See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/321106755

In-wheel motor design for an electric scooter

Article  in  Journal of Electrical Engineering and Technology · November 2017

DOI: 10.5370/JEET.2017.12.6.2307

CITATIONS READS

12 21,559

4 authors, including:

Byung-Chul Woo
Korea Electrotechnology Research Institute-KERI
102 PUBLICATIONS   1,339 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Byung-Chul Woo on 26 November 2018.

The user has requested enhancement of the downloaded file.

J Electr Eng Technol.2017; 12(6): 2307-2316 ISSN(Print) 1975-0102
http://doi.org/10.5370/JEET.2017.12.6.2307 ISSN(Online) 2093-7423

In-wheel Motor Design for an Electric Scooter

Ji-Young Lee†,**, Byung-Chul Woo*, Jong-Moo Kim* and Hong-Seok Oh***

Abstract – The aim of this paper is to provide an optimal design of in-wheel motor for an electric
scooter (E-scooter) considering economical production. The preliminary development in-wheel motor,
which has a direct-driven outer rotor type attached to the E-scooter’s rear wheel without any gear, is
introduced first. The objective of the optimal design of this in-wheel motor is to improve the output
characteristics of the motor and to have a stator form to facilitate automatic winding. Response surface
methodology was used for the optimal design and 2-dimensional finite element method was used for
electro-magnetic field analysis. Experimental results showed that the designed and fabricated in-wheel
motor could satisfy the required specifications in terms of speed, power, efficiency, and cogging
torque.

Keywords: Electric scooter (E-scooter), Finite element analysis, In-wheel motor, Response surface
methodology

1. Introduction winding. To use automatic winding machine, flat slot bottom

is basically necessary, but the preliminary development
Since the interest in using an electric motor drive system model’s slot bottom was semicircle shape. So, the company
as electric vehicles has been increasing, there have been required changes of stator core shape for automatic
lots of studies on them. Related products have begun to winding while maintaining the rotor specifications and
appear in the market. The in-wheel motor drive system that output characteristics. And the additional requirement is
places motors adjacent to the wheel or builds motors directly that the stator core should have minimum change to save
in the wheels can also be applied to various transportation production cost. To meet the company’s requirements,
system such as electric bicycle, electric powered wheelchair the preliminary development motor performances were
having two wheels, and electric cars having four wheels analytically and experimentally examined first. And then,
[1-19]. basis of the results the constraints of the motor design were
To survive in the market, products must meet the desired decided. The objective of the optimal design of this in-
performance on the market. From the standpoint of wheel motor is to improve the output characteristics of
propulsion electric motor design such as in-wheel motors, the motor and to have a stator form to facilitate automatic
cost saving must be taken into account, along with better winding. Response surface methodology (RSM) was
performances such as power density and efficiency. When used for the optimal design. Two-dimensional (2-D) finite
considering price factors, many efforts have been made to element analysis (FEA) method was used for the required
save rare earth permanent magnet (PM) while maintaining electro-magnetic field analysis during the design process.
basic properties [1, 2]. Researchers are working worldwide Experimental results showed that the designed and
to develop more efficient electrical drive system by motor fabricated in-wheel motor could satisfy the required
design to reduce its weight and losses in coils and magnetic specifications in terms of speed, power, efficiency, and
core materials [3-5]. These efforts are also related to cogging torque.
reduction in operating cost.
In this paper, brushless PM motor was used as in-wheel
motor for an electric scooter (E-scooter) having two wheels. 2. Preliminary Development In-wheel Motor
A commercial E-scooter of GMT [10] was considered for
this research. The armature winding of the preliminary To get the initial values and constraints for the optimal
development in-wheel motor in the E-scooter was made design, the preliminary development in-wheel motor and
manually, but for mass production, GMT wanted automatic its characteristics are described in this section.

† Corresponding Author: Electric Motor Research Center, Korea 2.1 System and motor configurations
Electrotechnology Research Institute, Korea.([email protected])
* Electric Motor Research Center, Korea Electrotechnology Research
Institute, Korea. It is usually called in-wheel motor when propulsion
** Dept. of Energy and Power Conversion Engineering, Korea motors are adjacent to or in the wheels. However,
University of Science and Technology, Korea. sometimes the names of rim-motor [4, 5] or hub-motor [8]
*** Star Group Industry Company (SGI), Korea.
Received: May 15, 2017; Accepted: August 16, 2017 are used depending on whether it is to be placed closer to

2307
Copyright The Korean Institute of Electrical Engineers
This is an Open-Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/
licenses/by-nc/3.0/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited.
 In-wheel Motor Design for an Electric Scooter

Fig. 4. The force acting on a scooter moving along a slop

Table 1. Specifications of the E-scooter

Fig. 1. Prefabricated structure of the in-wheel motor and a Contents Value Unit
wheel Dimensions Ls=1750, Ws=740, Hs=1275 mm
E-Scooter weight, Ms 105 kg
Max speed 60 km/h
Wheel size 12 inch
Battery Li-FePo4 (72V / 20Ah)
Controller Vector control (72V / 80A)
In-wheel motor Brushless PM motor

of mass Ms, proceeding at a velocity Vs, up a slop of angle

, as in Fig. 4, the total force propelling the scooter forward,
total tractive effort Ft, is the sum of all these forces [14-
18]:

Ft = Frr + Fad + Fhc + Fla

Frr = Cr Ms g cos
Fig. 2. Configurations of the preliminary development in- Fad = ( ACd Vs2) / 2 (1)
wheel motor Fhc = Ms g sin
Fla = Ms acc

where Frr is the rolling resistance force, Fad is the

aerodynamic drag, Fhc is the hill climbing force, and Fla is
the force required to give linear acceleration acc. Cr is the
Fig. 3. One period of repeated winding layout for 32-pole rolling resistance coefficient, and the typical values is
and 36-slot brushless PM motor 0.013 for car tires on concrete or asphalt as an example
[15]. The acceleration of gravity, air density, and frontal
one side of the rim or the hub of the wheel. The electric area of the scooter are g, , and A, respectively. Cd is the
motor of the E-scooter considered in this paper is in rear aerodynamic drag coefficient, and the typical value is 0.6-
wheel and placed near the rim as shown in Fig. 1. 0.7 for scooter [15, 18].
The electric motor can be called rim-motor because of The required axle torque, T, and angular speed, w, of the
its position. However, it is called in-wheel motor in this in-wheel motor are obtained from the equations related
paper. Fig. 2 and Fig. 3 show the configuration of the with radius of the tire, r as follows:
preliminary development in-wheel motor and its winding
layout, respectively. T=Ft r, and w=Vs /r (2)

2.2 System and motor specifications The output specifications of the in-wheel motor are
summarized in Table 2. They are considered as meeting the
The specifications of the E-scooter that can affect the in- requirements of the battery and the controller as well as the
wheel motor design are listed in Table 1. In the dimensions, own performances of the E-scooter. Although it is not
Ls, Ws, and Hs mean the whole length, width, and height of strictly in contrast with motors used for power train system
the E-scooter, respectively. When considering the scooter or wheel-steering in fancy cars, the cogging torque of this

2308 J Electr Eng Technol.2017; 12(6): 2307-2316

 Ji-Young Lee, Byung-Chul Woo, Jong-Moo Kim and Hong-Seok Oh

Table 2. Output specifications of the in-wheel motor Table 5. Temperature test results
Contents Value Unit Torque dt_r dt_d Tamb dTc
Rated speed 700 rpm (Nm) (minute) (minute) (oC) (oC)
Rated torque 27.3 Nm 30 125 130 10~30 5
Rated power 2 kW 40 68 130 10~30 5
Cogging torque Below 10 % 50 50 130 10~30 5
Max stacking length 45 mm 60 19 130 10~30 5
Cooling condition Natural cooling 70 14 130 10~30 5
80 9 125 10~30 5
90 6.3 120 10~30 5
Table 3. Specifications of materials and dimensions of in- 100 4.2 103 10~30 5
wheel motor 110 2.5 85 10~30 5
120 2 70 10~30 5
Contents Value Unit
130 1 50 10~30 5
Stator core material S18 (laminated silicon steel)
140 0.33 30 10~30 5
Rotor core material SM490 (solid steel)
PM material N38SH (Br=1.24T, Hcb=964 kA/m)
Coil material H-class copper
No. of poles and slots 32 / 36
Outer /Inner dia. of rotor 219 / 199 mm
Outer/ Inner dia. of stator 197.5/ 130 mm
Thickness of PM 3 mm
No. of wire per slot 8
No. of strands in a wire 18
Diameter of strand 0.63 mm

Table 4. Measured performance data Fig. 5. Temperature aspect for the duty type S2 (short-time
duty)
Contents Value Unit
Phase Resistance 26.31 mOhm
Phase Inductance 0.08 mH mances for the two load conditions, which are rated power
Phase EMF 33.23 V(rms) and max torque. These results satisfy the power conditions
Load condition Rated power Max torque of design specifications in Table 2. EMF stands for no-load
Speed 718 188 rpm electro-motive force.
Torque 30 130 Nm
Table 5 is the temperature test results under the condition
Efficiency 88.6 45.2 %
of duty type S2 (short-time duty), which is operation at
constant load for a stated time, less than that required to
in-wheel motor was restricted for riding comfort. The reach thermal equilibrium, followed by a rest and de-
cogging torque of % unit was calculated as the proportion energized period of sufficient duration to re-establish
of peak-to-peak cogging torque value in the rated torque. machine and heat exchanger temperature within ±2K of
The specifications of the materials and dimensions of the the final coolant temperature. The temperature graph
in-wheel motor are listed in Table 3. The outer diameter of aspect is the same as shown in Fig. 5. In Table 5 and Fig. 5,
rotor and the inner diameter of stator are decided by the the terms of dt_r, dt_d, Tamb, and dTc are the constant load
constraints of wheel size as shown in Fig. 1. And the time, the de-energized time, the ambient air temperature
stacking length of both stator and rotor cores is 45mm range, and the temperature difference, respectively. The
considering the maximum allowable condition. The winding dt_r was decided by the measured maximum temperature
layout is the same as shown in Fig. 3. The slot fill factor is of coil hot spot, Tc_max, which was 180 oC in this test.
49.3% as the ratio of the total conductor area versus the From these test results, it is clear that a possible driving for
total slot area. For the automatic winding, the specification two hours of continuous operation and one minute of short-
of wire or winding is required to change as well as core time maximum torque operation.
shape, because less number of strands are effective for the
automatic winding process. 2.4 Analytical assessment of the rotor core and PM

2.3 Experimental assessment of the motor perfor- When solid steel is used, it can save considerable
mance amount in expenses because the process to laminate with
several steel sheets is deleted. However, there is a side
The preliminary development motor was made in effect of reduction in output power or efficiency by
accordance with the specifications in Table 3, and the increasing the core losses. Therefore, laminated silicon
various performances were examined. steel should be used for stator core because there are many
Table 4 shows the electric parameters and its perfor- magnetic flux variation that can affect core losses. In the

http://www.jeet.or.kr 2309
 In-wheel Motor Design for an Electric Scooter

Table 6. Comparison of the performances depending on

rotor core material (at the rated speed and input
current = 32A)
Contents Values Unit
Rotor core S18 SM490
Torque 31.3 31.2 Nm
Torque ripple 4.83 4.83 %
Copper loss 77.9 77.9 W
Core loss 296.2 300.8 W
Efficiency 86.2 86.0 %

Fig. 8. Minimum value of radial-directional flux density in

PM depending on abnormal current values

that of laminated core model at 1.5%. There is a big

difference between the core loss data of the two material
samples in Fig. 7. In the analysis models depending on the
materials, however, the difference of core loss values,
Fig. 6. Measured B-H curves of the material S18 and which are calculated by the method in [20] and [21] using
SM490A the data in Fig.7, can be ignored. When calculating the
efficiency by considering core and copper losses, the
difference in efficiency is only 0.2%. Since there is no big
difference in characteristics between solid core and
laminated core as rotor core, solid core is continuing
considered as rotor core.
To confirm the possibility of the PM demagnetization, 2-
D FEA was performed depending on abnormal current
values and the results are shown in Fig. 8. The assumptions
are as follows: (1) current phase is balanced, (2) the
analysis point is when phase-A current is maximum, and
(3) both the center position and the magnetic polarity of
magneto-motive-force by current are same as those by
PM. Considering the analysis results it is expected that
the PM would not be demagnetized because the minimum
value of the radial-directional flux density among the
elements in PM is over 0.36T even when the current is
Fig. 7. Measured core losses; (upper) SM490A-93mm2
150A - the limit current by controller is 80A for the
cross section area- solid steel ring sample, and
(lower) S18-0.5mm thickness- laminated steel sheet normal driving, and the peak limit current is 150A for the
sample momentary overload. There is no knee point in the second
quadrant B-H curve of SH-grade PM until the operating
temperature 100 oC.
rotor core, there is generally very little magnetic flux
change. To determine whether the continuing use of solid
steel in the rotor part, it was conducted that the comparison
of characteristics and losses depending on whether solid or 3. Design of Stator Core for Automatic Winding
laminated steel is used. Electro-magnetic field analysis was
performed using 2-D FEA for both cases. Fig. 6 and Fig. 7 On the basis of the analysis results about the E-scooter
show the used core material data. and its preliminary development motor, the rotor core and
The analysis results are listed in Table 6. Results of the PM would be maintained as an existing state because
analysis under the conditions of the same current input and solid core is used for considering economical production.
the same rated speed, the torque of laminated core model Furthermore the characteristics are proper as verified in
was slightly higher than the torque of solid core model at section 2. Therefore stator core would be only redesigned
0.2 % and the core loss of solid core model was higher than considering automatic winding.

2310 J Electr Eng Technol.2017; 12(6): 2307-2316

 Ji-Young Lee, Byung-Chul Woo, Jong-Moo Kim and Hong-Seok Oh

3.1 Constraint conditions core design. The basic theory has been explained in [26-30].
The detailed process is displayed in diagrammatic form as
In order to automate armature winding, the specifications shown in Fig. 9. The optimal design procedure in this paper
of restrictions on the automatic winding machine should is based on [29] and as follows:
be checked. There are many kinds of automatic winding 1) Selection of design variables and setting the range of
machines in the market, and typically there are three types the variables: The selected design variables and levels are
which are nozzle drive winding machine, flyer winding shown in Fig. 10 and Table 7. Four design variables were
machine, and hook type winding machine [22-25]. For this selected considering the constraint conditions as previously
paper’s in-wheel motor, a modified nozzle drive winding mentioned in chapter 3.1. Tw stands for tooth width, Hso
machine type is considered. The maximum diameter of a stands for height of tooth tip, Tht stands for tooth height,
strand of wire is 1.0 mm when the modified nozzle drive and Bso stands for slot open width. The design variables in
winding machine is used. And the minimum slot open zero level are values of the preliminary development motor.
width of the stator is also limited as 2.0mm because of the The values of design variables in + level and - level
needle of automatic winding machine to move around the determined by each possible maximum and minimum
entrance slot. The rest of constraints for stator core optimal values.
design is as follows: 2) Design of experiment and collection of samples by
1) Constraints of performances: basically the design FEA: The central composite design (CCD) was adopted as
specification in the Table 2 should be satisfied. In the experimental design method. The total number of
additionally the torque constant, which is the ratio of the experiment was 25 - cube points are 16 due to the full
torque to the current, and efficiency should be equal or factorial considering the 4 design variables and 2 levels (-1
higher than those of the preliminary development motor. and +1), center points are 5, and the axial points are 4 by
The cogging torque should be reduced as much as possible. CCD. After calculating torque constant, torque ripple,
2) Constraints of stator winding: the slot fill factor is current density, and efficiency as response variables for
less than 50%. The current density is less than 6A/mm2 in each experiment model, the results are summarized in
the rated torque. The number of wire per slot is maintained Table 8. The response variables are calculated with electro-
as an existing state because the rated speed and torque of the magnetic field analysis results obtained by 2-D FEA.
preliminary development motor satisfied the specifications The torque constant is calculated as the ratio of the
in Table 1 and Table 2, and the number of wire per slot is generated torque to the current, which was the input used
too small to change. If the number of wire per slot is to generate rated torque. The torque ripple is calculated as
change, there is a large voltage variation. the ratio of peak to peak range of the torque to the average
3) Constraints of stator core: for automatic winding it is torque instead of cogging torque calculation directly. If
recommended that the bottom of slot is flat and the sides both the rated torque and cogging torque are calculated, the
of tooth is parallel. The magnetic flux density on the analysis needs to be conducted twice (one under loading
maximum saturated spot in the core is less than 1.8T in the condition and the other under no loading), and the analysis
rated torque condition for fair comparison. will be repeated 50 times for 25 models. To reduce the time

3.2 Optimum design process Table 7. Design variables and levels

Level of design variables
RSM was used as the optimal design process for stator Design variables
- -1 0 +1 +
Tw (mm) 7.5 8.0 8.5 9.0 9.5
Hso (mm) 0.3 0.6 0.9 1.2 1.5
Tht (mm) 13.0 14.0 15.0 16.0 17.0
Bso (mm) 2.0 2.5 3.0 3.5 4.0

Fig. 9. Optimization process Fig. 10. Design variables

http://www.jeet.or.kr 2311
 In-wheel Motor Design for an Electric Scooter

Table 8. Simulation experiment results (Speed = 710 rpm, needed for the analysis, only the torque ripple under loading
Input phase current = 32 A) condition was evaluated. This is because the cogging
Design variables Response variables torque is the most influential value for torque ripple [31].
No of
Experi- Tw Hso Tht Bso Toque Torque Current Effi- The current density was obtained by dividing the input
ment (mm) (mm) (mm) (mm) constant ripple density ciency current into the cross-sectional area of a pure conductor.
(Nm/A) (%) (A/mm2) (%)
The efficiency was calculated as the ratio of the output
M1 8 0.6 14 2.5 0.93 19.54 5.41 86.50
power to the sum of output power and losses. For the loss,
M2 9 0.6 14 2.5 0.97 14.88 6.08 86.59
M3 8 1.2 14 2.5 0.95 5.27 5.41 86.06
copper loss and core loss, which is calculated with the
M4 9 1.2 14 2.5 0.99 5.02 6.08 86.25 method in [20] and [21], were considered.
M5 8 0.6 16 2.5 0.94 8.44 4.82 85.72 3) Creation of response surface models and integrated
M6 9 0.6 16 2.5 0.98 7.32 5.41 86.01 analysis of different response surface models: RSM was
M7 8 1.2 16 2.5 0.95 6.40 4.82 85.51 applied to determine the relationship between multiple
M8 9 1.2 16 2.5 0.99 5.79 6.08 85.52 design variables and responses in interested area through
M9 8 0.6 14 3.5 0.93 5.70 5.41 86.47 statistical fitting method based on investigational data from
M10 9 0.6 14 3.5 0.97 3.59 6.08 86.61 designed machines [29, 30]. In this paper, a second-order
M11 8 1.2 14 3.5 0.95 4.70 5.41 86.16 response model was established. As in [28-30] each response
M12 9 1.2 14 3.5 0.99 4.17 6.08 86.39 could be represented with a three-dimensional curved
M13 8 0.6 16 3.5 0.94 3.93 4.82 85.92 surface or with two-dimensional curves as a contour map
M14 9 0.6 16 3.5 0.97 4.26 5.41 86.25
in projection plane of the three-dimensional curved surface.
M15 8 1.2 16 3.5 0.95 5.48 4.82 85.72
Fig. 11 shows the examples of the overlapped response
M16 9 1.2 16 3.5 0.99 4.49 6.08 85.77
M17 7.5 0.9 15 3 0.93 3.52 4.82 85.74
surface contour maps. Using the commercial software
M18 9.5 0.9 15 3 1.00 2.17 6.82 86.00
Minitab [32] it is possible to get the necessary regression
M19 8.5 0.3 15 3 0.95 13.44 5.41 86.35 equations and various graphs according to design variables.
M20 8.5 1.5 15 3 0.98 3.76 5.41 85.89 The optimal points, which are represented as dots in Fig.
M21 8.5 0.9 13 3 0.94 5.41 6.08 86.81 11, were determined so that torque constant and efficiency
M22 8.5 0.9 17 3 0.97 3.14 5.41 85.38 were maximized while the torque ripple and current
M23 8.5 0.9 15 2 0.97 6.16 5.41 85.79 density were minimized.
M24 8.5 0.9 15 4 0.96 7.44 5.41 86.23
M25 8.5 0.9 15 3 0.97 2.97 5.41 85.99

Fig. 12. Analysis models for 2-D FEA. (left half) model for
preliminary development motor, and (right half)
(Tht = 15 mm, Bso = 3 mm) (Tht = 15 mm, Bso = 3 mm) model for optimized motor

(Tw = 8.5 mm, Hso = 0.9 mm) (Tw = 8.5 mm, Hso = 0.9 mm)
Fig. 11. Response surface contour maps for two response
variables and two design variables Fig. 13. Flux density distribution for the optimized model

2312 J Electr Eng Technol.2017; 12(6): 2307-2316

 Ji-Young Lee, Byung-Chul Woo, Jong-Moo Kim and Hong-Seok Oh

3.3 Verification of optimal design results

By applying dimensions decided by the process of the

optimal design, a model for the magnetic field analysis was
built as shown in Fig. 12. For the easy comparison, the left
half is the preliminary development motor model, and the
right half is the optimized model. 2-D FEA was conducted
for both of the two models. Fig. 13 shows the magnetic
flux density distribution at the rated speed point on the
rated load torque. The comparison of the expected values
by RSM and calculating values by 2-D FEA is summarized, Fig. 15. Comparison of cogging torques between R1 and
R3
and the preliminary motor dimensions and performances
are also compared with the results of RSM and 2-D FEA as
the optimal model’s dimensions and performances in Table design was confirmed.
9. In the table, R1 stands for the 2-D FEA model of the The major change of between R1 and R3 are that the
preliminary development motor, R2 is the optimized model configuration and dimensions of slot were changed to
by RSM, and R3 is the 2-D FEA model of R2. The facilitate automatic winding and the performances related
differences in torque and efficiency between R2 and R3 torque were improved at the same time. Torque and torque
were less than 0.5%. Therefore, the accuracy of the optimal constant were slightly increased, and torque ripple was
particularly decreased. The cogging torque is dramatically
decreased more than 10 times even though torque ripple
Table 9. Comparison of performance (at the rated speed was considered instead of cogging torque during RSM
and input current 32A) process. For a fair comparison the mesh states of R1 and
Contents R1 R2 R3 Unit R3 are almost same as shown in Fig. 14 since cogging
Tw 8.7 8.8 mm torque value quite depends on the state and the number of
Hso 0.28 0.8 mm elements in the air-gap. Both the meshed air-gaps of R1
Tht 15.3 14 mm
and R3 have 8 layers, and the numbers of elements in air-
Bso 3 3.2 mm
Net weight 7.5 7.7 kg
gap of R1 and R2 are 8656 and 8722 respectively. The
Torque 31.2 31.4 31.5 Nm variation of cogging torque value is just about 1.5% even
Torque constant 0.98 0.98 0.99 Nm/A though the number of element in air-gap becomes over
Torque ripple 4.83 2.63 2.53 % 13000. Fig. 15 shows the cogging torque plot. One period
Cogging torque 0.92 0.06 Nm of cogging torque is 1.25 degree because the least common
-
(peak-to-peak) 2.9 0.2 %
Current density 5.7 5.95 6.08 A/mm2
multiple of 32-pole and 36-slots is 288 and it is the
Output power 2.32 2.33 2.34 kW frequency of cogging torque.
Output power density 0.31 0.30 0.30 kW/kg However current density and net weight were also
Copper loss 77.9 - 83.8 W slightly increased because slot area shrank and core area
Core loss 300.8 - 292.5 W was increased due to change the configuration of slot
No. of strands in a wire 18 - 7 bottom from arc form to flat form. Therefore the output
Diameter of strand 0.63 - 1.0
power density, which is the ratio of output power to the net
Efficiency 86.0 86.3 86.2 %
weight, was a little decreased. The specifications related to
strand was changed independently of optimal design
process. Considering the process of automatic winding, the
minimum number of strands in a wire and the maximum
diameter of strand, could be caught by a needle in the
automatic winding machine, were selected.

(a) R1 (2D FEA model of the preliminary development motor)

4. Experiment Results

The optimized model, R2, was fabricated as shown in

Fig. 16. The stator coils were wound by an automatic
winding machine, and the rotor has the solid core. Fig.
17 shows an overall look of a combination of the fabricated
in-wheel motor and a wheel, and the test rig. The load
(b) R3 (2D FEA model of the optimized model by RSM) testing of the in-wheel motor combined into the wheel
Fig. 14. Mesh plots in air-gap of R1 and R3 was conducted with the test rig, TB-500KSC, which was

http://www.jeet.or.kr 2313
 In-wheel Motor Design for an Electric Scooter

driven by torque control with Hysteresis brake [33]. The

test was performed according to load torque variation
under constant speed condition. The input and output
characteristics are shown in Fig. 18. The phase current is
rms (root-mean-square) value. The waveform of EMF is
shown in Fig. 19. Since the input current waveform is
sinusoidal, the motor was designed to have the sinusoidal
EMF waveform.
Fig. 19. Measured emf waveform for three phases
The comparison of the analysis results and tested results
is listed in Table 10. In the cases of EMF and torque, the
Table 10. Comparison of performances at the rated speed
analysis and measured values are similar each other, but
the input current has 8.6% difference. It is because the Contents FEA TEST Unit
measured input current includes the amount of current to EMF at 1000rpm 34.2 34.0 V (rms)
Phase resistance 0.02 0.02 Ohm
offset losses including core loss while analysis current
Torque 31.5 30.0 Nm
Output power 2342.1 2196.0 W
Torque constant 0.98 0.86 Nm/A
Torque ripple 2.53 - %
Phase current 32 35 A (rms)
Current density 6.08 6.65 A/mm2
Efficiency 86.2 85.8 %
Cogging torque 0.06 0.4 Nm
(peak-to-peak) 0.2 1.33 %

value is only for torque and core loss is calculated in the

post process just considered flux density and frequency.
Fig. 16. The fabricated stator (left) and rotor (right) of the Therefore, the efficiency could be similar because the
designed in-wheel motor copper and core losses as major losses were considered.
The small difference of the efficiency is presumed to be
because of mechanical loss, and calculation and measurement
error. The torque constant also has a big difference
between FEA and test results because of the current
difference.
The experimental cogging torque waveform includes
abnormal low frequency and harmonics because of a little
mechanical assembly problem, so the components except
lower frequency than ideal cogging torque frequency are
only considered as a measured value in Table 10. The
cogging torque value meets the specifications even though
Fig. 17. The in-wheel motor mounted on a tire (left) and the value is much more than expectation because of the
the test rig (right) additional harmonics.

5. Conclusion

So far it is presented that the design process and results

of in-wheel motor for the E-scooter considering economical
production. The most constraints for design were obtained
analytical and experimental assessments from the preli-
minary development motor. The design point was stator
core change for automatic winding while maintaining the
rotor specifications and output characteristics such as
power and efficiency. The design results meets those
requirements, and besides cogging torque was extremely
reduced. Hopefully, this paper would help develop in-
Fig. 18. Experimental data at the rated speed wheel motor for E-scooter.

2314 J Electr Eng Technol.2017; 12(6): 2307-2316

 Ji-Young Lee, Byung-Chul Woo, Jong-Moo Kim and Hong-Seok Oh

Acknowledgment Appl., vol. 8, iss. 1, pp. 1-12, 2014.

[14] I. Husan, Electric and Hybric Vehicles Design
This research was supported by the KERI Primary Fundamentals 2nd Edition, CRC Press, 2011.
research program of MSIP/NST (No. 17-12-N0101-20). [15] YM. Ehsani, Y. Gao, S. E. Gay, and A. Emadi,
Modern Electric, Hybrid Electric, and Fuel Cell
Vehicle, CRC Press, 2004.
References [16] L. Guzzella and A. Sciarretta, Vehicle Propulsion
Systems- Introduction to Modeling and Optimization
[1] K. Sone, M. Takemoto, S. Ogasawara, K. Takezaki, 2nd Edition, Springer, 2007.
and H. Akiyama, “A ferrite PM in-wheel motor [17] Y. Chen, X. Li, C. Wiet, and J. Wang, “Energy
without raer earth materials for electric city com- management and driving strategy for in-wheel motor
muters,” IEEE Trans. Magn., vol. 48, no. 11, pp. electric ground vehicles with terrain profile preview,”
2961-2964, Nov. 2012. IEEE Trans. Industrial Informatics, vol. 10, no. 3, pp.
[2] S. Chung, S. Moon, D. Kim, and J. Kim, “Develop- 1938-1947, 2014.
ment of a 20-pole-24-slot SPMSM with consequent [18] J. Larminie and J. Lowry, Electric Vehicle Tech-
pole rotor for in-wheel direct drive,” IEEE Trans. Ind. nology- Explained, Wiley, 2003.
Electron., vol. 63, no. 1, pp. 302-309, 2016. [19] J. Lee, J. Kim, and B. Woo, “Optimal design of in-
[3] C. J. Ifedi, B. C. Mecrow, S. T. M. Brockway, B. S. wheel motor for an E-bike,” in Proc. 2016 IEEE
Boast, G. J. Atkinson, and D. K. Perovic, “Fault- Transportation Electrification Conference and Expo
tolerant in-wheel motor topologies for high- (ITEC Asia-pacific), pp.441-443, 2016.
performance electric vehicle,” IEEE Trans. Ind. Appl., [20] D. Hong, J. Lee, B. Woo, D. Park, and B. Nam,
vol. 49, no. 3, pp. 1249-1257, May/June 2013. “Investigating a direct-drive PM type synchronous
[4] Y. P. Yang, H. C. Lin, F. C. Tsai, C. T. Lu, and K. H. machine for turret application using optimization,”
Tu, “Design and integration of dual power wheels IEEE Trans. Magn., vol. 48, no. 11, pp. 4491-4494,
with rim motors for a powered wheelchair,” IET November 2012.
Electr. Power Appl., vol. 6, no. 7, pp. 419-428, 2012. [21] Maxwell online help, Ansys Corporation, Maxwell
[5] Y. P. Yang, W. C. Huang, and C. W. Lai, “Optimal 16.0
design of rim motor for electric powered wheelchair,” [22] http://www.nittoku.co.jp/english/product/itiran/brush-
IET Electr. Power Appl., vol. 1, no. 5, pp. 825-832, 2007 inner.html
[6] Y. P. Yang, F. X. Ding, “Driving-scenario oriented [23] http://bestec-web.co.jp/en/product/product_list.
design of an axial-flux permanent-magnet synchronous [24] http://lutron-ind-en.weebly.com/motor-winding-
motor for a pedal electric cycle,” IET Electr. Power machine.html
Appl., vol. 9, no. 6, pp. 420-428, 2015 [25] http://www.unioncoil.com
[7] T. F. Chan, L. T. Yan, and S.Y. Fang, “In-wheel [26] R. H. Myers and D. C. Montgomery, Response
permanent-magnet brushless dc motor drive for an Surface Methodology: Process and Product Optimi-
electric bicycle,” IEEE Trans. Energy conversion, vol. zation Using Designed Experments, Nework, NY,
17, no. 2, pp. 229-233, June 2002. USA: Wiley, 2002.
[8] A. Muetze and Y. C. Tan, “Electric bicycles – a [27] S. Lee, J. Hong, S. Hwang, W. Lee, J. Lee, and Y.
performance evaluation,” IEEE Ind. Appl. Magazine, Kim, “Optimal design for noise reduction in interior
pp. 12-21, 2007. permanent-magnet motor,” IEEE Trnas. Ind. Appl.,
[9] Z. Q. Zhu and D. Howe, “Electrical machines and vol. 45, no. 6, pp. 1954-1960, Nov. 2009.
drives for electric, hybrid, and fuel cell vehicles,” [28] J. Lee, B. Park, and B. Woo, “Design and comparison
Proceedings of the IEEE, vol. 95, no. 4, pp. 746-765, of multistage axial flux permanent magnet machines
April 2007. for portable generating application,” Journal of
[10] http://www.greenmobility.kr/html/comm/main.asp Magnetics, vol. 20, no. 4, pp. 1-8, 2015.
[11] T. Kim, O. Vodyakho, and J. Yang, “Fuel cell hybrid [29] J. Xie, D. Kang, B. Woo, J. Lee, Z. Sha, and S. Zhao,
electric scooter,” IEEE Ind. Appl. Magazine, pp. 25- “Optimal design of transverse flux machine for high
31, Mar/Apr. 2011. contribution of permanent magnet to torque using
[12] D. Fodorean, L. Idoumghar, and L. Szabo, “Motori- response surface methology,” JEET, vol. 7, no. 5, pp.
zation for an electric scooter by using permanent- 745-752, 2012.
magnet machines optimized based on a hybrid [30] S. Chung, and J. Kim, “Double-sided iron-core
metaheuristic Algorithm,” IEEE Trans. Vehicular PMLSM mover teeth arrangement design for
Technology, vol. 62, no. 1, pp. 39-49, January 2013. reduction of detent force and speed ripple,” IEEE
[13] Y. Yang, C. Lee, and P. Hung, “Multi-objective Trans. Ind. Electron., vol. 63, no. 5, pp. 3000-3008,
optimal design of an axial-flux permanent-magnet May 2016.
wheel motor for electric scooters,” IET Electr. Power [31] D. Wang, X. Wang, and S. Jung, “Cogging torque

http://www.jeet.or.kr 2315
 In-wheel Motor Design for an Electric Scooter

minimization and torque ripple suppression in Hong-Seok Oh He received the B.S.,

surface-mounted permanent magnet synchronous M.S, and Ph.D degrees in electrical
machines using different magnet widths,” IEEE Trans. engineering from Yeungnam Univer-
Magn., vol. 49, no. 5, pp. 2295-2298, May 2013. sity, Korea, in 1992, 1994, and 2000
[32] https://www.minitab.com/en-us/. respectively. Now He is working as a
[33] http://www.sugawara-labs.co.jp/en/motors/torque/ managing director in Star Group
ta_tb_torque_meter_series Industry Company (SGI). His research
interests are design and analysis of
electromagnetic machines for transportation, and robot
actuators.
Ji-Young Lee She received the B.S.,
M.S, and Ph.D degrees in electrical
engineering from Changwon National
University, Korea, in 2000, 2002, and
2006 respectively. Now she is working
as a senior researcher in Korea Electro-
technology Research Institute (KERI),
and at the same time as an associate
professor at University of Science & Technology (UST).
Her research interests are design and analysis of electro-
magnetic machines for transportation, robot, and medical
systems.

Byung-Chul Woo He received the B.S.

degree in Mechanical engineering from
Youngnam University in 1989, and the
M.S. and Ph.D. degrees in Mechanical
Design from Kyungpook National
University in 1991 and 2000 respecti-
vely. He is working mechanical design
for electric machine in Korea Electro-
technology Research Institute (KERI). His research interests
are electric machines, actuator, robot, and automation
systems.

Jong-Moo Kim He received the B.S.,

M.S., and Ph.D. degrees in electrical
engineering from Yeungnam University,
Gyeongsan, Korea, in 1992, 1994, and
2009, respectively. He is currently a
Principal Researcher with the Electric
Motor Research Center, Korea Electro-
technology Research Institute, Changwon,
Korea. His research interests include high performance
motion and motor control, control theory, DSP applications,
contactless power supplies, and electric vehicles.

2316 J Electr Eng Technol.2017; 12(6): 2307-2316

View publication stats