Abstract: During the process that passengers gathering and distributing in stations of urban rail transit,bottlenecks of infrastructures can lead to congestion,which seriously affects efficiency and safety.In order to identify and classify the bottlenecks,a network for evacuation is developed based on partitioning space into cellular.Three methods are proposed to identify static bottlenecks under emergency evacuation,supported by complex network theory,network maximum flow theory,and blocking flow theory,respectively.With the software Anylogic,static bottlenecks are ranked by estimated e-vacuation time.For the identification and ranking of dynamic bottlenecks,an algorithm based on temporal and spatial con-gestion intensity is proposed.Finally,a case study of Xizhimen subway station in Beijing is carried out,which is used to test the proposed conceptual and simulated evacuation models.study results show that the bottleneck identified by bloc-king flow theory is most consistent with the observed operation among the three kind of static bottlenecks.The dynamic bottlenecks with the highest ranking importance are the inlets of stairs at the platform level.The bottlenecks generated during gathering and distributing processes mostly occurred at streamlined areas,and tend to propagate further.The reli-ability and applicability of the proposed method are also verified through the above case study.