Battery Based Diagnostic System for Engine Anomaly Detection in 48V Mild Hybrid Electric Vehicles (MHEVs)

This work demonstrates a new battery based diagnostic methodology for early indication of engine irregularities in 48V MHEVs. Rather than dependent on direct sensors of the engine, the innovative method utilizes the vehicle's battery voltage as an indirect indicator to monitor engine functioning and detect malfunctions. There is a correlation analysis between the engine RPM and battery voltage undertaken to detect essential patterns representing standard functioning or fault inception. Guided by these observations, a Support Vector Regression (SVR) model is constructed, using window-based statistical measures and peak feature engineering to identify subtle anomaly signatures. Smoothing filters and time-windowed mapping methods are used on the data stream to improve robustness and minimize noise, allowing for accurate anomaly detection. The final goal is to deploy this approach as a real-time cyber-physical diagnostic system for ongoing engine health monitoring. The paper concludes with a complete setup of the system's anomaly detection accuracy, consistency, and generalization ability under various driving conditions.