Towards the development of a hybrid methodology of head checks in railway infrastructure

In this paper, the first step towards the development of a hybrid methodology for the monitoring of head checks is discussed. The proposed hybrid method combines a data driven approach with physical modelling of the rail in order to obtain an early stage warning for head checks. Rail defect detection at an early stage of the growth can be challenging and the existence of the seed defects can be confused with non-defect objects on the rail. Thus, a physical model is proposed to investigate how head checks, in particular in curved tracks, initiate and evolve. Track characteristics and loading, e.g. track geometry and track tonnage, are considered to analyze crack initiation by using the Whole Life Rail Model (WLRM) for Rolling Contact Fatigue (RCF) relying on meta-models. The results of the physical modelling and the rail defect observations obtained from the data analysis on the eddy current (EC) measurements are then compared. The physics based model only suggests whether a crack will be initiated or not, it does not give information about the size of the crack. Hence, the next step is to develop an evolution model from the EC and Ultrasonic (US) measurements data, from which the crack size can be determined. This combination of physics based and data-driven evolution model is thus regarded as the hybrid method. This hybrid method can be a robust tool for the prediction of rail condition, as it eases the visualization of the rail degradation and keeps infrastructure managers informed of the actual rail condition that can be confirmed with rail inspections. Finally, real-life measurements from a track in the Dutch railway network are used to show the (potential) benefits of the proposed methodology.