Prediction Model for Track Quality Index Categories on the Northern and Southern Railway Lines of Java

Abstract

Track quality index (TQI) is a quality metric that objectively measures the geometric condition of railway tracks for maintenance planning. The TQI categories serve as the basis for proposed track maintenance. The TQI measured by the EM120 track recording car on Java island currently covers only 78.84% of the 5,634.363 km of railway tracks, indicating that there are still track sections without TQI category values. This study aims to model the maintenance of railway infrastructure based on TQI categories derived from both track recording car results and manual measurements across various sections of railway lines on the northern and southern routes of Java island. The analysis used is based on the standard deviation of railway track geometry, including superelevation, levelling, lining, and track gauge. Factors such as turnouts, bridges, crossings, straight sections, and curves were then classified as predictive factors. Machine learning techniques were adopted, with 80% of the data set randomly used for training and the remaining for testing to generate TQI category predictions. A total of 233,175 TQI data points from 2019–2022 were used to build and validate the model. The results indicate that the multinomial regression model for TQI Categories 1, 2, 3, and 4 is highly accurate, the rest is influenced by other factors. These results imply that the model has an exceptional fit and excellent predictive capability for TQI on the northern and southern railway lines of Java island.