Occupant characteristics in decision-making for simulation of fire evacuation
MetadataShow full item record
In 2008, fires resulted in 3,320 civilian fire fatalities, 16,705 civilian fire injuries and an estimated $15.5 billion in direct property loss. Most of the time, the unsafe designs of enclosure have been blamed as a main cause of such disasters. Consequently, many safety engineers have begun using computer models to test various designs as part of the decision-making of the building construction plan, to give insight into widths of doors and hallways, placement of exits, placement of other structural support objects (e.g. support columns), etc. Among these computer models, behavioral models present the capability to simulate complex human behaviors, which is another critical factor in fire loss, and help generate performance-based design solutions. Vacate is a behavioral model that was developed at the University of Buffalo, based on the heuristic optimization technique of Particle Swarm Optimization (PSO). The PSO method is used as a path finding algorithm that directs evacuees to the final exit as well as dynamically adjusts evacuees’ direction according to fire hazards and crowd movements. Although Vacate is superior to the other behavioral models on the market with respect to computational costs, a downfall is its rudimentary current occupant decision making procedure. Saunders (2001) showed that males and females exhibited differences in preparedness and response to fire emergencies. On the other hand, Bryan (1981) suggested social role or feeling of responsibility could override gender effect. A review of literature consequently needed to be carried out, to identify the occupants’ and environmental parameters involved in emergency situations decision making, and how they interact with each others. Then, a decision-making model has been built based on these parameters and mechanisms, and integrated into Vacate . The resulting algorithm has been successfully tested and confronted with various floor plans and crowd conditions. Topics for future works, such as assessing accurate values for actions durations or coding the fire fighting behavior, are also discussed.