinations p = Qs − k and p = Qs + k have equal winding factors, the
LCM is higher for p = Qs + k than for p = Qs − k.
74
Table 5.4: Lowest common multiple and cogging torque in % of the rated
torque of different SMPM motors.
Cogging
p, Qs , q LCM
torque [%]
p = 60, Qs = 72, q = 2/5 360 1.4
p = 64, Qs = 72, q = 3/8 576 0.3
p = 64, Qs = 60, q = 5/14 960 0.03
p = 64, Qs = 66, q = 11/32 2112 0.003
p = 62, Qs = 63, q = 21/62 3906 0.003
Torque ripple
FEM simulations were run to compute the torque for different winding lay-
outs. Thereby, an important aspect is to determine the initial position of the
rotor in relation to the applied currents. For a SMPM motor, the maximum
torque is achieved when the angle β between the current vector and the PM
flux vector is 90 electrical degrees. Choosing the current in phase A equal to
zero at the initial rotor position, means that the airgap flux density should
be at its maximum. Thus, the permanent magnet should be aligned to the
tooth or the slot that is the axis of symmetry of the phase A coils, as shown
in figure 5.8. Assuming that the first concentrated coil of phase A is around
tooth number 1, and the teeth are numbered according to figure 5.8, then
the tooth that should be aligned to a permanent magnet can be found in
table 5.5 for different winding configurations. When the tooth number is not
an integer, it means that the magnet faces a slot.
SMPM motors with different pole and slot numbers as those described in
table 5.5 were simulated at nominal load. The results show that they have a
torque ripple between 2 and 5 % of the rated torque (table 5.6). It can be no-
ticed that a design with a high cogging torque can have a lower torque ripple
than a design with a lower cogging torque, as shown in tables 5.4 and 5.6.
Whatever the investigated slot-pole number combination, the high pole-
number PM machines with concentrated windings will have a much lower
torque ripple than machines with distributed windings q = 1 without skew-
ing. These machines can have a torque ripple over 15 % as was shown in
chapter 4.
75
arbitrary position:
Ia
1 2 3 4 5 6 7
B'A A'A' A A A'A' A A A'C C'C' C...
Ia Bδ
Ψm
θ initial position
at t=0 with Ia=0: Ia
β=90 1 2 3 4 5 6 7
B'A A'A' A A A'A' A A A'C C'C' C...
Ψm
Figure 5.8: Initial position of a SMPM motor. Ia is the current in phase

A, Bδ is the airgap flux density.
Table 5.5: Determiniation of the initial rotor position.

Qs , p or q Tooth number
q = 2/5 4.5
q = 3/8 2
Qs = 12 + 6k, k = 0,1,2... Qs 1
12 + 2
p = Qs ± 2, p even
Qs = 12 + 6k, k = 0,1,2... Qs 1
3 + 2
p = Qs ± 2, p odd
Qs = 9 + 6k, k = 0,1,2... Qs 1
6 + 2
p = Qs ± 1
Table 5.6: Torque ripple of different SMPM motors.

Torque
p, Qs , q
ripple [%]
p = 60, Qs = 72, q = 2/5 2.6
p = 64, Qs = 72, q = 3/8 3.4
p = 64, Qs = 60, q = 5/14 3.2
p = 64, Qs = 66, q = 11/32 3.4
p = 62, Qs = 63, q = 21/62 4.4
76
5.2.4 Magnetic noise
Magnetic noise results from magnetic forces that make the stator vibrate.
Radial forces are undesirable attractive forces between the stator and the
rotor while tangential forces act on the rotor to produce the torque. The
forces that cause magnetic noise are mostly the radial forces [50]. FEM
simulations are used to calculate the radial magnetic forces.
The computation of the radial magnetic forces is based on the Maxwell

Stress method. For this computation, only the normal component of the
flux density is taken into account (5.9).
1
σ(θ,t) = Bn (θ,t)2 (5.9)
2μ0
where σ is the radial magnetic force density as a function of the angular
coordinate θ and the time t, Bn is the normal, i.e. radial component of the
airgap flux density.
The tangential component of the airgap flux density is neglected, since the
permeability of the iron material is much higher than the one of air. Indeed,
the flux lines enter and leave the stator respectively rotor surface almost
perpendicularly.
Ph Ph
as as
eA eA
se B
Pha
Ph
as
eB
C C
ase ase
Ph Ph
Figure 5.9: Radial magnetic forces on the stator of a SMPM motor with 68
poles and 69 slots, at different times.
77
Phase C
Phase B
Phase A
Phase C
Phase B
Phase A
Figure 5.10: Radial magnetic forces on the stator of a SMPM motor with
60 poles and 72 slots.
Figure 5.9 shows the radial magnetic force densities on the stator sur-
face of a SMPM motor with 68 poles and 69 slots at two different instances.
It can be seen that the distribution of the magnetic forces on the stator
is not symmetrical. The sum of these magnetic forces gives a resulting
unidirectional pulling force that rotates with time and generates noise and
vibration in the machine. This resulting force is due to the asymmetry in
the windings. The pole-slot number combinations giving winding layouts
without any symmetry such as combinations with Qs = 9 + 6k (k = 0,1,2...)
and p = Qs ± 1 are therefore not recommended. The radial magnetic
forces for an SMPM motor design with 60 poles and 72 slots are shown in
figure 5.10. As can be seen, its winding layout has a symmetry in one sixth
of the machine. Therefore, there is no resulting force for this design, as the
radial magnetic forces compensate each other.
78
5.3 Summary
5.3 Summary
Concentrated windings with q = 0.5 and irregular teeth distribution were

investigated for high pole numbers.
The winding factors for different pole-slot number combinations were

calculated. Some of the promising winding layouts with high winding factors
were described. By choosing one of these layouts, the performances of the
machine can be improved considerably. However, it should be taken into
account that the concentrated windings cause alternating magnetic fields
in the rotor giving rise to rotor losses. Depending on the pole-slot number
combinations, the cogging torque can be reduced. The torque ripple at
load is harder to predict from the slot-pole number combinations, but it
was always very low compared to a distributed winding machine without
skewing and q = 1. Magnetic noise is also an important issue for some PM
machines with concentrated windings. The machine designs with winding
layouts without any symmetry have unbalanced radial magnetic forces that
make the stator vibrate and produce noise. Slot-pole number combinations
with high winding factors and winding layouts with many symmetries give
the best performances and avoid high magnetic noise.
Using a carefully chosen slot-pole number combination is more advan-

tageous than using an irregular teeth distribution in terms of torque ripple
and winding factor. Therefore, only results of designs with suitable slot-pole
number combinations will be investigated in the next chapter.
79
80
6 Results and comparison of designs
with concentrated windings
In this chapter, the concentrated windings are tested on SMPM, outer-rotor

PM and tangentially-magnetized PM motors, which are the most promising
motor designs (refer to chapter 4). For this purpose, the design procedure
of chapter 3 is adapted to obtain a reliable analytical design:
• The number of slots being almost equal to the number of poles, the
teeth of motors with concentrated windings are much larger than those
with q = 1. As they saturate less easily, no flux leakage factor is used
for the calculation of the flux density in the teeth.
• The factor kleak , which is used to take the magnetic leakage into account
for the calculation of the airgap flux density, is slightly changed. The
FEM simulations give results that are between 4 and 8 % higher than
the analytically calculated results with the modified kleak .
• The slot fill factor fs is set to 0.6 (0.45 for q = 1).
• The factor kcoil is set to 0.93, as the end-windings are much shorter
than for designs with distributed windings. The winding factor now
depends on the number of poles and slots (refer to tables 5.4 and 5.5).
• The limit on the magnet weight is reduced to 5 kg.
As shown in the previous chapter, there are many suitable pole-slot number
combinations that give high winding factors. The tested combinations are
those with many symmetries in the winding layout. They can be found in
tables 5.4 and 5.5 and have the following number of slots per pole per phase
q and winding factor kw1 :
• q = 2/5 and 2/7, kw1 = 0.933
81
6 Results and comparison of designs with concentrated windings
• q = 3/8 and 3/10, kw1 = 0.9452
• q = 5/14 and 5/16, kw1 = 0.9514
To begin with, the obtained results of inner-rotor SMPM, outer-rotor SMPM

and tangentially-magnetized PM motor designs are analyzed separately. At
first, the motor weight for different winding layouts with both concentrated
and distributed windings is calculated as a function of the pole number.
Then, the active weight, the permanent magnet weight as well as the torque
and torque ripple of four 70-pole motors are calculated and simulated, among
which one has distributed windings. Finally, two geometries of 70-pole mo-
tors, one with q = 3/10 and one with distributed windings (q = 1), are
illustrated in scale 1:1.
6.1 SMPM motors
Figure 6.1 shows the active weight as a function of the pole number for
different values of q. The tendency for SMPM motors with concentrated
windings is that the active weight decreases with an increasing pole number.
Some points do not follow this tendency due to the fact that the active weight
also varies with the number of slots. Thus, the designs with q = 2/5 or 2/7
are slightly heavier than the others, because of the lower winding factor.
Figure 6.1 reveals also that the motors with concentrated windings are about
15 kg lighter than those with distributed windings (q = 1). Since there are
fewer teeth in designs with concentrated windings, both the slots and teeth
can be wider, as figure 6.3 shows. The permanent magnets are allowed to
be thicker and the airgap flux density to be higher without saturating the
teeth. This allows the motors with concentrated windings to be shorter and
therefore lighter. Furthermore, the motors with q = 1 were calculated with a
higher permanent magnet weight of 5.5 kg, which reduced their total weight
further (see figure 4.3).
Table 6.1 presents the results of the load simulations for four motors, of which
three have concentrated windings and one has distributed windings. All these
designs have equal copper losses (700 W). As can be seen, the torque ripple
is reduced from 9.3 % to under 3 % for concentrated windings. The torque
curves are displayed in figure 6.1.
82
6.1 SMPM motors
110
q=2/5, 2/7
105 q=3/8, 3/10
q=5/14, 5/16
100 q=1
Active weight [kg]

95
90
85
80
75
70
40 45 50 55 60 65 70
Pole number
Figure 6.1: Active motor weight as a function of the pole number for SMPM
motor designs with concentrated and distributed windings.
Table 6.1: Comparison of SMPM designs (FEM results).

Design Active PM Torque Torque
q, p, Qs weight [kg] weight [kg] [Nm] ripple [%]
q = 3/10, p = 70, Qs =63 76.1 5 833.6 2.4
q = 2/5, p = 70, Qs =60 78.6 5 842.6 2.4
q = 5/14, p = 70, Qs =75 73.0 5 825.2 2.6
q = 1, p = 70, Qs =210 91.4 5.5 833.0 9.3
1.05 q=3/8
q=2/5
1.04
q=5/14
1.03 q=1
1.02
Torque [p.u.]
1.01
0.99
0.98
0.97
0.96
0.002 0.004 0.006 0.008 0.01 0.012

Time [s]
Figure 6.2: Torque of SMPM motors with concentrated and distributed

windings at load conditions.
83
q=1
Active length:
q=1 L=187 mm
10 mm
10 mm radius
250 mm
q=3/10
L=124 mm
Figure 6.3: Geometries of two 70-pole SMPM motors, one with concen-
trated windings (q = 3/10) and one with distributed windings
(q = 1) scaled 1:1.
84
6.2 Outer-rotor SMPM motors
With distributed windings, the outer-rotor SMPM motor designs are

lighter than the inner-rotor SMPM motors, as shown in chapter 4. This
is also the case with concentrated windings. With concentrated windings,
the outer-rotor SMPM motors are approximately 20 kg lighter than with
distributed windings, (compare figure 6.1 to figure 6.4). Two geometries are
compared in figure 6.6. The drawback of the outer-rotor SMPM motors with
q = 1 is their high torque ripple (table 6.2). With concentrated windings, it
is successfully reduced from 26 % to under 5 %, as depicted in figure 6.5.
105
q=2/5, 2/7
100
q=3/8, 3/10
95 q=5/14, 5/16
q=1
Active weight [kg]
90
85
80
75
70
65
60
40 45 50 55 60 65 70
Pole number
Figure 6.4: Active motor weight as a function of the pole number for outer-
rotor SMPM motors with concentrated windings.
85
Table 6.2: Comparison of outer-rotor SMPM motor designs (FEM results).

Design Active PM Torque Torque
q, p, Qs weight [kg] weight [kg] [Nm] ripple [%]
q = 3/10, p = 70, Qs =63 61.9 5 887.0 4.6
q = 2/5, p = 70, Qs =60 63.9 5 899.9 4.2
q = 5/14, p = 70, Qs =75 60.5 5 866.2 4.4
q = 1, p = 70, Qs =210 83.0 5.5 839.6 25.7
q=3/8
1.1
q=2/5
q=5/14
1.05 q=1
Torque [p.u.]
0.95
0.9
0.85
0.002 0.004 0.006 0.008 0.01 0.012
Time [s]
Figure 6.5: Torque of outer-rotor SMPM motors with concentrated and dis-
tributed windings at load conditions.
86
Active length:
L=201 mm
10 mm
10 mm q=1
radius
250 mm
q=3/8
q=3/10
L=113 mm
Figure 6.6: Geometries of two 70-pole outer-rotor SMPM motors, one with
concentrated winding (q = 3/10) and one with distributed
windings (q = 1) scaled 1:1.
87
6.3 Tangentially-magnetized PM motors
As for the two previous configurations, a figure is used to present the active
weight of different motor designs with concentrated and distributed windings
(figure 6.7). However, for the tangentially-magnetized PM motor designs,
the difference is that the found optimized designs with concentrated wind-
ings have lower permanent magnet weights than the allowed 5 kg. Their
permanent magnet weight is only between 3.5 and 4.5 kg. This is due to the
constraint on the amplitude of the fundamental airgap flux density that is
not allowed to exceed 1.1 T. Due to the flux concentration, the airgap flux
density reaches easily 1.1 T with less than 5 kg permanent magnets. Fur-
thermore, the lower bound of the constraint on the machine length, which
is 100 mm, is reached for some designs. The external diameter is then re-
duced and becomes less than 500 mm. This can be observed in table 6.3 that
presents different features of 70-poles tangentially-magnetized PM motors.
Also in this case, the torque ripple is considerably reduced with concentrated
windings, as shown in table 6.3 and figure 6.8.
90
q=2/5, 2/7
85 q=3/8, 3/10
q=5/14, 5/16
80 q=1
Active weight [kg]
75
70
65
60
55
40 45 50 55 60 65 70
Pole number
Figure 6.7: Active motor weight as a function of the pole number for
tangentially-magnetized PM motors with concentrated wind-
ings.
88
6.3 Tangentially-magnetized PM motors
Table 6.3: Comparison of 70-pole tangentially-magnetized PM motor de-

signs (FEM results).
Active PM Torque Active Outer
Torque
q weight weight ripple length diameter
[N.m]
[kg] [kg] [%] [mm] [mm]
q = 3/10 55.9 3.4 876.3 4.2 101 500
q = 2/5 57.9 3.3 881.8 4.8 100 498
q = 5/14 56.2 4.1 868.6 4.2 119 500
q =1 76.4 5.5 863.1 41.7 163 500
1.25
1.2 q=3/8
q=2/5
1.15 q=5/14
1.1 q=1
Torque [p.u.]
1.05
0.95
0.9
0.85
0.8
0.75
0 0.002 0.004 0.006 0.008 0.01 0.012
Time [s]
Figure 6.8: Torque with concentrated and distributed windings of

tangentially-magnetized PM motors at load condition.
89
Active length:
L=163 mm
q=1
10 mm
radius
250 mm
q=3/10
q=3/8
L=101 mm
Figure 6.9: Geometries of two 70-pole tangentially-magnetized PM motors,

one with concentrated windings (q = 3/10) and one with dis-
tributed windings (q = 1) scaled 1:1.
90
6.4 Choice of a design
6.4 Choice of a design
A closer look is given on the 70-pole, 63-slot tangentially-magnetized PM

motor. This design is chosen because the tangentially-magnetized PM mo-
tors are lighter than the other configurations. The weight of the permanent
magnets is as well lower and their torque ripple, which is around 4 % of
the nominal torque, is satisfactory. Furthermore, the choice of q = 3/10 is
based on the number of symmetries in the winding layout that is 7 when
p = 70, q = 3/10 against 5 for the two other considered windings (q = 2/5
and q = 5/14).
The design fulfills the requirements with a good margin: the active
weight of 56 kg is far under 150 kg. With a 700 W value of copper losses,
490 W of other losses are allowed in order to keep the efficiency higher than
the 73 % of the induction motor and its gearbox. The iron losses reach
only 85 W according to the FEM calculations (appendix B), 78 W from the
stator and 7 W from the laminated rotor . Even the price of the motor looks
reasonable. If one kilogram of permanent magnet material costs 1000 SEK,
then the permanent magnets will cost 3400 SEK that is less than what the
gearbox costs.
6.5 Conclusions
Motors with distributed windings and q = 1 are unsatisfactory because of

the high torque ripple. This drawback is reduced by using concentrated
windings. For all motor configurations, the torque ripple is decreased under
5 %. The inner-rotor SMPM motor designs have the smallest torque ripple.
Furthermore, concentrated windings allow a higher airgap flux density

without the teeth getting saturated, as the tooth width is larger compared
to other motor designs. Thus, motor designs with concentrated windings are
lighter and contain less permanent magnet material.
91
The lightest motors are the tangentially-magnetized PM motors. They

allow designs with much lower permanent magnet weight than other motor
configurations without increasing the total active weight.
However, it should not be forgotten that concentrated windings in-

duce losses in the rotor. These losses have not been studied yet and
further work is needed. A thermal study should be made to show that
the permanent magnets are not subject to demagnetization due to high
temperature induced by eddy currents.
92
7 Design and results of axial-flux PM
motors
This chapter deals with the low-speed AFPM motors. The features used
for the design are first presented. The design procedure is then described
and followed by the results that are obtained analytically and from 3D-FEM
simulations.
7.1 Main features of the Torus AFPM motor
In this section, the geometry of the Torus AFPM motor is described by

defining its geometrical parameters. Then the flux densities are calculated.
Finally the windings are described, including analytical calculations of the
inductances, resistances and number of conductors. The equations given in
this section are based on [22] and [51].
7.1.1 Dimensions of the motor
The geometry of the axial-flux PM motor is described in figure 7.1. It is a

Torus AFPM machine without slots, as seen in section 2.2. The following
dimensions are used during the design procedure:
Ratio of inner to outer stator radius:

Ri
λ= (7.1)
Ro
Axial length of the stator:
Ls = 2lw + lsc (7.2)
93
7 Design and results of axial-flux PM motors
y Ro
Ri
z x
Ro : outer radius of the toroidal core
Ri : inner radius of the toroidal core
lsc : axial length of the stator core
2α
l rc lrc : axial length of the rotor core
l sc
lw lw : axial thickness of the winding
Wewo: outer radial thickness of the end-winding
Wewi Wewi : inner radial thickness of the end-winding
x
lm : axial length of the PM
y z δ: airgap length
p: number of poles
Wewo
lm δ
Figure 7.1: Definition of the geometrical parameters for the Torus AFPM
machines.
Axial length of a rotor disc:
Lr = lm + lrc (7.3)
Axial length of the machine:
L = Ls + 2Lr + 2δ (7.4)
Average diameter of the airgap:
Dg = Ro + Ri (7.5)
Average pole pitch:

τ = πDg /p (7.6)
Effective length of the stator core:
Ww = Ro − Ri (7.7)
94
Average length of the stator end connection:

1
Wew = (Wewo + Wewi ) (7.8)
2
7.1.2 Flux densities
Flux density in the airgap
The particularity of the Torus AFPM motor without teeth is its large effective
airgap, which includes the airgap δ and the axial thickness of the winding lw .
Owing to this large effective airgap, the airgap flux density created by the
permanent magnets usually does not exceed 0.65 T [22].
The maximum of the airgap flux density Bm is calculated as:
Br lm
Bm = (7.9)
lm + δ + lw
The amplitude of the fundamental airgap flux density Bδ is set equal to the
maximum airgap flux density Bm :
Bδ = Bm (7.10)
This assumption is based on 3D-FEM simulations. The obtained results are

gathered in figure 7.3 and table 7.1. Two components (By and Bz ) of the flux
density in the middle of the airgap length were investigated for three different
geometries with various permanent magnet widths. Bz is the axial compo-
nent of the airgap flux density whereas By is the ortho-radial component
(the directions of the x-, y- and z-axis are defined in figure 7.2). Appendix
B explains how the 3D-FEM simulations were conducted.
PM
x x-axis: radial
PM
y y-axis: ortho-radial
z z-axis: axial
Figure 7.2: Definition of the x-, y- and z-axis.
95
0.5
Bz
Airgap flux density (y and z components) [T]

0.4 By
0.3
0.2
0.1
0
-0.1
-0.2
-0.3
-0.4
-0.5
0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09

Path in the middle of the airgap [mm]
Figure 7.3: 3D-FEM simulated airgap flux densities, By , Bz components.
As shown in figure 7.3, the axial component of the airgap flux density Bz ,
which interacts with the current and generates the torque, is very smooth,
quite close to sinusoidal. This justifies the rough assumption made in equa-
tion (7.10).
The ortho-radial component of the airgap flux density By is the flux leaking
between two adjacent permanent magnets. Table 7.1 shows that the three
investigated designs have almost the same axial component of the 3D-FEM
simulated fundamental airgap flux density Bδz,F EM , whereas the permanent
magnet width is very different. Indeed, when the PM width lm is high, a
considerable part of the flux density from the permanent magnets is wasted
in the leakage flux between adjacent permament magnets contained in By .
Therefore, the permanent magnet thickness should not be too high in order
to utilize the permanent magnets as efficiently as possible. The upper limit
was chosen equal to 10 mm. As can be seen in table 7.1, the FEM simu-
lated value of the fundamental airgap flux density Bδz1,F EM is closed to the
analytically calculated value Bδ,an when lm = 10 mm.
Table 7.1: Airgap flux densities of different 30-pole AFPM machine designs.
lm δ lw Bδ,an Bδz1,F EM Bδy1,F EM
[mm] [mm] [mm] [T] [T] [T]
20 1 10.7 0.680 0.533 0.202
13 1 9.1 0.608 0.538 0.181
10 1 8.9 0.544 0.526 0.120
96
Flux density in the stator and rotor cores
The flux densities in the stator and rotor cores are calculated with equations
(7.11) and (7.12) respectively. As for the radial-flux PM machines, these flux
densities are subject to some constraints to avoid saturation.
πBδ Do (1 + λ)
Bsc = (7.11)
2p lsc
πBδ Do (1 + λ)
Brc = (7.12)
4p lrc
7.1.3 Windings
The windings are wound around the torus. As depicted in figure 7.1, the
thickness of the winding and end-windings is defined with three parameters:
lw for the ”working” part of the winding, Wewo and Wewi for the end-windings
which are not contributing in the generation of the torque. In order to
simplify the design, these three parameters are set equal to each other:
lw = Wewo = Wewi (7.13)
This assumption is possible because the outer and inner machine diameters
are large.
The values of the reactances, resistance and number of turns per phase
of the windings are needed for the design and thus their determination is
described below.
Winding reactances
The motor is not salient, therefore the d- and q-axis reactances are equal.
Xmd is the magnetizing reactance:
(Ns kw1 )2 τ Ww
Xmd = 4m1 μ0 f . (7.14)
πp δeq
97
where m1 is the number of phases, Ns the number of turns per phase and δeq
the equivalent aigap length given as:
δeq = 2δ + 2lm /μr (7.15)
Xl is the leakage reactance:

N 2 Ww Wew
Xl = 4πμ0 f s λle + λld (7.16)
pq Ww
where λle is the end-connection leakage permeance (q is the equivalent to the
number of slots per pole per phase):
λle = 0.3 q (7.17)
and λld the differential leakage permeance:
ml qτ kw12
λld = τdl (7.18)
π 2 δ kC ksat
τdl is the differential leakage factor:
! "
π 2 10q 2 + 2 30 ◦
τdl = sin −1 (7.19)
27 q
The d- axis synchronous reactance is then given by:
Xd = Xl + Xmd (7.20)
Winding resistance
The stator winding resistance per phase is given by:

Ns l1av ρcu
R= (7.21)
asa
where l1av is the average length of a stator turn (7.22), a is the number of
parallel paths or conductors and sa is the cross-section of a conductor (7.23).
l1av = 2Ww + Wewo + Wewi (7.22)
πd2w
sa = (7.23)
4
The wire diameter dw , is given by:
4 lw Di fs
dw = (7.24)
a m1 Ns
98
7.2 Design
Number of turns per phase
The number of turns per phase Ns of the Torus AFPM motor is calculated
with the same method as the SMPM motor (section 3.3.7, equation (3.50)).
The RMS value of the EMF is calculated as:
π
E = 2πf Ns kw1 Bδ Do2 1 − λ2 (7.25)
8p
7.2 Design
7.2.1 The design procedure
The design of the Torus AFPM machines follows the same procedure as the
radial-flux PM motors designs. The chart of figure 7.4 reminds of this design
procedure.
The design starts by choosing a set of values for the eight variables to
be optimized. These variables are p, Ro , Ri , lm , α, lrc , lsc , δ and lw . They
describe the entire motor geometry and allow to calculate the different
features of the motor.
The procedure then continues with the calculation of the flux densities in the
airgap and the rotor and stator cores. The ampere-turns are then computed
from the torque equation:
4T
Ns I = (7.26)
m1 kw1 Bδ Do2 (1 − λ2 )
Afterwards, the number of turns per phase can be deduced and the copper
losses can be calculated. Finally, the objective function which is the active
weight, is expressed as a function of the design variables. After some itera-
tions of the optimization procedure, a minimum for the objective function is
found.
99
Subject to
constraints on:
Vary the parameters: Outer

dimensions
Ro , Ri , lm , α
δ, lw , lrc , lsc Rigidity of
the structure
section 7.1.1
Express stator parameters
as a function of the variables
section 7.1.2
Maximum flux density
in the airgap
section 7.1.2
Flux density in the Magnetic

stator and rotor cores saturation
equation (7.26)
Ampere turns
Number of turns section 7.1.3

per phase
Copper losses Efficiency
Active weight
NO
Optimal
Result?
YES
Verification with
FEM
Figure 7.4: Followed procedure to optimize designs of low-speed Torus

AFPM motors.
100
7.2 Design
7.2.2 The design constraints
Geometry
The outer dimensions and the maximum motor weight are subject to the
same limits as the RFPM machines.
The rigidity of the structure is guaranteed by setting lower limits on the
geometrical parameters. Furthermore, the permanent magnet width should
not exceed 10 mm to limit the leakage between two adjacent permanent
magnets as seen in section 7.1.2. The effective airgap length is also limited
by having an upper boundary of 10 mm on the winding thickness lw .
The choice of the upper boundary for lw is confirmed by looking at the
magnetization curve of the PM material and the load lines (figure 7.5).
Indeed, the PM material is best utilized when operating at the point where
its energy product is maximum, thereby achieving minimum volume and
cost of the PM material.
1.1
load lines
1 l m =10 mm, lw =8.9 mm
0.9 l m =13 mm, lw =9.1 mm
0.8 l m =20 mm, lw =10.7 mm
Flux density [T]
0.7
0.6
0.5
0.4
0.3
(BH)max
0.2 operating point
at (BH)max
0.1
0
-700 000 -500 000 -300 000 -100 000 0
Magnetic field [A/m]
Figure 7.5: Magnetization curve of the PM material, its maximum energy

product and the load lines of three different designs.
101
The load lines of three 30-pole Torus AFPM designs with different lm and
lw (as given in table 7.1) are calculated from 3D-FEM simulations. The
flux density and magnetic field are varying depending on the point in the
permanent magnet where they are calculated. Therefore different points
were checked, which is illustrated in figure 7.5 by the arrows across the
load lines. The figure also shows the maximum energy product (BH)max .
For NdFeB permanent magnets in Torus AFPM machines, it corresponds
to the case where lw lm . Therefore, the load line corresponding to the
lowest permanent magnet width comes closest to the point where the energy
product of the permanent magnets is maximum.
Magnetic saturation
Magnetic saturation in the rotor and stator cores is avoided by limiting the
flux densities Bsc and Brc to a maximum of 1.6 T and 1.4 T respectively. The
winding temperature is assumed to be 80 ◦ C.
Efficiency and cost
The copper losses are limited to 700 W to start with. The maximum perma-
nent magnet weight is not subject to any constraints.
7.3 Results
7.3.1 Torus AFPM machine
At first, the design procedure for the Torus AFPM machine was run applying
the constraints presented above. No solutions that fulfill these constraints
were found. The number of poles was varied from p = 10 to 70. It was
noticed that the optimization algorithm was not as efficient as for design
of RFPM machines: The optimized solutions were depending a lot on the
initial values, probably due to many local minima in the objective function.
In order to find some solutions, the constraint on the outer diameter was
102
7.3 Results
adjusted. Optimized designs are then found when the maximum allowed
outer diameter is increased to 550 mm. The permanent magnet weight is
around 122 kg, which is very high compared to the radial-flux PM machines.
Another concession can be made by increasing the allowed copper losses
while the maximum outer diameter is set back to 500 mm. Solutions fulfilling
the requirements on the constraints are found for maximum copper losses
of 1030 W. These designs have a significantly decreased efficiency. The
permanent magnet weight is around 7 kg.
It can be concluded from these design attempts, that the torus AFPM
machine without slots, with two rotors and one stator is not appropriate for
the application. A way to achieve the nominal torque without exceeding the
allowed outer diameter may be to have a multi-disk Torus AFPM machine.
An investigation on such machines is made in the next section.
7.3.2 Multi-disk Torus AFPM machine
The multi-disk Torus AFPM machine is composed of at least two Torus

AFPM machines as those described previously. The studied example consists
thus of two stators and four rotor disks as shown in figure 7.6.
permanent magnet
rotor iron
stator iron
windings
Figure 7.6: Multi-disk torus AFPM machine.
103
The multi-disk Torus AFPM machine is designed using the same procedure
as for the Torus AFPM machine (refer to section 7.2.1): One of these two
Torus AFPM machines is optimized for half the nominal torque. The total
active weight of the machine is twice the active weight of the optimized
Torus AFPM machine.
Figure 7.7 shows the active weight of different multi-disk AFPM motors as
a function of the pole number. All the calculated designs have an outer
diameter equal to 500 mm, 700 W copper losses and permanent magnets
weight is 6 kg. As can be seen, the multi-disk AFPM machines are very
light, under 40 kg for most of them.
44
42
40
38
Active weight [kg]
36
34
32
30
28
26
24
20 25 30 35 40 45 50 55 60 65 70
Pole number
Figure 7.7: Active weight as a function of the pole number of multi-disk

Torus AFPM machines.
For RFPM machines with constant copper losses, a lower permanent

magnet weight, i.e. a lower airgap flux density from the permanent magnets,
is compensated by a higher current loading. Therefore the conductors
have a bigger diameter which causes the slots to be larger. For the AFPM
machines, another factor has to be taken into account if the permanent
magnet weight should be decreased. Indeed, a higher current loading is
obtained with a higher number of turns or a larger conductor diameter.
This implies an increase of the effective airgap, which leads to an even
lower airgap flux density. The nominal torque would then not be achieved.
Therefore, it was impossible to decrease the value of the permanent magnet
weight for constant copper losses.
104
7.3 Results
Figure 7.8 and 7.9 show the geometry of a calculated 36-pole multi-disk
Torus AFPM motor design. The permanent magnet weight is 6 kg, the
active weight is 30 kg and the axial length is 123 mm.
250
240
Radial direction [mm]
230
220
210
200
190
60 50 40 30 20 10 0 10 20 30 40 50 60
Axial direction [mm]
Figure 7.8: Part of the geometry of a 36-pole multi-disk Torus AFPM mo-
tor.
250
Radial direction [mm]
240
230
220
210
0 10 20 30 40 50 60 70 80
Tangential direction [mm]
Figure 7.9: Two permanent magnets of a 36-pole multi-disk Torus AFPM

motor.
105
7.3.3 Comparison with the investigated RFPM motors
The calculated multi-disk Torus AFPM motor designs are much lighter than
the RFPM motors. The difference in the active weight is approximately
20 kg. Another advantage of the multi-disk AFPM machine is that it has no
cogging torque due to the absence of teeth (but no attempt has been made
to simulate the torque ripple).
However, some RFPM machines with concentrated windings have much

lower permanent magnet weights than the multi-disk Torus AFPM machines:
the difference in PM material between radial- and axial-flux machines is
up to 2.5 kg. Table 7.2 recapitulates the obtained designs for the different
investigated machine configurations. In this table however, it does not
appear that the main drawback of the multi-disk Torus AFPM machine lies
in its construction. The assembly of the four disks and two stators is tricky
due to the attraction forces between the rotor discs [52].
Table 7.2: Comparison of the different investigated PM machines: Some

calculated and FEM-simulated results.
(CW stands for concentrated windings and DW for distributed windings)
Motor Winding Pole/Slot Active PM Torque
configuration type numbers weight [kg] weight [kg] ripple [%]
Inner-rotor q=1 70/210 91.4 5.5 9.3
SMPM CW 70/63 76.1 5 2.4
Outer-rotor q=1 70/210 83 5.5 25.7
SMPM CW 70/63 61.9 5 4.6
Inset PM q=1 70/210 90.7 5.5 22
V-shaped PM q=2 30/180 141.5 5.5 35
Tangentially- q=1 70/210 76.4 5.5 41.7
magnetized PM CW 70/63 55.9 3.4 4.2
Torus AFPM DW no solution
Multi-disk AFPM DW 36/108∗ 29.4 6 ∗∗
* number of coils (corresponds to q = 1).

** not calculated because it requires a transient 3D-FEM simulation.
106
7.4 Summary
7.4 Summary
The Torus AFPM machines were investigated for the low-speed direct-drive
application. At first, no designs that satisfy all the constraints were found.
The main reason is the restriction on the outer diameter.
The solution that was investigated to achieve the nominal torque with
a limited outer diameter was the multi-disk Torus AFPM machine. The
calculated designs have very attractive active weight compared to the RFPM
machine designs, but in contrast the permanent magnet weight is higher and
difficult to minimize. Furthermore, the assembly of the multi-disk Torus
AFPM machine is very delicate.
Another solution would be to consider a torus AFPM motor with teeth in

the stator. The advantage is a higher achievable airgap flux density, since
the effective airgap is much smaller with the teeth. However, this solution is
not investigated in this thesis.
107
108
8 Conclusions and future work
8.1 Conclusions
Different motor configurations have been investigated in an attempt to

replace an induction motor and its gearbox with a competing direct-driven
PM machine. A design procedure based on an optimization program has
been developed, taking the particularities of each configuration into account.
At first, designs with distributed windings and one slot per pole per
phase are calculated and simulated with FEM. The designs of the inner- and
outer-rotor SMPM motors and of the tangentially-magnetized PM motors
are the most promising, since they are satisfying the requirements on the
size, weight and efficiency. However, the torque ripple of these machines
is very high. The tangentially-magnetized PM motors are found to be the
lightest, while the V-shaped PM motor designs could not compete with the
other motor configurations.
In order to improve these designs, concentrated windings for high pole

number PM machines have been investigated. Winding layouts offering high
winding factors and limiting the rotor losses as well as the noise and vibra-
tions of the machine are identified and sorted out for further investigations
with FEM simulations. The designs with concentrated windings are much
lighter than the designs with distributed windings. The torque ripple is also
much lower, under 5 %. One of the best designs is a tangentially-magnetized
PM motor, with 70 poles, 63 slots and concentrated winding. This motor
has an active weight under 60 kg with only 3.4 kg of permanent magnet
material.
109
8 Conclusions and future work
At last, the Torus AFPM machine configuration was tested for the appli-
cation. No designs fulfilling the requirements have been found. Therefore,
multi-disk Torus AFPM machine designs were investigated. The design
procedure results in very light machines regarding the active weight, around
30 kg. However, the permanent magnet weight is slightly higher than for the
calculated RFPM machines. Additionally, the construction of this kind of
machine is problematic.
Several designs that fulfill the requirements have been both calculated
and simulated with FEM. Some of these designs are very promising, since
the weight of the permanent magnets is reasonable for a sufficient efficiency.
Furthermore, the active motor weight is far below the limit, leaving a
comfortable margin for the weight of the mechanical parts.
8.2 Suggestions for future work
Different theoretical designs have been found to be very promising for

the application. Thus, the future work could concentrate further on
one particular design for the purpose of building a prototype. The iron
losses, rotor losses and eddy-currents in the permanent magnets should be
investigated further. The constraints applied during the design procedure
should be improved, notably the constraints on the rigidity of the structure.
They have to be adapted to the application. If a prototype is built, a lot of
work would be required for designing the mechanical elements, testing the
prototype and verifying if the machine fulfills the expectations.
Besides, the AFPM machines require more work. The optimization

algorithm should be improved, as it was very dependent on the initial con-
ditions. Other AFPM machine configurations could fulfill the requirements.
A first one is the Torus AFPM machines with a slotted stator. A second
one is an AFPM machine with concentrated windings and stator cores
made of compact iron powder composites. These configurations could be an
alternative to the investigated multi-disk Torus AFPM machine.
110
Bibliography
[1] Grauers A., Design of direct-driven permanent-magnet generators for

wind turbines, Doctoral thesis, Chalmers University of Technology,
Göteborg, Sweden, 1996.
[2] Carlson O., Grauers A., Williamson A., Engström S., Spooner E., Design
and test of a 40 kw directly driven permanent-magnet generator with a
frequency converter, European Wind Energy Conference and Exhibition,
EWEC’99, Nice, France, 1999.
[3] Chalmers B., Wu W., Spooner E., An axial-flux permanent-magnet gen-

erator for a gearless wind energy system, IEEE Transactions on Energy
Conversion, Vol. 14, No. 2, pp. 251-257, 1999.
[4] Hartkopf T., Hofmann M., Jöckel S., Direct-drive generators for
megawatt wind turbine, European Union Wind Energy Conference,
EWEC’97, Dublin, 1997.
[5] Jöckel S., Gearless wind energy converters with permanent magnet gen-
erators. an option for the future?, European Union Wind Energy Con-
ference, EWEC’96, Göteborg, Sweden, 1996.
[6] Chen J., Nayar C., Xu L., Design and finite-element analysis of an outer-
rotor permanent-magnet generator for directly coupled wind turbines,
IEEE Transactions on Magnetics, Vol. 36, No. 5, pp. 3802-3809, 2000.
[7] Hakala H., Integration of motor and hoisting machine changes the ele-
vator business, ICEM 2000, Vol. 3, pp. 1242-1245, 2000.
[8] Caricchi F., Crescimbini F., Honorati O., Modular axial-flux permanent-
magnet motor for ship propulsion drives, IEEE Transactions on Energy
Conversion, Vol. 14, No. 3, pp. 673-679, 1999.
111
Bibliography
[9] Parker D., Hodge C., The electric warship, Power Engineering Journal,
Vol. 12, Issue 1, pp. 5-13, 1998.
[10] Ficheux R., Caricchi F., Crescimbini F., Onorato H., Axial-flux
permanent-magnet motor for direct-drive elevator systems without ma-
chine room, IEEE Transactions on Industry Applications, Vol. 37, No.
6, pp 1693-1701, 2001.
[11] ITT Flygt, Product catalogue, slow speed mixer, www.flygt.com.
[12] Sitapati K., Krishnan R., Performance comparisons of radial and ax-
ial field, permanent-magnet, brushless machines, IEEE Transactions on
Industry Applications, Vol. 37, No. 5, 2001.
[13] Haring T., Forsman K., Huhtanen T., Zawadzki M., Direct drive - open-
ing a new era in many applications, Pulp and Paper Industry Technical
Conference, pp. 171-179, 2003.
[14] Rosu M., Nahkuri V., Arkkio A., Jokinen T., Mantere J., Westerlund J.,
Permanent magnet synchronous motor for ship propulsion drive, Sympo-
sium on Power Electronics Electrical Drives Advanced Machines Power
Quality (SPEEDAM’98), Sorrento, Italy, pp. C3-7:C3-12, 1998.
[15] Lampola P., Perho J., Electromagnetic analysis of a low-speed

permanent-magnet wind generator, Opportunities and Advances in In-
ternational Power Generation, IEE, pp. 55-58, 1996.
[16] Spooner E., Williamson A., Direct coupled, permanent magnet genera-
tors for wind turbine applications, IEE Electric Power Applications, Vol.
143, No. 1, pp. 1-8, 1996.
[17] Chen J., Nayar C., A direct-coupled, wind-driven permanent magnet

generator, Energy Management and Power Delivery, EMPD’98 ,Vol. 2,
pp. 542-547, 1998.
[18] Wu W., Ramsden V., Crawford T., Hill G., A low-speed, high-torque,
direct-drive permanent magnet generator for wind turbines, IEEE In-
dustry Applications Conference, Vol. 1, pp. 147-154, 2000.
[19] Heikkilä T., Permanent magnet synchronous motor for industrial in-
verter applications - analysis and design, Doctoral thesis, Lappeenranta
University of Technology, 2002.
112
Bibliography
[20] Caricchi F., Crescimbini F., Axial-flux machines: design and applica-
tions, Overheads of the Seminar at KTH, 2000.
[21] Cirani M., Analysis of an innovative design of an axial flux torus-

machine, Licentiate Thesis, ISSN 1650-674X, 2002.
[22] Gieras J., Wing M., Permanent magnet motor technology, New York,
Marcel Dekker, ISBN 0-8247-0739-7, 2002.
[23] Muljadi E., Butterfield C., Wan Y., Axial flux, modular, permanent-
magnet generator with a toroidal winding for wind turbine applications,
IEEE Transactions on Industry Applications, Vol. 35, No. 4, pp. 831-836,
1999.
[24] Söderlund L., Eriksson J., Salonen J., H.Vihriälä, Perälä R., A
permanent-magnet generator for wind power applications, IEEE Trans-
actions on Magnetics, Vol. 32, No. 4, pp. 2389-2392, 1996.
[25] Crescimbini F., Carricchi F., Solero L., Chalmers B., Spooner E., Wu
W., Electrical equipment for a combined wind/pv isolated generating sys-
tem, Opportunities and Advances in International Power Generation
IEE, Conference publication No. 419, pp. 59-64, 1996.
[26] Mitcham A., Transverse flux motors for electric propulsion of ships, IEE
Colloquium on New Topologies for Permanent Magnet Machines, Digest
No. 1997/090, pp. 3/1-3/6, 1997.
[27] Henneberger G., Bork M., Development of a new transverse flux motor,
IEE Colloquium on New Topologies for Permanent Magnet Machines,
Digest No. 1997/090, pp. 1/1-1/6, 1997.
[28] Lefevre L., Soulard J., Nee H.P., Design procedure for line-start per-
manent magnet motors reversible modification of the no-load voltage of
a line-start permanent magnet synchronous motor, IEEE Nordic Work-
shop on Power and Industrial Electronics, NORpie 2000, Aalborg, Dan-
mark, 2000.
[29] Chen Z., Spooner E., A modular permanent-magnet generator for vari-
able speed wind turbine, Electrical Machines and Drives IEE, Conference
publication No. 412, pp. 453-457, 1995.
113
Bibliography
[30] Spooner E., Williamson A., Modular, permanent-magnet wind-turbine

generators, Industry Applications Conference, IAS’96., Vol. 1, pp. 497-
502, 1996.
[31] Williamson A., Spooner E., Thompson L., Large modular pm generators,
IEE Colloquium on New Topologies for Permanent Magnet Machines,
Digest No. 1997/090, pp. 4/1-4/6, 1997.
[32] Epskamp T., Hagenkort B., Hartkopf T., Jöckel S., No gearing, no con-
verter - assessing the idea of highly reliable permanent-magnet induction
generators, European Union Wind Energy Conference, EWEC’99, Nice,
France, 1999.
[33] Sadarangani C., Design and analysis of induction and permanent magnet
motors, Book, Royal Institute of Technology, Stockholm, ISBN 91-7170-
627-5, 2000.
[34] Zhu Z., Howe D., Xia Z., Prediction of open-circuit airgap field distri-
bution in brushless machines having an inset permanent magnet rotor
topology, IEEE Transactions on Magnetics, Vol. 30, No. 1, pp 98-107,
1994.
[35] Thelin P., Nee H.P., Analytical calculation of the airgap flux density
of pm synchronous motors with buried magnets including axial leakage,
tooth and yoke saturations, International Conference on Power Electron-
ics and Variable Speed Drives, London, 2000.
[36] Meier S., Theoretical design of surface-mounted permanent magnet mo-

tors with field-weakening capability, Master thesis, Royal Institute of
Technology, Departement of Electrical Machines and Power Electronics,
Stockholm, 2002.
[37] Magnussen F., C. Sadarangani T.
[38] Mohan N., Undeland T., Robbins W., Power electronics, converters,
applications, and design, Book, ISBN 0-471-58408-8, 1995.
[39] Matlab help on optimization toolbox.
[40] Say M., The performance and design of alternating current machines,
Book, Pitman, ISBN 999-184258-6, 1948.
114
Bibliography
[41] Jack A., Mecrow C., Dickinson P., Stephenson D., Burdess J., Fawcett
N., Evans J., Permanent-magnet machines with powdered iron cores and
prepressed windings, IEEE Transactions on Industry Applications, Vol.
36, Issue 4, pp. 1077-1084, 2000.
[42] Patent No. 92958, Mehrphasenmaschine mit ungleicher ankerspulen und

polzahl, Deutsches Reichspatent, 1895.
[43] Cros J., Viarouge P., Synthesis of high performance pm motors with
concentrated windings, IEEE Transactions on Energy Conversion, Vol.
17 Issue 2, pp. 248-253, 2002.
[44] Koch T., Binder A., Permanent magnet machines with fractional slot
winding for electric traction, International Conference on Electrical Ma-
chines, ICEM 2002, Brugge, Belgien, 2002.
[45] Magnussen F., Thelon P., Sadarangani C., Performance evaluation

of permanent magnet synchronous machines with concentrated and
distributed windings including the effect of field-weakening, IEE In-
ternational conference on Power Electronics and Electrical Machines
(PEMD), Edinburgh, United Kingdom, 2004.
[46] Ishak D., Zhu Z., Howe D., Comparative study of permanent magnet
brushless motors with all teeth and alternative teeth windings, IEE In-
ternational conference on Power Electronics and Electrical Machines
(PEMD), Edinburgh, United Kingdom, 2004.
[47] Jack A., Mecrow C., Evans S., Low cost smc brushless dc motors for high
volume applications in the automotive sector, IEE International confer-
ence on Power Electronics and Electrical Machines (PEMD), Edinburgh,
United Kingdom, 2004.
[48] Pellegrino G., F. Villata P.G., Vagati A., Design of direct-drive low-
speed pm machines, Industry Applications Conference, 38th IAS Annual
Meeting, Vol. 2, pp. 1421-1428, 2003.
[49] Cros J., Viarouge P., New structures of polyphase claw-pole machines,
IEEE Transactions on Indusrty Applications, Vol. 40, No. 1, pp. 113-120,
2004.
115
Bibliography
[50] Maliti K., Sadarangani C., Modelling magnetic noise in induction ma-
chines, International Conference on Electrical Machines, pp. 406-410,
Cambridge, U.K., 1997.
[51] Huang S., Luo J., Leonardi F., Lipo T., A comparison of power density
for axial flux machines based on general purpose sizing equations, IEEE
Transactions on Energy Conversion, Vol. 14, No. 2, pp. 185-192, 1999.
[52] Mbidi D., van der Westhuizen K., Wang R., Kamper M., Blom J., Me-
chanical design considerations of a double stage axial-flux permanent-
magnet machine, Industry Applications Conference, Conference record
of the 2000 IEEE, Vol. 1, pp. 198-201, 2000.
[53] Cedrat, www.cedrat.com.
[54] Surahammars Bruk, Non-oriented fully processed electrical steels.
[55] Lundgren J., Rönnqvist M., Värbrand P., Optimeringslära, Book, Stu-
dentlitteratur, ISBN 91-44-03104-1, 2003.
[56] Bianchi N., Bolognani S., Design optimisation of electrical motors by an

adaptive model based optimisation technique, 9th International Confer-
ence on Electrical Machines and Drives, Canterbury, 1999.
116
Glossary of symbols and acronyms
Symbols
a number of parallel paths or conductors in the windings -

Asl slot area [m2 ]
B̂δ amplitude of the fundamental airgap flux density [T ]
Bn radial component of the airgap flux density [T ]
Bm maximum value of the airgap flux density [T ]
Br remanence flux density of the permanent magnet [T ]
Brc flux density in rotor cores (AFPM machine) [T ]
Bry flux density in the rotor yoke [T ]
Bsat flux density in the saturated iron bridges of V-shaped PM motors [T ]
Bsc flux density in stator core (AFPM machine) [T ]
bso stator slot opening [m]
bss1 inner stator slot width [m]
bss2 outer stator slot width [m]
bt1 small tooth width (when the teeth are irregularly distributed) [m]
bt2 large tooth width (when the teeth are irregularly distributed) [m]
bts stator tooth width [m]
Bts flux density in a stator tooth [T ]
By airgap flux density, y- (orthoradial) direction (AFPM machine) [T ]
Bz airgap flux density, z- (axial) direction (AFPM machine) [T ]
d denominator of the fraction used for the winding layout design -
dw wire diameter [m]
D inner stator diameter [m]
Dg average diameter of the airgap (AFPM machine) [m]
Drc outer rotor diameter [m]
Di inner rotor diameter [m]
Do outer stator diameter [m]
E fundamental of the induced voltage (RMS value) [V ]
−
→
Ei EMF phasor of conductor i [V ]
f frequency [Hz]
fs slot fill factor -
117
Hm coercive magnetic field intensity of the PM [A/m]

hry rotor yoke height [m]
hss stator slot height [m]
hsw slot wedge height [m]
I line current [A]
Id d-axis current [A]
Iq q-axis current [A]
k integer -
kC Carter factor -
kcoil end-winding coefficient -
J current density [A/m2 ]
kcor correction factor for the calculation of the current loading -
kleak correction factor for the airgap flux density calculation -
kleakα correction factor for the airgap flux density calculation -
kleaktooth correction factor for the flux density in the teeth -
kopen ratio of the slot opening over the slot width -
kw1 fundamental winding factor -
L active length [m]
Ld d-axis inductance [H]
lF e length of the iron bridge (see figure 3.3) [m]
li airgap around the buried V-shaped permanent magnets [m]
liron width of a rotor iron piece [m]
of the tangentially-magnetized PM motor [m]
Ll leakage inductance [H]
lm magnet thickness [m]
Lmd d-axis magnetizing inductance [H]
Lmq q-axis magnetizing inductance [H]
Lq q-axis inductance [H]
Lr axial length of a rotor disc (AFPM machine) [m]
lrc axial length of a rotor core (AFPM machine) [m]
Ls axial length of the stator (AFPM machine) [m]
lsc axial length of the stator core (AFPM machine) [m]
lw winding axial thickness (AFPM machine) [m]
l1av average length of a stator turn [m]
ma amplitude modulation ratio -
m1 number of phases -
118
mtot total weight of the machine [kg]
n numerator of the fraction used for the winding layout design -
nl number of layers of the winding -
Ns number of turns per phase -
ns number of conductors per slot -
p number of poles -
q number of slots per pole per phase -
Qs number of stator slots -
R resistance of one phase of the stator winding [Ω]
Ri inner radius of the toroidal core (AFPM machine) [m]
m reluctance of the PM [At/Wb]
gap airgap reluctance [At/Wb]
Ro outer radius of the toroidal core (AFPM machine) [m]
sa conductor cross-section [m2 ]
S vector describing a sequence of conductors of phase A -
S1 current loading [A/m]
t time [s]
T torque [N m]
Tcu copper temperature [◦ C]
V external voltage [V ]
Vd rectified voltage across the filter capacitor [V ]
Wewi end-winding inner radial thickness (AFPM machine) [m]
Wewo end-winding outer radial thickness (AFPM machine) [m]
Ww effective length of the stator core (AFPM machine) [m]
wF e thickness of the iron bridge [m]
wm PM length [m]
Xmd armature reaction reactance [Ω]
Xd d-axis synchronous reactance [Ω]
Xl leakage reactance [Ω]
Greek symbols
α half pole angle in electrical degrees [◦ ]

αv ”V” angle between the permanent magnets [◦ ]
β angle between the d-axis and the current vector [rad]
δ airgap length [m]
δeq equivalent airgap [m]
119
γ current angle between the line current I

and the induced voltage E [rad]
λ ratio of inner to outer stator radius (AFPM machine) -
λ1 specific permeance coefficient of the slot opening -
λle end connection leakage permeance (AFPM machine) -
λld differential leakage permeance (AFPM machine) -
μ0 magnetic permeability of free space (4π.10−7 ) [H/m]
μr relative magnet permeability -
Φ flux [W b]
Ψm magnet flux linkage [W b]
ρcu copper resistivity [Ωm]
σ radial magnetic force density [N/m2 ]
τ average pole pitch (AFPM machine) [m]
τdl differential leakage factor (AFPM machine) -
τs stator slot pitch [m]
θ angular coordinate [rad]
ω electrical angular frequency [Hz]
Acronyms
AC alternative current
AFPM axial-flux permanent magnet
CW concentrated windings
DC direct current
DW distributed windings
EMF electro-motive force
FEM finite element method
LCM least common multiple
LSPM line start permanent magnet
MMF magneto-motive force
NdFeB neodymium iron boron
RFPM radial-flux permanent magnet
RMS root mean square value
SMPM surface-mounted permanent magnet
TFM transverse-flux machine
PM permanent magnet
p.u. per unit
2D two-dimensional
3D three-dimensional
120
A Direct-driven machines
manufacturers
Nowadays, low-speed direct-driven machines are both manufactured and sold. Some
of the manufacturers are presented here.
A.1 Wind turbines
Enercon, www.enercon.de, wind turbines with synchronous wound-rotor generators.
Bergey, www.bergey.com, small wind turbines (≤10 kW) with RFPM genera-
tors.
Westwind, www.westwind.com.au, small wind turbines (≤ 20kW) with outer-

rotor RFPM generators.
Jeumont Industry, www.jeumont-framatome.com, 750 kW wind turbines with

AFPM generators.
Zephyros, www.zephyros.com, 2 MW wind turbines with PM generators.
A.2 Other applications
Lynx Motion Technology, www.lynxmotiontechnology.com, modified AFPM motor

manufacturer.
Kone, www.kone.com, elevators with different kinds of AFPM generators.
121
A Direct-driven machines manufacturers
122
B Finite element methods simulations
All finite element simulations in this thesis were realized with the softwares
Flux2D and Flux3D from Cedrat [53].
B.1 2D-FEM simulations
Magneto-static simulations
Magneto-static simulations are used for the verification of the flux density
created by the permanent magnets in the airgap and the state of saturation in
the teeth as well as the stator and rotor yokes. Figure B.1 shows an example
of a meshed geometry used for the 2D simulations.
Simulations at no-load
No-load time-step simulations are used to verify the value of the induced
voltage. An external electrical circuit is coupled to the meshed geometry.
The coils representing the windings are connected to very high resistances.
Simulations at load
For time-step simulations at load conditions, the speed of the machine is

set constant to 50 rpm. As for the no-load simulations, an external circuit
is coupled to the meshed geometry. Thereby, current sources allow to de-
fine sinusoidal currents which are forced in the stator windings. With this
123
Figure B.1: Mesh of a 50-pole SMPM motor.
simulation, the torque and iron losses can be computed. The starting position
has to be defined carefully in order to match the imposed currents.
Iron losses
In order to be able to compute the iron losses, the iron loss coefficients
have to be calculated from the specific total loss data of the iron material,
supplied by the lamination manufacturer. It was not possible to obtain
this data for the iron material that is used in this thesis. Therefore, the
iron loss coefficients of another iron material were used. This material has
better magnetic properties, it saturates for higher values of the flux density.
According to [54], a material with a magnetization curve that is better than
another material, will also have higher iron losses. The iron losses that are
calculated in this thesis may thus be overestimated.
The iron losses calculations with FEM simulations include the hys-
teresis, copper and excess losses, equation (B.1). They also take the
124
non-sinusoidal shape of the flux density into account :

# ! 2 "
2 1 T d dB(t) 2 dB(t) 3/2
Piron = kh Bm kf f + σ + ke kf dt (B.1)
T 0 12 dt dt
where Bm is the maximum flux density in the concerned mesh element, f

is the frequency, σ is the conductivity of the iron material (4e6 (Ωm)−1 ), d
is the thickness of the lamination sheet (0.65 mm), kh is the coefficient of
hysteresis loss (279.81 W.s/T 2 /m3 ) and ke is the coefficient of excess loss
(1.28 W.(s/T )3/2 /m3 ).
3D-FEM simulations are needed for the designs of the AFPM motors. Only
static simulations are used since the time-step simulations demand a lot of
time for both their implementation and computation. An example of a ge-
ometry used for the computations is given in figure B.2. Because of the
symmetries, two poles are represented with only half the stator core and one
rotor disc.
rotor disk
PMs
half stator core
Figure B.2: Geometry of a 30-pole AFPM motor drawn in Flux3D.
125
126
C Optimization
Solving an optimization problem is to find the best possible values of the

problem variables that lead to an optimum of the objective function (i.e. a
minimum or a maximum value of the function). The objective function is the
function that is to be optimized; it depends on the problem variables. These
variables are subject to constraints and boundary conditions.
C.1 A simple example
The following example of an optimization problem illustrates the principles

on a simple two-variables optimization problem. Thereby, let x1 and x2
be the two variables. The objective function to maximize is x1 + x2 . The
constraints consist of four inequalities:
x1 + 3x2 ≤ 9
2x1 + x2 ≤ 8
x1 ≥ 0
x2 ≥ 0
In this simple case with two variables and both a linear objective function
and linear constraints, the solution can be found graphically (see figure C.1).
It can be found that x1 = 3 and x2 = 2 gives the maximum value of the
objective function which is 5.
C.2 Optimization program for the motor design
For the motor design, the optimization problem is much more complex:
127
C Optimization
x2
x 2 = -x1/3 +3
optimal
solution
2
vector
-(x1+x2 )
Area of possible x 2= -2x1 +8
1 solutions
0 x1
1 2 3
Figure C.1: Graphical solution of a two-variables optimization problem.
• There are 12 variables. Among these variables, two are integers, the
number of slots and the number of poles, which also should be even.
• The objective function, in this case the motor weight, is a non-linear

function of the variables.
• There are about 20 inequalities that describe the constraints. Some of

them are non-linear.
C.2.1 How to deal with the integer variables
Solving an optimization problem with both integer and non-integer variables

is tricky [55]. In order to avoid problems with integer variables, different
values for p and Qs are tested independently of the optimization problem,
using a loop. The value of p is varied in a range between 20 and 80 in a
loop exterior to the optimization program and is therefore a constant when
solving the optimization problem. Qs is chosen depending of the winding
type that is tested (for example if q = 1 then Qs is given by Qs = 3pq).
128
C.2 Optimization program for the motor design
C.2.2 How to deal with the non-linearities
Solving a non-linear optimization problem is also tricky. Different algorithms

exist in this context. A description of some algorithms and optimization
techniques including their advantages and drawbacks can be found in [56].
In Matlab, some algorithms are already implemented in the ”optimization
toolbox”. The algorithm applied in the motor design is based on sequential
quadratic programming. The method is described in detail in the Matlab
help [39]. An initial solution is needed in order to start the optimization
process. This can be seen as a drawback since the initial solution has to
be guessed. However, as the algorithm proved to be stable and the Matlab
toolbox very convenient, no other algorithms were investigated.
129
C Optimization
130
D Thermal study
The goal of the thermal study is to estimate the temperature in the

windings. During the design procedure, this temperature is used to calculate
the resistance of the windings.
A 62-pole SMPM motor with q = 1 is simulated in Flux2D. The sim-

ulation is static, magneto-thermal. Since the motor is placed under water in
the considered application, no forced cooling is considered. The temperature
of the water is assumed to be 25 ◦ C. The investigated motor geometry is
displayed in figure D.1. The housing is surrounded by water at 25 ◦ C. A
layer of air separates the stator iron and the housing. There is some air in
the airgap as well.
The heat is created in the windings by the copper losses. Hence, heat
sources are defined for each slot. They are constant and contribute equally
to the total power loss (i.e. 700/Qs W each). The heat is then transmitted by
conductivity from the copper into the stator iron. The transmission between
the stator iron and the surrounding air is done by natural convection and
conductivity. Most of the heat is then dissipated from the housing to the
water by convection.
The conductivities and convective exchange coefficients used in the simula-
tion are provided in tables D.1 and D.2 respectively.
The results of the simulation give a temperature in the winding of 81 ◦ C.
The winding temperature for the calculation of the winding resistance was
therefore set to 80 ◦ C. This value was compared with a measurement on a
mixer with an induction motor. With 2.55 kW of total losses the winding
temperature was measured to be 97 ◦ C, therefore the assumed temperature
of the windings seems to be reasonable.
131
D Thermal study
Boundary condition:
Air Dirichlet, 25 C
Air
Iron
P
M Iron H Water
o
I u
Air r s
o i
n P n
M g
Copper
Convective exchange coeff.

Convective exchange coeff.
Iron/Water
Air/Iron
Figure D.1: Definition of the thermal features for FEM simulation.
Table D.1: Thermal conductivity of the materials.

Thermal conductivity
Material
[W/mK]
Iron 43
Air 0.0287
Permanent magnet 9
Copper 385
Water 0.65
Table D.2: Convective exchange coefficient.

Convective exchange
Materials
coefficient [W/m2 K]
Air/Iron 10
Iron/Water 50
132

Design, Optimization and Comparison of Permanent Magnet Motors For A Low-Speed Direct-Driven Mixer

Uploaded by

Copyright:

Available Formats

Design, Optimization and Comparison of Permanent Magnet Motors For A Low-Speed Direct-Driven Mixer

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Design, Optimization and Comparison of Permanent Magnet Motors For A Low-Speed Direct-Driven Mixer

Uploaded by

Copyright:

Available Formats

Design, Optimization and Comparison

of Permanent Magnet Motors for a

Royal Institute of Technology

Copyright c Florence Libert, Sweden 2004

Induction motors are commonly coupled to a gearbox when an application

In this thesis, low-speed direct-drive PM motors are investigated for a

The designs of the PM machines are conducted by solving an opti-

Designs of low-speed radial-ﬂux PM machines with high pole numbers

The design of axial-ﬂux PM motors is treated separately in the thesis.

Since I started to work on my Ph.D., several people have been in-

I am grateful to Peter Lönn, who was helping me with my various

2 PM machines for low-speed direct drives 7

3 Properties of low-speed radial-flux PM motors 21

4 Optimization procedure and comparison of designs with dis-

6 Results and comparison of designs with concentrated wind-

7 Design and results of axial-flux PM motors 93

8 Conclusions and future work 109

Glossary of symbols and acronyms 117

A Direct-driven machines manufacturers 121

B Finite element methods simulations 123

D Thermal study 131

1.1 Permanent magnet synchronous machines

Nowadays, electrical motors account for 65 % of the worldwide energy

For low-speed applications, below 500 rpm, PM machines may further

1.2 Low-speed direct-drive applications

Figure 1.1: Some direct-drive applications: Wind turbines from Jeumont

1.3 The project application: The banana-blade mixer

Figure 1.2: ITT Flygt’s banana-blade mixer [11].

Table 1.1: Characteristics of the 2-pole induction motor.

1.4 Goal of the project

Figure 1.3: Description of the targeted PM drive.

Table 1.2: Design speciﬁcations of the direct-drive system.

1.5 Content of the thesis

A low-speed PM motor features many poles, which makes its design

Diﬀerent topologies of PM machines are adapted for low-speed direct

Chapter 3 deals with the geometrical, magnetic and electric properties

In chapter 4, the design procedure for the radial-ﬂux PM machines is

The concentrated windings for low-speed high-pole number PM ma-

In chapter 6, the optimized designs with concentrated windings are

2.1 Radial-ﬂux PM synchronous machine

Radial-ﬂux PM synchronous machines (RFPM) are the most conventional

flux direction Rotor Shaft

Figure 2.1: RFPM machine with ﬂux and current ﬂows.

2.1.1 Surface-mounted PM machines with inner rotor

The main advantage of the Surface-Mounted PM (SMPM) machine is

The main direct-drive application of the inner-rotor SMPM machine is

Figure 2.2: Cross-section of a surface-mounted PM motor with inner rotor

2.1.2 Surface-mounted PM machines with outer rotor

The machine consists of a stationary wound stator located in the centre of

Figure 2.3: Cross-section of a surface-mounted PM motor with outer rotor

Some advantages of this conﬁguration are:

• The rotor diameter is larger than for conventional radial-ﬂux machines,

2.1.3 Inset PM machines

As the SMPM machines, the inset PM machines have permanent magnets

Figure 2.4: Cross-section of an inset PM motor (one pole pair).

2.1.4 Buried PM machines

Many diﬀerent possibilities for the placement of the permanent mag-

V-shaped permanent magnets

Figure 2.5: Cross-section of a V-shaped buried PM motor (one pole pair).