Fire prevention strategies in electric vehicle batteries using machine learning

The increasing use of electric vehicles necessitates robust safety measures, particularly in battery management systems. This study emphasizes predictive maintenance by introducing a proactive approach to fire safety management using machine learning. The behavior of 60% Nickel, 20% Manganese, and 20% Cobalt (NMC 622) prismatic cells under mechanical impact was investigated, with CO and CO2 gas emissions monitored as early indicators of thermal runaway, a phenomenon that can lead to rapid and uncontrollable temperature increases if undetected. A real-scale experimental set-up simulated mechanical impacts, and the collected data were analyzed using MATLto derive meaningful insights. Four machine learning models - coarse tree, binary GLM-LR, efficient linear support vector machines, and Gaussian Naive Bayes-were trained and validated to predict the likelihood of thermal runaway based on gas emission patterns. This proactive approach enhances battery reliability and safety by enabling early intervention in critical areas, ensuring passenger safety. By addressing a significant gap in current research, this study contributes to the development of smarter and safer electric vehicles.