The maintenance and renewal of ballasted track can be optimized in terms of time and cost if a proper statistical model of track deterioration is derived from previous maintenance history and measurement data. In this regard, quite a few models with simplified assumptions on the parameters have been suggested for the deterioration of ballasted track. Meanwhile, data driven models such as the artificial neural network (ANN) and support vector regression (SVR), which are basic ingredients of machine learning (ML) technology, were introduced in this study to better represent the deterioration phenomena of track segments so that the results can be directly plugged into the optimization schemes. For this purpose, the influential parameters of track deterioration have been selected based on the maintenance history, and two ML models have been studied to find the best combination of input parameters. Through numerical experiments, it was found that at least 2 years of maintenance data were needed in our case to obtain a stable prediction of track deterioration.

Although reliability centered maintenance (RCM) has been used widely in the maintenance and renewal (M&R) of railway infrastructures, M&R methods and maintenance frequencies still require considerable study to obtain information on optimum decisions. Recently, various attempts have been made to optimize the maintenance works in railway engineering, although a decision support system on selecting the maintenance or renewal works has not been suggested distinctly. Therefore, a genetic algorithm on the basis of the nondominated sorting method (NSGA) was used in this study to optimize the biobjective functions on M&R interventions and costs. The results showed that the optimum frequency of M&R could be determined using the NSGA method with various constraints. Moreover, field application of the suggested method also was considered. Future work will be concentrated on the systematic approach of the decision support model with RCM and the enterprise resource planning (ERP) system of the infrastructure manager (IM).