Detecting Firefighter's Tenability Utilizing Machine Learning

Published

August 30, 2024

Author(s)

Qi Tong, Hongqiang FANG, Eugene Yujun Fu, Wai Cheong Tam, Thomas Gernay

Abstract

The proposed research aims to leverage machine learning to detect thermal operating classes and improve the tenability of firefighters in a commercial building. A total of 3000 simulations are run with FDS to collect temperature from heat detectors. The total simulation time is 900 s with a time step of 5 s. A total of 471,000 instances are obtained from FDS simulations. The temperatures obtained from FDS are later pre-processed for training, validation, and testing of a machine learning model with a ratio of 69:14:17. The machine learning model is fine-tuned, and results will be presented in the full paper. A multiple-input and single-output machine learning framework with time-series temperature information is developed to detect firefighter's thermal tenability. The results will show the machine learning model is able to detect three thermal operating classes (i.e., routine, ordinary, or emergency) of the firefighters at any specific location using the sensor temperatures at the ceiling as inputs. This study will demonstrate the potential of machine learning models in practical applications to improve firefighters' situational awareness and enhance the safety of firefighters.

Proceedings Title

International Conference on Automatic Fire Detection’ (AUBE ’24)

Conference Dates

September 24-26, 2024

Conference Location

Duisburg , DE

Pub Type

Conferences

Download Paper

Local Download

Keywords

thermal tenability, commercial building, smart firefighting, situational awareness

Citation

Tong, Q. , FANG, H. , Fu, E. , Tam, W. and Gernay, T. (2024), Detecting Firefighter's Tenability Utilizing Machine Learning, International Conference on Automatic Fire Detection’ (AUBE ’24), Duisburg , DE, [online], https://tsapps.nist.gov/publication/get_pdf.cfm?pub_id=957544 (Accessed December 15, 2024)