Modeling panic in ship fire evacuation using dynamic Bayesian network

In this paper, we model passengers' panic during a ship fire by considering its most influential factors. The qualitative factors are quantified, allowing us to study passengers' panic in a probabilistic manner. Considering the time-varying nature of these factors, we update the state of the factors over time. We utilize a dynamic Bayesian network (DBN) to model passengers' panic, this allows us to represent probabilistic and dynamic elements. By defining several worst-case scenarios and running the simulations, we demonstrate how panic can dynamically vary from passenger to passenger with different physical (mental) conditions. Furthermore, we show how this panic can threaten passengers' health during the evacuation process. The impact of panic on the evacuation time is also investigated. The results in this paper are valuable inputs for rescue teams and marine organizations that aim to mitigate property damages and human fatalities.