Sustainable gas tungsten arc welding of titanium alloys: machine learning-based process optimization for enhanced mechanical properties and reduced heat input

In this study, a concept of sustainable gas tungsten arc welding (GTAW) operation has been proposed and optimized to reduce the heat input – HE (environmental aspect) and improve ultimate tensile strength – UTS (social aspect) and deposition efficiency – DE (economic aspect). The process parameters considered are the welding current (I), gas flow rate (F), torch angle (T), and arc gap (A). The Gaussian process regression (GPR) and adaptive-network-based fuzzy inference system (ANFIS) approaches are used to develop the predictive models of GTAW responses due to their nonlinear characteristics. An efficient algorithm entitled multi-objective non-sorted dragonfly algorithm (MONSDA) is developed to generate optimal solutions. The criteria importance through intercriteria correlation (CRITIC) and multi-criteria optimization and compromise solution (VIKOR) are implemented to calculate the weights and select the proper data. The findings presented that the optimal I, F, T, and A are 84 A, 13.5 L/min, 50°, and 2.0 mm, respectively. The UTS and DE are enhanced by 11.9% and 1.2%, respectively, while the HE is reduced by 4.9%, as compared to the initial values. The HE and DE models are significantly affected by the I, F, and A, respectively. The UTS model is significantly affected by the I, F, T, and A, respectively. The GPR models provide higher precision for the prediction purpose, as compared to the ANFIS correlations. The MONSDA is a prominent solution to deal with complicated optimization issues. The outcomes can be applied to decrease HE and boost the mechanical characteristics of the titanium grade 3 joints.