فهرست نشریات علمی دسترسی آزاد ایران

Designing high-yield-strength steel alloys remains a key challenge in materials engineering, as traditional trial-and-error approaches are costly and inefficient. This study presents an intelligent two-stage framework integrating machine learning and metaheuristic optimization to accelerate alloy discovery. First, a Random Forest model was trained on experimental data, achieving a high predictive accuracy for yield strength (R² = 0.8194, MSE = 12445.02). This model was then employed as the objective function within the Cuckoo Optimization Algorithm (COA). After 100 iterations, COA identified an optimal alloy composition with a yield strength of 2456.46 MPa, significantly exceeding the maximum value in the original dataset. The optimized composition features substantial percentages of key strengthening elements: Cobalt (11.19%), Chromium (10.61%), Molybdenum (6.04%), and Tungsten (4.26%), aligning with known solid-solution and carbide precipitation mechanisms. These results confirm that combining machine learning with metaheuristic optimization provides a powerful, efficient pathway for designing novel alloys, promising to drastically shorten the materials development cycle.