Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

IEEE TRANSACTIONS ON MAGNETICS, VOL. 40, NO.

4, JULY 2004

1873

Design and Control of Axial-Flux Brushless DC Wheel Motors for Electric VehiclesPart I: Multiobjective Optimal Design and Analysis
Yee-Pien Yang, Member, IEEE, Yih-Ping Luh, and Cheng-Huei Cheung

AbstractWe have applied multiobjective optimal design to a brushless dc wheel motor. The resulting axial-flux permanent-magnet motor has high torque-to-weight ratio and motor efficiency and is suitable for direct-driven wheel applications. Because the disk-type wheel motor is built into the hub of the wheel, no transmission gears or mechanical differentials are necessary and overall efficiency is thereby increased and weight is reduced. The dedicated motor was modeled in magnetic circuits and designed to meet the specifications of an optimization scheme, subject to constraints such as limited space, current density, flux saturation, and driving voltage. In this paper, two different motor configurations of three and four phases are illustrated. Finite-element analyses are then carried out to obtain the electromagnetic, thermal, and modal characteristics of the motor for modification and verification of the preliminary design. The back-electromotive forces of prototypes are examined for control strategies of current driving waveforms. Index TermsAxial-flux wheel motor, electric vehicle, magnetic analysis, optimal design.

I. INTRODUCTION

GROWING interest in electric vehicles (EVs) has driven researchers and engineers to develop more efficient and reliable power systems under the pressure of the protection of natural environment. Traditional power systems for EVs are composed of batteries, electric motors with drives, and transmission gears to wheels. Each subsystem converts chemical, electrical, or mechanical energy into different forms, consuming energy through the dissipation components of windage and friction. It is quite essential for engineers to look for an approach to improve the overall efficiency of electric vehicles, and hence to increase their driving range. In addition to new battery technologies, new concepts for the design of motors and their optimal driving pattern have attracted substantial attention for the improvement of overall efficiency and reliability of EVs. Among the various researchers, Chang [1] provided an experts survey and concluded that induction motor drives were preferred for EV propulsion purpose, due to their low cost, high reliability, high speed, established converter and manufacturing technology, low torque ripple/noise, and absence of position

Manuscript received December 18, 2002; revised March 16, 2004. This work was supported by National Science Council of Taiwan, R.O.C., under Contract NSC90-2212-E002-218. The authors are with the Department of Mechanical Engineering, National Taiwan University, Taipei 106, Taiwan, R.O.C. (e-mail: ypyang@ntu.edu.tw). Digital Object Identifier 10.1109/TMAG.2004.828164

sensors. However, the permanent-magnet brushless dc motor featured compactness, low weight, and high efficiency, and therefore provided an alternative for EV propulsion. Several permanent-magnet motors have been developed for EVs to fulfill the special requirements, such as high power density, high efficiency, high starting torque, and high cruising speed. These motors can be classified as indirect-driven [2] and direct-driven motors [3][5]. The latter, also called wheel motors or hub-in motors, are directly mounted inside the wheels, thus eliminating transmission gears or mechanical differentials with their associated energy loss. The reduction of mechanical components in transmission chains or gears not only improves the overall efficiency, but also reduces the weight of the vehicle. In the class of direct-driven motors, the axial-flux motor competes the radial-flux motor with some advantages, such as balanced motor-stator attractive forces, better heat removal configuration, no rotor back iron, and adjustable air gap [6]. Zhang et al. [7] compared the several axial-flux permanent-magnet (AFPM) wheel motors for electric cars, and concluded that the ones with interior PM seemed to be the best compromise in terms of power density, efficiency, compactness, and capability characteristics. Their motor configuration had a stator core with slots on both sides, which was sandwiched by two steel rotor disks with arc-shaped poles and tangentially magnetized square magnets. Similar design of a two-stage AFPM machine with ironless water-cooled stator winding was presented by Caricchi et al., where the switching from series to parallel winding connections provided variable-speed operations [5]. Lovatt et al. [8] addressed the solar-powered vehicle and axial-flux in-wheel motor, which consisted of a magnetic array oriented along the direction of the flux flow and an ironless air-gap winding. Hredzak et al. also had their double-sided axial-flux direct wheel motor, and the torque pulsations were eliminated by a vector control scheme [9]. This paper proposes a systematic optimal design methodology on the permanent magnet axial-flux brushless dc wheel motor and its drive for EVs. Its specifications for motorcycles are proposed according to the policy of the environmental protection agency in this country. First, the electromagnetic properties are modeled in terms of motor geometries and electrical parameters with sensitivity analyses. Second, the preliminary motor shapes of three and four phases are obtained by optimizing a set of cost functional subjected to the constraints on the design parameters and physical properties. Third, the finite-element analyses are performed on the electromagnetic, thermal,

0018-9464/04$20.00 2004 IEEE

1874

IEEE TRANSACTIONS ON MAGNETICS, VOL. 40, NO. 4, JULY 2004

TABLE I SPECIFICATIONS OF ELECTRIC VEHICLE WITH WHEEL MOTOR

Fig. 1.

Explosive graph of the axial-flux disk-type wheel motor.

and dynamical characteristics. Finally, the prototypes are fabricated and experiments are performed to provide information for possible current driving patterns. II. MOTOR LAYOUT AND SPECIFICATIONS The novel design of the disk-type AFPM wheel motor presented in this paper is a prototype for electric vehicles, which was developed by the Electro-mechanical Power System Research Group at National Taiwan University, Taipei, Taiwan, R.O.C. The graph of the wheel motor is illustrated in Fig. 1. The rotor disk of the hub-in motor has 16 magnets, and is sandwiched between the two stator plates, each with 24 teeth to form a three-phase motor. For the same stator structure, 18 magnets in the rotor form a four-phase motor. The fractional number of slots per pole per phase is desired so as to yield a uniform magnetic force distribution between the stator and rotor, hence eliminating most of the cogging torque that usually occurs in permanent-magnet motors. The main magnetic flux flows through two air gaps between the stator and rotor along the axial direction. The tire is installed on the outer case, which rotates with the rotor. The final shape of this wheel motor is designed to meet the requested specifications of a multifunctional optimization scheme, with various constraints, such as limited space, current density of conductor, flux saturation, and driving voltage. Instead of being X-connected for four-phase motors or Y-connected for three-phase motors, the coils are independently wired on stator poles, and are grouped into required phases without a neutral. Since there is no neutral point for independent winding, the driving voltage is directly applied to each phase. Therefore, larger back-electromotive force (EMF) is induced and higher motor speed can be reached. Other specifications are listed in Table I. III. MAGNETIC CIRCUIT MODEL The torque of electric motors is produced by the rate of change of magnetic energy stored in the air gap. The magnetic energy comes from the magnetic field generated by the current flowing in the wires and/or permanent magnets. Both the

Fig. 2. Dual axial flux motor topology and its 2-D configuration.

sources generate magnetic flux forming flux loops in the magnetic materials of the motor. Based on the assumptions of material linearity and the collinearity of flux and field densities, the magnetic circuit model is used to describe the torque produced in the motor. It is also necessary to make three additional assumptions. 1) The motor is operated in the linear range of the curve of the magnetic material. 2) The air-gap reluctance of the slotted stator structure is approximated by the effective air-gap length with Carters coefficient [10]. 3) The flux flows straightly across the air gaps between the stator and rotor, ignoring the fringing flux for simplified analysis. The three-dimensional (3-D) motor structure can be simplified to a two-dimensional (2-D) configuration, and its two-side topology is cut in half for facilitating the magnetic circuit analysis, as shown in Fig. 2. The stator is laminated by sheets of electric steel, which are oriented in a stack coming out of the paper. The fan-shaped magnets are mapped into rectangular ones as the arc is transformed to a straight line in the 2-D linear motor mode. Therefore, the total flux through the same area of the surfaces of the magnet is unchanged. Take the three-phase motor as an example, in a section of 360 electrical degrees, the magnetic circuit of one flux loop is composed of three teeth on each side of the stator facing toward two permanent magnets embedded

YANG et al.: DESIGN AND CONTROL OF AXIAL-FLUX BRUSHLESS DC WHEEL MOTORS

1875

Fig. 4.

Parameters of stator and rotor of the wheel motor magnet flux.

. It is worth noting that the the average radius above functions used for the motor design are expressed explicitly in terms of the geometric dimensions and properties of magnetic materials of the wheel motor, as described in Appendix A. That inspires an optimal motor design so that better performance objectives are fulfilled. IV. DESIGN OPTIMIZATION The electrical and mechanical performances of the wheel motor depend on its geometry and properties of magnetic materials. In the past, basic electromagnetic theories and industrial experience are used for preliminary motor design, and the finite-element analysis contributes to the detailed modification and verification of the final shape of the motor. However, performance of a motor can be enhanced by the optimal design in terms of efficiency, weight, torque, and frequency response. These characteristics are described by the above electromagnetic equations, functions of motor dimensions, material parameters, and electrical constants. A. Sensitivity to Design Variables In most of the efficiency optimization methods, the design sensitivity analysis is required to determine the derivatives of the objective functions with respect to the parameters of interest. In this paper, the sensitivity derivatives are not obtained explicitly from the equations. Instead, the sensitivities of the motor torque, torque density, and efficiency to the design variables are numerically investigated. The purposes of the sensitivity analyses are as follows. 1) The designer may want to discard those design variables with the least sensitivities of the torque, torque density, and efficiency of the motor. 2) The designer may keep those design variables constant with sensitivities which are linear, or monotonic functions. 3) Only those design variables not included in the above two cases are retained for the subsequent optimal design. In this way, the number of the design variables can be kept minimal to expedite the design optimization of motors. On the other hand, decision makers can make a proper shift or modification of the final solutions according to the linearity of sensitivity with respect to a certain motor parameter. Fig. 4 illustrates some possible design variables in geometry. Other variables are also investigated in the sensitivity analysis, such as number of stator poles, number of winding layers, number of turns per layer, and copper wire diameter.

Fig. 3. motor.

Magnetic circuit model of an electrical period for the three-phase

in the rotor. As a matter of fact, the ferromagnetic material has very high permeability so that its reluctance can be ignored. The magnetic circuit model for one electrical period of half-side of the motor is depicted in Fig. 3 in terms of the air-gap reluctances, stator magnetomotive forces, and fluxes through the magnetic , and circuit. The air-gap reluctances are denoted by corresponding to three stator phases, while , and represent the reluctances of stator teeth, back iron and slot, , respectively, and the corresponding fluxes are given by and . The air-gap fluxes , and are calculated by the equivalent magetomotive force of the magnets and effective air-gap length. The magnetomotive forces of slots are rep, and , each of which contains resented by two phase coils. The Kirchhoffs voltage law calculates the magnetic flux at each branch, and the flux density of the air gap is (1) where is the overall magnetomotive force distribution is the air-gap from stator windings and rotor magnets, is the permeability of free space, and denotes the length, rotor shift. Therefore, the torque distribution is given by (2) where the coenergy is expressed as (3) the outer and inner radii of the stator are and , respectively, and represents the peripheral coordinate along the circle of

1876

IEEE TRANSACTIONS ON MAGNETICS, VOL. 40, NO. 4, JULY 2004

Fig. 5. Sensitivity of motor efficiency versus number of stator poles.

Fig. 7.

Sensitivity of motor efficiency versus inner radius.

Fig. 8. Fig. 6. Sensitivity of motor efficiency versus outer radius.

Sensitivity of motor efficiency versus air-gap length.

According to the geometric, material, and electric specifications and constraints, a set of nominal values of design variables is predetermined. For the three-phase motor configuration, the sensitivity curves of the motor efficiency with respect to each single design variable are obtained from the magnetic circuit model, while other variables are kept at their nominal values. It is not surprising that the increasing number of stator poles and larger outer radius produce larger torques and higher efficiencies, as shown in Figs. 5 and 6. However, these parameters are assigned with upper limits due to constrained wheel space for the motor. On the contrary, a smaller inner radius of stator poles allows larger magnetic flux distribution in the air gap to produce torque, and hence the motor efficiency as depicted in Fig. 7. It is also reasonable that longer air-gap length yields larger gap reluctance, and thereby lower efficiency, as proillustrated in Fig. 8. Fig. 9 shows that the thicker rotor duces more magnetomotive force and yields higher efficiency. However, the increasing inertia of the motor makes its efficiency reach a limit. Larger efficiency can be produced by thicker back for less magnetic density and core loss in the eleciron tric steel, as shown in Fig. 10. The motor efficiency is also , magnet fraction sensitive to the slot fraction , stator tooth fraction , and , as shown in Figs. 1114. shoe depth fraction The number of winding layer , the number of turns of layer , and the wire diameter , which affect the excitation of current and the magnetomotive force, have considerable influence on the efficiency, as displayed in Figs. 1517. The sensitivities of the motor torque and torque density, defined by the generated torque per unit motor weight, with respect to the motor parameters were also investigated. Different from the sensitivities on the motor efficiency, the torque is not sensitive to the thickness of the back iron, stator tooth fraction, and shoe depth fraction. The design variables are thus determined by the above sensitivity analyses and are listed in Table II.

Fig. 9.

Sensitivity of motor efficiency versus rotor thickness.

Fig. 10.

Sensitivity of motor efficiency versus back iron thickness.

Fig. 11.

Sensitivity of motor efficiency versus slot fraction.

B. Optimization The compromise programming method in the multifunctional optimization system tool (MOST) [11] is applied to search for

YANG et al.: DESIGN AND CONTROL OF AXIAL-FLUX BRUSHLESS DC WHEEL MOTORS

1877

Fig. 12.

Sensitivity of motor efficiency versus magnet fraction.

Fig. 17.

Sensitivity of motor efficiency versus wire diameter. TABLE II DESIGN PARAMETERS

Fig. 13.

Sensitivity of motor efficiency versus stator tooth fraction.

the optimal values of the design variables that maximize the following performance indexes: Motor torque: (4)
Fig. 14. Sensitivity of motor efficiency versus shoe depth fraction.

Torque density: (5) Motor efficiency: % (6)

Fig. 15.

Sensitivity of motor efficiency versus number of winding layers.

Fig. 16.

Sensitivity of motor efficiency versus number of turns per layer.

The average torque is a function of design variables implicitly, and is calculated from (2) by 60 equally spaced points of rotor shift over an electrical period. The driving current is a six-step three-phase-ON square wave. Likewise, the weight of , and the stray the motor , the ohmic loss , the core loss composed of windage, friction, noise, and other less loss dominant loss components, are all functions of design variables. The general formulations of the compromise programming method are described in Appendix B. The optimizer weighs these performance indexes to reach a satisfactory compromise among the design variables under the prescribed constraints. 1) The motor dimensions must be realized. 4. 2) The permeance coefficient of permanent magnets 3) The slot current density is less than 9 10 (A/m ). 0.42. 4) Conductor packing factor 5) The slot opening is 1.8 times larger than the air-gap length.

1878

IEEE TRANSACTIONS ON MAGNETICS, VOL. 40, NO. 4, JULY 2004

TABLE III MOTOR SPECIFICATIONS FROM THE MAGNETIC CIRCUIT MODEL

Fig. 18.

Flowchart of optimization process.

6) The peak value of back-EMF per phase should be less than the component of the driving voltage along the back-EMF vector. 7) The slot opening is 0.35 times less than the slot pitch. 8) The flux density in electrical steel is less than its saturation value 1.8 T. 9) The shoe depth fraction is confined between 0.25 and 0.5. The optimizer MOST can deal with real, integer, and discrete design variables simultaneously. In this design, the performance indexes, design functions, and prescribed constraints are all expressed in terms of design variables, in which the number of stator poles, the number of winding layers, and the number of turns per layer are integer design variables, while the wire diameter provided by manufacturers is a discrete design variable. The computational flow of the gradient-based optimization algorithm in MOST is composed of the following steps. 1) Give a set of initial guess of parameters in the design space, and evaluate the objective and constraint functions. 2) Calculate gradients of the objective and constraint functions with respect to each design variable. The gradients are not self-designed, but automatically calculate gradients using forward, backward, or central finite differences. 3) Based on the values of functions and gradients, determine a maximum descent direction, and the next set of design parameters are determined. 4) If any constraint is violated or convergence test or not satisfied, adjust the current point and repeat steps 1)4) until a final solution is obtained. Fig. 18 shows the flowchart of the optimization process. The motor parameters of the three and four phases from the optimal design and the technical data are listed in Table III, and the detailed geometrical features of the stator tooth and the rotor of the four-phase motor are illustrated in Fig. 19. Since the coils are independently wired on stator poles, eight parallel coils are

grouped into one phase for each side of the stator of the threephase motor, and six parallel coils are grouped into one phase for each side of the stator of the four-phase motor. Although the geometric dimensions between the three-phase motor and the four-phase motor are not very different, further investigation on the motor performance by the finite-element method becomes necessary for the final decision. V. FINITE-ELEMENT MAGNETIC ANALYSIS It is usually difficult to predict precisely the performance of the designed motor by the conventional magnetic circuit model. First, the proposed motor has a fractional number of slots per pole per phase, yielding additional assumptions and simplifications in the magnetic circuit analysis. Second, the nonlinear characteristics of magnetic materials produce saturation under overload currents in the condition when the electric vehicle is accelerating or climbing upslope. The finite-element analyses on the preliminary design prototype become necessary to provide detailed information on the magnetic flux and torque distribution, steady-state temperature distribution, and modal dynamics. These help designers investigate if additional performance specifications are satisfactory, otherwise iterative design and modification must be preceded. As opposed to the magnetic circuit analysis, the finite-element tool ANSOFT1 numerically calculates the magnetic field of the 3-D motor configuration. The finite-element mesh is
1ANSOFT

is a registered trademark of Ansoft Corporation.

YANG et al.: DESIGN AND CONTROL OF AXIAL-FLUX BRUSHLESS DC WHEEL MOTORS

1879

Fig. 19.

Detailed geometrical features of (a) stator tooth and (b) rotor.

(a)

(b)

Fig. 20.

(a) Three-phase and (b) four-phase torque patterns with respect to rotor shift.

automatically generated for the calculation of magnetic flux, flux density, and torque distributions. The boundary of the finite-element model is surrounded by air with enough thickness, whose magnetic permeability is much smaller than that of magnetic materials. First, the square-wave current excitation is assigned. The values of magnetic coenergy in the air gap of the motor are then calculated at equally spaced rotor shift angles, and their difference per unit angle determines the torque as shown in Fig. 20. As summarized in Table IV, for the maximum current of 4.53 A in the conductor due to the current density limit by its cross section, the maximum phase currents are, respectively, 72.4 and 54.3 A for three and four-phase motors, each with the average torque of 5.2 and 5.9 kg m. The maximum torques occur at 6.21 and 6.05 kg m, respectively, for the three- and four-phase motors. Therefore, the torque ripple of the four-phase motor generated by the reluctance components for the coils and magnets is 2.7%, which is much smaller than 18.9% of the three-phase motor. Moreover, the four-phase motor has a larger torque constant per phase current than the three-phase motor. The current distribution from the battery to multiple phases also relieves

TABLE IV MOTOR PERFORMANCE BY THE MAGNETIC ANALYSIS (SQUARE CURRENT WAVE)

the battery loading, thus increasing its life cycle. Better driving performance for the four-phase motor can be expected.

1880

IEEE TRANSACTIONS ON MAGNETICS, VOL. 40, NO. 4, JULY 2004

Fig. 22. Transient response of temperature in stator and rotor (diamond: stator teeth; square: outer case).

Fig. 21.

Steady-state temperature distribution.

VI. THERMAL AND MODAL ANALYSIS The thermal conduction analysis provided by COSMOS/M2 predicts the steady-state as well as the transient temperature distributions in the rotor, stator, and frame of the motor. Fig. 21 shows the steady-state temperature distribution of the wheel motor under a severe driving operation at 1000 rpm with a high phase current of 20 A, when the heat source comes from 35 W core loss and 350 W ohmic loss. In the steady state, the higher temperature holds near the coil at 123 C, which is below the upper temperature limit of 180 C for the insulation Class H. The lower temperature around 48 C happens at the frame surface, where the natural convection coefficient is set at 75 W/m K and the ambient temperature at 25 C. Also, the transient analysis in Fig. 22 reveals that it will take about 4 h for the motor to reach the steady-state temperature at the constant speed operation. For the normal operation of electric vehicles, the result of thermal analysis is quite conservative, therefore being satisfactory for the motor design. The natural frequencies of the motor are usually designed beyond the maximum speed 1700 rpm or 178 Hz to avoid undesirable motor resonance. In the modal analysis, the natural frequencies are examined for the rotor combined with the outer cases on which the tire is mounted. The boundaries are simply supported on the bearings of the motor shaft, while the bearings are assumed perfectly stiff. The first mode represents an axial vibration of the motor, and the second mode illustrates a torsional vibration along the axial direction. Each mode has a natural frequency of 261 and 293 Hz, respectively, and resonance may rarely occur below the maximum speed. The axial and torsional vibration modes are shown in Fig. 23. VII. PROTOTYPE AND CONTROL STRATEGY For a limited space of the wheel motor in the hub, the constraints on its outer diameter and thickness were specified for the optimization. Therefore, the dimensions of the stator
2COSMOS/M is a finite-element tool of Structural Research and Analysis Corporation.

components from the optimal design are similar between the three-phase and four-phase motors, as shown in Table III. The most apparent difference between the three-phase motor and the four-phase motor is the number of magnets on the rotor. That leads the final decision to fabricate the prototypes of threeand four-phase motors with the same stator, but different rotors with 16 and 18 magnets, respectively. Each magnet has the same size; hence, the only difference of the rotors is their pole pitch. The same stator with commutable rotors facilitates the manufacturing process and reduces the cost of the prototype, under allowable performance tolerance between the design and the prototype. The lamination material of 0.35 mm AISI M19 steel is selected for the stator teeth and yokes. The magnets of NdFeB 30SH are embedded in the rotor of the aluminum alloy 6061-T6. The shaft is made of stainless steel, and the cover of the motor is made of the aluminum alloy. The stator, rotor, and their assembly are provided in Fig. 24. Since the electric vehicles usually run in variable speed and load conditions, the control strategy must be made for high efficiency at high speed with enough torque for accelerating. To achieve this goal, the drive controller must have the following features. First, the motor is operated under an optimal driving pattern in terms of optimal current waveform for each independent phase. The square, triangular, or sinusoidal current patterns may not be the best for the dedicated design of the wheel motor. For most of the brushless dc motors, the maximum torque may be produced if the input current wave is proportional to the back-EMF. Fig. 25 shows the back-EMF of the prototype operated at 103 and 108 rpm, respectively, for the three- and four-phase wheel motors. The optimal current waveform can be obtained either by the minimization of the copper loss or by the maximization of the motor efficiency through the optimization process. VIII. SUMMARY AND CONCLUSION Dedicated disk-type axial-flux brushless three- and fourphase dc wheel motors have been successfully designed and fabricated. The sensitivity analysis with the magnetic circuit model provides an effective way to select the design parameters, which are iteratively tuned through the multiobjective optimal

YANG et al.: DESIGN AND CONTROL OF AXIAL-FLUX BRUSHLESS DC WHEEL MOTORS

1881

Fig. 23.

Axial (left) and torsional (right) vibration modes.

Fig. 24.

Stator, rotor, and motor assembly.

Fig. 25.

Back-EMF of the disk wheel motor: (a) three-phase and (b) four-phase.

design to maximize the output torque, efficiency, and torque density of the dedicated motor under prescribed constraints. A systematic procedure from magnetic circuit analyses to the finite-element modification and verification constitutes a complete design of the wheel motor. For square-wave current excitations, the four-phase motor presents satisfactory output torque with smaller ripple compared to the three-phase motor, and becomes a promising solution for the electric motorcycle of required specifications. Finally, the experiments on the

back-EMF of motor prototypes give information of optimal driving current patterns. APPENDIX A The distributions of the air-gap length , magnetomo, coenergy , and torque are fortive force mulated explicitly as functions of motor geometric and electric design variables.

1882

IEEE TRANSACTIONS ON MAGNETICS, VOL. 40, NO. 4, JULY 2004

Air-Gap Length: (A1) The effective air-gap length on the stator side is defined as [6]

nondominated points come to the ideal point is the family of matrices defined as

(B1) (A2) (A3) (A4) (A5) (A6) (A7) in which is the minimum air-gap length between the stator and the rotor, and is the slot width of the stator. Referring to Figs. 2 and 4, the air-gap length on the rotor side is a function of magnet fraction ; it is over the nonover the magnet material part, and is approximated by of the magnet, where its recoil permeability is . width Apparently, the distribution of the air-gap length is an explicit function of geometric design variables. Magnetomotive Force: (A8) The magnetomotive force produced by each stator , and is assumed to be equally distributed winding is is the number of turns on under the stator shoe, where each stator pole, and is the conductor current. Similarly, the of the magnet is approximated by magnetomotive force and equally distributed over each magnet, where is the coercivity of the magnet and is its thickness. Coenergy and Torque: In the optimization program, both the coenergy and torque are coded in discrete forms subject to , in which are weights, and are the minimum value and the worst value of the th objective function is the value of implementing the design varirespectively, is the feasible able with respect to the th objective, and design space. In the case of , all deviations from are taken into consideration in proportion to their magnitudes. For , the larger the deviation, the larger the weight in . In the limiting case of , only the largest deviation from the minimum objectives is considered. REFERENCES
[1] L. Chang, Comparison of AC drives for electric vehiclesA report on experts opinion survey, in IEEE Aerosp. Electron. Syst. Mag., Aug. 1994, pp. 710. [2] C. C. Chan, K. T. Chau, J. Z. Jiang, W. Xia, M. Zhu, and R. Zhang, Novel permanent magnet motor drives for electric vehicles, Int. Trans. Ind. Electron., vol. 43, no. 2, pp. 331338, 1996. [3] M. Terashima, T. Ashikaga, T. Mizuno, K. Natori, N. Fujiwara, and M. Yada, Novel motors and controllers for high-performance electric vehicle with four in-wheel motors, IEEE Trans. Ind. Electron., vol. 44, pp. 2837, Feb. 1997. [4] F. Caricchi, F. Crescimibini, O. Honorati, A. Di Napoli, and E. Santini, Compact wheel direct drive for EVs, IEEE Ind. Applicat. Mag., vol. 26, no. 6, pp. 2932, 1996. [5] F. Caricchi, F. Crescimibini, F. Mezzetti, and E. Santini, Multi-stage axial-flux PM machine for wheel direct drive, IEEE Trans. Ind. Applicat., vol. 32, pp. 882886, July/Aug. 1996. [6] D. C. Hanselman, Brushless Permanent-Magnet Motor Design. New York: McGraw-Hill, 1994. [7] Z. Zhang, F. Profumo, and A. Tenconi, Axial flux machine for electric vehicles, Elec. Mach. Power Syst., vol. 24, pp. 883896, 1996. [8] H. C. Lovatt, V. S. Ramsden, and B. C. Mecrow, Design of an in-wheel motor for a solar-powered electric vehicle, IEE Proc. Elec. Power Applicat., vol. 145, pp. 402408, Sept. 1998. [9] B. Hredzak, S. Gair, and J. F. Eastham, Elimination of torque pulsations in a direct drive EV wheel motor, IEEE Trans. Magn., vol. 32, pp. 50105012, Sept. 1996. [10] V. Ostovic, Computer-Aided Analysis of Electric Machines. Englewood Cliffs, NJ: Prentice-Hall, 1994. [11] C. H. Tseng, W. C. Liao, and T. C. Tang, MOST Users Manual, National Chiao-Tung University, Taiwan, R.O.C., Version 1.1, Tech. Rep. AODL-93-0, 1993. [12] M. Zeleny, Multiple Criteria Decision Making. New York: McGrawHill, 1982. [13] Y. P. Yang and C. H. Tseng, Multiobjective optimization of complex structures integrated with finite element software on workstations, J. Chin. Soc. Mech. Eng., vol. 16, no. 2, pp. 167179, 1995. [14] Y. P. Yang and C. C. Tu, Multiobjective optimization of hard disk suspension assemblies: Part IStructure design and sensitivity analysis, Comput. Struct., vol. 59, no. 4, pp. 757770, 1996.

(A9)

(A10) in which an electric period is divided by equally spaced apart while points, each with mechanical angle and . APPENDIX B The compromise programming [12][14] is to minimize the distance between the ideal solution and the optimal solution that is called the compromise solution. The distance measure used in compromise programming to evaluate how close the set of

Yee-Pien Yang (A92M02) was born in Taiwan, R.O.C., in 1957. He received the B.S. and M.S. degrees in mechanical engineering from National Cheng-Kung University, Taiwan, in 1979 and 1981, respectively, and the Ph.D. degree in mechanical, aerospace, and nuclear engineering from the University of California, Los Angeles, in 1988. He was an Associate Professor in the Department of Mechanical Engineering, National Taiwan University, from 1988 to 1996. He was promoted to Professor of Mechanical Engineering and led the Electromechanical System Research Group in 1996. His research interests are design and control of electromechanical systems, adaptive control of flexible structures, biomedical engineering, and assistive tool design for the disabled.

You might also like