A team of researchers at the University of Arizona has come up with something promising and claims to revolutionize the electric vehicle (EV) safety aspect by predicting as well as preventing dangerous temperature spikes in lithium-ion batteries. They have developed an innovative machine learning (ML) model under the guidance of doctoral student Basab Goswami.

The study is titled “Advancing Battery Safety” and was published in the Journal of Power Sources. It introduces a framework combining multiphysics and ML models to detect as well as predict battery overheating. It was funded by a $599,808 grant from the Department of Defense’s Defense Established Program to Stimulate Competitive Research. Goswami and his adviser, aerospace and mechanical engineering professor Vitaliy Yurkiv, believe the model could be integrated into EV battery management systems to prevent overheating.

Thermal runaway is a dangerous phenomenon where battery temperature rises exponentially and thereafter triggers a chain reaction within the battery pack. EV battery packs typically consist of thousands of cells. One cell overheating may lead to the heating up of the surrounding cells and this may cause explosion. Goswami’s team proposes using thermal sensors around battery cells that feed data into an ML algorithm. The algorithm can therefore predict when and where a runaway event might occur.

Yurkiv said that they didn’t expect machine learning to be so precise in predicting thermocouple temperatures and hotspot locations.

The research builds on earlier studies using thermal imaging and it is a lighter as well as more cost-effective to previous methods. The study comes at such a time when the Biden administration lately announced a $1.7 billion investment in EV manufacturing. It reflects the growing global demand for electric vehicles equipped with sales up by 35% in 2023 compared to the previous year.