Predicting A-TIG weld bead geometry of 304 stainless steel using artificial neural networks

The main issue that TIG welding experiences is limited weld penetration which eventually restricts its productivity. In order to overcome this difficulty and incomparable advantages of TIG welding, a method known as activating flux tungsten inert gas (A-TIG) welding was introduced, and is currently the focus of extensive research. When compared to traditional TIG welding, A-TIG welding is well capable of improving the aspect ratio by around 200% and weld penetration depth by about three times or even more. The present research tries to examine the impact of various fluxes on 8 mm thick grade 304 austenitic stainless steel regarding weld bead geometry. In this investigation, 3-factor 3-level RSM with Box-Behnken design method with 15 experimental runs were considered. In this experiment, a novel ternary flux mixture comprising of SiO2, TiO2 and Fe2O3 of different proportions such as 72:18:10, 65:25:10 and 45:45:10 were used. The sample with the optimal response, i.e. depth of penetration of 6.507 mm, bead width of 8.58 mm and reinforcement value as 0, had the welding parameters set as 160A current, 1.778 kJ/mm heat input, flux ratio of SiO2, TiO2 and Fe2O3 of 65:25:10 and root gap as 1.4 mm. A scanning electron microscope and an optical microscope are used to examine the microstructure of the weld zone. Additionally, hardness of the welding zone is assessed and the maximum value is measured as 66 HRC. Satisfactory results are obtained with ANOVA for validation and ANN as an estimation tool.