Machine Learning-Based Prediction of Impact Toughness in AISI 430-AISI 304 Friction-Welded Joints

Friction welding is a solid-state joining technique widely used for dissimilar materials due to its efficiency and cost-effectiveness. However, welding AISI 430 ferritic and AISI 304 austenitic stainless steels poses challenges due to their differences in chemical composition and mechanical properties, particularly in achieving optimal impact toughness. Existing studies focus on optimizing process parameters but lack predictive modeling for impact toughness using machine learning (ML). This study aims to bridge this gap by experimentally evaluating the impact toughness of friction-welded AISI 430-AISI 304 joints and developing ML models for accurate prediction. Experiments were designed using a Taguchi L32 orthogonal array, varying friction force, forge force, and burn-off. Charpy impact tests were performed to assess toughness, and the results were used to train Decision Tree, Random Forest, and Gradient Boosting regression models. Random Forest regression outperformed others with an R2 value of 0.98 and a mean squared error of 0.29. The predicted impact toughness (15.8 J) closely matched the value from the confirmation experiment (16 J) with only a 1.25% deviation. The findings demonstrate that ML can significantly enhance process optimization, reducing reliance on costly experimental runs. Future research should explore additional welding parameters and deep learning models for further improvements.