Sensor Fault Detection using Machine Learning Technique for Automobile Drive Applications

Electric motors and power electronic converters are the critical elements in industrial and automotive electric drives. It is critical to identify motor drive problems since they can prove to be a safety concern in applications like electric traction, propulsion, and aviation. In electrical drives, current sensor faults occur due to mechanical stress, signal conditioning, and hardware failure, hence this work is focused on sensor faults detection and identification. This paper presents a fault diagnostics system for detecting multiple faults in an electric motor using a machine learning technique. A direct torque control simulation model for a 4 kW induction motor has been developed. Different forms of current sensor faults have been incorporated into the system, including biasing and signal omission-based faults. The built-in machine learning framework is utilized to identify and classify the different failures. Each of the three families of classifiers employed achieves an accuracy of more than 90%. Matlab/Simulink is used for simulating the model with machine learning frameworks.