Vision-based weld line detection and real-time tracking in MAG fillet joint welding

Welding techniques are commonly used to join steel materials in structures such as architectural steel frames and bridges. Fillet joint welding is used in practical welded applications. Arc sensors detect the weld line based on that current or voltage change as the length of the arc changes during the fillet joint welding. However, the vertical plate efficiently gets heated because of the gap that exists between the vertical plate and the horizontal plate in fillet joint welding. It leads to undercut defects appearing on vertical plates. In this study, the authors proposed a method to visualize the melting state during Metal Arc Gas (MAG) welding using the visual sensor. The authors grab the weld pool images during welding by using the camera. The authors proposed an image processing method to determine the weld line position and wire tip position from the captured images. A Proportional-Integral (PI) controller was proposed to track the weld line in real-time and weld line imitation control was performed with good results. The proposed method has achieved a mean absolute error of 0.27 mm, and a root mean square error of 0.34 mm to demonstrate the effectiveness of the method. Furthermore, the base material with a 1-mm gap was verified, and the results showed that the control method has good adaptability and stability. Visual sensing technology and control optimization can be used even when arc sensors are difficult to apply. The camera is fixed on the welding torch to reduce the space required to be used in smaller welding area. It can improve efficiency and reliability in fillet joint welding processes.