RailCheck: Functional Safety for Wireless Condition Monitoring of Railway Turnouts and Level Crossings Jan Sramota, Mary Ann Lundteigen, Stig Petersen, Amund Skavhaug Department of Mechanical and Industrial Engineering Norwegian University of Science and Technology Trondheim, Norway Email: {jan.sramota, mary.a.lundteigen, stig.petersen, amund.skavhaug}@ntnu.no Abstract— Increasing demands for more cost-effective, reli- able and safer railway infrastructure unavoidably bring up the need for transitioning from current preventive mainten- ance strategies to more efficient, predictive, condition-based maintenance models. Such a change requires large installations of sensors that continuously monitor key infrastructure and aggregate captured data for post-processing in the cloud. Due to the amount of assets to be supervised, novel approaches must be studied in order to find a viable solution that is deployable on this scale. Continuous surveillance is then a desired goal, as it is closely associated with big-data analytics and allows to predict upcoming issues and react to unexpected events. Infrastructure managers will gain much better overview as a result of large amount of highly representative data set-in-context; moreover, they will benefit from having supportive algorithms simplifying their determinations. This paper describes the safety-related measures performed on one such system; eventually intended to replace the routine inspections currently being carried out on railway points and level crossings. I. I NTRODUCTION Increasing demands for better transportation systems in the 21st century resulted in a trend called smart transportation and facilitated the emergence of intelligent transportation systems. These systems aim to minimise traffic problems, enrich stakeholders with prior knowledge, reduce travel time and cost as well as enhance passengers’ comfort and their safety. Indeed, between today and 2050, major changes are expected due to previous increased activity in this area [1]. Speaking at the operational level, the European Rail Traffic Management System (ERTMS) [2], which provides a com- mon framework for all railway traffic in Europe, was adopted and is now being implemented. The ERTMS comprises the European Train Control System (ETCS), railway adaptation of the Global System for Mobile Communications (GSM-R) [3][4] and the European Traffic Management Layer. Interest- ingly, work that started on ETCS in the early 90s revealed number of challenges that either affected or resulted in several safety-related and technological standards, including e.g. IEC 61508 [5], EN 50159, GSM-R and the forthcoming LTE-R [6]. It remains an open question whether or not LTE-R will be launched as a successor to GSM-R (as happened in South Korea), or if the next generation ‘5G-R’ will be used instead. However, it is expected that ERTMS at level 3, having the potential to increase the capacity up to 40% on the current infrastructure [7], will revolutionise this sector. At the infrastructure level, the transition from preventive maintenance to a more targeted approach so-called predictive condition-based maintenance would dramatically alter this segment. Predictive maintenance strategies use sensors that continuously monitor crucial parameters and in conjunction with analysed historical trends evaluate the life-cycle stage of the monitored parts. This allows precisely predict impending failures and use the railway infrastructure with a higher efficiency, resulting in lower costs and enhanced safety. This paper describes one such system called RailCheck, developed and built at the Norwegian University of Science and Tech- nology (NTNU). This system monitors railway infrastructure by utilising remote sensors and big-data analytics to interpret approaching and imminent threats hard to detect otherwise. II. OVERALL RISKS &HAZARDS OVERVIEW Train derailment and collisions are the most severe situ- ations that may arise due to neglected maintenance and poor workmanship. They occur as a result of a number of distinct causes that can generally be classified as mechanical failure of track components (e.g. broken rails, cracked rails, broken gauge spreads), geometric failure of track components (e.g. rail climbing due to excessive wear, earthworks slip) and dy- namic failure of train/track interaction (e.g. extreme hunting, vertical bounce, track shift under the train). To prevent such events, railway tracks are regularly in- spected by equipped measurement trains that use a com- bination of cameras and laser-based systems. These tools automatically evaluate the condition of the track and help identify mentioned faults before they can negatively affect performance or become a safety issue. Located problematic spots can then be manually inspected by the infrastructure managers (IMs) either from the camera footage or personally by the inspection in the field. This is a very convenient way how to effectively monitor and maintain this large and dense network. Unfortunately, most of these tools are limited just to the tracks, excluding points and level crossings (P&C), which have to be then still inspected solely manually. This in combination with a large number of these units, estimated to be one P&C per km of track (EU27) [8], makes the associated tolerable hazard rate (THR) difficult to maintain and demands for lower maintenance costs, believed to be an equivalent of about 0.3 km of the plain tracks [8], unable to achieve.