AI could help save firefighters' lives

Flashovers are some of the most dangerous situations firefighters encounter, as they cause nearly all combustible items in a burning room to ignite at once, with little to no warning. 

With a view to reducing the firefighter deaths caused by this phenomenon, a team of scientists from the US National Institute of Standards and Technology (NIST), the Hong Kong Polytechnic University and other institutions have developed a Flashover Prediction Neural Network (FlashNet) model to forecast these events precious seconds before they erupt.

The model was able to predict flashovers with an accuracy of up to 92.1 per cent for over a dozen common residential floorplans in the US, according to a study published in Engineering Applications of Artificial Intelligence. 

Flashovers tend to suddenly flare up at approximately 600ºC and can then cause temperatures to shoot up further. Until now, most prediction tools rely on constant streams of temperature data from burning buildings or use machine learning to fill in the missing data in the event that heat detectors are damaged by the high temperatures. Because of this, they are only trained to operate in a single, familiar environment.

However, firefighters cannot predict the buildings in which they will be operating, and often fight fires in buildings they don't have floorplans for. 

“Our previous model only had to consider four or five rooms in one layout, but when the layout switches and you have 13 or 14 rooms, it can be a nightmare for the model,” said NIST mechanical engineer Wai Cheong Tam, co-author of the study. “For real-world application, we believe the key is to move to a generalised model that works for many different buildings.” 

In order to make it able to adapt to new environments, the researchers used a machine learning algorithm known as graph neural networks (GNN), often used to analyse road traffic and identify estimated time of arrival, or ETA. 

"It’s very complicated to properly make use of that kind of information simultaneously, so that’s where we got the idea to use GNNs,” said Eugene Yujun Fu, another of the study's authors. “Except for our application, we’re looking at rooms instead of roads and are predicting flashover events instead of ETA in traffic.” 

The researchers digitally simulated more than 41,000 fires in 17 kinds of buildings, representing a majority of the US residential building stock, with various types of furniture. They provided the GNN model with a set of nearly 25,000 fire cases to use as study material and then 16,000 for fine-tuning and final testing. 

Across the 17 kinds of homes, the new model’s accuracy depended on the amount of data it had to chew on and the lead time it sought to provide firefighters. However, the model’s accuracy outperformed five other machine-learning-based tools, including the authors’ previous model. Critically, the tool produced the least false negatives, dangerous cases where the models fail to predict an imminent flashover.

The authors still have a ways to go before they can take FlashNet to the rescue. As a next step, they plan to battle-test the model with real-world, rather than simulated, data.  

“In order to fully test our model’s performance, we actually need to build and burn our own structures and include some real sensors in them,” Tam said. “At the end of the day, that’s a must if we want to deploy this model in real fire scenarios.”