Real-Time EV Powertrain Fault Diagnostics Using IoT and FMEAGuided Machine Learning Strategy

Inhalt:
This paper introduces a retrofittable IoT-enabled Onboard Diagnostics (OBD) system for Electric Vehicles (EVs), designed to acquire real-time data from critical powertrain components, including the motor, controller, and battery. The system monitors a comprehensive set of parameters—such as bootstrap voltage, DC link and line voltages, battery and phase currents, MOSFET switching behavior, speed, motor vibration, vehicle lean angle, and temperatures—facilitating advanced condition monitoring and fault diagnostics through machine learning (ML). A decision tree algorithm, guided by Failure Modes and Effects Analysis (FMEA) principles, computes Risk Priority Numbers (RPN) to prioritize critical failure modes, enabling early fault detection. Experimental validation on an actual EV demonstrates the system's real-time prediction capabilities, achieving a 92% accuracy with a 100 Hz sampling rate and 50 ms. This approach enhances powertrain reliability by supporting proactive interventions and refined prognostic decision-making, surpassing the limitations of conventional threshold-based OBD systems. The integration of IoT and ML offers a scalable framework for dynamic fault classification and maintenance optimization, with potential applications in fleet-wide diagnostics.