Fabrice Mutabazi, Nicholas Larno
permanent magnet motor Analytical model of physical properties of a
high speed
Academic year 2016-2017 Faculty of Engineering and Architecture
Chair: Prof. dr. ir. Luc Dupré Department of Electrical Energy,
Systems and Automation
Master of Science in de industriële wetenschappen: elektrotechniek
Master's dissertation submitted in order to obtain the academic
degree of
Counsellor: Bert Hannon Supervisor: Prof. dr. ir. Peter
Sergeant
Copyright protection
The authors give permission to make this master dissertation
available for consultation and to copy parts of this master
dissertation for personal use. In the case of any other use, the
copyright terms have to be respected, in particular with regard to
the obliga- tion to state expressly the source when quoting results
from this master dissertation."
i
Fabrice Mutabazi, Nicholas Larno
permanent magnet motor Analytical model of physical properties of a
high speed
Academic year 2016-2017 Faculty of Engineering and Architecture
Chair: Prof. dr. ir. Luc Dupré Department of Electrical Energy,
Systems and Automation
Master of Science in de industriële wetenschappen: elektrotechniek
Master's dissertation submitted in order to obtain the academic
degree of
Counsellor: Bert Hannon Supervisor: Prof. dr. ir. Peter
Sergeant
Foreword and Acknowledgments
The subject of the master dissertation is the analytical study of
physical properties in high speed electrical machines.
We would like to thank our supervisor Prof.dr.ir. Peter Sergeant
and counselor ing. Bert Hannon for their insightful guidance,
advice and help during the entire year. This thesis would be less
accurate without their efforts. We would also like to express our
gratitude to everyone who has given us support to complete this
master dissertation. Especially Alyssa Vansteenbrugge and Ruben
Lammertyn for reading and correcting the script.
iii
Abstract
The goal of this master dissertation is to create an analytical
model to study physical properties of a high speed permanent magnet
motor. The demand for electrical ma- chines is changing. On the one
hand, the demand for more efficient electrical machines is
increasing due to an increasing ecological awareness. On the other
hand, there is an increase in demand of electrical machines with a
high power density due to a trend to- wards more flexible
applications which demands for smaller machines. This is driving
the interest for high-speed permanent magnet synchronous machines
(PMSMs). Design- ing high speed electrical machines is a
significantly more challenging task than tradi- tional machines.
Due to the high operating speed of the machines, rigorous
mechanical, thermal and electromagnetic conditions need to be
withstood. This work will focus on the electromagnetic aspect of
these machines such as eddy currents, eddy current losses and
torque.
the first step was to introduce an analytical method to study the
physical properties of an electrical machine. An analytical method
was preferred because of various reasons. First analytical models
have proven to require less computational time [1], [2] and are
more flexible in comparison with a finite element model. Secondly,
when using readily available numerical software, less machine
insight is re- quired while developing the model. Developing an
analytical model requires a more pro- found knowledge of the
machine’s operation and the underlaying physics. Thus using an
analytical model is in the authors best interest in gaining
knowledge on the subject. thirdly, analytical models are easier to
parametrize, which is a key feature in this work. The second step
was to understand the mathematical background of the physical prop-
erties and adjust them to the used analytical model. The discussed
physical properties are the machine’s torque and its components,
the eddy currents and the losses they pro- duce in the shielding
cylinder and the permanent magnets. The final step was to conduct a
parameter study in order to gain a better understand- ing of the
effects of the shielding cylinder on the machine’s performance. The
effect of the shielding cylinder’s thickness and conductivity on
the torque and the eddy current losses is studied.
v
Analytical model of physical properties of a high speed permanent
magnet motor
Nicholas Larno, Fabrice Mutabazi
Supervisor(s): Peter Sergeant, Bert Hannon
Abstract—The goal of this master dissertation is to create an
analytical model to study physical properties of a high speed
permanent magnet mo- tor. The demand for electrical machines is
changing. On the one hand, the demand for more efficient electrical
machines is increasing due to an increasing ecological awareness.
On the other hand, there is an increase in demand of electrical
machines with a high power density due to a trend towards more
flexible applications which demands for smaller machines. This is
driving the interest for high-speed permanent magnet synchronous
machines (PMSMs). Designing high speed electrical machines is a
signif- icantly more challenging task than traditional machines.
Due to the high operating speed of these machines, there are
rigorous mechanical, thermal and electromagnetic conditions that
need to be withstood. This work will fo- cus on the electromagnetic
aspect of these machines such as eddy currents, eddy current losses
and torque.
Keywords—Fourier-based modeling, eddy currents, eddy current
losses, torque, high speed permanent magnet motor, shielding
cylinder
I. INTRODUCTION
THE first step was to introduce an analytical method to study the
physical properties of an electrical machine. An ana-
lytical method was preferred because of various reasons. First,
analytical models have proven to require less computa- tional time
[1], [2] and are more flexible in comparison with a finite element
model. Secondly, when using readily available numerical software,
less machine insight is required while developing the model. Devel-
oping an analytical model requires a more profound knowledge of the
machine’s operation and the underlaying physics. Thus using an
analytical model is in the authors best interest in gain- ing
knowledge on the subject. Thirdly, analytical models are easier to
parametrize which is a key feature in this work. The second step
was to understand the mathematical back- ground of the physical
properties and adjust them to the used analytical model. The
discussed physical properties are the ma- chine’s torque and its
components, the eddy currents and the losses they produce in the
shielding cylinder and the permanent magnets. The final step was to
conduct a parameter study in order to gain a better understanding
of the effects of the shielding cylinder on the machine’s
performance. The effect of the shielding cylin- der’s thickness and
conductivity on the torque and the eddy cur- rent losses is
studied.
II. FOURIER-BASED MODELING
The goal of this technique is to solve Maxwell’s equations in the
entire geometry. To do so Maxwell’s equations are reformu- lated to
a single partial-differential equation using a magnetic
N. Larno and F. Mutabazi are with the Department of Electrical
energy, met- als, mechanical constructions and systems, Ghent
University (UGent), Gent, Belgium. (E-mail:
[email protected],
[email protected])
potential, that equation is referred to as the governing equation.
The machine is divided into subdomains in which the governing
equation is simplified and solved using the separation of vari-
ables technique. Using this technique implies that the result will
contain integration constants and eigenvalues. These are deter-
mined by imposing boundary conditions, these conditions also ensure
a coherent solution over the entire geometry.
III. TORQUE
In this chapter an analytical method for torque calculation is
introduced. The calculation is based on the Maxwell’s stress
tensor, which is briefly described in this chapter. Secondly the
implementation in the Fourier-based model is pre- sented and the
two considered torque components are described. The Maxwell’s
stress tensor is based on the total magnetic force F on a rigid
body with volume V and surface S which is placed in a magnetic
field. The derivation starts with a calculation of the total force
due to electromagnetic fields on the charges and currents within a
volume V calculated with the Lorentz force law:
F =
ρ(E + v × B)dV (1)
The total electromagnetic torque in a machine equipped with a
shielding cylinder is a combination of two torque producing
phenomena.
0 0.1 0.2 0.3 0.4 −2
1
4
7
T PM
T SC
Fig. 1. The torque and its components
First of which is the interaction of the permanent magnets and the
magnetic field due to the alternating stator current. This
interaction produces the dominant torque and is referred to as TPM
, the torque produced in the magnets. The other torque component is
produced by the interaction of eddy currents in the shielding
cylinder and the permanent magnet’s magnetic field. This component
is referred to as TSC , the torque produced in the shielding
cylinder. It should be noted that TPM and TSC
are fictitious components, they do not occur separately but are
regarded as such to provide a better understanding of the ma-
chine’s physics.
IV. EDDY CURRENTS
Knowledge of the eddy currents induced in the shielding cylinder
and the permanent magnets is crucial for the calcula- tion of the
eddy current losses, which is an important design factor for
electrical machines. In a moving conductive material, the current
density J is related to the electrical field E and the magnetic
field B surrounding the conductor:
J = σ (E + v × B) (2)
For the permanent magnets, a constant needs to be introduced to
ensure that the net current in each magnet is zero. In a 2D
approximation, it is possible for currents to be closed in another
magnet. Current flowing from one magnet to the other is of course
not possible. Because the shielding cylinder is made of one single
piece and thus short-circuited, it is possible for cur- rent loops
to be closed in an other region of the shielding cylin- der without
violating the law of conservation of energy. Thus it is not
necessary to use this spatially constant current density in the
shielding cylinder.
V. EDDY CURRENT LOSSES
To compute the eddy current losses in the shielding cylinder and
the permanent magnets, two different methods are used. The losses
in the permanent magnets can be calculated form Joule’s
formula:
P (t) =
dv (3)
Joule’s formula is not effective at computing the losses in the SC
because the r-dependent part of J3(r, φ, t) contains Bessel
functions, those functions can not be integrated analytically. To
cope with this problem, the losses in the shielding cylinder are
calculated using the Poynting vector (S):
S = E × H (4)
The Poynting vector represents the rate per unit area at which
energy crosses a surface. The integration of S over a closed
surface determines the energy flowing through that surface. In this
case, this implies that the total power P transmitted from the
stator to the rotor can be calculated by integrating the Poynting
vector over the surface of a cylinder that includes the
rotor.
VI. PARAMETER STUDY
A. Torque
As expected, TPM shows very little dependency of the shield- ing
cylinders conductivity. TPM is mainly caused by the inter-
action of synchronous field components and the expectation is that
these components are not affected by the shielding cylin- der’s
conductivity. Although when plotted separately, a small variation
of TPM can be observed. The reason for these small variations is
that a part of TPM is produced by the interaction be- tween
asynchronous field components at the boundary between the air gap
and the slotting region. The evolution of TSC is very similar to
the speed-torque charac- teristic of an induction machine and can
be explained by regard- ing the shielding cylinder as the squirrel
cage of an induction machine. Note that TSC is negative which
implies that the asyn- chronous components with a slower rotational
speed than the synchronous components are dominant.
0 2 4 6 8 10
x 10 7
T or
qu e
(N m
T SC
Fig. 2. Average torque as a function of the conductivity of the
SC
The second parameter studied is the thickness of the shielding
cylinder. The evolution of TSC can be explained by regarding the
shielding cylinder as an electrical conductor and applying
Pouillet’s law, which means that the shielding cylinder’s resis-
tance decreases as the thickness increases. This explains the
initial increase of TSC . As the thickness further increases, there
is a decrease of both components due to the increase of the ef-
fective air gap. The shielding cylinder proves to be beneficial in
reducing the ripple on the total torque as an increase of the
thick- ness shows a significant decrease of the torque ripple of
TPM . This can be interesting for applications which need a
constant torque.
B. Dissipated power
The shielding cylinder proves to be beneficial in reducing to- tal
dissipated power. Because of the shielding effect less eddy current
are induced in the permanent magnet which have a rel- atively high
resistance. The eddy currents instead are induced in the SC which
has a lower resistivity, thus the total dissipated power is
significantly reduced as depicted on Figure 4.
REFERENCES
[1] A. Hemeida, B. Hannon, H. Vansompel, and P. Sergeant,
“Comparison of three analytical methods for the precise calculation
of cogging torque and torque ripple in axial flux pm machines,”
MATHEMATICAL PROBLEMS IN ENGINEERING, vol. 2016, pp. 1–14,
2016.
[2] K. Ramakrishnan, M. Curti, D. Zarko, G. Mastinu, J. J. H.
Paulides, and E. A. Lomonova, “A comparison study of modelling
techniques for per- manent magnet machines,” in 2016 Eleventh
International Conference on Ecological Vehicles and Renewable
Energies (EVER), pp. 1–6, April 2016.
0 1 2 3 4 5
x 10 −3
T SC
Fig. 3. Average torque as a function of the thickness of the
SC
0 2.6
P ow
SC
Fig. 4. Dissipated power as a function of the conductivity of the
SC
[3] B. Hannon, P. Sergeant, and L. Dupre, “Voltage Sources in 2d
Fourier- Based Analytical Models of Electric Machines,”
Mathematical Problems in Engineering, vol. 2015, p. e195410, Oct.
2015.
[4] Z. Q. Zhu, L. J. Wu, and Z. P. Xia, “An accurate subdomain
model for mag- netic field computation in slotted surface-mounted
permanent-magnet ma- chines,” IEEE Transactions on Magnetics, vol.
46, pp. 1100–1115, April 2010.
[5] T. Lubin, S. Mezani, and A. Rezzoug, “2-d exact analytical
model for surface-mounted permanent-magnet motors with semi-closed
slots,” IEEE Transactions on Magnetics, vol. 47, pp. 479–492, Feb
2011.
[6] Z. J. Liu and J. T. Li, “Analytical solution of air-gap field
in permanent- magnet motors taking into account the effect of pole
transition over slots,” IEEE Transactions on Magnetics, vol. 43,
pp. 3872–3883, Oct 2007.
[7] K. Ramakrishnan, M. Curti, D. Zarko, G. Mastinu, J. J. H.
Paulides, and E. A. Lomonova, “Comparative analysis of various
methods for modelling surface permanent magnet machines,” IET
Electric Power Applications, vol. 11, no. 4, pp. 540–547,
2017.
[8] H. V. Xuan, D. Lahaye, H. Polinder, and J. A. Ferreira,
“Influence of stator slotting on the performance of
permanent-magnet machines with concen- trated windings,” IEEE
Transactions on Magnetics, vol. 49, pp. 929–938, Feb 2013.
[9] B. Hannon, P. Sergeant, and L. Dupre, “2d analytical torque
study of slot- less and slotted pmsm topologies at high-speed
operation,” in Proceedings of ElectrIMACS 2014, pp. 248–231,
Universitat Politecnica de Valencia, 2014.
[10] P. D. Pfister, X. Yin, and Y. Fang, “Slotted permanent-magnet
machines: General analytical model of magnetic fields, torque, eddy
currents, and permanent-magnet power losses including the diffusion
effect,” IEEE Transactions on Magnetics, vol. 52, pp. 1–13, May
2016.
[11] F. Zhou, J. Shen, W. Fei, and R. Lin, “Study of retaining
sleeve and con- ductive shield and their influence on rotor loss in
high-speed pm bldc mo- tors,” IEEE Transactions on Magnetics, vol.
42, pp. 3398–3400, Oct 2006.
[12] D. J. Griffiths, Introduction to Electrodynamics. Pearson
Education, 2007. [13] B. Hanon, P. Sergeant, and L. Dupre, “2d
analytical torque study of slotless
and slotted pmsm topologies at high-speed operation.” none. [14] B.
Hannon, P. Sergeant, and L. Dupre, “2-d analytical subdomain model
of
a slotted pmsm with shielding cylinder,” IEEE Transactions on
Magnetics, vol. 50, pp. 1–10, July 2014.
[15] B. Hanon, P. Sergeant, and L. Dupre, “Torque and torque
components in high-speed permanent-magnet synchronous machines with
a shielding cylinder.” none.
[16] A. Tenconi, S. Vaschetto, and A. Vigliani, “Electrical
machines for high- speed applications: Design considerations and
tradeoffs,” IEEE Transac- tions on Industrial Electronics, vol. 61,
pp. 3022–3029, June 2014.
[17] B. Hanon, P. Sergeant, and L. Dupre, “Two-dimensional
fourier-based modeling of electric machines.” none.
[18] Z. Tian, C. Zhang, and S. Zhang, “Analytical calculation of
magnetic field distribution and stator iron losses for
surface-mounted permanent magnet synchronous machines,” Energies,
2017.
[19] J. J. Lee, W. H. Kim, J. S. Yu, S. Y. Yun, S. M. Kim, J. J.
Lee, and J. Lee, “Comparison between concentrated and distributed
winding in ipmsm for traction application,” in 2010 International
Conference on Electrical Ma- chines and Systems, pp. 1172–1174, Oct
2010.
[20] M. Merdzan, J. J. H. Paulides, and E. A. Lomonova,
“Comparative analy- sis of rotor losses in high-speed permanent
magnet machines with different winding configurations considering
the influence of the inverter pwm,” in 2015 Tenth International
Conference on Ecological Vehicles and Renew- able Energies (EVER),
pp. 1–8, March 2015.
[21] S. Niu, S. L. Ho, W. N. Fu, and J. Zhu, “Eddy current
reduction in high-speed machines and eddy current loss analysis
with multislice time-stepping finite-element method,” IEEE
Transactions on Magnetics, vol. 48, pp. 1007–1010, Feb 2012.
[22] D. A. Co Huynh, Liping Zheng, “Losses in high speed permanent
magnet machines used in microturbine applications,” J. Eng. Gas
Turbines Power 131(2), 022301 (Dec 23, 2008) (6 pages), p. 6,
2008.
[23] J. Lim, Y. J. Kim, and S.-Y. Jung, “Numerical investigation on
permanent- magnet eddy current loss and harmonic iron loss for pm
skewed ipmsm,” Journal of Magnetics, 2011.
Analytisch model van fysische eigenschappen van een permanente
magneetbekrachtigde
hogesnelheidsmotor Nicholas Larno, Fabrice Mutabazi
Supervisor(s): Peter Sergeant, Bert Hannon
Abstract— Het doel van deze masterproef is het creeren van een ana-
lytisch model om de fysische eigenschappen van een hoge
snelheidsmotor met permanente magneten te bestuderen. De vraag naar
elektrische ma- chines is aan het veranderen. Aan de ene kant is de
vraag naar meer ef- ficiente elektrische machines aan het stijgen
door een stijgende ecologische bewustwording. Aan de andere kant is
er een stijgende vraag naar elektri- sche machines met een hoge
vermogendichtheid door een trend naar meer flexibele applicaties
die kleinere machines vereisen. Dit drijft de aandacht naar
permanent magneet synchrone hoge snelheidsmotoren (PMSMs). Het
ontwerpen van hoge snelheidsmotoren is een aanzienlijk uitdagendere
taak dan traditionele machines. Door de hoge snelheid van die
motoren zijn er extreme mechanische, thermische en
elektromagnetische condities die ze moeten kunnen weerstaan. Dit
werk zal zich verdiepen in het elektromag- netische aspect van deze
machines zoals de wervelstromen, wervelstroom- verliezen en het
koppel.
Keywords—Fourier-based modeling, eddy currents, eddy current
losses, torque, high speed permanent magnet motor, shielding
cylinder
I. INLEIDING
DE eerste stap was het invoeren van een analytisch model om de
fysische eigenschappen van een elektrische machine te
bestuderen. Een analytische methode was gekozen om verschil- lende
redenen. Ten eerste hebben analytische modellen bewezen minder
reken- tijd te vereisen ([1], [2]) en zijn ze meer flexibel in
vergelijking met een eindige elementen methode. Ten tweede, bij
gebruik van reeds beschikbare numerieke soft- ware is er minder
machine inzicht vereist tijdens het ontwikke- len van het model.
Het ontwikkelen van een analytisch model vereist een diepgaande
kennis van de werking van de machine en de onderliggende fysica. Zo
is het gebruik van een analytisch model het best voor de auteurs om
kennis over het onderwerp te verwerven. Ten derde zijn analytische
modellen eenvoudiger te parametri- seren, wat een belangrijk
onderdeel is in dit werk. De tweede stap was het begrijpen van de
wiskundige achter- grond van de fysische eigenschappen en deze
toepassen op het gebruikte analytisch model. De besproken fysische
eigenschap- pen van de machine zijn het koppel en zijn componenten,
de wervelstromen en de verliezen die ze produceren in de shielding
cylinder en de permanente magneten. De laatste stap was het
uitvoeren van een parameter studie om een beter inzicht te krijgen
in het effect van de shielding cylin- der op de performantie van de
machine. Het effect van de dikte en de geleidbaarheid van de
shielding cylinder op het koppel en de wervelstroom verliezen
werden bestudeerd.
N. Larno and F. Mutabazi are with the Department of Electrical
energy, me- tals, mechanical constructions and systems, Ghent
University (UGent), Gent, Belgium. (E-mail:
[email protected],
[email protected])
II. FOURIER-BASED MODELING
Het doel van deze techniek is het oplossen van de Max- well
vergelijkingen in de volledige geometrie. Om dit te doen worden de
Maxwell vergelijkingen geherformuleerd naar een partiele
differentiaalvergelijking gebruikmakend van een mag- netisch
potentiaal, deze vergelijking wordt de governing equa- tion
genoemd. De machine is in subdomeinen ingedeeld waarin de governing
equation vereenvoudigd is en kan opgelost wor- den door de
scheiding van veranderen techniek toe te passen. Het gebruik van
deze techniek zorgt ervoor dat het bekomen re- sultaat
integratieconstanten en eigenwaarden zal bevatten. Deze worden
bepaald door grensvoorwaarden in te stellen, deze voor- waarden
zorgen ook voor een coherente oplossing over de ge- hele
meetkunde.
III. KOPPEL
In dit hoofdstuk wordt een analytische methode voor het be- rekenen
van het koppel gentroduceerd. De berekening is geba- seerd op de
Maxwell stress tensor, welke uitvoerig besproken wordt in dit
hoofdstuk. Ten tweede wordt de implementatie in het Fourier-based
model besproken als ook de twee beschouwde koppel componenten. De
Maxwell stress tensor is gebaseerd op de totale magnetische kracht
F op een star lichaam met volume V en oppervlakte S dat zich in een
magnetisch veld bevindt. De afleiding start met de berekening van
de totale kracht veroorzaakt door elektromagne- tische velden op
ladingen en stromen in een volume V berekent met de Lorentz force
law:
F =
ρ(E + v × B)dV (1)
Het totale elektromagnetische koppel in een machine uitge- rust met
een shielding cylinder is een combinatie van twee kop- pel
producerende fenomenen.
Het eerste fenomeen is de interactie tussen de permanente magneten
en het magnetisch veld opgewekt door de wisselende stator stromen.
Deze interactie produceert het dominante kop- pel en wordt
aangeduid als TPM , het koppel geproduceerd in de magneten. De
andere component wordt geproduceerd door de interactie tussen de
wervelstromen in de shielding cylinder en het magnetisch veld van
de permanente magneten. Deze com- ponent wordt aangeduid als TSC ,
het koppel geproduceerd in de shielding cylinder. Merk op dat TPM
en TSC fictieve compo- nenten zijn, ze komen niet afzonderlijk voor
maar worden op
0 0.1 0.2 0.3 0.4 −2
1
4
7
T PM
T SC
Fig. 1. Het totale koppel en zijn componenten
deze manier beschouwd om de fysische achtergrond van de ma- chine
beter te begrijpen.
IV. WERVELSTROMEN
Kennis over de wervelstromen die genduceerd worden in de shielding
cylinder en de permanente magneten is cruciaal voor het berekenen
van de verliezen die ze teweegbrengen, deze ver- liezen zijn een
belangrijke ontwerpfactor voor elektrische ma- chines. In een
bewegende geleider is de stroomdichtheid J gerelateerd aan het
elektrisch veld E en het magnetisch veld B dat de gelei- der
omvat:
J = σ (E + v × B) (2)
Voor de permanente magneten moet een constante toege- voegd worden
om ervoor te zorgen dat de netto stroom in elke magneet nul is. In
een 2D benadering is het mogelijk dat stro- men zich sluiten in een
andere magneet. Stroom die vloeit van de ene magneet naar de andere
is natuurlijk onmogelijk. Door- dat de shielding cylinder een
geheel is en dus kortgesloten is, is het mogelijk dat er stromen
sluiten in een andere regio van de shielding cylinder zonder te
zondigen tegen de wet van be- houd van energie. Hierdoor is het
niet nodig deze constante in de shielding cylinder te
gebruiken.
Voor de permanente magneten moet een constante toege- voegd worden
om ervoor te zorgen dat de netto stroom in elke magneet nul is. In
een 2D benadering is het mogelijk dat stro- men zich sluiten in een
andere magneet. Stroom die vloeit van de ene magneet naar de andere
is natuurlijk niet mogelijk. Door- dat de shielding cylinder een
geheel is en dus kortgesloten is, is het mogelijk dat er stromen
sluiten in een andere regio van de shielding cylinder zonder te
zondigen tegen de wet van behoud van energie. Hierdoor is het niet
nodig deze constante in de shielding cylinder te gebruiken.
V. WERVELSTROOM VERLIEZEN
Om de wervelstroom verliezen in de shielding cylinder en de
permanent magneten te berekenen worden er twee methodes ge- bruikt.
De verliezen in de permante magneten worden berekend
met de formule van Joule:
P (t) =
dv (3)
Deze formule is echter niet effectief om de verliezen in de
shielding cylinder te berekenen doordat het r-afhankelijk deel van
J3(r, φ, t) Bessel functies bevat, deze functies kunnen niet
analytisch gentegreerd worden. Om dit probleem op te lossen worden
de verliezen in de shielding cylinder berekent door ge- bruik te
maken van de Poynting vector S:
S = E × H (4)
De Poynting vector representeert de hoeveelheid energie die per
tijdseenheid door een eenheidsoppervlakte stroomt. De in- tegratie
van S over een gesloten oppervlak bepaalt de energie die door dat
oppervlak stroomt. In dit geval betekent dit dat het totaal
vermogen P dat overgedragen wordt van de stator naar de rotor
berekend kan worden door de integratie van de Poynting vector over
een oppervlak dat de rotor omvat.
VI. PARAMETER STUDIE
A. Koppel
Zoals verwacht toont TPM slechts een beperkte afhankelijk- heid van
de geleidbaarheid van de shielding cylinder. TPM is vooral
veroorzaakt door de interactie van synchrone veldcom- ponenten en
de verwachting was dat deze componenten niet benvloed worden door
de geleidbaarheid van de shielding cy- linder. Maar wanneer deze
afzonderlijk wordt geplot kan er toch een kleine variatie
waargenomen worden. De reden voor deze variatie is dat een deel van
TPM geproduceerd wordt door de interactie tussen asynchrone
veldcomponenten aan de grens tussen de luchtspleet en de sloten. De
evolutie van TSC is gelijkaardig aan de koppel-toerental ka-
rakteristiek van een inductie machine en kan ook verklaard wor- den
door de shielding cylinder te beschouwen als de kooian- ker van een
inductie machine. Merk op dat TSC negatief is, dit houdt in dat
asynchrone componenten met een tragere draaisnel- heid dan de
synchrone componenten dominant zijn.
0 2 4 6 8 10
x 10 7
T or
qu e
(N m
T SC
Fig. 2. Het gemiddelde koppel in functie van de geleidbaarheid van
de SC
De tweede parameter die bestudeerd wordt is de dikte van de
shielding cylinder. De evolutie van TSC kan verklaard worden
door de shielding cylinder te bekijken als een elektrische gelei-
der en de wet van Pouillet toe te passen. Dit houdt in dat de
weerstand van de shielding cylinder daalt als de dikte vergroot.
Dit verklaart de initiele stijging van TSC . Als de dikte verder
vergroot is er een daling van beide componenten door het ver-
groten van de effectieve luchtspleet. De shielding cylinder blijkt
gunstig te zijn bij het verminderen van de rimpel op het totale
koppel aangezien een stijging van de dikte een significante da-
ling van de koppelrimpel van TPM teweeg brengt. Dit kan inte-
ressant zijn voor applicaties die een constant koppel
vereisen.
0 1 2 3 4 5
x 10 −3
T SC
Fig. 3. Het gemiddelde koppel in functie van de dike van de
SC
B. Gedissipeerd vermogen
De shielding cylinder blijkt ook gunstig te zijn in het verlagen
van het totaal gedissipeerd vermogen. Door het shielding effect
worden er minder wervelstromen genduceerd in de permanente magneten
die een relatief hoge weerstand hebben. Deze wer- velstromen worden
in plaats daarvan genduceerd in de shiel- ding cylinder, welke een
lagere weerstand heeft. Hierdoor wordt het totaal gedissipeerd
vermogen significant gereduceerd zoals te zien is op Figuur
4.
0 2.6
P ow
SC
Fig. 4. Het gedissipeerd vermogen in functie van de geleidbaarheid
van de SC
REFERENCES
[1] A. Hemeida, B. Hannon, H. Vansompel, and P. Sergeant,
“Comparison of three analytical methods for the precise calculation
of cogging torque and torque ripple in axial flux pm machines,”
MATHEMATICAL PROBLEMS IN ENGINEERING, vol. 2016, pp. 1–14,
2016.
[2] K. Ramakrishnan, M. Curti, D. Zarko, G. Mastinu, J. J. H.
Paulides, and E. A. Lomonova, “A comparison study of modelling
techniques for per- manent magnet machines,” in 2016 Eleventh
International Conference on Ecological Vehicles and Renewable
Energies (EVER), pp. 1–6, April 2016.
[3] B. Hannon, P. Sergeant, and L. Dupre, “Voltage Sources in 2d
Fourier- Based Analytical Models of Electric Machines,”
Mathematical Problems in Engineering, vol. 2015, p. e195410, Oct.
2015.
[4] Z. Q. Zhu, L. J. Wu, and Z. P. Xia, “An accurate subdomain
model for mag- netic field computation in slotted surface-mounted
permanent-magnet ma- chines,” IEEE Transactions on Magnetics, vol.
46, pp. 1100–1115, April 2010.
[5] T. Lubin, S. Mezani, and A. Rezzoug, “2-d exact analytical
model for surface-mounted permanent-magnet motors with semi-closed
slots,” IEEE Transactions on Magnetics, vol. 47, pp. 479–492, Feb
2011.
[6] Z. J. Liu and J. T. Li, “Analytical solution of air-gap field
in permanent- magnet motors taking into account the effect of pole
transition over slots,” IEEE Transactions on Magnetics, vol. 43,
pp. 3872–3883, Oct 2007.
[7] K. Ramakrishnan, M. Curti, D. Zarko, G. Mastinu, J. J. H.
Paulides, and E. A. Lomonova, “Comparative analysis of various
methods for modelling surface permanent magnet machines,” IET
Electric Power Applications, vol. 11, no. 4, pp. 540–547,
2017.
[8] H. V. Xuan, D. Lahaye, H. Polinder, and J. A. Ferreira,
“Influence of stator slotting on the performance of
permanent-magnet machines with concen- trated windings,” IEEE
Transactions on Magnetics, vol. 49, pp. 929–938, Feb 2013.
[9] B. Hannon, P. Sergeant, and L. Dupre, “2d analytical torque
study of slot- less and slotted pmsm topologies at high-speed
operation,” in Proceedings of ElectrIMACS 2014, pp. 248–231,
Universitat Politecnica de Valencia, 2014.
[10] P. D. Pfister, X. Yin, and Y. Fang, “Slotted permanent-magnet
machi- nes: General analytical model of magnetic fields, torque,
eddy currents, and permanent-magnet power losses including the
diffusion effect,” IEEE Transactions on Magnetics, vol. 52, pp.
1–13, May 2016.
[11] F. Zhou, J. Shen, W. Fei, and R. Lin, “Study of retaining
sleeve and con- ductive shield and their influence on rotor loss in
high-speed pm bldc mo- tors,” IEEE Transactions on Magnetics, vol.
42, pp. 3398–3400, Oct 2006.
[12] D. J. Griffiths, Introduction to Electrodynamics. Pearson
Education, 2007. [13] B. Hanon, P. Sergeant, and L. Dupre, “2d
analytical torque study of slotless
and slotted pmsm topologies at high-speed operation.” none. [14] B.
Hannon, P. Sergeant, and L. Dupre, “2-d analytical subdomain model
of
a slotted pmsm with shielding cylinder,” IEEE Transactions on
Magnetics, vol. 50, pp. 1–10, July 2014.
[15] B. Hanon, P. Sergeant, and L. Dupre, “Torque and torque
components in high-speed permanent-magnet synchronous machines with
a shielding cy- linder.” none.
[16] A. Tenconi, S. Vaschetto, and A. Vigliani, “Electrical
machines for high- speed applications: Design considerations and
tradeoffs,” IEEE Transac- tions on Industrial Electronics, vol. 61,
pp. 3022–3029, June 2014.
[17] B. Hanon, P. Sergeant, and L. Dupre, “Two-dimensional
fourier-based mo- deling of electric machines.” none.
[18] Z. Tian, C. Zhang, and S. Zhang, “Analytical calculation of
magnetic field distribution and stator iron losses for
surface-mounted permanent magnet synchronous machines,” Energies,
2017.
[19] J. J. Lee, W. H. Kim, J. S. Yu, S. Y. Yun, S. M. Kim, J. J.
Lee, and J. Lee, “Comparison between concentrated and distributed
winding in ipmsm for traction application,” in 2010 International
Conference on Electrical Ma- chines and Systems, pp. 1172–1174, Oct
2010.
[20] M. Merdzan, J. J. H. Paulides, and E. A. Lomonova,
“Comparative analy- sis of rotor losses in high-speed permanent
magnet machines with different winding configurations considering
the influence of the inverter pwm,” in 2015 Tenth International
Conference on Ecological Vehicles and Renewa- ble Energies (EVER),
pp. 1–8, March 2015.
[21] S. Niu, S. L. Ho, W. N. Fu, and J. Zhu, “Eddy current
reduction in high-speed machines and eddy current loss analysis
with multislice time-stepping finite-element method,” IEEE
Transactions on Magnetics, vol. 48, pp. 1007–1010, Feb 2012.
[22] D. A. Co Huynh, Liping Zheng, “Losses in high speed permanent
magnet machines used in microturbine applications,” J. Eng. Gas
Turbines Power 131(2), 022301 (Dec 23, 2008) (6 pages), p. 6,
2008.
Contents
2.2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . 9
3.3 Torque calculation and validation . . . . . . . . . . . . . . .
. . . . . . . . 13
3.3.1 Torque calculation . . . . . . . . . . . . . . . . . . . . .
. . . . . . 13
3.3.2 Torque components . . . . . . . . . . . . . . . . . . . . . .
. . . . . 14
4 Eddy currents 19
5 Eddy-current losses 25
6.3.1 Parameters . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . 40
8.3 Torque . . . . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . 54
8.5.1 Derivation . . . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . 57
3.1 radial flux density . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . 14
3.2 Harmonic content Br . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . 15
3.3 machine topology torque inner rotor - open slots - shielding
cylinder . . . 16
3.4 Validation total torque inner rotor - open slots - SC - no load
. . . . . . . 18
4.1 Eddy current density in the PM . . . . . . . . . . . . . . . .
. . . . . . . . 20
4.2 Eddy currents comparison ANA (a) vs FEM (b) in the SC . . . . .
. . . . 21
4.3 Eddy current density in the SC . . . . . . . . . . . . . . . .
. . . . . . . . 22
4.4 Eddy current density in r3 . . . . . . . . . . . . . . . . . .
. . . . . . . . . 23
4.5 Eddy current density in r2 . . . . . . . . . . . . . . . . . .
. . . . . . . . . 23
4.6 Eddy current density in the PM . . . . . . . . . . . . . . . .
. . . . . . . . 24
4.7 Eddy current density in r1 . . . . . . . . . . . . . . . . . .
. . . . . . . . . 24
6.1 Machine topology parameter study torque . . . . . . . . . . . .
. . . . . . 31
6.2 Harmonic content of the studied machine Br . . . . . . . . . .
. . . . . . . 32
6.3 Harmonic content of the studied machine Bφ . . . . . . . . . .
. . . . . . . 33
6.4 Average torque as a function of the conductivity of the SC . .
. . . . . . . 34
6.5 Illustration of the interaction between synchronous- and
asynchronous field components at the slot opening . . . . . . . . .
. . . . . . . . . . . . 34
6.6 Torque in the PM as a function of the conductivity of the SC .
. . . . . . . 35
6.7 Torque ripple in the PM as a function of the conductivity of
the SC . . . . 38
6.8 Average torque as a function of the thickness of the SC . . . .
. . . . . . . 39
6.9 Torque ripple as a function of the thickness of the SC . . . .
. . . . . . . . 40
xvii
6.10 machine topology parameter study Eddy current losses . . . . .
. . . . . . 41
6.11 harmonic content of the studied machine Br . . . . . . . . . .
. . . . . . . 42
6.12 harmonic content of the studied machine Bφ . . . . . . . . . .
. . . . . . . 43
6.13 Eddy current loss as a function of the conductivity of the SC
. . . . . . . . 44
6.14 Eddy current loss as a function of the conductivity of the SC
and the penetration depth of the dominant asynchronous field
components . . . . . 45
6.15 Induced eddy currents as a function of the conductivity of the
SC . . . . . 46
6.16 the square of the induced eddy currents as a function of the
conductivity of the SC . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . 47
6.17 Resistivity of the SC as a function of the conductivity of the
SC . . . . . . 47
6.18 Eddy current losses as a function of the thickness of the SC .
. . . . . . . . 48
6.19 Eddy current losses as a function of the thickness of the SC .
. . . . . . . . 48
xviii
xix
Symbol Definition δi,j Kronecker delta λ
(ν) k Eigenvalue related to the spatial-aspect of the solution λ(ν)
n Eigenvalue related to the time-aspect of the solution µ2
Permeability of the magnets µ3 Permeability of the sleeve φm
Angular span of a magnet φms Start angle of a magnet ρ Charge
density σ Conductivity σ2 Conductivity of the magnets σ3
Conductivity of the shielding cylinder σPM Conductivity of the
magnets σSC Conductivity of the shielding cylinder Jeddy Eddy
current density Jext Externally imposed current density A Magnetic
vector potential B Magnetic flux density Bm Residual flux density
of the magnets D Electric flux density E Electric field strength EC
Eddy currents F Total magnetic force f Force per unit volume fnom
Nominal frequency H Magnetic field strengh Inom Current density J
Current density ls Stack length
Symbol Definition m Number of phases MQS Magnetoquasistatic MV P
Magnetic vector potential N Number of windings per slot nnom
Nominal rotational speed Ns Number of slots p Number of pole pairs
PM Permanent magnets PMSM Permanent magnet synchronous machine r1
Outer rotor iron radius r2 Outer PM radius r3 Outer sleeve radius
r4 Outer airgap radius r5 Outer tooth tip radius r6 Outer winding
radius r7 Outer stator iron radius rac Radius at the center of the
airgap rmc Radius at the center of the magnet S Surface S Poynting
vector Sms Surface area of a single magnet SC Shielding cylinder SM
Surface-mounted T Maxwell stress tensor T Torque TPM Torque
produced in the permanent magnets TSC Torque produced in the
shielding cylinder Tt Time periodicity V Volume wτ tooth-tip width
wt Tooth width
Chapter 1
Introduction
Designing high-speed electrical machines is a significantly more
challenging task than traditional machines. Due to the high
operating speed of the machines, rigorous me- chanical, thermal and
electromagnetic conditions need to be withstood. In the field of
high-speed electrical machines the two most represented types of
machines are in- duction machines and Permanent Magnet Synchronous
Machines(PMSMs). Because of the higher power density achievable
with PMSM, especially at very high speed, the focus nowadays leans
towards PMSMs. These machine type will be the main focus of this
work, exclusively the Surface-Mounted (SM) PMSMs with a retaining
sleeve. This work will focus on the electromagnetic aspect of these
machines such as eddy currents, eddy current losses and torque.
Moreover a study of the effect of the so called Shielding Cylinder
on these properties will be conducted. A Fourier based modeling
technique is used to implement and study the mentioned
properties.
This work consists of three parts. First, the modeling technique is
shortly explained. The second part illustrates the implementation
of the studied physical properties in the Fourier based model. The
last part is the study itself where the considered properties are
studied in regard to varying parameters.
1
Chapter 2
Fourier-Based Modeling
2.1 Introduction
The main goal of this work is to perform an analytical study of
various physical proper- ties of high speed machines. An analytical
model is chosen because of various reasons. First, analytical
models have proven to require less computational time [1], [2],
although numerical approaches provide high accuracy results due to
their ability to include com- plex geometries [3]. Secondly, when
using readily available numerical software, less machine insight is
re- quired while developing the model. Developing an analytical
model requires a more pro- found knowledge of the machine’s
operation and the underlaying physics. Thus using an analytical
model is in the authors best interest in gaining knowledge on the
subject. Thirdly, analytical models are easier to parametrize,
which is a key feature in this work. This chapter summarizes some
of the basics of the analytical model used. It is mainly based on
the authors’ thesis counselor’s work [[4],[5],[6]].
2.2 Fourier-Based Modeling
The goal of this technique is to solve Maxwell’s equations in the
entire geometry. To do so Maxwell’s equations are reformulated to a
single partial-differential equation using a magnetic potential,
that equation is referred to as the governing equation. The machine
is divided into subdomains in which the governing equation is
simplified and solved us- ing the separation of variables
technique. Using this technique implies that the result will
contain integration constants and eigenvalues. These are determined
by imposing boundary conditions, these conditions also ensure a
coherent solution over the entire geometry. The modeled machines in
this work are radial-flux rotational machines. A cylindrical
coordinate system (r, φ, z) is used to describe the geometry, where
the z-axis is placed along the machine’s axis.
3
2.2.1 Assumptions
In order to enable an analytical solution of the problem with
limited mathematical complexity, multiple assumptions are made.
First, the situation is assumed to be magnetoquasistatic (MQS),
this implies two as- pects. The magneto aspect dictates that the
magnetic field is dominant in regard to the electric field. The
quasistatic aspect implies that time-dependent variations are
propa- gated instantly throughout the entire geometry. Secondly the
problem is assumed invariant along the z-axis, only two dimensions
are considered. Thirdly, the boundaries of the chosen subdomains
are either radial, i.e a constant φ, or circumferential, i.e a
constant r. This assumption is required to solve the problem
analytically, boundaries which aren’t radial or circumferential
need to be simplified. Fourthly, it is assumed that all movements
occur along the tangential direction. This means that all moving
parts rotate around the same axis. This axis has to coincide with
the longitudinal direction. Fifthly, all of the materials are
assumed linear and isotropic. Furthermore soft-magnetic material
are assumed to have an infinite magnetic permeability. Sixthly, the
conductivity (σ) of the magnet region is assumed zero, so that no
eddy cur- rent can be induced in the magnets. Finally, the machine
is assumed to be operating in steady-state, this implies that tran-
sient effects are not regarded, thus only time periodic components
are present. This as- sumption is required because the FBM is based
on periodicities in the magnetic field.
2.2.2 Potential formulation
The goal of Fourier-based modeling is to solve Maxwell’s equations,
These can be writ- ten as follow in the MQS situation:
∇× E = −∂B ∂t
∇×H = J (2.1b) ∇ ·B = 0 (2.1c)
Note that the influence of the electric charges is neglected,
because of the dominance of the magnetic field. Due to the
difficulty in solving these equations analytically, the problem is
reformulated using a magnetic potential; the Magnetic Vector
Potential (MVP). The MVP, which is a vector is defined by its
curl:
∇×A = B (2.2)
CHAPTER 2. FOURIER-BASED MODELING
For a complete description of MQS problems, the MVP has to be
combined with the electric scalar potential V, which is defined by
its gradient.
∇V = − (
) (2.3)
The following partial-differential equation for Maxwell’s equations
(2.1) can be obtained from the above definitions and the
constitutive relations:
∇2A−∇ ·A− µσ∂A ∂t
+ µσ(v× (∇×A)) = µσ∇V −∇×B0 (2.4)
It should be noted that the MVP is not uniquely defined,so gauge
fixing is necessary. the Coulomb gauge (∇ ·A = 0) is chosen as is
common in quasi static approximations.
Equation (2.4) can be physically interpreted by considering two
current densities, the eddy current density (Jeddy) and the
externally imposed current density (Jext). The eddy current density
is accounted for by the time-derivative and speed-dependent terms
of (2.4). The externally imposed current densities are accounted
for by the µσ∇V term. This implies that each of the machine’s
conductors should be modeled separately. As this would result in an
augmented computational burden, the externally imposed cur- rent
density is directly accounted for. This is done by substituting
Jext for σ∇V in (2.4). Resulting from these changes, a governing
equation for the MVP is obtained:
∇2A− µσ∂A ∂t
2.2.3 Subdomain technique
The governing equation (2.5) is too complex to be solved in the
entire geometry. In or- der to avoid that complexity, the geometry
is divided in regions called subdomains. In each chosen subdomain
the problem is simplified and can therefore be solved. To en- sure
a physically correct behavior of the magnetic vector potential,
boundary conditions are imposed. In choosing subdomains, two
aspects should be considered. First, the gov- erning equation
should be simplified as much as possible and secondly, boundary
con- ditions should be relatively easy to impose. These aspects are
further discussed in this section.
Subdomains
In order to simplify (2.5), dividing the geometry in subdomains
will reduce the num- bers of terms in the equation. Three types of
terms can be recognized in the governing equation (2.5). First, the
Laplacian term ∇2A is a general term that is present in every
subdomain.
5
CHAPTER 2. FOURIER-BASED MODELING
Secondly, the time-derivate and speed-dependent terms account for
the eddy-currents. These terms are only present in conductive
subdomains where eddy-currents are non- negligible. Last are the
terms on the right hand side, these are sources terms. Jext ac-
counts for externally imposed current densities and B0 accounts for
the permanent magnets. The equation is also greatly simplified if σ
and µ are constant, subdomain are chosen taking this in account.
The model used in this work does not consider sudo- mains with
multiple source terms or subdomains where both a source term and
eddy- currents are present. This implies that only the following
governing equations are con- sidered:
∇2A = 0 (2.6a) ∇2A = µJext (2.6b) ∇2A = −∇×B0 (2.6c)
∇2A = ∂A ∂t
+ µσ(v× (∇×A)) (2.6d)
Where the simplest equation (2.6a) applies to subdomains with no
source terms and no eddy currents. Equations (2.6b) and (2.6c)
applies to subdomains where source terms are present. Finally
(2.6d) applies to subdomains where the eddy-currents are non-
negligible.
Boundary conditions
The boundary conditions ensure a physically correct behavior of the
MVP. Initially a set of boundary conditions for B,D,H and E are
considered. Because the situation is assumed MQS and the focus is
thus on the magnetic field, only the conditions for B and H are
regarded. These are the following:
n · ( B(ν) −B(ν+1)
) = Js
(2.7)
where n is the normal vector from subdomain ν to ν + 1 and Js is
the current density on the boundary. In a 2-D model, Js is a line
current density which is not considered in this model, thus Js is
always zero. Using the magnetic vector potential definition (2.2)
and the constitutive relations in the MQS situation (see appendix
8.2) the boundary condition can be written as:
A(ν)(rν , φ, t)−A(ν+1)(rν , φ, t) = 0 (2.8a)( ∂A(ν)(r, φ, t)
µ(ν)∂r − ∂A(ν+1)(r, φ, t)
µ(ν+1)∂r
CHAPTER 2. FOURIER-BASED MODELING
Where (2.8a) imposes the continuity of A at the circumferential
boundary between sub- domains ν and ν + 1. Condition (2.8b)
expresses the continuous behavior of the tangen- tial component of
the magnetic field strength. As stated in assumption (2.2.1), a
subdomain is assumed to have four boundaries. Two of which are
constant along the normal r direction and two along the φ periodic
direc- tion. A subdomain is referred as periodic or non-periodic,
depending on whether or not they span the entire geometry along the
periodic direction. In the spherical coordinate system a periodic
subdomain is one which spans 2π radians.
2.2.4 Form of the solution
Equations (2.6) are solved using the separation of variables
technique. The result con- tains both eigenvalues and integration
constants. The eigenvalues, which determine the periodicity of the
solution are found by imposing the periodic boundary conditions.
The circumferential boundary conditions of all the boundaries form
a set of equations which determines the integration constants. The
solution in subdomain ν is written as:
Aν(r, φ, t) = ∑ λ
(ν) k
(r) is the r-dependent Fourier coefficient of the series, this
coefficient contains the integration constants. λ(ν)
n is the eigenvalue related to the time-aspect of the solution.
λ(ν)
k the eigenvalue related to the spatial-aspect of the solution. As
stated above the eigenvalues are determined by the subdomain’s
periodicity. The time period- icity (Tt) is the time the machine
needs to perform one revolution and is equal for all the
subdomains. The spatial periodicity of a subdomain is T (ν)
s radians. This periodic- ity is either 2π, for periodic
subdomains, or 2βν , for non-periodic subdomains with soft magnetic
boundaries. The eigenvalues can then be written as:
λ(ν) n = 2nπ
Tt = nω (2.10a)
(2.10b)
where ω is the angular speed of the machine, n and k are the time-
and spatial har- monic order respectively. The solution for a
periodic subdomain can be written as:
Aν(r, φ, t) = ∑ λ
and for a non periodic subdomain with soft-magnetic boundaries
as:
Aν(r, φ, t) = ∑ λ
j( kπβν φ−nωt) (2.12)
As the equation is a summation over every harmonic combination, the
governing equa- tion can be considered for every time- and
spatial-harmonic combination separately. This allows simplification
of the time-derivative and speed-dependent terms of the gov- erning
equations (2.6). From the final form of solution (2.9), the time
derivate can eas- ily be calculated for every time- and
spatial-harmonic combination:
∂A(ν) n,k
∂t = jnωA(ν)
n,k (2.13)
The time-derivate of A(ν) n,k can thus be written as the product of
a constant and A(ν)
n,k
itself. Another important term of the governing equation is the
speed-dependent term. As assumed in section (2.2.1), the speed of
subdomain υ only has a φ-dependent term, i.e. v(υ) = r(υ)eφ. The
speed-dependent term can then also be calculated for every harmonic
combination based on (2.9):
v(ν) × ( ∇×Aν
(ν)A(ν) n,k (2.14)
Similar to the time-derivate, the speed-dependent term can be
written as the product of a constant and A(ν)
n,k itself. Equations (2.6) can now be rewritten for an arbitrary
harmonic combination (n, k) as:
∂2A(ν) n,k
) A(ν) n,k (2.15d)
The governing equations (2.6) can now be solved using the
separation of variables tech- nique. As an example the first
equation (2.15a) is briefly discussed. Equation (2.15a) is a
Laplace equation, it is applicable to regions where there is no
magnetic field and the eddy-current density is negligible. These
conditions are met in the air gap region, which is a periodic
subdomain so Ts = 2π. The solution of the equa- tion follows the
form given in (2.9) and is written as:
Aν(r, φ, t) = ∞∑
8
C (ν) n,kr
where C(ν) n,k and D(ν)
n,k are the integration constants. The other solutions can be found
in [5].
2.2.5 Conclusion
This chapter covered the basics of the Fourier-based model that is
used in this work. The strategy of the technique, the necessary
assumptions and the solutions for this model were briefly
discussed. Note that there are many aspects which have not been
discussed for the sake of simplicity. Some additional information
can be found in the appendix. The authors made use of an existing
model provided by faculty researcher and thesis counselor ing. Bert
Hannon. The interested reader can find more in depth information
regarding analytical modeling technique in [[4],[6],[7],[8],
[9]].
9
3.1 Introduction
In this chapter an analytical method for torque calculation is
introduced. The calcula- tion is based on the Maxwell’s stress
tensor, which is briefly described in this chapter. Secondly the
implementation in the Fourier-based model is presented and the two
con- sidered torque components are described. Finally a validation
is provided by comparing the analytical results with finite element
calculations. This torque calculation method is later used in the
parameter study.
3.2 Maxwell stress tensor
The calculation of the Maxwell’s stress tensor is based on the
total magnetic force F on a rigid body with volume V and surface S
which is placed in a magnetic field. The derivation starts with a
calculation of the total force due to electromagnetic fields on the
charges and current within a volume V . Lorentz force law
states:
F = ˆ V
ρ(E + v×B)dV
(ρE + J×B)dV (3.1)
The integrand can be seen as a force per unit volume f in a region
with magnetic flux density B and current density J.
f = ρE + J x B (3.2)
11
CHAPTER 3. TORQUE
By using Maxwell’s equations, Faraday’s law and the product rule of
differential calcu- lus, f can be found as:
f = ε0[(∇ · E)E + (E · ∇)E] + 1 µ0
[(∇ ·B)B + (B · ∇)B]
(3.3)
The Maxwell stress tensor T can now be introduced, which is a 3 × 3
matrix whose components are defined as:
Ti,j = ε0(EiEj − 1 2δi,jE
where δi,j represents Kronecker delta.
δi,j = 1, if i = j,
0, if i 6= j.
The subscripts i and j indicate the ijth component of the matrix.
Those components each represent a unit of force per unit of area.
Which can be interpreted as the force along the ith axis
experienced by a surface normal to the jth axis, per unit of area.
The diagonal elements represent pressures and the non-diagonal
elements are shears stresses. The above equation can be used for
force calculations in a Cartesian coordinate sys- tem. However this
study is conducted in a cylindrical coordinate system, the equation
is therefore rewritten accordingly. Furthermore the field is mainly
magnetic so that a number of terms can be dropped and the equation
can be simplified to
Tr,t = 1 µ0
(BrBt − 1 2δr,tB
2) (3.5)
where r is the radial direction and t is the shear in the
tangential direction. The tan- gential component is the one that
expresses torque in a motor, it is that component that is focused
on later. The total force on volume V can be calculated by
integrating the force per unit of area for each element of the
Maxwell stress tensor over a surface S:
F = S
TdS (3.6)
Te = S
3.3.1 Torque calculation
Using the Maxwell stress tensor, the electromagnetic torque of a
rotating electrical mo- tor can be calculated. As described in the
previous section a surface integral must be chosen, the considered
surface must surround the studied motor. In a 2D model the surface
integral is reduced to a line integral along the air gap. If a
circle with radius r is considered as the integration path, the
torque can be calculated as:
Te = 1 µ0
0 r2BrBφdφ (3.8)
This result can now be implemented in the Fourier based model.
Where the radial (Br) and tangential (Bφ ) components of the flux
density are calculated by applying (2.2).
B (3i) r,k,n = 1
(3.9)
(3.10)
The flux components are calculated in the air gap region following
the surface integral chosen for the torque calculations.
Substituting (3.9) and (3.10) in (3.8) and accounting for the shaft
length ls.
T (t) = lsr 2 ac
µ0
2πˆ
0
φ (rac, φ, t) dφ (3.11)
By using the periodicity of the MVP, (3.11) can be simplified
to:
T (t) = 2π lsr 2 ac
µ0
∞∑ n=−∞
∞∑ s=−∞
Tn,se−j(n+s)t (3.12)
Where n is the time harmonic order related to tangential flux
component (B(3i) φ ) and s
the time harmonic order related to the radial flux component (B(3i)
φ ) and Tn,s is calcu-
lated as: Tn,s =
B (3i) r,k,n(rac)B(3i)
φ,−k,n(rac) (3.13)
Equation (3.12) is implemented in the Fourier based model and used
to calculate the electromagnetic torque.
13
CHAPTER 3. TORQUE
3.3.2 Torque components
The total electromagnetic torque in a machine equipped with a
shielding cylinder is a combination of two torque producing
phenomena. First of which is the interaction of the permanent
magnets and the magnetic field due to the alternating stator
current. This interaction produces the dominant torque and is
referred to as TPM , the torque produced in the magnets. TPM is
calculated by consider- ing the surface surrounding the rotor yoke
and the magnet ring. The other torque component is produced by the
interaction of eddy currents in the shielding cylinder and the
permanent magnet’s magnetic field. These eddy currents are
introduced by two phenomena. First, the magnetic flux density in
the air gap fluctuates as it passes from a point opposite to a
slot, to a point opposite to a stator tooth. This because of the
difference in permeability between the stator iron (high
permeability) and the air in the slot opening (low permeability).
This is known as the slotting effect.
0 1 2 3 4 5 6 −0.8
−0.6
−0.4
−0.2
0
0.2
0.4
0.6
0.8
Figure 3.1: radial flux density
These fluctuations can be observed on Figure 3.1, where the
magnetic field density is plotted in function of the angle φ. The
dips at the top are partially due to the slotting effect. With the
aid of Fourier analysis, the periodic wave of the flux density can
be de- composed in multiple harmonic components. this is shown in
Figure 3.2. It can be seen that beside the fundamental harmonics, a
number of asynchronous field components are presents, these
represent the effect of the slotting effect.
14
Figure 3.2: Harmonic content Br
These changes in magnetic flux density induces eddy currents in the
shielding cylinder, which in interaction with the magnets produces
torque. Secondly, eddy currents are induced due to the alternating
currents in the stator, this is similar to the principle of a
squirrel cage rotor in an induction machine. The resulting torque
component from both effects is referred to as TSC , the torque
produced in the shielding cylinder. It can be calculated as the
difference of the total torque T and the torque in the magnets TPM
:
TSC(t) = T (t)− TPM(t) (3.14)
It should be noted that TPM and TSC are fictitious components, they
do not occur sep- arately but are regarded as such to provide a
better understanding of the machine’s physics.
15
3.3.3 Finite element torque validation
Now that an analytical implementation of the torque has been
derived, it needs to be validated. The analytically calculated
torque is compared with a finite element torque calculation. The
software used to conduct the FEM analysis is COMSOL Multiphysics.
The considered topology is a surface mounted magnets PMSM with open
slots and a shielding cylinder, the machine topology is shown in
Figure 3.3.
Figure 3.3: machine topology torque inner rotor - open slots -
shielding cylinder
A FEM model of the depicted topology is created in COMSOL, along
with the various machine and material parameters. An overview of
the model parameters is listed in ta- ble 3.1.
16
Table 3.1: Parameters of the studied machines
Parameter Symbol Value Number of slots Ns 12 Number of pole pairs p
2 Number of phases m 3 Number of layers / 1 Residual flux density
of the magnets Bm 1.2 T Angular span of a magnet φm 80.00 Current
density Inom 5.106 A
m2
Number of windings per slot N 6 Nominal frequency fnom 1000 Hz
Nominal speed nnom 30,000 rpm Outer rotor iron radius r1 22.50 mm
Outer PM radius r2 25.00 mm Outer sleeve radius r3 26.50 mm Outer
airgap radius r4 28.50 mm Outer winding radius r5 40.40 mm Outer
stator iron radius r6 50.00 mm Stack length ls 200 mm Permeability
of the magnets µ2 1.0µ0
H M
17
1
4
7
T PM
T SC
Figure 3.4: Validation total torque inner rotor - open slots - SC -
no load
Figure 3.4 shows a comparison of the different torque components
between the FEM and the analytical calculations. A good agreement
can be observed for all the torque components (T , TPM and TSC).
From these results, it can be concluded that the im- plemented
torque calculation is valid. In Chapter 6, a parameter study of the
torque is conducted using the above described calculation
method.
18
Chapter 4
Eddy currents
4.1 Introduction
In this chapter an analytical eddy current density calculation is
presented. Knowledge of the eddy currents in a machine is crucial
for the calculation of eddy current losses, which is an important
design factor for electrical machines. The eddy current density is
calculated in the permanent magnets (PM) and the shielding cylinder
(SC). A valida- tion of these calculations is provided by comparing
the analytical results with a finite element calculation.
4.2 Eddy current calculation
In a moving conductive material, the current density J is related
to the electrical field E and the magnetic field B surrounding the
conductor.
J = σ (E + v×B) (4.1)
No external currents are applied in the magnets and the shielding
cylinder, so the elec- tric field E can be written as:
E = ∂A ∂t
(4.2)
By using (4.2) and the definition of the magnetic vector potential
A ( 2.2 ) the current density J can be written as:
J = σ
( ∂A ∂t
+ v× (∇×A) )
(4.3)
By using the properties of vector calculus and (2.2) for A , the
following is obtained for the SC:
19
kA (3) n,k(r, φ)ej(kφ−nωt) (4.4)
For the permanent magnets, a constant (C(t)) needs to be introduced
to ensure that the net current in the magnet region is zero. This
because in a 2D approximation, it is possible for currents to be
closed in another magnet. Current flowing from one magnet to the
other is not physically correct. When the net current through one
magnet is not zero, eddy currents from another magnet are closed
through the regarded magnet or eddy currents from the regarded
magnet are closed through another magnet. The con- stant C(t) is
the sum of the current flowing through all the magnets. Subtracting
this value from each magnet, ensures that the net current in the
magnet region is zero. Because the shielding cylinder is made of
one single piece and thus short-circuited, it is possible for
current loops to be closed in an other region of the shielding
cylinder with- out violating the law of conservation of energy.
Thus it is not necessary to implement this spatially constant
current density in the shielding cylinder.
Ct(t) = 1 Sms
ˆ φms(t)+φm
) rdrdφ (4.5)
Where φms(t) is the start angle of a magnet and φm is the magnet
arc angle. Sms is the surface area of a single magnet. The current
density J in the PM is given by:
J = jσ(2)ω (n− k) ∞∑
−6
−4
−2
0
2
4
6
20
(b)
Figure 4.2: Eddy currents comparison ANA (a) vs FEM (b) in the
SC
21
CHAPTER 4. EDDY CURRENTS
4.2.1 Eddy current validation
Yet again, the analytical implementation needs to be validated. To
validate the Eddy current density calculation, the same FE model is
used as in Section 3.3.3. The machine used for the calculations is
a three-phase permanent-magnet synchronous machine with
surface-mounted magnets and a shielding cylinder. The machine
topology is shown in Figure 3.3, an overview of the parameters is
listed in Table 3.1. In Figure 4.2, the con- tour plot of the eddy
currents in the SC calculated with the analytical model and with
the FE model is shown. It can be seen that these results show a
good agreement be- tween the analytical and the FEM results. A
visual comparison of only the contour plot is not a good
validation. To provide a more reliable validation, the eddy current
den- sity is calculated in a single point and compared with the
analytical result in that same point. These results are shown in
figures below. In the middle of the SC, the analyti- cal eddy
current density has an almost identical evolution as the one
calculated using FEM. At the boundaries, the two models have the
same waveform but the amplitude is different. This is because the
FE model has difficulties at calculations at boundaries. Figure 4.4
shows that if we calculate the eddy current density a posteriori
using the electrical field (E) of the FE model, the analytical
calculation is indeed confirmed. The same is noticed for the eddy
current density in de PM. Note that the eddy current den- sity
calculated with the FEM is zero because one of the assumptions made
is that the conductivity of the PM is zero, so the validation
occurs with the electrical field of the FE model.
0 0.05 0.1 0.15 0.2 0.25 0.3 −8
−6
−4
−2
0
2
4
6
22
−1
−0.5
0
0.5
1
1.5
0 0.05 0.1 0.15 0.2 0.25 0.3 −5
−4
−3
−2
−1
0
1
2
3
4
23
−3
−2
−1
0
1
2
3
4
−3
−2
−1
0
1
2
3
24
Chapter 5
Eddy-current losses
5.1 Introduction
To compute the eddy-current losses in the SC and the PM, two
different methods are used. The losses in the PM can be calculated
from Joule’s formula. Joule’s formula is not effective at computing
the losses in the SC because the r-dependent part of J (3)(r, φ, t)
contains Bessel functions, those functions can not be integrated
analytically. To cope with this problem, the Poynting vector (S)
can be used.
5.2 Shielding cylinder
A generalized form of Poynting’s theorem is derived in the
appendix. This section uses the obtained results to calculate the
power losses in the SC. The Poynting vector (S) is written
as:
S = E×H (5.1)
The Poynting vector represents the rate per unit area at which
energy crosses a surface. The integration of S over a closed
surface determines the energy flowing through that surface. In this
case, this implies that the total power P transmitted from the
stator to the rotor can be calculated by integrating the Poynting
vector over the surface of a cylinder that includes the
rotor.
25
Pn,se −j(n+s)ωt (5.4)
where n is the time-harmonic order related to E(4) z and s is the
time-harmonic order
related to H(4) φ . Pn,s is calculated as:
Pn,s = ∞∑
(5.5)
To find the losses in SC, the total power can now be divided in the
mechanical power (Pm) and the eddy-current losses in the SC (P
(3)
ec ). As the mechanical power can easily be derived from the
torque, the eddy-current losses in the SC can be calculated as the
difference of P and Pm:
P (3) ec (t) = P (t)− Pm(t) (5.6)
where
Pm(t) = ωT (t) (5.7)
The losses in the SC are validated using the same FE model as in
Section 3.3.3. A devi- ation of 2.58% is obtained, this deviation
is calculated as:
d = PdissSC,FEM − PdissSC,ANA PdissSC,FEM
5.3 Permanent magnets
The Poynting vector is calculated directly from the magnetic and
electric fields. As the eddy-currents in the magnets are not
directly included in the calculation of the mag- nets, their losses
can’t be accounted for by the integration of S. So the losses in
the magnets will be computed using Joule’s formula:
P (2) ec (t) =
dv (5.9)
This equation is used to calculate the eddy-current losses in the
PM, for a 2D model it can then be rewritten as:
P (2) ec (t) =
J (2)(r, φ, t)J (2)(r, φ, t) σ(2) ds (5.10)
The losses in the PM are validated in the same way as in the SC, A
deviation of 4.844% is obtained
27
Chapter 6
Parameter study
6.1 Introduction
Now that the analytical calculations have been validated, a
theoretical study of the calculated properties can be conducted.
One aspect of high-speed PMSMs that still requires a lot study is
the effect of the Shielding Cylinder (SC) on the machine’s per-
formance. This shielding cylinder is a conductive sleeve that is
wrapped around the magnets. The goal of the SC can either be to
keep the magnets in place at very high rotational speeds and/or to
protect the magnets from overheating. The latter is due the fact
that the resistance of the SC is much lower than that of the
magnets. Eddy- currents, which will be induced in the SC instead of
in the magnets, will therefore cause less losses. In the first
section, the goal is to study the effect of the shielding cylinder
on the electromagnetic torque production. Secondly, the effect of
the SC on the dissi- pated power is discussed. To do so two
properties of the shielding cylinder are studied, its thickness
(tSC) and its conductivity (σSC).
29
6.2 Torque
(6.1)
6.2.1 Parameters
Table (6.1) shows the parameters of the studied machine. The two
studied parameters thickness and conductivity although listed here
are subject to change when studied. The machine geometry is
depicted in Figure 6.1, the studied machine is a three-phase
permanent-magnet synchronous machine with surface-mounted magnets
and a shielding cylinder.
30
Table 6.1: Parameters of the studied machine
Parameter Symbol Value Number of slots Ns 12 Number of pole pairs p
2 Number of phases m 3 Number of layers / 1 Residual flux density
of the magnets Bm 1.2 T Angular span of a magnet φm 80.00 Current
density Inom 5.106 A
m2
Number of windings per slot N 6 Nominal frequency fnom 1000 Hz
Nominal speed nnom 30,000 rpm Outer rotor iron radius r1 22.50 mm
Outer PM radius r2 25.00 mm Outer sleeve radius r3 26.50 mm Outer
airgap radius r4 28.50 mm Outer winding radius r5 40.40 mm Outer
stator iron radius r6 50.00 mm Stack length ls 200 mm Permeability
of the magnets µ2 1.0µ0
H M
Figure 6.1: Machine topology parameter study torque
31
6.2.2 Harmonic content
In order to gain a better understanding of the evolution of the
studied properties, an overview of the harmonic content of the
machine’s magnetic field is given. It can be observed that the
dominant synchronous fields components are (n=2,k=2) and (n=6,k=6).
Beside these synchronous components, a couple of asynchronous field
com- ponents are mildly outspoken, these are (n=2,k=14),
(n=2,k=-10),(n=2,k=26) and (n=2,k=-22). As discussed in Section
3.3.2, the produced torque can be explained by regarding the
harmonic content of the magnetic field.
−50 −42
−34 −26
−18 −10
6.2.3 Conductivity
In this section, a constant shielding cylinder thickness of 1.5 mm
is maintained. The parameter that is varied is the shielding
cylinder’s conductivity. The goal of this section is to identify
trends rather than to model a real machine, so the studied range
contains
32
Figure 6.3: Harmonic content of the studied machine Bφ
extreme conductivities, which are not realistic. The range in which
σSC is regarded is [0,108].
Figure 6.4 shows the torque and its components as a function of σSC
. As expected, TPM shows little dependency of σSC . This is because
TPM is mainly caused by the interac- tion of synchronous field
components and the expectation is that these components are not
affected by the shielding cylinder’s conductivity. Although when
plotted sepa- rately, it can be seen that there is a variation on
TPM . Figure 6.6 shows an initial de- crease followed by a small
increase and ultimately a continuous decrease. The reason for these
small variations is that a part of TPM is produced by the
interaction between asynchronous field components. At the boundary
between the air gap and the slotting region, one space harmonic of
the air gap (k) has an effect on all the space harmonics in the
slotting region (m) with the same time harmonic (n) as shown in
figure 6.5. Be- cause of this interaction, each asynchronous field
component has an effect on the syn- chronous field
components.
33
x 10 7
T or
qu e
(N m
T SC
Figure 6.4: Average torque as a function of the conductivity of the
SC
Figure 6.5: Illustration of the interaction between synchronous-
and asynchronous field components at the slot opening
A variation of the asynchronous field components can thus influence
the synchronous field components. The evolution of TPM in Figure
6.6 is mostly due to synchronous field components (n=2,k=2) and
(n=6,k=6). The synchronous field component (n=2,k=2) produces a
positive torque for all considered conductivities. The other
significant syn-
34
CHAPTER 6. PARAMETER STUDY
chronous field component (n=6,k=6) produces a negative torque for
all the considered conductivities. These components are visualized
in Figure 6.2. In each region the evo- lution is caused by the
synchronous field components which varies the most. The ini- tial
decrease is caused by field component (n=2,k=2) which variation is
dominant in this region. Component (n=2,k=2) has a positive effect
on the torque so a decrease of (n=2,k=2) implies a decrease of TPM
. The small increase that is observed in region B is caused by a
decrease of field component (n=6,k=6). That decrease implies less
neg- ative torque produced by (n=6,k=6) and thus an increase of the
produced torque. In region C, the decrease of field component
(n=2,k=2) becomes greater than the decrease in (n=6,k=6) again,
thus a decrease in TPM is observed.
0 2 4 6 8 10
x 10 7
T or
qu e
(N m
(A) (B) (C)
Figure 6.6: Torque in the PM as a function of the conductivity of
the SC
The evolution of TSC with increasing σSC is plotted in Figure 6.4.
TSC is negative which implies that, in the studied machine, the
asynchronous components with a slower ro- tational speed than the
synchronous are dominant. In what follows the absolute value of
TSC,net will be discussed. The evolution of TSC,net is very similar
to the speed-torque characteristic of an induction machine.
Initially an almost linear increase is observed. However after
reaching its maximum, TSC,net becomes inversely proportional to
√σSC . As mentioned in the beginning of this chapter, the shielding
cylinder can be regarded as the squirrel cage of an induction
machine. Thus the evolution of TSC can be explained in a similar
way as the speed-torque characteristic. This means that for a given
com- ponent of the magnetic field (n, k), TSC depends on the
resistance of the SC (RSC), the stator resistance (Rs), the rotor
leakage inductance (Llr), the stator leakage inductance (Lls), the
magnetization inductance (Lm), the slip (sn,k) and the pulsation
(ωn,k) of the harmonic combination (n, k):
TSC,net,n,k ∼ RSC(
CHAPTER 6. PARAMETER STUDY
Following Pouillet’s law for low σSC , RSC is inversely
proportional to σSC :
RSC = ls σSCSSC
(6.3)
with ls the stack length and SSC the surface of the SC. At high
conductivities this is not longer true because the skin effect is
no longer negligible. The threshold value is the conductivity at
which the skin dept (δ) equals the thickness of the SC. If the skin
dept is smaller than the thickness of the shielding cylinder, the
conductive surface of the SC is no longer constant. Equation (6.4)
shows that the conductive Surface is in- versely proportional to
√σSC and according to Pouillet’s law, RSC will then be inversely
proportional to √σSC .
δn,k = √
(6.4)
At low conductivities of the shielding cylinder, RSC will be high
and will dominate the denominator of (6.2).
RSC
TSC,net,n,k ∼ RSC((
∼ { σSC if δ > tSC√ σSC if δ < tSC
In this study, the area in which TSC,net increases is situated
before the point at which δ = tSC . This confirms the linear
behavior at low σSC . At high σSC , RSC will be low and thus the
RSC term is negligible in the denominator of (6.2).
RSC
TSC,net,n,k ∼ RSC
if δ < tSC
This shows that until δ > tSC , TSC,net will decrease
proportionally to 1 σSC
and after the threshold value TSC,net becomes proportional to
1√
σSC . Note that the conductivity for
which δ = tSC is located in the transition area between an
increasing and a decreasing TSC,net. Which implies that only the
decrease inversely proportional to √σSC is visible. A final remark
is that the maximum of TSC,net, as (6.2) shows, depends on the
ratio be- tween the resistances and the inductances and thus on the
entire machine geometry.
In Figure 6.7, the percentage torque ripple as a function of σSC is
shown. It can be seen that there is a shift of the ripple from TPM
to TSC when σSC is increased. This is ex- plained by a stronger
shielding effect. The total torque ripple is almost constant. The
small variation of T% is explained by a similar variation of T
(σSC). The evolution of TSC%(σSC) is as expected. Increasing σSC
implies a better shielding effect. A better shielding effect
translates in an increasing torque ripple. When δ becomes smaller
than tSC , the SC already blocks most of the asynchronous content
and the increase of TSC% stagnates. Because the ripple is defined
as a percentage of T in (6.1), TSC% slightly decreases for higher
σSC because the total torque T increases for higher σSC .
37
x 10 7
T or
qu e
rip pl
SC
Figure 6.7: Torque ripple in the PM as a function of the
conductivity of the SC
6.2.4 Thickness
The second parameter to be studied is the thickness of the
shielding cylinder and it’s ef- fect on the different torque
components. This is done by varying the outer radius of the SC (r4)
and keeping the outer radius of the PM (r3) fixed. The width of the
elements beyond r4 are kept constant to ensure a fair comparison. A
first expectation of the ef- fect of an increasing tSC would
suggest a decrease in the net torque. As stated in 3.3.2, the net
torque is mainly associated with the torque component TPM . TPM is
produced by the interaction between the PM magnetic field and the
stator current. The amount of non-magnetic material between the
stator and the rotor is referred to as the effective air gap
(tAG,eff = r4−r2). Increasing the effective air gap tAG,eff reduces
the interaction which produces TPM and thereby decreasing the net
torque. On the other hand, TSC is determined by two effects in
regard to an increasing tSC . First, regarding the shielding
cylinder as an electrical conductor, and applying pouil- let’s law
(6.3), which dictates that the SC resistance’s decreases as tSC
increases. This decrease in resistance results in a higher eddy
current density and thus a rise of TSC . Secondly, TSC is
influenced by the effective air gap width (tAG). By increasing tAG,
a smaller amount of eddy currents are induced in the shielding
cylinder, resulting in a de- crease of TSC .
Figure 6.8 shows the evolution of TPM , TSC and the total torque T
in regard to a vary- ing tSC . It shows an initial increase in TSC
. Which means that the decreasing resistance
38
CHAPTER 6. PARAMETER STUDY
of the SC dominates the effect of a larger effective air gap. At a
certain tSC , the in- crease of the effective air gap width
dominates the evolution of TSC , which explains the continuous
decrease at the end. The amplitude and tSC at which the maximum
occurs depends on numerous factors, such as electrical frequency,
machine geometry and num- bers of pole pairs.
0 1 2 3 4 5
x 10 −3
T SC
Figure 6.8: Average torque as a function of the thickness of the
SC
Figure 6.9 shows the torque ripple with respect to Tnet as
described in (6.1). As tSC widen, the torque ripple on TPM rapidly
decreases. An increasing tSC also increases tAG, therefore less
asynchronous field can interact with the permanent magnets. As the
torque ripple is mainly caused by the slotting effect, this
explains why TPM% de- creases. Past a certain tAG, no more
asynchronous field components reach the magnet region and therefore
TPM% decreases to zero. The evolution of TSC% is similarly in-
fluenced by two same effects as TSC , the increasing surface area
of the SC and the in- creasing tAG. As tSC increases more eddy
currents are produced in the SC. When tSC becomes larger than the
penetration depth of the asynchronous fields, no more addi- tional
eddy currents are induced. As tSC further increases, this results
in a decrease of TSC .
39
x 10 −3
SC
Figure 6.9: Torque ripple as a function of the thickness of the
SC
6.3 Eddy current losses
In this section a study of the shielding cylinder’s effect on the
eddy current losses is conveyed. Eddy currents are induced due to
asynchronous field components in the air gap’s magnetic field,
which are a consequence of the slotting effect (see 6.5) and har-
monics in the stator currents. These eddy currents add losses to
the machine in the form of Joule losses (i2R), these losses
generate a temperature rise which can possibly cause
demagnetization of the permanent magnets. In order to reduce the
total losses and risk of demagnetization, a shielding cylinder is
added to the machine.
6.3.1 Parameters
To study the effect of the SC on the eddy current losses, two
aspects of the SC, the thickness and conductivity, are regarded.
Table 6.2 shows the parameters of the studied machine. A machine
with concentrated windings is studied because these machines are
more affected by eddy current losses in the PM at high speeds [10].
Figure 6.10 shows the studied machine geometry.
40
Table 6.2: Parameters of the studied machine
Parameter Symbol Value Number of slots Ns 6 Number of pole pairs p
1 Number of phases m 3 Number of layers / 2 Residual flux density
of the magnets Bm 1.2 T Angular span of a magnet φm 180 Current
density Inom 5.106 A
m2
Number of windings per slot N 2 Nominal frequency fnom 2500 Hz
Nominal speed nnom 150,000 rpm Outer rotor iron radius r1 7.00 mm
Outer PM radius r2 10.00 mm Outer sleeve radius r3 12.00 mm Outer
airgap radius r4 15.00 mm Outer tooth tip radius r5 17.00 mm Outer
winding radius r5 27.00 mm Outer stator iron radius r6 32.00 mm
Stack length ls 36 mm Permeability of the magnets µ2 1.0µ0
H M
Figure 6.10: machine topology parameter study Eddy current
losses
41
6.3.2 harmonic content
As with the previous machine an overview of the harmonic content of
the magnetic field is given to provide better understanding. It can
be observed that the primarily present synchronous fields are
(n=1,k=1), (n=3,k=3) a,d (n=5,k=5). Although less outspo- ken the
they are asynchronous fields components, these are (n=1,k=7)
(n=1,k=-5) (n=1,k=13) (n=1,k=19) (n=1,k=-11) (n=1,k=-17) and
(n=1,k=-23).
Figure 6.11: harmonic content of the studied machine Br
42
6.3.3 Conductivity
The first parameter to be studied is the conductivity of the SC.
For this study a con- stant shielding cylinder thickness of 1.5 mm
is considered. The range of the regarded σSC is [0− 13 · 106], as
in the previous study (6.2.3), the goal is to analyze trends rather
than model a real machine. Figure 6.13 shows the evolution of the
eddy current losses as a function of σSC . As the conductivity
increases, the shielding effect increases so less asynchronous
field compo- nents reach the magnet region. This implies that less
eddy currents are induced. This explains the parabolic decrease of
Pdiss,PM as the joule losses are proportional to I2. The evolution
of Pdiss,SC is plotted in Figure 6.14. As the conductivity
increases two aspects can be regarded, the increase in the amount
of eddy currents induced, and the decrease of the resistivity.
These changes are shown in Figure 6.15, 6.16 and 6.17. First, the
conductivity is zero, thus there are no induced eddy currents. As
the conductivity increases, there is an increase in the amount of
eddy currents induced. The percentage change of these induced eddy
currents is bigger than the percentage change of the resis- tivity
so there is an increase of the dissipated power. When the
conductivity increases beyond 2.5 · 105, there is a decrease of the
percentage change of the induced eddy cur- rents and the percentage
change of the resistivity becomes dominant. This results in a
43
CHAPTER 6. PARAMETER STUDY
decrease of the dissipated power. RSC is inversely proportional to
σSC as shown in Fig- ure 6.17. As σSC further increases, the
percentage change of the resistivity de