Multi-sensor mixed fault signal blind separation method based on edge calculation and machine learning

Abstract

The invention discloses a multi-sensor mixed fault signal blind separation method based on edge calculation and machine learning, and the method comprises the following steps: determining the number and positions of observation fault signals according to the number of existing faults, and collecting the observation signals through a signal collection device; preprocessing the collected mixed signal to obtain a preprocessed aliasing fault signal; separating the aliasing fault signals by adopting an independent component blind separation algorithm to obtain multiple paths of separated independent fault signals; performing normalization processing on the separated fault signals, performing spectrum and wavelet analysis, and extracting fault feature signals; and diagnosing and discriminating each extracted single fault characteristic signal. According to the invention, possible coexistence fault feature signals are effectively separated, complex coexistence fault diagnosis is converted into single fault diagnosis, the diagnosis efficiency and precision of composite faults can be greatly improved, powerful technical support is provided for safe operation and maintenance of mechanical equipment, and the prevention and treatment capacity for accidents is effectively improved.

Description

technical field [0001] The invention relates to a signal processing method in the technical field of mechanical engineering, in particular to a blind separation method of multi-sensor mixed fault signals based on edge computing and machine learning. Background technique [0002] In mechanical equipment, vibration is the source of sound signal generation, and sound signal is the continuation of vibration signal propagation, and the two are mutually unified whole. When a mechanical system such as a rolling bearing or a gear fails, its characteristic signal often has an obvious impact component, and at the same time the acoustic characteristics will also change, thus containing equipment status information. However, the actual sound field environment is complex and changeable, and the signal to be identified (fault source signal) is often mixed with other various interference signals or noises, which cannot be effectively identified. In order to accurately extract mechanical f...

AI Technical Summary

Problems solved by technology

[0004] In the noisy environment, the existing separation algorithm itself cannot recognize the noise, and the speech signal separated by the noise is either unable to be completed or distorted, but the various separation algorithms of the theoretical research on speech signal separation in the absence of noise environment have achieved satisfactory results. Effect

Images