Thermal Image-Based Wheel-Rail Contact Classification and Analysis Using Convolutional Neural Networks
[ 1 ] Instytut Transportu, Wydział Inżynierii Lądowej i Transportu, Politechnika Poznańska | [ SzD ] doctoral school student | [ P ] employee
2024
paper
english
- wheel-rail interface
- thermal-imaging
- convolutional neural networks
- image processing
- deep learning
- rail vehicle dynamics
EN The purpose of this paper is to outline the integration of thermal imaging measurements and Convolutional Neural Networks (CNNs) in studying the contact state of wheel-rail interaction under real operating conditions. The first part discusses the most important aspects of the mechanics of wheel-rail contact related to wear and heat generation in the context of vehicle operation and tram infrastructure. Special focus is placed on the origin and consequences of the multipoint contact between the wheel and rail. This is followed by a presentation of the methodology of thermal imaging measurement. The subsequent section focuses on the potential of CNNs to detect contact situations in the tram wheel-rail interface. The research concludes with a presentation of the most important research observations and outlines the further stages of the project.