Classification-learning-based framework for predicting railway track irregularities

Abstract

Railway track irregularities are key factors that affect the safety of trains and the comfort of passengers. In this study, a railway track is separated into consecutive track maintenance units, which are used as research objects. We propose a novel framework, which is referred to as the tree-augmented naïve Bayes-track quality index (TAN-TQI), to identify possible underlying patterns or rules for predicting track irregularities based on the characteristic deterioration of track maintenance units. The prediction framework is validated using track irregularity data measured using track geometry cars. Our evaluation shows that the prediction performance of the TAN-TQI framework is better than that of the state-of-the-art method, the short-range prediction method for track quality index.