The AI system detects a characteristic “click-hiss” sound produced when a battery releases gas as it approaches thermal runaway
Researchers at the National Institute of Standards and Technology (NIST), in collaboration with Xi’an University of Science and Technology, have developed an artificial intelligence (AI) system capable of detecting the distinct sound emitted by lithium-ion batteries moments before they fail. This breakthrough technology could significantly enhance safety measures, offering early warning against potential battery fires.
The AI system detects a characteristic “click-hiss” sound produced when a battery releases gas as it approaches thermal runaway. Thermal runaway is a hazardous chemical reaction that occurs when a lithium-ion battery overheats or sustains damage, potentially leading to fires or explosions.
During trials, the AI system demonstrated a 94 per cent accuracy rate in identifying the sound of battery failure. The researchers trained the algorithm using audio recordings from 38 battery explosions, incorporating over 1,000 unique audio variations to refine its ability to differentiate battery failure sounds from background noise.
Wai Cheong “Andy” Tam, a NIST researcher, explained how the idea emerged: “Right before the fire started, the safety valve in the battery broke, and it made this little noise. I thought we might be able to use that.” Testing this hypothesis with co-researcher Anthony Putorti confirmed that the sound was distinct enough to enable early detection.
Lithium-ion battery fires pose unique dangers due to their high heat and rapid ignition. Traditional smoke alarms often fail to detect these incidents early, as these batteries produce minimal smoke during the initial failure stages. This limitation has led to serious consequences. In New York City alone, 268 residential fires involving e-bike batteries in 2023 resulted in 150 injuries and 18 fatalities.
The AI-based detection system offers a solution by identifying the sound of battery failure up to two minutes before ignition, providing critical time for evacuation and emergency response. The researchers envision the technology being used in settings with significant lithium-ion battery usage, such as homes, offices, and warehouses.
The findings were presented at the 13th Asia-Oceania Symposium on Fire Science and Technology, where Tam highlighted the potential of sound-based detection to revolutionise battery safety. The team plans to expand testing to various battery types and experiment with different microphones to improve the system’s versatility. Efforts are also underway to secure a patent, paving the way for commercialisation.
By refining the AI system’s accuracy and adaptability, the researchers aim to provide an additional layer of safety in environments where lithium-ion batteries are integral to daily operations.