Environmental-Noise Analysis In Active-Vision Systems For Adaptive Welding

Active vision for the real-time sensing of the joint geometry in front of the weld-ing torch is considered by many as being one of the best sensors for arc welding process automation as well as to improve the quality of the resulting welds. It is currently used for seam tracking and work is being executed to use it efficiently for adaptative process control [1-4]. One of the main problems in active vision systems for adaptive welding is the pre-sence of strong environmental noise from arc light and molten-particle emission. An analysis of such noise source is presented in this paper. The arc light intensity is first characterized for the three (3) main welding processes used for industrial robotic welding: GMAW in short-circuit, spray and pulsed spray transfer mode. Light noise generated by spatter is then considered. The relevance of such an analysis in terms of the choice of effective noise-reduction techniques is pointed out. The frequency bandwidth of the arc-emission signal determines the optimum modulation range for a synchronous-detection technique within the limits imposed by the characteristics of the viewing device. Similar considerations are valid concerning the travelling speed of the projected particles. Examples are given of several noise-reduction techniques, and their implementation with either flying-spot or bidimentional sensors is discussed.