Laser scanning-based robot welding path planning for complex saddle-shaped weld seams

Saddle-shaped weld seams of the header-and-nozzle joints challenge both manual and robotic welding with their complex spatial characteristics, poor welding consistency, and the need for multi-layer and multi-pass welding. This paper proposes a multi-layer and multi-pass welding path planning method based on laser scanning and 3D reconstruction for complex saddle-shaped weld seams. A user coordinate system is established based on the workpiece benchmark determined by the continuous scanning of the scanner and the coordinate transformation. The 2D segmented weld seams feature is converted from the 3D feature obtained by a given scanning pose in the user coordinate system by a dimension reduction transformation. The path planning is completed by the equal-area method on the filled area determined by the contour of the welding layer obtained from the simplified curve with an accuracy within +/- 1 mm. The robot's 3D trajectory is formed by the 2D path points represented by the centroid of welding pass on each cross-section after inverse space transformation. Experimental results confirm the path planning's efficacy in robotic welding automation, ensuring good welding formation.