Single Phase Fault Detection of Induction Motor using Machine Learning Approaches

The induction motor (IM), an asynchronous type of AC electric motor, plays a crucial role in operational procedures in industrial sectors, which needs to be operated sophisticatedly without any error. This research investigates the occurrence of single-phase faults, which are commonly observed in induction motors, among various other types of failures, through detection and classification employing machine learning (ML) tools. This research addresses the machine's condition based on three operational modes, which include the healthy case, 5% fault and 10% fault of induction motors. In the case of generating the dataset for implementation of ML tools, simple d-axis and q-axis conversions are considered for a healthy situation of IM. However, on the other hand, Park's transformation is made in modeling the faulty IM by transforming it from a phase to a two-phase system for accumulating the faulty dataset. Several electrical features of IM are considered regarding generating healthy and faulty datasets for training the ML models so that they can detect and classify the operational mode of IM. Two well-known statistical features, namely the mean and standard deviation, are chosen to measure the performance of the ML models in detecting and identifying the motor operating conditions. Several ML models are implemented to the model of the machine in testing the robustness of the fault identifying and diagnosis procedure where the Random Forest algorithm shows the best performance with 99.9% accuracy.