Process modelling to establish control algorithms for automated GMAW
The feasibility of fully automatic GMAW processes may rely on the development of sophisticated equipment to emulate the manual welding torch oscillation pattern or on the development of high level methods of control to prevent the appearance of defects, especially the lack of sidewall fusion. An intermediate solution is to optimise the weaving parameters of a conventional pattern oscillator in such a way as to minimise the level of rejection. A prototype of a computerised system to work with Pulsed-GMAW equipment, in the vertical-up position, was proposed to produce a minimal level of rejection for welds in plates up to 25 mm thick. The system basically consists of optimised mode control algorithms, based on theoretical and experimental models of weld pool behaviour. Three tasks are performed by the system; the selection of parameters for an optimum working point, an off-line simulation of the operation and real-time error monitoring of the process. Statistical experimental modelling was applied in order to build most of the optimised models, because of the large number of variables to be treated and their complex inter-correlation. The welding variables were correlated with single responses. Partial and Correlation Analysis techniques were used to discover the relationship between the variables and the responses. Regression Analysis was then applied as a means of obtaining the 'weight' of the most significant variables. Finally, since some variables were found to be collinear, a corrective technique for biased variables was employed. Acceptance criteria for bead shapes were proposed and assessed. The effect of the oscillation parameters and other welding variables on the bead formation was analyzed and an operational 'envelope' for the parameters determined. A theoretical approach to predict the occurrence of poorly shaped beads, due to the lack of metal bridge between the joint walls, was successfully developed and applied in parallel with the statistical experimental methods. Equations for optimising the bead shape and for determining the operational envelope contours were subsequently generated and evaluated. An extension of the system to an actual adaptive control scheme was discussed and sensors and signals to be used were evaluated. Finally, a process instability phenomenon in long test plates was identified and investigated. This instability may prevent the use of GMA W in some conditions in the vertical-up position.