Abstract: To overcome the limitations of conventional 2D static topological models characterized by delayed hydrodynamic response and insufficient spatial attribute coupling, this paper simplifies underground mine tunnels to construct a graph structure model. Utilizing graph algorithms, it achieves rapid automatic generation of escape routes to improve miners’ survival probability during water inrush events. The research processes tunnel centerline point cloud data with 3D coordinate reconstruction and spatial topological relationships to establish a tunnel spatial graph G=（V, E） representing the three-dimensional mine structure. Practical computational requirements are addressed by incorporating engineering attributes such as volume and cross-sectional dimensions into the 3D tunnel model. Considering hydrodynamic impacts on escape processes, time-dependent edge weights are introduced to form an enhanced graph structure G=（V, E, W）, enabling dynamic weight calculation to quantify escape velocities under spatiotemporal variations. To evaluate personnel evacuation capacity, it enhances time-dependent processing in Dijkstra algorithm and propose a time-varying weighted network path planning method, overcoming traditional Dijkstra’s limitation in handling dynamic weights. A dynamic escape route planning system with 3D visualization is developed using VTK rendering engine, constructing a multilayer architecture system with visualization components for rapid optimal path generation through network weight assignment. Practical validation using engineering data from Maoping Lead-Zinc Mine in Yiliang, Yunnan Province demonstrates the validity of flooding simulation and escape route effectiveness in tunnel networks. This research provides a scientific safety assurance tool for mining operations through graph theory-based rapid escape path generation. The outcomes significantly enhance production safety in China’s deep mining industry.