A methodology based on Dijkstra's algorithm and mathematical programming for optimal evacuation in process plants in the event of major tank fires

•Thermal dose relationship has been modified.•Process plant is converted into a thermal dose graph.•Dijkstra's algorithm is used to find the shortest paths to safe spots on the graph.•Mathematical programming is used to allocate the evacuees to safe spots.•The methodology was demonstrated to outperform ad hoc evacuation plans. Effective evacuation plans are essential to protecting workers in the event of major fires in process plants. However, despite some general guidelines, there is no specific methodology available for identification of such plans. In the present study, we have developed a methodology for identifying optimal evacuation strategies in the event of single and multiple tank fires. In this study, considering heat flux as the main cause of casualties in outdoor fires, the conventional relationship for calculating thermal dose was first modified to estimate the values of thermal dose more accurately. The Dijkstra's algorithm was then employed to identify the evacuation paths to safe spots. Due to the limited capacity of safe spots, mathematical programming was used to determine how many people should be allocated to each safe spot and via which evacuation path so as to minimize the risk of casualties. The developed methodology was demonstrated to outperform the ad hoc evacuation plans in minimizing the casualties while allowing the analyst to update the evacuation plans in accordance with the evolution of fire scenarios.