Noise as a warning: AI can warn of exploding batteries in advance
When lithium-ion batteries explode, there is a hot flash almost without warning. Soon, a warning device with AI could urge people to flee in good time.
(Image: KarolGajewski/Shutterstock.com)
According to a team at the US research institute NIST, they have trained an AI technology that can provide a highly reliable warning before a lithium-ion battery explodes. Although work on such detectors already exists, the use of AI now ensures that early detection also works in unfavorable environments and that the risk of false alarms is not too great. Overall, the new algorithm recognized the noise as such 94 percent of the time, about two minutes before the explosion. If warning devices are developed based on this, this could be enough to enable people to escape, as is already the case with smoke detectors.
(Image:Â Jian Chen/Xi'an University of Science and Technology)
Very hot, very sudden, without smoke
Lithium-ion batteries are widely used because they can store a lot of energy in a small space. They are not only found in smartphones and laptops, but also in e-bikes and electric cars, for example. In the latter, they are significantly larger, making explosions much more dangerous. As the NIST explains, burning lithium-ion batteries are not only particularly dangerous because they burn at temperatures of up to 1100 degrees Celsius, but also because extremely little time elapses before they do. Within less than a second, the flame can become almost as hot as a blowtorch. This lack of warning time distinguishes them from most other fires. In addition, such fires produce little smoke that warning devices can detect.
Videos by heise
As the team now explains, their technology makes use of the fact that the batteries inflate shortly before the explosion. Many housings therefore have a safety valve through which the pressure is released. When this happens, a characteristic noise is produced. However, this can be confused with many others, especially in noisy environments. The group has therefore trained an algorithm to identify precisely this sound, using recordings of 38 exploding batteries. More than 1000 unique sound examples were created on the basis of these recordings. Connected to a microphone, the trained algorithm worked "remarkably well" and could not be confused by similar noises.
A new type of alarm soon?
The next step is to research the technology even more thoroughly using other types of batteries and other microphones. The team has also applied for a patent for its own technology. If it proves to be as powerful as it appears to be so far, it could form the basis for a new type of alarm, the team writes. These could be installed in homes, offices and buildings with a particularly large number of batteries –, such as warehouses or parking garages for electric cars. When they go off, this could give people enough time to get to safety. But there is still a long way to go before then.
Burning batteries, especially from e-bikes, repeatedly cause firefighting operations. The fire department therefore warns that such devices should never be charged in the home and certainly not at night or without supervision. Instead, it is recommended that such devices be charged on the balcony. Under no circumstances should the device be plugged into the socket in escape routes, for example in the stairwell. You should also not attempt to extinguish a potential fire yourself.
(mho)