Towards region-based robotic machining system from perspective of intelligent manufacturing: A technology framework with case study

Abstract

The region-based robotic machining system is mainly used for repair and remanufacturing of the complex components with local defects. Taking the automotive body manufacturing as an example, the resistance spot welding process is highly prone to randomly distributed and tiny-sized welding slag splashes, which poses a great challenge to the locally automated removal of the welding slag defects. In this paper, we construct a region-based robotic machining system that integrates the processes of defect detection, defect region division and positioning, and machining path decision-making, aiming to enhance the automation of locally grinding of automotive body welding slags. Both the framework of the system and the implementation process are proposed. In the framework, the improved YOLO v3 algorithm is put forward to accurately detect the automotive body welding slags by virtue of the machine vision and deep learning. Based on this, the K-means algorithm is used to divide the welding slags into the regions, so that the problem of positioning individual welding slag is transformed into the problem of positioning the boundary points of welding slag region. The binocular image-based region boundary points matching is then presented to accurately locate the welding slag regions accordingly. The Dijkstra algorithm is finally employed to make an autonomous decision on the pre-planned grinding path of automotive body welding slags based on the positioning results. The experiments on a case of locally robotic grinding of the welding slags on an automotive door frame are implemented to verify the effectiveness and practicality of the system. This study provides a valuable reference for the locally automated repair of automotive body welding slag defects, putty defects, paint defects, etc.