新型钴基与Nb-Si基高温合金扩散动力学研究进展

刘兴军, 魏振帮, 卢勇, 韩佳甲, 施荣沛, 王翠萍

Progress on the Diffusion Kinetics of Novel Co-based and Nb-Si-based Superalloys

LIU Xingjun, WEI Zhenbang, LU Yong, HAN Jiajia, SHI Rongpei, WANG Cuiping

图3 杂质扩散激活能(QI)预测模型中自变量的重要性排序，以及最优机器学习模型计算的自扩散激活能(Qs)、QI与实验结果的对比[67,84]

Fig.3 Ranking of the importance of features in the impurity diffusion activation energy (QI) machine learning model (a)[67], and comparisons among the results calculated by the self-diffusion activation energy (Qs) (b)[84] and QI (c)[67] machine learning models and experimental measurements (The features were classified as follows, 1) Electron configuration: numbers of electrons in closed-shell and s-, p-, d-, and f-orbits (CEC, Ns, Np, Nd, Nf); 2) Atomic properties: atomic radius (AR), atomic mass (AM), and electronegativity (EN); 3) Lattice parameters (including a, c, and γ) and atomic coordinate number (Z); 4) Cij; 5) Tm. The superscripts of the features M, I, Δ, and R denote matrix, impurity, matrix-impurity, and matrix/impurity, respectively)