Abstract
Extensive datasets containing the train-track interaction are essential for further progress and algorithm development required for an effective maintenance of railway turnouts. These crucial parts of railway superstructure are currently inspected primarily manually hence automated solution as in the case of nominal track is desired. This work presents the RailCheck dataset that was built with use of affordable wireless battery-powered sensors consisting of micro-electromechanical systems (MEMS) accelerometers which on-demand automatically records the train-track interaction on a double crossover turnout. The work discusses the challenges of using a consumer-grade inertial accelerometers instead of industrial ones, presents the difficulties of integrating the system with the CEN Transmodel (EN 12896) and introduces the collected dataset that is now available to research community. Knowledge provided can help to build better monitoring systems and reliable algorithms necessary for future transition from preventive to predictive maintenance.