Machine Learning-Driven Optimization of Electrochemical Corrosion Behavior in Y₂O₃-Reinforced AZ31B via Friction Stir Processing

Friction Stir Processing (FSP) was employed to reinforce AZ31B magnesium alloy with Y₂O₃ particles, leading to a significant enhancement in the electrochemical corrosion resistance. The electrochemical corrosion rate (ECR) decreased from 5.56 mm/year to 0.4511 mm/year. To predict and optimize these improvements, a Gaussian Process Regression (GPR) model was developed and trained using experimental data, achieving high accuracy (R² = 0.945). The model was coupled with the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to identify optimal processing parameters. Electrochemical behaviour was validated through Tafel analysis, which confirmed the improved corrosion resistance of optimized samples. Overall, this study demonstrates a synergistic approach combining surface engineering, machine learning, and evolutionary optimization for tailoring corrosion resistance in magnesium-based systems.