Abstract: In view of the metro turn back switch locking jamming problems that are prone to happen but difficult to predict, the S700K switch machine is taken as the research objective, a new selfadaptive locking jam critical state discrimination mechanism is established based on switch machine power from a statistical point of view. Then, the switch machine power change behavior before the locking jam fault is statistically analyzed, an adaptive segmentation model for critical power threshold of jamming based on cyclic Otsu is proposed, the application scenario of jamming critical state discrimination is designed according to the critical threshold. Empirical tests of the data obtained from multiple groups of switch locking faults on different Wuhan metro lines are conducted, which verified the validity of the proposed model and scheme.