A methodology for modeling and control of weld bead width in the GMAW process

The need to improve productivity and quality has led to the development and improvement of techniques and systems for monitoring and controlling welding processes. This work presents a methodology to perform the modeling, optimization and control of the weld bead width, enabling the adjustment of process parameters in real time. An integrated system was developed for image acquisition, modeling and control of the welding process, allowing a real-time response, through artificial neural networks. Parameters such as welding speed, wire feed velocity and arc voltage are predicted in the function of a desired weld bead width. To get the closed-loop control system, it was designed with a "fuzzy" controller, in which the difference between the width to be achieved and the actual width of the bead is taken as reference. This weld bead is measured through an acquisition system and images processed using a low-price webcam. The control action is carried out preferably at welding speed, a parameter that has the greatest influence on the weld bead width and has no influence on the metal transfer behavior. Weld beads with pre-defined width, good appearance and quality were obtained.