Keywords: wheel flat detection, wayside condition monitoring, train-track interaction, damage classification, unsupervised learning, machine learning.

Wheel flat defect is the most common type of damage on train wheels, which can cause a high impact on the railway infrastructure and vehicle components. Artificial intelligence techniques for detecting geometric defects in train wheels have significantly enhanced railway maintenance and safety efficiency. Artificial intelligence systems excel in analyzing intricate wheel rotation patterns, swiftly and accurately identifying potential geometric deformations that could lead to wheel flats. Compared to traditional methods, artificial intelligence-driven defect identification provides a faster and more reliable approach, ensuring the safety and reliability of railway operations. This study proposes a discrimination learning algorithm to identify railway wheel flats, which consists of two stages: i) wheel flat detection to distinguish a healthy wheel from a damaged one, ii) classification of wheel damage based on its severity. To validate the unsupervised learning method, synthetic data acquired from a virtual wayside monitoring system is used, considering freight train passages, including wheels afflicted by single or multiple defects. The developed methodology in this study represents effectiveness in detecting wheel flats and assessing the damage severity, regardless of the number of defective wheels.

download the full-text of this paper (PDF, 15 pages, 1070 Kb)

go to the previous paper
go to the next paper
return to the table of contents
return to the volume description