Artificial intelligence can help save firefighters’ lives by predicting deadly outbreaks during fires. Outbreaks occur when the combustible material in the room starts to burn all at the same time. This leads to a significant release of heat and gases, which can destroy walls and break windows.
For 10 years, such outbreaks led to 13% of cases of injuries and deaths of firefighters during work. Although firefighters must predict outbreaks based on their own experience, this remains a difficult task.
It was also not possible to simulate such an unstable phenomenon during the last decades. In the current research scientists from the US and China built a system with using graph neural networks to explore relationships between different data sources from simulated fires.
“GNNs are frequently used for estimated time of arrival, or ETA, in traffic where you can be analyzing 10 to 50 different roads. Except for our application, we’re looking at rooms instead of roads and are predicting flashover events instead of ETA in traffic,” says a co-author of the study.
The team simulated 41,000 fires in 17 building types, incorporating all kinds of data into the model: building layouts, surface materials, fire conditions, ventilation configurations, smoke detector locations, and room temperature profiles.
A neural network’s performance was judged by whether it could predict an outbreak 30 seconds before it happened. The model showed an accuracy of 92.1% at best.
“The focus of the research was to rely on building data that is or could easily be provided from available building sensors. One way to translate the research into reality is to integrate the model into a smart fire alarm control panel that would gather the temperature data from installed heat detectors and includes a computer module that can process the data and make the real-time predictions,” says the researcher.
Scientists hope that the forecast can be sent to the commander or individual firefighters in order to reduce the number of accidents.