A novel filling strategy for robotic multi-layer and multi-pass welding based on point clouds for saddle-shaped weld seams

The saddle-shaped weld seam of the boiler header and tube-seat presents challenges such as low efficiency, high labor intensity, poor post-welding consistency, and unstable quality. This paper developed a vision system based on a laser 3D camera for scanning groove zones and reconstructing point clouds, enabling functionalities such as point cloud acquisition, feature extraction, and adaptive planning of weld passes on the saddle-shaped weld. The saddle-shaped weld was theoretically analyzed based on the three-dimensional digital model of the header pipe seat. We optimized the weld trajectory characteristic curve and designed a layer-by-layer filling strategy for abdomen pre-filling and welding around the pipe axis. The effectiveness of the saddle-shaped weld point cloud Trimmed Iterative Closest Point (Tr-ICP) algorithm was validated through the running time of the point cloud registration algorithm and evaluation of registration error scores, with precise registration Euclidean distance being within 0.2 mm. The Random Sample Consensus (RANSAC) algorithm was utilized to determine the main pipe and branch pipe axis parameters, obtaining the user coordinate system origin pose of the saddle-shaped weld on the actual workpiece through the point cloud transformation matrix. The three-dimensional point cloud was segmented, projected, and provided with the point direction equation for single point clouds per 2°. We extracted the saddle-shaped groove's upper and lower edge points, corrected the workpiece processing and assembly errors, and finally realized multi-layer and multi-pass adaptive robot welding of the saddle-shaped weld of the header and tube-seat. Experimental results demonstrate that this method offers valuable ideas and technical approaches for robot welding of complex space intersecting line curves. Our results have been verified for reliability in complex spatial weld seams, which is crucial to the development of intelligent robotic welding.