Fault detection and diagnosis for railway switching points using fuzzy neural network

Switch, as one of the key equipment for railway, plays a vital role in the railway train operation safety and transportation efficiency. One method of ensuring high-level dependability is through the monitoring and recording from Centralized Signalling Monitoring system in China at present. This paper takes the switch action currents of the monitoring data as the research object to analyze and summarize their failure phenomenons and failure causes, and put forward to introduce the fuzzy neural network theory into the switch fault diagnosis. Based on the discussion of the effective feature extraction of the switch action current data, the fuzzy neural network model is established subsequently. The model is a T-S fuzzy one which selects four fault features as input and six kinds of typical fault type as output. This neural network model for switch fault diagnosis is proved effective by computer simulation and verification. The model also has a certain degree of accuracy and provides a good method for switch fault diagnosis.