International Conference on Power Electronics Systems andApplications (PESA 2009)
A Novel OuterRotor PermanentMagnet FluxSwitching Machine forUrban Electric Vehicle Propulsion
W. Fei
1
, P.C.K. Luk
1
, J. Shen
2
, Y. Wang
2
1
Department of Engineering Systems, Cranfield University, Shrivenham SN6 8LA, U.K.
2
College of Electrical Engineering, Zhejiang University, CHINA.Email: p.c.k.luk@cranfield.ac.uk
Abstract – A novel twelvestatorpole, twentytworotorpole(12/22) outerrotor permanentmagnet fluxswitching (PMFS)machine for electric propulsion in a lightweight electricvehicle is presented. Analytical equations are derived for thedimensioning of a 3phase 5kW inwheel motor. Optimisationtechniques are employed to maximise the performance of themachine. The validity of the analytical equations is verifiedby finite element analysis (FEA). The results show that thePMFS machine combines the key advantages of the PMmachines and the switched reluctance machine, and thusdemonstrate the viability of the proposed machine as asuitable candidate for inwheel electric propulsion.Keywords – Outerrotor permanentmagnet fluxswitchingmachines, electric vehicle propulsion, inwheel motor.
I. I
NTRODUCTION
Lightweight zeroemission electric vehicles, powered bybatteries or fuel cells, have been identified as an importantandrapidlygrowing market in providing personal transportin the urbanized parts of the developed and developingworlds [1]. The electric drivetrain system is central tosuccessful implementations of these lightweight electricvehicles. There are two main configurations suitable forthe electric drivetrain system: a) centralized motor drivewith reduction gears and a differential axle, b) inwheeldirect drives with independent control. Whilst thecentralized drive appears to be more popular partly due toits similarity with existing mechanicalbased systems, theinwheel direct drive configuration is gaining popularity asa more a electric motorbased system that offerspotentially superior features. Due to the absence of mechanical transmission, differential gears and drive belts,the inwheel direct drivetrain, shown in Fig.1, has higherefficiency, weight reduction, and increased vehicle space.It also helps to improve vehicle maneuverability andcontrol as a result of independent control of each wheel.Moreover, the increased unsprung mass, one of the keydisadvantages of inwheel drives, is less critical for lighturban vehicles traveling at reduced speed limits. Thismakes the inwheel configuration to be particularlyacceptable in the urban electric vehicle market. Permanentmagnet (PM) brushless machines with an outer rotor havebeen used as inwheel direct drive motors for electricvehicles, due to their high torque density, excellentefficiency and overload capability [2]. However, theremay be risks of demagnetization and mechanical damageof the rotor’s magnets in extreme driving conditions. Onthe other hand, the switched reluctance (SR) motor has avery simple and rugged rotor structure without magnets,which makes it particularly robust against mechanical andthermal impacts. However, SR motors are replete withlarge torque ripples which are not suitable for direct drive.Although special inwheel switched reluctance motorswith very small torque ripples have been proposed [3], thegenerated electromagnetic torque is reported to be smalleras a result of the special magnetic configuration used toreduce the torque ripple, making it less attractive as adirect drive without a very large rotor diameter.Although introduced more than half a century ago, thepermanentmagnet fluxswitching (PMFS) machine is arelatively less well known doublysalient PM brushlessmachine. It has a salient passive rotor that is exactly thesame as a SR machine, and has a stator with both windingsand magnets. Thus, through careful machine design, thePMFS machine can inherit selected advantages of both of the SR machine and PM synchronous machine [48]. It hasbeen shown that the machine can have the distinctattributes of high torque density, high efficiency, excellentfluxweakening capability and convenience of cooling [914]. Whilst it is noteworthy that some of these are highlydesirable features for motor drives in electric vehiclepropulsion, there appears no research study on PMFSmachines used as an inwheel direct drive, where an outerrotor configuration is preferably used. When comparedwith the conventional innerrotor machine, the outerrotorconfiguration is more suitable for direct drive in terms of higher generated torque at low speed. This paper proposesa novel 5kW outerrotor PMFS machine as an inwheelmotor drive, which meets the specifications of an urbanelectric vehicle. The topology and operation principle of the outerrotor PMFS machine are first introduced. Then,the sizing equations are derived for initial machinedimensions. Design details of a 12/22pole 3phase inwheel outerrotor PMFS machine are described. Finiteelement (FE) method is used for performance predicationand validation. The predicted results demonstrates theviability of the proposed machine as a promising candidatefor inwheel direct drive applications.
Figure 1: Inwheel direct drivetrain with independent control
International Conference on Power Electronics Systems andApplications (PESA 2009)
II. T
OPOLOGY AND
O
PERATION
P
RINCIPLE
1. Topology
The PMFS machine belongs to the family of fluxswitching machines, which are a form of doubly salientreluctance machine with a novel topology. The concept of the flux switching involves changing the polarity of theflux linking the armature winding due to the motion of therotor. All excitation sources are on the stator, with thearmature and field allocated to alternate stator teeth. Forthe PMFS machine, the field is provided by the permanentmagnets. Figure 2 shows the crosssection of the proposed12/22pole, 3phase outerrotor PMFS machine. Themagnets are inset in the middle of the stator poles,segmenting the machine stator yoke as a result. Heatextraction is easier to manage as all major heat sourceslocated in the stator. Moreover, the number of polepairsin the PMFS machine is the same as the number of rotorteeth, which facilitates a high number of polepairs andhence higher torque at low speed. The latter is a highlydesirable feature for direct drive traction motors. Toincrease sufficient winding area in the outerrotor PMFSmachine, the slotteeth ratio is proposed as
r
=
s
=
h
pm
=
h
slot
/5 as shown in Figure 2(b), instead of
r
=
s
=
h
pm
=
h
slot
commonly used in conventional innerrotor PMFS machines [7]. This is factored into themachine equation used to define a polyphased structure asfollows:
22
r s
n N N q
(1)where
N
r
and
N
s
are the respective machine rotor andstator pole numbers,
q
is the number of machine phases,and
n
is a natural number. The relationship between themachine’s mechanical rotation frequency
F
and theelectrical frequency
f
is:
r
f N F
(2)Due to the unique structure and zero net radial stress of themachine, both Nr and Ns should be made to be an evennumber. Thus, for the proposed machine,
q
=3,
N
s
=12,
N
r
=22 and
n
=1.
2. Operation principle
The concept of PMFS machine was firstly introduced [4].The basic operation principle of the outerrotor PMFSmachine is illustrated as in Figure 3. The upper part is thelaminated rotor similar to that of the SR machine. The lowerpart is the stator, where the PM and the armature windingare located. The PM is inset between two stator poles, andestablishes a selfexcited flux with a fixed direction withinitself. In Figure 3 (a), the two rotor poles align with the twostator teeth which are embraced by a concentrated windingcoil. The PM flux which is linked in the coil goes out of thecoil and into the rotor pole. When the rotor moves to theleft, the rotor pole leaves the current stator pole and alignswith the next stator tooth which belongs to the same coil, asshown in Figure 3 (b). As the rotor moves, both themagnitude and polarities of the fluxlinkage in the windingswill vary accordingly.
3. Finite Element Modelling
As a result of 2D FE modelling, the opencircuit fluxdistributions of four typical rotor positions of the machineare obtained as shown in Figure 4. In Figure 4(a),
r
=0
o
when the rotor poles are fully opposite to the respectivestator teeth belonging to PhaseA. In accordance with thepolarity of the magnet inset, the fluxes go into thecorresponding stator teeth. Thus, the PM fluxlinkage of PhaseA reaches the positive maximum when the flux goinginto the stator teeth is defined as positive. In Figure 4(b),with the rotor moves to
r
=4.09
o
anticlockwise, two of therotor poles leave the corresponding stator poles belongingto PhaseA, and other two of them align with the magnets inthe rest of PhaseA stator poles. Therefore, the PM fluxlinkage of PhaseA reduces to zero. In Figure 4 (c), the PMfluxlinkage of PhaseA increases to the maximum againwhen rotor reaches
r
=8.18
o
. However, the polarity changesto negative since the fluxes go out of the stator teeth. InFigure 4(d) where
r
=12.27
o
, the PM fluxlinkage of PhaseA reduces to zero again. If the rotor moves a further 4.09
o
, itwill return to the initial position as illustrated in Figure 4(a)to complete one electric cycle. It is evident from theforegoing analysis that the phase PM fluxlinkage in theouterrotor PMFS machine is bipolar, which brings aboutsome of the machine’s outstanding performance.
Figure 2:Topology of a 12/22pole threephase outerrotorPMFS machine, (a Crosssection. (b Main dimension.Figure 2:Topology of a 12/22pole threephase outerrotorPMFS machine. (a) Crosssection (b) Main dimensionFigure 3: Operation principle at two typical rotor positions
International Conference on Power Electronics Systems andApplications (PESA 2009)
III. M
ACHINE
M
ODELLING AND
D
ESIGN
1. Analytical Sizing Equations
Analytical sizing equations are employed in thepreliminary stages of machine design. They cansignificantly improve the machine design efficiency bysaving invaluable computational time otherwise necessaryin modelling methods. The analytical sizing equations of the outerrotor PMFS machine are derived from firstprinciples. When the stator outer radius
R
so
is given, thestator tooth width
s
, stator magnet thickness
h
pm
, backironthickness
h
b
and slot opening width
h
slot
are derived as:
5 4
slot sos pm bs
h Rh h N
(3)Then, the area of one stator slot is:
2 2
5sin tan8
s sos s
A R N N
(4)The electromagnetic torque
T
em
can be obtained as follows:
3 2
5sin16 8tan
r d p g peak sosems
N K K B J R l N T N
(5)where
K
d
and
K
p
are the leakage and winding packingfactors respectively,
B
g
is the peak airgap flux density atno load condition,
J
pea
k
is the peak current density of thecoils, and
l
is the active length of the machine. Theleakage factor in outerrotor machine is far biggercompared with that of the conventional innerrotormachine due to its large slot opening. Equation (5) showsthat the torque output is proportional to
R
so
3
. And
R
so
and
l
can be determined from equation (5) at preliminary designstage. In addition, the rotor pole height
h
pr
is chosen as 1/8of the stator outer radius
R
so
and rotor yoke thickness
h
yr
isdesigned as twice the stator back iron thickness
h
b
toalleviate possible mechanical vibration. Hence, themachine radial dimension can be expressed as:
98 2
so soos
R R R g N
(6)where
R
o
is the rotor outer radius, and
g
is the machine’sairgap.
2. Machine Optimization
For a three phase machine, a common stator pole numberof 12 is chosen. Higher stator pole number will result inhigher eddy current losses and increased assemblycomplexity. The basic machine dimensions, which areshown in Table 1, are determined by equations (3) to (6)with
T
em
=50N.m,
N
r
=22,
N
s
=10,
K
d
=0.75,
K
p
=50%,
B
g
=2.0T,
J
peak
=7.5A/mm
2
,
l
=50mm, and
g
=0.6mm.T
ABLE
1:
MACHINE PARAMETERSSYMBOL MACHINE PARAMETER VALUE UNIT
N
S
N
UMBER OF
S
TATOR
P
OLES
12
N
R
N
UMBER OF
R
OTOR
P
OLES
22
R
ro
R
OTOR OUTER RADIUS
94.8
MM
R
so
S
TATOR OUTER RADIUS
75
MM
U
DC
DCL
INK
V
OLTAGE
42 V
N
R
ATED
S
PEED
1000
RPM
P
EM
R
ATED
O
UTPUT
P
OWER
5.2
K
W
T
EM
R
ATED
O
UTPUT
T
ORQUE
50 N
M
L
D
D

AXIS
I
NDUCTANCE
62.7
H
L
Q
Q

AXIS
I
NDUCTANCE
72.0
H
Comprehensive FE models are constructed to validate theanalytical sizing equations. These models are then used todetermine other rest machine parameters, and to optimizethe machine performance. Initial FEA results show thatthat the backEMFs of the machine are essentiallysinusoidal. Thus, the machine performance can beanalyzed based on
dq
coordinates, and the machineelectromagnetic torque can then be expressed as:
2 2
8
r p sem pm d q d q
qN K AT J J
(7)where
pm
is the phase PM flux per turn,
d
and
q
arethe
dq
 axes permeance per turn, and
J
d
and
J
q
are the
dq
axes current density, respectively, which are defined by:
2 2 2
d q peak
J J J
(8)Machines with various rotor pole width
r
are studied tooptimize the machine backEMF waveform [11]. In thispaper, the rotor pole width is also employed to optimizethe machine performance. It should be noted that the rotorpole width aftermentioned is the normalized value
r
/
s
.
Based on FEA, the phase PM flux per turn can be directlyderived from open circuit field analysis and the
dq
axespermeance per turn can be calculated by the simplified twoposition method [13].
Figure 4: Opencircuit field distribution at four typical rotorpositions
(a)
r
=0
o
, (b)
r
=4.09
o
, (c)
r
=8.18
o
,(d)
r
=12.27
o
.
International Conference on Power Electronics Systems andApplications (PESA 2009)
0.00110.00130.00151 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Rotor Pole Width
P e r m e a n c e ( m W b A
 1
)
1.21.31.41.5
F l u x ( m W b )
daxis permeanceqaxis permeancePMflux
Figure 5: Variations of
dq
axes permeance per turn and phase PMflux per turn with rotor pole width.
434649521 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Rotor Pole Width
E l e c t r o m a g n e t i c T o r q u e ( N . m )
00.40.81.2
R e l u c t a n c e T o r q u e ( N . m )
ElectromagneticTorqueReluctance Torque
Figure 6:Variations of electromagnetic/reluctance torque withrotor pole width at rated current density
From the FEA results shown in Figure 5, the discrepancybetween
dq
 axes permeance diminishes as the rotor polewidth increases, as a result of machine saliency attenuation.And the phase PM flux reaches its maximum when the rotorpole width approaches 1.4. The maximum electromagnetictorque and the corresponding reluctance component at agiven current density can be deduced from equations (7)and (8). Figure 6 shows the variations of electromagneticand reluctance torque with rotor pole width at rated currentdensity, which are calculated based on the FEA results inFigure 5. It is noted that the reluctance torque isinsignificant compared to the total electromagnetic torque.Also, it diminishes as machine rotor pole width increasesand the saliency reduces. Furthermore, the generatedelectromagnetic torque at rated current density reaches themaximum with the rotor pole width equals 1.4.The backEMF obtained for each particular rotor pole widthis then analyzed and the belt (nontriplen) harmonicdistortion (BHD) in the phase backEMF is determined. Thefundamental amplitude of the phase backEMF and BHDfor different rotor pole widths are illustrated in Figure 7.Similar to the phase PM flux and electromagnetic torque,the fundamental amplitude of the phase backEMF reachesits maximum when the rotor pole width is 1.4. However,BHD reaches minimum with rotor pole width equals 1.3.Assessment of the cogging torque of the machine isparticularly important for the direct drive application. Thereis no reduction gear to absorb the undesirable effects of vibration, noises and startup hesitation due to coggingtorque. The peaktopeak cogging torque variation withdifferent rotor pole width is comprehensively studied asshown in Figure 8. The lowest cogging torque occurs withthe rotor pole width equals 1.2.
2.93.13.33.51 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2
Rotor Pole Width
E M F ( V )
0.0%1.5%3.0%4.5%
B H D %
FundamentalBHD
Figure 7: Variations of fundamental amplitude of phase backEMF and belt harmonics with rotor pole width.
048121 1.2 1.4 1.6 1.8 2
Rotor Pole Width
C o g g i n g T o r q u e ( N . m )
Figure 8: Variation of peaktopeak cogging torque with rotorpole width.
3. Machine Parameters
From the analysis of BHD, a rotor pole width of 1.3 ischosen to achieve the optimal machine performance. Nowthe only outstanding unknown parameter is the windingturns per coil
N
coil
, which can be calculated by the equationsin [12]. The PM fluxlinkage
pm
and
dq
axes inductances
L
d
/
L
q
can be derived by:
2 2
, ,
pm coil pm d coil d q coil q
N L N L N
(9)And the phase resistance can be evaluated by:
2
5 5sin8 8234sin
4
s sss
N N s coil so N N ph cu p s
N N R l Rqk A
(10)The waveforms of phase backEMF and the cogging torqueversus rotor mechanical position are shown in Figure 9.
25155515250 4.09 8.18 12.27 16.36
Rotor Position (Degree)
B a c k  E M F ( V )
31.501.53
C o g g i n g T o r q u e ( N . m )
backEMF Cogging Torque
Figure 9: Phase backEMF and cogging torque waveforms
.VI. M
ACHINE
L
OSSES
A
NALYSIS
International Conference on Power Electronics Systems andApplications (PESA 2009)
Machine losses are a complex function of speed and load. Ingeneral, electromagnetic losses dominate the total losses inlow speed machines. As such, only electromagnetic losseswill be analysed here. Electromagnetic losses can be brokendown into three distinct parts, copper loss in the windings,iron core losses in the stator and rotor laminations, and eddycurrent losses in the permanent magnets. The motor coilcopper loss can be computed from the estimated phaseresistance and the torquecurrent profile of the machinefrom the FEA static analysis. The estimated resistance of one phase of the machine at 100
o
C is 11.1 m
Ω
. Using thepredicted current densities required to achieve the requiredtorque, the copper losses for different load conditions aregiven in Table 2.Transient FEA can be used to calculate the core losses inelectrical steel laminations considering the harmonics. At agiven frequency, the core losses for electrical steel is:
2 1.52max max max
c h c e
P K B f K B f K B f
(11)where
K
h
,
K
c
, and
K
e
are respective coefficients forhysteresis losses, normal eddy current loss, and excess oranomalous eddy current loss, and
B
max
is the maximumamplitude of flux density. Figure 10 shows the sensitivity of the overall lamination core losses to the machine speed andload. If machine’s load and hence armature currentincreases, the peak flux density in the lamination will alsoincrease, leading to higher core losses. The peak load losseswill approximately double the core losses at no loadcondition. However, the core losses will not be significantcompared with copper loss or eddy current losses in the PM.The magnetic field in the PM will vary when the rotorrotates. Since the sintered NdFeB35 has conductivity of roughly 10% that of mild steel, it can generate considerableeddy current losses. Figure 11 shows that the eddy currentlosses in the PM of the proposed machine are significant,which are nearly two times the copper loss at ratedconditions. Reducing eddy current losses in the PM is notonly critical to ensuring the maximum working temperatureis not to be exceeded, but also because of the fact that theenergy product of the PM will reduce at elevatedtemperatures.From the electromagnetic loss data presented in thissection, it is found that the efficiency of the machine underrated load condition (50Nm) is just over 85%. The copperlosses in winding coils and eddy current losses in PM arefound to be the major components of electromagneticlosses. Furthermore, the copper losses can be reduced byimproving the machine winding packing factor. It is notedthat the eddy current losses in the PM will dominate if themachine is overload or the operating speed is increased.However, the eddy current losses can be significantlyreduced by segmenting the magnets into several sections.For example, it is estimated that the machine efficiencywill easily be improved to over 90% if each magnet issegmented into three parts.T
ABLE
2: C
OPPER
L
OSSES
Torque Phase Peak Current Loss25 N.m76 A 96 W50 N.m 152 A 383 W75 N.m228 A 862 W
0204060800 250 500 750 1000RotorSpeed (rpm)
C o r e L o s s e s ( W )
75N.m50N.m25N.m0N.m
Figure 10: Core losses in stator and rotor laminations
025050075010000 250 500 750 1000
Rotor Speed (rpm)
P M E d d y C u r r e n t L o s s e s ( W )
75N.m50N.m25N.m0N.m
Figure 11: Eddy current losses in permanent magnetsV. C
ONCLUSION
A novel outerrotor PMFS machine is proposed for anurban electric vehicle propulsion application. The machinecombines the advantages of the SR and PM motors,making it uniquely ideal as an inwheel directdrive motor.The machine sizing equations have been developed basedon fundamental 3phase machine theories, and arevalidated by FEA. A 5kW machine is designed based onthe analytical model. FEA is further employed to optimizethe machine performance and predict the electromagneticlosses. The machine efficiency found to be over 85%,which can be readily improved further by segmenting themagnets to reduce the eddy current losses. The predictedresults show that the proposed machine is a highly viablecandidate for inwheel directdrives.A
CKNOWLEDGMENT
The work was carried out under the collaborativeMemorandum of Understanding between CranfieldUniversity and Zhejiang University.