Fault diagnosis of roller bearing based on PCA and multi-class support vector machine

Abstract

This paper discusses the fault features selection using principal component analysis and using multi-class support vector machine (MSVM) for bearing faults classification. The bearings vibration signal is obtained from experiment in accordance with the following conditions: normal bearing, bearing with inner race fault, bearing with outer race fault and bearings with balls fault. Statistical parameters of vibration signal such as mean, standard deviation, sample variance, kurtosis, skewness, etc, are processed with principal component analysis (PCA) for extracting the optimal features and reducing the dimension of original features. The multi-class classification algorithm of support vector machine (SVM), one against one strategy, is used for bearing multi-class fault diagnosis. The performance of the method proposed was high accurate and efficient. © 2011 IFIP International Federation for Information Processing.