Robust Particle Swarm Optimization Algorithm for Modeling the Effect of Oxides Thermal Properties on AMIG 304L Stainless Steel Welds

There are several advantages to the MIG (Metal Inert Gas) process, which explains its increased use in various welding sectors, such as automotive, marine, and construction. A variant of the MIG process, where the same equipment is employed except for the deposition of a thin layer of flux before the welding operation, is the AMIG (Activated Metal Inert Gas) technique. This study focuses on investigating the impact of physical properties of individual metallic oxide fluxes for 304L stainless steel welding joint morphology and to what extent it can help determine a relationship among weld depth penetration, the aspect ratio, and the input physical properties of the oxides. Five types of oxides, TiO2, SiO2, Fe2O3, Cr2O3, and Mn2O3, are tested on butt joint design without preparation of the edges. A robust algorithm based on the particle swarm optimization (PSO) technique is applied to optimally tune the models' parameters, such as the quadratic error between the actual outputs (depth and aspect ratio), and the error estimated by the models' outputs is minimized. The results showed that the proposed PSO model is first and foremost robust against uncertainties in measurement devices and modeling errors, and second, that it is capable of accurately representing and quantifying the weld depth penetration and the weld aspect ratio to the oxides' thermal properties.