A probabilistic occupant response model for fire emergencies

Abstract

The present paper describes a probabilistic occupant response model for fire emergencies, which is integrated into a fire risk analysis model called CUrisk. Based on the PIA process, i.e., Perception, Interpretation and Action, the occupant response model predicts the probabilities of occupants perceiving fire signals due to direct perception, receiving fire alarms due to the activation of local alarms, sprinklers, the central alarm and the voice alarm, being warned by the other occupants and fire department, and taking actions including pulling the fire alarm, warning other occupants, calling the fire department, and commencing evacuation. The occupant response model is applied to predict the probabilities of evacuation initiation for a number of scenarios that consider combinations of fire detection and alarm systems for a mid-rise building. The results of the model show that asleep occupants need much longer response time to start evacuating and have lower probabilities of starting evacuation than awake occupants, which are consistent with what is observed in reality. Additionally, fire protection systems with only local alarms and only sprinklers connected to the central alarm can be improved significantly with systems with smoke detectors alone or combined with sprinklers connected to a central alarm, which result in higher probabilities of evacuation initiation with shorter delay times.