Detecting Firefighter's Thermal Risks in a Commercial Building Structure Using Machine Learning

Author(s)

Abstract

A multi-input and multi-output (MIMO) machine learning model is developed to simultaneously detect firefighter's thermal risks across a commercial building structure. A total of 2000 numerical experiments with a wide range of fire and ventilation scenarios are carried out using Fire Dynamics Simulator. Temperature data is obtained from sensors over a simulation duration of 900 s with a 5-s time step. A dataset consisting of 242,000 instances is constructed. The instances are labeled by four thermal operating conditions and are pre-processed for the purpose of training, validating, and testing a machine learning model. Model performance of the MIMO model is provided, and it is benchmarked against typical multi-input and single-output (MISO) machine learning models in terms of accuracy and computation time. Results show that the MIMO model can provide accurate detections at multiple locations simultaneously. This research demonstrates the potential of using machine learning methodologies to develop practical firefighting applications which can, in turn, enhance firefighters' situational awareness and improve their safety measures during firefighting and/or carrying out search-and-rescue in a large commercial building structure.

Download Paper

Fire risk reduction, Fire modeling, Fire fighting, Fire dynamics and science, Fire detection and Fire

Citation

Issues

If you have any questions about this publication or are having problems accessing it, please contact [email protected].