Affiliations: [a] Department of Automatic Control and Information Technology, Cracow University of Technology, Kraków, Poland | [b] Institute of Electromechanical Energy Conversion, Cracow University of Technology, Kraków, Poland
Correspondence:
[*]
Corresponding author: Mieczysław Zajac, Department of Automatic Control and Information Technology, Cracow University of Technology, Kraków, Poland. E-mail: mzaj@pk.edu.pl.
Abstract: The subject of this work concerns the problem of signal diagnostics of induction machines carried out in the time-and-frequency domain. It was assumed that the signal carrying the important information about the current state of the machine is the measured signal of the axial flux. The basis for the diagnostic decision making was the analysis of the energy of the detailed wavelet representation, obtained as a result of the use of a specially constructed orthonormal transformation basis. To ensure sufficient resolution of fault detection, the authors constructed orthonormal wavelet basis based on the recursive formulas for generating higher order Legendre polynomials. The construction of the orthonormal basis was carried out using the scaling function, which was designed using the appropriately scaled Legendre polynomial. The synthesis of the basic wavelet and the orthonormal basis was based on the recurrent dependencies of the multi-resolution analysis using the relationship between the elements of the quadrature mirror filters. Dependencies between elements from two consecutive levels of decomposition were used. By applying the constructed orthonormal basis, the research on the nature of the energy flow of an axial flux signal of an induction motor between adjacent frequency bandwidths in steady-state depending on the machine load was presented. Studies have shown that significant flows of the diagnostic signal energy occur between the adjacent frequency bandwidths periodically at frequency of about 2.5 Hz. This periodic energy flows of the axial flux signal hamper fault detection and diagnostic decision making. The analysis results confirmed that it is possible to ensure a sufficient level of selectivity of a diagnostic system by means of this method. To ensure a sufficient efficiency of diagnostic systems it is required to separate flow effects from fault symptoms.
