Automatic inspection of wheel surface defects using a combination of laser sensors and machine vision

Abstract

The safety of railways is ensured by the regular maintenance of tracks, signals and rolling stocks. This mainly involves manual inspection by skilled maintainers and automated inspection systems are not commonly used. The development of an automated inspection system is crucial for maintaining good rail conditions. In this study on automatic surface defect detection systems, we focused on wheels as the target because they are critical components of rolling stocks, and any wheel defects can cause severe safety problems. Wheel-tread-profile dimensions have been measured using laser instruments in previous studies. In this research, to validate the accuracy of the measurement equipment, special test pieces with surface defects are prepared. We propose a combined inspection system using laser sensors, and a machine vision technique to confirm the effectiveness of the proposed system. An artificial intelligence (AI)-based surface defect detection system is employed to improve the proposed system. Combining edge detection through image transformation with an AI-based surface defect detection system offers remarkable benefits. The effectiveness of the system is experimentally confirmed using test pieces that simulate different materials and passing speeds of wheels.

Keywords

• artificial intelligence
• automated inspection system
• image conversion
• laser sensor
• wheel surface defect
• wheel tread profile
• machine vision