Optimization of Eutectic Carbide Inhomogeneity in M42 High-Speed Steel through Machine Learning and Finite-Element Modeling

The optimization of forging processes in M42 high-speed steel is crucial for enhancing its microstructure and mechanical properties, particularly in reducing eutectic carbide inhomogeneity. In this study, machine learning (ML) is integrated with finite-element modeling (FEM) to address the challenges of process parameter optimization in large-sized ingots. The random forest algorithm is employed to predict the inhomogeneity of eutectic carbides using strain variables derived from FEM simulations as input features. The optimized process, validated through experimental analysis, demonstrates a significant improvement in carbide fragmentation, leading to a more uniform distribution of fine precipitates. The optimized M42 steel exhibits superior mechanical properties, with yield and compressive strengths increasing by ≈115 MPa and 305 MPa compared to the prior forging process. In the results, the efficacy of ML-driven optimization is underscored in achieving a refined microstructure and enhanced material performance, offering a promising approach for industrial applications of high-speed steel.