In this paper we examine the possibilities of using sensor data from trains in service to develop a real-time monitor of the adhesion conditions of the rail. In everyday railway operations, low adhesion conditions of the track are an important challenge for railway operators, since these may result in a loss of punctuality, an increase in wear of both wheel and rail, and in an increase of the risk for red signal passage in situations where trains are unable to stop in time. At the same time, prudent driving behavior while the adhesion conditions returned to normal, results in unnecessary train delays. A central issue here is that rail adhesion conditions vary across space and time. To date it is a challenge to give accurate real-time information to both train drivers and infrastructure managers. With real-time monitoring of the adhesion conditions, the drivers can adjust their (de)acceleration control to local adhesion conditions and thereby minimize wear, and the infrastructure managers could improve the track conditions by taking friction enhancing measures.
To monitor the adhesion conditions of the track, we used real-time sensor data from ~20 trains. We designed an algorithm that can diagnose track sections as having either slippery or normal adhesion conditions. Specifically, we trained a logistic classifier on a data set that contained reported rail adhesion conditions as well as sensor data from trains in service, such as information about traction, velocity, excessive wheel slip-/sliding detection, and weight. We then assessed the performance of this classifier using an independent test data set.
This first assessment shows a classification accuracy of approximately 77%, with a ~23% false positive rate and a ~23% false negative rate, when compared to the drivers reporting. Several improvements are proposed to increase the sensitivity, which outline the directions of our future research towards the implementation of a real-time monitor of railway track adhesion conditions.
How to Cite
Railways, adhesion conditions, condition monitoring, real-time, big data, logistic regression
The Prognostic and Health Management Society advocates open-access to scientific data and uses a Creative Commons license for publishing and distributing any papers. A Creative Commons license does not relinquish the author’s copyright; rather it allows them to share some of their rights with any member of the public under certain conditions whilst enjoying full legal protection. By submitting an article to the International Conference of the Prognostics and Health Management Society, the authors agree to be bound by the associated terms and conditions including the following:
As the author, you retain the copyright to your Work. By submitting your Work, you are granting anybody the right to copy, distribute and transmit your Work and to adapt your Work with proper attribution under the terms of the Creative Commons Attribution 3.0 United States license. You assign rights to the Prognostics and Health Management Society to publish and disseminate your Work through electronic and print media if it is accepted for publication. A license note citing the Creative Commons Attribution 3.0 United States License as shown below needs to be placed in the footnote on the first page of the article.
First Author et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 3.0 United States License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.