Research Of Bearing Fault Signal Based On Wavelet Transform And Recurrence Quantification Analysis

With the development of modern technology, rotating machinery is alsodeveloping in the direction of high-speed and automatic, which requires the higherrunning reliability of the critical components of rotating machinery. Therefore,increasing the accuracy of fault diagnosis becomes more important. Fault featureextraction is one of the major steps of fault diagnosis. Extracted fault signal featureis effective or not, which directly affects the result of fault diagnosis, so the researchof extracting fault feature is necessary.Firstly, this paper describes researches on the fault signal feature extraction ofrotating machinery at home and aboard, introduces phase space reconstructiontheory and discusses the selection of the delay time and the minimum embeddingdimension, and proposes using mutual information method to select the delay timeand using false nearest neighbors method to select minimum embedding dimension,meanwhile introduces recurrence plot and recurrence quantification analysisrespectively.In this paper, we took the bearing fault signals of rotating machinery as theresearch object. We can achieve the feature extraction of bearing fault signals byusing the methods of wavelet transform and recurrence quantification analysis,which is regarded as input vector, then we can achieve the classification of thebearing fault signals based on probabilistic neural network.Through the experiment of rotating machinery bearing failure, we test andverify that feature can be effective extracted based on recurrence quantificationanalysis. The results of fault classification experiment and fault estimationexperiment show that wavelet transform and recurrence quantification analysiscombind method can more effectively extract the features of the fault signals.Recurrence quantification analysis provides a new way to the study of faultdiagnosis for mechanical system, since it’s validity in fault feature extraction ofvibration signal. Fault signal information will be to represent by simple graphic,which makes the fault signal information visualizations, so it is of great significancein theory and practical application.