MOMEDA and compressed sensing rolling bearing fault diagnosis method

Abstract

The invention discloses an MOMEDA and compressed sensing rolling bearing fault diagnosis method, and relates to the technical field of intelligent detection. The method can perform intelligent diagnosis on an early fault bearing under the condition that the bearing is not disassembled, and huge pressure caused by data storage, signal transmission and subsequent data processing on related hardware can be relieved through the implementation of compressed sensing software. The specific scheme is as follows: the method comprises the following steps: S1, collecting an original signal; S2, carrying out noise reduction preprocessing on the original signal; S3, taking a K-SVD dictionary learning algorithm as a sparse representation method of the signals; S4, carrying out compression measurement on the sparse signal processed in the S3; S5, carrying out precise reconstruction on the compressed signal processed in the step S4, and obtaining a reconstruction error; S6, comparing the reconstruction errors of the same sample in different dictionaries, and selecting the dictionary type with the minimum error as the type of the test bearing; and S7, diagnosing the fault of the rolling bearing.

Description

technical field [0001] The invention relates to the technical field of intelligent detection, more specifically, it relates to a rolling bearing fault diagnosis method based on MOMEDA and compression sensing. Background technique [0002] The traditional signal processing method takes the stationary signal as the research object, and performs statistical analysis on the signal in the time domain, frequency domain or time-frequency domain. Among them, the time-domain analysis directly counts some time-domain indicators of the bearing vibration acceleration signal with time as the independent variable. Different time-domain indicators are used to reflect bearing status information from different angles. For example, root mean square value is used as a time-domain indicator to reflect signal energy. It shows an increasing trend; the kurtosis index can reflect the distribution characteristics of the vibration signal and is highly sensitive to the impact component, so it is ofte...

AI Technical Summary

Problems solved by technology

The MOMEDA algorithm does not need iteration, and can efficiently and accurately extract the fault pulse signal in the measured rolling bearing signal, but this algorithm is also not perfect: before using the MOMEDA algorithm, two parameters, the filter period and the filter length, need to be manually predefined. The settings are often subjective and unscientific: although the fault impact cycle can be identified by artificially viewing the multipoint kurtosis (MKurt) spectrum of the signal to be sought, the calculation of the Mkurt spectrum takes a long time and is complex High, in order to improve the operation efficiency, it is still necessary to set the calculation range of the Mkurt spectrum in advance

An unreasonable filter length will also directly affect the filtering effect of the algorithm: if L is too short, the filtering effect will be poor; if L is too long, a large amount of useful information in the original signal may be filtered out, and the gain outweighs the gain.

For now, there is also no definite basis for adaptive selection of the filter length L

Images