Mechanical multi-fault diagnosis method based on signal atomic-driven wavelet reproducing kernel machine learning

Abstract

The invention discloses a mechanical multi-fault diagnosis method based on signal atomic-driven wavelet reproducing kernel machine learning. The method comprises the following steps that: collecting the to-be-analyzed signal of mechanical equipment, carrying out fault setting on the mechanical equipment, and collecting the non-stationary mechanical vibration signal of the fault equipment; adoptinga Gabor atomic orthogonal matching tracking method based on an ant colony search algorithm to track and decompose the non-stationary mechanical vibration signal to obtain a plurality of atomic components and residual components optimally matched with the to-be-analyzed signal; and inputting the plurality of obtained atomic components into a machine learning-wavelet reproducing kernel support vector machine classifier as a characteristic sample to be subjected to training and testing learning so as to identify the type of a mechanical multi-fault mode. By use of the method, identification accuracy can be improved.

Description

technical field [0001] The invention relates to a mechanical multi-fault diagnosis method based on signal atom-driven wavelet regeneration kernel machine learning, and belongs to the technical field of mechanical fault diagnosis. Background technique [0002] Mechanical fault diagnosis is essentially a pattern recognition problem of the machine's operating state, and feature extraction and classifier design are the keys to pattern recognition. [0003] When diagnosing and identifying mechanical faults, it is first necessary to extract features from the fault signal. With the development of signal processing technology, various new signal time-frequency analysis has been introduced into the field of fault diagnosis. For example: Fourier transform method, wavelet transform method and atomic decomposition algorithm. Fourier transform is the most commonly used method to deal with stationary signals. However, when FF'I' algorithm is used to analyze non-stationary signals with m...

AI Technical Summary

Problems solved by technology

However, the regularization method is used to transform it into a well-posed problem to solve. The regenerating kernel and its corresponding regenerating kernel Hilbert space play an important role in function approximation and regularization theory, and the construction of regenerating kernel function is the core of the kernel sparse representation classification model. The key is that a perfect theory has not yet been formed

Images