Khizhnyak T. Diagnostics of semiconductor converters on the base of wavelet functions of m-ary argument

Object of investigation: process of semiconductor converter's diagnostics based on the time dependencies of currents and voltages. Aim: the development of new diagnostic methods for semiconductor converters based on the time-frequency representations of currents and voltages for degradation failure warning and nonrelevant failure' reasons identification. Methods: mathematical modelling, theoretical principles of multiresolution analysis, discrete spectral transforms on the finite time intervals, discrete wavelet transforms, classification theory, principles of the matched filters' development. Principal results: the new kernel of the wavelet transform based on the orthogonal m-ary argument functions' system was founded, which allowed to increase data volume of the semiconductor converter's processes with decreasing in 1.5 times the number of scales compared to Haar and Daubechies wavelet transorms. The new semiconductor converter's design methods based on combined using of wavelet transform and classification theory, also on the theory of the matched filters with finite impulse response in role of the m-ary argument wavelet function was developed; it allowed to numerically determine degree of current semiconductor converter's state to a failure state approach, and decrease the diagnostic's time and number of the parameters compared to current methods.