Abstract: Due to the fact that the troubleshooting of 25 Hz phase sensitive track circuits mainly relied on manual labor and could not implement fast and accurate fault location, this paper conducted research on fault diagnosis of 25 Hz phase sensitive track circuits based on logical reasoning and integrated learning. It built a track circuit model based on the working principle of the track circuit, and established a voltage signal monitoring scheme under common fault modes through internal logical reasoning analysis of signal transmission, used AdaBoost ensemble learning algorithm to construct a fault diagnosis model, and injected faults into the track circuit model to construct the corresponding fault model and obtain a fault dataset. Finally, it simulated faults in actual circuit environments and verified the performance of the fault diagnosis model. The verification results indicate that the model can diagnose track circuit faults more accurately than the general model.