2D analytical model for estimation of eddy current loss in the magnets of IPM machines considering the reaction field of the induced eddy currents Milind Paradkar, Student Member, IEEE and Joachim Böcker, Senior Member, IEEE Abstract-- In this paper a closed-form 2D analytical model developed for the estimation of eddy current loss in interior permanent magnet machines is presented. To check the generality of the developed model, two different types of commonly employed interior permanent magnet machines are considered. 2D finite-element analysis was used to benchmark the results from the analytical model. The proposed model can be directly used to evaluate the eddy current magnet loss for magnet segmentation in the circumferential direction. The influence of air-gap length, frequency of carrier harmonic, reaction field and non-linearity of iron due to saturation on the magnet loss have been studied in detail and the results are presented. The effect of axial segmentation has been considered by incorporating an end-effect factor in the analytical model. The results from analytical and 2D finite-element analyses are found to be in concurrence. Index Terms--permanent magnets, eddy current loss, magnet segmentation, analytical model, finite element method, interior permanent magnet synchronous motor I. INTRODUCTION I NTERIOR permanent magnet (IPM) motors are widely used in industrial applications where the ratio of nominal speed to maximum speed is in excess of 2 as in applications like hybrid and pure electric cars [1]. These motors are fed from an inverter which results in high frequency time harmonic currents. These harmonics cause eddy current loss in the magnets resulting in temperature rise and may result in irreversible demagnetization of the magnets. Hence the calculation of the eddy current loss and its mitigation through segmentation can ensure the safety of magnets of the IPM Milind Paradkar is a research co-worker at the Department of Power Electronics and Electrical Drives, University of Paderborn, Germany (email: paradkar@lea.upb.de). Joachim Böcker is a professor and head of the Department of Power Electronics and Electrical Drives, University of Paderborn, Germany (e-mail: boecker@lea.upb.de). 978-1-4799-7940-0/15/$31.00 ©2015 IEEE machines at different loads and speeds. Many previous publications have dealt with the calculations of eddy current loss in the magnets of PM machines, but mostly for surfacemounted PM machines [2]‐[8]. Magnetic field variations in the magnet are of the form travelling waves in the surfacemounted PM machines while being pulsating waves for IPM machines. Therefore the analytical calculations of eddy current loss in the surface mounted PM machines cannot be directly applied to IPM machines. The eddy current loss calculation for the IPM machine was investigated in [9]-[13]. Analytical and numerical finite element analyses were compared in [9]-[12] but did not show good concurrence. Consideration of air-gap effects on reaction fields of induced eddy currents is absent, which is necessary for accurate analysis of eddy currents in the magnets. In this paper, analytical and numerical finite element analyses of eddy current loss in the magnet of IPM machines are presented. To validate the model, two different types of IPM motors were considered. A 2D analytical model was developed for eddy current loss in the magnet with the consideration of the reaction field of induced eddy currents. Results from the 2D time-stepping finite element method are used to benchmark the results obtained from the analytical model. Analytical models help in quick estimation of eddy current loss and in determination of the appropriate magnet segmentation required for loss reduction. II. SPECIFICATIONS OF IPM MOTORS Fig. 1 shows the two types of IPM motors considered for analysis [14]. The motors have 8 poles and an internal and external stator diameter of 160 mm and 269 mm respectively. The air-gap length and axial length were 0.73 mm and 84 mm respectively. The magnet in motor 1 and motor 2 has a thickness of 6.5 mm while the width was 37.8 mm for motor 1and 18.9 mm for motor 2. The Cartesian co-ordinate system was chosen for the 1096 analysis as the magnets in an IPM machine are flat shaped. The following assumptions were made for analytical calculation of eddy currents in the magnets. 1) The permeability of iron is considered infinite ( r ∞ 2) Only the normal component of the flux density is 0 considered x z 3) Only the axial component of current density, z is considered x 0 y 4) The effect of stator slotting is neglected. Fig. 2 Simplified rectangular geometry for the analytical model 3 Fig. 1. IPM Motor 1 and Motor 2 considered for the analysis The simplified 2D model for the estimation of loss is shown in Fig. 2. The simplification is justified if the air-gap length is very much smaller than the air-gap radius. The air-gap length and magnet thickness is denoted by and respectively. For generalization, taking into account the magnet segmentation in circumferential direction, the magnet width is denoted by . Applying Ampere’s law for the loop shown in the Fig. 2 results in (1). ′ If we consider the source field is changing antiperiodically in axis direction resulting in (4). The variation of is shown in Fig. 3 [15] [16]. 4 · , 1 · , , , ,… · sin · 2 · cos · · 4 where, Bs,m is the maximum value of the Bs along the width of the magnet. 1 is the sum of source field from the stator where ′ and reaction field from the magnet eddy currents (all in is the effective air-gap length direction) and ( . If the magnitude of can be assumed to be constant along the direction, then (1) simplifies to (2). 2 Applying Faraday’s law leads us to (3). Fig. 3. Source field variation in a magnet segment The considered confines the induced eddy current within the magnet segment. Combining (2) and (3), the resulting differential equation for sinusoidal variation of as in (4), results as 5 1097 Equation (5) is solved using Fourier series method resulting in solution of as · , , , , ,… · · · · 2 · sin · cos · · 6 The induced eddy currents hence result as 4 · · , · · sin · · , , , ,… · sin · · 7 · 350 Analytical 300 250 200 150 Inductance -limited Resistance-limited 0 1 2 3 4 5 6 7 8 Circumferential magnet segment number, m Fig. 4. Loss comparison from FEM and Analytical for Motor 1 | | · FEM 50 450 · · , , , ,… · · · · · , FEM 400 8 Analytical 350 Eddy current magnet loss in W 4 is estimated as 15 mT. 400 100 · 2 If the length of the magnet in the axial direction axis) is represented by , the eddy current loss in each segmented magnet is then computed as · 2· , 450 · Eddy current loss in W 4 6.25 A. The value of The eddy current loss in magnet segment of an IPM machine as given by (8) is proportional to the width, thickness and magnet conductivity. It also varies as square of the flux density and frequency of the source field. The second term in , represents the effect of the denominator of (8), reaction field on the eddy current loss. If this is neglected it can lead to wrong results as will be seen in the next section. 300 250 200 150 100 50 0 1 III. RESULTS AND DISCUSSION The validity of the analytical model is tested by comparing the loss results obtained by time-stepping 2D finite element method. On a fundamental time harmonic current with an amplitude of 250 A at a frequency of 80 Hz at 1200 min-1, a carrier harmonic with an amplitude of 6.25 A (2.5 % of fundamental) at 10 kHz on the rotor side is analyzed for loss in the magnets of the two IPM motors 1 and 2. The stator and rotor lamination was considered linear with a relative permeability of 5000. The magnet length was taken as 84 mm, same as the lamination length. The conductivity of the magnet was considered 625000 S/m for the analysis. The flux density , is evaluated by a magneto-static FE analysis with d-flux excitation corresponding to current amplitude of 2 3 4 Circumferential magnet segment number, m Fig. 5. Loss comparison from FEM and Analytical for Motor 2 Fig. 4 and 5 show the comparison of the total eddy current loss in the magnets for various segmentation number, in the circumferential direction, as obtained from analytical and time stepping 2-D FE analyses for motor 1 and motor 2. There is a good concurrence with a maximum difference of 8 %. It is observed from Fig. 4 that with increase in magnet segmentation in the circumferential direction from 1 to 2 the total eddy current loss actually increases nearly by a factor of 2. This region is the inductance-limited region where the induced eddy currents in the magnets are limited by the magnet inductance [6]. Increasing the magnet segmentation, 1098 in this region leads to increase in eddy current loss. Beyond a certain limit, the eddy current loss starts decreasing with the increase in segmentation number. This region is the resistance-limited region where the induced eddy currents in the magnets are limited by the resistance. For motor 1, the resistance-limited region is for 2. It is obvious from the figure that it is better to remain the magnets unsegmented than segmenting if the chosen lies in range 2-5. The reduction in magnet loss due to segmentation can only be realized for 6. For the case of motor 2, increasing segmentation reduces the loss as it already in resistancelimited region for 1. Eddy current magnet loss in W 350 300 250 200 150 FEM 100 Analytical, Air-gap length considered Analytical, Air-gap length neglected 50 0 1 2 3 4 5 6 7 Circumferential magnet segment number, m Fig. 6. Influence of air-gap length - Motor 1 400 350 Eddy current magnet loss in W B. Influence of carrier frequency The region corresponding to inductance or resistance limitation is determined by the frequency of the carrier harmonic with all the other parameters remaining constant. Fig. 8 and 9 show the variation of the loss for different carrier frequencies for motor 1 and motor 2 respectively. With increasing harmonic frequency the loss curve shifts to the right. It can be seen from Fig. 10 that for a 20 kHz harmonic for motor 2, there is an inductance-limited region along with the resistance-limited which is not seen at 5 kHz and 10 kHz. Fig. 10 and 11 show the variation of the flux density and the induced current density in the magnet along the width of the magnet for a carrier amplitude of 6.25 A at different carrier frequencies for motor 1. The effect of the reaction field can be clearly noticed. At low frequencies where the reaction field effect is negligible, the flux density of the source field is uniformly distributed and the induced current density varies linearly 1 , both along the width of the magnet. However at higher frequencies the reaction field is much stronger thus causing the flux density and induced current density variation along the width of the magnet to be no longer uniform and linear along the magnet width. With increasing frequency both the fields are concentrated at the edges of the magnet. 400 8 FEM Analytical, Air-gap length considered Analytical, Air-gap length neglected 300 250 200 150 100 50 0 1 2 3 Circumferential magnet segment number, m Fig. 7. Influence of air-gap length - Motor 2 10 kHz Analytical 800 5 kHz Analytical 700 20 kHz FEM 600 10 kHz FEM 5 kHz FEM 500 400 300 200 100 0 1 2 3 4 5 6 7 8 Circumferential magnet segment number, m Fig. 8. Variation of loss with carrier frequency-Motor 1 1099 4 20 kHz Analytical 900 Eddy current magnet loss in W A. Influence of air-gap length In earlier works on eddy current loss estimation in IPM motors the effect of air-gap length was not considered [10] [12]. This leads to inaccurate results especially if the air-gap length is not too small when compared to the magnet thickness. The influence of air-gap length is studied by replacing the effective air-gap length by in (8). The effect on loss with and without the consideration of the air-gap length is seen in Fig. 6 and Fig. 7 for motor 1 and motor 2 respectively. Neglecting the air-gap length leads to under-estimation of eddy current loss as the reaction effect is smaller. 450 20 kHz Analytical 900 10 kHz Analytical Eddy current loss in W 800 C. Influence of the reaction field If the reaction field is neglected (3) results as in (9) 5 kHz Analytical 20 kHz FEM 700 · 10 kHz FEM 600 5 kHz FEM 500 9 defined as in (4), resulting in solution of With 400 4 300 · , · · sin · , , , ,… as · 2 200 100 · sin · · 10 0 The power loss in each magnet segment results as in (12) 1 2 3 4 Circumferential magnet segment number, m Fig. 9. Variation of loss with carrier frequency-Motor 2 · 2· 16 · | | 14 4 12 1 Hz Flux density in mT 10 5 kHz 6 10 kHz 4 20 kHz 2 0 -18.9 -12.6 -6.3 -2 0 6.3 12.6 18.9 Magnet width in mm Fig. 10. Flux density variation along magnet width- Motor 1 Induced current density in kA/mm2 3000 2000 1000 0 -18.9 -12.6 , , , ,… 1kHz 8 -6.3 0 -1000 6.3 12.6 18.9 1 Hz 1 kHz -2000 -3000 Magnet width in mm 5 kHz 10 kHz 20 kHz Fig. 11. Induced current density variation along magnet width- Motor 1 · · · · · · , 11 Fig. 12 and 13 show the effect on loss with and without the consideration of the reaction field on the evaluated total eddy current loss. It is observed that neglecting the effect of reaction field results in over-estimation of calculated loss, especially in the inductance-limited region. In the resistancelimited region, however, the difference is not significant. D. Influence of iron saturation The analytical model assumed the relative iron permeability as infinite and hence it was possible to obtain a closed form expression for the eddy current loss. 2D FE analysis has been used to calculate the loss taking the nonlinearity into consideration. For stator and rotor, lamination material M330-35A was used [17]. Fig 14 and 15 show the variation of loss with the consideration of non-linear effects of iron lamination for motor 1 and motor 2 respectively. It has to be noted that at the operating condition of 250A fundamental peak, the current loading corresponds to 1500 A/cm RMS. So at the chosen operating point, the motor is in a highly saturated condition. For motor 1, the eddy current loss with saturation increases when compared to the linear case in the range of 2-27 % depending on the segmentation number . For the motor 2, however, the loss decreases when compared to the linear case and is in the range of 3-10 %. Hence a generalization of the saturation effect on the loss cannot be made. It has to be noted that the pattern of variation of the loss with segmentation number in the circumferential direction remains unchanged. 1100 700 450 Analytical with reaction field Analytical without reaction field 500 400 300 200 100 1 FEM Non-linear 300 Analytical 250 200 150 100 50 0 2 3 4 5 6 7 8 Analytical with reaction field Analytical without reaction field 140 120 100 80 60 40 20 0 1 2 3 Circumferential magnet segment number, m 4 Fig. 13. Loss variation with and without reaction field – Motor 2 500 FEM-NonLinear Analytical 400 350 4 depends on the width and depth of a magnet segment [4]. Using the above method, for one segment in circumferential direction i.e. 1 and for various magnet segments in axial direction, , the loss results obtained by analytical model is presented. The number of magnet segments in axial direction, , was varied from 1 to 21. For each , the value of was computed using (9) and the modified magnet conductivity was calculated as FEM-Linear 450 3 E. Influence of axial segmentation As the model has been developed for 2D, the analytical model considers only the segmentation in the circumferential direction. This is also true for 2D FE analysis. In some cases it may be beneficial to have axially segmented magnets. In that case it is necessary to do a 3D analysis and a 2D analysis would not suffice. However, 3D FE analysis is time-intensive and demands huge computational power especially if the model size is large. Previous works have proposed the usage of an end-effect factor in 2D simulations to include the axial segmentation effect [4]. This factor, which is less than 1, is then used to modify the conductivity of the magnet to account for the magnet segmentation in the axial direction. Effectively, the magnet conductivity is reduced to account for the axial segmentation effect. The effect factor 3 9 4 180 160 2 Circumferential magnet segment number, m Fig. 15. Loss variation with non-linear iron – Motor 2 Circumferential magnet segment number, m Fig. 12. Loss variation with and without reaction field – Motor 1 Eddy current magnet loss in W 350 1 0 Eddy current magnet loss in W FEM Linear 400 Eddy current magnet loss in W Eddy current magnet loss in W 600 300 250 200 625000 S/m 150 100 50 0 1 2 3 4 5 6 7 Circumferential magnet segment number, m Fig. 14. Loss variation with non-linear iron – Motor 1 8 Fig. 16 shows the eddy current magnet loss in motor 1 and motor 2 for 1 and varying from 1 to 21. Only the 3D FE analysis of motor 1 is available at this time. It is observed that in the inductance-limited region, the magnet loss estimated using the end effect factor is higher than calculated using 3D FE analysis. In the resistance-limited region the difference reduces. 1101 V. ACKNOWLEDGMENT 450 Eddy current loss in W This work is funded by the Federal Ministry for Economic Affairs and Energy (BMWi) within the ATEM program. The authors are responsible for the contents of this publication. Motor 1 400 Motor 2 350 3D FEM - Motor 1 300 For circumferential segmentation, m =1 250 VI. REFERENCES 200 [1] 150 100 [2] 50 [3] 0 1 3 5 7 9 11 13 15 17 19 21 Axial magnet segment number, n Fig. 16. Loss variation with axial magnet segmentation, n for m =1 [4] F. Loss with inverter excitation The proposed analytical model can be extended to calculate the magnet loss under inverter excitation. For each known harmonic current the loss is computed and summed to get an initial estimate of the total eddy current loss. IV. CONCLUSIONS [5] [6] [7] This paper proposed a 2D analytical model for quick estimation of eddy current loss in the magnets of IPM motors taking into account the reaction field of the induced eddy currents. The proposed model was analyzed for two different types of commonly used IPM machines taking into account the effect of circumferential magnet segmentation. The analysis showed that an unskilled choice of the segmentation number can actually lead even increased magnet loss in the machine resulting in higher magnet temperature and thereby resulting in reduced performance. It was observed that the reaction field, if neglected, results in big errors in the estimation of magnet loss especially in the inductance-limited region. The effect of air-gap length, which was previously neglected, has a significant effect on the magnet loss. The actual loss was higher with the air-gap length considered. The effect of carrier frequency on the magnet loss was also analyzed. Choice of switching frequency can be a major factor in deciding the eddy current loss in the magnets. The effect of iron saturation on eddy current loss was not conclusive. Analysis showed that in one case it actually resulted in reduced loss. The effect of axial segmentation was considered by using an end-effect factor and results show good agreement in the resistance-limited region. However, considering the time taken by a 3D FE analysis, it is still a good bargain. Having proved the validity of the proposed analytical model at different switching frequencies, the analysis can be extended to evaluate the magnet eddy current loss with inverter excitation. [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] 1102 Powered by TCPDF (www.tcpdf.org) B. K. Bose, "A high performance inverter drive system of an interior permanent magnet synchronous machine," IEEE Trans. Industry Applications, vol. 24, pp. 987-997, 1988. Y. Aoyama, K. Miyata and K. Ohashi, "Simulations and experiments on eddy current in Nd-Fe-B magnet," IEEE Trans. Magnetics, vol. 10, pp. 3790-3792, 2005. P. Sergeant and A. Van den Bossche, "Segmentations of magnets to reduce losses in permanent magnet synchronous machines," IEEE Trans. Magnetics, vol. 44, pp. 4409-4412, 2008. S. Ruoho, T. Santa-Nokki, J. Kolehmainen and A. Arkkio, "Modeling magnet length in 2-D Finite-Element analysis of electric machines," IEEE Trans. Magnetics, vol. 45, pp. 3114-3120, 2009. C. Bode and W. R. Canders, "Advanced calculation of eddy current losses in PMSM with tooth windings," International Conference on Electric Machines (ICEM), pp. 1-6, 2010, Rome. M. Mirzaei, A. Binder, B. Funieru and M. Susic, "Analytical calculations of induced eddy current losses in the magnets of surface mounted PM machines with consideration of circumferential and axial segmentation effects," IEEE Trans. Magnetics, vol. 48, pp. 4831-4841, 2012. P. Zhang, G. Y. Sizov, Jiangbiao He, D.M. Ionel and N.A.O. Demerdash, "Calculation of Magnet losses in Concentrated-winding Permanent-Magnet Synchronous Machines using a Computationally Efficient Finite-element method," IEEE Trans. Industry Applications, vol. 49, pp. 2524-2532, 2013. P. Sergeant and A. Van den Bossche, "Segmentations of magnets to reduce losses in permanent magnet synchronous machines," IEEE Trans. Magnetics, vol. 44, pp. 4409-4412, 2008. K. Yamazaki and S. Watari, "Loss analysis of permanent-magnet motor considering carrier harmonics of PWM Inverter using combination of 2D and 3-D finite-element method," IEEE Trans. Magnetics, vol. 41, pp. 1980-1983, 2005. K. Yamazaki and A. Abe, "Loss investigation of Interior PermanentMagnet Motors Considering Carrier Harmonics and Magnet Eddy Currents," IEEE Trans. Industry Applications, vol. 45, pp. 659-665, 2009. T. J. E. Miller, M. I. McGilp and K. W. Klontz, "Approximate methods for calculating rotor losses in permanent-magnet brushless machines," IEEE International Electric Machines and Drives Conference (IEMDC), pp. 1-8, 2009, Miami. A. Bettayeb, X. Jannot and J. -C. Vannier, "Analytical calculation of rotor magnet eddy current losses for high speed IPMSM," International Conference on Electric Machines (ICEM), pp. 1-6, 2010, Rome. J. Blum, X. J. Merwerth and H. -G. Herzog, "Magnet eddy –current losses in interior permanent magnet machines with concentrated windings – analysis and reduction of the major source," IEEE International Conference on Power Electronics, Machines and Drives (PEMD), pp. 1-6, 2014, Manchester. Evaluation of the 2007 Toyota Camry Hybrid Synergy Drive System, Prepared by: Oak Ridge National Laboratory Mitch Olszewski, Program Manager, http://www.osti.gov/scitech/servlets/purl/928684/ R. Stoll, The Analysis of Eddy Currents, Clarendon Press, Oxford, 1974 K. J. Binns, and P. J. Lawrenson, Analysis and Computation of Electric and Magnetic Field Problems, Pergamon Press, 1973 http://www.sura.se/