Tvema, a Russian provider of railway infrastructure solutions, has developed a technology to predict the wear and tear on railway tracks. The data collected by diagnostic vehicles is automatically processed with the infrastructure diagnostic system of Russian state railway RZD and then further analysed by artificial intelligence (AI). The result of this analysis is the simulation of lateral rail wear, deviations in track geometry parameters and other defects. Experimental work realised on a section of the Moscow Railway showed a prediction accuracy of defects for one month in advance at a level of 92%. Based on this data, a complex for automatic planning of tasks for track repairs is being created. Since 2022, Tvema has been using machine learning for automatic acoustic and visual defect detection of tracks, currently determining 17 types of defects through linear video.
