Multivariable Adaptive Control of Thermal Properties During Welding

Complete solution of the welding control problem involves regulating weld geometry, residual stresses, distortion and weldment metallurgical characteristics. The latter has been largely ignored in closed-loop control work, yet it is a basic determinant of weld integrity. In this paper, a previously developed input-output model for continuous regulation of the critical thermal properties of heat affected zone and cooling rate is used to develop a closed-loop MIMO control scheme. Since the physical system is non-linear and non-stationary, and process disturbances are reflected as model parameter changes, an adaptive control method is used. The implementation of the controller is based on deadbeat adaptive control, and it is applied to several different forms of the basic welding process model. The different models reflect a need to simplify the control system to accommodate available process measurements and hardware capabilities. A series of experiments performed on a robotic Gas Metal Arc welding system are presented that confirm the basic stability, robustness, adaptivity, tracking and disturbance rejection properties of the scheme. Bandwidth limitations are identified and improvements (primarily hardware oriented) are suggested.