基于可解释机器学习模型的难熔高熵合金相预测  

Interpretable Machine Learning Model-Based Phase Prediction for Refractory High-Entropy Alloys

作　　者：赵凤媛 叶益聪[1] 张周然 李亚豪 王洁 唐宇[1] 李顺[1] 白书欣[1] Zhao Fengyuan;Ye Yicong;Zhang Zhouran;Li Yahao;Wang Jie;Tang Yu;Li Shun;Bai Shuxin(Department of Materials Science and Engineering,College of Aerospace Science and Engineering,National University of Defense Technology,Changsha 410073,China)

机构地区：[1]国防科技大学空天科学学院材料科学与工程系,湖南长沙410073

出　　处：《稀有金属材料与工程》2023年第4期1192-1200,共9页Rare Metal Materials and Engineering

基　　金：National Natural Science Foundation of China(U20A20231,11972372);College of Aerospace Science and Engineering Youth Talent Fund;National University of Defense Technology(KY0505072209)。

摘　　要：采用k近邻(KNN)、支持向量机(SVM)、决策树(DT)、随机森林(RF)和人工神经网络(ANN)5种机器学习(ML)方法对RHEAs中固溶体(SS)、混合固溶体和金属间化合物(SS+IM)进行了分类和预测。选择了5个输入相预测参数作为特征以及139组RHEAs数据以训练ML模型。结果表明,ANN模型的预测准确率最高,达到90.72%。9组新的四元和(TiVTa)_(x)Cr_(1-x)体系RHEAs的实验结果显示,RF和ANN的预测精度更高,精准预测了11个SS和3个SS+IM合金的相组成。采用了SHAP(SHapley Additive exPlanations)模型来解释精度最高的ANN模型,并研究每个特征对相形成的贡献。5个特征的重要性顺序是混合焓(ΔH_(mix))、原子尺寸差(δ)、价电子浓度(VEC)、混合熵(ΔS_(mix))和电负性差(Δ_(χ)),其中ΔH_(mix)的平均SHAP值大约是Δ_(χ)的5倍,是ΔS_(mix)的4倍。较大的ΔH_(mix)、较小的δ和VEC可能有助于RHEA中固溶体的形成。Five machine learning(ML)approaches,i.e.K-Nearest Neighbor(KNN),Support Vector Machine(SVM),Decision Tree(DT),Random Forest(RF)and Artificial Neural Network(ANN)were used to classify and to predict the combination of phases,i.e.solid solutions(SS)and mixed solid solution and intermetallic(SS+IM)in refractory high-entropy alloys(RHEAs).Five input characteristic phase predicting parameters and 139 RHEAs were selected to train these models.Results show that ANN model has the highest accuracy of 90.72%.Experimental results of 9 quaternary and(TiVTa)_(x)Cr_(1-x)RHEAs verify the accuracy of prediction and indicate that RF and ANN can predict more accurately,successfully predicting 11 SS and 3 SS+IM.SHAP(SHapley Additive exPlanations)model was used to interpret the ANN model which exhibits the highest accuracy and to investigate the contribution of each feature to phase formation.The order of importance of five features is enthalpy of mixing(ΔH_(mix)),atomic size difference(δ),valence electron concentration(VEC),entropy of mixing(ΔS_(mix)),and electronegativity difference(Δ_(χ)),where the mean SHAP value ofΔH_(mix) is approximately 5 times higher than that of Δ_(χ) and 4 times higher than that of ΔS_(mix).Less negative ΔH_(mix),smallerδand VEC may contribute to the formation of SS in RHEAs.

关 键 词：机器学习 难熔高熵合金 模型可解释性 固溶体 金属间化合物 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程] TG139[自动化与计算机技术—控制科学与工程]