non-overlapping windings
Johannes Ja obus Germishuizen
Dissertation approved for the degree of Do tor of Philosophy
in Ele tri al Engineering at Stellenbos h University
Promoters:
Prof. M.J. Kamper
Ele tri aland Ele troni Engineering
University of Stellenbos h
South Afri a
Dr. A. Jö kel
Te hnology and Innovation
Loher GmbH
By submitting this dissertation ele troni ally, I de lare that the entirety of the work
ontained thereinis my own, originalwork, that I amthe owner ofthe opyrightthereof
(unless to the extent expli itly otherwise stated) and that I have not previously in its
entirety or inpart submitted it for obtainingany quali ation.
Date:02.03.2009
Copyright ©2009 Stellenbos hUniversity
Analysis of interior permanent magnet motors with
non-overlapping windings
J.J.Germishuizen
Department of Ele tri al and Ele troni Engineering
University of Stellenbos h
Private Bag X1, Matieland,7602, South Afri a
PhD Dissertation
Mar h 2009
At present most of all existing variable speed drive systems are still based on indu tion
ma hines. In order to redu e energy loss, investment and maintenan e osts there is a
trendtorepla ethe indu tionmotorand gearboxwitha dire tdrivesystem. However, it
wasonly untilre ently that the progress in materials ien emade it possible to
e onom-i ally implementma hines with permanent magnet te hnology in a dire t drive system.
Interior permanent magnets motors with single layer non-overlapping windings have
ad-vantages,whi hmakeitattra tiveforthe useindire tdrives. Inthe presentdissertation
afast and a urate analysis methodology, suitablein an everyday designenvironment is
oered. The method makes use of both the analyti al and nite element method. Part
ofthe designpro edure requires asystemati algorithmtoallo atethe statorslots tothe
windingphase belts.
The a ura y of the proposed method was veried by means of a prototype tra tion
ma hine. Fromtheresultsitis on ludedthatthreetwo-dimensionalfun tions anbeused
to a urately al ulate the performan e of permanent magnet ma hines in an everyday
designenvironment. The analysis methodwhi h wasdeveloped isbased on real physi al
two-dimensional ma hine models. The unique ontributions of the present dissertation
are the performan e al ulation method and anexpression that denes a windingphase
Analise en ontwerp van elektiese motors met permanente
binnemagnete en nie-oorvleuelende wikkelings
J.J.Germishuizen
Departement van Elektrieseen Elektroniese Ingenieurswese
Universiteit van Stellenbos h
Privaatsak X1, Matieland,7602, Suid-Afrika
PhD Proefskrif
Maart 2009
Die meeste veranderlike spoed aandryfstelsels is tans gebaseer opinduksiemasjiene. Om
energie verliese, beleggings en onderhoudskoste te verminder, is daar 'n tendens om die
induksiemasjien en ratkas te vervang met n direkte aandryfstelsel. Dit is eers onlangs
dat vooruitgangin die materiaalwetenskappeditmoontlikgemaak het ommasjienemet
permanent magneet tegnologie ekonomies in n direkte aandryfstelsel te implementeer.
Elektiese motors met permanente binnemagnete en nie-oorvleuelende wikkelings het
vo-ordele wat ditbaie aantreklik maak vir die gebruik in direkte aandryfstelsels. In hierdie
proefskrif word nvinnige en akkurate analise metode geskik vir n daaglikse
ontwerps-omgewing voorgestel. Die metode maak gebruik van beide die analitise en die eindige
element metode. n Deel van die prosedure benodig n sistematiese algoritme om die
statorgleuwe aan die wikkeling-fase-gordelstoe teken.
Die akkuraatheid vandie metode word geverieer deur middelvannprototipe
trak-sie motor. Uit die resultate word die gevolgtrekking gemaak dat drie twee-dimensionele
funksies gebruik kan word om in ndaaglikse ontwerpsomgewing die werkverrigting van
permanentmagneet elektriesemasjiene akkuraat te kan bereken. Die analisemetode wat
ontwikkel is, is gebasseer op die werklike siese twee-dimensionele masjien model. Die
unieke bydrae van hierdie proefskrif is die werkverrigtingsberekengsmetode en 'n
I would like to thank my promoters, Prof. Maarten Kamper and Dr. Andreas Jö kel,
for their onstant support and guidan e throughout this proje t. I also a knowledge the
following persons and foundationfor their ontributions:
TheexternalexaminersProfessorsBerndPoni k(UniversityofHanover,Germany),
Konrad Rei hert (Swiss Federal Institute of Te hnology Züri h, Switzerland) and
Frikkie van der Merwe (Stellenbos h University, South Afri a) for their valuable
omments and advi e;
Prof. Heinri hvander Mes ht for ontinuous support with the dissertation layout;
My olleagues at Siemens AG (Vogelweiherstraÿe, Nürnberg), parti ularly Peter
E kert and ThomasS hmidt for their assistan e;
Prof.HansOttoSeins h(University ofHanover)and Prof.Zdeneke°ovský
(Te h-ni al University of Prague)who kindly provided me with detailson VilémKlíma;
IvanKlímaforproof-readingmyessayonhisfather,VilémKlíma,andforproviding
the photosof his father;
Willie Coetzeefor his un easing interest inmy work and
De laration ii Abstra t iii Uittreksel iv A knowledgements v Dedi ations vi Contents vii
List of Figures xii
List of Tables xv
Nomen lature xvi
1 Introdu tion 1
1.1 Ba kground tothe study . . . 1
1.2 Problem statement . . . 3
1.3 Overlapping and non-overlapping windings . . . 3
1.4 Approa h to the problem . . . 5
1.5 S ienti ontributions of this dissertation . . . 6
1.6 Delimitationsof the study . . . 6
1.7 Layout of the dissertation . . . 7
1.8 Di ulties en ountered during the study . . . 7
1.9 Notes tothe reader . . . 8
2 Literature overview 9 2.1 A histori alview of windingdesign . . . 9
2.2 Tenden ies inma hine analysis . . . 11
2.2.1 Introdu tory remarks . . . 11
2.2.2 The nite elementmethodas adesign tool . . . 12
2.2.3 The inuen e of saturation ontransient solutions . . . 12
2.2.4 Present approa hes iniron loss al ulation . . . 13
2.3 A short ex ursion . . . 14
2.3.1 Coming a ross avery interesting sour e . . . 14
3 Design and analysis of stator windings 17
3.1 Denition of the working harmoni . . . 17
3.2 Classi ation of symmetri alwindings . . . 17
3.2.1 Slots and oils perpoleand phase . . . 18
3.2.2 Average oilpit h . . . 18
3.2.3 Classi ation s heme . . . 19
3.3 Chara teristi s of symmetri alwindings . . . 20
3.3.1 Basi winding . . . 20
3.3.2 Winding symmetry . . . 22
3.3.3 Redu ed number of polepairs . . . 22
3.4 Rotating mmf . . . 23
3.4.1 The mmf of a single turn oil . . . 23
3.4.2 The mmf of three single turn oils . . . 24
3.4.3 Denition of the mmf envelope fun tions . . . 27
3.4.4 Phase beltdenition . . . 27
3.4.5 Higher order harmoni s. . . 27
3.5 Matrix representation of a winding . . . 28
3.5.1 Slot ve tor . . . 28
3.5.2 Phase belt onstraint . . . 29
3.5.3 Algorithm ow hart . . . 30
3.5.4 Matrix element assignment. . . 30
3.6 Properties of the windingmatrix . . . 32
3.6.1 Winding fa tor . . . 32
3.6.2 Current sheet anti-node axis . . . 33
3.6.3 Magneti axis . . . 33
3.6.4 Slot mmf. . . 34
3.7 Examples . . . 34
3.7.1 Winding fa tor table . . . 34
3.7.2 Slot mmf and urrent sheet . . . 36
3.7.2.1 Single layernon-overlapping . . . 36
3.7.2.2 Double layernon-overlapping . . . 37
3.7.2.3 Single layeroverlapping . . . 37
3.7.2.4 Double layeroverlapping . . . 41
3.8 Determination of the winding axes . . . 41
3.9 Summary . . . 41
4 Nonlinear magneti ir uit analysis 44 4.1 Introdu tory remarks . . . 44
4.2 Material properties . . . 45
4.2.1 Laminated steel . . . 45
4.2.1.1 Epstein framemeasurements . . . 45
4.2.1.2 Extrapolation of the B(H) Curve . . . 46
4.2.1.3 Three-term ore lossmodel . . . 48
4.2.1.4 Two-term ore loss model . . . 50
4.2.1.5 A ounting for the sta king fa tor in 2D FEA . . . 50
4.2.2 Permanentmagnets . . . 51
4.3 Nonlinear eld solutions . . . 52
4.3.1.1 Ampère'slaw . . . 53
4.3.1.2 Faraday's law . . . 53
4.3.2 Boundary setupand sour es . . . 53
4.3.2.1 Diri hlet boundary onditions . . . 53
4.3.2.2 Periodi boundary onditions . . . 54
4.3.2.3 Current sour es . . . 54
4.3.2.4 Voltage sour es . . . 55
4.3.3 Flux linkage al ulation . . . 56
4.3.4 Indu tan e al ulation . . . 56
4.3.4.1 Denitions for energy al ulation . . . 57
4.3.4.2 Indu tan e matrix . . . 57
4.4 Determination of the winding axis: se ond method . . . 59
4.5 Harmoni analysis. . . 60
4.5.1 Denition of the dis rete Fourier transform . . . 60
4.5.2 Time harmoni s . . . 61
4.5.3 Spatial harmoni s . . . 61
4.6 Example of a spatialharmoni analysis . . . 63
4.6.1 Air gap ux density . . . 63
4.6.1.1 Two-dimensionalairgap uxdensity fun tions . . . 63
4.6.1.2 Harmoni omponents . . . 65
4.6.2 Flux perpole . . . 68
4.7 Summary . . . 68
5 Tra tion ma hine ase study 69 5.1 Introdu tory remarks . . . 69
5.2 Torque versus speed hara teristi . . . 70
5.3 Con eptional design. . . 71
5.3.1 Winding type . . . 71
5.3.1.1 Form-wound versus random-wound oils . . . 71
5.3.1.2 Single versus double layernon-overlapping windings . . . . 73
5.3.2 Sizing equations . . . 73
5.3.3 Numberof stator slots . . . 74
5.3.4 Ele tromagneti design . . . 74
5.3.5 Winding properties . . . 75
5.3.5.1 Ee tivenumberof turns per slot . . . 75
5.3.5.2 Series numberof turns . . . 75
5.3.5.3 End windingleakage indu tan e. . . 76
5.3.5.4 Stator d. . windingresistan e . . . 76
5.3.5.5 Winding layout . . . 77
5.3.6 Geometri al dimensioning . . . 77
5.3.6.1 No-load ux linkage versus tooth width . . . 78
5.3.6.2 Torque pulsation . . . 78
5.3.6.3 Rotor design . . . 79
5.3.6.4 Staggered rotor design . . . 81
5.3.6.5 Stator parameters . . . 82
5.4 Loss al ulation . . . 82
5.4.1 Lo aleddy urrent loss . . . 82
5.4.1.2 Single layerwindings . . . 84
5.4.1.3 Round wire . . . 84
5.4.2 Cir ulating urrent loss. . . 86
5.4.3 Stator windinglosses . . . 86
5.4.4 Iron loss . . . 86
5.4.5 Eddy urrents inthe permanentmagnets . . . 87
5.5 Two-dimensional ma hine hara teristi fun tions . . . 88
5.5.1 Park transformation . . . 89
5.5.2 General motorvoltage and torque equations . . . 91
5.5.3 Equivalent ir uit . . . 92
5.5.4 Solution domain. . . 93
5.5.5 Torque versus frequen y hara teristi . . . 93
5.5.6 S aling of the solutiondomain . . . 96
5.6 Realisationof a prototype . . . 97
5.6.1 Manufa turedmotor . . . 97
5.6.2 Stator d. . resistan e measurement . . . 98
5.6.3 No-load measurements . . . 98
5.6.3.1 Open terminaltest . . . 98
5.6.3.2 Magnetising urrent . . . 100
5.6.4 Short ir uit test . . . 101
5.6.5 Removed rotor test . . . 101
5.6.6 Torque speed hara teristi . . . 103
5.6.7 Temperature rise test . . . 103
5.6.7.1 Measured results . . . 106
5.6.7.2 Adiabati rotor loss . . . 106
5.6.7.3 TransientFEA . . . 108
5.6.7.4 Loss al ulations . . . 108
5.7 Summary . . . 110
6 Con lusions and re ommendations 111 6.1 Introdu tory remarks . . . 111
6.2 Answering the resear h sub-questions . . . 111
6.2.1 What isthe mathemati alexpression to allo ate the stator slots? . 111 6.2.2 How an a windinglayout be represented ina ompa t form? . . . 112
6.3 Answering the main resear h question . . . 112
6.4 Re ommendations for further resear h . . . 113
Appendi es 114 A Klíma's losed expression 115 A.1 Introdu tory remarks . . . 115
A.2 Distribution fa tor . . . 115
B Slot ombinations 116 C Revision of the Air Gap Element 119 C.1 Introdu tory remarks . . . 119
C.2.1 AGE stiness matrix . . . 121
C.2.2 Time-savings heme. . . 121
C.2.2.1 Simplied Fourier oe ients . . . 122
C.2.2.2 Stepped AGE . . . 123
C.2.3 Periodi ity onditions . . . 123
C.2.4 Ele tromagneti torque . . . 124
C.2.5 Air gap ux density . . . 125
C.3 Cambridgesoftware . . . 125
C.3.1 History of the Cambridgesoftware . . . 126
C.3.2 Alternative formof the time-savings heme . . . 126
1.1 Double layeroverlapping and single layernon-overlapping windings . . . 4
1.2 Double and single layernon-overlapping windings . . . 5
2.1 Leakage ompensationinthe stator teeth . . . 11
2.2 VilémKlíma, 10.04.1906-06.10.1985 . . . 16
3.1 Classi ationof symmetri al
m
-phase windings . . . 213.2 Spatialmmf distribution fora
N
t
-turn oil . . . 233.3 The fundamentalspatial mmf distributionfor a three-phasema hine . . . 26
3.4 Rotationof the resultantmmf . . . 26
3.5 Star of slots . . . 29
3.6 Flow hart to allo atethe statorslots . . . 31
3.7 Non-overlapping single layer winding . . . 38
(a) Winding layout . . . 38
(b) Slot mmf and windingfa tors . . . 38
3.8 Non-overlapping double layerwinding. . . 39
(a) Winding layout . . . 39
(b) Slot mmf and windingfa tors . . . 39
3.9 Single layer overlapping winding . . . 40
(a) Single layeroverlapping winding layout . . . 40
(b) Slot mmf and windingfa tors . . . 40
3.10 Double layeroverlapping winding . . . 42
(a) Double layer overlapping windinglayout . . . 42
(b) Slot mmf and windingfa tors . . . 42
3.11 Determination of the windingaxes . . . 43
4.1 Measurement setup to measure
J(H)
for laminated steel . . . 464.2 The measured data for M470P65A laminationsteel . . . 47
(a) Frequen y dependen y of
J(H)
. . . 47(b) Spe i loss . . . 47
4.3 The extrapolationof the
B(H)
for M470P65A laminationsteel . . . 484.4 Determination of oe ients
k
h
andk
c
. . . 504.5 Permanent magnetmaterial properties . . . 51
4.6 Finiteelementmodel showing the dierent materialsand boundary onditions 54 4.7 Contours
C
1
andC
2
to al ulate the uxlinkage and indu tan e . . . 564.8 Denitionsfor energy al ulation . . . 57
4.9 Saved permeability for a given operating point . . . 59
4.11 Even and odd boundary onditions . . . 62
4.12
B
d
andB
q
as two-dimensional fun tions . . . 64(a)
B
d
asa fun tionofi
d
andi
q
. . . 64(b)
B
q
as afun tion ofi
d
andi
q
. . . 644.13 Spatialharmoni analysis with
I
1
= 0
. . . 66(a) Air gap ux density without statorslots . . . 66
(b) Air gap ux density in ludingthe stator slots . . . 66
4.14 Air gap harmoni analysis with
I
1
6= 0
. . . 67(a) Spatial harmoni analysis with
i
d
> 0
. . . 67(b) Spatial harmoni analysis with
i
q
> 0
. . . 675.1 Tra tion ma hine torque speed hara teristi . . . 71
5.2 Typi alsingle and double layerwinding layouts . . . 72
5.3 Geometri al ma hine parameters . . . 75
5.4 End winding parameters . . . 76
5.5 Winding layout . . . 77
5.6 No-load eld solutionusing FEMP . . . 79
5.7 Tooth widthas design parameter . . . 80
(a)
ψ
1
as a fun tionof tooth width . . . 80(b) Coggingtorque . . . 80
5.8 Air gap uxdensity harmoni s due to the magnets . . . 81
5.9 Staggered rotor . . . 82
5.10 Condu tors lo atedin a stator slotof anele tri alma hine . . . 85
(a) Single and doublelayer overlapping windings . . . 85
(b) Single layer on entrated oil . . . 85
5.11 Eddy urrentsin the permanent magnets . . . 88
5.12 Redu tion fa tor for the permanent magnet's ondu tivity . . . 89
5.13 Design loop . . . 90
(a) Existing pro ess. . . 90
(b) Proposed pro ess . . . 90
5.14 Cross se tionof ma hine and ve tor diagramfor
qd
-variables . . . 915.15 Steady-state
d
- andq
-axes equivalent ir uits . . . 925.16 Torque as a fun tionof
i
d
andi
q
. . . 945.17 Two-dimensional uxlinkage fun tions . . . 95
(a)
ψ
d
asa fun tion ofi
d
andi
q
. . . 95(b)
ψ
q
as afun tion ofi
d
andi
q
. . . 955.18 Findinga validoperating point . . . 96
5.19 Prototypetra tion ma hine . . . 97
5.20 Measured and al ulatedno-load voltageat
62.5 Hz
. . . 995.21 Measured and al ulatedno-load losses . . . 100
5.22 No-load magnetising urrent at
40 Hz
. . . 1015.23 Removed rotor test . . . 104
(a) Measured inputpower with
100 A
. . . 104(b) FEA resultsfor the removed rotor test . . . 104
5.24 Measured stator urrent for torque hara teristi . . . 105
5.25 Measured and al ulatedterminal voltage . . . 105
5.26 Measured and al ulatedtorque . . . 106
5.28 Cal ulated results fromTab. 5.13 (
U
p
= 337.8 V
) . . . 109B.1 Winding fa tors and slotmmf for
Q
s
= 24
andp = 8
. . . 116B.2 Winding fa tors and slotmmf for
Q
s
= 24
andp = 10
. . . 117B.3 Winding fa tors and slotmmf for
Q
s
= 30
andp = 13
. . . 117B.4 Winding fa tors and slotmmf for
Q
s
= 36
andp = 12
. . . 118B.5 Winding fa tors and slotmmf for
Q
s
= 36
andp = 15
. . . 118C.1 Finiteelementmesh with Air Gap Element. . . 120
C.2 Ve torpotential solutionwith positiveperiodi boundary onditions . . . 124
1.1 Dieren e between single and doublelayer windings . . . 4
1.2 Typi alma hine quantities and their German translation . . . 8
2.1 Klíma'sentry in the listof le turers inthe ghetto Theresienstadt . . . 14
3.1 Properties of singleand doublelayer windings . . . 18
3.2 Constraintsfor windingsymmetry. . . 22
3.3
ξ
p
× 10
−3
for single layernon-overlapping windings . . . 353.4
ξ
p
× 10
−3
for doublelayer non-overlapping windings . . . 353.5 Dierent polepair ombinationswith
Q
s
= 30
. . . 364.1 Multiple regression results forM470P65A laminationsteel . . . 49
4.2 Spe i lossfor M470P65A at
1 T
. . . 504.3 Properties of VAC677AP . . . 51
5.1 Drivespe i ation (maximum values) . . . 70
5.2 Maindieren es between form- and random-wound oils . . . 72
5.3 Summaryof single and double layerwinding oils . . . 73
5.4 Slot and pole pair ombinations for ase study design . . . 74
5.5 Winding fa tors when hangingthe slot pit h (
xτ
s
) . . . 785.6 Parameters of the prototype stator . . . 83
5.7 Ironloss al ulation at
20
(B
r
= 1.08 T
) . . . 875.8 Resistan e measurement of the windings . . . 98
5.9 No-load measured data with open terminals . . . 99
5.10 Short ir uittest measured results . . . 102
5.11 Removed rotor loss al ulations . . . 103
5.12 Temperaturerise test at
70 Hz
. . . 1075.13 Cal ulated FEAresults at
70 Hz
. . . 1085.14 Tempeaturerise test loss al ulations . . . 109
Variables
a
number of parallelbran hes, (5.3.2)a
w
number of parallelwires, (5.3.2)k
b
Boules's redu tion fa tor, Fig. 5.12k
r
average loss oe ient for lo aleddy urrents,(5.4.1) - (5.4.4)k
rc
average loss oe ient for ir ulating eddy urrents, (5.4.6)m
number of phases,se tion 3.2m
sl
number of wires arranged aboveea h other, (5.3.2)m
psl
parallel ondu tor wires at dierent heights, (5.4.8)n
1
outerloopvariable, Fig. 3.6n
2
inner loopvariable, Fig. 3.6n
sl
number of wires inslotwidth, (5.3.2)N
p
total number of phasebelts, (3.4.13)N
s
number of series turns perphase, (5.3.3)N
cs
number of oilsides ina slot, se tion 4.3.2.3N
t
number of turns per oil, se tion 3.4.1p
number of polepairs, se tion 3.2.1p
b
number of polepairs of the basi winding, (3.3.2)and (3.3.3)q
number of slots per poleand phase, se tion 3.2.1q
c
number of oils perpoleand phase, se tion3.2.1Q
b
number of slots of the basi winding,(3.3.2)Q
s
number of statorslots, se tion 3.2.1t
greatest ommondivisor, (3.3.1) and (3.3.4)y
d
oilpit h,(3.2.3)y
p
average oilpit h,(3.2.4)z
average numberof turns perslot, (5.3.2)ν
harmoni oder, (3.4.14)ξ
windingfa tor, (3.6.3)µ
r
relativepermeability, (4.2.12)Variables with units
A
Cu
opper ross se tion area,(5.3.7) . . . [m
2
b
c
wire width,Tab. 5.6 . . . [m
℄b
m
magnetwidth, Fig. 5.3 . . . [m
℄b
t
tooth width, Fig.5.3 . . . [m
℄B
r
remanentmagnetisation,Fig. 4.5 . . . [T
℄B
eq
equivalent ux density, (4.2.10) . . . [T
℄C
Esson number, (5.3.1) . . . [kW·m
−3
·min
−1
℄
f
s
samplingfrequen y, (4.5.2) . . . [Hz
℄F
magnetomotive for e, (3.4.3) . . . [A
℄h
1
ee tiveslot height, Fig. 5.3 . . . [m
℄h
c
wire thi kness, (5.4.5)and Tab. 5.6 . . . [m
℄h
m
magnetheight,Fig. 5.3 . . . [m
℄h
s
slotheight,Fig. 5.4. . . [m
℄H
magneti eld intensity, Fig. 4.5 . . . [A·m
−1
℄
H
c
magneti eld oer ivity intensity, Fig. 4.5 . . . [A·m
−1
℄
i
d
d
-axis urrent, Fig. 5.14 . . . [A
℄i
q
q
-axis urrent,Fig. 5.14 . . . [A
℄ˆ
I
1
peak phase urrent,Fig. 5.14 . . . [A
℄I
1
rms
phase urrent,(5.5.7) . . . [A
℄J
magneti polarisation, Fig. 4.5 . . . [T
℄J
s
magneti saturation polarisation, Fig. 4.5 . . . [T
℄l
av
average half-turnlength of a statorwire, (5.3.8) . . . [m
℄l
e
average lengthof a single end onne tion, Fig.5.4 . . . [m
℄l
o
lengthof the end windingoverhang, Fig.5.4 . . . [m
℄l
F e
sta k length,Fig. 5.4 . . . [m
℄l
m
magnetlength, Fig. 5.12 . . . [m
℄L
e
end-windingleakage indu tan e, (5.3.4) . . . [H
℄n
rotationalspeel, Fig. 5.1 . . . [min
−1
℄
P
Cu
opper loss, (5.4.8) . . . [W
℄p
F e
spe i iron loss, (4.2.6) . . . [W·kg
−1
℄
P
F e
iron loss, se tion 5.5.3 . . . [W
℄R
1
stator phaseresistan e, (5.3.7) and (5.4.7). . . [W℄R
b
resistan e of bars, (5.4.7) . . . [W℄R
c
ore lossresistan e, (5.5.6) . . . [W℄R
e
resistan e of the end onne tions, (5.4.7) . . . [W℄T
e
ele tromagneti torque, (5.5.5) and (C.2.22) . . . [N·m
℄T
s
samplingperiod,(4.5.4) . . . [s
℄ˆ
U
1
peak line-to-linevoltage, Fig. 5.14. . . [V
℄U
p
rms
phase voltage, (5.5.8) . . . [V
℄W
energy, (4.3.17) . . . [J
℄α
anglebetweenI
ˆ
1
and thed
-axis,Fig. 5.14 . . . [rad
℄β
anglebetweenU
ˆ
1
and theq
-axis, Fig.5.14. . . [rad
℄η
e ien y, (5.6.4) . . . [%
℄θ
an
urrent sheet anti-node axis,Fig. 3.11 . . . [rad
℄θ
m
magneti axisof a phase, Fig. 3.11 . . . [rad
℄θ
s
staggering angle,(5.3.11) . . . [rad
℄Θ
magnetomotive for e, Tab. 1.2 . . . [A
℄λ
wavelength, Fig. 4.11 . . . [m
℄λ
s
samplingwavelength, (4.5.7) . . . [m
℄σ
ele tri ondu tivity,(5.3.7) . . . [S·m
−1
℄
ρ
spe i mass density, Tab. 1.2 . . . [kg·m
−3
℄
ρ
spe i resistan e, Tab. 1.2 . . . [W·m
℄Φ
magneti ux, (4.6.4) . . . [V·s
℄ψ
uxlinkageψ = NΦ
, (4.3.16) . . . [V·s
℄µ
0
permeability of freespa eµ
0
= 4π10 × 10
−7
. . . [
H·m
−1
℄
τ
s
slotpit h,Fig. 3.6 . . . [m
℄τ
p
polepit h . . . [m
℄φ
anglebetweenU
ˆ
andI
ˆ
1
, Fig. 5.14 . . . [rad
℄ψ
δ
airgap ux linkage, Fig. 5.14 . . . [V·s
℄ψ
sl
stator slotleakage uxlinkage, Fig. 5.14 . . . [V·s
℄ψ
e
end-windingleakage ux linkage, Fig.5.14 . . . [V·s
℄Matri es
A
ve tor potentialat AGE nodes, (C.2.22)i
d
2Dd
-axis urrent matrix, (5.5.12)i
q
2Dq
-axis urrent matrix, (5.5.12)L
indu tan ematrix, se tion 4.3.4.2M
windingmatrix, (3.5.1)S
stiness matrix, (C.2.1)T
used in(C.2.22) fortorque al ulationT
2D torque matrix, (5.5.12)v
slotve tor,(3.5.4)X
regression matrix, (4.2.8)Symbols
C
omplex numbersN
positive integer numbers,n ∈ Z
+
Z
integer numbers. . . , −2, −1, 0, 1, 2, . . .
Z
+
integer numbers1, 2, 3, . . .
Z
∗
Introdu tion
Analysisofele tri alma hinesisaverywell-knownsubje tinele tri alengineeringandat
arstglan eonemightthinkthatthiseld annotbeembarkedonanyfurther. However,
this is not the ase, sin e advan es in materials su h as permanent magnets oer new
possibilities whi h might not have been e onomi al a few years ago. In addition to the
development of the materials used in ele tri al ma hines, the mathemati al and physi s
software toolsfor solving omplex problems andthe visualisationof the results have also
improved. Furthermore,the te hnologiesdriving themi ropro essorsand storagedevi es
usedinpersonal omputers have ontributed tosolve di ultmathemati alfun tions by
meansofnumeri almethods. Allthesefa torshaveinuen edthewayele tri alma hines
aredesignedandwillbedesignedinthefuture. Theresear hpresented inthisdissertation
is a typi al example of how advan es in other elds an lead to new ideas in ele tri al
engineering,and espe iallyele tri alma hines.
1.1 Ba kground to the study
At present most of all existing variable speed drive systems are still based on indu tion
ma hines. A typi alexample are tra tion drivesfor rail vehi les. Today, the state of the
artdrivesystem of metrotrainswith underoortra tionequipment onsistsof a3-phase
IGBT onverterfeedingtwoorfourindu tiontra tionma hinesinparallel. Forthetorque
transmissiontothe wheel-seta gearbox isneeded, be ausethe indu tionma hine annot
develop su ient torque in the available spa e. For the transport operator the gearbox
means investment, maintenan e, undesired noise and losses. Furthermore, Jö kel et al.
(2006a) mentioned that the very high torque developed during a terminal short ir uit,
whi harisesfromaninverterfailure,leads toover-sizedme hani al omponents. Inorder
to redu e energy loss, investment and maintenan e osts there is a trend to repla e the
indu tion motor and gearbox with a dire t drive system as pointed out by Jö kel et al.
(2006 ). This not only improvese ien y and performan e, but alsothe life y le osts,
asshown by Germishuizen et al. (2006).
The ma hine that is used in a dire t drive system requires a high torque to weight
ratio, high e ien y and a wide speed range for onstant power. The permanent
mag-net syn hronous ma hine fulllsthese requirements. Gieras and Gieras (2000) ompared
dierent ma hine types and showed that the permanent magnet syn hronous ma hine
has the highest torque density and highest e ien y. The ma hine an alsobe used for
ma hinethepermanentmagnetsyn hronous ma hine anbedesignedwithagreater
vari-etyof windinglayouts. Theseadvantagesmakeitthereforefeasible todesigndire tdrive
tra tion ma hines using permanent magnet te hnology. Usually su h designs lead to a
ma hine with a high polenumberand often with afra tional slotwinding.
Ko hand Binder (2001)investigatedthe performan eof permanentmagnetma hines
for high speed trains. The suggested ma hine has a double layer winding omprising
form-wound oils. This winding type has a high opper ll fa tor and although it was
foundthat thiswindingtypehas agoodperforman eitisstillquiteexpensivedue tothe
highnumberofform-wound oils. Analternativewindingtypeisanon-overlappingsingle
layer on entratedwinding. Here onlyevery se ondstatortoothhas a oilwound around
it. Advantages are the shorter end-winding overhang, the simplied winding insulation
and a redu ed number of stator oils as pointed out by Cros and Viarouge (2002) as
well as Huth and Qian (2004). The shorter overhang means that the a tive length an
be in reased. As a result the amount of opper in the end-windings is redu ed. These
advantages improve the e ien y and lead to redu ed manufa turing osts. Drawba ks
are the lower power fa tor and possible rotor heating due to the harmoni ontent of
the air gap magnetomotive for e (mmf). In general the higher harmoni s in the air gap
(andasa resultthe in reased leakage indu tan e) ause higher ore lossesas reportedby
Magnussen and Sadarangani(2003).
Thedesignof thewindingused inanele tri alma hine isoftenunderestimated. This
doesnotne essarilyneedtobeseenasnegative,sin etypi allyamanufa turerofma hines
willuse approved designs. Su h apoli yhas a lotto ommend it. Resorting toapproved
designs means saving development osts and even a redu ed design phase. With new
materials available, older ideas whi h seem to be forgotten, an be investigated. Only
then the real omplexity of designing windings omesto the fore.
The idea to repla e the indu tion motor and gearbox with a dire t drive is ertainly
not new. However, it was only until re ently that the progress in material s ien e made
it possible to e onomi ally implement su h a on ept. On 12 August 2008 the Muni h
publi transport ompany MVG started the operation of an underground train that is
equipped with dire t drive te hnology as reported by Tibudd (2008). It is the rst of
its kind and is a pra ti al example of an innovative strategy and approa h to integrate
tra tion, bogie and braking te hnology. This new motor bogie te hnology weighs
30 %
less than onventional state of the art motor bogies. The expe ted redu tion in energylossis estimated to be
20 %
less than the present system in operation and willbe tested over a periodof twoyears.Due to the in reasing demand to redu e manufa turing ost and weight, ma hine
designstendtobe omemore ompa t. Thisleadstoanin reaseinthetorquepervolume
ratio. As aresult,newmaterialswith highmagneti energydensitysu has
Neodymium-Iron-Boron(NdFeB)allowavery ompa tdesigninwhi h aseparti ularlythepermanent
magnetsyn hronous ma hinebenets. Inaddition,theemphasisona urateperforman e
estimations be omes essential and requires in many ases the use of numeri al methods
su h as the nite element method. Compa t designs alsomean that the laminated parts
are highly saturated and as a onsequen e the approximation of the material behaviour
be omes an essential design omponent. And be ause of the la k of a ura y analyti al
solutionsmay lead to inadequate designs. Moreover, redu ing osts may be a hieved by
the use of simple omponents whi h allow easy manufa turing. Therefore, a ma hine
1.2 Problem statement
Designingthe windingis froman ele tromagneti pointof viewthe mostimportantpart
inma hine design. In this step the oilsides of the phasesneed to be assignedto an
ap-propriatestatorslotwhi hallowsarotatingmagneti eldwhenphasedispla ed urrents
areappliedtothe winding. On e the assignment isdone,adja ent oilsides belongingto
the same phase are alled a phase belt. This is the denition used in modern textbooks
su h asthatwrittenby Fitzgerald et al.(1992)and wasalready inuseinearlypapers by
Kauders (1932) 1
.
When designing interior permanent magnet ma hines with non-overlapping
on en-tratedwindingsitisne essary tohaveaninexpensivemethodwhi ha urately al ulates
the ma hine's performan e over a wide speed range. In addition, it is desirable to use
theniteelementmethod(FEM)sin eittakesintoa ountthea tualma hinegeometry
aswell asthe non-linear materialproperties. Additionally, itshould be fast and suitable
inan everyday design environment. Therefore the main resear h question of the present
dissertationis:
How an an analysis method be des ribed to a urately al ulate
the performan e of interior permanent magnet motors with
non-overlapping windings in an everyday design environment?
In order to answer the main resear h question, it is ne essary to have a systemati
algorithm to allo ate the slots. Then, after doing this, the on entrated oils an be
inserted in the appropriate slots. The immediate and essential sub-question that rst
needsto be answered is:
What is the mathemati al expression to allo ate the stator slots
belonging to a phase belt?
Anothersub-questionarisesfromtherequirementtoallo atetheslotsthatbelongtoa
phasebelt. Espe iallywhen usingFEM toanalyse ma hines,the enteringof the winding
arrangementis a omplex pro ess. Theproblem isexa erbated if dierent windingtypes
aretobeanalysed. Furthermore,inthe on eptualdesignphasea ompa trepresentation
of the winding ould simplify the hoi e of the initialgeometri al parameters. Therefore
the se ondsub-question is:
How an a winding layout be represented in a ompa t form?
1.3 Overlapping and non-overlapping windings
Fig.1.1(a)shows adoublelayeroverlapping winding. Inthe drawingsomeofthe oilsare
removed whi hmakes iteasier to identify the two layers. In(b) anon-overlapping single
layerwindingis shown. Onlyea h se ond statortooth has a oilwound aroundit.
Non-overlapping windings are a sub-set of fra tional slot windings and need some
extra dis ussion. It is often ategorised as on entrated. This is not ne essarily wrong
but is in ongruent with the formaldenition. The opposite ofa on entrated winding is
a distributed winding. In the latter ase a the oils of a given phase are distributed in
1
PSfragrepla ements
doublelayer oil singlelayer oil
(a)Overlappingwinding (b)Non-overlappingwinding Statortooth
Figure 1.1: Double layer overlapping and single layernon-overlapping windings
severalslots. Whenreferringtooverlappingandnon-overlappingwindingsa on entrated
winding ould be dened as follows:
1. Formallya on entratedwindingisonewherethenumberofslotsperpoleandphase
equals one. In this ase the oil pit h equals the pole pit h and it is ategorised as
overlapping. This meansthatea h oilsideof thewindingispla edinasingleslot.
If the oilspans a polepit h, itis alled afull-pit h on entrated winding.
2. Non-overlapping ould also be lassied as on entrated, but then it is not done
in terms of the formal denition. Con entrated in this ase means that a oil is
on entrated around astator tooth as shown in Fig. 1.1(b).
Asimpliedillustrationofsingleanddoublelayernon-overlappingwindings 2
isshown
inFig. 1.2 and the dieren ebetween them is given inTab. 1.1. For the purposes of the
present dissertationa double layer windingis dened as follows:
Denition 1.3.1 Adoublelayer windinghastwo oilsidesperstatorslot. Doublelayer
in the ase of overlapping windingsmeans that the oil sides in a slot are pla ed radially
in two layers. In the ase of non-overlapping windings a double layer winding has two
oil sides side byside.
Table 1.1: Dieren e between single and double layerwindings
Single layer Double layer
Inthe ase ofthesinglelayerwindingea h
stator slot has only one oil side assigned
toitas shown in Fig. 1.2(a).
A double layer winding has two oil sides
assigned to a stator slot as shown in
Fig. 1.2(b).
In the literature dierent denitions of terms are in use for non-overlapping
on en-tratedwindings. The most ommon of themare:
2
InGermanthesewindingtypesare ommonlyreferredtoasZahnspulenwhi hmeanstooth oils.
Althoughtooth oilsseemtobeaverydes riptivenameforthesewindings,itisoften alled on entrated
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
0000000000000000000000000000
1111111111111111111111111111
1111111111111111111111111111
1111111111111111111111111111
1111111111111111111111111111
1111111111111111111111111111
1111111111111111111111111111
1111111111111111111111111111
1111111111111111111111111111
1111111111111111111111111111
1111111111111111111111111111
1111111111111111111111111111
0000000000000000000000000000000000
0000000000000000000000000000000000
0000000000000000000000000000000000
0000000000000000000000000000000000
0000000000000000000000000000000000
0000000000000000000000000000000000
0000000000000000000000000000000000
0000000000000000000000000000000000
0000000000000000000000000000000000
0000000000000000000000000000000000
0000000000000000000000000000000000
1111111111111111111111111111111111
1111111111111111111111111111111111
1111111111111111111111111111111111
1111111111111111111111111111111111
1111111111111111111111111111111111
1111111111111111111111111111111111
1111111111111111111111111111111111
1111111111111111111111111111111111
1111111111111111111111111111111111
1111111111111111111111111111111111
1111111111111111111111111111111111
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
0
0
0
0
0
0
0
0
0
0
0
1
1
1
1
1
1
1
1
1
1
1
PSfragrepla ements(a)Singlelayer (b)Doublelayer
xτ
s
2τ
s
τ
s
Figure1.2: Double and single layer non-overlapping windings
Magnussen and Sadarangani (2003): on entrated windings;
Magnussen et al. (2004): on entrated fra tionalpit hwindings;
Salminenet al.(2004a): fra tional slotwound; and
Bian hiand Bolognani (2004): fra tional slot.
Ifthe oils aretobeequallyinshape,whi h simpliesmanufa turing,itisre ognised
thatasinglelayer ouldeasilyhaveavariableslotpit h 3
. Furthermore,the oils ouldbe
implementedas either form-woundor round-wound. Returning to the variable oilpit h
ofthe singlelayer, itis importanttomentionthat this isanextra degreeof freedomthat
oerstobeanattra tivedesignparameter. Theairgapuxthatlinksthe oil ouldthus
bein reased and torque ripple an be improved.
It is therefore helpful to lassify a on entrated winding either as an overlapping
or non-overlapping on entrated winding. Cros and Viarouge (2002) ertainly aroused
interestin non-overlapping on entrated windings with theirpaperentitled Synthesis of
HighPerforman ePMMotorsWithCon entratedWindings,sin ethisisapaperwhi his
veryoftenusedasareferen eonthesewindingtypes. This ouldbeapossibleexplanation
forthe use of the term on entrated windings rather than tooth oilwindings.
The expansion of the lassi al winding types used in ma hines by non-overlapping
windingsoers newpossibilitiesinespe iallytra tionma hine design. Themainresear h
question in se tion 1.2 suggests a design algorithm that takes into a ount the
non-overlapping type. Obviouslya method that isvalidforall types is required. In addition,
itshould be easyto integrateintowhi heverpro ess isin use.
1.4 Approa h to the problem
As a rst step towards the analysis of interior permanent magnet syn hronous ma hines
withnon-overlapping on entratedwindings,anindepthoverviewofthesystemati sof
m
-phasewindingsisrequired. ThepaperofKauders(1932)andthebookofNürnberg(1952)are two very good examples 4
. A histori aloverview of winding design and tenden ies in
3
The on eptofavariable oilpit hisexplainedin hapter3. Alsoreferto Fig.3.5for regularand
irregulardistributed slots.
4
Thisliteratureis unfortunatelywrittenin German,whi hmeansthat itis noteasily a essiblefor
ma hine analyses are given in hapter 2.
Addressing the resear h questions in se tion 1.2, a good starting point would be to
derive the air gap mmf using the lassi al single oil approa h and then expand it for
three phases. It would be ideal to nd a systemati algorithm for assigning the oils to
the stator slots. The derivation of the air gap mmf and winding theory is presented in
hapter3.
The new perspe tive on the winding representation and its properties oered in the
present dissertation are then applied to a tra tion ma hine ase study. Here only the
degreeoffreedom thatsinglelayernon-overlappingwindings provide,isfo usedon. This
is explained in hapter 5. Some important remarks on non-linear ma hine analysis to
support the designresults are explainedin hapter4.
A prototypeand itsmeasured results toverify the design pro ess are given in se tion
5.6. Finally, the on lusions and re ommendations for further studies are presented in
hapter6.
1.5 S ienti ontributions of this dissertation
Thedesignof permanentmagnetmotorsoftenresultsinthe use ofafra tionalslot
wind-ing. Dierentmethodsare availabletodesignthesewindings(whi hare oftengraphi ally
presented),ofwhi hnonetakesintoa ountsinglelayernon-overlappingwindings(whi h
ould have a variableslot pit h). Moreover, due to the dierent slot ongurations,
ma-terialnon-linearitiesand saturationee ts, itis impossibleto nd analyti almethods to
fullltheserequirements. Inaneverydaydesignenvironmentitisne essary tohave
a u-rateand reliabletools. The s ienti ontributionofthis dissertation an be summarised
asfollows:
A general method todesign single layernon-overlapping windings with a xed and
variable slotpit h is oered. The winding design is presented in its most ompa t
form and has all the information on the physi al layout as well as the winding
harmoni s. Itwillbeshownthatthedevelopedmethodappliesto
m
-phasewindings in generalas well. Developingananalyti almodel forthe magneti ir uitof interiorpermanent
mag-netmotorsisnearlyimpossible. Abettermethodistomakeuseofboththe
analyti- alandniteelementmethod. Thema hineis hara terisedbythree(niteelement
analysis generated) two-dimensional fun tions and frequen y dependent losses are
al ulated analyti ally.
The methodology fromthe study is appliedto a
150 kW
prototypemotor designed for a tra tion appli ation. The manufa tured ma hine has30
stator slots and20
poles. Pra ti al measurements were done to verify the omputational resultspre-sented.
1.6 Delimitations of the study
Theaimof thisstudy isthe analysisofinteriorpermanentmagnetsyn hronous ma hines
with single layer non-overlapping windings. A omparison with other ma hine types is
namelythe variable oilpit h. Sin e thisis nottypi alforother 5
windingtypes, this will
betheonlydesign aspe t that willbefo used on. Adetailedma hineoptimisationisnot
provided.
1.7 Layout of the dissertation
The resear hresults are presented in six haptersas follows:
Chapter 2: A short literature overview whi h gives a brief histori al view of winding
design is presented. This is followed by general tenden iesin ma hine analyses.
Chapter 3: Analgorithmisderivedtopresentawindinginamatrixform. This ompa t
formallowsthe al ulationofthewindingfa torsforallharmoni s. Thebasisofthe
algorithmisthe phase beltsequen e and the phasebelt onstraint whi h isderived
from the airgap mmfenvelope fun tions.
Chapter 4: This hapterdes ribesthe nonlinearmagneti ir uitanalysisoftheinterior
permanent magnet ma hine. The hosen method is that of the nite elements.
The hapter starts o with an introdu tion to the nonlinear materials used in the
design, followed by a detailed al ulation of the relevant ma hine quantities whi h
are ne essary for aperforman e al ulation. The hapter on ludes with adetailed
explanation ofhow toperform aharmoni analysis.
Chapter 5: FEAand analyti almethodsare ombined tointrodu e adesign pro edure
suitable in anan everyday design environment. The hapter on ludes with a ase
study ofa prototype tra tionmotorwith non-overlapping singlelayer on entrated
windingsandinteriorpermanentmagnets. Pra ti almeasurementsontheprototype
are ompared to al ulated values and dis ussed.
Chapter 6: This hapter entails on lusion and re ommendations forfurther study.
Where appli ableallthe explanations are a ompaniedby anexample based onthe
pro-totype ma hine whi h has
30
stator slots and20
poles. A detailed des ription for this ombinationis given in hapter5.1.8 Di ulties en ountered during the study
Themajordi ultyen ounteredduringthestudywasthelimitationsof ommer ialnite
elementsoftware to generate and post-pro ess nite element analysis data in ane ient
way. Even the integration thereof in ustomised software tools was nearly not possible.
This di ulty was over ome by using non- ommer ial nite element software, FEMP
(Finite Element Method Program). However, it was rst ne essary (through intensive
Fortran programming) to adapt the software to in lude permanent magnets and allow
positive boundary onditions.
5
Whenreferringtootherwinding types,itmeansdoublelayernon-overlappingaswellas singleand
Initiallythewindingfa torandresultsfromthedis reteFouriertransformweretreated
asabsolutevalues. Keeping these values as omplexnumbers greatly simpliesthe
anal-ysis, sin e (through a proper hoi e of referen e axis) information on the winding axes
and the
dq
-variablesare given.1.9 Notes to the reader
The work presented in this thesis was done in Nürnberg, Germany. Consequently many
of the literature used were in German. To name only a few, books by Nürnberg (1952),
Klamt (1962) and Jordan et al. (1975) are still ommonly used in design o es. It is
noti eable that even thoughthere exists a list of symbols for various quantities, English
and German have in some ases dierent symbols for the same quantity. This is mainly
duetothe fa t thatlanguages develop individualand of ourse olloquiallanguagerules.
Even adire t translation does not ne essarilygive the typi alword used. An example is
theGermanwordFelderregerkurve. Adire ttranslationwouldbeeldex itation urve,
whi h of ourse is not wrong. However, the eld ex itation urve in ele tri al terms
is usually known as the magnetomotive for e (mmf). Tab. 1.2 gives some ommon
ele tri alma hine quantities in English and German,and the dierent symbols inuse.
Table1.2: Typi alma hinequantities,thereGermantranslationand ounterpartsymbols
Quantity English German Unit
magnetomotivefor e (Dur hutung)
F
Θ
A
uxlinkage (Flussverkettung)
λ
ψ
V·s
voltage (Spannung)
V
U
V
spe i resistan e (spezis her Widerstand)
ρ
̺
W·m
spe i weight(spezis her Masse)
ρ
γ
kg·m
−3
ondu tivity (Leitfähigkeit)
σ
κ
S·m
−1
ross se tionarea (Quers hnitt)
A
Q
m
2
turn number(Windungszahl) N W
-It is ommendablethat the Germanterminology onele tri alma hines allows a very
pre isedes riptionofalmostallaspe tsrelatedtothesubje t. Inmany asesitisdi ult
to nd an equivalent te hni al term in English, be ause it simply does not exist. The
problem of dierent terminologies even exists within a language: dierent s hools use
dierentterms whi hmakesthe study of ma hinerelated books not easy. Alsohistori al
hanges needto be kept inmind.
The term ma hine as used inthe present dissertation means it ould be a ma hine
that is operated either as a motor or as a generator. In the title the term motor is
used, sin e only the measured results of motor operation are presented. However, in the
theoreti alse tions, the word ma hine is preferred.
The do ument is optimised for a two-sided printout. A ip-book shows that the
Literature overview
In this hapter a brief overview of the history of stator windings (layout and design) is
given. Sin e this topi evolved over many years anin depth history is beyond the s ope
of the present dissertation. However, a fewimportantworks need tobementioned.
2.1 A histori al view of winding design
In the rst of his two remarkable papers Kauders (1932) explains the systemati s of
stator windings and the al ulation of the winding fa tors. The aim in this work was
to determine the parameters that hara terise the air gap mmf of the winding. Also in
this paper the indu ed voltage in the oil sides is already mentioned and represented
as a ve tor. The resultant ve tor diagram was alled the star of oil groups (German:
Spulengruppenstern). The adja ent ve tors onsu h a diagramthat belong tothe same
phaseis alled aphase belt (German: Zone).
Twoyears later,in these ond paper by Kauders(1934), thealgebrai methods
devel-oped in the rst paperwere visualised by means of Tingley's diagram. The latter ould
bereferred toas alinear representationof the now alled star ofslots (German:
Nuten-stern). The star of slots is onstru ted using the ele tri al angle between two adja ent
slots. Computerte hnologyasweknowittodaywasnotavailableatthe timeandtheuse
of graph paper ertainly was ommon. Furthermore, su h graphi al methods denitely
ontributed tothe subje tof stator windings.
Vilém Klíma (Wilhelm Kauders) died on 6 O tober 1985 and in an obituary by
Frohneand Seins h (1985) it is mentioned that Klíma's equation for the distribution
fa tor 6
of fra tionalslotwindings isnot found in textbooks. Another remarkin the
obit-uary is that in some referen es it is stated that it is not possible to nd a losed-form
expression for the winding fa tor of fra tional slot windings. In the rest of this
litera-ture overview, none of the authors (ex ept Kremser) refers to or makes use of Klíma's
losed-formexpression 7
.
MorethanftyyearslaterthanKauders(1934),Kremser(1989)introdu eda
ompre-hensivestudy offra tionalslotwindingsand the urrentsinthe parallelpaths ofrotating
ma hines. The reasonforthis bigtime gap doesnot meanthat nothinghad happened in
themeantimeinwindingdesign;itshouldbekept inmindthatthis overview onlypoints
out some important aspe ts. In the rst part of Kremser's do toral thesis a detailed
6
Alsoknownasthebreadthfa tor.
7
algebrai des ription of single and double layer fra tional slot windings is presented. In
ordertoallo atethe oils tothe statorslots,modulararithmeti ,whi huses theso- alled
ommutatorpit h,isused. This anbeseenasasteptowardusing omputerprogramsto
automati allyperformingwindingdesigns. Inthestar ofslotsthe ve torsstartwrapping,
andon eput onagraph,twove torshavinganglesof
45
°and405
°respe tivelylieexa tly onea h other. The starof slots isthusthe same asthe remainderafterdividing anangleby
360
° whi h an easily be implementedby means of the modulo fun tion.Wa h (1997) explains another algebrai method to design stator windings whi h has
mu hin ommonwith thatpresented by Kremser(1989) 8
. Themethodspresented inthe
paper is hara terised by the matrix representation of a winding. The winding matrix
simplifythe winding fa tor al ulation and avoids omplex equations. Basi ally the
ve -tors of the star of slots is used in matrix form. What is alled the windingmatrix ould
beseen as the matrix representation of Tingley's diagram. A drawba k of the method is
that ea h matrix entry for double layer windings has two values. In omputer
program-ming this should be solved by either a multi-dimensionalmatrix or two two-dimensional
matri es. In another paperbyWa h(1998),the fo us isontheoptimisationof fra tional
slot windings. In spite of the fa t that this paper is a very good sour e on the topi of
windingdesign,none of the literature inthe rest of this se tionrefers toit.
Cros and Viarouge(2002)showed thatthe useofstator oilswound aroundthestator
teeth has some attra tive advantages for manufa turing. The oils donot overlap whi h
means that the manufa turing is simplied and the winding overhang is less than that
of overlapping windings. Ele tri ally the shorter windingoverhang means the use of less
opper. Furthermore, the non-overlapping windings are derived from single or double
layer overlapping windings. It is also pointed out that in the ase of single layer
non-overlapping windings the oil pit h an be in reased and used as a design parameter 9
.
The latter, however, is not taken into a ount in the winding fa tor al ulation. This
paper is referen ed quite often, proving the interest among ma hine designers in
non-overlapping windings.
Salminen(2004)didanextensivestudyonfra tionalslotwindingsforlowspeed
appli- ations. Themainobje tiveinthedo toralthesisisto omparedierentslot ombinations
ofama hinewith axed airgap diameter inthe
45 kW
range. Althoughboth singleand doublenon-overlapping windings are taken into a ount,the fo us is onthe double layertype. Sin e round wire oils are used inthe investigated appli ation it is appropriate to
use a double layer rather than single layer. If form-wound oils were used the hoi e of
double layer windings would have aused some di ulties when inserting the oils into
the stator slots. Also, the stator had semi- losed slots. The star of slots isreferred to as
the voltage ve tor graph in Salminen'sthesis. A very important omment by the author
isthat areshouldbetakenwhenthewindingfa torsfordierentwindingtypesaretobe
al ulated. Dependingon the windingtype the rightset of equationsshould be hosen.
Skaar et al.(2006)proposedamethodto al ulatethewindingfa torsfor on entrated
oils without knowledge of the winding layout. This is of ourse su ient to qui kly
ompare dierent slot and pole ombinations. Using this method it is shown that the
best numberof slots per poleand phaseshould be inthe range ¼
≤ q ≤
½.The tutorial presented by Bian hiet al.(2007) is anin-depth study of fra tional slot
8
It should howeverbe noti ed that the authors Wa h and Kremserare from Poland andGermany
respe tively.
9
windingsingeneral. In theirtutorialnotesadistin tionismadebetween overlappingand
non-overlapping windings. Here too the non-overlapping type is derived from a double
layeroverlapping winding.
2.2 Tenden ies in ma hine analysis
When analysing permanent magnet ele tri al ma hines it is not ne essary to start from
s rat h. Sin e the indu tion ma hine has been known for more than
100
years the devel-oped methods serve as a good basis. A well-known book on indu tion ma hines is DieAsyn hronmas hine ( ommonly known as the indu tion ma hine) by Nürnberg (1952).
The book gives a very detailed des ription of modelingte hniques whi h an be used in
the design of indu tion ma hines (whi h is of ourse not limited to indu tion ma hines
only).
2.2.1 Introdu tory remarks
It is important to mention that Nürnberg (1952) uses equivalent
B(H)
urves for the al ulation of the mmf's in the teeth and yoke. This is not wrong and indeed helps tounderstandthephysi albehaviourofthema hine. However,the urvesevolvedovermany
years and are based onmeasured results and are only validfor steady-state operation.
Fig.2.1illustratestheprin ipleofleakage ompensation(German: Zahnentlastung).
Due to saturation, ux lines leave the teeth tips and are parallel to the slot wall. This
means that the ux density in the stator teeth de reases. Sin e the single layer
non-overlapping winding ould have avariable oilpit h the statorhas two tooth types. The
validity of the ompensation urves given by Nürnberg (1952) and applied to ma hines
with dierent stator slots was not investigated and ould be a very interesting topi for
furtherstudy. Theapproa hinthepresentdissertationistousetheniteelementmethod,
whi h al ulates the exa t ux densitiesby means ofa physi al model.
00000
00000
00000
00000
00000
00000
11111
11111
11111
11111
11111
11111
PSfragrepla ementsB
1
B
2
B
3
B
4
B
5
Fluxlines YokeH/A·m
−
1
B/T
B
1
B
5
H
1
H
5
Slotwall2.2.2 The nite element method as a design tool
Frommostre ent onferen e ontributionsthereisanoti eabletenden y tousenumeri al
methodsliketheniteelementmethodintheinitialdesignphase. Germishuizen and Kamper
(2007) proposed anew al ulation methodfor al ulatingthe ma hine performan e from
ux linkage and torque fun tions. These fun tions are obtained from a nite element
analysis (FEA).
Centner (2008)introdu ed a software toolfor the design of high-speed indu tion
ma- hines. Forsu hma hinestheloss al ulationandtheinuen eofthe materialproperties
are very important. Therefore, to a ount for these aspe ts the nite element method
is ombined with analyti al expressions. Hafner et al. (2008) suggested a similar design
strategyfor the same reasons.
Rei hert (2004) explains a simplied approa h to determine ma hine hara teristi s
by means of the nite element method. In doing so the non-linear materialbehaviour is
a ounted for and is still fast and a urate. Garbe et al. (2008)also ombines FEM and
analyti alexpressions to al ulate motor parameters.
2.2.3 The inuen e of saturation on transient solutions
Seifert and Strangmüller(1989)showedthattheuseofindu tionmotorparameters
al u-latedinthe lassi alwayleadstoina uratepredi tionsofma hinebehaviourunderfault
onditions su h as terminalshort ir uits. Experien ed based fa tors were introdu ed to
a ount for the la k of a ura y. Again, this is an example of a situation where results
must be known in order toobtain useful saturation fa tors.
Besides the saturation of the materials used the modelingthereof is alsoessential for
ma hine behaviour. The material behaviour up to saturation is explained in se tion 4.2
and,whereappli able, referen ed tothe literature. Poni k(1993) showed thatnegle ting
theee tofsaturationintransientsolutions anleadtoina urateresults. Itissuggested
thatnot onlytheindu tan esinuen ed by saturation,butthe hange in urrentaswell,
should be taken intoa ount. Avoiding saturation ould result in undersizing of riti al
me hani al omponents. Depending on the ma hine appli ation, undersized me hani al
parts an be ome asafety problem, forexample in tra tiondrives for railvehi les.
Retièreand Ivanès(1998)oeredstudymethodsofthree-phaseshort ir uitsof
indu -tionma hines. Thereversepeaktorqueduringashort ir uit an auseseriousme hani al
stress. The authors show that for indu tion ma hines the saturation and the skin ee t
in the rotor bars should be a ounted for in the al ulation of the reverse peak torque
and that it is ne essary to take into a ount the saturation ee ts. Brauer et al. (2004)
ompared the dynami short ir uits of indu tion ma hines (IM) to that of permanent
magnetsyn hronous ma hines (PMSM). An important out ome of their study is that in
the ase of the two-phase short ir uit a PMSM starts to os illate and ontinues to do
so until it is isolated from the inverter. It is alsoemphasised that these fault onditions
should be al ulated using FEM. Ei hler et al. (2008) alsodes ribes that linear ma hine
models annotbe used to simulatewind generators in the
5 MW
range.Thereisala k ofexperien e inthedevelopment ofpermanentmagnetma hines
om-paredtoindu tionma hines. Thismeansthattheuseofforexample ompensation urves
toa ountforsaturationshouldbeavoided,sin eitmightleadtoina uratedesign
al u-lations. Furthermore,itshouldbenoti edthatsu h ompensation urveswere developed
any-moreandthe useof stateofthe artte hnologyisadvisablefora urate predi tions. This
iseven more importantas prototyping (and measurementsto nd ompensationfa tors)
isto be kept to a minimum for e onomi reasons.
2.2.4 Present approa hes in iron loss al ulation
The huge number of s ienti ontributions found on iron loss al ulations emphasises
the omplexity of iron losses. Two aspe ts that denitely ontribute to this fa t is that
the material data annot be des ribed analyti ally and is usually presented in graphi al
form. Se ondly, the inuen e of the manufa turingpro ess isdi ult toa ount for.
It is important tomention that a goodapproximation of the spe i loss is found in
Nürnberg(1952). Here the spe i lossof one kilogramofdeburred laminatedsteel in
W
isgiven byp
F e
=
k
h
f
50
+ k
c
f
50
2
!
B
1 T
2
(2.2.1)The oe ients
k
h
andk
c
a ount for the hysteresis and lassi al eddy urrent losses respe tively. Sin e the frequen y in the stator is onstant for a given operating pointit is onvenient to separate the frequen y and ux density omponents of the loss as in
(2.2.1). In Nürnberg's book it isalsomentioned that the iron loss al ulated inthis way
isless than the measured loss using the Epstein 10
frame. To a ount for this ina ura y
it is suggested to multiply the al ulated iron loss in the yoke and teeth by
1.5
and3
respe tively. Thisholdsforstatorslotswhi hissemi- losed. Foropenstatorslotsthelossinthe teeth tends to in rease even further. Due to this un ertainty it is understandable
that Nürnberg (1952) advised the following 11
: Wegen der erhebli hen Unsi herheit bei
derEisenverlustbere hnung gehemanaberni htindieFeinheiten. This shows thelimits
of empiri almodels beforete hniques likethe nite elementmethodwere introdu ed.
Bertotti (1988) introdu ed what is alled the ex ess loss and thereby expanded the
analyti alexpression for spe i loss by a
k
e
(f B)
1.5
term. Thus, (2.2.1) hanges to
p
F e
= k
h
f + k
c
f
2
B
2
+ k
e
(f B)
1.5
(2.2.2)Linet al. (2003) used the expression (2.2.2) in a transient nite element analysis. The
advantageindoingsoisthat thea tualuxdensitiesaredire tlyobtained fromthenite
element solution. Germishuizen and Stanton (2008) showed that the loss al ulation in
(2.2.2) is only a theoreti al loss estimation based on extrapolated data obtained from
resultsmeasured inthe Epsteinframe. Instead of usingdierentfa tors forthe yokeand
teethtoa ountforthemanufa turinginuen e,asinglemanufa turingfa torshouldbe
applied to the overall result. For ma hines in the
200 kW
range a manufa turing fa tor of two is found to be adequate. Another out ome of this paper is that inadequate lossomputations ould easilylead to misinterpretation of results.
10
InNürnberg(1952)theEpsteinframeiswronglyspeltEppstein(whi hisasmalltowninGermany
notfarfromFrankfurtamMain).
11
2.3 A short ex ursion into the life of one of the
authors onsulted
AtthispointIwouldliketomakeashortex ursionandreportaveryinterestingin iden e
forthe following reasons:
Nearlynoneofthereferen es onsultedtooknoteofVilémKlíma's losedexpression
for the al ulationof the distribution fa tor.
Vilém Klíma'slife storyis aninspiration for young resear hers.
Published resear habout theintelle tuallifeoftheJewisheliteintheghetto
There-sienstadt led me to this dis overy.
If one takes a loser look at the referen es listed in the last paper by Klíma (1979),
there is an entry entitled Systematik der Drehstromwi klungen and the author is given
as V. Klíma (Kauders). To supply a name between bra kets is not typi al, and this
o urren emademe suspi ious, sin ethe onlypaperI ouldndwith thistitle iswritten
by Wilhelm Kauders. As I ould not explain the enigma, I started to ask questions
and never expe ted to nd answers about tragi events that o urred duringthe Se ond
World War. Sin e I grew up in South Afri a whi h, ompared to other ountries, was
littleee ted by the war,this dis overy made the experien e extraordinary.
2.3.1 Coming a ross a very interesting sour e
Nowadays it is typi al touse the well-known Google sear h engine. Tryinga few
ombi-nations of the names Kauders and Klíma, I ame a ross a list of le turers in the ghetto
of Terezin(Theresienstadt). The entry details forKauders are given inTab. 2.1.
Table 2.1: Klíma'sentry in the listof le turers inthe ghetto Theresienstadt
Name and Title Kauders (Klima) (Vilem)Dr.
Birth Date 10.04.1906
Deportedto Terezin 04.12.41
Deportedfrom Prague
Deportedfrom Terezin 01.11.44
Survived in Grossrosen
This led me to the book University Over The Abyss: The story behind 520 le turers
and 2,430 le tures in KZ Theresienstadt 1942-1944 by Makarova et al. (2004). A very
interesting detail from the book is that Dr. Golds hmied and Dr. Kauders were se retly
taken to Germany to improve the performan e of German radar 12
. A witness, Gerda
Haas, remembered the following:
12
A ordingto IvanKlíma (Klíma'sson),VilémKlímawassenttogetherwith twoother spe ialists:
Mr. Golds hmiedandMr. Kohntothe on entration ampin Grossrosen. Mr.Golds hmiedandKlíma
stayedonlyforafewweeks inGrossrosen. Asaresultoftheeva uationofthe ampbothofthemwere