Variable-based classification of ballast empirical settlement models for railway track transition zones

Abstract

Transition zones in railway track structures such as the interface between embankments and bridges are typically locations where significant ballast settlement occurs due to structural discontinuity and repeated dynamic impacts from wheel loads. This study aims to classify and compare empirical models used to predict ballast settlement in such zones, using a framework that groups models based on the level of input variables. These include global variables (number of axle load cycles, dynamic wheel load, train speed) and local variables (number of tamping cycles, ballast stress and void ratio), with the goal of evaluating the applicability of each model for use in real-world railway projects. Five empirical models with different structural forms were selected and applied to a case study. The predicted settlement results were normalized to allow trend comparison, and the influence of variables was evaluated using sensitivity analysis. The results indicate that models incorporating both global and local variables most accurately capture the localized behavior of transition zones. Moreover, dynamic wheel load and ballast void ratio consistently showed high sensitivity indices, highlighting their dominant influence on settlement prediction. In contrast, train speed and the number of tamping cycles exhibited low sensitivity, indicating limited influence. These findings provide a preliminary basis for selecting appropriate empirical models in project contexts where input data availability varies. The results support model-informed decision-making in track design, performance evaluation, and maintenance planning for transition zones with high vulnerability to long-term settlement.