A neuro-fuzzy approach for increasing productivity in gas metal arc welding processes

This paper focusses on a study carried out in order to increase productivity in gas metal arc welding (GMAW) processes by optimising the deposition rate of the filler metal. To reach this aim, a possible solution was found in developing an adaptive system that is able to control and keep the wire feed speed constant at a desired and optimal value. This control has been accomplished by regulating an opportune variable typical of the welding process; in this case, the attention was focussed on the welding current intensity. Typical difficulties of GMAW processes, due above all to the great number of main variables and to their interdependence, suggested the possible solution by modelling a fuzzy-logic-based system, whose elements were determined by training an artificial neural network (ANN) with experimental data, obtained from bead on plate welds. At the same time, mathematical models, based on multiple regression analysis, were developed from the same data, in order to provide a comparison term and to assess the effectiveness of the neuro-fuzzy approach versus the mathematical methods. The results of this study confirmed the effectiveness of the proposed approach in the development of an integrated welding system in order to increase productivity.