Title: Novel heuristics for identifying failures on railroad switches: a case study on Vale's railway

Authors: João Pedro Augusto Costa; Omar Andres Carmona Cortes; Adriano Rodrigues; André Machado Souza; Tiago Ferreira Souza; Valerio Nunes

Addresses: Applied Machine Learning and Optimization Group, Instituto Federal do Maranhão (IFMA), São Luís, MA, Brazil ' Departamento de Computação (DComp), Instituto Federal do Maranhão (IFMA), São Luís, MA, Brazil ' Vale SA, São Luís, MA, Brazil ' Vale SA, São Luís, MA, Brazil ' Vale SA, São Luís, MA, Brazil ' Vale SA, São Luís, MA, Brazil

Abstract: Railroad switches are essential mobile mechanisms to control trains, guiding them from one track to another. However, they are subject to failures over time because of wheel attrition, component wear, and obstacles during the train movement. In modern railways, switch operations are controlled by point machines, which allow the switches to be operated remotely, enabling a more robust operational schema. Electric point machines generally receive commands from PLCs, which keep historical information from both commands sent and indications received from the railroad equipment. This paper proposes two heuristics developed and used to identify the four most common types of occurrences that can lead to future failures on railroad switches and point machines, avoiding emergency maintenance that stops the railway operation and avoiding possible accidents. This system enables us to analyse the data coming from Vale's railway dataset and identify possible operational issues. The obtained results using these heuristics show that it is possible to decrease the number of failures by almost 50% when we use the information as the starting point to apply predictive maintenance.

Keywords: heuristics; failure; detection; railway.

DOI: 10.1504/IJMIC.2024.144032

International Journal of Modelling, Identification and Control, 2024 Vol.45 No.4, pp.232 - 244

Received: 12 Jan 2024
Accepted: 05 Jun 2024
Published online: 21 Jan 2025 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article