A State-of-Art Approach on Fault Detection in Three Phase Induction Motor using AI Techniques

Abstract

Three-phase induction motors are the key factors of electromechanical electricity conversion for a range of industrial sectors. The potential to discover motor faults earlier than they appear can minimize the dangers in choices involving computing device maintenance and decrease costs. In particular, the faults in the electrical machine should be resolved at an appropriate time to head off damages. The fault location depends on condition checking methods and utilization of Artificial Intelligence (AI) techniques. This is related with the classification problem, which normally is what Machine Learning algorithms are designed to solve. Three- phase induction motor intended to work on the most well-known fault is a broken rotor bar, which represents around 10% of complete induction motor issues. If a Broken rotor bar arises, it could restrict a current flow in that bar. This research work proposes a novel supervised classification method for Induction Motor (IM) faults that is based on the Logit boosting (Meta.logitboost) algorithm and uses an optimized sampling technique to deal with the unbalanced experimental dataset. To compose a data set of features from the frequency domain, the rotor current and vibration signal is used. The experimental results show that the method proposed achieves higher performance metrics for the incipient identification and classification of IM faults than other classifiers used in this field. The proposed approach can be utilized in industry for online monitoring and fault diagnosis of three-phase induction motors and the outcomes can be useful for the foundation of preventive upcoming plans in production lines.