Halbach Array for an In-Wheel Traction Motor I. Martinez Ocaña1, N. J. Baker1, B. C. Mecrow1, C. Hilton2, S. Brockway2 Abstract 1 - This paper compares a Halbach array to a more conventional surface mounted permanent magnet arrangement for an in wheel traction motor. Demagnetisation, physical air-gap and torque capability at continuous and overload operation is considered. Without increasing the machine operating space, using one transition per pole was found to increase rated torque by 5% and overload torque by 3%. A second comparison was done using 3 transitions per pole and this topology was found to give an increase in the overload torque of 4%. Finally, a novel trapezoidal Halbach array arrangement has been proposed to reduce the demagnetisation of the transition magnets and increase their permeance coefficient. performance [4]. This paper considers replacing the existing surface mounted rotor topology with permanent magnets arranged in a Halbach array. Halbach arrays have previously been proposed as traction motors for a solar powered electric vehicle in [5], [6] and [7] and for aerospace traction in [8]. Both design and demagnetisation of the permanent magnets is studied in the latter. This paper will investigate a rotor design using Halbach array and compare predicted torque capability against the surface mounted Reference Machine (RM). In addition the effect of physical air-gap and demagnetization of the topologies will be discussed. Index Terms – Halbach array, In-wheel, Permanent magnets, trapezoidal. T I. II. THE REFERENCE MACHINE The reference design shown in Fig. 2 is an existing outer rotor surface mounted permanent magnet machine [1]. This machine has been manufactured and extensively tested both on dynamometers and in road vehicles. Hence it is well characterised. Per-unit values of torque are scaled to this machine. The Reference Machine produces very high torque density, but it is insightful to investigate whether other topologies can be even more torque dense. Two critical points of operation have been chosen for design comparison: continuous steady state and short term overload. Dimensions of the proposed machines are constrained by the stator inner diameter and rotor outer diameter of the actual Reference Machine as well as the overall stack length – all fixed by the wheel size. The stator design of the proposed topologies has not been altered and only changes in the rotor design are considered here. Design constraints are shown in Table 1. INTRODUCTION he development of in-wheel motors has increased in recent years as a way of providing direct drive transmission in electric vehicles. By building the motor directly in the wheel, classical automotive components such as drive shafts, gears and differentials can be eliminated with a potential increase in the reliability of the system, see Fig. 1. Although the in-wheel motor increases the vehicle unsprung mass, this causes minimal steering and handling issues if the suspension system is designed to suit, whilst the removal of other components gives overall efficiency, weight and complexity gains. Control using true torque vectoring at each wheel, electronic differential, traction control and more efficient regeneration braking are inherent and usually software controlled. In addition, the integration of motor and inverter into the wheel also frees yet more space in the vehicle to be used in other ways. TABLE 1 DESIGN CONSTRAINS Fig. 1: 320 Discontinuous DC voltage supply [V] 400 Base speed [rpm] Top speed [rpm] 800 1600 The Reference Machine operates at a nominal speed of 800 rpm with a maximum speed of 1600 rpm. Maximum torque at overload condition is double that at continuous operation. Protean integrated drive This work is focused on the increase of torque density of the in-wheel motor. Previous work on this machine has been related to cost reduction [1], demagnetization analysis [2], torque density improvement [3] and fault tolerant III. ONE TRANSITION HALBACH ARRAY In this section, a Halbach array model using one transition magnet per pole is investigated. Following a parametrized design study, the most promising designs have been compared against the Reference Machine for different air-gap lengths and at the two operating points, corresponding to continuous operation (30Apk) and short term overload (79Apk). Simulations have been performed in This work has been supported by Protean Electric Ltd. All FEA simulations have been done using the software JMAG Designer. 1 School of Engineering, Newcastle University, Newcastle upon Tyne, United Kingdom 2 Protean Electric Limited, Farnham, Surrey, United Kingdom 978-1-5386-2477-7/18/$31.00 ©2018 IEEE Continuous DC voltage supply [V] 231 two dimensional FEA using an irreversible material so any effect of demagnetisation of the magnets is included on the torque predictions. For this initial study, the temperature of the magnet material has been set to 20℃ to get an overall impression of the torque capability of the Halbach array. In addition the permanent magnet material mass has been kept constant for all the machines. The Reference Machine uses NdFeB magnets with an energy product higher than 40MGOe. Fig. 4: Alteration of kw in the Halbach array design for different air-gaps at 20℃, continuous operation The demagnetisation shown indicates the percentage of magnet material that is demagnetised by more than 1%. This doesn’t mean that the whole area has completely lost demagnetisation but that there might be smaller areas which have. Fig. 2: 2D model of the Reference Machine (RM) Torque [p.u.] In the Reference Machine, physical air-gap variation was investigated by maintaining the permanent magnet height and reducing the thickness of the rotor core back, hence altering the air-gap without altering the volume of the permanent magnets. In the Halbach configuration, the permanent magnet height has been adjusted so the total amount of magnet material is constant and equal to that in the Reference Machine, regardless of air-gap length. The width of the pole is determined by the parameter kw, as defined in Fig. 3. This parameter was altered from 0.5 to 0.9. As values of kw less than 0.5 produce much poorer performance, only values of kw > 0.5 were used. Fig. 5: Fig. 3: Torque vs air-gap length for continuous operation at 80℃ Fig. 5 shows that, as expected for a kw = 0.7, higher rated torque than the Reference Machine can be achieved at 80℃. Although a Halbach array with a kw of 0.8 seem to obtain higher torques than the Reference Machine at 20℃, due to demagnetisation this is not met at realistic operating conditions. All three models have the same magnet height, however transition magnets for kw = 0.8 and kw = 0.9 are quite small, which reduces the magnet operating point, making it more susceptible to demagnetisation, which can be seen in Fig. 6. Pole pitch Halbach array As shown in Fig. 4, using a Halbach array allows an increase in the air-gap length of more than 20% whilst giving the same rated torque as the original surface mounted design. Continuous torque results take into account demagnetisation and the anticipated operation temperature of 80℃ for the magnets and are shown in Fig. 5 Demagnetisation simulations have also been undertaken at 125℃ for overload condition and they are compared in terms of torque and demagnetised magnet area in Fig. 6 and Fig. 7. The models include an irreversible material, hence when the operation point of the magnet reaches the “knee” on the BH curve, it recovers back in the recoil curve (parallel to BH curve) and shows a loss in magnetisation. Computed torque includes this non linearity of the material. When simulating demagnetisation of the Reference Machine, there was no loss in performance at continuous operation for any air-gap. This indicates that the magnets are working at a point which is adequate for this temperature. 232 3% RM k w = 0.7 2.5% k w = 0.8 Demagnetised area [%] Demagnetised area [%] k w = 0.9 2% 1.5% 1% 0.5% 0% Fig. 6: 1 1.2 1.4 1.6 Air gap [p.u.] 1.8 2 Fig. 8: Demagnetised area vs air-gap length in overload at 125℃ Demagnetised area vs air-gap for continuous operation at 80℃ IV. THREE TRANSITION HALBACH ARRAY Torque [p.u.] An alternative to the single transition Halbach configuration is the use of more transition magnets to generate a more sinusoidal air-gap flux density as the change in direction of magnetism within the magnets occursa in a more gradual manner. As torque can be expressed as: Fig. 7: (1) In order to increase torque, as current loading (Â), axial length (l), and rotor diameter (D) are fixed, only by increasing the air-gap flux density (B̂ g) the torque can be increased. Introducing more transitions increases the peak flux density in the gap (B̂ g) as shown in Fig. 9. A Fourier analysis reveals an increment of 8.8% in fundamental airgap flux density. As flux is better distributed, there is a reduction in space harmonics, which creates a more sinusoidal flux density distribution in the air-gap as seen in Fig. 10. Torque vs air-gap length in overload operation at 125℃ At overload condition, kw = 0.7 is still the topology that can deliver the highest torque. According to Fig. 7, the Halbach array topologies offer a better torque capability than the Reference Machine at overload conditions. In summary, the use of Halbach arrays could be beneficial for either obtaining higher torque or maintaining the torque developed by the Reference Machine but with an increase in the air-gap, which is beneficial for estimating the bearing tolerances and other mechanical aspects including eccentricity and other manufacturing issues. On the other hand these Halbach array topologies offer higher demagnetisation than the Reference Machine, as can be seen in Fig. 8. If no demagnetisation was occurring, higher torques could be obtained with this topology. Fig. 9: 233 Flux density in the air-gap for Halbach topologies Demagnetised area [%] Flux density [T] the equal segment model has a constant demagnetisation for the increase of air-gap length. However the unequal segment model has a better performance at air-gaps lower than 1.4 p.u, as shown in Fig. 12. Both suffer less demagnetisation than the one segment topology. Introducing more transitions reduces the rate of change between the segments direction of magnetisation (DOM). With fewer opposing magnetic domains between the segments the permeance coefficient is increased. Fig. 10: Harmonic distribution of the Flux density in the gap for studied topologies For this topology two models were used. In the first model, named the equal segment model, the pole-magnets and transition magnets are the same size. In the second model, termed, the unequal segment model, the transition magnets are the same size but the pole magnets are larger, as defined in Fig. 11. By increasing the width of the pole magnet, B̂ g can be increased further. Fig. 12: Demagnetisation comparison in continuous operation at 80℃ Fig. 13 shows that in overload operation, however, the unequal segment model has a lower demagnetisation than the equal segment model due to the bigger size of the pole magnet, which suffers less demagnetisation than the transitions. In the equal segment model, pole and transitions get demagnetised in a similar manner as they are all the same size. Fig. 11: Three transitions Halbach array topologies (equal segment model on the left and unequal segment model on the right) The results are compared in Table 2, which shows that the use of three transitions with equal segment size is of no benefit compared to the use of only one transition. However, in the unequal segment model, higher torque is obtained for both continuous and the overload condition. TABLE 2 TORQUE INCREASE COMPARISON AGAINST RM equal segment model Airgap [p.u.] 1 1.2 1.4 1.6 1.8 2 Conti nuous Overl oad 4% 4% 4% 4% 4% 4% 1% 1% 1% 0% 0% 0% unequal segment model Conti Over nuous load 5% 5% 6% 6% 6% 6% 4% 4% 4% 4% 4% 4% One transition kw = 0.7 Conti nuous Over load 5% 5% 5% 5% 5% 5% 3% 4% 4% 4% 3% 3% Fig. 13: Demagnetisation comparison in overload operation at 125℃ V. OVERCOMING DEMAGNETISATION One of the main concerns with Halbach array is the high demagnetisation due to the proximity of magnets with different polarities, hence other alternatives need to be investigated. Demagnetisation occurs due to the low operating point of the magnet with some areas working below the “knee point”. When the magnet works below this position, it cannot recover its remanence to the original The amount of permanent magnet material that is demagnetised is reduced by the use of three transitions. The magnets work at a higher operational point, further away from the “knee” of the BH curve. At continuous operation, 234 value, losing magnetisation. To try and resolve this issue, a new topology using trapezoidal magnets is investigated. The use of trapezoidal magnets in a Halbach array is not a new concept in itself, as it has been used before in other applications such as “fridge magnets” and linear actuators [9]. However, according to the literature, it has not previously been used in the rotor of a PMSM. By using trapezoidal magnets with a certain wall angle, more areas of the transition magnet work with a higher permeance coefficient, i.e. at higher operational points and hence less demagnetisation could occur. Fig. 14: Two poles trapezoidal Halbach array Fig. 14 shows a linear set of pole pairs consisting of trapezoidal magnets arranged in a Halbach array. Each pole magnet is separated by one transition magnet, where the angle between magnet walls is β and the angle of the change in direction of magnetisation across a wall is αc. The angle αc will depend on the number of transitions per pole and can be found using (2), being ns the number of segments per pole (pole and transitions). When using one transition, 2 segments are used, hence αc = 90°. (2) Fig. 15: Flux density in square Halbach array (top) and trapezoidal (bottom) Bknee = 0.3T in a linearized model Demagnetised area [%] If the direction of magnetisation of transition and pole magnets approaches the magnet wall with the same angle, the magnet will be working at a higher operating point. The angle β required to satisfy this requirement can be calculated using (3). In this one transition example, β will have a value of β = 90°. (3) Demagnetised area [%] The magnet material used in the Reference Machine has the “knee point” at 0.3T when operating at 125℃. Fig. 15 shows the flux density in a square Halbach array and in a trapezoidal Halbach array. The area in purple shows where the flux density is below the knee point, and hence the area affected by demagnetisation. As it is shown, trapezoidal magnets have a much smaller section working below the knee point. To study this behaviour in the actual geometry, the trapezoidal geometry has been compared to the kw = 0.7 single transition model and a model using kw = 0.5. Although kw = 0.5 couldn’t give the desired torque as shown previously, it represents a rotor with same size square pole and transition magnets. The magnet walls have been constrained to the adequate β angle in the trapezoidal topology. In addition, these models are compared against the Reference Machine in Fig. 16 and Fig. 17. A comparison with the 3 transition was intended to be done, however, due to the geometrical constraint a model using 3 transitions and trapezoidal magnets is not physically possible. As the number of transitions increases, the angle β increases as well, making it impossible to keep the same permanent magnet weight. Fig. 16: Demagnetisation area comparison at continuous operation (top) and overload (bottom) Demagnetisation results in Fig. 16 show that the trapezoidal magnet configuration offers lower demagnetisation than the square magnet and the rectangular pole one transition topology at rated current. At overload the trapezoidal has the lowest demagnetisation of all machines. 235 VI. CONCLUSION Torque [p.u.] Torque [p.u.] For a fixed volume of permanent magnet material, compared to the surface mounted machine, an increase in torque of 5% at continuous operation or 3% at overload can be achieved by using a single transition magnet Halbach array. Alternatively the same torque can be produced with a 20% air-gap increase, allowing for better tolerances for mechanical constraints. Using a three transitions Halbach array leads to very similar torque achievements, with a 5% torque increase in continuous operation and 4% in overload by having bigger pole than transition magnets. If increasing the torque is not intended, then again the air-gap distance can be increased by 20% while maintaining the torque for both topologies. A method has been introduced to reduce demagnetisation in Halbach arrays using trapezoidal magnets. By reducing the incident direction of magnetisation angle against the segments sides, this increases the permeance coefficient of the magnet and hence its operational point, having less area under the “knee” of the BH curve. This greatly reduces the overall demagnetised area both in continuous operation and overload, but doesn’t gives any significant improvements in torque. The topologies compared in this paper have been proposed for a very specific application with high geometrical constraints. Future work could be done using this topologies in a different application and with different working conditions. For this application the use of a Halbach array with a kw = 0.7 or the unequal segment model are good alternatives to the Reference Machine. Although both topologies suffer from demagnetisation, the produced torque is still higher than that of the Reference Machine. Fig. 17: Torque comparison at continuous operation (top) and overload (bottom) VII. [1] [2] [3] Fig. 18: Demagnetisation in square magnets (top) vs trapezoidal magnets (bottom) at overload condition [4] In terms of torque development, the trapezoidal magnet configuration does not produce as much torque as the square magnet topology, as seen in Fig. 17, as it has a lower peak gap flux density (B̂ g). [5] [6] The use of trapezoidal magnets in a Halbach array can´t give the same torque performance as using rectangular magnets in the adequate height to length ratio. However, the permanent magnet works in a much more secure environment when high fields or high temperatures occur. This trapezoidal configuration allows a more distributed permance across the magnet, resulting in using more effective the magnet material. The comparison in Fig. 18 clearly shows the demagnetised area in the square magnets with kw = 0.5, which reaches almost 10% of the total permanent magnet area, while this demagnetisation is less than 1% for the trapezoidal Halbach array. [7] [8] [9] REFERENCES S. Yang et al., “Cost reduction of a permanent magnet in-wheel electric vehicle traction motor,” Proc. - 2014 Int. Conf. Electr. Mach. ICEM 2014, pp. 443–449, 2014. S. Yang et al., “Magnet losses and demagnetisation in a permanent magnet in-wheel electric vehicle traction motor,” Proc. - 2015 IEEE Int. Electr. Mach. Drives Conf. IEMDC 2015, pp. 1831–1837, 2016. I. Martinez-Ocana, N. J. Baker, B. C. Mecrow, C. Hilton, and S. Brockway, “Transverse flux machines as an alternative to radial flux machines in an in-wheel motor,” in The 9th International Conference on Power Electronics, Machines and Drives, 2018, To be publish. C. J. Ifedi, B. C. Mecrow, S. T. M. Brockway, G. S. Boast, G. J. Atkinson, and D. Kostic-Perovic, “Fault tolerant in-wheel motor topologies for high performance electric vehicles,” 2011 IEEE Int. Electr. Mach. Drives Conf. IEMDC 2011, pp. 1310–1315, 2011. M. Munaro, N. Bianchi, and G. Meneghetti, “High Torque Density PM Motor for Racing Applications,” pp. 5826–5833, 2017. Q. Chen, G. Liu, W. Gong, and W. Zhao, “A new fault-tolerant permanent-magnet machine for electric vehicle applications,” IEEE Trans. Magn., vol. 47, no. 10, pp. 4183–4186, 2011. H. C. Lovatt, V. S. Ramsden, and B. C. Mecrow, “Design of an inwheel motor for a solar-powered electric vehicle,” IEE Proc. - Electr. Power Appl., vol. 145, no. 5, p. 402, 1998. M. Galea, T. Hamiti, and C. Gerada, “Torque density improvements for high performance machines,” Proc. 2013 IEEE Int. Electr. Mach. Drives Conf. IEMDC 2013, pp. 1066–1073, 2013. K. J. Meessen, B. L. J. Gysen, J. J. H. Paulides, and E. A. Lomonova, “Halbach permanent magnet shape selection for slotless tubular actuators,” IEEE Trans. Magn., vol. 44, no. 11 PART 2, pp. 4305– 4308, 2008. VIII. BIOGRAPHIES Iago Martinez Ocaña received the B.Eng. (Hons) in Electrical and Electronic Engineering in Glyndwr University, Wrexham, UK, in 2013 and the Technical Industrial Engineering in Industrial Electronics in Alcalá de 236 Henares University, Alcalá de Henares, Spain, in the same year. He received as well the M.Sc in Electrical Power in Newcastle University in 2014. He is currently pursuing the Ph.D. degree with the Electrical Power group at Newcastle University, Newcastle upon Tyne, UK. He has been working as an electromagnetic design engineer and has been involved in the Innovate UK project A.E.M.T.A. designing actuators for aerospace applications. His current research interests include in-wheel motors, permanent magnet machines and permanent magnet materials and assemblies. Nick J. Baker received the M.Eng. degree in mechanical engineering from the University of Birmingham, Birmingham, U.K., in 1999, and the Ph.D. degree in electrical machine design from Durham University, Durham, U.K., in 2003. He is currently a Lecturer with the Electrical Power Group, Newcastle University, Newcastle upon Tyne, U.K. He is a Machine Designer with research projects across the automotive, aerospace, and renewable energy sectors. Barrie Mecrow is Professor of Electrical Power Engineering at Newcastle University, UK. His research interests include fault tolerant drives, high performance PM machines and novel switched reluctance drives. He is actively involved with industry in the aerospace, automotive and consumer product sectors, who fund a large range of projects. Barrie commenced his career as a turbo-generator design engineer with NEI Parsons, England. He became a lecturer at the University of Newcastle in 1987 and a professor in 1998. In addition to his research interests he cares passionately about provision of high quality electrical engineering education in UK universities. Chris Hilton is the Chief Technology Officer at Protean Electric having previously held roles in the fields of communications electronics, satellite navigation and particle physics research. He holds a PhD in physics from the University of Manchester and a master's degree in mathematics from the University of Cambridge, UK. Simon Brockway is the Advanced Motor Research Manager at Protean Electric Ltd. He has more than 30 years' experience in permanent magnet electrical machine design and managing multi-disciplined engineering and product development teams. Having graduated in 1985 with a BSc (Hons) in Electrical & Electronic Engineering at Portsmouth Polytechnic he has been a developer of in-wheel motor technology since the mid 1990's. Apprentice trained he has experience in the servo, domestic appliance and automotive sectors having worked in low and high volume manufacturing, contractual research and start-up organisations. 237 Powered by TCPDF (www.tcpdf.org)