Researchers at the National Institute of Standards and Technology (NIST) have developed an innovative AI-powered system to detect impending lithium-ion battery fires using sound. Led by Wai Cheong “Andy” Tam and Anthony Putorti, the team has trained a machine learning algorithm to recognize the distinctive “click-hiss” sound made by a battery’s safety valve just before it fails.

This groundbreaking technology addresses a critical safety concern as lithium-ion batteries become increasingly prevalent in our daily lives. With the rising number of battery-related fires, particularly in urban areas, early detection is crucial.

Key features of the AI detection system include:

1. 94% accuracy in identifying the sound of an overheating battery
2. Ability to distinguish the safety valve sound from other common noises
3. Potential two-minute warning before catastrophic battery failure

The researchers used audio recordings from 38 exploding batteries, expanded to over 1,000 unique samples, to train the AI algorithm. They are now working on refining the technology for various battery types and exploring different microphone options.

This innovative approach could lead to a new generation of fire alarms specifically designed to detect lithium-ion battery fires in homes, offices, warehouses, and electric vehicle parking areas. By providing crucial early warning, this technology has the potential to save lives and protect property from the rapidly spreading fires associated with lithium-ion battery failures.

For more details on this groundbreaking research, visit the full article on the NIST website.