Research On Noise Reduction And Feature Extraction Algorithm For Rolling Bearing

Rolling bearings are widely used in large machinery such as rail transportation,wind power generation and aerospace.As a general component,the running state of rolling bearings directly affects the performance of the whole mechanical equipment.Due to the complex working condition of mechanical equipment,the process of bearing vibration data acquisition is inevitably interfered by measurement noise and environmental noise.If the noise is not processed,the health monitoring and fault diagnosis of the bearing can become difficult.Therefore,it is of great significance to reduce the noise of the bearing vibration signal.In the process of bearing fault diagnosis,the features of vibration signals need to be extracted.And different features have different sensitivity to fault classification.It is of great significance for bearing fault diagnosis to study how to select the subset of sensitive features with the best sensitivity to fault classification.This thesis focuses on the noise reduction and feature extraction of rolling bearing vibration signals.The main research contents are as follows:(1)For the noise reduction of rolling bearing vibration signals,an enhanced VMD wavelet threshold denoising method is proposed.Based on VMD wavelet threshold denoising algorithm,kurtosis index sensitive to fault impact is introduced in the proposed method.The components sensitive to fault impact are selected by the kurtosis values of the component signals.The sensitive components are subjected to wavelet threshold processing,and the heavy background noise of the bearing vibration signal is reduced efficiently.The effectiveness of the proposed method is verified by simulation and experiment.(2)For the sensitive feature selection of bearing faults,an improved distance evaluation technique is proposed.Firstly,the traditional distance evaluation technique is used to remove insensitive features,and the correlation analysis is carried out on the extracted sensitive features.Then,the sensitivity and redundancy of sensitive features are comprehensively considered,and finally the removal of insensitive features in the total feature set and redundant features in the sensitive feature set is realized.Based on Kmeans clustering algorithm,a bearing fault classification model is established to classify different types of bearing faults.The effectiveness and superiority of the proposed method in the extraction of sensitive features and fault classification are verified by data sets and experiments.(3)Based on the research of bearing vibration signal noise reduction and fault feature extraction algorithm,a monitoring system for motor bearing running state is compiled.The system includes four modules: data acquisition module,data processing module,data management module and interface display module.Multiple devices can be monitored simultaneously,and the functions of signal acquisition,signal noise reduction,fault classification,data storage and playback can be realized by the system.The application ability of the proposed noise reduction algorithm and feature extraction algorithm for bearing vibration signals in practical engineering is proved.