Abstract: In order to prevent building fires and reduce casualties in fires, reasoning and dynamic evaluation methods of building fire development process and evacuation response process are studied by using fire big data analysis technology and Bayesian network analysis method. Through the analysis of the response state of building fire fighting facilities and the psychological and behavioral characteristics of occupants in fire, the method of fire stage division and the classification of evacuation behavior are discussed and the trajectory intersection theory of casualties in fire is put forward, and framework on fire dynamic risk and evacuation safety assessment based on Bayesian network is constructed. The reasoning process of building fire development and evacuation response are discussed respectively, and the evacuation condition assessment process is discussed Research shows that the model can fuse large fire data such as building characteristics, fire facility status information and occupant response information, evacuation response process reasoning and dynamic risk assessment model in building fire are realized, so as to improve the level of building fire safety management.

Key words: fire, evacuation behavior, risk dynamic assessment, Bayesian network