1. Introduction

The automotive industry is focusing on environment friendly vehicle development for increasing fuel efficiency and reducing gas emission. The fact that hybrid vehicles should install the traction motor within a limited size of engine room requires batteries to be installed in small volume. Therefore, a novel design approach for vehicle system development is required. In terms of this design requirement, in-wheel system can solve conventional size problems by directly mounting the electric motor to the wheel. Additionally, it is possible to increase efficiency of the drive system by removing the power train elements. By installing traction motors to each wheel, it is possible to increase the system output power and maximize the amount of regenerated energy ^[1,2].

Considering the advantage that the electric motor can be inserted inside the wheel, concentrated winding is usually applied to in-wheel systems. Concentrated winding is superior to distributed winding in terms of easiness of manufacturing and output power density. However, the concentrated magnetic flux results in regional saturation of the stator core. As a result, the magnetic resistance increases, and the reluctance torque is small. Additionally, large eddy current loss of the permanent magnet results in heat demagnetization. Also, concentrated winding shows weakness in vibration/noise characteristics and shows poor performance at high speed drives as the space harmonics are larger compared to distributed winding configuration.

This paper is on design optimization for an electric motor applied in in-wheel motor systems. The design is based on concentrated winding for space minimization. The optimization objective is the permanent magnet configuration which shows good characteristics of reluctance torque and heat demagnetization. To validate our design result, a prototype was manufactured and several tests were performed with the prototype.

Fig. 1. Configuration of In-Wheel system

2. Specification and Load calculation

Design of general industrial motors is usually optimized within the constraint of constant loads. However, design optimization of traction motors should consider load variation and environment factors. In addition, full understanding of the system is essential.

The configuration of the In-wheel system is shown in Figure 1^[3]. The traction motors installed on the rear wheels of electric vehicles are composed of torsion beam suspension, motor control inverter, planetary gear reducer, drum brake, Hub bearing and in-wheel components.

In-wheel electric vehicles are expected to confront several types of loads. Several examples could be rolling load between the road and wheels, air-drag load induced by the air and slope load generated by the road slope and gravity. Therefore, the maximum value of required force for in-wheel electric vehicles should be computed by considering various factors such as vehicle mass, acceleration, friction and slope. The total thrust required by the In-wheel motor is shown in Equation (1).

As the system is a geared system, the motor load was selected by considering the required force and gear ratio. Figure 2 shows the speed-torque characteristics of the motor considering the gear ratio. The specification of the In-wheel system is shown in Table 1.

3. Motor design for In-Wheel Drive

In this paper, required output power and base speed was selected by considering the load characteristics, battery voltage, inverter voltage rating and the current characteristics. Based on the required power and base speed, the IPMSM was designed using the D$^2$L design approach. The design flow of IPMSM is shown in Figure 3.

Fig. 2. T-N curve of In-Wheel motor

As electric motors in hybrid systems are generally coupled to the engine, noise and vibration generated by the motors is not considered seriously during the motor design process. However, the noise and vibration generated from electric motors should be considered seriously in case of in-wheel system applications.

This paper applied Taguchi method to improve vibration and noise characteristics. Output power and reduction of EMF THD is selected as objective functions. Four design factors were selected such as configuration of shoe, configuration of permanent magnet, configuration of barrier and type of coil. The temperature of PM and tolerance of PM was defined as noise. This paper does not deal with the Taguchi method because it aims to reduce the eddy current loss in the permanent magnet and to increase the reluctance torque. This design uses a J-MAG, a finite element method tool.

Fig. 3. Flowchart for design of In-Wheel motor

4. Permanent magnet demagnetization characteristics

A typical cause of electric motor failure is permanent magnet demagnetization. Mechanism of demagnetization can be classified into two cases. The first case is irreversible demagnetization caused by momentary short circuit or excessive d-axis current induced during field weakening operation. The other case is heat demagnetization caused by high temperature. Permanent magnets show different B-H characteristics for different temperature. Although, the knee point does not exist on the second quadrant at room temperature, the knee point exists on the second quadrant at high temperature. This characteristic causes heat demagnetization ^[4-6].

During permanent magnet design process, it is general to consider the above mentioned characteristics. However, additional consideration on eddy current loss is required for design of in-wheel motors. The cause of eddy current loss of permanent magnet can be explained as the following. NdFeB magnet has conductivity, which is different from ferrite magnet, and results in eddy current loss. Especially, in-wheel system motors operate under low voltage level and high current level environments, and this results in high change rate of stator flux(dØ/dt). The change of magnetic flux induces voltage and causes current to flow inside permanent magnet which has conductivity. The eddy current loss of magnet results in heat and increases the magnet temperature. According to the author’s research on literature, there was no study case reported on the topic of electric motor design considering eddy current loss of permanent magnet. It is solved by increasing permanent magnet thickness.

Figure 4 shows the theoretic explanation of demagnetization and the eddy current distribution inside the permanent magnet according to flux variation.

Fig. 4. Cause of demagnetization at PM and eddy current in PM

5. Permanent Magnet Shape Optimization Design Considering Eddy Current Loss

Demagnetization of permanent magnets starts from the edge of magnets and spreads its region to the other sections of the motor. As the region of demagnetization increases, the back-emf, torque and power decreases. Generally, to solve this problem, the permanent magnet thickness is determined by considering the worst condition (maximum temperature, maximum d-axis current). During the design process, internal magnetic flux density of the permanent magnet should be higher than the maximum temperature knee point magnetic flux density^[7].

As it is hard to predict the amount of heat loss caused by eddy current loss of permanent magnet, the conventional design approach is either to minimize the eddy current loss itself or to assign margins to the thickness of magnet. However, increasing the permanent magnet thickness is not a desirable design because the cost of manufacturing the motor increases.

Figure 5 shows the eddy current loss distribution of the permanent magnet, and shows the conventional design and proposed design for improving demagnetization characteristics. As it is shown in Figure 5(a), the eddy current loss mostly takes place at the edge of the magnet. Therefore, demagnetization starts initially from the edge of magnets. To solve this problem, the magnet is divided or a thicker permanent magnet can be installed as shown in Figure 5(b) . However, this solution can result in an increase of manufacturing cost of the product. Instead of applying a fixed value of magnet thickness, this paper proposed a design of adopting a variable magnet thickness. In detail, a thicker magnet is applied for regions that have high eddy current loss and a less thick magnet is applied to regions that have a low eddy current loss. By introducing this design approach, the demagnetization improved from 0.42T to 0.54 T with a reduced usage of permanent magnet.

Fig. 5. Eddy current loss distribution of PM motor

Fig. 6. Q axis Flux path

Figure 6 shows the q axis flux path of both traditional model and proposed model. As shown in Figure 6(a) , the traditional model increases the thickness of all the permanent magnet, so that the q-axis flux path becomes narrow and saturation occurs. As a result, the q-axis inductance decreases as shown in Figure 7(a) . On the other hand, the proposed model has wider q-axis flux path through permanent magnet shape design considering eddy current loss. When the q-axis flux path is secured, the q-axis inductance becomes large, and the reluctance torque increases. Figure 7 shows the q-axis inductance of both conventional model and proposed model. As the q-axis inductance of the proposed model is higher. Figure 8 shows the torque characteristics of the existing model and the improved model. When the same current was applied, the improvement model increased the total torque due to the increase of the reluctance torque and the current phase angle also increased. The increased reluctance torque can enable increased power and efficiency at the high speed region. When the permanent magnet shape is designed in consideration of the eddy current generated in the permanent magnet, the performance of the motor is improved.

Fig. 7. Q axis Inductance

Fig. 8. Torque characteristics

6. Analysis and experiment

A prototype was made to validate the research results and major waveforms were compared with the simulation results. Figure 9 shows the electric motor embedded inside the wheel. It can be seen that the region with smaller PM thickness results in lower eddy current loss, while the larger PM thickness region shows larger eddy current loss. The no-load back-emf is measured from a constantly rotating motor. In this study, the emf was measured from a motor rotating at a speed of 1000rpm. The analysis result and experiment result of the improved model is shown in Figure 10. It can be seen that the results fairly match.

Fig. 9. Manufactured IPMSM

Figure 11 shows the torque characteristics for various current phase angles. The torque equation of IPMSM is shown in equation (2). The torque of IPMSM is an addition of magnetic torque and reluctance torque. As it is shown in equation (2), the magnetic torque is proportional to the current and the reluctance torque is proportional to the square of current. Therefore, it can be understood that the current phase angle for maximum torque increases as the current value increase. The difference between experimental values and computer simulation results can be explained by 1) not considering the iron loss and 2) Br value variation caused by temperature increase.

Fig. 10. EMF waveform of PM machine

Fig. 11. Torque characteristics as current angle

7. Conclusions

This paper introduced a novel permanent magnet motor design for in-wheel motor applications which minimized usage of magnet and shows good demagnetization characteristics. In addition, the high speed characteristics are improved by increasing the reluctance torque.

The output power and base speed is determined by considering the In-wheel system, load characteristic, battery voltage, inverter voltage rating and current condition. Based on the calculated base speed and output power, IPMSM design was done by applying D$^2$L method.

Although the concentrated winding method has an advantage of having a smaller coil- end length, the reluctance torque is relatively small and heat demagnetization can happen because of the eddy current loss. Instead of using a conventional approach to this problem, this paper introduced a variable magnet thickness concept based on the eddy current loss distribution. This resulted in decrease of permanent magnet usage and increase of demagnetization level.

Additional increase in reluctance torque was possible by having a q-axis flux path.

To validate the design, a prototype was manufactured and compared with computer simulation results. This design result can be applied not only to in-wheel drive systems but also to broad applications fields of HEV, EV and FCEV traction motor design.