Misalignment Fault Prediction of Motor-Shaft Using Multiscale Entropy and Support Vector Machine

Abstract

Rotating machines constitutes the major portion of the industrial sector. In case of rotating machines, misalignment has been observed to be one of the most common faults which can be regarded as a cause for decrease in efficiency and can also for the failure at a time. Till date the researchers have dealt only with the vibration samples for misalignment fault detection, whereas in the present work both stator current samples and vibration samples has been used as a diagnostic media for fault detection. Multiscale entropy (MSE) based statistical approach for feature extraction and support vector machine (SVM) classification makes the proposed algorithm more robust. Thus, any non-linear behavior in the diagnostic media is easily handled. The proposed work has depicted an approach to analyze features that distinguishes the vibration as well as current samples of a normal induction motor from that of a misaligned one. The result shows that the proposed novel approach is very effective to predict the misalignment fault for the induction motor.