Prediction and Optimization of Activated Tungsten Inert Gas Welding Process Parameters using Expert Systems

Abstract

The achievement of bead penetration in Tungsten Inert Gas Welding plays a crucial role in determining the ultimate weld quality. Multiple machining parameters exert an influence on weld bead penetration, hence, making it imperative to find the optimal combination of machining process variables. The present study analysed the optimization of machining variables for the Tungsten Inert Gas Welding method through the utilization of Response Surface Methodology (RSM) and Artificial Neural Network. A design matrix was created for experiment planning, employing the central composite design methodology. Input process parameters encompassed welding current, welding speed, and wire diameter, while the investigation focused on bead penetration as the output variable. Statistical models for response characteristics were established utilizing experimental data. A total of twenty tests were executed according to the design matrix, involving the welding of two mild steel plates sized at 60 x 40 x 10mm. Bead penetration, the primary response variable in this study, was subsequently measured and recorded on the samples. The findings reveals that a combination of 208.22A welding current, 3.02 mm/s welding speed, and a 2.55mm wire diameter yields an optimal bead penetration of 10.3054 mm. In the case of the Artificial Neural Network (ANN), 70% of the data was allocated for training, 15% for validation, and the remaining 15% for the actual test. Based on the results, a regression plot demonstrates an overall R-value of 0.88521. The Response Surface Methodology is chosen as the superior predictive model compared to the Artificial Neural Network because the RSM output aligns more closely with the experimental data than that of the ANN.