The technology will make inspections of train lines more efficient, the Federal Institute of Technology Lausanne (EPFL) said in a press release on Wednesday.
Railway lines are usually checked manually by experts who assess the condition of infrastructure such as walls or concrete sleepers based on pre-defined criteria. However, the EPFL says, this carries the risk of subjective assessments, with inspectors assessing damage differently at different times.
+ Get the most important news from Switzerland in your inbox
To solve this, researchers trained an AI algorithm to differentiate between images with cracks and those without cracks.
The algorithm can now be fed with images taken over several years of a railroad section and quantify the severity of cracks in the walls and sleepers over time. In a study published in the Automation in Construction scientific journal, the researchers showed that this method of monitoring works.
In the next tests, the technology will be tested on the lines between Zermatt and Brig and between Brig and Dissent’s, in southern Switzerland. These sections of track contain a number of retaining walls with different shapes and materials, which makes the task very challenging for the algorithm, according to the EPFL.
Adapted from German by DeepL/dos
