The Diagnostic Methods Of Bearing Failure Based On Correlation Analysis Of Output Signals

The work conditions of rolling bearing have a direct influence on the performance of the equipment and even the production and safety of the entire production line.Therefore,it is great theoretical and practical significance to study the corresponding technology on bearing fault identification in order to avoid the occurrence of major mechanical accidents.In this paper,the rolling bearing is taken as the research object.A series of research work is carried out on the two key issues:feature selection and fault identification,using the method and idea of correlation analysis.The main work is as follows:(1)A self-power spectral function classification method(SPSF)is proposed.Firstly,the self-power spectral function is adopted to reduce noise in order to enhance the frequency characteristics of the fault signals.Secondly,the correlation between the measured signal sequence and the priori signal sequence set is analyzed by similarity of self-power spectral functions and the values of similarity between signals are obtained.Finally,classify the type of fault signals according to the values of similarity.(2)A two-stage feature selection method based on MIC method is proposed.Firstly,the MIC method is used to analyze,measure and rank the features with strong expressive ability.Secondly,similarity analysis and measurement of selected strong feature subset is performed to achieve the purpose of redundancy.Finally,the performance of the feature subset is evaluated based on the classification results of given classifiers.(3)A MIC-k-means feature selection method is proposed.On the basis of selected feature subset with strong expressive ability,the MIC method is used to establish the similarity matrix between features.Then,using k-means clustering method to cluster features and get clustering centers.Finally,selecting one of the most representative features in each cluster and obtaining the optimal feature subset.