A new physics-based data-driven guideline for wear modelling and prediction of train wheels

Wear modelling of train wheels has long been an important research topic; both physics-based and data-based methods have some weaknesses. To bridge the gap between them and take their advantages, this paper proposes a new physics-based data-driven guideline for wheel wear modelling and prediction. First, based on wear mechanism analysis, the basic model of tread wear and flange wear are designed considering their correlation; wear models are established separately for different wheel positions considering the uncertainty in a wear process, and further trained mathematically with wear data. Second, wheel reprofiling is closely related to wheel wear and is modelled based on theoretical analysis and reprofiling data. Then, the numerical method for predicting wheel degradation is proposed based on the closed-loop alternation between wear and reprofiling; the remaining useful life (RUL) of wheels is further evaluated through point estimation and interval estimation. Finally, the good agreement between trained models and wear data validates the wear models; the proposed guideline is verified by measurement data to produce accurate prediction of wheel degradation and effective evaluation of wheel RUL. The proposed guideline has been applied to the prognostics and health management of wheels for various train types.