One-shot, integrated positioning for welding initial points via co-mapping of cross and parallel stripes


Robotic welding is gradually advancing towards intelligent integrated welding with integration of different seam types. In this process, the initial point positioning of weld seams is a foremost technique for ensuring a smooth subsequent welding process. However, existing studies on welding initial point positioning are not well integrated in terms of different groove shapes and generally require robot movement to search multiple times. This problem entails a high development cost in varying welding scenarios and with efficiency to be improved. Meanwhile, robustness of positioning is an ongoing challenge, represented by workpiece tilt and misalignment. To cope with these issues, we developed an integrated vision sensor based on co-mapping of cross and parallel stripes to achieve one-shot initial point positioning within a defined area for four typical seam types. Among them, the cross stripes are used to extract workpiece edge parameters and the parallel stripes to extract seam parameters, both are sequential. At the core, we proposed an interval-restricted search algorithms to extract the seam points, and combine it with the edge parameters to obtain the initial points. In addition, a series of parametric analyses are performed for detecting workpiece misalignment and determining the initial points. Experimental results show that the co-mapping of cross and parallel stripes achieves one-shot high-competitive accuracy for the initial point positioning of the four seam types even if the workpiece is tilted or misaligned. © 2023 Elsevier Ltd

Robotics and Computer-Integrated Manufacturing
Yuqian Lu
Yuqian Lu
Principle Investigator / Senior Lecturer

My research interests include smart manufacturing systems, industrial AI and robotics.