Comparison of Response Surface Methodology (RSM) and Artificial Neural Networks (ANN) in Optimisation of the Thermal Diffusivity of Mild Steel TIG Welding

This study compares the effectiveness of Response Surface Methodology (RSM) and Artificial Neural Networks (ANN) in optimizing the thermal diffusivity of mild steel Tungsten Inert Gas (TIG) welds. The analysis evaluates the predictive accuracy and optimization efficiency of both techniques, providing insights into their suitability for modeling thermal behavior in welding applications. The set of tools, including power hacksaw cutting and grinding machines, mechanical vice, emery (sand) paper and sander was used to prepare the mild steel coupons for welding. The produced coupons were evaluated for their Thermal Diffusivity. The two expert systems used to determine the effect of the interaction of welding current, welding voltage and gas flowrate on the Thermal Diffusivity were the Response Surface Methodology and Artificial Neural Network. The models were validated using the model summary values between the experimental results compared to RSM (R<sup>2</sup> = 94.49%) and ANN (R<sup>2</sup> = 97.83%) values. This shows that ANN is a better predictor as compared to RSM.