Progress in Electromagnetic Research Symposium 2004, Pisa, Italy, March 28 - 31 Comparison of Methods to Calculate Iron Losses Caused by Harmonic Fields in Rotating Machines K. Yamazaki and S. Watari Dept. of Electric Engineering, Chiba Institute of Technology, 2-17-1, Tsudanuma, Narashino, Chiba 275-0016, Japan e-mail: yamazaki@pf.it-chiba.ac.jp Abstract In this paper, we examine the accuracy of several calculation methods for iron loss of rotating machines using the finite element method. Two calculation methods are compared with the measurements. One is based on the Fourier transformation of the flux density waveforms at each finite element. The other is the direct calculation from the waveforms without Fourier transformation. It is clarified that the errors of the calculation methods vary due to the kinds of machines and that the method based on the Fourier transformation overestimate the hysteresis loss in case of highly saturated machines. I. Introduction Owing to the progress of the computers and the numerical methods, it has become possible to calculate the iron loss considering the harmonic and rotational electromagnetic fields in the rotating machines. Various calculation methods have been proposed and applied in recent years. There are two major methods for the harmonic losses calculation. One is based on the Fourier transformation of the waveforms of the flux density at each finite element. The other is the direct calculation from the waveforms without Fourier transformation (1)-(4). The first method is very useful to understand the each harmonic loss component. However, this method has one problem. In the strict scene, the hysteresis loss cannot be decomposed into harmonic components because of its nonlinear characteristics. Consequently, the total calculated iron loss may include error. In this paper, we compare the measured and the calculated iron losses of several rotating machines for the purpose of clarifying the accuracy of the calculation methods. II. Calculation Method A) Outline of the calculation Figure 1 shows the outline of the common procedure for the iron loss calculation of rotating machines. First, the electromagnetic field of the machine is calculated by the nonlinear time-stepping finite element method considering the magnetic saturation of the core and the movement of the rotor. The governing equation is as follows: ∇ × (ν∇ × A) = Jc − σ ∂A + v 0∇ × ∂t (1) where A is the magnetic vector potential, Jc is the current of the coil, M is the magnetization of the magnet, ν is the reluctivity and σ is the conductivity. At this step, the eddy current and the hysteresis phenomena in the core are neglected because it will require vast of computer resources to take the lamination structure of the core into account directly by the finite element analysis. The iron loss of the core, which are the eddy current and the hysteresis losses in the laminated electrical steel plates, are calculated in the second step with the time variation of flux density at each finite element. In this case, the effects of the reaction field caused by the iron loss in the laminated core are neglected in the total procedure. However it can be considered that they are small compared with the total field. B) Methods to calculate iron loss The core loss per weight induced by the single frequency alternating magnetic field can be expressed as follows: 329 Progress in Electromagnetic Research Symposium 2004, Pisa, Italy, March 28 - 31 w i = w ie + w ih = K e f 2 B max 2 + K h fB max γ (2) where wie, wih are the eddy current and the hysteresis losses per weight, Ke, Kh and γ are the experimental constants obtained by the Epstein frame, f is the frequency of the alternating magnetic field and Bmax is the maximum flux density during the time interval. However, as we will discuss in Chapter III, the magnetic field in rotating machines are not single frequency alternating field even when the sinusoidal power supply is utilized. As the stator and the rotor structures produce much time harmonics due to the rotation, the harmonic eddy current losses and the minor hysteresis losses are generated in the core. To consider these effects, the method based on the Fourier transformation of the flux density waveform at each finite element are often applied: We = ∫ ∑ K D(nf ) {B e iron 2 2 mr , n + Bmθ , n }dv (3) Wh = ∫ 2 h iron n ∑ K D(nf ){B mr , n γ + Bmθ , n }dv γ (4) n where We and Wh are the total eddy current and hysteresis losses, D is the density of the core, n is the order of the harmonics, Br,n and Bθ,n are the radial and the peripheral component of the nth harmonic flux density. On the other hand, the core loss can be calculated directly from the flux density waveform without the Fourier transformation (1)-(4). The eddy current loss in the steel plate can be expressed by the timevariation of the magnetic field as follows wie = h 2σcore 1 12 D T ∫ T 0 dB 2 dt dt (5) where h is the thickness, σcore is the conductivity of the steel plate and κ is the modified coefficient of the eddy current loss due to the effect of grains in the steel plate. In this case, the total eddy current loss can be calculated with the result of the finite element analysis and the coefficient Ke as. Wie = 1 N Brk + 1 − Brk 2 Bθk + 1 − Bθk 2 K eD ∑ ∆t + ∆t dv 2π 2 ∫core N k =1 (6) where N is the number of time step per one time period, ∆t is the time interval. This expression includes the effect of the time-harmonics. The hysteresis loss including harmonics can be also calculated by counting the local minimum and maximum peaks of the waveforms as Np θ KhD NE ∆Vi Npr ij 2 ( ) (Bmθij )2 Wih = B mr + ∑ ∑ ∑ T i =1 2 j =1 j =1 i i (7) where ∆Vi is the volume of i-th finite element, Bmrij, Bmθij are the amplitudes of each hysteresis loop, Npri, Npθi are the number of peaks in the time-variation of Br and Bθ. Fig.2 shows the outline of the method to determine the amplitudes of each hysteresis loop involving major and minor ones. Finite Element Method Iterative calculation for magnetic saturation of core Re-mesh at each time step due to movement of rotor Br,Bθ 1 2Bm1 (2Bmax) Time series data of flux density 5 2Bm5 2Bm6 6 Iron Loss Calculation 2Bm3 3 Eddy current loss of core considering harmonics Wie 2Bm4 (2Bmax) Hysteresis loss of core considering minor loops Wih 2 2 2Bm Iron loss considering harmonics Wi = Wie Wih 4 Figure 1. Outline of calculation Figure 2. Determination of amplitudes of major and minor hysteresis loops 330 Progress in Electromagnetic Research Symposium 2004, Pisa, Italy, March 28 - 31 III. Comparisons between Measured and Calculated Results The two calculation methods are applied to two kinds of machines, which are the surface permanent magnet motor and the induction motor. The measured and the calculated iron losses are compared to clarify the advantages and disadvantages of the calculation methods. Figure 3, 4 show the field distribution of the permanent magnet and induction motors. The motors are manufactured and measured by IEE-Japan. Although the outline of the permanent magnet motor is round shape and the induction motor is rectangular, the shape of the stator slot is identical. The electrical steel plates applied to the stators and the rotors cores are also identical. The number of the steel is 50A1300 defined by Japanese industrial standard. The iron loss coefficients determined by the Epstein frame test are Ke=3.098X10-4 and Kh=6.013X10-2. On the other hand, the rotor structures of the motors are much different. The permanent magnet motor has no rotor slots, whereas the induction motor has 34 rotor slots with the aluminum cage. As a result, the flux density waveform of the induction motor includes high order timeharmonics caused by the movement of the rotor, especially at the stator teeth top. On the contrary, the waveforms of the permanent magnet motor are nearly rectangular wave due to the magnetic saturation and they include no slot harmonic field. In this case, 3, 5, 7th harmonics are mainly included in the waveform. Figure 5, 6 show the experimental and the calculated iron losses of the motors. The experimental and calculated results agree well. In the experiment, the iron loss is obtained from the no-load loss. As a result, several components except for the eddy current and the hysteresis loss of the laminated core is included in the experimental iron loss, for instance, the harmonic eddy current losses of the magnet in the case of the permanent magnet motor and the harmonic eddy current losses of the aluminum cage in the case of the induction motor. These losses cannot be separated from the core losses by the measurement. Thus, these components are calculated directly by the finite element method and added to the calculated results. As the air gap of the permanent magnet is relatively large, the time-variation of the flux density at the rotor is nearly DC filed and the rotor iron loss is negligible. On the other hand, in the case of the induction motor, the rotor iron loss caused by the stator slot harmonics cannot be neglected because of the small air-gap. The underestimation of the calculated results in the case of the induction motor is mainly caused by the neglect of the inter-bar current, which pass through the rotor cage to the rotor core. Let us discuss the detailed difference between the results of the calculation methods. In the case of the permanent magnet motor, the result calculated with the Fourier transformation is larger than the direct calculation from the flux density waveform whereas the results are nearly same in the case of the induction motor. This error is caused by the misestimated hysteresis loss caused by the harmonic fields. Figure 7, 8 show the calculated hysteresis losses by the two methods. In the case of the permanent magnet motor, the fundamental component obtained by the FFT is already larger than the total hysteresis loss obtained by the direct calculation because the waveform is nearly rectangular wave. Furthermore, the 3, 5, 7th harmonic losses are added. However, they do not produce the hysteresis loss because they do not cause the minor loops. As a result, the calculation method using the Fourier transformation overestimates the total hysteresis loss in the case of the permanent magnet motor. IV. Conclusions The two calculation methods for the harmonic iron loss calculation of the rotating machines are applied to two kinds of machines. It is clarified that the errors of the calculation methods vary due to the kinds of machines. Although the method based on the Fourier transformation is useful to understand each harmonic loss component, it overestimates the hysteresis loss in case of highly saturated machines. It can be said that the method based on the direct calculation from the flux density waveform can be applied to various types of rotating machines. REFERENCES 1. 2. 3. 4. K.Atallah, Z. Q. Zhu and D. Howe. ”An Improved Method for Predicting Iron Losses in Brushless Permanent Magnet Motor ”, IEEE Trans. Magn., vol.28, no.5, pp.2997-2999, 1992. 1. S. L. Ho, W. N. Fu and H. C. Wong: “Estimation of Stray Losses of Skewed Rotor Induction Motors Using Coupled 2-D and 3-D Time Stepping Finite Element Methods”, IEEE Trans. Magn, vol.34, no.5, pp.3102-3105, 1998. Sadowski, J. P. A. Bastos, M. V. Ferreira Da Luz, P. Kuo-Peng and M. Lajoie-Mazenc, “The Rain-Flow Method for Evaluation of Iron Losses in Electrical Machines”, IEEE Trans. Magn, vol.36, no.4, pp.1923-1927, 2000. K. Yamazaki, “Torque and Efficiency Calculation of an Interior Permanent Magnet Motor Considering Harmonic Iron Losses of both Stator and Rotor”, IEEE Trans. Magn, vol.39, no.3, pp.1460-1463, 2003. 331 Progress in Electromagnetic Research Symposium 2004, Pisa, Italy, March 28 - 31 1 1 0 -1 0 45 90 135 180 Rotor Position (deg) -2 -3 0 -1 0 45 -2 -3 Stator Teeth Top Br 1 Br 2 Br T) 2 90 135 180 Flux Density(T) 3 Flux Density Br 2 Flux Density Flux Density 3 0 0 5 10 15 20 Tim e(m s) 1 0 0 5 10 -1 15 20 Time(ms) Rotor Positoin (deg) -1 Stator Teeth Center -2 Stator Teeth Top Stator Teeth Center Figure 4. Field of induction motor. Figure 3. Field of permanent magnet motor. 30 30 Magnet eddy current loss Magnet Eddy 20 Rotor Cage eddy current loss Magnet Eddy Iron Loss W) Iron Loss W) 20 Stator hysteresis loss 10 Rotor hysteresis loss Stator hysteresis loss 10 Rotor eddy current loss Stator eddy current loss 0 Exp. 1 (Average) Cal. 2 (Direct) Stator eddy current loss 0 Cal. 3 (FFT) Cal. 2 (Direct) Cal. 3 (FFT) Figure 6. Measured and calculated loss of IM. 18 9 16 8 14 7 Hysteresis loss (W) Hysteresis loss (W) Figure 5. Measured and calculated loss of PM. Exp. 1 (Average) 12 10 8 6 6 5 4 3 4 2 2 1 0 0 All1 2 31 4 53 6 75 8 9 1 All 10 12 16 11 18 2 Time Harmonics OrderOrder of time of harmonics Direct Cal. 31 4 5 3 6 75 8 9 10 11 12 16 18 OrderOrder of timeofharmonics Time Harmonics Using FFT Direct Cal. Figure 7. Calculated hysteresis loss of PM. Using FFT Figure 8. Calculated hysteresis loss of IM. 332