Intelligent multiple fault diagnosis for predictive maintenance of induction motor

This work proposes a machine learning-based bearing and rotor fault detection technique with an improved feature set and processing time. Predictive maintenance uses intelligent artificial learning techniques, which consist of problems such as high predictive time due to a massive amount of data and improper classifier parameters, which lead to less reliable and lower accuracy fault detection intelligent models. Both problems can lead to huge losses because of the high false positive results. The problems of massive data and more extended detection periods have been solved using the Univariate Feature Selection technique. The improper classifier parameter problem is solved using the Grid Search optimization technique. In this work, two majorly occurring faults, the broken rotor bar (BRB) and Bearing faults, have been considered. The acquired current signal for different conditions, such as healthy, bearing inner cage fault, bearing outer cage fault, one broken rotor bar, and three broken rotor bars, is divided using the Average Frequency Band Decomposition (AFBD) approach. An experimentally acquired data set is used to validate the proposed methodology. The data set is provided for training and testing the Support Vector Machine (SVM) classifier, whose parameters include ' regularization factor (C), ''gamma,' 'kernel,' and 'decision function shape' are tuned using Grid Search Algorithm. A resulting model with high accuracy of 99.7% was obtained with a 235 ms drop in detection time.