Characterization and real-time measurement of geometrical appearance of the weld pool

The weld pool contains important information about the welding process (Zhang et al., Welding J. 72, 463s-469s, 1993). In this study, a polar coordinate model is proposed to characterize the weld pool geometrically. The identification of its parameters involves complicated non-linear optimization which cannot be done in real time using conventional algorithms. A neural network is therefore proposed to identify the parameters in real time. By using pulsed laser illumination, clear images of the weld pool are captured. The developed image processing algorithm extracts the boundary of the weld pool in real time. Thus, a real-time system is developed to sense and process the image and identify the polar coordinate model. It is shown that the weld penetration can be accurately determined using the model parameters. Thus, a real-time weld penetration monitoring system is also achieved. Copyright (C) 1996 Elsevier Science Ltd.