An adaptive control system for off-line programming in robotic gas metal arc welding
The aim of this work was to develop an integration concept for using off-line programming in robotic gas metal arc welding of thin sheet steel. Off line-welding parameter optimization and on-line monitoring and adaptive control of process stability and torch-to-workpiece relative distance were used to ensure weld consistency. The concept developed included four main aspects: a) the use of a CAD system to design the workpiece; b) the use of a welding off-line programming system to design the welds, generate the welding parameters and to extract geometrical information from the CAD models to generate a robot program; c) the use of a graphical simulation system to simulate the robot movements; and d) the use of monitoring and adaptive control for ensuring that the required weld quality is delivered. The CAD system was chosen to be the basis for the development of the welding off-line programming system. The generation of optimized welding parameters was based on empirical welding models and the robot program generation was based on on-line programming experience. A PC based monitoring and control system was developed to provide on-line position and process control. The position control was carried out by pre-weld adjusting the initial position of the workpiece using a wire touch sensor and on-line adjusting the torch-to-workpiece distance by moving the workpiece based on the information provided by a through-the-arc sensor. The process control was carried out by automatically trimming the welding voltage such that the most stable process could be obtained. The stability of the process was estimated by using previously established monitoring indices. It was assumed that the off-line welding parameter optimization would provide the deposition rate necessary to produce the required weld quality. Successful welding control trials were performed showing the effectiveness of the adaptive control strategy. An off-line programming system has been developed and the programs generated have been tested by simulation. This showed that simulated positioning errors, produced by deliberate wrong path data, were successfully compensated for by the control system developed in this work.