Optimization of oxide-based activating flux combination formula in activated tungsten inert gas welding using a hybrid method including artificial neural networks and particle swarm optimization

The effects of the most important process-adjusting variables (welding current and welding speed) and the percentage of the combination of TiO$_2$ and SiO$_2$ activating fluxes on the most important quality characteristics (weld bead width, depth of penetration, and aspect ratio of these parameters) in welding of AISI316L austenite stainless steel parts are considered. Artificial neural networks (ANN) are used to determine the relations between the input variables and output responses of the activated tungsten inert gas (A-TIG) welding process. To determine the proper ANN architecture (the proper number of hidden layers and their corresponding neurons/nodes), the particle swarm optimization (PSO) method is used. Experimental tests are conducted to evaluate the proposed procedure performance. Based on the results, the proposed method is found to be efficient in modeling and optimization of the A-TIG welding process.