Machine Learning-Based Analysis of Surface Hardening in Shot-Peened Superalloys

This study presents a Bayesian-optimized machine learning approach for predicting grain size and hardness variations in shot-peened superalloys. Using combined DEM–FEM numerical simulations across various superalloys, a dataset was generated with an optimized dataspace. The resulting regression model exhibited strong predictive capability, achieving coefficient of determination (R²) values of 0.918 for grain size and 0.956 for hardness, with corresponding root mean square error values of 7.9 and 4.1%, respectively. It is suggested that the Bayesian optimization was instrumental in discerning complex patterns and dependencies among input variables and output targets, contributing to the model’s high performance. The results also revealed that the weight function of elastic modulus, yield strength, and hardness became more pronounced with increasing grain size and hardness ratios during shot peening, maximizing prediction accuracy. Additionally, serving as a case study, a comparative analysis was conducted between two superalloys, IN718 and Nimonic 80, showcasing the capacity to fine-tune input parameters for attaining a targeted surface hardness of 400 VHN and an average grain size of 5 μm following the shot peening process. Overall, the proposed model serves as a valuable tool for accurately predicting objectives and designing processing parameters to attain desired hardness and grain size refinements.