Rolling Bearing Fault Diagnosis using SVM,Frequency Kurtosis and Entropy Techniques

Using vibration signature analysis as the main and only source of information from the machining process, this paper presents a new procedure for rolling bearing fault diagnosis. In this procedure, firstly, in order to calculate the feature vector, we propose the use a combination of Frequency Kurtosis and Entropy techniques for determining the kurtosis and entropy values in each one of the all sub-bands of the last level of the fast kurtogram developed by Antoni that use the multi-rate filter-bank techniques. Lastly, using the calculated feature vector, the proposed procedure is based on Support Vector Machine SVM as a classifier system. In the experimental step, twelve different health bearing conditions were introduced to demonstrate the effectiveness and robustness of the proposed procedure.