Research on signal decomposition algorithm of mechanical system based on multiwavelet

Abstract

Wavelets are referred to as "mathematical microscopes" which can effectively extract the effective state information from the signals of mechanical systems. Multiwavelet is made up of two or more scale functions. It is easier to extract from mechanical system signals than single wavelets. However, in the decomposition of the multiwavelet mechanical system signal, it does not have obvious high and low pass characteristics, and its equivalent filter does not have the ideal frequency domain characteristics. As a result, a frequency band mixing problem will inevitably arise during the decomposition of the mechanical system signal. In order to solve the problem and get a purer decomposed signal, this paper proposes a harmonic wavelet-based band aliasing suppression method. The harmonic wavelet is tightly branched in the frequency domain. It has well-defined functional expression and possesses a fully boxed spectrum that allows the mechanical system signal to be processed in the frequency domain to obtain the desired frequency band information. Therefore, the harmonic wavelet transform can be performed on the mechanical system signals of different frequency bands after multiwavelet decomposition to obtain a purer mechanical system signal in that frequency band. The feasibility and validity of the method are demonstrated through the simulated and measured mechanical system signals, and the comparison with the decomposed mechanical system signals without harmonic wavelet transform shows the superiority of the method and its potential value for application in intelligent maintenance and health diagnosis of energy equipment.