Diagnostics for Automated Electric Vehicles: Batteries Short-Term Faults Predictions

The innovation in automotive industry is led by trends like vehicle electrification and automation. Following such trends, vehicle safety and diagnostics are always a primary focus when designing such vehicles and their systems. The limitations of current diagnostic systems raise concerns about the overall safety of automated electric vehicles. Electrical propulsion systems introduce new faults, while driverless automation eliminates manual monitoring of critical components. This paper proposes a novel solution for predictive diagnostics for automated electric vehicle battery systems. The approach presents a unique two-stage diagnostic process using SVM and CNN to predict various electrical faults in real-time. The solution is verified on the prototypical fully automated vehicle gave results with high precision of utilized models (96.43% for SVM and 99.85% for 1D CNNs). Implications made are that the prediction of specific battery faults in electric automated vehicles is possible up to 49 s before their real-time occurrence.