Biographie
Abstract
Rolling contact fatigue is the most frequent degradation mode observed on the rail. Many mechanical studies have revealed the main physical sizes that generate the crack initiation, which are all linked with the track geometry, the dynamic of the wheel and contact conditions. Except for extreme climate, exogenous phenomena such as ambient temperature have not been analysed. The main reason is that temperature variations induce stress cycles in the rail that are under the endurance limit of the material. Therefore, most fatigue life estimation methods do not consider these cycles.
In this paper, 6 years of daily temperature measurement over all the French railway network are used and integrated in a machine learning model developed for computing residual lifespan of each rail segment before crack initiation by rolling contact fatigue. Temperatures are aggregated using the continuum damage mechanics model developed by Lemaitre and Chaboche that gives the possibility to let low amplitude cycles to contribute to fatigue damage. Classification models are then implemented, and importance variables are finally quantified. Results show that daily temperature variations have more impact than seasonal variations and that its contribution to fatigue is significant.
