Original scientific paper

https://doi.org/10.17559/TV-20240420001479

Innovative Approaches in Railway Management: Leveraging Big Data and Artificial Intelligence for Predictive Maintenance of Track Geometry

Richárd Nagy ; Department of Transport Infrastructure, Faculty of Architecture, Civil Engineering and Transport Sciences, Széchenyi István University, Gyor, 9026, Hungary; Vehicle Industry Research Center, Széchenyi István University, Gyor, 9026, Hungary
Ferenc Horvát ; Department of Transport Infrastructure, Faculty of Architecture, Civil Engineering and Transport Sciences, Széchenyi István University, Gyor, 9026, Hungary; Vehicle Industry Research Center, Széchenyi István University, Gyor, 9026, Hungary
Szabolcs Fischer ; Department of Transport Infrastructure, Faculty of Architecture, Civil Engineering and Transport Sciences, Széchenyi István University, Gyor, 9026, Hungary; Vehicle Industry Research Center, Széchenyi István University, Gyor, 9026, Hungary

Abstract

This paper introduces and describes a method for extracting, processing, and analyzing large amounts of track geometrical data. It allows for a more accurate description of the orbital deterioration correlations than currently applied procedures, and it seems to be more valuable and efficient in practice. The initial data were the track geometry measurement and classification data for the whole national network provided by the Hungarian State Railways, i.e., the MÁV PLC. The MÁV provided data for the whole Hungarian railway network for 27 half-years, measured and recorded by the FMK-004 type special diesel locomotive (i.e., track geometry measuring car). The paper discusses the development of a procedure to automatically compute important condition ratings from the available data set of millions of units according to the algorithms created for railway industry colleagues, thus helping the maintenance and renewal decision-making process. Functions have been developed to classify the track geometry condition of a given railway line, to predict how long the service level can be maintained without intervention (i.e., e.g., lining, leveling, and tamping with a mechanized maintenance train), to determine the time of the necessary maintenance intervention, the time of the upgrade (rehabilitation or modernization), and to develop a track geometry prediction procedure that makes full use of the mathematical and computational possibilities of the present day.

Keywords

artificial neural networks (ANN); exponential predictive model; maintenance and renewal decision-making; mathematical and computational modeling; predictive maintenance; railway track geometry; track condition rating

Hrčak ID:

318486

URI

https://hrcak.srce.hr/318486

Publication date:

27.6.2024.

Visits: 0 *