Can the condition of a train's mechanical and electrical systems be sensed through its operating control system? This projects aims at providing an answer to this question.

Wear and maintenance in trains

With trains operating daily on diverse routes, wear and tear on wheels and other components is inevitable. After a time, train wheels need to be returned or replaced. Addressing and tracking these wear issues before the train arrives in the workshop would make maintenance more effective – this is called predictive maintenance, where condition is assessed by information from collected data. Conventional methods typically rely on fixed maintenance schedules.

A data-driven approach for wheel wear prediction

Custom monitoring systems have been installed on trains in traffic to collect real-time data during regular traffic operation. Mechanical operation parameter are collected from the trains driving system. This project also goes a step further by turning also utilizing the train’s power converter as a sensor. These electrical signals containing information about the trains continuous drive, from power gathered by the pantograph – to wheels turning on the tracks. This collected data can be used to spot anomalies in everyday operation and later be correlated to the accumulated wear on the wheels. 

Advanced analytics and machine learning techniques are used to examine key influence factors like weather conditions, localized features, and track quality. Over time, empirical models can be built in order to better predict the real-time condition of the wheel wear, and plan maintenance schedules to plan interventions such as turning or replacement. 

The project is conducted in close collaboration with Norske tog (funding body) as well as the industry partners Stadler and ABB.