Multimodal Data Analysis for Railway Switch Monitoring: Point Machine Motor Currents, Track Geometry, and Maintenance Records

Kurzfassung

Condition monitoring is an important tool in the maintenance of railway infrastructure, and data-driven approaches are gaining in popularity. Switches are complex systems that include both complex track superstructure - which is by design subject to increased dynamic load compared to open track - and point machines that must operate in concert to achieve optimal performance. Existing condition monitoring methods are often limited to a specific type of measurement data to monitor one of these subsystems of a switch and only certain faults. Research and industry are now moving toward integrating all available information into a single, unified system, enabling a clearer understanding of the interactions between subsystems, more efficient maintenance, and better fault root cause diagnosis. This paper investigates the relationships between motor current curves of electric point machines, track geometry data collected by measurement trains, and maintenance logs for 58 switches in the Netherlands over almost 3.5 years. The aim is to provide a foundation for the design of such holistic monitoring systems as well to derive insights and requirements for future data gathering. The analysis reveals that maintenance actions are the dominant factor influencing the measurement data. Major track geometric modifications appear to act as persistent boundary conditions for point machine operation. Given that switches require more frequent maintenance than open track and point machines exhibit dynamic behavior, continuous track geometry monitoring during operation is essential for accurate data analysis and maintenance assessment.