Interval model based robust control of weld joint penetration

Weld penetration sensing and control with a weld-face sensor are among the most relevant research issues in automated welding. Previous studies showed that the geometry of the weld pool contains accurate, instantaneous information about the weld penetration. In this study, the weld pool is measured in real-time to provide the feedback of the weld penetration, and the welding current is selected as the control variable. Analyses reveal that the influence of mandatory variations in welding conditions on the process dynamics can be described by an interval model that has bounded parameter intervals. A robust control algorithm with guaranteed closed-loop stability is used to overcome the interval uncertainty in the process dynamics. Dynamic experiments are performed using different welding conditions and varied welding parameters. From the experimental data the bounded parameter intervals are identified for the model of the process being controlled. Closed-loop control experiments are done under different perturbations. Experimentation shows that the variations encountered in practical welding can be overcome by the developed control system. In addition to penetration control, this work provides an example for developing robust manufacturing process control systems based on objective quantitative descriptions of the process uncertainty.