A fault diagnosis approach using SVM with data dimension reduction by PCA and LDA method

Abstract

As the development of modern industry, fault diagnosis technology is becoming one of the major issues. A novel fault diagnosis approach is presented for rolling bearing system based on support vector machine (SVM) in this paper. Data acquired for rotatory machinery has a large quantity as well as a high dimensionality, both principal component analysis (PCA) method and linear discriminant analysis (LDA) method are introduced to solve the problem of high dimensionality. In the proposed approach, PCA and LDA are separately applied to reduce data dimensions and extract data features from raw data first. Then the classic SVM method is used to classify fault types. With PCA and LDA methods applied, reduced number of inputs leads to fewer iterations and less time of training procedures. The experiment results show that accuracy of fault diagnosis has improved and cost of SVM training has decreased evidently, and prove that the proposed approach is practical and efficient for fault diagnosis of the rolling bearing system.