Abstract: The robot was used to weld round parts of a power plant and make welding procedure qualification. Finite element analysis method was adopted to optimize layout of weld pass and starting arc coordinates of each weld pass were established in coordinate system of workpiece. Feature points of workpiece were captured by laser vision of robot, and coordinate systems of workpiece and robot were fitted with three-point centering algorithm, so that parts placed arbitrarily in the range of robot workbench could be centered automatically. According to a large number of cladding forming tests, starting arc position of each weld pass was corrected, and the corrected starting arc coordinates and the optimized layout of weld pass were edited into robot welding path program in the fitted coordinate system. Combined with welding process program stored in the welder, parts in the range of machine’s workbench were intelligently positioned, and the robot could automatically walk and invoke welding process to realize intelligent welding. Finally, circular parts were tested by non-destructive inspection and physical and chemical inspection, and the results were qualified.Highlights: (1) For the distribution of layer and pass and arrangement order of weld in a specific part, finite element analysis method was used to optimize from a variety of paths.(2) Under conditions of specific parts and specific process, the optimized welding arc position and walking path were programmed to give the robot internal basis for realizing automatic welding path.(3) Feature points of parts were positioned visually by the robot, and external positioning of parts in the robot system was obtained according to specific algorithms, so as to realize the unification of the optimized welding path in the robot space system. It could avoid the process of locating and finding circle of parts.