Infrared Sensing Techniques for Adaptive Robotic Welding

"The objective of this research is to investigate the feasibility of using infrared sensors to monitor the welding process. Data were gathered using an infrared camera which was trained on the molten metal pool during the welding operation. Several types of process perturbations which result in weld defects were then intentionally induced and the resulting thermal images monitored. Gas tungsten arc using AC and DC currents and gas metal arc welding processes were investigated using steel, aluminum and stainless steel plate materials. The thermal images obtained in the three materials and different welding processes revealed nearly identical patterns for the same induced process perturbation. Based upon these results, infrared thermography is a method which may be very applicable to automation of the welding process.IntroductionAutomatic welding is commonplace in industry today. However, most, if not all, involves preprogramming for performance of a repetitive task. These systems are incapable of correcting for perturbations which arise during the welding process. Control of the welding process requires the identification and monitoring of perturbations in a wide variety of parameters. The varied nature of these parameters and the large number of variables involved have thwarted previous attempts at closed loop control of the welding process. The sensor(s) employed by a control system need to meet rather strenuous requirements. First and foremost of these is that such a sensor must be able to identify and discriminate between perturbations that might affect the quality of the weld, including such geometrical perturbations as a change of direction in the seam between the parts being welded, gaps in the seam or a misalignment of the parts being welded. The sensor should also be able to locate and identify contaminants and impurities in the weld puddle and ahead of it while monitoring the extent of the heat-affected zone and the cooling rate of the metal. The sensor(s) should, additionally, be able to perform these tasks in a time frame that is consistent with automatic control of the welding process. This paper presents results that could aid in the assembly of such a controlled welding system."