Abstract
With the integration of overlay welding and 3D printing technology, the possibilities of automatic cross-section repair have expanded significantly. However, there are still several challenges in perception and planning, such as uncertain and various geometric features (grooves, inner holes, and obstacles); particular welding process character, and molten pool fluidity. Conventional motion planning methods are inadequate to meet the complex requirements of the cross-section repair task. To address these challenges, this paper focuses on the visual perception and motion planning problem.Firstly, a Scan-Locate-Split-Merge (SLSM) workflow is proposed to deal with the perception problem.Based on the fact that the cross-section is a plane and all the geometric features are close to it, the cross-section plane can be identified from the point cloud and act as an anchor. Then all the geometric features can be clustered and classified according to their dimension and location. After merging these features, the task parameters such as plane boundary and groove radius and length can be extracted. Secondly, a hybrid overlay weld path planning method is proposed to deal with the planning problem. In order to eliminate the risk of overflow, an offset path is included in each layer; In order to improve the coverage rate, reciprocating paths is used in each layer. Furthermore, a special weld path connect pattern is proposed to reduce defects. The efficacy and reliability of the proposed method are validated through simulation analysis and physical experiments. The results demonstrate that the proposed method achieves an overall coverage rate of over 95% on the target surface. The repaired cross-section surface meets the flatness requirements.