Abstract:

In recent years, the poor shunting of the track circuit has been one of the serious problems in Railway Signaling & Telecommunication Department. Aiming at this problem in 25Hz phase detecting track circuit, we established the prewarning index system, designed the prewarning model by using the BP neural network in the poor shunting of the track circuit, assessed the faults’ risks and realized the prewarning. The prewarning result shows that, by inputting the normalized prewarning index data into the trained network, a credible prewarning result can be obtained in a short time. Compared with the traditional prewarning method that set several voltage lines to realize the alarm function in order to judge if the poor shunting of track circuit has happened, the proposed method has higher efficiency. It offers a certain reference value and guide significance for the maintenance of the track circuit.

Key words: track circuit；poor shunting；prewarning；BP neural network；LevenbergMarquardt algorithm