Urban rail component operation quality detection system based on deep learning and bilateral threshold algorithm

Abstract

As an important part of the bogie, the health of the rotating parts of urban rail vehicles directly determines the safety of vehicle operation. The traditional manual train inspection method can only judge whether there is a fault through the appearance state, which is time-consuming. This paper, for the starter, puts forward a scheme- that is to install trackside thermal imager in the return reservoir area to cover vehicle rotating parts, and to complete the screening gearbox, based on YoLo-v4 deep learning frame; for the second, a train inspection quality evaluation algorithm based on temperature is also proposed, effectively extracting fault features and evaluate the working condition of components.