Intelligent fault diagnosis and prediction technologies for condition based maintenance of track circuit

Condition based maintenance (CBM) is considered to be the most effective maintenance strategy in a performance contract. It significantly increases the system availability and efficiency, meanwhile reduces service affecting faults and maintenance cost, especially when combined with online monitoring of real-time condition data of assets. For the above reasons, CBM has been paid more attention in railway signal system. Under the basic principle of CBM, this paper studies the key maintenance technologies of ZPW-2000 track circuit, specifically relating to information collection, fault diagnosis and state prediction. Among them, the centralized signaling monitoring system is responsible for track circuit information collection as well as the preliminary state alarms. For advanced fault diagnosis on monitoring data, a fuzzy neural network (FNN) model is constructed based on the concept of fault tree. After training, the FNN model can realize fault diagnose in real-time. Moreover, a gray GM(1,1) predictor is further studied to estimate the evolution trend of system condition. Simulation experiments show that the proposed diagnosis and prediction methods are efficient, which makes it possible to realize the predictive maintenance of track circuit in real-time.