关键词: 人员行为, 决策, 自适应模糊神经网络, 应急疏散, 火灾, 问卷调查

Abstract: Questionnaire investigation was progressed on three fire cases in public domestic building in Hong Kong. Based on the adaptive neuro-fuzzy interference system and BP function, data training and checking were progressed to enforce the network learning upon the 150 sets of collected data. As a result, a personal emergency response model on evacuation decision behavior, PERM, was established. The prediction result of PERM and the questionnaire result were almost coincident. It showed that personal decision behavior before emergency evacuation is relevant to what she/he experienced in fire, initial status, residential conditions, etc. For instance, among the residents in public domestic housing in Hong Kong, half would not start evacuation for 10-20 minutes until they confirm the fire alarm/information. Only a quarter of the residents decided to evacuate immediately.