A screenshot from the NIST video "Predicting Deadly Backdrafts".
21 Oct 2022

"Predicting Deadly Backdrafts": NIST works on machine learning application to help firefighters avoid dangerous situations


Researchers at NIST are working on a model based on machine learning to help firefighters predict and avoid dangerous fire behavior related events in enclosure fires, including backdrafts.

FireEngineering.com reported on this in an article on October 17, where they claim researchers at the US National Institute of Standards and Technology (NIST) have a new plan for how to informing firefighters of what dangers lie behind the closed doors or a potentially under ventilated fire. 

According to the article, "currently firefighters look for visual indicators of a potential backdraft, including soot-stained windows, smoke puffing through small openings and the absence of flames. If the cues are present, they may vent the room by creating holes in its ceiling to reduce their risk".

However, the "Art of Reading the Fire" is not always and exact science. It is also not a guarantee it is always taught at fire schools. Some smaller brigades may perhaps also lack the resources to properly educate all fire fighters of the dangers which can occur if an under ventilated fire suddenly receives a rush of oxygen when a door or a window is opened. 

In order to create a more science based model, NIST wanted to record data from simulated situations which could later be used to anticipate in more reliable way when dangerous conditions for firefighters to entry can occur. 

NIST researcher Ryan Falkenstein-Smith and his colleagues have conducted experiments where they have created a gaseous, energy rich environment in a laboratory, simulating an enclosure fire. They starved the fire by keeping all doors and openings closed for several minutes to observe in what situations backdrafts or other fire gas related events would occur. 

They conducted nearly 500 experiments, where many parameters were recorded, including types of fuel, types of  combustible gases, temperatures, pressures and more. 

"The team obtained data from hundreds of backdrafts in the lab to use as a basis for a model that can predict backdrafts. The results of a new study, described at the 2022 Suppression, Detection and Signaling Research and Applications Conference.... In the future, the team seeks to implement the technology into small-scale devices that firefighters could deploy in the field to avoid or adapt to dangerous conditions", FireEngineering.com writes in their article. 

The idea is to use the results to create a "portable device" which can help firefighters learn and predict when dangerous conditions can occur. The NIST video below describes the project somewhat more in details, but doesn´t get in to precisely what their intended, final product will be.