1
An Optimized Seven-Phase Outer-Rotor Surface Mounted
Permanent Magnet Synchronous Machine for In-Wheel EMotorcycle Application
Hamidreza Ghorbani1, Mohammadreza Moradian1,2*, Mohamed Benbouzid3
1Department
2Smart
of Electrical Engineering, Najafabad Branch, Islamic Azad University, Najafabad, 85141-43131, Iran
Microgrid Research Center, Najafabad Branch, Islamic Azad University, Najafabad 85141-43131, Iran
3University of Brest, UMR CNRS 6027 IRDL, 29238 Brest, France
Abstract
Four outer rotor surface-mounted permanent magnet synchronous machines (SMPSM), supplied by a
seven-phase drive system, are proposed in this study, considering different q (number of stator slot per
phase per pole ratio) to achieve a satisfying value of electromagnetic torque and Back-Electromotive Force
(Back-EMF) with lower torque pulsation. Accordingly, the proposed configurations are investigated and
results are comparatively reported. So that based on the results, the best-performed configuration, the
Candidate Model, which presents the lowest torque pulsation with a desirable value of Tavg and Back-EMF
is selected. In order to demonstrate the advantages of this candidate model, an optimization analysis is
performed using 2D Finite Element Analysis (FEA). The resultant values of the variables are applied,
designing three optimized models. Performance results of the optimized models demonstrate that TCog
reduced noticeably and TRipple declined below 5%. The Artemis Drive-Cycles analysis results are also
included for the best-optimized model, considering E-Motorcycle requirements and properties for Urban,
Rural, and Motorway driving conditions. Accordingly, in terms of In-Wheel application of the optimized
machine, high torque/power density along with high values of PF and efficient performance is provided for
E-Motorcycle application.
Keywords-Artemis Drive-Cycle, Cogging Torque, E-Motorcycle, Optimization, SPMSM.
Electronic copy available at: https://ssrn.com/abstract=3981313
2
1. Introduction
In terms of Electric Vehicle (EV) propulsion application, enormous studies have been made on
transportation electrification technologies, requirements, policies, and electric machine topologies [1-4].
Permanent Magnet Synchronous Machine (PMSM) is one of the major alternatives widely investigated for
vehicular application [5-8]. PMSM topologies mostly benefit from a laminated cylindrical slotted stator
with a 3-Phase winding, supplied by an inverter set-up, and utilizes a Permanent Magnet (PM) rotor. In
comparison with Induction Machine (IM), PMSM provides advantages as [4], [5], [9-12]: (i) higher power
density, (ii) higher efficiency, (iii) high power factor, (iv) high ratio of output torque to weight and (v) more
reliability. PMSM rotor structure is mainly introduced as Interior/Surface Mounted PMSM (IPMSM and
SPMSM respectively), locating Inner or Outer side of the stator (IR or OR-SPMSM respectively) [13-14].
Due to features of the external location of the rotor and corresponding high inertia, OR-SPMSM
demonstrates constant speed operation rather than acceleration operation [4], [10], [15-16]. However,
SPMSM machines suffer from high values of torque ripple, particularly when operating at low speed and it
is an urgent issue to be tackled which is studied in the majority of literature considering topological design
optimizations [10], [14], [16]. Optimizing the dimension of PMs and slot entrance designs along with an
optimal selection of slot/pole combinations for OR-SPMSM have a high impact on the slotting effect and
air-gap flux density of the machine and hence effectively result in the cogging torque reduction. In
addition, with the advent of power electronic converters, the multiphase drive system with advantages such
as reduction of torque pulsation and providing higher torque pulsation frequencies can apply to electric
motor drive systems [17-24]. [3] studied the behaviour of the EV drivers and their impact on electricity and
transport electrification system. About 1000 EV agents have been investigated while being unaware/aware
of the test. Consequently, it has been found that their behaviour directly and indirectly affects both the road
transport system and the electricity grid, traffic of the roads, the distribution network’s stress and the usage
of the charging related infrastructure. [8] focuses on design considerations of an OR-SPMSM with
fractional concentrated winding configuration for In-Wheel applications, employing particle swarm
Electronic copy available at: https://ssrn.com/abstract=3981313
3
optimization (PSO). Eight rotor pole numbers for a 48-slot stator are considered in this study. According to
the applied PSO and 2D FEA results, the 48Slot/44Pole OR-SPMSM with about 10Nm of torque ripple and
0.5Nm of cogging torque provided the best performance in terms of EV application. A territory PSO
(TPSO) is proposed in [10] for design optimization of a SPMSM for unmanned aerial vehicle application
(UAV). Accordingly, the torque performance of the proposed SPMSM model is reduced, whereas it has to
be more precisely investigated because the model was applied to low power SPMSM (below 0.5kW) and
the optimum model still suffers from about 3% torque ripple for only 2.6Nm of average torque. A
numerical-based comparative analysis for OR brushless dc machine (BLDC) and SPMSM is studied in
[13], as a low voltage OR-BLDC and -SPMSM analyzed using 3D FEA. The results demonstrated that the
proposed OR-SPMSM presents lower value of torque fluctuations due to a reduction of stack length (torque
ripple reduced by 78%) in comparison with the proposed OR-BLDC. In [14] design and optimization of an
OR-SPMSM with reverse step shape PMs is studied. Based on theoretical results the air-gap flux
harmonics decreased in comparison with the conventional SPMSM. Although, torque ripple value is
reduced for the proposed OR-SPMSM, the optimized models faced drop in torque density. [21] and [23]
address the design of a 7-P SPMSM with tooth-concentrated winding using maximum torque per ampere
(MTPA) control strategy. [21] focuses on the design of two radially magnetized PM to reduce torque
ripple. In addition, by injecting the fifth harmonic current, the produced torque can develop without
increasing torque fluctuations. While [23] investigates the simplicity and fault-tolerant ability of a 7-Phase
SPMSM. According to [23], an appropriate segmentation of the PM layer is proposed, resulting in smooth
torque production even under strong distortion in stator current.
Therefore, to study tackling the high values of torque ripple in a SPMSM, this paper aims to investigate
an OR-SPMSM supplied by a 7-Phase drive system with four winding configurations, particularly for low
torque ripple In-Wheel E-Motorcycle application. Results are comparatively reported and the bestperformed configuration, providing a desirable amount of torque density with low fluctuations, is selected
to be precisely studied. Hence, to aim for this goal and for clearing the strengths of the machine, a 2D FEA
Electronic copy available at: https://ssrn.com/abstract=3981313
4
is applied considering the best-performed winding pattern. The final optimized models, meeting the
objective function constraints, are investigated and results have been compared. Finally, the model
presenting the best performance is analyzed under the standard Artemis Drive Cycle test and results
reported for Urban, Road, and Motorway operation.
2. System Equations Review
In the following sections, summarized design formulation and performance equations for multi-phase PM
machines and proposed models are presented.
2.1. Multiphase PM-Machine equations
Generally, for any multi-phase electrical machine, power, Pe, and torque, Te, can be expressed as [25]:
Pe 
Te 
m T
e  t i  t  dt
T 0
Pe m

 T
(1)
T
 e  t i  t  dt
(2)
0
where Ω is the rotational angular speed, m is the number of phases, T is the cycle period of Back-EMF, e(t)
and i(t) are one phase Back-EMF and current, respectively. To enhance torque density in the m-phase
system, current harmonic orders less than m can be injected into the system [17], [25]. Hence, assuming the
same harmonic orders for Back-EMF [25]:
m2
e  t    E Sin t 
(3)

m2
i  t    I Sin t 
(4)
 1
where Eυ and Iκ are υth and κth harmonic peak values of Back-EMF and current, respectively, and ω is the
electrical
m2
Te 
angular
frequency.
Accordingly,
the
torque
equation
can
be
 2 E I
m
rewritten
as
[25]:
(5)
 
 1
Electronic copy available at: https://ssrn.com/abstract=3981313
5
and Eκ, the peak value of Back-EMF, is given by:
(6)
E  K e B N t Dg L
where κth harmonic of winding factor and air-gap flux density are Keκ and Bκ, respectively. Nt is the number
of turns per phase and Dg and L are the air-gap diameter and stack length of the machine, respectively. In
order to achieve torque enhancements, the RMS value of phase current (including peak values of current
harmonics) can be given as [25]:
I rms
m2
 K
I1

2 K e B1
e
B 
2
(7)
 1
where I1 is the peak value of 1st current harmonic, Ke1 and B1 are the corresponding winding factor and airgap flux density harmonics, respectively. Inserting electrical loading, A, as [25]:
A  2mN t
I rms
 Dg
(8)
into (7), one can get:
 ADg K e1 B1
I1 
m2
 K
2mN t
e
(9)
B 
2
 1
and
I 
K e B
I1
K e1 B1
(10)
Thus, the electromagnetic torque can be expressed as [25]:
Te 
 ADg2 L
2 2
m2

 1
 K e B 
m2
 K
e
2
B 
(11)
2
 1
Calculating Do, outer diameter, by
o 
Dg
(12)
Do
Torque equation based on sizing equation, Do2L, for a multiphase PM machine can be expressed as Eq.
Electronic copy available at: https://ssrn.com/abstract=3981313
6
(13):
Te 
 Ao Do2 L
2 2
 K e B 
m2

 1
m2
 K
e
2
B 
(13)
2
 1
2.2. Winding types
Considering the value of q, the number of stator slots per rotor pole per phase, two main types of winding
can be introduced [26-27]:
(i) q is an integer, Distributed-Winding type: Conventionally used in electric machines, where greater
value of q results in more sinusoidal Magneto-motive Force (MMF) wave produced.
(ii) q is not an integer, Concentrated-Winding type: This type is mostly wounded in single- or doublelayer, which the comparison is given in Table I.
It can be found from Eq. 13. that the variation in air-gap flux density and the value of m has a high
impact on the torque performance of the machine. In addition, based on the q ratio as:
q
Ns
mN p
(14)
the values of Ns, Np, and m have a direct effect on the winding pattern and corresponding winding factor
and MMF. Overall, Bg influenced by the ratio of slot/pole combination along with the number of phases
and the value of PMs flux linkage in the rotor and therefore, it directly affects the electromagnetic torque of
the PM machine.
3. Proposed Model
In this study, based on the former introduction in Section II., four winding patterns and corresponding Ns
for a six-pole seven-phase outer-rotor SPM machine (Fig. 1.) are investigated for In-wheel application in an
E-Motorcycle. The proposed models are designed and analyzed under the same condition, assuming 40˚C
of ambient temperature.
Electronic copy available at: https://ssrn.com/abstract=3981313
7
3.1. Configuration and Design Parameters
Four Ns/Np ratios, namely 42/6, 28/6, 21/6, and 14/6, are considered in this study, resulting in q equal to
1, ⅔, ½, and ⅓, respectively. The initial values of design parameters are given in Table II. The winding
patterns and the objective function of this study are as follow:
3.2. Winding Pattern
According to the value of q in Table. II., four concentrated winding patterns and selected Ns values as
shown in Fig. 2 to Fig. 5 are considered. The corresponding winding factor and winding harmonics of each
q are presented in Fig. 6. and Fig. 7., respectively.
Fig. 1. 3D-View of an OR-SPMSM
Fig. 2. Winding pattern for NP=6, NS=42 (q=1)
Electronic copy available at: https://ssrn.com/abstract=3981313
8
Fig. 3. Winding pattern for NP=6, NS=28 (q=⅔)
Fig. 4. Winding pattern for NP=6, NS=21 (q=½)
Winding Factor
Fig. 5. Winding pattern for NP=6, NS=14 (q=⅓)
q=1
1
q=2/3
q=1/2
q=1/3
0.5
0
1
5
9
13 17 21 25 29 33 37 41 45 49
Harmonic Order
Fig. 6. Winding factor
Electronic copy available at: https://ssrn.com/abstract=3981313
9
MMF (Amp. Turns)
20
q=1
q=2/3
q=1/2
q=1/3
15
10
5
0
1
5
9
13 17 21 25 29 33 37 41 45 49
Harmonic Order
Fig. 7. Winding harmonics
3.3. Variables and Objective Function
The objective function of this study is to enhance torque performance of the machine for electric vehicle
application. Accordingly, following development processes are considered to be investigated using 2D
Finite Element Analysis (FEA). The initial design parameters are presented in Table II. and the variables
considered for sensitivity analysis of the candidate model are illustrated in Fig. 8.
 Initial Configurations
Objective Function:
Minimum Torque Pulsation, Satisfying Torque Density and Back-EMF: Determination through
q selection
Constraints:
Initial Design Parameters in Table II.
 Candidate Model
Objective Function:
Minimize Torque Pulsation and Maximize Torque: Determination for the best-performed model
through FE sensitivity analysis
Constraints:
Electronic copy available at: https://ssrn.com/abstract=3981313
10
WPM=4mm, LSSD=25mm, WST=5mm, TCog≤1Nm, TRipple<5%, Tavg≥80Nm
PM Dimension and slot entrance design variables:
0mm≤X1≤9mm, 0mm≤X2≤5mm, 5˚≤X3≤55˚
Fig. 8. Variables of sensitivity analysis
4. E-Magnetic Results
By applying 2D FEA to the proposed structures, the following results have been achieved, so that in
Section IV.A., initial models are investigated in order to select the candidate model, which provides a
satisfying and desirable amount of torque and Back-EMF amplitude. Then the resultant candidate model is
sensitively analyzed with constraints mentioned for X1, X2, and X3, and results illustrated in Section IV.B.
Optimized models, providing highest Tavg, lowest TCog and TRipple are introduced in Section IV.C., where
comparative results reported. In Section IV.D. efficiency and Power Factor map of the best-optimized
model are reported. Finally, Section IV.E. held results of Drive-Cycle Torque graph for In-Wheel
application of an E-Motorcycle adopting the best-performed model.
4.1. Analysis of the Initial Models
Fig. 9. to Fig. 12. demonstrate the analysis results of the average torque, cogging torque, torque ripple,
and back-EMF for the initial configurations, respectively.
Electronic copy available at: https://ssrn.com/abstract=3981313
11
As shown in Fig. 9. (a), the greater the value of q, the more average torque is produced by the proposed
7-P SPMSM. So that, according to torque FFT analysis shown in Fig. 9. (b), the amplitude of the
fundamental harmonic is increasing as q grew.
Based on TCog results, illustrated in Fig. 10., the cogging torque doesn’t follow the same path the average
torque did, thus, the growing value of q does not directly affect TCog. This means, for q=⅓ and ⅔, the peakto-peak cogging torque is approximately 50% less than q=1 and ½, respectively, where q=⅔ presents the
lowest TCog (less than 2Nm).
On the contrary, TRipple reduced for q≤⅔, reaching from 13% in q=⅓ to 7% in q=⅔, however, it is
suddenly doubled in q=1 (Fig. 11.).
The one-phase back-EMF and the corresponding FFT analysis results of the proposed models are
presented in Fig. 12. (a) and (b), respectively. Accordingly, the peak value of the phase back-EMF
increased along with the increasing value of q, from 50V to 70V for q=⅓ to q=1, respectively. The
amplitude of harmonics for q=⅔ continuously decreased to about zero for n≥7th harmonic, whereas other q
values suffer from n>7 harmonic orders. Thus, the phase back-EMF, produced by the q=⅔ model, presents
fewer fluctuations with desirable peak values.
Based on results discussed and demonstrated in Fig. 9. to Fig. 12., the proposed 28/6 initial model
with q=⅔ provides a satisfying average torque value of 95Nm, TCog less than 2Nm, 7% of TRipple and 67V of
phase back-EMF, is selected as the candidate model. Hence, in order to achieve the goal of this study,
reduction of torque pulsation, a sensitivity analysis is performed considering the variables shown in Fig. 8.
The flux density distribution of the initial models is illustrated in Appendix Section.
Electronic copy available at: https://ssrn.com/abstract=3981313
12
120
Torque (Nm)
90
60
q=1
q=2/3
q=1/2
q=1/3
30
0
0
60
120
180
240
300
360
Position (Elec. Deg.)
(a)
Amplitude (Nm)
120
q=1
q=2/3
q=1/2
q=1/3
9
11
13
15
90
60
30
0
1
3
5
7
Harmonic Order
(b)
Cogging Torque (Nm)
Fig. 9. Initial models: (a) Average torque, (b) Torque FFT analysis
q=1
q=2/3
q=1/2
q=1/3
10
5
0
-5
-10
0
60
120
Position (Elec. Deg.)
Fig. 10. Cogging torque of the initial models
Electronic copy available at: https://ssrn.com/abstract=3981313
180
13
Torque Ripple (%)
15
10
5
0
q=1
q=2/3
q=1/2
q=1/3
Number of Stator Slots/Phase/Pole
Phase Back-EMF (v)
Fig. 11. Torque ripple of the initial models
100
q=1
q=2/3
q=1/2
q=1/3
50
0
-50
-100
0
60
120
180
240
300
360
Position (Elec. Deg.)
(a)
Amplitude (v)
80
q=1
q=2/3
q=1/2
q=1/3
9
11
13
15
60
40
20
0
1
3
5
7
Harmonic Order
(b)
Fig. 12. Initial models: (a) Phase Back-EMF, (b) Back-EMF FFT
Electronic copy available at: https://ssrn.com/abstract=3981313
14
4.2. Sensitivity Analysis of the Candidate Model
As depicted in Fig. 8., the analysis variables, namely X1, X2, and X3, are sensitively analyzed for the
candidate model from Section IV.A., considering the constraints in the objective function. Torque, cogging
torque, and torque ripple results of this sensitivity analysis are illustrated in Fig. 13. to Fig. 15. As depicted
in Fig. 13. decreasing value of X1 and X2 along with an increase in X3 value increase Tavg, so that, Tavg-Max
produced equals 96.2Nm for X1=2.5mm, X2=3.5mm, and X3=45˚. TCog results are illustrated in Fig. 14.,
where X1 and X2 should move in opposite directions in order to provide lower cogging torque and for X3,
TCog fluctuates in the way that the lowest TCog produced is 0.018Nm as X1=0mm, X2=5mm and X3=5˚.
Lower TRipple is presented by increasing values of X2 and X3, where X1 causes fluctuations in torque ripple.
However, TRipple-Min is provided by values of X1=0mm, X2=4.5mm and X3=45˚, which equals to 1.7%. Based
on the results shown in Fig. 13. to Fig. 15. and resultant values for X1 to X3, the best performed models are
categorized, which provide maximum average torque (Model-I), minimum cogging torque (Model-II), and
minimum torque ripple (Model-III), respectively. Corresponding results for the optimized models are
reported in Table III.
100
X3(˚)
60
80
40
60
20
0
6
40
4
X2(mm)
2
0
0
2
4
6
8
X1(mm)
Fig. 13. Tavg (Nm) vs X1 vs X2 vs X3
Electronic copy available at: https://ssrn.com/abstract=3981313
10 20
0
15
6
X3(˚)
60
4
40
20
0
6
2
4
2
X2(mm)
0
0
2
4
6
8
10
0
X1(mm)
Fig. 14. TCog (Nm) vs X1 vs X2 vs X3
20
X3(˚)
60
15
40
20
0
6
10
4
2
X2(mm)
0
0
2
4
6
8
X1(mm)
10
5
0
Fig. 15. TRipple (%) vs X1 vs X2 vs X3
4.3. Analysis of the Optimized Models
With regard to the goal of this paper, a comparative study is done for the models designed by means of X
values in Table III. and the results are presented in Fig. 16. to Fig. 19.
It can be obtained from Fig. 16. (a) and (b) that the Model-II presents poor torque density, however, it
produced TCog, close to zero which is desirable. The Model-I provides the highest electromagnetic torque,
whereas, it suffers from high fluctuations in comparison with the Model-III, visiting the objective function
constraints (Tavg≥80Nm) with low pulsations.
Accordingly, TCog and TRipple are depicted in Fig. 17. and Fig. 18. respectively, where the Model-III
produced significantly low pulsations in comparison with other models and the initial q=⅔ configuration,
Electronic copy available at: https://ssrn.com/abstract=3981313
16
as desired in the objective function (TCog≤1Nm and TRipple≤5%). While Model-I and –II present the highest
TCog and TRipple, respectively.
Fig. 19. (a) and (b) demonstrate the phase back-EMF and corresponding FFT analysis of the proposed
models in Table III. It can be determined that in comparison with the candidate model, the amplitude of
phase back-EMF harmonics in the Model-I to -III faced a noticeable reduction. Although, the phase backEMF amplitude is low in the Model-II and -I, containing higher harmonics amplitude and even suffers from
9th harmonic order, the Model-III produces an acceptable peak-to-peak phase back-EMF value of 67.2V
with lower amplitude harmonic orders, damped for n≥5th harmonic.
Thus, the proposed Model-III provides the best performance and meets the established constraints,
producing 85.1Nm of torque with low pulsation, as TCog=0.26Nm and TRipple=1.7%. Flux density
distribution of the optimized models is shown in Appendix Section.
Torque (Nm)
120
90
60
Model-I
Model-II
Model-III
30
0
0
60
120
180
240
300
Position (Elec. Deg.)
(a)
Electronic copy available at: https://ssrn.com/abstract=3981313
360
17
Amplitude (Nm)
100
Model-I
Model-II
Model-III
50
0
1
3
5
7
9
11
13
15
Harmonic Order
(b)
Cogging Torque (Nm)
Fig. 16. Optimized models:(a) Average torque, (b) Torque FFT
5
2.5
0
Model-I
Model-II
Model-III
-2.5
-5
0
60
120
180
Position (Elec. Deg.)
Fig. 17. Cogging torque of the optimized models
Torque Ripple (%)
10
7.5
5
2.5
0
Model-I
Model-II
Model-III
Optimized Models
Fig. 18. Torque ripple of the optimized models
Electronic copy available at: https://ssrn.com/abstract=3981313
18
Amplitude (Nm)
100
Model-I
Model-II
Model-III
50
0
1
3
5
7
9
11
13
15
Harmonic Order
Amplitude (v)
(a)
Model-I
60
Model-II
Model-III
40
20
0
1
3
5
7
9
11
13
15
Harmonic Order
(b)
Fig. 19. Optimized models: (a) Phase Back-EMF, (b) Back-EMF FFT
4.4. Efficiency and Power Factor Map
As shown in Fig. 20., the efficiency map of the Model-III, considering Motoring (M) and Generating (G)
modes of operation, constant torque region is provided up to 1500rpm of speed with 85Nm of torque.
Where in generating mode, a wider region of high efficiency is determined. Hence, constant torque above
80Nm with an efficiency value of 80% is achievable when speed≥900rpm in M operation and
speed≥1100rpm in G mode. However, efficiency≥90% is provided for 500≤speed≤2200 and
500≤speed≤2500rpm in M and G mode respectively, producing at most 60Nm of average torque.
Electronic copy available at: https://ssrn.com/abstract=3981313
19
Based on the Model-III power factor (PF) map illustrated in Fig. 21., a higher power factor (above 0.7) is
determined for P≥5kW. As shown in Fig. 21., for both M and G modes of operation, on the contrary of
constant torque region, PF is increasing along with an increase in P for constant power region, wherein
constant torque region it declines from 0.99 to 0.7 and 0.5 in M and G mode, respectively, while in constant
power region it reaches from 0.15 to above 0.8 for both M and G operational modes.
Fig. 22. demonstrates the corresponding phase advance (PA) analysis results for the M mode of the
Model-III. Accordingly, for 0˚≤PA≤35˚ constant torque region is provided, where speed≤1500rpm and
TAvg≤85Nm. The constant power region is achieved for 35˚≤PA≤90˚ for speed values above 1500rpm
providing average torque below 80Nm.
Overall, based on the results reported, not only the 3rd Model designed based on the optimization
processes, has met the constraints mentioned in the objective function, it also presents desirable Torque,
Power, PF, and Efficiency performance, made Model-III the best-performed model being adapted for an EMotorcycle application.
25
30
35
40
545055
60
65
70
80
15
20
25
30
35
40
45
50
55
760
605
85
75
90
8580
-90
85
75 70650
6 550
545
2250
15
5
-60
10
-30
80
85
90
90
90
80 75
70
8090
85
80
85
8570
75
90
80 75
70
8090
85
80
85
90
85
85
90
80
8570 80 75
70 60
65
85
80 757055504540
6565
3530
25
20
15
65
25
20
15
555
10
2015 60105 7055504540
555
25
10
30
35
2015 105
40
25
45
30
50
35
40
7570 60
60
45
50
60
8085 3530
75
65
75
80
90
85
90
90
90
85
85
1000
2000
60
40
20
85
90
303540
0
100
85
90
75
30
0
75
10
60
5
Torque (Nm)
90
3000
Speed (rpm)
Fig. 20. Efficiency (%) map of the Model-III
Electronic copy available at: https://ssrn.com/abstract=3981313
0
20
15
1
Power (kW)
10
0.8
5
0.85
0.9
0
000.0.30.30.4.40.505.0.70.5805.95
5556.7
-5
0.9
0.9
0.95
0.95
0.8 0.85
0.6 0.75
5 0.7
0.9
0.9
0.95
0.95
0.95
0.9
0.9 0.8
5
0.8
5
0.7
0.7
5
0.6
0.6
0.55
5
0.50.4
.8.8 .70.
565
0.45
60.4
55 0.35
0.3 0.35
0.5
0.25
0.50.4
0.2
0.9 0000.70.
0.25
0.3
0.15
0.2
0.15
0.1
0.05
0.05
0.1
0.15
0.05
0.2
0.25
0.1 0.3
0.35
0.15
0.2
0.4
0.25
0.4
0.5
5
0.5
0.35
50.3
0.
0.
6
65
0.4
0.
0
0.45
7
.7
0.5
5
0.8
0.55
00.8
0.6
5
0.6
5
.9
0.7
0.7
0.85
0.95
0.8
0.95
0.95
0.95
0.95
0.9
0
0.8 .85
0.7
5
0.9
0.95
1000
0.95
0.8 0.85
0.9
0.8 0.85
0
0.9
0.95
0.95
5
0.9
-10
-15
5
0.9
0.8
0.7
0.6
0.9
2000
0.8
3000
0.5
Speed (rpm)
Fig. 21. Power factor map of the Model-III
90
35
90
35
55
35
60
25
60
55
50
70
20
70
30
20
25
15
15
10
10
525
525
525
535
535
10
10
10
1565 3090
1565 3090
1565 3090
20
20
20
40
40
40
45
45
45
50
50
50
55
55
55
60
60
60
75 7035
75 70
75 70
80
80
80
85
85
85
0
50
45
40
15
30
65
35
20
60
65
25
45
40
30
0
60
30
25
Torque (Nm)
50
30
45
40
30
1000
560
580
65
75
85
90
70
75
80
2000
85
90
3000
0
Speed (rpm)
Fig. 22. Phase Adv. (°) map of the Model-III
4.5. Artemis Drive-Cycle Analysis
Measuring vehicle speed over time as data series in form of Drive Cycles (DCs) are created theoretically
or based on Real-World Driving records. Due to developments of EV industries, DCs are mainly utilized in
order to test fuel consumption and CO2 emissions of a vehicle and recently being used to size electrical
powertrain components. [9].
The sizing of the powertrain is heavily dependent on vehicular and conditional parameters such as (Fig.
23.) [9]:
Electronic copy available at: https://ssrn.com/abstract=3981313
21
- Air Resistance, which is defined as Fair in Fig. 23, representing the main force to overcome when
driving on a flat road. Fair is proportional to air density (ρ), the square form of vehicle’s velocity
(υ2), a drag coefficient (Dd) and frontal cross-sectional area of the vehicle (A):
Fair 
1
 ACd 2
2
(15)
- Rolling Resistance, defined as Froll, occurs due to interaction between the surface of tires and road
surface. Froll is proportional to vehicle mass (m), gravity acceleration (g), road gradient (α), and a
rolling coefficient (Cr):
Froll  Cr mg cos 
(16)
- Grading Force, Fgrade, is the force required for up/downhill movement. Fgrade can be either be
Positive or Negative, thus it can be considered as resistive or supportive force and it can be
determined by multiplication of m, g, and sin(α):
Fgrade  mg sin 
(17)
- Acceleration Force, FAcc is a major parameter in electrical machine sizing and has a significant
contribution in total traction force, calculating by-product of m and derived velocity over time
(dυ/dt).
FAcc  m
d
dt
(18)
- Traction Force, Ftrac, is the summation of the above forces in which the desired torque of the
machine has a direct relation with the total traction force.
Ftrac  FAcc  Froll  Fgrade  Fair
(19)
A complete drive cycle study based on the Volvo XC90 and S60 cars can be found in [9] using a RealWorld database and dynamic model of the proposed cars.
Electronic copy available at: https://ssrn.com/abstract=3981313
22
Assessment and Reliability of Transport Emission Models and Inventory Systems (Artemis) is a
European standard project based on a large database of European Real-World driving patterns [28]. In this
study, standardized Artemis DC is utilized consisting of three driving conditions:
- Urban Cycle: 991 Sec. of operation with an average speed of 17.7 km/h, displacement value of 4.9
km, and Max. speed of 57.3 km/h
- Rural Road Cycle: 17.3 km rural road trip in 1080 Sec. with an average speed equal to 57.5 km/h
reaching its Max. at 111.1 km/h.
- Motor Way130 Cycle: 96.9 km/h average speed (Max. 131.4 km/h) in 1066 Sec. for 28.75 km
motorway trip.
The corresponding drive-cycle speed is illustrated in Fig. 24. (a). For an E-Motorcycle concept design
shown in Fig. 23. with properties listed in Table IV., by applying the Artemis DCs to the Model-III of this
study, DC torque and DC operational data points are demonstrated in Fig. 24 (b) and (c). Table V. reports
the drive cycle analysis results of each Artemis cycle applied to the Model-III.
Overall, according to Table V., for E-Motorcycle application the studied Model-III operates efficiently
with high values of torque/power density and power factor in different conditions of driving.
Fig. 23. Dynamic traction forces of E-Motorcycle (Honda concept)
Electronic copy available at: https://ssrn.com/abstract=3981313
Drive Cycle Speed (rpm)
23
1500
Urban
Road
MotorWay
1200
900
600
300
0
0
200
400
600
800
1000
800
1000
Time (s)
Drive Cycle Torque (Nm)
(a)
Urban
Road
MotorWay
100
0
-100
0
200
400
600
Time (s)
(b)
120
Torque (Nm)
80
40
0
-40
Urban
Road
MotorWay
-80
-120
0
500
1000
1500
Speed (rpm)
(c)
Fig. 24. The Artemis DC of the Model-III: (a) Speed, (b) Torque, and (c) Operational data points under
Artemis DCs
Electronic copy available at: https://ssrn.com/abstract=3981313
24
5. Conclusion
The goal of this paper is to investigate Outer-Rotor Surface Mounted Permanent Magnet Synchronous
Machine (OR-SPMSM) for E-Motorcycle application providing low torque fluctuations and high torquepower density. Four winding patterns (q=1, ⅔, ½, ⅓) are proposed and applied to the OR-SPMSM. Results
are comparably reported and demonstrated that OR-SPMSM with q = ⅔ presents the best performance in
terms of low torque pulsation. In order to investigate the advantages of this configuration and reaching
enhanced design, a parameter optimization process is considered with three design parameters involved,
resulting in three optimized models, producing Tavg-Max (Model-I), Tcog-Min (Model-II), and Tripple-Min (ModelIII). The results referencing each model are reported where the Model-III provides the best performance
meeting the objective function constraints with low torque fluctuations. The Artemis Drive Cycles are
applied to the Model-III in order to simulate the performance of the proposed machine for Urban, Road,
and Motorway applications in an E-Motorcycle. Accordingly, due to high torque density with low torque
pulsations and efficient operation under high values of power factor, the Model-III presents the best
performance and can be considered as a vital alternative for In-Wheel Light-Weight EV applications.
Author contributions
H.R. Ghorbani, M.R. Moradian: Conceptualization, Methodology. H.R. Ghorbani: Data curation,
Writing- Original draft preparation. H.R. Ghorbani, M.R. Moradian, M. Benbouzid: Visualization,
Investigation. M.R. Moradian, M. Benbouzid: Supervision. H.R. Ghorbani: Software. Validation. M.R.
Moradian, M. Benbouzid: Writing- Reviewing and Editing.
References
[1] M. Zeraoulia, M. Benbouzid, D. Diallo, 2006. Electric Motor Drive Selection Issues for HEV
Propulsion Systems: A Comparative Study. IEEE Trans. Vehicular Technology, Vol. 55, No. 6., pp.
1756-1764, DOI: 10.1109/TVT.2006.878719
Electronic copy available at: https://ssrn.com/abstract=3981313
25
[2] G. De Nunzio, et al., 2021. A general constrained optimization framework for the eco-routing problem:
Comparison and analysis of solution strategies for hybrid electric vehicles. Transportation Research
Part C: Emerging Technologies, Vol. 123, DOI: 10.1016/j.trc.2020.102935
[3] Ch. Marmaras, et al., 2017. Simulation of electric vehicle driver behaviour in road transport and electric
power networks. Transportation Research Part C: Emerging Technologies, Vol. 80, DOI:
10.1016/j.trc.2017.05.004
[4] H.R. Ghorbani, M.R. Moradian, 2021. Torque Pulsation Reduction in five-phase PMSyncRMs. Journal
of Power Electronics, DOI: 10.1007/s43236-021-00329-2
[5] M.R. Moradian, et al., 2019. Adaptive Torque and Flux Control of Sensorless IPMSM Drive in the
Stator Flux Field Oriented Reference Frame. IEEE Trans. Ind. Informatics, Vol. 15, Issue: 1, DOI:
10.1109/TII.2018.2808521
[6] T. Gundogdu, et al., 2020. Analysis of coil pitch in induction machines for electric vehicle applications.
IET Elec. Power App., Vol. 14., Issue 12., p. 2525-2536, DOI: 10.1049/iet-epa.2019.0980
[7] M. Tumbek and S. Kesler, 2019. Design and implementation of a low power outer-rotor line-start
permanent magnet synchronous motor for ultra-light electric vehicles. Energies, 12(16), 3174, DOI:
10.3390/en12163174
[8] D. Wu, et al., 2014. Design considerations of outer-rotor permanent magnet synchronous machines for
in-wheel electric vehicle applications using particle swarm optimization. 7th Int. Conf. Power Elec.
Mach. Drives (PEMD), Manchester, UK, DOI: 10.1049/cp.2014.0449
[9] B. Raghuraman and T. Gyllensten, 2019. Electric machine design based on drive cycle analysis. M.Sc.
thesis, Chalmers University of Technology, Gothenburg, Sweden.
[10]J.-M. Ahn, J.-Ch. Son and D.-K. Lim, 2021. Optimal design of outer-rotor surface mounted permanent
magnet synchronous motor for cogging torque reduction using territory particle swarm optimization.
Journal Elec. Eng. Tech. 16, 429–436, DOI: 10.1007/s42835-020-00599-z
Electronic copy available at: https://ssrn.com/abstract=3981313
26
[11]S. Fang, et al., 2019. High power density PMSM with lightweight structure and high-performance soft
magnetic alloy core. IEEE Trans. App. Supercond., Vol. 29., Issue: 2, March, DOI:
10.1109/TASC.2019.2891630
[12]H.R. Ghorbani, B. Majidi, 2019. Power Density Optimization Through Optimal Selection of PM
properties in a PM-SyncRM Using FEM Analysis. 10th Ann. Int. Power elec. Drive Sys. and Tech.
Conf. (PEDSTC), Shiraz University, Shiraz, Fars, Iran, DOI: 10.1109/PEDSTC.2019.8697258
[13]T.-Y. Lee, et al. 2016. Motor design and characteristics comparison of outer-rotor type BLDC motor
and BLAC motor based on numerical analysis. IEEE Trans. App. Supercond., Vol. 26., Issue: 4, DOI:
10.1109/TASC.2016.2548079
[14]J. Zhao, et al., 2019. Design and optimization of an outer-rotor SPMSM with permanent magnets of
revers
step
shape.
22nd
Int.
Conf.
Elec.
Mach.
(ICEM),
Harbin,
China,
DOI:
10.1109/ICEMS.2019.8922253
[15]Ch. He and T. Wu, 2019, Analysis and design of surface permanent magnet synchronous motor and
generator. CES Trans. Elec. Mach. Sys., Vol. 3, Issue: 1, DOI: 10.30941/CESTEMS.2019.00013
[16]S. S. R. Bonthu, et al., 2018. Optimal torque ripple reduction technique for outer rotor permanent
magnet synchronous reluctance motors. IEEE Trans. Energy Conv., Vol. 33., Issue: 3, DOI:
10.1109/TEC.2017.2781259
[17]S. S. R. Bonthu, S. Choi and J. Baek, 2016. Comparisons of three-phase and five-phase permanent
magnet assisted synchronous reluctance motors. IET Elec. Power App., Vol. 10., Issue 5, p. 347 – 355,
DOI:10.1049/iet-epa.2015.0268
[18]H. Toliyat, S. P. Waikar and T. A. Lipo, 1998. Analysis and simulation of five-phase synchronous
reluctance machines including third harmonic of airgap MMF. IEEE Trans. on Ind. App., Vol. 34.,
Issue: 2, p. 332 – 339, DOI:10.1109/28.663476
Electronic copy available at: https://ssrn.com/abstract=3981313
27
[19]H. Toliyat, L. Xu and T. A. Lipo, 1990. A five phase reluctance motor, with high specific torque. Conf.
Record
of
the
IEEE
Ind.
App.
Society
Annual
Meeting,
Seattle,
WA,
USA,
DOI:10.1109/IAS.1990.152188
[20]F. Barrero and J. Duran, 2016. Recent advances in the design, modelling and control of multiphase
machines-Part 1. IEEE Trans. Ind. Elec., Vol. 63., Issue: 1, DOI: 10.1109/TIE.2015.2447733
[21]F. Scuiller, 2018. Design of a 7-phase surface-mounted PM machine tooth-concentrated winding. XIII
Int. Conf. Elec. Mach. (ICEM), Alexandroupoli, Greece, DOI: 10.1109/ICELMACH.2018.8507205
[22]H. Zhang, et al., 2019. Torque optimization of a seven-phase Bi-harmonic PMSM in healthy and
degraded
mode.
22nd
Int.
Conf.
Elec.
Mach.
(ICEM),
Harbin,
China,
DOI:
10.1109/ICEMS.2019.8921839
[23]F. Scuiller, et al., 2020. Design of a Bi-harmonic 7-phase PM machine with tooth-concentrated
winding. IEEE Trans. Energy Conv., Vol. 35, Issue: 3, DOI: 10.1109/TEC.2020.2982041
[24]W. Levi, 2008. Multiphase electric machines for variable-speed application. IEEE Trans. Ind. Elec.,
Vol. 55, No. 5, DOI: 10.1109/TIE.2008.918488
[25]J. Wang, R. Qu and L. Zhou, 2012. Dual-rotor multiphase permanent magnet machine with harmonic
injection to enhance torque density. IEEE Trans. Appl. Supercond., Vol. 22, No. 3, DOI:
10.1109/TASC.2011.2179399
[26]H. Jussila, et al., 2007. Guidelines for designing concentrated winding fractional slot permanent
magnet machines. Int. Conf. Power Eng. Energy Elec. Drives, Setubal, Portugal, DOI:
10.1109/POWERENG.2007.4380186
[27]D. Martinez, 2012. Design of a permanent-magnet synchronous machine with non-overlapping
concentrated windings for the Shell Eco Marathon urban prototype,” M.Sc. thesis, Royal Institute of
Technology (KTH), Stockholm, Sweden.
[28]Common
Artemis
Driving
Cycles
(CADC)-Emission
https://dieselnet.com/standards/cycles/artemis.php, Revision: Dec. 2018
Electronic copy available at: https://ssrn.com/abstract=3981313
Test
Cycles.
28
APPENDIX
A. Initial Models
(a)
(b)
(c)
(d)
Fig. 25. Flux density distribution in the initial models: (a) 42/6 and q=1, (b) 28/6 and q=⅔, (c) 21/6 and
q=½, (d) 14/6 and q=⅓
B. Optimized Models
(a)
(b)
(c)
Fig. 26. Flux density distribution in the optimized models: (a) Model-I, (b) Model-II, (c) Model-III
Electronic copy available at: https://ssrn.com/abstract=3981313
29
Table I. Concentrated Single- and Double-layer winding Properties [27]
Winding layers
Single Double
Fundamental winding factor (kw) high
low
End Windings
long
short
Eddy Current
high
low
Over-load Torque Capability
high
low
Back-EMF Harmonics
high
low
Torque Ripple (TRipple)
high
low
Electronic copy available at: https://ssrn.com/abstract=3981313
30
Table II. Initial Design Parameters
Quantity
Sym. Value
Unit
Num. Phases
m
7
---
Rated Voltage
Vn
120
V
Rated Current
In
70
A
Rated Speed
nr
1500
rpm
Motor Length
LM
100
mm
Stator Outer Diameter
DSO
135
mm
Stator Slot Depth
LSSD
25
mm
Rotor Outer Diameter
DRO
165
mm
Rotor Back Iron
DRBI
10
mm
Axle Diameter
DA
55
mm
Num. Stator Slots
Ns
42, 28, 21, 14 ---
Num. Rotor Poles
NP
6
---
Ratios of Ns/mNp
q
1, ⅔, ½, ⅓
---
Air-gap Length
Lg
1
mm
Stator Tooth Width
WST
5
mm
Stator Slot Opening
LSO
2
mm
Stator Tooth Tip Depth LTTD
1
mm
PM Length
LPM
50
˚
PM Thickness
WPM
4
mm
PM Type
---
N42SH
---
Steel Type
---
M400-50A
---
Electronic copy available at: https://ssrn.com/abstract=3981313
31
Table III. Output Data of the Optimized Models
Quantity
Model-I Model-II Model-III
X1 (mm)
2.5
0
0
X2 (mm)
3
5
4.5
X3 (˚)
45
5
45
TAvg (Nm)
96.2
4.73
85.1
TCog (Nm)
8.5
0.018
0.26
TRipple (%)
7.8
8.53
1.7
Effi. (%)
88.5
22.94
86.75
PF
0.9
0.18
0.82
8.2
67.2
B-EMF (V) 114.7
Electronic copy available at: https://ssrn.com/abstract=3981313
32
Table IV. Drive-Cycle Properties
Vehicle Properties
Value
Mass (kg)
80
Rolling Resistance Coefficient 0.005
Air-Density (kg/m3)
1.225
Frontal Area (m2)
1.5
Drag Coefficient
0.26
Wheel Radius (m)
0.3
Mass Correction Factor
1.04
Motoring (M) Torque Ratio
1
Generating (G) Torque Ratio
1
Electronic copy available at: https://ssrn.com/abstract=3981313
33
Table V. Artemis Drive-Cycle Data for the Model-III
Drive Cycle
Urban Rural
M. Way
Data
Value
Value Value
Unit
Time
991
1080
1066
s
Distance
4.9
17.3
28.75
km
Avg. Speed
17.7
57.5
96.9
km/h
Max. Speed
57.3
111.1
131.4
km/h
Avg. Efficiency
71.65
85.2
82.33
%
Input Energy
80.71
558.5
2291.3
Wh
Output Energy
13.53
11.4
7.54
Wh
M. Energy
60.19
478.3
1887.4
Wh
G. Energy
22.17
16.35
10.3
Wh
Total Loss
29.17
85.1
406.6
Wh
Copper Loss
24.48
66.4
369.9
Wh
Iron Loss
2.54
11.04
23.99
Wh
Magnet Loss
0
0
0.02
Wh
Mechanical Loss 2.15
7.64
12.7
Wh
M. Operation
77.8
91.2
94.5
%
G. Operation
22.2
8.8
5.53
%
Electronic copy available at: https://ssrn.com/abstract=3981313