DCT PCA based Vibration Signal Analysis with Optimized SVM for Induction Motor Bearing Fault Detection and Classification

Abstract

In this paper an induction motor bearing fault detection algorithm based on PCA and optimized kernel SVM is presented.  The concept behind using the PCA is to express the vibration signal's large 1-D vector into the feature space's compact principal components. This is known as Eigen-space projection.  Eigen-space is determined by defining the covariance matrix's eigenvectors which are derived from a vibration signal collection (vectors). In our proposal first, the vibration signal is transformed into the frequency domain and then the PCA is used for feature extraction and finally, the SVM kernel optimization is used to get the best classification accuracy from SVM. Optimization increases the accuracy of SVM since the consistency of SVM models depends greatly on setting SVM kernel parameters correctly. Therefore, to determine the values of these parameters which lead to the best performance PSO-based approach is adopted for the optimization of SVM parameters. The findings show that the proposed model has great potential for bearings fault detection in induction.