Researchers from the National Institute of Standards and Technology (NIST) have found a way to use sound to detect when lithium-ion batteries are at risk of catching fire.
This new approach could help prevent dangerous fires caused by these common batteries, which power devices like phones, laptops, e-bikes, and electric cars.
Leading the project are scientists Wai Cheong “Andy” Tam and Anthony Putorti.
Lithium-ion batteries are popular because they store a lot of energy in a small space. However, this also makes them risky if they get too hot or become damaged.
In recent years, the danger has grown as more devices rely on these batteries.
For example, in 2023, New York City saw 268 residential fires caused by e-bike batteries, leading to 150 injuries and 18 deaths. Lithium-ion battery fires are dangerous because they ignite quickly and reach temperatures over 1,100°C (about 2,000°F) almost instantly, unlike typical fires that start slowly.
One of the problems with lithium-ion battery fires is that they don’t produce much smoke at first, which means smoke alarms might not detect them in time. Noticing this challenge, Tam saw an opportunity to use sound as an early warning.
When a battery is about to fail, a chemical reaction inside creates pressure, causing the battery to swell. If the battery has a hard casing, a safety valve may open to release the pressure, making a “click-hiss” sound, much like opening a soda bottle.
While other studies suggested that sound could be a warning system, many everyday sounds, like staplers or paper clips dropping, resemble this noise.
To avoid false alarms, the team trained an AI system to identify the exact sound of a battery’s safety valve breaking.
By collaborating with a lab at Xi’an University of Science and Technology, they recorded 38 battery explosions and tweaked these recordings to create over 1,000 samples, which they used to teach the AI to recognize the sound accurately.
The system worked well, with a 94% success rate in identifying overheating batteries. Even with background noises like footsteps and doors closing, the AI rarely made mistakes. Tam presented these results at a fire science symposium, and he and Putorti are working toward patenting the technology.
This new detection system could buy valuable time—up to two minutes—before a battery fails completely.
With further testing, it may be possible to create a fire alarm specifically for battery hazards, designed to detect dangerous situations in homes, offices, warehouses, and EV garages. This technology could offer peace of mind in a battery-powered world, where sometimes, simply “listening” can save lives.
