Seam tracking control of robotic pulsed mag welding by extraction of visual and arc features

Abstract

Currently, most of the robots used in arc welding applications are of the "teach and playback" type, that is they do the welding production by repeating predefined actions. So in order to meet the high quality requirement of the welding process, the demonstrated trajectory of the robot should be close to the weld seam as much as possible. However, this condition is difficult to realize because the seam position is often disturbed by machining error, misalignment error, welding distortion and so on. In this paper, the seam tracking technology for Pulsed MAG welding is studied. The vision sensor is used to collect the welding image, which can get the 2D seam-tracking information. Meanwhile, the arc sensor is used to get the information of the welding torch height, which is also important for the welding quality and difficult to acquire through the vision sensor. The composite sensors technology based on these two sensors can be used in the weld of 3D seam, which provides a possible solution for real-time tracking of 3D weld seam during robotic Pulsed MAG welding process. With the development of welding robots, the increase of the labor cost and the shortage of skilled welders, the seam tracking technology studied in this paper supply important theoretical significance and practical value for promoting the application of the welding robots. To realize the seam tracking function, this paper rebuilt the traditional "teach-playback" robot. On the base of analyzing the characteristics of Pulsed MAG welding and its arc spectrum, a set of visal system was designed to obtain clear welding images. Hall arc sensors were applied to detect the arc information during the welding process, and the program based on DSP was developed to collect and process the signals. The control software based on multithreading program was developed to realize the functions of collecting visual and arc information, information processing and the control of DSP and the robot. For the CCD camera system with one-reflex-ray-path, the relationship mapping was established between the image coordinates and the robotic coordinates. Then the required conditions were analyzed when such relationship mapping was effective, and the original transform formula was modified according the Cartesian coordinates and the joint coordinates. The modified transform formula could be used directly when the robot pose changed. Then in this paper, image features of the Pulsed MAG welding were researched. An adaptive window segmentation method was designed to acquire the arc and weld seam region from the background. An algorithm to extract the wire projection using threshold segmentation was developed by analyzing the gray scale distribution of the arc region. Since the single criterion was difficult to recognize the seam edge from the noise accurately, multi-criterion was proposed in this paper, and some prior knowledge was also introduced to assist the location of the seam edge. The designed algorithm can detect the weld seam edge fast, accurately and stably. After that, possible arc features relating to the welding torch height were proposed according to the characteristics of the pulse power supply and experiments. By analyzing arc information, a filter method based on characteristics of the noise signal was designed to reduce the interference of the noise and improve the accuracy and stability of the arc feature representing the welding torch height. The feature selection algorithm was applied to find the best arc feature and ideal processing parameters. Then the relational model between the arc feature and the torch height was established. To control the trajectory of the robot by the computer, this paper studied the algorithms of line interpolation and circle interpolation in robot working space, based on which a position adjustment method was designed to adjust the robot path in real time. A composite control strategy using adaptive fuzzy PID and fuzzy algorithms was designed to correct robot path. For the lag error introduced by image processing, this paper developed an error compensation algorithm based on modification of the wire projection on the workpiece, which can improve the accuracy and stability of the tracking process. Finally, on the base of above research, a robotic arc welding system with 3D seam tracking function was established and some seam tracking control experiments were conducted. According to the experiments, the system can track different type of seams and get good weld formation. The tracking accuracy can meet the practical requirements of the welding production.

Date of Award	2013
Original language	English