A fault diagnosis method for the tuning area of jointless track circuits based on a neural network

Abstract

This paper proposes a fault diagnosis method for the tuning area of jointless track circuits (JTCs) that is based on using a neural network. Based on the basic structure and working principle of a JTC and track circuit reader (TCR), the induced voltage amplitude envelope (IVAE) in a TCR under different typical fault modes of the tuning area is modelled using transmission line theory. Then, a quadratic function is used to implement piecewise fitting to the IVAE between the tuning area at the sending end of the track circuit and the fourth compensation capacitor counted from the sending end, for fault feature extraction. On the basis of the feature extracted, a back propagation neural network is constructed and trained for fault diagnosis of the tuning units. Experiments with real data show that this method has many advantages such as high detection accuracy, good adaptability and a wide applied range, etc. It can overcome the disadvantages of the current detection methods in aspects such as detection cost and timeliness. Furthermore, it also improves the safety and efficiency of train operation.