A new artificial intelligence model from researchers at the U.S. National Institute of Standards and Technology and Hong Kong Polytechnic University could warn firefighters of imminent combustion or flashovers.
The Flashover Prediction Neural Network (FlashNet) model can predict flashovers as early as 30 seconds before eruption, with up to 92.1% accuracy.
The researchers used graph neural networks to strengthen FlashNet and trained it on nearly 25,000 fire cases, then used another 16,000 for refinement and final testing.
In digital simulations of over 41,000 fires in 17 types of buildings representing common U.S. residential floorplans, FlashNet outperformed five other machine learning-based models.
From NIST News
View Full Article
Abstracts Copyright © 2022 SmithBucklin, Washington, DC, USA
No entries found