Common Crossing Fault Prediction with Track Based Inertial Measurements: Statistical vs. Mechanical Approach

Abstract

EN: The analysis of track based inertial measurements for common crossing fault detection and prediction is presented in the paper. The measurement of spatial acceleration in common crossing spike and impact position during overall lifecycle are studied regarding to rolling surface fatigue degradation. Two approaches for retrieving the relation of inertial parameters to common crossing lifetime are proposed. The first one is based on the statistical learning method - t-SNE algorithm that helps to find out similarities in measured dataset. The second one is a mechanical approach that handles the data with a fatigue and contact models. Both approaches allow the significant improvement of the common crossing fault detection as well as its early prediction.