Base On LMD Method In Roller Bearing Fault Diagnosis Research

Roller bearing is a common part of mechanical and electrical equipment. It is very importance to pay attention to fault monitoring and diagnosis during the operation of mechanical and electrical equipment. The vibration signal contains a wealth of fault information, analyze and process the information to obtain useful information about state changes, and then determine whether the mechanical equipment and parts run properly or not. Time-frequency analysis method is widely used in mechanical fault feature extraction. Based on this background, this paper introduces a new time-frequency analysis - the local mean decomposition, using a local mean decomposition and neural networks and SVMs and SVD method to study on roller bearing fault diagnosis.The main work is as follows:1. The applications and limitations of conventional time frequency analysis method in signal processing are briefly discussed. The analysis results show that the method is sensitive to the changes of transient signal in frequency and amplitude, besides, it is better than the EMD method in reducing the number of iterations and dealing with end effect problem.2. Combine LMD with neural networks, the method based on LMD and neural networks is proposed. Take the LMD method as a pre-processor, extract fault feature from the first few PF component which contains the main information in the roller bearing. Compared with extracting fault feature based wavelet packet decomposition, indicating that the based LMD and method can effectively identify the roller bearing fault and has a better recognition ability.3. Combine wavelet packet with LMD to improve the original method that is adopting LMD method to decompose the reconstructed signal after wavelet packet. Targeting the characterstics that periodic impulses signal often occur when the roller bearing exhibits local faults and the technical defects of the singular value decomposition, a method of feature extraction based on improved LMD and SVD is put forward. Vibration signal of the roller bearing with fault is analyzed using this method, experimental results show that the method based on SVD and improved LMD and SVMs have a high recognition rate in fault diagnosis, so as provides a new method to roller bearing fault diagnosis.