A method for detecting wear and damage on railcar wheel tread surface using a combination of laser measurement and machine vision

Abstract

Skilled technicians primarily use manual tools to maintain the geometry and surface conditions of railcar wheels. To ensure efficient wheel maintenance, the development of an automated inspection system is required. Herein, an inspection method for railcar wheels is proposed based on the assessment of wheel tread dimensions and surface conditions and investigation results of the method are presented. The parameters of wheel tread dimensions and surface conditions are considered because they are crucial for railway operation safety and manually inspecting them requires a large workforce. To improve the efficiency of the system, a combined inspection method through laser measurement and machine vision is used, both the techniques compensate for each other's shortcomings. The experiments are conducted using laser sensors and machine vision techniques, including defect detection using AI model based on YOLOv5, as well as image data captured via a digital video camera to obtain high-accuracy automatic measurements and determine the feasibility of such measurements. Additionally, the measurement accuracy of the proposed method is verified based on experiments on sample wheel having wheel tread surface defects. In this paper, sample wheel including various wheel tread surface defects is used.