Autonomous Welding Seam Detecting and Tracking Using Vision and Sound Sensors in Robotic Gas Metal Arc Welding

Abstract

At present, robotic welding requires the robot to be taught by a human operator using teach and playback methods. When the weldment or seam is changed, the prior off-line programming cannot be reused. The paper proposed an autonomous welding seam detecting and tracking system. The system can detect realistic weld joints and determine their position in the robot workspace by extracting seam feature with image processing algorithms. The robotic trajectory is planned based on the extracted seam feature. During the metal-inert gas (MIG) welding process, the system can locate the welding position and track seam path by processing multi-sensor information combined of dual- microphone array sound signals and passive vision image from a CCD camera. When the deviation between welding pool centre and seam centre line is detected, the system can eliminate the deviation by adjusting the welding torch position. The experiments show that the proposed path planning method can identify butt weld joints autonomously regardless of the base material, surface finish and surface imperfections. Meanwhile, the multi-sensor system showed its potential merit in quality monitoring and control for the MIG welding process, whose deviation range of welding seam tracking accuracy is within ±0.17 mm.