Abstract—Allocating limited resources in an optimal manner when rescuing victims from a hazard is a complex and error prone task, because the involved hazards are typically evolving over time; stagnating, building up or diminishing. Typical error sources are: miscalculation of resource availability and the victims’ condition. Thus, there is a need for decision support when it comes to rapidly predicting where the human fatalities are likely to occur to ensure timely rescue. This paper proposes a probabilistic model for tracking the condition of victims when exposed to fire hazards, using a Bayesian Network. The model is extracted from safety literature on human physiological and psychological responses against heat, thermal radiation and smoke. We simulate the state of victims under different fire scenarios and observe the likelihood of fatalities due to fire exposure. We show how our probabilistic approach can serve as the basis for improved decision support, providing real-time hazard and health assessments to the decision makers.
Index Terms—Bayesian networks, diagnostic model, emergency evacuation, human response in fire.
Jaziar Radianti and Ole-Christoffer Granmo are with Center of Emergency Management (CIEM), Dept. of Information and Communication Technology, Faculty of Engineering and Science, University of Agder, Grimstad, Norway (e-mail: firstname.lastname@example.org, email@example.com).
Cite:Jaziar Radianti and Ole-Christoffer Granmo, "A Framework for Assessing the Condition of Crowds Exposed to a Fire Hazard Using a Probabilistic Model," International Journal of Machine Learning and Computing vol.4, no. 1, pp. 14-20, 2014.