The steel-on-steel contact between railway wheels and rails results in extremely low rolling resistance, but the high contact stresses lead to large plastic deformations and complex crack networks. Currently, it is not possible to simulate these complex crack networks in rails and wheels. We aim to solve this challenge by using machine learning-enhanced modeling techniques, combined with state-of-the-art computational mechanics methods. This will make it possible to evaluate the influence of these complex crack networks on the rolling contact loading and ultimately the risk of rail and wheel failure.