A vision-based method for line-side switch rail condition monitoring and inspection

Abstract

In railway systems, switch rails are one of the key components of switches & crossings (S&C). They are controlled by switch machines to guide trains from one track to another. Due to the discontinuity in geometry, switch rails are exposed to high-impact loads as train wheels pass through. The long-term impact loads can cause local plastic deformation. These faults, and general alignment changes, can lead to the development of a gap between the switch rail and the stock rail known as a toe gap, as well as non-optimal contact with the wheel flange, both of which can endanger the safe operation of passing trains. Currently, periodic visual inspection is the main method for detecting these defects. This is not efficient or reliable enough to support the ever-shortening maintenance windows available in modern railway systems. The development of computer vision technologies and constantly improving processors make it possible to monitor the health status of such safety-critical components in real time. This research proposes a line-side condition monitoring approach for the switch rail. With the use of dedicated identification algorithms, the status of the switch rail, including movement, position, toe gap and the edge of the toes, can be monitored remotely in real time. This approach has been tested in a high-speed train testing centre in China. The results show a capability to further improve the safe operation of S&C while simultaneously reducing the cost and increasing the safety of inspection.