Professional Documents
Culture Documents
2014 - A Hybrid Model For Improved Hysteresis Loss Prediction in Electrical Machines 9p
2014 - A Hybrid Model For Improved Hysteresis Loss Prediction in Electrical Machines 9p
fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TIA.2014.2300160, IEEE Transactions on Industry Applications
Abstract— This paper presents a model for calculating the predict minor loop losses under a large variety of practically
hysteresis loss in electrical machines. The model can achieve encountered flux waveforms in electrical machines. The only
accurate and computationally efficient hysteresis loss calculations way these formulas can achieve accurate minor loop loss
by utilizing both analytical equations and the Energetic
prediction is to use variable coefficients based on a database
hysteresis model. The model results are verified experimentally
by comparing to a series of minor hysteresis loop measurements. of minor loop measurements [1], [5]. Such a database is
The hybrid model is then implemented to calculate the core losses difficult to provide by steel manufacturers, as the minor
in a switched reluctance (SR) machine using finite element (FE) hysteresis loop losses are dependent on multiple factors.
simulation. The results show that precise machine core loss Many hysteresis loss models are available for hysteresis
prediction requires having a model that is capable of calculating loop modeling, ranging from purely mathematical models [6]
the hysteresis losses under a variety of minor hysteresis loops.
to physics-based models such as the Preisach model [7, 8], the
Index Terms— Core loss, Hysteresis loss, Minor loops, Energetic Jiles-Atherton model [9], and the Energetic model [10]
model, Switched reluctance machine. Although the Preisach model can achieve accurate loss
prediction [11], the long computation time makes it unsuitable
I. INTRODUCTION for finite element (FE) electrical machine design, which
The design of high efficient electrical machines requires an requires fast core loss calculation in each mesh element. The
optimization procedure for minimizing the machine losses. In Jiles-Atherton model is adopted in [21] to calculate the static
order to achieve the optimum design, accurate core loss hysteresis losses of induction motors. The results demonstrate
calculations have to be performed for each candidate design. the importance of considering the harmonics initial phase in
Therefore, it is essential to have an accurate and fast core loss the machine core loss calculation.
model in order to optimize the machine within a convenient In this paper, a hybrid model is proposed to calculate the
computation time. hysteresis losses under any flux waveform. The model utilizes
One of the challenges of core loss prediction in electrical both analytical formulas and the Energetic hysteresis model to
machines is the calculation of the hysteresis losses under achieve accurate and computationally efficient hysteresis loss
distorted flux. The flux waveforms inside many machines, prediction.
e.g., permanent magnet (PM) machines and switched Section II describes the technique used for measuring minor
reluctance (SR) machines are naturally non-sinusoidal and hysteresis loops. Section III presents three models for
contain significant harmonic content and considerable DC hysteresis loss calculations; an analytical model, Energetic
component in some regions. Depending on the phase and hysteresis model and a hybrid model. In section IV, the
magnitude of these harmonics, the resulting flux waveforms in hysteresis losses of a 6/4 SR machine are examined using the
the machine core may contain local flux reversals causing three models.
minor hysteresis loops to occur inside the main loop. The
minor hysteresis loops can also be produced by non-sinusoidal II. MEASUREMENT CONCEPT
supplies, e.g., bipolar pulse width modulation (PWM) Since it is practically impossible to predetermine the
inverters [20]. general properties of the minor loops generated in electrical
Many empirical formulas are presented in the literature to machine laminations, the core loss model used should be able
evaluate minor loop hysteresis losses [1]-[4]. These formulas to calculate minor loop losses under all possible conditions.
can provide reasonable estimates of the hysteresis losses under As illustrated in Fig. 1, the hysteresis loss caused by a minor
certain conditions. However, they cannot be relied on to loop in a certain magnetic material is dependent on four
factors,
i. The magnitude of the minor loop B .
This work was supported by the Natural Sciences and Engineering ii. The position of the minor loop B m .
Research Council of Canada and by Hydro-Québec.
M. Ibrahim is with the Department of Electrical and Computer iii. The peak flux density of the major loop B p .
Engineering, Concordia University, Montreal, QC H3G 1M8, Canada, (e-
mail: ma_ib@encs.concordia.ca). iv. The quadrant in which the minor loop occurs.
P. Pillay is with the Department of Electrical and Computer Engineering,
While the shape of the minor loop on the ascending part of
Concordia University, Montreal, QC H3G 1M8, Canada, and also with the
University of Cape Town, Cape Town 7701, South Africa (e-mail: the hysteresis loop is different from the minor loop shape on
pillay@encs.concordia.ca). the descending part, it is observed from a series of minor loop
0093-9994 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TIA.2014.2300160, IEEE Transactions on Industry Applications
1.5 1.5
1 1
B B
Bp dB
0.5
Bm Bm Bp
0.5
Flux density (T)
dt
1.5
1.5
0
0
-1.5
-400 -300 -200 -100 0 100 200 300 400 -1.5
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Field intensity (A/m)
Time (s)
Fig. 1. Measured major and minor hysteresis loops.
Fig. 3. Flux density waveform required to generate a certain minor loop.
measurements that the influence of the minor loop quadrant on voltage is then increased until the desired peak flux density
the hysteresis loss is negligible. The same observation is also B p is reached. The hysteresis losses caused by the major and
reported in [1]. Therefore, a series of minor loop minor loops are then calculated by measuring the area
measurements is performed with all combinations of various enclosed by each loop.
values of the first three parameters B , B m and B p . The measurements are performed using an Epstein frame
The minor loops can be generated in the laminations by test fixture. The excitation signal is generated by Matlab
imposing high frequency harmonics to the applied sinusoidal Simulink interfaced with a dSPACE board, and applied to a
excitation waveform. The magnitude and position of the minor high bandwidth amplifier, which excites the Epstein frame
loop can be controlled by changing the magnitude and phase primary winding. The Epstein frame primary current and
shift of the harmonic frequency. This method is used in [1-3] secondary voltage are measured and sent back to Matlab
Simulink in order to calculate the hysteresis loops.
for minor loop measurements. However, the measured
hysteresis loops with this method represents the dynamic
III. HYSTERESIS LOSS MODELING
hysteresis loops, which includes both the hysteresis and eddy
current losses. Although the hysteresis loss component can A. Analytical
still be separated from the total measured loss, the separated For the cases where the flux waveforms in the machine core
minor loop hysteresis loss is less accurate, as it includes the are symmetric and contain only two flux reversals per cycle,
core loss separation errors in addition to the measurement the hysteresis energy loss can be represented by the modified
errors. In order to achieve accurate hysteresis loss Steinmetz equation [12] as,
measurements, the excitation waveform must have small
magnetization rate, so that the measured loops would represent a bBˆ cBˆ
Wh K h Bˆ p p p
2
(2).
the quasi-static hysteresis losses.
Neglecting skin effect, the instantaneous eddy current loss in a
the lamination can be represented by, The symmetric hysteresis losses for 0.5 mm-thick silicon
iron (M45G26) are measured and the parameters K h , a ,
2
dB b and c are then obtained from a curve fit of measured
Pe (t ) K e (1)
hysteresis loss data. The extracted parameters are shown in
dt
table I. As shown in Fig. 3, the calculated hysteresis energy
loss by equation (2) agrees well with the measured data.
where K e is dependent on the material electrical conductivity
The total hysteresis loss is then calculated by simply
and the lamination thickness. In order to keep the multiplying the static hysteresis energy loss by the
instantaneous eddy current loss constant throughout the operating frequency. However, this simplification is only
magnetization cycle, the measurements are performed under a valid at lower frequencies, where skin effect is negligible,
controlled rate of change of magnetization. To generate a as skin effect causes the peak flux density to vary across the
minor loop of a certain magnitude B , position B m and peak lamination causing the local hysteresis loops, and therefore
flux density B p , the flux reversal times are calculated based the local hysteresis energy loss per cycle to differ at
different points inside the lamination. The hysteresis loss at
on a constant dB/ dt and the corresponding flux waveform is high frequencies can be calculated by constructing the
generated in the laminations, as shown in Fig. 2. The applied magnetic field distribution inside the lamination, as shown
0093-9994 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TIA.2014.2300160, IEEE Transactions on Industry Applications
0.0467
bipolar PWM with modulation frequency ratio mf = 21 and
0.04 modulation index ma = 0.6.
0.0333 Figs. 6 and 7 show the errors of the calculated losses
0.0267
relative to the measured losses for B p equal to 1.2T and 1.6T,
0.02 respectively. It can be seen that the model results have
acceptable errors for only a few cases of minor loops, and the
0.0133
prediction errors can be as high as 150% for minor loops with
0.0067
different positions and magnitudes. Therefore, the model
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8
cannot be relied on to predict minor loop losses under a
variety of possible encountered flux waveforms in electrical
Flux density (T)
machines. Subsequently, an alternative method has to be
Fig. 4. Comparison between measured symmetric major loop hysteresis implemented in order to achieve accurate hysteresis loss
energy losses with the losses calculated by equation (2). prediction.
by the authors in [13]. However, this procedure is 0.04
Measured (Bm = Bp)
computationally expensive when the flux density waveforms 0.0356 Measured (Bm = 0.8 Bp)
are non-sinusoidal, as it becomes difficult to obtain the flux Measured (Bm = 0.6 Bp)
Hysteresis energy loss (J/kg)
0.0311
density distribution using the analytical models. Therefore, Calculated by Lavers formula
this method is not suitable for core loss determination in FE 0.0267
machine design, which requires fast core loss calculations at
0.0222
each mesh element. Therefore, the total hysteresis energy loss
is assumed to be only dependent on the flux density amplitude 0.0178
2 Bp (3).
0.5
Lavers suggested that a value of the coefficient k between
Flux density (T)
0.6 and 0.7 is suitable for the cases where B p is in the range
0
of 1.0T to 2.0T and the ratio of B / B p is relatively low. The
hysteresis losses calculated by (3) with k equal to 0.65 are 1.5
0093-9994 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TIA.2014.2300160, IEEE Transactions on Industry Applications
200
Bm = Bp In the Energetic model, the magnetic field H is calculated
Bm = 0.8 Bp from the relative magnetization ( m M / M s ) by,
150 Bm = 0.6 Bp
Bm = 0.4 Bp
Bm = 0.2 Bp
H N e M s m sgn( m ) h 1 m
1 m
1 m 1 m
g /2
1
k q
100 sgn( m m 0 ) C r H r 1 exp m m 0
% Error
0 s
M
(4).
50
The first term of equation (4) represents the linear material
behavior with N e , M s being the demagnetization factor and
0 saturation magnetization. The second term represents the no-
linear material behavior with h and g relating to saturation
field and anisotropy. The third term describes the hysteresis
50
0 0.5 1 1.5 2 effects, with k relating to hysteresis loss, q to the pinning site
B / Bp density, Cr to the grain geometry. The reversible field
Fig. 8. Errors in the minor loop hysteresis losses calculated using function H r is calculated by,
equation (3) relative to the measured losses for Bp = 1.6 T.
0093-9994 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TIA.2014.2300160, IEEE Transactions on Industry Applications
decreased until the lower field reversal point is reached, and Measured
and m 0 are recalculated. The same calculation procedure is 1 Simulated
0.5
Flux density (T)
TABLE II
Extracted Energetic model parameters
0
Ne 1.189e-5
1.5
Ms 1.432e6
h 7.332
0
g 9.957
k 82.800
-1.5
q 35.110 -400 -300 -200 -100 0 100 200 300 400
Field intensity (A/m)
Cr 0.342
Fig. 11. Comparison of measured and simulated hysteresis loops for
B 0.2B p , Bm 0.4Bp and Bp 1.4T .
1.5
2
Measured
Measured
Simulated Simulated
1 1.5
1
0.5
Flux density (T)
0.5
0
0
1.5 -0.5
-1
0
-1.5
-1.5
-400 -300 -200 -100 0 100 200 300 400 -2
Field intensity (A/m) -1000 -800 -600 -400 -200 0 200 400 600 800 1000
Field intensity (A/m)
Fig. 9. Comparison of measured and simulated hysteresis loops for
Fig. 12. Comparison of measured and simulated hysteresis loops for
B 0.4B p , Bm Bp and Bp 1.4T .
B 0.4Bp , Bm Bp and Bp 1.6T .
0093-9994 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TIA.2014.2300160, IEEE Transactions on Industry Applications
10
5
50
% Error
0 5
10
50
0 0.5 1 1.5 2
15
B / Bp
20
1 1.1 1.2 1.3 1.4 1.5 1.6 1.7
Fig. 14. Errors in the minor loop hysteresis losses calculated using the
Energetic model relative to the measured losses for Bp = 1.6 T.
Flux density (T)
Fig. 15. Comparison of the errors of symmetric major hysteresis loop
C. Hybrid model calculation by the energetic model and the modified Steinmtz equation.
Although the Energetic model is capable of simulating the
40
major hysteresis loops, the model results are not as accurate as Modified Steinmetz
the calculated losses by the modified Steinmetz equation. Fig. Energetic model
20
14 compares the errors of the calculated loss by the Energetic
model with equation (2) errors. It can be seen that precise 0
symmetric major loop loss prediction can be achieved by the
analytical formula without the need for the multiple iterations 20
% Error
0093-9994 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TIA.2014.2300160, IEEE Transactions on Industry Applications
Current (A)
Energetic model for improved major loop loss determination.
20
components. 0
Figs. 18 and 19 show the flux waveforms in some mesh
elements in the stator and rotor, respectively. It can be seen 0.5
minor hysteresis loops per cycle in the rotor core. Fig. 20 Rotor position [Degrees]
shows the simulated hysteresis loops by the Energetic model
Fig. 19. Flux density waveforms in the stator pole and the stator core.
in different points inside the machine core.
0093-9994 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TIA.2014.2300160, IEEE Transactions on Industry Applications
2 Energetic model
Rotor pole Modified Steinmetz
1.5 Rotor core Energetic model + modified Steinmetz
1
Flux density [T]
0.5
0.5
1
1.5
2
0 120 240 360 480 600 720
Rotor position [Degrees] Fig. 21. Spatial distribution of different calculation techniques used by
the hybrid model.
Fig. 20. Flux density waveforms in the rotor pole and the rotor core.
TABLE III
(a) Stator pole (b) Stator core Comparison between the machine hysteresis losses calculated by three models
2 2
1.5
1.5
1 Stator loss (W) Rotor loss (W) Total loss (W)
Flux density (T)
1
0.5
Hybrid model 6.8412 1.7705 8.6118
0
0.5
Analytical model 6.2585 1.8298 8.0883
-0.5
-1
Energetic model 7.0937 1.9940 9.0878
0
-1.5
-0.5 -2
-1000 0 1000 2000 3000 4000 5000 -1000 -500 0 500 1000 1500 2000
Field intensity (A/m) Field intensity (A/m)
hysteresis losses calculated by the analytical model are about
(c) Rotor pole (d) Rotor core 6.5% lower than the hybrid model losses. The main reason of
1.5 1.5
1 1
this divergence is that Lavers formula underestimates the
unipolar flux losses, as can be seen from the formula errors in
0.5 0.5
Flux density (T)
0093-9994 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI
10.1109/TIA.2014.2300160, IEEE Transactions on Industry Applications
calculated analytically are found to be more accurate. excitation waveforms”, IEEE Transactions on Energy Conversion, vol.
20, pp. 836 – 843, December 2005.
Consequently, a hybrid model is developed to calculate the
[19] S. D. Calverly, G. W. Jewell and J. R. Saunders “Prediction and
hysteresis losses in electrical machines using both the measurement of core losses in a high-speed switched-reluctance
analytical equations and the Energetic model. The hybrid machine” , IEEE Trans. Magn., Vol. 41 , pp. 4288 – 4298, Nov. 2005.
model is then applied to calculate the hysteresis losses in a SR [20] Z. Gmyrek, A. Boglietti, A. Cavagnino, "Iron Loss Prediction With
PWM Supply Using Low- and High-Frequency Measurements: Analysis
machine. The results show that having a model that is capable
and Results Comparison,", IEEE Transactions on Industrial Electronics,
of calculating the hysteresis losses under a variety of minor vol.55, no.4, pp.1722-1728, Apr. 2008.
hysteresis loops is essential for precise machine core loss [21] Z. Gmyrek, A. Boglietti, A. Cavagnino, "Estimation of Iron Losses in
prediction. Induction Motors: Calculation Method, Results, and Analysis," IEEE
Transactions on Industrial Electronics, vol.57, no.1, pp.161-171, Jan.
2010.
ACKNOWLEDGMENT
Maged Ibrahim (S’10) received the B.S. degree from
This project is part of the R&D program of the Natural Alexandria University, Alexandria, Egypt in 2008, and the
Sciences and Engineering Research Council Chair entitled M.S. degree in 2011 from Concordia University, Montreal,
“Energy Efficiency in Electrical Machines for Small Canada, where he is currently pursuing the Ph.D. degree in
electrical engineering.
Renewable Energy Production Systems” established in 2009 at His research interests include core loss modeling in
Concordia University. magnetic materials, design and control of electrical
machines, and power electronics.
REFERENCES Pragasen Pillay (F’05) received the B.S. degree from the
[1] T. Nakata, Y. Ishihara, and M. Nakano, “Iron losses of silicon steel core University of Kwa-Zulu Natal, Durban, South Africa, in
produced by distorted flux,” Elect. Eng. Jpn., vol. 90, pp. 10–20, 1970. 1981, the M.S. degree from the University of Kwa-Zulu,
[2] J. D. Lavers, P. P. Biringer, and H. Hollitscher, “A simple method of Natal, Durban, South Africa, in 1983, and the Ph.D.
estimating the minor loop hysteresis loss in thin laminations,” IEEE degree from Virginia Polytechnic Institute and State
Trans. Magn., vol. 14, pp. 386–388, Sept. 1978. University, Blacksburg, in 1987.
[3] P. Rupanagunta, J. S. Hsu, and W. F. Weldon, “Determination of iron Currently, he is a Professor in the Department of
core losses under influence of third-harmonic flux component,” IEEE Electrical and Computer Engineering, Concordia
Trans. Magn., vol. 27, pp. 768–777, Mar. 1991. University, Montreal, Canada, where he holds the NSERC/Hydro Quebec
[4] C. Cho, D. Son, and Y. Cho, “Core loss analysis of nonoriented Industrial Research Chair. From 1988 to 1990, he was with the University of
electrical steel under magnetic induction including higher harmonics,” J. Newcastle-upon-Tyne, Newcastle-upon-Tyne, U.K. From 1990 to 1995, he
Magn., vol. 6, p. 66, 2001. was with the University of New Orleans. From 1995 to 2007 he was with
[5] S.D. Calverley, G.W. Jewell, R.J. Saunders, “Prediction and Clarkson University, Potsdam, NY, where he held the Jean Newell
Measurement of Core Losses in a High-Speed Switched-Reluctance Distinguished Professorship in Engineering. He is also an Adjunct Professor
Machine,” IEEE Trans. Magn., Vol. 41 , pp. 4288 – 4298, Nov. 2005. at the University of Cape Town, Cape Town, South Africa. His research and
[6] M.L. Hodgdon, ”Mathematical theory and calculations of magnetic teaching interests are in modeling, design, and control of electric motors and
hysteresis curves”, IEEE Trans. Magn. 24(6) (1988). drives for industrial and alternate energy applications.
[7] F. Preisach, “Uber die magnetische nachwrikung,” Zeitschrift fur Dr. Pillay is a member of the IEEE Power Engineering, IEEE Industry
Physik, vol. B 94, pp. 277–302, 1935. Applications (IAS), IEEE Industrial Electronics, and IEEE Power Electronics
[8] I. D. Mayergoyz, Mathematical Models of Hysteresis. New York: Societies. He is a member of the Electric Machines Committee and Past
Springer-Verlag, 1991. Chairman of the IEEE Industrial Drives Committee of the IAS, Past Chairman
[9] D.C. Jiles, D.L. Atherton, “Theory of ferromagnetic hysteresis,” J. of the Induction Machinery Subcommittee of the IEEE Power Engineering
Magn. Magn. Mater. 61 (1986) 48. Society, Past Chairman of the Awards Committee of the IAS Industrial Power
[10] H. Hauser, “Energetic model of ferromagnetic hysteresis: Isotropic Conversion Department. He has organized and taught short courses in electric
magnetization,” J. Appl. Phys. 96, 2753 (2004). drives at IAS Annual Meetings. He is a Fellow of the Institution of Electrical
[11] E. Barbisio, F. Fiorillo, and C.Ragusa “Predicting loss in magnetic steels Engineers and Technologists, U.K., and a Chartered Electrical Engineer in the
under arbitrary induction waveform and with minor hysteresis loops,” U.K. He is also a Member of the Academy of Science of South Africa. He was
IEEE Trans. Magn., vol. 40, pp. 1810 - 1819, Jul. 2004. a recipient of a Fulbright Scholarship for his Ph.D and received the Order of
[12] Y. Chen and P. Pillay, “An improved formula for lamination core loss Mapungubwe from the President of South Africa in 2008 for contributions to
calculations in machines operating with high frequency and high flu South Africa in the area of energy conservation.
density excitation,” proc. of IEEE 37th IAS Annual Meeting, vol. 2, pp.
759– 766, 13-18 Oct 2002.
[13] M. Ibrahim and P.Pillay, “Advanced Testing and Modeling of Magnetic
Materials Including a New Method of Core Loss Separation for
Electrical Machines,” proc. of Energy Conversion Congress and
Exposition conference, pp. 706 -713, 17-22 Sep 2011.
[14] Y. Zhang, M. Cheng, P.Pillay and B. Helenbrook”High order finite
element model for core loss assessment in a hysteresis magnetic
lamination,”J. Appl. Phys. 106, 043911 (2009)
[15] Y. Hayashi and T. J. E. Miller, “A new approach to calculating core
losses in the SRM,” IEEE Trans on. Ind. Appl., vol. 31, no. 5, pp. 1039–
1046, Sep./Oct. 1995.
[16] P. Materu and R. Krishnan, “Estimation of switched reluctnace motor
losses,” in Proc. Conf. Rec. IEEE Ind. Appl. Soc. Annu. Meeting, vol. 1,
Oct. 2–7, 1988, pp. 79–90.
[17] J. Faiz and M. B. B. Sharafin, “Core losses estimation in a multiple teeth
per stator pole switched reluctance motor”, IEEE Trans. Magn., vol. 30,
no. 2, pp. 189–195, Mar. 1994.
[18] T. L. Mthombeni and P. Pillay, “Lamination core losses in motors with
nonsinusoidal excitation with particular reference to pwm and srm
0093-9994 (c) 2013 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.