Machine learning enhanced response surface methodology for hardness optimization in TIG welding of Al 6061-T6 alloy

This study presents an integrated approach combining machine learning and response surface methodology to optimize TIG welding parameters for enhancing the microhardness of Al 6061 T6 alloy weldments. Experimental trials were designed using central composite design with varying welding currents and filler rods. Gradient Boosting and Random Forest models were developed to predict microhardness based on process parameters. Among the models, Gradient Boosting showed superior accuracy with a high R-squared value and low prediction errors. Partial dependence analysis revealed that both welding current and filler rod type significantly influence hardness, with ER5356 showing the most consistent performance. The predicted optimum parameters - 65 amperes and ER5356 were validated experimentally, yielding a microhardness of 78 ± 2 Hv 0.3 . The close agreement between predicted and actual values confirms the effectiveness of the proposed approach. This study demonstrates the potential of machine learning enhanced response surface methodology for reliable optimization of welding processes returns.