Estimation of track modulus over long distances using artificial neural networks

Abstract Evaluating the railway track structure and identifying the problematic locations with the urgent need of repair along the thousands of miles of tracks have always been a challenge to the railroad industry. Track foundation modulus (also referred to as the track modulus) is one of the main parameters that affect the track performance, and thus, quantifying its magnitude and variation along the track has a potential to be a significant addition to the current methods for evaluating the track structure. Track modulus can be quantified by measuring the deflection of the rail when subjected to a known applied load. Hence, train-mounted vertical track deflection (VTD) measurement systems that have been developed over the past decades, present a great opportunity to estimate track modulus and its variation over long distances. This paper presents a new methodology for quantifying the track modulus average over track windows using VTD measurements. In this study, finite element modeling was used to simulate the track structure with stochastically varying track modulus. Various track modulus distributions were considered and the deflection of the rail under moving load was calculated at the predefined intervals along the track. The mathematical correlation between the rail deflection and track modulus over track windows was then studied using artificial neural networks. Numerical results suggest the average of track modulus over track windows can be successfully estimated using VTD measurements.