Deep Neural Network Based Bearing Fault Diagnosis of Induction Motor Using Fast Fourier Transform Analysis

Abstract

The demand of condition monitoring of Induction Motor is progressively increasing and the fault occurring must be considered as major issue because it prevents induction motor from failing and breaking down. Considering the maintenance cost and unscheduled downtime, the bearing fault has become the significant topic and many fault detection methods have been proposed. Predominantly, pitting is considered as a faulty factor in most of the cases. This paper is motivated by considering the practical fault occurrence, introducing the scratch on the outer raceway of the bearing. An online bearing diagnosis method is proposed using a deep learning (DL) based approach. A Convolutional Neural Network (CNN) architecture is originally used for fault characterization. Specifically, fast Fourier transform analysis is carried out using the load current of the stator, followed by the feature extraction of selected frequency components which are used to train the CNN algorithm. The effectiveness of the proposed approach is verified by series of experimental tests corresponding to different bearing fault conditions. The proposed method is also tested to detect the multiple faults and the application gets extended.