USE OF DISCRETE WAVELET ANALYSIS OF VIBRO-ACOUSTIC SIGNALS TO DETECT BEARING DEFECTS OF ROTATING ELECTRIC MACHINES

Abstract

The paper shows that one of the most effective methods for processing the time realization of a vibro-acoustic signal from electrical machines is a discrete wavelet transform. At the same time, due to the lack of built-in tools for separating periodic and aperiodic components in the latter, it significantly complicates the process of identifying periodic disturbing factors, which makes it necessary to develop new approaches to identifying periodic components of vibration based on discrete wavelet transform, which can be caused by a number of defects in rotating electrical machines. A new integral highly informative numerical criterion for the presence of bearing defects, obtained on the basis of a discrete wavelet transform of a vibro-acoustic signal of an electric machine, is proposed. It has been found that upon detection of the indicated defect, it is advisable to analyze the frequency bands, which include the rotor frequency and its first harmonic components. The expediency of using the parent wavelet function of higher orders upon detection of the indicated defect has been theoretically substantiated and experimentally proved. It is shown that, based on considerations of simplifying the mathematical apparatus of calculation and increasing the flexibility of the diagnostic system, it is advisable to use the Daubechies wavelet family, the calculation of the coefficients of which involves the use of a clear mathematical algorithm. The necessary duration of time realizations of vibration is theoretically substantiated and experimentally confirmed, which is advisable to use when determining the proposed highly informative numerical criterion. It is shown that the duration of such realizations should be significantly longer than the period of rotation of the rotor of the electric machine.