A new artificial intelligence system developed by researchers at the University of Waterloo, Canada, could help emergency responders, building safety professionals and evacuation planners better understand how fires behave in modern homes.
The framework combines AI, data science and advanced mathematics to analyze large volumes of fire sensor data, with the goal of predicting fire behavior and the gases released during residential fires.
Researchers said the work responds to a growing fire safety challenge: modern furnishings and increasingly airtight, energy-efficient homes can change the way fires develop compared with older, more ventilated buildings.
“The goal is to deepen our understanding so we can predict fire behavior and the gases it releases, enabling smart systems that support safer and more effective fire evacuations,” said Dr. Joshua Pulsipher, a chemical engineering professor at Waterloo.
To build the dataset, the research team conducted 15 experimental fires in a burn house on campus. Up to 175 sensors measured temperature, airflow, humidity, burn rate and a range of gases during the tests.
Even one sensor set in a single location, sampling four times per second, produced millions of data points. Across the full series of experiments, the volume and complexity of the information required a new analytical approach.
The researchers focused on modern furniture and energy-efficient architecture because foams and fabrics can produce different toxic gases when burned. Airtight homes can also create oxygen-starved fire conditions, increasing the amount of smoke and changing its chemical composition.
By applying machine-learning tools to the sensor data, the system can identify patterns linked to fire growth, ventilation conditions and smoke toxicity. The findings could support future improvements to building codes, fire response planning and evacuation route design.
“We want to create smart systems that model and anticipate what a fire will do and then route people to get out of the building safely,” said Dr. Beth Weckman, professor of mechanical and mechatronics engineering.
The system can help determine when a fire begins to under-ventilate, meaning it is running out of oxygen. That transition marks a critical change in combustion chemistry and is associated with the production of more toxic smoke.
“The input is the data from the experiments,” said Dr. Vinny Gupta, a mechanical and mechatronics engineering professor and member of Waterloo’s Fire Research Group. “The output is understanding what those experiments tell us to reveal fundamental insights about how the underlying fire evolves.”
