Fault Diagnosis for Automotive Electric Machines Based on a Combined Machine Learning and Parameter Estimation Method: An Approch for Predective Maintenance

Faults within the main electric machines of the vehicle affect directly its performance and its functional safety. This work introduces a novel approach to diagnose faults and identifies them according to type and location. The approach is a combination of parameter estimation and machine learning classification algorithm, which provides a complete mapping of the model for both the healthy and faulty states of the machine. For fault detection, training and testing the accuracy of the machine learning algorithms are relatively high and can be regarded as a successful implementation in this field. In addition to fault detection, this approach allows analyzing in depth the effect of faults on the performance of the machine in time domain, which makes it convenient for the digital twin concept.