Abstract: Maintenance due to wheel and rail damages is a significant part of the total cost of railway operation. To minimize these costs, it is desirable to only maintain when necessary but not too late either to not compromise safety. The KTH Railway Group has long experience of developing simulation tools to predict wear and Rolling Contact Fatigue on wheels and rails. These tools are now in many cases quite mature and validated and could be used as digital twins to predict maintenance needs due to a specific train operation. In the seminar examples of how simulations could be used to optimize rail grinding intervals or to predict the wheel life to forecast wheel turning needs will be presented. Also, some of our activities on using Machine Learning algorithms to detect vehicle component failures or to monitor track geometry will be shown.
