According to the researchers, this safety valve breaks and emits a distinctive hissing sound, much like the sound of a soda bottle cracking.

ML algorithm can recognize the sound of a safety valve breaking

Researchers from the National Institute of Standards and Technology (NIST) qualified a machine learning algorithm capable of recognizing the sound of a safety valve breaking.

The researchers first needed many sample sounds to make the algorithm work.

In collaboration with a laboratory at Xi’an University of Science and Technology, they recorded audio of 38 exploding batteries.

Then the researchers changed the speed and pitch of these recordings to expand them into more than 1,000 unique audio samples that they could use to teach the software what a rupturing safety valve does. as.

Algorithm detects the sound of an overheating battery 94% of the time

The researchers claimed that the algorithm works remarkably well because they detected the sound of an overheating battery 94% of the time using a camera-mounted microphone.

“I tried to confuse the algorithm by using all kinds of different noises, from recordings of people walking, to doors closing, to Coca-Cola cans opening,” explain Wai Cheong “Andy” Tam of NIST. “Only a few of them confused the detector.”

The researchers used 1,128 samples of acoustic data, including various human activities, to facilitate the development of a detection model that could be used in real-world settings.

“Using 10-second acoustic data as input and a convolutional neural network model structure as the backbone, the detection model has an overall accuracy of 93.9% with a precision and recall score of 91.6% and 97.7%, respectively,” explained the researchers of their study.

Parametric studies are carried out to evaluate the robustness of the proposed model structure and the effectiveness of the data augmentation methods. Additionally, according to the researchers, the model’s performance against two comprehensive tests is evaluated using test-free cross-validation.

It is also claimed that the proposed work could help develop a robust detection device capable of providing early warning of thermal runaway and giving users more time to mitigate potential extreme fire risks and/or evacuate in completely safe.