In order to solve the problem of combining plasma equipment with robots to process casting risers, where variations in riser dimensions due to mold accuracy necessitate precise trajectories and poses for proper arc initiation and operation, a method based on a process parameter library was employed for investigation. This method was comprised of process parameter library, analysis module, similarity measurement module, path generation module, and parameter module. Through the construction of two-stage comparison method, target parameters can be quickly filtered from the library. Using path generation algorithms and deep learning algorithms, working paths and process parameters matching the point cloud model were obtained. The results show that this method can accurately generate operational trajectories and poses based on the workpiece, enabling the reduction of manual intervention, lowering of operational complexity, and enhancement of the reliability and safety of the cleaning operations for coupler castings.