Comparative study of moving train hot boxes pre-detection and axles counting by in-situ implementation of two infrared cameras

Abstract : The present study addresses an in-situ infrared thermography method for detecting train's hot boxes during operation. Detecting axle overheat during train's rolling is a major concern for railway operators. In fact, such overheat can result in the axle journal break, damages on the track, a decrease of the tread etc.; possibly leading to important damages or even a derailment. Actual solutions are made of an infrared sensor, placed on the rail that will emit a signal if the measured temperature at wheel's level is to high. However, even if this solution has been used for many years now and standardized, those sensors are facing practical and economical difficulties. In fact, those hot boxes detectors are placed in the ballast of the railway track. Location that leads to non negligible civil engineering cost during its implementation and operative maintenance. As a result, using an infrared camera on the trackside would allow easier and less expansive maintenance cost. Moreover, by using all the information inside the images, infrared thermography opens up new perspectives towards monitoring by counting at the same time the axles number for example. The objective of this study is to check the feasibility of in-situ implementation of two technologies of infrared thermal cameras and exploit the gathered information for hot boxes detection. First numerical experiments were conducted using a 3D scene infrared software simulator studied and develop in the laboratory. It helped with the choice of the wavelength bands and acquisition frequency of the cameras to be used. Then in-situ experiments have been conducted and a comparison with the numerical data is made. Moreover, firsts automated axles counting and hot box detection algorithms are introduced. Finally, obtained results are presented and discussed with potential improvements. Some perspectives are proposed.
Document type :
Conference papers
Complete list of metadatas
Contributor : Thibaud Toullier <>
Submitted on : Wednesday, August 7, 2019 - 2:39:21 PM
Last modification on : Thursday, August 8, 2019 - 1:16:31 AM


  • HAL Id : hal-02264672, version 1



Thibaud Toullier, J. Dumoulin, V Bourgeois. Comparative study of moving train hot boxes pre-detection and axles counting by in-situ implementation of two infrared cameras. QIRT Asia 2019 - Quantitative InfraRed Thermography Conference, Jul 2019, Tokyo, Japan. ⟨hal-02264672⟩



Record views