Vision-based observation of keyhole geometry in plasma arc welding

Abstract

To understand the thermo-physical behavior of a keyhole and weld pool in plasma arc welding, it is essential to measure and monitor the dynamic variation information of the keyhole geometry in real-time. In this study, a vision system is developed to acquire the images from a keyhole from the underside of the work-piece. When a fully penetrated keyhole (open keyhole) is formed, clear keyhole images are captured. The characteristics of the keyhole images are analyzed according to the keyhole forming principles and the keyhole channel shape. The whole keyhole boundary is extracted by utilizing an image processing algorithm. With calibration of the vision system, the geometric parameters of the keyhole, including keyhole length and width, ratio of keyhole length to width, and keyhole displacement at the exit relative to the entrance, are obtained. Examining the variations of these parameters with time, the keyhole behavior is analyzed under different welding conditions. It is found that the keyhole shape at the exit is oblate, and its centre is deviated behind from the torch axis. The observation results lay a solid foundation for controlling keyhole stability and for validating numerical models of thermo-physical behaviors in keyhole plasma arc welding.