Professional Documents
Culture Documents
Design and Comparison of PMSM Vs PMaRSM
Design and Comparison of PMSM Vs PMaRSM
AVANCERAD NIVÅ, 30 HP
STOCKHOLM, SVERIGE 2018
VIKTOR BRIGGNER
KTH
SKOLAN FÖR ELEKTROTEKNIK OCH DATAVETENSKAP
TRITA TRITA-EECS-EX-2018:496
www.kth.se
Design and comparison of PMaSynRM versus PMSM for
pumping applications
VIKTOR BRIGGNER
TRITA-EECS-EX-2018:496
Design and comparison of PMaSynRM versus PMSM for pumping applications
VIKTOR BRIGGNER
ii
Sammanfattning
Detta examensarbete avsåg att designa en rotor till en permanentmagnetsassisterad
synkron reluktansmaskin (PMaSynRM) för pumpapplikationer, vilken skulle implement-
eras i en befintlig asynkronmaskin (IM) stator. Maskinen jämfördes ekonomiskt och pre-
standamässigt med en liknande synkronmaskin med permanentmagneter (PMSM) med
jämförbar vridmomentsproduktion.
Uppsatsen avhandlar teorin bakom synkrona reluktansmaskiner och konceptet kring
permanentmagnetassistans. Rotorn designades genom användandet av befintliga design-
metoder och simulering genom finit elementanalys (FEA). En avmagnetiseringsstudie
utfördes på de adderade magneterna för att undersöka rimligheten kring designen
Den slutgiltiga designen av PMaSynRMen jämfördes därefter mot den jämlika
PMSMen i termer om prestanda och kostnad. De undersökta prestandaparameterarna var
vridmoment, vridmomentsrippel, verkningsgrad och effektfaktor. Eftersom vridmoments-
densiteten i en PMaSynRM är lägre än hos en PMSM så visade sig PMSMen ha en 40%
kortare lamineringskropp än PMaSynRMen vid jämnlik vridmomentsproduktion.
Den ekonomiska utvärderingen resulterade i att vid användandet av ferritmagneter
i PMaSynRMen så blev den något billigare än PMSMen, upp till 20%. På grund av fluk-
tuerande priser hos NdFeB magneter, så finns det brytpunkter där PMaSynRMen faktiskt
blir dyrare än PMSMen eller då kostnadsreduktionen för PMaSynRMen kan bedömas att
vara för låg med tanke på den ökade längden och vridmomentsrippel. Däremot visades
det att PMaSynRMen är väldigt okänslig för prisvariationer och därför visades vara ett
kostnadsmässigt tryggare val än PMSMen.
iii
Acknowledgements
This master thesis has been carried out at the department of Research and Development
for electrical motors at Xylem Water Solutions in Stockholm, Sweden.
I would like to thank Xylem Water Solutions for giving me the opportunity to do my
master thesis for them and for the great experience that it has entailed. I would especially
like to thank Tanja Hedberg and Øystein Krogen for their supervision and help throughout
the duration of the project. Furthermore would I like to thank my co-workers at Xylem
Water Solutions for making my stay there even more enjoyable with their company.
I would also like to express a special thanks to Associate Professor Oskar Wallmark for
sparking my interest in electrical machines and for inspiring me to pursue this field of en-
gineering. Additionally I would like to thank him for acting as my examiner for this thesis.
I also want to give thanks to all of my friends here in Stockholm who has made my
years at KTH unforgettable to say the least. Thank you for all the memories and for your
friendship. Even if we eventually find ourselves in different parts of the world, I know
that we will always stay in touch.
Finally, I would like to give my deepest gratitude to my parents and my sister who always
have supported me and helped me whenever I needed it. I would also like to especially
thank my girlfriend, Saga Kubulenso, for her never-ending patience with me when my
studies has gotten the best of me and for always being there for me no matter what. Thank
you so much.
Viktor Briggner
Stockholm, Sweden
August 2018
iv
Contents
Abstract ii
Sammanfattning iii
Acknowledgements iv
Contents v
Acronyms 1
Nomenclature 3
1 Introduction 6
1.1 Background and objectives . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.2 Thesis outline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
v
Contents
3 Method of analysis 37
3.1 Modeling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
3.1.1 Performance parameters . . . . . . . . . . . . . . . . . . . . . . 37
3.2 Initial dimensions and target PMSM . . . . . . . . . . . . . . . . . . . . 38
3.2.1 Stator selection for PMaSynRM . . . . . . . . . . . . . . . . . . 39
3.2.2 Target PMSM . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.3 Design procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.3.1 Parametric study . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.3.2 SynRM base-line design . . . . . . . . . . . . . . . . . . . . . . 42
3.4 SynRM design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4.1 Insulation ratios . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
3.4.2 Air-gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.4.3 Rotor barrier end angles . . . . . . . . . . . . . . . . . . . . . . 44
3.4.4 Choice of barriers . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.4.5 Radial ribs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.5 PMaSynRM design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.5.1 Magnet sizes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.5.2 Steel grade . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.5.3 Demagnetization . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.6 Performance comparison and economic analysis . . . . . . . . . . . . . . 47
3.6.1 Comparing the machines . . . . . . . . . . . . . . . . . . . . . . 47
4 Results 50
4.1 Rotor design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.1 q-axis insulation ratio . . . . . . . . . . . . . . . . . . . . . . . . 50
4.1.2 d-axis insulation ratio . . . . . . . . . . . . . . . . . . . . . . . . 51
4.1.3 Air-gap . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.1.4 Rotor barrier end angles . . . . . . . . . . . . . . . . . . . . . . 54
4.1.5 Radial ribs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.1.6 Final SynRM rotor geometry . . . . . . . . . . . . . . . . . . . . 57
4.2 PMaSynRM design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2.1 Magnet addition . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.2.2 Without magnet in outermost barrier . . . . . . . . . . . . . . . . 59
vi
Contents
4.2.3 NdFeB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.2.4 Steel types . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.3 Demagnetization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4 PMSM versus PMaSynRM . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.4.1 Performance comparison . . . . . . . . . . . . . . . . . . . . . . 66
4.4.2 Cost comparison . . . . . . . . . . . . . . . . . . . . . . . . . . 66
A General calculations 74
A.1 Derivation of expression for IPF . . . . . . . . . . . . . . . . . . . . . . 74
A.2 Center of gravity of rotor segments . . . . . . . . . . . . . . . . . . . . . 75
B Data sheets 77
B.1 Steel . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
B.1.1 M400-50A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
B.1.2 M600-50A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
B.1.3 M800-50A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
B.2 Magnets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
B.2.1 Ferrite . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
B.2.2 NdFeB - N33EH . . . . . . . . . . . . . . . . . . . . . . . . . . 83
C Results 84
C.1 Demagnetization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84
References 87
vii
Acronyms
IM induction machine 6
IPF internal power factor 15
IPMSM interior PM synchronous machine 11
NdFeB neodymium-iron-boron 6
PF power factor 15
PM permanent magnet 10
PMaSynRM permanent magnet assisted synchronous machine 7
PMSM permanent magnet synchronous machine 6
SP salient pole 11
SynRM synchronous reluctance machine 7, 9
1
Acronyms
2
Nomenclature
3
Nomenclature
4
Nomenclature
5
Chapter 1
Introduction
6
1.1. Background and objectives
[6]. Furthermore, the extraction process of rare-earth magnets entails both a hazardous
environment for workers and for the people living nearby the extraction and refinement
plants, as well as the process used being very environmentally harmful [7].
synchronous reluctance machines (SynRMs), which utilize the anisotropy of the ro-
tor to produce torque known as reluctance torque, has been shown to perform better than
induction machines in terms of efficiency [8] while however falling short of equivalent
PMSMs by a large margin both in terms of efficiency and power factor, but also torque
density [9]. However, lately permanent magnet assisted synchronous machines (PMaSyn-
RMs) has been a source of interest in order to find a feasible competitor to the PMSMs.
The permanent magnets utilized in a PMaSynRM are either of far lesser quantities of
rare-earth magnets or alternatively of weaker, more abundant, and cheaper magnets such
as ferrite magnets [9]. High-efficiency PMSMs also generally perform better than PMa-
SynRM in terms of efficiency and torque density [5, 9], however this difference might be
small enough that the economical benefit can outweigh the reduction in performance. In
Fig. 1.1 a PMaSynRM and a surface-mounted PMSM is shown.
In the present thesis, a PMaSynRM will be designed and analyzed based on torque
density, efficiency and power factor and thereafter compared to an equivalent high-efficiency
surface-mounted PMSM. The machines are simulated and analyzed by means of finite el-
ement analysis (FEA). The work conducted in this thesis is in part based on the work
conducted by Adrian Ortega Dulanto as a master thesis [10]. However, the investigated
dimensions of the machines has been increased as to see if the PMaSynRM might dis-
play a relative increase in performance at greater dimensions as well as investigate the
scalability of the design developed in [10].
Furthermore, the choice to compare the PMaSynRM to a PMSM can be argued to be
a better comparison rather than to a line start PMSM (LSPM) or IM as both PMaSynRM
and PMSM requires a variable frequency drive (VFD) to operate as opposed to the other
7
1.2. Thesis outline
8
Chapter 2
As was briefly stated in Chapter 1 the synchronous reluctance machine (SynRM) and
PMaSynRM relied on the anisotropy of the rotor in order to produce torque. In this chapter
the theory of the SynRM will be discussed and how utilizing magnets to further improve
the operation of SynRM influence the operation, thus producing a PMaSynRM.
9
2.2. Synchronous reluctance machine
F
d F d
B
q q
F
F
a) b) c)
Object c) is added in Fig. 2.1 as to illustrate how the concept is transferred to rotat-
ing machines. Anisotropy is achieved by introducing air gaps, or flux barriers, in a rotor
structure and a torque is produced. However, note that in a SynRM the magnetic field will
be directed radially and rotating in order to produce a continuous torque, but the same
concept applies.
From this qualitative description of the reluctance torque it can be derived that the
torque production in a SynRM (and PMaSynRM) is dependent on and that there has to
be an optimal angle if it is sought to maximize the torque. The main flux in these machines
is induced by the current and therefore by controlling the current, the torque is controlled.
How the flux, current and reluctance is related is expanded on in the coming sections.
10
2.2. Synchronous reluctance machine
The main advantage of the SynRM when compared to a PMSM is generally the
lower price range as it doesn’t utilize expensive rare-earth magnets. However, there are
more advantages over the PMSM of the SynRM and PMaSynRM as outlined in [9] and
[12]. To name a few we have
• The SynRM is not as vulnerable to short-circuit conditions as the lack of magnets
means that no current is induced.
• The constant power speed range (CSPR) is very good for SynRM and especially
PMaSynRM
• The rotor saliency provides with easy rotor position detection at stand-still
The design of the SynRM rotor as it looks today is still conceptually based on the
work done by Kostko in 1923 where the rotor is divided into different segments with
flux barriers in order to achieve a high saliency [9, 13] as seen in Fig. 2.2b and c. Salient
pole machines can be constructed in a few different ways. First, there is the conventional
salient pole (SP) rotor, the axially laminated anisotropy (ALA) rotor and the transversally
laminated anisotropy (TLA) rotor [14] and these types can be seen in Fig. 2.2.
However, the SP design configuration has been shown to be sub-optimal for SynRM
drives and is more suitable for wound rotor synchronous machines. The ALA is more
theoretically appealing and is believed to be able to provide a higher saliency ratio than the
TLA configuration. However, the TLA configuration is much easier to mass produce as it
utilizes the same punching and assembly procedure as traditional electrical machines [15]
and therefore will be the focus of this thesis.
The geometry of a TLA SynRM is similar to that of object c) in Fig. 2.1. As was ex-
plained in Section 2.1 the SynRM produces its torque by differences in reluctance around
the rotor. Generally several flux barriers are introduced and it has been shown that the pole
number should be kept as low as possible where four poles are held as the most suitable
pole number. In [11] a thorough investigation regarding pole numbers are presented. Fig.
2.3 displays a common four-pole SynRM rotor design with 3 flux barriers.
11
2.2. Synchronous reluctance machine
Fig. 2.2: Different rotor designs for rotor saliency. a) Conventional salient pole. b) Axially lami-
nated anisotropy. c) Transversally laminated anisotropy. From [14]
along the flux from the PM, this distinction will prove important when discussing perma-
nent magnet assistance since the consequence will be that the axes are reversed in terms
of permanent magnet flux.
d
e= + j!e (2.1)
dt
Neglecting iron losses, the dynamics of the synchronous machine in dq-frame can
be described by
12
2.2. Synchronous reluctance machine
Flux barriers
Iron segments
d
Fig. 2.3: Cross-section of SynRM with defined dq-reference frame.
d d
v d = R s id + !e q (2.2a)
dt
d q
v q = R s iq + + !e d (2.2b)
dt
where vd and vq are the stator voltages, id and iq are the stator currents, d and q are the
machine flux linkages [9].
Fig. 2.4 can be utilized to draw a phasor diagram in the dq-plane as seen in Fig. 2.5.
Here the is the load angle, is the torque angle, is the current angle from the d-axis,
' and 'i are the power factor angle and internal power factor angle respectively.
The torque, ⌧em , can be described as
3
⌧em = p( d iq q id ) (2.3)
4
where p is the number of poles [9].
The d- and q-axis flux linkages are both very dependent on the operating point and
13
2.2. Synchronous reluctance machine
j!e
Rs is i
+
vs d
e Rc L
dt
ic
q
Rs id
vs !e Lq iq
e j!e Ld id
is ic
iq
i
Rs iq
'
'i jLq iq
Ld id
id d
Fig. 2.5: Phasor diagram for SynRM.
experience cross-coupling from currents in the adjacent axis [9, 11], i.e.
(
d = d (id , iq )
(2.4)
q = q (id , iq )
This cross-coupling occurs since the q-axis current cause a flux component in the d-axis
and vise verse. This not only contributes to the total flux in the respective axis but it also
affect the saturation level of the iron in the respective axes. Hence, when rewriting the
flux linkages as current times inductances it is very important to note that the inductances
(Ld , Lq ) are indeed not constant [9, 11]
(
d = Ld (id , iq )id
(2.5)
q = Lq (id , iq )iq
14
2.3. Saliency and performance
Note that this definition of the inductances is a simplification and for a more thor-
ough discussion see Chapter 4 in [9]. Note also that the inductances includes not only
magnetizing inductances but also the leakage inductances, which are not significantly in-
fluenced by the aforementioned cross-coupling and saturation [16].
Utilizing these relationships we find that we can write the internal power factor (IPF) as
⇣⇡ ⌘ ⇠ 1
IPF = cos 'i = cos + =r (2.9)
2 1 1
⇠2 +
sin2 cos2
for the derivation of this expression refer to Appendix A.1. And thus, it becomes obvious
that the saliency ratio influences the IPF heavily, as seen in Fig. 2.6 where the it is plotted
for different values of ⇠ as a function of the current vector angle. It is important to note
that IPF is not the same thing as power factor (PF) but they are, however, related and
a high value of IPF leads to a high value of PF since the difference between these two
are only governed by Rs and Rc which can be verified by looking at Fig. 2.5 and Fig.
2.4. Therefore, IPF is discussed here as it is quite straightforward to derive an analytical
expression from current angle and saliency.
From equation (2.9) it is obvious that for any given value of ⇠ there exist a value
of the current vector angle which allows for the optimal IPF. It can be shown that this
p ⇠ 1
occurs when tan = ⇠ and then the IPF is equal to . This operating point is often
⇠+1
called maximum torque per kVA (MTPkVA) [16] and thus correspond to the operating
point when the least amount of reactive power is required by the supply.
With the definition of flux in equation (2.5) and dropping the parentheses for sim-
plicity we find that equation (2.3) can be rewritten as
15
2.3. Saliency and performance
0.9 ⇠=2
⇠=5
0.8 ⇠=10
⇠=15
0.7
⇠=20
0.6 ⇠=100
IPF
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60 70 80 90
[deg]
3
⌧em = p(Ld Lq )id iq (2.10)
4
Thus, the torque is proportional to the difference in inductance between the two
axes. It is interesting to note that while the IPF is governed by the saliency ratio, the
torque production is dependent on the saliency difference. Even though these parameters
are related, it can prove difficult to maximize both parameters simultaneously when de-
veloping a rotor design [10]. This is due to the non-linear dependency of d- and q-axis
inductances on the rotor geometry [17].
Another conflict occurs when operating a SynRM and it becomes apparent when
rewriting equation (2.10) utilizing the phasor quantities presented in Fig. 2.5 in steady
state, i.e.
3
⌧em = p(Ld Lq )Is2 sin 2 (2.11)
4
where Is is the stator current magnitude in steady state. Here we see that for given (con-
stant) inductances and current magnitude, the maximum torque is achieved for a current
vector angle of 45 degrees. This operating point is referred to as maximum torque per
ampere (MTPA). Again, looking at Fig. 2.6 we see that this current vector angle does not
coincide with the angle which maximizes the IPF for moderately high values of ⇠.
16
2.4. Iron saturation
It should also be noted that saturation affect both the inductance difference and
saliency ratio such that in reality ⇠ and inductance difference decrease with increasing
current, which affect both the torque production and power factor. The effect of iron sat-
uration will be expanded upon below.
17
2.5. Permanent magnet assistance
100
Ld
90 Lq
80
70
Inductance [mH]
60
50
40
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
Current magnitude [A]
Fig. 2.7: Ld and Lq as function of current modulus for machine with 26 A rated current (RMS).
add a flux linkage component in the q-axis and thus equation (2.5) becomes
(
d = Ld (id , iq )id
(2.12)
q = Lq (id , iq )iq PM
where P M is the permanent magnet flux linkage addition. As can be seen in Fig. 2.9 the
phasor diagram is altered due to this addition. The effect of the PM flux is highlighted
in red. The voltage phasor is rotated towards the current vector effectively increasing the
power factor. Additionally, since the PM-flux is largely directed in the q-axis direction
this flux also help saturate the iron in the q-axis which reduces the q-axis inductance and
thus increases the saliency ratio [9].
To derive an analytical expression for the IPF as a function of the saliency ratio of a
PMaSynRM is not as straightforward as it were for the SynRM due to it being dependent
on the q-axis PM flux linkage aswell. However, when comparing the two phasor diagrams
in Fig. 2.5 and Fig. 2.9 we notice that the difference is the aforementioned rotation of
the voltage vector. This rotation occur because the added PM flux counteracts the q-axis
stator flux linkage and thus shifts the flux vector away from the current vector. For both
machines in the steady state, the emf (e) will be perpendicular to the flux vector ( ) and
therefore the emf will be rotated towards the current vector which increases the IPF and
hence PF.
18
2.5. Permanent magnet assistance
2.5
1.5
1
Flux linkage [Wb]
0.5
0
d (id , iq = 0)
0.5 d (id , iq = 50)
d (id , iq = 100)
1 q (iq , id = 0)
q (iq , id = 50)
1.5
q (iq , id = 100)
2
2.5
100 80 60 40 20 0 20 40 60 80 100
id or iq [A]
Fig. 2.8: d and q as function of different values of id and iq , for machine with rated current of
26 A (RMS).
The added PM flux also significantly alter the expression for the torque production
and equation (2.10) becomes
3
⌧em = p P M id + (Ld Lq )id iq (2.13)
4
i.e. we get two torque components, the PM induced torque (⌧P M ) and reluctance torque
(⌧rel ). Note that this equation is similar to that of a salient-pole PMSM, but the key dif-
ferences being that the torque contribution of the PM-flux is lower compared to the the
reluctance torque and that the reference frame is rotated.
Rewriting equation (2.13) in the same fashion as equation (2.11) we get
3
⌧em = p P M Is cos + (Ld Lq )Is2 sin 2 (2.14)
4
This implies that the PM torque and reluctance torque does not display coinciding
maxima with respect to . Hence, the optimal value of depends on the ratio between the
two but typically lies around 40 [9] when neglecting saturation.
In addition to the shifting of flux vectors and PM torque production, another feature
of permanent magnet assistance benefits the operation of the PMaSynRM. The PM flux
helps saturate the ribs within the rotor structure (which are further discussed in Section
19
2.5. Permanent magnet assistance
q
!e Lq iq R s id
j! PM vs
e
j!e Ld id
is
ic
iq
i
Rs iq
'
'i
jLq iq
id L d id
d
j PM
2.6.6). This hence helps reduce the flux between the iron segments in the rotor which
further reduces the q-axis inductance [18].
20
2.6. Geometry and performance of PMaSynRM
In fact, it is stated in [18] that the PM flux should be in the vicinity of 25-35% of the
nominal flux. This also implies that ferrite PMs might be sufficient in order to reach the
desired operating point.
• The barrier height of barrier k is described by Wk,q in the q-direction and Wk,d in
the d-direction.
• The height of the iron segment h between barriers is described in the q-direction as
Sh,q
• The angle that the barrier k makes at the periphery of the rotor with the d-axis is
defined as ✓b,k
• The angle between flux barrier arm and center of flux barrier is defined as ↵i
Often, the q-axis position of a rotor barrier along the q-axis is of interest. The dis-
tance to the n:th barrier is defined as
n
X n 1
X
D0,n = Sh,q + Wk,q (2.15)
h=1 k=1
Furthermore, in order to give an indication as to how much air versus iron there is in
the rotor in both the q- and d-direction the insulation ratios kw,q and kw,d are qualitatively
21
2.6. Geometry and performance of PMaSynRM
↵3
S4,q
↵2
W3,q
S3,q
↵1 ✓b,3
W3,d
W2,q
S2,q
W2,d ✓b,2
W1,q
S1,q
W1,d ✓b,1
Rsh
R1
defined as
Amount of air
kw,q = (2.16)
Amount of iron q axis
Amount of air
kw,d = (2.17)
Amount of iron d axis
which gives that a value below 1 of these ratios means that there is more iron than air in
the respective direction and conversely a value above 1 means that there is more air than
iron. Note that the path of calculation for the q-axis is easily defined along the axis. For
the d-axis it is a somewhat more complicated. The expressions for the insulation ratios
are given in Section 2.8.
22
2.6. Geometry and performance of PMaSynRM
23
2.6. Geometry and performance of PMaSynRM
presented in [23] as
nr = ns ± 4 (2.19)
where nr is the number of rotor barrier slots per pole and ns is the number of stator slots
per pole. Whether the equation should be treated with a plus or a minus is determined by
the feasibility of the structure albeit it is stated that +4 generates better results.
In [22] a thorough analysis of the behavior of different numbers of stator slots and
barriers is presented. There it was derived that different number of stator slots perform
at its best for different number of rotor barriers For instance was it shown that for a 48
slot machine, torque production was maximized for 4 or 6 number of barriers whereas
efficiency was maximized for 4 barriers and torque ripple was minimized for 6 barriers.
Whereas in [10] it was determined that for the 36 slot machine 3 barriers generated the
overall best performance in terms of torque production and power factor.
24
2.6. Geometry and performance of PMaSynRM
ripple won’t correspond to any major speed ripple. Hence, to optimize torque ripple is not
the focus of this paper but measures in terms of altering the placement of the rotor slots
will be made as long as it does not negatively impacts the torque production.
25
2.6. Geometry and performance of PMaSynRM
wr,3
wt,3
wr,2
wt,2
wr,1
wt,1
TLA, means that the rotor is punched and therefore require a continuous sheet of metal.
The width of this rib is in part determined by the tolerance of the punching machine, but
also by the expected tangential forces from torque ripple or load variations. However, the
calculation of the thickness of the tangential ribs are determined to be outside of the scope
of this project. In Fig. 2.12, these ribs are illustrated, and the parameters describing the
widths defined, wr,i is the widths of the radial ribs and wt,i is the width of the tangential
ribs.
Not all machines require radial ribs and it is rather a question of size of the rotor,
radial positioning of the flux barriers, and maximum allowable speed of the machine
which determines the need and widths of them. The width of the radial ribs can calculated
by diving the rotor into i segments and calculate the rotational force exerted on each
segment [26]. This is done as
Fc,i
wr,i = ⌫rib (2.20)
r Lstk
where Fc is the centrifugal force acting on the rotor, r is the tensile strength of the
material, Lstk is the total stack length of the rotor and ⌫rib is a safety factor usually in the
26
2.6. Geometry and performance of PMaSynRM
1 2 3
D0,3
D0,2
D0,1
where AF e is the area of the relevant rotor segment, RG is the center of gravity of the
rotor segment, !m is the mechanical angular frequency of the rotor and ⇢lam is the mass
density of the steel.
An overestimation of the radial ribs can be achieved by simplifying the geometry
as shown in Fig. 2.13 and use these sections to calculate the force that each rib will expe-
rience. Note that each area contains all steel within its boundaries, i.e. Area 1 includes 2
and 3 and so on. This is an overestimation since the air in the rotor structure is neglected
which means that the force in reality is lower provided that the center of gravity is not
increased dramatically. This also holds when magnets are added to the flux barriers since
the density of ferrite magnets is lower than the steel and NdFeB magnets are about com-
parable with the steel while they will likely not fill up the entirety of the flux barriers. For
the equations to derive the center of gravity refer to Appendix A.2.
Introduction of radial ribs in the barrier structure leads to an unwanted flux path
in the rotor which contributes to increase the q-axis inductance and therefore leads to a
torque reduction. In Fig. 2.14, the influence of the ribs on the q-axis flux is visible when
looking at the span -5 A to 5 A, the slope of the flux, and thus inductance, is significantly
higher in that span than elsewhere. This increased inductance is due to the flux path pro-
27
2.6. Geometry and performance of PMaSynRM
2.5
1.5
1
Flux linkage [Wb]
0.5
0.5
1
d (id , iq
= 0)
1.5
(i ,
q q d i = 0)
2
2.5
60 50 40 30 20 10 0 10 20 30 40 50 60
id or iq [A]
Fig. 2.14: Influence of radial ribs on flux, for machine with 26 A rated current.
vided by the radial ribs. However, it is also visible that for greater values of the current,
the ribs saturate and start to behave as air. In [15] it was shown that this reduction is in the
magnitude of a few percent of the nominal torque. In [27] an expression to estimate of the
magnitude of the torque reduction is presented where the torque reduction is proportional
to the number of poles in square times the width of the ribs assuming constant width of
the ribs. In [11] further analysis of the influence and design of radial ribs is conducted.
28
2.6. Geometry and performance of PMaSynRM
wma,3
wma,2
wmb,3
wma,1
wmb,2
wmb,1
Fig. 2.15: Magnet variables and magnet placement in rotor.
29
2.7. Permanent magnets
B
Br
Hc
H
for a machine in the range of 12 kW output power, the efficiency saw an increase of 9
percentage units between the lowest and highest loss steel grade. At the same time, the
output power increased by 8 percent when using the low loss grade steel compared to the
higher loss grade.
B = µ0 (H + M) (2.22)
30
2.7. Permanent magnets
B or Im
Br
Hc,i Hc,b
H
Fig. 2.17: Typical normal (dashed) and intrisic (line) curve for permanent magnet.
2.7.1 Demagnetization
Typically, permanent magnets do not lose its magnetization when the flux density is re-
duced to zero, i.e. when the coercivity is reached. Utilizing the above stated relationship
one can define the magnetic polarization, Im , as
I m = µ0 M = B µ0 H (2.23)
Both Im and B can be plotted in the same graph as is done in Fig. 2.17. The Im -H
plot is often referred to as the intrinsic curve and the B-H plot is called normal curve.
In these plots two coercivities appear, the intrinsic and normal coercivity Hc,i and Hc,b .
Analogous to the definition of Hc , Hc,i is the value at which the magnetization is forced
to zero and beyond this point the magnetization will start to shift polarity [31]. Note that
Hc,b = Hc as defined previously.
Demagnetization of the permanent magnet occurs when the magnetic field intensity
approaches the intrinsic coercivity. In fact, when the magnetic field passes the value close
to the knee of the intrinsic curve partial demagnetization start to occur. For most practical
situations in electrical machines, only the second quadrant of the hysteresis loops are of
interest. One can define the magnetic field knee value in the second quadrant as Hk and
when that value is exceeded and thereafter reduced to below that value again, the magne-
tization of the magnet will be reduced and therefore also the remanent flux density. It can
be shown that when Hk is exceeded, the new intrinsic curve follows the so called recoil
lines shown in Fig. 2.18. The slope of the recoil lines are similar to that of the slope of
31
2.8. Theoretical foundation of design approach
Br
Bk
Hc,i Hk
Fig. 2.18: Demagnetization curve for permanent magnet with recoil lines.
the original intrinsic curve when the magnetic field is zero [33]. This reduction of mag-
netization is what is referred to as demagnetization of the magnets and the consequences
of this is that the maximum energy product of the magnets is reduced which lowers the
magnetic torque and saturation flux of the magnets and is therefore undesirable.
As with all materials, magnetic materials are temperature sensitive where the tem-
perature of the magnet alter the magnetic characteristics. NdFeB magnets experience a
reduction in both remanent flux and coercivity for higher temperatures [31], meaning that
rated magnet values, which often are given at room-temperature, are slightly misleading
since the operating temperature tends to be higher. This, in turn, means that the NdFeB
magnets are more sensitive to demagnetization for higher temperatures. Ferrite magnets
are also affected by temperature differences, but as opposed to NdFeB the remanent flux
actually increases with decreasing temperature while the coercivity decreases [31]. This
means that the ferrite magnets are the most sensitive to demagnetization at lower temper-
atures.
32
2.8. Theoretical foundation of design approach
(2h 1)↵m
✓b,h = (2.24)
2
This is done as this part of the design is based on the work conducted in [23] which
advocates constant rotor slot pitch. However, in order to allow for a greater degree of free-
dom in order to minimize the torque ripple and other unwanted side-effects of rotor/stator
slotting as expanded upon in Section 2.6.4 the point (B) is introduced on the periphery of
the rotor which allows for altering the position of the rotor slots. The outermost barrier
slots is shifted further from the q-axis with the displacement angle s . Fig. 2.19 displays
the situation in a machine with three flux barriers. The addition of the displacement angle
entails that all rotor slots are shifted from each-other with equal angles except the slots
closest to the q-axis.
Hence, the rotor slot pitch angle can be calculated as
⇡
s
2p
↵m = (2.25)
1
k+
2
where k is the number of rotor barriers and s can be regarded as a design variable.
33
2.8. Theoretical foundation of design approach
s
(B)
(C)
↵m
↵m
↵m
↵m /2
d
Hence, assuming that there are k barriers the average values seen by the iron seg-
ments for a d- and q-axis MMF, fd,i and fq,i , represented by the steps in Fig. 2.20 can in
per unit be expressed as
✓Z
b,i+1
1 sin ✓b,i+1 sin ✓b,i
fd,i = cos ✓d✓ = i = 0, ..., k 1 (2.26a)
✓b,i ✓b,i+1 ✓b,i
✓b,i
✓Z
b,i+1
1 cos ✓b,i cos ✓b,i+1
fq,i = sin ✓d✓ = i = 0, ..., k 1 (2.26b)
✓b,i ✓b,i+1 ✓b,i
✓b,i
Where the angles are defined as in Fig. 2.10. Note that ✓b,0 =-✓b,1 since the iron segment
aligned with the d-axis is shared with another pole.
It was shown in [15] that if the ratio of the permeances across each flux barrier were
assumed to be constant for any barriers, the following relationship for the flux barrier
34
2.8. Theoretical foundation of design approach
0.5 0.5
0 0
0 10 20 30 40 0 10 20 30 40
Rotor periphery angle Rotor periphery angle
Fig. 2.20: Sinusoidal MMF as function rotor periphery, sinusoidal and averaged step function.
where i and j denotes any barriers in the structure and fq,i is the difference in q-axis
MMF across the i:th flux barrier. Note here that other assumptions can be made with
regards to the permeance ratio which would lead to other relationships, however the con-
stant permeance ratio was utilized as this would allow for a sinusoidal flux distribution in
the air gap [15].
However, if there are k barriers we find that equations (2.27) only gives k 1
equations. Recalling from previous section, the insulation ratio is defined as the ratio
between air and iron in the different directions. Hence, in order to complete the system of
equations we define the following relationship
k
X (R1 Rsh )
Wh,q = (2.28)
1
h=1 1+
kw,q
where R1 and R1 are the rotor radius and shaft radius respectively, as seen in Fig. 2.10.
Furthermore, in order to allow for a near-constant flux density in each iron segment
the iron segment width was set to be proportional to the average d-axis MMF seen by the
segment according to
2S1,q fd,1
= (2.29a)
S2,q fd,2
Si,q fd,i
= i = 2, ..., k (2.29b)
S(i+1),q fd,(i+1)
where the first equation is a product of the angle definition in equation (2.26a).
35
2.8. Theoretical foundation of design approach
Similarly as with (2.27) we have k + 1 unknowns and k equations and in the same
fashion we define the last equation from kw,q according to
k+1
X (R1 Rsh )
Sh,q = (2.30)
h=1
1 + kw,q
The d-axis flux barrier widths are determined by assuming them to be proportional
to their corresponding q-axis widths as
Wi,d Wi,q
= i = 1, ..., k 1 (2.31)
W(i+1),d W(i+1),q
where ↵m is the angle derived from equation (2.25) if s = ↵m /2 and has the value
⇡
2p(k + 1)
Hence, the rotor flux barrier placement and geometry of the SynRM is defined by
the three input variables kw,q , kw,d and s . It is important to note here that the displacement
angle affects each individual barrier width and which barrier is the thickest, this becomes
obvious when looking at the above equations and this should be kept in mind when alter-
ing s such that no barrier becomes unfeasible. Note also that this design approach only
fully defines the rotor geometry if the barrier arm angles ↵i are defined as constant or
considered known (see Fig. 2.10).
36
Chapter 3
Method of analysis
In this chapter, the design process is expanded upon based on the theory described in
the previous chapter. The parametric study is explained and assumptions are justified.
Additionally, the economic evaluation is expanded upon.
3.1 Modeling
The machines, both PMaSynRM and PMSM, were modelled utilizing the finite element
software FLUX. All models created were verified utilizing the analytical/finite element
software SPEED which provides a lesser accuracy at a much faster computational time.
When simulating in FLUX, all models were calculated with imposed pure sinusoidal cur-
rent.
• Average torque
• Torque ripple
• Power factor
• Efficency
Average torque production of a machine is the most obvious and crucial indicator
as, for a given speed, it indicates the power output of the machine and shows how much
load the machine can handle. The average torque calculated over the span of one elec-
trical period since over that period the rotor will see all possible rotor/stator slot relative
positions.
37
3.2. Initial dimensions and target PMSM
The torque ripple is another important aspect in terms of operation. While not a key
factor for pumping applications, as stated earlier, it is an undesirable aspect which should
be minimized if possible The torque ripple is calculated as the peak-peak torque ripple,
given in percent of the average torque.
The power factor is another important factor to control as a high power factor of the
motor can keep the VFD at lower rating and thereby reduce costs. In FLUX, it is possible
to add the end-winding inductances to the model, this in turn allows for the possibility
of retrieving the voltage as seen from the motor terminals from the time-derivative of the
flux. The fundamental voltage and angle was retrieved via FFT and since the current angle
was known, the power factor could be calculated.
The efficiency is an important factor since it determines how great the losses of the
machine is. It is calculated according to its definition as
Pmech
⌘= (3.1)
PM ech + PIron + PM agn + PF ric + PCu
where ⌘ is the efficiency. PM ech is the mechanical output power, PM agn is the losses in
the magnets, PF ric is the frictional losses from bearings and such, PCu is the copper
losses in the stator winding. The iron losses are a sum of three parts, namely Physt , Peddy
and Pstray which correspond to the hysteresis losses, eddy current losses and stray losses
respectively.
The hysteresis and eddy losses were calculated by use of a built-in function in
FLUX, which utilizes the Loss Surface (LS) model. The accuracy of the LS model is quite
good and can handle complex sinusoidal waveforms, but do require prior knowledge of
the material [34]. The hysteresis model utilized by the LS model is expanded upon in [35].
The stray losses, which are very difficult to predict analytically, were assumed constant
for all simulations. Therefore, based on data on stray losses from the utilized IM stator the
SynRM/PMaSynRM stray losses were set to be equal to 140 W. Similarly, for the PMSM
the stray losses were assumed to be 80 W.
Furthermore, the magnet losses in the PMaSynRM was assumed to be zero due to
the fact that the magnets would be mostly be buried in the rotor while the losses in the
surface mounted magnets on the PMSM were calculated in FLUX. The frictional losses
were based on the data from the IM and assumed equal for both the PMaSynRM and the
PMSM and was set to 50 W.
38
3.2. Initial dimensions and target PMSM
Number of slots 48
Outer diameter [mm] 250
Length [mm] 262
Number of turns 9
Number of parallel strands 10
Strand diameter [mm] 0.9
End-winding inductance [mH] 0.34
Stator steel grade M800-50
The current was set to a rated value of 26 A as this was the rated value for the
intended IM and provided a suitable current density value which corresponded to what
was found in literature [30] and the experience at Xylem.
39
3.3. Design procedure
up into segments. This is because it was of the type fractional slot concentrated wind-
ing (FSCW), and in this case it thus has single-tooth windings which gives it a very low
leakage inductance.
The primary rated parameters of the PMSM can be seen in Table 3.2
40
3.3. Design procedure
Barrier
Insulation center PF
Torque
ratios magnet Torque
width
Torque
Air-gap Barrier arm
Torque ripple PF
magnet
Torque
width
Rotor
Torque ripple Efficiency
slot pitch
Steel types (Torque)
(PF)
3 or 4
barriers
previous thesis [10]. Since the primary objective of this thesis were not to achieve an
optimal design for an PMaSynRM but to investigate the feasibility of replacing a PMSM
with a PMaSynRM this algorithm was deemed sufficient to pose as a basis for the design
and comparison. Based on this theory, a parametric study was conducted where the inputs
were varied in order to find the best results. The performance was evaluated at the MTPA
operating point, which shifted in terms of current angle when the geometry changed.
First the SynRM rotor was developed and when the performance of this was satisfactory,
magnets were added in order to further improve its performance. Fig. 3.2 displays the
design procedure in a flowchart together with the most relevant performance parameters
equipped to the different design variables, the blue boxes describes the SynRM design
while the green boxes describe the PMaSynRM design.
41
3.3. Design procedure
were
• q-axis insulation ratio, kw,q
• The tangential ribs were kept constant at 1 mm as their influence was out of the
scope of the analysis and kept at the minimum allowable tolerance from the manu-
facturer at 1 mm
The two base-line designs were based on the same input-values and can be seen in
Table 3.3. Note that the base-line value for s was 5.625 for three barriers and 4.5 for
four barriers.
42
3.4. SynRM design
Fig. 3.3: Impact of q-axis insulation ratio, 0.4 (left) and 0.9 (right).
43
3.4. SynRM design
Fig. 3.4: Impact of d-axis insulation ratio, 0.2 (left) and 0.7 (right).
Fig. 3.5: Impact of rotor slot displacement angle, 0 (left) and 20.625 (right).
3.4.2 Air-gap
While the torque production and power factor benefits of a shorter air-gap length, the
torque ripple increases with lower lengths. Therefore, the air-gap length was investigated
to find the most suitable value. While the mechanical limit was set to 0.6 mm for the
stator due to manufacturing limitations, the investigated interval was set to 0.5 mm to 1
mm in order to investigate how the performance was affected by a lower value for future
references.
44
3.5. PMaSynRM design
• Any calculated rib size smaller than 0.2 mm were neglected and no ribs was utilized
in that barrier
The values for the safety factor to be utilized was set to be ⌫ = {1.5, 2, 2.5, 3}
where the 1.5 value was motivated by the overestimation of the rib sizes.
45
3.5. PMaSynRM design
Ferrite NdFeB
Br [T] 0.430 1.165
Hc,b [kA/m] 310 867
BHmax [kJ/m3 ] 35 267
46
3.6. Performance comparison and economic analysis
were kept constant. For datasheets of the steels refer to Appendix B.1.
3.5.3 Demagnetization
In order to investigate demagnetization of the different magnets, the demagnetization
model as described in Section 2.7.1 was utilized in a built-in macro in FLUX where it
reduces the remanent flux when Hk was exceeded. The threshold for any part of the mag-
net being considered demagnetized was set to be 90% of the nominal remanent flux and
after the simulation the fraction of each magnet that violated that threshold was calculated.
In order to aptly account for as many demagnetization scenarios as possible the
current was set to be 1, 1.5, 2 and 2.5 times the nominal current levels. 2.5 was chosen
to be the upper limit because the drives normally utilized for machines of the intended
purpose has a current limit of around two times the nominal current.
Furthermore, the current angle also influences the demagnetization as the stator flux
should counteract the magnetization of the magnets. Since the magnets are magnetized in
different directions, different current angles were tested. Preliminary tests showed that
the most detrimental current angles were in the span 70 to 110 degrees. While these are
not commonplace current angles, it is important to take them into consideration since
the motor is intended for pumping applications which might lead to unintended clogging
which abruptly stops the machine and these angles might occur for short periods of time.
The demagnetization was tested at an operating temperature at which the magnets
was most at risk for experience demagnetization. For the ferrite magnets, this was deter-
mined to be at 0 C which was deemed the lowest feasible magnet temperature given a
start-up sequence in cold water. For the NdFeB the maximum operating temperature was
set to 120 C.
47
3.6. Performance comparison and economic analysis
production curves as function of current angle at rated current in order to assess the char-
acteristics of the machines. The machines were run at 1500 RPM. Any speed range tests
along with high-speed analysis was not deemed to be of great importance as the speed
range is in part limited by the DC link voltage which for this thesis was not a constraint.
Field-weakening operation was not of great relevance as the load torque for pumping ap-
plications tend to increase with the square of the rotational speed, and therefore to some
extent impractical in CPSR.
The economic analysis were made based on material costs and some production
costs of the stators due to the differences in production methods between the two ma-
chines. The costs of the machines are an important factor in this assessment as it com-
bined with the performance of the machines will help determine for which applications
the PMaSynRM is feasible.
For the economic analysis the following assumptions are made
• The PMaSynRM stator and rotor was assumed to be punched from a single sheet
of steel and excess steel is considered waste. The total steel usage thus is a square
with sides equal to the outer diameter of the machine.
• The PMSM stator segments are smaller and easier to punch and the shape of the
segments allows for less waste and therefore was calculated by assuming that a
rectangular segment with one side as wide as a stator segment and one side 50%
longer than one stator segment has space for two stator segment as exemplified in
Fig. 3.7
• The PMSM rotor is assumed to be punched from a single sheet of steel, given the
consumed material equal to the square of the diameter of the rotor.
• Only the cost per weight of the magnets were considered without taking the press-
ing, cutting or transport of the magnets into consideration.
• Only the manufacturing costs of the stators were taken into consideration as the
rotors are manufactured by a third party and thereby the manufacturing cost is diffi-
cult to estimate. Furthermore, the greatest difference manufacturing-wise is deemed
to be located to the stator. The manufacturing costs was given by the economics de-
partment at Xylem.
48
3.6. Performance comparison and economic analysis
49
Chapter 4
Results
In this chapter, all the results are presented. All relevant plots and figures are given and
discussed to some extent. The values which were carried over are given and the final
design is displayed.
kw,q kw,d g s
Variable Baseline Baseline Baseline
In Fig. 4.1 the results for torque, torque ripple, power factor and efficiency can be
seen respectively. It is useful to note that the torque doesn’t vary particularly much over
this interval, which can be attributed to the fact that the interval was too narrow to see the
detrimental effects of too low or too high values, and only the maxima was achieved.
We can also note that in terms of torque and power factor, the four barrier design
displayed better performance than three barriers whilst generally displaying worse perfor-
mance in terms of torque ripple. However, the torque ripple wasn’t the key performance
indicator for kw,q since the ripple was shown to be mostly governed by the location of
the rotor barrier ends. Thus, the optimum values for both barriers were deemed to be
50
4.1. Rotor design
160
Three barriers Three barriers
0
0.4 0.5 0.6 0.7 0.8 0.9 0.4 0.5 0.6 0.7 0.8 0.9
kw,q kw,q
0.8 0.96
Three barriers Three barriers
0.75 Four barriers 0.955 Four barriers
Power factor
Efficency
0.7 0.95
0.65 0.945
0.6 0.94
0.4 0.5 0.6 0.7 0.8 0.4 0.5 0.6 0.7 0.8 0.9
kw,q kw,q
0.7 and utilized in the upcoming simulation, since this value displayed the highest torque
production and acceptable levels of power factor and efficiency
kw,q kw,d g s
0.7 Variable Baseline Baseline
Here we see that the torque is maximized for both three and four barriers at kw,d =
0.3 and the detrimental impact on torque production of a too large insulation ratio becomes
apparent. Furthermore, we see that the torque ripple is affected negatively by increasing
51
4.1. Rotor design
160
Three barriers
0.8 0.96
Three barriers Three barriers
0.75 Four barriers 0.955 Four barriers
Power factor
Efficency
0.7 0.95
0.65 0.945
0.6 0.94
0.2 0.3 0.4 0.5 0.6 0.7 0.2 0.3 0.4 0.5 0.6 0.7
kw,d kw,d
insulation ratio. Once again, torque ripple wasn’t the key performance indicator for the
insulation ratio and since the increase in ripple is so low for the optimal value for torque
the slight increase is allowed. The same thing applies to the power factor, the maxima for
both power factors are situated at kw,d = 0.5 but since the variation is so small, especially
for the four barrier configuration this non-optimal value is allowed. Furthermore, we see
that the efficiency is negatively affected by increasing insulation ratio for four barriers
while higher values affect the efficiency negatively for three barriers. This can probably
be attributed to increased rotor saturation, which also can explain the decrease in torque
production for greater values of kw,d .
4.1.3 Air-gap
The input data for the air-gap sweep is given in Table 4.3. The air-gap sweep was con-
ducted between 0.4 mm to 1 mm. As can be seen from Fig. 4.3, the torque production
declines almost linearly with the air-gap length. Also, as expected the power factor re-
duces greatly. The efficiency declines, most likely as a product of the reduced mechanical
52
4.1. Rotor design
160
Three barriers Three barriers
0
0.4 0.6 0.8 1 0.4 0.6 0.8 1
Air-gap length [mm] Air-gap length [mm]
0.8
Three barriers Three barriers
Four barriers 0.96 Four barriers
Power factor
0.7
Efficency
0.94
0.6
0.5 0.92
0.4 0.6 0.8 1 0.4 0.6 0.8 1
Air-gap length [mm] Air-gap length [mm]
power. The torque ripple reduction is quite modest considering the drastic reduction of
torque production. Thus, it is visible that it is imperative to maintain the air-gap as short
as possible and a value below 0.6 mm would be beneficial. Given the steep decline of the
torque production and the low decrease in torque ripple, the minimum allowable air-gap
height was chosen, i.e. g = 0.6 mm.
kw,q kw,d g s
0.7 0.3 Variable Baseline
53
4.1. Rotor design
160 80
Three barriers Three barriers
120 40
20
100
0
0 5 10 15 20 0 5 10 15 20
s [deg] s [deg]
0.8 0.96
Three barriers Three barriers
0.75 Four barriers 0.955 Four barriers
Power factor
Efficency
0.7 0.95
0.65 0.945
0.6 0.94
0 5 10 15 20 0 5 10 15 20
s [deg] s [deg]
kw,q kw,d g s
0.7 0.3 0.6 mm Variable
As can be seen from Fig. 4.4 the torque production remains quite constant up until
the higher values of s where it starts to decline. As for the torque ripple however, we
observe a very pronounced influence. As was discussed in Chapter 2 the position of the
rotor slots are very important for the torque ripple which becomes very apparent here. The
power factor displays a slight variance in value, but not significantly so.
Note that the optimum value for three and four barriers are widely separated here,
for three barriers the optimum value is for 9.375 and for four barriers the optimum value
54
4.1. Rotor design
was 0 when considering both torque production and torque ripple. These differeces in
rotor angles are reasonable since the actual rotor slot positions are different for the same
value of s . With previous simulations, the basic SynRM design is concluded with the
performance as seen in Table. 4.5.
3 barriers 4 barriers
Torque 117.56 119.96
Torque ripple 10.32 6.85
Power factor 0.649 0.650
Efficiency 0.952 0.953
After these simulations it could be concluded that for this 48 slot stator, four bar-
riers were the best performing configuration in all regards. This resulted in the choice of
continuing the simulations with only the four-barrier alternative with the design parameter
values which can be seen in Table. 4.6
kw,q kw,d g s
0.7 0.3 0.6 mm 0
Table 4.7: Radial rib sizes for different values of safety factors
Safety factor, ⌫ wr,1 [mm] wr,2 [mm] wr,3 [mm] wr,4 [mm]
1.5 1.4986 1.0000* 1.0000* 0*
2 1.9981 1.1797 1.0000* 0*
2.5 2.4976 1.4747 1.0000* 0*
3 2.9971 1.7696 1.0000* 0*
The results can be seen in Fig.4.5. As expected, the torque production is affected by
the introduction of ribs and we also see a decrease in power factor. The torque reduction
55
4.1. Rotor design
10
4
100
2
0
No ribs 1.5 2 2.5 3 No ribs 1.5 2 2.5 3
⌫ ⌫
0.7 0.96
0.68 0.958
Power factor
Efficency
0.66 0.956
0.64 0.954
0.62 0.952
0.6 0.95
No ribs 1.5 2 2.5 3 No ribs 1.5 2 2.5 3
⌫ ⌫
Fig. 4.5: Simulation results for different values of ⌫ on four barrier rotor.
and PF reduction from the case where no ribs were utilized as a function of safety factor
can be seen in Tab. 4.8.
For the forthcoming simulations it was determined that the safety factor of 1.5 was
sufficient, following the discussion with regards to the overestimation of the rib width.
56
4.1. Rotor design
57
4.1. Rotor design
58
4.2. PMaSynRM design
59
4.2. PMaSynRM design
160
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Fraction of center barrier filled Fraction of center barrier filled
0.9 0.97
0.965
Power factor
0.8
Efficency
0.96
0.7
0.955
0.6 0.95
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Fraction of center barrier filled Fraction of center barrier filled
4.2.3 NdFeB
The NdFeB magnets were filled in all barriers including the outermost, as that topology
would provide the base for the economic calculations. The tolerance for the NdFeB flux
linkage versus the ferrite flux linkage was 1%. The magnet fill factor for the center barrier,
kmag,a , was set to 0.4 while for the barrier arms, kmag,b was set to 0.3625. The results of
the simulations can be seen in Table. 4.11
It is quite noticeable that the torque production is lower for the NdFeB, while the
torque ripple is higher. This can probably be attributed to the distribution of the magnets in
the barriers and the fact that the magnet flux does not affect all parts of the iron segments
in the same way since there are large portions of air in the flux barriers with NdFeB
magnets in the rotor. However, these results were deemed to be sufficient as basis for
the economic comparison since the performance of the two machines are similar to each
other.
60
4.3. Demagnetization
160
0
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Fraction of center barrier filled Fraction of center barrier filled
0.9 0.97
0.965
Power factor
0.8
Efficency
0.96
0.7
0.955
0.6 0.95
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
Fraction of center barrier filled Fraction of center barrier filled
4.3 Demagnetization
The demagnetization was studied for a variety of current magnitudes and angles. The basis
for this simulation was the ferrite rotor with all barriers filled. The current levels analyzed
61
4.3. Demagnetization
Ferrite NdFeB
Torque [Nm] 156.80 152.82
Torque ripple [%] 7.97 10.45
Power factor 0.85 0.86
Efficiency 0.962 0.962
were I = {In , 1.5In , 2In , 2.5In } and for every current value the current angles 70 , 80 ,
90 , 100 and 110 were investigated. The knee values utilized for the demagnetization
studies can be seen in Table 4.13
Due to limitations in FLUX, all magnet could not be investigated. Therefore only
half of the magnets in the center barriers were analyzed by assuming that the demagneti-
zation of these would be low due to their position and that any noticeable effect would be
fairly symmetrical. The analyzed magnets and their names can be seen in Fig. 4.9.
In Fig. 4.10 the amount of demagnetization is plotted for the percentually most
affected magnet for each current and angle. This figure serves as an indication for how
the demagnetization affected the machine, it is however not very descriptive for the state
of demagnetization throughout the machine as the higher current values demagnetized
more magnets a significant amount compared to the lower.
As it is hinted in Fig. 4.10, the demagnetization at nominal current was negligible
with values well below 1% demagnetization for all magnets. For the 1.5 times the nominal
current the situation changed somewhat but here the demagnetization really only affected
two magnets, namely L3 and R3 for all current angles except 90 . All demagnetization
values for nominal and 1.5 times nominal current can be seen in Table 4.14 where the
highest values for each current level is highlighted. However, the demagnetized areas for
the relevant magnets were not particularly worrisome since the values are not especially
high and the demagnetization is localized to the edges of the magnets as seen in Appendix
C, Fig. C.1, note that in these plots every color other than the main color signifies a value
below the threshold value.
Similarly, the demagnetization values for 2 and 2.5 times the nominal current can be
seen in Table 4.15 where the greatest values are highlighted. For these values a noticeable
trend appears which is that the most affected magnets are R3 and L3 and mostly around
62
4.3. Demagnetization
Ferrite NdFeB
Br [T] 0.45 1.025
Hk [kA/m] 360 1195
100 for R3 and 70 for L3 . This is reasonable since R3 and L3 are positioned such that the
stator flux direction is mostly counteracting the remanent flux of the magnets. Similarly,
it can be noticed that the demagnetization for barriers 2 and 1 increases with the angular
distance from 90 . Additionally we see that the center barriers are well protected from de-
magnetization as was stated in Chapter 2. For these two values, the demagnetization began
to be significant. However, utilizing the same argument as earlier, the demagnetization for
2 times the nominal current was deemed to be insignificant for producing a noticeable
performance reduction. For 2.5 times the nominal current, however, the demagnetization
of especially barrier 3 and 2 begin to be of significant magnitude. The demagnetization
for these two current levels can be seen in Appendix C, Fig. C.2.
A similar analysis were conducted on the NdFeB rotor aswell, and as expected
almost no demagnetization occurred as can be seen in Table. 4.16 where only the highest
current is displayed. The low demagnetization can be easily explained by the high values
of both remanent flux and coercive force that characterizes the NdFeB magnets.
63
4.3. Demagnetization
L1 L2 L3
CL4
CL3
CL2
CL1
R3
R2
R1
14 = 70
= 80
Amount demagnetized [%]
12
= 90
10 = 100
= 110
8
0
In 1.5In 2In 2.5In
64
4.3. Demagnetization
Table 4.14: Percent of ferrite magnet demagnetized below 0.41 T for In and 1.5In
Current In 1.5In
Angle 70 80 90 100 110 70 80 90 100 110
L1 0.39 0.29 0.00 0.02 0.05 1.05 0.77 0.31 0.19 0.36
L2 0.35 0.35 0.35 0.12 0.00 1.45 1.45 0.35 0.12 0.25
L3 0.00 0.00 0.00 0.00 0.00 3.89 3.89 0.00 0.00 0.00
R1 0.06 0.02 0.00 0.26 0.43 0.40 0.26 0.32 0.91 1.07
R2 0.00 0.11 0.33 0.33 0.33 0.35 0.45 0.67 1.73 2.08
R3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 3.88 3.88
CL1 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
CL2 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
CL3 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
CL4 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
Table 4.15: Percent of ferrite magnet demagnetized below 0.41 T for 2In and 2.5In
Current 2.5In
Angle 80 90 100 110
L1 0 0 0 0
L2 0.3221 0 0 0
L3 0 0 0 0
R1 0 0 0 0
R2 0 0 0.3273 0.3273
R3 0 0 0 0
65
4.4. PMSM versus PMaSynRM
PMaSynRM PMSM
Torque [Nm] 156.80 163.03
Torque ripple [%] 7.97 1.19
Efficiency 0.962 0.966
In Fig. 4.11 the torque and power factor of the machines are plotted as function of
the current vector angle. We see that the PMaSynRM maximum torque (MTPA operating
point) occurs at about 47 while the PMSM occurs at about 0 , even though none of the
machines display any particularly good power factor values at this point it is interesting
to note that the PMaSynRM in fact has a better value at this point. Both machines dis-
play a peak PF (MTPkVA operating point) of very close to 1. However, it is worth noting
that none of these operating points pose with a particularly high torque value. However,
it is very noticeable that the PF-curve is much flatter for the PMSM than for the PMa-
SynRM, meaning that for most torque values in these charts the PMSM is better than the
PMaSynRM.
66
4.4. PMSM versus PMaSynRM
160 1
140
120 0.8
Power Factor
Torque [Nm]
100
0.6
80
60 0.4
40
0.2
20
0 0
0 20 40 60 80
[deg]
(a) PMaSynRM
160 1
140
120 0.8
Power Factor
Torque [Nm]
100
0.6
80
60 0.4
40
0.2
20
0 0
0 20 40 60 80
[deg]
(b) PMSM
Fig. 4.11: Torque and power factor as function of current vector angle .
67
4.4. PMSM versus PMaSynRM
prices to the middle values of both these intervals, we find that the cost is divided up as
seen in Table. 4.19
Table 4.19: Cost breakdown of PMaSynRM and PMSM in middle of magnet cost intervals
[pu]
PMaSynRM PMSM
Magnets Ferrite NdFeB NdFeB
Steel M800 M600 M400 M800 M600 M400 M800/M600
Steel 0.188 0.193 0.195 0.188 0.193 0.195 0.088
Magnets 0.091 0.091 0.091 0.279 0.279 0.279 0.474
Copper 0.281 0.281 0.281 0.281 0.281 0.281 0.155
where the category ”Misc. mat.” is miscellaneous materials such as tape, zip ties,
stator housing and slot insulation. In the ”Assembly” category costs for operations such
as winding procedures, impregnation, insulation mounting etc are represented. As can be
seen from this table, the ferrite PMaSynRM with M800 steel became only slightly cheaper
than the PMSM whereas the NdFeB PMaSynRM proved to be more expensive than the
PMSM throughout the cost interval and was therefore deemed not cost-effective.
However, given that the prices for the magnets are by no means fixed and tend to
fluctuate this had to be accounted for in order to see if there were any breaking points
in terms of price differences. In Fig. 4.12 the price for both ferrite and NdFeB magnets
are varied for the M800 ferrite PMaSynRM and PMSM. Here we see that the PMSM is
in fact cheaper than the PMaSynRM for lower prices of NdFeB. Additionally, we notice
that the PMaSynRM cost is not particularly sensitive to price fluctuations of the ferrite
magnets as it only varies between 0.97-1.03 pu. Conversely, the PMSM is very dependent
on the magnet price as it varies between 0.89-1.20 pu. Thereby it can be seen that the
68
4.4. PMSM versus PMaSynRM
1.4
PMaSynRM
PMSM
1.2
0.8
0.6
0.1722 0.2066 0.2410 0.2755 0.3099 0.3443
NdFeB price [pu/kg]
breaking point occurs somewhere around an NdFeB price of 0.23 pu, depending on the
cost of the ferrite magnets. If the price of ferrite and NdFeB would follow each-other,
the PMaSynRM proves to be 14% cheaper than the PMSM at the maximum of the price
interval. If ferrite would be at its cheapest and NdFeB at its most expensive, the PMa-
SynRM would be almost 20% cheaper than the PMSM. On the other hand, if it were to
be the other way around, that NdFeB is at its cheapest and ferrite at its most expensive the
PMaSynRM would actually be 16% more expensive than the PMSM.
However, it is interesting to note that these prices are mostly due to the higher
production cost of the stator utilized for the PMaSynRM. As seen in Table 4.19, the cost
of assembly for the PMaSynRM stator is almost double that of the PMSM which is an
important factor. Related to this, in Table. 4.20 a cost breakdown in percent for each
machine is given when the magnets are at their most expensive. The magnets in the PMSM
Table 4.20: Cost breakdown for PMaSynRM and PMSM at peak magnet price [%]
PMaSynRM PMSM
Steel 18.77 8.41
Magnets 9.09 45.35
Copper 28.11 14.80
Misc. mat. 17.68 20.12
Assembly 26.34 11.32
correspond to almost half of the total cost of the machine in contrast to only 9% for the
69
4.4. PMSM versus PMaSynRM
PMaSynRM. For the PMaSynRM, the significant cost could be found mainly in the copper
usage and assembly cost. The copper cost is reasonable since the long end-windings and
longer motor adds to the overall cost.
The assembly is an interesting aspect to analyze, due to the PMSM having a single-
tooth winding with stator segments rather than a stator plate allows for much cheaper
assembly and winding cost.
70
Chapter 5
71
5.2. Economic feasibility of PMaSynRM
72
5.3. Future work
73
Appendix A
General calculations
⇡
+ = + 'i (A.1)
2
= + (A.2)
In the second equality in equation (2.9) we have that the IPF can be described as
⇡
cos ( + ). Consider the following expression
2
Ld id iq
+
⇡ Lq iq id
tan ( + ) = cot ( )= (A.3)
2 Ld
1
Lq
where the last equality holds because
cot cot + 1
cot ( )= (A.4)
cot cot
Hence, we find that the internal power factor can be described as
0 1
Ld id iq !
+
B Lq iq id C
IPF = cos arctan @ B C (A.5)
Ld A
1
Lq
where the definition of the angles in along with equation (2.5) has been used. Using the
relation that for any argument x we have that
1
cos (arctan x) = p (A.6)
1 + x2
74
A.2. Center of gravity of rotor segments
r is a point in the volume, dm is the infinitesimal mass of the volume. Since the rotor is
cylindrical and assuming that the mass density is constant (since we neglect the air in the
barriers) we find that we can write the following expression
RR
r2 cos ✓drd✓ num
Rc = RR = (A.13)
rdrd✓ den
where ✓ is the segment angles. Note here that the cos ✓-factor is because of the
symmetry of the segments, meaning that the center of gravity is along the q-axis.
75
A.2. Center of gravity of rotor segments
Z✓b,i ZR1
2 3
num = 2 r2 cos ✓drd✓ = R sin ✓b,i 3
D0,i tan ✓b,i (A.14)
3 1
0 D0,i
cos ✓
✓
Z ZR1
b,i
76
Appendix B
Data sheets
B.1 Steel
In this Appendix, all the steel data-sheets is presented. All data-sheets are retrieved from
[36].
77
B.1.1 M400-50A
AdhhVi&#*I!*%=o!L$`\ (!*,
AdhhVi&#%I!*%=o!L$`\ &!).
6c^hdigdend[adhh! -
BV\cZi^XedaVg^oVi^dcVi*%=o
=2'*%%6$b!I &!*.
=2*%%%6$b!I &!+-
=2&%%%%6$b!I &!,.
8dZgX^k^in98!6$b *%
GZaVi^kZeZgbZVW^a^inVi&#*I &%*%
GZh^hi^k^in!¥7Xb )'
N^ZaYhigZc\i]!C$bb ('*
IZch^aZhigZc\i]!C$bb )+*
Ndjc\»hbdYjajh!G9!C$bb '%%%%%
Ndjc\»hbdYjajh!I9!C$bb '&%%%%
=VgYcZhh=K*KEC &+*
AdhhVi&#*I!*%=o!L$`\ *!&,
AdhhVi&#%I!*%=o!L$`\ '!()
6c^hdigdend[adhh! +
BV\cZi^XedaVg^oVi^dcVi*%=o
=2'*%%6$b!I &!+(
=2*%%%6$b!I &!,'
=2&%%%%6$b!I &!-(
8dZgX^k^in98!6$b -*
GZaVi^kZeZgbZVW^a^inVi&#*I &++%
GZh^hi^k^in!¥7Xb (%
N^ZaYhigZc\i]!C$bb '-*
IZch^aZhigZc\i]!C$bb )%*
Ndjc\»hbdYjajh!G9!C$bb '&%%%%
Ndjc\»hbdYjajh!I9!C$bb ''%%%%
=VgYcZhh=K*KEC &'*
AdhhVi&#*I!*%=o!L$`\ +!+%
AdhhVi&#%I!*%=o!L$`\ (!%*
6c^hdigdend[adhh! *
BV\cZi^XedaVg^oVi^dcVi*%=o
=2'*%%6$b!I &!+*
=2*%%%6$b!I &!,)
=2&%%%%6$b!I &!-*
8dZgX^k^in98!6$b &%%
GZaVi^kZeZgbZVW^a^inVi&#*I &-&%
GZh^hi^k^in!¥7Xb '(
N^ZaYhigZc\i]!C$bb (%%
IZch^aZhigZc\i]!C$bb )&*
Ndjc\»hbdYjajh!G9!C$bb '&%%%%
Ndjc\»hbdYjajh!I9!C$bb ''%%%%
=VgYcZhh=K*KEC &(%
B.2 Magnets
In this section, all the data-sheets for the magnets are presented. The ferrite data-sheet is
retrieved via contact with the supplier [37]. The NdFeB data-sheet is retrieved from [38].
81
B.2. Magnets
B.2.1 Ferrite
82
B.2.2 NdFeB - N33EH
N33EH
Sintered Neodymium-Iron-Boron Magnets
Thermal Properties
of Induction, α(Br) %/ºC -0.120
of Coercivity, α(Hcj) %/ºC -0.420
Coefficient of Thermal Expansion (2)
-6
Characteristic Units min. nominal max. ∆L/L per ºCx10 7.5 -0.1
Gauss 11,300 11,650 12,000 Thermal Conductivity W / (m • K) 7.6
Br, Residual Induction (3)
mT 1130 1165 1200 Specific Heat J / (kg • K) 460
Magnetic Properties
Properties
Oersteds 30,000 MPa 285
HcJ, Intrinsic Coercivity
Other
kA/m 2,388 Density g/cm3 7.5
MGOe 31 34 36 Hardness, Vickers Hv 620
BHmax, Maximum Energy Product
kJ/m3 247 267 287 Electrical Resistivity, • cm 180
Notes: (1) Coefficients measured between 20 and 200 ºC
(2) Between 20 and 200 ºC (3) Between 20 and 140 ºC
kG Tesla
Pc = B H
0.5
Material: N33EH 0.75 1 1.5 2 3 5
15
14 1.4
13
12 1.2
-40°C
11
20°C
Flux Density, B
10 1.0
0.3 9
8 0.8
7
60°C
Polarization, J
6 0.6
80°C 5
100°C 4 0.4
120°C
0.1 3
150°C
2 0.2
180°C
200°C
220°C
1
0 0
30
kOe 28 26 24 22 20 18 16 14 12 10 8 6 4 2 0
kA/m 2230 2070 1910 1750 1590 1430 1275 1115 955 795 640 475 320 160 0
Demagnetizing Field, H
1 kA/m = 12.566 Oe 1 kOe = 79.577 kA/m
Notes The material data and demagnetization curves shown above represent typical properties that may vary due to product shape and size.
Magnets can be supplied thermally stabilized or magnetically calibrated to customer specifications.
Additional grades are available. Please contact the factory for information.
Results
C.1 Demagnetization
84
C.1. Demagnetization
85
C.1. Demagnetization
86
References
[1] R. Saidur, “A review on electrical motors energy use and energy savings,” Renewable
and Sustainable Energy Reviews, vol. 14, no. 3, pp. 877 – 898, 2010. [Online].
Available: http://www.sciencedirect.com/science/article/pii/S1364032109002494
[6] J. Rowlatt, “Rare earths: Neither rare, nor earths,” BBC News, Mar 2014, [Accessed
2018-03-13]. [Online]. Available: http://www.bbc.com/news/magazine-26687605
[7] C. Bontron, “Rare-earth mining in china comes at a heavy cost for local villages,”
The Guardian, Aug 2012, [Accessed 2018-03-13]. [Online]. Available: https://www.
theguardian.com/environment/2012/aug/07/china-rare-earth-village-pollution
87
References
[18] N. Bianchi, E. Fornasiero, and W. Soong, “Selection of pm flux linkage for max-
imum low-speed torque rating in a pm-assisted synchronous reluctance machine,”
IEEE Transactions on Industry Applications, vol. 51, no. 5, pp. 3600–3608, Sept
2015.
88
References
[26] C. Babetto, G. Bacco, and N. Bianchi, “Analytical approach to determine the power
limit of high-speed synchronous reluctance machines,” in 2017 IEEE International
Electric Machines and Drives Conference (IEMDC). IEEE, may 2017. [Online].
Available: https://doi.org/10.1109/iemdc.2017.8002316
89
References
[34] A. Krings and J. Soulard, “Overview and comparison of iron loss models for electri-
cal machines,” Journal of Electrical Engineering, vol. 10, no. 3, pp. 162–169, 2010.
90