نوع مقاله : مقاله پژوهشی کامل
عنوان مقاله English
نویسندگان English
The design of multi-principal element alloys (MPEAs), due to their vast compositional space and the complexity arising from chemical disorder, necessitates rapid and cost-effective methods for the preliminary estimation of mechanical properties. In this study, a data-driven machine learning framework was developed to predict the elastic constants (C11, C12 and C44) as well as the bulk and shear moduli for alloy systems comprising Cu, Ni, Au, Pd, and Pt. The dataset was extracted from reported binary alloys, and the atomic fractions of the elements along with binding energy were selected as the model input descriptors. Several regression algorithms, including tree-based models, Support Vector Regression, regularized regression models, Artificial Neural Networks, and ensemble models, were evaluated. The results demonstrated that the selected models achieve high predictive accuracy for elastic properties, yielding R2 values of 0.9989, 0.9832, 0.9754, 0.9642, and 0.9636 for bulk modulus, shear modulus, C11, C12 and C44, respectively. Correlation and SHAP analyses revealed that binding energy plays a dominant role in controlling the elastic response, whereas Pt is associated with an increase in bulk stiffness, and Au contributes to a reduction in shear resistance.
کلیدواژهها English