Jakub Eichler

Difficulty in identification of Preisach hysteresis model weighting function using first order reversal curves method in soft magnetic materials

The Preisach model can be used for detailed analysis of devices based on ferromagnetic materials, if its parameter, its weighting function, is well-known. Usually the weighting function is approximated by analytical formula. The second approach is to determine it directly from experimental data. Most widely used method to obtain the weighting function is the first order reversal curve method that is based on two partial derivatives of measured magnetization using a special excitation pattern beginning from deep material saturation. Since the derivative enhances the experimental error, a precision experiment is necessary. Furthermore, it is not easy to achieve the deep saturation with the required signal pattern. Therefore sophisticated data processing followed, in order to reduce experimental errors before performing the numerical derivative. The paper concerns measurements errors caused by insufficient saturation and also problems of negative values of the weighting function, partially due to the noise. Irrespective of measurement errors, the agreement between model and experiment is good and fully acceptable in technical praxis.

Improvements of Preisach model for soft magnetic materials, analysis of input excitation signal

Determination of the weighting function of the Preisach hysteresis model depends strongly on the quality of experimental data. The paper focuses on errors in the input excitation current. The analysis of experimental data and their processing were identified the main source of errors: RF noise, quantization noise at signal synthesis, power net interference and short-term faults of used power supply. Their reduction is possible by insertion of the filter in between the power source and measured sample. The filter was designed with respect to main frequencies of both the useful and disturbing signal. Its simulations are validated with measurements and the improvement of measured waveforms in the time domain is demonstrated.

Implementation of the first order reversal curve method for identification of weight function in Preisach model for ferromagnetics

Usually analytical weighting function is used in Preisach model application, since the use of derivation of experimental first order reverse curves exhibits large errors. To reduce the errors, extended experiments and methods for reduction of experimental errors are presented in the paper. Weighting function obtained by numeric derivation leads to acceptable agreement with experiments in Preisach model. Further improvement can be obtained by suggested reduction of remaining errors in experimental data.

Simple analysis and use of preisach model for ferromagnetic materials

Because of hysteresis the perfect modeling of ferromagnetic materials is difficult. The general purpose of Preisach model was applied in this area. The main problem is to determine the weighting function from measurement of hysteresis loops. Another approach, based on the method of trials and errors, was used. According to the nature of a material, the weighting function was searched in the form of normal distribution with constant background. Only several parameters should be found by the best fit with experimental data. The agreement with experiment is acceptable, especially for higher excitations. The numerical model can be used for semi quantitative analysis or simulation of electrical or electronic devices with a ferromagnetic core.

Differences between Preisach model and experiment for soft ferromagnetic materials, effect of instrument accuracy

The practical use of Preisach hysteresis model is based on proper identification of weighting function that characterizes the material. Obtaining of the weighting function needs two partial derivations of two-dimensional flux density function, called Everett surface, measured by specific procedure. Since the derivatives require exact measurements, the possible source of errors are studied. Main method is to compare experimental results and simulation. The comparison showed that the excitation current source accuracy influences the quality of Everett surface. Also the numeric integration drift plays an important role. The random or systematic failure in excitation current results in the minor secondary loops on the main hysteresis loop. Simulations of these effects confirmed the predicitons. As practical output it was found that the errors in excitation can be reduced by repeated measurements. The drift can be eliminated, if a more suitable wavefrom is used.

Computation speed of numeric Preisach model

Many applications of the Preisach model need a lot of repeated calculations. Therefore, the algorithm speed is an important program parameter. Two ways of calculation acceleration are considered: improvement of the algorithm and reduction of the Preisach matrix using its specific feature. In the first case, which is universal, the calculation time can decrease about 500 times. In the second case the time saving may be lower about 50 times without decrease of accuracy. The combination of both is possible.

Improvements of Preisach model for soft magnetic materials, experiments on filtering of input excitation signal

Identification of Preisach model supposes a perfect experiment. One condition is that prescribed harmonic current without any errors flows in the toroid primary winding. The condition can be ensured by a suitable filter that was designed after the complete analysis of sources of signal errors. The filter was analysed by the use of circuit theory and a very simple experiment was realized for its measurement. Good agreement between experiment and simulation was found and small deviations explained. The filter significantly reduces all the errors in primary part of apparatus and improves the Preisach model simulation results.

Experimental-numerical method for identification of weighting function in Preisach model for ferromagnetic materials

The Preisach model weighting function is given by partial derivations of a measured magnetic flux density that leads to high errors. Therefore other methods of the weighing function estimation are used frequently. In order to verify the direct approach the extended experiment was made and data were numerically processed to reduce known errors. It results in model of the hysteresis loop usable in a technical research. The possible errors reduction and further improvements are discussed and recommended.

Effective approximation of transfer characteristics of low frequency amplifier

A detailed knowledge of amplifier transmission characteristics is important in many technical areas, for instance in commercial modeling systems as the Simulink. The paper shows that a good approximation of measured amplitude and phase characteristics can be made numerically by a rational function. If there is enough computation time, the complex transfer characteristic can be well approximated. Then zeros and poles can be found. Preliminary calculation reveals that the tested amplifier is not a minimum phase network.