Welding robotic systems with visual sensing and real-time control of dynamic weld pool during pulsed GTAW

Real-time sensing and control of weld pool dynamics are important techniques to ensure expected weld quality because there are many uncertain effects during robotic arc welding. A current teaching play-back welding robot is not provided with this real-time function for sensing and control of the weld process. In our research, using composite filtering technology, a computer vision sensing system was established and clear weld pool images during pulsed Gas Tungsten Arc Welding (GTAW) were captured in robot moving process. A corresponding image-processing algorithm has been developed to acquire characteristic parameters of the weld pool in real time. Based on both PID controller and neuron self-learning PSD controller design, real-time control of dynamic weld pool for topside weld shape during robotic-pulsed GTAW process has been realized in this study.