Rolling Bearing Fault Diagnosis Based On Block Matching Pursuit And Data Enhancement

In modern society,the reliability of rolling bearing is getting more important.It has great significance to ensure the stable operation of equipment by analyzing the fault types and fault part.In actual engineering,such as high-speed rail and space shuttle,data collecting and processing are often not carried out in the different places.The huge data will cause enormous pressure on transmission and storage.Therefore,this paper applies the Compressed Sensing Algorithm into the bearing fault diagnosis.This measure has significance for remote fault diagnosis.In order to realize the remote synchronization fault diagnosis process of rolling bearings,a fault diagnosis technology of rolling bearings based on block matching tracking is proposed.The main research contents are as follows:Firstly,the block method in Compressed Sensing is studied.Researching an adaptive block method which is aimed to improve the problems existing in the method of blocking based on empirical values.This method searches the compound cycle of the fault signal by short-term auto-correlation function,uses the length of compound cycle as the block length to avoid the large disparity of information carried by different signal blocks.The improved adaptive block method can effectively equalize the sparsity of the signal block and ensure the overall reconstruction effect of the compressed signal during the reconstruction process.Secondly,the Stagewise Orthogonal Matching Pursuit has a poor reconstruction effect while the sparsity of the signal is large.In order to solve this problem,the Forward and Backward Stagewise Orthogonal Matching Pursuit algorithm is put forward.The FBSt OMP algorithm ameliorates the selection method of St OMP algorithm by adding the invalid support set atoms elimination process and effective support set atoms secondary selection process into the reconstruction support set.This step increased the probability that all effective support set atoms can be selected into the reconstruction support set,and effectively improves the algorithm's successful reconstruction rate.Thus,the FBStOMP algorithm can reconstruct the signal which has large sparsity.The difference between the signal reconstructed by the FBStOMP algorithm and the original signal is small,and thereconstructed signal retains more information of the original signal,which is conducive to the subsequent fault type diagnosis.Finally,aiming at the problem of low accuracy of fault diagnosis during reconstruction of fault signals with similar impact amplitudes for inner ring faults,rolling element faults and outer ring faults,this paper proposes a fault diagnosis method based on data enhancement.This method enhances the amplitude of the impact component and the time domain feature quantity of the signal through data enhancement,so as to ensure that the fault features in the data can be significantly enhanced,and promote significant difference between different types of fault feature quantities,which effectively improves the success rate of fault diagnosis.