VR FleetCare
The Service Product Developed by VR FleetCare to Automatically Detect the Condition of Trains in the Future
This article first appeared in the Railway-News magazine Issue 2 2021.
The Finnish rolling stock maintenance company VR FleetCare has launched Train Scanner, a new service product designed to automatically inspect the external condition of trains.
In addition to enhancing inspection activities and improving safety, the objective of using Train Scanner is to lower the lifecycle costs of rolling stock. With the help of automated scanning, inspecting the train’s condition will be quicker and the inspection quality will improve.
Up until now, VR FleetCare’s Train Scanner has only been used in passenger trains but due to its scalability, the device can be used to inspect any rolling stock. In border control, for instance, Train Scanner could be used at border stations to automatically detect whether the size of a wagon adheres to the local instructions or used to verify the safety of wagons.
Machine Doing the Work
A large part of rolling stock maintenance is manual inspection performed by human eyes, and the objective is to proceed towards automated and systematic processes.
Train Scanner is based on machine vision, shape recognition and artificial intelligence. The train’s sides and roof are scanned using a line camera with very high line rate speed and resolution. The resolution is approximately a few millimetres. The data from the scanner is instantly processed by means of edge computing that analyses the data and reports any deviations. The information is available to the owner and maintainer of the train within 10–20 minutes. The data is also stored in a cloud service for further analysis.
First Scanner for Customer Use in June
Train Scanner has now been piloted for one year and in June, the first customer will implement the device. In the future, the commuter FLIRT trains in Helsinki will be inspected by driving the trains past the scanner.
“The line rate speed and resolution of the camera could be compared to photo-finish cameras used in sports competitions, for instance. We have taught the device how a train should look so that it can detect any possible deviations. Artificial intelligence becomes increasingly more accurate based on the feedback it receives from humans. The final objective is to be able to reliably detect any rolling stock malfunctions in such a way that human- performed inspections are no longer needed. The time saved can be used to repair the malfunctions instead of searching for them. The algorithm’s accuracy is developed and improved every time a train passes by.”
