Research database

Continuous monitoring of freight trains can significantly increase the safety of the transportation of goods on railways and improve the management of the fleets of freight wagons, with big reduction of the maintenance costs. Nonetheless, a widespread solution for real-time monitoring is still not available,mainly because of the lack of electrical power on board freight wagons and the absence of wired communication between wagons. These issues could be tackled with the deployment of wireless sensor nodes powered through energy harvesting, which however cannot provide a sufficient and stable amount of power, thus limiting the efficiency and accuracy of the monitoring algorithms. The project overcomes the previous issues with the design and testing of a new low-power monitoring unit, whose computational efficiency is guaranteed by fast-to-evaluate surrogate models, acting as the digital twins of the wagon.The digital twins are built with machine learning (ML) and artificial intelligence (AI) techniques from the outputs of detailed numerical models of the wagon. The digital twins feature an accuracy level comparable with the one ensured by detailed numerical models, but they lead to a drop of the computational times by several orders of magnitude. This makes them suitable for the implementation on low-power and low-performance hardware architectures, thus enabling the installation on board freight wagons for real-time monitoring operations. The monitoring unit is intended to monitor the operation of the wagon considering the health status of the main wagon components and subsystems as well as the running behaviour of the vehicle, accounting for the dynamics of the whole train. The deployment of the monitoring system on a large scale can increase the attractiveness of railways over road vehicles, thus contributing to the reduction of emissions related to transports in the EU.