Dynamic risk assessment of emergency evacuation in large public buildings: A case study

Abstract

Large-scale public buildings are characterized by huge internal space, a large number of people, long evacuation paths, along with many internal risk factors and uncertain or unpredictable emergency events. If the condition of risk factors in large buildings could be monitored in real-time, to achieve dynamic assessment and prediction of personnel evacuation risk factors, it can provide early warning, give auxiliary support for emergency decision-making, and improve the risk management and control capability of large public buildings. For the importance of improving risk prediction and evaluation capability in practice, this paper proposes a dynamic risk assessment method that integrates dynamic object detection and risk assessment base on DNN (Deep Neural Network). A real-time detection method based on YOLO (You Only Look Once) algorithm is proposed to extract the risk factors about human factors, which have very important influence in the evacuation processes. Combining building structural features with these dynamic human factors as the input parameters, DNN is finally used as the risk assessment model and to realize the real-time evaluation of evacuation risk level.