3532 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 9, SEPTEMBER 2012 Design of a Five-Phase Brushless DC Motor for a Safety Critical Aerospace Application Xiaoyan Huang, Member, IEEE, Andrew Goodman, Member, IEEE, Chris Gerada, Member, IEEE, Youtong Fang, Member, IEEE, and Qinfen Lu, Member, IEEE Abstract—This paper describes a five-phase brushless dc (BLDC) motor designed for an electrohydrostatic actuation system (EHA) suited to the thin and optimized wings. The foundation of the design is a motor with fault tolerance and high reliability, compact structure, and low weight. The motor power rating is 12 kW at 12 000 rpm, and a “wet” form of construction is used where hydraulic oil is present in the motor to reduce the number of oil seals of the EHA for enhanced reliability and lifetime. The losses and thermal behavior are evaluated for an optimized design. Fault tolerance for BLDC motors is discussed. A five-phase motor has been manufactured, and test results are presented to validate the design. Index Terms—Brushless dc (BLDC) motor, electrohydrostatic actuation system (EHA), fault tolerant. I. I NTRODUCTION W ITH THE development of new composite materials thinner wings can be used for either improved aerodynamic efficiency or improved structural efficiency. This will however severely limit the installation space for flight control surface actuation systems. The limitation of space will make the current actuator assemblies infeasible. Since the more electric aircraft concept emerged, electrical actuation system has been intensively investigated due to advantages in terms of reduced weight, compact structure, easier maintenance, increased safety, and enhanced reliability [1], [2]. Proposed electrical solutions employ either an electrohydrostatic actuation system (EHA) or an electrical mechanical actuation system (EMA). The EMA uses mechanical gearing and a ball or roller screw to amplify the motor torque and provide linear actuation. This approach may lead to difficulty in realizing the safety requirements of a primary flight control surface where multiple Manuscript received November 3, 2010; revised April 9, 2011 and June 30, 2011; accepted September 23, 2011. Date of publication December 5, 2011; date of current version April 13, 2012. This work was supported in part by the U.K. Government under the DTI CARAD programme and by the National Natural Science Foundation of China under Grant 51007078. X. Huang was with the Power Electronics Machines and Control Group, Department of Electrical and Electronic Engineering, University of Nottingham, Nottingham NG7 2RD, U.K. She is now with the College of Electrical Engineering, Zhejiang University, Hangzhou 310027, China. A. Goodman and C. Gerada are with the Power Electronics Machines and Control Group, Department of Electrical and Electronic Engineering, University of Nottingham, Nottingham NG7 2RD, U.K. Y. Fang and Q. Lu are with the College of Electrical Engineering, Zhejiang University, Hangzhou 310027, China (e-mail: luqinfen@zju.edu.cn). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TIE.2011.2172170 actuators are used to ensure continued operation in the event of failure of one unit. When an EMA fails due to the mechanical gearing, the result can be a jammed actuator that may also prevent operation of the backup units. The EHA is a local, electrically driven hydraulic actuation system, which retains a similar function to the conventional hydraulic system. Where multiple EHA actuators are used in parallel on a flight control surface, failure of one unit can be accommodated using a simple hydraulic by-pass valve. This leaves the hydraulic ram of the failed unit free to move as the flight control surface moves under the control of the remaining healthy units. Thus, an EHA system is the basis of this study due to its fault tolerance, which is the key issue in aerospace applications. The aim of this paper is to design a fault-tolerant electrical motor for EHA which could be fitted in a thinner and optimized wing for better aerodynamics. In the past decades, switch reluctance (SR) motor, permanent magnet synchronous motor (PMSM), and brushless dc (BLDC) motor were investigated for aerospace applications [3], [4]. The SR motor is highlighted for safety critical applications due to its inherent high reliability. A 25-kW, four-phase, 8/6 SR has been designed and built by the University of Glasgow [5]. It was considered to offer the best compromise between performance, fault tolerance, and drive complexity for use in a large EMA system. A 22-kW, four-phase, 8/6 high power density SR drive for the MEA is reported in [6] which also discusses drives with different number of phases. The higher number of the phases, the more fault-tolerant the drive system is. However, the powerto-weight ratio reduces as the number of phases increases. Higher power density can be achieved with PMSM motors. Fault tolerance can be improved in these machines in a similar degree as in the SR motor drives [7]. A six-phase, eight-pole PMSM drive has been reported in [7] which can produce 29% more torque than its SR counterpart with high fault tolerance. A PMSM draws sinusoidal currents when supplied by sinusoidal voltages. This is an advantage of the PMSM over the SR motor, when a minimum current harmonic content is required [8]. PM motors with different structures had been designed and evaluated for fault tolerance and high reliability [9], [10]. The BLDC motor and PMSM share certain advantages and disadvantages. The BLDC motor has the advantage of a simpler control than the PMSM. No complex microprocessor is required, which to some degree could reduce the risk of failure. Multiphase BLDC motors with high power density and reliability for an EHA are reported [11]–[13]. The BLDC motor has the potential to provide the highest power density of any kind of motor because of the trapezoidal current supply 0278-0046/$26.00 © 2011 IEEE HUANG et al.: DESIGN OF A FIVE-PHASE BRUSHLESS DC MOTOR FOR A SAFETY CRITICAL AEROSPACE APPLICATION TABLE I M OTOR D ESIGN R EQUIREMENTS 3) 4) and back electromotive force (EMF). The disadvantage of the BLDC motor is the high input current harmonic content due to the nonsinusoidal back EMF and supply current. In safety critical applications, the fault tolerance and reliability of the whole electrical drive including the motor and power converter are equally of great importance. In most industrial applications, the type of the motor is selected first and followed by the power electronics and control strategy to reach various requirements. In this design, a different procedure is followed. A power converter with high reliability and simple control strategy is first determined. Then, the motor which suits the power converter and the control strategy with a specific power rating will be designed. From the power converter point of view, a matrix converter is highlighted for high reliability due to the absence of bulky capacitors which cannot operate reliably in the extremities of aerospace environmental conditions [14], [15]. The disadvantages are the higher risk of short circuits and more complicated commutation techniques which will lead to the use of a dedicated powerful microprocessor. In aerospace applications, the reliability of such a microprocessor can be a significant drawback. Therefore, the single-sided matrix converter (SSMC) which keeps the inherent advantages of a matrix converter but avoids the complex commutation problems is selected as the motor drive in this application; details are given in [16]. The SSMC can only provide unidirectional dc current. The BLDC and SR motors are all suitable candidates for use with the SSMC. Finally, the BLDC motor was selected over the SR motor due to its superior power density. In this paper, a BLDC motor will be designed to work with the SSMC and furthermore to meet the requirements of aerospace applications. 5) 6) 7) 3533 seen as limiting the diameter of the motor and limiting the space for the power electronics that is usually mounted close to the actuator. A compact structure is essential. “Wet” operation To minimize the number of sliding surface oil seals in the actuator for higher reliability and longer life time, the motor is immersed internally in Skydrol to avoid shaft seals. Viscous losses Skydrol produces a hydraulic power loss as it circulates through the relatively short air gap, and this loss must be minimized by controlling the rotor dimensions. The loss is greatest at high speed. Maximum temperature limit The maximum operating temperature for Skydrol is 120 ◦ C. Fast response For most of the time, the actuator is at the required position, and the motor is running at a low speed (2000 rpm) with minimum load so that the oil seals are continuously lubricated so keeping wear to a minimum and extending operational life. When the new position reference signal arrives, a higher motor power at high speed (12 000 rpm) will be produced and then transmitted by the fluid flow to the actuator to change the position until it reaches the required position. The motor should have a low inertia to respond rapidly to changes in load accelerating from 2000 to 12 000 rpm in about 0.5 s. The high-speed operation lasts only for as long as the high load is required before the motor returns to the low-speed condition. Accommodation to be driven from a SSMC The SSMC only provides a positive current half cycle in each stator coil. If the same design of motor were fed from a conventional converter, then it is obvious that for a given torque level, the peak current level would be half that of the SSMC fed machine. Correspondingly, the power loss would also be half that of the SSMC machine. The design of the motor is therefore required to accommodate the special form of SSMC power electronics which provides greater inherent commutation and control reliability for the converter and thus the complete drive system. II. R EQUIREMENTS FOR M OTOR D ESIGN The design objectives of the motor need to be identified before starting the motor design. The conventional design criteria for the complete drive are listed in Table I which includes the power supply, electrical performances, and mechanical performances. The key problems of the motor design are: 1) Fault tolerance and high reliability Fault tolerance and high reliability are of great importance in aerospace applications. A multiphase motor and drive will be designed toward this end. 2) High power density High power (torque)-to-weight (volume) ratio is a key criterion to save space for aerospace applications. The effect of the reduction of installation space on motor is III. M OTOR D ESIGN AND O PTIMIZATION In this paper, a typical motor design procedure is followed. First, the main dimensions are determined by the power rating, the empirical current loading, and magnetic loading in this power range. Then, the stator lamination dimensions can be calculated based on the specific number of the phases and poles, followed by the design of the rotor structure. Finally, electrical power losses and hydraulic losses are then evaluated. The temperature rise of the motor is investigated using Motorcad. A. Main Dimensions The initial design of the BLDC includes the determination of the main dimensions including stator inner diameter and 3534 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 9, SEPTEMBER 2012 Fig. 1. Comparison of output torques for motors with different number of phases under healthy and faulty conditions. effective length of the rotor. It should be noted that for the specific flux density, electric and magnetic loading, and speed, the output power is proportional to rotor volume. The choice of stack length and rotor diameter was influenced by the low inertia, the hydraulic loss, and the critical shaft speed. A small rotor diameter is required to ensure the low inertia and then the fast response. Hydraulic loss increases rapidly with diameter of the rotor and directly with its length. The hydraulic loss can easily dominate all other losses in the machine at 12 000 rpm. Therefore, the air gap length has a minimum value. Thus, magnet thickness is mostly defined and consequently is the diameter of the solid rotor shaft and rotor core. This in turn restricts the minimum rotor diameter both through increasing inter pole magnetic leakage flux and through the shaft critical speed. As the rotor diameter reduces, the core length and the distance between the bearing centers increase with a consequent reduction in the shaft critical speed. B. Number of Phases and Poles Since fault tolerance and high reliability are the foundations of the design, the number of phases should therefore be carefully selected. Generally, motors with a high number of phases have the potential advantage of higher fault tolerance and consequently reliability [17]–[19]. Finite-element method models of different phase BLDC motors were built to investigate their respective performances. In this section, the inner diameter, stack length, the rotor structure, and the rotating speed of multiphase BLDC motors are kept constant. Only the stator structures are affected when considering different numbers of phases. The comparison is based on the assumption that the same quantity of copper is used. The number of turns varies with the number of phases to keep the amount of copper constant. Fig. 1 shows the output torques for motors with different number of phases and unidirectional supply currents under healthy and faulty conditions. It can be seen that at least two phases should be conducting at any time for improved reliability of the drive. This allows a phase malfunction while still maintaining some drive functionality as there will always be at least one phase conducting current. The requirement for two conducting phases in an N-phase motor would imply that each phase current would conduct for 2/N of a cycle. Considering a four-phase motor, the current pulse conduction angle needs to be 180◦ , leading to large torque ripple. For five- and sixphase machines, the conduction angles are 144◦ and 120◦ , respectively, which are confirmed by the curve shape of back EMFs. The torque ripple can therefore be reduced significantly. The disadvantage of the five and six phase is of course the extra power devices. Compared to the six-phase drive, the torque produced by the five-phase drive is 2.98%, 4.19% less respectively at healthy and faulty condition. However, 20% extra power electronic components are required. Since the motor is fed from SSMC, the BLDC performance under the fault condition with one IGBT of the SSMC shortcircuit need to be investigated. The conduction period of current in the faulty phase is longer than the healthy phase which results in a large transient torque at the time. The torque can be double the rated torque if the faulty phase current cannot be switched off on time. The transient torque for all phase number drives rises by approximately 50% under this faulty condition. From the motor design point of view, the one IGBT short-circuit fault condition has similar impact on the BLDC motors with different number of phases. As a result, the five-phase BLDC motor was finally chosen as a compromise between fault tolerance, weight, and volume. Another important parameter that needs to be taken into consideration is the number of poles. The higher the number of poles, the smaller is the stator outer diameter for a given rotor diameter. Furthermore, from the motor point of view, the copper loss is reduced when increasing the number of poles, because the end windings become shorter. However, increasing the number of poles will lead to a higher basic switching frequency and hence an increase in iron losses. This increased loss is however mitigated due to the reduced stator back iron depth and thus reduces to some extent the iron losses. The risk of demagnetization for the PMs also reduces with the increasing number of poles. From the power converter point of view, a higher switching frequency is required for a higher number of poles. In this project, the switching frequency capability of the IGBTs on the power board is limited which can be easily reached if a large number of poles applied due to the low inductance of the BLDC. A pole number of four was a reasonable compromise. C. Stator Lamination According to the steel datasheet, the steel started to saturate severely around 1.8 T. As a result, the flux densities in stator teeth and air gap are 1.8 and 0.75 T, respectively. For specific flux density required in air gap and stator teeth, the width of stator teeth was determined by stator inner diameter and the number of slots. The slot area was determined by the current density, number of turns, and slot fill factor. In aerospace application, more space in the slots needs to be left for the insulation material for higher reliability. Consequently, a conservative slot fill factor was used. The width of stator back iron is one of the key factors determining the iron losses. The deeper the stator back iron is, the heavier the total stator laminations are, and in turn, the more HUANG et al.: DESIGN OF A FIVE-PHASE BRUSHLESS DC MOTOR FOR A SAFETY CRITICAL AEROSPACE APPLICATION 3535 Fig. 4. Peak cogging torque with various magnet arcs. Fig. 2. Flux density in the air gap with different magnet thickness. Fig. 3. Shapes of the PMs. Fig. 5. Flux density in the motor no-load (Ia = 0 A). iron losses. Contradictorily, the wider the stator back iron is, the smaller the flux density is in the back iron. The width of the back iron was limited to control the iron losses in the machine. D. Permanent Magnet Design The PM thickness is a significant parameter since it affects the flux density in the air gap, the demagnetization withstand capability, and the cost. A special feature of this BLDC motor is the “wet” operation. Therefore, a large air gap will be used to minimize viscous losses. The thickness of magnet is expected to be greater than traditional designs. The simulations are carried out with PMs of various thickness based on the same stator lamination dimensions and the rotor structure as shown in Fig. 2. It can be seen that flux density in the air gap (Bg ) increases as the thickness of the PMs increases. However, it reaches to the maximum 0.75 T at 8 mm before the core starts to saturate and interpole leakage becomes substantial. This confirms 8 mm as the maximum sensible value. A surface-mounted PM was used in this design due to higher power density, which was slightly modified to an octagon to prevent sliding between the rotor surface and the PMs as shown in Fig. 3. The arc of PMs is also optimized to minimize the torque ripple. The peak cogging torque with different magnet arcs is shown in Fig. 4. It can be seen that the peak cogging torque is relatively small, and therefore the magnets are not skewed in this application. The 75◦ magnet pole pitch produced the least cogging torque. However, the flat top area of the back EMF waveform reduces as the arc reduces. Large torque pulsation will occur during commutation when the arc of the PMs decreases, which will also reduce the average output torque. Fig. 6. Flux density in the motor full-load (Ia = 40 A). Consequently, an arc of 80◦ is a good compromise between the cogging torque ripple and the commutation torque ripple. IV. BLDC P ERFORMANCE A. Motor Operation FEM model of the BLDC motor is built. The flux densities in the motor under no-load and full-load conditions are shown in Figs. 5 and 6, respectively, which verify the motor design. The PMs can operate reliably when the current is less than 120 A without demagnetization according to the Fig. 7. If the motor winding has a turn-turn short circuit, the pulse of current may be greater than this level. Therefore, demagnetization might still be possible in this condition; however, the likelihood of it happening did not compromise the required system reliability level. However, the PMs have the potential risk of demagnetization which may result in the motor failure. Investigation of the resistance to demagnetization with armature reaction is thus of great importance to ensure high reliability. 3536 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 9, SEPTEMBER 2012 Fig. 7. Flux density in the motor when Ia = 120 A. Fig. 9. Stator iron loss versus speed. tions (12 000 rpm), respectively. However, the loss would be increased if the winding temperature rises above that which is predicted. B. Iron Losses Fig. 8. Flux density in the motor at 150 ◦ C. B. Motor Operation at High Temperatures Samarium cobalt is selected due to its low temperature coefficient. FEM simulations under no-load condition were performed to predict the performance of the motor at 150 ◦ C, as shown in Fig. 8. The peak value of flux density reduces from 1.57 T at 20 ◦ C to 1.49 T at 150 ◦ C. The average back EMF reduces from 28.72 to 28.0 V. The torque is proportional to the back EMF at the same speed and armature current, and therefore it reduces by 2.57%, which is within the tolerance range for the design and is easily accommodated with a small increase in phase current. V. L OSS E VALUATIONS The thermal design is critical for this application to get a compact structure. The losses need to be thoroughly evaluated for the thermal analysis [20]. The main losses in a BLDC motor consist of the copper, iron, mechanical losses, and stray load loss. The copper loss can be limited by reducing the current or the resistance of the winding. The resistance can be reduced by using conductors with large diameter and operating under cool temperatures. Clearly, large slot area is required as well as large machine size. A higher Bg is an advantage so that the armature current is reduced to deliver the same torque level. The increased Bg will lead to extra iron losses. The compromise between size and losses (copper and iron losses) needs to be made for best motor performance. A. Copper Loss For the ideal square wave current supply, the copper loss can be calculated simply by I 2 · R. The approximate resistance is estimated to be 0.121 Ω. The copper loss will be 6.05 and 387.2 W at minimum load (2000 rpm) and full-load condi- The iron losses in the BLDC include the iron losses in stator, rotor, shaft, retaining can, and the magnets. It is difficult to calculate the iron losses accurately. Several improved iron losses prediction methods have been presented based on the analytic or FEM method in the literature which involves complex magnetic field calculations [21]–[25]. The other method is to calculate the iron losses based on empirical formula using the empirical loss data provided by the manufacturer. This was applied in this design owing to its simplicity. The stator and rotor peak flux densities used were derived from the FEM calculations. The stator iron losses at different rotating speeds are shown in Fig. 9. It should be pointed out that the method used above is for sinusoidal current excitation. The BLDC produces nonsinusoidal current waveforms, and the stator winding is fed from the switched AC source from the SSMC, which will bring high frequency harmonics into the stator flux. As a result, there is expected to be modest correlation between the estimated value using the classic method and the measured iron losses. C. Retaining Can Eddy Current Losses The retaining can has generally low electric resistance, and the loss could be significant. Strength and high resistance are the important criteria in selecting the retaining can material. The losses in the can are derived from FEM software as shown in Figs. 10 and 11. During most of the operation time, the losses are small because the flux variation in the can is slight. The greatest losses occur during the armature current commutation, where the rapid change of the current in armature winding leads to a rapid flux variation in the retaining can. Under this condition, the can losses cannot be ignored. The average losses in one electrical cycle are 1.5 W and 81.7 W, respectively, under 2000 rpm and 12 000 rpm full-load conditions. The losses in the PMs are similar to those in the can. However, the relative longer distance from the stator tooth and HUANG et al.: DESIGN OF A FIVE-PHASE BRUSHLESS DC MOTOR FOR A SAFETY CRITICAL AEROSPACE APPLICATION 3537 TABLE II L OSSES S UMMARY Fig. 10. Retaining can eddy current loss at 2000 rpm. Fig. 12. Temperature rises of the components. Fig. 11. Retaining can eddy current loss at 12 000 rpm. the shielding due to the eddy currents generated in the can make these losses smaller.The PMs are made into segments to reduce these losses. The average losses in one electrical cycle are 0.6 W and 22.8 W, respectively, under 2000 rpm and 12 000 rpm fullload conditions. D. Rotor and Shaft Losses Owing to the low rate of flux cutting, the iron losses in the rotor and shaft have a similar waveform to those of the retaining can and the PMs, but the magnitudes are much smaller. The average losses in one electrical cycle are 0.029 W and 3.497 W, respectively, at 2000 rpm and 12 000 rpm, which are less than 1% of the total losses. E. Viscous Loss The Skydrol hydraulic fluid in contact with the rotor surface generates a viscous drag loss, which can be calculated using the formula P = 0.5 · π · ρ · ω 3 · Rr4 · l · Cm (1) where ρ is the density of the fluid, Rr is the rotor radius, ω is the angular velocity, Cm is the friction coefficient. Cm can be expressed as (2), when the fluid is turbulent, with a high Reynolds number [26] Cm = 0.065(lg /Rr )0.3 · Re−0.2 ρ · · R r · lg Re = μ here, lg is the air gap length and μ is the fluid viscosity. (2) (3) Consequently, the viscous loss is 6 W and 817 W at 2000 rpm and 12 000 rpm, respectively. The losses are summarized in the Table II. VI. T HERMAL D ESIGN The power-to-weight/volume ratio is mainly restrained by the operating environment temperature and the cooling method once the power rate is determined. The thermal performance of the motor should be carefully examined to maximize the ratio. In this design, the weight/volume is compromised by the requirement of fault tolerance and hydraulic losses. A Motorcad thermal model was built based on the motor dimensions and the losses mentioned above. The duty cycle of the motor varies with the actuation cycle, which cannot be predicted precisely. In this design, the maximum operation period of 60 s is expected at high speed under full-load conditions. The simulation considered the worst case scenario. That means 180 s per cycle; the motor started from 2000 rpm at 0 ∼ 20 s, then accelerated to 12 000 rpm in 0.5 s and maintained speed for 60 s, decelerated to 2000 rpm and maintained speed for another 100 s. The ambient temperature was set to 70 ◦ C at the beginning. The temperature rise waveforms of the components including stator winding and housing are shown in Fig. 12. The maximum temperature rise occurred at the rotor surface because of the most significant viscous loss. All the components were operated within the temperature limit, which verifies the thermal design of the motor. It can be seen that the maximum temperature designed for is around 100 ◦ C which is less than the temperature limits of the components used. The margin was left by the design for three 3538 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 9, SEPTEMBER 2012 TABLE III M EASURED R ESISTANCES AND I NDUCTANCE Fig. 13. BLDC motor prototype. (a) Stator. (b) Shaft, rotor with magnets. (c) Housing and end plates. (d) Assembled BLDC motor. reasons. The first reason is the actual losses may be greater than the predicted value due to the difficulty in predicting the iron losses accurately. Furthermore, it can be noted that the temperature of the Skydrol was set to 70 ◦ C during the simulations; however, in reality, a higher temperature than the ambient temperature is expected as this will carry the heat generated in the pump. The third reason is that the heat dissipation is greatly dependent on the leakage flow rate of the Skydrol from pump to the motor. The flow rate could not be precisely predicted at this stage of the design as it relies partly on the high pressure leakage past the pump pistons. Consequently, the maximum temperature of 100 ◦ C was expected as a worst case scenario. Maximum pump loss allowed in this design is 1197 W. Fig. 14. Five-phase back EMF waveforms at 2000 rpm. Skydrol oil, the radial fins are designed to accommodate the hydraulic system losses as well as the motor losses as shown in Fig. 13(d). VIII. T ESTING R ESULTS A. Results to Verify the Electromagnetic Design VII. M OTOR C ONSTRUCTION The stator stack is built from 0.35 mm thick laminations as shown in Fig. 13(a). The ideal lamination material should have low magnetic reluctance but high electrical resistance. The single layer winding is employed for higher reliability and smaller magnetic coupling between phases. Particular attention should be given to the insulation material, which must be Skydrol proof. Owing to the large effective air gap length in the design, a solid rotor could be employed for better mechanical integrity. The material used for the solid rotor is EN24 steel. The assembled rotor, shaft, the magnet segments, and the retaining can are shown in Fig. 13(b). The 1-mm retaining can is designed to withstand the centrifugal force of the magnet segments at high speeds. The material used for the can is nonmagnetic stainless steel. All the remaining components including the housing, two end plates, and the resolver housing are shown in Fig. 13(c). A special feature of the BLDC is the cold Skydrol oil fed from the pump along the shaft through ten holes in the housing and back to the reservoir. Owing to the method of cooling with the The first step of the test was measuring the five-phase resistances and inductance which were recorded in Table III. The resistances are 17.7% higher than the predicted value because the resistance in the end winding is difficult to be predicted accurately as well as the slot fill factor until the coils have been assembled in the motor due to the practical limitations of winding the machine. The end windings are actually longer than anticipated as can be seen in Fig. 13(a). The average inductance of the five-phase windings is 8.13% less than the predicted value. The inductances varied slightly with the rotor position because of the interaction between slot opening and the gap between the PMs. Consequently, the inductances are slightly different from each other at any rotor position. The measured resistances and inductances indicated that the windings were correctly constructed. The open circuit test was done at 2000 rpm. The measured five-phase back EMF waveforms at 2000 rpm were presented in Fig. 14 which verified that the designed flux density in the air gap was reached and the windings were correctly constructed. The data were recorded using two synchronized scopes. Each scope has a maximum of four channels. Therefore, HUANG et al.: DESIGN OF A FIVE-PHASE BRUSHLESS DC MOTOR FOR A SAFETY CRITICAL AEROSPACE APPLICATION 3539 Fig. 16. Comparison of tested and predicted viscous loss. Fig. 15. Tested iron losses, windage, and friction losses. two synchronized scopes are required to record the correct back EMF sequence. There should be two pulses in one mechanical revolution due to the two pole pairs in the BLDC motor. It can be seen from Fig. 14, there are two pulses in 30 ms which conformed to the 2000-rpm operation. The average value of the flat top area is 27.5 V. Compared to the simulation result of 28.7 V, it is 4.12% less. The reason is the unexpected large voltage dip in the middle of the flat top area which is caused by the gaps between the PMs and the slot opening. This voltage dip also introduces undesirable torque ripple. B. Motor Losses Determination 1) Iron Losses (Stator Iron, Retaining Can, Rotor and Shaft, and the PM Losses): It is difficult to separate the stator iron loss with the retaining can, rotor, shaft, and the PMs losses because they are all generated by the same source. During the test, they were all considered as one iron loss, which can be obtained from the open circuit test results. In the BLDC motor, the flux density is mainly determined by the PMs. Thus, once the motor was constructed, the iron losses mainly varied with the speed of the motor. Consequently, the iron losses cannot be separated from the friction and windage losses as in an induction motor. The mechanical power of the BLDC motor driven by the AC motor at different speeds without the Skydrol immersion was recorded as shown in Fig. 15. These included the iron losses, friction, and windage losses. The losses, as expected, increase more rapidly as the speed increases. It can be seen from the figure that the tested loss is greater than the predicted loss. This is mainly because of the difficulty in predicting the stator core losses accurately. The rotor and shaft losses are very small compared to other losses as simulated in the FEM software. However, the predicted losses do not include the bearing losses, which is not insignificant. Overall, therefore, the agreement is good between prediction and measurement. It should be noted that the rotor and shaft losses, the retaining can, and the PM losses both as tested here and as predicted are lower than for the fully loaded situation because there is no commutation of the current during the test. However, from the simulated results described above, large power loss pulses during commutation would be expected at commutation, which will increase the average losses above those measured. 2) Fluid Viscous Loss: The mechanical input power in the open circuit test with Skydrol immersion includes the viscous loss, iron losses, and friction and windage losses. As a consequence, the viscous loss could be found by subtracting noload mechanical losses for the nonimmersed condition from the immersed condition. The results were compared to the predicted viscous hydraulic loss in Fig. 16. The tested losses are greater than the predicted losses because the viscosity of the Skydrol is sensitive to the temperature. The viscosity at 20 ◦ C was used in the predicted losses. The actual room temperature during the test was 17.2 ◦ C which will increase the viscosity of the fluid and lead to a larger viscous loss. Furthermore, the measured power loss should be greater than the predicted loss because the prediction only deals with fluid in the air gap and ignores the ends of the rotor, the shaft, and the loss in the Skydrol flooded bearings. Overall, the predicted hydraulic loss can be seen to be in good agreement with predictions. 3) Thermal Design: It is clear, however, that loss prediction is in agreement with experimental results though in all cases the losses are underestimated. The tendency of the losses to increase with the speed was as expected. Most significant is the extra stator conductor loss on full load. However, a good safety margin was left for the thermal design. The actual maximum temperature rise predicted was 30 ◦ C with a 20 ◦ C margin left. The new actual power loss estimates suggest that 7 ◦ C to 8 ◦ C of the 20 ◦ C margin is consumed in the extra loss which still leaves a safety margin for underestimation of temperatures, cooling, and pump losses. It can be therefore seen that the BLDC motor is capable of providing the designed performance. IX. C ONCLUSION A 12-kW, 12 000-rpm, five-phase, four-pole BLDC motor was designed and manufactured for the EHA application. The performance of the motor was determined by magnetic circuit analysis and then FEM simulation. Although the motor performance in terms of torque per unit volume is sacrificed partly for the aerospace reliability requirements and partly for the SSMC limitations, the motor provides with the SSMC a highly faulttolerant drive system. Important design features based on the FEM simulation results include the 80◦ arc PMs. These are a 3540 IEEE TRANSACTIONS ON INDUSTRIAL ELECTRONICS, VOL. 59, NO. 9, SEPTEMBER 2012 compromise between considerations of cogging torque and the commutation torque ripple. The losses of the motor were investigated individually and controlled. At high speed, the viscous loss is the most significant loss and degrades the motor efficiency. The thermal design of the motor has been developed using Motorcad. Even when the motor operates at the maximum ambient temperature, the steady-state temperature will be within the limit set by the Skydrol. It should be noted that the motor dimensions have been optimized, but the size and volume of the housing particularly the cooling fins turned out to be a large proportion of the whole assembled motor. Further research into improved heat dissipation and more accurate estimation of the hydraulic pump losses considering the sizes of the heat dissipation components would be useful. [17] [18] [19] [20] [21] [22] [23] R EFERENCES [1] R. M. Crowder, “Electrically powered actuation for civil aircraft,” in Proc. IEE Colloq., Actuator Technol.: Current Pract. New Develop., 1996, pp. 5/1–5/3. [2] S. L. Botten, C. R. Whitley, and A. D. King, “Flight control actuation technology for next-generation all-electric aircraft,” Technol. Rev. J., vol. 23, no. 6, pp. 55–68, 2000. [3] M. E. Elbuluk and M. D. Kankam, “Motor drive technologies for the power-by-wire (PBW) program: Options, trends and tradeoffs. I. Motors and controllers,” IEEE Aerosp. Electron. Syst. Mag., vol. 10, no. 11, pp. 37–42, Nov. 1995. [4] A. Boglietti, A. Cavagnino, A. Tenconi, and S. Vaschetto, “The safety critical electric machines and drives in the more electric aircraft: A survey,” in Conf. Rec. IEEE IECON, 2009, pp. 2587–2594. [5] C. Cossar, L. Kelly, T. J. E. Miller, C. Whitley, C. Maxwell, and D. Moorhouse, “The design of a switched reluctance drive for aircraft flight control surface actuation,” in Proc. IEE Colloq. Elect. Mach. Syst. More Elect. Aircr., 1999, pp. 2/1–2/8. [6] R. Krishnan, D. Blanding, A. Bhanot, A. M. Staley, and N. S. Lobo, “High reliability SRM drive system for aerospace applications,” in Conf. Rec. IEEE IECON, 2003, vol. 2, pp. 1110–1115. [7] A. G. Jack, B. C. Mecrow, and J. A. Haylock, “A comparative study of permanent magnet and switched reluctance motors for high-performance fault-tolerant applications,” IEEE Trans. Ind. Appl., vol. 32, no. 4, pp. 889–895, Jul./Aug. 1996. [8] P. A. Robson, K. J. Bradley, P. Wheeler, J. Clare, L. de Lillo, C. Gerada, S. J. Pickering, D. Lampard, C. K. Goh, G. Towers, and C. Whitley, “The impact of matrix converter technology on motor design for an integrated flight control surface actuation system,” in Conf. Rec. IEMDC, 2003, vol. 2, pp. 1321–1327. [9] B. Vaseghi, N. Takorabet, J. P. Caron, B. Nahid-Mobarakeh, F. MeibodyTabar, and G. Humbert, “Study of different architectures of fault-tolerant actuator using a two-channel PM motor,” IEEE Trans. Ind. Appl., vol. 47, no. 1, pp. 47–54, Jan./Feb. 2011. [10] C. Gerada and K. J. Bradley, “Integrated PM machine design for an aircraft EMA,” IEEE Trans. Ind. Electron., vol. 55, no. 9, pp. 3300–3306, Sep. 2008. [11] P. M. Churn, C. J. Maxwell, N. Schofield, D. Howe, and D. J. Powell, “Electro-hydraulic actuation of primary flight control surfaces,” in Proc. IEE Colloq. All Elect. Aircr., 1998, pp. 3/1–3/5. [12] M. Villani, M. Tursini, G. Fabri, and L. Castellini, “Multi-phase fault tolerant drives for aircraft applications,” in Proc. ESARS, 2010, pp. 1–6. [13] J. W. Bennett, B. C. Mecrow, A. G. Jack, and D. J. Atkinson, “A prototype electrical actuator for aircraft flaps,” IEEE Trans. Ind. Appl., vol. 46, no. 3, pp. 915–921, May/Jun. 2010. [14] P. W. Wheeler, L. Empringham, M. Apap, L. de Lilo, J. C. Clare, and K. J. Bradley, “A matrix converter motor drive for an aircraft actuation system,” in Proc. Power Electron. Appl., EPE Conf., 2003, pp. 472–481. [15] M. Aten, G. Towers, C. Whitley, P. Wheeler, J. Clare, and K. Bradley, “Reliability comparison of matrix and other converter topologies,” IEEE Trans. Aerosp. Electron. Syst., vol. 42, no. 3, pp. 867–875, Jul. 2006. [16] X. Huang, K. Bradley, A. Goodman, C. Gerada, P. Wheeler, J. Clare, and C. Whitley, “Fault-tolerant brushless DC motor drive for electro hydro- [24] [25] [26] static actuation system in aerospace application,” in Conf. Rec. IEEE IAS Annu. Meeting, Oct. 2006, vol. 1, pp. 473–480. T. Gopalarathnam, H. A. Toliyat, and J. C. Moreira, “Multi-phase faulttolerant brushless DC motor drives,” in Conf. Rec. IEEE IAS Annu. Meeting, 2000, vol. 3, pp. 1683–1688. R. Kiani-Nezhad, B. Nehid-Mobarakeh, L. Baghli, F. Betin, and G.-A. Capolino, “Modeling and control of six-phase symmetrical induction machine under fault condition due to open phases,” IEEE Trans. Ind. Electron., vol. 55, no. 5, pp. 1966–1977, May 2008. E. Levi, “Multiphase electric machines for variable-speed applications,” IEEE Trans. Ind. Electron., vol. 55, no. 5, pp. 1893–1909, May 2008. A. Boglietti, A. Cavagnino, D. Staton, M. Shanel, M. Mueller, and C. Mejuto, “Evolution and modern approaches for thermal analysis of electrical machines,” IEEE Trans. Ind. Electron., vol. 56, no. 3, pp. 871– 882, Mar. 2009. A. Cassat, C. Espanet, and N. Wavre, “BLDC motor stator and rotor iron losses and thermal behavior based on lumped schemes and 3-D FEM analysis,” IEEE Trans. Ind. Appl., vol. 39, no. 5, pp. 1314–1322, Sep./Oct. 2003. F. Deng, “An improved iron loss estimation for permanent magnet brushless machines,” IEEE Trans. Energy Convers., vol. 14, no. 4, pp. 1391– 1395, Dec. 1999. K. Atallah, Z. Q. Zhu, and D. Howe, “An improved method for predicting iron losses in brushless permanent magnet dc drives,” IEEE Trans. Magn., vol. 28, no. 5, pp. 2997–2999, Sep. 1992. Z. Gmyrek, A. Boglietti, and A. Cavagnino, “Estimation of iron losses in induction motors: Calculation method, results, and analysis,” IEEE Trans. Ind. Electron., vol. 57, no. 1, pp. 161–171, Jan. 2010. K. Yamazaki and H. Ishigami, “Rotor-shape optimization of interiorpermanent-magnet motors to reduce harmonic iron losses,” IEEE Trans. Ind. Electron., vol. 57, no. 1, pp. 61–69, Jan. 2010. E. Bilgen and E. Boulos, “Functional dependence of torque coefficient of coaxial cylinders gap width and Reynolds numbers,” Trans. ASME, J. Fluids Eng., vol. 95, no. 1, pp. 122–126, Mar. 1973. Xiaoyan Huang (M’09) received the B.E. degree, from Zhejiang University, Hangzhou, China, in 2003, and the Ph.D. degree in electrical machines and drives from the University of Nottingham, Nottingham, U.K., in 2008. From 2008 to 2009, she was a Research Fellow with the University of Nottingham. Currently, she is a Lecturer with the College of Electrical Engineering, Zhejiang University, where she is working on electrical machines and drives. Her research interests are permanent magnet machines and drives for aerospace and traction applications, and generator system for urban networks. Andrew Goodman (M’09) received the M.S. and Ph.D. degrees in electrical and electronic engineering from the University of Nottingham, Nottingham, U.K., in 2002 and 2007, respectively. He is currently a Research Fellow in the Power Electronics, Machines, and Control Group at the University of Nottingham. His research interests include grid interface converters and motor drives. Chris Gerada (M’05) received the Ph.D. degree in numerical modelling of electrical machines from the University of Nottingham, Nottingham, U.K., in 2005. He subsequently worked as a Researcher at Nottingham on high performance electrical drives and on the design and modeling of electromagnetic actuators for aerospace applications. He is currently an Associate Professor in Electrical Machines within the Power Electronics Machine and Control research group at Nottingham. He is also the Project Manager of the GE Aviation Strategic Partnership, and his research interests include high-performance electric drives and machines. HUANG et al.: DESIGN OF A FIVE-PHASE BRUSHLESS DC MOTOR FOR A SAFETY CRITICAL AEROSPACE APPLICATION Youtong Fang (M’11) received the B.S. degree and Ph.D. degree in electrical engineering from Hebei University of Technology, Hebei, China, in 1984 and 2001, respectively. Currently, he is a professor with the College of Electrical Engineering, Zhejiang University, Hangzhou, China. His research interests include the application, control, and design of electrical machines. 3541 Qinfen Lu (M’10) received the B.E., M.E., and Ph.D. degrees from Zhejiang University, Hangzhou, China, in 1996, 1999, and 2005, respectively. Since 1999, she has been with the College of Electrical Engineering, Zhejiang University, where she is currently an Associate Professor. Her research interests include analysis and control of linear machines, permanent magnet machines, etc.