Development of FWD based hybrid back-analysis technique for railway track condition assessment

Abstract

Track substructure is a key component of the railway transportation system. Similar to the built environment of other surface transportation systems, track substructures are subjected to aging and deterioration. This, frequently leads to the failure and collapse of the transportation systems, resulting in the imposition of costly repairs and maintenance. At the same time, the emergence of high-speed trains and heavier axle loads, together with a need for sustainable designs, has put additional pressure on asset owners. It has been shown that frequent condition assessments of railway substructures can considerably reduce the overall annual maintenance costs. Furthermore, limited knowledge of the substructure condition leads to the employment of inefficient, time-consuming, and expensive maintenance actions. Therefore, development of time- and cost-efficient techniques to frequently monitor existing railway track substructures is vital. The falling weight deflectometer (FWD) is recognised as an effective non-destructive test used to survey ballasted railway substructures through a back-analysis process. This paper presents a novel hybrid back-analysis technique that includes an artificial neural network (ANN) and genetic algorithm (GA) to estimate the substructure layer moduli of railway tracks using FWD testing data. To this aim, firstly a dynamic finite element (FE) model was developed and validated against experimental data from the literature. This FE model then were employed to generate a reliable dataset to train the ANN. In the next step, GA was employed as an optimisation tool within the back-analysis technique/framework to optimise the layer moduli (the ANN's input). A comparison study was performed to evaluate the performance of the developed technique. The results of this comparison revealed excellent performance and robustness of the developed technique.