Real-time control of torch height in NG-GMAW process based on passive vision sensing technology

In narrow gap gas-shielded arc welding (NG-GMAW) for pipelines, maintaining a stable welding process and ensuring weld quality necessitates controlling the extension length of the welding wire (WWEL) within a specific range. However, when dealing with three-dimensional weld workpieces featuring height variations, welding defects are prone to occur due to changes in welding wire extension length. Therefore, real-time adjustment of the distance between the contact tip and workpiece (CTWD) is crucial during the welding process. To address this challenge, this paper proposes a welding torch height (WTH) control method based on passive vision sensing. The proposed method utilizes a wide dynamic range (WDR) camera to acquire distinct real-time welding images. An adaptive region of interest extraction method for the welding wire is then proposed based on the position relationship between the welding wire and arc. To address false edge issues in the welding wire profile, a cellular neural network (CNN) edge detection algorithm, optimized by particle swarm optimization, is employed to eliminate false edges. The extended length of the welding wire is subsequently extracted using an adaptive mask kernel morphology and corner detection method. Accordingly, a model predictive control (MPC) technique is developed to govern the height of the welding torch with the WWEL as input. The proposed MPC algorithm's tracking performance and robustness are validated through feedback control experiments. The results indicate that the tracking error of the WTH trajectory can be controlled within±0.41 mm, meeting the requirements of NG-GMAW welding torch height control for welding robots.