ELECTRICAL

MACHINE ANALYSIS
USING FINITE ELEMENTS
 POWER ELECTRONICS AND
A P P L I C AT I O N S S E R I E S
Muhammad H. Rashid, Series Editor
University of West Florida

PUBLISHED TITLES
Complex Behavior of Switching Power Converters
Chi Kong Tse

DSP-Based Electromechanical Motion Control

Hamid A. Toliyat and Steven Campbell

Advanced DC/DC Converters

Fang Lin Luo and Hong Ye

Renewable Energy Systems: Design and Analysis with

Induction Generators
M. Godoy Sim~oes and Felix A. Farret

Uninterruptible Power Supplies and Active Filters

Ali Emadi, Abdolhosein Nasiri, and Stoyan B. Bekiarov

Modern Electric, Hybrid Electric, and Fuel Cell Vehicles:

Fundamentals, Theory, and Design
Mehrdad Eshani, Yimin Gao, Sebastien E. Gay, and Ali Emadi

Electric Energy: An Introduction

Mohamed El-Sharkawi

Electrical Machine Analysis Using Finite Elements

Nicola Bianchi
ELECTRICAL
MACHINE ANALYSIS
USING FINITE ELEMENTS

NICOLA BIANCHI

Boca Raton London New York Singapore

A CRC title, part of the Taylor & Francis imprint, a member of the
Taylor & Francis Group, the academic division of T&F Informa plc.
CRC Press
Taylor & Francis Group
6000 Broken Sound Parkway NW, Suite 300
Boca Raton, FL 33487-2742
© 2005 by Taylor & Francis Group, LLC
CRC Press is an imprint of Taylor & Francis Group, an Informa business

No claim to original U.S. Government works

Version Date: 20110713

International Standard Book Number-13: 978-1-4200-5787-4 (eBook - PDF)

This book contains information obtained from authentic and highly regarded sources. Reasonable
efforts have been made to publish reliable data and information, but the author and publisher cannot
assume responsibility for the validity of all materials or the consequences of their use. The authors and
publishers have attempted to trace the copyright holders of all material reproduced in this publication
and apologize to copyright holders if permission to publish in this form has not been obtained. If any
copyright material has not been acknowledged please write and let us know so we may rectify in any
future reprint.

Except as permitted under U.S. Copyright Law, no part of this book may be reprinted, reproduced,
transmitted, or utilized in any form by any electronic, mechanical, or other means, now known or
hereafter invented, including photocopying, microfilming, and recording, or in any information stor-
age or retrieval system, without written permission from the publishers.

For permission to photocopy or use material electronically from this work, please access www.copy-
right.com (http://www.copyright.com/) or contact the Copyright Clearance Center, Inc. (CCC), 222
Rosewood Drive, Danvers, MA 01923, 978-750-8400. CCC is a not-for-profit organization that pro-
vides licenses and registration for a variety of users. For organizations that have been granted a pho-
tocopy license by the CCC, a separate system of payment has been arranged.

Trademark Notice: Product or corporate names may be trademarks or registered trademarks, and are
used only for identification and explanation without intent to infringe.
Visit the Taylor & Francis Web site at
http://www.taylorandfrancis.com
and the CRC Press Web site at
http://www.crcpress.com
Preface

For a long time the course Design Methodology of Electrical Machines has been
a teaching course for electrical engineering aimed at electrical applications.
This is due both to the subject that is dealt with and to the organization of
the course by Prof. Luciano Merigliano, based on many years’ experience
with electromechanical design.
The laboratory activities of the course started in 1996, directed by myself
along with Prof. Silverio Bolognani, who was the teacher in charge in those
years, and to whom the credit for course innovation has to be given.
This book reports some of the laboratory exercitations of the course Design
Methodology of Electrical Machines. The aim is to examine the key concepts of
the finite element method, as well as to focus attention on the applications
of such a method to electrical machine analysis.
Not long after the course began, the students expressed the need to have
a text illustrating the analysis of the electric and magnetic devices by means
of the finite element method. In addition, the electromechanical industries,
which require accurate analyses, also had need for such a text. This was a
result of the popularity of software programs for the finite element analysis
of electromagnetic problems, some of which are even available on the Web.
Among them, some well-known packages are ANSYS, CADEMA, FEMM,
FLUX, MAXWELL, MEGA, QFIELD, and VECTOR-FIELD.
A first draft of this text was published in Italian in 2001, for the electrical
engineering students at the University of Padova. It was also adopted at
other Italian universities and by some industries. The present text is essen-
tially a translation of the Italian version of the same text, with a few revisions.
The book is organized to meet the goal described above. Several examples
of finite element analysis of the electrical machines are pointed out. Some
repetitions in the various chapters are intentional, for the sake of a more
fluent treatment. Some theoretical concepts, numerical techniques, and sim-
ulation methods are reported gradually in the various chapters, in an effort
to make for an easier read and to avoid many boring introductory chapters.
To the aim of facilitating understanding, some electrotechnical principles
and some electrical machines concepts are sometimes reviewed. Moreover,
simple numerical algorithms, written using MATLAB or BASIC software,
are also reported. They are examples of processing the field analysis results
and of the automatic computation procedures.
Let me thank my two generous teachers, Prof. L. Merigliano and Prof. S.
Bolognani, for the unbroken respect and the constant teaching, and not only
in the technical field. Let me also thank my friends and colleagues Prof. F.
Dughiero and Dr. A. Tortella, whose help was essential for the fulfillment of
this text. Last, but not least, I would like to express my thanks to my students,
whose lively curiosity and diligent work contributed significantly to the
improvement of the text.

Nicola Bianchi
The Author

Nicola Bianchi, Ph.D., received the Laurea degree in electrical engineering

from the University of Padova in 1991, obtaining full marks and praise. His
Laurea thesis dealt with the electromechanical analysis and design of a
synchronous reluctance motor. The thesis was rewarded by the AEI (Italian
Electrical Association) with the “Stefano e Flora Bodoni” award for new
engineers in 1991. He received his Ph.D. degree in 1995, presenting a thesis
titled “PM Motors for AC Drives: Electromagnetic Analysis and Innovative
Design Procedures” (in Italian).
Since 1998, he has worked at the Electric Drives Laboratory at the Depart-
ment of Electric Engineering of the University of Padova, as an assistant
professor. His research activity is in the field of analysis of the performance,
the determination of the parameters, and the development of design proce-
dures for electrical motors, considering conventional and innovative
machines, design by using optimization techniques, together with finite
element analysis.
Since 1998, he has been an IEEE Member and Electric Machines Committee
member. He is a technical reviewer of technical transactions, journals, and
conferences. In addition, he has served as chairman at various conferences
in the same field.
Since 2002 he has been recognized as an associate professor in the scientific
area “Electrical Machines, Drives and Converters.”
He contributes to several research projects between the Electric Drives
Laboratory and local companies, concerning analysis and design of electrical
motors and drives.
Dr. Bianchi is author and coauthor of several papers on the subject of
electrical machines and drives. The paper “Design Criteria of Embedded PM
Motors for Squarewave Current Control” (with S. Bolognani), presented at
the International Conference OPTIM 2000 (sponsored by IEEE IAS), Brasov
Poiana, Romania, was given the First Award in the section “Electrical Machines
and Drives.”
Dedication

To Giacomo, Chiara, the little angel, and their wonderful mother

“…a man traveling to a far country called his own servants and delivered
his goods to them. And to one he gave five talents, to another two, and
to another one, to each according to his own ability…”

Mt. 25, 14–15

Contents

1 Outline of Electromagnetic Fields ................................................ 1

1.1 Vector Analysis...............................................................................................1
1.1.1 Operations Among Vectors..............................................................1
1.1.2 Linear Integral and Flux of a Vector Field....................................2
1.1.3 Differential Operators.......................................................................3
1.1.4 Integral Identities ..............................................................................5
1.1.5 Differential Identities ........................................................................6
1.1.6 Expression of the Differential Operators within Different
Coordinate Systems ..........................................................................7
1.1.6.1 Cartesian Coordinates........................................................7
1.1.6.2 Cylindrical Coordinates.....................................................8
1.1.6.3 Spherical Coordinates ........................................................8
1.1.7 Conservative, Irrotational, Solenoidal, and Harmonic Fields .... 9
1.1.8 Time Dependence............................................................................10
1.2 Electromagnetic Fields ................................................................................ 11
1.2.1 Electric Charge and Electric Charge Density ............................. 11
1.2.2 Electric Displacement Field ........................................................... 11
1.2.3 Current Density Field.....................................................................12
1.2.4 Magnetic Flux Density Field .........................................................13
1.2.5 Vector Magnetic Potential Field....................................................13
1.2.6 Magnetic Field Strength .................................................................13
1.2.7 Specific Electric Force, Electric Field............................................15
1.2.8 Electric Voltage and Electromotive Force....................................17
1.2.9 Poynting’s Vector ............................................................................18
1.2.10 Maxwell’s Stress Tensor .................................................................19
1.3 Fundamental Equations Summary ...........................................................21
1.3.1 Maxwell’s Equations.......................................................................22
1.3.2 Constitutive Relationships and Continuity Equation ...............22
1.3.3 Laplace, Poisson, Helmholtz Equations ......................................23
References...............................................................................................................24

2 Basic Principles of Finite Element Methods ............................. 25

2.1 Introduction ..................................................................................................25
2.2 Field Problems with Boundary Conditions ............................................26
2.2.1 Meaning of the Differential Operator L ......................................26
2.2.2 Boundary Conditions .....................................................................27
2.3 Classical Method for the Field Problem Solution ..................................28
2.4 The Classical Residual Method (Galerkin’s Method)............................29
2.5 The Classical Variational Method
(Rayleigh-Ritz’s Method)............................................................................31
2.5.1 A Field Problem Solution by Means of the Variational
Method ..............................................................................................31
2.5.2 Definition of the Modified Variationals ......................................33
2.5.3 Natural Boundary Conditions ......................................................33
2.6 The Finite Element Method .......................................................................34
2.6.1 Partition of the Domain .................................................................34
2.6.2 Choice of the Interpolating Function...........................................35
2.6.3 Formulation of the System ............................................................35
2.6.4 Solution of the Problem .................................................................36
References...............................................................................................................36

3 Applications of the Finite Element Method to

Two-Dimensional Fields .............................................................. 39
3.1 Introduction ..................................................................................................39
3.1.1 Statement of the Two-Dimensional Field Problem....................39
3.1.2 Application of the Variational Method .......................................40
3.2 Linear Interpolation of the Function φ.....................................................43
3.2.1 Computation on the Function φm(x,y) .........................................46
3.3 Application of the Variational Method ....................................................46
3.3.1 Functional Referred to the m-th Element ...................................47
3.3.2 Functional Referred to the Whole Domain.................................49
3.3.3 Assigning the Essential Boundary Conditions ..........................51
3.3.4 The Problem’s Solution ..................................................................51
3.4 Simple Descriptions of Electromagnetic Fields ......................................51
3.4.1 An Electrostatic Field......................................................................51
3.4.2 A Stationary Current Field ............................................................53
3.4.2.1 By Means of the Scalar Potential ...................................53
3.4.2.2 By Means of the Vector Potential...................................54
3.4.3 A Magnetostatic Field.....................................................................55
3.4.4 A Magnetostatic Field without Current ......................................57
3.4.5 A Magnetic Field with Permanent Magnets...............................57
3.5 Appendix: Integration in Triangular Elements.......................................59

4 The Analysis Procedure Using the Finite Element Method .... 61

4.1 Introduction ..................................................................................................61
4.2 Reduction of the Field Problem to a Two-Dimensional Problem .......61
4.3 Boundary Conditions..................................................................................63
4.3.1 Dirichlet’s Condition ......................................................................63
4.3.2 Neumann’s Condition ....................................................................64
4.3.3 Periodic Condition ..........................................................................64
4.4 Computation of the Solved Structure ......................................................66
 4.4.1 Drawing the Flux Lines .................................................................66
4.4.2 Magnetic Flux and Flux Linkage..................................................66
4.4.3 Joule Power Losses .........................................................................68
4.4.4 Magnetic Energy..............................................................................68
4.4.5 Magnetic Coenergy .........................................................................69
4.4.6 Magnetic Forces...............................................................................69
4.4.6.1 Tensor of Maxwell’s Strength .........................................69
4.4.6.2 The Virtual Works Method..............................................70
4.4.7 Determination of the Electrical Parameters................................71

5 Cylindrical Magnetic Devices ..................................................... 73

5.1 Introduction ..................................................................................................73
5.2 Analytical Study of the Magnetic Device................................................73
5.2.1 Computation of the Magnetic Quantities ...................................74
5.2.2 Magnetic Energy and Coenergy ...................................................76
5.2.3 Apparent Inductance and Differential Inductance....................77
5.2.4 Mechanical Forces ...........................................................................79
5.3 Finite Element Analysis ..............................................................................82
5.3.1 Formulation of the Problem for the Finite Element
Analysis.............................................................................................82
5.3.2 Computation of the Solved Structure..........................................84
5.3.2.1 Magnetic Flux Density .....................................................84
5.3.2.2 Flux Linkage ......................................................................84
5.3.2.3 Magnetic Energy and Coenergy.....................................85
5.3.2.4 Inductances ........................................................................86
5.3.2.5 Electromagnetic Forces ....................................................87
5.4 Example.........................................................................................................88
References...............................................................................................................94

6 The Single-Phase Transformer .................................................... 95

6.1 The Single-Phase Transformer...................................................................95
6.2 Equivalent Electric Circuit of the Transformer.......................................96
6.3 Computation of the No-Load Inductances .............................................99
6.3.1 Statement of the Problem ............................................................100
6.3.2 Computation on the Solved Structure.......................................101
6.3.2.1 Magnetic Flux Density ...................................................101
6.3.2.2 Magnetic Energy .............................................................101
6.3.2.3 Main Flux .........................................................................102
6.3.2.4 Flux Linkage ....................................................................102
6.3.2.5 Induced EMF ...................................................................103
6.3.2.6 Self- and Mutual Inductances.......................................103
6.3.2.7 Joule Power Losses .........................................................104
6.3.3 Effect of the Nonlinear B-H Curve ............................................104
 6.3.4Estimation of the Iron Losses......................................................105
6.3.5Example .........................................................................................105
6.3.5.1 Computation with Linear Iron .....................................108
6.3.5.2 Computation with Nonlinear Iron............................... 110
6.3.5.3 Effect of the Equivalent Air-Gap.................................. 111
6.4 Determination of the Leakage Inductances .......................................... 113
6.4.1 Statement of the Problem ............................................................ 114
6.4.2 Computation on the Solved Structure....................................... 115
6.4.2.1 Magnetic Flux Density ................................................... 115
6.4.2.2 Magnetic Energy ............................................................. 115
6.4.2.3 Flux Linkage .................................................................... 115
6.4.2.4 Induced EMF ................................................................... 116
6.4.2.5 Leakage Inductances ...................................................... 116
6.4.2.6 Electric Forces.................................................................. 117
6.4.3 Simulation with Induced Currents ............................................ 118
6.4.4 Example .......................................................................................... 119
6.5 Algorithm for the Construction of the Magnetizing Characteristic
of the Transformer .....................................................................................122
References.............................................................................................................124

7 Single-Phase Variable Reactance .............................................. 125

7.1 The Single-Phase Variable Reactance .....................................................125
7.2 Computation of the Reactance ................................................................125
7.2.1 Statement of the Field Problem ..................................................126
7.2.2 Computation on the Analyzed Structure..................................127
7.2.2.1 Magnetic Flux Density ...................................................127
7.2.2.2 Magnetic Energy .............................................................127
7.2.2.3 Flux Linkage ....................................................................128
7.2.2.4 Electrical Voltage.............................................................128
7.2.2.5 Self-Inductance ................................................................128
7.2.2.6 Electric Forces..................................................................129
7.2.3 Considerations with Nonlinear B-H Curve..............................130
7.3 Example.......................................................................................................131
7.3.1 Inductance Adjustment ................................................................132
7.3.2 Definition of the Winding............................................................133
7.3.3 Analysis...........................................................................................134
References.............................................................................................................140

8 Synchronous Generators ............................................................ 141

8.1 Introduction ................................................................................................141
8.2 Computation of the No-Load Characteristic ........................................143
8.2.1 Computations after the Field Solution ......................................144
8.2.1.1 Flux Linkage ....................................................................145
8.2.1.2 Induced EMF ...................................................................146
8.3 Computation of the Direct-Axis Inductance .........................................148
8.3.1 Computation on the Solved Structure.......................................150
8.3.1.1 Computation of Ld by Means of the Magnetic
Energy...............................................................................150
8.3.1.2 Computation of Ld by Means of the Flux Linkage ...151
8.3.1.3 Computation of Ld by Means of the Air-Gap Flux
Density..............................................................................151
8.4 Computation of the Quadrature Axis Inductance ...............................152
8.4.1 Computation of Lq by Means of the Magnetic Energy
(Linear Case) ..................................................................................153
8.4.2 Computation of Lq by Means of the Flux Linkage..................153
8.4.3 Computation of Lq by Means of the Air-Gap Flux Density...153
8.5 Self- and Mutual Inductances..................................................................154
8.6 Saturation Effect.........................................................................................154
8.7 Computation of Ld and Lq with Any Current ......................................155
8.8 Computation of the Machine Characteristics .......................................156
8.9 Example.......................................................................................................158
8.10 Appendix: The Transformation abc-dq..................................................161
References.............................................................................................................165

9 Surface-Mounted Permanent Magnet Motors ......................... 167

9.1 Introduction ................................................................................................167
9.2 Computation of the No-Load Characteristic ........................................170
9.2.1 Computation on the Solved Structure.......................................171
9.2.1.1 Flux Linkage ....................................................................171
9.2.1.2 Induced EMF ...................................................................171
9.2.2 Computation for Various Rotor Positions.................................171
9.2.2.1 Flux Linkage and Induced EMF ..................................171
9.2.2.2 Cogging Torque...............................................................172
9.3 Computation of the Inductances.............................................................175
9.4 Computation of the Torque......................................................................176
9.4.1 Computation by Means of Maxwell’s Stress Tensor...............176
9.4.2 Computation by Means of the Virtual Work Principle ..........177
9.4.3 Computation by Means of Flux Linkages and Currents........177
9.4.4 Dynamic Computation.................................................................177
9.5 Example.......................................................................................................179
9.5.1 No-Load Simulation .....................................................................179
9.5.2 Simulations of Operations Under Load ....................................181
References.............................................................................................................182
Further References on AC Drives ....................................................................183

10 Interior Permanent Magnet and Reluctance Synchronous

Motors .......................................................................................... 185
10.1 Introduction ................................................................................................185
10.2 Characteristic Motor Parameters ............................................................188
 10.2.1 Simplified Magnetic Model .........................................................188
10.2.2 Cross-Coupling Effect...................................................................191
10.3 Torque-Speed Characteristic ....................................................................192
10.3.1 Example ..........................................................................................193
10.4 Algorithm for an Automatic Computation ...........................................194
10.5 Synchronous Reluctance Motors.............................................................201
10.5.1 The Transversally Laminated Reluctance Motor .....................202
10.5.2 The Axially Laminated Reluctance Motor ................................203
References.............................................................................................................204

11 Self-Starting Single-Phase Synchronous Motors .................... 207

11.1 Introduction ................................................................................................207
11.2 Definition of the Motor Model ................................................................208
11.3 Computation of the Electrical Parameters ............................................209
11.3.1 Flux Linkage due to the Permanent Magnet............................210
11.3.2 Self-Inductance Coefficient ..........................................................210
11.4 Computation of the Torque......................................................................210
11.5 Analysis of the Dynamic Performance .................................................. 211
11.5.1 Simulation at Constant Speed.....................................................212
11.5.2 Simulation of the Full Dynamic Performance..........................213
11.6 Example.......................................................................................................214
11.6.1 Computation of the Flux Linkage due to the Permanent
Magnet ............................................................................................216
11.6.2 Computation of the Inductance..................................................216
11.6.3 The Torque Matrix ........................................................................216
11.6.4 Simulation of the Dynamic Performance..................................218
11.6.4.1 Constant Speed Simulation...........................................218
11.6.4.2 Simulation of the Rotor Starting ..................................219
11.7 Two-Dimensional Linear Interpolation..................................................220
11.7.1 Description of the Interpolating Algorithm..............................222
11.8 Numerical Codes for the Motor Analysis .............................................225
11.8.1 Code of the Computation of the Torque at Constant
Speed ...............................................................................................225
11.8.2 Code of the Computation of the Dynamic Performance .......226
References.............................................................................................................227

12 Switched Reluctance Motors ..................................................... 229

12.1 Introduction ................................................................................................229
12.2 Operating Principle ...................................................................................230
12.2.1 Linear Model..................................................................................231
12.2.2 Dynamic Performance ..................................................................232
12.3 Field Problem Statement ..........................................................................233
12.4 Computation on the Solved Structure ...................................................234
12.4.1 Dynamic Analysis .........................................................................234
12.5 Example.......................................................................................................236
12.5.1 Effect of the Number of Elements of the Mesh .......................238
12.5.2 Simulation of the Dynamic Performance..................................239
References.............................................................................................................241

13 Three-Phase Induction Motors .................................................. 243

13.1 Introduction ................................................................................................243
13.2 Simulations of the Indirect Tests.............................................................246
13.2.1 Equivalent Circuit .........................................................................246
13.2.2 No-Load Test..................................................................................247
13.2.3 Locked Rotor Test .........................................................................249
13.2.4 Construction of the Equivalent Circuit .....................................252
13.2.5 Example ..........................................................................................253
13.2.5.1 No-Load Simulation .......................................................253
13.2.5.2 Locked Rotor Simulation...............................................256
13.2.5.3 Motor Performance.........................................................257
13.3 Motor Analysis Using Simulations Under Load..................................258
13.3.1 Magnetic Field Equations Under Load .....................................259
13.3.2 Alternative Formulation...............................................................261
13.3.3 Nonlinearity of the Magnetic Materials ....................................263
13.3.4 Link to an External Circuit ..........................................................264
13.3.5 Computation of the Mechanical Torque....................................265
13.3.6 Example ..........................................................................................265
References.............................................................................................................265

Appendix .............................................................................................. 267

Index ..................................................................................................... 273

List of Symbols

A Tm magnetic vector potential

Am m2 surface of the m-th finite element
a — transformation ratio
B T magnetic flux density
BdM BqM T d-axis and q-axis fundamental harmonic flux
density
Bg T air-gap flux density
Br T radial component of the flux density
Bres T residual flux density of the permanent magnet
Bϑ T azimuthal component of the flux density
C F capacitance
D C/m2 electric displacement
D m diameter, air-gap diameter
De m outer diameter
D integration domain (surface or volume)
d m diameter
dCu m wire diameter
E V/m specific electric force
E, e V EMF
E1 E2 V EMF induced in primary or secondary winding
Erms V RMS voltage
Ec V/m Coulombian electric field
Ei V/m induced electric field
Em V/m motional specific electric force
F N force
Ft Fn N tangential and normal force (to a surface or to
a line)
F functional
f forcing functional
f Hz frequency
fr Hz rotor frequency
fa fb fc A magnetic voltage of the a, b, c phases
fd fq A d-axis and q-axis magnetic voltages
g m air-gap thickness
H A/m magnetic field strength
Hc A/m coercitive force of the permanent magnet
Hknee A/m magnetic field strength at the knee of the BH
curve
Hn Ht A/m normal and tangential component of the
magnetic field strength
hw m winding height
hm m permanent magnet height
ht m tooth (or slot) height
I, i A electric current intensity
i1 i2 A primary and secondary currents
I1o A primary current at no-load
Ic A current flowing in each coil
Ie A excitation current
Irms A RMS current
IL A line current
IM A maximum value of the current
Im A magnetizing current
Io A phase current at no-load
Iq A equivalent current within the q-th slot
ia ib ic A current of the a, b, and c phase
id iq A d-axis and q-axis current
J kgm2 inertia
J A/m2 electric current density
Jsd A/m direct axis linear current density
Jsq A/m quadrature axis linear current density
kB Nms coefficient of viscous friction
kf — lamination packing factor
kjq — fill factor of the q-th slot by the j-th phase
kw — winding factor
L — generic differential operator (e.g., laplacian
operator)
L m axial length of a structure with planar
symmetry
L H magnetic inductance
L1 L2 H primary and secondary self-inductance
L1m L2m H primary and secondary magnetizing
inductances
L1σ L2σ H primary and secondary leakage inductances
La Mba H self- and mutual inductances of each phase
Ld Lq H d-axis and q-axis synchronous inductances
Lapp H apparent inductance
Ldif H differential inductance
LFe m actual iron length of machine with planar
symmetry
LM H magnetizing inductances of an induction
motor
Lσ H leakage inductance
l m line
lg m line passing in the middle of the air-gap
lcv m ventilation channel length
lm m permanent magnet thickness
lave m average length of a turn
M A/m magnetization of a permanent magnet
M H mutual inductance
M — number of finite elements in the domain
N — number of nodes of the finite elements
N A/m electric potential vector
N — number of series conductors per phase
N1 N2 — number of turns of primary and secondary
winding
Nc — number of sampled points
Ncv — number of channels for cooling
Ne — number of turns of the excitation winding
Nsp — number of turns
n — normal unity vector (to a line or to a surface)
nb rpm base speed
nfw rpm flux-weakening speed (in the constant power
region)
npm — number of parallel paths
nq — number of conductors within a slot
P — dependence on the point P, e.g., on the
coordinate (x,y,z)
p — number of pole pairs
pFe W iron losses
pJ W Joule power losses
pn N/m2 magnetic pressure
ps W/kg specific iron losses
Q — number of total slots
Qr Qs — number of rotor and stator slots
q C electric charge
qs — number of stator slots per pole per phase
r, ϑ, z — cylindrical coordinates
R, r — residue
R Ω electric resistance
Rr Ω rotor resistance
Rg H–1 magnetic reluctance corresponding to the air-
gap
S m2 surface
SCu+ SCu– m2 equivalent surface of the conductors
SFe m2 actual iron surface
Sn VA nominal VA power
Sq m2 surface of the slot
s — slip
[SS] — stiffness matrix
[T] — known terms vector matrix
T Nm electromagnetic torque
Tb Nm base torque
Tcog Nm cogging torque
Tfw Nm flux-weakening torque (in the constant power
region)
Tabc/dq — transformation matrix abc/dq
t — tangent unity vector to a line or to a surface
t s time
t* s reference instant time
U0 V line voltage at no-load
Uνcos Uνsin V voltage harmonic component of order ν
u — coordinate unity vector
V, v V electric voltage
V V electric potential
Vdc V direct-current source voltage
v1 v2 V primary and secondary voltage
vd vq V d-axis and q-axis electric voltages
vm m/s speed of the medium on which the charge is
bound
vN V star-center potential
vq m/s speed of the electric charge
vρ m/s speed of the electric charge density
x, y, z — Cartesian coordinates
yq — slot pitch (in number of slots)
wi — i-th weight function
wm J/m3 magnetic energy density
wt m tooth width
Wm J magnetic energy
W′m J magnetic coenergy
Wel J electric work
Wmech J mechanic work

Greek symbols
αc rad electric slot angle
βr rad electric shortened angle
βr βs rad rotor and stator width angle
δ rad load angle
∆i A current variation
∆ϑ red mechanical angle variation
ε F/m electric permittivity
φ, ϕ, ψ — unknown potential to be determined
φ* — function that approaches the unknown
potential
Φ Wb magnetic flux
Φo Wb magnetic flux at no-load
ϕ rad power angle
ϕ rad initial voltage phase
Φj — j-th coefficient to approach the potential φ
[φ] — matrix vector of the unknown coefficients Φi
Γ — boundary of the domain (line or surface)
γFe kg/m3 specific iron weight
Λ, λ Wb, Vs flux linkage
Λd Λq Vs d-axis and q-axis flux linkages
Λrms Vs RMS flux linkage
Λj Vs flux linkage of the j-th phase (j = a, b, c)
λ1 λ2 Vs primary and secondary flux linkages
λ1σ λ2σ Vs primary and secondary leakage flux linkages
λpm Vs flux linkage due to the permanent magnet
λr λs rad rotor and stator pole pitch angle
µ H/m magnetic permeability
µr — relative magnetic permeability
h — efficiency
νj — j-th interpolanting function
ρ C/m3 volume density of electric charge
ρs C/m2 surface density of electric charge
ρ Ωm electric resistivity, ρ = σ −1
ϑ rad angular coordinate (electric angle)
ϑm rad angular coordinate (mechanic angle)
ϑoff rad switch-off angle
ϑon rad switch-on angle
ϑr rad angular coordinate referred to the rotor polar
axis
ϑs rad angular coordinate referred to the stator
a-phase axis
ξ — saliency ratio, ξ = Lq/Ld
ω rad/s electric frequency
ωm rad/s mechanic angular speed
ωr rad/s rotor electric angular speed ωr = 2πfr
σ S/m electric conductivity
τ m3 volume
τCu m3 conductive material volume
τm Nm instantaneous electromagnetic torque
τL Nm load torque
Ψ A magnetic voltage

Further Symbols
Im imaginary part
Re real part
~ conjugate complex
1
Outline of Electromagnetic Fields

In this chapter, some basic concepts of vector analysis and electromagnetic

fields are outlined. These concepts will be used in the remainder of the book.
The aim is to furnish a concise treatment and a practical formulary, but keep
in mind that it is not an exhaustive one. Any reader interested in mastering
this subject should look it up in the books listed in the References.

1.1 VECTOR ANALYSIS

In a three-dimensional space, let V = V(P) be a scalar quantity that defines
a scalar field, and A = A(P) be a vector quantity that defines a vector field.
The dependence on the point P is expressed by using a reference system
with curvilinear orthogonal coordinates: they are Cartesian (x,y,z), cylindri-
cal (r,ϑ,z), or spherical (r,ϑ,ϕ) coordinates.
In general, each of these quantities can be a function of the time, that is,
V = V(P,t) and A = A(P,t). This dependence is considered in Section 1.1.8.
However, in order to simplify the text, the explicit notation pointing out the
dependence of the various quantities on the point P and the time t is omitted.

1.1.1 Operations Among Vectors

Let A and B be two vectors, with
A
and
B
their magnitude respectively
and γ the angle between their oriented directions. The scalar product is the
real number given by
A · B =
A

B
cos γ (1.1)

The vector product is the vector given by

A × B = (
A

B
sin γ) nAB (1.2)

where nAB is the unity vector normal to the plane defined by A and B. The
vector (1.2) results normal to both vector A and vector B, and its direction

1
2 Electrical Machine Analysis Using Finite Elements

A
A⫻B A P2
A n
B t t
S
A P l P
l
γ
P1

(a) (b) (c) (d)

FIGURE 1.1
Operations among the vectors.

corresponds to the advancement of a right screw, rotating so as to place the

vector A on the vector B, as shown in Figure 1.1(a).
Using the Cartesian coordinates, the two vectors are A = (Ax, Ay , Az) =
Axux + Ayuy + Azuz and B = (Bx, By , Bz) = Bxux + Byuy + Bzuz, where ux, uy ,
and uz are the coordinate unity vectors. The scalar and the vector products
become

A ⋅ B = A x B x + A y B y + A zB z (1.3)

ux uy uz
A × B = Ax Ay Az
Bx By Bz (1.4)

( ) ( )
= A y B z − A zB y u x + A zB x − A xB z u y + A xB y − A y B x u z( )
The following properties exist:

2
A·A = A A·(B + C) = A· B + A· C

A· B = B ·A A × (B + C) = A × B + A × C (1.5)

A × B = −B × A A × (B × C) = (A· C) B − (A· B) C

1.1.2 Linear Integral and Flux of a Vector Field

The linear integral of a vector field A along a curve l, from point P1 to point
P2 and oriented in each point by the unity vector t, tangent to the line as
shown in Figure 1.1(b), is
P2
Γ 12 =
∫
P1
A ⋅ t dl (1.6)

The loop integral of the vector field corresponds to the linear integral along
a closed curve, shown in Figure 1.1(c), which is
Outline of Electromagnetic Fields 3

Γ=

∫ A ⋅ t dl
l
(1.7)

The flux of the vector field A through the surface S, oriented by the normal
unity vector n in each point, as shown in Figure 1.1(d), is given by

Φ=
∫
S
A ⋅ n dS (1.8)

In the case in which the surface is closed, and n is assumed conventionally

to have external direction, Φ defines the flux going out through the surface
S itself.

1.1.3 Differential Operators

The gradient of a scalar field V is a vector function defined in each point P
by a direction corresponding to the maximum directional derivative of V,
defined by the unity vector umax, and by a magnitude equal to this derivative,
given by

 ∂V 
gradV = max   u max (1.9)
 ∂l 

Along any direction defined by the unity vector u, the scalar product of
gradV by u gives the derivative of V along the direction of u, which is

∂V
gradV·u = (1.10)
∂l u

Finally, if the field V is not continuous through a surface Σ, characterized

by the normal unity vector nΣ , and it assumes the values V1 and V2 on the
two sides of Σ, then a surface gradient can be defined as

(
grad Σ V = V2 − V1 n Σ ) (1.11)

The divergence of a vector field A is a scalar function defined in each point

P by the ratio between the flux of A going out through a closed surface Sc
and the volume τ contained by Sc, when the volume tends to the point P
itself, which is

divA = lim

∫ Sc
A· n dS
(1.12)
τ→0 τ
4 Electrical Machine Analysis Using Finite Elements

The infinitesimal flux dΦ of the vector A going out through a closed

surface, that contains the infinitesimal volume dτ, is given by dΦ = divAdτ.
Then the divergence divA > 0 means a point where the lines of the field A
arise, while divA < 0 means a point where the lines of the field A terminate.

Finally, if the field A is not continuous through a surface Σ, with normal

unity vector nΣ , and assumes the values A1 and A2 on the two sides of Σ,
then a surface divergence can be defined as

( )
div Σ A = A 2 − A1 ⋅ n Σ (1.13)

The curl of a vector field A is a vector function defined in each point P by

the ratio between the integral of A along a closed loop l and the surface S
bordered by l, when S tends to the point P itself, which is

curlA = lim

∫ A ⋅ t dl
l
(1.14)
S→0 S

The infinitesimal loop integral dΓ of the vector A along a closed loop l

that borders the infinitesimal surface dS, oriented by n, is given by curlA·ndS.
Then the curl represents the vortexes around which the lines of the vector
A go round.
If the field A is not continuous through a surface Σ, with normal unity
vector nΣ , and assumes the values A1 and A2 on the two sides of Σ, then a
surface curl is defined as

(
curl Σ A = − A 2 − A1 × n Σ) (1.15)

The laplacian of a scalar field V is given by the divergence of the vector

gradient of the scalar function V, as

∇ 2 V = div gradV (1.16)

The laplacian of a vector field A is given by

∇ 2 A = grad divA − curl curlA (1.17)

Referring to the differential operators defined above, some properties are

reported in the following. U and V are scalar fields, A and B are vector fields,
k and h are generic constants.
Outline of Electromagnetic Fields 5

1.1.4 Integral Identities

The gradient theorem states that the linear integral from point P1 and point
P2 of a vector gradient gradV depends only on the value of V in the points
P1 and P2:

P2
Γ 12 =
∫ P1
gradV ⋅ t dl = V2 − V1 (1.18)

that, in the case of loop integral, is equal to zero. In addition, it is

∫ gradV dτ =
∫ Vn dS
τ S
(1.19)

The divergence theorem (or Gauss’s theorem) states that the volume integral
of a divergence divA is equal to the integral of the vector A through the
closed surface that contains the volume, which is the flux of A going out
through the closed surface:

Φ=

∫ Sc
A ⋅ n dS =
∫ (divA) dτ
τ
(1.20)

The curl theorem (or the Stokes theorem) states that the integral through
the surface S of a vector curlA is equal to the linear integral of the vector A
along a closed line l bordering the surface S itself. The unity vector t tangen-
tial to the line l has to be chosen in accordance with the unity vector n normal
to the surface S (with the rule of the right screw). Finally, it is

∫ A ⋅ t dl = ∫ curlA ⋅ n dS
l S
(1.21)

First Green’s scalar theorem states that

∂V
∫ U div(h gradV) + h gradU ⋅ gradV dτ =
∫
τ Sc
hU
∂n
dS (1.22)

Second Green’s scalar theorem states that

 ∂V ∂U 
∫ U div(a gradV) − Vdiv(a gradU) dτ =
∫
τ Sc
AU
 δn
−V
δn 
dS (1.23)
6 Electrical Machine Analysis Using Finite Elements

First Green’s vector theorem states that

∫ k curlA ⋅ curlB − A ⋅ curl(k curlB) dτ =
∫

τ Sc
k(A × curlB)⋅ n dS (1.24)

Second Green’s vector theorem states that

∫ B ⋅ curl(k curlA) − A ⋅ curl(k curlB) dτ

τ
(1.25)
=

∫ k ( A × curlB − B × curlA) ⋅ n dS
Sc

1.1.5 Differential Identities

Gradient:

( )
grad VU = V(gradU) + U(gradV) (1.26)

( )
grad kU + hV = k gradU + h gradV (1.27)

∂V
gradV U =( ) ∂U
gradU (1.28)

Divergence:

( )
div UA = U(divA) + gradU ⋅ A (1.29)

( ) ( )
div kA + hB = k div A + h div B ( ) (1.30)

( )
div A × B = − A ⋅ (curlB) + (curlA) ⋅ B
(1.31)
div ( curlA ) = 0
Curl:

( )
curl UA = U ⋅ (curlA) + (gradU) × A (1.32)

( ) ( )
curl kA + hB = k curl A + h curl B ( ) (1.33)

curl gradU = 0 (1.34)

Outline of Electromagnetic Fields 7

Laplacian:

( )
∇ 2 UV = U ⋅ ∇ 2 V + V∇ 2 U + 2 gradU ⋅ gradV (1.35)

( )
∇ 2 kU + hV = k ∇ 2 U + h ∇ 2 V (1.36)

1.1.6 Expression of the Differential Operators within Different

Coordinate Systems

1.1.6.1 Cartesian Coordinates

Using Cartesian coordinates, the point P is given by P(x, y, z) and the
coordinate unity vectors are ux, uy and uz as shown in Figure 1.2(a). The
differential operators are as follows:

∂V ∂V ∂V
gradV = ux + uy + uz (1.37)
∂x ∂y ∂z

∂A x ∂A y ∂A z
divA = + + (1.38)
∂x ∂y ∂z

ux uy uz
∂ ∂ ∂
curlA =
∂x ∂y ∂z
Ax Ay Az (1.39)

 ∂A z ∂A y   ∂A x ∂A z   ∂A y ∂A x 
= −  ux +  − uy +  − uz
 ∂y ∂z   ∂z ∂x 
 ∂x ∂y 

y y y
uy
uϑ uϑ uϕ
P P P
ux ϑ ux ux
ϑ
uz uz
ϕ

O x O r O r

z z z

FIGURE 1.2
Systems of orthogonal coordinates: Cartesian (a), cylindrical (b), and spherical (c) coordinates.
8 Electrical Machine Analysis Using Finite Elements

∂2 U ∂2 U ∂2 U
∇ 2 U = div gradU = + + (1.40)
∂x 2 ∂y 2 ∂z 2

( )
∇ 2 A = grad divA − curl curlA ( )
= ∇ 2 A x u x + ∇ 2 A y u y + ∇ 2 A zu z

 ∂2 A x ∂2 A x ∂2 A x   ∂2 A y ∂2 A y ∂2 A y  (1.41)
= + + 2  x
u + + +  uy +
 ∂x ∂y ∂z   ∂x ∂y 2 ∂z 2 
2 2 2

 ∂2 A z ∂2 A z ∂2 A z 
 ∂x 2 + ∂y 2 + ∂z 2  u z
 

1.1.6.2 Cylindrical Coordinates

Using cylindrical coordinates, the point P is given by P(r, ϑ, z) and the
coordinate unity vectors are ur , uϑ and uz , as shown in Figure 1.2(b). The
differential operators are

∂V 1 ∂V ∂V
gradV = ur + uθ + uz (1.42)
∂r r ∂θ ∂z

divA =
(
1 ∂ rA r
+
)1 ∂A θ ∂A z
+ (1.43)
r ∂r r ∂θ ∂z

 1 ∂A z ∂A θ 
curlA =  −
 ∂A r ∂A z 
ur +  −
1  ∂ rA θ
uθ +  −
(∂A r )
 u z (1.44)
 r ∂θ ∂z   ∂z ∂r  r  ∂r ∂θ 

1 ∂  ∂V  1 ∂2 V ∂2 V
∇2V = r + +
r ∂r  ∂r  r 2 ∂θ2 ∂z 2
(1.45)
1 ∂V ∂2 V 1 ∂2 V ∂2 V
= + + +
r ∂rr ∂r 2 r 2 ∂θ2 ∂z 2

1.1.6.3 Spherical Coordinates

Using spherical coordinates, the point P is given by P(r, ϑ, ϕ) and the coor-
dinate unity vectors are ur , uϑ and uϕ , as shown in Figure 1.2(c). The differential
operators are

∂V 1 ∂V 1 ∂V
gradV = ur + uθ + uϕ (1.46)
∂r r ∂θ r sin θ ∂ϕ
Outline of Electromagnetic Fields 9

divA =
(
1 ∂ r Ar
2

+
) (
1 ∂ sin θ A θ
+
)
1 ∂A ϕ
(1.47)
r 2
∂r r sin θ ∂θ n θ ∂ϕ
r sin

curlA = 
(
1  ∂ sin θ A ϕ
−
)
∂A θ 
 ur +
r sin θ  ∂θ ∂ϕ 
(1.48)

 −
(
1  ∂A r ∂ sin θ rA ϕ )  (
1  ∂ rA θ
 uθ + 
)−
∂A r 
 uϕ
r sin θ  ∂ϕ ∂r  r  ∂r ∂θ 

1 ∂  2 ∂V  1 ∂  ∂V  1 ∂2 V
∇2V = r + sin θ + (1.49)
r 2 ∂r  ∂r  r 2 sin θ ∂θ  ∂θ  r sin θ
( )
2
∂ϕ 2

1.1.7 Conservative, Irrotational, Solenoidal, and Harmonic Fields

A vector field B is conservative if it exhibits a null loop integral along each
closed line in the domain of definition. As a consequence, the linear integral
of B along a generic open line results dependent only on the end points of the
line and not on its path. It is possible to combine a scalar potential Φ with the
conservative field B. The potential Φ is defined so that B = −(gradΦ + C),
where C is an arbitrary constant. The surfaces where Φ assumes an equal
value are equipotential surfaces and are always normal to the lines of the
field B.
A vector field B is irrotational if its curl is null in all the domain of definition.
A conservative field is of course an irrotational field. Conversely an irrota-
tional field is a conservative field only if it is defined in a simply connected
domain (a domain is simply connected when each closed line along which the
integral is computed may be reduced to a point, remaining always within
the domain).
A vector field B is solenoidal if its divergence is null in the whole domain
of definition. As a consequence, the flux of B along any closed surface results
in null. It is possible to express a solenoidal field B by means of a curl of a
vector potential A. This vector potential A is defined so that B = curl(A + C),
where C is an arbitrary irrotational vector field. It follows that the loop
integral of the vector potential along the closed line l is equal to the flux of
the solenoidal field through any surfaces bordered by l.
Let B be a vector field with divergence and curl given by divB = f and
curlB = g, respectively. It is convenient to split the vector field B as the sum
of an irrotational field Birr and a solenoidal field Bsol. Such a splitting up is
called the resolution of Clebsh–Helmholtz. The field Birr has to exhibit a diver-
gence equal to f, and the field Bsol has to exhibit a curl equal to g. The solution
10 Electrical Machine Analysis Using Finite Elements

is unique if the field is normal at the infinite (a field whose magnitude tends
to 1/r2 when the distance r from the origin tends to infinite). Thus, it is

divB = divBirr
B = Birr + Bsol with  (1.50)
curlB = curlBsol

A scalar field is harmonic if it satisfies the Laplace equation:

( )
∇ 2φ = div gradφ = 0 (1.51)

Then, a vector field B that is obtained as gradient of a harmonic scalar

field Φ, i.e., B = gradΦ, is irrotational, since curl(gradΦ) = 0, and solenoidal.
A vector field A is harmonic if it is irrotational and solenoidal. Hence, it
results

∇2A = 0 (1.52)

Then, the vector field B that is obtained as curl of a harmonic vector field
A, i.e., B = curlA, is solenoidal because div(curlA) = 0, and irrotational.

1.1.8 Time Dependence

Up to now, the scalar fields and vector fields have been considered as a
function exclusively of the points of the space. However, they may be also a
function of the time. Of course, the space variables and the time variable t
are considered separately. Then, it is obtained

∂V ∂ ∂A ∂
grad = gradV div = divA
∂t ∂t ∂t ∂t
(1.53)
∂A ∂ ∂A ∂ 2
curl = curlA ∇2 = ∇A
∂t ∂t ∂t ∂t

It is usual to consider that the dependence on the time is sinusoidal,

characterized by the pulsation ω. In this case the complex notation can be
used: each quantity is represented by means of a complex phasor. For
instance, referring to the component Ax(P,t) along the x–axis of the vector
field A(P,t), it is possible to introduce the correspondence:

 = A e jα x
A x = A xM cos(ωt + α x ) ↔ A x xM

= A xr + jA xi (1.54)

= A xM cos α x + jA xM sin α x
Outline of Electromagnetic Fields 11

and analogously for the other components. The vector field A, whose com-
ponents are described by the symbolic function (1.54), is indicated as

 =A
A  u +A u +A
 u (1.55)
x x y y z z

1.2 Electromagnetic Fields

1.2.1 Electric Charge and Electric Charge Density
The electric charges in one region of the space may be distinguished as free
charges and charges of polarization. The former can carry out macroscopic
displacement, while the latter depend inevitably on the molecular structure
of the media. The phenomena of attraction and repulsion among the various
bodies, on which the electric charge are posed, are generated by the total
charge, the sum of the free charges and of those of polarization. Only the
free charges are dealt with hereafter, while the charges of polarization are
considered by defining conveniently the characteristics of the media.
In any point of the space, where the infinitesimal volume ∆τ contains the
(free) charge ∆q, it is possible to define a volume density of (free) charge ρ as

∆q
ρ = lim (1.56)
∆τ→0 ∆τ

that are measured in (C/m3) and generally variable with the time.
If the element on which the charge is distributed is characterized by a
dimension negligible with respect to the other two dimensions, it is possible
to define a surface density of charge ρs, measured in (C/m2), as

∆q
ρs = lim (1.57)
∆S→0 ∆S

1.2.2 Electric Displacement Field

The spatial description of the effects due to the free electric charges that are
in restricted regions is represented by the vector field D, called the electric
displacement field. The measure unit of its magnitude is (C/m2). The fun-
damental property of the field D, in differential form and in integral form
respectively, is
divD = ρ (1.58)
12 Electrical Machine Analysis Using Finite Elements

∫ Sc
D ⋅ n dS = q (1.59)

The second equation is the well-known Gauss’s law. It indicates that the
flux of the vector D going out through the closed surface Sc, oriented by the
unity vector n normal to the surface and with external direction, is equal to
the free charge q, which is contained by the surface Sc itself.

1.2.3 Current Density Field

The movement of the electric charges is described by means of the electric
current density. Referring to a volume charge density ρ moving at a velocity
vρ, the electric current density vector J can be defined as

J = ρ vρ (1.60)

whose reference positive direction is that of the positive charges. Its magni-
tude is measured in (A/m2). The vector J defines a vector field, called the
current field. Let S be an open surface, with normal unity vector n, then the
current intensity i measured in (A) is given by

i=
∫ J ⋅ n dS
S
(1.61)

The continuity equation of the current field in differential form is

∂ρ
divJ = − (1.62)
∂t

In integral form, letting Sc a closed surface, with normal unity vector n,

the current intensity leaving the surface Sc in the time interval ∆t corresponds
to the variation of the electric charge ∆q in the volume enclosed by Sc, which
is

∆q
i out =

∫ Sc
J ⋅ n dS = −
∆t
(1.63)

It is possible to define the total current density vector as

∂D
Jtot = J + (1.64)
∂t

which is the sum of the current density vector J and the displacement current
density vector ∂D/∂t. By property (1.58) about D and the continuity Equation
(1.62), the current field Jtot is solenoidal.
Outline of Electromagnetic Fields 13

1.2.4 Magnetic Flux Density Field

The movement of the electric charges causes effects in the points of the
surrounding space, so that it is possible to define the magnetic flux density
field B. Its magnitude is measured in (T). One can observe that a test charge
δq moving within the magnetic flux density field B at a velocity vq experi-
ences a force given by δF = δq vq × B. This force is called Lorentz’s force and
will be described hereafter.
The fundamental property of the field B is that it is a solenoidal field, i.e.,
the flux of B through any closed surface Sc is null. Thus, it is

divB = 0 (1.65)

∫
Sc
B ⋅ n dS = 0 (1.66)

in differential and integral forms, respectively.

1.2.5 Vector Magnetic Potential Field

Since the magnetic flux density field B is solenoidal in the whole space, it is
suitable to define a vector magnetic potential field A, whose magnitude is
measured in (Tm), such as

B = curlA (1.67)

This relationship defines the field A apart from a generic irrotational field.
The divergence of A can be defined in an arbitrary way; the positions that
are commonly adopted are as follows.

Coulomb’s position, mainly used in stationary or quasi-stationary magnetic

fields and also used in the book:

divA = 0 (1.68)

Lorentz’s position, which is well suited in the study of rapidly time-variable

electromagnetic fields:

∂V
divA = −µε (1.69)
∂t

1.2.6 Magnetic Field Strength

Together with the flux density vector B, the magnetic field strength vector
H is introduced. The measure unit of its magnitude is (A/m). The two vector
fields are linked by the constitutive law
14 Electrical Machine Analysis Using Finite Elements

B = µH (1.70)

where µ (H/m) is the magnetic permeability of the medium. Within an

uniform medium the two fields B and H are proportional (i.e., they have
same direction and proportional magnitude), while in an anisotropic
medium their link exhibits a tensorial nature.
The fundamental property of the field H is the definition of its curl
(Ampere’s law):

∂D
curlH = J + (1.71)
∂t

In stationary or quasi-stationary magnetic condition, i.e., when the dis-

placement current density may be neglected, Equation (1.71) is reduced to

curlH = J (1.72)

In integral form, the Ampere’s law is expressed by equating the line inte-
gral of H along an oriented closed line l to the current intensity i flowing
through the surface enclosed by the line l itself, which is

∫ H ⋅ t dl = i
l
(1.73)

which is called magnetomotive force (MMF). The line integral of H along

an oriented open line l defines the magnetic voltage between the end points
P and Q of the line:

Q
Ψ PQ =
∫P
H ⋅ t dl (1.74)

The magnetic field strength H and the magnetic flux density B are wholly
defined by Equation (1.64) and Equation (1.71) together with the constitutive
equation [Equation (1.70)]. From them, the divergence of H is

H
divH = − gradµ (1.75)
µ

that is of course null in any homogeneous medium, while the curl of B is

curlB = µJ + gradµ × H (1.76)

Outline of Electromagnetic Fields 15

If the field H is irrotational in a simply connected region, a scalar magnetic

potential Ψ may be defined in this region, so that

H = −gradΨ (1.77)

1.2.7 Specific Electric Force, Electric Field

Forces of various nature can exist on the electric charges. Let δFk be any force
on a test positive charge δq, the specific electric force Ek is defined as

δFk
E k = lim (1.78)
δq→0 δq

whose magnitude is measured in (N/C).

A useful classification of the specific electric forces is given in Table 1.1.
The Coulomb specific electric force Ec takes into account that the electric
charges tend to attract and repel each other. This is a property of the points
of the space; thus it defines a vector field. Its fundamental property is to be
conservative. Consequently it is irrotational, i.e., curlEc = 0. Then, it is pos-
sible to introduce a scalar field, the electric potential field V measured in (V),
such as

E c = − gradV (1.79)

Also the specific induced electric force Ei , produced by the rate of change
of the magnetic flux density field B with the time, is a function of the points
of the space, so that it defines a vector field as well. The fundamental
property of the induced electric field is the definition of its curl, which is

∂B
curlE i = − (1.80)
∂t

TABLE 1.1
Classification of the Specific Electric Forces

 conservative
 
Ec Coulomb
E i induced
} E electric field
 
electromagn.  
Ek   non-conservative E L Lorentz
  E motional
   m
 non-electromagn. E n.e.

16 Electrical Machine Analysis Using Finite Elements

Since there is no constraint on the divergence of Ei , it is generally assumed

that it is a solenoidal field, i.e., divEi = 0. By means of this assumption,
together with the Coulomb’s position [Equation (1.68)], i.e., divA = 0, the
induced electric field is given by

∂A
Ei = − (1.81)
∂t

This relationship is particularly useful in the computation of the induced

currents in conductive media.
The electric field E, also called Maxwell’s electric field, corresponds to the
sum of the Coulomb electric field and the induced electric field, i.e.,

∂A
E = E c + E i = − gradV − (1.82)
∂t

The electric field E is defined by its divergence and its curl, that are divE =
divEc and curlE = curlEi , respectively. In a dielectric medium with electric
permittivity ε (in general of tensorial nature), the electric field E is linked to
the electric displacement field D by means of the constitutive relationship

D = εE (1.83)

The Lorentz’s specific electric force EL acts on the electric charges moving
in a magnetic flux density field B at a velocity vq with respect to the adopted
reference system. It is

EL = vq × B (1.84)

The motional specific electric force Em acts on the electric charges posed
on a conductor moving at a velocity vm in a magnetic flux density field B. It is

Em = v m × B (1.85)

Both these two specific forces are not property of the points of the space,
but they depend on the velocity of the charges and of the conductor, with
respect to the reference system. Thus, they do not define a vector field.
However, by means of a change of the reference system it is possible to
consider the effects of the motional specific forces as induced specific forces.
An example will be shown in the study of the induction motor, in Chapter 13.
The specific electric force of nonelectromagnetic nature, Ene, can be of
chemical, piezoelectric, photovoltaic nature, and so on. They are all noncon-
servative. They do not define vector fields, but they must be considered as
specific forces external to the electromagnetic system.
Outline of Electromagnetic Fields 17

The total specific electric force Et is the sum of the field E and the Lorentz’s,
motional, and external specific forces, which is

Et = E + EL + Em + Ene (1.86)

In a conductive medium, characterized by the conductivity σ (in general

of tensorial nature), the field Et is linked to the current density vector field
J by means of the constitutive relationship

J = σE t (1.87)

The nonconservative specific electric forces Eemf are rotational specific elec-
tric forces. They are called specific electromotive forces. Among them, Ei , EL,
and Em are of electromagnetic nature.

1.2.8 Electric Voltage and Electromotive Force

The electric voltage vAB, between two points A (+) e B (–), along a line l,
oriented by the tangential unity vector t, and fixed in the adopted reference
system, is given by the line integral along l from point A to point B of the
electric field E. It is

B
v AB =
∫ A
E ⋅ t dl (1.88)

If the line l is moving at a velocity vl in the adopted reference system, the

electric voltage must also take into account the motional specific electric
force acting on the charges constrained to stay on the line. The electric voltage
becomes

∫ (E + v × B) ⋅ t dl
B
v AB = l (1.89)
A

Analogously, the electromotive force (EMF) eBA, between two points B (+)
e A (–), along a line l is given by the line integral along l from point A to
point B of the specific electromotive force Eemf. It is

B
eBA =
∫A
Eemf ⋅ t dl (1.90)

In analyzing electrical machines, the induced EMF and the motional EMF
are of particular interest. The induced EMF is caused by the time-variation
of the flux density B. The motional EMF is caused by the movement of the
line in a constant field B. The sum of the two EMFs, computed along a closed
18 Electrical Machine Analysis Using Finite Elements

line lc, represents the total EMF of electromagnetic nature. It is expressed by

means of the Faraday–Neumann law:

dλ

∫ (E + E ) ⋅ t dl = − dt ∫ B ⋅ n dS = − dt
d
elc = i m (1.91)
lc S

where S is the surface enclosed by lc and λ is the flux of the vector B linked
by the line lc. The subdivision of the EMF elc in its two components, induced
and motional EMF, depends by the adopted reference system only.

1.2.9 Poynting’s Vector

Poynting’s vector is defined as P = E × H and is a vector normal to the plane
defined by the vectors E and H. Its divergence corresponds to the power
density

divP = div(E × H)

= H ⋅ (curlE) − E ⋅ (curlH)

 ∂B   ∂D  (1.92)
= H⋅ −  − E ⋅ J +
 ∂t   ∂t 

∂B ∂D
= −H ⋅ − E⋅ − E⋅J
∂t ∂t

Then the integral of the divergence of the vector P throughout a volume

τ corresponds to the power that goes out the volume through the surface
enclosing the volume τ itself:

p=
∫ divP dτ
τ
(1.93)
 ∂B ∂D 
τ ∫
= − H⋅
 ∂t
+ E⋅
∂t 
dτ −
∫ (E ⋅ J)ddτ
τ

Assuming the parameters µ (between H and B) and ε (between D and E)

to be constant, it is possible to write

∂B 1 ∂H ⋅ B
H⋅ =
∂t 2 ∂t
(1.94)
∂D 1 ∂E ⋅ D
E⋅ =
∂t 2 ∂t
Outline of Electromagnetic Fields 19

H
n

H
S n
dS
θ Hn

τ Ht
dS t

FIGURE 1.3
Computation of the magnetic force by means of the Maxwell’s stress tensor.

Using Equation (1.94) and applying the divergence theorem to Equation

(1.93), it yields

p=
∫ τ
divP dτ

=

∫ P ⋅ n dS
S
(1.95)

∂  H⋅B E ⋅D
=−
∂t ∫ τ
 2 + 2  dτ − ∫τ
E ⋅ J dτ

1.2.10 Maxwell’s Stress Tensor

Maxwell’s stress tensor allows a rapid computation of the electromagnetic
forces acting on an object that is posed within an electromagnetic field. At
first, a suitable surface enclosing the object is selected; then the force is
computed as the integral over this surface of quantities obtained directly
from the potential that describes the field. It could be a scalar potential or a
vector potential depending on the study. In the following, let us refer to the
magnetic field, since it is more interesting in the applications that are dealt
with in this book.
Let a body occupy the volume τ. We suppose to know the value of the
vector field H over a surface S containing the volume τ and posed in a
material characterized by the permeability µo, as shown in Figure 1.3(a). In
any infinitesimal part dS of the surface S, Maxwell’s stress tensor is given by

µo 2
dF = − H ndS + µ o (H ⋅ ndS)H (1.96)
2

where n is the unity vector normal to the surface dS. By expressing the vector
H as the sum of its components tangential and normal to the surface dS, as
shown in Figure 1.3(b), as
20 Electrical Machine Analysis Using Finite Elements

H = Ht t + Hnn (1.97)

then the force is given by

dF = −
µo
2
( ) (
H t 2 + H n 2 dS t + µ o  H t t + H n n ⋅ n  H t t + H n n dS ) ( )
(1.98)
( ) µ
(
= µ oH t H n dSt + o H t 2 − H n 2 dSn
2
)
It is possible to identify the two components of the force along the two
preferential directions, as

(
dFt = µ oH t H n dS )
(1.99)
µ
dFn = o H n 2 − H t 2 dS
2
( )
The amplitude of the force is computed as

dF = (µ H H dS )
o t n
2 µ
( 
+  o H n 2 − H t 2 dS 
 2 
)
≈
1
2
(
µ o H t 2 + H n 2 dS ) (1.100)

1
≈ µ oH 2dS
2

The angle α between the force vector and the unity vector n normal to the
infinitesimal surface dS is computed as

Ht Hn
tan α =
1
2
(
Hn 2 − Ht 2 )
Ht / Hn
=2 (1.101)
( )
2
1 − Ht / Hn

tan θ
=2 = tan(2θ)
1 − tan 2 θ
Outline of Electromagnetic Fields 21

TABLE 1.2
Particular Cases of the Computation of the Magnetic
Force by Means of the Maxwell’s Stress Tensor
F

H Ht = 0 dFt = 0
Hn ≠ 0 dFn ≠ 0 (>0)

Hn = Ht
dFt ≠ 0
Ht = H n ≠ 0
F dFn = 0
Ht t

H
Ht ≠ 0 dFt = 0
F
Hn = 0 dFn ≠ 0 (<0)

where tanθ = Ht/Hn, that expresses the angle between the vector H and the
unity vector n. One can notice that the angle α results two times the angle
θ, i.e., α = 2θ.
Particular cases are shown in Table 1.2.

1.3 Fundamental Equations Summary

The main vector describing the electric and magnetic fields are

B(P,t) magnetic flux density

D(P,t) electric displacement
E(P,t) Maxwell’s electric field, sum of the Coulomb and induced field,
E(P,t) = Ec(P,t) + Ei(P,t)
H(P,t) magnetic field strength
J(P,t) current density

where P is the point where the vector is considered, while t is the time when
the vector is considered. In other words, P indicates the space dependence,
while t indicates the time dependence.
22 Electrical Machine Analysis Using Finite Elements

1.3.1 Maxwell’s Equations

The interactions among the vector fields are described by Maxwell’s equations

∂D(P,t)
curlH(P,t) = J(P,t) + (1.102)
∂t

∂B(P,t)
curlE(P,t) = − (1.103)
∂t

divB(P,t) = 0 (1.104)

divD(P,t) = ρ(P,t) (1.105)

Of course, in stationary field problems, the time-derivative of the vectors

D and B are null.

1.3.2 Constitutive Relationships and Continuity Equation

The constitutive relationships are

B(P,t) = µ(P) H(P,t)

D(P,t) = ε(P) E(P,t) (1.106)

J(P,t) = σ(P) E(P,t)

and the continuity equation is

∂ρ(P,t)
divJ(P,t) = − (1.107)
∂t

In general the media are not isotropic; then the parameters depending on
the material are of tensorial nature:
magnetic permeability electric permittivity electric conductivity

µ x 0 0 εx 0 0 σx 0 0
     
µ= 0 µy 0 ε=0 εy 0 σ= 0 σy 0
0 µ z  0 ε z  0 σ z 
 0  0  0
(1.108)
In general µi, εi, and σi are not constant along the generic i-th direction. In
fact they can be a function of the position (nonhomogeneous media) and/or
a function of the magnetic field H and the electric field E (nonlinear media).
Outline of Electromagnetic Fields 23

1.3.3 Laplace, Poisson, Helmholtz Equations

Let us refer to a magnetostatic field, described by Equation (1.65), Equation
(1.70), and Equation (1.72), where the time-derivative of the vector D is null.
By assuming the Coulomb’s position for the vector magnetic potential A,
Equation (1.68), the field problem is described by the quasi-harmonic equation

1 
curl  curlA  = J (1.109)
µ 

If the materials are homogeneous, the permeability µ is constant, then

Equation (1.109) is reduced to the Poisson equation, given by

∇ 2 A = −µJ (1.110)

If the current density field J is null in the considered domain, the problem
is described by the Laplace equation [Equation (1.52)], which is

∇2A = 0 (1.111)

When the magnetic field and the current density field are varying with
the time, they are mutually coupled. Let us suppose that such a variation is
sinusoidal with the time; then the field quantities can be expressed using
the symbolic notation, pointed out in Equation (1.54). A uniform medium is
considered, so that a constant electric conductivity σ and a constant magnetic
permeability µ are obtained. Finally, the time-derivative of the displacement
field D could be neglected in comparison with the other fields. Thus the
electric field E is the unique specific electric force. Hence the current density
J is obtained as the sum of the source current density Js and the induced
current density Ji, given by


∂A
J i = σE i = − σ 
= − jωσA (1.112)
∂t

that, substituted in Equation (1.110) gives rise to the Helmholtz equation,

combining the space variation of the vector A with its time variation, as

 − jωσA
∇2A  = −µJ (1.113)
s
24 Electrical Machine Analysis Using Finite Elements

References
1. J.A. Stratton, Electromagnetic Theory, McGraw-Hill, New York, 1941.
2. M. Abramowitz and I.A. Stegun, Handbook of Mathematical Functions, Dover
Publications, Inc., New York, 1965, pp. 374–376.
3. D.R. Corson and P. Lorrain, Introduction to Electromagnetic Field and Waves, W.H.
Freeman & Co., San Francisco–London, 1970.
4. J.D. Kraus and K.R. Carver, Electromagnetics, McGraw-Hill, New York, 1973.
5. G. Someda, Elementi di Elettrotecnica Generale, Patron Editore, Bologna, 1977 (in
Italian).
6. P. Lorrain and D.R. Corson, Electromagnetism. Principles and Applications, San
Francisco, 1978.
7. M.A. Plonus, Applied Electromagnetics, McGraw-Hill, New York, 1978.
8. R.E. Collin, Field Theory of Guided Waves, IEEE Press, New York, 1991.
9. D.K. Cheng, Fundamentals of Engineering Electromagnetics, Addison-Wesley,
Reading, MA, 1993.
10. M. Guarnieri and G. Malesani, Elettromagnetismo Stazionario e Quasi-Stazionario,
Edizioni Progetto, Padova, 1999 (in Italian).
11. J.M. Jin, Electromagnetic Analysis and Design in Magnetic Resonance Imaging, CRC
Press, Boca Raton, FL, 1998.
2
Basic Principles of Finite Element Methods

Some basic concepts of finite element method are dealt with in this chapter
and in the next. In this chapter, the differences of the finite element method
compared to the classical methods of field problem analysis are highlighted.
The mathematical notions of the method are presented. In the next chapter
the construction of the system of equations for the solution is investigated,
and the finite element method is applied to some two-dimensional field
problems. The concepts here reported are mainly quoted from two excellent
books (see References 16 and 17).
These two chapters are useful for acquiring a complete understanding of
how the method works. However, they might be passed over by the reader
primarily interested in the applications of the finite element method.

2.1 Introduction
The requirement of more and more accuracy during the process of design
and analysis of the electrical machines fostered the spreading of numerical
models appropriate for computing electric and magnetic fields. These
numerical methods are essentially based on the determination of the distri-
bution of the electric and magnetic fields in the structures under study, based
on the solution of Maxwell’s equations. An analytical solution is barely
achieved, because of the complex geometrical machine structures and the
nonlinear characteristic of the materials. Then, in most cases only a numerical
solution is possible.
The finite element method is a numerical technique that is suitable for this
purpose. It allows a field solution to be obtained, even with time-variable
fields and with materials that are nonhomogeneous, anisotropic, or non-
linear. Using the finite element method, the whole analysis domain is divided
into elementary subdomains, which are called finite elements, and the field
equations are applied to each of them.
This method was proposed in the 1940s, but it was firstly applied almost
ten years later in aeronautical design and in structural analysis. As years

25
26 Electrical Machine Analysis Using Finite Elements

went by, the finite element method was largely adopted in almost all physical
and mathematical problems. Today it is the most diffused method for the
solution of vector field problems.
The study of the field distributions, and in particular of electromagnetic
field problems, exhibits the following advantages. It allows a meticulous
local analysis to be carried out, highlighting dangerous field gradient, mag-
netic field strength, saturation, and so on. It allows a good estimation of the
performance of the electromagnetic devices under analysis (especially when
the classical methods of analysis give unsatisfactory results). Finally, it per-
mits one to reduce substantially the number of prototypes.
However, the method has some drawbacks, too. Because of its numerical
nature, the solution is necessarily approximate. Then, if the method is not
correctly applied, it might generate inaccurate results. Finally, since the com-
puted quantities are distributed in the space, the required computation time
is generally long.
In order to reduce the computation time, and to improve the analysis at
the same time, each periodicity and symmetry (both geometric and electro-
magnetic symmetry) of the structure is used. The resulting accuracy is influ-
enced by the dimension of the finite elements and by the uniformity of the
subdivision. To increase the accuracy, a fine subdivision of the structure is
carried out, adopting finite elements of smaller dimension. Nevertheless, an
excessive subdivision of the analysis domain causes an aggravation of the
computation time.

2.2 Field Problems with Boundary Conditions

Generally, a vector field problem is described by a differential equation,
defined in the domain D, as

L φ(P , t ) = f (P , t ) (2.1)

together with the boundary conditions. The latter constrain the fields along
the boundary Γ of the domain under analysis. In Equation (2.1) L is a
differential operator, φ is the unknown function to be determined, and f is
the forcing function. Equation (2.1) highlights that both φ and f are functions
of the position in the space, P(x,y,z), and of the time, t.

2.2.1 Meaning of the Differential Operator L

In general, L might be any differential operator. Commonly it represents a
linear operation, satisfying the property of additivity and the property of
product by a constant.
Basic Principles of Finite Element Methods 27

In the electromagnetic problems, Equation (2.1) is given by the Poisson,

Laplace, or Helmholtz equation, in which φ is a scalar or a vector field. As
an example, in the case of an electrostatic problem, φ indicates the scalar
electric potential V, and its distribution is described by the Poisson equation.
The forcing function is the distribution of the free charge density f = ρ. Then,
Equation (2.1) is rewritten as

( )
−div ε grad V = ρ (2.2)

in which a nonhomogeneous medium is considered, so that the electric

permittivity ε could be nonconstant. The differential operator L is then
expressed by

(
L = −div ε grad ) (2.3)

2.2.2 Boundary Conditions

The field problem admits a solution not only if the differential equation that
describes its distribution is known in all the points of the domain D, but also
if the unknown function φ is given on the boundary Γ of the domain D itself.
In addition, it can be verified that, once the solution has been found, this
solution is unique (this is the unicity theorem).
The conditions that express the behavior of the function φ on Γ are called
constraint, or boundary, conditions. Among these conditions, one can assign
a Dirichlet’s condition, which is when a given value of φ is assigned on the
boundary Γ, or a Neumann’s condition, which is when a given value of the
derivative of φ normal to the boundary Γ is assigned. In addition, periodicity
conditions can be assigned, by imposing equal values of φ in different parts
of the boundary Γ.
If we let Γ1 be a portion of the boundary Γ, the Dirichlet’s condition can be

1. Homogeneous condition (boundary condition of the first type):

φ=0 on Γ 1 (2.4)

2. Nonhomogeneous condition:

φ = φf on Γ 1 (2.5)

If we let Γ2 be the remaining portion of the total boundary Γ, the Neumann’s

conditions can be
28 Electrical Machine Analysis Using Finite Elements

1. Homogeneous condition (boundary condition of the second type):

∂φ
=0 on Γ 2 (2.6)
∂n

2. Homogeneous condition (boundary condition of the third type):

∂φ
+ kφ = 0 on Γ 2 (2.7)
∂n

3. Nonhomogeneous condition:

∂φ
+ kφ = φg on Γ 2 (2.8)
∂n

2.3 Classical Method for the Field Problem Solution

Let the field problem be expressed by Equation (2.1) and by suitable bound-
ary conditions, as given in the previous section. Some methods for solving
the field problems are now illustrated. In particular, the classic residual
method (or Galerkin’s method), the classic variational method (or Rayleigh-
Ritz’s method), and the finite element method.
All these methods aim to define a function φ* that approximates the
unknown function φ as closely as possible. Such a function is commonly
expressed as a linear combination of basic functions, as

φ * (P , t ) = ∑ Φ ν (P , t )
j=1
j j (2.9)

where νj are interpolating functions (that are also called expansion functions
or base functions), while Φj are unknown coefficients that have to be deter-
mined during the computation process. Such a combination has to approx-
imate appropriately the exact solution, satisfying the differential operator
[Equation (2.1)] and the boundary conditions at the same time.
The first two methods, the classical residual method and the classical
variational method, take into account the whole analysis domain. The func-
tions νj are defined on the whole domain. Conversely, in the finite element
method the whole domain is divided in subdomains; then the function φ* is
a combination of functions νj that are defined in the subdomains. Conse-
quently, since the subdomains are of reduced dimensions, the interpolating
functions νj can be very simple.
Basic Principles of Finite Element Methods 29

Before illustrating the different procedures, let us introduce the inner prod-
uct between two functions φ and ϕ. Let us refer to the volume τ; then the
inner product is defined as

φ, ϕ =
∫ φϕ
dτ
τ
(2.10)

where the symbol ~ indicates the complex conjugate. This inner product is
a linear operation, since the properties of additivity and the product by a
constant are satisfied:

φ1 + φ2 , ϕ = φ1 , ϕ + φ2 , ϕ (2.11)

αφ, ϕ = α φ, ϕ (2.12)

The definition of inner product given by Equation (2.10) could be used to

verify the properties of the differential operator L. In particular, it should be
verified if the operator L is positive defined:

> 0 φ≠0
Lφ, ϕ =  (2.13)
= 0 φ=0

and if it possible to change the argument of the operator L within the

operation of the inner product:

Lφ, ϕ = φ, Lϕ (2.14)

2.4 The Classical Residual Method (Galerkin’s Method)

The classical residual method deals with the differential equation (2.1)
directly. It solves the field problem by reducing the residual of the differential
equation (2.1). It is based on the following assumption: the function φ* that
better approaches the exact solution φ corresponds to a residual

r = Lφ * − f (2.15)

equal to zero (or at least very low) in the whole analysis domain. Fixing
some weight functions wi, the residual method forces the integral of the
residuals, weighed by wi to be zero over the domain volume τD. This is to
force the following condition:
30 Electrical Machine Analysis Using Finite Elements

Ri =
∫τD
(
w i L φ * − f dτ = 0 ) (2.16)

There are different weighed residual methods. The best known and most
used method is Galerkin’s method, where the weight functions wi are chosen
equal to the interpolating function νi, i.e.,

w i = νi i = 1, 2 , 3, … , N (2.17)

Generally this choice yields to a more accurate solution.

Then, using the approximation of Equation (2.9), Equation (2.16) becomes

 N 
Ri =
∫
τD
νi L 
 ∑j=1
Φ j ν j  − νi f dτ

i = 1, 2 , 3, … , N (2.18)

This equation yields to a system of equations that can be expressed as

 SS  φ  =  T  (2.19)

where [φ] is the column vector of the unknown coefficients Φi. [SS] is a matrix
vector that depends on the interpolating functions whose elements are given
by

∫ ( ν Lν + ν Lν ) dτ
1
s ij = i j j i (2.20)
2 τD

If the operator L satisfies the property (2.14), the matrix vector [SS] is
symmetrical, and its elements are

s ij =
∫τD
νi Lν j dτ (2.21)

In this case, the system of equations that is obtained by Galerkin’s method

is the same of that obtained by the variational method. At last, [T] in Equation
(2.19) is the column vector whose elements depend on the forcing function f.
They are given by

ti =
∫ τD
νi f dτ (2.22)
Basic Principles of Finite Element Methods 31

2.5 The Classical Variational Method

(Rayleigh-Ritz’s Method)
The variational method (also known as Rayleigh-Ritz’s method or simply as
Ritz’s method) solves the field problem by means of an integral approach.
Starting from the differential equation (2.1), a suitable functional is built, so
that its minimum corresponds to the solution of the field problem, which is
when the field equation (2.1) and the boundary conditions have been
matched. This functional is just called variational. The process of seeking
the solution of the differential problem becomes then a process of seeking
the minimum of the functional.
Then, since the minimum of the functional with respect to the function φ
corresponds to the solution of the Equation (2.1), it can be expressed as

1 1 1
F(φ) = Lφ, φ − φ, f − f , φ (2.23)
2 2 2

The function φ is substituted by the function φ* given by Equation (2.9),

where νj are again defined in the whole domain D. By substituting Equation
(2.9) in Equation (2.23), and posing to zero the derivatives of the variational
F with respect to the unknown coefficients Φj, i.e.,

δF
=0 i = 1, 2 , 3, … , N (2.24)
δΦ i

a system of linear equations, similar to that of Equation (2.19), is obtained.

The variational method requires that the property (2.14) is valid. As a con-
sequence the matrix vector [SS] is symmetrical and the system is identical
to that obtained by the residual Galerkin’s method.

2.5.1 A Field Problem Solution by Means of the Variational Method

Let us consider a field problem described by the differential equation (2.1),
with homogeneous boundary conditions, i.e., of the kind given in Equation
(2.4), Equation (2.6), or Equation (2.7). Then, the operation of the inner
product is defined as in Equation (2.10). The differential operator is assumed
to satisfy both the properties [Equation (2.13) and Equation (2.14)]. The aim
is to demonstrate that the solution of Equation (2.1) corresponds to the
minimum of the functional defined in Equation (2.23), i.e., to δF = 0 with
δ(δF) > 0.
32 Electrical Machine Analysis Using Finite Elements

It is

F + δF = F(φ + δφ)
(2.25)
=
1
2
( ) 1
L φ + δφ , φ + δφ − φ + δφ, f −
2
1
2
f , φ + δφ

Thanks to the additive property, neglecting the infinitesimal of higher

order, the following is obtained:

1 1 1 1
δF = Lφ, δφ + Lδφ, φ − δφ, f − f , δφ (2.26)
2 2 2 2

Since it has been assumed that L satisfies Equation (2.14), it is

1 1
δF = δφ, Lφ − f + Lφ − f , δφ (2.27)
2 2

Thus, from the definition of the inner product (2.10), the following results:

(
δF = Re δφ, Lφ − f ) (2.28)

where Re means the real part. By imposing the stationary condition δF = 0,

it is

(
Re δφ, Lφ − f = 0 ) (2.29)

Since δφ is an arbitrary variation of the function φ, one can immediately

conclude that φ has to satisfy the condition (2.1), in order that Equation (2.29)
is satisfied for any δφ. Then, to verify that it is a point of minimum, it is
found that

( ) ( ) ()
δ δF = δF φ + δφ − δF φ = Re δφ, Lδφ > 0 ( ) (2.30)

since it has been assumed that L is positive defined, as in Equation (2.13).

It is worth observing that the property (2.14) is essential in order that the
stationary point of the functional corresponds to the solution of the field
problem. Conversely, to satisfy the condition (2.13) means that this stationary
point corresponds to a minimum. Such a condition can be removed, since
the aim is to find a solution for Equation (2.1), no matter whether this solution
corresponds to a minimum, to a maximum, or to a flex point of the functional.
Basic Principles of Finite Element Methods 33

2.5.2 Definition of the Modified Variationals

When the field problem is described by Equation (2.1) and by nonhomoge-
neous boundary conditions, i.e., of the kind (2.5) and (2.8), it is essential to
modify the definition of the functional F given in Equation (2.23) in order
to apply again the method described above. In fact, in this case, the differ-
ential operator does not match the property (2.14).
In order to remedy this drawback, a new unknown function φ′ = φ – ψ is
introduced instead of φ, where ψ is any function that satisfies the prefixed
nonhomogeneous conditions. Therefore, the function φ′ satisfies the homo-
geneous conditions, and then the differential operator L satisfies the property
(2.14). Substituting φ′ instead of φ and f′ = f – Lψ instead of f in the definition
of the functional (2.23), it results in

1 1 1
F(φ′) = Lφ′ , φ′ − φ′ , f ′ − f ′ , φ′ (2.31)
2 2 2

From Equation (2.31) the modified functional is defined, taking into

account the nonhomogeneous boundary conditions, as

1 1 1 1 1
F(φ) = Lφ, φ − Lφ, ψ + φ ,Lψ − φ, f − f , φ (2.32)
2 2 2 2 2

In this equation, the second and the third term in the second member,
containing the function ψ, can be changed in integrals along the domain
boundary, by means of the Stokes theorem and the divergence theorem. Thus,
the function ψ disappears when the fixed boundary conditions are applied.
Further modifications can be applied to both the definition of the func-
tional and the definition of the inner product, which allow the method to be
generalized. In particular, it is possible to adapt the method to complex
differential operators (used in computation of media with losses) or to oper-
ators that do not satisfy the property (2.13) and (2.14). The reader who wants
to know more about these techniques may refer to the tests in the References.

2.5.3 Natural Boundary Conditions

In the variational method, the boundary conditions may be distinguished
between natural and essential conditions. The natural boundary conditions
are those that are automatically satisfied in the stationary point of the func-
tional. The essential boundary conditions have to be added to the definition
of the functional in order to obtain the exact solution of the field problem.
With the common definition of the functional, the Neumann conditions,
which are applied to the derivative of the function φ, are natural conditions,
i.e., are directly satisfied by the obtained solution. Conversely, the Dirichlet
conditions, which are applied directly to the value of the function φ, are
essential conditions, i.e., they have to be forced in the field solution.
34 Electrical Machine Analysis Using Finite Elements

2.6 The Finite Element Method

The finite element method is essentially based on the subdivision of the
whole domain in a fixed number of subdomains. Despite of the classical
methods described above, where the interpolating functions νi are defined
on the whole domain D, in the finite element method they are defined only
on each subdomain. It follows that, because of the small dimension of these
subdomains, the function φ is approximated by simple interpolating func-
tions whose coefficients are the unknown quantities. The solution of the field
problem is obtained when these unknown coefficients are found.
The finite element analysis is organized in the following steps:

1. Partition of the domain: The domain is divided in subdomains; they

are characterized by reduced dimensions.
2. Choice of the interpolating functions: The functions νi are chosen. As
said earlier, with the small dimension of the subdomains, these
functions can be very simple.
3. Formulation of the system to resolve the field problem: The system of
equations, representing the field solution, is developed indifferently
by means of Galerkin’s method or the Rayleigh-Ritz method.
4. Solution of the problem: The solution is obtained by solving the result-
ing system of equations.

2.6.1 Partition of the Domain

The first step of the finite element method is to divide the domain. The whole
domain D is subdivided in Nm elements Dm (m = 1, 2, 3, …, Nm). The way
to achieve such a subdivision greatly affects the solution accuracy. Moreover,
it influences the memory space required to the computer.
In one-dimensional problems, the domain is a curve and each subdomain
is a segment, as shown in Figure 2.1(a). The connection of the different
segments forms the original curve. In two-dimensional problems, the
domain is a surface and each subdomain is a polygon, usually a triangle or
a rectangle, as shown in Figure 2.1(b). In three-dimensional problems, the
domain is a volume and each subdomain is a tetrahedron, a triangular prism,
or a rectangular solid, as shown in Figure 2.1(c).

FIGURE 2.1
Elements for the partition of the domain.
Basic Principles of Finite Element Methods 35

2.6.2 Choice of the Interpolating Function

The second step consists of the choice of the interpolating function for
approximating the unknown function in each m-th element. If a first-order
polynomial is chosen, a linear interpolation is achieved. With a second-order
polynomial, a quadratic interpolation is achieved. Also a higher-order poly-
nomial can be chosen; however, although they yield to a higher accuracy in
the interpolation, they require a more complex formulation and, thus, are
barely adopted. Once the order of the polynomial is chosen, the unknown
solution in each m-th element is written as

φ m * (x , y , z , t ) = ∑Φ
j=1
mj νmj (x , y , z , t ) (2.33)

where n is the number of the nodes of the element, Φmj is the value of φ in
the j-th node of the m-th element. Finally, νmj is the interpolating function
referred to the j-th node of the m-th element. The highest order of the function
defines also the order of the element.

2.6.3 Formulation of the System

To solve the field problem, the values of Φmj have to be computed in the
nodes of each element. It is necessary to prepare a system of equations,
whose solution corresponds to the values Φmj. To develop this system, both
the variational method and the residual method may be adopted, in a similar
way as explained above, but applied to each element separately.
In the case of Galerkin’s method, Equation (2.16) is applied to each element.
The residual integral is put to zero. In the m-th element, the n integrals given
by

R im =
∫ ν ( Lφ
τ
i m )
* − fm dτ

 n  (2.34)

τ ∫
= νi L 
 ∑Φ
j=1
mj νmj  dτ −
 ∫ν f
τ
i m dτ i = 1, 2 , … , n

are posed equal to zero. A system of n equations with the n unknown Φmj
is obtained. Applying Equation (2.34) to all the Nm elements that form the
domain, and considering the relationships that link the adjacent elements, a
system of this kind is obtained:

 SS  φ  −  T  = 0 (2.35)

which is formed by Nn equations, with Nn unknown Φj.

36 Electrical Machine Analysis Using Finite Elements

In the case of the Rayleigh-Ritz method, the functional is given by

F(φ*) = ∑ F(φ* )
m=1
m

(2.36)
M
1 
= ∑ ∫
m=1
2
 τ
φm
* Lφm
* dτ −
1
2 ∫ τ
* dτ 
fm φm


that in matrix form results in

1 t t
F(φ*) = φ   SS  φ  − φ   T  (2.37)
2

The system is achieved by imposing the stationary condition to the func-

tional, which is to put to zero all the partial derivatives of Equation (2.37)
with respect to Φj, which is

∂F
=0 j = 1, 2 , 3, … , N (2.38)
∂Φ j

A system of Nn equations of the same form of Equation (2.35) is obtained.

However, using the variational method, the matrix [SS] is symmetrical, since
the condition (2.14) has to be satisfied. This is not verified in general with
Galerkin’s method.

2.6.4 Solution of the Problem

Once that the system (2.35) is developed, it is possible to compute the values
Φi in the Nn nodes of the domain. The system (2.35) is a system of equations
that may be solved by means of common numerical algorithms. Since the
matrix [SS] contains several zeros, suitable algorithms exist to resolve such
a system rapidly.

References
1. O.C. Zienkiewicz, The Finite Element Method in Engineering Science, McGraw
Hill, London, 1971.
2. H.C. Martin and G.F. Carey, Introduction to Finite Element Analysis: Theory and
Application, McGraw-Hill, New York, 1973.
3. D.H. Norrie and G. de Vries, The Finite Element Method: Fundamentals and Appli-
cations, Academic Press, New York, 1973.
Basic Principles of Finite Element Methods 37

4. M.V.K. Chari and P.P. Silvester, Finite Elements in Electrical and Magnetic Field
Problem, New York, John Wiley & Sons, 1980.
5. R.D. Cook, Concepts and Applications of Finite Element Analysis, Wiley, New York,
1981.
6. S.S. Rao, The Finite Element Method in Engineering, Pergamon Press, Oxford,
1982.
7. R.K. Livesley, Finite Element: an Introduction for Engineers, Cambridge University
Press, Cambridge, 1983.
8. P.P. Silvester and R.L. Ferrari, Finite Element Analysis and Design of Electromag-
netic Devices, Cambridge University Press, Cambridge, England, 1983.
9. J.N. Reddy, An Introduction to Finite Element Method, McGraw Hill, New York,
1984.
10. R. Wait and A.R. Mitchell, Finite Element Analysis and Applications, Wiley, New
York, 1985.
11. D.A. Lowther and P.P. Silvester, Computer Aided Design in Magnetics, Springer
Verlag, New York, 1986.
12. H. Grandin, Fundamentals of the Finite Element Method, Macmillan, New York,
1986.
13. D.S. Burnett, Finite Element Analysis: from Concepts to Applications, Addison-
Wesley Publishing Company, Reading, MA, 1987.
14. S.R.H. Hoole, Computer-Aided Analysis and Design of Electromagnetic Device,
Elsevier, New York, Amsterdam, London, 1989.
15. W.B. Bickford, A First Course in the Finite Element Method, Richard D. Irwin,
Homewood, 1990.
16. J.M. Jin, The Finite Element Method in Electromagnetics, John Wiley & Sons, New
York, 1992.
17. N. Ida and J.P.A. Bastos, Electromagnetics and Calculation of Fields, Springer-
Verlag, New York, 1992.
18. S.J. Salon, Finite Element Analysis of Electrical Machine, Kluwer Academic Pub-
lishers, Boston, MA, 1995.
19. A. Reece and T. Preston, Finite Element Method in Electric Power Engineering,
Oxford University Press, UK, 2000.
3
Applications of the Finite Element Method
to Two-Dimensional Fields

Although the three-dimensional codes are increasingly used, the majority of

the field problems concerning the analysis of electrical machines can be
carried out by a two-dimensional (2D) analysis. This provides several advan-
tages and results in an appreciable reduction of the computation time. The
phenomena omitted in the 2D analysis that are not negligible have to be
taken into account by means of suitable corrections to the obtained solution.
This chapter deals with the application of the finite element method to 2D
field problems. The latter will be described using Cartesian coordinates.

3.1 Introduction
3.1.1 Statement of the Two-Dimensional Field Problem
In 2D field problems, the considered domain is a surface S, and its boundary
is a curve. Let φ be the unknown function that is to be determined. It is a
scalar function of the space coordinates x and y, i.e., φ = φ(x,y). The time
dependence is omitted. Let f be the forcing function, which is also function
of x and y, and independent of the time. The 2D field problem is defined by
the differential equation of second order

∂  ∂φ  ∂  ∂φ 
( )
−div α ⋅ gradφ + βφ = − αx  − αy
∂x  ∂x  ∂y  ∂y 
  +βφ = f (3.1)

together with the boundary conditions that are imposed on the boundary Γ
of the domain. They are Dirichlet’s boundary conditions on the portion Γ1 of
the boundary:

φ = φf on Γ 1 (3.2)

39
40 Electrical Machine Analysis Using Finite Elements

together with Neumann’s boundary conditions on the remaining portion Γ2

of the boundary:

 ∂φ ∂φ 
 α x ∂x u x + α y ∂y u y  ⋅ n + kφ = φg on Γ 2 (3.3)

In Equation (3.1) and Equation (3.3), αx , αy, and β are known parameters
that are related to the physical property of the materials in the domain. In
Equation (3.2) and Equation (3.3), k, φf, and φg are known parameters that
are related to the physical property of the boundary. In particular, φf and φg
are functions describing the sources along the boundary curves.
The differential notation of Equation (3.1) is of a general nature, whose
Laplace’s, Poisson’s, and Helmholtz’s equations are the particular forms. It
corresponds to that given in Equation (2.1), where the differential operator
corresponds to

∂  ∂ ∂  ∂ 
L=− αx  − αy +β (3.4)
∂x  ∂x  ∂y  ∂y 
 

3.1.2 Application of the Variational Method

It is hereafter verified that the differential operator L, defined by Equation
(3.1) satisfies the property (2.14), if the following conditions are fulfilled:

1. The parameters αx and αy in Equation (3.1) are real, whether they

are numbers or functions.
2. The functions φ and ϕ satisfy the homogeneous Dirichlet’s condition
or the homogeneous Neumann’s condition.

From the inner product operation, defined in Equation (2.10), in a 2D

domain it is possible to write

Lφ, ϕ =
∫ −div (α ⋅ gradφ) + βφ  ϕ
dS
S
(3.5)
=
∫ −div (α ⋅ gradφ) ϕ
dS + β ∫ φϕ
dS
S S

Applying the second Green’s theorem (1.23) results in

 ∂φ ∂ϕ

Lφ, ϕ =
∫
S
( )
φ  −div α ⋅ gradϕ
 dS −
∫
Γ
α ⋅  ϕ

 ∂n
− φ  dl + β
∂n  ∫ φϕ
dS
S
(3.6)
Applications of the Finite Element Method to Two-Dimensional Fields 41

The second addendum in the second member is null, since it has been
assumed that both the functions φ and ϕ satisfy homogeneous conditions on
the boundary Γ. Then, it is

Lφ, ϕ =
∫ φ −div (α ⋅ gradϕ
) + βϕ
 dS
S
(3.7)
= φ, Lϕ

With the same assumptions, it is

Lφ, φ =
∫ −div (α ⋅ gradφ) + βφ  φ
dS
S
(3.8)
=
∫S
 (
 −div α ⋅ gradφ  φ
dS + β
 ) ∫ S
φ φ
dS

Applying the first Green’s theorem (1.22) results in

∂φ
Lφ, φ =
∫S
α ⋅ gradφ ⋅ gradφ
dS −
∫Γ
α ⋅ φ

∂n
dl + β
∫
S
φ φ
dS (3.9)

Substituting the homogeneous boundary conditions (2.4) and (2.7) results

in

∫ α gradφ ∫ ∫φ
2 2 2
Lφ, φ = dS + k φ dl + β dS
S Γ2 S
(3.10)

∫ α gradφ ∫
2 2 2
= +β φ dS + k φ dl
S Γ2

It is possible to see that, with positive α and k, the differential operator is

positive defined, i.e., the property (2.13) is satisfied: 〈Lφ,φ〉 > 0 if φ ≠ 0, and
〈Lφ,φ〉 = 0 if φ = 0.
The functional F(φ), defined in Equation (2.23), results in

1 1 1
F(φ) = Lφ, φ − φ, f − f , φ
2 2 2
(3.11)

∫ ∫ ∫( )
1 k 1
φf
+ fφ
dS
2 2 2
= α gradφ +β φ dS + φ dl −
2 S 2 Γ2 2 S

In case of nonhomogeneous boundary conditions, the technique illustrated

in Section 2.5.2 of Chapter 2 is applied. Always in this case the functional
becomes
42 Electrical Machine Analysis Using Finite Elements

∫ α gradφ ∫ ∫ (φf
+ fφ
) dS

1 2 2 k 2 1
F(φ) = + β φ dS + φ − φφ g dl − (3.12)
2 S Γ2 2 2 S

Given that φ and f are real functions, it is possible to obtain the following
formulation:

2 2
 ∂φ   ∂φ 
∫ ∫ ∫ f φ dS
1 k 2
F(φ) = α x   + α y   + βφ2 dS + φ − φφg dl − (3.13)
2 S  ∂x   ∂y  Γ2 2 S

It follows that the solution of the field problem, described by Equation

(3.1), Equation (3.2), and Equation (3.3), corresponds to that function φ that
satisfies the boundary conditions (3.2) and corresponds to a stationary point
of the functional (3.13).
The variation of F(φ) with respect to φ is

∂φ ∂δφ ∂φ ∂δφ
∫α
1
δF(φ) = x + αy + βφδφ dS +
2 S ∂x ∂x ∂y ∂y
(3.14)

∫ ( kφ − φ ) δφ dl − ∫
Γ2
g
S
f δφ dS

The two addenda of the first integral in the second member can be rewrit-
ten as

∂φ ∂δφ ∂  ∂φ  ∂  ∂φ 
αx = αx δφ − αx δφ
∂x ∂x ∂x  ∂x  ∂x  ∂x 
(3.15)
∂φ ∂δφ ∂  ∂φ  ∂  ∂φ 
αy = αy δφ − αy δφ
∂y ∂y ∂y  ∂y  ∂y  ∂y 
 

Let αx and αy be continuous in the whole domain. Then, by means the

divergence theorem (1.20) and the relationships (3.15), Equation (3.14) can
be rewritten as

 ∂  ∂φ  ∂  ∂φ  
δF(φ) =
∫ −  α x  −
S  ∂x 
 αy
∂x  ∂y  ∂y   + βφ − f  δφ dS +

(3.16)
 ∂φ ∂φ  
∫  α x
Γ  ∂x
ux + α y u y ⋅ n  δφ dl +
∂y   ∫Γ2
 kφ − φg  δφ dl

Since φ presents a fixed value on Γ1, equal to φf, as expressed in Equation

(3.2), then δφ becomes zero on the line Γ1. Consequently, the corresponding
integral on Γ1 becomes zero as well. The variation of the functional becomes
Applications of the Finite Element Method to Two-Dimensional Fields 43

 ∂  ∂φ  ∂  ∂φ  
δF(φ) =
∫ − ∂x  α
S
x − αy
∂x  ∂y  ∂y 
+ βφ − f  δφ dS +

(3.17)
 ∂φ ∂φ  
∫  α x
Γ2  ∂x
ux + α y u y  ⋅ n + kφ − φg  δφ dl
∂y  

The condition to have a stationary point of the functional corresponds to

force the variation (3.17) to be zero. This must be verified for each variation
of δφ, so that the surface integral and the line integral must be zero inde-
pendently. Therefore the results are

∂  ∂φ  ∂  ∂φ 
− αx  − αy + βφ − f = 0

∂x  ∂x  ∂y  ∂y 

(3.18)
 ∂φ ∂φ 
 α x ∂x u x + α y ∂y u y  ⋅ n + kφ − φg = 0

It is recognized that both the differential equation (3.1), which describes

the field problem, and the boundary condition (3.3), which forces the deriv-
ative of the function φ on the boundary, are satisfied. As stated in Chapter 2,
this condition is a natural condition, i.e., it is automatically satisfied by the
solution. Conversely the boundary condition (3.2) is an essential condition,
which must be forced explicitly in solving the problem.

3.2 Linear Interpolation of the Function φ

The 2D domain is subdivided in a finite and sufficiently high number of
elements. In the simplest case, they are elements of the triangular form, not
necessary equal, but not intersecting each other. Each vertex is called a node,
and all of them set up the mesh. Let us assume that the structure has been
divided in Nm finite elements, and the total number of node is Nn. Each of
them assumes the value Φi of the potential function φ.
Thanks to the small dimensions of the elements, the interpolating functions
νi(x,y) may be simple. In the following, a linear interpolation of the function
φ is assumed for each m-th triangular element, given by

φm (x , y) = a + bx + cy (3.19)

In particular, in the three nodes of the triangle, the three i-th values are
given by
44 Electrical Machine Analysis Using Finite Elements

Φ3 (x3, y3)
Φ2 (x2, y2)
Φ1 (x1, y1)
3(x3, y3)

Am 2 (x2, y2)

1 (x1, y1)

FIGURE 3.1
Linear interpolation of the potential function φm in the m-th triangular element.

 Φ1 = a + bx1 + cy1

Φ 2 = a + bx 2 + cy 2 (3.20)
 Φ = a + bx + cy
 3 3 3

From the knowledge of the value of the function in the nodes of each finite
element, i.e., Φ1, Φ2, and Φ3, by means of Equation (3.19), it is possible to
compute the potential function in any other point of the element. This is
represented in Figure 3.1.
If the three values of the potential are given in the three nodes of the
element, it is possible to solve the system (3.20) in the three unknowns a, b,
and c. At first, it is posed

1 x1 y1
1
Am = 1 x2 y2
2
1 x3 y3 (3.21)

=
1
2
( ) ( ) (
x 2 y 3 − x 3 y 2 + x 3 y1 − x1y 3 + x1y 2 − x 2 y1  )
that represents the area of the m-th triangular element, as shown in Figure
3.1. By Cramer’s rule, it results that

a=
1 
2A m 
( ) ( )
Φ1 x 2 y 3 − x 3 y 2 + Φ 2 x 3 y1 − x1y 3 + Φ 3 x1y 2 − x 2 y1  ( )
b=
1 
2A m 
( ) ( )
Φ1 y 2 − y 3 + Φ 2 y 3 − y1 + Φ 3 y1 − y 2  ( ) (3.22)

c=
1 
2A m 
( ) ( )
Φ1 x 3 − x 2 + Φ 2 x1 − x 3 + Φ 3 x 2 − x1  ( )
Applications of the Finite Element Method to Two-Dimensional Fields 45

It is

1
a=  Φ1p1 + Φ 2 p2 + Φ 3 p3 
2A m 

1
b=  Φ1q1 + Φ 2q2 + Φ 3q3  (3.23)
2A m 

1
c=  Φ1r1 + Φ 2 r2 + Φ 3r3 
2A m 

where

p1 = x 2 y 3 − x 3 y 2 q1 = y 2 − y 3 r1 = x 3 − x 2
p2 = x 3 y 1 − x 1 y 3 q2 = y 3 − y 1 r2 = x1 − x 3 (3.24)
p3 = x 1 y 2 − x 2 y 1 q3 = y 1 − y 2 r3 = x 2 − x1

Equation (3.19) expresses the unknown function φm(x,y), inside the m-th
triangular finite element. It can be rewritten as

 1 3   1 3   1 3 
φ m (x , y ) = 
 2 A m ∑
i =1
pi Φ i  + 
  2 A m ∑ i =1
qi Φ i  ⋅ x + 
  2 A m ∑ r Φ  ⋅ y
i =1
i i

3
(p + q x + r y) ⋅ Φ
= ∑
i =1
i

2A m
i i
i (3.25)

= ∑ ν (x , y ) ⋅ Φ
i =1
i i

It is worth noticing that is possible to define the function φm(x,y) in each

point of the triangle as linear combination of the values Φ1, Φ2, and Φ3 in

Φ3 = φm(x3, y3) Φ3 Φ3
Φ2 = φm(x2, y2) Φ2 Φ2
Φ1 = φm(x1, y1) Φ1 Φ1
3 (x3, y3) 3 (x3, y3) 3 (x3, y3)

Am Am Am
2 (x2, y2) 2 (x2, y2) 2 (x2, y2)

1 (x1, y1) 1 (x1, y1)

1 (x1, y1)

FIGURE 3.2
Representation of the linear interpolating functions of the function φm in the m-th triangular
element drawn in Figure 3.1.
46 Electrical Machine Analysis Using Finite Elements

the nodes of the triangle itself. It is then evident that, to get the field, it is
sufficient to compute the values of the unknown function φm(x,y) in the nodes
of each element that form the whole domain. In Equation (3.25), the inter-
polating functions νi(x,y), with i = 1, 2, 3, that interpolate the function φm(x,y)
have been highlighted. Each of them can be represented graphically, as
illustrated in Figure 3.2.

3.2.1 Computation on the Function φm(x,y)

Before going on, let us consider some operations on the function φm(x,y)
defined in Equation (3.19), which will be useful in the following.
If we let φm be a scalar function, its gradient is given by

gradφm = grad(a + bx + cy)

 ∂(a + bx + cy) ∂(a + bx + cy) 

= , , 0 (3.26)
 ∂x ∂y 
= ( b, c, 0)

By analogy, letting φm be a vector function, and with a z-axis component

only, as commonly occurs in 2D problems, its curl is given by

curlφm = curl  0, 0, (a + bx + cy) 

 ∂(a + bx + cy) ∂(a + bx + cy) 

= ,− , 0 (3.27)
 ∂y ∂x 
= (c, − b, 0)

The values (gradφm · gradφm) and (curlφm · curlφm), inside each triangular
element, are constant, since they do not depend on the coordinates x and y.
In fact, the results are

gradφm ⋅ gradφm = b2 + c 2 (3.28)

curlφm ⋅ curlφm = b2 + c 2 (3.29)

3.3 Application of the Variational Method

The field problem is solved by means of the integral formulation, computing
a functional F that is related to the field problem. It is expressed as a function
Applications of the Finite Element Method to Two-Dimensional Fields 47

of the Nn unknown values Φi of the potential φ. The field problem becomes

a Nn-dimensional vector of Φi that minimizes the functional and satisfies
Dirichlet’s boundary conditions (Neumann’s boundary conditions, if any,
are naturally satisfied, as seen in Section 3.1.2). In other words, it is necessary
to put to zero all the derivative of the functional F with respect to Φi in the
Nn nodes of the domain, that is

∂F
=0 with i = 1, 2 , 3, … , N n (3.30)
∂Φ i

Since the functional F is given by the sum of the functionals Fm associated

to all the Nm finite elements that form the domain

Nm

F= ∑F
m =1
m (3.31)

the solution of the field problem, expressed by Equation (3.30), can be

expressed as
Nm
∂F
∂Φ i
= ∑ ∂∂FΦ
m =1
m

i
=0 with i = 1, 2 , 3, … , N n (3.32)

3.3.1 Functional Referred to the m-th Element

The functional associated to the m-th finite element, say Fm, depends only
on the values Φi in the three nodes of the element. In fact, Fm depends on
the function φm in the element and the latter is a linear combination of the
potential in the nodes, as defined by Equation (3.25). Thus, all the derivatives
of Fm are null with respect to the unknown values Φi in the nodes that do
not belong to the m-th element.
For the sake of simplicity, the most general case of Helmholtz’s equation
(3.1) is disregarded, and we focus on Poisson’s equation (or Laplace’s equa-
tion in specific cases), that cover almost all of the problems about the elec-
trical machines. As a consequence, β = 0 in Equation (3.1) and in Equation
(3.4). In addition, on the boundary part Γ2 only the homogeneous Neumann’s
condition is considered, yielding to k = 0 and φg = 0 in Equation (3.3). Finally,
referring to Equation (3.1), it is assumed that in each finite element the forcing
term f is constant and equal to fm, and the values αx and αy are constant as well.
By means of these simplifications, the functional F of Equation (3.13),
referred to the m-th element, becomes

2 2
 ∂φ   ∂φ 
∫ ∫
1
Fm = α x  m  + α y  m  dS + fm φm dS (3.33)
2 Am  ∂x   ∂y  Am
48 Electrical Machine Analysis Using Finite Elements

With the interpolation of φm given in Equation (3.25), the partial derivatives

of φm are

3
∂φm
∂x
=
1
2A m ∑q Φ
i =1
i i

(3.34)
3
∂φm
∂y
=
1
2A m ∑r Φ
i =1
i i

so that

2 3 3
 ∂φm  1
 ∂x  = 4A 2
m
∑∑q q Φ Φ
i =1 j=1
i j i j

(3.35)
2 3 3
 ∂φm 
 ∂y 
=
1
4A m 2 ∑∑r rΦ Φ
i =1 j=1
i j i j

Both the terms in Equation (3.35) are constant inside the elements. They
can be put outside the integral of Equation (3.33). After some manipulations,
the functional Fm of the m-th element results in

3 3 3

Fm =
1
8A m ∑ ∑(
i =1 j=1
)
α xqiqj + α y ri rj Φ i Φ j + fm
Am
3 ∑Φ
i=1
i (3.36)

Using the vector notation, it is

 s11 s12 s13   Φ1   t1 

1     
Fm =  Φ1 Φ 2 Φ 3  s 21 s 22 s 23   Φ 2  +  Φ1 Φ 2 Φ 3   t 2 
2
s 31 s 32 s 33   Φ 3   t 3  (3.37)
  
1 t t
=  Φ123   S m   Φ123  +  Φ123   Tm 
2

where the terms of the matrix [Sm] and the vector [Tm] are

s ij =
1
4A m
(
α xqiqj + α y ri rj )
(3.38)
A
t i = fm m
3
Applications of the Finite Element Method to Two-Dimensional Fields 49

It is observed that the terms sij depend only on the geometry and the
material characteristic (expressed by the coefficients αx and αy). If the mate-
rial is linear, the terms sij do not depend on the value of the potentials in the
nodes (since αx and αy are constant with the field). By analogy, the terms ti
are only functions of the geometry of the element and of the forcing quantity
fm.
Since the research of the stationary point of the functional Fm is carried
out by posing to zero the derivative of Fm with respect to the values Φ1, Φ2,
and Φ3 in the three nodes of the m-th element. Deriving Fm in Equation (3.37)
with respect Φ1, Φ2, and Φ3 and equating the result to zero, the results are

t
 ∂Fm ∂Fm ∂Fm 
 ∂Φ , ∂Φ , ∂Φ  =  S m   Φ123  −  Tm  =  0  (3.39)
 1 2 3

Equation (3.39) shows that, for the generic m-th triangular element, a
system of three equations is obtained, with the three unknown potential
values Φ1, Φ2, and Φ3 in the nodes of the element. When this system of
equations is satisfied, the field problem solution is obtained in the m-th
element.

3.3.2 Functional Referred to the Whole Domain

The complete solution of the field problem consists of the solution of the
system (3.39) for each finite element that belongs to the domain. It is neces-
sary to pass from the local matrix [Sm] that refers to the single m-th element
(i.e., to the three values Φ1, Φ2, and Φ3 only) to the global matrix [SS] that
refers to the complete domain (i.e., to all the values Φi of the Nn nodes). This
operation is accomplished by combining all the functionals φm of the adjacent
elements.
As an example, let us consider two triangles that share the nodes 2 and 3,
where the function Φ assumes the values Φ2 and Φ3, as shown in Figure 3.3.

3 (x3, y3)

4 (x4, y4)
Fb
Fa

2 (x2, y2)
1 (x1, y1)

FIGURE 3.3
Adjacent elements with shared nodes 2 and 3.
50 Electrical Machine Analysis Using Finite Elements

The two triangles are characterized by the local functionals Fa and Fb, given
by

1 t t
Fa =  Φ123   Sa   Φ123  −  Φ123   Ta 
2
(3.40)
1 t t
Fb =  Φ 234   S b   Φ 234  −  Φ 234   Tb 
2

Then, the sum of the two functionals, Fa + Fb, is obtained as

1 t t
Fa + Fb =  Φ1234   Sab   Φ1234  −  Φ1234   Tab  = (3.41)
2

 Φ1 
t
 sa(11) sa(12 ) sa(13) 0   Φ   Φ  t  t a(1) 
    1 1
 
1  Φ2  sa( 21) sa( 22 ) + s b(11) sa( 23) + s b(12 ) s b(13)   Φ 2   Φ 2   t a( 2 ) + t b(1) 
s −
2  Φ3 
 a( 31) sa( 32 ) + s b( 21) sa( 33) + s b( 22 ) s b( 23)   Φ 3   Φ 3   t a( 3) + t b( 2 ) 
     
 Φ 4   0 s b( 31) s b( 32 ) s b( 33)   Φ 4   Φ 4   t b( 3) 


Proceeding in the same way for each triangular element with shared nodes,
the complete functional of the whole domain is obtained as

1 t t
F=  Φ   SS   Φ  −  Φ   T  (3.42)
2

where [Φ] is the column vector of all the Nn values of the function φ in all
the nodes of the domain. The dimension of the matrix [SS] is Nn × Nn. It is
called the “stiffness” matrix, due to the analogy with the structural analysis
to which the finite element method has been originally applied. Finally, [T]
is the column vector of the known terms, and its dimension is Nn.
The stiffness matrix is a sparse matrix, i.e., with several zeros. As observed
in Equation (3.38), in linear conditions, this matrix is built and is not modified
during the process of search of the solution. The stationary point of the
functional F (3.42) corresponds to

 SS   Φ  −  T  =  0  (3.43)

It is a linear system of Nn equations with Nn unknown Φi in the nodes of

the triangular finite elements. Thanks to the interpolation (3.25) valid in each
element, once the values Φi are computed, it is possible to go back to the
value of φ in each point of the domain.
Applications of the Finite Element Method to Two-Dimensional Fields 51

3.3.3 Assigning the Essential Boundary Conditions

The exact field solution has to satisfy the boundary conditions. Whereas
Neumann’s boundary conditions may be not imposed, since they are natural
conditions, it is necessary to impose Dirichlet’s boundary conditions, which
are essential conditions.
To impose Dirichlet’s boundary conditions, the stiffness matrix [SS] and
the column vector [T] of the known terms in Equation (3.43) must be mod-
ified. As an example, let us suppose that the potential value Φk has to be
assigned to the k-th node. In the stiffness matrix [SS], each term of the k-th
row is put to zero, except the term on the principal diagonal in which unity
is assigned. Then the value Φk is assigned to the k-th term of the column
vector [T].

3.3.4 The Problem’s Solution

Once the system is assembled, including the assignment of the essential
boundary conditions, it is still in the form (3.43). The system is solved by
one of the common numerical methods, among them:

1. The Gauss-Jordan direct method. This method is applied when the

physical properties of the materials are linear, which is when all
the parameters of the materials are constant.
2. The Newton-Rapson iterative method. This method is applied when
the materials are not linear, which is when the parameters of the
materials (αx and αy) are not constant, but they are functions of
the value of the magnitude of the magnetic fields. The common case
is the B-H curve of the ferromagnetic materials.

3.4 Simple Descriptions of Electromagnetic Fields

This section deals with some simple applications of the finite element method
to 2D electromagnetic fields, expressed in Cartesian coordinates. For all the
field problems, a convenient potential function φ is singled out and the
corresponding functional F is achieved. In particular, the various expressions
of the functional Fm will be reported, referred to as the m-th element.

3.4.1 An Electrostatic Field

Starting from Maxwell’s equations, reported in Chapter 1, the electrostatic
field problem is described by the relations divD = ρ, curlE = 0, and D = εE.
52 Electrical Machine Analysis Using Finite Elements

All the time derivatives are null and the current density J is null as well.
The unique specific electric force is Coulomb’s electric field, E = Ec. The
second equation highlights that the electric field is irrotational in the whole
domain, which is conservative. Thus, it is possible to define a scalar electric
potential V such as Ec = −gradV. From these equations, the field problem is
described by the following almost harmonic scalar equation:

divD = div(εE c ) = divε(− gradV) = ρ (3.44)

In case of a 2D field, using Cartesian coordinates, this equation becomes

∂ ∂V ∂ ∂V
εx + εy = −ρ (3.45)
∂x ∂x ∂y ∂y

If the medium is homogeneous and isotropic, i.e., the electric permittivity

ε is constant, Equation (3.45) is reduced to the scalar Poisson’s equation

∂2 V ∂2 V ρ
+ =− (3.46)
∂x 2 ∂y 2 ε

The functional related to the 2D electrostatic field, referred to the surface

S of the domain D, is given by

1 
F=
∫S
 2 εE c − ρV  dS
2
(3.47)

With reference to the m-th element, substituting the interpolation given in

Equation (3.19) in Equation (3.47) and using the relation (3.28), the functional
Fm is given by

∫ ∫
1
Fm = ε (gradV · gradV)dS − ρ V dS
2 Am Am

∫ ∫
1
= ε ( b2 + c 2 )dS − ρ (a + bx + cy) dS (3.48)
2 Am Am

1  x + x2 + x3 y + y2 + y3 
= εA m ( b2 + c 2 ) − ρA m  a + b 1 +c 1 
2  3 3

The coefficients a, b, and c are functions of the potential V1, V2, and V3 of
the nodes of the m-th triangular element, where the functional is computed.
Substituting the expressions (3.22), that have been obtained for a, b, and c
as a function of the potentials, into Equation (3.48), Fm can be written as
Applications of the Finite Element Method to Two-Dimensional Fields 53

1 t t
Fm =  V123   S m   V123  −  V123   Tm  (3.49)
2

where [V123] is the column vector of the potentials V1, V2, V3 of the three
nodes of the triangular element, and [Sm] is the stiffness matrix, with dimen-
sion 3 × 3. The generic component corresponding to the i-th row and j-th
column of [Sm] is given by

ε
s ij = (qiqj + ri rj ) (3.50)
4A m

Finally [Tm] is a column vector of known terms, which is [Tm] = [t1, t2, t3]t,
where

Am
ti = − ρ (3.51)
3

Of course, Equation (3.50) and Equation (3.51) coincide with those given
in Equation (3.38), once equating fm = −ρ and αx = αy = ε.

3.4.2 A Stationary Current Field

The equations describing the stationary current field are divJ = 0, rotE = 0,
and J = σE. All the time derivatives are null and the unique specific electric
force is Coulomb’s electric field, E = Ec. The study may be set out in two ways.

3.4.2.1 By Means of the Scalar Potential

Since the field is stationary, the electric field E is irrotational in the whole
domain; thus it is conservative. It is possible to define a scalar electric poten-
tial V such as E = −gradV. The field problem is described by the following,
almost harmonic scalar equation

divJ = div(σE) = divσ(− gradV) = 0 (3.52)

that, in a 2D field using Cartesian coordinates, becomes

∂ ∂V ∂ ∂V
σx + σy =0 (3.53)
∂x ∂x ∂y ∂y

Furthermore, if the electric conductivity σ is constant, Laplace’s equation

is achieved:

∂2 V ∂2 V
+ =0 (3.54)
∂x 2 ∂y 2
54 Electrical Machine Analysis Using Finite Elements

The functional Fm, referred to the surface Am of the m-th element, is

 1 2
Fm =
∫Am
 2 σE  dS

∫
1
= σ (gradV ⋅ gradV)dS (3.55)
2 Am

1
= σA m ( b2 + c 2 )
2

Since the coefficients a, b, and c are functions of the potentials V1, V2, and
V3 of the nodes of the triangular element, the functional Fm can be expressed
as

1 t
Fm =  V123   S m   V123  (3.56)
2

where [V123] = [V1, V2, V3]t, and the generic component of the matrix [Sm] is
given by

σ
s ij = (qiqj + ri rj ) (3.57)
4A m

3.4.2.2 By Means of the Vector Potential

Alternatively, in place of the scalar potential V, the stationary current field
can be studied by means of a suitable vector potential. Since the current
density vector J is solenoidal, it is possible to define an electric vector potential
N such as J = curlN, always being div(curlN) = 0. By handling these equa-
tions, the field problem is described by the almost harmonic vector equation

 J 1
curlE = curl   = curl curlN = 0
 σ σ
( ) (3.58)

The coordinate system is chosen so that the electric vector potential N has
only a z-axis component, which is N = [0, 0, Nz]. In this way, there are only
x-axis and y-axis components of the current density vector J:

 ∂N ∂Nz 
J = curl  0, 0, Nz  =  z , − , 0 (3.59)
 ∂y ∂x 
Applications of the Finite Element Method to Two-Dimensional Fields 55

With a constant conductivity σ Laplace’s equation is achieved:

∂2 Nz ∂2 Nz
+ =0 (3.60)
∂x 2 ∂y 2

The functional of the m-th triangular element is

 1 2
Fm =
∫ 
Am  2σ
J  dS


∫
1
= (curlN ⋅ curlN)dS (3.61)
2σ Am

1
= A m ( b2 + c 2 )
2σ

Since a, b, and c are functions of the potential values N1, N2, and N3 of the
three nodes of the triangular element, the functional Fm becomes

1 t
Fm =  N123   S m   N123  (3.62)
2

where [N123] = [N1, N2, N3]t, and [Sm] is the stiffness matrix. Its generic
component is

1
s ij = (qiqj + ri rj ) (3.63)
4σA m

3.4.3 A Magnetostatic Field

Starting from Maxwell’s equations given in Chapter 1, the magnetostatic
field problem is described by the equations divB = 0, curlH = J, and B = µH.
Since div(curlB) = 0, a magnetic vector potential A is defined such as B =
curlA. The field problem is described by the almost harmonic vector equation

B 1
curlH = curl = curl curlA = J (3.64)
µ µ

In the 2D field, using Cartesian coordinates, the current density vector J has
only a component normal to the plane (x, y), which is only a z-axis compo-
nent. Consequently, the magnetic vector potential A has only a z-axis com-
ponent, i.e., the vector A is parallel to the vector J. Then these vectors can
be expressed as J = [0, 0, Jz] and as A = [0, 0, Az]. Equation (3.64) becomes
56 Electrical Machine Analysis Using Finite Elements

∂ 1 ∂A z ∂ 1 ∂A z
+ = − Jz (3.65)
∂x µ x ∂x ∂y µ y ∂y

The magnetic flux-density vector B has components only on the plane

(x, y), because

 ∂A ∂A z 
B = curl  0, 0, A z  =  z , − , 0 (3.66)
 ∂y ∂x 

In homogeneous media, with magnetic permeability µx = µy = µ, Equation

(3.66) is reduced to a Poisson’s equation:

∂2 A z ∂2 A z
+ = − µJ z (3.67)
∂x 2 ∂y 2

By means of the linear interpolation of Equation (3.19), the functional

referred to the m-th finite element is given by

1 
Fm =
∫Am

 2
B ⋅ H − J ⋅ A  dS


∫ ∫
1
= (curlA ⋅ curlA)dS − J z A zdS (3.68)
2µ Am Am

1  x + x2 + x3 y + y2 + y3 
= A m ( b2 + c 2 ) − J z A m  a + b 1 +c 1 
2µ  3 3

Since a, b, and c are functions of the potentials A1, A2, and A3 of the three
nodes of the triangular element, Fm becomes

1 t t
Fm =  A 123   S m   A 123  −  A 123   Tm  (3.69)
2

where [A123] = [A1, A2, A3]t, and [Sm] is the stiffness matrix, whose generic
component is

1
s ij = (qiqj + ri rj ) (3.70)
4µA m

The column vector [Tm] is a function of the current density, which is

imposed as the source. Its generic component is

Am
ti = − Jz (3.71)
3
Applications of the Finite Element Method to Two-Dimensional Fields 57

3.4.4 A Magnetostatic Field without Current

Let us consider a magnetostatic field problem characterized by a null current
density vector J = 0 in the whole domain. In such a case, the equations
describing the field are divB = 0, curlH = 0, and B = µH. Since the magnetic
field strength H is irrotational, a magnetic scalar potential Ψ can be defined
such as H = −gradΨ, as curl(gradΨ) = 0. From these equations, the field
problem is described by the almost harmonic scalar equation

divB = div(µH) = divµ(− gradΨ ) = 0 (3.72)

In the 2D field, described using the Cartesian coordinates, and with a

constant magnetic permeability µ, Equation (3.72) becomes Laplace’s equa-
tion given by

∂2 Ψ ∂2 Ψ
+ =0 (3.73)
∂x 2 ∂y 2

The functional Fm of the m-th triangular element is given by

1 
Fm =
∫Am
 2 B ⋅ H dS

∫
1
= µ (gradΨ ⋅ gradΨ )dS (3.74)
2 Am

1
= µA m ( b2 + c 2 )
2

Since a, b, and c depend on the potentials Ψ1, Ψ2, and Ψ3 of the three nodes
of the triangular element, Equation (3.74) becomes

1 t
Fm =  Ψ 123   S m   Ψ 123  (3.75)
2

where [Ψ123]=[Ψ1, Ψ2, Ψ3]t, and [Sm] is the stiffness matrix whose generic
component corresponds to

µ
s ij = (qiqj + ri rj ) (3.76)
4A m

3.4.5 A Magnetic Field with Permanent Magnets

The equations that describe the field are curlH = 0, divB = 0, and B = Bres +
µH, where Bres is the residual flux density vector of the permanent magnet.
58 Electrical Machine Analysis Using Finite Elements

However, since div(curlB) = 0 and divB = 0, it is possible to define a magnetic

vector potential A so that B = curlA. From these equations, it results that

 B − Bres   B B 
curlH = curl  = curl   − curl  res 
 µ   µ  µ 
(3.77)
B 
1
( )
= curl curlA − curl  res  = 0
µ  µ 

The vector M = curl(Bres/µ) is the magnetization vector. The differential

equation that describes the field problem is

B 
curl
1
µ
(  µ 
)
curlA = curl  res  (3.78)

The functional that is associated with the field, referred to as the area Am
of the m-th triangular element, is given by

∫
1
( )
2
Fm = B − B res dS
A m 2µ

∫
1
= (B 2 − 2BB res + B res 2 )dS (3.79)
2µ Am

∫
1
= (curlA ⋅ curlA − 2B res curlA + B res 2 )dS
2µ Am

Substituting the linear interpolation given in Equation (3.19) in A = Azuz,

the functional Fm (3.79) becomes

1 t t
Fm =  A 123   S m   A 123  −  A 123   Tm  (3.80)
2

where [A123] = [A1, A2, A3]t, and the generic component of the stiffness matrix
[Sm] is

1
s ij = (qiqj + ri rj ) (3.81)
4µA m

The generic component of the column vector [Tm] is a function of the

residual flux density, and it is expressed as

ti =
1
µ
(
B res , yqi − B res ,x ri ) (3.82)
Applications of the Finite Element Method to Two-Dimensional Fields 59

3.5 Appendix: Integration in Triangular Elements

Let us consider a triangular element defined in the plane (x, y) by the three
points 1:(x1,y1), 2:(x2,y2), and 3:(x3,y3), so that its center of gravity is in the
origin of the coordinate system, i.e.,

x1 + x 2 + x 3 y1 + y 2 + y 3
= =0
3 3

Some surface integrals on the triangle area Am are reported as follows:

∫Am
x dx dy = 0

∫Am
y dx dy = 0

1 x1 y1

∫
1
dx dy = 1 x2 y2 = Am
Am 2
1 x3 y3

∫Am
x 2 dx dy =
Am
12
(
x 12 + x 2 2 + x 3 2 )
∫Am
y 2 dx dy =
Am
12
(
y 12 + y 2 2 + y 3 2 )
∫Am
xy dx dy =
Am
12
(
x1y1 + x 2 y 2 + x 3 y 3 )

References
See references in Chapter 2.
4
The Analysis Procedure Using the Finite
Element Method

4.1 Introduction
This chapter deals with the analysis of the electrical machines by means of the
finite element method. Instead of a generic problem, the magnetic field prob-
lem is investigated. Such a problem is the most common in the analysis of
electrical machines, including transformers, rotating machines, and actuators.
In any case, the study of the magnetic field alone is not sufficient for a
complete analysis of the machine. For instance, the evaluation of the electric
field between the turns of the coils requires an electrostatic field analysis;
the computation of the Joule losses in conductors with restricted sections,
especially in the terminals, requires a current field analysis. However, the
extension of the finite element analysis to vector fields of a nature different
from the magnetic field is left to the reader.

4.2 Reduction of the Field Problem

to a Two-Dimensional Problem
The electrical machine to be analyzed is obviously a three-dimensional (3D)
structure. However, a 3D analysis is required to subdivide the whole struc-
ture by means of 3D finite elements (as seen in Chapter 2, Figure 2.1(c)) and
requires heavy processing and long computation time. That is why, if pos-
sible, the field problem should be reduced to a 2D problem. This is carried
out by individuating any symmetry of the machine. Such a symmetry can
be of two kinds:

61
62 Electrical Machine Analysis Using Finite Elements

z
y
z ϑ
r

(a) (b)

FIGURE 4.1
Planar symmetry (a) and axial symmetry (b).

1. Planar symmetry (or xy symmetry): The magnetic phenomena are

assumed to be identical on each plane (x, y) normal to the z-axis, as
shown in Figure 4.1(a). The effects of the ending edges of the
machines are therefore neglected.
2. Axial symmetry (or rz symmetry): It is assumed that the magnetic
phenomena are repeated identical on each semiplane (r, z) obtained
as a rotation around the z-axis, which is called the symmetry axis,
as shown in Figure 4.1(b).

In this chapter, a field problem with planar symmetry is considered. The

study will be carried out in the plane (x, y) perpendicular to the z-axis, as
in Figure 4.1(a). In the next chapter, the finite element analysis of an axial-
symmetric structure will be presented and discussed.
As described in Chapter 3, the study of a 2D magnetostatic field problem
with planar symmetry is simplified as follows:

1. The current density vector J has a z-axis component only, which is

J = [0, 0, Jz].
2. The magnetic vector potential A is parallel to the vector J, thus it
has a z-axis component only, i.e., A = [0, 0, Az]. As regards the
divergence of the magnetic vector potential, Coulomb’s position
(1.68) is generally adopted, which is divA = 0.
3. The flux density vector B has components only on the plane (x, y),
as stated by Equation (3.66).
4. With a constant magnetic permeability µ, the field problem is
described by Poisson’s equation (3.67), which is reported again:

∂2 A z ∂2 A z
+ = − µJ z (4.1)
∂x 2 ∂y 2
The Analysis Procedure Using the Finite Element Method 63

4.3 Boundary Conditions

The magnetic vector potential Az can be determined from Poisson’s equation
(4.1), in each point of the domain D, once the current density Jz is fixed and
the value of Az is known on the boundary Γ of the domain itself. Going into
details, in a part of the boundary, say Γ1, the value of Az is assigned
(Dirichlet’s condition), and in the remanent part of the boundary, say Γ2, the
value of the derivative of Az normal to the boundary line is assigned (Neu-
mann’s condition). See Chapter 2, Section 2.2.
Therefore, the assignment of the boundary conditions is a thorough oper-
ation, which assumes a fundamental importance in the solution of the field
problem. The choice of the boundary conditions not only influences the final
solution, but also can further reduce the domain under study.
The boundary conditions are gathered in a group hereafter.

4.3.1 Dirichlet’s Condition

This condition corresponds to assign the value of the magnetic vector poten-
tial Az on a given part of the boundary. Generally, the value that is assigned
is constant, so that the boundary line assumes the same value of magnetic
vector potential Az. It follows that the flux lines are tangential to the bound-
ary itself, and no flux line crosses that boundary. It is common to assign the
homogeneous Dirichlet’s condition (2.4), fixing the magnetic vector potential
Az = 0 along all or part of the boundary.
Such a condition is equivalent to considering an external material with
null magnetic permeability, which is a magnetic insulating material just
outside the domain.
As an example, let us refer to a synchronous generator with salient poles.
By neglecting the edge effects (essentially due to the end windings), a planar
symmetry is recognized, so that only the section of the machine represented
in Figure 4.2 can be analyzed. Since the flux lines are confined within the

Homogeneous
Dirichlet
condition

FIGURE 4.2
Homogeneous Dirichlet’s condition along the external circumference of a synchronous generator.
64 Electrical Machine Analysis Using Finite Elements

stator back iron, the condition Az = 0 is assigned along all the external
circumference of the synchronous machine. Figure 4.2. highlights this homo-
geneous Dirichlet’s condition, by means of a bold line.

4.3.2 Neumann’s Condition

This condition corresponds to imposing a given value to the derivative of
Az normal to the boundary, so that the flux lines have a given incidence
angle with the boundary. In the case of homogeneous Neumann’s condition
(4.6), the flux lines are forced to be perpendicular to the boundary line. The
condition (4.6) applied to a magnetic field constrains the flux density vector
B to have components only normal to the boundary line.
This condition is equivalent to having an external material with infinite
magnetic permeability just outside the domain.
As an example, let us consider a single-phase reactance, with legs of
rectangular form. Neglecting the edge effects, a planar symmetry is consid-
ered, so that a section of the structure can be simply analyzed, as shown in
Figure 4.3(a). It is easy to notice that the structure of the reactance appears
to be symmetric with respect to the axis AA′, both geometrically and mag-
netically. Since the lower part of the machine is exactly the mirrored image
of the upper part, with respect to the axis AA′, and the flux density compo-
nents cannot be discontinuous, then the flux lines must be normal to the axis
AA′. The structure can be simplified as shown in Figure 4.3(b), imposing the
homogeneous Neumann’s condition along the boundary AA′. Thus the anal-
ysis is then carried out on such a portion of the structure only.

4.3.3 Periodic Condition

This condition corresponds to assigning a correspondence between the val-
ues of the magnetic vector potential along two (or more) boundary lines of
the structure. In general terms, two or more boundary lines are chosen. Among
them a principal line is selected, and the potential of the other (secondary)
lines is expressed as a function of the potential of the principal line.

A A' A A'
SCu+ SCu−
Neumann
condition

(a) (b)

FIGURE 4.3
Reduction of the analysis domain by means of Neumann’s conditions.
The Analysis Procedure Using the Finite Element Method 65

ϑ ϑ

Even periodicity Odd periodicity

(a) (b)

FIGURE 4.4
Reduction of the analysis domain by means of the periodic condition.

These boundary conditions are useful in structures that exhibit a repetition

of the electromagnetic fields, but neither Dirichlet’s nor Neumann’s bound-
ary conditions are appropriate. Assigning the periodic conditions along at
least two lines of the boundary, the structure can be reduced again and the
analysis carried out only on a part of it, the disregarded parts becoming
mirrored images.
As an example, let us refer to the section of the synchronous generator of
Figure 4.2. The machine exhibits a number of pole pairs equal to 3, i.e., a
recurrence with respect to the azimuthal coordinate ϑ equal to 3. It is possible
to study only a third of the machine, as represented in Figure 4.4(a), by
imposing on the boundary that

 π
( )
A z r , ϑ = + A z  r , ϑ + (2 k ) 
 p
k = 1, 2 , 3, … (4.2)

which is named the even periodic condition.

In addition, it is possible to study only a sixth of the machine, as shown
in Figure 4.4(b), by imposing on the boundary that

 π
( )
A z r , ϑ = − A z  r , ϑ + (2 k − 1) 
 p
k = 1, 2 , 3, … (4.3)

which is named the odd periodic condition.

Finally, along the remaining part of the external circumference, the value
of magnetic vector potential Az = 0 is assigned as in Figure 4.2.
As it is evident in the previous example, the process of locating of sym-
metry lines and periodic lines, together with the corresponding boundary
conditions, may reduce the study of a complex structure to only a part of it.
This conveys to a reduction of the field problem domain, with the twofold
advantages of shorter computing time and of greater accuracy in the analysis
of the remaining part.
66 Electrical Machine Analysis Using Finite Elements

y
x

Line of the equal

values of the
B potential Az

FIGURE 4.5
Flux lines, i.e., lines of the flux density vector B.

4.4 Computation of the Solved Structure

Once the structure is drawn and the boundary conditions are assigned, the
electric and magnetic properties of the materials are assigned together with
the field sources. The latter are current density or permanent magnet magne-
tization in the magnetostatic field problem. The solution of the field problem
(4.1) consists of the knowledge of the magnetic vector potential Az in each
point of the domain. The flux density and magnetic strength vectors are
derived from Az. Some considerations are reported in the following sections.

4.4.1 Drawing the Flux Lines

The flux lines are the lines to which the flux density vector B is parallel. They
correspond to the equipotential lines of the magnetic vector potential Az.
In Figure 4.5, the potential Az is assumed to be constant along a line. Let
us refer to a coordinate system Oxy such as the x-axis coincides with the
equipotential line. By applying the curl to the magnetic vector potential, the
components of the flux density vector are obtained as

∂A z
Bx =
∂y
(4.4)
∂A z
By = − =0
∂x

Equation (4.4) shows that the flux density vector has a component only in
the direction of the equipotential line of Az.

4.4.2 Magnetic Flux and Flux Linkage

The computation of the magnetic flux (1.8), through a surface S oriented by
the normal unity vector n, is given by
The Analysis Procedure Using the Finite Element Method 67

y B n y
B n
S

1 2 1 IS 2
L

z x x
(a) (b)

FIGURE 4.6
Surface integral in the 2D problem.

Φ=
∫ B ⋅ n dS = L ∫ B ⋅ n dl
S lS
(4.5)

In the 2D study, the surface integral is reduced to the computation of a

line integral: the magnetic flux is computed by integrating, along the given
oriented line lS, the flux density normal to the line itself; and the result is
multiplied by the length L. This is shown in Figure 4.6.
Otherwise, applying the Stokes theorem (1.21), the magnetic flux can be
obtained as the loop integral of the magnetic vector potential A along the
closed line l, bordering the surface S, which is

Φ=
∫ B ⋅ n dS = ∫ ( curlA) ⋅ n dS =
∫ A ⋅ t d l
S S l
(4.6)

In a 2D problem, such a computation results y

very simply. Referring to Figure 4.7, it is 1 2
observed that the line integral is null along the Az1
segments 1′2′ and 21, since the magnetic vector Az2
potential has the z-axis component only, which 1' 2'
x
is perpendicular to the segments. Along the
L
other two lines, which have length equal to L z
and are parallel to A, the value of Az is constant.
Then the magnetic flux is equal to the difference FIGURE 4.7
Computation of the magnetic
of Az in the two points 1 and 2, times the length
flux by means of the magnetic
L, which is vector potential.

Φ = (A z1 − A z 2 )L (4.7)

In the case of massive conductors with cross-section SCu, it is convenient

to consider the average value of the magnetic vector potential Az on the
conductor surface, which is

L  
Φ=
S Cu  ∫ S Cu +
A z dS −
∫ S Cu −
A z dS 

(4.8)
68 Electrical Machine Analysis Using Finite Elements

where SCu+ and SCu– are the conductor surfaces, positively oriented and
negatively oriented, respectively.
If the current is supposed to be uniformly distributed on the conductor
surface, the flux linkage Λ is obtained by multiplying the magnetic flux Φ
by the number of turns Nt that link the flux.

4.4.3 Joule Power Losses

If the current density is known in each point of the structure, at the time
instant t, the instantaneous Joule power losses are given by

pJ (t ) =
∫ ρJ
τ
2
dτ (4.9)

Of course, with planar symmetry in which J = Jzuz, the volume integral is

reduced to

pJ (t ) = L
∫ ρJ
S
2
z dS (4.10)

where L is the axial length of the structure.

In the case of stationary fields, the current density is not time dependent,
so that the power is also constant. In case of fields that are not constant, but
are sinusoidal in time, the symbolic representation with complex phasors
(1.54) is used. Again the current density has a z-axis component only, but is
now represented in complex form as

J = J zu z = J z e jϕu z (4.11)

The average power that is lost in a period due to Joule effect is given by

2
Jz
∫ ∫
1
pJ = ρ dτ = ρ J ⋅ J dτ (4.12)
τ 2 2 τ

where the factor 1/2 appears, because
Jz
denotes the maximum of the
sinusoidal variation.

4.4.4 Magnetic Energy

The magnetic energy stored in the structure can be computed as

B A
Wm =
∫∫
τ 0
H ⋅ dB dτ =
∫∫
τ 0
J ⋅ dA dτ (4.13)
The Analysis Procedure Using the Finite Element Method 69

that with linear media becomes

∫ H ⋅ B dτ = 2 ∫ µH
1 1
Wm = 2
dτ
2 τ τ
(4.14)

∫ ∫JA
1 1
= J ⋅ A dτ = z z dτ
2 τ 2 τ

In the case of stationary fields, the energy is constant, but in the case of
variable fields, this energy has to be considered as instantaneous, at the
time t. In particular, if the variable fields have a sinusoidal variation, and
they are described by using the symbolic notation, it is possible to compute
the average energy stored in a period as

B⋅H Jz ⋅ Az
∫ ∫
1 1
Wm = dτ = dτ (4.15)
2 τ 2 2 τ 2

where 1/2 appears because
B
and
H
represent the maximum value of the
flux density and the magnetic field strength, respectively, and also
Jz
and

Az
represent the maximum value of the current density and magnetic vector
potential, respectively.

4.4.5 Magnetic Coenergy

The magnetic coenergy can be computed as

Wm′ =
∫ H ⋅ B dτ − W
τ
m

(4.16)
=
∫τ
J ⋅ A dτ − Wm

4.4.6 Magnetic Forces

To compute the magnetic forces, which act on an object within the domain,
two methods are possible.

4.4.6.1 Tensor of Maxwell’s Strength

As discussed in Chapter 1, the method of Maxwell’s strength tensor allows
the magnetic forces acting on an object within the magnetic field to be
computed. At first a suitable surface that contains the object is selected. In
the 2D problem this surface is reduced to a line, named l. The tangential and
normal components of the force are computed by means of the integral along
the line l, as
70 Electrical Machine Analysis Using Finite Elements

Ft = Lµ o

∫ H H
l
t n dl
(4.17)

∫ ( H ) dl
µ
Fn = L o n
2
− Ht 2
2 l

The magnitude of the force and the direction of the force (which is
expressed with respect the direction normal to the surface) are given by
Equation (1.100) and Equation (1.101), respectively.
As far as the computation time is concerned, it is worth noticing that
Maxwell’s strength tensor requires only one field solution.
As far as the computation accuracy is concerned, it is observed that with
a correct magnetic field distribution, the method of Maxwell’s strength tensor
provides an exact value of the force, disregarding the choice of the surface
(i.e., the line) of integration. However, due to the intrinsic approximation of
the adopted numerical method, the continuity of the field components
among the adjacent elements is not guaranteed. As a result, the computed
forces depend on both the path of integration and the adopted mesh of the
domain.
Some preliminary tests are essential, in order to verify the accuracy obtain-
able with the adopted mesh of the structure and the selected path for the
integration. For instance, a possible check may be the evaluation of the force
in a circumstance where this force should be null. An example to check the
accuracy of the results by using the method of Maxwell’s strength tensor is
illustrated in Chapter 12.

4.4.6.2 The Virtual Works Method

As an alternative to Maxwell’s strength tensor, or just to verify its accuracy
in some conditions that are particularly critical, the virtual works method can
be adopted. Unfortunately, since this technique is based on the computation
of the variation of the magnetic energy between two different positions, it
requires the solution of at least two field problems, with an increase of the
computation time and with the not negligible problem of the choice of the
width of the position variation.
This method is based on the comparison of the energy balance between
two different positions, corresponding to a virtual change of position of the
object in the direction where the force is computed. As will be discussed in
the following chapter, the force is expressed as the derivative of the magnetic
coenergy with respect to the virtual movement, with constant electrical
sources. This implies the solution of the field problem at least with the
structure in two different configurations, with an increase of computation
and the problem of the width of the movement.
The Analysis Procedure Using the Finite Element Method 71

4.4.7 Determination of the Electrical Parameters

From the field solution, it could be interesting to calculate the parameters of
a zero-dimensional electrical network related to the electrical machine that
is analyzed. In particular, the electrical resistances, inductances, and capac-
itors are normally computed.
The resistance is computed from the Joule power losses, as

pJ
R= (4.18)
I2

In case of linearity, the inductance is obtained from the stored magnetic

energy and from the flux linkage, as

2 Wm Λ
L= = (4.19)
I2 I

Further observations are reported in the chapters that follow.

References
See references in Chapter 2.
5
Cylindrical Magnetic Devices

In this chapter, a cylindrical magnetic device is considered. A magnetostatic

axial-symmetric analysis is carried out. At first, an analytical study is pre-
sented. Then the basic concepts about the energy conversion are summarized.
Finally, the finite element method is applied for the analysis of the device.

5.1 Introduction
The cylindrical magnetic device is used in applications requiring high thrust
but small stroke, e.g., to move power switches, to drive valves, and so on.
A section of the cylindrical magnetic device is shown in Figure 5.1. It consists
of a fixed magnetic part, i.e., the core, and of a cylindrical magnetic plunger
moving axially within the core, guided by a nonmagnetic guide. A coil
composed by Nt turns is wound inside the core.
When no currents feed the coil, the plunger is in its rest position, at the
lower position, at which the air-gap thickness is maximum, i.e., g = gmax.
When the coil is fed by a current with a sufficiently high amplitude, the
plunger pops up, reaching the upper position, at which the air-gap is min-
imum, i.e., g = gmin.
The radial gap between the core and the plunger, corresponding to the
driver thickness, is constant and equal to t.

5.2 Analytical Study of the Magnetic Device

Before applying the finite element method, an analytical analysis of the
magnetic device is carried out. In spite of some convenient assumptions, this
study allows a first evaluation of the main electrical and magnetic quantities,
in such a way as to achieve their expected value. In the following finite
element analysis, it will be easy and rapid to detect any possible errors, if
any discrepancy will be found. Furthermore, the analytical and the numerical

73
74 Electrical Machine Analysis Using Finite Elements

fixed magnetic
part (core)
coil
g variable
airgap

ht non-magnetic
driver
r
cylindrical
plunger t d

FIGURE 5.1
Sketch of the cylindrical magnetic device.

results can be compared so as to evaluate the effects of the simplifications

of the analytical study.
In the analytical analysis, the flux leakages and the air-gap edge effects
are neglected. Thus, in correspondence of the air-gap the flux lines are
assumed to be normal to the two surfaces.

5.2.1 Computation of the Magnetic Quantities

Assuming that the coil carries a current i is feeding the coil, the flux density
in the two air-gaps is computed as follows. Referring to Figure 5.2, using
Ampere’s law and neglecting the magnetic voltage drop along the iron paths
yields

Nt i = H g g + H t t (5.1)

where Hg and Ht are the magnetic field strength in the air-gap g and t,
respectively. Because of the small air-gap thickness t, a constant value of Ht
is considered.
Gauss’s law yields

πd 2
Bg = B t h t π(d + t ) (5.2)
4

where Bg and Bt are the flux density in the air-gap g and t, respectively. In
Equation (5.2), π(d + t) is the average circumference in the middle of the air-
gap t. From the two equations (5.1) and (5.2), the flux density Bg is obtained as

Nt i Nt i
Bg = µo = µo (5.3)
d2t g + t'
g+
4 h t π(d + t )
Cylindrical Magnetic Devices 75

Nsp i
g Bg Hg

Rg ϕg

Ht Bt
Rt
t d

FIGURE 5.2
Definition of the magnetic strength and the flux density in the two air-gaps of the device (a)
and equivalent magnetic circuit (b).

with obvious meaning of the term t′.

Alternatively, the value of Bg can be obtained from the equivalent magnetic
circuit, shown in Figure 5.2(b). The magnetic reluctance corresponding to
the air-gap g is

g
Rg = (5.4)
πd 2
µo
4

and the magnetic reluctance corresponding to the air-gap t is

Rt =
(
ln 1 + 2 t / d ) (5.5)
µ o 2 πh t

With a small ratio t/d, so that ln(1 + 2t/d) is approximated by t/(d + t),
Equation (5.5) can be expressed as

t
Rt = (5.6)
µ oπh t (d + t )

The magnetic flux ϕg through the base surface of the plunger is computed
as

Nt i
ϕg = (5.7)
Rg + Rt

The flux density Bg is achieved as the ratio between ϕg and the air-gap
surface πd2/4, as
76 Electrical Machine Analysis Using Finite Elements

ϕg
Bg = (5.8)
 πd 2 
 4 
 

and therefore it results as in Equation (5.3).

The flux linkage with the coil results as follows:

Nt 2 i
λ = Nt ϕ g = µ oπd 2 (5.9)
4(g + t ')

It is assumed that the magnetic device is without hysteretic phenomena

and without losses. If needed, the latter may be considered as external losses.
In other words, the system is split in a part without losses (considered in
the following study) and a part with losses (not considered here). The mag-
netic device is then considered to be a conservative system, which can store
magnetic energy in a reversible way.
With these assumptions, the current i, the flux linkage λ, and the air-gap
thickness g are considered to be the state variables, which define completely
the state of the magnetic system. However, Equation (5.9) shows that there is
a link among these three quantities (each value of λ corresponds to precise
values of i and g). Since hysteresis is neglected, such a correspondence is
univocal. Thus, only two state variables are enough to describe completely
the state of the system, and they will be chosen on the basis of best convenience.

5.2.2 Magnetic Energy and Coenergy

The magnetic energy can be determined by the system geometry, by the
magnetic characteristic of the materials, and by the coil MMF. For a given
configuration, with a fixed position of the plunger, i.e., an air-gap g = go, the
energy stored in the magnetic field is computed as the energy furnished by
the electric source to bring the current from zero to the value i = io, to which
corresponds the flux linkage λ = λo = λ(io, go). It results that

λo
Wm =
∫ 0
i(λ)dλ (5.10)

Such an energy corresponds to the surface that is horizontally hatched in

Figure 5.3. With a reversible process, this magnetic energy is completely
returned to the electrical circuit, when the current is brought back to zero.
The surface that is vertically hatched, below the curve of Figure 5.3, cor-
responds to the magnetic coenergy and can be expressed as
Cylindrical Magnetic Devices 77

λ magnetic g = constant
energy magnetization
λo curve
Wm

W'm magnetic
coenergy
io i

FIGURE 5.3
Magnetization curve, magnetic energy, and coenergy.

λo
Wm′ =
∫0
λ(i)di
(5.11)
= λ oi o − Wm

This quantity, which is without an evident physical meaning, proves to be

very useful in the computation of the electromechanical quantities. Since the
system is conservative, both magnetic energy and coenergy may be consid-
ered as state functions. Thanks to the univocal correspondence among g, i,
and λ, the magnetic energy and coenergy are usually expressed as functions
of the current and the flux linkage, that is Wm(λ, i) e W′m (λ, i).
If the system is linear, i.e., the magnetic materials do not reach the satu-
ration, the magnetic energy and coenergy coincide and are computed as

1
Wm = Wm′ = λ oio (5.12)
2

In the equations above, the magnetic energy is expressed by means of

integral quantities that can be measured at the terminals of the device. The
magnetic energy can be also expressed as a function of the specific property
(or point property) of the magnetic field. Since the flux density has been
supposed to be constant in the two air-gaps, the edge effects to be negligible,
and the magnetic field strength to be null in the iron paths, the magnetic
energy results in

 1  πd 2  1 
Wm =  Bg 2  g+ B t 2  h t π(d + t )tt (5.13)
 2µ o  4  2µ o 

5.2.3 Apparent Inductance and Differential Inductance

The apparent inductance and the differential inductance are functions of the
air-gap g. Due to the nonlinear characteristic of the ferromagnetic material
78 Electrical Machine Analysis Using Finite Elements

dWm g = go
λo

di

io i

FIGURE 5.4
Current-flux linkage characteristic and surface corresponding to the apparent inductance com-
putation.

(the core and the plunger), they are also functions of the current i flowing
through the coil.
Let us consider the operating point (io, λo), defined by the air-gap go, the
current io, and the flux linkage λo = λ(io, go). The apparent inductance of the
coil is given by the ratio between the flux linkage λo and the current io, which
is

λo
(
Lapp i o , g o = ) io
(5.14)

The ratio between an infinitesimal variation of flux linkage, dλ, and the
corresponding variation of current, di, around the operating point (io ,λo),
defines the differential inductance of the coil, which is

dλ
(
Ldif i o , g o = ) di ( i
(5.15)
o ,λo )

In linear conditions, since the permeability of the materials is constant, the

two values of Lapp and Ldif are constant and coinciding: Lapp = Ldif = L.
Alternatively, the inductances can be computed from the energy quantity.
Referring to Figure 5.4, the apparent inductance can be computed as

Wm + Wm′ λ i
( )
Lapp i o , g o =
io2
= o 2o
io
(5.16)

The quantity (½λoio) has the same dimension of an energy, but it does not
have a physical meaning. It corresponds to the surface shown in Figure 5.4.
The same figure highlights the magnetic energy variation dWm corresponding
Cylindrical Magnetic Devices 79

λ B" g−dg λ λ B" g−dg

g−dg
B B coenergy B
λo B' g λo B' g g
A A A
energy

O io i O i O io i

FIGURE 5.5
Energy variations corresponding to the air-gap variation –dg.

to the current variation di, around the working point. From this variation,
since dWm ≈ dλio, the differential inductance can be roughly estimated as

( )
Ldif i o , g o ≈
1 dWm
i o di ( λ ,i )
(5.17)
o o

In linear conditions, the two inductances given in Equation (5.16) and

Equation (5.17) are simplified as

2 Wm
Lapp = Ldif = (5.18)
io2

5.2.4 Mechanical Forces

Since the magnetic device is a conservative system, the mechanical forces
that take effect can be determined by means of the principle of the energy
conservation. One has to estimate the variation of the energy corresponding
to a infinitesimal movement of the element in the direction along which the
force component has to be computed.
The force Fz on the plunger of the magnetic device of Figure 5.1 along the
direction of the z-axis is associated to a movement dz = −dg of the plunger
itself. According to the movement –dg, there is a variation of both the coil
current and the flux linkage. Figure 5.5(a) shows the starting working point
A, characterized by an air-gap length g, and the subsequent working point B,
characterized by an air-gap length g–dg. Because the movement –dg is
infinitesimal, the difference between the two curves is infinitesimal as well.
The force Fz realizes an infinitesimal mechanical work dWmech = Fzdz =
–Fzdg. It corresponds to the area OAB of Figure 5.5(a), as can be obtained
from the following energy balance:

dWmech + dWm = dWel (5.19)

80 Electrical Machine Analysis Using Finite Elements

λ λ λ λ
B B B B

A A A A

i i i i
dWmech Wm(A) Wm(B) dWel

FIGURE 5.6
Energy balance during the air-gap variation -dg.

where dWm is the variation of the magnetic energy stored in the magnetic
field, during the movement of the plunger from A to B, which is

dWm = Wm (B) − Wm (A)

λ o + dλ λo (5.20)
=
∫
0
i(λ , g − dg)dλ −
∫ 0
i(λ , g)dλ

and dWel is the electrical energy that is furnished by the electrical source via
the coil terminals, which is

λ o + dλ
dWel =
∫λo
i(λ)dλ (5.21)

These quantities can be drawn in the (i, λ) plane, as shown in Figure 5.6.
Neglecting the infinitesimal term of second order, the area OAB can be
approximated by the area OAB', shown in Figure 5.5(b), which corresponds
to the difference between the stored magnetic energy before and after the
movement –dg, with a constant flux linkage λ = λo. Then the force component
is computed as

(
Fz λ o , g = − ) dWm
=−
dWm λ , g ( ) (5.22)
dz λ = λo
−dg
λ = λo

Equation (5.22) highlights that the energy has to be expressed as a function

of λ. Since the flux linkage is constant, the electrical energy furnished to the
system is null. The mechanical work Fzdz = −Fzdg is accomplished at the
expense of the energy stored in the magnetic field.
In a similar way, the area OAB can be approximated by the area OAB″,
shown in Figure 5.5(c). It corresponds to the difference between the coenergy
of the system before and after the movement –dg, with a constant current
i = io. In this case, the force is given by
Cylindrical Magnetic Devices 81

λ λ λ
B B B

A A A

i i i

FIGURE 5.7
Mechanical work with different finite movements: although the starting and final points are
the same, the dependence on the way of variation of λ and i is evident.

( )
Fz i o , g = +
dWm′
=+
( )
dWm′ i , g
(5.23)
dz i = io
−dg
i = io

Equation (5.23) highlights that the energy has to be expressed as a function

of i. Since the current is constant, i.e., i = io, the electrical energy furnished
to the system is equal to iodλ. The mechanical work Fzdz = −Fzdg is computed
as the difference between the electrical energy, furnished by an external
source, and the variation of the magnetic energy stored in the magnetic field,
which is ½iodλ. This computation corresponds to that adopted using the
finite element method, in which the current is set as the source of the field.
Equation (5.22) and Equation (5.23) describe the fundamental relationships
of the electromechanical energy conversion. They are used to evaluate the
forces of magnetic nature that act on the mobile element. The force results
independent of how the current and flux linkage vary during the infinitesimal
movement –dg, since these variations are infinitesimal of the second order.
In other words, the force depends on the state variables (g, i, and λ) but is
independent of the way they change (as physically expected). However, this
occurs only if the movement is infinitesimal: with a finite movement, the
mechanical work is dependent on the way of variation, as shown in Figure 5.7.
In the particular case of the cylindrical magnetic device of Figure 5.1,
supposing linear operations, the magnetic coenergy is computed as W′m =
½λi, where λ is expressed by Equation (5.9), so that

( )
2
1 Nt i πd 2
Wm′ = µo g (5.24)
2 g + t′ 4

Using Equation (5.23), the force in direction of the infinitesimal movement

dz = −dg results in

πd 2
1
( ) 1
2
Fz = Nt i (5.25)
2µ o ( )
2
4 g + t′
82 Electrical Machine Analysis Using Finite Elements

A positive value of Fz means that the plunger is attracted upwards.

In the equations above, the magnetic force is expressed by means of inte-
gral quantities, as energy and coenergy. The same forces can be computed
starting from the specific properties of the magnetic field, as an example by
means of Maxwell’s stress tensor, applied to the magnetic field components.
In the particular case of the magnetic device, with the hypothesis that only
the z-axis component of the field strength exists in the air-gap g (see the first
row of Table 1.2, in Chapter 1), the magnetic pressure normal to the plunger
surface is given by

pn =
µo
2
( µ
) B 2
Hn 2 − Ht 2 = o Hn 2 = g
2 2µ o
(5.26)

where Hn and Ht are the normal and tangential components of the field
strength at the surface of the plunger in front of the air-gap. Then the total
force on the plunger is

1 πd 2
Fz = Bg 2 (5.27)
2µ o 4

By introducing the expression of Bg given in Equation (5.3) in Equation

(5.27), the same expression of the force given in Equation (5.25) is again
obtained.

5.3 Finite Element Analysis

5.3.1 Formulation of the Problem for the Finite Element Analysis
The device presents an axial symmetry (see Figure 4.1), thus only a section
of the magnetic device may be considered. The analysis is carried out in the
(r, z) plane as shown in Figure 5.8.
In fact, some considerations about the symmetry show that only the azi-
muthal component of the current density vector may exist, i.e., J = (0, Jϑ, 0).
As a consequence, the magnetic vector potential is characterized by only the
azimuthal component, i.e., A = (0, Aϑ, 0). Then the components of the mag-
netic flux density B = (Br , 0, Bz) are

∂A ϑ
Br = −
∂z
(5.28)
Bz =
1 ∂ rA ϑ( A
= ϑ+
)
∂A ϑ
r ∂r r ∂r
Cylindrical Magnetic Devices 83

z
Aϑ = 0

Aϑ = 0 Aϑ = 0

r
Aϑ = 0

FIGURE 5.8
Sketch of the section of the cylindrical magnetic device, which is used in the finite element
analysis and boundary conditions.

In addition, all the field quantities depend on the variables r and z only,
i.e., Jϑ = Jϑ(r,z), Aϑ = Aϑ(r,z), Br = Br(r,z), Bz = Bz(r,z). The components of the
magnetic field strength vector H = (Hr , 0, Hz) are connected to those of the
flux density vector by the constitutive law.
The field problem is magnetostatic and is expressed by the Poisson differ-
ential equation

1 ∂  ∂A ϑ  ∂2 A ϑ
∇2A ϑ = r + = −µJ ϑ (5.29)
r ∂r  ∂r  ∂z 2

Dirichlet’s boundary condition is assigned on the z-axis, with a reference

potential Aϑ = 0. The same value is assigned along a line at a suitable distance
from the device. The boundary conditions are shown in Figure 5.8.
A unity relative permeability is considered for nonmagnetic materials,
while mild steel B-H curve is used for the magnetic materials. If the flux
density is low, a linear behavior may be supposed for the iron paths, with
constant magnetic permeability. The typical value of the relative magnetic
permeability of nonsaturated iron materials is in the range from µr = 1000
to µr = 10,000.
The coil is formed by Nt turns, but it is modeled as a single conducting
bar, as shown in Figure 5.8. Thus, when the coil is fed by a current i = I, the
overall current NtI has to be assigned to the equivalent bar.
84 Electrical Machine Analysis Using Finite Elements

5.3.2 Computation of the Solved Structure

Once magnetic material characteristics, boundary conditions, and winding
current source are assigned, the field problem can be solved. As highlighted
by Equation (5.29), the solution consists of computing the azimuthal com-
ponent of the magnetic vector potential Aϑ(r, z) in each point of the domain.
All the other magnetic quantities of interest are derived from Aϑ(r, z), as
described later.

5.3.2.1 Magnetic Flux Density

The two components of the flux density vector are obtained from Equation
(5.28). From them, the flux lines are carried out. The magnitude of the flux
density results in

B = Br 2 + Bz 2 (5.30)

5.3.2.2 Flux Linkage

At first, a single dimensionless turn is considered. Its radius is ro and its
center is on the z-axis, as illustrated in Figure 5.9(a). The turn flux linkage
is computed as the integral of the flux density across the circular surface
limited by the turn itself, which is

ro 2π
λ ro =
∫ ∫
0 0
B z (r , z) r dϑ dz (5.31)

z
Bz uz r

ro

dϑ
dr = r dϑ
(a)

z
z
r r

ro
Aϑ uϑ
dϑ A = (0, Aϑ, 0)
SCu
dr = r dϑ

(b) (c)

FIGURE 5.9
Integral lines for the computation of the flux linkage.
Cylindrical Magnetic Devices 85

By means of the Stokes theorem, the flux linkage is conveniently expressed

as

λ ro =

∫ A (r, z) dr
ro
ϑ

2π
=
∫0
roA ϑ (r , z) dϑ (5.32)

= 2 πroA ϑ (ro , z)

as shown in Figure 5.9(b). If the turn is not dimensionless but has a section
SCu, as shown in Figure 5.9(c), the average flux linkage becomes

∫
1
λ= 2πrA ϑdS (5.33)
S Cu S Cu

At last, if Nt turns are considered that cover the whole surface SCu, the flux
linkage is

∫
Nt
λ= 2πrA ϑdS (5.34)
S Cu S Cu

5.3.2.3 Magnetic Energy and Coenergy

For computing the magnetic energy and coenergy, the equations reported in
Section 5.2.2 may be adopted. However, it is convenient to compute the energy
from the specific magnetic field quantities: the magnetic field strength H, the
flux density B, the current density J, and the magnetic vector potential A.
The magnetic energy Wm is computed as

B B
Wm =
∫∫
τ 0
HdB dτ = 2 π
∫ ∫
S
r
0
HdB dS (5.35)

where r is the dummy radius. Referring to the current density and the
magnetic vector potential, Wm is given by

Aϑ Aϑ
Wm =
∫∫
τ 0
J ϑdA ϑ dτ = 2 π
∫ ∫
S
r
0
J ϑdA ϑ dS (5.36)

The magnetic coenergy W′m, computed from the field quantities, is

Wm′ =
∫ BH dτ − W
τ
m (5.37)
86 Electrical Machine Analysis Using Finite Elements

In linear conditions, Equation (5.35) becomes

∫ r 2 BH dS
1
Wm = 2 π (5.38)
S

and Equation (5.36) becomes

∫ r2J A
1
Wm = 2 π ϑ ϑ dS (5.39)
S

The core and the plunger of the magnetic device show a nonlinear mag-
netic characteristic. However, when the magnitude of the flux density is low,
they work in the linear part of the B-H curve. Hence, a constant permeability
can be assigned even in these objects.

5.3.2.4 Inductances
The apparent inductance is obtained by dividing the flux linkage (5.34) by
the current:

∫
Nt
2πrA ϑdS
S Cu
Lapp =
S Cu
(5.40)
i

If the magnetic circuit is linear, the inductance Lapp can be computed from
the energy quantities as

1 

Lapp =
2
2 ∫ BH dτ = 2π ∫ rBH dS
τ S
(5.41)
i2 i2

The differential inductance has to be computed from two field solutions,

corresponding to different coils currents. For instance, let us suppose that
the flux linkages λ(i) and λ(i + ∆i) correspond to the currents i and i + ∆i,
computed as given in Equation (5.34). Then the differential inductance is
given by

λ(i + ∆i) − λ(i)

Ldif = (5.42)
∆i

The result depends on the value of the current variation ∆i. A more accurate
result is obtained if the field problem is solved for two current values around
i, which are i + ∆i and i – ∆i. Adopting the Taylor series expansion of the
Cylindrical Magnetic Devices 87

flux linkage around the current value i, and neglecting the terms of order
higher than the second order, it is

dλ(i) 1 d 2 λ( i ) 2
λ(i + ∆i) = λ(i) + ∆i + ∆i +…
di 2 di 2
(5.43)
dλ(i) 1 d 2 λ( i ) 2
λ(i − ∆i) = λ(i) − ∆i + ∆i −…
di 2 di 2

By subtracting the second equation from the first one, the second-order
terms disappear and the differential inductance results in

dλ(i) λ(i + ∆i) − λ(i − ∆i)

Ldif = = (5.44)
di 2 ∆i

5.3.2.5 Electromagnetic Forces

The force acting on the plunger can be evaluated using Maxwell’s stress
tensor. Let us refer to a surface containing the plunger, which is a line on
the (r,z) plane bounding the plunger. The total force Fz on the plunger in the
z-axis direction is given as the sum of the force components that are com-
puted on the different parts of the surface. The parts of the whole surface
that are normal to the unity vectors ur and uz are denoted Sr and Sz, respec-
tively. On the surface Sz, whose normal unity vector has a z-axis direction,
the force contribution to Fz corresponds to the force component normal to
Sz, which is

Fn =
∫Sz
µo
2
(
H n 2 − H t 2 dS ) (5.45)

On the surface Sr , whose normal unity vector has an r-axis direction, the
force contribution to Fz corresponds to the component of the force tangential
to Sr , which is

Ft =
∫Sr
µ oH n H t dS (5.46)

Alternatively, the force in the z-axis direction can be computed from the
magnetic coenergy variation, see Equation (5.23), as

Fz = −
( )
dWm′ i , g
dg
i = const
(5.47)

=−
( ) (
Wm′ i , g + ∆g − Wm′ i , g − ∆g )
2 ∆g
i = const
88 Electrical Machine Analysis Using Finite Elements

This region is defined

air when the plunger is
in the position g or g−∆g g g+∆g
g+∆g

This region is defined

air when the plunger is
in the position g + ∆ g
plunger

FIGURE 5.10
Definition of the regions that are interesting to the movement of the plunger; they are adopted
to avoid errors due to variation of the mesh.

where W′m (i, g + ∆g) and W′m (i, g – ∆g) are the magnetic coenergy values that
are computed from the field solutions, and correspond to the air-gap length
(g + ∆g) and (g – ∆g), respectively, and to the given current i.
The result of Equation (5.47) depends on the variation of the air-gap length
∆g. This has to be neither too small, so that the difference in the numerical
solution is not appreciable, nor too large, so that an excessive variation of
magnetic coenergy occurs.
In addition, the result may be influenced by the variation of the mesh of
the domain, since the geometry changes when the plunger is in the two
different positions. In order to avoid this numerical error, it is useful to define
the regions involved in the plunger movement, as shown in Figure 5.10.
When the plunger occupies these regions, they are considered as iron, assign-
ing the corresponding magnetic property. Conversely, when the plunger
does not occupy these regions, they are considered as air.

5.4 Example
This section illustrates the analysis of the cylindrical magnetic device shown
in Figure 5.11, together with its main dimensions. The coil is formed by Nt =
1000 turns, whose diameter is dc = 0.9 mm. The coil is represented by the
rectangle with sides 90 mm and 17.5 mm. The nonmagnetic driver, which
determines the lateral gap between the core and the plunger, is t = 0.5 mm.
When the plunger is in its lower position, the air-gap length is gmax = 15 mm,
while when the plunger is in its higher position, the air-gap length is gmin =
1 mm.
Since the problem is axial-symmetric, the section of the magnetic device
of Figure 5.8 is analyzed. The current density Jϑ is fixed as source, and the
magnetic vector potential Aϑ represents the field solution. The boundary
condition Aϑ = 0 is fixed on the boundary of the domain, as illustrated in
Figure 5.8. The mild steel B-H curve is used for the core and the plunger.
Cylindrical Magnetic Devices 89

15

150 100 90 1 – 15 mm

22

0.5 45
60
100
125

FIGURE 5.11
Main dimension of the cylindrical magnetic device.

FIGURE 5.12
Mesh of the domain: initial mesh (a) and refined mesh (b).

Figure 5.12 shows the mesh of the device: Figure 5.12(a) shows the initial
mesh, while Figure 5.12(b) shows the refined mesh. The mesh was mainly
refined in the air-gap region, where the highest field gradients take place.
The field solution consists of the knowledge of the magnetic vector poten-
tial Aϑ. The equipotential lines are shown in Figure 5.13(a). The flux density
90 Electrical Machine Analysis Using Finite Elements

FIGURE 5.13
Equipotential lines of Aϑ (a) and flux density vectors (b).

vector components are computed from the magnetic vector potential Aϑ, as
reported by Equation (5.28). The vector field B is shown in Figure 5.13(b).
As an example, three configurations of the magnetic device are analyzed,
corresponding to three air-gap lengths: g = 1 mm, g = 5 mm, and g = 15 mm.
The behavior of the z-axis component of the flux density in the middle of
the air-gap is shown in Figure 5.14, with a source current equal to i = 5 A.
The dependence of the flux density value on the air-gap length is evident.
When g = 1 mm, the effect of the plunger and core edges is also manifest.

Flux density component B (T)

2
1.8
1.6 g = 1 mm
1.4
1.2
1
0.8 g = 5 mm
0.6
0.4
0.2 g = 15 mm
0
0 5 10 15 20 25
Radius (mm)

FIGURE 5.14
Behavior of the flux density component Bz in the middle of the air-gap g (with i = 5 A).
Cylindrical Magnetic Devices 91

TABLE 5.1
Computation on the Solved Structure (with a
current i = 5 A)
g = 15 mm g = 5 mm g = 1 mm

∫τ Cu
A ϑ dτ 2.023⋅10–6 4.123⋅10–6 5.204⋅10–6

∫ J A dτ
τ
ϑ ϑ 6.422 13.09 16.52

∫ BHdτ
τ
6.415 13.16 16.79

∫∫
B

HdB dτ 3.217 6.119 3.978

τ 0

λ (Vs) 1.2843 2.6180 3.3042

Wm (J) 3.217 6.119 3.978
W′m (J) 3.198 7.041 12.812
Lapp (H) 0.257 0.524 0.661
Ldif (H) 0.257 0.306 0.117

Some magnetic quantities are computed, always considering the three

configurations, and reported in Table 5.1. It is worth noting that the magnetic
energy obtained by handling the quantities B and H, or the quantities Jϑ and
Aϑ, is substantially the same. Moreover, the magnetic energy and coenergy
are essentially coincident when g = 15 mm, i.e., when the flux density values
are low, as reported in Figure 5.14; therefore there is no saturation in any
part of the device. Conversely, they are very different when g = 1 mm, i.e.,
when the flux density values are high, and a high saturation occurs in the
iron paths.
In Table 5.2, the analytical and the finite element (FE) results are compared.
Concerning the flux density, the results coincide when the air-gap length is
high, while they are very different when the air-gap length is low. This is
because the analytical model does not consider the iron saturation, though
this hypothesis is not acceptable in case of magnetic fields so high. The

TABLE 5.2
Comparison between Analytical and Finite Element (FE) Results
g = 15 mm g = 5 mm g = 1 mm
Analytical FE Analytical FE Analytical FE
Bg (T) 0.415 0.407 1.219 1.097 5.435 (!) 1.87
λ (Vs) 0.659 1.284 1.939 2.618 — 3.304
Wm (J) 1.648 3.217 4.847 6.119 — 3.978
Fz (N) 217.5 169.4 1881 881.5 — 2086
92 Electrical Machine Analysis Using Finite Elements

Flux linkage (Vs)

4
g =1 mm
3.5
g = 5 mm
3
2.5 g = 15 mm
2
1.5
1
0.5
0
0 1 2 3 4 5 6 7 8 9 10
Current (A)

FIGURE 5.15
Flux linkage as a function of the current, with three air-gap lengths.

analytical model should be improved, taking the magnetic voltage drop in

the iron paths into account; however, this lies beyond the scope of this
example and we will not discuss it in any depth.
Table 5.2 also shows that the analytical values of flux linkage and magnetic
energy and those computed by the finite element method are quite different.
This is because the analytical model considers only the flux lines crossing
the air-gap. This assumption is valid only when the air-gap length is low.
Conversely, when the air-gap length is high, the flux lines go through the
lateral surface of the plunger, without crossing the air-gap, as highlighted in
Figure 5.13. The apparent inductance is computed by using Equation (5.40).
Figure 5.15 shows the flux linkage versus the coil current for the three air-
gap lengths. The different saturation level is evident in the three curves.
Keeping in mind the expressions for the magnetic energy (5.10) and magnetic
coenergy (5.11), and the graphical representation of Figure 5.3, the surfaces
representing Wm of Figure 5.15 are compared at the current i = 5 A. The
results are reported in Table 5.2.
The knowledge of the flux linkage, at different current amplitudes, allows
the differential inductance to be computed, by means of Equation (5.44). In
particular, around i = 5 A, with the three different air-gap lengths, the
computation result is reported in Table 5.1. With g = 15 mm, which is without
saturation, the inductances Lapp and Ldif are effectively coincident. With low
air-gap lengths, when saturation occurs in the iron parts, the inductances
are quite different.
The force on the plunger is computed with air-gap length g = 5 mm and
current i = 5 A. At first the force is computed by means of Maxwell’s stress
tensor. As shown in Table 5.3, three lines around the plunger are considered,
and the force density is integrated along these lines. The line A covers only
the upper part of the plunger, so that only the force on this surface is
considered. The line B is placed in the middle of the air-gap, covering a
Cylindrical Magnetic Devices 93

TABLE 5.3
Computation of the Force by Means of Maxwell’s Stress Tensor
Force (N)
g = 5 mm i=5A Line Points 100 Points 200 Points 400

line A line B line C

A 782.0 782.1 782.2

B 829.8 829.8 829.8

C 856.5 875.1 881.5

plunger plunger plunger

line A = on the upper surface of the plunger

line B = in the middle of the air-gap
line C = all around the plunger

surface higher than the plunger surface. Finally, the curve C completely
contains the plunger, and the effects of the lateral surface of the plunger are
also considered in the force computation. The numerical integration is car-
ried out with a different number of points. It is always a good rule to check
the influence of the number of points, to avoid errors due to numerical
integration.
By comparing the results of Table 5.3, obtained on the three different lines
of integration, the following conclusions arise. The attractive force takes
place essentially on the upper surface of the plunger: in fact the force com-
puted on the line A is almost 90% of the total force, considered equal to that
computed on the line C. A good approximation of the force, almost 94% of
the total force, is obtained computing the force on the line B.
The force acting on the plunger for different air-gap lengths is reported in
Table 5.3. These results refer to the integrals along the line C, with 400 points.
The force is also computed by means of the virtual work principle, in the
same conditions, i.e., i = 5 A and g = 5 mm. Two further field solutions are
carried out, considering a constant current i = 5 A, and two variations of the
air-gap length of ∆g = 0.5 mm, the first positive and the second negative.
The values obtained from the field solutions are reported in Table 5.4. Using
Equation (5.46) yields

Fz = −
( ) (
Wm′ i , g + ∆g − Wm′ i , g − ∆g ) =−
6.5799 − 7.477
= 898 N
2 ∆g 2 · 0.5 · 10−3
i = const

which coincides with the force computed by means of Maxwell’s stress tensor.
94 Electrical Machine Analysis Using Finite Elements

TABLE 5.4
Computation of Some Quantities Around g = 5 mm
(current i = 5 A)
g = 4.5 mm g = 5 mm g = 5.5 mm

∫τ Cu
A ϑ dτ 4.278 · 10–6 4.123 · 10–6 3.961 · 10–6

∫ J A dτ
τ
ϑ ϑ 13.58 13.09 12.58

∫ BHdτ
τ
13.63 13.16 12.59

∫∫
B

HdB dτ 6.153 6.119 6.011

τ 0

λ (Vs) 2.716 2.618 2.515

Wm (J) 6.153 6.119 6.011
W′m (J) 7.477 7.041 6.579

References
1. D.C. White and H.H. Woodson, Electromechanical Energy Conversion, John Wiley &
Sons, New York, 1959.
2. G. Xiong and S.A. Nasar, Analysis of Field and Forces in a Permanent Magnet
Linear Synchronous Machine Based on the Concept of Magnetic Charge, IEEE
Trans. on MAG, vol.25, no.3, pp. 2713–2719, 1989.
3. R. Akmese and J.F. Eastham, Dynamic performance of a brushless dc tubular
drive system, IEEE Trans, MAG, vol.25, no.5, pp. 3269–3271, 1989.
4. J. Hur, S.B. Yoon, D.Y. Hwang, and D.S. Hyun, Analysis of PMLSM Using Three
Dimensional Equivalent Magnetic Circuit Network Method, IEEE Trans, MAG,
vol.33, no.5, pp. 4143–4145, 1997.
5. I. Boldea and S.A. Nasar, Linear Electric Actuators and Generators, Cambridge
University Press, Cambridge, UK, 1997.
6. N. Bianchi, S. Bolognani, F. Tonel, Design Considerations for a Tubular Linear
PM Motor for Servo Drives, in Proc. EPE-PEMC 2000 International Conference,
Kosice, Slovak Republic, 5–7 September 2000.
7. N. Bianchi, Analytical Computation of Magnetic Fields and Forces of a Tubular
PM Linear Servo Motor, in Proc. di IEEE IAS Annual Meeting, Roma, Italy,
8–12 October 2000.
8. J.F. Gieras and Z.J. Piech, Linear Synchronous Motors. Transportation and Auto-
mation Systems, CRC Press, London–New York, 2000.
6
The Single-Phase Transformer

The aim of this chapter is to describe the finite element analysis of a low-
power, single-phase transformer. We start by defining the equivalent circuit
of the transformer; then we go on to describe the procedure to obtain its
parameters.

6.1 The Single-Phase Transformer

The iron core of the single-phase transformer may be of core type or shell
type. Figure 6.1(a) shows a section of the core-type transformer, whose pri-
mary and secondary windings are divided in two coils each and wound
around the two legs. Figure 6.1(b) shows the shell-type transformer, whose
primary and secondary windings are wound around the central leg.
The core of the low-power, single-phase transformer is obtained by stack-
ing silicon-steel laminations. The laminations are insulated to reduce the
eddy current losses. Generally a rectangular form of the leg is adopted.
Neglecting the edge effects, e.g., the end-winding flux leakage, the mag-
netic field is identical on each section of the transformer, along all the net
iron length LFe = kstkL, where kstk is the stacking factor and L is the total
length of the leg. Thus, the analysis is reduced to a two-dimensional problem
analysis.
To reduce mainly the field problem, the analysis of the transformer can be
carried out on a part of the section of Figure 6.1. In fact, both structures
feature geometric and magnetic symmetries. In the core-type transformer
drawn again in Figure 6.2(a), the flux lines must be symmetrical as regards
the axis BB′. Then the analysis can be carried out only on the part of the
structure drawn in Figure 6.2(b). Similarly, the segment AA′ is a symmetry
axis; thus the analysis is reduced to the section part illustrated in Figure 6.2(c).
On the symmetry lines, suitable boundary conditions have to be assigned,
as will be described in the following.
Similarly, in a shell-type transformer, the analysis can be simplified con-
sidering that the segments AA′ and BB′ are symmetry axes for both the

95
96 Electrical Machine Analysis Using Finite Elements

(a) (b)

FIGURE 6.1
Single-phase core-type (a) and shell-type (b) transformers.

B B B

A A' A A' A A'

B'

B' B'
(a) (b) (c)

FIGURE 6.2
Core-type transformer: simplification of the structure.

B B B

A A' A A' A A'

B'

B' B'
(a) (b) (c)

FIGURE 6.3
Shell-type transformer: simplification of the structure.

geometric structure and the flux lines. The section drawn in Figure 6.3(a)
can be reduced to the section part drawn in Figure 6.3(b) and then to that
drawn in Figure 6.3(c).

6.2 Equivalent Electric Circuit of the Transformer

In the following analysis a lossless transformer will be considered, whereas
winding resistances, iron losses, and saturation are not discussed.
The Single-Phase Transformer 97

The terminal voltage equations of such a transformer are given by

 dλ 1 di1 di 2
 v1 = dt = L1 dt + M dt

 (6.1)
 dλ 2 di di
v2 = = M 1 + L2 2
 dt dt dt

where v1, v2, i1, i2, λ1, and λ2 are the voltages, currents, and flux linkages at
the two terminal pairs. The sign notation is such as the positive direction of
the currents is against the voltage in both terminal pairs. Finally, L1, L2, and
M are the self- and mutual inductances of the transformer. With the assump-
tion that there is no saturation, these inductances are constant.
The flux linkages can be expressed as

 λ 1 = L1i1 + Mi 2

 (6.2)
λ 2 = Mi1 + L 2 i 2

For the sake of convenience, the self-inductances L1 and L2 are split into
two addenda, which are

L1 = L1σ + L1m
(6.3)
L2 = L2 σ + L2 m

so that the following relationship is satisfied:

L1m L2m = M2 (6.4)

The transformer formed by L1m, L2m, and M is an ideal transformer char-

acterized by a perfect coupling. The inductances L1σ and L2σ are the leakage
inductances referred to the primary and the secondary winding, respectively.
This split allows the transformer to be represented by means of the equiv-
alent electrical circuit shown in Figure 6.4, where a is the transformation
ratio, referred to the perfect link:

L1m M
a= = (6.5)
M L2 m

The voltage v′ and the current im are given by

 v′1 = a v′′2

 i2 (6.6)
 i m = i1 + a

98 Electrical Machine Analysis Using Finite Elements

L1σ a:1 L2σ

+ + im + +
i1 i2
L1m
v1 v'1 v"2 v2

_ _ _ _

FIGURE 6.4
Equivalent electric circuit of the real transformer, T representation.

L'2σ
a' : 1
+ + +
i1 i'm i2

v1 = v'1 L'1m v"2 v2

_ _ _

FIGURE 6.5
Equivalent electric circuit, Γ′ representation .

The choice of how the inductances L1 and L2 are separated in the two
addenda is arbitrary. It is sufficient that Equation (6.2) is satisfied and that
the inductances of the model are positive or, at least, null. The latter condition
limits the possible values of the transformation ratio:

M L
≤a≤ 1 (6.7)
L2 M

In the two limit cases, the following results are obtained:

1. With a′ = L1/M the primary leakage inductance is null, so that v1 =

v′1 . In this case, the equivalent circuit is shown in Figure 6.5, where

L′1m = L1

 L 1L 2 − M 2 (6.8)
L′2 σ =
 L1

2. With a″ = M/L2 the secondary leakage inductance is null, so that

v2 = v″2 . In this case, the equivalent electric circuit is shown in Figure
6.6, where
The Single-Phase Transformer 99

L"1σ
a" : 1
+ +
i1 i"m i2 +

v1 v'1 L"1m v"2 = v2

_ _ _

FIGURE 6.6
Equivalent electric circuit, Γ″ representation.

 M2
L′′1m =
 L2
 (6.9)
L′′ = L1L 2 − M
2

 1σ
L2

Another common choice is that considering the transformation ratio equal

to the ratio between the coils of the transformer, which is

N1
a= (6.10)
N2

With this choice, from Equation (6.5) it results that

2
L 1 m  N1 
= (6.11)
L 2 m  N2 

Of course the equivalent electric circuit is that in Figure 6.4. As will be

hereafter highlighted, the choice of this transformation ratio is useful to
impose the current sources in the finite element analysis of the leakage
inductances.
When the parameters of the equivalent circuit of Figure 6.4 are known, it
is possible to modify the circuit of the transformer using one of the two Γ
representations of Figure 6.5 or Figure 6.6, by means of the simple relation-
ships reported in Table 6.1.

6.3 Computation of the No-Load Inductances

In order to compute the magnetizing inductances, a simulation is carried
out that corresponds to the no-load test of the transformer. In the test, the
100 Electrical Machine Analysis Using Finite Elements

TABLE 6.1
Transformations of the Inductive Parameters of the Transformer
From T to Γ′ Representation From T to Γ″ Representation

L 1 m2
L′1 m = L 1 σ + L 1 m L′′1 m =
L1m + a2 L 2 σ

 L1m   L1σ + L1m 

2
a2 L 2 σ L1m
L′2 σ =  L 2 σ +
 2    −
1
2
(L 1σ
+ L1m ) L′′1 σ = L 1 σ +
a2 L 2 σ + L 1 m
a L1m a

primary winding is fed at the nominal frequency, with voltage equal to the
nominal voltage, measuring the primary current and the induced secondary
voltage. Similarly, in the simulation, the primary winding current is imposed
as the field problem source and the corresponding voltages are computed
in the two windings. It is also possible to compute the variation of the
induced voltage at no-load, corresponding to different primary current val-
ues. Of course, this makes sense when iron saturation occurs.

6.3.1 Statement of the Problem

The magnetic energy is mainly stored in the iron core; thus in the analysis
of the no-load operations, it is convenient to increase the mesh in the core
of the transformer.
Each part of the coil of the primary winding, concerning the considered
section of Figure 6.2(c) or Figure 6.3(c), is composed by a number of conductors:

N1/4, in case of core-type transformer, or

N1/2, in case of shell-type transformer.

A current source I1o is forced in the primary winding. For the sake of
simplicity, the latter is simplified as a unique conductive bar, as shown in
Figure 6.2(c) and Figure 6.3(c), fed by the total current:

N1I1o/4, in case of core-type transformer, or

N1I1o/2, in case of shell-type transformer

We must not forget that the current flows in the conductors and then a
free distribution of the current in the equivalent bar is not realistic. It has to
be imposed that the current is uniformly distributed in the bar. Of course,
if the current assigned in the simulation is constant, a magnetostatic problem
arises; then the problem of a nonuniform distribution of the current does
not exist. This kind of simulation should be preferred. In this case, the
constant current I1o denotes the maximum value of the sinusoidal current
waveform.
The Single-Phase Transformer 101

The secondary winding is open-circuited, and a null conductivity is

assigned to its corresponding equivalent bar.
At the beginning, the iron is assumed to be linear, with a constant relative
permeability µFe. Some considerations of nonlinear characteristics of the iron
will be reported in the following.
In order to assign correctly the boundary conditions, it is useful to estimate
the flux lines of the transformer. It is possible to believe that the flux density
vector B has only normal component along the symmetry axis AA′ (see Figure
6.2 and Figure 6.3). Moreover, in case of the core-type transformer, the flux
density vector B has only a normal component along the symmetry axis BB′
as well. Along these lines, Neumann’s boundary conditions are assigned.
In case of the shell-type transformer, the flux density vector B has only a
tangential component along the symmetry axis BB′. Along this line,
Dirichlet’s boundary condition is assigned, which is a constant value of the
magnetic vector potential, i.e., Az = 0. Along the other lines that delimit
externally the analysis domain, it is possible to assign a null magnetic vector
potential, i.e., Az = 0.

6.3.2 Computation on the Solved Structure

Once the problem is solved, the following magnetic quantities are computed.

6.3.2.1 Magnetic Flux Density

The components of the magnetic flux density in the iron core are obtained
by deriving the magnetic vector potential.

6.3.2.2 Magnetic Energy

The stored magnetic energy, corresponding to the current feeding the pri-
mary winding, can be computed as follows. Assuming linear materials, the
magnetic energy density is

1 1
wm = B ⋅ H = µH 2 (6.12)
2 2

Then the magnetic energy stored in the whole transformer is computed

by integrating the energy density over all the volume. Since a 2D problem
is analyzed, the result is obtained by means of the surface integral, multiplied
by the net iron length of the transformer LFe. Finally, the result has to be
multiplied by 4, since only a quarter of the transformer is analyzed. Then
the magnetic energy is

∫
1
Wm = 4L Fe µH 2 dS (6.13)
S 2
102 Electrical Machine Analysis Using Finite Elements

where S is the total surface of the domain, corresponding to Figure 6.2(c) or

Figure 6.3(c).
As an alternative, the magnetic energy can be computed as the energy that
is provided to the system by the source, as

∫ ∫
1 1
Wm = 4L Fe J ⋅ A dS = 4L Fe J z A z dS (6.14)
S 2 S 2

6.3.2.3 Main Flux

The main flux is obtained by integrating the component of the flux density
that is normal to the core section over the same surface. Since a 2D problem
is analyzed, the flux density is integrated along a line that crosses the iron
core, and then the result is multiplied by the net iron length LFe. It results that

∫
Φo = L Fe B ⋅ n dl ,
l
in case of core-type transformer; (6.15)

∫
Φo = 2 L Fe B ⋅ n dl ,
l
in case of shell-type transformer.

The ends of the lines are generally chosen within the windings.

6.3.2.4 Flux Linkage

The primary winding flux linkage may be computed by multiplying the
main flux by the number of turns of the winding:

Λ1o = N1 Φo (6.16)

A more reliable computation of the flux linkage is the integration of the

magnetic vector potential Az over the surfaces of the equivalent bars forming
the primary winding. Let SCu1+ be the surface of the bar where the current
direction is positive, and SCu1– be the surface of the bar where the current
direction is negative. Of course, the two surfaces are equal, that is SCu1+ =
SCu1– = SCu1. It results that

1  
Λ1o = N1L Fe
S Cu1  ∫ S Cu 1+
A z dS −
∫S Cu 1−
A z dS  in case of core-type transformer;

(6.17)
1  
Λ1o = 2 N1L Fe
S Cu1  ∫ S Cu 1+
A z dS 

in case of shell-type transformer.
The Single-Phase Transformer 103

The same computation is carried out for the secondary winding. The
relationships are the same, only subscript 2 will be used in place of 1.

6.3.2.5 Induced EMF

The EMF induced in the two windings are obtained by multiplying the flux
linkages by the electric frequency ω = 2πf, at which the transformer works.
It is obtained that

E1o = ω Λ1o
(6.18)
E2o = ω Λ 2o

We wish to point out that the simulation is carried out considering the
maximum value of the sinusoidal current, say I1o. Consequently, the quan-
tities Φo, Λ1o, and E1o, as well as the secondary winding quantities represent
the maximum value that is reached. Since the problem has been assumed to
be linear, the RMS values of these quantities are obtained by dividing the
maximum value by 2. This is not possible when the iron saturation occurs.

6.3.2.6 Self- and Mutual Inductances

The primary self-inductance, that is the no-load inductance of the trans-
former, is obtained from the first part of Equation (6.2), as the ratio between
the flux linkage Λ1o and the no-load current I1o. The mutual inductance is
obtained from the second part of Equation (6.2), as the ratio between the
flux linkage Λ2o and the no-load current I1o. They are

Λ1o
L1 =
I1o
(6.19)
Λ
M = 2o
I1o

Another way to compute the primary self-inductance is from the magnetic

energy, as

2 Wm
L1 = (6.20)
I1o 2

For obtaining the self-inductance L2, the secondary winding is fed by the
current I2o, and the primary winding is open-circuited (that is, it assigns a
null conductivity of the primary winding equivalent bars). Then L2 is given
by the ratio between the secondary flux linkage Λ2o and the current I2o.
104 Electrical Machine Analysis Using Finite Elements

6.3.2.7 Joule Power Losses

The Joule power losses of the coils and the corresponding resistances should
be computed by the classical analytical relationships, since the coils are
modeled as equivalent conducting bars and the simulations are carried out
with stationary current.
On the contrary, if the non-uniform current distribution in the winding
due to the source frequency has to be estimated, a different procedure must
be adopted. The coil is designed in detail, i.e., each turn is drawn in its actual
position. The simulation is carried out at the source frequency, always with
open-secondary winding (null conductibility). Then the Joule losses are com-
puted in the primary winding.
However, in low-power transformers, which work at the industry fre-
quency, the many turns of small diameter so that the non-uniform current
distribution is negligible.

6.3.3 Effect of the Nonlinear B-H Curve

When the flux-density in the iron is quite high, it is no longer possible to
consider a constant permeability, but it is necessary to consider the whole
B-H curve of the material. Neglecting the magnetic hysteresis, the first mag-
netization B-H curve is adopted, obtained by the dc test on the material.
A fixed number of simulations is carried out. The values of flux linkage
Λ1o are computed at different values of the current I1o. Then the flux linkage
Λ is drawn as a function of the current I, as shown in Figure 6.7. This curve
is called the dc magnetizing characteristic.
From this characteristic the magnetizing characteristic of the transformer
can be computed, as described in the following. A voltage is fixed, i.e., a flux

Sinusoidal Λ dc magnetizing Magnetizing

flux linkage characteristic characteristic
E
E 1o (rms)

ω Λ1o

dc magnetizing
I characteristic

I1o I1o(rms)
Corresponding
current

(a) (b)

FIGURE 6.7
Drawing of the current waveform (a) and the magnetizing characteristic of the transformer (b).
The Single-Phase Transformer 105

linkage, with a sinusoidal waveform. The corresponding waveform of the

current is achieved, by points, from the dc magnetizing characteristic. This
is shown in Figure 6.7(a).
The RMS value of the current waveform is computed, and the relationship
between E10(rms) e I10(rms) is achieved. Such a correspondence is less affected
by saturation, since the RMS value is computed taking into account the
whole current waveform. Such a characteristic is reported in Figure 6.7(b).
A simple algorithm for the computation of the characteristic is given at the
end of this chapter.
It is worthwhile to observe that the coil resistance is not taken into account.
However, the voltage drop on the coil resistance may be high, especially in
the low-power transformer. If the primary winding resistance is not negli-
gible, a more accurate computation has to be carried out. The dc magnetizing
characteristic of Figure 6.7(a) has to be modified by including the effect of
the voltage drop on the resistance.

6.3.4 Estimation of the Iron Losses

Let ps,Fe be the specific losses of the iron core, measured in (W/kg), corre-
sponding to a flux density of sinusoidal waveform, of amplitude BM = 1 T
and frequency f = 50 Hz. Then the corresponding losses in an infinitesimal
iron volume dτ, where the maximum value of the flux density is given by
BM, are computed by using Steimnetz’s relationship:

dpFe = ps ,Fe B M α γ Fe dτ (6.21)

where γFe is the specific weight of the iron(more or less γFe = 7700 kg/m3).
The value BM must be the maximum flux density amplitude, reached in the
volume dτ. This is the reason for using the maximum value of the current
I1o in the (magnetostatic) simulation. The power α of BM in Equation (6.21),
which ranges from 1.6 to 2.2 in Steimnetz’s formula, can be approximated
by 2.
Then the iron losses are given by

pFe =
∫
τ Fe
ps ,FeB M 2 γ Fedτ = 4L Fe
∫
S Fe
ps ,FeB M 2 γ FedS (6.22)

If the transformer works at a frequency f different from 50 Hz, the iron

losses have to be multiplied by (f/50)2.

6.3.5 Example
Let us consider a shell-type transformer, characterized by the following
nominal data:
106 Electrical Machine Analysis Using Finite Elements

25

125 71 75

25

25 25 50 25 25
150

20

50

FIGURE 6.8
Main dimensions of the transformer.

Sn = 200 VA — rated power, at the secondary in continuous duty operation

V1n = 220/127 V — primary voltages
V2n = 24/18 V — secondary voltages
f = 50 Hz — frequency
conductor insulation in E class
magnetic material TERNI 1350, with specific iron losses ps,Fe = 1.3 W/kg
(at BM = 1 T and f = 50 Hz)
rated magnetic flux density in the core BM ≈ 1.1 T

The aspect and the dimensions of the iron core are reported in Figure 6.8.
The rated currents of the secondary winding are

Sn
I2 n = = 8.33 A , at V2n = 24 V
V2 n

Sn
I2 n = = 11.1 A , at V2n = 18 V
V2 n

Assuming an efficiency η = 90% and a unity power factor, the rated currents
of the primary winding are

Sn
I1 n = = 1 A , at V1n = 220 V
ηV1n
The Single-Phase Transformer 107

Sn
I1 n = = 1.74 A , at V1n = 127 V
ηV1n

Such values have been used for the choice of the diameters of the coils,
fixing a current density about 2 A/mm2. Each coil is characterized by three
terminals, as shown in Figure 6.9. The number of conductors have been
chosen according to the iron flux density BM = 1.1 T in the iron core. In
addition, in the choice of the number of secondary winding turns, a 4%
voltage drop under full load has been assumed. With reference to Figure 6.9,
the following number of turns and diameters have been chosen for the
secondary winding:

N2′ = 35 turns, with dCu2′ = 2.938 mm

N2″ = 12 turns, with dCu2″ = 2.488 mm
N2 = 47 total turns

Assuming a 0.1 mm insulation thickness, the total dimension of the wind-

ing results:
2 layers with dCu2′ thickness 6.1 mm
1 layer with dCu2″ thickness 2.6 mm
thickness 8.7 mm

The following number of turns and diameters have been chosen for the
primary winding:

N1′ = 240 turns, with dCu1′ = 1.153 mm

N1″ = 176 turns, with dCu1″ = 0.885 mm
N1 = 416 total turns

The dimension of the primary winding results as follows:

5 layers with dCu2′ thickness 6.3 mm

3 layers with dCu1″ thickness 3.0 mm
thickness 9.3 mm

+ +
N1' + +
V1'
N2' V2'
_ _
V1 V2

N1'' N2''
_
_

FIGURE 6.9
Sketch of the transformer windings.
108 Electrical Machine Analysis Using Finite Elements

75

62.5

35 37.5

8.7 9.3
1.5 0.5

FIGURE 6.10
Section of the transformer that is used in the following simulations.

Finally, an insulation reel with a thickness of 1.5 mm, an insulation between

the primary and secondary winding with a thickness of 0.5 mm, and an
external insulation with a thickness of 0.5 mm have been adopted. The total
thickness of the windings and the corresponding insulation is then equal to
18.7 mm. In Figure 6.10 there are the dimensions of the two coils in a quarter
of the section of the transformer. The simulations are carried out referring
to this section part.

6.3.5.1 Computation with Linear Iron

Let us assume

• a constant current in the primary winding equal to I1o = 0.1 A, and

null current in the secondary winding. In the equivalent conductive
bar results N1I1o/2 = 20.8 A;
• a constant relative permeability of the iron equal to µFe = 5500.

Figure 6.11 shows the flux plot, obtained by drawing the equipotential
lines of the magnetic vector potential.
Figure 6.12 shows the flux density along the line AA″ (see Figure 6.11)
inside the magnetic iron. The average value of the flux density is approxi-
mately equal to 1.125 T. The flux density is not exactly constant, due to the
different length of the magnetic paths within the iron core. The total magnetic
energy stored in the transformer, using Equation (6.13) or Equation (6.14),
results in

Wm = 58.46 mJ
The Single-Phase Transformer 109

SCu2 SCu1

A A" A'
N1 I1o
2

FIGURE 6.11
Flux plot.

1.15

1.14
Flux density (T)

1.13

1.12

1.11

1.1
0 5 10 15 20 25
(A) Length (mm) (A")

FIGURE 6.12
Flux density within the iron core.

The integral of the magnetic energy density only over the iron core results
in

Wm(Fe) = 58.42 mJ

which represents almost the totality of the stored magnetic energy. Con-
versely, by applying Equation (6.14) to the iron core, the integral results null.
In fact this equation represents the energy of the transformer as work fur-
nished by the external sources. It is different from zero only in the coils of
the primary winding, where the current density is not null.
From the second part of Equation (6.15), the magnetic flux is

Φo = 2.7952 mWb

Then, from Equation (6.16), with N1 = 416 and N2 = 47, the flux linkages are

Λ1o = 1.1628 Vs
110 Electrical Machine Analysis Using Finite Elements

Λ2o = 0.1314 Vs

If Equation (6.17) is directly applied, where SCu1 = 3.255·10–4 m2 and SCu2 =

3.045·10–4 m2, one obtains

Λ1o = 1.1696 Vs

Λ2o = 0.1321 Vs

which differ from the previous values just of 0.5%. Finally the voltage
induced in the windings are

E1o = 367.4 V

E2o = 41.5 V

Since the system has been supposed to be linear, the RMS values of these
voltages result in

E1o(rms) = 220 V

E2o(rms) = 24.85 V

The self-inductance of the primary winding and the mutual inductance,

computed using Equation (6.19), are

L1 = 11.69 H

M = 1.321 H

From Equation (6.20), the same value for L1 is computed.

6.3.5.2 Computation with Nonlinear Iron

The B-H curve of the TERNI 1350 silicon-steel laminations is now adopted,
with a thickness of 0.5 mm and specific iron losses of ps,Fe = 1.3 W/kg (at
1 T and 50 Hz), which is not a grain-oriented steel. In the low-power trans-
former, grain-oriented laminations are not used both to reduce the cost and
to simplify the construction (E-shape or U-shape laminations are often used).
Table 6.2 reports the values of the MMF N1I1o/2, the iron flux density BFe,
the primary flux linkage Λ1o, and the induced voltage E1o related to the
various primary currents I1o. The magnetizing characteristic is shown in
Figure 6.13, highlighting the computed values by dots. In the same figure,
the magnetizing characteristic of the transformer is reported, obtained as
The Single-Phase Transformer 111

TABLE 6.2
Values Corresponding to the Field Analysis
I1o N1I1o/2 BFe Λ1o E1o
(mA) (A) (T) (Vs) (V)
25.0 5.2 0.36 0.3774 118.5
50.0 10.4 0.73 0.7542 237.0
75.0 15.6 0.94 0.9780 307.2
100.0 20.8 1.05 1.0977 344.9
112.5 23.4 1.09 1.1361 356.9
125.0 26.0 1.12 1.1684 367.1
137.5 28.6 1.15 1.1976 376.2
150.0 31.2 1.17 1.2202 383.3

(V) Magnetizing curves

400
Magnetizing curve
350 2 I1o(rms) vs 2 E1o(rms)
300
250
dc magnetizing curve
200
150
100
50
0
0 20 40 60 80 100 120 140
Current (A)

FIGURE 6.13
Current-voltage characteristics.

described in Section 6.3.3. Figure 6.14 shows the sinusoidal voltage wave-
form, corresponding to the rated value of E1o(rms) = 220 V, and the correspond-
ing current waveform. The distortion of the current from the sinusoidal
waveform is noticeable.

6.3.5.3 Effect of the Equivalent Air-Gap

To simplify the assembly of the low-power transformers, E-shape lamina-
tions are commonly used, as shown in Figure 6.15(a). The dimensions of the
lamination are defined by the standardization UNEL 82611, identified from
dimension C of the central leg.
In order to consider the small air-gap that occurs between the two parts
of the core, some analyses have been carried out with different values of the
air-gap. Since the symmetry between the upper and lower parts disappears,
112 Electrical Machine Analysis Using Finite Elements

Voltage (V) Current (mA)

350 87.5
Voltage
300 E 1o(rms) = 220 V 75.0

250 62.5
Current
200 50.0

150 37.5

100 25.0

50 12.5

0 0
0 2 4 6 8 10 12 14 16 18 20
Time (ms)

FIGURE 6.14
Voltage and current waveforms corresponding to the rated voltage E1o(rms) = 220 V.

C D F

FIGURE 6.15
Standardized lamination for a low-power shell-type transformer (a) and flux plot in half a
section of the transformer (b).

half a section of the transformer is considered, as illustrated in Figure 6.3(b).

The magnetic flux plot is shown in Figure 6.15(b).
Some magnetic quantities, corresponding to different values of the air-gap
thickness, are reported in Table 6.3. All the simulations are carried out with
µFe = 5500, N1I1 = 208 A, and I2 = 0. From Table 6.3 the effect of the air-gap
thickness is evident, even though it assumes low values.
The Single-Phase Transformer 113

TABLE 6.3
Effect of the Air-Gap on the Magnetic Quantities

g BFe
∫ S Cu 1+
A z dS
∫
S Cu 2 +
A z dS
∫ J A dS
S
z z

(mm) (T) (Vs/m) (Vs/m) (J/m)

0.05 2.13 17.32·10 –6 16.20·10 –6 11.074
0.10 1.19 9.674·10–6 9.043·10–6 6.1816
0.15 0.83 6.745·10–6 6.0303·10–6 4.3100

6.4 Determination of the Leakage Inductances

The leakage inductances L1σ and L2σ are determined from the short-circuit
test of the transformer, feeding the primary winding and connecting the
secondary winding in short circuit. The test is carried out at reduced voltage
and with currents almost equal to their rated value. It follows that the iron
core works at low magnetic flux density. It is commonly assumed that the
iron works in the linear part of its B-H curve.
As regards the equivalent electric circuit of the transformer, Figure 6.4 is
considered, with transformation ratio equal to the turn ratio, i.e., a = N1/N2
as in Equation (6.10). The simulation is carried out by forcing both winding
currents, that is, the secondary current is chosen according to the primary
current so that the two MMFs result in equal and opposite, i.e., N1i1 = –N2i2.
Then, from the second part of Equation (6.6), the magnetizing current im is
null, because of

i m = i1 +
i2
= i1 +
(
− N1 / N 2 i 1
=0
) (6.23)
a N1 / N 2

Thus the two flux linkages (6.2) correspond to the leakage flux linkages

 λ 1σ = L1σ i 1

 (6.24)
λ 2 σ = L 2 σ i 2

The current induced in the secondary winding is such as to have a null

main flux, so that only leakage magnetic fluxes appear in the transformer.
In practice, the assumption of the transformation ratio a = N1/N2 is unnec-
essarily restrictive in order to have a null magnetizing current. In fact, to
satisfy this condition, it is sufficient that i2 = –i1/a, for each value of a. The
114 Electrical Machine Analysis Using Finite Elements

choice of the transformation ratio equal to the turn ratio is particularly

convenient, since this allows the two coil currents to be easily set in the
simulation, and a physical meaning to be given to the flux linkages obtained
in the solved structure.
Once the leakage inductances L1σ and L2σ are obtained, the total leakage
inductance is computed, referring to the equivalent electric circuit in the Γ′
representation (Figure 6.5) or in the Γ″ representation (Figure 6.6).

6.4.1 Statement of the Problem

As far as the geometric symmetry of the structure is concerned, the consid-
eration given above can be repeated: the simulations can be carried out on
a quarter of the transformer section.
As far as the boundary conditions are concerned, they depend on the
transformer type. In the shell-type transformer (see Figure 6.3), the flux
density vectors have a normal component only along the symmetry line AA′,
and a tangential component only along the symmetry line BB′. Then the
boundary conditions remain the same as those assigned in the no-load sim-
ulation (Neumann’s condition in AA′ and Dirichlet’s condition in BB′).
In the core-type transformer, the flux density vectors have a normal com-
ponent only along the symmetry line AA′, and a tangential component only
along the symmetry line BB′, exactly as for the shell-type transformer. Then
the boundary conditions must be modified with respect to those assigned
at the no-load.
In this analysis, since only leakage fluxes occur, the flux lines are mainly
localized in the air and in the coils. It is appropriate to increase the mesh of
the domain in the coils and in the surrounding space.
The iron is assumed to be linear with a constant magnetic permeability µFe.
A magnetostatic simulation is conveniently carried out. Constant currents
are forced in the two windings: I1 in the primary coil and I2 = –I1(N1/N2) in
the secondary coil. Such currents are fixed to their rated value, even though,
since the problem is linear, the current value does not matter.
As in the no-load simulation, conductive bars are used, equivalent to the
whole coil, as indicated in Figure 6.2(c) and Figure 6.3(c), that are fed by the
total current. Since the field quantities are not time-dependent, the current
is uniformly distributed in the surface of the equivalent bar. Some con-
siderations about simulations with eddy currents will be described in
Section 4.3.
The current in the two parts of the primary and secondary coils are
imposed as follows. With a core-type transformer, they are

N1I1sc/4, in the primary coil, and

N2I2sc/4 = –N1I1sc/4, in the secondary coil
The Single-Phase Transformer 115

while, with a shell-type transformer, they are

N1I1sc/2, in the primary coil, and

N2I2sc/2 = –N1I1sc/2, in the secondary coil.

6.4.2 Computation on the Solved Structure

Once the field problem is solved, the magnetic quantities are computed as
follows.

6.4.2.1 Magnetic Flux Density

The components of the magnetic flux density are evaluated by deriving the
magnetic vector potential along the coordinate axes, as in Equation (3.66).

6.4.2.2 Magnetic Energy

The magnetic fields, corresponding to the assigned currents, are mainly
localized in the insulating space between the two windings and the two
windings themselves. Remembering that the material has been assumed to
be linear, the stored magnetic energy is computed by integrating the mag-
netic energy density (6.12) over all the volume of the transformer. Since the
problem is 2D and only a quarter of the section has been analyzed, the
magnetic energy results in

∫
1
Wm = 4l average µH 2 dS (6.25)
S 2

where S is the total surface of the domain and laverage indicates the average
length of a conductor (half a length of one turn). Let us highlight that the
effect of the end windings has been considered the same of the part of the
winding in the section under study. Such an approximation is reasonable,
since the fields are mainly localized between the two coils.
Alternatively, the magnetic energy is computed as the energy that is fur-
nished from the source to the system, as

∫
1 1
Wm = 4 l average J1z A z + J 2 z A z dS (6.26)
S 2 2

6.4.2.3 Flux Linkage

The flux linked with the primary winding is computed by integrating the
magnetic vector potential Az over the surface of the equivalent bar repre-
senting the primary coil. It results that
116 Electrical Machine Analysis Using Finite Elements

1  
Λ1sc = N1l average
S Cu1  ∫ S Cu 1+
A z dS −
∫
S Cu 1−
A z dS  , with core-type transformer


1  
Λ1sc = 2 N1l average
S Cu1  ∫ S Cu 1+
A z dS  , with shell-type transformer

(6.27)

Analogously for the secondary winding, substitute the subscript 2 for

subscript 1 in Equation (6.27).

6.4.2.4 Induced EMF

The EMFs induced in the two windings are obtained by multiplying the flux
linkages by the electric frequency ω. The EMFs are given by

E1sc = ω Λ1sc
(6.28)
E 2 sc = ω Λ 2 sc

Since the field problem is linear, the quantities I1sc, I2sc, Λ1sc, Λ2sc, and E1sc,
E2sc are in proportion. The maximum values are achieved by multiplying the
RMS values by the square root of two.

6.4.2.5 Leakage Inductances

The short-circuit inductances are obtained by dividing the flux linkages by
the respective winding currents. Since a = N1/N2 and N1I1sc = –N2I2sc, so that
Im = 0 from Equation (6.6), then Λ1sc = L1σI1sc and Λ2sc = L2σI2sc. These equations
are obtained by examining the electric circuit in Figure 6.4 and substituting
Equation (6.3) and Equation (6.23) in Equation (6.2) together with Equation
(6.11). In conclusion, it results that

Λ1sc
L1σ =
I1sc
(6.29)
Λ
L2 σ = 2 sc
I 2 sc

An alternative way to compute such inductances is to use the energy

furnished by the two sources. In the core-type transformer they are

2   
∫ ∫
1 1
L1σ = 2  average 
4l J1z A z dS − J1z A z dS  
I1sc   Scu 1+ 2 Scu 1− 2 
(6.30)
2   
∫ ∫
1 1
L2 σ = 2  average 
4l J 2 z A z dS − J 2 z A z dS  
I 2 sc   Scu 2 + 2 Scu 2 − 2 
The Single-Phase Transformer 117

while in the shell-type transformer they are

2  
∫
1
L1σ = 4l average J1z A z dS 
I1sc 2  Scu 1 2 
(6.31)
2  
∫
1
L2 σ = 4l average J 2 z A z dS 
I 2 sc 2  Scu 2 2 

that correspond to those given in Equation (6.29), observing that in the core-
type transformer

N1 I 1 N2 I 2
J 1z = J2z =
4S Cu1 4S Cu 2

and in the shell-type transformer

N1 I 1 N2 I 2
J 1z = J2z =
2 S Cu1 2 S Cu 2

The total leakage inductance L′2σ (or alternatively L″1σ) is obtained, apply-
ing the equivalences of Table 6.1.

6.4.2.6 Electric Forces

The forces on the coils are evaluated by using Maxwell’s stress tensor.
Because of the 2D problem, the coil is enclosed by a simple line. The tan-
gential and normal forces to each line are computed as

Ft = l average
∫ (µ H H ) dl
l
o t n

(6.32)

∫ ( )
µo
Fn = l average H n 2 − H t 2 dl
l 2

The x-axis force is the sum of the forces Ft of the vertical lines with the
forces Fn of the horizontal lines. Similarly, the y-axis force is the sum of the
forces Ft of the horizontal lines with the forces Fn of the vertical lines.
Due to the symmetry with respect to the axis BB′ that has been used in
the simulations, the forces refer to half a coil, and then the overall y-axis
force component is null while the overall x-axis force component is obtained
by doubling the force obtained in half a winding.
Alternatively, the force on the coils can be computed as Lorentz’s force.
According to Lorentz’s specific electric force (1.84) and the definition of the
118 Electrical Machine Analysis Using Finite Elements

current density vector (1.60), the x-axis and the y-axis that force components
in the half coil are

Fx = − l average
∫
S Cu 1
J zB y dS
(6.33)
Fy = l average
∫ S Cu1
J zB x dS

6.4.3 Simulation with Induced Currents

Instead of magnetostatic simulations, a variable field problem may be ana-
lyzed. The currents in primary and secondary windings are not forced simul-
taneously: a periodic sinewave current i1(t) is forced only in the primary
winding, while the current i2(t) results as the solution of the field problem.
It is a current induced in the secondary winding. The waveform of the two
currents is sinusoidal with a constant frequency, and the complex phasor
notation is adopted.
Such an approach is not recommended, however, when it is used it is
indispensable to avoid some possible errors. Some remarks are hereafter
underlined.

1. If the secondary coil is modeled by the equivalent conductive bar,

and eddy currents are computed in this bar, then the distribution of
the current in the bar is not uniform. In fact, the current are nonuni-
formly distributed as the bar size is large and the frequency is high.
To have a correct result, it must be imposed that the induced current
in the equivalent bar of the secondary winding be uniformly distrib-
uted over the whole section. It is not easy to assign such a constraint,
if not impossible.
Of course the same problem occurs for the equivalent bar of the
primary winding. However, since the primary current is a current
source, the current density can be defined to be uniformly distrib-
uted over the bar.
2. In order to obviate the problem described in the point 1, the second-
ary winding may be subdivided in the effective number of conductors.
However, because of the 2D problem, without a specific constraint
the induced current in each conductor may be different. It is neces-
sary to impose the constraint of equal current in each conductor that
forms the coil (series conductors). Unfortunately, it is not always
possible to assign this constraint.
3. With a core-type transformer, and the adopted simplifications, the
solution of the variable field problem is carried out as though the
secondary conductors are distinguished with a closing path at
The Single-Phase Transformer 119

the infinite. This is because of the 2D problem. Then, the bar Scu2+
links the flux produced by the primary currents, while Scu2– does not
link any flux, so that only Scu2+ is interested by eddy currents. It is
thus compulsory to impose that the sum of the currents induced in
Scu2+ and Scu2– be null, which is Icu2– = –Icu2+. A method to impose this
condition is to declare that the bars Scu2+ and Scu2– are interested by
a total current in any time. This is equivalent to imposing that Icu2+ +
Icu2– = 0.

6.4.4 Example
Let us consider the transformer described in Section 3.6, with the aim of
computing the leakage inductances.
The iron relative permeability is fixed to the constant value µFe = 5500.
The current I1sc = 1 A is forced in the primary winding, and then the total
current in the equivalent primary bar of the analyzed section part is

N1I1sc
= 208 A
2

and the total current in the equivalent secondary bar is

N2 I 2 sc NI
= − 1 1sc = −208 A
2 2

which is a current I2sc = –8.851 A. The flux plot is shown in Figure 6.16.
The flux density within the windings assumes the behavior of Figure 6.17(a)
with a maximum value B = 7.44 mT. This is in accord with the theoretical
behavior, as reported in Figure 6.17(b). For the sake of comparison, the
maximum flux density computed analytically results in

FIGURE 6.16
Flux plot during short-circuit operation.
120 Electrical Machine Analysis Using Finite Elements

10−3 Flux density (T)

8
B
7
6
5 N1I1sc
µo
4 hcoil
3
s2 s s1
2
1
0 s1 = 8.7 mm
25 30 35 40 45 50 s = 0.5 mm
Displacement (mm) s2 = 9.3 mm
hcoil = 70 mm
(a) (b)

FIGURE 6.17
Flux density behavior within the two windings.

N1I1sc
B = µo = 7.47 mT
h coil

which corresponds to the value computed from the field solution.

The length laverage is computed as the semi-circumference in the middle of
the two coils. According to Figure 6.8 and Figure 6.10, laverage = 132.8 mm.
The stored magnetic energy, computed by Equation (6.25), is

Wm = 2.547 mJ

and results are stored almost completely in the primary winding (45.59%),
secondary winding (42.87%), and the surrounding space (11.53%). In the
iron, it is lower than 0.01%.
From Equation (6.26) it results always that

Wm = 2.547 mJ

whose 0.9766 mJ in the primary winding and 1.5704 mJ in the secondary

winding.
The flux linkages, from Equation (6.27), are

Λ1sc = 1.9533 mVs

Λ2sc = –0.35488 mVs

The Single-Phase Transformer 121

The corresponding leakage inductances (6.29) are

L1σ = 1.9533 mH

L2σ = 0.040076 mH (that is, 3.1396 mH if referred to the primary winding)

If the total leakage inductance referred to the primary winding is approx-

imated by
2
N 
L1σ +  1  L 2 σ = 5.093 mH
 N2 

it is noted that it corresponds to that obtained as

2 Wm
Lσ = = 5.094 mH
I12

The same inductance, computed analytically, is given by

N1 2 s s 
Lσ = µo 2 l average  1 + s + 2  = 5.365 mH
h coil 3 3

The force on the coils is computed by means of Maxwell’s stress tensor.

The overall y-axis component of the force (which is in the direction parallel
to the axis of the leg) is null, for symmetry reasons. On the contrary, each
half a coil is compressed. The value of the compression force on the consid-
ered half a coil is Fy = 2.78 mN.
The overall x-axis component of the force (which is in the direction normal
to the axis of the leg) results in

Fx = 0.391 N

The x-axis force computed analytically is given by

1 2 l average
( )
2
Fx = µo N1I1sc = 0.413 N
2 h coil

which is slightly higher than that obtained from the field solution.
Note: the computation of the x-axis force component on the secondary coil
is illustrated as follows. A line surrounding the coil is fixed as shown in
Figure 6.18. The x-axis force Fx, is computed in the different paths as
122 Electrical Machine Analysis Using Finite Elements

Q R

P S
x
Secondary coil

FIGURE 6.18
Line surrounding the secondary coil to compute the x-axis force.

BxHx − B yH y
FxPQ = 4l average
∫
l PQ 2
dl = 1.22 ⋅ 10−3 N

FxQR = 4l average
∫
l QR
B x H ydl = 6.64 ⋅ 10−3 N

BxHx − B yH y
FxRS = 4l average
∫
l RS 2
dl = 383.1 ⋅ 10−3 N

The y-axis force component may be computed in a similar way.

6.5 Algorithm for the Construction of the Magnetizing

Characteristic of the Transformer
A simple algorithm is described in the following, using the BASIC language,
for the construction of the magnetizing characteristic of the transformer. The
saturation is considered, using the dc magnetizing characteristic I1o – Λ1o.
The FUNCTION current(FlxLnk) returns the value of the current cor-
responding to the value of the flux linkage, given as input. It interpolates
the value along the dc magnetizing characteristic. For the sake of simplicity,
a linear interpolation is used. Better results can be used by means of higher-
order interpolations.

DECLARE FUNCTION current (FlxLnk)

CLS
CLEAR

CONST pi = 3.14156
a$ = CHR$(34)
DIM f(205), i(205), ang(205), v(205)

omega = 2 * 50 * pi ‘electric frequency

Fm = 600 ‘maximum value of flux linkage
The Single-Phase Transformer 123

Vm = Fm * omega ‘maximum value of voltage

‘ computation of the current waveform

Imax = 0
FOR k = 1 TO 201
theta = 2 * pi * (k - 1)/ 200
ang(k) = theta
v(k) = Vm * SIN(theta)
f(k) = Fm * COS(theta)
i(k) = current(f(k))
IF i(k) > Imax THEN Imax = i(k)
NEXT k

‘ computation of the rms values

isum = 0
vsum = 0
FOR k = 1 TO 200
isum = isum + i(k) ^ 2
vsum = vsum + v(k) ^ 2
NEXT k
Vrms = SQR(vsum/ 200)
Irms = SQR(isum/ 200)

A possible structure of the FUNCTION current(FlxLnk), supposing

Nb=9 points of the dc magnetizing characteristic and a linear interpolation,
is as follows. The control of the sign of the input is required if the given
characteristic contains only positive values of the current and the flux linkages.
Refer to Figure 6.13 and Table 6.2.

FUNCTION current(FlxLnk)

‘sign evaluation

sign = 1
IF FlxLnk < 0 THEN
sign = -1
FlxLnk = - FlxLnk
END IF

‘Definition of the dc magnetizing characteristic

‘given by a number of Nb points

Nb = 9
DIM FluxVect(Nb)
DIM CurrVect(Nb)

FluxVect(1) = 0: CurrVect(1) = 0
FluxVect(2) = 0.3774: CurrVect(2) = .025
FluxVect(3) = 0.7543: CurrVect(3) = .050
FluxVect(4) = 0.9780: CurrVect(4) = .075
FluxVect(5) = 1.0977: CurrVect(5) = .100
FluxVect(6) = 1.1361: CurrVect(6) = .1125
FluxVect(7) = 1.1684: CurrVect(7) = .125
FluxVect(8) = 1.1976: CurrVect(8) = .1375
FluxVect(9) = 1.2202: CurrVect(9) = .150
124 Electrical Machine Analysis Using Finite Elements

‘Interpolation (linear)

FOR i = 0 TO (Nb - 1)
IF (FlxLnk >= FluxVect(i) AND FlxLnk < FluxVect(i + 1)) THEN
Cf =(CurrVect(i+1)-CurrVect(i))/ (FluxVect(i+1)-
FluxVect(i))
CurrAux = Cf * (FlxLnk - FluxVect(i)) + CurrVect(i)
END IF
NEXT i

IF (FlxLnk >= FluxVect(Nb)) THEN

Cf=(CurrVect(Nb)-CurrVect(Nb-1))/(FluxVect(Nb)-
FluxVect(Nb-1))
CurrAux = Cf * (FlxLnk - FluxVect(Nb)) +
CurrVect(Nb)
END IF

‘Sign correction

current = CurrAux * sign

FlxLnk = FlxLnk * sign

END FUNCTION

References
1. E. Arnold, Die Wechselstromtechnick. Die Transformatoren, Vol. 2, Julius Springer,
Berlin, 1904.
2. R. Richter, Elektrische Maschinen. Die Transformatoren, Vol. 3, Julius Springer,
Berlin, 1932.
3. A.S. Longsdorf, Theory of Alternating Current Machinery, McGraw-Hill, New
York, 1937.
4. T. Bödefeld and H. Segueuz, Elektrische Maschinen, Springer-Verlag, Wren, 1942.
5. A. Carrer, Macchine Elettriche. Parte prima: Trasformatori, Editrice Universitaria
Levrotto & Bella, Torino, 1954.
6. S. Ramo, J.R. Whinnery, and T. Van Duzer, Fields and Waves in Communication
Electronics, John Wiley & Sons, New York, 1965.
7. G.R. Slemon, Magnetoelectric Devices: Transducers, Transformers and Machines,
John Wiley & Sons, New York, London, Sydney, 1966.
8. A.E. Fitzgerald, C. Kingsley, Jr., and S.D. Umans, Electric Machinery, McGraw-
Hill, New York, 1983.
9. I.J. Nagrath and D.P. Kothari, Electric Machines, Tata McGraw-Hill Publishing
Company Limited, New Delhi, 1985.
7
Single-Phase Variable Reactance

This chapter deals with the finite element analysis of a single-phase variable
reactance. In particular the dependence of the reactance on the air-gap length
is analyzed. The results obtained by the finite element method are compared
with those computed analytically.

7.1 The Single-Phase Variable Reactance

Figure 7.1 shows some structures of single-phase reactance. Figure 7.1(a)
shows a shell-type reactance with constant air-gap length; Figure 7.1(b) shows
a shell-type reactance with variable air-gap; Figure 7.1(c) shows a core-type
reactance with variable air-gap.
Reactances are used in electrical circuits to filter or limit the current. Variable
reactances are connected to capacitors to control the reactive power flow
and/or to obtain electrical resonance, principally in laboratory tests.
According to the rated power of the reactance, the iron core may be formed
by silicon-steel laminations or by grain-oriented laminations. Considering a
plane parallel to the plane of the laminations and neglecting the end effects,
the study can be reduced to a two-dimensional problem.
In addition, as for the single-phase transformer, the analysis may be
reduced to a part of the section of Figure 7.1, after recognizing the symmetry
axes of the structure and imposing along them the suitable boundary con-
ditions. For instance, the study of the reactances of Figure 7.1 can be reduced
to that of their parts reported in Figure 7.2.

7.2 Computation of the Reactance

The aim of the study is the determination of the value of the self-inductance.
If the reactance is characterized by a variable air-gap, it is also important to

125
126 Electrical Machine Analysis Using Finite Elements

(a) (b) (c)

FIGURE 7.1
Magnetic structures of single-phase reactance.

B B B

A A'
B'

B' B'
(a) (b) (c)

FIGURE 7.2
Reduction of the domain of the structures of Figure 7.1 using their symmetry.

study the dependence of such an inductance on the air-gap length. This

dependence is achieved by means of several field analyses with different
values of the air-gap. Furthermore, it is possible to assess the attractive forces
between the movable part and the fixed part.

7.2.1 Statement of the Field Problem

The magnetic field strength is mainly localized in the air part of the magnetic
path. Thus, it is convenient to increase the mesh in the air-gap of the reactance.
A magnetostatic simulation is carried out; a constant current I is forced in
the winding as a field source. The flux linkage and the voltage are computed
from the field solution. Once again, each section of the primary winding is
represented by means of a unique conductive bar, through which the fol-
lowing total current flows:

NtI, with shell-type reactance, analyzing the whole structure of Figure

7.1(a) or Figure 7.1(b), or the half of the structure shown in Figure 7.2(b)
NtI/2, with shell-type reactance, analyzing a quarter of the structure, as
shown in Figure 7.2(a), or with a core-type reactance, as in Figure 7.1(c)
or in Figure 7.2(c)
Single-Phase Variable Reactance 127

where Nt is the total number of turns and I is the winding current.

Initially, a linear magnetic material is considered, with a constant magnetic
relative permeability µFe. The effect of the nonlinearity will be investigated
later.
In order to assign the boundary conditions correctly, the field lines have
to first be estimated along the boundaries. Along the symmetry axis AA′,
the magnetic flux density vector B presents only a normal component: Neu-
mann’s boundary condition is assigned. Along the symmetry axis BB′, with
shell-type reactances the magnetic flux density vector B presents only a
tangential component: Dirichlet’s boundary condition is assigned. Con-
versely, with core-type reactances the magnetic flux density vector B presents
only a normal component: Neumann’s boundary condition is assigned again.
Along the other boundaries that delimit the domain (that are suitably distant
lines), a null magnetic vector potential is assigned.

7.2.2 Computation on the Analyzed Structure

As the field solution has been obtained, the magnetic quantities of interest
can be computed. In the following, the study will be focused on the structure
of Figure 7.1(b) only. Thus, in order to extend the analysis to the other struc-
tures or parts of structure, the following equations have to be rearranged
appropriately.
In addition, a rectangular leg is assumed, whose net iron length is equal
to LFe.

7.2.2.1 Magnetic Flux Density

The flux lines are obtained directly by the magnetic vector potential. The
components of the flux density can be assessed by deriving the magnetic
vector potential, as described in Equation (3.66).

7.2.2.2 Magnetic Energy

Because of the use of all linear materials, the magnetic energy density is
everywhere given by

1 1
wm = B ⋅ H = µH 2 (7.1)
2 2

The magnetic energy stored in the whole reactance is obtained by integrat-

ing the magnetic energy density on the whole volume of the reactance. In a
2D problem, this is equivalent to a surface integral multiplied by the length
LFe. If S is the total surface of the domain, it results in

∫
1
Wm = L Fe µH 2 dS (7.2)
S 2
128 Electrical Machine Analysis Using Finite Elements

Alternatively, the magnetic energy can be computed as

∫
1
Wm = L Fe J z A z dS (7.3)
S 2

which results different from zero only on the surfaces of the coils.

7.2.2.3 Flux Linkage

The flux linked with the winding is computed by processing the magnetic
vector potential Az on the surface of the coil. Let SCu+ be the section of the
equivalent bar in which the current is positive (with the same direction of
the z-axis, i.e., leaving the sheet) and SCu– be the section of the equivalent
bar in which the current is negative (with direction opposite to the z-axis,
i.e., going into the sheet). Of course the two values are the same and equal
to SCu. It results that

1  
Λ = Nt L Fe
S Cu  ∫
S Cu +
A z dS −
∫
S Cu −
A z dS 

(7.4)

7.2.2.4 Electrical Voltage

With the assumption of a sinusoidal waveform of the flux linkage, the RMS
value of the voltage on the reactance is obtained by multiplying the RMS value
of the flux linkage by the electrical frequency ω, which is

Erms = ω Λrms (7.5)

The voltage at the terminal pairs of the reactance coils is obtained as the
geometric sum of the voltage given in Equation (7.5) with the voltage drop
on the resistance, computed analytically.

7.2.2.5 Self-Inductance
The value of the self-inductance is achieved by dividing the flux linkage by
the corresponding current, which is

Λ
L= (7.6)
I

Alternatively, always in the linear case, the self-inductance can be achieved

from the magnetic energy, resulting in

2 Wm
L= (7.7)
I2
Single-Phase Variable Reactance 129

P O

M N M N

x
(a) (b)

FIGURE 7.3
Lines along which the attractive force is computed.

7.2.2.6 Electric Forces

The force on the movable part of the reactance is assessed by means of
Maxwell’s stress tensor. A surface is initially chosen that surrounds the
movable part on which the force has to be computed. In the 2D problem,
the surface is reduced to a line, for instance the line MNOP of Figure 7.3(a).
The total force in the y-direction is given by all the force components with
direction parallel to the y-axis computed along the four segments of the line,
which is
Fy = FMN + FNO + FyOP + FyPM (7.8)

In Equation (7.8) there are the normal components of the force along the
segments MN and OP:

FyMN = − L fe
∫
MN
µo
2
(
H y 2 − H x 2 dl )
(7.9)
FyOP = L fe
∫
µo
OP 2
(
H y 2 − H x 2 dl )
and the tangential components of the force along the segments NO and PM:

FyNO = L fe
∫ (µ H H ) dl
NO
o y x

(7.10)
FyPM = − L fe
∫ (µ H H ) dl
PM
o y x

Since the magnetic field assumes a negligible value outside the structure,
the main addendum is the normal force on the segment MN. The attractive
force can be estimated from the integration of the magnetic pressure on the
line MN only, as illustrated in Figure 7.3(b).
For computing the force, an alternative of Maxwell’s stress tensor is the
method of the virtual works. Let Wm(y) be the magnetic energy computed
with the movable part in the generic position y. The movable part is shifted
130 Electrical Machine Analysis Using Finite Elements

a small quantity in the y-axis direction, maintaining constant the source

current. Let ∆y be the change of position. The magnetic energy Wm(y+∆y) is
computed with the movable part in the new position y+∆y. Then, since the
system is linear, the force is computed as

∆Wm ( y) W ( y + ∆y) − Wm ( y)
Fy = = m (7.11)
∆y i= const ∆y i = const

Of course, this result depends on the chosen value of the movement ∆y.
An improvement to the force computation is obtained considering the
changes of position +∆y and –∆y. The force yields

Wm ( y + ∆y) − Wm ( y + ∆y)
Fy = (7.12)
2 ∆y i = const

In this way, all the terms of the force due to the second derivative of the
energy are eliminated. This is verified by expanding the magnetic energy
around the position y via the Taylor series. Neglecting the upper-order
addenda, it is

dWm ( y) 1 d 2 Wm ( y) 2
Wm ( y + ∆y) = Wm ( y) + ∆y + ∆y +… (7.13)
dy 2 dy 2

dWm ( y) 1 d 2 Wm ( y) 2
Wm ( y − ∆y) = Wm ( y) − ∆y + ∆y +… (7.14)
dy 2 dy 2

By subtracting the second equation from the first one, it results that

dWm ( y) Wm ( y + ∆y) − Wm ( y − ∆y)

= (7.15)
dy 2 ∆y

7.2.3 Considerations with Nonlinear B-H Curve

As for the single-phase transformer, the dc magnetizing characteristic is
initially obtained from discrete values of current I and flux linkage Λ.
Neglecting the magnetic hysteresis and the resistive voltage drop, assuming
a sinusoidal waveform of the flux linkage, the corresponding current wave-
form is computed. Then, from the current waveform the RMS value of the
current is computed. The relationship between Erms and Irms is obtained, as
reported in Figure 7.4.
Single-Phase Variable Reactance 131

E Magnetizing
Erms characteristic

ωΛ dc magnetizing
characteristic

I Irms

FIGURE 7.4
Magnetizing characteristics.

7.3 Example
A single-phase variable reactance, designed for laboratory tests, is analyzed.
It has been designed for giving rise to resonance phenomena, when con-
nected with capacitor loads. The reactance is changed so that a very high
voltage is applied to the capacitor load under test, even though the voltage
source is lower. The air-gap length is adapted from time to time depending
on the capacitor load, in order to obtain the resonance.
According to a capacitor load in the range C = 10 ÷ 200 nF, the variation
of the inductance is in the range L = 50 ÷ 1000 H.
The rated data of the single-phase reactance are as follows:

1. f = 50Hz — frequency
2. Vn = 50 kV — RMS voltage in resonance conditions
3. In = 3.15 A — maximum RMS current
4. Sn = 160 kVA — apparent power
5. magnetic material — grain-oriented TERNI M5T30, with thickness
0.3 mm and specific iron losses ps,Fe = 0.5 W/kg (at BM = 1 T, f = 50 Hz)
6. maximum flux density between 1.4 and 1.5 T (to avoid saturation)
7. the reactance is in oil, so that a class A insulation is chosen

The leg is almost circular, with SFe = 4253 mm2 and external diameter D =
249.6 mm (the stacking factor is kstk = 0.93, and the utilization factor, due to
the steps of the leg, is 0.935). The distance between the iron leg and the
winding is 25 mm with two insulating layers of 5 mm thickness each and
three oil channels. The distance between the coil and the iron yoke is 100 mm,
and between the two external diameters of the coils it is 110 mm. The main
dimensions of the reactance are reported in Figure 7.5. The movable part is
chosen with a rectangular section with a width equal to the diameter D of
the leg and a height equal to 170.4 mm.
132 Electrical Machine Analysis Using Finite Elements

875.9

170.4
t
100
39.4
30

627.6
100

41.7 25

243.3
742.5

FIGURE 7.5
Main dimensions of the reactance.

7.3.1 Inductance Adjustment

The variation of the inductance is obtained operating on the air-gap length
and on the number of turns of the reactance.
Regarding the air-gap variation, the position that has to be analyzed with
the utmost care corresponds to the higher length, when the higher leakage
flux occurs. The leakage may be so high that may inhibit the regulation of
the reactance. The critical value of the air-gap length may be considered
more or less equal to 10 mm. The minimum value of the air-gap presents
some limitations as well, due to mechanical considerations, mainly caused
by nonperfect lamination joints. The minimum air-gap length has been esti-
mated between 1.5 and 2 mm.
As a guideline, an analytical expression of the inductance is displayed. It
can be obtained from the study of the magnetic circuit of the reactance. The
self-inductance of one coil of the winding, formed by Nt/2 turns (the other
coil is open-circuited), is given by

2
 Nt 
 2  µ S
L= ≈ Nt 2 o Fe (7.16)
2t l Fe ,average 8t
+
µ oS Fe µ Feµ oS Fe

The inductance is essentially inversely proportional to the air-gap length.

Thus, the constraints imposed on the variation of the air-gap correspond to
constraints on the variation of the inductance itself. With a variation of the
air-gap length between 2 mm and 10 mm, it yields a ratio between the
maximum and the minimum value equal to 5.
Single-Phase Variable Reactance 133

L L L M L
M

(a) (b)

FIGURE 7.6
Connection between the two coils.

Such a variation ratio is increased operating on the number of turns of the

reactance, i.e., connecting the two coils of the winding in series or in parallel.
Let us assume that each coil is characterized by a self-inductance equal to
L, expressed by Equation (7.16), and a mutual inductance M ≈ L. When the
two coils are connected in series, as in Figure 7.6(a), the total inductance is

µ oS Fe
Ls ≈ 2(L + M) ≈ 4L ≈ Nt 2 (7.17)
2t

When the two coils are connected in parallel, as in Figure 7.6(b), the total
inductance is reduced to a quarter, that is

1 µ S
Lp ≈ (L + M) ≈ L ≈ Nt 2 o Fe (7.18)
2 8t

The ratio between the maximum inductance, obtained with the minimum
air-gap length and with series-connected coils, and the minimum inductance,
obtained with the higher air-gap length and with parallel-connected coils,
becomes equal to 20. In conclusion, with the parallel connection, the resonance
is obtained with capacitor loads in the range C = 40 ÷ 200 nF, while with the
series connection, the resonance is obtained with C = 10 ÷ 50 nF.

7.3.2 Definition of the Winding

The maximum flux density is obtained with the parallel connection when
the higher voltage Vn is applied to each coil. The voltage applied to the coil
134 Electrical Machine Analysis Using Finite Elements

is halved with the series connection. The number of turns has been chosen
equal to 3780 for each coil, which is a total number of turns Nt = 7560.
The maximum value of the current occurs when the reactance is in reso-
nance with the capacitive load C = 200 nF with a current I = ωCVn = 3.15 A,
to which corresponds the apparent power Sn = 160 kVA. Such resonance is
obtained with the lowest value of inductance, which is with the parallel
connection. The current in the conductors is then halved, resulting in 1.575 A.
The diameter of the conductor corresponds to dCu = 1.093 mm with a
double insulating thickness (UNEL 01722-3-4).
Each coil is divided in three elementary coils of 1260 turns each. The turns
are composed of 35 layers of 36 turns each. Each layer is insulated with a
paper sheet of thickness 0.1 mm. The dimensions of each elementary coil
result: a 41.7 mm width and a 39.4 mm height, as shown in Figure 7.5.

7.3.3 Analysis
The analysis of the reactance is reduced to a 2D analysis on the plane (x,y).
To have a planar symmetry, the circular shape of the leg is modeled as it
would be rectangular, but preserving the same cross area. This is shown in
Figure 7.7. The circular leg has the diameter D = 249.6 mm, then keeping
the width of the rectangular leg equal to D, the z-axis length becomes LFe =
170.4 mm.
The height of the movable part has to be adapted as well, in order to refer
the magnetic quantities at the same machine z-axis length LFe. Such a height
becomes equal to the diameter D, as shown in Figure 7.7. Finally, with the
aim of considering the effective iron volume together with the effect of the
squared edges of the yoke, the horizontal length of the movable part is
increased, as shown by the dotted line in Figure 7.7.
With the required simplifications, the 2D analysis allows a good estimation
of the reactance behavior, with the advantage of a faster analysis.

D D

FIGURE 7.7
Equivalent structure of the reactance to apply the planar symmetry.
Single-Phase Variable Reactance 135

FIGURE 7.8
Mesh of the section of the reactance.

We will carry out the analysis on the whole section of the machine of Figure
7.5. However, the structure presents a symmetry axis, corresponding to the
BB′ axis of Figure 7.2(c). Then, only half a structure could be analyzed,
assigning Neumann’s boundary condition on the BB′ axis.
Since the magnetic field strength is higher in the air-gap, the mesh is
increased in this zone, as shown in Figure 7.8. The subdivision of the struc-
ture is carried out using some 2000 triangular elements. This choice proved
to be appropriate to obtain accurate results: a 20% increase of the number
of the finite elements generates a variation of the magnetic energy in the
field solution lower than 0.3%.
The magnetic B–H curve Terni M5T30 is used for the iron core.
Various simulations are carried out with different values of air-gap length t
and current I. Since the elementary coils are schematized by means of conduc-
tive bars, the equivalent current in each coil is obtained as the product of the
current I by the number of turns of each elementary coil, which is Nt/6 = 1260.
Figure 7.9 shows the flux plot. The detail of Figure 7.9(b) highlights that
the flux lines are not confined within the air-gap, but they expand at the
extremity of the legs. The existence of such flux lines points out that, at the
same current (and then at the same MMF), the flux linkage is higher than
that analytically predicted with the assumption of flux lines that are only
normal to the air-gap. Consequently, the inductance is higher. It is then
necessary to verify that the inductance assumes the required values, espe-
cially when the air-gap length is maximum.
Figure 7.10 shows the flux density magnitude in the middle of the air-gap,
corresponding to different values of air-gap length and current. The flux
density is drawn as a function of the distance from the symmetry axis of the
reactance. Figure 7.10(a) corresponds to an air-gap t = 1.5 mm and a current
I = 0.154 A. Figure 7.10(b) corresponds to an air-gap t = 9 mm and a current
I = 3.19 A.
136 Electrical Machine Analysis Using Finite Elements

(a) (b)

FIGURE 7.9
Flux plots.

B (T) B (T)
0.8 1.4
0.7 1.2
0.6 1
0.5
0.8
0.4
0.6
0.3
0.4
0.2
0.1 0.2
0 0
0 50 100 150 200 250 300 350 400 450 0 50 100 150 200 250 300 350 400 450
l (mm) l (mm)

FIGURE 7.10
Flux density in the middle of the air-gap as a function of the distance of the symmetry axis;
(a): t = 1.5 mm and I = 0.154 A; (b): t = 9 mm and I = 3.19 A.

In both figures, the edge effect is evident: in the central zone the value of
the flux density magnitude is essentially constant; near the edges it increases
about 10%.
Table 7.1 contains the values of the inductance obtained varying the air-
gap t. Low current is adopted to avoid saturation effect. The inductances are
computed by Equation (7.7) using the magnetic energy, given by Equation
(7.2) or Equation (7.3), since the problem is essentially linear. The same result
is obtained, however, if the inductances are computed by Equation (7.6)
using the flux linkages, given by Equation (7.4). The inductances with par-
allel-connected coil are obtained simply dividing by 4 the inductances with
series-connected coils.
For an easy comparison, in the same table the inductances computed
analytically are reported as well, as given by Equation (7.17) and Equation
Single-Phase Variable Reactance 137

TABLE 7.1
Values of the Self-Inductance, Computed at Low Current
t Ls Ls(analytic) Lp Lp(analytic)
(mm) (H) (H) (H) (H)
1.5 1047.0 1002.3 261.7 250.6
1.6 972.4 940.6 243.4 235.1
1.7 917.6 886.0 229.4 221.5
1.8 870.2 837.4 217.6 209.4
1.9 826.8 793.9 206.7 198.5
2.0 789.8 754.7 197.5 188.7
2.1 752.2 719.2 188.1 179.8
2.2 720.0 686.8 180.0 171.7
2.3 690.0 657.3 172.6 164.3
2.4 662.8 630.1 165.7 157.5
2.5 638.0 605.2 159.5 151.3
3.0 538.4 505.1 134.6 126.3
4.0 412.8 379.6 103.2 94.9
5.0 336.4 304.0 84.1 76.0
6.0 285.8 253.6 71.5 63.4
7.0 249.0 217.5 62.3 54.4
8.0 221.0 190.4 55.3 47.6
9.0 199.5 169.3 49.9 42.3
10.0 182.1 152.4 45.5 38.1

(7.18). As expected, the values obtained from the finite element method are
higher than those analytically computed. This is essentially due to the flux
lines external to the air-gap, as highlighted in Figure 7.9. The relative differ-
ence between the two inductances is as high as the air-gap increases. In fact,
with the higher values of t, the flux outer the air-gap increases.
Figure 7.11 illustrates graphically the results reported in Table 7.1.

1100
1000
900
800
Inductance (H)

700
600
500
400 Ls
Ls (analytical)
300
200
Lp
100
Lp (analytical)
0
2 3 4 5 6 7 8 9 10
Airgap length (mm)

FIGURE 7.11
Behavior of the inductance as a function of the air-gap, with constant current.
138 Electrical Machine Analysis Using Finite Elements

TABLE 7.2
Effect of the Saturation (t = 1.5 mm)
I Wm Λ Ls
(mA) (J) (Vs) (H)
58 1.76 60.71 1046.6
104 5.66 108.91 1047.2
150 11.78 157.12 1047.5
196 20.12 205.35 1047.7
242 30.68 253.58 1047.9
288 43.40 301.36 1046.4
334 58.19 348.46 1043.3
373 71.91 387.99 1040.2
419 89.41 432.30 1031.7
458 105.30 468.97 1024.0
504 123.10 506.53 1005.0
550 136.00 531.26 965.9
596 144.40 545.89 915.9
642 149.20 553.40 862.0
688 153.60 560.23 814.3
734 158.50 567.33 772.9
780 162.8 573.31 735.0
826 166.50 578.24 700.0
872 169.90 582.62 668.1
918 173.10 586.60 639.0
964 176.10 590.17 612.2

In order to highlight the saturation effect, the air-gap length is fixed to t =

1.5 mm and the current is changed between 0 and 1 A. The analysis is carried
out with current higher than the nominal current. In fact, with such an air-
gap length, the maximum current is 0.4 A (peak value).
Some results are given in Table 7.2, which reports the magnetic energy, the
flux linkage, and the inductance computed by means of Equation (7.6). Since
the problem does not remain linear, the inductance assumes the meaning of
apparent inductance. Let us remember that the computation by means of
the magnetic energy (7.7) is correct only in linear conditions. From Table 7.2
it is possible to verify that, with low currents, Equation (7.6) and Equation
(7.7) yield the same result, while they do not with high current.
Figure 7.12 shows the flux linkage Λ versus the current I. Since fixed values
have been used, this is the dc magnetizing curve. The same figure also shows
the apparent inductance (obtained as the ratio between Λ and I), as a function
of the current.
From the figure, it is easy to verify that the reactance works in linearity
during the normal operation, since the maximum nominal current is 0.4 A.
A sinusoidal waveform of the flux linkage is assumed, which is a sinuso-
idal waveform of the voltage with a negligible resistive voltage drop. Then
the current waveform is reconstructed from the dc magnetizing characteristic
of Figure 7.12.
Single-Phase Variable Reactance 139

Flux linkage (Vs) Inductance (H)

600 1200

500 1000
Inductance

400 800

300 600

200 400
Flux linkage
100 200

0 0
0 200 400 600 800 1000
Current (mA)

FIGURE 7.12
Flux linkage and inductance as functions of the current (at t = 1.5 mm).

The algorithm for the interpolation is described at the end of Chapter 6.

Figure 7.13 shows some waveforms of flux linkage and current, for different
peak values of the flux linkage. The saturation effect on the current waveform
is evident.
As described in the previous chapter, the magnetizing characteristic of the
reactance is obtained, computing the RMS value of the current in correspon-
dence to the flux linkage. As a consequence, a new value for the inductance
can be defined, say L*. It is obtained as the ratio between the RMS value of
the flux linkage and the RMS value of the current. Referring to Figure 7.13(d),
obtained with Nj points, here are the results:

ΛM 600
Λ rms = = = 424.3 Vs
2 2
Nj

∑i
2 π/ω
ω
∫
1
I rms = i(t ) dt =
2 2
= 0.556 A
2π
j
0 Nj j=1

Λ rms
L* = = 763.7 H
I rms
140 Electrical Machine Analysis Using Finite Elements

λ (Vs), i (mA) λ (Vs), i (mA)

400 500
λ (t) 400
300 i(t) λ (t)
i(t) 300
200
200
100 100
0 0
−100 −100
−200
−200
−300
−300 −400
−400 −500
0 π/2 π 3π/2 2π 0 π/2 π 3π/2 2π
ωt ωt
(a) (b)

λ (Vs), i (mA) λ (Vs), i (mA)

600 i(t) 1000
λ (t) i(t)
400
500
200
λ (t)
0 0
−200
−500
−400
−600 −1000
0 π/2 π 3π/2 2π 0 π/2 π 3π/2 2π
ωt ωt
(c) (d)

FIGURE 7.13
Current waveforms, corresponding to a sinusoidal flux linkage with peak value (a): 400 Vs; (b):
500 Vs; (c): 550 Vs; (d): 600 Vs.

References
1. G. Kron, “Equivalent Circuits to Represent the Electromagnetic Field Quanti-
ties,” The Physical Rev., vol. 64, Aug. 1–15, 1943, pp. 126–128.
2. L.F. Blume, A. Boyajian, Transformer Engineering, John Wiley, New York, 1951.
3. G. Rago, Costruzioni Elettromeccaniche e Disegno, Sansoni, Firenze, 1968.
4. N. Bianchi, F. Dughiero, “Optimal Design Techniques Applied to Transverse-
Flux Induction Heating Systems,” in IEEE Trans. on Magnetics, vol. 31, n. 3,
Maggio 1995, pp. 1992–1995.
5. N. Bianchi, F. Dughiero, S. Lupi, “Design of Induction Heating Systems by
Optimisation of Field Shape,” in Proceedings of the International Induction Heating
Seminar, IHS ’98, Padova, 13–15 Maggio, 1998, pp. 413–423.
6. W.T. McLyman, Transformer and Inductor Design Handbook, CRC Press, Boca
Raton, FL, 2004.
8
Synchronous Generators

This chapter describes the finite element analysis of a three-phase synchro-

nous generator. The no-load characteristic and the direct and quadrature axis
inductances are investigated.

8.1 Introduction
The synchronous generator is composed of a fixed part, the stator, and a
rotating part, the rotor. In both of them, magnetic materials are used to guide
the magnetic flux, and windings are employed to carry the current. The
structure of a 2p = 2 pole synchronous generator is represented in Figure 8.1.
The rotor winding, called field winding or magnetizing winding, is fed by
a direct current, by means of dragging contacts, i.e., brushes and conductive
rings. This is the primary source of the main flux of the machine. The polar
axis in the middle of the rotor pole, sketched in Figure 8.1, is called the direct
axis, or d-axis. The interpolar axis is 90 electrical degrees in advance with
respect to the rotating direction and is called the quadrature axis, or simply
q-axis. The rotor rotates at constant mechanical angular speed ωm (an elec-
trical angular speed ω = pωm), giving rise to a rotating magnetic field.
The stator winding, also called armature winding, links a variable mag-
netic flux so that a variable EMF is induced. The three phases of the winding
are named a, b, and c. Each phase winding is spaced out of 120 electrical
degrees. They are sketched in Figure 8.1 by means of a single turn and are
identified by the axis normal to the turn itself. Observe that the orientation
of the turns agrees with their axis directions, according to the right-handed
screw advance (a rotation in direction of the positive current would cause
the right-handed screw to advance in the direction of the turn axis). Finally,
the rotor position with respect to the stator is pointed out by the electrical
angle ϑ, between the polar axis and the a-phase axis. The angle ϑ is ϑ = pϑm
where ϑm is the mechanical angle between rotor and stator.
When the armature windings are open-circuited, no current flows and the
induced EMF can be measured at the winding terminals. This is the no-load

141
142 Electrical Machine Analysis Using Finite Elements

Rotor coordinate
b-phase axis ϑr
Interpolar axis Polar axis
(or quadrature (or direct axis)
axis)
+a
ϑm Angular position
between
−c −b rotor and stator
Field winding ω a-phase axis
(or magnetizing Rotation
+b
winding)
+c
−a
Armature
winding
c-phase axis

FIGURE 8.1
Structure and references of the three-phase synchronous generator.

voltage of the generator. On the contrary, when the armature windings are
connected to the load, variable currents flow in the windings, with the same
frequency of the EMF. Such currents cause a magnetic field that rotates at
the same speed (called the synchronous speed) of the rotor speed. The
electromechanical conversion is obtained by the interaction between the
stator and rotor magnetic fields.
Neglecting the edge effects and assuming an identical behavior along the
generator, a planar symmetry is supposed: only the plane (x, y) is considered,
the magnetic vector potential and the current density have a z-axis compo-
nent only, which is A = (0, 0, Az) and J = (0, 0, Jz). Then the flux density and
magnetic field strength vectors have a component on the plane (x, y) only,
which is B = (Bx, By , 0) and H = (Hx, Hy , 0).
The machine length is chosen equal to the effective length of the iron LFe,
considering the stacking factor kstk, which considers the insulation among
the laminations (kstk = 0.95 – 0.98) and the possible ventilation channels (e.g.,
Ncv channels, whose wideness is lcv). Then the net length is

( )
L Fe = L − Ncv l cv k stk (8.1)

where L is the total length of the machine.

In addition, a periodicity exists among the pole pairs, so that the study
may be carried out only on one pole pair, as in Figure 8.2(a). When a stator
symmetry exists among the poles (the number of slots is a multiple of the
number of poles), the analysis may be carried out on a single pole centered
along the polar axis, as in Figure 8.2(b), or centered along the interpolar axis,
as in Figure 8.2(c). Centering the drawing along the d- or the q-axis, as shown
in Figure 8.2, is not really necessary, since the periodic conditions may be
Synchronous Generators 143

q d q d q d q d

O O O O
(a) (b) (c) (d)

FIGURE 8.2
Sections of the machine and periodicity.

assigned no matter what axis is chosen. However, the choice of the d- or the
q-axis is particularly suitable in the assignment of the boundary conditions.
In fact, special operating conditions allow Dirichlet’s or Neumann’s bound-
ary conditions to be assigned, as will be hereafter described.
Finally, some cases exist in which the analysis is reduced to half a pole, as
shown in Figure 8.2(d).

8.2 Computation of the No-Load Characteristic

The field winding, formed by Ne turns, is fed by a constant current Ie. The
armature windings are open-circuited so that they do not carry any current.
They may be considered to have null conductivity. The stator and rotor
magnetic material is described by the suitable B-H curve.
The boundary conditions are assigned, considering that the flux lines of
half a pole are the mirrored flux lines of the other half of the pole. Then, a
value of the magnetic vector potential is fixed along the polar axis (Dirichlet’s
condition), while a normal flux density vector is assigned along the interpo-
lar axis (Neumann’s condition). Since the flux lines do not exit from the yoke,
a null magnetic vector potential is assigned, along the external circumference
of the stator. Figure 8.3 shows the boundary conditions assigned to the
different sections of the generator, corresponding to Figure 8.2(b–d), at the
no-load operation.
The field winding, composed by Ne turns and carrying the excitation
current Ie, is modeled by an equivalent conductive bar carrying the total
current NeIe.
The three-phase windings may be displaced into the stator slots, in what-
ever way with respect to the rotor. However, it is useful to take advantage
of the electrical symmetry. Figure 8.4 shows a part of a single-layer, non-
chorded winding, referred to one pole of the stator with 6 slots per pole.
Figure 8.4(a) and Figure 8.4(b) differ for the winding distribution; in the
figures the a-phase axis is highlighted as the stator reference.
144 Electrical Machine Analysis Using Finite Elements

Dirichlet Az = 0 Az = 0
Az = 0

dAz
dAz =0
=0 dn
dn

dAz
=0
dn Az = 0 Az = 0 Az = 0
Neumann

(a) (b) (c)

FIGURE 8.3
Boundary conditions in the no-load simulation.

a-phase axis a-phase axis

−b −b +a +a −c −c +a −c −c +b +b −a

FIGURE 8.4
Distribution in the slots of the three-phase windings: single-layer, nonchorded winding.

a-phase axis a-phase axis

−b +a −c +a −c +b −a

FIGURE 8.5
Distribution in the slots of the three-phase windings: double-layer, one-slot chorded winding.

Figure 8.5 shows a part of a double-layer, one-slot chorded winding,

referred to one pole of a stator with 6 slots per pole. In order to have a
complete electrical and geometrical symmetry, the stator pole is centered on
the middle of the a-phase axis. The stator reference is highlighted by means
of the a-phase axis. It is worth noticing that, in this case, the first and the
last slots have to be halved.

8.2.1 Computations after the Field Solution

Once the field problem is solved, the z-axis component of the magnetic vector
potential Az(x, y) is known in each point of the domain. The flux linked with
the three-phase windings and the induced EMFs are computed. Particular
care has to be kept to the turns distribution in the stator slots.
Synchronous Generators 145

8.2.1.1 Flux Linkage

The mean value of the magnetic vector potential in the q-th slot, whose
section is Sq, is

∫
1
A zdS (8.2)
Sq Sq

Let nq be the number of conductors per slot and npp be the number of
parallel paths of the machine. It follows that the total number of conductors
of the machine is npp times the number of series conductors. Considering
that the analysis domain is one pole only — see Figure 8.3(a) or Figure
8.3(b) — the j-th phase flux linkage is given by

Q/2 p

∑k ∫
nq 1
Λ j = 2 pL Fe jq A zdS j = a , b, c (8.3)
n pp q= 1
Sq Sq

where 2p is the number of poles and LFe is given by Equation 8.1; Q is the
number of the slots of the generator, so that Q/2p is the number of slots per
pole. Finally, kjq is the coefficient taking into account whether the conductors
in the q-th slot are of the j-th phase or not, as well as the conductor orientation.
Such coefficient assumes the following values.

Single-Layer Winding (One Coil Side in Each Slot)

kjq = 0 — if the coil side in the q-th slot does not belong to the j-th phase;
kjq = +1 — if the coil side in the q-th slot belongs to the j-th phase and
its orientation is positive with respect to the z-axis direction (leaving
the sheet with the references of Figure 8.1);
kjq = –1 — if the coil side in the q-th slot belongs to the j-th phase and
its orientation is negative with respect to the z-axis direction (going
into the sheet with the references of Figure 8.1).

Double-Layer Winding (Two Coil Sides in Each Slot)

kjq = 0 — if the coil sides in the q-th slot do not belong to the j-th phase;
kjq = +0.5 — if only one coil side in the q-th slot belongs to the j-th phase
and its orientation is positive with respect to the z-axis direction;
kjq = –0.5 — if only one coil side in the q-th slot belongs to the j-th phase
and its orientation is negative with respect to the z-axis direction;
kjq = +1 — if both two coil sides in the q-th slot belong to the j-th phase
and their orientation is positive with respect to the z-axis direction;
kjq = –1 — if both two coil sides in the q-th slot belong to the j-th phase
and their orientation is negative with respect to the z-axis direction.
146 Electrical Machine Analysis Using Finite Elements

TABLE 8.1
Values of kjq Referring to the
Distribution of Figure 8.4(a)
phase (j)
slot (q) a b c
0 –1 0
0 –1 0
+1 0 0
+1 0 0
0 0 –1
0 0 –1

TABLE 8.2
Values of kjq Referring to the
Distribution of Figure 8.5(a)
phase (j)
slot (q) a b c
1 0 –0.25 +0.25
2 0 –1 0
3 +0.5 –0.5 0
4 +1 0 0
5 +0.5 0 –0.5
6 0 0 –1
7 0 +0.25 –0.25

In other words, the magnitude of kjq specifies the relative filling of the q-th
slot by the j-th phase conductors. The sign of kjq considers the sign of the
scalar product A·t of the loop integral: the vector A has only the component
Az while the tangential unity vector t (which defines the turn’s orientation)
coincides with uz when a positive current flows in the z-axis direction, or
with –uz when the same positive current flows opposite to the z-axis direc-
tion. Thus, A·t = kjqAz is equal to +Az or to –Az.
For instance, with the single-layer winding of Figure 8.4(a), the values
reported in Table 8.1 are obtained, while with the double-layer winding of
Figure 8.5(a), the values reported in Table 8.2 are obtained. In the latter case,
the value 0.25 has been used in the slot q = 1 and in the slot q = 7, because
only half a slot is considered. Finally, it is worth noticing that the sum of the
magnitudes of the coefficients of each row (with whole slot) is always equal
to one.

8.2.1.2 Induced EMF

For the sake of convenience, let us consider the case of maximum flux linked
with the a-phase winding. This means choosing the a-phase axis parallel to
Synchronous Generators 147

−b −b +a +a −c −c
Uo(n)

+c −b −b +a +a −c

+c +c −b −b +a +a
0 π 2π
ϑ = ωt

FIGURE 8.6
Construction of the voltage waveform by dots.

the d-axis. Then, indicating Λma the maximum flux linkage, with constant
rotor speed, the RMS value of the induced EMF is

1
Ea = ωΛ ma (8.4)
2

which is the no-load line voltage with a wye winding connection. Con-
versely, with a star winding connection, the no-load line voltage is

U 0 = 3 Ea (8.5)

From the same field solution, supposing a step rotation of the coils of one
slot pitch, it is possible to achieve some points of the no-load voltage wave-
form. Assuming this rotation, different coil sides are associated to each slot,
as shown in Figure 8.6, the coefficients kjq are suitably modified. This is
possible because only the rotor is fed, which means the flux lines do not
depend on the distribution of the conductors into the slots.
The waveform Uo(n) is built by Nc points (e.g., it is common that Nc = Q/p).
The harmonic content can be approximately estimated. From the discrete
Fourier series expansion, the fundamental value is obtained as

Nc
 
U1sin =
2
Nc ∑ U ( n ) ⋅ sin  N2π n
n =1
o
c

Nc
 
U1 cos =
2
Nc ∑ U ( n ) ⋅ cos  N2π n
n =1
o
c
(8.6)

U1 = U1sin 2 + U1 cos 2
148 Electrical Machine Analysis Using Finite Elements

Then the ν-th order harmonic is

Nc
 
U ν sin =
2
Nc ∑ U ( n ) ⋅ sin  N2π νn
n =1
o
c
(8.7)
Nc
 
U ν cos =
2
Nc ∑ U ( n ) ⋅ cos  N2π νn
n =1
o
c

To avoid rough errors, the voltage harmonics can be evaluated up to the

maximum order equal to ν < Nc/2 (Shannon’s theorem).

8.3 Computation of the Direct-Axis Inductance

In the computation of the direct-axis inductance, Ld, the magnetizing current
is set to be zero and the stator windings are fed in such a way as to obtain
a MMF distribution with the maximum value coincident with the polar axis.
Since the magnetic field is synchronous with the rotor, a magnetostatic sim-
ulation is possible.
Thanks to the electrical and geometrical symmetry, only a portion of the
machine is analyzed, that is, a portion equal to one pole or to half a pole.
The boundary conditions are assigned to constrain the flux lines to be
tangential along the polar axis and normal to the interpolar axis. The boundary
conditions are those assigned in the no-load simulation, shown in Figure 8.3.
In order to define the phase current, let us refer to the phasor diagram of
Figure 8.7. With reference to the time instant t, it is posed

i a (t ) = I M sin ωt ( )
 2π 
i b (t ) = I M sin  ωt − (8.8)
 3 
 4π 
i c (t ) = I M sin  ωt −
 3 

Ic
Ia

ωt

Ib

FIGURE 8.7
Reference phasor diagram of the stator winding currents.
Synchronous Generators 149

a-phase axis
Ia
−b −b +a +a −c −c
ω t = π /2
Polar axis
(d-axis)
Ic Ib
(a)

a-phase axis

Ic
+a −c −c +b +b −a
Polar axis
(d-axis) Ia

ωt=0
Ib
(b)

FIGURE 8.8
Combinations of the winding distribution and time instant, for computing the direct axis
inductance.

Once the distribution of the winding is fixed, the three-phase currents are
chosen so that the maximum of the MMF distribution coincides with the
polar axis (d-axis). This corresponds to set Id = In and Iq = 0. Referring to one
pole of the generator, Figure 8.8 shows two suitable combinations of winding
distribution and time instant. For both of them, the boundary conditions of
Figure 8.3(b) can be assigned.
With the configuration of Figure 8.8(a) the time instant with ωt = π/2 has
been chosen, so that

ia = I M

IM
ib = − (8.9)
2
IM
ic = −
2

With the configuration of Figure 8.8(b) the time instant with ωt = 0 has
been chosen, so that

ia = 0

3
ib = − IM (8.10)
2

3
ic = + IM
2
150 Electrical Machine Analysis Using Finite Elements

Once the three-phase currents ia, ib, and ic, and the distribution of the
windings have been fixed, i.e., the coefficients kaq, kbq, and kcq for each slot,
the current flowing in the q-th slot can be expressed as

∑k i
nq
iq = jq q
n pp j= a , b , c
(8.11)

( )
n
= q k aqi a + k bqi b + k cqi c
n pp

8.3.1 Computation on the Solved Structure

Once the field solution is obtained, the direct axis inductance may be com-
puted in several ways.

8.3.1.1 Computation of Ld by Means of the Magnetic Energy

Let Wmd be the magnetic energy stored in the synchronous generator, due
to the stator d-axis currents only. In linear conditions it is

31 
Wmd =  Ld Id2  (8.12)
22 

With d-axis current only, it is Id = IM, and the synchronous d-axis inductance
results in

2 2 Wmd
Ld = (8.13)
3 I 2M

Slightly different results are obtained with the distributions of Figure 8.8(a)
and Figure 8.8(b) and the corresponding currents indicated in Equation (8.9)
and Equation (8.10). This is due to the different harmonic contents of the
two configurations. At last, we must bear in mind that this computation is
correct only in case of a linear magnetic circuit.

8.3.1.2 Computation of Ld by Means of the Flux Linkage

The synchronous direct axis inductance is defined as the inductance of that
phase whose axis coincides with the polar axis, when the three-phase wind-
ings are simultaneously fed. Let us refer to the combination of Figure 8.8(a).
It is worth noticing that the polar axis coincides with the a-phase axis, so
that ϑ = 0. Since Id = IM, forcing d-axis currents only, and Ia = IM, from Equation
(8.9), thus Id = Ia.
Synchronous Generators 151

The flux linked with the a-phase winding, Λa = Λd, is computed from
Equation (8.3). The inductance is given by

Λa Λd
Ld = = (8.14)
Ia IM

Similarly, the inductance can be computed from the winding distribution

of Figure 8.8(b) and the currents given by Equation (8.10). In this case, ϑ =
π/2, Ib = (√3/2)IM, and Λd = (√3/2)Λb (forcing d-axis currents only, always
Id = IM), then

Λb 2Λ b
Ld = = (8.15)
Ib 3IM

The difference between Equation (8.14) and Equation (8.15) indicates the
different harmonic contributions of the two solutions.

8.3.1.3 Computation of Ld by Means of the Air-Gap Flux Density

From the field solution, it is possible to obtain the radial component of the
flux density distribution in the air-gap, say Bgn(ϑr) where ϑr is the angular
position referred to the polar axis (which is the rotor coordinate, as shown
in Figure 8.1 expressed in electrical radiants). The fundamental harmonic of
the magnetic flux density with respect to the direct axis is given by

∫B
2
B1dM = (ϑ r )cos(ϑ r )dϑ r (8.16)
π
gn
ο

If the distribution of the radial component of the flux density Bgn(n) is

known in Np points, the computation becomes

Np

B1dM =
2
Np ∑B
n =1
gn ( n )cos( n∆ϑ)∆ϑ (8.17)

where ∆ϑ = π/Np (assuming a simulation of one pole of the generator).

From the flux density value (8.17), the flux is obtained as the product of
the mean value of the sinusoidal distribution (of the fundamental harmonic)
by the surface corresponding to one pole. Then, the peak value of the d-axis
flux linkage is

k w N DL Fe
Λ d = B1dM (8.18)
2 p
152 Electrical Machine Analysis Using Finite Elements

where D is the stator bore diameter, N is the number of series conductors

per phase, and kw is the winding factor. Finally

Λd
Ldm = (8.19)
IM

The last computation adopts the fundamental harmonic of the flux density
distribution, then is almost independent of the harmonic content.
In addition, while the leakage flux in the air-gap and in the slots is con-
sidered in Equation (8.13) and Equation (8.14), it is not considered when
Equation (8.19) is used. Thus, letting Lσ be the leakage inductance, the direct
axis inductance is obtained as

Ld = L σ + Ldm (8.20)

It is worth noticing that Equation (8.14) and Equation (8.19) can be applied
even without setting the magnetizing current equal to zero. In such a case,
the flux linkage due to the magnetizing current must be previously com-
puted and considered constant. It has to be subtracted from the total flux
linkage. With reference to the a-phase, Equation (8.14) and Equation (8.19)
become

Λ d − Λ dm Λ a − Λ ma
Ld = = (8.21)
IM IM

where Λdm = Λma indicates the a-phase flux linkage due to the magnetizing
current.

8.4 Computation of the Quadrature Axis Inductance

The computation of the quadrature axis inductance, Lq, is carried out in the
same way as the computation of the Ld inductance. The phase currents and
the winding distribution must be chosen so as to have an MMF distribution
with the maximum value coincident to the interpolar axis, i.e., the q-axis.
Referring to Figure 8.8, the two time instants (i.e., the angles ωt) that char-
acterized the three-phase currents ia, ib, and ic with respect the winding
distribution have to be exchanged. The boundary conditions are suitably
adjusted, as shown in Figure 8.9.
The q-axis inductance is then obtained as follows.
Synchronous Generators 153

Dirichlet Az = 0 Az = 0
Az = 0

Az = 0 dAz
Az = 0 =0
dAz dAz dn
=0 =0
Az = 0 dn dn
Neumann

(a) (b) (c)

FIGURE 8.9
Boundary conditions for computing the quadrature axis inductance.

8.4.1 Computation of Lq by Means of the Magnetic Energy (Linear Case)

2 2 Wmq
Lq = (8.22)
3 I 2M

8.4.2 Computation of Lq by Means of the Flux Linkage

Λq
Lq = (8.23)
IM

8.4.3 Computation of Lq by Means of the Air-Gap Flux Density

∫B
2
B1qM = (ϑ r )sin(ϑ r )dϑ r (8.24)
π
gn
ο

k w N DL Fe
Λ q = B1qM (8.25)
2 p

Λq
Lqm = (8.26)
IM
154 Electrical Machine Analysis Using Finite Elements

8.5 Self- and Mutual Inductances

The self- and mutual inductances are obtained by feeding only one phase
winding and computing the flux linked with the three-phase windings. For
instance, only the a-phase winding is fed with a current ia = Ia. The three-
phase flux linkages Λa, Λb, and Λc are computed as given in Equation (8.3).
Then

Λ a (ϑ )
L a (ϑ ) =
Ia

Λ b (ϑ )
M ba (ϑ) = (8.27)
Ia

Λ c (ϑ )
M ca (ϑ) =
Ia

As highlighted, the self- and mutual inductances depend on the angular

position ϑ. The most interesting values are in the position ϑ = 0 (d-axis
corresponding to the a-phase axis) and ϑ = −π/2 (q-axis corresponding to
the a-phase axis). The a-phase flux linkage Λa assumes the maximum value
when ϑ = 0, and minimum value when ϑ = ±π/2. It can be verified that Mba ≈
Mca ≈ M ≈ –La/2. Then it is

at ϑ = 0 Λa(0) maximum Lq ≈ La(0) – M(0) ≈ 3/2 La,max

at ϑ = ±π/2 Λa(π/2) minimum Lq ≈ La(π/2) – M(π/2) ≈ 3/2 La,min (8.28)

8.6 Saturation Effect

The inductances computed above are influenced by the nonlinearity of the
magnetic material of the rotor and stator. They depend on the values of the
currents that are imposed in the analysis. In addition, Equation (8.14) and
Equation (8.21) do not yield the same result, which is the computation with
null or non-null magnetizing current.
The inductances computed as described above assume the meaning of
apparent inductances. The differential inductances may be computed from
two field solutions, as the ratio between the flux linkage variation and the
corresponding current variation, as described in Chapter 5.
Synchronous Generators 155

8.7 Computation of L d and L q with Any Current

Until now the analysis of the flux linkages and the d- and q-axis inductances
has been carried out separately, feeding the windings so as to obtain the
MMF distribution with the maximum value coincident to the d-axis or the
q-axis. However, it is possible to carry out an analysis with stator currents
independent of the particular rotor position ϑ.
In this case, each section of Figure 8.2(a), (b), or (c) can be used, assigning
the periodic boundary conditions as shown in Figure 8.10. They are even
periodic boundary conditions in Figure 8.10(a), and odd periodic boundary
conditions in Figure 8.10(b) and Figure 8.10(c).
On the contrary, it is not possible to reduce the study to the section of
Figure 8.2(d), since the periodic boundary conditions cannot be assigned to
this section.
The d- and the q-axis flux linkage can be obtained from the three-phase
flux linkages or from the air-gap flux density distribution.
Starting from the three-phase flux linkages Λa, Λb, and Λc, computed as
given from Equation (8.3), the d- and the q-axis flux linkages Λd and Λq are
achieved by means of the transformation Tabc/dq (see the Appendix to this
chapter):

2  2π   4π  
Λd =  Λ a cos ϑ + Λ b cos  ϑ − + Λ c cos  ϑ − 
3  3   3  
(8.29)
2  2π   4π  
Λ q = −  Λ a sin ϑ + Λ b sin  ϑ −  + Λ c sin  ϑ − 
3  3   3  

Az = 0 Az = 0 Az = 0

+Az (r)
+Az (r) −Az (r)
+Az (r) +Az (r) −Az (r)

(a) (b) (c )

FIGURE 8.10
Periodic boundary conditions.
156 Electrical Machine Analysis Using Finite Elements

Alternatively, starting from the radial component of the air-gap flux density
distribution Bgn(ϑ), the two-axis fundamental harmonic components are
obtained from Equation (8.16) and Equation (8.24). The d- and q-axis flux
linkages Λ1d and Λ1q are then obtained from Equation (8.18) and Equation
(8.25). At this point, the d- and q-axis currents are computed from the three-
phase currents Ia, Ib, Ic, always by means of the transformation Tabc/dq (see
the Appendix to this chapter), as

2  2π   4π  
Id =  Ia cos ϑ + I b cos  ϑ −  + I c cos  ϑ − 
3  3   3  
(8.30)
2  2π   4π  
Iq = −  Ia sin ϑ + I b sin  ϑ −  + I c sin  ϑ − 
3  3   3  

Finally, neglecting the cross coupling between the d- and q-axis (that will
be considered in Chapter 10), the d- and q-axis inductances result in

Λd
Ld =
Id
(8.31)
Λ
Lq = q
Iq

8.8 Computation of the Machine Characteristics

The finite element method can be used to simulate the steady-state operation
of the synchronous generator. During steady state, the three-phase currents
and voltages vary sinusoidal with the time; thus the complex symbolic
notation can be used. According to the phasor diagram of Figure 8.11, the
simulation is carried out as follows:
Input quantities:

• Thanks to the synchronous speed of the generator, a magnetostatic

analysis is carried out, freezing a particular instant of its operation
(this corresponds to analyzing the machine in the rotor reference
frame). A constant synchronous speed ωm is considered; then the
operating electrical frequency is ω = pωm.
• The position of the rotor ϑ is fixed.
• The magnetizing current Ie is fixed.
• The maximum value of the stator current IM and the phase angle ωt
is chosen for the armature currents. From these values the three-
phase currents Ia, Ib, and Ic are obtained.
Synchronous Generators 157

XdId
Uo
RId
XqIq
RIq

Uq U
δ
I
ϕ
Iq γ
Id Ud
Polar axis
(d-axis) Interpolar axis
(q-axis)

FIGURE 8.11
Phasor diagram of the synchronous generator.

Computed quantities:

• From the field solution, the three-phase flux linkages Λa, Λb, and Λc
are computed.
• By means of the transformation Tabc/dq, the d- and q-axis components
of current, i.e., Id Iq, and flux linkages, i.e., Λd Λq, are computed.
• The d- and q-axis voltages are

Ud = − RId + ωΛ q
(8.32)
Uq = − RIq − ωΛ d

• The sign refers to the generating mode of the electrical machine, that
is, the positive current direction is in phase with the voltage. The
phasor Uo in Figure 8.11 is π/2 radiants lagging the polar axis (d-axis).
• The electrical angles are

Ud
δ = tan load angle
Uq

Iq
γ = tan current angle (8.33)
Id

π
ϕ= −γ −δ angle between voltage and current
2

In order to obtain the machine characteristics, various simulations should

be carried out, changing one or more input quantities.
158 Electrical Machine Analysis Using Finite Elements

ϑr d-axis

Stator
Infinitesimal 60°
conducting
sheet Rotor ϑs
y

200 480 500 800

x

Field
winding

200
400

FIGURE 8.12
Ideal synchronous machine.

8.9 Example
The aim of the following example is to show the flux plots and the air-gap
flux density distribution of the synchronous generator, during no-load and
full-load operations. The effects of the armature reaction are highlighted,
feeding the machine with d-axis currents only, q-axis currents only, and both
of them together.
Rather than analyzing an effective figure of the synchronous generator, let
us refer to the ideal structure drawn in Figure 8.12. This is a two-pole
machine, formed by a rotor with salient poles through which the field wind-
ings are wound, and by a slotless stator. In the inner surface of the stator
there is an infinitesimal conducting sheet carrying a linear current density
distribution. Such a distribution is considered to be sinusoidal, correspond-
ing to an ideally sinusoidal distribution of the winding.
In Figure 8.12, the dimensions of the generator are reported in mm. A
linear magnetic material is used, with a constant relative permeability µr =
5000.
The first simulation deals with the no-load operation, without stator cur-
rent and with a magnetizing current feeding the rotor winding. In the con-
ducting bar, equivalent to the magnetizing winding, the total current is set
NeIe = 13500 A. Figure 8.13 shows the corresponding flux plot and the air-
gap flux density distribution. Since the pole shoe is not shaped — which is
a constant air-gap under the pole — the flux density distribution looks like
a quasi-square distribution. The maximum flux density corresponds to the
Synchronous Generators 159

Air-gap flux density (T)

0.8
0.6
0.4
0.2

−0.2
−0.4
−0.6
−0.8
−1
0 90 180 240 360
Angle (electr. degrees)

FIGURE 8.13
No-load flux plot (a) and air-gap flux density distribution (b).

predictable value by Ampere’s law Bo = µ0NeIe/(2 g), with the air-gap length
g = 10 mm.
In the two following simulations, the machine is fed by a d-axis stator
current only and by a q-axis stator current only, respectively. In the stator
conducting sheet, a linear current density is imposed with the suitable dis-
tribution. In order to have a sinusoidal MMF distribution with the maximum
value right on the d-axis, the following linear current density must be
imposed:

( ) ( )
J sd ϑ r = J dM sin ϑ r (8.34)

where ϑr is the rotor coordinate and the positive sign means a current
direction equal to the z-axis direction.
Similarly, to have a sinusoidal MMF distribution with the maximum value
centered right on the q-axis, the following linear current density must be
imposed:

( ) ( )
J sq ϑ r = − J qM cos ϑ r (8.35)

Figure 8.14 shows the simulation with d-axis current density distribution.
It is useful to consider the stator coordinate ϑs = ϑm + ϑr (see Figure 8.12).
Since the rotor position is ϑm = π/2 (see Figure 8.1), a linear current density
Jsd(ϑs) = JdMcos(ϑs) has to be imposed. A negative current value has been
fixed, which is a demagnetizing current, so that JdM = −20000 A/m. Figure
8.14 shows the corresponding flux plot and the air-gap flux density distri-
bution. The flux density distribution looks like a sinusoidal distribution in
correspondence to the pole shoes, while it decreases elsewhere, due to the
decrease of the permeance of the magnetic paths.
160 Electrical Machine Analysis Using Finite Elements

Air-gap flux density (T)

1
0.8
0.6
0.4
0.2
0
−0.2
−0.4
−0.6
−0.8
−1
0 90 180 240 360
Angle (electr. degrees)

FIGURE 8.14
Flux plot (a) and the air-gap flux density distribution (b) with d-axis current only.

Air-gap flux density (T)

1
0.8
0.6
0.4
0.2
0
−0.2
−0.4
−0.6
−0.8
−1
0 90 180 240 360
Angle (electr. degrees)

FIGURE 8.15
Flux plot (a) and the air-gap flux density distribution (b) with q-axis current only.

Similarly, Figure 8.15 shows the flux plot and the air-gap flux density
distribution when the machine is fed by a q-axis current. A linear current
density Jsq(ϑs) = −JqMsin(ϑs) is assigned with JqM = 20000 A/m (positive). Once
again, the flux density distribution looks like a sinusoidal distribution in
correspondence with the pole shoes, while it decreases elsewhere. Compar-
ing the flux density distributions of Figure 8.14 and Figure 8.15, the reduction
of the flux density corresponding to the symmetry axis is almost 10 times.
Figure 8.16 shows the flux plots and the air-gap flux density distributions,
with magnetizing current in the rotor and d-axis current in the stator. Because
of the negative d-axis current, the demagnetizing effect of the stator current
is evident.
Figure 8.17 shows the flux plots and the air-gap flux density distributions,
with magnetizing current in the rotor and q-axis current in the stator. In this
Synchronous Generators 161

Air-gap flux density (T)

1
0.8
0.6
0.4
0.2
0
−0.2
−0.4
−0.6
−0.8
−1
0 90 180 240 360
Angle (electr. degrees)

FIGURE 8.16
Flux plot (a) and air-gap flux density distribution (b) with magnetizing and d-axis currents.

Air-gap flux density (T)

1.5

0.5

−0.5

−1

−1.5
0 90 180 240 360
Angle (electr. degrees)

FIGURE 8.17
Flux plot (a) and air-gap flux density distribution (b) with magnetizing and q-axis currents.

case, the distorting effect of the stator current is evident. Since the material
has a constant µr , the saturation does not occur.
Finally, Figure 8.18 shows the flux plot and the air-gap flux density distri-
bution, when the rotor winding is fed by the magnetizing current, and the
stator winding is fed by both d- and q-axis currents. Comparing Figure 8.18
and Figure 8.13, the effect of the armature reaction is evident.

8.10 Appendix: The Transformation abc-dq

The analysis of the three-phase machines is simplified using a change of
variables for the electrical quantities, as currents, voltages, and flux linkages.
162 Electrical Machine Analysis Using Finite Elements

Air-gap flux density (T)

1
0.8
0.6
0.4
0.2
0
−0.2
−0.4
−0.6
−0.8
−1
0 90 180 240 360
Angle (electr. degrees)

FIGURE 8.18
Flux plot (a) and air-gap flux density distribution (b) with magnetizing, d- and q-axis currents.

The physical meaning of the procedure is to replace the three windings,

fixed to the stator and displaced of 2π/3 electrical radiants (see Figure 8.1),
with two windings, rotating at the same speed of the rotor. They are dis-
placed of π/2 electrical radiants, and their axes correspond to the d- and the
q-axis respectively, as shown in Figure 8.19. These two windings are equiv-
alent to two stator windings connected to a collector, with the rotating
brushes along the lines of the d- and the q-axis. The new reference frame
(d, q) is called the synchronous rotating reference frame, while the initial refer-
ence frame (a, b, c) is called the stationary reference frame.
The transformation does not have to modify the magnetic quantities in
the air-gap of the machine, i.e., the MMF and flux density distributions. In
other words, an observer located in the rotor should not notice if the stator
is fed by three stationary windings a, b, and c, or by two rotating windings
d and q. The three currents flowing in the three-phase winding, i.e., ia, ib, ic,
yield to three MMFs fa, fb, fc that are distributed in the air-gap with the
maximum value along the axis of the phases a, b, and c, respectively (see
Figure 8.1). The resulting MMF distribution may be conveniently split in
two distributions, fd, fq, with the maximum value along the d- and the q-axis,
respectively (see Figure 8.18).
Let ϑ be the electrical angle individuating the position of the d-axis with
respect to the a-phase axis, i.e., ϑ = pϑm, the transformation required to obtain
the MMFs. fd, fq from the MMFs fa, fb, fc is given by

2  2π   4π  
fd =  fa cos ϑ + fb cos  ϑ −  + fc cos  ϑ − 
3  3   3  
(8.36)
2  2π   4π  
fq = −  fa sin ϑ + fb sin  ϑ −  + fc sin  ϑ − 
3  3   3  
Synchronous Generators 163

Rotor coordinate angle

b-phase axis ϑr Polar axis
Interpolar axis (d-axis)
(or quadrature axis) Angular position
−q ϑm between
+d rotor and stator
d-axis a-phase axis
winding ω
rotation
q-axis
winding +q −d

c-phase axis

FIGURE 8.19
Synchronous rotating d- and q-axis windings.

or, using vector notation, written as

  2π   4π  
 cos ϑ cos  ϑ − cos  ϑ −  f 
 fd  2   3   3    a 
 =  ⋅  fb  (8.37)
 fq  3  2π   4π    
 − sin ϑ − sin  ϑ − − sin  ϑ −  fc
  3   3    

The transformation matrix is indicated by Tabc/dq. The phasor fd + jfq is

called the space phasor. The same transformation (8.37) is applied to the
currents ia, ib, ic and to the voltages va, vb, vc obtaining id, iq and vd, vq,
respectively. It is possible to verify that the voltage equations in the synchro-
nous reference frame are given by

dλ d
vd = Ri d + − ωλ q
dt
(8.38)
dλ
vq = Ri q + q + ωλ d
dt

where R is the winding resistance, ω is the electrical speed, given by ω =

dϑ/dt. The signs are reported with the notation of the motor. Finally, λd and
λq are the d- and q-axis flux linkages. In the case of linear conditions, they
are expressed as

λ d = Ld i d + λ dm
(8.39)
λ q = Lqi q
164 Electrical Machine Analysis Using Finite Elements

Ld is the d-axis synchronous inductance, and Lq is the q-axis synchronous

inductance. The relationship between the Ld, Lq and the three-phase self-
inductance is given in Equation (8.28). Then, λdm is the flux linked by the
d-axis winding due to the magnetizing current only. Since the d-axis and the
q-axis are orthogonal, there is no q-axis flux linkage due to the magnetizing
current (see Figure 8.18).
At last, the electromagnetic torque developed by the 2p-pole synchronous
machine can be expressed as

T=
3
2
(
p λ d i q − λ qi d ) (8.40)

Some advantages of the transformation (8.36)–(8.37) are

1. Since the d- and q-axis are rotating with the rotor, and the stator
reluctance variation is negligible (apart from almost negligible reluc-
tance variation due to the stator slotting), the self-inductance of the
two d- and q-axis windings is constant, unlike for the three-phase
a, b, c windings in which it is a function of the rotor position ϑm.
2. Since the d- and q-axis are at π/2 electrical radiants and posed on
two symmetry machine axes, they are not mutually coupled, as is
clear by Equation (8.39). Conversely, the a-, b-, and c-phase windings
are mutually coupled and this coupling is a function of the rotor
position ϑm as well. In reality, with high saturation, the d- and q-
axis windings may show a mutual coupling (the so-called cross cou-
pling). This phenomenon will be analyzed in Chapter 10.
3. During steady-state operations, the electrical quantities in the sta-
tionary reference frame a, b, c are sinusoidal in time, while the
electrical quantities in the rotating reference frame d, q are constant.
For instance, observe the voltage components along the d- and
q-axis, reported in Equation (8.32).
4. As a consequence of the points (1) and (2), the dynamic analysis of
the machine in the synchronous reference frame is very simplified.

Finally, let us point out that the factor in Equation (8.36) and Equation
(8.37) could be not only 2/3, even though such a choice is the most used in
the literature. With this choice, it is

1. The rotating d- and q-axis windings are characterized by a number

of series conductors equal to (3/2)N, where N is the number of series
conductors of each a-, b-, or c-phase winding.
2. The maximum value of the electrical quantities (voltage, current,
and flux linkage) in the two reference frames, a, b, c and d, q, is the
same. In particular, during steady-state operations, the magnitude
Synchronous Generators 165

of the space phasor Id + jIq corresponds to the maximum value of

the sinusoidal waveforms ia, ib, ic.
3. The transformation is not conservative, as can be observed from the
presence of the factor 3/2 in the torque equation (8.40).

References
1. R.H. Park, “Two Reaction Theory of Synchronous Machines, Part I,” in Trans.
of American Institute of Electrical Engineers, vol. 48, p. 716, 1929.
2. R.H. Park, “Two Reaction Theory of Synchronous Machines, Part II,” in Trans.
of American Institute of Electrical Engineers, vol. 52, p. 352, 1933.
3. M. Liwschitz-Garik and C.C. Whipple, Electric Machinery, Vol. I–II, D: Van
Nostrand Co., New York, 1947.
4. C. Concordia, Synchronous Machines, John Wiley & Sons, New York, 1951.
5. G. Kron, Equivalent Circuits of Electric Machinery, John Wiley & Sons, New York,
1951.
6. W.V. Lyon, Transient Analysis of Alternating Current Machinery, John Wiley &
Sons, New York, 1954.
7. A.S. Langsdorf, Theory of Alternating Current Machinery, McGraw-Hill, New
York, 1955.
8. B.J. Chalmers, Electric Motor Handbook, Butterworth, London, 1988.
9. A.E. Fidzgerald, C. Kingsley, Jr., and S.D. Umans, Electric Machinery, McGraw
Hill, New York, 1983.
10. S.A. Nasar (editor), Handbook of Electric Machines, McGraw-Hill, New York,
1987.
9
Surface-Mounted Permanent Magnet Motors

This chapter deals with the study of a three-phase surface-mounted perma-

nent magnet motor. Both squarewave and sinewave current-fed motors are
considered. In particular, the torque ripple as a function of the rotor position
is investigated.

9.1 Introduction
The surface-mounted permanent magnet motor can be considered as a syn-
chronous machine where the magnetizing winding is replaced with a per-
manent magnet. Thanks to the actual performance of the recent permanent
magnet materials, such motors exhibit high efficiency and high torque-to-
volume ratio.
In addition, the permanent magnet motor may be designed in different
shapes. They may be built with high length-to-diameter ratio, when high
speed and low inertia are required, e.g., for machine tools, or with low
length-to-diameter ratio, when low speed and high torque are required, e.g.,
for direct drives and motor-wheel-in-traction applications. In recent years,
permanent magnet generators and motors were developed; however, the
following analysis deals with the permanent magnet motors.
The magnetic structure of two configurations of a four-pole surface-
mounted permanent magnet motor is shown in Figure 9.1. The first config-
uration represents the classical solution with an inner rotor and an outer
stator; the second configuration is with an outer rotor and an inner stator.
In both configurations, the permanent magnets are radially or parallel mag-
netized, and produce an almost rectangular waveform of air-gap flux density.
The d-axis corresponds to the polar axis in the middle of the rotor pole. The
q-axis is leading the d-axis of π/2 electrical radians.
In order to highlight the motor versatility to be designed in various shapes,
Figure 9.2 shows a radial-flux surface-mounted permanent magnet motor
configuration. This is formed by two flat cylinders over which the permanent
magnets, which are axially magnetized, are placed. Within the two permanent

167
168 Electrical Machine Analysis Using Finite Elements

d d
q q

FIGURE 9.1
Surface-mounted permanent magnet motor configurations: (a) with inner rotor and (b) outer
rotor.

(a) (b)

FIGURE 9.2
Axial flux surface-mounted permanent magnet motor configuration.

magnet cylinders, a stator cylinder has been inserted, holding the three-
phase armature windings. Figure 9.2(b) shows a detail of the rotor cylinder.
The surface-mounted permanent magnet motors fed by an electrical drive
are classified as: (1) squarewave current-fed motors (also called trapezoidal
brushless or dc brushless motors), and (2) sinewave current-fed motors (also
called sinusoidal brushless or ac brushless motors).
They are characterized by different winding distributions in such a way
as to obtain a different induced EMF waveform. In the squarewave current-
fed motors, the EMF waveform should be ideally trapezoidal and the windings
are fed by squarewave currents, which are synchronized with the EMFs. The
ideal waveforms of the three-phase EMFs and the corresponding forced
currents are shown in Figure 9.3. In the sinewave current-fed motors, the
induced EMF should be ideally sinusoidal and the windings are fed by
sinewave currents, synchronized with the EMFs. In both cases, the instan-
taneous motor torque is ideally constant.
Surface-Mounted Permanent Magnet Motors 169

ea t

eb t

ec t
(a)

ia t

ib t

ic t
(b)

FIGURE 9.3
Ideal EMF (a) and current (b) waveforms in a trapezoidal brushless motor.

Neglecting the edge effects and supposing an identical behavior of the

magnetic field along the whole machine length, a two-dimensional analysis
is carried out in the (x, y) plane. As for the synchronous generators (see
Chapter 8), the machine length is the net iron length LFe. The magnetic vector
potential and the current density have only a z-axis component, which is
A = (0, 0, Az) and J = (0, 0, Jz); then the flux density and the field strength
vectors have components only on the plane (x, y), which are B = (Bx, By, 0)
and H = (Hx , Hy , 0).
Along the magnetization direction, the permanent magnets are character-
ized by a B-H curve like that shown in Figure 9.4. This shows the residual
flux density Bres, the coercive force Hc, and the knee magnetic field strength
Hknee. A generic working point is represented by the couple of values (Ho, Bo).
In order to avoid an irreversible demagnetization, the magnetic field strength
cannot go down below Hknee. Since the permanent magnet material works
in the linear part of the characteristic, the curve is approximated with a
straight line, as also shown in Figure 9.4. Then, for the permanent magnet

B = Bres + µrµ0 H Bres

Bo

H
Hc Hknee Ho

FIGURE 9.4
B-H curve of the permanent magnet and straight-line approximation.
170 Electrical Machine Analysis Using Finite Elements

a constant relative differential permeability µr is assumed. The flux density

vector in the permanent magnets is then given by

B = Bres + µ rµ 0H (9.1)

where Bres = (Bres,x, Bres,y , 0). Then the two-dimensional magnetostatic problem
is described by the differential equation

∂  1 ∂A z  ∂  1 ∂A z  ∂  B res , y  ∂  B res ,x 
+ = −µ 0 J z − + (9.2)
   
∂x  µ r ∂x  ∂y  µ r ∂y  ∂x  µ r  ∂y  µ r 

The symmetry between the pole pairs and the corresponding periodic
boundary conditions are the same of the synchronous generator; see Chapter 8.

9.2 Computation of the No-Load Characteristic

The permanent magnet rotor produces an air-gap magnetic flux that links
the stator winding. When the rotor moves, an EMF is induced in the stator
windings that corresponds to the rate of change of the flux linkage.
For the computation of the no-load flux linkage, the permanent magnet is
magnetized (the B-H curve of Figure 9.4 is used), while the stator winding
is open-circuited, which means a null conductivity is considered. Stator and
rotor laminations are defined by the B-H curves of the used magnetic material.
As far as the boundary conditions are concerned, a null magnetic vector
potential is fixed along the external stator circumference and along the d-
axis (Dirichlet’s condition), while a null normal derivative of the magnetic
vector potential is assigned along the q-axis (Neumann’s condition). Such
boundary conditions are represented in Figure 9.5.

Dirichlet Az = 0 Az = 0
Az = 0

dAz dAz
=0 =0
dn dAz dn
=0 Az = 0
dn Az = 0
Az = 0
Neumann

(a) (b) (c)

FIGURE 9.5
Boundary conditions in the no-load simulation.
Surface-Mounted Permanent Magnet Motors 171

The three-phase winding is placed within the stator slots, and may be
single-layer, full-pitched winding or a double-layer, chorded winding. The
first solution is mainly adopted with squarewave current-fed motor, in order
to obtain EMFs with a trapezoidal waveform. The second solution is adopted
in sinewave current-fed motor, in order to obtain EMFs with a sinusoidal
waveform. Some examples of such windings are shown in Figure 8.4 and
Figure 8.5 of Chapter 8.

9.2.1 Computation on the Solved Structure

9.2.1.1 Flux Linkage
Once the field solution is obtained, the magnetic vector potential Az(x, y) is
known in each point of the domain. The magnetic flux due to the permanent
magnet, which is linked with the j-th phase winding, is

Q/2 p

∑k ∫
nq 1
Λ j, pm = 2 pL Fe jq A zdS j = a , b, c (9.3)
n pp q= 1
Sq Sq

where nq is the number of the conductors in the slot, npp is the number of
parallel paths of the winding, Q/2p is the number of slots per pole, and kjq
assumes the values 0, ±1, or ±0.5, depending on the orientation and whether
the coils in the q-th slot belong to the j-th phase (see Chapter 8, Section 8.2.1.1).

9.2.1.2 Induced EMF

In case of sinewave current-fed motors, the EMF waveform can be assumed
to be sinusoidal. Hence, the EMF may be easily estimated as follows. For
the sake of convenience, the a-phase is chosen so as to link the maximum
magnetic flux (the a-phase axis correspondent to the d-axis). Let Λa,pm be the
maximum flux linkage. Then, at constant rotor speed ωm = ω/p, the RMS
value of the a-phase EMF is

1
Ea = ωΛ a , pm (9.4)
2

From the same field solution, supposing a one-slot step rotation, it is

possible to find some points of the EMF waveform, as described in Chapter
8, Section 8.2.1.2.
In case of squarewave current-fed motors, the following procedure is
required.

9.2.2 Computation for Various Rotor Positions

9.2.2.1 Flux Linkage and Induced EMF
For a rapid estimation of the flux linkages, one field solution is enough. The
rotor is fixed in a convenient position with respect to the stator, usually with
172 Electrical Machine Analysis Using Finite Elements

one winding linking the maximum magnetic flux, as shown in the previous
section.
Conversely, a more accurate computation should make provision for an
analysis at various rotor positions. In this way, it is possible to built the flux
linkage waveform for each phase winding, as a function of the mechanical
angle ϑm, i.e., λpm(ϑm). Therefore, the fundamental and the higher-order
harmonics of the flux linkage can be obtained by means of the Fourier series
expansion. It yields

Nk

λ pm (ϑ m ) = ∑Λ
k =1
pm , k cos(kϑ m ) (9.5)

where the higher order of the series is Nk. In the series expansion, only cosi-
nusoidal waveforms have been adopted, thanks to a suitable choice of the
reference angle ϑm = 0. Really, this assumption is not valid in the case of
fractional-slot winding without symmetry between the North and South poles.
The no-load EMF is obtained by deriving the flux linkage with respect to
the time. Assuming a constant rotor speed, it is practical to derive the flux
linkage with respect to the mechanical angle ϑm, and to multiply by the
mechanical speed ωm, which is

dλ pm (ϑ m ) dλ pm (ϑ m ) dϑ m dλ pm (ϑ m )
e(ϑ m ) = = = ωm (9.6)
dt dϑ m dt dϑ m

Since the flux linkage in Equation (9.6) is derived numerically, this oper-
ation may give rise to errors. It is convenient to express the flux linkage
using the Fourier series expansion, as in Equation (9.5), and then to compute
the EMF as the series of the derivatives of each flux linkage harmonic, as in

Nk

e(ϑ m ) = ω m ∑ −k Λ
k =1
pm , k sin(kϑ m ) (9.7)

Of course, the EMF can be expressed as a function of the electrical angle

ϑ, replacing ϑm = ϑ/p, or as a function of the time, replacing ϑm = ωmt.

9.2.2.2 Cogging Torque

In the various positions of the rotor, it is also possible to compute the cogging
torque. This is due to the interaction between the permanent magnets and
the stator slotting. Such torque may be computed by means of Maxwell’s
stress tensor or as the derivative of the magnetic energy with the position.
In the first case, the torque is obtained by integrating Maxwell’s stress
tensor along a surface containing the rotor. Because of the two-dimensional
problem, this corresponds to integrating Maxwell’s stress tensor along a line
Surface-Mounted Permanent Magnet Motors 173

lg in the middle of the air-gap, and then multiplying the result by the active
length of the rotor, LFe. Assuming a 2p-pole machine, it yields

D − g L Fe
Tcog =
2 µ0
p
∫ B B dl
lg
r ϑ (9.8)

where Br is the radial component of the flux density (normal to the line lg),
Bϑ is the azimuthal component of the flux density (tangential to the line lg),
D is the inner stator diameter, and g is the air-gap length. The number of
pole pairs p is used assuming that the simulation has been carried out on
two pole-pieces of the machine only. A different factor has to be used with
different simulation (e.g., 2p should be used instead of p when only one
pole-piece is simulated).
Because of the numerical nature of the finite element method, the result
may depend on both the position of the integration line and the number of
points chosen for the numerical integration. Instead of Equation (9.8), it is
better to compute the average value of the torque over the entire air-gap
surface Sg. Then the torque is

∫
L Fe
Tcog = p rB rB ϑdS (9.9)
gµ 0 Sg

where r is the dummy radius.

In the second case, the cogging torque may be computed from the variation
of the magnetic energy Wm variation with respect to the rotor angular position
ϑm. On open-circuit the stator currents are null, hence there is no electrical
energy exchange between the motor and external sources. Corresponding to
an elementary rotation dϑm, the sum of the variation of the magnetic energy
stored in the system and the work done by the magnetic field is null: dWm +
Tcogdϑm = 0. Then, the cogging torque is equal to the negative derivative of
the magnetic energy with respect to the rotating angle ϑm.
The computation of the magnetic energy in the air and in the iron is carried
out in the classical way. The computation of the energy in permanent magnet
materials is discussed later in this chapter. It is convenient to omit the work
of magnetization (which is an irreversible process) and to refer the magnetic
energy density to the point (0, Bres) of the B-H curve, as reported in Figure
9.6(a). Since the torque depends on a difference of the energy, this useful
assumption does not cause any error. As a consequence, the energy in the
permanent magnet is defined as

B
Wm( pm ) =
∫ ∫
τ pm 0
H ⋅ dB dτ
(9.10)
B
= L Fe
∫ ∫
S pm 0
H ⋅ dB dS
174 Electrical Machine Analysis Using Finite Elements

B B B

Bres Bres Bres

Bo
∆B ∆B

B
H H H
Ho ∆H ∆H
(a) (b) (c)

FIGURE 9.6
Magnetic energy and coenergy density in permanent magnets.

Bearing in mind the linear approximation of the B-H curve, given by

Equation (9.1), supposing that H is parallel to Bres, then Equation (9.10)
becomes

∫ (B )
1
Wm( pm ) = L Fe 2
− B resB dS
2µ r µ 0 S pm
(9.11)
µµ
= r 0 L Fe
2 ∫ S pm
2
H dS

In an analogous way, the magnetic coenergy is defined as

H
Wm′ ( pm ) =
∫ ∫
τ pm 0
B ⋅ dH dτ

∫
1
= L Fe B 2 dS (9.12)
2µ r µ 0 S pm

2
µ rµ 0  B 
=
2
L Fe
∫ 
S pm 
H + res  dS
µ rµ 0 

The variation of the magnetic energy density corresponding to any rotation

of the rotor is shown in Figure 9.6(b). The variation of the magnetic coenergy
density in the permanent magnet is shown in Figure 9.6(c).
Finally, from the magnetic energy Wm of the whole system, the cogging
torque is given by

dWm
Tcog = − (9.13)
dϑ m

In the computation of the torque at various rotor positions the numerical

errors caused by the mesh changes should be avoided. Such errors may be
Surface-Mounted Permanent Magnet Motors 175

(a)

(b) (c)

FIGURE 9.7
Tricks to rotate the structure without mesh change, so as to avoid numerical errors.

so high that they significantly alter the solution, especially when low torques
are investigated (such as the cogging torque).
To avoid or to reduce such errors, two tricks are suggested.
The first is to divide the permanent magnet in various pieces, as shown
in Figure 9.7(a). The length of each piece should correspond to the minimum
rotation that has been fixed. According to the rotor position, the magnetic
B-H curve of the air or of the permanent magnet is assigned to each piece.
In this way, a fictitious rotation is obtained, except that it operates on the
material definition and not on the drawing of the motor. As a consequence,
the mesh of the domain remains always the same.
The second trick is to split the air-gap into two parts, as illustrated in
Figure 9.7(b). Then the rotor and the corresponding air-gap part is rotated,
keeping fixed the other air-gap part together with the stator, as shown in
Figure 9.7(c). The mesh remains the same, fixed to the structure, both in the
movable and in the fixed part. Of course, for each position, the boundary
conditions have to be adapted to the exact domain contour. It is worth
noticing that the boundary line between the two air-gap parts must be
divided properly, in such a way as the stator mesh correctly matches the
rotor mesh in each rotor position.

9.3 Computation of the Inductances

The computation of the inductances requires a magnetostatic field simula-
tion. By geometric and electric symmetry, only one pole-piece of the machine
is usually considered.
176 Electrical Machine Analysis Using Finite Elements

With the hypothesis of linear operating conditions (which is often verified

in the surface-mounted permanent magnet motors), the permanent magnet
is assumed to be demagnetized, i.e., with null residual flux density Bres. Thus,
the magnetic flux due to the permanent magnet results in being null. The
stator windings are fed so that the flux lines are normal or tangential to the
boundary of the domain, allowing only Neumann’s or Dirichlet’s boundary
conditions to be assigned.
The inductances can be computed from the magnetic energy, from the flux
linkage, or from the air-gap flux density distribution, with the same proce-
dures presented in Chapter 8, Sections 8.3 and 8.4.

9.4 Computation of the Torque

The motor torque is caused by the interaction between the permanent mag-
net and the stator currents. For the sake of generalization, the stator currents
are considered to be independent of the rotor position ϑ. The analysis is
reduced to a part of the machine, usually one pole or two poles according
to the stator geometry. Of course, suitable boundary conditions are assigned.
The computation of the torque is carried out in different ways:

1. by means of Maxwell’s stress tensor

2. by means of the virtual work principle, evaluating the changes of
magnetic energy in two different rotor positions
3. by means of the interaction between flux linkage and current com-
ponents in the synchronous rotating reference frame

The three ways for computing the motor torque are described next.

9.4.1 Computation by Means of Maxwell’s Stress Tensor

As explained in the previous chapter, Maxwell’s stress tensor is integrated
on an arc in the middle of the air-gap, and the result is multiplied by the
radius and the motor stack length LFe. However, as in Equation (9.9), to
reduce the numerical error, the torque is computed as the average value on
the whole air-gap surface Sg, as

∫
L Fe
T= p rB rB ϑdS (9.14)
gµ o Sg

The pole pairs number p is used in Equation (9.14), assuming that the
simulation is carried out again on two pole-pieces of the machine.
Surface-Mounted Permanent Magnet Motors 177

9.4.2 Computation by Means of the Virtual Work Principle

This method is based on the balance of the energy change corresponding to
an elementary rotation of the rotor. The different energy terms are computed
by integrating the volume magnetic energy density on the various compo-
nents of the domain.
In nonlinear conditions, the magnetic coenergy has to be adopted (see
Chapter 5). The torque is computed as

dWm′
T=+ (9.15)
dϑ m i = const

where the magnetic coenergy in the permanent magnet material is computed

as indicated by Equation (9.12).
As highlighted previously, this method requires at least two field simula-
tions, with the rotor in two different positions.

9.4.3 Computation by Means of Flux Linkages and Currents

At last, the motor torque can be computed from the interaction between the
flux linkages and the current. It is advantageous to consider these quantities
in the synchronous rotating reference frame, by means of the transformation
Tabc/dq (see the Appendix in Chapter 8). The torque is

T=
3
2
(
p Λ d I q − Λ qI d ) (9.16)

where the d- and q-axis currents are given by Equation (8.30), and the d-
and q-axis flux linkages are given by Equation (8.29).
The angle ϑ is the electrical angle between the reference a-phase axis and
the d-axis. The flux linked with the j-th phase (j = a, b, c) is obtained from
the average value of the magnetic vector potential Az over the slot surfaces SCu.
It is worth bearing in mind that the torque computed by Equation (9.16)
is obtained as the energy balance at the terminals of the motor windings.
Such a balance is based on the assumption of sinusoidal distributed wind-
ings. As a consequence, the average torque is correct, but the torque ripple
is lower. It could be verified that MMF harmonics are only partially included
in Equation (9.16). In addition, such a computation method does not consider
the cogging torque due to the interaction of the permanent magnet and the
stator slotting. In fact, the torque computed from Equation (9.16) is always
null when the stator currents are null (i.e., Id = Iq = 0).

9.4.4 Dynamic Computation

Up to now, the torque computation has been carried out referring to the
stator currents equal to their fixed ideal values. Because of the winding
178 Electrical Machine Analysis Using Finite Elements

3 2
2 1.5

Torque (Nm)
Current (A)

1
1
0 Reference (A)
0.5 ia(t)
2
−1 ia(t)
−2 0 0
ia(t) ib(t) ic(t)
−3 −0.5 −2
0 2 4 6 8 10 0 2 4 6 8 10
Time (ms) Time (ms)

FIGURE 9.8
Currents and torque waveforms.

inductances and the induced EMFs, together with the limited value of the
source voltage, the electrical drive is not always able to feed the motor with
ideal currents. On the contrary, the currents often follow their references
with a delay.
This delay is mainly appreciable in squarewave current-fed motors, since
the currents cannot exhibit rapid changes as those required by the ideal
currents shown in Figure 9.3(b). The current is as distorted as the rotor speed
is high. Consequently, a constant torque is not obtained, but a torque ripple
takes place, which is called the commutation effect. Figure 9.8(a) shows the
typical three-phase current waveforms, obtained by a dynamic simulation
at a supply frequency f = 100 Hz. In the figure, the rise and fall times of the
current are clear. The delay of the phase current with respect to its reference,
highlighted in Figure 9.8(b), causes the torque ripple, shown in Figure 9.8(b)
as well.
For the computation of the instantaneous torque, a simple procedure is
adopted. All the motor parameters are known, i.e., the resistance R, self- and
mutual inductances L and M, the induced EMF waveform e(ϑm). With fixed
motor speed, the latter may be expressed as a function of the time, as ϑm =
ωmt. Using a two-level inverter, the voltage forced at the motor terminals
can assume the two values v = −Vdc or v = +Vdc (v = 0 using a three-level
inverter), where Vdc is the dc bus voltage. The voltage level is chosen by the
control comparing the current reference (ideal current) with the effective
current.
The dynamic analysis is carried out, from the step integration of the dif-
ferential equation

di j (t )
v j (t ) = Ri j (t ) + (L − M) + e j (t ) + v N (t ) (9.17)
dt

for each j-th phase winding (j = a, b, c). In Equation (9.17), vN(t) indicates
the star-center potential, which is computed at each step from the forced
terminal voltages.
Surface-Mounted Permanent Magnet Motors 179

9.5 Example
This section discusses some results of finite element analysis of a surface-
mounted permanent magnet motor. The motor data are reported in Table 9.1;
the magnetic structure and the winding are sketched in Figure 9.9. It is a
squarewave current-fed brushless motor with a low number of slots per pole
and a chorded winding. Such a solution is adopted in low-power applica-
tions, with the aim of having minimum end-winding length and higher
motor efficiency, in spite of a higher torque ripple.
With such a motor structure, a two-pole section of the machine is simu-
lated, using the periodic conditions on the two boundaries. Different rotor
positions are simulated splitting the permanent magnet in elementary pieces,
as shown in Figure 9.7(a).
A rotation of 40 mechanical degrees (120 electrical degrees) is considered
although a rotation of 30 mechanical degrees would be enough.

9.5.1 No-Load Simulation

Figure 9.10 shows the flux lines at no-load when the rotor is in the position
ϑm = 10 mechanical degrees.

TABLE 9.1
Brushless Motor Data

Stator Data

De = 125 (mm) stator external diameter

D = 61 (mm) stator inner diameter
LFe = 38 (mm) stack length
2p = 6 number of poles
Q=9 number of slots
N = 1458 number of series conductors per phase

Stator Slot Data

wso = 3.55 (mm) width of slot opening

wt = 7.5 (mm) width of stator tooth
wso = 4 (mm) height of slot opening
hs = 22 (mm) total height of slot

Rotor Data

Dr = 44.7 (mm) rotor external diameter

Dsh = 18 (mm) shaft diameter

Permanent Magnet Data

lm = 7.5 (mm) permanent magnet thickness

2αpm = 55 (deg) mechanical permanent magnet angle
Bres = 0.4 (T) residual flux density (Ferrite)
µr = 1.05 (-) differential relative permeability
180 Electrical Machine Analysis Using Finite Elements

a1
b1
c3 a1 b1 c1

c1

b3 a2 b2 c2

a2
a3 b3 c3
a3

b2
c2

FIGURE 9.9
Brushless motor magnetic structure.

FIGURE 9.10
Flux lines at no-load operation; a rotation of ϑm = 10 mechanical degrees is considered.

The three-phase flux linkages are shown in Figure 9.11(a). They are approx-
imately flat for a portion of 40 mechanical degrees (120 electrical degrees).
It follows that, with constant speed, the back EMFs exhibit a constant value
for a corresponding period of time (one third of the electrical period), as
required in a trapezoidal brushless drive; see Figure 9.3(a).
The EMF waveform is not exactly trapezoidal because of the chorded
winding. Thus, a constant torque is not achieved even when ideal square-
wave currents of Figure 9.3(b) feed the stator windings.
The magnetic energy is shown in Figure 9.11(b). The energy variation
indicates that, during the rotation, there is a cogging torque due to the
interaction of the permanent magnet with the stator slotting.
Surface-Mounted Permanent Magnet Motors 181

0.2 0.229
0.15 0.228

Magnetic energy (J)

Flux linkages (Vs)

0.1 0.227
0.05 0.226
0.225
0
0.224
−0.05 0.223
−0.1 0.222
−0.15 0.221
−0.2 0.22
0° 5° 10° 15° 20° 25° 30° 35° 40° 0° 5° 10° 15° 20° 25° 30° 35° 40°
Mechanical angle (degrees) Mechanical angle (degrees)

FIGURE 9.11
Flux linkages and magnetic energy during the no-load operation.

Maxwell
0.1 Energy 0.1
0.08 Coenergy
Cogging torque (Nm)

Cogging torque (Nm)

0.08
0.06 0.06
0.04 0.04
0.02 0.02
0 0
−0.02 −0.02
−0.04 −0.04
−0.06 −0.06
−0.08 −0.08
−0.1 −0.1
0° 5° 10° 15° 20° 25° 30° 35° 40° 0 1 2 3 4 5 6
Mechanical angle (degrees) Time (s)

FIGURE 9.12
Cogging torque: simulated (a) and measured (b) values.

The cogging torque behavior is shown in Figure 9.12(a). It has been com-
puted in different ways by means of Maxwell’s stress tensor, using Equation
(9.9), drawn in a thin line; by means of the derivative of the magnetic energy,
using Equation (9.13), drawn in a bold line (almost indistinguishable from
the previous one); and by means of the derivative of the magnetic coenergy,
using Equation (9.15) with null currents, drawn using dots. It is worth
noticing the good match among the three methods.
At last, Figure 9.12(b) shows the comparison between the simulated and
the measured cogging torque. The measure system has been regulated to
complete a whole rotation in 6 seconds. The figure shows a satisfactory
agreement between simulated results and measurements. A small difference
can be attributed to a nonperfect symmetry of the rotor poles and to a
residual flux density slightly lower than that used in the simulations.

9.5.2 Simulations of Operations Under Load

The motor is fed by ideal squarewave currents, as depicted in Figure 9.3.
They are centered on the constant value of the corresponding EMFs and their
maximum value is 2 A.
182 Electrical Machine Analysis Using Finite Elements

2.5

Electromechanical torque (Nm)

2

1.5
Maxwell stress tensor
1 Balance in d-q ref. frame
Magnetic coenergy variation

0.5
Cogging torque
0
0° 5° 10° 15° 20° 25° 30° 35° 40°
Mechanical angle (degrees)

FIGURE 9.13
Electromechanic torque with ideal squarewave currents, I = 2 A.

The electromechanical torque behavior is shown in Figure 9.13. It is com-

puted by means of the methods described above. The thin line refers to the
integration of Maxwell’s stress tensor, as given by Equation (9.14). The circles
refer to the torque computed by the derivative of the magnetic coenergy
with constant currents, as in Equation (9.15). With squarewave currents, the
stator currents are constant for 20 mechanical degrees, so that this method
is easily applied.
Finally, the bold line refers to the torque computed from the flux linkages
and currents in the synchronous reference frame, as in Equation (9.16). As
pointed out earlier, it is observable that the torque computed by the latter
method shows the same average value, but a lower ripple. For a comparison,
the same Figure 9.13 shows the cogging torque, which has been computed
before.

References
1. N. Demerdash et al., “Comparison between features and performance charac-
teristics of fifteen-hp samarium-cobalt and ferrite-based brushless d.c. motors,”
IEEE Trans. on PAS, vol. 102, pp. 104–112, 1983.
2. H. Weh, H. Mosebach, H. May, “Design concepts and force generation in
inverter-fed synchronous machines with permanent magnet excitation,” in
IEEE Trans on Magnetics, vol. MAG-20, no. 5, pp. 1756–1761, 1984.
3. B.J. Chalmers, S.A. Hamed, G.D. Baines, “Parameters and performances of a
high-field permanent-magnet synchronous motor for variable-frequency oper-
ation,” IEEE Proc. Pt. B, vol. 132, no. 3, 1985.
4. N. Boules, “Prediction of No-Load Flux Density Distribution in Permanent
Magnet Machines,” IEEE Trans. on IA, vol. 21, no. 4, pp. 633–643, 1985.
5. K. Kobayashi and M. Goto, “A brushless DC motor of a new structure with
reduced torque fluctuations,” Elec. Eng. Jpn, vol. 105, no. 3, pp. 104–112, 1985.
Surface-Mounted Permanent Magnet Motors 183

6. T. Kenjo and S. Nagamori, Permanent-Magnet and Brushless DC Motors, Oxford

University Press, 1985.
7. T. Nehl, N. Demerdash, and F. Fouad, “Impact of winding inductances and
other parameters on the design and performance of brushless d.c. motors,”
IEEE Trans. on PAS, vol. 104, pp. 2206–2213, 1985.
8. T. Li and G. Slemon, “Reduction of cogging torque in permanent magnet
motors,” IEEE Trans. on MAG, vol. 24, no. 6, pp. 2901–2903, 1988.
9. B.J. Chalmers, “Influence of saturation in brushless permanent-magnet motor
drives,” in IEEE Proc. Pt. B, vol. 139, pp. 51–52, January 1992.
10. Z.Q. Zhu, D. Howe, E. Bolte, and B. Ackermann, “Istantaneous Magnetic Field
Distribution in Brushless Permanent Magnet DC Motors, Part I: Open-Circuit
Field,” IEEE Trans. on MAG, vol. 29, no. 1, pp. 124–135, 1993.
11. N. Bianchi, S. Bolognani, “A Procedure for the Electromagnetic and Thermal
Integrated Design of PM Brushless Motors,” SPEEDAM, Taormina, 8–10
Giugno 1994, pp. 13–18.
12. N. Bianchi, S. Bolognani, “Optimal design of a PM brushless motor,” in Proc.
of International Conference on Electrical Machine, ICEM ’94, Parigi, 5–8 September
1994, pp. 204–-209.
13. N. Bianchi, S. Bolognani, “Analysis and design of a PM synchronous magnetic
drive,” in Proc. of International Conference on Electrical Machine, ICEM ’94, Parigi,
5–8 September 1994, pp. 341–346.
14. D.C. Hanselman, Brushless Permanent Magnet Motor Design, McGraw-Hill, Inc.,
1994.
15. J.M.D. Coey (editor), Rare-Heart Iron Permanent Magnet, Monographs on the
Physics and Chemistry of Materials, Oxford Science Publications, Claredon
Press, Oxford, 1996.
16. R.P. Deodhar, D.A. Staton, T.M. Jahns, and T.J.E. Miller, “Prediction of cogging
torque using flux-mmf diagram technique,” IEEE Trans. on IA, vol. 32, no. 3,
pp. 569–575, 1996.
17. N. Bianchi, A. Lorenzoni, “Permanent Magnet Generators for Wind Power
Industry: an Overall Comparison with Traditional Generators,” in Proceeding
of IEEE International Conference on Opportunities and Advance in Electric Power
Generation, Durham (UK), 1996, pp. 49–54.
18. N. Bianchi, S. Bolognani, “Design Techniques for Reducing the Cogging Torque
in Surface-Mounted PM Motors,” accettata per presentazione e pubblicazione
sui Proc. della conferenza internazionale IEEE IAS Annual Meeting, Rome, Italy,
8–12 October 2000.
19. R. Krishnan, Electric Motor Drives, Prentice-Hall, Englewood Cliffs, NJ, 2001.

Further References on AC Drives

1. W. Schuisky, Berechnung Elektrischer Machinen, Springer Verlag, Wien, 1960.
2. G.R. Slemon, A. Straughen, Electric Machines, Addison-Wesley, New York, 1980.
3. P. Sen, Thyristor DC Drives, John Wiley & Sons, New York, 1981.
4. E. Levi, Polyphase Motors, John Wiley & Sons, New York, 1984.
5. W. Leonard, Control of Electrical Drives, Springer-Verlag, Berlin, 1985 (first edi-
tion), 1996 (second edition).
184 Electrical Machine Analysis Using Finite Elements

6. P. Vas, Vector Control of AC Machines, Oxford Science Publications, Clarendon

Press, Oxford, 1990.
7. I. Boldea and S.A. Nasar, Vector Control of AC Drives, CRC Press, Boca Raton,
FL, 1992.
8. D.W. Novotny, T.A. Lipo, Vector Control and Dynamics of AC Drives, Oxford
Science Publications, Clarendon Press, Oxford, 1996.
9. J.F. Gieras and M. Wing, Permanent Magnet Motors Technology: Design and Appli-
cations, Marcel Dekker, Inc., New York, 1996.
10. I. Boldea and S.A. Nasar, Electric Drives, CRC Press, Boca Raton, FL, 1999.
10
Interior Permanent Magnet and Reluctance
Synchronous Motors

This chapter deals with the finite element analysis of an interior permanent
magnet motor. The interior permanent magnet motor is used in traction
applications or when high dynamic performance is required. It is a synchro-
nous motor, fed by sinewave currents via an electrical drive; therefore the
analysis of such a motor is carried out by means of the direct and the
quadrature axis theory. Two magnetic models of the motor are presented,
and the procedures for obtaining them are illustrated. An automatic algo-
rithm is also described to process the data obtained from the finite element
analysis. The synchronous reluctance motor is discussed at the end of the
chapter, considering such a motor as a particular case of the interior perma-
nent magnet motor.

10.1 Introduction
The key characteristic of the interior permanent magnet motor is that the
permanent magnets are not on the surface of the rotor but are buried in the
rotor laminations. Apposite holes are stamped in the rotor laminations to
hold the permanent magnets and to obstruct the d-axis flux due to the stator
current. For this reason, they are also called flux barriers.
Since the differential relative permeability of the permanent magnet is only
slightly higher than the air permeability, the rotor exhibits magnetic paths
with different magnetic permeance. Let the d-axis correspond to the polar
axis (i.e., the permanent magnet axis). Thus the d-axis permeance is lower
than the q-axis permeance, and the inductances are Ld < Lq. This is opposite
of the usual situation of the salient pole synchronous machines with wound
rotor (it is useful to compare with Chapter 8).
As a consequence of the rotor anisotropy, the motor takes advantage of
the reluctance torque and the permanent magnet torque, showing a favorable
torque-to-volume ratio. In addition, the interior permanent magnet motor is

185
186 Electrical Machine Analysis Using Finite Elements

d d
q q

N
N N S
S S S S
N N
S S
N N
S S
N N N
S N
S N
S

FIGURE 10.1
Different rotor structures of the interior permanent magnet motors.

appropriate for flux-weakening operations, when a constant power torque-

speed characteristic is required, since it maintains a high power up to high
operating speeds.
Although the stator is similar to that of an induction motor, the rotor may
assume different topologies, according to the permanent magnet position.
These configurations may be divided into two groups: rotor with tangential
(or azimuthal) magnetized permanent magnets and with radial magnetized
permanent magnets. They are shown for a 6-pole motor in Figure 10.1(a)
and Figure 10.1(b), respectively.
In the first case, the magnetic flux of each pole is given by the sum of the
flux of two permanent magnets. This configuration type is appropriate with
permanent magnets of low residual flux density and with machines of high
number of poles. The two permanent magnet surfaces are chosen so as their
sum is higher than the air-gap pole surface, in order to obtain a higher air-
gap flux density. In addition, the rotor shaft should be nonmagnetic to avoid
flux loops.
In the second case, since the permanent magnet surface is lower than the
pole surface, the air-gap flux density is lower than the permanent magnet
flux density. Moreover, such a configuration shows a high q-axis permeance
and a low d-axis permeance, which is a high rotor anisotropy.
In both configurations of Figure 10.1, there are magnetic bridges that are
essential for support the rotor structure against the centrifugal forces. How-
ever, their thickness should be as narrow as admitted by mechanical consid-
erations, since they are magnetic short-circuits of the PM flux.
Referring to a tangential (azimuthal) magnetization permanent magnet
motor, some flux density waveforms are reported in Figure 10.2.
Figure 10.2(a) shows the air-gap flux density distribution due to the per-
manent magnet only. The average value is highlighted by the dashed line.
Interior Permanent Magnet and Reluctance Synchronous Motors 187

Airgap flux density (T) Airgap flux density (T)

0.35 1
0.8
0.3
0.6
0.25 0.4
0.2
0.2 0
−0.2
0.15
−0.4
0.1 −0.6
−0.8
0.05 −1
0° 60° 120° 180° 0° 60° 120° 180°
Electric angle (deg) Electric angle (deg)
(a) (b)

Flux density along the

Airgap flux density (T) permanent magnet (T)
0
0.4 −0.02
0.2 −0.04
−0.06
0
−0.08
−0.2 −0.1
−0.12
−0.4
−0.14
−0.6 −0.16
−0.18
0° 60° 120° 180° 0 hm/3 2hm/3 hm
Electric angle (deg)
(c) (d)

FIGURE 10.2
Air-gap flux density distributions due to the permanent magnet (a), d-axis current (b), q-axis
current (c), and flux density distribution along the permanent magnet with negative d-axis
current (d).

The effect of the slot opening is well manifested. Figure 10.2(b) and Figure
10.2(c) illustrate the air-gap flux density distribution due to the stator cur-
rents only (the permanent magnet has been demagnetized, i.e., Bres = 0), with
only a d-axis current and only a q-axis current, respectively (the PM is
unmagnetized). The fundamental harmonic of the distribution is highlighted
by dashed lines. The slotting effect is again manifested, together with the
effect of the winding distribution. Because of the flux barriers, when a d-
axis current flows in the stator windings, some parts of the rotor assume a
magnetic potential different from zero. These rotor parts are the iron
“islands,” bordered by the flux barriers and the air-gap. As a consequence,
188 Electrical Machine Analysis Using Finite Elements

the maximum value of the flux density due to the d-axis current of Figure
10.2(c) is lower than that due to the q-axis current of Figure 10.2(b), although
the stator is fed by the same current.
Finally, Figure 10.2(d) shows the flux density distribution on the surface
of the permanent magnet, caused by a negative d-axis current only. The flux
density is negative (demagnetizing) and is uniformly distributed on the
whole surface. This check should be carried out at the maximum stator
current and is necessary to ensure that the permanent magnets are not
demagnetized irreversibly.

10.2 Characteristic Motor Parameters

The steady-state analysis of the interior permanent magnet motor is carried
out in the synchronous rotating reference frame. The motor torque is
expressed by Equation (9.16) and is reported as

T=
3
2
(
p Λ d I q − Λ qI d ) (10.1)

The d- and q-axis steady-state voltages are

Vd = RId − ωΛ q
(10.2)
Vq = RIq + ωΛ d

Neglecting the resistive voltage drop, the voltage magnitude is

(
V 2 = ω 2 Λd 2 + Λq2 ) (10.3)

The flux linkages Λd and Λq are always expressed by using the flux linkage
due to the permanent magnet Λm and the two axis inductances Ld and Lq.
In general, they are dependent on the stator current, due to the saturation.
These parameters represent the magnetic model of the motor and are easily
obtained from the field solution.

10.2.1 Simplified Magnetic Model

The flux linkage due to the permanent magnet is obtained from the no-load
simulation, as shown in Equation (9.3). Let Λm be its maximum value. It is
considered constant and independent of the stator currents.
Interior Permanent Magnet and Reluctance Synchronous Motors 189

d- and q-axis flux linkage (Vs)

0.14
Λd (Id)
0.12
#4 Λq (Iq)
0.1 #3
Lqs
0.08 Λm Ld
0.06 #1
#2 #2
0.04 Λqs

0.02 Lqu
0
−200 −100 0 100 200
d- and q-axis current (A)

FIGURE 10.3
Simplified magnetic model of the interior permanent magnet motor.

For computing the d- and q-axis inductances, the same procedure

described for the synchronous generator in Chapter 8 is adopted. They are
indicated as Ld and Lq (eventually corrected to consider the three-dimen-
sional effects, i.e., Ld = Lσ + Ldm and Lq = Lσ + Lqm). As explained earlier, due
to the particular rotor structure, it is Ld < Lq.
The d-axis inductance Ld is essentially constant, with the current. Then,
the d-axis flux linkage becomes

Λ d (I d ) = Λ m + L d I d (10.4)

On the contrary, the quadrature axis inductance Lq depends on the stator

current, in particular on the q-axis current, which is Lq(Iq). The q-axis flux
linkage is Λq = Lq(Iq)·Iq. For the sake of convenience, this relationship is
approximated using piece-wise lines. Letting Iqs be the q-axis current at the
beginning of the saturation, it is useful to define an unsaturated inductance
Lqu with Iq < Iqs and a saturated (differential) inductance Lqs with Iq > Iqs. The
q-axis flux linkage becomes

Λ q (Iq ) = Lqu Iq Iq ≤ Iqs

(10.5)
= Λ qs + Lqs Iq Iq ≥ Iqs

The magnetic characteristics that describe the dependence of the flux link-
age on the currents are shown in Figure 10.3. An advantage of this magnetic
model is that simple equations are employed. These are particularly advan-
tageous in the design step of the interior permanent magnet motor or of the
current control. Moreover, it is rapidly obtained from only four field solutions:
190 Electrical Machine Analysis Using Finite Elements

1. A first simulation with null stator currents allows the flux linkage
due to the permanent magnet, Λm, to be obtained.
2. A second simulation, with low d- and q-axis current values (an
indicative value could be Id = Iq equal to the 10% of the rated stator
current), allows the two unsaturated inductances to be computed.
Letting Λd and Λq be the flux linkages corresponding to the currents
Id and Iq, the two-axis inductances are

Λd − Λ m Λq
Ld = Lqu = (10.6)
Id Iq

3. Two further simulations, with a high value of the q-axis currents

only (for instance, equal to 80% and 100% of the rated current), allow
the differential saturated inductance Lqs to be evaluated. If Λ′q and
Λ″q are the q-axis flux linkages obtained with the two q-axis currents
I′q and I″q , respectively, the inductance results

Λ′q − Λ′′q
Lqs = (10.7)
I′q − I′′q

Then, referring to Figure 10.3, the remaining parameters are given by

Λ qs = Λ′q − Lqs I′q (10.8)

Λ qs
Iqs =
Lqu − Lqs

Figure 10.3 refers to the simulation of an interior permanent magnet motor

with radial magnetization, whose configuration is reported in Figure 10.1(b).
The values obtained from the field solutions are highlighted by dots. The
number close to each dot indicates the corresponding simulation step. The
same values are reported in Table 10.1.

TABLE 10.1
Computed Values for the Interior Permanent Magnet
Motor Model
Id (A) Iq (A) Λd (mVs) Λq (mVs) Motor Parameter
#1 0 0 67.5 — Λm = 67.5 (mVs)
#2 –85 85 46.8 46.2 Ld = 0.243 (mH)
Lqu = 0.543 (mH)
#3 0 192 — 95.05
#4 0 240 — 108.9 Lqs = 0.288 (mH)
Interior Permanent Magnet and Reluctance Synchronous Motors 191

10.2.2 Cross-Coupling Effect

The cross-coupling is the magnetic interaction between the d- and the q-axis,
which is essentially due to the saturation. This phenomenon may be inter-
preted as follows: the magnetic flux produced by the d-axis current saturates
some zones of the machine, distorting the flux lines produced by the q-axis
current, and vice versa.
The d- and q-axis flux linkages are considered to be functions of both the
two axis currents, which is

Λ d = Λ d (I d , I q )
(10.9)
Λ q = Λ q (I d , I q )

The flux linkages may be computed by integrating the magnetic vector

potential or by processing the air-gap flux density distribution, as described
in Chapter 8. An automatic procedure is also reported in Section 10.4.
In the case of linear magnetic circuit, the d- and q-axis flux linkages Λd
and Λq vary linearly with the d- and q-axis currents, which is

Λ d (Id , Iq ) = Λ d (Id ) = Λ m +Ld Id

(10.10)
Λ q (I d , I q ) = Λ q (I q ) = L q I q

The dependence of the flux linkages on the currents, given in Equation

(10.9) is reported in Figure 10.4. They refer always to the motor of Figure
10.1(b). It is worth noticing that the two flux linkages are not described by
a single curve, but feel the effects of both the d- and q-axis currents. It is
interesting to compare the result with that of Figure 10.3 (it refers to the
same motor). The knowledge of the correct dependence of the flux linkages

Flux linkage Λd (Vs) Flux linkage Λq (Vs)

0.08 0.16
0.14
0.06
0.12
0.04 Iq 0.1
|Id|
0.08
0.02
0.06
0 Iq 0.04
0.02
−0.02
0
−0.04 −0.02
−350 −300 −250 −200 −150 −100 −50 0 0 50 100 150 200 250 300 350
Current Id (A) Current Id (A)

FIGURE 10.4
Flux linkages Λd and Λq vs. currents Id and Iq.
192 Electrical Machine Analysis Using Finite Elements

on the currents is particularly useful for the design of electrical drive with
very precise positioning and very high dynamic performance.

10.3 Torque-Speed Characteristic

Thanks to its anisotropic structure, the interior permanent magnet motor is
well suited in applications requiring flux-weakening operations, that is, a
torque-speed characteristic with a constant torque region at low speeds, and
a constant power region at higher speeds. During steady-state operations,
the limit operating region may be divided into two regions:

Constant torque region — At low speed, the motor operates at the rated
current, say In, with Id and Iq components chosen in such a way to
obtain the maximum torque. It is the base torque Tb. This operating
point can be maintained from null speed up to the base speed, say
nb, at which the terminal voltage reaches the rated value Vn.
Constant power region — The motor operates at speed higher than nb
with a demagnetizing (i.e., negative) d-axis current. Such a current
weakens the flux linked by the stator winding, so that it is possible
to increase the motor speed without exceeding the nominal voltage,
as indicated by Equation (10.3) where ω = (2π/60)np. On the other
hand, a decrease of the torque is achieved, since the Id and Iq currents
move from the operating point of maximum torque. The operating
region presents a torque decreasing with the speed and is so called
the “constant power” region, even though it only approximates this
behavior. Referring to the operating limit, the current phasor Id + jIq
maintains a constant magnitude and describes a circular trajectory,
always increasing the negative value of the d-axis current. A wide
operating region is obtained if the interior permanent magnet motor
exhibits Λm ≈ LdIn.

In order to individuate the operating regions by means of finite element

analysis, the following procedure is suggested:

1. The value of the rated current In is fixed. Some simulations are

carried out with different combinations of the base operating point
defined by the current phasor Id + jIq. From the field solutions the
flux linkages Λd and Λq are computed, and then the torque from
Equation (10.1). The maximum torque Tb and the corresponding
operating point are found. Because Ld < Lq, then Id < 0 has to be
imposed. In addition, it can be verified that the maximum torque is
found with Iq >
Id
.
Interior Permanent Magnet and Reluctance Synchronous Motors 193

2. Once the maximum torque operating point has been found, impos-
ing the nominal voltage V = Vn in Equation (10.3), the base speed is
computed as nb = (60/2π)ω/p, where

Vn
ω= (10.11)
Λd 2 + Λq2

The constant torque region is then completely defined by Tb and nb.

3. The d- and q-axis current components, Id and Iq, are modified,
increasing the negative value of Id and decreasing the value of Iq,
but always with a constant magnitude equal to In. From the corre-
sponding field solution, the flux linkages Λd and Λq are computed,
and thus the torque T from Equation (10.1) and the speed ω from
Equation (10.11), and thus n = (60/2π)ω/p. The values of torque T
and speed n correspond to a point of the mechanical characteristic,
in the decreasing torque region.
4. The procedure described in the step 3 is repeated until the required
maximum speed is reached or, alternatively, when a null torque is
reached.

10.3.1 Example
An interior permanent magnet motor is designed for an application requiring
a constant torque region with base torque Tb = 10 Nm and base speed nb =
2000 rpm, and a constant power region up to the maximum speed nfw =
6000 rpm, at which the torque should be Tfw = 3.33 Nm.
The motor structure and the main data are reported in Figure 10.5. A single-
layer, full-pitched winding is used. The rotor exhibits an anisotropy ratio
(also called the saliency ratio) ξ = Lq/Ld almost equal to 4, confirming the
high anisotropy of this kind of rotor configuration.
Figure 10.6 shows the flux lines in one section of the motor pole. Figure
10.6(a) refers to the no-load operation. The flux leakage through the rotor

wt

ht
De g S
2 lm
N

hm
2 S N
D/2

FIGURE 10.5
Motor structure and main data.
194 Electrical Machine Analysis Using Finite Elements

FIGURE 10.6
Flux lines with only permanent magnet (a) and only q-axis currents (b).

magnetic bridges is evident. The corresponding air-gap flux density distri-

bution is shown in Figure 10.7(a1). Figure 10.7(a2) and Figure 10.7(a3) show
the air-gap flux density distribution corresponding to the d-axis current equal
to −30 A and −70 A, respectively, highlighting their demagnetizing effect.
Figure 10.6(b) shows the flux lines with only a q-axis current (the perma-
nent magnet is assumed to be demagnetized, i.e., Bres = 0). The flux lines
enter in the rotor, centered along the interpolar axis and flow by the side of
the barriers. The air-gap flux density distribution with only a q-axis current
Iq = 30 A is shown in Figure 10.7(b1). Figure 10.7(b2) shows the air-gap flux
density distribution due to both permanent magnet and Iq = 30 A. Finally,
Figure 10.7(b3) shows the flux density distribution increasing the q-axis
current to Iq = 70 A.
The motor torque computed from the field solutions at different currents
is shown in Figure 10.8. In the same figure the experimental data on a motor
prototype are reported, confirming a good agreement between simulated
and tested values.
Since the rated dc voltage is 48 V, reduced to 42 V for considering the voltage
drop, a voltage limit Vn = 24.2 V (peak value of the sinusoidal waveform).
According to the procedure illustrated above, the mechanical characteristics
of Figure 10.9 are obtained, with different currents In and given voltage Vn.
The dashed line in Figure 10.9 confirms that the requirements are satisfied
with a rated current of 70 A, peak value. The interior permanent magnet
motor exhibits a constant power region in a speed range of 1:3.

10.4 Algorithm for an Automatic Computation

The procedures illustrated above are well suited to be implemented in a code
for the automatic computation of the motor characteristics. These procedures
interact with the finite element algorithms, leading each step of the simulation.
Interior Permanent Magnet and Reluctance Synchronous Motors 195

Flux density (T) Flux density (T)

1 0.5

0.5 0

0 −0.5

−0.5 −1 Iq = 30 (A)
no PM
only PM
−1 −1.5
0° 60° 120° 180° 0° 60° 120° 180°
Electrical angle (deg) Electrical angle (deg)
(a1) (b1)

Flux density (T)

1
Flux density (T)
1 0.5 Iq = 30 (A)
0.5 0
Id = −30 (A)
0 −0.5

−0.5 −1

−1 −1.5
0° 60° 120° 180° 0° 60° 120° 180°
Electrical angle (deg) Electrical angle (deg)
(a2) (b2)

Flux density (T)

1
Flux density (T)
1 0.5
Iq = 70 (A)
0.5 0
Id = −70 (A)
0 −0.5

−0.5 −1

−1 −1.5
0° 60° 120° 180° 0° 60° 120° 180°
Electrical angle (deg) Electrical angle (deg)
(a3) (b3)

FIGURE 10.7
Effect of different currents on the air-gap flux density distribution.

At first, the code is reported for the automatic computation of the d- and
q-axis flux linkages, corresponding to a generic current value.
In the example, the motor structure of Figure 10.10 is considered. A double-
layer, one-slot chorded winding is adopted. The portion used in the simula-
tion is that shown in Figure 8.5(b) in Chapter 8.
196 Electrical Machine Analysis Using Finite Elements

12
70 (A) peak Simulated
10 Measured

Torque (Nm)
50 (A) peak
6
30 (A) peak
4

0
90° 100° 110° 120° 130° 140° 150° 160° 170° 180°
Current vector angle (elect. degrees)

FIGURE 10.8
Motor torque versus the angle of the current vector (Id, Iq).

Torque (Nm) Power (kW)

14 3.5
90 A
12 90 A Vn = 24.2 (V)
3
70 A
10 70 A 2.5
8 50 A B F 2
6 1.5
4 30 A 50 A 1
2 30 A 0.5
0 0
0 1000 2000 3000 4000 5000 6000 7000
Speed (rpm)

FIGURE 10.9
Torque and power versus speed, at different stator current values.

7 d
6
5
4
3

2
1
g
D
De

FIGURE 10.10
Structure of the interior permanent magnet motor and references.
Interior Permanent Magnet and Reluctance Synchronous Motors 197

Some preliminary variables are initially defined:

' Preliminary definitions

DIM Bn1(180), Bn2(180), Bn3(180)
pi = 3.1415926536
dtheta = pi / 180
'-----------------------------------------
' Current values in the simulation
Id = -113
Iq = 185

Some FUNCTIONs are employed to assign the currents, the material char-
acteristics, and the boundary conditions. Then the field problem is solved.
From the field solution the magnetic quantities are processed. Let us assume
that the air-gap flux density distribution is stored in the file FileInduz$,
and the average magnetic vector potential within the stator slots are stored
in files FilePotential$, one for each slot.
' Name of the file for flux density distribution
FileInduz$ = “bgload”
' Name of the file for average magnetic vector potential
FilePotential$ = “MagnVectPot”
' Assignment of the currents
CALL AssignCurrent (Id, Iq)
' Assignment of the iron permeability
CALL AssignMu
' Assignment of the permanent magnet parameters
CALL AssignPM (mu, Br, ang)
' The field problem is solved
CALL Solution
' the flux density distribution is stored in the file
CALL Induction (FileInduz$)
' the flux linkage is computed by integrating Az
CALL GlobalFlux

The FUNCTION that is used to assign the current in the stator slots is
reported next. This FUNCTION is based on the distribution of the coils as
shown in Figure 10.10(b). In order to impose the current components Id and
Iq, since the polar axis has been fixed corresponding to the a-phase axis, so
that ϑ = 0, the three-phase currents are given by

Ia = Id Ib = −
1
2(Id − 3 Iq ) Ic = −
1
2(Id + 3 Iq ) (10.12)

At this point, the current in the slot can be assigned as

( ) ( ) ( )
n qs n qs n qs
I1 = − Ia − Ia I2 = − Ia + I c I3 = Ic + Ic
4 2 2

( ) ( ) ( )
n n n
I 4 = qs I c − I b I 5 = qs − I b − I b I6 = qs − I b + Ia (10.13)
2 2 2

( )
n
I7 = qs Ia + Ia
4
198 Electrical Machine Analysis Using Finite Elements

where nqs is the number of series conductors per slot, given by nq/npp, with
npp the number of parallel paths (see Chapter 8). The current I1 and I7 are
divided by 4, since only half a slot is simulated. Let N be the number of
series conductors per phase, 2p the number of poles, and qs the number of
slots per pole per phase. Then, it results in

N 3N
n qs = = (10.14)
2 pqs Q

SUB AssignCurrent (Id, Iq)

Ia = Id
Ib = -0.5 * (Id - SQR(3) * Iq)
Ic = -0.5 * (Id + SQR(3) * Iq)
Icava1 = (nqs / 4) * (-Ia - Ia)
Icava2 = (nqs / 2) * (-Ia + Ic)
Icava3 = (nqs / 2) * (Ic + Ic)
Icava4 = (nqs / 2) * (Ic - Ib)
Icava5 = (nqs / 2) * (-Ib - Ib)
Icava6 = (nqs / 2) * (-Ib + Ia)
Icava7 = (nqs / 4) * (Ia + Ia)

After the field solution has been obtained, the results stored in the files
are processed. At first the air-gap flux density distribution is managed. The
distribution of the air-gap flux density is given by 180 values; it is read from
the file FileInduz$. These 180 values are stored in the variable Bn1(nnn).
The two fundamental harmonics are computed, corresponding to the direct
axis (cosϑr) and the quadrature axis (sinϑr) as depicted by Equation (8.16)
and Equation (8.24). The d- and q-axis flux linkages Λd and Λq are obtained
as depicted by Equation (8.18) and Equation (8.25).

' Processing of the obtained results

'
' 1: Method of the airgap flux density
' Reading the file
OPEN FileInduz$ FOR INPUT ACCESS READ AS #1
FOR nnn = 1 TO 180
INPUT #1, a, b, c
' The flux density value is in the third column
Bn1(nnn) = c
NEXT nnn
CLOSE #1
' The two–axis fundamental components are computed
Bid = 0
Biq = 0
FOR nnn = 1 TO 180
theta = dtheta * nnn
Bid = Bid + Bn1(nnn) * SIN(theta) * dtheta
Biq = Biq - Bn1(nnn) * COS(theta) * dtheta
NEXT nnn
B1d = Bid * 2 / pi
B1q = Biq * 2 / pi
F1d = (B1d * D * L / p) * (kw * n / 2)
F1q = (B1q * D * L / p) * (kw * n / 2)
Interior Permanent Magnet and Reluctance Synchronous Motors 199

The second method is based on the processing of the magnetic vector

potential Az on each coil section. As in Equation (8.2), the average magnetic
vector potential in each q-th slot is computed as

∫
1
A slot (q) = A z dS (10.15)
Sq Sq

and it is stored in the files FilePotential$, one for each slot. Then these
values are read from the files, for each slots of the domain (7 in the present
example) and assigned to the variables fslot(nnn).
From the distribution of the coil sections in the slots and from the sym-
metry conditions, the three-phase flux linkages Λa, Λb, and Λc are computed.
In the specific example they are given by

( ) ( )
n qs
Λa = 2 p  − Λ1 + Λ 2 + Λ 6 + Λ 7 
2  

( )
n qs
Λ b = 2p  − Λ 4 + 2 Λ 5 + Λ6  (10.16)
2  

( )
n qs
Λc = 2p  Λ2 + 2Λ3 + Λ4 
2  

The two-axis flux linkages Λd and Λq are then computed, using the trans-
formation Tabc/dq (see the Appendix in Chapter 8). Since ϑ = 0, the d- and q-axis
flux linkage components are given by

2 Λ + Λc 
Λd = 
3
Λa − b
2 
Λq =
1
3
(Λ b − Λc ) (10.17)

Then it results:

' Processing of the obtained results

'
' 2: Method of the flux linkages
' Reading the files corresponding to the 7 slots
FOR nnn = 1 TO 7
File$ = FilePotential$ +
+ LTRIM$(STR$(nnn)) + “.dat”
OPEN file$ FOR INPUT ACCESS READ AS #5
INPUT #5, fcava(nnn)
fcava(nnn) = fcava(nnn)
CLOSE #5
NEXT nnn
' The three–phase flux linkages are computed
Fa0 = (fcava(1)+fcava(2))+(fcava(6)+fcava(7))
Fb0 = -(fcava(4) + 2 * fcava(5) + fcava(6))
200 Electrical Machine Analysis Using Finite Elements

Fc0 = (fcava(2) + 2 * fcava(3) + fcava(4))

Fa = 2 * p * nqs / 2 * L * Fa0
Fb = 2 * p * nqs / 2 * L * Fb0
Fc = 2 * p * nqs / 2 * L * Fc0
' The two–axis flux linkages are computed
Fd = 2 / 3 * (Fa - (Fb + Fc) / 2)
Fq = 1 / SQR(3) * (Fb - Fc)

The flux lines corresponding to one pole of the machine are shown in
Figure 10.11, referring to the no-load and the full-load operation, respectively.
Some results obtained by applying the two proposed methods are reported
in Figure 10.12. Figure 10.12 shows a good agreement between the two
computation methods. In addition, the results are compared with experi-
mental results obtained by feeding the motor at industry frequency 50 Hz
and different loads.

FIGURE 10.11
Flux lines during no-load (a) and the full-load (b) operations.

Voltage (V) @ f = 50 (Hz) Torque (Nm)

100 180
33.5° 33.5°
38.3° 160
90 38.3°
39.5°
32.3° 140
80 39.5°
31.4° 120 32.3°
70 100
29.9° 31.4°
60 Measured 80 Measured
27.6° 29.9°

γ = 23.6°
* Method 1 60 27.6° * Method 1
50 x Method 2 x Method 2
40
γ = 23.6°
40 20
50 150 250 350 450 50 150 250 350 450
Current: magnitude (A), phase (deg) Current: magnitude (A), phase (deg)

FIGURE 10.12
Comparison between simulated and measured results.
Interior Permanent Magnet and Reluctance Synchronous Motors 201

10.5 Synchronous Reluctance Motors

The synchronous reluctance motor may be considered as a particular case
of the interior permanent magnet motor, obtained by taking the permanent
magnet away from the rotor. The synchronous reluctance rotor is similar to
the interior permanent magnet rotor, but generally it is designed with several
barriers per pole. This is because the motor torque is only due to the rotor
anisotropy, so that the saliency ratio has to be as high as possible.
Two reluctance motor configurations are shown in Figure 10.13. The first
one refers to a transversally laminated rotor, and the second one refers to an
axially laminated rotor. The commonly used notation for the d- and the q-axis
of synchronous reluctance motor is adopted in Figure 10.13, that is, the d-axis
is posed along the higher permeance path. This notation is opposite to the
notation adopted above for the interior permanent magnet motor. In the
reluctance motors, Ld > Lq is obtained and the saliency ratio is defined by
ξ = Ld/Lq.
In the case of the transversally laminated reluctance motor, shown in
Figure 10.13(a), the rotor is formed by a stack of stamped laminations. The
flux barriers are directly formed during the stamping process. There are a
number of barriers per pole in the range between 2 and 5. The rotor structure
is supported by thin magnetic bridges, whose thickness is designed by means
of mechanical considerations. These magnetic bridges are saturated by the
q-axis flux. This rotor configuration shows a saliency ratio up to 10.
The axially laminated reluctance rotor, illustrated in Figure 10.13(b), is
formed by staking laminations and nonmagnetic insulating layers, which

d d

q q

(a) (b)

FIGURE 10.13
Synchronous reluctance motor configurations.
202 Electrical Machine Analysis Using Finite Elements

are bended and placed along the axial direction. The structure is supported
by nonmagnetic spacings. The shaft is nonmagnetic as well. This rotor con-
figuration shows a saliency ratio more than 20, in linear conditions. However,
the rotor assembly is difficult and iron rotor losses occur, caused by the flux
fluctuation in the rotor laminations due to the stator slotting.
Sometimes, for increasing the power factor or the performance in the con-
stant power region, permanent magnets are inset within the flux barriers. They
are plastic bonded low-energy permanent magnets, which are shaped easily.

10.5.1 The Transversally Laminated Reluctance Motor

In the simulation of the reluctance motor, particular care has to be taken
with the rotor. The mesh of the domain should be mainly refined in the rotor
and in the air-gap. In addition, attention should be given to the magnetic
bridges, which work extremely saturated.
The motor may be fed by only a d-axis or only a q-axis current component,
reducing the analysis to only half a pole-piece. Otherwise, the motor may
be fed by whatever couple of current components Id and Iq, so that the
analysis is on one pole of the motor. The latter analysis allows a more
accurate magnetic model to be developed, considering the cross-coupling
effect, which greatly affects the reluctance motor performance.
As an example, Figure 10.14 shows the field lines of one pole of the
synchronous reluctance motor with different combinations of the currents Id
and Iq. The a-phase axis has been chosen in correspondence with the direct
axis.
Figure 10.14(a) shows the flux lines when the motor is fed by d-axis current
only (Id = 10 A, Iq = 0 A), imposing the three-phase currents Ia =10 A, Ib =
–5 A, Ic = −5 A.
In Figure 10.14(b) the motor is fed by a q-axis current only (Id = 0 A, Iq =
10 A), imposing the three-phase currents Ia = 0 A, Ib = 8.66 A, Ic = −8.66 A.
At last, Figure 10.14(c) refers to the motor fed by a current phasor with a
45 electrical degree (Id = Iq = 5 2 A), which is Ia = 7.07 A, Ib = 2.59 A, Ic =
–9.66 A.

FIGURE 10.14
Flux lines in a synchronous reluctance motor fed by only d-axis current (a), only q-axis current
(b), and two-axis current (c).
Interior Permanent Magnet and Reluctance Synchronous Motors 203

FIGURE 10.15
Flux plots with d-axis current only (a) and q-axis current only (b).

10.5.2 The Axially Laminated Reluctance Motor

Because of the several laminations that are thin and divided one to the other,
the finite element simulation requires a mesh with a very large number of
elements. For this reason, it is necessary to reduce the analyzed domain as
much as possible, making use of any symmetry of the machine. In the
following, only a quarter of pole is analyzed, feeding the stator with only
d-axis or only q-axis current. As a consequence, the flux linkages Λd and Λq
are computed separately.
With d-axis current only, the magnetic vector potential Az is null along the
direct axis. Homogeneous Dirichlet’s boundary conditions are assigned
along the d-axis, and homogeneous Neumann’s boundary conditions along
the q-axis, along which the flux density vector has only normal components.
In the same way, with q-axis current only, Dirichlet’s and Neumann’s
boundary conditions are assigned along the q-axis and the d-axis, respectively.
An example is reported in Figure 10.15. A three-phase, four-pole 36-slot
motor is considered with a single-layer, full-pitched winding. The rotor is
composed by 58 magnetic laminations with thickness tl = 0.328 mm spaced
by an insulating sheet with thickness ti = 0.194 mm.
Due to the heavy simulation of such a machine, the analysis may be limited
to the rotor only. The stator is substituted by an infinitesimal conducting
sheet placed along the periphery of the bore diameter D. In this sheet a
surface current density is forced. It could be a d-axis surface current density
as

mNIˆ
( )
J sd ϑ r =
πD
sin(ϑ r ) (10.18)

or a q-axis surface current density as

mNIˆ
( )
J sq ϑ r = −
πD
cos(ϑ r ) (10.19)
204 Electrical Machine Analysis Using Finite Elements

FIGURE 10.16
Flux lines with (a) d-axis and (b) q-axis surface current density distribution.

Airgap flux density (T) Airgap flux density (T)

0.3 0.03

0.25 0.02
0.01
0.2
0
0.15
−0.01
0.1
−0.02
0.05 −0.03 Current Iq = 1 A
Current Id = 1 A
0 −0.04
−90° −80° −70° −60° −50° −40° −30° −20° −10° 0° −90° −80° −70° −60° −50° −40° −30° −20° −10° 0°
Electric angle (deg) Electrical angle (deg)

FIGURE 10.17
Air-gap flux density distributions in simulations of rotor only.

where m is the number of phases, N is the number of series conductors, and

ϑr is the electrical angle referred to the rotor d-axis (as shown in Figure 8.19).
Figure 10.16 shows the flux lines with a d-axis and q-axis current distri-
bution, respectively. In the first case, the flux lines are parallel to the lami-
nations, while in the second case they are normal to the laminations. The
corresponding air-gap flux density distributions are shown in Figure 10.17.

References
1. V.B. Honsinger, “The field and parameters of interior type AC permanent
magnet machines,” in IEEE Trans. on PAS, 101, 1981, pp. 867–876.
2. B. Sneyers, D.W. Novotny, T.A. Lipo, “Field weakening in buried permanent
magnet AC motor drives,” in IEEE Trans. on Ind. Appl., vol. 21, no. 2, pp.
398–407, March/April 1985.
3. T.M. Jahns, G.B. Kliman, T.W. Neumann, “Interior permanent-magnet synchro-
nous motors for adjustable-speed drives,” in IEEE Trans. on Ind. Appl., vol. 22,
no. 4, pp. 738–747, July/August 1986.
4. R.F. Schiferl and T.A. Lipo, “Power capability of salient pole permanent magnet
synchronous motors in variable speed drive applications,” in IEEE Trans. on
Ind. Appl., vol. 26, no. 1, pp. 115–123, January/February 1990.
Interior Permanent Magnet and Reluctance Synchronous Motors 205

5. S. Morimoto, Y. Takeda, T. Hirasa, K. Taniguchi, “Expansion of operating limits

for permanent magnet motor by current vector control considering inverter
capability,” in IEEE Trans. on Ind. Appl., vol. 26, no. 5, pp. 866–871, September/
October 1990.
6. A. Fratta, A. Vagati, F. Villata, “Design Criteria of an IPM Machine Suitable for
Field-Weakened Operation,” Proc. of International Conference on Electrical Ma-
chines, pp. 1059–1065, 1990.
7. A.K. Adnanes and T.M. Undeland, “Optimum torque performance in PMSM
drives above rated speed,” in Ind. Appl. Society Annual Meeting Records, pp.
169–175, 1991.
8. W.L. Soong and T.J.E. Miller, “Theoretical limitations to the field-weakening
performance of the five classes of brushless synchronous AC motor drive,” in
Proc. of Electrical Machines and Drives Conference, pp. 127–132, Oxford (UK), 1993.
9. B.J. Chalmers, L. Musaba, D.F. Gosden, “Variable Frequency Synchronous
Motor Drives for Electric Vehicles,” IEEE Trans. on Ind. Appl., vol. 32, no. 4,
1996, pp. 896–903.
10. E. Chiricozzi, F. Parasiliti, M. Villani, T. Higuchi, J. Oyama, E. Yamada, “Buried
PM Synchronous Motor Design Optimization,” in 2nd EPE Chapter Symposium
on Electric Drive Design and Applications, Nancy (F), June 1996, pp. 41–45.
11. N. Bianchi, S. Bolognani, “Parameters and Volt-Ampere Ratings of a Synchro-
nous Motor Drive for Flux-Weakening Applications,” in IEEE Trans. on Power
Electronics, vol. 12, no. 5, September 1997, pp. 895–903.
12. N. Bianchi, S. Bolognani, B.J. Chalmers, “Comparison of Different Synchronous
Motor Drives for Flux-Weakening Applications,” in Proc. of International Con-
ference on Electrical Machine, ICEM ’98, Istanbul, Turkey, 2–4 September 1998,
pp. 946–951.
13. N. Bianchi, S. Bolognani, “Magnetic Models of Saturated Interior Permanent
Magnet Motors based on Finite Element Analysis,” in IEEE Industrial Applica-
tions Conference Records, IAS ’98, St. Louis, 12–15 October 1998, vol. 1, pp. 27–34.
14. B.J. Chalmers, R. Akmese, L. Musaba, “Design and field-weakening perfor-
mance of permanent-magnet/reluctance motor with two-part rotor,” in IEEE
Proc. Pt. B, vol. 145, no. 2, 1998, pp. 133–139.
15. D.M. Ionel, J.F. Eastham, T.J.E. Miller, E. Demeter, “Design Considerations for
Permanent Magnet Synchronous Motors for Flux-Weakening Applications,”
IEEE Proc. Pt. B, vol. 145, no. 5, 1998, pp. 435–440.
16. N. Bianchi, S. Bolognani, F. Parasiliti, M. Villani, “Prediction of Overload and
Flux-Weakening Performance of an IPM Motor Drive: Analytical versus Finite
Element Approach,” in Proc. of 8th European Conference on Power Electronics and
Applications, EPE99, Lausanne, 7–9 September 1999, DS 1.6, topic 5, pp. 1–8.
17. N. Bianchi, S. Bolognani, B.J. Chalmers, “Design Considerations for a PM
Synchronous Motor with Rotor Saliency for High Speed Drives,” in IEEE Indus-
trial Applications Conference Records, IAS ’99, Phoenix, Arizona, 3–7 October 1999,
vol. 1, pp. 117–124.
18. N. Bianchi, S. Bolognani, M. Zigliotto, “Design and Development of a PM
Synchronous Motor Drive for an Electrical Scooter,” in Proc. of ICEM 2000
International Conference, Helsinki, Finland, 28–30 August 2000.
19. N. Bianchi, S. Bolognani, M. Zigliotto, “High Performance PM Synchronous
Motor Drive for an Electrical Scooter,” in IEEE IAS Annual Meeting Records,
Rome, Italy, 8–12 October 2000.
206 Electrical Machine Analysis Using Finite Elements

20. A. Fratta and A. Vagati, “A Reluctance Motor Drive for High Dynamic Perfor-
mance Applications,” in IEEE IAS Annual Meeting, Atlanta, pp. 295–302, 1987.
21. R.E. Betz, “Control of Synchronous Reluctance Machines,” in Ind. Appl. Society
Annual Meeting Records, pp. 456–462, 1991.
22. R.E. Betz, M. Javanovic, R. Lagerquist, T.J.E. Miller, “Aspects of the control of
synchronous reluctance machines including saturation and iron losses,” in Ind.
Appl. Society Annual Meeting Records, pp. 456–463, 1992.
23. A. Vagati, G. Franceschini, I. Marongiu, G.P. Troglia, “Design criteria of high
performance synchronous reluctance motors,” in Ind. Appl. Society Annu. Meet.,
1992, vol. I, pp. 66–73.
24. N. Bianchi, S. Bolognani, “Synchronous Reluctance Motor Drives: Overload
and Flux-Weakening Performance Taking into Account Iron Saturation,” in
Proc. of International Conference on Electrical Machines, ICEM ’96, Vigo (Spagna),
10–12 September 1996, vol. I, pp. 360–365.
25. N. Bianchi, S. Bolognani, “Parameters and Volt-Ampere Ratings of a Synchro-
nous Reluctance Motor Drives for Flux-Weakening Applications Taking into
Account Iron Saturation,” in Proc. of Conference on Power Electronics and Appli-
cations, EPE ’97, Trondheim (Norvey), 8–10 September 1997, vol. 3, pp. 631–636.
26. B.J. Chalmers, L. Musaba, “Design and Field-Weakening Performance of a
Synchronous Reluctance Motor with Axially-Laminated Rotor,” IEEE Ind. Appl.
Society Annual Meeting Conference Records, 1997, pp. 271–278.
27. N. Bianchi, B.J. Chalmers, “Effect of the Distribution of the Laminations in an
Axially Laminated Relactance Motor,” in Proc. of IEEE 9th International Confer-
ence of Electrical Machines and Drives, EMD ’99, Canterbury, UK, 1–3 September
1999, pp. 376–380.
28. N. Bianchi and T.M. Jahns, Eds., “Design, Analysis and Control of Interior PM
Synchronous Machines,” Tutorial Course Notes, IEEE IAS Meeting, Seattle,
WA, Oct. 3, 2004, CLEUP, Padova.
11
Self-Starting Single-Phase
Synchronous Motors

This chapter deals with the analysis of a self-starting single-phase synchro-

nous motor. At first, finite element magnetostatic analysis is carried out to
obtain the motor parameters as a function of rotor position and current.
Then, these parameters are used in the analysis of dynamic performance of
the motor.
In this particular case, a single-phase permanent magnet micro-motor is
analyzed. It is characterized by an anisotropic stator that is necessary to
develop a starting torque. Its stationary and dynamic operations are simu-
lated. The algorithms for the prediction of the dynamic performance are
reported.

11.1 Introduction
Electrical motors of very small dimensions are used in many devices, espe-
cially in civil applications. Since an alternative voltage is commonly avail-
able, the motors are often ac motors. Among them, the most utilized is the
single-phase induction motor, with an auxiliary winding or a shaded pole
to allow the rotor starting. In some applications, where a synchronous speed
is required, hysteresis motors, reluctance motors, or permanent magnet
motors are adopted. Additionally, when reduced dimensions are explicitly
required, the use of permanent magnets allows miniaturized motors to be
obtained.
Apart from the hysteresis motor, the main problem of the synchronous
motors is the rotor starting. In order to obviate such a problem, a conducting
cage may be inset in the rotor, so that an asynchronous starting is achieved.
Also, an anisotropic stator may be designed, so as the rest positions of the
rotor are different according to the stator current. The analysis and the design
of such motors are laborious and require remarkable work.

207
208 Electrical Machine Analysis Using Finite Elements

N N
S S
N
y S S
N
z x

FIGURE 11.1
Structure of the single-phase permanent magnet motor.

11.2 Definition of the Motor Model

Some structures of the single-phase permanent magnet motor are shown in
Figure 11.1. The stator core is obtained by magnetic laminations. The stator
coil is wound around the stator core. A planar symmetry is recognized, and
a two-dimensional analysis is carried out.
If the motor does not show any other symmetry axis, as easily observed
from Figure 11.1, then there are further feasible simplifications of the domain.
In Figure 11.1(a), the rotor axis is not aligned with the stator laminations,
while in Figure 11.1(b) and Figure 11.1(c) there is an asymmetry between the
rotor and stator poles.
Such a misalignment is required for achieving the starting torque. The rest
position of the rotor with null currents is different from that position when
a stator current feeds the coil. In other words, the magnetic field strength
due to the stator current produces a force that tends to rotate the permanent
magnet from its natural rest position.
The analysis of the stationary performance of the motor is carried out by
means of magnetostatic field solutions. Neglecting the edge effects, the fron-
tal section is drawn in the plane (x, y). The results of the magnetostatic field
solution are then processed for studying the dynamic behavior of the motor.
Silicon-steel laminations are used for stator core. Particular care is taken
with modeling the permanent magnet: the characteristic and the magneti-
zation directions of the various rotor sectors have to be exactly defined.
Ferrite permanent magnet composites are usually employed. Plastic
bonded permanent magnet materials reduce dimensions, keeping restrictive
mechanical tolerances. In addition, the brittleness is lower than in sintered
permanent magnet materials.
The B-H curve is essentially linear in the second quadrant, as seen in Figure
9.4. It is characterized by the residual flux density Bres and the coercive field
strength Hc, or alternatively by the differential relative permeability µr . The
rotor poles are defined as separate objects, with the form of a cylinder sector.
Self-Starting Single-Phase Synchronous Motors 209

They are identical but characterized by the appropriate direction of magne-

tization. The resulting rotor is isotropic, since µr ≈ 1.
The boundary conditions are assigned assuming that the flux lines do not
go out of the fixed domain. Hence, a null magnetic vector potential can be
assigned along all the boundary of the domain.
The stator coil is formed by Nt turns, and is modeled by an equivalent
conducting rectangular bar. The current flowing within the equivalent bar
corresponds to Nt times the effective current of the winding.
According to the reference coordinate system adopted in Figure 11.1, the
current is positive when its direction matches the z-axis direction (leaving
the sheet), and negative when its direction is opposite to the z-axis direction
(going into the sheet).

11.3 Computation of the Electrical Parameters

The stator flux linkage is a function of the stator current and of the rotor
angular position, because of the permanent magnet, which is λ = λ(ϑm, i).
Let us assume that the flux density is low so that the magnetic circuit is not
saturated. Then, since the rotor is isotropic, the flux linkage can be expressed
as
λ(ϑm, i) = La i + λpm(ϑm) (11.1)

where La is the self-inductance of the winding. It is constant with the assump-

tions given above (linear magnetic circuit and isotropic rotor). Then, λpm is
the flux linkage due to the permanent magnet; it is a function of the rotor
angular position ϑm. With no hysteresis, Equation (11.1) defines a univocal
correspondence among λ, i, and ϑm. Two of these quantities are chosen to
be the state variables of the system (see Chapter 5).
The computation of the flux linkage, in a fixed rotor position and with a
fixed stator current, is carried out by integrating the magnetic vector poten-
tial A over the equivalent bar. Since the magnetic vector potential has only
the z-axis component, the flux linkage is given by

 1 
( )
λ ϑ m , i = Nt L Fe 
∫ A zdS −
1
∫ A zdS  (11.2)
 S Cu S Cu + S Cu S Cu − 

where LFe is the motor net length, SCu+ is the equivalent conductive bar
carrying a positive current (which is a current with direction corresponding
to the z-axis), and SCu– is the equivalent conductive bar carrying a negative
current (which is a current with direction opposite to the z-axis). The ratio
between the integral of Az and the bar surface gives the average value of the
magnetic vector potential.
210 Electrical Machine Analysis Using Finite Elements

11.3.1 Flux Linkage due to the Permanent Magnet

The dependence of the flux linkage due to the permanent magnet on the
rotor angle, which is λpm(ϑm), is analyzed by means of a series of simulations
with different rotor positions ϑm and null stator current, as Equation (11.1)
highlights.
The magnetization of the permanent magnet elements has to be defined
from time to time, at each rotor position ϑm. The flux linkage has to be
analyzed in an angular interval 2π/p, where p is the pole pairs number. It
is useful to choose the angular positions at regular intervals, e.g., fixing an
angular interval ∆ϑm.

11.3.2 Self-Inductance Coefficient

In the computation of the self-inductance of the stator winding, only the flux
linkage due to the stator current is considered. It is

La =
( )
λ ϑ m , i − λ pm ϑ m ( ) (11.3)
i

and can be computed from two field solutions at the same rotor angle ϑm,
the former with null current, to obtain the flux linkage λpm(ϑm), and the latter
with a stator current i, to obtain the flux linkage λ(ϑm,i).
Alternatively, the permanent magnet is demagnetized, so that λpm(ϑm) = 0.
This is achieved by assigning a null residual flux density, i.e., Bres = 0 to the
permanent magnet elements, but without varying the relative permeability
µr . It results in

La =
(
λ ϑm , i ) (11.4)
i
λ pm = 0

With the hypothesis of linear magnetic circuit, La is constant and one

computation is enough to determine it.

11.4 Computation of the Torque

The motor torque is a state function of the rotor angular position ϑm and the
stator current i, i.e., τm = τm(ϑm,i).
A complete study of the torque should be carried out varying ϑm and i.
The torque is evaluated in a number Nϑ of angular positions placed at
Self-Starting Single-Phase Synchronous Motors 211

intervals ∆ϑm for a complete angle 2π/p, corresponding to two motor poles.
Similarly, a number of Ni currents with fixed increment is chosen. Positive
currents only may be studied. It can be assumed that the motor torque for
positive current i and angle ϑm is the same torque developed with a negative
current and a rotation of π/p, i.e., τm(ϑm + π/p, –i) = τm(ϑm,i). Such an
observation allows the number of magnetostatic field analysis to be reduced.
Computing the motor torque in all the Nϑ rotor positions and with all the
Ni stator currents, a matrix with Ni column and Nϑ rows is built, containing
the numerical evaluation of the state function τm = τm(ϑm,i). It is called the
torque matrix.

11.5 Analysis of the Dynamic Performance

The stationary behavior of the synchronous permanent magnet motor is
analyzed, by means of the field solutions. The flux linkage and the torque
are expressed as functions of the current and the rotor position. These results
are processed to carry out the analysis of the dynamic performance of the
motor.
To do that, the equations describing the dynamic performance of the elec-
trical and mechanical quantities are used:

v(t ) = Ri(t ) + La
di(t ) dλ pm t
+
() (11.5)
dt dt

dω m (t )
τ m (t ) = τ L + k Bω m (t ) + J (11.6)
dt

where v(t) is the forcing voltage source and R is the coil resistance. In
Equation (11.6), τm(t) is the electromechanical torque developed by the motor,
τL is the load torque, kB is the friction coefficient, and ωm(t) is the mechanical
speed.
The dynamic simulation requires that the motor performance be known
with any rotor position and with any stator current. Since in the previous
field analyses, the flux linkage and the torque have been obtained in a
discrete number of points, these values have been interpolated.
The flux linkage vector has been computed at Nϑ different rotor positions.
A one-dimensional interpolation is required to obtain the value correspond-
ing at each angular position.
Similarly, the torque matrix contains the torque computed at Nϑ rotor
position with Ni current values. A 2D interpolation is needed to obtain the
motor torque at any position and current.
212 Electrical Machine Analysis Using Finite Elements

In Section 11.7, a simple algorithm for the linear interpolation is presented.

In Section 11.8, the algorithm for the dynamic analysis of the motor is
described.

11.5.1 Simulation at Constant Speed

The first dynamic analysis is carried out at a constant rotor speed ωm, sup-
posing a rotor inertia so high as any speed variation is neglected. Hence, in
the mechanical differential equation (11.6) the derivative term may be omit-
ted. The problem is reduced to the integration of the electrical differential
equation (11.5).
Let us assume a voltage supply with sinusoidal waveform of amplitude
VM and frequency f, while the constant rotor speed is fixed to ωm = ω/p =
(2πf)/p.
A simple procedure for the integration of the electrical differential equation
(11.5) is described in the following. Eulero’s method is used for the numerical
integration. The algorithm is reported in Section 11.8. Because of the sim-
plicity of the integration method, a short integration time has to be chosen
to obtain a good result.
The time derivative of the flux linkage due to the permanent magnet is
given by

dλ pm dλ pm dϑ m dλ pm
= = ωm (11.7)
dt dϑ m dt dϑ m

since the variation of λpm with ϑm has been computed by the magnetostatic
field analysis.
Assuming that the voltage v, and the derivative dλpm/dt are given in the
n-th instant, the time variation of the current is computed from Equation
(11.6) as

 di  1  dλ pm 
 dt  = L  v − Ri( n − 1) − dt  (11.8)
a  

In Equation (11.8), i(n–1) is the current value at the (n–1)-th instant, and
the current variation refers to the n-th instant. Then the current at the n-th
instant is given by

 di 
i( n ) = i( n − 1) +   dt (11.9)
 dt 

The corresponding torque is obtained by means of the 2D interpolation

(see Section 11.7), considering an average value of current given by
Self-Starting Single-Phase Synchronous Motors 213

i( n ) + i( n − 1)
i m ( n − 1) = (11.10)
2

in each interval dt.

Since the mechanical speed ωm is constant, the rotor position varies linearly:

ϑ m ( n ) = ϑ m ( n − 1) + ω mdt (11.11)

11.5.2 Simulation of the Full Dynamic Performance

The full dynamic performance of the motor is obtained by integrating at the
same time both the electrical and mechanical differential equations (11.5)
and (11.6). This is particularly convenient for evaluating the starting perfor-
mance of the rotor. Once again Eulero’s integration method is used. The
angular acceleration of the rotor is computed from the torque at the n-th
instant and the angular speed at the (n–1)-th instant, which is

 dω m  1
 dt  = J  τ m ( n ) − τ L − k Bω m ( n − 1)  (11.12)

then the angular speed at the n-th instant is

 dω m 
ω m ( n ) = ω m ( n − 1) +  dt (11.13)
 dt 

The angular position at the end of each integration time is

dω m dt 2
ϑ m ( n ) = ϑ m ( n − 1) + ω m ( n − 1)dt + (11.14)
dt 2

The rotor starting is simulated by imposing a sinusoidal voltage with

amplitude VM and frequency f, with different values of the initial voltage
phase ϕ and with different initial angular rotor positions ϑm(0). In particular,
the rotor starting has to be verified, assuming the rotor to be in its rest
positions at null current, which is in the positions where τm(ϑm,0) = 0.
By means of this study one can verify if the rotor is able to start and
synchronize itself at the fixed supply frequency. Section 11.8 deals with the
algorithm of such an integration.
214 Electrical Machine Analysis Using Finite Elements

3.4
0.4 2.6 0.4
0.25
0.4
0.95
1.1 1.9
0.05
N
1.0 S S
N
2.65

FIGURE 11.2
Sketch and main dimensions of the motor (measured in mm).

11.6 Example
The motor used in this example is a micro-motor, whose main dimensions
are shown in Figure 11.2. The rotor is formed by an isotropic permanent
magnet, which is a ceramic Barium-Ferrite magnet, characterized by a resid-
ual flux density Bres = 80 mT and a coercive field strength Hc = –60 kA/m.
The external diameter is 1 mm and the inner diameter is 0.15 mm. Then the
permanent magnet thickness is 0.425 mm.
The stator is formed by magnetic steel laminations with a thickness of
0.4 mm, and a stack length LFe = 5 mm. The coil is placed on one side of the
lamination and is formed by Nt = 1000 turns with dc = 25 µm copper wire
diameter.
The four poles of the rotor are modeled by four different objects; the shape
of each object is a circular sector. Each is characterized by a suitable magne-
tization direction. The flux lines due to the permanent magnet only are
shown in Figure 11.3.

S S

FIGURE 11.3
Flux lines due to the permanent magnet only.
Self-Starting Single-Phase Synchronous Motors 215

The reference initial position ϑm = 0 is chosen when the North pole is in

front of the upper expansion of the stator expansion, as shown in Figure 11.2.
Figure 11.4 and Figure 11.5 show the flux lines changing the rotor position
ϑm, starting from the initial position ϑm = 0 up to the position ϑm = π/3, with
a counter-clockwise rotation.

0 deg 45 deg

30 deg 60 deg

FIGURE 11.4
Flux lines at different rotor positions ϑm (at no-load).

0 deg 30 deg

45 deg 60 deg

FIGURE 11.5
Flux lines at different rotor positions ϑm (under load).
216 Electrical Machine Analysis Using Finite Elements

⫻ 10−5
1.5

Flux linkage (Vs) 0.5

−0.5

−1

−1.5
0 0.5 1 1.5 2 2.5 3
Angular position (rad)

FIGURE 11.6
Flux linkage due to the permanent magnet only, simulated values, and interpolating function.

11.6.1 Computation of the Flux Linkage due to the Permanent Magnet

The flux linkage due to the permanent magnet has been computed at differ-
ent rotor positions at a constant angular interval ∆ϑm = π/36 from ϑm = 0 to
ϑm = π. The value of λpm is obtained from Equation (11.2), with SCu+ = SCu– =
65 · 10–8 m2, by means of simulations with null stator current.
Figure 11.6. shows the flux linkage λpm as a function of the rotor position
ϑm. Since the behavior is essentially sinusoidal, it is possible to approximate
it by a sinusoidal waveform, given by

  π 
λ pm (ϑ m ) = −1.431 ⋅ cos  2  ϑ m −   (11.15)
  36  

In Figure 11.6, stars indicate the simulated values while the solid line
represents the interpolating function.

11.6.2 Computation of the Inductance

The inductance is computed by Equation (11.4), from a field analysis with a
demagnetized permanent magnet (Bres = 0 T) and with a stator current 20 mA.
The flux linkage is computed by Equation (11.2). The integration of magnetic
vector potential Az on the section SCu+ is equal to 2.79·10–11 Vsm, while the
integration on the section SCu– is equal to –0.0771·10–11 Vsm. The flux linkage
is then λ = 2.2055·10–5 Vs, so that the inductance results in La = 1.6541 mH.

11.6.3 The Torque Matrix

The torque motor contains the values of the motor torque τm = τm(ϑm,i) at
different rotor positions ϑm and stator currents i. The angular position ranges
Self-Starting Single-Phase Synchronous Motors 217

TABLE 11.1
Torque Matrix (Torque as a Function of the Rotor Position and Current)

Torque · 10-7 (Nm)

ϑm Current (mA)
(rad) 0 2.5 5 7.5 10 12.5 15 17.5 20
0 –0.020 –0.118 –0.199 –0.282 –0.354 –0.435 –0.521 –0.602 –0.689
0.131 –0.370 –0.340 –0.313 –0.283 –0.253 –0.218 –0.189 –0.183 –0.164
0.262 –0.595 –0.453 –0.315 –0.172 –0.035 0.113 0.247 0.301 0.425
0.396 –0.659 –0.411 –0.166 0.085 0.335 0.585 0.750 0.978 1.207
0.524 –0.558 –0.217 0.130 0.478 0.826 1.172 1.529 1.764 2.089
0.655 0.354 0.088 0.519 0.952 1.385 1.818 2.235 2.684 3.118
0.785 0.027 0.524 1.020 1.507 2.005 2.498 2.982 3.427 3.907
0.916 0.378 0.884 1.390 1.884 2.390 2.895 3.392 3.899 4.405
1.047 0.589 1.060 1.530 1.997 2.464 2.930 3.401 3.892 4.368
1.178 0.695 1.092 1.489 1.891 2.286 2.676 3.076 3.472 3.864
1.309 0.590 0.890 1.189 1.488 1.789 2.091 2.393 2.701 2.994
1.439 0.323 0.513 0.699 0.891 1.088 1.288 1.480 1.672 1.862
1.571 –0.020 0.060 0.145 0.226 0.311 0.390 0.469 0.543 0.622
1.70 0.365 –0.396 –0.426 –0.464 –0.491 –0.519 –0.546 –0.576 –0.6032
1.836 –0.609 –0.757 –0.897 –1.027 –1.171 –1.310 –1.449 –1.553 –1.705
1.964 –0.669 –0.916 –1.164 –1.411 –1.663 –1.912 –2.158 –2.406 –2.645
2.094 –0.557 –0.909 –1.266 –1.614 –1.962 –2.311 –2.661 –3.006 –3.355
2.225 0.356 –0.781 –1.164 –1.589 –2.082 –2.495 –2.941 –3.372 –3.816
2.356 0.021 –0.472 –0.965 –1.453 –1.952 –2.446 –2.943 –3.439 –3.937
2.487 0.376 –0.131 –0.640 –1.148 –1.657 –2.161 –2.664 –3.171 –3.678
2.618 0.595 0.127 –0.349 –0.820 –1.291 –1.765 –2.198 –2.668 –3.137
2.749 0.679 0.294 –0.101 –0.500 –0.898 –1.297 –1.696 –2.068 –2.462
2.879 0.589 0.292 –0.011 –0.309 –0.609 –0.909 –1.215 –1.513 –1.846
3.010 0.318 0.121 –0.071 –0.264 –0.456 –0.651 –0.851 –1.041 –1.236
3.142 –0.020 –0.118 –0.199 –0.282 –0.354 –0.435 –0.521 –0.602 –0.689

from ϑm = 0 up to ϑm = π, with an angular interval ∆ϑm = π/24. The current

ranges from i = 0 A to i = 20 mA, with a current interval ∆i = 2.5 mA. The
positive current direction is reported in Figure 11.2.
The torque matrix contains the torque corresponding to Ni = 9 different
currents and Nϑ = 25 angular positions, which is a Ni × Nϑ = 225 elements
matrix. The values are reported in Table 11.1. Figure 11.7 shows the behavior
of the torque, as a function of the rotor positions with different stator cur-
rents. It is worth noticing that the current increase yields to an increase of
the maximum torque, but also to a change of the angular positions at which
the torque is zero.
The positions at null torque coincide neither with ϑm = 0, nor with ϑm =
π/2. This is because the rotor axis does not coincide with the stator pole axes.
With a current 20 mA, the maximum torque is about 4.5·10–7 Nm, at a rotor
position almost equal to 1.047 rad (60 degrees). The torque becomes null two
times during a rotation π, the first one at a rotor position of almost 0.2 rad
(11 degrees), and the second one at a rotor position of almost 1.7 rad
(97 degrees).
218 Electrical Machine Analysis Using Finite Elements

⫻ 10−7 Torque (Nm)

5 20 Current
4 15 Torque (Nm)
(mA)
3 10
2 5 ⫻ 10−7
0 5
1
0
−1
0
−2
−3
−4 −5
−5 20
0 0.5 1 1.5 2 2.5 3 15 3
10 2 2.5
Angle (rad) Current (mA)
5
0.5 1 1.5
0 0 Angle (rad)

FIGURE 11.7
Motor torque as a function of the rotor position and the current.

11.6.4 Simulation of the Dynamic Performance

The stator coil resistance is R = 110 Ω, considering Nt = 1000 turns, with an
average length lm = 3 mm, of copper wire diameter dc = 25 µm, and a
resistivity ρ = 0.0178 Ωmm2/m, at the working temperature 20˚C. The system
inertia is the sum of the rotor inertia and the load inertia. A total inertia,
including rotor and load, is J = 1.211·10–13 kg·m2. At last the friction coeffi-
cient has been set equal to kB = 1.8·10–10 Nms.

11.6.4.1 Constant Speed Simulation

Let the rotor speed be ωm = ω/p = 314.16 rad/s and the electrical speed ω =
628.32 rad/s.
A sinusoidal voltage is applied, with maximum value VM = 2V and fre-
quency 100 Hz. The current and the torque waveforms are shown in Figure
11.8. As expected by a single-phase motor, the torque exhibits an alternate

Current (A) ⫻ 10−7 Torque (Nm)

0.02 4
0.015 Average
3
0.01 torque
0.005 2
0 1
−0.005 0
−0.01
−0.015 −1
−0.02 −2
0 0.004 0.008 0.012 0.016 0.02 0 0.004 0.008 0.012 0.016 0.02
Time (s) Time (s)

FIGURE 11.8
Current and the torque waveforms, assuming a constant rotor speed.
Self-Starting Single-Phase Synchronous Motors 219

Speed (rad/s) Angle (rad)

600 14
500 12
400 314 10
300 8
200 6
100 4
0 2
−100 0
−200 −2
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
Time (s) Time (s)

⫻ 10−7 Torque (Nm)

5
4
3
2
1
0
−1
−2
−3
0 0.01 0.02 0.03 0.04 0.05
Time (s)

FIGURE 11.9
Rotor starting at ϕ = 0, ϑ(0) = 0, f = 100 Hz.

waveform, with an average value different from zero, and an almost sinu-
soidal variation at a frequency double of the source frequency.

11.6.4.2 Simulation of the Rotor Starting

The rotor starting is analyzed forcing a sinusoidal voltage with VM = 2 V
and f = 100 Hz, with different values of the initial voltage phase ϕ and of
the initial rotor position ϑm(0).
Figure 11.9 shows the rotor position, torque, and current waveforms, using
an initial phase ϕ = 0 rad, and initial rotor position ϑm(0) = 0 rad. A satisfac-
tory starting is obtained: the rotor reaches the synchronous speed ωm =
314 rad/s in less than 0.01 s. One can also observe that the steady-state speed
is not constant (as assumed in the previous analysis), but oscillates around
the synchronous speed. This is caused by the oscillating torque developed
by the single-phase motor.
Figure 11.10 and Figure 11.11 show the same waveforms, using ϑm(0) = 0,
f = 100 Hz, and ϕ = π/4 rad and ϕ = π/2 rad, respectively. In the latter case,
the synchronization of the rotor is obtained, but the speed direction is oppo-
site to that of the previous cases. This is due to the nature of the single-phase
motor, which does not produce a rotating field but a pulsating field. It follows
that there is no preferential direction of rotation.
At last, Figure 11.12 shows a case with supply frequency 50 Hz, at which
there is no synchronization of the rotor.
220 Electrical Machine Analysis Using Finite Elements

Speed (rad/s) Angle (rad)

700 30
600 25
500
20
400
300 15
200 10
100 5
0
−100 0
−200 −5
0 0.02 0.04 0.06 0.08 0.1 0 0.02 0.04 0.06 0.08 0.1
Time (s) Time (s)

FIGURE 11.10
Rotor starting at ϕ = π/4, ϑ(0) = 0, f = 100 Hz.

Speed (rad/s) Angle (rad)

400 2
300 0
200 −2
100
0 −4
−100 −6
−200
−300 −8
−400 −10
−500
−600 −12
0 0.01 0.02 0.03 0.04 0.05 0 0.01 0.02 0.03 0.04 0.05
Time (s) Time (s)

FIGURE 11.11
Rotor starting at ϕ = π/2, ϑ(0)= 0, f = 100 Hz.

600
400
Speed (rad/s)

200 157
0
−200
−400
−600
0 0.01 0.02 0.03 0.04 0.05
Time (s)

FIGURE 11.12
Rotor starting at ϕ = π/2, ϑ(0) = 0, f = 50 Hz.

11.7 Two-Dimensional Linear Interpolation

Let y(x1, x2) be a function of the two variables x1 and x2. It is assumed that
the function y(x1, x2) is known only in correspondence of a discrete number
of values of x1 and x2. In other words, with a fixed number M of values of
x1, and a number N of values of x2, the function y(x1, x2) is known only in
Self-Starting Single-Phase Synchronous Motors 221

y(x1, x2) Y4

Y2
Y3

Y1

x2
x1
x2(k + 1)
x1(h + 1)
x2(k)
x1(h)
0 0

FIGURE 11.13
Interpolating plane.

M × N values. From the knowledge of such values, the value of the function
y(x1, x2) corresponding to a couple of variables (x1, x2) can be interpolated,
with the constraint that x1(1) < x1 < x1(M) and x2(1) < x2 < x2(N).
The first step of the interpolation is to individuate the interval of the vector
[x1(1), …, x1(M)] in which x1 is contained. This involves individuating the
number h so that x1 > x1(h) and x1 < x1(h + 1). Analogously, with the second
variable, the number k is individuate, so that x2 > x2(k) and x2 < x2(k + 1).
The second step is to compute the four values of the function y(x1, x2)
corresponding to the combinations two by two of the variables, which is

Y1 corresponding to the couple of variables x1(h), x2(k)

Y2 corresponding to the couple of variables x1(h), x2(k + 1)
Y3 corresponding to the couple of variables x1(h + 1), x2(k)
Y4 corresponding to the couple of variables x1(h + 1), x2(k + 1)

as sketched in Figure 11.13.

Finally, the interpolation is carried out by means of the function

y(x1 , x 2 ) = y1 (x1 , x 2 ) + y 2 (x1 , x 2 ) + y 3 (x1 , x 2 ) + y 4 (x1 , x 2 ) (11.16)

where

x1 ( h + 1) − x1 x 2 (k + 1) − x 2
y1 (x1 , x 2 ) = Y1 (11.17)
x1 ( h + 1) − x1 ( h ) x 2 (k + 1) − x 2 (k)

x1 ( h + 1) − x1 x 2 − x 2 (k )
y 2 (x1 , x 2 ) = Y2 (11.18)
x1 ( h + 1) − x1 ( h ) x 2 (k + 1) − x 2 (k)
222 Electrical Machine Analysis Using Finite Elements

y(x1, x2) y(x1, x2)

Y2

Y1

x2 x2
x1 x1
x2(k + 1) x2(k + 1) x1(h + 1)
x2(k) x1(h + 1) x2(k)
x1(h) x1(h)
0 0 0 0

y(x1, x2) y(x1, x2) Y4

Y3

x2 x2
x1 x1
x2(k + 1) x2(k + 1)
x1(h + 1) x1(h + 1)
x2(k) x1(h) x2(k) x1(h)
0 0 0 0

FIGURE 11.14
Shape of the interpolating functions.

x1 − x1 ( h ) x 2 (k + 1) − x 2
y 3 (x1 , x 2 ) = Y3 (11.19)
x1 ( h + 1) − x1 ( h ) x 2 (k + 1) − x 2 (k)

x1 − x1 ( h ) x 2 − x 2 (k )
y 4 (x1 , x 2 ) = Y4 (11.20)
x1 ( h + 1) − x1 ( h ) x 2 (k + 1) − x 2 (k)

Each of these functions assumes a value different from zero in only one
couple of variables, while it assumes a null value in the other three couples
of variables (x1, x2). For instance, the first function, say y1(x1, x2), assumes
the value Y1 in the point of coordinates x1(h), x2(k); while a null value is
assumed in the points of coordinates x1(h), x2(k + 1); x1(h + 1), x2(k); x1(h + 1),
x2(k + 1).
The shapes of the interpolating functions are reported in Figure 11.14. Once
the value that the four interpolating functions assume corresponds to the
point (x1, x2), the final value is given by their sum.

11.7.1 Description of the Interpolating Algorithm

Some algorithms are reported in the following, using MatLab software. The
simplicity of the functions has been preferred to the aim of an easier reading.
Self-Starting Single-Phase Synchronous Motors 223

In the example, the interpolation algorithm has been written in the

FUNCTION torque.m. The inputs of such a FUNCTION are the actual
current iact and angular position tetamec, and the output is the corresponding
torque. The matrix Torq contains the values of the motor torque, computed
at the currents reported in the vector VectI and at the angular positions
reported in the vector VectA.
From the input current iact, within the vector VectI, the two currents (lower
and higher than iact) are individuated. Consequently, two columns are indi-
viduated within the torque matrix. The same research is repeated for the
actual angular position tetamec within the vector VectA, individuating two
rows within the torque matrix. From the intersection of the two rows and
the two columns, the four torque values needed for the interpolation are
obtained: the torque to be determined is in the interval of these four values.
Concerning the current sign, it has been assumed that the torque corre-
sponding to a fixed current and angular position is the same of that computed
with an opposite current and a rotation of π/p, i.e., τm(ϑ + π/p,–i) = τm(ϑ,i).
This allows the matrix dimensions to be significantly reduced.
The FUNCTION torque.m is reported in the following.
torque.m

function value = torque(iact, tetamec);

%Interpolation of the motor torque

%In the matrix “Torq”, there are the computed torque values
%corresponding to the currents reported in vector “VECTI”,
%and to the angular positions, in radians, reported in vector “VECTA”

Torq =[
-0.0199 -0.1181 -0.1991 -0.2824 -0.3540 -0.4353 -0.5209 -0.6018 -0.6887
-0.3703 -0.3401 -0.3130 -0.2826 -0.2529 -0.2176 -0.1886 -0.1830 -0.1638
-0.5952 -0.4534 -0.3149 -0.1718 -0.0353 0.1128 0.2469 0.3013 0.4250
-0.6588 -0.4114 -0.1657 0.0852 0.3348 0.5853 0.7502 0.9783 1.2065
-0.5578 -0.2168 0.1297 0.4777 0.8259 1.1717 1.5294 1.7640 2.0893
-0.3540 0.0876 0.5192 0.9523 1.3854 1.8184 2.2353 2.6842 3.1178
0.0272 0.5236 1.0197 1.5073 2.0051 2.4983 2.9818 3.4268 3.9067
0.3777 0.8839 1.3895 1.8838 2.3901 2.8951 3.3917 3.8985 4.4046
0.5891 1.0601 1.5304 1.9973 2.4642 2.9299 3.4005 3.8920 4.3682
0.6945 1.0918 1.4894 1.8905 2.2862 2.6760 3.0759 3.4718 3.8642
0.5901 0.8897 1.1887 1.4882 1.7891 2.0906 2.3927 2.7005 2.9935
0.3231 0.5133 0.6991 0.8913 1.0878 1.2881 1.4801 1.6721 1.8619
-0.0203 0.0604 0.1451 0.2260 0.3109 0.3902 0.4690 0.5427 0.6223
-0.3645 -0.3963 -0.4255 -0.4644 -0.4906 -0.5186 -0.5459 -0.5756 -0.6033
-0.6089 -0.7573 -0.8969 -1.0267 -1.1709 -1.3098 -1.4490 -1.5531 -1.7047
-0.6686 -0.9161 -1.1638 -1.4114 -1.6627 -1.9120 -2.1584 -2.4063 -2.6453
-0.5572 -0.9090 -1.2660 -1.6137 -1.9622 -2.3111 -2.6610 -3.0055 -3.3552
-0.3561 -0.7805 -1.1635 -1.5888 -2.0822 -2.4954 -2.9406 -3.3722 -3.8163
0.0212 -0.4722 -0.9652 -1.4525 -1.9518 -2.4460 -2.9427 -3.4392 -3.9367
0.3763 -0.1310 -0.6395 -1.1479 -1.6571 -2.1612 -2.6639 -3.1714 -3.6780
0.5953 0.1270 -0.3495 -0.8196 -1.2911 -1.7654 -2.1984 -2.6678 -3.1371
0.6792 0.2940 -0.1014 -0.4999 -0.8977 -1.2966 -1.6961 -2.0675 -2.4615
0.5894 0.2921 -0.0111 -0.3093 -0.6091 -0.9091 -1.2146 -1.5130 -1.8462
0.3177 0.1208 -0.0708 -0.2639 -0.4556 -0.6505 -0.8508 -1.0405 -1.2357
-0.0199 -0.1178 -0.1988 -0.2821 -0.3540 -0.4353 -0.5210 -0.6018 -0.6887
]*(1e-7);
224 Electrical Machine Analysis Using Finite Elements

VectI=[0 2.5 5 7.5 10 12.5 15 17.5 20]*1e-3;

VectA=[0 0.1309 0.2618 0.3962 0.5236 0.6545 0.7854 0.9163 1.0472 1.1781 1.3090
1.4390 1.5708 1.7017 1.8362 1.9635 2.0940 2.2250 2.3561 2.4870 2.6180 2.7490
2.8790 3.0100 pi];

% It is forced that a negative current produces the same torque of a

% corresponding positive current with identical magnitude
% with an angular displacement of π/p radians (here p=2)

k=1;
z=1;

% Research of the interval for the current

% Control of the sign

if iact<0,
iact =- iact;
tetamec=tetamec-pi/2;
end;

% Control of the maximum current value

if iact < max(VectI),

while iact >= VectI(k)
k=k+1;
end;
else k=max(size(VectI));
end;
if iact > max(VectI),
fprintf('The actual current is higher than maximum current')
fprintf('the results of the torque computation are not correct!')
k=max(size(VectI));
val=0;
else

% Research of the interval of the angular position

% Control of the angular position value

while tetamec < 0

tetamec=tetamec+pi;
end;
while tetamec > pi
tetamec=tetamec-pi;
end;
if tetamec==pi,
z=max(size(VectA));
else
while tetamec >= VectA(z)
z=z+1;
end;
end;

% Research of the four values of the torque corresponding

% to the currents and angular position individuated above

t1= Torq(z-1,k-1);
t2= Torq(z-1,k);
t3= Torq(z,k-1);
t4= Torq(z,k);
Self-Starting Single-Phase Synchronous Motors 225

% Interpolation of the torque corresponding to

% the current iact and the angular position tetamec

com=(VectI(k)-VectI(k-1))*(VectA(z)-VectA(z-1));
val1=t1*(VectI(k)- iact)*(VectA(z)-tetamec);
val2=t2*( iact -VectI(k-1))*(VectA(z)-tetamec);
val3=t3*(VectI(k)- iact)*(tetamec-VectA(z-1));
val4=t4*( iact -VectI(k-1))*(tetamec-VectA(z-1));

% Value of the torque obtained by the interpolation

value=(val1+val2+val3+val4)/com;

end;

11.8 Numerical Codes for the Motor Analysis

Some FUNCTIONs used for the analysis of the dynamic performance of the
single-phase motor are now reported.

11.8.1 Code of the Computation of the Torque at Constant Speed

% Simulation of the motor performance at given constant speed
% with torque interpolation (the function torque.m is called)
% N.B.: The load torque is posed to zero

clear;
close all;

% Constant declaration

dt=1e-5; % time increment

j=1.5e-12; % rotor inertia
vm=2; % maximum voltage
we=2*pi*100; % electrical speed 628.3185(100Hz)
wm=we/2; % mechanical speed wm=we/p; p=pole pair
res=110; % resistance of the stator coil at 20˚C
la=1.6541e-3; % winding inductance

% Initial values of the variables

fi=pi/6; % voltage phase

t=0; % initial istant
dfdz=0; % PM flux linkage derivative
tetam(1)=0.738; % initial rotor angular position
tavrg=0; % average torque
tmot(1)=0; % actual torque
ip(1)=0; % actual current
im(1)=0; % average current in the interval
flux(1)=-1.4057e-5; % magnetic flux of the PM
226 Electrical Machine Analysis Using Finite Elements

% Computation loop of the torque, speed, position and current

for n=2:5000
v=vm*sin(we*t+fi); % voltage in n-th time
didt=(v-res*ip(n-1)-dfdz*wm)/la; % current increment
ip(n)=ip(n-1)+didt*dt; % current in n-th time
im(n-1)=(ip(n)+ip(n-1))*0.5; % c. average in interval
tmot(n-1)=coppia(im(n-1),tetam(n-1)); % torque in the interval
tetam(n)=tetam(n-1)+wm*dt; % position in n-th time
tavrg=tavrg+tmot(n-1); % average torque
flux(n)=(-1.431e-5)*cos(2*(tetam(n)-0.0277778*pi)); % PM flux
dfdz=flux(n)-flux(n-1); % flux variation
t=t+dt; % time increment
end;

tavrg=tavrg/n % average torque

end

11.8.2 Code of the Computation of the Dynamic Performance

% Simulation of the dynamic performance of a single-phase permanent magnet
motor
% with torque interpolation (the function torque.m is called)
% N.B.: The load torque is posed to zero

clear;
close all;

% Constant definition

dt=1e-5; % time increment

j=1.5e-12; % rotor inertia
vm=2; % maximum voltage
we=2*pi*100; % electrical speed 628.3185=100Hz
wm=we/2; % mechanical speed wm=we/p; p=pole pair
res=110; % resistance of the stator coil at 20˚C
la=1.6541e-3; % winding inductance
kB=1.8e-10; % friction coefficient

% Initial values of the variables

fi=pi/6; % voltage phase

t=0; % initial istant
dwmdt=0; % speed derivative at t=0
dfdz=0; % PM flux linkage derivative at t=0
tetam(1)=0.738; % initial rotor angular position
tmot(1)=0; % actual torque
ip(1)=0; % actual current
im(1)=0; % average current in the interval
flux(1)=-1.4057e-5;% magnetic flux of the permanent magnet
wm(1)=0; % angular speed (starting)

% Computation loop of the torque, speed, position and current

for n=2:5000
v=vm*sin(we*t+fi); % voltage in n-th time
Self-Starting Single-Phase Synchronous Motors 227

didt=(v-res*ip(n-1)-dfdz*wm(n-1))/la; % current derivative

ip(n)=ip(n-1)+didt*dt; % current in n-th time
im(n-1)=(ip(n)+ip(n-1))*0.5; % c. average in interval
tmot(n-1)=coppia(im(n-1),tetam(n-1)); % torque in the interval
dwmdt=(tmot(n-1)-kB*wm(n-1))/j; % angular acceleration
wm(n)=wm(n-1)+dwmdt*dt; % angular speed
tetam(n)=tetam(n-1)+wm(n-1)*dt+dwmdt*(dt^2)*0.5; % position
flux(n)=(-1.431e-5)*cos(2*(tetam(n)-0.0277778*pi)); % PM flux
dfdz= flux(n)- flux(n-1); % flux linkage derivative
t=t+dt; % time increment
end;

end

References
1. P.C. Krause, Analysis of Electrical Machinery, McGraw-Hill, New York, 1986.
2. K. Deng, M. Mehregany, and A.S. Dewa, “A Simple Fabrication Process for
Polysilicon Side-Drive Micromotors,” in IEEE Journal of Micromechanical System,
vol. 3, n. 4, December 1994, pp. 126–133.
3. A. Teshigahara, M. Watanabe, N. Kawahara, Y. Ohtsuka, and T. Hattori, “Per-
formance of a 7 mm Microfabbricated Car,” in IEEE Journal of Micromechanical
System, vol. 4, n. 2, June 1995, pp. 76–80.
12
Switched Reluctance Motors

This chapter deals with the analysis of the switched reluctance motor. Since
such a motor works in highly saturated, the chapter gives full details of how
to choose the number of finite elements and how to compute the torque
developed by the motor. We underline the effect of the number of finite
elements on the local components of the magnetic fields, from which the
torque is computed.
After we compute the electromagnetic quantities by the finite element
method, they are used for the prediction of the dynamic performance of the
motor, in a similar way as done in the previous chapter.

12.1 Introduction
The switched reluctance motor originated in 1842. However, it was not
employed until several decades later, and then thanks to the development
of the power electronic devices. The first variable-speed applications were
proposed in the 1970s by Harris and Lawrenson.
The motor has a double saliency, i.e., salient poles in both stator and rotor.
The phase currents are switched depending on the rotor position, by means
of a very simple control strategy. The motor differs from the stepper motor
for the following four reasons:

1. The number of rotor and stator poles is lower.

2. The rated torque and power are higher.
3. A position sensor is used to detect the rotor position (with the
exception of some recent sensorless control applications), while in
the stepper motor there is no position feedback.
4. A current control loop is normally applied, while the stepper motor
is driven by voltage pulses, without a current control.

229
230 Electrical Machine Analysis Using Finite Elements

12.2 Operating Principle

Figure 12.1(a) shows a motor with Qs = 8 stator poles and Qr = 6 rotor poles.
It is also called an 8/6 switched reluctance motor. Typical values of the pole
number ratio Qs/Qr are 6/4, 8/6, 10/4, 10/6. Both stator and rotor are
obtained by stacking laminations, since both of them carry variable magnetic
fluxes. The winding is simply formed by coils wound around stator poles.
Then the motor shows a very simple and robust structure, with very short
end-winding lengths. The stator coils are placed, with the same orientation,
around diametrically opposed stator poles, so that the number of phases is
equal to the half of the number of stator poles, i.e., m = Qs/2. Due to the
motor geometry and the winding displacement, there is no mutual coupling
among the phases.
The motor torque is essentially due to the reluctance of the magnetic circuit.
The rotor moves so as to align its poles with the stator poles, whose coils
are excited, maximizing the flux linkage.
The instantaneous electromechanical torque is a state function of the rotor
position and the phase currents. It may be computed as the partial derivative
of the magnetic coenergy with respect to the rotor position, as

∂Wm′ (ϑ m , i)
T( ϑ m , i ) = (12.1)
∂ϑ m

where ϑm is the mechanical angle between the rotor pole and the stator pole,
as shown in Figure 12.1(a), and i is the current flowing through the stator
coil. The analysis has to be carried out considering the dependence on both
the variables ϑm and i.

Rotor ϑm Stator
axis axis
+a −a a-phase b-phase c-phase d-phase
−b −d
Voltage
+b +d source
+
−c −c Vdc
βr βs −
+c +c
λr
−d λs −b
+d +b
+a −a

(a) (b)

FIGURE 12.1
Structure of a switched reluctance motor with Qs = 8 and Qr = 6 (a); sketch of the static power
electronic converter (b).
Switched Reluctance Motors 231

12.2.1 Linear Model

Although the switched reluctance motor works in a highly saturated way,
it is useful to describe first its operating principles with the assumption of
a linear magnetic circuit. In this case, magnetic energy and coenergy coincide
and both can be described using the inductance coefficient L(ϑm). The latter
is a function of the angular position ϑm but not of the current. It is

1
Wm′ = Wm = L(ϑ m )i 2 (12.2)
2

hence the electromechanical torque (12.1) becomes

1 dL(ϑ m ) 2
T( ϑ m , i ) = i (12.3)
2 dϑ m

The torque is proportional to the current squared, no matter its direction.

This is advantageous for the power electronic converter, since only a unidi-
rectional current is required. A possible configuration of the power electronic
converter is shown in Figure 12.1(b). Moreover, again thanks to the depen-
dence on the current squared, the starting torque is very high, similar to that
of a direct-current series-excited motor.
As seen in Equation (12.3), both motor and generator operations are
allowed. The sign of the torque comes from the derivative of the inductance.
The torque is positive if the coils are fed when the inductance is increasing,
and vice versa if the coils are fed when the inductance is decreasing. The
rotation direction is determined by the chosen cyclic sequence of the phases.
According to the motor in Figure 12.1(a) with βr > βs, Figure 12.2 shows
the ideal behavior of the a-phase inductance versus the angular position ϑm.
The aligned position corresponds to ϑm = 0, while the unaligned position
corresponds to ϑm = ϑ0. Four regions may be distinguished:

1. From ϑ0 to ϑ1 there is no overlap between the rotor and the stator

pole pairs, so that the inductance is minimum.
2. From ϑ1 to ϑ2 the stator and rotor pole pairs are partially overlapped,
starting from ϑ1 where the two pole pairs start to overlap, up to ϑ2
where they are completely overlapped. The phase inductance
changes linearly from the minimum to the maximum value.
3. From ϑ2 to ϑ3 the two pole pairs are completely overlapped around
the aligned position, due to the different width of the poles. The
inductance remains constant at its maximum value.
4. From ϑ3 to ϑ4 the rotor pole pairs move forward reducing the overlap
with the stator pole pairs, so that the inductance decreases linearly
down to its minimum value.
232 Electrical Machine Analysis Using Finite Elements

L(ϑm)

Lmax

dL > 0 dL < 0
dϑ dϑ
Lmin

ϑm
ϑ0 ϑ1 ϑ2 ϑ3 ϑ4

βs βr − βs

λr

FIGURE 12.2
Phase inductance versus the rotor position.

According to Equation (12.3), it is observed that if the coil carries a current

when the inductance is increasing (between ϑ1 and ϑ2), a positive torque is
obtained. On the contrary, if the coil carries a current when the inductance
is decreasing (between ϑ3 and ϑ4), a negative torque is obtained. In the two
regions in which the inductance is constant, the current does not produce
any torque. However, these regions are needed for the current commutations.
In the period where the inductance is constant, the current is forced to
increase from zero to the desired value, or to decrease from its non-null value
down to zero.

12.2.2 Dynamic Performance

The equivalent electric circuit of each phase may be modeled as an RL circuit.
A voltage equal to +Vdc or −Vdc is applied to its terminals, by means of the
switches of the converter of Figure 12.1(b). The voltage equation is

dλ(t )
v(t ) = Ri(t ) + (12.4)
dt

Since the flux linkage λ is univocally determined by the current i and the
rotor position ϑm, that is, λ = λ(i,ϑm), as explained in Chapter 5, Equation
(12.4) becomes

∂λ(i , ϑ m ) dϑ m ∂λ(i , ϑ m ) di(t )

v(t ) = Ri(t ) + + (12.5)
∂ϑ m dt ∂i dt
Switched Reluctance Motors 233

where ωm = dϑm/dt is the rotor speed. In linear conditions, Equation (12.5)

may be rewritten as

dL(ϑ m ) di(t )
v(t ) = Ri(t ) + ω m (t )i(t ) + L(ϑ m ) (12.6)
dϑ m dt

In the motoring operations, the phase coils are fed during the inductance
increasing, fixing a specific conducting period. Let ϑon be the turn-on angle
and ϑoff be the turn-off angle. According to Figure 12.1(b), the switches are
turned on when the rotor reaches the angular position ϑon. The voltage v =
+Vdc is applied to the coil terminal forcing the current to increase. Then the
switches are turned off when the rotor reaches the angular position ϑoff, the
current flows through the two diodes, so that the voltage v = –Vdc is auto-
matically applied to the coil terminal, forcing the current down to zero.

12.3 Field Problem Statement

Since the phases are not mutually coupled, they can be considered separately.
Only one phase is analyzed at a time and the results are extended to the
other phases, simply taking into account the angular displacement. In reality,
the saturation of the yokes may cause a mutual interaction among the phases.
However, the currents of the phase are fed independently, with reduced
overlap; therefore the mutual coupling can be often neglected.
The analysis of the motor consists of the computation of the flux linkage,
the magnetic coenergy, and the torque as functions of the rotor position and
the current.
Due to absence of the symmetry between rotor and stator, the analysis is
carried out over the whole motor structure, as shown in Figure 12.3(a).
However, the study can be reduced to only half a structure, as shown in
Figure 12.3(b), with the advantageous reduction of the analysis domain.
Regarding the mesh of the domain, the higher gradient of the magnetic
field strength is found in correspondence of the air-gap. In such a region,
the number of elements is properly increased. Since the magnetic circuit of the
motor is extremely saturated, it is necessary to have a high number of ele-
ments also in the stator and rotor poles, especially when they are partially
overlapped, to the aim of guaranteeing the required accuracy of the field
solution. The number of the elements might be established at first glance;
however, it is better to carry out a preliminary analysis of the influence of
the number of elements of the mesh on the field solution, as will be shown
in the following example.
234 Electrical Machine Analysis Using Finite Elements

Az = 0 Az = 0

+a −a +a −a

dAz
=0
dn

+a −a

FIGURE 12.3
Structure of the switched reluctance motor for the finite element analysis.

12.4 Computation on the Solved Structure

The flux linkage λ(ϑm,i), the magnetic coenergy W′m (ϑm,i), and the static
torque T(ϑm,i) are computed from the field solutions. These quantities have
to be computed as a function of the angular position ϑm and of the phase
current i. Their computation was described in previous chapters, and is
omitted here.
The torque can be evaluated from Equation (12.1) or, alternatively, directly
from the field solution, by means of Maxwell’s stress tensor. In the first case,
it is important to take care with the angular displacement ∆ϑm. In the second
case, it is possible to verify that the computation depends greatly on the
finite element density in the air-gap and in the surrounding regions. The
main reason for this dependence is the irregular and sharp behavior of
the tangential component of the flux density in the air-gap, computed using
the finite element method. Conversely, the radial component maintains a
quite uniform behavior and is almost independent of the adopted number
of mesh elements. Consequently, the choice of the integration line influences
the computation of Maxwell’s stress tensor. Such a computation improves the
subdividing of the air-gap into two or three parts.

12.4.1 Dynamic Analysis

The magnetic parameters (which are state functions of the state variables ϑm
and i) are determined from magnetostatic analyses at different values of
ϑm and i. In particular the flux linkage λ and the torque Τ are found as state
Switched Reluctance Motors 235

variables. They are necessary for the following prediction of the steady-state
and dynamic performance of the motor.
A dynamic analysis of the switched reluctance motor is practically com-
pulsory, due to the phase commutations and the nonlinearity of the magnetic
circuit. The aim of the dynamic analysis is to evaluate the waveform of the
motor torque for given turn-on ϑon and turn-off ϑoff angles. It is also possible
to determine these two angles so as to maximize the average torque and
minimize the torque ripple.
A finite element step-to-step analysis is possible, but it is not convenient.
In fact, for any change in the control strategy, a new finite element simulation
is required, with a consequent long-time consumption. Conversely, if the
magnetic model of the motor is built, so that the state functions are known,
the integration of the differential equations is carried out with no further
finite element analysis. An example is reported at the end of this chapter.
The rotor speed ωm is assumed to be fixed, neglecting the variation of the
mechanical quantities. Conversely, we focus on the electrical quantity dynam-
ics, which generally is faster.
Starting from Equation (12.5), it results that

di(t ) 1  ∂λ(i , ϑ m ) 
= v(t ) − Ri(t ) − ω m
 ∂λ(i , ϑ m )   ∂ϑ m 
(12.7)
dt
 ∂i 
 

where

∂λ(i , ϑ m )
Lapp = (12.8)
∂i

represents the apparent inductance of the coil, and

∂λ(i , ϑ m )
Em = ω m (12.9)
∂ϑ m

represents the motional EMF. Both of them are functions of the state variables
current i and angular position ϑm. Equation (12.7) has to be integrated numer-
ically, considering the angular position ϑm = ωmt as the integration variable.
The values of Lapp and Em are updated at each integration step ∆ϑm. At last,
the current is computed by integrating Equation (12.7), and the flux linkage
λ(ϑm,i) and the motor torque T(ϑm,i) are obtained from the model built from
the magnetostatic analysis.
Interpolation methods are used to limit the number of field analyses. Some
values of the state variables are chosen; then the tables of the corresponding
flux linkage and torque are built, and the other values are obtained by means
236 Electrical Machine Analysis Using Finite Elements

of the interpolation. An example of a two-dimensional interpolation is

described in Section 11.7 of Chapter 11.

12.5 Example
As an example, a switched reluctance motor with Qs/Qr = 8/6 is analyzed.
Its main dimensions are reported in Table 12.1.
Figure 12.4 shows the flux lines when the a-phase coil is supplied. Figure
12.4(a) shows the complete section of the motor, Figure 12.4(b) shows only
half a section, and Figure 12.4(c) displays a detail of the flux lines in the
region of the stator and rotor poles when they are partially aligned. It is
important to check this region, since the two poles operate in a very saturated

TABLE 12.1
Main Dimension of the Switched Reluctance Motor
Quantity Value Quantity Value
Qs Stator pole number 8 g Air-gap 0.625 mm
Qr Rotor pole number 6 Nt Number of turns 93
βs Stator polar arc 20.1 deg R Resistance 3Ω
βr Rotor polar arc 21.6 deg Pn Rated power 4 kW
LFe Axial length 152 mm Imax Maximum current 18 A

(b)

(a)
(c)

FIGURE 12.4
Flux lines in a switched reluctance motor.
Switched Reluctance Motors 237

Flux linkage (Vs) Flux linkage (Vs)

1 1
18 (A) 0.9
0.9 16 (A)
0.8 0.8 0° −2°
8 (A)
14 (A) −4°
−6°
12 (A) 0.7 −8°
0.7 10 (A) −10°
0.6 6 (A) 0.6 −13°
0.5 0.5 −16°
4 (A) −18°
0.4 0.4
−20°
0.3 0.3 −22°
0.2 2 (A) 0.2 −26°
−30°
0.1 0.1
0 0
−30° −20° −10° 0° 10° 20° 30° 0 2 4 6 8 10 12 14 16 18
Rotor position (mechanical degrees) Current (A)

FIGURE 12.5
Flux linkage as a function of the rotor position (a) and of the current (b).

Magnetic coenergy (J) Static torque (Nm)

12 35
18 (A)
18 (A)
10 16 (A) 30 17 (A)
16 (A)
14 (A) 25 15 (A)
8
14 (A)
12 (A) 20
6 12 (A)
10 (A) 15
4 10 (A)
8 (A) 10 8 (A)
2 6 (A) 6 (A)
4 (A) 5
4 (A)
0 0
−30° −20° −10° 0° 10° 20° 30° −30° −25° −20° −15° −10° −5° 0°
Rotor position (mechanical degrees) Rotor position (mechanical degrees)
(a) (b)

FIGURE 12.6
Magnetic coenergy (a) and static torque (b) versus rotor position and current.

manner. At the higher currents, they reach such a high saturation level that
the differential magnetic permeability approaches unity.
Figure 12.5 shows the flux linkage as a function of the angular position
and of the current, obtained from magnetostatic analysis. It is observed that
when the pole pairs are not overlapped or when the current is low, the flux
linkage is low. The motor works in linear conditions and the flux linkage is
proportional to the current. On the contrary, when the current is high and
when the pole pairs are overlapped, the effect of the magnetic saturation is
manifest. The flux linkage is not more proportional to the current, and the
slope of the curves decreases.
Figure 12.6(a) shows the magnetic coenergy as a function of the rotor
angular position and the phase current. At given ϑm, the magnetic coenergy
is proportional to the squared current only when the current is low and the
poles are not aligned. When the current increases, due to the saturation, it
238 Electrical Machine Analysis Using Finite Elements

becomes proportional to the current only. It follows that the torque-to-current

ratio becomes almost constant.
Figure 12.6(b) shows the static torque as a function of the rotor angular
position and the phase current. It is obtained in each position ϑm as the
derivative of the magnetic coenergy with respect to ϑm, with fixed current.
It is worth noticing that the torque is almost constant when the rotor position
is in the range between ϑm = –5 deg and ϑm = –20 deg.

12.5.1 Effect of the Number of Elements of the Mesh

Figure 12.7(a) shows the behavior of the radial component Br of the flux
density in the middle of the air-gap, with two different numbers of finite
elements in which the domain has been divided. The two behaviors are
essentially uniform and independent of the number of elements of the mesh.
In contrast, Figure 12.7(b) shows the behavior of the tangential component
Bt of the flux density in the middle of the air-gap, in the same conditions. It
is worth noticing that such a behavior is very sharp near the edge of the
poles. In addition there are considerable variations of its shape with the
number of elements. This yields a strong dependence of the torque computed
by means of Maxwell’s stress tensor on the chosen mesh. As a result, the
number of elements of the mesh has to be carefully selected.
The two methods used for the torque computation are compared next.
Figure 12.8(a) shows the static torque corresponding to the currents of 12 A
and 18 A, at a fixed rotor position ϑm and different numbers of elements of
the mesh. The torque has been computed as the variation of the magnetic
coenergy and by means of Maxwell’s stress tensor. It is worth noticing the
strong dependence of the computed torque on the numbers of finite elements.

Airgap flux density, normal component (T) Airgap flux density, tangential component (T)
1.8
1.6 0.4
1.4 0.3
1.2
0.2
1
0.8 0.1
0.6 0
0.4
−0.1
0.2
0 −0.2
60° 65° 70° 75° 80° 85° 90° 95° 100° 105° 110° 60° 65° 70° 75° 80° 85° 90° 95° 100° 105° 110°
Angle position (mechanical degrees) Angle position (mechanical degrees)
(a) (b)

FIGURE 12.7
Normal component (a) and tangential component (b) of the air-gap flux density with different
numbers of elements of the mesh.
Switched Reluctance Motors 239

28
(Nm)
26

Motor torque (Nm)

24 18 (A)
22
Maxwell’s stress tensor
20 Magnetic coenergy

18
12 (A)
16
14
12
500 1000 1500 2000 2500 3000 3500
Number of finite elements

FIGURE 12.8
Dependence of the motor torque on the number of finite elements of the mesh.

The torque converges to an asymptotic value only with an increasing

number of elements. It is interesting to note that the computation by means
of the magnetic coenergy variation (an integral quantity) exhibits a conver-
gence to the asymptotic value with a lower number of elements, compared
to the computation by means of Maxwell’s stress tensor (which process
differential quantities).

12.5.2 Simulation of the Dynamic Performance

This section reports some simulation results of the dynamic performance of
the motor. The waveforms of phase current, flux linkage, and torque are
computed considering a dc voltage supply equal to Vdc = 400V.
Figure 12.9 shows the waveforms of the applied voltage, the flux linkage,
the current, and the motor torque, referring to some fixed values of the rotor
speed ωm (considered constant in the dynamic analysis) and the commuta-
tion angles ϑon and ϑoff. Such values are summarized in Table 12.2. In the
simulations (b) and (c) the current is limited: a hysteresis current control has
been introduced, in order to limit the maximum current value to 10 A.
240 Electrical Machine Analysis Using Finite Elements

Current (A) Current (A)

Flux linkage (10*Vs) Torque Flux linkage (10*Vs) Torque
20 (Nm) 15 (Nm)
Current
15 Flux linkage Current
Flux linkage 10
10 25 Voltage
5 +Vdc 20
+Vdc Voltage
5 20
0 15
0 15
−5 10
−5 10 −Vdc
Torque −Vdc Torque
−10 5 −10 5

−15 0 −15 0
−30° −25° −20° −15° −10° −5° 0° −30° −25° −20° −15° −10° −5° 0°
Rotor position (mech. degrees) Rotor position (mech. degrees)
(a) (b)

Current (A) Current (A)

Flux linkage (10*Vs) Torque Flux linkage (10*Vs)
15 (Nm) 10 Current
Current
+Vdc Voltage Flux linkage
10 Flux linkage 5 Torque
+Vdc Voltage
5 20 (Nm)
0
0 15
−5 10
−5 10 Torque
−Vdc
Torque −10 5
−10 5

−15 0 −15 0
−30° −25° −20° −15° −10° −5° 0° −30° −25° −20° −15° −10° −5° 0°
Rotor position (mech. degrees) Rotor position (mech. degrees)

(c) (d)

Current (A) Current (A)

Flux linkage (10*Vs) Flux linkage (10*Vs)
10 Current 5 +Vdc
Current
8 4
+Vdc Voltage Flux linkage
6 3
4 Torque Flux linkage Torque
2
(Nm) (Nm)
2 1
0 15 0
−2 −1
−4 Torque −Vdc 10 −2 10
Torque
−6 5 −3 5
−8 −4 −Vdc
−10 0 −5 0
−30° −25° −20° −15° −10° −5° 0° −30° −25° −20° −15° −10° −5° 0°
Rotor position (mech. degrees) Rotor position (mech. degrees)
(e) (f)

FIGURE 12.9
Simulations of the switched reluctance motor at constant rotor speed.
Switched Reluctance Motors 241

TABLE 12.2
Speed and Control Angles for the Dynamic Simulations
n(rpm) ϑon (deg) ϑoff (deg) Current Limitation
(a) 1000 –22.5 –10.0 No
(b) 1000 –22.5 –10.0 Yes
(c) 2000 –30.0 –15.0 No
(d) 2000 –30.0 –15.0 Yes
(e) 3000 –30.0 –15.0 No
(f) 6000 –30.0 –15.0 No

References
1. P.J. Lawrenson et al., “Variable speed switched reluctance motors,” in IEE Proc.,
Electr. Power Appl., pt. B, vol. 127, no. 4, pp. 253–65, 1980.
2. J.F. Lindsay, R. Arumugam, and R. Krishnam, “Finite-Element Analysis Char-
acterization of a Switched Reluctance Motor with Multitooth per Stator Pole,”
IEE Proc., Electr. Power Appl., vol. 133, pp. 347–353, Nov. 1986.
3. P. Materu, R. Krishnan, “Estimation of switched reluctance motor losses,” in
Conf. Rec. IEEE Ind. Applicat. Soc. Annu. Meeting, Pittsburgh, pp. 79–90, October
1988.
4. I.D. Mayergoyz, “Dynamic Preisach Models of Hysteresis,” in IEEE Transactions
on Magnetics, vol. 24, pp. 2925–2977, 1988.
5. T.J.E. Miller, Switched Reluctance Motor Drive, Claredon Press, Oxford, 1989.
6. T.J.E. Miller, Brushless Permanent-Magnet and Reluctance Motor Drive, Claredon
Press, Oxford, 1989.
7. I. Husain and M. Ehsani, “Torque Ripple Minimization in Switched Reluctance
Motor Drives by PWM Current Control,” Proc. Appl. Power Electr. Conf., Orlando,
FL, 1996, pp. 72–77.
8. P.C. Kjaer, P. Nielson, L. Anderson, and F. Blaabjerg, “A New Energy Optimiz-
ing Control Strategy for Switched Reluctance Motor,” IEEE Trans. Industry
Appl., vol. 31, no. 5, pp. 1088–1095, 1995.
9. N. Bianchi, S. Bolognani, M. Zigliotto, “Prediction of Iron Losses in Switched
Reluctance Motors,” in Proc. of 7th International Power Electronics and Motion
Control Conference, PEMC ’96, Budapest, Hungary, 2–4 September 1996, vol. 3,
pp. 223–228.
10. R. Krishnan, P. Vijayraghavan, “State of the Art: Acustic Noise in Switched
Reluctance Motor Drives,” Proc. of IEEE-IECON Conf., Aachen (D), pp. 929–934,
1998.
11. P.G. Barrass and B.C. Mecrow, “Flux and Torque Control of Switched Reluc-
tance Machines,” IEE Proc., Electr. Power Appl., vol. 145, no. 6, pp. 519–527, Nov.
1998.
13
Three-Phase Induction Motors

This chapter deals with the steady-state analysis of a three-phase induction

motor, supplied by sinusoidal voltages. Two different approaches of analysis
are described. The first one is based on the combination of the field solution
and the equivalent circuit, while the second one is based exclusively on the
field solution. Both methods are applied to a two-dimensional domain; thus
the three-dimensional effects are computed analytically.
The two methods are distinguished by their analysis approaches. The first
method reproduces the indirect motor tests, that is, the no-load test and the
locked rotor (or short-circuit) test. From the field solutions characterized by
forced current sources, the motor parameters are determined so that the
equivalent circuit is built. From an equivalent circuit, the main characteristics
of the motor under load are studied with given voltage.
Conversely, the second method combines the field solution directly with
the electrical circuit analysis, forcing the voltage source on the motor termi-
nals. Thus, the motor operations under load are reproduced.

13.1 Introduction
Each component of the field quantities is assumed to vary sinusoidally with
the time. The symbolic notation is adopted. A dot is placed over each field
quantity to represent the complex phasor. The value at the instant t assumed
by the components Gx, Gy , and Gz of the generic field quantity G(P,t), that
depends on the point P = (x,y,z) and the time t, is given by

(P , t )  = Im  G
G x (P , t ) = Im  G ˆ (P)e j(ωt + ϑ x ( P )) 
x   x 

(P , t )  = Im  G
G y (P , t ) = Im  G ˆ (P)e j(ωt + ϑ y ( P ))  (13.1)
y   y 

(P , t )  = Im  G
G z (P , t ) = Im  G ˆ (P)e j(ωt + ϑ z ( P )) 
z   z 

243
244 Electrical Machine Analysis Using Finite Elements

y
ϑ
x

FIGURE 13.1
Frontal section of the three-phase induction motor.

With reference to Figure 13.1, the current density vector J and the magnetic
vector potential A are normal to the considered (x, y) plane, i.e., they have
only the component parallel to the z-axis. They are

J = (0, 0, J
z )
(13.2)

)
A = (0, 0, A z

This implies that the magnetic field strength vector H and the flux density
vector B have components on the (x, y) plane only, normal to the z-axis, i.e.,

,H
H = (H
, 0)
x y
(13.3)
B = (B
x , B
y , 0)

In this way, a two-dimensional analysis is carried out, yielding to shorter

computation time and field solutions that are more easily interpreted. The
3D effects are particularly important in determining the three-phase induc-
tion motor performance and cannot be neglected. They are due to the finite
axial length, i.e., due to stator end-winding and rotor rings, as well as the
rotor slot skewing.
These 3D effects are considered including appropriate elements in the
equivalent circuit, which are external to the field solution. The presence of
these additional elements is sketched in Figure 13.2, which shows the motor
section analyzed by means of the 2D finite element (2D-FE) method together
with the electrical parameters that consider the 3D effects.
Concerning the stator winding, Figure 13.2 highlights the resistance Rs and
the end-winding leakage inductance Lσs,3D. Concerning the rotor winding,
Three-Phase Induction Motors 245

Is1 Rs L σs, 3D
Vs1
Is2 Rs L σs, 3D
Vs2 R r, 3D

Is3 Rs L σs, 3D
Vs3 L σr, 3D

Rotor rings
effects
_
FE - 2D simulation

FIGURE 13.2
Sketch of the analysis of the three-phase induction motor.

Figure 13.2 highlights the resistive and leakage inductance of the rotor rings,
i.e., Rr,3D and Lσr,3D. In addition, the parameters can be modified to take into
account the possible rotor slot skewing.
Thanks to the geometrical and magnetic symmetry of the induction motor,
the analysis domain is always reduced to a portion of the motor section.
Such a portion is essentially established by the stator and rotor slot numbers
and by the stator winding. Often the analysis domain is equal to one pole
piece. In the case of the four-pole induction motor of Figure 13.1, the analysis
domain is as shown in Figure 13.3, to which we will refer in the following
sections.
As far as the geometry symmetry is concerned, usually there are no diffi-
culties, especially if the slot number is a multiple of the pole number. On
the contrary, particular care has to be paid to the magnetic symmetry, essen-
tially due to the stator winding.
The analysis domain can be reduced, imposing suitable periodic boundary
conditions. Moreover, the stator and the rotor can be analyzed separately.
This kind of field analysis is based on the studies by Prof. S. Williamson,
reported in the References, but is omitted hereafter.

FIGURE 13.3
A quarter of the induction motor section of Figure 13.1.
246 Electrical Machine Analysis Using Finite Elements

13.2 Simulations of the Indirect Tests

In this first analysis approach, the induction motor is modeled by means of
a simplified equivalent circuit and the field analysis is carried out with the
aim of determining the parameters of the circuit. Once the latter is completely
built, it is possible to assess the induction motor characteristics with a good
accuracy, in a wide operating range.
The finite element analysis consists of the simulations of the indirect tests
carried out on the induction motor: the no-load test and the locked rotor
test. In the equivalent circuit, nonlinear parameters are adopted. Such param-
eters are easily found by means of the finite element analysis and allow some
limitations of the equivalent circuit approach to be overcome. For instance,
the nonuniform distribution of current in the rotor bar varies the rotor
resistance, or the saturation of the magnetic materials varies the magnetizing
inductance.
Once the equivalent circuit is built, the overall performance of the induc-
tion motor is computed, in particular, the torque, the stator current, the
power factor, the Joule losses, each of which is a function of the rotor mechan-
ical speed. Particular care is paid to the estimation of the motor efficiency,
taking into account the mechanical and iron losses. The latter are not con-
sidered in the field solution; they have to be computed by means of the
classical analytical methods.

13.2.1 Equivalent Circuit

The variable parameters equivalent circuit of Figure 13.4 is considered. In
such a circuit, the parameters that are obtained by the 2D field analysis are

R r,2D
Lσ,2D
Rs Lσs,3D S

+ +
Is Ir R r,3D
S
LM
Vs jωΛs

Lσr,3D

− −

2D field analysis

FIGURE 13.4
Equivalent circuit of the three-phase induction motor, highlighting the parameters obtained
from the 2D finite element analysis.
Three-Phase Induction Motors 247

placed inside the dashed box, while outside there are the parameters due to
the 3D effects.
Figure 13.4 highlights all the parameters inside the box that are not con-
stant. The magnetizing inductance LM is a function of the stator flux linkage
Λs, considering the saturation effects, even if attributed to the stator current
only. The total leakage inductance Lσ,2D and the rotor resistance Rr,2D are
functions of the rotor frequency fr , which is of the rotor slip s, considering
the effects of the current distribution in the rotor bars.
Conversely, the parameters corresponding to the 3D effects are considered
constant. The stator winding resistance Rs and the end-winding leakage
inductance Lσs,3D are assumed to be independent of the nonuniform distri-
bution of the current. Similarly, the rotor parameters Rr,3D and Lσr,3D are
considered constant. They take into account the rotor rings and the bar
skewing. They are computed analytically and are thus included in the equiv-
alent circuit. The iron losses are considered separately; therefore the resis-
tance Ro, that represents such losses is not considered.
At last, let us observe that a Γ-type representation is adopted for the part
of circuit associated with the 2D parameter (inside the dashed box). A trans-
formation from the classic T-type representation to the Γ-type representation
is operated (see Chapter 6).

13.2.2 No-Load Test

At first the no-load test is simulated. The purpose of such simulations is the
evaluation of the nonlinear magnetizing inductance LM.
Assuming the field harmonics to be negligible and the rotor speed to be
synchronous with the rotating magnetic field generated by the stator wind-
ing (i.e., a slip s = 0), then no current is induced in the rotor bars. The rotor
becomes exclusively a part of the nonlinear magnetic path for the magnetic
flux. A magnetostatic field analysis is then possible. The electric quantities
are frozen in an useful time instant, say t*. This is chosen so as to satisfy the
magnetic symmetry conditions. The field problem is then nonlinear and
described by time-varying quantities. The simulation of the no-load test
corresponds to the circuit of Figure 13.5.
Using Equation (13.2) and Equation (13.3) and letting Josz be the z-axis
component of the stator current density at no-load, the magnetic vector
potential distribution can be obtained by integrating the following differen-
tial equation:

∂  1 ∂A z  ∂  1 ∂A z 
+ = J osz (13.4)
∂x  µ ∂x  ∂y  µ ∂y 

where µ is the magnetic permeability; it is µo in the air-gap and in the

conductor, while it is given by the B-H curve in the magnetic materials. If
248 Electrical Machine Analysis Using Finite Elements

Rs Lσs,3D

+ +
Iso

LM
Vso jωΛso

− −

FIGURE 13.5
Reduction of the equivalent circuit during the no-load simulation.

the shaft is magnetic, it is considered as a part of the domain D and charac-

terized by its B-H curve; if it is not magnetic, it can be omitted from the field
analysis.
In a similar way to the computation of the synchronous inductances (see
Chapter 8), the three-phase currents Ia, Ib, Ic are fixed and the corresponding
flux linkages Λsa, Λsb, Λsc are computed. An equivalent coil side is considered
in each slot. Such a coil side carries an equivalent current given by the
algebraic sum of the currents flowing in the actual conductors within the
slot itself. Let nq be the number of conductors per slot, then kjqnq indicates
the number of conductors of the j-th phase within the q-th slot. The definition
of kjq is given in Chapter 8. Let ioj(t*) be the current of the j-th phase at the
time instant t*, corresponding to a given RMS phase current Iso. Such a current
may be referred to as the line current, using the ratio 1 or 1/ 3 according
to the stator winding connection, star of delta connection, respectively.
Then the equivalent current in the q-th slot is given by

∑n
nq
Iq = k jqi oj (t *) q = 1, 2 , … , Q s /2 p (13.5)
j= a , b , c pp

where npp is the number of parallel paths of the stator winding.

The currents Iq (q = 1, …, Qs/2p) are forced within the slots and the field
problem is solved. Hence, the stator flux linkages are obtained by means of
the integral loop of the magnetic vector potential Az. Using the equivalent
coil sides, all of them with the same section S = Sq, the instantaneous value
of the j-th phase flux linkage during the no-load test is given by

Qs /2 p

∑k ∫
nq 1
Λ oj = 2 pL Fe jq A zdS j = a , b, c (13.6)
n pp q= 1
Sq Sq
Three-Phase Induction Motors 249

where LFe is the net axial length of the lamination stack. From the three-
phase flux linkages, the RMS value of the no-load flux linkage Λso is obtained.
The same stator flux linkage may be also obtained starting from the air-
gap flux density distribution. It is necessary to evaluate the radial component
of the flux density Br(ϑ,t*) = B(ϑ,t*)·ur along a circumference in the middle
of the air-gap. Using the Fourier series expansion, from the distribution
Br(ϑ,t*), the amplitude of the fundamental harmonic Br1 is computed. Then,
the average value of the fundamental harmonic distribution in a polar pitch
is (2/π)Br1. The corresponding average magnetic flux per pole is

2  πDL Fe DL Fe
Φ o1 =  B r 1  = B r1 (13.7)
π  2p p

and then the corresponding stator flux linkage (the peak value) becomes

k wN
Λ̂ so = Φ o1 (13.8)
p

where kw is the winding factor of the fundamental harmonic and N is the

number of series conductors per phase. The flux linkage is variable in a
sinusoidal way. Then the RMS value of the induced EMF, due to the only
fundamental component of the flux density, is given by

ˆ
Λ π
Eo = ωΛ so = ω so
= f k w NΦo1 (13.9)
2 2

At last, the nonlinear magnetizing inductance LM of Figure 13.4 is

Λ so
( )
L M Λ so =
Iso
(13.10)

It is assumed that the dependence of LM on the saturation is the same at

no-load and under load, that is, LM(Λs) = LM(Λso).

13.2.3 Locked Rotor Test

The locked rotor test is simulated. During the test, the rotor is locked in a
fixed position and three-phase sinusoidal currents are forced in the stator
windings. The dependence of the parameters Rr,2D and Lσ,2D of Figure 13.4
on the rotor frequency fr can be assessed by changing the source frequency.
The rotor frequency is the same as the source frequency f, because of the
locked rotor.
250 Electrical Machine Analysis Using Finite Elements

In the locked rotor test, the rated currents correspond to relatively low
source voltages, then to low magnetic fluxes. As a consequence, the magnetic
materials are assumed to be linear. The field problem is then linear and
characterized by field quantities varying sinusoidally, at the electrical fre-
quency ω = ωr = 2πfr . The magnetic vector potential has only the z-axis
component, as shown in Equation (13.2), that may be expressed by means
of the phasor notation of Equation (13.1), which is

=A
A ˆ e j(ωt + ϑ ) (13.11)
z z

with ω = ωr .
Thus, with constant magnetic permeability µ, the field problem is
described by the differential equation

1  ∂2 A

∂2 A
z
+ z

= J
sz − jω r σA (13.12)

µ  ∂x 2
∂y 2 
z

where σ is the electrical conductivity and ωr is the source electrical frequency.

The equivalent circuit corresponds to that reported in Figure 13.6. As
shown in Figure 13.6, from the field analysis the rotor 3D effects are not
taken into account. They will be analytically computed and added in the
equivalent circuit later on. In other words, the parameters computed by the
field analysis are only Rr,2D and Lσ,2D. Considering the problem to be linear,
the inductance LM assumes a constant value, which was computed previ-
ously during the no-load test. The analysis is focused on the rotor parameters.
From the locked rotor simulations, carried out at different frequencies fr , it
is possible to assess the functions Rr,2D(fr) and Lσ,2D(fr), which is to consider
the effects of a nonuniform distribution of the current in the rotor bars at
the different operating speeds.

Is Ir Lσ,2D
Rs Lσs,3D A
+ +

Vs jωΛs LM R r,2D
S

− −
A′
2D Simulation

FIGURE 13.6
Equivalent circuit corresponding to the locked rotor test.
Three-Phase Induction Motors 251

Also in this case, the field problem is solved referring to the equivalent coil
sides. The shaft may be disregarded in the field analysis. In fact, the currents
induced in the rotor bars shield the inner parts of the rotor significantly.
From the field solution, the Joule losses in the rotor bars are

∫
1
PJr = 2 pL J
z ⋅ J zdS (13.13)
2 σ Al S Al

where L is the total length of the rotor bars, σAl is the Aluminium conduc-
tivity, and SAl is the overall cross-section of the rotor bars in the simulated
pole piece. The factor 2p considers that one pole piece only is simulated.
The average value of the magnetic energy in the magnetic material is

Bˆ H
ˆ
∫
1
Wm = 2 pL Fe dS (13.14)
2 D 2 2

With the hypothesis of a rotating magnetic field, exactly sine-distributed,

both PJr and Wm are constant in time. The real and imaginary components
of the equivalent impedance at the terminals AA′ of the nonlinear circuit of
Figure 13.6, indicated by Req + jωrLeq in Figure 13.7, are obtained. Such
parameters represent the connection elements between the field solution and
their representation by means of the lumped-parameters circuit. They are
computed as

PJr 2 Wm
R eq = Leq = (13.15)
3I s 2 3I s 2

where Is is the RMS value of the current used in the test.

At this point, since LM is known from the no-load simulation and equal to
its unsaturated value, the resistance Rr,2D(fr) and the inductance Lσ,2D(fr) of
the circuit in Figure 13.6 can be computed.

Is Leq
A

jωΛs Req

A′ 2D Simulation

FIGURE 13.7
Electrical parameters corresponding to the locked rotor test.
252 Electrical Machine Analysis Using Finite Elements

13.2.4 Construction of the Equivalent Circuit

The parameter LM(Λs), obtained from the no-load test, together with Rr,2D(fr)
and Lσ,2D(fr), obtained from the locked rotor test, constitute the electrical
parameters within the dashed box of the nonlinear circuit of Figure 13.4. In
order to complete the circuit, the parameters corresponding to the stator end-
winding, the rotor bar skewing, and the rotor ring effects have to be included
in the circuit.
The leakage inductance Lσs,3D includes the stator end-winding leakage
inductance and the leakage inductance caused by skewing. An estimation
of the leakage inductance of the stator end-winding is

N2
L σs ,ew = µ o Lew λ ew (13.16)
2p

where Lew is the effective end-winding length, and λew is the specific per-
meance coefficient. Its value ranges between 0.35 and 0.55, according to the
winding type.
The leakage inductance caused by skewing is estimated by

(
L σs ,sk = L M 1 − k sk
2
) (13.17)

where the magnetizing inductance LM is given by Equation (13.10) and ksk

is given by

k sk =
(
sin ε sk 2 ) (13.18)
ε sk 2

and εsk is the skewing angle in electrical radians.

The short-circuited rings resistance is given by

( )
2
3 k wN
R r ,3D = ρAl (13.19)
2 πp2 S ring

Once the model of the induction motor is obtained, i.e., all the parameters
of the equivalent circuit of Figure 13.4 are found, it is possible to predict the
motor performance at different operating conditions. The motor character-
istics are computed, for instance the mechanical characteristic, the current-
speed curve, and so on. As explained earlier, the iron and mechanical losses,
needed for the computation of the efficiency, have to be considered separately.
Three-Phase Induction Motors 253

TABLE 13.1
Main Data of the Three-Phase
Induction Motor
Rated power 22000W
Rated voltage 280/485V
Line rated current 32.7 A
Rated frequency 60 Hz
Rated speed 1750 rpm
Number of poles 4
Stack length 300 mm
Air-gap length 0.5 mm
Rotor bar skewing 1 slot
Service S1
Protection IP55
Insulation Class F
Cooling IC 41

13.2.5 Example
The finite element method is applied to a commercial three-phase induction
motor with a double bar rotor. The geometry is shown in Figure 13.1. The
main data of the induction motor are reported in Table 13.1.
The number of stator slots is Qs = 36 so that the number of slots per pole
per phase is qs = 3 and the electrical slot angle is αc = 20 degrees. The stator
winding is a double-layer winding, with two coil sides per slot, each of them
formed by 14 copper conductors, thus nq = 28. The coil pitch is yq = 9,
corresponding to one slot chording. At last there are npp = 2 parallel paths.
A delta connection is adopted. The number of rotor aluminum bars is Qr = 28.

13.2.5.1 No-Load Simulation

In the no-load simulation, a large number of elements of the mesh is used
in the air-gap and in the iron regions, where the highest field gradients are
expected. With no or low saturation of the iron paths, it is observed that the
magnetic energy in the air-gap equals almost 80% of the total energy. The
magnetic material is a TERNI 2650.
Because of the delta connection, the phase current is Io = ILo/√3 where ILo
is the line no-load current. The RMS values are considered. The current
flowing in each coil, since there are npp = 2 parallel paths, is given by Ic =
Io/2 = ILo/(2 3 ).
The no-load test may be carried out in different ways, choosing a suitable
reference time t* and coherent boundary conditions. Even if a smaller part
of the motor may be simulated for the no-load analysis, the simulation is
carried out on one pole piece of the motor, so as to use the same domain for
the no-load and the locked rotor analysis with fixed winding distribution,
254 Electrical Machine Analysis Using Finite Elements

Dirichlet Dirichlet

+B −C −C +B −C −C
+B −C +B −C
−A +B +B +B −C −C +A −A +B +B +B −C −C +A
−A +B −C +A −A +B −C +A
−A +A −A +A

Neumann Neumann Dirichlet

Dirichlet
Ia
Ic Ia

Ic Ib
Ib

FIGURE 13.8
Two solutions of assignment of the stator currents and the boundary conditions for the no-load
test: solution (a) and solution (b).

TABLE 13.2
Three-Phase Currents at the Time t*
for the No-Load Simulation
No-load simulation
Phase Solution (a) Solution (b)

A + 2 Ic 0

B 2 3
− Ic + 2 Ic
2 2

C 2 3
− Ic − 2 Ic
2 2

–A − 2 Ic 0

–B 2 3
+ Ic − 2 Ic
2 2

–C 2 3
+ Ic + 2 Ic
2 2

the currents and the boundary conditions may be assigned as illustrated in

Figure 13.8. They correspond to those used in Chapter 8. Equivalent coils
are used: the corresponding current distribution is directly found once the
instant t* is fixed, as reported in Table 13.2, where Ic is the RMS current
flowing in each coil.
The flux lines and the air-gap flux density corresponding to the two no-
load tests are shown in Figure 13.9 and Figure 13.10. It is easy to observe,
in both simulations, the flux plot symmetry. This symmetry might be properly
Three-Phase Induction Motors 255

FIGURE 13.9
Flux plots in the two no-load simulations: referring to Table 13.2, solution (a) and solution (b).

Flux density (T) Flux density (T)

0 0.8
−0.1 0.6
−0.2 0.4
−0.3 0.2
−0.4
0
−0.5
−0.2
−0.6
−0.7 −0.4
−0.8 −0.6
−0.9 −0.8
0° 20° 40° 60° 80° 100° 120° 140° 160° 180° 0° 20° 40° 60° 80° 100° 120° 140° 160° 180°
Electrical angle (degrees) Electrical angle (degrees)

FIGURE 13.10
Air-gap flux density distributions in the two no-load simulations: referring to Table 13.2,
solution (a) and solution (b).

used to halve the analysis domain. Finally, the effect of the slot openings is
clear on the distribution of the air-gap flux density.
The magnetization curve is illustrated in Figure 13.11, showing the mag-
netizing current as a function of the source voltage. In order to stress the
effect of the air-gap length, the no-load simulation is carried out on induction
motor with different air-gaps.
At last, referring to the equivalent electric circuit of Figure 13.5, the mag-
netizing inductance LM is computed as a function of the stator flux linkage
Λs = Λso. It is shown in Figure 13.12.
It is worth noticing that using a current source allows the obtained result
to be applied to motors of different stack length, but with the same lamina-
tions and the same winding distribution. In fact, the result is extended to
motors with different stack length maintaining the same slot current. Of
course, the values of line current and voltage have to be arranged according
to the number of the actual conductors.
256 Electrical Machine Analysis Using Finite Elements

14
(c) (b) (a)
(a) g = 0.4 mm
12
(b) g = 0.5 mm
(c) g = 0.6 mm
10
Current (A)
8

2
100 150 200 250 300 350 400 450 500
Voltage (V)

FIGURE 13.11
Magnetizing current versus source voltage with different air-gap lengths.

0.16
0.15
Inductance (H)

0.14
0.13
0.12
0.11
0.1
0.09
0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
Flux linkage (Vs)

FIGURE 13.12
Magnetizing inductance LM versus stator flux linkage Λs.

13.2.5.2 Locked Rotor Simulation

In the locked rotor simulation, a large number of finite elements is chosen
in the region surrounding the air-gap and in the rotor bars. In fact, in the
air-gap there are high field gradients, and in the rotor bars there is a non-
uniform distribution of the current, caused by the operating frequency.
As depicted above, to the aim of freezing the value of the magnetizing
inductance, the magnetic material is described by a constant relative mag-
netic permeability. The latter is fixed equal to the initial permeability of the
used B-H curve, i.e., µr = 4000. The conductivity of the aluminum is fixed to
σAl = 2.549·107 S/m, corresponding to a temperature of 120˚C. An adequate
value of the rotor cage conductivity is crucial for a correct analysis of the
induction motor.
The induction motor is locked in a position. However, now the forced
currents are sinusoidal, and are described by means of the complex notation.
Let IL be a fixed RMS value of the line current, then the RMS value of the
Three-Phase Induction Motors 257

TABLE 13.3
Three-Phase Currents for the
Locked Rotor Simulation
Phase Locked Rotor Simulation

A + 2 Ic + j0

B 2 6
− Ic − j Ic
2 2

C 2 6
− Ic + j Ic
2 2

–a − 2 Ic − j0

–b 2 6
+ Ic + j Ic
2 2

–c 2 6
+ Ic − j Ic
2 2

current is each coil is given by Ic = IL/(2√3), because of npp = 2 parallel paths.

The three-phase currents are directly computed, as reported in Table 13.3.
As far as the boundary conditions are concerned, Dirichlet’s boundary
condition is assigned to the external circumference of the stator, forcing the
flux lines to remain confined within the stator yoke. Similarly the same
boundary condition is assigned between the shaft and the rotor lamination
(the shaft is excluded from the analysis domain). Finally, along the lateral
borders the periodic boundary conditions are assigned.
Figure 13.13 shows the flux plots relative to the locked rotor simulations.
Figure 13.13(a) corresponds to a frequency f = fr = 60 Hz, while Figure
13.13(b) corresponds to a frequency f = fr = 5 Hz. The shielding effect of the
rotor bars is as evident as the frequency increases.
From these simulations, the behavior of the resistance Rr,2D(fr) and the total
leakage inductance Lσ,2D(fr), of the equivalent circuit of Figure 13.4 are
obtained as a function of the rotor frequency fr . As explained above, the
latter corresponds to the supply frequency, since the rotor is locked. The two
behaviors are shown in Figure 13.14.

13.2.5.3 Motor Performance

Once the variable lumped-parameter equivalent circuit is built, it is possible
to assess the overall performance of the motor. In particular, the mechanical
characteristic is computed at nominal voltage and nominal frequency, as
reported in Figure 13.15 by the solid line.
258 Electrical Machine Analysis Using Finite Elements

FIGURE 13.13
Flux plots of the locked rotor simulations, with source frequency (a) f = fr = 60 Hz, and (b) f =
fr = 5 Hz.

0.9 1.6

0.8 1.4
Resistance (Ohm)

Inductance (H)

0.7 1.2

0.6 1

0.5 0.8

0.4 0.6

0.3 0.4
0 10 20 30 40 50 60 0 10 20 30 40 50 60
Frequency (Hz) Frequency (Hz)

FIGURE 13.14
Variation of the resistance Rr,2D and of the total leakage inductance Lσ,2D with the rotor frequency fr .

In the same Figure 13.15, some experimental results are reported, by circles.
Comparing simulated and measured data, the agreement is satisfactory,
during operations both at standstill and under load.
Analogously, Figure 13.16 shows the stator current as a function of the
motor speed, while Figure 13.17 shows the power factor as a function of the
phase current.

13.3 Motor Analysis Using Simulations Under Load

The second approach, alternative to that proposed in Section 13.2, is based
on the simulations of the indirect tests. The motor operations under load are
simulated directly. The aim of such an analysis approach is to assess each
Three-Phase Induction Motors 259

400
350

Mechanical torque (Nm) 300

250

200

150 Simulations of indirect test

100 Simulations of direct test
Experimental results
50
0
0 200 400 600 800 1000 1200 1400 1600 1800
Speed (rpm)

FIGURE 13.15
Torque-speed characteristic.

Current (A)
160

140

120

100

80

60 Simulations of indirect test

40 Simulations of direct test
Experimental results
20
0
0 200 400 600 800 1000 1200 1400 1600 1800
Speed (rpm)

FIGURE 13.16
Stator current versus speed.

operating point, considering a sinusoidal voltage source. The motor charac-

teristics are obtained by a set of simulations of the motor.

13.3.1 Magnetic Field Equations Under Load

Since a generic operating point is investigated, the simulations have to
consider both the nonlinearity of the magnetic material and the nonuniform
distribution of the current in the rotor bars, as well as the rotor movement.
The field problem is nonlinear and time-variable, and requires that we
compute an induced current due to motional effects. The current density
may be expressed by means of the constitutive law in σ given in Equation
260 Electrical Machine Analysis Using Finite Elements

Power factor
0.9

0.8

0.7

0.6

0.5

0.4

0.3 Simulations of indirect test

Simulations of direct test
0.2
Experimental results
0.1
0 20 40 60 80 100 120 140 160
Current (A)

FIGURE 13.17
Power factor versus stator current.

(1.87), as a function of the specific electric force. Substituting for Et in Equa-

tion (1.87) the expressions of the induced and motional specific electric forces,
Ei of Equation (1.81) and Em of Equation (1.85), respectively, and considering
the forced current density, that can be expressed as Js = σ(Ec + Ene), it results
that

∂A
J = Js − σ + σv × B (13.20)
∂t

Because of the 2D problem, with only the z-axis components of the current
density vector J and the magnetic vector potential A, as indicated in Equation
(13.2), the field problem is described by the following differential equation:

 ∂  1 ∂A
∂  1 ∂A
 ∂A

z
+ z
= J
sz − σ z
)
+ σ v × curl(0, 0, A (13.21)
  
∂x  µ ∂x  ∂y  µ ∂y  ∂t
z

In the latter, please note that

∂A

J
i = −σ z
∂t

corresponds to the induced current density, while

)
J
m = σ v × curl(0, 0, A z

corresponds to the motional current density.

Three-Phase Induction Motors 261

13.3.2 Alternative Formulation

The motional term, which appears essentially in the rotor, does not define a
vector field. However, it depends on the particular reference system that is
adopted. It is useful to change the reference system, so as to convert the
motional term in a corresponding induced term.
Let us remember that, as in Equation (13.1), the z-axis component of the
current density source is a sinusoidal function of the time, according to the
relation

J
sz = ˆJ sze jωt (13.22)

In the same way, the z-axis component of the magnetic vector potential is

=A
A ˆ e j(ωt − ϑ ) (13.23)
z z

The time derivative of the magnetic vector potential becomes

∂A

z ˆ e j(ωt − ϑ ) = jωA
= jωA
(13.24)
∂t
z z

From the first Maxwell’s equation (1.102), in which the displacement cur-
rent density is negligible because of the low frequency, it is

curlH = J (13.25)

Substituting Equation (13.20) into Equation (13.25), it is

∂A
curlH + σ − σv × B = Js (13.26)
∂t

Using the complex notation, and in particular Equation (13.24), it becomes

curlH + jωσA − σv × B = Js (13.27)

In the cylindrical coordinate system of Figure 13.18, defined by the coor-

dinate unity vectors (ur , uϑ, uz), the speed v of the generic point P of the
rotor is given by

v = rω ru ϑ (13.28)

where ωr is the rotor speed.

262 Electrical Machine Analysis Using Finite Elements

z
P

uz uϑ

ϑ
ur
r

FIGURE 13.18
Cylindrical coordinate system.

Consequently, since B = curlA, the motional specific electric force Em can

be expressed as

(
v × B = rω ru ϑ × curlA ⋅ u r ⋅ u r ) (13.29)

bearing in mind that the inner product ur · ur = 1.

Taking into account that the magnetic vector potential A has only the z-axis
component, the curl of A in cylindrical coordinates is

1 ∂A ∂A

curlA = z
ur − z
uϑ (13.30)
r ∂ϑ ∂r

Substituting Equation (13.30) into Equation (13.29), results in

 1 ∂A
∂A
 
v × B = rω ru ϑ ×  z
ur − z
uϑ  ⋅ ur  ⋅ ur (13.31)
 r ∂ϑ ∂r  

Since ur ·ur = 1 and uϑ × ur = –uz, Equation (13.31) is simplified as

∂A

v × B = −ω r z
uz (13.32)
∂ϑ

Now, the z-axis component of the magnetic vector potential A is assumed

to be a sinusoidal function of the time and of the rotor coordinate, according
to Equation (13.13), where ϑ is the azimuthal coordinate of Figure 13.18.
Equation (13.13) can be interpreted as the description of a magnetic field
sine-distributed along the air-gap and rotating at speed ω.
Substituting Equation (13.13) into Equation (13.27) results in

curlH + jωσA − jω r σA = Js (13.33)

Three-Phase Induction Motors 263

ω − ωr
Finally, introducing the equivalent conductivity σ eq = σ , it is possi-
ble to write ω

curlH + jωσ eqA = Js (13.34)

Expressing Equation (13.34) by means of the magnetic vector potential A

only, it results that

 ∂  1 ∂A
∂  1 ∂A


z
 + z
∂x  µ ∂x  ∂y  µ ∂y 
 ( )
= J
sz − j ωσ
eq


A z (13.35)

In conclusion, the magneto-dynamic complex field problem expressed by

Equation (13.21) is transformed in a magnetostatic complex field problem
with induced currents: it would be just as simple to assign a suitable value
to the term (ωσ)eq at each operating condition.
The term (ωσ)eq corresponds to (sωσ), where s is the rotor slip. It can be
obtained changing the conductivity of the rotor bars (assuming an equivalent
conductivity σeq = sσ) or changing the source frequency (imposing in the
simulation an equivalent frequency feq = sf), in any case depending on the
particular operating condition of the motor (distinguished by the slip s).
In the first case, the analysis frequency is equated to the stator frequency
f and the conductivity σeq = sσ is assigned to the rotor bars. In the second
case, the analysis frequency is equated to the rotor frequency feq = sf, keeping
constant the conductivity of the materials. In other words, with the first
approach the rotor is fixed in the stator reference frame, while with the
second approach the stator is considered rotating, at the speed of the rotor
reference frame.

13.3.3 Nonlinearity of the Magnetic Materials

The nonlinearity of the magnetic materials greatly affects the field solution.
Since the field quantities are time-variable quantities, the problem (13.37) is
sometimes solved using a step-to-step method. However, this method is very
time consuming.
As an alternative, it is possible to take advantage of the fact that the
variation of the field quantities is sinusoidal. To do that, the B-H curve is
suitably adjusted so as to consider the variation of the permeability µ in a
period. Starting from the dc magnetizing characteristic Bm(Hm) of the mate-
rial, an equivalent curve Beq(H) is computed, corresponding to a magnetic
field H = Hmsin(ωt), which is sinusoidal with the time. This is achieved
imposing the average value in a time period T of the magnetic energy density
computed on the dc B-H curve to be equal to the magnetic energy density
computed on the equivalent curve Beq(Hm), as reported in Figure 13.19.
264 Electrical Machine Analysis Using Finite Elements

B
Beq (Hm)

Bm (Hm )

FIGURE 13.19
Equivalent B-H curve assuming a sinusoidal waveform of the magnetic field strength.

The corresponding magnetic energy densities are

Beq T B(t)

∫ ∫∫
1 1
H m dBeq = H dB dt (13.36)
2 0 T 0 0

Applying Equation (13.36) with different values of Hm the curve Beq(Hm)

is obtained, which is the curve µeq(B). The latter curve allows the time
harmonics to be modeled.

13.3.4 Link to an External Circuit

The field problem described by Equation (13.35) has to be linked to an
external circuit to study the exact operating point.
In other words, the field solution should consider at the same time the
quantities of the field solution, corresponding to the central block of Figure
13.2, and the lumped-parameters of the external circuit, i.e., the elements
external to the central block of Figure 13.2. It is necessary to fix a correspon-
dence between the different region of the domain D and the lumped-param-
eters of the external circuit, which are not considered in the finite element
analysis. Among them the 3D effects (end-windings and skewing) and the
stator winding resistance.
It is possible to introduce directly in the matrix describing the field problem
some integral conditions, which would otherwise be applicable only after
the field solution is achieved.
The use of an external circuit allows the simulation of a voltage-fed induc-
tion motor to be carried out, describing the field problem by means of the
differential equation

 ∂  1 ∂A
∂  1 ∂A


∂V

z
 + 
z
∂x  µ ∂x  ∂y  µ ∂y  =σ
∂z
− j ωσ ( ) eq


A z (13.37)
Three-Phase Induction Motors 265

13.3.5 Computation of the Mechanical Torque

Contrary to the previous approach in which the torque is assessed by the
equivalent circuit, in this case the torque is obtained directly from the field
solution. Either Maxwell’s stress tensor or the virtual work principle may
be used. Maxwell’s stress tensor requires an integration on all the finite
elements located between the fixed part (the stator) and the moving part
(the rotor); the virtual work principle is based on a virtual angular displace-
ment dϑ of the rotor.

13.3.6 Example
The second analysis approach is applied on the same induction motor of the
previous example. The simulation is carried out forcing three symmetric
sinusoidal voltages with the RMS value Vs = 385 V at the motor terminals,
which is in Figure 13.2. The various electrical lumped-parameters are added,
to consider the 3D effects, and the magnetization characteristic is set up.
Each simulation is carried out at a fixed rotor speed.
To have an easy comparison with the previous approach, the results are
reported in Figure 13.14 to Figure 13.16, using stars. It is observed that the
results obtained by means of this approach agree with the measured one.
The highest error is observed on the mechanical characteristic comparing
the maximum torque. The value of the maximum torque is the same with
the two analysis approaches, while the speed at which this value is reached
is different.

References
1. T. Trnhuvud, K. Reichert, “Accuracy problems of force and torque calculation
in FE-systems,” IEEE Trans. Industry Appl., vol. 24, n. 1, pp. 443–446, 1988.
2. G.R. Slemon, “Modelling of induction machines for electric drives,” IEEE Trans.
on Industry Applications, vol. 25, n. 6, pp. 1126–1131, 1989.
3. E. Vassent, J.C. Sabonnadiere, “Simulation of induction machine operating
using complex magnetodynamic finite elements,” IEEE Trans. on Magnetics,
vol. 25, n. 4, pp. 3064–3066, 1989.
4. R. Belmans, D. Verdyck, T.B. Johansson, W. Geysen, R.D. Findlay, “Calculation
of the no-load and torque speed characteristic of induction motors using finite
elements,” Int. Conf. on Electrical Machines, Boston, 1990, pp. 2724–2729.
5. S. Williamson, M.J. Robinson, “Calculation of induction motor equivalent cir-
cuit parameters using finite elements,” IEE Proc., Pt. B, vol. 138, n. 5, pp.
264–276, 1991.
6. P.L. Alger, Induction Machines. Their Behavior and Uses, 2nd ed., Gordon and
Breach, Basel, 1995.
7. S. Williamson, D.R. Gersh, “Finite element calculation of double-cage rotor
equivalent circuit parameters,” IEEE Trans. on Energy Conversion, vol. 11, n. 1,
pp. 41–48, 1996.
266 Electrical Machine Analysis Using Finite Elements

8. N. Bianchi, S. Bolognani, “Design Procedure of a Vector-Controlled Induction

Motor for Flux-Weakening Applications,” in IEEE Industrial Applications Con-
ference Records, IAS ’97, New Orleans, Lousiana, 4–9 October 1997, vol. 1, pp.
104–111.
9. S.L. Ho and W.N. Fu, “Review and Future Application of Finite Eelment Meth-
ods in Induction Motors,” in Electrical Machine and Power System, vol. 26, pp.
111–125, 1998.
10. N. Bianchi, S. Bolognani, G. Comelato, “Finite Element Analysis of Three-Phase
Induction Motors: Comparison of Two Different Approaches,” IEEE Trans. on
Energy Conversion, vol. 14, no. 4, December 1999, pp. 1523–1528.
APPENDIX
Material Data

This appendix reports some characteristic data of conductive material, insu-

lating materials and magnetic materials, useful for simulations.

Main Data of the More Common Conductive Materials

Specific Specific Thermal Expansion Breaking
Resistivity Weight Heat Conduct. Coeff. Load
10–8 Ωm α ˚C–1 kg/m3 J/kg ˚C W/m ˚C 10–6 ˚C–1 kg/mm2
Copper 1.72 0.0043 8 900 383 389 16.5 21–24
Aluminum 2.78 0.0042 2 700 904 229 23.8 7–10
Iron 9.8 0.0065 7 800 452 66 11.7 30–45
Nickel 7.3 0.0065 8 900 444 88 13.3 45–50
Platinum 10.5 0.0039 21 400 134 71 8.9 34
Gold 2.4 0.0038 19 300 126 314 14.2 25
Silver 1.6 0.0040 10 500 234 420 19.3 30
Tungsten 5.5 0.0046 19 300 138 130 4.4 60–100
Molibden 4.8 0.0046 10 200 264 151 5.1 80
Tin 12.0 0.0044 7 300 226 65 23.0 1.6–3.8
Lead 21.0 0.0040 11 400 128 35 28.7 1.6
Zinc 5.9 0.0037 7 100 385 110 30.7 7–12
Aldrey 3.2 0.0036 2 700 920 188 23.0 30–35

267
268 Electrical Machine Analysis Using Finite Elements

Main Data of the More Common Insulating Materials

Breaking Electric Relative Maximum Operating
Material Field (kV/mm) Permittivity Temperature (˚C)
Asbestos 2.5–3.5 — 200
Air 2–3 1 —
Distillate water 5–10 80 —
Asphalt 12–16 2.5 40
Bachelite 10–18 5–9 120
Paper 6–11 1.6–5 85
Paper bachelised 5–15 5 95
Wax paper 40–50 2.5–4 85
Presspan cardboard 8–10 2.5–4 90
Cellonite 8–10 4–5 95
Ebonyte 5–25 1.4 100
Fiber 2–10 2.1–2.3 100
Vulcanized rubber 8–20 2.5–3 60
Rubber (varnish) 10–35 1 90
Impregnated wood 8–30 3–3.5 90
Mica 60–160 5–5.5 700–800
Micalex 13–15 7–8 600
Micanite 20–40 3–3.5 90–120
Flax oil 8–18 3.5 —
Mineral oil 10–16 2–2.5 100
Apiroil 20–25 4.5–5 —
Solid paraffin 14–45 2 45–50
Porcelain 20–40 4.5–6 200
Steatite 14–16 5.6–6.5 600
Bachelised cloth 10–20 4.5–6 95
Sterlingata cloth 25–50 3.5–5.5 95
Glass fiber 45–50 4.5 150
Glass Pirex 30–150 5 400
Note: The maximum electric field varies with the material quality. In addition, it
decreases when the dimensions of the sample are used in the test.
Material Data 269

Magnetic Behaviors of Some Massive Magnetic Materials

Flux Forged Steel Gray Cast Iron
Density and Melted Steel and Tempered Steel Iron Steel Nickel
B (T) H (A/m) µr H (A/m) µr H (A/m) µr H (A/m) µr
0.1 70 1140 200 400 25 3200 215 372
0.2 90 1780 450 355 50 3200 420 380
0.3 100 2400 800 300 85 2820 1020 235
0.4 120 2660 1300 246 115 2780 2660 120
0.5 140 2860 2000 200 160 2500 8000 50
0.6 170 2820 2800 171 200 2400 24000 20
0.7 220 2500 4000 140 250 2240 88000 6.4
0.8 270 2370 5500 117 310 2060
0.9 320 2260 8000 90 380 1900
1.0 400 2000 11000 73 490 1630
1.1 500 1750 15000 58 640 1380
1.2 620 1550 20000 48 850 1130
1.3 850 1230 — — 1480 705
1.4 1200 930 — — 2800 400
1.5 2000 600 — — 5000 240
1.6 3500 365 — — 9300 138
1.7 6000 226 — — 17000 80
1.8 10000 144 — — 29000 50
1.9 16000 95
2.0 25000 64

Magnetic Behaviors of Some Magnetic Laminations

Normal Iron Silicon Iron Silicon Iron Grain-Oriented
Thickness Thickness Thickness Thickness
Flux 0,5 mm 0,5 mm 0,35 mm 0,35 mm
Density cp = 3,6 W/kg cp = 2,5 W/kg cp = 1,1 W/kg cp = 0,6 W/kg
B (T) H (A/m) µr H (A/m) µr H (A/m) µr H (A/m) µr
0.8 230 2770 120 5305 140 4545 40 15920
0.9 330 2170 160 4475 200 3580 80 8950
1.0 470 1695 240 3315 290 2745 140 5685
1.1 630 1390 340 2575 430 2035 230 3805
1.2 800 1195 510 1870 680 1405 320 2985
1.3 1050 985 780 1325 1150 900 460 2250
1.4 1350 825 1320 845 1960 568 630 1770
1.5 1750 680 2360 505 3240 370 840 1420
1.6 3500 365 3720 340 5430 235 1110 1145
1.7 6200 220 5670 240 7160 190 1560 865
1.8 9000 160 8450 170 9000 160 2060 695
270 Electrical Machine Analysis Using Finite Elements

2.0
1.9
1.8
1.7
1.6
Induzione tesla

1.5
1.4
1.3
1.2 TERNI-ARMCO M5T30
1.1 Curva di magnetizzazione inc.c.
D-C magnetization curve
1.0 Prova dopo ricottura di distensione
0.9 Test: SRA; parallel; A 341

0.8
0.7
0.6
0.5
1 2 3 5 7 10 2 3 5 7 102 2 3 5 7 103 2 3 5 7 104
Forza di magnetizzazione- Asp/m

FIGURE A.1
Grain-oriented lamination.

2.0 8
10−3

1.8 7
10−3

1.6 6
10−3
Permeabilita’ µ-H/m

1.4 5
10−3
Induzione tesla

TERNI-ARMCO DI MAX 2350-2365 CR

1.2 Curva di magnetizzazione e della permeabilita in c.c. 4
10−3
Rilevata su provino come cesolato 50/50
D.C. magnetizzation and permeability curve
1.0
Test: AS; 50/50; A 341
3
10−3

0.8 2
10−3

0.6 1
10−3

0.4 0
10 2 3 5 7 102 2 3 5 7 103 2 3 5 7 104 2 3 5 7 105
Forza magnetica-Asp/m

FIGURE A.2
Silicon iron laminations for rotating machines.
Material Data 271

Main Data of the More Common Permanent Magnets

Thermal Thermal Specific
Residual Coercive Factor for Factor for Temperature Weight
Br (T) Hc (kA/m) Br (%/˚C) Hc (%/˚C) ϑmax (˚C) (kg/m3)
Ferrite 0.40 300 –0.2 0.4 200 4800
AlNiCo5 1.30 60 –0.02 0.02 250 7300
AlNiCo8 1.15 110 –0.02 0.02 250 7300
MnAlC 0.57 180 –0.12 0.25 300 5100
SmCo5 0.90 700 –0.045 –0.250 250 8200
Sm2Co17 1.00 600 –0.035 –0.200 300 8300
NdFeB 1.20 900 –0.12 –0.70 150 7400
Note: The data refer to samples, obtained by sinterised magnetic material. Ferrite, SmCo, and
NdFeB could be found bonded in plastic material. In this case, the values of Br , Hc, and
specific weight are practically halved. The maximum operating temperature ϑmax is
determined by the plastic material.

B (T)

1.3
3
1.1
1 Ferrite di Bario
2 AINiCo 5 2 0.9
3 AINiCo 8
0.7
4 MnAIC 7
5 SmCo5 4 0.5
5
6 Sm2Co17
1 0.3
7 NdFeB
6
0.1
H (kA/m)
900 700 500 300 100

FIGURE A.3
B-H curves of some permanent magnet materials.
272 Electrical Machine Analysis Using Finite Elements

1.5 2 2.5
500
B·H
B
kJ/m330 mT
−µ0 · H 20 10 5

1 400

KOEROX 300

300
KOEROX 330

KOEROX 360
0.5 200 B
KOEROX 300 K
KOEROX 380

KOEROX 330 K
100
KOEROX 150

KOEROX 100
0
−300 −250 −200 −150 −100 −50 kA/m 0
H

FIGURE A.4
Ferrite permanent magnet.

B-H B (kG) (mT)

0.6 0.7 0.8 1.0 1.5 2.0 3.0 5.0
14 1400
0.5
12 1200
24°C
0.4 50°C
1000
10
75°C 100°C
0.3 8 800

6 600
0.2
4 400
0.1
2 200

0
KOe −26 −24 −22 −20 −18 −16 −14 −12 −10 −8 −6 −4 −2 −0
2000 1800 1600 1400 1200 1000 800 600 400 200 0
H (kA/m)

FIGURE A.5
Neodimium–iron–boron permanent magnet.