Rolling bearing health state evaluation method based on Hankel matrix

Abstract

The invention discloses a rolling bearing health state evaluation method based on a Hankel matrix. The rolling bearing health state evaluation method comprises the steps of segmenting a vibration timedomain signal in an existing health state and establishing a corresponding Hankel matrix for an obtained segmented signal; secondly, establishing a Hankel matrix of each segmented signal and calculating a feature vector of an average Hankel matrix; carrying out matrix decomposition on the average Hankel matrix by utilizing the feature vector and decomposing the decomposed matrix into a diagonal matrix and a non-diagonal matrix; calculating a 1-norm of the non-diagonal matrix and setting 10 times of the 1-norm of the non-diagonal matrix as a threshold value; establishing a corresponding Hankelmatrix for a to-be-detected time domain signal, performing matrix decomposition on the graph connection matrix by using the feature vectors obtained in the previous process, and calculating an abnormal value; and if the time domain signal is greater than the threshold value, considering the rolling bearing corresponding to the time domain signal to be in an unhealthy state. According to the invention, the problems of poor learning efficiency and the like of a traditional machine learning method are solved, and the rolling bearing performance health state can be quickly identified.

Description

technical field [0001] The invention relates to the technical field of equipment operation state evaluation, in particular to a method for evaluating the health state of rolling bearings based on Hankel matrix. Background technique [0002] Mechanical system fault detection is of great significance for reducing the downtime of mechanical systems and preventing catastrophic failures. Currently, many algorithms have been proposed for fault feature extraction, but how to monitor the state of a mechanical system from a signal containing a large amount of interference noise is still a challenge. Feature extraction in time domain, frequency domain, and time-frequency domain is an important means to realize the health status evaluation of rolling bearings. In order to robustly evaluate the health status of rolling bearings, a variety of machine learning methods have been proposed and achieved good results. However, the incomplete life cycle data leads to low learning efficiency o...

AI Technical Summary

Problems solved by technology

However, the incomplete life cycle data leads to low learning efficiency of machine learning methods

Images