Intelligent modeling and optimization of activated gas tungsten arc welding process

The activated gas tungsten arc welding (A-GTAW) process, employing an activating flux on the workpiece surface, enhances weld penetration compared with conventional GTAW. This study presents a modeling and optimization approach for A-GTAW of AISI316L stainless steel, considering welding speed (S), current (C), and flux combination (F) as input variables, and depth of penetration (DOP), weld bead width (WBW), and aspect ratio (ASR) as outputs. Data were generated using response surface methodology and modeled with a backpropagation neural network (BPNN). Particle swarm optimization and simulated annealing were applied to optimize the BPNN-driven model. For the optimized conditions, key weld quality indicators including: heat-affected zone (HAZ) width, ultimate tensile strength (UTS), microhardness, and elongation were evaluated for micro- and nano-based A-GTAW process. The nano-based A-GTAW achieved significant improvements over conventional GTAW: DOP (93%), UTS (61%), HAZ width (50%), elongation (45%), and WBW (24%). The optimal flux coating was 1.8 mg/cm² with 75% SiO₂ + 25% TiO₂. The proposed methodology demonstrates efficient multi-objective modeling and optimization of A-GTAW, with errors below 4%, providing potential benefits for industrial welding applications.