Upload
others
View
3
Download
0
Embed Size (px)
Citation preview
Purdue UniversityPurdue e-Pubs
Open Access Theses Theses and Dissertations
4-2016
Mixed mesh/nodal magnetic equivalent circuitmodeling of a six-phase claw-pole automotivealternatorDaniel C. HorvathPurdue University
Follow this and additional works at: https://docs.lib.purdue.edu/open_access_theses
Part of the Applied Mechanics Commons, and the Power and Energy Commons
This document has been made available through Purdue e-Pubs, a service of the Purdue University Libraries. Please contact [email protected] foradditional information.
Recommended CitationHorvath, Daniel C., "Mixed mesh/nodal magnetic equivalent circuit modeling of a six-phase claw-pole automotive alternator" (2016).Open Access Theses. 775.https://docs.lib.purdue.edu/open_access_theses/775
Graduate School Form30 Updated
PURDUE UNIVERSITYGRADUATE SCHOOL
Thesis/Dissertation Acceptance
This is to certify that the thesis/dissertation prepared
By
Entitled
For the degree of
Is approved by the final examining committee:
To the best of my knowledge and as understood by the student in the Thesis/Dissertation Agreement, Publication Delay, and Certification Disclaimer (Graduate School Form 32), this thesis/dissertation adheres to the provisions of Purdue University’s “Policy of Integrity in Research” and the use of copyright material.
Approved by Major Professor(s):
Approved by:Head of the Departmental Graduate Program Date
Daniel C. Horvath
MIXED MESH/NODAL MAGNETIC EQUIVALENT CIRCUIT MODELING OF A SIX-PHASE CLAW-POLEAUTOMOTIVE ALTERNATOR
Master of Science in Electrical and Computer Engineering
Steven D. PekarekChair
Gregory M. Shaver
Scott D. Sudhoff
Steven D. Pekarek
Venkataramanan Balakrishnan 4/18/2016
MIXED MESH/NODAL MAGNETIC EQUIVALENT CIRCUIT MODELING OF A
SIX-PHASE CLAW-POLE AUTOMOTIVE ALTERNATOR
A Thesis
Submitted to the Faculty
of
Purdue University
by
Daniel C. Horvath
In Partial Fulfillment of the
Requirements for the Degree
of
Master of Science in Electrical and Computer Engineering
May 2016
Purdue University
West Lafayette, Indiana
ii
ACKNOWLEDGEMENTS
First I would like to thank my Advisor, Dr. Steven Pekarek. I am frequently
amazed with his patience both in teaching and in advising. I have been fortunate to work
for him these past years in my undergraduate and graduate career and because of this I
have had fantastic research opportunities and projects that complement my coursework. I
may not be in graduate school had he not asked me to do undergraduate research under
him years ago.
Thank you to my family. My parents David and Kimberly Horvath have always
fostered my curiosity as I grew up – they have always allowed me to freely pursue
projects that would provide a learning opportunity, and I couldn’t ask for better parents.
Sally, my best friend, thank you for your love and support during the past nine years.
Of course I must thank all the students in Power and Energy research area. I am
glad to have such a great group of friends/co-workers. I think our comradery is unique
and special; it is something that every research community needs. Whenever I wish to
discuss a concept or ask a question, I know that there isn’t a single person in our group
who I cannot go to.
Finally, thank you to Remy International, Inc./Borgwarner for the continued
support of this work, especially Dave Fulton, Katie Riley and Greg Holbrook.
iii
TABLE OF CONTENTS
Page
LIST OF TABLES .............................................................................................................. v
LIST OF FIGURES ........................................................................................................... vi
ABSTRACT ..................................................................................................................... viii
INTRODUCTION ....................................................................................................... 1
GEOMETRY OF LUNDELL ALTERNATOR AND
RELUCTANCE/PERMEANCE DERIVATION ........................................................ 3
2.1 Claw Pole Geometry Introduction and Construction .................................................3
2.2 Flux Tube Fundamentals ............................................................................................5
2.3 Stator Flux Tubes .......................................................................................................8
2.3.1 Yoke Flux Tube ............................................................................................. 10
2.3.2 Tooth Flux Tube ............................................................................................ 11
2.3.3 Slot Leakage Flux Tube ................................................................................. 12
2.4 Rotor Flux Tubes ......................................................................................................13
2.4.1 Shaft and Core Flux Tubes ............................................................................ 15
2.4.2 End Disk Flux Tube ....................................................................................... 17
2.4.3 End Piece Flux Tube ...................................................................................... 18
2.4.4 Claw Section Flux Tube ................................................................................ 19
2.4.5 Rotor Leakage Flux Tubes ............................................................................. 22
2.4.6 Permanent Magnet ......................................................................................... 25
2.5 Air Gap Flux Tubes ..................................................................................................26
ALGEBRAIC SYSTEM OF MAGNETIC CIRCUIT EQUATIONS ....................... 33
3.1 Stator Teeth Mesh Equations ...................................................................................33
iv
Page
3.2 Nodal Analysis at Stator Tooth Tip Nodes ..............................................................34
3.3 Mesh/Nodal Analysis at Claw Section Nodes .........................................................35
3.4 Relating Loop Flux to Phase Flux ............................................................................37
3.5 Mesh Analysis of Rotor ...........................................................................................37
3.6 Overall Algebraic System ........................................................................................39
3.7 Solution of Nonlinear Algebraic System .................................................................40
MODELS FOR DIFFERENT LOADING CONDITIONS ....................................... 42
4.1 Passive Rectifier State Model ..................................................................................42
4.2 Open Circuit Model ..................................................................................................49
4.3 Active Rectifier Model .............................................................................................51
VALIDATION OF MODELS ................................................................................... 55
5.1 Passive Rectifier State Model Validation ................................................................55
5.2 Open-Circuit Model Validation ...............................................................................58
5.3 Active Rectifier Model Validation ...........................................................................59
CONCLUSION .......................................................................................................... 60
LIST OF REFERENCES .................................................................................................. 61
APPENDICES
A. AIRGAP PERMEANCE EXPRESSIONS ............................................................. 62
A.1. Triangular Region 1 .................................................................................................63
A.2. Rectangular Region ..................................................................................................64
A.3. Triangular Region 2 .................................................................................................65
A.4. Chamfer Region .......................................................................................................66
B. MATRIX EXPANSION ......................................................................................... 78
v
LIST OF TABLES
Table .............................................................................................................................. Page
2.1: Variable Identifiers and Descriptions, Stator Quantities. ............................................ 9
2.2: Variable Identifiers and Descriptions, Rotor Quantities ............................................ 15
2.3: Variable Identifiers and Descriptions, Air Gap Quantities ........................................ 28
vi
LIST OF FIGURES
Figure ............................................................................................................................. Page
Figure 2.1: Example rotor of a claw-pole machine, without field coil. .............................. 4
Figure 2.2: Simple cylindrical flux tube. ............................................................................ 5
Figure 2.3: Flux tube with non-uniform cross-sectional area. ............................................ 6
Figure 2.4: Flux tube with non-uniform length. ................................................................. 7
Figure 2.5: Generic stator MEC. ......................................................................................... 9
Figure 2.6: Illustration of yoke flux tube. ......................................................................... 10
Figure 2.7: Stator tooth. .................................................................................................... 11
Figure 2.8: Coffin-shaped slot leakage permeance. .......................................................... 12
Figure 2.9: Rotor Section. ................................................................................................. 14
Figure 2.10: Physical Rotor. ............................................................................................. 14
Figure 2.11: Shaft and core flux tubes. ............................................................................. 16
Figure 2.12: End disk of rotor. .......................................................................................... 17
Figure 2.13: End piece, or “knuckle,” of rotor. ................................................................ 18
Figure 2.14: Detail of rotor claws. .................................................................................... 20
Figure 2.15: Rotor field leakage path detail. ..................................................................... 22
Figure 2.16: Detail of the claw-claw leakage permeance. ................................................ 25
Figure 2.17: Permanent magnet in between claws. ........................................................... 26
vii
Figure Page
Figure 2.18: Formation of airgap flux tubes. .................................................................... 27
Figure 2.19: Two cases for determining airgap length over the chamfer region. ............. 30
Figure 3.1: MEC structure detailing the discretization of the airgap and stator into respective flux tubes. ..................................................................................... 34
Figure 3.2: Discretization of claws into sections. ............................................................. 36
Figure 3.3: MEC structure detailing discretization of rotor into respective flux tubes. ... 38
Figure 4.1: Passive rectifier connection to machine. ........................................................ 43
Figure 4.2: Lower diode voltage. ...................................................................................... 44
Figure 4.3: Solution procedure block diagram. ................................................................ 49
Figure 4.4: Active rectifier connection to machine. ......................................................... 52
Figure 5.1: Comparison of MEC and hardware AC quantities for rectified alternator with resistive load. ..........................................................................................56
Figure 5.2: Comparison of MEC and hardware AC quantities for rectified alternator with resistive load. ..........................................................................................57
Figure 5.3: Comparison between open-circuit line-line voltage prediction of MEC, FEA models and with experimental data. ......................................................58
Figure 5.4: Torque vs speed profiles for FEA and MEC. ................................................. 59
Appendix Figure
Figure A.1: Formation of airgap flux tubes. ..................................................................... 62
Figure A.2: Two cases for determining flux tubes over chamfer region. ......................... 66
viii
ABSTRACT
Horvath, Daniel C. M.S.E.C.E., Purdue University, May 2016. Mixed Mesh/Nodal Magnetic Equivalent Circuit Modeling of a Six-Phase Claw-Pole Automotive Alternator. Major Professor: Steven Pekarek.
Claw-pole, or Lundell alternators, are used as the charging source in most
commercial ground vehicles. Increasingly, manufacturers of these machines are being
required to meet strict acoustic noise and vibration specifications. This has led to
alternator designs with more than the traditional three phases. Increasing phase count
above three has been shown to reduce torque ripple, which is an acknowledged source of
the acoustic noise.
In this research, a magnetic equivalent circuit is used to first establish a model of
the nonlinear relationship between stator and field winding current and magnetic flux in a
claw-pole machine. The flux/current relationship is coupled with a state model to predict
the electrical dynamics of the machine connected to a passive rectifier/battery.
Subsequently, two variants of the model are derived to facilitate design optimization
under alternative excitation strategies. In one, the inputs are selected as the stator currents
in order to consider the performance of the machine connected to an active rectifier. In
another, the model is structured to compute open-circuit performance, which facilitates
relatively quick evaluation of model accuracy. Validation of the models is performed
using both FEA and laboratory experiments.
1
INTRODUCTION
Magnetic equivalent circuits (MECs) have been employed by many researchers to
model the relationship between magnetic flux and current in electromagnetic systems
such as electric machines, transformers and inductors [1] ,[2]. Magnetic circuits are
analogous to electric circuits where voltage, current, resistance and conductance are the
respective counterparts of magneto-motive force (MMF), magnetic flux, reluctance and
permeance. The solution of MECs can be accomplished with the plethora of techniques
developed for electrical circuit analysis. Specifically, mesh analysis, based on Kirchoff’s
Voltage Law (KVL), and nodal analysis, based on Kirchoffs Current Law (KCL), are two
very common solution techniques. Once an MEC is established, the question is often of
which circuit analysis technique should be applied in order to minimize computational
effort.
For linear circuits, there is little advantage to using mesh over nodal analysis.
Using one method may yield a system with fewer equations, but for most problems the
difference in unknowns is insignificant. When analyzing nonlinear magnetic systems,
researchers have noted a significant difference in mesh versus nodal analysis. Derbas et al
have noted that for nonlinear MECs a mesh analysis reduces the number of iterations
required to solve the nonlinear system using a Newton-Raphson method [3]. It was
further shown that for strong nonlinearities caused by magnetic saturation, a nodal-based
solution will often fail to converge whereas a mesh-based solution will converge.
2
It is relatively easy to apply MEC analysis to stationary magnetic systems.
However, modeling electric machinery with MECs can be challenging since the circuit
structure can depend on the position of the rotor. Specifically, in the case in which mesh-
based solution techniques are applied, the circuit components representing the airgap will
tend to infinite values as stator/rotor structures (i.e. teeth) come into and out of alignment.
As a result, one must eliminate these elements and establish new KVL loops with the
remaining non-infinite components. Researchers have developed algorithms to automate
the loop construction process [4]. However, the algorithms require one to categorize all
potential overlap positions, which is a challenge for claw-pole machines. One does not
experience this issue in nodal-based solution techniques. However, since machines tend
to operate in saturation, issues of convergence are often encountered.
In this research, an alternative solution technique is provided in which mesh
analysis is used in all magnetically nonlinear flux tubes while nodal analysis is used to
solve for quantities in the airgap. This has the potential to use the advantage of each
solution technique. This research builds upon that presented in [5], in which permeance
expressions for all flux tubes were developed for a nodal MEC model of a three-phase
claw-pole alternator with a delta-connected stator. In addition to the mixed mesh/nodal
system a second focus is to explore new model configurations including six-phase
machines with wye-connected stator and permanent magnets on the rotor. Validation of
the models that are proposed is performed using both FEA models and experimental data
from commercially-available alternators.
3
GEOMETRY OF LUNDELL ALTERNATOR AND RELUCTANCE/PERMEANCE DERIVATION
The claw-pole machine will require a more complex geometrical description than
most machines encountered because its magnetic structure must be considered in three
dimensions. In this chapter, a focus is to first introduce the structure of the stator/rotor.
Subsequently, the circuit structure and circuit elements are defined and parameterized
which is based on the derivation in [5], however modifications are made where necessary
for the six-phase, P pole machine.
2.1 Claw Pole Geometry Introduction and Construction
The claw-pole machine is a wound rotor synchronous machine (WRSM) with a
stator structure that is similar to that used on most salient-pole machines. In most
automotive applications, the stator winding is unity slot/pole/phase, single-layer that
allows the stator to be constructed in an automated and cost-effective manner.
The rotor of a claw-pole alternator is this machine’s defining feature. The name
“claw-pole” comes from the shape of the rotor pole faces, which resemble claws and as
can be seen in Figure 2.1. The vast majority of machines encountered in the literature can
be sufficiently analyzed in two dimensions because they have a consistent cross section
when traveling along the rotational axis. Due to the complex shape of the rotor, claw-pole
machines do not have a consistent cross section and so the traditional two-dimensional
analysis techniques are not sufficient to capture the detail of the magnetic flux paths.
4
Figure 2.1: Example rotor of a claw-pole machine, without field coil.
The rotor is constructed using two halves each with a set of tapered “claws.” Each half of
the rotor can be quickly constructed using an automated forging process. The main source
of rotor MMF is a single bobbin-wound concentrated coil which occupies the toroidal
cavity beneath the claws radially, with the coil’s axis coinciding with the rotational axis.
The coil can be wound on a bobbin as a separate step before being placed between the
claws. These simple components of the rotor are significantly different from the rotors of
salient-pole WRSMs which are constructed using lamination stacks, individual field coils
per pole, and damper bars. The minimal cost of construction of the claw-pole machine is
an important factor that has led to its widespread use for power generation in ground
vehicles.
5
2.2 Flux Tube Fundamentals
Magnetic Equivalent Circuit theory is based upon the flux tube which is defined as
a volume of a material or space which has magnetic flux enter (in a normal direction) at a
single surface of equal magnetic potential and leave (in a normal direction) the flux tube
volume only at a second surface of equal magnetic potential. The flux tube is analogous
to an electrical resistor. Figure 2.2 shows a simple representative cylindrical flux tube
example.
Figure 2.2: Simple cylindrical flux tube.
The flux tube parametric property is its reluctance R , which is defined as the
ratio of magnetic potential across its terminals to the flux flowing through it
1 2f fRφ−
= (2.1)
Similar to resistance, the reluctance can be defined only in terms of its dimensions and
material properties. In particular, reluctance is proportional to its length tl and inversely
proportional to its cross-sectional area tA . It is expressed as
t tt
t
lRA
ν= (2.2)
where tν is the reluctivity. It is a popular choice to instead express the relationship in
terms of the material’s magnetic permeability tµ which is the reciprocal of reluctivity
6
t
t t
lRAµ
= (2.3)
For many situations it is useful to work with a quantity that is the reciprocal of reluctance,
called the permeance
1PR
= (2.4)
t t
t
APl
µ= (2.5)
which is analogous to conductance in electrical circuits.
Figure 2.3: Flux tube with non-uniform cross-sectional area.
Flux tubes may have non-uniform cross-sectional area, as shown in Figure 2.3. If
encountered, the entire flux tube can be thought of as a collection of smaller reluctances
in series, each with its own constant cross-sectional area and length. The differential
reluctance can be expressed in terms of the differential length dx and area as a function
of position as in [2],
( ) ( )
dxdRx A xµ
= (2.6)
7
In order to determine the total reluctance of the non-uniform flux tube, the differential
reluctance is integrated which represents the series combination.
( ) ( )
dxRx A xµ
= ∫ (2.7)
Flux tubes with non-uniform length between the equipotential surfaces are often
encountered, as shown in Figure 2.4. In this case it is desired to work with differential
permeances. Each differential permeance is assumed to have constant cross-sectional area,
and the length is a function of position.
( ),
dAdPl x yµ
= (2.8)
The total permeance between the equipotential planes can then be thought of as the
parallel combination of the differential permeances. For permeances, like conductances in
electrical circuits, this amounts to a summation or rather an integration in this instance
( )( ),,A A
x y dAP dP
l x yµ
= =∫ ∫ (2.9)
Figure 2.4: Flux tube with non-uniform length.
8
As a general rule, in order to determine reluctance, an analytical expression is required
for the cross-sectional area, and for permeance an analytical expression is required for the
flux tube length.
In order to establish a magnetic potential difference across a flux tube, free
current must flow in a nearby conductor. It can be shown that conductors carrying current
can be thought of as an “MMF source,” by using Ampere’s law
enclosedH dl i⋅ =∫
(2.10)
where H
is the magnetic field intensity and enclosedi is the current enclosed by the closed
path. Windings are often composed of a collection of coils with each coil being
composed of a series of turns. If the closed path encloses all N conductors, then
Ampere’s law may be expressed as
H dl Ni⋅ =∫
(2.11)
The MMF source F in the magnetic circuit is defined as
F Ni≡ (2.12)
2.3 Stator Flux Tubes
A generic machine’s stator can be modeled using a simple magnetic circuit as
shown in Figure 2.5. Therein, a reluctance or permeance can represent each yoke, tooth
and slot leakage flux tube, while the conductors in the stator slots form a complex system
of MMF sources. The location of the MMF sources is subject to a discussion which is
detailed in section 3.1.
9
Figure 2.5: Generic stator MEC.
A list of the variables used in the derivation of stator elements is provided in Table 1
Table 2.1: Variable Identifiers and Descriptions, Stator Quantities.
Description
DBS stator yoke depth
GLS Stator stack length
SFL Stacking factor
ID Stator inner diameter
LT Length of tooth shank
OD Stator outside diameter
SL Number of stator slots
SLTINS Height of conductor above bottom of slot
STTW Stator tooth tip width
STW Stator tooth width at shank
tµ Permeability of stator tooth material
yµ Permeability of stator yoke material
0µ Permeability of free space
10
2.3.1 Yoke Flux Tube
Each yoke can be thought of as a pie-section of a toroid, in which the flux travels in the
tangential direction as shown in Figure 2.6. Since length of the flux paths is not constant,
an expression for reluctance is obtained by integrating differential permeances. Taking
the angular span as 2SLπ
, the length of each differential permeance is then 2 rSLπ
. The
differential area isGLS SFL dr⋅ ⋅ , and the permeance of the yoke can be expressed using
cylindrical coordinates by integrating along the radial direction.
Figure 2.6: Illustration of yoke flux tube.
2
2
2
OD
yy
OD DBS
P drr
SL
GLS SFLµπ
−
⋅= ∫ (2.13)
The final expression for the permeance of the stator yoke is
ln2 2
yy
GLS SFL SL ODPOD DBS
µπ⋅ ⋅ = −
(2.14)
where the magnetic permeability yµ is the product of the relative permeability of the
yoke material and the permeability of free space
11
, 0y y rµ µ µ= (2.15)
The reluctance of the yoke flux tube is then obtained using the reciprocal relationship
(2.4).
2.3.2 Tooth Flux Tube
Figure 2.7: Stator tooth.
Stator teeth, shown in Figure 2.7, provide a low-reluctance path from the yoke to
the airgap. The mean path taken by the flux includes 12
DBS for a total length
12
2OD ID DBSl − −
= . Neglecting the tooth tip, the cross-sectional area of the tooth is
GLS FSTW S L⋅ ⋅ . Using (2.3), the stator tooth reluctance is expressed as
2T
T
OD ID DBSTW SFL
SRGLSµ ⋅ ⋅
− −= (2.16)
2.3.3 Slot Leakage Flux Tube
The spaces between the stator teeth form the slots, which are cavities in which
windings are placed to produce stator MMF. Though the majority of the flux produced by
the MMF flows through the yoke and tooth, some will take a path across the stator slot
tangentially. This is referred to as leakage flux, and is modeled by placing a permeance
across the tooth tip potentials which span the slot. The leakage permeance for the coffin-
shaped slot shown in Figure 2.8 can be expressed as [6]
Figure 2.8: Coffin-shaped slot leakage permeance.
13
2 4
0
22
2 2 1 1 0log log ...2 1 1 1 0 0 0
2 1 2 2 3log4 43
2 21 1
TL
H B H B HB B B B B B B
B B BP GLS BS BS BSH
BS B BBS BS
µ
+ + + − − − − − = ⋅ − ⋅ −
(2.17)
where, in terms of the set of variables being used here,
3 0 1 22 2
OD IDH DBS SLTINS H H H = − − − + + +
(2.18)
( )21 0 12
IDB H H STWSLπ = + + −
(2.19)
( )22 0 1 22
IDB H H H STWSLπ = + + + −
(2.20)
( )2 0 1 2 32
IDBS H H H H STWSLπ = + + + + −
(2.21)
Many different tooth/slot shape combinations are encountered in practice. An example of
rectangular teeth with coffin-shaped slots has been presented here, but in reality the
shapes will vary considerably. The specific alternator used for the majority of the
validation had rather peculiar tooth/slot shape and so custom expressions were derived in
place of (2.16) and (2.17).
2.4 Rotor Flux Tubes
A simplified view of the rotor is shown in Figure 2.9 where the flux paths are
illustrated as well as the orientation of the field winding. This differs from most
synchronous machines where the rotor flux is primarily oriented in the radial and
14
tangential directions, but in the claw-pole rotor, the flux is concentrated at the axis of the
machine in the axial direction. Flux tubes and their reluctance or permeance will now be
established for the paths shown in Figure 2.9. A list of variables used in rotor derivations
is provided in Table 2.2.
Figure 2.9: Rotor Section.
Figure 2.10: Physical Rotor.
15
2.4.1 Shaft and Core Flux Tubes
Table 2.2: Variable Identifiers and Descriptions, Rotor Quantities
Description
CL core length
COILH field coil height
CSN section number
DC core diameter
DD end disk diameter
G1 main airgap
G3 gap between claw sections
GLP length of rotor pole
HPR rotor tooth height at root
HPT rotor tooth height at tip
PITCH angle of pitch in of side of claw
RID inside diameter of claw
RISE angle of rise of claw underside
COD Field coil plastic slot outer diameter
ROD outside diameter of claw
RP number of poles in machine
SD shaft diameter
SPC sections per claw
TAPER angle of claw taper from base to tip
TED end piece thickness
WPR rotor tooth width at root
WPT rotor tooth width at tip
β1 half of angular width of narrow end of claw section
β2 half of angular width of wide end of claw section
µc rotor claw permeability
µcor rotor core permeability
16
Table 2.2 Continued
µd rotor end disk permeability
µend rotor end piece permeability
µsha rotor shaft permeability
µ0 permeability of free space
Figure 2.11 shows the simple cylindrical flux tubes for the rotor’s shaft and core.
The expressions for these flux tubes are relatively straightforward. The area of the shaft is
222
SDRP
π
and the length is CL TED+ . A coefficient of 2
RP is needed in the cross-
sectional area to account for the fact that the MEC is only a model of a single pole pair of
the machine. The extra term TED in the length is to account for the extra length that the
flux travels into the end disks on either side, which is a result of applying the mean path
approximation.
( )SHA 2
cor
CL TEDRPR2 SDμ π
2
+=
(2.22)
Figure 2.11: Shaft and core flux tubes.
17
The core is a cylindrical shell of a flux tube where the machine shaft resides in the cavity
in its center. The cross-sectional area is 2 22
2 2DC SD
RPπ
− and the length is the
same as the shaft flux tube. The reluctance is then
COR 2 2
o cor
RP CL+TEDR2 DC SDμ μ π
2 2
= −
(2.23)
2.4.2 End Disk Flux Tube
Figure 2.12: End disk of rotor.
In the shaft and core, the flux is in the axial direction. In the end disk shown in
Figure 2.12, the flux changes direction and then travels radially toward the stator. Though
the transition in direction is not truly abrupt, for modeling purposes an abrupt transition is
assumed. Looking into the disk radially the cross-sectional area is 4πA(r) r TEDRP
= ⋅ ⋅ .
With a differential length dr , the end disk differential reluctance is
RPπ
18
4d
DPdrdRr TEDµ π
=⋅
(2.24)
The total reluctance is obtained by integrating (2.24) from the inner to outer radius,
2
4
4
DD
SD
d DTEDP drR
rπµ= ∫ (2.25)
ln 24D
d
P DDRTED SDπ µ
=⋅⋅
(2.26)
2.4.3 End Piece Flux Tube
The intermediate path from the end disk to the claws is referred to as the end
piece. If the pole faces are the claws, then the end pieces can be thought of as the
knuckles, as shown in Figure 2.13.
Figure 2.13: End piece, or “knuckle,” of rotor.
The flux tube representing the knuckle is modeled using two series reluctances
which are perpendicular. Using a mean path approximation argument, the flux first
19
travels a distance 1 ( )2
ROD DD HPR− − radially from the end disk and then travels
axially toward the claw a distance 12
TED . The radial flux tube has cross-sectional area
TED WPR⋅ and the axial flux tube has cross-sectional area WPR HPR⋅ .
22 END END
ENDROD DD HPR TEDR
TED WPR WPR HPRµ µ− −
= +⋅ ⋅
(2.27)
2.4.4 Claw Section Flux Tube
The claws are a rather complex portion of the MEC for several reasons. Not only
are the claws difficult to define geometrically, but also the condition that flux must only
enter or exit an equipotential surfaces is violated as the flux leaves the pole face and
traverses the airgap. In order to model the effect of flux leaving the pole face, the length
of the claw is divided into several sections, described by the parameter SPC , the
“sections per claw.” Throughout the derivation of the MEC 4SPC = is used. At each
claw section junction, a node is placed in the MEC where an airgap flux tube is
connected which provides a path to traverse the airgap. The parameters for claw width
and claw height are assumed to be parameters of a particular claw section rather than the
full length of the claw.
As shown in the side view in Figure 2.14, the claw can be thought of as a
cylindrical shell whose inner radius changes linearly along the axial direction, described
by the parameter RISE . In addition to the underside radius changing, there is a tapering
of the angular span of the claw, as shown in the top view of Figure 2.14. This is described
by the parameter TAPER .
20
Figure 2.14: Detail of rotor claws.
The intermediate parameter TAPER can be expressed in terms of the geometrical
descriptors in Table 2-2 as
1tan2
WPR WPTTAPERGLP
− − = ⋅ (2.28)
Using a similar geometrical argument, the intermediate parameter RISE can be
expressed as
1tan HPR HPTRISEGLP
− − =
(2.29)
Noticing that the cross-sectional area of the claw section flux tube varies along the length
of the machine, the approach taken is to use differential reluctances as before
( )c
dzdRA zµ
= (2.30)
21
where the cross-section area of the claw as a function of position z down the axis of the
machine can be thought of as the difference in area of an outer and inner pie-section,
( ) ( ) (z)outer innerA z A z A= − , as detailed in the back view of Figure 2.14. The areas of the
pie-sections are functions of the angular span and radius. The radius of the inner pie-
section as well as the angular span of each section are also functions of z . First, ( )outerA z
can be expressed in terms of a circular area and a proportion of angular span to the whole
circle,
2
( ( ))2outer
RODA z zβππ
=
(2.31)
where ( )zβ can be expressed as the linear transition from 2β to 1β over the distance
/GLP SPC
2( )z m zβ β= ⋅ + (2.32)
and ( )1 2 SPCm
GLPβ β− ⋅
= , the slope of the linear transition. A similar approach is taken
for the inner pie-section, but it must be considered that the inner radius changes with
distance down the z axis. The claw section index CSN is included in the expression to
account for distance in the z direction that is taken up by claw sections preceding the
claw section of interest
( ) ( ) ( )2
1( ) 2
2inner
CSN GLP RIDA z m z z TAN RISESPC
β − ⋅
= ⋅ + ⋅ + ⋅ + (2.33)
where m is the same slope as before. Combining (2.30) – (2.33) yields
22
( ) ( ) ( )22 1
22 2o c
dzdRCSN GLPROD RIDm z z TAN RISE
SPCµ µ β
= − ⋅ ⋅ + ⋅ − + ⋅ +
(2.34)
In the MATLAB code implementation, (2.34) is integrated numerically to obtain the final
reluctance.
2.4.5 Rotor Leakage Flux Tubes
Some of the flux through the core and shaft does not go to the claws/airgap and
instead takes a path back through the space occupied by the field winding beneath the
claws. This path is herein referred to as the field leakage path. The geometry of the field
leakage path is shown in Figure 2.15.
Figure 2.15: Rotor field leakage path detail.
The field leakage path is modeled by two parallel paths. Flux that travels through
the winding is partially-linked flux and must be treated with an energy-based approach of
determining leakage permeances as detailed in [7]. Specifically, the energy is first
determined using the volume integral
23
212 V
E H dvµ= ∫ (2.35)
Subsequently another expression for energy
2 212
E N i P= (2.36)
is equated to (2.35) and then the permeance P determined. To do so, Ampere’s law is
applied along the core and back through the field winding, the same path the flux takes in
Figure 2.15. The H
field in the magnetic material is neglected as the MMF drop is
expected to be small there, and the current enclosed is then a function of radius. When
2 2CID CIDr COILH≤ ≤ + ,
2( ) CLfd
z
CIDTRC i rH r
COILH
⋅ − ⋅ = (2.37)
and when 2
CIDr COILH≥ + , the winding is fully-linked and ( ) CLz fdH r TRC i⋅ = ⋅ .
Substituting (2.37) into (2.35),
2
1 22
fdCIDTRC i r
CLE drCL COILH RP
π ⋅ − =
⋅
⋅
∫ (2.38)
where RP is used to represent the pole count so as to avoid mistaking it for permeance.
Evaluating the integral over 2
CIDr = to2
CID COILH+ yields
( )2 021 2 32 3fd
COILHE TRC i CID COILHCL RP
πµ ⋅= +
⋅ ⋅ (2.39)
Equating (2.39) to (2.36) and cancelling terms,
24
( )0 2 33coil
COILHP CID COILHCL RP
πµ ⋅= +
⋅ ⋅ (2.40)
The flux also occupies the space between the field coil and the underside of the claws,
where the permeance is that of a cylindrical shell
( )( )22
0AIR
RID CID 2COILHπ μP2 RP CL
− +⋅=
⋅ (2.41)
Then the parallel combination of these two permeances is
FLD COIL AIRP P P= + (2.42)
Magnetic flux will also leak from the claw on one rotor half to the adjacent claw
on the other rotor half, and it will do this on both sides. To model this path, it is assumed
that the radial component of the permeance cross-section uses the average of the claw
underside radius as denoted by _under rad in Figure 2.16.
1_2 2 2
RID RODunder rad HPT = + − (2.43)
( )1GLPh
COS TAPER= (2.44)
Substituting (2.44) into into (2.5) and noting the cross-sectional area is the product of 1h
and _2
R radOD under−
, the leakage permeance from claw to claw is
( )
0 _2
3CL
RODGLP under radP
COS TAPER SPC G
µ ⋅ ⋅ − =
⋅ ⋅ (2.45)
where 3G is the distance from claw to claw perpendicular to the claw edge, which can be
expressed as
25
( ) ( )WPR WPT2π RODG3 COS TAPERRP 2 2
− = − ⋅
(2.46)
Figure 2.16: Detail of the claw-claw leakage permeance.
2.4.6 Permanent Magnet
A second source of rotor MMF is utilized in some alternators in order to reduce
the claw-to-claw leakage flux. The direction is critical in that the magnets are oriented to
oppose the leakage flux that would occur with positive field current. If permanent
magnets are used in the alternator, the permanent magnet’s MEC in Figure 2.17 replaces
the leakage permeance (2.45).
0PM
PMH PMLPPMW SPC
µ ⋅ ⋅=
⋅ (2.47)
PM PMPMC PMW P= ⋅ ⋅Φ (2.48)
26
Figure 2.17: Permanent magnet in between claws.
2.5 Air Gap Flux Tubes
To model the complex collection of airgap flux tubes, it is assumed that a flux tube
is formed between the rotor and stator tooth only when there is some amount of angular
overlap between the two, and if there is not an angular overlap logic must be used to
determine if the rotor pole face is near the stator tooth, in which case a fringing flux tube
is described. If the rotor pole face is not near the stator tooth, then the airgap flux tube’s
permeance is zero. More specifically, when any portion of the claw is within the angular
range of the tooth tip’s edges, there is a direct path of length 1G , and on the approach to
angular overlap there is a more complex fringing path. An additional feature of the
airgap model is the representation of a chamfer along the tapered edge of the claw which
is a common feature in existing alternators as a means to control electrical and torque
harmonics and thus audible noise. A claw face with a trailing edge chamfer is shown in
27
Figure 2.18 where Region 4 represents the chamfer. This chamfer introduces an
additional airgap that is non-constant and must be dealt with accordingly. As described
earlier, the claw face area is divided into sections down the length of the machine’s axis
described by the parameter SPC . Table 2-3 lists the parameters used to describe the
airgap flux tubes along with descriptions of each variable.
φ
θφs
r
r
−αα
Tooth
GLPRe
gion
1 Region 2 Region 3Region 4
CW
z
l( )φr
β1β2 −β1 −β2
Figure 2.18: Formation of airgap flux tubes.
In Figure 2.18 the variable sφ is an angular location on the stator relative to a
reference axis that is fixed on the stator. Similarly, rφ is the angular location on the rotor
relative to an axis that is fixed on the rotor. For a chosen angular location, the two angles
can be related by
rr sφ θφ= − (2.49)
where rθ is the angle between the axis fixed on the stator and the axis fixed on the rotor.
28
Table 2.3: Variable Identifiers and Descriptions, Air Gap Quantities
Description
G1 main airgap length
GLP length of rotor pole
ID inside radius of stator stack
CD chamfer depth
CW chamfer width
β1 half of angular width of narrow end of claw section
β2 half of angular width of wide end of claw section
µ0 permeability of free space
θr mechanical position of rotor
From Figure 2.18 the angular location of the tooth tips in rotor angles are
( )r rα θφ α− = − − (2.50)
( )r rα θφ α= − (2.51)
The length of the overlapping edge can be expressed as
( ) 2r r
1 2 1 2
β GLPGLPβ β β β
l φ φ ⋅= −
− − (2.52)
When the tooth tip’s edge ( )α− begins entering the Region 1, i.e. when 1 ( ) 2β βr αφ −≤ ≤
and ( ) 2βr αφ ≥ , then the area of overlap is expressed as
( )2
r
β
r rα θ
A ID dl φ φ− −
= ∫ (2.53)
To consider the next two cases, it is noted that for some designs the total width of the
tooth tip can exceed the width of Region 1, or the opposite can occur – the width of
29
Region 1 can exceed the width of the stator tooth tip. When Region 1 spans the entirety
of the tooth width, or when 1 ( ) 2β βr αφ −≤ ≤ and 1 ( ) 2β βr αφ≤ ≤ , the area of overlap is
( )r
r
α θ
r rα θ
A ID dl φ φ−
− −
= ∫ (2.54)
When the tooth tip spans the entirety of Region 1, or when ( ) 1βr αφ − ≤ and ( ) 2βr αφ ≥ , the
area of overlap can be expressed as
( )2
1
β
r rβ
A ID dl φ φ= ∫ (2.55)
The next case that must be considered is for when the tooth is transitioning out of Region
1. This is described by the inequalities ( ) 1r αφ β− ≤ and 1 ( ) 2r αβ βφ≤ ≤ . In this situation, the
area of overlap is expressed as
( ) ( )( )r
1
α θ
r r 1 r 1 2 r2 1β
ID GLPA ID d β α θ α β 2β θ2 β β
l φ φ−
= = − + + − −−∫ (2.56)
A similar analysis can be applied to the remaining regions 2 and 3. The details and results
are provided in Appendix A.
Now the chamfer region with non-constant airgap is analyzed which is Region 4
in Figure 2.18. The parameter CW describes the angular width of the chamfer along the
tip and base of the tooth. Two cases arise when dealing with the chamfer, as shown in
Figure 2.19. On the left side of Figure 2.19, the corner point of the claw tip, described by
the angle 1β− , when projected onto the base is outside the angular span CW at the base
of the claw. The other case is illustrated on the right side of Figure 2.19 where the
projection of the corner point is within the angular span CW at the base of the claw. More
compactly, the inequality for the left side of Figure 2.19 is 1 2 CWβ β− ≥ − + and the
30
inequality is 1 2 CWβ β− ≤ − + for the right side. The chamfer analysis is further broken
down into three regions as shown in Figure 2.19.
GLPGLPCWCW
zz
φrφr
1 1
2 2
3 3
z1
z2 z2
z1
−β2−β1 −β2−β1
Figure 2.19: Two cases for determining airgap length over the chamfer region.
The airgap length in the chamfer region is varying with respect to position, therefore the
total permeance over the region can be determined by integrating differential permeances
where airgap length is a function of position. Using cylindrical coordinates, the
permeance of the chamfer region is
( )
rchamfer 0
r
dzdIDP μ2 g , z
φφ
= ∫∫ (2.57)
where ( , )rg zφ is the airgap length as a function of position. The chamfer is described
geometrically as a plane which contains the points ( ,, )g zφ : ( )21, ,0G CWβ− + ,
( )11, ,G CW GLPβ− + and ( )11 , ,G CD GLPβ+ − . An expression for g is then
( ) ( )2 1 2r r
CD CDCDg ,z z CD G1CW GLP CW CW
β β βφ φ− ⋅
= − + + −⋅
(2.58)
An expression for the line 1z which is the edge of the pole face where 1g G= is
( ) ( )1 r r 22 1
GLPz CWφ φ ββ β
= + −−
(2.59)
31
and 2z where 1g G CD= + is
( ) ( )22
2 r r1
GLPz φ φ ββ β
= +−
(2.60)
These z lines are used to bound the integration (2.57). Again with a case by case
treatment, the various expressions for airgap permeance can be expressed. When the
stator tooth is transitioning into Region 1 for both 1 2 CWβ β− ≥ − + and 1 2 CWβ β− +≤ −
(both left and right sides of Figure 2.19), the airgap permeance of the chamfer region is
( )
1
r 1
CW GLPr
chamfer 0rα θ z
dzdIDP μ2 g ,z
β φφ
− +
− −
= ∫ ∫ (2.61)
As before, in some designs the stator tooth width can entirely span the region or the
region can entirely span the tooth. When the tooth width is larger than the width of
Region 1, or 1 rβ α θ− ≥ − − and 1 rCWβ α θ− + ≤ −
( )
1 1
1 CW GLPr
chamfer 0rz
dzdIDP μ2 g ,z
β
β
φφ
− +
−
= ∫ ∫ (2.62)
When the tooth width is smaller than the width of Region 1, or
1 1r CWβ α θ β≤ − ≤ − +− − and 1 1r CWβ α θ β− ≤ ≤ − +− ,
( )
1
GLPr
chamfer 0rz
dzdIDP μ2 g ,z
r
r
α θ
α θ
φφ
+
− +
= ∫ ∫ (2.63)
When the tooth is transitioning out of Region 1, the limits of integration are again
changed accordingly and the airgap permeance is
( )
1 1
GLPr
chamfer 0rz
dzdIDP μ2 g ,z
rα θ
β
φφ−
+
= ∫ ∫ (2.64)
32
The same analysis techniques can be used to determine the airgap permeance over the
chamfer for Regions 2 and 3 for both 1 2 CWβ β− ≥ − + and 1 2 CWβ β− +≤ − (both left
and right sides of Figure 2.19). The results of all integrations in this section are
summarized in Appendix A.
A simple means of approximating fringing effects has been used in the MEC
whereby the airgap 1G is lengthened by Carter’s coefficient and the tooth width is
lengthened to account for fringing. Carter’s coefficient is
4 1 ln 1
4 1
ss sts
ssst
G ww wc
wG
ππ
+=
+ + ⋅
⋅ (2.65)
where 0ssw b= is the width of the slot, 0b is an input dimension of the machine and
0stID bwSL
π= − is the tooth tip width. The modified airgap 1 1G c Gs= ⋅′ is used in place of
1G in all airgap permeance expressions. To compensate for the lengthened airgap, a
modified tooth width is also used so that the fringing path begins as soon as the edge of
the rotor passes into the halfway-point of the slot width. Specifically the tooth tip width
α replaced by 'SLπα = in the airgap permeance expressions.
33
ALGEBRAIC SYSTEM OF MAGNETIC CIRCUIT EQUATIONS
Using the expressions for each reluctance or permeance, a system of equations is
formed and a solution method is presented. Within this research, nodal analysis is
employed in the airgap, where permeances will become zero-valued, and mesh analysis is
used elsewhere in the machine. Herein, a loop flux emanates from a node which is
nonlinear material and terminates when it encounters an air path.
3.1 Stator Teeth Mesh Equations
Referring to Figure 3.1, traversing the loop of mesh flux 1LΦ , the mesh equation
for the first stator slot can be expressed as
1 1 2 2 1 2 1 12 2 1 2 1( ) ( ) 0T L L T L L L y T TR R R f f F F− + − + + − + −Φ Φ Φ Φ =Φ (3.1)
The other 11 loop equations for the remaining teeth are similar. The MMF sources in
teeth 1 and 2 can be related to the total current in the slot by
2 1asTC i F F⋅ = − (3.2)
The turns count is the variable TC and is common to all windings. Similar equations can
be written for the remaining slots. In matrix form, the entire set of stator mesh equations
is
(12,12) (12,1) (12,12) (12,1) (12,6) (6,1)LT T LL L IP abcxyzs+ + = 0A f A A iΦ (3.3)
All matrices are expanded in Appendix A.
34
Figure 3.1: MEC structure detailing the discretization of the airgap and stator into respective flux tubes.
3.2 Nodal Analysis at Stator Tooth Tip Nodes
Summing all flux entering the node at stator tooth tip 1, including connections to all
the claw section nodes (10 claw section nodes when 4SPC = ), though most of these
permeances will be zero-valued for a given rotor position indicating there is not a path for
them at this rotor position, ones can use KCL to express
6 1 2 1 12 1
10 1 1 10 9 1 1 9 8 1 1 8 7 1 1 7 6 1 1 6
5 1 1 5 4 1 1 4 3 1 1 3 2 1 1 2 1 1 1 1
) )) ) ) ) )
) ) ) )
( (( ( (
)( (
( ( ( ( (0
L L T T TL T T TL
C T T C C T T C C T T C C T T C C T T C
C T T C C T T C C T T C C T T C C T T C
f P f Pf
f ff f f f ff f f f
P f P f P f P f Pf P f P f P f P f Pf
+ ++ + + + ++ +
Φ − Φ − −− − − − −
− − − −+ + +=
− (3.4)
The remaining 12 stator tooth tips have an identical form. Placing all of them together
yields a matrix expression,
35
(12,12) (12,1) (12,10) (10,1) TL(12,12) (12,1)TT T TC C L+ + = 0A f A f A Φ (3.5)
For the purposes of determining magnitude of flux density in each tooth, which is
required information for determining reluctance, the flux in each tooth is the difference of
the two loop fluxes present in a particular tooth, since they are opposite in direction. For
example, the flux density in tooth 1 is
1 1 12T L LΦ = Φ − Φ (3.6)
and the remaining teeth are similar.
3.3 Mesh/Nodal Analysis at Claw Section Nodes
Since the rotor’s interaction with the airgap is at the nodes between each claw
section, nodal analysis is employed. As an example, claw nodes 1 and 2 are similar in
form,
13 1 1 1 1 2 1 2 1 3 1 3 1 4 1 4 1
5 1 5 1 6 1 6 1 7 1 7 1 8 1 8 1
9 1 9 1 10 1 10 1 11 1 11 1 12 1 12 1
10
( ) P ( ) P ( ) P ( ) P( ) P ( ) P ( ) P ( ) P( ) P ( ) P ( ) P ( ) P2(
CS T C T C T C T C T C T C T C T C
T C T C T C T C T C T C T C T C
T C T C T C T C T C T C T C T C
C
f f f f f f f ff f f f f f f ff f f f f f f ff f
+ − + − + − + −
+ − + − + − + −
+ − + − + − + −
+
Φ
− 1 1,10)C CL
PM
P= Φ−
(3.7)
having only one mesh flux as they are on the tips of the claws (most narrow portions).
The inner claw section nodes have an additional mesh flux, for example, the claw section
node at potential 3Cf can be expressed as
35 13 1 3 1 3 2 3 2 3 3 3 3 3 4 3 4 3
5 3 5 3 6 3 6 3 7 3 7 3 8 3 8 3
9 3 9 3 10 3 10 3 11 3 11 3 12 3 12 3
( ) P ( ) P ( ) P ( ) P( ) P ( ) P ( ) P ( ) P( ) P ( ) P ( ) P ( ) P2(
CS CS T C T C T C T C T C T C T C T C
T C T C T C T C T C T C T C T C
T C T C T C T C T C T C T C T C
f f f f f f f ff f f f f f f ff f f f f f f f
Φ − Φ + − + − + − + −
+ − + − + − + −
+ − + − + − + −
+ 8 3 3,8)C C CL
PM
f Pf= Φ−
− (3.8)
36
The remaining nodes take on a similar form, although the nodes at potentials 9Cf and
10Cf have an additional ENDΦ term from the knuckles. In matrix form, the set of 10
(when 4SPC = ) equations is
(10,12) (12,1) (10,10) (10,1) (10,1) (1,1) (10,8) (8,1) (10,17) (17,1)CT T CC C CE END CS CS PM+ + + = CA f A f A A bΦ Φ (3.9)
where PMC is a connection matrix that relates the magnetic circuit equations to system
inputs in the vector b , such as permanent magnet flux sources. If permanent magnets are
present between the claws, the leakage permeances CLP are replaced with (2.47).
Figure 3.2: Discretization of claws into sections.
In addition to the set of nodal equations for the claw section nodes, a set of mesh
equations is needed to solve for the mesh fluxes 1,3CSΦ – 8,10CSΦ . Referring to Figure 3.2,
for example, the flux in the claw section between nodes at potentials 1Cf and 3Cf is
1,3 1,3 3 1( ) 0CS CS C CR f f− − =Φ (3.10)
In matrix form the set of mesh fluxes can be expressed as
10Cf
CLP
8Cf
6Cf
4Cf
2Cf9Cf
7Cf
5Cf
3Cf
1Cf
7,9CSR
5,7CSR
3,5CSR
1,3CSR8,10CSR
6,8CSR
4,6CSR
2,4CSR1,3CSΦ
3,5CSΦ
5,7CSΦ
7,9CSΦ
9Cf
7Cf
5Cf
3Cf
1Cf
2,4CSΦ
4,6CSΦ
6,8CSΦ
8,10CSΦ
37
FS(8,10) (10,1) (8,8) (8,1)C FD CS+ = 0A f A Φ (3.11)
3.4 Relating Loop Flux to Phase Flux
The state variables of the system are the flux linkages of each winding. Phase flux
linkage is related to the phase flux by a pole count factor 2P
and a turns ratio for an
integer slot/pole/phase winding scenario. The tooth fluxes constituting a phase flux are
the sum of the teeth which are spanned by the respective winding. Subsequently, since
the a phase conductors span teeth 2-7, the phase flux is related to stator tooth fluxes by
, 2 3 4 5 6 7as p T T T T T TΦ Φ Φ Φ= + + + Φ +Φ Φ+ (3.12)
where the p denotes the flux in a pole pair. Using the relationship between tooth and
mesh flux, as in (3.6), for example in the a-phase
, 7 1as p L LΦ Φ Φ= − (3.13)
In matrix form,
(6,12) (12,1) (6,17) (17,1)FF L F=A C bΦ (3.14)
where the phase fluxes are members of the input vector b .
3.5 Mesh Analysis of Rotor
Mesh analysis is employed in the rotor of the alternator. This is accomplished using
KVL loops in each of the three circuital paths shown in Figure 3.3. The end disk DR is
divided into two series reluctances, an inner and outer disk, denoted DIR and DOR
respectively. Correspondingly, the limits of integration in (2.25) are adjusted and the
38
integral divided into two parts. The inner disk is integrated from 4
SD to
4SD DC+
and
the outer is integrated from 4
SD DC+ to
2DD
.
Figure 3.3: MEC structure detailing discretization of rotor into respective flux tubes.
9 102 ( ) 0END END END FD FD C CR R f f+ − + −Φ Φ =Φ (3.15)
( ) (2 )FD DO FD END FD FD SHA COR fdTRC iR R RΦ + Φ − Φ Φ = ⋅− Φ+ (3.16)
( )2 0SHA DI SHA RHA SHA FD CORR R RΦ + Φ + Φ − Φ = (3.17)
In matrix form, the three mesh equations describing the rotor can be placed into the
tableau
[ ]0
0
c
ENDRFC REND RFD RS fd
FD
SHA
f
A A A A TRC i
= ⋅
ΦΦΦ
(3.18)
ENDΦ
FDΦ
SHAΦ
9Cf 10Cf
SHAR
CORR
DIR
DOR
ENDR
FDR
39
3.6 Overall Algebraic System
Each subset of magnetic circuit equations can be combined into an overall system
of the form =Ax Cb where
(12,10) (12,1) (12,1) (12,12) (12,8) (12,6) (12,1)
(12,10) (12,1) (12,1) (12,12) (12,8) (12,6) (12,1)
(12,12)
(
(10,12) (10,1) (10,1) (10,12) (10,8) (10,6) (10,1)
12,12)
(10,10)
LT
TT
CC
LL IP
TC TL
CT CE CS=
0 0 0 0 00 0 0 0 0
0 0 0 00
A A AA A AA A A A
A(8,12) (8,10) (8,1) (8,1) (8,12) (8,8) (8,6) (8,1)
(6,12) (6,10) (6,1) (6,1) (6,12) (6,8) (6,6) (6,1)
(3,12) (3,10) (3,1) (3,1) (3,12) (3,8) (3,6) (3,1)
FS FD
FF
RFC REND RS RFD
0 0 0 0 00 0 0 0 0 0 00 0 0 0
A AA
A A A A
(3.19)
(1,12) (1,10) (1,1) (1,1) (1,12) (1,8) (1,6) (1,1)
T
T C END S L CS abcxyzs fdiΦ Φ ix = f f Φ Φ (3.20)
(17,12) (17,12) PM(17,10) (17,8) (17,6) (17,3)
T
F RFD = 0 0 0C C C C (3.21)
(1,10) (1,6) (1,1)
T
PM abcxyzs fd Φ b = Φ Φ (3.22)
When solving circuits in terms of node potentials, the potential is always relative
to the user-selected reference node. In this research, the reference node was chosen to be
stator tooth tip 1, which has potential 1Tf . To implement this, the equation corresponding
to the row which solves for 1Tf is deleted, and the variable is removed from (3.20).
When developing the system of equations for a wye-connected stator, it was noted
that one must account for the physical constraint that the sum of the currents in each wye
is zero. Therefore, the third is redundant, and the stator teeth mesh equations and the
relationship between loop and phase flux was altered to eliminate csi and zsi from the
system. This modification affects matrices IPA , FFA and FC . Details are provided in
40
Appendix A for the expanded matrices, including changes for wye and delta-connected
stators.
3.7 Solution of Nonlinear Algebraic System
From the MEC algebraic system, it is desired to solve for the vector x in
=Ax Cb . At first glance, the answer is simply 1−=x A Cb but the matrix A is a function
of x . More specifically, in each nonlinear flux tube, the relative permeability is a
function of the average flux density in each flux tube. The well-known Newton-Raphson
method is used to solve the nonlinear system of equations
) ( )( = −A xx xf Cb (3.23)
for the roots x . The Newton-Raphson iterator is
( ) ( ) ( ) ( )11 )( ( )i i i i−+ − =x x J x f x (3.24)
where ( ) )( iJ x is the Jacobian matrix of f and the superscript i denotes the iteration
number. In this research ( )J x is approximated as ( )A x , as was done in [3], for two
reasons. First, the nodal system which used this approximation in [3] converged in more
cases than with the full analytical Jacobian matrix. Second, less computation is needed to
express the approximation than the full Jacobian matrix, speeding up the time to solution.
As verification, a numerically-computed full Jacobian matrix was tested and compared to
(( ) )= AJ xx , and the numerically obtained Jacobian matrix did not improve
convergence. To further ensure convergence, the Newton-Raphson method is relaxed by
the constant α , 0 1α≤ ≤ and the iterator becomes
( ) ( ) ( ) ( )11 ( ) )(i i i iα−+ = − x x J x f x (3.25)
41
To initiate the Newton-Raphson method, initial values of phase and field flux linkages
are required which are user-selected. From the flux linkages, magnetic flux values are
determined and fed into the MEC algebraic system as the input vector b (3.22). The
Newton-Raphson method is then iterated until error criteria are met. Specifically, since
the members of x are MMF, current and flux as can be seen from (3.20), three error
tolerances are checked corresponding to the three types of variables Fe , eΦ and Ie
where
( ) ( ) ( ) ( )1 1
2 2
i i i iF r aF F F Fe x x K x x K+ += − − + − (3.26)
( ) ( ) ( ) ( )1 1
2 2r ai i i ie x x K x x K+
Φ Φ Φ+
Φ Φ= − − + − (3.27)
( ) ( ) ( ) ( )1 1
2 2
i i i iI I r I I aIe x x K x x K+ += − − + − (3.28)
This is similar to the nonlinear solver in [7]. The Newton-Raphson solver stops iterating
when each respective error is less than zero. rK and aK are small constants that describe
the relative and absolute error tolerance respectively. In other words, the iteration stops
when the solution difference between successive iterations is small relative to the
magnitude of the variables and relative to the absolute error.
42
MODELS FOR DIFFERENT LOADING CONDITIONS
In this chapter, several model structures are developed to enable coupling of
common loads to the MEC algebraic system. First a passive rectifier state model is
presented, next a model for stator open-circuit conditions is presented and then a
simplified way of simulating an active rectifier is presented without the need to model
individual switch states.
4.1 Passive Rectifier State Model
For the six phase machine being modeled, the first three phase set is labeled abc ,
which is connected in a wye configuration, and the second three-phase set is labelled
xyz and is also in a wye configuration, but it is displaced physically in the stator by 30°
electrically, or by one slot. The neutral points of the wye terminations are not connected.
Figure 4.1 illustrates the machine connection to the rectifier. Faraday’s law is used as the
starting point for the state model derivation as in [8]
abcxyzs s abcxyzs abcxyzsr p= +v i λ (4.1)
fd fd fdv r i= (4.2)
where dpdt
= .
43
Figure 4.1: Passive rectifier connection to machine.
Rearranging (4.1) – (4.2) and expanding,
as as s asv rp iλ = − (4.3)
bs bs s bsv rp iλ = − (4.4)
cs cs s csv rp iλ = − (4.5)
xs xs s xsv rp iλ = − (4.6)
ys ys s ysv rp iλ = − (4.7)
zs zs s zsv rp iλ = − (4.8)
fd fd fd fdv rp iλ = − (4.9)
fdv
Load
aibi
ci
dci
dcv
+
−
bsv
asv
csv
+
−
+
− +
−
+
−
+
−
+−xsvysv
zsv
xsi
ysizsi
g
g
agv+
−
bgv+
−
cgv+
−
xgv+
−
ygv+
−
zgv+
−
n
n
44
r rpθ ω= (4.10)
State models typically take the form = +x Ax Bu where x is the vector of state
variables and u is the vector of system inputs. Since the right hand sides of (4.3) –(4.10)
are not in terms of the variables being differentiated on the left-hand side, the model is in
a “quasi-state” model form. The quasi-state model could consist of (4.3) – (4.10) , but as
will be shown, the state model must be solved in the arbitrary reference frame in order to
avoid making an assumption about the stator voltages.
After the Newton-Raphson iteration has converged to a solution, the stator
currents, members of x (3.20), are input into the rectifier model, which determines
terminal voltages on the stator. The diode bridge model is illustrated in Figure 4.2. The
diode voltage as a function of diode current ( )D Dv i is modeled using the Shockley diode
equation which is solved for voltage, namely
1 n( ) l 1DD D
iv iD Dβ α
= +
(4.11)
where the diode parameters Dα and Dβ are selected for the particular diode being
modeled.
vag
vdc/2
v + vdc d
ialε−ε-vd
Figure 4.2: Lower diode voltage.
45
When the phase current Xsi is greater than , the upper diode is conducting and
the lower diode voltage ( )Xg dc D Xsv v v i= + , where [ ]X abcxyz∈ . If the phase current is in
the range [ ],− , a simple transitioning between the upper diode in full conduction and
the lower diode in full conduction is assumed, where the lower diode voltage is the linear
interpolant of the points ( )(, )Dv− and ( )(, )dc Dv v+ . When the phase current is less
than − , the lower diode is fully conducting and its voltage Xgv is ( )D Xsv i− . Once all
the lower diode voltages are established, a relationship to the phase voltages can be
derived. The analysis is performed on the first three phase set abc and the same analysis
is applied to the second three phase set xyz . Referring to Figure 4.1, KVL loops can be
made through the lower diode, phase winding and across potential ngv .
0as ag ngv v v− + = (4.12)
0bs bg ngv v v− + = (4.13)
0cs cg ngv v v− + = (4.14)
Summing (4.12)-(4.14) yields
( ) ( )1 13 3ag bg cg as bs cng sv v v v v v v+ + − + += (4.15)
At this point, it is common to make the assumption that the sum of the phase voltages
will be zero in each wye connection, however back-emf harmonics exist in machines with
unity slot/pole/phase windings and this causes the instantaneous sum of the phase
voltages in each three phase set to be nonzero. Substituting (4.15) into (4.12) - (4.14) and
rearranging yields
46
( ) ( )1 13 3as as bs cs ag ag bg cgv v v v v v v v− = + +−+ + (4.16)
( ) ( )1 13 3bs as bs cs bg ag bg cgv v v v v v v v− = + +−+ + (4.17)
( ) ( )1 13 3cs as bs cs cg ag bg cgv v v v v v v v− = + +−+ + (4.18)
Representing (4.16) - (4.18) in matrix form,
abcs abcg=Av Av (4.19)
where
2 1 1
1 1 2 13
1 1 2
− − = − − − −
A (4.20)
Substituting the subset abc of (4.1) into (4.19),
s as as
s bs bs abcg
s cs cs
r i pr i pr i p
λλλ
+ + = +
A Av (4.21)
abcs abcg s abcsp v r= −A A Aiλ (4.22)
In order to use this relationship in a state model, it is necessary to solve for each time-
derivative of flux linkage. Since A is not invertible, this cannot be done directly. When a
reference frame transformation is applied to (4.22), the differential equation then
becomes
[ ]( ) [ ]( )1 10 0s qd s abcg s s qd sp r− −= −A K Av A K iλ (4.23)
where sK is the arbitrary reference frame transformation matrix.
47
( )
( )
2 23 3
2 2
cos cos cos
2 sin sin sin3
1 1 12 2 2
3 3s
π πθ θ θ
π πθ θ θ
=
− +
+
−K (4.24)
For the second three phase set the analysis is similar except that the transformation matrix
takes on a 30° phase shift. Multiplying (4.23) by sK yields
[ ]( ) [ ]( )1 10 0s s qd s s abcg s s s qd sp r− −= −K A K A A K iK v Kλ (4.25)
Expanding the left-hand side of (4.25) yields
[ ]( ) [ ] ( ) [ ]( )1 1 10 0 0s s qd s s s qd s s abcg s s s qd sp p r− − −+ = −K A K K A K K K A K iAvλ λ (4.26)
where
[ ]( )10 1 01 0 00 0 0
s sp ω− = −
K A K (4.27)
[ ] 11 0 00 1 00 0 0
s s−
=
K A K (4.28)
Substituting (4.27) and (4.28) into (4.26), the time-derivatives of flux-linkages can be
solved for in the arbitrary reference frame
1 0 00 1 0
dsqs qss abcg s
qsds ds
ip r
iλλ
ωλλ
= −
− +
K Av (4.29)
The 2 3× matrix multiplying the voltage term extracts the upper two rows from [ ]sK A .
Electrical dynamics are simulated by integrating the quasi-state model derivatives in
(4.29) and (4.9) using the Forward Euler method. Mechanical dynamics are not modeled;
48
the model simulates constant rotor speed and the rotor position is updated using (4.10)
and the Forward Euler method. With (4.29), the Forward Euler update equations for qsλ
and dsλ are
( ) ( ) ( )1k k kqs qs qsp tλ λ λ+ = + ∆ (4.30)
( ) ( ) ( )1k k kds ds dsp tλ λ λ+ = + ∆ (4.31)
where the superscript k denotes the time step in the simulation. Since the bottom row of
(4.28) is composed of zeros, there is no explicit expression for 0spλ and it is not included
in the state model. From (4.30) and (4.31) the inverse reference frame transformation can
be applied.
( ) [ ]
( )
( )
( )
1
11 1
0
kqs
k kabcs s ds
ks
λ
λ
λ
+
−+ +
=
Kλ (4.32)
The input to the MEC (3.22) requires updated phase flux values which are obtained from
(4.32)
( ) ( )1 11k kabcs abcsTC
+ += λΦ (4.33)
Note that an approximation is made in (4.32) where 0sλ is obtained from the solution to
the Newton-Raphson method from time step k
( ) ( ) ( ) ( )( )013
k k k ks as bs csTCλ = Φ + Φ + Φ (4.34)
With updated phase fluxes, the next simulation time step is computed. The block diagram
in Figure 4.3 illustrates the overall solution procedure. Simulation results are presented in
the next chapter. The simulation output is compared to data from lab experiments.
49
Figure 4.3: Solution procedure block diagram.
4.2 Open Circuit Model
Open circuit modeling of the machine does not require the solution of a state
model, but solves a similar nonlinear MEC algebraic system as before using the Newton-
Raphson method. Again the overall system is described by some =Ax Cb . In open-circuit
conditions the stator current is zero and thus the algebraic system of equations established
in section 3.6 still holds, but some modifications must be made. Terms involving stator
current are removed as well as the set of equations relating loop to phase flux (3.14). This
corresponds to column 7 and row 5 being removed from (3.19).
MEC/Newton-Raphson
,TC TRC÷ ,TC TRC÷
,abcxyzs fdΦΦ
Rectifier Model
,abcxyzs fdi i
abcxyzgv
(4.29)
Forward Euler Integration
0qd spλ
sK
0qd sλ
[ ] 1s
−K
,abcxyzs fdλλ Initial Guess
0qd sλabcxyzsλ
abcxyzsifdi
0qd si
abcxyzsλ
50
(12,10) (12,1) (12,1) (12,12) (12,8) (12,1)
(12,10) (12,1) (12,1) (12,12
(12,12)
(12,12)
(10
) (12,8) (12,1)
(10,12) (10,1) (10,1) (10,12) (10,8) (10,1)
(8,12)
,
(8,10) (8,1
1
)
0)
(
LL
TC TL
CT
LT
TT
CE CS
FS
CC=
0 0 0 0 00 0 0 0
0 0 00 0 0
A AA A AA A A AA
A 8,1) (8,12) (8,8) (8,1)
(3,12) (3,10) (3,1) (3,1) (3,12) (3,8) (3,1)
FD
RFC REND RS RFD
0 00 0 0
AA A A A
(4.35)
Since stator currents no longer need to be solved for, they are removed from the solution
vector. Additionally, the flux through the field winding is added to the solution vector.
(1,12) (1,10) (1,1) (1,1) (1,12) (1,8) (1,1)
T
T C END S L CS fd Φ Φ Φx = f f Φ Φ (4.36)
The matrix (3.21) is now
(11,12) (11,12) (11,10) (11,8) (11,3)
T
PM RFD = 0 0 C 0C C (4.37)
The input to the system is now field current, the sole excitation of the system, thus (3.22)
becomes
(1,10) (1,1)
T
PM fd= i b Φ (4.38)
as the phase fluxes abcxyzsΦ are no longer driven by state of the system. When the
Newton-Raphson solver reaches a solution, the mesh flux vector LΦ is extracted from x ,
then the relationship between phase flux and mesh flux in the stator teeth (3.14) is applied
to yield phase flux abcxyzsΦ . Multiplying the phase flux by the stator turn count TC will
yield a set of stator flux linkages. The phase voltages are then
abcxyzs abcxyzsp=v λ (4.39)
which are computed by numerically differentiating the flux linkages, using the average of
forward and backward derivatives. Simulation results are presented in the next chapter.
The simulation output is compared to FEA predictions as well as data from lab
experiments. Matrices are expanded in Appendix B.
51
4.3 Active Rectifier Model
Automotive alternators have historically been used as generators which are
connected to passive rectifiers to power electrical loads and charge the on-board battery.
In recent years the automotive industry has manufactured vehicles with increasing use of
fuel-saving technology such as start-stop. A variant of the start-stop technology is
alternator-based start/stop, where an alternator is connected to an active rectifier in order
to provide starting torque to an internal combustion engine. In this section, an alternative
method of using the MEC to predict motoring torque is shown. The MEC is used to
predict the maximum achievable torque vs speed characteristic subject to maximum
current and bus voltage constraints. Figure 4.4 shows the circuit topology of the
machine/rectifier connection.
The MEC model can again be rearranged so that its inputs are field and stator
currents. It is similar in structure to the open circuit system. The system matrix A is
given by (4.35) and the solution vector x is given by (4.36). Now (4.37) becomes
(12,6)
(10,12)
(17,12) PM(17,10) (17,8) (17,3)
(1,12)
IP
T
TRFD
= −
00 C 0
0
C A C (4.40)
52
Figure 4.4: Active rectifier connection to machine.
Stator currents are added to the input vector and it becomes
(1,10) (1,6) (1,1)
T
PM abcxyzs fd= i b iΦ (4.41)
Electromagnetic torque is expressed in terms of quantities that are readily
obtained from the MEC. Analytical expressions for the airgap permeances derived in
Chapter 2 can be differentiated with respect to rθ and the electromagnetic torque can
then be directly calculated by [9]
( )2
1
12
nj
jej
r
r
dPT F
dθ
θ=
= ∑ (4.42)
Where n is the number of airgap elements.
fdv
Load
aibi
ci
dci
dcv
+
−
bsv
asv
csv
+
−
+
− +
−
+
−
+
−
+−xsvysv
zsv
xsi
ysizsi
53
In order to identify a maximum performance envelope for the machine, a genetic
algorithm (GA) optimization is employed to maximize torque over a three variable search
space r rdq fs dsi i i subject to maximum current and voltage constraints. Specifically, at
each rotor speed of interest, a GA performs the single-objective optimization subject to
the operating constraints
( ) ( )2 2
,2r rqs s Rateds dIi i ≤ ⋅+ (4.43)
( ) ( )2 2 13
r rqs ds dcv v v≤+ (4.44)
In order to determine the voltages so that constraint (4.44) may be checked, first the
phase currents in physical variables are given by
( ) ( )
3 3
3
cos sin2 2cos sin
2 2co sin3
s
rqs
r r
abcs r r rds
r r
ii
θ θπ πθ θ
π πθ θ
= − −
+ +
i (4.45)
cos sin6 6
2 2cos si3 3
n6 6
2 2cos sin6 63 3
r r
xyz
rqsrds
s r r
r r
ii
π πθ θ
π π π πθ θ
π π π πθ θ
− − = − − − −
+ − + −
i (4.46)
Similar to the previous section, the phase fluxes can be obtained from the stator teeth
mesh flux values. Then flux linkages are obtained for each phase. The flux linkages are
numerically differentiated using the average of the forward and backward numerical
54
derivatives. The phase voltages are then obtained by (4.1). The phase voltages are
transformed to the rotor reference frame by
0r
abcr
qd s s s=v K v (4.47)
Note that individual switch states are not simulated. Whether the modulation strategy for
the active rectifier is Space Vector Modulation, Sin-Triangle with third harmonic
injection or a current control method, sinusoidal currents result in the stator while (4.44)
is valid. The fitness function of the GA is simply electromagnetic torque (4.42). The
MEC was solved for several rotor positions spanning one slot pitch and then the torque
and voltages were averaged in order to include slotting effects. The maximal torque vs
rotor speed predicted by the optimization are presented in the next chapter. Matrices are
expanded in Appendix B.
55
VALIDATION OF MODELS
In order to validate the MEC models, three-dimensional finite element models
were created that accurately model the geometrical features and electrical output of a
commercially-available claw-pole alternator. One such FEA model was used to quantify
the flux linkage in the unloaded stator winding with varied rotor position which then
allowed the calculation of open-circuit voltages. In addition to the open-circuit FEA
model, the commercially-available alternator was tested in laboratory experiments where
it was affixed to a dynamometer test stand and loaded according to the test.
5.1 Passive Rectifier State Model Validation
Figure 5.1 and Figure 5.2 show a comparison between the prediction capability of
the MEC model and a laboratory test done on the commercially-available alternator. A
resistive load of .2 Ω was connected to the output of the alternator, the rotor was set to a
speed of 1750 RPM and had a field voltage applied such that rated field current was
achieved at 5 A. The stator current shows agreement in the harmonic content, but the
amplitudes of the harmonics don’t appear to match very well. The fundamental
component of stator current predicted by the MEC has a 4.7% difference from the
experimental data. However, when comparing RMS values, the difference increases to
10.4%. It has been observed that the harmonics are sensitive to parameters such as the
modified tooth width α′ used in the fringing airgap permeance calculation, as well as
56
claw-claw leakage permeance CLP . In addition, the amplitude of the stator currents are
sensitive to the diode parameters Dα and Dβ .
Figure 5.1: Comparison of MEC and hardware AC quantities for rectified alternator with resistive load.
The line-line voltage waveforms differ on the slope, but have a similarity in
magnitude near the peak. The chattering that is present in the slope of the MEC line-line
voltage is believed to be due to the diode model. Specifically, the transition region
between full lower diode conduction to full upper diode conduction is an abrupt change
57
in the voltage, which can cause oscillation in the state, accompanied by voltage
oscillation.
In Figure 5.2, the various DC quantities of interest are compared between the
MEC and lab test data. Field current matches well, but the DC output current has
approximately a 3.9 A difference. Inaccuracy in the stator current harmonics causes
significant discrepancy in the output current as the harmonics are rectified. There is
disagreement in both the amplitude and the harmonics of the DC voltage waveform.
Figure 5.2: Comparison of MEC and hardware AC quantities for rectified alternator with resistive load.
58
The disagreement in amplitude can be explained by the error in DC current for a given
resistance, but the disagreement in harmonics is believed to be due to the switching
events of an electronic load used in the experimental setup.
5.2 Open-Circuit Model Validation
Figure 5.3 illustrates the comparison between open circuit predictions of the MEC,
FEA and lab test data. In this comparison, the conditions were the same as for the
rectified testing (rotor speed 1750 rpm, field current 5A) except that the stator windings
have been disconnected from the rectifier and the line-line voltage is measured. The
amplitudes of all three waveforms in Figure 5.3 show agreement, however the MEC
predicts slightly higher amplitude.
Figure 5.3: Comparison between open-circuit line-line voltage prediction of MEC, FEA models and with experimental data.
59
Specifically, the fundamental component of the waveform predicted by the MEC
is 29.7 V whereas the FEA and Lab Test waveforms both predict a fundamental
component of 25.5 V. This corresponds to a 20% difference.
5.3 Active Rectifier Model Validation
In order to validate the torque prediction of the MEC, a comparison was done
between an FEA model and the MEC model for the same current magnitude, current
angle, field current and rotor speed. The normalized torque comparison is shown in
Figure 5.4. The GA coupled with the MEC was then run, as detailed in section 4.3, in
order to explore whether a higher torque value could be achieved while staying within the
machine current rating and obeying bus voltage limits. From Figure 5.4 higher torque is
indeed achievable than was shown for the FEA model. The torque has been normalized to
the maximum torque predicted by the FEA model.
Figure 5.4: Torque vs speed profiles for FEA and MEC.
0 200 400 600 800 1000 1200 1400 1600 1800 20000
0.2
0.4
0.6
0.8
1
rpm
Nor
mal
ized
Tor
que
FEAMECGA/MEC
60
CONCLUSION
In this research, the MEC model originally derived in [5] was recast as a mixed
mesh/nodal system in order to explore potential numerical improvements that were
expected based on well-performing mesh-analysis-based MEC models in [3] and [10].
Additional models and features were added and validated such as active rectification,
wye connection, permanent magnets between rotor claws and a higher phase count.
The MEC predictions show reasonable agreement with FEA and hardware testing,
in both harmonic content as well as amplitude. The error for loaded conditions is
acceptable, but the error can be larger in the open circuit voltage. The error is acceptable
since the computation time of MEC simulation is several orders of magnitude less than
FEA.
Structuring the circuit equations in a particular mixed mesh/nodal arrangement
was done with the intent that the Newton-Raphson method would have robust
convergence properties, and would not require the use of a user-tuned relaxation constant.
Despite this, the system matrix was ill-conditioned which caused poor convergence
properties and did not rid the problem of the relaxation constant. Further exploration of
nonlinear solution methods in future work is warranted in order to attain consistent
convergence with a minimal iteration count.
LIST OF REFERENCES
61
LIST OF REFERENCES
[1] C. J. Carpenter, “Magnetic equivalent circuits,” Proc. Inst. Electr. Eng., vol. 115, no. 10, pp. 1503–1511, Oct. 1968.
[2] V. Ostović, Dynamics of Saturated Electric Machines. New York, NY: Springer New York, 1989.
[3] H. W. Derbas, J. M. Williams, A. C. Koenig, and S. D. Pekarek, “A Comparison of Nodal- and Mesh-Based Magnetic Equivalent Circuit Models,” IEEE Trans. Energy Convers., vol. 24, no. 2, pp. 388–396, Jun. 2009.
[4] M. L. Bash, J. M. Williams, and S. D. Pekarek, “Incorporating Motion in Mesh-Based Magnetic Equivalent Circuits,” IEEE Trans. Energy Convers., vol. 25, no. 2, pp. 329–338, Jun. 2010.
[5] J. Williams, “Derivation of a magnetic equivalent circuit model for analysis and design of claw-pole alternator based automotive charging systems,” Masters Theses, Jan. 2001.
[6] “9780974547022: Introduction to AC Machine Design - AbeBooks - Thomas A. Lipo: 0974547026.” [Online]. Available: http://www.abebooks.com/9780974547022/Introduction-AC-Machine-Design-Thomas-0974547026/plp. [Accessed: 04-Apr-2016].
[7] S. D. Sudhoff, Ed., “Magnetics and Magnetic Equivalent Circuits,” in Power Magnetic Devices, Hoboken, New Jersey: John Wiley & Sons, Inc., 2014, pp. 45–112.
[8] P. Krause, O. Wasynczuk, S. Sudhoff, and S. Pekarek, Eds., “Distributed Windings in ac Machinery,” in Analysis of Electric Machinery and Drive Systems, Hoboken, NJ, USA: John Wiley & Sons, Inc., 2013, pp. 53–85.
[9] S. D. Sudhoff, Ed., “Force and Torque,” in Power Magnetic Devices, Hoboken, New Jersey: John Wiley & Sons, Inc., 2014, pp. 133–153.
[10] M. L. Bash and S. D. Pekarek, “Modeling of Salient-Pole Wound-Rotor Synchronous Machines for Population-Based Design,” IEEE Trans. Energy Convers., vol. 26, no. 2, pp. 381–392, Jun. 2011.
APPENDICES
62
A. AIRGAP PERMEANCE EXPRESSIONS
Airgap permeance derivations for Regions 1-4 are considered one by one. Areas of
overlap based on conditional statements are listed. To compute the airgap permeance in
Regions 1-3, the various area expressions can be substituted into
0
1RegionX
RegionX
AP
Gµ
= (A.1)
Explicit permeance expressions are derived for Region 4 because of the non-constant
airgap length.
φ
θφs
r
r
−αα
Tooth
GLP
Regi
on 1 Region 2 Region 3
Region 4
CW
z
l( )φr
β1β2 −β1 −β2
Figure A.1: Formation of airgap flux tubes.
Repeated for convenience,
rr sφ θφ= − (A.2)
( )r rα θφ α− = − − (A.3)
63
( )r rα θφ α= − (A.4)
A.1. Triangular Region 1
( ) 2r r
1 2 1 2
β GLPGLPβ β β β
l φ φ ⋅= −
− − (A.5)
Tooth edge ( α− ) transitioning into Region 1, or when ( ) ( )1 ( ) 2 ( ) 2r rα αβ φ β φ β−≤ ≤ ∧ ≥ ,
( ) ( )2
r
β2
r r r 21
12α θ
ID GLPA ID l d α θ β4 β βRegion φ φ
− −
= = + +−∫ (A.6)
Tooth spans Region 1 or Region 1 spans tooth,
( )
( )1
r
2
r 1 ( ) 2 1 ( ) 2
1
( ) 2 ( )
α θ
r r 1
rID l d , ( ) ( )2
A
ID l d , ( ) ( )2
r
r r
Region
r r
α θ
α α
β
α αβ
φ φ β φ β β φ β
φ φ θ β θ β
−
−
−
− −
≤ ≤ ∧ ≤ ≤
= ≥ ∧ ≤
∫
∫
(A.7)
( )( ) ( ) ( )
( )
2( ) ( ) 2
2 1
1
2 1( ) 2 ( 1
1 2 1
)
GLP ID,
AGLP ID
, ( ) ( )4
rr r
Region
r r
α α
α α
α β θβ φ β β φ β
β β
β βθ β θ β
−
−
⋅ +≤ ≤ ∧ ≤ ≤ −=
⋅ − ≥ ∧ ≤
(A.8)
Tooth transitioning out of Region 1, or when ( ) ( )( ) 1 ( )2 1r rα αβ φ β φ β−≥ ≥ ∧ ≤ ,
( )1
1 r rID l d2
r
RegionAα θ
β
φ φ−
= ∫ (A.9)
64
( ) ( ) ( )2 2
1 2 1 2124Region r
GLP IDA β β α θ ββ β
⋅ = − ++− − (A.10)
A.2. Rectangular Region
Area calculations consist of the product of arc lengths and l GLP= , no integration
necessary.
Tooth Transitioning into Region 2, or when ( ) ( )1 ( ) 1 ( ) 2r rα αβ φ β φ β−− ∧≤ ≤ ≥ ,
( )2 12Region rIA D GLP β α θ⋅
+= + (A.11)
Tooth spans Region 2 or Region 2 spans tooth,
( ) ( )
( ) ( )
1 ( ) 1 1 ( ) 1
2
1 ( ) 1 ( ) 1
,
,
r r
Region
r r
GLP ID
GLP
CW CW
I CD
A
W
α α
α α
α β φ β β φ β
β φ β φ β
−
−
⋅ ⋅ − ≤ ≤ ∧ − ≤ ≤ + +=
⋅ ⋅ ≥ ∧ ≤ − +
(A.12)
Tooth Transitioning out of Region 2, or when
( ) ( )1 2 1r rCW CWβ α θ β α θ β≥ − ≥ − − ≤+ ∧ − + ,
( )2 12Region rIA GLPD α θ β⋅
+−= (A.13)
65
A.3. Triangular Region 2
( ) ( )2r r
1 2 1 2
β GLPGLPβ β β β
CWl φ φ
− ⋅= − −
− − (A.14)
Tooth transitioning into Region 3, or when
( ) ( )2 ( ) 1 ( ) 1r rCCW CWWα αβ φ β φ β−≤ ≤ − + ≥− + ∧ + ,
( ) ( ) ( ) ( )1
r
-
r r2
2 23 2 1
1α2
θ
ID ID GLPA d2 4 β βReg r
W
i
C
on l CWβ
φ φ β β φ β+
− −
= = − − + − −⋅
∫ (A.15)
Tooth spans Region 3 or Region 3 spans tooth,
( ) ( ) ( )
( ) ( ) ( )1
2
r 2 ( ) 1 2 ( ) 1
3
( ) 1 ( )
r
2r
ID d ,2
A
ID d ,2
r
r
r r
RegioCW
n
r rCW
r
CW Cl CW CW
l CW C
W
W
α θ
α αα θ
β
α αβ
φ φ β φ β β φ β
φ φ φ β φ β
−
−−
+
+
−
−
−−
− + ≤ ≤ − + ≤ ≤ − +
≥ − ≤ −
∧ − += + ∧ +
∫
∫ (A.16)
( )
( ) ( )
( )
( ) ( )
22 1
2 ( ) 1 2 ( ) 1
3
2 1
( ) 1 ( ) 2
ID ,
AID GLP ,
4
r
r r
Region
r r
GLP CW
CW CW
CW CW
CW CWα α
α α
α θ ββ β
β φ β β φ β
β β
φ β φ β
−
−
− + − − + ∧ − +=
⋅ ⋅
≤ ≤ − + ≤ ≤ − +
⋅
≥
− + ∧ +
− ≤ −
(A.17)
Transitioning out of Region 3, or when
( ) ( )1 ( ) 2 ( ) 2r rCW CW CWα αβ φ β φ β−− + + ≤ −∧− +≥ ≥ ,
66
( )2
23 2
2 1
ID( )2
r
RegioCW
n r r rIDA l d G CWLPα θ
β
φ φ α θ ββ β
−
− +
= = +− −⋅−∫ (A.18)
A.4. Chamfer Region
The pole face is subdivided further into sub-regions of the chamfer region, as
shown in Figure A.2. Each sub-region is considered on a case-by-case basis, for both the
left and right possibilities in Figure A.2, i.e. 2 1CW β β≤ − and 2 1CW β β≥ − .
GLPGLPCWCW
zz
φrφr
1 1
2 2
3 3
z1
z2 z2
z1
−β2−β1 −β2−β1
Figure A.2: Two cases for determining flux tubes over chamfer region.
( ) ( )2 1 2r r
CD CDCDg ,z z CD G1CW GLP CW CW
β β βφ φ− ⋅
= − + + −⋅
(A.19)
( ) ( )1 r r 22 1
GLPz CWφ φ ββ β
= + −−
(A.20)
( ) ( )22
2 r r1
GLPz φ φ ββ β
= +−
(A.21)
Each case is considered first for the left possibility in Figure A.2, or when
2 1CW β β≤ − . For a tooth transitioning into sub-region 1, or when
( ) ( )1 ( ) 1 ( ) 1r rCW CWα αβ φ β φ β−≤ ≤ − ≥ −+ +− ∧ ,
67
( )
r 1
1 CW GLPr
0rα θ z
dz dIDμ2 g ,zchamferP
β φφ
− +
− −
= ∫ ∫ (A.22)
( )lnchamfer a b c dP = − (A.23)
where
( )0 2
2 1
CWGLPaCD
µβ β
=−
⋅
1CD +CD CW+CW g0+CD +CD rb β α θ= − ⋅ ⋅ ⋅ ⋅ ⋅
CW g0
bc =⋅
( )1CD + +CWrd α θ β= − −
For a tooth fully inside sub-region 1, or when
( ) ( )1 ( ) 1 1 ( ) 1r rCW CWα αβ φ β β φ β−− + ∧ ≤ ≤ −−− +≤ ≤ ,
( )
r
r 1
α θ GLPr
0rα θ z
dz dIDμ2 g ,zchamferP φ
φ
−
− −
= ∫ ∫ (A.24)
[ ]log( ) log( ) 2 log(CW/C ) 11Dchamfer GP a b c d e α= − −⋅ ⋅ ⋅ −⋅ (A.25)
where
0
2 1
GLP CW( - ) CD
a µβ β ⋅
=⋅
1( / 1 )rb CW CD G CWα θ β ⋅= + − + +
1+ - +CW/CD 1+CWrc Gα θ β= ⋅
1( )1/rd CW CD G CWα θ β ⋅= − + − + +
68
When sub-region 1 is fully-spanned by the tooth, or when
( ) ( )( ) 1 ( ) 1r rCWα αφ β φ β−+≥ − ≤ −∧ ,
( )
1
1
1
GLPr
0rz
dz dIDμ2 g ,zchamfe
C
r
W
Pβ
β
φφ
− +
−
= ∫ ∫ (A.26)
2
02
2 1
1( 1) ln( )CD 1chamfer
GLP CW G CDP CD G CDG
µβ β
⋅ + + − − = (A.27)
When tooth is transitioning out of sub-region 1, or when
( ) ( )1 ( ) 1 ( ) 1r rCWα αβ φ β φ β−− + ∧≤ ≤ − ≤ −
( )
1
( )
1
GLPr
0rz
dz dIDμ2 g ,z
r
chamferPαφ
β
φφ−
= ∫ ∫ (A.28)
( ln( ) ln( 1) ln( ) )chamfer b b c CW G d dP a e= − + ⋅ + + (A.29)
Where
02
2 1( )GLP CW
CDa µ
β β=
−⋅
1 1rb CD CD CD CW G CD CWα θ β= − + − + ⋅ + ⋅
( )1rc CD θ β α−= −
( )1CW Gd CD= +
( )1re CD α θ β+= − −
69
For the right possibility in Figure A.2, or when 2 1CW β β≥ − , each angular location is
considered again on a case-by-case basis. For a tooth transitioning into sub-region 1, or
when ( ) ( )1 ( ) 1 ( ) 1r rCW CWα αβ φ β φ β−≤ ≤ − ≥ −+ +− ∧ , then (A.22) and (A.23) apply again.
When the tooth is fully inside sub-region 1, or when
( ) ( )2 ( ) 1 2 ( ) 1r rCCW W WCW Cα αβ φ β β φ β−− + ∧ −≤ ≤ − + ≤ ≤ − ++ , then (A.24) and
(A.25) apply again.
When sub-region 1 is fully-spanned by the tooth, or when
( ) ( )1 ( ) ( ) 2r rCW CWα αβ φ φ β−− ≤ −∧≤ ++ ,
( )
1
1
2
GLPr
0rz
dz dIDμ2 g ,zchamf
CW
CWerP
β
β
φφ
− +
− +
= ∫ ∫ (A.30)
02
2 1( )1lnchamfer
CW Ga ba
GLP CWPh
µβ β
= − + ⋅ ⋅
− (A.31)
where
1 2CD -CD -CW G1a β β= ⋅ ⋅ ⋅
2 1-b CD CDβ β= +⋅ ⋅
When tooth is transitioning out of sub-region 1, or when
( ) ( )2 ( ) 1 ( ) 2r rCW CWCW α αβ φ β φ β−≤ ≤ − + ≤ −∧ +− +
( )
12
GLPr
0rz
dz dIDμ2 g ,z
r
chamfeW
rC
Pα θ
β
φφ− +
−
= ∫ ∫ (A.32)
[ ]02
2 1(ln( ) ln( 1) l
)n( )chamfer
GLP CWP a a b CW G cCD
cµβ β
= − + +⋅
⋅−
(A.33)
1CD -CD +CD -CW G1-CD CWra α θ β⋅ ⋅ ⋅ ⋅ ⋅=
70
2- -rb CD CD CD CD CWθ α β= +⋅ ⋅ ⋅ ⋅
1 2-CD +CW G1+CDc β β⋅ ⋅ ⋅=
Sub-region 2 is now considered. First, when 2 1CW β β≤ − , when the tooth is
transitioning into sub-region 2, or when ( ) ( )( ) 2 (1 ) 1r rCWα αβ φ β φ β−∧ +− ≤ − ≤ ≤ − ,
( )
1
1 2zr
0rz
dz dIDμ2 g ,z
r
cham
CW
ferPβ
α θ
φφ
−
− −
+
= ∫ ∫ (A.34)
01
2 1
CW GLP CD( + )ln 1+( )CD G1chamfer rP µ α β θβ β
⋅ = − − (A.35)
When the tooth is fully inside sub-region 2, or when
( ) ( )2 ( ) 1 2 ( ) 1r rCW CWα αβ φ β β φ β−− ≤ ≤ − − ≤+ ≤ −+ ∧ ,
( )
1
2zr
0rz
dz dIDμ2 g ,z
r
r
chamferPα θ
α θ
φφ
−
− −
= ∫ ∫ (A.36)
0
2 1
GLP CW G1ln( ) 1+CD
2CD GchamferP αµ
β β = −
⋅⋅
⋅ (A.37)
When sub-region 2 is fully-spanned by the tooth, or when
( ) ( )( ) 2 (1 ) 1r rCWα αβ φ β φ β−∧ +− ≤ − ≤ ≤ − ,
( )
1
1 2
2
zr
0rz
dz dIDμ2 g ,zcha r
Cmfe
W
Pβ
β
φφ−
−
+
= ∫ ∫ (A.38)
02 1
2 1
CW GLP CD( CW)ln 1+( )CD G1chamferP µ β ββ β
= − −
⋅
− (A.39)
When transitioning out of sub-region 2, or when
( ) ( )2 ( ) 1 ( ) 2r rCW CWα αβ φ β φ β−≤ ≤ − ≤ −− + ∧ + ,
71
( )
1
2
2
zr
0rz
dz dIDμ2 g ,z
r
chamf rW
eC
Pα θ
β
φφ+
−
−
= ∫ ∫ (A.40)
01
2 1
CW GLP CD( )ln 1+( )CD G1chamfer rP CWµ α θ ββ β
⋅ = + − − − (A.41)
Now, when 2 1CW β β≥ − , when the tooth is transitioning into sub-region 2, or when
( ) ( )( ) 2 1 ( ) 2r rCW CWα αφ β β φ β−∧ ≤ ≤+ +≥ − ,
( )
2r
0r0
dz dIDμ2 g ,z
r
GLP
cham
CW
ferPβ
α θ
φφ
−
− −
+
= ∫ ∫ (A.42)
[ ]02
2 1
ln(- ) * 1*ln( * 1) ln( ) - l ( )(
n)
CW GLP a a CW G CCD
D G b b c cµβ β
⋅+ +
− (A.43)
where
4 - 3- 1a CD b CD b CW G⋅ ⋅ ⋅=
1- 1rb CD CD CD CW CD CW Gα θ β⋅ ⋅ ⋅ +⋅+ ⋅= +
2- 1rc CD CD CD CW CD CW Gα θ β⋅ ⋅ +⋅+ ⋅= +
When the tooth is fully inside sub-region 2, or when
( ) ( )1 ( ) 2 1 ( ) 2r rCW CWα αβ φ β β φ β−− ≤ ≤ − + − ≤ ≤ − +∧ ,
( )
r0
r0
dz dIDμ2 g ,z
r
r
GLP
chamferPα θ
α θ
φφ
−
− −
= ∫ ∫ (A.44)
( ) ( )0
2 1
GLP CW( )CD
ln ln ln( ) ln( )chamfer a a b b c d dP cµβ β
⋅+ + −
−= (A.45)
where
72
11
ra CW GCWCD
α θ β ⋅− + − + +=
11
rCW GCW
CDb α θ β ⋅
+ += − +
21
rc CW GCWCD
α θ β ⋅− + − + +=
2G1
rCWd CW
CDα θ β ⋅
= + − + +
When sub-region 2 is fully-spanned by tooth, or when ( ) ( )( ) 2 ( ) 1r rCWα αφ β φ β−+≥ − ≤ −∧
( )
2
1
r0
r0
dz dIDμ2 g ,z
CW GLP
chamferPβ
β
φφ
− +
−
= ∫ ∫ (A.46)
[ ]02
2 1
ln( ) ln(b) ln( ) lC n( )W G( )CD
LPchamfer a a b c cP dµ
β β⋅
− + + +−
= (A.47)
Where
2 1 1CD CD CW Ga β β− ⋅ + ⋅ − ⋅=
1b CW G= ⋅
( 1 )CW Gc CD⋅ +=
1 21d CD CW G CD CD CWβ β= − ⋅ − ⋅ + ⋅ − ⋅
When tooth is transitioning out of sub-region 2, or when
( ) ( )1 ( ) 2 ( ) 1r rCWα αβ φ β φ β−+ ∧− ≤ ≤ − ≤ − ,
( )
1
r0
r0
dz dIDμ2 g ,z
r GLP
chamferPα θ
β
φφ
−
−
= ∫ ∫ (A.48)
73
( ) [ ]0
22 1
ln( ) ln( ) ln( ) ln( )chamferGLP a aCWPCD
b b c c d dµβ β
⋅− + + − +
−= (A.49)
where
2 1 1CD CD CW G CDa CWβ β⋅ ⋅ − ⋅ − ⋅= −
( 1 )CW Gb CD⋅ +=
1 1rCD CD Cc D CW CD CW Gα θ β⋅ − ⋅ − ⋅ + ⋅ − ⋅=
2 1rCD CD CD CW CD W Gd Cα θ β− ⋅ + ⋅ + ⋅ − ⋅ + ⋅=
Sub-region 3 is now considered. First, the case when 2 1CW β β≤ − is considered for all
angular location possibilities. When the tooth is transitioning into sub-region 3, or when
( ) ( )2 ( ) 2 2 ( )r rCW CWα αβ φ β β φ− ≤ ≤ − − + ≤∧+ ,
( )
2 2
( )
r0
r0
dz dIDμ2 g ,z
r
z
cha
C
mfe
W
rPα
β
φ
φφ
−
− +
= ∫ ∫ (A.50)
( )( ) ( )02
2 1
l( )
n 1 lnchamferCW GLPP a
Dc
Cb dµ
β β =
⋅+ + −
(A.51)
where
2 ra CD CD CW CD CDβ α θ⋅ ⋅ ⋅− + + ⋅= +
2
CW G1+CW CDCD +CD -CD +CD CW+CW G1r
bα θ β
⋅ ⋅⋅ ⋅ ⋅ ⋅ ⋅
=
1c CW G= ⋅
2
CW G1CD +CD -CD +CD CW+CW G1r
dα θ β
⋅⋅ ⋅ ⋅ ⋅
=
74
When the tooth is fully inside sub-region 3, or when
( ) ( )2 ( ) 2 2 ( ) 2r rCW CWα αβ φ β β φ β−− ≤ ≤ − + − ≤ ≤ − +∧ ,
( )
( ) 2
( )
r0
r0
dz dIDμ2 g ,z
r
r
z
chamferPα
α
φ
φ
φφ
−
= ∫ ∫ (A.52)
[ ]0
2 1
GLP CW 2 (ln( ) 1) - ln( ) ln( )( ) CDchamferP CD a a b bµ αβ β
= + +⋅ ⋅− ⋅
(A.53)
where
2- - -1r
CWa CWCD G
α θ β+⋅
=
2- - - -1r
CWb CWCD G
α θ β= +⋅
When sub-region 3 is fully-spanned by the tooth, or when
( ) ( )2 ( ) ( ) 2r rCW α αβ φ φ β−+ ∧− ≤ ≤ − ,
( )
2 2
2
r0
r0
dz dIDμ2 g ,z
z
cha f
W
m er
C
Pβ
β
φφ
− +
−
= ∫ ∫ (A.54)
[ ]2
02
2 1
1(ln( 1) ln( 1 ))*( )chamfer
CW GLP G CW G CW G CW CD CWD
PCµ
β β⋅
⋅ − ⋅ + ⋅ +=−
(A.55)
When the tooth is transitioning out of sub-region 3, or when
( ) ( )2 ( ) 2 ( ) 2r rCWα αβ φ β φ β−+ ∧− ≤ ≤ − ≤ − ,
( )
( ) 2
2
r0
r0
dz dIDμ2 g ,z
r z
chamferPαφ
β
φφ−
= ∫ ∫ (A.56)
[ ]022
2 1( )ln( ) ( - )chamfer r
CW GLPP a b DCD
Cµ α β θβ β
⋅⋅
−= + + (A.57)
where
75
2-CD +CD -CD +CD CW+G1 CWra α θ β⋅ ⋅ ⋅ ⋅ ⋅=
2CD (- + - +CW)+CW G1CW G1+CW CD
rb α θ β⋅ ⋅⋅ ⋅
=
Now, the case when 2 1CW β β≥ − is considered for all angular location possibilities,
again for sub-region 3. When the tooth is transitioning into sub-region 3, or when
( ) ( )2 ( ) 1 1 ( )r rα αβ φ β β φ∧− ≤ ≤ − − ≤ ,
( )
1 2
( )
r0
r0
dz dIDμ2 g ,z
r
z
chamferPα
β
φ
φφ
−
−
= ∫ ∫ (A.58)
[ ]02 ln(b) ln( ) - ln( )
( 3 4)chamfer b b CCW GLPP a c c d
Dd eµ ⋅
−= + + (A.59)
where
1- CD+CD +CD ra β α θ⋅ ⋅ ⋅=
1b CW G CW CD= +⋅ ⋅
1 21c CW CW G CD CD CWβ β ⋅= − +⋅ ⋅+⋅
21rd CD CD CW G CD CD CWα θ β⋅ ⋅ ⋅ ⋅+ − + ⋅= +
1re CD CD CDα θ β⋅ ⋅ − ⋅= +
When the tooth is fully inside sub-region 3, or when
( ) ( )2 ( ) 1 2 ( ) 1r rα αβ φ β β φ β−− ≤ ≤ − − ≤ ≤ −∧ ,
( )
( ) 2
( )
r0
r0
dz dIDμ2 g ,z
r
r
z
chamferPα
α
φ
φ
φφ
−
= ∫ ∫ (A.60)
[ ]0
2 1
2 (1 ln( )) - ln( ) ln( )( )chamfer
GLP CWP CD a a b bCD
µ αβ β
= + +−
⋅ ⋅⋅
(A.61)
76
where
2- - -1r
CWa CWCD G
α θ β+⋅
=
2- - - -1r
CWb CWCD G
α θ β= +⋅
When the sub-region 3 is fully-spanned by the tooth, or when
( ) ( )1 ( ) ( ) 2r rα αβ φ φ β−− ∧≤ ≤ − ,
( )
1 2
2
r0
r0
dz dIDμ2 g ,z
z
chamferPβ
β
φφ
−
−
= ∫ ∫ (A.62)
02
2 1
ln( )
CW GLP baCD
dc
µβ β
⋅+ −
(A.63)
where
2 1- CD+CW G1+CD CW+ CDa β β⋅= ⋅ ⋅ ⋅
CW G1+CW CDb = ⋅ ⋅
2 1- CD+CW G1+CD CW+ CDc β β⋅= ⋅ ⋅ ⋅
2-CD -CD*b3d β=
When the tooth is transitioning out of sub-region 3, or when
( ) ( )2 ( ) 1 ( ) 2r rα αβ φ β φ β−− ≤ ≤ ≤ −∧− ,
( )
( ) 2
2
r0
r0
dz dIDμ2 g ,z
r z
chamferPαφ
β
φφ−
= ∫ ∫ (A.64)
[ ]02
2 1(ln( ) ln(
))chamfer
CW GLPP a b c cCD
dµβ β
+⋅
−= + (A.65)
where
77
2CD -CD + CD-CW G1-CD CWra α θ β⋅ ⋅ ⋅ ⋅ ⋅=
CW G1+CW CDb = ⋅ ⋅
2-CD +CD - CD+CW G1+CD CWrc α θ β⋅ ⋅ ⋅ ⋅ ⋅=
2CD -CD +CDrd β θ α⋅= ⋅ ⋅
78
B. MATRIX EXPANSION
The matrices for the system of MEC equations are expanded here. Where applicable, a
matrix may appear as several different versions depending on which model it is used in or
which stator phase connection is used.
(12,12)
1 1 0 00 1 1 0
0 1 1 00 1 1 0
0 1 1 00 1 1 0
0 1 1 00 1 1 0
0 1 1 00 1 1 0
0 0 0 1 11 0 0 1
LT
− − − − −
− = −
− − − − −
A (2.1)
(12,10) , 1 12, 1 10TC TiCjP i j= = = A (2.2)
(10,12) (12,10)CT TCT=A A (2.3)
(12,12) (12,12)TLTTL =A A (2.4)
79
(3,10)
0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0
FDL FDL
RFC
P PA
− =
(2.5)
(3,1)2 0
2
T
END FDL FDLREND
FDL COR DO FDL
R R RR R R R
+ −= − − −
A (2.6)
(3,1)20
2
T
COR SHA COR DIRS
COR DO FDL COR
R R R RR R R R
− + += − − −
A (2.7)
(3,1) 0 02
T
RFDCOR DO FDL
TRCR R R
−= − − −
A (2.8)
(3,7)
0 0 10 0 10 0 1
RFD
=
C (2.9)
80
80
(12,12)
122 1
1 2
232 3
2 3
343 4
3 4
454 5
4 5
565 6
5 6
676 7
6 7
787
7
0 0
0 0 0
0 0
0 0
0 0
0 0
0
LL
YT T
T T
YT T
T T
YT T
T T
YT T
T T
YT T
T T
YT T
T T
YT
T T
A
RR R
R R
RR R
R R
RR R
R R
RR R
R R
RR R
R R
RR R
R R
RR
R R
=
+ − − +
+− −
+
+− −
+
+− −
+
+− −
+
+− −
+
+−
+
88
898 9
8 9
9109 10
9 10
101110 11
10 11
111211 12
11 12
1211 12
12 1
0
0 0
0 0
0 0
0 0 0
0 0
T
YT T
T T
YT T
T T
YT T
T T
YT T
T T
YT T
T T
R
RR R
R R
RR R
R R
RR R
R R
RR R
R R
RR R
R R
− + − − +
+− −
+
+− −
+
+
− −+
+−
+ −
(2.10)
81
81
1
1
1
1
1
2
3
4
5
(12,12
1
)
20 0
20 0 0
20 0
20 0
20 0
20
c
c
c
c
c
N
i
N
i
N
TL
TL TLT Ci
TL
TL TLT Ci
TL
TL TLT C
ii
TL
TL TLT Ci
TL
TL TLT Ci
N
i
N
i
TL
TL
TT
PP P
P
PP P
P
PP P
P
PP P
P
PP P
P
PP
P
=
=
=
=
=
− −
− −
− −
− −
− −
−
−
=
∑
∑
∑
∑
∑
A6
7
8
9
10
11
1
1
1
1
1
1
0
20 0
20 0
20 0
20 0
20 0 0
c
c
c
c
c
c
TLT Ci
TL
TL TLT Ci
TL
TL TLT Ci
TL
TL
N
i
N
i
N
i
N
i
N
TLT Ci
TL
TL TLT Ci
TL
TLT Ci
i
N
i
P
PP P
P
PP P
P
PP P
P
PP P
P
PP
P
=
=
=
=
=
=
− −
− −
− −
− −
− −
∑
∑
∑
∑
∑
∑
11
2
20 0 c
TL
TL
TL TN
i
LT Ci
P
PP P
P=
− −
∑
(2.11)
82
82
1
110
1101
29
292
38
383
47
474
56
(10,
1
1
)
1
10
20 0 0 0 0 0 0 0 2
20 0 0 0 0 0 0 2 0
20 0 0 0 0 0 2 0 0
20 0 0 0 0 2 0 0 0
20 0 0 0
t
t
t
t
N
i
N
i
N
CL
CLTiC
CL
CLTiC
CL
CLTiC
CL
CLTiC
CL
Ti
C
i
i
C
N
PP
P
PP
P
PP
P
PP
P
P
P
=
=
=
=
− −
− −
− −
− −
−
−
=
∑
∑
∑
∑
A 1
1
1
1
565
56
566
47
477
38
388
29
299
1
2 0 0 0 0
20 0 0 0 2 0 0 0 0
20 0 0 2 0 0 0 0 0
20 0 2 0 0 0 0 0 0
20 2 0 0 0 0 0 0 0
t
t
t
t
t
N
i
N
i
N
CLC
CL
CLTiC
CL
CLTiC
CL
CLTiC
CL
CL
i
N
i
T
N
iCi
P
PP
P
PP
P
PP
P
PP
P
=
=
=
=
=
− −
− −
− −
− −
∑
∑
∑
∑
∑
110
11
1
010
22 0 0 0 0 0 0 0 0 tN
CL
CLTiC
i
PP
P=
− −
∑ (2.12)
83
For wye connection,
(12,4)
1 0 0 00 0 1 01 1 0 00 0 1 10 1 0 00 0 0 11 0 0 0
0 0 1 01 1 0 0
0 0 1 10 1 0 00 0 0 1
IP TC
= ⋅ −
− − − − − − −
A (2.13)
(4,12)
1 0 1 0 0 0 1 0 1 0 0 00 0 1 0 1 0 0 0 1 0 1 00 1 0 1 0 0 0 1 0 1 0 00 0 0 1 0 1 0 0 0 1 0 1
FF
− − − − = − − − − −
A (2.14)
(4,7)
1 0 0 0 0 0 00 1 0 0 0 0 00 0 0 1 0 0 00 0 0 0 1 0 0
F
=
C (2.15)
84
For delta connection,
(12,6)
1 0 0 0 0 00 0 0 1 0 00 0 1 0 0 00 0 0 0 0 10 1 0 0 0 00 0 0 0 1 01 0 0 0 0 0
0 0 0 1 0 00 0 1 0 0 00 0 0 0 0 10 1 0 0 0 00 0 0 0 1 0
IP TC
− − = ⋅ −
− − −
A (2.16)
(6,12)
1 0 0 0 0 0 1 0 0 0 0 00 0 0 0 1 0 0 0 0 0 1 00 0 1 0 0 0 0 0 1 0 0 00 1 0 0 0 0 0 1 0 0 0 00 0 0 0 0 1 0 0 0 0 0 10 0 0 1 0 0 0 0 0 1 0 0
FF
− − −
= − −
−
A (2.17)
(4,7)
1 0 0 0 0 0 00 1 0 0 0 0 00 0 1 0 0 0 00 0 0 1 0 0 00 0 0 0 1 0 00 0 0 0 0 1 0
F
=
C (2.18)
85
(12,8)
1 0 0 0 0 0 0 00 1 0 0 0 0 0 01 0 1 0 0 0 0 0
0 1 0 1 0 0 0 00 0 1 0 1 0 0 00 0 0 1 0 1 0 00 0 0 0 1 0 1 00 0 0 0 0 1 0 10 0 0 0 0 0 1 00 0 0 0 0 0 0 1
CS
− − − −
= −
− −
A (2.19)
[ ](10,1) 0 0 1 1 TCE = −A (2.20)
(8,10)
1 0 1 0 00 1 0 1 0
0 1 0 1 00 1 0 1 0
0 1 0 1 00 1 0 1 0
0 1 0 1 00 0 1 0 1
FS
− − − − = −
− −
−
A (2.21)
13
24
35
46(8,8)
57
68
79
810
0 00 0
0 00 0
0 00 0
0 00 0
C
CS
CS
CSFD
CS
CS
CS
CS
RR
RR
RR
RR
=
A (2.22)