Building fire simulation and BIM visual emergency response model combined with deep learning

Building fire poses a serious threat to public safety, and the existing fire simulation technology has limitations, so it is difficult to truly reflect the fire dynamics in complex scenes. In this study, a building fire simulation and visual emergency response model combining deep learning and BIM is proposed to improve fire prediction accuracy and emergency response efficiency. By constructing a multi-source data fusion framework and combining the hybrid simulation engine of physical information neural network (PINN) and traditional fire dynamics solver (FDS), the high-precision dynamic simulation of complex coupled scenes such as earthquake secondary fire and earthquake damage of sprinkler system is realized. At the same time, the space-time graph convolution network (ST-GCN) is used for dynamic risk assessment, and the hierarchical reinforcement learning algorithm is used to optimize the evacuation path, which significantly improves the evacuation efficiency. In addition, the BIM-based visual emergency response platform integrates 3D situational awareness, dynamic scheduling of emergency resources and other functions, and realizes real-time visualization of fire scenes and optimal scheduling of resources through WebGL engine and AR technology. The experimental verification with the inpatient building of a 3A hospital as a case shows that the model is superior to the traditional method in smoke diffusion prediction accuracy, evacuation efficiency and emergency resource scheduling optimization, and shows stronger robustness in complex scenes, which provides a new technical solution for building fire emergency response.