Data-driven optimization of material composition and casting processes for large ferritic nodular cast iron components in wind turbines

Abstract This study addresses the challenges of producing large-scale wind turbine components by proposing an optimization scheme for the material properties and casting processes of ductile cast iron. The study integrates experimental data with JMatPro calculations to establish a comprehensive database of 1, 139 ferritic nodular cast iron samples. Using machine learning and multi-objective optimization, an optimized composition was developed, and Pareto front solutions were achieved with performance significantly surpassing QT400-18. The AI-CAST platform was then applied to refine the casting process for a 6 MW wind turbine hub, reducing internal defects to 2, 738.52 mm³ and decreasing loose defects by 83.68%. This integrated approach not only enhances material performance and casting quality but also offers significant cost savings, suggesting broad applicability for improved production efficiency in the casting industry.