Detection of Track Damage in Railways – Train-Based Prototype Field Tests

Developing an automated system for condition monitoring that supports efficient maintenance decisions for the Swedish railway network and facilitating optimisation of the assets' availability and utilisation.

Goal

The main goal of this project is to develop an automated system for condition monitoring that support efficient maintenance decision for the Swedish railway network and to facilitate optimisation of the assets availability and utilisation.

Project status and results

The project will implement a scalable solution for railway fastener inspection through three main work packages. The first phase focuses on installing the railway fastener measurement system on a locomotive. This will be followed by extended testing of the measuring system. Concurrently, we will address the integration of measurement systems into operational trains, considering various stakeholders' interests and how to manage and combine information from multiple systems.

Figure 1 illustrates the differential eddy current sensors mounted on both sides of a freight train. These sensors successfully captured fastener signatures from 110mm above the railhead during train measurements. Signal processing and feature extraction techniques were employed to analyze the raw signals. Machine learning algorithms for unsupervised anomaly detection effectively identified and isolated anomalous data points from healthy fasteners. The clustering model accurately detected missing clamps and weld joints, distinguishing them with clear boundaries.

This innovative project promises to revolutionize railway maintenance by enhancing the reliability, efficiency, and safety of railway operations through advanced monitoring and data analysis techniques.